Re: [PATCH] Fix read-only superblock in case of subvol RO remount

From: Graham Cobb <g.btrfs@cobb.uk.net>
To: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: linux-btrfs@vger.kernel.org, linux-fsdevel@vger.kernel.org,
	dhowells@redhat.com, fvogt@suse.com
Subject: Re: [PATCH] Fix read-only superblock in case of subvol RO remount
Date: Thu, 10 Feb 2022 22:03:26 +0000	[thread overview]
Message-ID: <938de929-d63f-2f04-ec0a-9005ba013a2f@cobb.uk.net> (raw)
In-Reply-To: <20220210213058.m7kukfryrk4cgsye@fiona>

On 10/02/2022 21:30, Goldwyn Rodrigues wrote:
> On 19:54 10/02, Graham Cobb wrote:
>> On 10/02/2022 16:51, Goldwyn Rodrigues wrote:
>>> If a read-write root mount is remounted as read-only, the subvolume
>>> is also set to read-only.
>>
>> Errrr... Isn't that exactly what I want?
>>
>> If I have a btrfs filesystem with hundreds of subvols, some of which may
>> be mounted into various places in my filesystem, I would expect that if
>> I remount the main mountpoint as RO, that all the subvols become RO as
>> well. I actually don't mind if the behaviour went further and remounting
>> ANY of the mount points as RO would make them all RO.
>>
>> My mental model is that mounting a subvol somewhere is rather like a
>> bind mount. And a bind mount goes RO if the underlying fs goes RO -
>> doesn't it?
>>
> 
> If we want bind mount, we would use bind mount. subvolume mounts and bind
> mounts are different and should be treated as different features.

Yes that's a good point. However, I am still not convinced that this is
a change in behaviour that is obvious enough to justify the risk of
disruption to existing systems, admin scripts or system managers.

> 
>> Or am I just confused about what this patch is discussing?
> 
> Root can also be considered as a unique subvolume with a unique
> subvolume id and a unique name=/

But with an important special property that is different from all other
subvolumes: all other subvolumes are reachable from it.