Greg,
I know it's late in the -rc cycle but I'd like to get this fix into 3.13.
Although it's only likely to happen at shutdown/reboot, the hang frequency
could be as often as 1 in 10000.
The patch fixes the count value returned when the cmpxchg() has
successfully changed the count. Only one code path checks the
returned count when the cmpxchg() is successful; down_read_failed().
After failed down_read attempt is reversed but before the reader waits
for the lock, the new count is checked to ensure _someone_ has the lock:
/* if there are no active locks, wake the new lock owner(s) */
if ((count & LDSEM_ACTIVE_MASK) == 0)
__ldsem_wake(sem);
Because ldsem_cmpxchg() was returning the _old_ value on success, this
was checking the wrong count value.
No other code paths are impacted by the patch. The equivalent diff below
also fixes the problem; however, I feel the intent is less clear.
| diff --git a/drivers/tty/tty_ldsem.c b/drivers/tty/tty_ldsem.c
| index 22fad8a..29d9e7c 100644
| --- a/drivers/tty/tty_ldsem.c
| +++ b/drivers/tty/tty_ldsem.c
| @@ -222,7 +222,7 @@ down_read_failed(struct ld_semaphore *sem, long count,
long timeout)
| get_task_struct(tsk);
|
| /* if there are no active locks, wake the new lock owner(s) */
| - if ((count & LDSEM_ACTIVE_MASK) == 0)
| + if ((count + adjust & LDSEM_ACTIVE_MASK) == 0)
| __ldsem_wake(sem);
|
| raw_spin_unlock_irq(&sem->wait_lock);
Regards,
Peter Hurley (1):
tty: Fix hang at ldsem_down_read()
drivers/tty/tty_ldsem.c | 16 +++++++++++++---
1 file changed, 13 insertions(+), 3 deletions(-)
--
1.8.1.2
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
