• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *  SR-IPv6 implementation
4  *
5  *  Authors:
6  *  David Lebrun <david.lebrun@uclouvain.be>
7  *  eBPF support: Mathieu Xhonneux <m.xhonneux@gmail.com>
8  */
9 
10 #include <linux/types.h>
11 #include <linux/skbuff.h>
12 #include <linux/net.h>
13 #include <linux/module.h>
14 #include <net/ip.h>
15 #include <net/lwtunnel.h>
16 #include <net/netevent.h>
17 #include <net/netns/generic.h>
18 #include <net/ip6_fib.h>
19 #include <net/route.h>
20 #include <net/seg6.h>
21 #include <linux/seg6.h>
22 #include <linux/seg6_local.h>
23 #include <net/addrconf.h>
24 #include <net/ip6_route.h>
25 #include <net/dst_cache.h>
26 #include <net/ip_tunnels.h>
27 #ifdef CONFIG_IPV6_SEG6_HMAC
28 #include <net/seg6_hmac.h>
29 #endif
30 #include <net/seg6_local.h>
31 #include <linux/etherdevice.h>
32 #include <linux/bpf.h>
33 #include <linux/netfilter.h>
34 
35 #define SEG6_F_ATTR(i)		BIT(i)
36 
37 struct seg6_local_lwt;
38 
39 /* callbacks used for customizing the creation and destruction of a behavior */
40 struct seg6_local_lwtunnel_ops {
41 	int (*build_state)(struct seg6_local_lwt *slwt, const void *cfg,
42 			   struct netlink_ext_ack *extack);
43 	void (*destroy_state)(struct seg6_local_lwt *slwt);
44 };
45 
46 struct seg6_action_desc {
47 	int action;
48 	unsigned long attrs;
49 
50 	/* The optattrs field is used for specifying all the optional
51 	 * attributes supported by a specific behavior.
52 	 * It means that if one of these attributes is not provided in the
53 	 * netlink message during the behavior creation, no errors will be
54 	 * returned to the userspace.
55 	 *
56 	 * Each attribute can be only of two types (mutually exclusive):
57 	 * 1) required or 2) optional.
58 	 * Every user MUST obey to this rule! If you set an attribute as
59 	 * required the same attribute CANNOT be set as optional and vice
60 	 * versa.
61 	 */
62 	unsigned long optattrs;
63 
64 	int (*input)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
65 	int static_headroom;
66 
67 	struct seg6_local_lwtunnel_ops slwt_ops;
68 };
69 
70 struct bpf_lwt_prog {
71 	struct bpf_prog *prog;
72 	char *name;
73 };
74 
75 enum seg6_end_dt_mode {
76 	DT_INVALID_MODE	= -EINVAL,
77 	DT_LEGACY_MODE	= 0,
78 	DT_VRF_MODE	= 1,
79 };
80 
81 struct seg6_end_dt_info {
82 	enum seg6_end_dt_mode mode;
83 
84 	struct net *net;
85 	/* VRF device associated to the routing table used by the SRv6
86 	 * End.DT4/DT6 behavior for routing IPv4/IPv6 packets.
87 	 */
88 	int vrf_ifindex;
89 	int vrf_table;
90 
91 	/* tunneled packet family (IPv4 or IPv6).
92 	 * Protocol and header length are inferred from family.
93 	 */
94 	u16 family;
95 };
96 
97 struct pcpu_seg6_local_counters {
98 	u64_stats_t packets;
99 	u64_stats_t bytes;
100 	u64_stats_t errors;
101 
102 	struct u64_stats_sync syncp;
103 };
104 
105 /* This struct groups all the SRv6 Behavior counters supported so far.
106  *
107  * put_nla_counters() makes use of this data structure to collect all counter
108  * values after the per-CPU counter evaluation has been performed.
109  * Finally, each counter value (in seg6_local_counters) is stored in the
110  * corresponding netlink attribute and sent to user space.
111  *
112  * NB: we don't want to expose this structure to user space!
113  */
114 struct seg6_local_counters {
115 	__u64 packets;
116 	__u64 bytes;
117 	__u64 errors;
118 };
119 
120 #define seg6_local_alloc_pcpu_counters(__gfp)				\
121 	__netdev_alloc_pcpu_stats(struct pcpu_seg6_local_counters,	\
122 				  ((__gfp) | __GFP_ZERO))
123 
124 #define SEG6_F_LOCAL_COUNTERS	SEG6_F_ATTR(SEG6_LOCAL_COUNTERS)
125 
126 struct seg6_local_lwt {
127 	int action;
128 	struct ipv6_sr_hdr *srh;
129 	int table;
130 	struct in_addr nh4;
131 	struct in6_addr nh6;
132 	int iif;
133 	int oif;
134 	struct bpf_lwt_prog bpf;
135 #ifdef CONFIG_NET_L3_MASTER_DEV
136 	struct seg6_end_dt_info dt_info;
137 #endif
138 	struct pcpu_seg6_local_counters __percpu *pcpu_counters;
139 
140 	int headroom;
141 	struct seg6_action_desc *desc;
142 	/* unlike the required attrs, we have to track the optional attributes
143 	 * that have been effectively parsed.
144 	 */
145 	unsigned long parsed_optattrs;
146 };
147 
seg6_local_lwtunnel(struct lwtunnel_state * lwt)148 static struct seg6_local_lwt *seg6_local_lwtunnel(struct lwtunnel_state *lwt)
149 {
150 	return (struct seg6_local_lwt *)lwt->data;
151 }
152 
get_and_validate_srh(struct sk_buff * skb)153 static struct ipv6_sr_hdr *get_and_validate_srh(struct sk_buff *skb)
154 {
155 	struct ipv6_sr_hdr *srh;
156 
157 	srh = seg6_get_srh(skb, IP6_FH_F_SKIP_RH);
158 	if (!srh)
159 		return NULL;
160 
161 #ifdef CONFIG_IPV6_SEG6_HMAC
162 	if (!seg6_hmac_validate_skb(skb))
163 		return NULL;
164 #endif
165 
166 	return srh;
167 }
168 
decap_and_validate(struct sk_buff * skb,int proto)169 static bool decap_and_validate(struct sk_buff *skb, int proto)
170 {
171 	struct ipv6_sr_hdr *srh;
172 	unsigned int off = 0;
173 
174 	srh = seg6_get_srh(skb, 0);
175 	if (srh && srh->segments_left > 0)
176 		return false;
177 
178 #ifdef CONFIG_IPV6_SEG6_HMAC
179 	if (srh && !seg6_hmac_validate_skb(skb))
180 		return false;
181 #endif
182 
183 	if (ipv6_find_hdr(skb, &off, proto, NULL, NULL) < 0)
184 		return false;
185 
186 	if (!pskb_pull(skb, off))
187 		return false;
188 
189 	skb_postpull_rcsum(skb, skb_network_header(skb), off);
190 
191 	skb_reset_network_header(skb);
192 	skb_reset_transport_header(skb);
193 	if (iptunnel_pull_offloads(skb))
194 		return false;
195 
196 	return true;
197 }
198 
advance_nextseg(struct ipv6_sr_hdr * srh,struct in6_addr * daddr)199 static void advance_nextseg(struct ipv6_sr_hdr *srh, struct in6_addr *daddr)
200 {
201 	struct in6_addr *addr;
202 
203 	srh->segments_left--;
204 	addr = srh->segments + srh->segments_left;
205 	*daddr = *addr;
206 }
207 
208 static int
seg6_lookup_any_nexthop(struct sk_buff * skb,struct in6_addr * nhaddr,u32 tbl_id,bool local_delivery)209 seg6_lookup_any_nexthop(struct sk_buff *skb, struct in6_addr *nhaddr,
210 			u32 tbl_id, bool local_delivery)
211 {
212 	struct net *net = dev_net(skb->dev);
213 	struct ipv6hdr *hdr = ipv6_hdr(skb);
214 	int flags = RT6_LOOKUP_F_HAS_SADDR;
215 	struct dst_entry *dst = NULL;
216 	struct rt6_info *rt;
217 	struct flowi6 fl6;
218 	int dev_flags = 0;
219 
220 	fl6.flowi6_iif = skb->dev->ifindex;
221 	fl6.daddr = nhaddr ? *nhaddr : hdr->daddr;
222 	fl6.saddr = hdr->saddr;
223 	fl6.flowlabel = ip6_flowinfo(hdr);
224 	fl6.flowi6_mark = skb->mark;
225 	fl6.flowi6_proto = hdr->nexthdr;
226 
227 	if (nhaddr)
228 		fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH;
229 
230 	if (!tbl_id) {
231 		dst = ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags);
232 	} else {
233 		struct fib6_table *table;
234 
235 		table = fib6_get_table(net, tbl_id);
236 		if (!table)
237 			goto out;
238 
239 		rt = ip6_pol_route(net, table, 0, &fl6, skb, flags);
240 		dst = &rt->dst;
241 	}
242 
243 	/* we want to discard traffic destined for local packet processing,
244 	 * if @local_delivery is set to false.
