2009-06-25 09:05:54 -06:00
|
|
|
#include <sys/types.h>
|
|
|
|
#include <unistd.h>
|
|
|
|
#include <stdio.h>
|
|
|
|
#include <stdlib.h>
|
2009-11-10 20:51:03 -07:00
|
|
|
#include <linux/list.h>
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
#include "util.h"
|
|
|
|
#include "header.h"
|
2009-10-06 15:36:47 -06:00
|
|
|
#include "../perf.h"
|
|
|
|
#include "trace-event.h"
|
2009-12-13 14:50:25 -07:00
|
|
|
#include "session.h"
|
2009-11-10 20:51:03 -07:00
|
|
|
#include "symbol.h"
|
2009-11-10 20:51:05 -07:00
|
|
|
#include "debug.h"
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
/*
|
2009-09-19 05:36:30 -06:00
|
|
|
* Create new perf.data header attribute:
|
2009-06-25 09:05:54 -06:00
|
|
|
*/
|
perf: Do the big rename: Performance Counters -> Performance Events
Bye-bye Performance Counters, welcome Performance Events!
In the past few months the perfcounters subsystem has grown out its
initial role of counting hardware events, and has become (and is
becoming) a much broader generic event enumeration, reporting, logging,
monitoring, analysis facility.
Naming its core object 'perf_counter' and naming the subsystem
'perfcounters' has become more and more of a misnomer. With pending
code like hw-breakpoints support the 'counter' name is less and
less appropriate.
All in one, we've decided to rename the subsystem to 'performance
events' and to propagate this rename through all fields, variables
and API names. (in an ABI compatible fashion)
The word 'event' is also a bit shorter than 'counter' - which makes
it slightly more convenient to write/handle as well.
Thanks goes to Stephane Eranian who first observed this misnomer and
suggested a rename.
User-space tooling and ABI compatibility is not affected - this patch
should be function-invariant. (Also, defconfigs were not touched to
keep the size down.)
This patch has been generated via the following script:
FILES=$(find * -type f | grep -vE 'oprofile|[^K]config')
sed -i \
-e 's/PERF_EVENT_/PERF_RECORD_/g' \
-e 's/PERF_COUNTER/PERF_EVENT/g' \
-e 's/perf_counter/perf_event/g' \
-e 's/nb_counters/nb_events/g' \
-e 's/swcounter/swevent/g' \
-e 's/tpcounter_event/tp_event/g' \
$FILES
for N in $(find . -name perf_counter.[ch]); do
M=$(echo $N | sed 's/perf_counter/perf_event/g')
mv $N $M
done
FILES=$(find . -name perf_event.*)
sed -i \
-e 's/COUNTER_MASK/REG_MASK/g' \
-e 's/COUNTER/EVENT/g' \
-e 's/\<event\>/event_id/g' \
-e 's/counter/event/g' \
-e 's/Counter/Event/g' \
$FILES
... to keep it as correct as possible. This script can also be
used by anyone who has pending perfcounters patches - it converts
a Linux kernel tree over to the new naming. We tried to time this
change to the point in time where the amount of pending patches
is the smallest: the end of the merge window.
Namespace clashes were fixed up in a preparatory patch - and some
stylistic fallout will be fixed up in a subsequent patch.
( NOTE: 'counters' are still the proper terminology when we deal
with hardware registers - and these sed scripts are a bit
over-eager in renaming them. I've undone some of that, but
in case there's something left where 'counter' would be
better than 'event' we can undo that on an individual basis
instead of touching an otherwise nicely automated patch. )
Suggested-by: Stephane Eranian <eranian@google.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Paul Mackerras <paulus@samba.org>
Reviewed-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: <linux-arch@vger.kernel.org>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-21 04:02:48 -06:00
|
|
|
struct perf_header_attr *perf_header_attr__new(struct perf_event_attr *attr)
|
2009-06-25 09:05:54 -06:00
|
|
|
{
|
|
|
|
struct perf_header_attr *self = malloc(sizeof(*self));
|
|
|
|
|
2009-11-16 14:30:26 -07:00
|
|
|
if (self != NULL) {
|
|
|
|
self->attr = *attr;
|
|
|
|
self->ids = 0;
|
|
|
|
self->size = 1;
|
|
|
|
self->id = malloc(sizeof(u64));
|
|
|
|
if (self->id == NULL) {
|
|
|
|
free(self);
|
|
|
|
self = NULL;
|
|
|
|
}
|
|
|
|
}
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
return self;
|
|
|
|
}
|
|
|
|
|
2009-11-16 20:18:09 -07:00
|
|
|
void perf_header_attr__delete(struct perf_header_attr *self)
|
|
|
|
{
|
|
|
|
free(self->id);
|
|
|
|
free(self);
|
|
|
|
}
|
|
|
|
|
2009-11-16 20:18:10 -07:00
|
|
|
int perf_header_attr__add_id(struct perf_header_attr *self, u64 id)
|
2009-06-25 09:05:54 -06:00
|
|
|
{
|
|
|
|
int pos = self->ids;
|
|
|
|
|
|
|
|
self->ids++;
|
|
|
|
if (self->ids > self->size) {
|
2009-11-16 20:18:10 -07:00
|
|
|
int nsize = self->size * 2;
|
|
|
|
u64 *nid = realloc(self->id, nsize * sizeof(u64));
|
|
|
|
|
|
|
|
if (nid == NULL)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
self->size = nsize;
|
|
|
|
self->id = nid;
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
|
|
|
self->id[pos] = id;
|
2009-11-16 20:18:10 -07:00
|
|
|
return 0;
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
|
|
|
|
2009-12-11 16:24:02 -07:00
|
|
|
int perf_header__init(struct perf_header *self)
|
2009-06-25 09:05:54 -06:00
|
|
|
{
|
2009-12-11 16:24:02 -07:00
