From 75a282ace767a8a36b3a9c40d7ace96e55575b68 Mon Sep 17 00:00:00 2001 From: coolsnowwolf Date: Thu, 3 Jan 2019 23:16:39 +0800 Subject: [PATCH] kernel: add SFE kernel patch --- ...k-events-support-multiple-registrant.patch | 325 +++++++++++++++++ ...-linux-kernel-to-support-shortcut-fe.patch | 140 +++++++ ...ridge-APIs-needed-for-network-HW-acc.patch | 74 ++++ ...k-events-support-multiple-registrant.patch | 345 ++++++++++++++++++ 4 files changed, 884 insertions(+) create mode 100644 target/linux/generic/hack-4.14/952-net-conntrack-events-support-multiple-registrant.patch create mode 100644 target/linux/generic/hack-4.9/950-net-patch-linux-kernel-to-support-shortcut-fe.patch create mode 100644 target/linux/generic/hack-4.9/951-bridge-Add-new-bridge-APIs-needed-for-network-HW-acc.patch create mode 100644 target/linux/generic/hack-4.9/952-net-conntrack-events-support-multiple-registrant.patch diff --git a/target/linux/generic/hack-4.14/952-net-conntrack-events-support-multiple-registrant.patch b/target/linux/generic/hack-4.14/952-net-conntrack-events-support-multiple-registrant.patch new file mode 100644 index 000000000..f844328db --- /dev/null +++ b/target/linux/generic/hack-4.14/952-net-conntrack-events-support-multiple-registrant.patch @@ -0,0 +1,325 @@ +diff --git a/include/net/netfilter/nf_conntrack_ecache.h b/include/net/netfilter/nf_conntrack_ecache.h +index 12d967b..c2b98b6 100644 +--- a/include/net/netfilter/nf_conntrack_ecache.h ++++ b/include/net/netfilter/nf_conntrack_ecache.h +@@ -71,6 +71,10 @@ struct nf_ct_event { + int report; + }; + ++#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); ++#else + struct nf_ct_event_notifier { + int (*fcn)(unsigned int events, struct nf_ct_event *item); + }; +@@ -79,7 +83,7 @@ int nf_conntrack_register_notifier(struc + struct nf_ct_event_notifier *nb); + void nf_conntrack_unregister_notifier(struct net *net, + struct nf_ct_event_notifier *nb); +- ++#endif + void nf_ct_deliver_cached_events(struct nf_conn *ct); + int nf_conntrack_eventmask_report(unsigned int eventmask, struct nf_conn *ct, + u32 portid, int report); +@@ -87,12 +91,15 @@ int nf_conntrack_eventmask_report(unsign + static inline void + nf_conntrack_event_cache(enum ip_conntrack_events event, struct nf_conn *ct) + { +- struct net *net = nf_ct_net(ct); + struct nf_conntrack_ecache *e; ++#ifndef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++ struct net *net = nf_ct_net(ct); + + if (!rcu_access_pointer(net->ct.nf_conntrack_event_cb)) + return; + ++#endif ++ + e = nf_ct_ecache_find(ct); + if (e == NULL) + return; +@@ -104,10 +111,12 @@ static inline int + nf_conntrack_event_report(enum ip_conntrack_events event, struct nf_conn *ct, + u32 portid, int report) + { ++#ifndef CONFIG_NF_CONNTRACK_CHAIN_EVENTS + const struct net *net = nf_ct_net(ct); + + if (!rcu_access_pointer(net->ct.nf_conntrack_event_cb)) + return 0; ++#endif + + return nf_conntrack_eventmask_report(1 << event, ct, portid, report); + } +@@ -115,11 +124,14 @@ nf_conntrack_event_report(enum ip_conntr + static inline int + nf_conntrack_event(enum ip_conntrack_events event, struct nf_conn *ct) + { ++#ifndef CONFIG_NF_CONNTRACK_CHAIN_EVENTS + const struct net *net = nf_ct_net(ct); + + if (!rcu_access_pointer(net->ct.nf_conntrack_event_cb)) + return 0; + ++#endif ++ + return nf_conntrack_eventmask_report(1 << event, ct, 0, 0); + } + +diff --git a/include/net/netns/conntrack.h b/include/net/netns/conntrack.h +index e469e85..1d31db8 100644 +--- a/include/net/netns/conntrack.h ++++ b/include/net/netns/conntrack.h +@@ -114,7 +114,11 @@ struct netns_ct { + + struct ct_pcpu __percpu *pcpu_lists; + struct ip_conntrack_stat __percpu *stat; ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++ struct atomic_notifier_head nf_conntrack_chain; ++#else + struct nf_ct_event_notifier __rcu *nf_conntrack_event_cb; ++#endif + struct nf_exp_event_notifier __rcu *nf_expect_event_cb; + struct nf_ip_net nf_ct_proto; + #if defined(CONFIG_NF_CONNTRACK_LABELS) +diff --git a/net/netfilter/Kconfig b/net/netfilter/Kconfig +index 63073be..08d7aab 100644 +--- a/net/netfilter/Kconfig ++++ b/net/netfilter/Kconfig +@@ -118,6 +118,14 @@ config NF_CONNTRACK_EVENTS + + If unsure, say `N'. + ++config NF_CONNTRACK_CHAIN_EVENTS ++ bool "Register multiple callbacks to ct events" ++ depends on NF_CONNTRACK_EVENTS ++ help ++ Support multiple registrations. ++ ++ If unsure, say `N'. ++ + config NF_CONNTRACK_TIMEOUT + bool 'Connection tracking timeout' + depends on NETFILTER_ADVANCED +diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c +index 6bd1508..9b81c7c 100644 +--- a/net/netfilter/nf_conntrack_core.c ++++ b/net/netfilter/nf_conntrack_core.c +@@ -2167,6 +2167,10 @@ int nf_conntrack_init_net(struct net *ne + ret = nf_conntrack_proto_pernet_init(net); + if (ret < 0) + goto err_proto; ++ ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++ ATOMIC_INIT_NOTIFIER_HEAD(&net->ct.nf_conntrack_chain); ++#endif + return 0; + + err_proto: +diff --git a/net/netfilter/nf_conntrack_ecache.c b/net/netfilter/nf_conntrack_ecache.c +index da9df2d..e0e2a8f 100644 +--- a/net/netfilter/nf_conntrack_ecache.c ++++ b/net/netfilter/nf_conntrack_ecache.c +@@ -18,6 +18,9 @@ + #include + #include + #include ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++#include ++#endif + #include + #include + #include +@@ -117,6 +120,38 @@ static void ecache_work(struct work_stru + schedule_delayed_work(&ctnet->ecache_dwork, delay); + } + ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++int ++nf_conntrack_eventmask_report(unsigned int eventmask, ++ struct nf_conn *ct, ++ u32 portid, ++ int report) ++{ ++ struct nf_conntrack_ecache *e; ++ struct net *net = nf_ct_net(ct); ++ ++ e = nf_ct_ecache_find(ct); ++ if (e == NULL) ++ return 0; ++ ++ if (nf_ct_is_confirmed(ct)) { ++ struct nf_ct_event item = { ++ .ct = ct, ++ .portid = e->portid ? e->portid : portid, ++ .report = report ++ }; ++ /* This is a resent of a destroy event? If so, skip missed */ ++ unsigned long missed = e->portid ? 0 : e->missed; ++ ++ if (!((eventmask | missed) & e->ctmask)) ++ return 0; ++ ++ atomic_notifier_call_chain(&net->ct.nf_conntrack_chain, eventmask | missed, &item); ++ } ++ ++ return 0; ++} ++#else + int nf_conntrack_eventmask_report(unsigned int eventmask, struct nf_conn *ct, + u32 portid, int report) + { +@@ -171,10 +206,52 @@ out_unlock: + rcu_read_unlock(); + return ret; + } ++#endif + EXPORT_SYMBOL_GPL(nf_conntrack_eventmask_report); + + /* deliver cached events and clear cache entry - must be called with locally + * disabled softirqs */ ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++void nf_ct_deliver_cached_events(struct nf_conn *ct) ++{ ++ unsigned long events, missed; ++ struct nf_conntrack_ecache *e; ++ struct nf_ct_event item; ++ struct net *net = nf_ct_net(ct); ++ ++ e = nf_ct_ecache_find(ct); ++ if (e == NULL) ++ return; ++ ++ events = xchg(&e->cache, 0); ++ ++ if (!nf_ct_is_confirmed(ct) || nf_ct_is_dying(ct) || !events) ++ return; ++ ++ /* We make a copy of the missed event cache without taking ++ * the lock, thus we may send missed events twice. However, ++ * this does not harm and it happens very rarely. */ ++ missed = e->missed; ++ ++ if (!