Linux Audio

Check our new training course

Yocto / OpenEmbedded training

Feb 10-13, 2025
Register
Loading...
v3.1
  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
 25#include <net/pkt_cls.h>
 26#include <net/ip.h>
 27#include <net/route.h>
 28#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 
 
 29#include <net/netfilter/nf_conntrack.h>
 30#endif
 31
 32struct flow_head {
 33	struct list_head	filters;
 
 34};
 35
 36struct flow_filter {
 37	struct list_head	list;
 38	struct tcf_exts		exts;
 39	struct tcf_ematch_tree	ematches;
 
 40	struct timer_list	perturb_timer;
 41	u32			perturb_period;
 42	u32			handle;
 43
 44	u32			nkeys;
 45	u32			keymask;
 46	u32			mode;
 47	u32			mask;
 48	u32			xor;
 49	u32			rshift;
 50	u32			addend;
 51	u32			divisor;
 52	u32			baseclass;
 53	u32			hashrnd;
 54};
 55
 56static const struct tcf_ext_map flow_ext_map = {
 57	.action	= TCA_FLOW_ACT,
 58	.police	= TCA_FLOW_POLICE,
 59};
 60
 61static inline u32 addr_fold(void *addr)
 62{
 63	unsigned long a = (unsigned long)addr;
 64
 65	return (a & 0xFFFFFFFF) ^ (BITS_PER_LONG > 32 ? a >> 32 : 0);
 66}
 67
 68static u32 flow_get_src(struct sk_buff *skb)
 69{
 70	switch (skb->protocol) {
 71	case htons(ETH_P_IP):
 72		if (pskb_network_may_pull(skb, sizeof(struct iphdr)))
 73			return ntohl(ip_hdr(skb)->saddr);
 74		break;
 75	case htons(ETH_P_IPV6):
 76		if (pskb_network_may_pull(skb, sizeof(struct ipv6hdr)))
 77			return ntohl(ipv6_hdr(skb)->saddr.s6_addr32[3]);
 78		break;
 79	}
 80
 81	return addr_fold(skb->sk);
 82}
 83
 84static u32 flow_get_dst(struct sk_buff *skb)
 85{
 86	switch (skb->protocol) {
 87	case htons(ETH_P_IP):
 88		if (pskb_network_may_pull(skb, sizeof(struct iphdr)))
 89			return ntohl(ip_hdr(skb)->daddr);
 90		break;
 91	case htons(ETH_P_IPV6):
 92		if (pskb_network_may_pull(skb, sizeof(struct ipv6hdr)))
 93			return ntohl(ipv6_hdr(skb)->daddr.s6_addr32[3]);
 94		break;
 95	}
 96
 97	return addr_fold(skb_dst(skb)) ^ (__force u16)skb->protocol;
 98}
 99
100static u32 flow_get_proto(struct sk_buff *skb)
 
101{
102	switch (skb->protocol) {
103	case htons(ETH_P_IP):
104		return pskb_network_may_pull(skb, sizeof(struct iphdr)) ?
105		       ip_hdr(skb)->protocol : 0;
106	case htons(ETH_P_IPV6):
107		return pskb_network_may_pull(skb, sizeof(struct ipv6hdr)) ?
108		       ipv6_hdr(skb)->nexthdr : 0;
109	default:
110		return 0;
111	}
112}
113
114static u32 flow_get_proto_src(struct sk_buff *skb)
 
115{
116	switch (skb->protocol) {
117	case htons(ETH_P_IP): {
118		struct iphdr *iph;
119		int poff;
120
121		if (!pskb_network_may_pull(skb, sizeof(*iph)))
122			break;
123		iph = ip_hdr(skb);
124		if (ip_is_fragment(iph))
125			break;
126		poff = proto_ports_offset(iph->protocol);
127		if (poff >= 0 &&
128		    pskb_network_may_pull(skb, iph->ihl * 4 + 2 + poff)) {
129			iph = ip_hdr(skb);
130			return ntohs(*(__be16 *)((void *)iph + iph->ihl * 4 +
131						 poff));
132		}
133		break;
134	}
135	case htons(ETH_P_IPV6): {
136		struct ipv6hdr *iph;
137		int poff;
138
139		if (!pskb_network_may_pull(skb, sizeof(*iph)))
140			break;
141		iph = ipv6_hdr(skb);
142		poff = proto_ports_offset(iph->nexthdr);
143		if (poff >= 0 &&
144		    pskb_network_may_pull(skb, sizeof(*iph) + poff + 2)) {
145			iph = ipv6_hdr(skb);
146			return ntohs(*(__be16 *)((void *)iph + sizeof(*iph) +
147						 poff));
148		}
149		break;
150	}
151	}
152
153	return addr_fold(skb->sk);
154}
155
156static u32 flow_get_proto_dst(struct sk_buff *skb)
 
