2 * net/sched/cls_api.c Packet classifier API.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
13 * Eduardo J. Blanco <ejbs@netlabs.com.uy> :990222: kmod support
17 #include <linux/module.h>
18 #include <linux/types.h>
19 #include <linux/kernel.h>
20 #include <linux/string.h>
21 #include <linux/errno.h>
22 #include <linux/err.h>
23 #include <linux/skbuff.h>
24 #include <linux/init.h>
25 #include <linux/kmod.h>
26 #include <linux/slab.h>
27 #include <net/net_namespace.h>
29 #include <net/netlink.h>
30 #include <net/pkt_sched.h>
31 #include <net/pkt_cls.h>
33 /* The list of all installed classifier types */
34 static LIST_HEAD(tcf_proto_base
);
36 /* Protects list of registered TC modules. It is pure SMP lock. */
37 static DEFINE_RWLOCK(cls_mod_lock
);
39 /* Find classifier type by string name */
41 static const struct tcf_proto_ops
*tcf_proto_lookup_ops(const char *kind
)
43 const struct tcf_proto_ops
*t
, *res
= NULL
;
46 read_lock(&cls_mod_lock
);
47 list_for_each_entry(t
, &tcf_proto_base
, head
) {
48 if (strcmp(kind
, t
->kind
) == 0) {
49 if (try_module_get(t
->owner
))
54 read_unlock(&cls_mod_lock
);
59 /* Register(unregister) new classifier type */
61 int register_tcf_proto_ops(struct tcf_proto_ops
*ops
)
63 struct tcf_proto_ops
*t
;
66 write_lock(&cls_mod_lock
);
67 list_for_each_entry(t
, &tcf_proto_base
, head
)
68 if (!strcmp(ops
->kind
, t
->kind
))
71 list_add_tail(&ops
->head
, &tcf_proto_base
);
74 write_unlock(&cls_mod_lock
);
77 EXPORT_SYMBOL(register_tcf_proto_ops
);
79 static struct workqueue_struct
*tc_filter_wq
;
81 int unregister_tcf_proto_ops(struct tcf_proto_ops
*ops
)
83 struct tcf_proto_ops
*t
;
86 /* Wait for outstanding call_rcu()s, if any, from a
87 * tcf_proto_ops's destroy() handler.
90 flush_workqueue(tc_filter_wq
);
92 write_lock(&cls_mod_lock
);
93 list_for_each_entry(t
, &tcf_proto_base
, head
) {
100 write_unlock(&cls_mod_lock
);
103 EXPORT_SYMBOL(unregister_tcf_proto_ops
);
105 bool tcf_queue_work(struct work_struct
*work
)
107 return queue_work(tc_filter_wq
, work
);
109 EXPORT_SYMBOL(tcf_queue_work
);
111 /* Select new prio value from the range, managed by kernel. */
113 static inline u32
tcf_auto_prio(struct tcf_proto
*tp
)
115 u32 first
= TC_H_MAKE(0xC0000000U
, 0U);
118 first
= tp
->prio
- 1;
120 return TC_H_MAJ(first
);
123 static struct tcf_proto
*tcf_proto_create(const char *kind
, u32 protocol
,
124 u32 prio
, u32 parent
, struct Qdisc
*q
,
125 struct tcf_chain
*chain
)
127 struct tcf_proto
*tp
;
130 tp
= kzalloc(sizeof(*tp
), GFP_KERNEL
);
132 return ERR_PTR(-ENOBUFS
);
135 tp
->ops
= tcf_proto_lookup_ops(kind
);
137 #ifdef CONFIG_MODULES
139 request_module("cls_%s", kind
);
141 tp
->ops
= tcf_proto_lookup_ops(kind
);
142 /* We dropped the RTNL semaphore in order to perform
143 * the module load. So, even if we succeeded in loading
144 * the module we have to replay the request. We indicate
145 * this using -EAGAIN.
