[PATCH] nvmet: fix workqueue MEM_RECLAIM flushing dependency

[PATCH] nvmet: fix workqueue MEM_RECLAIM flushing dependency

[Sagi Grimberg]

The keep alive timer needs to stay on nvmet_wq, and not
modified to reschedule on the system_wq.

This fixes a warning:
------------[ cut here ]------------
workqueue: WQ_MEM_RECLAIM
nvmet-wq:nvmet_rdma_release_queue_work [nvmet_rdma] is flushing
!WQ_MEM_RECLAIM events:nvmet_keep_alive_timer [nvmet]
WARNING: CPU: 3 PID: 1086 at kernel/workqueue.c:2628
check_flush_dependency+0x16c/0x1e0

Reported-by: Yi Zhang <yi.zhang@redhat.com>
Fixes: 8832cf922151 ("nvmet: use a private workqueue instead of the
system workqueue")
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/target/core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c
index 7f4083cf953a..428317fe6da0 100644
--- a/drivers/nvme/target/core.c
+++ b/drivers/nvme/target/core.c
@@ -1175,7 +1175,7 @@ static void nvmet_start_ctrl(struct nvmet_ctrl *ctrl)
 	 * reset the keep alive timer when the controller is enabled.
 	 */
 	if (ctrl->kato)
-		mod_delayed_work(system_wq, &ctrl->ka_work, ctrl->kato * HZ);
+		mod_delayed_work(nvmet_wq, &ctrl->ka_work, ctrl->kato * HZ);
 }
 
 static void nvmet_clear_ctrl(struct nvmet_ctrl *ctrl)
-- 
2.34.1

Re: [PATCH] nvmet: fix workqueue MEM_RECLAIM flushing dependency

[Chaitanya Kulkarni]

> 
> On Sep 27, 2022, at 11:40 PM, Sagi Grimberg <sagi@grimberg.me> wrote:
> 
> The keep alive timer needs to stay on nvmet_wq, and not
> modified to reschedule on the system_wq.
> 
> This fixes a warning:
> ------------[ cut here ]------------
> workqueue: WQ_MEM_RECLAIM
> nvmet-wq:nvmet_rdma_release_queue_work [nvmet_rdma] is flushing
> !WQ_MEM_RECLAIM events:nvmet_keep_alive_timer [nvmet]
> WARNING: CPU: 3 PID: 1086 at kernel/workqueue.c:2628
> check_flush_dependency+0x16c/0x1e0
> 
> Reported-by: Yi Zhang <yi.zhang@redhat.com>
> Fixes: 8832cf922151 ("nvmet: use a private workqueue instead of the
> system workqueue")
> Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
> ---
> 

Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>

-ck

Re: [PATCH] nvmet: fix workqueue MEM_RECLAIM flushing dependency

[Christoph Hellwig]

On Wed, Sep 28, 2022 at 09:39:10AM +0300, Sagi Grimberg wrote:
> The keep alive timer needs to stay on nvmet_wq, and not
> modified to reschedule on the system_wq.

Does it?  The keep alive timeout is required for making progress
under heavy memory pressure.

Re: [PATCH] nvmet: fix workqueue MEM_RECLAIM flushing dependency

[Sagi Grimberg]

>> The keep alive timer needs to stay on nvmet_wq, and not
>> modified to reschedule on the system_wq.
> 
> Does it?  The keep alive timeout is required for making progress
> under heavy memory pressure.

nvmet_wq has WQ_MEM_RECLAIM. Also, note that keep_alive is initially
scheduled in nvmet_wq but mod_delayed_work incorrectly moved it to
the system_wq...

Re: [PATCH] nvmet: fix workqueue MEM_RECLAIM flushing dependency

[Christoph Hellwig]

Thanks,

applied to nvme-6.1.