thr3ads.net - Linux Virtualization - [PATCH v4 2/5] vsock/virtio: reduce credit update messages [Sep 2019]

If this information is useful, please help other people find it:
Share via:

Michael S. Tsirkin

2019-Sep-03 04:38 UTC

[PATCH v4 2/5] vsock/virtio: reduce credit update messages

On Wed, Jul 17, 2019 at 01:30:27PM +0200, Stefano Garzarella
wrote:> In order to reduce the number of credit update messages,
> we send them only when the space available seen by the
> transmitter is less than VIRTIO_VSOCK_MAX_PKT_BUF_SIZE.
> 
> Signed-off-by: Stefano Garzarella <sgarzare at redhat.com>
> ---
>  include/linux/virtio_vsock.h            |  1 +
>  net/vmw_vsock/virtio_transport_common.c | 16 +++++++++++++---
>  2 files changed, 14 insertions(+), 3 deletions(-)
> 
> diff --git a/include/linux/virtio_vsock.h b/include/linux/virtio_vsock.h
> index 7d973903f52e..49fc9d20bc43 100644
> --- a/include/linux/virtio_vsock.h
> +++ b/include/linux/virtio_vsock.h
> @@ -41,6 +41,7 @@ struct virtio_vsock_sock {
>  
>  	/* Protected by rx_lock */
>  	u32 fwd_cnt;
> +	u32 last_fwd_cnt;
>  	u32 rx_bytes;
>  	struct list_head rx_queue;
>  };
> diff --git a/net/vmw_vsock/virtio_transport_common.c
b/net/vmw_vsock/virtio_transport_common.c
> index 095221f94786..a85559d4d974 100644
> --- a/net/vmw_vsock/virtio_transport_common.c
> +++ b/net/vmw_vsock/virtio_transport_common.c
> @@ -211,6 +211,7 @@ static void virtio_transport_dec_rx_pkt(struct
virtio_vsock_sock *vvs,
>  void virtio_transport_inc_tx_pkt(struct virtio_vsock_sock *vvs, struct
virtio_vsock_pkt *pkt)
>  {
>  	spin_lock_bh(&vvs->tx_lock);
> +	vvs->last_fwd_cnt = vvs->fwd_cnt;
>  	pkt->hdr.fwd_cnt = cpu_to_le32(vvs->fwd_cnt);
>  	pkt->hdr.buf_alloc = cpu_to_le32(vvs->buf_alloc);
>  	spin_unlock_bh(&vvs->tx_lock);
> @@ -261,6 +262,7 @@ virtio_transport_stream_do_dequeue(struct vsock_sock
*vsk,
>  	struct virtio_vsock_sock *vvs = vsk->trans;
>  	struct virtio_vsock_pkt *pkt;
>  	size_t bytes, total = 0;
> +	u32 free_space;
>  	int err = -EFAULT;
>  
>  	spin_lock_bh(&vvs->rx_lock);
> @@ -291,11 +293,19 @@ virtio_transport_stream_do_dequeue(struct vsock_sock
*vsk,
>  			virtio_transport_free_pkt(pkt);
>  		}
>  	}
> +
> +	free_space = vvs->buf_alloc - (vvs->fwd_cnt -
vvs->last_fwd_cnt);
> +
>  	spin_unlock_bh(&vvs->rx_lock);
>  
> -	/* Send a credit pkt to peer */
> -	virtio_transport_send_credit_update(vsk, VIRTIO_VSOCK_TYPE_STREAM,
> -					    NULL);
> +	/* We send a credit update only when the space available seen
> +	 * by the transmitter is less than VIRTIO_VSOCK_MAX_PKT_BUF_SIZE
This is just repeating what code does though.
Please include the *reason* for the condition.
E.g. here's a better comment:

	/* To reduce number of credit update messages,
	 * don't update credits as long as lots of space is available.
	 * Note: the limit chosen here is arbitrary. Setting the limit
	 * too high causes extra messages. Too low causes transmitter
	 * stalls. As stalls are in theory more expensive than extra
	 * messages, we set the limit to a high value. TODO: experiment
	 * with different values.
	 */

> +	 */
> +	if (free_space < VIRTIO_VSOCK_MAX_PKT_BUF_SIZE) {
> +		virtio_transport_send_credit_update(vsk,
> +						    VIRTIO_VSOCK_TYPE_STREAM,
> +						    NULL);
> +	}
>  
>  	return total;
>  
> -- 
> 2.20.1

