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 <linux/idr.h>
28 #include <net/net_namespace.h>
30 #include <net/netlink.h>
31 #include <net/pkt_sched.h>
32 #include <net/pkt_cls.h>
34 /* The list of all installed classifier types */
35 static LIST_HEAD(tcf_proto_base);
37 /* Protects list of registered TC modules. It is pure SMP lock. */
38 static DEFINE_RWLOCK(cls_mod_lock);
40 /* Find classifier type by string name */
42 static const struct tcf_proto_ops *__tcf_proto_lookup_ops(const char *kind)
44 const struct tcf_proto_ops *t, *res = NULL;
47 read_lock(&cls_mod_lock);
48 list_for_each_entry(t, &tcf_proto_base, head) {
49 if (strcmp(kind, t->kind) == 0) {
50 if (try_module_get(t->owner))
55 read_unlock(&cls_mod_lock);
60 static const struct tcf_proto_ops *
61 tcf_proto_lookup_ops(const char *kind, struct netlink_ext_ack *extack)
63 const struct tcf_proto_ops *ops;
65 ops = __tcf_proto_lookup_ops(kind);
70 request_module("cls_%s", kind);
72 ops = __tcf_proto_lookup_ops(kind);
73 /* We dropped the RTNL semaphore in order to perform
74 * the module load. So, even if we succeeded in loading
75 * the module we have to replay the request. We indicate
79 module_put(ops->owner);
80 return ERR_PTR(-EAGAIN);
83 NL_SET_ERR_MSG(extack, "TC classifier not found");
84 return ERR_PTR(-ENOENT);
87 /* Register(unregister) new classifier type */
89 int register_tcf_proto_ops(struct tcf_proto_ops *ops)
91 struct tcf_proto_ops *t;
94 write_lock(&cls_mod_lock);
95 list_for_each_entry(t, &tcf_proto_base, head)
96 if (!strcmp(ops->kind, t->kind))
99 list_add_tail(&ops->head, &tcf_proto_base);
102 write_unlock(&cls_mod_lock);
105 EXPORT_SYMBOL(register_tcf_proto_ops);
107 static struct workqueue_struct *tc_filter_wq;
109 int unregister_tcf_proto_ops(struct tcf_proto_ops *ops)
111 struct tcf_proto_ops *t;
114 /* Wait for outstanding call_rcu()s, if any, from a
115 * tcf_proto_ops's destroy() handler.
118 flush_workqueue(tc_filter_wq);
120 write_lock(&cls_mod_lock);
121 list_for_each_entry(t, &tcf_proto_base, head) {
128 write_unlock(&cls_mod_lock);
131 EXPORT_SYMBOL(unregister_tcf_proto_ops);
133 bool tcf_queue_work(struct rcu_work *rwork, work_func_t func)
135 INIT_RCU_WORK(rwork, func);
136 return queue_rcu_work(tc_filter_wq, rwork);
138 EXPORT_SYMBOL(tcf_queue_work);
140 /* Select new prio value from the range, managed by kernel. */
142 static inline u32 tcf_auto_prio(struct tcf_proto *tp)
144 u32 first = TC_H_MAKE(0xC0000000U, 0U);
147 first = tp->prio - 1;
149 return TC_H_MAJ(first);
152 static struct tcf_proto *tcf_proto_create(const char *kind, u32 protocol,
153 u32 prio, struct tcf_chain *chain,
154 struct netlink_ext_ack *extack)
156 struct tcf_proto *tp;
159 tp = kzalloc(sizeof(*tp), GFP_KERNEL);
161 return ERR_PTR(-ENOBUFS);
163 tp->ops = tcf_proto_lookup_ops(kind, extack);
164 if (IS_ERR(tp->ops)) {
165 err = PTR_ERR(tp->ops);
168 tp->classify = tp->ops->classify;
169 tp->protocol = protocol;
173 err = tp->ops->init(tp);
175 module_put(tp->ops->owner);
185 static void tcf_proto_destroy(struct tcf_proto *tp,
186 struct netlink_ext_ack *extack)
188 tp->ops->destroy(tp, extack);
189 module_put(tp->ops->owner);
193 struct tcf_filter_chain_list_item {
194 struct list_head list;
195 tcf_chain_head_change_t *chain_head_change;
196 void *chain_head_change_priv;
199 static struct tcf_chain *tcf_chain_create(struct tcf_block *block,
202 struct tcf_chain *chain;
204 chain = kzalloc(sizeof(*chain), GFP_KERNEL);
207 list_add_tail(&chain->list, &block->chain_list);
208 chain->block = block;
209 chain->index = chain_index;
212 block->chain0.chain = chain;
216 static void tcf_chain_head_change_item(struct tcf_filter_chain_list_item *item,
217 struct tcf_proto *tp_head)
219 if (item->chain_head_change)
220 item->chain_head_change(tp_head, item->chain_head_change_priv);
223 static void tcf_chain0_head_change(struct tcf_chain *chain,
224 struct tcf_proto *tp_head)
226 struct tcf_filter_chain_list_item *item;
227 struct tcf_block *block = chain->block;
231 list_for_each_entry(item, &block->chain0.filter_chain_list, list)
232 tcf_chain_head_change_item(item, tp_head);
235 static void tcf_chain_flush(struct tcf_chain *chain)
237 struct tcf_proto *tp = rtnl_dereference(chain->filter_chain);
239 tcf_chain0_head_change(chain, NULL);
241 RCU_INIT_POINTER(chain->filter_chain, tp->next);
242 tcf_proto_destroy(tp, NULL);
243 tp = rtnl_dereference(chain->filter_chain);
244 tcf_chain_put(chain);
248 static void tcf_chain_destroy(struct tcf_chain *chain)
250 struct tcf_block *block = chain->block;
252 list_del(&chain->list);
254 block->chain0.chain = NULL;
256 if (list_empty(&block->chain_list) && block->refcnt == 0)
260 static void tcf_chain_hold(struct tcf_chain *chain)
265 static struct tcf_chain *tcf_chain_lookup(struct tcf_block *block,
268 struct tcf_chain *chain;
270 list_for_each_entry(chain, &block->chain_list, list) {
271 if (chain->index == chain_index)
277 static int tc_chain_notify(struct tcf_chain *chain, struct sk_buff *oskb,
278 u32 seq, u16 flags, int event, bool unicast);
280 struct tcf_chain *tcf_chain_get(struct tcf_block *block, u32 chain_index,
283 struct tcf_chain *chain = tcf_chain_lookup(block, chain_index);
286 tcf_chain_hold(chain);
292 chain = tcf_chain_create(block, chain_index);
295 tc_chain_notify(chain, NULL, 0, NLM_F_CREATE | NLM_F_EXCL,
296 RTM_NEWCHAIN, false);
299 EXPORT_SYMBOL(tcf_chain_get);
301 static void tc_chain_tmplt_del(struct tcf_chain *chain);
303 void tcf_chain_put(struct tcf_chain *chain)
305 if (--chain->refcnt == 0) {
306 tc_chain_notify(chain, NULL, 0, 0, RTM_DELCHAIN, false);
307 tc_chain_tmplt_del(chain);
308 tcf_chain_destroy(chain);
311 EXPORT_SYMBOL(tcf_chain_put);
313 static void tcf_chain_put_explicitly_created(struct tcf_chain *chain)
315 if (chain->explicitly_created)
316 tcf_chain_put(chain);
319 static bool tcf_block_offload_in_use(struct tcf_block *block)
321 return block->offloadcnt;
324 static int tcf_block_offload_cmd(struct tcf_block *block,
325 struct net_device *dev,
326 struct tcf_block_ext_info *ei,
327 enum tc_block_command command,
328 struct netlink_ext_ack *extack)
330 struct tc_block_offload bo = {};
332 bo.command = command;
333 bo.binder_type = ei->binder_type;
336 return dev->netdev_ops->ndo_setup_tc(dev, TC_SETUP_BLOCK, &bo);
339 static int tcf_block_offload_bind(struct tcf_block *block, struct Qdisc *q,
340 struct tcf_block_ext_info *ei,
341 struct netlink_ext_ack *extack)
343 struct net_device *dev = q->dev_queue->dev;
346 if (!dev->netdev_ops->ndo_setup_tc)
347 goto no_offload_dev_inc;
349 /* If tc offload feature is disabled and the block we try to bind
350 * to already has some offloaded filters, forbid to bind.
