1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * Copyright (c) 2008, Intel Corporation.
4 *
5 * Author: Alexander Duyck <alexander.h.duyck@intel.com>
6 */
7
8 #include <linux/module.h>
9 #include <linux/init.h>
10 #include <linux/kernel.h>
11 #include <linux/skbuff.h>
12 #include <linux/rtnetlink.h>
13 #include <net/netlink.h>
14 #include <net/pkt_sched.h>
15 #include <net/ip.h>
16 #include <net/ipv6.h>
17 #include <net/dsfield.h>
18 #include <net/pkt_cls.h>
19
20 #include <linux/tc_act/tc_skbedit.h>
21 #include <net/tc_act/tc_skbedit.h>
22
23 static unsigned int skbedit_net_id;
24 static struct tc_action_ops act_skbedit_ops;
25
tcf_skbedit_act(struct sk_buff * skb,const struct tc_action * a,struct tcf_result * res)26 static int tcf_skbedit_act(struct sk_buff *skb, const struct tc_action *a,
27 struct tcf_result *res)
28 {
29 struct tcf_skbedit *d = to_skbedit(a);
30 struct tcf_skbedit_params *params;
31 int action;
32
33 tcf_lastuse_update(&d->tcf_tm);
34 bstats_cpu_update(this_cpu_ptr(d->common.cpu_bstats), skb);
35
36 params = rcu_dereference_bh(d->params);
37 action = READ_ONCE(d->tcf_action);
38
39 if (params->flags & SKBEDIT_F_PRIORITY)
40 skb->priority = params->priority;
41 if (params->flags & SKBEDIT_F_INHERITDSFIELD) {
42 int wlen = skb_network_offset(skb);
43
44 switch (skb_protocol(skb, true)) {
45 case htons(ETH_P_IP):
46 wlen += sizeof(struct iphdr);
47 if (!pskb_may_pull(skb, wlen))
48 goto err;
49 skb->priority = ipv4_get_dsfield(ip_hdr(skb)) >> 2;
50 break;
51
52 case htons(ETH_P_IPV6):
53 wlen += sizeof(struct ipv6hdr);
54 if (!pskb_may_pull(skb, wlen))
55 goto err;
56 skb->priority = ipv6_get_dsfield(ipv6_hdr(skb)) >> 2;
57 break;
58 }
59 }
60 if (params->flags & SKBEDIT_F_QUEUE_MAPPING &&
61 skb->dev->real_num_tx_queues > params->queue_mapping)
62 skb_set_queue_mapping(skb, params->queue_mapping);
63 if (params->flags & SKBEDIT_F_MARK) {
64 skb->mark &= ~params->mask;
65 skb->mark |= params->mark & params->mask;
66 }
67 if (params->flags & SKBEDIT_F_PTYPE)
68 skb->pkt_type = params->ptype;
69 return action;
70
71 err:
72 qstats_drop_inc(this_cpu_ptr(d->common.cpu_qstats));
73 return TC_ACT_SHOT;
74 }
75
tcf_skbedit_stats_update(struct tc_action * a,u64 bytes,u64 packets,u64 drops,u64 lastuse,bool hw)76 static void tcf_skbedit_stats_update(struct tc_action *a, u64 bytes,
77 u64 packets, u64 drops,
78 u64 lastuse, bool hw)
79 {
80 struct tcf_skbedit *d = to_skbedit(a);
81 struct tcf_t *tm = &d->tcf_tm;
82
83 tcf_action_update_stats(a, bytes, packets, drops, hw);
84 tm->lastuse = max_t(u64, tm->lastuse, lastuse);
85 }
86
87 static const struct nla_policy skbedit_policy[TCA_SKBEDIT_MAX + 1] = {
88 [TCA_SKBEDIT_PARMS] = { .len = sizeof(struct tc_skbedit) },
89 [TCA_SKBEDIT_PRIORITY] = { .len = sizeof(u32) },
90 [TCA_SKBEDIT_QUEUE_MAPPING] = { .len = sizeof(u16) },
91 [TCA_SKBEDIT_MARK] = { .len = sizeof(u32) },
92 [TCA_SKBEDIT_PTYPE] = { .len = sizeof(u16) },
93 [TCA_SKBEDIT_MASK] = { .len = sizeof(u32) },
94 [TCA_SKBEDIT_FLAGS] = { .len = sizeof(u64) },
95 };
