1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * SR-IPv6 implementation
6 * David Lebrun <david.lebrun@uclouvain.be>
7 * eBPF support: Mathieu Xhonneux <m.xhonneux@gmail.com>
10 #include <linux/filter.h>
11 #include <linux/types.h>
12 #include <linux/skbuff.h>
13 #include <linux/net.h>
14 #include <linux/module.h>
16 #include <net/lwtunnel.h>
17 #include <net/netevent.h>
18 #include <net/netns/generic.h>
19 #include <net/ip6_fib.h>
20 #include <net/route.h>
22 #include <linux/seg6.h>
23 #include <linux/seg6_local.h>
24 #include <net/addrconf.h>
25 #include <net/ip6_route.h>
26 #include <net/dst_cache.h>
27 #include <net/ip_tunnels.h>
28 #ifdef CONFIG_IPV6_SEG6_HMAC
29 #include <net/seg6_hmac.h>
31 #include <net/seg6_local.h>
32 #include <linux/etherdevice.h>
33 #include <linux/bpf.h>
34 #include <linux/netfilter.h>
36 #define SEG6_F_ATTR(i) BIT(i)
38 struct seg6_local_lwt;
40 /* callbacks used for customizing the creation and destruction of a behavior */
41 struct seg6_local_lwtunnel_ops {
42 int (*build_state)(struct seg6_local_lwt *slwt, const void *cfg,
43 struct netlink_ext_ack *extack);
44 void (*destroy_state)(struct seg6_local_lwt *slwt);
47 struct seg6_action_desc {
51 /* The optattrs field is used for specifying all the optional
52 * attributes supported by a specific behavior.
53 * It means that if one of these attributes is not provided in the
54 * netlink message during the behavior creation, no errors will be
55 * returned to the userspace.
57 * Each attribute can be only of two types (mutually exclusive):
58 * 1) required or 2) optional.
59 * Every user MUST obey to this rule! If you set an attribute as
60 * required the same attribute CANNOT be set as optional and vice
63 unsigned long optattrs;
65 int (*input)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
68 struct seg6_local_lwtunnel_ops slwt_ops;
72 struct bpf_prog *prog;
76 enum seg6_end_dt_mode {
77 DT_INVALID_MODE = -EINVAL,
82 struct seg6_end_dt_info {
83 enum seg6_end_dt_mode mode;
86 /* VRF device associated to the routing table used by the SRv6
87 * End.DT4/DT6 behavior for routing IPv4/IPv6 packets.
92 /* tunneled packet family (IPv4 or IPv6).
93 * Protocol and header length are inferred from family.
98 struct pcpu_seg6_local_counters {
103 struct u64_stats_sync syncp;
106 /* This struct groups all the SRv6 Behavior counters supported so far.
108 * put_nla_counters() makes use of this data structure to collect all counter
109 * values after the per-CPU counter evaluation has been performed.
110 * Finally, each counter value (in seg6_local_counters) is stored in the
111 * corresponding netlink attribute and sent to user space.
113 * NB: we don't want to expose this structure to user space!
115 struct seg6_local_counters {
121 #define seg6_local_alloc_pcpu_counters(__gfp) \
122 __netdev_alloc_pcpu_stats(struct pcpu_seg6_local_counters, \
123 ((__gfp) | __GFP_ZERO))
125 #define SEG6_F_LOCAL_COUNTERS SEG6_F_ATTR(SEG6_LOCAL_COUNTERS)
127 struct seg6_local_lwt {
129 struct ipv6_sr_hdr *srh;
135 struct bpf_lwt_prog bpf;
136 #ifdef CONFIG_NET_L3_MASTER_DEV
137 struct seg6_end_dt_info dt_info;
139 struct pcpu_seg6_local_counters __percpu *pcpu_counters;
142 struct seg6_action_desc *desc;
143 /* unlike the required attrs, we have to track the optional attributes
144 * that have been effectively parsed.
146 unsigned long parsed_optattrs;
149 static struct seg6_local_lwt *seg6_local_lwtunnel(struct lwtunnel_state *lwt)
151 return (struct seg6_local_lwt *)lwt->data;
154 static struct ipv6_sr_hdr *get_and_validate_srh(struct sk_buff *skb)
156 struct ipv6_sr_hdr *srh;
158 srh = seg6_get_srh(skb, IP6_FH_F_SKIP_RH);
162 #ifdef CONFIG_IPV6_SEG6_HMAC
163 if (!seg6_hmac_validate_skb(skb))
170 static bool decap_and_validate(struct sk_buff *skb, int proto)
172 struct ipv6_sr_hdr *srh;
173 unsigned int off = 0;
175 srh = seg6_get_srh(skb, 0);
176 if (srh && srh->segments_left > 0)
179 #ifdef CONFIG_IPV6_SEG6_HMAC
180 if (srh && !seg6_hmac_validate_skb(skb))
184 if (ipv6_find_hdr(skb, &off, proto, NULL, NULL) < 0)
187 if (!pskb_pull(skb, off))
190 skb_postpull_rcsum(skb, skb_network_header(skb), off);
192 skb_reset_network_header(skb);
193 skb_reset_transport_header(skb);
194 if (iptunnel_pull_offloads(skb))
200 static void advance_nextseg(struct ipv6_sr_hdr *srh, struct in6_addr *daddr)
202 struct in6_addr *addr;
204 srh->segments_left--;
205 addr = srh->segments + srh->segments_left;
210 seg6_lookup_any_nexthop(struct sk_buff *skb, struct in6_addr *nhaddr,
211 u32 tbl_id, bool local_delivery)
213 struct net *net = dev_net(skb->dev);
214 struct ipv6hdr *hdr = ipv6_hdr(skb);
215 int flags = RT6_LOOKUP_F_HAS_SADDR;
216 struct dst_entry *dst = NULL;
221 memset(&fl6, 0, sizeof(fl6));
222 fl6.flowi6_iif = skb->dev->ifindex;
223 fl6.daddr = nhaddr ? *nhaddr : hdr->daddr;
224 fl6.saddr = hdr->saddr;
225 fl6.flowlabel = ip6_flowinfo(hdr);
226 fl6.flowi6_mark = skb->mark;
227 fl6.flowi6_proto = hdr->nexthdr;
230 fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH;
233 dst = ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags);
235 struct fib6_table *table;
237 table = fib6_get_table(net, tbl_id);
241 rt = ip6_pol_route(net, table, 0, &fl6, skb, flags);
245 /* we want to discard traffic destined for local packet processing,
246 * if @local_delivery is set to false.
