From c530665c31c0140b74ca7689e7f836177796e5bd Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 3 Mar 2010 07:16:16 +0100 Subject: perf: Take a hot regs snapshot for trace events We are taking a wrong regs snapshot when a trace event triggers. Either we use get_irq_regs(), which gives us the interrupted registers if we are in an interrupt, or we use task_pt_regs() which gives us the state before we entered the kernel, assuming we are lucky enough to be no kernel thread, in which case task_pt_regs() returns the initial set of regs when the kernel thread was started. What we want is different. We need a hot snapshot of the regs, so that we can get the instruction pointer to record in the sample, the frame pointer for the callchain, and some other things. Let's use the new perf_fetch_caller_regs() for that. Comparison with perf record -e lock: -R -a -f -g Before: perf [kernel] [k] __do_softirq | --- __do_softirq | |--55.16%-- __open | --44.84%-- __write_nocancel After: perf [kernel] [k] perf_tp_event | --- perf_tp_event | |--41.07%-- lock_acquire | | | |--39.36%-- _raw_spin_lock | | | | | |--7.81%-- hrtimer_interrupt | | | smp_apic_timer_interrupt | | | apic_timer_interrupt The old case was producing unreliable callchains. Now having right frame and instruction pointers, we have the trace we want. Also syscalls and kprobe events already have the right regs, let's use them instead of wasting a retrieval. v2: Follow the rename perf_save_regs() -> perf_fetch_caller_regs() Signed-off-by: Frederic Weisbecker Cc: Ingo Molnar Cc: Thomas Gleixner Cc: H. Peter Anvin Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Steven Rostedt Cc: Arnaldo Carvalho de Melo Cc: Masami Hiramatsu Cc: Jason Baron Cc: Archs --- include/trace/ftrace.h | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'include/trace') diff --git a/include/trace/ftrace.h b/include/trace/ftrace.h index 0804cd594803..f31bb8b9777c 100644 --- a/include/trace/ftrace.h +++ b/include/trace/ftrace.h @@ -764,6 +764,7 @@ ftrace_profile_templ_##call(struct ftrace_event_call *event_call, \ struct ftrace_raw_##call *entry; \ u64 __addr = 0, __count = 1; \ unsigned long irq_flags; \ + struct pt_regs *__regs; \ int __entry_size; \ int __data_size; \ int rctx; \ @@ -784,8 +785,11 @@ ftrace_profile_templ_##call(struct ftrace_event_call *event_call, \ \ { assign; } \ \ + __regs = &__get_cpu_var(perf_trace_regs); \ + perf_fetch_caller_regs(__regs, 2); \ + \ ftrace_perf_buf_submit(entry, __entry_size, rctx, __addr, \ - __count, irq_flags); \ + __count, irq_flags, __regs); \ } #undef DEFINE_EVENT -- cgit v1.2.3 From 97d5a22005f38057b4bc0d95f81cd26510268794 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Fri, 5 Mar 2010 05:35:37 +0100 Subject: perf: Drop the obsolete profile naming for trace events Drop the obsolete "profile" naming used by perf for trace events. Perf can now do more than simple events counting, so generalize the API naming. Signed-off-by: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Steven Rostedt Cc: Masami Hiramatsu Cc: Jason Baron --- include/trace/ftrace.h | 38 +++++++++++++++++++------------------- include/trace/syscall.h | 8 ++++---- 2 files changed, 23 insertions(+), 23 deletions(-) (limited to 'include/trace') diff --git a/include/trace/ftrace.h b/include/trace/ftrace.h index f31bb8b9777c..25ab56f75d65 100644 --- a/include/trace/ftrace.h +++ b/include/trace/ftrace.h @@ -401,18 +401,18 @@ static inline notrace int ftrace_get_offsets_##call( \ #undef DEFINE_EVENT #define DEFINE_EVENT(template, name, proto, args) \ \ -static void ftrace_profile_##name(proto); \ +static void perf_trace_##name(proto); \ \ static notrace int \ -ftrace_profile_enable_##name(struct ftrace_event_call *unused) \ +perf_trace_enable_##name(struct ftrace_event_call *unused) \ { \ - return register_trace_##name(ftrace_profile_##name); \ + return register_trace_##name(perf_trace_##name); \ } \ \ static notrace void \ -ftrace_profile_disable_##name(struct ftrace_event_call *unused) \ +perf_trace_disable_##name(struct ftrace_event_call *unused) \ { \ - unregister_trace_##name(ftrace_profile_##name); \ + unregister_trace_##name(perf_trace_##name); \ } #undef DEFINE_EVENT_PRINT @@ -507,12 +507,12 @@ ftrace_profile_disable_##name(struct ftrace_event_call *unused) \ #ifdef CONFIG_PERF_EVENTS -#define _TRACE_PROFILE_INIT(call) \ - .profile_enable = ftrace_profile_enable_##call, \ - .profile_disable = ftrace_profile_disable_##call, +#define _TRACE_PERF_INIT(call) \ + .perf_event_enable = perf_trace_enable_##call, \ + .perf_event_disable = perf_trace_disable_##call, #else -#define _TRACE_PROFILE_INIT(call) +#define _TRACE_PERF_INIT(call) #endif /* CONFIG_PERF_EVENTS */ #undef __entry @@ -638,7 +638,7 @@ __attribute__((section("_ftrace_events"))) event_##call = { \ .unregfunc = ftrace_raw_unreg_event_##call, \ .print_fmt = print_fmt_##template, \ .define_fields = ftrace_define_fields_##template, \ - _TRACE_PROFILE_INIT(call) \ + _TRACE_PERF_INIT(call) \ } #undef DEFINE_EVENT_PRINT @@ -657,18 +657,18 @@ __attribute__((section("_ftrace_events"))) event_##call = { \ .unregfunc = ftrace_raw_unreg_event_##call, \ .print_fmt = print_fmt_##call, \ .define_fields = ftrace_define_fields_##template, \ - _TRACE_PROFILE_INIT(call) \ + _TRACE_PERF_INIT(call) \ } #include TRACE_INCLUDE(TRACE_INCLUDE_FILE) /* - * Define the insertion callback to profile events + * Define the insertion callback to perf events * * The job is very similar to ftrace_raw_event_ except that we don't * insert in the ring buffer but in a perf counter. * - * static void ftrace_profile_(proto) + * static void ftrace_perf_(proto) * { * struct ftrace_data_offsets_ __maybe_unused __data_offsets; * struct ftrace_event_call *event_call = &event_; @@ -757,7 +757,7 @@ __attribute__((section("_ftrace_events"))) event_##call = { \ #undef DECLARE_EVENT_CLASS #define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print) \ static notrace void \ -ftrace_profile_templ_##call(struct ftrace_event_call *event_call, \ +perf_trace_templ_##call(struct ftrace_event_call *event_call, \ proto) \ { \ struct ftrace_data_offsets_##call __maybe_unused __data_offsets;\ @@ -774,10 +774,10 @@ ftrace_profile_templ_##call(struct ftrace_event_call *event_call, \ sizeof(u64)); \ __entry_size -= sizeof(u32); \ \ - if (WARN_ONCE(__entry_size > FTRACE_MAX_PROFILE_SIZE, \ + if (WARN_ONCE(__entry_size > PERF_MAX_TRACE_SIZE, \ "profile buffer not large enough")) \ return; \ - entry = (struct ftrace_raw_##call *)ftrace_perf_buf_prepare( \ + entry = (struct ftrace_raw_##call *)perf_trace_buf_prepare( \ __entry_size, event_call->id, &rctx, &irq_flags); \ if (!entry) \ return; \ @@ -788,17 +788,17 @@ ftrace_profile_templ_##call(struct ftrace_event_call *event_call, \ __regs = &__get_cpu_var(perf_trace_regs); \ perf_fetch_caller_regs(__regs, 2); \ \ - ftrace_perf_buf_submit(entry, __entry_size, rctx, __addr, \ + perf_trace_buf_submit(entry, __entry_size, rctx, __addr, \ __count, irq_flags, __regs); \ } #undef DEFINE_EVENT #define DEFINE_EVENT(template, call, proto, args) \ -static notrace void ftrace_profile_##call(proto) \ +static notrace void perf_trace_##call(proto) \ { \ struct ftrace_event_call *event_call = &event_##call; \ \ - ftrace_profile_templ_##template(event_call, args); \ + perf_trace_templ_##template(event_call, args); \ } #undef DEFINE_EVENT_PRINT diff --git a/include/trace/syscall.h b/include/trace/syscall.h index 0387100752f0..e5e5f48dbfb3 100644 --- a/include/trace/syscall.h +++ b/include/trace/syscall.h @@ -47,10 +47,10 @@ enum print_line_t print_syscall_exit(struct trace_iterator *iter, int flags); #endif #ifdef CONFIG_PERF_EVENTS -int prof_sysenter_enable(struct ftrace_event_call *call); -void prof_sysenter_disable(struct ftrace_event_call *call); -int prof_sysexit_enable(struct ftrace_event_call *call); -void prof_sysexit_disable(struct ftrace_event_call *call); +int perf_sysenter_enable(struct ftrace_event_call *call); +void perf_sysenter_disable(struct ftrace_event_call *call); +int perf_sysexit_enable(struct ftrace_event_call *call); +void perf_sysexit_disable(struct ftrace_event_call *call); #endif #endif /* _TRACE_SYSCALL_H */ -- cgit v1.2.3