157{
158	switch (skb->protocol) {
159	case htons(ETH_P_IP): {
160		struct iphdr *iph;
161		int poff;
162
163		if (!pskb_network_may_pull(skb, sizeof(*iph)))
164			break;
165		iph = ip_hdr(skb);
166		if (ip_is_fragment(iph))
167			break;
168		poff = proto_ports_offset(iph->protocol);
169		if (poff >= 0 &&
170		    pskb_network_may_pull(skb, iph->ihl * 4 + 4 + poff)) {
171			iph = ip_hdr(skb);
172			return ntohs(*(__be16 *)((void *)iph + iph->ihl * 4 +
173						 2 + poff));
174		}
175		break;
176	}
177	case htons(ETH_P_IPV6): {
178		struct ipv6hdr *iph;
179		int poff;
180
181		if (!pskb_network_may_pull(skb, sizeof(*iph)))
182			break;
183		iph = ipv6_hdr(skb);
184		poff = proto_ports_offset(iph->nexthdr);
185		if (poff >= 0 &&
186		    pskb_network_may_pull(skb, sizeof(*iph) + poff + 4)) {
187			iph = ipv6_hdr(skb);
188			return ntohs(*(__be16 *)((void *)iph + sizeof(*iph) +
189						 poff + 2));
190		}
191		break;
192	}
193	}
194
195	return addr_fold(skb_dst(skb)) ^ (__force u16)skb->protocol;
196}
197
198static u32 flow_get_iif(const struct sk_buff *skb)
199{
200	return skb->skb_iif;
201}
202
203static u32 flow_get_priority(const struct sk_buff *skb)
204{
205	return skb->priority;
206}
207
208static u32 flow_get_mark(const struct sk_buff *skb)
209{
210	return skb->mark;
211}
212
213static u32 flow_get_nfct(const struct sk_buff *skb)
214{
215#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
216	return addr_fold(skb->nfct);
217#else
218	return 0;
219#endif
220}
221
222#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
223#define CTTUPLE(skb, member)						\
224({									\
225	enum ip_conntrack_info ctinfo;					\
226	struct nf_conn *ct = nf_ct_get(skb, &ctinfo);			\
227	if (ct == NULL)							\
228		goto fallback;						\
229	ct->tuplehash[CTINFO2DIR(ctinfo)].tuple.member;			\
230})
231#else
232#define CTTUPLE(skb, member)						\
233({									\
234	goto fallback;							\
235	0;								\
236})
237#endif
238
239static u32 flow_get_nfct_src(struct sk_buff *skb)
 
240{
241	switch (skb->protocol) {
242	case htons(ETH_P_IP):
243		return ntohl(CTTUPLE(skb, src.u3.ip));
244	case htons(ETH_P_IPV6):
245		return ntohl(CTTUPLE(skb, src.u3.ip6[3]));
246	}
247fallback:
248	return flow_get_src(skb);
249}
250
251static u32 flow_get_nfct_dst(struct sk_buff *skb)
 
252{
253	switch (skb->protocol) {
254	case htons(ETH_P_IP):
255		return ntohl(CTTUPLE(skb, dst.u3.ip));
256	case htons(ETH_P_IPV6):
257		return ntohl(CTTUPLE(skb, dst.u3.ip6[3]));
258	}
259fallback:
260	return flow_get_dst(skb);
261}
262
263static u32 flow_get_nfct_proto_src(struct sk_buff *skb)
 
264{
265	return ntohs(CTTUPLE(skb, src.u.all));
266fallback:
267	return flow_get_proto_src(skb);
268}
269
270static u32 flow_get_nfct_proto_dst(struct sk_buff *skb)
 
