Received-SPF: pass (google.com: domain of linux-ext4+bounces-204-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Date: Mon, 27 Nov 2023 17:25:44 +0000
From: Al Viro <viro@zeniv.linux.org.uk>
To: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Gabriel Krisman Bertazi <gabriel@krisman.be>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Christian Brauner <brauner@kernel.org>, tytso@mit.edu,
	linux-f2fs-devel@lists.sourceforge.net, ebiggers@kernel.org,
	linux-fsdevel@vger.kernel.org, jaegeuk@kernel.org,
	linux-ext4@vger.kernel.org, Miklos Szeredi <miklos@szeredi.hu>
Subject: Re: fun with d_invalidate() vs. d_splice_alias() was Re: [f2fs-dev]
 [PATCH v6 0/9] Support negative dentries on case-insensitive ext4 and f2fs
Message-ID: <20231127172544.GJ38156@ZenIV>
References: <CAHk-=wh5WYPN7BLSUjUr_VBsPTxHOcMHo1gOH2P4+5NuXAsCKA@mail.gmail.com>
 <20231123171255.GN38156@ZenIV>
 <20231123182426.GO38156@ZenIV>
 <20231123215234.GQ38156@ZenIV>
 <20231125220136.GB38156@ZenIV>
 <20231126045219.GD38156@ZenIV>
 <20231126184141.GF38156@ZenIV>
 <20231127063842.GG38156@ZenIV>
 <87jzq3nqos.fsf@email.froward.int.ebiederm.org>
 <878r6jnq1t.fsf@email.froward.int.ebiederm.org>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <878r6jnq1t.fsf@email.froward.int.ebiederm.org>
Sender: Al Viro <viro@ftp.linux.org.uk>

On Mon, Nov 27, 2023 at 10:01:34AM -0600, Eric W. Biederman wrote:
> "Eric W. Biederman" <ebiederm@xmission.com> writes:
> 
> > I am confused what is going on with ext4 and f2fs.  I think they
> > are calling d_invalidate when all they need to call is d_drop.
> 
> ext4 and f2f2 are buggy in how they call d_invalidate, if I am reading
> the code correctly.
> 
> d_invalidate calls detach_mounts.
> 
> detach_mounts relies on setting D_CANT_MOUNT on the top level dentry to
> prevent races with new mounts.
>
> ext4 and f2fs (in their case insensitive code) are calling d_invalidate
> before dont_mount has been called to set D_CANT_MOUNT.

Not really - note that the place where we check cant_mount() is under
the lock on the mountpoint's inode, so anything inside ->unlink() or
->rmdir() is indistinguishable from the places where we do dont_mount()
in vfs_{unlink,rmdir}.