net/sched/sch_codel.c

   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
  53
  54 #define DEFAULT_CODEL_LIMIT 1000
  55
  56 struct codel_sched_data {
  57         struct codel_params     params;
  58         struct codel_vars       vars;
  59         struct codel_stats      stats;
  60         u32                     drop_overlimit;
  61 };
  62
  63 /* This is the specific function called from codel_dequeue()
  64  * to dequeue a packet from queue. Note: backlog is handled in
  65  * codel, we dont need to reduce it here.
  66  */
  67 static struct sk_buff *dequeue(struct codel_vars *vars, struct Qdisc *sch)
  68 {
  69         struct sk_buff *skb = __skb_dequeue(&sch->q);
  70
  71         prefetch(&skb->end); /* we'll need skb_shinfo() */
  72         return skb;
  73 }
  74
  75 static struct sk_buff *codel_qdisc_dequeue(struct Qdisc *sch)
  76 {
  77         struct codel_sched_data *q = qdisc_priv(sch);
  78         struct sk_buff *skb;
  79
  80         skb = codel_dequeue(sch, &q->params, &q->vars, &q->stats, dequeue);
  81
  82         /* We cant call qdisc_tree_decrease_qlen() if our qlen is 0,
  83          * or HTB crashes. Defer it for next round.
  84          */
  85         if (q->stats.drop_count && sch->q.qlen) {
  86                 qdisc_tree_decrease_qlen(sch, q->stats.drop_count);
  87                 q->stats.drop_count = 0;
  88         }
  89         if (skb)
  90                 qdisc_bstats_update(sch, skb);
  91         return skb;
  92 }
  93
  94 static int codel_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch)
  95 {
  96         struct codel_sched_data *q;
  97
  98         if (likely(qdisc_qlen(sch) < sch->limit)) {
  99                 codel_set_enqueue_time(skb);
 100                 return qdisc_enqueue_tail(skb, sch);
 101         }
 102         q = qdisc_priv(sch);
 103         q->drop_overlimit++;
 104         return qdisc_drop(skb, sch);
 105 }
 106
 107 static const struct nla_policy codel_policy[TCA_CODEL_MAX + 1] = {
 108         [TCA_CODEL_TARGET]      = { .type = NLA_U32 },
 109         [TCA_CODEL_LIMIT]       = { .type = NLA_U32 },
 110         [TCA_CODEL_INTERVAL]    = { .type = NLA_U32 },
 111         [TCA_CODEL_ECN]         = { .type = NLA_U32 },
 112         [TCA_CODEL_CE_THRESHOLD]= { .type = NLA_U32 },
 113 };
 114
 115 static int codel_change(struct Qdisc *sch, struct nlattr *opt)
 116 {
 117         struct codel_sched_data *q = qdisc_priv(sch);
 118         struct nlattr *tb[TCA_CODEL_MAX + 1];
 119         unsigned int qlen;
 120         int err;
 121
 122         if (!opt)
 123                 return -EINVAL;
 124
 125         err = nla_parse_nested(tb, TCA_CODEL_MAX, opt, codel_policy);
 126         if (err < 0)
 127                 return err;
 128
 129         sch_tree_lock(sch);
 130
 131         if (tb[TCA_CODEL_TARGET]) {
 132                 u32 target = nla_get_u32(tb[TCA_CODEL_TARGET]);
 133
 134                 q->params.target = ((u64)target * NSEC_PER_USEC) >> CODEL_SHIFT;
 135         }
 136
 137         if (tb[TCA_CODEL_CE_THRESHOLD]) {
 138                 u64 val = nla_get_u32(tb[TCA_CODEL_CE_THRESHOLD]);
 139
 140                 q->params.ce_threshold = (val * NSEC_PER_USEC) >> CODEL_SHIFT;
 141         }
 142
 143         if (tb[TCA_CODEL_INTERVAL]) {
 144                 u32 interval = nla_get_u32(tb[TCA_CODEL_INTERVAL]);
 145
 146                 q->params.interval = ((u64)interval * NSEC_PER_USEC) >> CODEL_SHIFT;
 147         }
 148
 149         if (tb[TCA_CODEL_LIMIT])
 150                 sch->limit = nla_get_u32(tb[TCA_CODEL_LIMIT]);
 151
 152         if (tb[TCA_CODEL_ECN])
 153                 q->params.ecn = !!nla_get_u32(tb[TCA_CODEL_ECN]);
 154
 155         qlen = sch->q.qlen;
 156         while (sch->q.qlen > sch->limit) {
 157                 struct sk_buff *skb = __skb_dequeue(&sch->q);
 158
 159                 qdisc_qstats_backlog_dec(sch, skb);
 160                 qdisc_drop(skb, sch);
 161         }
 162         qdisc_tree_decrease_qlen(sch, qlen - sch->q.qlen);
 163
 164         sch_tree_unlock(sch);
 165         return 0;
 166 }
 167
 168 static int codel_init(struct Qdisc *sch, struct nlattr *opt)
 169 {
 170         struct codel_sched_data *q = qdisc_priv(sch);