96
tcf_skbedit_init(struct net * net,struct nlattr * nla,struct nlattr * est,struct tc_action ** a,int ovr,int bind,bool rtnl_held,struct tcf_proto * tp,u32 act_flags,struct netlink_ext_ack * extack)97 static int tcf_skbedit_init(struct net *net, struct nlattr *nla,
98 struct nlattr *est, struct tc_action **a,
99 int ovr, int bind, bool rtnl_held,
100 struct tcf_proto *tp, u32 act_flags,
101 struct netlink_ext_ack *extack)
102 {
103 struct tc_action_net *tn = net_generic(net, skbedit_net_id);
104 struct tcf_skbedit_params *params_new;
105 struct nlattr *tb[TCA_SKBEDIT_MAX + 1];
106 struct tcf_chain *goto_ch = NULL;
107 struct tc_skbedit *parm;
108 struct tcf_skbedit *d;
109 u32 flags = 0, *priority = NULL, *mark = NULL, *mask = NULL;
110 u16 *queue_mapping = NULL, *ptype = NULL;
111 bool exists = false;
112 int ret = 0, err;
113 u32 index;
114
115 if (nla == NULL)
116 return -EINVAL;
117
118 err = nla_parse_nested_deprecated(tb, TCA_SKBEDIT_MAX, nla,
119 skbedit_policy, NULL);
120 if (err < 0)
121 return err;
122
123 if (tb[TCA_SKBEDIT_PARMS] == NULL)
124 return -EINVAL;
125
126 if (tb[TCA_SKBEDIT_PRIORITY] != NULL) {
127 flags |= SKBEDIT_F_PRIORITY;
128 priority = nla_data(tb[TCA_SKBEDIT_PRIORITY]);
129 }
130
131 if (tb[TCA_SKBEDIT_QUEUE_MAPPING] != NULL) {
132 flags |= SKBEDIT_F_QUEUE_MAPPING;
133 queue_mapping = nla_data(tb[TCA_SKBEDIT_QUEUE_MAPPING]);
134 }
135
136 if (tb[TCA_SKBEDIT_PTYPE] != NULL) {
137 ptype = nla_data(tb[TCA_SKBEDIT_PTYPE]);
138 if (!skb_pkt_type_ok(*ptype))
139 return -EINVAL;
140 flags |= SKBEDIT_F_PTYPE;
141 }
142
143 if (tb[TCA_SKBEDIT_MARK] != NULL) {
144 flags |= SKBEDIT_F_MARK;
145 mark = nla_data(tb[TCA_SKBEDIT_MARK]);
146 }
147
148 if (tb[TCA_SKBEDIT_MASK] != NULL) {
149 flags |= SKBEDIT_F_MASK;
150 mask = nla_data(tb[TCA_SKBEDIT_MASK]);
151 }
152
153 if (tb[TCA_SKBEDIT_FLAGS] != NULL) {
154 u64 *pure_flags = nla_data(tb[TCA_SKBEDIT_FLAGS]);
155
156 if (*pure_flags & SKBEDIT_F_INHERITDSFIELD)
157 flags |= SKBEDIT_F_INHERITDSFIELD;
158 }
159
160 parm = nla_data(tb[TCA_SKBEDIT_PARMS]);
161 index = parm->index;
162 err = tcf_idr_check_alloc(tn, &index, a, bind);
163 if (err < 0)
164 return err;
165 exists = err;
166 if (exists && bind)
167 return 0;
168
169 if (!flags) {
170 if (exists)
171 tcf_idr_release(*a, bind);
172 else
173 tcf_idr_cleanup(tn, index);
174 return -EINVAL;
175 }
176
177 if (!exists) {
178 ret = tcf_idr_create(tn, index, est, a,
179 &act_skbedit_ops, bind, true, 0);
180 if (ret) {
181 tcf_idr_cleanup(tn, index);
182 return ret;
183 }
184
185 d = to_skbedit(*a);
186 ret = ACT_P_CREATED;
187 } else {
188 d = to_skbedit(*a);
189 if (!ovr) {
190 tcf_idr_release(*a, bind);
191 return -EEXIST;
192 }
193 }
194 err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
195 if (err < 0)
196 goto release_idr;
197
198 params_new = kzalloc(sizeof(*params_new), GFP_KERNEL);
199 if (unlikely(!params_new)) {
200 err = -ENOMEM;
201 goto put_chain;
202 }
203
204 params_new->flags = flags;
205 if (flags & SKBEDIT_F_PRIORITY)
206 params_new->priority = *priority;
