Back to home page

OSCL-LXR
 
 

     

0001 // SPDX-License-Identifier: GPL-2.0-only
0002 /*
0003  * Copyright (c) 2007-2014 Nicira, Inc.
0004  */
0005 
0006 #include <linux/uaccess.h>
0007 #include <linux/netdevice.h>
0008 #include <linux/etherdevice.h>
0009 #include <linux/if_ether.h>
0010 #include <linux/if_vlan.h>
0011 #include <net/llc_pdu.h>
0012 #include <linux/kernel.h>
0013 #include <linux/jhash.h>
0014 #include <linux/jiffies.h>
0015 #include <linux/llc.h>
0016 #include <linux/module.h>
0017 #include <linux/in.h>
0018 #include <linux/rcupdate.h>
0019 #include <linux/cpumask.h>
0020 #include <linux/if_arp.h>
0021 #include <linux/ip.h>
0022 #include <linux/ipv6.h>
0023 #include <linux/mpls.h>
0024 #include <linux/sctp.h>
0025 #include <linux/smp.h>
0026 #include <linux/tcp.h>
0027 #include <linux/udp.h>
0028 #include <linux/icmp.h>
0029 #include <linux/icmpv6.h>
0030 #include <linux/rculist.h>
0031 #include <net/ip.h>
0032 #include <net/ip_tunnels.h>
0033 #include <net/ipv6.h>
0034 #include <net/mpls.h>
0035 #include <net/ndisc.h>
0036 #include <net/nsh.h>
0037 #include <net/pkt_cls.h>
0038 #include <net/netfilter/nf_conntrack_zones.h>
0039 
0040 #include "conntrack.h"
0041 #include "datapath.h"
0042 #include "flow.h"
0043 #include "flow_netlink.h"
0044 #include "vport.h"
0045 
0046 u64 ovs_flow_used_time(unsigned long flow_jiffies)
0047 {
0048     struct timespec64 cur_ts;
0049     u64 cur_ms, idle_ms;
0050 
0051     ktime_get_ts64(&cur_ts);
0052     idle_ms = jiffies_to_msecs(jiffies - flow_jiffies);
0053     cur_ms = (u64)(u32)cur_ts.tv_sec * MSEC_PER_SEC +
0054          cur_ts.tv_nsec / NSEC_PER_MSEC;
0055 
0056     return cur_ms - idle_ms;
0057 }
0058 
0059 #define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF))
0060 
0061 void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags,
0062                const struct sk_buff *skb)
0063 {
0064     struct sw_flow_stats *stats;
0065     unsigned int cpu = smp_processor_id();
0066     int len = skb->len + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
0067 
0068     stats = rcu_dereference(flow->stats[cpu]);
0069 
0070     /* Check if already have CPU-specific stats. */
0071     if (likely(stats)) {
0072         spin_lock(&stats->lock);
0073         /* Mark if we write on the pre-allocated stats. */
0074         if (cpu == 0 && unlikely(flow->stats_last_writer != cpu))
0075             flow->stats_last_writer = cpu;
0076     } else {
0077         stats = rcu_dereference(flow->stats[0]); /* Pre-allocated. */
0078         spin_lock(&stats->lock);
0079 
0080         /* If the current CPU is the only writer on the
0081          * pre-allocated stats keep using them.
0082          */
0083         if (unlikely(flow->stats_last_writer != cpu)) {
0084             /* A previous locker may have already allocated the
0085              * stats, so we need to check again.  If CPU-specific
0086              * stats were already allocated, we update the pre-
0087              * allocated stats as we have already locked them.
0088              */
0089             if (likely(flow->stats_last_writer != -1) &&
0090                 likely(!rcu_access_pointer(flow->stats[cpu]))) {
0091                 /* Try to allocate CPU-specific stats. */
0092                 struct sw_flow_stats *new_stats;
0093 
0094                 new_stats =
0095                     kmem_cache_alloc_node(flow_stats_cache,
0096                                   GFP_NOWAIT |
0097                                   __GFP_THISNODE |
0098                                   __GFP_NOWARN |
0099                                   __GFP_NOMEMALLOC,
0100                                   numa_node_id());
0101                 if (likely(new_stats)) {
0102                     new_stats->used = jiffies;
0103                     new_stats->packet_count = 1;
0104                     new_stats->byte_count = len;
0105                     new_stats->tcp_flags = tcp_flags;
0106                     spin_lock_init(&new_stats->lock);
0107 
0108                     rcu_assign_pointer(flow->stats[cpu],
0109                                new_stats);
0110                     cpumask_set_cpu(cpu, &flow->cpu_used_mask);
0111                     goto unlock;
0112                 }
0113             }
0114             flow->stats_last_writer = cpu;
0115         }
0116     }
0117 
0118     stats->used = jiffies;
0119     stats->packet_count++;
0120     stats->byte_count += len;
0121     stats->tcp_flags |= tcp_flags;
0122 unlock:
0123     spin_unlock(&stats->lock);
0124 }
0125 
0126 /* Must be called with rcu_read_lock or ovs_mutex. */
0127 void ovs_flow_stats_get(const struct sw_flow *flow,
0128             struct ovs_flow_stats *ovs_stats,
0129             unsigned long *used, __be16 *tcp_flags)
0130 {
0131     int cpu;
0132 
0133     *used = 0;
0134     *tcp_flags = 0;
0135     memset(ovs_stats, 0, sizeof(*ovs_stats));
0136 
0137     /* We open code this to make sure cpu 0 is always considered */
0138     for (cpu = 0; cpu < nr_cpu_ids; cpu = cpumask_next(cpu, &flow->cpu_used_mask)) {
0139         struct sw_flow_stats *stats = rcu_dereference_ovsl(flow->stats[cpu]);
0140 
0141         if (stats) {
0142             /* Local CPU may write on non-local stats, so we must
0143              * block bottom-halves here.
