summaryrefslogtreecommitdiffstats
path: root/net/sched/sch_red.c
blob: 7845d045eec4d1434037eaf0a964f33f70fcb0e9 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
/*
 * net/sched/sch_red.c	Random Early Detection queue.
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
 *
 * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
 *
 * Changes:
 * J Hadi Salim <hadi@nortel.com> 980914:	computation fixes
 * Alexey Makarenko <makar@phoenix.kharkov.ua> 990814: qave on idle link was calculated incorrectly.
 * J Hadi Salim <hadi@nortelnetworks.com> 980816:  ECN support	
 */

#include <linux/config.h>
#include <linux/module.h>
#include <asm/uaccess.h>
#include <asm/system.h>
#include <linux/bitops.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/string.h>
#include <linux/mm.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/in.h>
#include <linux/errno.h>
#include <linux/interrupt.h>
#include <linux/if_ether.h>
#include <linux/inet.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/notifier.h>
#include <net/ip.h>
#include <net/route.h>
#include <linux/skbuff.h>
#include <net/sock.h>
#include <net/pkt_sched.h>
#include <net/inet_ecn.h>
#include <net/dsfield.h>


/*	Random Early Detection (RED) algorithm.
	=======================================

	Source: Sally Floyd and Van Jacobson, "Random Early Detection Gateways
	for Congestion Avoidance", 1993, IEEE/ACM Transactions on Networking.

	This file codes a "divisionless" version of RED algorithm
	as written down in Fig.17 of the paper.

Short description.
------------------

	When a new packet arrives we calculate the average queue length:

	avg = (1-W)*avg + W*current_queue_len,

	W is the filter time constant (chosen as 2^(-Wlog)), it controls
	the inertia of the algorithm. To allow larger bursts, W should be
	decreased.

	if (avg > th_max) -> packet marked (dropped).
	if (avg < th_min) -> packet passes.
	if (th_min < avg < th_max) we calculate probability:

	Pb = max_P * (avg - th_min)/(th_max-th_min)

	and mark (drop) packet with this probability.
	Pb changes from 0 (at avg==th_min) to max_P (avg==th_max).
	max_P should be small (not 1), usually 0.01..0.02 is good value.

	max_P is chosen as a number, so that max_P/(th_max-th_min)
	is a negative power of two in order arithmetics to contain
	only shifts.


	Parameters, settable by user:
	-----------------------------

	limit		- bytes (must be > qth_max + burst)

	Hard limit on queue length, should be chosen >qth_max
	to allow packet bursts. This parameter does not
	affect the algorithms behaviour and can be chosen
	arbitrarily high (well, less than ram size)
	Really, this limit will never be reached
	if RED works correctly.

	qth_min		- bytes (should be < qth_max/2)
	qth_max		- bytes (should be at least 2*qth_min and less limit)
	Wlog	       	- bits (<32) log(1/W).
	Plog	       	- bits (<32)

	Plog is related to max_P by formula:

	max_P = (qth_max-qth_min)/2^Plog;

	F.e. if qth_max=128K and qth_min=32K, then Plog=22
	corresponds to max_P=0.02

	Scell_log
	Stab

	Lookup table for log((1-W)^(t/t_ave).


NOTES:

Upper bound on W.
-----------------

	If you want to allow bursts of L packets of size S,
	you should choose W:

	L + 1 - th_min/S < (1-(1-W)^L)/W

	th_min/S = 32         th_min/S = 4
			                       
	log(W)	L
	-1	33
	-2	35
	-3	39
	-4	46
	-5	57
	-6	75
	-7	101
	-8	135
	-9	190
	etc.
 */

struct red_sched_data
{
/* Parameters */
	u32		limit;		/* HARD maximal queue length	*/
	u32		qth_min;	/* Min average length threshold: A scaled */
	u32		qth_max;	/* Max average length threshold: A scaled */
	u32		Rmask;
	u32		Scell_max;
	unsigned char	flags;
	char		Wlog;		/* log(W)		*/
	char		Plog;		/* random number bits	*/
	char		Scell_log;
	u8		Stab[256];

/* Variables */
	unsigned long	qave;		/* Average queue length: A scaled */
	int		qcount;		/* Packets since last random number generation */
	u32		qR;		/* Cached random number */

	psched_time_t	qidlestart;	/* Start of idle period		*/
	struct tc_red_xstats st;
};

static int red_ecn_mark(struct sk_buff *skb)
{
	if (skb->nh.raw + 20 > skb->tail)
		return 0;

	switch (skb->protocol) {
	case __constant_htons(ETH_P_IP):
		if (INET_ECN_is_not_ect(skb->nh.iph->tos))
			return 0;
		IP_ECN_set_ce(skb->nh.iph);
		return 1;
	case __constant_htons(ETH_P_IPV6):
		if (INET_ECN_is_not_ect(ipv6_get_dsfield(skb->nh.ipv6h)))
			return 0;
		IP6_ECN_set_ce(skb->nh.ipv6h);
		return 1;
	default:
		return 0;
	}
}

static int
red_enqueue(struct sk_buff *skb, struct Qdisc* sch)
{
	struct red_sched_data *q = qdisc_priv(sch);

	psched_time_t now;

	if (!PSCHED_IS_PASTPERFECT(q->qidlestart)) {
		long us_idle;
		int  shift;

