Re: [PATCH] bridge: fix device delete race.

2006-01-31 Thread David S. Miller
From: Herbert Xu <[EMAIL PROTECTED]> Date: Wed, 1 Feb 2006 12:53:45 +1100 > Sorry which original fix is this? The very first one is still open > to the same race that started this thread. The one adding the deleted flag and removing the del_timer_sync() calls which you said was fine. - To unsubsc

Re: [PATCH] bridge: fix device delete race.

2006-01-31 Thread Herbert Xu
On Tue, Jan 31, 2006 at 05:43:46PM -0800, David S. Miller wrote: > > I'm going to apply Stephen's original fix, and I'd like you guys > to work on a more comprehensive version of the bridge netfilter > race fix, ok? Sorry which original fix is this? The very first one is still open to the same rac

Re: [PATCH] bridge: fix device delete race.

2006-01-31 Thread David S. Miller
From: Herbert Xu <[EMAIL PROTECTED]> Date: Wed, 1 Feb 2006 12:05:58 +1100 > On Tue, Jan 31, 2006 at 04:19:11PM -0800, Stephen Hemminger wrote: > > > > This should close a race where bridge interface is deleted while CPU > > was away doing it's netfilter stuff... > > This patch looks good to me.

Re: [PATCH] bridge: fix device delete race.

2006-01-31 Thread Herbert Xu
On Tue, Jan 31, 2006 at 04:19:11PM -0800, Stephen Hemminger wrote: > > This should close a race where bridge interface is deleted while CPU > was away doing it's netfilter stuff... This patch looks good to me. BTW, there are a lot more places where we enter netfilter in the bridging code (especi

Re: [PATCH] bridge: fix device delete race.

2006-01-31 Thread Stephen Hemminger
> BTW, can NF_BR_PRE_ROUTING return with NF_QUEUE? If it can, then we'll > need to recheck the validity of br_port in br_handle_frame_finish. This should close a race where bridge interface is deleted while CPU was away doing it's netfilter stuff... Untested.. --- sky2-1.0.orig/net/bridge/br_

Re: [PATCH] bridge: fix device delete race.

2006-01-31 Thread Herbert Xu
On Tue, Jan 31, 2006 at 02:41:13PM -0800, Stephen Hemminger wrote: > This is a simpler fix for the two races in bridge device removal. > The Xen race of delif and notify is managed now by a new deleted flag. > No need for barriers or other locking because of rtnl mutex. > > The del_timer_sync()'s

[PATCH] bridge: fix device delete race.

2006-01-31 Thread Stephen Hemminger
This is a simpler fix for the two races in bridge device removal. The Xen race of delif and notify is managed now by a new deleted flag. No need for barriers or other locking because of rtnl mutex. The del_timer_sync()'s are unnecessary, because br_stp_disable_port delete's the timers, and they wi