0144              */
0145             spin_lock_bh(&stats->lock);
0146             if (!*used || time_after(stats->used, *used))
0147                 *used = stats->used;
0148             *tcp_flags |= stats->tcp_flags;
0149             ovs_stats->n_packets += stats->packet_count;
0150             ovs_stats->n_bytes += stats->byte_count;
0151             spin_unlock_bh(&stats->lock);
0152         }
0153     }
0154 }
0155 
0156 /* Called with ovs_mutex. */
0157 void ovs_flow_stats_clear(struct sw_flow *flow)
0158 {
0159     int cpu;
0160 
0161     /* We open code this to make sure cpu 0 is always considered */
0162     for (cpu = 0; cpu < nr_cpu_ids; cpu = cpumask_next(cpu, &flow->cpu_used_mask)) {
0163         struct sw_flow_stats *stats = ovsl_dereference(flow->stats[cpu]);
0164 
0165         if (stats) {
0166             spin_lock_bh(&stats->lock);
0167             stats->used = 0;
0168             stats->packet_count = 0;
0169             stats->byte_count = 0;
0170             stats->tcp_flags = 0;
0171             spin_unlock_bh(&stats->lock);
0172         }
0173     }
0174 }
0175 
0176 static int check_header(struct sk_buff *skb, int len)
0177 {
0178     if (unlikely(skb->len < len))
0179         return -EINVAL;
0180     if (unlikely(!pskb_may_pull(skb, len)))
0181         return -ENOMEM;
0182     return 0;
0183 }
0184 
0185 static bool arphdr_ok(struct sk_buff *skb)
0186 {
0187     return pskb_may_pull(skb, skb_network_offset(skb) +
0188                   sizeof(struct arp_eth_header));
0189 }
0190 
0191 static int check_iphdr(struct sk_buff *skb)
0192 {
0193     unsigned int nh_ofs = skb_network_offset(skb);
0194     unsigned int ip_len;
0195     int err;
0196 
0197     err = check_header(skb, nh_ofs + sizeof(struct iphdr));
0198     if (unlikely(err))
0199         return err;
0200 
0201     ip_len = ip_hdrlen(skb);
0202     if (unlikely(ip_len < sizeof(struct iphdr) ||
0203              skb->len < nh_ofs + ip_len))
0204         return -EINVAL;
0205 
0206     skb_set_transport_header(skb, nh_ofs + ip_len);
0207     return 0;
0208 }
0209 
0210 static bool tcphdr_ok(struct sk_buff *skb)
0211 {
0212     int th_ofs = skb_transport_offset(skb);
0213     int tcp_len;
0214 
0215     if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr))))
0216         return false;
0217 
0218     tcp_len = tcp_hdrlen(skb);
0219     if (unlikely(tcp_len < sizeof(struct tcphdr) ||
0220              skb->len < th_ofs + tcp_len))
0221         return false;
0222 
0223     return true;
0224 }
0225 
0226 static bool udphdr_ok(struct sk_buff *skb)
0227 {
0228     return pskb_may_pull(skb, skb_transport_offset(skb) +
0229                   sizeof(struct udphdr));
0230 }
0231 
0232 static bool sctphdr_ok(struct sk_buff *skb)
0233 {
0234     return pskb_may_pull(skb, skb_transport_offset(skb) +
0235                   sizeof(struct sctphdr));
0236 }
0237 
0238 static bool icmphdr_ok(struct sk_buff *skb)
0239 {
0240     return pskb_may_pull(skb, skb_transport_offset(skb) +
0241                   sizeof(struct icmphdr));
0242 }
0243 
0244 /**
0245  * get_ipv6_ext_hdrs() - Parses packet and sets IPv6 extension header flags.
0246  *
0247  * @skb: buffer where extension header data starts in packet
0248  * @nh: ipv6 header
0249  * @ext_hdrs: flags are stored here
0250  *
0251  * OFPIEH12_UNREP is set if more than one of a given IPv6 extension header
0252  * is unexpectedly encountered. (Two destination options headers may be
0253  * expected and would not cause this bit to be set.)
