patch: tc action cleanups

patch: tc action cleanups

[jamal]

[-- Attachment #1: Type: text/plain, Size: 267 bytes --]


Heres one i think you didnt resurect Patrick. If ok, Dave please apply.
Other than that i cant seem to find anything at issue with the core
(havent looked at the individual actions other than to follow changed
code from the core).
Very well done.

cheers,
jamal





[-- Attachment #2: bl_p --]
[-- Type: text/plain, Size: 336 bytes --]

--- a/include/net/pkt_act.h	2005/01/16 13:13:47	1.1
+++ b/include/net/pkt_act.h	2005/01/16 13:15:12
@@ -223,8 +223,10 @@
 	struct tcf_st *p = NULL;
 	if (index && (p = tcf_hash_lookup(index)) != NULL) {
 		if (bind) {
+			spin_lock_bh(&p->lock);
 			p->bindcnt++;
 			p->refcnt++;
+			spin_unlock_bh(&p->lock);
 		}
 		a->priv = p;
 	}

Re: patch: tc action cleanups

[Patrick McHardy]

jamal wrote:

>Heres one i think you didnt resurect Patrick. If ok, Dave please apply.
>Other than that i cant seem to find anything at issue with the core
>(havent looked at the individual actions other than to follow changed
>code from the core).
>
bindcnt/refcnt changes are protected by the rtnl, no need for extra locking.
The lock only needs to protect data shared with the "act" function.

>Very well done.
>
A couple additional fixes are coming up later tonight.

Regards
Patrick

Re: patch: tc action cleanups

[jamal]

On Sun, 2005-01-16 at 13:29, Patrick McHardy wrote:
> jamal wrote:
> 
> >Heres one i think you didnt resurect Patrick. If ok, Dave please apply.
> >Other than that i cant seem to find anything at issue with the core
> >(havent looked at the individual actions other than to follow changed
> >code from the core).
> >
> bindcnt/refcnt changes are protected by the rtnl, no need for extra locking.
> The lock only needs to protect data shared with the "act" function.

Very clever ;-> maybe add a comment then to that effect in your next
batch? Dave ignore that patch.

cheers,
jamal