2009-08-17 22:18:08 +08:00
|
|
|
#include "builtin.h"
|
|
|
|
|
2010-08-05 21:51:38 +08:00
|
|
|
#include "perf.h"
|
2009-08-17 22:18:08 +08:00
|
|
|
#include "util/cache.h"
|
2010-08-05 21:51:38 +08:00
|
|
|
#include "util/debug.h"
|
2015-12-15 23:39:39 +08:00
|
|
|
#include <subcmd/exec-cmd.h>
|
2010-08-05 21:51:38 +08:00
|
|
|
#include "util/header.h"
|
2015-12-15 23:39:39 +08:00
|
|
|
#include <subcmd/parse-options.h>
|
2015-09-01 00:41:10 +08:00
|
|
|
#include "util/perf_regs.h"
|
2010-08-05 21:51:38 +08:00
|
|
|
#include "util/session.h"
|
2011-11-28 18:30:20 +08:00
|
|
|
#include "util/tool.h"
|
2009-08-17 22:18:08 +08:00
|
|
|
#include "util/symbol.h"
|
|
|
|
#include "util/thread.h"
|
2009-11-28 17:11:00 +08:00
|
|
|
#include "util/trace-event.h"
|
2010-08-05 21:51:38 +08:00
|
|
|
#include "util/util.h"
|
2011-03-10 13:23:28 +08:00
|
|
|
#include "util/evlist.h"
|
|
|
|
#include "util/evsel.h"
|
2012-09-07 16:42:24 +08:00
|
|
|
#include "util/sort.h"
|
2013-10-15 22:27:32 +08:00
|
|
|
#include "util/data.h"
|
2015-04-09 23:53:56 +08:00
|
|
|
#include "util/auxtrace.h"
|
2016-01-06 05:09:06 +08:00
|
|
|
#include "util/cpumap.h"
|
|
|
|
#include "util/thread_map.h"
|
|
|
|
#include "util/stat.h"
|
perf script: Add callindent option
Based on patches from Andi Kleen.
When printing PT instruction traces with perf script it is rather useful
to see some indentation for the call tree. This patch adds a new
callindent field to perf script that prints spaces for the function call
stack depth.
We already have code to track the function call stack for PT, that we
can reuse with minor modifications.
The resulting output is not quite as nice as ftrace yet, but a lot
better than what was there before.
Note there are some corner cases when the thread stack gets code
confused and prints incorrect indentation. Even with that it is fairly
useful.
When displaying kernel code traces it is recommended to run as root, as
otherwise perf doesn't understand the kernel addresses properly, and may
not reset the call stack correctly on kernel boundaries.
Example output:
sudo perf-with-kcore record eg2 -a -e intel_pt// -- sleep 1
sudo perf-with-kcore script eg2 --ns -F callindent,time,comm,pid,sym,ip,addr,flags,cpu --itrace=cre | less
...
swapper 0 [000] 5830.389116586: call irq_exit ffffffff8104d620 smp_call_function_single_interrupt+0x30 => ffffffff8107e720 irq_exit
swapper 0 [000] 5830.389116586: call idle_cpu ffffffff8107e769 irq_exit+0x49 => ffffffff810a3970 idle_cpu
swapper 0 [000] 5830.389116586: return idle_cpu ffffffff810a39b7 idle_cpu+0x47 => ffffffff8107e76e irq_exit
swapper 0 [000] 5830.389116586: call tick_nohz_irq_exit ffffffff8107e7bd irq_exit+0x9d => ffffffff810f2fc0 tick_nohz_irq_exit
swapper 0 [000] 5830.389116919: call __tick_nohz_idle_enter ffffffff810f2fe0 tick_nohz_irq_exit+0x20 => ffffffff810f28d0 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call ktime_get ffffffff810f28f1 __tick_nohz_idle_enter+0x21 => ffffffff810e9ec0 ktime_get
swapper 0 [000] 5830.389116919: call read_tsc ffffffff810e9ef6 ktime_get+0x36 => ffffffff81035070 read_tsc
swapper 0 [000] 5830.389116919: return read_tsc ffffffff81035084 read_tsc+0x14 => ffffffff810e9efc ktime_get
swapper 0 [000] 5830.389116919: return ktime_get ffffffff810e9f46 ktime_get+0x86 => ffffffff810f28f6 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call sched_clock_idle_sleep_event ffffffff810f290b __tick_nohz_idle_enter+0x3b => ffffffff810a7380 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: call sched_clock_cpu ffffffff810a738b sched_clock_idle_sleep_event+0xb => ffffffff810a72e0 sched_clock_cpu
swapper 0 [000] 5830.389116919: call sched_clock ffffffff810a734d sched_clock_cpu+0x6d => ffffffff81035750 sched_clock
swapper 0 [000] 5830.389116919: call native_sched_clock ffffffff81035754 sched_clock+0x4 => ffffffff81035640 native_sched_clock
swapper 0 [000] 5830.389116919: return native_sched_clock ffffffff8103568c native_sched_clock+0x4c => ffffffff81035759 sched_clock
swapper 0 [000] 5830.389116919: return sched_clock ffffffff8103575c sched_clock+0xc => ffffffff810a7352 sched_clock_cpu
swapper 0 [000] 5830.389116919: return sched_clock_cpu ffffffff810a7356 sched_clock_cpu+0x76 => ffffffff810a7390 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: return sched_clock_idle_sleep_event ffffffff810a7391 sched_clock_idle_sleep_event+0x11 => ffffffff810f2910 __tick_nohz_idle_enter
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-4-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:58 +08:00
|
|
|
#include "util/thread-stack.h"
|
2011-07-04 19:57:50 +08:00
|
|
|
#include <linux/bitmap.h>
|
2016-04-15 05:15:18 +08:00
|
|
|
#include <linux/stringify.h>
|
2016-01-06 05:09:06 +08:00
|
|
|
#include "asm/bug.h"
|
2016-02-24 16:46:54 +08:00
|
|
|
#include "util/mem-events.h"
|
2009-08-17 22:18:08 +08:00
|
|
|
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
static char const *script_name;
|
|
|
|
static char const *generate_script_lang;
|
2010-05-27 22:27:47 +08:00
|
|
|
static bool debug_mode;
|
2010-04-24 07:55:09 +08:00
|
|
|
static u64 last_timestamp;
|
2010-05-27 21:46:25 +08:00
|
|
|
static u64 nr_unordered;
|
2011-03-10 13:23:27 +08:00
|
|
|
static bool no_callchain;
|
2013-06-04 13:20:28 +08:00
|
|
|
static bool latency_format;
|
2011-11-25 22:05:25 +08:00
|
|
|
static bool system_wide;
|
perf script: Add field option 'flags' to print sample flags
Instruction tracing will typically have access to information about the
instruction being executed for a particular ip sample. Some of that
information will be available in the 'flags' member of struct
perf_sample.
With the addition of transactions events synthesis to Instruction
Tracing options, there is a need to be able easily to see the flags
because they show whether the ip is at the start, commit or abort of a
tranasaction.
Consequently add an option to display the flags.
The flags are "bcrosyiABEx" which stand for branch, call, return,
conditional, system, asynchronous, interrupt, transaction abort, trace
begin, trace end, and in transaction, respectively.
Example using Intel PT:
perf script -fip,time,event,sym,addr,flags
...
1288.721584105: branches:u: bo 401146 main => 401152 main
1288.721584105: transactions: x 0 401164 main
1288.721584105: branches:u: bx 40117c main => 40119b main
1288.721584105: branches:u: box 4011a4 main => 40117e main
1288.721584105: branches:u: bcx 401187 main => 401094 g
...
1288.721591645: branches:u: bx 4010c4 g => 4010cb g
1288.721591645: branches:u: brx 4010cc g => 401189 main
1288.721591645: transactions: 0 4011a6 main
1288.721593199: branches:u: b 4011a9 main => 4011af main
1288.721593199: branches:u: bo 4011bc main => 40113e main
1288.721593199: branches:u: b 401150 main => 40115a main
1288.721593199: transactions: x 0 401164 main
1288.721593199: branches:u: bx 40117c main => 40119b main
1288.721593199: branches:u: box 4011a4 main => 40117e main
1288.721593199: branches:u: bcx 401187 main => 40105e f
...
1288.722284747: branches:u: brx 401093 f => 401189 main
1288.722284747: branches:u: box 4011a4 main => 40117e main
1288.722284747: branches:u: bcx 401187 main => 40105e f
1288.722285883: transactions: bA 0 401071 f
1288.722285883: branches:u: bA 401071 f => 40116a main
1288.722285883: branches:u: bE 40116a main => 0 [unknown]
1288.722297174: branches:u: bB 0 [unknown] => 40116a main
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1428594864-29309-26-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-09 23:54:05 +08:00
|
|
|
static bool print_flags;
|
2015-09-25 21:15:36 +08:00
|
|
|
static bool nanosecs;
|
2011-07-04 19:57:50 +08:00
|
|
|
static const char *cpu_list;
|
|
|
|
static DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
|
2016-01-06 05:09:07 +08:00
|
|
|
static struct perf_stat_config stat_config;
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
|
2015-09-25 21:15:50 +08:00
|
|
|
unsigned int scripting_max_stack = PERF_MAX_STACK_DEPTH;
|
2015-09-25 21:15:49 +08:00
|
|
|
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
enum perf_output_field {
|
|
|
|
PERF_OUTPUT_COMM = 1U << 0,
|
|
|
|
PERF_OUTPUT_TID = 1U << 1,
|
|
|
|
PERF_OUTPUT_PID = 1U << 2,
|
|
|
|
PERF_OUTPUT_TIME = 1U << 3,
|
|
|
|
PERF_OUTPUT_CPU = 1U << 4,
|
|
|
|
PERF_OUTPUT_EVNAME = 1U << 5,
|
|
|
|
PERF_OUTPUT_TRACE = 1U << 6,
|
2011-05-28 04:28:43 +08:00
|
|
|
PERF_OUTPUT_IP = 1U << 7,
|
|
|
|
PERF_OUTPUT_SYM = 1U << 8,
|
2011-05-28 04:28:44 +08:00
|
|
|
PERF_OUTPUT_DSO = 1U << 9,
|
2011-05-31 03:08:23 +08:00
|
|
|
PERF_OUTPUT_ADDR = 1U << 10,
|
2012-01-30 12:43:15 +08:00
|
|
|
PERF_OUTPUT_SYMOFFSET = 1U << 11,
|
2013-12-06 15:42:57 +08:00
|
|
|
PERF_OUTPUT_SRCLINE = 1U << 12,
|
2014-08-25 22:45:42 +08:00
|
|
|
PERF_OUTPUT_PERIOD = 1U << 13,
|
2015-09-01 00:41:10 +08:00
|
|
|
PERF_OUTPUT_IREGS = 1U << 14,
|
2015-09-01 00:41:13 +08:00
|
|
|
PERF_OUTPUT_BRSTACK = 1U << 15,
|
|
|
|
PERF_OUTPUT_BRSTACKSYM = 1U << 16,
|
2016-02-15 16:34:51 +08:00
|
|
|
PERF_OUTPUT_DATA_SRC = 1U << 17,
|
|
|
|
PERF_OUTPUT_WEIGHT = 1U << 18,
|
perf script: Print bpf-output events in 'perf script'
This patch allows 'perf script' output messages from BPF program. For
example, use test_bpf_output_3.c at the end of this commit message,
# ./perf record -e bpf-output/no-inherit,name=evt/ \
-e ./test_bpf_output_3.c/map:channel.event=evt/ \
usleep 100000
# ./perf script
usleep 4882 21384.532523: evt: ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
usleep 4882 21384.632606: evt: ffffffff8105c609 kretprobe_trampoline_holder ([kernel.kallsyms
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
Two samples from BPF output are printed by both binary and string
format.
If BPF program output something unprintable, string format is
suppressed.
/************************ BEGIN **************************/
#include <uapi/linux/bpf.h>
struct bpf_map_def {
unsigned int type;
unsigned int key_size;
unsigned int value_size;
unsigned int max_entries;
};
#define SEC(NAME) __attribute__((section(NAME), used))
static u64 (*ktime_get_ns)(void) =
(void *)BPF_FUNC_ktime_get_ns;
static int (*trace_printk)(const char *fmt, int fmt_size, ...) =
(void *)BPF_FUNC_trace_printk;
static int (*get_smp_processor_id)(void) =
(void *)BPF_FUNC_get_smp_processor_id;
static int (*perf_event_output)(void *, struct bpf_map_def *, int, void *, unsigned long) =
(void *)BPF_FUNC_perf_event_output;
struct bpf_map_def SEC("maps") channel = {
.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
.key_size = sizeof(int),
.value_size = sizeof(u32),
.max_entries = __NR_CPUS__,
};
static inline int __attribute__((always_inline))
func(void *ctx, int type)
{
char output_str[] = "Raise a BPF event!";
perf_event_output(ctx, &channel, get_smp_processor_id(),
&output_str, sizeof(output_str));
return 0;
}
SEC("func_begin=sys_nanosleep")
int func_begin(void *ctx) {return func(ctx, 1);}
SEC("func_end=sys_nanosleep%return")
int func_end(void *ctx) { return func(ctx, 2);}
char _license[] SEC("license") = "GPL";
int _version SEC("version") = LINUX_VERSION_CODE;
/************************* END ***************************/
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456312845-111583-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-02-24 19:20:45 +08:00
|
|
|
PERF_OUTPUT_BPF_OUTPUT = 1U << 19,
|
perf script: Add callindent option
Based on patches from Andi Kleen.
When printing PT instruction traces with perf script it is rather useful
to see some indentation for the call tree. This patch adds a new
callindent field to perf script that prints spaces for the function call
stack depth.
We already have code to track the function call stack for PT, that we
can reuse with minor modifications.
The resulting output is not quite as nice as ftrace yet, but a lot
better than what was there before.
Note there are some corner cases when the thread stack gets code
confused and prints incorrect indentation. Even with that it is fairly
useful.
When displaying kernel code traces it is recommended to run as root, as
otherwise perf doesn't understand the kernel addresses properly, and may
not reset the call stack correctly on kernel boundaries.
Example output:
sudo perf-with-kcore record eg2 -a -e intel_pt// -- sleep 1
sudo perf-with-kcore script eg2 --ns -F callindent,time,comm,pid,sym,ip,addr,flags,cpu --itrace=cre | less
...
swapper 0 [000] 5830.389116586: call irq_exit ffffffff8104d620 smp_call_function_single_interrupt+0x30 => ffffffff8107e720 irq_exit
swapper 0 [000] 5830.389116586: call idle_cpu ffffffff8107e769 irq_exit+0x49 => ffffffff810a3970 idle_cpu
swapper 0 [000] 5830.389116586: return idle_cpu ffffffff810a39b7 idle_cpu+0x47 => ffffffff8107e76e irq_exit
swapper 0 [000] 5830.389116586: call tick_nohz_irq_exit ffffffff8107e7bd irq_exit+0x9d => ffffffff810f2fc0 tick_nohz_irq_exit
swapper 0 [000] 5830.389116919: call __tick_nohz_idle_enter ffffffff810f2fe0 tick_nohz_irq_exit+0x20 => ffffffff810f28d0 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call ktime_get ffffffff810f28f1 __tick_nohz_idle_enter+0x21 => ffffffff810e9ec0 ktime_get
swapper 0 [000] 5830.389116919: call read_tsc ffffffff810e9ef6 ktime_get+0x36 => ffffffff81035070 read_tsc
swapper 0 [000] 5830.389116919: return read_tsc ffffffff81035084 read_tsc+0x14 => ffffffff810e9efc ktime_get
swapper 0 [000] 5830.389116919: return ktime_get ffffffff810e9f46 ktime_get+0x86 => ffffffff810f28f6 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call sched_clock_idle_sleep_event ffffffff810f290b __tick_nohz_idle_enter+0x3b => ffffffff810a7380 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: call sched_clock_cpu ffffffff810a738b sched_clock_idle_sleep_event+0xb => ffffffff810a72e0 sched_clock_cpu
swapper 0 [000] 5830.389116919: call sched_clock ffffffff810a734d sched_clock_cpu+0x6d => ffffffff81035750 sched_clock
swapper 0 [000] 5830.389116919: call native_sched_clock ffffffff81035754 sched_clock+0x4 => ffffffff81035640 native_sched_clock
swapper 0 [000] 5830.389116919: return native_sched_clock ffffffff8103568c native_sched_clock+0x4c => ffffffff81035759 sched_clock
swapper 0 [000] 5830.389116919: return sched_clock ffffffff8103575c sched_clock+0xc => ffffffff810a7352 sched_clock_cpu
swapper 0 [000] 5830.389116919: return sched_clock_cpu ffffffff810a7356 sched_clock_cpu+0x76 => ffffffff810a7390 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: return sched_clock_idle_sleep_event ffffffff810a7391 sched_clock_idle_sleep_event+0x11 => ffffffff810f2910 __tick_nohz_idle_enter
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-4-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:58 +08:00
|
|
|
PERF_OUTPUT_CALLINDENT = 1U << 20,
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
struct output_option {
|
|
|
|
const char *str;
|
|
|
|
enum perf_output_field field;
|
|
|
|
} all_output_options[] = {
|
|
|
|
{.str = "comm", .field = PERF_OUTPUT_COMM},
|
|
|
|
{.str = "tid", .field = PERF_OUTPUT_TID},
|
|
|
|
{.str = "pid", .field = PERF_OUTPUT_PID},
|
|
|
|
{.str = "time", .field = PERF_OUTPUT_TIME},
|
|
|
|
{.str = "cpu", .field = PERF_OUTPUT_CPU},
|
|
|
|
{.str = "event", .field = PERF_OUTPUT_EVNAME},
|
|
|
|
{.str = "trace", .field = PERF_OUTPUT_TRACE},
|
2011-05-28 04:28:43 +08:00
|
|
|
{.str = "ip", .field = PERF_OUTPUT_IP},
|
2011-03-10 13:23:27 +08:00
|
|
|
{.str = "sym", .field = PERF_OUTPUT_SYM},
|
2011-05-28 04:28:44 +08:00
|
|
|
{.str = "dso", .field = PERF_OUTPUT_DSO},
|
2011-05-31 03:08:23 +08:00
|
|
|
{.str = "addr", .field = PERF_OUTPUT_ADDR},
|
2012-01-30 12:43:15 +08:00
|
|
|
{.str = "symoff", .field = PERF_OUTPUT_SYMOFFSET},
|
2013-12-06 15:42:57 +08:00
|
|
|
{.str = "srcline", .field = PERF_OUTPUT_SRCLINE},
|
2014-08-25 22:45:42 +08:00
|
|
|
{.str = "period", .field = PERF_OUTPUT_PERIOD},
|
2015-09-01 00:41:10 +08:00
|
|
|
{.str = "iregs", .field = PERF_OUTPUT_IREGS},
|
2015-09-01 00:41:13 +08:00
|
|
|
{.str = "brstack", .field = PERF_OUTPUT_BRSTACK},
|
|
|
|
{.str = "brstacksym", .field = PERF_OUTPUT_BRSTACKSYM},
|
2016-02-15 16:34:51 +08:00
|
|
|
{.str = "data_src", .field = PERF_OUTPUT_DATA_SRC},
|
|
|
|
{.str = "weight", .field = PERF_OUTPUT_WEIGHT},
|
perf script: Print bpf-output events in 'perf script'
This patch allows 'perf script' output messages from BPF program. For
example, use test_bpf_output_3.c at the end of this commit message,
# ./perf record -e bpf-output/no-inherit,name=evt/ \
-e ./test_bpf_output_3.c/map:channel.event=evt/ \
usleep 100000
# ./perf script
usleep 4882 21384.532523: evt: ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
usleep 4882 21384.632606: evt: ffffffff8105c609 kretprobe_trampoline_holder ([kernel.kallsyms
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
Two samples from BPF output are printed by both binary and string
format.
If BPF program output something unprintable, string format is
suppressed.
/************************ BEGIN **************************/
#include <uapi/linux/bpf.h>
struct bpf_map_def {
unsigned int type;
unsigned int key_size;
unsigned int value_size;
unsigned int max_entries;
};
#define SEC(NAME) __attribute__((section(NAME), used))
static u64 (*ktime_get_ns)(void) =
(void *)BPF_FUNC_ktime_get_ns;
static int (*trace_printk)(const char *fmt, int fmt_size, ...) =
(void *)BPF_FUNC_trace_printk;
static int (*get_smp_processor_id)(void) =
(void *)BPF_FUNC_get_smp_processor_id;
static int (*perf_event_output)(void *, struct bpf_map_def *, int, void *, unsigned long) =
(void *)BPF_FUNC_perf_event_output;
struct bpf_map_def SEC("maps") channel = {
.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
.key_size = sizeof(int),
.value_size = sizeof(u32),
.max_entries = __NR_CPUS__,
};
static inline int __attribute__((always_inline))
func(void *ctx, int type)
{
char output_str[] = "Raise a BPF event!";
perf_event_output(ctx, &channel, get_smp_processor_id(),
&output_str, sizeof(output_str));
return 0;
}
SEC("func_begin=sys_nanosleep")
int func_begin(void *ctx) {return func(ctx, 1);}
SEC("func_end=sys_nanosleep%return")
int func_end(void *ctx) { return func(ctx, 2);}
char _license[] SEC("license") = "GPL";
int _version SEC("version") = LINUX_VERSION_CODE;
/************************* END ***************************/
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456312845-111583-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-02-24 19:20:45 +08:00
|
|
|
{.str = "bpf-output", .field = PERF_OUTPUT_BPF_OUTPUT},
|
perf script: Add callindent option
Based on patches from Andi Kleen.
When printing PT instruction traces with perf script it is rather useful
to see some indentation for the call tree. This patch adds a new
callindent field to perf script that prints spaces for the function call
stack depth.
We already have code to track the function call stack for PT, that we
can reuse with minor modifications.
The resulting output is not quite as nice as ftrace yet, but a lot
better than what was there before.
Note there are some corner cases when the thread stack gets code
confused and prints incorrect indentation. Even with that it is fairly
useful.
When displaying kernel code traces it is recommended to run as root, as
otherwise perf doesn't understand the kernel addresses properly, and may
not reset the call stack correctly on kernel boundaries.
Example output:
sudo perf-with-kcore record eg2 -a -e intel_pt// -- sleep 1
sudo perf-with-kcore script eg2 --ns -F callindent,time,comm,pid,sym,ip,addr,flags,cpu --itrace=cre | less
...
swapper 0 [000] 5830.389116586: call irq_exit ffffffff8104d620 smp_call_function_single_interrupt+0x30 => ffffffff8107e720 irq_exit
swapper 0 [000] 5830.389116586: call idle_cpu ffffffff8107e769 irq_exit+0x49 => ffffffff810a3970 idle_cpu
swapper 0 [000] 5830.389116586: return idle_cpu ffffffff810a39b7 idle_cpu+0x47 => ffffffff8107e76e irq_exit
swapper 0 [000] 5830.389116586: call tick_nohz_irq_exit ffffffff8107e7bd irq_exit+0x9d => ffffffff810f2fc0 tick_nohz_irq_exit
swapper 0 [000] 5830.389116919: call __tick_nohz_idle_enter ffffffff810f2fe0 tick_nohz_irq_exit+0x20 => ffffffff810f28d0 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call ktime_get ffffffff810f28f1 __tick_nohz_idle_enter+0x21 => ffffffff810e9ec0 ktime_get
swapper 0 [000] 5830.389116919: call read_tsc ffffffff810e9ef6 ktime_get+0x36 => ffffffff81035070 read_tsc
swapper 0 [000] 5830.389116919: return read_tsc ffffffff81035084 read_tsc+0x14 => ffffffff810e9efc ktime_get
swapper 0 [000] 5830.389116919: return ktime_get ffffffff810e9f46 ktime_get+0x86 => ffffffff810f28f6 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call sched_clock_idle_sleep_event ffffffff810f290b __tick_nohz_idle_enter+0x3b => ffffffff810a7380 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: call sched_clock_cpu ffffffff810a738b sched_clock_idle_sleep_event+0xb => ffffffff810a72e0 sched_clock_cpu
swapper 0 [000] 5830.389116919: call sched_clock ffffffff810a734d sched_clock_cpu+0x6d => ffffffff81035750 sched_clock
swapper 0 [000] 5830.389116919: call native_sched_clock ffffffff81035754 sched_clock+0x4 => ffffffff81035640 native_sched_clock
swapper 0 [000] 5830.389116919: return native_sched_clock ffffffff8103568c native_sched_clock+0x4c => ffffffff81035759 sched_clock
swapper 0 [000] 5830.389116919: return sched_clock ffffffff8103575c sched_clock+0xc => ffffffff810a7352 sched_clock_cpu
swapper 0 [000] 5830.389116919: return sched_clock_cpu ffffffff810a7356 sched_clock_cpu+0x76 => ffffffff810a7390 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: return sched_clock_idle_sleep_event ffffffff810a7391 sched_clock_idle_sleep_event+0x11 => ffffffff810f2910 __tick_nohz_idle_enter
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-4-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:58 +08:00
|
|
|
{.str = "callindent", .field = PERF_OUTPUT_CALLINDENT},
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
/* default set to maintain compatibility with current format */
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
static struct {
|
|
|
|
bool user_set;
|
2011-04-07 11:54:20 +08:00
|
|
|
bool wildcard_set;
|
2013-08-08 10:50:51 +08:00
|
|
|
unsigned int print_ip_opts;
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
u64 fields;
|
|
|
|
u64 invalid_fields;
|
|
|
|
} output[PERF_TYPE_MAX] = {
|
|
|
|
|
|
|
|
[PERF_TYPE_HARDWARE] = {
|
|
|
|
.user_set = false,
|
|
|
|
|
|
|
|
.fields = PERF_OUTPUT_COMM | PERF_OUTPUT_TID |
|
|
|
|
PERF_OUTPUT_CPU | PERF_OUTPUT_TIME |
|
2011-05-28 04:28:43 +08:00
|
|
|
PERF_OUTPUT_EVNAME | PERF_OUTPUT_IP |
|
2014-08-25 22:45:43 +08:00
|
|
|
PERF_OUTPUT_SYM | PERF_OUTPUT_DSO |
|
|
|
|
PERF_OUTPUT_PERIOD,
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
|
perf script: Print bpf-output events in 'perf script'
This patch allows 'perf script' output messages from BPF program. For
example, use test_bpf_output_3.c at the end of this commit message,
# ./perf record -e bpf-output/no-inherit,name=evt/ \
-e ./test_bpf_output_3.c/map:channel.event=evt/ \
usleep 100000
# ./perf script
usleep 4882 21384.532523: evt: ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
usleep 4882 21384.632606: evt: ffffffff8105c609 kretprobe_trampoline_holder ([kernel.kallsyms
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
Two samples from BPF output are printed by both binary and string
format.
If BPF program output something unprintable, string format is
suppressed.
/************************ BEGIN **************************/
#include <uapi/linux/bpf.h>
struct bpf_map_def {
unsigned int type;
unsigned int key_size;
unsigned int value_size;
unsigned int max_entries;
};
#define SEC(NAME) __attribute__((section(NAME), used))
static u64 (*ktime_get_ns)(void) =
(void *)BPF_FUNC_ktime_get_ns;
static int (*trace_printk)(const char *fmt, int fmt_size, ...) =
(void *)BPF_FUNC_trace_printk;
static int (*get_smp_processor_id)(void) =
(void *)BPF_FUNC_get_smp_processor_id;
static int (*perf_event_output)(void *, struct bpf_map_def *, int, void *, unsigned long) =
(void *)BPF_FUNC_perf_event_output;
struct bpf_map_def SEC("maps") channel = {
.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
.key_size = sizeof(int),
.value_size = sizeof(u32),
.max_entries = __NR_CPUS__,
};
static inline int __attribute__((always_inline))
func(void *ctx, int type)
{
char output_str[] = "Raise a BPF event!";
perf_event_output(ctx, &channel, get_smp_processor_id(),
&output_str, sizeof(output_str));
return 0;
}
SEC("func_begin=sys_nanosleep")
int func_begin(void *ctx) {return func(ctx, 1);}
SEC("func_end=sys_nanosleep%return")
int func_end(void *ctx) { return func(ctx, 2);}
char _license[] SEC("license") = "GPL";
int _version SEC("version") = LINUX_VERSION_CODE;
/************************* END ***************************/
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456312845-111583-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-02-24 19:20:45 +08:00
|
|
|
.invalid_fields = PERF_OUTPUT_TRACE | PERF_OUTPUT_BPF_OUTPUT,
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
},
|
|
|
|
|
|
|
|
[PERF_TYPE_SOFTWARE] = {
|
|
|
|
.user_set = false,
|
|
|
|
|
|
|
|
.fields = PERF_OUTPUT_COMM | PERF_OUTPUT_TID |
|
|
|
|
PERF_OUTPUT_CPU | PERF_OUTPUT_TIME |
|
2011-05-28 04:28:43 +08:00
|
|
|
PERF_OUTPUT_EVNAME | PERF_OUTPUT_IP |
|
2014-08-25 22:45:43 +08:00
|
|
|
PERF_OUTPUT_SYM | PERF_OUTPUT_DSO |
|
perf script: Print bpf-output events in 'perf script'
This patch allows 'perf script' output messages from BPF program. For
example, use test_bpf_output_3.c at the end of this commit message,
# ./perf record -e bpf-output/no-inherit,name=evt/ \
-e ./test_bpf_output_3.c/map:channel.event=evt/ \
usleep 100000
# ./perf script
usleep 4882 21384.532523: evt: ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
usleep 4882 21384.632606: evt: ffffffff8105c609 kretprobe_trampoline_holder ([kernel.kallsyms
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
Two samples from BPF output are printed by both binary and string
format.
If BPF program output something unprintable, string format is
suppressed.