249 dev_flags |= IFF_LOOPBACK;
251 if (dst && (dst->dev->flags & dev_flags) && !dst->error) {
258 rt = net->ipv6.ip6_blk_hole_entry;
264 skb_dst_set(skb, dst);
268 int seg6_lookup_nexthop(struct sk_buff *skb,
269 struct in6_addr *nhaddr, u32 tbl_id)
271 return seg6_lookup_any_nexthop(skb, nhaddr, tbl_id, false);
274 /* regular endpoint function */
275 static int input_action_end(struct sk_buff *skb, struct seg6_local_lwt *slwt)
277 struct ipv6_sr_hdr *srh;
279 srh = get_and_validate_srh(skb);
283 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
285 seg6_lookup_nexthop(skb, NULL, 0);
287 return dst_input(skb);
294 /* regular endpoint, and forward to specified nexthop */
295 static int input_action_end_x(struct sk_buff *skb, struct seg6_local_lwt *slwt)
297 struct ipv6_sr_hdr *srh;
299 srh = get_and_validate_srh(skb);
303 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
305 seg6_lookup_nexthop(skb, &slwt->nh6, 0);
307 return dst_input(skb);
314 static int input_action_end_t(struct sk_buff *skb, struct seg6_local_lwt *slwt)
316 struct ipv6_sr_hdr *srh;
318 srh = get_and_validate_srh(skb);
322 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
324 seg6_lookup_nexthop(skb, NULL, slwt->table);
326 return dst_input(skb);
333 /* decapsulate and forward inner L2 frame on specified interface */
334 static int input_action_end_dx2(struct sk_buff *skb,
335 struct seg6_local_lwt *slwt)
337 struct net *net = dev_net(skb->dev);
338 struct net_device *odev;
341 if (!decap_and_validate(skb, IPPROTO_ETHERNET))
344 if (!pskb_may_pull(skb, ETH_HLEN))
347 skb_reset_mac_header(skb);
348 eth = (struct ethhdr *)skb->data;
350 /* To determine the frame's protocol, we assume it is 802.3. This avoids
351 * a call to eth_type_trans(), which is not really relevant for our
354 if (!eth_proto_is_802_3(eth->h_proto))
357 odev = dev_get_by_index_rcu(net, slwt->oif);
361 /* As we accept Ethernet frames, make sure the egress device is of
364 if (odev->type != ARPHRD_ETHER)
367 if (!(odev->flags & IFF_UP) || !netif_carrier_ok(odev))
372 if (skb_warn_if_lro(skb))
375 skb_forward_csum(skb);
377 if (skb->len - ETH_HLEN > odev->mtu)
381 skb->protocol = eth->h_proto;
383 return dev_queue_xmit(skb);
390 static int input_action_end_dx6_finish(struct net *net, struct sock *sk,
393 struct dst_entry *orig_dst = skb_dst(skb);
394 struct in6_addr *nhaddr = NULL;
395 struct seg6_local_lwt *slwt;
397 slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
399 /* The inner packet is not associated to any local interface,
400 * so we do not call netif_rx().
402 * If slwt->nh6 is set to ::, then lookup the nexthop for the
403 * inner packet's DA. Otherwise, use the specified nexthop.
405 if (!ipv6_addr_any(&slwt->nh6))
408 seg6_lookup_nexthop(skb, nhaddr, 0);
410 return dst_input(skb);
413 /* decapsulate and forward to specified nexthop */
414 static int input_action_end_dx6(struct sk_buff *skb,
415 struct seg6_local_lwt *slwt)
417 /* this function accepts IPv6 encapsulated packets, with either
418 * an SRH with SL=0, or no SRH.
421 if (!decap_and_validate(skb, IPPROTO_IPV6))
424 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
427 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
430 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
431 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
432 dev_net(skb->dev), NULL, skb, NULL,
433 skb_dst(skb)->dev, input_action_end_dx6_finish);
435 return input_action_end_dx6_finish(dev_net(skb->dev), NULL, skb);
441 static int input_action_end_dx4_finish(struct net *net, struct sock *sk,
444 struct dst_entry *orig_dst = skb_dst(skb);
445 struct seg6_local_lwt *slwt;
450 slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
454 nhaddr = slwt->nh4.s_addr ?: iph->daddr;
458 err = ip_route_input(skb, nhaddr, iph->saddr, 0, skb->dev);
464 return dst_input(skb);
467 static int input_action_end_dx4(struct sk_buff *skb,
468 struct seg6_local_lwt *slwt)
470 if (!decap_and_validate(skb, IPPROTO_IPIP))
473 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
476 skb->protocol = htons(ETH_P_IP);
477 skb_set_transport_header(skb, sizeof(struct iphdr));
480 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
481 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING,
482 dev_net(skb->dev), NULL, skb, NULL,
483 skb_dst(skb)->dev, input_action_end_dx4_finish);
485 return input_action_end_dx4_finish(dev_net(skb->dev), NULL, skb);
491 #ifdef CONFIG_NET_L3_MASTER_DEV
492 static struct net *fib6_config_get_net(const struct fib6_config *fib6_cfg)
494 const struct nl_info *nli = &fib6_cfg->fc_nlinfo;
499 static int __seg6_end_dt_vrf_build(struct seg6_local_lwt *slwt, const void *cfg,
500 u16 family, struct netlink_ext_ack *extack)
502 struct seg6_end_dt_info *info = &slwt->dt_info;
506 net = fib6_config_get_net(cfg);
508 /* note that vrf_table was already set by parse_nla_vrftable() */
509 vrf_ifindex = l3mdev_ifindex_lookup_by_table_id(L3MDEV_TYPE_VRF, net,
511 if (vrf_ifindex < 0) {
512 if (vrf_ifindex == -EPERM) {
513 NL_SET_ERR_MSG(extack,
514 "Strict mode for VRF is disabled");
515 } else if (vrf_ifindex == -ENODEV) {
516 NL_SET_ERR_MSG(extack,
517 "Table has no associated VRF device");
519 pr_debug("seg6local: SRv6 End.DT* creation error=%d\n",
527 info->vrf_ifindex = vrf_ifindex;
529 info->family = family;
530 info->mode = DT_VRF_MODE;
535 /* The SRv6 End.DT4/DT6 behavior extracts the inner (IPv4/IPv6) packet and
536 * routes the IPv4/IPv6 packet by looking at the configured routing table.
