Re: drivers/char/random.c line 728 BUG

[Andrew Morton <akpm@linux-foundation.org>]

On Fri, 29 Aug 2008 15:31:55 -0700
Aaron Straus <aaron@merfinllc.com> wrote:

> > -	if (r == &input_pool &&
> > -	    r->entropy_count >= random_read_wakeup_thresh) {
> > +	if (r == &input_pool && entropy_count >= random_read_wakeup_thresh) {
> >  		wake_up_interruptible(&random_read_wait);
> >  		kill_fasync(&fasync, SIGIO, POLL_IN);
> >  	}
> > -
> > +	r->entropy_count = entropy_count;
> 
> Here you set the entropy_count in the pool *after* waking up the reader.
> I was a little worried the reader would find the entropy_count unchanged
> and go back to sleep.
> 

Probably it doesn't matter because of the coverage of the lock but yes,
let's do the assignment first.

--- a/drivers/char/random.c~drivers-char-randomc-fix-a-race-which-can-lead-to-a-bogus-bug-fix-fix
+++ a/drivers/char/random.c
@@ -535,13 +535,13 @@ static void credit_entropy_bits(struct e
 		entropy_count = 0;
 	} else if (entropy_count > r->poolinfo->POOLBITS)
 		entropy_count = r->poolinfo->POOLBITS;
+	r->entropy_count = entropy_count;
 
 	/* should we wake readers? */
 	if (r == &input_pool && entropy_count >= random_read_wakeup_thresh) {
 		wake_up_interruptible(&random_read_wait);
 		kill_fasync(&fasync, SIGIO, POLL_IN);
 	}
-	r->entropy_count = entropy_count;
 	spin_unlock_irqrestore(&r->lock, flags);
 }
 
_