--- /dev/null
+From: Felix Fietkau <nbd@nbd.name>
+Date: Tue, 23 Apr 2024 12:35:21 +0200
+Subject: [PATCH] net: enable fraglist GRO by default
+
+This can significantly improve performance for packet forwarding/bridging
+
+Signed-off-by: Felix Fietkau <nbd@nbd.name>
+---
+
+--- a/include/linux/netdev_features.h
++++ b/include/linux/netdev_features.h
+@@ -242,10 +242,10 @@ static inline int find_next_netdev_featu
+ #define NETIF_F_UPPER_DISABLES NETIF_F_LRO
+
+ /* changeable features with no special hardware requirements */
+-#define NETIF_F_SOFT_FEATURES (NETIF_F_GSO | NETIF_F_GRO)
++#define NETIF_F_SOFT_FEATURES (NETIF_F_GSO | NETIF_F_GRO | NETIF_F_GRO_FRAGLIST)
+
+ /* Changeable features with no special hardware requirements that defaults to off. */
+-#define NETIF_F_SOFT_FEATURES_OFF (NETIF_F_GRO_FRAGLIST | NETIF_F_GRO_UDP_FWD)
++#define NETIF_F_SOFT_FEATURES_OFF (NETIF_F_GRO_UDP_FWD)
+
+ #define NETIF_F_VLAN_FEATURES (NETIF_F_HW_VLAN_CTAG_FILTER | \
+ NETIF_F_HW_VLAN_CTAG_RX | \
const struct header_ops *header_ops;
unsigned char operstate;
-@@ -2206,6 +2213,10 @@ struct net_device {
+@@ -2204,6 +2211,10 @@ struct net_device {
struct mctp_dev __rcu *mctp_ptr;
#endif
*/
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
-@@ -3046,6 +3046,10 @@ static inline int pskb_trim(struct sk_bu
+@@ -3045,6 +3045,10 @@ static inline int pskb_trim(struct sk_bu
return (len < skb->len) ? __pskb_trim(skb, len) : 0;
}
/**
* pskb_trim_unique - remove end from a paged unique (not cloned) buffer
* @skb: buffer to alter
-@@ -3195,16 +3199,6 @@ static inline struct sk_buff *dev_alloc_
+@@ -3194,16 +3198,6 @@ static inline struct sk_buff *dev_alloc_
}
{
--- a/net/ethernet/eth.c
+++ b/net/ethernet/eth.c
-@@ -171,6 +171,12 @@ __be16 eth_type_trans(struct sk_buff *sk
+@@ -159,6 +159,12 @@ __be16 eth_type_trans(struct sk_buff *sk
const struct ethhdr *eth;
skb->dev = dev;
--- /dev/null
+From: Felix Fietkau <nbd@nbd.name>
+Date: Tue, 23 Apr 2024 12:35:21 +0200
+Subject: [PATCH] net: enable fraglist GRO by default
+
+This can significantly improve performance for packet forwarding/bridging
+
+Signed-off-by: Felix Fietkau <nbd@nbd.name>
+---
+
+--- a/include/linux/netdev_features.h
++++ b/include/linux/netdev_features.h
+@@ -242,10 +242,10 @@ static inline int find_next_netdev_featu
+ #define NETIF_F_UPPER_DISABLES NETIF_F_LRO
+
+ /* changeable features with no special hardware requirements */
+-#define NETIF_F_SOFT_FEATURES (NETIF_F_GSO | NETIF_F_GRO)
++#define NETIF_F_SOFT_FEATURES (NETIF_F_GSO | NETIF_F_GRO | NETIF_F_GRO_FRAGLIST)
+
+ /* Changeable features with no special hardware requirements that defaults to off. */
+-#define NETIF_F_SOFT_FEATURES_OFF (NETIF_F_GRO_FRAGLIST | NETIF_F_GRO_UDP_FWD)
++#define NETIF_F_SOFT_FEATURES_OFF (NETIF_F_GRO_UDP_FWD)
+
+ #define NETIF_F_VLAN_FEATURES (NETIF_F_HW_VLAN_CTAG_FILTER | \
+ NETIF_F_HW_VLAN_CTAG_RX | \
const struct header_ops *header_ops;
unsigned char operstate;
-@@ -2259,6 +2266,10 @@ struct net_device {
+@@ -2257,6 +2264,10 @@ struct net_device {
struct mctp_dev __rcu *mctp_ptr;
#endif
*/
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
-@@ -3081,6 +3081,10 @@ static inline int pskb_trim(struct sk_bu
+@@ -3080,6 +3080,10 @@ static inline int pskb_trim(struct sk_bu
return (len < skb->len) ? __pskb_trim(skb, len) : 0;
}
/**
* pskb_trim_unique - remove end from a paged unique (not cloned) buffer
* @skb: buffer to alter
-@@ -3246,16 +3250,6 @@ static inline struct sk_buff *dev_alloc_
+@@ -3245,16 +3249,6 @@ static inline struct sk_buff *dev_alloc_
}
{
--- a/net/ethernet/eth.c
+++ b/net/ethernet/eth.c
-@@ -171,6 +171,12 @@ __be16 eth_type_trans(struct sk_buff *sk
+@@ -159,6 +159,12 @@ __be16 eth_type_trans(struct sk_buff *sk
const struct ethhdr *eth;
skb->dev = dev;
+++ /dev/null
-From: Felix Fietkau <nbd@nbd.name>
-Subject: net: replace GRO optimization patch with a new one that supports VLANs/bridges with different MAC addresses
-
-Signed-off-by: Felix Fietkau <nbd@nbd.name>
----
- include/linux/netdevice.h | 2 ++
- include/linux/skbuff.h | 3 ++-
- net/core/dev.c | 48 +++++++++++++++++++++++++++++++++++++++++++++++
- net/ethernet/eth.c | 18 +++++++++++++++++-
- 4 files changed, 69 insertions(+), 2 deletions(-)
-
---- a/include/linux/netdevice.h
-+++ b/include/linux/netdevice.h
-@@ -2157,6 +2157,8 @@ struct net_device {
- struct netdev_hw_addr_list mc;
- struct netdev_hw_addr_list dev_addrs;
-
-+ unsigned char local_addr_mask[MAX_ADDR_LEN];
-+
- #ifdef CONFIG_SYSFS
- struct kset *queues_kset;
- #endif
---- a/include/linux/skbuff.h
-+++ b/include/linux/skbuff.h
-@@ -967,6 +967,7 @@ struct sk_buff {
- #ifdef CONFIG_IPV6_NDISC_NODETYPE
- __u8 ndisc_nodetype:2;
- #endif
-+ __u8 gro_skip:1;
-
- __u8 ipvs_property:1;
- __u8 inner_protocol_type:1;
---- a/net/core/gro.c
-+++ b/net/core/gro.c
-@@ -492,6 +492,9 @@ static enum gro_result dev_gro_receive(s
- int same_flow;
- int grow;
-
-+ if (skb->gro_skip)
-+ goto normal;
-+
- if (netif_elide_gro(skb->dev))
- goto normal;
-
