Re: [PATCH] Tracing events with GPIOs

[Tom Zanussi <tom.zanussi@linux.intel.com>]

On Fri, 2013-12-27 at 15:54 +0100, Jean-Jacques Hiblot wrote:
> Hi Tom, Steven,
> 
> I'm currently doing some tests with the event triggers
> (git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace.git
> for-next).
> At the moment I can't use event triggers with probes dynamically
> created (with perf probe or via
> /sys/kernel/debug/tracing/kprobe_events)
> Predefined probes like
> /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger are working
> fine.
> Am I missing something ?
> 

No, it's just that kprobes hasn't been hooked up yet to actually invoke
the triggers.

Let me look at doing that over the weekend...

Tom

> example:
> 
> ----------------------------------------------------------------
> with a predefined probe
> ----------------------------------------------------------------
> 
> prompt> echo 'stacktrace' >
> /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
> prompt> ls
> prompt> cat /sys/kernel/debug/tracing/trace
> # tracer: nop
> #
> # entries-in-buffer/entries-written: 14/14   #P:1
> #
> #                              _-----=> irqs-off
> #                             / _----=> need-resched
> #                            | / _---=> hardirq/softirq
> #                            || / _--=> preempt-depth
> #                            ||| /     delay
> #           TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
> #              | |       |   ||||       |         |
>               ls-971   [000] ....   298.050000: <stack trace>
>  => ftrace_raw_event_kmem_alloc
>  => kmem_cache_alloc_trace
>  => do_execve
>  => SyS_execve
>  => ret_fast_syscall
>               ls-971   [000] ....   298.050000: <stack trace>
>  => ftrace_raw_event_kmem_alloc
>  => kmem_cache_alloc_trace
>  => load_elf_binary
> [...]
> 
> ----------------------------------------------------------------
> with a dynamically defined probe
> ----------------------------------------------------------------
> 
> prompt> perf probe -a in_kmalloc=__kmalloc
> prompt> echo 'stacktrace' >
> /sys/kernel/debug/tracing/events/probe/in_kmalloc/trigger
> prompt> perf record -e probe:in_kmalloc -aR sleep 10&
> [... do something here ...]
> prompt> perf report
> [...]
> # Samples: 24  of event 'probe:in_kmalloc'
> # Event count (approx.): 24
> #
> # Overhead  Command      Shared Object         Symbol
> # ........  .......  .................  .............
> #
>     50.00%       ls  [kernel.kallsyms]  [k] __kmalloc
>     12.50%    sleep  [kernel.kallsyms]  [k] __kmalloc
>     12.50%       sh  [kernel.kallsyms]  [k] __kmalloc
>     12.50%     perf  [kernel.kallsyms]  [k] __kmalloc
>     12.50%     more  [kernel.kallsyms]  [k] __kmalloc
> [...]
> 
> prompt> cat /sys/kernel/debug/tracing/trace
> # tracer: nop
> #
> # entries-in-buffer/entries-written: 0/0   #P:1
> #
> #                              _-----=> irqs-off
> #                             / _----=> need-resched
> #                            | / _---=> hardirq/softirq
> #                            || / _--=> preempt-depth
> #              random: nonblocking pool is initialized
>               ||| /     delay
> #           TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
> #              | |       |   ||||       |         |