[PATCH 1/2] fs/super.c: introduce get_tree_bdev_by_dev()

[Gao Xiang]

Hi Christoph,

On 2024/10/8 19:49, Christoph Hellwig wrote:
> On Tue, Oct 08, 2024 at 05:56:05PM +0800, Gao Xiang wrote:
>> As Allison reported [1], currently get_tree_bdev() will store
>> "Can't lookup blockdev" error message.  Although it makes sense for
>> pure bdev-based fses, this message may mislead users who try to use
>> EROFS file-backed mounts since get_tree_nodev() is used as a fallback
>> then.
>>
>> Add get_tree_bdev_by_dev() to specify a device number explicitly
>> instead of the hardcoded fc->source as mentioned in [2], there are
>> other benefits like:
>>    - Filesystems can have other ways to get a bdev-based sb
>>      in addition to the current hard-coded source path;
>>
>>    - Pseudo-filesystems can utilize this method to generate a
>>      filesystem from given device numbers too.
>>
>>    - Like get_tree_nodev(), it doesn't strictly tie to fc->source
>>      either.
> 
> Do you have concrete plans for any of those?  If so send pointers.

Thanks for your comment.

I don't have any pointer for now, just listing the potential use
cases for reference.

But the error message out of get_tree_bdev() is inflexible and
IMHO it's too coupled to `fc->source`.

> Otherwise just passing a quiet flag of some form feels like a much
> saner interface.

I'm fine with this way, but that will be a treewide change, I
will send out a version with a flag later.

Thanks,
Gao Xiang