- 29 May, 2009 13 commits
-
-
Arnaldo Carvalho de Melo authored
[acme@emilia linux-2.6-tip]$ pwd /home/acme/git/linux-2.6-tip Before (still available using -P/--full-paths) [acme@emilia linux-2.6-tip]$ perf report -P | head -10 11.48% perf: 7454 [kernel]: clear_page_c 4.89% perf: 7454 [kernel]: vsnprintf 4.61% perf: 7454 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: dso__find_symbol 4.09% perf: 7454 [kernel]: number 4.06% perf: 7454 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: dso__fprintf 4.00% perf: 7454 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: symbol_filter New default: [acme@emilia linux-2.6-tip]$ perf report | head -10 11.48% perf: 7454 [kernel]: clear_page_c 4.89% perf: 7454 [kernel]: vsnprintf 4.61% perf: 7454 ./Documentation/perf_counter/perf: dso__find_symbol 4.09% perf: 7454 [kernel]: number 4.06% perf: 7454 ./Documentation/perf_counter/perf: dso__fprintf 4.00% perf: 7454 ./Documentation/perf_counter/perf: symbol_filter Suggested-by:Ingo Molnar <mingo@elte.hu> Signed-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <20090529164859.GN4747@ghostprotocols.net> Signed-off-by:
-
Peter Zijlstra authored
When fork() fails we cannot use perf_counter_exit_task() since that assumes to operate on current. Write a new helper that cleans up unused/clean contexts. Signed-off-by:
Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: John Kacur <jkacur@redhat.com> LKML-Reference: <new-submission> Signed-off-by:
-
Peter Zijlstra authored
Remove the local_irq_save() etc.. in routines that are smp function calls, or have IRQs disabled by other means. Then change the COMM, MMAP, and swcounter context iteration to current->perf_counter_ctxp and RCU, since it really doesn't matter which context they iterate, they're all folded. Signed-off-by:
-
Peter Zijlstra authored
Commit a63eaf34 ("perf_counter: Dynamically allocate tasks' perf_counter_context struct") broke COMM and MMAP notification for cpu wide counters by dropping out early if there was no task context, thereby also not iterating the cpu context. Signed-off-by:
-
Ingo Molnar authored
This fixes a nasty crash and highlights a bug that we were freeing failed-fork() counters incorrectly. (the fix for that will come separately) [ Impact: fix crashes/lockups with inherited counters ] Acked-by:
-
Ingo Molnar authored
Peter noticed that we are sometimes reading cpuctx->task_ctx with interrupts enabled. Noticed-by:
-
Paul Mackerras authored
Peter Zijlstra pointed out that under some circumstances, we can take the mutex in a context or a counter and then swap that context or counter to another task, potentially leading to lock order inversions or the mutexes not protecting what they are supposed to protect. This fixes the problem by making sure that we never take a mutex in a context or counter which could get swapped to another task. Most of the cases where we take a mutex is on a top-level counter or context, i.e. a counter which has an fd associated with it or a context that contains such a counter. This adds WARN_ON_ONCE statements to verify that. The two cases where we need to take the mutex on a context that is a clone of another are in perf_counter_exit_task and perf_counter_init_task. The perf_counter_exit_task case is solved by uncloning the context before starting to remove the counters from it. The perf_counter_init_task is a little trickier; we temporarily disable context swapping for the parent (forking) task by setting its ctx->parent_gen to the all-1s value after locking the context, if it is a cloned context, and restore the ctx->parent_gen value at the end if the context didn't get uncloned in the meantime. This also moves the increment of the context generation count to be within the same critical section, protected by the context mutex, that adds the new counter to the context. That way, taking the mutex is sufficient to ensure that both the counter list and the generation count are stable. [ Impact: fix hangs, races with inherited and PID counters ] Signed-off-by:
Paul Mackerras <paulus@samba.org> Acked-by:
-
Ingo Molnar authored
