Re: [PATCH] erofs: add error log in erofs_fc_parse_param

[Gao Xiang <hsiangkao@linux.alibaba.com>]

On 2025/1/17 18:00, Chen Linxuan wrote:
> On Fri, 2025-01-17 at 17:54 +0800, Gao Xiang wrote:
>>
>> On 2025/1/17 17:50, Chen Linxuan wrote:
>>> On Fri, 2025-01-17 at 17:28 +0800, Gao Xiang wrote:
>>>> Hi Linxuan,
>>>>
>>>> On 2025/1/17 16:52, Chen Linxuan wrote:
>>>>> While reading erofs code, I notice that `erofs_fc_parse_param` will
>>>>> return -ENOPARAM, which means that erofs do not support this option,
>>>>> without report anything when `fs_parse` return an unknown `opt`.
>>>>>
>>>>> But if an option is unknown to erofs, I mean that option not in
>>>>> `erofs_fs_parameters` at all, `fs_parse` will return -ENOPARAM,
>>>>> which means that `erofs_fs_parameters` should has returned earlier.
>>>>>
>>>>> Entering `default` means `fs_parse` return something we unexpected.
>>>>> I am not sure about it but I think we should return -EINVAL here,
>>>>> just like `xfs_fs_parse_param`.
>>>>>
>>>>> Signed-off-by: Chen Linxuan <chenlinxuan@uniontech.com>
>>>>
>>>> I think the default branch is actually deadcode here, see
>>>> erofs_fc_parse_param() -> fs_parse() -> fs_lookup_key() -> -ENOPARAM
>>>>
>>>> then vfs_parse_fs_param() will show "Unknown parameter".
>>>>
>>>> Maybe we could just kill `default:` branch...
>>>
>>> ext4 do not have a `default:` branch, but xfs return -EINVAL.
>>>
>>> I think `default:` branch can report error when `fs_parse` or
>>> `erofs_fs_parameters` goes wrong.
>>
>> How can it go wrong?
> 
> What if we forget to update the switch branch for a new option?

Then it's clearly a bug (we don't even handle the new option),
I think we shouldn't consider it as a normal case.

Thanks,
Gao Xiang

> 
>>
>> Thanks,
>> Gao Xiang
>>
>>