148 module_put(tp
->ops
->owner
);
156 tp
->classify
= tp
->ops
->classify
;
157 tp
->protocol
= protocol
;
159 tp
->classid
= parent
;
163 err
= tp
->ops
->init(tp
);
165 module_put(tp
->ops
->owner
);
175 static void tcf_proto_destroy(struct tcf_proto
*tp
)
177 tp
->ops
->destroy(tp
);
178 module_put(tp
->ops
->owner
);
182 static struct tcf_chain
*tcf_chain_create(struct tcf_block
*block
,
185 struct tcf_chain
*chain
;
187 chain
= kzalloc(sizeof(*chain
), GFP_KERNEL
);
190 list_add_tail(&chain
->list
, &block
->chain_list
);
191 chain
->block
= block
;
192 chain
->index
= chain_index
;
197 static void tcf_chain_head_change(struct tcf_chain
*chain
,
198 struct tcf_proto
*tp_head
)
200 if (chain
->chain_head_change
)
201 chain
->chain_head_change(tp_head
,
202 chain
->chain_head_change_priv
);
205 static void tcf_chain_flush(struct tcf_chain
*chain
)
207 struct tcf_proto
*tp
= rtnl_dereference(chain
->filter_chain
);
209 tcf_chain_head_change(chain
, NULL
);
211 RCU_INIT_POINTER(chain
->filter_chain
, tp
->next
);
212 tcf_proto_destroy(tp
);
213 tp
= rtnl_dereference(chain
->filter_chain
);
214 tcf_chain_put(chain
);
218 static void tcf_chain_destroy(struct tcf_chain
*chain
)
220 list_del(&chain
->list
);
224 static void tcf_chain_hold(struct tcf_chain
*chain
)
229 struct tcf_chain
*tcf_chain_get(struct tcf_block
*block
, u32 chain_index
,
232 struct tcf_chain
*chain
;
234 list_for_each_entry(chain
, &block
->chain_list
, list
) {
235 if (chain
->index
== chain_index
) {
236 tcf_chain_hold(chain
);
241 return create
? tcf_chain_create(block
, chain_index
) : NULL
;
243 EXPORT_SYMBOL(tcf_chain_get
);
245 void tcf_chain_put(struct tcf_chain
*chain
)
247 if (--chain
->refcnt
== 0)
248 tcf_chain_destroy(chain
);
250 EXPORT_SYMBOL(tcf_chain_put
);
252 static void tcf_block_offload_cmd(struct tcf_block
*block
, struct Qdisc
*q
,
253 struct tcf_block_ext_info
*ei
,
254 enum tc_block_command command
)
256 struct net_device
*dev
= q
->dev_queue
->dev
;
257 struct tc_block_offload bo
= {};
259 if (!dev
->netdev_ops
->ndo_setup_tc
)
261 bo
.command
= command
;
262 bo
.binder_type
= ei
->binder_type
;
264 dev
->netdev_ops
->ndo_setup_tc(dev
, TC_SETUP_BLOCK
, &bo
);
267 static void tcf_block_offload_bind(struct tcf_block
*block
, struct Qdisc
*q
,
268 struct tcf_block_ext_info
*ei
)
270 tcf_block_offload_cmd(block
, q
, ei
, TC_BLOCK_BIND
);
273 static void tcf_block_offload_unbind(struct tcf_block
*block
, struct Qdisc
*q
,
274 struct tcf_block_ext_info
*ei
)
276 tcf_block_offload_cmd(block
, q
, ei
, TC_BLOCK_UNBIND
);
279 int tcf_block_get_ext(struct tcf_block
**p_block
, struct Qdisc
*q
,
280 struct tcf_block_ext_info
*ei
)
282 struct tcf_block
*block
= kzalloc(sizeof(*block
), GFP_KERNEL
);
283 struct tcf_chain
*chain
;
288 INIT_LIST_HEAD(&block
->chain_list
);
289 INIT_LIST_HEAD(&block
->cb_list
);
291 /* Create chain 0 by default, it has to be always present. */
292 chain
= tcf_chain_create(block
, 0);
295 goto err_chain_create
;
297 WARN_ON(!ei
->chain_head_change
);
298 chain
->chain_head_change
= ei
->chain_head_change
;
299 chain
->chain_head_change_priv
= ei
->chain_head_change_priv
;
300 block
->net
= qdisc_net(q
);
302 tcf_block_offload_bind(block
, q
, ei
);
310 EXPORT_SYMBOL(tcf_block_get_ext
);
312 static void tcf_chain_head_change_dflt(struct tcf_proto
*tp_head
, void *priv
)
314 struct tcf_proto __rcu
**p_filter_chain
= priv
;
316 rcu_assign_pointer(*p_filter_chain
, tp_head
);
319 int tcf_block_get(struct tcf_block
**p_block
,
320 struct tcf_proto __rcu
**p_filter_chain
, struct Qdisc
*q
)
322 struct tcf_block_ext_info ei
= {
323 .chain_head_change
= tcf_chain_head_change_dflt
,
324 .chain_head_change_priv
= p_filter_chain
,
327 WARN_ON(!p_filter_chain
);
328 return tcf_block_get_ext(p_block
, q
, &ei
);
330 EXPORT_SYMBOL(tcf_block_get
);
332 static void tcf_block_put_final(struct work_struct
*work
)
334 struct tcf_block
*block
= container_of(work
, struct tcf_block
, work
);
335 struct tcf_chain
*chain
, *tmp
;
339 /* At this point, all the chains should have refcnt == 1. */
340 list_for_each_entry_safe(chain
, tmp
, &block
->chain_list
, list
)
341 tcf_chain_put(chain
);
346 /* XXX: Standalone actions are not allowed to jump to any chain, and bound
347 * actions should be all removed after flushing.
