* [PATCH 2/2] dm-snapshot: suspend merging snapshot when doing exception handover
@ 2015-02-26 16:41 Mikulas Patocka
2015-02-27 19:23 ` Mikulas Patocka
2015-02-27 20:02 ` Mike Snitzer
0 siblings, 2 replies; 3+ messages in thread
From: Mikulas Patocka @ 2015-02-26 16:41 UTC (permalink / raw)
To: Mike Snitzer, Alasdair G. Kergon; +Cc: dm-devel
There was a bug when that resulted in a crash when there were pending
exceptions and snapshot exception store handover was performed at the
same time - and there was a patch that fixed it.
However, a similar problem exists in snapshot merging. When snapshot
merging is in progress, we use the target "snapshot-merge" instead of
"snapshot-origin". Consequently, during exception store handover, we must
find the snapshot-merge target and suspend it's associated md.
To avoid lockdep warnings, the target must be suspended and resumed
without holding _origins_lock.
This patch introduces a function dm_hold that grabs a reference on
mapped_device, but unlike dm_get, it doesn't crash if the devices has the
flag DMF_FREEING, it returns and error in this case.
In snapshot_resume we grab the reference to the origin device using
dm_hold while holding _origins_lock (_origins_lock guarantees that the
device won't disappear). Then we release _origins_lock, suspend the
device and grab _origins_lock again.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Index: linux-2.6-debug/drivers/md/dm-snap.c
===================================================================
--- linux-2.6-debug.orig/drivers/md/dm-snap.c
+++ linux-2.6-debug/drivers/md/dm-snap.c
@@ -1889,20 +1889,39 @@ static int snapshot_preresume(struct dm_
static void snapshot_resume(struct dm_target *ti)
{
struct dm_snapshot *s = ti->private;
- struct dm_snapshot *snap_src = NULL, *snap_dest = NULL;
+ struct dm_snapshot *snap_src = NULL, *snap_dest = NULL, *snap_merging = NULL;
struct dm_origin *o;
struct mapped_device *origin_md = NULL;
+ bool must_restart_merging = false;
down_read(&_origins_lock);
o = __lookup_dm_origin(s->origin->bdev);
if (o)
origin_md = dm_table_get_md(o->ti->table);
+ if (!origin_md) {
+ (void) __find_snapshots_sharing_cow(s, NULL, NULL, &snap_merging);
+ if (snap_merging)
+ origin_md = dm_table_get_md(snap_merging->ti->table);
+ }
if (origin_md == dm_table_get_md(ti->table))
origin_md = NULL;
+ if (origin_md) {
+ if (dm_hold(origin_md))
+ origin_md = NULL;
+ }
+
+ up_read(&_origins_lock);
+
+ if (origin_md) {
+ dm_internal_suspend_fast(origin_md);
+ if (snap_merging && test_bit(RUNNING_MERGE, &snap_merging->state_bits)) {
+ must_restart_merging = true;
+ stop_merge(snap_merging);
+ }
+ }
- if (origin_md)
- dm_internal_suspend_fast(origin_md);
+ down_read(&_origins_lock);
(void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL);
if (snap_src && snap_dest) {
@@ -1913,11 +1932,15 @@ static void snapshot_resume(struct dm_ta
up_write(&snap_src->lock);
}
- if (origin_md)
- dm_internal_resume_fast(origin_md);
-
up_read(&_origins_lock);
+ if (origin_md) {
+ if (must_restart_merging)
+ start_merge(snap_merging);
+ dm_internal_resume_fast(origin_md);
+ dm_put(origin_md);
+ }
+
/* Now we have correct chunk size, reregister */
reregister_snapshot(s);
Index: linux-2.6-debug/include/linux/device-mapper.h
===================================================================
--- linux-2.6-debug.orig/include/linux/device-mapper.h
+++ linux-2.6-debug/include/linux/device-mapper.h
@@ -368,6 +368,7 @@ int dm_create(int minor, struct mapped_d
*/
struct mapped_device *dm_get_md(dev_t dev);
void dm_get(struct mapped_device *md);
+int dm_hold(struct mapped_device *md);
void dm_put(struct mapped_device *md);
/*
Index: linux-2.6-debug/drivers/md/dm.c
===================================================================
--- linux-2.6-debug.orig/drivers/md/dm.c
+++ linux-2.6-debug/drivers/md/dm.c
@@ -2507,6 +2507,19 @@ void dm_get(struct mapped_device *md)
BUG_ON(test_bit(DMF_FREEING, &md->flags));
}
+int dm_hold(struct mapped_device *md)
+{
+ spin_lock(&_minor_lock);
+ if (test_bit(DMF_FREEING, &md->flags)) {
+ spin_unlock(&_minor_lock);
+ return -EBUSY;
+ }
+ dm_get(md);
+ spin_unlock(&_minor_lock);
+ return 0;
+}
+EXPORT_SYMBOL_GPL(dm_hold);
+
const char *dm_device_name(struct mapped_device *md)
{
return md->name;
@@ -2526,10 +2539,16 @@ static void __dm_destroy(struct mapped_d
set_bit(DMF_FREEING, &md->flags);
spin_unlock(&_minor_lock);
+ /*
+ * Take suspend_lock so that presuspend and postsuspend methods
+ * do not race with internal suspend.
