Linux Audio

Check our new training course

Loading...
v6.13.7
  1// SPDX-License-Identifier: GPL-2.0-or-later
  2/*
  3 * net/sched/cls_flow.c		Generic flow classifier
  4 *
  5 * Copyright (c) 2007, 2008 Patrick McHardy <kaber@trash.net>
 
 
 
 
 
  6 */
  7
  8#include <linux/kernel.h>
  9#include <linux/init.h>
 10#include <linux/list.h>
 11#include <linux/jhash.h>
 12#include <linux/random.h>
 13#include <linux/pkt_cls.h>
 14#include <linux/skbuff.h>
 15#include <linux/in.h>
 16#include <linux/ip.h>
 17#include <linux/ipv6.h>
 18#include <linux/if_vlan.h>
 19#include <linux/slab.h>
 20#include <linux/module.h>
 21#include <net/inet_sock.h>
 22
 23#include <net/pkt_cls.h>
 24#include <net/ip.h>
 25#include <net/route.h>
 26#include <net/flow_dissector.h>
 27#include <net/tc_wrapper.h>
 28
 29#if IS_ENABLED(CONFIG_NF_CONNTRACK)
 30#include <net/netfilter/nf_conntrack.h>
 31#endif
 32
 33struct flow_head {
 34	struct list_head	filters;
 35	struct rcu_head		rcu;
 36};
 37
 38struct flow_filter {
 39	struct list_head	list;
 40	struct tcf_exts		exts;
 41	struct tcf_ematch_tree	ematches;
 42	struct tcf_proto	*tp;
 43	struct timer_list	perturb_timer;
 44	u32			perturb_period;
 45	u32			handle;
 46
 47	u32			nkeys;
 48	u32			keymask;
 49	u32			mode;
 50	u32			mask;
 51	u32			xor;
 52	u32			rshift;
 53	u32			addend;
 54	u32			divisor;
 55	u32			baseclass;
 56	u32			hashrnd;
 57	struct rcu_work		rwork;
 
 
 