|
|
|
self->size = 1;
|
|
|
|
self->attr = malloc(sizeof(void *));
|
|
|
|
return self->attr == NULL ? -ENOMEM : 0;
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
|
|
|
|
2009-12-11 16:24:02 -07:00
|
|
|
void perf_header__exit(struct perf_header *self)
|
2009-11-19 09:55:55 -07:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
for (i = 0; i < self->attrs; ++i)
|
2009-12-11 16:24:02 -07:00
|
|
|
perf_header_attr__delete(self->attr[i]);
|
2009-11-19 09:55:55 -07:00
|
|
|
free(self->attr);
|
|
|
|
}
|
|
|
|
|
2009-11-16 20:18:09 -07:00
|
|
|
int perf_header__add_attr(struct perf_header *self,
|
|
|
|
struct perf_header_attr *attr)
|
2009-06-25 09:05:54 -06:00
|
|
|
{
|
|
|
|
if (self->frozen)
|
2009-11-16 20:18:09 -07:00
|
|
|
return -1;
|
2009-06-25 09:05:54 -06:00
|
|
|
|
2009-11-19 09:55:55 -07:00
|
|
|
if (self->attrs == self->size) {
|
2009-11-16 20:18:09 -07:00
|
|
|
int nsize = self->size * 2;
|
|
|
|
struct perf_header_attr **nattr;
|
|
|
|
|
|
|
|
nattr = realloc(self->attr, nsize * sizeof(void *));
|
|
|
|
if (nattr == NULL)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
self->size = nsize;
|
|
|
|
self->attr = nattr;
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
2009-11-19 09:55:55 -07:00
|
|
|
|
|
|
|
self->attr[self->attrs++] = attr;
|
2009-11-16 20:18:09 -07:00
|
|
|
return 0;
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
|
|
|
|
2009-09-19 05:36:30 -06:00
|
|
|
#define MAX_EVENT_NAME 64
|
|
|
|
|
2009-09-11 23:52:51 -06:00
|
|
|
struct perf_trace_event_type {
|
|
|
|
u64 event_id;
|
2009-09-19 05:36:30 -06:00
|
|
|
char name[MAX_EVENT_NAME];
|
2009-09-11 23:52:51 -06:00
|
|
|
};
|
|
|
|
|
|
|
|
static int event_count;
|
|
|
|
static struct perf_trace_event_type *events;
|
|
|
|
|
|
|
|
void perf_header__push_event(u64 id, const char *name)
|
|
|
|
{
|
2009-09-19 05:36:30 -06:00
|
|
|
if (strlen(name) > MAX_EVENT_NAME)
|
2009-10-21 13:34:06 -06:00
|
|
|
pr_warning("Event %s will be truncated\n", name);
|
2009-09-11 23:52:51 -06:00
|
|
|
|
|
|
|
if (!events) {
|
|
|
|
events = malloc(sizeof(struct perf_trace_event_type));
|
|
|
|
if (!events)
|
|
|
|
die("nomem");
|
|
|
|
} else {
|
|
|
|
events = realloc(events, (event_count + 1) * sizeof(struct perf_trace_event_type));
|
|
|
|
if (!events)
|
|
|
|
die("nomem");
|
|
|
|
}
|
|
|
|
memset(&events[event_count], 0, sizeof(struct perf_trace_event_type));
|
|
|
|
events[event_count].event_id = id;
|
2009-09-19 05:36:30 -06:00
|
|
|
strncpy(events[event_count].name, name, MAX_EVENT_NAME - 1);
|
2009-09-11 23:52:51 -06:00
|
|
|
event_count++;
|
|
|
|
}
|
|
|
|
|
|
|
|
char *perf_header__find_event(u64 id)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
for (i = 0 ; i < event_count; i++) {
|
|
|
|
if (events[i].event_id == id)
|
|
|
|
return events[i].name;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2009-06-25 09:05:54 -06:00
|
|
|
static const char *__perf_magic = "PERFFILE";
|
|
|
|
|
|
|
|
#define PERF_MAGIC (*(u64 *)__perf_magic)
|
|
|
|
|
|
|
|
struct perf_file_attr {
|
perf: Do the big rename: Performance Counters -> Performance Events
Bye-bye Performance Counters, welcome Performance Events!
In the past few months the perfcounters subsystem has grown out its
initial role of counting hardware events, and has become (and is
becoming) a much broader generic event enumeration, reporting, logging,
monitoring, analysis facility.
Naming its core object 'perf_counter' and naming the subsystem
'perfcounters' has become more and more of a misnomer. With pending
code like hw-breakpoints support the 'counter' name is less and
less appropriate.
All in one, we've decided to rename the subsystem to 'performance
events' and to propagate this rename through all fields, variables
and API names. (in an ABI compatible fashion)
The word 'event' is also a bit shorter than 'counter' - which makes
it slightly more convenient to write/handle as well.
Thanks goes to Stephane Eranian who first observed this misnomer and
suggested a rename.
User-space tooling and ABI compatibility is not affected - this patch
should be function-invariant. (Also, defconfigs were not touched to
keep the size down.)
This patch has been generated via the following script:
FILES=$(find * -type f | grep -vE 'oprofile|[^K]config')
sed -i \
-e 's/PERF_EVENT_/PERF_RECORD_/g' \
-e 's/PERF_COUNTER/PERF_EVENT/g' \
-e 's/perf_counter/perf_event/g' \
-e 's/nb_counters/nb_events/g' \
-e 's/swcounter/swevent/g' \
-e 's/tpcounter_event/tp_event/g' \
$FILES
for N in $(find . -name perf_counter.[ch]); do
M=$(echo $N | sed 's/perf_counter/perf_event/g')
mv $N $M
done
FILES=$(find . -name perf_event.*)
sed -i \
-e 's/COUNTER_MASK/REG_MASK/g' \
-e 's/COUNTER/EVENT/g' \
-e 's/\<event\>/event_id/g' \
-e 's/counter/event/g' \
-e 's/Counter/Event/g' \
$FILES
... to keep it as correct as possible. This script can also be
used by anyone who has pending perfcounters patches - it converts
a Linux kernel tree over to the new naming. We tried to time this
change to the point in time where the amount of pending patches
is the smallest: the end of the merge window.