271{
272	return ntohs(CTTUPLE(skb, dst.u.all));
273fallback:
274	return flow_get_proto_dst(skb);
275}
276
277static u32 flow_get_rtclassid(const struct sk_buff *skb)
278{
279#ifdef CONFIG_IP_ROUTE_CLASSID
280	if (skb_dst(skb))
281		return skb_dst(skb)->tclassid;
282#endif
283	return 0;
284}
285
286static u32 flow_get_skuid(const struct sk_buff *skb)
287{
288	if (skb->sk && skb->sk->sk_socket && skb->sk->sk_socket->file)
289		return skb->sk->sk_socket->file->f_cred->fsuid;
 
 
 
 
 
290	return 0;
291}
292
293static u32 flow_get_skgid(const struct sk_buff *skb)
294{
295	if (skb->sk && skb->sk->sk_socket && skb->sk->sk_socket->file)
296		return skb->sk->sk_socket->file->f_cred->fsgid;
 
 
 
 
 
297	return 0;
298}
299
300static u32 flow_get_vlan_tag(const struct sk_buff *skb)
301{
302	u16 uninitialized_var(tag);
303
304	if (vlan_get_tag(skb, &tag) < 0)
305		return 0;
306	return tag & VLAN_VID_MASK;
307}
308
309static u32 flow_get_rxhash(struct sk_buff *skb)
310{
311	return skb_get_rxhash(skb);
312}
313
314static u32 flow_key_get(struct sk_buff *skb, int key)
315{
316	switch (key) {
317	case FLOW_KEY_SRC:
318		return flow_get_src(skb);
319	case FLOW_KEY_DST:
320		return flow_get_dst(skb);
321	case FLOW_KEY_PROTO:
322		return flow_get_proto(skb);
323	case FLOW_KEY_PROTO_SRC:
324		return flow_get_proto_src(skb);
325	case FLOW_KEY_PROTO_DST:
326		return flow_get_proto_dst(skb);
327	case FLOW_KEY_IIF:
328		return flow_get_iif(skb);
329	case FLOW_KEY_PRIORITY:
330		return flow_get_priority(skb);
331	case FLOW_KEY_MARK:
332		return flow_get_mark(skb);
333	case FLOW_KEY_NFCT:
334		return flow_get_nfct(skb);
335	case FLOW_KEY_NFCT_SRC:
336		return flow_get_nfct_src(skb);
337	case FLOW_KEY_NFCT_DST:
338		return flow_get_nfct_dst(skb);
339	case FLOW_KEY_NFCT_PROTO_SRC:
340		return flow_get_nfct_proto_src(skb);
341	case FLOW_KEY_NFCT_PROTO_DST:
342		return flow_get_nfct_proto_dst(skb);
343	case FLOW_KEY_RTCLASSID:
344		return flow_get_rtclassid(skb);
345	case FLOW_KEY_SKUID:
346		return flow_get_skuid(skb);
347	case FLOW_KEY_SKGID:
348		return flow_get_skgid(skb);
349	case FLOW_KEY_VLAN_TAG:
350		return flow_get_vlan_tag(skb);
351	case FLOW_KEY_RXHASH:
352		return flow_get_rxhash(skb);
353	default:
354		WARN_ON(1);
355		return 0;
356	}
357}
358
 
 
 
 
 
 
 
 
 
 
359static int flow_classify(struct sk_buff *skb, const struct tcf_proto *tp,
360			 struct tcf_result *res)
361{
362	struct flow_head *head = tp->root;
363	struct flow_filter *f;
364	u32 keymask;
365	u32 classid;
366	unsigned int n, key;
367	int r;
368
369	list_for_each_entry(f, &head->filters, list) {
370		u32 keys[f->nkeys];
 
371
372		if (!tcf_em_tree_match(skb, &f->ematches, NULL))
373			continue;
374
375		keymask = f->keymask;
 