0254  *
0255  * OFPIEH12_UNSEQ is set if IPv6 extension headers were not in the order
0256  * preferred (but not required) by RFC 2460:
0257  *
0258  * When more than one extension header is used in the same packet, it is
0259  * recommended that those headers appear in the following order:
0260  *      IPv6 header
0261  *      Hop-by-Hop Options header
0262  *      Destination Options header
0263  *      Routing header
0264  *      Fragment header
0265  *      Authentication header
0266  *      Encapsulating Security Payload header
0267  *      Destination Options header
0268  *      upper-layer header
0269  */
0270 static void get_ipv6_ext_hdrs(struct sk_buff *skb, struct ipv6hdr *nh,
0271                   u16 *ext_hdrs)
0272 {
0273     u8 next_type = nh->nexthdr;
0274     unsigned int start = skb_network_offset(skb) + sizeof(struct ipv6hdr);
0275     int dest_options_header_count = 0;
0276 
0277     *ext_hdrs = 0;
0278 
0279     while (ipv6_ext_hdr(next_type)) {
0280         struct ipv6_opt_hdr _hdr, *hp;
0281 
0282         switch (next_type) {
0283         case IPPROTO_NONE:
0284             *ext_hdrs |= OFPIEH12_NONEXT;
0285             /* stop parsing */
0286             return;
0287 
0288         case IPPROTO_ESP:
0289             if (*ext_hdrs & OFPIEH12_ESP)
0290                 *ext_hdrs |= OFPIEH12_UNREP;
0291             if ((*ext_hdrs & ~(OFPIEH12_HOP | OFPIEH12_DEST |
0292                        OFPIEH12_ROUTER | IPPROTO_FRAGMENT |
0293                        OFPIEH12_AUTH | OFPIEH12_UNREP)) ||
0294                 dest_options_header_count >= 2) {
0295                 *ext_hdrs |= OFPIEH12_UNSEQ;
0296             }
0297             *ext_hdrs |= OFPIEH12_ESP;
0298             break;
0299 
0300         case IPPROTO_AH:
0301             if (*ext_hdrs & OFPIEH12_AUTH)
0302                 *ext_hdrs |= OFPIEH12_UNREP;
0303             if ((*ext_hdrs &
0304                  ~(OFPIEH12_HOP | OFPIEH12_DEST | OFPIEH12_ROUTER |
0305                    IPPROTO_FRAGMENT | OFPIEH12_UNREP)) ||
0306                 dest_options_header_count >= 2) {
0307                 *ext_hdrs |= OFPIEH12_UNSEQ;
0308             }
0309             *ext_hdrs |= OFPIEH12_AUTH;
0310             break;
0311 
0312         case IPPROTO_DSTOPTS:
0313             if (dest_options_header_count == 0) {
0314                 if (*ext_hdrs &
0315                     ~(OFPIEH12_HOP | OFPIEH12_UNREP))
0316                     *ext_hdrs |= OFPIEH12_UNSEQ;
0317                 *ext_hdrs |= OFPIEH12_DEST;
0318             } else if (dest_options_header_count == 1) {
0319                 if (*ext_hdrs &
0320                     ~(OFPIEH12_HOP | OFPIEH12_DEST |
0321                       OFPIEH12_ROUTER | OFPIEH12_FRAG |
0322                       OFPIEH12_AUTH | OFPIEH12_ESP |
0323                       OFPIEH12_UNREP)) {
0324                     *ext_hdrs |= OFPIEH12_UNSEQ;
0325                 }
0326             } else {
0327                 *ext_hdrs |= OFPIEH12_UNREP;
0328             }
0329             dest_options_header_count++;
0330             break;
0331 
0332         case IPPROTO_FRAGMENT:
0333             if (*ext_hdrs & OFPIEH12_FRAG)
0334                 *ext_hdrs |= OFPIEH12_UNREP;
0335             if ((*ext_hdrs & ~(OFPIEH12_HOP |
0336                        OFPIEH12_DEST |
0337                        OFPIEH12_ROUTER |
0338                        OFPIEH12_UNREP)) ||
0339                 dest_options_header_count >= 2) {
0340                 *ext_hdrs |= OFPIEH12_UNSEQ;
0341             }
0342             *ext_hdrs |= OFPIEH12_FRAG;
0343             break;
0344 
0345         case IPPROTO_ROUTING:
0346             if (*ext_hdrs & OFPIEH12_ROUTER)
0347                 *ext_hdrs |= OFPIEH12_UNREP;
0348             if ((*ext_hdrs & ~(OFPIEH12_HOP |
0349                        OFPIEH12_DEST |
0350                        OFPIEH12_UNREP)) ||
0351                 dest_options_header_count >= 2) {
0352                 *ext_hdrs |= OFPIEH12_UNSEQ;
0353             }
0354             *ext_hdrs |= OFPIEH12_ROUTER;
0355             break;
0356 
0357         case IPPROTO_HOPOPTS:
0358             if (*ext_hdrs & OFPIEH12_HOP)
0359                 *ext_hdrs |= OFPIEH12_UNREP;
0360             /* OFPIEH12_HOP is set to 1 if a hop-by-hop IPv6
0361              * extension header is present as the first
0362              * extension header in the packet.
0363              */
0364             if (*ext_hdrs == 0)
0365                 *ext_hdrs |= OFPIEH12_HOP;
0366             else
0367                 *ext_hdrs |= OFPIEH12_UNSEQ;
0368             break;
0369 
0370         default:
0371             return;
0372         }
0373 
0374         hp = skb_header_pointer(skb, start, sizeof(_hdr), &_hdr);
0375         if (!hp)
0376             break;
0377         next_type = hp->nexthdr;
0378         start += ipv6_optlen(hp);
0379     }
0380 }
0381 
0382 static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key)
0383 {
0384     unsigned short frag_off;
0385     unsigned int payload_ofs = 0;
0386     unsigned int nh_ofs = skb_network_offset(skb);
0387     unsigned int nh_len;
0388     struct ipv6hdr *nh;
0389     int err, nexthdr, flags = 0;
0390 
0391     err = check_header(skb, nh_ofs + sizeof(*nh));
0392     if (unlikely(err))
0393         return err;
0394 
0395     nh = ipv6_hdr(skb);
0396 
0397     get_ipv6_ext_hdrs(skb, nh, &key->ipv6.exthdrs);
0398 
0399     key->ip.proto = NEXTHDR_NONE;
0400     key->ip.tos = ipv6_get_dsfield(nh);
0401     key->ip.ttl = nh->hop_limit;
0402     key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL);
0403     key->ipv6.addr.src = nh->saddr;
0404     key->ipv6.addr.dst = nh->daddr;
0405 
0406     nexthdr = ipv6_find_hdr(skb, &payload_ofs, -1, &frag_off, &flags);
0407     if (flags & IP6_FH_F_FRAG) {
0408         if (frag_off) {
0409             key->ip.frag = OVS_FRAG_TYPE_LATER;
0410             key->ip.proto = NEXTHDR_FRAGMENT;
0411             return 0;
0412         }
0413         key->ip.frag = OVS_FRAG_TYPE_FIRST;
0414     } else {
0415         key->ip.frag = OVS_FRAG_TYPE_NONE;
0416     }
0417 
0418     /* Delayed handling of error in ipv6_find_hdr() as it
0419      * always sets flags and frag_off to a valid value which may be
0420      * used to set key->ip.frag above.