---- a/net/core/dev.c
-+++ b/net/core/dev.c
-@@ -7628,6 +7628,48 @@ static void __netdev_adjacent_dev_unlink
- &upper_dev->adj_list.lower);
- }
-
-+static void __netdev_addr_mask(unsigned char *mask, const unsigned char *addr,
-+ struct net_device *dev)
-+{
-+ int i;
-+
-+ for (i = 0; i < dev->addr_len; i++)
-+ mask[i] |= addr[i] ^ dev->dev_addr[i];
-+}
-+
-+static void __netdev_upper_mask(unsigned char *mask, struct net_device *dev,
-+ struct net_device *lower)
-+{
-+ struct net_device *cur;
-+ struct list_head *iter;
-+
-+ netdev_for_each_upper_dev_rcu(dev, cur, iter) {
-+ __netdev_addr_mask(mask, cur->dev_addr, lower);
-+ __netdev_upper_mask(mask, cur, lower);
-+ }
-+}
-+
-+static void __netdev_update_addr_mask(struct net_device *dev)
-+{
-+ unsigned char mask[MAX_ADDR_LEN];
-+ struct net_device *cur;
-+ struct list_head *iter;
-+
-+ memset(mask, 0, sizeof(mask));
-+ __netdev_upper_mask(mask, dev, dev);
-+ memcpy(dev->local_addr_mask, mask, dev->addr_len);
-+
-+ netdev_for_each_lower_dev(dev, cur, iter)
-+ __netdev_update_addr_mask(cur);
-+}
-+
-+static void netdev_update_addr_mask(struct net_device *dev)
-+{
-+ rcu_read_lock();
-+ __netdev_update_addr_mask(dev);
-+ rcu_read_unlock();
-+}
-+
- static int __netdev_upper_dev_link(struct net_device *dev,
- struct net_device *upper_dev, bool master,
- void *upper_priv, void *upper_info,
-@@ -7679,6 +7721,7 @@ static int __netdev_upper_dev_link(struc
- if (ret)
- return ret;
-
-+ netdev_update_addr_mask(dev);
- ret = call_netdevice_notifiers_info(NETDEV_CHANGEUPPER,
- &changeupper_info.info);
- ret = notifier_to_errno(ret);
-@@ -7775,6 +7818,7 @@ static void __netdev_upper_dev_unlink(st
-
- __netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
-
-+ netdev_update_addr_mask(dev);
- call_netdevice_notifiers_info(NETDEV_CHANGEUPPER,
- &changeupper_info.info);
-
-@@ -8827,6 +8871,7 @@ int dev_set_mac_address(struct net_devic
- if (err)
- return err;
- dev->addr_assign_type = NET_ADDR_SET;
-+ netdev_update_addr_mask(dev);
- call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
- add_device_randomness(dev->dev_addr, dev->addr_len);
- return 0;
---- a/net/ethernet/eth.c
-+++ b/net/ethernet/eth.c
-@@ -143,6 +143,18 @@ u32 eth_get_headlen(const struct net_dev
- }
- EXPORT_SYMBOL(eth_get_headlen);
-
-+static inline bool
-+eth_check_local_mask(const void *addr1, const void *addr2, const void *mask)
-+{
-+ const u16 *a1 = addr1;
-+ const u16 *a2 = addr2;
-+ const u16 *m = mask;
-+
-+ return (((a1[0] ^ a2[0]) & ~m[0]) |
-+ ((a1[1] ^ a2[1]) & ~m[1]) |
-+ ((a1[2] ^ a2[2]) & ~m[2]));
-+}
-+
- /**
- * eth_type_trans - determine the packet's protocol ID.
- * @skb: received socket data
-@@ -174,6 +186,10 @@ __be16 eth_type_trans(struct sk_buff *sk
- } else {
- skb->pkt_type = PACKET_OTHERHOST;
- }
-+
-+ if (eth_check_local_mask(eth->h_dest, dev->dev_addr,
-+ dev->local_addr_mask))
-+ skb->gro_skip = 1;
- }
-
- /*
--- /dev/null
+From: Felix Fietkau <nbd@nbd.name>
+Date: Tue, 23 Apr 2024 11:23:03 +0200
+Subject: [PATCH] net: add TCP fraglist GRO support
+
+When forwarding TCP after GRO, software segmentation is very expensive,
+especially when the checksum needs to be recalculated.
+One case where that's currently unavoidable is when routing packets over
+PPPoE. Performance improves significantly when using fraglist GRO
+implemented in the same way as for UDP.
+
+Here's a measurement of running 2 TCP streams through a MediaTek MT7622
+device (2-core Cortex-A53), which runs NAT with flow offload enabled from
+one ethernet port to PPPoE on another ethernet port + cake qdisc set to
+1Gbps.
+
+rx-gro-list off: 630 Mbit/s, CPU 35% idle
+rx-gro-list on: 770 Mbit/s, CPU 40% idle
+
+Signe-off-by: Felix Fietkau <nbd@nbd.name>
+---
+
+--- a/include/net/gro.h
++++ b/include/net/gro.h
+@@ -424,6 +424,7 @@ static inline __wsum ip6_gro_compute_pse
+ }
+
+ int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
++int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
+
+ /* Pass the currently batched GRO_NORMAL SKBs up to the stack. */
+ static inline void gro_normal_list(struct napi_struct *napi)
+@@ -446,5 +447,48 @@ static inline void gro_normal_one(struct
+ gro_normal_list(napi);
+ }
+
++/* This function is the alternative of 'inet_iif' and 'inet_sdif'
++ * functions in case we can not rely on fields of IPCB.
++ *
++ * The caller must verify skb_valid_dst(skb) is false and skb->dev is initialized.
++ * The caller must hold the RCU read lock.
++ */
++static inline void inet_get_iif_sdif(const struct sk_buff *skb, int *iif, int *sdif)
++{
++ *iif = inet_iif(skb) ?: skb->dev->ifindex;
++ *sdif = 0;
++
++#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
++ if (netif_is_l3_slave(skb->dev)) {
++ struct net_device *master = netdev_master_upper_dev_get_rcu(skb->dev);
++
++ *sdif = *iif;
++ *iif = master ? master->ifindex : 0;
++ }
++#endif
++}
++
++/* This function is the alternative of 'inet6_iif' and 'inet6_sdif'
++ * functions in case we can not rely on fields of IP6CB.