Namespace clashes were fixed up in a preparatory patch - and some
stylistic fallout will be fixed up in a subsequent patch.
( NOTE: 'counters' are still the proper terminology when we deal
with hardware registers - and these sed scripts are a bit
over-eager in renaming them. I've undone some of that, but
in case there's something left where 'counter' would be
better than 'event' we can undo that on an individual basis
instead of touching an otherwise nicely automated patch. )
Suggested-by: Stephane Eranian <eranian@google.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Paul Mackerras <paulus@samba.org>
Reviewed-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: <linux-arch@vger.kernel.org>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-21 04:02:48 -06:00
|
|
|
struct perf_event_attr attr;
|
2009-06-25 09:05:54 -06:00
|
|
|
struct perf_file_section ids;
|
|
|
|
};
|
|
|
|
|
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>
2009-11-04 13:50:43 -07:00
|
|
|
void perf_header__set_feat(struct perf_header *self, int feat)
|
|
|
|
{
|
|
|
|
set_bit(feat, self->adds_features);
|
|
|
|
}
|
|
|
|
|
|
|
|
bool perf_header__has_feat(const struct perf_header *self, int feat)
|
|
|
|
{
|
|
|
|
return test_bit(feat, self->adds_features);
|
|
|
|
}
|
|
|
|
|
2009-11-16 20:18:12 -07:00
|
|
|
static int do_write(int fd, const void *buf, size_t size)
|
2009-06-25 09:05:54 -06:00
|
|
|
{
|
|
|
|
while (size) {
|
|
|
|
int ret = write(fd, buf, size);
|
|
|
|
|
|
|
|
if (ret < 0)
|
2009-11-19 09:55:56 -07:00
|
|
|
return -errno;
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
size -= ret;
|
|
|
|
buf += ret;
|
|
|
|
}
|
2009-11-16 20:18:12 -07:00
|
|
|
|
|
|
|
return 0;
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
|
|
|
|
perf record: Introduce a symtab cache
Now a cache will be created in a ~/.debug debuginfo like
hierarchy, so that at the end of a 'perf record' session all the
binaries (with build-ids) involved get collected and indexed by
their build-ids, so that perf report can find them.
This is interesting when developing software where you want to
do a 'perf diff' with the previous build and opens avenues for
lots more interesting tools, like a 'perf diff --graph' that
takes more than two binaries into account.
Tunables for collecting just the symtabs can be added if one
doesn't want to have the full binary, but having the full binary
allows things like 'perf rerecord' or other tools that can
re-run the tests by having access to the exact binary in some
perf.data file, so it may well be interesting to keep the full
binary there.
Space consumption is minimised by trying to use hard links, a
'perf cache' tool to manage the space used, a la ccache is
required to purge older entries.
With this in place it will be possible also to introduce new
commands, 'perf archive' and 'perf restore' (or some more
suitable and future proof names) to create a cpio/tar file with
the perf data and the files in the cache that _had_ perf hits of
interest.
There are more aspects to polish, like finding the right vmlinux
file to cache, etc, but this is enough for a first step.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-27 16:37:06 -07:00
|
|
|
#define dsos__for_each_with_build_id(pos, head) \
|
|
|
|
list_for_each_entry(pos, head, node) \
|
|
|
|
if (!pos->has_build_id) \
|
|
|
|
continue; \
|
|
|
|
else
|
|
|
|
|
2009-11-27 11:29:14 -07:00
|
|
|
static int __dsos__write_buildid_table(struct list_head *head, int fd)
|
2009-11-10 20:51:03 -07:00
|
|
|
{
|
2009-12-06 04:10:49 -07:00
|
|
|
#define NAME_ALIGN 64
|
2009-11-18 15:20:51 -07:00
|
|
|
struct dso *pos;
|
2009-12-06 04:10:49 -07:00
|
|
|
static const char zero_buf[NAME_ALIGN];
|
2009-11-18 15:20:51 -07:00
|
|
|
|
perf record: Introduce a symtab cache
Now a cache will be created in a ~/.debug debuginfo like
hierarchy, so that at the end of a 'perf record' session all the
binaries (with build-ids) involved get collected and indexed by
their build-ids, so that perf report can find them.
This is interesting when developing software where you want to
do a 'perf diff' with the previous build and opens avenues for
lots more interesting tools, like a 'perf diff --graph' that
takes more than two binaries into account.
Tunables for collecting just the symtabs can be added if one
doesn't want to have the full binary, but having the full binary
allows things like 'perf rerecord' or other tools that can
re-run the tests by having access to the exact binary in some
perf.data file, so it may well be interesting to keep the full
binary there.
Space consumption is minimised by trying to use hard links, a
'perf cache' tool to manage the space used, a la ccache is
required to purge older entries.
With this in place it will be possible also to introduce new
commands, 'perf archive' and 'perf restore' (or some more
suitable and future proof names) to create a cpio/tar file with
the perf data and the files in the cache that _had_ perf hits of
interest.
There are more aspects to polish, like finding the right vmlinux
file to cache, etc, but this is enough for a first step.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-27 16:37:06 -07:00
|
|
|
dsos__for_each_with_build_id(pos, head) {
|
2009-11-19 09:55:56 -07:00
|
|
|
int err;
|
2009-11-18 15:20:51 -07:00
|
|
|
struct build_id_event b;
|
perf record: Introduce a symtab cache
Now a cache will be created in a ~/.debug debuginfo like
hierarchy, so that at the end of a 'perf record' session all the
binaries (with build-ids) involved get collected and indexed by
their build-ids, so that perf report can find them.
This is interesting when developing software where you want to
do a 'perf diff' with the previous build and opens avenues for
lots more interesting tools, like a 'perf diff --graph' that
takes more than two binaries into account.
