xref: /linux/net/ipv6/ioam6.c (revision fd7d598270724cc787982ea48bbe17ad383a8b7f)
1 // SPDX-License-Identifier: GPL-2.0+
2 /*
3  *  IPv6 IOAM implementation
4  *
5  *  Author:
6  *  Justin Iurman <justin.iurman@uliege.be>
7  */
8 
9 #include <linux/errno.h>
10 #include <linux/types.h>
11 #include <linux/kernel.h>
12 #include <linux/net.h>
13 #include <linux/ioam6.h>
14 #include <linux/ioam6_genl.h>
15 #include <linux/rhashtable.h>
16 #include <linux/netdevice.h>
17 
18 #include <net/addrconf.h>
19 #include <net/genetlink.h>
20 #include <net/ioam6.h>
21 #include <net/sch_generic.h>
22 
23 static void ioam6_ns_release(struct ioam6_namespace *ns)
24 {
25 	kfree_rcu(ns, rcu);
26 }
27 
28 static void ioam6_sc_release(struct ioam6_schema *sc)
29 {
30 	kfree_rcu(sc, rcu);
31 }
32 
33 static void ioam6_free_ns(void *ptr, void *arg)
34 {
35 	struct ioam6_namespace *ns = (struct ioam6_namespace *)ptr;
36 
37 	if (ns)
38 		ioam6_ns_release(ns);
39 }
40 
41 static void ioam6_free_sc(void *ptr, void *arg)
42 {
43 	struct ioam6_schema *sc = (struct ioam6_schema *)ptr;
44 
45 	if (sc)
46 		ioam6_sc_release(sc);
47 }
48 
49 static int ioam6_ns_cmpfn(struct rhashtable_compare_arg *arg, const void *obj)
50 {
51 	const struct ioam6_namespace *ns = obj;
52 
53 	return (ns->id != *(__be16 *)arg->key);
54 }
55 
56 static int ioam6_sc_cmpfn(struct rhashtable_compare_arg *arg, const void *obj)
57 {
58 	const struct ioam6_schema *sc = obj;
59 
60 	return (sc->id != *(u32 *)arg->key);
61 }
62 
63 static const struct rhashtable_params rht_ns_params = {
64 	.key_len		= sizeof(__be16),
65 	.key_offset		= offsetof(struct ioam6_namespace, id),
66 	.head_offset		= offsetof(struct ioam6_namespace, head),
67 	.automatic_shrinking	= true,
68 	.obj_cmpfn		= ioam6_ns_cmpfn,
69 };
70 
71 static const struct rhashtable_params rht_sc_params = {
72 	.key_len		= sizeof(u32),
73 	.key_offset		= offsetof(struct ioam6_schema, id),
74 	.head_offset		= offsetof(struct ioam6_schema, head),
75 	.automatic_shrinking	= true,
76 	.obj_cmpfn		= ioam6_sc_cmpfn,
77 };
78 
79 static struct genl_family ioam6_genl_family;
80 
81 static const struct nla_policy ioam6_genl_policy_addns[] = {
82 	[IOAM6_ATTR_NS_ID]	= { .type = NLA_U16 },
83 	[IOAM6_ATTR_NS_DATA]	= { .type = NLA_U32 },
84 	[IOAM6_ATTR_NS_DATA_WIDE] = { .type = NLA_U64 },
85 };
86 
87 static const struct nla_policy ioam6_genl_policy_delns[] = {
88 	[IOAM6_ATTR_NS_ID]	= { .type = NLA_U16 },
89 };
90 
91 static const struct nla_policy ioam6_genl_policy_addsc[] = {
92 	[IOAM6_ATTR_SC_ID]	= { .type = NLA_U32 },
93 	[IOAM6_ATTR_SC_DATA]	= { .type = NLA_BINARY,
94 				    .len = IOAM6_MAX_SCHEMA_DATA_LEN },
95 };
96 
97 static const struct nla_policy ioam6_genl_policy_delsc[] = {
98 	[IOAM6_ATTR_SC_ID]	= { .type = NLA_U32 },
99 };
100 
101 static const struct nla_policy ioam6_genl_policy_ns_sc[] = {
102 	[IOAM6_ATTR_NS_ID]	= { .type = NLA_U16 },
103 	[IOAM6_ATTR_SC_ID]	= { .type = NLA_U32 },
104 	[IOAM6_ATTR_SC_NONE]	= { .type = NLA_FLAG },
105 };
106 
107 static int ioam6_genl_addns(struct sk_buff *skb, struct genl_info *info)
108 {
109 	struct ioam6_pernet_data *nsdata;
110 	struct ioam6_namespace *ns;
111 	u64 data64;
112 	u32 data32;
113 	__be16 id;
114 	int err;
115 
116 	if (!info->attrs[IOAM6_ATTR_NS_ID])
117 		return -EINVAL;
