[PATCH 08/19] coresight: trbe: Use PERF_AUX_FLAG_PARTIAL instead of PERF_AUX_FLAG_COLLISION
Anshuman Khandual
anshuman.khandual at arm.com
Thu Dec 4 20:28:02 PST 2025
On 01/12/25 4:51 PM, Leo Yan wrote:
> When tracing stops, there is no collision with other samples, so using
> PERF_AUX_FLAG_COLLISION does not accurately reflect the trace state and
> may mislead userspace.
>
> Use PERF_AUX_FLAG_PARTIAL instead to indicate that tracing stopped and
> the record may contain gaps.
>
> Signed-off-by: Leo Yan <leo.yan at arm.com>
> ---
> drivers/hwtracing/coresight/coresight-trbe.c | 19 ++++++-------------
> 1 file changed, 6 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/hwtracing/coresight/coresight-trbe.c b/drivers/hwtracing/coresight/coresight-trbe.c
> index b06885a08e082fd34f68d9588518807b5c47c86e..0caa4a6b437a3aa39fc6bcc72a23711b54f7c598 100644
> --- a/drivers/hwtracing/coresight/coresight-trbe.c
> +++ b/drivers/hwtracing/coresight/coresight-trbe.c
> @@ -669,21 +669,14 @@ static enum trbe_fault_action trbe_get_fault_act(struct perf_output_handle *hand
> }
>
> /*
> - * Mark the buffer to indicate that there was a WRAP event by
> - * setting the COLLISION flag. This indicates to the user that
> - * the TRBE trace collection was stopped without stopping the
> - * ETE and thus there might be some amount of trace that was
> - * lost between the time the WRAP was detected and the IRQ
> - * was consumed by the CPU.
> - *
> - * Setting the TRUNCATED flag would move the event to STOPPED
> - * state unnecessarily, even when there is space left in the
> - * ring buffer. Using the COLLISION flag doesn't have this side
> - * effect. We only set TRUNCATED flag when there is no space
> - * left in the ring buffer.
> + * Mark the buffer to indicate that the trace is stopped by setting
> + * the PARTIAL flag. This indicates to the user that the TRBE trace
> + * collection was stopped without stopping the ETE and thus there
> + * might be some amount of trace that was lost between the time the
> + * TRBE event was detected and the IRQ was consumed by the CPU.
> */
> if (!is_trbe_running(trbsr))
> - perf_aux_output_flag(handle, PERF_AUX_FLAG_COLLISION);
> + perf_aux_output_flag(handle, PERF_AUX_FLAG_PARTIAL);
This is interesting. So there is a no possibility for the records to be
overridden in the perf ring buffer and incomplete record is all that can
happen when TRBE is stopped, without stopping the ETE first ? Could both
of these scenarios might be possible as well ?
>
> if (is_trbe_wrap(trbsr))
> return TRBE_FAULT_ACT_WRAP;
>
More information about the linux-arm-kernel
mailing list