[net-next] tcp: avoid frag allocation for small frames

tcp_sendmsg() uses select_size() helper to choose skb head size when a
new skb must be allocated.

If GSO is enabled for the socket, current strategy is to force all
payload data to be outside of headroom, in PAGE fragments.

This strategy is not welcome for small packets, wasting memory.

Experiments show that best results are obtained when using 2048 bytes
for skb head (This includes the skb overhead and various headers)

This patch provides better len/truesize ratios for packets sent to
loopback device, and reduce memory needs for in-flight loopback packets,
particularly on arches with big pages.

If a sender sends many 1-byte packets to an unresponsive application,
receiver rmem_alloc will grow faster and will stop queuing these packets
sooner, or will collapse its receive queue to free excess memory.

netperf -t TCP_RR results are improved by ~4 %, and many workloads are
improved as well (tbench, mysql...)

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
---
 net/ipv4/tcp.c |    9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)



--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Sat, 22 Oct 2011 14:23:57 +0200

> tcp_sendmsg() uses select_size() helper to choose skb head size when a
> new skb must be allocated.
> 
> If GSO is enabled for the socket, current strategy is to force all
> payload data to be outside of headroom, in PAGE fragments.
> 
> This strategy is not welcome for small packets, wasting memory.
> 
> Experiments show that best results are obtained when using 2048 bytes
> for skb head (This includes the skb overhead and various headers)
> 
> This patch provides better len/truesize ratios for packets sent to
> loopback device, and reduce memory needs for in-flight loopback packets,
> particularly on arches with big pages.
> 
> If a sender sends many 1-byte packets to an unresponsive application,
> receiver rmem_alloc will grow faster and will stop queuing these packets
> sooner, or will collapse its receive queue to free excess memory.
> 
> netperf -t TCP_RR results are improved by ~4 %, and many workloads are
> improved as well (tbench, mysql...)
> 
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>

Applied, thanks Eric.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On 22 October 2011 05:23, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> tcp_sendmsg() uses select_size() helper to choose skb head size when a
> new skb must be allocated.
>
> If GSO is enabled for the socket, current strategy is to force all
> payload data to be outside of headroom, in PAGE fragments.
>
> This strategy is not welcome for small packets, wasting memory.
>
> Experiments show that best results are obtained when using 2048 bytes
> for skb head (This includes the skb overhead and various headers)
>
> This patch provides better len/truesize ratios for packets sent to
> loopback device, and reduce memory needs for in-flight loopback packets,
> particularly on arches with big pages.
>
> If a sender sends many 1-byte packets to an unresponsive application,
> receiver rmem_alloc will grow faster and will stop queuing these packets
> sooner, or will collapse its receive queue to free excess memory.
>
> netperf -t TCP_RR results are improved by ~4 %, and many workloads are
> improved as well (tbench, mysql...)
>
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> ---
>  net/ipv4/tcp.c |    9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> index 704adad..cd45b44 100644
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -897,9 +897,12 @@ static inline int select_size(const struct sock *sk, int sg)
>        int tmp = tp->mss_cache;
>
>        if (sg) {
> -               if (sk_can_gso(sk))
> -                       tmp = 0;
> -               else {
> +               if (sk_can_gso(sk)) {
> +                       /* Small frames wont use a full page:
> +                        * Payload will immediately follow tcp header.
> +                        */
> +                       tmp = SKB_WITH_OVERHEAD(2048 - MAX_TCP_HEADER);
> +               } else {
>                        int pgbreak = SKB_MAX_HEAD(MAX_TCP_HEADER);
>
>                        if (tmp >= pgbreak &&
>
>



This patch from Eric fixing select_size in tcp.c was queued to be
applied but does not seem to be in net-next tree. Was this somehow
overlooked or have I missed something?

Regards,
Vijay Subramanian
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Le lundi 28 novembre 2011 à 17:04 -0800, Vijay Subramanian a écrit :
> On 22 October 2011 05:23, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> > tcp_sendmsg() uses select_size() helper to choose skb head size when a
> > new skb must be allocated.
> >
> > If GSO is enabled for the socket, current strategy is to force all
> > payload data to be outside of headroom, in PAGE fragments.
> >
> > This strategy is not welcome for small packets, wasting memory.
> >
> > Experiments show that best results are obtained when using 2048 bytes
> > for skb head (This includes the skb overhead and various headers)
> >
> > This patch provides better len/truesize ratios for packets sent to
> > loopback device, and reduce memory needs for in-flight loopback packets,
> > particularly on arches with big pages.
> >
> > If a sender sends many 1-byte packets to an unresponsive application,
> > receiver rmem_alloc will grow faster and will stop queuing these packets
> > sooner, or will collapse its receive queue to free excess memory.
> >
> > netperf -t TCP_RR results are improved by ~4 %, and many workloads are
> > improved as well (tbench, mysql...)
> >
> > Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> > ---
> >  net/ipv4/tcp.c |    9 ++++++---
> >  1 file changed, 6 insertions(+), 3 deletions(-)
> >
> > diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> > index 704adad..cd45b44 100644
> > --- a/net/ipv4/tcp.c
> > +++ b/net/ipv4/tcp.c
> > @@ -897,9 +897,12 @@ static inline int select_size(const struct sock *sk, int sg)
> >        int tmp = tp->mss_cache;
> >
> >        if (sg) {
> > -               if (sk_can_gso(sk))
> > -                       tmp = 0;
> > -               else {
> > +               if (sk_can_gso(sk)) {
> > +                       /* Small frames wont use a full page:
> > +                        * Payload will immediately follow tcp header.
> > +                        */
> > +                       tmp = SKB_WITH_OVERHEAD(2048 - MAX_TCP_HEADER);
> > +               } else {
> >                        int pgbreak = SKB_MAX_HEAD(MAX_TCP_HEADER);
> >
> >                        if (tmp >= pgbreak &&
> >
> >
> 
> 
> 
> This patch from Eric fixing select_size in tcp.c was queued to be
> applied but does not seem to be in net-next tree. Was this somehow
> overlooked or have I missed something?
> 

Good catch Vijay, thanks for noticing !

David, do you want me to respin this ?



--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Tue, 29 Nov 2011 08:51:50 +0100

> David, do you want me to respin this ?

Please do, thank you.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html