aboutsummaryrefslogtreecommitdiffstats
path: root/net/sched
diff options
context:
space:
mode:
Diffstat (limited to 'net/sched')
-rw-r--r--net/sched/Kconfig11
-rw-r--r--net/sched/Makefile1
-rw-r--r--net/sched/sch_codel.c275
3 files changed, 287 insertions, 0 deletions
diff --git a/net/sched/Kconfig b/net/sched/Kconfig
index 75b58f81d53d..fadd2522053d 100644
--- a/net/sched/Kconfig
+++ b/net/sched/Kconfig
@@ -250,6 +250,17 @@ config NET_SCH_QFQ
250 250
251 If unsure, say N. 251 If unsure, say N.
252 252
253config NET_SCH_CODEL
254 tristate "Controlled Delay AQM (CODEL)"
255 help
256 Say Y here if you want to use the Controlled Delay (CODEL)
257 packet scheduling algorithm.
258
259 To compile this driver as a module, choose M here: the module
260 will be called sch_codel.
261
262 If unsure, say N.
263
253config NET_SCH_INGRESS 264config NET_SCH_INGRESS
254 tristate "Ingress Qdisc" 265 tristate "Ingress Qdisc"
255 depends on NET_CLS_ACT 266 depends on NET_CLS_ACT
diff --git a/net/sched/Makefile b/net/sched/Makefile
index 8cdf4e2b51d3..30fab03b8516 100644
--- a/net/sched/Makefile
+++ b/net/sched/Makefile
@@ -37,6 +37,7 @@ obj-$(CONFIG_NET_SCH_PLUG) += sch_plug.o
37obj-$(CONFIG_NET_SCH_MQPRIO) += sch_mqprio.o 37obj-$(CONFIG_NET_SCH_MQPRIO) += sch_mqprio.o
38obj-$(CONFIG_NET_SCH_CHOKE) += sch_choke.o 38obj-$(CONFIG_NET_SCH_CHOKE) += sch_choke.o
39obj-$(CONFIG_NET_SCH_QFQ) += sch_qfq.o 39obj-$(CONFIG_NET_SCH_QFQ) += sch_qfq.o
40obj-$(CONFIG_NET_SCH_CODEL) += sch_codel.o
40 41
41obj-$(CONFIG_NET_CLS_U32) += cls_u32.o 42obj-$(CONFIG_NET_CLS_U32) += cls_u32.o
42obj-$(CONFIG_NET_CLS_ROUTE4) += cls_route.o 43obj-$(CONFIG_NET_CLS_ROUTE4) += cls_route.o
diff --git a/net/sched/sch_codel.c b/net/sched/sch_codel.c
new file mode 100644
index 000000000000..b4a1a81e757e
--- /dev/null
+++ b/net/sched/sch_codel.c
@@ -0,0 +1,275 @@
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 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 <net/pkt_sched.h>
50#include <net/codel.h>
51
52
53#define DEFAULT_CODEL_LIMIT 1000
54
55struct codel_sched_data {
56 struct codel_params params;
57 struct codel_vars vars;
58 struct codel_stats stats;
59 u32 drop_overlimit;
60};
61
62/* This is the specific function called from codel_dequeue()
63 * to dequeue a packet from queue. Note: backlog is handled in
64 * codel, we dont need to reduce it here.
65 */
66static struct sk_buff *dequeue(struct codel_vars *vars, struct Qdisc *sch)
67{
68 struct sk_buff *skb = __skb_dequeue(&sch->q);
69
70 prefetch(&skb->end); /* we'll need skb_shinfo() */
71 return skb;
72}
73
74static struct sk_buff *codel_qdisc_dequeue(struct Qdisc *sch)
75{
76 struct codel_sched_data *q = qdisc_priv(sch);
77 struct sk_buff *skb;
78
79 skb = codel_dequeue(sch, &q->params, &q->vars, &q->stats,
80 dequeue, &sch->qstats.backlog);
81 /* We cant call qdisc_tree_decrease_qlen() if our qlen is 0,
82 * or HTB crashes. Defer it for next round.