++ *
++ * The caller must verify skb_valid_dst(skb) is false and skb->dev is initialized.
++ * The caller must hold the RCU read lock.
++ */
++static inline void inet6_get_iif_sdif(const struct sk_buff *skb, int *iif, int *sdif)
++{
++ /* using skb->dev->ifindex because skb_dst(skb) is not initialized */
++ *iif = skb->dev->ifindex;
++ *sdif = 0;
++
++#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
++ if (netif_is_l3_slave(skb->dev)) {
++ struct net_device *master = netdev_master_upper_dev_get_rcu(skb->dev);
++
++ *sdif = *iif;
++ *iif = master ? master->ifindex : 0;
++ }
++#endif
++}
++
+
+ #endif /* _NET_IPV6_GRO_H */
+--- a/include/net/tcp.h
++++ b/include/net/tcp.h
+@@ -2057,7 +2057,8 @@ void tcp_v4_destroy_sock(struct sock *sk
+
+ struct sk_buff *tcp_gso_segment(struct sk_buff *skb,
+ netdev_features_t features);
+-struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb);
++struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb,
++ bool fraglist);
+ INDIRECT_CALLABLE_DECLARE(int tcp4_gro_complete(struct sk_buff *skb, int thoff));
+ INDIRECT_CALLABLE_DECLARE(struct sk_buff *tcp4_gro_receive(struct list_head *head, struct sk_buff *skb));
+ INDIRECT_CALLABLE_DECLARE(int tcp6_gro_complete(struct sk_buff *skb, int thoff));
+--- a/net/core/gro.c
++++ b/net/core/gro.c
+@@ -290,6 +290,33 @@ done:
+ return 0;
+ }
+
++int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb)
++{
++ if (unlikely(p->len + skb->len >= 65536))
++ return -E2BIG;
++
++ if (NAPI_GRO_CB(p)->last == p)
++ skb_shinfo(p)->frag_list = skb;
++ else
++ NAPI_GRO_CB(p)->last->next = skb;
++
++ skb_pull(skb, skb_gro_offset(skb));
++
++ NAPI_GRO_CB(p)->last = skb;
++ NAPI_GRO_CB(p)->count++;
++ p->data_len += skb->len;
++
++ /* sk ownership - if any - completely transferred to the aggregated packet */
++ skb->destructor = NULL;
++ skb->sk = NULL;
++ p->truesize += skb->truesize;
++ p->len += skb->len;
++
++ NAPI_GRO_CB(skb)->same_flow = 1;
++
++ return 0;
++}
++
+
+ static void napi_gro_complete(struct napi_struct *napi, struct sk_buff *skb)
+ {
+--- a/net/ipv4/tcp_offload.c
++++ b/net/ipv4/tcp_offload.c
+@@ -27,6 +27,74 @@ static void tcp_gso_tstamp(struct sk_buf
+ }
+ }
+
++static void __tcpv4_gso_segment_csum(struct sk_buff *seg,
++ __be32 *oldip, __be32 *newip,
++ __be16 *oldport, __be16 *newport)
++{
++ struct tcphdr *th;
++ struct iphdr *iph;
++
++ if (*oldip == *newip && *oldport == *newport)
++ return;
++
++ th = tcp_hdr(seg);
++ iph = ip_hdr(seg);
++
++ if (th->check) {
++ inet_proto_csum_replace4(&th->check, seg, *oldip, *newip,
++ true);
++ inet_proto_csum_replace2(&th->check, seg, *oldport, *newport,
++ false);
++ if (!th->check)
++ th->check = CSUM_MANGLED_0;
++ }
++ *oldport = *newport;
++
++ csum_replace4(&iph->check, *oldip, *newip);
++ *oldip = *newip;
++}
++
++static struct sk_buff *__tcpv4_gso_segment_list_csum(struct sk_buff *segs)
++{
++ struct sk_buff *seg;
++ struct tcphdr *uh, *uh2;
++ struct iphdr *iph, *iph2;
++
++ seg = segs;
++ uh = tcp_hdr(seg);
++ iph = ip_hdr(seg);
++
++ if ((tcp_hdr(seg)->dest == tcp_hdr(seg->next)->dest) &&
++ (tcp_hdr(seg)->source == tcp_hdr(seg->next)->source) &&
++ (ip_hdr(seg)->daddr == ip_hdr(seg->next)->daddr) &&
++ (ip_hdr(seg)->saddr == ip_hdr(seg->next)->saddr))
++ return segs;
++
++ while ((seg = seg->next)) {
++ uh2 = tcp_hdr(seg);
++ iph2 = ip_hdr(seg);
++
++ __tcpv4_gso_segment_csum(seg,
++ &iph2->saddr, &iph->saddr,
++ &uh2->source, &uh->source);
++ __tcpv4_gso_segment_csum(seg,
++ &iph2->daddr, &iph->daddr,
++ &uh2->dest, &uh->dest);
++ }
++
++ return segs;
++}
++
++static struct sk_buff *__tcp_gso_segment_list(struct sk_buff *skb,
++ netdev_features_t features)
++{
++ skb = skb_segment_list(skb, features, skb_mac_header_len(skb));
++ if (IS_ERR(skb))
++ return skb;
++
++ return __tcpv4_gso_segment_list_csum(skb);
++}
++
+ static struct sk_buff *tcp4_gso_segment(struct sk_buff *skb,
+ netdev_features_t features)
+ {
+@@ -36,6 +104,9 @@ static struct sk_buff *tcp4_gso_segment(
+ if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
+ return ERR_PTR(-EINVAL);
+
++ if (skb_shinfo(skb)->gso_type & SKB_GSO_FRAGLIST)
++ return __tcp_gso_segment_list(skb, features);
++
+ if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
+ const struct iphdr *iph = ip_hdr(skb);
+ struct tcphdr *th = tcp_hdr(skb);
+@@ -177,7 +248,8 @@ out:
+ return segs;
+ }
+
+-struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb)
++struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb,
++ bool fraglist)
+ {
+ struct sk_buff *pp = NULL;
+ struct sk_buff *p;
+@@ -214,6 +286,7 @@ struct sk_buff *tcp_gro_receive(struct l
+ len = skb_gro_len(skb);
+ flags = tcp_flag_word(th);
+
++ NAPI_GRO_CB(skb)->is_flist = fraglist;
+ list_for_each_entry(p, head, list) {
+ if (!NAPI_GRO_CB(p)->same_flow)
+ continue;