+ */
+ mutex_lock(&md->suspend_lock);
if (!dm_suspended_md(md)) {
dm_table_presuspend_targets(map);
dm_table_postsuspend_targets(map);
}
+ mutex_unlock(&md->suspend_lock);
/* dm_put_live_table must be before msleep, otherwise deadlock is possible */
dm_put_live_table(md, srcu_idx);
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH 2/2] dm-snapshot: suspend merging snapshot when doing exception handover
2015-02-26 16:41 [PATCH 2/2] dm-snapshot: suspend merging snapshot when doing exception handover Mikulas Patocka
@ 2015-02-27 19:23 ` Mikulas Patocka
2015-02-27 20:02 ` Mike Snitzer
1 sibling, 0 replies; 3+ messages in thread
From: Mikulas Patocka @ 2015-02-27 19:23 UTC (permalink / raw)
To: Mike Snitzer, Alasdair G. Kergon; +Cc: dm-devel
On Thu, 26 Feb 2015, Mikulas Patocka wrote:
> There was a bug when that resulted in a crash when there were pending
> exceptions and snapshot exception store handover was performed at the
> same time - and there was a patch that fixed it.
>
> However, a similar problem exists in snapshot merging. When snapshot
> merging is in progress, we use the target "snapshot-merge" instead of
> "snapshot-origin". Consequently, during exception store handover, we must
> find the snapshot-merge target and suspend it's associated md.
>
> To avoid lockdep warnings, the target must be suspended and resumed
> without holding _origins_lock.
>
> This patch introduces a function dm_hold that grabs a reference on
> mapped_device, but unlike dm_get, it doesn't crash if the devices has the
> flag DMF_FREEING, it returns and error in this case.
>
> In snapshot_resume we grab the reference to the origin device using
> dm_hold while holding _origins_lock (_origins_lock guarantees that the
> device won't disappear). Then we release _origins_lock, suspend the
> device and grab _origins_lock again.
>
> Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
I forgot to add
Cc: stable@vger.kernel.org
When backporting to kernels 3.18 and older, use dm_internal_suspend and
dm_internal_resume instead of dm_internal_suspend_fast and
dm_internal_resume_fast.
Mikulas
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH 2/2] dm-snapshot: suspend merging snapshot when doing exception handover
2015-02-26 16:41 [PATCH 2/2] dm-snapshot: suspend merging snapshot when doing exception handover Mikulas Patocka
2015-02-27 19:23 ` Mikulas Patocka
@ 2015-02-27 20:02 ` Mike Snitzer
1 sibling, 0 replies; 3+ messages in thread
From: Mike Snitzer @ 2015-02-27 20:02 UTC (permalink / raw)
To: Mikulas Patocka; +Cc: dm-devel, Alasdair G. Kergon
On Thu, Feb 26 2015 at 11:41am -0500,
Mikulas Patocka <mpatocka@redhat.com> wrote:
> There was a bug when that resulted in a crash when there were pending
> exceptions and snapshot exception store handover was performed at the
> same time - and there was a patch that fixed it.
>
> However, a similar problem exists in snapshot merging. When snapshot
> merging is in progress, we use the target "snapshot-merge" instead of
> "snapshot-origin". Consequently, during exception store handover, we must
> find the snapshot-merge target and suspend it's associated md.
>
> To avoid lockdep warnings, the target must be suspended and resumed
> without holding _origins_lock.
>
> This patch introduces a function dm_hold that grabs a reference on
> mapped_device, but unlike dm_get, it doesn't crash if the devices has the
> flag DMF_FREEING, it returns and error in this case.
>
> In snapshot_resume we grab the reference to the origin device using
> dm_hold while holding _origins_lock (_origins_lock guarantees that the
> device won't disappear). Then we release _origins_lock, suspend the
> device and grab _origins_lock again.
>
> Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Staged for 4.0 here (again bumped target version and tweaked header):
https://git.kernel.org/cgit/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-for-4.0&id=09ee96b21456883e108c3b00597bb37ec512151b
> Index: linux-2.6-debug/drivers/md/dm.c
> ===================================================================
> --- linux-2.6-debug.orig/drivers/md/dm.c
> +++ linux-2.6-debug/drivers/md/dm.c
> @@ -2526,10 +2539,16 @@ static void __dm_destroy(struct mapped_d
> set_bit(DMF_FREEING, &md->flags);
> spin_unlock(&_minor_lock);
>
> + /*
> + * Take suspend_lock so that presuspend and postsuspend methods
> + * do not race with internal suspend.
> + */
> + mutex_lock(&md->suspend_lock);
> if (!dm_suspended_md(md)) {
> dm_table_presuspend_targets(map);
> dm_table_postsuspend_targets(map);
> }
> + mutex_unlock(&md->suspend_lock);
>
> /* dm_put_live_table must be before msleep, otherwise deadlock is possible */
> dm_put_live_table(md, srcu_idx);
I split this chunk out to a new commit for 4.0 here:
https://git.kernel.org/cgit/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-for-4.0&id=ab7c7bb6f4ab95dbca96fcfc4463cd69843e3e24
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2015-02-27 20:02 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-02-26 16:41 [PATCH 2/2] dm-snapshot: suspend merging snapshot when doing exception handover Mikulas Patocka
2015-02-27 19:23 ` Mikulas Patocka
2015-02-27 20:02 ` Mike Snitzer
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.