linux-kernel - Re: [RFC PATCH 1/6] fs: invoke LSM file_open hook in do_dentry_open for O

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20250313-dompteur-dachten-bb695fcbebf1@brauner>
Date: Thu, 13 Mar 2025 09:50:31 +0100
From: Christian Brauner <brauner@...nel.org>
To: Al Viro <viro@...iv.linux.org.uk>, Ryan Lee <ryan.lee@...onical.com>
Cc: linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org, 
	apparmor@...ts.ubuntu.com, linux-security-module@...r.kernel.org, selinux@...r.kernel.org, 
	Jan Kara <jack@...e.cz>, John Johansen <john.johansen@...onical.com>, 
	Paul Moore <paul@...l-moore.com>, James Morris <jmorris@...ei.org>, 
	"Serge E. Hallyn" <serge@...lyn.com>, Mickaël Salaün <mic@...ikod.net>, 
	Günther Noack <gnoack@...gle.com>, Stephen Smalley <stephen.smalley.work@...il.com>, 
	Ondrej Mosnacek <omosnace@...hat.com>, Casey Schaufler <casey@...aufler-ca.com>, 
	Kentaro Takeda <takedakn@...data.co.jp>, Tetsuo Handa <penguin-kernel@...ove.sakura.ne.jp>
Subject: Re: [RFC PATCH 1/6] fs: invoke LSM file_open hook in do_dentry_open
 for O_PATH fds as well

On Wed, Mar 12, 2025 at 09:37:14PM +0000, Al Viro wrote:
> On Wed, Mar 12, 2025 at 02:21:41PM -0700, Ryan Lee wrote:
> > Currently, opening O_PATH file descriptors completely bypasses the LSM
> > infrastructure. Invoking the LSM file_open hook for O_PATH fds will
> > be necessary for e.g. mediating the fsmount() syscall.

LSM mediation for the mount api should be done by adding appropriate
hooks to the new mount api.

> > 
> > Signed-off-by: Ryan Lee <ryan.lee@...onical.com>
> > ---
> >  fs/open.c | 7 ++++++-
> >  1 file changed, 6 insertions(+), 1 deletion(-)
> > 
> > diff --git a/fs/open.c b/fs/open.c
> > index 30bfcddd505d..0f8542bf6cd4 100644
> > --- a/fs/open.c
> > +++ b/fs/open.c
> > @@ -921,8 +921,13 @@ static int do_dentry_open(struct file *f,
> >  	if (unlikely(f->f_flags & O_PATH)) {
> >  		f->f_mode = FMODE_PATH | FMODE_OPENED;
> >  		file_set_fsnotify_mode(f, FMODE_NONOTIFY);
> >  		f->f_op = &empty_fops;
> > -		return 0;
> > +		/*
> > +		 * do_o_path in fs/namei.c unconditionally invokes path_put
> > +		 * after this function returns, so don't path_put the path
> > +		 * upon LSM rejection of O_PATH opening
> > +		 */
> > +		return security_file_open(f);
> 
> Unconditional path_put() in do_o_path() has nothing to do with that -
> what gets dropped there is the reference acquired there; the reference
> acquired (and not dropped) here is the one that went into ->f_path.
> Since you are leaving FMODE_OPENED set, you will have __fput() drop
> that reference.
> 
> Basically, you are simulating behaviour on the O_DIRECT open of
> something that does not support O_DIRECT - return an error, with
> ->f_path and FMODE_OPENED left in place.
> 
> Said that, what I do not understand is the point of that exercise -
> why does LSM need to veto anything for those and why is security_file_open()

I really think this is misguided. This should be done via proper hooks
into apis that use O_PATH file descriptors for specific purposes but not
for the generic open() path.

> the right place for such checks?

It isn't.

> 
> The second part is particularly interesting...