Re: [v2 PATCH 0/2] macvlan: Avoid unnecessary multicast cloning

2016-06-01 Thread David Miller
From: Herbert Xu Date: Wed, 1 Jun 2016 11:42:18 +0800 > This patch tries to improve macvlan multicast performance by > maintaining a filter hash at the macvlan_port level so that we > can quickly determine whether a given packet is needed or not. > > It is preceded

[v2 PATCH 0/2] macvlan: Avoid unnecessary multicast cloning

2016-05-31 Thread Herbert Xu
On Tue, May 31, 2016 at 02:07:13PM -0700, David Miller wrote: > > I think you need to set the vlan->port->mc_filter to all 1's in the > PROMISC/ALLMUTI branch here. > > Otherwise packets won't properly pass your new hash test. Good point. Here's v2. This patch tries to improve macvlan

Re: [PATCH 0/2] macvlan: Avoid unnecessary multicast cloning

2016-05-31 Thread David Miller
From: Herbert Xu Date: Tue, 31 May 2016 08:49:45 +0800 > On Mon, May 30, 2016 at 07:27:59PM +0300, Lennert Buytenhek wrote: >> >> That and stack switches to kworker threads and serialisation on >> the bc_queue queue lock. > > My patch should resolve these problems

Re: [PATCH 0/2] macvlan: Avoid unnecessary multicast cloning

2016-05-30 Thread Herbert Xu
On Mon, May 30, 2016 at 07:27:59PM +0300, Lennert Buytenhek wrote: > > That and stack switches to kworker threads and serialisation on > the bc_queue queue lock. My patch should resolve these problems too since the packet is discarded if nobody is interested in it. Cheers, -- Email: Herbert Xu

Re: [PATCH 0/2] macvlan: Avoid unnecessary multicast cloning

2016-05-30 Thread Lennert Buytenhek
On Mon, May 30, 2016 at 04:17:52PM +0800, Herbert Xu wrote: > > Commit 412ca1550cbecb2c ("macvlan: Move broadcasts into a work queue") > > moved processing of all macvlan multicasts into a work queue. This > > causes a noticable performance regression when there is heavy multicast > > traffic on

[PATCH 0/2] macvlan: Avoid unnecessary multicast cloning

2016-05-30 Thread Herbert Xu
On Fri, May 27, 2016 at 02:44:33AM +0300, Lennert Buytenhek wrote: > Commit 412ca1550cbecb2c ("macvlan: Move broadcasts into a work queue") > moved processing of all macvlan multicasts into a work queue. This > causes a noticable performance regression when there is heavy multicast > traffic on