Re: kernel panic in fib_rules_lookup [2.6.27.7 vendor-patched]

[Eric Dumazet <eric.dumazet@gmail.com>]

Le samedi 23 octobre 2010 à 17:11 +0200, Eric Dumazet a écrit :
> Le samedi 23 octobre 2010 à 05:44 +0200, Eric Dumazet a écrit :
> > Le vendredi 22 octobre 2010 à 16:30 -0400, Joe Buehler a écrit :
> > > Eric Dumazet wrote:
> > > 
> > > > Could you provide a disassembly of function fib_rules_lookup ?
> > > 
> > > Try looking in http://68.100.141.95:3000/linux-crash/.  There should be
> > > the source file I am using (not current release if you recall), the .o,
> > > the disassembly, and a -S compile that makes deducing the line numbers a
> > > little easier.
> > > 
> > 
> > Hmm, I'll take a look sometime in the future, thanks
> 
> Did that... Hmm...
> 
> I am wondering if smp_rcu_assign_pointer() (or more precisely smp_wmb())
> is correctly implemented on octeon platform.
> 
> Try to add in fib_nl_newrule() right after the kzalloc bloc :
> 
> 	rule = kzalloc(ops->rule_size, GFP_KERNEL);
> 	if (rule == NULL) {
> 		err = -ENOMEM;
> 		goto errout;
> 	}
> +       rule->list.next = LIST_POISON1;
> +       rule->list.prev = LIST_POISON2;
> 
> 
> So that we can actually see if the NULL dereference bug you hit becomes
> a "LIST_POISON1" dereference bug...
> 
> 

Reading commit 500c2e1fdbcc2b273bd is interesting...

David Daney added a nudge_writes(), actually doing a "syncw"
instruction, and this seems to be the smp_wmb() this platform should be
using in the first place, not a pure compiler barrier (barrier()) 

So Joe, you might want to change the smp_wmb() call in
rcu_assign_pointer() by the nudge_writes() call, and see what happens...