0421      */
0422     if (unlikely(nexthdr < 0))
0423         return -EPROTO;
0424 
0425     nh_len = payload_ofs - nh_ofs;
0426     skb_set_transport_header(skb, nh_ofs + nh_len);
0427     key->ip.proto = nexthdr;
0428     return nh_len;
0429 }
0430 
0431 static bool icmp6hdr_ok(struct sk_buff *skb)
0432 {
0433     return pskb_may_pull(skb, skb_transport_offset(skb) +
0434                   sizeof(struct icmp6hdr));
0435 }
0436 
0437 /**
0438  * parse_vlan_tag - Parse vlan tag from vlan header.
0439  * @skb: skb containing frame to parse
0440  * @key_vh: pointer to parsed vlan tag
0441  * @untag_vlan: should the vlan header be removed from the frame
0442  *
0443  * Return: ERROR on memory error.
0444  * %0 if it encounters a non-vlan or incomplete packet.
0445  * %1 after successfully parsing vlan tag.
0446  */
0447 static int parse_vlan_tag(struct sk_buff *skb, struct vlan_head *key_vh,
0448               bool untag_vlan)
0449 {
0450     struct vlan_head *vh = (struct vlan_head *)skb->data;
0451 
0452     if (likely(!eth_type_vlan(vh->tpid)))
0453         return 0;
0454 
0455     if (unlikely(skb->len < sizeof(struct vlan_head) + sizeof(__be16)))
0456         return 0;
0457 
0458     if (unlikely(!pskb_may_pull(skb, sizeof(struct vlan_head) +
0459                  sizeof(__be16))))
0460         return -ENOMEM;
0461 
0462     vh = (struct vlan_head *)skb->data;
0463     key_vh->tci = vh->tci | htons(VLAN_CFI_MASK);
0464     key_vh->tpid = vh->tpid;
0465 
0466     if (unlikely(untag_vlan)) {
0467         int offset = skb->data - skb_mac_header(skb);
0468         u16 tci;
0469         int err;
0470 
0471         __skb_push(skb, offset);
0472         err = __skb_vlan_pop(skb, &tci);
0473         __skb_pull(skb, offset);
0474         if (err)
0475             return err;
0476         __vlan_hwaccel_put_tag(skb, key_vh->tpid, tci);
0477     } else {
0478         __skb_pull(skb, sizeof(struct vlan_head));
0479     }
0480     return 1;
0481 }
0482 
0483 static void clear_vlan(struct sw_flow_key *key)
0484 {
0485     key->eth.vlan.tci = 0;
0486     key->eth.vlan.tpid = 0;
0487     key->eth.cvlan.tci = 0;
0488     key->eth.cvlan.tpid = 0;
0489 }
0490 
0491 static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key)
0492 {
0493     int res;
0494 
0495     if (skb_vlan_tag_present(skb)) {
0496         key->eth.vlan.tci = htons(skb->vlan_tci) | htons(VLAN_CFI_MASK);
0497         key->eth.vlan.tpid = skb->vlan_proto;
0498     } else {
0499         /* Parse outer vlan tag in the non-accelerated case. */
0500         res = parse_vlan_tag(skb, &key->eth.vlan, true);
0501         if (res <= 0)
0502             return res;
0503     }
0504 
0505     /* Parse inner vlan tag. */
0506     res = parse_vlan_tag(skb, &key->eth.cvlan, false);
0507     if (res <= 0)
0508         return res;
0509 
0510     return 0;
0511 }
0512 
0513 static __be16 parse_ethertype(struct sk_buff *skb)
0514 {
0515     struct llc_snap_hdr {
0516         u8  dsap;  /* Always 0xAA */
0517         u8  ssap;  /* Always 0xAA */
0518         u8  ctrl;
0519         u8  oui[3];
0520         __be16 ethertype;
0521     };
0522     struct llc_snap_hdr *llc;
0523     __be16 proto;
0524 
0525     proto = *(__be16 *) skb->data;
0526     __skb_pull(skb, sizeof(__be16));
0527 
0528     if (eth_proto_is_802_3(proto))
0529         return proto;
0530 
0531     if (skb->len < sizeof(struct llc_snap_hdr))
0532         return htons(ETH_P_802_2);
0533 
0534     if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr))))
0535         return htons(0);
0536 
0537     llc = (struct llc_snap_hdr *) skb->data;
0538     if (llc->dsap != LLC_SAP_SNAP ||
0539         llc->ssap != LLC_SAP_SNAP ||
0540         (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0)
0541         return htons(ETH_P_802_2);
0542 
0543     __skb_pull(skb, sizeof(struct llc_snap_hdr));
0544 
0545     if (eth_proto_is_802_3(llc->ethertype))
0546         return llc->ethertype;
0547 
0548     return htons(ETH_P_802_2);
0549 }
0550 
0551 static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
0552             int nh_len)
0553 {
0554     struct icmp6hdr *icmp = icmp6_hdr(skb);
0555 
0556     /* The ICMPv6 type and code fields use the 16-bit transport port
0557      * fields, so we need to store them in 16-bit network byte order.
