From ff52cd74910cb21ff3e612ddb359101ce7ad274e Mon Sep 17 00:00:00 2001 From: W_Y_CPP <383152993@qq.com> Date: Wed, 19 Jan 2022 04:56:01 -0500 Subject: [PATCH] net-patch-linux-kernel-to-support-shortcut-fe --- include/linux/if_bridge.h | 3 +++ include/linux/skbuff.h | 4 +++ include/linux/timer.h | 4 +++ include/net/netfilter/nf_conntrack_ecache.h | 2 ++ net/Kconfig | 3 +++ net/bridge/br_if.c | 22 ++++++++++++++++ net/core/dev.c | 10 +++++++++++++++++++-- net/netfilter/nf_conntrack_ecache.c | 25 ++++++++++++++++-- 8 files changed, 115 insertions(+), 7 deletions(-) diff --git a/include/linux/if_bridge.h b/include/linux/if_bridge.h index 509e18c7e..d0ea4afbb 100644 --- a/include/linux/if_bridge.h +++ b/include/linux/if_bridge.h @@ -68,6 +68,9 @@ void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br, int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd, struct ifreq *ifr, void __user *uarg); +extern void br_dev_update_stats(struct net_device *dev, + struct rtnl_link_stats64 *nlstats); + #if IS_ENABLED(CONFIG_BRIDGE) && IS_ENABLED(CONFIG_BRIDGE_IGMP_SNOOPING) int br_multicast_list_adjacent(struct net_device *dev, struct list_head *br_ip_list); diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h index b8c273af2..2e238037c 100644 --- a/include/linux/skbuff.h +++ b/include/linux/skbuff.h @@ -873,6 +873,10 @@ struct sk_buff { #endif __u8 slow_gro:1; +#ifdef CONFIG_SHORTCUT_FE + __u8 fast_forwarded:1; +#endif + #ifdef CONFIG_NET_SCHED __u16 tc_index; /* traffic control index */ #endif diff --git a/include/linux/timer.h b/include/linux/timer.h index fda13c9d1..b69492fd4 100644 --- a/include/linux/timer.h +++ b/include/linux/timer.h @@ -18,6 +18,10 @@ struct timer_list { void (*function)(struct timer_list *); u32 flags; +#ifdef CONFIG_SHORTCUT_FE + unsigned long cust_data; +#endif + #ifdef CONFIG_LOCKDEP struct lockdep_map lockdep_map; #endif diff --git a/include/net/netfilter/nf_conntrack_ecache.h b/include/net/netfilter/nf_conntrack_ecache.h index c45657013..3c02392ea 100644 --- a/include/net/netfilter/nf_conntrack_ecache.h +++ b/include/net/netfilter/nf_conntrack_ecache.h @@ -86,6 +86,8 @@ struct nf_ct_event_notifier { #ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS extern int nf_conntrack_register_notifier(struct net *net, struct notifier_block *nb); extern int nf_conntrack_unregister_notifier(struct net *net, struct notifier_block *nb); +extern int nf_conntrack_register_chain_notifier(struct net *net, struct notifier_block *nb); +extern int nf_conntrack_unregister_chain_notifier(struct net *net, struct notifier_block *nb); #else int nf_conntrack_register_notifier(struct net *net, const struct nf_ct_event_notifier *nb); diff --git a/net/Kconfig b/net/Kconfig index fb13460c6..2e926a2ed 100644 --- a/net/Kconfig +++ b/net/Kconfig @@ -447,6 +447,9 @@ config FAILOVER migration of VMs with direct attached VFs by failing over to the paravirtual datapath when the VF is unplugged. +config SHORTCUT_FE + bool "Enables kernel network stack path for Shortcut Forwarding Engine" + config ETHTOOL_NETLINK bool "Netlink interface for ethtool" default y diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c index 4a02f8bb2..8a85ab557 100644 --- a/net/bridge/br_if.c +++ b/net/bridge/br_if.c @@ -773,6 +773,28 @@ void br_port_flags_change(struct net_bridge_port *p, unsigned long mask) br_recalculate_neigh_suppress_enabled(br); } +void br_dev_update_stats(struct net_device *dev, + struct rtnl_link_stats64 *nlstats) +{ + + struct pcpu_sw_netstats *stats; + + /* Is this a bridge? */ + if (!