perf does not properly handle monitoring of processes with named threads.
For example:
$ ps -C myapp -L
PID LWP TTY TIME CMD
25118 25118 ? 00:00:00 myapp
25118 25119 ? 00:00:00 myapp:worker
perf record -e cs -c 1 -fo /tmp/perf.data -p 25118 -- sleep 10
perf report --stdio -i /tmp/perf.data
100.00% myapp:worker [kernel.kallsyms] [k] perf_event_task_sched_out
The process name is set to the name of the last thread it finds for the
process.
The Problem:
perf-top and perf-record both create a thread_map of threads to be
monitored. That map is used in perf_event__synthesize_thread_map which
loops over the entries in thread_map and calls __event__synthesize_thread
to generate COMM and MMAP events.
__event__synthesize_thread calls perf_event__synthesize_comm which opens
/proc/pid/status, reads the name of the task and its thread group id.
That's all fine. The problem is that it then reads /proc/pid/task and
generates COMM events for each task it finds - but using the name found
in /proc/pid/status where pid is the thread of interest.
The end result (looping over thread_map + synthesizing comm events for
each thread each time) means the name of the last thread processed sets
the name for all threads in the process - which is not good for
multithreaded processes with named threads.
The Fix:
perf_event__synthesize_comm has an input argument (full) that decides
whether to process task entries for each pid it is passed. It currently
never set to 0 (perf_event__synthesize_comm has a single caller and it
always passes the value 1). Let's fix that.
Add the full input argument to __event__synthesize_thread which passes
it to perf_event__synthesize_comm. For thread/process monitoring set full
to 0 which means COMM and MMAP events are only generated for the pid
passed to it. For system wide monitoring set full to 1 so that COMM events
are generated for all threads in a process.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1324578603-12762-2-git-send-email-dsahern@gmail.com
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Use local variable 'dso' to reduce typing a bit and rearrange the if
condition. Also NULL check of al->map in the condition is not necessary.
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/1323703017-6060-7-git-send-email-namhyung@gmail.com
Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
So that tools like 'perf test' can print the events when in verbose
mode, for instance.
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-xnovdqfi25nc48gy6604k7yp@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To better reflect that it became the base class for all tools, that must
be in each tool struct and where common stuff will be put.
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-qgpc4msetqlwr8y2k7537cxe@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Reducing the exposure of perf_session further, so that we can use the
classes in cases where no perf.data file is created.
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-stua66dcscsezzrcdugvbmvd@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
So that we don't need to have that many globals.
Next steps will remove the 'session' pointer, that in most cases is
not needed.
Then we can rename perf_event_ops to 'perf_tool' that better describes
this class hierarchy.
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-wp4djox7x6w1i2bab1pt4xxp@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf_event__synthesize_mmap_events does not create anonymous mmap events
even though the kernel does. As a result an already running application
with dynamically created code will not get profiled - all samples end up
in the unknown bucket.
This patch skips any entries with '[' in the name to avoid adding events
for special regions (eg the vsyscall page). All other executable mmaps
are assumed to be anonymous and an event is synthesized.
Acked-by: Pekka Enberg <penberg@kernel.org>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Link: http://lkml.kernel.org/r/20110830091506.60b51fe8@kryten
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fixes two more cases where the python binding would not load:
. Not finding die(), which it shouldn't anyway, not good to just stop the
world because some particular perf.data file is invalid, just propagate
the error to the caller.
. Not finding perf_sample_size: fix it by moving it from event.c to evsel,
where it belongs, as most cases are moving to operate on an evsel object.o
One of the fixed problems:
[root@emilia ~]# python
>>> import perf
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ImportError: /home/acme/git/build/perf/python/perf.so: undefined symbol: perf_sample_size
>>>
[root@emilia ~]#
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-1hkj7b2cvgbfnoizsekjb6c9@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Perf uses /proc/modules to figure out where kernel modules are loaded.
