[PATCH v3 17/23] tty: Document lock requirements to halt ldisc

From: Peter Hurley
Date: Tue Feb 05 2013 - 15:45:31 EST


Although the tty_lock() - aka legacy_mutex - protects concurrent
close/re-open of the tty->ldisc, the lock cannot be held while
waiting for the line discipline to halt. The ldisc_mutex acts as
a surrogate lock to prevent the ldisc close while waiting for halt.
Unfortunately, to reacquire the tty_lock() requires dropping the
ldisc_mutex to maintain lock order. This accounts for the complicated
sequence of:

/* holding tty_lock() for atomic tty operation */

mutex_lock();
tty_unlock();
tty_ldisc_halt();
mutex_unlock();

tty_lock();
mutex_lock();

/* check for tty state change which may have occurred
while tty_lock() was dropped */

/* close old ldisc & open new ldisc */

tty_ldisc_enable();

mutex_unlock();
tty_unlock();

tty_ldisc_release() is exempt from locking the halt, as the only
other possible concurrent user, tty_ldisc_hangup(), checks tty
state to prevent this concurrent use.

Signed-off-by: Peter Hurley <peter@xxxxxxxxxxxxxxxxxx>
---
drivers/tty/tty_ldisc.c | 11 +++++++----
1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/drivers/tty/tty_ldisc.c b/drivers/tty/tty_ldisc.c
index c2837b2..9599b05 100644
--- a/drivers/tty/tty_ldisc.c
+++ b/drivers/tty/tty_ldisc.c
@@ -505,6 +505,9 @@ static void tty_ldisc_restore(struct tty_struct *tty, struct tty_ldisc *old)
*
* Wait for the line discipline to become idle. The discipline must
* have been halted for this to guarantee it remains idle.
+ *
+ * Callers other than tty_ldisc_release() must hold ->ldisc_mutex and
+ * cannot be holding tty_lock() (aka legacy_mutex).
*/
static int tty_ldisc_wait_idle(struct tty_struct *tty, long timeout)
{
@@ -533,6 +536,9 @@ static int tty_ldisc_wait_idle(struct tty_struct *tty, long timeout)
* released, which in turn, guarantees that no future buffer work
* can be rescheduled.
*
+ * Callers other than tty_ldisc_release() must hold ->ldisc_mutex and
+ * cannot be holding tty_lock() (aka legacy_mutex).
+ *
* You need to do a 'flush_scheduled_work()' (outside the ldisc_mutex)
* in order to make sure any currently executing ldisc work is also
* flushed.
@@ -713,10 +719,7 @@ int tty_set_ldisc(struct tty_struct *tty, int ldisc)

retval = tty_ldisc_halt(tty, o_tty, &work, &o_work, 5 * HZ);

- /*
- * Wait for hangup to complete, if pending.
- * We must drop the mutex here in case a hangup is also in process.
- */
+ /* Wait for hangup to complete, if pending. */

mutex_unlock(&tty->ldisc_mutex);

--
1.8.1.2

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/