118 
119 	id = cpu_to_be16(nla_get_u16(info->attrs[IOAM6_ATTR_NS_ID]));
120 	nsdata = ioam6_pernet(genl_info_net(info));
121 
122 	mutex_lock(&nsdata->lock);
123 
124 	ns = rhashtable_lookup_fast(&nsdata->namespaces, &id, rht_ns_params);
125 	if (ns) {
126 		err = -EEXIST;
127 		goto out_unlock;
128 	}
129 
130 	ns = kzalloc(sizeof(*ns), GFP_KERNEL);
131 	if (!ns) {
132 		err = -ENOMEM;
133 		goto out_unlock;
134 	}
135 
136 	ns->id = id;
137 
138 	if (!info->attrs[IOAM6_ATTR_NS_DATA])
139 		data32 = IOAM6_U32_UNAVAILABLE;
140 	else
141 		data32 = nla_get_u32(info->attrs[IOAM6_ATTR_NS_DATA]);
142 
143 	if (!info->attrs[IOAM6_ATTR_NS_DATA_WIDE])
144 		data64 = IOAM6_U64_UNAVAILABLE;
145 	else
146 		data64 = nla_get_u64(info->attrs[IOAM6_ATTR_NS_DATA_WIDE]);
147 
148 	ns->data = cpu_to_be32(data32);
149 	ns->data_wide = cpu_to_be64(data64);
150 
151 	err = rhashtable_lookup_insert_fast(&nsdata->namespaces, &ns->head,
152 					    rht_ns_params);
153 	if (err)
154 		kfree(ns);
155 
156 out_unlock:
157 	mutex_unlock(&nsdata->lock);
158 	return err;
159 }
160 
161 static int ioam6_genl_delns(struct sk_buff *skb, struct genl_info *info)
162 {
163 	struct ioam6_pernet_data *nsdata;
164 	struct ioam6_namespace *ns;
165 	struct ioam6_schema *sc;
166 	__be16 id;
167 	int err;
168 
169 	if (!info->attrs[IOAM6_ATTR_NS_ID])
170 		return -EINVAL;
171 
172 	id = cpu_to_be16(nla_get_u16(info->attrs[IOAM6_ATTR_NS_ID]));
173 	nsdata = ioam6_pernet(genl_info_net(info));
174 
175 	mutex_lock(&nsdata->lock);
176 
177 	ns = rhashtable_lookup_fast(&nsdata->namespaces, &id, rht_ns_params);
178 	if (!ns) {
179 		err = -ENOENT;
180 		goto out_unlock;
181 	}
182 
183 	sc = rcu_dereference_protected(ns->schema,
184 				       lockdep_is_held(&nsdata->lock));
185 
186 	err = rhashtable_remove_fast(&nsdata->namespaces, &ns->head,
187 				     rht_ns_params);
188 	if (err)
189 		goto out_unlock;
190 
191 	if (sc)
192 		rcu_assign_pointer(sc->ns, NULL);
193 
194 	ioam6_ns_release(ns);
195 
196 out_unlock:
197 	mutex_unlock(&nsdata->lock);
198 	return err;
199 }
200 
201 static int __ioam6_genl_dumpns_element(struct ioam6_namespace *ns,
202 				       u32 portid,
203 				       u32 seq,
204 				       u32 flags,
205 				       struct sk_buff *skb,
206 				       u8 cmd)
207 {
208 	struct ioam6_schema *sc;
209 	u64 data64;
210 	u32 data32;
211 	void *hdr;
212 
213 	hdr = genlmsg_put(skb, portid, seq, &ioam6_genl_family, flags, cmd);
214 	if (!hdr)
215 		return -ENOMEM;
216 
217 	data32 = be32_to_cpu(ns->data);
218 	data64 = be64_to_cpu(ns->data_wide);
219 
220 	if (nla_put_u16(skb, IOAM6_ATTR_NS_ID, be16_to_cpu(ns->id)) ||
221 	    (data32 != IOAM6_U32_UNAVAILABLE &&
222 	     nla_put_u32(skb, IOAM6_ATTR_NS_DATA, data32)) ||
223 	    (data64 != IOAM6_U64_UNAVAILABLE &&
224 	     nla_put_u64_64bit(skb, IOAM6_ATTR_NS_DATA_WIDE,
225 			       data64, IOAM6_ATTR_PAD)))
226 		goto nla_put_failure;
227 
228 	rcu_read_lock();
229 
230 	sc = rcu_dereference(ns->schema);
231 	if (sc && nla_put_u32(skb, IOAM6_ATTR_SC_ID, sc->id)) {
232 		rcu_read_unlock();
233 		goto nla_put_failure;
234 	}
235 
236 	rcu_read_unlock();
237 
238 	genlmsg_end(skb, hdr);
239 	return 0;
240 
241 nla_put_failure:
242 	genlmsg_cancel(skb, hdr);
243 	return -EMSGSIZE;
244 }
245 
