| /* |
| * Fastpath Learner |
| * |
| * This program is free software; you can redistribute it and/or |
| * modify it under the terms of the GNU FP_ERR( Public License |
| * as published by the Free Software Foundation; either version |
| * 2 of the License, or (at your option) any later version. |
| */ |
| #define pr_fmt(fmt) "mfp" " learner:%s:%d: " fmt, __func__, __LINE__ |
| |
| #include <br_private.h> |
| #include <net/addrconf.h> |
| #include <linux/inetdevice.h> |
| #include "fp_common.h" |
| #include "fp_database.h" |
| #include "fp_device.h" |
| #include "fp_core.h" |
| #include "fp_netlink.h" |
| |
| #define RTMGRP_IPV4_ROUTE 0x40 |
| #define RTMGRP_IPV4_RULE 0x80 |
| #define RTMGRP_IPV6_ROUTE 0x400 |
| #define RTNETLINK_GRP (RTMGRP_IPV4_ROUTE | RTMGRP_IPV4_RULE | RTMGRP_IPV6_ROUTE) |
| |
| #define NFLGRP2MASK(group) ((((group) > NFNLGRP_NONE) && \ |
| ((group) < __NFNLGRP_MAX)) ? \ |
| (0x1UL << ((group) - 1)) : 0) |
| |
| #define NFNETLINK_GRP \ |
| NFLGRP2MASK(NFNLGRP_CONNTRACK_NEW) | \ |
| NFLGRP2MASK(NFNLGRP_CONNTRACK_UPDATE) | \ |
| NFLGRP2MASK(NFNLGRP_CONNTRACK_DESTROY) | \ |
| NFLGRP2MASK(NFNLGRP_CONNTRACK_EXP_NEW) | \ |
| NFLGRP2MASK(NFNLGRP_CONNTRACK_EXP_UPDATE) | \ |
| NFLGRP2MASK(NFNLGRP_CONNTRACK_EXP_DESTROY) |
| |
| /* ipv6 special flags always rejected (RTF values > 64K) */ |
| #define RT6_REJECT_MASK ~(RTF_UP | RTF_GATEWAY | RTF_HOST | \ |
| RTF_REINSTATE | RTF_DYNAMIC | RTF_MODIFIED | \ |
| RTF_DEFAULT | RTF_ADDRCONF | RTF_CACHE) |
| |
| #define DEFAULT_LOOKUPS_DELAY_MS (5) |
| #define DEFAULT_LOOKUPS_RETRIES (10) |
| |
| #define NETIF_INVALID(x) (!(x) || !netif_device_present(x) || \ |
| !netif_running(x) || !netif_carrier_ok(x)) |
| |
| static inline struct net_device * |
| get_netdev_from_br(struct net_device *br, struct nf_conntrack_tuple *tuple); |
| |
| static bool fp_learner_wq = FP_LEARNER_WQ_DEFAULT; |
| |
| struct policy_entry { |
| struct list_head list; |
| unsigned int port; |
| }; |
| |
| struct fp_learner { |
| spinlock_t lock; |
| struct list_head work_items_list; |
| struct list_head policy_list; |
| struct workqueue_struct *wq; |
| struct work_struct update_work; |
| struct socket *rt_nl_sock; |
| struct socket *nf_nl_sock; |
| struct notifier_block netdev_notifier; |
| struct notifier_block netevent_notifier; |
| struct notifier_block inet6addr_notifier; |
| |
| unsigned int lookups_retries; |
| unsigned int lookups_delay; |
| unsigned int fp_rmmoding; |
| }; |
| |
| struct learner_work { |
| struct list_head list; |
| struct fp_learner *priv; |
| struct delayed_work work; |
| /* add new connection data*/ |
| struct nf_conn *ct; |
| }; |
| |
| struct nf_conn * |
| __get_conntrack_from_nlmsg(struct sk_buff *skb, struct nlmsghdr *nlh); |
| struct nf_conntrack_expect * |
| __get_expect_from_nlmsg(struct sk_buff *skb, struct nlmsghdr *nlh); |
| |
| void learner_nc_dump_conntrack_tuple(char *msg, struct nf_conn *ct) |
| { |
| struct nf_conntrack_tuple *orig_tuple = |
| &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple; |
| |
| struct nf_conntrack_tuple *reply_tuple = |
| &ct->tuplehash[IP_CT_DIR_REPLY].tuple; |
| |
| char buf[MAX_DEBUG_PRINT_SIZE]; |
| int len = 0; |
| |
| if (msg) |
| len = sprintf(buf, "%s", msg); |
| |
| len += sprintf(buf + len, "tuple orig:\n"); |
| len += fp_dump_tuple(buf + len, orig_tuple); |
| len += sprintf(buf + len, "\ntuple reply:\n"); |
| len += fp_dump_tuple(buf + len, reply_tuple); |
| |
| pr_err("%s\n", buf); |
| } |
| |
| static inline bool policy_check_port(u_int8_t protocol, __be16 port) |
| { |
| if (protocol == IPPROTO_UDP) { |
| switch (ntohs(port)) { |
| case 53: /* DNS */ |
| case 67: /* bootps */ |
| case 68: /* bootpc */ |
| case 69: /* Trivial File Transfer Protocol (TFTP) */ |
| case 135: /* DHCP server, DNS server and WINS. Also used by DCOM */ |
| case 137: /* NetBIOS NetBIOS Name Service */ |
| case 138: /* NetBIOS NetBIOS Datagram Service */ |
| case 139: /* NetBIOS NetBIOS Session Service */ |
| case 161: /* SNMP */ |
| case 162: /* SNMPTRAP */ |
| case 199: /* SMUX, SNMP Unix Multiplexer */ |
| case 517: /* Talk */ |
| case 518: /* NTalk */ |
| case 546: /* DHCPv6 client*/ |
| case 547: /* DHCPv6 server*/ |
| case 953: /* Domain Name System (DNS) RNDC Service */ |
| case 1719: /* H.323 Registration and alternate communication */ |
| case 1723: /* Microsoft Point-to-Point Tunneling Protocol (PPTP) */ |
| case 5060: /* Session Initiation Protocol (SIP) */ |
| case 5353: /* Multicast DNS (mDNS) */ |
| case 6566: /* SANE (Scanner Access Now Easy) */ |
| case 20480: /* emwavemsg (emWave Message Service) */ |
| return false; |
| } |
| } else { /* TCP */ |
| switch (ntohs(port)) { |
| case 21: /* FTP control (command) */ |
| case 53: /* DNS */ |
| case 135: /* DHCP server, DNS server and WINS. Also used by DCOM */ |
| case 137: /* NetBIOS NetBIOS Name Service */ |
| case 138: /* NetBIOS NetBIOS Datagram Service */ |
| case 139: /* NetBIOS NetBIOS Session Service */ |
| case 162: /* SNMPTRAP */ |
| case 199: /* SMUX, SNMP Unix Multiplexer */ |
| case 546: /* DHCPv6 client*/ |
| case 547: /* DHCPv6 server*/ |
| case 953: /* Domain Name System (DNS) RNDC Service */ |
| case 1720: /* H.323 Call signalling */ |
| case 1723: /* Microsoft Point-to-Point Tunneling Protocol (PPTP) */ |
| case 5060: /* Session Initiation Protocol (SIP) */ |
| case 6566: /* SANE (Scanner Access Now Easy) */ |
| case 6667: /* Internet Relay Chat (IRC) */ |
| case 20480: /* emwavemsg (emWave Message Service) */ |
| return false; |
| } |
| } |
| |
| return true; |
| } |
| |
| |
| |
| static bool learner_policy_check(struct fp_learner *priv, struct nf_conn *ct) |
| { |
| const struct nf_conntrack_l4proto *l4proto; |
| struct policy_entry *itr; |
| struct nf_conntrack_tuple *orig_tuple; |
| |
| orig_tuple = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple; |
| |
| l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct)); |
| NF_CT_ASSERT(l4proto); |
| |
| if (!l4proto->l4proto) |
| goto fail; |
| |
| /* check protocol is UDP/TCP */ |
| if (l4proto->l4proto != IPPROTO_UDP && |
| l4proto->l4proto != IPPROTO_TCP) |
| goto fail; |
| |
| if (!policy_check_port(l4proto->l4proto, orig_tuple->dst.u.all)) |
| goto fail; |
| |
| if (!policy_check_port(l4proto->l4proto, orig_tuple->src.u.all)) |
