Re: [PATCH v2 0/5] btrfs: use the super_block as bdev holder

[Qu Wenruo <wqu@suse.com>]

在 2025/6/13 15:29, hch 写道:
> On Fri, Jun 13, 2025 at 07:51:25AM +0930, Qu Wenruo wrote:
>>> -       if (fs_info->fs_devices) {
>>> +       if (fs_info && fs_info->fs_devices) {
>>> +               ASSERT(fs_info->fs_devices->is_open);
>>>                    btrfs_close_devices(fs_info->fs_devices);
>>>                    fs_info->fs_devices = NULL;
>>>            }
>>>
>>> So if we end up in btrfs_reconfigure_for_mount() and fs_info and
>>> fs_info->fs_devices are set, I see the is_open flag being set as
>>> well. But the fstests run isn't 100% finished yet (and it's only
>>> been a -g quick run anyways).
>>
>> Since I'm also working on cleaning up the mount process, I'm getting a
>> little familiar with this part, but if HCH can comment on this, it will be
>> a great help.
> 
> I wish I could.  A lot of this mount stuff has been entirely paged out
> of my memory, I'm sorry.  Note that Christian did some fairly big rework
> in this area, and now the new mount API came in as well.  So things
> around it looks pretty different.

That's totally fine.

And since I'm already working on the surrounding code, I'll definitely 
make the involved code easier to read.

> 
> I think the parts of the series that are valueable as is are the
> "open read-only for scanning" and split the inuse counter bits,
> which are pretty obvious.  Everything else might need a more or less
> big redo with all the surrounding changes.
> 

And the "open block devices after super block creation" patch.

It's the existing code making the sb creation way harder to grasp.

With that improved, the benefit of delaying device open should be very 
obvious.

If Johannes is fine, I can integrate those patches into my upcoming 
get_tree cleanup.

Thanks,
Qu