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