246 static int ioam6_genl_dumpns_start(struct netlink_callback *cb)
247 {
248 	struct ioam6_pernet_data *nsdata = ioam6_pernet(sock_net(cb->skb->sk));
249 	struct rhashtable_iter *iter = (struct rhashtable_iter *)cb->args[0];
250 
251 	if (!iter) {
252 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
253 		if (!iter)
254 			return -ENOMEM;
255 
256 		cb->args[0] = (long)iter;
257 	}
258 
259 	rhashtable_walk_enter(&nsdata->namespaces, iter);
260 
261 	return 0;
262 }
263 
264 static int ioam6_genl_dumpns_done(struct netlink_callback *cb)
265 {
266 	struct rhashtable_iter *iter = (struct rhashtable_iter *)cb->args[0];
267 
268 	rhashtable_walk_exit(iter);
269 	kfree(iter);
270 
271 	return 0;
272 }
273 
274 static int ioam6_genl_dumpns(struct sk_buff *skb, struct netlink_callback *cb)
275 {
276 	struct rhashtable_iter *iter;
277 	struct ioam6_namespace *ns;
278 	int err;
279 
280 	iter = (struct rhashtable_iter *)cb->args[0];
281 	rhashtable_walk_start(iter);
282 
283 	for (;;) {
284 		ns = rhashtable_walk_next(iter);
285 
286 		if (IS_ERR(ns)) {
287 			if (PTR_ERR(ns) == -EAGAIN)
288 				continue;
289 			err = PTR_ERR(ns);
290 			goto done;
291 		} else if (!ns) {
292 			break;
293 		}
294 
295 		err = __ioam6_genl_dumpns_element(ns,
296 						  NETLINK_CB(cb->skb).portid,
297 						  cb->nlh->nlmsg_seq,
298 						  NLM_F_MULTI,
299 						  skb,
300 						  IOAM6_CMD_DUMP_NAMESPACES);
301 		if (err)
302 			goto done;
303 	}
304 
305 	err = skb->len;
306 
307 done:
308 	rhashtable_walk_stop(iter);
309 	return err;
310 }
311 
312 static int ioam6_genl_addsc(struct sk_buff *skb, struct genl_info *info)
313 {
314 	struct ioam6_pernet_data *nsdata;
315 	int len, len_aligned, err;
316 	struct ioam6_schema *sc;
317 	u32 id;
318 
319 	if (!info->attrs[IOAM6_ATTR_SC_ID] || !info->attrs[IOAM6_ATTR_SC_DATA])
320 		return -EINVAL;
321 
322 	id = nla_get_u32(info->attrs[IOAM6_ATTR_SC_ID]);
323 	nsdata = ioam6_pernet(genl_info_net(info));
324 
325 	mutex_lock(&nsdata->lock);
326 
327 	sc = rhashtable_lookup_fast(&nsdata->schemas, &id, rht_sc_params);
328 	if (sc) {
329 		err = -EEXIST;
330 		goto out_unlock;
331 	}
332 
333 	len = nla_len(info->attrs[IOAM6_ATTR_SC_DATA]);
334 	len_aligned = ALIGN(len, 4);
335 
336 	sc = kzalloc(sizeof(*sc) + len_aligned, GFP_KERNEL);
337 	if (!sc) {
338 		err = -ENOMEM;
339 		goto out_unlock;
340 	}
341 
342 	sc->id = id;
343 	sc->len = len_aligned;
344 	sc->hdr = cpu_to_be32(sc->id | ((u8)(sc->len / 4) << 24));
345 	nla_memcpy(sc->data, info->attrs[IOAM6_ATTR_SC_DATA], len);
346 
347 	err = rhashtable_lookup_insert_fast(&nsdata->schemas, &sc->head,
348 					    rht_sc_params);
349 	if (err)
350 		goto free_sc;
351 
352 out_unlock:
353 	mutex_unlock(&nsdata->lock);
354 	return err;
355 free_sc:
356 	kfree(sc);
357 	goto out_unlock;
358 }
359 
360 static int ioam6_genl_delsc(struct sk_buff *skb, struct genl_info *info)
361 {
362 	struct ioam6_pernet_data *nsdata;
363 	struct ioam6_namespace *ns;
364 	struct ioam6_schema *sc;
365 	int err;
366 	u32 id;
367 
368 	if (!info->attrs[IOAM6_ATTR_SC_ID])
369 		return -EINVAL;
370 
371 	id = nla_get_u32(info->attrs[IOAM6_ATTR_SC_ID]);
372 	nsdata = ioam6_pernet(genl_info_net(info));
373 
374 	mutex_lock(&nsdata->lock);
375 
376 	sc = rhashtable_lookup_fast(&nsdata->schemas, &id, rht_sc_params);
