Skip to contentSkip to conversation

Fix several data races in tty

PATCH In review Dmitry Vyukov 2 participants 7 messages Sep 17, 2015 TTY layer and serial drivers

lore
Share & tools
lore
#0 lore raw

v5 cover Fix several data races in tty

v5: Combine the 3 patches into single series

v4: Corrected commit log and patch revision notes

v3: Added code comment re: paired smp barrier in "tty: fix data race in tty_buffer_flush"

v2: Split "tty: fix data race in tty_buffer_flush" from "tty: fix data races on tty_buffer.commit". Removed WRITE_ONCE when updating port->itty in "tty: fix data race in flush_to_ldisc".

Dmitry Vyukov (3):
tty: fix data race in flush_to_ldisc
tty: fix data race in tty_buffer_flush
tty: fix data race on tty_buffer.commit

 drivers/tty/tty_buffer.c | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)
Signature · 1 line
2.6.0.rc0.131.gf624c3d
#1 lore raw

PATCH v5 1/3 tty: fix data race in flush_to_ldisc

flush_to_ldisc reads port->itty and checks that it is not NULL, concurrently release_tty sets port->itty to NULL. It is possible that flush_to_ldisc loads port->itty once, ensures that it is not NULL, but then reloads it again and uses. The second load can already return NULL, which will cause a crash.

Use READ_ONCE to read port->itty.

The data race was found with KernelThreadSanitizer (KTSAN).

---

Changed since first version:
- remove WRITE_ONCE when updating port->itty

---
 drivers/tty/tty_buffer.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/tty/tty_buffer.c +1 -1
index 5a3fa89..23de97d 100644
@@ -467,7 +467,7 @@ static void flush_to_ldisc(struct work_struct *work)
 	struct tty_struct *tty;
 	struct tty_ldisc *disc;
 
-	tty = port->itty;
+	tty = READ_ONCE(port->itty);
 	if (tty == NULL)
 		return;
 
Signature · 1 line
2.6.0.rc0.131.gf624c3d
Reviewed-by on v5 1/3
#2 lore raw

Re: [PATCH v5 1/3] tty: fix data race in flush_to_ldisc

Quoted 9 lines · Dmitry Vyukov

flush_to_ldisc reads port->itty and checks that it is not NULL, concurrently release_tty sets port->itty to NULL. It is possible that flush_to_ldisc loads port->itty once, ensures that it is not NULL, but then reloads it again and uses. The second load can already return NULL, which will cause a crash.

Use READ_ONCE to read port->itty.

The data race was found with KernelThreadSanitizer (KTSAN).

#3 lore raw

PATCH v5 2/3 tty: fix data race in tty_buffer_flush

In reply to Dmitry Vyukov

tty_buffer_flush frees not acquired buffers.
As the result, for example, read of b->size in tty_buffer_free can return garbage value which will lead to a huge buffer hanging in the freelist. This is just the benignest manifestation of freeing of a not acquired object. If the object is passed to kfree, heap can be corrupted.

Acquire visibility over the buffer before freeing it.

The data race was found with KernelThreadSanitizer (KTSAN).

---

v4: Corrected commit log and patch revision notes

v3: Added code comment re: paired smp barrier

v2: Split from 'tty: fix data races on tty_buffer.commit'

---
 drivers/tty/tty_buffer.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/tty/tty_buffer.c +4 -1
index 5a3fa89..a2a8cd0 100644
@@ -242,7 +242,10 @@ void tty_buffer_flush(struct tty_struct *tty, struct tty_ldisc *ld)
 	atomic_inc(&buf->priority);
 
 	mutex_lock(&buf->lock);
-	while ((next = buf->head->next) != NULL) {
+	/* paired w/ release in __tty_buffer_request_room; ensures there are
+	 * no pending memory accesses to the freed buffer
+	 */
+	while ((next = smp_load_acquire(&buf->head->next)) != NULL) {
 		tty_buffer_free(port, buf->head);
 		buf->head = next;
 	}
Signature · 1 line
2.6.0.rc0.131.gf624c3d
Reviewed-by on v5 2/3
#4 lore raw

Re: [PATCH v5 2/3] tty: fix data race in tty_buffer_flush

Quoted 10 lines · Dmitry Vyukov

tty_buffer_flush frees not acquired buffers.
As the result, for example, read of b->size in tty_buffer_free can return garbage value which will lead to a huge buffer hanging in the freelist. This is just the benignest manifestation of freeing of a not acquired object. If the object is passed to kfree, heap can be corrupted.