/************************ BEGIN **************************/
#include <uapi/linux/bpf.h>
struct bpf_map_def {
unsigned int type;
unsigned int key_size;
unsigned int value_size;
unsigned int max_entries;
};
#define SEC(NAME) __attribute__((section(NAME), used))
static u64 (*ktime_get_ns)(void) =
(void *)BPF_FUNC_ktime_get_ns;
static int (*trace_printk)(const char *fmt, int fmt_size, ...) =
(void *)BPF_FUNC_trace_printk;
static int (*get_smp_processor_id)(void) =
(void *)BPF_FUNC_get_smp_processor_id;
static int (*perf_event_output)(void *, struct bpf_map_def *, int, void *, unsigned long) =
(void *)BPF_FUNC_perf_event_output;
struct bpf_map_def SEC("maps") channel = {
.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
.key_size = sizeof(int),
.value_size = sizeof(u32),
.max_entries = __NR_CPUS__,
};
static inline int __attribute__((always_inline))
func(void *ctx, int type)
{
char output_str[] = "Raise a BPF event!";
perf_event_output(ctx, &channel, get_smp_processor_id(),
&output_str, sizeof(output_str));
return 0;
}
SEC("func_begin=sys_nanosleep")
int func_begin(void *ctx) {return func(ctx, 1);}
SEC("func_end=sys_nanosleep%return")
int func_end(void *ctx) { return func(ctx, 2);}
char _license[] SEC("license") = "GPL";
int _version SEC("version") = LINUX_VERSION_CODE;
/************************* END ***************************/
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456312845-111583-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-02-24 19:20:45 +08:00
|
|
|
PERF_OUTPUT_PERIOD | PERF_OUTPUT_BPF_OUTPUT,
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
|
|
|
|
.invalid_fields = PERF_OUTPUT_TRACE,
|
|
|
|
},
|
|
|
|
|
|
|
|
[PERF_TYPE_TRACEPOINT] = {
|
|
|
|
.user_set = false,
|
|
|
|
|
|
|
|
.fields = PERF_OUTPUT_COMM | PERF_OUTPUT_TID |
|
|
|
|
PERF_OUTPUT_CPU | PERF_OUTPUT_TIME |
|
perf script: Print bpf-output events in 'perf script'
This patch allows 'perf script' output messages from BPF program. For
example, use test_bpf_output_3.c at the end of this commit message,
# ./perf record -e bpf-output/no-inherit,name=evt/ \
-e ./test_bpf_output_3.c/map:channel.event=evt/ \
usleep 100000
# ./perf script
usleep 4882 21384.532523: evt: ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
usleep 4882 21384.632606: evt: ffffffff8105c609 kretprobe_trampoline_holder ([kernel.kallsyms
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
Two samples from BPF output are printed by both binary and string
format.
If BPF program output something unprintable, string format is
suppressed.
/************************ BEGIN **************************/
#include <uapi/linux/bpf.h>
struct bpf_map_def {
unsigned int type;
unsigned int key_size;
unsigned int value_size;
unsigned int max_entries;
};
#define SEC(NAME) __attribute__((section(NAME), used))
static u64 (*ktime_get_ns)(void) =
(void *)BPF_FUNC_ktime_get_ns;
static int (*trace_printk)(const char *fmt, int fmt_size, ...) =
(void *)BPF_FUNC_trace_printk;
static int (*get_smp_processor_id)(void) =
(void *)BPF_FUNC_get_smp_processor_id;
static int (*perf_event_output)(void *, struct bpf_map_def *, int, void *, unsigned long) =
(void *)BPF_FUNC_perf_event_output;
struct bpf_map_def SEC("maps") channel = {
.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
.key_size = sizeof(int),
.value_size = sizeof(u32),
.max_entries = __NR_CPUS__,
};
static inline int __attribute__((always_inline))
func(void *ctx, int type)
{
char output_str[] = "Raise a BPF event!";
perf_event_output(ctx, &channel, get_smp_processor_id(),
&output_str, sizeof(output_str));
return 0;
}
SEC("func_begin=sys_nanosleep")
int func_begin(void *ctx) {return func(ctx, 1);}
SEC("func_end=sys_nanosleep%return")
int func_end(void *ctx) { return func(ctx, 2);}
char _license[] SEC("license") = "GPL";
int _version SEC("version") = LINUX_VERSION_CODE;
/************************* END ***************************/
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456312845-111583-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-02-24 19:20:45 +08:00
|
|
|
PERF_OUTPUT_EVNAME | PERF_OUTPUT_TRACE
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
},
|
2011-04-15 01:38:18 +08:00
|
|
|
|
|
|
|
[PERF_TYPE_RAW] = {
|
|
|
|
.user_set = false,
|
|
|
|
|
|
|
|
.fields = PERF_OUTPUT_COMM | PERF_OUTPUT_TID |
|
|
|
|
PERF_OUTPUT_CPU | PERF_OUTPUT_TIME |
|
2011-05-28 04:28:43 +08:00
|
|
|
PERF_OUTPUT_EVNAME | PERF_OUTPUT_IP |
|
2014-08-25 22:45:43 +08:00
|
|
|
PERF_OUTPUT_SYM | PERF_OUTPUT_DSO |
|
2016-02-15 16:34:52 +08:00
|
|
|
PERF_OUTPUT_PERIOD | PERF_OUTPUT_ADDR |
|
|
|
|
PERF_OUTPUT_DATA_SRC | PERF_OUTPUT_WEIGHT,
|
2011-04-15 01:38:18 +08:00
|
|
|
|
perf script: Print bpf-output events in 'perf script'
This patch allows 'perf script' output messages from BPF program. For
example, use test_bpf_output_3.c at the end of this commit message,
# ./perf record -e bpf-output/no-inherit,name=evt/ \
-e ./test_bpf_output_3.c/map:channel.event=evt/ \
usleep 100000
# ./perf script
usleep 4882 21384.532523: evt: ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
usleep 4882 21384.632606: evt: ffffffff8105c609 kretprobe_trampoline_holder ([kernel.kallsyms
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
Two samples from BPF output are printed by both binary and string
format.
If BPF program output something unprintable, string format is
suppressed.
/************************ BEGIN **************************/
#include <uapi/linux/bpf.h>
struct bpf_map_def {
unsigned int type;
unsigned int key_size;
unsigned int value_size;
unsigned int max_entries;
};
#define SEC(NAME) __attribute__((section(NAME), used))
static u64 (*ktime_get_ns)(void) =
(void *)BPF_FUNC_ktime_get_ns;
static int (*trace_printk)(const char *fmt, int fmt_size, ...) =
(void *)BPF_FUNC_trace_printk;
static int (*get_smp_processor_id)(void) =
(void *)BPF_FUNC_get_smp_processor_id;
static int (*perf_event_output)(void *, struct bpf_map_def *, int, void *, unsigned long) =
(void *)BPF_FUNC_perf_event_output;
struct bpf_map_def SEC("maps") channel = {
.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
.key_size = sizeof(int),
.value_size = sizeof(u32),
.max_entries = __NR_CPUS__,
};
static inline int __attribute__((always_inline))
func(void *ctx, int type)
{
char output_str[] = "Raise a BPF event!";
perf_event_output(ctx, &channel, get_smp_processor_id(),
&output_str, sizeof(output_str));
return 0;
}
SEC("func_begin=sys_nanosleep")
int func_begin(void *ctx) {return func(ctx, 1);}
SEC("func_end=sys_nanosleep%return")
int func_end(void *ctx) { return func(ctx, 2);}
char _license[] SEC("license") = "GPL";
int _version SEC("version") = LINUX_VERSION_CODE;
/************************* END ***************************/
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456312845-111583-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-02-24 19:20:45 +08:00
|
|
|
.invalid_fields = PERF_OUTPUT_TRACE | PERF_OUTPUT_BPF_OUTPUT,
|
2011-04-15 01:38:18 +08:00
|
|
|
},
|
perf script: Add support for PERF_TYPE_BREAKPOINT
Useful for getting stack traces for hardware breakpoint events.
Test result:
Before this patch:
# ~/perf record -g -e mem:0x600980 ./sample
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.011 MB perf.data (12 samples) ]
# ~/perf script
# ~/perf script -F comm,tid,pid,time,event,ip,sym,dso
sample 22520/22520 97457.836294: mem:0x600980:
5a4ad8 __clear_user (/lib/modules/4.3.0-rc4+/build/vmlinux)
...
3f41ba sys_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
979395 return_from_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
7f1b59719cf7 [unknown] ([unknown])
sample 22520/22520 97457.836648: mem:0x600980:
532 main (/home/w00229757/DataBreakpoints/sample)
21bd5 __libc_start_main (/tmp/oxygen_root-root/lib64/libc-2.18.so)
...
After this patch:
# ~/perf script
sample 22520 97457.836294: mem:0x600980:
5a4ad8 __clear_user (/lib/modules/4.3.0-rc4+/build/vmlinux)
...
3f41ba sys_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
979395 return_from_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
7f1b59719cf7 [unknown] ([unknown])
sample 22520 97457.836648: mem:0x600980:
532 main (/home/w00229757/DataBreakpoints/sample)
21bd5 __libc_start_main (/tmp/oxygen_root-root/lib64/libc-2.18.so)
Committer note:
So, further testing, lets do it for a kernel global variable,
tcp_hashinfo:
# grep -w tcp_hashinfo /proc/kallsyms
ffffffff8202fc00 B tcp_hashinfo
#
Note: allow specifying mem:tcp_hashinfo:
# perf record -g -e mem:0xffffffff81c65ac0 -a
^C[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.790 MB perf.data ]
#
# perf evlist
mem:0xffffffff8202fc00
# perf evlist -v
mem:0xffffffff8202fc00: type: 5, size: 112, { sample_period, sample_freq }: 1, sample_type: IP|TID|TIME|CALLCHAIN|CPU, disabled: 1, inherit: 1, mmap: 1, comm: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, bp_type: 3, { bp_addr, config1 }: 0xffffffff8202fc00, { bp_len, config2 }: 0x4
#
Then, after this patch:
# perf script
swapper 0 [000] 171036.986988: mem:0xffffffff8202fc00:
8a0fb5 __inet_lookup_established (/lib/modules/4.3.0+/build/vmlinux)
8bc09d tcp_v4_early_demux (/lib/modules/4.3.0+/build/vmlinux)
896def ip_rcv_finish (/lib/modules/4.3.0+/build/vmlinux)
8976c2 ip_rcv (/lib/modules/4.3.0+/build/vmlinux)
855eba __netif_receive_skb_core (/lib/modules/4.3.0+/build/vmlinux)
8565d8 __netif_receive_skb (/lib/modules/4.3.0+/build/vmlinux)
8572a8 process_backlog (/lib/modules/4.3.0+/build/vmlinux)
856b11 net_rx_action (/lib/modules/4.3.0+/build/vmlinux)
2a284b __do_softirq (/lib/modules/4.3.0+/build/vmlinux)
2a2ba3 irq_exit (/lib/modules/4.3.0+/build/vmlinux)
96b7a4 do_IRQ (/lib/modules/4.3.0+/build/vmlinux)
969807 ret_from_intr (/lib/modules/4.3.0+/build/vmlinux)
804c27 cpuidle_enter (/lib/modules/4.3.0+/build/vmlinux)
2ded22 call_cpuidle (/lib/modules/4.3.0+/build/vmlinux)
2defb6 cpu_startup_entry (/lib/modules/4.3.0+/build/vmlinux)
95d5bc rest_init (/lib/modules/4.3.0+/build/vmlinux)
1163ffa start_kernel ([kernel.vmlinux].init.text)
11634d7 x86_64_start_reservations ([kernel.vmlinux].init.text)
1163623 x86_64_start_kernel ([kernel.vmlinux].init.text)
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1449541544-67621-16-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-12-08 10:25:43 +08:00
|
|
|
|
|
|
|
[PERF_TYPE_BREAKPOINT] = {
|
|
|
|
.user_set = false,
|
|
|
|
|
|
|
|
.fields = PERF_OUTPUT_COMM | PERF_OUTPUT_TID |
|
|
|
|
PERF_OUTPUT_CPU | PERF_OUTPUT_TIME |
|
|
|
|
PERF_OUTPUT_EVNAME | PERF_OUTPUT_IP |
|
|
|
|
PERF_OUTPUT_SYM | PERF_OUTPUT_DSO |
|
|
|
|
PERF_OUTPUT_PERIOD,
|
|
|
|
|
perf script: Print bpf-output events in 'perf script'
This patch allows 'perf script' output messages from BPF program. For
example, use test_bpf_output_3.c at the end of this commit message,
# ./perf record -e bpf-output/no-inherit,name=evt/ \
-e ./test_bpf_output_3.c/map:channel.event=evt/ \
usleep 100000
# ./perf script
usleep 4882 21384.532523: evt: ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
usleep 4882 21384.632606: evt: ffffffff8105c609 kretprobe_trampoline_holder ([kernel.kallsyms
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
Two samples from BPF output are printed by both binary and string
format.
If BPF program output something unprintable, string format is
suppressed.
/************************ BEGIN **************************/
#include <uapi/linux/bpf.h>
struct bpf_map_def {
unsigned int type;
unsigned int key_size;
unsigned int value_size;
unsigned int max_entries;
};
#define SEC(NAME) __attribute__((section(NAME), used))
static u64 (*ktime_get_ns)(void) =
(void *)BPF_FUNC_ktime_get_ns;
static int (*trace_printk)(const char *fmt, int fmt_size, ...) =
(void *)BPF_FUNC_trace_printk;
static int (*get_smp_processor_id)(void) =
(void *)BPF_FUNC_get_smp_processor_id;
static int (*perf_event_output)(void *, struct bpf_map_def *, int, void *, unsigned long) =
(void *)BPF_FUNC_perf_event_output;
struct bpf_map_def SEC("maps") channel = {
.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
.key_size = sizeof(int),
.value_size = sizeof(u32),
.max_entries = __NR_CPUS__,
};
static inline int __attribute__((always_inline))
func(void *ctx, int type)
{
char output_str[] = "Raise a BPF event!";
perf_event_output(ctx, &channel, get_smp_processor_id(),
&output_str, sizeof(output_str));
return 0;
}
SEC("func_begin=sys_nanosleep")
int func_begin(void *ctx) {return func(ctx, 1);}
SEC("func_end=sys_nanosleep%return")
int func_end(void *ctx) { return func(ctx, 2);}
char _license[] SEC("license") = "GPL";
int _version SEC("version") = LINUX_VERSION_CODE;
/************************* END ***************************/
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456312845-111583-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-02-24 19:20:45 +08:00
|
|
|
.invalid_fields = PERF_OUTPUT_TRACE | PERF_OUTPUT_BPF_OUTPUT,
|
perf script: Add support for PERF_TYPE_BREAKPOINT
Useful for getting stack traces for hardware breakpoint events.
Test result:
Before this patch:
# ~/perf record -g -e mem:0x600980 ./sample
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.011 MB perf.data (12 samples) ]
# ~/perf script
# ~/perf script -F comm,tid,pid,time,event,ip,sym,dso
sample 22520/22520 97457.836294: mem:0x600980:
5a4ad8 __clear_user (/lib/modules/4.3.0-rc4+/build/vmlinux)
...
3f41ba sys_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
979395 return_from_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
7f1b59719cf7 [unknown] ([unknown])
sample 22520/22520 97457.836648: mem:0x600980:
532 main (/home/w00229757/DataBreakpoints/sample)
21bd5 __libc_start_main (/tmp/oxygen_root-root/lib64/libc-2.18.so)
...
After this patch:
# ~/perf script
sample 22520 97457.836294: mem:0x600980:
5a4ad8 __clear_user (/lib/modules/4.3.0-rc4+/build/vmlinux)
...
3f41ba sys_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
979395 return_from_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
7f1b59719cf7 [unknown] ([unknown])
sample 22520 97457.836648: mem:0x600980:
532 main (/home/w00229757/DataBreakpoints/sample)
21bd5 __libc_start_main (/tmp/oxygen_root-root/lib64/libc-2.18.so)
Committer note:
So, further testing, lets do it for a kernel global variable,
tcp_hashinfo:
# grep -w tcp_hashinfo /proc/kallsyms
ffffffff8202fc00 B tcp_hashinfo
#
Note: allow specifying mem:tcp_hashinfo:
# perf record -g -e mem:0xffffffff81c65ac0 -a
^C[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.790 MB perf.data ]
#
# perf evlist
mem:0xffffffff8202fc00
# perf evlist -v
mem:0xffffffff8202fc00: type: 5, size: 112, { sample_period, sample_freq }: 1, sample_type: IP|TID|TIME|CALLCHAIN|CPU, disabled: 1, inherit: 1, mmap: 1, comm: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, bp_type: 3, { bp_addr, config1 }: 0xffffffff8202fc00, { bp_len, config2 }: 0x4
#
Then, after this patch:
# perf script
swapper 0 [000] 171036.986988: mem:0xffffffff8202fc00:
8a0fb5 __inet_lookup_established (/lib/modules/4.3.0+/build/vmlinux)
8bc09d tcp_v4_early_demux (/lib/modules/4.3.0+/build/vmlinux)
896def ip_rcv_finish (/lib/modules/4.3.0+/build/vmlinux)
8976c2 ip_rcv (/lib/modules/4.3.0+/build/vmlinux)
855eba __netif_receive_skb_core (/lib/modules/4.3.0+/build/vmlinux)
8565d8 __netif_receive_skb (/lib/modules/4.3.0+/build/vmlinux)
8572a8 process_backlog (/lib/modules/4.3.0+/build/vmlinux)
856b11 net_rx_action (/lib/modules/4.3.0+/build/vmlinux)
2a284b __do_softirq (/lib/modules/4.3.0+/build/vmlinux)
2a2ba3 irq_exit (/lib/modules/4.3.0+/build/vmlinux)
96b7a4 do_IRQ (/lib/modules/4.3.0+/build/vmlinux)
969807 ret_from_intr (/lib/modules/4.3.0+/build/vmlinux)
804c27 cpuidle_enter (/lib/modules/4.3.0+/build/vmlinux)
2ded22 call_cpuidle (/lib/modules/4.3.0+/build/vmlinux)
2defb6 cpu_startup_entry (/lib/modules/4.3.0+/build/vmlinux)
95d5bc rest_init (/lib/modules/4.3.0+/build/vmlinux)
1163ffa start_kernel ([kernel.vmlinux].init.text)
11634d7 x86_64_start_reservations ([kernel.vmlinux].init.text)
1163623 x86_64_start_kernel ([kernel.vmlinux].init.text)
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1449541544-67621-16-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-12-08 10:25:43 +08:00
|
|
|
},
|
2011-03-10 13:23:28 +08:00
|
|
|
};
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
static bool output_set_by_user(void)
|
|
|
|
{
|
|
|
|
int j;
|
|
|
|
for (j = 0; j < PERF_TYPE_MAX; ++j) {
|
|
|
|
if (output[j].user_set)
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
|
2011-04-07 11:54:20 +08:00
|
|
|
static const char *output_field2str(enum perf_output_field field)
|
|
|
|
{
|
|
|
|
int i, imax = ARRAY_SIZE(all_output_options);
|
|
|
|
const char *str = "";
|
|
|
|
|
|
|
|
for (i = 0; i < imax; ++i) {
|
|
|
|
if (all_output_options[i].field == field) {
|
|
|
|
str = all_output_options[i].str;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return str;
|
|
|
|
}
|
|
|
|
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
#define PRINT_FIELD(x) (output[attr->type].fields & PERF_OUTPUT_##x)
|
2011-03-10 13:23:28 +08:00
|
|
|
|
2015-04-25 03:29:44 +08:00
|
|
|
static int perf_evsel__do_check_stype(struct perf_evsel *evsel,
|
|
|
|
u64 sample_type, const char *sample_msg,
|
|
|
|
enum perf_output_field field,
|
|
|
|
bool allow_user_set)
|
2011-03-10 13:23:28 +08:00
|
|
|
{
|
2012-06-13 00:35:44 +08:00
|
|
|
struct perf_event_attr *attr = &evsel->attr;
|
2011-04-07 11:54:20 +08:00
|
|
|
int type = attr->type;
|
|
|
|
const char *evname;
|
|
|
|
|
|
|
|
if (attr->sample_type & sample_type)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (output[type].user_set) {
|
2015-04-25 03:29:44 +08:00
|
|
|
if (allow_user_set)
|
|
|
|
return 0;
|
2012-06-13 00:35:44 +08:00
|
|
|
evname = perf_evsel__name(evsel);
|
2011-04-07 11:54:20 +08:00
|
|
|
pr_err("Samples for '%s' event do not have %s attribute set. "
|
|
|
|
"Cannot print '%s' field.\n",
|
|
|
|
evname, sample_msg, output_field2str(field));
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* user did not ask for it explicitly so remove from the default list */
|
|
|
|
output[type].fields &= ~field;
|
2012-06-13 00:35:44 +08:00
|
|
|
evname = perf_evsel__name(evsel);
|
2011-04-07 11:54:20 +08:00
|
|
|
pr_debug("Samples for '%s' event do not have %s attribute set. "
|
|
|
|
"Skipping '%s' field.\n",
|
|
|
|
evname, sample_msg, output_field2str(field));
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2015-04-25 03:29:44 +08:00
|
|
|
static int perf_evsel__check_stype(struct perf_evsel *evsel,
|
|
|
|
u64 sample_type, const char *sample_msg,
|
|
|
|
enum perf_output_field field)
|
|
|
|
{
|
|
|
|
return perf_evsel__do_check_stype(evsel, sample_type, sample_msg, field,
|
|
|
|
false);
|
|
|
|
}
|
|
|
|
|
2011-04-07 11:54:20 +08:00
|
|
|
static int perf_evsel__check_attr(struct perf_evsel *evsel,
|
|
|
|
struct perf_session *session)
|
|
|
|
{
|
|
|
|
struct perf_event_attr *attr = &evsel->attr;
|
2015-04-25 03:29:44 +08:00
|
|
|
bool allow_user_set;
|
|
|
|
|
2016-01-06 05:09:09 +08:00
|
|
|
if (perf_header__has_feat(&session->header, HEADER_STAT))
|
|
|
|
return 0;
|
|
|
|
|
2015-04-25 03:29:44 +08:00
|
|
|
allow_user_set = perf_header__has_feat(&session->header,
|
|
|
|
HEADER_AUXTRACE);
|
2011-04-07 11:54:20 +08:00
|
|
|
|
2011-03-10 13:23:28 +08:00
|
|
|
if (PRINT_FIELD(TRACE) &&
|
|
|
|
!perf_session__has_traces(session, "record -R"))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2011-05-28 04:28:43 +08:00
|
|
|
if (PRINT_FIELD(IP)) {
|
2012-06-13 00:35:44 +08:00
|
|
|
if (perf_evsel__check_stype(evsel, PERF_SAMPLE_IP, "IP",
|
|
|
|
PERF_OUTPUT_IP))
|
2011-03-10 13:23:28 +08:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
2011-05-31 03:08:23 +08:00
|
|
|
|
|
|
|
if (PRINT_FIELD(ADDR) &&
|
2015-04-25 03:29:44 +08:00
|
|
|
perf_evsel__do_check_stype(evsel, PERF_SAMPLE_ADDR, "ADDR",
|
|
|
|
PERF_OUTPUT_ADDR, allow_user_set))
|
2011-05-31 03:08:23 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
2016-02-15 16:34:51 +08:00
|
|
|
if (PRINT_FIELD(DATA_SRC) &&
|
|
|
|
perf_evsel__check_stype(evsel, PERF_SAMPLE_DATA_SRC, "DATA_SRC",
|
|
|
|
PERF_OUTPUT_DATA_SRC))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (PRINT_FIELD(WEIGHT) &&
|
|
|
|
perf_evsel__check_stype(evsel, PERF_SAMPLE_WEIGHT, "WEIGHT",
|
|
|
|
PERF_OUTPUT_WEIGHT))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2011-05-31 03:08:23 +08:00
|
|
|
if (PRINT_FIELD(SYM) && !PRINT_FIELD(IP) && !PRINT_FIELD(ADDR)) {
|
|
|
|
pr_err("Display of symbols requested but neither sample IP nor "
|
|
|
|
"sample address\nis selected. Hence, no addresses to convert "
|
|
|
|
"to symbols.\n");
|
2011-05-28 04:28:43 +08:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
2012-01-30 12:43:15 +08:00
|
|
|
if (PRINT_FIELD(SYMOFFSET) && !PRINT_FIELD(SYM)) {
|
|
|
|
pr_err("Display of offsets requested but symbol is not"
|
|
|
|
"selected.\n");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
2011-05-31 03:08:23 +08:00
|
|
|
if (PRINT_FIELD(DSO) && !PRINT_FIELD(IP) && !PRINT_FIELD(ADDR)) {
|
|
|
|
pr_err("Display of DSO requested but neither sample IP nor "
|
|
|
|
"sample address\nis selected. Hence, no addresses to convert "
|
|
|
|
"to DSO.\n");
|
2011-05-28 04:28:44 +08:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
2013-12-06 15:42:57 +08:00
|
|
|
if (PRINT_FIELD(SRCLINE) && !PRINT_FIELD(IP)) {
|
|
|
|
pr_err("Display of source line number requested but sample IP is not\n"
|
|
|
|
"selected. Hence, no address to lookup the source line number.\n");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
2011-03-10 13:23:28 +08:00
|
|
|
|
|
|
|
if ((PRINT_FIELD(PID) || PRINT_FIELD(TID)) &&
|
2012-06-13 00:35:44 +08:00
|
|
|
perf_evsel__check_stype(evsel, PERF_SAMPLE_TID, "TID",
|
|
|
|
PERF_OUTPUT_TID|PERF_OUTPUT_PID))
|
2011-03-10 13:23:28 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (PRINT_FIELD(TIME) &&
|
2012-06-13 00:35:44 +08:00
|
|
|
perf_evsel__check_stype(evsel, PERF_SAMPLE_TIME, "TIME",
|
|
|
|
PERF_OUTPUT_TIME))
|
2011-03-10 13:23:28 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (PRINT_FIELD(CPU) &&
|
2015-04-25 03:29:44 +08:00
|
|
|
perf_evsel__do_check_stype(evsel, PERF_SAMPLE_CPU, "CPU",
|
|
|
|
PERF_OUTPUT_CPU, allow_user_set))
|
2011-03-10 13:23:28 +08:00
|
|
|
return -EINVAL;
|
2011-04-07 11:54:20 +08:00
|
|
|
|
2014-08-25 22:45:42 +08:00
|
|
|
if (PRINT_FIELD(PERIOD) &&
|
|
|
|
perf_evsel__check_stype(evsel, PERF_SAMPLE_PERIOD, "PERIOD",
|
|
|
|
PERF_OUTPUT_PERIOD))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2015-09-01 00:41:10 +08:00
|
|
|
if (PRINT_FIELD(IREGS) &&
|
|
|
|
perf_evsel__check_stype(evsel, PERF_SAMPLE_REGS_INTR, "IREGS",
|
|
|
|
PERF_OUTPUT_IREGS))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2011-04-07 11:54:20 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-11-01 21:51:30 +08:00
|
|
|
static void set_print_ip_opts(struct perf_event_attr *attr)
|
|
|
|
{
|
|
|
|
unsigned int type = attr->type;
|
|
|
|
|
|
|
|
output[type].print_ip_opts = 0;
|
|
|
|
if (PRINT_FIELD(IP))
|
2016-04-13 02:16:15 +08:00
|
|
|
output[type].print_ip_opts |= EVSEL__PRINT_IP;
|
2013-11-01 21:51:30 +08:00
|
|
|
|
|
|
|
if (PRINT_FIELD(SYM))
|
2016-04-13 02:16:15 +08:00
|
|
|
output[type].print_ip_opts |= EVSEL__PRINT_SYM;
|
2013-11-01 21:51:30 +08:00
|
|
|
|
|
|
|
if (PRINT_FIELD(DSO))
|
2016-04-13 02:16:15 +08:00
|
|
|
output[type].print_ip_opts |= EVSEL__PRINT_DSO;
|
2013-11-01 21:51:30 +08:00
|
|
|
|
|
|
|
if (PRINT_FIELD(SYMOFFSET))
|
2016-04-13 02:16:15 +08:00
|
|
|
output[type].print_ip_opts |= EVSEL__PRINT_SYMOFFSET;
|
2013-12-06 15:42:57 +08:00
|
|
|
|
|
|
|
if (PRINT_FIELD(SRCLINE))
|
2016-04-13 02:16:15 +08:00
|
|
|
output[type].print_ip_opts |= EVSEL__PRINT_SRCLINE;
|
2013-11-01 21:51:30 +08:00
|
|
|
}
|
|
|
|
|
2011-04-07 11:54:20 +08:00
|
|
|
/*
|
|
|
|
* verify all user requested events exist and the samples
|
|
|
|
* have the expected data
|
|
|
|
*/
|
|
|
|
static int perf_session__check_output_opt(struct perf_session *session)
|
|
|
|
{
|
2016-05-16 12:51:19 +08:00
|
|
|
unsigned int j;
|
2011-04-07 11:54:20 +08:00
|
|
|
struct perf_evsel *evsel;
|
|
|
|
|
|
|
|
for (j = 0; j < PERF_TYPE_MAX; ++j) {
|
|
|
|
evsel = perf_session__find_first_evtype(session, j);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* even if fields is set to 0 (ie., show nothing) event must
|
|
|
|
* exist if user explicitly includes it on the command line
|
|
|
|
*/
|
|
|
|
if (!evsel && output[j].user_set && !output[j].wildcard_set) {
|
|
|
|
pr_err("%s events do not exist. "
|
|
|
|
"Remove corresponding -f option to proceed.\n",
|
|
|
|
event_type(j));
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (evsel && output[j].fields &&
|
|
|
|
perf_evsel__check_attr(evsel, session))
|
|
|
|
return -1;
|
2013-08-08 10:50:51 +08:00
|
|
|
|
|
|
|
if (evsel == NULL)
|
|
|
|
continue;
|
|
|
|
|
2013-11-01 21:51:30 +08:00
|
|
|
set_print_ip_opts(&evsel->attr);
|
2011-03-10 13:23:28 +08:00
|
|
|
}
|
|
|
|
|
2014-07-31 14:00:59 +08:00
|
|
|
if (!no_callchain) {
|
|
|
|
bool use_callchain = false;
|
2016-08-04 19:25:43 +08:00
|
|
|
bool not_pipe = false;
|
2014-07-31 14:00:59 +08:00
|
|
|
|
2016-06-23 22:26:15 +08:00
|
|
|
evlist__for_each_entry(session->evlist, evsel) {
|
2016-08-04 19:25:43 +08:00
|
|
|
not_pipe = true;
|
2014-07-31 14:00:59 +08:00
|
|
|
if (evsel->attr.sample_type & PERF_SAMPLE_CALLCHAIN) {
|
|
|
|
use_callchain = true;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2016-08-04 19:25:43 +08:00
|
|
|
if (not_pipe && !use_callchain)
|
2014-07-31 14:00:59 +08:00
|
|
|
symbol_conf.use_callchain = false;
|
|
|
|
}
|
|
|
|
|
2013-11-20 12:07:37 +08:00
|
|
|
/*
|
|
|
|
* set default for tracepoints to print symbols only
|
|
|
|
* if callchains are present
|
|
|
|
*/
|
|
|
|
if (symbol_conf.use_callchain &&
|
|
|
|
!output[PERF_TYPE_TRACEPOINT].user_set) {
|
|
|
|
struct perf_event_attr *attr;
|
|
|
|
|
|
|
|
j = PERF_TYPE_TRACEPOINT;
|
|
|
|
|
2016-06-23 22:26:15 +08:00
|
|
|
evlist__for_each_entry(session->evlist, evsel) {
|
2016-05-16 12:51:19 +08:00
|
|
|
if (evsel->attr.type != j)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
attr = &evsel->attr;
|
2013-11-20 12:07:37 +08:00
|
|
|
|
2016-05-16 12:51:19 +08:00
|
|
|
if (attr->sample_type & PERF_SAMPLE_CALLCHAIN) {
