linux-kernel - Re: [PATCH] keys: Fix UAF in key

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: <20250318113655.GB14792@redhat.com>
Date: Tue, 18 Mar 2025 12:36:55 +0100
From: Oleg Nesterov <oleg@...hat.com>
To: David Howells <dhowells@...hat.com>
Cc: Jarkko Sakkinen <jarkko@...nel.org>, Kees Cook <kees@...nel.org>,
	Greg KH <gregkh@...uxfoundation.org>,
	Josh Drake <josh@...phoslabs.com>,
	Suraj Sonawane <surajsonawane0215@...il.com>,
	keyrings@...r.kernel.org, linux-security-module@...r.kernel.org,
	security@...nel.org, stable@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] keys: Fix UAF in key_put()

Either way... I know nothing about security/key, but grep -w key_put
finds

 *  When it is no longer required, the key should be released using::

	void key_put(struct key *key);

    Or::

	void key_ref_put(key_ref_t key_ref);

    These can be called from interrupt context.
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

in Documentation/security/keys/core.rst

and since key_user_put() takes key_user_lock with irqs enabled,
key_put()->key_user_put() doesn't look correct...

Oleg.

On 03/18, Oleg Nesterov wrote:
>
> On 03/18, David Howells wrote:
> >
> > --- a/security/keys/key.c
> > +++ b/security/keys/key.c
> > @@ -645,21 +645,30 @@ EXPORT_SYMBOL(key_reject_and_link);
> >   */
> >  void key_put(struct key *key)
> >  {
> > +	int quota_flag;
> > +	unsigned short len;
> > +	struct key_user *user;
> > +
> >  	if (key) {
> >  		key_check(key);
> >
> > +		quota_flag = test_bit(KEY_FLAG_IN_QUOTA, &key->flags);
> > +		len = key->quotalen;
> > +		user = key->user;
> > +		refcount_inc(&user->usage);
> >  		if (refcount_dec_and_test(&key->usage)) {
> >  			unsigned long flags;
> >
> >  			/* deal with the user's key tracking and quota */
> > -			if (test_bit(KEY_FLAG_IN_QUOTA, &key->flags)) {
> > -				spin_lock_irqsave(&key->user->lock, flags);
> > -				key->user->qnkeys--;
> > -				key->user->qnbytes -= key->quotalen;
> > -				spin_unlock_irqrestore(&key->user->lock, flags);
> > +			if (quota_flag) {
> > +				spin_lock_irqsave(&user->lock, flags);
> > +				user->qnkeys--;
> > +				user->qnbytes -= len;
> > +				spin_unlock_irqrestore(&user->lock, flags);
> >  			}
> >  			schedule_work(&key_gc_work);
> >  		}
> > +		key_user_put(user);
>
> Do we really need the unconditional refcount_inc / key_user_put ?
>
> 	void key_put(struct key *key)
> 	{
> 		if (key) {
> 			struct key_user *user = NULL;
> 			unsigned short len;
>
> 			key_check(key);
>
> 			if (test_bit(KEY_FLAG_IN_QUOTA, &key->flags)) {
> 				len = key->quotalen;
> 				user = key->user;
> 				refcount_inc(&user->usage);
> 			}
>
> 			if (refcount_dec_and_test(&key->usage)) {
> 				unsigned long flags;
>
> 				/* deal with the user's key tracking and quota */
> 				if (user) {
> 					spin_lock_irqsave(&user->lock, flags);
> 					user->qnkeys--;
> 					user->qnbytes -= len;
> 					spin_unlock_irqrestore(&user->lock, flags);
> 				}
> 				schedule_work(&key_gc_work);
> 			}
>
> 			if (user)
> 				key_user_put(user);
> 		}
> 	}
>
> looks a bit more clear/simple to me...
>
> Oleg.