 58};
 59
 60static inline u32 addr_fold(void *addr)
 61{
 62	unsigned long a = (unsigned long)addr;
 63
 64	return (a & 0xFFFFFFFF) ^ (BITS_PER_LONG > 32 ? a >> 32 : 0);
 65}
 66
 67static u32 flow_get_src(const struct sk_buff *skb, const struct flow_keys *flow)
 68{
 69	__be32 src = flow_get_u32_src(flow);
 70
 71	if (src)
 72		return ntohl(src);
 73
 74	return addr_fold(skb->sk);
 75}
 76
 77static u32 flow_get_dst(const struct sk_buff *skb, const struct flow_keys *flow)
 78{
 79	__be32 dst = flow_get_u32_dst(flow);
 80
 81	if (dst)
 82		return ntohl(dst);
 83
 84	return addr_fold(skb_dst(skb)) ^ (__force u16)skb_protocol(skb, true);
 85}
 86
 87static u32 flow_get_proto(const struct sk_buff *skb,
 88			  const struct flow_keys *flow)
 89{
 90	return flow->basic.ip_proto;
 91}
 92
 93static u32 flow_get_proto_src(const struct sk_buff *skb,
 94			      const struct flow_keys *flow)
 95{
 96	if (flow->ports.ports)
 97		return ntohs(flow->ports.src);
 98
 99	return addr_fold(skb->sk);
100}
101
102static u32 flow_get_proto_dst(const struct sk_buff *skb,
103			      const struct flow_keys *flow)
104{
105	if (flow->ports.ports)
106		return ntohs(flow->ports.dst);
107
108	return addr_fold(skb_dst(skb)) ^ (__force u16)skb_protocol(skb, true);
109}
110
111static u32 flow_get_iif(const struct sk_buff *skb)
112{
113	return skb->skb_iif;
114}
115
116static u32 flow_get_priority(const struct sk_buff *skb)
117{
118	return skb->priority;
119}
120
121static u32 flow_get_mark(const struct sk_buff *skb)
122{
123	return skb->mark;
124}
125
126static u32 flow_get_nfct(const struct sk_buff *skb)
127{
128#if IS_ENABLED(CONFIG_NF_CONNTRACK)
129	return addr_fold(skb_nfct(skb));
130#else
131	return 0;
132#endif
133}
134
135#if IS_ENABLED(CONFIG_NF_CONNTRACK)
136#define CTTUPLE(skb, member)						\
137({									\
138	enum ip_conntrack_info ctinfo;					\
139	const struct nf_conn *ct = nf_ct_get(skb, &ctinfo);		\
140	if (ct == NULL)							\
141		goto fallback;						\
142	ct->tuplehash[CTINFO2DIR(ctinfo)].tuple.member;			\
143})
144#else
145#define CTTUPLE(skb, member)						\
146({									\
147	goto fallback;							\
148	0;								\
149})
150#endif
151
152static u32 flow_get_nfct_src(const struct sk_buff *skb,
153			     const struct flow_keys *flow)
154{
155	switch (skb_protocol(skb, true)) {
156	case htons(ETH_P_IP):
157		return ntohl(CTTUPLE(skb, src.u3.ip));
158	case htons(ETH_P_IPV6):
159		return ntohl(CTTUPLE(skb, src.u3.ip6[3]));
160	}
161fallback:
162	return flow_get_src(skb, flow);
163}
164
165static u32 flow_get_nfct_dst(const struct sk_buff *skb,
166			     const struct flow_keys *flow)
167{
168	switch (skb_protocol(skb, true)) {
169	case htons(ETH_P_IP):
170		return ntohl(CTTUPLE(skb, dst.u3.ip));
171	case htons(ETH_P_IPV6):
172		return ntohl(CTTUPLE(skb, dst.u3.ip6[3]));
173	}
174fallback:
175	return flow_get_dst(skb, flow);
176}
177
178static u32 flow_get_nfct_proto_src(const struct sk_buff *skb,
179				   const struct flow_keys *flow)
180{
181	return ntohs(CTTUPLE(skb, src.u.all));
182fallback:
183	return flow_get_proto_src(skb, flow);
184}
185
186static u32 flow_get_nfct_proto_dst(const struct sk_buff *skb,
187				   const struct flow_keys *flow)
188{
189	return ntohs(CTTUPLE(skb, dst.u.all));
190fallback:
191	return flow_get_proto_dst(skb, flow);
192}
193
194static u32 flow_get_rtclassid(const struct sk_buff *skb)
195{
196#ifdef CONFIG_IP_ROUTE_CLASSID
197	if (skb_dst(skb))
198		return skb_dst(skb)->tclassid;
199#endif
200	return 0;
201}
202
203static u32 flow_get_skuid(const struct sk_buff *skb)
204{
205	struct sock *sk = skb_to_full_sk(skb);
206
207	if (sk && sk->sk_socket && sk->sk_socket->file) {
208		kuid_t skuid = sk->sk_socket->file->f_cred->fsuid;
209
210		return from_kuid(&init_user_ns, skuid);
211	}
212	return 0;
213}
214
215static u32 flow_get_skgid(const struct sk_buff *skb)
216{
217	struct sock *sk = skb_to_full_sk(skb);
218
219	if (sk && sk->sk_socket && sk->sk_socket->file) {
220		kgid_t skgid = sk->sk_socket->file->f_cred->fsgid;
221
222		return from_kgid(&init_user_ns, skgid);
223	}
224	return 0;
225}
226
227static u32 flow_get_vlan_tag(const struct sk_buff *skb)
228{
229	u16 tag;
230
231	if (vlan_get_tag(skb, &tag) < 0)
232		return 0;
233	return tag & VLAN_VID_MASK;
234}
235
236static u32 flow_get_rxhash(struct sk_buff *skb)
237{
238	return skb_get_hash(skb);
239}
240
