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/filter.h> |
11 | #include <linux/types.h> |
12 | #include <linux/skbuff.h> |
13 | #include <linux/net.h> |
14 | #include <linux/module.h> |
15 | #include <net/ip.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> |
21 | #include <net/seg6.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> |
30 | #endif |
31 | #include <net/seg6_local.h> |
32 | #include <linux/etherdevice.h> |
33 | #include <linux/bpf.h> |
34 | #include <linux/netfilter.h> |
35 | |
36 | #define SEG6_F_ATTR(i) BIT(i) |
37 | |
38 | struct seg6_local_lwt; |
39 | |
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); |
45 | }; |
46 | |
47 | struct seg6_action_desc { |
48 | int action; |
49 | unsigned long attrs; |
50 | |
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. |
56 | * |
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 |
61 | * versa. |
62 | */ |
63 | unsigned long optattrs; |
64 | |
65 | int (*input)(struct sk_buff *skb, struct seg6_local_lwt *slwt); |
66 | int static_headroom; |
67 | |
68 | struct seg6_local_lwtunnel_ops slwt_ops; |
69 | }; |
70 | |
71 | struct bpf_lwt_prog { |
72 | struct bpf_prog *prog; |
73 | char *name; |
74 | }; |
75 | |
76 | /* default length values (expressed in bits) for both Locator-Block and |
77 | * Locator-Node Function. |
78 | * |
79 | * Both SEG6_LOCAL_LCBLOCK_DBITS and SEG6_LOCAL_LCNODE_FN_DBITS *must* be: |
80 | * i) greater than 0; |
81 | * ii) evenly divisible by 8. In other terms, the lengths of the |
82 | * Locator-Block and Locator-Node Function must be byte-aligned (we can |
83 | * relax this constraint in the future if really needed). |
84 | * |
85 | * Moreover, a third condition must hold: |
86 | * iii) SEG6_LOCAL_LCBLOCK_DBITS + SEG6_LOCAL_LCNODE_FN_DBITS <= 128. |
87 | * |
88 | * The correctness of SEG6_LOCAL_LCBLOCK_DBITS and SEG6_LOCAL_LCNODE_FN_DBITS |
89 | * values are checked during the kernel compilation. If the compilation stops, |
90 | * check the value of these parameters to see if they meet conditions (i), (ii) |
91 | * and (iii). |
92 | */ |
93 | #define SEG6_LOCAL_LCBLOCK_DBITS 32 |
94 | #define SEG6_LOCAL_LCNODE_FN_DBITS 16 |
95 | |
96 | /* The following next_csid_chk_{cntr,lcblock,lcblock_fn}_bits macros can be |
97 | * used directly to check whether the lengths (in bits) of Locator-Block and |
98 | * Locator-Node Function are valid according to (i), (ii), (iii). |
99 | */ |
100 | #define next_csid_chk_cntr_bits(blen, flen) \ |
101 | ((blen) + (flen) > 128) |
102 | |
103 | #define next_csid_chk_lcblock_bits(blen) \ |
104 | ({ \ |
105 | typeof(blen) __tmp = blen; \ |
106 | (!__tmp || __tmp > 120 || (__tmp & 0x07)); \ |
107 | }) |
108 | |
109 | #define next_csid_chk_lcnode_fn_bits(flen) \ |
110 | next_csid_chk_lcblock_bits(flen) |
111 | |
112 | /* flag indicating that flavors are set up for a given End* behavior */ |
113 | #define SEG6_F_LOCAL_FLAVORS SEG6_F_ATTR(SEG6_LOCAL_FLAVORS) |
114 | |
115 | #define SEG6_F_LOCAL_FLV_OP(flvname) BIT(SEG6_LOCAL_FLV_OP_##flvname) |
116 | #define SEG6_F_LOCAL_FLV_NEXT_CSID SEG6_F_LOCAL_FLV_OP(NEXT_CSID) |
117 | #define SEG6_F_LOCAL_FLV_PSP SEG6_F_LOCAL_FLV_OP(PSP) |
118 | |
119 | /* Supported RFC8986 Flavor operations are reported in this bitmask */ |
120 | #define SEG6_LOCAL_FLV8986_SUPP_OPS SEG6_F_LOCAL_FLV_PSP |
121 | |
122 | #define SEG6_LOCAL_END_FLV_SUPP_OPS (SEG6_F_LOCAL_FLV_NEXT_CSID | \ |
123 | SEG6_LOCAL_FLV8986_SUPP_OPS) |
124 | #define SEG6_LOCAL_END_X_FLV_SUPP_OPS SEG6_F_LOCAL_FLV_NEXT_CSID |
125 | |
126 | struct seg6_flavors_info { |
127 | /* Flavor operations */ |
128 | __u32 flv_ops; |
129 | |
130 | /* Locator-Block length, expressed in bits */ |
131 | __u8 lcblock_bits; |
132 | /* Locator-Node Function length, expressed in bits*/ |
133 | __u8 lcnode_func_bits; |
134 | }; |
135 | |
136 | enum seg6_end_dt_mode { |
137 | DT_INVALID_MODE = -EINVAL, |
138 | DT_LEGACY_MODE = 0, |
139 | DT_VRF_MODE = 1, |
140 | }; |
141 | |
142 | struct seg6_end_dt_info { |
143 | enum seg6_end_dt_mode mode; |
144 | |
145 | struct net *net; |
146 | /* VRF device associated to the routing table used by the SRv6 |
147 | * End.DT4/DT6 behavior for routing IPv4/IPv6 packets. |
148 | */ |
149 | int vrf_ifindex; |
150 | int vrf_table; |
151 | |
152 | /* tunneled packet family (IPv4 or IPv6). |
153 | * Protocol and header length are inferred from family. |
154 | */ |
155 | u16 family; |
156 | }; |
157 | |
158 | struct pcpu_seg6_local_counters { |
159 | u64_stats_t packets; |
160 | u64_stats_t bytes; |
161 | u64_stats_t errors; |
162 | |
163 | struct u64_stats_sync syncp; |
164 | }; |
165 | |
166 | /* This struct groups all the SRv6 Behavior counters supported so far. |
167 | * |
168 | * put_nla_counters() makes use of this data structure to collect all counter |
169 | * values after the per-CPU counter evaluation has been performed. |
170 | * Finally, each counter value (in seg6_local_counters) is stored in the |
171 | * corresponding netlink attribute and sent to user space. |
172 | * |
173 | * NB: we don't want to expose this structure to user space! |
174 | */ |
175 | struct seg6_local_counters { |
176 | __u64 packets; |
177 | __u64 bytes; |
178 | __u64 errors; |
179 | }; |
180 | |
181 | #define seg6_local_alloc_pcpu_counters(__gfp) \ |
182 | __netdev_alloc_pcpu_stats(struct pcpu_seg6_local_counters, \ |
183 | ((__gfp) | __GFP_ZERO)) |
184 | |
185 | #define SEG6_F_LOCAL_COUNTERS SEG6_F_ATTR(SEG6_LOCAL_COUNTERS) |
186 | |
187 | struct seg6_local_lwt { |
188 | int action; |
189 | struct ipv6_sr_hdr *srh; |
190 | int table; |
191 | struct in_addr nh4; |
192 | struct in6_addr nh6; |
193 | int iif; |
194 | int oif; |
195 | struct bpf_lwt_prog bpf; |
196 | #ifdef CONFIG_NET_L3_MASTER_DEV |
197 | struct seg6_end_dt_info dt_info; |
198 | #endif |
199 | struct seg6_flavors_info flv_info; |
200 | |
201 | struct pcpu_seg6_local_counters __percpu *pcpu_counters; |
202 | |
203 | int headroom; |
204 | struct seg6_action_desc *desc; |
205 | /* unlike the required attrs, we have to track the optional attributes |
206 | * that have been effectively parsed. |
207 | */ |
208 | unsigned long parsed_optattrs; |
209 | }; |
210 | |
211 | static struct seg6_local_lwt *seg6_local_lwtunnel(struct lwtunnel_state *lwt) |
212 | { |
213 | return (struct seg6_local_lwt *)lwt->data; |
214 | } |
215 | |
216 | static struct ipv6_sr_hdr *get_and_validate_srh(struct sk_buff *skb) |
217 | { |
218 | struct ipv6_sr_hdr *srh; |
219 | |
220 | srh = seg6_get_srh(skb, flags: IP6_FH_F_SKIP_RH); |
221 | if (!srh) |
222 | return NULL; |
223 | |
224 | #ifdef CONFIG_IPV6_SEG6_HMAC |
225 | if (!seg6_hmac_validate_skb(skb)) |
226 | return NULL; |
227 | #endif |
228 | |
229 | return srh; |
230 | } |
231 | |
232 | static bool decap_and_validate(struct sk_buff *skb, int proto) |
233 | { |
234 | struct ipv6_sr_hdr *srh; |
235 | unsigned int off = 0; |
236 | |
237 | srh = seg6_get_srh(skb, flags: 0); |
238 | if (srh && srh->segments_left > 0) |
239 | return false; |
240 | |
241 | #ifdef CONFIG_IPV6_SEG6_HMAC |
242 | if (srh && !seg6_hmac_validate_skb(skb)) |
243 | return false; |
244 | #endif |
245 | |
246 | if (ipv6_find_hdr(skb, offset: &off, target: proto, NULL, NULL) < 0) |
247 | return false; |
248 | |
249 | if (!pskb_pull(skb, len: off)) |
250 | return false; |
251 | |
252 | skb_postpull_rcsum(skb, start: skb_network_header(skb), len: off); |
253 | |
254 | skb_reset_network_header(skb); |
255 | skb_reset_transport_header(skb); |
256 | if (iptunnel_pull_offloads(skb)) |
257 | return false; |
258 | |
259 | return true; |
260 | } |
261 | |
262 | static void advance_nextseg(struct ipv6_sr_hdr *srh, struct in6_addr *daddr) |
263 | { |
264 | struct in6_addr *addr; |
265 | |
266 | srh->segments_left--; |
267 | addr = srh->segments + srh->segments_left; |
268 | *daddr = *addr; |
269 | } |
270 | |
271 | static int |
272 | seg6_lookup_any_nexthop(struct sk_buff *skb, struct in6_addr *nhaddr, |
273 | u32 tbl_id, bool local_delivery) |
274 | { |
275 | struct net *net = dev_net(dev: skb->dev); |
276 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
277 | int flags = RT6_LOOKUP_F_HAS_SADDR; |
278 | struct dst_entry *dst = NULL; |
279 | struct rt6_info *rt; |
280 | struct flowi6 fl6; |
281 | int dev_flags = 0; |
282 | |
283 | memset(&fl6, 0, sizeof(fl6)); |
284 | fl6.flowi6_iif = skb->dev->ifindex; |
285 | fl6.daddr = nhaddr ? *nhaddr : hdr->daddr; |
286 | fl6.saddr = hdr->saddr; |
287 | fl6.flowlabel = ip6_flowinfo(hdr); |
288 | fl6.flowi6_mark = skb->mark; |
289 | fl6.flowi6_proto = hdr->nexthdr; |
290 | |
291 | if (nhaddr) |
292 | fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH; |
293 | |
294 | if (!tbl_id) { |
295 | dst = ip6_route_input_lookup(net, dev: skb->dev, fl6: &fl6, skb, flags); |
296 | } else { |
297 | struct fib6_table *table; |
298 | |
299 | table = fib6_get_table(net, id: tbl_id); |
300 | if (!table) |
301 | goto out; |
302 | |
303 | rt = ip6_pol_route(net, table, ifindex: 0, fl6: &fl6, skb, flags); |
304 | dst = &rt->dst; |
305 | } |
306 | |
307 | /* we want to discard traffic destined for local packet processing, |
308 | * if @local_delivery is set to false. |
309 | */ |
310 | if (!local_delivery) |
311 | dev_flags |= IFF_LOOPBACK; |
312 | |
313 | if (dst && (dst->dev->flags & dev_flags) && !dst->error) { |
314 | dst_release(dst); |
315 | dst = NULL; |
316 | } |
317 | |
318 | out: |
319 | if (!dst) { |
320 | rt = net->ipv6.ip6_blk_hole_entry; |
321 | dst = &rt->dst; |
322 | dst_hold(dst); |
323 | } |
324 | |
325 | skb_dst_drop(skb); |
326 | skb_dst_set(skb, dst); |
327 | return dst->error; |
328 | } |
329 | |
330 | int seg6_lookup_nexthop(struct sk_buff *skb, |
331 | struct in6_addr *nhaddr, u32 tbl_id) |
332 | { |
333 | return seg6_lookup_any_nexthop(skb, nhaddr, tbl_id, local_delivery: false); |
334 | } |
335 | |
336 | static __u8 seg6_flv_lcblock_octects(const struct seg6_flavors_info *finfo) |
337 | { |
338 | return finfo->lcblock_bits >> 3; |
339 | } |
340 | |
341 | static __u8 seg6_flv_lcnode_func_octects(const struct seg6_flavors_info *finfo) |
342 | { |
343 | return finfo->lcnode_func_bits >> 3; |
344 | } |
345 | |
346 | static bool seg6_next_csid_is_arg_zero(const struct in6_addr *addr, |
347 | const struct seg6_flavors_info *finfo) |
348 | { |
349 | __u8 fnc_octects = seg6_flv_lcnode_func_octects(finfo); |
350 | __u8 blk_octects = seg6_flv_lcblock_octects(finfo); |
351 | __u8 arg_octects; |
352 | int i; |
353 | |
354 | arg_octects = 16 - blk_octects - fnc_octects; |
355 | for (i = 0; i < arg_octects; ++i) { |
356 | if (addr->s6_addr[blk_octects + fnc_octects + i] != 0x00) |
357 | return false; |
358 | } |
359 | |
360 | return true; |
361 | } |
362 | |
363 | /* assume that DA.Argument length > 0 */ |
364 | static void seg6_next_csid_advance_arg(struct in6_addr *addr, |
365 | const struct seg6_flavors_info *finfo) |
366 | { |
367 | __u8 fnc_octects = seg6_flv_lcnode_func_octects(finfo); |
368 | __u8 blk_octects = seg6_flv_lcblock_octects(finfo); |
369 | |
370 | /* advance DA.Argument */ |
371 | memmove(&addr->s6_addr[blk_octects], |
372 | &addr->s6_addr[blk_octects + fnc_octects], |
373 | 16 - blk_octects - fnc_octects); |
374 | |
375 | memset(&addr->s6_addr[16 - fnc_octects], 0x00, fnc_octects); |
