projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'mlx5-updates-2023-06-09' of git://git.kernel.org/pub/scm/linux/kernel...
[linux-2.6-block.git] / net / ipv6 / seg6_local.c
CommitLineData
2874c5fd 1// SPDX-License-Identifier: GPL-2.0-or-later
d1df6fd8
DL		 2/*
3 * SR-IPv6 implementation
4 *
004d4b27 5 * Authors:
d1df6fd8 6 * David Lebrun <david.lebrun@uclouvain.be>
004d4b27 7 * eBPF support: Mathieu Xhonneux <m.xhonneux@gmail.com>
d1df6fd8
DL		 8 */
9
b6459415 10#include <linux/filter.h>
d1df6fd8
DL		 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>
62ebaeae 27#include <net/ip_tunnels.h>
d1df6fd8
DL		 28#ifdef CONFIG_IPV6_SEG6_HMAC
29#include <net/seg6_hmac.h>
30#endif
1c1e761e 31#include <net/seg6_local.h>
891ef8dd 32#include <linux/etherdevice.h>
004d4b27 33#include <linux/bpf.h>
7a3f5b0d 34#include <linux/netfilter.h>
d1df6fd8 35
300a0fd8
AM		 36#define SEG6_F_ATTR(i) BIT(i)
37
d1df6fd8
DL		 38struct seg6_local_lwt;
39
cfdf64a0
AM		 40/* callbacks used for customizing the creation and destruction of a behavior */
41struct 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
d1df6fd8
DL		 47struct seg6_action_desc {
48 int action;
49 unsigned long attrs;
0a3021f1
AM		 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
d1df6fd8
DL		 65 int (*input)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
66 int static_headroom;
cfdf64a0
AM		 67
68 struct seg6_local_lwtunnel_ops slwt_ops;
d1df6fd8
DL		 69};
70
004d4b27
MX		 71struct bpf_lwt_prog {
72 struct bpf_prog *prog;
73 char *name;
74};
75
848f3c0d
AM		 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
bdf3c0b9
AM		 112#define SEG6_F_LOCAL_FLV_OP(flvname) BIT(SEG6_LOCAL_FLV_OP_##flvname)
113#define SEG6_F_LOCAL_FLV_PSP SEG6_F_LOCAL_FLV_OP(PSP)
114
115/* Supported RFC8986 Flavor operations are reported in this bitmask */
116#define SEG6_LOCAL_FLV8986_SUPP_OPS SEG6_F_LOCAL_FLV_PSP
117
848f3c0d 118/* Supported Flavor operations are reported in this bitmask */
bdf3c0b9
AM		 119#define SEG6_LOCAL_FLV_SUPP_OPS (SEG6_F_LOCAL_FLV_OP(NEXT_CSID) | \
120 SEG6_LOCAL_FLV8986_SUPP_OPS)
848f3c0d
AM		 121
122struct seg6_flavors_info {
123 /* Flavor operations */
124 __u32 flv_ops;
125
126 /* Locator-Block length, expressed in bits */
127 __u8 lcblock_bits;
128 /* Locator-Node Function length, expressed in bits*/
129 __u8 lcnode_func_bits;
130};
131
664d6f86
AM		 132enum seg6_end_dt_mode {
133 DT_INVALID_MODE = -EINVAL,
134 DT_LEGACY_MODE = 0,
135 DT_VRF_MODE = 1,
136};
137
138struct seg6_end_dt_info {
139 enum seg6_end_dt_mode mode;
140
141 struct net *net;
142 /* VRF device associated to the routing table used by the SRv6
143 * End.DT4/DT6 behavior for routing IPv4/IPv6 packets.
144 */
145 int vrf_ifindex;
146 int vrf_table;
147
8b532109
AM		 148 /* tunneled packet family (IPv4 or IPv6).
149 * Protocol and header length are inferred from family.
150 */
664d6f86 151 u16 family;
664d6f86
AM		 152};
153
94604548
AM		 154struct pcpu_seg6_local_counters {
155 u64_stats_t packets;
156 u64_stats_t bytes;
157 u64_stats_t errors;
158
159 struct u64_stats_sync syncp;
160};
161
162/* This struct groups all the SRv6 Behavior counters supported so far.
163 *
164 * put_nla_counters() makes use of this data structure to collect all counter
165 * values after the per-CPU counter evaluation has been performed.
166 * Finally, each counter value (in seg6_local_counters) is stored in the
167 * corresponding netlink attribute and sent to user space.
168 *
169 * NB: we don't want to expose this structure to user space!
170 */
171struct seg6_local_counters {
172 __u64 packets;
173 __u64 bytes;
174 __u64 errors;
175};
176
177#define seg6_local_alloc_pcpu_counters(__gfp) \
178 __netdev_alloc_pcpu_stats(struct pcpu_seg6_local_counters, \
179 ((__gfp) | __GFP_ZERO))
180
181#define SEG6_F_LOCAL_COUNTERS SEG6_F_ATTR(SEG6_LOCAL_COUNTERS)
182
d1df6fd8
DL		 183struct seg6_local_lwt {
184 int action;
185 struct ipv6_sr_hdr *srh;
186 int table;
187 struct in_addr nh4;
188 struct in6_addr nh6;
189 int iif;
190 int oif;
004d4b27 191 struct bpf_lwt_prog bpf;
664d6f86
AM		 192#ifdef CONFIG_NET_L3_MASTER_DEV
193 struct seg6_end_dt_info dt_info;
194#endif
848f3c0d
AM		 195 struct seg6_flavors_info flv_info;
196
94604548 197 struct pcpu_seg6_local_counters __percpu *pcpu_counters;
d1df6fd8
DL		 198
199 int headroom;
200 struct seg6_action_desc *desc;
0a3021f1
AM		 201 /* unlike the required attrs, we have to track the optional attributes
202 * that have been effectively parsed.
203 */
204 unsigned long parsed_optattrs;
d1df6fd8
DL		 205};
206
207static struct seg6_local_lwt *seg6_local_lwtunnel(struct lwtunnel_state *lwt)
208{
209 return (struct seg6_local_lwt *)lwt->data;
210}
211
140f04c3
DL		 212static struct ipv6_sr_hdr *get_and_validate_srh(struct sk_buff *skb)
213{
214 struct ipv6_sr_hdr *srh;
215
fa55a7d7 216 srh = seg6_get_srh(skb, IP6_FH_F_SKIP_RH);
140f04c3
DL		 217 if (!srh)
218 return NULL;
219
140f04c3
DL		 220#ifdef CONFIG_IPV6_SEG6_HMAC
221 if (!seg6_hmac_validate_skb(skb))
222 return NULL;
223#endif
224
225 return srh;
226}
227
d7a669dd
DL		 228static bool decap_and_validate(struct sk_buff *skb, int proto)
229{
230 struct ipv6_sr_hdr *srh;
231 unsigned int off = 0;
232
fa55a7d7 233 srh = seg6_get_srh(skb, 0);
d7a669dd
DL		 234 if (srh && srh->segments_left > 0)
235 return false;
236
237#ifdef CONFIG_IPV6_SEG6_HMAC
238 if (srh && !seg6_hmac_validate_skb(skb))
239 return false;
240#endif
241
242 if (ipv6_find_hdr(skb, &off, proto, NULL, NULL) < 0)
243 return false;
244
245 if (!pskb_pull(skb, off))
246 return false;
247
248 skb_postpull_rcsum(skb, skb_network_header(skb), off);
249
250 skb_reset_network_header(skb);
251 skb_reset_transport_header(skb);
62ebaeae
YT		 252 if (iptunnel_pull_offloads(skb))
253 return false;
d7a669dd
DL		 254
255 return true;
256}
257
258static void advance_nextseg(struct ipv6_sr_hdr *srh, struct in6_addr *daddr)
259{
260 struct in6_addr *addr;
261
262 srh->segments_left--;
263 addr = srh->segments + srh->segments_left;
264 *daddr = *addr;
265}
266
fd1fef0c
AM		 267static int
268seg6_lookup_any_nexthop(struct sk_buff *skb, struct in6_addr *nhaddr,
269 u32 tbl_id, bool local_delivery)
d7a669dd
DL		 270{
271 struct net *net = dev_net(skb->dev);
272 struct ipv6hdr *hdr = ipv6_hdr(skb);
273 int flags = RT6_LOOKUP_F_HAS_SADDR;
274 struct dst_entry *dst = NULL;
275 struct rt6_info *rt;
276 struct flowi6 fl6;
fd1fef0c 277 int dev_flags = 0;
d7a669dd 278
a3bd2102 279 memset(&fl6, 0, sizeof(fl6));
d7a669dd
DL		 280 fl6.flowi6_iif = skb->dev->ifindex;
281 fl6.daddr = nhaddr ? *nhaddr : hdr->daddr;
282 fl6.saddr = hdr->saddr;
283 fl6.flowlabel = ip6_flowinfo(hdr);
284 fl6.flowi6_mark = skb->mark;
285 fl6.flowi6_proto = hdr->nexthdr;
286
287 if (nhaddr)
288 fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH;
289
290 if (!tbl_id) {
b75cc8f9 291 dst = ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags);
d7a669dd
DL		 292 } else {
293 struct fib6_table *table;
294
295 table = fib6_get_table(net, tbl_id);
296 if (!table)
297 goto out;
298
b75cc8f9 299 rt = ip6_pol_route(net, table, 0, &fl6, skb, flags);
d7a669dd
DL		 300 dst = &rt->dst;
301 }
302
fd1fef0c
AM		 303 /* we want to discard traffic destined for local packet processing,
304 * if @local_delivery is set to false.
305 */
306 if (!local_delivery)
307 dev_flags |= IFF_LOOPBACK;
308
309 if (dst && (dst->dev->flags & dev_flags) && !dst->error) {
d7a669dd
DL		 310 dst_release(dst);
311 dst = NULL;
312 }
313
314out:
315 if (!dst) {
316 rt = net->ipv6.ip6_blk_hole_entry;
317 dst = &rt->dst;
318 dst_hold(dst);
319 }
320
321 skb_dst_drop(skb);
322 skb_dst_set(skb, dst);
1c1e761e 323 return dst->error;
d7a669dd
DL		 324}
325
fd1fef0c
AM		 326int seg6_lookup_nexthop(struct sk_buff *skb,
327 struct in6_addr *nhaddr, u32 tbl_id)
328{
329 return seg6_lookup_any_nexthop(skb, nhaddr, tbl_id, false);
330}
331
848f3c0d
AM		 332static __u8 seg6_flv_lcblock_octects(const struct seg6_flavors_info *finfo)
333{
334 return finfo->lcblock_bits >> 3;
335}
336
337static __u8 seg6_flv_lcnode_func_octects(const struct seg6_flavors_info *finfo)
338{
339 return finfo->lcnode_func_bits >> 3;
340}
341
342static bool seg6_next_csid_is_arg_zero(const struct in6_addr *addr,
343 const struct seg6_flavors_info *finfo)
344{
345 __u8 fnc_octects = seg6_flv_lcnode_func_octects(finfo);
346 __u8 blk_octects = seg6_flv_lcblock_octects(finfo);
347 __u8 arg_octects;
348 int i;
349
350 arg_octects = 16 - blk_octects - fnc_octects;
351 for (i = 0; i < arg_octects; ++i) {
352 if (addr->s6_addr[blk_octects + fnc_octects + i] != 0x00)
353 return false;
354 }
355
356 return true;
357}
358
359/* assume that DA.Argument length > 0 */
360static void seg6_next_csid_advance_arg(struct in6_addr *addr,
361 const struct seg6_flavors_info *finfo)
362{
363 __u8 fnc_octects = seg6_flv_lcnode_func_octects(finfo);
364 __u8 blk_octects = seg6_flv_lcblock_octects(finfo);
365
366 /* advance DA.Argument */
367 memmove(&addr->s6_addr[blk_octects],
368 &addr->s6_addr[blk_octects + fnc_octects],
369 16 - blk_octects - fnc_octects);
370
371 memset(&addr->s6_addr[16 - fnc_octects], 0x00, fnc_octects);
372}
373
525c65ff
AM		 374static int input_action_end_finish(struct sk_buff *skb,
375 struct seg6_local_lwt *slwt)
376{
377 seg6_lookup_nexthop(skb, NULL, 0);
378
379 return dst_input(skb);
380}
381
848f3c0d
AM		 382static int input_action_end_core(struct sk_buff *skb,
383 struct seg6_local_lwt *slwt)
140f04c3
DL		 384{
385 struct ipv6_sr_hdr *srh;
140f04c3
DL		 386
387 srh = get_and_validate_srh(skb);
388 if (!srh)
389 goto drop;
390
d7a669dd 391 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
140f04c3 392
525c65ff 393 return input_action_end_finish(skb, slwt);
140f04c3
DL		 394
395drop:
396 kfree_skb(skb);
397 return -EINVAL;
398}
399
848f3c0d
AM		 400static int end_next_csid_core(struct sk_buff *skb, struct seg6_local_lwt *slwt)
401{
402 const struct seg6_flavors_info *finfo = &slwt->flv_info;
403 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr;
404
405 if (seg6_next_csid_is_arg_zero(daddr, finfo))
406 return input_action_end_core(skb, slwt);
407
408 /* update DA */
409 seg6_next_csid_advance_arg(daddr, finfo);
410
525c65ff 411 return input_action_end_finish(skb, slwt);
848f3c0d
AM		 412}
413
414static bool seg6_next_csid_enabled(__u32 fops)
415{
416 return fops & BIT(SEG6_LOCAL_FLV_OP_NEXT_CSID);
417}
418
bdf3c0b9
AM		 419/* We describe the packet state in relation to the absence/presence of the SRH
420 * and the Segment Left (SL) field.