241static u32 flow_key_get(struct sk_buff *skb, int key, struct flow_keys *flow)
242{
243	switch (key) {
244	case FLOW_KEY_SRC:
245		return flow_get_src(skb, flow);
246	case FLOW_KEY_DST:
247		return flow_get_dst(skb, flow);
248	case FLOW_KEY_PROTO:
249		return flow_get_proto(skb, flow);
250	case FLOW_KEY_PROTO_SRC:
251		return flow_get_proto_src(skb, flow);
252	case FLOW_KEY_PROTO_DST:
253		return flow_get_proto_dst(skb, flow);
254	case FLOW_KEY_IIF:
255		return flow_get_iif(skb);
256	case FLOW_KEY_PRIORITY:
257		return flow_get_priority(skb);
258	case FLOW_KEY_MARK:
259		return flow_get_mark(skb);
260	case FLOW_KEY_NFCT:
261		return flow_get_nfct(skb);
262	case FLOW_KEY_NFCT_SRC:
263		return flow_get_nfct_src(skb, flow);
264	case FLOW_KEY_NFCT_DST:
265		return flow_get_nfct_dst(skb, flow);
266	case FLOW_KEY_NFCT_PROTO_SRC:
267		return flow_get_nfct_proto_src(skb, flow);
268	case FLOW_KEY_NFCT_PROTO_DST:
269		return flow_get_nfct_proto_dst(skb, flow);
270	case FLOW_KEY_RTCLASSID:
271		return flow_get_rtclassid(skb);
272	case FLOW_KEY_SKUID:
273		return flow_get_skuid(skb);
274	case FLOW_KEY_SKGID:
275		return flow_get_skgid(skb);
276	case FLOW_KEY_VLAN_TAG:
277		return flow_get_vlan_tag(skb);
278	case FLOW_KEY_RXHASH:
279		return flow_get_rxhash(skb);
280	default:
281		WARN_ON(1);
282		return 0;
283	}
284}
285
286#define FLOW_KEYS_NEEDED ((1 << FLOW_KEY_SRC) | 		\
287			  (1 << FLOW_KEY_DST) |			\
288			  (1 << FLOW_KEY_PROTO) |		\
289			  (1 << FLOW_KEY_PROTO_SRC) |		\
290			  (1 << FLOW_KEY_PROTO_DST) | 		\
291			  (1 << FLOW_KEY_NFCT_SRC) |		\
292			  (1 << FLOW_KEY_NFCT_DST) |		\
293			  (1 << FLOW_KEY_NFCT_PROTO_SRC) |	\
294			  (1 << FLOW_KEY_NFCT_PROTO_DST))
295
296TC_INDIRECT_SCOPE int flow_classify(struct sk_buff *skb,
297				    const struct tcf_proto *tp,
298				    struct tcf_result *res)
299{
300	struct flow_head *head = rcu_dereference_bh(tp->root);
301	struct flow_filter *f;
302	u32 keymask;
303	u32 classid;
304	unsigned int n, key;
305	int r;
306
307	list_for_each_entry_rcu(f, &head->filters, list) {
308		u32 keys[FLOW_KEY_MAX + 1];
309		struct flow_keys flow_keys;
310
311		if (!tcf_em_tree_match(skb, &f->ematches, NULL))
312			continue;
313
314		keymask = f->keymask;
315		if (keymask & FLOW_KEYS_NEEDED)
316			skb_flow_dissect_flow_keys(skb, &flow_keys, 0);
317
318		for (n = 0; n < f->nkeys; n++) {
319			key = ffs(keymask) - 1;
320			keymask &= ~(1 << key);
321			keys[n] = flow_key_get(skb, key, &flow_keys);
322		}
323
324		if (f->mode == FLOW_MODE_HASH)
325			classid = jhash2(keys, f->nkeys, f->hashrnd);
326		else {
327			classid = keys[0];
328			classid = (classid & f->mask) ^ f->xor;
329			classid = (classid >> f->rshift) + f->addend;
330		}
331
332		if (f->divisor)
333			classid %= f->divisor;
334
335		res->class   = 0;
336		res->classid = TC_H_MAKE(f->baseclass, f->baseclass + classid);
337
338		r = tcf_exts_exec(skb, &f->exts, res);
339		if (r < 0)
340			continue;
341		return r;
342	}
343	return -1;
344}
345
346static void flow_perturbation(struct timer_list *t)
347{
348	struct flow_filter *f = from_timer(f, t, perturb_timer);
349
350	get_random_bytes(&f->hashrnd, 4);
351	if (f->perturb_period)
352		mod_timer(&f->perturb_timer, jiffies + f->perturb_period);
353}
354
355static const struct nla_policy flow_policy[TCA_FLOW_MAX + 1] = {
356	[TCA_FLOW_KEYS]		= { .type = NLA_U32 },
357	[TCA_FLOW_MODE]		= { .type = NLA_U32 },
358	[TCA_FLOW_BASECLASS]	= { .type = NLA_U32 },
359	[TCA_FLOW_RSHIFT]	= NLA_POLICY_MAX(NLA_U32,
360						 31 /* BITS_PER_U32 - 1 */),
361	[TCA_FLOW_ADDEND]	= { .type = NLA_U32 },
362	[TCA_FLOW_MASK]		= { .type = NLA_U32 },
363	[TCA_FLOW_XOR]		= { .type = NLA_U32 },
364	[TCA_FLOW_DIVISOR]	= { .type = NLA_U32 },
365	[TCA_FLOW_ACT]		= { .type = NLA_NESTED },
366	[TCA_FLOW_POLICE]	= { .type = NLA_NESTED },
367	[TCA_FLOW_EMATCHES]	= { .type = NLA_NESTED },
368	[TCA_FLOW_PERTURB]	= { .type = NLA_U32 },
369};
370
371static void __flow_destroy_filter(struct flow_filter *f)
372{
373	timer_shutdown_sync(&f->perturb_timer);
374	tcf_exts_destroy(&f->exts);
375	tcf_em_tree_destroy(&f->ematches);
376	tcf_exts_put_net(&f->exts);
377	kfree(f);
378}
379
380static void flow_destroy_filter_work(struct work_struct *work)
381{
382	struct flow_filter *f = container_of(to_rcu_work(work),
383					     struct flow_filter,
384					     rwork);
385	rtnl_lock();
386	__flow_destroy_filter(f);
387	rtnl_unlock();
388}
389
 