245 	 */
246 	if (!local_delivery)
247 		dev_flags |= IFF_LOOPBACK;
248 
249 	if (dst && (dst->dev->flags & dev_flags) && !dst->error) {
250 		dst_release(dst);
251 		dst = NULL;
252 	}
253 
254 out:
255 	if (!dst) {
256 		rt = net->ipv6.ip6_blk_hole_entry;
257 		dst = &rt->dst;
258 		dst_hold(dst);
259 	}
260 
261 	skb_dst_drop(skb);
262 	skb_dst_set(skb, dst);
263 	return dst->error;
264 }
265 
seg6_lookup_nexthop(struct sk_buff * skb,struct in6_addr * nhaddr,u32 tbl_id)266 int seg6_lookup_nexthop(struct sk_buff *skb,
267 			struct in6_addr *nhaddr, u32 tbl_id)
268 {
269 	return seg6_lookup_any_nexthop(skb, nhaddr, tbl_id, false);
270 }
271 
272 /* regular endpoint function */
input_action_end(struct sk_buff * skb,struct seg6_local_lwt * slwt)273 static int input_action_end(struct sk_buff *skb, struct seg6_local_lwt *slwt)
274 {
275 	struct ipv6_sr_hdr *srh;
276 
277 	srh = get_and_validate_srh(skb);
278 	if (!srh)
279 		goto drop;
280 
281 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
282 
283 	seg6_lookup_nexthop(skb, NULL, 0);
284 
285 	return dst_input(skb);
286 
287 drop:
288 	kfree_skb(skb);
289 	return -EINVAL;
290 }
291 
292 /* regular endpoint, and forward to specified nexthop */
input_action_end_x(struct sk_buff * skb,struct seg6_local_lwt * slwt)293 static int input_action_end_x(struct sk_buff *skb, struct seg6_local_lwt *slwt)
294 {
295 	struct ipv6_sr_hdr *srh;
296 
297 	srh = get_and_validate_srh(skb);
298 	if (!srh)
299 		goto drop;
300 
301 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
302 
303 	seg6_lookup_nexthop(skb, &slwt->nh6, 0);
304 
305 	return dst_input(skb);
306 
307 drop:
308 	kfree_skb(skb);
309 	return -EINVAL;
310 }
311 
input_action_end_t(struct sk_buff * skb,struct seg6_local_lwt * slwt)312 static int input_action_end_t(struct sk_buff *skb, struct seg6_local_lwt *slwt)
313 {
314 	struct ipv6_sr_hdr *srh;
315 
316 	srh = get_and_validate_srh(skb);
317 	if (!srh)
318 		goto drop;
319 
320 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
321 
322 	seg6_lookup_nexthop(skb, NULL, slwt->table);
323 
324 	return dst_input(skb);
325 
326 drop:
327 	kfree_skb(skb);
328 	return -EINVAL;
329 }
330 
331 /* decapsulate and forward inner L2 frame on specified interface */
input_action_end_dx2(struct sk_buff * skb,struct seg6_local_lwt * slwt)332 static int input_action_end_dx2(struct sk_buff *skb,
333 				struct seg6_local_lwt *slwt)
334 {
335 	struct net *net = dev_net(skb->dev);
336 	struct net_device *odev;
337 	struct ethhdr *eth;
338 
339 	if (!decap_and_validate(skb, IPPROTO_ETHERNET))
340 		goto drop;
341 
342 	if (!pskb_may_pull(skb, ETH_HLEN))
343 		goto drop;
344 
345 	skb_reset_mac_header(skb);
346 	eth = (struct ethhdr *)skb->data;
347 
348 	/* To determine the frame's protocol, we assume it is 802.3. This avoids
349 	 * a call to eth_type_trans(), which is not really relevant for our
350 	 * use case.
351 	 */
352 	if (!eth_proto_is_802_3(eth->h_proto))
353 		goto drop;
354 
355 	odev = dev_get_by_index_rcu(net, slwt->oif);
356 	if (!odev)
357 		goto drop;
358 
359 	/* As we accept Ethernet frames, make sure the egress device is of
360 	 * the correct type.
361 	 */
362 	if (odev->type != ARPHRD_ETHER)
363 		goto drop;
364 
365 	if (!(odev->flags & IFF_UP) || !netif_carrier_ok(odev))
366 		goto drop;
367 
368 	skb_orphan(skb);
369 
370 	if (skb_warn_if_lro(skb))
371 		goto drop;
372 
373 	skb_forward_csum(skb);
374 
375 	if (skb->len - ETH_HLEN > odev->mtu)
376 		goto drop;
377 
378 	skb->dev = odev;
379 	skb->protocol = eth->h_proto;
380 
381 	return dev_queue_xmit(skb);
382 
383 drop:
384 	kfree_skb(skb);
385 	return -EINVAL;
386 }
387 
input_action_end_dx6_finish(struct net * net,struct sock * sk,struct sk_buff * skb)388 static int input_action_end_dx6_finish(struct net *net, struct sock *sk,
389 				       struct sk_buff *skb)
390 {
391 	struct dst_entry *orig_dst = skb_dst(skb);
392 	struct in6_addr *nhaddr = NULL;
393 	struct seg6_local_lwt *slwt;
394 
395 	slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
396 
397 	/* The inner packet is not associated to any local interface,
398 	 * so we do not call netif_rx().
399 	 *
400 	 * If slwt->nh6 is set to ::, then lookup the nexthop for the
401 	 * inner packet's DA. Otherwise, use the specified nexthop.
402 	 */
403 	if (!ipv6_addr_any(&slwt->nh6))
404 		nhaddr = &slwt->nh6;
405 
406 	seg6_lookup_nexthop(skb, nhaddr, 0);
407 
408 	return dst_input(skb);
409 }
410 
411 /* decapsulate and forward to specified nexthop */
input_action_end_dx6(struct sk_buff * skb,struct seg6_local_lwt * slwt)412 static int input_action_end_dx6(struct sk_buff *skb,
413 				struct seg6_local_lwt *slwt)
414 {
415 	/* this function accepts IPv6 encapsulated packets, with either
416 	 * an SRH with SL=0, or no SRH.
417 	 */
418 
419 	if (!decap_and_validate(skb, IPPROTO_IPV6))
420 		goto drop;
421 
422 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
423 		goto drop;
424 
425 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
426 	nf_reset_ct(skb);
427 
428 	if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
429 		return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
430 			       dev_net(skb->dev), NULL, skb, NULL,
431 			       skb_dst(skb)->dev, input_action_end_dx6_finish);
432 
433 	return input_action_end_dx6_finish(dev_net(skb->dev), NULL, skb);
434 drop:
435 	kfree_skb(skb);
436 	return -EINVAL;
437 }
438 
input_action_end_dx4_finish(struct net * net,struct sock * sk,struct sk_buff * skb)439 static int input_action_end_dx4_finish(struct net *net, struct sock *sk,
440 				       struct sk_buff *skb)
441 {
442 	struct dst_entry *orig_dst = skb_dst(skb);
443 	struct seg6_local_lwt *slwt;
444 	struct iphdr *iph;
445 	__be32 nhaddr;
446 	int err;
447 
448 	slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
449 
450 	iph = ip_hdr(skb);
451 
452 	nhaddr = slwt->nh4.s_addr ?: iph->daddr;
453 
454 	skb_dst_drop(skb);
455 
456 	err = ip_route_input(skb, nhaddr, iph->saddr, 0, skb->dev);
457 	if (err) {
458 		kfree_skb(skb);
459 		return -EINVAL;
460 	}
461 
462 	return dst_input(skb);
463 }
464 
input_action_end_dx4(struct sk_buff * skb,struct seg6_local_lwt * slwt)465 static int input_action_end_dx4(struct sk_buff *skb,
466 				struct seg6_local_lwt *slwt)
467 {
468 	if (!decap_and_validate(skb, IPPROTO_IPIP))
469 		goto drop;
470 
471 	if (!pskb_may_pull(skb, sizeof(struct iphdr)))
472 		goto drop;
473 
474 	skb->protocol = htons(ETH_P_IP);
475 	skb_set_transport_header(skb, sizeof(struct iphdr));
476 	nf_reset_ct(skb);
477 
478 	if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
479 		return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING,
480 			       dev_net(skb->dev), NULL, skb, NULL,
481 			       skb_dst(skb)->dev, input_action_end_dx4_finish);
482 
483 	return input_action_end_dx4_finish(dev_net(skb->dev), NULL, skb);
484 drop:
485 	kfree_skb(skb);
486 	return -EINVAL;
487 }
488 
489 #ifdef CONFIG_NET_L3_MASTER_DEV
fib6_config_get_net(const struct fib6_config * fib6_cfg)490 static struct net *fib6_config_get_net(const struct fib6_config *fib6_cfg)
491 {
492 	const struct nl_info *nli = &fib6_cfg->fc_nlinfo;
493 
494 	return nli->nl_net;
495 }
496 
__seg6_end_dt_vrf_build(struct seg6_local_lwt * slwt,const void * cfg,u16 family,struct netlink_ext_ack * extack)497 static int __seg6_end_dt_vrf_build(struct seg6_local_lwt *slwt, const void *cfg,
498 				   u16 family, struct netlink_ext_ack *extack)
499 {
500 	struct seg6_end_dt_info *info = &slwt->dt_info;
501 	int vrf_ifindex;
502 	struct net *net;
503 
504 	net = fib6_config_get_net(cfg);
505 
506 	/* note that vrf_table was already set by parse_nla_vrftable() */
507 	vrf_ifindex = l3mdev_ifindex_lookup_by_table_id(L3MDEV_TYPE_VRF, net,
508 							info->vrf_table);
509 	if (vrf_ifindex < 0) {
510 		if (vrf_ifindex == -EPERM) {
511 			NL_SET_ERR_MSG(extack,
512 				       "Strict mode for VRF is disabled");
513 		} else if (vrf_ifindex == -ENODEV) {
514 			NL_SET_ERR_MSG(extack,
515 				       "Table has no associated VRF device");
516 		} else {
517 			pr_debug("seg6local: SRv6 End.DT* creation error=%d\n",
518 				 vrf_ifindex);
519 		}
520 
521 		return vrf_ifindex;
522 	}
523 
524 	info->net = net;
525 	info->vrf_ifindex = vrf_ifindex;
526 
527 	info->family = family;
528 	info->mode = DT_VRF_MODE;
529 
530 	return 0;
531 }
532 
533 /* The SRv6 End.DT4/DT6 behavior extracts the inner (IPv4/IPv6) packet and
534  * routes the IPv4/IPv6 packet by looking at the configured routing table.
