* Re: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
2024-07-29 12:47 [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands Heng Qi
@ 2024-07-30 1:11 ` Xuan Zhuo
2024-07-30 11:04 ` Eugenio Perez Martin
` (2 subsequent siblings)
3 siblings, 0 replies; 8+ messages in thread
From: Xuan Zhuo @ 2024-07-30 1:11 UTC (permalink / raw)
To: Heng Qi
Cc: virtualization, Eugenio Pérez, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, netdev, Michael S. Tsirkin,
Jason Wang
On Mon, 29 Jul 2024 20:47:55 +0800, Heng Qi <hengqi@linux.alibaba.com> wrote:
> From the virtio spec:
>
> The driver MUST have negotiated the VIRTIO_NET_F_VQ_NOTF_COAL
> feature when issuing commands VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET
> and VIRTIO_NET_CTRL_NOTF_COAL_VQ_GET.
>
> The driver must not send vq notification coalescing commands if
> VIRTIO_NET_F_VQ_NOTF_COAL is not negotiated. This limitation of course
> applies to vq resize.
>
> Fixes: f61fe5f081cf ("virtio-net: fix the vq coalescing setting for vq resize")
> Signed-off-by: Heng Qi <hengqi@linux.alibaba.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
> ---
> drivers/net/virtio_net.c | 29 +++++++++++++++++------------
> 1 file changed, 17 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index 0383a3e136d6..eb115e807882 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -3708,6 +3708,7 @@ static int virtnet_set_ringparam(struct net_device *dev,
> u32 rx_pending, tx_pending;
> struct receive_queue *rq;
> struct send_queue *sq;
> + u32 pkts, usecs;
> int i, err;
>
> if (ring->rx_mini_pending || ring->rx_jumbo_pending)
> @@ -3740,11 +3741,13 @@ static int virtnet_set_ringparam(struct net_device *dev,
> * through the VIRTIO_NET_CTRL_NOTF_COAL_TX_SET command, or, if the driver
> * did not set any TX coalescing parameters, to 0.
> */
> - err = virtnet_send_tx_ctrl_coal_vq_cmd(vi, i,
> - vi->intr_coal_tx.max_usecs,
> - vi->intr_coal_tx.max_packets);
> - if (err)
> - return err;
> + if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL)) {
> + usecs = vi->intr_coal_tx.max_usecs;
> + pkts = vi->intr_coal_tx.max_packets;
> + err = virtnet_send_tx_ctrl_coal_vq_cmd(vi, i, usecs, pkts);
> + if (err)
> + return err;
> + }
> }
>
> if (ring->rx_pending != rx_pending) {
> @@ -3753,13 +3756,15 @@ static int virtnet_set_ringparam(struct net_device *dev,
> return err;
>
> /* The reason is same as the transmit virtqueue reset */
> - mutex_lock(&vi->rq[i].dim_lock);
> - err = virtnet_send_rx_ctrl_coal_vq_cmd(vi, i,
> - vi->intr_coal_rx.max_usecs,
> - vi->intr_coal_rx.max_packets);
> - mutex_unlock(&vi->rq[i].dim_lock);
> - if (err)
> - return err;
> + if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL)) {
> + usecs = vi->intr_coal_rx.max_usecs;
> + pkts = vi->intr_coal_rx.max_packets;
> + mutex_lock(&vi->rq[i].dim_lock);
> + err = virtnet_send_rx_ctrl_coal_vq_cmd(vi, i, usecs, pkts);
> + mutex_unlock(&vi->rq[i].dim_lock);
> + if (err)
> + return err;
> + }
> }
> }
>
> --
> 2.32.0.3.g01195cf9f
>
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
2024-07-29 12:47 [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands Heng Qi
2024-07-30 1:11 ` Xuan Zhuo
@ 2024-07-30 11:04 ` Eugenio Perez Martin
2024-07-31 1:20 ` Jakub Kicinski
2024-07-31 3:11 ` Jason Wang
3 siblings, 0 replies; 8+ messages in thread
From: Eugenio Perez Martin @ 2024-07-30 11:04 UTC (permalink / raw)
To: Heng Qi
Cc: netdev, Michael S. Tsirkin, Jason Wang, virtualization,
David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Xuan Zhuo
On Mon, Jul 29, 2024 at 2:48 PM Heng Qi <hengqi@linux.alibaba.com> wrote:
>
> From the virtio spec:
>
> The driver MUST have negotiated the VIRTIO_NET_F_VQ_NOTF_COAL
> feature when issuing commands VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET
> and VIRTIO_NET_CTRL_NOTF_COAL_VQ_GET.
