[PATCH 0/2] perf: Order SPE and TRBE padding before publishing AUX head

[PATCH 0/2] perf: Order SPE and TRBE padding before publishing AUX head

[Leo Yan]

SPE and TRBE write padding into their AUX buffers before advancing
aux_head. Add the missing write barriers so that a userspace reader on
another CPU cannot observe the new head before the padding stores
become visible.

AUX producers must make their data visible before publishing aux_head.
The required synchronization depends on the trace hardware and CPU
memory model, so the barriers belong in the drivers rather than in
perf_aux_output_skip() or perf_aux_output_end(). The TRBE change also
covers padding inserted by the overwrite erratum workaround before
perf_aux_output_end().

Signed-off-by: Leo Yan <leo.yan@arm.com>
---
Leo Yan (2):
      perf: arm_spe: Order padding before publishing AUX head
      coresight: trbe: Order padding before publishing AUX head

 drivers/hwtracing/coresight/coresight-trbe.c | 2 ++
 drivers/perf/arm_spe_pmu.c                   | 5 ++++-
 2 files changed, 6 insertions(+), 1 deletion(-)
---
base-commit: 0fcb05b72344f40806cbe810de8872f1698bf053
change-id: 20261015-add_barriers_for_perf_aux_output_skip-70356815b2bd

Best regards,
-- 
Leo Yan <leo.yan@arm.com>

[PATCH 1/2] perf: arm_spe: Order padding before publishing AUX head

[Leo Yan]

The driver fills alignment gaps and calls perf_aux_output_skip(), which
publishes the new aux_head without a memory barrier. A userspace
consumer on another CPU can therefore observe the new head before the
padding stores become visible.

Add smp_wmb() between the padding stores and perf_aux_output_skip().

Fixes: d5d9696b0380 ("drivers/perf: Add support for ARMv8.2 Statistical Profiling Extension")
Assisted-by: LLM
Signed-off-by: Leo Yan <leo.yan@arm.com>
---
 drivers/perf/arm_spe_pmu.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/perf/arm_spe_pmu.c b/drivers/perf/arm_spe_pmu.c
index b70f3fedf0282baba1395ce1116272ec5f961ea8..e105462ec5f38eab4776b0a654ba7c8a84f70f71 100644
--- a/drivers/perf/arm_spe_pmu.c
+++ b/drivers/perf/arm_spe_pmu.c
@@ -487,8 +487,11 @@ static void arm_spe_pmu_pad_buf(struct perf_output_handle *handle, int len)
 	u64 head = PERF_IDX2OFF(handle->head, buf);
 
 	memset(buf->base + head, ARM_SPE_BUF_PAD_BYTE, len);
-	if (!buf->snapshot)
+	if (!buf->snapshot) {
+		/* Order the padding stores before publishing the new aux_head. */
+		smp_wmb();
 		perf_aux_output_skip(handle, len);
+	}
 }
 
 static u64 arm_spe_pmu_next_snapshot_off(struct perf_output_handle *handle)

-- 
2.34.1

[PATCH 2/2] coresight: trbe: Order padding before publishing AUX head

[Leo Yan]

TRBE fills gaps in the AUX buffer before advancing aux_head. Neither
perf_aux_output_skip() nor perf_aux_output_end() guarantees ordering of
these CPU stores before publishing the new head.

Add smp_wmb() after memset() in __trbe_pad_buf(). This covers alignment
padding published by perf_aux_output_skip() and the overwrite erratum
padding written by trbe_get_trace_size() before perf_aux_output_end().

Fixes: 3fbf7f011f24 ("coresight: sink: Add TRBE driver")
Assisted-by: LLM
Signed-off-by: Leo Yan <leo.yan@arm.com>
---
 drivers/hwtracing/coresight/coresight-trbe.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/hwtracing/coresight/coresight-trbe.c b/drivers/hwtracing/coresight/coresight-trbe.c
index c7cbca45f2debd4047b93283ea9fe5dd9e1f2ebf..b605121b1dcfed9837c15cd92ff4c42515f39eaf 100644
--- a/drivers/hwtracing/coresight/coresight-trbe.c
+++ b/drivers/hwtracing/coresight/coresight-trbe.c
@@ -362,6 +362,8 @@ static void trbe_stop_and_truncate_event(struct perf_output_handle *handle)
 static void __trbe_pad_buf(struct trbe_buf *buf, u64 offset, int len)
 {
 	memset((void *)buf->trbe_base + offset, ETE_IGNORE_PACKET, len);
+	/* Order the padding stores before publishing the new aux_head. */
+	smp_wmb();
 }
 
 static void trbe_pad_buf(struct perf_output_handle *handle, int len)

-- 
2.34.1

Re: [PATCH 1/2] perf: arm_spe: Order padding before publishing AUX head

[sashiko-bot]

> The driver fills alignment gaps and calls perf_aux_output_skip(), which
> publishes the new aux_head without a memory barrier. A userspace
> consumer on another CPU can therefore observe the new head before the
> padding stores become visible.
> 
> Add smp_wmb() between the padding stores and perf_aux_output_skip().
> 
> Fixes: d5d9696b0380 ("drivers/perf: Add support for ARMv8.2 Statistical Profiling Extension")
> Assisted-by: LLM
> Signed-off-by: Leo Yan <leo.yan@arm.com>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261015-add_barriers_for_perf_aux_output_skip-v1-0-5847d0578859@arm.com?part=1

Re: [PATCH 2/2] coresight: trbe: Order padding before publishing AUX head

[sashiko-bot]

> TRBE fills gaps in the AUX buffer before advancing aux_head. Neither
> perf_aux_output_skip() nor perf_aux_output_end() guarantees ordering of
> these CPU stores before publishing the new head.
> 
> Add smp_wmb() after memset() in __trbe_pad_buf(). This covers alignment
> padding published by perf_aux_output_skip() and the overwrite erratum
> padding written by trbe_get_trace_size() before perf_aux_output_end().
> 
> Fixes: 3fbf7f011f24 ("coresight: sink: Add TRBE driver")
> Assisted-by: LLM
> Signed-off-by: Leo Yan <leo.yan@arm.com>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261015-add_barriers_for_perf_aux_output_skip-v1-0-5847d0578859@arm.com?part=2