Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 1 | /* |
Andy Zhou | 971427f3 | 2014-09-15 19:37:25 -0700 | [diff] [blame] | 2 | * Copyright (c) 2007-2014 Nicira, Inc. |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 3 | * |
| 4 | * This program is free software; you can redistribute it and/or |
| 5 | * modify it under the terms of version 2 of the GNU General Public |
| 6 | * License as published by the Free Software Foundation. |
| 7 | * |
| 8 | * This program is distributed in the hope that it will be useful, but |
| 9 | * WITHOUT ANY WARRANTY; without even the implied warranty of |
| 10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 11 | * General Public License for more details. |
| 12 | * |
| 13 | * You should have received a copy of the GNU General Public License |
| 14 | * along with this program; if not, write to the Free Software |
| 15 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA |
| 16 | * 02110-1301, USA |
| 17 | */ |
| 18 | |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 19 | #include <linux/uaccess.h> |
| 20 | #include <linux/netdevice.h> |
| 21 | #include <linux/etherdevice.h> |
| 22 | #include <linux/if_ether.h> |
| 23 | #include <linux/if_vlan.h> |
| 24 | #include <net/llc_pdu.h> |
| 25 | #include <linux/kernel.h> |
| 26 | #include <linux/jhash.h> |
| 27 | #include <linux/jiffies.h> |
| 28 | #include <linux/llc.h> |
| 29 | #include <linux/module.h> |
| 30 | #include <linux/in.h> |
| 31 | #include <linux/rcupdate.h> |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 32 | #include <linux/cpumask.h> |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 33 | #include <linux/if_arp.h> |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 34 | #include <linux/ip.h> |
| 35 | #include <linux/ipv6.h> |
Simon Horman | 25cd9ba | 2014-10-06 05:05:13 -0700 | [diff] [blame] | 36 | #include <linux/mpls.h> |
Joe Stringer | a175a72 | 2013-08-22 12:30:48 -0700 | [diff] [blame] | 37 | #include <linux/sctp.h> |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 38 | #include <linux/smp.h> |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 39 | #include <linux/tcp.h> |
| 40 | #include <linux/udp.h> |
| 41 | #include <linux/icmp.h> |
| 42 | #include <linux/icmpv6.h> |
| 43 | #include <linux/rculist.h> |
| 44 | #include <net/ip.h> |
Pravin B Shelar | 7d5437c | 2013-06-17 17:50:18 -0700 | [diff] [blame] | 45 | #include <net/ip_tunnels.h> |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 46 | #include <net/ipv6.h> |
Simon Horman | 25cd9ba | 2014-10-06 05:05:13 -0700 | [diff] [blame] | 47 | #include <net/mpls.h> |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 48 | #include <net/ndisc.h> |
| 49 | |
Pravin B Shelar | a581b96 | 2015-08-29 17:44:08 -0700 | [diff] [blame] | 50 | #include "conntrack.h" |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 51 | #include "datapath.h" |
| 52 | #include "flow.h" |
| 53 | #include "flow_netlink.h" |
Pravin B Shelar | a581b96 | 2015-08-29 17:44:08 -0700 | [diff] [blame] | 54 | #include "vport.h" |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 55 | |
Pravin B Shelar | e644571 | 2013-10-03 18:16:47 -0700 | [diff] [blame] | 56 | u64 ovs_flow_used_time(unsigned long flow_jiffies) |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 57 | { |
Pravin B Shelar | e644571 | 2013-10-03 18:16:47 -0700 | [diff] [blame] | 58 | struct timespec cur_ts; |
| 59 | u64 cur_ms, idle_ms; |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 60 | |
Pravin B Shelar | e644571 | 2013-10-03 18:16:47 -0700 | [diff] [blame] | 61 | ktime_get_ts(&cur_ts); |
| 62 | idle_ms = jiffies_to_msecs(jiffies - flow_jiffies); |
| 63 | cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC + |
| 64 | cur_ts.tv_nsec / NSEC_PER_MSEC; |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 65 | |
Pravin B Shelar | e644571 | 2013-10-03 18:16:47 -0700 | [diff] [blame] | 66 | return cur_ms - idle_ms; |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 67 | } |
| 68 | |
Jarno Rajahalme | df23e9f | 2013-10-23 01:40:44 -0700 | [diff] [blame] | 69 | #define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF)) |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 70 | |
Ben Pfaff | ad55200 | 2014-05-06 16:48:38 -0700 | [diff] [blame] | 71 | void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags, |
Thomas Graf | 12eb18f | 2014-11-06 06:58:52 -0800 | [diff] [blame] | 72 | const struct sk_buff *skb) |
Andy Zhou | 5828cd9 | 2013-08-27 13:02:21 -0700 | [diff] [blame] | 73 | { |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 74 | struct flow_stats *stats; |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 75 | int node = numa_node_id(); |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 76 | int cpu = smp_processor_id(); |
Jiri Pirko | df8a39d | 2015-01-13 17:13:44 +0100 | [diff] [blame] | 77 | int len = skb->len + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0); |
Andy Zhou | 5828cd9 | 2013-08-27 13:02:21 -0700 | [diff] [blame] | 78 | |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 79 | stats = rcu_dereference(flow->stats[cpu]); |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 80 | |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 81 | /* Check if already have CPU-specific stats. */ |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 82 | if (likely(stats)) { |
| 83 | spin_lock(&stats->lock); |
| 84 | /* Mark if we write on the pre-allocated stats. */ |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 85 | if (cpu == 0 && unlikely(flow->stats_last_writer != cpu)) |
| 86 | flow->stats_last_writer = cpu; |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 87 | } else { |
| 88 | stats = rcu_dereference(flow->stats[0]); /* Pre-allocated. */ |
| 89 | spin_lock(&stats->lock); |
