The kvm kernel module includes a number of trace events which can be
useful when debugging system behavior. Even on production systems these
trace events can be used to observe guest behavior and identify the
source of problems.
The kvm_flightrecorder script is a command-line wrapper for the
/sys/kernel/debug/tracing interface. Kernel symbols do not need to be
installed.
This script captures a fixed-size buffer of KVM trace events. Recent
events overwrite the oldest events when the buffer size is exceeded and
it is possible to leave KVM tracing enabled for any period of time with
just a fixed-size buffer. If the buffer is large enough this script is
a useful tool for collecting detailed information after an issue occurs
with a guest. Hence the name "flight recorder".
The script can also be used in 'tail' mode to simply view KVM trace
events as they occur. This is handy for development and to ensure that
the guest is indeed running.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
It has happened more than once that patches that look perfectly sane
and work with simpletrace broke systemtap because they use 'next' as an
argument name for a tracing function. However, 'next' is a keyword for
systemtap, so we shouldn't use it.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
* stefanha/tracing:
tracetool: dtrace: handle in and next reserved words
tracetool: dtrace disabled-events fix
Makefile.target: code stp dependency on trace-events
Some locale settings let make fail or create wrong results
because tr '[:lower:]' '[:upper:]' which is used to convert
from lower to upper case depends on the locale.
With locale tr_TR.UTF-8, lower case 'i' is not converted to 'I'.
This results in wrong entries in config-host.h like these ones:
#define CONFIG_QEMU_PREFiX "/usr/local"
#define CONFIG_QEMU_BiNDiR "/usr/local/bin"
This problem was reported by Emre Ersin.
The same problem occurs when configure creates the target specific
files config-target.mak. They get wrong declarations:
TARGET_CRiS=y
TARGET_i386=y
TARGET_MiCROBLAZE=y
TARGET_MiPS64=y
TARGET_MiPS=y
TARGET_UNiCORE32=y
It is sufficient to restrict the conversion to the characters a-z.
Using this explicit range avoids the dependency on the locale
settings and is also shorter.
v2:
POSIX says that 'tr a-z' is unspecified outside of the POSIX
locale, so we must set LC_ALL=C to make sure that we are using
POSIX (hint from Eric Blake, thanks).
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
The idea behind qtest is pretty simple. Instead of executing a CPU via TCG or
KVM, rely on an external process to send events to the device model that the CPU
would normally generate.
qtest presents itself as an accelerator. In addition, a new option is added to
establish a qtest server (-qtest) that takes a character device. This is what
allows the external process to send CPU events to the device model.
qtest uses a simple line based protocol to send the events. Documentation of
that protocol is in qtest.c.
I considered reusing the monitor for this job. Adding interrupts would be a bit
difficult. In addition, logging would also be difficult.
qtest has extensive logging support. All protocol commands are logged with
time stamps using a new command line option (-qtest-log). Logging is important
since ultimately, this is a feature for debugging.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This introduces new test reporting infrastructure based on
gtester and gtester-report.
Also, all existing tests are moved to tests/, and tests/Makefile
is reorganized to factor out the commonalities in the rules.
Signed-off-by: Anthony Liguori <aliguori@linux.vnet.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
If there are "disabled" entries in the trace-events file then
linetod_nop() is called if the backend is dtrace, it's currently
not present. Also equivalent fix for stap.
Signed-off-by: Lee Essen <lee.essen@nowonline.co.uk>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Right now, the semantics of next_list are complicated. The caller must:
* call start_list
* call next_list for each element *including the first*
* on the first call to next_list, the second argument should point to
NULL and the result is the head of the list. On subsequent calls,
the second argument should point to the last node (last result of
next_list) and next_list itself tacks the element at the tail of the
list.
This works for both input and output visitor, but having the visitor
write memory when it is only reading the list is ugly. Plus, relying
on *list to detect the first call is tricky and undocumented.
We can initialize so->entry in next_list instead of start_list, leaving
it NULL in start_list. This way next_list sees clearly whether it is
on the first call---as a bonus, it discriminates the cases based on
internal state of the visitor rather than external state. We can
also pull the assignment of the list head from generated code up to
next_list.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Objects going through the dealloc visitor can be only partially allocated.
Detect the situation and avoid a segfault. This also helps with the
input visitor, when there are errors.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
We can exit very soon if we enter a visitor with a preexisting error.
This simplifies some cases because we will not have to deal with
obj being non-NULL while *obj is NULL.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
We've at least one UTF8 char in the qemu texi doc:
$ grep Tibor qemu-doc.texi
by Tibor "TS" Schütz.
$ man ./qemu.1 | grep Tibor
by Tibor "TS" SchA~Xtz.
This patch allows utf8 in man/pod docs.
Initially it was split into two parts and sent on 2012-02-02.
