?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
mm, x86/mm: Fix performance regression in get_user_pages_fast()
[linux-2.6/btrfs-unstable.git] / net / sched / sch_red.c
blob11292adce4121022a86aff0aed537be3d9319490
1 /*
2 * net/sched/sch_red.c Random Early Detection queue.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 *
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 *
11 * Changes:
12 * J Hadi Salim 980914: computation fixes
13 * Alexey Makarenko <makar@phoenix.kharkov.ua> 990814: qave on idle link was calculated incorrectly.
14 * J Hadi Salim 980816: ECN support
15 */
16
17 #include <linux/module.h>
18 #include <linux/types.h>
19 #include <linux/kernel.h>
20 #include <linux/skbuff.h>
21 #include <net/pkt_sched.h>
22 #include <net/inet_ecn.h>
23 #include <net/red.h>
24
25
26 /* Parameters, settable by user:
27 -----------------------------
28
29 limit - bytes (must be > qth_max + burst)
30
31 Hard limit on queue length, should be chosen >qth_max
32 to allow packet bursts. This parameter does not
33 affect the algorithms behaviour and can be chosen
34 arbitrarily high (well, less than ram size)
35 Really, this limit will never be reached
36 if RED works correctly.
37 */
38
39 struct red_sched_data {
40 u32 limit; /* HARD maximal queue length */
41 unsigned char flags;
42 struct timer_list adapt_timer;
43 struct red_parms parms;
44 struct red_vars vars;
45 struct red_stats stats;
46 struct Qdisc *qdisc;
47 };
48
49 static inline int red_use_ecn(struct red_sched_data *q)
50 {
51 return q->flags & TC_RED_ECN;
52 }
53
54 static inline int red_use_harddrop(struct red_sched_data *q)
55 {
56 return q->flags & TC_RED_HARDDROP;
57 }
58
59 static int red_enqueue(struct sk_buff *skb, struct Qdisc *sch,
60 struct sk_buff **to_free)
61 {
62 struct red_sched_data *q = qdisc_priv(sch);
63 struct Qdisc *child = q->qdisc;
64 int ret;
65
66 q->vars.qavg = red_calc_qavg(&q->parms,
67 &q->vars,
68 child->qstats.backlog);
69
70 if (red_is_idling(&q->vars))
71 red_end_of_idle_period(&q->vars);
72
73 switch (red_action(&q->parms, &q->vars, q->vars.qavg)) {
74 case RED_DONT_MARK:
75 break;
76
77 case RED_PROB_MARK:
78 qdisc_qstats_overlimit(sch);
79 if (!red_use_ecn(q) || !INET_ECN_set_ce(skb)) {
80 q->stats.prob_drop++;
81 goto congestion_drop;
82 }
83
84 q->stats.prob_mark++;
85 break;
86
87 case RED_HARD_MARK:
88 qdisc_qstats_overlimit(sch);
89 if (red_use_harddrop(q) || !red_use_ecn(q) ||
90 !INET_ECN_set_ce(skb)) {
91 q->stats.forced_drop++;
92 goto congestion_drop;
93 }
94
95 q->stats.forced_mark++;
96 break;
97 }
98
99 ret = qdisc_enqueue(skb, child, to_free);
100 if (likely(ret == NET_XMIT_SUCCESS)) {
101 qdisc_qstats_backlog_inc(sch, skb);
102 sch->q.qlen++;
103 } else if (net_xmit_drop_count(ret)) {
104 q->stats.pdrop++;
105 qdisc_qstats_drop(sch);
106 }
107 return ret;
108
109 congestion_drop:
110 qdisc_drop(skb, sch, to_free);
111 return NET_XMIT_CN;
112 }
113
114 static struct sk_buff *red_dequeue(struct Qdisc *sch)
115 {
116 struct sk_buff *skb;
117 struct red_sched_data *q = qdisc_priv(sch);
118 struct Qdisc *child = q->qdisc;
119
120 skb = child->dequeue(child);
121 if (skb) {
122 qdisc_bstats_update(sch, skb);
123 qdisc_qstats_backlog_dec(sch, skb);