Add new column that normalizes counter results by 'nanoseconds spent running' unit. Before: Performance counter stats for '/home/mingo/hackbench': 10469.403605 task clock ticks (msecs) 75502 context switches (events) 9501 CPU migrations (events) 36158 pagefaults (events) 31975676185 CPU cycles (events) 26257738659 instructions (events) 108740581 cache references (events) 54606088 cache misses (events) Wall-clock time elapsed: 810.514504 msecs After: Performance counter stats for '/home/mingo/hackbench': 10469.403605 task clock ticks (msecs) 75502 context switches # 0.007 M/sec 9501 CPU migrations # 0.001 M/sec 36158 pagefaults # 0.003 M/sec 31975676185 CPU cycles # 3054.202 M/sec 26257738659 instructions # 2508.045 M/sec 108740581 cache references # 10.387 M/sec 54606088 cache misses # 5.216 M/sec Wall-clock time elapsed: 810.514504 msecs The advantage of that column is that it is characteristic of the execution workflow, regardless of runtime. Hence 'hackbench 10' will look similar to 'hackbench 15' - while the absolute counter values are very different. Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: John Kacur <jkacur@redhat.com> LKML-Reference: <new-submission> Signed-off-by: -
Ingo Molnar authored
We introduce the extra pass to allow the print-out to possibly rely on already read counters. [ Impact: cleanup ] Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: John Kacur <jkacur@redhat.com> LKML-Reference: <new-submission> Signed-off-by: -
Ingo Molnar authored
[ Impact: cleanup ] Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: John Kacur <jkacur@redhat.com> LKML-Reference: <new-submission> Signed-off-by: -
Yong Wang authored
Always use NMI for performance-monitoring interrupt as there could be racy situations if we switch between irq and nmi mode frequently. Signed-off-by:
Yong Wang <yong.y.wang@intel.com> LKML-Reference: <20090529052835.GA13657@ywang-moblin2.bj.intel.com> Signed-off-by:
-
Mike Galbraith authored
Signed-off-by:
Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> LKML-Reference: <new-submission> Signed-off-by:
-
Mike Galbraith authored
Signed-off-by:
-
- 28 May, 2009 10 commits
-
-
Mike Galbraith authored
Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Now both perf top and report use the same routines. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Will be used by perf top. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Now one has just to use dso__load_kernel() optionally passing a vmlinux filename. Will make things easier for perf top that will want to pass a callback to filter some symbols. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
When creating a dso instance allow asking that all symbols in this dso have a private area just before the symbol. perf top will use this for its counters, etc. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Will be used by perf top as well. Signed-off-by:
-
Paul Mackerras authored
Commit 564c2b21 ("perf_counter: Optimize context switch between identical inherited contexts") introduced a race where it is possible that a counter being attached to a task could get attached to the wrong task, if the task is one that has inherited its context from another task via fork. This happens because the optimized context switch could switch the context to another task after find_get_context has read task->perf_counter_ctxp. In fact, it's possible that the context could then get freed, if the other task then exits. This fixes the problem by protecting both the context switch and the critical code in find_get_context with spinlocks. The context switch locks the cxt->lock of both the outgoing and incoming contexts before swapping them. That means that once code such as find_get_context has obtained the spinlock for the context associated with a task, the context can't get swapped to another task. However, the context may have been swapped in the interval between reading task->perf_counter_ctxp and getting the lock, so it is necessary to check and retry. To make sure that none of the contexts being looked at in find_get_context can get freed, this changes the context freeing code to use RCU. Thus an rcu_read_lock() is sufficient to ensure that no contexts can get freed. This part of the patch is lifted from a patch posted by Peter Zijlstra. This also adds a check to make sure that we can't add a counter to a task that is exiting. There is also a race between perf_counter_exit_task and find_get_context; this solves the race by moving the get_ctx that was in perf_counter_alloc into the locked region in find_get_context, so that once find_get_context has got the context for a task, it won't get freed even if the task calls perf_counter_exit_task. It doesn't matter if new top-level (non-inherited) counters get attached to the context after perf_counter_exit_task has detached the context from the task. They will just stay there and never get scheduled in until the counters' fds get closed, and then perf_release will remove them from the context and eventually free the context. With this, we are now doing the unclone in find_get_context rather than when a counter was added to or removed from a context (actually, we were missing the unclone_ctx() call when adding a counter to a context). We don't need to unclone when removing a counter from a context because we have no way to remove a counter from a cloned context. This also takes out the smp_wmb() in find_get_context, which Peter Zijlstra pointed out was unnecessary because the cmpxchg implies a full barrier anyway. Signed-off-by:
