Linux Audio

Check our new training course

Loading...
v4.6
  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 defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 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	struct rcu_head		rcu;
 61};
 62
 63static inline u32 addr_fold(void *addr)
 64{
 65	unsigned long a = (unsigned long)addr;
 66
 67	return (a & 0xFFFFFFFF) ^ (BITS_PER_LONG > 32 ? a >> 32 : 0);
 68}
 69
 70static u32 flow_get_src(const struct sk_buff *skb, const struct flow_keys *flow)
 71{
 72	__be32 src = flow_get_u32_src(flow);
 73
 74	if (src)
 75		return ntohl(src);
 76
 77	return addr_fold(skb->sk);
 78}
 79
 80static u32 flow_get_dst(const struct sk_buff *skb, const struct flow_keys *flow)
 81{
 82	__be32 dst = flow_get_u32_dst(flow);
 83
 84	if (dst)
 85		return ntohl(dst);
 86
 87	return addr_fold(skb_dst(skb)) ^ (__force u16) tc_skb_protocol(skb);
 88}
 89
 90static u32 flow_get_proto(const struct sk_buff *skb, const struct flow_keys *flow)
 
 91{
 92	return flow->basic.ip_proto;
 93}
 94
 95static u32 flow_get_proto_src(const struct sk_buff *skb, const struct flow_keys *flow)
 
 96{
 97	if (flow->ports.ports)
 98		return ntohs(flow->ports.src);
 99
100	return addr_fold(skb->sk);
101}
102
103static u32 flow_get_proto_dst(const struct sk_buff *skb, 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) tc_skb_protocol(skb);
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 defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
129	return addr_fold(skb->nfct);
130#else
131	return 0;
132#endif
133}
134
135#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
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, const struct flow_keys *flow)
 
153{
154	switch (tc_skb_protocol(skb)) {
155	case htons(ETH_P_IP):
156		return ntohl(CTTUPLE(skb, src.u3.ip));
157	case htons(ETH_P_IPV6):
158		return ntohl(CTTUPLE(skb, src.u3.ip6[3]));
159	}
160fallback:
161	return flow_get_src(skb, flow);
162}
163
164static u32 flow_get_nfct_dst(const struct sk_buff *skb, const struct flow_keys *flow)
 
165{
166	switch (tc_skb_protocol(skb)) {
167	case htons(ETH_P_IP):
168		return ntohl(CTTUPLE(skb, dst.u3.ip));
169	case htons(ETH_P_IPV6):
170		return ntohl(CTTUPLE(skb, dst.u3.ip6[3]));
171	}
172fallback:
173	return flow_get_dst(skb, flow);
174}
175
176static u32 flow_get_nfct_proto_src(const struct sk_buff *skb, const struct flow_keys *flow)
 
177{
178	return ntohs(CTTUPLE(skb, src.u.all));
179fallback:
180	return flow_get_proto_src(skb, flow);
181}
182
183static u32 flow_get_nfct_proto_dst(const struct sk_buff *skb, const struct flow_keys *flow)
 
184{
185	return ntohs(CTTUPLE(skb, dst.u.all));
186fallback:
187	return flow_get_proto_dst(skb, flow);
188}
189
190static u32 flow_get_rtclassid(const struct sk_buff *skb)
191{
192#ifdef CONFIG_IP_ROUTE_CLASSID
193	if (skb_dst(skb))
194		return skb_dst(skb)->tclassid;
195#endif
196	return 0;
197}
198
199static u32 flow_get_skuid(const struct sk_buff *skb)
200{
201	struct sock *sk = skb_to_full_sk(skb);
202
203	if (sk && sk->sk_socket && sk->sk_socket->file) {
204		kuid_t skuid = sk->sk_socket->file->f_cred->fsuid;
205
206		return from_kuid(&init_user_ns, skuid);
207	}
208	return 0;
209}
210
211static u32 flow_get_skgid(const struct sk_buff *skb)
212{
213	struct sock *sk = skb_to_full_sk(skb);
214
215	if (sk && sk->sk_socket && sk->sk_socket->file) {
216		kgid_t skgid = sk->sk_socket->file->f_cred->fsgid;
217
218		return from_kgid(&init_user_ns, skgid);
219	}
220	return 0;
221}
222
223static u32 flow_get_vlan_tag(const struct sk_buff *skb)
224{
225	u16 uninitialized_var(tag);
226
227	if (vlan_get_tag(skb, &tag) < 0)
228		return 0;
229	return tag & VLAN_VID_MASK;
230}
231
232static u32 flow_get_rxhash(struct sk_buff *skb)
233{
234	return skb_get_hash(skb);
235}
236
237static u32 flow_key_get(struct sk_buff *skb, int key, struct flow_keys *flow)
238{
239	switch (key) {
240	case FLOW_KEY_SRC:
241		return flow_get_src(skb, flow);
242	case FLOW_KEY_DST:
243		return flow_get_dst(skb, flow);
244	case FLOW_KEY_PROTO:
245		return flow_get_proto(skb, flow);
246	case FLOW_KEY_PROTO_SRC:
247		return flow_get_proto_src(skb, flow);
248	case FLOW_KEY_PROTO_DST:
249		return flow_get_proto_dst(skb, flow);
250	case FLOW_KEY_IIF:
251		return flow_get_iif(skb);
252	case FLOW_KEY_PRIORITY:
253		return flow_get_priority(skb);
254	case FLOW_KEY_MARK:
255		return flow_get_mark(skb);
256	case FLOW_KEY_NFCT:
257		return flow_get_nfct(skb);
258	case FLOW_KEY_NFCT_SRC:
259		return flow_get_nfct_src(skb, flow);
260	case FLOW_KEY_NFCT_DST:
261		return flow_get_nfct_dst(skb, flow);
262	case FLOW_KEY_NFCT_PROTO_SRC:
263		return flow_get_nfct_proto_src(skb, flow);
264	case FLOW_KEY_NFCT_PROTO_DST:
265		return flow_get_nfct_proto_dst(skb, flow);
266	case FLOW_KEY_RTCLASSID:
267		return flow_get_rtclassid(skb);
268	case FLOW_KEY_SKUID:
269		return flow_get_skuid(skb);
270	case FLOW_KEY_SKGID:
271		return flow_get_skgid(skb);
272	case FLOW_KEY_VLAN_TAG:
273		return flow_get_vlan_tag(skb);
274	case FLOW_KEY_RXHASH:
275		return flow_get_rxhash(skb);
276	default:
277		WARN_ON(1);
278		return 0;
279	}
280}
281
282#define FLOW_KEYS_NEEDED ((1 << FLOW_KEY_SRC) | 		\
283			  (1 << FLOW_KEY_DST) |			\
284			  (1 << FLOW_KEY_PROTO) |		\
285			  (1 << FLOW_KEY_PROTO_SRC) |		\
286			  (1 << FLOW_KEY_PROTO_DST) | 		\
287			  (1 << FLOW_KEY_NFCT_SRC) |		\
288			  (1 << FLOW_KEY_NFCT_DST) |		\
289			  (1 << FLOW_KEY_NFCT_PROTO_SRC) |	\
290			  (1 << FLOW_KEY_NFCT_PROTO_DST))
291
292static int flow_classify(struct sk_buff *skb, const struct tcf_proto *tp,
293			 struct tcf_result *res)
294{
295	struct flow_head *head = rcu_dereference_bh(tp->root);
296	struct flow_filter *f;
297	u32 keymask;
298	u32 classid;
299	unsigned int n, key;
300	int r;
301
302	list_for_each_entry_rcu(f, &head->filters, list) {
303		u32 keys[FLOW_KEY_MAX + 1];
304		struct flow_keys flow_keys;
305
306		if (!tcf_em_tree_match(skb, &f->ematches, NULL))
307			continue;
308
309		keymask = f->keymask;
310		if (keymask & FLOW_KEYS_NEEDED)
311			skb_flow_dissect_flow_keys(skb, &flow_keys, 0);
312
313		for (n = 0; n < f->nkeys; n++) {
314			key = ffs(keymask) - 1;
315			keymask &= ~(1 << key);
316			keys[n] = flow_key_get(skb, key, &flow_keys);
317		}
318
319		if (f->mode == FLOW_MODE_HASH)
320			classid = jhash2(keys, f->nkeys, f->hashrnd);
321		else {
322			classid = keys[0];
323			classid = (classid & f->mask) ^ f->xor;
324			classid = (classid >> f->rshift) + f->addend;
325		}
326
327		if (f->divisor)
328			classid %= f->divisor;
329
330		res->class   = 0;
331		res->classid = TC_H_MAKE(f->baseclass, f->baseclass + classid);
332
333		r = tcf_exts_exec(skb, &f->exts, res);
334		if (r < 0)
335			continue;
336		return r;
337	}
338	return -1;
339}
340
341static void flow_perturbation(unsigned long arg)
342{
343	struct flow_filter *f = (struct flow_filter *)arg;
344
345	get_random_bytes(&f->hashrnd, 4);
346	if (f->perturb_period)
347		mod_timer(&f->perturb_timer, jiffies + f->perturb_period);
348}
349
350static const struct nla_policy flow_policy[TCA_FLOW_MAX + 1] = {
351	[TCA_FLOW_KEYS]		= { .type = NLA_U32 },
352	[TCA_FLOW_MODE]		= { .type = NLA_U32 },
353	[TCA_FLOW_BASECLASS]	= { .type = NLA_U32 },
354	[TCA_FLOW_RSHIFT]	= { .type = NLA_U32 },
355	[TCA_FLOW_ADDEND]	= { .type = NLA_U32 },
356	[TCA_FLOW_MASK]		= { .type = NLA_U32 },
357	[TCA_FLOW_XOR]		= { .type = NLA_U32 },
358	[TCA_FLOW_DIVISOR]	= { .type = NLA_U32 },
359	[TCA_FLOW_ACT]		= { .type = NLA_NESTED },
360	[TCA_FLOW_POLICE]	= { .type = NLA_NESTED },
361	[TCA_FLOW_EMATCHES]	= { .type = NLA_NESTED },
362	[TCA_FLOW_PERTURB]	= { .type = NLA_U32 },
363};
364
365static void flow_destroy_filter(struct rcu_head *head)
366{
367	struct flow_filter *f = container_of(head, struct flow_filter, rcu);
368
369	del_timer_sync(&f->perturb_timer);
370	tcf_exts_destroy(&f->exts);
371	tcf_em_tree_destroy(&f->ematches);
372	kfree(f);
373}
374
375static int flow_change(struct net *net, struct sk_buff *in_skb,
376		       struct tcf_proto *tp, unsigned long base,
377		       u32 handle, struct nlattr **tca,
378		       unsigned long *arg, bool ovr)
379{
380	struct flow_head *head = rtnl_dereference(tp->root);
381	struct flow_filter *fold, *fnew;
382	struct nlattr *opt = tca[TCA_OPTIONS];
383	struct nlattr *tb[TCA_FLOW_MAX + 1];
384	struct tcf_exts e;
385	struct tcf_ematch_tree t;
386	unsigned int nkeys = 0;
387	unsigned int perturb_period = 0;
388	u32 baseclass = 0;
389	u32 keymask = 0;
390	u32 mode;
391	int err;
392
393	if (opt == NULL)
394		return -EINVAL;
395
396	err = nla_parse_nested(tb, TCA_FLOW_MAX, opt, flow_policy);