377 	if (!sc) {
378 		err = -ENOENT;
379 		goto out_unlock;
380 	}
381 
382 	ns = rcu_dereference_protected(sc->ns, lockdep_is_held(&nsdata->lock));
383 
384 	err = rhashtable_remove_fast(&nsdata->schemas, &sc->head,
385 				     rht_sc_params);
386 	if (err)
387 		goto out_unlock;
388 
389 	if (ns)
390 		rcu_assign_pointer(ns->schema, NULL);
391 
392 	ioam6_sc_release(sc);
393 
394 out_unlock:
395 	mutex_unlock(&nsdata->lock);
396 	return err;
397 }
398 
399 static int __ioam6_genl_dumpsc_element(struct ioam6_schema *sc,
400 				       u32 portid, u32 seq, u32 flags,
401 				       struct sk_buff *skb, u8 cmd)
402 {
403 	struct ioam6_namespace *ns;
404 	void *hdr;
405 
406 	hdr = genlmsg_put(skb, portid, seq, &ioam6_genl_family, flags, cmd);
407 	if (!hdr)
408 		return -ENOMEM;
409 
410 	if (nla_put_u32(skb, IOAM6_ATTR_SC_ID, sc->id) ||
411 	    nla_put(skb, IOAM6_ATTR_SC_DATA, sc->len, sc->data))
412 		goto nla_put_failure;
413 
414 	rcu_read_lock();
415 
416 	ns = rcu_dereference(sc->ns);
417 	if (ns && nla_put_u16(skb, IOAM6_ATTR_NS_ID, be16_to_cpu(ns->id))) {
418 		rcu_read_unlock();
419 		goto nla_put_failure;
420 	}
421 
422 	rcu_read_unlock();
423 
424 	genlmsg_end(skb, hdr);
425 	return 0;
426 
427 nla_put_failure:
428 	genlmsg_cancel(skb, hdr);
429 	return -EMSGSIZE;
430 }
431 
432 static int ioam6_genl_dumpsc_start(struct netlink_callback *cb)
433 {
434 	struct ioam6_pernet_data *nsdata = ioam6_pernet(sock_net(cb->skb->sk));
435 	struct rhashtable_iter *iter = (struct rhashtable_iter *)cb->args[0];
436 
437 	if (!iter) {
438 		iter = kmalloc(sizeof(*iter), GFP_KERNEL);
439 		if (!iter)
440 			return -ENOMEM;
441 
442 		cb->args[0] = (long)iter;
443 	}
444 
445 	rhashtable_walk_enter(&nsdata->schemas, iter);
446 
447 	return 0;
448 }
449 
450 static int ioam6_genl_dumpsc_done(struct netlink_callback *cb)
451 {
452 	struct rhashtable_iter *iter = (struct rhashtable_iter *)cb->args[0];
453 
454 	rhashtable_walk_exit(iter);
455 	kfree(iter);
456 
457 	return 0;
458 }
459 
460 static int ioam6_genl_dumpsc(struct sk_buff *skb, struct netlink_callback *cb)
461 {
462 	struct rhashtable_iter *iter;
463 	struct ioam6_schema *sc;
464 	int err;
465 
466 	iter = (struct rhashtable_iter *)cb->args[0];
467 	rhashtable_walk_start(iter);
468 
469 	for (;;) {
470 		sc = rhashtable_walk_next(iter);
471 
472 		if (IS_ERR(sc)) {
473 			if (PTR_ERR(sc) == -EAGAIN)
474 				continue;
475 			err = PTR_ERR(sc);
476 			goto done;
477 		} else if (!sc) {
478 			break;
479 		}
480 
481 		err = __ioam6_genl_dumpsc_element(sc,
482 						  NETLINK_CB(cb->skb).portid,
483 						  cb->nlh->nlmsg_seq,
484 						  NLM_F_MULTI,
485 						  skb,
486 						  IOAM6_CMD_DUMP_SCHEMAS);
487 		if (err)
488 			goto done;
489 	}
490 
491 	err = skb->len;
492 
493 done:
494 	rhashtable_walk_stop(iter);
495 	return err;
496 }
497 
498 static int ioam6_genl_ns_set_schema(struct sk_buff *skb, struct genl_info *info)
499 {
500 	struct ioam6_namespace *ns, *ns_ref;
501 	struct ioam6_schema *sc, *sc_ref;
502 	struct ioam6_pernet_data *nsdata;
503 	__be16 ns_id;
504 	u32 sc_id;
505 	int err;
506 
507 	if (!info->attrs[IOAM6_ATTR_NS_ID] ||
508 	    (!info->attrs[IOAM6_ATTR_SC_ID] &&
509 	     !info->attrs[IOAM6_ATTR_SC_NONE]))
510 		return -EINVAL;
511 
