uring_cmd implementations should not try to guess issue_flags, use a freshly added helper io_uring_cmd_complete() instead. Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx> Link: https://lore.kernel.org/r/c661cc48f3dd4a09ace5f9d93f5d498cbf3de583.1710514702.git.asml.silence@xxxxxxxxx Signed-off-by: Jens Axboe <axboe@xxxxxxxxx> --- drivers/nvme/host/ioctl.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/drivers/nvme/host/ioctl.c b/drivers/nvme/host/ioctl.c index 3dfd5ae99ae0..1a7b5af42dbc 100644 --- a/drivers/nvme/host/ioctl.c +++ b/drivers/nvme/host/ioctl.c @@ -426,10 +426,13 @@ static enum rq_end_io_ret nvme_uring_cmd_end_io(struct request *req, * For iopoll, complete it directly. * Otherwise, move the completion to task work. */ - if (blk_rq_is_poll(req)) - nvme_uring_task_cb(ioucmd, IO_URING_F_UNLOCKED); - else + if (blk_rq_is_poll(req)) { + if (pdu->bio) + blk_rq_unmap_user(pdu->bio); + io_uring_cmd_complete(ioucmd, pdu->status, pdu->result); + } else { io_uring_cmd_do_in_task_lazy(ioucmd, nvme_uring_task_cb); + } return RQ_END_IO_FREE; } -- 2.44.0
