1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * net/sched/act_pedit.c Generic packet editor
5 * Authors: Jamal Hadi Salim (2002-4)
8 #include <linux/types.h>
9 #include <linux/kernel.h>
10 #include <linux/string.h>
11 #include <linux/errno.h>
12 #include <linux/skbuff.h>
13 #include <linux/rtnetlink.h>
14 #include <linux/module.h>
15 #include <linux/init.h>
17 #include <linux/ipv6.h>
18 #include <linux/slab.h>
20 #include <net/netlink.h>
21 #include <net/pkt_sched.h>
22 #include <linux/tc_act/tc_pedit.h>
23 #include <net/tc_act/tc_pedit.h>
24 #include <uapi/linux/tc_act/tc_pedit.h>
25 #include <net/pkt_cls.h>
27 static struct tc_action_ops act_pedit_ops;
29 static const struct nla_policy pedit_policy[TCA_PEDIT_MAX + 1] = {
30 [TCA_PEDIT_PARMS] = { .len = sizeof(struct tc_pedit) },
31 [TCA_PEDIT_PARMS_EX] = { .len = sizeof(struct tc_pedit) },
32 [TCA_PEDIT_KEYS_EX] = { .type = NLA_NESTED },
35 static const struct nla_policy pedit_key_ex_policy[TCA_PEDIT_KEY_EX_MAX + 1] = {
36 [TCA_PEDIT_KEY_EX_HTYPE] = { .type = NLA_U16 },
37 [TCA_PEDIT_KEY_EX_CMD] = { .type = NLA_U16 },
40 static struct tcf_pedit_key_ex *tcf_pedit_keys_ex_parse(struct nlattr *nla,
43 struct tcf_pedit_key_ex *keys_ex;
44 struct tcf_pedit_key_ex *k;
45 const struct nlattr *ka;
52 keys_ex = kcalloc(n, sizeof(*k), GFP_KERNEL);
54 return ERR_PTR(-ENOMEM);
58 nla_for_each_nested(ka, nla, rem) {
59 struct nlattr *tb[TCA_PEDIT_KEY_EX_MAX + 1];
67 if (nla_type(ka) != TCA_PEDIT_KEY_EX) {
72 err = nla_parse_nested_deprecated(tb, TCA_PEDIT_KEY_EX_MAX,
73 ka, pedit_key_ex_policy,
78 if (!tb[TCA_PEDIT_KEY_EX_HTYPE] ||
79 !tb[TCA_PEDIT_KEY_EX_CMD]) {
84 k->htype = nla_get_u16(tb[TCA_PEDIT_KEY_EX_HTYPE]);
85 k->cmd = nla_get_u16(tb[TCA_PEDIT_KEY_EX_CMD]);
87 if (k->htype > TCA_PEDIT_HDR_TYPE_MAX ||
88 k->cmd > TCA_PEDIT_CMD_MAX) {
108 static int tcf_pedit_key_ex_dump(struct sk_buff *skb,
109 struct tcf_pedit_key_ex *keys_ex, int n)
111 struct nlattr *keys_start = nla_nest_start_noflag(skb,
117 struct nlattr *key_start;
119 key_start = nla_nest_start_noflag(skb, TCA_PEDIT_KEY_EX);
123 if (nla_put_u16(skb, TCA_PEDIT_KEY_EX_HTYPE, keys_ex->htype) ||
124 nla_put_u16(skb, TCA_PEDIT_KEY_EX_CMD, keys_ex->cmd))
127 nla_nest_end(skb, key_start);
132 nla_nest_end(skb, keys_start);
136 nla_nest_cancel(skb, keys_start);
140 static void tcf_pedit_cleanup_rcu(struct rcu_head *head)
142 struct tcf_pedit_parms *parms =
143 container_of(head, struct tcf_pedit_parms, rcu);
145 kfree(parms->tcfp_keys_ex);
146 kfree(parms->tcfp_keys);
151 static int tcf_pedit_init(struct net *net, struct nlattr *nla,
152 struct nlattr *est, struct tc_action **a,
153 struct tcf_proto *tp, u32 flags,
154 struct netlink_ext_ack *extack)
156 struct tc_action_net *tn = net_generic(net, act_pedit_ops.net_id);
157 bool bind = flags & TCA_ACT_FLAGS_BIND;
158 struct tcf_chain *goto_ch = NULL;
159 struct tcf_pedit_parms *oparms, *nparms;
160 struct nlattr *tb[TCA_PEDIT_MAX + 1];
161 struct tc_pedit *parm;
162 struct nlattr *pattr;
169 NL_SET_ERR_MSG_MOD(extack, "Pedit requires attributes to be passed");
