xref: /openbmc/linux/net/sched/act_skbedit.c (revision 8cb08174) 
1 /*
2  * Copyright (c) 2008, Intel Corporation.
3  *
4  * This program is free software; you can redistribute it and/or modify it
5  * under the terms and conditions of the GNU General Public License,
6  * version 2, as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope it will be useful, but WITHOUT
9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
11  * more details.
12  *
13  * You should have received a copy of the GNU General Public License along with
14  * this program; if not, see <http://www.gnu.org/licenses/>.
15  *
16  * Author: Alexander Duyck <alexander.h.duyck@intel.com>
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 #include <net/pkt_cls.h>
30 
31 #include <linux/tc_act/tc_skbedit.h>
32 #include <net/tc_act/tc_skbedit.h>
33 
34 static unsigned int skbedit_net_id;
35 static struct tc_action_ops act_skbedit_ops;
36 
37 static int tcf_skbedit_act(struct sk_buff *skb, const struct tc_action *a,
38 			   struct tcf_result *res)
39 {
40 	struct tcf_skbedit *d = to_skbedit(a);
41 	struct tcf_skbedit_params *params;
42 	int action;
43 
44 	tcf_lastuse_update(&d->tcf_tm);
45 	bstats_cpu_update(this_cpu_ptr(d->common.cpu_bstats), skb);
46 
47 	params = rcu_dereference_bh(d->params);
48 	action = READ_ONCE(d->tcf_action);
49 
50 	if (params->flags & SKBEDIT_F_PRIORITY)
51 		skb->priority = params->priority;
52 	if (params->flags & SKBEDIT_F_INHERITDSFIELD) {
53 		int wlen = skb_network_offset(skb);
54 
55 		switch (tc_skb_protocol(skb)) {
56 		case htons(ETH_P_IP):
57 			wlen += sizeof(struct iphdr);
58 			if (!pskb_may_pull(skb, wlen))
59 				goto err;
60 			skb->priority = ipv4_get_dsfield(ip_hdr(skb)) >> 2;
61 			break;
62 
63 		case htons(ETH_P_IPV6):
64 			wlen += sizeof(struct ipv6hdr);
65 			if (!pskb_may_pull(skb, wlen))
66 				goto err;
67 			skb->priority = ipv6_get_dsfield(ipv6_hdr(skb)) >> 2;
68 			break;
69 		}
70 	}
71 	if (params->flags & SKBEDIT_F_QUEUE_MAPPING &&
72 	    skb->dev->real_num_tx_queues > params->queue_mapping)
73 		skb_set_queue_mapping(skb, params->queue_mapping);
74 	if (params->flags & SKBEDIT_F_MARK) {
75 		skb->mark &= ~params->mask;
76 		skb->mark |= params->mark & params->mask;
77 	}
78 	if (params->flags & SKBEDIT_F_PTYPE)
79 		skb->pkt_type = params->ptype;
80 	return action;
81 
82 err:
83 	qstats_drop_inc(this_cpu_ptr(d->common.cpu_qstats));
84 	return TC_ACT_SHOT;
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 
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,
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 
114 	if (nla == NULL)
115 		return -EINVAL;
116 
117 	err = nla_parse_nested_deprecated(tb, TCA_SKBEDIT_MAX, nla,
118 					  skbedit_policy, NULL);
119 	if (err < 0)
120 		return err;
121 
122 	if (tb[TCA_SKBEDIT_PARMS] == NULL)
123 		return -EINVAL;
124 
125 	if (tb[TCA_SKBEDIT_PRIORITY] != NULL) {
126 		flags |= SKBEDIT_F_PRIORITY;
127 		priority = nla_data(tb[TCA_SKBEDIT_PRIORITY]);
128 	}
129 
130 	if (tb[TCA_SKBEDIT_QUEUE_MAPPING] != NULL) {
131 		flags |= SKBEDIT_F_QUEUE_MAPPING;
132 		queue_mapping = nla_data(tb[TCA_SKBEDIT_QUEUE_MAPPING]);
133 	}
134 
135 	if (tb[TCA_SKBEDIT_PTYPE] != NULL) {
136 		ptype = nla_data(tb[TCA_SKBEDIT_PTYPE]);
137 		if (!skb_pkt_type_ok(*ptype))
138 			return -EINVAL;
139 		flags |= SKBEDIT_F_PTYPE;
140 	}
141 
142 	if (tb[TCA_SKBEDIT_MARK] != NULL) {