352 if (!tc_can_offload(dev) && tcf_block_offload_in_use(block)) {
353 NL_SET_ERR_MSG(extack, "Bind to offloaded block failed as dev has offload disabled");
357 err = tcf_block_offload_cmd(block, dev, ei, TC_BLOCK_BIND, extack);
358 if (err == -EOPNOTSUPP)
359 goto no_offload_dev_inc;
363 if (tcf_block_offload_in_use(block))
365 block->nooffloaddevcnt++;
369 static void tcf_block_offload_unbind(struct tcf_block *block, struct Qdisc *q,
370 struct tcf_block_ext_info *ei)
372 struct net_device *dev = q->dev_queue->dev;
375 if (!dev->netdev_ops->ndo_setup_tc)
376 goto no_offload_dev_dec;
377 err = tcf_block_offload_cmd(block, dev, ei, TC_BLOCK_UNBIND, NULL);
378 if (err == -EOPNOTSUPP)
379 goto no_offload_dev_dec;
383 WARN_ON(block->nooffloaddevcnt-- == 0);
387 tcf_chain0_head_change_cb_add(struct tcf_block *block,
388 struct tcf_block_ext_info *ei,
389 struct netlink_ext_ack *extack)
391 struct tcf_chain *chain0 = block->chain0.chain;
392 struct tcf_filter_chain_list_item *item;
394 item = kmalloc(sizeof(*item), GFP_KERNEL);
396 NL_SET_ERR_MSG(extack, "Memory allocation for head change callback item failed");
399 item->chain_head_change = ei->chain_head_change;
400 item->chain_head_change_priv = ei->chain_head_change_priv;
401 if (chain0 && chain0->filter_chain)
402 tcf_chain_head_change_item(item, chain0->filter_chain);
403 list_add(&item->list, &block->chain0.filter_chain_list);
408 tcf_chain0_head_change_cb_del(struct tcf_block *block,
409 struct tcf_block_ext_info *ei)
411 struct tcf_chain *chain0 = block->chain0.chain;
412 struct tcf_filter_chain_list_item *item;
414 list_for_each_entry(item, &block->chain0.filter_chain_list, list) {
415 if ((!ei->chain_head_change && !ei->chain_head_change_priv) ||
416 (item->chain_head_change == ei->chain_head_change &&
417 item->chain_head_change_priv == ei->chain_head_change_priv)) {
419 tcf_chain_head_change_item(item, NULL);
420 list_del(&item->list);
432 static unsigned int tcf_net_id;
434 static int tcf_block_insert(struct tcf_block *block, struct net *net,
435 struct netlink_ext_ack *extack)
437 struct tcf_net *tn = net_generic(net, tcf_net_id);
439 return idr_alloc_u32(&tn->idr, block, &block->index, block->index,
443 static void tcf_block_remove(struct tcf_block *block, struct net *net)
445 struct tcf_net *tn = net_generic(net, tcf_net_id);
447 idr_remove(&tn->idr, block->index);
450 static struct tcf_block *tcf_block_create(struct net *net, struct Qdisc *q,
452 struct netlink_ext_ack *extack)
454 struct tcf_block *block;
456 block = kzalloc(sizeof(*block), GFP_KERNEL);
458 NL_SET_ERR_MSG(extack, "Memory allocation for block failed");
459 return ERR_PTR(-ENOMEM);
461 INIT_LIST_HEAD(&block->chain_list);
462 INIT_LIST_HEAD(&block->cb_list);
463 INIT_LIST_HEAD(&block->owner_list);
464 INIT_LIST_HEAD(&block->chain0.filter_chain_list);
468 block->index = block_index;
470 /* Don't store q pointer for blocks which are shared */
471 if (!tcf_block_shared(block))
476 static struct tcf_block *tcf_block_lookup(struct net *net, u32 block_index)
478 struct tcf_net *tn = net_generic(net, tcf_net_id);
480 return idr_find(&tn->idr, block_index);
484 * Set q, parent, cl when appropriate.
487 static struct tcf_block *tcf_block_find(struct net *net, struct Qdisc **q,
488 u32 *parent, unsigned long *cl,
489 int ifindex, u32 block_index,
490 struct netlink_ext_ack *extack)
492 struct tcf_block *block;
494 if (ifindex == TCM_IFINDEX_MAGIC_BLOCK) {
495 block = tcf_block_lookup(net, block_index);
497 NL_SET_ERR_MSG(extack, "Block of given index was not found");
498 return ERR_PTR(-EINVAL);
501 const struct Qdisc_class_ops *cops;
502 struct net_device *dev;
505 dev = __dev_get_by_index(net, ifindex);
507 return ERR_PTR(-ENODEV);
512 *parent = (*q)->handle;
514 *q = qdisc_lookup(dev, TC_H_MAJ(*parent));
516 NL_SET_ERR_MSG(extack, "Parent Qdisc doesn't exists");
517 return ERR_PTR(-EINVAL);
521 /* Is it classful? */
522 cops = (*q)->ops->cl_ops;
524 NL_SET_ERR_MSG(extack, "Qdisc not classful");
525 return ERR_PTR(-EINVAL);
528 if (!cops->tcf_block) {
529 NL_SET_ERR_MSG(extack, "Class doesn't support blocks");
530 return ERR_PTR(-EOPNOTSUPP);
533 /* Do we search for filter, attached to class? */
534 if (TC_H_MIN(*parent)) {
535 *cl = cops->find(*q, *parent);
537 NL_SET_ERR_MSG(extack, "Specified class doesn't exist");
538 return ERR_PTR(-ENOENT);
542 /* And the last stroke */
543 block = cops->tcf_block(*q, *cl, extack);
545 return ERR_PTR(-EINVAL);
546 if (tcf_block_shared(block)) {
547 NL_SET_ERR_MSG(extack, "This filter block is shared. Please use the block index to manipulate the filters");
548 return ERR_PTR(-EOPNOTSUPP);
555 struct tcf_block_owner_item {
556 struct list_head list;
558 enum tcf_block_binder_type binder_type;
562 tcf_block_owner_netif_keep_dst(struct tcf_block *block,
564 enum tcf_block_binder_type binder_type)
566 if (block->keep_dst &&
567 binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS &&
568 binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_EGRESS)
569 netif_keep_dst(qdisc_dev(q));
572 void tcf_block_netif_keep_dst(struct tcf_block *block)
574 struct tcf_block_owner_item *item;
576 block->keep_dst = true;
577 list_for_each_entry(item, &block->owner_list, list)
578 tcf_block_owner_netif_keep_dst(block, item->q,
581 EXPORT_SYMBOL(tcf_block_netif_keep_dst);
583 static int tcf_block_owner_add(struct tcf_block *block,
585 enum tcf_block_binder_type binder_type)
587 struct tcf_block_owner_item *item;
589 item = kmalloc(sizeof(*item), GFP_KERNEL);
593 item->binder_type = binder_type;
594 list_add(&item->list, &block->owner_list);
598 static void tcf_block_owner_del(struct tcf_block *block,
600 enum tcf_block_binder_type binder_type)
602 struct tcf_block_owner_item *item;
604 list_for_each_entry(item, &block->owner_list, list) {
605 if (item->q == q && item->binder_type == binder_type) {
606 list_del(&item->list);
614 int tcf_block_get_ext(struct tcf_block **p_block, struct Qdisc *q,
615 struct tcf_block_ext_info *ei,
616 struct netlink_ext_ack *extack)
618 struct net *net = qdisc_net(q);
619 struct tcf_block *block = NULL;
620 bool created = false;
623 if (ei->block_index) {
624 /* block_index not 0 means the shared block is requested */
