Re: performance bug in virtio net xdp

2020-05-06 Thread Michael S. Tsirkin
On Wed, May 06, 2020 at 04:37:41PM +0800, Jason Wang wrote: > > On 2020/5/6 下午4:08, Michael S. Tsirkin wrote: > > So for mergeable bufs, we use ewma machinery to guess the correct buffer > > size. If we don't guess correctly, XDP has to do aggressive copies. > > > > Problem is, xdp paths do not

Re: performance bug in virtio net xdp

2020-05-06 Thread Michael S. Tsirkin
On Wed, May 06, 2020 at 10:37:57AM +0200, Jesper Dangaard Brouer wrote: > On Wed, 6 May 2020 04:08:27 -0400 > "Michael S. Tsirkin" wrote: > > > So for mergeable bufs, we use ewma machinery to guess the correct buffer > > size. If we don't guess correctly, XDP has to do aggressive copies. > > >

Re: performance bug in virtio net xdp

2020-05-06 Thread Jason Wang
On 2020/5/6 下午4:08, Michael S. Tsirkin wrote: So for mergeable bufs, we use ewma machinery to guess the correct buffer size. If we don't guess correctly, XDP has to do aggressive copies. Problem is, xdp paths do not update the ewma at all, except sometimes with XDP_PASS. So whatever we happen

Re: performance bug in virtio net xdp

2020-05-06 Thread Jesper Dangaard Brouer
On Wed, 6 May 2020 04:08:27 -0400 "Michael S. Tsirkin" wrote: > So for mergeable bufs, we use ewma machinery to guess the correct buffer > size. If we don't guess correctly, XDP has to do aggressive copies. > > Problem is, xdp paths do not update the ewma at all, except > sometimes with

performance bug in virtio net xdp

2020-05-06 Thread Michael S. Tsirkin
So for mergeable bufs, we use ewma machinery to guess the correct buffer size. If we don't guess correctly, XDP has to do aggressive copies. Problem is, xdp paths do not update the ewma at all, except sometimes with XDP_PASS. So whatever we happen to have before we attach XDP, will mostly stay