397	if (err < 0)
398		return err;
399
400	if (tb[TCA_FLOW_BASECLASS]) {
401		baseclass = nla_get_u32(tb[TCA_FLOW_BASECLASS]);
402		if (TC_H_MIN(baseclass) == 0)
403			return -EINVAL;
404	}
405
406	if (tb[TCA_FLOW_KEYS]) {
407		keymask = nla_get_u32(tb[TCA_FLOW_KEYS]);
408
409		nkeys = hweight32(keymask);
410		if (nkeys == 0)
411			return -EINVAL;
412
413		if (fls(keymask) - 1 > FLOW_KEY_MAX)
414			return -EOPNOTSUPP;
415
416		if ((keymask & (FLOW_KEY_SKUID|FLOW_KEY_SKGID)) &&
417		    sk_user_ns(NETLINK_CB(in_skb).sk) != &init_user_ns)
418			return -EOPNOTSUPP;
419	}
420
421	tcf_exts_init(&e, TCA_FLOW_ACT, TCA_FLOW_POLICE);
 
 
422	err = tcf_exts_validate(net, tp, tb, tca[TCA_RATE], &e, ovr);
423	if (err < 0)
424		return err;
425
426	err = tcf_em_tree_validate(tp, tb[TCA_FLOW_EMATCHES], &t);
427	if (err < 0)
428		goto err1;
429
430	err = -ENOBUFS;
431	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
432	if (!fnew)
433		goto err2;
434
435	tcf_exts_init(&fnew->exts, TCA_FLOW_ACT, TCA_FLOW_POLICE);
 
 
436
437	fold = (struct flow_filter *)*arg;
438	if (fold) {
439		err = -EINVAL;
440		if (fold->handle != handle && handle)
441			goto err2;
442
443		/* Copy fold into fnew */
444		fnew->tp = fold->tp;
445		fnew->handle = fold->handle;
446		fnew->nkeys = fold->nkeys;
447		fnew->keymask = fold->keymask;
448		fnew->mode = fold->mode;
449		fnew->mask = fold->mask;
450		fnew->xor = fold->xor;
451		fnew->rshift = fold->rshift;
452		fnew->addend = fold->addend;
453		fnew->divisor = fold->divisor;
454		fnew->baseclass = fold->baseclass;
455		fnew->hashrnd = fold->hashrnd;
456
457		mode = fold->mode;
458		if (tb[TCA_FLOW_MODE])
459			mode = nla_get_u32(tb[TCA_FLOW_MODE]);
460		if (mode != FLOW_MODE_HASH && nkeys > 1)
461			goto err2;
462
463		if (mode == FLOW_MODE_HASH)
464			perturb_period = fold->perturb_period;
465		if (tb[TCA_FLOW_PERTURB]) {
466			if (mode != FLOW_MODE_HASH)
467				goto err2;
468			perturb_period = nla_get_u32(tb[TCA_FLOW_PERTURB]) * HZ;
469		}
470	} else {
471		err = -EINVAL;
472		if (!handle)
473			goto err2;
474		if (!tb[TCA_FLOW_KEYS])
475			goto err2;
476
477		mode = FLOW_MODE_MAP;
478		if (tb[TCA_FLOW_MODE])
479			mode = nla_get_u32(tb[TCA_FLOW_MODE]);
480		if (mode != FLOW_MODE_HASH && nkeys > 1)
481			goto err2;
482
483		if (tb[TCA_FLOW_PERTURB]) {
484			if (mode != FLOW_MODE_HASH)
485				goto err2;
486			perturb_period = nla_get_u32(tb[TCA_FLOW_PERTURB]) * HZ;
487		}
488
489		if (TC_H_MAJ(baseclass) == 0)
490			baseclass = TC_H_MAKE(tp->q->handle, baseclass);
491		if (TC_H_MIN(baseclass) == 0)
492			baseclass = TC_H_MAKE(baseclass, 1);
493
494		fnew->handle = handle;
495		fnew->mask  = ~0U;
496		fnew->tp = tp;
497		get_random_bytes(&fnew->hashrnd, 4);
498	}
499
500	fnew->perturb_timer.function = flow_perturbation;
501	fnew->perturb_timer.data = (unsigned long)fnew;
502	init_timer_deferrable(&fnew->perturb_timer);
503
504	tcf_exts_change(tp, &fnew->exts, &e);
505	tcf_em_tree_change(tp, &fnew->ematches, &t);
506
507	netif_keep_dst(qdisc_dev(tp->q));
508
509	if (tb[TCA_FLOW_KEYS]) {
510		fnew->keymask = keymask;
511		fnew->nkeys   = nkeys;
512	}
513
514	fnew->mode = mode;
515
516	if (tb[TCA_FLOW_MASK])
517		fnew->mask = nla_get_u32(tb[TCA_FLOW_MASK]);
518	if (tb[TCA_FLOW_XOR])
519		fnew->xor = nla_get_u32(tb[TCA_FLOW_XOR]);
520	if (tb[TCA_FLOW_RSHIFT])
521		fnew->rshift = nla_get_u32(tb[TCA_FLOW_RSHIFT]);
522	if (tb[TCA_FLOW_ADDEND])
523		fnew->addend = nla_get_u32(tb[TCA_FLOW_ADDEND]);
524
525	if (tb[TCA_FLOW_DIVISOR])
526		fnew->divisor = nla_get_u32(tb[TCA_FLOW_DIVISOR]);
527	if (baseclass)
528		fnew->baseclass = baseclass;
529
530	fnew->perturb_period = perturb_period;
531	if (perturb_period)
532		mod_timer(&fnew->perturb_timer, jiffies + perturb_period);
533
534	if (*arg == 0)
535		list_add_tail_rcu(&fnew->list, &head->filters);
536	else
537		list_replace_rcu(&fold->list, &fnew->list);
538
539	*arg = (unsigned long)fnew;
540
541	if (fold)
542		call_rcu(&fold->rcu, flow_destroy_filter);
543	return 0;
544
 
 
545err2:
546	tcf_em_tree_destroy(&t);
547	kfree(fnew);
548err1:
549	tcf_exts_destroy(&e);
550	return err;
551}
552
553static int flow_delete(struct tcf_proto *tp, unsigned long arg)
554{
555	struct flow_filter *f = (struct flow_filter *)arg;
556
557	list_del_rcu(&f->list);