349 void tcf_block_put_ext(struct tcf_block
*block
, struct Qdisc
*q
,
350 struct tcf_block_ext_info
*ei
)
352 struct tcf_chain
*chain
;
354 /* Hold a refcnt for all chains, except 0, so that they don't disappear
355 * while we are iterating.
357 list_for_each_entry(chain
, &block
->chain_list
, list
)
359 tcf_chain_hold(chain
);
361 list_for_each_entry(chain
, &block
->chain_list
, list
)
362 tcf_chain_flush(chain
);
364 tcf_block_offload_unbind(block
, q
, ei
);
366 INIT_WORK(&block
->work
, tcf_block_put_final
);
367 /* Wait for existing RCU callbacks to cool down, make sure their works
368 * have been queued before this. We can not flush pending works here
369 * because we are holding the RTNL lock.
372 tcf_queue_work(&block
->work
);
374 EXPORT_SYMBOL(tcf_block_put_ext
);
376 void tcf_block_put(struct tcf_block
*block
)
378 struct tcf_block_ext_info ei
= {0, };
382 tcf_block_put_ext(block
, block
->q
, &ei
);
385 EXPORT_SYMBOL(tcf_block_put
);
387 struct tcf_block_cb
{
388 struct list_head list
;
395 void *tcf_block_cb_priv(struct tcf_block_cb
*block_cb
)
397 return block_cb
->cb_priv
;
399 EXPORT_SYMBOL(tcf_block_cb_priv
);
401 struct tcf_block_cb
*tcf_block_cb_lookup(struct tcf_block
*block
,
402 tc_setup_cb_t
*cb
, void *cb_ident
)
403 { struct tcf_block_cb
*block_cb
;
405 list_for_each_entry(block_cb
, &block
->cb_list
, list
)
406 if (block_cb
->cb
== cb
&& block_cb
->cb_ident
== cb_ident
)
410 EXPORT_SYMBOL(tcf_block_cb_lookup
);
412 void tcf_block_cb_incref(struct tcf_block_cb
*block_cb
)
416 EXPORT_SYMBOL(tcf_block_cb_incref
);
418 unsigned int tcf_block_cb_decref(struct tcf_block_cb
*block_cb
)
420 return --block_cb
->refcnt
;
422 EXPORT_SYMBOL(tcf_block_cb_decref
);
424 struct tcf_block_cb
*__tcf_block_cb_register(struct tcf_block
*block
,
425 tc_setup_cb_t
*cb
, void *cb_ident
,
428 struct tcf_block_cb
*block_cb
;
430 block_cb
= kzalloc(sizeof(*block_cb
), GFP_KERNEL
);
434 block_cb
->cb_ident
= cb_ident
;
435 block_cb
->cb_priv
= cb_priv
;
436 list_add(&block_cb
->list
, &block
->cb_list
);
439 EXPORT_SYMBOL(__tcf_block_cb_register
);
441 int tcf_block_cb_register(struct tcf_block
*block
,
442 tc_setup_cb_t
*cb
, void *cb_ident
,
445 struct tcf_block_cb
*block_cb
;
447 block_cb
= __tcf_block_cb_register(block
, cb
, cb_ident
, cb_priv
);
448 return block_cb
? 0 : -ENOMEM
;
450 EXPORT_SYMBOL(tcf_block_cb_register
);
452 void __tcf_block_cb_unregister(struct tcf_block_cb
*block_cb
)
454 list_del(&block_cb
->list
);
457 EXPORT_SYMBOL(__tcf_block_cb_unregister
);
459 void tcf_block_cb_unregister(struct tcf_block
*block
,
460 tc_setup_cb_t
*cb
, void *cb_ident
)
462 struct tcf_block_cb
*block_cb
;
464 block_cb
= tcf_block_cb_lookup(block
, cb
, cb_ident
);
467 __tcf_block_cb_unregister(block_cb
);
469 EXPORT_SYMBOL(tcf_block_cb_unregister
);
471 static int tcf_block_cb_call(struct tcf_block
*block
, enum tc_setup_type type
,
472 void *type_data
, bool err_stop
)
474 struct tcf_block_cb
*block_cb
;
478 list_for_each_entry(block_cb
, &block
->cb_list
, list
) {
479 err
= block_cb
->cb(type
, type_data
, block_cb
->cb_priv
);
490 /* Main classifier routine: scans classifier chain attached
491 * to this qdisc, (optionally) tests for protocol and asks
492 * specific classifiers.