143 		flags |= SKBEDIT_F_MARK;
144 		mark = nla_data(tb[TCA_SKBEDIT_MARK]);
145 	}
146 
147 	if (tb[TCA_SKBEDIT_MASK] != NULL) {
148 		flags |= SKBEDIT_F_MASK;
149 		mask = nla_data(tb[TCA_SKBEDIT_MASK]);
150 	}
151 
152 	if (tb[TCA_SKBEDIT_FLAGS] != NULL) {
153 		u64 *pure_flags = nla_data(tb[TCA_SKBEDIT_FLAGS]);
154 
155 		if (*pure_flags & SKBEDIT_F_INHERITDSFIELD)
156 			flags |= SKBEDIT_F_INHERITDSFIELD;
157 	}
158 
159 	parm = nla_data(tb[TCA_SKBEDIT_PARMS]);
160 
161 	err = tcf_idr_check_alloc(tn, &parm->index, a, bind);
162 	if (err < 0)
163 		return err;
164 	exists = err;
165 	if (exists && bind)
166 		return 0;
167 
168 	if (!flags) {
169 		if (exists)
170 			tcf_idr_release(*a, bind);
171 		else
172 			tcf_idr_cleanup(tn, parm->index);
173 		return -EINVAL;
174 	}
175 
176 	if (!exists) {
177 		ret = tcf_idr_create(tn, parm->index, est, a,
178 				     &act_skbedit_ops, bind, true);
179 		if (ret) {
180 			tcf_idr_cleanup(tn, parm->index);
181 			return ret;
182 		}
183 
184 		d = to_skbedit(*a);
185 		ret = ACT_P_CREATED;
186 	} else {
187 		d = to_skbedit(*a);
188 		if (!ovr) {
189 			tcf_idr_release(*a, bind);
190 			return -EEXIST;
191 		}
192 	}
193 	err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
194 	if (err < 0)
195 		goto release_idr;
196 
197 	params_new = kzalloc(sizeof(*params_new), GFP_KERNEL);
198 	if (unlikely(!params_new)) {
199 		err = -ENOMEM;
200 		goto put_chain;
201 	}
202 
203 	params_new->flags = flags;
204 	if (flags & SKBEDIT_F_PRIORITY)
205 		params_new->priority = *priority;
206 	if (flags & SKBEDIT_F_QUEUE_MAPPING)
207 		params_new->queue_mapping = *queue_mapping;
208 	if (flags & SKBEDIT_F_MARK)
209 		params_new->mark = *mark;
210 	if (flags & SKBEDIT_F_PTYPE)
211 		params_new->ptype = *ptype;
212 	/* default behaviour is to use all the bits */
213 	params_new->mask = 0xffffffff;
214 	if (flags & SKBEDIT_F_MASK)
215 		params_new->mask = *mask;
216 
217 	spin_lock_bh(&d->tcf_lock);
218 	goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
219 	rcu_swap_protected(d->params, params_new,
220 			   lockdep_is_held(&d->tcf_lock));
221 	spin_unlock_bh(&d->tcf_lock);
222 	if (params_new)
223 		kfree_rcu(params_new, rcu);
224 	if (goto_ch)
225 		tcf_chain_put_by_act(goto_ch);
226 
227 	if (ret == ACT_P_CREATED)
228 		tcf_idr_insert(tn, *a);
229 	return ret;
230 put_chain:
231 	if (goto_ch)
232 		tcf_chain_put_by_act(goto_ch);
233 release_idr:
234 	tcf_idr_release(*a, bind);
235 	return err;
236 }
237 
238 static int tcf_skbedit_dump(struct sk_buff *skb, struct tc_action *a,
239 			    int bind, int ref)
240 {
241 	unsigned char *b = skb_tail_pointer(skb);
242 	struct tcf_skbedit *d = to_skbedit(a);
243 	struct tcf_skbedit_params *params;
244 	struct tc_skbedit opt = {
245 		.index   = d->tcf_index,
246 		.refcnt  = refcount_read(&d->tcf_refcnt) - ref,
247 		.bindcnt = atomic_read(&d->tcf_bindcnt) - bind,
248 	};
249 	u64 pure_flags = 0;
250 	struct tcf_t t;
251 
252 	spin_lock_bh(&d->tcf_lock);
253 	params = rcu_dereference_protected(d->params,
254 					   lockdep_is_held(&d->tcf_lock));
255 	opt.action = d->tcf_action;
256 
257 	if (nla_put(skb, TCA_SKBEDIT_PARMS, sizeof(opt), &opt))
258 		goto nla_put_failure;
259 	if ((params->flags & SKBEDIT_F_PRIORITY) &&
260 	    nla_put_u32(skb, TCA_SKBEDIT_PRIORITY, params->priority))
261 		goto nla_put_failure;
