drbd: avoid potential deadlock during handshake

During handshake communication, we also reconsider our device size,
using drbd_determine_dev_size(). Just in case we need to change the
offsets or layout of our on-disk metadata, we lock out application
and other meta data IO, and wait for the activity log to be "idle"
(no more referenced extents).

If this handshake happens just after a connection loss, with a fencing
policy of "resource-and-stonith", we have frozen IO.

If, additionally, the activity log was "starving" (too many incoming
random writes at that point in time), it won't become idle, ever,
because of the frozen IO, and this would be a lockup of the receiver
thread, and consquentially of DRBD.

Previous logic (re-)initialized with a special "empty" transaction
block, which required the activity log to fully drain first.

Instead, write out some standard activity log transactions.
Using lc_try_lock_for_transaction() instead of lc_try_lock() does not
care about pending activity log references, avoiding the potential
deadlock.

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

1 files changed, 11 insertions, 8 deletions

diff --git a/drivers/block/drbd/drbd_actlog.c b/drivers/block/drbd/drbd_actlog.c
index 4b484ac1d8cb..10459a145062 100644
--- a/drivers/block/drbd/drbd_actlog.c
+++ b/drivers/block/drbd/drbd_actlog.c
@@ -614,21 +614,24 @@ void drbd_al_shrink(struct drbd_device *device)
 	wake_up(&device->al_wait);
 }
 
-int drbd_initialize_al(struct drbd_device *device, void *buffer)
+int drbd_al_initialize(struct drbd_device *device, void *buffer)
 {
 	struct al_transaction_on_disk *al = buffer;
 	struct drbd_md *md = &device->ldev->md;
-	sector_t al_base = md->md_offset + md->al_offset;
 	int al_size_4k = md->al_stripes * md->al_stripe_size_4k;
 	int i;
 
-	memset(al, 0, 4096);
-	al->magic = cpu_to_be32(DRBD_AL_MAGIC);
-	al->transaction_type = cpu_to_be16(AL_TR_INITIALIZED);
-	al->crc32c = cpu_to_be32(crc32c(0, al, 4096));
+	__al_write_transaction(device, al);
+	/* There may or may not have been a pending transaction. */
+	spin_lock_irq(&device->al_lock);
+	lc_committed(device->act_log);
+	spin_unlock_irq(&device->al_lock);
 
-	for (i = 0; i < al_size_4k; i++) {
-		int err = drbd_md_sync_page_io(device, device->ldev, al_base + i * 8, WRITE);
+	/* The rest of the transactions will have an empty "updates" list, and
+	 * are written out only to provide the context, and to initialize the
+	 * on-disk ring buffer. */
+	for (i = 1; i < al_size_4k; i++) {
+		int err = __al_write_transaction(device, al);
 		if (err)
 			return err;
 	}