2009-10-07 18:47:31 +08:00
|
|
|
#include "data_map.h"
|
|
|
|
#include "symbol.h"
|
|
|
|
#include "util.h"
|
|
|
|
#include "debug.h"
|
|
|
|
|
|
|
|
|
|
|
|
static struct perf_file_handler *curr_handler;
|
|
|
|
static unsigned long mmap_window = 32;
|
|
|
|
static char __cwd[PATH_MAX];
|
|
|
|
|
|
|
|
static int
|
|
|
|
process_event_stub(event_t *event __used,
|
|
|
|
unsigned long offset __used,
|
|
|
|
unsigned long head __used)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void register_perf_file_handler(struct perf_file_handler *handler)
|
|
|
|
{
|
|
|
|
if (!handler->process_sample_event)
|
|
|
|
handler->process_sample_event = process_event_stub;
|
|
|
|
if (!handler->process_mmap_event)
|
|
|
|
handler->process_mmap_event = process_event_stub;
|
|
|
|
if (!handler->process_comm_event)
|
|
|
|
handler->process_comm_event = process_event_stub;
|
|
|
|
if (!handler->process_fork_event)
|
|
|
|
handler->process_fork_event = process_event_stub;
|
|
|
|
if (!handler->process_exit_event)
|
|
|
|
handler->process_exit_event = process_event_stub;
|
|
|
|
if (!handler->process_lost_event)
|
|
|
|
handler->process_lost_event = process_event_stub;
|
|
|
|
if (!handler->process_read_event)
|
|
|
|
handler->process_read_event = process_event_stub;
|
|
|
|
if (!handler->process_throttle_event)
|
|
|
|
handler->process_throttle_event = process_event_stub;
|
|
|
|
if (!handler->process_unthrottle_event)
|
|
|
|
handler->process_unthrottle_event = process_event_stub;
|
|
|
|
|
|
|
|
curr_handler = handler;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
process_event(event_t *event, unsigned long offset, unsigned long head)
|
|
|
|
{
|
|
|
|
trace_event(event);
|
|
|
|
|
|
|
|
switch (event->header.type) {
|
|
|
|
case PERF_RECORD_SAMPLE:
|
|
|
|
return curr_handler->process_sample_event(event, offset, head);
|
|
|
|
case PERF_RECORD_MMAP:
|
|
|
|
return curr_handler->process_mmap_event(event, offset, head);
|
|
|
|
case PERF_RECORD_COMM:
|
|
|
|
return curr_handler->process_comm_event(event, offset, head);
|
|
|
|
case PERF_RECORD_FORK:
|
|
|
|
return curr_handler->process_fork_event(event, offset, head);
|
|
|
|
case PERF_RECORD_EXIT:
|
|
|
|
return curr_handler->process_exit_event(event, offset, head);
|
|
|
|
case PERF_RECORD_LOST:
|
|
|
|
return curr_handler->process_lost_event(event, offset, head);
|
|
|
|
case PERF_RECORD_READ:
|
|
|
|
return curr_handler->process_read_event(event, offset, head);
|
|
|
|
case PERF_RECORD_THROTTLE:
|
|
|
|
return curr_handler->process_throttle_event(event, offset, head);
|
|
|
|
case PERF_RECORD_UNTHROTTLE:
|
|
|
|
return curr_handler->process_unthrottle_event(event, offset, head);
|
|
|
|
default:
|
|
|
|
curr_handler->total_unknown++;
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-11-11 11:51:07 +08:00
|
|
|
int perf_header__read_build_ids(int input, off_t size)
|
perf symbols: Use the buildids if present
With this change 'perf record' will intercept PERF_RECORD_MMAP
calls, creating a linked list of DSOs, then when the session
finishes, it will traverse this list and read the buildids,
stashing them at the end of the file and will set up a new
feature bit in the header bitmask.
'perf report' will then notice this feature and populate the
'dsos' list and set the build ids.
When reading the symtabs it will refuse to load from a file that
doesn't have the same build id. This improves the
reliability of the profiler output, as symbols and profiling
data is more guaranteed to match.