421 * For our purposes, it is not necessary to record the exact value of the SL
422 * when the SID List consists of two or more segments.
423 */
424enum seg6_local_pktinfo {
425 /* the order really matters! */
426 SEG6_LOCAL_PKTINFO_NOHDR = 0,
427 SEG6_LOCAL_PKTINFO_SL_ZERO,
428 SEG6_LOCAL_PKTINFO_SL_ONE,
429 SEG6_LOCAL_PKTINFO_SL_MORE,
430 __SEG6_LOCAL_PKTINFO_MAX,
431};
432
433#define SEG6_LOCAL_PKTINFO_MAX (__SEG6_LOCAL_PKTINFO_MAX - 1)
434
435static enum seg6_local_pktinfo seg6_get_srh_pktinfo(struct ipv6_sr_hdr *srh)
436{
437 __u8 sgl;
438
439 if (!srh)
440 return SEG6_LOCAL_PKTINFO_NOHDR;
441
442 sgl = srh->segments_left;
443 if (sgl < 2)
444 return SEG6_LOCAL_PKTINFO_SL_ZERO + sgl;
445
446 return SEG6_LOCAL_PKTINFO_SL_MORE;
447}
448
449enum seg6_local_flv_action {
450 SEG6_LOCAL_FLV_ACT_UNSPEC = 0,
451 SEG6_LOCAL_FLV_ACT_END,
452 SEG6_LOCAL_FLV_ACT_PSP,
453 SEG6_LOCAL_FLV_ACT_USP,
454 SEG6_LOCAL_FLV_ACT_USD,
455 __SEG6_LOCAL_FLV_ACT_MAX
456};
457
458#define SEG6_LOCAL_FLV_ACT_MAX (__SEG6_LOCAL_FLV_ACT_MAX - 1)
459
460/* The action table for RFC8986 flavors (see the flv8986_act_tbl below)
461 * contains the actions (i.e. processing operations) to be applied on packets
462 * when flavors are configured for an End* behavior.
463 * By combining the pkinfo data and from the flavors mask, the macro
464 * computes the index used to access the elements (actions) stored in the
465 * action table. The index is structured as follows:
466 *
467 * index
468 * _______________/\________________
469 * / \
470 * +----------------+----------------+
471 * | pf | afm |
472 * +----------------+----------------+
473 * ph-1 ... p1 p0 fk-1 ... f1 f0
474 * MSB LSB
475 *
476 * where:
477 * - 'afm' (adjusted flavor mask) is the mask containing a combination of the
478 * RFC8986 flavors currently supported. 'afm' corresponds to the @fm
479 * argument of the macro whose value is righ-shifted by 1 bit. By doing so,
480 * we discard the SEG6_LOCAL_FLV_OP_UNSPEC flag (bit 0 in @fm) which is
481 * never used here;
482 * - 'pf' encodes the packet info (pktinfo) regarding the presence/absence of
483 * the SRH, SL = 0, etc. 'pf' is set with the value of @pf provided as
484 * argument to the macro.
485 */
486#define flv8986_act_tbl_idx(pf, fm) \
487 ((((pf) << bits_per(SEG6_LOCAL_FLV8986_SUPP_OPS)) | \
488 ((fm) & SEG6_LOCAL_FLV8986_SUPP_OPS)) >> SEG6_LOCAL_FLV_OP_PSP)
489
490/* We compute the size of the action table by considering the RFC8986 flavors
491 * actually supported by the kernel. In this way, the size is automatically
492 * adjusted when new flavors are supported.
493 */
494#define FLV8986_ACT_TBL_SIZE \
495 roundup_pow_of_two(flv8986_act_tbl_idx(SEG6_LOCAL_PKTINFO_MAX, \
496 SEG6_LOCAL_FLV8986_SUPP_OPS))
497
498/* tbl_cfg(act, pf, fm) macro is used to easily configure the action
499 * table; it accepts 3 arguments:
500 * i) @act, the suffix from SEG6_LOCAL_FLV_ACT_{act} representing
501 * the action that should be applied on the packet;
502 * ii) @pf, the suffix from SEG6_LOCAL_PKTINFO_{pf} reporting the packet
503 * info about the lack/presence of SRH, SRH with SL = 0, etc;
504 * iii) @fm, the mask of flavors.
505 */
506#define tbl_cfg(act, pf, fm) \
507 [flv8986_act_tbl_idx(SEG6_LOCAL_PKTINFO_##pf, \
508 (fm))] = SEG6_LOCAL_FLV_ACT_##act
509
510/* shorthand for improving readability */
511#define F_PSP SEG6_F_LOCAL_FLV_PSP
512
513/* The table contains, for each combination of the pktinfo data and
514 * flavors, the action that should be taken on a packet (e.g.
515 * "standard" Endpoint processing, Penultimate Segment Pop, etc).
516 *
517 * By default, table entries not explicitly configured are initialized with the
518 * SEG6_LOCAL_FLV_ACT_UNSPEC action, which generally has the effect of
519 * discarding the processed packet.
520 */
521static const u8 flv8986_act_tbl[FLV8986_ACT_TBL_SIZE] = {
522 /* PSP variant for packet where SRH with SL = 1 */
523 tbl_cfg(PSP, SL_ONE, F_PSP),
524 /* End for packet where the SRH with SL > 1*/
525 tbl_cfg(END, SL_MORE, F_PSP),
526};
527
528#undef F_PSP
529#undef tbl_cfg
530
531/* For each flavor defined in RFC8986 (or a combination of them) an action is
532 * performed on the packet. The specific action depends on:
533 * - info extracted from the packet (i.e. pktinfo data) regarding the
534 * lack/presence of the SRH, and if the SRH is available, on the value of
535 * Segment Left field;
536 * - the mask of flavors configured for the specific SRv6 End* behavior.
537 *
538 * The function combines both the pkinfo and the flavors mask to evaluate the
539 * corresponding action to be taken on the packet.
540 */
541static enum seg6_local_flv_action
542seg6_local_flv8986_act_lookup(enum seg6_local_pktinfo pinfo, __u32 flvmask)
543{
544 unsigned long index;
545
546 /* check if the provided mask of flavors is supported */
547 if (unlikely(flvmask & ~SEG6_LOCAL_FLV8986_SUPP_OPS))
548 return SEG6_LOCAL_FLV_ACT_UNSPEC;
549
550 index = flv8986_act_tbl_idx(pinfo, flvmask);
551 if (unlikely(index >= FLV8986_ACT_TBL_SIZE))
552 return SEG6_LOCAL_FLV_ACT_UNSPEC;
553
554 return flv8986_act_tbl[index];
555}
556
557/* skb->data must be aligned with skb->network_header */
558static bool seg6_pop_srh(struct sk_buff *skb, int srhoff)
559{
560 struct ipv6_sr_hdr *srh;
561 struct ipv6hdr *iph;
562 __u8 srh_nexthdr;
563 int thoff = -1;
564 int srhlen;
565 int nhlen;
566
567 if (unlikely(srhoff < sizeof(*iph) ||
568 !pskb_may_pull(skb, srhoff + sizeof(*srh))))
569 return false;
570
571 srh = (struct ipv6_sr_hdr *)(skb->data + srhoff);
572 srhlen = ipv6_optlen(srh);
573
574 /* we are about to mangle the pkt, let's check if we can write on it */
575 if (unlikely(skb_ensure_writable(skb, srhoff + srhlen)))
576 return false;
577
578 /* skb_ensure_writable() may change skb pointers; evaluate srh again */
579 srh = (struct ipv6_sr_hdr *)(skb->data + srhoff);
580 srh_nexthdr = srh->nexthdr;
581
582 if (unlikely(!skb_transport_header_was_set(skb)))
583 goto pull;
584
585 nhlen = skb_network_header_len(skb);
586 /* we have to deal with the transport header: it could be set before
587 * the SRH, after the SRH, or within it (which is considered wrong,
588 * however).
589 */
590 if (likely(nhlen <= srhoff))
591 thoff = nhlen;
592 else if (nhlen >= srhoff + srhlen)
593 /* transport_header is set after the SRH */
594 thoff = nhlen - srhlen;
595 else
596 /* transport_header falls inside the SRH; hence, we can't
597 * restore the transport_header pointer properly after
598 * SRH removing operation.
599 */
600 return false;
601pull:
602 /* we need to pop the SRH:
603 * 1) first of all, we pull out everything from IPv6 header up to SRH
604 * (included) evaluating also the rcsum;
605 * 2) we overwrite (and then remove) the SRH by properly moving the
606 * IPv6 along with any extension header that precedes the SRH;
607 * 3) At the end, we push back the pulled headers (except for SRH,
608 * obviously).
609 */
610 skb_pull_rcsum(skb, srhoff + srhlen);
611 memmove(skb_network_header(skb) + srhlen, skb_network_header(skb),
612 srhoff);
613 skb_push(skb, srhoff);
614
615 skb_reset_network_header(skb);
616 skb_mac_header_rebuild(skb);
617 if (likely(thoff >= 0))
618 skb_set_transport_header(skb, thoff);
619
620 iph = ipv6_hdr(skb);
621 if (iph->nexthdr == NEXTHDR_ROUTING) {
622 iph->nexthdr = srh_nexthdr;
623 } else {
624 /* we must look for the extension header (EXTH, for short) that
625 * immediately precedes the SRH we have just removed.
626 * Then, we update the value of the EXTH nexthdr with the one
627 * contained in the SRH nexthdr.
628 */
629 unsigned int off = sizeof(*iph);
630 struct ipv6_opt_hdr *hp, _hdr;
631 __u8 nexthdr = iph->nexthdr;
632
633 for (;;) {
634 if (unlikely(!ipv6_ext_hdr(nexthdr) ||
635 nexthdr == NEXTHDR_NONE))
636 return false;
637
638 hp = skb_header_pointer(skb, off, sizeof(_hdr), &_hdr);
639 if (unlikely(!hp))
640 return false;
641
642 if (hp->nexthdr == NEXTHDR_ROUTING) {
643 hp->nexthdr = srh_nexthdr;
644 break;
645 }
646
647 switch (nexthdr) {
648 case NEXTHDR_FRAGMENT:
649 fallthrough;
650 case NEXTHDR_AUTH:
651 /* we expect SRH before FRAG and AUTH */
652 return false;
653 default:
654 off += ipv6_optlen(hp);
655 break;
656 }
657
658 nexthdr = hp->nexthdr;
659 }
660 }
661
662 iph->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
663
664 skb_postpush_rcsum(skb, iph, srhoff);
665
666 return true;
667}
668
669/* process the packet on the basis of the RFC8986 flavors set for the given
670 * SRv6 End behavior instance.