538 * In the SRv6 End.DT4/DT6 use case, we can receive traffic (IPv6+Segment
539 * Routing Header packets) from several interfaces and the outer IPv6
540 * destination address (DA) is used for retrieving the specific instance of the
541 * End.DT4/DT6 behavior that should process the packets.
543 * However, the inner IPv4/IPv6 packet is not really bound to any receiving
544 * interface and thus the End.DT4/DT6 sets the VRF (associated with the
545 * corresponding routing table) as the *receiving* interface.
546 * In other words, the End.DT4/DT6 processes a packet as if it has been received
547 * directly by the VRF (and not by one of its slave devices, if any).
548 * In this way, the VRF interface is used for routing the IPv4/IPv6 packet in
549 * according to the routing table configured by the End.DT4/DT6 instance.
551 * This design allows you to get some interesting features like:
552 * 1) the statistics on rx packets;
553 * 2) the possibility to install a packet sniffer on the receiving interface
554 * (the VRF one) for looking at the incoming packets;
555 * 3) the possibility to leverage the netfilter prerouting hook for the inner
558 * This function returns:
559 * - the sk_buff* when the VRF rcv handler has processed the packet correctly;
560 * - NULL when the skb is consumed by the VRF rcv handler;
561 * - a pointer which encodes a negative error number in case of error.
562 * Note that in this case, the function takes care of freeing the skb.
564 static struct sk_buff *end_dt_vrf_rcv(struct sk_buff *skb, u16 family,
565 struct net_device *dev)
567 /* based on l3mdev_ip_rcv; we are only interested in the master */
568 if (unlikely(!netif_is_l3_master(dev) && !netif_has_l3_rx_handler(dev)))
571 if (unlikely(!dev->l3mdev_ops->l3mdev_l3_rcv))
574 /* the decap packet IPv4/IPv6 does not come with any mac header info.
575 * We must unset the mac header to allow the VRF device to rebuild it,
576 * just in case there is a sniffer attached on the device.
578 skb_unset_mac_header(skb);
580 skb = dev->l3mdev_ops->l3mdev_l3_rcv(dev, skb, family);
582 /* the skb buffer was consumed by the handler */
585 /* when a packet is received by a VRF or by one of its slaves, the
586 * master device reference is set into the skb.
588 if (unlikely(skb->dev != dev || skb->skb_iif != dev->ifindex))
595 return ERR_PTR(-EINVAL);
598 static struct net_device *end_dt_get_vrf_rcu(struct sk_buff *skb,
599 struct seg6_end_dt_info *info)
601 int vrf_ifindex = info->vrf_ifindex;
602 struct net *net = info->net;
604 if (unlikely(vrf_ifindex < 0))
607 if (unlikely(!net_eq(dev_net(skb->dev), net)))
610 return dev_get_by_index_rcu(net, vrf_ifindex);
616 static struct sk_buff *end_dt_vrf_core(struct sk_buff *skb,
617 struct seg6_local_lwt *slwt, u16 family)
619 struct seg6_end_dt_info *info = &slwt->dt_info;
620 struct net_device *vrf;
624 vrf = end_dt_get_vrf_rcu(skb, info);
630 protocol = htons(ETH_P_IP);
631 hdrlen = sizeof(struct iphdr);
634 protocol = htons(ETH_P_IPV6);
635 hdrlen = sizeof(struct ipv6hdr);
643 if (unlikely(info->family != AF_UNSPEC && info->family != family)) {
644 pr_warn_once("seg6local: SRv6 End.DT* family mismatch");
648 skb->protocol = protocol;
652 skb_set_transport_header(skb, hdrlen);
655 return end_dt_vrf_rcv(skb, family, vrf);
659 return ERR_PTR(-EINVAL);
662 static int input_action_end_dt4(struct sk_buff *skb,
663 struct seg6_local_lwt *slwt)
668 if (!decap_and_validate(skb, IPPROTO_IPIP))
671 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
674 skb = end_dt_vrf_core(skb, slwt, AF_INET);
676 /* packet has been processed and consumed by the VRF */
684 err = ip_route_input(skb, iph->daddr, iph->saddr, 0, skb->dev);
688 return dst_input(skb);
695 static int seg6_end_dt4_build(struct seg6_local_lwt *slwt, const void *cfg,
696 struct netlink_ext_ack *extack)
698 return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET, extack);
702 seg6_end_dt_mode seg6_end_dt6_parse_mode(struct seg6_local_lwt *slwt)
704 unsigned long parsed_optattrs = slwt->parsed_optattrs;
705 bool legacy, vrfmode;
707 legacy = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE));
708 vrfmode = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE));
710 if (!(legacy ^ vrfmode))
711 /* both are absent or present: invalid DT6 mode */
712 return DT_INVALID_MODE;
714 return legacy ? DT_LEGACY_MODE : DT_VRF_MODE;
717 static enum seg6_end_dt_mode seg6_end_dt6_get_mode(struct seg6_local_lwt *slwt)
719 struct seg6_end_dt_info *info = &slwt->dt_info;
724 static int seg6_end_dt6_build(struct seg6_local_lwt *slwt, const void *cfg,
725 struct netlink_ext_ack *extack)
727 enum seg6_end_dt_mode mode = seg6_end_dt6_parse_mode(slwt);
728 struct seg6_end_dt_info *info = &slwt->dt_info;
732 info->mode = DT_LEGACY_MODE;
735 return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET6, extack);
737 NL_SET_ERR_MSG(extack, "table or vrftable must be specified");
743 static int input_action_end_dt6(struct sk_buff *skb,
744 struct seg6_local_lwt *slwt)
746 if (!decap_and_validate(skb, IPPROTO_IPV6))
749 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
752 #ifdef CONFIG_NET_L3_MASTER_DEV
753 if (seg6_end_dt6_get_mode(slwt) == DT_LEGACY_MODE)
757 skb = end_dt_vrf_core(skb, slwt, AF_INET6);
759 /* packet has been processed and consumed by the VRF */
765 /* note: this time we do not need to specify the table because the VRF
766 * takes care of selecting the correct table.