+@@ -233,6 +306,7 @@ struct sk_buff *tcp_gro_receive(struct l
+ found:
+ /* Include the IP ID check below from the inner most IP hdr */
+ flush = NAPI_GRO_CB(p)->flush;
++ flush |= fraglist != NAPI_GRO_CB(p)->is_flist;
+ flush |= (__force int)(flags & TCP_FLAG_CWR);
+ flush |= (__force int)((flags ^ tcp_flag_word(th2)) &
+ ~(TCP_FLAG_CWR | TCP_FLAG_FIN | TCP_FLAG_PSH));
+@@ -268,6 +342,19 @@ found:
+ flush |= p->decrypted ^ skb->decrypted;
+ #endif
+
++ if (fraglist) {
++ flush |= (__force int)(flags ^ tcp_flag_word(th2));
++ flush |= skb->ip_summed != p->ip_summed;
++ flush |= skb->csum_level != p->csum_level;
++ flush |= !pskb_may_pull(skb, skb_gro_offset(skb));
++ flush |= NAPI_GRO_CB(p)->count >= 64;
++
++ if (flush || skb_gro_receive_list(p, skb))
++ mss = 1;
++
++ goto out_check_final;
++ }
++
+ if (flush || skb_gro_receive(p, skb)) {
+ mss = 1;
+ goto out_check_final;
+@@ -315,6 +402,49 @@ int tcp_gro_complete(struct sk_buff *skb
+ }
+ EXPORT_SYMBOL(tcp_gro_complete);
+
++static bool tcp4_check_fraglist_gro(struct sk_buff *skb)
++{
++ const struct iphdr *iph = skb_gro_network_header(skb);
++ struct net *net = dev_net(skb->dev);
++ unsigned int off, hlen, thlen;
++ struct tcphdr *th;
++ struct sock *sk;
++ int iif, sdif;
++
++ if (!(skb->dev->features & NETIF_F_GRO_FRAGLIST))
++ return false;
++
++ inet_get_iif_sdif(skb, &iif, &sdif);
++
++ off = skb_gro_offset(skb);
++ hlen = off + sizeof(*th);
++ th = skb_gro_header(skb, hlen, off);
++ if (unlikely(!th))
++ return false;
++
++ thlen = th->doff * 4;
++ if (thlen < sizeof(*th))
++ return false;
++
++ hlen = off + thlen;
++ if (skb_gro_header_hard(skb, hlen)) {
++ th = skb_gro_header_slow(skb, hlen, off);
++ if (unlikely(!th))
++ return false;
++ }
++
++ sk = __inet_lookup_established(net, net->ipv4.tcp_death_row.hashinfo,
++ iph->saddr, th->source,
++ iph->daddr, ntohs(th->dest),
++ iif, sdif);
++ if (!sk)
++ return true;
++
++ sock_put(sk);
++
++ return false;
++}
++
+ INDIRECT_CALLABLE_SCOPE
+ struct sk_buff *tcp4_gro_receive(struct list_head *head, struct sk_buff *skb)
+ {
+@@ -326,7 +456,7 @@ struct sk_buff *tcp4_gro_receive(struct
+ return NULL;
+ }
+
+- return tcp_gro_receive(head, skb);
++ return tcp_gro_receive(head, skb, tcp4_check_fraglist_gro(skb));
+ }
+
+ INDIRECT_CALLABLE_SCOPE int tcp4_gro_complete(struct sk_buff *skb, int thoff)
+@@ -334,6 +464,15 @@ INDIRECT_CALLABLE_SCOPE int tcp4_gro_com
+ const struct iphdr *iph = ip_hdr(skb);
+ struct tcphdr *th = tcp_hdr(skb);
+
++ if (NAPI_GRO_CB(skb)->is_flist) {
++ skb_shinfo(skb)->gso_type |= SKB_GSO_FRAGLIST | SKB_GSO_TCPV4;
++ skb_shinfo(skb)->gso_segs = NAPI_GRO_CB(skb)->count;
++
++ __skb_incr_checksum_unnecessary(skb);
++
++ return 0;
++ }
++
+ th->check = ~tcp_v4_check(skb->len - thoff, iph->saddr,
+ iph->daddr, 0);
+ skb_shinfo(skb)->gso_type |= SKB_GSO_TCPV4;
+--- a/net/ipv4/udp_offload.c
++++ b/net/ipv4/udp_offload.c
+@@ -425,33 +425,6 @@ out:
+ return segs;
+ }
+
+-static int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb)
+-{
+- if (unlikely(p->len + skb->len >= 65536))
+- return -E2BIG;
+-
+- if (NAPI_GRO_CB(p)->last == p)
+- skb_shinfo(p)->frag_list = skb;
+- else
+- NAPI_GRO_CB(p)->last->next = skb;
+-
+- skb_pull(skb, skb_gro_offset(skb));
+-
+- NAPI_GRO_CB(p)->last = skb;
+- NAPI_GRO_CB(p)->count++;
+- p->data_len += skb->len;
+-
+- /* sk ownership - if any - completely transferred to the aggregated packet */
+- skb->destructor = NULL;
+- skb->sk = NULL;
+- p->truesize += skb->truesize;
+- p->len += skb->len;
+-
+- NAPI_GRO_CB(skb)->same_flow = 1;
+-
+- return 0;
+-}
+-
+
+ #define UDP_GRO_CNT_MAX 64
+ static struct sk_buff *udp_gro_receive_segment(struct list_head *head,
+--- a/net/ipv6/tcpv6_offload.c
++++ b/net/ipv6/tcpv6_offload.c
+@@ -7,12 +7,56 @@
+ */
+ #include <linux/indirect_call_wrapper.h>
+ #include <linux/skbuff.h>
++#include <net/inet6_hashtables.h>
+ #include <net/gro.h>
+ #include <net/protocol.h>
+ #include <net/tcp.h>
+ #include <net/ip6_checksum.h>
+ #include "ip6_offload.h"
+
++static bool tcp6_check_fraglist_gro(struct sk_buff *skb)
++{
++ const struct ipv6hdr *hdr = skb_gro_network_header(skb);
++ struct net *net = dev_net(skb->dev);
++ unsigned int off, hlen, thlen;
++ struct tcphdr *th;
++ struct sock *sk;
++ int iif, sdif;
++
++ if (!(skb->dev->features & NETIF_F_GRO_FRAGLIST))
++ return false;
++
++ inet6_get_iif_sdif(skb, &iif, &sdif);
++
++ off = skb_gro_offset(skb);
++ hlen = off + sizeof(*th);
++ th = skb_gro_header(skb, hlen, off);
++ if (unlikely(!th))
++ return false;
++
++ thlen = th->doff * 4;
++ if (thlen < sizeof(*th))
++ return false;
++
++ hlen = off + thlen;
++ if (skb_gro_header_hard(skb, hlen)) {
++ th = skb_gro_header_slow(skb, hlen, off);
++ if (unlikely(!th))
++ return false;
++ }
++
++ sk = __inet6_lookup_established(net, net->ipv4.tcp_death_row.hashinfo,
