Messages in this thread | | | Date | Thu, 17 Jun 2010 19:50:10 -0300 | From | Arnaldo Carvalho de Melo <> | Subject | Re: [RESEND][PATCH v2] perf probe: add kernel source path option |
| |
Corrected Masami-san address,
Em Mon, Jun 14, 2010 at 03:26:30PM -0400, Chase Douglas escreveu: > The probe plugin requires access to the source code for some operations. > The source code must be in the exact same location as specified by the > DWARF tags, but sometimes the location is an absolute path that cannot > be replicated by a normal user. This change adds the -s|--source option > to allow the user to specify the root of the kernel source tree. > > Signed-off-by: Chase Douglas <chase.douglas@canonical.com>
Just tested it, works as advertised:
[root@emilia mingo]# perf probe -L __schedule | head -22 Debuginfo analysis failed. (-2) Error: Failed to show lines. (-2) [root@emilia mingo]# perf probe -s /usr/src/debug/kernel-rt-2.6.33.4-rt20.19.el5rt/linux-2.6.33.4.x86_64 -L __schedule | head -22 <__schedule:0> 0 { struct task_struct *prev, *next; unsigned long *switch_count; struct rq *rq; int cpu; 6 cpu = smp_processor_id(); 7 rq = cpu_rq(cpu); 8 rcu_sched_qs(cpu); 9 prev = rq->curr; 10 switch_count = &prev->nivcsw; 12 release_kernel_lock(prev); schedule_debug(prev); 16 preempt_disable(); 18 if (sched_feat(HRTICK)) 19 hrtick_clear(rq); [root@emilia mingo]# cat /sys/kernel/debug/sched_features FAIR_SLEEPERS GENTLE_FAIR_SLEEPERS NO_NORMALIZED_SLEEPER START_DEBIT WAKEUP_PREEMPT ADAPTIVE_GRAN ASYM_GRAN NO_WAKEUP_SYNC NO_WAKEUP_OVERLAP SYNC_WAKEUPS AFFINE_WAKEUPS SYNC_LESS NO_SYNC_MORE NO_NEXT_BUDDY LAST_BUDDY CACHE_HOT_BUDDY NO_ARCH_POWER NO_HRTICK NO_DOUBLE_TICK LB_BIAS LB_SHARES_UPDATE ASYM_EFF_LOAD OWNER_SPIN [root@emilia mingo]# perf probe __schedule:19 Add new event: probe:__schedule (on __schedule:19)
You can now use it on all perf tools, such as:
perf record -e probe:__schedule -aR sleep 1
[root@emilia mingo]# perf record -e probe:__schedule -aR sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.061 MB perf.data (~2684 samples) ] [root@emilia mingo]# perf report | head -15 # # (For a higher level overview, try: perf report --sort comm,dso) # [root@emilia mingo]# echo HRTICK > /sys/kernel/debug/sched_features [root@emilia mingo]# cat /sys/kernel/debug/sched_features FAIR_SLEEPERS GENTLE_FAIR_SLEEPERS NO_NORMALIZED_SLEEPER START_DEBIT WAKEUP_PREEMPT ADAPTIVE_GRAN ASYM_GRAN NO_WAKEUP_SYNC NO_WAKEUP_OVERLAP SYNC_WAKEUPS AFFINE_WAKEUPS SYNC_LESS NO_SYNC_MORE NO_NEXT_BUDDY LAST_BUDDY CACHE_HOT_BUDDY NO_ARCH_POWER HRTICK NO_DOUBLE_TICK LB_BIAS LB_SHARES_UPDATE ASYM_EFF_LOAD OWNER_SPIN [root@emilia mingo]# perf record -e probe:__schedule -aR sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 1.528 MB perf.data (~66766 samples) ] [root@emilia mingo]# perf report | head -15 # Events: 19K cycles # # Overhead Command Shared Object Symbol # ........ ............... .................. ...... # 38.42% init 0 [k] 0000000000000000 5.42% sirq-timer/0 [kernel.kallsyms] [k] kernel_thread_helper 5.41% sirq-timer/5 [kernel.kallsyms] [k] kernel_thread_helper 5.41% sirq-timer/4 [kernel.kallsyms] [k] kernel_thread_helper 5.41% sirq-timer/3 [kernel.kallsyms] [k] kernel_thread_helper 5.41% sirq-timer/2 [kernel.kallsyms] [k] kernel_thread_helper 5.41% sirq-timer/1 [kernel.kallsyms] [k] kernel_thread_helper 5.40% sirq-timer/7 [kernel.kallsyms] [k] kernel_thread_helper 5.40% sirq-timer/6 [kernel.kallsyms] [k] kernel_thread_helper 5.26% swapper 0 [k] 0000000000000000 [root@emilia mingo]#
Really cool stuff, thanks for your patch and to Masami for his amazing work on 'perf probe'.
I think we should look at usual places like debuginfo packages and /usr/src/linux-`uname`, ~/git/linux-2.6/, etc, but that is something for a follow on patch.
Best Regards,
- Arnaldo
| |