* [PATCH] dm-raid: fix use-after-free in raid_dtr()
@ 2026-11-12 9:12 Li Youhong
2026-11-12 20:35 ` Mikulas Patocka
0 siblings, 1 reply; 3+ messages in thread
From: Li Youhong @ 2026-11-12 9:12 UTC (permalink / raw)
To: agk, snitzer, mpatocka, bmarzins
Cc: yukuai, dm-devel, linux-kernel, Li Youhong, stable
From: Li Youhong <liyouhong@kylinos.cn>
raid_dtr() currently flushes event_work only if work_pending() is true.
This is insufficient because once a worker has dequeued the work, the
PENDING bit is cleared while do_table_event() is still running. In that
window raid_set_free() can free the raid_set, leading to a
use-after-free when do_table_event() accesses rs->ti->table.
Flush event_work whenever it has been initialized, so raid_set is not
freed while the event handler is still running. The .func guard skips the
flush for mddev instances where INIT_WORK has not run yet.
Fixes: db29d79b34d9 ("dm-raid: delay flushing event_work() after reconfig_mutex is released")
Cc: stable@vger.kernel.org
Signed-off-by: Li Youhong <liyouhong@kylinos.cn>
---
v2:
- Fix incorrect Fixes tag; the work_pending() check in raid_dtr() was
introduced by db29d79b34d9, not a17d1bca5aba.
- v1: https://lore.kernel.org/all/20261112075741.1449168-1-dayou5941@163.com/
drivers/md/dm-raid.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c
index ac019a0f5a92..c1f546e8c65d 100644
--- a/drivers/md/dm-raid.c
+++ b/drivers/md/dm-raid.c
@@ -3340,7 +3340,7 @@ static void raid_dtr(struct dm_target *ti)
rs->md.dm_gendisk = NULL;
mddev_unlock(&rs->md);
- if (work_pending(&rs->md.event_work))
+ if (rs->md.event_work.func)
flush_work(&rs->md.event_work);
raid_set_free(rs);
}
--
2.25.1
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] dm-raid: fix use-after-free in raid_dtr()
2026-11-12 9:12 [PATCH] dm-raid: fix use-after-free in raid_dtr() Li Youhong
@ 2026-11-12 20:35 ` Mikulas Patocka
0 siblings, 0 replies; 3+ messages in thread
From: Mikulas Patocka @ 2026-11-12 20:35 UTC (permalink / raw)
To: Li Youhong
Cc: agk, snitzer, bmarzins, yukuai, dm-devel, linux-kernel,
Li Youhong, stable
OK. I applied the patch, thanks.
Mikulas
On Thu, 12 Nov 2026, Li Youhong wrote:
> From: Li Youhong <liyouhong@kylinos.cn>
>
> raid_dtr() currently flushes event_work only if work_pending() is true.
> This is insufficient because once a worker has dequeued the work, the
> PENDING bit is cleared while do_table_event() is still running. In that
> window raid_set_free() can free the raid_set, leading to a
> use-after-free when do_table_event() accesses rs->ti->table.
>
> Flush event_work whenever it has been initialized, so raid_set is not
> freed while the event handler is still running. The .func guard skips the
> flush for mddev instances where INIT_WORK has not run yet.
>
> Fixes: db29d79b34d9 ("dm-raid: delay flushing event_work() after reconfig_mutex is released")
> Cc: stable@vger.kernel.org
> Signed-off-by: Li Youhong <liyouhong@kylinos.cn>
> ---
> v2:
> - Fix incorrect Fixes tag; the work_pending() check in raid_dtr() was
> introduced by db29d79b34d9, not a17d1bca5aba.
> - v1: https://lore.kernel.org/all/20261112075741.1449168-1-dayou5941@163.com/
>
> drivers/md/dm-raid.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c
> index ac019a0f5a92..c1f546e8c65d 100644
> --- a/drivers/md/dm-raid.c
> +++ b/drivers/md/dm-raid.c
> @@ -3340,7 +3340,7 @@ static void raid_dtr(struct dm_target *ti)
> rs->md.dm_gendisk = NULL;
> mddev_unlock(&rs->md);
>
> - if (work_pending(&rs->md.event_work))
> + if (rs->md.event_work.func)
> flush_work(&rs->md.event_work);
> raid_set_free(rs);
> }
>
> --
> 2.25.1
>
^ permalink raw reply [flat|nested] 3+ messages in thread
* [PATCH] dm-raid: fix use-after-free in raid_dtr()
@ 2026-11-12 7:57 Li Youhong
0 siblings, 0 replies; 3+ messages in thread
From: Li Youhong @ 2026-11-12 7:57 UTC (permalink / raw)
To: agk, snitzer, mpatocka, bmarzins
Cc: dm-devel, linux-kernel, Li Youhong, stable
From: Li Youhong <liyouhong@kylinos.cn>
raid_dtr() currently flushes event_work only if work_pending() is true.
This is insufficient because once a worker has dequeued the work, the
PENDING bit is cleared while do_table_event() is still running. In that
window raid_set_free() can free the raid_set, leading to a
use-after-free when do_table_event() accesses rs->ti->table.
Flush event_work whenever it has been initialized, so raid_set is not
freed while the event handler is still running. The .func guard skips the
flush for mddev instances where INIT_WORK has not run yet.
Fixes: a17d1bca5aba ("md: fix stopping sync thread")
Cc: stable@vger.kernel.org
Signed-off-by: Li Youhong <liyouhong@kylinos.cn>
---
drivers/md/dm-raid.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c
index ac019a0f5a92..c1f546e8c65d 100644
--- a/drivers/md/dm-raid.c
+++ b/drivers/md/dm-raid.c
@@ -3340,7 +3340,7 @@ static void raid_dtr(struct dm_target *ti)
rs->md.dm_gendisk = NULL;
mddev_unlock(&rs->md);
- if (work_pending(&rs->md.event_work))
+ if (rs->md.event_work.func)
flush_work(&rs->md.event_work);
raid_set_free(rs);
}
--
2.25.1
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-11-12 20:36 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-12 9:12 [PATCH] dm-raid: fix use-after-free in raid_dtr() Li Youhong
2026-11-12 20:35 ` Mikulas Patocka
-- strict thread matches above, loose matches on Subject: below --
2026-11-12 7:57 Li Youhong
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®