]> git.ipfire.org Git - thirdparty/kernel/stable.git/blame - net/sched/cls_api.c
vhost: synchronize IOTLB message with dev cleanup
[thirdparty/kernel/stable.git] / net / sched / cls_api.c
CommitLineData
1da177e4
LT
1/*
2 * net/sched/cls_api.c Packet classifier API.
3 *
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.
8 *
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 *
11 * Changes:
12 *
13 * Eduardo J. Blanco <ejbs@netlabs.com.uy> :990222: kmod support
14 *
15 */
16
1da177e4
LT
17#include <linux/module.h>
18#include <linux/types.h>
19#include <linux/kernel.h>
1da177e4 20#include <linux/string.h>
1da177e4 21#include <linux/errno.h>
33a48927 22#include <linux/err.h>
1da177e4 23#include <linux/skbuff.h>
1da177e4
LT
24#include <linux/init.h>
25#include <linux/kmod.h>
5a0e3ad6 26#include <linux/slab.h>
48617387 27#include <linux/idr.h>
b854272b
DL
28#include <net/net_namespace.h>
29#include <net/sock.h>
dc5fc579 30#include <net/netlink.h>
1da177e4
LT
31#include <net/pkt_sched.h>
32#include <net/pkt_cls.h>
33
1da177e4 34/* The list of all installed classifier types */
36272874 35static LIST_HEAD(tcf_proto_base);
1da177e4
LT
36
37/* Protects list of registered TC modules. It is pure SMP lock. */
38static DEFINE_RWLOCK(cls_mod_lock);
39
40/* Find classifier type by string name */
41
33a48927 42static const struct tcf_proto_ops *tcf_proto_lookup_ops(const char *kind)
1da177e4 43{
dcd76081 44 const struct tcf_proto_ops *t, *res = NULL;
1da177e4
LT
45
46 if (kind) {
47 read_lock(&cls_mod_lock);
36272874 48 list_for_each_entry(t, &tcf_proto_base, head) {
33a48927 49 if (strcmp(kind, t->kind) == 0) {
dcd76081
ED
50 if (try_module_get(t->owner))
51 res = t;
1da177e4
LT
52 break;
53 }
54 }
55 read_unlock(&cls_mod_lock);
56 }
dcd76081 57 return res;
1da177e4
LT
58}
59
60/* Register(unregister) new classifier type */
61
62int register_tcf_proto_ops(struct tcf_proto_ops *ops)
63{
36272874 64 struct tcf_proto_ops *t;
1da177e4
LT
65 int rc = -EEXIST;
66
67 write_lock(&cls_mod_lock);
36272874 68 list_for_each_entry(t, &tcf_proto_base, head)
1da177e4
LT
69 if (!strcmp(ops->kind, t->kind))
70 goto out;
71
36272874 72 list_add_tail(&ops->head, &tcf_proto_base);
1da177e4
LT
73 rc = 0;
74out:
75 write_unlock(&cls_mod_lock);
76 return rc;
77}
aa767bfe 78EXPORT_SYMBOL(register_tcf_proto_ops);
1da177e4 79
7aa0045d
CW
80static struct workqueue_struct *tc_filter_wq;
81
1da177e4
LT
82int unregister_tcf_proto_ops(struct tcf_proto_ops *ops)
83{
36272874 84 struct tcf_proto_ops *t;
1da177e4
LT
85 int rc = -ENOENT;
86
c78e1746
DB
87 /* Wait for outstanding call_rcu()s, if any, from a
88 * tcf_proto_ops's destroy() handler.
89 */
90 rcu_barrier();
7aa0045d 91 flush_workqueue(tc_filter_wq);
c78e1746 92
1da177e4 93 write_lock(&cls_mod_lock);
dcd76081
ED
94 list_for_each_entry(t, &tcf_proto_base, head) {
95 if (t == ops) {
96 list_del(&t->head);
97 rc = 0;
1da177e4 98 break;
dcd76081
ED
99 }
100 }
1da177e4
LT
101 write_unlock(&cls_mod_lock);
102 return rc;
103}
aa767bfe 104EXPORT_SYMBOL(unregister_tcf_proto_ops);
1da177e4 105
7aa0045d
CW
106bool tcf_queue_work(struct work_struct *work)
107{
108 return queue_work(tc_filter_wq, work);
109}
110EXPORT_SYMBOL(tcf_queue_work);
111
1da177e4
LT
112/* Select new prio value from the range, managed by kernel. */
113
aa767bfe 114static inline u32 tcf_auto_prio(struct tcf_proto *tp)
1da177e4 115{
aa767bfe 116 u32 first = TC_H_MAKE(0xC0000000U, 0U);
1da177e4
LT
117
118 if (tp)
cc7ec456 119 first = tp->prio - 1;
1da177e4 120
7961973a 121 return TC_H_MAJ(first);
1da177e4
LT
122}
123
33a48927 124static struct tcf_proto *tcf_proto_create(const char *kind, u32 protocol,
c35a4acc
AA
125 u32 prio, struct tcf_chain *chain,
126 struct netlink_ext_ack *extack)
33a48927
JP
127{
128 struct tcf_proto *tp;
129 int err;
130
131 tp = kzalloc(sizeof(*tp), GFP_KERNEL);
132 if (!tp)
133 return ERR_PTR(-ENOBUFS);
134
135 err = -ENOENT;
136 tp->ops = tcf_proto_lookup_ops(kind);
137 if (!tp->ops) {
138#ifdef CONFIG_MODULES
139 rtnl_unlock();
140 request_module("cls_%s", kind);
141 rtnl_lock();
142 tp->ops = tcf_proto_lookup_ops(kind);
143 /* We dropped the RTNL semaphore in order to perform
144 * the module load. So, even if we succeeded in loading
145 * the module we have to replay the request. We indicate
146 * this using -EAGAIN.
147 */
148 if (tp->ops) {
149 module_put(tp->ops->owner);
150 err = -EAGAIN;
151 } else {
c35a4acc 152 NL_SET_ERR_MSG(extack, "TC classifier not found");
33a48927
JP
153 err = -ENOENT;
154 }
33a48927 155#endif
d68d75fd 156 goto errout;
33a48927
JP
157 }
158 tp->classify = tp->ops->classify;
159 tp->protocol = protocol;
160 tp->prio = prio;
5bc17018 161 tp->chain = chain;
33a48927
JP
162
163 err = tp->ops->init(tp);
164 if (err) {
165 module_put(tp->ops->owner);
166 goto errout;
167 }
168 return tp;
169
170errout:
171 kfree(tp);
172 return ERR_PTR(err);
173}
174
715df5ec
JK
175static void tcf_proto_destroy(struct tcf_proto *tp,
176 struct netlink_ext_ack *extack)
cf1facda 177{
715df5ec 178 tp->ops->destroy(tp, extack);
763dbf63
WC
179 module_put(tp->ops->owner);
180 kfree_rcu(tp, rcu);
cf1facda
JP
181}
182
a9b19443
JP
183struct tcf_filter_chain_list_item {
184 struct list_head list;
185 tcf_chain_head_change_t *chain_head_change;
186 void *chain_head_change_priv;
187};
188
5bc17018
JP
189static struct tcf_chain *tcf_chain_create(struct tcf_block *block,
190 u32 chain_index)
2190d1d0 191{
5bc17018
JP
192 struct tcf_chain *chain;
193
194 chain = kzalloc(sizeof(*chain), GFP_KERNEL);
195 if (!chain)
196 return NULL;
a9b19443 197 INIT_LIST_HEAD(&chain->filter_chain_list);
5bc17018
JP
198 list_add_tail(&chain->list, &block->chain_list);
199 chain->block = block;
200 chain->index = chain_index;
e2ef7544 201 chain->refcnt = 1;
5bc17018 202 return chain;
2190d1d0
JP
203}
204
a9b19443
JP
205static void tcf_chain_head_change_item(struct tcf_filter_chain_list_item *item,
206 struct tcf_proto *tp_head)
207{
208 if (item->chain_head_change)
209 item->chain_head_change(tp_head, item->chain_head_change_priv);
210}
c7eb7d72
JP
211static void tcf_chain_head_change(struct tcf_chain *chain,
212 struct tcf_proto *tp_head)
213{
a9b19443
JP
214 struct tcf_filter_chain_list_item *item;
215
216 list_for_each_entry(item, &chain->filter_chain_list, list)
217 tcf_chain_head_change_item(item, tp_head);
c7eb7d72
JP
218}
219
f93e1cdc 220static void tcf_chain_flush(struct tcf_chain *chain)
cf1facda 221{
d7aa04a5 222 struct tcf_proto *tp = rtnl_dereference(chain->filter_chain);
cf1facda 223
c7eb7d72 224 tcf_chain_head_change(chain, NULL);
d7aa04a5 225 while (tp) {
2190d1d0 226 RCU_INIT_POINTER(chain->filter_chain, tp->next);
715df5ec 227 tcf_proto_destroy(tp, NULL);
d7aa04a5
RK
228 tp = rtnl_dereference(chain->filter_chain);
229 tcf_chain_put(chain);
cf1facda 230 }
f93e1cdc
JP
231}
232
233static void tcf_chain_destroy(struct tcf_chain *chain)
234{
efbf7897
CW
235 struct tcf_block *block = chain->block;
236
e2ef7544
CW
237 list_del(&chain->list);
238 kfree(chain);
efbf7897
CW
239 if (list_empty(&block->chain_list))
240 kfree(block);
e2ef7544 241}
744a4cf6 242
e2ef7544
CW
243static void tcf_chain_hold(struct tcf_chain *chain)
244{
245 ++chain->refcnt;
2190d1d0
JP
246}
247
367a8ce8
WC
248struct tcf_chain *tcf_chain_get(struct tcf_block *block, u32 chain_index,
249 bool create)
5bc17018
JP
250{
251 struct tcf_chain *chain;
252
253 list_for_each_entry(chain, &block->chain_list, list) {
e2ef7544
CW
254 if (chain->index == chain_index) {
255 tcf_chain_hold(chain);
256 return chain;
257 }
5bc17018 258 }
80532384 259
e2ef7544 260 return create ? tcf_chain_create(block, chain_index) : NULL;
5bc17018
JP
261}
262EXPORT_SYMBOL(tcf_chain_get);
263
264void tcf_chain_put(struct tcf_chain *chain)
265{
e2ef7544 266 if (--chain->refcnt == 0)
5bc17018
JP
267 tcf_chain_destroy(chain);
268}
269EXPORT_SYMBOL(tcf_chain_put);
270
caa72601
JP
271static bool tcf_block_offload_in_use(struct tcf_block *block)
272{
273 return block->offloadcnt;
274}
275
276static int tcf_block_offload_cmd(struct tcf_block *block,
277 struct net_device *dev,
278 struct tcf_block_ext_info *ei,
279 enum tc_block_command command)
8c4083b3 280{
8c4083b3
JP
281 struct tc_block_offload bo = {};
282
8c4083b3
JP
283 bo.command = command;
284 bo.binder_type = ei->binder_type;
285 bo.block = block;
caa72601 286 return dev->netdev_ops->ndo_setup_tc(dev, TC_SETUP_BLOCK, &bo);
8c4083b3
JP
287}
288
caa72601
JP
289static int tcf_block_offload_bind(struct tcf_block *block, struct Qdisc *q,
290 struct tcf_block_ext_info *ei)
8c4083b3 291{
caa72601
JP
292 struct net_device *dev = q->dev_queue->dev;
293 int err;
294
295 if (!dev->netdev_ops->ndo_setup_tc)
296 goto no_offload_dev_inc;
297
298 /* If tc offload feature is disabled and the block we try to bind
299 * to already has some offloaded filters, forbid to bind.
300 */
301 if (!tc_can_offload(dev) && tcf_block_offload_in_use(block))
302 return -EOPNOTSUPP;
303
304 err = tcf_block_offload_cmd(block, dev, ei, TC_BLOCK_BIND);
305 if (err == -EOPNOTSUPP)
306 goto no_offload_dev_inc;
307 return err;
308
309no_offload_dev_inc:
310 if (tcf_block_offload_in_use(block))
311 return -EOPNOTSUPP;
312 block->nooffloaddevcnt++;
313 return 0;
8c4083b3
JP
314}
315
316static void tcf_block_offload_unbind(struct tcf_block *block, struct Qdisc *q,
317 struct tcf_block_ext_info *ei)
318{
caa72601
JP
319 struct net_device *dev = q->dev_queue->dev;
320 int err;
321
322 if (!dev->netdev_ops->ndo_setup_tc)
323 goto no_offload_dev_dec;
324 err = tcf_block_offload_cmd(block, dev, ei, TC_BLOCK_UNBIND);
325 if (err == -EOPNOTSUPP)
326 goto no_offload_dev_dec;
327 return;
328
329no_offload_dev_dec:
330 WARN_ON(block->nooffloaddevcnt-- == 0);
8c4083b3
JP
331}
332
a9b19443
JP
333static int
334tcf_chain_head_change_cb_add(struct tcf_chain *chain,
335 struct tcf_block_ext_info *ei,
336 struct netlink_ext_ack *extack)
337{
338 struct tcf_filter_chain_list_item *item;
339
340 item = kmalloc(sizeof(*item), GFP_KERNEL);
341 if (!item) {
342 NL_SET_ERR_MSG(extack, "Memory allocation for head change callback item failed");
343 return -ENOMEM;
344 }
345 item->chain_head_change = ei->chain_head_change;
346 item->chain_head_change_priv = ei->chain_head_change_priv;
347 if (chain->filter_chain)
348 tcf_chain_head_change_item(item, chain->filter_chain);
349 list_add(&item->list, &chain->filter_chain_list);
350 return 0;
351}
352
353static void
354tcf_chain_head_change_cb_del(struct tcf_chain *chain,
355 struct tcf_block_ext_info *ei)
356{
357 struct tcf_filter_chain_list_item *item;
358
359 list_for_each_entry(item, &chain->filter_chain_list, list) {
360 if ((!ei->chain_head_change && !ei->chain_head_change_priv) ||
361 (item->chain_head_change == ei->chain_head_change &&
362 item->chain_head_change_priv == ei->chain_head_change_priv)) {
363 tcf_chain_head_change_item(item, NULL);
364 list_del(&item->list);
365 kfree(item);
366 return;
367 }
368 }
369 WARN_ON(1);
370}
371
48617387
JP
372struct tcf_net {
373 struct idr idr;
374};
375
376static unsigned int tcf_net_id;
377
378static int tcf_block_insert(struct tcf_block *block, struct net *net,
bb047ddd 379 struct netlink_ext_ack *extack)
a9b19443 380{
48617387 381 struct tcf_net *tn = net_generic(net, tcf_net_id);
48617387 382
bb047ddd
JP
383 return idr_alloc_u32(&tn->idr, block, &block->index, block->index,
384 GFP_KERNEL);
a9b19443
JP
385}
386
48617387
JP
387static void tcf_block_remove(struct tcf_block *block, struct net *net)
388{
389 struct tcf_net *tn = net_generic(net, tcf_net_id);
390
9c160941 391 idr_remove(&tn->idr, block->index);
48617387
JP
392}
393
394static struct tcf_block *tcf_block_create(struct net *net, struct Qdisc *q,
bb047ddd 395 u32 block_index,
48617387 396 struct netlink_ext_ack *extack)
6529eaba 397{
48617387 398 struct tcf_block *block;
5bc17018 399 struct tcf_chain *chain;
2190d1d0 400 int err;
6529eaba 401
48617387 402 block = kzalloc(sizeof(*block), GFP_KERNEL);
8d1a77f9
AA
403 if (!block) {
404 NL_SET_ERR_MSG(extack, "Memory allocation for block failed");
48617387 405 return ERR_PTR(-ENOMEM);
8d1a77f9 406 }
5bc17018 407 INIT_LIST_HEAD(&block->chain_list);
acb67442 408 INIT_LIST_HEAD(&block->cb_list);
f36fe1c4 409 INIT_LIST_HEAD(&block->owner_list);
acb67442 410
5bc17018
JP
411 /* Create chain 0 by default, it has to be always present. */
412 chain = tcf_chain_create(block, 0);
413 if (!chain) {
8d1a77f9 414 NL_SET_ERR_MSG(extack, "Failed to create new tcf chain");
2190d1d0
JP
415 err = -ENOMEM;
416 goto err_chain_create;
417 }
48617387
JP
418 block->refcnt = 1;
419 block->net = net;
bb047ddd
JP
420 block->index = block_index;
421
422 /* Don't store q pointer for blocks which are shared */
423 if (!tcf_block_shared(block))
424 block->q = q;
48617387
JP
425 return block;
426
427err_chain_create:
428 kfree(block);
429 return ERR_PTR(err);
430}
431
432static struct tcf_block *tcf_block_lookup(struct net *net, u32 block_index)
433{
434 struct tcf_net *tn = net_generic(net, tcf_net_id);
435
322d884b 436 return idr_find(&tn->idr, block_index);
48617387
JP
437}
438
439static struct tcf_chain *tcf_block_chain_zero(struct tcf_block *block)
440{
441 return list_first_entry(&block->chain_list, struct tcf_chain, list);
442}
443
f36fe1c4
JP
444struct tcf_block_owner_item {
445 struct list_head list;
446 struct Qdisc *q;
447 enum tcf_block_binder_type binder_type;
448};
449
450static void
451tcf_block_owner_netif_keep_dst(struct tcf_block *block,
452 struct Qdisc *q,
453 enum tcf_block_binder_type binder_type)
454{
455 if (block->keep_dst &&
456 binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS &&
457 binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_EGRESS)
458 netif_keep_dst(qdisc_dev(q));
459}
460
461void tcf_block_netif_keep_dst(struct tcf_block *block)
462{
463 struct tcf_block_owner_item *item;
464
465 block->keep_dst = true;
466 list_for_each_entry(item, &block->owner_list, list)
467 tcf_block_owner_netif_keep_dst(block, item->q,
468 item->binder_type);
469}
470EXPORT_SYMBOL(tcf_block_netif_keep_dst);
471
472static int tcf_block_owner_add(struct tcf_block *block,
473 struct Qdisc *q,
474 enum tcf_block_binder_type binder_type)
475{
476 struct tcf_block_owner_item *item;
477
478 item = kmalloc(sizeof(*item), GFP_KERNEL);
479 if (!item)
480 return -ENOMEM;
481 item->q = q;
482 item->binder_type = binder_type;
483 list_add(&item->list, &block->owner_list);
484 return 0;
485}
486
487static void tcf_block_owner_del(struct tcf_block *block,
488 struct Qdisc *q,
489 enum tcf_block_binder_type binder_type)
490{
491 struct tcf_block_owner_item *item;
492
493 list_for_each_entry(item, &block->owner_list, list) {
494 if (item->q == q && item->binder_type == binder_type) {
495 list_del(&item->list);
496 kfree(item);
497 return;
498 }
499 }
500 WARN_ON(1);
501}
502
48617387
JP
503int tcf_block_get_ext(struct tcf_block **p_block, struct Qdisc *q,
504 struct tcf_block_ext_info *ei,
505 struct netlink_ext_ack *extack)
506{
507 struct net *net = qdisc_net(q);
508 struct tcf_block *block = NULL;
509 bool created = false;
510 int err;
511
512 if (ei->block_index) {
513 /* block_index not 0 means the shared block is requested */
514 block = tcf_block_lookup(net, ei->block_index);
515 if (block)
516 block->refcnt++;
517 }
518
519 if (!block) {
bb047ddd 520 block = tcf_block_create(net, q, ei->block_index, extack);
48617387
JP
521 if (IS_ERR(block))
522 return PTR_ERR(block);
523 created = true;
bb047ddd
JP
524 if (tcf_block_shared(block)) {
525 err = tcf_block_insert(block, net, extack);
48617387
JP
526 if (err)
527 goto err_block_insert;
528 }
529 }
530
f36fe1c4
JP
531 err = tcf_block_owner_add(block, q, ei->binder_type);
532 if (err)
533 goto err_block_owner_add;
534
535 tcf_block_owner_netif_keep_dst(block, q, ei->binder_type);
536
a9b19443
JP
537 err = tcf_chain_head_change_cb_add(tcf_block_chain_zero(block),
538 ei, extack);
539 if (err)
540 goto err_chain_head_change_cb_add;
caa72601
JP
541
542 err = tcf_block_offload_bind(block, q, ei);
543 if (err)
544 goto err_block_offload_bind;
545
6529eaba
JP
546 *p_block = block;
547 return 0;
2190d1d0 548
caa72601
JP
549err_block_offload_bind:
550 tcf_chain_head_change_cb_del(tcf_block_chain_zero(block), ei);
a9b19443 551err_chain_head_change_cb_add:
f36fe1c4
JP
552 tcf_block_owner_del(block, q, ei->binder_type);
553err_block_owner_add:
48617387
JP
554 if (created) {
555 if (tcf_block_shared(block))
556 tcf_block_remove(block, net);
557err_block_insert:
558 kfree(tcf_block_chain_zero(block));
559 kfree(block);
560 } else {
561 block->refcnt--;
562 }
2190d1d0 563 return err;
6529eaba 564}
8c4083b3
JP
565EXPORT_SYMBOL(tcf_block_get_ext);
566
c7eb7d72
JP
567static void tcf_chain_head_change_dflt(struct tcf_proto *tp_head, void *priv)
568{
569 struct tcf_proto __rcu **p_filter_chain = priv;
570
571 rcu_assign_pointer(*p_filter_chain, tp_head);
572}
573
8c4083b3 574int tcf_block_get(struct tcf_block **p_block,
8d1a77f9
AA
575 struct tcf_proto __rcu **p_filter_chain, struct Qdisc *q,
576 struct netlink_ext_ack *extack)
8c4083b3 577{
c7eb7d72
JP
578 struct tcf_block_ext_info ei = {
579 .chain_head_change = tcf_chain_head_change_dflt,
580 .chain_head_change_priv = p_filter_chain,
581 };
8c4083b3 582
c7eb7d72 583 WARN_ON(!p_filter_chain);
8d1a77f9 584 return tcf_block_get_ext(p_block, q, &ei, extack);
8c4083b3 585}
6529eaba
JP
586EXPORT_SYMBOL(tcf_block_get);
587
7aa0045d 588/* XXX: Standalone actions are not allowed to jump to any chain, and bound
a60b3f51 589 * actions should be all removed after flushing.
7aa0045d 590 */
c7eb7d72 591void tcf_block_put_ext(struct tcf_block *block, struct Qdisc *q,
e1ea2f98 592 struct tcf_block_ext_info *ei)
7aa0045d 593{
efbf7897 594 struct tcf_chain *chain, *tmp;
1697c4bb 595
c30abd5e
DM
596 if (!block)
597 return;
a9b19443 598 tcf_chain_head_change_cb_del(tcf_block_chain_zero(block), ei);
f36fe1c4 599 tcf_block_owner_del(block, q, ei->binder_type);
a60b3f51 600
48617387
JP
601 if (--block->refcnt == 0) {
602 if (tcf_block_shared(block))
603 tcf_block_remove(block, block->net);
604
605 /* Hold a refcnt for all chains, so that they don't disappear
606 * while we are iterating.
607 */
608 list_for_each_entry(chain, &block->chain_list, list)
609 tcf_chain_hold(chain);
610
611 list_for_each_entry(chain, &block->chain_list, list)
612 tcf_chain_flush(chain);
613 }
e2ef7544 614
4bb1b116
JP
615 tcf_block_offload_unbind(block, q, ei);
616
48617387
JP
617 if (block->refcnt == 0) {
618 /* At this point, all the chains should have refcnt >= 1. */
619 list_for_each_entry_safe(chain, tmp, &block->chain_list, list)
620 tcf_chain_put(chain);
df45bf84 621
48617387
JP
622 /* Finally, put chain 0 and allow block to be freed. */
623 tcf_chain_put(tcf_block_chain_zero(block));
624 }
6529eaba 625}
8c4083b3
JP
626EXPORT_SYMBOL(tcf_block_put_ext);
627
628void tcf_block_put(struct tcf_block *block)
629{
630 struct tcf_block_ext_info ei = {0, };
631
4853f128
JP
632 if (!block)
633 return;
c7eb7d72 634 tcf_block_put_ext(block, block->q, &ei);
8c4083b3 635}
e1ea2f98 636
6529eaba 637EXPORT_SYMBOL(tcf_block_put);
cf1facda 638
acb67442
JP
639struct tcf_block_cb {
640 struct list_head list;
641 tc_setup_cb_t *cb;
642 void *cb_ident;
643 void *cb_priv;
644 unsigned int refcnt;
645};
646
647void *tcf_block_cb_priv(struct tcf_block_cb *block_cb)
648{
649 return block_cb->cb_priv;
650}
651EXPORT_SYMBOL(tcf_block_cb_priv);
652
653struct tcf_block_cb *tcf_block_cb_lookup(struct tcf_block *block,
654 tc_setup_cb_t *cb, void *cb_ident)
655{ struct tcf_block_cb *block_cb;
656
657 list_for_each_entry(block_cb, &block->cb_list, list)
658 if (block_cb->cb == cb && block_cb->cb_ident == cb_ident)
659 return block_cb;
660 return NULL;
661}
662EXPORT_SYMBOL(tcf_block_cb_lookup);
663
664void tcf_block_cb_incref(struct tcf_block_cb *block_cb)
665{
666 block_cb->refcnt++;
667}
668EXPORT_SYMBOL(tcf_block_cb_incref);
669
670unsigned int tcf_block_cb_decref(struct tcf_block_cb *block_cb)
671{
672 return --block_cb->refcnt;
673}
674EXPORT_SYMBOL(tcf_block_cb_decref);
675
676struct tcf_block_cb *__tcf_block_cb_register(struct tcf_block *block,
677 tc_setup_cb_t *cb, void *cb_ident,
678 void *cb_priv)
679{
680 struct tcf_block_cb *block_cb;
681
caa72601
JP
682 /* At this point, playback of previous block cb calls is not supported,
683 * so forbid to register to block which already has some offloaded
684 * filters present.
685 */
686 if (tcf_block_offload_in_use(block))
687 return ERR_PTR(-EOPNOTSUPP);
688
acb67442
JP
689 block_cb = kzalloc(sizeof(*block_cb), GFP_KERNEL);
690 if (!block_cb)
caa72601 691 return ERR_PTR(-ENOMEM);
acb67442
JP
692 block_cb->cb = cb;
693 block_cb->cb_ident = cb_ident;
694 block_cb->cb_priv = cb_priv;
695 list_add(&block_cb->list, &block->cb_list);
696 return block_cb;
697}
698EXPORT_SYMBOL(__tcf_block_cb_register);
699
700int tcf_block_cb_register(struct tcf_block *block,
701 tc_setup_cb_t *cb, void *cb_ident,
702 void *cb_priv)
703{
704 struct tcf_block_cb *block_cb;
705
706 block_cb = __tcf_block_cb_register(block, cb, cb_ident, cb_priv);
caa72601 707 return IS_ERR(block_cb) ? PTR_ERR(block_cb) : 0;
acb67442
JP
708}
709EXPORT_SYMBOL(tcf_block_cb_register);
710
711void __tcf_block_cb_unregister(struct tcf_block_cb *block_cb)
712{
713 list_del(&block_cb->list);
714 kfree(block_cb);
715}
716EXPORT_SYMBOL(__tcf_block_cb_unregister);
717
718void tcf_block_cb_unregister(struct tcf_block *block,
719 tc_setup_cb_t *cb, void *cb_ident)
720{
721 struct tcf_block_cb *block_cb;
722
723 block_cb = tcf_block_cb_lookup(block, cb, cb_ident);
724 if (!block_cb)
725 return;
726 __tcf_block_cb_unregister(block_cb);
727}
728EXPORT_SYMBOL(tcf_block_cb_unregister);
729
730static int tcf_block_cb_call(struct tcf_block *block, enum tc_setup_type type,
731 void *type_data, bool err_stop)
732{
733 struct tcf_block_cb *block_cb;
734 int ok_count = 0;
735 int err;
736
caa72601
JP
737 /* Make sure all netdevs sharing this block are offload-capable. */
738 if (block->nooffloaddevcnt && err_stop)
739 return -EOPNOTSUPP;
740
acb67442
JP
741 list_for_each_entry(block_cb, &block->cb_list, list) {
742 err = block_cb->cb(type, type_data, block_cb->cb_priv);
743 if (err) {
744 if (err_stop)
745 return err;
746 } else {
747 ok_count++;
748 }
749 }
750 return ok_count;
751}
752
87d83093
JP
753/* Main classifier routine: scans classifier chain attached
754 * to this qdisc, (optionally) tests for protocol and asks
755 * specific classifiers.
756 */
757int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
758 struct tcf_result *res, bool compat_mode)
759{
760 __be16 protocol = tc_skb_protocol(skb);
761#ifdef CONFIG_NET_CLS_ACT
762 const int max_reclassify_loop = 4;
ee538dce
JP
763 const struct tcf_proto *orig_tp = tp;
764 const struct tcf_proto *first_tp;
87d83093
JP
765 int limit = 0;
766
767reclassify:
768#endif
769 for (; tp; tp = rcu_dereference_bh(tp->next)) {
770 int err;
771
772 if (tp->protocol != protocol &&
773 tp->protocol != htons(ETH_P_ALL))
774 continue;
775
776 err = tp->classify(skb, tp, res);
777#ifdef CONFIG_NET_CLS_ACT
db50514f 778 if (unlikely(err == TC_ACT_RECLASSIFY && !compat_mode)) {
ee538dce 779 first_tp = orig_tp;
87d83093 780 goto reset;
db50514f 781 } else if (unlikely(TC_ACT_EXT_CMP(err, TC_ACT_GOTO_CHAIN))) {
ee538dce 782 first_tp = res->goto_tp;
db50514f
JP
783 goto reset;
784 }
87d83093
JP
785#endif
786 if (err >= 0)
787 return err;
788 }
789
790 return TC_ACT_UNSPEC; /* signal: continue lookup */
791#ifdef CONFIG_NET_CLS_ACT
792reset:
793 if (unlikely(limit++ >= max_reclassify_loop)) {
9d3aaff3
JP
794 net_notice_ratelimited("%u: reclassify loop, rule prio %u, protocol %02x\n",
795 tp->chain->block->index,
796 tp->prio & 0xffff,
87d83093
JP
797 ntohs(tp->protocol));
798 return TC_ACT_SHOT;
799 }
800
ee538dce 801 tp = first_tp;
87d83093
JP
802 protocol = tc_skb_protocol(skb);
803 goto reclassify;
804#endif
805}
806EXPORT_SYMBOL(tcf_classify);
807
2190d1d0
JP
808struct tcf_chain_info {
809 struct tcf_proto __rcu **pprev;
810 struct tcf_proto __rcu *next;
811};
812
813static struct tcf_proto *tcf_chain_tp_prev(struct tcf_chain_info *chain_info)
814{
815 return rtnl_dereference(*chain_info->pprev);
816}
817
818static void tcf_chain_tp_insert(struct tcf_chain *chain,
819 struct tcf_chain_info *chain_info,
820 struct tcf_proto *tp)
821{
c7eb7d72
JP
822 if (*chain_info->pprev == chain->filter_chain)
823 tcf_chain_head_change(chain, tp);
2190d1d0
JP
824 RCU_INIT_POINTER(tp->next, tcf_chain_tp_prev(chain_info));
825 rcu_assign_pointer(*chain_info->pprev, tp);
e2ef7544 826 tcf_chain_hold(chain);
2190d1d0
JP
827}
828
829static void tcf_chain_tp_remove(struct tcf_chain *chain,
830 struct tcf_chain_info *chain_info,
831 struct tcf_proto *tp)
832{
833 struct tcf_proto *next = rtnl_dereference(chain_info->next);
834
c7eb7d72
JP
835 if (tp == chain->filter_chain)
836 tcf_chain_head_change(chain, next);
2190d1d0 837 RCU_INIT_POINTER(*chain_info->pprev, next);
e2ef7544 838 tcf_chain_put(chain);
2190d1d0
JP
839}
840
841static struct tcf_proto *tcf_chain_tp_find(struct tcf_chain *chain,
842 struct tcf_chain_info *chain_info,
843 u32 protocol, u32 prio,
844 bool prio_allocate)
845{
846 struct tcf_proto **pprev;
847 struct tcf_proto *tp;
848
849 /* Check the chain for existence of proto-tcf with this priority */
850 for (pprev = &chain->filter_chain;
851 (tp = rtnl_dereference(*pprev)); pprev = &tp->next) {
852 if (tp->prio >= prio) {
853 if (tp->prio == prio) {
854 if (prio_allocate ||
855 (tp->protocol != protocol && protocol))
856 return ERR_PTR(-EINVAL);
857 } else {
858 tp = NULL;
859 }
860 break;
861 }
862 }
863 chain_info->pprev = pprev;
864 chain_info->next = tp ? tp->next : NULL;
865 return tp;
866}
867
7120371c 868static int tcf_fill_node(struct net *net, struct sk_buff *skb,
7960d1da
JP
869 struct tcf_proto *tp, struct tcf_block *block,
870 struct Qdisc *q, u32 parent, void *fh,
871 u32 portid, u32 seq, u16 flags, int event)
7120371c
WC
872{
873 struct tcmsg *tcm;
874 struct nlmsghdr *nlh;
875 unsigned char *b = skb_tail_pointer(skb);
876
877 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*tcm), flags);
878 if (!nlh)
879 goto out_nlmsg_trim;
880 tcm = nlmsg_data(nlh);
881 tcm->tcm_family = AF_UNSPEC;
882 tcm->tcm__pad1 = 0;
883 tcm->tcm__pad2 = 0;
7960d1da
JP
884 if (q) {
885 tcm->tcm_ifindex = qdisc_dev(q)->ifindex;
886 tcm->tcm_parent = parent;
887 } else {
888 tcm->tcm_ifindex = TCM_IFINDEX_MAGIC_BLOCK;
889 tcm->tcm_block_index = block->index;
890 }
7120371c
WC
891 tcm->tcm_info = TC_H_MAKE(tp->prio, tp->protocol);
892 if (nla_put_string(skb, TCA_KIND, tp->ops->kind))
893 goto nla_put_failure;
894 if (nla_put_u32(skb, TCA_CHAIN, tp->chain->index))
895 goto nla_put_failure;
896 if (!fh) {
897 tcm->tcm_handle = 0;
898 } else {
899 if (tp->ops->dump && tp->ops->dump(net, tp, fh, skb, tcm) < 0)
900 goto nla_put_failure;
901 }
902 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
903 return skb->len;
904
905out_nlmsg_trim:
906nla_put_failure:
907 nlmsg_trim(skb, b);
908 return -1;
909}
910
911static int tfilter_notify(struct net *net, struct sk_buff *oskb,
912 struct nlmsghdr *n, struct tcf_proto *tp,
7960d1da
JP
913 struct tcf_block *block, struct Qdisc *q,
914 u32 parent, void *fh, int event, bool unicast)
7120371c
WC
915{
916 struct sk_buff *skb;
917 u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
918
919 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
920 if (!skb)
921 return -ENOBUFS;
922
7960d1da
JP
923 if (tcf_fill_node(net, skb, tp, block, q, parent, fh, portid,
924 n->nlmsg_seq, n->nlmsg_flags, event) <= 0) {
7120371c
WC
925 kfree_skb(skb);
926 return -EINVAL;
927 }
928
929 if (unicast)
930 return netlink_unicast(net->rtnl, skb, portid, MSG_DONTWAIT);
931
932 return rtnetlink_send(skb, net, portid, RTNLGRP_TC,
933 n->nlmsg_flags & NLM_F_ECHO);
934}
935
936static int tfilter_del_notify(struct net *net, struct sk_buff *oskb,
937 struct nlmsghdr *n, struct tcf_proto *tp,
7960d1da 938 struct tcf_block *block, struct Qdisc *q,
c35a4acc
AA
939 u32 parent, void *fh, bool unicast, bool *last,
940 struct netlink_ext_ack *extack)
7120371c
WC
941{
942 struct sk_buff *skb;
943 u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
944 int err;
945
946 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
947 if (!skb)
948 return -ENOBUFS;
949
7960d1da
JP
950 if (tcf_fill_node(net, skb, tp, block, q, parent, fh, portid,
951 n->nlmsg_seq, n->nlmsg_flags, RTM_DELTFILTER) <= 0) {
c35a4acc 952 NL_SET_ERR_MSG(extack, "Failed to build del event notification");
7120371c
WC
953 kfree_skb(skb);
954 return -EINVAL;
955 }
956
571acf21 957 err = tp->ops->delete(tp, fh, last, extack);
7120371c
WC
958 if (err) {
959 kfree_skb(skb);
960 return err;
961 }
962
963 if (unicast)
964 return netlink_unicast(net->rtnl, skb, portid, MSG_DONTWAIT);
965
c35a4acc
AA
966 err = rtnetlink_send(skb, net, portid, RTNLGRP_TC,
967 n->nlmsg_flags & NLM_F_ECHO);
968 if (err < 0)
969 NL_SET_ERR_MSG(extack, "Failed to send filter delete notification");
970 return err;
7120371c
WC
971}
972
973static void tfilter_notify_chain(struct net *net, struct sk_buff *oskb,
7960d1da
JP
974 struct tcf_block *block, struct Qdisc *q,
975 u32 parent, struct nlmsghdr *n,
7120371c
WC
976 struct tcf_chain *chain, int event)
977{
978 struct tcf_proto *tp;
979
980 for (tp = rtnl_dereference(chain->filter_chain);
981 tp; tp = rtnl_dereference(tp->next))
7960d1da
JP
982 tfilter_notify(net, oskb, n, tp, block,
983 q, parent, 0, event, false);
7120371c
WC
984}
985
1da177e4
LT
986/* Add/change/delete/get a filter node */
987
c21ef3e3
DA
988static int tc_ctl_tfilter(struct sk_buff *skb, struct nlmsghdr *n,
989 struct netlink_ext_ack *extack)
1da177e4 990{
3b1e0a65 991 struct net *net = sock_net(skb->sk);
add93b61 992 struct nlattr *tca[TCA_MAX + 1];
1da177e4
LT
993 struct tcmsg *t;
994 u32 protocol;
995 u32 prio;
9d36d9e5 996 bool prio_allocate;
1da177e4 997 u32 parent;
5bc17018 998 u32 chain_index;
7960d1da 999 struct Qdisc *q = NULL;
2190d1d0 1000 struct tcf_chain_info chain_info;
5bc17018 1001 struct tcf_chain *chain = NULL;
6529eaba 1002 struct tcf_block *block;
1da177e4 1003 struct tcf_proto *tp;
1da177e4 1004 unsigned long cl;
8113c095 1005 void *fh;
1da177e4 1006 int err;
628185cf 1007 int tp_created;
1da177e4 1008
4e8bbb81 1009 if ((n->nlmsg_type != RTM_GETTFILTER) &&
5f013c9b 1010 !netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
dfc47ef8 1011 return -EPERM;
de179c8c 1012
1da177e4 1013replay:
628185cf
DB
1014 tp_created = 0;
1015
c21ef3e3 1016 err = nlmsg_parse(n, sizeof(*t), tca, TCA_MAX, NULL, extack);
de179c8c
H
1017 if (err < 0)
1018 return err;
1019
942b8165 1020 t = nlmsg_data(n);
1da177e4
LT
1021 protocol = TC_H_MIN(t->tcm_info);
1022 prio = TC_H_MAJ(t->tcm_info);
9d36d9e5 1023 prio_allocate = false;
1da177e4
LT
1024 parent = t->tcm_parent;
1025 cl = 0;
1026
1027 if (prio == 0) {
ea7f8277
DB
1028 switch (n->nlmsg_type) {
1029 case RTM_DELTFILTER:
c35a4acc
AA
1030 if (protocol || t->tcm_handle || tca[TCA_KIND]) {
1031 NL_SET_ERR_MSG(extack, "Cannot flush filters with protocol, handle or kind set");
ea7f8277 1032 return -ENOENT;
c35a4acc 1033 }
ea7f8277
DB
1034 break;
1035 case RTM_NEWTFILTER:
1036 /* If no priority is provided by the user,
1037 * we allocate one.
1038 */
1039 if (n->nlmsg_flags & NLM_F_CREATE) {
1040 prio = TC_H_MAKE(0x80000000U, 0U);
9d36d9e5 1041 prio_allocate = true;
ea7f8277
DB
1042 break;
1043 }
1044 /* fall-through */
1045 default:
c35a4acc 1046 NL_SET_ERR_MSG(extack, "Invalid filter command with priority of zero");
1da177e4 1047 return -ENOENT;
ea7f8277 1048 }
1da177e4
LT
1049 }
1050
1051 /* Find head of filter chain. */
1052
7960d1da
JP
1053 if (t->tcm_ifindex == TCM_IFINDEX_MAGIC_BLOCK) {
1054 block = tcf_block_lookup(net, t->tcm_block_index);
1055 if (!block) {
1056 NL_SET_ERR_MSG(extack, "Block of given index was not found");
1057 err = -EINVAL;
1058 goto errout;
1059 }
aa767bfe 1060 } else {
7960d1da
JP
1061 const struct Qdisc_class_ops *cops;
1062 struct net_device *dev;
1063
1064 /* Find link */
1065 dev = __dev_get_by_index(net, t->tcm_ifindex);
1066 if (!dev)
1067 return -ENODEV;
1068
1069 /* Find qdisc */
1070 if (!parent) {
1071 q = dev->qdisc;
1072 parent = q->handle;
1073 } else {
1074 q = qdisc_lookup(dev, TC_H_MAJ(t->tcm_parent));
c35a4acc
AA
1075 if (!q) {
1076 NL_SET_ERR_MSG(extack, "Parent Qdisc doesn't exists");
7960d1da 1077 return -EINVAL;
c35a4acc 1078 }
7960d1da 1079 }
1da177e4 1080
7960d1da
JP
1081 /* Is it classful? */
1082 cops = q->ops->cl_ops;
c35a4acc
AA
1083 if (!cops) {
1084 NL_SET_ERR_MSG(extack, "Qdisc not classful");
7960d1da 1085 return -EINVAL;
c35a4acc 1086 }
1da177e4 1087
c35a4acc
AA
1088 if (!cops->tcf_block) {
1089 NL_SET_ERR_MSG(extack, "Class doesn't support blocks");
7960d1da 1090 return -EOPNOTSUPP;
c35a4acc 1091 }
71ebe5e9 1092
7960d1da
JP
1093 /* Do we search for filter, attached to class? */
1094 if (TC_H_MIN(parent)) {
1095 cl = cops->find(q, parent);
c35a4acc
AA
1096 if (cl == 0) {
1097 NL_SET_ERR_MSG(extack, "Specified class doesn't exist");
7960d1da 1098 return -ENOENT;
c35a4acc 1099 }
7960d1da 1100 }
1da177e4 1101
7960d1da
JP
1102 /* And the last stroke */
1103 block = cops->tcf_block(q, cl, extack);
1104 if (!block) {
1105 err = -EINVAL;
1106 goto errout;
1107 }
1108 if (tcf_block_shared(block)) {
1109 NL_SET_ERR_MSG(extack, "This filter block is shared. Please use the block index to manipulate the filters");
1110 err = -EOPNOTSUPP;
1111 goto errout;
1112 }
6bb16e7a 1113 }
5bc17018
JP
1114
1115 chain_index = tca[TCA_CHAIN] ? nla_get_u32(tca[TCA_CHAIN]) : 0;
1116 if (chain_index > TC_ACT_EXT_VAL_MASK) {
c35a4acc 1117 NL_SET_ERR_MSG(extack, "Specified chain index exceeds upper limit");
5bc17018
JP
1118 err = -EINVAL;
1119 goto errout;
1120 }
367a8ce8
WC
1121 chain = tcf_chain_get(block, chain_index,
1122 n->nlmsg_type == RTM_NEWTFILTER);
5bc17018 1123 if (!chain) {
c35a4acc 1124 NL_SET_ERR_MSG(extack, "Cannot find specified filter chain");
367a8ce8 1125 err = n->nlmsg_type == RTM_NEWTFILTER ? -ENOMEM : -EINVAL;
5bc17018
JP
1126 goto errout;
1127 }
6529eaba 1128
ea7f8277 1129 if (n->nlmsg_type == RTM_DELTFILTER && prio == 0) {
7960d1da 1130 tfilter_notify_chain(net, skb, block, q, parent, n,
a10fa201 1131 chain, RTM_DELTFILTER);
f93e1cdc 1132 tcf_chain_flush(chain);
ea7f8277
DB
1133 err = 0;
1134 goto errout;
1135 }
1da177e4 1136
2190d1d0
JP
1137 tp = tcf_chain_tp_find(chain, &chain_info, protocol,
1138 prio, prio_allocate);
1139 if (IS_ERR(tp)) {
c35a4acc 1140 NL_SET_ERR_MSG(extack, "Filter with specified priority/protocol not found");
2190d1d0
JP
1141 err = PTR_ERR(tp);
1142 goto errout;
1da177e4
LT
1143 }
1144
1145 if (tp == NULL) {
1146 /* Proto-tcf does not exist, create new one */
1147
6bb16e7a 1148 if (tca[TCA_KIND] == NULL || !protocol) {
c35a4acc 1149 NL_SET_ERR_MSG(extack, "Filter kind and protocol must be specified");
6bb16e7a 1150 err = -EINVAL;
1da177e4 1151 goto errout;
6bb16e7a 1152 }
1da177e4 1153
cc7ec456 1154 if (n->nlmsg_type != RTM_NEWTFILTER ||
6bb16e7a 1155 !(n->nlmsg_flags & NLM_F_CREATE)) {
c35a4acc 1156 NL_SET_ERR_MSG(extack, "Need both RTM_NEWTFILTER and NLM_F_CREATE to create a new filter");
6bb16e7a 1157 err = -ENOENT;
1da177e4 1158 goto errout;
6bb16e7a 1159 }
1da177e4 1160
9d36d9e5 1161 if (prio_allocate)
2190d1d0 1162 prio = tcf_auto_prio(tcf_chain_tp_prev(&chain_info));
1da177e4 1163
33a48927 1164 tp = tcf_proto_create(nla_data(tca[TCA_KIND]),
c35a4acc 1165 protocol, prio, chain, extack);
33a48927
JP
1166 if (IS_ERR(tp)) {
1167 err = PTR_ERR(tp);
1da177e4
LT
1168 goto errout;
1169 }
12186be7 1170 tp_created = 1;
6bb16e7a 1171 } else if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], tp->ops->kind)) {
c35a4acc 1172 NL_SET_ERR_MSG(extack, "Specified filter kind does not match existing one");
6bb16e7a 1173 err = -EINVAL;
1da177e4 1174 goto errout;
6bb16e7a 1175 }
1da177e4
LT
1176
1177 fh = tp->ops->get(tp, t->tcm_handle);
1178
8113c095 1179 if (!fh) {
1da177e4 1180 if (n->nlmsg_type == RTM_DELTFILTER && t->tcm_handle == 0) {
2190d1d0 1181 tcf_chain_tp_remove(chain, &chain_info, tp);
7960d1da 1182 tfilter_notify(net, skb, n, tp, block, q, parent, fh,
fa59b27c 1183 RTM_DELTFILTER, false);
715df5ec 1184 tcf_proto_destroy(tp, extack);
1da177e4
LT
1185 err = 0;
1186 goto errout;
1187 }
1188
aa767bfe 1189 if (n->nlmsg_type != RTM_NEWTFILTER ||
6bb16e7a 1190 !(n->nlmsg_flags & NLM_F_CREATE)) {
c35a4acc 1191 NL_SET_ERR_MSG(extack, "Need both RTM_NEWTFILTER and NLM_F_CREATE to create a new filter");
6bb16e7a 1192 err = -ENOENT;
1da177e4 1193 goto errout;
6bb16e7a 1194 }
1da177e4 1195 } else {
763dbf63
WC
1196 bool last;
1197
1da177e4 1198 switch (n->nlmsg_type) {
10297b99 1199 case RTM_NEWTFILTER:
12186be7
MU
1200 if (n->nlmsg_flags & NLM_F_EXCL) {
1201 if (tp_created)
715df5ec 1202 tcf_proto_destroy(tp, NULL);
c35a4acc 1203 NL_SET_ERR_MSG(extack, "Filter already exists");
6bb16e7a 1204 err = -EEXIST;
1da177e4 1205 goto errout;
12186be7 1206 }
1da177e4
LT
1207 break;
1208 case RTM_DELTFILTER:
7960d1da 1209 err = tfilter_del_notify(net, skb, n, tp, block,
c35a4acc
AA
1210 q, parent, fh, false, &last,
1211 extack);
40c81b25
JP
1212 if (err)
1213 goto errout;
763dbf63 1214 if (last) {
2190d1d0 1215 tcf_chain_tp_remove(chain, &chain_info, tp);
715df5ec 1216 tcf_proto_destroy(tp, extack);
763dbf63 1217 }
d7cf52c2 1218 goto errout;
1da177e4 1219 case RTM_GETTFILTER:
7960d1da
JP
1220 err = tfilter_notify(net, skb, n, tp, block, q, parent,
1221 fh, RTM_NEWTFILTER, true);
c35a4acc
AA
1222 if (err < 0)
1223 NL_SET_ERR_MSG(extack, "Failed to send filter notify message");
1da177e4
LT
1224 goto errout;
1225 default:
c35a4acc 1226 NL_SET_ERR_MSG(extack, "Invalid netlink message type");
1da177e4
LT
1227 err = -EINVAL;
1228 goto errout;
1229 }
1230 }
1231
2f7ef2f8 1232 err = tp->ops->change(net, skb, tp, cl, t->tcm_handle, tca, &fh,
7306db38
AA
1233 n->nlmsg_flags & NLM_F_CREATE ? TCA_ACT_NOREPLACE : TCA_ACT_REPLACE,
1234 extack);
12186be7 1235 if (err == 0) {
2190d1d0
JP
1236 if (tp_created)
1237 tcf_chain_tp_insert(chain, &chain_info, tp);
7960d1da 1238 tfilter_notify(net, skb, n, tp, block, q, parent, fh,
a10fa201 1239 RTM_NEWTFILTER, false);
12186be7
MU
1240 } else {
1241 if (tp_created)
715df5ec 1242 tcf_proto_destroy(tp, NULL);
12186be7 1243 }
1da177e4
LT
1244
1245errout:
5bc17018
JP
1246 if (chain)
1247 tcf_chain_put(chain);
1da177e4
LT
1248 if (err == -EAGAIN)
1249 /* Replay the request. */
1250 goto replay;
1251 return err;
1252}
1253
aa767bfe 1254struct tcf_dump_args {
1da177e4
LT
1255 struct tcf_walker w;
1256 struct sk_buff *skb;
1257 struct netlink_callback *cb;
7960d1da 1258 struct tcf_block *block;
a10fa201
JP
1259 struct Qdisc *q;
1260 u32 parent;
1da177e4
LT
1261};
1262
8113c095 1263static int tcf_node_dump(struct tcf_proto *tp, void *n, struct tcf_walker *arg)
1da177e4 1264{
aa767bfe 1265 struct tcf_dump_args *a = (void *)arg;
832d1d5b 1266 struct net *net = sock_net(a->skb->sk);
1da177e4 1267
7960d1da 1268 return tcf_fill_node(net, a->skb, tp, a->block, a->q, a->parent,
a10fa201 1269 n, NETLINK_CB(a->cb->skb).portid,
5a7a5555
JHS
1270 a->cb->nlh->nlmsg_seq, NLM_F_MULTI,
1271 RTM_NEWTFILTER);
1da177e4
LT
1272}
1273
a10fa201
JP
1274static bool tcf_chain_dump(struct tcf_chain *chain, struct Qdisc *q, u32 parent,
1275 struct sk_buff *skb, struct netlink_callback *cb,
acb31fae
JP
1276 long index_start, long *p_index)
1277{
1278 struct net *net = sock_net(skb->sk);
7960d1da 1279 struct tcf_block *block = chain->block;
acb31fae
JP
1280 struct tcmsg *tcm = nlmsg_data(cb->nlh);
1281 struct tcf_dump_args arg;
1282 struct tcf_proto *tp;
1283
1284 for (tp = rtnl_dereference(chain->filter_chain);
1285 tp; tp = rtnl_dereference(tp->next), (*p_index)++) {
1286 if (*p_index < index_start)
1287 continue;
1288 if (TC_H_MAJ(tcm->tcm_info) &&
1289 TC_H_MAJ(tcm->tcm_info) != tp->prio)
1290 continue;
1291 if (TC_H_MIN(tcm->tcm_info) &&
1292 TC_H_MIN(tcm->tcm_info) != tp->protocol)
1293 continue;
1294 if (*p_index > index_start)
1295 memset(&cb->args[1], 0,
1296 sizeof(cb->args) - sizeof(cb->args[0]));
1297 if (cb->args[1] == 0) {
7960d1da 1298 if (tcf_fill_node(net, skb, tp, block, q, parent, 0,
acb31fae
JP
1299 NETLINK_CB(cb->skb).portid,
1300 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1301 RTM_NEWTFILTER) <= 0)
5bc17018 1302 return false;
acb31fae
JP
1303
1304 cb->args[1] = 1;
1305 }
1306 if (!tp->ops->walk)
1307 continue;
1308 arg.w.fn = tcf_node_dump;
1309 arg.skb = skb;
1310 arg.cb = cb;
7960d1da 1311 arg.block = block;
a10fa201
JP
1312 arg.q = q;
1313 arg.parent = parent;
acb31fae
JP
1314 arg.w.stop = 0;
1315 arg.w.skip = cb->args[1] - 1;
1316 arg.w.count = 0;
1317 tp->ops->walk(tp, &arg.w);
1318 cb->args[1] = arg.w.count + 1;
1319 if (arg.w.stop)
5bc17018 1320 return false;
acb31fae 1321 }
5bc17018 1322 return true;
acb31fae
JP
1323}
1324
bd27a875 1325/* called with RTNL */
1da177e4
LT
1326static int tc_dump_tfilter(struct sk_buff *skb, struct netlink_callback *cb)
1327{
3b1e0a65 1328 struct net *net = sock_net(skb->sk);
5bc17018 1329 struct nlattr *tca[TCA_MAX + 1];
7960d1da 1330 struct Qdisc *q = NULL;
6529eaba 1331 struct tcf_block *block;
2190d1d0 1332 struct tcf_chain *chain;
942b8165 1333 struct tcmsg *tcm = nlmsg_data(cb->nlh);
acb31fae
JP
1334 long index_start;
1335 long index;
a10fa201 1336 u32 parent;
5bc17018 1337 int err;
1da177e4 1338
573ce260 1339 if (nlmsg_len(cb->nlh) < sizeof(*tcm))
1da177e4 1340 return skb->len;
5bc17018
JP
1341
1342 err = nlmsg_parse(cb->nlh, sizeof(*tcm), tca, TCA_MAX, NULL, NULL);
1343 if (err)
1344 return err;
1345
7960d1da
JP
1346 if (tcm->tcm_ifindex == TCM_IFINDEX_MAGIC_BLOCK) {
1347 block = tcf_block_lookup(net, tcm->tcm_block_index);
1348 if (!block)
1349 goto out;
d680b352
JP
1350 /* If we work with block index, q is NULL and parent value
1351 * will never be used in the following code. The check
1352 * in tcf_fill_node prevents it. However, compiler does not
1353 * see that far, so set parent to zero to silence the warning
1354 * about parent being uninitialized.
1355 */
1356 parent = 0;
a10fa201 1357 } else {
7960d1da
JP
1358 const struct Qdisc_class_ops *cops;
1359 struct net_device *dev;
1360 unsigned long cl = 0;
1361
1362 dev = __dev_get_by_index(net, tcm->tcm_ifindex);
1363 if (!dev)
1364 return skb->len;
1365
1366 parent = tcm->tcm_parent;
1367 if (!parent) {
1368 q = dev->qdisc;
1369 parent = q->handle;
1370 } else {
1371 q = qdisc_lookup(dev, TC_H_MAJ(tcm->tcm_parent));
1372 }
1373 if (!q)
1374 goto out;
1375 cops = q->ops->cl_ops;
1376 if (!cops)
143976ce 1377 goto out;
7960d1da
JP
1378 if (!cops->tcf_block)
1379 goto out;
1380 if (TC_H_MIN(tcm->tcm_parent)) {
1381 cl = cops->find(q, tcm->tcm_parent);
1382 if (cl == 0)
1383 goto out;
1384 }
1385 block = cops->tcf_block(q, cl, NULL);
1386 if (!block)
1387 goto out;
1388 if (tcf_block_shared(block))
1389 q = NULL;
1da177e4 1390 }
1da177e4 1391
acb31fae
JP
1392 index_start = cb->args[0];
1393 index = 0;
5bc17018
JP
1394
1395 list_for_each_entry(chain, &block->chain_list, list) {
1396 if (tca[TCA_CHAIN] &&
1397 nla_get_u32(tca[TCA_CHAIN]) != chain->index)
1398 continue;
a10fa201 1399 if (!tcf_chain_dump(chain, q, parent, skb, cb,
5ae437ad
RK
1400 index_start, &index)) {
1401 err = -EMSGSIZE;
5bc17018 1402 break;
5ae437ad 1403 }
5bc17018
JP
1404 }
1405
acb31fae 1406 cb->args[0] = index;
1da177e4 1407
1da177e4 1408out:
5ae437ad
RK
1409 /* If we did no progress, the error (EMSGSIZE) is real */
1410 if (skb->len == 0 && err)
1411 return err;
1da177e4
LT
1412 return skb->len;
1413}
1414
18d0264f 1415void tcf_exts_destroy(struct tcf_exts *exts)
1da177e4
LT
1416{
1417#ifdef CONFIG_NET_CLS_ACT
22dc13c8
WC
1418 LIST_HEAD(actions);
1419
2d132eba 1420 ASSERT_RTNL();
22dc13c8
WC
1421 tcf_exts_to_list(exts, &actions);
1422 tcf_action_destroy(&actions, TCA_ACT_UNBIND);
1423 kfree(exts->actions);
1424 exts->nr_actions = 0;
1da177e4
LT
1425#endif
1426}
aa767bfe 1427EXPORT_SYMBOL(tcf_exts_destroy);
1da177e4 1428
c1b52739 1429int tcf_exts_validate(struct net *net, struct tcf_proto *tp, struct nlattr **tb,
50a56190
AA
1430 struct nlattr *rate_tlv, struct tcf_exts *exts, bool ovr,
1431 struct netlink_ext_ack *extack)
1da177e4 1432{
1da177e4
LT
1433#ifdef CONFIG_NET_CLS_ACT
1434 {
1da177e4 1435 struct tc_action *act;
d04e6990 1436 size_t attr_size = 0;
1da177e4 1437
5da57f42 1438 if (exts->police && tb[exts->police]) {
9fb9f251
JP
1439 act = tcf_action_init_1(net, tp, tb[exts->police],
1440 rate_tlv, "police", ovr,
aea0d727 1441 TCA_ACT_BIND, extack);
ab27cfb8
PM
1442 if (IS_ERR(act))
1443 return PTR_ERR(act);
1da177e4 1444
33be6271 1445 act->type = exts->type = TCA_OLD_COMPAT;
22dc13c8
WC
1446 exts->actions[0] = act;
1447 exts->nr_actions = 1;
5da57f42 1448 } else if (exts->action && tb[exts->action]) {
22dc13c8
WC
1449 LIST_HEAD(actions);
1450 int err, i = 0;
1451
9fb9f251
JP
1452 err = tcf_action_init(net, tp, tb[exts->action],
1453 rate_tlv, NULL, ovr, TCA_ACT_BIND,
d04e6990 1454 &actions, &attr_size, extack);
33be6271
WC
1455 if (err)
1456 return err;
22dc13c8
WC
1457 list_for_each_entry(act, &actions, list)
1458 exts->actions[i++] = act;
1459 exts->nr_actions = i;
1da177e4 1460 }
e4b95c41 1461 exts->net = net;
1da177e4 1462 }
1da177e4 1463#else
5da57f42 1464 if ((exts->action && tb[exts->action]) ||
50a56190
AA
1465 (exts->police && tb[exts->police])) {
1466 NL_SET_ERR_MSG(extack, "Classifier actions are not supported per compile options (CONFIG_NET_CLS_ACT)");
1da177e4 1467 return -EOPNOTSUPP;
50a56190 1468 }
1da177e4
LT
1469#endif
1470
1471 return 0;
1472}
aa767bfe 1473EXPORT_SYMBOL(tcf_exts_validate);
1da177e4 1474
9b0d4446 1475void tcf_exts_change(struct tcf_exts *dst, struct tcf_exts *src)
1da177e4
LT
1476{
1477#ifdef CONFIG_NET_CLS_ACT
22dc13c8
WC
1478 struct tcf_exts old = *dst;
1479
9b0d4446 1480 *dst = *src;
22dc13c8 1481 tcf_exts_destroy(&old);
1da177e4
LT
1482#endif
1483}
aa767bfe 1484EXPORT_SYMBOL(tcf_exts_change);
1da177e4 1485
22dc13c8
WC
1486#ifdef CONFIG_NET_CLS_ACT
1487static struct tc_action *tcf_exts_first_act(struct tcf_exts *exts)
1488{
1489 if (exts->nr_actions == 0)
1490 return NULL;
1491 else
1492 return exts->actions[0];
1493}
1494#endif
33be6271 1495
5da57f42 1496int tcf_exts_dump(struct sk_buff *skb, struct tcf_exts *exts)
1da177e4
LT
1497{
1498#ifdef CONFIG_NET_CLS_ACT
9cc63db5
CW
1499 struct nlattr *nest;
1500
978dfd8d 1501 if (exts->action && tcf_exts_has_actions(exts)) {
1da177e4
LT
1502 /*
1503 * again for backward compatible mode - we want
1504 * to work with both old and new modes of entering
1505 * tc data even if iproute2 was newer - jhs
1506 */
33be6271 1507 if (exts->type != TCA_OLD_COMPAT) {
22dc13c8
WC
1508 LIST_HEAD(actions);
1509
5da57f42 1510 nest = nla_nest_start(skb, exts->action);
4b3550ef
PM
1511 if (nest == NULL)
1512 goto nla_put_failure;
22dc13c8
WC
1513
1514 tcf_exts_to_list(exts, &actions);
1515 if (tcf_action_dump(skb, &actions, 0, 0) < 0)
add93b61 1516 goto nla_put_failure;
4b3550ef 1517 nla_nest_end(skb, nest);
5da57f42 1518 } else if (exts->police) {
33be6271 1519 struct tc_action *act = tcf_exts_first_act(exts);
5da57f42 1520 nest = nla_nest_start(skb, exts->police);
63acd680 1521 if (nest == NULL || !act)
4b3550ef 1522 goto nla_put_failure;
33be6271 1523 if (tcf_action_dump_old(skb, act, 0, 0) < 0)
add93b61 1524 goto nla_put_failure;
4b3550ef 1525 nla_nest_end(skb, nest);
1da177e4
LT
1526 }
1527 }
1da177e4 1528 return 0;
9cc63db5
CW
1529
1530nla_put_failure:
1531 nla_nest_cancel(skb, nest);
1da177e4 1532 return -1;
9cc63db5
CW
1533#else
1534 return 0;
1535#endif
1da177e4 1536}
aa767bfe 1537EXPORT_SYMBOL(tcf_exts_dump);
1da177e4 1538
aa767bfe 1539
5da57f42 1540int tcf_exts_dump_stats(struct sk_buff *skb, struct tcf_exts *exts)
1da177e4
LT
1541{
1542#ifdef CONFIG_NET_CLS_ACT
33be6271 1543 struct tc_action *a = tcf_exts_first_act(exts);
b057df24 1544 if (a != NULL && tcf_action_copy_stats(skb, a, 1) < 0)
33be6271 1545 return -1;
1da177e4
LT
1546#endif
1547 return 0;
1da177e4 1548}
aa767bfe 1549EXPORT_SYMBOL(tcf_exts_dump_stats);
1da177e4 1550
717503b9
JP
1551static int tc_exts_setup_cb_egdev_call(struct tcf_exts *exts,
1552 enum tc_setup_type type,
1553 void *type_data, bool err_stop)
b3f55bdd
JP
1554{
1555 int ok_count = 0;
1556#ifdef CONFIG_NET_CLS_ACT
1557 const struct tc_action *a;
1558 struct net_device *dev;
9d452ceb 1559 int i, ret;
b3f55bdd
JP
1560
1561 if (!tcf_exts_has_actions(exts))
1562 return 0;
1563
9d452ceb
OG
1564 for (i = 0; i < exts->nr_actions; i++) {
1565 a = exts->actions[i];
b3f55bdd
JP
1566 if (!a->ops->get_dev)
1567 continue;
1568 dev = a->ops->get_dev(a);
7612fb03 1569 if (!dev)
b3f55bdd
JP
1570 continue;
1571 ret = tc_setup_cb_egdev_call(dev, type, type_data, err_stop);
1572 if (ret < 0)
1573 return ret;
1574 ok_count += ret;
1575 }
1576#endif
1577 return ok_count;
1578}
717503b9 1579
208c0f4b
JP
1580int tc_setup_cb_call(struct tcf_block *block, struct tcf_exts *exts,
1581 enum tc_setup_type type, void *type_data, bool err_stop)
717503b9 1582{
208c0f4b
JP
1583 int ok_count;
1584 int ret;
1585
1586 ret = tcf_block_cb_call(block, type, type_data, err_stop);
1587 if (ret < 0)
1588 return ret;
1589 ok_count = ret;
1590
1591 if (!exts)
1592 return ok_count;
1593 ret = tc_exts_setup_cb_egdev_call(exts, type, type_data, err_stop);
1594 if (ret < 0)
1595 return ret;
1596 ok_count += ret;
1597
1598 return ok_count;
717503b9
JP
1599}
1600EXPORT_SYMBOL(tc_setup_cb_call);
b3f55bdd 1601
48617387
JP
1602static __net_init int tcf_net_init(struct net *net)
1603{
1604 struct tcf_net *tn = net_generic(net, tcf_net_id);
1605
1606 idr_init(&tn->idr);
1607 return 0;
1608}
1609
1610static void __net_exit tcf_net_exit(struct net *net)
1611{
1612 struct tcf_net *tn = net_generic(net, tcf_net_id);
1613
1614 idr_destroy(&tn->idr);
1615}
1616
1617static struct pernet_operations tcf_net_ops = {
1618 .init = tcf_net_init,
1619 .exit = tcf_net_exit,
1620 .id = &tcf_net_id,
1621 .size = sizeof(struct tcf_net),
1622};
1623
1da177e4
LT
1624static int __init tc_filter_init(void)
1625{
48617387
JP
1626 int err;
1627
7aa0045d
CW
1628 tc_filter_wq = alloc_ordered_workqueue("tc_filter_workqueue", 0);
1629 if (!tc_filter_wq)
1630 return -ENOMEM;
1631
48617387
JP
1632 err = register_pernet_subsys(&tcf_net_ops);
1633 if (err)
1634 goto err_register_pernet_subsys;
1635
b97bac64
FW
1636 rtnl_register(PF_UNSPEC, RTM_NEWTFILTER, tc_ctl_tfilter, NULL, 0);
1637 rtnl_register(PF_UNSPEC, RTM_DELTFILTER, tc_ctl_tfilter, NULL, 0);
82623c0d 1638 rtnl_register(PF_UNSPEC, RTM_GETTFILTER, tc_ctl_tfilter,
b97bac64 1639 tc_dump_tfilter, 0);
1da177e4 1640
1da177e4 1641 return 0;
48617387
JP
1642
1643err_register_pernet_subsys:
1644 destroy_workqueue(tc_filter_wq);
1645 return err;
1da177e4
LT
1646}
1647
1648subsys_initcall(tc_filter_init);