RE: [PATCH 1/3] nfsd: use __fput_sync() to avoid delayed closing of files.

From: "NeilBrown" <neilb@suse.de>
To: "David Laight" <David.Laight@ACULAB.COM>
Cc: "Al Viro" <viro@zeniv.linux.org.uk>,
	"Chuck Lever" <chuck.lever@oracle.com>,
	"Christian Brauner" <brauner@kernel.org>,
	"Jens Axboe" <axboe@kernel.dk>, "Oleg Nesterov" <oleg@redhat.com>,
	"Jeff Layton" <jlayton@kernel.org>,
	"linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-nfs@vger.kernel.org" <linux-nfs@vger.kernel.org>
Subject: RE: [PATCH 1/3] nfsd: use __fput_sync() to avoid delayed closing of files.
Date: Sat, 16 Dec 2023 09:36:25 +1100	[thread overview]
Message-ID: <170267978502.12910.3767924819236993323@noble.neil.brown.name> (raw)
In-Reply-To: <ac74bdb82e114d71b26864fe51f6433b@AcuMS.aculab.com>

On Sat, 16 Dec 2023, David Laight wrote:
> ...
> > > PS: put it that way - I can buy "nfsd is doing that only to regular
> > > files and not on an arbitrary filesystem, at that; having the thread
> > > wait on that sucker is not going to cause too much trouble"; I do *not*
> > > buy turning it into a thing usable outside of a very narrow set of
> > > circumstances.
> > >
> > 
> > Can you say more about "not on an arbitrary filesystem" ?
> > I guess you means that procfs and/or sysfs might be problematic as may
> > similar virtual filesystems (nfsd maybe).
> 
> Can nfs export an ext4 fs that is on a loopback mount on a file
> that is remotely nfs (or other) mounted?

Sure.  There is no reason this would cause a problem.
If the nfs mount were also a loopback mount, that might be interesting.
i.e.  You have a local filesystem, containing a file with a filesystem
image.
You nfs-export that local filesystem, and nfs mount it on the same host.
Then you loop-mount that file in the nfs-mounted filesystem.  Now you
are getting into uncharted waters.  I've testing loop-back NFS mounting
and assure that it works.  I haven't tried the double-loop.
But if that caused problem, I though it would be fput.  It would be
fsync or writeback which causes the problem.

> 
> As soon as you get loops like that you might find that fput() starts
> being problematic.

When calling fput on a regular file there are, I think, only two problem
areas.  One is that the fput might lead to a lazy-filesystem unmount
completing.  That only applies to MNT_INTERNAL filesystems, and they are
unlikely to be exported (probably it's impossible, but I haven't
checked).
The other is synchronous (or even async) IO in the filesystem code,
maybe completing an unlink or a truncate.  This is no worse than any
other synchronous IO that nfsd does.

Thanks,
NeilBrown

> 
> I'm also sure I remember that nfs wasn't supposed to respond to a write
> until it had issued the actual disk write - but maybe no one do that
> any more because it really is too slow.
> (Especially if the 'disk' is a USB stick.)
> 
> 	David
> 
> -
> Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
> Registration No: 1397386 (Wales)
> 