((events | missed) & e->ctmask)) ++ return; ++ ++ item.ct = ct; ++ item.portid = 0; ++ item.report = 0; ++ ++ atomic_notifier_call_chain(&net->ct.nf_conntrack_chain, ++ events | missed, ++ &item); ++ ++ if (likely(!missed)) ++ return; ++ ++ spin_lock_bh(&ct->lock); ++ e->missed &= ~missed; ++ spin_unlock_bh(&ct->lock); ++} ++#else + void nf_ct_deliver_cached_events(struct nf_conn *ct) + { + struct net *net = nf_ct_net(ct); +@@ -225,6 +302,7 @@ void nf_ct_deliver_cached_events(struct + out_unlock: + rcu_read_unlock(); + } ++#endif + EXPORT_SYMBOL_GPL(nf_ct_deliver_cached_events); + + void nf_ct_expect_event_report(enum ip_conntrack_expect_events event, +@@ -257,6 +335,12 @@ out_unlock: + rcu_read_unlock(); + } + ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++int nf_conntrack_register_notifier(struct net *net, struct notifier_block *nb) ++{ ++ return atomic_notifier_chain_register(&net->ct.nf_conntrack_chain, nb); ++} ++#else + int nf_conntrack_register_notifier(struct net *net, + struct nf_ct_event_notifier *new) + { +@@ -277,8 +361,15 @@ out_unlock: + mutex_unlock(&nf_ct_ecache_mutex); + return ret; + } ++#endif + EXPORT_SYMBOL_GPL(nf_conntrack_register_notifier); + ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++int nf_conntrack_unregister_notifier(struct net *net, struct notifier_block *nb) ++{ ++ return atomic_notifier_chain_unregister(&net->ct.nf_conntrack_chain, nb); ++} ++#else + void nf_conntrack_unregister_notifier(struct net *net, + struct nf_ct_event_notifier *new) + { +@@ -292,6 +383,7 @@ void nf_conntrack_unregister_notifier(st + mutex_unlock(&nf_ct_ecache_mutex); + /* synchronize_rcu() is called from ctnetlink_exit. */ + } ++#endif + EXPORT_SYMBOL_GPL(nf_conntrack_unregister_notifier); + + int nf_ct_expect_register_notifier(struct net *net, +diff --git a/net/netfilter/nf_conntrack_netlink.c b/net/netfilter/nf_conntrack_netlink.c +index 04111c1..8c741f7 100644 +--- a/net/netfilter/nf_conntrack_netlink.c ++++ b/net/netfilter/nf_conntrack_netlink.c +@@ -28,6 +28,11 @@ + #include + #include + #include ++ ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++#include ++#endif ++ + #include + + #include +@@ -618,14 +623,22 @@ static size_t ctnetlink_nlmsg_size(const + ; + } + ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++static int ctnetlink_conntrack_event(struct notifier_block *this, ++ unsigned long events, void *ptr) ++#else + static int + ctnetlink_conntrack_event(unsigned int events, struct nf_ct_event *item) ++#endif + { + const struct nf_conntrack_zone *zone; + struct net *net; + struct nlmsghdr *nlh; + struct nfgenmsg *nfmsg; + struct nlattr *nest_parms; ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++ struct nf_ct_event *item = (struct nf_ct_event *)ptr; ++#endif + struct nf_conn *ct = item->ct; + struct sk_buff *skb; + unsigned int type; +@@ -3290,9 +3303,15 @@ static int ctnetlink_stat_exp_cpu(struct + } + + #ifdef CONFIG_NF_CONNTRACK_EVENTS ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++static struct notifier_block ctnl_notifier = { ++ .notifier_call = ctnetlink_conntrack_event, ++}; ++#else + static struct nf_ct_event_notifier ctnl_notifier = { + .fcn = ctnetlink_conntrack_event, + }; ++#endif + + static struct nf_exp_event_notifier ctnl_notifier_exp = { + .fcn = ctnetlink_expect_event, diff --git a/target/linux/generic/hack-4.9/950-net-patch-linux-kernel-to-support-shortcut-fe.patch