Tunables for collecting just the symtabs can be added if one
doesn't want to have the full binary, but having the full binary
allows things like 'perf rerecord' or other tools that can
re-run the tests by having access to the exact binary in some
perf.data file, so it may well be interesting to keep the full
binary there.
Space consumption is minimised by trying to use hard links, a
'perf cache' tool to manage the space used, a la ccache is
required to purge older entries.
With this in place it will be possible also to introduce new
commands, 'perf archive' and 'perf restore' (or some more
suitable and future proof names) to create a cpio/tar file with
the perf data and the files in the cache that _had_ perf hits of
interest.
There are more aspects to polish, like finding the right vmlinux
file to cache, etc, but this is enough for a first step.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-27 16:37:06 -07:00
|
|
|
size_t len = pos->long_name_len + 1;
|
2009-11-18 15:20:51 -07:00
|
|
|
|
2009-12-06 04:10:49 -07:00
|
|
|
len = ALIGN(len, NAME_ALIGN);
|
2009-11-18 15:20:51 -07:00
|
|
|
memset(&b, 0, sizeof(b));
|
|
|
|
memcpy(&b.build_id, pos->build_id, sizeof(pos->build_id));
|
|
|
|
b.header.size = sizeof(b) + len;
|
2009-11-19 09:55:56 -07:00
|
|
|
err = do_write(fd, &b, sizeof(b));
|
|
|
|
if (err < 0)
|
|
|
|
return err;
|
2009-12-06 04:10:49 -07:00
|
|
|
err = do_write(fd, pos->long_name, pos->long_name_len + 1);
|
|
|
|
if (err < 0)
|
|
|
|
return err;
|
2009-12-06 21:07:15 -07:00
|
|
|
err = do_write(fd, zero_buf, len - pos->long_name_len - 1);
|
2009-11-19 09:55:56 -07:00
|
|
|
if (err < 0)
|
|
|
|
return err;
|
2009-11-10 20:51:04 -07:00
|
|
|
}
|
2009-11-16 20:18:12 -07:00
|
|
|
|
|
|
|
return 0;
|
2009-11-10 20:51:03 -07:00
|
|
|
}
|
|
|
|
|
2009-11-27 11:29:14 -07:00
|
|
|
static int dsos__write_buildid_table(int fd)
|
|
|
|
{
|
|
|
|
int err = __dsos__write_buildid_table(&dsos__kernel, fd);
|
|
|
|
if (err == 0)
|
|
|
|
err = __dsos__write_buildid_table(&dsos__user, fd);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
perf record: Introduce a symtab cache
Now a cache will be created in a ~/.debug debuginfo like
hierarchy, so that at the end of a 'perf record' session all the
binaries (with build-ids) involved get collected and indexed by
their build-ids, so that perf report can find them.
This is interesting when developing software where you want to
do a 'perf diff' with the previous build and opens avenues for
lots more interesting tools, like a 'perf diff --graph' that
takes more than two binaries into account.
Tunables for collecting just the symtabs can be added if one
doesn't want to have the full binary, but having the full binary
allows things like 'perf rerecord' or other tools that can
re-run the tests by having access to the exact binary in some
perf.data file, so it may well be interesting to keep the full
binary there.
Space consumption is minimised by trying to use hard links, a
'perf cache' tool to manage the space used, a la ccache is
required to purge older entries.
With this in place it will be possible also to introduce new
commands, 'perf archive' and 'perf restore' (or some more
suitable and future proof names) to create a cpio/tar file with
the perf data and the files in the cache that _had_ perf hits of
interest.
There are more aspects to polish, like finding the right vmlinux
file to cache, etc, but this is enough for a first step.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-27 16:37:06 -07:00
|
|
|
static int dso__cache_build_id(struct dso *self, const char *debugdir)
|
|
|
|
{
|
|
|
|
const size_t size = PATH_MAX;
|
|
|
|
char *filename = malloc(size),
|
|
|
|
*linkname = malloc(size), *targetname, *sbuild_id;
|
|
|
|
int len, err = -1;
|
|
|
|
|
|
|
|
if (filename == NULL || linkname == NULL)
|
|
|
|
goto out_free;
|
|
|
|
|
|
|
|
len = snprintf(filename, size, "%s%s", debugdir, self->long_name);
|
|
|
|
if (mkdir_p(filename, 0755))
|
|
|
|
goto out_free;
|
|
|
|
|
|
|
|
len += snprintf(filename + len, sizeof(filename) - len, "/");
|
|
|
|
sbuild_id = filename + len;
|
|
|
|
build_id__sprintf(self->build_id, sizeof(self->build_id), sbuild_id);
|
|
|
|
|
|
|
|
if (access(filename, F_OK) && link(self->long_name, filename) &&
|
|
|
|
copyfile(self->long_name, filename))
|
|
|
|
goto out_free;
|
|
|
|
|
|
|
|
len = snprintf(linkname, size, "%s/.build-id/%.2s",
|
|
|
|
debugdir, sbuild_id);
|
|
|
|
|
|
|
|
if (access(linkname, X_OK) && mkdir_p(linkname, 0755))
|
|
|
|
goto out_free;
|
|
|
|
|
|
|
|
snprintf(linkname + len, size - len, "/%s", sbuild_id + 2);
|
|
|
|