| goto fail; |
| |
| /* Check dynamic policy */ |
| spin_lock_bh(&priv->lock); |
| list_for_each_entry(itr, &priv->policy_list, list) |
| if (itr && ((itr->port == ntohs(orig_tuple->dst.u.all)) || |
| (itr->port == ntohs(orig_tuple->src.u.all)))) { |
| spin_unlock_bh(&priv->lock); |
| goto fail; |
| } |
| spin_unlock_bh(&priv->lock); |
| |
| return true; |
| fail: |
| pr_debug("connection %p failed police check\n", ct); |
| return false; |
| } |
| |
| static inline void flowi_init(struct flowi *fl, int iif, |
| __u8 scope, __u8 proto, |
| __be32 daddr, __be32 saddr, |
| __be16 dport, __be16 sport, |
| __u32 mark) |
| { |
| memset(fl, 0, sizeof(*fl)); |
| |
| #if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 39) |
| fl->flowi_iif = iif; |
| fl->flowi_scope = scope; |
| fl->flowi_proto = proto; |
| fl->u.ip4.daddr = daddr; |
| fl->u.ip4.saddr = saddr; |
| fl->u.ip4.fl4_dport = dport; |
| fl->u.ip4.fl4_sport = sport; |
| #ifdef CONFIG_NF_CONNTRACK_MARK |
| fl->flowi_mark = mark; |
| #endif |
| #else |
| fl->iif = iif; |
| fl->fl4_scope = scope; |
| fl->proto = proto; |
| fl->fl4_dst = daddr; |
| fl->fl4_src = saddr; |
| fl->fl_ip_dport = dport; |
| fl->fl_ip_sport = sport; |
| #endif |
| } |
| |
| static inline bool invert_tuple(struct nf_conntrack_tuple *inverse, |
| struct nf_conntrack_tuple *orig) |
| { |
| return nf_ct_invert_tuple(inverse, orig); |
| } |
| |
| static inline bool ipv6_check_special_addr(const struct in6_addr *addr) |
| { |
| int addr_type = ipv6_addr_type(addr); |
| /* TODO: check if we need to filter other types - such as Link Local */ |
| return ((addr_type & IPV6_ADDR_MULTICAST) || |
| (addr_type & IPV6_ADDR_LOOPBACK) || |
| (addr_type & IPV6_ADDR_ANY)); |
| } |
| |
| static struct net_device *fp_get_route_ipv6(struct nf_conn *ct, |
| struct nf_conntrack_tuple *tuple, |
| unsigned int *route) |
| { |
| struct net_device *dev = NULL; |
| struct flowi6 fl6 = { |
| .flowi6_oif = 0, |
| .daddr = tuple->dst.u3.in6, |
| }; |
| int flags = RT6_LOOKUP_F_IFACE; |
| struct fib6_result res = {}; |
| int ret = 0; |
| |
| if (ipv6_check_special_addr(&tuple->dst.u3.in6) || |
| ipv6_check_special_addr(&tuple->src.u3.in6)) { |
| pr_debug("Filter special address (saddr=%pI6c, daddr=%pI6c)\n", |
| &tuple->src.u3.in6, &tuple->dst.u3.in6); |
| return NULL; |
| } |
| |
| //if (&tuple->src.u3.in6) { |
| memcpy(&fl6.saddr, &tuple->src.u3.in6, sizeof(tuple->src.u3.in6)); |
| flags |= RT6_LOOKUP_F_HAS_SADDR; |
| //} |
| |
| #ifdef CONFIG_NF_CONNTRACK_MARK |
| fl6.flowi6_mark = ct->mark; |
| #endif |
| |
| ret = ip6_route_lookup_fastpath(nf_ct_net(ct), &fl6, &res, flags); |
| if (ret){ |
| pr_debug("rt6_lookup failed\n"); |
| goto out; |
| } |
| |
| /* check if route is usable*/ |
| if (res.fib6_flags & RTF_UP) { |
| if (res.fib6_flags & RT6_REJECT_MASK) { |
| pr_debug("route rejected (rt6i_flags = 0x%08x)\n", res.fib6_flags); |
| goto out; |
| } |
| /* accepted in fastpath */ |
| dev = res.nh->fib_nh_dev; |
| *route = res.fib6_flags; |
| } |
| |
| out: |
| return dev; |
| } |
| |
| static inline bool ipv4_check_special_addr(const __be32 addr) |
| { |
| /* Filter multicast, broadcast, loopback and zero net*/ |
| return (ipv4_is_loopback(addr) || ipv4_is_multicast(addr) || |
| ipv4_is_lbcast(addr) || ipv4_is_zeronet(addr)); |
| } |
| |
| static inline struct net_device *fp_get_dev_by_ipaddr(struct nf_conntrack_tuple *tuple) |
| { |
| struct net *net; |
| struct net_device *dev; |
| struct in_device *in_dev; |
| struct in_ifaddr *ifa; |
| |
| for_each_net(net) { |
| for_each_netdev(net, dev) { |
| in_dev = __in_dev_get_rcu(dev); |
| if (!in_dev) |
| continue; |
| |
| in_dev_for_each_ifa_rcu(ifa, in_dev) { |
| if (tuple->src.u3.ip == ifa->ifa_local) |
| return dev; |
| } |
| } |
| } |
| |
| return NULL; |
| } |
| |
| static struct net_device *fp_get_route_ipv4(struct nf_conn *ct, |
| struct nf_conntrack_tuple *tuple, |
| unsigned int *route) |
| { |
| struct fib_result res; |
| struct flowi flp; |
| struct net_device *dev = NULL; |
| |
| if (ipv4_check_special_addr(tuple->dst.u3.ip) || |
| ipv4_check_special_addr(tuple->src.u3.ip)) { |
| pr_debug("Filter special address (saddr=%pI4, daddr=%pI4)\n", |
| &tuple->src.u3.ip, &tuple->dst.u3.ip); |
| return NULL; |
| } |
| |
| #ifdef CONFIG_NF_CONNTRACK_MARK |
| flowi_init(&flp, 0, 0, tuple->dst.protonum, tuple->dst.u3.ip, |
| tuple->src.u3.ip, tuple->src.u.all, tuple->dst.u.all, ct->mark); |
| #else |
| flowi_init(&flp, 0, 0, tuple->dst.protonum, tuple->dst.u3.ip, |
| tuple->src.u3.ip, tuple->src.u.all, tuple->dst.u.all, 0); |
| #endif |
| |
| rcu_read_lock_bh(); |
| if (rt4_lookup(nf_ct_net(ct), &flp, &res) < 0) { |
| pr_debug("Getting route failed\n"); |
| rcu_read_unlock_bh(); |
| return NULL; |
| } |
| |
| if (res.type == RTN_BROADCAST) { |
| pr_err("Route = RTN_BROADCAST\n"); |
| goto out; |
| } |
| |
| if (res.type == RTN_MULTICAST) { |
| pr_err("Route = RTN_MULTICAST\n"); |
| goto out; |
| } |
| |
| if (res.type == RTN_LOCAL) { |
| pr_debug("Route = RTN_LOCAL\n"); |
| goto out; |
| } |
| |
| *route = res.type; |
| dev = res.fi->fib_nh->fib_nh_dev; |
| |
| if (NF_CT_NAT(ct)) |
| dev = fp_get_dev_by_ipaddr(tuple) ? fp_get_dev_by_ipaddr(tuple) : dev; |
| out: |
| ip4_rt_put(&res); |
| rcu_read_unlock_bh(); |
| return dev; |
| } |
| |
| static struct fp_net_device *fp_get_route(struct nf_conn *ct, |
| struct nf_conntrack_tuple *tuple, |
| u32 *route, int retries, int delay) |
| { |
| struct fp_net_device *fdev; |
| struct net_device *dev, *br = NULL; |
| |
| dev = (tuple->src.l3num == AF_INET6) ? |
| fp_get_route_ipv6(ct, tuple, route) : |
| fp_get_route_ipv4(ct, tuple, route); |
| if (!dev) |
| return NULL; |
| |
| if (dev->priv_flags & IFF_EBRIDGE) { |
| br = dev; |
| do { |
| dev = get_netdev_from_br(br, tuple); |
| if (dev) |
| break; |
| if (delay) |
| msleep(delay); |
| } while (retries--); |
| |
| if (!dev) { |
| pr_debug("Unable to get net device from bridge IP\n"); |
| return NULL; |
| } |
| } |
| |
| if (dev->reg_state != NETREG_REGISTERED) { |
| pr_debug("device %s not registred (reg_state=%d)\n", dev->name, |
| dev->reg_state); |
| return NULL; |
| } |
| |
| if (unlikely(NETIF_INVALID(dev)) || !(dev->flags & IFF_UP)) { |