		PSCHED_GET_TIME(now);
		us_idle = PSCHED_TDIFF_SAFE(now, q->qidlestart, q->Scell_max);
		PSCHED_SET_PASTPERFECT(q->qidlestart);

/*
   The problem: ideally, average length queue recalcultion should
   be done over constant clock intervals. This is too expensive, so that
   the calculation is driven by outgoing packets.
   When the queue is idle we have to model this clock by hand.

   SF+VJ proposed to "generate" m = idletime/(average_pkt_size/bandwidth)
   dummy packets as a burst after idle time, i.e.

          q->qave *= (1-W)^m

   This is an apparently overcomplicated solution (f.e. we have to precompute
   a table to make this calculation in reasonable time)
   I believe that a simpler model may be used here,
   but it is field for experiments.
*/
		shift = q->Stab[us_idle>>q->Scell_log];

		if (shift) {
			q->qave >>= shift;
		} else {
			/* Approximate initial part of exponent
			   with linear function:
			   (1-W)^m ~= 1-mW + ...

			   Seems, it is the best solution to
			   problem of too coarce exponent tabulation.
			 */

			us_idle = (q->qave * us_idle)>>q->Scell_log;
			if (us_idle < q->qave/2)
				q->qave -= us_idle;
			else
				q->qave >>= 1;
		}
	} else {
		q->qave += sch->qstats.backlog - (q->qave >> q->Wlog);
		/* NOTE:
		   q->qave is fixed point number with point at Wlog.
		   The formulae above is equvalent to floating point
		   version:

		   qave = qave*(1-W) + sch->qstats.backlog*W;
		                                           --ANK (980924)
		 */
	}

	if (q->qave < q->qth_min) {
		q->qcount = -1;
enqueue:
		if (sch->qstats.backlog + skb->len <= q->limit) {
			__skb_queue_tail(&sch->q, skb);
			sch->qstats.backlog += skb->len;
			sch->bstats.bytes += skb->len;
			sch->bstats.packets++;
			return NET_XMIT_SUCCESS;
		} else {
			q->st.pdrop++;
		}
		kfree_skb(skb);
		sch->qstats.drops++;
		return NET_XMIT_DROP;
	}
	if (q->qave >= q->qth_max) {
		q->qcount = -1;
		sch->qstats.overlimits++;
mark:
		if  (!(q->flags&TC_RED_ECN) || !red_ecn_mark(skb)) {
			q->st.early++;
			goto drop;
		}
		q->st.marked++;
		goto enqueue;
	}

	if (++q->qcount) {
		/* The formula used below causes questions.

		   OK. qR is random number in the interval 0..Rmask
		   i.e. 0..(2^Plog). If we used floating point
		   arithmetics, it would be: (2^Plog)*rnd_num,
		   where rnd_num is less 1.

		   Taking into account, that qave have fixed
		   point at Wlog, and Plog is related to max_P by
		   max_P = (qth_max-qth_min)/2^Plog; two lines
		   below have the following floating point equivalent:
		   
		   max_P*(qave - qth_min)/(qth_max-qth_min) < rnd/qcount

		   Any questions? --ANK (980924)
		 */
		if (((q->qave - q->qth_min)>>q->Wlog)*q->qcount < q->qR)
			goto enqueue;
		q->qcount = 0;
		q->qR = net_random()&q->Rmask;
		sch->qstats.overlimits++;
		goto mark;
	}
	q->qR = net_random()&q->Rmask;
	goto enqueue;

drop:
	kfree_skb(skb);
	sch->qstats.drops++;
	return NET_XMIT_CN;
}

static int
red_requeue(struct sk_buff *skb, struct Qdisc* sch)
{
	struct red_sched_data *q = qdisc_priv(sch);

	PSCHED_SET_PASTPERFECT(q->qidlestart);

