We were relying on the info in pevent, but since we have it in
perf_evsel, set up by the perf_session routine if read from a perf.data
file or by whoever creates the evsels, use it.
New 'perf test' entries will use it to parse locally generated events,
in a non perf.data centered workflow.
As well as use byteswap.h to get per arch optimized swap routines, like
other parts of perf (header, perf_evsel__parse_sample, symbol, etc)
already do.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-8tjuxk09mlsfmh7macgkxsip@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Just like was done for parse_events__set_leader.
Also we need to have the list_entry set_leader method in evlist.c so that we
don't grow another dep in the python binding:
# ~acme/git/linux/tools/perf/python/twatch.py
Traceback (most recent call last):
File "/home/acme/git/linux/tools/perf/python/twatch.py", line 16, in <module>
import perf
ImportError: /home/acme/git/build/perf/python/perf.so: undefined symbol: parse_events__set_leader
And also remove a pr_debug from evsel.c so that we avoid this one too:
# ~acme/git/linux/tools/perf/python/twatch.py
Traceback (most recent call last):
File "/home/acme/git/linux/tools/perf/python/twatch.py", line 16, in <module>
import perf
ImportError: /home/acme/git/build/perf/python/perf.so: undefined symbol: eprintf
Acked-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Namhyung Kim <namhyung@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-0hk9dazg9pora9jylkqngovm@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This patch adds a functionality that allows to create event groups
based on the way they are specified on the command line. Adding
functionality to the '{}' group syntax introduced in earlier patch.
The current '--group/-g' option behaviour remains intact. If you
specify it for record/stat/top command, all the specified events
become members of a single group with the first event as a group
leader.
With the new '{}' group syntax you can create group like:
# perf record -e '{cycles,faults}' ls
resulting in single event group containing 'cycles' and 'faults'
events, with cycles event as group leader.
All groups are created with regards to threads and cpus. Thus
recording an event group within a 2 threads on server with
4 CPUs will create 8 separate groups.
Examples (first event in brackets is group leader):
# 1 group (cpu-clock,task-clock)
perf record --group -e cpu-clock,task-clock ls
perf record -e '{cpu-clock,task-clock}' ls
# 2 groups (cpu-clock,task-clock) (minor-faults,major-faults)
perf record -e '{cpu-clock,task-clock},{minor-faults,major-faults}' ls
# 1 group (cpu-clock,task-clock,minor-faults,major-faults)
perf record --group -e cpu-clock,task-clock -e minor-faults,major-faults ls
perf record -e '{cpu-clock,task-clock,minor-faults,major-faults}' ls
# 2 groups (cpu-clock,task-clock) (minor-faults,major-faults)
perf record -e '{cpu-clock,task-clock} -e '{minor-faults,major-faults}' \
-e instructions ls
# 1 group
# (cpu-clock,task-clock,minor-faults,major-faults,instructions)
perf record --group -e cpu-clock,task-clock \
-e minor-faults,major-faults -e instructions ls perf record -e
'{cpu-clock,task-clock,minor-faults,major-faults,instructions}' ls
It's possible to use standard event modifier for a group, which spans
over all events in the group and updates each event modifier settings,
for example:
# perf record -r '{faults:k,cache-references}:p'
resulting in ':kp' modifier being used for 'faults' and ':p' modifier
being used for 'cache-references' event.
Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ulrich Drepper <drepper@gmail.com>
Link: http://lkml.kernel.org/n/tip-ho42u0wcr8mn1otkalqi13qp@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The commit c410431cef ("perf tools: Reconstruct event with modifiers
from perf_event_attr") added the line, but it's broken since it needs to
go up 3 directories to get to the kernel root directory, not 2.
However host gcc contains /usr/local/include in its search path, so that
it can find the perf_event.h in /usr/include. This why we didn't notice
the problem yet. But when I tried to cross compile it appears like:
CC util/evsel.o
util/evsel.c:18:44: error: ../../include/linux/perf_event.h: No such file or directory
make: *** [util/evsel.o] Error 1
Looking at the source, it isn't needed at all as evsel.h already
included the perf_event.h. So simply remove it would solve the problem.
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1340268772-5737-1-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
We swap the sample_id_all header by u64 pointers. Some members of the
header happen to be 32 bit values. We need to handle them separatelly.
Together with other endianity patches, this change fixies perf report
discrepancies on origin and target systems as described in test 1 below,
e.g. following perf report diff:
...