494 int tcf_classify(struct sk_buff
*skb
, const struct tcf_proto
*tp
,
495 struct tcf_result
*res
, bool compat_mode
)
497 __be16 protocol
= tc_skb_protocol(skb
);
498 #ifdef CONFIG_NET_CLS_ACT
499 const int max_reclassify_loop
= 4;
500 const struct tcf_proto
*orig_tp
= tp
;
501 const struct tcf_proto
*first_tp
;
506 for (; tp
; tp
= rcu_dereference_bh(tp
->next
)) {
509 if (tp
->protocol
!= protocol
&&
510 tp
->protocol
!= htons(ETH_P_ALL
))
513 err
= tp
->classify(skb
, tp
, res
);
514 #ifdef CONFIG_NET_CLS_ACT
515 if (unlikely(err
== TC_ACT_RECLASSIFY
&& !compat_mode
)) {
518 } else if (unlikely(TC_ACT_EXT_CMP(err
, TC_ACT_GOTO_CHAIN
))) {
519 first_tp
= res
->goto_tp
;
527 return TC_ACT_UNSPEC
; /* signal: continue lookup */
528 #ifdef CONFIG_NET_CLS_ACT
530 if (unlikely(limit
++ >= max_reclassify_loop
)) {
531 net_notice_ratelimited("%s: reclassify loop, rule prio %u, protocol %02x\n",
532 tp
->q
->ops
->id
, tp
->prio
& 0xffff,
533 ntohs(tp
->protocol
));
538 protocol
= tc_skb_protocol(skb
);
542 EXPORT_SYMBOL(tcf_classify
);
544 struct tcf_chain_info
{
545 struct tcf_proto __rcu
**pprev
;
546 struct tcf_proto __rcu
*next
;
549 static struct tcf_proto
*tcf_chain_tp_prev(struct tcf_chain_info
*chain_info
)
551 return rtnl_dereference(*chain_info
->pprev
);
554 static void tcf_chain_tp_insert(struct tcf_chain
*chain
,
555 struct tcf_chain_info
*chain_info
,
556 struct tcf_proto
*tp
)
558 if (*chain_info
->pprev
== chain
->filter_chain
)
559 tcf_chain_head_change(chain
, tp
);
560 RCU_INIT_POINTER(tp
->next
, tcf_chain_tp_prev(chain_info
));
561 rcu_assign_pointer(*chain_info
->pprev
, tp
);
562 tcf_chain_hold(chain
);
565 static void tcf_chain_tp_remove(struct tcf_chain
*chain
,
566 struct tcf_chain_info
*chain_info
,
567 struct tcf_proto
*tp
)
569 struct tcf_proto
*next
= rtnl_dereference(chain_info
->next
);
571 if (tp
== chain
->filter_chain
)
572 tcf_chain_head_change(chain
, next
);
573 RCU_INIT_POINTER(*chain_info
->pprev
, next
);
574 tcf_chain_put(chain
);
577 static struct tcf_proto
*tcf_chain_tp_find(struct tcf_chain
*chain
,
578 struct tcf_chain_info
*chain_info
,
579 u32 protocol
, u32 prio
,
582 struct tcf_proto
**pprev
;
583 struct tcf_proto
*tp
;
585 /* Check the chain for existence of proto-tcf with this priority */
586 for (pprev
= &chain
->filter_chain
;
587 (tp
= rtnl_dereference(*pprev
)); pprev
= &tp
->next
) {
588 if (tp
->prio
>= prio
) {
589 if (tp
->prio
== prio
) {
591 (tp
->protocol
!= protocol
&& protocol
))
592 return ERR_PTR(-EINVAL
);
599 chain_info
->pprev
= pprev
;
600 chain_info
->next
= tp
? tp
->next
: NULL
;
604 static int tcf_fill_node(struct net
*net
, struct sk_buff
*skb
,
605 struct tcf_proto
*tp
, struct Qdisc
*q
, u32 parent
,
606 void *fh
, u32 portid
, u32 seq
, u16 flags
, int event
)
609 struct nlmsghdr
*nlh
;
610 unsigned char *b
= skb_tail_pointer(skb
);
612 nlh
= nlmsg_put(skb
, portid
, seq
, event
, sizeof(*tcm
), flags
);
615 tcm
= nlmsg_data(nlh
);
616 tcm
->tcm_family
= AF_UNSPEC
;
619 tcm
->tcm_ifindex
= qdisc_dev(q
