On 5/12/22 2:44 AM, Ming Lei wrote: > On Thu, May 12, 2022 at 10:09:12AM +0200, Christoph Hellwig wrote: >> On Thu, May 12, 2022 at 01:25:24PM +0800, Ming Lei wrote: >>> This way may cause nested plugging, and breaks xfstests generic/131. >>> Also may cause io hang since request can't be polled before flushing >>> plug in blk_execute_rq(). >> >> Looking at this again, yes blk_mq_request_bypass_insert is probably the >> wrong place. >> >>> I'd suggest to apply the plug in blk_execute_rq_nowait(), such as: >> >> Do we really need the use_plug parameter and the extra helper? If >> someone holds a plug over passthrough command submission I think >> we can assume they actually do want to use it. Otherwise this does >> indeed look like the better plan. > > use_plug is just for avoiding hang in blk_rq_poll_completion(), so > I think we can bypass plug if one polled rq is executed inside > blk_execute_rq(). Agree, and good catch. Do you want to send out a patch for this? -- Jens Axboe