625 block = tcf_block_lookup(net, ei->block_index);
631 block = tcf_block_create(net, q, ei->block_index, extack);
633 return PTR_ERR(block);
635 if (tcf_block_shared(block)) {
636 err = tcf_block_insert(block, net, extack);
638 goto err_block_insert;
642 err = tcf_block_owner_add(block, q, ei->binder_type);
644 goto err_block_owner_add;
646 tcf_block_owner_netif_keep_dst(block, q, ei->binder_type);
648 err = tcf_chain0_head_change_cb_add(block, ei, extack);
650 goto err_chain0_head_change_cb_add;
652 err = tcf_block_offload_bind(block, q, ei, extack);
654 goto err_block_offload_bind;
659 err_block_offload_bind:
660 tcf_chain0_head_change_cb_del(block, ei);
661 err_chain0_head_change_cb_add:
662 tcf_block_owner_del(block, q, ei->binder_type);
665 if (tcf_block_shared(block))
666 tcf_block_remove(block, net);
674 EXPORT_SYMBOL(tcf_block_get_ext);
676 static void tcf_chain_head_change_dflt(struct tcf_proto *tp_head, void *priv)
678 struct tcf_proto __rcu **p_filter_chain = priv;
680 rcu_assign_pointer(*p_filter_chain, tp_head);
683 int tcf_block_get(struct tcf_block **p_block,
684 struct tcf_proto __rcu **p_filter_chain, struct Qdisc *q,
685 struct netlink_ext_ack *extack)
687 struct tcf_block_ext_info ei = {
688 .chain_head_change = tcf_chain_head_change_dflt,
689 .chain_head_change_priv = p_filter_chain,
692 WARN_ON(!p_filter_chain);
693 return tcf_block_get_ext(p_block, q, &ei, extack);
695 EXPORT_SYMBOL(tcf_block_get);
697 /* XXX: Standalone actions are not allowed to jump to any chain, and bound
698 * actions should be all removed after flushing.
700 void tcf_block_put_ext(struct tcf_block *block, struct Qdisc *q,
701 struct tcf_block_ext_info *ei)
703 struct tcf_chain *chain, *tmp;
707 tcf_chain0_head_change_cb_del(block, ei);
708 tcf_block_owner_del(block, q, ei->binder_type);
710 if (block->refcnt == 1) {
711 if (tcf_block_shared(block))
712 tcf_block_remove(block, block->net);
714 /* Hold a refcnt for all chains, so that they don't disappear
715 * while we are iterating.
717 list_for_each_entry(chain, &block->chain_list, list)
718 tcf_chain_hold(chain);
720 list_for_each_entry(chain, &block->chain_list, list)
721 tcf_chain_flush(chain);
724 tcf_block_offload_unbind(block, q, ei);
726 if (block->refcnt == 1) {
727 /* At this point, all the chains should have refcnt >= 1. */
728 list_for_each_entry_safe(chain, tmp, &block->chain_list, list) {
729 tcf_chain_put_explicitly_created(chain);
730 tcf_chain_put(chain);
734 if (list_empty(&block->chain_list))
738 EXPORT_SYMBOL(tcf_block_put_ext);
740 void tcf_block_put(struct tcf_block *block)
742 struct tcf_block_ext_info ei = {0, };
746 tcf_block_put_ext(block, block->q, &ei);
749 EXPORT_SYMBOL(tcf_block_put);
751 struct tcf_block_cb {
752 struct list_head list;
759 void *tcf_block_cb_priv(struct tcf_block_cb *block_cb)
761 return block_cb->cb_priv;
763 EXPORT_SYMBOL(tcf_block_cb_priv);
765 struct tcf_block_cb *tcf_block_cb_lookup(struct tcf_block *block,
766 tc_setup_cb_t *cb, void *cb_ident)
767 { struct tcf_block_cb *block_cb;
769 list_for_each_entry(block_cb, &block->cb_list, list)
770 if (block_cb->cb == cb && block_cb->cb_ident == cb_ident)
774 EXPORT_SYMBOL(tcf_block_cb_lookup);
776 void tcf_block_cb_incref(struct tcf_block_cb *block_cb)
780 EXPORT_SYMBOL(tcf_block_cb_incref);
782 unsigned int tcf_block_cb_decref(struct tcf_block_cb *block_cb)
784 return --block_cb->refcnt;
786 EXPORT_SYMBOL(tcf_block_cb_decref);
789 tcf_block_playback_offloads(struct tcf_block *block, tc_setup_cb_t *cb,
790 void *cb_priv, bool add, bool offload_in_use,
791 struct netlink_ext_ack *extack)
793 struct tcf_chain *chain;
794 struct tcf_proto *tp;
797 list_for_each_entry(chain, &block->chain_list, list) {
798 for (tp = rtnl_dereference(chain->filter_chain); tp;
799 tp = rtnl_dereference(tp->next)) {
800 if (tp->ops->reoffload) {
801 err = tp->ops->reoffload(tp, add, cb, cb_priv,
804 goto err_playback_remove;
805 } else if (add && offload_in_use) {
807 NL_SET_ERR_MSG(extack, "Filter HW offload failed - classifier without re-offloading support");
808 goto err_playback_remove;
816 tcf_block_playback_offloads(block, cb, cb_priv, false, offload_in_use,
821 struct tcf_block_cb *__tcf_block_cb_register(struct tcf_block *block,
822 tc_setup_cb_t *cb, void *cb_ident,
824 struct netlink_ext_ack *extack)
826 struct tcf_block_cb *block_cb;
829 /* Replay any already present rules */
830 err = tcf_block_playback_offloads(block, cb, cb_priv, true,
831 tcf_block_offload_in_use(block),
836 block_cb = kzalloc(sizeof(*block_cb), GFP_KERNEL);
838 return ERR_PTR(-ENOMEM);
840 block_cb->cb_ident = cb_ident;
841 block_cb->cb_priv = cb_priv;
842 list_add(&block_cb->list, &block->cb_list);
845 EXPORT_SYMBOL(__tcf_block_cb_register);
847 int tcf_block_cb_register(struct tcf_block *block,
848 tc_setup_cb_t *cb, void *cb_ident,
849 void *cb_priv, struct netlink_ext_ack *extack)
851 struct tcf_block_cb *block_cb;
853 block_cb = __tcf_block_cb_register(block, cb, cb_ident, cb_priv,
855 return PTR_ERR_OR_ZERO(block_cb);
857 EXPORT_SYMBOL(tcf_block_cb_register);
859 void __tcf_block_cb_unregister(struct tcf_block *block,
860 struct tcf_block_cb *block_cb)
862 tcf_block_playback_offloads(block, block_cb->cb, block_cb->cb_priv,
863 false, tcf_block_offload_in_use(block),
865 list_del(&block_cb->list);
868 EXPORT_SYMBOL(__tcf_block_cb_unregister);
870 void tcf_block_cb_unregister(struct tcf_block *block,
871 tc_setup_cb_t *cb, void *cb_ident)
873 struct tcf_block_cb *block_cb;
875 block_cb = tcf_block_cb_lookup(block, cb, cb_ident);
878 __tcf_block_cb_unregister(block, block_cb);
880 EXPORT_SYMBOL(tcf_block_cb_unregister);
882 static int tcf_block_cb_call(struct tcf_block *block, enum tc_setup_type type,
883 void *type_data, bool err_stop)
885 struct tcf_block_cb *block_cb;
889 /* Make sure all netdevs sharing this block are offload-capable. */
890 if (block->nooffloaddevcnt && err_stop)
893 list_for_each_entry(block_cb, &block->cb_list, list) {
894 err = block_cb->cb(type, type_data, block_cb->cb_priv);
905 /* Main classifier routine: scans classifier chain attached
906 * to this qdisc, (optionally) tests for protocol and asks
907 * specific classifiers.