>
> The driver must not send vq notification coalescing commands if
> VIRTIO_NET_F_VQ_NOTF_COAL is not negotiated. This limitation of course
> applies to vq resize.
>
> Fixes: f61fe5f081cf ("virtio-net: fix the vq coalescing setting for vq resize")
Acked-by: Eugenio Pérez <eperezma@redhat.com>
> Signed-off-by: Heng Qi <hengqi@linux.alibaba.com>
> ---
> drivers/net/virtio_net.c | 29 +++++++++++++++++------------
> 1 file changed, 17 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index 0383a3e136d6..eb115e807882 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -3708,6 +3708,7 @@ static int virtnet_set_ringparam(struct net_device *dev,
> u32 rx_pending, tx_pending;
> struct receive_queue *rq;
> struct send_queue *sq;
> + u32 pkts, usecs;
> int i, err;
>
> if (ring->rx_mini_pending || ring->rx_jumbo_pending)
> @@ -3740,11 +3741,13 @@ static int virtnet_set_ringparam(struct net_device *dev,
> * through the VIRTIO_NET_CTRL_NOTF_COAL_TX_SET command, or, if the driver
> * did not set any TX coalescing parameters, to 0.
> */
> - err = virtnet_send_tx_ctrl_coal_vq_cmd(vi, i,
> - vi->intr_coal_tx.max_usecs,
> - vi->intr_coal_tx.max_packets);
> - if (err)
> - return err;
> + if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL)) {
> + usecs = vi->intr_coal_tx.max_usecs;
> + pkts = vi->intr_coal_tx.max_packets;
> + err = virtnet_send_tx_ctrl_coal_vq_cmd(vi, i, usecs, pkts);
> + if (err)
> + return err;
> + }
> }
>
> if (ring->rx_pending != rx_pending) {
> @@ -3753,13 +3756,15 @@ static int virtnet_set_ringparam(struct net_device *dev,
> return err;
>
> /* The reason is same as the transmit virtqueue reset */
> - mutex_lock(&vi->rq[i].dim_lock);
> - err = virtnet_send_rx_ctrl_coal_vq_cmd(vi, i,
> - vi->intr_coal_rx.max_usecs,
> - vi->intr_coal_rx.max_packets);
> - mutex_unlock(&vi->rq[i].dim_lock);
> - if (err)
> - return err;
> + if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL)) {
> + usecs = vi->intr_coal_rx.max_usecs;
> + pkts = vi->intr_coal_rx.max_packets;
> + mutex_lock(&vi->rq[i].dim_lock);
> + err = virtnet_send_rx_ctrl_coal_vq_cmd(vi, i, usecs, pkts);
> + mutex_unlock(&vi->rq[i].dim_lock);
> + if (err)
> + return err;
> + }
> }
> }
>
> --
> 2.32.0.3.g01195cf9f
>
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
2024-07-29 12:47 [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands Heng Qi
2024-07-30 1:11 ` Xuan Zhuo
2024-07-30 11:04 ` Eugenio Perez Martin
@ 2024-07-31 1:20 ` Jakub Kicinski
2024-07-31 3:13 ` Jason Wang
2024-07-31 11:30 ` Heng Qi
2024-07-31 3:11 ` Jason Wang
3 siblings, 2 replies; 8+ messages in thread
From: Jakub Kicinski @ 2024-07-31 1:20 UTC (permalink / raw)
To: Heng Qi
Cc: netdev, Michael S. Tsirkin, Jason Wang, virtualization,
Eugenio Pérez, David S. Miller, Eric Dumazet, Paolo Abeni,
Xuan Zhuo
On Mon, 29 Jul 2024 20:47:55 +0800 Heng Qi wrote:
> Subject: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
subject currently reads like this is an optimization, could you
rephrase?
> From the virtio spec:
>
> The driver MUST have negotiated the VIRTIO_NET_F_VQ_NOTF_COAL
> feature when issuing commands VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET
> and VIRTIO_NET_CTRL_NOTF_COAL_VQ_GET.