Example:
[root@doppio ~]# perf report | head
/home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols
# Samples: 2621434559
#
# Overhead Command Shared Object Symbol
# ........ ............... ............................. ......
#
7.91% init [kernel] [k] read_hpet
7.64% init [kernel] [k] mwait_idle_with_hints
7.60% swapper [kernel] [k] read_hpet
7.60% swapper [kernel] [k] mwait_idle_with_hints
3.65% init [kernel] [k] 0xffffffffa02339d9
[root@doppio ~]#
In this case the 'perf' binary was an older one, vanished,
so its symbols probably wouldn't match or would cause subtly
different (and misleading) output.
Next patches will support the kernel as well, reading the build
id notes for it and the modules from /sys.
Another patch should also introduce a new plumbing command:
'perf list-buildids'
that will then be used in porcelain that is distro specific to
fetch -debuginfo packages where such buildids are present. This
will in turn allow for one to run 'perf record' in one machine
and 'perf report' in another.
Future work on having the buildid sent directly from the kernel
in the PERF_RECORD_MMAP event is needed to close races, as the
DSO can be changed during a 'perf record' session, but this
patch at least helps with non-corner cases and current/older
kernels.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Frank Ch. Eigler <fche@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jason Baron <jbaron@redhat.com>
Cc: Jim Keniston <jkenisto@us.ibm.com>
Cc: K. Prasad <prasad@linux.vnet.ibm.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Roland McGrath <roland@redhat.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
|
|
|
{
|
|
|
|
struct build_id_event bev;
|
|
|
|
char filename[PATH_MAX];
|
2009-11-11 11:51:07 +08:00
|
|
|
off_t offset = lseek(input, 0, SEEK_CUR);
|
|
|
|
off_t limit = offset + size;
|
perf symbols: Use the buildids if present
With this change 'perf record' will intercept PERF_RECORD_MMAP
calls, creating a linked list of DSOs, then when the session
finishes, it will traverse this list and read the buildids,
stashing them at the end of the file and will set up a new
feature bit in the header bitmask.
'perf report' will then notice this feature and populate the
'dsos' list and set the build ids.
When reading the symtabs it will refuse to load from a file that
doesn't have the same build id. This improves the
reliability of the profiler output, as symbols and profiling
data is more guaranteed to match.
Example:
[root@doppio ~]# perf report | head
/home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols
# Samples: 2621434559
#
# Overhead Command Shared Object Symbol
# ........ ............... ............................. ......
#
7.91% init [kernel] [k] read_hpet
7.64% init [kernel] [k] mwait_idle_with_hints
7.60% swapper [kernel] [k] read_hpet
7.60% swapper [kernel] [k] mwait_idle_with_hints
3.65% init [kernel] [k] 0xffffffffa02339d9
[root@doppio ~]#
In this case the 'perf' binary was an older one, vanished,
so its symbols probably wouldn't match or would cause subtly
different (and misleading) output.
Next patches will support the kernel as well, reading the build
id notes for it and the modules from /sys.
Another patch should also introduce a new plumbing command:
'perf list-buildids'
that will then be used in porcelain that is distro specific to
fetch -debuginfo packages where such buildids are present. This
will in turn allow for one to run 'perf record' in one machine
and 'perf report' in another.
Future work on having the buildid sent directly from the kernel
in the PERF_RECORD_MMAP event is needed to close races, as the
DSO can be changed during a 'perf record' session, but this
patch at least helps with non-corner cases and current/older
kernels.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Frank Ch. Eigler <fche@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jason Baron <jbaron@redhat.com>
Cc: Jim Keniston <jkenisto@us.ibm.com>
Cc: K. Prasad <prasad@linux.vnet.ibm.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Roland McGrath <roland@redhat.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
|
|
|
int err = -1;
|
|
|
|
|
2009-11-11 11:51:07 +08:00
|
|
|
while (offset < limit) {
|
perf symbols: Use the buildids if present
With this change 'perf record' will intercept PERF_RECORD_MMAP
calls, creating a linked list of DSOs, then when the session
finishes, it will traverse this list and read the buildids,
stashing them at the end of the file and will set up a new
feature bit in the header bitmask.
'perf report' will then notice this feature and populate the
'dsos' list and set the build ids.