671 */
672static int end_flv8986_core(struct sk_buff *skb, struct seg6_local_lwt *slwt)
673{
674 const struct seg6_flavors_info *finfo = &slwt->flv_info;
675 enum seg6_local_flv_action action;
676 enum seg6_local_pktinfo pinfo;
677 struct ipv6_sr_hdr *srh;
678 __u32 flvmask;
679 int srhoff;
680
681 srh = seg6_get_srh(skb, 0);
682 srhoff = srh ? ((unsigned char *)srh - skb->data) : 0;
683 pinfo = seg6_get_srh_pktinfo(srh);
684#ifdef CONFIG_IPV6_SEG6_HMAC
685 if (srh && !seg6_hmac_validate_skb(skb))
686 goto drop;
687#endif
688 flvmask = finfo->flv_ops;
689 if (unlikely(flvmask & ~SEG6_LOCAL_FLV8986_SUPP_OPS)) {
690 pr_warn_once("seg6local: invalid RFC8986 flavors\n");
691 goto drop;
692 }
693
694 /* retrieve the action triggered by the combination of pktinfo data and
695 * the flavors mask.
696 */
697 action = seg6_local_flv8986_act_lookup(pinfo, flvmask);
698 switch (action) {
699 case SEG6_LOCAL_FLV_ACT_END:
700 /* process the packet as the "standard" End behavior */
701 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
702 break;
703 case SEG6_LOCAL_FLV_ACT_PSP:
704 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
705
706 if (unlikely(!seg6_pop_srh(skb, srhoff)))
707 goto drop;
708 break;
709 case SEG6_LOCAL_FLV_ACT_UNSPEC:
710 fallthrough;
711 default:
712 /* by default, we drop the packet since we could not find a
713 * suitable action.
714 */
715 goto drop;
716 }
717
718 return input_action_end_finish(skb, slwt);
719
720drop:
721 kfree_skb(skb);
722 return -EINVAL;
723}
724
848f3c0d
AM		 725/* regular endpoint function */
726static int input_action_end(struct sk_buff *skb, struct seg6_local_lwt *slwt)
727{
728 const struct seg6_flavors_info *finfo = &slwt->flv_info;
bdf3c0b9 729 __u32 fops = finfo->flv_ops;
848f3c0d 730
bdf3c0b9
AM		 731 if (!fops)
732 return input_action_end_core(skb, slwt);
733
734 /* check for the presence of NEXT-C-SID since it applies first */
735 if (seg6_next_csid_enabled(fops))
848f3c0d
AM		 736 return end_next_csid_core(skb, slwt);
737
bdf3c0b9
AM		 738 /* the specific processing function to be performed on the packet
739 * depends on the combination of flavors defined in RFC8986 and some
740 * information extracted from the packet, e.g. presence/absence of SRH,
741 * Segment Left = 0, etc.
742 */
743 return end_flv8986_core(skb, slwt);
848f3c0d
AM		 744}
745
140f04c3
DL		 746/* regular endpoint, and forward to specified nexthop */
747static int input_action_end_x(struct sk_buff *skb, struct seg6_local_lwt *slwt)
748{
140f04c3 749 struct ipv6_sr_hdr *srh;
140f04c3
DL		 750
751 srh = get_and_validate_srh(skb);
752 if (!srh)
753 goto drop;
754
d7a669dd 755 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
140f04c3 756
1c1e761e 757 seg6_lookup_nexthop(skb, &slwt->nh6, 0);
140f04c3
DL		 758
759 return dst_input(skb);
760
761drop:
762 kfree_skb(skb);
763 return -EINVAL;
764}
765
891ef8dd
DL		 766static int input_action_end_t(struct sk_buff *skb, struct seg6_local_lwt *slwt)
767{
768 struct ipv6_sr_hdr *srh;
769
770 srh = get_and_validate_srh(skb);
771 if (!srh)
772 goto drop;
773
774 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
775
1c1e761e 776 seg6_lookup_nexthop(skb, NULL, slwt->table);
891ef8dd
DL		 777
778 return dst_input(skb);
779
780drop:
781 kfree_skb(skb);
782 return -EINVAL;
783}
784
785/* decapsulate and forward inner L2 frame on specified interface */
786static int input_action_end_dx2(struct sk_buff *skb,
787 struct seg6_local_lwt *slwt)
788{
789 struct net *net = dev_net(skb->dev);
790 struct net_device *odev;
791 struct ethhdr *eth;
792
26776253 793 if (!decap_and_validate(skb, IPPROTO_ETHERNET))
891ef8dd
DL		 794 goto drop;
795
796 if (!pskb_may_pull(skb, ETH_HLEN))
797 goto drop;
798
799 skb_reset_mac_header(skb);
800 eth = (struct ethhdr *)skb->data;
801
802 /* To determine the frame's protocol, we assume it is 802.3. This avoids
803 * a call to eth_type_trans(), which is not really relevant for our
804 * use case.
805 */
806 if (!eth_proto_is_802_3(eth->h_proto))
807 goto drop;
808
809 odev = dev_get_by_index_rcu(net, slwt->oif);
810 if (!odev)
811 goto drop;
812
813 /* As we accept Ethernet frames, make sure the egress device is of
814 * the correct type.
815 */
816 if (odev->type != ARPHRD_ETHER)
817 goto drop;
818
819 if (!(odev->flags & IFF_UP) || !netif_carrier_ok(odev))
820 goto drop;
821
822 skb_orphan(skb);
823
824 if (skb_warn_if_lro(skb))
825 goto drop;
826
827 skb_forward_csum(skb);
828
829 if (skb->len - ETH_HLEN > odev->mtu)
830 goto drop;
831
832 skb->dev = odev;
833 skb->protocol = eth->h_proto;
834
835 return dev_queue_xmit(skb);
836
837drop:
838 kfree_skb(skb);
839 return -EINVAL;
840}
841
7a3f5b0d
RS		 842static int input_action_end_dx6_finish(struct net *net, struct sock *sk,
843 struct sk_buff *skb)
844{
845 struct dst_entry *orig_dst = skb_dst(skb);
846 struct in6_addr *nhaddr = NULL;
847 struct seg6_local_lwt *slwt;
848
849 slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
850
851 /* The inner packet is not associated to any local interface,
852 * so we do not call netif_rx().
853 *
854 * If slwt->nh6 is set to ::, then lookup the nexthop for the
855 * inner packet's DA. Otherwise, use the specified nexthop.
856 */
857 if (!ipv6_addr_any(&slwt->nh6))
858 nhaddr = &slwt->nh6;
859
860 seg6_lookup_nexthop(skb, nhaddr, 0);
861
862 return dst_input(skb);
863}
864
140f04c3
DL		 865/* decapsulate and forward to specified nexthop */
866static int input_action_end_dx6(struct sk_buff *skb,
867 struct seg6_local_lwt *slwt)
868{
140f04c3
DL		 869 /* this function accepts IPv6 encapsulated packets, with either
870 * an SRH with SL=0, or no SRH.
871 */
872
d7a669dd 873 if (!decap_and_validate(skb, IPPROTO_IPV6))
140f04c3 874 goto drop;
140f04c3 875
d7a669dd 876 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
140f04c3
DL		 877 goto drop;
878
c71644d0 879 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
7a3f5b0d 880 nf_reset_ct(skb);
c71644d0 881
7a3f5b0d
RS		 882 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
883 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
884 dev_net(skb->dev), NULL, skb, NULL,
885 skb_dst(skb)->dev, input_action_end_dx6_finish);
140f04c3 886
7a3f5b0d 887 return input_action_end_dx6_finish(dev_net(skb->dev), NULL, skb);
140f04c3
DL		 888drop:
889 kfree_skb(skb);
890 return -EINVAL;
891}
892
7a3f5b0d
RS		 893static int input_action_end_dx4_finish(struct net *net, struct sock *sk,
894 struct sk_buff *skb)
891ef8dd 895{
7a3f5b0d
RS		 896 struct dst_entry *orig_dst = skb_dst(skb);
897 struct seg6_local_lwt *slwt;
891ef8dd
DL		 898 struct iphdr *iph;
899 __be32 nhaddr;
900 int err;
901
7a3f5b0d 902 slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
891ef8dd
DL		 903
904 iph = ip_hdr(skb);
905
906 nhaddr = slwt->nh4.s_addr ?: iph->daddr;
907
908 skb_dst_drop(skb);
909
910 err = ip_route_input(skb, nhaddr, iph->saddr, 0, skb->dev);
7a3f5b0d
RS		 911 if (err) {
912 kfree_skb(skb);
913 return -EINVAL;
914 }
891ef8dd
DL		 915
916 return dst_input(skb);
7a3f5b0d
RS		 917}
918
919static int input_action_end_dx4(struct sk_buff *skb,
920 struct seg6_local_lwt *slwt)
921{
922 if (!decap_and_validate(skb, IPPROTO_IPIP))
923 goto drop;
924
925 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
926 goto drop;
927
928 skb->protocol = htons(ETH_P_IP);
929 skb_set_transport_header(skb, sizeof(struct iphdr));
930 nf_reset_ct(skb);
931
932 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
933 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING,
934 dev_net(skb->dev), NULL, skb, NULL,
935 skb_dst(skb)->dev, input_action_end_dx4_finish);
891ef8dd 936
7a3f5b0d 937 return input_action_end_dx4_finish(dev_net(skb->dev), NULL, skb);
891ef8dd
DL		 938drop:
939 kfree_skb(skb);
940 return -EINVAL;
941}
942
664d6f86
AM		 943#ifdef CONFIG_NET_L3_MASTER_DEV
944static struct net *fib6_config_get_net(const struct fib6_config *fib6_cfg)
945{
946 const struct nl_info *nli = &fib6_cfg->fc_nlinfo;
947
948 return nli->nl_net;
949}
950
951static int __seg6_end_dt_vrf_build(struct seg6_local_lwt *slwt, const void *cfg,
952 u16 family, struct netlink_ext_ack *extack)
953{
954 struct seg6_end_dt_info *info = &slwt->dt_info;
955 int vrf_ifindex;
956 struct net *net;
957
958 net = fib6_config_get_net(cfg);
959
960 /* note that vrf_table was already set by parse_nla_vrftable() */
961 vrf_ifindex = l3mdev_ifindex_lookup_by_table_id(L3MDEV_TYPE_VRF, net,
962 info->vrf_table);
963 if (vrf_ifindex < 0) {
964 if (vrf_ifindex == -EPERM) {
965 NL_SET_ERR_MSG(extack,
966 "Strict mode for VRF is disabled");
967 } else if (vrf_ifindex == -ENODEV) {
968 NL_SET_ERR_MSG(extack,
969 "Table has no associated VRF device");
970 } else {
971 pr_debug("seg6local: SRv6 End.DT* creation error=%d\n",
972 vrf_ifindex);
973 }
974
975 return vrf_ifindex;
976 }
977
978 info->net = net;
979 info->vrf_ifindex = vrf_ifindex;
980
664d6f86
AM		 981 info->family = family;
982 info->mode = DT_VRF_MODE;
983
984 return 0;
985}
986
987/* The SRv6 End.DT4/DT6 behavior extracts the inner (IPv4/IPv6) packet and
988 * routes the IPv4/IPv6 packet by looking at the configured routing table.
989 *
990 * In the SRv6 End.DT4/DT6 use case, we can receive traffic (IPv6+Segment
991 * Routing Header packets) from several interfaces and the outer IPv6
992 * destination address (DA) is used for retrieving the specific instance of the
993 * End.DT4/DT6 behavior that should process the packets.
994 *
995 * However, the inner IPv4/IPv6 packet is not really bound to any receiving
996 * interface and thus the End.DT4/DT6 sets the VRF (associated with the
997 * corresponding routing table) as the *receiving* interface.
998 * In other words, the End.DT4/DT6 processes a packet as if it has been received
999 * directly by the VRF (and not by one of its slave devices, if any).
1000 * In this way, the VRF interface is used for routing the IPv4/IPv6 packet in
1001 * according to the routing table configured by the End.DT4/DT6 instance.
1002 *
1003 * This design allows you to get some interesting features like:
1004 * 1) the statistics on rx packets;
1005 * 2) the possibility to install a packet sniffer on the receiving interface
1006 * (the VRF one) for looking at the incoming packets;
1007 * 3) the possibility to leverage the netfilter prerouting hook for the inner
1008 * IPv4 packet.
1009 *
1010 * This function returns:
1011 * - the sk_buff* when the VRF rcv handler has processed the packet correctly;
1012 * - NULL when the skb is consumed by the VRF rcv handler;
1013 * - a pointer which encodes a negative error number in case of error.
