Stop using this python/OOP convention, doesn't really helps. Will do
more from time to time till we get it cleaned up in all of /perf.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
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-91i56jwnzq9edhsj9y2y9l3b@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The stack/vdso/heap memory maps dont have any dso file. Setting the
perf dso objects as 'loaded' for these maps, we avoid unnecessary
warnings like:
"Failed to open [stack], continuing without symbols"
All map__find_* functions still return NULL when searching for symbols
in these maps.
Link: http://lkml.kernel.org/r/20110824131834.GA2007@jolsa.brq.redhat.com
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
In the past we tried to avoid printing the name of the event when just
one event was found in the perf.data file, after some refactorings it
ended up not printing the event name if just one hist_entry was found in
one of the events.
Fix it by always printing the name of the event, even if just one is
found.
Reported-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/n/tip-kikr0c7ou55bd9caok8569rf@git.kernel.org
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>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Add -M option to report/annotate to pass directly to objdump. This
allows to use -M intel for intel style disassembler syntax, which is
useful for people who are very used to the Intel syntax.
Link: http://lkml.kernel.org/r/1316122302-24306-2-git-send-email-andi@firstfloor.org
[committer note: Add missing Documentation bits, fixup conflicts with 3e6a2a7]
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Stephane Eranian <eranian@google.com>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When a program crashes under perf there is no message about it, unlike
when running it from bash. This can be confusing and lead to wrong
actions during debugging.
Print fatal signals in perf stat/record.
Thanks to Furat Afram for finding the problem originally
Link: http://lkml.kernel.org/r/1316122302-24306-1-git-send-email-andi@firstfloor.org
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Stephane Eranian <eranian@google.com>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
If option -x '\t' is given, convert '\t' to "\t". This makes cvs
printing more flexible.
Link: http://lkml.kernel.org/r/1315437244-3788-5-git-send-email-jim.cromie@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
For pretty output only (preserve column for cvs output), dont print
std-deviation when its 0.00. Do this based upon value, instead of
checking for --no-aggr, since the stats could conceivably be computed
over the runs on each CPU, and theres no reason to preclude that.
Link: http://lkml.kernel.org/r/1315437244-3788-4-git-send-email-jim.cromie@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Without this patch, running:
$ sudo ./perf stat -r20 --no-aggr -a perl -e '$i++ for 1..100000'
I get computations like this:
CPU0 12.488247 task-clock # 1.224 CPUs utilized ( +- -nan% )
CPU1 12.488909 task-clock # 1.225 CPUs utilized ( +- -nan% )
CPU2 12.500221 task-clock # 1.226 CPUs utilized ( +- -nan% )
CPU3 12.481713 task-clock # 1.224 CPUs utilized ( +- -nan% )
but with patch, I get:
CPU0 8.233682 task-clock # 0.754 CPUs utilized ( +- 0.00% )
CPU1 8.226318 task-clock # 0.754 CPUs utilized ( +- 0.00% )
CPU2 8.210737 task-clock # 0.752 CPUs utilized ( +- 0.00% )
CPU3 8.201691 task-clock # 0.751 CPUs utilized ( +- 0.00% )
Note that without --no-aggr, I get non-0 statistics both before and after patch:
231.986022 task-clock # 4.030 CPUs utilized ( +- 0.97% )
212 context-switches # 0.001 M/sec ( +- 12.07% )
9 CPU-migrations # 0.000 M/sec ( +- 25.80% )
466 page-faults # 0.002 M/sec ( +- 3.23% )
174,318,593 cycles # 0.751 GHz ( +- 1.06% )
I couldnt see anything wrong in the caller, so fixed it in
stddev_stats(). ISTM that 0.00 is better than nan, since perf stat was
passed -A (--no-aggr) so no standard deviation should be expected, and
nan is suggestive of a deeper error.
When running with --no-aggr, perhaps we should suppress the statistics
printing entirely, or do so when they are 0.00.
Link: http://lkml.kernel.org/r/1315437244-3788-3-git-send-email-jim.cromie@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This perf stat option emulates valgrind's --log-fd option, allowing the
user to send perf results elsewhere, and leaving stderr for use by the
program under test. This complements --output file option, and is
mutually exclusive with it.