558	call_rcu(&f->rcu, flow_destroy_filter);
559	return 0;
560}
561
562static int flow_init(struct tcf_proto *tp)
563{
564	struct flow_head *head;
565
566	head = kzalloc(sizeof(*head), GFP_KERNEL);
567	if (head == NULL)
568		return -ENOBUFS;
569	INIT_LIST_HEAD(&head->filters);
570	rcu_assign_pointer(tp->root, head);
571	return 0;
572}
573
574static bool flow_destroy(struct tcf_proto *tp, bool force)
575{
576	struct flow_head *head = rtnl_dereference(tp->root);
577	struct flow_filter *f, *next;
578
579	if (!force && !list_empty(&head->filters))
580		return false;
581
582	list_for_each_entry_safe(f, next, &head->filters, list) {
583		list_del_rcu(&f->list);
584		call_rcu(&f->rcu, flow_destroy_filter);
585	}
586	RCU_INIT_POINTER(tp->root, NULL);
587	kfree_rcu(head, rcu);
588	return true;
589}
590
591static unsigned long flow_get(struct tcf_proto *tp, u32 handle)
592{
593	struct flow_head *head = rtnl_dereference(tp->root);
594	struct flow_filter *f;
595
596	list_for_each_entry(f, &head->filters, list)
597		if (f->handle == handle)
598			return (unsigned long)f;
599	return 0;
600}
601
602static int flow_dump(struct net *net, struct tcf_proto *tp, unsigned long fh,
603		     struct sk_buff *skb, struct tcmsg *t)
604{
605	struct flow_filter *f = (struct flow_filter *)fh;
606	struct nlattr *nest;
607
608	if (f == NULL)
609		return skb->len;
610
611	t->tcm_handle = f->handle;
612
613	nest = nla_nest_start(skb, TCA_OPTIONS);
614	if (nest == NULL)
615		goto nla_put_failure;
616
617	if (nla_put_u32(skb, TCA_FLOW_KEYS, f->keymask) ||
618	    nla_put_u32(skb, TCA_FLOW_MODE, f->mode))
619		goto nla_put_failure;
620
621	if (f->mask != ~0 || f->xor != 0) {
622		if (nla_put_u32(skb, TCA_FLOW_MASK, f->mask) ||
623		    nla_put_u32(skb, TCA_FLOW_XOR, f->xor))
624			goto nla_put_failure;
625	}
626	if (f->rshift &&
627	    nla_put_u32(skb, TCA_FLOW_RSHIFT, f->rshift))
628		goto nla_put_failure;
629	if (f->addend &&
630	    nla_put_u32(skb, TCA_FLOW_ADDEND, f->addend))
631		goto nla_put_failure;
632
633	if (f->divisor &&
634	    nla_put_u32(skb, TCA_FLOW_DIVISOR, f->divisor))
635		goto nla_put_failure;
636	if (f->baseclass &&
637	    nla_put_u32(skb, TCA_FLOW_BASECLASS, f->baseclass))
638		goto nla_put_failure;
639
640	if (f->perturb_period &&
641	    nla_put_u32(skb, TCA_FLOW_PERTURB, f->perturb_period / HZ))
642		goto nla_put_failure;
643
644	if (tcf_exts_dump(skb, &f->exts) < 0)
645		goto nla_put_failure;
646#ifdef CONFIG_NET_EMATCH
647	if (f->ematches.hdr.nmatches &&
648	    tcf_em_tree_dump(skb, &f->ematches, TCA_FLOW_EMATCHES) < 0)
649		goto nla_put_failure;
650#endif
651	nla_nest_end(skb, nest);
652
653	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
654		goto nla_put_failure;
655
656	return skb->len;
657
658nla_put_failure:
659	nla_nest_cancel(skb, nest);
660	return -1;
661}
662
663static void flow_walk(struct tcf_proto *tp, struct tcf_walker *arg)
664{
665	struct flow_head *head = rtnl_dereference(tp->root);
666	struct flow_filter *f;
667
668	list_for_each_entry(f, &head->filters, list) {
669		if (arg->count < arg->skip)
670			goto skip;
671		if (arg->fn(tp, (unsigned long)f, arg) < 0) {
672			arg->stop = 1;
673			break;
674		}
675skip:
676		arg->count++;
677	}
678}
679
680static struct tcf_proto_ops cls_flow_ops __read_mostly = {
681	.kind		= "flow",
682	.classify	= flow_classify,
683	.init		= flow_init,
684	.destroy	= flow_destroy,
685	.change		= flow_change,
686	.delete		= flow_delete,
687	.get		= flow_get,
688	.dump		= flow_dump,
689	.walk		= flow_walk,
690	.owner		= THIS_MODULE,
691};
692
693static int __init cls_flow_init(void)
694{
695	return register_tcf_proto_ops(&cls_flow_ops);
696}
697
698static void __exit cls_flow_exit(void)
699{
700	unregister_tcf_proto_ops(&cls_flow_ops);
701}
702
703module_init(cls_flow_init);
704module_exit(cls_flow_exit);
705
706MODULE_LICENSE("GPL");
707MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
708MODULE_DESCRIPTION("TC flow classifier");
v4.10.11
  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	struct rcu_head		rcu;
 61};
 62
 63static inline u32 addr_fold(void *addr)
 64{
 65	unsigned long a = (unsigned long)addr;
 66
 67	return (a & 0xFFFFFFFF) ^ (BITS_PER_LONG > 32 ? a >> 32 : 0);
 68}
 69
 70static u32 flow_get_src(const struct sk_buff *skb, const struct flow_keys *flow)
 71{
