dm snapshot: fix header corruption race on invalidation
If a persistent snapshot fills up, a race can corrupt the on-disk header
which causes a crash on any future attempt to activate the snapshot
(typically while booting). This patch fixes the race.
When the snapshot overflows, __invalidate_snapshot is called, which calls
snapshot store method drop_snapshot. It goes to persistent_drop_snapshot that
calls write_header. write_header constructs the new header in the "area"
location.
Concurrently, an existing kcopyd job may finish, call copy_callback
and commit_exception method, that goes to persistent_commit_exception.
persistent_commit_exception doesn't do locking, relying on the fact that
callbacks are single-threaded, but it can race with snapshot invalidation and
overwrite the header that is just being written while the snapshot is being
invalidated.
The result of this race is a corrupted header being written that can
lead to a crash on further reactivation (if chunk_size is zero in the
corrupted header).
The fix is to use separate memory areas for each.
See the bug: https://bugzilla.redhat.com/show_bug.cgi?id=461506
Cc: stable@kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
diff --git a/drivers/md/dm-snap-persistent.c b/drivers/md/dm-snap-persistent.c
index 2a3d626..5d1a975 100644
--- a/drivers/md/dm-snap-persistent.c
+++ b/drivers/md/dm-snap-persistent.c
@@ -106,6 +106,13 @@
void *zero_area;
/*
+ * An area used for header. The header can be written
+ * concurrently with metadata (when invalidating the snapshot),
+ * so it needs a separate buffer.
+ */
+ void *header_area;
+
+ /*
* Used to keep track of which metadata area the data in
* 'chunk' refers to.
*/
@@ -148,16 +155,27 @@
*/
ps->area = vmalloc(len);
if (!ps->area)
- return r;
+ goto err_area;
ps->zero_area = vmalloc(len);
- if (!ps->zero_area) {
- vfree(ps->area);
- return r;
- }
+ if (!ps->zero_area)
+ goto err_zero_area;
memset(ps->zero_area, 0, len);
+ ps->header_area = vmalloc(len);
+ if (!ps->header_area)
+ goto err_header_area;
+
return 0;
+
+err_header_area:
+ vfree(ps->zero_area);
+
+err_zero_area:
+ vfree(ps->area);
+
+err_area:
+ return r;
}
static void free_area(struct pstore *ps)
@@ -169,6 +187,10 @@
if (ps->zero_area)
vfree(ps->zero_area);
ps->zero_area = NULL;
+
+ if (ps->header_area)
+ vfree(ps->header_area);
+ ps->header_area = NULL;
}
struct mdata_req {
@@ -285,11 +307,11 @@
if (r)
return r;
- r = chunk_io(ps, ps->area, 0, READ, 1);
+ r = chunk_io(ps, ps->header_area, 0, READ, 1);
if (r)
goto bad;
- dh = (struct disk_header *) ps->area;
+ dh = ps->header_area;
if (le32_to_cpu(dh->magic) == 0) {
*new_snapshot = 1;
@@ -339,15 +361,15 @@
{
struct disk_header *dh;
- memset(ps->area, 0, ps->store->chunk_size << SECTOR_SHIFT);
+ memset(ps->header_area, 0, ps->store->chunk_size << SECTOR_SHIFT);
- dh = (struct disk_header *) ps->area;
+ dh = ps->header_area;
dh->magic = cpu_to_le32(SNAP_MAGIC);
dh->valid = cpu_to_le32(ps->valid);
dh->version = cpu_to_le32(ps->version);
dh->chunk_size = cpu_to_le32(ps->store->chunk_size);
- return chunk_io(ps, ps->area, 0, WRITE, 1);
+ return chunk_io(ps, ps->header_area, 0, WRITE, 1);
}
/*
@@ -667,6 +689,8 @@
ps->valid = 1;
ps->version = SNAPSHOT_DISK_VERSION;
ps->area = NULL;
+ ps->zero_area = NULL;
+ ps->header_area = NULL;
ps->next_free = 2; /* skipping the header and first area */
ps->current_committed = 0;