]> git.ipfire.org Git - thirdparty/kernel/linux.git/commit
md: wake raid456 reshape waiters before suspend
authorYu Kuai <yukuai@fnnas.com>
Fri, 27 Mar 2026 14:07:29 +0000 (22:07 +0800)
committerYu Kuai <yukuai@fnnas.com>
Tue, 7 Apr 2026 05:09:22 +0000 (13:09 +0800)
commitcf86bb53b9c92354904a328e947a05ffbfdd1840
tree542c9f03c0db0d848eef3481d52232e95ecdc313
parent808cec74601cfddea87b6970134febfdc7f574b9
md: wake raid456 reshape waiters before suspend

During raid456 reshape, direct IO across the reshape position can sleep
in raid5_make_request() waiting for reshape progress while still
holding an active_io reference. If userspace then freezes reshape and
writes md/suspend_lo or md/suspend_hi, mddev_suspend() kills active_io
and waits for all in-flight IO to drain.

This can deadlock: the IO needs reshape progress to continue, but the
reshape thread is already frozen, so the active_io reference is never
dropped and suspend never completes.

raid5_prepare_suspend() already wakes wait_for_reshape for dm-raid. Do
the same for normal md suspend when reshape is already interrupted, so
waiting raid456 IO can abort, drop its reference, and let suspend
finish.

The mdadm test tests/25raid456-reshape-deadlock reproduces the hang.

Fixes: 714d20150ed8 ("md: add new helpers to suspend/resume array")
Link: https://lore.kernel.org/linux-raid/20260327140729.2030564-1-yukuai@fnnas.com/
Signed-off-by: Yu Kuai <yukuai@fnnas.com>
drivers/md/md.c