[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170406091604.37poma56ptmcn7de@hirez.programming.kicks-ass.net>
Date: Thu, 6 Apr 2017 11:16:04 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Rabin Vincent <rabin.vincent@...s.com>
Cc: ralf@...ux-mips.org, linux-mips@...ux-mips.org,
linux-kernel@...r.kernel.org, mingo@...hat.com, acme@...nel.org,
alexander.shishkin@...ux.intel.com, Rabin Vincent <rabinv@...s.com>
Subject: Re: [PATCH] MIPS: perf: fix deadlock
On Wed, Apr 05, 2017 at 03:14:08PM +0200, Rabin Vincent wrote:
That lock is disgusting... but yes patch looks about right.
I'll leave it to the MIPS people though.
> ---
> arch/mips/kernel/perf_event_mipsxx.c | 9 +++++----
> 1 file changed, 5 insertions(+), 4 deletions(-)
>
> diff --git a/arch/mips/kernel/perf_event_mipsxx.c b/arch/mips/kernel/perf_event_mipsxx.c
> index 8c35b31..9452b02 100644
> --- a/arch/mips/kernel/perf_event_mipsxx.c
> +++ b/arch/mips/kernel/perf_event_mipsxx.c
> @@ -1446,6 +1446,11 @@ static int mipsxx_pmu_handle_shared_irq(void)
> HANDLE_COUNTER(0)
> }
>
> +#ifdef CONFIG_MIPS_PERF_SHARED_TC_COUNTERS
> + read_unlock(&pmuint_rwlock);
> +#endif
> + resume_local_counters();
> +
> /*
> * Do all the work for the pending perf events. We can do this
> * in here because the performance counter interrupt is a regular
> @@ -1454,10 +1459,6 @@ static int mipsxx_pmu_handle_shared_irq(void)
> if (handled == IRQ_HANDLED)
> irq_work_run();
>
> -#ifdef CONFIG_MIPS_PERF_SHARED_TC_COUNTERS
> - read_unlock(&pmuint_rwlock);
> -#endif
> - resume_local_counters();
> return handled;
> }
>
> --
> 2.7.0
>
Powered by blists - more mailing lists