0558      */
0559     key->tp.src = htons(icmp->icmp6_type);
0560     key->tp.dst = htons(icmp->icmp6_code);
0561     memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd));
0562 
0563     if (icmp->icmp6_code == 0 &&
0564         (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION ||
0565          icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) {
0566         int icmp_len = skb->len - skb_transport_offset(skb);
0567         struct nd_msg *nd;
0568         int offset;
0569 
0570         /* In order to process neighbor discovery options, we need the
0571          * entire packet.
0572          */
0573         if (unlikely(icmp_len < sizeof(*nd)))
0574             return 0;
0575 
0576         if (unlikely(skb_linearize(skb)))
0577             return -ENOMEM;
0578 
0579         nd = (struct nd_msg *)skb_transport_header(skb);
0580         key->ipv6.nd.target = nd->target;
0581 
0582         icmp_len -= sizeof(*nd);
0583         offset = 0;
0584         while (icmp_len >= 8) {
0585             struct nd_opt_hdr *nd_opt =
0586                  (struct nd_opt_hdr *)(nd->opt + offset);
0587             int opt_len = nd_opt->nd_opt_len * 8;
0588 
0589             if (unlikely(!opt_len || opt_len > icmp_len))
0590                 return 0;
0591 
0592             /* Store the link layer address if the appropriate
0593              * option is provided.  It is considered an error if
0594              * the same link layer option is specified twice.
0595              */
0596             if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR
0597                 && opt_len == 8) {
0598                 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll)))
0599                     goto invalid;
0600                 ether_addr_copy(key->ipv6.nd.sll,
0601                         &nd->opt[offset+sizeof(*nd_opt)]);
0602             } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR
0603                    && opt_len == 8) {
0604                 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll)))
0605                     goto invalid;
0606                 ether_addr_copy(key->ipv6.nd.tll,
0607                         &nd->opt[offset+sizeof(*nd_opt)]);
0608             }
0609 
0610             icmp_len -= opt_len;
0611             offset += opt_len;
0612         }
0613     }
0614 
0615     return 0;
0616 
0617 invalid:
0618     memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target));
0619     memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll));
0620     memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll));
0621 
0622     return 0;
0623 }
0624 
0625 static int parse_nsh(struct sk_buff *skb, struct sw_flow_key *key)
0626 {
0627     struct nshhdr *nh;
0628     unsigned int nh_ofs = skb_network_offset(skb);
0629     u8 version, length;
0630     int err;
0631 
0632     err = check_header(skb, nh_ofs + NSH_BASE_HDR_LEN);
0633     if (unlikely(err))
0634         return err;
0635 
0636     nh = nsh_hdr(skb);
0637     version = nsh_get_ver(nh);
0638     length = nsh_hdr_len(nh);
0639 
0640     if (version != 0)
0641         return -EINVAL;
0642 
0643     err = check_header(skb, nh_ofs + length);
0644     if (unlikely(err))
0645         return err;
0646 
0647     nh = nsh_hdr(skb);
0648     key->nsh.base.flags = nsh_get_flags(nh);
0649     key->nsh.base.ttl = nsh_get_ttl(nh);
0650     key->nsh.base.mdtype = nh->mdtype;
0651     key->nsh.base.np = nh->np;
0652     key->nsh.base.path_hdr = nh->path_hdr;
0653     switch (key->nsh.base.mdtype) {
0654     case NSH_M_TYPE1:
0655         if (length != NSH_M_TYPE1_LEN)
0656             return -EINVAL;
0657         memcpy(key->nsh.context, nh->md1.context,
0658                sizeof(nh->md1));
0659         break;
0660     case NSH_M_TYPE2:
0661         memset(key->nsh.context, 0,
0662                sizeof(nh->md1));
0663         break;
0664     default:
0665         return -EINVAL;
0666     }
0667 
0668     return 0;
0669 }
0670 
0671 /**
0672  * key_extract_l3l4 - extracts L3/L4 header information.
0673  * @skb: sk_buff that contains the frame, with skb->data pointing to the
0674  *       L3 header
0675  * @key: output flow key
0676  *
0677  * Return: %0 if successful, otherwise a negative errno value.