 
 
 
 
 
 
 
390static int flow_change(struct net *net, struct sk_buff *in_skb,
391		       struct tcf_proto *tp, unsigned long base,
392		       u32 handle, struct nlattr **tca,
393		       void **arg, u32 flags,
394		       struct netlink_ext_ack *extack)
395{
396	struct flow_head *head = rtnl_dereference(tp->root);
397	struct flow_filter *fold, *fnew;
398	struct nlattr *opt = tca[TCA_OPTIONS];
399	struct nlattr *tb[TCA_FLOW_MAX + 1];
400	unsigned int nkeys = 0;
401	unsigned int perturb_period = 0;
402	u32 baseclass = 0;
403	u32 keymask = 0;
404	u32 mode;
405	int err;
406
407	if (opt == NULL)
408		return -EINVAL;
409
410	err = nla_parse_nested_deprecated(tb, TCA_FLOW_MAX, opt, flow_policy,
411					  NULL);
412	if (err < 0)
413		return err;
414
415	if (tb[TCA_FLOW_BASECLASS]) {
416		baseclass = nla_get_u32(tb[TCA_FLOW_BASECLASS]);
417		if (TC_H_MIN(baseclass) == 0)
418			return -EINVAL;
419	}
420
421	if (tb[TCA_FLOW_KEYS]) {
422		keymask = nla_get_u32(tb[TCA_FLOW_KEYS]);
423
424		nkeys = hweight32(keymask);
425		if (nkeys == 0)
426			return -EINVAL;
427
428		if (fls(keymask) - 1 > FLOW_KEY_MAX)
429			return -EOPNOTSUPP;
430
431		if ((keymask & (FLOW_KEY_SKUID|FLOW_KEY_SKGID)) &&
432		    sk_user_ns(NETLINK_CB(in_skb).sk) != &init_user_ns)
433			return -EOPNOTSUPP;
434	}
435
436	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
437	if (!fnew)
438		return -ENOBUFS;
439
440	err = tcf_em_tree_validate(tp, tb[TCA_FLOW_EMATCHES], &fnew->ematches);
441	if (err < 0)
442		goto err1;
443
444	err = tcf_exts_init(&fnew->exts, net, TCA_FLOW_ACT, TCA_FLOW_POLICE);
445	if (err < 0)
446		goto err2;
447
448	err = tcf_exts_validate(net, tp, tb, tca[TCA_RATE], &fnew->exts, flags,
449				extack);
450	if (err < 0)
451		goto err2;
452
453	fold = *arg;
454	if (fold) {
455		err = -EINVAL;
456		if (fold->handle != handle && handle)
457			goto err2;
458
459		/* Copy fold into fnew */
460		fnew->tp = fold->tp;
461		fnew->handle = fold->handle;
462		fnew->nkeys = fold->nkeys;
463		fnew->keymask = fold->keymask;
464		fnew->mode = fold->mode;
465		fnew->mask = fold->mask;
466		fnew->xor = fold->xor;
467		fnew->rshift = fold->rshift;
468		fnew->addend = fold->addend;
469		fnew->divisor = fold->divisor;
470		fnew->baseclass = fold->baseclass;
471		fnew->hashrnd = fold->hashrnd;
472
473		mode = fold->mode;
474		if (tb[TCA_FLOW_MODE])
475			mode = nla_get_u32(tb[TCA_FLOW_MODE]);
476		if (mode != FLOW_MODE_HASH && nkeys > 1)
477			goto err2;
478
479		if (mode == FLOW_MODE_HASH)
480			perturb_period = fold->perturb_period;
481		if (tb[TCA_FLOW_PERTURB]) {
482			if (mode != FLOW_MODE_HASH)
483				goto err2;
484			perturb_period = nla_get_u32(tb[TCA_FLOW_PERTURB]) * HZ;
485		}
486	} else {
487		err = -EINVAL;
488		if (!handle)
489			goto err2;
490		if (!tb[TCA_FLOW_KEYS])
491			goto err2;
492
493		mode = FLOW_MODE_MAP;
494		if (tb[TCA_FLOW_MODE])
495			mode = nla_get_u32(tb[TCA_FLOW_MODE]);
496		if (mode != FLOW_MODE_HASH && nkeys > 1)
497			goto err2;
498
499		if (tb[TCA_FLOW_PERTURB]) {
500			if (mode != FLOW_MODE_HASH)
501				goto err2;
502			perturb_period = nla_get_u32(tb[TCA_FLOW_PERTURB]) * HZ;
503		}
504
505		if (TC_H_MAJ(baseclass) == 0) {
506			struct Qdisc *q = tcf_block_q(tp->chain->block);
507
508			baseclass = TC_H_MAKE(q->handle, baseclass);
509		}
510		if (TC_H_MIN(baseclass) == 0)
511			baseclass = TC_H_MAKE(baseclass, 1);
512
513		fnew->handle = handle;
514		fnew->mask  = ~0U;
515		fnew->tp = tp;
516		get_random_bytes(&fnew->hashrnd, 4);
517	}
518
519	timer_setup(&fnew->perturb_timer, flow_perturbation, TIMER_DEFERRABLE);
520
521	tcf_block_netif_keep_dst(tp->chain->block);
522
523	if (tb[TCA_FLOW_KEYS]) {
524		fnew->keymask = keymask;
525		fnew->nkeys   = nkeys;
526	}
527
528	fnew->mode = mode;
529
530	if (tb[TCA_FLOW_MASK])
531		fnew->mask = nla_get_u32(tb[TCA_FLOW_MASK]);
532	if (tb[TCA_FLOW_XOR])
533		fnew->xor = nla_get_u32(tb[TCA_FLOW_XOR]);
534	if (tb[TCA_FLOW_RSHIFT])
535		fnew->rshift = nla_get_u32(tb[TCA_FLOW_RSHIFT]);
536	if (tb[TCA_FLOW_ADDEND])
537		fnew->addend = nla_get_u32(tb[TCA_FLOW_ADDEND]);
538
539	if (tb[TCA_FLOW_DIVISOR])
540		fnew->divisor = nla_get_u32(tb[TCA_FLOW_DIVISOR]);
541	if (baseclass)
542		fnew->baseclass = baseclass;
543
544	fnew->perturb_period = perturb_period;
545	if (perturb_period)
546		mod_timer(&fnew->perturb_timer, jiffies + perturb_period);
547
548	if (!*arg)
549		list_add_tail_rcu(&fnew->list, &head->filters);
550	else
551		list_replace_rcu(&fold->list, &fnew->list);
552
553	*arg = fnew;
554
555	if (fold) {
556		tcf_exts_get_net(&fold->exts);
557		tcf_queue_work(&fold->rwork, flow_destroy_filter_work);
558	}
559	return 0;
560
561err2:
562	tcf_exts_destroy(&fnew->exts);
563	tcf_em_tree_destroy(&fnew->ematches);
564err1:
565	kfree(fnew);
566	return err;
567}
568
569static int flow_delete(struct tcf_proto *tp, void *arg, bool *last,
570		       bool rtnl_held, struct netlink_ext_ack *extack)
571{
572	struct flow_head *head = rtnl_dereference(tp->root);
573	struct flow_filter *f = arg;
574
575	list_del_rcu(&f->list);
576	tcf_exts_get_net(&f->exts);
577	tcf_queue_work(&f->rwork, flow_destroy_filter_work);
578	*last = list_empty(&head->filters);
579	return 0;
580}
581
582static int flow_init(struct tcf_proto *tp)
583{
584	struct flow_head *head;
585
586	head = kzalloc(sizeof(*head), GFP_KERNEL);
587	if (head == NULL)
588		return -ENOBUFS;
589	INIT_LIST_HEAD(&head->filters);
590	rcu_assign_pointer(tp->root, head);
591	return 0;
592}
593
594static void flow_destroy(struct tcf_proto *tp, bool rtnl_held,
595			 struct netlink_ext_ack *extack)
596{
597	struct flow_head *head = rtnl_dereference(tp->root);
598	struct flow_filter *f, *next;
599
600	list_for_each_entry_safe(f, next, &head->filters, list) {
601		list_del_rcu(&f->list);
602		if (tcf_exts_get_net(&f->exts))
603			tcf_queue_work(&f->rwork, flow_destroy_filter_work);
604		else
605			__flow_destroy_filter(f);
606	}
607	kfree_rcu(head, rcu);
608}
609
610static void *flow_get(struct tcf_proto *tp, u32 handle)
611{
612	struct flow_head *head = rtnl_dereference(tp->root);
613	struct flow_filter *f;
614
615	list_for_each_entry(f, &head->filters, list)
616		if (f->handle == handle)
617			return f;
618	return NULL;
619}
620
621static int flow_dump(struct net *net, struct tcf_proto *tp, void *fh,
622		     struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
623{
624	struct flow_filter *f = fh;
625	struct nlattr *nest;
626
627	if (f == NULL)
628		return skb->len;
629
630	t->tcm_handle = f->handle;
631
632	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
633	if (nest == NULL)
634		goto nla_put_failure;
635
636	if (nla_put_u32(skb, TCA_FLOW_KEYS, f->keymask) ||
637	    nla_put_u32(skb, TCA_FLOW_MODE, f->mode))
638		goto nla_put_failure;
639
640	if (f->mask != ~0 || f->xor != 0) {
641		if (nla_put_u32(skb, TCA_FLOW_MASK, f->mask) ||
642		    nla_put_u32(skb, TCA_FLOW_XOR, f->xor))
643			goto nla_put_failure;
644	}
645	if (f->rshift &&
646	    nla_put_u32(skb, TCA_FLOW_RSHIFT, f->rshift))
647		goto nla_put_failure;
648	if (f->addend &&
649	    nla_put_u32(skb, TCA_FLOW_ADDEND, f->addend))
650		goto nla_put_failure;
651
652	if (f->divisor &&
653	    nla_put_u32(skb, TCA_FLOW_DIVISOR, f->divisor))
654		goto nla_put_failure;
655	if (f->baseclass &&
656	    nla_put_u32(skb, TCA_FLOW_BASECLASS, f->baseclass))
657		goto nla_put_failure;
658
659	if (f->perturb_period &&
660	    nla_put_u32(skb, TCA_FLOW_PERTURB, f->perturb_period / HZ))
661		goto nla_put_failure;
662
663	if (tcf_exts_dump(skb, &f->exts) < 0)
664		goto nla_put_failure;
665#ifdef CONFIG_NET_EMATCH
666	if (f->ematches.hdr.nmatches &&
667	    tcf_em_tree_dump(skb, &f->ematches, TCA_FLOW_EMATCHES) < 0)
668		goto nla_put_failure;
669#endif
670	nla_nest_end(skb, nest);
671
672	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
673		goto nla_put_failure;
674
675	return skb->len;
676
677nla_put_failure:
678	nla_nest_cancel(skb, nest);
679	return -1;
680}
681
682static void flow_walk(struct tcf_proto *tp, struct tcf_walker *arg,
683		      bool rtnl_held)
684{
685	struct flow_head *head = rtnl_dereference(tp->root);
686	struct flow_filter *f;
687
688	list_for_each_entry(f, &head->filters, list) {
689		if (!tc_cls_stats_dump(tp, arg, f))
 