targetname = filename + strlen(debugdir) - 5;
|
|
|
|
memcpy(targetname, "../..", 5);
|
|
|
|
|
|
|
|
if (symlink(targetname, linkname) == 0)
|
|
|
|
err = 0;
|
|
|
|
out_free:
|
|
|
|
free(filename);
|
|
|
|
free(linkname);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __dsos__cache_build_ids(struct list_head *head, const char *debugdir)
|
|
|
|
{
|
|
|
|
struct dso *pos;
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
dsos__for_each_with_build_id(pos, head)
|
|
|
|
if (dso__cache_build_id(pos, debugdir))
|
|
|
|
err = -1;
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int dsos__cache_build_ids(void)
|
|
|
|
{
|
|
|
|
int err_kernel, err_user;
|
|
|
|
char debugdir[PATH_MAX];
|
|
|
|
|
|
|
|
snprintf(debugdir, sizeof(debugdir), "%s/%s", getenv("HOME"),
|
|
|
|
DEBUG_CACHE_DIR);
|
|
|
|
|
|
|
|
if (mkdir(debugdir, 0755) != 0 && errno != EEXIST)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
err_kernel = __dsos__cache_build_ids(&dsos__kernel, debugdir);
|
|
|
|
err_user = __dsos__cache_build_ids(&dsos__user, debugdir);
|
|
|
|
return err_kernel || err_user ? -1 : 0;
|
|
|
|
}
|
|
|
|
|
2009-11-19 09:55:56 -07:00
|
|
|
static int perf_header__adds_write(struct perf_header *self, int fd)
|
2009-10-17 09:12:34 -06:00
|
|
|
{
|
2009-11-10 20:51:07 -07:00
|
|
|
int nr_sections;
|
|
|
|
struct perf_file_section *feat_sec;
|
|
|
|
int sec_size;
|
|
|
|
u64 sec_start;
|
2009-11-19 09:55:56 -07:00
|
|
|
int idx = 0, err;
|
2009-11-10 20:51:07 -07:00
|
|
|
|
2009-11-18 15:20:51 -07:00
|
|
|
if (dsos__read_build_ids())
|
2009-11-10 20:51:07 -07:00
|
|
|
perf_header__set_feat(self, HEADER_BUILD_ID);
|
|
|
|
|
|
|
|
nr_sections = bitmap_weight(self->adds_features, HEADER_FEAT_BITS);
|
|
|
|
if (!nr_sections)
|
2009-11-19 09:55:56 -07:00
|
|
|
return 0;
|
2009-11-10 20:51:07 -07:00
|
|
|
|
|
|
|
feat_sec = calloc(sizeof(*feat_sec), nr_sections);
|
2009-11-19 09:55:56 -07:00
|
|
|
if (feat_sec == NULL)
|
|
|
|
return -ENOMEM;
|
2009-11-10 20:51:07 -07:00
|
|
|
|
|
|
|
sec_size = sizeof(*feat_sec) * nr_sections;
|
|
|
|
|
|
|
|
sec_start = self->data_offset + self->data_size;
|
|
|
|
lseek(fd, sec_start + sec_size, SEEK_SET);
|
2009-10-17 09:12:34 -06:00
|
|
|
|
2009-11-10 20:51:06 -07:00
|
|
|
if (perf_header__has_feat(self, HEADER_TRACE_INFO)) {
|
2009-11-10 20:51:07 -07:00
|
|
|
struct perf_file_section *trace_sec;
|
|
|
|
|
|
|
|
trace_sec = &feat_sec[idx++];
|
|
|
|
|
2009-10-17 09:12:34 -06:00
|
|
|
/* Write trace info */
|
2009-11-10 20:51:07 -07:00
|
|
|
trace_sec->offset = lseek(fd, 0, SEEK_CUR);
|
2009-10-17 09:12:34 -06:00
|
|
|
read_tracing_data(fd, attrs, nr_counters);
|
2009-11-10 20:51:07 -07:00
|
|
|
trace_sec->size = lseek(fd, 0, SEEK_CUR) - trace_sec->offset;
|
2009-10-17 09:12:34 -06:00
|
|
|
}
|
2009-11-10 20:51:03 -07:00
|
|
|
|
2009-11-10 20:51:04 -07:00
|
|
|
|
2009-11-10 20:51:07 -07:00
|
|
|
if (perf_header__has_feat(self, HEADER_BUILD_ID)) {
|
|
|
|
struct perf_file_section *buildid_sec;
|
|
|
|
|
|
|
|
buildid_sec = &feat_sec[idx++];
|
|
|
|
|
|
|
|
/* Write build-ids */
|
|
|
|
buildid_sec->offset = lseek(fd, 0, SEEK_CUR);
|
2009-11-19 09:55:56 -07:00
|
|
|
err = dsos__write_buildid_table(fd);
|
|
|
|
if (err < 0) {
|
|
|
|
pr_debug("failed to write buildid table\n");
|
|
|
|
goto out_free;
|
|
|
|
}
|
2009-11-10 20:51:07 -07:00
|
|
|
buildid_sec->size = lseek(fd, 0, SEEK_CUR) - buildid_sec->offset;
|
perf record: Introduce a symtab cache
Now a cache will be created in a ~/.debug debuginfo like
hierarchy, so that at the end of a 'perf record' session all the
binaries (with build-ids) involved get collected and indexed by
their build-ids, so that perf report can find them.
This is interesting when developing software where you want to
do a 'perf diff' with the previous build and opens avenues for
lots more interesting tools, like a 'perf diff --graph' that
takes more than two binaries into account.
Tunables for collecting just the symtabs can be added if one
doesn't want to have the full binary, but having the full binary
allows things like 'perf rerecord' or other tools that can
re-run the tests by having access to the exact binary in some
perf.data file, so it may well be interesting to keep the full
binary there.
Space consumption is minimised by trying to use hard links, a
'perf cache' tool to manage the space used, a la ccache is
required to purge older entries.
With this in place it will be possible also to introduce new
commands, 'perf archive' and 'perf restore' (or some more
suitable and future proof names) to create a cpio/tar file with
the perf data and the files in the cache that _had_ perf hits of
interest.