When reading the symtabs it will refuse to load from a file that
doesn't have the same build id. This improves the
reliability of the profiler output, as symbols and profiling
data is more guaranteed to match.
Example:
[root@doppio ~]# perf report | head
/home/acme/bin/perf with build id b1ea544ac3746e7538972548a09aadecc5753868 not found, continuing without symbols
# Samples: 2621434559
#
# Overhead Command Shared Object Symbol
# ........ ............... ............................. ......
#
7.91% init [kernel] [k] read_hpet
7.64% init [kernel] [k] mwait_idle_with_hints
7.60% swapper [kernel] [k] read_hpet
7.60% swapper [kernel] [k] mwait_idle_with_hints
3.65% init [kernel] [k] 0xffffffffa02339d9
[root@doppio ~]#
In this case the 'perf' binary was an older one, vanished,
so its symbols probably wouldn't match or would cause subtly
different (and misleading) output.
Next patches will support the kernel as well, reading the build
id notes for it and the modules from /sys.
Another patch should also introduce a new plumbing command:
'perf list-buildids'
that will then be used in porcelain that is distro specific to
fetch -debuginfo packages where such buildids are present. This
will in turn allow for one to run 'perf record' in one machine
and 'perf report' in another.
Future work on having the buildid sent directly from the kernel
in the PERF_RECORD_MMAP event is needed to close races, as the
DSO can be changed during a 'perf record' session, but this
patch at least helps with non-corner cases and current/older
kernels.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Frank Ch. Eigler <fche@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jason Baron <jbaron@redhat.com>
Cc: Jim Keniston <jkenisto@us.ibm.com>
Cc: K. Prasad <prasad@linux.vnet.ibm.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Roland McGrath <roland@redhat.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1257367843-26224-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-05 04:50:43 +08:00
|
|
|
struct dso *dso;
|
|
|
|
ssize_t len;
|
|
|
|
|
|
|
|
if (read(input, &bev, sizeof(bev)) != sizeof(bev))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
len = bev.header.size - sizeof(bev);
|
|
|
|
if (read(input, filename, len) != len)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
dso = dsos__findnew(filename);
|
|
|
|
if (dso != NULL)
|
|
|
|
dso__set_build_id(dso, &bev.build_id);
|
|
|
|
|
|
|
|
offset += bev.header.size;
|
|
|
|
}
|
|
|
|
err = 0;
|
|
|
|
out:
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2009-10-07 18:47:31 +08:00
|
|
|
int mmap_dispatch_perf_file(struct perf_header **pheader,
|
|
|
|
const char *input_name,
|
|
|
|
int force,
|
|
|
|
int full_paths,
|
|
|
|
int *cwdlen,
|
|
|
|
char **cwd)
|
|
|
|
{
|
|
|
|
int ret, rc = EXIT_FAILURE;
|
|
|
|
struct perf_header *header;
|
|
|
|
unsigned long head, shift;
|
|
|
|
unsigned long offset = 0;
|
|
|
|
struct stat input_stat;
|
|
|
|
size_t page_size;
|
|
|
|
u64 sample_type;
|
|
|
|
event_t *event;
|
|
|
|
uint32_t size;
|
|
|
|
int input;
|
|
|
|
char *buf;
|
|
|
|
|
|
|
|
if (!curr_handler)
|
|
|
|
die("Forgot to register perf file handler");
|
|
|
|
|
|
|
|
page_size = getpagesize();
|
|
|
|
|
|
|
|
input = open(input_name, O_RDONLY);
|
|
|
|
if (input < 0) {
|
|
|
|
fprintf(stderr, " failed to open file: %s", input_name);
|
|
|
|
if (!strcmp(input_name, "perf.data"))
|
|
|
|
