diff options
-rw-r--r-- | include/linux/pkt_sched.h | 26 | ||||
-rw-r--r-- | include/net/codel.h | 332 | ||||
-rw-r--r-- | net/sched/Kconfig | 11 | ||||
-rw-r--r-- | net/sched/Makefile | 1 | ||||
-rw-r--r-- | net/sched/sch_codel.c | 275 |
5 files changed, 645 insertions, 0 deletions
diff --git a/include/linux/pkt_sched.h b/include/linux/pkt_sched.h index ffe975c3f1d8..cde56c22bdab 100644 --- a/include/linux/pkt_sched.h +++ b/include/linux/pkt_sched.h | |||
@@ -655,4 +655,30 @@ struct tc_qfq_stats { | |||
655 | __u32 lmax; | 655 | __u32 lmax; |
656 | }; | 656 | }; |
657 | 657 | ||
658 | /* CODEL */ | ||
659 | |||
660 | enum { | ||
661 | TCA_CODEL_UNSPEC, | ||
662 | TCA_CODEL_TARGET, | ||
663 | TCA_CODEL_LIMIT, | ||
664 | TCA_CODEL_INTERVAL, | ||
665 | TCA_CODEL_ECN, | ||
666 | __TCA_CODEL_MAX | ||
667 | }; | ||
668 | |||
669 | #define TCA_CODEL_MAX (__TCA_CODEL_MAX - 1) | ||
670 | |||
671 | struct tc_codel_xstats { | ||
672 | __u32 maxpacket; /* largest packet we've seen so far */ | ||
673 | __u32 count; /* how many drops we've done since the last time we | ||
674 | * entered dropping state | ||
675 | */ | ||
676 | __u32 lastcount; /* count at entry to dropping state */ | ||
677 | __u32 ldelay; /* in-queue delay seen by most recently dequeued packet */ | ||
678 | __s32 drop_next; /* time to drop next packet */ | ||
679 | __u32 drop_overlimit; /* number of time max qdisc packet limit was hit */ | ||
680 | __u32 ecn_mark; /* number of packets we ECN marked instead of dropped */ | ||
681 | __u32 dropping; /* are we in dropping state ? */ | ||
682 | }; | ||
683 | |||
658 | #endif | 684 | #endif |
diff --git a/include/net/codel.h b/include/net/codel.h new file mode 100644 index 000000000000..bce2cefa8c94 --- /dev/null +++ b/include/net/codel.h | |||
@@ -0,0 +1,332 @@ | |||
1 | #ifndef __NET_SCHED_CODEL_H | ||
2 | #define __NET_SCHED_CODEL_H | ||
3 | |||
4 | /* | ||
5 | * Codel - The Controlled-Delay Active Queue Management algorithm | ||
6 | * | ||
7 | * Copyright (C) 2011-2012 Kathleen Nichols <nichols@pollere.com> | ||
8 | * Copyright (C) 2011-2012 Van Jacobson <van@pollere.net> | ||
9 | * Copyright (C) 2012 Michael D. Taht <dave.taht@bufferbloat.net> | ||
10 | * Copyright (C) 2012 Eric Dumazet <edumazet@google.com> | ||
11 | * | ||
12 | * Redistribution and use in source and binary forms, with or without | ||
13 | * modification, are permitted provided that the following conditions | ||
14 | * are met: | ||
15 | * 1. Redistributions of source code must retain the above copyright | ||
16 | * notice, this list of conditions, and the following disclaimer, | ||
17 | * without modification. | ||
18 | * 2. Redistributions in binary form must reproduce the above copyright | ||
19 | * notice, this list of conditions and the following disclaimer in the | ||
20 | * documentation and/or other materials provided with the distribution. | ||
21 | * 3. The names of the authors may not be used to endorse or promote products | ||
22 | * derived from this software without specific prior written permission. | ||
23 | * | ||
24 | * Alternatively, provided that this notice is retained in full, this | ||
25 | * software may be distributed under the terms of the GNU General | ||
26 | * Public License ("GPL") version 2, in which case the provisions of the | ||