768 seg6_lookup_any_nexthop(skb, NULL, 0, true);
770 return dst_input(skb);
774 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
776 seg6_lookup_any_nexthop(skb, NULL, slwt->table, true);
778 return dst_input(skb);
785 #ifdef CONFIG_NET_L3_MASTER_DEV
786 static int seg6_end_dt46_build(struct seg6_local_lwt *slwt, const void *cfg,
787 struct netlink_ext_ack *extack)
789 return __seg6_end_dt_vrf_build(slwt, cfg, AF_UNSPEC, extack);
792 static int input_action_end_dt46(struct sk_buff *skb,
793 struct seg6_local_lwt *slwt)
795 unsigned int off = 0;
798 nexthdr = ipv6_find_hdr(skb, &off, -1, NULL, NULL);
799 if (unlikely(nexthdr < 0))
804 return input_action_end_dt4(skb, slwt);
806 return input_action_end_dt6(skb, slwt);
815 /* push an SRH on top of the current one */
816 static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
818 struct ipv6_sr_hdr *srh;
821 srh = get_and_validate_srh(skb);
825 err = seg6_do_srh_inline(skb, slwt->srh);
829 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
830 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
832 seg6_lookup_nexthop(skb, NULL, 0);
834 return dst_input(skb);
841 /* encapsulate within an outer IPv6 header and a specified SRH */
842 static int input_action_end_b6_encap(struct sk_buff *skb,
843 struct seg6_local_lwt *slwt)
845 struct ipv6_sr_hdr *srh;
848 srh = get_and_validate_srh(skb);
852 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
854 skb_reset_inner_headers(skb);
855 skb->encapsulation = 1;
857 err = seg6_do_srh_encap(skb, slwt->srh, IPPROTO_IPV6);
861 ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
862 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
864 seg6_lookup_nexthop(skb, NULL, 0);
866 return dst_input(skb);
873 DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states);
875 bool seg6_bpf_has_valid_srh(struct sk_buff *skb)
877 struct seg6_bpf_srh_state *srh_state =
878 this_cpu_ptr(&seg6_bpf_srh_states);
879 struct ipv6_sr_hdr *srh = srh_state->srh;
881 if (unlikely(srh == NULL))
884 if (unlikely(!srh_state->valid)) {
885 if ((srh_state->hdrlen & 7) != 0)
888 srh->hdrlen = (u8)(srh_state->hdrlen >> 3);
889 if (!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3, true))
892 srh_state->valid = true;
898 static int input_action_end_bpf(struct sk_buff *skb,
899 struct seg6_local_lwt *slwt)
901 struct seg6_bpf_srh_state *srh_state =
902 this_cpu_ptr(&seg6_bpf_srh_states);
903 struct ipv6_sr_hdr *srh;
906 srh = get_and_validate_srh(skb);
911 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
913 /* preempt_disable is needed to protect the per-CPU buffer srh_state,
914 * which is also accessed by the bpf_lwt_seg6_* helpers
917 srh_state->srh = srh;
918 srh_state->hdrlen = srh->hdrlen << 3;
919 srh_state->valid = true;
922 bpf_compute_data_pointers(skb);
923 ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb);
933 pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret);
937 if (srh_state->srh && !seg6_bpf_has_valid_srh(skb))
941 if (ret != BPF_REDIRECT)
942 seg6_lookup_nexthop(skb, NULL, 0);
944 return dst_input(skb);
952 static struct seg6_action_desc seg6_action_table[] = {
954 .action = SEG6_LOCAL_ACTION_END,
956 .optattrs = SEG6_F_LOCAL_COUNTERS,
957 .input = input_action_end,
960 .action = SEG6_LOCAL_ACTION_END_X,
961 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6),
962 .optattrs = SEG6_F_LOCAL_COUNTERS,
963 .input = input_action_end_x,
966 .action = SEG6_LOCAL_ACTION_END_T,
967 .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE),
968 .optattrs = SEG6_F_LOCAL_COUNTERS,
969 .input = input_action_end_t,
972 .action = SEG6_LOCAL_ACTION_END_DX2,
973 .attrs = SEG6_F_ATTR(SEG6_LOCAL_OIF),
974 .optattrs = SEG6_F_LOCAL_COUNTERS,
975 .input = input_action_end_dx2,
978 .action = SEG6_LOCAL_ACTION_END_DX6,
979 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6),
980 .optattrs = SEG6_F_LOCAL_COUNTERS,
981 .input = input_action_end_dx6,
984 .action = SEG6_LOCAL_ACTION_END_DX4,
985 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH4),
986 .optattrs = SEG6_F_LOCAL_COUNTERS,
987 .input = input_action_end_dx4,
990 .action = SEG6_LOCAL_ACTION_END_DT4,
991 .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE),
992 .optattrs = SEG6_F_LOCAL_COUNTERS,
993 #ifdef CONFIG_NET_L3_MASTER_DEV
994 .input = input_action_end_dt4,
996 .build_state = seg6_end_dt4_build,
1001 .action = SEG6_LOCAL_ACTION_END_DT6,
1002 #ifdef CONFIG_NET_L3_MASTER_DEV
1004 .optattrs = SEG6_F_LOCAL_COUNTERS |
1005 SEG6_F_ATTR(SEG6_LOCAL_TABLE) |
1006 SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE),
1008 .build_state = seg6_end_dt6_build,
1011 .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE),
1012 .optattrs = SEG6_F_LOCAL_COUNTERS,
1014 .input = input_action_end_dt6,
1017 .action = SEG6_LOCAL_ACTION_END_DT46,
1018 .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE),
1019 .optattrs = SEG6_F_LOCAL_COUNTERS,
1020 #ifdef CONFIG_NET_L3_MASTER_DEV
1021 .input = input_action_end_dt46,
1023 .build_state = seg6_end_dt46_build,
1028 .action = SEG6_LOCAL_ACTION_END_B6,
1029 .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH),