376 | } |
377 | |
378 | static int input_action_end_finish(struct sk_buff *skb, |
379 | struct seg6_local_lwt *slwt) |
380 | { |
381 | seg6_lookup_nexthop(skb, NULL, tbl_id: 0); |
382 | |
383 | return dst_input(skb); |
384 | } |
385 | |
386 | static int input_action_end_core(struct sk_buff *skb, |
387 | struct seg6_local_lwt *slwt) |
388 | { |
389 | struct ipv6_sr_hdr *srh; |
390 | |
391 | srh = get_and_validate_srh(skb); |
392 | if (!srh) |
393 | goto drop; |
394 | |
395 | advance_nextseg(srh, daddr: &ipv6_hdr(skb)->daddr); |
396 | |
397 | return input_action_end_finish(skb, slwt); |
398 | |
399 | drop: |
400 | kfree_skb(skb); |
401 | return -EINVAL; |
402 | } |
403 | |
404 | static int end_next_csid_core(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
405 | { |
406 | const struct seg6_flavors_info *finfo = &slwt->flv_info; |
407 | struct in6_addr *daddr = &ipv6_hdr(skb)->daddr; |
408 | |
409 | if (seg6_next_csid_is_arg_zero(addr: daddr, finfo)) |
410 | return input_action_end_core(skb, slwt); |
411 | |
412 | /* update DA */ |
413 | seg6_next_csid_advance_arg(addr: daddr, finfo); |
414 | |
415 | return input_action_end_finish(skb, slwt); |
416 | } |
417 | |
418 | static int input_action_end_x_finish(struct sk_buff *skb, |
419 | struct seg6_local_lwt *slwt) |
420 | { |
421 | seg6_lookup_nexthop(skb, nhaddr: &slwt->nh6, tbl_id: 0); |
422 | |
423 | return dst_input(skb); |
424 | } |
425 | |
426 | static int input_action_end_x_core(struct sk_buff *skb, |
427 | struct seg6_local_lwt *slwt) |
428 | { |
429 | struct ipv6_sr_hdr *srh; |
430 | |
431 | srh = get_and_validate_srh(skb); |
432 | if (!srh) |
433 | goto drop; |
434 | |
435 | advance_nextseg(srh, daddr: &ipv6_hdr(skb)->daddr); |
436 | |
437 | return input_action_end_x_finish(skb, slwt); |
438 | |
439 | drop: |
440 | kfree_skb(skb); |
441 | return -EINVAL; |
442 | } |
443 | |
444 | static int end_x_next_csid_core(struct sk_buff *skb, |
445 | struct seg6_local_lwt *slwt) |
446 | { |
447 | const struct seg6_flavors_info *finfo = &slwt->flv_info; |
448 | struct in6_addr *daddr = &ipv6_hdr(skb)->daddr; |
449 | |
450 | if (seg6_next_csid_is_arg_zero(addr: daddr, finfo)) |
451 | return input_action_end_x_core(skb, slwt); |
452 | |
453 | /* update DA */ |
454 | seg6_next_csid_advance_arg(addr: daddr, finfo); |
455 | |
456 | return input_action_end_x_finish(skb, slwt); |
457 | } |
458 | |
459 | static bool seg6_next_csid_enabled(__u32 fops) |
460 | { |
461 | return fops & SEG6_F_LOCAL_FLV_NEXT_CSID; |
462 | } |
463 | |
464 | /* Processing of SRv6 End, End.X, and End.T behaviors can be extended through |
465 | * the flavors framework. These behaviors must report the subset of (flavor) |
466 | * operations they currently implement. In this way, if a user specifies a |
467 | * flavor combination that is not supported by a given End* behavior, the |
468 | * kernel refuses to instantiate the tunnel reporting the error. |
469 | */ |
470 | static int seg6_flv_supp_ops_by_action(int action, __u32 *fops) |
471 | { |
472 | switch (action) { |
473 | case SEG6_LOCAL_ACTION_END: |
474 | *fops = SEG6_LOCAL_END_FLV_SUPP_OPS; |
475 | break; |
476 | case SEG6_LOCAL_ACTION_END_X: |
477 | *fops = SEG6_LOCAL_END_X_FLV_SUPP_OPS; |
478 | break; |
479 | default: |
480 | return -EOPNOTSUPP; |
481 | } |
482 | |
483 | return 0; |
484 | } |
485 | |
486 | /* We describe the packet state in relation to the absence/presence of the SRH |
487 | * and the Segment Left (SL) field. |
488 | * For our purposes, it is not necessary to record the exact value of the SL |
489 | * when the SID List consists of two or more segments. |
490 | */ |
491 | enum seg6_local_pktinfo { |
492 | /* the order really matters! */ |
493 | SEG6_LOCAL_PKTINFO_NOHDR = 0, |
494 | SEG6_LOCAL_PKTINFO_SL_ZERO, |
495 | SEG6_LOCAL_PKTINFO_SL_ONE, |
496 | SEG6_LOCAL_PKTINFO_SL_MORE, |
497 | __SEG6_LOCAL_PKTINFO_MAX, |
498 | }; |
499 | |
500 | #define SEG6_LOCAL_PKTINFO_MAX (__SEG6_LOCAL_PKTINFO_MAX - 1) |
501 | |
502 | static enum seg6_local_pktinfo seg6_get_srh_pktinfo(struct ipv6_sr_hdr *srh) |
503 | { |
504 | __u8 sgl; |
505 | |
506 | if (!srh) |
507 | return SEG6_LOCAL_PKTINFO_NOHDR; |
508 | |
509 | sgl = srh->segments_left; |
510 | if (sgl < 2) |
511 | return SEG6_LOCAL_PKTINFO_SL_ZERO + sgl; |
512 | |
513 | return SEG6_LOCAL_PKTINFO_SL_MORE; |
514 | } |
515 | |
516 | enum seg6_local_flv_action { |
517 | SEG6_LOCAL_FLV_ACT_UNSPEC = 0, |
518 | SEG6_LOCAL_FLV_ACT_END, |
519 | SEG6_LOCAL_FLV_ACT_PSP, |
520 | SEG6_LOCAL_FLV_ACT_USP, |
521 | SEG6_LOCAL_FLV_ACT_USD, |
522 | __SEG6_LOCAL_FLV_ACT_MAX |
523 | }; |
524 | |
525 | #define SEG6_LOCAL_FLV_ACT_MAX (__SEG6_LOCAL_FLV_ACT_MAX - 1) |
526 | |
527 | /* The action table for RFC8986 flavors (see the flv8986_act_tbl below) |
528 | * contains the actions (i.e. processing operations) to be applied on packets |
529 | * when flavors are configured for an End* behavior. |
530 | * By combining the pkinfo data and from the flavors mask, the macro |
531 | * computes the index used to access the elements (actions) stored in the |
532 | * action table. The index is structured as follows: |
533 | * |
534 | * index |
535 | * _______________/\________________ |
536 | * / \ |
537 | * +----------------+----------------+ |
538 | * | pf | afm | |
539 | * +----------------+----------------+ |
540 | * ph-1 ... p1 p0 fk-1 ... f1 f0 |
541 | * MSB LSB |
542 | * |
543 | * where: |
544 | * - 'afm' (adjusted flavor mask) is the mask containing a combination of the |
545 | * RFC8986 flavors currently supported. 'afm' corresponds to the @fm |
546 | * argument of the macro whose value is righ-shifted by 1 bit. By doing so, |
547 | * we discard the SEG6_LOCAL_FLV_OP_UNSPEC flag (bit 0 in @fm) which is |
548 | * never used here; |
549 | * - 'pf' encodes the packet info (pktinfo) regarding the presence/absence of |
550 | * the SRH, SL = 0, etc. 'pf' is set with the value of @pf provided as |
551 | * argument to the macro. |
552 | */ |
553 | #define flv8986_act_tbl_idx(pf, fm) \ |
554 | ((((pf) << bits_per(SEG6_LOCAL_FLV8986_SUPP_OPS)) | \ |
555 | ((fm) & SEG6_LOCAL_FLV8986_SUPP_OPS)) >> SEG6_LOCAL_FLV_OP_PSP) |
556 | |
557 | /* We compute the size of the action table by considering the RFC8986 flavors |
558 | * actually supported by the kernel. In this way, the size is automatically |
559 | * adjusted when new flavors are supported. |
560 | */ |
561 | #define FLV8986_ACT_TBL_SIZE \ |
562 | roundup_pow_of_two(flv8986_act_tbl_idx(SEG6_LOCAL_PKTINFO_MAX, \ |
563 | SEG6_LOCAL_FLV8986_SUPP_OPS)) |
564 | |
565 | /* tbl_cfg(act, pf, fm) macro is used to easily configure the action |
566 | * table; it accepts 3 arguments: |
567 | * i) @act, the suffix from SEG6_LOCAL_FLV_ACT_{act} representing |
568 | * the action that should be applied on the packet; |
569 | * ii) @pf, the suffix from SEG6_LOCAL_PKTINFO_{pf} reporting the packet |
570 | * info about the lack/presence of SRH, SRH with SL = 0, etc; |
571 | * iii) @fm, the mask of flavors. |
572 | */ |
573 | #define tbl_cfg(act, pf, fm) \ |
574 | [flv8986_act_tbl_idx(SEG6_LOCAL_PKTINFO_##pf, \ |
575 | (fm))] = SEG6_LOCAL_FLV_ACT_##act |
576 | |
577 | /* shorthand for improving readability */ |
578 | #define F_PSP SEG6_F_LOCAL_FLV_PSP |
579 | |
580 | /* The table contains, for each combination of the pktinfo data and |
581 | * flavors, the action that should be taken on a packet (e.g. |
582 | * "standard" Endpoint processing, Penultimate Segment Pop, etc). |
583 | * |
584 | * By default, table entries not explicitly configured are initialized with the |
585 | * SEG6_LOCAL_FLV_ACT_UNSPEC action, which generally has the effect of |
586 | * discarding the processed packet. |
587 | */ |
588 | static const u8 flv8986_act_tbl[FLV8986_ACT_TBL_SIZE] = { |
589 | /* PSP variant for packet where SRH with SL = 1 */ |
590 | tbl_cfg(PSP, SL_ONE, F_PSP), |
591 | /* End for packet where the SRH with SL > 1*/ |
592 | tbl_cfg(END, SL_MORE, F_PSP), |
593 | }; |
594 | |
595 | #undef F_PSP |
596 | #undef tbl_cfg |
597 | |
598 | /* For each flavor defined in RFC8986 (or a combination of them) an action is |
599 | * performed on the packet. The specific action depends on: |
600 | * - info extracted from the packet (i.e. pktinfo data) regarding the |
601 | * lack/presence of the SRH, and if the SRH is available, on the value of |
602 | * Segment Left field; |
603 | * - the mask of flavors configured for the specific SRv6 End* behavior. |
604 | * |
605 | * The function combines both the pkinfo and the flavors mask to evaluate the |
606 | * corresponding action to be taken on the packet. |
607 | */ |
608 | static enum seg6_local_flv_action |
609 | seg6_local_flv8986_act_lookup(enum seg6_local_pktinfo pinfo, __u32 flvmask) |
610 | { |
611 | unsigned long index; |
612 | |
613 | /* check if the provided mask of flavors is supported */ |
614 | if (unlikely(flvmask & ~SEG6_LOCAL_FLV8986_SUPP_OPS)) |
615 | return SEG6_LOCAL_FLV_ACT_UNSPEC; |
616 | |
617 | index = flv8986_act_tbl_idx(pinfo, flvmask); |
618 | if (unlikely(index >= FLV8986_ACT_TBL_SIZE)) |
619 | return SEG6_LOCAL_FLV_ACT_UNSPEC; |
620 | |
621 | return flv8986_act_tbl[index]; |
622 | } |
623 | |
624 | /* skb->data must be aligned with skb->network_header */ |
625 | static bool seg6_pop_srh(struct sk_buff *skb, int srhoff) |
626 | { |
627 | struct ipv6_sr_hdr *srh; |
628 | struct ipv6hdr *iph; |
629 | __u8 srh_nexthdr; |
630 | int thoff = -1; |
631 | int srhlen; |
632 | int nhlen; |
633 | |
634 | if (unlikely(srhoff < sizeof(*iph) || |
635 | !pskb_may_pull(skb, srhoff + sizeof(*srh)))) |
636 | return false; |
637 | |
638 | srh = (struct ipv6_sr_hdr *)(skb->data + srhoff); |
639 | srhlen = ipv6_optlen(srh); |
640 | |
641 | /* we are about to mangle the pkt, let's check if we can write on it */ |
642 | if (unlikely(skb_ensure_writable(skb, srhoff + srhlen))) |
643 | return false; |
644 | |
645 | /* skb_ensure_writable() may change skb pointers; evaluate srh again */ |
646 | srh = (struct ipv6_sr_hdr *)(skb->data + srhoff); |
647 | srh_nexthdr = srh->nexthdr; |
648 | |
649 | if (unlikely(!skb_transport_header_was_set(skb))) |
650 | goto pull; |
651 | |
652 | nhlen = skb_network_header_len(skb); |
653 | /* we have to deal with the transport header: it could be set before |
654 | * the SRH, after the SRH, or within it (which is considered wrong, |
655 | * however). |
656 | */ |
657 | if (likely(nhlen <= srhoff)) |
658 | thoff = nhlen; |
659 | else if (nhlen >= srhoff + srhlen) |
660 | /* transport_header is set after the SRH */ |
661 | thoff = nhlen - srhlen; |
662 | else |
663 | /* transport_header falls inside the SRH; hence, we can't |
664 | * restore the transport_header pointer properly after |
665 | * SRH removing operation. |
666 | */ |
667 | return false; |
668 | pull: |
669 | /* we need to pop the SRH: |
670 | * 1) first of all, we pull out everything from IPv6 header up to SRH |
671 | * (included) evaluating also the rcsum; |
672 | * 2) we overwrite (and then remove) the SRH by properly moving the |
673 | * IPv6 along with any extension header that precedes the SRH; |
674 | * 3) At the end, we push back the pulled headers (except for SRH, |
675 | * obviously). |
676 | */ |