Resending it again (3rd time) now in merged form. If any
other generalizations of $(POD2MAN) are needed it can be done
in a separate patch. Current form of $(POD2MAN) is choosen
to be able to easily change it if some implementation does
not support utf8 or resulting output has issues with local
man(1) program/macros.
First, add @documentencoding in scripts/texi2pod.pl:
Currently our texi2pod ignores @documentencoding even if it is set
properly in *.texi files. This results in a mojibake in documents
generated from qemu.pod (which is generated from qemu-doc.texi by
texi2pod), because the rest of the tools assumes ASCII encoding.
This patch recognizes first @documentencoding in input and places
it at the beginning of output as =encoding directive.
Second, run pod2man with --utf8 option to enable utf8 in manpages:
This option makes no difference for manpages which contains only
ascii chars. But for manpages with actual UTF8 characters (qemu
docs contains these), this change allows to see real characters
instead of mojibakes or substitutes.
Signed-off-By: Michael Tokarev <mjt@tls.msk.ru>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Add a script that enhances gdb to be aware of QEMU data structures.
This patch adds a single gdb command, 'qemu mtree'. The command is
similar to the monitor's 'info mtree', except that it prints MemoryRegion
addresses, and except for working from a core dump as well as a live instance.
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
* kwolf/for-anthony:
test-coroutine: add performance test for nesting
coroutine: adding configure option for sigaltstack coroutine backend
coroutine: adding configure choose mechanism for coroutine backend
coroutine: adding sigaltstack method (.c source)
qcow2: Reduce number of I/O requests
qcow2: Add qcow2_alloc_clusters_at()
qcow2: Factor out count_cow_clusters
qmp: convert blockdev-snapshot-sync to a wrapper around transactions
add mode field to blockdev-snapshot-sync transaction item
rename blockdev-group-snapshot-sync
qapi: complete implementation of unions
use QSIMPLEQ_FOREACH_SAFE when freeing list elements
Add 'make check-block'
make check: Add qemu-iotests subset
qemu-iotests: Mark some tests as quick
qcow2: Add error messages in qcow2_truncate
block: handle -EBUSY in bdrv_commit_all()
qcow2: Add some tracing
qed: do not evict in-use L2 table cache entries
Group snapshot: Fix format name for backing file
SystemTap provides a "semaphore" that can optionally be tested before
executing a trace event. The purpose of this mechanism is to skip
expensive tracing code when the trace event is disabled.
For example, some applications may have trace events that format or
convert strings for trace events. This expensive processing should only
be done in the case where the trace event is enabled.
Since QEMU's generated trace events never have such special-purpose
code, there is no reason to add the semaphore check.
Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Adds a 'TRACE_${NAME}_ENABLED' preprocessor define for each tracing event in
"trace.h".
This lets the user conditionally compile code with a relatively high execution
cost that is only necessary when producing the tracing information for an event
that is enabled.
Note that events using this define will probably have the "disable" property by
default, in order to avoid such costs on regular builds.
Signed-off-by: Lluís Vilanova <vilanova@ac.upc.edu>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
We're supposed to keep qerror definitions and table entries in
alphabetical order. In practice this is not checked.
I haven't found a nice way to integrate this into the makefile yet but
we can at least have this script which verifies that qerrors are in
alphabetical order.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
The fixes to qapi code generation had multiple bugs:
- the Null class used to drop output was missing some methods
- in some scripts it was never instantiated, leading to a None return,
which is missing even more methods
- the --source and --header options were swapped
Luckily, all those bugs were hidden by a makefile bug which caused the
old behaviour (with the race) to be invoked.
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Make's multiple output syntax
x.c x.h: x.template
gen < x.template
actually invokes the command once for x.c and once for x.h (with differing $@
in each invocation). During a parallel build, the two commands may be invoked
in parallel; this opens up a race, where the second invocation trashes a file
supposedly produced during the first, and now in use by a dependent command.
The various qapi code generators are susceptible to this; fix by making them
generate just one file per invocation.
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Currently, we just print the numerical value of 9p operation identifier in
case of RERROR which is less meaningful for readability. Mapping 9p
operation ids to symbolic names provides a better tracelog:
RERROR (tag = 1 , id = TWALK , err = " No such file or directory ")
RERROR (tag = 1 , id = TUNLINKAT , err = " Directory not empty ")
This patch provides a dictionary of all possible 9p operation symbols mapped
to their numerical identifiers which are likely to be used in future at
various places in this script.
Signed-off-by: Harsh Prateek Bora <harsh@linux.vnet.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Only print options in the help output that are accepted by our arch.
This is less confusing for users and also for other programs that
consume the help output.