>
> The driver must not send vq notification coalescing commands if
> VIRTIO_NET_F_VQ_NOTF_COAL is not negotiated. This limitation of course
> applies to vq resize.
>
> Fixes: f61fe5f081cf ("virtio-net: fix the vq coalescing setting for vq resize")
> Signed-off-by: Heng Qi <hengqi@linux.alibaba.com>
> ---
> drivers/net/virtio_net.c | 29 +++++++++++++++++------------
> 1 file changed, 17 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index 0383a3e136d6..eb115e807882 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -3708,6 +3708,7 @@ static int virtnet_set_ringparam(struct net_device *dev,
> u32 rx_pending, tx_pending;
> struct receive_queue *rq;
> struct send_queue *sq;
> + u32 pkts, usecs;
> int i, err;
>
> if (ring->rx_mini_pending || ring->rx_jumbo_pending)
> @@ -3740,11 +3741,13 @@ static int virtnet_set_ringparam(struct net_device *dev,
> * through the VIRTIO_NET_CTRL_NOTF_COAL_TX_SET command, or, if the driver
> * did not set any TX coalescing parameters, to 0.
> */
> - err = virtnet_send_tx_ctrl_coal_vq_cmd(vi, i,
> - vi->intr_coal_tx.max_usecs,
> - vi->intr_coal_tx.max_packets);
> - if (err)
> - return err;
> + if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL)) {
> + usecs = vi->intr_coal_tx.max_usecs;
> + pkts = vi->intr_coal_tx.max_packets;
> + err = virtnet_send_tx_ctrl_coal_vq_cmd(vi, i, usecs, pkts);
> + if (err)
> + return err;
Can you check the feature inside the
virtnet_send_.x_ctrl_coal_vq_cmd() helpers?
5 levels of indentation is a bit much
--
pw-bot: cr
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
2024-07-31 1:20 ` Jakub Kicinski
@ 2024-07-31 3:13 ` Jason Wang
2024-07-31 11:22 ` Heng Qi
2024-07-31 11:30 ` Heng Qi
1 sibling, 1 reply; 8+ messages in thread
From: Jason Wang @ 2024-07-31 3:13 UTC (permalink / raw)
To: Jakub Kicinski
Cc: Heng Qi, netdev, Michael S. Tsirkin, virtualization,
Eugenio Pérez, David S. Miller, Eric Dumazet, Paolo Abeni,
Xuan Zhuo
On Wed, Jul 31, 2024 at 9:20 AM Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Mon, 29 Jul 2024 20:47:55 +0800 Heng Qi wrote:
> > Subject: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
>
> subject currently reads like this is an optimization, could you
> rephrase?
It might be "virtio-net: unbreak vq resizing when coalescing is not
negotiated" ?
Thanks
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
2024-07-31 3:13 ` Jason Wang
@ 2024-07-31 11:22 ` Heng Qi
0 siblings, 0 replies; 8+ messages in thread
From: Heng Qi @ 2024-07-31 11:22 UTC (permalink / raw)
To: Jason Wang
Cc: netdev, Michael S. Tsirkin, virtualization, Eugenio Pérez,
David S. Miller, Eric Dumazet, Paolo Abeni, Xuan Zhuo,
Jakub Kicinski
On Wed, 31 Jul 2024 11:13:03 +0800, Jason Wang <jasowang@redhat.com> wrote:
> On Wed, Jul 31, 2024 at 9:20 AM Jakub Kicinski <kuba@kernel.org> wrote:
> >
> > On Mon, 29 Jul 2024 20:47:55 +0800 Heng Qi wrote:
> > > Subject: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
> >
> > subject currently reads like this is an optimization, could you
> > rephrase?
>
> It might be "virtio-net: unbreak vq resizing when coalescing is not
> negotiated" ?
It's great.
Thanks.
>
> Thanks
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
2024-07-31 1:20 ` Jakub Kicinski
2024-07-31 3:13 ` Jason Wang
@ 2024-07-31 11:30 ` Heng Qi
1 sibling, 0 replies; 8+ messages in thread
From: Heng Qi @ 2024-07-31 11:30 UTC (permalink / raw)
To: Jakub Kicinski
Cc: netdev, Michael S. Tsirkin, Jason Wang, virtualization,
EugenioPérez, David S. Miller, Eric Dumazet, Paolo Abeni,
Xuan Zhuo
On Tue, 30 Jul 2024 18:20:20 -0700, Jakub Kicinski <kuba@kernel.org> wrote:
> On Mon, 29 Jul 2024 20:47:55 +0800 Heng Qi wrote:
> > Subject: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
>
> subject currently reads like this is an optimization, could you
> rephrase?
