On Sun, Dec 09, 2001 at 05:52:37PM +0100, Bart De Schuymer wrote:
> I had a look at the code. All looks logical, and better readable than the
> old code.
OK. I'll submit it for 2.4 as soon as I hear back from DaveM about the
other submitted patches.
> But I think there is still a race condition in it on an smp box. I'm not
> sure if you know about this, so here it goes:
Yes, I've noticed. In general it's never really safe to grab a pointer
without holding a proper lock.. it really wants a nicer solution.
> To solve this I think we need a spinlock in the net_device struct.
That sounds a bit baroque. What if we change the rules so that br_del_if
(and more generally, any code that modifies port_list) has to run under
BR_NETPROTO_LOCK? That sounds clean enough.. and would also enable us to
get rid of all the "called under bridge lock" mess in the spanning tree
code.
cheers,
Lennert
_______________________________________________
Bridge mailing list
[EMAIL PROTECTED]
http://www.math.leidenuniv.nl/mailman/listinfo/bridge