1014 * Note that in this case, the function takes care of freeing the skb.
1015 */
1016static struct sk_buff *end_dt_vrf_rcv(struct sk_buff *skb, u16 family,
1017 struct net_device *dev)
1018{
1019 /* based on l3mdev_ip_rcv; we are only interested in the master */
1020 if (unlikely(!netif_is_l3_master(dev) && !netif_has_l3_rx_handler(dev)))
1021 goto drop;
1022
1023 if (unlikely(!dev->l3mdev_ops->l3mdev_l3_rcv))
1024 goto drop;
1025
1026 /* the decap packet IPv4/IPv6 does not come with any mac header info.
1027 * We must unset the mac header to allow the VRF device to rebuild it,
1028 * just in case there is a sniffer attached on the device.
1029 */
1030 skb_unset_mac_header(skb);
1031
1032 skb = dev->l3mdev_ops->l3mdev_l3_rcv(dev, skb, family);
1033 if (!skb)
1034 /* the skb buffer was consumed by the handler */
1035 return NULL;
1036
1037 /* when a packet is received by a VRF or by one of its slaves, the
1038 * master device reference is set into the skb.
1039 */
1040 if (unlikely(skb->dev != dev || skb->skb_iif != dev->ifindex))
1041 goto drop;
1042
1043 return skb;
1044
1045drop:
1046 kfree_skb(skb);
1047 return ERR_PTR(-EINVAL);
1048}
1049
1050static struct net_device *end_dt_get_vrf_rcu(struct sk_buff *skb,
1051 struct seg6_end_dt_info *info)
1052{
1053 int vrf_ifindex = info->vrf_ifindex;
1054 struct net *net = info->net;
1055
1056 if (unlikely(vrf_ifindex < 0))
1057 goto error;
1058
1059 if (unlikely(!net_eq(dev_net(skb->dev), net)))
1060 goto error;
1061
1062 return dev_get_by_index_rcu(net, vrf_ifindex);
1063
1064error:
1065 return NULL;
1066}
1067
1068static struct sk_buff *end_dt_vrf_core(struct sk_buff *skb,
8b532109 1069 struct seg6_local_lwt *slwt, u16 family)
664d6f86
AM		 1070{
1071 struct seg6_end_dt_info *info = &slwt->dt_info;
1072 struct net_device *vrf;
8b532109
AM		 1073 __be16 protocol;
1074 int hdrlen;
664d6f86
AM		 1075
1076 vrf = end_dt_get_vrf_rcu(skb, info);
1077 if (unlikely(!vrf))
1078 goto drop;
1079
8b532109
AM		 1080 switch (family) {
1081 case AF_INET:
1082 protocol = htons(ETH_P_IP);
1083 hdrlen = sizeof(struct iphdr);
1084 break;
1085 case AF_INET6:
1086 protocol = htons(ETH_P_IPV6);
1087 hdrlen = sizeof(struct ipv6hdr);
1088 break;
1089 case AF_UNSPEC:
1090 fallthrough;
1091 default:
1092 goto drop;
1093 }
1094
1095 if (unlikely(info->family != AF_UNSPEC && info->family != family)) {
1096 pr_warn_once("seg6local: SRv6 End.DT* family mismatch");
1097 goto drop;
1098 }
1099
1100 skb->protocol = protocol;
664d6f86
AM		 1101
1102 skb_dst_drop(skb);
1103
8b532109 1104 skb_set_transport_header(skb, hdrlen);
7a3f5b0d 1105 nf_reset_ct(skb);
664d6f86 1106
8b532109 1107 return end_dt_vrf_rcv(skb, family, vrf);
664d6f86
AM		 1108
1109drop:
1110 kfree_skb(skb);
1111 return ERR_PTR(-EINVAL);
1112}
1113
1114static int input_action_end_dt4(struct sk_buff *skb,
1115 struct seg6_local_lwt *slwt)
1116{
1117 struct iphdr *iph;
1118 int err;
1119
1120 if (!decap_and_validate(skb, IPPROTO_IPIP))
1121 goto drop;
1122
1123 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1124 goto drop;
1125
8b532109 1126 skb = end_dt_vrf_core(skb, slwt, AF_INET);
664d6f86
AM		 1127 if (!skb)
1128 /* packet has been processed and consumed by the VRF */
1129 return 0;
1130
1131 if (IS_ERR(skb))
1132 return PTR_ERR(skb);
1133
1134 iph = ip_hdr(skb);
1135
1136 err = ip_route_input(skb, iph->daddr, iph->saddr, 0, skb->dev);
1137 if (unlikely(err))
1138 goto drop;
1139
1140 return dst_input(skb);
1141
1142drop:
1143 kfree_skb(skb);
1144 return -EINVAL;
1145}
1146
1147static int seg6_end_dt4_build(struct seg6_local_lwt *slwt, const void *cfg,
1148 struct netlink_ext_ack *extack)
1149{
1150 return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET, extack);
1151}
20a081b7
AM		 1152
1153static enum
1154seg6_end_dt_mode seg6_end_dt6_parse_mode(struct seg6_local_lwt *slwt)
1155{
1156 unsigned long parsed_optattrs = slwt->parsed_optattrs;
1157 bool legacy, vrfmode;
1158
300a0fd8
AM		 1159 legacy = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE));
1160 vrfmode = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE));
20a081b7
AM		 1161
1162 if (!(legacy ^ vrfmode))
1163 /* both are absent or present: invalid DT6 mode */
1164 return DT_INVALID_MODE;
1165
1166 return legacy ? DT_LEGACY_MODE : DT_VRF_MODE;
1167}
1168
1169static enum seg6_end_dt_mode seg6_end_dt6_get_mode(struct seg6_local_lwt *slwt)
1170{
1171 struct seg6_end_dt_info *info = &slwt->dt_info;
1172
1173 return info->mode;
1174}
1175
1176static int seg6_end_dt6_build(struct seg6_local_lwt *slwt, const void *cfg,
1177 struct netlink_ext_ack *extack)
1178{
1179 enum seg6_end_dt_mode mode = seg6_end_dt6_parse_mode(slwt);
1180 struct seg6_end_dt_info *info = &slwt->dt_info;
1181
1182 switch (mode) {
1183 case DT_LEGACY_MODE:
1184 info->mode = DT_LEGACY_MODE;
1185 return 0;
1186 case DT_VRF_MODE:
1187 return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET6, extack);
1188 default:
1189 NL_SET_ERR_MSG(extack, "table or vrftable must be specified");
1190 return -EINVAL;
1191 }
1192}
664d6f86
AM		 1193#endif
1194
891ef8dd
DL		 1195static int input_action_end_dt6(struct sk_buff *skb,
1196 struct seg6_local_lwt *slwt)
1197{
1198 if (!decap_and_validate(skb, IPPROTO_IPV6))
1199 goto drop;
1200
1201 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1202 goto drop;
1203
20a081b7
AM		 1204#ifdef CONFIG_NET_L3_MASTER_DEV
1205 if (seg6_end_dt6_get_mode(slwt) == DT_LEGACY_MODE)
1206 goto legacy_mode;
1207
1208 /* DT6_VRF_MODE */
8b532109 1209 skb = end_dt_vrf_core(skb, slwt, AF_INET6);
20a081b7
AM		 1210 if (!skb)
1211 /* packet has been processed and consumed by the VRF */
1212 return 0;
1213
1214 if (IS_ERR(skb))
1215 return PTR_ERR(skb);
1216
1217 /* note: this time we do not need to specify the table because the VRF
1218 * takes care of selecting the correct table.
1219 */
1220 seg6_lookup_any_nexthop(skb, NULL, 0, true);
1221
1222 return dst_input(skb);
1223
1224legacy_mode:
1225#endif
c71644d0
AM		 1226 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
1227
fd1fef0c 1228 seg6_lookup_any_nexthop(skb, NULL, slwt->table, true);
891ef8dd
DL		 1229
1230 return dst_input(skb);
1231
1232drop:
1233 kfree_skb(skb);
1234 return -EINVAL;
1235}
1236
8b532109
AM		 1237#ifdef CONFIG_NET_L3_MASTER_DEV
1238static int seg6_end_dt46_build(struct seg6_local_lwt *slwt, const void *cfg,
1239 struct netlink_ext_ack *extack)
1240{
1241 return __seg6_end_dt_vrf_build(slwt, cfg, AF_UNSPEC, extack);
1242}
1243
1244static int input_action_end_dt46(struct sk_buff *skb,
1245 struct seg6_local_lwt *slwt)
1246{
1247 unsigned int off = 0;
1248 int nexthdr;
1249
1250 nexthdr = ipv6_find_hdr(skb, &off, -1, NULL, NULL);
1251 if (unlikely(nexthdr < 0))
1252 goto drop;
1253
1254 switch (nexthdr) {
1255 case IPPROTO_IPIP:
1256 return input_action_end_dt4(skb, slwt);
1257 case IPPROTO_IPV6:
1258 return input_action_end_dt6(skb, slwt);
1259 }
1260
1261drop:
1262 kfree_skb(skb);
1263 return -EINVAL;
1264}
1265#endif
1266
140f04c3
DL		 1267/* push an SRH on top of the current one */
1268static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1269{
1270 struct ipv6_sr_hdr *srh;
1271 int err = -EINVAL;
1272
1273 srh = get_and_validate_srh(skb);
1274 if (!srh)
1275 goto drop;
1276
1277 err = seg6_do_srh_inline(skb, slwt->srh);
1278 if (err)
1279 goto drop;
1280
140f04c3
DL		 1281 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
1282
1c1e761e 1283 seg6_lookup_nexthop(skb, NULL, 0);
140f04c3
DL		 1284
1285 return dst_input(skb);
1286
1287drop:
1288 kfree_skb(skb);
1289 return err;
1290}
1291
1292/* encapsulate within an outer IPv6 header and a specified SRH */
1293static int input_action_end_b6_encap(struct sk_buff *skb,
1294 struct seg6_local_lwt *slwt)
1295{
1296 struct ipv6_sr_hdr *srh;
140f04c3
DL		 1297 int err = -EINVAL;
1298
1299 srh = get_and_validate_srh(skb);
1300 if (!srh)
1301 goto drop;
1302
d7a669dd 1303 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
140f04c3
DL		 1304
1305 skb_reset_inner_headers(skb);
1306 skb->encapsulation = 1;
1307
32d99d0b 1308 err = seg6_do_srh_encap(skb, slwt->srh, IPPROTO_IPV6);
140f04c3
DL		 1309 if (err)
1310 goto drop;
1311
140f04c3
DL		 1312 skb_set_transport_header(skb, sizeof(struct ipv6hdr));
1313
1c1e761e 1314 seg6_lookup_nexthop(skb, NULL, 0);
140f04c3
DL		 1315
1316 return dst_input(skb);
1317
1318drop:
1319 kfree_skb(skb);
1320 return err;
1321}
1322
fe94cc29
MX		 1323DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states);
1324
486cdf21
MX		 1325bool seg6_bpf_has_valid_srh(struct sk_buff *skb)
1326{
1327 struct seg6_bpf_srh_state *srh_state =
1328 this_cpu_ptr(&seg6_bpf_srh_states);
1329 struct ipv6_sr_hdr *srh = srh_state->srh;
1330
1331 if (unlikely(srh == NULL))
1332 return false;
1333
1334 if (unlikely(!srh_state->valid)) {
1335 if ((srh_state->hdrlen & 7) != 0)
1336 return false;
1337
1338 srh->hdrlen = (u8)(srh_state->hdrlen >> 3);
bb986a50 1339 if (!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3, true))
486cdf21
MX		 1340 return false;
1341
1342 srh_state->valid = true;
1343 }
1344
1345 return true;
1346}
1347
004d4b27
MX		 1348static int input_action_end_bpf(struct sk_buff *skb,
1349 struct seg6_local_lwt *slwt)
1350{
1351 struct seg6_bpf_srh_state *srh_state =
1352 this_cpu_ptr(&seg6_bpf_srh_states);
004d4b27 1353 struct ipv6_sr_hdr *srh;
004d4b27
MX		 1354 int ret;
1355
1356 srh = get_and_validate_srh(skb);
486cdf21
MX		 1357 if (!srh) {
1358 kfree_skb(skb);
1359 return -EINVAL;
1360 }
004d4b27
MX		 1361 advance_nextseg(srh, &ipv6_hdr(skb)->daddr);
1362
1363 /* preempt_disable is needed to protect the per-CPU buffer srh_state,
1364 * which is also accessed by the bpf_lwt_seg6_* helpers
1365 */
1366 preempt_disable();
486cdf21 1367 srh_state->srh = srh;