++ &hdr->saddr, th->source,
++ &hdr->daddr, ntohs(th->dest),
++ iif, sdif);
++ if (!sk)
++ return true;
++
++ sock_put(sk);
++
++ return false;
++}
++
+ INDIRECT_CALLABLE_SCOPE
+ struct sk_buff *tcp6_gro_receive(struct list_head *head, struct sk_buff *skb)
+ {
+@@ -24,7 +68,7 @@ struct sk_buff *tcp6_gro_receive(struct
+ return NULL;
+ }
+
+- return tcp_gro_receive(head, skb);
++ return tcp_gro_receive(head, skb, tcp6_check_fraglist_gro(skb));
+ }
+
+ INDIRECT_CALLABLE_SCOPE int tcp6_gro_complete(struct sk_buff *skb, int thoff)
+@@ -32,6 +76,15 @@ INDIRECT_CALLABLE_SCOPE int tcp6_gro_com
+ const struct ipv6hdr *iph = ipv6_hdr(skb);
+ struct tcphdr *th = tcp_hdr(skb);
+
++ if (NAPI_GRO_CB(skb)->is_flist) {
++ skb_shinfo(skb)->gso_type |= SKB_GSO_FRAGLIST | SKB_GSO_TCPV6;
++ skb_shinfo(skb)->gso_segs = NAPI_GRO_CB(skb)->count;
++
++ __skb_incr_checksum_unnecessary(skb);
++
++ return 0;
++ }
++
+ th->check = ~tcp_v6_check(skb->len - thoff, &iph->saddr,
+ &iph->daddr, 0);
+ skb_shinfo(skb)->gso_type |= SKB_GSO_TCPV6;
+@@ -50,6 +103,9 @@ static struct sk_buff *tcp6_gso_segment(
+ if (!pskb_may_pull(skb, sizeof(*th)))
+ return ERR_PTR(-EINVAL);
+
++ if (skb_shinfo(skb)->gso_type & SKB_GSO_FRAGLIST)
++ return skb_segment_list(skb, features, skb_mac_header_len(skb));
++
+ if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
+ const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
+ struct tcphdr *th = tcp_hdr(skb);
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
-@@ -2192,7 +2192,7 @@ struct net_device {
+@@ -2190,7 +2190,7 @@ struct net_device {
#if IS_ENABLED(CONFIG_AX25)
void *ax25_ptr;
#endif
/**
* napi_disable - prevent NAPI from scheduling
-@@ -3152,6 +3153,7 @@ struct softnet_data {
+@@ -3150,6 +3151,7 @@ struct softnet_data {
unsigned int processed;
unsigned int time_squeeze;
unsigned int received_rps;
void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
int (*poll)(struct napi_struct *, int), int weight)
{
-@@ -11171,6 +11242,9 @@ static int dev_cpu_dead(unsigned int old
+@@ -11126,6 +11197,9 @@ static int dev_cpu_dead(unsigned int old
raise_softirq_irqoff(NET_TX_SOFTIRQ);
local_irq_enable();
#ifdef CONFIG_RPS
remsd = oldsd->rps_ipi_list;
oldsd->rps_ipi_list = NULL;
-@@ -11483,6 +11557,7 @@ static int __init net_dev_init(void)
+@@ -11438,6 +11512,7 @@ static int __init net_dev_init(void)
INIT_CSD(&sd->defer_csd, trigger_rx_softirq, sd);
spin_lock_init(&sd->defer_lock);
+++ /dev/null
-From: Felix Fietkau <nbd@nbd.name>
-Subject: net: replace GRO optimization patch with a new one that supports VLANs/bridges with different MAC addresses
-
-Signed-off-by: Felix Fietkau <nbd@nbd.name>
----
- include/linux/netdevice.h | 2 ++
- include/linux/skbuff.h | 3 ++-
- net/core/dev.c | 48 +++++++++++++++++++++++++++++++++++++++++++++++
- net/ethernet/eth.c | 18 +++++++++++++++++-
- 4 files changed, 69 insertions(+), 2 deletions(-)
-
---- a/include/linux/netdevice.h
-+++ b/include/linux/netdevice.h
-@@ -2210,6 +2210,8 @@ struct net_device {
- struct netdev_hw_addr_list mc;
- struct netdev_hw_addr_list dev_addrs;
-
-+ unsigned char local_addr_mask[MAX_ADDR_LEN];
-+
- #ifdef CONFIG_SYSFS
- struct kset *queues_kset;
- #endif
---- a/include/linux/skbuff.h
-+++ b/include/linux/skbuff.h
-@@ -959,6 +959,7 @@ struct sk_buff {
- #ifdef CONFIG_IPV6_NDISC_NODETYPE
- __u8 ndisc_nodetype:2;
- #endif
-+ __u8 gro_skip:1;
-
- #if IS_ENABLED(CONFIG_IP_VS)
- __u8 ipvs_property:1;
---- a/net/core/gro.c
-+++ b/net/core/gro.c
-@@ -446,6 +446,9 @@ static enum gro_result dev_gro_receive(s
- enum gro_result ret;
- int same_flow;
-
-+ if (skb->gro_skip)
-+ goto normal;
-+
- if (netif_elide_gro(skb->dev))
- goto normal;
-
---- a/net/core/dev.c
-+++ b/net/core/dev.c
-@@ -7689,6 +7689,48 @@ static void __netdev_adjacent_dev_unlink
- &upper_dev->adj_list.lower);
- }
-
-+static void __netdev_addr_mask(unsigned char *mask, const unsigned char *addr,
-+ struct net_device *dev)
-+{
-+ int i;
-+
-+ for (i = 0; i < dev->addr_len; i++)
-+ mask[i] |= addr[i] ^ dev->dev_addr[i];
-+}
-+
-+static void __netdev_upper_mask(unsigned char *mask, struct net_device *dev,
-+ struct net_device *lower)
-+{
-+ struct net_device *cur;
-+ struct list_head *iter;
-+
-+ netdev_for_each_upper_dev_rcu(dev, cur, iter) {
-+ __netdev_addr_mask(mask, cur->dev_addr, lower);
-+ __netdev_upper_mask(mask, cur, lower);
-+ }
-+}
-+
-+static void __netdev_update_addr_mask(struct net_device *dev)
-+{
-+ unsigned char mask[MAX_ADDR_LEN];
-+ struct net_device *cur;
-+ struct list_head *iter;
-+
-+ memset(mask, 0, sizeof(mask));
-+ __netdev_upper_mask(mask, dev, dev);
-+ memcpy(dev->local_addr_mask, mask, dev->addr_len);
-+
-+ netdev_for_each_lower_dev(dev, cur, iter)
-+ __netdev_update_addr_mask(cur);
-+}
-+
-+static void netdev_update_addr_mask(struct net_device *dev)
