[PATCH v3 06/10] coresight: trbe: Fix handling of spurious interrupts

[Mathieu Poirier]

On Tue, Sep 21, 2021 at 10:29:20PM +0100, Suzuki K Poulose wrote:
> On 21/09/2021 18:24, Mathieu Poirier wrote:
> > On Tue, Sep 14, 2021 at 11:26:37AM +0100, Suzuki K Poulose wrote:
> > > On a spurious IRQ, right now we disable the TRBE and then re-enable
> > > it back, resetting the "buffer" pointers(i.e BASE, LIMIT and more
> > > importantly WRITE) to the original pointers from the AUX handle.
> > > This implies that we overwrite any trace that was written so far,
> > > (by overwriting TRBPTR) while we should have ignored the IRQ.
> > > 
> > > This patch cleans the behavior, by only stopping the TRBE if the
> > > IRQ was indeed raised, as we can read the TRBSR without stopping
> > > the TRBE (Only writes to the TRBSR requires the TRBE disabled).
> > > And also, on detecting a spurious IRQ after examining the TRBSR,
> > > we simply re-enable the TRBE without touching the other parameters.
> > > 
> > > Cc: Anshuman Khandual <anshuman.khandual at arm.com>
> > > Cc: Mathieu Poirier <mathieu.poirier at linaro.org>
> > > Cc: Mike Leach <mike.leach at linaro.org>
> > > Cc: Leo Yan <leo.yan at linaro.org>
> > > Signed-off-by: Suzuki K Poulose <suzuki.poulose at arm.com>
> > > ---
> > >   drivers/hwtracing/coresight/coresight-trbe.c | 30 ++++++++++----------
> > >   1 file changed, 15 insertions(+), 15 deletions(-)
> > > 
> > > diff --git a/drivers/hwtracing/coresight/coresight-trbe.c b/drivers/hwtracing/coresight/coresight-trbe.c
> > > index 5297b11f26b7..de99dd0aecd3 100644
> > > --- a/drivers/hwtracing/coresight/coresight-trbe.c
> > > +++ b/drivers/hwtracing/coresight/coresight-trbe.c
> > > @@ -677,16 +677,16 @@ static int arm_trbe_disable(struct coresight_device *csdev)
> > >   static void trbe_handle_spurious(struct perf_output_handle *handle)
> > >   {
> > > -	struct trbe_buf *buf = etm_perf_sink_config(handle);
> > > +	u64 limitr = read_sysreg_s(SYS_TRBLIMITR_EL1);
> > > -	buf->trbe_limit = compute_trbe_buffer_limit(handle);
> > > -	buf->trbe_write = buf->trbe_base + PERF_IDX2OFF(handle->head, buf);
> > > -	if (buf->trbe_limit == buf->trbe_base) {
> > > -		trbe_drain_and_disable_local();
> > > -		perf_aux_output_flag(handle, PERF_AUX_FLAG_TRUNCATED);
> > > -		return;
> > > -	}
> > > -	trbe_enable_hw(buf);
> > > +	/*
> > > +	 * If the IRQ was spurious, simply re-enable the TRBE
> > > +	 * back without modifying the buffer parameters to
> > > +	 * retain the trace collected so far.
> > > +	 */
> > > +	limitr |= TRBLIMITR_ENABLE;
> > > +	write_sysreg_s(limitr, SYS_TRBLIMITR_EL1);
> > > +	isb();
> > 
> > I understand (and agree with) this part of the patch...
> > 
> > >   }
> > >   static void trbe_handle_overflow(struct perf_output_handle *handle)
> > > @@ -759,12 +759,7 @@ static irqreturn_t arm_trbe_irq_handler(int irq, void *dev)
> > >   	enum trbe_fault_action act;
> > >   	u64 status;
> > > -	/*
> > > -	 * Ensure the trace is visible to the CPUs and
> > > -	 * any external aborts have been resolved.
> > > -	 */
> > > -	trbe_drain_and_disable_local();
> > > -
> > > +	/* Reads to TRBSR_EL1 is fine when TRBE is active */
> > >   	status = read_sysreg_s(SYS_TRBSR_EL1);
> > >   	/*
> > >   	 * If the pending IRQ was handled by update_buffer callback
> 
> [0] See below
> 
> > > @@ -773,6 +768,11 @@ static irqreturn_t arm_trbe_irq_handler(int irq, void *dev)
> > >   	if (!is_trbe_irq(status))
> > >   		return IRQ_NONE;
> > > +	/*
> > > +	 * Ensure the trace is visible to the CPUs and
> > > +	 * any external aborts have been resolved.
> > > +	 */
> > > +	trbe_drain_and_disable_local();
> > 
> > But not this part...  I can see why you'd want to move this after the check for
> > is_trbe_irq(), but not how it relates to spurious interrupts.  To me it seems
> > like it is addressing another issue.  If those code snippets are related then a
> > good dose of comments is missing.
> 
> This step is to make sure that we stop the TRBE only when there was
> really something to process. (i.e, TRBSR indicates an IRQ was raised).
> Also, there is a comment [0] above, for handling a case where the TRBE
> event was consumed by the "update_buffer()" due to a race with IRQ
> handler. Thus we stop the TRBE only when we need to analyse the cause
> and take an action. I agree there is a bit of disconnect.
> 
> I can think of the following options:
> 
>  - Split the patch to 2. with
>    1. Don't stop the trbe if there is no IRQ (the bit explained above)
>    2. Don't reset the TRBE ptrs on spurious IRQ

Please do two patches.

Other than this patch I commented on 07 and picked up 08.  Patches 09 and 10
won't apply if 06 and 07 aren't present so please address comments for 06 and 07
and resend all 4 patches (06, 07, 09, 10).

Thanks,
Mathieu

> 
>    OR
> 
>  - Add the above comment to the section.
> 
> The commit description has a hint, "This patch cleans the behavior, by only
> stopping the TRBE if the IRQ was indeed raised", but I agree that
> the code could be documented too.
> 
> Let me know what you think.
> 
> Thanks for the review.
> 
> Suzuki