909 int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
910 struct tcf_result *res, bool compat_mode)
912 __be16 protocol = tc_skb_protocol(skb);
913 #ifdef CONFIG_NET_CLS_ACT
914 const int max_reclassify_loop = 4;
915 const struct tcf_proto *orig_tp = tp;
916 const struct tcf_proto *first_tp;
921 for (; tp; tp = rcu_dereference_bh(tp->next)) {
924 if (tp->protocol != protocol &&
925 tp->protocol != htons(ETH_P_ALL))
928 err = tp->classify(skb, tp, res);
929 #ifdef CONFIG_NET_CLS_ACT
930 if (unlikely(err == TC_ACT_RECLASSIFY && !compat_mode)) {
933 } else if (unlikely(TC_ACT_EXT_CMP(err, TC_ACT_GOTO_CHAIN))) {
934 first_tp = res->goto_tp;
942 return TC_ACT_UNSPEC; /* signal: continue lookup */
943 #ifdef CONFIG_NET_CLS_ACT
945 if (unlikely(limit++ >= max_reclassify_loop)) {
946 net_notice_ratelimited("%u: reclassify loop, rule prio %u, protocol %02x\n",
947 tp->chain->block->index,
949 ntohs(tp->protocol));
954 protocol = tc_skb_protocol(skb);
958 EXPORT_SYMBOL(tcf_classify);
960 struct tcf_chain_info {
961 struct tcf_proto __rcu **pprev;
962 struct tcf_proto __rcu *next;
965 static struct tcf_proto *tcf_chain_tp_prev(struct tcf_chain_info *chain_info)
967 return rtnl_dereference(*chain_info->pprev);
970 static void tcf_chain_tp_insert(struct tcf_chain *chain,
971 struct tcf_chain_info *chain_info,
972 struct tcf_proto *tp)
974 if (*chain_info->pprev == chain->filter_chain)
975 tcf_chain0_head_change(chain, tp);
976 RCU_INIT_POINTER(tp->next, tcf_chain_tp_prev(chain_info));
977 rcu_assign_pointer(*chain_info->pprev, tp);
978 tcf_chain_hold(chain);
981 static void tcf_chain_tp_remove(struct tcf_chain *chain,
982 struct tcf_chain_info *chain_info,
983 struct tcf_proto *tp)
985 struct tcf_proto *next = rtnl_dereference(chain_info->next);
987 if (tp == chain->filter_chain)
988 tcf_chain0_head_change(chain, next);
989 RCU_INIT_POINTER(*chain_info->pprev, next);
990 tcf_chain_put(chain);
993 static struct tcf_proto *tcf_chain_tp_find(struct tcf_chain *chain,
994 struct tcf_chain_info *chain_info,
995 u32 protocol, u32 prio,
998 struct tcf_proto **pprev;
999 struct tcf_proto *tp;
1001 /* Check the chain for existence of proto-tcf with this priority */
1002 for (pprev = &chain->filter_chain;
1003 (tp = rtnl_dereference(*pprev)); pprev = &tp->next) {
1004 if (tp->prio >= prio) {
1005 if (tp->prio == prio) {
1006 if (prio_allocate ||
1007 (tp->protocol != protocol && protocol))
1008 return ERR_PTR(-EINVAL);
1015 chain_info->pprev = pprev;
1016 chain_info->next = tp ? tp->next : NULL;
1020 static int tcf_fill_node(struct net *net, struct sk_buff *skb,
1021 struct tcf_proto *tp, struct tcf_block *block,
1022 struct Qdisc *q, u32 parent, void *fh,
1023 u32 portid, u32 seq, u16 flags, int event)
1026 struct nlmsghdr *nlh;
1027 unsigned char *b = skb_tail_pointer(skb);
1029 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*tcm), flags);
1031 goto out_nlmsg_trim;
1032 tcm = nlmsg_data(nlh);
1033 tcm->tcm_family = AF_UNSPEC;
1037 tcm->tcm_ifindex = qdisc_dev(q)->ifindex;
1038 tcm->tcm_parent = parent;
1040 tcm->tcm_ifindex = TCM_IFINDEX_MAGIC_BLOCK;
1041 tcm->tcm_block_index = block->index;
1043 tcm->tcm_info = TC_H_MAKE(tp->prio, tp->protocol);
1044 if (nla_put_string(skb, TCA_KIND, tp->ops->kind))
1045 goto nla_put_failure;
1046 if (nla_put_u32(skb, TCA_CHAIN, tp->chain->index))
1047 goto nla_put_failure;
1049 tcm->tcm_handle = 0;
1051 if (tp->ops->dump && tp->ops->dump(net, tp, fh, skb, tcm) < 0)
1052 goto nla_put_failure;
1054 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
1063 static int tfilter_notify(struct net *net, struct sk_buff *oskb,
1064 struct nlmsghdr *n, struct tcf_proto *tp,
1065 struct tcf_block *block, struct Qdisc *q,
1066 u32 parent, void *fh, int event, bool unicast)
1068 struct sk_buff *skb;
1069 u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
1071 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1075 if (tcf_fill_node(net, skb, tp, block, q, parent, fh, portid,
1076 n->nlmsg_seq, n->nlmsg_flags, event) <= 0) {
1082 return netlink_unicast(net->rtnl, skb, portid, MSG_DONTWAIT);
1084 return rtnetlink_send(skb, net, portid, RTNLGRP_TC,
1085 n->nlmsg_flags & NLM_F_ECHO);
1088 static int tfilter_del_notify(struct net *net, struct sk_buff *oskb,
1089 struct nlmsghdr *n, struct tcf_proto *tp,
1090 struct tcf_block *block, struct Qdisc *q,
1091 u32 parent, void *fh, bool unicast, bool *last,
1092 struct netlink_ext_ack *extack)
1094 struct sk_buff *skb;
1095 u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
1098 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1102 if (tcf_fill_node(net, skb, tp, block, q, parent, fh, portid,
1103 n->nlmsg_seq, n->nlmsg_flags, RTM_DELTFILTER) <= 0) {
1104 NL_SET_ERR_MSG(extack, "Failed to build del event notification");
1109 err = tp->ops->delete(tp, fh, last, extack);
1116 return netlink_unicast(net->rtnl, skb, portid, MSG_DONTWAIT);
1118 err = rtnetlink_send(skb, net, portid, RTNLGRP_TC,
1119 n->nlmsg_flags & NLM_F_ECHO);
1121 NL_SET_ERR_MSG(extack, "Failed to send filter delete notification");
1125 static void tfilter_notify_chain(struct net *net, struct sk_buff *oskb,
1126 struct tcf_block *block, struct Qdisc *q,
1127 u32 parent, struct nlmsghdr *n,
1128 struct tcf_chain *chain, int event)
1130 struct tcf_proto *tp;
1132 for (tp = rtnl_dereference(chain->filter_chain);
1133 tp; tp = rtnl_dereference(tp->next))
1134 tfilter_notify(net, oskb, n, tp, block,
1135 q, parent, NULL, event, false);
1138 static int tc_new_tfilter(struct sk_buff *skb, struct nlmsghdr *n,
1139 struct netlink_ext_ack *extack)
1141 struct net *net = sock_net(skb->sk);
1142 struct nlattr *tca[TCA_MAX + 1];
1149 struct Qdisc *q = NULL;
1150 struct tcf_chain_info chain_info;
1151 struct tcf_chain *chain = NULL;
1152 struct tcf_block *block;
1153 struct tcf_proto *tp;
1159 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
1165 err = nlmsg_parse(n, sizeof(*t), tca, TCA_MAX, NULL, extack);
1170 protocol = TC_H_MIN(t->tcm_info);
1171 prio = TC_H_MAJ(t->tcm_info);
1172 prio_allocate = false;
1173 parent = t->tcm_parent;
1177 /* If no priority is provided by the user,
1180 if (n->nlmsg_flags & NLM_F_CREATE) {
1181 prio = TC_H_MAKE(0x80000000U, 0U);
1182 prio_allocate = true;
1184 NL_SET_ERR_MSG(extack, "Invalid filter command with priority of zero");
1189 /* Find head of filter chain. */
1191 block = tcf_block_find(net, &q, &parent, &cl,
1192 t->tcm_ifindex, t->tcm_block_index, extack);
1193 if (IS_ERR(block)) {
1194 err = PTR_ERR(block);
1198 chain_index = tca[TCA_CHAIN] ? nla_get_u32(tca[TCA_CHAIN]) : 0;
1199 if (chain_index > TC_ACT_EXT_VAL_MASK) {
1200 NL_SET_ERR_MSG(extack, "Specified chain index exceeds upper limit");
1204 chain = tcf_chain_get(block, chain_index, true);
1206 NL_SET_ERR_MSG(extack, "Cannot find specified filter chain");
1211 tp = tcf_chain_tp_find(chain, &chain_info, protocol,
1212 prio, prio_allocate);
1214 NL_SET_ERR_MSG(extack, "Filter with specified priority/protocol not found");
1220 /* Proto-tcf does not exist, create new one */
1222 if (tca[TCA_KIND] == NULL || !protocol) {
1223 NL_SET_ERR_MSG(extack, "Filter kind and protocol must be specified");
1228 if (!(n->nlmsg_flags & NLM_F_CREATE)) {
1229 NL_SET_ERR_MSG(extack, "Need both RTM_NEWTFILTER and NLM_F_CREATE to create a new filter");
1235 prio = tcf_auto_prio(tcf_chain_tp_prev(&chain_info));
1237 tp = tcf_proto_create(nla_data(tca[TCA_KIND]),
1238 protocol, prio, chain, extack);
1244 } else if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], tp->ops->kind)) {
1245 NL_SET_ERR_MSG(extack, "Specified filter kind does not match existing one");
1250 fh = tp->ops->get(tp, t->tcm_handle);
1253 if (!(n->nlmsg_flags & NLM_F_CREATE)) {
1254 NL_SET_ERR_MSG(extack, "Need both RTM_NEWTFILTER and NLM_F_CREATE to create a new filter");
1258 } else if (n->nlmsg_flags & NLM_F_EXCL) {
1259 NL_SET_ERR_MSG(extack, "Filter already exists");
1264 if (chain->tmplt_ops && chain->tmplt_ops != tp->ops) {
1265 NL_SET_ERR_MSG(extack, "Chain template is set to a different filter kind");
1270 err = tp->ops->change(net, skb, tp, cl, t->tcm_handle, tca, &fh,
1271 n->nlmsg_flags & NLM_F_CREATE ? TCA_ACT_NOREPLACE : TCA_ACT_REPLACE,
1275 tcf_chain_tp_insert(chain, &chain_info, tp);
1276 tfilter_notify(net, skb, n, tp, block, q, parent, fh,
1277 RTM_NEWTFILTER, false);
1280 tcf_proto_destroy(tp, NULL);
1285 tcf_chain_put(chain);
1287 /* Replay the request. */
1292 static int tc_del_tfilter(struct sk_buff *skb, struct nlmsghdr *n,
1293 struct netlink_ext_ack *extack)
1295 struct net *net = sock_net(skb->sk);
1296 struct nlattr *tca[TCA_MAX + 1];
1302 struct Qdisc *q = NULL;
1303 struct tcf_chain_info chain_info;
1304 struct tcf_chain *chain = NULL;
1305 struct tcf_block *block;
1306 struct tcf_proto *tp = NULL;
1307 unsigned long cl = 0;
1311 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
1314 err = nlmsg_parse(n, sizeof(*t), tca, TCA_MAX, NULL, extack);
1319 protocol = TC_H_MIN(t->tcm_info);
1320 prio = TC_H_MAJ(t->tcm_info);
1321 parent = t->tcm_parent;
1323 if (prio == 0 && (protocol || t->tcm_handle || tca[TCA_KIND])) {
1324 NL_SET_ERR_MSG(extack, "Cannot flush filters with protocol, handle or kind set");
1328 /* Find head of filter chain. */
1330 block = tcf_block_find(net, &q, &parent, &cl,
1331 t->tcm_ifindex, t->tcm_block_index, extack);
1332 if (IS_ERR(block)) {
1333 err = PTR_ERR(block);
1337 chain_index = tca[TCA_CHAIN] ? nla_get_u32(tca[TCA_CHAIN]) : 0;
1338 if (chain_index > TC_ACT_EXT_VAL_MASK) {
1339 NL_SET_ERR_MSG(extack, "Specified chain index exceeds upper limit");
1343 chain = tcf_chain_get(block, chain_index, false);
1345 NL_SET_ERR_MSG(extack, "Cannot find specified filter chain");
1351 tfilter_notify_chain(net, skb, block, q, parent, n,
1352 chain, RTM_DELTFILTER);
1353 tcf_chain_flush(chain);
1358 tp = tcf_chain_tp_find(chain, &chain_info, protocol,
1360 if (!tp || IS_ERR(tp)) {
1361 NL_SET_ERR_MSG(extack, "Filter with specified priority/protocol not found");
1362 err = tp ? PTR_ERR(tp) : -ENOENT;
1364 } else if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], tp->ops->kind)) {
1365 NL_SET_ERR_MSG(extack, "Specified filter kind does not match existing one");
1370 fh = tp->ops->get(tp, t->tcm_handle);
1373 if (t->tcm_handle == 0) {
1374 tcf_chain_tp_remove(chain, &chain_info, tp);
1375 tfilter_notify(net, skb, n, tp, block, q, parent, fh,
1376 RTM_DELTFILTER, false);
1377 tcf_proto_destroy(tp, extack);
1380 NL_SET_ERR_MSG(extack, "Specified filter handle not found");
1386 err = tfilter_del_notify(net, skb, n, tp, block,
1387 q, parent, fh, false, &last,
1392 tcf_chain_tp_remove(chain, &chain_info, tp);
1393 tcf_proto_destroy(tp, extack);
1399 tcf_chain_put(chain);
1403 static int tc_get_tfilter(struct sk_buff *skb, struct nlmsghdr *n,
1404 struct netlink_ext_ack *extack)
1406 struct net *net = sock_net(skb->sk);
1407 struct nlattr *tca[TCA_MAX + 1];
1413 struct Qdisc *q = NULL;
1414 struct tcf_chain_info chain_info;
1415 struct tcf_chain *chain = NULL;
1416 struct tcf_block *block;
1417 struct tcf_proto *tp = NULL;
1418 unsigned long cl = 0;
1422 err = nlmsg_parse(n, sizeof(*t), tca, TCA_MAX, NULL, extack);
1427 protocol = TC_H_MIN(t->tcm_info);
1428 prio = TC_H_MAJ(t->tcm_info);
1429 parent = t->tcm_parent;
1432 NL_SET_ERR_MSG(extack, "Invalid filter command with priority of zero");
1436 /* Find head of filter chain. */
1438 block = tcf_block_find(net, &q, &parent, &cl,
1439 t->tcm_ifindex, t->tcm_block_index, extack);
1440 if (IS_ERR(block)) {
1441 err = PTR_ERR(block);
1445 chain_index = tca[TCA_CHAIN] ? nla_get_u32(tca[TCA_CHAIN]) : 0;
1446 if (chain_index > TC_ACT_EXT_VAL_MASK) {
1447 NL_SET_ERR_MSG(extack, "Specified chain index exceeds upper limit");
1451 chain = tcf_chain_get(block, chain_index, false);
1453 NL_SET_ERR_MSG(extack, "Cannot find specified filter chain");
1458 tp = tcf_chain_tp_find(chain, &chain_info, protocol,
1460 if (!tp || IS_ERR(tp)) {
1461 NL_SET_ERR_MSG(extack, "Filter with specified priority/protocol not found");
1462 err = tp ? PTR_ERR(tp) : -ENOENT;
1464 } else if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], tp->ops->kind)) {
1465 NL_SET_ERR_MSG(extack, "Specified filter kind does not match existing one");
1470 fh = tp->ops->get(tp, t->tcm_handle);
1473 NL_SET_ERR_MSG(extack, "Specified filter handle not found");
1476 err = tfilter_notify(net, skb, n, tp, block, q, parent,
1477 fh, RTM_NEWTFILTER, true);
1479 NL_SET_ERR_MSG(extack, "Failed to send filter notify message");
1484 tcf_chain_put(chain);
1488 struct tcf_dump_args {
1489 struct tcf_walker w;
1490 struct sk_buff *skb;
1491 struct netlink_callback *cb;
1492 struct tcf_block *block;
1497 static int tcf_node_dump(struct tcf_proto *tp, void *n, struct tcf_walker *arg)
1499 struct tcf_dump_args *a = (void *)arg;
1500 struct net *net = sock_net(a->skb->sk);
1502 return tcf_fill_node(net, a->skb, tp, a->block, a->q, a->parent,
1503 n, NETLINK_CB(a->cb->skb).portid,
1504 a->cb->nlh->nlmsg_seq, NLM_F_MULTI,
1508 static bool tcf_chain_dump(struct tcf_chain *chain, struct Qdisc *q, u32 parent,
1509 struct sk_buff *skb, struct netlink_callback *cb,
1510 long index_start, long *p_index)
1512 struct net *net = sock_net(skb->sk);
1513 struct tcf_block *block = chain->block;
1514 struct tcmsg *tcm = nlmsg_data(cb->nlh);
1515 struct tcf_dump_args arg;
1516 struct tcf_proto *tp;
1518 for (tp = rtnl_dereference(chain->filter_chain);
1519 tp; tp = rtnl_dereference(tp->next), (*p_index)++) {
1520 if (*p_index < index_start)
1522 if (TC_H_MAJ(tcm->tcm_info) &&
1523 TC_H_MAJ(tcm->tcm_info) != tp->prio)
1525 if (TC_H_MIN(tcm->tcm_info) &&
1526 TC_H_MIN(tcm->tcm_info) != tp->protocol)
1528 if (*p_index > index_start)
1529 memset(&cb->args[1], 0,
1530 sizeof(cb->args) - sizeof(cb->args[0]));
1531 if (cb->args[1] == 0) {
1532 if (tcf_fill_node(net, skb, tp, block, q, parent, NULL,
1533 NETLINK_CB(cb->skb).portid,
1534 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1535 RTM_NEWTFILTER) <= 0)
1542 arg.w.fn = tcf_node_dump;
1547 arg.parent = parent;
1549 arg.w.skip = cb->args[1] - 1;
1551 arg.w.cookie = cb->args[2];
1552 tp->ops->walk(tp, &arg.w);
1553 cb->args[2] = arg.w.cookie;
1554 cb->args[1] = arg.w.count + 1;
1561 /* called with RTNL */
1562 static int tc_dump_tfilter(struct sk_buff *skb, struct netlink_callback *cb)
1564 struct net *net = sock_net(skb->sk);
1565 struct nlattr *tca[TCA_MAX + 1];
1566 struct Qdisc *q = NULL;
1567 struct tcf_block *block;
1568 struct tcf_chain *chain;
1569 struct tcmsg *tcm = nlmsg_data(cb->nlh);
1575 if (nlmsg_len(cb->nlh) < sizeof(*tcm))
1578 err = nlmsg_parse(cb->nlh, sizeof(*tcm), tca, TCA_MAX, NULL, NULL);
1582 if (tcm->tcm_ifindex == TCM_IFINDEX_MAGIC_BLOCK) {
1583 block = tcf_block_lookup(net, tcm->tcm_block_index);
1586 /* If we work with block index, q is NULL and parent value
1587 * will never be used in the following code. The check
1588 * in tcf_fill_node prevents it. However, compiler does not
1589 * see that far, so set parent to zero to silence the warning
1590 * about parent being uninitialized.
1594 const struct Qdisc_class_ops *cops;
1595 struct net_device *dev;
1596 unsigned long cl = 0;
1598 dev = __dev_get_by_index(net, tcm->tcm_ifindex);
1602 parent = tcm->tcm_parent;
1607 q = qdisc_lookup(dev, TC_H_MAJ(tcm->tcm_parent));
1611 cops = q->ops->cl_ops;
1614 if (!cops->tcf_block)
1616 if (TC_H_MIN(tcm->tcm_parent)) {
1617 cl = cops->find(q, tcm->tcm_parent);
1621 block = cops->tcf_block(q, cl, NULL);
1624 if (tcf_block_shared(block))
1628 index_start = cb->args[0];
1631 list_for_each_entry(chain, &block->chain_list, list) {
1632 if (tca[TCA_CHAIN] &&
1633 nla_get_u32(tca[TCA_CHAIN]) != chain->index)
1635 if (!tcf_chain_dump(chain, q, parent, skb, cb,
1636 index_start, &index)) {
1642 cb->args[0] = index;
1645 /* If we did no progress, the error (EMSGSIZE) is real */
1646 if (skb->len == 0 && err)
1651 static int tc_chain_fill_node(struct tcf_chain *chain, struct net *net,
1652 struct sk_buff *skb, struct tcf_block *block,
1653 u32 portid, u32 seq, u16 flags, int event)
1655 unsigned char *b = skb_tail_pointer(skb);
1656 const struct tcf_proto_ops *ops;
1657 struct nlmsghdr *nlh;
1661 ops = chain->tmplt_ops;
1662 priv = chain->tmplt_priv;
1664 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*tcm), flags);
1666 goto out_nlmsg_trim;
1667 tcm = nlmsg_data(nlh);
1668 tcm->tcm_family = AF_UNSPEC;
1671 tcm->tcm_handle = 0;
1673 tcm->tcm_ifindex = qdisc_dev(block->q)->ifindex;
1674 tcm->tcm_parent = block->q->handle;
1676 tcm->tcm_ifindex = TCM_IFINDEX_MAGIC_BLOCK;
1677 tcm->tcm_block_index = block->index;
1680 if (nla_put_u32(skb, TCA_CHAIN, chain->index))
1681 goto nla_put_failure;
1684 if (nla_put_string(skb, TCA_KIND, ops->kind))
1685 goto nla_put_failure;
1686 if (ops->tmplt_dump(skb, net, priv) < 0)
1687 goto nla_put_failure;
1690 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
1699 static int tc_chain_notify(struct tcf_chain *chain, struct sk_buff *oskb,
1700 u32 seq, u16 flags, int event, bool unicast)
1702 u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
1703 struct tcf_block *block = chain->block;
1704 struct net *net = block->net;
1705 struct sk_buff *skb;
1707 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1711 if (tc_chain_fill_node(chain, net, skb, block, portid,
1712 seq, flags, event) <= 0) {
1718 return netlink_unicast(net->rtnl, skb, portid, MSG_DONTWAIT);
1720 return rtnetlink_send(skb, net, portid, RTNLGRP_TC, flags & NLM_F_ECHO);
1723 static int tc_chain_tmplt_add(struct tcf_chain *chain, struct net *net,
1724 struct nlattr **tca,
1725 struct netlink_ext_ack *extack)
1727 const struct tcf_proto_ops *ops;
1730 /* If kind is not set, user did not specify template. */
1734 ops = tcf_proto_lookup_ops(nla_data(tca[TCA_KIND]), extack);
1736 return PTR_ERR(ops);
1737 if (!ops->tmplt_create || !ops->tmplt_destroy || !ops->tmplt_dump) {
1738 NL_SET_ERR_MSG(extack, "Chain templates are not supported with specified classifier");
1742 tmplt_priv = ops->tmplt_create(net, chain, tca, extack);
1743 if (IS_ERR(tmplt_priv)) {
1744 module_put(ops->owner);
1745 return PTR_ERR(tmplt_priv);
1747 chain->tmplt_ops = ops;
1748 chain->tmplt_priv = tmplt_priv;
1752 static void tc_chain_tmplt_del(struct tcf_chain *chain)
1754 const struct tcf_proto_ops *ops = chain->tmplt_ops;
1756 /* If template ops are set, no work to do for us. */
1760 ops->tmplt_destroy(chain->tmplt_priv);
1761 module_put(ops->owner);
1764 /* Add/delete/get a chain */
1766 static int tc_ctl_chain(struct sk_buff *skb, struct nlmsghdr *n,
1767 struct netlink_ext_ack *extack)
1769 struct net *net = sock_net(skb->sk);
1770 struct nlattr *tca[TCA_MAX + 1];
1774 struct Qdisc *q = NULL;
1775 struct tcf_chain *chain = NULL;
1776 struct tcf_block *block;
1780 if (n->nlmsg_type != RTM_GETCHAIN &&
1781 !netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
1785 err = nlmsg_parse(n, sizeof(*t), tca, TCA_MAX, NULL, extack);
1790 parent = t->tcm_parent;
1793 block = tcf_block_find(net, &q, &parent, &cl,
1794 t->tcm_ifindex, t->tcm_block_index, extack);
1796 return PTR_ERR(block);
1798 chain_index = tca[TCA_CHAIN] ? nla_get_u32(tca[TCA_CHAIN]) : 0;
1799 if (chain_index > TC_ACT_EXT_VAL_MASK) {
1800 NL_SET_ERR_MSG(extack, "Specified chain index exceeds upper limit");
1803 chain = tcf_chain_lookup(block, chain_index);
1804 if (n->nlmsg_type == RTM_NEWCHAIN) {
1806 NL_SET_ERR_MSG(extack, "Filter chain already exists");
1809 if (!(n->nlmsg_flags & NLM_F_CREATE)) {
1810 NL_SET_ERR_MSG(extack, "Need both RTM_NEWCHAIN and NLM_F_CREATE to create a new chain");
1813 chain = tcf_chain_create(block, chain_index);
1815 NL_SET_ERR_MSG(extack, "Failed to create filter chain");
1820 NL_SET_ERR_MSG(extack, "Cannot find specified filter chain");
1823 tcf_chain_hold(chain);
1826 switch (n->nlmsg_type) {
1828 err = tc_chain_tmplt_add(chain, net, tca, extack);
1831 /* In case the chain was successfully added, take a reference
1832 * to the chain. This ensures that an empty chain
1833 * does not disappear at the end of this function.
1835 tcf_chain_hold(chain);
1836 chain->explicitly_created = true;
1837 tc_chain_notify(chain, NULL, 0, NLM_F_CREATE | NLM_F_EXCL,
1838 RTM_NEWCHAIN, false);
1841 /* Flush the chain first as the user requested chain removal. */
1842 tcf_chain_flush(chain);
1843 /* In case the chain was successfully deleted, put a reference
1844 * to the chain previously taken during addition.
1846 tcf_chain_put_explicitly_created(chain);
1847 chain->explicitly_created = false;
1850 err = tc_chain_notify(chain, skb, n->nlmsg_seq,
1851 n->nlmsg_seq, n->nlmsg_type, true);
1853 NL_SET_ERR_MSG(extack, "Failed to send chain notify message");
1857 NL_SET_ERR_MSG(extack, "Unsupported message type");
1862 tcf_chain_put(chain);
1864 /* Replay the request. */
1869 /* called with RTNL */
1870 static int tc_dump_chain(struct sk_buff *skb, struct netlink_callback *cb)
1872 struct net *net = sock_net(skb->sk);
1873 struct nlattr *tca[TCA_MAX + 1];
1874 struct Qdisc *q = NULL;
1875 struct tcf_block *block;
1876 struct tcf_chain *chain;
1877 struct tcmsg *tcm = nlmsg_data(cb->nlh);
1883 if (nlmsg_len(cb->nlh) < sizeof(*tcm))
1886 err = nlmsg_parse(cb->nlh, sizeof(*tcm), tca, TCA_MAX, NULL, NULL);
1890 if (tcm->tcm_ifindex == TCM_IFINDEX_MAGIC_BLOCK) {
1891 block = tcf_block_lookup(net, tcm->tcm_block_index);
1894 /* If we work with block index, q is NULL and parent value
1895 * will never be used in the following code. The check
1896 * in tcf_fill_node prevents it. However, compiler does not
1897 * see that far, so set parent to zero to silence the warning
1898 * about parent being uninitialized.
1902 const struct Qdisc_class_ops *cops;
1903 struct net_device *dev;
1904 unsigned long cl = 0;
1906 dev = __dev_get_by_index(net, tcm->tcm_ifindex);
1910 parent = tcm->tcm_parent;
1915 q = qdisc_lookup(dev, TC_H_MAJ(tcm->tcm_parent));
1919 cops = q->ops->cl_ops;
1922 if (!cops->tcf_block)
1924 if (TC_H_MIN(tcm->tcm_parent)) {
1925 cl = cops->find(q, tcm->tcm_parent);
1929 block = cops->tcf_block(q, cl, NULL);
1932 if (tcf_block_shared(block))
1936 index_start = cb->args[0];
1939 list_for_each_entry(chain, &block->chain_list, list) {
1940 if ((tca[TCA_CHAIN] &&
1941 nla_get_u32(tca[TCA_CHAIN]) != chain->index))
1943 if (index < index_start) {
1947 err = tc_chain_fill_node(chain, net, skb, block,
1948 NETLINK_CB(cb->skb).portid,
1949 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1956 cb->args[0] = index;
1959 /* If we did no progress, the error (EMSGSIZE) is real */
1960 if (skb->len == 0 && err)
1965 void tcf_exts_destroy(struct tcf_exts *exts)
1967 #ifdef CONFIG_NET_CLS_ACT
1968 tcf_action_destroy(exts->actions, TCA_ACT_UNBIND);
1969 kfree(exts->actions);
1970 exts->nr_actions = 0;
1973 EXPORT_SYMBOL(tcf_exts_destroy);
1975 int tcf_exts_validate(struct net *net, struct tcf_proto *tp, struct nlattr **tb,
1976 struct nlattr *rate_tlv, struct tcf_exts *exts, bool ovr,
1977 struct netlink_ext_ack *extack)
1979 #ifdef CONFIG_NET_CLS_ACT
1981 struct tc_action *act;
1982 size_t attr_size = 0;
1984 if (exts->police && tb[exts->police]) {
1985 act = tcf_action_init_1(net, tp, tb[exts->police],
1986 rate_tlv, "police", ovr,
1987 TCA_ACT_BIND, true, extack);
1989 return PTR_ERR(act);
1991 act->type = exts->type = TCA_OLD_COMPAT;
1992 exts->actions[0] = act;
1993 exts->nr_actions = 1;
1994 } else if (exts->action && tb[exts->action]) {
1997 err = tcf_action_init(net, tp, tb[exts->action],
1998 rate_tlv, NULL, ovr, TCA_ACT_BIND,
1999 exts->actions, &attr_size, true,
2003 exts->nr_actions = err;
2008 if ((exts->action && tb[exts->action]) ||
2009 (exts->police && tb[exts->police])) {
2010 NL_SET_ERR_MSG(extack, "Classifier actions are not supported per compile options (CONFIG_NET_CLS_ACT)");
2017 EXPORT_SYMBOL(tcf_exts_validate);
2019 void tcf_exts_change(struct tcf_exts *dst, struct tcf_exts *src)
2021 #ifdef CONFIG_NET_CLS_ACT
2022 struct tcf_exts old = *dst;
2025 tcf_exts_destroy(&old);
2028 EXPORT_SYMBOL(tcf_exts_change);
2030 #ifdef CONFIG_NET_CLS_ACT
2031 static struct tc_action *tcf_exts_first_act(struct tcf_exts *exts)
2033 if (exts->nr_actions == 0)
2036 return exts->actions[0];
2040 int tcf_exts_dump(struct sk_buff *skb, struct tcf_exts *exts)
2042 #ifdef CONFIG_NET_CLS_ACT
2043 struct nlattr *nest;
2045 if (exts->action && tcf_exts_has_actions(exts)) {
2047 * again for backward compatible mode - we want
2048 * to work with both old and new modes of entering
2049 * tc data even if iproute2 was newer - jhs
2051 if (exts->type != TCA_OLD_COMPAT) {
2052 nest = nla_nest_start(skb, exts->action);
2054 goto nla_put_failure;
2056 if (tcf_action_dump(skb, exts->actions, 0, 0) < 0)
2057 goto nla_put_failure;
2058 nla_nest_end(skb, nest);
2059 } else if (exts->police) {
2060 struct tc_action *act = tcf_exts_first_act(exts);
2061 nest = nla_nest_start(skb, exts->police);
2062 if (nest == NULL || !act)
2063 goto nla_put_failure;
2064 if (tcf_action_dump_old(skb, act, 0, 0) < 0)
2065 goto nla_put_failure;
2066 nla_nest_end(skb, nest);
2072 nla_nest_cancel(skb, nest);
2078 EXPORT_SYMBOL(tcf_exts_dump);
2081 int tcf_exts_dump_stats(struct sk_buff *skb, struct tcf_exts *exts)
2083 #ifdef CONFIG_NET_CLS_ACT
2084 struct tc_action *a = tcf_exts_first_act(exts);
2085 if (a != NULL && tcf_action_copy_stats(skb, a, 1) < 0)
2090 EXPORT_SYMBOL(tcf_exts_dump_stats);
2092 static int tc_exts_setup_cb_egdev_call(struct tcf_exts *exts,
2093 enum tc_setup_type type,
2094 void *type_data, bool err_stop)
2097 #ifdef CONFIG_NET_CLS_ACT
2098 const struct tc_action *a;
2099 struct net_device *dev;
2102 if (!tcf_exts_has_actions(exts))
2105 for (i = 0; i < exts->nr_actions; i++) {
2106 a = exts->actions[i];
2107 if (!a->ops->get_dev)
2109 dev = a->ops->get_dev(a);
2112 ret = tc_setup_cb_egdev_call(dev, type, type_data, err_stop);
2121 int tc_setup_cb_call(struct tcf_block *block, struct tcf_exts *exts,
2122 enum tc_setup_type type, void *type_data, bool err_stop)
2127 ret = tcf_block_cb_call(block, type, type_data, err_stop);
2132 if (!exts || ok_count)
2134 ret = tc_exts_setup_cb_egdev_call(exts, type, type_data, err_stop);
2141 EXPORT_SYMBOL(tc_setup_cb_call);
2143 static __net_init int tcf_net_init(struct net *net)
2145 struct tcf_net *tn = net_generic(net, tcf_net_id);
2151 static void __net_exit tcf_net_exit(struct net *net)
2153 struct tcf_net *tn = net_generic(net, tcf_net_id);
2155 idr_destroy(&tn->idr);
2158 static struct pernet_operations tcf_net_ops = {
2159 .init = tcf_net_init,
2160 .exit = tcf_net_exit,
2162 .size = sizeof(struct tcf_net),
2165 static int __init tc_filter_init(void)
2169 tc_filter_wq = alloc_ordered_workqueue("tc_filter_workqueue", 0);
2173 err = register_pernet_subsys(&tcf_net_ops);
2175 goto err_register_pernet_subsys;
2177 rtnl_register(PF_UNSPEC, RTM_NEWTFILTER, tc_new_tfilter, NULL, 0);
2178 rtnl_register(PF_UNSPEC, RTM_DELTFILTER, tc_del_tfilter, NULL, 0);
2179 rtnl_register(PF_UNSPEC, RTM_GETTFILTER, tc_get_tfilter,
2180 tc_dump_tfilter, 0);
2181 rtnl_register(PF_UNSPEC, RTM_NEWCHAIN, tc_ctl_chain, NULL, 0);
2182 rtnl_register(PF_UNSPEC, RTM_DELCHAIN, tc_ctl_chain, NULL, 0);
2183 rtnl_register(PF_UNSPEC, RTM_GETCHAIN, tc_ctl_chain,
2188 err_register_pernet_subsys:
2189 destroy_workqueue(tc_filter_wq);
2193 subsys_initcall(tc_filter_init);