| pr_debug("dev (%s) state invalid (state: %lu) or is not up (flags: 0x%x)\n", dev->name, dev->state, dev->flags); |
| return NULL; |
| } |
| |
| fdev = fpdev_get_if(dev); |
| if (!fdev) { |
| pr_err("no fastpath device for %s\n", dev->name); |
| return NULL; |
| } |
| |
| fdev->br = br; |
| return fdev; |
| } |
| |
| static inline int ipv4_gw_addr(struct nf_conn *ct, struct net_device *dev, |
| __be32 saddr, __be32 daddr, __be32 *gw) |
| { |
| struct fib_result res; |
| int ret = 0; |
| struct flowi flp; |
| |
| #ifdef CONFIG_NF_CONNTRACK_MARK |
| flowi_init(&flp, dev->ifindex, RT_SCOPE_UNIVERSE, 0, daddr, saddr, 0, 0, ct->mark); |
| #else |
| flowi_init(&flp, dev->ifindex, RT_SCOPE_UNIVERSE, 0, daddr, saddr, 0, 0, 0); |
| #endif |
| rcu_read_lock_bh(); |
| ret = rt4_lookup(dev_net(dev), &flp, &res); |
| if (ret != 0) { |
| pr_err("rt4_lookup failed, ret = %d\n", ret); |
| rcu_read_unlock_bh(); |
| return ret; |
| } |
| |
| if (res.type == RTN_BROADCAST || res.type == RTN_MULTICAST || |
| res.type == RTN_LOCAL) { |
| pr_debug("gw not found - res.type = %d\n", res.type); |
| ret = -EFAULT; |
| } else { |
| *gw = res.fi->fib_nh->fib_nh_gw4; |
| pr_debug("gw found (%pI4)\n", gw); |
| } |
| |
| ip4_rt_put(&res); |
| rcu_read_unlock_bh(); |
| return ret; |
| } |
| |
| static inline int ipv6_gw_addr(struct nf_conn *ct, struct net_device *dev, struct in6_addr *saddr, |
| struct in6_addr *daddr, struct in6_addr *gw) |
| { |
| int ret = 0; |
| struct flowi6 fl6 = { |
| .flowi6_oif = 0, |
| .daddr = *daddr, |
| }; |
| int flags = RT6_LOOKUP_F_IFACE; |
| struct fib6_result res = {}; |
| |
| if (saddr) { |
| memcpy(&fl6.saddr, saddr, sizeof(*saddr)); |
| flags |= RT6_LOOKUP_F_HAS_SADDR; |
| } |
| |
| #ifdef CONFIG_NF_CONNTRACK_MARK |
| fl6.flowi6_mark = ct->mark; |
| #endif |
| |
| ret = ip6_route_lookup_fastpath(dev_net(dev), &fl6, &res, flags); |
| if (ret) { |
| pr_err("rt6_lookup failed\n"); |
| ret = -ENETUNREACH; |
| goto out; |
| } |
| |
| /* check if route is usable*/ |
| if (res.fib6_flags & RTF_UP) { |
| if (res.nh->fib_nh_gw_family) |
| *gw = res.nh->fib_nh_gw6; |
| } else { |
| pr_debug("gw found but route is not up\n"); |
| ret = -EFAULT; |
| } |
| |
| out: |
| return ret; |
| } |
| |
| /* copied from br_fdb.c */ |
| static inline |
| struct net_bridge_fdb_entry *fp_br_fdb_find(struct hlist_head *head, |
| const unsigned char *addr) |
| { |
| struct net_bridge_fdb_entry *fdb; |
| |
| hlist_for_each_entry_rcu(fdb, head, fdb_node) { |
| if (ether_addr_equal(fdb->key.addr.addr, addr)) |
| return fdb; |
| } |
| |
| return NULL; |
| } |
| |
| static inline |
| struct net_device *fp_br_get_netdev_by_mac(struct net_bridge *br, |
| const unsigned char *mac) |
| { |
| unsigned int i; |
| struct net_bridge_fdb_entry *fdb; |
| |
| BUG_ON(!br); |
| |
| rcu_read_lock_bh(); |
| for (i = 0; i < BR_HASH_SIZE; i++) { |
| fdb = fp_br_fdb_find(&br->fdb_list, mac); |
| if (fdb) { |
| pr_debug("br: %s fdb[%u]: %pIM , port:%s\n", |
| br->dev->name, i, fdb->key.addr.addr, |
| fdb->dst->dev->name); |
| rcu_read_unlock_bh(); |
| return fdb->dst->dev; |
| } |
| } |
| rcu_read_unlock_bh(); |
| pr_debug("no match found in fdb (%pM)\n", mac); |
| |
| return NULL; |
| } |
| |
| static inline |
| struct net_device *get_netdev_from_br(struct net_device *br, |
| struct nf_conntrack_tuple *tuple) |
| { |
| struct neighbour *neigh; |
| struct neigh_table *tbl; |
| struct net_device *dev = NULL; |
| |
| BUG_ON(!tuple); |
| |
| tbl = (tuple->src.l3num == AF_INET6) ? &nd_tbl : &arp_tbl; |
| |
| neigh = neigh_lookup(tbl, tuple->dst.u3.all, br); |
| if (neigh) { |
| dev = fp_br_get_netdev_by_mac(netdev_priv(br), neigh->ha); |
| neigh_release(neigh); |
| } |
| |
| return dev; |
| } |
| |
| static int fp_hh_init(struct nf_conn *ct, struct nf_conntrack_tuple *t, |
| struct fp_net_device *dst, struct hh_cache *hh) |
| { |
| struct net_device *dev = dst->br ? dst->br : dst->dev; |
| __be16 prot; |
| struct neighbour *n; |
| const struct header_ops *header_ops; |
| |
| if (is_vlan_dev(dev)) |
| header_ops = vlan_dev_real_dev(dev)->header_ops; |
| else |
| header_ops = dev->header_ops; |
| |
| memset(hh, 0, sizeof(*hh)); |
| |
| if (!header_ops) { |
| pr_debug("device %s has no header ops\n", dev->name); |
| return 0; /* device does not have L2 header*/ |
| } |
| |
| if (!header_ops->cache || !header_ops->cache_update) { |
| pr_debug("device %s has no header cache ops\n", dev->name); |
| return -ENOTSUPP; |
| } |
| |
| if (t->src.l3num == AF_INET) { |
| __be32 gw; |
| prot = htons(ETH_P_IP); |
| |
| n = __ipv4_neigh_lookup(dev, t->dst.u3.ip); |
| if (!n) { |
| if (ipv4_gw_addr(ct, dev, t->src.u3.ip, t->dst.u3.ip, &gw)) |
| goto not_found; |
| n = __ipv4_neigh_lookup(dev, gw); |
| if (!n) |
| goto not_found; |
| } |
| } else if (t->src.l3num == AF_INET6) { |
| struct in6_addr gw6; |
| prot = htons(ETH_P_IPV6); |
| |
| n = __ipv6_neigh_lookup(dev, &t->dst.u3.in6); |
| if (!n) { |
| if (ipv6_gw_addr(ct, dev, &t->src.u3.in6, &t->dst.u3.in6, &gw6)) |
| goto not_found; |
| n = __ipv6_neigh_lookup(dev, &t->dst.u3.in6); |
| if (!n) |
| goto not_found; |
| } |
| } else { |
| BUG(); |
| } |
| |
| if (n->nud_state & NUD_VALID) { |
| int err = header_ops->cache(n, hh, prot); |
| neigh_release(n); |
| pr_debug("device %s hh_cache initialized: hh_len=%d, hh_data=%pM\n", |
| dev->name, hh->hh_len, hh->hh_data); |
| return err; |
| } |
| |
| pr_debug("neighbour state invalid (%02x)\n", n->nud_state); |
| neigh_release(n); |
| not_found: |
| /* we get here in 2 cases, both are NOT considered as error: |
| * 1. Neighbour lookup failed - we will be notified when the neighbour |
| * will be finally created |
| * 2. Neighbour state not valid - we will be notified when the neighbour |
| * state changes |
| * Both are handled by netdev_event - where the entry's hh_cache will be |
| * updated. Untill this happens, all packets matching this entry will be |
| * classified as slow by the fp_classifier. |
| */ |
| pr_debug("No neighbour found or neighbour state invalid\n"); |
| return 0; |
| } |
| |
| static struct fpdb_entry *connection_to_entry(struct fp_learner *priv, |
| struct nf_conn *ct, |
| enum ip_conntrack_dir dir, |
| gfp_t flags) |
| { |
| struct fp_net_device *dst = NULL, *src = NULL; |
| struct nf_conntrack_tuple *orig_tuple, *reply_tuple; |