With the advent of kptr_restrict, non root users get zeroes for all module
start addresses.
So check if kptr_restrict is non zero and don't generate the syntethic
PERF_RECORD_MMAP events for them.
Warn the user about it in perf record and in perf report.
In perf report the reference relocation symbol being zero means that
kptr_restrict was set, thus /proc/kallsyms has only zeroed addresses, so don't
use it to fixup symbol addresses when using a valid kallsyms (in the buildid
cache) or vmlinux (in the vmlinux path) build-id located automatically or
specified by the user.
Provide an explanation about it in 'perf report' if kernel samples were taken,
checking if a suitable vmlinux or kallsyms was found/specified.
Restricted /proc/kallsyms don't go to the buildid cache anymore.
Example:
[acme@emilia ~]$ perf record -F 100000 sleep 1
WARNING: Kernel address maps (/proc/{kallsyms,modules}) are restricted, check
/proc/sys/kernel/kptr_restrict.
Samples in kernel functions may not be resolved if a suitable vmlinux file is
not found in the buildid cache or in the vmlinux path.
Samples in kernel modules won't be resolved at all.
If some relocation was applied (e.g. kexec) symbols may be misresolved even
with a suitable vmlinux or kallsyms file.
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.005 MB perf.data (~231 samples) ]
[acme@emilia ~]$
[acme@emilia ~]$ perf report --stdio
Kernel address maps (/proc/{kallsyms,modules}) were restricted,
check /proc/sys/kernel/kptr_restrict before running 'perf record'.
If some relocation was applied (e.g. kexec) symbols may be misresolved.
Samples in kernel modules can't be resolved as well.
# Events: 13 cycles
#
# Overhead Command Shared Object Symbol
# ........ ....... ................. .....................
#
20.24% sleep [kernel.kallsyms] [k] page_fault
20.04% sleep [kernel.kallsyms] [k] filemap_fault
19.78% sleep [kernel.kallsyms] [k] __lru_cache_add
19.69% sleep ld-2.12.so [.] memcpy
14.71% sleep [kernel.kallsyms] [k] dput
4.70% sleep [kernel.kallsyms] [k] flush_signal_handlers
0.73% sleep [kernel.kallsyms] [k] perf_event_comm
0.11% sleep [kernel.kallsyms] [k] native_write_msr_safe
#
# (For a higher level overview, try: perf report --sort comm,dso)
#
[acme@emilia ~]$
This is because it found a suitable vmlinux (build-id checked) in
/lib/modules/2.6.39-rc7+/build/vmlinux (use -v in perf report to see the long
file name).
If we remove that file from the vmlinux path:
[root@emilia ~]# mv /lib/modules/2.6.39-rc7+/build/vmlinux \
/lib/modules/2.6.39-rc7+/build/vmlinux.OFF
[acme@emilia ~]$ perf report --stdio
[kernel.kallsyms] with build id 57298cdbe0131f6871667ec0eaab4804dcf6f562
not found, continuing without symbols
Kernel address maps (/proc/{kallsyms,modules}) were restricted, check
/proc/sys/kernel/kptr_restrict before running 'perf record'.
As no suitable kallsyms nor vmlinux was found, kernel samples can't be
resolved.
Samples in kernel modules can't be resolved as well.
# Events: 13 cycles
#
# Overhead Command Shared Object Symbol
# ........ ....... ................. ......
#
80.31% sleep [kernel.kallsyms] [k] 0xffffffff8103425a
19.69% sleep ld-2.12.so [.] memcpy
#
# (For a higher level overview, try: perf report --sort comm,dso)
#
[acme@emilia ~]$
Reported-by: Stephane Eranian <eranian@google.com>
Suggested-by: David Miller <davem@davemloft.net>
Cc: Dave Jones <davej@redhat.com>
Cc: David Miller <davem@davemloft.net>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Kees Cook <kees.cook@canonical.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Link: http://lkml.kernel.org/n/tip-mt512joaxxbhhp1odop04yit@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The shift used here to count the number of bits set in
the mask doesn't work above the low part for archs that
are not 64 bits.
