Re: [PATCH 0/3] fanotify support for btrfs sub-volumes

From: Amir Goldstein <amir73il@gmail.com>
To: Christian Brauner <brauner@kernel.org>
Cc: Christoph Hellwig <hch@infradead.org>,
	Josef Bacik <josef@toxicpanda.com>,
	 Qu Wenruo <quwenruo.btrfs@gmx.com>, Jan Kara <jack@suse.cz>,
	Chris Mason <clm@fb.com>,  David Sterba <dsterba@suse.com>,
	linux-btrfs@vger.kernel.org,  linux-fsdevel@vger.kernel.org
Subject: Re: [PATCH 0/3] fanotify support for btrfs sub-volumes
Date: Fri, 10 Nov 2023 12:31:04 +0200	[thread overview]
Message-ID: <CAOQ4uxheWsR5EWtCvZq33r1+LLj6ANQLi9OJKYBpDP92a6ZTkw@mail.gmail.com> (raw)
In-Reply-To: <20231110-vorleben-unvorbereitet-fe3b302c5079@brauner>

On Fri, Nov 10, 2023 at 11:33 AM Christian Brauner <brauner@kernel.org> wrote:
>
> > you hit a mount point or another (nested) subvolume.  Can't comment
> > on overlayfs.  But if it keeps mixing things forth and back what would
>
> Overlayfs shows that this st_dev switching happens on things other than
> btrfs. It has nothing to do with subvolumes.

IMO, overlayfs is a good example of Pragmatism.

On some filesystems, following POSIX standard verbatim is not possible.
We care more about the users and about how the POSIX standard affects
real life applications, than about actually following the standard.

This table is complex, but it explains the tradeoffs that overlayfs
does when following strict POSIX is not possible:
https://docs.kernel.org/filesystems/overlayfs.html#inode-properties

In the Legacy case, where Uniform st_dev is not possible,
overlayfs preserve these important *practical* rules:
1. Unique st_dev,st_ino
2. Uniform st_dev across all directories (non-persistent st_ino)
3. Persistent st_ino for non-directories (non-uniform st_dev)

Rule #2 is important for traversal boundaries (e.g. find -xdev, du -x)

It's not mentioned in this table, but overlayfs f_fsid is and always
was uniform. Since v6.6, overlayfs f_fsid is also unique among
overlayfs instances.

I don't know of a good way to stop a thread where all that needs
to be said has been said, but my opinion is that this thread has lost
focus a long time ago.

I will post a new patch for fanotify with Jan's proposal to address
Christoph's concerns:
- no ->get_fsid() method
- no sb/mount mark on btrfs subvol

Thanks,
Amir.