Received-SPF: pass (google.com: best guess record for domain of linux-nfs-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: NFS issues about aio and dio test
To:     "J. Bruce Fields" <bfields@fieldses.org>
CC:     <linux-nfs@vger.kernel.org>, <cuiyue-fnst@cn.fujitsu.com>
References: <975395cc-62f2-843f-cc71-82339b2869cd@cn.fujitsu.com>
 <65ea4ea2-548f-1546-059a-af901bba2e87@cn.fujitsu.com>
 <20190828172849.GA29148@fieldses.org>
From:   Su Yanjun <suyj.fnst@cn.fujitsu.com>
Message-ID: <25e5e2a0-c95a-acd7-04da-d48e8400efe3@cn.fujitsu.com>
Date:   Mon, 2 Sep 2019 16:37:20 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:60.0) Gecko/20100101
 Thunderbird/60.7.0
MIME-Version: 1.0
In-Reply-To: <20190828172849.GA29148@fieldses.org>
Content-Type: text/plain; charset="utf-8"; format=flowed
Content-Transfer-Encoding: 8bit
Sender: linux-nfs-owner@vger.kernel.org
Precedence: bulk


在 2019/8/29 1:28, J. Bruce Fields 写道:
> On Mon, Aug 26, 2019 at 08:37:41AM +0800, Su Yanjun wrote:
>> Any ping?
>>
>> 在 2019/8/6 14:08, Su Yanjun 写道:
>>> Hi,
>>>
>>> When I tested xfstests generic/465 with NFS, there was something
>>> unexpected.
>>>
>>> When memory of NFS server was 10G, test passed.
>>> But when memory of NFS server was 4G, test failed.
>>>
>>> Fail message was as below.
>>>      non-aio dio test
>>>      encounter an error: block 4 offset 0, content 62
>>>      aio-dio test
>>>      encounter an error: block 1 offset 0, content 62
>>>
>>> All of the NFS versions(v3 v4.0 v4.1 v4.2) have  this problem.
>>> Maybe something is wrong about NFS's I/O operation.
>>>
>>> Thanks in advance.
> Off the top of my head it doesn't look familiar.
>
> What kernel version are you running on client and server?
client:4.18
server:4.18 (latest version:5.3.0-rc6+ also has this problem)
>
> Did this test previously pass on an older kernel (so, was this a recent
> regression?)
This test previously passed on 3.10 kernel.
> Have you looked at generic/465 to see what exactly is happening here?

generic/465 tests for  aio and dio function.

The memory usage is very low for th test case.

Not quite like a memory problem.

>
> --b.
>
Thanks in advance.

su