 
 
690			break;
 
 
 
691	}
692}
693
694static struct tcf_proto_ops cls_flow_ops __read_mostly = {
695	.kind		= "flow",
696	.classify	= flow_classify,
697	.init		= flow_init,
698	.destroy	= flow_destroy,
699	.change		= flow_change,
700	.delete		= flow_delete,
701	.get		= flow_get,
702	.dump		= flow_dump,
703	.walk		= flow_walk,
704	.owner		= THIS_MODULE,
705};
706MODULE_ALIAS_NET_CLS("flow");
707
708static int __init cls_flow_init(void)
709{
710	return register_tcf_proto_ops(&cls_flow_ops);
711}
712
713static void __exit cls_flow_exit(void)
714{
715	unregister_tcf_proto_ops(&cls_flow_ops);
716}
717
718module_init(cls_flow_init);
719module_exit(cls_flow_exit);
720
721MODULE_LICENSE("GPL");
722MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
723MODULE_DESCRIPTION("TC flow classifier");
v4.17
 
  1/*
  2 * net/sched/cls_flow.c		Generic flow classifier
  3 *
  4 * Copyright (c) 2007, 2008 Patrick McHardy <kaber@trash.net>
  5 *
  6 * This program is free software; you can redistribute it and/or
  7 * modify it under the terms of the GNU General Public License
  8 * as published by the Free Software Foundation; either version 2
  9 * of the License, or (at your option) any later version.
 10 */
 11
 12#include <linux/kernel.h>
 13#include <linux/init.h>
 14#include <linux/list.h>
 15#include <linux/jhash.h>
 16#include <linux/random.h>
 17#include <linux/pkt_cls.h>
 18#include <linux/skbuff.h>
 19#include <linux/in.h>
 20#include <linux/ip.h>
 21#include <linux/ipv6.h>
 22#include <linux/if_vlan.h>
 23#include <linux/slab.h>
 24#include <linux/module.h>
 25#include <net/inet_sock.h>
 26
 27#include <net/pkt_cls.h>
 28#include <net/ip.h>
 29#include <net/route.h>
 30#include <net/flow_dissector.h>
 
 31
 32#if IS_ENABLED(CONFIG_NF_CONNTRACK)
 33#include <net/netfilter/nf_conntrack.h>
 34#endif
 35
 36struct flow_head {
 37	struct list_head	filters;
 38	struct rcu_head		rcu;
 39};
 40
 41struct flow_filter {
 42	struct list_head	list;
 43	struct tcf_exts		exts;
 44	struct tcf_ematch_tree	ematches;
 45	struct tcf_proto	*tp;
 46	struct timer_list	perturb_timer;
 47	u32			perturb_period;
 48	u32			handle;
 49
 50	u32			nkeys;
 51	u32			keymask;
 52	u32			mode;
 53	u32			mask;
 54	u32			xor;
 55	u32			rshift;
 56	u32			addend;
 57	u32			divisor;
 58	u32			baseclass;
 59	u32			hashrnd;
 60	union {
 61		struct work_struct	work;
 62		struct rcu_head		rcu;
 63	};
 64};
 65
 66static inline u32 addr_fold(void *addr)
 67{
 68	unsigned long a = (unsigned long)addr;
 69
 70	return (a & 0xFFFFFFFF) ^ (BITS_PER_LONG > 32 ? a >> 32 : 0);
 71}
 72
 73static u32 flow_get_src(const struct sk_buff *skb, const struct flow_keys *flow)
 74{
 75	__be32 src = flow_get_u32_src(flow);
 76
 77	if (src)
 78		return ntohl(src);
 79
 80	return addr_fold(skb->sk);
 81}
 82
 83static u32 flow_get_dst(const struct sk_buff *skb, const struct flow_keys *flow)
 84{
 85	__be32 dst = flow_get_u32_dst(flow);
 86
 87	if (dst)
 88		return ntohl(dst);
 89
 90	return addr_fold(skb_dst(skb)) ^ (__force u16) tc_skb_protocol(skb);
 91}
 92
 93static u32 flow_get_proto(const struct sk_buff *skb,
 94			  const struct flow_keys *flow)
 95{
 96	return flow->basic.ip_proto;
 97}
 98
 99static u32 flow_get_proto_src(const struct sk_buff *skb,
100			      const struct flow_keys *flow)
101{
102	if (flow->ports.ports)
103		return ntohs(flow->ports.src);
104
105	return addr_fold(skb->sk);
106}
107
108static u32 flow_get_proto_dst(const struct sk_buff *skb,
109			      const struct flow_keys *flow)
110{
111	if (flow->ports.ports)
112		return ntohs(flow->ports.dst);
113
114	return addr_fold(skb_dst(skb)) ^ (__force u16) tc_skb_protocol(skb);
115}
116
117static u32 flow_get_iif(const struct sk_buff *skb)
118{
119	return skb->skb_iif;
120}
121
122static u32 flow_get_priority(const struct sk_buff *skb)
123{
124	return skb->priority;
125}
126
127static u32 flow_get_mark(const struct sk_buff *skb)
128{
129	return skb->mark;
130}
131
132static u32 flow_get_nfct(const struct sk_buff *skb)
133{
134#if IS_ENABLED(CONFIG_NF_CONNTRACK)
135	return addr_fold(skb_nfct(skb));
136#else
137	return 0;
138#endif
139}
140
141#if IS_ENABLED(CONFIG_NF_CONNTRACK)
142#define CTTUPLE(skb, member)						\
143({									\
144	enum ip_conntrack_info ctinfo;					\
145	const struct nf_conn *ct = nf_ct_get(skb, &ctinfo);		\
146	if (ct == NULL)							\
147		goto fallback;						\
148	ct->tuplehash[CTINFO2DIR(ctinfo)].tuple.member;			\
149})
150#else
151#define CTTUPLE(skb, member)						\
152({									\
153	goto fallback;							\
154	0;								\
155})
156#endif
157
158static u32 flow_get_nfct_src(const struct sk_buff *skb,
159			     const struct flow_keys *flow)
160{
161	switch (tc_skb_protocol(skb)) {
162	case htons(ETH_P_IP):
163		return ntohl(CTTUPLE(skb, src.u3.ip));
164	case htons(ETH_P_IPV6):
165		return ntohl(CTTUPLE(skb, src.u3.ip6[3]));
166	}
167fallback:
168	return flow_get_src(skb, flow);
169}
170
171static u32 flow_get_nfct_dst(const struct sk_buff *skb,
172			     const struct flow_keys *flow)
173{
174	switch (tc_skb_protocol(skb)) {
175	case htons(ETH_P_IP):
176		return ntohl(CTTUPLE(skb, dst.u3.ip));
177	case htons(ETH_P_IPV6):
178		return ntohl(CTTUPLE(skb, dst.u3.ip6[3]));
179	}
180fallback:
181	return flow_get_dst(skb, flow);
182}
183
184static u32 flow_get_nfct_proto_src(const struct sk_buff *skb,
185				   const struct flow_keys *flow)
186{
187	return ntohs(CTTUPLE(skb, src.u.all));
188fallback:
189	return flow_get_proto_src(skb, flow);
190}
191
192static u32 flow_get_nfct_proto_dst(const struct sk_buff *skb,
193				   const struct flow_keys *flow)
194{
195	return ntohs(CTTUPLE(skb, dst.u.all));
196fallback:
197	return flow_get_proto_dst(skb, flow);
198}
199
200static u32 flow_get_rtclassid(const struct sk_buff *skb)
201{
202#ifdef CONFIG_IP_ROUTE_CLASSID
203	if (skb_dst(skb))
204		return skb_dst(skb)->tclassid;
205#endif
206	return 0;
207}
208
209static u32 flow_get_skuid(const struct sk_buff *skb)
210{
211	struct sock *sk = skb_to_full_sk(skb);
212
213	if (sk && sk->sk_socket && sk->sk_socket->file) {
214		kuid_t skuid = sk->sk_socket->file->f_cred->fsuid;
215
216		return from_kuid(&init_user_ns, skuid);
217	}
218	return 0;
219}
220
221static u32 flow_get_skgid(const struct sk_buff *skb)
222{
223	struct sock *sk = skb_to_full_sk(skb);
224
225	if (sk && sk->sk_socket && sk->sk_socket->file) {
226		kgid_t skgid = sk->sk_socket->file->f_cred->fsgid;
227
228		return from_kgid(&init_user_ns, skgid);
229	}
230	return 0;
231}
232
233static u32 flow_get_vlan_tag(const struct sk_buff *skb)
234{
235	u16 uninitialized_var(tag);
236
237	if (vlan_get_tag(skb, &tag) < 0)
238		return 0;
239	return tag & VLAN_VID_MASK;
240}
241
242static u32 flow_get_rxhash(struct sk_buff *skb)
243{
244	return skb_get_hash(skb);
245}
246
247static u32 flow_key_get(struct sk_buff *skb, int key, struct flow_keys *flow)
248{
249	switch (key) {
250	case FLOW_KEY_SRC:
251		return flow_get_src(skb, flow);
252	case FLOW_KEY_DST:
253		return flow_get_dst(skb, flow);
254	case FLOW_KEY_PROTO:
255		return flow_get_proto(skb, flow);
256	case FLOW_KEY_PROTO_SRC:
257		return flow_get_proto_src(skb, flow);
258	case FLOW_KEY_PROTO_DST:
259		return flow_get_proto_dst(skb, flow);
260	case FLOW_KEY_IIF:
261		return flow_get_iif(skb);
262	case FLOW_KEY_PRIORITY:
263		return flow_get_priority(skb);
264	case FLOW_KEY_MARK:
265		return flow_get_mark(skb);
266	case FLOW_KEY_NFCT:
267		return flow_get_nfct(skb);
268	case FLOW_KEY_NFCT_SRC:
269		return flow_get_nfct_src(skb, flow);
270	case FLOW_KEY_NFCT_DST:
271		return flow_get_nfct_dst(skb, flow);
272	case FLOW_KEY_NFCT_PROTO_SRC:
273		return flow_get_nfct_proto_src(skb, flow);
274	case FLOW_KEY_NFCT_PROTO_DST:
275		return flow_get_nfct_proto_dst(skb, flow);
276	case FLOW_KEY_RTCLASSID:
277		return flow_get_rtclassid(skb);
278	case FLOW_KEY_SKUID:
279		return flow_get_skuid(skb);
280	case FLOW_KEY_SKGID:
281		return flow_get_skgid(skb);
282	case FLOW_KEY_VLAN_TAG:
283		return flow_get_vlan_tag(skb);
284	case FLOW_KEY_RXHASH:
285		return flow_get_rxhash(skb);
286	default:
287		WARN_ON(1);
288		return 0;
289	}
290}
291
292#define FLOW_KEYS_NEEDED ((1 << FLOW_KEY_SRC) | 		\
293			  (1 << FLOW_KEY_DST) |			\
294			  (1 << FLOW_KEY_PROTO) |		\
295			  (1 << FLOW_KEY_PROTO_SRC) |		\
296			  (1 << FLOW_KEY_PROTO_DST) | 		\
297			  (1 << FLOW_KEY_NFCT_SRC) |		\
298			  (1 << FLOW_KEY_NFCT_DST) |		\
299			  (1 << FLOW_KEY_NFCT_PROTO_SRC) |	\
300			  (1 << FLOW_KEY_NFCT_PROTO_DST))
301
302static int flow_classify(struct sk_buff *skb, const struct tcf_proto *tp,
303			 struct tcf_result *res)
 