b/target/linux/generic/hack-4.9/950-net-patch-linux-kernel-to-support-shortcut-fe.patch new file mode 100644 index 000000000..e7844effd --- /dev/null +++ b/target/linux/generic/hack-4.9/950-net-patch-linux-kernel-to-support-shortcut-fe.patch @@ -0,0 +1,140 @@ +From 5c43b9811dd88dd2d18d88a253212474bb665896 Mon Sep 17 00:00:00 2001 +From: Xiaoping Fan +Date: Fri, 26 Feb 2016 15:01:53 -0800 +Subject: [PATCH 1/3] net: patch linux kernel to support shortcut-fe + +1, add a new flag 'fast_forwarded' in skb structure. +2, put a hook in '__netif_receive_skb_core' to + deliver packet to shortcut-fe. + +Change-Id: Icaa7c172a06df1c3bc89ff89814d1136772fe217 +Signed-off-by: Xiaoping Fan + +msm: ipq806x: exporting TCP sequence check parameters + +This is for use in NSS connection manager. + +Change-Id: I01d30c0ab552308c439353c0d51d3d0ab3aa7699 +Signed-off-by: Pamidipati, Vijay +Signed-off-by: Murat Sezgin +--- + include/linux/skbuff.h | 3 +++ + net/Kconfig | 3 +++ + net/core/dev.c | 25 +++++++++++++++++++++++++ + net/netfilter/nf_conntrack_proto_tcp.c | 10 ++++++++++ + 4 files changed, 41 insertions(+) + +diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h +index 9a0c945..56f4007 100644 +--- a/include/linux/skbuff.h ++++ b/include/linux/skbuff.h +@@ -738,6 +738,9 @@ struct sk_buff { + #endif + __u8 ipvs_property:1; + __u8 inner_protocol_type:1; ++#ifdef CONFIG_SHORTCUT_FE ++ __u8 fast_forwarded:1; ++#endif + __u8 remcsum_offload:1; + #ifdef CONFIG_NET_SWITCHDEV + __u8 offload_fwd_mark:1; +diff --git a/net/Kconfig b/net/Kconfig +index b4621e1..6ba72cf 100644 +--- a/net/Kconfig ++++ b/net/Kconfig +@@ -444,3 +444,6 @@ config HAVE_CBPF_JIT + # Extended BPF JIT (eBPF) + config HAVE_EBPF_JIT + bool ++ ++config SHORTCUT_FE ++ bool "Enables kernel network stack path for Shortcut Forwarding Engine" +diff --git a/net/core/dev.c b/net/core/dev.c +index ca39cd2..74cd937 100644 +--- a/net/core/dev.c ++++ b/net/core/dev.c +@@ -2929,8 +2929,17 @@ 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 (!list_empty(&ptype_all) || !list_empty(&dev->ptype_all)) + dev_queue_xmit_nit(skb, dev); ++#ifdef CONFIG_SHORTCUT_FE ++ } ++#endif + + #ifdef CONFIG_ETHERNET_PACKET_MANGLE + if (!dev->eth_mangle_tx || +@@ -4061,6 +4070,11 @@ void netdev_rx_handler_unregister(struct net_device *dev) + } + EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister); + ++#ifdef CONFIG_SHORTCUT_FE ++int (*fast_nat_recv)(struct sk_buff *skb) __rcu __read_mostly; ++EXPORT_SYMBOL_GPL(fast_nat_recv); ++#endif ++ + /* + * Limit the use of PFMEMALLOC reserves to those protocols that implement + * the special handling of PFMEMALLOC skbs. +@@ -4108,6 +4122,9 @@ static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc) + bool deliver_exact = false; + 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); + +@@ -4134,6 +4151,14 @@ static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc) + goto out; + } + ++#ifdef CONFIG_SHORTCUT_FE ++ fast_recv = rcu_dereference(fast_nat_recv); ++ if (fast_recv && fast_recv(skb)) { ++ ret = NET_RX_SUCCESS; ++ goto out; ++ } ++#endif ++ + #ifdef CONFIG_NET_CLS_ACT + if (skb->tc_verd & TC_NCLS) { + skb->tc_verd = CLR_TC_NCLS(skb->tc_verd); +diff --git a/net/netfilter/nf_conntrack_proto_tcp.c b/net/netfilter/nf_conntrack_proto_tcp.c +index f24b626..4581481 100644 +--- a/net/netfilter/nf_conntrack_proto_tcp.c ++++ b/net/netfilter/nf_conntrack_proto_tcp.c +@@ -34,12 +34,22 @@ + #include + + /* Do not check the TCP window for incoming packets */ ++#ifdef