677 | skb_pull_rcsum(skb, len: srhoff + srhlen); |
678 | memmove(skb_network_header(skb) + srhlen, skb_network_header(skb), |
679 | srhoff); |
680 | skb_push(skb, len: srhoff); |
681 | |
682 | skb_reset_network_header(skb); |
683 | skb_mac_header_rebuild(skb); |
684 | if (likely(thoff >= 0)) |
685 | skb_set_transport_header(skb, offset: thoff); |
686 | |
687 | iph = ipv6_hdr(skb); |
688 | if (iph->nexthdr == NEXTHDR_ROUTING) { |
689 | iph->nexthdr = srh_nexthdr; |
690 | } else { |
691 | /* we must look for the extension header (EXTH, for short) that |
692 | * immediately precedes the SRH we have just removed. |
693 | * Then, we update the value of the EXTH nexthdr with the one |
694 | * contained in the SRH nexthdr. |
695 | */ |
696 | unsigned int off = sizeof(*iph); |
697 | struct ipv6_opt_hdr *hp, _hdr; |
698 | __u8 nexthdr = iph->nexthdr; |
699 | |
700 | for (;;) { |
701 | if (unlikely(!ipv6_ext_hdr(nexthdr) || |
702 | nexthdr == NEXTHDR_NONE)) |
703 | return false; |
704 | |
705 | hp = skb_header_pointer(skb, offset: off, len: sizeof(_hdr), buffer: &_hdr); |
706 | if (unlikely(!hp)) |
707 | return false; |
708 | |
709 | if (hp->nexthdr == NEXTHDR_ROUTING) { |
710 | hp->nexthdr = srh_nexthdr; |
711 | break; |
712 | } |
713 | |
714 | switch (nexthdr) { |
715 | case NEXTHDR_FRAGMENT: |
716 | fallthrough; |
717 | case NEXTHDR_AUTH: |
718 | /* we expect SRH before FRAG and AUTH */ |
719 | return false; |
720 | default: |
721 | off += ipv6_optlen(hp); |
722 | break; |
723 | } |
724 | |
725 | nexthdr = hp->nexthdr; |
726 | } |
727 | } |
728 | |
729 | iph->payload_len = htons(skb->len - sizeof(struct ipv6hdr)); |
730 | |
731 | skb_postpush_rcsum(skb, start: iph, len: srhoff); |
732 | |
733 | return true; |
734 | } |
735 | |
736 | /* process the packet on the basis of the RFC8986 flavors set for the given |
737 | * SRv6 End behavior instance. |
738 | */ |
739 | static int end_flv8986_core(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
740 | { |
741 | const struct seg6_flavors_info *finfo = &slwt->flv_info; |
742 | enum seg6_local_flv_action action; |
743 | enum seg6_local_pktinfo pinfo; |
744 | struct ipv6_sr_hdr *srh; |
745 | __u32 flvmask; |
746 | int srhoff; |
747 | |
748 | srh = seg6_get_srh(skb, flags: 0); |
749 | srhoff = srh ? ((unsigned char *)srh - skb->data) : 0; |
750 | pinfo = seg6_get_srh_pktinfo(srh); |
751 | #ifdef CONFIG_IPV6_SEG6_HMAC |
752 | if (srh && !seg6_hmac_validate_skb(skb)) |
753 | goto drop; |
754 | #endif |
755 | flvmask = finfo->flv_ops; |
756 | if (unlikely(flvmask & ~SEG6_LOCAL_FLV8986_SUPP_OPS)) { |
757 | pr_warn_once("seg6local: invalid RFC8986 flavors\n" ); |
758 | goto drop; |
759 | } |
760 | |
761 | /* retrieve the action triggered by the combination of pktinfo data and |
762 | * the flavors mask. |
763 | */ |
764 | action = seg6_local_flv8986_act_lookup(pinfo, flvmask); |
765 | switch (action) { |
766 | case SEG6_LOCAL_FLV_ACT_END: |
767 | /* process the packet as the "standard" End behavior */ |
768 | advance_nextseg(srh, daddr: &ipv6_hdr(skb)->daddr); |
769 | break; |
770 | case SEG6_LOCAL_FLV_ACT_PSP: |
771 | advance_nextseg(srh, daddr: &ipv6_hdr(skb)->daddr); |
772 | |
773 | if (unlikely(!seg6_pop_srh(skb, srhoff))) |
774 | goto drop; |
775 | break; |
776 | case SEG6_LOCAL_FLV_ACT_UNSPEC: |
777 | fallthrough; |
778 | default: |
779 | /* by default, we drop the packet since we could not find a |
780 | * suitable action. |
781 | */ |
782 | goto drop; |
783 | } |
784 | |
785 | return input_action_end_finish(skb, slwt); |
786 | |
787 | drop: |
788 | kfree_skb(skb); |
789 | return -EINVAL; |
790 | } |
791 | |
792 | /* regular endpoint function */ |
793 | static int input_action_end(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
794 | { |
795 | const struct seg6_flavors_info *finfo = &slwt->flv_info; |
796 | __u32 fops = finfo->flv_ops; |
797 | |
798 | if (!fops) |
799 | return input_action_end_core(skb, slwt); |
800 | |
801 | /* check for the presence of NEXT-C-SID since it applies first */ |
802 | if (seg6_next_csid_enabled(fops)) |
803 | return end_next_csid_core(skb, slwt); |
804 | |
805 | /* the specific processing function to be performed on the packet |
806 | * depends on the combination of flavors defined in RFC8986 and some |
807 | * information extracted from the packet, e.g. presence/absence of SRH, |
808 | * Segment Left = 0, etc. |
809 | */ |
810 | return end_flv8986_core(skb, slwt); |
811 | } |
812 | |
813 | /* regular endpoint, and forward to specified nexthop */ |
814 | static int input_action_end_x(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
815 | { |
816 | const struct seg6_flavors_info *finfo = &slwt->flv_info; |
817 | __u32 fops = finfo->flv_ops; |
818 | |
819 | /* check for the presence of NEXT-C-SID since it applies first */ |
820 | if (seg6_next_csid_enabled(fops)) |
821 | return end_x_next_csid_core(skb, slwt); |
822 | |
823 | return input_action_end_x_core(skb, slwt); |
824 | } |
825 | |
826 | static int input_action_end_t(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
827 | { |
828 | struct ipv6_sr_hdr *srh; |
829 | |
830 | srh = get_and_validate_srh(skb); |
831 | if (!srh) |
832 | goto drop; |
833 | |
834 | advance_nextseg(srh, daddr: &ipv6_hdr(skb)->daddr); |
835 | |
836 | seg6_lookup_nexthop(skb, NULL, tbl_id: slwt->table); |
837 | |
838 | return dst_input(skb); |
839 | |
840 | drop: |
841 | kfree_skb(skb); |
842 | return -EINVAL; |
843 | } |
844 | |
845 | /* decapsulate and forward inner L2 frame on specified interface */ |
846 | static int input_action_end_dx2(struct sk_buff *skb, |
847 | struct seg6_local_lwt *slwt) |
848 | { |
849 | struct net *net = dev_net(dev: skb->dev); |
850 | struct net_device *odev; |
851 | struct ethhdr *eth; |
852 | |
853 | if (!decap_and_validate(skb, IPPROTO_ETHERNET)) |
854 | goto drop; |
855 | |
856 | if (!pskb_may_pull(skb, ETH_HLEN)) |
857 | goto drop; |
858 | |
859 | skb_reset_mac_header(skb); |
860 | eth = (struct ethhdr *)skb->data; |
861 | |
862 | /* To determine the frame's protocol, we assume it is 802.3. This avoids |
863 | * a call to eth_type_trans(), which is not really relevant for our |
864 | * use case. |
865 | */ |
866 | if (!eth_proto_is_802_3(proto: eth->h_proto)) |
867 | goto drop; |
868 | |
869 | odev = dev_get_by_index_rcu(net, ifindex: slwt->oif); |
870 | if (!odev) |
871 | goto drop; |
872 | |
873 | /* As we accept Ethernet frames, make sure the egress device is of |
874 | * the correct type. |
875 | */ |
876 | if (odev->type != ARPHRD_ETHER) |
877 | goto drop; |
878 | |
879 | if (!(odev->flags & IFF_UP) || !netif_carrier_ok(dev: odev)) |
880 | goto drop; |
881 | |
882 | skb_orphan(skb); |
883 | |
884 | if (skb_warn_if_lro(skb)) |
885 | goto drop; |
886 | |
887 | skb_forward_csum(skb); |
888 | |
889 | if (skb->len - ETH_HLEN > odev->mtu) |
890 | goto drop; |
891 | |
892 | skb->dev = odev; |
893 | skb->protocol = eth->h_proto; |
894 | |
895 | return dev_queue_xmit(skb); |
896 | |
897 | drop: |
898 | kfree_skb(skb); |
899 | return -EINVAL; |
900 | } |
901 | |
902 | static int input_action_end_dx6_finish(struct net *net, struct sock *sk, |
903 | struct sk_buff *skb) |
904 | { |
905 | struct dst_entry *orig_dst = skb_dst(skb); |
906 | struct in6_addr *nhaddr = NULL; |
907 | struct seg6_local_lwt *slwt; |
908 | |
909 | slwt = seg6_local_lwtunnel(lwt: orig_dst->lwtstate); |
910 | |
911 | /* The inner packet is not associated to any local interface, |
912 | * so we do not call netif_rx(). |
913 | * |
914 | * If slwt->nh6 is set to ::, then lookup the nexthop for the |
915 | * inner packet's DA. Otherwise, use the specified nexthop. |
916 | */ |
917 | if (!ipv6_addr_any(a: &slwt->nh6)) |
918 | nhaddr = &slwt->nh6; |
919 | |
920 | seg6_lookup_nexthop(skb, nhaddr, tbl_id: 0); |
921 | |
922 | return dst_input(skb); |
923 | } |
924 | |
925 | /* decapsulate and forward to specified nexthop */ |
926 | static int input_action_end_dx6(struct sk_buff *skb, |
927 | struct seg6_local_lwt *slwt) |
928 | { |
929 | /* this function accepts IPv6 encapsulated packets, with either |
930 | * an SRH with SL=0, or no SRH. |
931 | */ |
932 | |
933 | if (!decap_and_validate(skb, IPPROTO_IPV6)) |
934 | goto drop; |
935 | |
936 | if (!pskb_may_pull(skb, len: sizeof(struct ipv6hdr))) |
937 | goto drop; |
938 | |
939 | skb_set_transport_header(skb, offset: sizeof(struct ipv6hdr)); |
940 | nf_reset_ct(skb); |
941 | |
942 | if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) |
943 | return NF_HOOK(pf: NFPROTO_IPV6, hook: NF_INET_PRE_ROUTING, |
944 | net: dev_net(dev: skb->dev), NULL, skb, NULL, |
945 | out: skb_dst(skb)->dev, okfn: input_action_end_dx6_finish); |
946 | |
947 | return input_action_end_dx6_finish(net: dev_net(dev: skb->dev), NULL, skb); |
948 | drop: |
949 | kfree_skb(skb); |
950 | return -EINVAL; |
951 | } |
952 | |
953 | static int input_action_end_dx4_finish(struct net *net, struct sock *sk, |
954 | struct sk_buff *skb) |
955 | { |
956 | struct dst_entry *orig_dst = skb_dst(skb); |
957 | struct seg6_local_lwt *slwt; |
958 | struct iphdr *iph; |
959 | __be32 nhaddr; |
960 | int err; |
961 | |
962 | slwt = seg6_local_lwtunnel(lwt: orig_dst->lwtstate); |
963 | |
964 | iph = ip_hdr(skb); |
965 | |
966 | nhaddr = slwt->nh4.s_addr ?: iph->daddr; |
967 | |
968 | skb_dst_drop(skb); |
969 | |
970 | err = ip_route_input(skb, dst: nhaddr, src: iph->saddr, tos: 0, devin: skb->dev); |
971 | if (err) { |
972 | kfree_skb(skb); |
973 | return -EINVAL; |
974 | } |
975 | |
976 | return dst_input(skb); |
977 | } |
978 | |
979 | static int input_action_end_dx4(struct sk_buff *skb, |
980 | struct seg6_local_lwt *slwt) |
981 | { |
982 | if (!decap_and_validate(skb, IPPROTO_IPIP)) |
983 | goto drop; |
984 | |
985 | if (!pskb_may_pull(skb, len: sizeof(struct iphdr))) |
986 | goto drop; |
987 | |
988 | skb->protocol = htons(ETH_P_IP); |
989 | skb_set_transport_header(skb, offset: sizeof(struct iphdr)); |
990 | nf_reset_ct(skb); |
991 | |
992 | if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) |
993 | return NF_HOOK(pf: NFPROTO_IPV4, hook: NF_INET_PRE_ROUTING, |
994 | net: dev_net(dev: skb->dev), NULL, skb, NULL, |
995 | out: skb_dst(skb)->dev, okfn: input_action_end_dx4_finish); |
996 | |
997 | return input_action_end_dx4_finish(net: dev_net(dev: skb->dev), NULL, skb); |
998 | drop: |
999 | kfree_skb(skb); |
1000 | return -EINVAL; |
1001 | } |
1002 | |
1003 | #ifdef CONFIG_NET_L3_MASTER_DEV |
1004 | static struct net *fib6_config_get_net(const struct fib6_config *fib6_cfg) |
1005 | { |
1006 | const struct nl_info *nli = &fib6_cfg->fc_nlinfo; |
1007 | |
1008 | return nli->nl_net; |
1009 | } |
1010 | |
1011 | static int __seg6_end_dt_vrf_build(struct seg6_local_lwt *slwt, const void *cfg, |
1012 | u16 family, struct netlink_ext_ack *extack) |
1013 | { |
1014 | struct seg6_end_dt_info *info = &slwt->dt_info; |
1015 | int vrf_ifindex; |
1016 | struct net *net; |
1017 | |
1018 | net = fib6_config_get_net(fib6_cfg: cfg); |
1019 | |
1020 | /* note that vrf_table was already set by parse_nla_vrftable() */ |
1021 | vrf_ifindex = l3mdev_ifindex_lookup_by_table_id(l3type: L3MDEV_TYPE_VRF, net, |
1022 | table_id: info->vrf_table); |
1023 | if (vrf_ifindex < 0) { |