| 90 | |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 91 | /* If the current CPU is the only writer on the |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 92 | * pre-allocated stats keep using them. |
| 93 | */ |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 94 | if (unlikely(flow->stats_last_writer != cpu)) { |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 95 | /* A previous locker may have already allocated the |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 96 | * stats, so we need to check again. If CPU-specific |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 97 | * stats were already allocated, we update the pre- |
| 98 | * allocated stats as we have already locked them. |
| 99 | */ |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 100 | if (likely(flow->stats_last_writer != -1) && |
| 101 | likely(!rcu_access_pointer(flow->stats[cpu]))) { |
| 102 | /* Try to allocate CPU-specific stats. */ |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 103 | struct flow_stats *new_stats; |
| 104 | |
| 105 | new_stats = |
| 106 | kmem_cache_alloc_node(flow_stats_cache, |
David Rientjes | 4167e9b | 2015-04-14 15:46:55 -0700 | [diff] [blame] | 107 | GFP_NOWAIT | |
| 108 | __GFP_THISNODE | |
| 109 | __GFP_NOWARN | |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 110 | __GFP_NOMEMALLOC, |
| 111 | node); |
| 112 | if (likely(new_stats)) { |
| 113 | new_stats->used = jiffies; |
| 114 | new_stats->packet_count = 1; |
Ben Pfaff | 24cc59d | 2014-12-31 08:45:46 -0800 | [diff] [blame] | 115 | new_stats->byte_count = len; |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 116 | new_stats->tcp_flags = tcp_flags; |
| 117 | spin_lock_init(&new_stats->lock); |
| 118 | |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 119 | rcu_assign_pointer(flow->stats[cpu], |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 120 | new_stats); |
| 121 | goto unlock; |
| 122 | } |
| 123 | } |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 124 | flow->stats_last_writer = cpu; |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 125 | } |
| 126 | } |
| 127 | |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 128 | stats->used = jiffies; |
| 129 | stats->packet_count++; |
Ben Pfaff | 24cc59d | 2014-12-31 08:45:46 -0800 | [diff] [blame] | 130 | stats->byte_count += len; |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 131 | stats->tcp_flags |= tcp_flags; |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 132 | unlock: |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 133 | spin_unlock(&stats->lock); |
| 134 | } |
| 135 | |
Jarno Rajahalme | 86ec8db | 2014-05-05 14:17:28 -0700 | [diff] [blame] | 136 | /* Must be called with rcu_read_lock or ovs_mutex. */ |
| 137 | void ovs_flow_stats_get(const struct sw_flow *flow, |
| 138 | struct ovs_flow_stats *ovs_stats, |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 139 | unsigned long *used, __be16 *tcp_flags) |
| 140 | { |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 141 | int cpu; |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 142 | |
| 143 | *used = 0; |
| 144 | *tcp_flags = 0; |
| 145 | memset(ovs_stats, 0, sizeof(*ovs_stats)); |
| 146 | |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 147 | /* We open code this to make sure cpu 0 is always considered */ |
| 148 | for (cpu = 0; cpu < nr_cpu_ids; cpu = cpumask_next(cpu, cpu_possible_mask)) { |
| 149 | struct flow_stats *stats = rcu_dereference_ovsl(flow->stats[cpu]); |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 150 | |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 151 | if (stats) { |
| 152 | /* Local CPU may write on non-local stats, so we must |
| 153 | * block bottom-halves here. |
| 154 | */ |
| 155 | spin_lock_bh(&stats->lock); |
| 156 | if (!*used || time_after(stats->used, *used)) |
| 157 | *used = stats->used; |
| 158 | *tcp_flags |= stats->tcp_flags; |
| 159 | ovs_stats->n_packets += stats->packet_count; |
| 160 | ovs_stats->n_bytes += stats->byte_count; |
| 161 | spin_unlock_bh(&stats->lock); |
| 162 | } |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 163 | } |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 164 | } |
| 165 | |
Jarno Rajahalme | 86ec8db | 2014-05-05 14:17:28 -0700 | [diff] [blame] | 166 | /* Called with ovs_mutex. */ |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 167 | void ovs_flow_stats_clear(struct sw_flow *flow) |
| 168 | { |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 169 | int cpu; |
Pravin B Shelar | e298e50 | 2013-10-29 17:22:21 -0700 | [diff] [blame] | 170 | |
Thadeu Lima de Souza Cascardo | db74a33 | 2016-09-15 19:11:53 -0300 | [diff] [blame] | 171 | /* We open code this to make sure cpu 0 is always considered */ |
| 172 | for (cpu = 0; cpu < nr_cpu_ids; cpu = cpumask_next(cpu, cpu_possible_mask)) { |
| 173 | struct flow_stats *stats = ovsl_dereference(flow->stats[cpu]); |
Jarno Rajahalme | 23dabf8 | 2014-03-27 12:35:23 -0700 | [diff] [blame] | 174 | |
Jarno Rajahalme | 63e7959 | 2014-03-27 12:42:54 -0700 | [diff] [blame] | 175 | if (stats) { |
| 176 | spin_lock_bh(&stats->lock); |
| 177 | stats->used = 0; |
| 178 | stats->packet_count = 0; |
| 179 | stats->byte_count = 0; |
| 180 | stats->tcp_flags = 0; |
| 181 | spin_unlock_bh(&stats->lock); |
| 182 | } |
| 183 | } |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 184 | } |
| 185 | |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 186 | static int check_header(struct sk_buff *skb, int len) |
| 187 | { |
| 188 | if (unlikely(skb->len < len)) |
| 189 | return -EINVAL; |
| 190 | if (unlikely(!pskb_may_pull(skb, len))) |
| 191 | return -ENOMEM; |
| 192 | return 0; |
| 193 | } |
| 194 | |
| 195 | static bool arphdr_ok(struct sk_buff *skb) |
| 196 | { |