| struct nf_conntrack_tuple orig_tuple_inverse, reply_tuple_inverse; |
| struct fpdb_entry *entry; |
| struct hh_cache hh; |
| unsigned int in_route_type, out_route_type; |
| int retries = flags != GFP_ATOMIC ? priv->lookups_retries : 0; |
| int delay = flags != GFP_ATOMIC ? priv->lookups_delay : 0; |
| |
| if (unlikely(priv->fp_rmmoding)) |
| goto failed; |
| |
| /* For reply connections -> switch tuples */ |
| if (dir == IP_CT_DIR_REPLY) { |
| orig_tuple = &ct->tuplehash[IP_CT_DIR_REPLY].tuple; |
| reply_tuple = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple; |
| } else { |
| orig_tuple = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple; |
| reply_tuple = &ct->tuplehash[IP_CT_DIR_REPLY].tuple; |
| } |
| |
| if (NF_CT_NAT(ct)) { |
| if (!invert_tuple(&orig_tuple_inverse, orig_tuple)) { |
| pr_err("Inverting tuple failed\n"); |
| goto failed; |
| } |
| |
| if (!invert_tuple(&reply_tuple_inverse, reply_tuple)) { |
| pr_err("Inverting tuple failed\n"); |
| goto failed; |
| } |
| |
| orig_tuple = &reply_tuple_inverse; |
| reply_tuple = &orig_tuple_inverse; |
| pr_debug( "NAT connection was detected\n"); |
| } |
| |
| /* Check destination route */ |
| dst = fp_get_route(ct, orig_tuple, &in_route_type, retries, delay); |
| if (!dst) { |
| pr_debug("Connection routing failed\n"); |
| goto failed; |
| } |
| |
| /* Check source route */ |
| src = fp_get_route(ct, reply_tuple, &out_route_type, retries, delay); |
| if (!src) { |
| pr_debug("Connection routing failed (local)\n"); |
| goto failed; |
| } |
| |
| if (fp_hh_init(ct, orig_tuple, dst, &hh)) { |
| pr_debug("fp_hh_init failed \n"); |
| goto failed; |
| } |
| |
| entry = fpdb_alloc(flags); |
| if (!entry) { |
| pr_debug("Allocating entry failed\n"); |
| goto failed; |
| } |
| |
| /* Restore the original tuples */ |
| if (dir == IP_CT_DIR_REPLY) { |
| orig_tuple = &ct->tuplehash[IP_CT_DIR_REPLY].tuple; |
| reply_tuple = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple; |
| } else { |
| orig_tuple = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple; |
| reply_tuple = &ct->tuplehash[IP_CT_DIR_REPLY].tuple; |
| } |
| |
| /* |
| * if interface is going down, and we are updating an entry refering |
| * to this interface, we might accidently route the connection to its |
| * source device. Block this entry until it is updated again. |
| */ |
| if (src->dev == dst->dev) { |
| pr_debug("Connection created with src == dst for (%s)\n", |
| src->dev->name); |
| entry->block = 1; |
| } |
| |
| /* Fill in entry */ |
| entry->dir = dir; |
| entry->in_tuple = *orig_tuple; |
| entry->out_tuple = *reply_tuple; |
| entry->ct = ct; |
| entry->out_dev = dst; |
| entry->in_dev = src; |
| entry->hit_counter = 0; |
| entry->debug.in_route_type = in_route_type; |
| entry->debug.out_route_type = out_route_type; |
| entry->hh = hh; |
| |
| /* Succced */ |
| pr_debug("connection added (ct=%p, dir=%d)\n", ct, dir); |
| FP_DEBUG_DUMP_CONTRACK(NULL, ct); |
| return entry; |
| |
| failed: |
| /* Failed */ |
| fpdev_put(src); |
| fpdev_put(dst); |
| pr_debug("connection refused (ct=%p, dir=%d)\n", ct, dir); |
| FP_DEBUG_DUMP_CONTRACK(NULL, ct); |
| return NULL; |
| } |
| |
| |
| static inline int __add_new_connection(struct fp_learner *priv, |
| struct nf_conn *ct, gfp_t flags) |
| { |
| struct nf_conn_fastpath *fastpath = nfct_fastpath(ct); |
| struct fpdb_entry *e; |
| |
| rcu_read_lock_bh(); |
| /* original fastpath connection */ |
| if (!fastpath) { |
| e = connection_to_entry(priv, ct, IP_CT_DIR_ORIGINAL, flags); |
| if (!e) { |
| rcu_read_unlock_bh(); |
| return -EINVAL; |
| } |
| |
| set_bit(IPS_FASTPATH_BIT, &ct->status); |
| fastpath = nf_ct_ext_add(ct, NF_CT_EXT_FASTPATH, flags); |
| BUG_ON(!fastpath); |
| |
| fastpath->fpd_el[IP_CT_DIR_ORIGINAL] = e; |
| fastpath->fpd_el[IP_CT_DIR_REPLY] = NULL; |
| fpdb_add(e); |
| goto del_entry; |
| } |
| |
| /* reply fastpath connection */ |
| BUG_ON(!test_bit(IPS_FASTPATH_BIT, &ct->status)); |
| if (fastpath->fpd_el[IP_CT_DIR_REPLY] == NULL && |
| test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { |
| e = connection_to_entry(priv, ct, IP_CT_DIR_REPLY, flags); |
| if (!e) { |
| rcu_read_unlock_bh(); |
| return -EINVAL; |
| } |
| |
| fastpath->fpd_el[IP_CT_DIR_REPLY] = e; |
| fpdb_add(e); |
| goto del_entry; |
| } |
| rcu_read_unlock_bh(); |
| return 0; |
| |
| del_entry: |
| if (unlikely((NETIF_INVALID(e->in_dev->dev)) || |
| !(e->in_dev->dev->flags & IFF_UP) || priv->fp_rmmoding)) { |
| pr_err("in_dev (%s) state invalid or is rmmoding, del entry!\n", e->in_dev->dev->name); |
| fpdb_del_by_dev(e->in_dev->dev); |
| } |
| |
| if (unlikely((NETIF_INVALID(e->out_dev->dev)) || |
| !(e->out_dev->dev->flags & IFF_UP) || priv->fp_rmmoding)) { |
| pr_err("out_dev (%s) state invalid or is rmmoding, del entry!\n", e->out_dev->dev->name); |
| fpdb_del_by_dev(e->out_dev->dev); |
| } |
| rcu_read_unlock_bh(); |
| return 0; |
| } |
| |
| static void new_connection_work(struct work_struct *w) |
| { |
| struct learner_work *work; |
| |
| work = container_of(w, struct learner_work, work.work); |
| BUG_ON(!work); |
| |
| __add_new_connection(work->priv, work->ct, GFP_KERNEL); |
| |
| /* release work */ |
| spin_lock_bh(&work->priv->lock); |
| list_del(&work->list); |
| spin_unlock_bh(&work->priv->lock); |
| kfree(work); |
| } |
| |
| static inline int add_new_connection_work(struct fp_learner *priv, |
| struct nf_conn *ct) |
| { |
| struct learner_work *work; |
| |
| if (!learner_policy_check(priv, ct)) |
| return -EINVAL; |
| |
| work = kzalloc(sizeof(*work), GFP_ATOMIC); |
| if (!work) |
| return -ENOMEM; |
| |
| work->ct = ct; |
| work->priv = priv; |
| INIT_LIST_HEAD(&work->list); |
| INIT_DELAYED_WORK(&work->work, new_connection_work); |
| |
| spin_lock_bh(&priv->lock); |
| list_add_tail(&work->list, &priv->work_items_list); |
| spin_unlock_bh(&priv->lock); |
| |
| queue_delayed_work(priv->wq, &work->work, 0); |
| |
| return 0; |
| } |
| |
| |
| static inline int add_new_connection_noblock(struct fp_learner *priv, |
| struct nf_conn *ct) |
| { |
| if (!learner_policy_check(priv, ct)) |
| return -EINVAL; |
| return __add_new_connection(priv, ct, GFP_ATOMIC); |
| } |
| |
| static inline int add_new_connection(struct fp_learner *priv, |