1024 | if (vrf_ifindex == -EPERM) { |
1025 | NL_SET_ERR_MSG(extack, |
1026 | "Strict mode for VRF is disabled" ); |
1027 | } else if (vrf_ifindex == -ENODEV) { |
1028 | NL_SET_ERR_MSG(extack, |
1029 | "Table has no associated VRF device" ); |
1030 | } else { |
1031 | pr_debug("seg6local: SRv6 End.DT* creation error=%d\n" , |
1032 | vrf_ifindex); |
1033 | } |
1034 | |
1035 | return vrf_ifindex; |
1036 | } |
1037 | |
1038 | info->net = net; |
1039 | info->vrf_ifindex = vrf_ifindex; |
1040 | |
1041 | info->family = family; |
1042 | info->mode = DT_VRF_MODE; |
1043 | |
1044 | return 0; |
1045 | } |
1046 | |
1047 | /* The SRv6 End.DT4/DT6 behavior extracts the inner (IPv4/IPv6) packet and |
1048 | * routes the IPv4/IPv6 packet by looking at the configured routing table. |
1049 | * |
1050 | * In the SRv6 End.DT4/DT6 use case, we can receive traffic (IPv6+Segment |
1051 | * Routing Header packets) from several interfaces and the outer IPv6 |
1052 | * destination address (DA) is used for retrieving the specific instance of the |
1053 | * End.DT4/DT6 behavior that should process the packets. |
1054 | * |
1055 | * However, the inner IPv4/IPv6 packet is not really bound to any receiving |
1056 | * interface and thus the End.DT4/DT6 sets the VRF (associated with the |
1057 | * corresponding routing table) as the *receiving* interface. |
1058 | * In other words, the End.DT4/DT6 processes a packet as if it has been received |
1059 | * directly by the VRF (and not by one of its slave devices, if any). |
1060 | * In this way, the VRF interface is used for routing the IPv4/IPv6 packet in |
1061 | * according to the routing table configured by the End.DT4/DT6 instance. |
1062 | * |
1063 | * This design allows you to get some interesting features like: |
1064 | * 1) the statistics on rx packets; |
1065 | * 2) the possibility to install a packet sniffer on the receiving interface |
1066 | * (the VRF one) for looking at the incoming packets; |
1067 | * 3) the possibility to leverage the netfilter prerouting hook for the inner |
1068 | * IPv4 packet. |
1069 | * |
1070 | * This function returns: |
1071 | * - the sk_buff* when the VRF rcv handler has processed the packet correctly; |
1072 | * - NULL when the skb is consumed by the VRF rcv handler; |
1073 | * - a pointer which encodes a negative error number in case of error. |
1074 | * Note that in this case, the function takes care of freeing the skb. |
1075 | */ |
1076 | static struct sk_buff *end_dt_vrf_rcv(struct sk_buff *skb, u16 family, |
1077 | struct net_device *dev) |
1078 | { |
1079 | /* based on l3mdev_ip_rcv; we are only interested in the master */ |
1080 | if (unlikely(!netif_is_l3_master(dev) && !netif_has_l3_rx_handler(dev))) |
1081 | goto drop; |
1082 | |
1083 | if (unlikely(!dev->l3mdev_ops->l3mdev_l3_rcv)) |
1084 | goto drop; |
1085 | |
1086 | /* the decap packet IPv4/IPv6 does not come with any mac header info. |
1087 | * We must unset the mac header to allow the VRF device to rebuild it, |
1088 | * just in case there is a sniffer attached on the device. |
1089 | */ |
1090 | skb_unset_mac_header(skb); |
1091 | |
1092 | skb = dev->l3mdev_ops->l3mdev_l3_rcv(dev, skb, family); |
1093 | if (!skb) |
1094 | /* the skb buffer was consumed by the handler */ |
1095 | return NULL; |
1096 | |
1097 | /* when a packet is received by a VRF or by one of its slaves, the |
1098 | * master device reference is set into the skb. |
1099 | */ |
1100 | if (unlikely(skb->dev != dev || skb->skb_iif != dev->ifindex)) |
1101 | goto drop; |
1102 | |
1103 | return skb; |
1104 | |
1105 | drop: |
1106 | kfree_skb(skb); |
1107 | return ERR_PTR(error: -EINVAL); |
1108 | } |
1109 | |
1110 | static struct net_device *end_dt_get_vrf_rcu(struct sk_buff *skb, |
1111 | struct seg6_end_dt_info *info) |
1112 | { |
1113 | int vrf_ifindex = info->vrf_ifindex; |
1114 | struct net *net = info->net; |
1115 | |
1116 | if (unlikely(vrf_ifindex < 0)) |
1117 | goto error; |
1118 | |
1119 | if (unlikely(!net_eq(dev_net(skb->dev), net))) |
1120 | goto error; |
1121 | |
1122 | return dev_get_by_index_rcu(net, ifindex: vrf_ifindex); |
1123 | |
1124 | error: |
1125 | return NULL; |
1126 | } |
1127 | |
1128 | static struct sk_buff *end_dt_vrf_core(struct sk_buff *skb, |
1129 | struct seg6_local_lwt *slwt, u16 family) |
1130 | { |
1131 | struct seg6_end_dt_info *info = &slwt->dt_info; |
1132 | struct net_device *vrf; |
1133 | __be16 protocol; |
1134 | int hdrlen; |
1135 | |
1136 | vrf = end_dt_get_vrf_rcu(skb, info); |
1137 | if (unlikely(!vrf)) |
1138 | goto drop; |
1139 | |
1140 | switch (family) { |
1141 | case AF_INET: |
1142 | protocol = htons(ETH_P_IP); |
1143 | hdrlen = sizeof(struct iphdr); |
1144 | break; |
1145 | case AF_INET6: |
1146 | protocol = htons(ETH_P_IPV6); |
1147 | hdrlen = sizeof(struct ipv6hdr); |
1148 | break; |
1149 | case AF_UNSPEC: |
1150 | fallthrough; |
1151 | default: |
1152 | goto drop; |
1153 | } |
1154 | |
1155 | if (unlikely(info->family != AF_UNSPEC && info->family != family)) { |
1156 | pr_warn_once("seg6local: SRv6 End.DT* family mismatch" ); |
1157 | goto drop; |
1158 | } |
1159 | |
1160 | skb->protocol = protocol; |
1161 | |
1162 | skb_dst_drop(skb); |
1163 | |
1164 | skb_set_transport_header(skb, offset: hdrlen); |
1165 | nf_reset_ct(skb); |
1166 | |
1167 | return end_dt_vrf_rcv(skb, family, dev: vrf); |
1168 | |
1169 | drop: |
1170 | kfree_skb(skb); |
1171 | return ERR_PTR(error: -EINVAL); |
1172 | } |
1173 | |
1174 | static int input_action_end_dt4(struct sk_buff *skb, |
1175 | struct seg6_local_lwt *slwt) |
1176 | { |
1177 | struct iphdr *iph; |
1178 | int err; |
1179 | |
1180 | if (!decap_and_validate(skb, IPPROTO_IPIP)) |
1181 | goto drop; |
1182 | |
1183 | if (!pskb_may_pull(skb, len: sizeof(struct iphdr))) |
1184 | goto drop; |
1185 | |
1186 | skb = end_dt_vrf_core(skb, slwt, AF_INET); |
1187 | if (!skb) |
1188 | /* packet has been processed and consumed by the VRF */ |
1189 | return 0; |
1190 | |
1191 | if (IS_ERR(ptr: skb)) |
1192 | return PTR_ERR(ptr: skb); |
1193 | |
1194 | iph = ip_hdr(skb); |
1195 | |
1196 | err = ip_route_input(skb, dst: iph->daddr, src: iph->saddr, tos: 0, devin: skb->dev); |
1197 | if (unlikely(err)) |
1198 | goto drop; |
1199 | |
1200 | return dst_input(skb); |
1201 | |
1202 | drop: |
1203 | kfree_skb(skb); |
1204 | return -EINVAL; |
1205 | } |
1206 | |
1207 | static int seg6_end_dt4_build(struct seg6_local_lwt *slwt, const void *cfg, |
1208 | struct netlink_ext_ack *extack) |
1209 | { |
1210 | return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET, extack); |
1211 | } |
1212 | |
1213 | static enum |
1214 | seg6_end_dt_mode seg6_end_dt6_parse_mode(struct seg6_local_lwt *slwt) |
1215 | { |
1216 | unsigned long parsed_optattrs = slwt->parsed_optattrs; |
1217 | bool legacy, vrfmode; |
1218 | |
1219 | legacy = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE)); |
1220 | vrfmode = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE)); |
1221 | |
1222 | if (!(legacy ^ vrfmode)) |
1223 | /* both are absent or present: invalid DT6 mode */ |
1224 | return DT_INVALID_MODE; |
1225 | |
1226 | return legacy ? DT_LEGACY_MODE : DT_VRF_MODE; |
1227 | } |
1228 | |
1229 | static enum seg6_end_dt_mode seg6_end_dt6_get_mode(struct seg6_local_lwt *slwt) |
1230 | { |
1231 | struct seg6_end_dt_info *info = &slwt->dt_info; |
1232 | |
1233 | return info->mode; |
1234 | } |
1235 | |
1236 | static int seg6_end_dt6_build(struct seg6_local_lwt *slwt, const void *cfg, |
1237 | struct netlink_ext_ack *extack) |
1238 | { |
1239 | enum seg6_end_dt_mode mode = seg6_end_dt6_parse_mode(slwt); |
1240 | struct seg6_end_dt_info *info = &slwt->dt_info; |
1241 | |
1242 | switch (mode) { |
1243 | case DT_LEGACY_MODE: |
1244 | info->mode = DT_LEGACY_MODE; |
1245 | return 0; |
1246 | case DT_VRF_MODE: |
1247 | return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET6, extack); |
1248 | default: |
1249 | NL_SET_ERR_MSG(extack, "table or vrftable must be specified" ); |
1250 | return -EINVAL; |
1251 | } |
1252 | } |
1253 | #endif |
1254 | |
1255 | static int input_action_end_dt6(struct sk_buff *skb, |
1256 | struct seg6_local_lwt *slwt) |
1257 | { |
1258 | if (!decap_and_validate(skb, IPPROTO_IPV6)) |
1259 | goto drop; |
1260 | |
1261 | if (!pskb_may_pull(skb, len: sizeof(struct ipv6hdr))) |
1262 | goto drop; |
1263 | |
1264 | #ifdef CONFIG_NET_L3_MASTER_DEV |
1265 | if (seg6_end_dt6_get_mode(slwt) == DT_LEGACY_MODE) |
1266 | goto legacy_mode; |
1267 | |
1268 | /* DT6_VRF_MODE */ |
1269 | skb = end_dt_vrf_core(skb, slwt, AF_INET6); |
1270 | if (!skb) |
1271 | /* packet has been processed and consumed by the VRF */ |
1272 | return 0; |
1273 | |
1274 | if (IS_ERR(ptr: skb)) |
1275 | return PTR_ERR(ptr: skb); |
1276 | |
1277 | /* note: this time we do not need to specify the table because the VRF |
1278 | * takes care of selecting the correct table. |
1279 | */ |
1280 | seg6_lookup_any_nexthop(skb, NULL, tbl_id: 0, local_delivery: true); |
1281 | |
1282 | return dst_input(skb); |
1283 | |
1284 | legacy_mode: |
1285 | #endif |
1286 | skb_set_transport_header(skb, offset: sizeof(struct ipv6hdr)); |
1287 | |
1288 | seg6_lookup_any_nexthop(skb, NULL, tbl_id: slwt->table, local_delivery: true); |
1289 | |
1290 | return dst_input(skb); |
1291 | |
1292 | drop: |
1293 | kfree_skb(skb); |
1294 | return -EINVAL; |
1295 | } |
1296 | |
1297 | #ifdef CONFIG_NET_L3_MASTER_DEV |
1298 | static int seg6_end_dt46_build(struct seg6_local_lwt *slwt, const void *cfg, |
1299 | struct netlink_ext_ack *extack) |
1300 | { |
1301 | return __seg6_end_dt_vrf_build(slwt, cfg, AF_UNSPEC, extack); |
1302 | } |
1303 | |
1304 | static int input_action_end_dt46(struct sk_buff *skb, |
1305 | struct seg6_local_lwt *slwt) |
1306 | { |
1307 | unsigned int off = 0; |
1308 | int nexthdr; |
1309 | |
1310 | nexthdr = ipv6_find_hdr(skb, offset: &off, target: -1, NULL, NULL); |
1311 | if (unlikely(nexthdr < 0)) |
1312 | goto drop; |
1313 | |
1314 | switch (nexthdr) { |
1315 | case IPPROTO_IPIP: |
1316 | return input_action_end_dt4(skb, slwt); |
1317 | case IPPROTO_IPV6: |
1318 | return input_action_end_dt6(skb, slwt); |
1319 | } |
1320 | |
1321 | drop: |
1322 | kfree_skb(skb); |
1323 | return -EINVAL; |
1324 | } |
1325 | #endif |
1326 | |
1327 | /* push an SRH on top of the current one */ |
1328 | static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
1329 | { |
1330 | struct ipv6_sr_hdr *srh; |
1331 | int err = -EINVAL; |
1332 | |
1333 | srh = get_and_validate_srh(skb); |
1334 | if (!srh) |
1335 | goto drop; |
1336 | |
1337 | err = seg6_do_srh_inline(skb, osrh: slwt->srh); |
1338 | if (err) |
1339 | goto drop; |
1340 | |
1341 | skb_set_transport_header(skb, offset: sizeof(struct ipv6hdr)); |
1342 | |
1343 | seg6_lookup_nexthop(skb, NULL, tbl_id: 0); |
1344 | |
1345 | return dst_input(skb); |
1346 | |
1347 | drop: |
1348 | kfree_skb(skb); |
1349 | return err; |
1350 | } |
1351 | |
1352 | /* encapsulate within an outer IPv6 header and a specified SRH */ |
1353 | static int input_action_end_b6_encap(struct sk_buff *skb, |
1354 | struct seg6_local_lwt *slwt) |
1355 | { |
1356 | struct ipv6_sr_hdr *srh; |
1357 | int err = -EINVAL; |
1358 | |
1359 | srh = get_and_validate_srh(skb); |