Acquire visibility over the buffer before freeing it.

The data race was found with KernelThreadSanitizer (KTSAN).

#5 lore raw

PATCH v5 3/3 tty: fix data race on tty_buffer.commit

In reply to Dmitry Vyukov

Race on buffer data happens when newly committed data is picked up by an old flush work in the following scenario:
__tty_buffer_request_room does a plain write of tail->commit, no barriers were executed before that.
At this point flush_to_ldisc reads this new value of commit, and reads buffer data, no barriers in between.
The committed buffer data is not necessary visible to flush_to_ldisc.

Similar bug happens when tty_schedule_flip commits data.

Update commit with smp_store_release and read commit with smp_load_acquire, as it is commit that signals data readiness. This is orthogonal to the existing synchronization on tty_buffer.next, which is required to not dismiss a buffer with unconsumed data.

The data race was found with KernelThreadSanitizer (KTSAN).

---

v3: Added patch revision notes

v2: Removed unrelated changed

---
 drivers/tty/tty_buffer.c | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)
drivers/tty/tty_buffer.c +12 -3
index 5a3fa89..7ed2006 100644
@@ -290,7 +290,10 @@ static int __tty_buffer_request_room(struct tty_port *port, size_t size,
 		if (n != NULL) {
 			n->flags = flags;
 			buf->tail = n;
-			b->commit = b->used;
+			/* paired w/ acquire in flush_to_ldisc(); ensures
+			 * flush_to_ldisc() sees buffer data.
+			 */
+			smp_store_release(&b->commit, b->used);
 			/* paired w/ acquire in flush_to_ldisc(); ensures the
 			 * latest commit value can be read before the head is
 			 * advanced to the next buffer
@@ -393,7 +396,10 @@ void tty_schedule_flip(struct tty_port *port)
 {
 	struct tty_bufhead *buf = &port->buf;
 
-	buf->tail->commit = buf->tail->used;
+	/* paired w/ acquire in flush_to_ldisc(); ensures
+	 * flush_to_ldisc() sees buffer data.
+	 */
+	smp_store_release(&buf->tail->commit, buf->tail->used);
 	schedule_work(&buf->work);
 }
 EXPORT_SYMBOL(tty_schedule_flip);
@@ -491,7 +497,10 @@ static void flush_to_ldisc(struct work_struct *work)
 		 * is advancing to the next buffer
 		 */
 		next = smp_load_acquire(&head->next);
-		count = head->commit - head->read;
+		/* paired w/ release in __tty_buffer_request_room() or in
+		 * tty_buffer_flush(); ensures we see the committed buffer data
+		 */
+		count = smp_load_acquire(&head->commit) - head->read;
 		if (!count) {
 			if (next == NULL) {
 				check_other_closed(tty);
Signature · 1 line
2.6.0.rc0.131.gf624c3d
Reviewed-by on v5 3/3
#6 lore raw

Re: [PATCH v5 3/3] tty: fix data race on tty_buffer.commit

Quoted 8 lines · Dmitry Vyukov

Race on buffer data happens when newly committed data is picked up by an old flush work in the following scenario:
__tty_buffer_request_room does a plain write of tail->commit, no barriers were executed before that.
At this point flush_to_ldisc reads this new value of commit, and reads buffer data, no barriers in between.
The committed buffer data is not necessary visible to flush_to_ldisc.

Quoted 8 lines

Similar bug happens when tty_schedule_flip commits data.

Update commit with smp_store_release and read commit with smp_load_acquire, as it is commit that signals data readiness. This is orthogonal to the existing synchronization on tty_buffer.next, which is required to not dismiss a buffer with unconsumed data.

The data race was found with KernelThreadSanitizer (KTSAN).