Stefano Garzarella

2019-Sep-03 07:31 UTC

head link

[PATCH v4 2/5] vsock/virtio: reduce credit update messages

On Tue, Sep 03, 2019 at 12:38:02AM -0400, Michael S. Tsirkin
wrote:> On Wed, Jul 17, 2019 at 01:30:27PM +0200, Stefano Garzarella wrote:
> > In order to reduce the number of credit update messages,
> > we send them only when the space available seen by the
> > transmitter is less than VIRTIO_VSOCK_MAX_PKT_BUF_SIZE.
> > 
> > Signed-off-by: Stefano Garzarella <sgarzare at redhat.com>
> > ---
> >  include/linux/virtio_vsock.h            |  1 +
> >  net/vmw_vsock/virtio_transport_common.c | 16 +++++++++++++---
> >  2 files changed, 14 insertions(+), 3 deletions(-)
> > 
> > diff --git a/include/linux/virtio_vsock.h
b/include/linux/virtio_vsock.h
> > index 7d973903f52e..49fc9d20bc43 100644
> > --- a/include/linux/virtio_vsock.h
> > +++ b/include/linux/virtio_vsock.h
> > @@ -41,6 +41,7 @@ struct virtio_vsock_sock {
> >  
> >  	/* Protected by rx_lock */
> >  	u32 fwd_cnt;
> > +	u32 last_fwd_cnt;
> >  	u32 rx_bytes;
> >  	struct list_head rx_queue;
> >  };
> > diff --git a/net/vmw_vsock/virtio_transport_common.c
b/net/vmw_vsock/virtio_transport_common.c
> > index 095221f94786..a85559d4d974 100644
> > --- a/net/vmw_vsock/virtio_transport_common.c
> > +++ b/net/vmw_vsock/virtio_transport_common.c
> > @@ -211,6 +211,7 @@ static void virtio_transport_dec_rx_pkt(struct
virtio_vsock_sock *vvs,
> >  void virtio_transport_inc_tx_pkt(struct virtio_vsock_sock *vvs,
struct virtio_vsock_pkt *pkt)
> >  {
> >  	spin_lock_bh(&vvs->tx_lock);
> > +	vvs->last_fwd_cnt = vvs->fwd_cnt;
> >  	pkt->hdr.fwd_cnt = cpu_to_le32(vvs->fwd_cnt);
> >  	pkt->hdr.buf_alloc = cpu_to_le32(vvs->buf_alloc);
> >  	spin_unlock_bh(&vvs->tx_lock);
> > @@ -261,6 +262,7 @@ virtio_transport_stream_do_dequeue(struct
vsock_sock *vsk,
> >  	struct virtio_vsock_sock *vvs = vsk->trans;
> >  	struct virtio_vsock_pkt *pkt;
> >  	size_t bytes, total = 0;
> > +	u32 free_space;
> >  	int err = -EFAULT;
> >  
> >  	spin_lock_bh(&vvs->rx_lock);
> > @@ -291,11 +293,19 @@ virtio_transport_stream_do_dequeue(struct
vsock_sock *vsk,
> >  			virtio_transport_free_pkt(pkt);
> >  		}
> >  	}
> > +
> > +	free_space = vvs->buf_alloc - (vvs->fwd_cnt -
vvs->last_fwd_cnt);
> > +
> >  	spin_unlock_bh(&vvs->rx_lock);
> >  
> > -	/* Send a credit pkt to peer */
> > -	virtio_transport_send_credit_update(vsk, VIRTIO_VSOCK_TYPE_STREAM,
> > -					    NULL);
> > +	/* We send a credit update only when the space available seen
> > +	 * by the transmitter is less than VIRTIO_VSOCK_MAX_PKT_BUF_SIZE
> 
> This is just repeating what code does though.
> Please include the *reason* for the condition.
> E.g. here's a better comment:
> 
> 	/* To reduce number of credit update messages,
> 	 * don't update credits as long as lots of space is available.
> 	 * Note: the limit chosen here is arbitrary. Setting the limit
> 	 * too high causes extra messages. Too low causes transmitter
> 	 * stalls. As stalls are in theory more expensive than extra
> 	 * messages, we set the limit to a high value. TODO: experiment
> 	 * with different values.
> 	 */
> 
Yes, it is better, sorry for that. I'll try to avoid unnecessary comments,
explaining the reason for certain changes.

Since this patch is already queued in net-next, should I send another
patch to fix the comment?