| 197 | return pskb_may_pull(skb, skb_network_offset(skb) + |
| 198 | sizeof(struct arp_eth_header)); |
| 199 | } |
| 200 | |
| 201 | static int check_iphdr(struct sk_buff *skb) |
| 202 | { |
| 203 | unsigned int nh_ofs = skb_network_offset(skb); |
| 204 | unsigned int ip_len; |
| 205 | int err; |
| 206 | |
| 207 | err = check_header(skb, nh_ofs + sizeof(struct iphdr)); |
| 208 | if (unlikely(err)) |
| 209 | return err; |
| 210 | |
| 211 | ip_len = ip_hdrlen(skb); |
| 212 | if (unlikely(ip_len < sizeof(struct iphdr) || |
| 213 | skb->len < nh_ofs + ip_len)) |
| 214 | return -EINVAL; |
| 215 | |
| 216 | skb_set_transport_header(skb, nh_ofs + ip_len); |
| 217 | return 0; |
| 218 | } |
| 219 | |
| 220 | static bool tcphdr_ok(struct sk_buff *skb) |
| 221 | { |
| 222 | int th_ofs = skb_transport_offset(skb); |
| 223 | int tcp_len; |
| 224 | |
| 225 | if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr)))) |
| 226 | return false; |
| 227 | |
| 228 | tcp_len = tcp_hdrlen(skb); |
| 229 | if (unlikely(tcp_len < sizeof(struct tcphdr) || |
| 230 | skb->len < th_ofs + tcp_len)) |
| 231 | return false; |
| 232 | |
| 233 | return true; |
| 234 | } |
| 235 | |
| 236 | static bool udphdr_ok(struct sk_buff *skb) |
| 237 | { |
| 238 | return pskb_may_pull(skb, skb_transport_offset(skb) + |
| 239 | sizeof(struct udphdr)); |
| 240 | } |
| 241 | |
Joe Stringer | a175a72 | 2013-08-22 12:30:48 -0700 | [diff] [blame] | 242 | static bool sctphdr_ok(struct sk_buff *skb) |
| 243 | { |
| 244 | return pskb_may_pull(skb, skb_transport_offset(skb) + |
| 245 | sizeof(struct sctphdr)); |
| 246 | } |
| 247 | |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 248 | static bool icmphdr_ok(struct sk_buff *skb) |
| 249 | { |
| 250 | return pskb_may_pull(skb, skb_transport_offset(skb) + |
| 251 | sizeof(struct icmphdr)); |
| 252 | } |
| 253 | |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 254 | static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key) |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 255 | { |
| 256 | unsigned int nh_ofs = skb_network_offset(skb); |
| 257 | unsigned int nh_len; |
| 258 | int payload_ofs; |
| 259 | struct ipv6hdr *nh; |
| 260 | uint8_t nexthdr; |
| 261 | __be16 frag_off; |
| 262 | int err; |
| 263 | |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 264 | err = check_header(skb, nh_ofs + sizeof(*nh)); |
| 265 | if (unlikely(err)) |
| 266 | return err; |
| 267 | |
| 268 | nh = ipv6_hdr(skb); |
| 269 | nexthdr = nh->nexthdr; |
| 270 | payload_ofs = (u8 *)(nh + 1) - skb->data; |
| 271 | |
| 272 | key->ip.proto = NEXTHDR_NONE; |
| 273 | key->ip.tos = ipv6_get_dsfield(nh); |
| 274 | key->ip.ttl = nh->hop_limit; |
| 275 | key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); |
| 276 | key->ipv6.addr.src = nh->saddr; |
| 277 | key->ipv6.addr.dst = nh->daddr; |
| 278 | |
| 279 | payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 280 | |
| 281 | if (frag_off) { |
| 282 | if (frag_off & htons(~0x7)) |
| 283 | key->ip.frag = OVS_FRAG_TYPE_LATER; |
| 284 | else |
| 285 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
Pravin B Shelar | 25ef132 | 2014-10-17 13:56:31 -0700 | [diff] [blame] | 286 | } else { |
| 287 | key->ip.frag = OVS_FRAG_TYPE_NONE; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 288 | } |
| 289 | |
Simon Horman | c30da49 | 2015-08-29 09:02:21 +0900 | [diff] [blame] | 290 | /* Delayed handling of error in ipv6_skip_exthdr() as it |
| 291 | * always sets frag_off to a valid value which may be |
| 292 | * used to set key->ip.frag above. |
| 293 | */ |
| 294 | if (unlikely(payload_ofs < 0)) |
| 295 | return -EPROTO; |
| 296 | |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 297 | nh_len = payload_ofs - nh_ofs; |
| 298 | skb_set_transport_header(skb, nh_ofs + nh_len); |
| 299 | key->ip.proto = nexthdr; |
| 300 | return nh_len; |
| 301 | } |
| 302 | |
| 303 | static bool icmp6hdr_ok(struct sk_buff *skb) |
| 304 | { |
| 305 | return pskb_may_pull(skb, skb_transport_offset(skb) + |
| 306 | sizeof(struct icmp6hdr)); |
| 307 | } |
| 308 | |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 309 | /** |
| 310 | * Parse vlan tag from vlan header. |
| 311 | * Returns ERROR on memory error. |
| 312 | * Returns 0 if it encounters a non-vlan or incomplete packet. |
| 313 | * Returns 1 after successfully parsing vlan tag. |
| 314 | */ |
pravin shelar | df30f74 | 2016-12-26 08:31:27 -0800 | [diff] [blame] | 315 | static int parse_vlan_tag(struct sk_buff *skb, struct vlan_head *key_vh, |
| 316 | bool untag_vlan) |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 317 | { |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 318 | struct vlan_head *vh = (struct vlan_head *)skb->data; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 319 | |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 320 | if (likely(!eth_type_vlan(vh->tpid))) |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 321 | return 0; |
| 322 | |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 323 | if (unlikely(skb->len < sizeof(struct vlan_head) + sizeof(__be16))) |
| 324 | return 0; |
| 325 | |
| 326 | if (unlikely(!pskb_may_pull(skb, sizeof(struct vlan_head) + |
| 327 | sizeof(__be16)))) |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 328 | return -ENOMEM; |
| 329 | |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 330 | vh = (struct vlan_head *)skb->data; |
| 331 | key_vh->tci = vh->tci | htons(VLAN_TAG_PRESENT); |
| 332 | key_vh->tpid = vh->tpid; |
| 333 | |
pravin shelar | df30f74 | 2016-12-26 08:31:27 -0800 | [diff] [blame] | 334 | if (unlikely(untag_vlan)) { |
| 335 | int offset = skb->data - skb_mac_header(skb); |
| 336 | u16 tci; |
| 337 | int err; |
| 338 | |
| 339 | __skb_push(skb, offset); |
| 340 | err = __skb_vlan_pop(skb, &tci); |
