Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Date:   Tue, 17 May 2022 22:32:38 +0000
From:   Al Viro <viro@zeniv.linux.org.uk>
To:     Daniil Lunev <dlunev@chromium.org>
Cc:     linux-fsdevel@vger.kernel.org, hch@infradead.org,
        fuse-devel@lists.sourceforge.net, tytso@mit.edu, miklos@szeredi.hu,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 2/2] FUSE: Retire superblock on force unmount
Message-ID: <YoQihi4OMjJj2Mj0@zeniv-ca.linux.org.uk>
References: <20220511222910.635307-1-dlunev@chromium.org>
 <20220512082832.v2.2.I692165059274c30b59bed56940b54a573ccb46e4@changeid>
 <YoQfls6hFcP3kCaH@zeniv-ca.linux.org.uk>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YoQfls6hFcP3kCaH@zeniv-ca.linux.org.uk>
Sender: Al Viro <viro@ftp.linux.org.uk>
Precedence: bulk

On Tue, May 17, 2022 at 10:20:06PM +0000, Al Viro wrote:
> On Thu, May 12, 2022 at 08:29:10AM +1000, Daniil Lunev wrote:
> > Force unmount of FUSE severes the connection with the user space, even
> > if there are still open files. Subsequent remount tries to re-use the
> > superblock held by the open files, which is meaningless in the FUSE case
> > after disconnect - reused super block doesn't have userspace counterpart
> > attached to it and is incapable of doing any IO.
> 
> 	Why not simply have those simply rejected by fuse_test_super()?
> Looks like that would be much smaller and less invasive patch...
> Confused...

... because Miklos had suggested that, apparently ;-/  I disagree -
that approach has more side effects.  "mount will skip that sucker" is,
AFAICS, the only effect of modiyfing test_super callback(s); yours, OTOH...

Note that generic_shutdown_super() is *not* called while superblock is
mounted anywhere.  And it doesn't get to eviction from the list while it still
has live dentries.  Or inodes, for that matter.

So this
        if (sb->s_bdi != &noop_backing_dev_info) {
		if (sb->s_iflags & SB_I_PERSB_BDI)
			bdi_unregister(sb->s_bdi);
		bdi_put(sb->s_bdi);
		sb->s_bdi = &noop_backing_dev_info;
	}
is almost certainly not safe to be done on a live superblock.