-+{
-+ rcu_read_lock();
-+ __netdev_update_addr_mask(dev);
-+ rcu_read_unlock();
-+}
-+
- static int __netdev_upper_dev_link(struct net_device *dev,
- struct net_device *upper_dev, bool master,
- void *upper_priv, void *upper_info,
-@@ -7740,6 +7782,7 @@ static int __netdev_upper_dev_link(struc
- if (ret)
- return ret;
-
-+ netdev_update_addr_mask(dev);
- ret = call_netdevice_notifiers_info(NETDEV_CHANGEUPPER,
- &changeupper_info.info);
- ret = notifier_to_errno(ret);
-@@ -7836,6 +7879,7 @@ static void __netdev_upper_dev_unlink(st
-
- __netdev_adjacent_dev_unlink_neighbour(dev, upper_dev);
-
-+ netdev_update_addr_mask(dev);
- call_netdevice_notifiers_info(NETDEV_CHANGEUPPER,
- &changeupper_info.info);
-
-@@ -8892,6 +8936,7 @@ int dev_set_mac_address(struct net_devic
- return err;
- }
- dev->addr_assign_type = NET_ADDR_SET;
-+ netdev_update_addr_mask(dev);
- call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
- add_device_randomness(dev->dev_addr, dev->addr_len);
- return 0;
---- a/net/ethernet/eth.c
-+++ b/net/ethernet/eth.c
-@@ -143,6 +143,18 @@ u32 eth_get_headlen(const struct net_dev
- }
- EXPORT_SYMBOL(eth_get_headlen);
-
-+static inline bool
-+eth_check_local_mask(const void *addr1, const void *addr2, const void *mask)
-+{
-+ const u16 *a1 = addr1;
-+ const u16 *a2 = addr2;
-+ const u16 *m = mask;
-+
-+ return (((a1[0] ^ a2[0]) & ~m[0]) |
-+ ((a1[1] ^ a2[1]) & ~m[1]) |
-+ ((a1[2] ^ a2[2]) & ~m[2]));
-+}
-+
- /**
- * eth_type_trans - determine the packet's protocol ID.
- * @skb: received socket data
-@@ -174,6 +186,10 @@ __be16 eth_type_trans(struct sk_buff *sk
- } else {
- skb->pkt_type = PACKET_OTHERHOST;
- }
-+
-+ if (eth_check_local_mask(eth->h_dest, dev->dev_addr,
-+ dev->local_addr_mask))
-+ skb->gro_skip = 1;
- }
-
- /*
--- /dev/null
+From: Felix Fietkau <nbd@nbd.name>
+Date: Tue, 23 Apr 2024 11:23:03 +0200
+Subject: [PATCH] net: add TCP fraglist GRO support
+
+When forwarding TCP after GRO, software segmentation is very expensive,
+especially when the checksum needs to be recalculated.
+One case where that's currently unavoidable is when routing packets over
+PPPoE. Performance improves significantly when using fraglist GRO
+implemented in the same way as for UDP.
+
+Here's a measurement of running 2 TCP streams through a MediaTek MT7622
+device (2-core Cortex-A53), which runs NAT with flow offload enabled from
+one ethernet port to PPPoE on another ethernet port + cake qdisc set to
+1Gbps.
+
+rx-gro-list off: 630 Mbit/s, CPU 35% idle
+rx-gro-list on: 770 Mbit/s, CPU 40% idle
+
+Signe-off-by: Felix Fietkau <nbd@nbd.name>
+---
+
+--- a/include/net/gro.h
++++ b/include/net/gro.h
+@@ -430,6 +430,7 @@ static inline __wsum ip6_gro_compute_pse
+ }
+
+ int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
++int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
+
+ /* Pass the currently batched GRO_NORMAL SKBs up to the stack. */
+ static inline void gro_normal_list(struct napi_struct *napi)
+--- a/include/net/tcp.h
++++ b/include/net/tcp.h
+@@ -2082,7 +2082,8 @@ void tcp_v4_destroy_sock(struct sock *sk
+
+ struct sk_buff *tcp_gso_segment(struct sk_buff *skb,
+ netdev_features_t features);
+-struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb);
++struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb,
++ bool fraglist);
+ INDIRECT_CALLABLE_DECLARE(int tcp4_gro_complete(struct sk_buff *skb, int thoff));
+ INDIRECT_CALLABLE_DECLARE(struct sk_buff *tcp4_gro_receive(struct list_head *head, struct sk_buff *skb));
+ INDIRECT_CALLABLE_DECLARE(int tcp6_gro_complete(struct sk_buff *skb, int thoff));
+--- a/net/core/gro.c
++++ b/net/core/gro.c
+@@ -233,6 +233,33 @@ done:
+ return 0;
+ }
+
++int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb)
++{
++ if (unlikely(p->len + skb->len >= 65536))
++ return -E2BIG;
++
++ if (NAPI_GRO_CB(p)->last == p)
++ skb_shinfo(p)->frag_list = skb;
++ else
++ NAPI_GRO_CB(p)->last->next = skb;
++
++ skb_pull(skb, skb_gro_offset(skb));
++
++ NAPI_GRO_CB(p)->last = skb;
++ NAPI_GRO_CB(p)->count++;
++ p->data_len += skb->len;
++
++ /* sk ownership - if any - completely transferred to the aggregated packet */
++ skb->destructor = NULL;
++ skb->sk = NULL;
++ p->truesize += skb->truesize;
++ p->len += skb->len;
++
++ NAPI_GRO_CB(skb)->same_flow = 1;
++
++ return 0;
++}
++
+
+ static void napi_gro_complete(struct napi_struct *napi, struct sk_buff *skb)
+ {
+--- a/net/ipv4/tcp_offload.c
++++ b/net/ipv4/tcp_offload.c
+@@ -28,6 +28,74 @@ static void tcp_gso_tstamp(struct sk_buf
+ }
+ }
+
++static void __tcpv4_gso_segment_csum(struct sk_buff *seg,
++ __be32 *oldip, __be32 *newip,
++ __be16 *oldport, __be16 *newport)
++{
++ struct tcphdr *th;
++ struct iphdr *iph;
++
++ if (*oldip == *newip && *oldport == *newport)
++ return;
++
++ th = tcp_hdr(seg);
++ iph = ip_hdr(seg);
++