|
|
|
|
output[j].fields |= PERF_OUTPUT_IP;
|
|
|
|
output[j].fields |= PERF_OUTPUT_SYM;
|
|
|
|
output[j].fields |= PERF_OUTPUT_DSO;
|
|
|
|
set_print_ip_opts(attr);
|
|
|
|
goto out;
|
|
|
|
}
|
2013-11-20 12:07:37 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
2011-03-10 13:23:28 +08:00
|
|
|
return 0;
|
|
|
|
}
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
|
2016-03-23 22:55:24 +08:00
|
|
|
static void print_sample_iregs(struct perf_sample *sample,
|
2015-09-01 00:41:10 +08:00
|
|
|
struct perf_event_attr *attr)
|
|
|
|
{
|
|
|
|
struct regs_dump *regs = &sample->intr_regs;
|
|
|
|
uint64_t mask = attr->sample_regs_intr;
|
|
|
|
unsigned i = 0, r;
|
|
|
|
|
|
|
|
if (!regs)
|
|
|
|
return;
|
|
|
|
|
|
|
|
for_each_set_bit(r, (unsigned long *) &mask, sizeof(mask) * 8) {
|
|
|
|
u64 val = regs->regs[i++];
|
|
|
|
printf("%5s:0x%"PRIx64" ", perf_reg_name(r), val);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-08-07 20:58:03 +08:00
|
|
|
static void print_sample_start(struct perf_sample *sample,
|
2011-03-10 13:23:28 +08:00
|
|
|
struct thread *thread,
|
2012-06-13 00:35:44 +08:00
|
|
|
struct perf_evsel *evsel)
|
2011-03-10 13:23:25 +08:00
|
|
|
{
|
2012-06-13 00:35:44 +08:00
|
|
|
struct perf_event_attr *attr = &evsel->attr;
|
2011-03-10 13:23:25 +08:00
|
|
|
unsigned long secs;
|
|
|
|
unsigned long usecs;
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
unsigned long long nsecs;
|
|
|
|
|
|
|
|
if (PRINT_FIELD(COMM)) {
|
|
|
|
if (latency_format)
|
2013-09-11 20:46:56 +08:00
|
|
|
printf("%8.8s ", thread__comm_str(thread));
|
2011-05-28 04:28:43 +08:00
|
|
|
else if (PRINT_FIELD(IP) && symbol_conf.use_callchain)
|
2013-09-11 20:46:56 +08:00
|
|
|
printf("%s ", thread__comm_str(thread));
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
else
|
2013-09-11 20:46:56 +08:00
|
|
|
printf("%16s ", thread__comm_str(thread));
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
}
|
2011-03-10 13:23:25 +08:00
|
|
|
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
if (PRINT_FIELD(PID) && PRINT_FIELD(TID))
|
|
|
|
printf("%5d/%-5d ", sample->pid, sample->tid);
|
|
|
|
else if (PRINT_FIELD(PID))
|
|
|
|
printf("%5d ", sample->pid);
|
|
|
|
else if (PRINT_FIELD(TID))
|
|
|
|
printf("%5d ", sample->tid);
|
|
|
|
|
|
|
|
if (PRINT_FIELD(CPU)) {
|
|
|
|
if (latency_format)
|
|
|
|
printf("%3d ", sample->cpu);
|
|
|
|
else
|
|
|
|
printf("[%03d] ", sample->cpu);
|
|
|
|
}
|
2011-03-10 13:23:25 +08:00
|
|
|
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
if (PRINT_FIELD(TIME)) {
|
|
|
|
nsecs = sample->time;
|
|
|
|
secs = nsecs / NSECS_PER_SEC;
|
|
|
|
nsecs -= secs * NSECS_PER_SEC;
|
|
|
|
usecs = nsecs / NSECS_PER_USEC;
|
2015-09-25 21:15:36 +08:00
|
|
|
if (nanosecs)
|
|
|
|
printf("%5lu.%09llu: ", secs, nsecs);
|
|
|
|
else
|
|
|
|
printf("%5lu.%06lu: ", secs, usecs);
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
}
|
2011-03-10 13:23:25 +08:00
|
|
|
}
|
|
|
|
|
2015-09-01 00:41:13 +08:00
|
|
|
static inline char
|
|
|
|
mispred_str(struct branch_entry *br)
|
|
|
|
{
|
|
|
|
if (!(br->flags.mispred || br->flags.predicted))
|
|
|
|
return '-';
|
|
|
|
|
|
|
|
return br->flags.predicted ? 'P' : 'M';
|
|
|
|
}
|
|
|
|
|
2016-03-23 22:55:24 +08:00
|
|
|
static void print_sample_brstack(struct perf_sample *sample)
|
2015-09-01 00:41:13 +08:00
|
|
|
{
|
|
|
|
struct branch_stack *br = sample->branch_stack;
|
|
|
|
u64 i;
|
|
|
|
|
|
|
|
if (!(br && br->nr))
|
|
|
|
return;
|
|
|
|
|
|
|
|
for (i = 0; i < br->nr; i++) {
|
|
|
|
printf(" 0x%"PRIx64"/0x%"PRIx64"/%c/%c/%c/%d ",
|
|
|
|
br->entries[i].from,
|
|
|
|
br->entries[i].to,
|
|
|
|
mispred_str( br->entries + i),
|
|
|
|
br->entries[i].flags.in_tx? 'X' : '-',
|
|
|
|
br->entries[i].flags.abort? 'A' : '-',
|
|
|
|
br->entries[i].flags.cycles);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-03-23 22:55:24 +08:00
|
|
|
static void print_sample_brstacksym(struct perf_sample *sample,
|
|
|
|
struct thread *thread)
|
2015-09-01 00:41:13 +08:00
|
|
|
{
|
|
|
|
struct branch_stack *br = sample->branch_stack;
|
|
|
|
struct addr_location alf, alt;
|
|
|
|
u64 i, from, to;
|
|
|
|
|
|
|
|
if (!(br && br->nr))
|
|
|
|
return;
|
|
|
|
|
|
|
|
for (i = 0; i < br->nr; i++) {
|
|
|
|
|
|
|
|
memset(&alf, 0, sizeof(alf));
|
|
|
|
memset(&alt, 0, sizeof(alt));
|
|
|
|
from = br->entries[i].from;
|
|
|
|
to = br->entries[i].to;
|
|
|
|
|
2016-03-23 05:23:43 +08:00
|
|
|
thread__find_addr_map(thread, sample->cpumode, MAP__FUNCTION, from, &alf);
|
2015-09-01 00:41:13 +08:00
|
|
|
if (alf.map)
|
|
|
|
alf.sym = map__find_symbol(alf.map, alf.addr, NULL);
|
|
|
|
|
2016-03-23 05:23:43 +08:00
|
|
|
thread__find_addr_map(thread, sample->cpumode, MAP__FUNCTION, to, &alt);
|
2015-09-01 00:41:13 +08:00
|
|
|
if (alt.map)
|
|
|
|
alt.sym = map__find_symbol(alt.map, alt.addr, NULL);
|
|
|
|
|
|
|
|
symbol__fprintf_symname_offs(alf.sym, &alf, stdout);
|
|
|
|
putchar('/');
|
|
|
|
symbol__fprintf_symname_offs(alt.sym, &alt, stdout);
|
|
|
|
printf("/%c/%c/%c/%d ",
|
|
|
|
mispred_str( br->entries + i),
|
|
|
|
br->entries[i].flags.in_tx? 'X' : '-',
|
|
|
|
br->entries[i].flags.abort? 'A' : '-',
|
|
|
|
br->entries[i].flags.cycles);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2016-03-23 22:55:24 +08:00
|
|
|
static void print_sample_addr(struct perf_sample *sample,
|
2011-05-31 03:08:23 +08:00
|
|
|
struct thread *thread,
|
|
|
|
struct perf_event_attr *attr)
|
|
|
|
{
|
|
|
|
struct addr_location al;
|
|
|
|
|
|
|
|
printf("%16" PRIx64, sample->addr);
|
|
|
|
|
|
|
|
if (!sample_addr_correlates_sym(attr))
|
|
|
|
return;
|
|
|
|
|
2016-03-23 05:44:46 +08:00
|
|
|
thread__resolve(thread, &al, sample);
|
2011-05-31 03:08:23 +08:00
|
|
|
|
|
|
|
if (PRINT_FIELD(SYM)) {
|
2012-01-30 12:42:57 +08:00
|
|
|
printf(" ");
|
2012-01-30 12:43:15 +08:00
|
|
|
if (PRINT_FIELD(SYMOFFSET))
|
|
|
|
symbol__fprintf_symname_offs(al.sym, &al, stdout);
|
|
|
|
else
|
|
|
|
symbol__fprintf_symname(al.sym, stdout);
|
2011-05-31 03:08:23 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
if (PRINT_FIELD(DSO)) {
|
2012-01-30 12:42:57 +08:00
|
|
|
printf(" (");
|
|
|
|
map__fprintf_dsoname(al.map, stdout);
|
|
|
|
printf(")");
|
2011-05-31 03:08:23 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
perf script: Add callindent option
Based on patches from Andi Kleen.
When printing PT instruction traces with perf script it is rather useful
to see some indentation for the call tree. This patch adds a new
callindent field to perf script that prints spaces for the function call
stack depth.
We already have code to track the function call stack for PT, that we
can reuse with minor modifications.
The resulting output is not quite as nice as ftrace yet, but a lot
better than what was there before.
Note there are some corner cases when the thread stack gets code
confused and prints incorrect indentation. Even with that it is fairly
useful.
When displaying kernel code traces it is recommended to run as root, as
otherwise perf doesn't understand the kernel addresses properly, and may
not reset the call stack correctly on kernel boundaries.
Example output:
sudo perf-with-kcore record eg2 -a -e intel_pt// -- sleep 1
sudo perf-with-kcore script eg2 --ns -F callindent,time,comm,pid,sym,ip,addr,flags,cpu --itrace=cre | less
...
swapper 0 [000] 5830.389116586: call irq_exit ffffffff8104d620 smp_call_function_single_interrupt+0x30 => ffffffff8107e720 irq_exit
swapper 0 [000] 5830.389116586: call idle_cpu ffffffff8107e769 irq_exit+0x49 => ffffffff810a3970 idle_cpu
swapper 0 [000] 5830.389116586: return idle_cpu ffffffff810a39b7 idle_cpu+0x47 => ffffffff8107e76e irq_exit
swapper 0 [000] 5830.389116586: call tick_nohz_irq_exit ffffffff8107e7bd irq_exit+0x9d => ffffffff810f2fc0 tick_nohz_irq_exit
swapper 0 [000] 5830.389116919: call __tick_nohz_idle_enter ffffffff810f2fe0 tick_nohz_irq_exit+0x20 => ffffffff810f28d0 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call ktime_get ffffffff810f28f1 __tick_nohz_idle_enter+0x21 => ffffffff810e9ec0 ktime_get
swapper 0 [000] 5830.389116919: call read_tsc ffffffff810e9ef6 ktime_get+0x36 => ffffffff81035070 read_tsc
swapper 0 [000] 5830.389116919: return read_tsc ffffffff81035084 read_tsc+0x14 => ffffffff810e9efc ktime_get
swapper 0 [000] 5830.389116919: return ktime_get ffffffff810e9f46 ktime_get+0x86 => ffffffff810f28f6 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call sched_clock_idle_sleep_event ffffffff810f290b __tick_nohz_idle_enter+0x3b => ffffffff810a7380 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: call sched_clock_cpu ffffffff810a738b sched_clock_idle_sleep_event+0xb => ffffffff810a72e0 sched_clock_cpu
swapper 0 [000] 5830.389116919: call sched_clock ffffffff810a734d sched_clock_cpu+0x6d => ffffffff81035750 sched_clock
swapper 0 [000] 5830.389116919: call native_sched_clock ffffffff81035754 sched_clock+0x4 => ffffffff81035640 native_sched_clock
swapper 0 [000] 5830.389116919: return native_sched_clock ffffffff8103568c native_sched_clock+0x4c => ffffffff81035759 sched_clock
swapper 0 [000] 5830.389116919: return sched_clock ffffffff8103575c sched_clock+0xc => ffffffff810a7352 sched_clock_cpu
swapper 0 [000] 5830.389116919: return sched_clock_cpu ffffffff810a7356 sched_clock_cpu+0x76 => ffffffff810a7390 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: return sched_clock_idle_sleep_event ffffffff810a7391 sched_clock_idle_sleep_event+0x11 => ffffffff810f2910 __tick_nohz_idle_enter
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-4-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:58 +08:00
|
|
|
static void print_sample_callindent(struct perf_sample *sample,
|
|
|
|
struct perf_evsel *evsel,
|
|
|
|
struct thread *thread,
|
|
|
|
struct addr_location *al)
|
|
|
|
{
|
|
|
|
struct perf_event_attr *attr = &evsel->attr;
|
|
|
|
size_t depth = thread_stack__depth(thread);
|
|
|
|
struct addr_location addr_al;
|
|
|
|
const char *name = NULL;
|
|
|
|
static int spacing;
|
|
|
|
int len = 0;
|
|
|
|
u64 ip = 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The 'return' has already been popped off the stack so the depth has
|
|
|
|
* to be adjusted to match the 'call'.
|
|
|
|
*/
|
|
|
|
if (thread->ts && sample->flags & PERF_IP_FLAG_RETURN)
|
|
|
|
depth += 1;
|
|
|
|
|
|
|
|
if (sample->flags & (PERF_IP_FLAG_CALL | PERF_IP_FLAG_TRACE_BEGIN)) {
|
|
|
|
if (sample_addr_correlates_sym(attr)) {
|
|
|
|
thread__resolve(thread, &addr_al, sample);
|
|
|
|
if (addr_al.sym)
|
|
|
|
name = addr_al.sym->name;
|
|
|
|
else
|
|
|
|
ip = sample->addr;
|
|
|
|
} else {
|
|
|
|
ip = sample->addr;
|
|
|
|
}
|
|
|
|
} else if (sample->flags & (PERF_IP_FLAG_RETURN | PERF_IP_FLAG_TRACE_END)) {
|
|
|
|
if (al->sym)
|
|
|
|
name = al->sym->name;
|
|
|
|
else
|
|
|
|
ip = sample->ip;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (name)
|
|
|
|
len = printf("%*s%s", (int)depth * 4, "", name);
|
|
|
|
else if (ip)
|
|
|
|
len = printf("%*s%16" PRIx64, (int)depth * 4, "", ip);
|
|
|
|
|
|
|
|
if (len < 0)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to keep the output length from changing frequently so that the
|
|
|
|
* output lines up more nicely.
|
|
|
|
*/
|
|
|
|
if (len > spacing || (len && len < spacing - 52))
|
|
|
|
spacing = round_up(len + 4, 32);
|
|
|
|
|
|
|
|
if (len < spacing)
|
|
|
|
printf("%*s", spacing - len, "");
|
|
|
|
}
|
|
|
|
|
2016-03-23 22:55:24 +08:00
|
|
|
static void print_sample_bts(struct perf_sample *sample,
|
2012-01-30 12:43:09 +08:00
|
|
|
struct perf_evsel *evsel,
|
2013-12-04 22:16:36 +08:00
|
|
|
struct thread *thread,
|
|
|
|
struct addr_location *al)
|
2012-01-30 12:43:09 +08:00
|
|
|
{
|
|
|
|
struct perf_event_attr *attr = &evsel->attr;
|
2014-07-22 21:17:15 +08:00
|
|
|
bool print_srcline_last = false;
|
2012-01-30 12:43:09 +08:00
|
|
|
|
perf script: Add callindent option
Based on patches from Andi Kleen.
When printing PT instruction traces with perf script it is rather useful
to see some indentation for the call tree. This patch adds a new
callindent field to perf script that prints spaces for the function call
stack depth.
We already have code to track the function call stack for PT, that we
can reuse with minor modifications.
The resulting output is not quite as nice as ftrace yet, but a lot
better than what was there before.
Note there are some corner cases when the thread stack gets code
confused and prints incorrect indentation. Even with that it is fairly
useful.
When displaying kernel code traces it is recommended to run as root, as
otherwise perf doesn't understand the kernel addresses properly, and may
not reset the call stack correctly on kernel boundaries.
Example output:
sudo perf-with-kcore record eg2 -a -e intel_pt// -- sleep 1
sudo perf-with-kcore script eg2 --ns -F callindent,time,comm,pid,sym,ip,addr,flags,cpu --itrace=cre | less
...
swapper 0 [000] 5830.389116586: call irq_exit ffffffff8104d620 smp_call_function_single_interrupt+0x30 => ffffffff8107e720 irq_exit
swapper 0 [000] 5830.389116586: call idle_cpu ffffffff8107e769 irq_exit+0x49 => ffffffff810a3970 idle_cpu
swapper 0 [000] 5830.389116586: return idle_cpu ffffffff810a39b7 idle_cpu+0x47 => ffffffff8107e76e irq_exit
swapper 0 [000] 5830.389116586: call tick_nohz_irq_exit ffffffff8107e7bd irq_exit+0x9d => ffffffff810f2fc0 tick_nohz_irq_exit
swapper 0 [000] 5830.389116919: call __tick_nohz_idle_enter ffffffff810f2fe0 tick_nohz_irq_exit+0x20 => ffffffff810f28d0 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call ktime_get ffffffff810f28f1 __tick_nohz_idle_enter+0x21 => ffffffff810e9ec0 ktime_get
swapper 0 [000] 5830.389116919: call read_tsc ffffffff810e9ef6 ktime_get+0x36 => ffffffff81035070 read_tsc
swapper 0 [000] 5830.389116919: return read_tsc ffffffff81035084 read_tsc+0x14 => ffffffff810e9efc ktime_get
swapper 0 [000] 5830.389116919: return ktime_get ffffffff810e9f46 ktime_get+0x86 => ffffffff810f28f6 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call sched_clock_idle_sleep_event ffffffff810f290b __tick_nohz_idle_enter+0x3b => ffffffff810a7380 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: call sched_clock_cpu ffffffff810a738b sched_clock_idle_sleep_event+0xb => ffffffff810a72e0 sched_clock_cpu
swapper 0 [000] 5830.389116919: call sched_clock ffffffff810a734d sched_clock_cpu+0x6d => ffffffff81035750 sched_clock
swapper 0 [000] 5830.389116919: call native_sched_clock ffffffff81035754 sched_clock+0x4 => ffffffff81035640 native_sched_clock
swapper 0 [000] 5830.389116919: return native_sched_clock ffffffff8103568c native_sched_clock+0x4c => ffffffff81035759 sched_clock
swapper 0 [000] 5830.389116919: return sched_clock ffffffff8103575c sched_clock+0xc => ffffffff810a7352 sched_clock_cpu
swapper 0 [000] 5830.389116919: return sched_clock_cpu ffffffff810a7356 sched_clock_cpu+0x76 => ffffffff810a7390 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: return sched_clock_idle_sleep_event ffffffff810a7391 sched_clock_idle_sleep_event+0x11 => ffffffff810f2910 __tick_nohz_idle_enter
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-4-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:58 +08:00
|
|
|
if (PRINT_FIELD(CALLINDENT))
|
|
|
|
print_sample_callindent(sample, evsel, thread, al);
|
|
|
|
|
2012-01-30 12:43:09 +08:00
|
|
|
/* print branch_from information */
|
|
|
|
if (PRINT_FIELD(IP)) {
|
2014-07-22 21:17:15 +08:00
|
|
|
unsigned int print_opts = output[attr->type].print_ip_opts;
|
2016-04-20 10:32:11 +08:00
|
|
|
struct callchain_cursor *cursor = NULL;
|
2014-07-22 21:17:15 +08:00
|
|
|
|
2016-04-15 04:45:51 +08:00
|
|
|
if (symbol_conf.use_callchain && sample->callchain &&
|
2016-04-20 10:32:11 +08:00
|
|
|
thread__resolve_callchain(al->thread, &callchain_cursor, evsel,
|
2016-04-15 04:45:51 +08:00
|
|
|
sample, NULL, NULL, scripting_max_stack) == 0)
|
2016-04-20 10:32:11 +08:00
|
|
|
cursor = &callchain_cursor;
|
2016-04-15 04:45:51 +08:00
|
|
|
|
|
|
|
if (cursor == NULL) {
|
|
|
|
putchar(' ');
|
2016-04-13 02:16:15 +08:00
|
|
|
if (print_opts & EVSEL__PRINT_SRCLINE) {
|
2014-07-22 21:17:15 +08:00
|
|
|
print_srcline_last = true;
|
2016-04-13 02:16:15 +08:00
|
|
|
print_opts &= ~EVSEL__PRINT_SRCLINE;
|
2014-07-22 21:17:15 +08:00
|
|
|
}
|
2016-04-15 04:45:51 +08:00
|
|
|
} else
|
|
|
|
putchar('\n');
|
|
|
|
|
|
|
|
sample__fprintf_sym(sample, al, 0, print_opts, cursor, stdout);
|
2012-01-30 12:43:09 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* print branch_to information */
|
2013-10-18 20:29:14 +08:00
|
|
|
if (PRINT_FIELD(ADDR) ||
|
|
|
|
((evsel->attr.sample_type & PERF_SAMPLE_ADDR) &&
|
2014-07-22 21:17:16 +08:00
|
|
|
!output[attr->type].user_set)) {
|
|
|
|
printf(" => ");
|
2016-03-23 22:55:24 +08:00
|
|
|
print_sample_addr(sample, thread, attr);
|
2014-07-22 21:17:16 +08:00
|
|
|
}
|
2012-01-30 12:43:09 +08:00
|
|
|
|
2014-07-22 21:17:15 +08:00
|
|
|
if (print_srcline_last)
|
|
|
|
map__fprintf_srcline(al->map, al->addr, "\n ", stdout);
|
|
|
|
|
2012-01-30 12:43:09 +08:00
|
|
|
printf("\n");
|
|
|
|
}
|
|
|
|
|
perf script: Print sample flags more nicely
The flags field is synthesized and may have a value when Instruction
Trace decoding. The flags are "bcrosyiABEx" which stand for branch,
call, return, conditional, system, asynchronous, interrupt, transaction
abort, trace begin, trace end, and in transaction, respectively.
Change the display so that known combinations of flags are printed more
nicely e.g.: "call" for "bc", "return" for "br", "jcc" for "bo", "jmp"
for "b", "int" for "bci", "iret" for "bri", "syscall" for "bcs",
"sysret" for "brs", "async" for "by", "hw int" for "bcyi", "tx abrt" for
"bA", "tr strt" for "bB", "tr end" for "bE".
However the "x" flag will be displayed separately in those cases e.g.
"jcc (x)" for a condition branch within a transaction.
Example:
perf record -e intel_pt//u ls
perf script --ns -F comm,cpu,pid,tid,time,ip,addr,sym,dso,symoff,flags
...
ls 3689/3689 [001] 2062.020965237: jcc 7f06a958847a _dl_sysdep_start+0xfa (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9588450 _dl_sysdep_start+0xd0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965237: jmp 7f06a9588461 _dl_sysdep_start+0xe1 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95885a0 _dl_sysdep_start+0x220 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965237: jmp 7f06a95885a4 _dl_sysdep_start+0x224 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9588470 _dl_sysdep_start+0xf0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965904: call 7f06a95884c3 _dl_sysdep_start+0x143 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9589140 brk+0x0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965904: syscall 7f06a958914a brk+0xa (/lib/x86_64-linux-gnu/ld-2.19.so) => 0 [unknown] ([unknown])
ls 3689/3689 [001] 2062.020966237: tr strt 0 [unknown] ([unknown]) => 7f06a958914c brk+0xc (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: return 7f06a9589165 brk+0x25 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95884c8 _dl_sysdep_start+0x148 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: jcc 7f06a95884d7 _dl_sysdep_start+0x157 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95885f0 _dl_sysdep_start+0x270 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: call 7f06a95885f0 _dl_sysdep_start+0x270 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a958ac50 strlen+0x0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: jcc 7f06a958ac6e strlen+0x1e (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a958ac60 strlen+0x10 (/lib/x86_64-linux-gnu/ld-2.19.so)
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-2-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:56 +08:00
|
|
|
static struct {
|
|
|
|
u32 flags;
|
|
|
|
const char *name;
|
|
|
|
} sample_flags[] = {
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CALL, "call"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_RETURN, "return"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CONDITIONAL, "jcc"},
|
|
|
|
{PERF_IP_FLAG_BRANCH, "jmp"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CALL | PERF_IP_FLAG_INTERRUPT, "int"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_RETURN | PERF_IP_FLAG_INTERRUPT, "iret"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CALL | PERF_IP_FLAG_SYSCALLRET, "syscall"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_RETURN | PERF_IP_FLAG_SYSCALLRET, "sysret"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_ASYNC, "async"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CALL | PERF_IP_FLAG_ASYNC | PERF_IP_FLAG_INTERRUPT, "hw int"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_TX_ABORT, "tx abrt"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_TRACE_BEGIN, "tr strt"},
|
|
|
|
{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_TRACE_END, "tr end"},
|
|
|
|
{0, NULL}
|
|
|
|
};
|
|
|
|
|
perf script: Add field option 'flags' to print sample flags
Instruction tracing will typically have access to information about the
instruction being executed for a particular ip sample. Some of that
information will be available in the 'flags' member of struct
perf_sample.
With the addition of transactions events synthesis to Instruction
Tracing options, there is a need to be able easily to see the flags
because they show whether the ip is at the start, commit or abort of a
tranasaction.
Consequently add an option to display the flags.
The flags are "bcrosyiABEx" which stand for branch, call, return,
conditional, system, asynchronous, interrupt, transaction abort, trace
begin, trace end, and in transaction, respectively.
Example using Intel PT:
perf script -fip,time,event,sym,addr,flags
...
1288.721584105: branches:u: bo 401146 main => 401152 main
1288.721584105: transactions: x 0 401164 main
1288.721584105: branches:u: bx 40117c main => 40119b main
1288.721584105: branches:u: box 4011a4 main => 40117e main
1288.721584105: branches:u: bcx 401187 main => 401094 g
...
1288.721591645: branches:u: bx 4010c4 g => 4010cb g
1288.721591645: branches:u: brx 4010cc g => 401189 main
1288.721591645: transactions: 0 4011a6 main
1288.721593199: branches:u: b 4011a9 main => 4011af main
1288.721593199: branches:u: bo 4011bc main => 40113e main
1288.721593199: branches:u: b 401150 main => 40115a main
1288.721593199: transactions: x 0 401164 main
1288.721593199: branches:u: bx 40117c main => 40119b main
1288.721593199: branches:u: box 4011a4 main => 40117e main
1288.721593199: branches:u: bcx 401187 main => 40105e f
...
1288.722284747: branches:u: brx 401093 f => 401189 main
1288.722284747: branches:u: box 4011a4 main => 40117e main
1288.722284747: branches:u: bcx 401187 main => 40105e f
1288.722285883: transactions: bA 0 401071 f
1288.722285883: branches:u: bA 401071 f => 40116a main
1288.722285883: branches:u: bE 40116a main => 0 [unknown]
1288.722297174: branches:u: bB 0 [unknown] => 40116a main
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1428594864-29309-26-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-09 23:54:05 +08:00
|
|
|
static void print_sample_flags(u32 flags)
|
|
|
|
{
|
|
|
|
const char *chars = PERF_IP_FLAG_CHARS;
|
|
|
|
const int n = strlen(PERF_IP_FLAG_CHARS);
|
perf script: Print sample flags more nicely
The flags field is synthesized and may have a value when Instruction
Trace decoding. The flags are "bcrosyiABEx" which stand for branch,
call, return, conditional, system, asynchronous, interrupt, transaction
abort, trace begin, trace end, and in transaction, respectively.
Change the display so that known combinations of flags are printed more
nicely e.g.: "call" for "bc", "return" for "br", "jcc" for "bo", "jmp"
for "b", "int" for "bci", "iret" for "bri", "syscall" for "bcs",
"sysret" for "brs", "async" for "by", "hw int" for "bcyi", "tx abrt" for
"bA", "tr strt" for "bB", "tr end" for "bE".
However the "x" flag will be displayed separately in those cases e.g.
"jcc (x)" for a condition branch within a transaction.
Example:
perf record -e intel_pt//u ls
perf script --ns -F comm,cpu,pid,tid,time,ip,addr,sym,dso,symoff,flags
...
ls 3689/3689 [001] 2062.020965237: jcc 7f06a958847a _dl_sysdep_start+0xfa (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9588450 _dl_sysdep_start+0xd0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965237: jmp 7f06a9588461 _dl_sysdep_start+0xe1 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95885a0 _dl_sysdep_start+0x220 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965237: jmp 7f06a95885a4 _dl_sysdep_start+0x224 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9588470 _dl_sysdep_start+0xf0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965904: call 7f06a95884c3 _dl_sysdep_start+0x143 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9589140 brk+0x0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965904: syscall 7f06a958914a brk+0xa (/lib/x86_64-linux-gnu/ld-2.19.so) => 0 [unknown] ([unknown])
ls 3689/3689 [001] 2062.020966237: tr strt 0 [unknown] ([unknown]) => 7f06a958914c brk+0xc (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: return 7f06a9589165 brk+0x25 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95884c8 _dl_sysdep_start+0x148 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: jcc 7f06a95884d7 _dl_sysdep_start+0x157 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95885f0 _dl_sysdep_start+0x270 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: call 7f06a95885f0 _dl_sysdep_start+0x270 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a958ac50 strlen+0x0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: jcc 7f06a958ac6e strlen+0x1e (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a958ac60 strlen+0x10 (/lib/x86_64-linux-gnu/ld-2.19.so)
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-2-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:56 +08:00
|
|
|
bool in_tx = flags & PERF_IP_FLAG_IN_TX;
|
|
|
|
const char *name = NULL;
|
perf script: Add field option 'flags' to print sample flags
Instruction tracing will typically have access to information about the
instruction being executed for a particular ip sample. Some of that
information will be available in the 'flags' member of struct
perf_sample.