535  *
536  * In the SRv6 End.DT4/DT6 use case, we can receive traffic (IPv6+Segment
537  * Routing Header packets) from several interfaces and the outer IPv6
538  * destination address (DA) is used for retrieving the specific instance of the
539  * End.DT4/DT6 behavior that should process the packets.
540  *
541  * However, the inner IPv4/IPv6 packet is not really bound to any receiving
542  * interface and thus the End.DT4/DT6 sets the VRF (associated with the
543  * corresponding routing table) as the *receiving* interface.
544  * In other words, the End.DT4/DT6 processes a packet as if it has been received
545  * directly by the VRF (and not by one of its slave devices, if any).
546  * In this way, the VRF interface is used for routing the IPv4/IPv6 packet in
547  * according to the routing table configured by the End.DT4/DT6 instance.
548  *
549  * This design allows you to get some interesting features like:
550  *  1) the statistics on rx packets;
551  *  2) the possibility to install a packet sniffer on the receiving interface
552  *     (the VRF one) for looking at the incoming packets;
553  *  3) the possibility to leverage the netfilter prerouting hook for the inner
554  *     IPv4 packet.
555  *
556  * This function returns:
557  *  - the sk_buff* when the VRF rcv handler has processed the packet correctly;
558  *  - NULL when the skb is consumed by the VRF rcv handler;
559  *  - a pointer which encodes a negative error number in case of error.
560  *    Note that in this case, the function takes care of freeing the skb.
561  */
end_dt_vrf_rcv(struct sk_buff * skb,u16 family,struct net_device * dev)562 static struct sk_buff *end_dt_vrf_rcv(struct sk_buff *skb, u16 family,
563 				      struct net_device *dev)
564 {
565 	/* based on l3mdev_ip_rcv; we are only interested in the master */
566 	if (unlikely(!netif_is_l3_master(dev) && !netif_has_l3_rx_handler(dev)))
567 		goto drop;
568 
569 	if (unlikely(!dev->l3mdev_ops->l3mdev_l3_rcv))
570 		goto drop;
571 
572 	/* the decap packet IPv4/IPv6 does not come with any mac header info.
573 	 * We must unset the mac header to allow the VRF device to rebuild it,
574 	 * just in case there is a sniffer attached on the device.
575 	 */
576 	skb_unset_mac_header(skb);
577 
578 	skb = dev->l3mdev_ops->l3mdev_l3_rcv(dev, skb, family);
579 	if (!skb)
580 		/* the skb buffer was consumed by the handler */
581 		return NULL;
582 
583 	/* when a packet is received by a VRF or by one of its slaves, the
584 	 * master device reference is set into the skb.
585 	 */
586 	if (unlikely(skb->dev != dev || skb->skb_iif != dev->ifindex))
587 		goto drop;
588 
589 	return skb;
590 
591 drop:
592 	kfree_skb(skb);
593 	return ERR_PTR(-EINVAL);
594 }
595 
end_dt_get_vrf_rcu(struct sk_buff * skb,struct seg6_end_dt_info * info)596 static struct net_device *end_dt_get_vrf_rcu(struct sk_buff *skb,
597 					     struct seg6_end_dt_info *info)
598 {
599 	int vrf_ifindex = info->vrf_ifindex;
600 	struct net *net = info->net;
601 
602 	if (unlikely(vrf_ifindex < 0))
603 		goto error;
604 
605 	if (unlikely(!net_eq(dev_net(skb->dev), net)))
606 		goto error;
607 
608 	return dev_get_by_index_rcu(net, vrf_ifindex);
609 
610 error:
611 	return NULL;
612 }
613 
end_dt_vrf_core(struct sk_buff * skb,struct seg6_local_lwt * slwt,u16 family)614 static struct sk_buff *end_dt_vrf_core(struct sk_buff *skb,
615 				       struct seg6_local_lwt *slwt, u16 family)
616 {
617 	struct seg6_end_dt_info *info = &slwt->dt_info;
618 	struct net_device *vrf;
619 	__be16 protocol;
620 	int hdrlen;
621 
622 	vrf = end_dt_get_vrf_rcu(skb, info);
623 	if (unlikely(!vrf))
624 		goto drop;
625 
626 	switch (family) {
627 	case AF_INET:
628 		protocol = htons(ETH_P_IP);
629 		hdrlen = sizeof(struct iphdr);
630 		break;
631 	case AF_INET6:
632 		protocol = htons(ETH_P_IPV6);
633 		hdrlen = sizeof(struct ipv6hdr);
634 		break;
635 	case AF_UNSPEC:
636 		fallthrough;
637 	default:
638 		goto drop;
639 	}
640 
641 	if (unlikely(info->family != AF_UNSPEC && info->family != family)) {
642 		pr_warn_once("seg6local: SRv6 End.DT* family mismatch");
643 		goto drop;
644 	}
645 
646 	skb->protocol = protocol;
647 
648 	skb_dst_drop(skb);
649 
650 	skb_set_transport_header(skb, hdrlen);
651 	nf_reset_ct(skb);
652 
653 	return end_dt_vrf_rcv(skb, family, vrf);
654 
655 drop:
656 	kfree_skb(skb);
657 	return ERR_PTR(-EINVAL);
658 }
659 
input_action_end_dt4(struct sk_buff * skb,struct seg6_local_lwt * slwt)660 static int input_action_end_dt4(struct sk_buff *skb,
661 				struct seg6_local_lwt *slwt)
662 {
663 	struct iphdr *iph;
664 	int err;
665 
666 	if (!decap_and_validate(skb, IPPROTO_IPIP))
667 		goto drop;
668 
669 	if (!pskb_may_pull(skb, sizeof(struct iphdr)))
670 		goto drop;
671 
672 	skb = end_dt_vrf_core(skb, slwt, AF_INET);
673 	if (!skb)
674 		/* packet has been processed and consumed by the VRF */
675 		return 0;
676 
677 	if (IS_ERR(skb))
678 		return PTR_ERR(skb);
679 
680 	iph = ip_hdr(skb);
681 
682 	err = ip_route_input(skb, iph->daddr, iph->saddr, 0, skb->dev);
683 	if (unlikely(err))
684 		goto drop;
685 
686 	return dst_input(skb);
687 
688 drop:
689 	kfree_skb(skb);
690 	return -EINVAL;
691 }
692 
seg6_end_dt4_build(struct seg6_local_lwt * slwt,const void * cfg,struct netlink_ext_ack * extack)693 static int seg6_end_dt4_build(struct seg6_local_lwt *slwt, const void *cfg,
694 			      struct netlink_ext_ack *extack)
695 {
696 	return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET, extack);
697 }
698 
699 static enum
seg6_end_dt6_parse_mode(struct seg6_local_lwt * slwt)700 seg6_end_dt_mode seg6_end_dt6_parse_mode(struct seg6_local_lwt *slwt)
701 {
702 	unsigned long parsed_optattrs = slwt->parsed_optattrs;
703 	bool legacy, vrfmode;
704 
705 	legacy	= !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE));
706 	vrfmode	= !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE));
707 
708 	if (!(legacy ^ vrfmode))
709 		/* both are absent or present: invalid DT6 mode */
710 		return DT_INVALID_MODE;
711 
712 	return legacy ? DT_LEGACY_MODE : DT_VRF_MODE;
713 }
714 
seg6_end_dt6_get_mode(struct seg6_local_lwt * slwt)715 static enum seg6_end_dt_mode seg6_end_dt6_get_mode(struct seg6_local_lwt *slwt)
716 {
717 	struct seg6_end_dt_info *info = &slwt->dt_info;
718 
719 	return info->mode;
720 }
721 
seg6_end_dt6_build(struct seg6_local_lwt * slwt,const void * cfg,struct netlink_ext_ack * extack)722 static int seg6_end_dt6_build(struct seg6_local_lwt *slwt, const void *cfg,
723 			      struct netlink_ext_ack *extack)
724 {
725 	enum seg6_end_dt_mode mode = seg6_end_dt6_parse_mode(slwt);
726 	struct seg6_end_dt_info *info = &slwt->dt_info;
727 
728 	switch (mode) {
729 	case DT_LEGACY_MODE:
730 		info->mode = DT_LEGACY_MODE;
731 		return 0;
732 	case DT_VRF_MODE:
733 		return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET6, extack);
734 	default:
735 		NL_SET_ERR_MSG(extack, "table or vrftable must be specified");
736 		return -EINVAL;
737 	}
738 }
739 #endif
740 
input_action_end_dt6(struct sk_buff * skb,struct seg6_local_lwt * slwt)741 static int input_action_end_dt6(struct sk_buff *skb,
742 				struct seg6_local_lwt *slwt)
743 {
744 	if (!decap_and_validate(skb, IPPROTO_IPV6))
745 		goto drop;
746 
747 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
748 		goto drop;
749 
750 #ifdef CONFIG_NET_L3_MASTER_DEV
751 	if (seg6_end_dt6_get_mode(slwt) == DT_LEGACY_MODE)
752 		goto legacy_mode;
753 
754 	/* DT6_VRF_MODE */
755 	skb = end_dt_vrf_core(skb, slwt, AF_INET6);
756 	if (!skb)
757 		/* packet has been processed and consumed by the VRF */
758 		return 0;
759 
760 	if (IS_ERR(skb))
761 		return PTR_ERR(skb);
762 
763 	/* note: this time we do not need to specify the table because the VRF
764 	 * takes care of selecting the correct table.
