sch_fq_codel.c 18.9 KB
Newer Older
E
Eric Dumazet 已提交
1 2 3 4 5 6 7 8
/*
 * Fair Queue CoDel discipline
 *
 *	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.
 *
E
Eric Dumazet 已提交
9
 *  Copyright (C) 2012,2015 Eric Dumazet <edumazet@google.com>
E
Eric Dumazet 已提交
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
 */

#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/jiffies.h>
#include <linux/string.h>
#include <linux/in.h>
#include <linux/errno.h>
#include <linux/init.h>
#include <linux/skbuff.h>
#include <linux/jhash.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>
#include <net/netlink.h>
#include <net/pkt_sched.h>
26
#include <net/pkt_cls.h>
E
Eric Dumazet 已提交
27
#include <net/codel.h>
M
Michal Kazior 已提交
28 29
#include <net/codel_impl.h>
#include <net/codel_qdisc.h>
E
Eric Dumazet 已提交
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

/*	Fair Queue CoDel.
 *
 * Principles :
 * Packets are classified (internal classifier or external) on flows.
 * This is a Stochastic model (as we use a hash, several flows
 *			       might be hashed on same slot)
 * Each flow has a CoDel managed queue.
 * Flows are linked onto two (Round Robin) lists,
 * so that new flows have priority on old ones.
 *
 * For a given flow, packets are not reordered (CoDel uses a FIFO)
 * head drops only.
 * ECN capability is on by default.
 * Low memory footprint (64 bytes per flow)
 */

struct fq_codel_flow {
	struct sk_buff	  *head;
	struct sk_buff	  *tail;
	struct list_head  flowchain;
	int		  deficit;
	u32		  dropped; /* number of drops (or ECN marks) on this flow */
	struct codel_vars cvars;
}; /* please try to keep this structure <= 64 bytes */

struct fq_codel_sched_data {
J
John Fastabend 已提交
57
	struct tcf_proto __rcu *filter_list; /* optional external classifier */
58
	struct tcf_block *block;
E
Eric Dumazet 已提交
59 60 61 62
	struct fq_codel_flow *flows;	/* Flows table [flows_cnt] */
	u32		*backlogs;	/* backlog table [flows_cnt] */
	u32		flows_cnt;	/* number of flows */
	u32		quantum;	/* psched_mtu(qdisc_dev(sch)); */
63
	u32		drop_batch_size;
64
	u32		memory_limit;
E
Eric Dumazet 已提交
65 66
	struct codel_params cparams;
	struct codel_stats cstats;
67 68
	u32		memory_usage;
	u32		drop_overmemory;
E
Eric Dumazet 已提交
69 70 71 72 73 74 75 76
	u32		drop_overlimit;
	u32		new_flow_count;

	struct list_head new_flows;	/* list of new flows */
	struct list_head old_flows;	/* list of old flows */
};

static unsigned int fq_codel_hash(const struct fq_codel_sched_data *q,
77
				  struct sk_buff *skb)
E
Eric Dumazet 已提交
78
{
79
	return reciprocal_scale(skb_get_hash(skb), q->flows_cnt);
E
Eric Dumazet 已提交
80 81 82 83 84 85
}

static unsigned int fq_codel_classify(struct sk_buff *skb, struct Qdisc *sch,
				      int *qerr)
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
J
John Fastabend 已提交
86
	struct tcf_proto *filter;
E
Eric Dumazet 已提交
87 88 89 90 91 92 93 94
	struct tcf_result res;
	int result;

	if (TC_H_MAJ(skb->priority) == sch->handle &&
	    TC_H_MIN(skb->priority) > 0 &&
	    TC_H_MIN(skb->priority) <= q->flows_cnt)
		return TC_H_MIN(skb->priority);

95
	filter = rcu_dereference_bh(q->filter_list);
J
John Fastabend 已提交
96
	if (!filter)
E
Eric Dumazet 已提交
97 98 99
		return fq_codel_hash(q, skb) + 1;