27 | * GPL apply INSTEAD OF those given above. | ||
28 | * | ||
29 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | ||
30 | * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | ||
31 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | ||
32 | * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT | ||
33 | * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, | ||
34 | * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT | ||
35 | * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, | ||
36 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY | ||
37 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT | ||
38 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE | ||
39 | * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH | ||
40 | * DAMAGE. | ||
41 | * | ||
42 | */ | ||
43 | |||
44 | #include <linux/types.h> | ||
45 | #include <linux/ktime.h> | ||
46 | #include <linux/skbuff.h> | ||
47 | #include <net/pkt_sched.h> | ||
48 | #include <net/inet_ecn.h> | ||
49 | |||
50 | /* Controlling Queue Delay (CoDel) algorithm | ||
51 | * ========================================= | ||
52 | * Source : Kathleen Nichols and Van Jacobson | ||
53 | * http://queue.acm.org/detail.cfm?id=2209336 | ||
54 | * | ||
55 | * Implemented on linux by Dave Taht and Eric Dumazet | ||
56 | */ | ||
57 | |||
58 | |||
59 | /* CoDel uses a 1024 nsec clock, encoded in u32 | ||
60 | * This gives a range of 2199 seconds, because of signed compares | ||
61 | */ | ||
62 | typedef u32 codel_time_t; | ||
63 | typedef s32 codel_tdiff_t; | ||
64 | #define CODEL_SHIFT 10 | ||
65 | #define MS2TIME(a) ((a * NSEC_PER_MSEC) >> CODEL_SHIFT) | ||
66 | |||
67 | static inline codel_time_t codel_get_time(void) | ||
68 | { | ||
69 | u64 ns = ktime_to_ns(ktime_get()); | ||
70 | |||
71 | return ns >> CODEL_SHIFT; | ||
72 | } | ||
73 | |||
74 | #define codel_time_after(a, b) ((s32)(a) - (s32)(b) > 0) | ||
75 | #define codel_time_after_eq(a, b) ((s32)(a) - (s32)(b) >= 0) | ||
76 | #define codel_time_before(a, b) ((s32)(a) - (s32)(b) < 0) | ||
77 | #define codel_time_before_eq(a, b) ((s32)(a) - (s32)(b) <= 0) | ||
78 | |||
79 | /* Qdiscs using codel plugin must use codel_skb_cb in their own cb[] */ | ||
80 | struct codel_skb_cb { | ||
81 | codel_time_t enqueue_time; | ||
82 | }; | ||
83 | |||
84 | static struct codel_skb_cb *get_codel_cb(const struct sk_buff *skb) | ||
85 | { | ||
86 | qdisc_cb_private_validate(skb, sizeof(struct codel_skb_cb)); | ||
87 | return (struct codel_skb_cb *)qdisc_skb_cb(skb)->data; | ||
88 | } | ||
89 | |||
90 | static codel_time_t codel_get_enqueue_time(const struct sk_buff *skb) | ||
91 | { | ||
92 | return get_codel_cb(skb)->enqueue_time; | ||
93 | } | ||
94 | |||
95 | static void codel_set_enqueue_time(struct sk_buff *skb) | ||
96 | { | ||
97 | get_codel_cb(skb)->enqueue_time = codel_get_time(); | ||
98 | } | ||
99 | |||
100 | static inline u32 codel_time_to_us(codel_time_t val) | ||
101 | { | ||
102 | u64 valns = ((u64)val << CODEL_SHIFT); | ||
103 | |||
104 | do_div(valns, NSEC_PER_USEC); | ||
105 | return (u32)valns; | ||
106 | } | ||
107 | |||
108 | /** | ||
109 | * struct codel_params - contains codel parameters | ||
110 | * @target: target queue size (in time units) | ||
111 | * @interval: width of moving time window | ||
112 | * @ecn: is Explicit Congestion Notification enabled | ||