765 	 */
766 	seg6_lookup_any_nexthop(skb, NULL, 0, true);
767 
768 	return dst_input(skb);
769 
770 legacy_mode:
771 #endif
772 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
773 
774 	seg6_lookup_any_nexthop(skb, NULL, slwt->table, true);
775 
776 	return dst_input(skb);
777 
778 drop:
779 	kfree_skb(skb);
780 	return -EINVAL;
781 }
782 
783 #ifdef CONFIG_NET_L3_MASTER_DEV
seg6_end_dt46_build(struct seg6_local_lwt * slwt,const void * cfg,struct netlink_ext_ack * extack)784 static int seg6_end_dt46_build(struct seg6_local_lwt *slwt, const void *cfg,
785 			       struct netlink_ext_ack *extack)
786 {
787 	return __seg6_end_dt_vrf_build(slwt, cfg, AF_UNSPEC, extack);
788 }
789 
input_action_end_dt46(struct sk_buff * skb,struct seg6_local_lwt * slwt)790 static int input_action_end_dt46(struct sk_buff *skb,
791 				 struct seg6_local_lwt *slwt)
792 {
793 	unsigned int off = 0;
794 	int nexthdr;
795 
796 	nexthdr = ipv6_find_hdr(skb, &off, -1, NULL, NULL);
797 	if (unlikely(nexthdr < 0))
798 		goto drop;
799 
800 	switch (nexthdr) {
801 	case IPPROTO_IPIP:
802 		return input_action_end_dt4(skb, slwt);
803 	case IPPROTO_IPV6:
804 		return input_action_end_dt6(skb, slwt);
805 	}
806 
807 drop:
808 	kfree_skb(skb);
809 	return -EINVAL;
810 }
811 #endif
812 
813 /* push an SRH on top of the current one */
input_action_end_b6(struct sk_buff * skb,struct seg6_local_lwt * slwt)814 static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
815 {
816 	struct ipv6_sr_hdr *srh;
817 	int err = -EINVAL;
818 
819 	srh = get_and_validate_srh(skb);
820 	if (!srh)
821 		goto drop;
822 
823 	err = seg6_do_srh_inline(skb, slwt->srh);
824 	if (err)
825 		goto drop;
826 
827 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
828 
829 	seg6_lookup_nexthop(skb, NULL, 0);
830 
831 	return dst_input(skb);
832 
833 drop:
834 	kfree_skb(skb);
835 	return err;
836 }
837 
838 /* encapsulate within an outer IPv6 header and a specified SRH */
input_action_end_b6_encap(struct sk_buff * skb,struct seg6_local_lwt * slwt)839 static int input_action_end_b6_encap(struct sk_buff *skb,
840 				     struct seg6_local_lwt *slwt)
841 {
842 	struct ipv6_sr_hdr *srh;
843 	int err = -EINVAL;
844 
845 	srh = get_and_validate_srh(skb);
846 	if (!srh)
847 		goto drop;
848 
849 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
850 
851 	skb_reset_inner_headers(skb);
852 	skb->encapsulation = 1;
853 
854 	err = seg6_do_srh_encap(skb, slwt->srh, IPPROTO_IPV6);
855 	if (err)
856 		goto drop;
857 
858 	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
859 
860 	seg6_lookup_nexthop(skb, NULL, 0);
861 
862 	return dst_input(skb);
863 
864 drop:
865 	kfree_skb(skb);
866 	return err;
867 }
868 
869 DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states);
870 
seg6_bpf_has_valid_srh(struct sk_buff * skb)871 bool seg6_bpf_has_valid_srh(struct sk_buff *skb)
872 {
873 	struct seg6_bpf_srh_state *srh_state =
874 		this_cpu_ptr(&seg6_bpf_srh_states);
875 	struct ipv6_sr_hdr *srh = srh_state->srh;
876 
877 	if (unlikely(srh == NULL))
878 		return false;
879 
880 	if (unlikely(!srh_state->valid)) {
881 		if ((srh_state->hdrlen & 7) != 0)
882 			return false;
883 
884 		srh->hdrlen = (u8)(srh_state->hdrlen >> 3);
885 		if (!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3, true))
886 			return false;
887 
888 		srh_state->valid = true;
889 	}
890 
891 	return true;
892 }
893 
input_action_end_bpf(struct sk_buff * skb,struct seg6_local_lwt * slwt)894 static int input_action_end_bpf(struct sk_buff *skb,
895 				struct seg6_local_lwt *slwt)
896 {
897 	struct seg6_bpf_srh_state *srh_state =
898 		this_cpu_ptr(&seg6_bpf_srh_states);
899 	struct ipv6_sr_hdr *srh;
900 	int ret;
901 
902 	srh = get_and_validate_srh(skb);
903 	if (!srh) {
904 		kfree_skb(skb);
905 		return -EINVAL;
906 	}
907 	advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
908 
909 	/* preempt_disable is needed to protect the per-CPU buffer srh_state,
910 	 * which is also accessed by the bpf_lwt_seg6_* helpers
911 	 */
912 	preempt_disable();
913 	srh_state->srh = srh;
914 	srh_state->hdrlen = srh->hdrlen << 3;
915 	srh_state->valid = true;
916 
917 	rcu_read_lock();
918 	bpf_compute_data_pointers(skb);
919 	ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb);
920 	rcu_read_unlock();
921 
922 	switch (ret) {
923 	case BPF_OK:
924 	case BPF_REDIRECT:
925 		break;
926 	case BPF_DROP:
927 		goto drop;
928 	default:
929 		pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret);
930 		goto drop;
931 	}
932 
933 	if (srh_state->srh && !seg6_bpf_has_valid_srh(skb))
934 		goto drop;
935 
936 	preempt_enable();
937 	if (ret != BPF_REDIRECT)
938 		seg6_lookup_nexthop(skb, NULL, 0);
939 
940 	return dst_input(skb);
941 
942 drop:
943 	preempt_enable();
944 	kfree_skb(skb);
945 	return -EINVAL;
946 }
947 
948 static struct seg6_action_desc seg6_action_table[] = {
949 	{
950 		.action		= SEG6_LOCAL_ACTION_END,
951 		.attrs		= 0,
952 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
953 		.input		= input_action_end,
954 	},
955 	{
956 		.action		= SEG6_LOCAL_ACTION_END_X,
957 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_NH6),
958 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
959 		.input		= input_action_end_x,
960 	},
961 	{
962 		.action		= SEG6_LOCAL_ACTION_END_T,
963 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_TABLE),
964 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
965 		.input		= input_action_end_t,
966 	},
967 	{
968 		.action		= SEG6_LOCAL_ACTION_END_DX2,
969 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_OIF),
970 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
971 		.input		= input_action_end_dx2,
972 	},
973 	{
974 		.action		= SEG6_LOCAL_ACTION_END_DX6,
975 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_NH6),
976 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
977 		.input		= input_action_end_dx6,
978 	},
979 	{
980 		.action		= SEG6_LOCAL_ACTION_END_DX4,
981 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_NH4),
982 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
983 		.input		= input_action_end_dx4,
984 	},
985 	{
986 		.action		= SEG6_LOCAL_ACTION_END_DT4,
987 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE),
988 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
989 #ifdef CONFIG_NET_L3_MASTER_DEV
990 		.input		= input_action_end_dt4,
991 		.slwt_ops	= {
992 					.build_state = seg6_end_dt4_build,
993 				  },
994 #endif
995 	},
996 	{
997 		.action		= SEG6_LOCAL_ACTION_END_DT6,
998 #ifdef CONFIG_NET_L3_MASTER_DEV
999 		.attrs		= 0,
1000 		.optattrs	= SEG6_F_LOCAL_COUNTERS		|
1001 				  SEG6_F_ATTR(SEG6_LOCAL_TABLE) |
1002 				  SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE),
1003 		.slwt_ops	= {
1004 					.build_state = seg6_end_dt6_build,
1005 				  },
1006 #else
1007 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_TABLE),
1008 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
1009 #endif
1010 		.input		= input_action_end_dt6,
1011 	},
1012 	{
1013 		.action		= SEG6_LOCAL_ACTION_END_DT46,
1014 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE),
1015 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
1016 #ifdef CONFIG_NET_L3_MASTER_DEV
1017 		.input		= input_action_end_dt46,
1018 		.slwt_ops	= {
1019 					.build_state = seg6_end_dt46_build,
1020 				  },
1021 #endif
1022 	},
1023 	{
1024 		.action		= SEG6_LOCAL_ACTION_END_B6,
1025 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_SRH),
1026 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
1027 		.input		= input_action_end_b6,
1028 	},
1029 	{
1030 		.action		= SEG6_LOCAL_ACTION_END_B6_ENCAP,
1031 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_SRH),
1032 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
1033 		.input		= input_action_end_b6_encap,
1034 		.static_headroom	= sizeof(struct ipv6hdr),
1035 	},
1036 	{
1037 		.action		= SEG6_LOCAL_ACTION_END_BPF,
1038 		.attrs		= SEG6_F_ATTR(SEG6_LOCAL_BPF),
1039 		.optattrs	= SEG6_F_LOCAL_COUNTERS,
1040 		.input		= input_action_end_bpf,
1041 	},
1042 
1043 };
1044 
__get_action_desc(int action)1045 static struct seg6_action_desc *__get_action_desc(int action)
1046 {