CONFIG_SHORTCUT_FE ++int nf_ct_tcp_no_window_check __read_mostly = 0; ++EXPORT_SYMBOL_GPL(nf_ct_tcp_no_window_check); ++#else + static int nf_ct_tcp_no_window_check __read_mostly = 1; ++#endif + + /* "Be conservative in what you do, + be liberal in what you accept from others." + If it's non-zero, we mark only out of window RST segments as INVALID. */ ++#ifdef CONFIG_SHORTCUT_FE ++int nf_ct_tcp_be_liberal __read_mostly = 0; ++EXPORT_SYMBOL_GPL(nf_ct_tcp_be_liberal); ++#else + static int nf_ct_tcp_be_liberal __read_mostly = 0; ++#endif + + /* If it is set to zero, we disable picking up already established + connections. */ +-- +2.7.4 + diff --git a/target/linux/generic/hack-4.9/951-bridge-Add-new-bridge-APIs-needed-for-network-HW-acc.patch b/target/linux/generic/hack-4.9/951-bridge-Add-new-bridge-APIs-needed-for-network-HW-acc.patch new file mode 100644 index 000000000..5cc7dabad --- /dev/null +++ b/target/linux/generic/hack-4.9/951-bridge-Add-new-bridge-APIs-needed-for-network-HW-acc.patch @@ -0,0 +1,74 @@ +From cdabe46cd62d22266aed7fdae36f72a36e3558e7 Mon Sep 17 00:00:00 2001 +From: Murat Sezgin +Date: Tue, 25 Nov 2014 17:22:24 -0800 +Subject: [PATCH 2/3] bridge: Add new bridge APIs needed for network HW + acceleration + +Bridge acceleration hardware needs to perform certain operations, + currently unsupported by the existing bridge code: + + ** cut ** + + *update bridge interface statistics from outside the bridge code: + once acceleration is enabled on a connection, packets will not flow + through the host CPU, so we need the hardware accelerator driver to + maintain the statistics on the host and update them and add whatever + flows through the hardware. + These change adds the corresponding functions, and make it available + to other through EXPORT_SYMBOLS(). + +Change-Id: I67afb325796004053897d9916e2df91827b65139 +Signed-off-by: Murat Sezgin +--- + include/linux/if_bridge.h | 1 + + net/bridge/br_if.c | 25 +++++++++++++++++++++++++ + 2 files changed, 26 insertions(+) + +diff --git a/include/linux/if_bridge.h b/include/linux/if_bridge.h +index 0911c8c..4473f62 100644 +--- a/include/linux/if_bridge.h ++++ b/include/linux/if_bridge.h +@@ -52,6 +52,7 @@ struct br_ip_list { + #define BR_DEFAULT_AGEING_TIME (300 * HZ) + + extern void brioctl_set(int (*ioctl_hook)(struct net *, unsigned int, void __user *)); ++extern void br_dev_update_stats(struct net_device *dev, struct rtnl_link_stats64 *nlstats); + + typedef int br_should_route_hook_t(struct sk_buff *skb); + extern br_should_route_hook_t __rcu *br_should_route_hook; +diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c +index ed0dd33..342b2d9 100644 +--- a/net/bridge/br_if.c ++++ b/net/bridge/br_if.c +@@ -655,3 +655,28 @@ void br_port_flags_change(struct net_bridge_port *p, unsigned long mask) + if (mask & BR_AUTO_MASK) + nbp_update_port_count(br); + } ++ ++/* Update bridge statistics for bridge packets processed by offload engines */ ++void br_dev_update_stats(struct net_device *dev, struct rtnl_link_stats64 *nlstats) ++{ ++ struct net_bridge *br; ++ struct pcpu_sw_netstats *stats; ++ ++ /* ++ * Is this a bridge? ++ */ ++ if (!