fprintf(stderr, " (try 'perf record' first)");
|
|
|
|
fprintf(stderr, "\n");
|
|
|
|
exit(-1);
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = fstat(input, &input_stat);
|
|
|
|
if (ret < 0) {
|
|
|
|
perror("failed to stat file");
|
|
|
|
exit(-1);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!force && input_stat.st_uid && (input_stat.st_uid != geteuid())) {
|
|
|
|
fprintf(stderr, "file: %s not owned by current user or root\n",
|
|
|
|
input_name);
|
|
|
|
exit(-1);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!input_stat.st_size) {
|
|
|
|
fprintf(stderr, "zero-sized file, nothing to do!\n");
|
|
|
|
exit(0);
|
|
|
|
}
|
|
|
|
|
|
|
|
*pheader = perf_header__read(input);
|
|
|
|
header = *pheader;
|
|
|
|
head = header->data_offset;
|
|
|
|
|
|
|
|
sample_type = perf_header__sample_type(header);
|
|
|
|
|
|
|
|
if (curr_handler->sample_type_check)
|
|
|
|
if (curr_handler->sample_type_check(sample_type) < 0)
|
|
|
|
exit(-1);
|
|
|
|
|
2009-10-31 02:28:24 +08:00
|
|
|
if (load_kernel(NULL) < 0) {
|
2009-10-07 18:47:31 +08:00
|
|
|
perror("failed to load kernel symbols");
|
|
|
|
return EXIT_FAILURE;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!full_paths) {
|
|
|
|
if (getcwd(__cwd, sizeof(__cwd)) == NULL) {
|
|
|
|
perror("failed to get the current directory");
|
|
|
|
return EXIT_FAILURE;
|
|
|
|
}
|
|
|
|
*cwd = __cwd;
|
|
|
|
*cwdlen = strlen(*cwd);
|
|
|
|
} else {
|
|
|
|
*cwd = NULL;
|
|
|
|
*cwdlen = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
shift = page_size * (head / page_size);
|
|
|
|
offset += shift;
|
|
|
|
head -= shift;
|
|
|
|
|
|
|
|
remap:
|
|
|
|
buf = (char *)mmap(NULL, page_size * mmap_window, PROT_READ,
|
|
|
|
MAP_SHARED, input, offset);
|
|
|
|
if (buf == MAP_FAILED) {
|
|
|
|
perror("failed to mmap file");
|
|
|
|
exit(-1);
|
|
|
|
}
|
|
|
|
|
|
|
|
more:
|
|
|
|
event = (event_t *)(buf + head);
|
|
|
|
|
|
|
|
size = event->header.size;
|
|
|
|
if (!size)
|
|
|
|
size = 8;
|
|
|
|
|
|
|
|
if (head + event->header.size >= page_size * mmap_window) {
|
|
|
|
int munmap_ret;
|
|
|
|
|
|
|
|
shift = page_size * (head / page_size);
|
|
|
|
|
|
|
|
munmap_ret = munmap(buf, page_size * mmap_window);
|
|
|
|
assert(munmap_ret == 0);
|
|
|
|
|
|
|
|
offset += shift;
|
|
|
|
head -= shift;
|
|
|
|
goto remap;
|
|
|
|
}
|
|
|
|
|
|
|
|
size = event->header.size;
|
|
|
|
|
|
|
|
dump_printf("\n%p [%p]: event: %d\n",
|
|
|
|
(void *)(offset + head),
|
|
|
|
(void *)(long)event->header.size,
|
|
|
|
event->header.type);
|
|
|
|
|
|
|
|
if (!size || process_event(event, offset, head) < 0) {
|
|
|
|
|
|
|
|
dump_printf("%p [%p]: skipping unknown header type: %d\n",
|
|
|
|
(void *)(offset + head),
|
|
|
|
(void *)(long)(event->header.size),
|
|
|
|
event->header.type);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* assume we lost track of the stream, check alignment, and
|
|
|
|
* increment a single u64 in the hope to catch on again 'soon'.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (unlikely(head & 7))
|
|
|
|
head &= ~7ULL;
|
|
|
|
|
|
|
|
size = 8;
|
|
|
|
}
|
|
|
|
|
|
|
|
head += size;
|
|
|
|
|
|
|
|
if (offset + head >= header->data_offset + header->data_size)
|
|
|
|
goto done;
|
|
|
|
|
|
|
|
if (offset + head < (unsigned long)input_stat.st_size)
|
|
|
|
goto more;
|
|
|
|
|
|
|
|
done:
|
|
|
|
rc = EXIT_SUCCESS;
|
|
|
|
close(input);
|
|
|
|
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
|