1047 	struct seg6_action_desc *desc;
1048 	int i, count;
1049 
1050 	count = ARRAY_SIZE(seg6_action_table);
1051 	for (i = 0; i < count; i++) {
1052 		desc = &seg6_action_table[i];
1053 		if (desc->action == action)
1054 			return desc;
1055 	}
1056 
1057 	return NULL;
1058 }
1059 
seg6_lwtunnel_counters_enabled(struct seg6_local_lwt * slwt)1060 static bool seg6_lwtunnel_counters_enabled(struct seg6_local_lwt *slwt)
1061 {
1062 	return slwt->parsed_optattrs & SEG6_F_LOCAL_COUNTERS;
1063 }
1064 
seg6_local_update_counters(struct seg6_local_lwt * slwt,unsigned int len,int err)1065 static void seg6_local_update_counters(struct seg6_local_lwt *slwt,
1066 				       unsigned int len, int err)
1067 {
1068 	struct pcpu_seg6_local_counters *pcounters;
1069 
1070 	pcounters = this_cpu_ptr(slwt->pcpu_counters);
1071 	u64_stats_update_begin(&pcounters->syncp);
1072 
1073 	if (likely(!err)) {
1074 		u64_stats_inc(&pcounters->packets);
1075 		u64_stats_add(&pcounters->bytes, len);
1076 	} else {
1077 		u64_stats_inc(&pcounters->errors);
1078 	}
1079 
1080 	u64_stats_update_end(&pcounters->syncp);
1081 }
1082 
seg6_local_input_core(struct net * net,struct sock * sk,struct sk_buff * skb)1083 static int seg6_local_input_core(struct net *net, struct sock *sk,
1084 				 struct sk_buff *skb)
1085 {
1086 	struct dst_entry *orig_dst = skb_dst(skb);
1087 	struct seg6_action_desc *desc;
1088 	struct seg6_local_lwt *slwt;
1089 	unsigned int len = skb->len;
1090 	int rc;
1091 
1092 	slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
1093 	desc = slwt->desc;
1094 
1095 	rc = desc->input(skb, slwt);
1096 
1097 	if (!seg6_lwtunnel_counters_enabled(slwt))
1098 		return rc;
1099 
1100 	seg6_local_update_counters(slwt, len, rc);
1101 
1102 	return rc;
1103 }
1104 
seg6_local_input(struct sk_buff * skb)1105 static int seg6_local_input(struct sk_buff *skb)
1106 {
1107 	if (skb->protocol != htons(ETH_P_IPV6)) {
1108 		kfree_skb(skb);
1109 		return -EINVAL;
1110 	}
1111 
1112 	if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
1113 		return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
1114 			       dev_net(skb->dev), NULL, skb, skb->dev, NULL,
1115 			       seg6_local_input_core);
1116 
1117 	return seg6_local_input_core(dev_net(skb->dev), NULL, skb);
1118 }
1119 
1120 static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = {
1121 	[SEG6_LOCAL_ACTION]	= { .type = NLA_U32 },
1122 	[SEG6_LOCAL_SRH]	= { .type = NLA_BINARY },
1123 	[SEG6_LOCAL_TABLE]	= { .type = NLA_U32 },
1124 	[SEG6_LOCAL_VRFTABLE]	= { .type = NLA_U32 },
1125 	[SEG6_LOCAL_NH4]	= { .type = NLA_BINARY,
1126 				    .len = sizeof(struct in_addr) },
1127 	[SEG6_LOCAL_NH6]	= { .type = NLA_BINARY,
1128 				    .len = sizeof(struct in6_addr) },
1129 	[SEG6_LOCAL_IIF]	= { .type = NLA_U32 },
1130 	[SEG6_LOCAL_OIF]	= { .type = NLA_U32 },
1131 	[SEG6_LOCAL_BPF]	= { .type = NLA_NESTED },
1132 	[SEG6_LOCAL_COUNTERS]	= { .type = NLA_NESTED },
1133 };
1134 
parse_nla_srh(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1135 static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1136 {
1137 	struct ipv6_sr_hdr *srh;
1138 	int len;
1139 
1140 	srh = nla_data(attrs[SEG6_LOCAL_SRH]);
1141 	len = nla_len(attrs[SEG6_LOCAL_SRH]);
1142 
1143 	/* SRH must contain at least one segment */
1144 	if (len < sizeof(*srh) + sizeof(struct in6_addr))
1145 		return -EINVAL;
1146 
1147 	if (!seg6_validate_srh(srh, len, false))
1148 		return -EINVAL;
1149 
1150 	slwt->srh = kmemdup(srh, len, GFP_KERNEL);
1151 	if (!slwt->srh)
1152 		return -ENOMEM;
1153 
1154 	slwt->headroom += len;
1155 
1156 	return 0;
1157 }
1158 
put_nla_srh(struct sk_buff * skb,struct seg6_local_lwt * slwt)1159 static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1160 {
1161 	struct ipv6_sr_hdr *srh;
1162 	struct nlattr *nla;
1163 	int len;
1164 
1165 	srh = slwt->srh;
1166 	len = (srh->hdrlen + 1) << 3;
1167 
1168 	nla = nla_reserve(skb, SEG6_LOCAL_SRH, len);
1169 	if (!nla)
1170 		return -EMSGSIZE;
1171 
1172 	memcpy(nla_data(nla), srh, len);
1173 
1174 	return 0;
1175 }
1176 
cmp_nla_srh(struct seg6_local_lwt * a,struct seg6_local_lwt * b)1177 static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1178 {
1179 	int len = (a->srh->hdrlen + 1) << 3;
1180 
1181 	if (len != ((b->srh->hdrlen + 1) << 3))
1182 		return 1;
1183 
1184 	return memcmp(a->srh, b->srh, len);
1185 }
1186 
destroy_attr_srh(struct seg6_local_lwt * slwt)1187 static void destroy_attr_srh(struct seg6_local_lwt *slwt)
1188 {
1189 	kfree(slwt->srh);
1190 }
1191 
parse_nla_table(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1192 static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1193 {
1194 	slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]);
1195 
1196 	return 0;
1197 }
1198 
put_nla_table(struct sk_buff * skb,struct seg6_local_lwt * slwt)1199 static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1200 {
1201 	if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table))
1202 		return -EMSGSIZE;
1203 
1204 	return 0;
1205 }
1206 
cmp_nla_table(struct seg6_local_lwt * a,struct seg6_local_lwt * b)1207 static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1208 {
1209 	if (a->table != b->table)
1210 		return 1;
1211 
1212 	return 0;
1213 }
1214 
1215 static struct
seg6_possible_end_dt_info(struct seg6_local_lwt * slwt)1216 seg6_end_dt_info *seg6_possible_end_dt_info(struct seg6_local_lwt *slwt)
1217 {
1218 #ifdef CONFIG_NET_L3_MASTER_DEV
1219 	return &slwt->dt_info;
1220 #else
1221 	return ERR_PTR(-EOPNOTSUPP);
1222 #endif
1223 }
1224 
parse_nla_vrftable(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1225 static int parse_nla_vrftable(struct nlattr **attrs,
1226 			      struct seg6_local_lwt *slwt)
1227 {
1228 	struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt);
1229 
1230 	if (IS_ERR(info))
1231 		return PTR_ERR(info);
1232 
1233 	info->vrf_table = nla_get_u32(attrs[SEG6_LOCAL_VRFTABLE]);
1234 
1235 	return 0;
1236 }
1237 
put_nla_vrftable(struct sk_buff * skb,struct seg6_local_lwt * slwt)1238 static int put_nla_vrftable(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1239 {
1240 	struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt);
1241 
1242 	if (IS_ERR(info))
1243 		return PTR_ERR(info);
1244 
1245 	if (nla_put_u32(skb, SEG6_LOCAL_VRFTABLE, info->vrf_table))
1246 		return -EMSGSIZE;
1247 
1248 	return 0;
1249 }
1250 
cmp_nla_vrftable(struct seg6_local_lwt * a,struct seg6_local_lwt * b)1251 static int cmp_nla_vrftable(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1252 {
1253 	struct seg6_end_dt_info *info_a = seg6_possible_end_dt_info(a);
1254 	struct seg6_end_dt_info *info_b = seg6_possible_end_dt_info(b);
1255 
1256 	if (info_a->vrf_table != info_b->vrf_table)
1257 		return 1;
1258 
1259 	return 0;
1260 }
1261 
parse_nla_nh4(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1262 static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1263 {
1264 	memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]),
1265 	       sizeof(struct in_addr));
1266 
1267 	return 0;
1268 }
1269 
put_nla_nh4(struct sk_buff * skb,struct seg6_local_lwt * slwt)1270 static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1271 {
1272 	struct nlattr *nla;
1273 
1274 	nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr));
1275 	if (!nla)
1276 		return -EMSGSIZE;
1277 
1278 	memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr));
1279 
1280 	return 0;
1281 }
1282 
cmp_nla_nh4(struct seg6_local_lwt * a,struct seg6_local_lwt * b)1283 static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1284 {
1285 	return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr));
1286 }
1287 
parse_nla_nh6(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1288 static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1289 {
1290 	memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]),
1291 	       sizeof(struct in6_addr));
1292 
1293 	return 0;
1294 }
1295 
put_nla_nh6(struct sk_buff * skb,struct seg6_local_lwt * slwt)1296 static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1297 {
1298 	struct nlattr *nla;
1299 
1300 	nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr));
1301 	if (!nla)
1302 		return -EMSGSIZE;