| struct nf_conn *ct) |
| { |
| if (fp_learner_wq) |
| return add_new_connection_work(priv, ct); |
| else |
| return add_new_connection_noblock(priv, ct); |
| } |
| |
| /* check if this connection is waiting in our workqueue |
| * and cancle it if it is. |
| */ |
| static inline int |
| new_connection_cancle(struct fp_learner *priv, struct nf_conn *ct) |
| { |
| struct learner_work *work; |
| |
| if (!fp_learner_wq) |
| return 0; |
| |
| spin_lock_bh(&priv->lock); |
| list_for_each_entry(work, &priv->work_items_list, list) { |
| if (work->ct == ct) { |
| if (cancel_delayed_work(&work->work)) { |
| pr_debug("cancle connection add %p\n", ct); |
| list_del(&work->list); |
| kfree(work); |
| } |
| break; |
| } |
| } |
| spin_unlock_bh(&priv->lock); |
| |
| return 0; |
| } |
| |
| static int learner_ct_event(struct fp_learner *priv, struct nf_conn *ct, |
| unsigned int type, unsigned int flags) |
| { |
| if (type == IPCTNL_MSG_CT_DELETE) { |
| pr_debug("delete connection (%p)\n", ct); |
| return new_connection_cancle(priv, ct); |
| } else if (type == IPCTNL_MSG_CT_NEW) { |
| pr_debug("new connection (%p)\n", ct); |
| return add_new_connection(priv, ct); |
| } |
| |
| pr_debug("Unhandled type=%u\n", type); |
| FP_DEBUG_DUMP_CONTRACK(NULL, ct); |
| |
| return -ENOTSUPP; |
| } |
| |
| static int fpdev_del_gb6(struct net_device *dev) |
| { |
| struct fp_net_device *fpdev; |
| |
| fpdev = fpdev_get_if(dev); |
| if (unlikely(!fpdev)) |
| return 0; |
| |
| memset(&fpdev->ll6addr, 0, sizeof(struct in6_addr)); |
| memset(&fpdev->gb6addr, 0, sizeof(struct in6_addr)); |
| fpdev->prefixlen = 0; |
| fpdev->mtu = 0; |
| fpdev_clear_ll6(fpdev); |
| fpdev_clear_gb6(fpdev); |
| fpdev_clear_mtu(fpdev); |
| |
| fpdev_put(fpdev); |
| |
| return 0; |
| } |
| |
| /** |
| * handle netdevice events. |
| * |
| * NETDEV_REGISTER |
| * new net_device is registered. A fastpath device is created |
| * and associated to it. |
| * |
| * NETDEV_UNREGISTER |
| * net_device unregistered, delete the associated fastpath device. In |
| * addition, remove all conntracks related to this device - this will |
| * cause all the related fastpath database entries to be deleted thus allowing |
| * the device to be safely removed. |
| * |
| * @note We can safely ignore NETDEV_UP / NETDEV_DOWN since it is |
| * checked in the classifier anyway. Regarding other events - |
| * will be added in the future if needed. |
| * @param dev |
| * @param event |
| * |
| * @return NOTIFY_DONE |
| */ |
| static int |
| __learner_netdev_event(struct net_device *dev, unsigned long event) |
| { |
| switch (event) { |
| case NETDEV_REGISTER: |
| pr_debug("received netdev (%s) register, event %lu, state: 0x%lx, flags: 0x%x, invalid: %d\n", |
| dev->name, event, dev->state, dev->flags, NETIF_INVALID(dev)); |
| fpdev_add_if(dev); |
| break; |
| case NETDEV_UNREGISTER: |
| printk(KERN_DEBUG "received netdev (%s) unregister, event %lu, state: 0x%lx, flags: 0x%x, invalid: %d\n", |
| dev->name, event, dev->state, dev->flags, NETIF_INVALID(dev)); |
| fpdb_del_by_dev(dev); |
| fpdb_iterate(&fpdb_del_block_entry_by_dev, (void *)dev); |
| fpdev_del_if(dev); |
| break; |
| case NETDEV_DOWN: |
| fpdev_del_gb6(dev); |
| break; |
| default: |
| pr_debug("ignoring netdev %s event %lu, state: 0x%lx, flags: 0x%x, invalid: %d\n", |
| dev->name, event, dev->state, dev->flags, NETIF_INVALID(dev)); |
| } |
| |
| return NOTIFY_DONE; |
| } |
| |
| /* main dispatcher for netdev events - bridge and loopback ignored */ |
| static int learner_netdev_event(struct notifier_block *nb, |
| unsigned long event, void *ptr) |
| { |
| struct net_device *dev; |
| |
| #if LINUX_VERSION_CODE < KERNEL_VERSION(3, 11, 0) |
| dev = ptr; |
| #else |
| dev = ((struct netdev_notifier_info*)ptr)->dev; |
| #endif |
| |
| if ((dev->priv_flags & IFF_EBRIDGE) || (dev->flags & IFF_LOOPBACK)) |
| return NOTIFY_DONE; |
| |
| return __learner_netdev_event(dev, event); |
| } |
| |
| static void learner_netdev_cleanup(struct notifier_block *nb) |
| { |
| #if LINUX_VERSION_CODE < KERNEL_VERSION(3, 4, 0) |
| struct net_device *dev; |
| struct net *net; |
| |
| rtnl_lock(); |
| for_each_net(net) { |
| for_each_netdev(net, dev) { |
| if (dev->flags & IFF_UP) { |
| nb->notifier_call(nb, NETDEV_GOING_DOWN, dev); |
| nb->notifier_call(nb, NETDEV_DOWN, dev); |
| } |
| nb->notifier_call(nb, NETDEV_UNREGISTER, dev); |
| nb->notifier_call(nb, NETDEV_UNREGISTER_BATCH, dev); |
| } |
| } |
| rtnl_unlock(); |
| #endif |
| } |
| |
| static int fp_inet6addr_event(struct notifier_block *nb, |
| unsigned long event, void *ptr) |
| { |
| struct inet6_ifaddr *ifa = (struct inet6_ifaddr *)ptr; |
| struct net_device *dev = ifa->idev->dev; |
| struct fp_net_device *fpdev; |
| int addr_type; |
| |
| if (event != NETDEV_UP) |
| return NOTIFY_DONE; |
| |
| addr_type = ipv6_addr_type(&ifa->addr); |
| if (!(addr_type & IPV6_ADDR_LINKLOCAL)) |
| return NOTIFY_DONE; |
| |
| fpdev = fpdev_get_if(dev); |
| if (unlikely(!fpdev)) |
| return NOTIFY_DONE; |
| |
| if (!fpdev_is_ll6_set(fpdev)) { |
| memcpy(&fpdev->ll6addr, &ifa->addr, sizeof(ifa->addr)); |
| fpdev_set_ll6(fpdev); |
| } |
| |
| fpdev_put(fpdev); |
| |
| return NOTIFY_DONE; |
| } |
| |
| static int update_entry(struct fpdb_entry *e, void *data) |
| { |
| struct fpdb_entry *ne = NULL; |
| struct fp_learner *fpl = (struct fp_learner *)data; |
| struct nf_conn_fastpath *fastpath; |
| |
| spin_lock_bh(&e->lock); |
| fpdb_lock_bh(); |
| |
| /* Exit if CT destroied, in case fpdb get a wrong ct info */ |
| if (e->state == ENTRY_DYING) |
| goto done; |
| |
| fastpath = nfct_fastpath(e->ct); |
| if (unlikely(!fastpath)) |
| goto done; |
| |
| ne = connection_to_entry(fpl, e->ct, e->dir, GFP_ATOMIC); |
| if (!ne) { |
| /* The connection may become local but we do not want |
| to remove it from STACK so just block it */ |
| e->block = 1; |
| } else { |
| if (ne->out_dev == e->out_dev && |
| ne->in_dev == e->in_dev && |
| nf_ct_tuple_equal(&ne->in_tuple, &e->in_tuple) && |
| nf_ct_tuple_equal(&ne->out_tuple, &e->out_tuple) && |
| !memcmp(&ne->hh, &e->hh, sizeof(struct hh_cache))) { |
| pr_debug("new fp entry equal old,no update\n"); |
| fpdb_free(ne); |
| goto done; |
| } |
| |
| /* if the old connection is blocked keep it blocked */ |