113 | */ | ||
114 | struct codel_params { | ||
115 | codel_time_t target; | ||
116 | codel_time_t interval; | ||
117 | bool ecn; | ||
118 | }; | ||
119 | |||
120 | /** | ||
121 | * struct codel_vars - contains codel variables | ||
122 | * @count: how many drops we've done since the last time we | ||
123 | * entered dropping state | ||
124 | * @lastcount: count at entry to dropping state | ||
125 | * @dropping: set to true if in dropping state | ||
126 | * @first_above_time: when we went (or will go) continuously above target | ||
127 | * for interval | ||
128 | * @drop_next: time to drop next packet, or when we dropped last | ||
129 | * @ldelay: sojourn time of last dequeued packet | ||
130 | */ | ||
131 | struct codel_vars { | ||
132 | u32 count; | ||
133 | u32 lastcount; | ||
134 | bool dropping; | ||
135 | codel_time_t first_above_time; | ||
136 | codel_time_t drop_next; | ||
137 | codel_time_t ldelay; | ||
138 | }; | ||
139 | |||
140 | /** | ||
141 | * struct codel_stats - contains codel shared variables and stats | ||
142 | * @maxpacket: largest packet we've seen so far | ||
143 | * @drop_count: temp count of dropped packets in dequeue() | ||
144 | * ecn_mark: number of packets we ECN marked instead of dropping | ||
145 | */ | ||
146 | struct codel_stats { | ||
147 | u32 maxpacket; | ||
148 | u32 drop_count; | ||
149 | u32 ecn_mark; | ||
150 | }; | ||
151 | |||
152 | static void codel_params_init(struct codel_params *params) | ||
153 | { | ||
154 | params->interval = MS2TIME(100); | ||
155 | params->target = MS2TIME(5); | ||
156 | params->ecn = false; | ||
157 | } | ||
158 | |||
159 | static void codel_vars_init(struct codel_vars *vars) | ||
160 | { | ||
161 | vars->drop_next = 0; | ||
162 | vars->first_above_time = 0; | ||
163 | vars->dropping = false; /* exit dropping state */ | ||
164 | vars->count = 0; | ||
165 | vars->lastcount = 0; | ||
166 | } | ||
167 | |||
168 | static void codel_stats_init(struct codel_stats *stats) | ||
169 | { | ||
170 | stats->maxpacket = 256; | ||
171 | } | ||
172 | |||
173 | /* return interval/sqrt(x) with good precision | ||
174 | * relies on int_sqrt(unsigned long x) kernel implementation | ||
175 | */ | ||
176 | static u32 codel_inv_sqrt(u32 _interval, u32 _x) | ||
177 | { | ||
178 | u64 interval = _interval; | ||
179 | unsigned long x = _x; | ||
180 | |||
181 | /* Scale operands for max precision */ | ||
182 | |||
183 | #if BITS_PER_LONG == 64 | ||
184 | x <<= 32; /* On 64bit arches, we can prescale x by 32bits */ | ||
185 | interval <<= 16; | ||
186 | #endif | ||
187 | |||
188 | while (x < (1UL << (BITS_PER_LONG - 2))) { | ||
189 | x <<= 2; | ||
190 | interval <<= 1; | ||
191 | } | ||
192 | do_div(interval, int_sqrt(x)); | ||
193 | return (u32)interval; | ||
194 | } | ||
195 | |||
196 | static codel_time_t codel_control_law(codel_time_t t, | ||
197 | codel_time_t interval, | ||
198 | u32 count) | ||
199 | { | ||
200 | return t + codel_inv_sqrt(interval, count); | ||
201 | } | ||
202 | |||
203 | |||
204 | static bool codel_should_drop(struct sk_buff *skb, | ||
205 | unsigned int *backlog, | ||
206 | struct codel_vars *vars, | ||
207 | struct codel_params *params, | ||
208 | struct codel_stats *stats, | ||
209 | codel_time_t now) | ||
210 | { | ||
211 | bool ok_to_drop; | ||
212 | |||
213 | if (!skb) { | ||
