1 /*
2  * Codel - The Controlled-Delay Active Queue Management algorithm
3  *
4  *  Copyright (C) 2011-2012 Kathleen Nichols <nichols@pollere.com>
5  *  Copyright (C) 2011-2012 Van Jacobson <van@pollere.net>
6  *
7  *  Implemented on linux by :
8  *  Copyright (C) 2012 Michael D. Taht <dave.taht@bufferbloat.net>
9  *  Copyright (C) 2012,2015 Eric Dumazet <edumazet@google.com>
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions, and the following disclaimer,
16  *    without modification.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. The names of the authors may not be used to endorse or promote products
21  *    derived from this software without specific prior written permission.
22  *
23  * Alternatively, provided that this notice is retained in full, this
24  * software may be distributed under the terms of the GNU General
25  * Public License ("GPL") version 2, in which case the provisions of the
26  * GPL apply INSTEAD OF those given above.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
29  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
30  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
31  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
32  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
33  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
34  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
35  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
36  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
37  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
38  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
39  * DAMAGE.
40  *
41  */
42 
43 #include <linux/module.h>
44 #include <linux/slab.h>
45 #include <linux/types.h>
46 #include <linux/kernel.h>
47 #include <linux/errno.h>
48 #include <linux/skbuff.h>
49 #include <linux/prefetch.h>
50 #include <net/pkt_sched.h>
51 #include <net/codel.h>
52 #include <net/codel_impl.h>
53 #include <net/codel_qdisc.h>
54 
55 
56 #define DEFAULT_CODEL_LIMIT 1000
57 
58 struct codel_sched_data {
59 	struct codel_params	params;
60 	struct codel_vars	vars;
61 	struct codel_stats	stats;
62 	u32			drop_overlimit;
63 };
64 
65 /* This is the specific function called from codel_dequeue()
66  * to dequeue a packet from queue. Note: backlog is handled in
67  * codel, we dont need to reduce it here.
68  */
dequeue_func(struct codel_vars * vars,void * ctx)69 static struct sk_buff *dequeue_func(struct codel_vars *vars, void *ctx)
70 {
71 	struct Qdisc *sch = ctx;
72 	struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);
73 
74 	if (skb) {
75 		sch->qstats.backlog -= qdisc_pkt_len(skb);
76 		prefetch(&skb->end); /* we'll need skb_shinfo() */
77 	}
78 	return skb;
79 }
80 
drop_func(struct sk_buff * skb,void * ctx)81 static void drop_func(struct sk_buff *skb, void *ctx)
82 {
83 	struct Qdisc *sch = ctx;
84 
85 	kfree_skb(skb);
86 	qdisc_qstats_drop(sch);
87 }
88 
codel_qdisc_dequeue(struct Qdisc * sch)89 static struct sk_buff *codel_qdisc_dequeue(struct Qdisc *sch)
90 {
91 	struct codel_sched_data *q = qdisc_priv(sch);
92 	struct sk_buff *skb;
93 
94 	skb = codel_dequeue(sch, &sch->qstats.backlog, &q->params, &q->vars,
95 			    &q->stats, qdisc_pkt_len, codel_get_enqueue_time,
96 			    drop_func, dequeue_func);
97 
98 	/* We cant call qdisc_tree_reduce_backlog() if our qlen is 0,
99 	 * or HTB crashes. Defer it for next round.
