dm space map metadata: fix extending the space map
authorJoe Thornber <ejt@redhat.com>
Tue, 7 Jan 2014 15:49:02 +0000 (15:49 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 13 Feb 2014 21:48:01 +0000 (13:48 -0800)
commit 7e664b3dec431eebf0c5df5ff704d6197634cf35 upstream.

When extending a metadata space map we should do the first commit whilst
still in bootstrap mode -- a mode where all blocks get allocated in the
new area.

That way the commit overhead is allocated from the newly added space.
Otherwise we risk running out of space.

With this fix, and the previous commit "dm space map common: make sure
new space is used during extend", the following device mapper testsuite
test passes:
 dmtest run --suite thin-provisioning -n /resize_metadata_no_io/

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/md/persistent-data/dm-space-map-metadata.c

index 58fc1ee..f04d92f 100644 (file)
@@ -608,20 +608,28 @@ static int sm_metadata_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
         * Flick into a mode where all blocks get allocated in the new area.
         */
        smm->begin = old_len;
-       memcpy(&smm->sm, &bootstrap_ops, sizeof(smm->sm));
+       memcpy(sm, &bootstrap_ops, sizeof(*sm));
 
        /*
         * Extend.
         */
        r = sm_ll_extend(&smm->ll, extra_blocks);
+       if (r)
+               goto out;
+
+       for (i = old_len; !r && i < smm->begin; i++) {
+               r = sm_ll_inc(&smm->ll, i, &ev);
+               if (r)
+                       goto out;
+       }
 
+       r = sm_metadata_commit(sm);
+
+out:
        /*
         * Switch back to normal behaviour.
         */
-       memcpy(&smm->sm, &ops, sizeof(smm->sm));
-       for (i = old_len; !r && i < smm->begin; i++)
-               r = sm_ll_inc(&smm->ll, i, &ev);
-
+       memcpy(sm, &ops, sizeof(*sm));
        return r;
 }