With the addition of transactions events synthesis to Instruction
Tracing options, there is a need to be able easily to see the flags
because they show whether the ip is at the start, commit or abort of a
tranasaction.
Consequently add an option to display the flags.
The flags are "bcrosyiABEx" which stand for branch, call, return,
conditional, system, asynchronous, interrupt, transaction abort, trace
begin, trace end, and in transaction, respectively.
Example using Intel PT:
perf script -fip,time,event,sym,addr,flags
...
1288.721584105: branches:u: bo 401146 main => 401152 main
1288.721584105: transactions: x 0 401164 main
1288.721584105: branches:u: bx 40117c main => 40119b main
1288.721584105: branches:u: box 4011a4 main => 40117e main
1288.721584105: branches:u: bcx 401187 main => 401094 g
...
1288.721591645: branches:u: bx 4010c4 g => 4010cb g
1288.721591645: branches:u: brx 4010cc g => 401189 main
1288.721591645: transactions: 0 4011a6 main
1288.721593199: branches:u: b 4011a9 main => 4011af main
1288.721593199: branches:u: bo 4011bc main => 40113e main
1288.721593199: branches:u: b 401150 main => 40115a main
1288.721593199: transactions: x 0 401164 main
1288.721593199: branches:u: bx 40117c main => 40119b main
1288.721593199: branches:u: box 4011a4 main => 40117e main
1288.721593199: branches:u: bcx 401187 main => 40105e f
...
1288.722284747: branches:u: brx 401093 f => 401189 main
1288.722284747: branches:u: box 4011a4 main => 40117e main
1288.722284747: branches:u: bcx 401187 main => 40105e f
1288.722285883: transactions: bA 0 401071 f
1288.722285883: branches:u: bA 401071 f => 40116a main
1288.722285883: branches:u: bE 40116a main => 0 [unknown]
1288.722297174: branches:u: bB 0 [unknown] => 40116a main
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1428594864-29309-26-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-09 23:54:05 +08:00
|
|
|
char str[33];
|
|
|
|
int i, pos = 0;
|
|
|
|
|
perf script: Print sample flags more nicely
The flags field is synthesized and may have a value when Instruction
Trace decoding. The flags are "bcrosyiABEx" which stand for branch,
call, return, conditional, system, asynchronous, interrupt, transaction
abort, trace begin, trace end, and in transaction, respectively.
Change the display so that known combinations of flags are printed more
nicely e.g.: "call" for "bc", "return" for "br", "jcc" for "bo", "jmp"
for "b", "int" for "bci", "iret" for "bri", "syscall" for "bcs",
"sysret" for "brs", "async" for "by", "hw int" for "bcyi", "tx abrt" for
"bA", "tr strt" for "bB", "tr end" for "bE".
However the "x" flag will be displayed separately in those cases e.g.
"jcc (x)" for a condition branch within a transaction.
Example:
perf record -e intel_pt//u ls
perf script --ns -F comm,cpu,pid,tid,time,ip,addr,sym,dso,symoff,flags
...
ls 3689/3689 [001] 2062.020965237: jcc 7f06a958847a _dl_sysdep_start+0xfa (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9588450 _dl_sysdep_start+0xd0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965237: jmp 7f06a9588461 _dl_sysdep_start+0xe1 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95885a0 _dl_sysdep_start+0x220 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965237: jmp 7f06a95885a4 _dl_sysdep_start+0x224 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9588470 _dl_sysdep_start+0xf0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965904: call 7f06a95884c3 _dl_sysdep_start+0x143 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9589140 brk+0x0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965904: syscall 7f06a958914a brk+0xa (/lib/x86_64-linux-gnu/ld-2.19.so) => 0 [unknown] ([unknown])
ls 3689/3689 [001] 2062.020966237: tr strt 0 [unknown] ([unknown]) => 7f06a958914c brk+0xc (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: return 7f06a9589165 brk+0x25 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95884c8 _dl_sysdep_start+0x148 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: jcc 7f06a95884d7 _dl_sysdep_start+0x157 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95885f0 _dl_sysdep_start+0x270 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: call 7f06a95885f0 _dl_sysdep_start+0x270 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a958ac50 strlen+0x0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: jcc 7f06a958ac6e strlen+0x1e (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a958ac60 strlen+0x10 (/lib/x86_64-linux-gnu/ld-2.19.so)
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-2-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:56 +08:00
|
|
|
for (i = 0; sample_flags[i].name ; i++) {
|
|
|
|
if (sample_flags[i].flags == (flags & ~PERF_IP_FLAG_IN_TX)) {
|
|
|
|
name = sample_flags[i].name;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
perf script: Add field option 'flags' to print sample flags
Instruction tracing will typically have access to information about the
instruction being executed for a particular ip sample. Some of that
information will be available in the 'flags' member of struct
perf_sample.
With the addition of transactions events synthesis to Instruction
Tracing options, there is a need to be able easily to see the flags
because they show whether the ip is at the start, commit or abort of a
tranasaction.
Consequently add an option to display the flags.
The flags are "bcrosyiABEx" which stand for branch, call, return,
conditional, system, asynchronous, interrupt, transaction abort, trace
begin, trace end, and in transaction, respectively.
Example using Intel PT:
perf script -fip,time,event,sym,addr,flags
...
1288.721584105: branches:u: bo 401146 main => 401152 main
1288.721584105: transactions: x 0 401164 main
1288.721584105: branches:u: bx 40117c main => 40119b main
1288.721584105: branches:u: box 4011a4 main => 40117e main
1288.721584105: branches:u: bcx 401187 main => 401094 g
...
1288.721591645: branches:u: bx 4010c4 g => 4010cb g
1288.721591645: branches:u: brx 4010cc g => 401189 main
1288.721591645: transactions: 0 4011a6 main
1288.721593199: branches:u: b 4011a9 main => 4011af main
1288.721593199: branches:u: bo 4011bc main => 40113e main
1288.721593199: branches:u: b 401150 main => 40115a main
1288.721593199: transactions: x 0 401164 main
1288.721593199: branches:u: bx 40117c main => 40119b main
1288.721593199: branches:u: box 4011a4 main => 40117e main
1288.721593199: branches:u: bcx 401187 main => 40105e f
...
1288.722284747: branches:u: brx 401093 f => 401189 main
1288.722284747: branches:u: box 4011a4 main => 40117e main
1288.722284747: branches:u: bcx 401187 main => 40105e f
1288.722285883: transactions: bA 0 401071 f
1288.722285883: branches:u: bA 401071 f => 40116a main
1288.722285883: branches:u: bE 40116a main => 0 [unknown]
1288.722297174: branches:u: bB 0 [unknown] => 40116a main
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1428594864-29309-26-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-09 23:54:05 +08:00
|
|
|
for (i = 0; i < n; i++, flags >>= 1) {
|
|
|
|
if (flags & 1)
|
|
|
|
str[pos++] = chars[i];
|
|
|
|
}
|
|
|
|
for (; i < 32; i++, flags >>= 1) {
|
|
|
|
if (flags & 1)
|
|
|
|
str[pos++] = '?';
|
|
|
|
}
|
|
|
|
str[pos] = 0;
|
perf script: Print sample flags more nicely
The flags field is synthesized and may have a value when Instruction
Trace decoding. The flags are "bcrosyiABEx" which stand for branch,
call, return, conditional, system, asynchronous, interrupt, transaction
abort, trace begin, trace end, and in transaction, respectively.
Change the display so that known combinations of flags are printed more
nicely e.g.: "call" for "bc", "return" for "br", "jcc" for "bo", "jmp"
for "b", "int" for "bci", "iret" for "bri", "syscall" for "bcs",
"sysret" for "brs", "async" for "by", "hw int" for "bcyi", "tx abrt" for
"bA", "tr strt" for "bB", "tr end" for "bE".
However the "x" flag will be displayed separately in those cases e.g.
"jcc (x)" for a condition branch within a transaction.
Example:
perf record -e intel_pt//u ls
perf script --ns -F comm,cpu,pid,tid,time,ip,addr,sym,dso,symoff,flags
...
ls 3689/3689 [001] 2062.020965237: jcc 7f06a958847a _dl_sysdep_start+0xfa (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9588450 _dl_sysdep_start+0xd0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965237: jmp 7f06a9588461 _dl_sysdep_start+0xe1 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95885a0 _dl_sysdep_start+0x220 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965237: jmp 7f06a95885a4 _dl_sysdep_start+0x224 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9588470 _dl_sysdep_start+0xf0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965904: call 7f06a95884c3 _dl_sysdep_start+0x143 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a9589140 brk+0x0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020965904: syscall 7f06a958914a brk+0xa (/lib/x86_64-linux-gnu/ld-2.19.so) => 0 [unknown] ([unknown])
ls 3689/3689 [001] 2062.020966237: tr strt 0 [unknown] ([unknown]) => 7f06a958914c brk+0xc (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: return 7f06a9589165 brk+0x25 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95884c8 _dl_sysdep_start+0x148 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: jcc 7f06a95884d7 _dl_sysdep_start+0x157 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a95885f0 _dl_sysdep_start+0x270 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: call 7f06a95885f0 _dl_sysdep_start+0x270 (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a958ac50 strlen+0x0 (/lib/x86_64-linux-gnu/ld-2.19.so)
ls 3689/3689 [001] 2062.020966237: jcc 7f06a958ac6e strlen+0x1e (/lib/x86_64-linux-gnu/ld-2.19.so) => 7f06a958ac60 strlen+0x10 (/lib/x86_64-linux-gnu/ld-2.19.so)
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-2-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:56 +08:00
|
|
|
|
|
|
|
if (name)
|
|
|
|
printf(" %-7s%4s ", name, in_tx ? "(x)" : "");
|
|
|
|
else
|
|
|
|
printf(" %-11s ", str);
|
perf script: Add field option 'flags' to print sample flags
Instruction tracing will typically have access to information about the
instruction being executed for a particular ip sample. Some of that
information will be available in the 'flags' member of struct
perf_sample.
With the addition of transactions events synthesis to Instruction
Tracing options, there is a need to be able easily to see the flags
because they show whether the ip is at the start, commit or abort of a
tranasaction.
Consequently add an option to display the flags.
The flags are "bcrosyiABEx" which stand for branch, call, return,
conditional, system, asynchronous, interrupt, transaction abort, trace
begin, trace end, and in transaction, respectively.
Example using Intel PT:
perf script -fip,time,event,sym,addr,flags
...
1288.721584105: branches:u: bo 401146 main => 401152 main
1288.721584105: transactions: x 0 401164 main
1288.721584105: branches:u: bx 40117c main => 40119b main
1288.721584105: branches:u: box 4011a4 main => 40117e main
1288.721584105: branches:u: bcx 401187 main => 401094 g
...
1288.721591645: branches:u: bx 4010c4 g => 4010cb g
1288.721591645: branches:u: brx 4010cc g => 401189 main
1288.721591645: transactions: 0 4011a6 main
1288.721593199: branches:u: b 4011a9 main => 4011af main
1288.721593199: branches:u: bo 4011bc main => 40113e main
1288.721593199: branches:u: b 401150 main => 40115a main
1288.721593199: transactions: x 0 401164 main
1288.721593199: branches:u: bx 40117c main => 40119b main
1288.721593199: branches:u: box 4011a4 main => 40117e main
1288.721593199: branches:u: bcx 401187 main => 40105e f
...
1288.722284747: branches:u: brx 401093 f => 401189 main
1288.722284747: branches:u: box 4011a4 main => 40117e main
1288.722284747: branches:u: bcx 401187 main => 40105e f
1288.722285883: transactions: bA 0 401071 f
1288.722285883: branches:u: bA 401071 f => 40116a main
1288.722285883: branches:u: bE 40116a main => 0 [unknown]
1288.722297174: branches:u: bB 0 [unknown] => 40116a main
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1428594864-29309-26-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-09 23:54:05 +08:00
|
|
|
}
|
|
|
|
|
perf script: Print bpf-output events in 'perf script'
This patch allows 'perf script' output messages from BPF program. For
example, use test_bpf_output_3.c at the end of this commit message,
# ./perf record -e bpf-output/no-inherit,name=evt/ \
-e ./test_bpf_output_3.c/map:channel.event=evt/ \
usleep 100000
# ./perf script
usleep 4882 21384.532523: evt: ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
usleep 4882 21384.632606: evt: ffffffff8105c609 kretprobe_trampoline_holder ([kernel.kallsyms
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
Two samples from BPF output are printed by both binary and string
format.
If BPF program output something unprintable, string format is
suppressed.
/************************ BEGIN **************************/
#include <uapi/linux/bpf.h>
struct bpf_map_def {
unsigned int type;
unsigned int key_size;
unsigned int value_size;
unsigned int max_entries;
};
#define SEC(NAME) __attribute__((section(NAME), used))
static u64 (*ktime_get_ns)(void) =
(void *)BPF_FUNC_ktime_get_ns;
static int (*trace_printk)(const char *fmt, int fmt_size, ...) =
(void *)BPF_FUNC_trace_printk;
static int (*get_smp_processor_id)(void) =
(void *)BPF_FUNC_get_smp_processor_id;
static int (*perf_event_output)(void *, struct bpf_map_def *, int, void *, unsigned long) =
(void *)BPF_FUNC_perf_event_output;
struct bpf_map_def SEC("maps") channel = {
.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
.key_size = sizeof(int),
.value_size = sizeof(u32),
.max_entries = __NR_CPUS__,
};
static inline int __attribute__((always_inline))
func(void *ctx, int type)
{
char output_str[] = "Raise a BPF event!";
perf_event_output(ctx, &channel, get_smp_processor_id(),
&output_str, sizeof(output_str));
return 0;
}
SEC("func_begin=sys_nanosleep")
int func_begin(void *ctx) {return func(ctx, 1);}
SEC("func_end=sys_nanosleep%return")
int func_end(void *ctx) { return func(ctx, 2);}
char _license[] SEC("license") = "GPL";
int _version SEC("version") = LINUX_VERSION_CODE;
/************************* END ***************************/
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456312845-111583-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-02-24 19:20:45 +08:00
|
|
|
struct printer_data {
|
|
|
|
int line_no;
|
|
|
|
bool hit_nul;
|
|
|
|
bool is_printable;
|
|
|
|
};
|
|
|
|
|
|
|
|
static void
|
|
|
|
print_sample_bpf_output_printer(enum binary_printer_ops op,
|
|
|
|
unsigned int val,
|
|
|
|
void *extra)
|
|
|
|
{
|
|
|
|
unsigned char ch = (unsigned char)val;
|
|
|
|
struct printer_data *printer_data = extra;
|
|
|
|
|
|
|
|
switch (op) {
|
|
|
|
case BINARY_PRINT_DATA_BEGIN:
|
|
|
|
printf("\n");
|
|
|
|
break;
|
|
|
|
case BINARY_PRINT_LINE_BEGIN:
|
|
|
|
printf("%17s", !printer_data->line_no ? "BPF output:" :
|
|
|
|
" ");
|
|
|
|
break;
|
|
|
|
case BINARY_PRINT_ADDR:
|
|
|
|
printf(" %04x:", val);
|
|
|
|
break;
|
|
|
|
case BINARY_PRINT_NUM_DATA:
|
|
|
|
printf(" %02x", val);
|
|
|
|
break;
|
|
|
|
case BINARY_PRINT_NUM_PAD:
|
|
|
|
printf(" ");
|
|
|
|
break;
|
|
|
|
case BINARY_PRINT_SEP:
|
|
|
|
printf(" ");
|
|
|
|
break;
|
|
|
|
case BINARY_PRINT_CHAR_DATA:
|
|
|
|
if (printer_data->hit_nul && ch)
|
|
|
|
printer_data->is_printable = false;
|
|
|
|
|
|
|
|
if (!isprint(ch)) {
|
|
|
|
printf("%c", '.');
|
|
|
|
|
|
|
|
if (!printer_data->is_printable)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (ch == '\0')
|
|
|
|
printer_data->hit_nul = true;
|
|
|
|
else
|
|
|
|
printer_data->is_printable = false;
|
|
|
|
} else {
|
|
|
|
printf("%c", ch);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case BINARY_PRINT_CHAR_PAD:
|
|
|
|
printf(" ");
|
|
|
|
break;
|
|
|
|
case BINARY_PRINT_LINE_END:
|
|
|
|
printf("\n");
|
|
|
|
printer_data->line_no++;
|
|
|
|
break;
|
|
|
|
case BINARY_PRINT_DATA_END:
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void print_sample_bpf_output(struct perf_sample *sample)
|
|
|
|
{
|
|
|
|
unsigned int nr_bytes = sample->raw_size;
|
|
|
|
struct printer_data printer_data = {0, false, true};
|
|
|
|
|
|
|
|
print_binary(sample->raw_data, nr_bytes, 8,
|
|
|
|
print_sample_bpf_output_printer, &printer_data);
|
|
|
|
|
|
|
|
if (printer_data.is_printable && printer_data.hit_nul)
|
|
|
|
printf("%17s \"%s\"\n", "BPF string:",
|
|
|
|
(char *)(sample->raw_data));
|
|
|
|
}
|
|
|
|
|
2015-11-27 01:55:21 +08:00
|
|
|
struct perf_script {
|
|
|
|
struct perf_tool tool;
|
|
|
|
struct perf_session *session;
|
|
|
|
bool show_task_events;
|
|
|
|
bool show_mmap_events;
|
|
|
|
bool show_switch_events;
|
2016-01-06 05:09:06 +08:00
|
|
|
bool allocated;
|
|
|
|
struct cpu_map *cpus;
|
|
|
|
struct thread_map *threads;
|
2016-01-07 17:14:05 +08:00
|
|
|
int name_width;
|
2015-11-27 01:55:21 +08:00
|
|
|
};
|
|
|
|
|
2016-01-07 17:14:05 +08:00
|
|
|
static int perf_evlist__max_name_len(struct perf_evlist *evlist)
|
|
|
|
{
|
|
|
|
struct perf_evsel *evsel;
|
|
|
|
int max = 0;
|
|
|
|
|
2016-06-23 22:26:15 +08:00
|
|
|
evlist__for_each_entry(evlist, evsel) {
|
2016-01-07 17:14:05 +08:00
|
|
|
int len = strlen(perf_evsel__name(evsel));
|
|
|
|
|
|
|
|
max = MAX(len, max);
|
|
|
|
}
|
|
|
|
|
|
|
|
return max;
|
|
|
|
}
|
|
|
|
|
2016-02-24 16:46:54 +08:00
|
|
|
static size_t data_src__printf(u64 data_src)
|
|
|
|
{
|
|
|
|
struct mem_info mi = { .data_src.val = data_src };
|
|
|
|
char decode[100];
|
|
|
|
char out[100];
|
|
|
|
static int maxlen;
|
|
|
|
int len;
|
|
|
|
|
|
|
|
perf_script__meminfo_scnprintf(decode, 100, &mi);
|
|
|
|
|
|
|
|
len = scnprintf(out, 100, "%16" PRIx64 " %s", data_src, decode);
|
|
|
|
if (maxlen < len)
|
|
|
|
maxlen = len;
|
|
|
|
|
|
|
|
return printf("%-*s", maxlen, out);
|
|
|
|
}
|
|
|
|
|
2016-03-23 22:55:24 +08:00
|
|
|
static void process_event(struct perf_script *script,
|
2015-11-27 01:55:21 +08:00
|
|
|
struct perf_sample *sample, struct perf_evsel *evsel,
|
|
|
|
struct addr_location *al)
|
2011-03-10 13:23:23 +08:00
|
|
|
{
|
2015-04-02 00:29:25 +08:00
|
|
|
struct thread *thread = al->thread;
|
2011-03-16 02:44:01 +08:00
|
|
|
struct perf_event_attr *attr = &evsel->attr;
|
2011-03-10 13:23:28 +08:00
|
|
|
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
if (output[attr->type].fields == 0)
|
2011-03-10 13:23:28 +08:00
|
|
|
return;
|
|
|
|
|
2012-08-07 20:58:03 +08:00
|
|
|
print_sample_start(sample, thread, evsel);
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
|
2014-08-25 22:45:42 +08:00
|
|
|
if (PRINT_FIELD(PERIOD))
|
|
|
|
printf("%10" PRIu64 " ", sample->period);
|
|
|
|
|
2013-11-18 13:34:52 +08:00
|
|
|
if (PRINT_FIELD(EVNAME)) {
|
|
|
|
const char *evname = perf_evsel__name(evsel);
|
2016-01-07 17:14:05 +08:00
|
|
|
|
|
|
|
if (!script->name_width)
|
|
|
|
script->name_width = perf_evlist__max_name_len(script->session->evlist);
|
|
|
|
|
|
|
|
printf("%*s: ", script->name_width,
|
|
|
|
evname ? evname : "[unknown]");
|
2013-11-18 13:34:52 +08:00
|
|
|
}
|
|
|
|
|
perf script: Add field option 'flags' to print sample flags
Instruction tracing will typically have access to information about the
instruction being executed for a particular ip sample. Some of that
information will be available in the 'flags' member of struct
perf_sample.
With the addition of transactions events synthesis to Instruction
Tracing options, there is a need to be able easily to see the flags
because they show whether the ip is at the start, commit or abort of a
tranasaction.
Consequently add an option to display the flags.
The flags are "bcrosyiABEx" which stand for branch, call, return,
conditional, system, asynchronous, interrupt, transaction abort, trace
begin, trace end, and in transaction, respectively.
Example using Intel PT:
perf script -fip,time,event,sym,addr,flags
...
1288.721584105: branches:u: bo 401146 main => 401152 main
1288.721584105: transactions: x 0 401164 main
1288.721584105: branches:u: bx 40117c main => 40119b main
1288.721584105: branches:u: box 4011a4 main => 40117e main
1288.721584105: branches:u: bcx 401187 main => 401094 g
...
1288.721591645: branches:u: bx 4010c4 g => 4010cb g
1288.721591645: branches:u: brx 4010cc g => 401189 main
1288.721591645: transactions: 0 4011a6 main
1288.721593199: branches:u: b 4011a9 main => 4011af main
1288.721593199: branches:u: bo 4011bc main => 40113e main
1288.721593199: branches:u: b 401150 main => 40115a main
1288.721593199: transactions: x 0 401164 main
1288.721593199: branches:u: bx 40117c main => 40119b main
1288.721593199: branches:u: box 4011a4 main => 40117e main
1288.721593199: branches:u: bcx 401187 main => 40105e f
...
1288.722284747: branches:u: brx 401093 f => 401189 main
1288.722284747: branches:u: box 4011a4 main => 40117e main
1288.722284747: branches:u: bcx 401187 main => 40105e f
1288.722285883: transactions: bA 0 401071 f
1288.722285883: branches:u: bA 401071 f => 40116a main
1288.722285883: branches:u: bE 40116a main => 0 [unknown]
1288.722297174: branches:u: bB 0 [unknown] => 40116a main
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1428594864-29309-26-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-09 23:54:05 +08:00
|
|
|
if (print_flags)
|
|
|
|
print_sample_flags(sample->flags);
|
|
|
|
|
2012-01-30 12:43:09 +08:00
|
|
|
if (is_bts_event(attr)) {
|
2016-03-23 22:55:24 +08:00
|
|
|
print_sample_bts(sample, evsel, thread, al);
|
2012-01-30 12:43:09 +08:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
if (PRINT_FIELD(TRACE))
|
2012-08-07 20:58:03 +08:00
|
|
|
event_format__print(evsel->tp_format, sample->cpu,
|
|
|
|
sample->raw_data, sample->raw_size);
|
2011-05-31 03:08:23 +08:00
|
|
|
if (PRINT_FIELD(ADDR))
|
2016-03-23 22:55:24 +08:00
|
|
|
print_sample_addr(sample, thread, attr);
|
2011-05-31 03:08:23 +08:00
|
|
|
|
2016-02-15 16:34:51 +08:00
|
|
|
if (PRINT_FIELD(DATA_SRC))
|
2016-02-24 16:46:54 +08:00
|
|
|
data_src__printf(sample->data_src);
|
2016-02-15 16:34:51 +08:00
|
|
|
|
|
|
|
if (PRINT_FIELD(WEIGHT))
|
|
|
|
printf("%16" PRIu64, sample->weight);
|
|
|
|
|
2011-05-28 04:28:43 +08:00
|
|
|
if (PRINT_FIELD(IP)) {
|
2016-04-20 10:32:11 +08:00
|
|
|
struct callchain_cursor *cursor = NULL;
|
2016-04-15 04:45:51 +08:00
|
|
|
|
2016-04-18 22:31:46 +08:00
|
|
|
if (symbol_conf.use_callchain && sample->callchain &&
|
2016-04-20 10:32:11 +08:00
|
|
|
thread__resolve_callchain(al->thread, &callchain_cursor, evsel,
|
2016-04-15 04:45:51 +08:00
|
|
|
sample, NULL, NULL, scripting_max_stack) == 0)
|
2016-04-20 10:32:11 +08:00
|
|
|
cursor = &callchain_cursor;
|
2013-08-08 10:50:51 +08:00
|
|
|
|
2016-04-15 04:45:51 +08:00
|
|
|
putchar(cursor ? '\n' : ' ');
|
|
|
|
sample__fprintf_sym(sample, al, 0, output[attr->type].print_ip_opts, cursor, stdout);
|
2011-03-10 13:23:27 +08:00
|
|
|
}
|
|
|
|
|
2015-09-01 00:41:10 +08:00
|
|
|
if (PRINT_FIELD(IREGS))
|
2016-03-23 22:55:24 +08:00
|
|
|
print_sample_iregs(sample, attr);
|
2015-09-01 00:41:10 +08:00
|
|
|
|
2015-09-01 00:41:13 +08:00
|
|
|
if (PRINT_FIELD(BRSTACK))
|
2016-03-23 22:55:24 +08:00
|
|
|
print_sample_brstack(sample);
|
2015-09-01 00:41:13 +08:00
|
|
|
else if (PRINT_FIELD(BRSTACKSYM))
|
2016-03-23 22:55:24 +08:00
|
|
|
print_sample_brstacksym(sample, thread);
|
2015-09-01 00:41:13 +08:00
|
|
|
|
perf script: Print bpf-output events in 'perf script'
This patch allows 'perf script' output messages from BPF program. For
example, use test_bpf_output_3.c at the end of this commit message,
# ./perf record -e bpf-output/no-inherit,name=evt/ \
-e ./test_bpf_output_3.c/map:channel.event=evt/ \
usleep 100000
# ./perf script
usleep 4882 21384.532523: evt: ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
usleep 4882 21384.632606: evt: ffffffff8105c609 kretprobe_trampoline_holder ([kernel.kallsyms
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
Two samples from BPF output are printed by both binary and string
format.
If BPF program output something unprintable, string format is
suppressed.
/************************ BEGIN **************************/
#include <uapi/linux/bpf.h>
struct bpf_map_def {
unsigned int type;
unsigned int key_size;
unsigned int value_size;
unsigned int max_entries;
};
#define SEC(NAME) __attribute__((section(NAME), used))
static u64 (*ktime_get_ns)(void) =
(void *)BPF_FUNC_ktime_get_ns;
static int (*trace_printk)(const char *fmt, int fmt_size, ...) =
(void *)BPF_FUNC_trace_printk;
static int (*get_smp_processor_id)(void) =
(void *)BPF_FUNC_get_smp_processor_id;
static int (*perf_event_output)(void *, struct bpf_map_def *, int, void *, unsigned long) =
(void *)BPF_FUNC_perf_event_output;
struct bpf_map_def SEC("maps") channel = {
.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
.key_size = sizeof(int),
.value_size = sizeof(u32),
.max_entries = __NR_CPUS__,
};
static inline int __attribute__((always_inline))
func(void *ctx, int type)
{
char output_str[] = "Raise a BPF event!";
perf_event_output(ctx, &channel, get_smp_processor_id(),
&output_str, sizeof(output_str));
return 0;
}
SEC("func_begin=sys_nanosleep")
int func_begin(void *ctx) {return func(ctx, 1);}
SEC("func_end=sys_nanosleep%return")
int func_end(void *ctx) { return func(ctx, 2);}
char _license[] SEC("license") = "GPL";
int _version SEC("version") = LINUX_VERSION_CODE;
/************************* END ***************************/
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456312845-111583-3-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-02-24 19:20:45 +08:00
|
|
|
if (perf_evsel__is_bpf_output(evsel) && PRINT_FIELD(BPF_OUTPUT))
|
|
|
|
print_sample_bpf_output(sample);
|
|
|
|
|
2011-03-10 13:23:25 +08:00
|
|
|
printf("\n");
|
2011-03-10 13:23:23 +08:00
|
|
|
}
|
|
|
|
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
static struct scripting_ops *scripting_ops;
|
|
|
|
|
2016-01-06 18:49:56 +08:00
|
|
|
static void __process_stat(struct perf_evsel *counter, u64 tstamp)
|
|
|
|
{
|
|
|
|
int nthreads = thread_map__nr(counter->threads);
|
|
|
|
int ncpus = perf_evsel__nr_cpus(counter);
|
|
|
|
int cpu, thread;
|
|
|
|
static int header_printed;
|
|
|
|
|
|
|
|
if (counter->system_wide)
|
|
|
|
nthreads = 1;
|
|
|
|
|
|
|
|
if (!header_printed) {
|
|
|
|
printf("%3s %8s %15s %15s %15s %15s %s\n",
|
|
|
|
"CPU", "THREAD", "VAL", "ENA", "RUN", "TIME", "EVENT");
|
|
|
|
header_printed = 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (thread = 0; thread < nthreads; thread++) {
|
|
|
|
for (cpu = 0; cpu < ncpus; cpu++) {
|
|
|
|
struct perf_counts_values *counts;
|
|
|
|
|
|
|
|
counts = perf_counts(counter->counts, cpu, thread);
|
|
|
|
|
|
|
|
printf("%3d %8d %15" PRIu64 " %15" PRIu64 " %15" PRIu64 " %15" PRIu64 " %s\n",
|
|
|
|
counter->cpus->map[cpu],
|
|
|
|
thread_map__pid(counter->threads, thread),
|
|
|
|
counts->val,
|
|
|
|
counts->ena,
|
|
|
|
counts->run,
|
|
|
|
tstamp,
|
|
|
|
perf_evsel__name(counter));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-01-06 05:09:09 +08:00
|
|
|
static void process_stat(struct perf_evsel *counter, u64 tstamp)
|
|
|
|
{
|
|
|
|
if (scripting_ops && scripting_ops->process_stat)
|
|
|
|
scripting_ops->process_stat(&stat_config, counter, tstamp);
|
2016-01-06 18:49:56 +08:00
|
|
|
else
|
|
|
|
__process_stat(counter, tstamp);
|
2016-01-06 05:09:09 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static void process_stat_interval(u64 tstamp)
|
|
|
|
{
|
|
|
|
if (scripting_ops && scripting_ops->process_stat_interval)
|
|
|
|
scripting_ops->process_stat_interval(tstamp);
|
|
|
|
}
|
|
|
|
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
static void setup_scripting(void)
|
|
|
|
{
|
2009-11-25 15:15:48 +08:00
|
|
|
setup_perl_scripting();
|
2010-01-27 16:27:57 +08:00
|
|
|
setup_python_scripting();
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
}
|
|
|
|
|
2014-08-16 03:08:37 +08:00
|
|
|
static int flush_scripting(void)
|
|
|
|
{
|
2015-11-26 21:55:23 +08:00
|
|
|
return scripting_ops ? scripting_ops->flush_script() : 0;
|
2014-08-16 03:08:37 +08:00
|
|
|
}
|
|
|
|
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
static int cleanup_scripting(void)
|
|
|
|
{
|
2010-11-17 01:45:39 +08:00
|
|
|
pr_debug("\nperf script stopped\n");
|
2010-05-10 12:46:57 +08:00
|
|
|
|
2015-11-26 21:55:23 +08:00
|
|
|
return scripting_ops ? scripting_ops->stop_script() : 0;
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
}
|
|
|
|
|
2015-11-27 01:55:21 +08:00
|
|
|
static int process_sample_event(struct perf_tool *tool,
|
2011-11-25 18:19:45 +08:00
|
|
|
union perf_event *event,
|
2011-01-30 00:01:45 +08:00
|
|
|
struct perf_sample *sample,
|
2011-03-16 02:44:01 +08:00
|
|
|
struct perf_evsel *evsel,
|
2011-11-28 17:56:39 +08:00
|
|
|
struct machine *machine)
|
2009-08-17 22:18:08 +08:00
|
|
|
{
|
2015-11-27 01:55:21 +08:00
|
|
|
struct perf_script *scr = container_of(tool, struct perf_script, tool);
|
2011-11-22 01:02:52 +08:00
|
|
|
struct addr_location al;
|
2009-08-17 22:18:08 +08:00
|
|
|
|
2011-03-10 13:23:28 +08:00
|
|
|
if (debug_mode) {
|
|
|
|
if (sample->time < last_timestamp) {
|
|
|
|
pr_err("Samples misordered, previous: %" PRIu64
|
|
|
|
" this: %" PRIu64 "\n", last_timestamp,
|
|
|
|
sample->time);
|
|
|
|
nr_unordered++;
|
2010-04-24 07:55:09 +08:00
|
|
|
}
|
2011-03-10 13:23:28 +08:00
|
|
|
last_timestamp = sample->time;
|
|
|
|
return 0;
|
2009-08-17 22:18:08 +08:00
|
|
|
}
|
2011-07-04 19:57:50 +08:00
|
|
|
|
2016-03-23 05:39:09 +08:00
|
|
|
if (machine__resolve(machine, &al, sample) < 0) {
|
2011-11-22 01:02:52 +08:00
|
|
|
pr_err("problem processing %d event, skipping it.\n",
|
|
|
|
event->header.type);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (al.filtered)
|
perf machine: Protect the machine->threads with a rwlock
In addition to using refcounts for the struct thread lifetime
management, we need to protect access to machine->threads from
concurrent access.
