Message ID | 1534680686-3108-8-git-send-email-xiangxia.m.yue@gmail.com |
---|---|
State | Deferred, archived |
Delegated to: | David Miller |
Headers | show |
Series | net: vhost: improve performance when enable busyloop | expand |
On 2018年08月19日 20:11, xiangxia.m.yue@gmail.com wrote: > From: Tonghao Zhang <xiangxia.m.yue@gmail.com> > > The patch uses vhost_has_work_pending() to check if > the specified handler is scheduled, because in the most case, > vhost_has_work() return true when other side handler is added > to worker list. Use the vhost_has_work_pending() insead of > vhost_has_work(). > > Topology: > [Host] ->linux bridge -> tap vhost-net ->[Guest] > > TCP_STREAM (netperf): > * Without the patch: 38035.39 Mbps, 3.37 us mean latency > * With the patch: 38409.44 Mbps, 3.34 us mean latency The improvement is not obvious as last version. Do you imply there's some recent changes of vhost that make it faster? Thanks > > Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com> > --- > drivers/vhost/net.c | 9 ++++++--- > 1 file changed, 6 insertions(+), 3 deletions(-) > > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c > index db63ae2..b6939ef 100644 > --- a/drivers/vhost/net.c > +++ b/drivers/vhost/net.c > @@ -487,10 +487,8 @@ static void vhost_net_busy_poll(struct vhost_net *net, > endtime = busy_clock() + busyloop_timeout; > > while (vhost_can_busy_poll(endtime)) { > - if (vhost_has_work(&net->dev)) { > - *busyloop_intr = true; > + if (vhost_has_work(&net->dev)) > break; > - } > > if ((sock_has_rx_data(sock) && > !vhost_vq_avail_empty(&net->dev, rvq)) || > @@ -513,6 +511,11 @@ static void vhost_net_busy_poll(struct vhost_net *net, > !vhost_has_work_pending(&net->dev, VHOST_NET_VQ_RX)) > vhost_net_enable_vq(net, rvq); > > + if (vhost_has_work_pending(&net->dev, > + poll_rx ? > + VHOST_NET_VQ_RX: VHOST_NET_VQ_TX)) > + *busyloop_intr = true; > + > mutex_unlock(&vq->mutex); > } >
On 2018年08月21日 08:33, Jason Wang wrote: > > > On 2018年08月19日 20:11, xiangxia.m.yue@gmail.com wrote: >> From: Tonghao Zhang <xiangxia.m.yue@gmail.com> >> >> The patch uses vhost_has_work_pending() to check if >> the specified handler is scheduled, because in the most case, >> vhost_has_work() return true when other side handler is added >> to worker list. Use the vhost_has_work_pending() insead of >> vhost_has_work(). >> >> Topology: >> [Host] ->linux bridge -> tap vhost-net ->[Guest] >> >> TCP_STREAM (netperf): >> * Without the patch: 38035.39 Mbps, 3.37 us mean latency >> * With the patch: 38409.44 Mbps, 3.34 us mean latency > > The improvement is not obvious as last version. Do you imply there's > some recent changes of vhost that make it faster? > I misunderstood the numbers, please ignore this. It shows less than 1% improvement. I'm not sure it's worth to do so. Can you try bi-directional pktgen to see if it has more obvious effect? Thanks > Thanks > >> >> Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com> >> --- >> drivers/vhost/net.c | 9 ++++++--- >> 1 file changed, 6 insertions(+), 3 deletions(-) >> >> diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c >> index db63ae2..b6939ef 100644 >> --- a/drivers/vhost/net.c >> +++ b/drivers/vhost/net.c >> @@ -487,10 +487,8 @@ static void vhost_net_busy_poll(struct vhost_net >> *net, >> endtime = busy_clock() + busyloop_timeout; >> while (vhost_can_busy_poll(endtime)) { >> - if (vhost_has_work(&net->dev)) { >> - *busyloop_intr = true; >> + if (vhost_has_work(&net->dev)) >> break; >> - } >> if ((sock_has_rx_data(sock) && >> !vhost_vq_avail_empty(&net->dev, rvq)) || >> @@ -513,6 +511,11 @@ static void vhost_net_busy_poll(struct vhost_net >> *net, >> !vhost_has_work_pending(&net->dev, VHOST_NET_VQ_RX)) >> vhost_net_enable_vq(net, rvq); >> + if (vhost_has_work_pending(&net->dev, >> + poll_rx ? >> + VHOST_NET_VQ_RX: VHOST_NET_VQ_TX)) >> + *busyloop_intr = true; >> + >> mutex_unlock(&vq->mutex); >> } > > _______________________________________________ > Virtualization mailing list > Virtualization@lists.linux-foundation.org > https://lists.linuxfoundation.org/mailman/listinfo/virtualization