There are more aspects to polish, like finding the right vmlinux
file to cache, etc, but this is enough for a first step.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1261957026-15580-10-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-27 16:37:06 -07:00
|
|
|
dsos__cache_build_ids();
|
2009-11-10 20:51:03 -07:00
|
|
|
}
|
2009-11-10 20:51:07 -07:00
|
|
|
|
|
|
|
lseek(fd, sec_start, SEEK_SET);
|
2009-11-19 09:55:56 -07:00
|
|
|
err = do_write(fd, feat_sec, sec_size);
|
|
|
|
if (err < 0)
|
|
|
|
pr_debug("failed to write feature section\n");
|
|
|
|
out_free:
|
2009-11-10 20:51:07 -07:00
|
|
|
free(feat_sec);
|
2009-11-19 09:55:56 -07:00
|
|
|
return err;
|
2009-11-10 20:51:07 -07:00
|
|
|
}
|
2009-10-17 09:12:34 -06:00
|
|
|
|
2009-11-19 09:55:56 -07:00
|
|
|
int perf_header__write(struct perf_header *self, int fd, bool at_exit)
|
2009-06-25 09:05:54 -06:00
|
|
|
{
|
|
|
|
struct perf_file_header f_header;
|
|
|
|
struct perf_file_attr f_attr;
|
|
|
|
struct perf_header_attr *attr;
|
2009-11-19 09:55:56 -07:00
|
|
|
int i, err;
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
lseek(fd, sizeof(f_header), SEEK_SET);
|
|
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < self->attrs; i++) {
|
|
|
|
attr = self->attr[i];
|
|
|
|
|
|
|
|
attr->id_offset = lseek(fd, 0, SEEK_CUR);
|
2009-11-19 09:55:56 -07:00
|
|
|
err = do_write(fd, attr->id, attr->ids * sizeof(u64));
|
|
|
|
if (err < 0) {
|
|
|
|
pr_debug("failed to write perf header\n");
|
|
|
|
return err;
|
|
|
|
}
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
self->attr_offset = lseek(fd, 0, SEEK_CUR);
|
|
|
|
|
|
|
|
for (i = 0; i < self->attrs; i++) {
|
|
|
|
attr = self->attr[i];
|
|
|
|
|
|
|
|
f_attr = (struct perf_file_attr){
|
|
|
|
.attr = attr->attr,
|
|
|
|
.ids = {
|
|
|
|
.offset = attr->id_offset,
|
|
|
|
.size = attr->ids * sizeof(u64),
|
|
|
|
}
|
|
|
|
};
|
2009-11-19 09:55:56 -07:00
|
|
|
err = do_write(fd, &f_attr, sizeof(f_attr));
|
|
|
|
if (err < 0) {
|
|
|
|
pr_debug("failed to write perf header attribute\n");
|
|
|
|
return err;
|
|
|
|
}
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
|
|
|
|
2009-09-11 23:52:51 -06:00
|
|
|
self->event_offset = lseek(fd, 0, SEEK_CUR);
|
|
|
|
self->event_size = event_count * sizeof(struct perf_trace_event_type);
|
2009-11-19 09:55:56 -07:00
|
|
|
if (events) {
|
|
|
|
err = do_write(fd, events, self->event_size);
|
|
|
|
if (err < 0) {
|
|
|
|
pr_debug("failed to write perf header events\n");
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
}
|
2009-09-11 23:52:51 -06:00
|
|
|
|
2009-06-25 09:05:54 -06:00
|
|
|
self->data_offset = lseek(fd, 0, SEEK_CUR);
|
|
|
|
|
2009-11-19 09:55:56 -07:00
|
|
|
if (at_exit) {
|
|
|
|
err = perf_header__adds_write(self, fd);
|
|
|
|
if (err < 0)
|
|
|
|
return err;
|
|
|
|
}
|
2009-11-10 20:51:07 -07:00
|
|
|
|
2009-06-25 09:05:54 -06:00
|
|
|
f_header = (struct perf_file_header){
|
|
|
|
.magic = PERF_MAGIC,
|
|
|
|
.size = sizeof(f_header),
|
|
|
|
.attr_size = sizeof(f_attr),
|
|
|
|
.attrs = {
|
|
|
|
.offset = self->attr_offset,
|
|
|
|
.size = self->attrs * sizeof(f_attr),
|
|
|
|
},
|
|
|
|
.data = {
|
|
|
|
.offset = self->data_offset,
|
|
|
|
.size = self->data_size,
|
|
|
|
},
|
2009-09-11 23:52:51 -06:00
|
|
|
.event_types = {
|
|
|
|
.offset = self->event_offset,
|
|
|
|
.size = self->event_size,
|
|
|
|
},
|
2009-06-25 09:05:54 -06:00
|
|
|
};
|
|
|
|
|
2009-10-17 09:57:18 -06:00
|
|
|
memcpy(&f_header.adds_features, &self->adds_features, sizeof(self->adds_features));
|
2009-10-17 09:12:34 -06:00
|
|
|
|
2009-06-25 09:05:54 -06:00
|
|
|
lseek(fd, 0, SEEK_SET);
|
2009-11-19 09:55:56 -07:00
|
|
|
err = do_write(fd, &f_header, sizeof(f_header));
|
|
|
|
if (err < 0) {
|
|
|
|
pr_debug("failed to write perf header\n");
|
|
|
|
return err;
|
|
|
|
}
|
2009-06-25 09:05:54 -06:00
|
|
|
lseek(fd, self->data_offset + self->data_size, SEEK_SET);
|
|
|
|
|
|
|
|
self->frozen = 1;
|
2009-11-19 09:55:56 -07:00
|
|
|
return 0;
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
static void do_read(int fd, void *buf, size_t size)
|
|
|
|
{
|
|
|
|
while (size) {
|
|
|
|
int ret = read(fd, buf, size);
|
|
|
|
|
|
|
|
if (ret < 0)
|
|
|
|
die("failed to read");
|
2009-08-07 06:16:00 -06:00
|
|
|
if (ret == 0)
|
|
|
|
die("failed to read: missing data");
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
size -= ret;
|
|
|
|
buf += ret;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-11-16 11:32:43 -07:00
|
|
|
int perf_header__process_sections(struct perf_header *self, int fd,
|
|
|
|
int (*process)(struct perf_file_section *self,
|
|
|
|
int feat, int fd))
|
2009-10-17 09:12:34 -06:00
|
|
|
{
|
2009-11-10 20:51:07 -07:00
|
|
|
struct perf_file_section *feat_sec;