 72	__be32 src = flow_get_u32_src(flow);
 73
 74	if (src)
 75		return ntohl(src);
 76
 77	return addr_fold(skb->sk);
 78}
 79
 80static u32 flow_get_dst(const struct sk_buff *skb, const struct flow_keys *flow)
 81{
 82	__be32 dst = flow_get_u32_dst(flow);
 83
 84	if (dst)
 85		return ntohl(dst);
 86
 87	return addr_fold(skb_dst(skb)) ^ (__force u16) tc_skb_protocol(skb);
 88}
 89
 90static u32 flow_get_proto(const struct sk_buff *skb,
 91			  const struct flow_keys *flow)
 92{
 93	return flow->basic.ip_proto;
 94}
 95
 96static u32 flow_get_proto_src(const struct sk_buff *skb,
 97			      const struct flow_keys *flow)
 98{
 99	if (flow->ports.ports)
100		return ntohs(flow->ports.src);
101
102	return addr_fold(skb->sk);
103}
104
105static u32 flow_get_proto_dst(const struct sk_buff *skb,
106			      const struct flow_keys *flow)
107{
108	if (flow->ports.ports)
109		return ntohs(flow->ports.dst);
110
111	return addr_fold(skb_dst(skb)) ^ (__force u16) tc_skb_protocol(skb);
112}
113
114static u32 flow_get_iif(const struct sk_buff *skb)
115{
116	return skb->skb_iif;
117}
118
119static u32 flow_get_priority(const struct sk_buff *skb)
120{
121	return skb->priority;
122}
123
124static u32 flow_get_mark(const struct sk_buff *skb)
125{
126	return skb->mark;
127}
128
129static u32 flow_get_nfct(const struct sk_buff *skb)
130{
131#if IS_ENABLED(CONFIG_NF_CONNTRACK)
132	return addr_fold(skb->nfct);
133#else
134	return 0;
135#endif
136}
137
138#if IS_ENABLED(CONFIG_NF_CONNTRACK)
139#define CTTUPLE(skb, member)						\
140({									\
141	enum ip_conntrack_info ctinfo;					\
142	const struct nf_conn *ct = nf_ct_get(skb, &ctinfo);		\
143	if (ct == NULL)							\
144		goto fallback;						\
145	ct->tuplehash[CTINFO2DIR(ctinfo)].tuple.member;			\
146})
147#else
148#define CTTUPLE(skb, member)						\
149({									\
150	goto fallback;							\
151	0;								\
152})
153#endif
154
155static u32 flow_get_nfct_src(const struct sk_buff *skb,
156			     const struct flow_keys *flow)
157{
158	switch (tc_skb_protocol(skb)) {
159	case htons(ETH_P_IP):
160		return ntohl(CTTUPLE(skb, src.u3.ip));
161	case htons(ETH_P_IPV6):
162		return ntohl(CTTUPLE(skb, src.u3.ip6[3]));
163	}
164fallback:
165	return flow_get_src(skb, flow);
166}
167
168static u32 flow_get_nfct_dst(const struct sk_buff *skb,
169			     const struct flow_keys *flow)
170{
171	switch (tc_skb_protocol(skb)) {
172	case htons(ETH_P_IP):
173		return ntohl(CTTUPLE(skb, dst.u3.ip));
174	case htons(ETH_P_IPV6):
175		return ntohl(CTTUPLE(skb, dst.u3.ip6[3]));
176	}
177fallback:
178	return flow_get_dst(skb, flow);
179}
180
181static u32 flow_get_nfct_proto_src(const struct sk_buff *skb,
182				   const struct flow_keys *flow)
183{
184	return ntohs(CTTUPLE(skb, src.u.all));
185fallback:
186	return flow_get_proto_src(skb, flow);
187}
188
189static u32 flow_get_nfct_proto_dst(const struct sk_buff *skb,
190				   const struct flow_keys *flow)
191{
192	return ntohs(CTTUPLE(skb, dst.u.all));
193fallback:
194	return flow_get_proto_dst(skb, flow);
195}
196
197static u32 flow_get_rtclassid(const struct sk_buff *skb)
198{
199#ifdef CONFIG_IP_ROUTE_CLASSID
200	if (skb_dst(skb))
201		return skb_dst(skb)->tclassid;
202#endif
203	return 0;
204}
205
206static u32 flow_get_skuid(const struct sk_buff *skb)
207{
208	struct sock *sk = skb_to_full_sk(skb);
209
210	if (sk && sk->sk_socket && sk->sk_socket->file) {
211		kuid_t skuid = sk->sk_socket->file->f_cred->fsuid;
212
213		return from_kuid(&init_user_ns, skuid);
214	}
215	return 0;
216}
217
218static u32 flow_get_skgid(const struct sk_buff *skb)
219{
220	struct sock *sk = skb_to_full_sk(skb);
221
222	if (sk && sk->sk_socket && sk->sk_socket->file) {
223		kgid_t skgid = sk->sk_socket->file->f_cred->fsgid;
224
225		return from_kgid(&init_user_ns, skgid);
226	}
227	return 0;
228}
229
230static u32 flow_get_vlan_tag(const struct sk_buff *skb)
231{
232	u16 uninitialized_var(tag);
233
234	if (vlan_get_tag(skb, &tag) < 0)
235		return 0;
236	return tag & VLAN_VID_MASK;
237}
238
239static u32 flow_get_rxhash(struct sk_buff *skb)
240{
241	return skb_get_hash(skb);
242}
243
244static u32 flow_key_get(struct sk_buff *skb, int key, struct flow_keys *flow)
245{
246	switch (key) {
247	case FLOW_KEY_SRC:
248		return flow_get_src(skb, flow);
249	case FLOW_KEY_DST:
250		return flow_get_dst(skb, flow);