(dev->priv_flags & IFF_EBRIDGE)) { ++ return; ++ } ++ ++ br = netdev_priv(dev); ++ stats = this_cpu_ptr(br->stats); ++ ++ 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); +-- +2.7.4 + diff --git a/target/linux/generic/hack-4.9/952-net-conntrack-events-support-multiple-registrant.patch b/target/linux/generic/hack-4.9/952-net-conntrack-events-support-multiple-registrant.patch new file mode 100644 index 000000000..a898c4a69 --- /dev/null +++ b/target/linux/generic/hack-4.9/952-net-conntrack-events-support-multiple-registrant.patch @@ -0,0 +1,345 @@ +From 56294a39f14712f868a495f28be6b2e92f0a48c6 Mon Sep 17 00:00:00 2001 +From: Zhi Chen +Date: Tue, 13 Jan 2015 14:28:18 -0800 +Subject: [PATCH 3/3] net: conntrack events, support multiple registrant + +Merging this patch from kernel 3.4: +This was supported by old (.28) kernel versions but removed +because of it's overhead. +But we need this feature for NA connection manager. Both ipv4 +and ipv6 modules needs to register themselves to ct events. + +Change-Id: Iebfb254590fb594f5baf232f849d1b7ae45ef757 +Signed-off-by: Zhi Chen +--- + include/net/netfilter/nf_conntrack_ecache.h | 13 +++- + include/net/netns/conntrack.h | 4 ++ + net/netfilter/Kconfig | 8 +++ + net/netfilter/nf_conntrack_core.c | 4 ++ + net/netfilter/nf_conntrack_ecache.c | 92 +++++++++++++++++++++++++++++ + net/netfilter/nf_conntrack_netlink.c | 19 ++++++ + 6 files changed, 139 insertions(+), 1 deletion(-) + +diff --git a/include/net/netfilter/nf_conntrack_ecache.h b/include/net/netfilter/nf_conntrack_ecache.h +index 12d967b..c2b98b6 100644 +--- a/include/net/netfilter/nf_conntrack_ecache.h ++++ b/include/net/netfilter/nf_conntrack_ecache.h +@@ -70,6 +70,10 @@ struct nf_ct_event { + int report; + }; + ++#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); ++#else + struct nf_ct_event_notifier { + int (*fcn)(unsigned int events, struct nf_ct_event *item); + }; +@@ -78,6 +82,7 @@ int nf_conntrack_register_notifier(struct net *net, + struct nf_ct_event_notifier *nb); + void nf_conntrack_unregister_notifier(struct net *net, + struct nf_ct_event_notifier *nb); ++#endif + + void nf_ct_deliver_cached_events(struct nf_conn *ct); + int nf_conntrack_eventmask_report(unsigned int eventmask, struct nf_conn *ct, +@@ -86,11 +91,13 @@ int nf_conntrack_eventmask_report(unsigned int eventmask, struct nf_conn *ct, + static inline void + nf_conntrack_event_cache(enum ip_conntrack_events event, struct nf_conn *ct) + { +- struct net *net = nf_ct_net(ct); + struct nf_conntrack_ecache *e; ++#ifndef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++ struct net *net = nf_ct_net(ct); + + if (!rcu_access_pointer(net->ct.nf_conntrack_event_cb)) + return; ++#endif + + e = nf_ct_ecache_find(ct); + if (e == NULL) +@@ -103,10 +110,12 @@ static inline int + nf_conntrack_event_report(enum ip_conntrack_events event, struct nf_conn *ct, + u32 portid, int report) + { ++#ifndef CONFIG_NF_CONNTRACK_CHAIN_EVENTS + const struct net *net = nf_ct_net(ct); + + if (!rcu_access_pointer(net->ct.nf_conntrack_event_cb)) + return 0; ++#endif + + return nf_conntrack_eventmask_report(1 << event, ct, portid, report); + } +@@ -114,10 +123,12 @@ nf_conntrack_event_report(enum ip_conntrack_events event, struct nf_conn *ct, + static inline int + nf_conntrack_event(enum ip_conntrack_events event, struct nf_conn *ct) + { ++#ifndef CONFIG_NF_CONNTRACK_CHAIN_EVENTS + const struct net *net = nf_ct_net(ct); + + if (!rcu_access_pointer(net->ct.nf_conntrack_event_cb)) + return 0; ++#endif + + return nf_conntrack_eventmask_report(1 << event, ct, 0, 0); + } +diff --git a/include/net/netns/conntrack.h b/include/net/netns/conntrack.h +index e469e85..1d31db8 100644 +--- a/include/net/netns/conntrack.h ++++ b/include/net/netns/conntrack.h +@@ -86,7 +86,11 @@ struct netns_ct { + + struct ct_pcpu __percpu *pcpu_lists; + struct ip_conntrack_stat __percpu *stat; ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++ struct atomic_notifier_head nf_conntrack_chain; ++#else + struct nf_ct_event_notifier __rcu *nf_conntrack_event_cb; ++#endif + struct nf_exp_event_notifier __rcu *nf_expect_event_cb; + struct nf_ip_net nf_ct_proto; + #if defined(CONFIG_NF_CONNTRACK_LABELS) +diff --git a/net/netfilter/Kconfig b/net/netfilter/Kconfig +index 63073be..08d7aab 100644 +--- a/net/netfilter/Kconfig ++++ b/net/netfilter/Kconfig +@@ -136,6 +136,14 @@ config NF_CONNTRACK_TIMEOUT + + If unsure, say `N'. + ++config NF_CONNTRACK_CHAIN_EVENTS ++ bool "Register multiple callbacks to ct events" ++ depends on NF_CONNTRACK_EVENTS ++ help ++ Support multiple registrations. ++ ++ If unsure, say `N'. ++ + config NF_CONNTRACK_TIMESTAMP + bool 'Connection tracking timestamping' + depends on NETFILTER_ADVANCED +diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c +index 6bd1508..9b81c7c 100644 +--- a/net/netfilter/nf_conntrack_core.c ++++ b/net/netfilter/nf_conntrack_core.c +@@ -1998,6 +1998,10 @@ int nf_conntrack_init_net(struct net *net) + ret = nf_conntrack_proto_pernet_init(net); + if (ret < 0) + goto err_proto; ++ ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++ ATOMIC_INIT_NOTIFIER_HEAD(&net->ct.nf_conntrack_chain); ++#endif + return 0; + + err_proto: +diff --git a/net/netfilter/nf_conntrack_ecache.c b/net/netfilter/nf_conntrack_ecache.c +index da9df2d..e0e2a8f 100644 +--- a/net/netfilter/nf_conntrack_ecache.c ++++ b/net/netfilter/nf_conntrack_ecache.c +@@ -18,6 +18,9 @@ + #include + #include + #include ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++#include ++#endif + #include + #include + #include +@@ -117,6 +120,38 @@ static void ecache_work(struct work_struct *work) + schedule_delayed_work(&ctnet->ecache_dwork, delay); + } + ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++int ++nf_conntrack_eventmask_report(unsigned int eventmask, ++ struct nf_conn *ct, ++ u32 portid, ++ int report) ++{ ++ struct nf_conntrack_ecache *e; ++ struct net *net = nf_ct_net(ct); ++ ++ e = nf_ct_ecache_find(ct); ++ if (e == NULL) ++ return 0; ++ ++ if (nf_ct_is_confirmed(ct)) { ++ struct nf_ct_event item = { ++ .ct = ct, ++ .portid = e->portid ? e->portid : portid, ++ .report = report ++ }; ++ /* This is a resent of a destroy event? If so, skip missed */ ++ unsigned long missed = e->portid ? 0 : e->missed; ++ ++ if (!((eventmask | missed) & e->ctmask)) ++ return 0; ++ ++ atomic_notifier_call_chain(&net->ct.nf_conntrack_chain, eventmask | missed, &item); ++ } ++ ++ return 0; ++} ++#else + int nf_conntrack_eventmask_report(unsigned int eventmask, struct nf_conn *ct, + u32 portid, int report) + { +@@ -171,10 +206,52 @@ int nf_conntrack_eventmask_report(unsigned int eventmask, struct nf_conn *ct, + rcu_read_unlock(); + return ret; + } ++#endif + EXPORT_SYMBOL_GPL(nf_conntrack_eventmask_report); + + /* deliver cached events and clear cache entry - must be called with locally + * disabled softirqs */ ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++void nf_ct_deliver_cached_events(struct nf_conn *ct) ++{ ++ unsigned long events, missed; ++ struct nf_conntrack_ecache *e; ++ struct nf_ct_event item; ++ struct net *net = nf_ct_net(ct); ++ ++ e = nf_ct_ecache_find(ct); ++ if (e == NULL) ++ return; ++ ++ events = xchg(&e->cache, 0); ++ ++ if (!nf_ct_is_confirmed(ct) || nf_ct_is_dying(ct) || !events) ++ return; ++ ++ /* We make a copy of the missed event cache without taking ++ * the lock, thus we may send missed events twice. However, ++ * this does not harm and it happens very rarely. */ ++ missed = e->missed; ++ ++ if (!