1030 .optattrs = SEG6_F_LOCAL_COUNTERS,
1031 .input = input_action_end_b6,
1034 .action = SEG6_LOCAL_ACTION_END_B6_ENCAP,
1035 .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH),
1036 .optattrs = SEG6_F_LOCAL_COUNTERS,
1037 .input = input_action_end_b6_encap,
1038 .static_headroom = sizeof(struct ipv6hdr),
1041 .action = SEG6_LOCAL_ACTION_END_BPF,
1042 .attrs = SEG6_F_ATTR(SEG6_LOCAL_BPF),
1043 .optattrs = SEG6_F_LOCAL_COUNTERS,
1044 .input = input_action_end_bpf,
1049 static struct seg6_action_desc *__get_action_desc(int action)
1051 struct seg6_action_desc *desc;
1054 count = ARRAY_SIZE(seg6_action_table);
1055 for (i = 0; i < count; i++) {
1056 desc = &seg6_action_table[i];
1057 if (desc->action == action)
1064 static bool seg6_lwtunnel_counters_enabled(struct seg6_local_lwt *slwt)
1066 return slwt->parsed_optattrs & SEG6_F_LOCAL_COUNTERS;
1069 static void seg6_local_update_counters(struct seg6_local_lwt *slwt,
1070 unsigned int len, int err)
1072 struct pcpu_seg6_local_counters *pcounters;
1074 pcounters = this_cpu_ptr(slwt->pcpu_counters);
1075 u64_stats_update_begin(&pcounters->syncp);
1078 u64_stats_inc(&pcounters->packets);
1079 u64_stats_add(&pcounters->bytes, len);
1081 u64_stats_inc(&pcounters->errors);
1084 u64_stats_update_end(&pcounters->syncp);
1087 static int seg6_local_input_core(struct net *net, struct sock *sk,
1088 struct sk_buff *skb)
1090 struct dst_entry *orig_dst = skb_dst(skb);
1091 struct seg6_action_desc *desc;
1092 struct seg6_local_lwt *slwt;
1093 unsigned int len = skb->len;
1096 slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
1099 rc = desc->input(skb, slwt);
1101 if (!seg6_lwtunnel_counters_enabled(slwt))
1104 seg6_local_update_counters(slwt, len, rc);
1109 static int seg6_local_input(struct sk_buff *skb)
1111 if (skb->protocol != htons(ETH_P_IPV6)) {
1116 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
1117 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
1118 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
1119 seg6_local_input_core);
1121 return seg6_local_input_core(dev_net(skb->dev), NULL, skb);
1124 static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = {
1125 [SEG6_LOCAL_ACTION] = { .type = NLA_U32 },
1126 [SEG6_LOCAL_SRH] = { .type = NLA_BINARY },
1127 [SEG6_LOCAL_TABLE] = { .type = NLA_U32 },
1128 [SEG6_LOCAL_VRFTABLE] = { .type = NLA_U32 },
1129 [SEG6_LOCAL_NH4] = { .type = NLA_BINARY,
1130 .len = sizeof(struct in_addr) },
1131 [SEG6_LOCAL_NH6] = { .type = NLA_BINARY,
1132 .len = sizeof(struct in6_addr) },
1133 [SEG6_LOCAL_IIF] = { .type = NLA_U32 },
1134 [SEG6_LOCAL_OIF] = { .type = NLA_U32 },
1135 [SEG6_LOCAL_BPF] = { .type = NLA_NESTED },
1136 [SEG6_LOCAL_COUNTERS] = { .type = NLA_NESTED },
1139 static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1141 struct ipv6_sr_hdr *srh;
1144 srh = nla_data(attrs[SEG6_LOCAL_SRH]);
1145 len = nla_len(attrs[SEG6_LOCAL_SRH]);
1147 /* SRH must contain at least one segment */
1148 if (len < sizeof(*srh) + sizeof(struct in6_addr))
1151 if (!seg6_validate_srh(srh, len, false))
1154 slwt->srh = kmemdup(srh, len, GFP_KERNEL);
1158 slwt->headroom += len;
1163 static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1165 struct ipv6_sr_hdr *srh;
1170 len = (srh->hdrlen + 1) << 3;
1172 nla = nla_reserve(skb, SEG6_LOCAL_SRH, len);
1176 memcpy(nla_data(nla), srh, len);
1181 static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1183 int len = (a->srh->hdrlen + 1) << 3;
1185 if (len != ((b->srh->hdrlen + 1) << 3))
1188 return memcmp(a->srh, b->srh, len);
1191 static void destroy_attr_srh(struct seg6_local_lwt *slwt)
1196 static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1198 slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]);
1203 static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1205 if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table))
1211 static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1213 if (a->table != b->table)
1220 seg6_end_dt_info *seg6_possible_end_dt_info(struct seg6_local_lwt *slwt)
1222 #ifdef CONFIG_NET_L3_MASTER_DEV
1223 return &slwt->dt_info;
1225 return ERR_PTR(-EOPNOTSUPP);
1229 static int parse_nla_vrftable(struct nlattr **attrs,
1230 struct seg6_local_lwt *slwt)
1232 struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt);
1235 return PTR_ERR(info);
1237 info->vrf_table = nla_get_u32(attrs[SEG6_LOCAL_VRFTABLE]);
1242 static int put_nla_vrftable(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1244 struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt);
1247 return PTR_ERR(info);
1249 if (nla_put_u32(skb, SEG6_LOCAL_VRFTABLE, info->vrf_table))
1255 static int cmp_nla_vrftable(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1257 struct seg6_end_dt_info *info_a = seg6_possible_end_dt_info(a);
1258 struct seg6_end_dt_info *info_b = seg6_possible_end_dt_info(b);
1260 if (info_a->vrf_table != info_b->vrf_table)
1266 static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1268 memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]),
1269 sizeof(struct in_addr));
1274 static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1278 nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr));