 171
 172         sch->limit = DEFAULT_CODEL_LIMIT;
 173
 174         codel_params_init(&q->params, sch);
 175         codel_vars_init(&q->vars);
 176         codel_stats_init(&q->stats);
 177
 178         if (opt) {
 179                 int err = codel_change(sch, opt);
 180
 181                 if (err)
 182                         return err;
 183         }
 184
 185         if (sch->limit >= 1)
 186                 sch->flags |= TCQ_F_CAN_BYPASS;
 187         else
 188                 sch->flags &= ~TCQ_F_CAN_BYPASS;
 189
 190         return 0;
 191 }
 192
 193 static int codel_dump(struct Qdisc *sch, struct sk_buff *skb)
 194 {
 195         struct codel_sched_data *q = qdisc_priv(sch);
 196         struct nlattr *opts;
 197
 198         opts = nla_nest_start(skb, TCA_OPTIONS);
 199         if (opts == NULL)
 200                 goto nla_put_failure;
 201
 202         if (nla_put_u32(skb, TCA_CODEL_TARGET,
 203                         codel_time_to_us(q->params.target)) ||
 204             nla_put_u32(skb, TCA_CODEL_LIMIT,
 205                         sch->limit) ||
 206             nla_put_u32(skb, TCA_CODEL_INTERVAL,
 207                         codel_time_to_us(q->params.interval)) ||
 208             nla_put_u32(skb, TCA_CODEL_ECN,
 209                         q->params.ecn))
 210                 goto nla_put_failure;
 211         if (q->params.ce_threshold != CODEL_DISABLED_THRESHOLD &&
 212             nla_put_u32(skb, TCA_CODEL_CE_THRESHOLD,
 213                         codel_time_to_us(q->params.ce_threshold)))
 214                 goto nla_put_failure;
 215         return nla_nest_end(skb, opts);
 216
 217 nla_put_failure:
 218         nla_nest_cancel(skb, opts);
 219         return -1;
 220 }
 221
 222 static int codel_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
 223 {
 224         const struct codel_sched_data *q = qdisc_priv(sch);
 225         struct tc_codel_xstats st = {
 226                 .maxpacket      = q->stats.maxpacket,
 227                 .count          = q->vars.count,
 228                 .lastcount      = q->vars.lastcount,
 229                 .drop_overlimit = q->drop_overlimit,
 230                 .ldelay         = codel_time_to_us(q->vars.ldelay),
 231                 .dropping       = q->vars.dropping,
 232                 .ecn_mark       = q->stats.ecn_mark,
 233                 .ce_mark        = q->stats.ce_mark,
 234         };
 235
 236         if (q->vars.dropping) {
 237                 codel_tdiff_t delta = q->vars.drop_next - codel_get_time();
 238
 239                 if (delta >= 0)
 240                         st.drop_next = codel_time_to_us(delta);
 241                 else
 242                         st.drop_next = -codel_time_to_us(-delta);
 243         }
 244
 245         return gnet_stats_copy_app(d, &st, sizeof(st));
 246 }
 247
 248 static void codel_reset(struct Qdisc *sch)
 249 {
 250         struct codel_sched_data *q = qdisc_priv(sch);
 251
 252         qdisc_reset_queue(sch);
 253         codel_vars_init(&q->vars);
 254 }
 255
 256 static struct Qdisc_ops codel_qdisc_ops __read_mostly = {
 257         .id             =       "codel",
 258         .priv_size      =       sizeof(struct codel_sched_data),
 259
 260         .enqueue        =       codel_qdisc_enqueue,
 261         .dequeue        =       codel_qdisc_dequeue,
 262         .peek           =       qdisc_peek_dequeued,
 263         .init           =       codel_init,
 264         .reset          =       codel_reset,
 265         .change         =       codel_change,
 266         .dump           =       codel_dump,
 267         .dump_stats     =       codel_dump_stats,
 268         .owner          =       THIS_MODULE,
 269 };
 270
 271 static int __init codel_module_init(void)
 272 {
 273         return register_qdisc(&codel_qdisc_ops);
 274 }
 275
 276 static void __exit codel_module_exit(void)
 277 {
 278         unregister_qdisc(&codel_qdisc_ops);
 279 }
 280
 281 module_init(codel_module_init)
 282 module_exit(codel_module_exit)
 283
 284 MODULE_DESCRIPTION("Controlled Delay queue discipline");
 285 MODULE_AUTHOR("Dave Taht");
 286 MODULE_AUTHOR("Eric Dumazet");
 287 MODULE_LICENSE("Dual BSD/GPL");