[PATCH 5.10 075/193] ptp: annotate data-race around q->head and q->tail

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



5.10-stable review patch.  If anyone has any objections, please let me know.

------------------

From: Eric Dumazet <edumazet@xxxxxxxxxx>

[ Upstream commit 73bde5a3294853947252cd9092a3517c7cb0cd2d ]

As I was working on a syzbot report, I found that KCSAN would
probably complain that reading q->head or q->tail without
barriers could lead to invalid results.

Add corresponding READ_ONCE() and WRITE_ONCE() to avoid
load-store tearing.

Fixes: d94ba80ebbea ("ptp: Added a brand new class driver for ptp clocks.")
Signed-off-by: Eric Dumazet <edumazet@xxxxxxxxxx>
Acked-by: Richard Cochran <richardcochran@xxxxxxxxx>
Link: https://lore.kernel.org/r/20231109174859.3995880-1-edumazet@xxxxxxxxxx
Signed-off-by: Jakub Kicinski <kuba@xxxxxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---
 drivers/ptp/ptp_chardev.c | 3 ++-
 drivers/ptp/ptp_clock.c   | 5 +++--
 drivers/ptp/ptp_private.h | 8 ++++++--
 drivers/ptp/ptp_sysfs.c   | 3 ++-
 4 files changed, 13 insertions(+), 6 deletions(-)

diff --git a/drivers/ptp/ptp_chardev.c b/drivers/ptp/ptp_chardev.c
index af3bc65c4595d..9311f3d09c8fc 100644
--- a/drivers/ptp/ptp_chardev.c
+++ b/drivers/ptp/ptp_chardev.c
@@ -487,7 +487,8 @@ ssize_t ptp_read(struct posix_clock *pc,
 
 	for (i = 0; i < cnt; i++) {
 		event[i] = queue->buf[queue->head];
-		queue->head = (queue->head + 1) % PTP_MAX_TIMESTAMPS;
+		/* Paired with READ_ONCE() in queue_cnt() */
+		WRITE_ONCE(queue->head, (queue->head + 1) % PTP_MAX_TIMESTAMPS);
 	}
 
 	spin_unlock_irqrestore(&queue->lock, flags);
diff --git a/drivers/ptp/ptp_clock.c b/drivers/ptp/ptp_clock.c
index 21c4c34c52d8d..ed766943a3563 100644
--- a/drivers/ptp/ptp_clock.c
+++ b/drivers/ptp/ptp_clock.c
@@ -55,10 +55,11 @@ static void enqueue_external_timestamp(struct timestamp_event_queue *queue,
 	dst->t.sec = seconds;
 	dst->t.nsec = remainder;
 
+	/* Both WRITE_ONCE() are paired with READ_ONCE() in queue_cnt() */
 	if (!queue_free(queue))
-		queue->head = (queue->head + 1) % PTP_MAX_TIMESTAMPS;
+		WRITE_ONCE(queue->head, (queue->head + 1) % PTP_MAX_TIMESTAMPS);
 
-	queue->tail = (queue->tail + 1) % PTP_MAX_TIMESTAMPS;
+	WRITE_ONCE(queue->tail, (queue->tail + 1) % PTP_MAX_TIMESTAMPS);
 
 	spin_unlock_irqrestore(&queue->lock, flags);
 }
diff --git a/drivers/ptp/ptp_private.h b/drivers/ptp/ptp_private.h
index 6b97155148f11..d2cb956706763 100644
--- a/drivers/ptp/ptp_private.h
+++ b/drivers/ptp/ptp_private.h
@@ -55,9 +55,13 @@ struct ptp_clock {
  * that a writer might concurrently increment the tail does not
  * matter, since the queue remains nonempty nonetheless.
  */
-static inline int queue_cnt(struct timestamp_event_queue *q)
+static inline int queue_cnt(const struct timestamp_event_queue *q)
 {
-	int cnt = q->tail - q->head;
+	/*
+	 * Paired with WRITE_ONCE() in enqueue_external_timestamp(),
+	 * ptp_read(), extts_fifo_show().
+	 */
+	int cnt = READ_ONCE(q->tail) - READ_ONCE(q->head);
 	return cnt < 0 ? PTP_MAX_TIMESTAMPS + cnt : cnt;
 }
 
diff --git a/drivers/ptp/ptp_sysfs.c b/drivers/ptp/ptp_sysfs.c
index 8cd59e8481631..8d52815e05b31 100644
--- a/drivers/ptp/ptp_sysfs.c
+++ b/drivers/ptp/ptp_sysfs.c
@@ -78,7 +78,8 @@ static ssize_t extts_fifo_show(struct device *dev,
 	qcnt = queue_cnt(queue);
 	if (qcnt) {
 		event = queue->buf[queue->head];
-		queue->head = (queue->head + 1) % PTP_MAX_TIMESTAMPS;
+		/* Paired with READ_ONCE() in queue_cnt() */
+		WRITE_ONCE(queue->head, (queue->head + 1) % PTP_MAX_TIMESTAMPS);
 	}
 	spin_unlock_irqrestore(&queue->lock, flags);
 
-- 
2.42.0







[Index of Archives]     [Linux Kernel]     [Kernel Development Newbies]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite Hiking]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux