netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP
@ 2013-10-19 18:42 Eric Dumazet
  2013-10-19 18:42 ` [PATCH net-next 1/6] ipv4: gso: send_check() & segment() cleanups Eric Dumazet
                   ` (6 more replies)
  0 siblings, 7 replies; 10+ messages in thread
From: Eric Dumazet @ 2013-10-19 18:42 UTC (permalink / raw)
  To: David S. Miller; +Cc: netdev, Jerry Chu, Tom Herbert, Eric Dumazet

This patch serie implements GSO/TSO support for IPIP

David, please note it applies after "ipv4: gso: send_check() & segment() cleanups"
( http://patchwork.ozlabs.org/patch/284714/ )

Broadcom bnx2x driver is now enabled for TSO support of IPIP traffic

Before patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      3357.88   5.09     3.70     2.983   2.167

After patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      8532.40   2.55     7.73     0.588   1.781
    

Eric Dumazet (6):
  ipv4: gso: send_check() & segment() cleanups
  net: generalize skb_segment()
  ipv4: generalize gre_handle_offloads
  ipv4: gso: make inet_gso_segment() stackable
  ipip: add GSO/TSO support
  bnx2x: add TSO support for IPIP

 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c |  4 +-
 include/linux/netdev_features.h                  |  2 +
 include/linux/skbuff.h                           | 13 ++++--
 include/net/gre.h                                |  8 +++-
 include/net/ip_tunnels.h                         |  3 ++
 net/core/dev.c                                   |  2 +
 net/core/ethtool.c                               |  1 +
 net/core/skbuff.c                                | 22 ++-------
 net/ipv4/af_inet.c                               | 58 ++++++++++++++++--------
 net/ipv4/gre_demux.c                             | 29 ------------
 net/ipv4/gre_offload.c                           |  3 +-
 net/ipv4/ip_tunnel_core.c                        | 33 ++++++++++++++
 net/ipv4/ipip.c                                  | 11 +++--
 net/ipv4/tcp_offload.c                           |  1 +
 net/ipv4/udp_offload.c                           |  1 +
 net/ipv6/ip6_offload.c                           |  1 +
 net/ipv6/udp_offload.c                           |  1 +
 net/mpls/mpls_gso.c                              |  1 +
 18 files changed, 116 insertions(+), 78 deletions(-)

-- 
1.8.4

^ permalink raw reply	[flat|nested] 10+ messages in thread

* [PATCH net-next 1/6] ipv4: gso: send_check() & segment() cleanups
  2013-10-19 18:42 [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP Eric Dumazet
@ 2013-10-19 18:42 ` Eric Dumazet
  2013-10-19 18:42 ` [PATCH net-next 2/6] net: generalize skb_segment() Eric Dumazet
                   ` (5 subsequent siblings)
  6 siblings, 0 replies; 10+ messages in thread
From: Eric Dumazet @ 2013-10-19 18:42 UTC (permalink / raw)
  To: David S. Miller; +Cc: netdev, Jerry Chu, Tom Herbert, Eric Dumazet

inet_gso_segment() and inet_gso_send_check() are called by
skb_mac_gso_segment() under rcu lock, no need to use
rcu_read_lock() / rcu_read_unlock()

Avoid calling ip_hdr() twice per function.

We can use ip_send_check() helper.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 net/ipv4/af_inet.c | 24 +++++++++++-------------
 1 file changed, 11 insertions(+), 13 deletions(-)

diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 35913fb..4f8cd4f 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -1254,20 +1254,19 @@ static int inet_gso_send_check(struct sk_buff *skb)
 	if (ihl < sizeof(*iph))
 		goto out;
 
+	proto = iph->protocol;
+
+	/* Warning: after this point, iph might be no longer valid */
 	if (unlikely(!pskb_may_pull(skb, ihl)))
 		goto out;
-
 	__skb_pull(skb, ihl);
+
 	skb_reset_transport_header(skb);
-	iph = ip_hdr(skb);
-	proto = iph->protocol;
 	err = -EPROTONOSUPPORT;
 
-	rcu_read_lock();
 	ops = rcu_dereference(inet_offloads[proto]);
 	if (likely(ops && ops->callbacks.gso_send_check))
 		err = ops->callbacks.gso_send_check(skb);
-	rcu_read_unlock();
 
 out:
 	return err;
@@ -1305,23 +1304,23 @@ static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
 	if (ihl < sizeof(*iph))
 		goto out;
 
+	id = ntohs(iph->id);
+	proto = iph->protocol;
+
+	/* Warning: after this point, iph might be no longer valid */
 	if (unlikely(!pskb_may_pull(skb, ihl)))
 		goto out;
+	__skb_pull(skb, ihl);
 
 	tunnel = !!skb->encapsulation;
 
-	__skb_pull(skb, ihl);
 	skb_reset_transport_header(skb);
-	iph = ip_hdr(skb);
-	id = ntohs(iph->id);
-	proto = iph->protocol;
+
 	segs = ERR_PTR(-EPROTONOSUPPORT);
 
-	rcu_read_lock();
 	ops = rcu_dereference(inet_offloads[proto]);
 	if (likely(ops && ops->callbacks.gso_segment))
 		segs = ops->callbacks.gso_segment(skb, features);
-	rcu_read_unlock();
 
 	if (IS_ERR_OR_NULL(segs))
 		goto out;
@@ -1339,8 +1338,7 @@ static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
 			iph->id = htons(id++);
 		}
 		iph->tot_len = htons(skb->len - skb->mac_len);
-		iph->check = 0;
-		iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
+		ip_send_check(iph);
 	} while ((skb = skb->next));
 
 out:
-- 
1.8.4

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH net-next 2/6] net: generalize skb_segment()
  2013-10-19 18:42 [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP Eric Dumazet
  2013-10-19 18:42 ` [PATCH net-next 1/6] ipv4: gso: send_check() & segment() cleanups Eric Dumazet
@ 2013-10-19 18:42 ` Eric Dumazet
  2013-10-19 18:42 ` [PATCH net-next 3/6] ipv4: generalize gre_handle_offloads Eric Dumazet
                   ` (4 subsequent siblings)
  6 siblings, 0 replies; 10+ messages in thread
From: Eric Dumazet @ 2013-10-19 18:42 UTC (permalink / raw)
  To: David S. Miller; +Cc: netdev, Jerry Chu, Tom Herbert, Eric Dumazet

While implementing GSO/TSO support for IPIP, I found skb_segment()
was assuming network header was immediately following mac header.

Its not really true in the case inet_gso_segment() is stacked :
By the time tcp_gso_segment() is called, network header points
to the inner IP header.

Let's instead assume nothing and pick the current offsets found in
original skb, we have skb_headers_offset_update() helper for that.

Also move the csum_start update inside skb_headers_offset_update()

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 net/core/skbuff.c | 22 +++++-----------------
 1 file changed, 5 insertions(+), 17 deletions(-)

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 8ead744..0ab32fa 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -903,6 +903,9 @@ EXPORT_SYMBOL(skb_clone);
 
 static void skb_headers_offset_update(struct sk_buff *skb, int off)
 {
+	/* Only adjust this if it actually is csum_start rather than csum */
+	if (skb->ip_summed == CHECKSUM_PARTIAL)
+		skb->csum_start += off;
 	/* {transport,network,mac}_header and tail are relative to skb->head */
 	skb->transport_header += off;
 	skb->network_header   += off;
@@ -1109,9 +1112,6 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,
 #endif
 	skb->tail	      += off;
 	skb_headers_offset_update(skb, nhead);
-	/* Only adjust this if it actually is csum_start rather than csum */
-	if (skb->ip_summed == CHECKSUM_PARTIAL)
-		skb->csum_start += nhead;
 	skb->cloned   = 0;
 	skb->hdr_len  = 0;
 	skb->nohdr    = 0;
@@ -1176,7 +1176,6 @@ struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
 					NUMA_NO_NODE);
 	int oldheadroom = skb_headroom(skb);
 	int head_copy_len, head_copy_off;
-	int off;
 
 	if (!n)
 		return NULL;
@@ -1200,11 +1199,7 @@ struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
 
 	copy_skb_header(n, skb);
 
-	off                  = newheadroom - oldheadroom;
-	if (n->ip_summed == CHECKSUM_PARTIAL)
-		n->csum_start += off;
-
-	skb_headers_offset_update(n, off);
+	skb_headers_offset_update(n, newheadroom - oldheadroom);
 
 	return n;
 }
@@ -2837,14 +2832,7 @@ struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features)
 		__copy_skb_header(nskb, skb);
 		nskb->mac_len = skb->mac_len;
 
-		/* nskb and skb might have different headroom */
-		if (nskb->ip_summed == CHECKSUM_PARTIAL)
-			nskb->csum_start += skb_headroom(nskb) - headroom;
-
-		skb_reset_mac_header(nskb);
-		skb_set_network_header(nskb, skb->mac_len);
-		nskb->transport_header = (nskb->network_header +
-					  skb_network_header_len(skb));
+		skb_headers_offset_update(nskb, skb_headroom(nskb) - headroom);
 
 		skb_copy_from_linear_data_offset(skb, -tnl_hlen,
 						 nskb->data - tnl_hlen,
-- 
1.8.4

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH net-next 3/6] ipv4: generalize gre_handle_offloads
  2013-10-19 18:42 [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP Eric Dumazet
  2013-10-19 18:42 ` [PATCH net-next 1/6] ipv4: gso: send_check() & segment() cleanups Eric Dumazet
  2013-10-19 18:42 ` [PATCH net-next 2/6] net: generalize skb_segment() Eric Dumazet
@ 2013-10-19 18:42 ` Eric Dumazet
  2013-10-19 18:42 ` [PATCH net-next 4/6] ipv4: gso: make inet_gso_segment() stackable Eric Dumazet
                   ` (3 subsequent siblings)
  6 siblings, 0 replies; 10+ messages in thread
From: Eric Dumazet @ 2013-10-19 18:42 UTC (permalink / raw)
  To: David S. Miller; +Cc: netdev, Jerry Chu, Tom Herbert, Eric Dumazet

This patch makes gre_handle_offloads() more generic
and rename it to iptunnel_handle_offloads()

This will be used to add GSO/TSO support to IPIP tunnels.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 include/net/gre.h         |  8 +++++++-
 include/net/ip_tunnels.h  |  3 +++
 net/ipv4/gre_demux.c      | 29 -----------------------------
 net/ipv4/ip_tunnel_core.c | 33 +++++++++++++++++++++++++++++++++
 4 files changed, 43 insertions(+), 30 deletions(-)

diff --git a/include/net/gre.h b/include/net/gre.h
index 57e4afd..dcd9ae3 100644
--- a/include/net/gre.h
+++ b/include/net/gre.h
@@ -38,7 +38,13 @@ void gre_offload_exit(void);
 
 void gre_build_header(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
 		      int hdr_len);
-struct sk_buff *gre_handle_offloads(struct sk_buff *skb, bool gre_csum);
+
+static inline struct sk_buff *gre_handle_offloads(struct sk_buff *skb,
+						  bool gre_csum)
+{
+	return iptunnel_handle_offloads(skb, gre_csum, SKB_GSO_GRE);
+}
+
 
 static inline int ip_gre_calc_hlen(__be16 o_flags)
 {
diff --git a/include/net/ip_tunnels.h b/include/net/ip_tunnels.h
index a0a4a10..732f8c6 100644
--- a/include/net/ip_tunnels.h
+++ b/include/net/ip_tunnels.h
@@ -150,6 +150,9 @@ int iptunnel_xmit(struct rtable *rt, struct sk_buff *skb,
 		  __be32 src, __be32 dst, __u8 proto,
 		  __u8 tos, __u8 ttl, __be16 df, bool xnet);
 
+struct sk_buff *iptunnel_handle_offloads(struct sk_buff *skb, bool gre_csum,
+					 int gso_type_mask);
+
 static inline void iptunnel_xmit_stats(int err,
 				       struct net_device_stats *err_stats,
 				       struct pcpu_tstats __percpu *stats)
diff --git a/net/ipv4/gre_demux.c b/net/ipv4/gre_demux.c
index 736c9fc3..5893e99 100644
--- a/net/ipv4/gre_demux.c
+++ b/net/ipv4/gre_demux.c
@@ -93,35 +93,6 @@ void gre_build_header(struct sk_buff *skb, const struct tnl_ptk_info *tpi,
 }
 EXPORT_SYMBOL_GPL(gre_build_header);
 
-struct sk_buff *gre_handle_offloads(struct sk_buff *skb, bool gre_csum)
-{
-	int err;
-
-	if (likely(!skb->encapsulation)) {
-		skb_reset_inner_headers(skb);
-		skb->encapsulation = 1;
-	}
-
-	if (skb_is_gso(skb)) {
-		err = skb_unclone(skb, GFP_ATOMIC);
-		if (unlikely(err))
-			goto error;
-		skb_shinfo(skb)->gso_type |= SKB_GSO_GRE;
-		return skb;
-	} else if (skb->ip_summed == CHECKSUM_PARTIAL && gre_csum) {
-		err = skb_checksum_help(skb);
-		if (unlikely(err))
-			goto error;
-	} else if (skb->ip_summed != CHECKSUM_PARTIAL)
-		skb->ip_summed = CHECKSUM_NONE;
-
-	return skb;
-error:
-	kfree_skb(skb);
-	return ERR_PTR(err);
-}
-EXPORT_SYMBOL_GPL(gre_handle_offloads);
-
 static __sum16 check_checksum(struct sk_buff *skb)
 {
 	__sum16 csum = 0;
diff --git a/net/ipv4/ip_tunnel_core.c b/net/ipv4/ip_tunnel_core.c
index c31e3ad..42ffbc8 100644
--- a/net/ipv4/ip_tunnel_core.c
+++ b/net/ipv4/ip_tunnel_core.c
@@ -116,3 +116,36 @@ int iptunnel_pull_header(struct sk_buff *skb, int hdr_len, __be16 inner_proto)
 	return 0;
 }
 EXPORT_SYMBOL_GPL(iptunnel_pull_header);
+
+struct sk_buff *iptunnel_handle_offloads(struct sk_buff *skb,
+					 bool csum_help,
+					 int gso_type_mask)
+{
+	int err;
+
+	if (likely(!skb->encapsulation)) {
+		skb_reset_inner_headers(skb);
+		skb->encapsulation = 1;
+	}
+
+	if (skb_is_gso(skb)) {
+		err = skb_unclone(skb, GFP_ATOMIC);
+		if (unlikely(err))
+			goto error;
+		skb_shinfo(skb)->gso_type |= gso_type_mask;
+		return skb;
+	}
+
+	if (skb->ip_summed == CHECKSUM_PARTIAL && csum_help) {
+		err = skb_checksum_help(skb);
+		if (unlikely(err))
+			goto error;
+	} else if (skb->ip_summed != CHECKSUM_PARTIAL)
+		skb->ip_summed = CHECKSUM_NONE;
+
+	return skb;
+error:
+	kfree_skb(skb);
+	return ERR_PTR(err);
+}
+EXPORT_SYMBOL_GPL(iptunnel_handle_offloads);
-- 
1.8.4

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH net-next 4/6] ipv4: gso: make inet_gso_segment() stackable
  2013-10-19 18:42 [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP Eric Dumazet
                   ` (2 preceding siblings ...)
  2013-10-19 18:42 ` [PATCH net-next 3/6] ipv4: generalize gre_handle_offloads Eric Dumazet
@ 2013-10-19 18:42 ` Eric Dumazet
  2013-10-19 18:42 ` [PATCH net-next 5/6] ipip: add GSO/TSO support Eric Dumazet
                   ` (2 subsequent siblings)
  6 siblings, 0 replies; 10+ messages in thread
From: Eric Dumazet @ 2013-10-19 18:42 UTC (permalink / raw)
  To: David S. Miller; +Cc: netdev, Jerry Chu, Tom Herbert, Eric Dumazet

In order to support GSO on IPIP, we need to make
inet_gso_segment() stackable.

It should not assume network header starts right after mac
header.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 include/linux/skbuff.h |  7 +++++--
 net/core/dev.c         |  2 ++
 net/ipv4/af_inet.c     | 25 ++++++++++++++++++-------
 3 files changed, 25 insertions(+), 9 deletions(-)

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index ba74474..cad1e0c 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -2722,9 +2722,12 @@ static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
 /* Keeps track of mac header offset relative to skb->head.
  * It is useful for TSO of Tunneling protocol. e.g. GRE.
  * For non-tunnel skb it points to skb_mac_header() and for
- * tunnel skb it points to outer mac header. */
+ * tunnel skb it points to outer mac header.
+ * Keeps track of level of encapsulation of network headers.
+ */
 struct skb_gso_cb {
-	int mac_offset;
+	int	mac_offset;
+	int	encap_level;
 };
 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
 
diff --git a/net/core/dev.c b/net/core/dev.c
index 1b6eadf..0918aad 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2377,6 +2377,8 @@ struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
 	}
 
 	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
+	SKB_GSO_CB(skb)->encap_level = 0;
+
 	skb_reset_mac_header(skb);
 	skb_reset_mac_len(skb);
 
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 4f8cd4f..5783ab5 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -1273,16 +1273,17 @@ out:
 }
 
 static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
-	netdev_features_t features)
+					netdev_features_t features)
 {
 	struct sk_buff *segs = ERR_PTR(-EINVAL);
 	const struct net_offload *ops;
+	unsigned int offset = 0;
 	struct iphdr *iph;
+	bool tunnel;
 	int proto;
+	int nhoff;
 	int ihl;
 	int id;
-	unsigned int offset = 0;
-	bool tunnel;
 
 	if (unlikely(skb_shinfo(skb)->gso_type &
 		     ~(SKB_GSO_TCPV4 |
@@ -1296,6 +1297,8 @@ static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
 		       0)))
 		goto out;
 
+	skb_reset_network_header(skb);
+	nhoff = skb_network_header(skb) - skb_mac_header(skb);
 	if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
 		goto out;
 
@@ -1312,7 +1315,10 @@ static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
 		goto out;
 	__skb_pull(skb, ihl);
 
-	tunnel = !!skb->encapsulation;
+	tunnel = SKB_GSO_CB(skb)->encap_level > 0;
+	if (tunnel)
+		features = skb->dev->hw_enc_features & netif_skb_features(skb);
+	SKB_GSO_CB(skb)->encap_level += ihl;
 
 	skb_reset_transport_header(skb);
 
@@ -1327,18 +1333,23 @@ static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
 
 	skb = segs;
 	do {
-		iph = ip_hdr(skb);
+		iph = (struct iphdr *)(skb_mac_header(skb) + nhoff);
 		if (!tunnel && proto == IPPROTO_UDP) {
 			iph->id = htons(id);
 			iph->frag_off = htons(offset >> 3);
 			if (skb->next != NULL)
 				iph->frag_off |= htons(IP_MF);
-			offset += (skb->len - skb->mac_len - iph->ihl * 4);
+			offset += skb->len - nhoff - ihl;
 		} else  {
 			iph->id = htons(id++);
 		}
-		iph->tot_len = htons(skb->len - skb->mac_len);
+		iph->tot_len = htons(skb->len - nhoff);
 		ip_send_check(iph);
+		if (tunnel) {
+			skb_reset_inner_headers(skb);
+			skb->encapsulation = 1;
+		}
+		skb->network_header = (u8 *)iph - skb->head;
 	} while ((skb = skb->next));
 
 out:
-- 
1.8.4

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH net-next 5/6] ipip: add GSO/TSO support
  2013-10-19 18:42 [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP Eric Dumazet
                   ` (3 preceding siblings ...)
  2013-10-19 18:42 ` [PATCH net-next 4/6] ipv4: gso: make inet_gso_segment() stackable Eric Dumazet
@ 2013-10-19 18:42 ` Eric Dumazet
  2013-10-19 21:00   ` Dmitry Kravkov
  2013-10-19 18:42 ` [PATCH net-next 6/6] bnx2x: add TSO support for IPIP Eric Dumazet
  2013-10-19 18:59 ` [PATCH net-next 0/6] net: Implement GSO/TSO " Eric Dumazet
  6 siblings, 1 reply; 10+ messages in thread
From: Eric Dumazet @ 2013-10-19 18:42 UTC (permalink / raw)
  To: David S. Miller; +Cc: netdev, Jerry Chu, Tom Herbert, Eric Dumazet

Now inet_gso_segment() is stackable, its relatively easy to
implement GSO/TSO support for IPIP

Performance results, when segmentation is done after tunnel
device (as no NIC is yet enabled for TSO IPIP support) :

Before patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      3357.88   5.09     3.70     2.983   2.167

After patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      7710.19   4.52     6.62     1.152   1.687

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 include/linux/netdev_features.h |  2 ++
 include/linux/skbuff.h          |  6 ++++--
 net/core/ethtool.c              |  1 +
 net/ipv4/af_inet.c              |  9 +++++++++
 net/ipv4/gre_offload.c          |  3 ++-
 net/ipv4/ipip.c                 | 11 ++++++-----
 net/ipv4/tcp_offload.c          |  1 +
 net/ipv4/udp_offload.c          |  1 +
 net/ipv6/ip6_offload.c          |  1 +
 net/ipv6/udp_offload.c          |  1 +
 net/mpls/mpls_gso.c             |  1 +
 11 files changed, 29 insertions(+), 8 deletions(-)

diff --git a/include/linux/netdev_features.h b/include/linux/netdev_features.h
index a2a89a5..8dad68c 100644
--- a/include/linux/netdev_features.h
+++ b/include/linux/netdev_features.h
@@ -42,6 +42,7 @@ enum {
 	NETIF_F_TSO6_BIT,		/* ... TCPv6 segmentation */
 	NETIF_F_FSO_BIT,		/* ... FCoE segmentation */
 	NETIF_F_GSO_GRE_BIT,		/* ... GRE with TSO */
+	NETIF_F_GSO_IPIP_BIT,		/* ... IPIP tunnel with TSO */
 	NETIF_F_GSO_UDP_TUNNEL_BIT,	/* ... UDP TUNNEL with TSO */
 	NETIF_F_GSO_MPLS_BIT,		/* ... MPLS segmentation */
 	/**/NETIF_F_GSO_LAST =		/* last bit, see GSO_MASK */
@@ -107,6 +108,7 @@ enum {
 #define NETIF_F_RXFCS		__NETIF_F(RXFCS)
 #define NETIF_F_RXALL		__NETIF_F(RXALL)
 #define NETIF_F_GSO_GRE		__NETIF_F(GSO_GRE)
+#define NETIF_F_GSO_IPIP	__NETIF_F(GSO_IPIP)
 #define NETIF_F_GSO_UDP_TUNNEL	__NETIF_F(GSO_UDP_TUNNEL)
 #define NETIF_F_GSO_MPLS	__NETIF_F(GSO_MPLS)
 #define NETIF_F_HW_VLAN_STAG_FILTER __NETIF_F(HW_VLAN_STAG_FILTER)
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index cad1e0c..6072913 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -318,9 +318,11 @@ enum {
 
