Re: [PATCHSET 0/2] io_uring: handle short reads internally

[Jens Axboe <axboe@kernel.dk>]

On 8/18/20 10:55 AM, Jeff Moyer wrote:
> Jens Axboe <axboe@kernel.dk> writes:
> 
>> On 8/17/20 1:55 PM, Jeff Moyer wrote:
>>> Jens Axboe <axboe@kernel.dk> writes:
>>>
>>>> On 8/13/20 4:21 PM, Jeff Moyer wrote:
>>>>> Jens Axboe <axboe@kernel.dk> writes:
>>>>>
>>>>>>>>> BTW, what git sha did you run?
>>>>>>>>
>>>>>>>> I do see a failure with dm on that, I'll take a look.
>>>>>>>
>>>>>>> I ran it on a file system atop nvme with 8 poll queues.
>>>>>>>
>>>>>>> liburing head: 9e1d69e078ee51f253a829ff421b17cfc996d158
>>>>>>> linux-block head: ff1353802d86a9d8e40ef1377efb12a1d3000a20
>>>>>>
>>>>>> Fixed it, and actually enabled a further cleanup.
>>>>>
>>>>> Great, thanks!  Did you push that out somewhere?
>>>>
>>>> It's pushed to io_uring-5.9, current sha is:
>>>>
>>>> ee6ac2d3d5cc50d58ca55a5967671c9c1f38b085
>>>>
>>>> FWIW, the issue was just for fixed buffers. It's running through the
>>>> usual testing now.
>>>
>>> OK.  Since it was an unrelated problem, I was expecting a separate
>>> commit for it.  What was the exact issue?  Is it something that needs
>>> backporting to -stable?
>>
>> No, it was a bug in the posted patch, so I just folded in the fix.
> 
> We must be hitting different problems, then.  I just tested your
> 5.7-stable branch (running the test suite from an xfs file system on an
> nvme partition with polling enabled), and the read-write test fails:
> 
> Running test read-write:
> Non-vectored IO not supported, skipping
> cqe res -22, wanted 2048
> test_buf_select_short vec failed
> Test read-write failed with ret 1
> 
> That's with this head: a451911d530075352fbc7ef9bb2df68145a747ad

Not sure what this is, haven't seen that here and my regular liburing
runs include both xfs-on-nvme(with poll queues) as one of the test
points. Seems to me like there's two oddities in the above:

1) Saying that Non-vectored isn't supported, that is not true on 5.7.
   This is due to an -EINVAL return.
2) The test_buf_select_short_vec failure

I'll see if I can reproduce this. Anything special otherwise enabled?
Scheduler on the nvme device? nr_requests? XFS options?

-- 
Jens Axboe