207 if (flags & SKBEDIT_F_QUEUE_MAPPING)
208 params_new->queue_mapping = *queue_mapping;
209 if (flags & SKBEDIT_F_MARK)
210 params_new->mark = *mark;
211 if (flags & SKBEDIT_F_PTYPE)
212 params_new->ptype = *ptype;
213 /* default behaviour is to use all the bits */
214 params_new->mask = 0xffffffff;
215 if (flags & SKBEDIT_F_MASK)
216 params_new->mask = *mask;
217
218 spin_lock_bh(&d->tcf_lock);
219 goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
220 params_new = rcu_replace_pointer(d->params, params_new,
221 lockdep_is_held(&d->tcf_lock));
222 spin_unlock_bh(&d->tcf_lock);
223 if (params_new)
224 kfree_rcu(params_new, rcu);
225 if (goto_ch)
226 tcf_chain_put_by_act(goto_ch);
227
228 return ret;
229 put_chain:
230 if (goto_ch)
231 tcf_chain_put_by_act(goto_ch);
232 release_idr:
233 tcf_idr_release(*a, bind);
234 return err;
235 }
236
tcf_skbedit_dump(struct sk_buff * skb,struct tc_action * a,int bind,int ref)237 static int tcf_skbedit_dump(struct sk_buff *skb, struct tc_action *a,
238 int bind, int ref)
239 {
240 unsigned char *b = skb_tail_pointer(skb);
241 struct tcf_skbedit *d = to_skbedit(a);
242 struct tcf_skbedit_params *params;
243 struct tc_skbedit opt = {
244 .index = d->tcf_index,
245 .refcnt = refcount_read(&d->tcf_refcnt) - ref,
246 .bindcnt = atomic_read(&d->tcf_bindcnt) - bind,
247 };
248 u64 pure_flags = 0;
249 struct tcf_t t;
250
251 spin_lock_bh(&d->tcf_lock);
252 params = rcu_dereference_protected(d->params,
253 lockdep_is_held(&d->tcf_lock));
254 opt.action = d->tcf_action;
255
256 if (nla_put(skb, TCA_SKBEDIT_PARMS, sizeof(opt), &opt))
257 goto nla_put_failure;
258 if ((params->flags & SKBEDIT_F_PRIORITY) &&
259 nla_put_u32(skb, TCA_SKBEDIT_PRIORITY, params->priority))
260 goto nla_put_failure;
261 if ((params->flags & SKBEDIT_F_QUEUE_MAPPING) &&
262 nla_put_u16(skb, TCA_SKBEDIT_QUEUE_MAPPING, params->queue_mapping))
263 goto nla_put_failure;
264 if ((params->flags & SKBEDIT_F_MARK) &&
265 nla_put_u32(skb, TCA_SKBEDIT_MARK, params->mark))
266 goto nla_put_failure;
267 if ((params->flags & SKBEDIT_F_PTYPE) &&
268 nla_put_u16(skb, TCA_SKBEDIT_PTYPE, params->ptype))
269 goto nla_put_failure;
270 if ((params->flags & SKBEDIT_F_MASK) &&
271 nla_put_u32(skb, TCA_SKBEDIT_MASK, params->mask))
272 goto nla_put_failure;
273 if (params->flags & SKBEDIT_F_INHERITDSFIELD)
274 pure_flags |= SKBEDIT_F_INHERITDSFIELD;
275 if (pure_flags != 0 &&
276 nla_put(skb, TCA_SKBEDIT_FLAGS, sizeof(pure_flags), &pure_flags))
277 goto nla_put_failure;
278
279 tcf_tm_dump(&t, &d->tcf_tm);
280 if (nla_put_64bit(skb, TCA_SKBEDIT_TM, sizeof(t), &t, TCA_SKBEDIT_PAD))
281 goto nla_put_failure;
282 spin_unlock_bh(&d->tcf_lock);
283
284 return skb->len;
285
286 nla_put_failure:
287 spin_unlock_bh(&d->tcf_lock);
288 nlmsg_trim(skb, b);
289 return -1;
290 }
291
tcf_skbedit_cleanup(struct tc_action * a)292 static void tcf_skbedit_cleanup(struct tc_action *a)
293 {
294 struct tcf_skbedit *d = to_skbedit(a);
295 struct tcf_skbedit_params *params;
296
297 params = rcu_dereference_protected(d->params, 1);