 	SKB_GSO_GRE = 1 << 6,
 
-	SKB_GSO_UDP_TUNNEL = 1 << 7,
+	SKB_GSO_IPIP = 1 << 7,
 
-	SKB_GSO_MPLS = 1 << 8,
+	SKB_GSO_UDP_TUNNEL = 1 << 8,
+
+	SKB_GSO_MPLS = 1 << 9,
 };
 
 #if BITS_PER_LONG > 32
diff --git a/net/core/ethtool.c b/net/core/ethtool.c
index 78e9d92..8cab774 100644
--- a/net/core/ethtool.c
+++ b/net/core/ethtool.c
@@ -81,6 +81,7 @@ static const char netdev_features_strings[NETDEV_FEATURE_COUNT][ETH_GSTRING_LEN]
 	[NETIF_F_TSO6_BIT] =             "tx-tcp6-segmentation",
 	[NETIF_F_FSO_BIT] =              "tx-fcoe-segmentation",
 	[NETIF_F_GSO_GRE_BIT] =		 "tx-gre-segmentation",
+	[NETIF_F_GSO_IPIP_BIT] =	 "tx-ipip-segmentation",
 	[NETIF_F_GSO_UDP_TUNNEL_BIT] =	 "tx-udp_tnl-segmentation",
 	[NETIF_F_GSO_MPLS_BIT] =	 "tx-mpls-segmentation",
 
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 5783ab5..4049906 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -1291,6 +1291,7 @@ static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
 		       SKB_GSO_DODGY |
 		       SKB_GSO_TCP_ECN |
 		       SKB_GSO_GRE |
+		       SKB_GSO_IPIP |
 		       SKB_GSO_TCPV6 |
 		       SKB_GSO_UDP_TUNNEL |
 		       SKB_GSO_MPLS |
@@ -1656,6 +1657,13 @@ static struct packet_offload ip_packet_offload __read_mostly = {
 	},
 };
 
+static const struct net_offload ipip_offload = {
+	.callbacks = {
+		.gso_send_check = inet_gso_send_check,
+		.gso_segment	= inet_gso_segment,
+	},
+};
+
 static int __init ipv4_offload_init(void)
 {
 	/*
@@ -1667,6 +1675,7 @@ static int __init ipv4_offload_init(void)
 		pr_crit("%s: Cannot add TCP protocol offload\n", __func__);
 
 	dev_add_offload(&ip_packet_offload);
+	inet_add_offload(&ipip_offload, IPPROTO_IPIP);
 	return 0;
 }
 
diff --git a/net/ipv4/gre_offload.c b/net/ipv4/gre_offload.c
index 55e6bfb..e5d4361 100644
--- a/net/ipv4/gre_offload.c
+++ b/net/ipv4/gre_offload.c
@@ -39,7 +39,8 @@ static struct sk_buff *gre_gso_segment(struct sk_buff *skb,
 				  SKB_GSO_UDP |
 				  SKB_GSO_DODGY |
 				  SKB_GSO_TCP_ECN |
-				  SKB_GSO_GRE)))
+				  SKB_GSO_GRE |
+				  SKB_GSO_IPIP)))
 		goto out;
 
 	if (unlikely(!pskb_may_pull(skb, sizeof(*greh))))
diff --git a/net/ipv4/ipip.c b/net/ipv4/ipip.c
index 7f80fb4..fe3e9f7 100644
--- a/net/ipv4/ipip.c
+++ b/net/ipv4/ipip.c
@@ -220,17 +220,17 @@ static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 	if (unlikely(skb->protocol != htons(ETH_P_IP)))
 		goto tx_error;
 
-	if (likely(!skb->encapsulation)) {
-		skb_reset_inner_headers(skb);
-		skb->encapsulation = 1;
-	}
+	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
+	if (IS_ERR(skb))
+		goto out;
 
 	ip_tunnel_xmit(skb, dev, tiph, tiph->protocol);
 	return NETDEV_TX_OK;
 
 tx_error:
-	dev->stats.tx_errors++;
 	dev_kfree_skb(skb);
+out:
+	dev->stats.tx_errors++;
 	return NETDEV_TX_OK;
 }
 