124 sch->q.qlen--;
125 } else {
126 if (!red_is_idling(&q->vars))
127 red_start_of_idle_period(&q->vars);
128 }
129 return skb;
130 }
131
132 static struct sk_buff *red_peek(struct Qdisc *sch)
133 {
134 struct red_sched_data *q = qdisc_priv(sch);
135 struct Qdisc *child = q->qdisc;
136
137 return child->ops->peek(child);
138 }
139
140 static void red_reset(struct Qdisc *sch)
141 {
142 struct red_sched_data *q = qdisc_priv(sch);
143
144 qdisc_reset(q->qdisc);
145 sch->qstats.backlog = 0;
146 sch->q.qlen = 0;
147 red_restart(&q->vars);
148 }
149
150 static void red_destroy(struct Qdisc *sch)
151 {
152 struct red_sched_data *q = qdisc_priv(sch);
153
154 del_timer_sync(&q->adapt_timer);
155 qdisc_destroy(q->qdisc);
156 }
157
158 static const struct nla_policy red_policy[TCA_RED_MAX + 1] = {
159 [TCA_RED_PARMS] = { .len = sizeof(struct tc_red_qopt) },
160 [TCA_RED_STAB] = { .len = RED_STAB_SIZE },
161 [TCA_RED_MAX_P] = { .type = NLA_U32 },
162 };
163
164 static int red_change(struct Qdisc *sch, struct nlattr *opt)
165 {
166 struct red_sched_data *q = qdisc_priv(sch);
167 struct nlattr *tb[TCA_RED_MAX + 1];
168 struct tc_red_qopt *ctl;
169 struct Qdisc *child = NULL;
170 int err;
171 u32 max_P;
172
173 if (opt == NULL)
174 return -EINVAL;
175
176 err = nla_parse_nested(tb, TCA_RED_MAX, opt, red_policy, NULL);
177 if (err < 0)
178 return err;
179
180 if (tb[TCA_RED_PARMS] == NULL ||
181 tb[TCA_RED_STAB] == NULL)
182 return -EINVAL;
183
184 max_P = tb[TCA_RED_MAX_P] ? nla_get_u32(tb[TCA_RED_MAX_P]) : 0;
185
186 ctl = nla_data(tb[TCA_RED_PARMS]);
187
188 if (ctl->limit > 0) {
189 child = fifo_create_dflt(sch, &bfifo_qdisc_ops, ctl->limit);
190 if (IS_ERR(child))
191 return PTR_ERR(child);
192 }
193
194 if (child != &noop_qdisc)
195 qdisc_hash_add(child, true);
196 sch_tree_lock(sch);
197 q->flags = ctl->flags;
198 q->limit = ctl->limit;
199 if (child) {
200 qdisc_tree_reduce_backlog(q->qdisc, q->qdisc->q.qlen,
201 q->qdisc->qstats.backlog);
202 qdisc_destroy(q->qdisc);
203 q->qdisc = child;
204 }
205
206 red_set_parms(&q->parms,
207 ctl->qth_min, ctl->qth_max, ctl->Wlog,
208 ctl->Plog, ctl->Scell_log,
209 nla_data(tb[TCA_RED_STAB]),
210 max_P);
211 red_set_vars(&q->vars);
212
213 del_timer(&q->adapt_timer);
214 if (ctl->flags & TC_RED_ADAPTATIVE)
215 mod_timer(&q->adapt_timer, jiffies + HZ/2);
216
217 if (!q->qdisc->q.qlen)
218 red_start_of_idle_period(&q->vars);
219
220 sch_tree_unlock(sch);
221 return 0;
222 }
223
224 static inline void red_adaptative_timer(unsigned long arg)
225 {
226 struct Qdisc *sch = (struct Qdisc *)arg;
227 struct red_sched_data *q = qdisc_priv(sch);
228 spinlock_t *root_lock = qdisc_lock(qdisc_root_sleeping(sch));
229
230 spin_lock(root_lock);
231 red_adaptative_algo(&q->parms, &q->vars);
232 mod_timer(&q->adapt_timer, jiffies + HZ/2);
233 spin_unlock(root_lock);
234 }
235
236 static int red_init(struct Qdisc *sch, struct nlattr *opt)
237 {
238 struct red_sched_data *q = qdisc_priv(sch);
239
240 q->qdisc = &noop_qdisc;
241 setup_timer(&q->adapt_timer, red_adaptative_timer, (unsigned long)sch);
242 return red_change(sch, opt);
243 }
244
245 static int red_dump(struct Qdisc *sch, struct sk_buff *skb)
246 {
247 struct red_sched_data *q = qdisc_priv(sch);