On Tue, Aug 21, 2018 at 08:47:35AM +0800, Jason Wang wrote: > > > On 2018年08月21日 08:33, Jason Wang wrote: > > > > > > On 2018年08月19日 20:11, xiangxia.m.yue@gmail.com wrote: > > > From: Tonghao Zhang <xiangxia.m.yue@gmail.com> > > > > > > The patch uses vhost_has_work_pending() to check if > > > the specified handler is scheduled, because in the most case, > > > vhost_has_work() return true when other side handler is added > > > to worker list. Use the vhost_has_work_pending() insead of > > > vhost_has_work(). > > > > > > Topology: > > > [Host] ->linux bridge -> tap vhost-net ->[Guest] > > > > > > TCP_STREAM (netperf): > > > * Without the patch: 38035.39 Mbps, 3.37 us mean latency > > > * With the patch: 38409.44 Mbps, 3.34 us mean latency > > > > The improvement is not obvious as last version. Do you imply there's > > some recent changes of vhost that make it faster? > > > > I misunderstood the numbers, please ignore this. > > It shows less than 1% improvement. I'm not sure it's worth to do so. Can you > try bi-directional pktgen to see if it has more obvious effect? > > Thanks Right, this kind of gain is in the noise. Try measuring CPU utilization? > > Thanks > > > > > > > > Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com> > > > --- > > > drivers/vhost/net.c | 9 ++++++--- > > > 1 file changed, 6 insertions(+), 3 deletions(-) > > > > > > diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c > > > index db63ae2..b6939ef 100644 > > > --- a/drivers/vhost/net.c > > > +++ b/drivers/vhost/net.c > > > @@ -487,10 +487,8 @@ static void vhost_net_busy_poll(struct > > > vhost_net *net, > > > endtime = busy_clock() + busyloop_timeout; > > > while (vhost_can_busy_poll(endtime)) { > > > - if (vhost_has_work(&net->dev)) { > > > - *busyloop_intr = true; > > > + if (vhost_has_work(&net->dev)) > > > break; > > > - } > > > if ((sock_has_rx_data(sock) && > > > !vhost_vq_avail_empty(&net->dev, rvq)) || > > > @@ -513,6 +511,11 @@ static void vhost_net_busy_poll(struct > > > vhost_net *net, > > > !vhost_has_work_pending(&net->dev, VHOST_NET_VQ_RX)) > > > vhost_net_enable_vq(net, rvq); > > > + if (vhost_has_work_pending(&net->dev, > > > + poll_rx ? > > > + VHOST_NET_VQ_RX: VHOST_NET_VQ_TX)) > > > + *busyloop_intr = true; > > > + > > > mutex_unlock(&vq->mutex); > > > } > > > > _______________________________________________ > > Virtualization mailing list > > Virtualization@lists.linux-foundation.org > > https://lists.linuxfoundation.org/mailman/listinfo/virtualization
diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c index db63ae2..b6939ef 100644 --- a/drivers/vhost/net.c +++ b/drivers/vhost/net.c @@ -487,10 +487,8 @@ static void vhost_net_busy_poll(struct vhost_net *net, endtime = busy_clock() + busyloop_timeout; while (vhost_can_busy_poll(endtime)) { - if (vhost_has_work(&net->dev)) { - *busyloop_intr = true; + if (vhost_has_work(&net->dev)) break; - } if ((sock_has_rx_data(sock) && !vhost_vq_avail_empty(&net->dev, rvq)) || @@ -513,6 +511,11 @@ static void vhost_net_busy_poll(struct vhost_net *net, !vhost_has_work_pending(&net->dev, VHOST_NET_VQ_RX)) vhost_net_enable_vq(net, rvq); + if (vhost_has_work_pending(&net->dev, + poll_rx ? + VHOST_NET_VQ_RX: VHOST_NET_VQ_TX)) + *busyloop_intr = true; + mutex_unlock(&vq->mutex); }