173 err = nla_parse_nested_deprecated(tb, TCA_PEDIT_MAX, nla,
178 pattr = tb[TCA_PEDIT_PARMS];
180 pattr = tb[TCA_PEDIT_PARMS_EX];
182 NL_SET_ERR_MSG_MOD(extack, "Missing required TCA_PEDIT_PARMS or TCA_PEDIT_PARMS_EX pedit attribute");
186 parm = nla_data(pattr);
189 err = tcf_idr_check_alloc(tn, &index, a, bind);
191 ret = tcf_idr_create_from_flags(tn, index, est, a,
192 &act_pedit_ops, bind, flags);
194 tcf_idr_cleanup(tn, index);
198 } else if (err > 0) {
201 if (!(flags & TCA_ACT_FLAGS_REPLACE)) {
210 NL_SET_ERR_MSG_MOD(extack, "Pedit requires keys to be passed");
214 ksize = parm->nkeys * sizeof(struct tc_pedit_key);
215 if (nla_len(pattr) < sizeof(*parm) + ksize) {
216 NL_SET_ERR_MSG_ATTR(extack, pattr, "Length of TCA_PEDIT_PARMS or TCA_PEDIT_PARMS_EX pedit attribute is invalid");
221 nparms = kzalloc(sizeof(*nparms), GFP_KERNEL);
227 nparms->tcfp_keys_ex =
228 tcf_pedit_keys_ex_parse(tb[TCA_PEDIT_KEYS_EX], parm->nkeys);
229 if (IS_ERR(nparms->tcfp_keys_ex)) {
230 ret = PTR_ERR(nparms->tcfp_keys_ex);
234 err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
240 nparms->tcfp_off_max_hint = 0;
241 nparms->tcfp_flags = parm->flags;
242 nparms->tcfp_nkeys = parm->nkeys;
244 nparms->tcfp_keys = kmalloc(ksize, GFP_KERNEL);
245 if (!nparms->tcfp_keys) {
250 memcpy(nparms->tcfp_keys, parm->keys, ksize);
252 for (i = 0; i < nparms->tcfp_nkeys; ++i) {
253 u32 cur = nparms->tcfp_keys[i].off;
255 /* sanitize the shift value for any later use */
256 nparms->tcfp_keys[i].shift = min_t(size_t,
257 BITS_PER_TYPE(int) - 1,
258 nparms->tcfp_keys[i].shift);
260 /* The AT option can read a single byte, we can bound the actual
261 * value with uchar max.
263 cur += (0xff & nparms->tcfp_keys[i].offmask) >> nparms->tcfp_keys[i].shift;
265 /* Each key touches 4 bytes starting from the computed offset */
266 nparms->tcfp_off_max_hint =
267 max(nparms->tcfp_off_max_hint, cur + 4);
272 spin_lock_bh(&p->tcf_lock);
273 goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
274 oparms = rcu_replace_pointer(p->parms, nparms, 1);
275 spin_unlock_bh(&p->tcf_lock);
278 call_rcu(&oparms->rcu, tcf_pedit_cleanup_rcu);
281 tcf_chain_put_by_act(goto_ch);
287 tcf_chain_put_by_act(goto_ch);
289 kfree(nparms->tcfp_keys_ex);
293 tcf_idr_release(*a, bind);
297 static void tcf_pedit_cleanup(struct tc_action *a)
299 struct tcf_pedit *p = to_pedit(a);
300 struct tcf_pedit_parms *parms;
302 parms = rcu_dereference_protected(p->parms, 1);
305 call_rcu(&parms->rcu, tcf_pedit_cleanup_rcu);
308 static bool offset_valid(struct sk_buff *skb, int offset)
310 if (offset > 0 && offset > skb->len)
313 if (offset < 0 && -offset > skb_headroom(skb))
319 static int pedit_l4_skb_offset(struct sk_buff *skb, int *hoffset, const int header_type)
321 const int noff = skb_network_offset(skb);
325 switch (skb->protocol) {
326 case htons(ETH_P_IP): {
327 const struct iphdr *iph = skb_header_pointer(skb, noff, sizeof(_iph), &_iph);
331 *hoffset = noff + iph->ihl * 4;
335 case htons(ETH_P_IPV6):
336 ret = ipv6_find_hdr(skb, hoffset, header_type, NULL, NULL) == header_type ? 0 : -EINVAL;