004d4b27 1368 srh_state->hdrlen = srh->hdrlen << 3;
486cdf21 1369 srh_state->valid = true;
004d4b27
MX		 1370
1371 rcu_read_lock();
1372 bpf_compute_data_pointers(skb);
1373 ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb);
1374 rcu_read_unlock();
1375
004d4b27
MX		 1376 switch (ret) {
1377 case BPF_OK:
1378 case BPF_REDIRECT:
1379 break;
1380 case BPF_DROP:
1381 goto drop;
1382 default:
1383 pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret);
1384 goto drop;
1385 }
1386
486cdf21 1387 if (srh_state->srh && !seg6_bpf_has_valid_srh(skb))
004d4b27
MX		 1388 goto drop;
1389
486cdf21 1390 preempt_enable();
004d4b27
MX		 1391 if (ret != BPF_REDIRECT)
1392 seg6_lookup_nexthop(skb, NULL, 0);
1393
1394 return dst_input(skb);
1395
1396drop:
486cdf21 1397 preempt_enable();
004d4b27
MX		 1398 kfree_skb(skb);
1399 return -EINVAL;
1400}
1401
d1df6fd8
DL		 1402static struct seg6_action_desc seg6_action_table[] = {
1403 {
1404 .action = SEG6_LOCAL_ACTION_END,
1405 .attrs = 0,
848f3c0d
AM		 1406 .optattrs = SEG6_F_LOCAL_COUNTERS |
1407 SEG6_F_ATTR(SEG6_LOCAL_FLAVORS),
140f04c3
DL		 1408 .input = input_action_end,
1409 },
1410 {
1411 .action = SEG6_LOCAL_ACTION_END_X,
300a0fd8 1412 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6),
94604548 1413 .optattrs = SEG6_F_LOCAL_COUNTERS,
140f04c3 1414 .input = input_action_end_x,
d1df6fd8 1415 },
891ef8dd
DL		 1416 {
1417 .action = SEG6_LOCAL_ACTION_END_T,
300a0fd8 1418 .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE),
94604548 1419 .optattrs = SEG6_F_LOCAL_COUNTERS,
891ef8dd
DL		 1420 .input = input_action_end_t,
1421 },
1422 {
1423 .action = SEG6_LOCAL_ACTION_END_DX2,
300a0fd8 1424 .attrs = SEG6_F_ATTR(SEG6_LOCAL_OIF),
94604548 1425 .optattrs = SEG6_F_LOCAL_COUNTERS,
891ef8dd
DL		 1426 .input = input_action_end_dx2,
1427 },
140f04c3
DL		 1428 {
1429 .action = SEG6_LOCAL_ACTION_END_DX6,
300a0fd8 1430 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6),
94604548 1431 .optattrs = SEG6_F_LOCAL_COUNTERS,
140f04c3
DL		 1432 .input = input_action_end_dx6,
1433 },
891ef8dd
DL		 1434 {
1435 .action = SEG6_LOCAL_ACTION_END_DX4,
300a0fd8 1436 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH4),
94604548 1437 .optattrs = SEG6_F_LOCAL_COUNTERS,
891ef8dd
DL		 1438 .input = input_action_end_dx4,
1439 },
664d6f86
AM		 1440 {
1441 .action = SEG6_LOCAL_ACTION_END_DT4,
300a0fd8 1442 .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE),
94604548 1443 .optattrs = SEG6_F_LOCAL_COUNTERS,
664d6f86
AM		 1444#ifdef CONFIG_NET_L3_MASTER_DEV
1445 .input = input_action_end_dt4,
1446 .slwt_ops = {
1447 .build_state = seg6_end_dt4_build,
1448 },
1449#endif
1450 },
891ef8dd
DL		 1451 {
1452 .action = SEG6_LOCAL_ACTION_END_DT6,
20a081b7
AM		 1453#ifdef CONFIG_NET_L3_MASTER_DEV
1454 .attrs = 0,
94604548
AM		 1455 .optattrs = SEG6_F_LOCAL_COUNTERS |
1456 SEG6_F_ATTR(SEG6_LOCAL_TABLE) |
300a0fd8 1457 SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE),
20a081b7
AM		 1458 .slwt_ops = {
1459 .build_state = seg6_end_dt6_build,
1460 },
1461#else
300a0fd8 1462 .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE),
94604548 1463 .optattrs = SEG6_F_LOCAL_COUNTERS,
20a081b7 1464#endif
891ef8dd
DL		 1465 .input = input_action_end_dt6,
1466 },
8b532109
AM		 1467 {
1468 .action = SEG6_LOCAL_ACTION_END_DT46,
1469 .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE),
1470 .optattrs = SEG6_F_LOCAL_COUNTERS,
1471#ifdef CONFIG_NET_L3_MASTER_DEV
1472 .input = input_action_end_dt46,
1473 .slwt_ops = {
1474 .build_state = seg6_end_dt46_build,
1475 },
1476#endif
1477 },
140f04c3
DL		 1478 {
1479 .action = SEG6_LOCAL_ACTION_END_B6,
300a0fd8 1480 .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH),
94604548 1481 .optattrs = SEG6_F_LOCAL_COUNTERS,
140f04c3
DL		 1482 .input = input_action_end_b6,
1483 },
1484 {
1485 .action = SEG6_LOCAL_ACTION_END_B6_ENCAP,
300a0fd8 1486 .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH),
94604548 1487 .optattrs = SEG6_F_LOCAL_COUNTERS,
140f04c3
DL		 1488 .input = input_action_end_b6_encap,
1489 .static_headroom = sizeof(struct ipv6hdr),
004d4b27
MX		 1490 },
1491 {
1492 .action = SEG6_LOCAL_ACTION_END_BPF,
300a0fd8 1493 .attrs = SEG6_F_ATTR(SEG6_LOCAL_BPF),
94604548 1494 .optattrs = SEG6_F_LOCAL_COUNTERS,
004d4b27
MX		 1495 .input = input_action_end_bpf,
1496 },
1497
d1df6fd8
DL		 1498};
1499
1500static struct seg6_action_desc *__get_action_desc(int action)
1501{
1502 struct seg6_action_desc *desc;
1503 int i, count;
1504
709af180 1505 count = ARRAY_SIZE(seg6_action_table);
d1df6fd8
DL		 1506 for (i = 0; i < count; i++) {
1507 desc = &seg6_action_table[i];
1508 if (desc->action == action)
1509 return desc;
1510 }
1511
1512 return NULL;
1513}
1514
94604548
AM		 1515static bool seg6_lwtunnel_counters_enabled(struct seg6_local_lwt *slwt)
1516{
1517 return slwt->parsed_optattrs & SEG6_F_LOCAL_COUNTERS;
1518}
1519
1520static void seg6_local_update_counters(struct seg6_local_lwt *slwt,
1521 unsigned int len, int err)
1522{
1523 struct pcpu_seg6_local_counters *pcounters;
1524
1525 pcounters = this_cpu_ptr(slwt->pcpu_counters);
1526 u64_stats_update_begin(&pcounters->syncp);
1527
1528 if (likely(!err)) {
1529 u64_stats_inc(&pcounters->packets);
1530 u64_stats_add(&pcounters->bytes, len);
1531 } else {
1532 u64_stats_inc(&pcounters->errors);
1533 }
1534
1535 u64_stats_update_end(&pcounters->syncp);
1536}
1537
7a3f5b0d
RS		 1538static int seg6_local_input_core(struct net *net, struct sock *sk,
1539 struct sk_buff *skb)
d1df6fd8
DL		 1540{
1541 struct dst_entry *orig_dst = skb_dst(skb);
1542 struct seg6_action_desc *desc;
1543 struct seg6_local_lwt *slwt;
94604548
AM		 1544 unsigned int len = skb->len;
1545 int rc;
d1df6fd8
DL		 1546
1547 slwt = seg6_local_lwtunnel(orig_dst->lwtstate);
1548 desc = slwt->desc;
1549
94604548
AM		 1550 rc = desc->input(skb, slwt);
1551
1552 if (!seg6_lwtunnel_counters_enabled(slwt))
1553 return rc;
1554
1555 seg6_local_update_counters(slwt, len, rc);
1556
1557 return rc;
d1df6fd8
DL		 1558}
1559
7a3f5b0d
RS		 1560static int seg6_local_input(struct sk_buff *skb)
1561{
1562 if (skb->protocol != htons(ETH_P_IPV6)) {
1563 kfree_skb(skb);
1564 return -EINVAL;
1565 }
1566
1567 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled))
1568 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN,
1569 dev_net(skb->dev), NULL, skb, skb->dev, NULL,
1570 seg6_local_input_core);
1571
1572 return seg6_local_input_core(dev_net(skb->dev), NULL, skb);
1573}
1574
d1df6fd8
DL		 1575static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = {
1576 [SEG6_LOCAL_ACTION] = { .type = NLA_U32 },
1577 [SEG6_LOCAL_SRH] = { .type = NLA_BINARY },
1578 [SEG6_LOCAL_TABLE] = { .type = NLA_U32 },
664d6f86 1579 [SEG6_LOCAL_VRFTABLE] = { .type = NLA_U32 },
d1df6fd8
DL		 1580 [SEG6_LOCAL_NH4] = { .type = NLA_BINARY,
1581 .len = sizeof(struct in_addr) },
1582 [SEG6_LOCAL_NH6] = { .type = NLA_BINARY,
1583 .len = sizeof(struct in6_addr) },
1584 [SEG6_LOCAL_IIF] = { .type = NLA_U32 },
1585 [SEG6_LOCAL_OIF] = { .type = NLA_U32 },
004d4b27 1586 [SEG6_LOCAL_BPF] = { .type = NLA_NESTED },
94604548 1587 [SEG6_LOCAL_COUNTERS] = { .type = NLA_NESTED },
848f3c0d 1588 [SEG6_LOCAL_FLAVORS] = { .type = NLA_NESTED },
d1df6fd8
DL		 1589};
1590
e2a8ecc4
AM		 1591static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt,
1592 struct netlink_ext_ack *extack)
2d9cc60a
DL		 1593{
1594 struct ipv6_sr_hdr *srh;
1595 int len;
1596
1597 srh = nla_data(attrs[SEG6_LOCAL_SRH]);
1598 len = nla_len(attrs[SEG6_LOCAL_SRH]);
1599
1600 /* SRH must contain at least one segment */
1601 if (len < sizeof(*srh) + sizeof(struct in6_addr))
1602 return -EINVAL;
1603
bb986a50 1604 if (!seg6_validate_srh(srh, len, false))
2d9cc60a
DL		 1605 return -EINVAL;
1606
7fa41efa 1607 slwt->srh = kmemdup(srh, len, GFP_KERNEL);
2d9cc60a
DL		 1608 if (!slwt->srh)
1609 return -ENOMEM;
1610
2d9cc60a
DL		 1611 slwt->headroom += len;
1612
1613 return 0;
1614}
1615
1616static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1617{
1618 struct ipv6_sr_hdr *srh;
1619 struct nlattr *nla;
1620 int len;
1621
1622 srh = slwt->srh;
1623 len = (srh->hdrlen + 1) << 3;
1624
1625 nla = nla_reserve(skb, SEG6_LOCAL_SRH, len);
1626 if (!nla)
1627 return -EMSGSIZE;
1628
1629 memcpy(nla_data(nla), srh, len);
1630
1631 return 0;
1632}
1633
1634static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1635{
1636 int len = (a->srh->hdrlen + 1) << 3;
1637
1638 if (len != ((b->srh->hdrlen + 1) << 3))
1639 return 1;
1640
1641 return memcmp(a->srh, b->srh, len);
1642}
1643
964adce5
AM		 1644static void destroy_attr_srh(struct seg6_local_lwt *slwt)
1645{
1646 kfree(slwt->srh);
1647}
1648
e2a8ecc4
AM		 1649static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt,
1650 struct netlink_ext_ack *extack)
2d9cc60a
DL		 1651{
1652 slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]);
1653
1654 return 0;
1655}
1656
1657static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1658{
1659 if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table))
1660 return -EMSGSIZE;
1661
1662 return 0;
1663}
1664
1665static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1666{
1667 if (a->table != b->table)
1668 return 1;
1669
1670 return 0;
1671}
1672
664d6f86
AM		 1673static struct
1674seg6_end_dt_info *seg6_possible_end_dt_info(struct seg6_local_lwt *slwt)
1675{
1676#ifdef CONFIG_NET_L3_MASTER_DEV
1677 return &slwt->dt_info;
1678#else
1679 return ERR_PTR(-EOPNOTSUPP);
1680#endif
1681}
1682
1683static int parse_nla_vrftable(struct nlattr **attrs,
e2a8ecc4
AM		 1684 struct seg6_local_lwt *slwt,
1685 struct netlink_ext_ack *extack)
664d6f86
AM		 1686{
1687 struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt);
1688
1689 if (IS_ERR(info))
1690 return PTR_ERR(info);
1691
1692 info->vrf_table = nla_get_u32(attrs[SEG6_LOCAL_VRFTABLE]);