| 341 | __skb_pull(skb, offset); |
| 342 | if (err) |
| 343 | return err; |
| 344 | __vlan_hwaccel_put_tag(skb, key_vh->tpid, tci); |
| 345 | } else { |
| 346 | __skb_pull(skb, sizeof(struct vlan_head)); |
| 347 | } |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 348 | return 1; |
| 349 | } |
| 350 | |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 351 | static void clear_vlan(struct sw_flow_key *key) |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 352 | { |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 353 | key->eth.vlan.tci = 0; |
| 354 | key->eth.vlan.tpid = 0; |
| 355 | key->eth.cvlan.tci = 0; |
| 356 | key->eth.cvlan.tpid = 0; |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 357 | } |
| 358 | |
| 359 | static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key) |
| 360 | { |
| 361 | int res; |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 362 | |
Jiri Benc | 20ecf1e | 2016-10-10 17:02:42 +0200 | [diff] [blame] | 363 | if (skb_vlan_tag_present(skb)) { |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 364 | key->eth.vlan.tci = htons(skb->vlan_tci); |
| 365 | key->eth.vlan.tpid = skb->vlan_proto; |
| 366 | } else { |
| 367 | /* Parse outer vlan tag in the non-accelerated case. */ |
pravin shelar | df30f74 | 2016-12-26 08:31:27 -0800 | [diff] [blame] | 368 | res = parse_vlan_tag(skb, &key->eth.vlan, true); |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 369 | if (res <= 0) |
| 370 | return res; |
| 371 | } |
| 372 | |
| 373 | /* Parse inner vlan tag. */ |
pravin shelar | df30f74 | 2016-12-26 08:31:27 -0800 | [diff] [blame] | 374 | res = parse_vlan_tag(skb, &key->eth.cvlan, false); |
Eric Garver | 018c1dd | 2016-09-07 12:56:59 -0400 | [diff] [blame] | 375 | if (res <= 0) |
| 376 | return res; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 377 | |
| 378 | return 0; |
| 379 | } |
| 380 | |
| 381 | static __be16 parse_ethertype(struct sk_buff *skb) |
| 382 | { |
| 383 | struct llc_snap_hdr { |
| 384 | u8 dsap; /* Always 0xAA */ |
| 385 | u8 ssap; /* Always 0xAA */ |
| 386 | u8 ctrl; |
| 387 | u8 oui[3]; |
| 388 | __be16 ethertype; |
| 389 | }; |
| 390 | struct llc_snap_hdr *llc; |
| 391 | __be16 proto; |
| 392 | |
| 393 | proto = *(__be16 *) skb->data; |
| 394 | __skb_pull(skb, sizeof(__be16)); |
| 395 | |
Alexander Duyck | 6713fc9 | 2015-05-04 14:34:05 -0700 | [diff] [blame] | 396 | if (eth_proto_is_802_3(proto)) |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 397 | return proto; |
| 398 | |
| 399 | if (skb->len < sizeof(struct llc_snap_hdr)) |
| 400 | return htons(ETH_P_802_2); |
| 401 | |
| 402 | if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr)))) |
| 403 | return htons(0); |
| 404 | |
| 405 | llc = (struct llc_snap_hdr *) skb->data; |
| 406 | if (llc->dsap != LLC_SAP_SNAP || |
| 407 | llc->ssap != LLC_SAP_SNAP || |
| 408 | (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0) |
| 409 | return htons(ETH_P_802_2); |
| 410 | |
| 411 | __skb_pull(skb, sizeof(struct llc_snap_hdr)); |
Rich Lane | 17b682a | 2013-02-19 11:10:30 -0800 | [diff] [blame] | 412 | |
Alexander Duyck | 6713fc9 | 2015-05-04 14:34:05 -0700 | [diff] [blame] | 413 | if (eth_proto_is_802_3(llc->ethertype)) |
Rich Lane | 17b682a | 2013-02-19 11:10:30 -0800 | [diff] [blame] | 414 | return llc->ethertype; |
| 415 | |
| 416 | return htons(ETH_P_802_2); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 417 | } |
| 418 | |
| 419 | static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key, |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 420 | int nh_len) |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 421 | { |
| 422 | struct icmp6hdr *icmp = icmp6_hdr(skb); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 423 | |
| 424 | /* The ICMPv6 type and code fields use the 16-bit transport port |
| 425 | * fields, so we need to store them in 16-bit network byte order. |
| 426 | */ |
Jarno Rajahalme | 1139e24 | 2014-05-05 09:54:49 -0700 | [diff] [blame] | 427 | key->tp.src = htons(icmp->icmp6_type); |
| 428 | key->tp.dst = htons(icmp->icmp6_code); |
Pravin B Shelar | 25ef132 | 2014-10-17 13:56:31 -0700 | [diff] [blame] | 429 | memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd)); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 430 | |
| 431 | if (icmp->icmp6_code == 0 && |
| 432 | (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION || |
| 433 | icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) { |
| 434 | int icmp_len = skb->len - skb_transport_offset(skb); |
| 435 | struct nd_msg *nd; |
| 436 | int offset; |
| 437 | |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 438 | /* In order to process neighbor discovery options, we need the |
| 439 | * entire packet. |
| 440 | */ |
| 441 | if (unlikely(icmp_len < sizeof(*nd))) |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 442 | return 0; |
| 443 | |
| 444 | if (unlikely(skb_linearize(skb))) |
| 445 | return -ENOMEM; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 446 | |
| 447 | nd = (struct nd_msg *)skb_transport_header(skb); |
| 448 | key->ipv6.nd.target = nd->target; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 449 | |
| 450 | icmp_len -= sizeof(*nd); |
| 451 | offset = 0; |
| 452 | while (icmp_len >= 8) { |
| 453 | struct nd_opt_hdr *nd_opt = |
| 454 | (struct nd_opt_hdr *)(nd->opt + offset); |
| 455 | int opt_len = nd_opt->nd_opt_len * 8; |
| 456 | |
| 457 | if (unlikely(!opt_len || opt_len > icmp_len)) |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 458 | return 0; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 459 | |
| 460 | /* Store the link layer address if the appropriate |
| 461 | * option is provided. It is considered an error if |
| 462 | * the same link layer option is specified twice. |
