Received-SPF: pass (google.com: domain of linux-kernel+bounces-131794-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Message-ID: <14e66691a65e3d05d3d8d50e74dfb366@manguebit.com>
From: Paulo Alcantara <pc@manguebit.com>
To: David Howells <dhowells@redhat.com>, Steve French <sfrench@samba.org>
Cc: dhowells@redhat.com, Shyam Prasad N <sprasad@microsoft.com>,
 linux-cifs@vger.kernel.org, linux-fsdevel@vger.kernel.org,
 linux-kernel@vger.kernel.org
Subject: Re: [PATCH] cifs: Fix reacquisition of volume cookie on still-live
 connection
In-Reply-To: <3756406.1712244064@warthog.procyon.org.uk>
References: <3756406.1712244064@warthog.procyon.org.uk>
Date: Thu, 04 Apr 2024 13:07:24 -0300
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain

Hi Dave,

David Howells <dhowells@redhat.com> writes:

> During mount, cifs_mount_get_tcon() gets a tcon resource connection record
> and then attaches an fscache volume cookie to it.  However, it does this
> irrespective of whether or not the tcon returned from cifs_get_tcon() is a
> new record or one that's already in use.  This leads to a warning about a
> volume cookie collision and a leaked volume cookie because tcon->fscache
> gets reset.
>
> Fix this be adding a mutex and a "we've already tried this" flag and only
> doing it once for the lifetime of the tcon.

Can't we just move the cookie acquisition to cifs_get_tcon() before it
gets added to list @ses->tcon_list.  This way we'll guarantee that the
cookie is set only once for the new tcon.

Besides, do we want to share a tcon with two different superblocks that
have 'fsc' and 'nofsc', respectively?  If not, it would be better to fix
match_tcon() as well to handle such case.

> [!] Note: Looking at cifs_mount_get_tcon(), a more general solution may
> actually be required.  Reacquiring the volume cookie isn't the only thing
> that function does: it also partially reinitialises the tcon record without
> any locking - which may cause live filesystem ops already using the tcon
> through a previous mount to malfunction.

Agreed.