[PATCH 04/47] block: provide a new BLK_EH_QUIESCED timeout return value

[Christoph Hellwig]

On Tue, Nov 24, 2015 at 01:12:52PM -0500, Jeff Moyer wrote:
> >> blk_complete_request():
> >>         if (!blk_mark_rq_complete(req) ||  // this fails, as it's already marked complete
> >>             test_and_clear_bit(REQ_ATOM_QUIESCED, &req->atomic_flags))  // this succeeds
> >
> > and clears the flag, so we'd need a race betweem this call to
> > blk_mq_complete request and the later completion of all outstanding
> > commands from reset.  For NVMe we ensure this by not taking completions
> 
> But we're completing the request, so the request will actually be freed.
> Any references to this request are bogus at this point, no?

For blk-mq it won't be freed.  For the non blk-mq case the tag
to request lookup will have to handle that case, but that path isn't
exercised by nvme.