Re: [PATCH v4 02/27] ntsync: Introduce NTSYNC_IOC_WAIT_ALL.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wednesday, 17 April 2024 06:37:03 CDT Peter Zijlstra wrote:
> On Mon, Apr 15, 2024 at 08:08:12PM -0500, Elizabeth Figura wrote:
> > +	if (atomic_read(&sem->all_hint) > 0) {
> > +		spin_lock(&dev->wait_all_lock);
> > +		spin_lock_nest_lock(&sem->lock, &dev->wait_all_lock);
> >  
> > +		prev_count = sem->u.sem.count;
> > +		ret = post_sem_state(sem, args);
> > +		if (!ret) {
> > +			try_wake_all_obj(dev, sem);
> > +			try_wake_any_sem(sem);
> > +		}
> >  
> > +		spin_unlock(&sem->lock);
> > +		spin_unlock(&dev->wait_all_lock);
> > +	} else {
> > +		spin_lock(&sem->lock);
> > +
> > +		prev_count = sem->u.sem.count;
> > +		ret = post_sem_state(sem, args);
> > +		if (!ret)
> > +			try_wake_any_sem(sem);
> > +
> > +		spin_unlock(&sem->lock);
> > +	}
> >  
> >  	if (!ret && put_user(prev_count, user_args))
> >  		ret = -EFAULT;
> 
> vs.
> 
> > +	/* queue ourselves */
> > +
> > +	spin_lock(&dev->wait_all_lock);
> > +
> > +	for (i = 0; i < args.count; i++) {
> > +		struct ntsync_q_entry *entry = &q->entries[i];
> > +		struct ntsync_obj *obj = entry->obj;
> > +
> > +		atomic_inc(&obj->all_hint);
> > +
> > +		/*
> > +		 * obj->all_waiters is protected by dev->wait_all_lock rather
> > +		 * than obj->lock, so there is no need to acquire obj->lock
> > +		 * here.
> > +		 */
> > +		list_add_tail(&entry->node, &obj->all_waiters);
> > +	}
> 
> This looks racy, consider:
> 
> 	atomic_read(all_hints) /* 0 */
> 
> 				spin_lock(wait_all_lock)
> 				atomic_inc(all_hint)	/* 1 */
> 				list_add_tail()
> 
> 	spin_lock(sem->lock)
> 	/* try_wake_all_obj() missing */
> 
> 
> 
> 
> I've not yet thought about if this is harmful or not, but if not, it
> definitely needs a comment.
> 
> Anyway, I need a break, maybe more this evening.

Ach. I wrote this with the idea that the race isn't meaningful, but
looking at it again you're right—there is a harmful race here.

I think it should be fixable by moving the atomic_read inside the lock,
though.







[Index of Archives]     [Kernel Newbies]     [Security]     [Netfilter]     [Bugtraq]     [Linux FS]     [Yosemite Forum]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Video 4 Linux]     [Device Mapper]     [Linux Resources]

  Powered by Linux