343 static int pedit_skb_hdr_offset(struct sk_buff *skb,
344 enum pedit_header_type htype, int *hoffset)
347 /* 'htype' is validated in the netlink parsing */
349 case TCA_PEDIT_KEY_EX_HDR_TYPE_ETH:
350 if (skb_mac_header_was_set(skb)) {
351 *hoffset = skb_mac_offset(skb);
355 case TCA_PEDIT_KEY_EX_HDR_TYPE_NETWORK:
356 case TCA_PEDIT_KEY_EX_HDR_TYPE_IP4:
357 case TCA_PEDIT_KEY_EX_HDR_TYPE_IP6:
358 *hoffset = skb_network_offset(skb);
361 case TCA_PEDIT_KEY_EX_HDR_TYPE_TCP:
362 ret = pedit_l4_skb_offset(skb, hoffset, IPPROTO_TCP);
364 case TCA_PEDIT_KEY_EX_HDR_TYPE_UDP:
365 ret = pedit_l4_skb_offset(skb, hoffset, IPPROTO_UDP);
373 static int tcf_pedit_act(struct sk_buff *skb, const struct tc_action *a,
374 struct tcf_result *res)
376 enum pedit_header_type htype = TCA_PEDIT_KEY_EX_HDR_TYPE_NETWORK;
377 enum pedit_cmd cmd = TCA_PEDIT_KEY_EX_CMD_SET;
378 struct tcf_pedit *p = to_pedit(a);
379 struct tcf_pedit_key_ex *tkey_ex;
380 struct tcf_pedit_parms *parms;
381 struct tc_pedit_key *tkey;
385 parms = rcu_dereference_bh(p->parms);
387 max_offset = (skb_transport_header_was_set(skb) ?
388 skb_transport_offset(skb) :
389 skb_network_offset(skb)) +
390 parms->tcfp_off_max_hint;
391 if (skb_ensure_writable(skb, min(skb->len, max_offset)))
394 tcf_lastuse_update(&p->tcf_tm);
395 tcf_action_update_bstats(&p->common, skb);
397 tkey = parms->tcfp_keys;
398 tkey_ex = parms->tcfp_keys_ex;
400 for (i = parms->tcfp_nkeys; i > 0; i--, tkey++) {
401 int offset = tkey->off;
408 htype = tkey_ex->htype;
414 rc = pedit_skb_hdr_offset(skb, htype, &hoffset);
416 pr_info_ratelimited("tc action pedit unable to extract header offset for header type (0x%x)\n", htype);
423 if (!offset_valid(skb, hoffset + tkey->at)) {
424 pr_info("tc action pedit 'at' offset %d out of bounds\n",
428 d = skb_header_pointer(skb, hoffset + tkey->at,
432 offset += (*d & tkey->offmask) >> tkey->shift;
436 pr_info("tc action pedit offset must be on 32 bit boundaries\n");
440 if (!offset_valid(skb, hoffset + offset)) {
441 pr_info("tc action pedit offset %d out of bounds\n",
446 ptr = skb_header_pointer(skb, hoffset + offset,
447 sizeof(hdata), &hdata);
450 /* just do it, baby */
452 case TCA_PEDIT_KEY_EX_CMD_SET:
455 case TCA_PEDIT_KEY_EX_CMD_ADD:
456 val = (*ptr + tkey->val) & ~tkey->mask;
459 pr_info("tc action pedit bad command (%d)\n",
464 *ptr = ((*ptr & tkey->mask) ^ val);
466 skb_store_bits(skb, hoffset + offset, ptr, 4);
472 spin_lock(&p->tcf_lock);
473 p->tcf_qstats.overlimits++;
474 spin_unlock(&p->tcf_lock);
476 return p->tcf_action;
479 static void tcf_pedit_stats_update(struct tc_action *a, u64 bytes, u64 packets,
480 u64 drops, u64 lastuse, bool hw)
482 struct tcf_pedit *d = to_pedit(a);
483 struct tcf_t *tm = &d->tcf_tm;
485 tcf_action_update_stats(a, bytes, packets, drops, hw);
486 tm->lastuse = max_t(u64, tm->lastuse, lastuse);
489 static int tcf_pedit_dump(struct sk_buff *skb, struct tc_action *a,
492 unsigned char *b = skb_tail_pointer(skb);
493 struct tcf_pedit *p = to_pedit(a);
494 struct tcf_pedit_parms *parms;