248 struct nlattr *opts = NULL;
249 struct tc_red_qopt opt = {
250 .limit = q->limit,
251 .flags = q->flags,
252 .qth_min = q->parms.qth_min >> q->parms.Wlog,
253 .qth_max = q->parms.qth_max >> q->parms.Wlog,
254 .Wlog = q->parms.Wlog,
255 .Plog = q->parms.Plog,
256 .Scell_log = q->parms.Scell_log,
257 };
258
259 sch->qstats.backlog = q->qdisc->qstats.backlog;
260 opts = nla_nest_start(skb, TCA_OPTIONS);
261 if (opts == NULL)
262 goto nla_put_failure;
263 if (nla_put(skb, TCA_RED_PARMS, sizeof(opt), &opt) ||
264 nla_put_u32(skb, TCA_RED_MAX_P, q->parms.max_P))
265 goto nla_put_failure;
266 return nla_nest_end(skb, opts);
267
268 nla_put_failure:
269 nla_nest_cancel(skb, opts);
270 return -EMSGSIZE;
271 }
272
273 static int red_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
274 {
275 struct red_sched_data *q = qdisc_priv(sch);
276 struct tc_red_xstats st = {
277 .early = q->stats.prob_drop + q->stats.forced_drop,
278 .pdrop = q->stats.pdrop,
279 .other = q->stats.other,
280 .marked = q->stats.prob_mark + q->stats.forced_mark,
281 };
282
283 return gnet_stats_copy_app(d, &st, sizeof(st));
284 }
285
286 static int red_dump_class(struct Qdisc *sch, unsigned long cl,
287 struct sk_buff *skb, struct tcmsg *tcm)
288 {
289 struct red_sched_data *q = qdisc_priv(sch);
290
291 tcm->tcm_handle |= TC_H_MIN(1);
292 tcm->tcm_info = q->qdisc->handle;
293 return 0;
294 }
295
296 static int red_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
297 struct Qdisc **old)
298 {
299 struct red_sched_data *q = qdisc_priv(sch);
300
301 if (new == NULL)
302 new = &noop_qdisc;
303
304 *old = qdisc_replace(sch, new, &q->qdisc);
305 return 0;
306 }
307
308 static struct Qdisc *red_leaf(struct Qdisc *sch, unsigned long arg)
309 {
310 struct red_sched_data *q = qdisc_priv(sch);
311 return q->qdisc;
312 }
313
314 static unsigned long red_get(struct Qdisc *sch, u32 classid)
315 {
316 return 1;
317 }
318
319 static void red_put(struct Qdisc *sch, unsigned long arg)
320 {
321 }
322
323 static void red_walk(struct Qdisc *sch, struct qdisc_walker *walker)
324 {
325 if (!walker->stop) {
326 if (walker->count >= walker->skip)
327 if (walker->fn(sch, 1, walker) < 0) {
328 walker->stop = 1;
329 return;
330 }
331 walker->count++;
332 }
333 }
334
335 static const struct Qdisc_class_ops red_class_ops = {
336 .graft = red_graft,
337 .leaf = red_leaf,
338 .get = red_get,
339 .put = red_put,
340 .walk = red_walk,
341 .dump = red_dump_class,
342 };
343
344 static struct Qdisc_ops red_qdisc_ops __read_mostly = {
345 .id = "red",
346 .priv_size = sizeof(struct red_sched_data),
347 .cl_ops = &red_class_ops,
348 .enqueue = red_enqueue,
349 .dequeue = red_dequeue,
350 .peek = red_peek,
351 .init = red_init,
352 .reset = red_reset,
353 .destroy = red_destroy,
354 .change = red_change,
355 .dump = red_dump,
356 .dump_stats = red_dump_stats,
357 .owner = THIS_MODULE,
358 };
359
360 static int __init red_module_init(void)
361 {
362 return register_qdisc(&red_qdisc_ops);
363 }
364
365 static void __exit red_module_exit(void)
366 {
367 unregister_qdisc(&red_qdisc_ops);
368 }
369
370 module_init(red_module_init)
371 module_exit(red_module_exit)
372
373 MODULE_LICENSE("GPL");
(deprecated) Btrfs devel tree