0678  */
0679 static int key_extract_l3l4(struct sk_buff *skb, struct sw_flow_key *key)
0680 {
0681     int error;
0682 
0683     /* Network layer. */
0684     if (key->eth.type == htons(ETH_P_IP)) {
0685         struct iphdr *nh;
0686         __be16 offset;
0687 
0688         error = check_iphdr(skb);
0689         if (unlikely(error)) {
0690             memset(&key->ip, 0, sizeof(key->ip));
0691             memset(&key->ipv4, 0, sizeof(key->ipv4));
0692             if (error == -EINVAL) {
0693                 skb->transport_header = skb->network_header;
0694                 error = 0;
0695             }
0696             return error;
0697         }
0698 
0699         nh = ip_hdr(skb);
0700         key->ipv4.addr.src = nh->saddr;
0701         key->ipv4.addr.dst = nh->daddr;
0702 
0703         key->ip.proto = nh->protocol;
0704         key->ip.tos = nh->tos;
0705         key->ip.ttl = nh->ttl;
0706 
0707         offset = nh->frag_off & htons(IP_OFFSET);
0708         if (offset) {
0709             key->ip.frag = OVS_FRAG_TYPE_LATER;
0710             memset(&key->tp, 0, sizeof(key->tp));
0711             return 0;
0712         }
0713         if (nh->frag_off & htons(IP_MF) ||
0714             skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
0715             key->ip.frag = OVS_FRAG_TYPE_FIRST;
0716         else
0717             key->ip.frag = OVS_FRAG_TYPE_NONE;
0718 
0719         /* Transport layer. */
0720         if (key->ip.proto == IPPROTO_TCP) {
0721             if (tcphdr_ok(skb)) {
0722                 struct tcphdr *tcp = tcp_hdr(skb);
0723                 key->tp.src = tcp->source;
0724                 key->tp.dst = tcp->dest;
0725                 key->tp.flags = TCP_FLAGS_BE16(tcp);
0726             } else {
0727                 memset(&key->tp, 0, sizeof(key->tp));
0728             }
0729 
0730         } else if (key->ip.proto == IPPROTO_UDP) {
0731             if (udphdr_ok(skb)) {
0732                 struct udphdr *udp = udp_hdr(skb);
0733                 key->tp.src = udp->source;
0734                 key->tp.dst = udp->dest;
0735             } else {
0736                 memset(&key->tp, 0, sizeof(key->tp));
0737             }
0738         } else if (key->ip.proto == IPPROTO_SCTP) {
0739             if (sctphdr_ok(skb)) {
0740                 struct sctphdr *sctp = sctp_hdr(skb);
0741                 key->tp.src = sctp->source;
0742                 key->tp.dst = sctp->dest;
0743             } else {
0744                 memset(&key->tp, 0, sizeof(key->tp));
0745             }
0746         } else if (key->ip.proto == IPPROTO_ICMP) {
0747             if (icmphdr_ok(skb)) {
0748                 struct icmphdr *icmp = icmp_hdr(skb);
0749                 /* The ICMP type and code fields use the 16-bit
0750                  * transport port fields, so we need to store
0751                  * them in 16-bit network byte order. */
0752                 key->tp.src = htons(icmp->type);
0753                 key->tp.dst = htons(icmp->code);
0754             } else {
0755                 memset(&key->tp, 0, sizeof(key->tp));
0756             }
0757         }
0758 
0759     } else if (key->eth.type == htons(ETH_P_ARP) ||
0760            key->eth.type == htons(ETH_P_RARP)) {
0761         struct arp_eth_header *arp;
0762         bool arp_available = arphdr_ok(skb);
0763 
0764         arp = (struct arp_eth_header *)skb_network_header(skb);
0765 
0766         if (arp_available &&
0767             arp->ar_hrd == htons(ARPHRD_ETHER) &&
0768             arp->ar_pro == htons(ETH_P_IP) &&
0769             arp->ar_hln == ETH_ALEN &&
0770             arp->ar_pln == 4) {
0771 
0772             /* We only match on the lower 8 bits of the opcode. */
0773             if (ntohs(arp->ar_op) <= 0xff)
0774                 key->ip.proto = ntohs(arp->ar_op);
0775             else
0776                 key->ip.proto = 0;
0777 
0778             memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src));
0779             memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst));
0780             ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha);
0781             ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha);
0782         } else {
0783             memset(&key->ip, 0, sizeof(key->ip));
0784             memset(&key->ipv4, 0, sizeof(key->ipv4));
0785         }
0786     } else if (eth_p_mpls(key->eth.type)) {
0787         u8 label_count = 1;
0788 
0789         memset(&key->mpls, 0, sizeof(key->mpls));
0790         skb_set_inner_network_header(skb, skb->mac_len);
0791         while (1) {
0792             __be32 lse;
0793 
0794             error = check_header(skb, skb->mac_len +
0795                          label_count * MPLS_HLEN);
0796             if (unlikely(error))
0797                 return 0;
0798 
0799             memcpy(&lse, skb_inner_network_header(skb), MPLS_HLEN);
0800 
0801             if (label_count <= MPLS_LABEL_DEPTH)
0802                 memcpy(&key->mpls.lse[label_count - 1], &lse,
0803                        MPLS_HLEN);
0804 
0805             skb_set_inner_network_header(skb, skb->mac_len +
0806                              label_count * MPLS_HLEN);
0807             if (lse & htonl(MPLS_LS_S_MASK))
0808                 break;
0809 
0810             label_count++;
0811         }
0812         if (label_count > MPLS_LABEL_DEPTH)
0813             label_count = MPLS_LABEL_DEPTH;
0814 
0815         key->mpls.num_labels_mask = GENMASK(label_count - 1, 0);
0816     } else if (key->eth.type == htons(ETH_P_IPV6)) {
