On Thu, Feb 13, 2025 at 07:32:14AM +0100, Christoph Hellwig wrote: > On Thu, Feb 13, 2025 at 09:29:53AM +0800, Ming Lei wrote: > > Yeah, turns out oops is triggered in initializing req_iterator for > > discard req, and the following patch should be enough: > > How do we end up in blk_rq_map_sg for a discard request here? > dma-mapping doesn't make sense for a non-special pyaload discard > as used by xxen-blkfront, and xen-blkfront also only calls > blk_rq_map_sg from blkif_queue_rw_req and not blkif_queue_discard_req. I think we're probably dealing with a flush command, as that's the only request that doesn't have a bio except for empty passthrough commands. xen-blkfront is a bit weird in calling into these data transfer helpers despite not having data to transfer, but I guess something like your patch to safeguard against it should be fine. But add a comment as well please.