1693
1694 return 0;
1695}
1696
1697static int put_nla_vrftable(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1698{
1699 struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt);
1700
1701 if (IS_ERR(info))
1702 return PTR_ERR(info);
1703
1704 if (nla_put_u32(skb, SEG6_LOCAL_VRFTABLE, info->vrf_table))
1705 return -EMSGSIZE;
1706
1707 return 0;
1708}
1709
1710static int cmp_nla_vrftable(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1711{
1712 struct seg6_end_dt_info *info_a = seg6_possible_end_dt_info(a);
1713 struct seg6_end_dt_info *info_b = seg6_possible_end_dt_info(b);
1714
1715 if (info_a->vrf_table != info_b->vrf_table)
1716 return 1;
1717
1718 return 0;
1719}
1720
e2a8ecc4
AM		 1721static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt,
1722 struct netlink_ext_ack *extack)
2d9cc60a
DL		 1723{
1724 memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]),
1725 sizeof(struct in_addr));
1726
1727 return 0;
1728}
1729
1730static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1731{
1732 struct nlattr *nla;
1733
1734 nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr));
1735 if (!nla)
1736 return -EMSGSIZE;
1737
1738 memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr));
1739
1740 return 0;
1741}
1742
1743static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1744{
1745 return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr));
1746}
1747
e2a8ecc4
AM		 1748static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt,
1749 struct netlink_ext_ack *extack)
2d9cc60a
DL		 1750{
1751 memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]),
1752 sizeof(struct in6_addr));
1753
1754 return 0;
1755}
1756
1757static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1758{
1759 struct nlattr *nla;
1760
1761 nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr));
1762 if (!nla)
1763 return -EMSGSIZE;
1764
1765 memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr));
1766
1767 return 0;
1768}
1769
1770static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1771{
1772 return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr));
1773}
1774
e2a8ecc4
AM		 1775static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt,
1776 struct netlink_ext_ack *extack)
2d9cc60a
DL		 1777{
1778 slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]);
1779
1780 return 0;
1781}
1782
1783static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1784{
1785 if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif))
1786 return -EMSGSIZE;
1787
1788 return 0;
1789}
1790
1791static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1792{
1793 if (a->iif != b->iif)
1794 return 1;
1795
1796 return 0;
1797}
1798
e2a8ecc4
AM		 1799static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt,
1800 struct netlink_ext_ack *extack)
2d9cc60a
DL		 1801{
1802 slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]);
1803
1804 return 0;
1805}
1806
1807static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1808{
1809 if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif))
1810 return -EMSGSIZE;
1811
1812 return 0;
1813}
1814
1815static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1816{
1817 if (a->oif != b->oif)
1818 return 1;
1819
1820 return 0;
1821}
1822
004d4b27
MX		 1823#define MAX_PROG_NAME 256
1824static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = {
1825 [SEG6_LOCAL_BPF_PROG] = { .type = NLA_U32, },
1826 [SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING,
1827 .len = MAX_PROG_NAME },
1828};
1829
e2a8ecc4
AM		 1830static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt,
1831 struct netlink_ext_ack *extack)
004d4b27
MX		 1832{
1833 struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1];
1834 struct bpf_prog *p;
1835 int ret;
1836 u32 fd;
1837
8cb08174
JB		 1838 ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX,
1839 attrs[SEG6_LOCAL_BPF],
1840 bpf_prog_policy, NULL);
004d4b27
MX		 1841 if (ret < 0)
1842 return ret;
1843
1844 if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME])
1845 return -EINVAL;
1846
1847 slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL);
1848 if (!slwt->bpf.name)
1849 return -ENOMEM;
1850
1851 fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]);
1852 p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL);
1853 if (IS_ERR(p)) {
1854 kfree(slwt->bpf.name);
1855 return PTR_ERR(p);
1856 }
1857
1858 slwt->bpf.prog = p;
1859 return 0;
1860}
1861
1862static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1863{
1864 struct nlattr *nest;
1865
1866 if (!slwt->bpf.prog)
1867 return 0;
1868
ae0be8de 1869 nest = nla_nest_start_noflag(skb, SEG6_LOCAL_BPF);
004d4b27
MX		 1870 if (!nest)
1871 return -EMSGSIZE;
1872
1873 if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id))
1874 return -EMSGSIZE;
1875
1876 if (slwt->bpf.name &&
1877 nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name))
1878 return -EMSGSIZE;
1879
1880 return nla_nest_end(skb, nest);
1881}
1882
1883static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1884{
1885 if (!a->bpf.name && !b->bpf.name)
1886 return 0;
1887
1888 if (!a->bpf.name || !b->bpf.name)
1889 return 1;
1890
1891 return strcmp(a->bpf.name, b->bpf.name);
1892}
1893
964adce5
AM		 1894static void destroy_attr_bpf(struct seg6_local_lwt *slwt)
1895{
1896 kfree(slwt->bpf.name);
1897 if (slwt->bpf.prog)
1898 bpf_prog_put(slwt->bpf.prog);
1899}
1900
94604548
AM		 1901static const struct
1902nla_policy seg6_local_counters_policy[SEG6_LOCAL_CNT_MAX + 1] = {
1903 [SEG6_LOCAL_CNT_PACKETS] = { .type = NLA_U64 },
1904 [SEG6_LOCAL_CNT_BYTES] = { .type = NLA_U64 },
1905 [SEG6_LOCAL_CNT_ERRORS] = { .type = NLA_U64 },
1906};
1907
1908static int parse_nla_counters(struct nlattr **attrs,
e2a8ecc4
AM		 1909 struct seg6_local_lwt *slwt,
1910 struct netlink_ext_ack *extack)
94604548
AM		 1911{
1912 struct pcpu_seg6_local_counters __percpu *pcounters;
1913 struct nlattr *tb[SEG6_LOCAL_CNT_MAX + 1];
1914 int ret;
1915
1916 ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_CNT_MAX,
1917 attrs[SEG6_LOCAL_COUNTERS],
1918 seg6_local_counters_policy, NULL);
1919 if (ret < 0)
1920 return ret;
1921
1922 /* basic support for SRv6 Behavior counters requires at least:
1923 * packets, bytes and errors.
1924 */
1925 if (!tb[SEG6_LOCAL_CNT_PACKETS] || !tb[SEG6_LOCAL_CNT_BYTES] ||
1926 !tb[SEG6_LOCAL_CNT_ERRORS])
1927 return -EINVAL;
1928
1929 /* counters are always zero initialized */
1930 pcounters = seg6_local_alloc_pcpu_counters(GFP_KERNEL);
1931 if (!pcounters)
1932 return -ENOMEM;
1933
1934 slwt->pcpu_counters = pcounters;
1935
1936 return 0;
1937}
1938
1939static int seg6_local_fill_nla_counters(struct sk_buff *skb,
1940 struct seg6_local_counters *counters)
1941{
1942 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_PACKETS, counters->packets,
1943 SEG6_LOCAL_CNT_PAD))
1944 return -EMSGSIZE;
1945
1946 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_BYTES, counters->bytes,
1947 SEG6_LOCAL_CNT_PAD))
1948 return -EMSGSIZE;
1949
1950 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_ERRORS, counters->errors,
1951 SEG6_LOCAL_CNT_PAD))
1952 return -EMSGSIZE;
1953
1954 return 0;
1955}
1956
1957static int put_nla_counters(struct sk_buff *skb, struct seg6_local_lwt *slwt)
1958{
1959 struct seg6_local_counters counters = { 0, 0, 0 };
1960 struct nlattr *nest;
1961 int rc, i;
1962
1963 nest = nla_nest_start(skb, SEG6_LOCAL_COUNTERS);
1964 if (!nest)
1965 return -EMSGSIZE;
1966
1967 for_each_possible_cpu(i) {
1968 struct pcpu_seg6_local_counters *pcounters;
1969 u64 packets, bytes, errors;
1970 unsigned int start;
1971
1972 pcounters = per_cpu_ptr(slwt->pcpu_counters, i);
1973 do {
d120d1a6 1974 start = u64_stats_fetch_begin(&pcounters->syncp);
94604548
AM		 1975
1976 packets = u64_stats_read(&pcounters->packets);
1977 bytes = u64_stats_read(&pcounters->bytes);
1978 errors = u64_stats_read(&pcounters->errors);
1979
d120d1a6 1980 } while (u64_stats_fetch_retry(&pcounters->syncp, start));
94604548
AM		 1981
1982 counters.packets += packets;
1983 counters.bytes += bytes;
1984 counters.errors += errors;
1985 }
1986
1987 rc = seg6_local_fill_nla_counters(skb, &counters);
1988 if (rc < 0) {
1989 nla_nest_cancel(skb, nest);
1990 return rc;
1991 }
1992
1993 return nla_nest_end(skb, nest);
1994}
1995
1996static int cmp_nla_counters(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
1997{
1998 /* a and b are equal if both have pcpu_counters set or not */
1999 return (!!((unsigned long)a->pcpu_counters)) ^
2000 (!!((unsigned long)b->pcpu_counters));
2001}
2002
2003static void destroy_attr_counters(struct seg6_local_lwt *slwt)
2004{
2005 free_percpu(slwt->pcpu_counters);
2006}
2007
848f3c0d
AM		 2008static const
2009struct nla_policy seg6_local_flavors_policy[SEG6_LOCAL_FLV_MAX + 1] = {
2010 [SEG6_LOCAL_FLV_OPERATION] = { .type = NLA_U32 },
2011 [SEG6_LOCAL_FLV_LCBLOCK_BITS] = { .type = NLA_U8 },
2012 [SEG6_LOCAL_FLV_LCNODE_FN_BITS] = { .type = NLA_U8 },
2013};
2014
2015/* check whether the lengths of the Locator-Block and Locator-Node Function
2016 * are compatible with the dimension of a C-SID container.
2017 */
2018static int seg6_chk_next_csid_cfg(__u8 block_len, __u8 func_len)
2019{
2020 /* Locator-Block and Locator-Node Function cannot exceed 128 bits
2021 * (i.e. C-SID container lenghts).
2022 */
2023 if (next_csid_chk_cntr_bits(block_len, func_len))
2024 return -EINVAL;
2025
2026 /* Locator-Block length must be greater than zero and evenly divisible
2027 * by 8. There must be room for a Locator-Node Function, at least.
2028 */
2029 if (next_csid_chk_lcblock_bits(block_len))
2030 return -EINVAL;
2031
2032 /* Locator-Node Function length must be greater than zero and evenly
2033 * divisible by 8. There must be room for the Locator-Block.