3>results perf stat --log-fd 3 -- $cmd
3>>results perf stat --log-fd 3 --append -- $cmd
The perl distro's make test.valgrind target uses valgrind's --log-fd
option, I've adapted it to invoke perf also, and tested this patch
there.
Link: http://lkml.kernel.org/r/1315437244-3788-2-git-send-email-jim.cromie@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Now it warns everytime that new events are lost.
And the TUI also warns now.
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-w1n168yrvrppnq6887s4u0wx@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fixing an artifact where the last 3 chars of a long DSO name would
remain on the screen sometimes.
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-dkiakcl3z69dh1bt9uegaktv@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Try first reading the build id, validating that it is an ELF file, etc.
Cheap as libelf will bail out as soon as the magic number check fails.
Useful when investigating debuginfo packaging problems like this one:
[root@emilia ~]# perf buildid-list -i /usr/lib/debug/lib/modules/`uname -r`/vmlinux
77bb4ea591a602d455ace759a377c9adfff1aba3
[root@emilia ~]# perf buildid-list -k
07b0c016a2b30004e86132d0239945b1e88f5d75
[root@emilia ~]#
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-4elot9oxwa0rr0d90dshca3a@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
A while back I created the dropmonitor protocol, which allowed users to get
reports of dropped frames communicated to them via a netlink socket.
While useful, several people have now asked that I integrate the ability
to do drop monitoring with perf, so they don't have to run additional
tools.
This patch adds a drop monitor script to the perf suite, and provides
the same output that the netlink socket does.
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/1309801217-22450-1-git-send-email-nhorman@tuxdriver.com
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Stop using this python/OOP convention, doesn't really helps. Will do
more from time to time till we get it cleaned up in all of /perf.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
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-rl9e690y60vnuyng05yp1zd3@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Wrong pointer is being passed for raw data sanity checking, when parsing
sample event.
This ends up with invalid event and perf record being stuck in
__perf_session__process_events function during processing build IDs
(process_buildids function).
Following command hangs up in my setup:
./perf record -e raw_syscalls:sys_enter ls
The fix is to use proper pointer to the raw data instead of the 'u'
union.
Reviewed-by: David Ahern <dsahern@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1317308709-9474-2-git-send-email-jolsa@redhat.com
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Problem introduced in 936be50, that missed one perf_event__parse_sample
user, the python binding.
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
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-ja4phms9618ggi657plyuch2@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
GCC often introduces new warnings with lots of false positives -
breaking -Werror builds. WERROR=0 allows one to build perf without much
fuss - while still encouraging people to send patches to avoid the fuss
of having to type WERROR=0.
Bisecting back to commits that produce a (mostly harmless) warning on
some compilers is more difficult. With WERROR=0 one could bisect without
worrying about harmless warnings.
Cc: Ingo Molnar <mingo@elte.hu>
Link: http://lkml.kernel.org/r/eac06c7cc4920e5d4830417d466161fb26c7359c.1315514559.git.dvhart@linux.intel.com
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The 'perf top' tool came from the kernel where we had each DSO (vmlinux,
modules) loaded just once at a time.
But userspace may have DSOs loaded in multiple addresses (shared
libraries), requiring that we use the just resolved map instead of the
first one found.
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-ag53wz0yllpgers0n2w7hchp@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Buildid can vary in size. According to the man page of ld, buildid can
be 160 bits (sha1) or 128 bits (md5, uuid). Perf assumes buildid size of
20 bytes (160 bits) regardless. When dealing with md5 buildids, it would
thus read more than needed and that would cause mismatches and samples
without symbols.
This patch fixes this by taking into account the actual buildid size as
encoded int he section header. The leftover bytes are also cleared.
This second version fixes a minor issue with the memset() base position.