214 | vars->first_above_time = 0; | ||
215 | return false; | ||
216 | } | ||
217 | |||
218 | vars->ldelay = now - codel_get_enqueue_time(skb); | ||
219 | *backlog -= qdisc_pkt_len(skb); | ||
220 | |||
221 | if (unlikely(qdisc_pkt_len(skb) > stats->maxpacket)) | ||
222 | stats->maxpacket = qdisc_pkt_len(skb); | ||
223 | |||
224 | if (codel_time_before(vars->ldelay, params->target) || | ||
225 | *backlog <= stats->maxpacket) { | ||
226 | /* went below - stay below for at least interval */ | ||
227 | vars->first_above_time = 0; | ||
228 | return false; | ||
229 | } | ||
230 | ok_to_drop = false; | ||
231 | if (vars->first_above_time == 0) { | ||
232 | /* just went above from below. If we stay above | ||
233 | * for at least interval we'll say it's ok to drop | ||
234 | */ | ||
235 | vars->first_above_time = now + params->interval; | ||
236 | } else if (codel_time_after(now, vars->first_above_time)) { | ||
237 | ok_to_drop = true; | ||
238 | } | ||
239 | return ok_to_drop; | ||
240 | } | ||
241 | |||
242 | typedef struct sk_buff * (*codel_skb_dequeue_t)(struct codel_vars *vars, | ||
243 | struct Qdisc *sch); | ||
244 | |||
245 | static struct sk_buff *codel_dequeue(struct Qdisc *sch, | ||
246 | struct codel_params *params, | ||
247 | struct codel_vars *vars, | ||
248 | struct codel_stats *stats, | ||
249 | codel_skb_dequeue_t dequeue_func, | ||
250 | u32 *backlog) | ||
251 | { | ||
252 | struct sk_buff *skb = dequeue_func(vars, sch); | ||
253 | codel_time_t now; | ||
254 | bool drop; | ||
255 | |||
256 | if (!skb) { | ||
257 | vars->dropping = false; | ||
258 | return skb; | ||
259 | } | ||
260 | now = codel_get_time(); | ||
261 | drop = codel_should_drop(skb, backlog, vars, params, stats, now); | ||
262 | if (vars->dropping) { | ||
263 | if (!drop) { | ||
264 | /* sojourn time below target - leave dropping state */ | ||
265 | vars->dropping = false; | ||
266 | } else if (codel_time_after_eq(now, vars->drop_next)) { | ||
267 | /* It's time for the next drop. Drop the current | ||
268 | * packet and dequeue the next. The dequeue might | ||
269 | * take us out of dropping state. | ||
270 | * If not, schedule the next drop. | ||
271 | * A large backlog might result in drop rates so high | ||
272 | * that the next drop should happen now, | ||
273 | * hence the while loop. | ||
274 | */ | ||
275 | while (vars->dropping && | ||
276 | codel_time_after_eq(now, vars->drop_next)) { | ||
277 | if (++vars->count == 0) /* avoid zero divides */ | ||
278 | vars->count = ~0U; | ||
279 | if (params->ecn && INET_ECN_set_ce(skb)) { | ||
280 | stats->ecn_mark++; | ||
281 | vars->drop_next = | ||
282 | codel_control_law(vars->drop_next, | ||
283 | params->interval, | ||
284 | vars->count); | ||
285 | goto end; | ||
286 | } | ||
287 | qdisc_drop(skb, sch); | ||
288 | stats->drop_count++; | ||
289 | skb = dequeue_func(vars, sch); | ||
290 | if (!codel_should_drop(skb, backlog, | ||
291 | vars, params, stats, now)) { | ||
292 | /* leave dropping state */ | ||
293 | vars->dropping = false; | ||
294 | } else { | ||
295 | /* and schedule the next drop */ | ||
296 | vars->drop_next = | ||
297 | codel_control_law(vars->drop_next, | ||
298 | params->interval, | ||
299 | vars->count); | ||
300 | } | ||
301 | } | ||
302 | } | ||
303 | } else if (drop) { | ||
304 | if (params->ecn && INET_ECN_set_ce(skb)) { | ||