The options affected are:
-g and -prom-env only displayed on PPC or SPARC
-win2k-hack, -rtc-td-hack, -no-fd-bootchk, -no-acpi, -no-hpet,
-acpitable, -smbios only displayed on i386
-semihosting only displayed on ARM, M68K or XTENSA
-old-param only displayed on ARM
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
accidently->accidentally
annother->another
choosen->chosen
consideres->considers
decriptor->descriptor
developement->development
paramter->parameter
preceed->precede
preceeding->preceding
priviledge->privilege
propogation->propagation
substraction->subtraction
throught->through
upto->up to
usefull->useful
Fix also grammar in posix-aio-compat.c
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Fix a bug in handling dotted paths, and exclude directory prefixes
from generated guardnames to avoid odd/pseudo-random guardnames in
generated headers.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
We should only claim that something is a cast if we did not encouter a
token before, that did set av_pending.
This fixes the operator * in the line below to be detected as binary (vs
unary).
kmalloc(sizeof(struct alphatrack_ocmd) * true_size, GFP_KERNEL);
Reported-by: Peter Chubb <nicta.com.au>
Signed-off-by: Florian Mickler <florian@mickler.org>
(cherry-picked from Linux kernel commit c023e4734c3e8801e0ecb5e81b831d42a374d861)
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Cc: Blue Swirl <blauwirbel@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
There are 508 non-indented (non-default) labels, and 511 that are
indented. So the rule is debatable at least. Actually, in the
common case of labels at the outermost scope, there is really just
one place where to put the label, so the rule is just wrong IMHO.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
this patch fix multiple issues with VirtFS tracing.
a) Add tracepoint to the correct code path. We handle error in complete_pdu
b) Fix indentation in python script
c) Fix variable naming issue in python script
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Today we generate something like this:
int qmp_marshal_input_query_foo(...)
...
retval = qmp_query_foo(errp);
qmp_marshal_output_query_foo(retval, ret, errp);
...
However, if qmp_query_foo() fails 'retval' will probably be NULL,
which can cause a segfault as not all visitors check if 'retval'
is valid.
This commit fixes that by changing the code generator to only
call the output marshal if qmp_query_foo() succeeds, like this:
retval = qmp_query_foo(errp);
if (!error_is_set(errp)) {
qmp_marshal_output_query_foo(retval, ret, errp);
}
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Otherwise if we have something like 'foo-bar' in the schema,
it will be generated as 'foo_bar' in the string lookup table.
c_var() is good for C variables, but not for enum strings.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Modify logic such that we never assign values to the list head argument
to progress through the list on subsequent iterations, instead rely only
on having our return value passed back in as an argument on the next
call. Also update QMP I/O visitors and test cases accordingly, and add a
missing test case for QmpOutputVisitor.
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
To get the ball rolling merging QAPI, this patch introduces a "middle mode" to
the code generator. In middle mode, the code generator generates marshalling
functions that are compatible with the current QMP server. We absolutely need
to replace the current QMP server in order to support proper asynchronous
commands but using a middle mode provides a middle-ground that lets us start
converting commands in tree.
Note that all of the commands have been converted already in my glib branch.
Middle mode only exists until we finish merging them from my branch into the
main tree.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
The tracetool parser only picks up PRI*64 and other format string macros
when enclosed between double quoted strings. Lift this restriction by
extracting everything after the closing ')' as the format string:
cpu_set_apic_base(uint64_t val) "%016"PRIx64
^^ ^^
One trick here: it turns out that backslashes in the format string like
"\n" were being interpreted by echo(1). Fix this by using the POSIX
printf(1) command instead. Although it normally does not make sense to
include backslashes in trace event format strings, an injected newline
causes tracetool to emit a broken header file and I want to eliminate
cases where broken output is emitted, even if the input was bad.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
The simpletrace.process() function invokes analyzer methods with the
wrong number of arguments if a timestamp should be included. This patch
fixes the issue so that trace analysis scripts can make use of
timestamps.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Uses the generic interface provided in "trace/control.h" in order to provide
a programmatic interface as well as command line and monitor controls.
Signed-off-by: Fabien Chouteau <chouteau@adacore.com>
Signed-off-by: Lluís Vilanova <vilanova@ac.upc.edu>
Note that this refers to the backend-specific state (whether the output must be
generated), not the event "disabled" property (which always uses the "nop"
backend).
Signed-off-by: Lluís Vilanova <vilanova@ac.upc.edu>
Any event with the keyword/property "disable" generates an empty trace event
using the "nop" backend, regardless of the current backend.
Signed-off-by: Lluís Vilanova <vilanova@ac.upc.edu>
This adds/modifies the following functions:
* get_name: Get _only_ the event name
* has_property: Return whether an event has a property (keyword before the event
name)
Signed-off-by: Lluís Vilanova <vilanova@ac.upc.edu>