0.12% ps [kernel.kallsyms] [k] clear_page
- 0.12% awk bash [.] alloc_word_desc
+ 0.12% awk bash [.] yyparse
0.11% beah-rhts-task libpython2.6.so.1.0 [.] 0x5560e
0.10% perf libc-2.12.so [.] __ctype_toupper_loc
- 0.09% rhts-test-runne bash [.] maybe_make_export_env
+ 0.09% rhts-test-runne bash [.] 0x385a0
0.09% ps [kernel.kallsyms] [k] page_fault
...
Note, running following to test perf endianity handling:
test 1)
- origin system:
# perf record -a -- sleep 10 (any perf record will do)
# perf report > report.origin
# perf archive perf.data
- copy the perf.data, report.origin and perf.data.tar.bz2
to a target system and run:
# tar xjvf perf.data.tar.bz2 -C ~/.debug
# perf report > report.target
# diff -u report.origin report.target
- the diff should produce no output
(besides some white space stuff and possibly different
date/TZ output)
test 2)
- origin system:
# perf record -ag -fo /tmp/perf.data -- sleep 1
- mount origin system root to the target system on /mnt/origin
- target system:
# perf script --symfs /mnt/origin -I -i /mnt/origin/tmp/perf.data \
--kallsyms /mnt/origin/proc/kallsyms
- complete perf.data header is displayed
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Tested-by: David Ahern <dsahern@gmail.com>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1338380624-7443-4-git-send-email-jolsa@redhat.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The modifiers:
k kernel space
u user space
h hypervisor
G guest
H host
p, pp, ppp precision level (PEBS)
that can be suffixed to an event were lost when tools used event_name()
to reconstruct them from the perf_event_attr entries in a perf.data
file.
Fix it by following the defaults used for these modifiers in the current
codebase, so:
$ perf record -e instructions:u usleep 1 2> /dev/null
$ perf evlist
instructions:u
$ perf record -e cycles:k usleep 1 2> /dev/null
$ perf evlist
cycles:k
$ perf record -e cycles:kh usleep 1 2> /dev/null
$ perf evlist
cycles:kh
$ perf record -e cache-misses:G usleep 1 2> /dev/null
$ perf evlist
cache-misses:G
$ perf record -e cycles:ppk usleep 1 2> /dev/null
$ perf evlist
cycles:kpp
$
Also works with 'top', 'report', etc.
More work needed to cover tracepoints and software events while not
dragging lots of baggage to the python binding, this is a minimal fix
for v3.5.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-4hl5glle0hxlklw4usva1mkt@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
764e16a changed perf-record to create events disabled by default and
enable them once perf initializations are done. This setting was dropped
by 0f82ebc. Now perf events are once again generated during perf's
initialization phase (e.g., generating maps).
As an example, perf opens a lot of files at startup. Unpatched:
perf record -e syscalls:sys_enter_open -ga -fo /tmp/perf.data -- sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.087 MB /tmp/perf.data (~3798 samples) ]
Using perf-script to look at the samples shows the perf command generating
563 of the 566 total events.
Patched:
perf record -e syscalls:sys_enter_open -ga -fo /tmp/perf.data -- sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.028 MB /tmp/perf.data (~1206 samples) ]
Using perf-script to look at the samples does not show perf command.
Signed-off-by: David Ahern <dsahern@gmail.com>
Link: http://lkml.kernel.org/r/1336968088-11531-1-git-send-email-dsahern@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
In perf_event__parse_sample(), the array variable was not incremented
by the amount of data used by the raw_data.
That was okay until we added PERF_SAMPLE_BRANCH_STACK which depends on
the array variable pointing to the beginning of the branch stack data.
But that was not the case if branch stack was combined with raw mode
sampling. That led to bogus branch stack addresses and count.
The bug would show up with:
$ perf record -R -b foo
This patch fixes the problem by correctly moving the array pointer
forward for RAW samples.
Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20120317222317.GA8803@quad
[ committer note: Fix also later submitted by Jiri Olsa ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The perf diff command is broken since:
perf hists: Threaded addition and sorting of entries
commit 1980c2ebd7
Several places were broken:
- hists data need to be collected into opened sessions instead
of into events
- session's hists data need to be initialized properly when the
session is created
- hist_entry__pcnt_snprintf: the percentage and displacement
buffer preparation must not use 'ret' because it's used
as a pointer to the final buffer
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20120322133726.GB1601@m.brq.redhat.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When event group is enabled for forked task (i.e. no target task/cpu
was specified) all events were disabled and marked ->enable_on_exec.
However they wouldn't be counted at all since only group leader will
be enabled on exec actually.
