Re: file metadata via fs API (was: [GIT PULL] Filesystem Information)

From: Miklos Szeredi <miklos@szeredi.hu>
To: Al Viro <viro@zeniv.linux.org.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
	Jann Horn <jannh@google.com>,
	Casey Schaufler <casey@schaufler-ca.com>,
	Andy Lutomirski <luto@amacapital.net>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	David Howells <dhowells@redhat.com>, Karel Zak <kzak@redhat.com>,
	Jeff Layton <jlayton@redhat.com>,
	Miklos Szeredi <mszeredi@redhat.com>,
	Nicolas Dichtel <nicolas.dichtel@6wind.com>,
	Christian Brauner <christian@brauner.io>,
	Lennart Poettering <lennart@poettering.net>,
	Linux API <linux-api@vger.kernel.org>,
	Ian Kent <raven@themaw.net>,
	LSM <linux-security-module@vger.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: file metadata via fs API (was: [GIT PULL] Filesystem Information)
Date: Tue, 18 Aug 2020 11:41:29 +0200	[thread overview]
Message-ID: <CAJfpeguyQhfxrSnaH1mZkncgfiLFB2yM2ZgdMBzuhAdKdmmuxA@mail.gmail.com> (raw)
In-Reply-To: <20200812213041.GV1236603@ZenIV.linux.org.uk>

On Wed, Aug 12, 2020 at 11:30 PM Al Viro <viro@zeniv.linux.org.uk> wrote:
>
> On Wed, Aug 12, 2020 at 07:33:26PM +0100, Al Viro wrote:
>
> > BTW, what would such opened files look like from /proc/*/fd/* POV?  And
> > what would happen if you walk _through_ that symlink, with e.g. ".."
> > following it?  Or with names of those attributes, for that matter...
> > What about a normal open() of such a sucker?  It won't know where to
> > look for your ->private_data...
> >
> > FWIW, you keep refering to regularity of this stuff from the syscall
> > POV, but it looks like you have no real idea of what subset of the
> > things available for normal descriptors will be available for those.
>
> Another question: what should happen with that sucker on umount of
> the filesystem holding the underlying object?  Should it be counted
> as pinning that fs?

Obviously yes.

> Who controls what's in that tree?

It could be several entities:

 - global (like mount info)
 - per inode (like xattr)
 - per fs (fs specific inode attributes)
 - etc..

>  If we plan to have xattrs there,
> will they be in a flat tree, or should it mirror the hierarchy of
> xattrs?  When is it populated?  open() time?  What happens if we
> add/remove an xattr after that point?

From the interface perspective it would be dynamic (i.e. would get
updated on open or read).  From an implementation POV it could have
caching, but that's not how I'd start out.

> If we open the same file several times, what should we get?  A full
> copy of the tree every time, with all coherency being up to whatever's
> putting attributes there?
>
> What are the permissions needed to do lookups in that thing?

That would depend on what would need to be looked up.  Top level would
be world readable, otherwise it would be up to the attribute/group.

>
> All of that is about semantics and the answers are needed before we
> start looking into implementations.  "Whatever my implementation
> does" is _not_ a good way to go, especially since that'll be cast
> in stone as soon as API becomes exposed to userland...

Fine.

Thanks,
Miklos