)->ifindex
;
620 tcm
->tcm_parent
= parent
;
621 tcm
->tcm_info
= TC_H_MAKE(tp
->prio
, tp
->protocol
);
622 if (nla_put_string(skb
, TCA_KIND
, tp
->ops
->kind
))
623 goto nla_put_failure
;
624 if (nla_put_u32(skb
, TCA_CHAIN
, tp
->chain
->index
))
625 goto nla_put_failure
;
629 if (tp
->ops
->dump
&& tp
->ops
->dump(net
, tp
, fh
, skb
, tcm
) < 0)
630 goto nla_put_failure
;
632 nlh
->nlmsg_len
= skb_tail_pointer(skb
) - b
;
641 static int tfilter_notify(struct net
*net
, struct sk_buff
*oskb
,
642 struct nlmsghdr
*n
, struct tcf_proto
*tp
,
643 struct Qdisc
*q
, u32 parent
,
644 void *fh
, int event
, bool unicast
)
647 u32 portid
= oskb
? NETLINK_CB(oskb
).portid
: 0;
649 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
653 if (tcf_fill_node(net
, skb
, tp
, q
, parent
, fh
, portid
, n
->nlmsg_seq
,
654 n
->nlmsg_flags
, event
) <= 0) {
660 return netlink_unicast(net
->rtnl
, skb
, portid
, MSG_DONTWAIT
);
662 return rtnetlink_send(skb
, net
, portid
, RTNLGRP_TC
,
663 n
->nlmsg_flags
& NLM_F_ECHO
);
666 static int tfilter_del_notify(struct net
*net
, struct sk_buff
*oskb
,
667 struct nlmsghdr
*n
, struct tcf_proto
*tp
,
668 struct Qdisc
*q
, u32 parent
,
669 void *fh
, bool unicast
, bool *last
)
672 u32 portid
= oskb
? NETLINK_CB(oskb
).portid
: 0;
675 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
679 if (tcf_fill_node(net
, skb
, tp
, q
, parent
, fh
, portid
, n
->nlmsg_seq
,
680 n
->nlmsg_flags
, RTM_DELTFILTER
) <= 0) {
685 err
= tp
->ops
->delete(tp
, fh
, last
);
692 return netlink_unicast(net
->rtnl
, skb
, portid
, MSG_DONTWAIT
);
694 return rtnetlink_send(skb
, net
, portid
, RTNLGRP_TC
,
695 n
->nlmsg_flags
& NLM_F_ECHO
);
698 static void tfilter_notify_chain(struct net
*net
, struct sk_buff
*oskb
,
699 struct Qdisc
*q
, u32 parent
,
701 struct tcf_chain
*chain
, int event
)
703 struct tcf_proto
*tp
;
705 for (tp
= rtnl_dereference(chain
->filter_chain
);
706 tp
; tp
= rtnl_dereference(tp
->next
))
707 tfilter_notify(net
, oskb
, n
, tp
, q
, parent
, 0, event
, false);
710 /* Add/change/delete/get a filter node */
712 static int tc_ctl_tfilter(struct sk_buff
*skb
, struct nlmsghdr
*n
,
713 struct netlink_ext_ack
*extack
)
715 struct net
*net
= sock_net(skb
->sk
);
716 struct nlattr
*tca
[TCA_MAX
+ 1];
723 struct net_device
*dev
;
725 struct tcf_chain_info chain_info
;
726 struct tcf_chain
*chain
= NULL
;
727 struct tcf_block
*block
;
728 struct tcf_proto
*tp
;
729 const struct Qdisc_class_ops
*cops
;
735 if ((n
->nlmsg_type
!= RTM_GETTFILTER
) &&
736 !netlink_ns_capable(skb
, net
->user_ns
, CAP_NET_ADMIN
))
742 err
= nlmsg_parse(n
, sizeof(*t
), tca
, TCA_MAX
, NULL
, extack
);
747 protocol
= TC_H_MIN(t
->tcm_info
);
748 prio
= TC_H_MAJ(t
->tcm_info
);
749 prio_allocate
= false;
750 parent
= t
->tcm_parent
;
754 switch (n
->nlmsg_type
) {
756 if (protocol
|| t
->tcm_handle
|| tca
[TCA_KIND
])
760 /* If no priority is provided by the user,
763 if (n
->nlmsg_flags
& NLM_F_CREATE
) {
764 prio
= TC_H_MAKE(0x80000000U
, 0U);
765 prio_allocate
= true;