Jason's rephrase will be used.
>
> > From the virtio spec:
> >
> > The driver MUST have negotiated the VIRTIO_NET_F_VQ_NOTF_COAL
> > feature when issuing commands VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET
> > and VIRTIO_NET_CTRL_NOTF_COAL_VQ_GET.
> >
> > The driver must not send vq notification coalescing commands if
> > VIRTIO_NET_F_VQ_NOTF_COAL is not negotiated. This limitation of course
> > applies to vq resize.
> >
> > Fixes: f61fe5f081cf ("virtio-net: fix the vq coalescing setting for vq resize")
> > Signed-off-by: Heng Qi <hengqi@linux.alibaba.com>
> > ---
> > drivers/net/virtio_net.c | 29 +++++++++++++++++------------
> > 1 file changed, 17 insertions(+), 12 deletions(-)
> >
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > index 0383a3e136d6..eb115e807882 100644
> > --- a/drivers/net/virtio_net.c
> > +++ b/drivers/net/virtio_net.c
> > @@ -3708,6 +3708,7 @@ static int virtnet_set_ringparam(struct net_device *dev,
> > u32 rx_pending, tx_pending;
> > struct receive_queue *rq;
> > struct send_queue *sq;
> > + u32 pkts, usecs;
> > int i, err;
> >
> > if (ring->rx_mini_pending || ring->rx_jumbo_pending)
> > @@ -3740,11 +3741,13 @@ static int virtnet_set_ringparam(struct net_device *dev,
> > * through the VIRTIO_NET_CTRL_NOTF_COAL_TX_SET command, or, if the driver
> > * did not set any TX coalescing parameters, to 0.
> > */
> > - err = virtnet_send_tx_ctrl_coal_vq_cmd(vi, i,
> > - vi->intr_coal_tx.max_usecs,
> > - vi->intr_coal_tx.max_packets);
> > - if (err)
> > - return err;
> > + if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_VQ_NOTF_COAL)) {
> > + usecs = vi->intr_coal_tx.max_usecs;
> > + pkts = vi->intr_coal_tx.max_packets;
> > + err = virtnet_send_tx_ctrl_coal_vq_cmd(vi, i, usecs, pkts);
> > + if (err)
> > + return err;
>
> Can you check the feature inside the
> virtnet_send_.x_ctrl_coal_vq_cmd() helpers?
> 5 levels of indentation is a bit much
Makes sense. Will update in the next version.
Thanks.
> --
> pw-bot: cr
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands
2024-07-29 12:47 [PATCH net] virtio_net: Avoid sending unnecessary vq coalescing commands Heng Qi
` (2 preceding siblings ...)
2024-07-31 1:20 ` Jakub Kicinski
@ 2024-07-31 3:11 ` Jason Wang
3 siblings, 0 replies; 8+ messages in thread
From: Jason Wang @ 2024-07-31 3:11 UTC (permalink / raw)
To: Heng Qi
Cc: netdev, Michael S. Tsirkin, virtualization, Eugenio Pérez,
David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Xuan Zhuo
On Mon, Jul 29, 2024 at 8:48 PM Heng Qi <hengqi@linux.alibaba.com> wrote:
>
> From the virtio spec:
>
> The driver MUST have negotiated the VIRTIO_NET_F_VQ_NOTF_COAL
> feature when issuing commands VIRTIO_NET_CTRL_NOTF_COAL_VQ_SET
> and VIRTIO_NET_CTRL_NOTF_COAL_VQ_GET.
>
> The driver must not send vq notification coalescing commands if
> VIRTIO_NET_F_VQ_NOTF_COAL is not negotiated. This limitation of course
> applies to vq resize.
>
> Fixes: f61fe5f081cf ("virtio-net: fix the vq coalescing setting for vq resize")
> Signed-off-by: Heng Qi <hengqi@linux.alibaba.com>
> ---
Acked-by: Jason Wang <jasowang@redhat.com>
Thanks
^ permalink raw reply [flat|nested] 8+ messages in thread