512 	ns_id = cpu_to_be16(nla_get_u16(info->attrs[IOAM6_ATTR_NS_ID]));
513 	nsdata = ioam6_pernet(genl_info_net(info));
514 
515 	mutex_lock(&nsdata->lock);
516 
517 	ns = rhashtable_lookup_fast(&nsdata->namespaces, &ns_id, rht_ns_params);
518 	if (!ns) {
519 		err = -ENOENT;
520 		goto out_unlock;
521 	}
522 
523 	if (info->attrs[IOAM6_ATTR_SC_NONE]) {
524 		sc = NULL;
525 	} else {
526 		sc_id = nla_get_u32(info->attrs[IOAM6_ATTR_SC_ID]);
527 		sc = rhashtable_lookup_fast(&nsdata->schemas, &sc_id,
528 					    rht_sc_params);
529 		if (!sc) {
530 			err = -ENOENT;
531 			goto out_unlock;
532 		}
533 	}
534 
535 	sc_ref = rcu_dereference_protected(ns->schema,
536 					   lockdep_is_held(&nsdata->lock));
537 	if (sc_ref)
538 		rcu_assign_pointer(sc_ref->ns, NULL);
539 	rcu_assign_pointer(ns->schema, sc);
540 
541 	if (sc) {
542 		ns_ref = rcu_dereference_protected(sc->ns,
543 						   lockdep_is_held(&nsdata->lock));
544 		if (ns_ref)
545 			rcu_assign_pointer(ns_ref->schema, NULL);
546 		rcu_assign_pointer(sc->ns, ns);
547 	}
548 
549 	err = 0;
550 
551 out_unlock:
552 	mutex_unlock(&nsdata->lock);
553 	return err;
554 }
555 
556 static const struct genl_ops ioam6_genl_ops[] = {
557 	{
558 		.cmd	= IOAM6_CMD_ADD_NAMESPACE,
559 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
560 		.doit	= ioam6_genl_addns,
561 		.flags	= GENL_ADMIN_PERM,
562 		.policy	= ioam6_genl_policy_addns,
563 		.maxattr = ARRAY_SIZE(ioam6_genl_policy_addns) - 1,
564 	},
565 	{
566 		.cmd	= IOAM6_CMD_DEL_NAMESPACE,
567 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
568 		.doit	= ioam6_genl_delns,
569 		.flags	= GENL_ADMIN_PERM,
570 		.policy	= ioam6_genl_policy_delns,
571 		.maxattr = ARRAY_SIZE(ioam6_genl_policy_delns) - 1,
572 	},
573 	{
574 		.cmd	= IOAM6_CMD_DUMP_NAMESPACES,
575 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
576 		.start	= ioam6_genl_dumpns_start,
577 		.dumpit	= ioam6_genl_dumpns,
578 		.done	= ioam6_genl_dumpns_done,
579 		.flags	= GENL_ADMIN_PERM,
580 	},
581 	{
582 		.cmd	= IOAM6_CMD_ADD_SCHEMA,
583 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
584 		.doit	= ioam6_genl_addsc,
585 		.flags	= GENL_ADMIN_PERM,
586 		.policy	= ioam6_genl_policy_addsc,
587 		.maxattr = ARRAY_SIZE(ioam6_genl_policy_addsc) - 1,
588 	},
589 	{
590 		.cmd	= IOAM6_CMD_DEL_SCHEMA,
591 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
592 		.doit	= ioam6_genl_delsc,
593 		.flags	= GENL_ADMIN_PERM,
594 		.policy	= ioam6_genl_policy_delsc,
595 		.maxattr = ARRAY_SIZE(ioam6_genl_policy_delsc) - 1,
596 	},
597 	{
598 		.cmd	= IOAM6_CMD_DUMP_SCHEMAS,
599 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
600 		.start	= ioam6_genl_dumpsc_start,
601 		.dumpit	= ioam6_genl_dumpsc,
602 		.done	= ioam6_genl_dumpsc_done,
603 		.flags	= GENL_ADMIN_PERM,
604 	},
605 	{
606 		.cmd	= IOAM6_CMD_NS_SET_SCHEMA,
607 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
608 		.doit	= ioam6_genl_ns_set_schema,
609 		.flags	= GENL_ADMIN_PERM,
610 		.policy	= ioam6_genl_policy_ns_sc,
611 		.maxattr = ARRAY_SIZE(ioam6_genl_policy_ns_sc) - 1,
612 	},
613 };
614 
615 static struct genl_family ioam6_genl_family __ro_after_init = {
616 	.name		= IOAM6_GENL_NAME,
617 	.version	= IOAM6_GENL_VERSION,
618 	.netnsok	= true,
619 	.parallel_ops	= true,
620 	.ops		= ioam6_genl_ops,
