drbd: Fixed a deadlock, probably only affected UP machines

After disconnect (most likely mdev->net_cnt == 0) and we are
still in an unstable state (!drbd_state_is_stable()). When we
get an IO request in drbd_get_max_buffers() (called from
__inc_ap_bio_cond(), called from inc_ap_bio()) we wake up
misc_wait. Misc_wait is also used in inc_ap_bio() to sleep
until the outcome of __inc_ap_bio_cond() changes. => Busy loop!

Solution: Have a dedicated wait queue for get_net_conf() and
put_net_conf().

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>

1 files changed, 2 insertions, 1 deletions

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index e0e08f5e0a76..aa9bb213fe70 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -947,6 +947,7 @@ struct drbd_conf {
         union drbd_state state;
         wait_queue_head_t misc_wait;
         wait_queue_head_t state_wait;  /* upon each state change. */
+        wait_queue_head_t net_cnt_wait;
         unsigned int send_cnt;
         unsigned int recv_cnt;
         unsigned int read_cnt;
@@ -2018,7 +2019,7 @@ static inline void inc_unacked(struct drbd_conf *mdev)
 static inline void put_net_conf(struct drbd_conf *mdev)
 {
         if (atomic_dec_and_test(&mdev->net_cnt))
-                wake_up(&mdev->misc_wait);
+                wake_up(&mdev->net_cnt_wait);
 }
 
 /**