Fix the constant used for the shift.
This fixes a 32-bit perf top failure reported by Eric Dumazet:
Can't parse sample, err = -14
Can't parse sample, err = -14
...
Reported-and-tested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com
Link: http://lkml.kernel.org/r/1306200686-17317-1-git-send-email-fweisbec@gmail.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
What we want is to count the number of bits in the mask,
not some other random operation written in the middle
of the night.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1306148788-6179-2-git-send-email-fweisbec@gmail.com
[ Fixed perf_event__names[] alignment which was nearby and hurting my eyes ... ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Check that the total size of the sample fields having a fixed
size do not exceed the one of the whole event. This robustifies
the sample parsing.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
Perf can't currently trace into the vsyscall page. It looks like it was
meant to work.
Tested on 2.6.38 and today's -git.
The bug is easy to reproduce. Compile this:
int main()
{
int i;
struct timespec t;
for(i = 0; i < 10000000; i++)
clock_gettime(CLOCK_MONOTONIC, &t);
return 0;
}
and run it through perf record; perf report. The top entry shows
"[unknown]" and you can't zoom in.
It looks like there are two issues. The first is a that a test for user
mode executing in kernel space is backwards. (That's the first hunk
below). The second (I think) is that something's wrong with the code
that generates lots of little struct dso objects for different sections
-- when it runs on vmlinux it results in bogus long_name values which
cause objdump to fail.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LPU-Reference: <AANLkTikxSw5+wJZUWNz++nL7mgivCh_Zf=2Kq6=f9Ce_@mail.gmail.com>
Signed-off-by: Andy Lutomirski <luto@mit.edu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To support multiple events we need to do these calcs per 'struct hists'
instance, and it turns out we already do that at:
__hists__add_entry
hists__inc_nr_entries
hists__calc_col_len
for all the unfiltered hist_entry instances we stash in the rb tree, so
trow away the dead code.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fixups due to rename of event_t routines from event__ to perf_event__
done in perf/core.
Conflicts:
tools/perf/builtin-record.c
tools/perf/builtin-top.c
tools/perf/util/event.c
tools/perf/util/event.h
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Jeff Moyer reported these messages:
Warning: ... trying to fall back to cpu-clock-ticks
couldn't open /proc/-1/status
couldn't open /proc/-1/maps
[ls output]
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.008 MB perf.data (~363 samples) ]
That lead me and David Ahern to see that something was fishy on the thread
synthesizing routines, at least for the case where the workload is started
from 'perf record', as -1 is the default for target_tid in 'perf record --tid'
parameter, so somehow we were trying to synthesize the PERF_RECORD_MMAP and
PERF_RECORD_COMM events for the thread -1, a bug.
So I investigated this and noticed that when we introduced support for
recording a process and its threads using --pid some bugs were introduced and
that the way to fix it was to instead of passing the target_tid to the event
synthesizing routines we should better pass the thread_map that has the list of
threads for a --pid or just the single thread for a --tid.