621 	.n_ops		= ARRAY_SIZE(ioam6_genl_ops),
622 	.resv_start_op	= IOAM6_CMD_NS_SET_SCHEMA + 1,
623 	.module		= THIS_MODULE,
624 };
625 
626 struct ioam6_namespace *ioam6_namespace(struct net *net, __be16 id)
627 {
628 	struct ioam6_pernet_data *nsdata = ioam6_pernet(net);
629 
630 	return rhashtable_lookup_fast(&nsdata->namespaces, &id, rht_ns_params);
631 }
632 
633 static void __ioam6_fill_trace_data(struct sk_buff *skb,
634 				    struct ioam6_namespace *ns,
635 				    struct ioam6_trace_hdr *trace,
636 				    struct ioam6_schema *sc,
637 				    u8 sclen, bool is_input)
638 {
639 	struct timespec64 ts;
640 	ktime_t tstamp;
641 	u64 raw64;
642 	u32 raw32;
643 	u16 raw16;
644 	u8 *data;
645 	u8 byte;
646 
647 	data = trace->data + trace->remlen * 4 - trace->nodelen * 4 - sclen * 4;
648 
649 	/* hop_lim and node_id */
650 	if (trace->type.bit0) {
651 		byte = ipv6_hdr(skb)->hop_limit;
652 		if (is_input)
653 			byte--;
654 
655 		raw32 = dev_net(skb_dst(skb)->dev)->ipv6.sysctl.ioam6_id;
656 
657 		*(__be32 *)data = cpu_to_be32((byte << 24) | raw32);
658 		data += sizeof(__be32);
659 	}
660 
661 	/* ingress_if_id and egress_if_id */
662 	if (trace->type.bit1) {
663 		if (!skb->dev)
664 			raw16 = IOAM6_U16_UNAVAILABLE;
665 		else
666 			raw16 = (__force u16)__in6_dev_get(skb->dev)->cnf.ioam6_id;
667 
668 		*(__be16 *)data = cpu_to_be16(raw16);
669 		data += sizeof(__be16);
670 
671 		if (skb_dst(skb)->dev->flags & IFF_LOOPBACK)
672 			raw16 = IOAM6_U16_UNAVAILABLE;
673 		else
674 			raw16 = (__force u16)__in6_dev_get(skb_dst(skb)->dev)->cnf.ioam6_id;
675 
676 		*(__be16 *)data = cpu_to_be16(raw16);
677 		data += sizeof(__be16);
678 	}
679 
680 	/* timestamp seconds */
681 	if (trace->type.bit2) {
682 		if (!skb->dev) {
683 			*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
684 		} else {
685 			tstamp = skb_tstamp_cond(skb, true);
686 			ts = ktime_to_timespec64(tstamp);
687 
688 			*(__be32 *)data = cpu_to_be32((u32)ts.tv_sec);
689 		}
690 		data += sizeof(__be32);
691 	}
692 
693 	/* timestamp subseconds */
694 	if (trace->type.bit3) {
695 		if (!skb->dev) {
696 			*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
697 		} else {
698 			if (!trace->type.bit2) {
699 				tstamp = skb_tstamp_cond(skb, true);
700 				ts = ktime_to_timespec64(tstamp);
701 			}
702 
703 			*(__be32 *)data = cpu_to_be32((u32)(ts.tv_nsec / NSEC_PER_USEC));
704 		}
705 		data += sizeof(__be32);
706 	}
707 
708 	/* transit delay */
709 	if (trace->type.bit4) {
710 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
711 		data += sizeof(__be32);
712 	}
713 
714 	/* namespace data */
715 	if (trace->type.bit5) {
716 		*(__be32 *)data = ns->data;
717 		data += sizeof(__be32);
718 	}
719 
720 	/* queue depth */
721 	if (trace->type.bit6) {
722 		struct netdev_queue *queue;
723 		struct Qdisc *qdisc;
724 		__u32 qlen, backlog;
725 
726 		if (skb_dst(skb)->dev->flags & IFF_LOOPBACK) {
727 			*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
728 		} else {
729 			queue = skb_get_tx_queue(skb_dst(skb)->dev, skb);
730 			qdisc = rcu_dereference(queue->qdisc);
731 			qdisc_qstats_qlen_backlog(qdisc, &qlen, &backlog);
732 
733 			*(__be32 *)data = cpu_to_be32(backlog);
734 		}
735 		data += sizeof(__be32);
736 	}
737 
738 	/* checksum complement */