Cc: David S. Miller <davem@davemloft.net>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@gmail.com>
Link: http://lkml.kernel.org/r/4cc1af3c.8ee7d80a.5a28.ffff868e@mx.google.com
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Currently, analyzing PPC data files on x86 the cpu field is always 0 and
the tid and pid are backwards. For example, analyzing a PPC file on PPC
the pid/tid fields show:
rsyslogd 1210/1212
and analyzing the same PPC file using an x86 perf binary shows:
rsyslogd 1212/1210
The problem is that the swap_op method for samples is
perf_event__all64_swap which assumes all elements in the sample_data
struct are u64s. cpu, tid and pid are u32s and need to be handled
individually. Given that the swap is done before the sample is parsed,
the simplest solution is to undo the 64-bit swap of those elements when
the sample is parsed and do the proper swap.
The RAW data field is generic and perf cannot have programmatic knowledge
of how to treat that data. Instead a warning is given to the user.
Thanks to Anton Blanchard for providing a data file for a mult-CPU
PPC system so I could verify the fix for the CPU fields.
v3 -> v4:
- fixed use of WARN_ONCE
v2 -> v3:
- used WARN_ONCE for message regarding raw data
- removed struct wrapper around union
- fixed whitespace issues
v1 -> v2:
- added a union for undoing the byte-swap on u64 and redoing swap on
u32's to address compiler errors (see git commit 65014ab3)
Cc: Anton Blanchard <anton@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1315321946-16993-1-git-send-email-dsahern@gmail.com
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
I took a profile that suggested 60% of total CPU time was in the
hypervisor:
...
60.20% [H] 0x33d43c
4.43% [k] ._spin_lock_irqsave
1.07% [k] ._spin_lock
Using perf stat to get the user/kernel/hypervisor breakdown contradicted
this.
The problem is we merge all unresolved samples into the one unknown
bucket. If add a comparison by sample type to sort__sym_cmp we get the
real picture:
...
57.11% [.] 0x80fbf63c
4.43% [k] ._spin_lock_irqsave
1.07% [k] ._spin_lock
0.65% [H] 0x33d43c
So it was almost all userspace, not hypervisor as the initial profile
suggested.
I found another issue while adding this. Symbol sorting sometimes shows
multiple entries for the unknown bucket:
...
16.65% [.] 0x6cd3a8
7.25% [.] 0x422460
5.37% [.] yylex
4.79% [.] malloc
4.78% [.] _int_malloc
4.03% [.] _int_free
3.95% [.] hash_source_code_string
2.82% [.] 0x532908
2.64% [.] 0x36b538
0.94% [H] 0x8000000000e132a4
0.82% [H] 0x800000000000e8b0
This happens because we aren't consistent with our sorting. On
one hand we check to see if both symbols match and for two unresolved
samples sym is NULL so we match:
if (left->ms.sym == right->ms.sym)
return 0;
On the other hand we use sample IP for unresolved samples when
comparing against a symbol:
ip_l = left->ms.sym ? left->ms.sym->start : left->ip;
ip_r = right->ms.sym ? right->ms.sym->start : right->ip;
This means unresolved samples end up spread across the rbtree and we
can't merge them all.
If we use cmp_null all unresolved samples will end up in the one bucket
and the output makes more sense:
...
39.12% [.] 0x36b538
5.37% [.] yylex
4.79% [.] malloc
4.78% [.] _int_malloc
4.03% [.] _int_free
3.95% [.] hash_source_code_string
2.26% [H] 0x800000000000e8b0
Acked-by: Eric B Munson <emunson@mgebm.net>
Cc: Eric B Munson <emunson@mgebm.net>
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>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Link: http://lkml.kernel.org/r/20110831115145.4f598ab2@kryten
Signed-off-by: Anton Blanchard <anton@samba.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>
perf-record currently creates events enabled. When doing a system wide
collection (-a arg) this causes data collection for perf's
initialization activities -- eg., perf_event__synthesize_threads().
