Re: [PATCH] Avoid null-pointer access in w1/slaves/w1_therm

[David Fries <david@fries.net>]

On Sun, Mar 01, 2015 at 04:48:22AM +0300, ??????? ??????? wrote:
> Hi everyone
> 
> 28.02.2015, 23:18, "David Fries" <David@Fries.net>:
> > Thanks for preparing the patch, it looks like it will go another
> > round, but that happens to everyone.
> 
> Patch itself doesn't solve the problem, it only adds a workaround.
> There is a reference counter issue - reading data via sysfs only holds
> a reference to device, w1 slave structure will be deleted by w1 search core.
> 
> I believe it only accidentally doesn't crash because of kernel memory allocator
> hasn't yet reclaimed memory.

Good point, it does look like that's the case.

w1_unref_slave calls w1_family_notify(BUS_NOTIFY_DEL_DEVICE, sl);
which calls sl->family->fops->remove_slave(sl);
which kfree and sl->family_data = NULL;

In that case what do you think about in w1_slave_show after getting
bus_mutex,
atomic_inc(&sl->refcnt);

then add
w1_unref_slave(sl);
in each return path, or do a goto and consolidate them.

or just increment it while sleeping, which is when it's needed, which
also looks simpler.

			if (external_power) {
+				int refcnt;
				mutex_unlock(&dev->bus_mutex);

+				/* prevent the slave from going away */
+				atomic_inc(&sl->refcnt);
				sleep_rem = msleep_interruptible(tm);
+				refcnt = w1_unref_slave(sl);
-				if (sleep_rem != 0)
+				if (sleep_rem != 0 || !refcnt)
					return -EINTR;

				i = mutex_lock_interruptible(&dev->bus_mutex);
				if (i != 0)
					return i;
			} else if (!w1_strong_pullup) {


-- 
David Fries <david@fries.net>    PGP pub CB1EE8F0
http://fries.net/~david/