| 463 | */ |
| 464 | if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR |
| 465 | && opt_len == 8) { |
| 466 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll))) |
| 467 | goto invalid; |
Joe Perches | 8c63ff0 | 2014-02-18 11:15:45 -0800 | [diff] [blame] | 468 | ether_addr_copy(key->ipv6.nd.sll, |
| 469 | &nd->opt[offset+sizeof(*nd_opt)]); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 470 | } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR |
| 471 | && opt_len == 8) { |
| 472 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll))) |
| 473 | goto invalid; |
Joe Perches | 8c63ff0 | 2014-02-18 11:15:45 -0800 | [diff] [blame] | 474 | ether_addr_copy(key->ipv6.nd.tll, |
| 475 | &nd->opt[offset+sizeof(*nd_opt)]); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 476 | } |
| 477 | |
| 478 | icmp_len -= opt_len; |
| 479 | offset += opt_len; |
| 480 | } |
| 481 | } |
| 482 | |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 483 | return 0; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 484 | |
| 485 | invalid: |
| 486 | memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target)); |
| 487 | memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll)); |
| 488 | memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll)); |
| 489 | |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 490 | return 0; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 491 | } |
| 492 | |
| 493 | /** |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 494 | * key_extract - extracts a flow key from an Ethernet frame. |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 495 | * @skb: sk_buff that contains the frame, with skb->data pointing to the |
| 496 | * Ethernet header |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 497 | * @key: output flow key |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 498 | * |
| 499 | * The caller must ensure that skb->len >= ETH_HLEN. |
| 500 | * |
| 501 | * Returns 0 if successful, otherwise a negative errno value. |
| 502 | * |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 503 | * Initializes @skb header fields as follows: |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 504 | * |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 505 | * - skb->mac_header: the L2 header. |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 506 | * |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 507 | * - skb->network_header: just past the L2 header, or just past the |
| 508 | * VLAN header, to the first byte of the L2 payload. |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 509 | * |
Lorand Jakab | 34d94f2 | 2013-06-03 10:01:14 -0700 | [diff] [blame] | 510 | * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6 |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 511 | * on output, then just past the IP header, if one is present and |
| 512 | * of a correct length, otherwise the same as skb->network_header. |
Lorand Jakab | 34d94f2 | 2013-06-03 10:01:14 -0700 | [diff] [blame] | 513 | * For other key->eth.type values it is left untouched. |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 514 | * |
| 515 | * - skb->protocol: the type of the data starting at skb->network_header. |
| 516 | * Equals to key->eth.type. |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 517 | */ |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 518 | static int key_extract(struct sk_buff *skb, struct sw_flow_key *key) |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 519 | { |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 520 | int error; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 521 | struct ethhdr *eth; |
| 522 | |
Jesse Gross | f579668 | 2014-10-03 15:35:33 -0700 | [diff] [blame] | 523 | /* Flags are always used as part of stats */ |
| 524 | key->tp.flags = 0; |
| 525 | |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 526 | skb_reset_mac_header(skb); |
| 527 | |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 528 | /* Link layer. */ |
| 529 | clear_vlan(key); |
Yi-Hung Wei | 6f56f61 | 2017-03-30 12:36:03 -0700 | [diff] [blame] | 530 | if (ovs_key_mac_proto(key) == MAC_PROTO_NONE) { |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 531 | if (unlikely(eth_type_vlan(skb->protocol))) |
| 532 | return -EINVAL; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 533 | |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 534 | skb_reset_network_header(skb); |
| 535 | } else { |
| 536 | eth = eth_hdr(skb); |
| 537 | ether_addr_copy(key->eth.src, eth->h_source); |
| 538 | ether_addr_copy(key->eth.dst, eth->h_dest); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 539 | |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 540 | __skb_pull(skb, 2 * ETH_ALEN); |
| 541 | /* We are going to push all headers that we pull, so no need to |
| 542 | * update skb->csum here. |
| 543 | */ |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 544 | |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 545 | if (unlikely(parse_vlan(skb, key))) |
| 546 | return -ENOMEM; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 547 | |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 548 | skb->protocol = parse_ethertype(skb); |
| 549 | if (unlikely(skb->protocol == htons(0))) |
| 550 | return -ENOMEM; |
| 551 | |
| 552 | skb_reset_network_header(skb); |
| 553 | __skb_push(skb, skb->data - skb_mac_header(skb)); |
| 554 | } |
Simon Horman | 25cd9ba | 2014-10-06 05:05:13 -0700 | [diff] [blame] | 555 | skb_reset_mac_len(skb); |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 556 | key->eth.type = skb->protocol; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 557 | |
| 558 | /* Network layer. */ |
| 559 | if (key->eth.type == htons(ETH_P_IP)) { |