0817         int nh_len;             /* IPv6 Header + Extensions */
0818 
0819         nh_len = parse_ipv6hdr(skb, key);
0820         if (unlikely(nh_len < 0)) {
0821             switch (nh_len) {
0822             case -EINVAL:
0823                 memset(&key->ip, 0, sizeof(key->ip));
0824                 memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr));
0825                 fallthrough;
0826             case -EPROTO:
0827                 skb->transport_header = skb->network_header;
0828                 error = 0;
0829                 break;
0830             default:
0831                 error = nh_len;
0832             }
0833             return error;
0834         }
0835 
0836         if (key->ip.frag == OVS_FRAG_TYPE_LATER) {
0837             memset(&key->tp, 0, sizeof(key->tp));
0838             return 0;
0839         }
0840         if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
0841             key->ip.frag = OVS_FRAG_TYPE_FIRST;
0842 
0843         /* Transport layer. */
0844         if (key->ip.proto == NEXTHDR_TCP) {
0845             if (tcphdr_ok(skb)) {
0846                 struct tcphdr *tcp = tcp_hdr(skb);
0847                 key->tp.src = tcp->source;
0848                 key->tp.dst = tcp->dest;
0849                 key->tp.flags = TCP_FLAGS_BE16(tcp);
0850             } else {
0851                 memset(&key->tp, 0, sizeof(key->tp));
0852             }
0853         } else if (key->ip.proto == NEXTHDR_UDP) {
0854             if (udphdr_ok(skb)) {
0855                 struct udphdr *udp = udp_hdr(skb);
0856                 key->tp.src = udp->source;
0857                 key->tp.dst = udp->dest;
0858             } else {
0859                 memset(&key->tp, 0, sizeof(key->tp));
0860             }
0861         } else if (key->ip.proto == NEXTHDR_SCTP) {
0862             if (sctphdr_ok(skb)) {
0863                 struct sctphdr *sctp = sctp_hdr(skb);
0864                 key->tp.src = sctp->source;
0865                 key->tp.dst = sctp->dest;
0866             } else {
0867                 memset(&key->tp, 0, sizeof(key->tp));
0868             }
0869         } else if (key->ip.proto == NEXTHDR_ICMP) {
0870             if (icmp6hdr_ok(skb)) {
0871                 error = parse_icmpv6(skb, key, nh_len);
0872                 if (error)
0873                     return error;
0874             } else {
0875                 memset(&key->tp, 0, sizeof(key->tp));
0876             }
0877         }
0878     } else if (key->eth.type == htons(ETH_P_NSH)) {
0879         error = parse_nsh(skb, key);
0880         if (error)
0881             return error;
0882     }
0883     return 0;
0884 }
0885 
0886 /**
0887  * key_extract - extracts a flow key from an Ethernet frame.
0888  * @skb: sk_buff that contains the frame, with skb->data pointing to the
0889  * Ethernet header
0890  * @key: output flow key
0891  *
0892  * The caller must ensure that skb->len >= ETH_HLEN.
0893  *
0894  * Initializes @skb header fields as follows:
0895  *
0896  *    - skb->mac_header: the L2 header.
0897  *
0898  *    - skb->network_header: just past the L2 header, or just past the
0899  *      VLAN header, to the first byte of the L2 payload.
0900  *
0901  *    - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6
0902  *      on output, then just past the IP header, if one is present and
0903  *      of a correct length, otherwise the same as skb->network_header.
0904  *      For other key->eth.type values it is left untouched.
0905  *
0906  *    - skb->protocol: the type of the data starting at skb->network_header.
0907  *      Equals to key->eth.type.
0908  *
0909  * Return: %0 if successful, otherwise a negative errno value.
0910  */
0911 static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
0912 {
0913     struct ethhdr *eth;
0914 
0915     /* Flags are always used as part of stats */
0916     key->tp.flags = 0;
0917 
0918     skb_reset_mac_header(skb);
0919 
0920     /* Link layer. */
0921     clear_vlan(key);
0922     if (ovs_key_mac_proto(key) == MAC_PROTO_NONE) {
0923         if (unlikely(eth_type_vlan(skb->protocol)))
0924             return -EINVAL;
0925 
0926         skb_reset_network_header(skb);
0927         key->eth.type = skb->protocol;
0928     } else {
0929         eth = eth_hdr(skb);
0930         ether_addr_copy(key->eth.src, eth->h_source);
0931         ether_addr_copy(key->eth.dst, eth->h_dest);
0932 
0933         __skb_pull(skb, 2 * ETH_ALEN);
0934         /* We are going to push all headers that we pull, so no need to
0935          * update skb->csum here.
0936          */
0937 
0938         if (unlikely(parse_vlan(skb, key)))
0939             return -ENOMEM;
0940 
0941         key->eth.type = parse_ethertype(skb);
0942         if (unlikely(key->eth.type == htons(0)))
0943             return -ENOMEM;
0944 
0945         /* Multiple tagged packets need to retain TPID to satisfy
0946          * skb_vlan_pop(), which will later shift the ethertype into
0947          * skb->protocol.
0948          */
0949         if (key->eth.cvlan.tci & htons(VLAN_CFI_MASK))
0950             skb->protocol = key->eth.cvlan.tpid;
0951         else
0952             skb->protocol = key->eth.type;
0953 
0954         skb_reset_network_header(skb);
0955         __skb_push(skb, skb->data - skb_mac_header(skb));
0956     }
0957 
0958     skb_reset_mac_len(skb);
0959 
0960     /* Fill out L3/L4 key info, if any */
0961     return key_extract_l3l4(skb, key);
0962 }
0963 
0964 /* In the case of conntrack fragment handling it expects L3 headers,
0965  * add a helper.
