linux-kernel - Re: sparse warnings related to kref_put_lock() and refcount_dec_and

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]

Date:   Tue, 28 Jun 2022 10:58:21 +0200
From:   Luc Van Oostenryck <luc.vanoostenryck@...il.com>
To:     Alexander Aring <aahringo@...hat.com>
Cc:     jacob.e.keller@...el.com, akpm@...ux-foundation.org,
        thunder.leizhen@...wei.com, linux-sparse@...r.kernel.org,
        cluster-devel <cluster-devel@...hat.com>,
        linux-kernel@...r.kernel.org
Subject: Re: sparse warnings related to kref_put_lock() and
 refcount_dec_and_lock()

On Mon, Jun 27, 2022 at 09:06:43PM -0400, Alexander Aring wrote:
> >
> > If we change the refcount code to _never_ calling unlock() for the
> > specific lock, then all those foo_and_lock_bar() functions can be
> > annotated with "__acquires()". This should also end in the same code?
> 
> sorry, this will not work because of the first condition of "if
> (refcount_dec_not_one(r))" which will never hold the lock if true...
> that's what the optimization is all about. However, maybe somebody has
> another idea...

I would certainly not recommend this but ...
if it's OK to cheat and lie then you can do:
+	bool refcount_dec_and_lock(refcount_t *r, spinlock_t *lock) __acquires(lock);
+
	bool refcount_dec_and_lock(refcount_t *r, spinlock_t *lock)
	{
-		if (refcount_dec_not_one(r))
-			return false;
+		if (refcount_dec_not_one(r)) {
+			__acquire(lock);
+			return false;
+		}
	
		spin_lock(lock);
		if (!refcount_dec_and_test(r)) {
			spin_unlock(lock);
+			__acquire(lock);
			return false;
		}
	
		return true;
	}

In other word, pretend that the lock is always taken but ...
1) it's ugly
2) it's lying and can be confusing
3) now all the users of this function will have an imbalance problem
   (but they probably already have one since refcount_dec_and_lock()
    is not annotated).

What is needed is some kind of annotation for conditional locks.
I've tried a few time and in itself it was working but in most
cases the usage pattern was so that there was a imbalance anyway.

-- Luc