|
|
|
|
int nr_sections;
|
|
|
|
int sec_size;
|
|
|
|
int idx = 0;
|
2009-11-16 11:32:43 -07:00
|
|
|
int err = 0, feat = 1;
|
2009-11-10 20:51:07 -07:00
|
|
|
|
|
|
|
nr_sections = bitmap_weight(self->adds_features, HEADER_FEAT_BITS);
|
|
|
|
if (!nr_sections)
|
2009-11-16 11:32:43 -07:00
|
|
|
return 0;
|
2009-11-10 20:51:07 -07:00
|
|
|
|
|
|
|
feat_sec = calloc(sizeof(*feat_sec), nr_sections);
|
|
|
|
if (!feat_sec)
|
2009-11-16 11:32:43 -07:00
|
|
|
return -1;
|
2009-11-10 20:51:07 -07:00
|
|
|
|
|
|
|
sec_size = sizeof(*feat_sec) * nr_sections;
|
|
|
|
|
|
|
|
lseek(fd, self->data_offset + self->data_size, SEEK_SET);
|
|
|
|
|
|
|
|
do_read(fd, feat_sec, sec_size);
|
|
|
|
|
2009-11-16 11:32:43 -07:00
|
|
|
while (idx < nr_sections && feat < HEADER_LAST_FEATURE) {
|
|
|
|
if (perf_header__has_feat(self, feat)) {
|
|
|
|
struct perf_file_section *sec = &feat_sec[idx++];
|
2009-10-17 09:12:34 -06:00
|
|
|
|
2009-11-16 11:32:43 -07:00
|
|
|
err = process(sec, feat, fd);
|
|
|
|
if (err < 0)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
++feat;
|
2009-10-17 09:12:34 -06:00
|
|
|
}
|
2009-11-10 20:51:05 -07:00
|
|
|
|
2009-11-16 11:32:43 -07:00
|
|
|
free(feat_sec);
|
|
|
|
return err;
|
|
|
|
};
|
2009-11-10 20:51:05 -07:00
|
|
|
|
2009-11-16 11:32:43 -07:00
|
|
|
int perf_file_header__read(struct perf_file_header *self,
|
|
|
|
struct perf_header *ph, int fd)
|
|
|
|
{
|
|
|
|
lseek(fd, 0, SEEK_SET);
|
|
|
|
do_read(fd, self, sizeof(*self));
|
|
|
|
|
|
|
|
if (self->magic != PERF_MAGIC ||
|
|
|
|
self->attr_size != sizeof(struct perf_file_attr))
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
if (self->size != sizeof(*self)) {
|
|
|
|
/* Support the previous format */
|
|
|
|
if (self->size == offsetof(typeof(*self), adds_features))
|
|
|
|
bitmap_zero(self->adds_features, HEADER_FEAT_BITS);
|
|
|
|
else
|
|
|
|
return -1;
|
2009-11-10 20:51:05 -07:00
|
|
|
}
|
2009-11-10 20:51:07 -07:00
|
|
|
|
2009-11-16 11:32:43 -07:00
|
|
|
memcpy(&ph->adds_features, &self->adds_features,
|
|
|
|
sizeof(self->adds_features));
|
|
|
|
|
|
|
|
ph->event_offset = self->event_types.offset;
|
|
|
|
ph->event_size = self->event_types.size;
|
|
|
|
ph->data_offset = self->data.offset;
|
|
|
|
ph->data_size = self->data.size;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int perf_file_section__process(struct perf_file_section *self,
|
|
|
|
int feat, int fd)
|
|
|
|
{
|
|
|
|
if (lseek(fd, self->offset, SEEK_SET) < 0) {
|
|
|
|
pr_debug("Failed to lseek to %Ld offset for feature %d, "
|
|
|
|
"continuing...\n", self->offset, feat);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
switch (feat) {
|
|
|
|
case HEADER_TRACE_INFO:
|
|
|
|
trace_report(fd);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case HEADER_BUILD_ID:
|
|
|
|
if (perf_header__read_build_ids(fd, self->offset, self->size))
|
|
|
|
pr_debug("Failed to read buildids, continuing...\n");
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
pr_debug("unknown feature %d, continuing...\n", feat);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2009-10-17 09:12:34 -06:00
|
|
|
|
2009-11-19 09:55:55 -07:00
|
|
|
int perf_header__read(struct perf_header *self, int fd)
|
2009-06-25 09:05:54 -06:00
|
|
|
{
|
|
|
|
struct perf_file_header f_header;
|
|
|
|
struct perf_file_attr f_attr;
|
|
|
|
u64 f_id;
|
|
|
|
int nr_attrs, nr_ids, i, j;
|
|
|
|
|
2009-11-19 09:55:55 -07:00
|
|
|
if (perf_file_header__read(&f_header, self, fd) < 0) {
|
|
|
|
pr_debug("incompatible file format\n");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
nr_attrs = f_header.attrs.size / sizeof(f_attr);
|
|
|
|
lseek(fd, f_header.attrs.offset, SEEK_SET);
|
|
|
|
|
|
|
|
for (i = 0; i < nr_attrs; i++) {
|
|
|
|
struct perf_header_attr *attr;
|
2009-08-06 12:57:41 -06:00
|
|
|
off_t tmp;
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
do_read(fd, &f_attr, sizeof(f_attr));
|
2009-08-06 12:57:41 -06:00
|
|
|
tmp = lseek(fd, 0, SEEK_CUR);
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
attr = perf_header_attr__new(&f_attr.attr);
|
2009-11-16 14:30:26 -07:00
|
|
|
if (attr == NULL)
|
2009-11-19 09:55:55 -07:00
|
|
|
return -ENOMEM;
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
nr_ids = f_attr.ids.size / sizeof(u64);
|
|
|
|
lseek(fd, f_attr.ids.offset, SEEK_SET);
|
|
|
|
|
|
|
|
for (j = 0; j < nr_ids; j++) {
|
|
|
|
do_read(fd, &f_id, sizeof(f_id));
|
|
|
|
|
2009-11-19 09:55:55 -07:00
|
|
|
if (perf_header_attr__add_id(attr, f_id) < 0) {
|
|
|
|
perf_header_attr__delete(attr);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (perf_header__add_attr(self, attr) < 0) {
|
|
|
|
perf_header_attr__delete(attr);
|
|
|
|
return -ENOMEM;