0966  */
0967 int ovs_flow_key_update_l3l4(struct sk_buff *skb, struct sw_flow_key *key)
0968 {
0969     return key_extract_l3l4(skb, key);
0970 }
0971 
0972 int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key)
0973 {
0974     int res;
0975 
0976     res = key_extract(skb, key);
0977     if (!res)
0978         key->mac_proto &= ~SW_FLOW_KEY_INVALID;
0979 
0980     return res;
0981 }
0982 
0983 static int key_extract_mac_proto(struct sk_buff *skb)
0984 {
0985     switch (skb->dev->type) {
0986     case ARPHRD_ETHER:
0987         return MAC_PROTO_ETHERNET;
0988     case ARPHRD_NONE:
0989         if (skb->protocol == htons(ETH_P_TEB))
0990             return MAC_PROTO_ETHERNET;
0991         return MAC_PROTO_NONE;
0992     }
0993     WARN_ON_ONCE(1);
0994     return -EINVAL;
0995 }
0996 
0997 int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
0998              struct sk_buff *skb, struct sw_flow_key *key)
0999 {
1000 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
1001     struct tc_skb_ext *tc_ext;
1002 #endif
1003     bool post_ct = false, post_ct_snat = false, post_ct_dnat = false;
1004     int res, err;
1005     u16 zone = 0;
1006 
1007     /* Extract metadata from packet. */
1008     if (tun_info) {
1009         key->tun_proto = ip_tunnel_info_af(tun_info);
1010         memcpy(&key->tun_key, &tun_info->key, sizeof(key->tun_key));
1011 
1012         if (tun_info->options_len) {
1013             BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) *
1014                            8)) - 1
1015                     > sizeof(key->tun_opts));
1016 
1017             ip_tunnel_info_opts_get(TUN_METADATA_OPTS(key, tun_info->options_len),
1018                         tun_info);
1019             key->tun_opts_len = tun_info->options_len;
1020         } else {
1021             key->tun_opts_len = 0;
1022         }
1023     } else  {
1024         key->tun_proto = 0;
1025         key->tun_opts_len = 0;
1026         memset(&key->tun_key, 0, sizeof(key->tun_key));
1027     }
1028 
1029     key->phy.priority = skb->priority;
1030     key->phy.in_port = OVS_CB(skb)->input_vport->port_no;
1031     key->phy.skb_mark = skb->mark;
1032     key->ovs_flow_hash = 0;
1033     res = key_extract_mac_proto(skb);
1034     if (res < 0)
1035         return res;
1036     key->mac_proto = res;
1037 
1038 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
1039     if (tc_skb_ext_tc_enabled()) {
1040         tc_ext = skb_ext_find(skb, TC_SKB_EXT);
1041         key->recirc_id = tc_ext ? tc_ext->chain : 0;
1042         OVS_CB(skb)->mru = tc_ext ? tc_ext->mru : 0;
1043         post_ct = tc_ext ? tc_ext->post_ct : false;
1044         post_ct_snat = post_ct ? tc_ext->post_ct_snat : false;
1045         post_ct_dnat = post_ct ? tc_ext->post_ct_dnat : false;
1046         zone = post_ct ? tc_ext->zone : 0;
1047     } else {
1048         key->recirc_id = 0;
1049     }
1050 #else
1051     key->recirc_id = 0;
1052 #endif
1053 
1054     err = key_extract(skb, key);
1055     if (!err) {
1056         ovs_ct_fill_key(skb, key, post_ct);   /* Must be after key_extract(). */
1057         if (post_ct) {
1058             if (!skb_get_nfct(skb)) {
1059                 key->ct_zone = zone;
1060             } else {
1061                 if (!post_ct_dnat)
1062                     key->ct_state &= ~OVS_CS_F_DST_NAT;
1063                 if (!post_ct_snat)
1064                     key->ct_state &= ~OVS_CS_F_SRC_NAT;
1065             }
1066         }
1067     }
1068     return err;
1069 }
1070 
1071 int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
1072                    struct sk_buff *skb,
1073                    struct sw_flow_key *key, bool log)
1074 {
1075     const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1076     u64 attrs = 0;
1077     int err;
1078 
1079     err = parse_flow_nlattrs(attr, a, &attrs, log);
1080     if (err)
1081         return -EINVAL;
1082 
1083     /* Extract metadata from netlink attributes. */
1084     err = ovs_nla_get_flow_metadata(net, a, attrs, key, log);
1085     if (err)
1086         return err;
1087 
1088     /* key_extract assumes that skb->protocol is set-up for
1089      * layer 3 packets which is the case for other callers,
1090      * in particular packets received from the network stack.
1091      * Here the correct value can be set from the metadata
1092      * extracted above.
1093      * For L2 packet key eth type would be zero. skb protocol
1094      * would be set to correct value later during key-extact.
1095      */
1096 
1097     skb->protocol = key->eth.type;
1098     err = key_extract(skb, key);
1099     if (err)
1100         return err;
1101 
1102     /* Check that we have conntrack original direction tuple metadata only
1103      * for packets for which it makes sense.  Otherwise the key may be
1104      * corrupted due to overlapping key fields.
1105      */
1106     if (attrs & (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4) &&
1107         key->eth.type != htons(ETH_P_IP))
1108         return -EINVAL;
1109     if (attrs & (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6) &&
1110         (key->eth.type != htons(ETH_P_IPV6) ||
1111          sw_flow_key_is_nd(key)))
1112         return -EINVAL;
1113 
1114     return 0;
1115 }
[ Source navigation ] [ Diff markup ] [ Identifier search ] [ general search ]

This page was automatically generated by the 2.1.0 LXR engine.

The LXR team
Valid CSS 2.1! Valid HTML 4.01!