- 16 Dec, 2009 11 commits
-
-
Peter Zijlstra authored
In order to allow for per-task-per-cpu counters, useful for scalability when profiling task hierarchies, we allow installing events with event->cpu != -1 in task contexts. __perf_event_sched_in() already skips events where ->cpu mis-matches the current cpu, fix up __perf_install_in_context() and __perf_event_enable() to also respect this filter. This does lead to vary hard to interpret enabled/running times for such counters, but I don't see a simple solution for that. Signed-off-by:
Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: fweisbec@gmail.com Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <20091216165904.831451147@chello.nl> Signed-off-by:
Ingo Molnar <mingo@elte.hu>
-
Arnaldo Carvalho de Melo authored
Otherwise we do integer math and the delta values round up to multiples of 1.0%. Also, calculate absolute values. Things look precise now: $ perf report -i perf.data.old --sort dso,symbol | head -13 9.02% libc-2.10.1.so [.] _IO_vfprintf_internal 4.88% find [.] 0x00000000014af0 2.91% [kernel] [k] __kmalloc 2.85% [kernel] [k] ext4_htree_store_dirent 2.50% libc-2.10.1.so [.] __GI_memmove 2.44% [kernel] [k] half_md4_transform 2.43% [kernel] [k] _spin_lock 2.33% [kernel] [k] system_call $ perf report -i perf.data --sort dso,symbol | head -13 8.55% libc-2.10.1.so [.] _IO_vfprintf_internal 3.11% [kernel] [k] __kmalloc 3.07% [kernel] [k] ext4_htree_store_dirent 2.66% find [.] 0x00000000016bcf 2.61% [kernel] [k] _atomic_dec_and_lock 2.46% [kernel] [k] half_md4_transform 2.41% libc-2.10.1.so [.] __GI_memmove 2.30% find [.] 0x00000000009219 $ perf diff | head -13 9.02% -0.47% libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% +0.20% [kernel] [k] __kmalloc 2.85% +0.23% [kernel] [k] ext4_htree_store_dirent 1.99% +0.62% [kernel] [k] _atomic_dec_and_lock 2.44% +0.02% [kernel] [k] half_md4_transform 2.50% -0.09% libc-2.10.1.so [.] __GI_memmove 1.88% +0.01% [kernel] [k] __d_lookup 2.43% -0.75% [kernel] [k] _spin_lock 0.97% +0.62% [kernel] [k] path_get 1.99% -0.42% libc-2.10.1.so [.] _int_malloc $ Signed-off-by:Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260981109-2621-1-git-send-email-acme@infradead.org> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
This is a more intuitive / more meaningful default: $ perf diff | head -8 9.02% +1.00% libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% [kernel] [k] __kmalloc 2.85% -1.00% [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% [kernel] [k] _atomic_dec_and_lock 2.44% [kernel] $ Suggested-by: -
Arnaldo Carvalho de Melo authored
That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ perf record -f find / > /dev/null [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.062 MB perf.data (~2699 samples) ] $ perf record -f find / > /dev/null [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.062 MB perf.data (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: -
Arnaldo Carvalho de Melo authored
Introduced in: d599db3f "perf report: Generalize perf_session__fprintf_hists()" Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Those don't make sense for tools such as 'perf diff'. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Will be used in other tools such as 'perf diff'. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Pull it out of builtin-report - further changes will be made and it will then be reusable in 'perf diff' as well. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
So that --dsos, --comm, --symbols can bem used in more tools, like in perf diff: $ perf record -f find / > /dev/null $ perf record -f find / > /dev/null $ perf diff --dsos /lib64/libc-2.10.1.so | head -5 1 +22392124 /lib64/libc-2.10.1.so _IO_vfprintf_internal 2 +6410655 /lib64/libc-2.10.1.so __GI_memmove 3 +1 +9192692 /lib64/libc-2.10.1.so _int_malloc 4 -1 -15158605 /lib64/libc-2.10.1.so _int_free 5 +45669 /lib64/libc-2.10.1.so _IO_new_file_xsputn $ Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Will be used in perf diff too. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
This simplifies a lot of functions, less stuff to be done by tool writers. Signed-off-by:
-
- 15 Dec, 2009 29 commits
-
-
Masami Hiramatsu authored