That happens in 'perf top', where a thread processes events, inserting
and deleting entries from that rb_tree while another thread decays
hist_entries, that end up dropping references and ultimately deleting
threads from the rb_tree and releasing its resources when no further
hist_entry (or other data structures, like in 'perf sched') references
it.
So the rule is the same for refcounts + protected trees in the kernel,
get the tree lock, find object, bump the refcount, drop the tree lock,
return, use object, drop the refcount if no more use of it is needed,
keep it if storing it in some other data structure, drop when releasing
that data structure.
I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and
"perf_event__preprocess_sample(&al)" with "addr_location__put(&al)".
The addr_location__put() one is because as we return references to
several data structures, we may end up adding more reference counting
for the other data structures and then we'll drop it at
addr_location__put() time.
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
|
|
|
goto out_put;
|
2011-11-22 01:02:52 +08:00
|
|
|
|
2011-07-04 19:57:50 +08:00
|
|
|
if (cpu_list && !test_bit(sample->cpu, cpu_bitmap))
|
perf machine: Protect the machine->threads with a rwlock
In addition to using refcounts for the struct thread lifetime
management, we need to protect access to machine->threads from
concurrent access.
That happens in 'perf top', where a thread processes events, inserting
and deleting entries from that rb_tree while another thread decays
hist_entries, that end up dropping references and ultimately deleting
threads from the rb_tree and releasing its resources when no further
hist_entry (or other data structures, like in 'perf sched') references
it.
So the rule is the same for refcounts + protected trees in the kernel,
get the tree lock, find object, bump the refcount, drop the tree lock,
return, use object, drop the refcount if no more use of it is needed,
keep it if storing it in some other data structure, drop when releasing
that data structure.
I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and
"perf_event__preprocess_sample(&al)" with "addr_location__put(&al)".
The addr_location__put() one is because as we return references to
several data structures, we may end up adding more reference counting
for the other data structures and then we'll drop it at
addr_location__put() time.
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
|
|
|
goto out_put;
|
2011-07-04 19:57:50 +08:00
|
|
|
|
2015-11-26 21:55:23 +08:00
|
|
|
if (scripting_ops)
|
|
|
|
scripting_ops->process_event(event, sample, evsel, &al);
|
|
|
|
else
|
2016-03-23 22:55:24 +08:00
|
|
|
process_event(scr, sample, evsel, &al);
|
2015-11-26 21:55:23 +08:00
|
|
|
|
perf machine: Protect the machine->threads with a rwlock
In addition to using refcounts for the struct thread lifetime
management, we need to protect access to machine->threads from
concurrent access.
That happens in 'perf top', where a thread processes events, inserting
and deleting entries from that rb_tree while another thread decays
hist_entries, that end up dropping references and ultimately deleting
threads from the rb_tree and releasing its resources when no further
hist_entry (or other data structures, like in 'perf sched') references
it.
So the rule is the same for refcounts + protected trees in the kernel,
get the tree lock, find object, bump the refcount, drop the tree lock,
return, use object, drop the refcount if no more use of it is needed,
keep it if storing it in some other data structure, drop when releasing
that data structure.
I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and
"perf_event__preprocess_sample(&al)" with "addr_location__put(&al)".
The addr_location__put() one is because as we return references to
several data structures, we may end up adding more reference counting
for the other data structures and then we'll drop it at
addr_location__put() time.
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
|
|
|
out_put:
|
|
|
|
addr_location__put(&al);
|
2009-08-17 22:18:08 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-11-01 21:51:30 +08:00
|
|
|
static int process_attr(struct perf_tool *tool, union perf_event *event,
|
|
|
|
struct perf_evlist **pevlist)
|
|
|
|
{
|
|
|
|
struct perf_script *scr = container_of(tool, struct perf_script, tool);
|
|
|
|
struct perf_evlist *evlist;
|
|
|
|
struct perf_evsel *evsel, *pos;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = perf_event__process_attr(tool, event, pevlist);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
evlist = *pevlist;
|
|
|
|
evsel = perf_evlist__last(*pevlist);
|
|
|
|
|
|
|
|
if (evsel->attr.type >= PERF_TYPE_MAX)
|
|
|
|
return 0;
|
|
|
|
|
2016-06-23 22:26:15 +08:00
|
|
|
evlist__for_each_entry(evlist, pos) {
|
2013-11-01 21:51:30 +08:00
|
|
|
if (pos->attr.type == evsel->attr.type && pos != evsel)
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
set_print_ip_opts(&evsel->attr);
|
|
|
|
|
2015-10-16 18:41:25 +08:00
|
|
|
if (evsel->attr.sample_type)
|
|
|
|
err = perf_evsel__check_attr(evsel, scr->session);
|
|
|
|
|
|
|
|
return err;
|
2013-11-01 21:51:30 +08:00
|
|
|
}
|
|
|
|
|
2013-11-26 16:51:12 +08:00
|
|
|
static int process_comm_event(struct perf_tool *tool,
|
|
|
|
union perf_event *event,
|
|
|
|
struct perf_sample *sample,
|
|
|
|
struct machine *machine)
|
|
|
|
{
|
|
|
|
struct thread *thread;
|
|
|
|
struct perf_script *script = container_of(tool, struct perf_script, tool);
|
|
|
|
struct perf_session *session = script->session;
|
2015-07-21 17:44:05 +08:00
|
|
|
struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
|
2013-11-26 16:51:12 +08:00
|
|
|
int ret = -1;
|
|
|
|
|
|
|
|
thread = machine__findnew_thread(machine, event->comm.pid, event->comm.tid);
|
|
|
|
if (thread == NULL) {
|
|
|
|
pr_debug("problem processing COMM event, skipping it.\n");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (perf_event__process_comm(tool, event, sample, machine) < 0)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (!evsel->attr.sample_id_all) {
|
|
|
|
sample->cpu = 0;
|
|
|
|
sample->time = 0;
|
|
|
|
sample->tid = event->comm.tid;
|
|
|
|
sample->pid = event->comm.pid;
|
|
|
|
}
|
|
|
|
print_sample_start(sample, thread, evsel);
|
|
|
|
perf_event__fprintf(event, stdout);
|
|
|
|
ret = 0;
|
|
|
|
out:
|
perf machine: Protect the machine->threads with a rwlock
In addition to using refcounts for the struct thread lifetime
management, we need to protect access to machine->threads from
concurrent access.
That happens in 'perf top', where a thread processes events, inserting
and deleting entries from that rb_tree while another thread decays
hist_entries, that end up dropping references and ultimately deleting
threads from the rb_tree and releasing its resources when no further
hist_entry (or other data structures, like in 'perf sched') references
it.
So the rule is the same for refcounts + protected trees in the kernel,
get the tree lock, find object, bump the refcount, drop the tree lock,
return, use object, drop the refcount if no more use of it is needed,
keep it if storing it in some other data structure, drop when releasing
that data structure.
I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and
"perf_event__preprocess_sample(&al)" with "addr_location__put(&al)".
The addr_location__put() one is because as we return references to
several data structures, we may end up adding more reference counting
for the other data structures and then we'll drop it at
addr_location__put() time.
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
|
|
|
thread__put(thread);
|
2013-11-26 16:51:12 +08:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int process_fork_event(struct perf_tool *tool,
|
|
|
|
union perf_event *event,
|
|
|
|
struct perf_sample *sample,
|
|
|
|
struct machine *machine)
|
|
|
|
{
|
|
|
|
struct thread *thread;
|
|
|
|
struct perf_script *script = container_of(tool, struct perf_script, tool);
|
|
|
|
struct perf_session *session = script->session;
|
2015-07-21 17:44:05 +08:00
|
|
|
struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
|
2013-11-26 16:51:12 +08:00
|
|
|
|
|
|
|
if (perf_event__process_fork(tool, event, sample, machine) < 0)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
thread = machine__findnew_thread(machine, event->fork.pid, event->fork.tid);
|
|
|
|
if (thread == NULL) {
|
|
|
|
pr_debug("problem processing FORK event, skipping it.\n");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!evsel->attr.sample_id_all) {
|
|
|
|
sample->cpu = 0;
|
|
|
|
sample->time = event->fork.time;
|
|
|
|
sample->tid = event->fork.tid;
|
|
|
|
sample->pid = event->fork.pid;
|
|
|
|
}
|
|
|
|
print_sample_start(sample, thread, evsel);
|
|
|
|
perf_event__fprintf(event, stdout);
|
perf machine: Protect the machine->threads with a rwlock
In addition to using refcounts for the struct thread lifetime
management, we need to protect access to machine->threads from
concurrent access.
That happens in 'perf top', where a thread processes events, inserting
and deleting entries from that rb_tree while another thread decays
hist_entries, that end up dropping references and ultimately deleting
threads from the rb_tree and releasing its resources when no further
hist_entry (or other data structures, like in 'perf sched') references
it.
So the rule is the same for refcounts + protected trees in the kernel,
get the tree lock, find object, bump the refcount, drop the tree lock,
return, use object, drop the refcount if no more use of it is needed,
keep it if storing it in some other data structure, drop when releasing
that data structure.
I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and
"perf_event__preprocess_sample(&al)" with "addr_location__put(&al)".
The addr_location__put() one is because as we return references to
several data structures, we may end up adding more reference counting
for the other data structures and then we'll drop it at
addr_location__put() time.
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
|
|
|
thread__put(thread);
|
2013-11-26 16:51:12 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
static int process_exit_event(struct perf_tool *tool,
|
|
|
|
union perf_event *event,
|
|
|
|
struct perf_sample *sample,
|
|
|
|
struct machine *machine)
|
|
|
|
{
|
perf machine: Protect the machine->threads with a rwlock
In addition to using refcounts for the struct thread lifetime
management, we need to protect access to machine->threads from
concurrent access.
That happens in 'perf top', where a thread processes events, inserting
and deleting entries from that rb_tree while another thread decays
hist_entries, that end up dropping references and ultimately deleting
threads from the rb_tree and releasing its resources when no further
hist_entry (or other data structures, like in 'perf sched') references
it.
So the rule is the same for refcounts + protected trees in the kernel,
get the tree lock, find object, bump the refcount, drop the tree lock,
return, use object, drop the refcount if no more use of it is needed,
keep it if storing it in some other data structure, drop when releasing
that data structure.
I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and
"perf_event__preprocess_sample(&al)" with "addr_location__put(&al)".
The addr_location__put() one is because as we return references to
several data structures, we may end up adding more reference counting
for the other data structures and then we'll drop it at
addr_location__put() time.
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
|
|
|
int err = 0;
|
2013-11-26 16:51:12 +08:00
|
|
|
struct thread *thread;
|
|
|
|
struct perf_script *script = container_of(tool, struct perf_script, tool);
|
|
|
|
struct perf_session *session = script->session;
|
2015-07-21 17:44:05 +08:00
|
|
|
struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
|
2013-11-26 16:51:12 +08:00
|
|
|
|
|
|
|
thread = machine__findnew_thread(machine, event->fork.pid, event->fork.tid);
|
|
|
|
if (thread == NULL) {
|
|
|
|
pr_debug("problem processing EXIT event, skipping it.\n");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!evsel->attr.sample_id_all) {
|
|
|
|
sample->cpu = 0;
|
|
|
|
sample->time = 0;
|
2015-08-18 17:07:05 +08:00
|
|
|
sample->tid = event->fork.tid;
|
|
|
|
sample->pid = event->fork.pid;
|
2013-11-26 16:51:12 +08:00
|
|
|
}
|
|
|
|
print_sample_start(sample, thread, evsel);
|
|
|
|
perf_event__fprintf(event, stdout);
|
|
|
|
|
|
|
|
if (perf_event__process_exit(tool, event, sample, machine) < 0)
|
perf machine: Protect the machine->threads with a rwlock
In addition to using refcounts for the struct thread lifetime
management, we need to protect access to machine->threads from
concurrent access.
That happens in 'perf top', where a thread processes events, inserting
and deleting entries from that rb_tree while another thread decays
hist_entries, that end up dropping references and ultimately deleting
threads from the rb_tree and releasing its resources when no further
hist_entry (or other data structures, like in 'perf sched') references
it.
So the rule is the same for refcounts + protected trees in the kernel,
get the tree lock, find object, bump the refcount, drop the tree lock,
return, use object, drop the refcount if no more use of it is needed,
keep it if storing it in some other data structure, drop when releasing
that data structure.
I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and
"perf_event__preprocess_sample(&al)" with "addr_location__put(&al)".
The addr_location__put() one is because as we return references to
several data structures, we may end up adding more reference counting
for the other data structures and then we'll drop it at
addr_location__put() time.
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
|
|
|
err = -1;
|
2013-11-26 16:51:12 +08:00
|
|
|
|
perf machine: Protect the machine->threads with a rwlock
In addition to using refcounts for the struct thread lifetime
management, we need to protect access to machine->threads from
concurrent access.
That happens in 'perf top', where a thread processes events, inserting
and deleting entries from that rb_tree while another thread decays
hist_entries, that end up dropping references and ultimately deleting
threads from the rb_tree and releasing its resources when no further
hist_entry (or other data structures, like in 'perf sched') references
it.
So the rule is the same for refcounts + protected trees in the kernel,
get the tree lock, find object, bump the refcount, drop the tree lock,
return, use object, drop the refcount if no more use of it is needed,
keep it if storing it in some other data structure, drop when releasing
that data structure.
I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and
"perf_event__preprocess_sample(&al)" with "addr_location__put(&al)".
The addr_location__put() one is because as we return references to
several data structures, we may end up adding more reference counting
for the other data structures and then we'll drop it at
addr_location__put() time.
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
|
|
|
thread__put(thread);
|
|
|
|
return err;
|
2013-11-26 16:51:12 +08:00
|
|
|
}
|
|
|
|
|
2013-11-26 16:54:26 +08:00
|
|
|
static int process_mmap_event(struct perf_tool *tool,
|
|
|
|
union perf_event *event,
|
|
|
|
struct perf_sample *sample,
|
|
|
|
struct machine *machine)
|
|
|
|
{
|
|
|
|
struct thread *thread;
|
|
|
|
struct perf_script *script = container_of(tool, struct perf_script, tool);
|
|
|
|
struct perf_session *session = script->session;
|
2015-07-21 17:44:05 +08:00
|
|
|
struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
|
2013-11-26 16:54:26 +08:00
|
|
|
|
|
|
|
if (perf_event__process_mmap(tool, event, sample, machine) < 0)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
thread = machine__findnew_thread(machine, event->mmap.pid, event->mmap.tid);
|
|
|
|
if (thread == NULL) {
|
|
|
|
pr_debug("problem processing MMAP event, skipping it.\n");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!evsel->attr.sample_id_all) {
|
|
|
|
sample->cpu = 0;
|
|
|
|
sample->time = 0;
|
|
|
|
sample->tid = event->mmap.tid;
|
|
|
|
sample->pid = event->mmap.pid;
|
|
|
|
}
|
|
|
|
print_sample_start(sample, thread, evsel);
|
|
|
|
perf_event__fprintf(event, stdout);
|
perf machine: Protect the machine->threads with a rwlock
In addition to using refcounts for the struct thread lifetime
management, we need to protect access to machine->threads from
concurrent access.
That happens in 'perf top', where a thread processes events, inserting
and deleting entries from that rb_tree while another thread decays
hist_entries, that end up dropping references and ultimately deleting
threads from the rb_tree and releasing its resources when no further
hist_entry (or other data structures, like in 'perf sched') references
it.
So the rule is the same for refcounts + protected trees in the kernel,
get the tree lock, find object, bump the refcount, drop the tree lock,
return, use object, drop the refcount if no more use of it is needed,
keep it if storing it in some other data structure, drop when releasing
that data structure.
I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and
"perf_event__preprocess_sample(&al)" with "addr_location__put(&al)".
The addr_location__put() one is because as we return references to
several data structures, we may end up adding more reference counting
for the other data structures and then we'll drop it at
addr_location__put() time.
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
|
|
|
thread__put(thread);
|
2013-11-26 16:54:26 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int process_mmap2_event(struct perf_tool *tool,
|
|
|
|
union perf_event *event,
|
|
|
|
struct perf_sample *sample,
|
|
|
|
struct machine *machine)
|
|
|
|
{
|
|
|
|
struct thread *thread;
|
|
|
|
struct perf_script *script = container_of(tool, struct perf_script, tool);
|
|
|
|
struct perf_session *session = script->session;
|
2015-07-21 17:44:05 +08:00
|
|
|
struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
|
2013-11-26 16:54:26 +08:00
|
|
|
|
|
|
|
if (perf_event__process_mmap2(tool, event, sample, machine) < 0)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
thread = machine__findnew_thread(machine, event->mmap2.pid, event->mmap2.tid);
|
|
|
|
if (thread == NULL) {
|
|
|
|
pr_debug("problem processing MMAP2 event, skipping it.\n");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!evsel->attr.sample_id_all) {
|
|
|
|
sample->cpu = 0;
|
|
|
|
sample->time = 0;
|
|
|
|
sample->tid = event->mmap2.tid;
|
|
|
|
sample->pid = event->mmap2.pid;
|
|
|
|
}
|
|
|
|
print_sample_start(sample, thread, evsel);
|
|
|
|
perf_event__fprintf(event, stdout);
|
perf machine: Protect the machine->threads with a rwlock
In addition to using refcounts for the struct thread lifetime
management, we need to protect access to machine->threads from
concurrent access.
That happens in 'perf top', where a thread processes events, inserting
and deleting entries from that rb_tree while another thread decays
hist_entries, that end up dropping references and ultimately deleting
threads from the rb_tree and releasing its resources when no further
hist_entry (or other data structures, like in 'perf sched') references
it.
So the rule is the same for refcounts + protected trees in the kernel,
get the tree lock, find object, bump the refcount, drop the tree lock,
return, use object, drop the refcount if no more use of it is needed,
keep it if storing it in some other data structure, drop when releasing
that data structure.
I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and
"perf_event__preprocess_sample(&al)" with "addr_location__put(&al)".
The addr_location__put() one is because as we return references to
several data structures, we may end up adding more reference counting
for the other data structures and then we'll drop it at
addr_location__put() time.
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
|
|
|
thread__put(thread);
|
2013-11-26 16:54:26 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2015-07-21 17:44:06 +08:00
|
|
|
static int process_switch_event(struct perf_tool *tool,
|
|
|
|
union perf_event *event,
|
|
|
|
struct perf_sample *sample,
|
|
|
|
struct machine *machine)
|
|
|
|
{
|
|
|
|
struct thread *thread;
|
|
|
|
struct perf_script *script = container_of(tool, struct perf_script, tool);
|
|
|
|
struct perf_session *session = script->session;
|
|
|
|
struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
|
|
|
|
|
|
|
|
if (perf_event__process_switch(tool, event, sample, machine) < 0)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
thread = machine__findnew_thread(machine, sample->pid,
|
|
|
|
sample->tid);
|
|
|
|
if (thread == NULL) {
|
|
|
|
pr_debug("problem processing SWITCH event, skipping it.\n");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
print_sample_start(sample, thread, evsel);
|
|
|
|
perf_event__fprintf(event, stdout);
|
|
|
|
thread__put(thread);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-09-11 06:15:03 +08:00
|
|
|
static void sig_handler(int sig __maybe_unused)
|
2010-04-02 12:59:18 +08:00
|
|
|
{
|
|
|
|
session_done = 1;
|
|
|
|
}
|
|
|
|
|
2013-10-22 15:34:07 +08:00
|
|
|
static int __cmd_script(struct perf_script *script)
|
2009-08-17 22:18:08 +08:00
|
|
|
{
|
2010-05-27 21:46:25 +08:00
|
|
|
int ret;
|
|
|
|
|
2010-04-02 12:59:18 +08:00
|
|
|
signal(SIGINT, sig_handler);
|
|
|
|
|
2013-11-26 16:51:12 +08:00
|
|
|
/* override event processing functions */
|
|
|
|
if (script->show_task_events) {
|
|
|
|
script->tool.comm = process_comm_event;
|
|
|
|
script->tool.fork = process_fork_event;
|
|
|
|
script->tool.exit = process_exit_event;
|
|
|
|
}
|
2013-11-26 16:54:26 +08:00
|
|
|
if (script->show_mmap_events) {
|
|
|
|
script->tool.mmap = process_mmap_event;
|
|
|
|
script->tool.mmap2 = process_mmap2_event;
|
|
|
|
}
|
2015-07-21 17:44:06 +08:00
|
|
|
if (script->show_switch_events)
|
|
|
|
script->tool.context_switch = process_switch_event;
|
2013-11-26 16:51:12 +08:00
|
|
|
|
2015-03-03 22:58:45 +08:00
|
|
|
ret = perf_session__process_events(script->session);
|
2010-05-27 21:46:25 +08:00
|
|
|
|
2011-01-05 02:27:30 +08:00
|
|
|
if (debug_mode)
|
2011-01-23 06:37:02 +08:00
|
|
|
pr_err("Misordered timestamps: %" PRIu64 "\n", nr_unordered);
|
2010-05-27 21:46:25 +08:00
|
|
|
|
|
|
|
return ret;
|
2009-08-17 22:18:08 +08:00
|
|
|
}
|
|
|
|
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
struct script_spec {
|
|
|
|
struct list_head node;
|
|
|
|
struct scripting_ops *ops;
|
|
|
|
char spec[0];
|
|
|
|
};
|
|
|
|
|
2011-01-05 02:32:52 +08:00
|
|
|
static LIST_HEAD(script_specs);
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
|
|
|
|
static struct script_spec *script_spec__new(const char *spec,
|
|
|
|
struct scripting_ops *ops)
|
|
|
|
{
|
|
|
|
struct script_spec *s = malloc(sizeof(*s) + strlen(spec) + 1);
|
|
|
|
|
|
|
|
if (s != NULL) {
|
|
|
|
strcpy(s->spec, spec);
|
|
|
|
s->ops = ops;
|
|
|
|
}
|
|
|
|
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void script_spec__add(struct script_spec *s)
|
|
|
|
{
|
|
|
|
list_add_tail(&s->node, &script_specs);
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct script_spec *script_spec__find(const char *spec)
|
|
|
|
{
|
|
|
|
struct script_spec *s;
|
|
|
|
|
|
|
|
list_for_each_entry(s, &script_specs, node)
|
|
|
|
if (strcasecmp(s->spec, spec) == 0)
|
|
|
|
return s;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
int script_spec_register(const char *spec, struct scripting_ops *ops)
|
|
|
|
{
|
|
|
|
struct script_spec *s;
|
|
|
|
|
|
|
|
s = script_spec__find(spec);
|
|
|
|
if (s)
|
|
|
|
return -1;
|
|
|
|
|
2016-02-25 23:13:10 +08:00
|
|
|
s = script_spec__new(spec, ops);
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
if (!s)
|
|
|
|
return -1;
|
2016-02-25 23:13:10 +08:00
|
|
|
else
|
|
|
|
script_spec__add(s);
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct scripting_ops *script_spec__lookup(const char *spec)
|
|
|
|
{
|
|
|
|
struct script_spec *s = script_spec__find(spec);
|
|
|
|
if (!s)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
return s->ops;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void list_available_languages(void)
|
|
|
|
{
|
|
|
|
struct script_spec *s;
|
|
|
|
|
|
|
|
fprintf(stderr, "\n");
|
|
|
|
fprintf(stderr, "Scripting language extensions (used in "
|
2010-11-17 01:45:39 +08:00
|
|
|
"perf script -s [spec:]script.[spec]):\n\n");
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
|
|
|
|
list_for_each_entry(s, &script_specs, node)
|
|
|
|
fprintf(stderr, " %-42s [%s]\n", s->spec, s->ops->name);
|
|
|
|
|
|
|
|
fprintf(stderr, "\n");
|
|
|
|
}
|
|
|
|
|
2012-09-11 06:15:03 +08:00
|
|
|
static int parse_scriptname(const struct option *opt __maybe_unused,
|
|
|
|
const char *str, int unset __maybe_unused)
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
{
|
|
|
|
char spec[PATH_MAX];
|
|
|
|
const char *script, *ext;
|
|
|
|
int len;
|
|
|
|
|
2010-01-27 16:27:52 +08:00
|
|
|
if (strcmp(str, "lang") == 0) {
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
list_available_languages();
|
2010-01-27 16:27:52 +08:00
|
|
|
exit(0);
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
script = strchr(str, ':');
|
|
|
|
if (script) {
|
|
|
|
len = script - str;
|
|
|
|
if (len >= PATH_MAX) {
|
|
|
|
fprintf(stderr, "invalid language specifier");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
strncpy(spec, str, len);
|
|
|
|
spec[len] = '\0';
|
|
|
|
scripting_ops = script_spec__lookup(spec);
|
|
|
|
if (!scripting_ops) {
|
|
|
|
fprintf(stderr, "invalid language specifier");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
script++;
|
|
|
|
} else {
|
|
|
|
script = str;
|
2010-10-10 23:11:02 +08:00
|
|
|
ext = strrchr(script, '.');
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
if (!ext) {
|
|
|
|
fprintf(stderr, "invalid script extension");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
scripting_ops = script_spec__lookup(++ext);
|
|
|
|
if (!scripting_ops) {
|
|
|
|
fprintf(stderr, "invalid script extension");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
script_name = strdup(script);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-09-11 06:15:03 +08:00
|
|
|
static int parse_output_fields(const struct option *opt __maybe_unused,
|
|
|
|
const char *arg, int unset __maybe_unused)
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
{
|
|
|
|
char *tok;
|
2012-12-21 03:11:19 +08:00
|
|
|
int i, imax = ARRAY_SIZE(all_output_options);
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
int j;
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
int rc = 0;
|
|
|
|
char *str = strdup(arg);
|
2011-03-10 13:23:28 +08:00
|
|
|
int type = -1;
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
|
|
|
|
if (!str)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
/* first word can state for which event type the user is specifying
|
|
|
|
* the fields. If no type exists, the specified fields apply to all
|
|
|
|
* event types found in the file minus the invalid fields for a type.
|
2011-03-10 13:23:28 +08:00
|
|
|
*/
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
tok = strchr(str, ':');
|
|
|
|
if (tok) {
|
|
|
|
*tok = '\0';
|
|
|
|
tok++;
|
|
|
|
if (!strcmp(str, "hw"))
|
|
|
|
type = PERF_TYPE_HARDWARE;
|
|
|
|
else if (!strcmp(str, "sw"))
|
|
|
|
type = PERF_TYPE_SOFTWARE;
|
|
|
|
else if (!strcmp(str, "trace"))
|
|
|
|
type = PERF_TYPE_TRACEPOINT;
|
2011-04-15 01:38:18 +08:00
|
|
|
else if (!strcmp(str, "raw"))
|
|
|
|
type = PERF_TYPE_RAW;
|
perf script: Add support for PERF_TYPE_BREAKPOINT
Useful for getting stack traces for hardware breakpoint events.
Test result:
Before this patch:
# ~/perf record -g -e mem:0x600980 ./sample
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.011 MB perf.data (12 samples) ]
# ~/perf script
# ~/perf script -F comm,tid,pid,time,event,ip,sym,dso
sample 22520/22520 97457.836294: mem:0x600980:
5a4ad8 __clear_user (/lib/modules/4.3.0-rc4+/build/vmlinux)
...
