Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [Bug] fio hang when running multiple job io_uring/hipri over nvme
From:   Jens Axboe <axboe@kernel.dk>
To:     Ming Lei <ming.lei@redhat.com>, linux-kernel@vger.kernel.org,
        io-uring@vger.kernel.org
References: <CAFj5m9+ckHjfMVW_O20NBAPvnauPdABa8edPy--dSEf=XdhYRA@mail.gmail.com>
 <6691cf72-3a26-a1bb-228d-ddec8391620f@kernel.dk>
Message-ID: <1b56a4f7-ce56-ee32-67d5-0fcd5dc6c0cb@kernel.dk>
Date:   Thu, 17 Jun 2021 10:56:53 -0600
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.10.0
MIME-Version: 1.0
In-Reply-To: <6691cf72-3a26-a1bb-228d-ddec8391620f@kernel.dk>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 6/17/21 10:48 AM, Jens Axboe wrote:
> On 6/17/21 5:17 AM, Ming Lei wrote:
>> Hello,
>>
>> fio hangs when running the test[1], and doesn't observe this issue
>> when running a
>> such single job test.
>>
>> v5.12 is good, both v5.13-rc3 and the latest v5.13-rc6 are bad.
>>
>>
>> [1] fio test script and log
>> + fio --bs=4k --ioengine=io_uring --fixedbufs --registerfiles --hipri
>> --iodepth=64 --iodepth_batch_submit=16
>> --iodepth_batch_complete_min=16 --filename=/dev/nvme0n1 --direct=1
>> --runtime=20 --numjobs=4 --rw=randread
>> --name=test --group_reporting
>>
>> test: (g=0): rw=randread, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T)
>> 4096B-4096B, ioengine=io_uring, iodepth=64
>> ...
>> fio-3.25
>> Starting 4 processes
>> fio: filehash.c:64: __lookup_file_hash: Assertion `f->fd != -1' failed.
>> fio: pid=1122, got signal=6
>> ^Cbs: 3 (f=0): [f(1),r(1),K(1),r(1)][63.6%][eta 00m:20s]
> 
> Funky, would it be possible to bisect this? I'll see if I can reproduce.

Actually, this looks like a fio bug, that assert is a bit too trigger
happy. Current -git should work, please test and see if things work.
I believe it's just kernel timing that causes this, not a kernel issue.

-- 
Jens Axboe