[PATCH] nvmet: Don't queue fatal error work if csts.cfs is set

[PATCH] nvmet: Don't queue fatal error work if csts.cfs is set

[Sagi Grimberg]

In the transport, in case of an interal queue error like
error completion in rdma we trigger a fatal error. However,
multiple queues in the same controller can serr error completions
and we don't want to trigger fatal error work more than once.

Signed-off-by: Sagi Grimberg <sagi at grimberg.me>
---
 drivers/nvme/target/core.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c
index 4d5be2de25f5..ffc8445a1ac3 100644
--- a/drivers/nvme/target/core.c
+++ b/drivers/nvme/target/core.c
@@ -840,7 +840,9 @@ static void nvmet_fatal_error_handler(struct work_struct *work)
 
 void nvmet_ctrl_fatal_error(struct nvmet_ctrl *ctrl)
 {
-	ctrl->csts |= NVME_CSTS_CFS;
+	if (test_and_set_bit(NVME_CSTS_CFS, (unsigned long *)&ctrl->csts))
+		return;
+
 	INIT_WORK(&ctrl->fatal_err_work, nvmet_fatal_error_handler);
 	schedule_work(&ctrl->fatal_err_work);
 }
-- 
2.7.4

[PATCH] nvmet: Don't queue fatal error work if csts.cfs is set

[Christoph Hellwig]

>  {
> -	ctrl->csts |= NVME_CSTS_CFS;
> +	if (test_and_set_bit(NVME_CSTS_CFS, (unsigned long *)&ctrl->csts))
> +		return;
> +

This can't work - test_and_set_bit takes a bit index and NVME_CSTS_CFS
is the actual value.  I think we'll need a lock to protect ->csts
instead.

[PATCH] nvmet: Don't queue fatal error work if csts.cfs is set

[Sagi Grimberg]

>>  {
>> -	ctrl->csts |= NVME_CSTS_CFS;
>> +	if (test_and_set_bit(NVME_CSTS_CFS, (unsigned long *)&ctrl->csts))
>> +		return;
>> +
>
> This can't work - test_and_set_bit takes a bit index and NVME_CSTS_CFS
> is the actual value.  I think we'll need a lock to protect ->csts
> instead.

Oops... it still prevents from fatal_err work from requeueing
so at least I got 50% :)

I'll send a proper patch with a lock in place...

[PATCH] nvmet: Don't queue fatal error work if csts.cfs is set

[Steve Wise]

> >>  {
> >> -	ctrl->csts |= NVME_CSTS_CFS;
> >> +	if (test_and_set_bit(NVME_CSTS_CFS, (unsigned long *)&ctrl->csts))
> >> +		return;
> >> +
> >
> > This can't work - test_and_set_bit takes a bit index and NVME_CSTS_CFS
> > is the actual value.  I think we'll need a lock to protect ->csts
> > instead.
> 
> Oops... it still prevents from fatal_err work from requeueing
> so at least I got 50% :)
> 
> I'll send a proper patch with a lock in place...
> 

I wonder if this is causing my most recent target crashes during kato
reconnect/recovery testing?

[PATCH] nvmet: Don't queue fatal error work if csts.cfs is set

[J Freyensee]

On Fri, 2016-11-04@01:01 +0100, Christoph Hellwig wrote:
> > 
> > ?{
> > -	ctrl->csts |= NVME_CSTS_CFS;
> > +	if (test_and_set_bit(NVME_CSTS_CFS, (unsigned long
> > *)&ctrl->csts))
> > +		return;
> > +
> 
> This can't work - test_and_set_bit takes a bit index and
> NVME_CSTS_CFS
> is the actual value.??I think we'll need a lock to protect ->csts
> instead.

Yeah, I agree, it feels more like a lock solution.

> 
> 
> _______________________________________________
> Linux-nvme mailing list
> Linux-nvme at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-nvme