251	case FLOW_KEY_PROTO:
252		return flow_get_proto(skb, flow);
253	case FLOW_KEY_PROTO_SRC:
254		return flow_get_proto_src(skb, flow);
255	case FLOW_KEY_PROTO_DST:
256		return flow_get_proto_dst(skb, flow);
257	case FLOW_KEY_IIF:
258		return flow_get_iif(skb);
259	case FLOW_KEY_PRIORITY:
260		return flow_get_priority(skb);
261	case FLOW_KEY_MARK:
262		return flow_get_mark(skb);
263	case FLOW_KEY_NFCT:
264		return flow_get_nfct(skb);
265	case FLOW_KEY_NFCT_SRC:
266		return flow_get_nfct_src(skb, flow);
267	case FLOW_KEY_NFCT_DST:
268		return flow_get_nfct_dst(skb, flow);
269	case FLOW_KEY_NFCT_PROTO_SRC:
270		return flow_get_nfct_proto_src(skb, flow);
271	case FLOW_KEY_NFCT_PROTO_DST:
272		return flow_get_nfct_proto_dst(skb, flow);
273	case FLOW_KEY_RTCLASSID:
274		return flow_get_rtclassid(skb);
275	case FLOW_KEY_SKUID:
276		return flow_get_skuid(skb);
277	case FLOW_KEY_SKGID:
278		return flow_get_skgid(skb);
279	case FLOW_KEY_VLAN_TAG:
280		return flow_get_vlan_tag(skb);
281	case FLOW_KEY_RXHASH:
282		return flow_get_rxhash(skb);
283	default:
284		WARN_ON(1);
285		return 0;
286	}
287}
288
289#define FLOW_KEYS_NEEDED ((1 << FLOW_KEY_SRC) | 		\
290			  (1 << FLOW_KEY_DST) |			\
291			  (1 << FLOW_KEY_PROTO) |		\
292			  (1 << FLOW_KEY_PROTO_SRC) |		\
293			  (1 << FLOW_KEY_PROTO_DST) | 		\
294			  (1 << FLOW_KEY_NFCT_SRC) |		\
295			  (1 << FLOW_KEY_NFCT_DST) |		\
296			  (1 << FLOW_KEY_NFCT_PROTO_SRC) |	\
297			  (1 << FLOW_KEY_NFCT_PROTO_DST))
298
299static int flow_classify(struct sk_buff *skb, const struct tcf_proto *tp,
300			 struct tcf_result *res)
301{
302	struct flow_head *head = rcu_dereference_bh(tp->root);
303	struct flow_filter *f;
304	u32 keymask;
305	u32 classid;
306	unsigned int n, key;
307	int r;
308
309	list_for_each_entry_rcu(f, &head->filters, list) {
310		u32 keys[FLOW_KEY_MAX + 1];
311		struct flow_keys flow_keys;
312
313		if (!tcf_em_tree_match(skb, &f->ematches, NULL))
314			continue;
315
316		keymask = f->keymask;
317		if (keymask & FLOW_KEYS_NEEDED)
318			skb_flow_dissect_flow_keys(skb, &flow_keys, 0);
319
320		for (n = 0; n < f->nkeys; n++) {
321			key = ffs(keymask) - 1;
322			keymask &= ~(1 << key);
323			keys[n] = flow_key_get(skb, key, &flow_keys);
324		}
325
326		if (f->mode == FLOW_MODE_HASH)
327			classid = jhash2(keys, f->nkeys, f->hashrnd);
328		else {
329			classid = keys[0];
330			classid = (classid & f->mask) ^ f->xor;
331			classid = (classid >> f->rshift) + f->addend;
332		}
333
334		if (f->divisor)
335			classid %= f->divisor;
336
337		res->class   = 0;
338		res->classid = TC_H_MAKE(f->baseclass, f->baseclass + classid);
339
340		r = tcf_exts_exec(skb, &f->exts, res);
341		if (r < 0)
342			continue;
343		return r;
344	}
345	return -1;
346}
347
348static void flow_perturbation(unsigned long arg)
349{
350	struct flow_filter *f = (struct flow_filter *)arg;
351
352	get_random_bytes(&f->hashrnd, 4);
353	if (f->perturb_period)
354		mod_timer(&f->perturb_timer, jiffies + f->perturb_period);
355}
356
357static const struct nla_policy flow_policy[TCA_FLOW_MAX + 1] = {
358	[TCA_FLOW_KEYS]		= { .type = NLA_U32 },
359	[TCA_FLOW_MODE]		= { .type = NLA_U32 },
360	[TCA_FLOW_BASECLASS]	= { .type = NLA_U32 },
361	[TCA_FLOW_RSHIFT]	= { .type = NLA_U32 },
362	[TCA_FLOW_ADDEND]	= { .type = NLA_U32 },
363	[TCA_FLOW_MASK]		= { .type = NLA_U32 },
364	[TCA_FLOW_XOR]		= { .type = NLA_U32 },
365	[TCA_FLOW_DIVISOR]	= { .type = NLA_U32 },
366	[TCA_FLOW_ACT]		= { .type = NLA_NESTED },
367	[TCA_FLOW_POLICE]	= { .type = NLA_NESTED },
368	[TCA_FLOW_EMATCHES]	= { .type = NLA_NESTED },
369	[TCA_FLOW_PERTURB]	= { .type = NLA_U32 },
370};
371
372static void flow_destroy_filter(struct rcu_head *head)
373{
374	struct flow_filter *f = container_of(head, struct flow_filter, rcu);
375
376	del_timer_sync(&f->perturb_timer);
377	tcf_exts_destroy(&f->exts);
378	tcf_em_tree_destroy(&f->ematches);
379	kfree(f);
380}
381
382static int flow_change(struct net *net, struct sk_buff *in_skb,
383		       struct tcf_proto *tp, unsigned long base,
384		       u32 handle, struct nlattr **tca,
385		       unsigned long *arg, bool ovr)
386{
387	struct flow_head *head = rtnl_dereference(tp->root);
388	struct flow_filter *fold, *fnew;
389	struct nlattr *opt = tca[TCA_OPTIONS];
390	struct nlattr *tb[TCA_FLOW_MAX + 1];