1360 | if (!srh) |
1361 | goto drop; |
1362 | |
1363 | advance_nextseg(srh, daddr: &ipv6_hdr(skb)->daddr); |
1364 | |
1365 | skb_reset_inner_headers(skb); |
1366 | skb->encapsulation = 1; |
1367 | |
1368 | err = seg6_do_srh_encap(skb, osrh: slwt->srh, IPPROTO_IPV6); |
1369 | if (err) |
1370 | goto drop; |
1371 | |
1372 | skb_set_transport_header(skb, offset: sizeof(struct ipv6hdr)); |
1373 | |
1374 | seg6_lookup_nexthop(skb, NULL, tbl_id: 0); |
1375 | |
1376 | return dst_input(skb); |
1377 | |
1378 | drop: |
1379 | kfree_skb(skb); |
1380 | return err; |
1381 | } |
1382 | |
1383 | DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states); |
1384 | |
1385 | bool seg6_bpf_has_valid_srh(struct sk_buff *skb) |
1386 | { |
1387 | struct seg6_bpf_srh_state *srh_state = |
1388 | this_cpu_ptr(&seg6_bpf_srh_states); |
1389 | struct ipv6_sr_hdr *srh = srh_state->srh; |
1390 | |
1391 | if (unlikely(srh == NULL)) |
1392 | return false; |
1393 | |
1394 | if (unlikely(!srh_state->valid)) { |
1395 | if ((srh_state->hdrlen & 7) != 0) |
1396 | return false; |
1397 | |
1398 | srh->hdrlen = (u8)(srh_state->hdrlen >> 3); |
1399 | if (!seg6_validate_srh(srh, len: (srh->hdrlen + 1) << 3, reduced: true)) |
1400 | return false; |
1401 | |
1402 | srh_state->valid = true; |
1403 | } |
1404 | |
1405 | return true; |
1406 | } |
1407 | |
1408 | static int input_action_end_bpf(struct sk_buff *skb, |
1409 | struct seg6_local_lwt *slwt) |
1410 | { |
1411 | struct seg6_bpf_srh_state *srh_state = |
1412 | this_cpu_ptr(&seg6_bpf_srh_states); |
1413 | struct ipv6_sr_hdr *srh; |
1414 | int ret; |
1415 | |
1416 | srh = get_and_validate_srh(skb); |
1417 | if (!srh) { |
1418 | kfree_skb(skb); |
1419 | return -EINVAL; |
1420 | } |
1421 | advance_nextseg(srh, daddr: &ipv6_hdr(skb)->daddr); |
1422 | |
1423 | /* preempt_disable is needed to protect the per-CPU buffer srh_state, |
1424 | * which is also accessed by the bpf_lwt_seg6_* helpers |
1425 | */ |
1426 | preempt_disable(); |
1427 | srh_state->srh = srh; |
1428 | srh_state->hdrlen = srh->hdrlen << 3; |
1429 | srh_state->valid = true; |
1430 | |
1431 | rcu_read_lock(); |
1432 | bpf_compute_data_pointers(skb); |
1433 | ret = bpf_prog_run_save_cb(prog: slwt->bpf.prog, skb); |
1434 | rcu_read_unlock(); |
1435 | |
1436 | switch (ret) { |
1437 | case BPF_OK: |
1438 | case BPF_REDIRECT: |
1439 | break; |
1440 | case BPF_DROP: |
1441 | goto drop; |
1442 | default: |
1443 | pr_warn_once("bpf-seg6local: Illegal return value %u\n" , ret); |
1444 | goto drop; |
1445 | } |
1446 | |
1447 | if (srh_state->srh && !seg6_bpf_has_valid_srh(skb)) |
1448 | goto drop; |
1449 | |
1450 | preempt_enable(); |
1451 | if (ret != BPF_REDIRECT) |
1452 | seg6_lookup_nexthop(skb, NULL, tbl_id: 0); |
1453 | |
1454 | return dst_input(skb); |
1455 | |
1456 | drop: |
1457 | preempt_enable(); |
1458 | kfree_skb(skb); |
1459 | return -EINVAL; |
1460 | } |
1461 | |
1462 | static struct seg6_action_desc seg6_action_table[] = { |
1463 | { |
1464 | .action = SEG6_LOCAL_ACTION_END, |
1465 | .attrs = 0, |
1466 | .optattrs = SEG6_F_LOCAL_COUNTERS | |
1467 | SEG6_F_LOCAL_FLAVORS, |
1468 | .input = input_action_end, |
1469 | }, |
1470 | { |
1471 | .action = SEG6_LOCAL_ACTION_END_X, |
1472 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6), |
1473 | .optattrs = SEG6_F_LOCAL_COUNTERS | |
1474 | SEG6_F_LOCAL_FLAVORS, |
1475 | .input = input_action_end_x, |
1476 | }, |
1477 | { |
1478 | .action = SEG6_LOCAL_ACTION_END_T, |
1479 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE), |
1480 | .optattrs = SEG6_F_LOCAL_COUNTERS, |
1481 | .input = input_action_end_t, |
1482 | }, |
1483 | { |
1484 | .action = SEG6_LOCAL_ACTION_END_DX2, |
1485 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_OIF), |
1486 | .optattrs = SEG6_F_LOCAL_COUNTERS, |
1487 | .input = input_action_end_dx2, |
1488 | }, |
1489 | { |
1490 | .action = SEG6_LOCAL_ACTION_END_DX6, |
1491 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6), |
1492 | .optattrs = SEG6_F_LOCAL_COUNTERS, |
1493 | .input = input_action_end_dx6, |
1494 | }, |
1495 | { |
1496 | .action = SEG6_LOCAL_ACTION_END_DX4, |
1497 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH4), |
1498 | .optattrs = SEG6_F_LOCAL_COUNTERS, |
1499 | .input = input_action_end_dx4, |
1500 | }, |
1501 | { |
1502 | .action = SEG6_LOCAL_ACTION_END_DT4, |
1503 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), |
1504 | .optattrs = SEG6_F_LOCAL_COUNTERS, |
1505 | #ifdef CONFIG_NET_L3_MASTER_DEV |
1506 | .input = input_action_end_dt4, |
1507 | .slwt_ops = { |
1508 | .build_state = seg6_end_dt4_build, |
1509 | }, |
1510 | #endif |
1511 | }, |
1512 | { |
1513 | .action = SEG6_LOCAL_ACTION_END_DT6, |
1514 | #ifdef CONFIG_NET_L3_MASTER_DEV |
1515 | .attrs = 0, |
1516 | .optattrs = SEG6_F_LOCAL_COUNTERS | |
1517 | SEG6_F_ATTR(SEG6_LOCAL_TABLE) | |
1518 | SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), |
1519 | .slwt_ops = { |
1520 | .build_state = seg6_end_dt6_build, |
1521 | }, |
1522 | #else |
1523 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE), |
1524 | .optattrs = SEG6_F_LOCAL_COUNTERS, |
1525 | #endif |
1526 | .input = input_action_end_dt6, |
1527 | }, |
1528 | { |
1529 | .action = SEG6_LOCAL_ACTION_END_DT46, |
1530 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), |
1531 | .optattrs = SEG6_F_LOCAL_COUNTERS, |
1532 | #ifdef CONFIG_NET_L3_MASTER_DEV |
1533 | .input = input_action_end_dt46, |
1534 | .slwt_ops = { |
1535 | .build_state = seg6_end_dt46_build, |
1536 | }, |
1537 | #endif |
1538 | }, |
1539 | { |
1540 | .action = SEG6_LOCAL_ACTION_END_B6, |
1541 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH), |
1542 | .optattrs = SEG6_F_LOCAL_COUNTERS, |
1543 | .input = input_action_end_b6, |
1544 | }, |
1545 | { |
1546 | .action = SEG6_LOCAL_ACTION_END_B6_ENCAP, |
1547 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH), |
1548 | .optattrs = SEG6_F_LOCAL_COUNTERS, |
1549 | .input = input_action_end_b6_encap, |
1550 | .static_headroom = sizeof(struct ipv6hdr), |
1551 | }, |
1552 | { |
1553 | .action = SEG6_LOCAL_ACTION_END_BPF, |
1554 | .attrs = SEG6_F_ATTR(SEG6_LOCAL_BPF), |
1555 | .optattrs = SEG6_F_LOCAL_COUNTERS, |
1556 | .input = input_action_end_bpf, |
1557 | }, |
1558 | |
1559 | }; |
1560 | |
1561 | static struct seg6_action_desc *__get_action_desc(int action) |
1562 | { |
1563 | struct seg6_action_desc *desc; |
1564 | int i, count; |
1565 | |
1566 | count = ARRAY_SIZE(seg6_action_table); |
1567 | for (i = 0; i < count; i++) { |
1568 | desc = &seg6_action_table[i]; |
1569 | if (desc->action == action) |
1570 | return desc; |
1571 | } |
1572 | |
1573 | return NULL; |
1574 | } |
1575 | |
1576 | static bool seg6_lwtunnel_counters_enabled(struct seg6_local_lwt *slwt) |
1577 | { |
1578 | return slwt->parsed_optattrs & SEG6_F_LOCAL_COUNTERS; |
1579 | } |
1580 | |
1581 | static void seg6_local_update_counters(struct seg6_local_lwt *slwt, |
1582 | unsigned int len, int err) |
1583 | { |
1584 | struct pcpu_seg6_local_counters *pcounters; |
1585 | |
1586 | pcounters = this_cpu_ptr(slwt->pcpu_counters); |
1587 | u64_stats_update_begin(syncp: &pcounters->syncp); |
1588 | |
1589 | if (likely(!err)) { |
1590 | u64_stats_inc(p: &pcounters->packets); |
1591 | u64_stats_add(p: &pcounters->bytes, val: len); |
1592 | } else { |
1593 | u64_stats_inc(p: &pcounters->errors); |
1594 | } |
1595 | |
1596 | u64_stats_update_end(syncp: &pcounters->syncp); |
1597 | } |
1598 | |
1599 | static int seg6_local_input_core(struct net *net, struct sock *sk, |
1600 | struct sk_buff *skb) |
1601 | { |
1602 | struct dst_entry *orig_dst = skb_dst(skb); |
1603 | struct seg6_action_desc *desc; |
1604 | struct seg6_local_lwt *slwt; |
1605 | unsigned int len = skb->len; |
1606 | int rc; |
1607 | |
1608 | slwt = seg6_local_lwtunnel(lwt: orig_dst->lwtstate); |
1609 | desc = slwt->desc; |
1610 | |
1611 | rc = desc->input(skb, slwt); |
1612 | |
1613 | if (!seg6_lwtunnel_counters_enabled(slwt)) |
1614 | return rc; |
1615 | |
1616 | seg6_local_update_counters(slwt, len, err: rc); |
1617 | |
1618 | return rc; |
1619 | } |
1620 | |
1621 | static int seg6_local_input(struct sk_buff *skb) |
1622 | { |
1623 | if (skb->protocol != htons(ETH_P_IPV6)) { |
1624 | kfree_skb(skb); |
1625 | return -EINVAL; |
1626 | } |
1627 | |
1628 | if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) |
1629 | return NF_HOOK(pf: NFPROTO_IPV6, hook: NF_INET_LOCAL_IN, |
1630 | net: dev_net(dev: skb->dev), NULL, skb, in: skb->dev, NULL, |
1631 | okfn: seg6_local_input_core); |
1632 | |
1633 | return seg6_local_input_core(net: dev_net(dev: skb->dev), NULL, skb); |
1634 | } |
1635 | |
1636 | static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = { |
1637 | [SEG6_LOCAL_ACTION] = { .type = NLA_U32 }, |
1638 | [SEG6_LOCAL_SRH] = { .type = NLA_BINARY }, |
1639 | [SEG6_LOCAL_TABLE] = { .type = NLA_U32 }, |
1640 | [SEG6_LOCAL_VRFTABLE] = { .type = NLA_U32 }, |
1641 | [SEG6_LOCAL_NH4] = { .type = NLA_BINARY, |
1642 | .len = sizeof(struct in_addr) }, |
1643 | [SEG6_LOCAL_NH6] = { .type = NLA_BINARY, |
1644 | .len = sizeof(struct in6_addr) }, |
1645 | [SEG6_LOCAL_IIF] = { .type = NLA_U32 }, |
1646 | [SEG6_LOCAL_OIF] = { .type = NLA_U32 }, |
1647 | [SEG6_LOCAL_BPF] = { .type = NLA_NESTED }, |
1648 | [SEG6_LOCAL_COUNTERS] = { .type = NLA_NESTED }, |
1649 | [SEG6_LOCAL_FLAVORS] = { .type = NLA_NESTED }, |
1650 | }; |
1651 | |
1652 | static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt, |
1653 | struct netlink_ext_ack *extack) |
1654 | { |
1655 | struct ipv6_sr_hdr *srh; |
1656 | int len; |
1657 | |
1658 | srh = nla_data(nla: attrs[SEG6_LOCAL_SRH]); |
1659 | len = nla_len(nla: attrs[SEG6_LOCAL_SRH]); |
1660 | |
1661 | /* SRH must contain at least one segment */ |
1662 | if (len < sizeof(*srh) + sizeof(struct in6_addr)) |
1663 | return -EINVAL; |
1664 | |
1665 | if (!seg6_validate_srh(srh, len, reduced: false)) |
1666 | return -EINVAL; |
1667 | |
1668 | slwt->srh = kmemdup(p: srh, size: len, GFP_KERNEL); |
1669 | if (!slwt->srh) |
1670 | return -ENOMEM; |
1671 | |
1672 | slwt->headroom += len; |
1673 | |
1674 | return 0; |
1675 | } |
1676 | |
1677 | static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
1678 | { |
1679 | struct ipv6_sr_hdr *srh; |
1680 | struct nlattr *nla; |
1681 | int len; |
1682 | |
1683 | srh = slwt->srh; |
1684 | len = (srh->hdrlen + 1) << 3; |
1685 | |
1686 | nla = nla_reserve(skb, attrtype: SEG6_LOCAL_SRH, attrlen: len); |
1687 | if (!nla) |
1688 | return -EMSGSIZE; |
1689 | |
1690 | memcpy(nla_data(nla), srh, len); |
1691 | |
1692 | return 0; |
1693 | } |
1694 | |
1695 | static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b) |
1696 | { |
1697 | int len = (a->srh->hdrlen + 1) << 3; |
1698 | |
1699 | if (len != ((b->srh->hdrlen + 1) << 3)) |
1700 | return 1; |
1701 | |
1702 | return memcmp(p: a->srh, q: b->srh, size: len); |
1703 | } |
1704 | |
1705 | static void destroy_attr_srh(struct seg6_local_lwt *slwt) |
1706 | { |
1707 | kfree(objp: slwt->srh); |
1708 | } |
1709 | |
1710 | static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt, |
1711 | struct netlink_ext_ack *extack) |
1712 | { |
1713 | slwt->table = nla_get_u32(nla: attrs[SEG6_LOCAL_TABLE]); |
1714 | |
1715 | return 0; |