Checked in the following ways:
On a 8-way machine run cyclictest:
[root@emilia ~]# perf record cyclictest -a -t -n -p99 -i100 -d50
policy: fifo: loadavg: 0.00 0.13 0.31 2/139 28798
T: 0 (28791) P:99 I:100 C: 25072 Min: 4 Act: 5 Avg: 6 Max: 122
T: 1 (28792) P:98 I:150 C: 16715 Min: 4 Act: 6 Avg: 5 Max: 27
T: 2 (28793) P:97 I:200 C: 12534 Min: 4 Act: 5 Avg: 4 Max: 8
T: 3 (28794) P:96 I:250 C: 10028 Min: 4 Act: 5 Avg: 5 Max: 96
T: 4 (28795) P:95 I:300 C: 8357 Min: 5 Act: 6 Avg: 5 Max: 12
T: 5 (28796) P:94 I:350 C: 7163 Min: 5 Act: 6 Avg: 5 Max: 12
T: 6 (28797) P:93 I:400 C: 6267 Min: 4 Act: 5 Avg: 5 Max: 9
T: 7 (28798) P:92 I:450 C: 5571 Min: 4 Act: 5 Avg: 5 Max: 9
^C[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.108 MB perf.data (~4719 samples) ]
[root@emilia ~]#
This will create one extra thread per CPU:
[root@emilia ~]# tuna -t cyclictest -CP
thread ctxt_switches
pid SCHED_ rtpri affinity voluntary nonvoluntary cmd
28825 OTHER 0 0xff 2169 671 cyclictest
28832 FIFO 93 6 52338 1 cyclictest
28833 FIFO 92 7 46524 1 cyclictest
28826 FIFO 99 0 209360 1 cyclictest
28827 FIFO 98 1 139577 1 cyclictest
28828 FIFO 97 2 104686 0 cyclictest
28829 FIFO 96 3 83751 1 cyclictest
28830 FIFO 95 4 69794 1 cyclictest
28831 FIFO 94 5 59825 1 cyclictest
[root@emilia ~]#
So we should expect only samples for the above 9 threads when using the
--dump-raw-trace|-D perf report switch to look at the column with the tid:
[root@emilia ~]# perf report -D | grep RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort | uniq -c
629 28825
110 28826
491 28827
308 28828
198 28829
621 28830
225 28831
203 28832
89 28833
[root@emilia ~]#
So for workloads started by 'perf record' seems to work, now for existing workloads,
just run cyclictest first, without 'perf record':
[root@emilia ~]# tuna -t cyclictest -CP
thread ctxt_switches
pid SCHED_ rtpri affinity voluntary nonvoluntary cmd
28859 OTHER 0 0xff 594 200 cyclictest
28864 FIFO 95 4 16587 1 cyclictest
28865 FIFO 94 5 14219 1 cyclictest
28866 FIFO 93 6 12443 0 cyclictest
28867 FIFO 92 7 11062 1 cyclictest
28860 FIFO 99 0 49779 1 cyclictest
28861 FIFO 98 1 33190 1 cyclictest
28862 FIFO 97 2 24895 1 cyclictest
28863 FIFO 96 3 19918 1 cyclictest
[root@emilia ~]#
and then later did:
[root@emilia ~]# perf record --pid 28859 sleep 3
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.027 MB perf.data (~1195 samples) ]
[root@emilia ~]#
To collect 3 seconds worth of samples for pid 28859 and its children:
[root@emilia ~]# perf report -D | grep RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort | uniq -c
15 28859
33 28860
19 28861
13 28862
13 28863
10 28864
11 28865
9 28866
255 28867
[root@emilia ~]#
Works, last thing is to check if looking at just one of those threads also works:
[root@emilia ~]# perf record --tid 28866 sleep 3
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.006 MB perf.data (~242 samples) ]
[root@emilia ~]# perf report -D | grep RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort | uniq -c
3 28866
[root@emilia ~]#
Works too.
Reported-by: Jeff Moyer <jmoyer@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jeff Moyer <jmoyer@redhat.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>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
And move the event_t methods to the perf_event__ too.
No code changes, just namespace consistency.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Right now this function is only used by perf top, that uses PROT_READ
only, i.e. overwrite mode, so no PERF_RECORD_LOST events are generated,
but don't forget those events.
The patch that moved this out of perf top was made so that this routine
could be used by 'perf probe' in the uprobes patchset, so perhaps there
they need to check for LOST events and warn the user, as will be done in
the following patches that will switch 'perf top' to non overwrite mode
(mmap with PROT_READ|PROT_WRITE).
