Re: [PATCH 4/3] perf_counter: Correct PERF_SAMPLE_RAW output

From: Frederic Weisbecker
Date: Mon Aug 10 2009 - 05:48:27 EST


On Mon, Aug 10, 2009 at 11:27:27AM +0200, Peter Zijlstra wrote:
> Subject: perf_counter: Correct PERF_SAMPLE_RAW output
> From: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
> Date: Mon Aug 10 11:16:52 CEST 2009
>
> PERF_SAMPLE_* output switches should unconditionally output the
> correct format, as they are the only way to unambiguously parse the
> PERF_EVENT_SAMPLE data.
>
> Signed-off-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>


I economized this part because I thought ftrace events could
be parsed the same way we parse binary events through debugfs.
The size of an ftrace event can be deduced from its format.

But I must confess that would require a pre-parsing of dynamic
strings.

Acked-by: Frederic Weisbecker <fweisbec@xxxxxxxxx>


> ---
> include/linux/perf_counter.h | 2 ++
> include/trace/ftrace.h | 3 ++-
> kernel/perf_counter.c | 30 ++++++++++++++++++++++++------
> 3 files changed, 28 insertions(+), 7 deletions(-)
>
> Index: linux-2.6/include/linux/perf_counter.h
> ===================================================================
> --- linux-2.6.orig/include/linux/perf_counter.h
> +++ linux-2.6/include/linux/perf_counter.h
> @@ -369,6 +369,8 @@ enum perf_event_type {
> *
> * { u64 nr,
> * u64 ips[nr]; } && PERF_SAMPLE_CALLCHAIN
> + * { u32 size;
> + * char data[size];}&& PERF_SAMPLE_RAW
> * };
> */
> PERF_EVENT_SAMPLE = 9,
> Index: linux-2.6/include/trace/ftrace.h
> ===================================================================
> --- linux-2.6.orig/include/trace/ftrace.h
> +++ linux-2.6/include/trace/ftrace.h
> @@ -687,7 +687,8 @@ static void ftrace_profile_##call(proto)
> pc = preempt_count(); \
> \
> __data_size = ftrace_get_offsets_##call(&__data_offsets, args); \
> - __entry_size = ALIGN(__data_size + sizeof(*entry), sizeof(u64));\
> + __entry_size = ALIGN(__data_size + sizeof(*entry) + sizeof(u32),\
> + sizeof(u64)); \
> \
> do { \
> char raw_data[__entry_size]; \
> Index: linux-2.6/kernel/perf_counter.c
> ===================================================================
> --- linux-2.6.orig/kernel/perf_counter.c
> +++ linux-2.6/kernel/perf_counter.c
> @@ -2647,7 +2647,6 @@ void perf_counter_output(struct perf_cou
> u64 counter;
> } group_entry;
> struct perf_callchain_entry *callchain = NULL;
> - struct perf_raw_record *raw = NULL;
> int callchain_size = 0;
> u64 time;
> struct {
> @@ -2717,9 +2716,15 @@ void perf_counter_output(struct perf_cou
> }
>
> if (sample_type & PERF_SAMPLE_RAW) {
> - raw = data->raw;
> - if (raw)
> - header.size += raw->size;
> + int size = sizeof(u32);
> +
> + if (data->raw)
> + size += data->raw->size;
> + else
> + size += sizeof(u32);
> +
> + WARN_ON_ONCE(size & (sizeof(u64)-1));
> + header.size += size;
> }
>
> ret = perf_output_begin(&handle, counter, header.size, nmi, 1);
> @@ -2785,8 +2790,21 @@ void perf_counter_output(struct perf_cou
> }
> }
>
> - if ((sample_type & PERF_SAMPLE_RAW) && raw)
> - perf_output_copy(&handle, raw->data, raw->size);
> + if (sample_type & PERF_SAMPLE_RAW) {
> + if (data->raw) {
> + perf_output_put(&handle, data->raw->size);
> + perf_output_copy(&handle, data->raw->data, data->raw->size);
> + } else {
> + struct {
> + u32 size;
> + u32 data;
> + } raw = {
> + .size = sizeof(u32),
> + .data = 0,
> + };
> + perf_output_put(&handle, raw);
> + }
> + }
>
> perf_output_end(&handle);
> }
>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/