| /* if ne->block is 1 and e->block is 0, there will be issue --yhuang 20160617*/ |
| if (ne->block != 1) |
| ne->block = e->block; |
| |
| if (ne->dir == IP_CT_DIR_REPLY) |
| fastpath->fpd_el[IP_CT_DIR_REPLY] = ne; |
| else |
| fastpath->fpd_el[IP_CT_DIR_ORIGINAL] = ne; |
| |
| fpdb_replace(e, ne); |
| } |
| |
| done: |
| fpdb_unlock_bh(); |
| spin_unlock_bh(&e->lock); |
| |
| if (ne) { |
| if (unlikely((NETIF_INVALID(ne->in_dev->dev)) || |
| !(ne->in_dev->dev->flags & IFF_UP) || fpl->fp_rmmoding)) { |
| pr_err("in_dev (%s) state invalid or rmmoding, del!\n", |
| ne->in_dev->dev->name); |
| fpdb_del_by_dev(ne->in_dev->dev); |
| } |
| |
| if (unlikely((NETIF_INVALID(ne->out_dev->dev)) || |
| !(ne->out_dev->dev->flags & IFF_UP) || fpl->fp_rmmoding)) { |
| pr_err("out_dev (%s) state invalid or rmmoding, del!\n", |
| ne->out_dev->dev->name); |
| fpdb_del_by_dev(ne->out_dev->dev); |
| } |
| } |
| return 0; |
| } |
| |
| static int block_entry(struct fpdb_entry *e, void *ptr) |
| { |
| spin_lock_bh(&e->lock); |
| e->block = 1; |
| spin_unlock_bh(&e->lock); |
| |
| return 0; |
| } |
| |
| static void learner_ct_update_work(struct work_struct *work) |
| { |
| struct fp_learner *fpl = container_of(work, |
| struct fp_learner, update_work); |
| |
| fpdb_iterate(&update_entry, (void *)fpl); |
| } |
| |
| void __learner_ct_update_all(struct fp_learner *fpl) |
| { |
| schedule_work(&fpl->update_work); |
| } |
| |
| static int learner_netevent(struct notifier_block *nb, unsigned long event, void *ctx) |
| { |
| struct fp_learner *fpl = container_of(nb, struct fp_learner, netevent_notifier); |
| |
| BUG_ON(!fpl); |
| |
| if (event == NETEVENT_NEIGH_UPDATE) { |
| struct neighbour *n = ctx; |
| |
| pr_debug("neighbor update received (state=%d, dev=%s)\n", |
| n->nud_state, n->dev->name); |
| __learner_ct_update_all(fpl); |
| } else if (event == NETEVENT_REDIRECT) { |
| |
| __learner_ct_update_all(fpl); |
| pr_debug("neighbor redirect received\n"); |
| } else { |
| pr_debug("mfp received netevent %lu, which no need to update ct\n", event); |
| WARN_ON(1); |
| } |
| |
| return 0; |
| } |
| |
| static void learner_rtnetlink_rcv(struct nlmsghdr *nlh, void *ptr) |
| { |
| struct fp_learner *fpl = (struct fp_learner *)ptr; |
| switch (nlh->nlmsg_type) { |
| case RTM_NEWROUTE: |
| case RTM_DELROUTE: |
| pr_debug( "%s\n", |
| nlh->nlmsg_type == RTM_NEWROUTE ? "RTM_NEWROUTE" : |
| "RTM_DELROUTE"); |
| __learner_ct_update_all(fpl); |
| break; |
| case RTM_NEWTFILTER: |
| pr_debug( "RTM_NEWTFILTER\n"); |
| /* TODO: check if we need update in this case*/ |
| break; |
| case RTM_DELTFILTER: |
| pr_debug( "RTM_DELTFILTER\n"); |
| /* TODO: check if we need update in this case*/ |
| break; |
| case RTM_GETTFILTER: |
| pr_debug( "RTM_GETTFILTER\n"); |
| break; |
| } |
| pr_debug("handle routing netlink message, type=%d\n", nlh->nlmsg_type); |
| //TODO - add support for update_all_connections |
| } |
| |
| static void learner_nfnetlink_rcv(struct sk_buff *skb, |
| struct nlmsghdr *nlh, void *ptr) |
| { |
| struct fp_learner *priv = (struct fp_learner *)ptr; |
| unsigned int type = NFNL_MSG_TYPE(nlh->nlmsg_type); |
| unsigned int ssid = NFNL_SUBSYS_ID(nlh->nlmsg_type); |
| struct nf_conn *ct; |
| struct nf_conntrack_expect *exp; |
| unsigned int flags = nlh->nlmsg_flags; |
| int ret; |
| |
| if (ssid == NFNL_SUBSYS_CTNETLINK) { |
| ct = __get_conntrack_from_nlmsg(skb, nlh); |
| if (ct == NULL) { |
| pr_debug("can't get nf conn type=%u, ssid=%u\n", type, ssid); |
| return; |
| } |
| pr_debug("found CTNETLINK connection %p, type=%u, ssid=%u\n", ct, type, ssid); |
| } else if (ssid == NFNL_SUBSYS_CTNETLINK_EXP) { |
| exp = __get_expect_from_nlmsg(skb, nlh); |
| if (exp == NULL) { |
| pr_err("can't get expect\n"); |
| return; |
| } |
| ct = exp->master; |
| pr_debug("found CTNETLINK_EXP exp %p, master connection %p, type=%u, ssid=%u\n", exp, ct, type, ssid); |
| } else { |
| pr_err("unexpected ssid (%d)\n", ssid); |
| return; |
| } |
| |
| /* dispatch events */ |
| ret = learner_ct_event(priv, ct, type, flags); |
| if (ret < 0) |
| pr_debug("learner_ct_event failed with error code %d\n" |
| "ct=%p, type=%u, flags=%u\n", ret, ct, type, flags); |
| } |
| |
| /* Receive message from netlink and pass information to relevant function. */ |
| static void learner_nl_data_ready(struct sock *sk) |
| { |
| int ret = 0; |
| int len; |
| struct sk_buff *skb; |
| struct nlmsghdr *nlh; |
| |
| BUG_ON(!sk); |
| pr_debug("got a message (socket protocol=%d)\n", sk->sk_protocol); |
| |
| while ((skb = skb_recv_datagram(sk, 0, 1, &ret)) == NULL) { |
| if (ret == -EAGAIN || ret == -ENOBUFS) { |
| pr_err("recvfrom() error %d\n", -ret); |
| return; |
| } |
| } |
| |
| len = skb->len; |
| for (nlh = (struct nlmsghdr *)skb->data; NLMSG_OK(nlh, len); |
| nlh = NLMSG_NEXT(nlh, len)) { |
| pr_debug("nlmsg_len %u, nlmsg_type %u\n", nlh->nlmsg_len, nlh->nlmsg_type); |
| |
| /* Finish of reading. */ |
| if (nlh->nlmsg_type == NFNL_MSG_TYPE(NLMSG_DONE)) |
| goto out; |
| |
| /* Error handling. */ |
| if (nlh->nlmsg_type == NFNL_MSG_TYPE(NLMSG_ERROR)) { |
| pr_err("nl message error\n"); |
| goto out; |
| } |
| |
| if (sk->sk_protocol == NETLINK_ROUTE) { |
| learner_rtnetlink_rcv(nlh, sk->sk_user_data); |
| } else if (sk->sk_protocol == NETLINK_NETFILTER) { |
| learner_nfnetlink_rcv(skb, nlh, sk->sk_user_data); |
| } else { |
| pr_err("unrecognized sk_protocol (%u)\n", sk->sk_protocol); |
| goto out; |
| } |
| } |
| out: |
| skb_orphan(skb); |
| kfree_skb(skb); |
| return; |
| } |
| |
| static int learner_nl_open(void *priv, struct socket **s, int proto, int groups) |
| { |
| struct socket *sock; |
| struct sockaddr_nl addr; |
| int rc, val = 1; |
| |
| rc = sock_create_kern(&init_net, AF_NETLINK , SOCK_RAW, proto, &sock); |
| if (rc < 0) { |
| pr_err("create err (rc=%d)\n", rc); |
| return rc; |
| } |
| |
| memset((void *)&addr, 0, sizeof(addr)); |
| addr.nl_family = AF_NETLINK; |
| addr.nl_pid = 0; |
| addr.nl_groups = groups; |
| sock->sk->sk_user_data = priv; |
| sock->sk->sk_data_ready = learner_nl_data_ready; |
| sock->sk->sk_allocation = GFP_ATOMIC; |
| |
| rc = kernel_bind(sock, (struct sockaddr *)&addr, sizeof(addr)); |
| if (rc < 0) { |
| pr_err("bind err (rc=%d)\n", rc); |
| goto sock_err; |