1303 
1304 	memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr));
1305 
1306 	return 0;
1307 }
1308 
cmp_nla_nh6(struct seg6_local_lwt * a,struct seg6_local_lwt * b)1309 static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1310 {
1311 	return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr));
1312 }
1313 
parse_nla_iif(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1314 static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1315 {
1316 	slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]);
1317 
1318 	return 0;
1319 }
1320 
put_nla_iif(struct sk_buff * skb,struct seg6_local_lwt * slwt)1321 static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1322 {
1323 	if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif))
1324 		return -EMSGSIZE;
1325 
1326 	return 0;
1327 }
1328 
cmp_nla_iif(struct seg6_local_lwt * a,struct seg6_local_lwt * b)1329 static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1330 {
1331 	if (a->iif != b->iif)
1332 		return 1;
1333 
1334 	return 0;
1335 }
1336 
parse_nla_oif(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1337 static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1338 {
1339 	slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]);
1340 
1341 	return 0;
1342 }
1343 
put_nla_oif(struct sk_buff * skb,struct seg6_local_lwt * slwt)1344 static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1345 {
1346 	if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif))
1347 		return -EMSGSIZE;
1348 
1349 	return 0;
1350 }
1351 
cmp_nla_oif(struct seg6_local_lwt * a,struct seg6_local_lwt * b)1352 static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1353 {
1354 	if (a->oif != b->oif)
1355 		return 1;
1356 
1357 	return 0;
1358 }
1359 
1360 #define MAX_PROG_NAME 256
1361 static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = {
1362 	[SEG6_LOCAL_BPF_PROG]	   = { .type = NLA_U32, },
1363 	[SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING,
1364 				       .len = MAX_PROG_NAME },
1365 };
1366 
parse_nla_bpf(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1367 static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1368 {
1369 	struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1];
1370 	struct bpf_prog *p;
1371 	int ret;
1372 	u32 fd;
1373 
1374 	ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX,
1375 					  attrs[SEG6_LOCAL_BPF],
1376 					  bpf_prog_policy, NULL);
1377 	if (ret < 0)
1378 		return ret;
1379 
1380 	if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME])
1381 		return -EINVAL;
1382 
1383 	slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL);
1384 	if (!slwt->bpf.name)
1385 		return -ENOMEM;
1386 
1387 	fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]);
1388 	p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL);
1389 	if (IS_ERR(p)) {
1390 		kfree(slwt->bpf.name);
1391 		return PTR_ERR(p);
1392 	}
1393 
1394 	slwt->bpf.prog = p;
1395 	return 0;
1396 }
1397 
put_nla_bpf(struct sk_buff * skb,struct seg6_local_lwt * slwt)1398 static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1399 {
1400 	struct nlattr *nest;
1401 
1402 	if (!slwt->bpf.prog)
1403 		return 0;
1404 
1405 	nest = nla_nest_start_noflag(skb, SEG6_LOCAL_BPF);
1406 	if (!nest)
1407 		return -EMSGSIZE;
1408 
1409 	if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id))
1410 		return -EMSGSIZE;
1411 
1412 	if (slwt->bpf.name &&
1413 	    nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name))
1414 		return -EMSGSIZE;
1415 
1416 	return nla_nest_end(skb, nest);
1417 }
1418 
cmp_nla_bpf(struct seg6_local_lwt * a,struct seg6_local_lwt * b)1419 static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1420 {
1421 	if (!a->bpf.name && !b->bpf.name)
1422 		return 0;
1423 
1424 	if (!a->bpf.name || !b->bpf.name)
1425 		return 1;
1426 
1427 	return strcmp(a->bpf.name, b->bpf.name);
1428 }
1429 
destroy_attr_bpf(struct seg6_local_lwt * slwt)1430 static void destroy_attr_bpf(struct seg6_local_lwt *slwt)
1431 {
1432 	kfree(slwt->bpf.name);
1433 	if (slwt->bpf.prog)
1434 		bpf_prog_put(slwt->bpf.prog);
1435 }
1436 
1437 static const struct
1438 nla_policy seg6_local_counters_policy[SEG6_LOCAL_CNT_MAX + 1] = {
1439 	[SEG6_LOCAL_CNT_PACKETS]	= { .type = NLA_U64 },
1440 	[SEG6_LOCAL_CNT_BYTES]		= { .type = NLA_U64 },
1441 	[SEG6_LOCAL_CNT_ERRORS]		= { .type = NLA_U64 },
1442 };
1443 
parse_nla_counters(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1444 static int parse_nla_counters(struct nlattr **attrs,
1445 			      struct seg6_local_lwt *slwt)
1446 {
1447 	struct pcpu_seg6_local_counters __percpu *pcounters;
1448 	struct nlattr *tb[SEG6_LOCAL_CNT_MAX + 1];
1449 	int ret;
1450 
1451 	ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_CNT_MAX,
1452 					  attrs[SEG6_LOCAL_COUNTERS],
1453 					  seg6_local_counters_policy, NULL);
1454 	if (ret < 0)
1455 		return ret;
1456 
1457 	/* basic support for SRv6 Behavior counters requires at least:
1458 	 * packets, bytes and errors.
1459 	 */
1460 	if (!tb[SEG6_LOCAL_CNT_PACKETS] || !tb[SEG6_LOCAL_CNT_BYTES] ||
1461 	    !tb[SEG6_LOCAL_CNT_ERRORS])
1462 		return -EINVAL;
1463 
1464 	/* counters are always zero initialized */
1465 	pcounters = seg6_local_alloc_pcpu_counters(GFP_KERNEL);
1466 	if (!pcounters)
1467 		return -ENOMEM;
1468 
1469 	slwt->pcpu_counters = pcounters;
1470 
1471 	return 0;
1472 }
1473 
seg6_local_fill_nla_counters(struct sk_buff * skb,struct seg6_local_counters * counters)1474 static int seg6_local_fill_nla_counters(struct sk_buff *skb,
1475 					struct seg6_local_counters *counters)
1476 {
1477 	if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_PACKETS, counters->packets,
1478 			      SEG6_LOCAL_CNT_PAD))
1479 		return -EMSGSIZE;
1480 
1481 	if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_BYTES, counters->bytes,
1482 			      SEG6_LOCAL_CNT_PAD))
1483 		return -EMSGSIZE;
1484 
1485 	if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_ERRORS, counters->errors,
1486 			      SEG6_LOCAL_CNT_PAD))
1487 		return -EMSGSIZE;
1488 
1489 	return 0;
1490 }
1491 
put_nla_counters(struct sk_buff * skb,struct seg6_local_lwt * slwt)1492 static int put_nla_counters(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1493 {
1494 	struct seg6_local_counters counters = { 0, 0, 0 };
1495 	struct nlattr *nest;
1496 	int rc, i;
1497 
1498 	nest = nla_nest_start(skb, SEG6_LOCAL_COUNTERS);
1499 	if (!nest)
1500 		return -EMSGSIZE;
1501 
1502 	for_each_possible_cpu(i) {
1503 		struct pcpu_seg6_local_counters *pcounters;
1504 		u64 packets, bytes, errors;
1505 		unsigned int start;
1506 
1507 		pcounters = per_cpu_ptr(slwt->pcpu_counters, i);
1508 		do {
1509 			start = u64_stats_fetch_begin_irq(&pcounters->syncp);
1510 
1511 			packets = u64_stats_read(&pcounters->packets);
1512 			bytes = u64_stats_read(&pcounters->bytes);
1513 			errors = u64_stats_read(&pcounters->errors);
1514 
1515 		} while (u64_stats_fetch_retry_irq(&pcounters->syncp, start));
1516 
1517 		counters.packets += packets;
1518 		counters.bytes += bytes;
1519 		counters.errors += errors;
1520 	}
1521 
1522 	rc = seg6_local_fill_nla_counters(skb, &counters);
1523 	if (rc < 0) {
1524 		nla_nest_cancel(skb, nest);
1525 		return rc;
1526 	}
1527 
1528 	return nla_nest_end(skb, nest);
1529 }
1530 
cmp_nla_counters(struct seg6_local_lwt * a,struct seg6_local_lwt * b)1531 static int cmp_nla_counters(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1532 {
1533 	/* a and b are equal if both have pcpu_counters set or not */
1534 	return (!!((unsigned long)a->pcpu_counters)) ^
1535 		(!!((unsigned long)b->pcpu_counters));
1536 }
1537 
destroy_attr_counters(struct seg6_local_lwt * slwt)1538 static void destroy_attr_counters(struct seg6_local_lwt *slwt)
1539 {
1540 	free_percpu(slwt->pcpu_counters);
1541 }
1542 
1543 struct seg6_action_param {
1544 	int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt);
1545 	int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
1546 	int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b);
1547 
1548 	/* optional destroy() callback useful for releasing resources which
1549 	 * have been previously acquired in the corresponding parse()
1550 	 * function.