1716 | } |
1717 | |
1718 | static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
1719 | { |
1720 | if (nla_put_u32(skb, attrtype: SEG6_LOCAL_TABLE, value: slwt->table)) |
1721 | return -EMSGSIZE; |
1722 | |
1723 | return 0; |
1724 | } |
1725 | |
1726 | static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b) |
1727 | { |
1728 | if (a->table != b->table) |
1729 | return 1; |
1730 | |
1731 | return 0; |
1732 | } |
1733 | |
1734 | static struct |
1735 | seg6_end_dt_info *seg6_possible_end_dt_info(struct seg6_local_lwt *slwt) |
1736 | { |
1737 | #ifdef CONFIG_NET_L3_MASTER_DEV |
1738 | return &slwt->dt_info; |
1739 | #else |
1740 | return ERR_PTR(-EOPNOTSUPP); |
1741 | #endif |
1742 | } |
1743 | |
1744 | static int parse_nla_vrftable(struct nlattr **attrs, |
1745 | struct seg6_local_lwt *slwt, |
1746 | struct netlink_ext_ack *extack) |
1747 | { |
1748 | struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt); |
1749 | |
1750 | if (IS_ERR(ptr: info)) |
1751 | return PTR_ERR(ptr: info); |
1752 | |
1753 | info->vrf_table = nla_get_u32(nla: attrs[SEG6_LOCAL_VRFTABLE]); |
1754 | |
1755 | return 0; |
1756 | } |
1757 | |
1758 | static int put_nla_vrftable(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
1759 | { |
1760 | struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt); |
1761 | |
1762 | if (IS_ERR(ptr: info)) |
1763 | return PTR_ERR(ptr: info); |
1764 | |
1765 | if (nla_put_u32(skb, attrtype: SEG6_LOCAL_VRFTABLE, value: info->vrf_table)) |
1766 | return -EMSGSIZE; |
1767 | |
1768 | return 0; |
1769 | } |
1770 | |
1771 | static int cmp_nla_vrftable(struct seg6_local_lwt *a, struct seg6_local_lwt *b) |
1772 | { |
1773 | struct seg6_end_dt_info *info_a = seg6_possible_end_dt_info(slwt: a); |
1774 | struct seg6_end_dt_info *info_b = seg6_possible_end_dt_info(slwt: b); |
1775 | |
1776 | if (info_a->vrf_table != info_b->vrf_table) |
1777 | return 1; |
1778 | |
1779 | return 0; |
1780 | } |
1781 | |
1782 | static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt, |
1783 | struct netlink_ext_ack *extack) |
1784 | { |
1785 | memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]), |
1786 | sizeof(struct in_addr)); |
1787 | |
1788 | return 0; |
1789 | } |
1790 | |
1791 | static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
1792 | { |
1793 | struct nlattr *nla; |
1794 | |
1795 | nla = nla_reserve(skb, attrtype: SEG6_LOCAL_NH4, attrlen: sizeof(struct in_addr)); |
1796 | if (!nla) |
1797 | return -EMSGSIZE; |
1798 | |
1799 | memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr)); |
1800 | |
1801 | return 0; |
1802 | } |
1803 | |
1804 | static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b) |
1805 | { |
1806 | return memcmp(p: &a->nh4, q: &b->nh4, size: sizeof(struct in_addr)); |
1807 | } |
1808 | |
1809 | static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt, |
1810 | struct netlink_ext_ack *extack) |
1811 | { |
1812 | memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]), |
1813 | sizeof(struct in6_addr)); |
1814 | |
1815 | return 0; |
1816 | } |
1817 | |
1818 | static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
1819 | { |
1820 | struct nlattr *nla; |
1821 | |
1822 | nla = nla_reserve(skb, attrtype: SEG6_LOCAL_NH6, attrlen: sizeof(struct in6_addr)); |
1823 | if (!nla) |
1824 | return -EMSGSIZE; |
1825 | |
1826 | memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr)); |
1827 | |
1828 | return 0; |
1829 | } |
1830 | |
1831 | static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b) |
1832 | { |
1833 | return memcmp(p: &a->nh6, q: &b->nh6, size: sizeof(struct in6_addr)); |
1834 | } |
1835 | |
1836 | static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt, |
1837 | struct netlink_ext_ack *extack) |
1838 | { |
1839 | slwt->iif = nla_get_u32(nla: attrs[SEG6_LOCAL_IIF]); |
1840 | |
1841 | return 0; |
1842 | } |
1843 | |
1844 | static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
1845 | { |
1846 | if (nla_put_u32(skb, attrtype: SEG6_LOCAL_IIF, value: slwt->iif)) |
1847 | return -EMSGSIZE; |
1848 | |
1849 | return 0; |
1850 | } |
1851 | |
1852 | static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b) |
1853 | { |
1854 | if (a->iif != b->iif) |
1855 | return 1; |
1856 | |
1857 | return 0; |
1858 | } |
1859 | |
1860 | static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt, |
1861 | struct netlink_ext_ack *extack) |
1862 | { |
1863 | slwt->oif = nla_get_u32(nla: attrs[SEG6_LOCAL_OIF]); |
1864 | |
1865 | return 0; |
1866 | } |
1867 | |
1868 | static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
1869 | { |
1870 | if (nla_put_u32(skb, attrtype: SEG6_LOCAL_OIF, value: slwt->oif)) |
1871 | return -EMSGSIZE; |
1872 | |
1873 | return 0; |
1874 | } |
1875 | |
1876 | static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b) |
1877 | { |
1878 | if (a->oif != b->oif) |
1879 | return 1; |
1880 | |
1881 | return 0; |
1882 | } |
1883 | |
1884 | #define MAX_PROG_NAME 256 |
1885 | static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = { |
1886 | [SEG6_LOCAL_BPF_PROG] = { .type = NLA_U32, }, |
1887 | [SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING, |
1888 | .len = MAX_PROG_NAME }, |
1889 | }; |
1890 | |
1891 | static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt, |
1892 | struct netlink_ext_ack *extack) |
1893 | { |
1894 | struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1]; |
1895 | struct bpf_prog *p; |
1896 | int ret; |
1897 | u32 fd; |
1898 | |
1899 | ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX, |
1900 | nla: attrs[SEG6_LOCAL_BPF], |
1901 | policy: bpf_prog_policy, NULL); |
1902 | if (ret < 0) |
1903 | return ret; |
1904 | |
1905 | if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME]) |
1906 | return -EINVAL; |
1907 | |
1908 | slwt->bpf.name = nla_memdup(src: tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL); |
1909 | if (!slwt->bpf.name) |
1910 | return -ENOMEM; |
1911 | |
1912 | fd = nla_get_u32(nla: tb[SEG6_LOCAL_BPF_PROG]); |
1913 | p = bpf_prog_get_type(ufd: fd, type: BPF_PROG_TYPE_LWT_SEG6LOCAL); |
1914 | if (IS_ERR(ptr: p)) { |
1915 | kfree(objp: slwt->bpf.name); |
1916 | return PTR_ERR(ptr: p); |
1917 | } |
1918 | |
1919 | slwt->bpf.prog = p; |
1920 | return 0; |
1921 | } |
1922 | |
1923 | static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
1924 | { |
1925 | struct nlattr *nest; |
1926 | |
1927 | if (!slwt->bpf.prog) |
1928 | return 0; |
1929 | |
1930 | nest = nla_nest_start_noflag(skb, attrtype: SEG6_LOCAL_BPF); |
1931 | if (!nest) |
1932 | return -EMSGSIZE; |
1933 | |
1934 | if (nla_put_u32(skb, attrtype: SEG6_LOCAL_BPF_PROG, value: slwt->bpf.prog->aux->id)) |
1935 | return -EMSGSIZE; |
1936 | |
1937 | if (slwt->bpf.name && |
1938 | nla_put_string(skb, attrtype: SEG6_LOCAL_BPF_PROG_NAME, str: slwt->bpf.name)) |
1939 | return -EMSGSIZE; |
1940 | |
1941 | return nla_nest_end(skb, start: nest); |
1942 | } |
1943 | |
1944 | static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b) |
1945 | { |
1946 | if (!a->bpf.name && !b->bpf.name) |
1947 | return 0; |
1948 | |
1949 | if (!a->bpf.name || !b->bpf.name) |
1950 | return 1; |
1951 | |
1952 | return strcmp(a->bpf.name, b->bpf.name); |
1953 | } |
1954 | |
1955 | static void destroy_attr_bpf(struct seg6_local_lwt *slwt) |
1956 | { |
1957 | kfree(objp: slwt->bpf.name); |
1958 | if (slwt->bpf.prog) |
1959 | bpf_prog_put(prog: slwt->bpf.prog); |
1960 | } |
1961 | |
1962 | static const struct |
1963 | nla_policy seg6_local_counters_policy[SEG6_LOCAL_CNT_MAX + 1] = { |
1964 | [SEG6_LOCAL_CNT_PACKETS] = { .type = NLA_U64 }, |
1965 | [SEG6_LOCAL_CNT_BYTES] = { .type = NLA_U64 }, |
1966 | [SEG6_LOCAL_CNT_ERRORS] = { .type = NLA_U64 }, |
1967 | }; |
1968 | |
1969 | static int parse_nla_counters(struct nlattr **attrs, |
1970 | struct seg6_local_lwt *slwt, |
1971 | struct netlink_ext_ack *extack) |
1972 | { |
1973 | struct pcpu_seg6_local_counters __percpu *pcounters; |
1974 | struct nlattr *tb[SEG6_LOCAL_CNT_MAX + 1]; |
1975 | int ret; |
1976 | |
1977 | ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_CNT_MAX, |
1978 | nla: attrs[SEG6_LOCAL_COUNTERS], |
1979 | policy: seg6_local_counters_policy, NULL); |
1980 | if (ret < 0) |
1981 | return ret; |
1982 | |
1983 | /* basic support for SRv6 Behavior counters requires at least: |
1984 | * packets, bytes and errors. |
1985 | */ |
1986 | if (!tb[SEG6_LOCAL_CNT_PACKETS] || !tb[SEG6_LOCAL_CNT_BYTES] || |
1987 | !tb[SEG6_LOCAL_CNT_ERRORS]) |
1988 | return -EINVAL; |
1989 | |
1990 | /* counters are always zero initialized */ |
1991 | pcounters = seg6_local_alloc_pcpu_counters(GFP_KERNEL); |
1992 | if (!pcounters) |
1993 | return -ENOMEM; |
1994 | |
1995 | slwt->pcpu_counters = pcounters; |
1996 | |
1997 | return 0; |
1998 | } |
1999 | |
2000 | static int seg6_local_fill_nla_counters(struct sk_buff *skb, |
2001 | struct seg6_local_counters *counters) |
2002 | { |
2003 | if (nla_put_u64_64bit(skb, attrtype: SEG6_LOCAL_CNT_PACKETS, value: counters->packets, |
2004 | padattr: SEG6_LOCAL_CNT_PAD)) |
2005 | return -EMSGSIZE; |
2006 | |
2007 | if (nla_put_u64_64bit(skb, attrtype: SEG6_LOCAL_CNT_BYTES, value: counters->bytes, |
2008 | padattr: SEG6_LOCAL_CNT_PAD)) |
2009 | return -EMSGSIZE; |
2010 | |
2011 | if (nla_put_u64_64bit(skb, attrtype: SEG6_LOCAL_CNT_ERRORS, value: counters->errors, |
2012 | padattr: SEG6_LOCAL_CNT_PAD)) |
2013 | return -EMSGSIZE; |
2014 | |
2015 | return 0; |
2016 | } |
2017 | |
2018 | static int put_nla_counters(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
2019 | { |
2020 | struct seg6_local_counters counters = { 0, 0, 0 }; |
2021 | struct nlattr *nest; |
2022 | int rc, i; |
2023 | |
2024 | nest = nla_nest_start(skb, attrtype: SEG6_LOCAL_COUNTERS); |
2025 | if (!nest) |
2026 | return -EMSGSIZE; |
2027 | |
2028 | for_each_possible_cpu(i) { |
2029 | struct pcpu_seg6_local_counters *pcounters; |
2030 | u64 packets, bytes, errors; |
2031 | unsigned int start; |
2032 | |
2033 | pcounters = per_cpu_ptr(slwt->pcpu_counters, i); |
2034 | do { |
2035 | start = u64_stats_fetch_begin(syncp: &pcounters->syncp); |
2036 | |
2037 | packets = u64_stats_read(p: &pcounters->packets); |
2038 | bytes = u64_stats_read(p: &pcounters->bytes); |
2039 | errors = u64_stats_read(p: &pcounters->errors); |
2040 | |
2041 | } while (u64_stats_fetch_retry(syncp: &pcounters->syncp, start)); |
2042 | |
2043 | counters.packets += packets; |
2044 | counters.bytes += bytes; |
2045 | counters.errors += errors; |
2046 | } |
2047 | |
2048 | rc = seg6_local_fill_nla_counters(skb, counters: &counters); |
2049 | if (rc < 0) { |
2050 | nla_nest_cancel(skb, start: nest); |
2051 | return rc; |
2052 | } |
2053 | |
2054 | return nla_nest_end(skb, start: nest); |
2055 | } |
2056 | |
2057 | static int cmp_nla_counters(struct seg6_local_lwt *a, struct seg6_local_lwt *b) |
2058 | { |
2059 | /* a and b are equal if both have pcpu_counters set or not */ |
2060 | return (!!((unsigned long)a->pcpu_counters)) ^ |
2061 | (!!((unsigned long)b->pcpu_counters)); |
2062 | } |
2063 | |
2064 | static void destroy_attr_counters(struct seg6_local_lwt *slwt) |
2065 | { |
2066 | free_percpu(pdata: slwt->pcpu_counters); |
2067 | } |
2068 | |
2069 | static const |