1282 memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr));
1287 static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1289 return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr));
1292 static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1294 memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]),
1295 sizeof(struct in6_addr));
1300 static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1304 nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr));
1308 memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr));
1313 static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1315 return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr));
1318 static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1320 slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]);
1325 static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1327 if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif))
1333 static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1335 if (a->iif != b->iif)
1341 static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1343 slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]);
1348 static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1350 if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif))
1356 static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1358 if (a->oif != b->oif)
1364 #define MAX_PROG_NAME 256
1365 static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = {
1366 [SEG6_LOCAL_BPF_PROG] = { .type = NLA_U32, },
1367 [SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING,
1368 .len = MAX_PROG_NAME },
1371 static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1373 struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1];
1378 ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX,
1379 attrs[SEG6_LOCAL_BPF],
1380 bpf_prog_policy, NULL);
1384 if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME])
1387 slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL);
1388 if (!slwt->bpf.name)
1391 fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]);
1392 p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL);
1394 kfree(slwt->bpf.name);
1402 static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1404 struct nlattr *nest;
1406 if (!slwt->bpf.prog)
1409 nest = nla_nest_start_noflag(skb, SEG6_LOCAL_BPF);
1413 if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id))
1416 if (slwt->bpf.name &&
1417 nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name))
1420 return nla_nest_end(skb, nest);
1423 static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1425 if (!a->bpf.name && !b->bpf.name)
1428 if (!a->bpf.name || !b->bpf.name)
1431 return strcmp(a->bpf.name, b->bpf.name);
1434 static void destroy_attr_bpf(struct seg6_local_lwt *slwt)
1436 kfree(slwt->bpf.name);
1438 bpf_prog_put(slwt->bpf.prog);
1442 nla_policy seg6_local_counters_policy[SEG6_LOCAL_CNT_MAX + 1] = {
1443 [SEG6_LOCAL_CNT_PACKETS] = { .type = NLA_U64 },
1444 [SEG6_LOCAL_CNT_BYTES] = { .type = NLA_U64 },
1445 [SEG6_LOCAL_CNT_ERRORS] = { .type = NLA_U64 },
1448 static int parse_nla_counters(struct nlattr **attrs,
1449 struct seg6_local_lwt *slwt)
1451 struct pcpu_seg6_local_counters __percpu *pcounters;
1452 struct nlattr *tb[SEG6_LOCAL_CNT_MAX + 1];
1455 ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_CNT_MAX,
1456 attrs[SEG6_LOCAL_COUNTERS],
1457 seg6_local_counters_policy, NULL);
1461 /* basic support for SRv6 Behavior counters requires at least:
1462 * packets, bytes and errors.
1464 if (!tb[SEG6_LOCAL_CNT_PACKETS] || !tb[SEG6_LOCAL_CNT_BYTES] ||
1465 !tb[SEG6_LOCAL_CNT_ERRORS])
1468 /* counters are always zero initialized */
1469 pcounters = seg6_local_alloc_pcpu_counters(GFP_KERNEL);
1473 slwt->pcpu_counters = pcounters;
1478 static int seg6_local_fill_nla_counters(struct sk_buff *skb,
1479 struct seg6_local_counters *counters)
1481 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_PACKETS, counters->packets,
1482 SEG6_LOCAL_CNT_PAD))
1485 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_BYTES, counters->bytes,
1486 SEG6_LOCAL_CNT_PAD))
1489 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_ERRORS, counters->errors,
1490 SEG6_LOCAL_CNT_PAD))
1496 static int put_nla_counters(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1498 struct seg6_local_counters counters = { 0, 0, 0 };
1499 struct nlattr *nest;
1502 nest = nla_nest_start(skb, SEG6_LOCAL_COUNTERS);
1506 for_each_possible_cpu(i) {
1507 struct pcpu_seg6_local_counters *pcounters;
1508 u64 packets, bytes, errors;
1511 pcounters = per_cpu_ptr(slwt->pcpu_counters, i);
1513 start = u64_stats_fetch_begin_irq(&pcounters->syncp);
1515 packets = u64_stats_read(&pcounters->packets);
1516 bytes = u64_stats_read(&pcounters->bytes);
1517 errors = u64_stats_read(&pcounters->errors);
1519 } while (u64_stats_fetch_retry_irq(&pcounters->syncp, start));
1521 counters.packets += packets;
1522 counters.bytes += bytes;
1523 counters.errors += errors;
1526 rc = seg6_local_fill_nla_counters(skb, &counters);
1528 nla_nest_cancel(skb, nest);