100 	 */
101 	if (q->stats.drop_count && sch->q.qlen) {
102 		qdisc_tree_reduce_backlog(sch, q->stats.drop_count, q->stats.drop_len);
103 		q->stats.drop_count = 0;
104 		q->stats.drop_len = 0;
105 	}
106 	if (skb)
107 		qdisc_bstats_update(sch, skb);
108 	return skb;
109 }
110 
codel_qdisc_enqueue(struct sk_buff * skb,struct Qdisc * sch,struct sk_buff ** to_free)111 static int codel_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch,
112 			       struct sk_buff **to_free)
113 {
114 	struct codel_sched_data *q;
115 
116 	if (likely(qdisc_qlen(sch) < sch->limit)) {
117 		codel_set_enqueue_time(skb);
118 		return qdisc_enqueue_tail(skb, sch);
119 	}
120 	q = qdisc_priv(sch);
121 	q->drop_overlimit++;
122 	return qdisc_drop(skb, sch, to_free);
123 }
124 
125 static const struct nla_policy codel_policy[TCA_CODEL_MAX + 1] = {
126 	[TCA_CODEL_TARGET]	= { .type = NLA_U32 },
127 	[TCA_CODEL_LIMIT]	= { .type = NLA_U32 },
128 	[TCA_CODEL_INTERVAL]	= { .type = NLA_U32 },
129 	[TCA_CODEL_ECN]		= { .type = NLA_U32 },
130 	[TCA_CODEL_CE_THRESHOLD]= { .type = NLA_U32 },
131 };
132 
codel_change(struct Qdisc * sch,struct nlattr * opt,struct netlink_ext_ack * extack)133 static int codel_change(struct Qdisc *sch, struct nlattr *opt,
134 			struct netlink_ext_ack *extack)
135 {
136 	struct codel_sched_data *q = qdisc_priv(sch);
137 	struct nlattr *tb[TCA_CODEL_MAX + 1];
138 	unsigned int qlen, dropped = 0;
139 	int err;
140 
141 	if (!opt)
142 		return -EINVAL;
143 
144 	err = nla_parse_nested_deprecated(tb, TCA_CODEL_MAX, opt,
145 					  codel_policy, NULL);
146 	if (err < 0)
147 		return err;
148 
149 	sch_tree_lock(sch);
150 
151 	if (tb[TCA_CODEL_TARGET]) {
152 		u32 target = nla_get_u32(tb[TCA_CODEL_TARGET]);
153 
154 		q->params.target = ((u64)target * NSEC_PER_USEC) >> CODEL_SHIFT;
155 	}
156 
157 	if (tb[TCA_CODEL_CE_THRESHOLD]) {
158 		u64 val = nla_get_u32(tb[TCA_CODEL_CE_THRESHOLD]);
159 
160 		q->params.ce_threshold = (val * NSEC_PER_USEC) >> CODEL_SHIFT;
161 	}
162 
163 	if (tb[TCA_CODEL_INTERVAL]) {
164 		u32 interval = nla_get_u32(tb[TCA_CODEL_INTERVAL]);
165 
166 		q->params.interval = ((u64)interval * NSEC_PER_USEC) >> CODEL_SHIFT;
167 	}
168 
169 	if (tb[TCA_CODEL_LIMIT])
170 		sch->limit = nla_get_u32(tb[TCA_CODEL_LIMIT]);
171 
172 	if (tb[TCA_CODEL_ECN])
173 		q->params.ecn = !!nla_get_u32(tb[TCA_CODEL_ECN]);
174 
175 	qlen = sch->q.qlen;
176 	while (sch->q.qlen > sch->limit) {
177 		struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);
178 
179 		dropped += qdisc_pkt_len(skb);
180 		qdisc_qstats_backlog_dec(sch, skb);
181 		rtnl_qdisc_drop(skb, sch);
182 	}
183 	qdisc_tree_reduce_backlog(sch, qlen - sch->q.qlen, dropped);
184 
185 	sch_tree_unlock(sch);
186 	return 0;
187 }
188 
codel_init(struct Qdisc * sch,struct nlattr * opt,struct netlink_ext_ack * extack)189 static int codel_init(struct Qdisc *sch, struct nlattr *opt,
190 		      struct netlink_ext_ack *extack)
191 {
192 	struct codel_sched_data *q = qdisc_priv(sch);
193 
194 	sch->limit = DEFAULT_CODEL_LIMIT;
195 
196 	codel_params_init(&q->params);
197 	codel_vars_init(&q->vars);
198 	codel_stats_init(&q->stats);
199 	q->params.mtu = psched_mtu(qdisc_dev(sch));
200 
201 	if (opt) {
