similar to: [PATCH] net: vhost: improve performance when enable busyloop

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patch improves the guest receive performance from host. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. For avoiding deadlock, change the code to lock the vq one by one and use the VHOST_NET_VQ_XX as a subclass for mutex_lock_nested. With the patch, qemu can set differently the

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patch improves the guest receive performance from host. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. For avoiding deadlock, change the code to lock the vq one by one and use the VHOST_NET_VQ_XX as a subclass for mutex_lock_nested. With the patch, qemu can set differently the

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patches improve the guest receive and transmit performance. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. This patches are splited from previous big patch: http://patchwork.ozlabs.org/patch/934673/ For more performance report, see patch 4. Tonghao Zhang (4): net: vhost:

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patches improve the guest receive and transmit performance. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. For more performance report, see patch 4. v3 -> v4: fix some issues v2 -> v3: This patches are splited from previous big patch:

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patches improve the guest receive and transmit performance. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. For more performance report, see patch 4. v4 -> v5: fix some issues v3 -> v4: fix some issues v2 -> v3: This patches are splited from previous big patch:

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patches improve the guest receive and transmit performance. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. For more performance report, see patch 4. v4 -> v5: fix some issues v3 -> v4: fix some issues v2 -> v3: This patches are splited from previous big patch:

On 2018?07?03? 15:31, Toshiaki Makita wrote: > We may run out of avail rx ring descriptor under heavy load but busypoll > did not detect it so busypoll may have exited prematurely. Avoid this by > checking rx ring full during busypoll. Actually, we're checking whether it was empty in fact? > > Signed-off-by: Toshiaki Makita <makita.toshiaki at lab.ntt.co.jp> > ---

On 2018?07?03? 15:31, Toshiaki Makita wrote: > We may run out of avail rx ring descriptor under heavy load but busypoll > did not detect it so busypoll may have exited prematurely. Avoid this by > checking rx ring full during busypoll. Actually, we're checking whether it was empty in fact? > > Signed-off-by: Toshiaki Makita <makita.toshiaki at lab.ntt.co.jp> > ---

On Wed, Jun 27, 2018 at 10:24:43PM +0800, Jason Wang wrote: > > > On 2018?06?26? 13:17, xiangxia.m.yue at gmail.com wrote: > > From: Tonghao Zhang <xiangxia.m.yue at gmail.com> > > > > This patch improves the guest receive performance from > > host. On the handle_tx side, we poll the sock receive > > queue at the same time. handle_rx do that in the

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patches improve the guest receive performance. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. For more performance report, see patch 4. v6->v7: fix issue and rebase codes: 1. on tx, busypoll will vhost_net_disable/enable_vq rx vq. [This is suggested by Toshiaki Makita

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patches improve the guest receive performance. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. For more performance report, see patch 4. v5->v6: rebase the codes. Tonghao Zhang (4): net: vhost: lock the vqs one by one net: vhost: replace magic number of lock annotation

Under heavy load vhost tx busypoll tend not to suppress vq kicks, which causes poor guest tx performance. The detailed scenario is described in commitlog of patch 2. Rx seems not to have that serious problem, but for consistency I made a similar change on rx to avoid rx wakeups (patch 3). Additionary patch 4 is to avoid rx kicks under heavy load during busypoll. Tx performance is greatly improved

Under heavy load vhost busypoll may run without suppressing notification. For example tx zerocopy callback can push tx work while handle_tx() is running, then busyloop exits due to vhost_has_work() condition and enables notification but immediately reenters handle_tx() because the pushed work was tx. In this case handle_tx() tries to disable notification again, but when using event_idx it by

Under heavy load vhost busypoll may run without suppressing notification. For example tx zerocopy callback can push tx work while handle_tx() is running, then busyloop exits due to vhost_has_work() condition and enables notification but immediately reenters handle_tx() because the pushed work was tx. In this case handle_tx() tries to disable notification again, but when using event_idx it by

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patches improve the guest receive performance. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. For more performance report, see patch 4, 5, 6 Tonghao Zhang (6): net: vhost: lock the vqs one by one net: vhost: replace magic number of lock annotation net: vhost: factor out

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patches improve the guest receive performance. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. For more performance report, see patch 4, 5, 6 Tonghao Zhang (6): net: vhost: lock the vqs one by one net: vhost: replace magic number of lock annotation net: vhost: factor out

From: Tonghao Zhang <xiangxia.m.yue at gmail.com> This patches improve the guest receive performance. On the handle_tx side, we poll the sock receive queue at the same time. handle_rx do that in the same way. For more performance report, see patch 4 Tonghao Zhang (4): net: vhost: lock the vqs one by one net: vhost: replace magic number of lock annotation net: vhost: factor out busy

Hi Jason, On 2018/06/29 18:30, Jason Wang wrote: > On 2018?06?29? 16:09, Toshiaki Makita wrote: ... >> To fix this, poll the work instead of enabling notification when >> busypoll is interrupted by something. IMHO signal_pending() and >> vhost_has_work() are kind of interruptions rather than signals to >> completely cancel the busypoll, so let's run busypoll after

Hi Jason, On 2018/06/29 18:30, Jason Wang wrote: > On 2018?06?29? 16:09, Toshiaki Makita wrote: ... >> To fix this, poll the work instead of enabling notification when >> busypoll is interrupted by something. IMHO signal_pending() and >> vhost_has_work() are kind of interruptions rather than signals to >> completely cancel the busypoll, so let's run busypoll after

On 2018/07/22 3:04, xiangxia.m.yue at gmail.com wrote: > From: Tonghao Zhang <xiangxia.m.yue at gmail.com> > > Factor out generic busy polling logic and will be > used for in tx path in the next patch. And with the patch, > qemu can set differently the busyloop_timeout for rx queue. > > Signed-off-by: Tonghao Zhang <xiangxia.m.yue at gmail.com> > --- ... >