 
376
377		for (n = 0; n < f->nkeys; n++) {
378			key = ffs(keymask) - 1;
379			keymask &= ~(1 << key);
380			keys[n] = flow_key_get(skb, key);
381		}
382
383		if (f->mode == FLOW_MODE_HASH)
384			classid = jhash2(keys, f->nkeys, f->hashrnd);
385		else {
386			classid = keys[0];
387			classid = (classid & f->mask) ^ f->xor;
388			classid = (classid >> f->rshift) + f->addend;
389		}
390
391		if (f->divisor)
392			classid %= f->divisor;
393
394		res->class   = 0;
395		res->classid = TC_H_MAKE(f->baseclass, f->baseclass + classid);
396
397		r = tcf_exts_exec(skb, &f->exts, res);
398		if (r < 0)
399			continue;
400		return r;
401	}
402	return -1;
403}
404
405static void flow_perturbation(unsigned long arg)
406{
407	struct flow_filter *f = (struct flow_filter *)arg;
408
409	get_random_bytes(&f->hashrnd, 4);
410	if (f->perturb_period)
411		mod_timer(&f->perturb_timer, jiffies + f->perturb_period);
412}
413
414static const struct nla_policy flow_policy[TCA_FLOW_MAX + 1] = {
415	[TCA_FLOW_KEYS]		= { .type = NLA_U32 },
416	[TCA_FLOW_MODE]		= { .type = NLA_U32 },
417	[TCA_FLOW_BASECLASS]	= { .type = NLA_U32 },
418	[TCA_FLOW_RSHIFT]	= { .type = NLA_U32 },
419	[TCA_FLOW_ADDEND]	= { .type = NLA_U32 },
420	[TCA_FLOW_MASK]		= { .type = NLA_U32 },
421	[TCA_FLOW_XOR]		= { .type = NLA_U32 },
422	[TCA_FLOW_DIVISOR]	= { .type = NLA_U32 },
423	[TCA_FLOW_ACT]		= { .type = NLA_NESTED },
424	[TCA_FLOW_POLICE]	= { .type = NLA_NESTED },
425	[TCA_FLOW_EMATCHES]	= { .type = NLA_NESTED },
426	[TCA_FLOW_PERTURB]	= { .type = NLA_U32 },
427};
428
429static int flow_change(struct tcf_proto *tp, unsigned long base,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
430		       u32 handle, struct nlattr **tca,
431		       unsigned long *arg)
432{
433	struct flow_head *head = tp->root;
434	struct flow_filter *f;
435	struct nlattr *opt = tca[TCA_OPTIONS];
436	struct nlattr *tb[TCA_FLOW_MAX + 1];
437	struct tcf_exts e;
438	struct tcf_ematch_tree t;
439	unsigned int nkeys = 0;
440	unsigned int perturb_period = 0;
441	u32 baseclass = 0;
442	u32 keymask = 0;
443	u32 mode;
444	int err;
445
446	if (opt == NULL)
447		return -EINVAL;
448
449	err = nla_parse_nested(tb, TCA_FLOW_MAX, opt, flow_policy);
450	if (err < 0)
451		return err;
452
453	if (tb[TCA_FLOW_BASECLASS]) {
454		baseclass = nla_get_u32(tb[TCA_FLOW_BASECLASS]);
455		if (TC_H_MIN(baseclass) == 0)
456			return -EINVAL;
457	}
458
459	if (tb[TCA_FLOW_KEYS]) {
460		keymask = nla_get_u32(tb[TCA_FLOW_KEYS]);
461
462		nkeys = hweight32(keymask);
463		if (nkeys == 0)
464			return -EINVAL;
465
466		if (fls(keymask) - 1 > FLOW_KEY_MAX)
467			return -EOPNOTSUPP;
 
 
 
 
468	}
469
470	err = tcf_exts_validate(tp, tb, tca[TCA_RATE], &e, &flow_ext_map);
471	if (err < 0)
472		return err;
473
474	err = tcf_em_tree_validate(tp, tb[TCA_FLOW_EMATCHES], &t);
475	if (err < 0)
476		goto err1;
477
478	f = (struct flow_filter *)*arg;
479	if (f != NULL) {
 
 
 
 
 
 
 
 
 