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To avoid linking more stuff in the python binding I'm working on, future
csets will make the sample type be taken from the evsel itself, but for
that we need to first have one file per cpu and per sample_type, not a
single perf.data file.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Using %L[uxd] has issues in some architectures, like on ppc64. Fix it
by making our 64 bit integers typedefs of stdint.h types and using
PRI[ux]64 like, for instance, git does.
Reported by Denis Kirjanov that provided a patch for one case, I went
and changed all cases.
Reported-by: Denis Kirjanov <dkirjanov@kernel.org>
Tested-by: Denis Kirjanov <dkirjanov@kernel.org>
LKML-Reference: <20110120093246.GA8031@hera.kernel.org>
Cc: Denis Kirjanov <dkirjanov@kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Pingtian Han <phan@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
For kallsyms we don't have the symbol address end, so we do an extra pass and
set the symbol end addr as being the start of the next minus one.
But this was being done just after we filtered the symbols of a
particular type (functions, variables), so the symbol end was sometimes
after what it really is.
Fixing up symbol end also was falling apart when we have symbol aliases,
then the end address of all but the last alias was being set to be
before its start.
Fix it up by checking for symbol aliases and making the kallsyms__parse
routine use the next symbol, whatever its type, as the limit for the
previous symbol, passing that end address to the callback.
This was detected by the 'perf test' synthetic paranoid regression
tests, fix it up so that even that case doesn't mislead us.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
event__name[] is missing an entry for PERF_RECORD_FINISHED_ROUND, but we
happily access the array from the dump code.
Make event__name[] static and provide an accessor function, fix up all
callers and add the missing string.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <20101207124550.432593943@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
If a 32bit userspace perf is running on a 64bit kernel, the end of the final
map in the kernel would incorrectly be set to 2^32-1 rather than 2^64-1.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1290658375-10342-1-git-send-email-imunsie@au1.ibm.com>
Signed-off-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The event__process function is useful in processing /proc/<pid>/maps. All of
the functions that are called from event__process are defined in util/event.c.
Though its defined in builtin-top.c, it could be reused for perf probe for
uprobes. Hence moving it to util/event.c and exporting the function.
LKML-Reference: <20100802123851.GD22812@linux.vnet.ibm.com>
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fix buggy-looking code which unnecessarily adjusts the file offset
fields read from /proc/*/maps.
This may have gone unnoticed since the offset is usually 0 (and the
logic in util/symbol.c may work incorrectly for other offset values).
Commiter note:
This fixes a bug introduced in 4af8b35, there is no need to shift pgoff
twice, the show_map_vma routine in fs/proc/task_mmu.c already converts
it from the number of pages to the size in bytes, and that is what
appears in /proc/PID/map.
Cc: Nicolas Pitre <nicolas.pitre@linaro.org>
Cc: Will Deacon <Will.Deacon@arm.com>
LKML-Reference: <1280836116-6654-2-git-send-email-dave.martin@linaro.org>
Signed-off-by: Dave Martin <dave.martin@linaro.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
So that we reduce the noise when looking for leaks using tools such as
valgrind.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This avoids around some problems where the full path is executables and DSOs it
needed for finding debug symbols on platforms with separated debug symbol files
such as Ubuntu. This is simpler than tracking an extra name for each image.
The only impact should be that paths in verbose output from the perf tools
become absolute, instead of relative to .
LKML-Reference: <new-submission>
Signed-off-by: Dave Martin <dave.martin@linaro.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
They were globals, and since we support multiple hists and sessions
at the same time, it doesn't make sense to calculate those values
considereing all symbols in all sessions.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Move them to a session->dead_threads list just like we do with maps that
are replaced, because we may have hist_entries pointing to them.
This fixes a bug when inserting maps for a new thread that reused the
TID, mixing maps for two different threads, causing an endless loop.
The code for insering maps should be made more robust but for .35 this
is the minimalistic patch.
Reported-by: Ingo Molnar <mingo@elte.hu>
Cc: David S. Miller <davem@davemloft.net>
Cc: Frédéric 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>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
In a shared multi-core environment, users want to analyze why their
program was slow. In particular, if the code ran slower only on certain
CPUs due to interference from other programs or kernel threads, the user
should be able to notice that.
