From 8cbd9cc6254065c97c4bac42daa55ba1abe73a8e Mon Sep 17 00:00:00 2001 From: David Sharp Date: Tue, 13 Nov 2012 12:18:21 -0800 Subject: tracing,x86: Add a TSC trace_clock In order to promote interoperability between userspace tracers and ftrace, add a trace_clock that reports raw TSC values which will then be recorded in the ring buffer. Userspace tracers that also record TSCs are then on exactly the same time base as the kernel and events can be unambiguously interlaced. Tested: Enabled a tracepoint and the "tsc" trace_clock and saw very large timestamp values. v2: Move arch-specific bits out of generic code. v3: Rename "x86-tsc", cleanups v7: Generic arch bits in Kbuild. Google-Bug-Id: 6980623 Link: http://lkml.kernel.org/r/1352837903-32191-1-git-send-email-dhsharp@google.com Acked-by: Ingo Molnar Cc: Masami Hiramatsu Cc: Ingo Molnar Cc: Thomas Gleixner Cc: "H. Peter Anvin" Signed-off-by: David Sharp Signed-off-by: Steven Rostedt --- include/asm-generic/trace_clock.h | 16 ++++++++++++++++ include/linux/trace_clock.h | 2 ++ 2 files changed, 18 insertions(+) create mode 100644 include/asm-generic/trace_clock.h (limited to 'include') diff --git a/include/asm-generic/trace_clock.h b/include/asm-generic/trace_clock.h new file mode 100644 index 0000000..6726f1b --- /dev/null +++ b/include/asm-generic/trace_clock.h @@ -0,0 +1,16 @@ +#ifndef _ASM_GENERIC_TRACE_CLOCK_H +#define _ASM_GENERIC_TRACE_CLOCK_H +/* + * Arch-specific trace clocks. + */ + +/* + * Additional trace clocks added to the trace_clocks + * array in kernel/trace/trace.c + * None if the architecture has not defined it. + */ +#ifndef ARCH_TRACE_CLOCKS +# define ARCH_TRACE_CLOCKS +#endif + +#endif /* _ASM_GENERIC_TRACE_CLOCK_H */ diff --git a/include/linux/trace_clock.h b/include/linux/trace_clock.h index 4eb4902..d563f37 100644 --- a/include/linux/trace_clock.h +++ b/include/linux/trace_clock.h @@ -12,6 +12,8 @@ #include #include +#include + extern u64 notrace trace_clock_local(void); extern u64 notrace trace_clock(void); extern u64 notrace trace_clock_global(void); -- cgit v1.1 From 8be0709f10e3dd5d7d07933ad61a9f18c4b93ca5 Mon Sep 17 00:00:00 2001 From: David Sharp Date: Tue, 13 Nov 2012 12:18:22 -0800 Subject: tracing: Format non-nanosec times from tsc clock without a decimal point. With the addition of the "tsc" clock, formatting timestamps to look like fractional seconds is misleading. Mark clocks as either in nanoseconds or not, and format non-nanosecond timestamps as decimal integers. Tested: $ cd /sys/kernel/debug/tracing/ $ cat trace_clock [local] global tsc $ echo sched_switch > set_event $ echo 1 > tracing_on ; sleep 0.0005 ; echo 0 > tracing_on $ cat trace -0 [000] 6330.555552: sched_switch: prev_comm=swapper prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=bash next_pid=29964 next_prio=120 sleep-29964 [000] 6330.555628: sched_switch: prev_comm=bash prev_pid=29964 prev_prio=120 prev_state=S ==> next_comm=swapper next_pid=0 next_prio=120 ... $ echo 1 > options/latency-format $ cat trace -0 0 4104553247us+: sched_switch: prev_comm=swapper prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=bash next_pid=29964 next_prio=120 sleep-29964 0 4104553322us+: sched_switch: prev_comm=bash prev_pid=29964 prev_prio=120 prev_state=S ==> next_comm=swapper next_pid=0 next_prio=120 ... $ echo tsc > trace_clock $ cat trace $ echo 1 > tracing_on ; sleep 0.0005 ; echo 0 > tracing_on $ echo 0 > options/latency-format $ cat trace -0 [000] 16490053398357: sched_switch: prev_comm=swapper prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=bash next_pid=31128 next_prio=120 sleep-31128 [000] 16490053588518: sched_switch: prev_comm=bash prev_pid=31128 prev_prio=120 prev_state=S ==> next_comm=swapper next_pid=0 next_prio=120 ... echo 1 > options/latency-format $ cat trace -0 0 91557653238+: sched_switch: prev_comm=swapper prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=bash next_pid=31128 next_prio=120 sleep-31128 0 91557843399+: sched_switch: prev_comm=bash prev_pid=31128 prev_prio=120 prev_state=S ==> next_comm=swapper next_pid=0 next_prio=120 ... v2: Move arch-specific bits out of generic code. v4: Fix x86_32 build due to 64-bit division. Google-Bug-Id: 6980623 Link: http://lkml.kernel.org/r/1352837903-32191-2-git-send-email-dhsharp@google.com Cc: Masami Hiramatsu Signed-off-by: David Sharp Signed-off-by: Steven Rostedt --- include/linux/ftrace_event.h | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'include') diff --git a/include/linux/ftrace_event.h b/include/linux/ftrace_event.h index b80c8dd..a3d4895 100644 --- a/include/linux/ftrace_event.h +++ b/include/linux/ftrace_event.h @@ -86,6 +86,12 @@ struct trace_iterator { cpumask_var_t started; }; +enum trace_iter_flags { + TRACE_FILE_LAT_FMT = 1, + TRACE_FILE_ANNOTATE = 2, + TRACE_FILE_TIME_IN_NS = 4, +}; + struct trace_event; -- cgit v1.1 From 1c7d66732458dc187008e3f5b2f71e019e320fc2 Mon Sep 17 00:00:00 2001 From: Shan Wei Date: Sat, 3 Nov 2012 12:38:33 +0800 Subject: tracing: Kill unused and puzzled sample code in ftrace.h When doing per-cpu helper optimizing work, find that this code is so puzzled. 1. It's mark as comment text, maybe a sample function for guidelines or a todo work. 2. But, this sample code is odd where struct perf_trace_buf is nonexistent. commit ce71b9 delete struct perf_trace_buf definition. Author: Frederic Weisbecker Date: Sun Nov 22 05:26:55 2009 +0100 tracing: Use the perf recursion protection from trace event Is it necessary to keep there? just compile test. Link: http://lkml.kernel.org/r/50949FC9.6050202@gmail.com Signed-off-by: Shan Wei Signed-off-by: Steven Rostedt --- include/trace/ftrace.h | 73 -------------------------------------------------- 1 file changed, 73 deletions(-) (limited to 'include') diff --git a/include/trace/ftrace.h b/include/trace/ftrace.h index 698f2a8..40dc5e8 100644 --- a/include/trace/ftrace.h +++ b/include/trace/ftrace.h @@ -619,79 +619,6 @@ __attribute__((section("_ftrace_events"))) *__event_##call = &event_##call #include TRACE_INCLUDE(TRACE_INCLUDE_FILE) -/* - * 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_perf_(proto) - * { - * struct ftrace_data_offsets_ __maybe_unused __data_offsets; - * struct ftrace_event_call *event_call = &event_; - * extern void perf_tp_event(int, u64, u64, void *, int); - * struct ftrace_raw_##call *entry; - * struct perf_trace_buf *trace_buf; - * u64 __addr = 0, __count = 1; - * unsigned long irq_flags; - * struct trace_entry *ent; - * int __entry_size; - * int __data_size; - * int __cpu - * int pc; - * - * pc = preempt_count(); - * - * __data_size = ftrace_get_offsets_(&__data_offsets, args); - * - * // Below we want to get the aligned size by taking into account - * // the u32 field that will later store the buffer size - * __entry_size = ALIGN(__data_size + sizeof(*entry) + sizeof(u32), - * sizeof(u64)); - * __entry_size -= sizeof(u32); - * - * // Protect the non nmi buffer - * // This also protects the rcu read side - * local_irq_save(irq_flags); - * __cpu = smp_processor_id(); - * - * if (in_nmi()) - * trace_buf = rcu_dereference_sched(perf_trace_buf_nmi); - * else - * trace_buf = rcu_dereference_sched(perf_trace_buf); - * - * if (!trace_buf) - * goto end; - * - * trace_buf = per_cpu_ptr(trace_buf, __cpu); - * - * // Avoid recursion from perf that could mess up the buffer - * if (trace_buf->recursion++) - * goto end_recursion; - * - * raw_data = trace_buf->buf; - * - * // Make recursion update visible before entering perf_tp_event - * // so that we protect from perf recursions. - * - * barrier(); - * - * //zero dead bytes from alignment to avoid stack leak to userspace: - * *(u64 *)(&raw_data[__entry_size - sizeof(u64)]) = 0ULL; - * entry = (struct ftrace_raw_ *)raw_data; - * ent = &entry->ent; - * tracing_generic_entry_update(ent, irq_flags, pc); - * ent->type = event_call->id; - * - * <- do some jobs with dynamic arrays - * - * <- affect our values - * - * perf_tp_event(event_call->id, __addr, __count, entry, - * __entry_size); <- submit them to perf counter - * - * } - */ #ifdef CONFIG_PERF_EVENTS -- cgit v1.1