2070 | struct nla_policy seg6_local_flavors_policy[SEG6_LOCAL_FLV_MAX + 1] = { |
2071 | [SEG6_LOCAL_FLV_OPERATION] = { .type = NLA_U32 }, |
2072 | [SEG6_LOCAL_FLV_LCBLOCK_BITS] = { .type = NLA_U8 }, |
2073 | [SEG6_LOCAL_FLV_LCNODE_FN_BITS] = { .type = NLA_U8 }, |
2074 | }; |
2075 | |
2076 | /* check whether the lengths of the Locator-Block and Locator-Node Function |
2077 | * are compatible with the dimension of a C-SID container. |
2078 | */ |
2079 | static int seg6_chk_next_csid_cfg(__u8 block_len, __u8 func_len) |
2080 | { |
2081 | /* Locator-Block and Locator-Node Function cannot exceed 128 bits |
2082 | * (i.e. C-SID container lenghts). |
2083 | */ |
2084 | if (next_csid_chk_cntr_bits(block_len, func_len)) |
2085 | return -EINVAL; |
2086 | |
2087 | /* Locator-Block length must be greater than zero and evenly divisible |
2088 | * by 8. There must be room for a Locator-Node Function, at least. |
2089 | */ |
2090 | if (next_csid_chk_lcblock_bits(block_len)) |
2091 | return -EINVAL; |
2092 | |
2093 | /* Locator-Node Function length must be greater than zero and evenly |
2094 | * divisible by 8. There must be room for the Locator-Block. |
2095 | */ |
2096 | if (next_csid_chk_lcnode_fn_bits(func_len)) |
2097 | return -EINVAL; |
2098 | |
2099 | return 0; |
2100 | } |
2101 | |
2102 | static int seg6_parse_nla_next_csid_cfg(struct nlattr **tb, |
2103 | struct seg6_flavors_info *finfo, |
2104 | struct netlink_ext_ack *extack) |
2105 | { |
2106 | __u8 func_len = SEG6_LOCAL_LCNODE_FN_DBITS; |
2107 | __u8 block_len = SEG6_LOCAL_LCBLOCK_DBITS; |
2108 | int rc; |
2109 | |
2110 | if (tb[SEG6_LOCAL_FLV_LCBLOCK_BITS]) |
2111 | block_len = nla_get_u8(nla: tb[SEG6_LOCAL_FLV_LCBLOCK_BITS]); |
2112 | |
2113 | if (tb[SEG6_LOCAL_FLV_LCNODE_FN_BITS]) |
2114 | func_len = nla_get_u8(nla: tb[SEG6_LOCAL_FLV_LCNODE_FN_BITS]); |
2115 | |
2116 | rc = seg6_chk_next_csid_cfg(block_len, func_len); |
2117 | if (rc < 0) { |
2118 | NL_SET_ERR_MSG(extack, |
2119 | "Invalid Locator Block/Node Function lengths" ); |
2120 | return rc; |
2121 | } |
2122 | |
2123 | finfo->lcblock_bits = block_len; |
2124 | finfo->lcnode_func_bits = func_len; |
2125 | |
2126 | return 0; |
2127 | } |
2128 | |
2129 | static int parse_nla_flavors(struct nlattr **attrs, struct seg6_local_lwt *slwt, |
2130 | struct netlink_ext_ack *extack) |
2131 | { |
2132 | struct seg6_flavors_info *finfo = &slwt->flv_info; |
2133 | struct nlattr *tb[SEG6_LOCAL_FLV_MAX + 1]; |
2134 | int action = slwt->action; |
2135 | __u32 fops, supp_fops; |
2136 | int rc; |
2137 | |
2138 | rc = nla_parse_nested_deprecated(tb, SEG6_LOCAL_FLV_MAX, |
2139 | nla: attrs[SEG6_LOCAL_FLAVORS], |
2140 | policy: seg6_local_flavors_policy, NULL); |
2141 | if (rc < 0) |
2142 | return rc; |
2143 | |
2144 | /* this attribute MUST always be present since it represents the Flavor |
2145 | * operation(s) to be carried out. |
2146 | */ |
2147 | if (!tb[SEG6_LOCAL_FLV_OPERATION]) |
2148 | return -EINVAL; |
2149 | |
2150 | fops = nla_get_u32(nla: tb[SEG6_LOCAL_FLV_OPERATION]); |
2151 | rc = seg6_flv_supp_ops_by_action(action, fops: &supp_fops); |
2152 | if (rc < 0 || (fops & ~supp_fops)) { |
2153 | NL_SET_ERR_MSG(extack, "Unsupported Flavor operation(s)" ); |
2154 | return -EOPNOTSUPP; |
2155 | } |
2156 | |
2157 | finfo->flv_ops = fops; |
2158 | |
2159 | if (seg6_next_csid_enabled(fops)) { |
2160 | /* Locator-Block and Locator-Node Function lengths can be |
2161 | * provided by the user space. Otherwise, default values are |
2162 | * applied. |
2163 | */ |
2164 | rc = seg6_parse_nla_next_csid_cfg(tb, finfo, extack); |
2165 | if (rc < 0) |
2166 | return rc; |
2167 | } |
2168 | |
2169 | return 0; |
2170 | } |
2171 | |
2172 | static int seg6_fill_nla_next_csid_cfg(struct sk_buff *skb, |
2173 | struct seg6_flavors_info *finfo) |
2174 | { |
2175 | if (nla_put_u8(skb, attrtype: SEG6_LOCAL_FLV_LCBLOCK_BITS, value: finfo->lcblock_bits)) |
2176 | return -EMSGSIZE; |
2177 | |
2178 | if (nla_put_u8(skb, attrtype: SEG6_LOCAL_FLV_LCNODE_FN_BITS, |
2179 | value: finfo->lcnode_func_bits)) |
2180 | return -EMSGSIZE; |
2181 | |
2182 | return 0; |
2183 | } |
2184 | |
2185 | static int put_nla_flavors(struct sk_buff *skb, struct seg6_local_lwt *slwt) |
2186 | { |
2187 | struct seg6_flavors_info *finfo = &slwt->flv_info; |
2188 | __u32 fops = finfo->flv_ops; |
2189 | struct nlattr *nest; |
2190 | int rc; |
2191 | |
2192 | nest = nla_nest_start(skb, attrtype: SEG6_LOCAL_FLAVORS); |
2193 | if (!nest) |
2194 | return -EMSGSIZE; |
2195 | |
2196 | if (nla_put_u32(skb, attrtype: SEG6_LOCAL_FLV_OPERATION, value: fops)) { |
2197 | rc = -EMSGSIZE; |
2198 | goto err; |
2199 | } |
2200 | |
2201 | if (seg6_next_csid_enabled(fops)) { |
2202 | rc = seg6_fill_nla_next_csid_cfg(skb, finfo); |
2203 | if (rc < 0) |
2204 | goto err; |
2205 | } |
2206 | |
2207 | return nla_nest_end(skb, start: nest); |
2208 | |
2209 | err: |
2210 | nla_nest_cancel(skb, start: nest); |
2211 | return rc; |
2212 | } |
2213 | |
2214 | static int seg6_cmp_nla_next_csid_cfg(struct seg6_flavors_info *finfo_a, |
2215 | struct seg6_flavors_info *finfo_b) |
2216 | { |
2217 | if (finfo_a->lcblock_bits != finfo_b->lcblock_bits) |
2218 | return 1; |
2219 | |
2220 | if (finfo_a->lcnode_func_bits != finfo_b->lcnode_func_bits) |
2221 | return 1; |
2222 | |
2223 | return 0; |
2224 | } |
2225 | |
2226 | static int cmp_nla_flavors(struct seg6_local_lwt *a, struct seg6_local_lwt *b) |
2227 | { |
2228 | struct seg6_flavors_info *finfo_a = &a->flv_info; |
2229 | struct seg6_flavors_info *finfo_b = &b->flv_info; |
2230 | |
2231 | if (finfo_a->flv_ops != finfo_b->flv_ops) |
2232 | return 1; |
2233 | |
2234 | if (seg6_next_csid_enabled(fops: finfo_a->flv_ops)) { |
2235 | if (seg6_cmp_nla_next_csid_cfg(finfo_a, finfo_b)) |
2236 | return 1; |
2237 | } |
2238 | |
2239 | return 0; |
2240 | } |
2241 | |
2242 | static int encap_size_flavors(struct seg6_local_lwt *slwt) |
2243 | { |
2244 | struct seg6_flavors_info *finfo = &slwt->flv_info; |
2245 | int nlsize; |
2246 | |
2247 | nlsize = nla_total_size(payload: 0) + /* nest SEG6_LOCAL_FLAVORS */ |
2248 | nla_total_size(payload: 4); /* SEG6_LOCAL_FLV_OPERATION */ |
2249 | |
2250 | if (seg6_next_csid_enabled(fops: finfo->flv_ops)) |
2251 | nlsize += nla_total_size(payload: 1) + /* SEG6_LOCAL_FLV_LCBLOCK_BITS */ |
2252 | nla_total_size(payload: 1); /* SEG6_LOCAL_FLV_LCNODE_FN_BITS */ |
2253 | |
2254 | return nlsize; |
2255 | } |
2256 | |
2257 | struct seg6_action_param { |
2258 | int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt, |
2259 | struct netlink_ext_ack *extack); |
2260 | int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt); |
2261 | int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b); |
2262 | |
2263 | /* optional destroy() callback useful for releasing resources which |
2264 | * have been previously acquired in the corresponding parse() |
2265 | * function. |
2266 | */ |
2267 | void (*destroy)(struct seg6_local_lwt *slwt); |
2268 | }; |
2269 | |
2270 | static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = { |
2271 | [SEG6_LOCAL_SRH] = { .parse = parse_nla_srh, |
2272 | .put = put_nla_srh, |
2273 | .cmp = cmp_nla_srh, |
2274 | .destroy = destroy_attr_srh }, |
2275 | |
2276 | [SEG6_LOCAL_TABLE] = { .parse = parse_nla_table, |
2277 | .put = put_nla_table, |
2278 | .cmp = cmp_nla_table }, |
2279 | |
2280 | [SEG6_LOCAL_NH4] = { .parse = parse_nla_nh4, |
2281 | .put = put_nla_nh4, |
2282 | .cmp = cmp_nla_nh4 }, |
2283 | |
2284 | [SEG6_LOCAL_NH6] = { .parse = parse_nla_nh6, |
2285 | .put = put_nla_nh6, |
2286 | .cmp = cmp_nla_nh6 }, |
2287 | |
2288 | [SEG6_LOCAL_IIF] = { .parse = parse_nla_iif, |
2289 | .put = put_nla_iif, |
2290 | .cmp = cmp_nla_iif }, |
2291 | |
2292 | [SEG6_LOCAL_OIF] = { .parse = parse_nla_oif, |
2293 | .put = put_nla_oif, |
2294 | .cmp = cmp_nla_oif }, |
2295 | |
2296 | [SEG6_LOCAL_BPF] = { .parse = parse_nla_bpf, |
2297 | .put = put_nla_bpf, |
2298 | .cmp = cmp_nla_bpf, |
2299 | .destroy = destroy_attr_bpf }, |
2300 | |
2301 | [SEG6_LOCAL_VRFTABLE] = { .parse = parse_nla_vrftable, |
2302 | .put = put_nla_vrftable, |
2303 | .cmp = cmp_nla_vrftable }, |
2304 | |
2305 | [SEG6_LOCAL_COUNTERS] = { .parse = parse_nla_counters, |
2306 | .put = put_nla_counters, |
2307 | .cmp = cmp_nla_counters, |
2308 | .destroy = destroy_attr_counters }, |
2309 | |
2310 | [SEG6_LOCAL_FLAVORS] = { .parse = parse_nla_flavors, |
2311 | .put = put_nla_flavors, |
2312 | .cmp = cmp_nla_flavors }, |
2313 | }; |
2314 | |
2315 | /* call the destroy() callback (if available) for each set attribute in |
2316 | * @parsed_attrs, starting from the first attribute up to the @max_parsed |
2317 | * (excluded) attribute. |
2318 | */ |
2319 | static void __destroy_attrs(unsigned long parsed_attrs, int max_parsed, |
2320 | struct seg6_local_lwt *slwt) |
2321 | { |
2322 | struct seg6_action_param *param; |
2323 | int i; |
2324 | |
2325 | /* Every required seg6local attribute is identified by an ID which is |
2326 | * encoded as a flag (i.e: 1 << ID) in the 'attrs' bitmask; |
2327 | * |
2328 | * We scan the 'parsed_attrs' bitmask, starting from the first attribute |
2329 | * up to the @max_parsed (excluded) attribute. |
2330 | * For each set attribute, we retrieve the corresponding destroy() |
2331 | * callback. If the callback is not available, then we skip to the next |
2332 | * attribute; otherwise, we call the destroy() callback. |
2333 | */ |
2334 | for (i = SEG6_LOCAL_SRH; i < max_parsed; ++i) { |
2335 | if (!(parsed_attrs & SEG6_F_ATTR(i))) |
2336 | continue; |
2337 | |
2338 | param = &seg6_action_params[i]; |
2339 | |
2340 | if (param->destroy) |
2341 | param->destroy(slwt); |
2342 | } |
2343 | } |
2344 | |
2345 | /* release all the resources that may have been acquired during parsing |
2346 | * operations. |
2347 | */ |
2348 | static void destroy_attrs(struct seg6_local_lwt *slwt) |
2349 | { |
2350 | unsigned long attrs = slwt->desc->attrs | slwt->parsed_optattrs; |
2351 | |
2352 | __destroy_attrs(parsed_attrs: attrs, SEG6_LOCAL_MAX + 1, slwt); |
2353 | } |
2354 | |
2355 | static int parse_nla_optional_attrs(struct nlattr **attrs, |
2356 | struct seg6_local_lwt *slwt, |
2357 | struct netlink_ext_ack *extack) |
2358 | { |
2359 | struct seg6_action_desc *desc = slwt->desc; |
2360 | unsigned long parsed_optattrs = 0; |
2361 | struct seg6_action_param *param; |
2362 | int err, i; |
2363 | |
2364 | for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; ++i) { |
2365 | if (!(desc->optattrs & SEG6_F_ATTR(i)) || !attrs[i]) |
2366 | continue; |
2367 | |
2368 | /* once here, the i-th attribute is provided by the |
2369 | * userspace AND it is identified optional as well. |
2370 | */ |
2371 | param = &seg6_action_params[i]; |
2372 | |
2373 | err = param->parse(attrs, slwt, extack); |
2374 | if (err < 0) |
2375 | goto parse_optattrs_err; |
2376 | |
2377 | /* current attribute has been correctly parsed */ |
2378 | parsed_optattrs |= SEG6_F_ATTR(i); |
2379 | } |
2380 | |
2381 | /* store in the tunnel state all the optional attributed successfully |
2382 | * parsed. |
2383 | */ |
2384 | slwt->parsed_optattrs = parsed_optattrs; |
2385 | |
2386 | return 0; |
2387 | |
2388 | parse_optattrs_err: |
2389 | __destroy_attrs(parsed_attrs: parsed_optattrs, max_parsed: i, slwt); |