739 	if (trace->type.bit7) {
740 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
741 		data += sizeof(__be32);
742 	}
743 
744 	/* hop_lim and node_id (wide) */
745 	if (trace->type.bit8) {
746 		byte = ipv6_hdr(skb)->hop_limit;
747 		if (is_input)
748 			byte--;
749 
750 		raw64 = dev_net(skb_dst(skb)->dev)->ipv6.sysctl.ioam6_id_wide;
751 
752 		*(__be64 *)data = cpu_to_be64(((u64)byte << 56) | raw64);
753 		data += sizeof(__be64);
754 	}
755 
756 	/* ingress_if_id and egress_if_id (wide) */
757 	if (trace->type.bit9) {
758 		if (!skb->dev)
759 			raw32 = IOAM6_U32_UNAVAILABLE;
760 		else
761 			raw32 = __in6_dev_get(skb->dev)->cnf.ioam6_id_wide;
762 
763 		*(__be32 *)data = cpu_to_be32(raw32);
764 		data += sizeof(__be32);
765 
766 		if (skb_dst(skb)->dev->flags & IFF_LOOPBACK)
767 			raw32 = IOAM6_U32_UNAVAILABLE;
768 		else
769 			raw32 = __in6_dev_get(skb_dst(skb)->dev)->cnf.ioam6_id_wide;
770 
771 		*(__be32 *)data = cpu_to_be32(raw32);
772 		data += sizeof(__be32);
773 	}
774 
775 	/* namespace data (wide) */
776 	if (trace->type.bit10) {
777 		*(__be64 *)data = ns->data_wide;
778 		data += sizeof(__be64);
779 	}
780 
781 	/* buffer occupancy */
782 	if (trace->type.bit11) {
783 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
784 		data += sizeof(__be32);
785 	}
786 
787 	/* bit12 undefined: filled with empty value */
788 	if (trace->type.bit12) {
789 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
790 		data += sizeof(__be32);
791 	}
792 
793 	/* bit13 undefined: filled with empty value */
794 	if (trace->type.bit13) {
795 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
796 		data += sizeof(__be32);
797 	}
798 
799 	/* bit14 undefined: filled with empty value */
800 	if (trace->type.bit14) {
801 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
802 		data += sizeof(__be32);
803 	}
804 
805 	/* bit15 undefined: filled with empty value */
806 	if (trace->type.bit15) {
807 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
808 		data += sizeof(__be32);
809 	}
810 
811 	/* bit16 undefined: filled with empty value */
812 	if (trace->type.bit16) {
813 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
814 		data += sizeof(__be32);
815 	}
816 
817 	/* bit17 undefined: filled with empty value */
818 	if (trace->type.bit17) {
819 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
820 		data += sizeof(__be32);
821 	}
822 
823 	/* bit18 undefined: filled with empty value */
824 	if (trace->type.bit18) {
825 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
826 		data += sizeof(__be32);
827 	}
828 
829 	/* bit19 undefined: filled with empty value */
830 	if (trace->type.bit19) {
831 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
832 		data += sizeof(__be32);
833 	}
834 
835 	/* bit20 undefined: filled with empty value */
836 	if (trace->type.bit20) {
837 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
838 		data += sizeof(__be32);
839 	}
840 
841 	/* bit21 undefined: filled with empty value */
842 	if (trace->type.bit21) {
843 		*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE);
844 		data += sizeof(__be32);
845 	}
846 
847 	/* opaque state snapshot */
848 	if (trace->type.bit22) {
849 		if (!sc) {
850 			*(__be32 *)data = cpu_to_be32(IOAM6_U32_UNAVAILABLE >> 8);
851 		} else {