++ if (th->check) {
++ inet_proto_csum_replace4(&th->check, seg, *oldip, *newip,
++ true);
++ inet_proto_csum_replace2(&th->check, seg, *oldport, *newport,
++ false);
++ if (!th->check)
++ th->check = CSUM_MANGLED_0;
++ }
++ *oldport = *newport;
++
++ csum_replace4(&iph->check, *oldip, *newip);
++ *oldip = *newip;
++}
++
++static struct sk_buff *__tcpv4_gso_segment_list_csum(struct sk_buff *segs)
++{
++ struct sk_buff *seg;
++ struct tcphdr *uh, *uh2;
++ struct iphdr *iph, *iph2;
++
++ seg = segs;
++ uh = tcp_hdr(seg);
++ iph = ip_hdr(seg);
++
++ if ((tcp_hdr(seg)->dest == tcp_hdr(seg->next)->dest) &&
++ (tcp_hdr(seg)->source == tcp_hdr(seg->next)->source) &&
++ (ip_hdr(seg)->daddr == ip_hdr(seg->next)->daddr) &&
++ (ip_hdr(seg)->saddr == ip_hdr(seg->next)->saddr))
++ return segs;
++
++ while ((seg = seg->next)) {
++ uh2 = tcp_hdr(seg);
++ iph2 = ip_hdr(seg);
++
++ __tcpv4_gso_segment_csum(seg,
++ &iph2->saddr, &iph->saddr,
++ &uh2->source, &uh->source);
++ __tcpv4_gso_segment_csum(seg,
++ &iph2->daddr, &iph->daddr,
++ &uh2->dest, &uh->dest);
++ }
++
++ return segs;
++}
++
++static struct sk_buff *__tcp_gso_segment_list(struct sk_buff *skb,
++ netdev_features_t features)
++{
++ skb = skb_segment_list(skb, features, skb_mac_header_len(skb));
++ if (IS_ERR(skb))
++ return skb;
++
++ return __tcpv4_gso_segment_list_csum(skb);
++}
++
+ static struct sk_buff *tcp4_gso_segment(struct sk_buff *skb,
+ netdev_features_t features)
+ {
+@@ -37,6 +105,9 @@ static struct sk_buff *tcp4_gso_segment(
+ if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
+ return ERR_PTR(-EINVAL);
+
++ if (skb_shinfo(skb)->gso_type & SKB_GSO_FRAGLIST)
++ return __tcp_gso_segment_list(skb, features);
++
+ if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
+ const struct iphdr *iph = ip_hdr(skb);
+ struct tcphdr *th = tcp_hdr(skb);
+@@ -178,7 +249,8 @@ out:
+ return segs;
+ }
+
+-struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb)
++struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb,
++ bool fraglist)
+ {
+ struct sk_buff *pp = NULL;
+ struct sk_buff *p;
+@@ -215,6 +287,7 @@ struct sk_buff *tcp_gro_receive(struct l
+ len = skb_gro_len(skb);
+ flags = tcp_flag_word(th);
+
++ NAPI_GRO_CB(skb)->is_flist = fraglist;
+ list_for_each_entry(p, head, list) {
+ if (!NAPI_GRO_CB(p)->same_flow)
+ continue;
+@@ -234,6 +307,7 @@ struct sk_buff *tcp_gro_receive(struct l
+ found:
+ /* Include the IP ID check below from the inner most IP hdr */
+ flush = NAPI_GRO_CB(p)->flush;
++ flush |= fraglist != NAPI_GRO_CB(p)->is_flist;
+ flush |= (__force int)(flags & TCP_FLAG_CWR);
+ flush |= (__force int)((flags ^ tcp_flag_word(th2)) &
+ ~(TCP_FLAG_CWR | TCP_FLAG_FIN | TCP_FLAG_PSH));
+@@ -269,6 +343,19 @@ found:
+ flush |= p->decrypted ^ skb->decrypted;
+ #endif
+
++ if (fraglist) {
++ flush |= (__force int)(flags ^ tcp_flag_word(th2));
++ flush |= skb->ip_summed != p->ip_summed;
++ flush |= skb->csum_level != p->csum_level;
++ flush |= !pskb_may_pull(skb, skb_gro_offset(skb));
++ flush |= NAPI_GRO_CB(p)->count >= 64;
++
++ if (flush || skb_gro_receive_list(p, skb))
++ mss = 1;
++
++ goto out_check_final;
++ }
++
+ if (flush || skb_gro_receive(p, skb)) {
+ mss = 1;
+ goto out_check_final;
+@@ -314,6 +401,49 @@ void tcp_gro_complete(struct sk_buff *sk
+ }
+ EXPORT_SYMBOL(tcp_gro_complete);
+
++static bool tcp4_check_fraglist_gro(struct sk_buff *skb)
++{
++ const struct iphdr *iph = skb_gro_network_header(skb);
++ struct net *net = dev_net(skb->dev);
++ unsigned int off, hlen, thlen;
++ struct tcphdr *th;
++ struct sock *sk;
++ int iif, sdif;
++
++ if (!(skb->dev->features & NETIF_F_GRO_FRAGLIST))
++ return false;
++
++ inet_get_iif_sdif(skb, &iif, &sdif);
++
++ off = skb_gro_offset(skb);
++ hlen = off + sizeof(*th);
++ th = skb_gro_header(skb, hlen, off);
++ if (unlikely(!th))
++ return false;
++
++ thlen = th->doff * 4;
++ if (thlen < sizeof(*th))
++ return false;
++
++ hlen = off + thlen;
++ if (skb_gro_header_hard(skb, hlen)) {
++ th = skb_gro_header_slow(skb, hlen, off);
++ if (unlikely(!th))
++ return false;
++ }
++
++ sk = __inet_lookup_established(net, net->ipv4.tcp_death_row.hashinfo,
++ iph->saddr, th->source,
++ iph->daddr, ntohs(th->dest),
++ iif, sdif);
++ if (!sk)
++ return true;
++
++ sock_put(sk);
++
++ return false;
++}
++
+ INDIRECT_CALLABLE_SCOPE
+ struct sk_buff *tcp4_gro_receive(struct list_head *head, struct sk_buff *skb)
+ {
+@@ -325,7 +455,7 @@ struct sk_buff *tcp4_gro_receive(struct
+ return NULL;
+ }
+
+- return tcp_gro_receive(head, skb);
++ return tcp_gro_receive(head, skb, tcp4_check_fraglist_gro(skb));
+ }
+
+ INDIRECT_CALLABLE_SCOPE int tcp4_gro_complete(struct sk_buff *skb, int thoff)
+@@ -333,6 +463,15 @@ INDIRECT_CALLABLE_SCOPE int tcp4_gro_com
+ const struct iphdr *iph = ip_hdr(skb);
+ struct tcphdr *th = tcp_hdr(skb);
+