480		err = -EINVAL;
481		if (f->handle != handle && handle)
482			goto err2;
483
484		mode = f->mode;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
485		if (tb[TCA_FLOW_MODE])
486			mode = nla_get_u32(tb[TCA_FLOW_MODE]);
487		if (mode != FLOW_MODE_HASH && nkeys > 1)
488			goto err2;
489
490		if (mode == FLOW_MODE_HASH)
491			perturb_period = f->perturb_period;
492		if (tb[TCA_FLOW_PERTURB]) {
493			if (mode != FLOW_MODE_HASH)
494				goto err2;
495			perturb_period = nla_get_u32(tb[TCA_FLOW_PERTURB]) * HZ;
496		}
497	} else {
498		err = -EINVAL;
499		if (!handle)
500			goto err2;
501		if (!tb[TCA_FLOW_KEYS])
502			goto err2;
503
504		mode = FLOW_MODE_MAP;
505		if (tb[TCA_FLOW_MODE])
506			mode = nla_get_u32(tb[TCA_FLOW_MODE]);
507		if (mode != FLOW_MODE_HASH && nkeys > 1)
508			goto err2;
509
510		if (tb[TCA_FLOW_PERTURB]) {
511			if (mode != FLOW_MODE_HASH)
512				goto err2;
513			perturb_period = nla_get_u32(tb[TCA_FLOW_PERTURB]) * HZ;
514		}
515
516		if (TC_H_MAJ(baseclass) == 0)
517			baseclass = TC_H_MAKE(tp->q->handle, baseclass);
 
 
 
518		if (TC_H_MIN(baseclass) == 0)
519			baseclass = TC_H_MAKE(baseclass, 1);
520
521		err = -ENOBUFS;
522		f = kzalloc(sizeof(*f), GFP_KERNEL);
523		if (f == NULL)
524			goto err2;
525
526		f->handle = handle;
527		f->mask	  = ~0U;
528
529		get_random_bytes(&f->hashrnd, 4);
530		f->perturb_timer.function = flow_perturbation;
531		f->perturb_timer.data = (unsigned long)f;
532		init_timer_deferrable(&f->perturb_timer);
533	}
534
535	tcf_exts_change(tp, &f->exts, &e);
536	tcf_em_tree_change(tp, &f->ematches, &t);
537
538	tcf_tree_lock(tp);
539
540	if (tb[TCA_FLOW_KEYS]) {
541		f->keymask = keymask;
542		f->nkeys   = nkeys;
543	}
544
545	f->mode = mode;
546
547	if (tb[TCA_FLOW_MASK])
548		f->mask = nla_get_u32(tb[TCA_FLOW_MASK]);
549	if (tb[TCA_FLOW_XOR])
550		f->xor = nla_get_u32(tb[TCA_FLOW_XOR]);
551	if (tb[TCA_FLOW_RSHIFT])
552		f->rshift = nla_get_u32(tb[TCA_FLOW_RSHIFT]);
553	if (tb[TCA_FLOW_ADDEND])
554		f->addend = nla_get_u32(tb[TCA_FLOW_ADDEND]);
555
556	if (tb[TCA_FLOW_DIVISOR])
557		f->divisor = nla_get_u32(tb[TCA_FLOW_DIVISOR]);
558	if (baseclass)
559		f->baseclass = baseclass;
560
561	f->perturb_period = perturb_period;
562	del_timer(&f->perturb_timer);
563	if (perturb_period)
564		mod_timer(&f->perturb_timer, jiffies + perturb_period);
565
566	if (*arg == 0)
567		list_add_tail(&f->list, &head->filters);
568
569	tcf_tree_unlock(tp);
570
571	*arg = (unsigned long)f;
 
 
 
 
 
572	return 0;
573
574err2:
575	tcf_em_tree_destroy(tp, &t);
 
576err1:
577	tcf_exts_destroy(tp, &e);
578	return err;
579}
580
581static void flow_destroy_filter(struct tcf_proto *tp, struct flow_filter *f)
 
582{
583	del_timer_sync(&f->perturb_timer);
584	tcf_exts_destroy(tp, &f->exts);
585	tcf_em_tree_destroy(tp, &f->ematches);
586	kfree(f);
587}
588
589static int flow_delete(struct tcf_proto *tp, unsigned long arg)
590{
591	struct flow_filter *f = (struct flow_filter *)arg;
592
593	tcf_tree_lock(tp);
594	list_del(&f->list);
595	tcf_tree_unlock(tp);
596	flow_destroy_filter(tp, f);
597	return 0;
598}
599
600static int flow_init(struct tcf_proto *tp)
601{
602	struct flow_head *head;
603
604	head = kzalloc(sizeof(*head), GFP_KERNEL);
605	if (head == NULL)
606		return -ENOBUFS;
607	INIT_LIST_HEAD(&head->filters);
608	tp->root = head;
609	return 0;
610}
611
612static void flow_destroy(struct tcf_proto *tp)
613{
614	struct flow_head *head = tp->root;
615	struct flow_filter *f, *next;
616
617	list_for_each_entry_safe(f, next, &head->filters, list) {
618		list_del(&f->list);
619		flow_destroy_filter(tp, f);
 