304{
305	struct flow_head *head = rcu_dereference_bh(tp->root);
306	struct flow_filter *f;
307	u32 keymask;
308	u32 classid;
309	unsigned int n, key;
310	int r;
311
312	list_for_each_entry_rcu(f, &head->filters, list) {
313		u32 keys[FLOW_KEY_MAX + 1];
314		struct flow_keys flow_keys;
315
316		if (!tcf_em_tree_match(skb, &f->ematches, NULL))
317			continue;
318
319		keymask = f->keymask;
320		if (keymask & FLOW_KEYS_NEEDED)
321			skb_flow_dissect_flow_keys(skb, &flow_keys, 0);
322
323		for (n = 0; n < f->nkeys; n++) {
324			key = ffs(keymask) - 1;
325			keymask &= ~(1 << key);
326			keys[n] = flow_key_get(skb, key, &flow_keys);
327		}
328
329		if (f->mode == FLOW_MODE_HASH)
330			classid = jhash2(keys, f->nkeys, f->hashrnd);
331		else {
332			classid = keys[0];
333			classid = (classid & f->mask) ^ f->xor;
334			classid = (classid >> f->rshift) + f->addend;
335		}
336
337		if (f->divisor)
338			classid %= f->divisor;
339
340		res->class   = 0;
341		res->classid = TC_H_MAKE(f->baseclass, f->baseclass + classid);
342
343		r = tcf_exts_exec(skb, &f->exts, res);
344		if (r < 0)
345			continue;
346		return r;
347	}
348	return -1;
349}
350
351static void flow_perturbation(struct timer_list *t)
352{
353	struct flow_filter *f = from_timer(f, t, perturb_timer);
354
355	get_random_bytes(&f->hashrnd, 4);
356	if (f->perturb_period)
357		mod_timer(&f->perturb_timer, jiffies + f->perturb_period);
358}
359
360static const struct nla_policy flow_policy[TCA_FLOW_MAX + 1] = {
361	[TCA_FLOW_KEYS]		= { .type = NLA_U32 },
362	[TCA_FLOW_MODE]		= { .type = NLA_U32 },
363	[TCA_FLOW_BASECLASS]	= { .type = NLA_U32 },
364	[TCA_FLOW_RSHIFT]	= { .type = NLA_U32 },
 