298 if (params)
299 kfree_rcu(params, rcu);
300 }
301
tcf_skbedit_walker(struct net * net,struct sk_buff * skb,struct netlink_callback * cb,int type,const struct tc_action_ops * ops,struct netlink_ext_ack * extack)302 static int tcf_skbedit_walker(struct net *net, struct sk_buff *skb,
303 struct netlink_callback *cb, int type,
304 const struct tc_action_ops *ops,
305 struct netlink_ext_ack *extack)
306 {
307 struct tc_action_net *tn = net_generic(net, skbedit_net_id);
308
309 return tcf_generic_walker(tn, skb, cb, type, ops, extack);
310 }
311
tcf_skbedit_search(struct net * net,struct tc_action ** a,u32 index)312 static int tcf_skbedit_search(struct net *net, struct tc_action **a, u32 index)
313 {
314 struct tc_action_net *tn = net_generic(net, skbedit_net_id);
315
316 return tcf_idr_search(tn, a, index);
317 }
318
tcf_skbedit_get_fill_size(const struct tc_action * act)319 static size_t tcf_skbedit_get_fill_size(const struct tc_action *act)
320 {
321 return nla_total_size(sizeof(struct tc_skbedit))
322 + nla_total_size(sizeof(u32)) /* TCA_SKBEDIT_PRIORITY */
323 + nla_total_size(sizeof(u16)) /* TCA_SKBEDIT_QUEUE_MAPPING */
324 + nla_total_size(sizeof(u32)) /* TCA_SKBEDIT_MARK */
325 + nla_total_size(sizeof(u16)) /* TCA_SKBEDIT_PTYPE */
326 + nla_total_size(sizeof(u32)) /* TCA_SKBEDIT_MASK */
327 + nla_total_size_64bit(sizeof(u64)); /* TCA_SKBEDIT_FLAGS */
328 }
329
330 static struct tc_action_ops act_skbedit_ops = {
331 .kind = "skbedit",
332 .id = TCA_ID_SKBEDIT,
333 .owner = THIS_MODULE,
334 .act = tcf_skbedit_act,
335 .stats_update = tcf_skbedit_stats_update,
336 .dump = tcf_skbedit_dump,
337 .init = tcf_skbedit_init,
338 .cleanup = tcf_skbedit_cleanup,
339 .walk = tcf_skbedit_walker,
340 .get_fill_size = tcf_skbedit_get_fill_size,
341 .lookup = tcf_skbedit_search,
342 .size = sizeof(struct tcf_skbedit),
343 };
344
skbedit_init_net(struct net * net)345 static __net_init int skbedit_init_net(struct net *net)
346 {
347 struct tc_action_net *tn = net_generic(net, skbedit_net_id);
348
349 return tc_action_net_init(net, tn, &act_skbedit_ops);
350 }
351
skbedit_exit_net(struct list_head * net_list)352 static void __net_exit skbedit_exit_net(struct list_head *net_list)
353 {
354 tc_action_net_exit(net_list, skbedit_net_id);
355 }
356
357 static struct pernet_operations skbedit_net_ops = {
358 .init = skbedit_init_net,
359 .exit_batch = skbedit_exit_net,
360 .id = &skbedit_net_id,
361 .size = sizeof(struct tc_action_net),
362 };
363
364 MODULE_AUTHOR("Alexander Duyck, <alexander.h.duyck@intel.com>");
365 MODULE_DESCRIPTION("SKB Editing");
366 MODULE_LICENSE("GPL");
367
skbedit_init_module(void)368 static int __init skbedit_init_module(void)
369 {
370 return tcf_register_action(&act_skbedit_ops, &skbedit_net_ops);
371 }
372
skbedit_cleanup_module(void)373 static void __exit skbedit_cleanup_module(void)
374 {
375 tcf_unregister_action(&act_skbedit_ops, &skbedit_net_ops);
376 }
377
378 module_init(skbedit_init_module);
379 module_exit(skbedit_cleanup_module);
380