xfstests failed test cases

xfstests failed test cases

[ChenXiaoSong]

I have reproduced the failures of these test cases. For detailed 
information, please check the link: 
https://chenxiaosong.com/en/smb-buildbot.html (I will ensure this link 
is always accessible).

Perhaps sending these issues to the mailing list could help resolve them 
more quickly.

Thanks,
ChenXiaoSong <chenxiaosong@kylinos.cn>

On 1/25/26 03:45, Steve French wrote:
> I did see a crash in cifs/103
> (http://smb311-linux-testing.southcentralus.cloudapp.azure.com/#/ 
> builders/12/builds/38/steps/19/logs/stdio)
> running mchan test run to Windows.  Ideas?

On 1/25/26 12:36, Steve French wrote:
> Any thoughts on the failures with current for-next (to WIndows, fast
> VM, with multichannel) for tests generic/533 and generic/565
> 
> +++ /data/xfstests-dev/results//smb3mcmfs/generic/565.out.bad
> 2026-01-24 22:13:17.098765412 -0600
> @@ -1,4 +1,6 @@
> QA output created by 565
> +mkdir: cannot create directory '/mnt/test/test-565': Input/output error
> +/mnt/scratch/copy: Input/output error cmp: /mnt/test/test-565/file:
> No such file or directory
> md5sums after xdev copy:
> -81615449a98aaaad8dc179b3bec87f38 TEST_DIR/test-565/file
> -81615449a98aaaad8dc179b3bec87f38 SCRATCH_MNT/copy
> +md5sum: /mnt/test/test-565/file: No such file or directory
> 
> http://smb311-linux-testing.southcentralus.cloudapp.azure.com/#/builders/12/builds/46/steps/138/logs/stdio
> 
> and
> 
> http://smb311-linux-testing.southcentralus.cloudapp.azure.com/#/builders/12/builds/46/steps/136/logs/stdio
> 
> and also git/0003
> 
> http://smb311-linux-testing.southcentralus.cloudapp.azure.com/#/builders/12/builds/46/steps/160/logs/stdio
> 
> Does this fail for you as well?

Re: xfstests failed test cases

[ChenXiaoSong]

I will identify the cause of these test case failures as soon as possible.

Thanks,
ChenXiaoSong <chenxiaosong@kylinos.cn>

On 1/28/26 23:37, Steve French wrote:
> This is helpful. Any luck finding out if fails on both mainline and also 
> for-next and if so can we bisect it?

Re: xfstests failed test cases

[ChenXiaoSong]

I found the key factor that triggers kmemleak in cifs/103, the directory 
exported by the server must be read-only.

When sharing a folder on Windows, do not check "Allow Change", set the 
permissions of the shared directory to read-only.

Alternatively, `smb.conf` of Samba is configured as follows:
```
[test]
     ...
     read only = yes
[test2]
     ...
     read only = yes
```

For detailed information, please check the link: 
https://chenxiaosong.com/en/smb-buildbot.html#cifs-103

I will first try to analyze the code, and if that doesn't work, I will 
try to bisect it.

Thanks,
ChenXiaoSong <chenxiaosong@kylinos.cn>

On 1/28/26 23:29, ChenXiaoSong wrote:
> I have reproduced the failures of these test cases. For detailed 
> information, please check the link: https://chenxiaosong.com/en/smb- 
> buildbot.html (I will ensure this link is always accessible).
> 
> Perhaps sending these issues to the mailing list could help resolve them 
> more quickly.

Re: xfstests failed test cases

[ChenXiaoSong]

Steve and I have completed a git bisect and found the first bad commit:
[e255612b5ed9f179abe8196df7c2ba09dd227900] cifs: Add fallback for SMB2 
CREATE without FILE_READ_ATTRIBUTES

For detailed information, please check the link: 
https://chenxiaosong.com/en/smb-buildbot.html#cifs-103-bisect

Thanks,
ChenXiaoSong <chenxiaosong@kylinos.cn>

On 1/29/26 8:20 PM, ChenXiaoSong wrote:
> I found the key factor that triggers kmemleak in cifs/103, the directory 
> exported by the server must be read-only.
> 
> When sharing a folder on Windows, do not check "Allow Change", set the 
> permissions of the shared directory to read-only.
> 
> Alternatively, `smb.conf` of Samba is configured as follows:
> ```
> [test]
>      ...
>      read only = yes
> [test2]
>      ...
>      read only = yes
> ```
> 
> For detailed information, please check the link: https:// 
> chenxiaosong.com/en/smb-buildbot.html#cifs-103
> 
> I will first try to analyze the code, and if that doesn't work, I will 
> try to bisect it.
> 
> Thanks,
> ChenXiaoSong <chenxiaosong@kylinos.cn>
> 
> On 1/28/26 23:29, ChenXiaoSong wrote:
>> I have reproduced the failures of these test cases. For detailed 
>> information, please check the link: https://chenxiaosong.com/en/smb- 
>> buildbot.html (I will ensure this link is always accessible).
>>
>> Perhaps sending these issues to the mailing list could help resolve 
>> them more quickly.