852 			*(__be32 *)data = sc->hdr;
853 			data += sizeof(__be32);
854 
855 			memcpy(data, sc->data, sc->len);
856 		}
857 	}
858 }
859 
860 /* called with rcu_read_lock() */
861 void ioam6_fill_trace_data(struct sk_buff *skb,
862 			   struct ioam6_namespace *ns,
863 			   struct ioam6_trace_hdr *trace,
864 			   bool is_input)
865 {
866 	struct ioam6_schema *sc;
867 	u8 sclen = 0;
868 
869 	/* Skip if Overflow flag is set
870 	 */
871 	if (trace->overflow)
872 		return;
873 
874 	/* NodeLen does not include Opaque State Snapshot length. We need to
875 	 * take it into account if the corresponding bit is set (bit 22) and
876 	 * if the current IOAM namespace has an active schema attached to it
877 	 */
878 	sc = rcu_dereference(ns->schema);
879 	if (trace->type.bit22) {
880 		sclen = sizeof_field(struct ioam6_schema, hdr) / 4;
881 
882 		if (sc)
883 			sclen += sc->len / 4;
884 	}
885 
886 	/* If there is no space remaining, we set the Overflow flag and we
887 	 * skip without filling the trace
888 	 */
889 	if (!trace->remlen || trace->remlen < trace->nodelen + sclen) {
890 		trace->overflow = 1;
891 		return;
892 	}
893 
894 	__ioam6_fill_trace_data(skb, ns, trace, sc, sclen, is_input);
895 	trace->remlen -= trace->nodelen + sclen;
896 }
897 
898 static int __net_init ioam6_net_init(struct net *net)
899 {
900 	struct ioam6_pernet_data *nsdata;
901 	int err = -ENOMEM;
902 
903 	nsdata = kzalloc(sizeof(*nsdata), GFP_KERNEL);
904 	if (!nsdata)
905 		goto out;
906 
907 	mutex_init(&nsdata->lock);
908 	net->ipv6.ioam6_data = nsdata;
909 
910 	err = rhashtable_init(&nsdata->namespaces, &rht_ns_params);
911 	if (err)
912 		goto free_nsdata;
913 
914 	err = rhashtable_init(&nsdata->schemas, &rht_sc_params);
915 	if (err)
916 		goto free_rht_ns;
917 
918 out:
919 	return err;
920 free_rht_ns:
921 	rhashtable_destroy(&nsdata->namespaces);
922 free_nsdata:
923 	kfree(nsdata);
924 	net->ipv6.ioam6_data = NULL;
925 	goto out;
926 }
927 
928 static void __net_exit ioam6_net_exit(struct net *net)
929 {
930 	struct ioam6_pernet_data *nsdata = ioam6_pernet(net);
931 
932 	rhashtable_free_and_destroy(&nsdata->namespaces, ioam6_free_ns, NULL);
933 	rhashtable_free_and_destroy(&nsdata->schemas, ioam6_free_sc, NULL);
934 
935 	kfree(nsdata);
936 }
937 
938 static struct pernet_operations ioam6_net_ops = {
939 	.init = ioam6_net_init,
940 	.exit = ioam6_net_exit,
941 };
942 
943 int __init ioam6_init(void)
944 {
945 	int err = register_pernet_subsys(&ioam6_net_ops);
946 	if (err)
947 		goto out;
948 
949 	err = genl_register_family(&ioam6_genl_family);
950 	if (err)
951 		goto out_unregister_pernet_subsys;
952 
953 #ifdef CONFIG_IPV6_IOAM6_LWTUNNEL
954 	err = ioam6_iptunnel_init();
955 	if (err)
956 		goto out_unregister_genl;
957 #endif
958 
959 	pr_info("In-situ OAM (IOAM) with IPv6\n");
960 
961 out:
962 	return err;
963 #ifdef CONFIG_IPV6_IOAM6_LWTUNNEL
964 out_unregister_genl:
965 	genl_unregister_family(&ioam6_genl_family);
966 #endif
967 out_unregister_pernet_subsys:
968 	unregister_pernet_subsys(&ioam6_net_ops);
969 	goto out;
970 }
971 
972 void ioam6_exit(void)
973 {
974 #ifdef CONFIG_IPV6_IOAM6_LWTUNNEL
975 	ioam6_iptunnel_exit();
976 #endif
977 	genl_unregister_family(&ioam6_genl_family);
978 	unregister_pernet_subsys(&ioam6_net_ops);
979 }
980