[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20250501114546.GL551819@e132581.arm.com>
Date: Thu, 1 May 2025 12:45:46 +0100
From: Leo Yan <leo.yan@....com>
To: James Clark <james.clark@...aro.org>
Cc: Suzuki K Poulose <suzuki.poulose@....com>,
Mike Leach <mike.leach@...aro.org>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
coresight@...ts.linaro.org, linux-arm-kernel@...ts.infradead.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] coresight: etm4x: Remove redundant claim register setting
On Thu, May 01, 2025 at 11:54:58AM +0100, James Clark wrote:
>
> On 01/05/2025 11:42 am, Leo Yan wrote:
> > The claim register is set twice in the restore flow; remove the
> > duplicate operation.
> >
> > Signed-off-by: Leo Yan <leo.yan@....com>
> > ---
> > drivers/hwtracing/coresight/coresight-etm4x-core.c | 2 --
> > 1 file changed, 2 deletions(-)
> >
> > diff --git a/drivers/hwtracing/coresight/coresight-etm4x-core.c b/drivers/hwtracing/coresight/coresight-etm4x-core.c
> > index 5c20ed4cf4ed..228317991ec2 100644
> > --- a/drivers/hwtracing/coresight/coresight-etm4x-core.c
> > +++ b/drivers/hwtracing/coresight/coresight-etm4x-core.c
> > @@ -1958,8 +1958,6 @@ static void __etm4_cpu_restore(struct etmv4_drvdata *drvdata)
> > if (drvdata->numvmidc > 4)
> > etm4x_relaxed_write32(csa, state->trcvmidcctlr0, TRCVMIDCCTLR1);
> > - etm4x_relaxed_write32(csa, state->trcclaimset, TRCCLAIMSET);
> > -
> > if (!drvdata->skip_power_up)
> > etm4x_relaxed_write32(csa, state->trcpdcr, TRCPDCR);
>
> Reviewed-by: James Clark <james.clark@...aro.org>
>
> Semi-related to this patch but probably not worth fixing: There's
> technically a race condition with an external debugger here. It would be
> better to use coresight_claim_device_unlocked() and report the error that's
> returned instead of manually writing to the whole claim register without
> checking it before and after like the claim function does.
I thought again and conclude the current patch would be fine (at least
not bad ;)
The reason is the ETM driver has acquired the OS lock during idle
states, this prevents the race condition with external debuggers.
And if a CPU detects pending interrupts and abort a low power flow,
the CoreSight self claim tag is still set. In this case,
coresight_claim_device_unlocked() will report unnecessary warnings.
Thanks,
Leo
Powered by blists - more mailing lists