2390 | |
2391 | return err; |
2392 | } |
2393 | |
2394 | /* call the custom constructor of the behavior during its initialization phase |
2395 | * and after that all its attributes have been parsed successfully. |
2396 | */ |
2397 | static int |
2398 | seg6_local_lwtunnel_build_state(struct seg6_local_lwt *slwt, const void *cfg, |
2399 | struct netlink_ext_ack *extack) |
2400 | { |
2401 | struct seg6_action_desc *desc = slwt->desc; |
2402 | struct seg6_local_lwtunnel_ops *ops; |
2403 | |
2404 | ops = &desc->slwt_ops; |
2405 | if (!ops->build_state) |
2406 | return 0; |
2407 | |
2408 | return ops->build_state(slwt, cfg, extack); |
2409 | } |
2410 | |
2411 | /* call the custom destructor of the behavior which is invoked before the |
2412 | * tunnel is going to be destroyed. |
2413 | */ |
2414 | static void seg6_local_lwtunnel_destroy_state(struct seg6_local_lwt *slwt) |
2415 | { |
2416 | struct seg6_action_desc *desc = slwt->desc; |
2417 | struct seg6_local_lwtunnel_ops *ops; |
2418 | |
2419 | ops = &desc->slwt_ops; |
2420 | if (!ops->destroy_state) |
2421 | return; |
2422 | |
2423 | ops->destroy_state(slwt); |
2424 | } |
2425 | |
2426 | static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt, |
2427 | struct netlink_ext_ack *extack) |
2428 | { |
2429 | struct seg6_action_param *param; |
2430 | struct seg6_action_desc *desc; |
2431 | unsigned long invalid_attrs; |
2432 | int i, err; |
2433 | |
2434 | desc = __get_action_desc(action: slwt->action); |
2435 | if (!desc) |
2436 | return -EINVAL; |
2437 | |
2438 | if (!desc->input) |
2439 | return -EOPNOTSUPP; |
2440 | |
2441 | slwt->desc = desc; |
2442 | slwt->headroom += desc->static_headroom; |
2443 | |
2444 | /* Forcing the desc->optattrs *set* and the desc->attrs *set* to be |
2445 | * disjoined, this allow us to release acquired resources by optional |
2446 | * attributes and by required attributes independently from each other |
2447 | * without any interference. |
2448 | * In other terms, we are sure that we do not release some the acquired |
2449 | * resources twice. |
2450 | * |
2451 | * Note that if an attribute is configured both as required and as |
2452 | * optional, it means that the user has messed something up in the |
2453 | * seg6_action_table. Therefore, this check is required for SRv6 |
2454 | * behaviors to work properly. |
2455 | */ |
2456 | invalid_attrs = desc->attrs & desc->optattrs; |
2457 | if (invalid_attrs) { |
2458 | WARN_ONCE(1, |
2459 | "An attribute cannot be both required AND optional" ); |
2460 | return -EINVAL; |
2461 | } |
2462 | |
2463 | /* parse the required attributes */ |
2464 | for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { |
2465 | if (desc->attrs & SEG6_F_ATTR(i)) { |
2466 | if (!attrs[i]) |
2467 | return -EINVAL; |
2468 | |
2469 | param = &seg6_action_params[i]; |
2470 | |
2471 | err = param->parse(attrs, slwt, extack); |
2472 | if (err < 0) |
2473 | goto parse_attrs_err; |
2474 | } |
2475 | } |
2476 | |
2477 | /* parse the optional attributes, if any */ |
2478 | err = parse_nla_optional_attrs(attrs, slwt, extack); |
2479 | if (err < 0) |
2480 | goto parse_attrs_err; |
2481 | |
2482 | return 0; |
2483 | |
2484 | parse_attrs_err: |
2485 | /* release any resource that may have been acquired during the i-1 |
2486 | * parse() operations. |
2487 | */ |
2488 | __destroy_attrs(parsed_attrs: desc->attrs, max_parsed: i, slwt); |
2489 | |
2490 | return err; |
2491 | } |
2492 | |
2493 | static int seg6_local_build_state(struct net *net, struct nlattr *nla, |
2494 | unsigned int family, const void *cfg, |
2495 | struct lwtunnel_state **ts, |
2496 | struct netlink_ext_ack *extack) |
2497 | { |
2498 | struct nlattr *tb[SEG6_LOCAL_MAX + 1]; |
2499 | struct lwtunnel_state *newts; |
2500 | struct seg6_local_lwt *slwt; |
2501 | int err; |
2502 | |
2503 | if (family != AF_INET6) |
2504 | return -EINVAL; |
2505 | |
2506 | err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla, |
2507 | policy: seg6_local_policy, extack); |
2508 | |
2509 | if (err < 0) |
2510 | return err; |
2511 | |
2512 | if (!tb[SEG6_LOCAL_ACTION]) |
2513 | return -EINVAL; |
2514 | |
2515 | newts = lwtunnel_state_alloc(hdr_len: sizeof(*slwt)); |
2516 | if (!newts) |
2517 | return -ENOMEM; |
2518 | |
2519 | slwt = seg6_local_lwtunnel(lwt: newts); |
2520 | slwt->action = nla_get_u32(nla: tb[SEG6_LOCAL_ACTION]); |
2521 | |
2522 | err = parse_nla_action(attrs: tb, slwt, extack); |
2523 | if (err < 0) |
2524 | goto out_free; |
2525 | |
2526 | err = seg6_local_lwtunnel_build_state(slwt, cfg, extack); |
2527 | if (err < 0) |
2528 | goto out_destroy_attrs; |
2529 | |
2530 | newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL; |
2531 | newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT; |
2532 | newts->headroom = slwt->headroom; |
2533 | |
2534 | *ts = newts; |
2535 | |
2536 | return 0; |
2537 | |
2538 | out_destroy_attrs: |
2539 | destroy_attrs(slwt); |
2540 | out_free: |
2541 | kfree(objp: newts); |
2542 | return err; |
2543 | } |
2544 | |
2545 | static void seg6_local_destroy_state(struct lwtunnel_state *lwt) |
2546 | { |
2547 | struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); |
2548 | |
2549 | seg6_local_lwtunnel_destroy_state(slwt); |
2550 | |
2551 | destroy_attrs(slwt); |
2552 | |
2553 | return; |
2554 | } |
2555 | |
2556 | static int seg6_local_fill_encap(struct sk_buff *skb, |
2557 | struct lwtunnel_state *lwt) |
2558 | { |
2559 | struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); |
2560 | struct seg6_action_param *param; |
2561 | unsigned long attrs; |
2562 | int i, err; |
2563 | |
2564 | if (nla_put_u32(skb, attrtype: SEG6_LOCAL_ACTION, value: slwt->action)) |
2565 | return -EMSGSIZE; |
2566 | |
2567 | attrs = slwt->desc->attrs | slwt->parsed_optattrs; |
2568 | |
2569 | for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { |
2570 | if (attrs & SEG6_F_ATTR(i)) { |
2571 | param = &seg6_action_params[i]; |
2572 | err = param->put(skb, slwt); |
2573 | if (err < 0) |
2574 | return err; |
2575 | } |
2576 | } |
2577 | |
2578 | return 0; |
2579 | } |
2580 | |
2581 | static int seg6_local_get_encap_size(struct lwtunnel_state *lwt) |
2582 | { |
2583 | struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); |
2584 | unsigned long attrs; |
2585 | int nlsize; |
2586 | |
2587 | nlsize = nla_total_size(payload: 4); /* action */ |
2588 | |
2589 | attrs = slwt->desc->attrs | slwt->parsed_optattrs; |
2590 | |
2591 | if (attrs & SEG6_F_ATTR(SEG6_LOCAL_SRH)) |
2592 | nlsize += nla_total_size(payload: (slwt->srh->hdrlen + 1) << 3); |
2593 | |
2594 | if (attrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE)) |
2595 | nlsize += nla_total_size(payload: 4); |
2596 | |
2597 | if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH4)) |
2598 | nlsize += nla_total_size(payload: 4); |
2599 | |
2600 | if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH6)) |
2601 | nlsize += nla_total_size(payload: 16); |
2602 | |
2603 | if (attrs & SEG6_F_ATTR(SEG6_LOCAL_IIF)) |
2604 | nlsize += nla_total_size(payload: 4); |
2605 | |
2606 | if (attrs & SEG6_F_ATTR(SEG6_LOCAL_OIF)) |
2607 | nlsize += nla_total_size(payload: 4); |
2608 | |
2609 | if (attrs & SEG6_F_ATTR(SEG6_LOCAL_BPF)) |
2610 | nlsize += nla_total_size(payload: sizeof(struct nlattr)) + |
2611 | nla_total_size(MAX_PROG_NAME) + |
2612 | nla_total_size(payload: 4); |
2613 | |
2614 | if (attrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE)) |
2615 | nlsize += nla_total_size(payload: 4); |
2616 | |
2617 | if (attrs & SEG6_F_LOCAL_COUNTERS) |
2618 | nlsize += nla_total_size(payload: 0) + /* nest SEG6_LOCAL_COUNTERS */ |
2619 | /* SEG6_LOCAL_CNT_PACKETS */ |
2620 | nla_total_size_64bit(payload: sizeof(__u64)) + |
2621 | /* SEG6_LOCAL_CNT_BYTES */ |
2622 | nla_total_size_64bit(payload: sizeof(__u64)) + |
2623 | /* SEG6_LOCAL_CNT_ERRORS */ |
2624 | nla_total_size_64bit(payload: sizeof(__u64)); |
2625 | |
2626 | if (attrs & SEG6_F_ATTR(SEG6_LOCAL_FLAVORS)) |
2627 | nlsize += encap_size_flavors(slwt); |
2628 | |
2629 | return nlsize; |
2630 | } |
2631 | |
2632 | static int seg6_local_cmp_encap(struct lwtunnel_state *a, |
2633 | struct lwtunnel_state *b) |
2634 | { |
2635 | struct seg6_local_lwt *slwt_a, *slwt_b; |
2636 | struct seg6_action_param *param; |
2637 | unsigned long attrs_a, attrs_b; |
2638 | int i; |
2639 | |
2640 | slwt_a = seg6_local_lwtunnel(lwt: a); |
2641 | slwt_b = seg6_local_lwtunnel(lwt: b); |
2642 | |
2643 | if (slwt_a->action != slwt_b->action) |
2644 | return 1; |
2645 | |
2646 | attrs_a = slwt_a->desc->attrs | slwt_a->parsed_optattrs; |
2647 | attrs_b = slwt_b->desc->attrs | slwt_b->parsed_optattrs; |
2648 | |
2649 | if (attrs_a != attrs_b) |
2650 | return 1; |
2651 | |
2652 | for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { |
2653 | if (attrs_a & SEG6_F_ATTR(i)) { |
2654 | param = &seg6_action_params[i]; |
2655 | if (param->cmp(slwt_a, slwt_b)) |
2656 | return 1; |
2657 | } |
2658 | } |
2659 | |
2660 | return 0; |
2661 | } |
2662 | |
2663 | static const struct lwtunnel_encap_ops seg6_local_ops = { |
2664 | .build_state = seg6_local_build_state, |
2665 | .destroy_state = seg6_local_destroy_state, |
2666 | .input = seg6_local_input, |
2667 | .fill_encap = seg6_local_fill_encap, |
2668 | .get_encap_size = seg6_local_get_encap_size, |
2669 | .cmp_encap = seg6_local_cmp_encap, |
2670 | .owner = THIS_MODULE, |
2671 | }; |
2672 | |
2673 | int __init seg6_local_init(void) |
2674 | { |
2675 | /* If the max total number of defined attributes is reached, then your |
2676 | * kernel build stops here. |
2677 | * |
2678 | * This check is required to avoid arithmetic overflows when processing |
2679 | * behavior attributes and the maximum number of defined attributes |
2680 | * exceeds the allowed value. |
2681 | */ |
2682 | BUILD_BUG_ON(SEG6_LOCAL_MAX + 1 > BITS_PER_TYPE(unsigned long)); |
2683 | |
2684 | /* Check whether the number of defined flavors exceeds the maximum |
2685 | * allowed value. |
2686 | */ |
2687 | BUILD_BUG_ON(SEG6_LOCAL_FLV_OP_MAX + 1 > BITS_PER_TYPE(__u32)); |
2688 | |
2689 | /* If the default NEXT-C-SID Locator-Block/Node Function lengths (in |
2690 | * bits) have been changed with invalid values, kernel build stops |
2691 | * here. |
2692 | */ |
2693 | BUILD_BUG_ON(next_csid_chk_cntr_bits(SEG6_LOCAL_LCBLOCK_DBITS, |
2694 | SEG6_LOCAL_LCNODE_FN_DBITS)); |
2695 | BUILD_BUG_ON(next_csid_chk_lcblock_bits(SEG6_LOCAL_LCBLOCK_DBITS)); |
2696 | BUILD_BUG_ON(next_csid_chk_lcnode_fn_bits(SEG6_LOCAL_LCNODE_FN_DBITS)); |
2697 | |
2698 | /* To be memory efficient, we use 'u8' to represent the different |
2699 | * actions related to RFC8986 flavors. If the kernel build stops here, |
2700 | * it means that it is not possible to correctly encode these actions |
2701 | * with the data type chosen for the action table. |
2702 | */ |
2703 | BUILD_BUG_ON(SEG6_LOCAL_FLV_ACT_MAX > (typeof(flv8986_act_tbl[0]))~0U); |
2704 | |
2705 | return lwtunnel_encap_add_ops(op: &seg6_local_ops, |
2706 | num: LWTUNNEL_ENCAP_SEG6_LOCAL); |
2707 | } |
2708 | |
2709 | void seg6_local_exit(void) |
2710 | { |
2711 | lwtunnel_encap_del_ops(op: &seg6_local_ops, num: LWTUNNEL_ENCAP_SEG6_LOCAL); |
2712 | } |
2713 | |