For some events (e.g., context switch S/W event or tracepoints like
syscalls) perf's initialization causes a lot of events to be captured
frequently generating "Check IO/CPU overload!" warnings on larger
systems (e.g., 2 socket, quad core, hyperthreading).
perf's initialization phase can be skipped by creating events
disabled and then enabling them once the initialization is done.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1314289075-14706-1-git-send-email-dsahern@gmail.com
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Try and pick the best symbol based on a few heuristics:
- Prefer a non weak symbol over a weak one
- Prefer a global symbol over a non global one
- Prefer a symbol with less underscores (idea taken from kallsyms.c)
- If all else fails, choose the symbol with the longest name
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>
Link: http://lkml.kernel.org/r/20110824065243.161953371@samba.org
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
kallsyms__parse capitalises the symbol type, so every symbol is marked
global. Remove this and fix symbol_type__is_a to handle both local and
global symbols.
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>
Link: http://lkml.kernel.org/r/20110824065243.077125989@samba.org
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
kallsyms__parse assumes that /proc/kallsyms is sorted and sets the end
of the previous symbol to the start of the current one.
Unfortunately module symbols are not sorted, eg:
ffffffffa0081f30 t e1000_clean_rx_irq [e1000e]
ffffffffa00817a0 t e1000_alloc_rx_buffers [e1000e]
Some symbols end up with a negative length and others have a length
larger than they should. This results in confusing perf output.
We already have a function to fixup the end of zero length symbols so
use that instead.
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>
Link: http://lkml.kernel.org/r/20110824065242.969681349@samba.org
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
64bit PowerPC debuginfo files have an empty function descriptor section.
I hit a SEGV when perf tried to use this section for symbol resolution.
To fix this we need to check the section is valid and we can do this by
checking for type SHT_PROGBITS.
Cc: <stable@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Eric B Munson <emunson@mgebm.net>
Link: http://lkml.kernel.org/r/20110824065242.895239970@samba.org
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fix to call convert_variable() if previous call does not fail.
To call convert_variable, it ensures "ret" is 0. However, since
"ret" has the return value of synthesize_perf_probe_arg() which
always returns positive value if it succeeded, perf probe doesn't
call convert_variable(). This will cause a SEGV when we add an
event with arguments.
This has to be fixed as it ensures "ret" is greater than 0
(or not negative).
This regression has been introduced by my previous patch, f182e3e1.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20110820053922.3286.65805.stgit@fedora15
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
* 'fixes' of master.kernel.org:/home/rmk/linux-2.6-arm:
ARM: pm: avoid writing the auxillary control register for ARMv7
ARM: pm: some ARMv7 requires a dsb in resume to ensure correctness
ARM: pm: arm920/926: fix number of registers saved
ARM: pm: CPU specific code should not overwrite r1 (v:p offset)
ARM: 7066/1: proc-v7: disable SCTLR.TE when disabling MMU
ARM: 7065/1: kexec: ensure new kernel is entered in ARM state
ARM: 7003/1: vexpress: Add clock definition for the SP805.
ARM: 7051/1: cpuimx* boards: fix mach-types errors
ARM: 7019/1: Footbridge: select CLKEVT_I8253 for ARCH_NETWINDER
ARM: 7015/1: ARM errata: Possible cache data corruption with hit-under-miss enabled
ARM: 7014/1: cache-l2x0: Fix L2 Cache size calculation.
ARM: 6967/1: ep93xx: ts72xx: fix board model detection
ARM: 6965/1: ep93xx: add model detection for ts-7300 and ts-7400 boards
ARM: cache: detect VIPT aliasing I-cache on ARMv6
ARM: twd: register clockevents device before enabling PPI
ARM: realview: ensure visibility of writes during reset
ARM: perf: make name of arm_pmu_type consistent
ARM: perf: fix prototype of release_pmu
ARM: fix perf build with uclibc toolchains
This patch adds an option (-o) to save the output of perf stat into a
file. You could do this with perf record but not with perf stat.
Instead, you had to fiddle with stderr to save the counts into a
separate file.
The patch also adds the --append option so that results can be
concatenated into a single file across runs. Each run of the tool is
clearly separated by a comment line starting with a hash mark. The -A
option of perf record is already used by perf stat, so we only add a
long option.