1532 return nla_nest_end(skb, nest);
1535 static int cmp_nla_counters(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1537 /* a and b are equal if both have pcpu_counters set or not */
1538 return (!!((unsigned long)a->pcpu_counters)) ^
1539 (!!((unsigned long)b->pcpu_counters));
1542 static void destroy_attr_counters(struct seg6_local_lwt *slwt)
1544 free_percpu(slwt->pcpu_counters);
1547 struct seg6_action_param {
1548 int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt);
1549 int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
1550 int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b);
1552 /* optional destroy() callback useful for releasing resources which
1553 * have been previously acquired in the corresponding parse()
1556 void (*destroy)(struct seg6_local_lwt *slwt);
1559 static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = {
1560 [SEG6_LOCAL_SRH] = { .parse = parse_nla_srh,
1563 .destroy = destroy_attr_srh },
1565 [SEG6_LOCAL_TABLE] = { .parse = parse_nla_table,
1566 .put = put_nla_table,
1567 .cmp = cmp_nla_table },
1569 [SEG6_LOCAL_NH4] = { .parse = parse_nla_nh4,
1571 .cmp = cmp_nla_nh4 },
1573 [SEG6_LOCAL_NH6] = { .parse = parse_nla_nh6,
1575 .cmp = cmp_nla_nh6 },
1577 [SEG6_LOCAL_IIF] = { .parse = parse_nla_iif,
1579 .cmp = cmp_nla_iif },
1581 [SEG6_LOCAL_OIF] = { .parse = parse_nla_oif,
1583 .cmp = cmp_nla_oif },
1585 [SEG6_LOCAL_BPF] = { .parse = parse_nla_bpf,
1588 .destroy = destroy_attr_bpf },
1590 [SEG6_LOCAL_VRFTABLE] = { .parse = parse_nla_vrftable,
1591 .put = put_nla_vrftable,
1592 .cmp = cmp_nla_vrftable },
1594 [SEG6_LOCAL_COUNTERS] = { .parse = parse_nla_counters,
1595 .put = put_nla_counters,
1596 .cmp = cmp_nla_counters,
1597 .destroy = destroy_attr_counters },
1600 /* call the destroy() callback (if available) for each set attribute in
1601 * @parsed_attrs, starting from the first attribute up to the @max_parsed
1602 * (excluded) attribute.
1604 static void __destroy_attrs(unsigned long parsed_attrs, int max_parsed,
1605 struct seg6_local_lwt *slwt)
1607 struct seg6_action_param *param;
1610 /* Every required seg6local attribute is identified by an ID which is
1611 * encoded as a flag (i.e: 1 << ID) in the 'attrs' bitmask;
1613 * We scan the 'parsed_attrs' bitmask, starting from the first attribute
1614 * up to the @max_parsed (excluded) attribute.
1615 * For each set attribute, we retrieve the corresponding destroy()
1616 * callback. If the callback is not available, then we skip to the next
1617 * attribute; otherwise, we call the destroy() callback.
1619 for (i = 0; i < max_parsed; ++i) {
1620 if (!(parsed_attrs & SEG6_F_ATTR(i)))
1623 param = &seg6_action_params[i];
1626 param->destroy(slwt);
1630 /* release all the resources that may have been acquired during parsing
1633 static void destroy_attrs(struct seg6_local_lwt *slwt)
1635 unsigned long attrs = slwt->desc->attrs | slwt->parsed_optattrs;
1637 __destroy_attrs(attrs, SEG6_LOCAL_MAX + 1, slwt);
1640 static int parse_nla_optional_attrs(struct nlattr **attrs,
1641 struct seg6_local_lwt *slwt)
1643 struct seg6_action_desc *desc = slwt->desc;
1644 unsigned long parsed_optattrs = 0;
1645 struct seg6_action_param *param;
1648 for (i = 0; i < SEG6_LOCAL_MAX + 1; ++i) {
1649 if (!(desc->optattrs & SEG6_F_ATTR(i)) || !attrs[i])
1652 /* once here, the i-th attribute is provided by the
1653 * userspace AND it is identified optional as well.
1655 param = &seg6_action_params[i];
1657 err = param->parse(attrs, slwt);
1659 goto parse_optattrs_err;
1661 /* current attribute has been correctly parsed */
1662 parsed_optattrs |= SEG6_F_ATTR(i);
1665 /* store in the tunnel state all the optional attributed successfully
1668 slwt->parsed_optattrs = parsed_optattrs;
1673 __destroy_attrs(parsed_optattrs, i, slwt);
1678 /* call the custom constructor of the behavior during its initialization phase
1679 * and after that all its attributes have been parsed successfully.
1682 seg6_local_lwtunnel_build_state(struct seg6_local_lwt *slwt, const void *cfg,
1683 struct netlink_ext_ack *extack)
1685 struct seg6_action_desc *desc = slwt->desc;
1686 struct seg6_local_lwtunnel_ops *ops;
1688 ops = &desc->slwt_ops;
1689 if (!ops->build_state)
1692 return ops->build_state(slwt, cfg, extack);
1695 /* call the custom destructor of the behavior which is invoked before the
1696 * tunnel is going to be destroyed.
1698 static void seg6_local_lwtunnel_destroy_state(struct seg6_local_lwt *slwt)
1700 struct seg6_action_desc *desc = slwt->desc;
1701 struct seg6_local_lwtunnel_ops *ops;
1703 ops = &desc->slwt_ops;
1704 if (!ops->destroy_state)
1707 ops->destroy_state(slwt);
1710 static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1712 struct seg6_action_param *param;
1713 struct seg6_action_desc *desc;
1714 unsigned long invalid_attrs;
1717 desc = __get_action_desc(slwt->action);
1725 slwt->headroom += desc->static_headroom;
1727 /* Forcing the desc->optattrs *set* and the desc->attrs *set* to be
1728 * disjoined, this allow us to release acquired resources by optional
1729 * attributes and by required attributes independently from each other
1730 * without any interference.
