Re: [PATCH] perf_events: fix event scheduling issues introduced by transactional API (take 2)

[stephane eranian]

On Tue, May 25, 2010 at 3:35 PM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> On Tue, 2010-05-25 at 15:20 +0200, Stephane Eranian wrote:
>
>> With this patch, you can now overcommit the PMU even with pinned
>> system-wide events present and still get valid counts.
>
> Does this patch differ from the one you send earlier?
>
Yes, it does. It changes the way n_added is updated.

The first version was buggy (perf top would crash the machine).
You cannot delay updating n_added until commit, because there
are paths where you don't go through transactions. This version
instead keeps the number of events last added and speculatively
updates n_added (assuming success). If the transaction succeeds,
then no correction is done (subtracting 0), if it fails, then the number
of events just added to n_added is subtracted.
--
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/