3f41ba sys_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
979395 return_from_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
7f1b59719cf7 [unknown] ([unknown])
sample 22520/22520 97457.836648: mem:0x600980:
532 main (/home/w00229757/DataBreakpoints/sample)
21bd5 __libc_start_main (/tmp/oxygen_root-root/lib64/libc-2.18.so)
...
After this patch:
# ~/perf script
sample 22520 97457.836294: mem:0x600980:
5a4ad8 __clear_user (/lib/modules/4.3.0-rc4+/build/vmlinux)
...
3f41ba sys_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
979395 return_from_execve (/lib/modules/4.3.0-rc4+/build/vmlinux)
7f1b59719cf7 [unknown] ([unknown])
sample 22520 97457.836648: mem:0x600980:
532 main (/home/w00229757/DataBreakpoints/sample)
21bd5 __libc_start_main (/tmp/oxygen_root-root/lib64/libc-2.18.so)
Committer note:
So, further testing, lets do it for a kernel global variable,
tcp_hashinfo:
# grep -w tcp_hashinfo /proc/kallsyms
ffffffff8202fc00 B tcp_hashinfo
#
Note: allow specifying mem:tcp_hashinfo:
# perf record -g -e mem:0xffffffff81c65ac0 -a
^C[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.790 MB perf.data ]
#
# perf evlist
mem:0xffffffff8202fc00
# perf evlist -v
mem:0xffffffff8202fc00: type: 5, size: 112, { sample_period, sample_freq }: 1, sample_type: IP|TID|TIME|CALLCHAIN|CPU, disabled: 1, inherit: 1, mmap: 1, comm: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, bp_type: 3, { bp_addr, config1 }: 0xffffffff8202fc00, { bp_len, config2 }: 0x4
#
Then, after this patch:
# perf script
swapper 0 [000] 171036.986988: mem:0xffffffff8202fc00:
8a0fb5 __inet_lookup_established (/lib/modules/4.3.0+/build/vmlinux)
8bc09d tcp_v4_early_demux (/lib/modules/4.3.0+/build/vmlinux)
896def ip_rcv_finish (/lib/modules/4.3.0+/build/vmlinux)
8976c2 ip_rcv (/lib/modules/4.3.0+/build/vmlinux)
855eba __netif_receive_skb_core (/lib/modules/4.3.0+/build/vmlinux)
8565d8 __netif_receive_skb (/lib/modules/4.3.0+/build/vmlinux)
8572a8 process_backlog (/lib/modules/4.3.0+/build/vmlinux)
856b11 net_rx_action (/lib/modules/4.3.0+/build/vmlinux)
2a284b __do_softirq (/lib/modules/4.3.0+/build/vmlinux)
2a2ba3 irq_exit (/lib/modules/4.3.0+/build/vmlinux)
96b7a4 do_IRQ (/lib/modules/4.3.0+/build/vmlinux)
969807 ret_from_intr (/lib/modules/4.3.0+/build/vmlinux)
804c27 cpuidle_enter (/lib/modules/4.3.0+/build/vmlinux)
2ded22 call_cpuidle (/lib/modules/4.3.0+/build/vmlinux)
2defb6 cpu_startup_entry (/lib/modules/4.3.0+/build/vmlinux)
95d5bc rest_init (/lib/modules/4.3.0+/build/vmlinux)
1163ffa start_kernel ([kernel.vmlinux].init.text)
11634d7 x86_64_start_reservations ([kernel.vmlinux].init.text)
1163623 x86_64_start_kernel ([kernel.vmlinux].init.text)
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1449541544-67621-16-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-12-08 10:25:43 +08:00
|
|
|
else if (!strcmp(str, "break"))
|
|
|
|
type = PERF_TYPE_BREAKPOINT;
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
else {
|
|
|
|
fprintf(stderr, "Invalid event type in field string.\n");
|
2011-11-25 18:38:40 +08:00
|
|
|
rc = -EINVAL;
|
|
|
|
goto out;
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
if (output[type].user_set)
|
|
|
|
pr_warning("Overriding previous field request for %s events.\n",
|
|
|
|
event_type(type));
|
|
|
|
|
|
|
|
output[type].fields = 0;
|
|
|
|
output[type].user_set = true;
|
2011-04-07 11:54:20 +08:00
|
|
|
output[type].wildcard_set = false;
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
|
|
|
|
} else {
|
|
|
|
tok = str;
|
|
|
|
if (strlen(str) == 0) {
|
|
|
|
fprintf(stderr,
|
|
|
|
"Cannot set fields to 'none' for all event types.\n");
|
|
|
|
rc = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (output_set_by_user())
|
|
|
|
pr_warning("Overriding previous field request for all events.\n");
|
|
|
|
|
|
|
|
for (j = 0; j < PERF_TYPE_MAX; ++j) {
|
|
|
|
output[j].fields = 0;
|
|
|
|
output[j].user_set = true;
|
2011-04-07 11:54:20 +08:00
|
|
|
output[j].wildcard_set = true;
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
}
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
}
|
|
|
|
|
perf script: Add field option 'flags' to print sample flags
Instruction tracing will typically have access to information about the
instruction being executed for a particular ip sample. Some of that
information will be available in the 'flags' member of struct
perf_sample.
With the addition of transactions events synthesis to Instruction
Tracing options, there is a need to be able easily to see the flags
because they show whether the ip is at the start, commit or abort of a
tranasaction.
Consequently add an option to display the flags.
The flags are "bcrosyiABEx" which stand for branch, call, return,
conditional, system, asynchronous, interrupt, transaction abort, trace
begin, trace end, and in transaction, respectively.
Example using Intel PT:
perf script -fip,time,event,sym,addr,flags
...
1288.721584105: branches:u: bo 401146 main => 401152 main
1288.721584105: transactions: x 0 401164 main
1288.721584105: branches:u: bx 40117c main => 40119b main
1288.721584105: branches:u: box 4011a4 main => 40117e main
1288.721584105: branches:u: bcx 401187 main => 401094 g
...
1288.721591645: branches:u: bx 4010c4 g => 4010cb g
1288.721591645: branches:u: brx 4010cc g => 401189 main
1288.721591645: transactions: 0 4011a6 main
1288.721593199: branches:u: b 4011a9 main => 4011af main
1288.721593199: branches:u: bo 4011bc main => 40113e main
1288.721593199: branches:u: b 401150 main => 40115a main
1288.721593199: transactions: x 0 401164 main
1288.721593199: branches:u: bx 40117c main => 40119b main
1288.721593199: branches:u: box 4011a4 main => 40117e main
1288.721593199: branches:u: bcx 401187 main => 40105e f
...
1288.722284747: branches:u: brx 401093 f => 401189 main
1288.722284747: branches:u: box 4011a4 main => 40117e main
1288.722284747: branches:u: bcx 401187 main => 40105e f
1288.722285883: transactions: bA 0 401071 f
1288.722285883: branches:u: bA 401071 f => 40116a main
1288.722285883: branches:u: bE 40116a main => 0 [unknown]
1288.722297174: branches:u: bB 0 [unknown] => 40116a main
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1428594864-29309-26-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-09 23:54:05 +08:00
|
|
|
for (tok = strtok(tok, ","); tok; tok = strtok(NULL, ",")) {
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
for (i = 0; i < imax; ++i) {
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
if (strcmp(tok, all_output_options[i].str) == 0)
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
break;
|
|
|
|
}
|
perf script: Add field option 'flags' to print sample flags
Instruction tracing will typically have access to information about the
instruction being executed for a particular ip sample. Some of that
information will be available in the 'flags' member of struct
perf_sample.
With the addition of transactions events synthesis to Instruction
Tracing options, there is a need to be able easily to see the flags
because they show whether the ip is at the start, commit or abort of a
tranasaction.
Consequently add an option to display the flags.
The flags are "bcrosyiABEx" which stand for branch, call, return,
conditional, system, asynchronous, interrupt, transaction abort, trace
begin, trace end, and in transaction, respectively.
Example using Intel PT:
perf script -fip,time,event,sym,addr,flags
...
1288.721584105: branches:u: bo 401146 main => 401152 main
1288.721584105: transactions: x 0 401164 main
1288.721584105: branches:u: bx 40117c main => 40119b main
1288.721584105: branches:u: box 4011a4 main => 40117e main
1288.721584105: branches:u: bcx 401187 main => 401094 g
...
1288.721591645: branches:u: bx 4010c4 g => 4010cb g
1288.721591645: branches:u: brx 4010cc g => 401189 main
1288.721591645: transactions: 0 4011a6 main
1288.721593199: branches:u: b 4011a9 main => 4011af main
1288.721593199: branches:u: bo 4011bc main => 40113e main
1288.721593199: branches:u: b 401150 main => 40115a main
1288.721593199: transactions: x 0 401164 main
1288.721593199: branches:u: bx 40117c main => 40119b main
1288.721593199: branches:u: box 4011a4 main => 40117e main
1288.721593199: branches:u: bcx 401187 main => 40105e f
...
1288.722284747: branches:u: brx 401093 f => 401189 main
1288.722284747: branches:u: box 4011a4 main => 40117e main
1288.722284747: branches:u: bcx 401187 main => 40105e f
1288.722285883: transactions: bA 0 401071 f
1288.722285883: branches:u: bA 401071 f => 40116a main
1288.722285883: branches:u: bE 40116a main => 0 [unknown]
1288.722297174: branches:u: bB 0 [unknown] => 40116a main
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1428594864-29309-26-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-09 23:54:05 +08:00
|
|
|
if (i == imax && strcmp(tok, "flags") == 0) {
|
|
|
|
print_flags = true;
|
|
|
|
continue;
|
|
|
|
}
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
if (i == imax) {
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
fprintf(stderr, "Invalid field requested.\n");
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
rc = -EINVAL;
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
goto out;
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
}
|
|
|
|
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
if (type == -1) {
|
|
|
|
/* add user option to all events types for
|
|
|
|
* which it is valid
|
|
|
|
*/
|
|
|
|
for (j = 0; j < PERF_TYPE_MAX; ++j) {
|
|
|
|
if (output[j].invalid_fields & all_output_options[i].field) {
|
|
|
|
pr_warning("\'%s\' not valid for %s events. Ignoring.\n",
|
|
|
|
all_output_options[i].str, event_type(j));
|
|
|
|
} else
|
|
|
|
output[j].fields |= all_output_options[i].field;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if (output[type].invalid_fields & all_output_options[i].field) {
|
|
|
|
fprintf(stderr, "\'%s\' not valid for %s events.\n",
|
|
|
|
all_output_options[i].str, event_type(type));
|
|
|
|
|
|
|
|
rc = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
output[type].fields |= all_output_options[i].field;
|
|
|
|
}
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
}
|
|
|
|
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
if (type >= 0) {
|
|
|
|
if (output[type].fields == 0) {
|
|
|
|
pr_debug("No fields requested for %s type. "
|
|
|
|
"Events will not be displayed.\n", event_type(type));
|
|
|
|
}
|
|
|
|
}
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
out:
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
free(str);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2010-11-22 00:09:39 +08:00
|
|
|
/* Helper function for filesystems that return a dent->d_type DT_UNKNOWN */
|
|
|
|
static int is_directory(const char *base_path, const struct dirent *dent)
|
|
|
|
{
|
|
|
|
char path[PATH_MAX];
|
|
|
|
struct stat st;
|
|
|
|
|
|
|
|
sprintf(path, "%s/%s", base_path, dent->d_name);
|
|
|
|
if (stat(path, &st))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
return S_ISDIR(st.st_mode);
|
|
|
|
}
|
|
|
|
|
2016-04-08 22:25:59 +08:00
|
|
|
#define for_each_lang(scripts_path, scripts_dir, lang_dirent) \
|
|
|
|
while ((lang_dirent = readdir(scripts_dir)) != NULL) \
|
|
|
|
if ((lang_dirent->d_type == DT_DIR || \
|
|
|
|
(lang_dirent->d_type == DT_UNKNOWN && \
|
|
|
|
is_directory(scripts_path, lang_dirent))) && \
|
|
|
|
(strcmp(lang_dirent->d_name, ".")) && \
|
|
|
|
(strcmp(lang_dirent->d_name, "..")))
|
|
|
|
|
|
|
|
#define for_each_script(lang_path, lang_dir, script_dirent) \
|
|
|
|
while ((script_dirent = readdir(lang_dir)) != NULL) \
|
|
|
|
if (script_dirent->d_type != DT_DIR && \
|
|
|
|
(script_dirent->d_type != DT_UNKNOWN || \
|
|
|
|
!is_directory(lang_path, script_dirent)))
|
2009-12-15 16:53:38 +08:00
|
|
|
|
|
|
|
|
|
|
|
#define RECORD_SUFFIX "-record"
|
|
|
|
#define REPORT_SUFFIX "-report"
|
|
|
|
|
|
|
|
struct script_desc {
|
|
|
|
struct list_head node;
|
|
|
|
char *name;
|
|
|
|
char *half_liner;
|
|
|
|
char *args;
|
|
|
|
};
|
|
|
|
|
2011-01-05 02:32:52 +08:00
|
|
|
static LIST_HEAD(script_descs);
|
2009-12-15 16:53:38 +08:00
|
|
|
|
|
|
|
static struct script_desc *script_desc__new(const char *name)
|
|
|
|
{
|
|
|
|
struct script_desc *s = zalloc(sizeof(*s));
|
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
if (s != NULL && name)
|
2009-12-15 16:53:38 +08:00
|
|
|
s->name = strdup(name);
|
|
|
|
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void script_desc__delete(struct script_desc *s)
|
|
|
|
{
|
2013-12-28 03:55:14 +08:00
|
|
|
zfree(&s->name);
|
|
|
|
zfree(&s->half_liner);
|
|
|
|
zfree(&s->args);
|
2009-12-15 16:53:38 +08:00
|
|
|
free(s);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void script_desc__add(struct script_desc *s)
|
|
|
|
{
|
|
|
|
list_add_tail(&s->node, &script_descs);
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct script_desc *script_desc__find(const char *name)
|
|
|
|
{
|
|
|
|
struct script_desc *s;
|
|
|
|
|
|
|
|
list_for_each_entry(s, &script_descs, node)
|
|
|
|
if (strcasecmp(s->name, name) == 0)
|
|
|
|
return s;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct script_desc *script_desc__findnew(const char *name)
|
|
|
|
{
|
|
|
|
struct script_desc *s = script_desc__find(name);
|
|
|
|
|
|
|
|
if (s)
|
|
|
|
return s;
|
|
|
|
|
|
|
|
s = script_desc__new(name);
|
|
|
|
if (!s)
|
|
|
|
goto out_delete_desc;
|
|
|
|
|
|
|
|
script_desc__add(s);
|
|
|
|
|
|
|
|
return s;
|
|
|
|
|
|
|
|
out_delete_desc:
|
|
|
|
script_desc__delete(s);
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2010-12-03 23:52:01 +08:00
|
|
|
static const char *ends_with(const char *str, const char *suffix)
|
2009-12-15 16:53:38 +08:00
|
|
|
{
|
|
|
|
size_t suffix_len = strlen(suffix);
|
2010-12-03 23:52:01 +08:00
|
|
|
const char *p = str;
|
2009-12-15 16:53:38 +08:00
|
|
|
|
|
|
|
if (strlen(str) > suffix_len) {
|
|
|
|
p = str + strlen(str) - suffix_len;
|
|
|
|
if (!strncmp(p, suffix, suffix_len))
|
|
|
|
return p;
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int read_script_info(struct script_desc *desc, const char *filename)
|
|
|
|
{
|
|
|
|
char line[BUFSIZ], *p;
|
|
|
|
FILE *fp;
|
|
|
|
|
|
|
|
fp = fopen(filename, "r");
|
|
|
|
if (!fp)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
while (fgets(line, sizeof(line), fp)) {
|
|
|
|
p = ltrim(line);
|
|
|
|
if (strlen(p) == 0)
|
|
|
|
continue;
|
|
|
|
if (*p != '#')
|
|
|
|
continue;
|
|
|
|
p++;
|
|
|
|
if (strlen(p) && *p == '!')
|
|
|
|
continue;
|
|
|
|
|
|
|
|
p = ltrim(p);
|
|
|
|
if (strlen(p) && p[strlen(p) - 1] == '\n')
|
|
|
|
p[strlen(p) - 1] = '\0';
|
|
|
|
|
|
|
|
if (!strncmp(p, "description:", strlen("description:"))) {
|
|
|
|
p += strlen("description:");
|
|
|
|
desc->half_liner = strdup(ltrim(p));
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!strncmp(p, "args:", strlen("args:"))) {
|
|
|
|
p += strlen("args:");
|
|
|
|
desc->args = strdup(ltrim(p));
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
fclose(fp);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-11-25 18:38:40 +08:00
|
|
|
static char *get_script_root(struct dirent *script_dirent, const char *suffix)
|
|
|
|
{
|
|
|
|
char *script_root, *str;
|
|
|
|
|
|
|
|
script_root = strdup(script_dirent->d_name);
|
|
|
|
if (!script_root)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
str = (char *)ends_with(script_root, suffix);
|
|
|
|
if (!str) {
|
|
|
|
free(script_root);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
*str = '\0';
|
|
|
|
return script_root;
|
|
|
|
}
|
|
|
|
|
2012-09-11 06:15:03 +08:00
|
|
|
static int list_available_scripts(const struct option *opt __maybe_unused,
|
|
|
|
const char *s __maybe_unused,
|
|
|
|
int unset __maybe_unused)
|
2009-12-15 16:53:38 +08:00
|
|
|
{
|
2016-04-08 22:25:59 +08:00
|
|
|
struct dirent *script_dirent, *lang_dirent;
|
2009-12-15 16:53:38 +08:00
|
|
|
char scripts_path[MAXPATHLEN];
|
|
|
|
DIR *scripts_dir, *lang_dir;
|
|
|
|
char script_path[MAXPATHLEN];
|
|
|
|
char lang_path[MAXPATHLEN];
|
|
|
|
struct script_desc *desc;
|
|
|
|
char first_half[BUFSIZ];
|
|
|
|
char *script_root;
|
|
|
|
|
2015-12-15 23:39:37 +08:00
|
|
|
snprintf(scripts_path, MAXPATHLEN, "%s/scripts", get_argv_exec_path());
|
2009-12-15 16:53:38 +08:00
|
|
|
|
|
|
|
scripts_dir = opendir(scripts_path);
|
2016-08-04 19:25:42 +08:00
|
|
|
if (!scripts_dir) {
|
|
|
|
fprintf(stdout,
|
|
|
|
"open(%s) failed.\n"
|
|
|
|
"Check \"PERF_EXEC_PATH\" env to set scripts dir.\n",
|
|
|
|
scripts_path);
|
|
|
|
exit(-1);
|
|
|
|
}
|
2009-12-15 16:53:38 +08:00
|
|
|
|
2016-04-08 22:25:59 +08:00
|
|
|
for_each_lang(scripts_path, scripts_dir, lang_dirent) {
|
2009-12-15 16:53:38 +08:00
|
|
|
snprintf(lang_path, MAXPATHLEN, "%s/%s/bin", scripts_path,
|
2016-04-08 22:25:59 +08:00
|
|
|
lang_dirent->d_name);
|
2009-12-15 16:53:38 +08:00
|
|
|
lang_dir = opendir(lang_path);
|
|
|
|
if (!lang_dir)
|
|
|
|
continue;
|
|
|
|
|
2016-04-08 22:25:59 +08:00
|
|
|
for_each_script(lang_path, lang_dir, script_dirent) {
|
|
|
|
script_root = get_script_root(script_dirent, REPORT_SUFFIX);
|
2011-11-25 18:38:40 +08:00
|
|
|
if (script_root) {
|
2009-12-15 16:53:38 +08:00
|
|
|
desc = script_desc__findnew(script_root);
|
|
|
|
snprintf(script_path, MAXPATHLEN, "%s/%s",
|
2016-04-08 22:25:59 +08:00
|
|
|
lang_path, script_dirent->d_name);
|
2009-12-15 16:53:38 +08:00
|
|
|
read_script_info(desc, script_path);
|
2011-11-25 18:38:40 +08:00
|
|
|
free(script_root);
|
2009-12-15 16:53:38 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
fprintf(stdout, "List of available trace scripts:\n");
|
|
|
|
list_for_each_entry(desc, &script_descs, node) {
|
|
|
|
sprintf(first_half, "%s %s", desc->name,
|
|
|
|
desc->args ? desc->args : "");
|
|
|
|
fprintf(stdout, " %-36s %s\n", first_half,
|
|
|
|
desc->half_liner ? desc->half_liner : "");
|
|
|
|
}
|
|
|
|
|
|
|
|
exit(0);
|
|
|
|
}
|
|
|
|
|
2012-10-30 11:56:03 +08:00
|
|
|
/*
|
|
|
|
* Some scripts specify the required events in their "xxx-record" file,
|
|
|
|
* this function will check if the events in perf.data match those
|
|
|
|
* mentioned in the "xxx-record".
|
|
|
|
*
|
|
|
|
* Fixme: All existing "xxx-record" are all in good formats "-e event ",
|
|
|
|
* which is covered well now. And new parsing code should be added to
|
|
|
|
* cover the future complexing formats like event groups etc.
|
|
|
|
*/
|
|
|
|
static int check_ev_match(char *dir_name, char *scriptname,
|
|
|
|
struct perf_session *session)
|
|
|
|
{
|
|
|
|
char filename[MAXPATHLEN], evname[128];
|
|
|
|
char line[BUFSIZ], *p;
|
|
|
|
struct perf_evsel *pos;
|
|
|
|
int match, len;
|
|
|
|
FILE *fp;
|
|
|
|
|
|
|
|
sprintf(filename, "%s/bin/%s-record", dir_name, scriptname);
|
|
|
|
|
|
|
|
fp = fopen(filename, "r");
|
|
|
|
if (!fp)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
while (fgets(line, sizeof(line), fp)) {
|
|
|
|
p = ltrim(line);
|
|
|
|
if (*p == '#')
|
|
|
|
continue;
|
|
|
|
|
|
|
|
while (strlen(p)) {
|
|
|
|
p = strstr(p, "-e");
|
|
|
|
if (!p)
|
|
|
|
break;
|
|
|
|
|
|
|
|
p += 2;
|
|
|
|
p = ltrim(p);
|
|
|
|
len = strcspn(p, " \t");
|
|
|
|
if (!len)
|
|
|
|
break;
|
|
|
|
|
|
|
|
snprintf(evname, len + 1, "%s", p);
|
|
|
|
|
|
|
|
match = 0;
|
2016-06-23 22:26:15 +08:00
|
|
|
evlist__for_each_entry(session->evlist, pos) {
|
2012-10-30 11:56:03 +08:00
|
|
|
if (!strcmp(perf_evsel__name(pos), evname)) {
|
|
|
|
match = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!match) {
|
|
|
|
fclose(fp);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
fclose(fp);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-09-07 16:42:26 +08:00
|
|
|
/*
|
|
|
|
* Return -1 if none is found, otherwise the actual scripts number.
|
|
|
|
*
|
|
|
|
* Currently the only user of this function is the script browser, which
|
|
|
|
* will list all statically runnable scripts, select one, execute it and
|
|
|
|
* show the output in a perf browser.
|
|
|
|
*/
|
|
|
|
int find_scripts(char **scripts_array, char **scripts_path_array)
|
|
|
|
{
|
2016-04-08 22:25:59 +08:00
|
|
|
struct dirent *script_dirent, *lang_dirent;
|
2012-10-30 11:56:03 +08:00
|
|
|
char scripts_path[MAXPATHLEN], lang_path[MAXPATHLEN];
|
2012-09-07 16:42:26 +08:00
|
|
|
DIR *scripts_dir, *lang_dir;
|
2012-10-30 11:56:03 +08:00
|
|
|
struct perf_session *session;
|
2013-10-15 22:27:32 +08:00
|
|
|
struct perf_data_file file = {
|
|
|
|
.path = input_name,
|
|
|
|
.mode = PERF_DATA_MODE_READ,
|
|
|
|
};
|
2012-09-07 16:42:26 +08:00
|
|
|
char *temp;
|
|
|
|
int i = 0;
|
|
|
|
|
2013-10-15 22:27:32 +08:00
|
|
|
session = perf_session__new(&file, false, NULL);
|
2012-10-30 11:56:03 +08:00
|
|
|
if (!session)
|
|
|
|
return -1;
|
|
|
|
|
2015-12-15 23:39:37 +08:00
|
|
|
snprintf(scripts_path, MAXPATHLEN, "%s/scripts", get_argv_exec_path());
|
2012-09-07 16:42:26 +08:00
|
|
|
|
|
|
|
scripts_dir = opendir(scripts_path);
|
2012-10-30 11:56:03 +08:00
|
|
|
if (!scripts_dir) {
|
|
|
|
perf_session__delete(session);
|
2012-09-07 16:42:26 +08:00
|
|
|
return -1;
|
2012-10-30 11:56:03 +08:00
|
|
|
}
|
2012-09-07 16:42:26 +08:00
|
|
|
|
2016-04-08 22:25:59 +08:00
|
|
|
for_each_lang(scripts_path, scripts_dir, lang_dirent) {
|
2012-09-07 16:42:26 +08:00
|
|
|
snprintf(lang_path, MAXPATHLEN, "%s/%s", scripts_path,
|
2016-04-08 22:25:59 +08:00
|
|
|
lang_dirent->d_name);
|
2012-09-07 16:42:26 +08:00
|
|
|
#ifdef NO_LIBPERL
|
|
|
|
if (strstr(lang_path, "perl"))
|
|
|
|
continue;
|
|
|
|
#endif
|
|
|
|
#ifdef NO_LIBPYTHON
|
|
|
|
if (strstr(lang_path, "python"))
|
|
|
|
continue;
|
|
|
|
#endif
|
|
|
|
|
|
|
|
lang_dir = opendir(lang_path);
|
|
|
|
if (!lang_dir)
|
|
|
|
continue;
|
|
|
|
|
2016-04-08 22:25:59 +08:00
|
|
|
for_each_script(lang_path, lang_dir, script_dirent) {
|
2012-09-07 16:42:26 +08:00
|
|
|
/* Skip those real time scripts: xxxtop.p[yl] */
|
2016-04-08 22:25:59 +08:00
|
|
|
if (strstr(script_dirent->d_name, "top."))