@@ -275,6 +275,7 @@ static const struct net_device_ops ipip_netdev_ops = {
 #define IPIP_FEATURES (NETIF_F_SG |		\
 		       NETIF_F_FRAGLIST |	\
 		       NETIF_F_HIGHDMA |	\
+		       NETIF_F_GSO_SOFTWARE |	\
 		       NETIF_F_HW_CSUM)
 
 static void ipip_tunnel_setup(struct net_device *dev)
diff --git a/net/ipv4/tcp_offload.c b/net/ipv4/tcp_offload.c
index 8e3113f..dfc96b0 100644
--- a/net/ipv4/tcp_offload.c
+++ b/net/ipv4/tcp_offload.c
@@ -56,6 +56,7 @@ struct sk_buff *tcp_gso_segment(struct sk_buff *skb,
 			       SKB_GSO_TCP_ECN |
 			       SKB_GSO_TCPV6 |
 			       SKB_GSO_GRE |
+			       SKB_GSO_IPIP |
 			       SKB_GSO_MPLS |
 			       SKB_GSO_UDP_TUNNEL |
 			       0) ||
diff --git a/net/ipv4/udp_offload.c b/net/ipv4/udp_offload.c
index f35ecca..83206de 100644
--- a/net/ipv4/udp_offload.c
+++ b/net/ipv4/udp_offload.c
@@ -52,6 +52,7 @@ static struct sk_buff *udp4_ufo_fragment(struct sk_buff *skb,
 
 		if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY |
 				      SKB_GSO_UDP_TUNNEL |
+				      SKB_GSO_IPIP |
 				      SKB_GSO_GRE | SKB_GSO_MPLS) ||
 			     !(type & (SKB_GSO_UDP))))
 			goto out;
diff --git a/net/ipv6/ip6_offload.c b/net/ipv6/ip6_offload.c
index d82de72..548036a 100644
--- a/net/ipv6/ip6_offload.c
+++ b/net/ipv6/ip6_offload.c
@@ -98,6 +98,7 @@ static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb,
 		       SKB_GSO_DODGY |
 		       SKB_GSO_TCP_ECN |
 		       SKB_GSO_GRE |
+		       SKB_GSO_IPIP |
 		       SKB_GSO_UDP_TUNNEL |
 		       SKB_GSO_MPLS |
 		       SKB_GSO_TCPV6 |
diff --git a/net/ipv6/udp_offload.c b/net/ipv6/udp_offload.c
index 6055951..f63780f 100644
--- a/net/ipv6/udp_offload.c
+++ b/net/ipv6/udp_offload.c
@@ -64,6 +64,7 @@ static struct sk_buff *udp6_ufo_fragment(struct sk_buff *skb,
 				      SKB_GSO_DODGY |
 				      SKB_GSO_UDP_TUNNEL |
 				      SKB_GSO_GRE |
+				      SKB_GSO_IPIP |
 				      SKB_GSO_MPLS) ||
 			     !(type & (SKB_GSO_UDP))))
 			goto out;
diff --git a/net/mpls/mpls_gso.c b/net/mpls/mpls_gso.c
index 1bec121..851cd88 100644
--- a/net/mpls/mpls_gso.c
+++ b/net/mpls/mpls_gso.c
@@ -33,6 +33,7 @@ static struct sk_buff *mpls_gso_segment(struct sk_buff *skb,
 				  SKB_GSO_DODGY |
 				  SKB_GSO_TCP_ECN |
 				  SKB_GSO_GRE |
+				  SKB_GSO_IPIP |
 				  SKB_GSO_MPLS)))
 		goto out;
 
-- 
1.8.4

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* [PATCH net-next 6/6] bnx2x: add TSO support for IPIP
  2013-10-19 18:42 [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP Eric Dumazet
                   ` (4 preceding siblings ...)
  2013-10-19 18:42 ` [PATCH net-next 5/6] ipip: add GSO/TSO support Eric Dumazet
@ 2013-10-19 18:42 ` Eric Dumazet
  2013-10-19 18:59 ` [PATCH net-next 0/6] net: Implement GSO/TSO " Eric Dumazet
  6 siblings, 0 replies; 10+ messages in thread
From: Eric Dumazet @ 2013-10-19 18:42 UTC (permalink / raw)
  To: David S. Miller; +Cc: netdev, Jerry Chu, Tom Herbert, Eric Dumazet

bnx2x driver already handles TSO for GRE, current code
is the same for IPIP.

Performance results : (Note we are now limited by receiver,
as it does not support GRO for IPIP yet)

Before patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      7710.19   4.52     6.62     1.152   1.687

After patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      8532.40   2.55     7.73     0.588   1.781

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
index 8fd3432..6e5b35f 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c
@@ -12260,10 +12260,12 @@ static int bnx2x_init_dev(struct bnx2x *bp, struct pci_dev *pdev,
 		NETIF_F_RXCSUM | NETIF_F_LRO | NETIF_F_GRO |
 		NETIF_F_RXHASH | NETIF_F_HW_VLAN_CTAG_TX;
 	if (!CHIP_IS_E1x(bp)) {
-		dev->hw_features |= NETIF_F_GSO_GRE | NETIF_F_GSO_UDP_TUNNEL;
+		dev->hw_features |= NETIF_F_GSO_GRE | NETIF_F_GSO_UDP_TUNNEL |
+				    NETIF_F_GSO_IPIP;
 		dev->hw_enc_features =
 			NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_SG |
 			NETIF_F_TSO | NETIF_F_TSO_ECN | NETIF_F_TSO6 |
+			NETIF_F_GSO_IPIP |
 			NETIF_F_GSO_GRE | NETIF_F_GSO_UDP_TUNNEL;
 	}
 
