1 // SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
2 /* Copyright (C) 2019 Netronome Systems, Inc. */
4 #include <linux/if_arp.h>
5 #include <linux/init.h>
6 #include <linux/kernel.h>
7 #include <linux/module.h>
8 #include <linux/mpls.h>
9 #include <linux/rtnetlink.h>
10 #include <linux/skbuff.h>
11 #include <linux/tc_act/tc_mpls.h>
13 #include <net/netlink.h>
14 #include <net/pkt_sched.h>
15 #include <net/pkt_cls.h>
16 #include <net/tc_act/tc_mpls.h>
17 #include <net/tc_wrapper.h>
19 static struct tc_action_ops act_mpls_ops;
21 #define ACT_MPLS_TTL_DEFAULT 255
23 static __be32 tcf_mpls_get_lse(struct mpls_shim_hdr *lse,
24 struct tcf_mpls_params *p, bool set_bos)
29 new_lse = be32_to_cpu(lse->label_stack_entry);
31 if (p->tcfm_label != ACT_MPLS_LABEL_NOT_SET) {
32 new_lse &= ~MPLS_LS_LABEL_MASK;
33 new_lse |= p->tcfm_label << MPLS_LS_LABEL_SHIFT;
36 new_lse &= ~MPLS_LS_TTL_MASK;
37 new_lse |= p->tcfm_ttl << MPLS_LS_TTL_SHIFT;
39 if (p->tcfm_tc != ACT_MPLS_TC_NOT_SET) {
40 new_lse &= ~MPLS_LS_TC_MASK;
41 new_lse |= p->tcfm_tc << MPLS_LS_TC_SHIFT;
43 if (p->tcfm_bos != ACT_MPLS_BOS_NOT_SET) {
44 new_lse &= ~MPLS_LS_S_MASK;
45 new_lse |= p->tcfm_bos << MPLS_LS_S_SHIFT;
47 new_lse |= 1 << MPLS_LS_S_SHIFT;
50 return cpu_to_be32(new_lse);
53 TC_INDIRECT_SCOPE int tcf_mpls_act(struct sk_buff *skb,
54 const struct tc_action *a,
55 struct tcf_result *res)
57 struct tcf_mpls *m = to_mpls(a);
58 struct tcf_mpls_params *p;
62 tcf_lastuse_update(&m->tcf_tm);
63 bstats_update(this_cpu_ptr(m->common.cpu_bstats), skb);
65 /* Ensure 'data' points at mac_header prior calling mpls manipulating
68 if (skb_at_tc_ingress(skb)) {
69 skb_push_rcsum(skb, skb->mac_len);
70 mac_len = skb->mac_len;
72 mac_len = skb_network_offset(skb);
75 ret = READ_ONCE(m->tcf_action);
77 p = rcu_dereference_bh(m->mpls_p);
79 switch (p->tcfm_action) {
80 case TCA_MPLS_ACT_POP:
81 if (skb_mpls_pop(skb, p->tcfm_proto, mac_len,
82 skb->dev && skb->dev->type == ARPHRD_ETHER))
85 case TCA_MPLS_ACT_PUSH:
86 new_lse = tcf_mpls_get_lse(NULL, p, !eth_p_mpls(skb_protocol(skb, true)));
87 if (skb_mpls_push(skb, new_lse, p->tcfm_proto, mac_len,
88 skb->dev && skb->dev->type == ARPHRD_ETHER))
91 case TCA_MPLS_ACT_MAC_PUSH:
92 if (skb_vlan_tag_present(skb)) {
93 if (__vlan_insert_inner_tag(skb, skb->vlan_proto,
94 skb_vlan_tag_get(skb),
98 skb->protocol = skb->vlan_proto;
99 __vlan_hwaccel_clear_tag(skb);
102 new_lse = tcf_mpls_get_lse(NULL, p, mac_len ||
103 !eth_p_mpls(skb->protocol));
105 if (skb_mpls_push(skb, new_lse, p->tcfm_proto, 0, false))
108 case TCA_MPLS_ACT_MODIFY:
109 if (!pskb_may_pull(skb,
110 skb_network_offset(skb) + MPLS_HLEN))
112 new_lse = tcf_mpls_get_lse(mpls_hdr(skb), p, false);
113 if (skb_mpls_update_lse(skb, new_lse))
116 case TCA_MPLS_ACT_DEC_TTL:
117 if (skb_mpls_dec_ttl(skb))
122 if (skb_at_tc_ingress(skb))
123 skb_pull_rcsum(skb, skb->mac_len);
128 qstats_drop_inc(this_cpu_ptr(m->common.cpu_qstats));
132 static int valid_label(const struct nlattr *attr,
133 struct netlink_ext_ack *extack)
135 const u32 *label = nla_data(attr);
137 if (nla_len(attr) != sizeof(*label)) {
138 NL_SET_ERR_MSG_MOD(extack, "Invalid MPLS label length");
142 if (*label & ~MPLS_LABEL_MASK || *label == MPLS_LABEL_IMPLNULL) {
143 NL_SET_ERR_MSG_MOD(extack, "MPLS label out of range");
150 static const struct nla_policy mpls_policy[TCA_MPLS_MAX + 1] = {
151 [TCA_MPLS_PARMS] = NLA_POLICY_EXACT_LEN(sizeof(struct tc_mpls)),
152 [TCA_MPLS_PROTO] = { .type = NLA_U16 },
153 [TCA_MPLS_LABEL] = NLA_POLICY_VALIDATE_FN(NLA_BINARY,
155 [TCA_MPLS_TC] = NLA_POLICY_RANGE(NLA_U8, 0, 7),
156 [TCA_MPLS_TTL] = NLA_POLICY_MIN(NLA_U8, 1),
157 [TCA_MPLS_BOS] = NLA_POLICY_RANGE(NLA_U8, 0, 1),
160 static int tcf_mpls_init(struct net *net, struct nlattr *nla,
161 struct nlattr *est, struct tc_action **a,
162 struct tcf_proto *tp, u32 flags,
163 struct netlink_ext_ack *extack)
165 struct tc_action_net *tn = net_generic(net, act_mpls_ops.net_id);
166 bool bind = flags & TCA_ACT_FLAGS_BIND;
167 struct nlattr *tb[TCA_MPLS_MAX + 1];
168 struct tcf_chain *goto_ch = NULL;
169 struct tcf_mpls_params *p;
170 struct tc_mpls *parm;
178 NL_SET_ERR_MSG_MOD(extack, "Missing netlink attributes");
182 err = nla_parse_nested(tb, TCA_MPLS_MAX, nla, mpls_policy, extack);
186 if (!tb[TCA_MPLS_PARMS]) {
187 NL_SET_ERR_MSG_MOD(extack, "No MPLS params");
190 parm = nla_data(tb[TCA_MPLS_PARMS]);
193 err = tcf_idr_check_alloc(tn, &index, a, bind);
201 ret = tcf_idr_create(tn, index, est, a, &act_mpls_ops, bind,
204 tcf_idr_cleanup(tn, index);
209 } else if (!(flags & TCA_ACT_FLAGS_REPLACE)) {
210 tcf_idr_release(*a, bind);
214 /* Verify parameters against action type. */
215 switch (parm->m_action) {
216 case TCA_MPLS_ACT_POP:
217 if (!tb[TCA_MPLS_PROTO]) {
218 NL_SET_ERR_MSG_MOD(extack, "Protocol must be set for MPLS pop");
222 if (!eth_proto_is_802_3(nla_get_be16(tb[TCA_MPLS_PROTO]))) {
223 NL_SET_ERR_MSG_MOD(extack, "Invalid protocol type for MPLS pop");
227 if (tb[TCA_MPLS_LABEL] || tb[TCA_MPLS_TTL] || tb[TCA_MPLS_TC] ||
229 NL_SET_ERR_MSG_MOD(extack, "Label, TTL, TC or BOS cannot be used with MPLS pop");
234 case TCA_MPLS_ACT_DEC_TTL:
235 if (tb[TCA_MPLS_PROTO] || tb[TCA_MPLS_LABEL] ||
236 tb[TCA_MPLS_TTL] || tb[TCA_MPLS_TC] || tb[TCA_MPLS_BOS]) {
237 NL_SET_ERR_MSG_MOD(extack, "Label, TTL, TC, BOS or protocol cannot be used with MPLS dec_ttl");
242 case TCA_MPLS_ACT_PUSH:
243 case TCA_MPLS_ACT_MAC_PUSH:
244 if (!tb[TCA_MPLS_LABEL]) {
245 NL_SET_ERR_MSG_MOD(extack, "Label is required for MPLS push");
249 if (tb[TCA_MPLS_PROTO] &&
250 !eth_p_mpls(nla_get_be16(tb[TCA_MPLS_PROTO]))) {
251 NL_SET_ERR_MSG_MOD(extack, "Protocol must be an MPLS type for MPLS push");
252 err = -EPROTONOSUPPORT;
255 /* Push needs a TTL - if not specified, set a default value. */
256 if (!tb[TCA_MPLS_TTL]) {
257 #if IS_ENABLED(CONFIG_MPLS)
258 mpls_ttl = net->mpls.default_ttl ?