((events | missed) & e->ctmask)) ++ return; ++ ++ item.ct = ct; ++ item.portid = 0; ++ item.report = 0; ++ ++ atomic_notifier_call_chain(&net->ct.nf_conntrack_chain, ++ events | missed, ++ &item); ++ ++ if (likely(!missed)) ++ return; ++ ++ spin_lock_bh(&ct->lock); ++ e->missed &= ~missed; ++ spin_unlock_bh(&ct->lock); ++} ++#else + void nf_ct_deliver_cached_events(struct nf_conn *ct) + { + struct net *net = nf_ct_net(ct); +@@ -225,6 +302,7 @@ void nf_ct_deliver_cached_events(struct nf_conn *ct) + out_unlock: + rcu_read_unlock(); + } ++#endif + EXPORT_SYMBOL_GPL(nf_ct_deliver_cached_events); + + void nf_ct_expect_event_report(enum ip_conntrack_expect_events event, +@@ -257,6 +335,12 @@ void nf_ct_expect_event_report(enum ip_conntrack_expect_events event, + rcu_read_unlock(); + } + ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++int nf_conntrack_register_notifier(struct net *net, struct notifier_block *nb) ++{ ++ return atomic_notifier_chain_register(&net->ct.nf_conntrack_chain, nb); ++} ++#else + int nf_conntrack_register_notifier(struct net *net, + struct nf_ct_event_notifier *new) + { +@@ -277,8 +361,15 @@ int nf_conntrack_register_notifier(struct net *net, + mutex_unlock(&nf_ct_ecache_mutex); + return ret; + } ++#endif + EXPORT_SYMBOL_GPL(nf_conntrack_register_notifier); + ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++int nf_conntrack_unregister_notifier(struct net *net, struct notifier_block *nb) ++{ ++ return atomic_notifier_chain_unregister(&net->ct.nf_conntrack_chain, nb); ++} ++#else + void nf_conntrack_unregister_notifier(struct net *net, + struct nf_ct_event_notifier *new) + { +@@ -291,6 +382,7 @@ void nf_conntrack_unregister_notifier(struct net *net, + RCU_INIT_POINTER(net->ct.nf_conntrack_event_cb, NULL); + mutex_unlock(&nf_ct_ecache_mutex); + } ++#endif + EXPORT_SYMBOL_GPL(nf_conntrack_unregister_notifier); + + int nf_ct_expect_register_notifier(struct net *net, +diff --git a/net/netfilter/nf_conntrack_netlink.c b/net/netfilter/nf_conntrack_netlink.c +index 04111c1..8c741f7 100644 +--- a/net/netfilter/nf_conntrack_netlink.c ++++ b/net/netfilter/nf_conntrack_netlink.c +@@ -28,6 +28,11 @@ + #include + #include + #include ++ ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++#include ++#endif ++ + #include + + #include +@@ -615,14 +620,22 @@ static size_t ctnetlink_nlmsg_size(const struct nf_conn *ct) + ; + } + ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++static int ctnetlink_conntrack_event(struct notifier_block *this, ++ unsigned long events, void *ptr) ++#else + static int + ctnetlink_conntrack_event(unsigned int events, struct nf_ct_event *item) ++#endif + { + const struct nf_conntrack_zone *zone; + struct net *net; + struct nlmsghdr *nlh; + struct nfgenmsg *nfmsg; + struct nlattr *nest_parms; ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++ struct nf_ct_event *item = (struct nf_ct_event *)ptr; ++#endif + struct nf_conn *ct = item->ct; + struct sk_buff *skb; + unsigned int type; +@@ -3260,9 +3273,15 @@ static int ctnetlink_stat_exp_cpu(struct net *net, struct sock *ctnl, + } + + #ifdef CONFIG_NF_CONNTRACK_EVENTS ++#ifdef CONFIG_NF_CONNTRACK_CHAIN_EVENTS ++static struct notifier_block ctnl_notifier = { ++ .notifier_call = ctnetlink_conntrack_event, ++}; ++#else + static struct nf_ct_event_notifier ctnl_notifier = { + .fcn = ctnetlink_conntrack_event, + }; ++#endif + + static struct nf_exp_event_notifier ctnl_notifier_exp = { + .fcn = ctnetlink_expect_event, +-- +2.7.4 +