-
Ingo Molnar authored
Acked-by:
-
Peter Zijlstra authored
Implement this style of header: # # Overhead Command File: Symbol # ........ ....... ............ # for the various --sort variants as well. Signed-off-by:
-
Ingo Molnar authored
Pointed out by compiler warnings: tip/include/linux/perf_counter.h:644: warning: no return statement in function returning non-void Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: John Kacur <jkacur@redhat.com> LKML-Reference: <new-submission> Signed-off-by:
-
- 27 May, 2009 17 commits
-
-
Ingo Molnar authored
This error condition: aldebaran:~/linux/linux/Documentation/perf_counter> perf report dso__load_sym: cannot get elf header. failed to open: /etc/ld.so.cache problem processing PERF_EVENT_MMAP, bailing out caused the profile to be very short - as the error was at the beginning of the file and we bailed out completely. Be more permissive and consider the event broken instead. Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: John Kacur <jkacur@redhat.com> LKML-Reference: <new-submission> Signed-off-by: -
Ingo Molnar authored
Old default output: 3.12% perf-report [.] ./perf-report: dsos__find 2.44% perf-report [k] kernel: kallsyms_expand_symbol 2.28% :4483 [.] <unknown>: <unknown> 2.05% :4174 [k] kernel: _spin_lock_irqsave 2.01% perf-report [k] kernel: vsnprintf 1.92% perf-report [k] kernel: format_decode 1.92% :4438 [k] kernel: _spin_lock New default output: # # Overhead Command File: Symbol # ........ ....... ............ # 6.54% perf [k] kernel: kallsyms_expand_symbol 6.26% perf [.] /home/mingo/tip/Documentation/perf_counter/perf: dso__insert_symbol 4.76% perf [.] /home/mingo/tip/Documentation/perf_counter/perf: hex2long 4.55% perf [k] kernel: number 4.48% perf [k] kernel: format_decode 4.09% perf [k] kernel: vsnprintf Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: John Kacur <jkacur@redhat.com> Cc: Mike Galbraith <efault@gmx.de> LKML-Reference: <20090527182101.229504802@chello.nl> Signed-off-by: -
Peter Zijlstra authored
Signed-off-by:
-
Peter Zijlstra authored
Signed-off-by:
-
Peter Zijlstra authored
option parsing for dynamic sorting. Signed-off-by:
-
Peter Zijlstra authored
Make the sorting and printing dynamic. Signed-off-by:
-
Peter Zijlstra authored
In preparation for configurable sorting, rework the histgram code a bit. Signed-off-by:
-
Peter Zijlstra authored
Allow to use vmlinux instead of kallsyms. Signed-off-by:
-
Peter Zijlstra authored
Some distros seem to store debuginfo in weird places. Signed-off-by:
-
Ingo Molnar authored
Pekka reported build failure in builtin-report.c: CC builtin-report.o In file included from builtin-report.c:7: /usr/include/ctype.h:102: error: expected expression before token And observed: | Removing #include <ctype.h> from builtin-report.c makes the problem | go away. I am running Ubuntu 9.04 that has gcc 4.3.3 and libc 2.9. Reported-by:Pekka J Enberg <penberg@cs.helsinki.fi> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: John Kacur <jkacur@redhat.com> LKML-Reference: <new-submission> Signed-off-by:
-
Mike Galbraith authored
Perf record bails if no command argument is provided, so you can't use naked -a or -p to profile a running task or the whole box. Allow foreground profiling of an existing pid or the entire system. [ Impact: fix command option handling bug ] Signed-off-by:
-
Ingo Molnar authored
Add Git's pager.c (and sigchain) code. A command only has to call setup_pager() to get paged interactive output. Non-interactive (redirected, command-piped, etc.) uses are not affected. Update perf-report to make use of this. [ Impact: new feature ] Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: John Kacur <jkacur@redhat.com> LKML-Reference: <new-submission> Signed-off-by: -
Ingo Molnar authored
output.perf is only output to perf-record - it's input to perf-report. So change it to a more direction-neutral name. Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: John Kacur <jkacur@redhat.com> LKML-Reference: <new-submission> Signed-off-by: -