259 net->mpls.default_ttl : ACT_MPLS_TTL_DEFAULT;
261 mpls_ttl = ACT_MPLS_TTL_DEFAULT;
265 case TCA_MPLS_ACT_MODIFY:
266 if (tb[TCA_MPLS_PROTO]) {
267 NL_SET_ERR_MSG_MOD(extack, "Protocol cannot be used with MPLS modify");
273 NL_SET_ERR_MSG_MOD(extack, "Unknown MPLS action");
278 err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
284 p = kzalloc(sizeof(*p), GFP_KERNEL);
290 p->tcfm_action = parm->m_action;
291 p->tcfm_label = tb[TCA_MPLS_LABEL] ? nla_get_u32(tb[TCA_MPLS_LABEL]) :
292 ACT_MPLS_LABEL_NOT_SET;
293 p->tcfm_tc = tb[TCA_MPLS_TC] ? nla_get_u8(tb[TCA_MPLS_TC]) :
295 p->tcfm_ttl = tb[TCA_MPLS_TTL] ? nla_get_u8(tb[TCA_MPLS_TTL]) :
297 p->tcfm_bos = tb[TCA_MPLS_BOS] ? nla_get_u8(tb[TCA_MPLS_BOS]) :
298 ACT_MPLS_BOS_NOT_SET;
299 p->tcfm_proto = tb[TCA_MPLS_PROTO] ? nla_get_be16(tb[TCA_MPLS_PROTO]) :
300 htons(ETH_P_MPLS_UC);
302 spin_lock_bh(&m->tcf_lock);
303 goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
304 p = rcu_replace_pointer(m->mpls_p, p, lockdep_is_held(&m->tcf_lock));
305 spin_unlock_bh(&m->tcf_lock);
308 tcf_chain_put_by_act(goto_ch);
315 tcf_chain_put_by_act(goto_ch);
317 tcf_idr_release(*a, bind);
321 static void tcf_mpls_cleanup(struct tc_action *a)
323 struct tcf_mpls *m = to_mpls(a);
324 struct tcf_mpls_params *p;
326 p = rcu_dereference_protected(m->mpls_p, 1);
331 static int tcf_mpls_dump(struct sk_buff *skb, struct tc_action *a,
334 unsigned char *b = skb_tail_pointer(skb);
335 struct tcf_mpls *m = to_mpls(a);
336 struct tcf_mpls_params *p;
337 struct tc_mpls opt = {
338 .index = m->tcf_index,
339 .refcnt = refcount_read(&m->tcf_refcnt) - ref,
340 .bindcnt = atomic_read(&m->tcf_bindcnt) - bind,
344 spin_lock_bh(&m->tcf_lock);
345 opt.action = m->tcf_action;
346 p = rcu_dereference_protected(m->mpls_p, lockdep_is_held(&m->tcf_lock));
347 opt.m_action = p->tcfm_action;
349 if (nla_put(skb, TCA_MPLS_PARMS, sizeof(opt), &opt))
350 goto nla_put_failure;
352 if (p->tcfm_label != ACT_MPLS_LABEL_NOT_SET &&
353 nla_put_u32(skb, TCA_MPLS_LABEL, p->tcfm_label))
354 goto nla_put_failure;
356 if (p->tcfm_tc != ACT_MPLS_TC_NOT_SET &&
357 nla_put_u8(skb, TCA_MPLS_TC, p->tcfm_tc))
358 goto nla_put_failure;
360 if (p->tcfm_ttl && nla_put_u8(skb, TCA_MPLS_TTL, p->tcfm_ttl))
361 goto nla_put_failure;
363 if (p->tcfm_bos != ACT_MPLS_BOS_NOT_SET &&
364 nla_put_u8(skb, TCA_MPLS_BOS, p->tcfm_bos))
365 goto nla_put_failure;
367 if (nla_put_be16(skb, TCA_MPLS_PROTO, p->tcfm_proto))
368 goto nla_put_failure;
370 tcf_tm_dump(&t, &m->tcf_tm);
372 if (nla_put_64bit(skb, TCA_MPLS_TM, sizeof(t), &t, TCA_MPLS_PAD))
373 goto nla_put_failure;
375 spin_unlock_bh(&m->tcf_lock);
380 spin_unlock_bh(&m->tcf_lock);
385 static int tcf_mpls_offload_act_setup(struct tc_action *act, void *entry_data,