1551 	 */
1552 	void (*destroy)(struct seg6_local_lwt *slwt);
1553 };
1554 
1555 static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = {
1556 	[SEG6_LOCAL_SRH]	= { .parse = parse_nla_srh,
1557 				    .put = put_nla_srh,
1558 				    .cmp = cmp_nla_srh,
1559 				    .destroy = destroy_attr_srh },
1560 
1561 	[SEG6_LOCAL_TABLE]	= { .parse = parse_nla_table,
1562 				    .put = put_nla_table,
1563 				    .cmp = cmp_nla_table },
1564 
1565 	[SEG6_LOCAL_NH4]	= { .parse = parse_nla_nh4,
1566 				    .put = put_nla_nh4,
1567 				    .cmp = cmp_nla_nh4 },
1568 
1569 	[SEG6_LOCAL_NH6]	= { .parse = parse_nla_nh6,
1570 				    .put = put_nla_nh6,
1571 				    .cmp = cmp_nla_nh6 },
1572 
1573 	[SEG6_LOCAL_IIF]	= { .parse = parse_nla_iif,
1574 				    .put = put_nla_iif,
1575 				    .cmp = cmp_nla_iif },
1576 
1577 	[SEG6_LOCAL_OIF]	= { .parse = parse_nla_oif,
1578 				    .put = put_nla_oif,
1579 				    .cmp = cmp_nla_oif },
1580 
1581 	[SEG6_LOCAL_BPF]	= { .parse = parse_nla_bpf,
1582 				    .put = put_nla_bpf,
1583 				    .cmp = cmp_nla_bpf,
1584 				    .destroy = destroy_attr_bpf },
1585 
1586 	[SEG6_LOCAL_VRFTABLE]	= { .parse = parse_nla_vrftable,
1587 				    .put = put_nla_vrftable,
1588 				    .cmp = cmp_nla_vrftable },
1589 
1590 	[SEG6_LOCAL_COUNTERS]	= { .parse = parse_nla_counters,
1591 				    .put = put_nla_counters,
1592 				    .cmp = cmp_nla_counters,
1593 				    .destroy = destroy_attr_counters },
1594 };
1595 
1596 /* call the destroy() callback (if available) for each set attribute in
1597  * @parsed_attrs, starting from the first attribute up to the @max_parsed
1598  * (excluded) attribute.
1599  */
__destroy_attrs(unsigned long parsed_attrs,int max_parsed,struct seg6_local_lwt * slwt)1600 static void __destroy_attrs(unsigned long parsed_attrs, int max_parsed,
1601 			    struct seg6_local_lwt *slwt)
1602 {
1603 	struct seg6_action_param *param;
1604 	int i;
1605 
1606 	/* Every required seg6local attribute is identified by an ID which is
1607 	 * encoded as a flag (i.e: 1 << ID) in the 'attrs' bitmask;
1608 	 *
1609 	 * We scan the 'parsed_attrs' bitmask, starting from the first attribute
1610 	 * up to the @max_parsed (excluded) attribute.
1611 	 * For each set attribute, we retrieve the corresponding destroy()
1612 	 * callback. If the callback is not available, then we skip to the next
1613 	 * attribute; otherwise, we call the destroy() callback.
1614 	 */
1615 	for (i = 0; i < max_parsed; ++i) {
1616 		if (!(parsed_attrs & SEG6_F_ATTR(i)))
1617 			continue;
1618 
1619 		param = &seg6_action_params[i];
1620 
1621 		if (param->destroy)
1622 			param->destroy(slwt);
1623 	}
1624 }
1625 
1626 /* release all the resources that may have been acquired during parsing
1627  * operations.
1628  */
destroy_attrs(struct seg6_local_lwt * slwt)1629 static void destroy_attrs(struct seg6_local_lwt *slwt)
1630 {
1631 	unsigned long attrs = slwt->desc->attrs | slwt->parsed_optattrs;
1632 
1633 	__destroy_attrs(attrs, SEG6_LOCAL_MAX + 1, slwt);
1634 }
1635 
parse_nla_optional_attrs(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1636 static int parse_nla_optional_attrs(struct nlattr **attrs,
1637 				    struct seg6_local_lwt *slwt)
1638 {
1639 	struct seg6_action_desc *desc = slwt->desc;
1640 	unsigned long parsed_optattrs = 0;
1641 	struct seg6_action_param *param;
1642 	int err, i;
1643 
1644 	for (i = 0; i < SEG6_LOCAL_MAX + 1; ++i) {
1645 		if (!(desc->optattrs & SEG6_F_ATTR(i)) || !attrs[i])
1646 			continue;
1647 
1648 		/* once here, the i-th attribute is provided by the
1649 		 * userspace AND it is identified optional as well.
1650 		 */
1651 		param = &seg6_action_params[i];
1652 
1653 		err = param->parse(attrs, slwt);
1654 		if (err < 0)
1655 			goto parse_optattrs_err;
1656 
1657 		/* current attribute has been correctly parsed */
1658 		parsed_optattrs |= SEG6_F_ATTR(i);
1659 	}
1660 
1661 	/* store in the tunnel state all the optional attributed successfully
1662 	 * parsed.
1663 	 */
1664 	slwt->parsed_optattrs = parsed_optattrs;
1665 
1666 	return 0;
1667 
1668 parse_optattrs_err:
1669 	__destroy_attrs(parsed_optattrs, i, slwt);
1670 
1671 	return err;
1672 }
1673 
1674 /* call the custom constructor of the behavior during its initialization phase
1675  * and after that all its attributes have been parsed successfully.
1676  */
1677 static int
seg6_local_lwtunnel_build_state(struct seg6_local_lwt * slwt,const void * cfg,struct netlink_ext_ack * extack)1678 seg6_local_lwtunnel_build_state(struct seg6_local_lwt *slwt, const void *cfg,
1679 				struct netlink_ext_ack *extack)
1680 {
1681 	struct seg6_action_desc *desc = slwt->desc;
1682 	struct seg6_local_lwtunnel_ops *ops;
1683 
1684 	ops = &desc->slwt_ops;
1685 	if (!ops->build_state)
1686 		return 0;
1687 
1688 	return ops->build_state(slwt, cfg, extack);
1689 }
1690 
1691 /* call the custom destructor of the behavior which is invoked before the
1692  * tunnel is going to be destroyed.
1693  */
seg6_local_lwtunnel_destroy_state(struct seg6_local_lwt * slwt)1694 static void seg6_local_lwtunnel_destroy_state(struct seg6_local_lwt *slwt)
1695 {
1696 	struct seg6_action_desc *desc = slwt->desc;
1697 	struct seg6_local_lwtunnel_ops *ops;
1698 
1699 	ops = &desc->slwt_ops;
1700 	if (!ops->destroy_state)
1701 		return;
1702 
1703 	ops->destroy_state(slwt);
1704 }
1705 
parse_nla_action(struct nlattr ** attrs,struct seg6_local_lwt * slwt)1706 static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt)
1707 {
1708 	struct seg6_action_param *param;
1709 	struct seg6_action_desc *desc;
1710 	unsigned long invalid_attrs;
1711 	int i, err;
1712 
1713 	desc = __get_action_desc(slwt->action);
1714 	if (!desc)
1715 		return -EINVAL;
1716 
1717 	if (!desc->input)
1718 		return -EOPNOTSUPP;
1719 
1720 	slwt->desc = desc;
1721 	slwt->headroom += desc->static_headroom;
1722 
1723 	/* Forcing the desc->optattrs *set* and the desc->attrs *set* to be
1724 	 * disjoined, this allow us to release acquired resources by optional
1725 	 * attributes and by required attributes independently from each other
1726 	 * without any interference.