Fix perf probe to show which probe point is not found. With out this patch, it shows just "No probe point found." This doesn't help users if they specify several probes. e.g. # perf probe -f --add schedule --add test Fatal: No probe point found. This patch makes error message more helpful as below. # perf probe --add schedule --add test Fatal: Probe point 'test' not found. - probe not added. Signed-off-by:
Masami Hiramatsu <mhiramat@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Jason Baron <jbaron@redhat.com> Cc: K.Prasad <prasad@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: systemtap <systemtap@sources.redhat.com> Cc: DLE <dle-develop@lists.sourceforge.net> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20091215153247.17436.49068.stgit@dhcp-100-2-132.bos.redhat.com> Signed-off-by:
-
Masami Hiramatsu authored
Check symbols in symtab/kallsyms when no debuginfo is available. e.g. # perf probe test Fatal: Kernel symbol 'test' not found - probe not added. Signed-off-by:
-
Masami Hiramatsu authored
Check build-id of vmlinux by using functions in symbol.c. This also exposes map__load() for getting vmlinux path, and removes vmlinux path list in builtin-probe.c, because symbol.c already has that. Checking build-id prevents users to open old or different debuginfo from current running kernel. Signed-off-by:
-
Masami Hiramatsu authored
Reject second attempt of adding same-name event. This patch also provides --force option which allows user to add additional probe events on the same-name event. e.g. (the first attempt : success) ./perf probe schedule Added new event: probe:schedule (on schedule+0) (the second attempt : failure) ./perf probe schedule:11 Error: event "schedule" already exists. (Use -f to force duplicates.) Fatal: Can't add new event. (the second attempt with -f : successfully added) ./perf probe -f schedule:11 Added new event: probe:schedule_1 (on schedule+45) Signed-off-by:
-
Masami Hiramatsu authored
Support event name syntax for --add option. This allows users to specify event name for each new event. The --add syntax is: perf probe --add '[EVENT=]SRC:LINE ARGS' or perf probe --add '[EVENT=]FUNC[+OFFS|%return|:RLN][@SRC] ARGS' e.g. ./perf probe --add myprobe1=schedule Note: currently group name is not supported yet, because it can cause name-space confliction with other tracepoint/ hw-breakpoint events. Signed-off-by:
-
Masami Hiramatsu authored
Add glob-expression matching support on --del option. You can use wildcards for specifying deleting events. e.g. Clear all probe events: # perf probe --del '*' Clear probes on schedule(): # perf probe --del 'schedule*' Signed-off-by:
-
Masami Hiramatsu authored
Use strlist__for_each macros instead of using strlist__entry() and index variable. Signed-off-by:
-
Masami Hiramatsu authored
Add for_each iteration macros for strlist. This patch introduces strlist__for_each() and strlist__for_each_safe(), both are similar to list_for_each() and list_for_each_safe(). Signed-off-by:
-
Masami Hiramatsu authored
Fix --del option to update current existing event list after perf probe deleted an event. Signed-off-by:
-
Masami Hiramatsu authored
Fix --del option to show info message instead of warning if failing to find specified event. Signed-off-by:
-
Masami Hiramatsu authored
Show need-dwarf message only if the probe is really requires debuginfo analysis. This also use pr_debug for debugging message instead of pr_warning. Signed-off-by:
-
Masami Hiramatsu authored
Check hyphen only command argument, because perf-probe doesn't support event recording feature yet. e.g. # perf probe - Error: '-' is not supported. Signed-off-by:
-
Masami Hiramatsu authored
Fix show_perf_probe_event() to check the result of e_snprintf(). Signed-off-by:
-
Masami Hiramatsu authored
Clean up struct session in builtin-probe.c, including change need_dwarf to bool and move listing flag into struct session as list_events flag. This also changes parse_perf_probe_event() interface due to code readability. Signed-off-by:
-
Peter Zijlstra authored