| 560 | struct iphdr *nh; |
| 561 | __be16 offset; |
| 562 | |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 563 | error = check_iphdr(skb); |
| 564 | if (unlikely(error)) { |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 565 | memset(&key->ip, 0, sizeof(key->ip)); |
| 566 | memset(&key->ipv4, 0, sizeof(key->ipv4)); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 567 | if (error == -EINVAL) { |
| 568 | skb->transport_header = skb->network_header; |
| 569 | error = 0; |
| 570 | } |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 571 | return error; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 572 | } |
| 573 | |
| 574 | nh = ip_hdr(skb); |
| 575 | key->ipv4.addr.src = nh->saddr; |
| 576 | key->ipv4.addr.dst = nh->daddr; |
| 577 | |
| 578 | key->ip.proto = nh->protocol; |
| 579 | key->ip.tos = nh->tos; |
| 580 | key->ip.ttl = nh->ttl; |
| 581 | |
| 582 | offset = nh->frag_off & htons(IP_OFFSET); |
| 583 | if (offset) { |
| 584 | key->ip.frag = OVS_FRAG_TYPE_LATER; |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 585 | return 0; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 586 | } |
| 587 | if (nh->frag_off & htons(IP_MF) || |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 588 | skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 589 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 590 | else |
| 591 | key->ip.frag = OVS_FRAG_TYPE_NONE; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 592 | |
| 593 | /* Transport layer. */ |
| 594 | if (key->ip.proto == IPPROTO_TCP) { |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 595 | if (tcphdr_ok(skb)) { |
| 596 | struct tcphdr *tcp = tcp_hdr(skb); |
Jarno Rajahalme | 1139e24 | 2014-05-05 09:54:49 -0700 | [diff] [blame] | 597 | key->tp.src = tcp->source; |
| 598 | key->tp.dst = tcp->dest; |
| 599 | key->tp.flags = TCP_FLAGS_BE16(tcp); |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 600 | } else { |
| 601 | memset(&key->tp, 0, sizeof(key->tp)); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 602 | } |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 603 | |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 604 | } else if (key->ip.proto == IPPROTO_UDP) { |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 605 | if (udphdr_ok(skb)) { |
| 606 | struct udphdr *udp = udp_hdr(skb); |
Jarno Rajahalme | 1139e24 | 2014-05-05 09:54:49 -0700 | [diff] [blame] | 607 | key->tp.src = udp->source; |
| 608 | key->tp.dst = udp->dest; |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 609 | } else { |
| 610 | memset(&key->tp, 0, sizeof(key->tp)); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 611 | } |
Joe Stringer | a175a72 | 2013-08-22 12:30:48 -0700 | [diff] [blame] | 612 | } else if (key->ip.proto == IPPROTO_SCTP) { |
| 613 | if (sctphdr_ok(skb)) { |
| 614 | struct sctphdr *sctp = sctp_hdr(skb); |
Jarno Rajahalme | 1139e24 | 2014-05-05 09:54:49 -0700 | [diff] [blame] | 615 | key->tp.src = sctp->source; |
| 616 | key->tp.dst = sctp->dest; |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 617 | } else { |
| 618 | memset(&key->tp, 0, sizeof(key->tp)); |
Joe Stringer | a175a72 | 2013-08-22 12:30:48 -0700 | [diff] [blame] | 619 | } |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 620 | } else if (key->ip.proto == IPPROTO_ICMP) { |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 621 | if (icmphdr_ok(skb)) { |
| 622 | struct icmphdr *icmp = icmp_hdr(skb); |
| 623 | /* The ICMP type and code fields use the 16-bit |
| 624 | * transport port fields, so we need to store |
| 625 | * them in 16-bit network byte order. */ |
Jarno Rajahalme | 1139e24 | 2014-05-05 09:54:49 -0700 | [diff] [blame] | 626 | key->tp.src = htons(icmp->type); |
| 627 | key->tp.dst = htons(icmp->code); |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 628 | } else { |
| 629 | memset(&key->tp, 0, sizeof(key->tp)); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 630 | } |
| 631 | } |
| 632 | |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 633 | } else if (key->eth.type == htons(ETH_P_ARP) || |
| 634 | key->eth.type == htons(ETH_P_RARP)) { |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 635 | struct arp_eth_header *arp; |
Li RongQing | 389f489 | 2014-10-17 14:03:08 +0800 | [diff] [blame] | 636 | bool arp_available = arphdr_ok(skb); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 637 | |
| 638 | arp = (struct arp_eth_header *)skb_network_header(skb); |
| 639 | |
Li RongQing | 389f489 | 2014-10-17 14:03:08 +0800 | [diff] [blame] | 640 | if (arp_available && |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 641 | arp->ar_hrd == htons(ARPHRD_ETHER) && |
| 642 | arp->ar_pro == htons(ETH_P_IP) && |
| 643 | arp->ar_hln == ETH_ALEN && |
| 644 | arp->ar_pln == 4) { |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 645 | |
| 646 | /* We only match on the lower 8 bits of the opcode. */ |
| 647 | if (ntohs(arp->ar_op) <= 0xff) |
| 648 | key->ip.proto = ntohs(arp->ar_op); |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 649 | else |
| 650 | key->ip.proto = 0; |
| 651 | |
Mehak Mahajan | d04d382 | 2012-10-30 15:50:28 -0700 | [diff] [blame] | 652 | memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src)); |
| 653 | memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst)); |
Joe Perches | 8c63ff0 | 2014-02-18 11:15:45 -0800 | [diff] [blame] | 654 | ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha); |
| 655 | ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha); |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 656 | } else { |
| 657 | memset(&key->ip, 0, sizeof(key->ip)); |
| 658 | memset(&key->ipv4, 0, sizeof(key->ipv4)); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 659 | } |