2034 */
2035 if (next_csid_chk_lcnode_fn_bits(func_len))
2036 return -EINVAL;
2037
2038 return 0;
2039}
2040
2041static int seg6_parse_nla_next_csid_cfg(struct nlattr **tb,
2042 struct seg6_flavors_info *finfo,
2043 struct netlink_ext_ack *extack)
2044{
2045 __u8 func_len = SEG6_LOCAL_LCNODE_FN_DBITS;
2046 __u8 block_len = SEG6_LOCAL_LCBLOCK_DBITS;
2047 int rc;
2048
2049 if (tb[SEG6_LOCAL_FLV_LCBLOCK_BITS])
2050 block_len = nla_get_u8(tb[SEG6_LOCAL_FLV_LCBLOCK_BITS]);
2051
2052 if (tb[SEG6_LOCAL_FLV_LCNODE_FN_BITS])
2053 func_len = nla_get_u8(tb[SEG6_LOCAL_FLV_LCNODE_FN_BITS]);
2054
2055 rc = seg6_chk_next_csid_cfg(block_len, func_len);
2056 if (rc < 0) {
2057 NL_SET_ERR_MSG(extack,
2058 "Invalid Locator Block/Node Function lengths");
2059 return rc;
2060 }
2061
2062 finfo->lcblock_bits = block_len;
2063 finfo->lcnode_func_bits = func_len;
2064
2065 return 0;
2066}
2067
2068static int parse_nla_flavors(struct nlattr **attrs, struct seg6_local_lwt *slwt,
2069 struct netlink_ext_ack *extack)
2070{
2071 struct seg6_flavors_info *finfo = &slwt->flv_info;
2072 struct nlattr *tb[SEG6_LOCAL_FLV_MAX + 1];
2073 unsigned long fops;
2074 int rc;
2075
2076 rc = nla_parse_nested_deprecated(tb, SEG6_LOCAL_FLV_MAX,
2077 attrs[SEG6_LOCAL_FLAVORS],
2078 seg6_local_flavors_policy, NULL);
2079 if (rc < 0)
2080 return rc;
2081
2082 /* this attribute MUST always be present since it represents the Flavor
2083 * operation(s) to be carried out.
2084 */
2085 if (!tb[SEG6_LOCAL_FLV_OPERATION])
2086 return -EINVAL;
2087
2088 fops = nla_get_u32(tb[SEG6_LOCAL_FLV_OPERATION]);
2089 if (fops & ~SEG6_LOCAL_FLV_SUPP_OPS) {
2090 NL_SET_ERR_MSG(extack, "Unsupported Flavor operation(s)");
2091 return -EOPNOTSUPP;
2092 }
2093
2094 finfo->flv_ops = fops;
2095
2096 if (seg6_next_csid_enabled(fops)) {
2097 /* Locator-Block and Locator-Node Function lengths can be
2098 * provided by the user space. Otherwise, default values are
2099 * applied.
2100 */
2101 rc = seg6_parse_nla_next_csid_cfg(tb, finfo, extack);
2102 if (rc < 0)
2103 return rc;
2104 }
2105
2106 return 0;
2107}
2108
2109static int seg6_fill_nla_next_csid_cfg(struct sk_buff *skb,
2110 struct seg6_flavors_info *finfo)
2111{
2112 if (nla_put_u8(skb, SEG6_LOCAL_FLV_LCBLOCK_BITS, finfo->lcblock_bits))
2113 return -EMSGSIZE;
2114
2115 if (nla_put_u8(skb, SEG6_LOCAL_FLV_LCNODE_FN_BITS,
2116 finfo->lcnode_func_bits))
2117 return -EMSGSIZE;
2118
2119 return 0;
2120}
2121
2122static int put_nla_flavors(struct sk_buff *skb, struct seg6_local_lwt *slwt)
2123{
2124 struct seg6_flavors_info *finfo = &slwt->flv_info;
2125 __u32 fops = finfo->flv_ops;
2126 struct nlattr *nest;
2127 int rc;
2128
2129 nest = nla_nest_start(skb, SEG6_LOCAL_FLAVORS);
2130 if (!nest)
2131 return -EMSGSIZE;
2132
2133 if (nla_put_u32(skb, SEG6_LOCAL_FLV_OPERATION, fops)) {
2134 rc = -EMSGSIZE;
2135 goto err;
2136 }
2137
2138 if (seg6_next_csid_enabled(fops)) {
2139 rc = seg6_fill_nla_next_csid_cfg(skb, finfo);
2140 if (rc < 0)
2141 goto err;
2142 }
2143
2144 return nla_nest_end(skb, nest);
2145
2146err:
2147 nla_nest_cancel(skb, nest);
2148 return rc;
2149}
2150
2151static int seg6_cmp_nla_next_csid_cfg(struct seg6_flavors_info *finfo_a,
2152 struct seg6_flavors_info *finfo_b)
2153{
2154 if (finfo_a->lcblock_bits != finfo_b->lcblock_bits)
2155 return 1;
2156
2157 if (finfo_a->lcnode_func_bits != finfo_b->lcnode_func_bits)
2158 return 1;
2159
2160 return 0;
2161}
2162
2163static int cmp_nla_flavors(struct seg6_local_lwt *a, struct seg6_local_lwt *b)
2164{
2165 struct seg6_flavors_info *finfo_a = &a->flv_info;
2166 struct seg6_flavors_info *finfo_b = &b->flv_info;
2167
2168 if (finfo_a->flv_ops != finfo_b->flv_ops)
2169 return 1;
2170
2171 if (seg6_next_csid_enabled(finfo_a->flv_ops)) {
2172 if (seg6_cmp_nla_next_csid_cfg(finfo_a, finfo_b))
2173 return 1;
2174 }
2175
2176 return 0;
2177}
2178
2179static int encap_size_flavors(struct seg6_local_lwt *slwt)
2180{
2181 struct seg6_flavors_info *finfo = &slwt->flv_info;
2182 int nlsize;
2183
2184 nlsize = nla_total_size(0) + /* nest SEG6_LOCAL_FLAVORS */
2185 nla_total_size(4); /* SEG6_LOCAL_FLV_OPERATION */
2186
2187 if (seg6_next_csid_enabled(finfo->flv_ops))
2188 nlsize += nla_total_size(1) + /* SEG6_LOCAL_FLV_LCBLOCK_BITS */
2189 nla_total_size(1); /* SEG6_LOCAL_FLV_LCNODE_FN_BITS */
2190
2191 return nlsize;
2192}
2193
d1df6fd8 2194struct seg6_action_param {
e2a8ecc4
AM		 2195 int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt,
2196 struct netlink_ext_ack *extack);
d1df6fd8
DL		 2197 int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt);
2198 int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b);
964adce5
AM		 2199
2200 /* optional destroy() callback useful for releasing resources which
2201 * have been previously acquired in the corresponding parse()
2202 * function.
2203 */
2204 void (*destroy)(struct seg6_local_lwt *slwt);
d1df6fd8
DL		 2205};
2206
2207static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = {
2d9cc60a
DL		 2208 [SEG6_LOCAL_SRH] = { .parse = parse_nla_srh,
2209 .put = put_nla_srh,
964adce5
AM		 2210 .cmp = cmp_nla_srh,
2211 .destroy = destroy_attr_srh },
d1df6fd8 2212
2d9cc60a
DL		 2213 [SEG6_LOCAL_TABLE] = { .parse = parse_nla_table,
2214 .put = put_nla_table,
2215 .cmp = cmp_nla_table },
d1df6fd8 2216
2d9cc60a
DL		 2217 [SEG6_LOCAL_NH4] = { .parse = parse_nla_nh4,
2218 .put = put_nla_nh4,
2219 .cmp = cmp_nla_nh4 },
d1df6fd8 2220
2d9cc60a
DL		 2221 [SEG6_LOCAL_NH6] = { .parse = parse_nla_nh6,
2222 .put = put_nla_nh6,
2223 .cmp = cmp_nla_nh6 },
d1df6fd8 2224
2d9cc60a
DL		 2225 [SEG6_LOCAL_IIF] = { .parse = parse_nla_iif,
2226 .put = put_nla_iif,
2227 .cmp = cmp_nla_iif },
d1df6fd8 2228
2d9cc60a
DL		 2229 [SEG6_LOCAL_OIF] = { .parse = parse_nla_oif,
2230 .put = put_nla_oif,
2231 .cmp = cmp_nla_oif },
004d4b27
MX		 2232
2233 [SEG6_LOCAL_BPF] = { .parse = parse_nla_bpf,
2234 .put = put_nla_bpf,
964adce5
AM		 2235 .cmp = cmp_nla_bpf,
2236 .destroy = destroy_attr_bpf },
004d4b27 2237
664d6f86
AM		 2238 [SEG6_LOCAL_VRFTABLE] = { .parse = parse_nla_vrftable,
2239 .put = put_nla_vrftable,
2240 .cmp = cmp_nla_vrftable },
2241
94604548
AM		 2242 [SEG6_LOCAL_COUNTERS] = { .parse = parse_nla_counters,
2243 .put = put_nla_counters,
2244 .cmp = cmp_nla_counters,
2245 .destroy = destroy_attr_counters },
848f3c0d
AM		 2246
2247 [SEG6_LOCAL_FLAVORS] = { .parse = parse_nla_flavors,
2248 .put = put_nla_flavors,
2249 .cmp = cmp_nla_flavors },
d1df6fd8
DL		 2250};
2251
964adce5 2252/* call the destroy() callback (if available) for each set attribute in
0a3021f1
AM		 2253 * @parsed_attrs, starting from the first attribute up to the @max_parsed
2254 * (excluded) attribute.
964adce5 2255 */
0a3021f1
AM		 2256static void __destroy_attrs(unsigned long parsed_attrs, int max_parsed,
2257 struct seg6_local_lwt *slwt)
964adce5 2258{
964adce5
AM		 2259 struct seg6_action_param *param;
2260 int i;
2261
2262 /* Every required seg6local attribute is identified by an ID which is
2263 * encoded as a flag (i.e: 1 << ID) in the 'attrs' bitmask;
2264 *
0a3021f1 2265 * We scan the 'parsed_attrs' bitmask, starting from the first attribute
964adce5
AM		 2266 * up to the @max_parsed (excluded) attribute.
2267 * For each set attribute, we retrieve the corresponding destroy()
2268 * callback. If the callback is not available, then we skip to the next
2269 * attribute; otherwise, we call the destroy() callback.
2270 */
ac0dbed9 2271 for (i = SEG6_LOCAL_SRH; i < max_parsed; ++i) {
300a0fd8 2272 if (!(parsed_attrs & SEG6_F_ATTR(i)))
964adce5
AM		 2273 continue;
2274
2275 param = &seg6_action_params[i];
2276
2277 if (param->destroy)
2278 param->destroy(slwt);
2279 }
2280}
2281
2282/* release all the resources that may have been acquired during parsing
2283 * operations.
2284 */
2285static void destroy_attrs(struct seg6_local_lwt *slwt)
2286{
0a3021f1
AM		 2287 unsigned long attrs = slwt->desc->attrs | slwt->parsed_optattrs;
2288
2289 __destroy_attrs(attrs, SEG6_LOCAL_MAX + 1, slwt);
2290}
2291
2292static int parse_nla_optional_attrs(struct nlattr **attrs,
e2a8ecc4
AM		 2293 struct seg6_local_lwt *slwt,
2294 struct netlink_ext_ack *extack)
0a3021f1
AM		 2295{
2296 struct seg6_action_desc *desc = slwt->desc;
2297 unsigned long parsed_optattrs = 0;
2298 struct seg6_action_param *param;
2299 int err, i;
2300
ac0dbed9 2301 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; ++i) {
300a0fd8 2302 if (!(desc->optattrs & SEG6_F_ATTR(i)) || !attrs[i])
0a3021f1
AM		 2303 continue;
2304
2305 /* once here, the i-th attribute is provided by the
2306 * userspace AND it is identified optional as well.
2307 */
2308 param = &seg6_action_params[i];
2309
e2a8ecc4 2310 err = param->parse(attrs, slwt, extack);
0a3021f1
AM		 2311 if (err < 0)
2312 goto parse_optattrs_err;
2313
2314 /* current attribute has been correctly parsed */
300a0fd8 2315 parsed_optattrs |= SEG6_F_ATTR(i);
0a3021f1
AM		 2316 }
2317
2318 /* store in the tunnel state all the optional attributed successfully
2319 * parsed.
2320 */
2321 slwt->parsed_optattrs = parsed_optattrs;
2322
2323 return 0;
2324
2325parse_optattrs_err:
2326 __destroy_attrs(parsed_optattrs, i, slwt);
2327
2328 return err;
964adce5
AM		 2329}
2330
cfdf64a0
AM		 2331/* call the custom constructor of the behavior during its initialization phase
2332 * and after that all its attributes have been parsed successfully.