|
2012-09-07 16:42:26 +08:00
|
|
|
continue;
|
|
|
|
sprintf(scripts_path_array[i], "%s/%s", lang_path,
|
2016-04-08 22:25:59 +08:00
|
|
|
script_dirent->d_name);
|
|
|
|
temp = strchr(script_dirent->d_name, '.');
|
2012-09-07 16:42:26 +08:00
|
|
|
snprintf(scripts_array[i],
|
2016-04-08 22:25:59 +08:00
|
|
|
(temp - script_dirent->d_name) + 1,
|
|
|
|
"%s", script_dirent->d_name);
|
2012-10-30 11:56:03 +08:00
|
|
|
|
|
|
|
if (check_ev_match(lang_path,
|
|
|
|
scripts_array[i], session))
|
|
|
|
continue;
|
|
|
|
|
2012-09-07 16:42:26 +08:00
|
|
|
i++;
|
|
|
|
}
|
2012-10-30 11:56:03 +08:00
|
|
|
closedir(lang_dir);
|
2012-09-07 16:42:26 +08:00
|
|
|
}
|
|
|
|
|
2012-10-30 11:56:03 +08:00
|
|
|
closedir(scripts_dir);
|
|
|
|
perf_session__delete(session);
|
2012-09-07 16:42:26 +08:00
|
|
|
return i;
|
|
|
|
}
|
|
|
|
|
2009-12-15 16:53:39 +08:00
|
|
|
static char *get_script_path(const char *script_root, const char *suffix)
|
|
|
|
{
|
2016-04-08 22:25:59 +08:00
|
|
|
struct dirent *script_dirent, *lang_dirent;
|
2009-12-15 16:53:39 +08:00
|
|
|
char scripts_path[MAXPATHLEN];
|
|
|
|
char script_path[MAXPATHLEN];
|
|
|
|
DIR *scripts_dir, *lang_dir;
|
|
|
|
char lang_path[MAXPATHLEN];
|
2011-11-25 18:38:40 +08:00
|
|
|
char *__script_root;
|
2009-12-15 16:53:39 +08:00
|
|
|
|
2015-12-15 23:39:37 +08:00
|
|
|
snprintf(scripts_path, MAXPATHLEN, "%s/scripts", get_argv_exec_path());
|
2009-12-15 16:53:39 +08:00
|
|
|
|
|
|
|
scripts_dir = opendir(scripts_path);
|
|
|
|
if (!scripts_dir)
|
|
|
|
return NULL;
|
|
|
|
|
2016-04-08 22:25:59 +08:00
|
|
|
for_each_lang(scripts_path, scripts_dir, lang_dirent) {
|
2009-12-15 16:53:39 +08:00
|
|
|
snprintf(lang_path, MAXPATHLEN, "%s/%s/bin", scripts_path,
|
2016-04-08 22:25:59 +08:00
|
|
|
lang_dirent->d_name);
|
2009-12-15 16:53:39 +08:00
|
|
|
lang_dir = opendir(lang_path);
|
|
|
|
if (!lang_dir)
|
|
|
|
continue;
|
|
|
|
|
2016-04-08 22:25:59 +08:00
|
|
|
for_each_script(lang_path, lang_dir, script_dirent) {
|
|
|
|
__script_root = get_script_root(script_dirent, suffix);
|
2011-11-25 18:38:40 +08:00
|
|
|
if (__script_root && !strcmp(script_root, __script_root)) {
|
|
|
|
free(__script_root);
|
2012-01-08 01:25:25 +08:00
|
|
|
closedir(lang_dir);
|
|
|
|
closedir(scripts_dir);
|
2009-12-15 16:53:39 +08:00
|
|
|
snprintf(script_path, MAXPATHLEN, "%s/%s",
|
2016-04-08 22:25:59 +08:00
|
|
|
lang_path, script_dirent->d_name);
|
2011-11-25 18:38:40 +08:00
|
|
|
return strdup(script_path);
|
2009-12-15 16:53:39 +08:00
|
|
|
}
|
|
|
|
free(__script_root);
|
|
|
|
}
|
2012-01-08 01:25:25 +08:00
|
|
|
closedir(lang_dir);
|
2009-12-15 16:53:39 +08:00
|
|
|
}
|
2012-01-08 01:25:25 +08:00
|
|
|
closedir(scripts_dir);
|
2009-12-15 16:53:39 +08:00
|
|
|
|
2011-11-25 18:38:40 +08:00
|
|
|
return NULL;
|
2009-12-15 16:53:39 +08:00
|
|
|
}
|
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
static bool is_top_script(const char *script_path)
|
|
|
|
{
|
2010-12-03 23:52:01 +08:00
|
|
|
return ends_with(script_path, "top") == NULL ? false : true;
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static int has_required_arg(char *script_path)
|
|
|
|
{
|
|
|
|
struct script_desc *desc;
|
|
|
|
int n_args = 0;
|
|
|
|
char *p;
|
|
|
|
|
|
|
|
desc = script_desc__new(NULL);
|
|
|
|
|
|
|
|
if (read_script_info(desc, script_path))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (!desc->args)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
for (p = desc->args; *p; p++)
|
|
|
|
if (*p == '<')
|
|
|
|
n_args++;
|
|
|
|
out:
|
|
|
|
script_desc__delete(desc);
|
|
|
|
|
|
|
|
return n_args;
|
|
|
|
}
|
|
|
|
|
2012-10-02 02:20:58 +08:00
|
|
|
static int have_cmd(int argc, const char **argv)
|
|
|
|
{
|
|
|
|
char **__argv = malloc(sizeof(const char *) * argc);
|
|
|
|
|
|
|
|
if (!__argv) {
|
|
|
|
pr_err("malloc failed\n");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
memcpy(__argv, argv, sizeof(const char *) * argc);
|
|
|
|
argc = parse_options(argc, (const char **)__argv, record_options,
|
|
|
|
NULL, PARSE_OPT_STOP_AT_NON_OPTION);
|
|
|
|
free(__argv);
|
2009-08-17 22:18:08 +08:00
|
|
|
|
2012-10-02 02:20:58 +08:00
|
|
|
system_wide = (argc == 0);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2015-08-13 15:17:24 +08:00
|
|
|
static void script__setup_sample_type(struct perf_script *script)
|
|
|
|
{
|
|
|
|
struct perf_session *session = script->session;
|
|
|
|
u64 sample_type = perf_evlist__combined_sample_type(session->evlist);
|
|
|
|
|
|
|
|
if (symbol_conf.use_callchain || symbol_conf.cumulate_callchain) {
|
|
|
|
if ((sample_type & PERF_SAMPLE_REGS_USER) &&
|
|
|
|
(sample_type & PERF_SAMPLE_STACK_USER))
|
|
|
|
callchain_param.record_mode = CALLCHAIN_DWARF;
|
|
|
|
else if (sample_type & PERF_SAMPLE_BRANCH_STACK)
|
|
|
|
callchain_param.record_mode = CALLCHAIN_LBR;
|
|
|
|
else
|
|
|
|
callchain_param.record_mode = CALLCHAIN_FP;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-01-06 05:09:09 +08:00
|
|
|
static int process_stat_round_event(struct perf_tool *tool __maybe_unused,
|
|
|
|
union perf_event *event,
|
|
|
|
struct perf_session *session)
|
|
|
|
{
|
|
|
|
struct stat_round_event *round = &event->stat_round;
|
|
|
|
struct perf_evsel *counter;
|
|
|
|
|
2016-06-23 22:26:15 +08:00
|
|
|
evlist__for_each_entry(session->evlist, counter) {
|
2016-01-06 05:09:09 +08:00
|
|
|
perf_stat_process_counter(&stat_config, counter);
|
|
|
|
process_stat(counter, round->time);
|
|
|
|
}
|
|
|
|
|
|
|
|
process_stat_interval(round->time);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2016-01-06 05:09:07 +08:00
|
|
|
static int process_stat_config_event(struct perf_tool *tool __maybe_unused,
|
|
|
|
union perf_event *event,
|
|
|
|
struct perf_session *session __maybe_unused)
|
|
|
|
{
|
|
|
|
perf_event__read_stat_config(&stat_config, &event->stat_config);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2016-01-06 05:09:06 +08:00
|
|
|
static int set_maps(struct perf_script *script)
|
|
|
|
{
|
|
|
|
struct perf_evlist *evlist = script->session->evlist;
|
|
|
|
|
|
|
|
if (!script->cpus || !script->threads)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (WARN_ONCE(script->allocated, "stats double allocation\n"))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
perf_evlist__set_maps(evlist, script->cpus, script->threads);
|
|
|
|
|
|
|
|
if (perf_evlist__alloc_stats(evlist, true))
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
script->allocated = true;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static
|
|
|
|
int process_thread_map_event(struct perf_tool *tool,
|
|
|
|
union perf_event *event,
|
|
|
|
struct perf_session *session __maybe_unused)
|
|
|
|
{
|
|
|
|
struct perf_script *script = container_of(tool, struct perf_script, tool);
|
|
|
|
|
|
|
|
if (script->threads) {
|
|
|
|
pr_warning("Extra thread map event, ignoring.\n");
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
script->threads = thread_map__new_event(&event->thread_map);
|
|
|
|
if (!script->threads)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
return set_maps(script);
|
|
|
|
}
|
|
|
|
|
|
|
|
static
|
|
|
|
int process_cpu_map_event(struct perf_tool *tool __maybe_unused,
|
|
|
|
union perf_event *event,
|
|
|
|
struct perf_session *session __maybe_unused)
|
|
|
|
{
|
|
|
|
struct perf_script *script = container_of(tool, struct perf_script, tool);
|
|
|
|
|
|
|
|
if (script->cpus) {
|
|
|
|
pr_warning("Extra cpu map event, ignoring.\n");
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
script->cpus = cpu_map__new_data(&event->cpu_map.data);
|
|
|
|
if (!script->cpus)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
return set_maps(script);
|
|
|
|
}
|
|
|
|
|
2012-10-02 02:20:58 +08:00
|
|
|
int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)
|
|
|
|
{
|
|
|
|
bool show_full_info = false;
|
2013-12-09 18:02:50 +08:00
|
|
|
bool header = false;
|
|
|
|
bool header_only = false;
|
2014-08-12 14:40:33 +08:00
|
|
|
bool script_started = false;
|
2012-10-02 02:20:58 +08:00
|
|
|
char *rec_script_path = NULL;
|
|
|
|
char *rep_script_path = NULL;
|
|
|
|
struct perf_session *session;
|
2015-04-09 23:53:56 +08:00
|
|
|
struct itrace_synth_opts itrace_synth_opts = { .set = false, };
|
2012-10-02 02:20:58 +08:00
|
|
|
char *script_path = NULL;
|
|
|
|
const char **__argv;
|
2014-08-12 14:40:33 +08:00
|
|
|
int i, j, err = 0;
|
2013-10-22 15:34:07 +08:00
|
|
|
struct perf_script script = {
|
|
|
|
.tool = {
|
|
|
|
.sample = process_sample_event,
|
|
|
|
.mmap = perf_event__process_mmap,
|
|
|
|
.mmap2 = perf_event__process_mmap2,
|
|
|
|
.comm = perf_event__process_comm,
|
|
|
|
.exit = perf_event__process_exit,
|
|
|
|
.fork = perf_event__process_fork,
|
2013-11-01 21:51:30 +08:00
|
|
|
.attr = process_attr,
|
perf script: Process event update events
Andreas reported following command produces no output:
# cat test.py
#!/usr/bin/env python
def stat__krava(cpu, thread, time, val, ena, run):
print "event %s cpu %d, thread %d, time %d, val %d, ena %d, run %d" % \
("krava", cpu, thread, time, val, ena, run)
# perf stat -a -I 1000 -e cycles,"cpu/config=0x6530160,name=krava/" record | perf script -s test.py
^C
#
The reason is that 'perf script' does not process event update events and
will never get the event name update thus the python callback is never
called.
The fix is just to add already existing callback we use in 'perf stat
report'.
Committer note:
After the patch:
# perf stat -a -I 1000 -e cycles,"cpu/config=0x6530160,name=krava/" record | perf script -s test.py
event krava cpu -1, thread -1, time 1000239179, val 1789051, ena 4000690920, run 4000690920
event krava cpu -1, thread -1, time 2000479061, val 2391338, ena 4000879596, run 4000879596
event krava cpu -1, thread -1, time 3000740802, val 1939121, ena 4000977209, run 4000977209
event krava cpu -1, thread -1, time 4001006730, val 2356115, ena 4001000489, run 4001000489
^C
#
Reported-by: Andreas Hollmann <hollmann@in.tum.de>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Milian Wolff <milian.wolff@kdab.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1460013073-18444-3-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-04-07 15:11:13 +08:00
|
|
|
.event_update = perf_event__process_event_update,
|
2013-10-22 15:34:07 +08:00
|
|
|
.tracing_data = perf_event__process_tracing_data,
|
|
|
|
.build_id = perf_event__process_build_id,
|
2015-04-09 23:53:56 +08:00
|
|
|
.id_index = perf_event__process_id_index,
|
|
|
|
.auxtrace_info = perf_event__process_auxtrace_info,
|
|
|
|
.auxtrace = perf_event__process_auxtrace,
|
|
|
|
.auxtrace_error = perf_event__process_auxtrace_error,
|
2016-01-06 05:09:09 +08:00
|
|
|
.stat = perf_event__process_stat_event,
|
|
|
|
.stat_round = process_stat_round_event,
|
2016-01-06 05:09:07 +08:00
|
|
|
.stat_config = process_stat_config_event,
|
2016-01-06 05:09:06 +08:00
|
|
|
.thread_map = process_thread_map_event,
|
|
|
|
.cpu_map = process_cpu_map_event,
|
2014-07-06 20:18:21 +08:00
|
|
|
.ordered_events = true,
|
2013-10-22 15:34:07 +08:00
|
|
|
.ordering_requires_timestamps = true,
|
|
|
|
},
|
|
|
|
};
|
perf script: Support using -f to override perf.data file ownership
Enable perf script to use perf.data when it is not owned by current user
or root. Change the short option name of --fields to -F to avoid confusion
with --force.
Example:
# perf record ls
# chown Yunlong.Song:Yunlong.Song perf.data
# ls -al perf.data
-rw------- 1 Yunlong.Song Yunlong.Song 28360 Apr 2 14:53 perf.data
# id
uid=0(root) gid=0(root) groups=0(root),64(pkcs11)
Before this patch:
# perf script
File perf.data not owned by current user or root (use -f to override)
# perf script -f
Error: switch `f' requires a value
usage: perf script [<options>]
or: perf script [<options>] record <script> [<record-options>] <command>
or: perf script [<options>] report <script> [script-args]
or: perf script [<options>] <script> [<record-options>] <command>
or: perf script [<options>] <top-script> [script-args]
-f, --fields <str> comma separated output fields prepend with
'type:'. Valid types: hw,sw,trace,raw. Fields:
comm,tid,pid,time,cpu,event,trace,ip,sym,dso,addr,symoff,period
As shown above, the -f option does not work at all. And -f is already
taken up by --fields, which makes --force confused, so change the short
option name of --fields to -F like what other perf commands do (e.g.
perf report -F) and use -f as the short option name of --force.
After this patch:
# perf script
File perf.data not owned by current user or root (use -f to override)
# perf script -f
:41298 41298 2590086.564226: 1 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
:41298 41298 2590086.564244: 1 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
:41298 41298 2590086.564249: 7 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
:41298 41298 2590086.564255: 176 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
ls 41298 2590086.567346: 4059 cycles: ffffffff8105a592
raise_softirq ([kernel.kallsyms])
ls 41298 2590086.567353: 3717 cycles: ffffffff8105a592
raise_softirq ([kernel.kallsyms])
ls 41298 2590086.567358: 63058 cycles: ffffffff8105a592
raise_softirq ([kernel.kallsyms])
ls 41298 2590086.567448: 1706255 cycles: 406ae0
[unknown] (/usr/bin/ls)
As shown above, the -f option really works now.
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1427982439-27388-8-git-send-email-yunlong.song@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-02 21:47:16 +08:00
|
|
|
struct perf_data_file file = {
|
|
|
|
.mode = PERF_DATA_MODE_READ,
|
|
|
|
};
|
2012-10-02 02:20:58 +08:00
|
|
|
const struct option options[] = {
|
2009-08-17 22:18:08 +08:00
|
|
|
OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
|
|
|
|
"dump raw trace in ASCII"),
|
2010-04-13 16:37:33 +08:00
|
|
|
OPT_INCR('v', "verbose", &verbose,
|
2012-10-02 02:20:58 +08:00
|
|
|
"be more verbose (show symbol address, etc)"),
|
2009-12-15 16:53:38 +08:00
|
|
|
OPT_BOOLEAN('L', "Latency", &latency_format,
|
2009-10-15 03:43:42 +08:00
|
|
|
"show latency attributes (irqs/preemption disabled, etc)"),
|
2009-12-15 16:53:38 +08:00
|
|
|
OPT_CALLBACK_NOOPT('l', "list", NULL, NULL, "list available scripts",
|
|
|
|
list_available_scripts),
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
OPT_CALLBACK('s', "script", NULL, "name",
|
|
|
|
"script file name (lang:script name, script name, or *)",
|
|
|
|
parse_scriptname),
|
|
|
|
OPT_STRING('g', "gen-script", &generate_script_lang, "lang",
|
2010-11-17 01:45:39 +08:00
|
|
|
"generate perf-script.xx script in specified language"),
|
2012-10-02 02:20:58 +08:00
|
|
|
OPT_STRING('i', "input", &input_name, "file", "input file name"),
|
2010-05-27 22:27:47 +08:00
|
|
|
OPT_BOOLEAN('d', "debug-mode", &debug_mode,
|
|
|
|
"do various checks like samples ordering and lost events"),
|
2013-12-09 18:02:50 +08:00
|
|
|
OPT_BOOLEAN(0, "header", &header, "Show data header."),
|
|
|
|
OPT_BOOLEAN(0, "header-only", &header_only, "Show only data header."),
|
2011-03-10 13:23:27 +08:00
|
|
|
OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
|
|
|
|
"file", "vmlinux pathname"),
|
|
|
|
OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name,
|
|
|
|
"file", "kallsyms pathname"),
|
|
|
|
OPT_BOOLEAN('G', "hide-call-graph", &no_callchain,
|
|
|
|
"When printing symbols do not display call chain"),
|
2016-05-19 19:47:37 +08:00
|
|
|
OPT_CALLBACK(0, "symfs", NULL, "directory",
|
|
|
|
"Look for files with symbols relative to this directory",
|
|
|
|
symbol__config_symfs),
|
perf script: Support using -f to override perf.data file ownership
Enable perf script to use perf.data when it is not owned by current user
or root. Change the short option name of --fields to -F to avoid confusion
with --force.
Example:
# perf record ls
# chown Yunlong.Song:Yunlong.Song perf.data
# ls -al perf.data
-rw------- 1 Yunlong.Song Yunlong.Song 28360 Apr 2 14:53 perf.data
# id
uid=0(root) gid=0(root) groups=0(root),64(pkcs11)
Before this patch:
# perf script
File perf.data not owned by current user or root (use -f to override)
# perf script -f
Error: switch `f' requires a value
usage: perf script [<options>]
or: perf script [<options>] record <script> [<record-options>] <command>
or: perf script [<options>] report <script> [script-args]
or: perf script [<options>] <script> [<record-options>] <command>
or: perf script [<options>] <top-script> [script-args]
-f, --fields <str> comma separated output fields prepend with
'type:'. Valid types: hw,sw,trace,raw. Fields:
comm,tid,pid,time,cpu,event,trace,ip,sym,dso,addr,symoff,period
As shown above, the -f option does not work at all. And -f is already
taken up by --fields, which makes --force confused, so change the short
option name of --fields to -F like what other perf commands do (e.g.
perf report -F) and use -f as the short option name of --force.
After this patch:
# perf script
File perf.data not owned by current user or root (use -f to override)
# perf script -f
:41298 41298 2590086.564226: 1 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
:41298 41298 2590086.564244: 1 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
:41298 41298 2590086.564249: 7 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
:41298 41298 2590086.564255: 176 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
ls 41298 2590086.567346: 4059 cycles: ffffffff8105a592
raise_softirq ([kernel.kallsyms])
ls 41298 2590086.567353: 3717 cycles: ffffffff8105a592
raise_softirq ([kernel.kallsyms])
ls 41298 2590086.567358: 63058 cycles: ffffffff8105a592
raise_softirq ([kernel.kallsyms])
ls 41298 2590086.567448: 1706255 cycles: 406ae0
[unknown] (/usr/bin/ls)
As shown above, the -f option really works now.
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1427982439-27388-8-git-send-email-yunlong.song@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-02 21:47:16 +08:00
|
|
|
OPT_CALLBACK('F', "fields", NULL, "str",
|
2012-01-30 12:43:15 +08:00
|
|
|
"comma separated output fields prepend with 'type:'. "
|
|
|
|
"Valid types: hw,sw,trace,raw. "
|
|
|
|
"Fields: comm,tid,pid,time,cpu,event,trace,ip,sym,dso,"
|
perf script: Add callindent option
Based on patches from Andi Kleen.
When printing PT instruction traces with perf script it is rather useful
to see some indentation for the call tree. This patch adds a new
callindent field to perf script that prints spaces for the function call
stack depth.
We already have code to track the function call stack for PT, that we
can reuse with minor modifications.
The resulting output is not quite as nice as ftrace yet, but a lot
better than what was there before.
Note there are some corner cases when the thread stack gets code
confused and prints incorrect indentation. Even with that it is fairly
useful.
When displaying kernel code traces it is recommended to run as root, as
otherwise perf doesn't understand the kernel addresses properly, and may
not reset the call stack correctly on kernel boundaries.
Example output:
sudo perf-with-kcore record eg2 -a -e intel_pt// -- sleep 1
sudo perf-with-kcore script eg2 --ns -F callindent,time,comm,pid,sym,ip,addr,flags,cpu --itrace=cre | less
...
swapper 0 [000] 5830.389116586: call irq_exit ffffffff8104d620 smp_call_function_single_interrupt+0x30 => ffffffff8107e720 irq_exit
swapper 0 [000] 5830.389116586: call idle_cpu ffffffff8107e769 irq_exit+0x49 => ffffffff810a3970 idle_cpu
swapper 0 [000] 5830.389116586: return idle_cpu ffffffff810a39b7 idle_cpu+0x47 => ffffffff8107e76e irq_exit
swapper 0 [000] 5830.389116586: call tick_nohz_irq_exit ffffffff8107e7bd irq_exit+0x9d => ffffffff810f2fc0 tick_nohz_irq_exit
swapper 0 [000] 5830.389116919: call __tick_nohz_idle_enter ffffffff810f2fe0 tick_nohz_irq_exit+0x20 => ffffffff810f28d0 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call ktime_get ffffffff810f28f1 __tick_nohz_idle_enter+0x21 => ffffffff810e9ec0 ktime_get
swapper 0 [000] 5830.389116919: call read_tsc ffffffff810e9ef6 ktime_get+0x36 => ffffffff81035070 read_tsc
swapper 0 [000] 5830.389116919: return read_tsc ffffffff81035084 read_tsc+0x14 => ffffffff810e9efc ktime_get
swapper 0 [000] 5830.389116919: return ktime_get ffffffff810e9f46 ktime_get+0x86 => ffffffff810f28f6 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call sched_clock_idle_sleep_event ffffffff810f290b __tick_nohz_idle_enter+0x3b => ffffffff810a7380 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: call sched_clock_cpu ffffffff810a738b sched_clock_idle_sleep_event+0xb => ffffffff810a72e0 sched_clock_cpu
swapper 0 [000] 5830.389116919: call sched_clock ffffffff810a734d sched_clock_cpu+0x6d => ffffffff81035750 sched_clock
swapper 0 [000] 5830.389116919: call native_sched_clock ffffffff81035754 sched_clock+0x4 => ffffffff81035640 native_sched_clock
swapper 0 [000] 5830.389116919: return native_sched_clock ffffffff8103568c native_sched_clock+0x4c => ffffffff81035759 sched_clock
swapper 0 [000] 5830.389116919: return sched_clock ffffffff8103575c sched_clock+0xc => ffffffff810a7352 sched_clock_cpu
swapper 0 [000] 5830.389116919: return sched_clock_cpu ffffffff810a7356 sched_clock_cpu+0x76 => ffffffff810a7390 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: return sched_clock_idle_sleep_event ffffffff810a7391 sched_clock_idle_sleep_event+0x11 => ffffffff810f2910 __tick_nohz_idle_enter
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-4-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:58 +08:00
|
|
|
"addr,symoff,period,iregs,brstack,brstacksym,flags,"
|
2016-08-03 10:47:49 +08:00
|
|
|
"bpf-output,callindent", parse_output_fields),
|
2011-11-25 22:05:25 +08:00
|
|
|
OPT_BOOLEAN('a', "all-cpus", &system_wide,
|
2012-10-02 02:20:58 +08:00
|
|
|
"system-wide collection from all CPUs"),
|
2012-09-07 16:42:24 +08:00
|
|
|
OPT_STRING('S', "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",
|
|
|
|
"only consider these symbols"),
|
2011-11-14 02:30:08 +08:00
|
|
|
OPT_STRING('C', "cpu", &cpu_list, "cpu", "list of cpus to profile"),
|
2011-11-22 01:02:52 +08:00
|
|
|
OPT_STRING('c', "comms", &symbol_conf.comm_list_str, "comm[,comm...]",
|
|
|
|
"only display events for these comms"),
|
2015-03-24 23:52:41 +08:00
|
|
|
OPT_STRING(0, "pid", &symbol_conf.pid_list_str, "pid[,pid...]",
|
|
|
|
"only consider symbols in these pids"),
|
|
|
|
OPT_STRING(0, "tid", &symbol_conf.tid_list_str, "tid[,tid...]",
|
|
|
|
"only consider symbols in these tids"),
|
2016-04-15 05:15:18 +08:00
|
|
|
OPT_UINTEGER(0, "max-stack", &scripting_max_stack,
|
|
|
|
"Set the maximum stack depth when parsing the callchain, "
|
|
|
|
"anything beyond the specified depth will be ignored. "
|
2016-04-27 21:16:24 +08:00
|
|
|
"Default: kernel.perf_event_max_stack or " __stringify(PERF_MAX_STACK_DEPTH)),
|
perf tools: Make perf.data more self-descriptive (v8)
The goal of this patch is to include more information about the host
environment into the perf.data so it is more self-descriptive. Overtime,
profiles are captured on various machines and it becomes hard to track
what was recorded, on what machine and when.
This patch provides a way to solve this by extending the perf.data file
with basic information about the host machine. To add those extensions,
we leverage the feature bits capabilities of the perf.data format. The
change is backward compatible with existing perf.data files.
We define the following useful new extensions:
- HEADER_HOSTNAME: the hostname
- HEADER_OSRELEASE: the kernel release number
- HEADER_ARCH: the hw architecture
- HEADER_CPUDESC: generic CPU description
- HEADER_NRCPUS: number of online/avail cpus
- HEADER_CMDLINE: perf command line
- HEADER_VERSION: perf version
- HEADER_TOPOLOGY: cpu topology
- HEADER_EVENT_DESC: full event description (attrs)
- HEADER_CPUID: easy-to-parse low level CPU identication
The small granularity for the entries is to make it easier to extend
without breaking backward compatiblity. Many entries are provided as
ASCII strings.
Perf report/script have been modified to print the basic information as
easy-to-parse ASCII strings. Extended information about CPU and NUMA
topology may be requested with the -I option.
Thanks to David Ahern for reviewing and testing the many versions of
this patch.
$ perf report --stdio
# ========
# captured on : Mon Sep 26 15:22:14 2011
# hostname : quad
# os release : 3.1.0-rc4-tip
# perf version : 3.1.0-rc4
# arch : x86_64
# nrcpus online : 4
# nrcpus avail : 4
# cpudesc : Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz
# cpuid : GenuineIntel,6,15,11
# total memory : 8105360 kB
# cmdline : /home/eranian/perfmon/official/tip/build/tools/perf/perf record date
# event : name = cycles, type = 0, config = 0x0, config1 = 0x0, config2 = 0x0, excl_usr = 0, excl_kern = 0, id = { 29, 30, 31,
# HEADER_CPU_TOPOLOGY info available, use -I to display
# HEADER_NUMA_TOPOLOGY info available, use -I to display
# ========
#
...
$ perf report --stdio -I
# ========
# captured on : Mon Sep 26 15:22:14 2011
# hostname : quad
# os release : 3.1.0-rc4-tip
# perf version : 3.1.0-rc4
# arch : x86_64
# nrcpus online : 4
# nrcpus avail : 4
# cpudesc : Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz
# cpuid : GenuineIntel,6,15,11
# total memory : 8105360 kB
# cmdline : /home/eranian/perfmon/official/tip/build/tools/perf/perf record date
# event : name = cycles, type = 0, config = 0x0, config1 = 0x0, config2 = 0x0, excl_usr = 0, excl_kern = 0, id = { 29, 30, 31,
# sibling cores : 0-3
# sibling threads : 0
# sibling threads : 1
# sibling threads : 2
# sibling threads : 3
# node0 meminfo : total = 8320608 kB, free = 7571024 kB
# node0 cpu list : 0-3
# ========
#
...
Reviewed-by: David Ahern <dsahern@gmail.com>
Tested-by: David Ahern <dsahern@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Andi Kleen <ak@linux.intel.com>
Link: http://lkml.kernel.org/r/20110930134040.GA5575@quad
Signed-off-by: Stephane Eranian <eranian@google.com>
[ committer notes: Use --show-info in the tools as was in the docs, rename
perf_header_fprintf_info to perf_file_section__fprintf_info, fixup
conflict with f69b64f7 "perf: Support setting the disassembler style" ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-09-30 21:40:40 +08:00
|
|
|
OPT_BOOLEAN('I', "show-info", &show_full_info,
|
|
|
|
"display extended information from perf.data file"),
|
2012-01-30 12:43:20 +08:00
|
|
|
OPT_BOOLEAN('\0', "show-kernel-path", &symbol_conf.show_kernel_path,
|
|
|
|
"Show the path of [kernel.kallsyms]"),
|
2013-11-26 16:51:12 +08:00
|
|
|
OPT_BOOLEAN('\0', "show-task-events", &script.show_task_events,
|
|
|
|
"Show the fork/comm/exit events"),
|
2013-11-26 16:54:26 +08:00
|
|
|
OPT_BOOLEAN('\0', "show-mmap-events", &script.show_mmap_events,
|
|
|
|
"Show the mmap events"),
|
2015-07-21 17:44:06 +08:00
|
|
|
OPT_BOOLEAN('\0', "show-switch-events", &script.show_switch_events,
|
|
|
|
"Show context switch events (if recorded)"),
|
perf script: Support using -f to override perf.data file ownership
Enable perf script to use perf.data when it is not owned by current user
or root. Change the short option name of --fields to -F to avoid confusion
with --force.
Example:
# perf record ls
# chown Yunlong.Song:Yunlong.Song perf.data
# ls -al perf.data
-rw------- 1 Yunlong.Song Yunlong.Song 28360 Apr 2 14:53 perf.data
# id
uid=0(root) gid=0(root) groups=0(root),64(pkcs11)
Before this patch:
# perf script
File perf.data not owned by current user or root (use -f to override)
# perf script -f
Error: switch `f' requires a value
usage: perf script [<options>]
or: perf script [<options>] record <script> [<record-options>] <command>
or: perf script [<options>] report <script> [script-args]
or: perf script [<options>] <script> [<record-options>] <command>
or: perf script [<options>] <top-script> [script-args]
-f, --fields <str> comma separated output fields prepend with
'type:'. Valid types: hw,sw,trace,raw. Fields:
comm,tid,pid,time,cpu,event,trace,ip,sym,dso,addr,symoff,period
As shown above, the -f option does not work at all. And -f is already
taken up by --fields, which makes --force confused, so change the short
option name of --fields to -F like what other perf commands do (e.g.
perf report -F) and use -f as the short option name of --force.
After this patch:
# perf script
File perf.data not owned by current user or root (use -f to override)
# perf script -f
:41298 41298 2590086.564226: 1 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
:41298 41298 2590086.564244: 1 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
:41298 41298 2590086.564249: 7 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
:41298 41298 2590086.564255: 176 cycles: ffffffff8103efc6
native_write_msr_safe ([kernel.kallsyms])
ls 41298 2590086.567346: 4059 cycles: ffffffff8105a592
raise_softirq ([kernel.kallsyms])
ls 41298 2590086.567353: 3717 cycles: ffffffff8105a592
raise_softirq ([kernel.kallsyms])
ls 41298 2590086.567358: 63058 cycles: ffffffff8105a592
raise_softirq ([kernel.kallsyms])
ls 41298 2590086.567448: 1706255 cycles: 406ae0
[unknown] (/usr/bin/ls)
As shown above, the -f option really works now.
