dm cache metadata: save in-core policy_hint_size to on-disk superblock
authorMike Snitzer <snitzer@redhat.com>
Thu, 2 Aug 2018 20:08:52 +0000 (16:08 -0400)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 9 Sep 2018 17:55:55 +0000 (19:55 +0200)
commitb7227e6044be9e0a0617346cd54baa384298b18b
tree45f1cfbf6723e788e8dce8e0c1623bbdccbee012
parent3bef88257145fd4f8abf0c0f92ecdeafbd96739e
dm cache metadata: save in-core policy_hint_size to on-disk superblock

commit fd2fa95416188a767a63979296fa3e169a9ef5ec upstream.

policy_hint_size starts as 0 during __write_initial_superblock().  It
isn't until the policy is loaded that policy_hint_size is set in-core
(cmd->policy_hint_size).  But it never got recorded in the on-disk
superblock because __commit_transaction() didn't deal with transfering
the in-core cmd->policy_hint_size to the on-disk superblock.

The in-core cmd->policy_hint_size gets initialized by metadata_open()'s
__begin_transaction_flags() which re-reads all superblock fields.
Because the superblock's policy_hint_size was never properly stored, when
the cache was created, hints_array_available() would always return false
when re-activating a previously created cache.  This means
__load_mappings() always considered the hints invalid and never made use
of the hints (these hints served to optimize).

Another detremental side-effect of this oversight is the cache_check
utility would fail with: "invalid hint width: 0"

Cc: stable@vger.kernel.org
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/md/dm-cache-metadata.c