365	[TCA_FLOW_ADDEND]	= { .type = NLA_U32 },
366	[TCA_FLOW_MASK]		= { .type = NLA_U32 },
367	[TCA_FLOW_XOR]		= { .type = NLA_U32 },
368	[TCA_FLOW_DIVISOR]	= { .type = NLA_U32 },
369	[TCA_FLOW_ACT]		= { .type = NLA_NESTED },
370	[TCA_FLOW_POLICE]	= { .type = NLA_NESTED },
371	[TCA_FLOW_EMATCHES]	= { .type = NLA_NESTED },
372	[TCA_FLOW_PERTURB]	= { .type = NLA_U32 },
373};
374
375static void __flow_destroy_filter(struct flow_filter *f)
376{
377	del_timer_sync(&f->perturb_timer);
378	tcf_exts_destroy(&f->exts);
379	tcf_em_tree_destroy(&f->ematches);
380	tcf_exts_put_net(&f->exts);
381	kfree(f);
382}
383
384static void flow_destroy_filter_work(struct work_struct *work)
385{
386	struct flow_filter *f = container_of(work, struct flow_filter, work);
387
 
388	rtnl_lock();
389	__flow_destroy_filter(f);
390	rtnl_unlock();
391}
392
393static void flow_destroy_filter(struct rcu_head *head)
394{
395	struct flow_filter *f = container_of(head, struct flow_filter, rcu);
396
397	INIT_WORK(&f->work, flow_destroy_filter_work);
398	tcf_queue_work(&f->work);
399}
400
401static int flow_change(struct net *net, struct sk_buff *in_skb,
402		       struct tcf_proto *tp, unsigned long base,
403		       u32 handle, struct nlattr **tca,
404		       void **arg, bool ovr, struct netlink_ext_ack *extack)
 
405{
406	struct flow_head *head = rtnl_dereference(tp->root);
407	struct flow_filter *fold, *fnew;
408	struct nlattr *opt = tca[TCA_OPTIONS];
409	struct nlattr *tb[TCA_FLOW_MAX + 1];
410	unsigned int nkeys = 0;
411	unsigned int perturb_period = 0;
412	u32 baseclass = 0;
413	u32 keymask = 0;
414	u32 mode;
415	int err;
416
417	if (opt == NULL)
418		return -EINVAL;
419
420	err = nla_parse_nested(tb, TCA_FLOW_MAX, opt, flow_policy, NULL);
 
421	if (err < 0)
422		return err;
423
424	if (tb[TCA_FLOW_BASECLASS]) {
425		baseclass = nla_get_u32(tb[TCA_FLOW_BASECLASS]);
426		if (TC_H_MIN(baseclass) == 0)
427			return -EINVAL;
428	}
429
430	if (tb[TCA_FLOW_KEYS]) {
431		keymask = nla_get_u32(tb[TCA_FLOW_KEYS]);
432
433		nkeys = hweight32(keymask);
434		if (nkeys == 0)
435			return -EINVAL;
436
437		if (fls(keymask) - 1 > FLOW_KEY_MAX)
438			return -EOPNOTSUPP;
439
440		if ((keymask & (FLOW_KEY_SKUID|FLOW_KEY_SKGID)) &&
441		    sk_user_ns(NETLINK_CB(in_skb).sk) != &init_user_ns)
442			return -EOPNOTSUPP;
443	}
444
445	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
446	if (!fnew)
447		return -ENOBUFS;
448
449	err = tcf_em_tree_validate(tp, tb[TCA_FLOW_EMATCHES], &fnew->ematches);
450	if (err < 0)
451		goto err1;
452
453	err = tcf_exts_init(&fnew->exts, TCA_FLOW_ACT, TCA_FLOW_POLICE);
454	if (err < 0)
455		goto err2;
456
457	err = tcf_exts_validate(net, tp, tb, tca[TCA_RATE], &fnew->exts, ovr,
458				extack);
459	if (err < 0)
460		goto err2;
461
462	fold = *arg;
463	if (fold) {
464		err = -EINVAL;
465		if (fold->handle != handle && handle)
466			goto err2;
467
468		/* Copy fold into fnew */
469		fnew->tp = fold->tp;
470		fnew->handle = fold->handle;
471		fnew->nkeys = fold->nkeys;
472		fnew->keymask = fold->keymask;
473		fnew->mode = fold->mode;
474		fnew->mask = fold->mask;
475		fnew->xor = fold->xor;
476		fnew->rshift = fold->rshift;
477		fnew->addend = fold->addend;
478		fnew->divisor = fold->divisor;
479		fnew->baseclass = fold->baseclass;
480		fnew->hashrnd = fold->hashrnd;
481
482		mode = fold->mode;
483		if (tb[TCA_FLOW_MODE])
484			mode = nla_get_u32(tb[TCA_FLOW_MODE]);
485		if (mode != FLOW_MODE_HASH && nkeys > 1)
486			goto err2;
487
488		if (mode == FLOW_MODE_HASH)
489			perturb_period = fold->perturb_period;
490		if (tb[TCA_FLOW_PERTURB]) {
491			if (mode != FLOW_MODE_HASH)
492				goto err2;
493			perturb_period = nla_get_u32(tb[TCA_FLOW_PERTURB]) * HZ;
494		}
495	} else {
496		err = -EINVAL;
497		if (!handle)
498			goto err2;
499		if (!tb[TCA_FLOW_KEYS])
500			goto err2;
501
502		mode = FLOW_MODE_MAP;
503		if (tb[TCA_FLOW_MODE])
504			mode = nla_get_u32(tb[TCA_FLOW_MODE]);
505		if (mode != FLOW_MODE_HASH && nkeys > 1)
506			goto err2;
507
508		if (tb[TCA_FLOW_PERTURB]) {
509			if (mode != FLOW_MODE_HASH)
510				goto err2;
511			perturb_period = nla_get_u32(tb[TCA_FLOW_PERTURB]) * HZ;
512		}
513
514		if (TC_H_MAJ(baseclass) == 0) {
515			struct Qdisc *q = tcf_block_q(tp->chain->block);
516
517			baseclass = TC_H_MAKE(q->handle, baseclass);
518		}
519		if (TC_H_MIN(baseclass) == 0)
520			baseclass = TC_H_MAKE(baseclass, 1);
521
522		fnew->handle = handle;
523		fnew->mask  = ~0U;
524		fnew->tp = tp;
525		get_random_bytes(&fnew->hashrnd, 4);
526	}
527
528	timer_setup(&fnew->perturb_timer, flow_perturbation, TIMER_DEFERRABLE);
529
530	tcf_block_netif_keep_dst(tp->chain->block);
531
532	if (tb[TCA_FLOW_KEYS]) {
533		fnew->keymask = keymask;
534		fnew->nkeys   = nkeys;
535	}
536
537	fnew->mode = mode;
538
539	if (tb[TCA_FLOW_MASK])
540		fnew->mask = nla_get_u32(tb[TCA_FLOW_MASK]);
541	if (tb[TCA_FLOW_XOR])
542		fnew->xor = nla_get_u32(tb[TCA_FLOW_XOR]);
543	if (tb[TCA_FLOW_RSHIFT])
544		fnew->rshift = nla_get_u32(tb[TCA_FLOW_RSHIFT]);
545	if (tb[TCA_FLOW_ADDEND])
546		fnew->addend = nla_get_u32(tb[TCA_FLOW_ADDEND]);
547
548	if (tb[TCA_FLOW_DIVISOR])
549		fnew->divisor = nla_get_u32(tb[TCA_FLOW_DIVISOR]);
550	if (baseclass)
551		fnew->baseclass = baseclass;
552
553	fnew->perturb_period = perturb_period;
554	if (perturb_period)
555		mod_timer(&fnew->perturb_timer, jiffies + perturb_period);
556
557	if (!*arg)
558		list_add_tail_rcu(&fnew->list, &head->filters);
559	else
560		list_replace_rcu(&fold->list, &fnew->list);
561
562	*arg = fnew;
563
564	if (fold) {
565		tcf_exts_get_net(&fold->exts);
566		call_rcu(&fold->rcu, flow_destroy_filter);
567	}
568	return 0;
569
570err2:
571	tcf_exts_destroy(&fnew->exts);
572	tcf_em_tree_destroy(&fnew->ematches);
573err1:
574	kfree(fnew);
575	return err;
576}
577
578static int flow_delete(struct tcf_proto *tp, void *arg, bool *last,
579		       struct netlink_ext_ack *extack)
580{
581	struct flow_head *head = rtnl_dereference(tp->root);
582	struct flow_filter *f = arg;
583
584	list_del_rcu(&f->list);
585	tcf_exts_get_net(&f->exts);
586	call_rcu(&f->rcu, flow_destroy_filter);
587	*last = list_empty(&head->filters);
588	return 0;
589}
590
591static int flow_init(struct tcf_proto *tp)
592{
593	struct flow_head *head;
594
595	head = kzalloc(sizeof(*head), GFP_KERNEL);
596	if (head == NULL)
597		return -ENOBUFS;
598	INIT_LIST_HEAD(&head->filters);
599	rcu_assign_pointer(tp->root, head);
600	return 0;
601}
602
603static void flow_destroy(struct tcf_proto *tp, struct netlink_ext_ack *extack)
 