Simon Horman | 25cd9ba | 2014-10-06 05:05:13 -0700 | [diff] [blame] | 660 | } else if (eth_p_mpls(key->eth.type)) { |
| 661 | size_t stack_len = MPLS_HLEN; |
| 662 | |
Jiri Benc | f7d49bc | 2016-09-30 19:08:05 +0200 | [diff] [blame] | 663 | skb_set_inner_network_header(skb, skb->mac_len); |
Simon Horman | 25cd9ba | 2014-10-06 05:05:13 -0700 | [diff] [blame] | 664 | while (1) { |
| 665 | __be32 lse; |
| 666 | |
| 667 | error = check_header(skb, skb->mac_len + stack_len); |
| 668 | if (unlikely(error)) |
| 669 | return 0; |
| 670 | |
Jiri Benc | f7d49bc | 2016-09-30 19:08:05 +0200 | [diff] [blame] | 671 | memcpy(&lse, skb_inner_network_header(skb), MPLS_HLEN); |
Simon Horman | 25cd9ba | 2014-10-06 05:05:13 -0700 | [diff] [blame] | 672 | |
| 673 | if (stack_len == MPLS_HLEN) |
| 674 | memcpy(&key->mpls.top_lse, &lse, MPLS_HLEN); |
| 675 | |
Jiri Benc | f7d49bc | 2016-09-30 19:08:05 +0200 | [diff] [blame] | 676 | skb_set_inner_network_header(skb, skb->mac_len + stack_len); |
Simon Horman | 25cd9ba | 2014-10-06 05:05:13 -0700 | [diff] [blame] | 677 | if (lse & htonl(MPLS_LS_S_MASK)) |
| 678 | break; |
| 679 | |
| 680 | stack_len += MPLS_HLEN; |
| 681 | } |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 682 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
| 683 | int nh_len; /* IPv6 Header + Extensions */ |
| 684 | |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 685 | nh_len = parse_ipv6hdr(skb, key); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 686 | if (unlikely(nh_len < 0)) { |
Simon Horman | c30da49 | 2015-08-29 09:02:21 +0900 | [diff] [blame] | 687 | switch (nh_len) { |
| 688 | case -EINVAL: |
| 689 | memset(&key->ip, 0, sizeof(key->ip)); |
| 690 | memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr)); |
| 691 | /* fall-through */ |
| 692 | case -EPROTO: |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 693 | skb->transport_header = skb->network_header; |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 694 | error = 0; |
Simon Horman | c30da49 | 2015-08-29 09:02:21 +0900 | [diff] [blame] | 695 | break; |
| 696 | default: |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 697 | error = nh_len; |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 698 | } |
| 699 | return error; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 700 | } |
| 701 | |
| 702 | if (key->ip.frag == OVS_FRAG_TYPE_LATER) |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 703 | return 0; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 704 | if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
| 705 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
| 706 | |
| 707 | /* Transport layer. */ |
| 708 | if (key->ip.proto == NEXTHDR_TCP) { |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 709 | if (tcphdr_ok(skb)) { |
| 710 | struct tcphdr *tcp = tcp_hdr(skb); |
Jarno Rajahalme | 1139e24 | 2014-05-05 09:54:49 -0700 | [diff] [blame] | 711 | key->tp.src = tcp->source; |
| 712 | key->tp.dst = tcp->dest; |
| 713 | key->tp.flags = TCP_FLAGS_BE16(tcp); |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 714 | } else { |
| 715 | memset(&key->tp, 0, sizeof(key->tp)); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 716 | } |
| 717 | } else if (key->ip.proto == NEXTHDR_UDP) { |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 718 | if (udphdr_ok(skb)) { |
| 719 | struct udphdr *udp = udp_hdr(skb); |
Jarno Rajahalme | 1139e24 | 2014-05-05 09:54:49 -0700 | [diff] [blame] | 720 | key->tp.src = udp->source; |
| 721 | key->tp.dst = udp->dest; |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 722 | } else { |
| 723 | memset(&key->tp, 0, sizeof(key->tp)); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 724 | } |
Joe Stringer | a175a72 | 2013-08-22 12:30:48 -0700 | [diff] [blame] | 725 | } else if (key->ip.proto == NEXTHDR_SCTP) { |
| 726 | if (sctphdr_ok(skb)) { |
| 727 | struct sctphdr *sctp = sctp_hdr(skb); |
Jarno Rajahalme | 1139e24 | 2014-05-05 09:54:49 -0700 | [diff] [blame] | 728 | key->tp.src = sctp->source; |
| 729 | key->tp.dst = sctp->dest; |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 730 | } else { |
| 731 | memset(&key->tp, 0, sizeof(key->tp)); |
Joe Stringer | a175a72 | 2013-08-22 12:30:48 -0700 | [diff] [blame] | 732 | } |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 733 | } else if (key->ip.proto == NEXTHDR_ICMP) { |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 734 | if (icmp6hdr_ok(skb)) { |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 735 | error = parse_icmpv6(skb, key, nh_len); |
| 736 | if (error) |
| 737 | return error; |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 738 | } else { |
| 739 | memset(&key->tp, 0, sizeof(key->tp)); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 740 | } |
| 741 | } |
| 742 | } |
Andy Zhou | 03f0d91 | 2013-08-07 20:01:00 -0700 | [diff] [blame] | 743 | return 0; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 744 | } |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 745 | |
Andy Zhou | 971427f3 | 2014-09-15 19:37:25 -0700 | [diff] [blame] | 746 | int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key) |
| 747 | { |
Yi-Hung Wei | 6f56f61 | 2017-03-30 12:36:03 -0700 | [diff] [blame] | 748 | int res; |
| 749 | |
| 750 | res = key_extract(skb, key); |
| 751 | if (!res) |
| 752 | key->mac_proto &= ~SW_FLOW_KEY_INVALID; |
| 753 | |
| 754 | return res; |
Andy Zhou | 971427f3 | 2014-09-15 19:37:25 -0700 | [diff] [blame] | 755 | } |
| 756 | |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 757 | static int key_extract_mac_proto(struct sk_buff *skb) |
| 758 | { |
| 759 | switch (skb->dev->type) { |
| 760 | case ARPHRD_ETHER: |
| 761 | return MAC_PROTO_ETHERNET; |
| 762 | case ARPHRD_NONE: |
| 763 | if (skb->protocol == htons(ETH_P_TEB)) |
| 764 | return MAC_PROTO_ETHERNET; |