305 | stats->ecn_mark++; | ||
306 | } else { | ||
307 | qdisc_drop(skb, sch); | ||
308 | stats->drop_count++; | ||
309 | |||
310 | skb = dequeue_func(vars, sch); | ||
311 | drop = codel_should_drop(skb, backlog, vars, params, | ||
312 | stats, now); | ||
313 | } | ||
314 | vars->dropping = true; | ||
315 | /* if min went above target close to when we last went below it | ||
316 | * assume that the drop rate that controlled the queue on the | ||
317 | * last cycle is a good starting point to control it now. | ||
318 | */ | ||
319 | if (codel_time_before(now - vars->drop_next, | ||
320 | 16 * params->interval)) { | ||
321 | vars->count = (vars->count - vars->lastcount) | 1; | ||
322 | } else { | ||
323 | vars->count = 1; | ||
324 | } | ||
325 | vars->lastcount = vars->count; | ||
326 | vars->drop_next = codel_control_law(now, params->interval, | ||
327 | vars->count); | ||
328 | } | ||
329 | end: | ||
330 | return skb; | ||
331 | } | ||
332 | #endif | ||
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 | ||
253 | config 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 | |||
253 | config NET_SCH_INGRESS | 264 | config 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-$(CONFIG_NET_SCH_PLUG) += sch_plug.o | |||
37 | obj-$(CONFIG_NET_SCH_MQPRIO) += sch_mqprio.o | 37 | obj-$(CONFIG_NET_SCH_MQPRIO) += sch_mqprio.o |
38 | obj-$(CONFIG_NET_SCH_CHOKE) += sch_choke.o | 38 | obj-$(CONFIG_NET_SCH_CHOKE) += sch_choke.o |
39 | obj-$(CONFIG_NET_SCH_QFQ) += sch_qfq.o | 39 | obj-$(CONFIG_NET_SCH_QFQ) += sch_qfq.o |
40 | obj-$(CONFIG_NET_SCH_CODEL) += sch_codel.o | ||
40 | 41 | ||
41 | obj-$(CONFIG_NET_CLS_U32) += cls_u32.o | 42 | obj-$(CONFIG_NET_CLS_U32) += cls_u32.o |
42 | obj-$(CONFIG_NET_CLS_ROUTE4) += cls_route.o | 43 | obj-$(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 | |||
55 | struct 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 | */ | ||
66 | static 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 | |||
74 | static 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 | |||
93 | static 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 | |||
106 | static 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 | |||
113 | static 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 | |||
160 | static 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 | |||
185 | static 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 | |||
206 | nla_put_failure: | ||
207 | nla_nest_cancel(skb, opts); | ||
208 | return -1; | ||
209 | } | ||
210 | |||
211 | static 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 | |||
236 | static 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 | |||
244 | static 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 | |||
259 | static int __init codel_module_init(void) | ||
260 | { | ||
261 | return register_qdisc(&codel_qdisc_ops); | ||
262 | } | ||
263 | |||
264 | static void __exit codel_module_exit(void) | ||
265 | { | ||
266 | unregister_qdisc(&codel_qdisc_ops); | ||
267 | } | ||
268 | |||
269 | module_init(codel_module_init) | ||
270 | module_exit(codel_module_exit) | ||
271 | |||
272 | MODULE_DESCRIPTION("Controlled Delay queue discipline"); | ||
273 | MODULE_AUTHOR("Dave Taht"); | ||
274 | MODULE_AUTHOR("Eric Dumazet"); | ||
275 | MODULE_LICENSE("Dual BSD/GPL"); | ||