$ perf stat -o res.txt date
$ cat res.txt
Performance counter stats for 'date':
0.791306 task-clock # 0.668 CPUs utilized
2 context-switches # 0.003 M/sec
0 CPU-migrations # 0.000 M/sec
197 page-faults # 0.249 M/sec
1878143 cycles # 2.373 GHz
<not supported> stalled-cycles-frontend
<not supported> stalled-cycles-backend
1083367 instructions # 0.58 insns per cycle
193027 branches # 243.935 M/sec
9014 branch-misses # 4.67% of all branches
0.001184746 seconds time elapsed
The option can be combined with -x to make the output file much easier
to parse.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20110815202233.GA18535@quad
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
If you have --symfs in perf report, then you also need it for perf
annotate. This allows off-box assembly level analysis of perf.data
samples.
This patch complements:
commit ec5761eab3
Author: David Ahern <daahern@cisco.com>
Date: Thu Dec 9 13:27:07 2010 -0700
perf symbols: Add symfs option for off-box analysis using specified tree
Acked-by: David Ahern <daahern@cisco.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: David Ahern <daahern@cisco.com>
Link: http://lkml.kernel.org/r/20110729232040.GA21838@quad
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This patch adds two new options to perf annotate:
- --no-asm-raw : Do not display raw instruction encodings
- --no-source : Do not interleave source code with assembly code
We believe those options make the output of annotate more readable.
Systematically displaying source can make it hard to follow code and
especially optimized code.
Raw encodings are not useful in most cases.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20110517153207.GA9834@quad
Signed-off-by: Stephane Eranian <eranian@google.com>
[committer note: Use the 'no-' option inverting logic]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Group event scheduling command line option is missing in perf
record/stat.
Add it to perf record/stat, which is same as in perf top.
Reported-by: Andi Kleen <andi@firstfloor.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1313577727.2754.5.camel@hp6530s
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Upstream glibc commit 295e904 added a definition for __attribute_const__
to cdefs.h. This causes the following error when building perf:
util/include/linux/compiler.h:8:0: error: "__attribute_const__"
redefined [-Werror] /usr/include/sys/cdefs.h:226:0: note: this is the
location of the previous definition
Wrap __attribute_const__ in #ifndef as we do for __always_inline.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20110818113720.GL2227@zod.bos.redhat.com
Signed-off-by: Josh Boyer <jwboyer@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
There was a problem with the parse_events() code not printing the
correct event name when an event was unknown and starting with an 'r'.
The source of the problem was the way raw notation was parsed.
Without the patch:
$ perf stat -e retired_foo
invalid event modifier: 'tired_foo'
With the patch:
$ perf stat -e retired_foo
invalid or unsupported event: 'retired_foo'
This also covers the case where the name of the event was not printed at
all when perf was linked with libpfm4.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20110723021043.GA20178@quad
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When no event is given to perf record, perf top, a default event is
initialized (cycles). However, perf_evlist__add_default() was not
setting the symbolic name for the event. Perf top worked simply because
it was reconstructing the name from the event code. But it should not
have to do this. This patch initializes the evsel->name field properly.
This second version improves the code flow on the non error path.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20110607161936.GA8163@quad
Signed-off-by: Stephane Eranian <eranian@google.com>
[committer note: Use perf_evsel__delete() instead of plain free()]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
libio.h is not provided by uClibc, in order to be able to test the
definition of __UCLIBC__ we need to include stdlib.h, which also
includes stddef.h, providing the definition of 'NULL'.
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
With gcc4.6, some instances of concrete inlined function looks redundant
and broken, because it appears inside of a concrete instance and its
call_file and call_line are same as the original abstruct's decl_file
and decl_line respectively.
e.g.
[ d1aa] subprogram
external (flag) Yes
name (strp) "add_timer"
decl_file (data1) 2 ;here is original
decl_line (data2) 847 ;line and file
prototyped (flag) Yes
inline (data1) inlined (1)
sibling (ref4) [ d1c6]
...