1727 	 * In other terms, we are sure that we do not release some the acquired
1728 	 * resources twice.
1729 	 *
1730 	 * Note that if an attribute is configured both as required and as
1731 	 * optional, it means that the user has messed something up in the
1732 	 * seg6_action_table. Therefore, this check is required for SRv6
1733 	 * behaviors to work properly.
1734 	 */
1735 	invalid_attrs = desc->attrs & desc->optattrs;
1736 	if (invalid_attrs) {
1737 		WARN_ONCE(1,
1738 			  "An attribute cannot be both required AND optional");
1739 		return -EINVAL;
1740 	}
1741 
1742 	/* parse the required attributes */
1743 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1744 		if (desc->attrs & SEG6_F_ATTR(i)) {
1745 			if (!attrs[i])
1746 				return -EINVAL;
1747 
1748 			param = &seg6_action_params[i];
1749 
1750 			err = param->parse(attrs, slwt);
1751 			if (err < 0)
1752 				goto parse_attrs_err;
1753 		}
1754 	}
1755 
1756 	/* parse the optional attributes, if any */
1757 	err = parse_nla_optional_attrs(attrs, slwt);
1758 	if (err < 0)
1759 		goto parse_attrs_err;
1760 
1761 	return 0;
1762 
1763 parse_attrs_err:
1764 	/* release any resource that may have been acquired during the i-1
1765 	 * parse() operations.
1766 	 */
1767 	__destroy_attrs(desc->attrs, i, slwt);
1768 
1769 	return err;
1770 }
1771 
seg6_local_build_state(struct net * net,struct nlattr * nla,unsigned int family,const void * cfg,struct lwtunnel_state ** ts,struct netlink_ext_ack * extack)1772 static int seg6_local_build_state(struct net *net, struct nlattr *nla,
1773 				  unsigned int family, const void *cfg,
1774 				  struct lwtunnel_state **ts,
1775 				  struct netlink_ext_ack *extack)
1776 {
1777 	struct nlattr *tb[SEG6_LOCAL_MAX + 1];
1778 	struct lwtunnel_state *newts;
1779 	struct seg6_local_lwt *slwt;
1780 	int err;
1781 
1782 	if (family != AF_INET6)
1783 		return -EINVAL;
1784 
1785 	err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla,
1786 					  seg6_local_policy, extack);
1787 
1788 	if (err < 0)
1789 		return err;
1790 
1791 	if (!tb[SEG6_LOCAL_ACTION])
1792 		return -EINVAL;
1793 
1794 	newts = lwtunnel_state_alloc(sizeof(*slwt));
1795 	if (!newts)
1796 		return -ENOMEM;
1797 
1798 	slwt = seg6_local_lwtunnel(newts);
1799 	slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]);
1800 
1801 	err = parse_nla_action(tb, slwt);
1802 	if (err < 0)
1803 		goto out_free;
1804 
1805 	err = seg6_local_lwtunnel_build_state(slwt, cfg, extack);
1806 	if (err < 0)
1807 		goto out_destroy_attrs;
1808 
1809 	newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL;
1810 	newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT;
1811 	newts->headroom = slwt->headroom;
1812 
1813 	*ts = newts;
1814 
1815 	return 0;
1816 
1817 out_destroy_attrs:
1818 	destroy_attrs(slwt);
1819 out_free:
1820 	kfree(newts);
1821 	return err;
1822 }
1823 
seg6_local_destroy_state(struct lwtunnel_state * lwt)1824 static void seg6_local_destroy_state(struct lwtunnel_state *lwt)
1825 {
1826 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1827 
1828 	seg6_local_lwtunnel_destroy_state(slwt);
1829 
1830 	destroy_attrs(slwt);
1831 
1832 	return;
1833 }
1834 
seg6_local_fill_encap(struct sk_buff * skb,struct lwtunnel_state * lwt)1835 static int seg6_local_fill_encap(struct sk_buff *skb,
1836 				 struct lwtunnel_state *lwt)
1837 {
1838 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1839 	struct seg6_action_param *param;
1840 	unsigned long attrs;
1841 	int i, err;
1842 
1843 	if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action))
1844 		return -EMSGSIZE;
1845 
1846 	attrs = slwt->desc->attrs | slwt->parsed_optattrs;
1847 
1848 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1849 		if (attrs & SEG6_F_ATTR(i)) {
1850 			param = &seg6_action_params[i];
1851 			err = param->put(skb, slwt);
1852 			if (err < 0)
1853 				return err;
1854 		}
1855 	}
1856 
1857 	return 0;
1858 }
1859 
seg6_local_get_encap_size(struct lwtunnel_state * lwt)1860 static int seg6_local_get_encap_size(struct lwtunnel_state *lwt)
1861 {
1862 	struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
1863 	unsigned long attrs;
1864 	int nlsize;
1865 
1866 	nlsize = nla_total_size(4); /* action */
1867 
1868 	attrs = slwt->desc->attrs | slwt->parsed_optattrs;
1869 
1870 	if (attrs & SEG6_F_ATTR(SEG6_LOCAL_SRH))
1871 		nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3);
1872 
1873 	if (attrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE))
1874 		nlsize += nla_total_size(4);
1875 
1876 	if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH4))
1877 		nlsize += nla_total_size(4);
1878 
1879 	if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH6))
1880 		nlsize += nla_total_size(16);
1881 
1882 	if (attrs & SEG6_F_ATTR(SEG6_LOCAL_IIF))
1883 		nlsize += nla_total_size(4);
1884 
1885 	if (attrs & SEG6_F_ATTR(SEG6_LOCAL_OIF))
1886 		nlsize += nla_total_size(4);
1887 
1888 	if (attrs & SEG6_F_ATTR(SEG6_LOCAL_BPF))
1889 		nlsize += nla_total_size(sizeof(struct nlattr)) +
1890 		       nla_total_size(MAX_PROG_NAME) +
1891 		       nla_total_size(4);
1892 
1893 	if (attrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE))
1894 		nlsize += nla_total_size(4);
1895 
1896 	if (attrs & SEG6_F_LOCAL_COUNTERS)
1897 		nlsize += nla_total_size(0) + /* nest SEG6_LOCAL_COUNTERS */
1898 			  /* SEG6_LOCAL_CNT_PACKETS */
1899 			  nla_total_size_64bit(sizeof(__u64)) +
1900 			  /* SEG6_LOCAL_CNT_BYTES */
1901 			  nla_total_size_64bit(sizeof(__u64)) +
1902 			  /* SEG6_LOCAL_CNT_ERRORS */
1903 			  nla_total_size_64bit(sizeof(__u64));
1904 
1905 	return nlsize;
1906 }
1907 
seg6_local_cmp_encap(struct lwtunnel_state * a,struct lwtunnel_state * b)1908 static int seg6_local_cmp_encap(struct lwtunnel_state *a,
1909 				struct lwtunnel_state *b)
1910 {
1911 	struct seg6_local_lwt *slwt_a, *slwt_b;
1912 	struct seg6_action_param *param;
1913 	unsigned long attrs_a, attrs_b;
1914 	int i;
1915 
1916 	slwt_a = seg6_local_lwtunnel(a);
1917 	slwt_b = seg6_local_lwtunnel(b);
1918 
1919 	if (slwt_a->action != slwt_b->action)
1920 		return 1;
1921 
1922 	attrs_a = slwt_a->desc->attrs | slwt_a->parsed_optattrs;
1923 	attrs_b = slwt_b->desc->attrs | slwt_b->parsed_optattrs;
1924 
1925 	if (attrs_a != attrs_b)
1926 		return 1;
1927 
1928 	for (i = 0; i < SEG6_LOCAL_MAX + 1; i++) {
1929 		if (attrs_a & SEG6_F_ATTR(i)) {
1930 			param = &seg6_action_params[i];
1931 			if (param->cmp(slwt_a, slwt_b))
1932 				return 1;
1933 		}
1934 	}
1935 
1936 	return 0;
1937 }
1938 
1939 static const struct lwtunnel_encap_ops seg6_local_ops = {
1940 	.build_state	= seg6_local_build_state,
1941 	.destroy_state	= seg6_local_destroy_state,
1942 	.input		= seg6_local_input,
1943 	.fill_encap	= seg6_local_fill_encap,
1944 	.get_encap_size	= seg6_local_get_encap_size,
1945 	.cmp_encap	= seg6_local_cmp_encap,
1946 	.owner		= THIS_MODULE,
1947 };
1948 
seg6_local_init(void)1949 int __init seg6_local_init(void)
1950 {
1951 	/* If the max total number of defined attributes is reached, then your
1952 	 * kernel build stops here.
1953 	 *
1954 	 * This check is required to avoid arithmetic overflows when processing
1955 	 * behavior attributes and the maximum number of defined attributes
1956 	 * exceeds the allowed value.
1957 	 */
1958 	BUILD_BUG_ON(SEG6_LOCAL_MAX + 1 > BITS_PER_TYPE(unsigned long));
1959 
1960 	return lwtunnel_encap_add_ops(&seg6_local_ops,
1961 				      LWTUNNEL_ENCAP_SEG6_LOCAL);
1962 }
1963 
seg6_local_exit(void)1964 void seg6_local_exit(void)
1965 {
1966 	lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL);
1967 }
1968