drbd: Consider the disk-timeout also for meta-data IO operations

If the backing device is already frozen during attach, we failed
to recognize that. The current disk-timeout code works on top
of the drbd_request objects. During attach we do not allow IO
and therefore never generate a drbd_request object but block
before that in drbd_make_request().

This patch adds the timeout to all drbd_md_sync_page_io().

Before this patch we used to go from D_ATTACHING directly
to D_DISKLESS if IO failed during attach. We can no longer
do this since we have to stay in D_FAILED until all IO
ops issued to the backing device returned.

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

drivers/block/drbd/drbd_actlog.c

@@ -88,9 +88,16 @@ static bool md_io_allowed(struct drbd_conf *mdev)
 	return ds >= D_NEGOTIATING || ds == D_ATTACHING;
 }
 
-void wait_until_done_or_disk_failure(struct drbd_conf *mdev, unsigned int *done)
+void wait_until_done_or_disk_failure(struct drbd_conf *mdev, struct drbd_backing_dev *bdev,
+				     unsigned int *done)
 {
-	wait_event(mdev->misc_wait, *done || !md_io_allowed(mdev));
+	long dt = bdev->dc.disk_timeout * HZ / 10;
+	if (dt == 0)
+		dt = MAX_SCHEDULE_TIMEOUT;
+
+	dt = wait_event_timeout(mdev->misc_wait, *done || !md_io_allowed(mdev), dt);
+	if (dt == 0)
+		dev_err(DEV, "meta-data IO operation timed out\n");
 }
 
 static int _drbd_md_sync_page_io(struct drbd_conf *mdev,
@@ -130,7 +137,7 @@ static int _drbd_md_sync_page_io(struct drbd_conf *mdev,
 		bio_endio(bio, -EIO);
 	else
 		submit_bio(rw, bio);
-	wait_until_done_or_disk_failure(mdev, &mdev->md_io.done);
+	wait_until_done_or_disk_failure(mdev, bdev, &mdev->md_io.done);
 	ok = bio_flagged(bio, BIO_UPTODATE) && mdev->md_io.error == 0;
 
  out:

drivers/block/drbd/drbd_bitmap.c

@@ -1088,7 +1088,7 @@ static int bm_rw(struct drbd_conf *mdev, int rw, unsigned lazy_writeout_upper_id
 	 * "in_flight reached zero, all done" event.
 	 */
 	if (!atomic_dec_and_test(&ctx->in_flight))
-		wait_until_done_or_disk_failure(mdev, &ctx->done);
+		wait_until_done_or_disk_failure(mdev, mdev->ldev, &ctx->done);
 	else
 		kref_put(&ctx->kref, &bm_aio_ctx_destroy);
 
@@ -1195,7 +1195,7 @@ int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(loc
 	}
 
 	bm_page_io_async(ctx, idx, WRITE_SYNC);
-	wait_until_done_or_disk_failure(mdev, &ctx->done);
+	wait_until_done_or_disk_failure(mdev, mdev->ldev, &ctx->done);
 
 	if (ctx->error)
 		drbd_chk_io_error(mdev, 1, true);

drivers/block/drbd/drbd_int.h

@@ -1541,7 +1541,8 @@ extern void *drbd_md_get_buffer(struct drbd_conf *mdev);
 extern void drbd_md_put_buffer(struct drbd_conf *mdev);
 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
 				struct drbd_backing_dev *bdev, sector_t sector, int rw);
-extern void wait_until_done_or_disk_failure(struct drbd_conf *mdev, unsigned int *done);
+extern void wait_until_done_or_disk_failure(struct drbd_conf *mdev, struct drbd_backing_dev *bdev,
+					    unsigned int *done);
 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
 extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
 

drivers/block/drbd/drbd_main.c

@@ -916,11 +916,6 @@ static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state
 	if (ns.disk == D_FAILED && os.disk == D_DISKLESS)
 		ns.disk = D_DISKLESS;
 
-	/* if we are only D_ATTACHING yet,
-	 * we can (and should) go directly to D_DISKLESS. */
-	if (ns.disk == D_FAILED && os.disk == D_ATTACHING)
-		ns.disk = D_DISKLESS;
-
 	/* After C_DISCONNECTING only C_STANDALONE may follow */
 	if (os.conn == C_DISCONNECTING && ns.conn != C_STANDALONE)
 		ns.conn = os.conn;
@@ -1592,11 +1587,13 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os,
 	/* first half of local IO error, failure to attach,
 	 * or administrative detach */
 	if (os.disk != D_FAILED && ns.disk == D_FAILED) {
-		enum drbd_io_error_p eh;
-		int was_io_error;
+		enum drbd_io_error_p eh = EP_PASS_ON;
+		int was_io_error = 0;
 		/* corresponding get_ldev was in __drbd_set_state, to serialize
-		 * our cleanup here with the transition to D_DISKLESS,
-		 * so it is safe to dreference ldev here. */
+		 * our cleanup here with the transition to D_DISKLESS.
+		 * But is is still not save to dreference ldev here, since
+		 * we might come from an failed Attach before ldev was set. */
+		if (mdev->ldev) {
 			eh = mdev->ldev->dc.on_io_error;
 			was_io_error = test_and_clear_bit(WAS_IO_ERROR, &mdev->flags);
 
@@ -1621,6 +1618,7 @@ static void after_state_ch(struct drbd_conf *mdev, union drbd_state os,
 			 * this may be the last chance.
 			 * Following put_ldev may transition to D_DISKLESS. */
 			drbd_md_sync(mdev);
+		}
 		put_ldev(mdev);
 
 		if (was_io_error && eh == EP_CALL_HELPER)