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1427982439-27388-8-git-send-email-yunlong.song@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-02 21:47:16 +08:00
|
|
|
OPT_BOOLEAN('f', "force", &file.force, "don't complain, do it"),
|
2015-09-25 21:15:36 +08:00
|
|
|
OPT_BOOLEAN(0, "ns", &nanosecs,
|
|
|
|
"Use 9 decimal places when displaying time"),
|
2015-04-09 23:53:56 +08:00
|
|
|
OPT_CALLBACK_OPTARG(0, "itrace", &itrace_synth_opts, NULL, "opts",
|
|
|
|
"Instruction Tracing options",
|
|
|
|
itrace_parse_synth_opts),
|
2015-08-08 06:24:05 +08:00
|
|
|
OPT_BOOLEAN(0, "full-source-path", &srcline_full_filename,
|
|
|
|
"Show full source file name path for source lines"),
|
2015-08-27 03:18:15 +08:00
|
|
|
OPT_BOOLEAN(0, "demangle", &symbol_conf.demangle,
|
|
|
|
"Enable symbol demangling"),
|
|
|
|
OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
|
|
|
|
"Enable kernel symbol demangling"),
|
|
|
|
|
2009-08-22 02:56:03 +08:00
|
|
|
OPT_END()
|
2012-10-02 02:20:58 +08:00
|
|
|
};
|
2015-03-18 21:35:54 +08:00
|
|
|
const char * const script_subcommands[] = { "record", "report", NULL };
|
|
|
|
const char *script_usage[] = {
|
2012-10-02 02:20:58 +08:00
|
|
|
"perf script [<options>]",
|
|
|
|
"perf script [<options>] record <script> [<record-options>] <command>",
|
|
|
|
"perf script [<options>] report <script> [script-args]",
|
|
|
|
"perf script [<options>] <script> [<record-options>] <command>",
|
|
|
|
"perf script [<options>] <top-script> [script-args]",
|
|
|
|
NULL
|
|
|
|
};
|
2009-12-15 16:53:39 +08:00
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
setup_scripting();
|
|
|
|
|
2015-03-18 21:35:54 +08:00
|
|
|
argc = parse_options_subcommand(argc, argv, options, script_subcommands, script_usage,
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
PARSE_OPT_STOP_AT_NON_OPTION);
|
|
|
|
|
2013-10-15 22:27:32 +08:00
|
|
|
file.path = input_name;
|
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
if (argc > 1 && !strncmp(argv[0], "rec", strlen("rec"))) {
|
|
|
|
rec_script_path = get_script_path(argv[1], RECORD_SUFFIX);
|
|
|
|
if (!rec_script_path)
|
|
|
|
return cmd_record(argc, argv, NULL);
|
2009-12-15 16:53:39 +08:00
|
|
|
}
|
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
if (argc > 1 && !strncmp(argv[0], "rep", strlen("rep"))) {
|
|
|
|
rep_script_path = get_script_path(argv[1], REPORT_SUFFIX);
|
|
|
|
if (!rep_script_path) {
|
2009-12-15 16:53:39 +08:00
|
|
|
fprintf(stderr,
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
"Please specify a valid report script"
|
2010-11-17 01:45:39 +08:00
|
|
|
"(see 'perf script -l' for listing)\n");
|
2009-12-15 16:53:39 +08:00
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-09-25 21:15:51 +08:00
|
|
|
if (itrace_synth_opts.callchain &&
|
|
|
|
itrace_synth_opts.callchain_sz > scripting_max_stack)
|
|
|
|
scripting_max_stack = itrace_synth_opts.callchain_sz;
|
|
|
|
|
2010-10-10 23:10:03 +08:00
|
|
|
/* make sure PERF_EXEC_PATH is set for scripts */
|
2015-12-15 23:39:37 +08:00
|
|
|
set_argv_exec_path(get_argv_exec_path());
|
2010-10-10 23:10:03 +08:00
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
if (argc && !script_name && !rec_script_path && !rep_script_path) {
|
2010-04-02 12:59:25 +08:00
|
|
|
int live_pipe[2];
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
int rep_args;
|
2010-04-02 12:59:25 +08:00
|
|
|
pid_t pid;
|
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
rec_script_path = get_script_path(argv[0], RECORD_SUFFIX);
|
|
|
|
rep_script_path = get_script_path(argv[0], REPORT_SUFFIX);
|
|
|
|
|
|
|
|
if (!rec_script_path && !rep_script_path) {
|
2015-10-24 23:49:27 +08:00
|
|
|
usage_with_options_msg(script_usage, options,
|
|
|
|
"Couldn't find script `%s'\n\n See perf"
|
2010-11-17 01:45:39 +08:00
|
|
|
" script -l for available scripts.\n", argv[0]);
|
2010-04-02 12:59:25 +08:00
|
|
|
}
|
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
if (is_top_script(argv[0])) {
|
|
|
|
rep_args = argc - 1;
|
|
|
|
} else {
|
|
|
|
int rec_args;
|
|
|
|
|
|
|
|
rep_args = has_required_arg(rep_script_path);
|
|
|
|
rec_args = (argc - 1) - rep_args;
|
|
|
|
if (rec_args < 0) {
|
2015-10-24 23:49:27 +08:00
|
|
|
usage_with_options_msg(script_usage, options,
|
|
|
|
"`%s' script requires options."
|
2010-11-17 01:45:39 +08:00
|
|
|
"\n\n See perf script -l for available "
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
"scripts and options.\n", argv[0]);
|
|
|
|
}
|
2010-04-02 12:59:25 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
if (pipe(live_pipe) < 0) {
|
|
|
|
perror("failed to create pipe");
|
2012-08-27 02:24:46 +08:00
|
|
|
return -1;
|
2010-04-02 12:59:25 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
pid = fork();
|
|
|
|
if (pid < 0) {
|
|
|
|
perror("failed to fork");
|
2012-08-27 02:24:46 +08:00
|
|
|
return -1;
|
2010-04-02 12:59:25 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!pid) {
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
j = 0;
|
|
|
|
|
2010-04-02 12:59:25 +08:00
|
|
|
dup2(live_pipe[1], 1);
|
|
|
|
close(live_pipe[0]);
|
|
|
|
|
2011-11-25 22:05:25 +08:00
|
|
|
if (is_top_script(argv[0])) {
|
|
|
|
system_wide = true;
|
|
|
|
} else if (!system_wide) {
|
2012-08-27 02:24:46 +08:00
|
|
|
if (have_cmd(argc - rep_args, &argv[rep_args]) != 0) {
|
|
|
|
err = -1;
|
|
|
|
goto out;
|
|
|
|
}
|
2011-11-25 22:05:25 +08:00
|
|
|
}
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
|
|
|
|
__argv = malloc((argc + 6) * sizeof(const char *));
|
2012-08-27 02:24:46 +08:00
|
|
|
if (!__argv) {
|
|
|
|
pr_err("malloc failed\n");
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
2010-11-10 21:52:32 +08:00
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
__argv[j++] = "/bin/sh";
|
|
|
|
__argv[j++] = rec_script_path;
|
|
|
|
if (system_wide)
|
|
|
|
__argv[j++] = "-a";
|
|
|
|
__argv[j++] = "-q";
|
|
|
|
__argv[j++] = "-o";
|
|
|
|
__argv[j++] = "-";
|
|
|
|
for (i = rep_args + 1; i < argc; i++)
|
|
|
|
__argv[j++] = argv[i];
|
|
|
|
__argv[j++] = NULL;
|
2010-04-02 12:59:25 +08:00
|
|
|
|
|
|
|
execvp("/bin/sh", (char **)__argv);
|
2010-11-10 21:52:32 +08:00
|
|
|
free(__argv);
|
2010-04-02 12:59:25 +08:00
|
|
|
exit(-1);
|
|
|
|
}
|
|
|
|
|
|
|
|
dup2(live_pipe[0], 0);
|
|
|
|
close(live_pipe[1]);
|
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
__argv = malloc((argc + 4) * sizeof(const char *));
|
2012-08-27 02:24:46 +08:00
|
|
|
if (!__argv) {
|
|
|
|
pr_err("malloc failed\n");
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
j = 0;
|
|
|
|
__argv[j++] = "/bin/sh";
|
|
|
|
__argv[j++] = rep_script_path;
|
|
|
|
for (i = 1; i < rep_args + 1; i++)
|
|
|
|
__argv[j++] = argv[i];
|
|
|
|
__argv[j++] = "-i";
|
|
|
|
__argv[j++] = "-";
|
|
|
|
__argv[j++] = NULL;
|
2010-04-02 12:59:25 +08:00
|
|
|
|
|
|
|
execvp("/bin/sh", (char **)__argv);
|
2010-11-10 21:52:32 +08:00
|
|
|
free(__argv);
|
2010-04-02 12:59:25 +08:00
|
|
|
exit(-1);
|
|
|
|
}
|
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
if (rec_script_path)
|
|
|
|
script_path = rec_script_path;
|
|
|
|
if (rep_script_path)
|
|
|
|
script_path = rep_script_path;
|
perf trace record: handle commands correctly
Because the perf-trace shell scripts hard-coded the use of the
perf-record system-wide param, a perf trace record session was always
system wide, even if it was given a command.
If given a command, perf trace record now only records the events for
the command, as users expect.
If no command is given, or if the '-a' option is used, the recorded
events are system-wide, as before.
root@tropicana:~# perf trace record syscall-counts ls -al
root@tropicana:~# perf trace
ls-23152 [000] 39984.890387: sys_enter: NR 12 (0, 0, 0, 0, 0, 0)
ls-23152 [000] 39984.890404: sys_enter: NR 9 (0, 0, 0, 0, 0, 0)
root@tropicana:~# perf trace record syscall-counts -a ls -al
root@tropicana:~# perf trace
npviewer.bin-22297 [000] 39831.102709: sys_enter: NR 168 (0, 0, 0, 0, 0, 0)
ls-23111 [000] 39831.107679: sys_enter: NR 59 (0, 0, 0, 0, 0, 0)
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:15:43 +08:00
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
if (script_path) {
|
|
|
|
j = 0;
|
2009-12-15 16:53:39 +08:00
|
|
|
|
2011-11-25 22:05:25 +08:00
|
|
|
if (!rec_script_path)
|
|
|
|
system_wide = false;
|
2012-08-27 02:24:46 +08:00
|
|
|
else if (!system_wide) {
|
|
|
|
if (have_cmd(argc - 1, &argv[1]) != 0) {
|
|
|
|
err = -1;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
}
|
perf trace record: handle commands correctly
Because the perf-trace shell scripts hard-coded the use of the
perf-record system-wide param, a perf trace record session was always
system wide, even if it was given a command.
If given a command, perf trace record now only records the events for
the command, as users expect.
If no command is given, or if the '-a' option is used, the recorded
events are system-wide, as before.
root@tropicana:~# perf trace record syscall-counts ls -al
root@tropicana:~# perf trace
ls-23152 [000] 39984.890387: sys_enter: NR 12 (0, 0, 0, 0, 0, 0)
ls-23152 [000] 39984.890404: sys_enter: NR 9 (0, 0, 0, 0, 0, 0)
root@tropicana:~# perf trace record syscall-counts -a ls -al
root@tropicana:~# perf trace
npviewer.bin-22297 [000] 39831.102709: sys_enter: NR 168 (0, 0, 0, 0, 0, 0)
ls-23111 [000] 39831.107679: sys_enter: NR 59 (0, 0, 0, 0, 0, 0)
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:15:43 +08:00
|
|
|
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
__argv = malloc((argc + 2) * sizeof(const char *));
|
2012-08-27 02:24:46 +08:00
|
|
|
if (!__argv) {
|
|
|
|
pr_err("malloc failed\n");
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
perf trace record: handle commands correctly
Because the perf-trace shell scripts hard-coded the use of the
perf-record system-wide param, a perf trace record session was always
system wide, even if it was given a command.
If given a command, perf trace record now only records the events for
the command, as users expect.
If no command is given, or if the '-a' option is used, the recorded
events are system-wide, as before.
root@tropicana:~# perf trace record syscall-counts ls -al
root@tropicana:~# perf trace
ls-23152 [000] 39984.890387: sys_enter: NR 12 (0, 0, 0, 0, 0, 0)
ls-23152 [000] 39984.890404: sys_enter: NR 9 (0, 0, 0, 0, 0, 0)
root@tropicana:~# perf trace record syscall-counts -a ls -al
root@tropicana:~# perf trace
npviewer.bin-22297 [000] 39831.102709: sys_enter: NR 168 (0, 0, 0, 0, 0, 0)
ls-23111 [000] 39831.107679: sys_enter: NR 59 (0, 0, 0, 0, 0, 0)
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:15:43 +08:00
|
|
|
__argv[j++] = "/bin/sh";
|
|
|
|
__argv[j++] = script_path;
|
|
|
|
if (system_wide)
|
|
|
|
__argv[j++] = "-a";
|
perf trace: live-mode command-line cleanup
This patch attempts to make the perf trace command-line for live-mode
commands more user-friendly and consistent with other perf commands.
The main change it makes is to allow <commands> to be run as part of
perf trace live-mode commands, as other perf commands do, instead of
the system-wide traces they're currently hard-coded to by the shell
scripts.
With this patch, the following live-mode trace now works as expected:
$ perf trace rw-by-pid ls -al
The previous system-wide behavior for this command would still be
available by explicitly specifying -a:
$ perf trace rw-by-pid -a ls -al
and if no <command> is specified, the output is also system-wide:
$ perf trace rw-by-pid
Because live-mode requires both record and report steps to be invoked,
it isn't always possible to know which args to send to the report and
which to send to the record steps - mainly this is the case for report
scripts with optional args - in those cases it would be necessary to
use separate 'perf trace record' and 'perf trace report' steps.
For example:
$ perf trace syscall-counts ls
Here we can't decide whether ls should be passed as a param to the
syscall-counts script or whether we should invoke ls as a <command>.
In these cases, we just say that we'll ignore optional script params
and always interpret the extra arguments as a <command>.
If the user instead wants the other interpretation, that can be
accomplished by using separate record and report commands explicitly:
$ perf trace record syscall-counts
$ perf trace report syscall-counts ls
So the rules that this patch implements, which seem to make the most
intuitive sense for live-mode commands:
- for commands with optional args and commands with no args, no args
are sent to the report script, all are sent to the record step
- for 'top' commands i.e. that end with 'top', <commands> can't be
used - all extra args are send to the report script as params
- for commands with required args, the n required args are taken to be
the first n args after the script name and sent to the report
script, and the rest are sent to the record step
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:16:51 +08:00
|
|
|
for (i = 2; i < argc; i++)
|
perf trace record: handle commands correctly
Because the perf-trace shell scripts hard-coded the use of the
perf-record system-wide param, a perf trace record session was always
system wide, even if it was given a command.
If given a command, perf trace record now only records the events for
the command, as users expect.
If no command is given, or if the '-a' option is used, the recorded
events are system-wide, as before.
root@tropicana:~# perf trace record syscall-counts ls -al
root@tropicana:~# perf trace
ls-23152 [000] 39984.890387: sys_enter: NR 12 (0, 0, 0, 0, 0, 0)
ls-23152 [000] 39984.890404: sys_enter: NR 9 (0, 0, 0, 0, 0, 0)
root@tropicana:~# perf trace record syscall-counts -a ls -al
root@tropicana:~# perf trace
npviewer.bin-22297 [000] 39831.102709: sys_enter: NR 168 (0, 0, 0, 0, 0, 0)
ls-23111 [000] 39831.107679: sys_enter: NR 59 (0, 0, 0, 0, 0, 0)
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
2010-11-10 22:15:43 +08:00
|
|
|
__argv[j++] = argv[i];
|
|
|
|
__argv[j++] = NULL;
|
2009-12-15 16:53:39 +08:00
|
|
|
|
|
|
|
execvp("/bin/sh", (char **)__argv);
|
2010-11-10 21:52:32 +08:00
|
|
|
free(__argv);
|
2009-12-15 16:53:39 +08:00
|
|
|
exit(-1);
|
|
|
|
}
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
|
2010-03-03 15:04:33 +08:00
|
|
|
if (!script_name)
|
|
|
|
setup_pager();
|
2009-08-17 22:18:08 +08:00
|
|
|
|
2013-10-22 15:34:07 +08:00
|
|
|
session = perf_session__new(&file, false, &script.tool);
|
2009-12-14 05:50:24 +08:00
|
|
|
if (session == NULL)
|
2014-09-24 09:33:37 +08:00
|
|
|
return -1;
|
2009-12-14 05:50:24 +08:00
|
|
|
|
2013-12-09 18:02:50 +08:00
|
|
|
if (header || header_only) {
|
|
|
|
perf_session__fprintf_info(session, stdout, show_full_info);
|
|
|
|
if (header_only)
|
2014-08-12 14:40:33 +08:00
|
|
|
goto out_delete;
|
2013-12-09 18:02:50 +08:00
|
|
|
}
|
|
|
|
|
2014-08-12 14:40:45 +08:00
|
|
|
if (symbol__init(&session->header.env) < 0)
|
2014-08-12 14:40:42 +08:00
|
|
|
goto out_delete;
|
|
|
|
|
2013-10-22 15:34:07 +08:00
|
|
|
script.session = session;
|
2015-08-13 15:17:24 +08:00
|
|
|
script__setup_sample_type(&script);
|
2013-10-22 15:34:07 +08:00
|
|
|
|
perf script: Add callindent option
Based on patches from Andi Kleen.
When printing PT instruction traces with perf script it is rather useful
to see some indentation for the call tree. This patch adds a new
callindent field to perf script that prints spaces for the function call
stack depth.
We already have code to track the function call stack for PT, that we
can reuse with minor modifications.
The resulting output is not quite as nice as ftrace yet, but a lot
better than what was there before.
Note there are some corner cases when the thread stack gets code
confused and prints incorrect indentation. Even with that it is fairly
useful.
When displaying kernel code traces it is recommended to run as root, as
otherwise perf doesn't understand the kernel addresses properly, and may
not reset the call stack correctly on kernel boundaries.
Example output:
sudo perf-with-kcore record eg2 -a -e intel_pt// -- sleep 1
sudo perf-with-kcore script eg2 --ns -F callindent,time,comm,pid,sym,ip,addr,flags,cpu --itrace=cre | less
...
swapper 0 [000] 5830.389116586: call irq_exit ffffffff8104d620 smp_call_function_single_interrupt+0x30 => ffffffff8107e720 irq_exit
swapper 0 [000] 5830.389116586: call idle_cpu ffffffff8107e769 irq_exit+0x49 => ffffffff810a3970 idle_cpu
swapper 0 [000] 5830.389116586: return idle_cpu ffffffff810a39b7 idle_cpu+0x47 => ffffffff8107e76e irq_exit
swapper 0 [000] 5830.389116586: call tick_nohz_irq_exit ffffffff8107e7bd irq_exit+0x9d => ffffffff810f2fc0 tick_nohz_irq_exit
swapper 0 [000] 5830.389116919: call __tick_nohz_idle_enter ffffffff810f2fe0 tick_nohz_irq_exit+0x20 => ffffffff810f28d0 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call ktime_get ffffffff810f28f1 __tick_nohz_idle_enter+0x21 => ffffffff810e9ec0 ktime_get
swapper 0 [000] 5830.389116919: call read_tsc ffffffff810e9ef6 ktime_get+0x36 => ffffffff81035070 read_tsc
swapper 0 [000] 5830.389116919: return read_tsc ffffffff81035084 read_tsc+0x14 => ffffffff810e9efc ktime_get
swapper 0 [000] 5830.389116919: return ktime_get ffffffff810e9f46 ktime_get+0x86 => ffffffff810f28f6 __tick_nohz_idle_enter
swapper 0 [000] 5830.389116919: call sched_clock_idle_sleep_event ffffffff810f290b __tick_nohz_idle_enter+0x3b => ffffffff810a7380 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: call sched_clock_cpu ffffffff810a738b sched_clock_idle_sleep_event+0xb => ffffffff810a72e0 sched_clock_cpu
swapper 0 [000] 5830.389116919: call sched_clock ffffffff810a734d sched_clock_cpu+0x6d => ffffffff81035750 sched_clock
swapper 0 [000] 5830.389116919: call native_sched_clock ffffffff81035754 sched_clock+0x4 => ffffffff81035640 native_sched_clock
swapper 0 [000] 5830.389116919: return native_sched_clock ffffffff8103568c native_sched_clock+0x4c => ffffffff81035759 sched_clock
swapper 0 [000] 5830.389116919: return sched_clock ffffffff8103575c sched_clock+0xc => ffffffff810a7352 sched_clock_cpu
swapper 0 [000] 5830.389116919: return sched_clock_cpu ffffffff810a7356 sched_clock_cpu+0x76 => ffffffff810a7390 sched_clock_idle_sleep_event
swapper 0 [000] 5830.389116919: return sched_clock_idle_sleep_event ffffffff810a7391 sched_clock_idle_sleep_event+0x11 => ffffffff810f2910 __tick_nohz_idle_enter
...
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1466689258-28493-4-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2016-06-23 21:40:58 +08:00
|
|
|
if (output[PERF_TYPE_HARDWARE].fields & PERF_OUTPUT_CALLINDENT)
|
|
|
|
itrace_synth_opts.thread_stack = true;
|
|
|
|
|
2015-04-09 23:53:56 +08:00
|
|
|
session->itrace_synth_opts = &itrace_synth_opts;
|
|
|
|
|
2011-07-04 19:57:50 +08:00
|
|
|
if (cpu_list) {
|
2014-08-12 14:40:33 +08:00
|
|
|
err = perf_session__cpu_bitmap(session, cpu_list, cpu_bitmap);
|
|
|
|
if (err < 0)
|
|
|
|
goto out_delete;
|
2011-07-04 19:57:50 +08:00
|
|
|
}
|
|
|
|
|
2011-03-10 13:23:28 +08:00
|
|
|
if (!no_callchain)
|
2011-03-10 13:23:27 +08:00
|
|
|
symbol_conf.use_callchain = true;
|
|
|
|
else
|
|
|
|
symbol_conf.use_callchain = false;
|
|
|
|
|
2015-08-04 03:27:40 +08:00
|
|
|
if (session->tevent.pevent &&
|
|
|
|
pevent_set_function_resolver(session->tevent.pevent,
|
2015-07-23 03:43:37 +08:00
|
|
|
machine__resolve_kernel_addr,
|
|
|
|
&session->machines.host) < 0) {
|
|
|
|
pr_err("%s: failed to set libtraceevent function resolver\n", __func__);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
if (generate_script_lang) {
|
|
|
|
struct stat perf_stat;
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
int input;
|
|
|
|
|
perf script: If type not given fields apply to all event types
Allow:
perf script -f <fields>
to be equivalent to:
perf script -f trace:<fields> -f sw:<fields> -f hw:<fields>
i.e., the specified fields apply to all event types if the type string
is not given.
The field (-f) arguments are processed in the order received. A later
usage can reset a prior request. e.g.,
-f trace: -f comm,tid,time,sym
The first -f suppresses trace events (field list is ""), but then the second
invocation sets the fields to comm,tid,time,sym. In this case a warning is
given to the user:
"Overriding previous field request for all events."
Alternativey, consider the order:
-f comm,tid,time,sym -f trace:
The first -f sets the fields for all events and the second -f suppresses trace
events. The user is given a warning message about the override, and the result
of the above is that only S/W and H/W events are displayed with the given
fields.
For the 'wildcard' option if a user selected field is invalid for an event
type, a message is displayed to the user that the option is ignored for that
type. For example:
perf script -f comm,tid,trace 2>&1 | less
'trace' not valid for hardware events. Ignoring.
'trace' not valid for software events. Ignoring.
Alternatively, if the type is given an invalid field is specified it is an
error. For example:
perf script -v -f sw:comm,tid,trace 2>&1 | less
'trace' not valid for software events.
At this point usage is displayed, and perf-script exits.
Finally, a user may not set fields to none for all event types.
i.e., -f "" is not allowed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
LPU-Reference: <1300377801-27246-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-18 00:03:21 +08:00
|
|
|
if (output_set_by_user()) {
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
fprintf(stderr,
|
|
|
|
"custom fields not supported for generated scripts");
|
2014-08-12 14:40:33 +08:00
|
|
|
err = -EINVAL;
|
|
|
|
goto out_delete;
|
perf script: Support custom field selection for output
Allow a user to select which fields to print to stdout for event data.
Options include comm (command name), tid (thread id), pid (process id),
time (perf timestamp), cpu, event (for event name), and trace (for
trace data).
Default is set to maintain compatibility with current output; this
feature does alter output format slightly -- no '-' between command
and pid/tid.
Thanks to Frederic Weisbecker for detailed suggestions on this approach.
Examples (output compressed)
1. trace, default format
perf record -ga -e sched:sched_switch
perf script
swapper 0 [000] 537.037184: sched_switch: prev_comm=swapper prev_pid=0...
sshd 1675 [000] 537.037309: sched_switch: prev_comm=sshd prev_pid=1675...
netstat 1692 [001] 537.038664: sched_switch: prev_comm=netstat prev_pid=1692...
2. trace, custom format
perf record -ga -e sched:sched_switch
perf script -f comm,pid,time,trace <--- omitting cpu and event name
swapper 0 537.037184: prev_comm=swapper prev_pid=0 prev_prio=120 ...
sshd 1675 537.037309: prev_comm=sshd prev_pid=1675 prev_prio=120 ...
netstat 1692 537.038664: prev_comm=netstat prev_pid=1692 prev_prio=120 ...
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1299734608-5223-5-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-03-10 13:23:26 +08:00
|
|
|
}
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
|
2013-10-15 22:27:34 +08:00
|
|
|
input = open(file.path, O_RDONLY); /* input_name */
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
if (input < 0) {
|
2014-08-12 14:40:33 +08:00
|
|
|
err = -errno;
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
perror("failed to open file");
|
2014-08-12 14:40:33 +08:00
|
|
|
goto out_delete;
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
err = fstat(input, &perf_stat);
|
|
|
|
if (err < 0) {
|
|
|
|
perror("failed to stat file");
|
2014-08-12 14:40:33 +08:00
|
|
|
goto out_delete;
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!perf_stat.st_size) {
|
|
|
|
fprintf(stderr, "zero-sized file, nothing to do!\n");
|
2014-08-12 14:40:33 +08:00
|
|
|
goto out_delete;
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
scripting_ops = script_spec__lookup(generate_script_lang);
|
|
|
|
if (!scripting_ops) {
|
|
|
|
fprintf(stderr, "invalid language specifier");
|
2014-08-12 14:40:33 +08:00
|
|
|
err = -ENOENT;
|
|
|
|
goto out_delete;
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
}
|
|
|
|
|
2013-12-03 21:09:23 +08:00
|
|
|
err = scripting_ops->generate_script(session->tevent.pevent,
|
2012-06-28 00:08:42 +08:00
|
|
|
"perf-script");
|
2014-08-12 14:40:33 +08:00
|
|
|
goto out_delete;
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
if (script_name) {
|
2009-12-15 16:53:35 +08:00
|
|
|
err = scripting_ops->start_script(script_name, argc, argv);
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
if (err)
|
2014-08-12 14:40:33 +08:00
|
|
|
goto out_delete;
|
2010-11-17 01:45:39 +08:00
|
|
|
pr_debug("perf script started with script %s\n\n", script_name);
|
2014-08-12 14:40:33 +08:00
|
|
|
script_started = true;
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
}
|
|
|
|
|
2011-04-07 11:54:20 +08:00
|
|
|
|
|
|
|
err = perf_session__check_output_opt(session);
|
|
|
|
if (err < 0)
|
2014-08-12 14:40:33 +08:00
|
|
|
goto out_delete;
|
2011-04-07 11:54:20 +08:00
|
|
|
|
2013-10-22 15:34:07 +08:00
|
|
|
err = __cmd_script(&script);
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
|
2014-08-16 03:08:37 +08:00
|
|
|
flush_scripting();
|
|
|
|
|
2014-08-12 14:40:33 +08:00
|
|
|
out_delete:
|
2016-01-06 05:09:06 +08:00
|
|
|
perf_evlist__free_stats(session->evlist);
|
2009-12-14 05:50:24 +08:00
|
|
|
perf_session__delete(session);
|
2014-08-12 14:40:33 +08:00
|
|
|
|
|
|
|
if (script_started)
|
|
|
|
cleanup_scripting();
|
perf trace: Add scripting ops
Adds an interface, scripting_ops, that when implemented for a
particular scripting language enables built-in support for trace
stream processing using that language.
The interface is designed to enable full-fledged language
interpreters to be embedded inside the perf executable and
thereby make the full capabilities of the supported languages
available for trace processing.
See below for details on the interface.
This patch also adds a couple command-line options to 'perf
trace':
The -s option option is used to specify the script to be run.
Script names that can be used with -s take the form:
[language spec:]scriptname[.ext]
Scripting languages register a set of 'language specs' that can
be used to specify scripts for the registered languages. The
specs can be used either as prefixes or extensions.
If [language spec:] is used, the script is taken as a script of
the matching language regardless of any extension it might have.
If [language spec:] is not used, [.ext] is used to look up the
language it corresponds to. Language specs are case
insensitive.
e.g. Perl scripts can be specified in the following ways:
Perl:scriptname
pl:scriptname.py # extension ignored
PL:scriptname
scriptname.pl
scriptname.perl
The -g [language spec] option gives users an easy starting point
for writing scripts in the specified language. Scripting
support for a particular language can implement a
generate_script() scripting op that outputs an empty (or
near-empty) set of handlers for all the events contained in a
given perf.data trace file - this option gives users a direct
way to access that.
Adding support for a scripting language
---------------------------------------
The main thing that needs to be done do add support for a new
language is to implement the scripting_ops interface:
It consists of the following four functions:
start_script()
stop_script()
process_event()
generate_script()
start_script() is called before any events are processed, and is
meant to give the scripting language support an opportunity to
set things up to receive events e.g. create and initialize an
instance of a language interpreter.
stop_script() is called after all events are processed, and is
meant to give the scripting language support an opportunity to
clean up e.g. destroy the interpreter instance, etc.
process_event() is called once for each event and takes as its
main parameter a pointer to the binary trace event record to be
processed. The implementation is responsible for picking out the
binary fields from the event record and sending them to the
script handler function associated with that event e.g. a
function derived from the event name it's meant to handle e.g.
'sched::sched_switch()'. The 'format' information for trace
events can be used to parse the binary data and map it into a
form usable by a given scripting language; see the Perl
implemention in subsequent patches for one possible way to
leverage the existing trace format parsing code in perf and map
that info into specific scripting language types.
generate_script() should generate a ready-to-run script for the
current set of events in the trace, preferably with bodies that
print out every field for each event. Again, look at the Perl
implementation for clues as to how that can be done. This is an
optional, but very useful op.
Support for a given language should also add a language-specific
setup function and call it from setup_scripting(). The
language-specific setup function associates the the scripting
ops for that language with one or more 'language specifiers'
(see below) using script_spec_register(). When a script name is
specified on the command line, the scripting ops associated with
the specified language are used to instantiate and use the
appropriate interpreter to process the trace stream.
In general, it should be relatively easy to add support for a
new language, especially if the language implementation supports
an interface allowing an interpreter to be 'embedded' inside
another program (in this case the containing program will be
'perf trace'). If so, it should be relatively straightforward to
translate trace events into invocations of user-defined script
functions where e.g. the function name corresponds to the event
type and the function parameters correspond to the event fields.
The event and field type information exported by the event
tracing infrastructure (via the event 'format' files) should be
enough to parse and send any piece of trace data to the user
script. The easiest way to see how this can be done would be to
look at the Perl implementation contained in
perf/util/trace-event-perl.c/.h.
There are a couple of other things that aren't covered by the
scripting_ops or setup interface and are technically optional,
but should be implemented if possible. One of these is support
for 'flag' and 'symbolic' fields e.g. being able to use more
human-readable values such as 'GFP_KERNEL' or
HI/BLOCK_IOPOLL/TASKLET in place of raw flag values. See the
Perl implementation to see how this can be done. The other thing
is support for 'calling back' into the perf executable to access
e.g. uncommon fields not passed by default into handler
functions, or any metadata the implementation might want to make
available to users via the language interface. Again, see the
Perl implementation for examples.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: fweisbec@gmail.com
Cc: rostedt@goodmis.org
Cc: anton@samba.org
Cc: hch@infradead.org
LKML-Reference: <1259133352-23685-2-git-send-email-tzanussi@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-25 15:15:46 +08:00
|
|
|
out:
|
|
|
|
return err;
|
2009-08-17 22:18:08 +08:00
|
|
|
}
|