Skip to contentSkip to conversation

tty: pty: don't flush line discipline of another side in pty_flush_buffer

PATCH New Wang YanQing 1 participant 1 message Feb 6, 2017 TTY layer and serial drivers

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tty: pty: don't flush line discipline of another side in pty_flush_buffer

Don't flush line discipline of another side, because user could already see the data in line discipline through poll, TIOCINQ or FIONREAD. Current behavior break the expection of userspace application in below situations.

Example1:
Thread A Thread B
-------- --------
n_tty_poll return POLLIN
CTRL-C trigger pty_flush_buffer
tty_buffer_flush
n_tty_flush_buffer
TIOCINQ or FIONREAD
return zero number of bytes

Example2:
Thread A Thread B
-------- --------
TIOCINQ or FIONREAD
return valid number of bytes

CTRL-C trigger pty_flush_buffer
tty_buffer_flush
n_tty_flush_buffer
n_tty_read still block due to no data

I meet this problem(Example1) in konsole, current behavior cause konsole to hangup forever.

It has high possibility to trigger the trouble I meet to execute below commands a few times in konsole with kernel has commit 1d1d14da12e79a6c05fbe1a975401f0f56c93316 ("pty: Fix buffer flush deadlock") which bring this behavior:
1:cat BigFile
2:CTRL-C

---
 drivers/tty/pty.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)
drivers/tty/pty.c +1 -6
index a23fa5e..2b90738 100644
@@ -216,16 +216,11 @@ static int pty_signal(struct tty_struct *tty, int sig)
 static void pty_flush_buffer(struct tty_struct *tty)
 {
 	struct tty_struct *to = tty->link;
-	struct tty_ldisc *ld;
 
 	if (!to)
 		return;
 
-	ld = tty_ldisc_ref(to);
-	tty_buffer_flush(to, ld);
-	if (ld)
-		tty_ldisc_deref(ld);
-
+	tty_buffer_flush(to, NULL);
 	if (to->packet) {
 		spin_lock_irq(&tty->ctrl_lock);
 		tty->ctrl_status |= TIOCPKT_FLUSHWRITE;
Signature · 1 line
1.8.5.6.2.g3d8a54e.dirty