ceph: avoid putting the realm twice when decoding snaps fails
authorXiubo Li <xiubli@redhat.com>
Wed, 9 Nov 2022 03:00:39 +0000 (11:00 +0800)
committerIlya Dryomov <idryomov@gmail.com>
Mon, 14 Nov 2022 09:29:05 +0000 (10:29 +0100)
When decoding the snaps fails it maybe leaving the 'first_realm'
and 'realm' pointing to the same snaprealm memory. And then it'll
put it twice and could cause random use-after-free, BUG_ON, etc
issues.

Cc: stable@vger.kernel.org
Link: https://tracker.ceph.com/issues/57686
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
fs/ceph/snap.c

index 864cdaa..e415185 100644 (file)
@@ -763,7 +763,7 @@ int ceph_update_snap_trace(struct ceph_mds_client *mdsc,
        struct ceph_mds_snap_realm *ri;    /* encoded */
        __le64 *snaps;                     /* encoded */
        __le64 *prior_parent_snaps;        /* encoded */
-       struct ceph_snap_realm *realm = NULL;
+       struct ceph_snap_realm *realm;
        struct ceph_snap_realm *first_realm = NULL;
        struct ceph_snap_realm *realm_to_rebuild = NULL;
        int rebuild_snapcs;
@@ -774,6 +774,7 @@ int ceph_update_snap_trace(struct ceph_mds_client *mdsc,
 
        dout("%s deletion=%d\n", __func__, deletion);
 more:
+       realm = NULL;
        rebuild_snapcs = 0;
        ceph_decode_need(&p, e, sizeof(*ri), bad);
        ri = p;