2333 */
2334static int
2335seg6_local_lwtunnel_build_state(struct seg6_local_lwt *slwt, const void *cfg,
2336 struct netlink_ext_ack *extack)
2337{
2338 struct seg6_action_desc *desc = slwt->desc;
2339 struct seg6_local_lwtunnel_ops *ops;
2340
2341 ops = &desc->slwt_ops;
2342 if (!ops->build_state)
2343 return 0;
2344
2345 return ops->build_state(slwt, cfg, extack);
2346}
2347
2348/* call the custom destructor of the behavior which is invoked before the
2349 * tunnel is going to be destroyed.
2350 */
2351static void seg6_local_lwtunnel_destroy_state(struct seg6_local_lwt *slwt)
2352{
2353 struct seg6_action_desc *desc = slwt->desc;
2354 struct seg6_local_lwtunnel_ops *ops;
2355
2356 ops = &desc->slwt_ops;
2357 if (!ops->destroy_state)
2358 return;
2359
2360 ops->destroy_state(slwt);
2361}
2362
e2a8ecc4
AM		 2363static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt,
2364 struct netlink_ext_ack *extack)
d1df6fd8
DL		 2365{
2366 struct seg6_action_param *param;
2367 struct seg6_action_desc *desc;
0a3021f1 2368 unsigned long invalid_attrs;
d1df6fd8
DL		 2369 int i, err;
2370
2371 desc = __get_action_desc(slwt->action);
2372 if (!desc)
2373 return -EINVAL;
2374
2375 if (!desc->input)
2376 return -EOPNOTSUPP;
2377
2378 slwt->desc = desc;
2379 slwt->headroom += desc->static_headroom;
2380
0a3021f1
AM		 2381 /* Forcing the desc->optattrs *set* and the desc->attrs *set* to be
2382 * disjoined, this allow us to release acquired resources by optional
2383 * attributes and by required attributes independently from each other
0d770360 2384 * without any interference.
0a3021f1
AM		 2385 * In other terms, we are sure that we do not release some the acquired
2386 * resources twice.
2387 *
2388 * Note that if an attribute is configured both as required and as
2389 * optional, it means that the user has messed something up in the
2390 * seg6_action_table. Therefore, this check is required for SRv6
2391 * behaviors to work properly.
2392 */
2393 invalid_attrs = desc->attrs & desc->optattrs;
2394 if (invalid_attrs) {
2395 WARN_ONCE(1,
2396 "An attribute cannot be both required AND optional");
2397 return -EINVAL;
2398 }
2399
2400 /* parse the required attributes */
ac0dbed9 2401 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) {
300a0fd8 2402 if (desc->attrs & SEG6_F_ATTR(i)) {
d1df6fd8
DL		 2403 if (!attrs[i])
2404 return -EINVAL;
2405
2406 param = &seg6_action_params[i];
2407
e2a8ecc4 2408 err = param->parse(attrs, slwt, extack);
d1df6fd8 2409 if (err < 0)
0a3021f1 2410 goto parse_attrs_err;
d1df6fd8
DL		 2411 }
2412 }
2413
0a3021f1 2414 /* parse the optional attributes, if any */
e2a8ecc4 2415 err = parse_nla_optional_attrs(attrs, slwt, extack);
0a3021f1
AM		 2416 if (err < 0)
2417 goto parse_attrs_err;
2418
d1df6fd8 2419 return 0;
964adce5 2420
0a3021f1 2421parse_attrs_err:
964adce5
AM		 2422 /* release any resource that may have been acquired during the i-1
2423 * parse() operations.
2424 */
0a3021f1 2425 __destroy_attrs(desc->attrs, i, slwt);
964adce5
AM		 2426
2427 return err;
d1df6fd8
DL		 2428}
2429
faee6769
AA		 2430static int seg6_local_build_state(struct net *net, struct nlattr *nla,
2431 unsigned int family, const void *cfg,
2432 struct lwtunnel_state **ts,
d1df6fd8
DL		 2433 struct netlink_ext_ack *extack)
2434{
2435 struct nlattr *tb[SEG6_LOCAL_MAX + 1];
2436 struct lwtunnel_state *newts;
2437 struct seg6_local_lwt *slwt;
2438 int err;
2439
6285217f
DL		 2440 if (family != AF_INET6)
2441 return -EINVAL;
2442
8cb08174
JB		 2443 err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla,
2444 seg6_local_policy, extack);
d1df6fd8
DL		 2445
2446 if (err < 0)
2447 return err;
2448
2449 if (!tb[SEG6_LOCAL_ACTION])
2450 return -EINVAL;
2451
2452 newts = lwtunnel_state_alloc(sizeof(*slwt));
2453 if (!newts)
2454 return -ENOMEM;
2455
2456 slwt = seg6_local_lwtunnel(newts);
2457 slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]);
2458
e2a8ecc4 2459 err = parse_nla_action(tb, slwt, extack);
d1df6fd8
DL		 2460 if (err < 0)
2461 goto out_free;
2462
cfdf64a0
AM		 2463 err = seg6_local_lwtunnel_build_state(slwt, cfg, extack);
2464 if (err < 0)
2465 goto out_destroy_attrs;
2466
d1df6fd8
DL		 2467 newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL;
2468 newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT;
2469 newts->headroom = slwt->headroom;
2470
2471 *ts = newts;
2472
2473 return 0;
2474
cfdf64a0
AM		 2475out_destroy_attrs:
2476 destroy_attrs(slwt);
d1df6fd8 2477out_free:
d1df6fd8
DL		 2478 kfree(newts);
2479 return err;
2480}
2481
2482static void seg6_local_destroy_state(struct lwtunnel_state *lwt)
2483{
2484 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
2485
cfdf64a0
AM		 2486 seg6_local_lwtunnel_destroy_state(slwt);
2487
964adce5 2488 destroy_attrs(slwt);
004d4b27
MX		 2489
2490 return;
d1df6fd8
DL		 2491}
2492
2493static int seg6_local_fill_encap(struct sk_buff *skb,
2494 struct lwtunnel_state *lwt)
2495{
2496 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
2497 struct seg6_action_param *param;
0a3021f1 2498 unsigned long attrs;
d1df6fd8
DL		 2499 int i, err;
2500
2501 if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action))
2502 return -EMSGSIZE;
2503
0a3021f1
AM		 2504 attrs = slwt->desc->attrs | slwt->parsed_optattrs;
2505
ac0dbed9 2506 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) {
300a0fd8 2507 if (attrs & SEG6_F_ATTR(i)) {
d1df6fd8
DL		 2508 param = &seg6_action_params[i];
2509 err = param->put(skb, slwt);
2510 if (err < 0)
2511 return err;
2512 }
2513 }
2514
2515 return 0;
2516}
2517
2518static int seg6_local_get_encap_size(struct lwtunnel_state *lwt)
2519{
2520 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt);
2521 unsigned long attrs;
2522 int nlsize;
2523
2524 nlsize = nla_total_size(4); /* action */
2525
0a3021f1 2526 attrs = slwt->desc->attrs | slwt->parsed_optattrs;
d1df6fd8 2527
300a0fd8 2528 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_SRH))
d1df6fd8
DL		 2529 nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3);
2530
300a0fd8 2531 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE))
d1df6fd8
DL		 2532 nlsize += nla_total_size(4);
2533
300a0fd8 2534 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH4))
d1df6fd8
DL		 2535 nlsize += nla_total_size(4);
2536
300a0fd8 2537 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH6))
d1df6fd8
DL		 2538 nlsize += nla_total_size(16);
2539
300a0fd8 2540 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_IIF))
d1df6fd8
DL		 2541 nlsize += nla_total_size(4);
2542
300a0fd8 2543 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_OIF))
d1df6fd8
DL		 2544 nlsize += nla_total_size(4);
2545
300a0fd8 2546 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_BPF))
004d4b27
MX		 2547 nlsize += nla_total_size(sizeof(struct nlattr)) +
2548 nla_total_size(MAX_PROG_NAME) +
2549 nla_total_size(4);
2550
300a0fd8 2551 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE))
664d6f86
AM		 2552 nlsize += nla_total_size(4);
2553
94604548
AM		 2554 if (attrs & SEG6_F_LOCAL_COUNTERS)
2555 nlsize += nla_total_size(0) + /* nest SEG6_LOCAL_COUNTERS */
2556 /* SEG6_LOCAL_CNT_PACKETS */
2557 nla_total_size_64bit(sizeof(__u64)) +
2558 /* SEG6_LOCAL_CNT_BYTES */
2559 nla_total_size_64bit(sizeof(__u64)) +
2560 /* SEG6_LOCAL_CNT_ERRORS */
2561 nla_total_size_64bit(sizeof(__u64));
2562
848f3c0d
AM		 2563 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_FLAVORS))
2564 nlsize += encap_size_flavors(slwt);
2565
d1df6fd8
DL		 2566 return nlsize;
2567}
2568
2569static int seg6_local_cmp_encap(struct lwtunnel_state *a,
2570 struct lwtunnel_state *b)
2571{
2572 struct seg6_local_lwt *slwt_a, *slwt_b;
2573 struct seg6_action_param *param;
0a3021f1 2574 unsigned long attrs_a, attrs_b;
d1df6fd8
DL		 2575 int i;
2576
2577 slwt_a = seg6_local_lwtunnel(a);
2578 slwt_b = seg6_local_lwtunnel(b);
2579
2580 if (slwt_a->action != slwt_b->action)
2581 return 1;
2582
0a3021f1
AM		 2583 attrs_a = slwt_a->desc->attrs | slwt_a->parsed_optattrs;
2584 attrs_b = slwt_b->desc->attrs | slwt_b->parsed_optattrs;
2585
2586 if (attrs_a != attrs_b)
d1df6fd8
DL		 2587 return 1;
2588
ac0dbed9 2589 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) {
300a0fd8 2590 if (attrs_a & SEG6_F_ATTR(i)) {
d1df6fd8
DL		 2591 param = &seg6_action_params[i];
2592 if (param->cmp(slwt_a, slwt_b))
2593 return 1;
2594 }
2595 }
2596
2597 return 0;
2598}
2599
2600static const struct lwtunnel_encap_ops seg6_local_ops = {
2601 .build_state = seg6_local_build_state,
2602 .destroy_state = seg6_local_destroy_state,
2603 .input = seg6_local_input,
2604 .fill_encap = seg6_local_fill_encap,
2605 .get_encap_size = seg6_local_get_encap_size,
2606 .cmp_encap = seg6_local_cmp_encap,
2607 .owner = THIS_MODULE,
2608};
2609
2610int __init seg6_local_init(void)
2611{
300a0fd8
AM		 2612 /* If the max total number of defined attributes is reached, then your
2613 * kernel build stops here.
2614 *
2615 * This check is required to avoid arithmetic overflows when processing
2616 * behavior attributes and the maximum number of defined attributes
2617 * exceeds the allowed value.
2618 */
2619 BUILD_BUG_ON(SEG6_LOCAL_MAX + 1 > BITS_PER_TYPE(unsigned long));
2620
848f3c0d
AM		 2621 /* If the default NEXT-C-SID Locator-Block/Node Function lengths (in
2622 * bits) have been changed with invalid values, kernel build stops
2623 * here.
2624 */
2625 BUILD_BUG_ON(next_csid_chk_cntr_bits(SEG6_LOCAL_LCBLOCK_DBITS,
2626 SEG6_LOCAL_LCNODE_FN_DBITS));
2627 BUILD_BUG_ON(next_csid_chk_lcblock_bits(SEG6_LOCAL_LCBLOCK_DBITS));
2628 BUILD_BUG_ON(next_csid_chk_lcnode_fn_bits(SEG6_LOCAL_LCNODE_FN_DBITS));
2629
bdf3c0b9
AM		 2630 /* To be memory efficient, we use 'u8' to represent the different
2631 * actions related to RFC8986 flavors. If the kernel build stops here,
2632 * it means that it is not possible to correctly encode these actions
2633 * with the data type chosen for the action table.
2634 */
2635 BUILD_BUG_ON(SEG6_LOCAL_FLV_ACT_MAX > (typeof(flv8986_act_tbl[0]))~0U);
2636
d1df6fd8
DL		 2637 return lwtunnel_encap_add_ops(&seg6_local_ops,
2638 LWTUNNEL_ENCAP_SEG6_LOCAL);
2639}
2640
2641void seg6_local_exit(void)
2642{
2643 lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL);
2644}
Linux 6.x block layer and io_uring tree(s)