202 		int err = codel_change(sch, opt, extack);
203 
204 		if (err)
205 			return err;
206 	}
207 
208 	if (sch->limit >= 1)
209 		sch->flags |= TCQ_F_CAN_BYPASS;
210 	else
211 		sch->flags &= ~TCQ_F_CAN_BYPASS;
212 
213 	return 0;
214 }
215 
codel_dump(struct Qdisc * sch,struct sk_buff * skb)216 static int codel_dump(struct Qdisc *sch, struct sk_buff *skb)
217 {
218 	struct codel_sched_data *q = qdisc_priv(sch);
219 	struct nlattr *opts;
220 
221 	opts = nla_nest_start_noflag(skb, TCA_OPTIONS);
222 	if (opts == NULL)
223 		goto nla_put_failure;
224 
225 	if (nla_put_u32(skb, TCA_CODEL_TARGET,
226 			codel_time_to_us(q->params.target)) ||
227 	    nla_put_u32(skb, TCA_CODEL_LIMIT,
228 			sch->limit) ||
229 	    nla_put_u32(skb, TCA_CODEL_INTERVAL,
230 			codel_time_to_us(q->params.interval)) ||
231 	    nla_put_u32(skb, TCA_CODEL_ECN,
232 			q->params.ecn))
233 		goto nla_put_failure;
234 	if (q->params.ce_threshold != CODEL_DISABLED_THRESHOLD &&
235 	    nla_put_u32(skb, TCA_CODEL_CE_THRESHOLD,
236 			codel_time_to_us(q->params.ce_threshold)))
237 		goto nla_put_failure;
238 	return nla_nest_end(skb, opts);
239 
240 nla_put_failure:
241 	nla_nest_cancel(skb, opts);
242 	return -1;
243 }
244 
codel_dump_stats(struct Qdisc * sch,struct gnet_dump * d)245 static int codel_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
246 {
247 	const struct codel_sched_data *q = qdisc_priv(sch);
248 	struct tc_codel_xstats st = {
249 		.maxpacket	= q->stats.maxpacket,
250 		.count		= q->vars.count,
251 		.lastcount	= q->vars.lastcount,
252 		.drop_overlimit = q->drop_overlimit,
253 		.ldelay		= codel_time_to_us(q->vars.ldelay),
254 		.dropping	= q->vars.dropping,
255 		.ecn_mark	= q->stats.ecn_mark,
256 		.ce_mark	= q->stats.ce_mark,
257 	};
258 
259 	if (q->vars.dropping) {
260 		codel_tdiff_t delta = q->vars.drop_next - codel_get_time();
261 
262 		if (delta >= 0)
263 			st.drop_next = codel_time_to_us(delta);
264 		else
265 			st.drop_next = -codel_time_to_us(-delta);
266 	}
267 
268 	return gnet_stats_copy_app(d, &st, sizeof(st));
269 }
270 
codel_reset(struct Qdisc * sch)271 static void codel_reset(struct Qdisc *sch)
272 {
273 	struct codel_sched_data *q = qdisc_priv(sch);
274 
275 	qdisc_reset_queue(sch);
276 	codel_vars_init(&q->vars);
277 }
278 
279 static struct Qdisc_ops codel_qdisc_ops __read_mostly = {
280 	.id		=	"codel",
281 	.priv_size	=	sizeof(struct codel_sched_data),
282 
283 	.enqueue	=	codel_qdisc_enqueue,
284 	.dequeue	=	codel_qdisc_dequeue,
285 	.peek		=	qdisc_peek_dequeued,
286 	.init		=	codel_init,
287 	.reset		=	codel_reset,
288 	.change 	=	codel_change,
289 	.dump		=	codel_dump,
290 	.dump_stats	=	codel_dump_stats,
291 	.owner		=	THIS_MODULE,
292 };
293 
codel_module_init(void)294 static int __init codel_module_init(void)
295 {
296 	return register_qdisc(&codel_qdisc_ops);
297 }
298 
codel_module_exit(void)299 static void __exit codel_module_exit(void)
300 {
301 	unregister_qdisc(&codel_qdisc_ops);
302 }
303 
304 module_init(codel_module_init)
305 module_exit(codel_module_exit)
306 
307 MODULE_DESCRIPTION("Controlled Delay queue discipline");
308 MODULE_AUTHOR("Dave Taht");
309 MODULE_AUTHOR("Eric Dumazet");
310 MODULE_LICENSE("Dual BSD/GPL");
311