Re: xfstests failed test cases

[ChenXiaoSong]

I have already submitted a patch to fix this issue: 
https://lore.kernel.org/linux-cifs/20260201081017.998628-1-chenxiaosong.chenxiaosong@linux.dev/

Thanks,
ChenXiaoSong <chenxiaosong@kylinos.cn>

On 2/1/26 2:30 PM, ChenXiaoSong wrote:
> Steve and I have completed a git bisect and found the first bad commit:
> [e255612b5ed9f179abe8196df7c2ba09dd227900] cifs: Add fallback for SMB2 
> CREATE without FILE_READ_ATTRIBUTES
> 
> For detailed information, please check the link: https:// 
> chenxiaosong.com/en/smb-buildbot.html#cifs-103-bisect
> 
> Thanks,
> ChenXiaoSong <chenxiaosong@kylinos.cn>
> 
> On 1/29/26 8:20 PM, ChenXiaoSong wrote:
>> I found the key factor that triggers kmemleak in cifs/103, the 
>> directory exported by the server must be read-only.
>>
>> When sharing a folder on Windows, do not check "Allow Change", set the 
>> permissions of the shared directory to read-only.
>>
>> Alternatively, `smb.conf` of Samba is configured as follows:
>> ```
>> [test]
>>      ...
>>      read only = yes
>> [test2]
>>      ...
>>      read only = yes
>> ```
>>
>> For detailed information, please check the link: https:// 
>> chenxiaosong.com/en/smb-buildbot.html#cifs-103
>>
>> I will first try to analyze the code, and if that doesn't work, I will 
>> try to bisect it.
>>
>> Thanks,
>> ChenXiaoSong <chenxiaosong@kylinos.cn>
>>
>> On 1/28/26 23:29, ChenXiaoSong wrote:
>>> I have reproduced the failures of these test cases. For detailed 
>>> information, please check the link: https://chenxiaosong.com/en/smb- 
>>> buildbot.html (I will ensure this link is always accessible).
>>>
>>> Perhaps sending these issues to the mailing list could help resolve 
>>> them more quickly.
> 

Re: xfstests failed test cases

[ChenXiaoSong]

I have submitted another patch to fix the SMB1 kmemleak issue: 
https://lore.kernel.org/linux-cifs/20260202064928.1879323-1-chenxiaosong.chenxiaosong@linux.dev/

Thanks,
ChenXiaoSong <chenxiaosong@kylinos.cn>

On 2/1/26 4:14 PM, ChenXiaoSong wrote:
> I have already submitted a patch to fix this issue: https:// 
> lore.kernel.org/linux-cifs/20260201081017.998628-1- 
> chenxiaosong.chenxiaosong@linux.dev/
> 
> Thanks,
> ChenXiaoSong <chenxiaosong@kylinos.cn>
> 
> On 2/1/26 2:30 PM, ChenXiaoSong wrote:
>> Steve and I have completed a git bisect and found the first bad commit:
>> [e255612b5ed9f179abe8196df7c2ba09dd227900] cifs: Add fallback for SMB2 
>> CREATE without FILE_READ_ATTRIBUTES
>>
>> For detailed information, please check the link: https:// 
>> chenxiaosong.com/en/smb-buildbot.html#cifs-103-bisect
>>
>> Thanks,
>> ChenXiaoSong <chenxiaosong@kylinos.cn>
>>
>> On 1/29/26 8:20 PM, ChenXiaoSong wrote:
>>> I found the key factor that triggers kmemleak in cifs/103, the 
>>> directory exported by the server must be read-only.
>>>
>>> When sharing a folder on Windows, do not check "Allow Change", set 
>>> the permissions of the shared directory to read-only.
>>>
>>> Alternatively, `smb.conf` of Samba is configured as follows:
>>> ```
>>> [test]
>>>      ...
>>>      read only = yes
>>> [test2]
>>>      ...
>>>      read only = yes
>>> ```
>>>
>>> For detailed information, please check the link: https:// 
>>> chenxiaosong.com/en/smb-buildbot.html#cifs-103
>>>
>>> I will first try to analyze the code, and if that doesn't work, I 
>>> will try to bisect it.
>>>
>>> Thanks,
>>> ChenXiaoSong <chenxiaosong@kylinos.cn>
>>>
>>> On 1/28/26 23:29, ChenXiaoSong wrote:
>>>> I have reproduced the failures of these test cases. For detailed 
>>>> information, please check the link: https://chenxiaosong.com/en/smb- 
>>>> buildbot.html (I will ensure this link is always accessible).
>>>>
>>>> Perhaps sending these issues to the mailing list could help resolve 
>>>> them more quickly.
>>
>