 
 
620	}
621	kfree(head);
622}
623
624static unsigned long flow_get(struct tcf_proto *tp, u32 handle)
625{
626	struct flow_head *head = tp->root;
627	struct flow_filter *f;
628
629	list_for_each_entry(f, &head->filters, list)
630		if (f->handle == handle)
631			return (unsigned long)f;
632	return 0;
633}
634
635static void flow_put(struct tcf_proto *tp, unsigned long f)
636{
637}
638
639static int flow_dump(struct tcf_proto *tp, unsigned long fh,
640		     struct sk_buff *skb, struct tcmsg *t)
641{
642	struct flow_filter *f = (struct flow_filter *)fh;
643	struct nlattr *nest;
644
645	if (f == NULL)
646		return skb->len;
647
648	t->tcm_handle = f->handle;
649
650	nest = nla_nest_start(skb, TCA_OPTIONS);
651	if (nest == NULL)
652		goto nla_put_failure;
653
654	NLA_PUT_U32(skb, TCA_FLOW_KEYS, f->keymask);
655	NLA_PUT_U32(skb, TCA_FLOW_MODE, f->mode);
 
656
657	if (f->mask != ~0 || f->xor != 0) {
658		NLA_PUT_U32(skb, TCA_FLOW_MASK, f->mask);
659		NLA_PUT_U32(skb, TCA_FLOW_XOR, f->xor);
 
660	}
661	if (f->rshift)
662		NLA_PUT_U32(skb, TCA_FLOW_RSHIFT, f->rshift);
663	if (f->addend)
664		NLA_PUT_U32(skb, TCA_FLOW_ADDEND, f->addend);
665
666	if (f->divisor)
667		NLA_PUT_U32(skb, TCA_FLOW_DIVISOR, f->divisor);
668	if (f->baseclass)
669		NLA_PUT_U32(skb, TCA_FLOW_BASECLASS, f->baseclass);
670
671	if (f->perturb_period)
672		NLA_PUT_U32(skb, TCA_FLOW_PERTURB, f->perturb_period / HZ);
 
 
 
 
 
 
 
 
673
674	if (tcf_exts_dump(skb, &f->exts, &flow_ext_map) < 0)
675		goto nla_put_failure;
676#ifdef CONFIG_NET_EMATCH
677	if (f->ematches.hdr.nmatches &&
678	    tcf_em_tree_dump(skb, &f->ematches, TCA_FLOW_EMATCHES) < 0)
679		goto nla_put_failure;
680#endif
681	nla_nest_end(skb, nest);
682
683	if (tcf_exts_dump_stats(skb, &f->exts, &flow_ext_map) < 0)
684		goto nla_put_failure;
685
686	return skb->len;
687
688nla_put_failure:
689	nlmsg_trim(skb, nest);
690	return -1;
691}
692
693static void flow_walk(struct tcf_proto *tp, struct tcf_walker *arg)
694{
695	struct flow_head *head = tp->root;
696	struct flow_filter *f;
697
698	list_for_each_entry(f, &head->filters, list) {
699		if (arg->count < arg->skip)
700			goto skip;
701		if (arg->fn(tp, (unsigned long)f, arg) < 0) {
702			arg->stop = 1;
703			break;
704		}
705skip:
706		arg->count++;
707	}
708}
709
710static struct tcf_proto_ops cls_flow_ops __read_mostly = {
711	.kind		= "flow",
712	.classify	= flow_classify,
713	.init		= flow_init,
714	.destroy	= flow_destroy,
715	.change		= flow_change,
716	.delete		= flow_delete,
717	.get		= flow_get,
718	.put		= flow_put,
719	.dump		= flow_dump,
720	.walk		= flow_walk,
721	.owner		= THIS_MODULE,
722};
723
724static int __init cls_flow_init(void)
725{
726	return register_tcf_proto_ops(&cls_flow_ops);
727}
728
729static void __exit cls_flow_exit(void)
730{
731	unregister_tcf_proto_ops(&cls_flow_ops);
732}
733
734module_init(cls_flow_init);
735module_exit(cls_flow_exit);
736
737MODULE_LICENSE("GPL");
738MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
739MODULE_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");