262 	if ((params->flags & SKBEDIT_F_QUEUE_MAPPING) &&
263 	    nla_put_u16(skb, TCA_SKBEDIT_QUEUE_MAPPING, params->queue_mapping))
264 		goto nla_put_failure;
265 	if ((params->flags & SKBEDIT_F_MARK) &&
266 	    nla_put_u32(skb, TCA_SKBEDIT_MARK, params->mark))
267 		goto nla_put_failure;
268 	if ((params->flags & SKBEDIT_F_PTYPE) &&
269 	    nla_put_u16(skb, TCA_SKBEDIT_PTYPE, params->ptype))
270 		goto nla_put_failure;
271 	if ((params->flags & SKBEDIT_F_MASK) &&
272 	    nla_put_u32(skb, TCA_SKBEDIT_MASK, params->mask))
273 		goto nla_put_failure;
274 	if (params->flags & SKBEDIT_F_INHERITDSFIELD)
275 		pure_flags |= SKBEDIT_F_INHERITDSFIELD;
276 	if (pure_flags != 0 &&
277 	    nla_put(skb, TCA_SKBEDIT_FLAGS, sizeof(pure_flags), &pure_flags))
278 		goto nla_put_failure;
279 
280 	tcf_tm_dump(&t, &d->tcf_tm);
281 	if (nla_put_64bit(skb, TCA_SKBEDIT_TM, sizeof(t), &t, TCA_SKBEDIT_PAD))
282 		goto nla_put_failure;
283 	spin_unlock_bh(&d->tcf_lock);
284 
285 	return skb->len;
286 
287 nla_put_failure:
288 	spin_unlock_bh(&d->tcf_lock);
289 	nlmsg_trim(skb, b);
290 	return -1;
291 }
292 
293 static void tcf_skbedit_cleanup(struct tc_action *a)
294 {
295 	struct tcf_skbedit *d = to_skbedit(a);
296 	struct tcf_skbedit_params *params;
297 
298 	params = rcu_dereference_protected(d->params, 1);
299 	if (params)
300 		kfree_rcu(params, rcu);
301 }
302 
303 static int tcf_skbedit_walker(struct net *net, struct sk_buff *skb,
304 			      struct netlink_callback *cb, int type,
305 			      const struct tc_action_ops *ops,
306 			      struct netlink_ext_ack *extack)
307 {
308 	struct tc_action_net *tn = net_generic(net, skbedit_net_id);
309 
310 	return tcf_generic_walker(tn, skb, cb, type, ops, extack);
311 }
312 
313 static int tcf_skbedit_search(struct net *net, struct tc_action **a, u32 index)
314 {
315 	struct tc_action_net *tn = net_generic(net, skbedit_net_id);
316 
317 	return tcf_idr_search(tn, a, index);
318 }
319 
320 static struct tc_action_ops act_skbedit_ops = {
321 	.kind		=	"skbedit",
322 	.id		=	TCA_ID_SKBEDIT,
323 	.owner		=	THIS_MODULE,
324 	.act		=	tcf_skbedit_act,
325 	.dump		=	tcf_skbedit_dump,
326 	.init		=	tcf_skbedit_init,
327 	.cleanup	=	tcf_skbedit_cleanup,
328 	.walk		=	tcf_skbedit_walker,
329 	.lookup		=	tcf_skbedit_search,
330 	.size		=	sizeof(struct tcf_skbedit),
331 };
332 
333 static __net_init int skbedit_init_net(struct net *net)
334 {
335 	struct tc_action_net *tn = net_generic(net, skbedit_net_id);
336 
337 	return tc_action_net_init(tn, &act_skbedit_ops);
338 }
339 
340 static void __net_exit skbedit_exit_net(struct list_head *net_list)
341 {
342 	tc_action_net_exit(net_list, skbedit_net_id);
343 }
344 
345 static struct pernet_operations skbedit_net_ops = {
346 	.init = skbedit_init_net,
347 	.exit_batch = skbedit_exit_net,
348 	.id   = &skbedit_net_id,
349 	.size = sizeof(struct tc_action_net),
350 };
351 
352 MODULE_AUTHOR("Alexander Duyck, <alexander.h.duyck@intel.com>");
353 MODULE_DESCRIPTION("SKB Editing");
354 MODULE_LICENSE("GPL");
355 
356 static int __init skbedit_init_module(void)
357 {
358 	return tcf_register_action(&act_skbedit_ops, &skbedit_net_ops);
359 }
360 
361 static void __exit skbedit_cleanup_module(void)
362 {
363 	tcf_unregister_action(&act_skbedit_ops, &skbedit_net_ops);
364 }
365 
366 module_init(skbedit_init_module);
367 module_exit(skbedit_cleanup_module);
368