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
2009-11-16 20:18:09 -07:00
|
|
|
|
2009-06-25 09:05:54 -06:00
|
|
|
lseek(fd, tmp, SEEK_SET);
|
|
|
|
}
|
|
|
|
|
2009-09-11 23:52:51 -06:00
|
|
|
if (f_header.event_types.size) {
|
|
|
|
lseek(fd, f_header.event_types.offset, SEEK_SET);
|
|
|
|
events = malloc(f_header.event_types.size);
|
2009-11-19 09:55:55 -07:00
|
|
|
if (events == NULL)
|
|
|
|
return -ENOMEM;
|
2009-09-11 23:52:51 -06:00
|
|
|
do_read(fd, events, f_header.event_types.size);
|
|
|
|
event_count = f_header.event_types.size / sizeof(struct perf_trace_event_type);
|
|
|
|
}
|
2009-10-06 15:36:47 -06:00
|
|
|
|
2009-11-16 11:32:43 -07:00
|
|
|
perf_header__process_sections(self, fd, perf_file_section__process);
|
2009-11-10 20:51:05 -07:00
|
|
|
|
2009-09-03 08:21:11 -06:00
|
|
|
lseek(fd, self->data_offset, SEEK_SET);
|
2009-06-25 09:05:54 -06:00
|
|
|
|
|
|
|
self->frozen = 1;
|
2009-11-19 09:55:55 -07:00
|
|
|
return 0;
|
2009-06-25 09:05:54 -06:00
|
|
|
}
|
2009-08-16 12:56:37 -06:00
|
|
|
|
|
|
|
u64 perf_header__sample_type(struct perf_header *header)
|
|
|
|
{
|
|
|
|
u64 type = 0;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < header->attrs; i++) {
|
|
|
|
struct perf_header_attr *attr = header->attr[i];
|
|
|
|
|
|
|
|
if (!type)
|
|
|
|
type = attr->attr.sample_type;
|
|
|
|
else if (type != attr->attr.sample_type)
|
|
|
|
die("non matching sample_type");
|
|
|
|
}
|
|
|
|
|
|
|
|
return type;
|
|
|
|
}
|
|
|
|
|
perf: Do the big rename: Performance Counters -> Performance Events
Bye-bye Performance Counters, welcome Performance Events!
In the past few months the perfcounters subsystem has grown out its
initial role of counting hardware events, and has become (and is
becoming) a much broader generic event enumeration, reporting, logging,
monitoring, analysis facility.
Naming its core object 'perf_counter' and naming the subsystem
'perfcounters' has become more and more of a misnomer. With pending
code like hw-breakpoints support the 'counter' name is less and
less appropriate.
All in one, we've decided to rename the subsystem to 'performance
events' and to propagate this rename through all fields, variables
and API names. (in an ABI compatible fashion)
The word 'event' is also a bit shorter than 'counter' - which makes
it slightly more convenient to write/handle as well.
Thanks goes to Stephane Eranian who first observed this misnomer and
suggested a rename.
User-space tooling and ABI compatibility is not affected - this patch
should be function-invariant. (Also, defconfigs were not touched to
keep the size down.)
This patch has been generated via the following script:
FILES=$(find * -type f | grep -vE 'oprofile|[^K]config')
sed -i \
-e 's/PERF_EVENT_/PERF_RECORD_/g' \
-e 's/PERF_COUNTER/PERF_EVENT/g' \
-e 's/perf_counter/perf_event/g' \
-e 's/nb_counters/nb_events/g' \
-e 's/swcounter/swevent/g' \
-e 's/tpcounter_event/tp_event/g' \
$FILES
for N in $(find . -name perf_counter.[ch]); do
M=$(echo $N | sed 's/perf_counter/perf_event/g')
mv $N $M
done
FILES=$(find . -name perf_event.*)
sed -i \
-e 's/COUNTER_MASK/REG_MASK/g' \
-e 's/COUNTER/EVENT/g' \
-e 's/\<event\>/event_id/g' \
-e 's/counter/event/g' \
-e 's/Counter/Event/g' \
$FILES
... to keep it as correct as possible. This script can also be
used by anyone who has pending perfcounters patches - it converts
a Linux kernel tree over to the new naming. We tried to time this
change to the point in time where the amount of pending patches
is the smallest: the end of the merge window.
Namespace clashes were fixed up in a preparatory patch - and some
stylistic fallout will be fixed up in a subsequent patch.
( NOTE: 'counters' are still the proper terminology when we deal
with hardware registers - and these sed scripts are a bit
over-eager in renaming them. I've undone some of that, but
in case there's something left where 'counter' would be
better than 'event' we can undo that on an individual basis
instead of touching an otherwise nicely automated patch. )
Suggested-by: Stephane Eranian <eranian@google.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Paul Mackerras <paulus@samba.org>
Reviewed-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: <linux-arch@vger.kernel.org>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-21 04:02:48 -06:00
|
|
|
struct perf_event_attr *
|
2009-08-16 12:56:37 -06:00
|
|
|
perf_header__find_attr(u64 id, struct perf_header *header)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < header->attrs; i++) {
|
|
|
|
struct perf_header_attr *attr = header->attr[i];
|
|
|
|
int j;
|
|
|
|
|
|
|
|
for (j = 0; j < attr->ids; j++) {
|
|
|
|
if (attr->id[j] == id)
|
|
|
|
return &attr->attr;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|