The miss-alignment of bp_addr created a 32bit hole, causing different structure packings on 32 and 64 bit machines. Fix that by moving __reserve_2 into that hole. Further, remove the useless struct and redundant __bp_reserve muck. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Add a newline do fix this problem: ERROR: perf-diff.txt: line 31: closing [blockdef-listing] delimiter expected Signed-off-by:
-
Ingo Molnar authored
Fix the short line displayed by 'perf' and also fix some other details in the longer text. Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <new-submission> Signed-off-by: -
Paul Mackerras authored
It is quite legitimate for CPUs to be numbered sparsely, meaning that it possible for an online CPU to have a number which is greater than the total count of possible CPUs. Currently find_get_context() has a sanity check on the cpu number where it checks it against num_possible_cpus(). This test can fail for a legitimate cpu number if the cpu_possible_mask is sparsely populated. This fixes the problem by checking the CPU number against nr_cpumask_bits instead, since that is the appropriate check to ensure that the cpu number is same to pass to cpu_isset() subsequently. Reported-by:
Michael Neuling <mikey@neuling.org> Signed-off-by:
Paul Mackerras <paulus@samba.org> Tested-by:
Peter Zijlstra <peterz@infradead.org> Cc: <stable@kernel.org> LKML-Reference: <20091215084032.GA18661@brick.ozlabs.ibm.com> Signed-off-by:
-
Tom Zanussi authored
Update the perf-trace page with new and missing options and remove some unused ones. Signed-off-by:
Tom Zanussi <tzanussi@gmail.com> Cc: fweisbec@gmail.com Cc: rostedt@goodmis.org LKML-Reference: <1260867220-15699-7-git-send-email-tzanussi@gmail.com> Signed-off-by:
-
Tom Zanussi authored
Allow scripts to be recorded/executed by simply specifying the script root name (the script name minus extension) along with 'record' or 'report' to 'perf trace'. The script names shown by 'perf trace -l' can be directly used to run the command-line contained within the corresponding '-record' and '-report' versions of scripts in the scripts/*/bin directories. For example, to record the trace data needed to run the wakeup-latency.pl script, the user can easily find the name of the corresponding script from the script list and invoke it using 'perf trace record', without having to remember the details of how to do the same thing using the lower-level perf trace command-line options: root@tropicana:~# perf trace -l List of available trace scripts: workqueue-stats workqueue stats (ins/exe/create/destroy) wakeup-latency system-wide min/max/avg wakeup latency rw-by-file <comm> r/w activity for a program, by file check-perf-trace useless but exhaustive test script rw-by-pid system-wide r/w activity root@tropicana:~# perf trace record wakeup-latency ^C[ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.296 MB perf.data (~12931 samples) ] To run the wakeup-latency.pl script using the captured data, change 'record' to 'report' in the command-line: root@tropicana:~# perf trace report wakeup-latency wakeup_latency stats: total_wakeups: 65 avg_wakeup_latency (ns): 22417 min_wakeup_latency (ns): 3470 max_wakeup_latency (ns): 223311 perf trace Perl script stopped If the script takes options, thay can be simply added to the end of the 'report' invocation: root@tropicana:~# perf trace record rw-by-file ^C[ perf record: Woken up 2 times to write data ] [ perf record: Captured and wrote 0.782 MB perf.data (~34171 samples) ] root@tropicana:~# perf trace report rw-by-file perf file read counts for perf: fd # reads bytes_requested ------ ---------- ----------- 122 1934 1980416 120 1 32 file write counts for perf: fd # writes bytes_written ------ ---------- ----------- 3 4006 280568 perf trace Perl script stopped Signed-off-by: -
Tom Zanussi authored
