[PATCH net-next 0/2] Performances improvement for IPv6 Segment Routing

[PATCH net-next 0/2] Performances improvement for IPv6 Segment Routing

[David Lebrun]

This patch series improves the performances of IPv6 SR by optimizing skb head
reallocation and extending the use of dst_cache. The overall performances improve
by 35%.

Before patch series (SRH encap):
Result: OK: 7348320(c7347271+d1048) usec, 5000000 (1000byte,0frags)
  680427pps 5443Mb/sec (5443416000bps) errors: 0

After patch series (SRH encap):
Result: OK: 4774543(c4774084+d459) usec, 5000000 (1000byte,0frags)
  1047220pps 8377Mb/sec (8377760000bps) errors: 0

Baseline for plain IPv6 forwarding:
Result: OK: 4244144(c4243722+d422) usec, 5000000 (1000byte,0frags)
  1178093pps 9424Mb/sec (9424744000bps) errors: 0

David Lebrun (2):
  ipv6: sr: expand skb head only if necessary
  ipv6: sr: use dst_cache in seg6_input

 net/ipv6/seg6_iptunnel.c | 31 ++++++++++++++++++++++++++++---
 1 file changed, 28 insertions(+), 3 deletions(-)

-- 
2.10.2

[PATCH net-next 1/2] ipv6: sr: expand skb head only if necessary

[David Lebrun]

To insert or encapsulate a packet with an SRH, we need a large enough skb
headroom. Currently, we are using pskb_expand_head to inconditionally increase
the size of the headroom by the amount needed by the SRH (and IPv6 header).
If this reallocation is performed by another CPU than the one that initially
allocated the skb, then when the initial CPU kfree the skb, it will enter the
__slab_free slowpath, impacting performances.

This patch replaces pskb_expand_head with skb_cow_head, that will reallocate the
skb head only if the headroom is not large enough.

Performances for SRH encapsulation before the patch:
Result: OK: 7348320(c7347271+d1048) usec, 5000000 (1000byte,0frags)
  680427pps 5443Mb/sec (5443416000bps) errors: 0

Performances after the patch:
Result: OK: 5656067(c5655678+d388) usec, 5000000 (1000byte,0frags)
  884006pps 7072Mb/sec (7072048000bps) errors: 0

Signed-off-by: David Lebrun <david.lebrun@uclouvain.be>
---
 net/ipv6/seg6_iptunnel.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/ipv6/seg6_iptunnel.c b/net/ipv6/seg6_iptunnel.c
index 8558225..dda2c51 100644
--- a/net/ipv6/seg6_iptunnel.c
+++ b/net/ipv6/seg6_iptunnel.c
@@ -105,7 +105,7 @@ static int seg6_do_srh_encap(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
 	hdrlen = (osrh->hdrlen + 1) << 3;
 	tot_len = hdrlen + sizeof(*hdr);
 
-	err = pskb_expand_head(skb, tot_len, 0, GFP_ATOMIC);
+	err = skb_cow_head(skb, tot_len);
 	if (unlikely(err))
 		return err;
 
@@ -156,7 +156,7 @@ static int seg6_do_srh_inline(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
 
 	hdrlen = (osrh->hdrlen + 1) << 3;
 
-	err = pskb_expand_head(skb, hdrlen, 0, GFP_ATOMIC);
+	err = skb_cow_head(skb, hdrlen);
 	if (unlikely(err))
 		return err;
 
-- 
2.10.2

[PATCH net-next 2/2] ipv6: sr: use dst_cache in seg6_input

[David Lebrun]

We already use dst_cache in seg6_output, when handling locally generated
packets. We extend it in seg6_input, to also handle forwarded packets, and avoid
unnecessary fib lookups.

Performances for SRH encapsulation before the patch:
Result: OK: 5656067(c5655678+d388) usec, 5000000 (1000byte,0frags)
  884006pps 7072Mb/sec (7072048000bps) errors: 0

Performances after the patch:
Result: OK: 4774543(c4774084+d459) usec, 5000000 (1000byte,0frags)
  1047220pps 8377Mb/sec (8377760000bps) errors: 0

Signed-off-by: David Lebrun <david.lebrun@uclouvain.be>
---
 net/ipv6/seg6_iptunnel.c | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)

diff --git a/net/ipv6/seg6_iptunnel.c b/net/ipv6/seg6_iptunnel.c
index dda2c51..30ef1d1 100644
--- a/net/ipv6/seg6_iptunnel.c
+++ b/net/ipv6/seg6_iptunnel.c
@@ -237,6 +237,9 @@ static int seg6_do_srh(struct sk_buff *skb)
 
 static int seg6_input(struct sk_buff *skb)
 {
+	struct dst_entry *orig_dst = skb_dst(skb);
+	struct dst_entry *dst = NULL;
+	struct seg6_lwt *slwt;
 	int err;
 
 	err = seg6_do_srh(skb);
@@ -245,8 +248,30 @@ static int seg6_input(struct sk_buff *skb)
 		return err;
 	}
 
+	slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
+
+#ifdef CONFIG_DST_CACHE
+	preempt_disable();
+	dst = dst_cache_get(&slwt->cache);
+	preempt_enable();
+#endif
+
 	skb_dst_drop(skb);
-	ip6_route_input(skb);
+
+	if (!dst) {
+		ip6_route_input(skb);
+#ifdef CONFIG_DST_CACHE
+		dst = skb_dst(skb);
+		if (!dst->error) {
+			preempt_disable();
+			dst_cache_set_ip6(&slwt->cache, dst,
+					  &ipv6_hdr(skb)->saddr);
+			preempt_enable();
+		}
+#endif
+	} else {
+		skb_dst_set(skb, dst);
+	}
 
 	return dst_input(skb);
 }
-- 
2.10.2

Re: [PATCH net-next 0/2] Performances improvement for IPv6 Segment Routing

[David Miller]

From: David Lebrun <david.lebrun@uclouvain.be>
Date: Fri, 24 Mar 2017 10:46:25 +0100

> This patch series improves the performances of IPv6 SR by optimizing skb head
> reallocation and extending the use of dst_cache. The overall performances improve
> by 35%.

Series applied, thanks.

I think you should make SEG6_IPTUNNEL select DST_CACHE and then remove all of
these ugly ifdefs.

Thanks.

Re: [PATCH net-next 0/2] Performances improvement for IPv6 Segment Routing

[David Lebrun]

[-- Attachment #1.1: Type: text/plain, Size: 243 bytes --]

On 03/24/2017 10:48 PM, David Miller wrote:
> Series applied, thanks.
> 
> I think you should make SEG6_IPTUNNEL select DST_CACHE and then remove all of
> these ugly ifdefs.

Thanks

Agreed, I'll submit a patch for that :)

David


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 195 bytes --]