774 /* Find head of filter chain. */
777 dev
= __dev_get_by_index(net
, t
->tcm_ifindex
);
786 q
= qdisc_lookup(dev
, TC_H_MAJ(t
->tcm_parent
));
791 /* Is it classful? */
792 cops
= q
->ops
->cl_ops
;
796 if (!cops
->tcf_block
)
799 /* Do we search for filter, attached to class? */
800 if (TC_H_MIN(parent
)) {
801 cl
= cops
->find(q
, parent
);
806 /* And the last stroke */
807 block
= cops
->tcf_block(q
, cl
);
813 chain_index
= tca
[TCA_CHAIN
] ? nla_get_u32(tca
[TCA_CHAIN
]) : 0;
814 if (chain_index
> TC_ACT_EXT_VAL_MASK
) {
818 chain
= tcf_chain_get(block
, chain_index
,
819 n
->nlmsg_type
== RTM_NEWTFILTER
);
821 err
= n
->nlmsg_type
== RTM_NEWTFILTER
? -ENOMEM
: -EINVAL
;
825 if (n
->nlmsg_type
== RTM_DELTFILTER
&& prio
== 0) {
826 tfilter_notify_chain(net
, skb
, q
, parent
, n
,
827 chain
, RTM_DELTFILTER
);
828 tcf_chain_flush(chain
);
833 tp
= tcf_chain_tp_find(chain
, &chain_info
, protocol
,
834 prio
, prio_allocate
);
841 /* Proto-tcf does not exist, create new one */
843 if (tca
[TCA_KIND
] == NULL
|| !protocol
) {
848 if (n
->nlmsg_type
!= RTM_NEWTFILTER
||
849 !(n
->nlmsg_flags
& NLM_F_CREATE
)) {
855 prio
= tcf_auto_prio(tcf_chain_tp_prev(&chain_info
));
857 tp
= tcf_proto_create(nla_data(tca
[TCA_KIND
]),
858 protocol
, prio
, parent
, q
, chain
);
864 } else if (tca
[TCA_KIND
] && nla_strcmp(tca
[TCA_KIND
], tp
->ops
->kind
)) {
869 fh
= tp
->ops
->get(tp
, t
->tcm_handle
);
872 if (n
->nlmsg_type
== RTM_DELTFILTER
&& t
->tcm_handle
== 0) {
873 tcf_chain_tp_remove(chain
, &chain_info
, tp
);
874 tfilter_notify(net
, skb
, n
, tp
, q
, parent
, fh
,
875 RTM_DELTFILTER
, false);
876 tcf_proto_destroy(tp
);
881 if (n
->nlmsg_type
!= RTM_NEWTFILTER
||
882 !(n
->nlmsg_flags
& NLM_F_CREATE
)) {
889 switch (n
->nlmsg_type
) {
891 if (n
->nlmsg_flags
& NLM_F_EXCL
) {
893 tcf_proto_destroy(tp
);
899 err
= tfilter_del_notify(net
, skb
, n
, tp
, q
, parent
,
904 tcf_chain_tp_remove(chain
, &chain_info
, tp
);
905 tcf_proto_destroy(tp
);
909 err
= tfilter_notify(net
, skb
, n
, tp
, q
, parent
, fh
,
910 RTM_NEWTFILTER
, true);
918 err
= tp
->ops
->change(net
, skb
, tp
, cl
, t
->tcm_handle
, tca
, &fh
,
919 n
->nlmsg_flags
& NLM_F_CREATE
? TCA_ACT_NOREPLACE
: TCA_ACT_REPLACE
);
922 tcf_chain_tp_insert(chain
, &chain_info
, tp
);
923 tfilter_notify(net
, skb
, n
, tp
, q
, parent
, fh
,
924 RTM_NEWTFILTER
, false);
927 tcf_proto_destroy(tp
);
932 tcf_chain_put(chain
);
934 /* Replay the request. */
939 struct tcf_dump_args
{
942 struct netlink_callback
*cb
;
947 static int tcf_node_dump(struct tcf_proto
*tp
, void *n
, struct tcf_walker
*arg
)
949 struct tcf_dump_args
*a
= (void *)arg
;
950 struct net
*net
= sock_net(a
->skb
->sk
);
952 return tcf_fill_node(net
, a
->skb
, tp
, a
->q
, a
->parent
,
953 n
, NETLINK_CB(a
->cb
->skb
).portid
,
954 a
->cb
->nlh
->nlmsg_seq
, NLM_F_MULTI
,
958 static bool tcf_chain_dump(struct tcf_chain
*chain
, struct Qdisc
*q
, u32 parent
,
959 struct sk_buff
*skb
, struct netlink_callback
*cb
,
960 long index_start
, long *p_index
)
962 struct net