Sample usage:
perf record -f -a -- sleep 3
perf report --sort cpu,comm
Workload:
program is running on 16 CPUs
Experiencing interference from an antagonist only on 4 CPUs.
Samples: 106218177676 cycles
Overhead CPU Command
........ ... ...............
6.25% 2 program
6.24% 6 program
6.24% 11 program
6.24% 5 program
6.24% 9 program
6.24% 10 program
6.23% 15 program
6.23% 7 program
6.23% 3 program
6.23% 14 program
6.22% 1 program
6.20% 13 program
3.17% 12 program
3.15% 8 program
3.14% 0 program
3.13% 4 program
3.11% 4 antagonist
3.11% 0 antagonist
3.10% 8 antagonist
3.07% 12 antagonist
Cc: David S. Miller <davem@davemloft.net>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <20100505181612.GA5091@sharma-home.net>
Signed-off-by: Arun Sharma <aruns@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Simplifying the tools that were using both in sequence and allowing
upcoming simplifications, such as Arun's patch to sort by cpus.
Cc: David S. Miller <davem@davemloft.net>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
It is useful to know on which CPU a sample was captured on.
The information is captured with perf record -R but it was
not printed out by perf report -D. This patch adds this.
When -R is not used, cpu is set to -1to indicate that
the CPU is unknown (it is not captured).
Cc: David S. Miller <davem@davemloft.net>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <4bff964c.e88cd80a.3106.7d31@mx.google.com>
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
event__process_task() doesn't propagate the comm copy on clone,
but only on process fork. So we loose all the tid:comm resolution
for tasks that aren't a main process thread.
Progragate the per thread granularity to event__process_task for
pid resolution.
This fixes various unresolved pids in perf sched, especially when
we trace multithread processes. The problem is quickly reproducible
with the messaging benchmark using the multithread mode "-t" :
perf sched record perf bench sched messaging -t
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Stephane Eranian <eranian@google.com>
When we synthetize the existing running tasks though procfs,
we walk through every threads of a process, queuing one comm
events per tid.
But then on report time, event__process_comm() only creates and
sets the comm on a per process granularity. This is the right
thing for comm events that came from the kernel, as they are
only created on exec. Sub-threads then inherit their comm
from fork events. But that doesn't work with our synthetized
comm events taken from procfs informations as the per thread
granularity is done on comm events directly there.
Hence we need event__process_comm() to work with the tid rather
than the pid. It won't change anything for comm events coming
from the kernel but this will fix the synthetized ones.
Before:
$ ./perf report -D | grep COMM | grep firefox
0x2c7b8 [0x18]: PERF_RECORD_COMM: firefox:5297
0x2c7d0 [0x18]: PERF_RECORD_COMM: firefox:5297
0x2c7e8 [0x18]: PERF_RECORD_COMM: firefox:5297
0x2c800 [0x18]: PERF_RECORD_COMM: firefox:5297
0x2c818 [0x18]: PERF_RECORD_COMM: firefox:5297
0x2c830 [0x18]: PERF_RECORD_COMM: firefox:5297
After:
$ ./perf report -D | grep COMM | grep firefox
0x2c7b8 [0x18]: PERF_RECORD_COMM: firefox:5297
0x2c7d0 [0x18]: PERF_RECORD_COMM: firefox:5299
0x2c7e8 [0x18]: PERF_RECORD_COMM: firefox:5300
0x2c800 [0x18]: PERF_RECORD_COMM: firefox:5308
0x2c818 [0x18]: PERF_RECORD_COMM: firefox:5309
0x2c830 [0x18]: PERF_RECORD_COMM: firefox:5312
This fixes various unresolved pid on perf sched.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Stephane Eranian <eranian@google.com>
The events_stats.total field is too generic, rename it to .total_period,
and also add a comment explaining that it is the sum of all the .period
fields in samples, that is needed because we use auto-freq to avoid
sampling artifacts.