| 765 | return MAC_PROTO_NONE; |
| 766 | } |
| 767 | WARN_ON_ONCE(1); |
| 768 | return -EINVAL; |
| 769 | } |
| 770 | |
Thomas Graf | 1d8fff9 | 2015-07-21 10:43:54 +0200 | [diff] [blame] | 771 | int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info, |
Pravin B Shelar | 8c8b1b8 | 2014-09-15 19:28:44 -0700 | [diff] [blame] | 772 | struct sk_buff *skb, struct sw_flow_key *key) |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 773 | { |
Jarno Rajahalme | 9dd7f89 | 2017-02-09 11:21:59 -0800 | [diff] [blame] | 774 | int res, err; |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 775 | |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 776 | /* Extract metadata from packet. */ |
Jesse Gross | f579668 | 2014-10-03 15:35:33 -0700 | [diff] [blame] | 777 | if (tun_info) { |
Jiri Benc | 00a93ba | 2015-10-05 13:09:46 +0200 | [diff] [blame] | 778 | key->tun_proto = ip_tunnel_info_af(tun_info); |
Thomas Graf | 1d8fff9 | 2015-07-21 10:43:54 +0200 | [diff] [blame] | 779 | memcpy(&key->tun_key, &tun_info->key, sizeof(key->tun_key)); |
Jesse Gross | f579668 | 2014-10-03 15:35:33 -0700 | [diff] [blame] | 780 | |
Pravin B Shelar | 4c22279 | 2015-08-30 18:09:38 -0700 | [diff] [blame] | 781 | if (tun_info->options_len) { |
Jesse Gross | f579668 | 2014-10-03 15:35:33 -0700 | [diff] [blame] | 782 | BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) * |
| 783 | 8)) - 1 |
| 784 | > sizeof(key->tun_opts)); |
Pravin B Shelar | 4c22279 | 2015-08-30 18:09:38 -0700 | [diff] [blame] | 785 | |
| 786 | ip_tunnel_info_opts_get(TUN_METADATA_OPTS(key, tun_info->options_len), |
| 787 | tun_info); |
Jesse Gross | f579668 | 2014-10-03 15:35:33 -0700 | [diff] [blame] | 788 | key->tun_opts_len = tun_info->options_len; |
| 789 | } else { |
| 790 | key->tun_opts_len = 0; |
| 791 | } |
| 792 | } else { |
Jiri Benc | 00a93ba | 2015-10-05 13:09:46 +0200 | [diff] [blame] | 793 | key->tun_proto = 0; |
Jesse Gross | f579668 | 2014-10-03 15:35:33 -0700 | [diff] [blame] | 794 | key->tun_opts_len = 0; |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 795 | memset(&key->tun_key, 0, sizeof(key->tun_key)); |
Jesse Gross | f579668 | 2014-10-03 15:35:33 -0700 | [diff] [blame] | 796 | } |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 797 | |
| 798 | key->phy.priority = skb->priority; |
| 799 | key->phy.in_port = OVS_CB(skb)->input_vport->port_no; |
| 800 | key->phy.skb_mark = skb->mark; |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 801 | key->ovs_flow_hash = 0; |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 802 | res = key_extract_mac_proto(skb); |
| 803 | if (res < 0) |
| 804 | return res; |
| 805 | key->mac_proto = res; |
Jesse Gross | 0714812 | 2014-10-03 15:35:29 -0700 | [diff] [blame] | 806 | key->recirc_id = 0; |
| 807 | |
Jarno Rajahalme | 9dd7f89 | 2017-02-09 11:21:59 -0800 | [diff] [blame] | 808 | err = key_extract(skb, key); |
| 809 | if (!err) |
| 810 | ovs_ct_fill_key(skb, key); /* Must be after key_extract(). */ |
| 811 | return err; |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 812 | } |
| 813 | |
Joe Stringer | c2ac667 | 2015-08-26 11:31:52 -0700 | [diff] [blame] | 814 | int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr, |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 815 | struct sk_buff *skb, |
Jarno Rajahalme | 05da589 | 2014-11-06 07:03:05 -0800 | [diff] [blame] | 816 | struct sw_flow_key *key, bool log) |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 817 | { |
Jarno Rajahalme | 9dd7f89 | 2017-02-09 11:21:59 -0800 | [diff] [blame] | 818 | const struct nlattr *a[OVS_KEY_ATTR_MAX + 1]; |
| 819 | u64 attrs = 0; |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 820 | int err; |
| 821 | |
Jarno Rajahalme | 9dd7f89 | 2017-02-09 11:21:59 -0800 | [diff] [blame] | 822 | err = parse_flow_nlattrs(attr, a, &attrs, log); |
| 823 | if (err) |
| 824 | return -EINVAL; |
| 825 | |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 826 | /* Extract metadata from netlink attributes. */ |
Jarno Rajahalme | 9dd7f89 | 2017-02-09 11:21:59 -0800 | [diff] [blame] | 827 | err = ovs_nla_get_flow_metadata(net, a, attrs, key, log); |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 828 | if (err) |
| 829 | return err; |
| 830 | |
pravin shelar | df30f74 | 2016-12-26 08:31:27 -0800 | [diff] [blame] | 831 | /* key_extract assumes that skb->protocol is set-up for |
| 832 | * layer 3 packets which is the case for other callers, |
| 833 | * in particular packets received from the network stack. |
| 834 | * Here the correct value can be set from the metadata |
| 835 | * extracted above. |
| 836 | * For L2 packet key eth type would be zero. skb protocol |
| 837 | * would be set to correct value later during key-extact. |
| 838 | */ |
Jiri Benc | 5108bba | 2016-11-10 16:28:21 +0100 | [diff] [blame] | 839 | |
pravin shelar | df30f74 | 2016-12-26 08:31:27 -0800 | [diff] [blame] | 840 | skb->protocol = key->eth.type; |
Jarno Rajahalme | 9dd7f89 | 2017-02-09 11:21:59 -0800 | [diff] [blame] | 841 | err = key_extract(skb, key); |
| 842 | if (err) |
| 843 | return err; |
| 844 | |
| 845 | /* Check that we have conntrack original direction tuple metadata only |
| 846 | * for packets for which it makes sense. Otherwise the key may be |
| 847 | * corrupted due to overlapping key fields. |
| 848 | */ |
| 849 | if (attrs & (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4) && |
| 850 | key->eth.type != htons(ETH_P_IP)) |
| 851 | return -EINVAL; |
| 852 | if (attrs & (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6) && |
| 853 | (key->eth.type != htons(ETH_P_IPV6) || |
| 854 | sw_flow_key_is_nd(key))) |
| 855 | return -EINVAL; |
| 856 | |
| 857 | return 0; |
Pravin B Shelar | 83c8df2 | 2014-09-15 19:20:31 -0700 | [diff] [blame] | 858 | } |