In contrast to perf stat, perf record doesn't have a real problem
as it enables all the event before proceeding. But it needs to be
fixed anyway IMHO.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1331887340-32448-2-git-send-email-namhyung.kim@lge.com
Signed-off-by: Namhyung Kim <namhyung.kim@lge.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
So that we can get the perf bench exec stack fixes and then apply the
remaining fix for the files added after what is in perf/urgent.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This patch fixes an issue where perf report shows nan% for certain
perf.data files. The below is from a report for a do_fork probe:
-nan% sshd [kernel.kallsyms] [k] do_fork
-nan% packagekitd [kernel.kallsyms] [k] do_fork
-nan% dbus-daemon [kernel.kallsyms] [k] do_fork
-nan% bash [kernel.kallsyms] [k] do_fork
A git bisect shows commit f3bda2c as the cause. However, looking back
through the git history, I saw commit 640c03c which seems to have
removed the required initialization for perf_sample->period. The problem
only started showing after commit f3bda2c. The below patch re-introduces
the initialization and it fixes the problem for me.
With the below patch, for the same perf.data:
73.08% bash [kernel.kallsyms] [k] do_fork
8.97% 11-dhclient [kernel.kallsyms] [k] do_fork
6.41% sshd [kernel.kallsyms] [k] do_fork
3.85% 20-chrony [kernel.kallsyms] [k] do_fork
2.56% sendmail [kernel.kallsyms] [k] do_fork
This patch applies over current linux-tip commit 9949284.
Problem introduced in:
$ git describe 640c03c
v2.6.37-rc3-83-g640c03c
Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: stable@kernel.org
Link: http://lkml.kernel.org/r/20120203170113.5190.25558.stgit@localhost6.localdomain6
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The problem is that when SAMPLE_PERIOD is not set, the kernel generates
a number of samples in proportion to an event's period. Number of these
samples may be too big and the kernel throttles all samples above a
defined limit.
E.g.: I want to trace when a process sleeps. I created a process which
sleeps for 1ms and for 4ms. perf got 100 events in both cases.
swapper 0 [000] 1141.371830: sched_stat_sleep: comm=foo pid=1801 delay=1386750 [ns]
swapper 0 [000] 1141.369444: sched_stat_sleep: comm=foo pid=1801 delay=4499585 [ns]
In the first case a kernel want to send 4499585 events and in the second
case it wants to send 1386750 events. perf-reports shows that process
sleeps in both places equal time.
Instead of this we can get only one sample with an attribute period. As
result we have less data transferring between kernel and user-space and
we avoid throttling of samples.
The patch "events: Don't divide events if it has field period" added a
kernel part of this functionality.
Acked-by: Arun Sharma <asharma@fb.com>
Cc: Arun Sharma <asharma@fb.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: devel@openvz.org
Link: http://lkml.kernel.org/r/1324391565-1369947-1-git-send-email-avagin@openvz.org
Signed-off-by: Andrew Vagin <avagin@openvz.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Recently we made perf_evsel__init call hists__init, which broke the perf
python binding:
[root@emilia linux]# ./tools/perf/python/twatch.py
Traceback (most recent call last):
File "./tools/perf/python/twatch.py", line 16, in <module>
import perf
ImportError: /home/acme/git/build/perf/python/perf.so: undefined symbol: hists__init
Fix it by moving the hists__init function to its only caller, evsel.c.
This way we avoid dragging in other parts of tools/perf/util/ to the
perf python binding.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-5nffmdt5mu6ozxgj54oi4qon@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The __perf_evsel__open routing was grouping just the threads for that
specific events per cpu when we want to group all threads in all events
to the first fd opened on that cpu.
So pass the xyarray with the first event, where the other events will be
able to get that first per cpu fd.
At some point top and record will switch to using perf_evlist__open that
takes care of this detail and probably will also handle the fallback
from hw to soft counters, etc.
Reported-by: Deng-Cheng Zhu <dczhu@mips.com>
Tested-by: Deng-Cheng Zhu <dczhu@mips.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-ebm34rh098i9y9v4cytfdp0x@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
By using a mutex just for inserting and rotating two hist_entry rb
trees, so that when sorting we can get the last batch of entries created
from the ring buffer, merge it with whatever we have processed so far
and show the output while new entries are being added.
The 'report' tool continues, for now, to do it without threading, but
will use this in the future to allow visualization of results in long
perf.data sessions while the entries are being processed.
The new 'top' tool will be the first user.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-9b05atsn0q6m7fqgrug8fk2i@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