Ditto for events_stats.lost, that is the sum of all lost_event.lost
fields, i.e. the number of events the kernel dropped.
Looking at the users, builtin-sched.c can make use of these fields and
stop doing it again.
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This is one more thing that started global but are more useful per hist
or per session.
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
In cbbc79a we introduced support for multiple events by introducing a
new "event_stat_id" struct and then made several perf_session methods
receive a point to it instead of a pointer to perf_session, and kept the
event_stats and hists rb_tree in perf_session.
While working on the new newt based browser, I realised that it would be
better to introduce a new class, "hists" (short for "histograms"),
renaming the "event_stat_id" struct and the perf_session methods that
were really "hists" methods, as they manipulate only struct hists
members, not touching anything in the other perf_session members.
Other optimizations, such as calculating the maximum lenght of a symbol
name present in an hists instance will be possible as we add them,
avoiding a re-traversal just for finding that information.
The rationale for the name "hists" to replace "event_stat_id" is that we
may have multiple sets of hists for the same event_stat id, as, for
instance, the 'perf diff' tool has, so event stat id is not what
characterizes what this struct and the functions that manipulate it do.
Cc: Eric B Munson <ebmunson@us.ibm.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
We have just one host on a given session, and that is the most common
setup right now, so embed a ->host_machine struct machine instance
directly in the perf_session class, check if we're looking for it before
going to the rb_tree.
This also fixes a problem found when we try to process old perf.data
files where we didn't have MMAP events for the kernel and modules and
thus don't create the kernel maps, do it in event__preprocess_sample if
it wasn't already.
Reported-by: Ingo Molnar <mingo@elte.hu>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
Cc: Zhang, Yanmin <yanmin_zhang@linux.intel.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Which can happen when processing old files that had no fake kernel MMAP,
events.
That shouldn't result in perf_session__create_kernel_maps not being
called, this will be fixed in a followup patch, for now do these checks
to avoid segfaulting.
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
By using BITS_PER_LONG / 4, that is the number of chars that will be
used in such cases as the DSO "name".
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The first was always using the ->long_name, while the later used
->short_name if verbose was not set, resulting in the dso column to be
much wider than needed most of the time.
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
On a large machine we spend a lot of time in perf_header__find_attr when
running perf report.
If we are parsing a file without PERF_SAMPLE_ID then for each sample we call
perf_header__find_attr and loop through all counter IDs, never finding a match.
As the machine gets larger there are more per cpu counters and we spend an
awful lot of time in there.
The patch below initialises each sample id to -1ULL and checks for this in
perf_header__find_attr. We may need to do something more intelligent eventually
(eg a hash lookup from counter id to attr) but this at least fixes the most
common usage of perf report.
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Eric B Munson <ebmunson@us.ibm.com>
Acked-by: Eric B Munson <ebmunson@us.ibm.com>
LKML-Reference: <20100504111915.GB14636@kryten>
Signed-off-by: Anton Blanchard <anton@samba.org>
--
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Those functions operated on members now grouped in 'struct machine', so
move those methods to this new class.
The changes made to 'perf probe' shows that using this abstraction
inserting probes on guests almost got supported for free.
Cc: Avi Kivity <avi@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Zhang, Yanmin <yanmin_zhang@linux.intel.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Don't blindly assume that the size of the buffer is enough, use
snprintf.
Cc: Avi Kivity <avi@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>
Cc: Zhang, Yanmin <yanmin_zhang@linux.intel.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
struct kernel_info and kerninfo__ are too vague, what they really
describe are machines, virtual ones or hosts.
There are more changes to introduce helpers to shorten function calls
and to make more clear what is really being done, but I left that for
subsequent patches.
Cc: Avi Kivity <avi@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>
Cc: Zhang, Yanmin <yanmin_zhang@linux.intel.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