604{
605	struct flow_head *head = rtnl_dereference(tp->root);
606	struct flow_filter *f, *next;
607
608	list_for_each_entry_safe(f, next, &head->filters, list) {
609		list_del_rcu(&f->list);
610		if (tcf_exts_get_net(&f->exts))
611			call_rcu(&f->rcu, flow_destroy_filter);
612		else
613			__flow_destroy_filter(f);
614	}
615	kfree_rcu(head, rcu);
616}
617
618static void *flow_get(struct tcf_proto *tp, u32 handle)
619{
620	struct flow_head *head = rtnl_dereference(tp->root);
621	struct flow_filter *f;
622
623	list_for_each_entry(f, &head->filters, list)
624		if (f->handle == handle)
625			return f;
626	return NULL;
627}
628
629static int flow_dump(struct net *net, struct tcf_proto *tp, void *fh,
630		     struct sk_buff *skb, struct tcmsg *t)
631{
632	struct flow_filter *f = fh;
633	struct nlattr *nest;
634
635	if (f == NULL)
636		return skb->len;
637
638	t->tcm_handle = f->handle;
639
640	nest = nla_nest_start(skb, TCA_OPTIONS);
641	if (nest == NULL)
642		goto nla_put_failure;
643
644	if (nla_put_u32(skb, TCA_FLOW_KEYS, f->keymask) ||
645	    nla_put_u32(skb, TCA_FLOW_MODE, f->mode))
646		goto nla_put_failure;
647
648	if (f->mask != ~0 || f->xor != 0) {
649		if (nla_put_u32(skb, TCA_FLOW_MASK, f->mask) ||
650		    nla_put_u32(skb, TCA_FLOW_XOR, f->xor))
651			goto nla_put_failure;
652	}
653	if (f->rshift &&
654	    nla_put_u32(skb, TCA_FLOW_RSHIFT, f->rshift))
655		goto nla_put_failure;
656	if (f->addend &&
657	    nla_put_u32(skb, TCA_FLOW_ADDEND, f->addend))
658		goto nla_put_failure;
659
660	if (f->divisor &&
661	    nla_put_u32(skb, TCA_FLOW_DIVISOR, f->divisor))
662		goto nla_put_failure;
663	if (f->baseclass &&
664	    nla_put_u32(skb, TCA_FLOW_BASECLASS, f->baseclass))
665		goto nla_put_failure;
666
667	if (f->perturb_period &&
668	    nla_put_u32(skb, TCA_FLOW_PERTURB, f->perturb_period / HZ))
669		goto nla_put_failure;
670
671	if (tcf_exts_dump(skb, &f->exts) < 0)
672		goto nla_put_failure;
673#ifdef CONFIG_NET_EMATCH
674	if (f->ematches.hdr.nmatches &&
675	    tcf_em_tree_dump(skb, &f->ematches, TCA_FLOW_EMATCHES) < 0)
676		goto nla_put_failure;
677#endif
678	nla_nest_end(skb, nest);
679
680	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
681		goto nla_put_failure;
682
683	return skb->len;
684
685nla_put_failure:
686	nla_nest_cancel(skb, nest);
687	return -1;
688}
689
690static void flow_walk(struct tcf_proto *tp, struct tcf_walker *arg)
 
691{
692	struct flow_head *head = rtnl_dereference(tp->root);
693	struct flow_filter *f;
694
695	list_for_each_entry(f, &head->filters, list) {
696		if (arg->count < arg->skip)
697			goto skip;
698		if (arg->fn(tp, f, arg) < 0) {
699			arg->stop = 1;
700			break;
701		}
702skip:
703		arg->count++;
704	}
705}
706
707static struct tcf_proto_ops cls_flow_ops __read_mostly = {
708	.kind		= "flow",
709	.classify	= flow_classify,
710	.init		= flow_init,
711	.destroy	= flow_destroy,
712	.change		= flow_change,
713	.delete		= flow_delete,
714	.get		= flow_get,
715	.dump		= flow_dump,
716	.walk		= flow_walk,
717	.owner		= THIS_MODULE,
718};
 
719
720static int __init cls_flow_init(void)
721{
722	return register_tcf_proto_ops(&cls_flow_ops);
723}
724
725static void __exit cls_flow_exit(void)
726{
727	unregister_tcf_proto_ops(&cls_flow_ops);
728}
729
730module_init(cls_flow_init);
731module_exit(cls_flow_exit);
732
733MODULE_LICENSE("GPL");
734MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
735MODULE_DESCRIPTION("TC flow classifier");