83 */
84 if (q->stats.drop_count && sch->q.qlen) {
85 qdisc_tree_decrease_qlen(sch, q->stats.drop_count);
86 q->stats.drop_count = 0;
87 }
88 if (skb)
89 qdisc_bstats_update(sch, skb);
90 return skb;
91}
92
93static int codel_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch)
94{
95 struct codel_sched_data *q;
96
97 if (likely(qdisc_qlen(sch) < sch->limit)) {
98 codel_set_enqueue_time(skb);
99 return qdisc_enqueue_tail(skb, sch);
100 }
101 q = qdisc_priv(sch);
102 q->drop_overlimit++;
103 return qdisc_drop(skb, sch);
104}
105
106static const struct nla_policy codel_policy[TCA_CODEL_MAX + 1] = {
107 [TCA_CODEL_TARGET] = { .type = NLA_U32 },
108 [TCA_CODEL_LIMIT] = { .type = NLA_U32 },
109 [TCA_CODEL_INTERVAL] = { .type = NLA_U32 },
110 [TCA_CODEL_ECN] = { .type = NLA_U32 },
111};
112
113static int codel_change(struct Qdisc *sch, struct nlattr *opt)
114{
115 struct codel_sched_data *q = qdisc_priv(sch);
116 struct nlattr *tb[TCA_CODEL_MAX + 1];
117 unsigned int qlen;
118 int err;
119
120 if (!opt)
121 return -EINVAL;
122
123 err = nla_parse_nested(tb, TCA_CODEL_MAX, opt, codel_policy);
124 if (err < 0)
125 return err;
126
127 sch_tree_lock(sch);
128
129 if (tb[TCA_CODEL_TARGET]) {
130 u32 target = nla_get_u32(tb[TCA_CODEL_TARGET]);
131
132 q->params.target = ((u64)target * NSEC_PER_USEC) >> CODEL_SHIFT;
133 }
134
135 if (tb[TCA_CODEL_INTERVAL]) {
136 u32 interval = nla_get_u32(tb[TCA_CODEL_INTERVAL]);
137
138 q->params.interval = ((u64)interval * NSEC_PER_USEC) >> CODEL_SHIFT;
139 }
140
141 if (tb[TCA_CODEL_LIMIT])
142 sch->limit = nla_get_u32(tb[TCA_CODEL_LIMIT]);
143
144 if (tb[TCA_CODEL_ECN])
145 q->params.ecn = !!nla_get_u32(tb[TCA_CODEL_ECN]);
146
147 qlen = sch->q.qlen;
148 while (sch->q.qlen > sch->limit) {
149 struct sk_buff *skb = __skb_dequeue(&sch->q);
150
151 sch->qstats.backlog -= qdisc_pkt_len(skb);
152 qdisc_drop(skb, sch);
153 }
154 qdisc_tree_decrease_qlen(sch, qlen - sch->q.qlen);
155
156 sch_tree_unlock(sch);
157 return 0;
158}
159
160static int codel_init(struct Qdisc *sch, struct nlattr *opt)
161{
162 struct codel_sched_data *q = qdisc_priv(sch);
163
164 sch->limit = DEFAULT_CODEL_LIMIT;
165
166 codel_params_init(&q->params);
167 codel_vars_init(&q->vars);
168 codel_stats_init(&q->stats);
169
170 if (opt) {
171 int err = codel_change(sch, opt);
172
173 if (err)
174 return err;
175 }
176
177 if (sch->limit >= 1)
178 sch->flags |= TCQ_F_CAN_BYPASS;
179 else
180 sch->flags &= ~TCQ_F_CAN_BYPASS;
181
182 return 0;
183}
184
185static int codel_dump(struct Qdisc *sch, struct sk_buff *skb)
186{
187 struct codel_sched_data *q = qdisc_priv(sch);
188 struct nlattr *opts;
189
190 opts = nla_nest_start(skb, TCA_OPTIONS);
191 if (opts == NULL)
192 goto nla_put_failure;
193
194 if (nla_put_u32(skb, TCA_CODEL_TARGET,
195 codel_time_to_us(q->params.target)) ||
196 nla_put_u32(skb, TCA_CODEL_LIMIT,
197 sch->limit) ||
198 nla_put_u32(skb, TCA_CODEL_INTERVAL,
199 codel_time_to_us(q->params.interval)) ||
200 nla_put_u32(skb, TCA_CODEL_ECN,
201 q->params.ecn))
202 goto nla_put_failure;
203
204 return nla_nest_end(skb, opts);
205
206nla_put_failure:
207 nla_nest_cancel(skb, opts);
208 return -1;
209}
210
211static int codel_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
212{
213 const struct codel_sched_data *q = qdisc_priv(sch);
214 struct tc_codel_xstats st = {
215 .maxpacket = q->stats.maxpacket,
216 .count = q->vars.count,
217 .lastcount = q->vars.lastcount,
218 .drop_overlimit = q->drop_overlimit,
219 .ldelay = codel_time_to_us(q->vars.ldelay),
220 .dropping = q->vars.dropping,
221 .ecn_mark = q->stats.ecn_mark,
222 };
223
224 if (q->vars.dropping) {
225 codel_tdiff_t delta = q->vars.drop_next - codel_get_time();
226
227 if (delta >= 0)
228 st.drop_next = codel_time_to_us(delta);
229 else
230 st.drop_next = -codel_time_to_us(-delta);
231 }
232
233 return gnet_stats_copy_app(d, &st, sizeof(st));
234}
235
236static void codel_reset(struct Qdisc *sch)
237{
238 struct codel_sched_data *q = qdisc_priv(sch);
239
240 qdisc_reset_queue(sch);
241 codel_vars_init(&q->vars);
242}
243
244static struct Qdisc_ops codel_qdisc_ops __read_mostly = {
245 .id = "codel",
246 .priv_size = sizeof(struct codel_sched_data),
247
248 .enqueue = codel_qdisc_enqueue,
249 .dequeue = codel_qdisc_dequeue,
250 .peek = qdisc_peek_dequeued,
251 .init = codel_init,
252 .reset = codel_reset,
253 .change = codel_change,
254 .dump = codel_dump,
255 .dump_stats = codel_dump_stats,
256 .owner = THIS_MODULE,
257};
258
259static int __init codel_module_init(void)
260{
261 return register_qdisc(&codel_qdisc_ops);
262}
263
264static void __exit codel_module_exit(void)
265{
266 unregister_qdisc(&codel_qdisc_ops);
267}
268
269module_init(codel_module_init)
270module_exit(codel_module_exit)
271
272MODULE_DESCRIPTION("Controlled Delay queue discipline");
273MODULE_AUTHOR("Dave Taht");
274MODULE_AUTHOR("Eric Dumazet");
275MODULE_LICENSE("Dual BSD/GPL");