Thanks,
Stefano

Michael S. Tsirkin

2019-Sep-03 07:38 UTC

head link

[PATCH v4 2/5] vsock/virtio: reduce credit update messages

On Tue, Sep 03, 2019 at 09:31:20AM +0200, Stefano Garzarella
wrote:> On Tue, Sep 03, 2019 at 12:38:02AM -0400, Michael S. Tsirkin wrote:
> > On Wed, Jul 17, 2019 at 01:30:27PM +0200, Stefano Garzarella wrote:
> > > In order to reduce the number of credit update messages,
> > > we send them only when the space available seen by the
> > > transmitter is less than VIRTIO_VSOCK_MAX_PKT_BUF_SIZE.
> > > 
> > > Signed-off-by: Stefano Garzarella <sgarzare at redhat.com>
> > > ---
> > >  include/linux/virtio_vsock.h            |  1 +
> > >  net/vmw_vsock/virtio_transport_common.c | 16 +++++++++++++---
> > >  2 files changed, 14 insertions(+), 3 deletions(-)
> > > 
> > > diff --git a/include/linux/virtio_vsock.h
b/include/linux/virtio_vsock.h
> > > index 7d973903f52e..49fc9d20bc43 100644
> > > --- a/include/linux/virtio_vsock.h
> > > +++ b/include/linux/virtio_vsock.h
> > > @@ -41,6 +41,7 @@ struct virtio_vsock_sock {
> > >  
> > >  	/* Protected by rx_lock */
> > >  	u32 fwd_cnt;
> > > +	u32 last_fwd_cnt;
> > >  	u32 rx_bytes;
> > >  	struct list_head rx_queue;
> > >  };
> > > diff --git a/net/vmw_vsock/virtio_transport_common.c
b/net/vmw_vsock/virtio_transport_common.c
> > > index 095221f94786..a85559d4d974 100644
> > > --- a/net/vmw_vsock/virtio_transport_common.c
> > > +++ b/net/vmw_vsock/virtio_transport_common.c
> > > @@ -211,6 +211,7 @@ static void
virtio_transport_dec_rx_pkt(struct virtio_vsock_sock *vvs,
> > >  void virtio_transport_inc_tx_pkt(struct virtio_vsock_sock *vvs,
struct virtio_vsock_pkt *pkt)
> > >  {
> > >  	spin_lock_bh(&vvs->tx_lock);
> > > +	vvs->last_fwd_cnt = vvs->fwd_cnt;
> > >  	pkt->hdr.fwd_cnt = cpu_to_le32(vvs->fwd_cnt);
> > >  	pkt->hdr.buf_alloc = cpu_to_le32(vvs->buf_alloc);
> > >  	spin_unlock_bh(&vvs->tx_lock);
> > > @@ -261,6 +262,7 @@ virtio_transport_stream_do_dequeue(struct
vsock_sock *vsk,
> > >  	struct virtio_vsock_sock *vvs = vsk->trans;
> > >  	struct virtio_vsock_pkt *pkt;
> > >  	size_t bytes, total = 0;
> > > +	u32 free_space;
> > >  	int err = -EFAULT;
> > >  
> > >  	spin_lock_bh(&vvs->rx_lock);
> > > @@ -291,11 +293,19 @@ virtio_transport_stream_do_dequeue(struct
vsock_sock *vsk,
> > >  			virtio_transport_free_pkt(pkt);
> > >  		}
> > >  	}
> > > +
> > > +	free_space = vvs->buf_alloc - (vvs->fwd_cnt -
vvs->last_fwd_cnt);
> > > +
> > >  	spin_unlock_bh(&vvs->rx_lock);
> > >  
> > > -	/* Send a credit pkt to peer */
> > > -	virtio_transport_send_credit_update(vsk,
VIRTIO_VSOCK_TYPE_STREAM,
> > > -					    NULL);
> > > +	/* We send a credit update only when the space available seen
> > > +	 * by the transmitter is less than
VIRTIO_VSOCK_MAX_PKT_BUF_SIZE
> > 
> > This is just repeating what code does though.
> > Please include the *reason* for the condition.
> > E.g. here's a better comment:
> > 
> > 	/* To reduce number of credit update messages,
> > 	 * don't update credits as long as lots of space is available.
> > 	 * Note: the limit chosen here is arbitrary. Setting the limit
> > 	 * too high causes extra messages. Too low causes transmitter
> > 	 * stalls. As stalls are in theory more expensive than extra
> > 	 * messages, we set the limit to a high value. TODO: experiment
> > 	 * with different values.
> > 	 */
> > 
> 
> Yes, it is better, sorry for that. I'll try to avoid unnecessary
comments,
> explaining the reason for certain changes.
> 
> Since this patch is already queued in net-next, should I send another
> patch to fix the comment?
> 
> Thanks,
> Stefano
I just sent a patch like that, pls ack it.

-- 
MST

Maybe Matching Threads

Search for more maybe matching threads

Linux Virtualization - Sep 2019 - [PATCH v4 2/5] vsock/virtio: reduce credit update messages

[PATCH v4 2/5] vsock/virtio: reduce credit update messages

[PATCH v4 2/5] vsock/virtio: reduce credit update messages

[PATCH v4 2/5] vsock/virtio: reduce credit update messages

Maybe Matching Threads