rbd: return errors for mapped but deleted snapshot

When a snapshot is deleted, the OSD will return ENOENT when reading
from it. This is normally interpreted as a hole by rbd, which will
return zeroes. To minimize the time in which this can happen, stop
requests early when we are notified that our snapshot no longer
exists.

[elder@inktank.com: updated __rbd_init_snaps_header() logic]

Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>

diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
index b124442dab3a1156436ae23dc4502084aee03047..730d0ce505e1c4210eca3e198e602bff911a8422 100644 (file)
--- a/drivers/block/rbd.c
+++ b/drivers/block/rbd.c
@@ -172,9 +172,13 @@ struct rbd_device {
 
        /* protects updating the header */
        struct rw_semaphore     header_rwsem;
+       /* name of the snapshot this device reads from */
        char                    *snap_name;
+       /* id of the snapshot this device reads from */
        u64                     snap_id;        /* current snapshot id */
-       int read_only;
+       /* whether the snap_id this device reads from still exists */
+       bool                    snap_exists;
+       int                     read_only;
 
        struct list_head        node;
 
@@ -597,6 +601,7 @@ static int rbd_header_set_snap(struct rbd_device *rbd_dev, u64 *size)
                else
                        snapc->seq = 0;
                rbd_dev->snap_id = CEPH_NOSNAP;
+               rbd_dev->snap_exists = false;
                rbd_dev->read_only = 0;
                if (size)
                        *size = header->image_size;
@@ -606,6 +611,7 @@ static int rbd_header_set_snap(struct rbd_device *rbd_dev, u64 *size)
                if (ret < 0)
                        goto done;
                rbd_dev->snap_id = snapc->seq;
+               rbd_dev->snap_exists = true;
                rbd_dev->read_only = 1;
        }
 
@@ -1468,6 +1474,21 @@ static void rbd_rq_fn(struct request_queue *q)
 
                spin_unlock_irq(q->queue_lock);
 
+               if (rbd_dev->snap_id != CEPH_NOSNAP) {
+                       bool snap_exists;
+
+                       down_read(&rbd_dev->header_rwsem);
+                       snap_exists = rbd_dev->snap_exists;
+                       up_read(&rbd_dev->header_rwsem);
+
+                       if (!snap_exists) {
+                               dout("request for non-existent snapshot");
+                               spin_lock_irq(q->queue_lock);
+                               __blk_end_request_all(rq, -ENXIO);
+                               continue;
+                       }
+               }
+
                dout("%s 0x%x bytes at 0x%llx\n",
                     do_write ? "write" : "read",
                     size, blk_rq_pos(rq) * SECTOR_SIZE);
@@ -2088,7 +2109,14 @@ static int __rbd_init_snaps_header(struct rbd_device *rbd_dev)
                        cur_id = rbd_dev->header.snapc->snaps[i - 1];
 
                if (!i || old_snap->id < cur_id) {
-                       /* old_snap->id was skipped, thus was removed */
+                       /*
+                        * old_snap->id was skipped, thus was
+                        * removed.  If this rbd_dev is mapped to
+                        * the removed snapshot, record that it no
+                        * longer exists, to prevent further I/O.
+                        */
+                       if (rbd_dev->snap_id == old_snap->id)
+                               rbd_dev->snap_exists = false;
                        __rbd_remove_snap_dev(rbd_dev, old_snap);
                        continue;
                }