	__skb_queue_head(&sch->q, skb);
	sch->qstats.backlog += skb->len;
	sch->qstats.requeues++;
	return 0;
}

static struct sk_buff *
red_dequeue(struct Qdisc* sch)
{
	struct sk_buff *skb;
	struct red_sched_data *q = qdisc_priv(sch);

	skb = __skb_dequeue(&sch->q);
	if (skb) {
		sch->qstats.backlog -= skb->len;
		return skb;
	}
	PSCHED_GET_TIME(q->qidlestart);
	return NULL;
}

static unsigned int red_drop(struct Qdisc* sch)
{
	struct sk_buff *skb;
	struct red_sched_data *q = qdisc_priv(sch);

	skb = __skb_dequeue_tail(&sch->q);
	if (skb) {
		unsigned int len = skb->len;
		sch->qstats.backlog -= len;
		sch->qstats.drops++;
		q->st.other++;
		kfree_skb(skb);
		return len;
	}
	PSCHED_GET_TIME(q->qidlestart);
	return 0;
}

static void red_reset(struct Qdisc* sch)
{
	struct red_sched_data *q = qdisc_priv(sch);

	__skb_queue_purge(&sch->q);
	sch->qstats.backlog = 0;
	PSCHED_SET_PASTPERFECT(q->qidlestart);
	q->qave = 0;
	q->qcount = -1;
}

static int red_change(struct Qdisc *sch, struct rtattr *opt)
{
	struct red_sched_data *q = qdisc_priv(sch);
	struct rtattr *tb[TCA_RED_STAB];
	struct tc_red_qopt *ctl;

	if (opt == NULL ||
	    rtattr_parse_nested(tb, TCA_RED_STAB, opt) ||
	    tb[TCA_RED_PARMS-1] == 0 || tb[TCA_RED_STAB-1] == 0 ||
	    RTA_PAYLOAD(tb[TCA_RED_PARMS-1]) < sizeof(*ctl) ||
	    RTA_PAYLOAD(tb[TCA_RED_STAB-1]) < 256)
		return -EINVAL;

	ctl = RTA_DATA(tb[TCA_RED_PARMS-1]);

	sch_tree_lock(sch);
	q->flags = ctl->flags;
	q->Wlog = ctl->Wlog;
	q->Plog = ctl->Plog;
	q->Rmask = ctl->Plog < 32 ? ((1<<ctl->Plog) - 1) : ~0UL;
	q->Scell_log = ctl->Scell_log;
	q->Scell_max = (255<<q->Scell_log);
	q->qth_min = ctl->qth_min<<ctl->Wlog;
	q->qth_max = ctl->qth_max<<ctl->Wlog;
	q->limit = ctl->limit;
	memcpy(q->Stab, RTA_DATA(tb[TCA_RED_STAB-1]), 256);

	q->qcount = -1;
	if (skb_queue_empty(&sch->q))
		PSCHED_SET_PASTPERFECT(q->qidlestart);
	sch_tree_unlock(sch);
	return 0;
}

static int red_init(struct Qdisc* sch, struct rtattr *opt)
{
	return red_change(sch, opt);
}

static int red_dump(struct Qdisc *sch, struct sk_buff *skb)
{
	struct red_sched_data *q = qdisc_priv(sch);
	unsigned char	 *b = skb->tail;
	struct rtattr *rta;
	struct tc_red_qopt opt;

	rta = (struct rtattr*)b;
	RTA_PUT(skb, TCA_OPTIONS, 0, NULL);
	opt.limit = q->limit;
	opt.qth_min = q->qth_min>>q->Wlog;
	opt.qth_max = q->qth_max>>q->Wlog;
	opt.Wlog = q->Wlog;
	opt.Plog = q->Plog;
	opt.Scell_log = q->Scell_log;
	opt.flags = q->flags;
	RTA_PUT(skb, TCA_RED_PARMS, sizeof(opt), &opt);
	rta->rta_len = skb->tail - b;

	return skb->len;

rtattr_failure:
	skb_trim(skb, b - skb->data);
	return -1;
}

static int red_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
{
	struct red_sched_data *q = qdisc_priv(sch);

	return gnet_stats_copy_app(d, &q->st, sizeof(q->st));
}

static struct Qdisc_ops red_qdisc_ops = {
	.next		=	NULL,
	.cl_ops		=	NULL,
	.id		=	"red",
	.priv_size	=	sizeof(struct red_sched_data),
	.enqueue	=	red_enqueue,
	.dequeue	=	red_dequeue,
	.requeue	=	red_requeue,
	.drop		=	red_drop,
	.init		=	red_init,
	.reset		=	red_reset,
	.change		=	red_change,
	.dump		=	red_dump,
	.dump_stats	=	red_dump_stats,
	.owner		=	THIS_MODULE,
};

static int __init red_module_init(void)
{
	return register_qdisc(&red_qdisc_ops);
}
static void __exit red_module_exit(void) 
{
	unregister_qdisc(&red_qdisc_ops);
}
module_init(red_module_init)
module_exit(red_module_exit)
MODULE_LICENSE("GPL");
OpenPOWER on IntegriCloud