-- 
1.8.4

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* Re: [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP
  2013-10-19 18:42 [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP Eric Dumazet
                   ` (5 preceding siblings ...)
  2013-10-19 18:42 ` [PATCH net-next 6/6] bnx2x: add TSO support for IPIP Eric Dumazet
@ 2013-10-19 18:59 ` Eric Dumazet
  2013-10-19 23:37   ` David Miller
  6 siblings, 1 reply; 10+ messages in thread
From: Eric Dumazet @ 2013-10-19 18:59 UTC (permalink / raw)
  To: Eric Dumazet; +Cc: David S. Miller, netdev, Jerry Chu, Tom Herbert

On Sat, 2013-10-19 at 11:42 -0700, Eric Dumazet wrote:
> This patch serie implements GSO/TSO support for IPIP
> 
> David, please note it applies after "ipv4: gso: send_check() & segment() cleanups"
> ( http://patchwork.ozlabs.org/patch/284714/ )

Oh well, I meant to say the above patch was _included_ in this patch
set.

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH net-next 5/6] ipip: add GSO/TSO support
  2013-10-19 18:42 ` [PATCH net-next 5/6] ipip: add GSO/TSO support Eric Dumazet
@ 2013-10-19 21:00   ` Dmitry Kravkov
  0 siblings, 0 replies; 10+ messages in thread
From: Dmitry Kravkov @ 2013-10-19 21:00 UTC (permalink / raw)
  To: Eric Dumazet
  Cc: David S. Miller, netdev@vger.kernel.org, Jerry Chu, Tom Herbert

> diff --git a/net/ipv4/gre_offload.c b/net/ipv4/gre_offload.c
> index 55e6bfb..e5d4361 100644
> --- a/net/ipv4/gre_offload.c
> +++ b/net/ipv4/gre_offload.c
> @@ -39,7 +39,8 @@ static struct sk_buff *gre_gso_segment(struct sk_buff *skb,
>                                   SKB_GSO_UDP |
>                                   SKB_GSO_DODGY |
>                                   SKB_GSO_TCP_ECN |
> -                                 SKB_GSO_GRE)))
> +                                 SKB_GSO_GRE |
> +                                 SKB_GSO_IPIP)))
>                 goto out;

Why GRE needs this bit?

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP
  2013-10-19 18:59 ` [PATCH net-next 0/6] net: Implement GSO/TSO " Eric Dumazet
@ 2013-10-19 23:37   ` David Miller
  0 siblings, 0 replies; 10+ messages in thread
From: David Miller @ 2013-10-19 23:37 UTC (permalink / raw)
  To: eric.dumazet; +Cc: edumazet, netdev, hkchu, therbert

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Sat, 19 Oct 2013 11:59:18 -0700

> On Sat, 2013-10-19 at 11:42 -0700, Eric Dumazet wrote:
>> This patch serie implements GSO/TSO support for IPIP
>> 
>> David, please note it applies after "ipv4: gso: send_check() & segment() cleanups"
>> ( http://patchwork.ozlabs.org/patch/284714/ )
> 
> Oh well, I meant to say the above patch was _included_ in this patch
> set.

I understood what you meant, especially because I just applied the patch
in question :-)

This looks fantastic, all applied, thanks!

^ permalink raw reply	[flat|nested] 10+ messages in thread

end of thread, other threads:[~2013-10-19 23:37 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-10-19 18:42 [PATCH net-next 0/6] net: Implement GSO/TSO support for IPIP Eric Dumazet
2013-10-19 18:42 ` [PATCH net-next 1/6] ipv4: gso: send_check() & segment() cleanups Eric Dumazet
2013-10-19 18:42 ` [PATCH net-next 2/6] net: generalize skb_segment() Eric Dumazet
2013-10-19 18:42 ` [PATCH net-next 3/6] ipv4: generalize gre_handle_offloads Eric Dumazet
2013-10-19 18:42 ` [PATCH net-next 4/6] ipv4: gso: make inet_gso_segment() stackable Eric Dumazet
2013-10-19 18:42 ` [PATCH net-next 5/6] ipip: add GSO/TSO support Eric Dumazet
2013-10-19 21:00   ` Dmitry Kravkov
2013-10-19 18:42 ` [PATCH net-next 6/6] bnx2x: add TSO support for IPIP Eric Dumazet
2013-10-19 18:59 ` [PATCH net-next 0/6] net: Implement GSO/TSO " Eric Dumazet
2013-10-19 23:37   ` David Miller

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).