[PATCH 2/2] nvme-pci: handle persistent internal error AER from NVMe controller

Christoph Hellwig hch at lst.de
Wed Jun 1 10:08:47 PDT 2022


On Wed, Jun 01, 2022 at 03:56:59PM +0000, Michael Kelley (LINUX) wrote:
> If there is a persistent error that does a controller reset, it looks
> like we should *not* queue async_event_work at the end of
> nvme_complete_async_event().  The controller reset will
> submit an AER on the admin queue, and so presumably
> we don't want nvme_async_event_work() to also try to submit
> another AER, which may or may not succeed depending on the
> timing of when the controller state shows LIVE again.
> Agreed?

Yes, that makes sense.  I guess we can just check the return value
from nvme_reset_ctrl and propagate this to nvme_async_event_work
and skip the rearming for that case.



More information about the Linux-nvme mailing list