[ 11d84] subprogram
abstract_origin (ref4) [ d1aa] ; concrete instance
low_pc (addr) .text+0x000000000000246f <add_timer>
high_pc (addr) .text+0x000000000000248b <mod_timer_pending>
frame_base (block1) [ 0] call_frame_cfa
sibling (ref4) [ 11dd9]
[ 11d9f] formal_parameter
abstract_origin (ref4) [ d1b9]
location (data4) location list [ 701b]
[ 11da8] inlined_subroutine
abstract_origin (ref4) [ d1aa] ; redundant instance
low_pc (addr) .text+0x000000000000247e <add_timer+0xf>
high_pc (addr) .text+0x0000000000002480 <add_timer+0x11>
call_file (data1) 2 ; call line and file
call_line (data2) 847 ; are same as above
Those redundant instances leads unwilling results;
e.g. find probe points inside of functions even if we specify
a function entry as below;
$ perf probe -V add_timer
Available variables at add_timer
@<add_timer+0>
struct timer_list* timer
@<add_timer+15>
(No matched variables)
So, this filters out those redundant instances based on call-site and
decl-site information.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20110811110317.19900.59525.stgit@fedora15
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
gcc 4.6 generates a concrete out-of-line instance when there is a
function which is implicitly inlined somewhere but also has its own
instance. The concrete out-of-line instance means that it has an
abstract origin of the function which is referred by not only
inlined-subroutines but also a concrete subprogram.
Since current dwarf_func_inline_instances() can find only instances of
inlined-subroutines, this introduces new die_walk_instances() to find
both of subprogram and inlined-subroutines.
e.g. without this,
Available variables at sched_group_rt_period
@<cpu_rt_period_read_uint+9>
struct task_group* tg
perf probe failed to find actual subprogram instance of
sched_group_rt_period().
With this,
Available variables at sched_group_rt_period
@<cpu_rt_period_read_uint+9>
struct task_group* tg
@<sched_group_rt_period+0>
struct task_group* tg
Now it found the sched_group_rt_period() itself.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20110811110311.19900.63997.stgit@fedora15
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fix variable searching logic to search one in inner than local scope or
global(CU) scope. In the other words, skip searching in intermediate
scopes.
e.g., in the following code,
int var1;
void inline infunc(int i)
{
i++; <--- [A]
}
void func(void)
{
int var1, var2;
infunc(var2);
}
At [A], "var1" should point the global variable "var1", however, if user
mis-typed as "var2", variable search should be failed. However, current
logic searches variable infunc() scope, global scope, and then func()
scope. Thus, it can find "var2" variable in func() scope. This may not
be what user expects.
So, it would better not search outer scopes except outermost (compile
unit) scope which contains only global variables, when it failed to find
given variable in local scope.
E.g.
Without this:
$ perf probe -V pre_schedule --externs > without.vars
With this:
$ perf probe -V pre_schedule --externs > with.vars
Check the diff:
$ diff without.vars with.vars
88d87
< int cpu
133d131
< long unsigned int* switch_count
These vars are actually in the scope of schedule(), the caller of
pre_schedule().
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20110811110305.19900.94374.stgit@fedora15
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fix perf probe to search local variables in appropriate local inlined
function scope. For example, pre_schedule() has only 2 local variables,
as below;
$ perf probe -L pre_schedule
<pre_schedule@/home/mhiramat/ksrc/linux-2.6/kernel/sched.c:0>
0 static inline void pre_schedule(struct rq *rq, struct task_struct *prev)
{
2 if (prev->sched_class->pre_schedule)
3 prev->sched_class->pre_schedule(rq, prev);
}
However, current perf probe shows 4 local variables on pre_schedule(),
because it searches variables in the caller(schedule()) scope.
$ perf probe -V pre_schedule
Available variables at pre_schedule
@<schedule+445>
int cpu
long unsigned int* switch_count
struct rq* rq
struct task_struct* prev
This patch fixes this issue by searching variables in the local scope of
the instance of inlined function. Here is the result.
$ perf probe -V pre_schedule
Available variables at pre_schedule
@<schedule+445>
struct rq* rq
struct task_struct* prev
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: yrl.pp-manager.tt@hitachi.com
Link: http://lkml.kernel.org/r/20110811110259.19900.85664.stgit@fedora15
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
