Re: [PATCH] io_uring: uring_cmd: add multishot support without poll

[Ming Lei <ming.lei@redhat.com>]

On Tue, Aug 19, 2025 at 08:43:58AM -0600, Jens Axboe wrote:
> On 8/19/25 8:40 AM, Ming Lei wrote:
> > On Tue, Aug 19, 2025 at 08:31:32AM -0600, Jens Axboe wrote:
> >> On 8/19/25 8:28 AM, Ming Lei wrote:
> >>> On Tue, Aug 19, 2025 at 08:01:18AM -0600, Jens Axboe wrote:
> >>>> On 8/19/25 5:19 AM, Ming Lei wrote:
> >>>>>>> @@ -251,6 +264,11 @@ int io_uring_cmd(struct io_kiocb *req, unsigned int issue_flags)
> >>>>>>>  	}
> >>>>>>>  
> >>>>>>>  	ret = file->f_op->uring_cmd(ioucmd, issue_flags);
> >>>>>>> +	if (ioucmd->flags & IORING_URING_CMD_MULTISHOT) {
> >>>>>>> +		if (ret >= 0)
> >>>>>>> +			return IOU_ISSUE_SKIP_COMPLETE;
> >>>>>>> +		io_kbuf_recycle(req, issue_flags);
> >>>>>>> +	}
> >>>>>>>  	if (ret == -EAGAIN) {
> >>>>>>>  		ioucmd->flags |= IORING_URING_CMD_REISSUE;
> >>>>>>>  		return ret;
> >>>>>>
> >>>>>> Missing recycle for -EAGAIN?
> >>>>>
> >>>>> io_kbuf_recycle() is done above if `ret < 0`
> >>>>
> >>>> Inside the multishot case. I don't see anywhere where it's forbidden to
> >>>> use IOSQE_BUFFER_SELECT without having multishot set? Either that needs
> >>>
> >>> REQ_F_BUFFER_SELECT is supposed to be allowed for IORING_URING_CMD_MULTISHOT
> >>> only, and it is checked in io_uring_cmd_prep().
> >>>
> >>>> to be explicit for now, or the recycling should happen generically.
> >>>> Probably the former I would suspect.
> >>>
> >>> Yes, the former is exactly what the patch is doing.
> >>
> >> Is it? Because looking at v2, you check if IORING_URING_CMD_FIXED is
> >> set, and you fail for that case if REQ_F_BUFFER_SELECT is set. Then you
> >> have a IORING_URING_CMD_MULTISHOT where the opposite is true, which
> >> obviously makes sense.
> >>
> >> But no checks if neither is set?
> > 
> > Indeed, thanks for the catch, and the REQ_F_BUFFER_SELECT check in IORING_URING_CMD_FIXED
> > branch can be moved to the branch of !IORING_URING_CMD_MULTISHOT.
> > 
> >>
> >> You could add that in io_uring_cmd_select_buffer(), eg fail if
> >> IORING_URING_CMD_MULTISHOT isn't set. Which if done, then the prep side
> >> checking could probably just go away.
> > 
> > Looks this way is good too.
> 
> Want to spin a v3 for this?

Yeah, will send V3 out after test is done.

Thanks,
Ming