391	struct tcf_exts e;
392	struct tcf_ematch_tree t;
393	unsigned int nkeys = 0;
394	unsigned int perturb_period = 0;
395	u32 baseclass = 0;
396	u32 keymask = 0;
397	u32 mode;
398	int err;
399
400	if (opt == NULL)
401		return -EINVAL;
402
403	err = nla_parse_nested(tb, TCA_FLOW_MAX, opt, flow_policy);
404	if (err < 0)
405		return err;
406
407	if (tb[TCA_FLOW_BASECLASS]) {
408		baseclass = nla_get_u32(tb[TCA_FLOW_BASECLASS]);
409		if (TC_H_MIN(baseclass) == 0)
410			return -EINVAL;
411	}
412
413	if (tb[TCA_FLOW_KEYS]) {
414		keymask = nla_get_u32(tb[TCA_FLOW_KEYS]);
415
416		nkeys = hweight32(keymask);
417		if (nkeys == 0)
418			return -EINVAL;
419
420		if (fls(keymask) - 1 > FLOW_KEY_MAX)
421			return -EOPNOTSUPP;
422
423		if ((keymask & (FLOW_KEY_SKUID|FLOW_KEY_SKGID)) &&
424		    sk_user_ns(NETLINK_CB(in_skb).sk) != &init_user_ns)
425			return -EOPNOTSUPP;
426	}
427
428	err = tcf_exts_init(&e, TCA_FLOW_ACT, TCA_FLOW_POLICE);
429	if (err < 0)
430		goto err1;
431	err = tcf_exts_validate(net, tp, tb, tca[TCA_RATE], &e, ovr);
432	if (err < 0)
433		goto err1;
434
435	err = tcf_em_tree_validate(tp, tb[TCA_FLOW_EMATCHES], &t);
436	if (err < 0)
437		goto err1;
438
439	err = -ENOBUFS;
440	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
441	if (!fnew)
442		goto err2;
443
444	err = tcf_exts_init(&fnew->exts, TCA_FLOW_ACT, TCA_FLOW_POLICE);
445	if (err < 0)
446		goto err3;
447
448	fold = (struct flow_filter *)*arg;
449	if (fold) {
450		err = -EINVAL;
451		if (fold->handle != handle && handle)
452			goto err3;
453
454		/* Copy fold into fnew */
455		fnew->tp = fold->tp;
456		fnew->handle = fold->handle;
457		fnew->nkeys = fold->nkeys;
458		fnew->keymask = fold->keymask;
459		fnew->mode = fold->mode;
460		fnew->mask = fold->mask;
461		fnew->xor = fold->xor;
462		fnew->rshift = fold->rshift;
463		fnew->addend = fold->addend;
464		fnew->divisor = fold->divisor;
465		fnew->baseclass = fold->baseclass;
466		fnew->hashrnd = fold->hashrnd;
467
468		mode = fold->mode;
469		if (tb[TCA_FLOW_MODE])
470			mode = nla_get_u32(tb[TCA_FLOW_MODE]);
471		if (mode != FLOW_MODE_HASH && nkeys > 1)
472			goto err3;
473
474		if (mode == FLOW_MODE_HASH)
475			perturb_period = fold->perturb_period;
476		if (tb[TCA_FLOW_PERTURB]) {
477			if (mode != FLOW_MODE_HASH)
478				goto err3;
479			perturb_period = nla_get_u32(tb[TCA_FLOW_PERTURB]) * HZ;
480		}
481	} else {
482		err = -EINVAL;
483		if (!handle)
484			goto err3;
485		if (!tb[TCA_FLOW_KEYS])
486			goto err3;
487
488		mode = FLOW_MODE_MAP;
489		if (tb[TCA_FLOW_MODE])
490			mode = nla_get_u32(tb[TCA_FLOW_MODE]);
491		if (mode != FLOW_MODE_HASH && nkeys > 1)
492			goto err3;
493
494		if (tb[TCA_FLOW_PERTURB]) {
495			if (mode != FLOW_MODE_HASH)
496				goto err3;
497			perturb_period = nla_get_u32(tb[TCA_FLOW_PERTURB]) * HZ;
498		}
499
500		if (TC_H_MAJ(baseclass) == 0)
501			baseclass = TC_H_MAKE(tp->q->handle, baseclass);
502		if (TC_H_MIN(baseclass) == 0)
503			baseclass = TC_H_MAKE(baseclass, 1);
504
505		fnew->handle = handle;
506		fnew->mask  = ~0U;
507		fnew->tp = tp;
508		get_random_bytes(&fnew->hashrnd, 4);
509	}
510
511	fnew->perturb_timer.function = flow_perturbation;
512	fnew->perturb_timer.data = (unsigned long)fnew;
513	init_timer_deferrable(&fnew->perturb_timer);
514
515	tcf_exts_change(tp, &fnew->exts, &e);
516	tcf_em_tree_change(tp, &fnew->ematches, &t);
517
518	netif_keep_dst(qdisc_dev(tp->q));
519
520	if (tb[TCA_FLOW_KEYS]) {
521		fnew->keymask = keymask;
522		fnew->nkeys   = nkeys;
523	}
524
525	fnew->mode = mode;
526
527	if (tb[TCA_FLOW_MASK])
528		fnew->mask = nla_get_u32(tb[TCA_FLOW_MASK]);
529	if (tb[TCA_FLOW_XOR])
530		fnew->xor = nla_get_u32(tb[TCA_FLOW_XOR]);
531	if (tb[TCA_FLOW_RSHIFT])
532		fnew->rshift = nla_get_u32(tb[TCA_FLOW_RSHIFT]);
533	if (tb[TCA_FLOW_ADDEND])
534		fnew->addend = nla_get_u32(tb[TCA_FLOW_ADDEND]);
535
536	if (tb[TCA_FLOW_DIVISOR])
537		fnew->divisor = nla_get_u32(tb[TCA_FLOW_DIVISOR]);
538	if (baseclass)
539		fnew->baseclass = baseclass;
540
541	fnew->perturb_period = perturb_period;
542	if (perturb_period)
543		mod_timer(&fnew->perturb_timer, jiffies + perturb_period);
544
545	if (*arg == 0)
546		list_add_tail_rcu(&fnew->list, &head->filters);
547	else
548		list_replace_rcu(&fold->list, &fnew->list);
549
550	*arg = (unsigned long)fnew;
551
552	if (fold)
