Re: [PATCH 3/4] MIPS/Perf-events: temporarily connect event to its pmu at event init

[Deng-Cheng Zhu <dczhu@mips.com>]

On 10/24/2011 06:56 PM, Deng-Cheng Zhu wrote:
> When arch level event init is called, the event is not yet connected to
> the PMU, thereby causing validate_group() to always do dummy work. On MIPS,
> this is due to the following lines in validate_event() called by
> validate_group():
>
> if (event->pmu !=&pmu || event->state<= PERF_EVENT_STATE_OFF)
>          return 1;
>
> This patch fixes it.
>
> Signed-off-by: Deng-Cheng Zhu<dczhu@mips.com>
> Cc: Peter Zijlstra<a.p.zijlstra@chello.nl>
> Cc: Paul Mackerras<paulus@samba.org>
> Cc: Ingo Molnar<mingo@elte.hu>
> Cc: Arnaldo Carvalho de Melo<acme@ghostprotocols.net>
> Cc: David Daney<david.daney@cavium.com>
> ---
>   arch/mips/kernel/perf_event_mipsxx.c |   17 +++++++++++++----
>   1 files changed, 13 insertions(+), 4 deletions(-)
>
> diff --git a/arch/mips/kernel/perf_event_mipsxx.c 
> b/arch/mips/kernel/perf_event_mipsxx.c
> index 1f654ca..c804fdd 100644
> --- a/arch/mips/kernel/perf_event_mipsxx.c
> +++ b/arch/mips/kernel/perf_event_mipsxx.c
> @@ -548,6 +548,7 @@ static int __hw_perf_event_init(struct perf_event *event)
>         struct perf_event_attr *attr =&event->attr;
>         struct hw_perf_event *hwc =&event->hw;
>         const struct mips_perf_event *pev;
> +       struct pmu *tmp;
>         int err;
>
>         /* Returning MIPS event descriptor for generic perf event. */
> @@ -611,11 +612,19 @@ static int __hw_perf_event_init(struct perf_event *event)
>         }
>
>         err = 0;
> -       if (event->group_leader != event) {
> +
> +       /*
> +        * we temporarily connect event to its pmu such that
> +        * validate_event() in validate_group() can classify
> +        * it as a MIPS event by passing (event->pmu ==&pmu).
> +        */
> +       tmp = event->pmu;
> +       event->pmu =&pmu;
> +
> +       if (event->group_leader != event)
>                 err = validate_group(event);
> -               if (err)
> -                       return -EINVAL;
> -       }
> +
> +       event->pmu = tmp;
>
>         event->destroy = hw_perf_event_destroy;
After looking at David Ahern's reply to my another patch (link
provided below), I started to think whether the PMU and event state
checks are redundant in validate_event() on MIPS (ARM may also have
the same consideration).

The related patch link:
http://www.spinics.net/lists/kernel/msg1252452.html

_If_ the state fix goes to group checks instead of to the perf tool,
then the following line in validate_event() on MIPS seems redundant:

if (event->pmu !=&pmu || event->state <= PERF_EVENT_STATE_OFF)
        return 1;

This is because validate_event() is only called by validate_group()
which is called only by __hw_perf_event_init(), and the issue of the
pmu check is already addressed in this patch, and we don't want the
grouped events which are in PERF_EVENT_STATE_OFF (by "attr->disabled =
1") to be filtered out here.

Also, _if_ the state fix goes to group checks instead of to the perf
tool, then X86 may need a patch to allow collect_events() to do real
work in validate_group().

Any comments?


Deng-Cheng