1731 * In other terms, we are sure that we do not release some the acquired
1734 * Note that if an attribute is configured both as required and as
1735 * optional, it means that the user has messed something up in the
1736 * seg6_action_table. Therefore, this check is required for SRv6
1737 * behaviors to work properly.
1739 invalid_attrs = desc->attrs & desc->optattrs;
1740 if (invalid_attrs) {
1742 "An attribute cannot be both required AND optional");
1746 /* parse the required attributes */
1747 for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1748 if (desc->attrs & SEG6_F_ATTR(i)) {
1752 param = &seg6_action_params[i];
1754 err = param->parse(attrs, slwt);
1756 goto parse_attrs_err;
1760 /* parse the optional attributes, if any */
1761 err = parse_nla_optional_attrs(attrs, slwt);
1763 goto parse_attrs_err;
1768 /* release any resource that may have been acquired during the i-1
1769 * parse() operations.
1771 __destroy_attrs(desc->attrs, i, slwt);
1776 static int seg6_local_build_state(struct net *net, struct nlattr *nla,
1777 unsigned int family, const void *cfg,
1778 struct lwtunnel_state **ts,
1779 struct netlink_ext_ack *extack)
1781 struct nlattr *tb[SEG6_LOCAL_MAX + 1];
1782 struct lwtunnel_state *newts;
1783 struct seg6_local_lwt *slwt;
1786 if (family != AF_INET6)
1789 err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla,
1790 seg6_local_policy, extack);
1795 if (!tb[SEG6_LOCAL_ACTION])
1798 newts = lwtunnel_state_alloc(sizeof(*slwt));
1802 slwt = seg6_local_lwtunnel(newts);
1803 slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]);
1805 err = parse_nla_action(tb, slwt);
1809 err = seg6_local_lwtunnel_build_state(slwt, cfg, extack);
1811 goto out_destroy_attrs;
1813 newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL;
1814 newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT;
1815 newts->headroom = slwt->headroom;
1822 destroy_attrs(slwt);
1828 static void seg6_local_destroy_state(struct lwtunnel_state *lwt)
1830 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1832 seg6_local_lwtunnel_destroy_state(slwt);
1834 destroy_attrs(slwt);
1839 static int seg6_local_fill_encap(struct sk_buff *skb,
1840 struct lwtunnel_state *lwt)
1842 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1843 struct seg6_action_param *param;
1844 unsigned long attrs;
1847 if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action))
1850 attrs = slwt->desc->attrs | slwt->parsed_optattrs;
1852 for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1853 if (attrs & SEG6_F_ATTR(i)) {
1854 param = &seg6_action_params[i];
1855 err = param->put(skb, slwt);
1864 static int seg6_local_get_encap_size(struct lwtunnel_state *lwt)
1866 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1867 unsigned long attrs;
1870 nlsize = nla_total_size(4); /* action */
1872 attrs = slwt->desc->attrs | slwt->parsed_optattrs;
1874 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_SRH))
1875 nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3);
1877 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE))
1878 nlsize += nla_total_size(4);
1880 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH4))
1881 nlsize += nla_total_size(4);
1883 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH6))
1884 nlsize += nla_total_size(16);
1886 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_IIF))
1887 nlsize += nla_total_size(4);
1889 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_OIF))
1890 nlsize += nla_total_size(4);
1892 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_BPF))
1893 nlsize += nla_total_size(sizeof(struct nlattr)) +
1894 nla_total_size(MAX_PROG_NAME) +
1897 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE))
1898 nlsize += nla_total_size(4);
1900 if (attrs & SEG6_F_LOCAL_COUNTERS)
1901 nlsize += nla_total_size(0) + /* nest SEG6_LOCAL_COUNTERS */
1902 /* SEG6_LOCAL_CNT_PACKETS */
1903 nla_total_size_64bit(sizeof(__u64)) +
1904 /* SEG6_LOCAL_CNT_BYTES */
1905 nla_total_size_64bit(sizeof(__u64)) +
1906 /* SEG6_LOCAL_CNT_ERRORS */
1907 nla_total_size_64bit(sizeof(__u64));
1912 static int seg6_local_cmp_encap(struct lwtunnel_state *a,
1913 struct lwtunnel_state *b)
1915 struct seg6_local_lwt *slwt_a, *slwt_b;
1916 struct seg6_action_param *param;
1917 unsigned long attrs_a, attrs_b;
1920 slwt_a = seg6_local_lwtunnel(a);
1921 slwt_b = seg6_local_lwtunnel(b);
1923 if (slwt_a->action != slwt_b->action)
1926 attrs_a = slwt_a->desc->attrs | slwt_a->parsed_optattrs;
1927 attrs_b = slwt_b->desc->attrs | slwt_b->parsed_optattrs;
1929 if (attrs_a != attrs_b)
1932 for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1933 if (attrs_a & SEG6_F_ATTR(i)) {
1934 param = &seg6_action_params[i];
1935 if (param->cmp(slwt_a, slwt_b))
1943 static const struct lwtunnel_encap_ops seg6_local_ops = {
1944 .build_state = seg6_local_build_state,
1945 .destroy_state = seg6_local_destroy_state,
1946 .input = seg6_local_input,
1947 .fill_encap = seg6_local_fill_encap,
1948 .get_encap_size = seg6_local_get_encap_size,
1949 .cmp_encap = seg6_local_cmp_encap,
1950 .owner = THIS_MODULE,
1953 int __init seg6_local_init(void)
1955 /* If the max total number of defined attributes is reached, then your
1956 * kernel build stops here.
1958 * This check is required to avoid arithmetic overflows when processing
1959 * behavior attributes and the maximum number of defined attributes
1960 * exceeds the allowed value.
1962 BUILD_BUG_ON(SEG6_LOCAL_MAX + 1 > BITS_PER_TYPE(unsigned long));
1964 return lwtunnel_encap_add_ops(&seg6_local_ops,
1965 LWTUNNEL_ENCAP_SEG6_LOCAL);
1968 void seg6_local_exit(void)
1970 lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL);