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