495 struct tc_pedit *opt;
499 spin_lock_bh(&p->tcf_lock);
500 parms = rcu_dereference_protected(p->parms, 1);
501 s = struct_size(opt, keys, parms->tcfp_nkeys);
503 opt = kzalloc(s, GFP_ATOMIC);
504 if (unlikely(!opt)) {
505 spin_unlock_bh(&p->tcf_lock);
509 memcpy(opt->keys, parms->tcfp_keys,
510 flex_array_size(opt, keys, parms->tcfp_nkeys));
511 opt->index = p->tcf_index;
512 opt->nkeys = parms->tcfp_nkeys;
513 opt->flags = parms->tcfp_flags;
514 opt->action = p->tcf_action;
515 opt->refcnt = refcount_read(&p->tcf_refcnt) - ref;
516 opt->bindcnt = atomic_read(&p->tcf_bindcnt) - bind;
518 if (parms->tcfp_keys_ex) {
519 if (tcf_pedit_key_ex_dump(skb, parms->tcfp_keys_ex,
521 goto nla_put_failure;
523 if (nla_put(skb, TCA_PEDIT_PARMS_EX, s, opt))
524 goto nla_put_failure;
526 if (nla_put(skb, TCA_PEDIT_PARMS, s, opt))
527 goto nla_put_failure;
530 tcf_tm_dump(&t, &p->tcf_tm);
531 if (nla_put_64bit(skb, TCA_PEDIT_TM, sizeof(t), &t, TCA_PEDIT_PAD))
532 goto nla_put_failure;
533 spin_unlock_bh(&p->tcf_lock);
539 spin_unlock_bh(&p->tcf_lock);
545 static int tcf_pedit_offload_act_setup(struct tc_action *act, void *entry_data,
546 u32 *index_inc, bool bind,
547 struct netlink_ext_ack *extack)
550 struct flow_action_entry *entry = entry_data;
553 for (k = 0; k < tcf_pedit_nkeys(act); k++) {
554 switch (tcf_pedit_cmd(act, k)) {
555 case TCA_PEDIT_KEY_EX_CMD_SET:
556 entry->id = FLOW_ACTION_MANGLE;
558 case TCA_PEDIT_KEY_EX_CMD_ADD:
559 entry->id = FLOW_ACTION_ADD;
562 NL_SET_ERR_MSG_MOD(extack, "Unsupported pedit command offload");
565 entry->mangle.htype = tcf_pedit_htype(act, k);
566 entry->mangle.mask = tcf_pedit_mask(act, k);
567 entry->mangle.val = tcf_pedit_val(act, k);
568 entry->mangle.offset = tcf_pedit_offset(act, k);
569 entry->hw_stats = tc_act_hw_stats(act->hw_stats);
580 static struct tc_action_ops act_pedit_ops = {
583 .owner = THIS_MODULE,
584 .act = tcf_pedit_act,
585 .stats_update = tcf_pedit_stats_update,
586 .dump = tcf_pedit_dump,
587 .cleanup = tcf_pedit_cleanup,
588 .init = tcf_pedit_init,
589 .offload_act_setup = tcf_pedit_offload_act_setup,
590 .size = sizeof(struct tcf_pedit),
593 static __net_init int pedit_init_net(struct net *net)
595 struct tc_action_net *tn = net_generic(net, act_pedit_ops.net_id);
597 return tc_action_net_init(net, tn, &act_pedit_ops);
600 static void __net_exit pedit_exit_net(struct list_head *net_list)
602 tc_action_net_exit(net_list, act_pedit_ops.net_id);
605 static struct pernet_operations pedit_net_ops = {
606 .init = pedit_init_net,
607 .exit_batch = pedit_exit_net,
608 .id = &act_pedit_ops.net_id,
609 .size = sizeof(struct tc_action_net),
612 MODULE_AUTHOR("Jamal Hadi Salim(2002-4)");
613 MODULE_DESCRIPTION("Generic Packet Editor actions");
614 MODULE_LICENSE("GPL");
616 static int __init pedit_init_module(void)
618 return tcf_register_action(&act_pedit_ops, &pedit_net_ops);
621 static void __exit pedit_cleanup_module(void)
623 tcf_unregister_action(&act_pedit_ops, &pedit_net_ops);
626 module_init(pedit_init_module);
627 module_exit(pedit_cleanup_module);