Ingo Molnar authored
Tighten up our C code requirements: - disallow warnings - disallow declarations-mixed-with-statements - require proper prototypes - require C99 (with gcc extensions) Fix up a ton of problems these measures unearth: - unused functions - needlessly global functions - missing prototypes - code mixed with declarations 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: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: John Kacur <jkacur@redhat.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <20090526222155.GJ4424@ghostprotocols.net> Signed-off-by: -
Arnaldo Carvalho de Melo authored
perf: report should show the IP only in --verbose mode [acme@emilia ~]$ perf report | head 4.95 find [k] _spin_lock 2.19 find [k] ext3fs_dirhash [ext3] 1.87 find [k] __rcu_read_lock 1.86 find [k] _atomic_dec_and_lock 1.86 find [.] /lib64/libc-2.5.so: __GI_strlen 1.85 find [k] __kmalloc 1.62 find [.] /lib64/libc-2.5.so: vfprintf 1.59 find [k] __rcu_read_unlock 1.55 find [k] __d_lookup Signed-off-by:
-
Ingo Molnar authored
- allow 'W' symbols too - Convert initializations to C99 style - whitespace cleanups 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: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: Marcelo Tosatti <mtosatti@redhat.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: John Kacur <jkacur@redhat.com> Cc: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <20090526222155.GJ4424@ghostprotocols.net> Signed-off-by: -
Arnaldo Carvalho de Melo authored
Just like we do for userspace when reading the symtab, reducing the number of entries we insert on the symbols rbtree. Before: [acme@emilia ~]$ rm -f perf_report.perf ; perf record -o perf_report.perf perf stat perf report > /dev/null Performance counter stats for 'perf': 218.138382 task clock ticks (msecs) 4 context switches (events) 8 CPU migrations (events) 2136 pagefaults (events) 32746212 CPU cycles (events) (scaled from 67.04%) 11961102 instructions (events) (scaled from 66.19%) 49841 cache references (events) (scaled from 21.96%) 13777 cache misses (events) (scaled from 21.98%) Wall-clock time elapsed: 218.702477 msecs [acme@emilia ~]$ perf report -i perf_report.perf | head 11.06 perf [.] 0x00000000000057cb /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: dso__find_symbol 9.15 perf [.] 0x00000000000056a0 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: dso__insert_symbol 8.72 perf [k] 0xffffffff8101b1d2 intel_pmu_enable_all 8.51 perf [.] 0x0000000000006672 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: thread__symbol_incnew 3.83 perf [k] 0xffffffff811cfc5a vsnprintf 3.40 perf [.] 0x0000000000005e33 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: hex 3.40 perf [.] 0x0000000000005ec7 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: hex2long 3.19 perf [k] 0xffffffff811ce1c1 number 2.77 perf [.] 0x0000000000006869 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: threads__findnew 2.77 perf [.] 0x000000000000fde3 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: rb_insert_color [acme@emilia ~]$ After: acme@emilia ~]$ rm -f perf_report.perf ; perf record -o perf_report.perf perf stat perf report > /dev/null Performance counter stats for 'perf': 190.228511 task clock ticks (msecs) 4 context switches (events) 7 CPU migrations (events) 1625 pagefaults (events) 29578745 CPU cycles (events) (scaled from 66.92%) 10516914 instructions (events) (scaled from 66.47%) 44015 cache references (events) (scaled from 22.04%) 8248 cache misses (events) (scaled from 22.07%) Wall-clock time elapsed: 190.816096 msecs [acme@emilia ~]$ perf report -i perf_report.perf | head 15.99 perf [.] 0x00000000000057a9 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: dso__find_symbol 10.87 perf [.] 0x000000000000674d /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: thread__symbol_incnew 8.74 perf [k] 0xffffffff8101b1d2 intel_pmu_enable_all 5.54 perf [.] 0x0000000000005e42 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: hex 4.48 perf [.] 0x0000000000005ebe /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: hex2long 4.48 perf [k] 0xffffffff811cfba0 vsnprintf 3.84 perf [.] 0x00000000000056b4 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: dso__insert_symbol 3.62 perf [.] 0x00000000000068d0 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: threads__findnew 3.20 perf [k] 0xffffffff811ce0b3 number 2.56 perf [.] 0x0000000000006d78 /home/acme/git/linux-2.6-tip/Documentation/perf_counter/perf: __cmd_report [acme@emilia ~]$ [ Impact: optimization ] Signed-off-by:
-