*net
= sock_net(skb
->sk
);
963 struct tcmsg
*tcm
= nlmsg_data(cb
->nlh
);
964 struct tcf_dump_args arg
;
965 struct tcf_proto
*tp
;
967 for (tp
= rtnl_dereference(chain
->filter_chain
);
968 tp
; tp
= rtnl_dereference(tp
->next
), (*p_index
)++) {
969 if (*p_index
< index_start
)
971 if (TC_H_MAJ(tcm
->tcm_info
) &&
972 TC_H_MAJ(tcm
->tcm_info
) != tp
->prio
)
974 if (TC_H_MIN(tcm
->tcm_info
) &&
975 TC_H_MIN(tcm
->tcm_info
) != tp
->protocol
)
977 if (*p_index
> index_start
)
978 memset(&cb
->args
[1], 0,
979 sizeof(cb
->args
) - sizeof(cb
->args
[0]));
980 if (cb
->args
[1] == 0) {
981 if (tcf_fill_node(net
, skb
, tp
, q
, parent
, 0,
982 NETLINK_CB(cb
->skb
).portid
,
983 cb
->nlh
->nlmsg_seq
, NLM_F_MULTI
,
984 RTM_NEWTFILTER
) <= 0)
991 arg
.w
.fn
= tcf_node_dump
;
997 arg
.w
.skip
= cb
->args
[1] - 1;
999 tp
->ops
->walk(tp
, &arg
.w
);
1000 cb
->args
[1] = arg
.w
.count
+ 1;
1007 /* called with RTNL */
1008 static int tc_dump_tfilter(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1010 struct net
*net
= sock_net(skb
->sk
);
1011 struct nlattr
*tca
[TCA_MAX
+ 1];
1012 struct net_device
*dev
;
1014 struct tcf_block
*block
;
1015 struct tcf_chain
*chain
;
1016 struct tcmsg
*tcm
= nlmsg_data(cb
->nlh
);
1017 unsigned long cl
= 0;
1018 const struct Qdisc_class_ops
*cops
;
1024 if (nlmsg_len(cb
->nlh
) < sizeof(*tcm
))
1027 err
= nlmsg_parse(cb
->nlh
, sizeof(*tcm
), tca
, TCA_MAX
, NULL
, NULL
);
1031 dev
= __dev_get_by_index(net
, tcm
->tcm_ifindex
);
1035 parent
= tcm
->tcm_parent
;
1040 q
= qdisc_lookup(dev
, TC_H_MAJ(tcm
->tcm_parent
));
1044 cops
= q
->ops
->cl_ops
;
1047 if (!cops
->tcf_block
)
1049 if (TC_H_MIN(tcm
->tcm_parent
)) {
1050 cl
= cops
->find(q
, tcm
->tcm_parent
);
1054 block
= cops
->tcf_block(q
, cl
);
1058 index_start
= cb
->args
[0];
1061 list_for_each_entry(chain
, &block
->chain_list
, list
) {
1062 if (tca
[TCA_CHAIN
] &&
1063 nla_get_u32(tca
[TCA_CHAIN
]) != chain
->index
)
1065 if (!tcf_chain_dump(chain
, q
, parent
, skb
, cb
,
1066 index_start
, &index
))
1070 cb
->args
[0] = index
;
1076 void tcf_exts_destroy(struct tcf_exts
*exts
)
1078 #ifdef CONFIG_NET_CLS_ACT
1082 tcf_exts_to_list(exts
, &actions
);
1083 tcf_action_destroy(&actions
, TCA_ACT_UNBIND
);
1084 kfree(exts
->actions
);
1085 exts
->nr_actions
= 0;
1088 EXPORT_SYMBOL(tcf_exts_destroy
);
1090 int tcf_exts_validate(struct net
*net
, struct tcf_proto
*tp
, struct nlattr
**tb
,
1091 struct nlattr
*rate_tlv
, struct tcf_exts
*exts
, bool ovr
)
1093 #ifdef CONFIG_NET_CLS_ACT
1095 struct tc_action
*act
;
1097 if (exts
->police
&& tb
[exts
->police
]) {
1098 act
= tcf_action_init_1(net
, tp
, tb
[exts
->police
],
1099 rate_tlv
, "police", ovr
,
1102 return PTR_ERR(act
);
1104 act
->type
= exts
->type
= TCA_OLD_COMPAT
;
1105 exts
->actions
[0] = act
;
1106 exts
->nr_actions
= 1;
1107 } else if (exts
->action
&& tb
[exts
->action
]) {
1111 err
= tcf_action_init(net
, tp
, tb
[exts
->action
],
1112 rate_tlv
, NULL
, ovr
, TCA_ACT_BIND
,
1116 list_for_each_entry(act
, &actions
, list
)
1117 exts
