Re: [PATCH] pipe: read/write_iter() handler should check for IOCB_NOWAIT

[Jens Axboe]

On 4/30/20 10:14 PM, Jens Axboe wrote:
> On 4/30/20 9:58 PM, Al Viro wrote:
>> On Thu, Apr 30, 2020 at 10:24:46AM -0600, Jens Axboe wrote:
>>> Pipe read/write only checks for the file O_NONBLOCK flag, but we should
>>> also check for IOCB_NOWAIT for whether or not we should handle this read
>>> or write in a non-blocking fashion. If we don't, then we will block on
>>> data or space for iocbs that explicitly asked for non-blocking
>>> operation. This messes up callers that explicitly ask for non-blocking
>>> operations.
>>
>> Why does io_uring allow setting IOCB_NOWAIT without FMODE_NOWAIT, anyway?
> 
> To do per-io non-blocking operations. It's not practical or convenient
> to flip the file flag, nor does it work if you have multiple of them
> going. If pipes honor the flag for the read/write iter handlers, then
> we can handle them a lot more efficiently instead of requiring async
> offload.

Sorry, I think I misread you and the answer, while correct by itself, is
not the answer to the question you are asking. You're saying we should
not be using IOCB_NOWAIT if FMODE_NOWAIT isn't set, which is fair. I'll
re-do the patch, we can probably just use FMODE_NOWAIT for the final
check in io_uring instead.

For pipes, we should include the setting of FMODE_NOWAIT for fifo_open()
with the patch, at least.

-- 
Jens Axboe