553		call_rcu(&fold->rcu, flow_destroy_filter);
554	return 0;
555
556err3:
557	tcf_exts_destroy(&fnew->exts);
558err2:
559	tcf_em_tree_destroy(&t);
560	kfree(fnew);
561err1:
562	tcf_exts_destroy(&e);
563	return err;
564}
565
566static int flow_delete(struct tcf_proto *tp, unsigned long arg)
567{
568	struct flow_filter *f = (struct flow_filter *)arg;
569
570	list_del_rcu(&f->list);
571	call_rcu(&f->rcu, flow_destroy_filter);
572	return 0;
573}
574
575static int flow_init(struct tcf_proto *tp)
576{
577	struct flow_head *head;
578
579	head = kzalloc(sizeof(*head), GFP_KERNEL);
580	if (head == NULL)
581		return -ENOBUFS;
582	INIT_LIST_HEAD(&head->filters);
583	rcu_assign_pointer(tp->root, head);
584	return 0;
585}
586
587static bool flow_destroy(struct tcf_proto *tp, bool force)
588{
589	struct flow_head *head = rtnl_dereference(tp->root);
590	struct flow_filter *f, *next;
591
592	if (!force && !list_empty(&head->filters))
593		return false;
594
595	list_for_each_entry_safe(f, next, &head->filters, list) {
596		list_del_rcu(&f->list);
597		call_rcu(&f->rcu, flow_destroy_filter);
598	}
 
599	kfree_rcu(head, rcu);
600	return true;
601}
602
603static unsigned long flow_get(struct tcf_proto *tp, u32 handle)
604{
605	struct flow_head *head = rtnl_dereference(tp->root);
606	struct flow_filter *f;
607
608	list_for_each_entry(f, &head->filters, list)
609		if (f->handle == handle)
610			return (unsigned long)f;
611	return 0;
612}
613
614static int flow_dump(struct net *net, struct tcf_proto *tp, unsigned long fh,
615		     struct sk_buff *skb, struct tcmsg *t)
616{
617	struct flow_filter *f = (struct flow_filter *)fh;
618	struct nlattr *nest;
619
620	if (f == NULL)
621		return skb->len;
622
623	t->tcm_handle = f->handle;
624
625	nest = nla_nest_start(skb, TCA_OPTIONS);
626	if (nest == NULL)
627		goto nla_put_failure;
628
629	if (nla_put_u32(skb, TCA_FLOW_KEYS, f->keymask) ||
630	    nla_put_u32(skb, TCA_FLOW_MODE, f->mode))
631		goto nla_put_failure;
632
633	if (f->mask != ~0 || f->xor != 0) {
634		if (nla_put_u32(skb, TCA_FLOW_MASK, f->mask) ||
635		    nla_put_u32(skb, TCA_FLOW_XOR, f->xor))
636			goto nla_put_failure;
637	}
638	if (f->rshift &&
639	    nla_put_u32(skb, TCA_FLOW_RSHIFT, f->rshift))
640		goto nla_put_failure;
641	if (f->addend &&
642	    nla_put_u32(skb, TCA_FLOW_ADDEND, f->addend))
643		goto nla_put_failure;
644
645	if (f->divisor &&
646	    nla_put_u32(skb, TCA_FLOW_DIVISOR, f->divisor))
647		goto nla_put_failure;
648	if (f->baseclass &&
649	    nla_put_u32(skb, TCA_FLOW_BASECLASS, f->baseclass))
650		goto nla_put_failure;
651
652	if (f->perturb_period &&
653	    nla_put_u32(skb, TCA_FLOW_PERTURB, f->perturb_period / HZ))
654		goto nla_put_failure;
655
656	if (tcf_exts_dump(skb, &f->exts) < 0)
657		goto nla_put_failure;
658#ifdef CONFIG_NET_EMATCH
659	if (f->ematches.hdr.nmatches &&
660	    tcf_em_tree_dump(skb, &f->ematches, TCA_FLOW_EMATCHES) < 0)
661		goto nla_put_failure;
662#endif
663	nla_nest_end(skb, nest);
664
665	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
666		goto nla_put_failure;
667
668	return skb->len;
669
670nla_put_failure:
671	nla_nest_cancel(skb, nest);
672	return -1;
673}
674
675static void flow_walk(struct tcf_proto *tp, struct tcf_walker *arg)
676{
677	struct flow_head *head = rtnl_dereference(tp->root);
678	struct flow_filter *f;
679
680	list_for_each_entry(f, &head->filters, list) {
681		if (arg->count < arg->skip)
682			goto skip;
683		if (arg->fn(tp, (unsigned long)f, arg) < 0) {
684			arg->stop = 1;
685			break;
686		}
687skip:
688		arg->count++;
689	}
690}
691
692static struct tcf_proto_ops cls_flow_ops __read_mostly = {
693	.kind		= "flow",
694	.classify	= flow_classify,
695	.init		= flow_init,
696	.destroy	= flow_destroy,
697	.change		= flow_change,
698	.delete		= flow_delete,
699	.get		= flow_get,
700	.dump		= flow_dump,
701	.walk		= flow_walk,
702	.owner		= THIS_MODULE,
703};
704
705static int __init cls_flow_init(void)
706{
707	return register_tcf_proto_ops(&cls_flow_ops);
708}
709
710static void __exit cls_flow_exit(void)
711{
712	unregister_tcf_proto_ops(&cls_flow_ops);
713}
714
715module_init(cls_flow_init);
716module_exit(cls_flow_exit);
717
718MODULE_LICENSE("GPL");
719MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
720MODULE_DESCRIPTION("TC flow classifier");