->actions
[i
++] = act
;
1118 exts
->nr_actions
= i
;
1123 if ((exts
->action
&& tb
[exts
->action
]) ||
1124 (exts
->police
&& tb
[exts
->police
]))
1130 EXPORT_SYMBOL(tcf_exts_validate
);
1132 void tcf_exts_change(struct tcf_exts
*dst
, struct tcf_exts
*src
)
1134 #ifdef CONFIG_NET_CLS_ACT
1135 struct tcf_exts old
= *dst
;
1138 tcf_exts_destroy(&old
);
1141 EXPORT_SYMBOL(tcf_exts_change
);
1143 #ifdef CONFIG_NET_CLS_ACT
1144 static struct tc_action
*tcf_exts_first_act(struct tcf_exts
*exts
)
1146 if (exts
->nr_actions
== 0)
1149 return exts
->actions
[0];
1153 int tcf_exts_dump(struct sk_buff
*skb
, struct tcf_exts
*exts
)
1155 #ifdef CONFIG_NET_CLS_ACT
1156 struct nlattr
*nest
;
1158 if (exts
->action
&& tcf_exts_has_actions(exts
)) {
1160 * again for backward compatible mode - we want
1161 * to work with both old and new modes of entering
1162 * tc data even if iproute2 was newer - jhs
1164 if (exts
->type
!= TCA_OLD_COMPAT
) {
1167 nest
= nla_nest_start(skb
, exts
->action
);
1169 goto nla_put_failure
;
1171 tcf_exts_to_list(exts
, &actions
);
1172 if (tcf_action_dump(skb
, &actions
, 0, 0) < 0)
1173 goto nla_put_failure
;
1174 nla_nest_end(skb
, nest
);
1175 } else if (exts
->police
) {
1176 struct tc_action
*act
= tcf_exts_first_act(exts
);
1177 nest
= nla_nest_start(skb
, exts
->police
);
1178 if (nest
== NULL
|| !act
)
1179 goto nla_put_failure
;
1180 if (tcf_action_dump_old(skb
, act
, 0, 0) < 0)
1181 goto nla_put_failure
;
1182 nla_nest_end(skb
, nest
);
1188 nla_nest_cancel(skb
, nest
);
1194 EXPORT_SYMBOL(tcf_exts_dump
);
1197 int tcf_exts_dump_stats(struct sk_buff
*skb
, struct tcf_exts
*exts
)
1199 #ifdef CONFIG_NET_CLS_ACT
1200 struct tc_action
*a
= tcf_exts_first_act(exts
);
1201 if (a
!= NULL
&& tcf_action_copy_stats(skb
, a
, 1) < 0)
1206 EXPORT_SYMBOL(tcf_exts_dump_stats
);
1208 static int tc_exts_setup_cb_egdev_call(struct tcf_exts
*exts
,
1209 enum tc_setup_type type
,
1210 void *type_data
, bool err_stop
)
1213 #ifdef CONFIG_NET_CLS_ACT
1214 const struct tc_action
*a
;
1215 struct net_device
*dev
;
1218 if (!tcf_exts_has_actions(exts
))
1221 for (i
= 0; i
< exts
->nr_actions
; i
++) {
1222 a
= exts
->actions
[i
];
1223 if (!a
->ops
->get_dev
)
1225 dev
= a
->ops
->get_dev(a
);
1228 ret
= tc_setup_cb_egdev_call(dev
, type
, type_data
, err_stop
);
1237 int tc_setup_cb_call(struct tcf_block
*block
, struct tcf_exts
*exts
,
1238 enum tc_setup_type type
, void *type_data
, bool err_stop
)
1243 ret
= tcf_block_cb_call(block
, type
, type_data
, err_stop
);
1250 ret
= tc_exts_setup_cb_egdev_call(exts
, type
, type_data
, err_stop
);
1257 EXPORT_SYMBOL(tc_setup_cb_call
);
1259 static int __init
tc_filter_init(void)
1261 tc_filter_wq
= alloc_ordered_workqueue("tc_filter_workqueue", 0);
1265 rtnl_register(PF_UNSPEC
, RTM_NEWTFILTER
, tc_ctl_tfilter
, NULL
, 0);
1266 rtnl_register(PF_UNSPEC
, RTM_DELTFILTER
, tc_ctl_tfilter
, NULL
, 0);
1267 rtnl_register(PF_UNSPEC
, RTM_GETTFILTER
, tc_ctl_tfilter
,
1268 tc_dump_tfilter
, 0);
1273 subsys_initcall(tc_filter_init
);