| } |
| |
| rc = kernel_setsockopt(sock, SOL_NETLINK, NETLINK_NO_ENOBUFS, (char *)&val, sizeof(val)); |
| if (rc < 0) { |
| pr_err("setsockopt err (rc=%d)", rc); |
| goto sock_err; |
| } |
| |
| pr_debug("netlink socket opened (proto=%u, groups=%u)\n", proto, groups); |
| *s = sock; |
| return 0; |
| |
| sock_err: |
| kernel_sock_shutdown(sock, SHUT_RDWR); |
| sock_release(sock); |
| return rc; |
| } |
| |
| static void learner_nl_close(struct socket *sk) |
| { |
| BUG_ON(!sk); |
| kernel_sock_shutdown(sk, SHUT_RDWR); |
| sock_release(sk); |
| } |
| |
| static unsigned int fp_learner_nf_hook(void *priv, |
| struct sk_buff *skb, |
| const struct nf_hook_state *state) |
| { |
| enum ip_conntrack_info ctinfo; |
| struct nf_conn *ct = nf_ct_get(skb, &ctinfo); |
| struct nf_conn_fastpath *f; |
| struct fpdb_entry *e; |
| struct net_device *el_src, *el_dst; |
| |
| if (!ct) |
| goto out; |
| |
| f = nfct_fastpath(ct); |
| if (!f) |
| goto out; |
| |
| rcu_read_lock_bh(); |
| e = rcu_dereference(f->fpd_el[CTINFO2DIR(ctinfo)]); |
| /* Here we can not clear block simply. We need to check */ |
| /* whether src dev equals to dst dev yhuang 20160622*/ |
| if (unlikely(e && (e->block == true))) { |
| if ((e->state != ENTRY_ALIVE) |
| || (e->dir >= IP_CT_DIR_MAX)) { |
| rcu_read_unlock_bh(); |
| goto out; |
| } |
| |
| /* skb->dev already set to destiniation device */ |
| el_src = e->in_dev->dev; |
| el_dst = e->out_dev->dev; |
| if(unlikely((el_src == NULL) || (el_dst == NULL))) { |
| rcu_read_unlock_bh(); |
| goto out; |
| } |
| if (unlikely(NETIF_INVALID(el_src) || NETIF_INVALID(el_dst))) { |
| rcu_read_unlock_bh(); |
| goto out; |
| } |
| |
| if (likely(el_src != el_dst)) { |
| spin_lock_bh(&e->lock); |
| e->block = 0; |
| spin_unlock_bh(&e->lock); |
| pr_debug("unblock entry:ct=%x dir=%d bucket=%x %x %x\n", |
| (unsigned int)e->ct, e->dir, e->bucket, |
| (unsigned int)(&e->in_tuple), |
| (unsigned int)(&e->out_tuple)); |
| } |
| } |
| rcu_read_unlock_bh(); |
| out: |
| return NF_ACCEPT; |
| } |
| |
| /** learner's netfilter hook */ |
| static struct nf_hook_ops nf_learner_hook_data[] __read_mostly = { |
| { |
| .hook = fp_learner_nf_hook, |
| .pf = NFPROTO_IPV4, |
| .hooknum = NF_INET_POST_ROUTING, |
| .priority = NF_IP_PRI_LAST, |
| }, |
| { |
| .hook = fp_learner_nf_hook, |
| .pf = NFPROTO_IPV6, |
| .hooknum = NF_INET_POST_ROUTING, |
| .priority = NF_IP_PRI_LAST, |
| }, |
| }; |
| |
| void __learner_ct_block_all(struct fastpath_module *m) |
| { |
| fpdb_iterate(&block_entry, m->priv); |
| } |
| |
| static ssize_t |
| learner_ct_update_all(struct fastpath_module *m, const char *buf, size_t count) |
| { |
| __learner_ct_update_all((struct fp_learner *)m->priv); |
| return count; |
| } |
| |
| static ssize_t |
| learner_ct_block_all(struct fastpath_module *m, const char *buf, size_t count) |
| { |
| __learner_ct_block_all(m); |
| return count; |
| } |
| |
| static ssize_t learner_policy_show(struct fastpath_module *m, char *buf) |
| { |
| struct policy_entry *itr; |
| struct fp_learner *priv = m->priv; |
| int len; |
| |
| len = scnprintf(buf, PAGE_SIZE, "dynamic policy restricted ports:\n"); |
| |
| spin_lock_bh(&priv->lock); |
| list_for_each_entry(itr, &priv->policy_list, list) |
| len += scnprintf(buf + len, PAGE_SIZE - len, "%d, ", itr->port); |
| spin_unlock_bh(&priv->lock); |
| |
| len += scnprintf(buf + len, PAGE_SIZE - len, "\n"); |
| |
| return len; |
| } |
| |
| static ssize_t learner_policy_store(struct fastpath_module *m, |
| const char *buf, size_t count) |
| { |
| char op; |
| unsigned int port; |
| struct policy_entry *entry, *tmp; |
| struct fp_learner *priv = m->priv; |
| if (sscanf(buf, "%c%u", &op, &port) != 2 || port > 0xFFFF) |
| return -EINVAL; |
| pr_err("Enter learner_policy_store:op=%c\n", op); |
| if (op == '-') { |
| /* remove port from the restricted list*/ |
| spin_lock_bh(&priv->lock); |
| list_for_each_entry_safe(entry, tmp, &priv->policy_list, list) |
| if (entry && entry->port == port) { |
| list_del(&entry->list); |
| kfree(entry); |
| } |
| spin_unlock_bh(&priv->lock); |
| } else if (op == '+') { |
| /* add port to the restricted list*/ |
| entry = kzalloc(sizeof(struct policy_entry), GFP_KERNEL); |
| if (!entry) |
| return -ENOMEM; |
| |
| INIT_LIST_HEAD(&entry->list); |
| entry->port = port; |
| |
| spin_lock_bh(&priv->lock); |
| list_add(&entry->list, &priv->policy_list); |
| spin_unlock_bh(&priv->lock); |
| |
| fpdb_del_by_port(port); |
| } else { |
| return -EINVAL; |
| } |
| |
| return count; |
| } |
| |
| static ssize_t learner_lookup_retries_store(struct fastpath_module *m, |
| const char *buf, size_t count) |
| { |
| unsigned int retries; |
| struct fp_learner *priv = m->priv; |
| |
| if (sscanf(buf, "%u", &retries) != 1) |
| return -EINVAL; |
| |
| priv->lookups_retries = retries; |
| |
| return count; |
| } |
| |
| static ssize_t learner_lookup_retries_show(struct fastpath_module *m, char *buf) |
| { |
| struct fp_learner *priv = m->priv; |
| return scnprintf(buf, PAGE_SIZE, "%u\n", priv->lookups_retries); |
| } |
| |
| static ssize_t learner_lookup_delay_store(struct fastpath_module *m, |
| const char *buf, size_t count) |
| { |
| unsigned int delay; |
| struct fp_learner *priv = m->priv; |
| |
| if (sscanf(buf, "%u", &delay) != 1) |
| return -EINVAL; |
| |
| priv->lookups_delay = delay; |
| |
| return count; |
| } |
| |
| static ssize_t learner_lookup_delay_show(struct fastpath_module *m, char *buf) |
| { |
| struct fp_learner *priv = m->priv; |
| return scnprintf(buf, PAGE_SIZE, "%u[ms]\n", priv->lookups_delay); |
| } |
| |
| static ssize_t learner_nfnl_groups_show(struct fastpath_module *m, char *buf) |
| { |
| struct fp_learner *priv = m->priv; |
| struct sockaddr addr; |
| int rc; |
| |
| rc = kernel_getsockname(priv->nf_nl_sock, &addr); |
| if (rc < 0) |
| return scnprintf(buf, PAGE_SIZE, "ERROR\n"); |
| |
| return scnprintf(buf, PAGE_SIZE, "0x%08x\n", ((struct sockaddr_nl *)&addr)->nl_groups); |
| } |
| |
| static ssize_t learner_rtnl_groups_show(struct fastpath_module *m, char *buf) |
| { |
| struct fp_learner *priv = m->priv; |
| struct sockaddr addr; |
| int rc; |
| |
| rc = kernel_getsockname(priv->rt_nl_sock, &addr); |
| if (rc < 0) |
| return scnprintf(buf, PAGE_SIZE, "ERROR\n"); |
| |