386 u32 *index_inc, bool bind,
387 struct netlink_ext_ack *extack)
390 struct flow_action_entry *entry = entry_data;
392 switch (tcf_mpls_action(act)) {
393 case TCA_MPLS_ACT_PUSH:
394 entry->id = FLOW_ACTION_MPLS_PUSH;
395 entry->mpls_push.proto = tcf_mpls_proto(act);
396 entry->mpls_push.label = tcf_mpls_label(act);
397 entry->mpls_push.tc = tcf_mpls_tc(act);
398 entry->mpls_push.bos = tcf_mpls_bos(act);
399 entry->mpls_push.ttl = tcf_mpls_ttl(act);
401 case TCA_MPLS_ACT_POP:
402 entry->id = FLOW_ACTION_MPLS_POP;
403 entry->mpls_pop.proto = tcf_mpls_proto(act);
405 case TCA_MPLS_ACT_MODIFY:
406 entry->id = FLOW_ACTION_MPLS_MANGLE;
407 entry->mpls_mangle.label = tcf_mpls_label(act);
408 entry->mpls_mangle.tc = tcf_mpls_tc(act);
409 entry->mpls_mangle.bos = tcf_mpls_bos(act);
410 entry->mpls_mangle.ttl = tcf_mpls_ttl(act);
412 case TCA_MPLS_ACT_DEC_TTL:
413 NL_SET_ERR_MSG_MOD(extack, "Offload not supported when \"dec_ttl\" option is used");
415 case TCA_MPLS_ACT_MAC_PUSH:
416 NL_SET_ERR_MSG_MOD(extack, "Offload not supported when \"mac_push\" option is used");
419 NL_SET_ERR_MSG_MOD(extack, "Unsupported MPLS mode offload");
424 struct flow_offload_action *fl_action = entry_data;
426 switch (tcf_mpls_action(act)) {
427 case TCA_MPLS_ACT_PUSH:
428 fl_action->id = FLOW_ACTION_MPLS_PUSH;
430 case TCA_MPLS_ACT_POP:
431 fl_action->id = FLOW_ACTION_MPLS_POP;
433 case TCA_MPLS_ACT_MODIFY:
434 fl_action->id = FLOW_ACTION_MPLS_MANGLE;
444 static struct tc_action_ops act_mpls_ops = {
447 .owner = THIS_MODULE,
449 .dump = tcf_mpls_dump,
450 .init = tcf_mpls_init,
451 .cleanup = tcf_mpls_cleanup,
452 .offload_act_setup = tcf_mpls_offload_act_setup,
453 .size = sizeof(struct tcf_mpls),
456 static __net_init int mpls_init_net(struct net *net)
458 struct tc_action_net *tn = net_generic(net, act_mpls_ops.net_id);
460 return tc_action_net_init(net, tn, &act_mpls_ops);
463 static void __net_exit mpls_exit_net(struct list_head *net_list)
465 tc_action_net_exit(net_list, act_mpls_ops.net_id);
468 static struct pernet_operations mpls_net_ops = {
469 .init = mpls_init_net,
470 .exit_batch = mpls_exit_net,
471 .id = &act_mpls_ops.net_id,
472 .size = sizeof(struct tc_action_net),
475 static int __init mpls_init_module(void)
477 return tcf_register_action(&act_mpls_ops, &mpls_net_ops);
480 static void __exit mpls_cleanup_module(void)
482 tcf_unregister_action(&act_mpls_ops, &mpls_net_ops);
485 module_init(mpls_init_module);
486 module_exit(mpls_cleanup_module);
488 MODULE_SOFTDEP("post: mpls_gso");
489 MODULE_AUTHOR("Netronome Systems <oss-drivers@netronome.com>");
490 MODULE_LICENSE("GPL");
491 MODULE_DESCRIPTION("MPLS manipulation actions");