[tip: perf/core] perf/x86/intel/pt: Fix sampling synchronization

[tip-bot2 for Adrian Hunter]

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     d92792a4b26e50b96ab734cbe203d8a4c932a7a9
Gitweb:        https://git.kernel.org/tip/d92792a4b26e50b96ab734cbe203d8a4c932a7a9
Author:        Adrian Hunter <adrian.hunter@xxxxxxxxx>
AuthorDate:    Mon, 15 Jul 2024 19:07:00 +03:00
Committer:     Peter Zijlstra <peterz@xxxxxxxxxxxxx>
CommitterDate: Mon, 29 Jul 2024 12:16:24 +02:00

perf/x86/intel/pt: Fix sampling synchronization

pt_event_snapshot_aux() uses pt->handle_nmi to determine if tracing
needs to be stopped, however tracing can still be going because
pt->handle_nmi is set to zero before tracing is stopped in pt_event_stop,
whereas pt_event_snapshot_aux() requires that tracing must be stopped in
order to copy a sample of trace from the buffer.

Instead call pt_config_stop() always, which anyway checks config for
RTIT_CTL_TRACEEN and does nothing if it is already clear.

Note pt_event_snapshot_aux() can continue to use pt->handle_nmi to
determine if the trace needs to be restarted afterwards.

Fixes: 25e8920b301c ("perf/x86/intel/pt: Add sampling support")
Signed-off-by: Adrian Hunter <adrian.hunter@xxxxxxxxx>
Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
Cc: stable@xxxxxxxxxxxxxxx
Link: https://lkml.kernel.org/r/20240715160712.127117-2-adrian.hunter@xxxxxxxxx
---
 arch/x86/events/intel/pt.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/arch/x86/events/intel/pt.c b/arch/x86/events/intel/pt.c
index b4aa8da..2959970 100644
--- a/arch/x86/events/intel/pt.c
+++ b/arch/x86/events/intel/pt.c
@@ -1606,6 +1606,7 @@ static void pt_event_stop(struct perf_event *event, int mode)
 	 * see comment in intel_pt_interrupt().
 	 */
 	WRITE_ONCE(pt->handle_nmi, 0);
+	barrier();
 
 	pt_config_stop(event);
 
@@ -1657,11 +1658,10 @@ static long pt_event_snapshot_aux(struct perf_event *event,
 		return 0;
 
 	/*
-	 * Here, handle_nmi tells us if the tracing is on
+	 * There is no PT interrupt in this mode, so stop the trace and it will
+	 * remain stopped while the buffer is copied.
 	 */
-	if (READ_ONCE(pt->handle_nmi))
-		pt_config_stop(event);
-
+	pt_config_stop(event);
 	pt_read_offset(buf);
 	pt_update_head(pt);
 
@@ -1673,11 +1673,10 @@ static long pt_event_snapshot_aux(struct perf_event *event,
 	ret = perf_output_copy_aux(&pt->handle, handle, from, to);
 
 	/*
-	 * If the tracing was on when we turned up, restart it.
-	 * Compiler barrier not needed as we couldn't have been
-	 * preempted by anything that touches pt->handle_nmi.
+	 * Here, handle_nmi tells us if the tracing was on.
+	 * If the tracing was on, restart it.
 	 */
-	if (pt->handle_nmi)
+	if (READ_ONCE(pt->handle_nmi))
 		pt_config_start(event);
 
 	return ret;