++ if (NAPI_GRO_CB(skb)->is_flist) {
++ skb_shinfo(skb)->gso_type |= SKB_GSO_FRAGLIST | SKB_GSO_TCPV4;
++ skb_shinfo(skb)->gso_segs = NAPI_GRO_CB(skb)->count;
++
++ __skb_incr_checksum_unnecessary(skb);
++
++ return 0;
++ }
++
+ th->check = ~tcp_v4_check(skb->len - thoff, iph->saddr,
+ iph->daddr, 0);
+ skb_shinfo(skb)->gso_type |= SKB_GSO_TCPV4;
+--- a/net/ipv4/udp_offload.c
++++ b/net/ipv4/udp_offload.c
+@@ -433,33 +433,6 @@ out:
+ return segs;
+ }
+
+-static int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb)
+-{
+- if (unlikely(p->len + skb->len >= 65536))
+- return -E2BIG;
+-
+- if (NAPI_GRO_CB(p)->last == p)
+- skb_shinfo(p)->frag_list = skb;
+- else
+- NAPI_GRO_CB(p)->last->next = skb;
+-
+- skb_pull(skb, skb_gro_offset(skb));
+-
+- NAPI_GRO_CB(p)->last = skb;
+- NAPI_GRO_CB(p)->count++;
+- p->data_len += skb->len;
+-
+- /* sk ownership - if any - completely transferred to the aggregated packet */
+- skb->destructor = NULL;
+- skb->sk = NULL;
+- p->truesize += skb->truesize;
+- p->len += skb->len;
+-
+- NAPI_GRO_CB(skb)->same_flow = 1;
+-
+- return 0;
+-}
+-
+
+ #define UDP_GRO_CNT_MAX 64
+ static struct sk_buff *udp_gro_receive_segment(struct list_head *head,
+--- a/net/ipv6/tcpv6_offload.c
++++ b/net/ipv6/tcpv6_offload.c
+@@ -7,12 +7,56 @@
+ */
+ #include <linux/indirect_call_wrapper.h>
+ #include <linux/skbuff.h>
++#include <net/inet6_hashtables.h>
+ #include <net/gro.h>
+ #include <net/protocol.h>
+ #include <net/tcp.h>
+ #include <net/ip6_checksum.h>
+ #include "ip6_offload.h"
+
++static bool tcp6_check_fraglist_gro(struct sk_buff *skb)
++{
++ const struct ipv6hdr *hdr = skb_gro_network_header(skb);
++ struct net *net = dev_net(skb->dev);
++ unsigned int off, hlen, thlen;
++ struct tcphdr *th;
++ struct sock *sk;
++ int iif, sdif;
++
++ if (!(skb->dev->features & NETIF_F_GRO_FRAGLIST))
++ return false;
++
++ inet6_get_iif_sdif(skb, &iif, &sdif);
++
++ off = skb_gro_offset(skb);
++ hlen = off + sizeof(*th);
++ th = skb_gro_header(skb, hlen, off);
++ if (unlikely(!th))
++ return false;
++
++ thlen = th->doff * 4;
++ if (thlen < sizeof(*th))
++ return false;
++
++ hlen = off + thlen;
++ if (skb_gro_header_hard(skb, hlen)) {
++ th = skb_gro_header_slow(skb, hlen, off);
++ if (unlikely(!th))
++ return false;
++ }
++
++ sk = __inet6_lookup_established(net, net->ipv4.tcp_death_row.hashinfo,
++ &hdr->saddr, th->source,
++ &hdr->daddr, ntohs(th->dest),
++ iif, sdif);
++ if (!sk)
++ return true;
++
++ sock_put(sk);
++
++ return false;
++}
++
+ INDIRECT_CALLABLE_SCOPE
+ struct sk_buff *tcp6_gro_receive(struct list_head *head, struct sk_buff *skb)
+ {
+@@ -24,7 +68,7 @@ struct sk_buff *tcp6_gro_receive(struct
+ return NULL;
+ }
+
+- return tcp_gro_receive(head, skb);
++ return tcp_gro_receive(head, skb, tcp6_check_fraglist_gro(skb));
+ }
+
+ INDIRECT_CALLABLE_SCOPE int tcp6_gro_complete(struct sk_buff *skb, int thoff)
+@@ -32,6 +76,15 @@ INDIRECT_CALLABLE_SCOPE int tcp6_gro_com
+ const struct ipv6hdr *iph = ipv6_hdr(skb);
+ struct tcphdr *th = tcp_hdr(skb);
+
++ if (NAPI_GRO_CB(skb)->is_flist) {
++ skb_shinfo(skb)->gso_type |= SKB_GSO_FRAGLIST | SKB_GSO_TCPV6;
++ skb_shinfo(skb)->gso_segs = NAPI_GRO_CB(skb)->count;
++
++ __skb_incr_checksum_unnecessary(skb);
++
++ return 0;
++ }
++
+ th->check = ~tcp_v6_check(skb->len - thoff, &iph->saddr,
+ &iph->daddr, 0);
+ skb_shinfo(skb)->gso_type |= SKB_GSO_TCPV6;
+@@ -51,6 +104,9 @@ static struct sk_buff *tcp6_gso_segment(
+ if (!pskb_may_pull(skb, sizeof(*th)))
+ return ERR_PTR(-EINVAL);
+
++ if (skb_shinfo(skb)->gso_type & SKB_GSO_FRAGLIST)
++ return skb_segment_list(skb, features, skb_mac_header_len(skb));
++
+ if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
+ const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
+ struct tcphdr *th = tcp_hdr(skb);
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
-@@ -2245,7 +2245,7 @@ struct net_device {
+@@ -2243,7 +2243,7 @@ struct net_device {
#if IS_ENABLED(CONFIG_AX25)
void *ax25_ptr;
#endif
/**
* napi_disable - prevent NAPI from scheduling
-@@ -3238,6 +3239,7 @@ struct softnet_data {
+@@ -3236,6 +3237,7 @@ struct softnet_data {
/* stats */
unsigned int processed;
unsigned int time_squeeze;
void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
int (*poll)(struct napi_struct *, int), int weight)
{
-@@ -11351,6 +11422,9 @@ static int dev_cpu_dead(unsigned int old
+@@ -11306,6 +11377,9 @@ static int dev_cpu_dead(unsigned int old
raise_softirq_irqoff(NET_TX_SOFTIRQ);
local_irq_enable();
#ifdef CONFIG_RPS
remsd = oldsd->rps_ipi_list;
oldsd->rps_ipi_list = NULL;
-@@ -11666,6 +11740,7 @@ static int __init net_dev_init(void)
+@@ -11621,6 +11695,7 @@ static int __init net_dev_init(void)
INIT_CSD(&sd->defer_csd, trigger_rx_softirq, sd);
spin_lock_init(&sd->defer_lock);
projects / openwrt / staging / nbd.git / commitdiff