(dev->priv_flags & IFF_EBRIDGE)) + return; + + + stats = this_cpu_ptr(dev->tstats); + + u64_stats_update_begin(&stats->syncp); + stats->rx_packets += nlstats->rx_packets; + stats->rx_bytes += nlstats->rx_bytes; + stats->tx_packets += nlstats->tx_packets; + stats->tx_bytes += nlstats->tx_bytes; + u64_stats_update_end(&stats->syncp); +} +EXPORT_SYMBOL_GPL(br_dev_update_stats); + bool br_port_flag_is_set(const struct net_device *dev, unsigned long flag) { struct net_bridge_port *p; diff --git a/net/core/dev.c b/net/core/dev.c index e0878a500..4a9846b64 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -3574,9 +3574,17 @@ static int xmit_one(struct sk_buff *skb, struct net_device *dev, { unsigned int len; int rc; - +#ifdef CONFIG_SHORTCUT_FE + /* If this skb has been fast forwarded then we don't want it to + * go to any taps (by definition we're trying to bypass them). + */ + if (!skb->fast_forwarded) { +#endif if (dev_nit_active(dev)) dev_queue_xmit_nit(skb, dev); +#ifdef CONFIG_SHORTCUT_FE + } +#endif #ifdef CONFIG_ETHERNET_PACKET_MANGLE if (dev->eth_mangle_tx && !(skb = dev->eth_mangle_tx(dev, skb))) @@ -5198,6 +5205,11 @@ void netdev_rx_handler_unregister(struct net_device *dev) } EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister); +#ifdef CONFIG_SHORTCUT_FE +int (*athrs_fast_nat_recv)(struct sk_buff *skb) __rcu __read_mostly; +EXPORT_SYMBOL_GPL(athrs_fast_nat_recv); +#endif + /* * Limit the use of PFMEMALLOC reserves to those protocols that implement * the special handling of PFMEMALLOC skbs. @@ -5246,6 +5258,10 @@ static int __netif_receive_skb_core(struct sk_buff **pskb, bool pfmemalloc, int ret = NET_RX_DROP; __be16 type; +#ifdef CONFIG_SHORTCUT_FE + int (*fast_recv)(struct sk_buff *skb); +#endif + net_timestamp_check(!netdev_tstamp_prequeue, skb); trace_netif_receive_skb(skb); @@ -5283,6 +5299,15 @@ another_round: goto out; } +#ifdef CONFIG_SHORTCUT_FE + fast_recv = rcu_dereference(athrs_fast_nat_recv); + if (fast_recv) { + if (fast_recv(skb)) { + ret = NET_RX_SUCCESS; + goto out; + } + } +#endif if (skb_skip_tc_classify(skb)) goto skip_classify; diff --git a/net/netfilter/nf_conntrack_ecache.c b/net/netfilter/nf_conntrack_ecache.c index 9d8075db0..fc49b77e4 100644 --- a/net/netfilter/nf_conntrack_ecache.c +++ b/net/netfilter/nf_conntrack_ecache.c @@ -149,12 +149,23 @@ static int __nf_conntrack_eventmask_report(struct nf_conntrack_ecache *e, rcu_read_lock(); notify = rcu_dereference(net->ct.nf_conntrack_event_cb); - if (!notify) { +#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS + if (!notify && !rcu_dereference_raw(net->ct.nf_conntrack_chain.head)) +#else + if (!notify) +#endif + { rcu_read_unlock(); return 0; } - +#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS + ret = atomic_notifier_call_chain(&net->ct.nf_conntrack_chain, + events | missed, &item); + if (notify) + ret = notify->ct_event(events | missed, item); +#else ret = notify->ct_event(events | missed, item); +#endif rcu_read_unlock(); if (likely(ret >= 0 && missed == 0)) @@ -345,6 +356,11 @@ int nf_conntrack_register_notifier(struct net *net, { return atomic_notifier_chain_register(&net->ct.nf_conntrack_chain, nb); } +int nf_conntrack_register_chain_notifier(struct net *net, struct notifier_block *nb) +{ + return atomic_notifier_chain_register(&net->ct.nf_conntrack_chain, nb); +} +EXPORT_SYMBOL_GPL(nf_conntrack_register_chain_notifier); #else int nf_conntrack_register_notifier(struct net *net, const struct nf_ct_event_notifier *new) @@ -375,6 +391,11 @@ int nf_conntrack_unregister_notifier(struct net *net, struct notifier_block *nb) { return atomic_notifier_chain_unregister(&net->ct.nf_conntrack_chain, nb); } +int nf_conntrack_unregister_chain_notifier(struct net *net, struct notifier_block *nb) +{ + return atomic_notifier_chain_unregister(&net->ct.nf_conntrack_chain, nb); +} +EXPORT_SYMBOL_GPL(nf_conntrack_unregister_chain_notifier); #else void nf_conntrack_unregister_notifier(struct net *net) {