[dm-devel] [PATCH] dm: Avoid flush_scheduled_work() usage

[Tetsuo Handa]

Please see commit c4f135d643823a86 ("workqueue: Wrap flush_workqueue()
using a macro") for background.

What should we do for this module?

On 2022/04/20 14:12, Tetsuo Handa wrote:
> Flushing system-wide workqueues is dangerous and will be forbidden.
> Remove flush_scheduled_work() from local_init() in dm.c and instead
> use local workqueue to dm-mpath.c, dm-raid1.c, dm-stripe.c.
> 
> Link: https://lkml.kernel.org/r/49925af7-78a8-a3dd-bce6-cfc02e1a9236@I-love.SAKURA.ne.jp
> Signed-off-by: Tetsuo Handa <penguin-kernel at I-love.SAKURA.ne.jp>
> ---
> Note: This patch was made blindly and completely untested. Maybe simply
> removing flush_scheduled_work() from local_init() is sufficient. But I
> don't know dependency across all files. Therefore, please carefully check
> which schedule_work() needs to be converted to queue_work().
> 
> As far as I can see, dm-stripe.c which contains dm_stripe_exit() which is
> called via _exits[] before local_exit() calls flush_scheduled_work() is
> the only file which calls schedule_work(). Therefore, I used dm_stripe_wq
> for dm-stripe.c.
> 
> Since I don't know dependency, and dm-raid1.c and dm-mpath.c are also calling
> schedule_work(), I used dm_raid1_wq and dm_mpath_wq. But these changes might be
> unnecessary because of not calling flush_scheduled_work().
> 
>  drivers/md/dm-mpath.c  | 17 +++++++++++++----
>  drivers/md/dm-raid1.c  | 14 +++++++++++---
>  drivers/md/dm-stripe.c | 12 ++++++++++--
>  drivers/md/dm.c        |  1 -
>  4 files changed, 34 insertions(+), 10 deletions(-)
>