Lists the available perf trace scripts, one per line e.g.: root@tropicana:~# perf trace -l List of available trace scripts: workqueue-stats workqueue stats (ins/exe/create/destroy) wakeup-latency system-wide min/max/avg wakeup latency rw-by-file <comm> r/w activity for a program, by file check-perf-trace useless but exhaustive test script rw-by-pid system-wide r/w activity To be consistent with the other listing options in perf, the current latency trace option was changed to '-L', and '-l' is now used to access the script listing as: To create the list, it searches each scripts/*/bin directory for files ending with "-report" and reads information found in certain comment lines contained in those shell scripts: - if the comment line starts with "description:", the rest of the line is used as a 'half-line' description. To keep each line in the list to a single line, the description should be limited to 40 characters (the rest of the line contains the script name and args) - if the comment line starts with "args:", the rest of the line names the args the script supports. Required args should be surrounded by <> brackets, optional args by [] brackets. The current scripts in scripts/perl/bin have also been updated with description: and args: comments. Signed-off-by: -
Tom Zanussi authored
The return value from perl_run() is currently ignored, but it should be checked and used to exit perf if there are problems loading the script. Signed-off-by:
-
Tom Zanussi authored
README and Makefile.PL don't need to be installed for Perl run-time support. Signed-off-by:
-
Tom Zanussi authored
One oversight of the original scripting_ops patch was a lack of support for passing args to handler scripts. This adds argc/argv to the start_script() scripting_op, and changes the rw-by-file script to take 'comm' arg rather than the 'perf' value currently hard-coded. It also takes the opportunity to do some related minor cleanup. Signed-off-by:
-
Frederic Weisbecker authored
perf top, report and annotate all define their own symbol_conf, it should be static. Signed-off-by:
Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1260843322-6602-1-git-send-regression-fweisbec@gmail.com> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
I guess it is enough to show some examples: [root@doppio linux-2.6-tip]# rm -f perf.data* [root@doppio linux-2.6-tip]# ls -la perf.data* ls: cannot access perf.data*: No such file or directory [root@doppio linux-2.6-tip]# perf record -f find / > /dev/null [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.062 MB perf.data (~2699 samples) ] [root@doppio linux-2.6-tip]# ls -la perf.data* -rw------- 1 root root 74440 2009-12-14 20:03 perf.data [root@doppio linux-2.6-tip]# perf record -f find / > /dev/null [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.062 MB perf.data (~2692 samples) ] [root@doppio linux-2.6-tip]# ls -la perf.data* -rw------- 1 root root 74280 2009-12-14 20:03 perf.data -rw------- 1 root root 74440 2009-12-14 20:03 perf.data.old [root@doppio linux-2.6-tip]# perf diff | head -5 1 -34994580 /lib64/libc-2.10.1.so _IO_vfprintf_internal 2 -15307806 [kernel.kallsyms] __kmalloc 3 +1 +3665941 /lib64/libc-2.10.1.so __GI_memmove 4 +4 +23508995 /lib64/libc-2.10.1.so _int_malloc 5 +7 +38538813 [kernel.kallsyms] __d_lookup [root@doppio linux-2.6-tip]# perf diff -p | head -5 1 +1.00% /lib64/libc-2.10.1.so _IO_vfprintf_internal 2 [kernel.kallsyms] __kmalloc 3 +1 /lib64/libc-2.10.1.so __GI_memmove 4 +4 /lib64/libc-2.10.1.so _int_malloc 5 +7 -1.00% [kernel.kallsyms] __d_lookup [root@doppio linux-2.6-tip]# perf diff -v | head -5 1 361449551 326454971 -34994580 /lib64/libc-2.10.1.so _IO_vfprintf_internal 2 151009241 135701435 -15307806 [kernel.kallsyms] __kmalloc 3 +1 101805328 105471269 +3665941 /lib64/libc-2.10.1.so __GI_memmove 4 +4 78041440 101550435 +23508995 /lib64/libc-2.10.1.so _int_malloc 5 +7 59536172 98074985 +38538813 [kernel.kallsyms] __d_lookup [root@doppio linux-2.6-tip]# perf diff -vp | head -5 1 9.00% 8.00% +1.00% /lib64/libc-2.10.1.so _IO_vfprintf_internal 2 3.00% 3.00% [kernel.kallsyms] __kmalloc 3 +1 2.00% 2.00% /lib64/libc-2.10.1.so __GI_memmove 4 +4 2.00% 2.00% /lib64/libc-2.10.1.so _int_malloc 5 +7 1.00% 2.00% -1.00% [kernel.kallsyms] __d_lookup [root@doppio linux-2.6-tip]# This should be enough for diffs where the system is non volatile, i.e. when one doesn't updates binaries. For volatile environments, stay tuned for the next perf tool feature: a buildid cache populated by 'perf record', managed by 'perf buildid-cache' a-la ccache, and used by all the report tools. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Suggested-by:
Steven Rostedt <rostedt@goodmis.org> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
And it is also needed by 'perf diff'. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Signed-off-by:
-