| return scnprintf(buf, PAGE_SIZE, "0x%08x\n", ((struct sockaddr_nl *)&addr)->nl_groups); |
| } |
| |
| static FP_ATTR(policy, S_IRUGO|S_IWUSR, learner_policy_show, learner_policy_store); |
| static FP_ATTR(lookup_retries, S_IRUGO|S_IWUSR, learner_lookup_retries_show, learner_lookup_retries_store); |
| static FP_ATTR(lookup_delay, S_IRUGO|S_IWUSR, learner_lookup_delay_show, learner_lookup_delay_store); |
| static FP_ATTR(nfnl_groups, S_IRUGO, learner_nfnl_groups_show, NULL); |
| static FP_ATTR(rtnl_groups, S_IRUGO, learner_rtnl_groups_show, NULL); |
| static FP_ATTR(update, S_IWUSR, NULL, learner_ct_update_all); |
| static FP_ATTR(block, S_IWUSR, NULL, learner_ct_block_all); |
| |
| static struct attribute *fp_learner_attrs[] = { |
| &fp_attr_policy.attr, |
| &fp_attr_lookup_retries.attr, |
| &fp_attr_lookup_delay.attr, |
| &fp_attr_nfnl_groups.attr, |
| &fp_attr_rtnl_groups.attr, |
| &fp_attr_update.attr, |
| &fp_attr_block.attr, |
| NULL, /* need to NULL terminate the list of attributes */ |
| }; |
| |
| static int |
| fp_learner_ioctl(struct fastpath_module *m, unsigned int cmd, void *data) |
| { |
| BUG_ON(!m); |
| |
| switch (cmd) { |
| case FASTPATH_NL_C_IPT_NOTIFY: |
| __learner_ct_block_all(m); |
| break; |
| default: |
| pr_err("unsupported command %u\n", cmd); |
| return -ENOTSUPP; |
| } |
| |
| return 0; |
| } |
| |
| static void fp_learner_release(struct kobject *kobj) |
| { |
| struct fastpath_module *module = to_fpmod(kobj); |
| struct fp_learner *priv = module->priv; |
| |
| priv->fp_rmmoding = 1; |
| nf_unregister_net_hooks(&init_net, nf_learner_hook_data, ARRAY_SIZE(nf_learner_hook_data)); |
| unregister_inet6addr_notifier(&priv->inet6addr_notifier); |
| unregister_netevent_notifier(&priv->netevent_notifier); |
| learner_netdev_cleanup(&priv->netdev_notifier); |
| unregister_netdevice_notifier(&priv->netdev_notifier); |
| learner_nl_close(priv->nf_nl_sock); |
| learner_nl_close(priv->rt_nl_sock); |
| |
| cancel_work_sync(&priv->update_work); |
| if (fp_learner_wq) { |
| flush_workqueue(priv->wq); |
| destroy_workqueue(priv->wq); |
| } |
| |
| kfree(priv); |
| kfree(module); |
| |
| pr_debug("fp_learner released\n"); |
| } |
| |
| static struct kobj_type ktype_learner = { |
| .sysfs_ops = &fp_sysfs_ops, |
| .default_attrs = fp_learner_attrs, |
| .release = fp_learner_release, |
| }; |
| |
| static int fp_learner_probe(struct fastpath_module *module) |
| { |
| struct fp_learner *priv; |
| int ret; |
| |
| priv = kzalloc(sizeof(*priv), GFP_KERNEL); |
| if (!priv) { |
| pr_err("no memeory\n"); |
| return -ENOMEM; |
| } |
| |
| module->priv = priv; |
| snprintf(module->name, sizeof(module->name),"fp_learner"); |
| |
| spin_lock_init(&priv->lock); |
| INIT_LIST_HEAD(&priv->policy_list); |
| priv->lookups_retries = DEFAULT_LOOKUPS_RETRIES; |
| priv->fp_rmmoding = 0; |
| |
| if (fp_learner_wq) { |
| INIT_LIST_HEAD(&priv->work_items_list); |
| priv->lookups_delay = DEFAULT_LOOKUPS_DELAY_MS; |
| priv->wq = create_singlethread_workqueue(module->name); |
| if (!priv->wq) { |
| pr_err("create workqueue failed\n"); |
| ret = -EBUSY; |
| goto priv_kfree; |
| } |
| } |
| |
| INIT_WORK(&priv->update_work, (void *)learner_ct_update_work); |
| |
| rtnl_lock(); |
| ret = learner_nl_open(priv, &priv->rt_nl_sock, NETLINK_ROUTE, RTNETLINK_GRP); |
| rtnl_unlock(); |
| if (ret < 0) { |
| pr_err("learner_nl_open(NETLINK_ROUTE) failed (%d)\n", ret); |
| goto wq_destroy; |
| } |
| |
| nfnl_lock(NFNL_SUBSYS_CTNETLINK); |
| ret = learner_nl_open(priv, &priv->nf_nl_sock, NETLINK_NETFILTER, NFNETLINK_GRP); |
| nfnl_unlock(NFNL_SUBSYS_CTNETLINK); |
| if (ret < 0) { |
| pr_err("learner_nl_open(NETLINK_NETFILTER) failed (%d)\n", ret); |
| goto nl_close_rt; |
| } |
| |
| priv->netdev_notifier.notifier_call = learner_netdev_event; |
| ret = register_netdevice_notifier(&priv->netdev_notifier); |
| if (ret < 0) { |
| pr_err("register_netdevice_notifier failed (%d)\n", ret); |
| goto nl_close_nf; |
| } |
| |
| priv->netevent_notifier.notifier_call = learner_netevent; |
| ret = register_netevent_notifier(&priv->netevent_notifier); |
| if (ret < 0) { |
| pr_err("register_netevent_notifier failed (%d)\n", ret); |
| goto netdev_notifier_unreg; |
| } |
| |
| priv->inet6addr_notifier.notifier_call = fp_inet6addr_event; |
| ret = register_inet6addr_notifier(&priv->inet6addr_notifier); |
| if (ret < 0) { |
| pr_err("register_inet6addr_notifier failed (%d)\n", ret); |
| goto netdev_netevent_unreg; |
| } |
| |
| ret = nf_register_net_hooks(&init_net, nf_learner_hook_data, ARRAY_SIZE(nf_learner_hook_data)); |
| if (ret < 0) { |
| pr_err("nf_register_hooks failed (%d)\n", ret); |
| goto in6_notifier_err; |
| } |
| |
| kobject_init(&module->kobj, &ktype_learner); |
| ret = kobject_add(&module->kobj, module->fastpath->kobj, "%s", module->name); |
| if (ret < 0) { |
| pr_err("kobject_add failed (%d)\n", ret); |
| goto nf_hooks_unreg; |
| } |
| |
| kobject_uevent(&module->kobj, KOBJ_ADD); |
| |
| pr_debug("fp_learner probed\n"); |
| return 0; |
| |
| nf_hooks_unreg: |
| kobject_put(&module->kobj); |
| nf_unregister_net_hooks(&init_net, nf_learner_hook_data, ARRAY_SIZE(nf_learner_hook_data)); |
| in6_notifier_err: |
| unregister_inet6addr_notifier(&priv->inet6addr_notifier); |
| netdev_netevent_unreg: |
| unregister_netevent_notifier(&priv->netevent_notifier); |
| netdev_notifier_unreg: |
| learner_netdev_cleanup(&priv->netdev_notifier); |
| unregister_netdevice_notifier(&priv->netdev_notifier); |
| nl_close_nf: |
| learner_nl_close(priv->nf_nl_sock); |
| nl_close_rt: |
| learner_nl_close(priv->rt_nl_sock); |
| wq_destroy: |
| if (fp_learner_wq) { |
| flush_workqueue(priv->wq); |
| destroy_workqueue(priv->wq); |
| } |
| priv_kfree: |
| kfree(priv); |
| |
| return ret; |
| } |
| |
| static int fp_learner_remove(struct fastpath_module *module) |
| { |
| kobject_put(&module->kobj); |
| |
| pr_debug("fp_learner removed\n"); |
| return 0; |
| } |
| |
| struct fastpath_module_ops fp_learner_ops = { |
| .probe = fp_learner_probe, |
| .remove = fp_learner_remove, |
| .ioctl = fp_learner_ioctl, |
| }; |
| |
| module_param(fp_learner_wq, bool, S_IRUGO); |
| MODULE_PARM_DESC(fp_learner_wq, "fastpath learner worqueue mode (default=" |
| __MODULE_STRING(FP_LEARNER_WQ_DEFAULT) ")"); |