	*qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
100
	result = tcf_classify(skb, filter, &res, false);
E
Eric Dumazet 已提交
101 102 103 104 105
	if (result >= 0) {
#ifdef CONFIG_NET_CLS_ACT
		switch (result) {
		case TC_ACT_STOLEN:
		case TC_ACT_QUEUED:
106
		case TC_ACT_TRAP:
E
Eric Dumazet 已提交
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
			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
		case TC_ACT_SHOT:
			return 0;
		}
#endif
		if (TC_H_MIN(res.classid) <= q->flows_cnt)
			return TC_H_MIN(res.classid);
	}
	return 0;
}

/* helper functions : might be changed when/if skb use a standard list_head */

/* remove one skb from head of slot queue */
static inline struct sk_buff *dequeue_head(struct fq_codel_flow *flow)
{
	struct sk_buff *skb = flow->head;

	flow->head = skb->next;
	skb->next = NULL;
	return skb;
}

/* add skb to flow queue (tail add) */
static inline void flow_queue_add(struct fq_codel_flow *flow,
				  struct sk_buff *skb)
{
	if (flow->head == NULL)
		flow->head = skb;
	else
		flow->tail->next = skb;
	flow->tail = skb;
	skb->next = NULL;
}

142 143
static unsigned int fq_codel_drop(struct Qdisc *sch, unsigned int max_packets,
				  struct sk_buff **to_free)
E
Eric Dumazet 已提交
144 145 146 147 148
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	struct sk_buff *skb;
	unsigned int maxbacklog = 0, idx = 0, i, len;
	struct fq_codel_flow *flow;
149
	unsigned int threshold;
150
	unsigned int mem = 0;
E
Eric Dumazet 已提交
151

152
	/* Queue is full! Find the fat flow and drop packet(s) from it.
E
Eric Dumazet 已提交
153 154 155
	 * This might sound expensive, but with 1024 flows, we scan
	 * 4KB of memory, and we dont need to handle a complex tree
	 * in fast path (packet queue/enqueue) with many cache misses.
156 157
	 * In stress mode, we'll try to drop 64 packets from the flow,
	 * amortizing this linear lookup to one cache line per drop.
E
Eric Dumazet 已提交
158 159 160 161 162 163 164
	 */
	for (i = 0; i < q->flows_cnt; i++) {
		if (q->backlogs[i] > maxbacklog) {
			maxbacklog = q->backlogs[i];
			idx = i;
		}
	}
165 166 167 168

	/* Our goal is to drop half of this fat flow backlog */
	threshold = maxbacklog >> 1;

E
Eric Dumazet 已提交
169
	flow = &q->flows[idx];
170 171 172 173 174
	len = 0;
	i = 0;
	do {
		skb = dequeue_head(flow);
		len += qdisc_pkt_len(skb);
175
		mem += get_codel_cb(skb)->mem_usage;
176
		__qdisc_drop(skb, to_free);
177 178 179
	} while (++i < max_packets && len < threshold);

	flow->dropped += i;
E
Eric Dumazet 已提交
180
	q->backlogs[idx] -= len;
181
	q->memory_usage -= mem;
182 183 184
	sch->qstats.drops += i;
	sch->qstats.backlog -= len;
	sch->q.qlen -= i;
E
Eric Dumazet 已提交
185 186 187
	return idx;
}

188 189
static int fq_codel_enqueue(struct sk_buff *skb, struct Qdisc *sch,
			    struct sk_buff **to_free)
E
Eric Dumazet 已提交
190 191
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
192
	unsigned int idx, prev_backlog, prev_qlen;
E
Eric Dumazet 已提交
193 194
	struct fq_codel_flow *flow;
	int uninitialized_var(ret);
195
	unsigned int pkt_len;
196
	bool memory_limited;
E
Eric Dumazet 已提交
197 198 199 200

	idx = fq_codel_classify(skb, sch, &ret);
	if (idx == 0) {
		if (ret & __NET_XMIT_BYPASS)
201
			qdisc_qstats_drop(sch);
202
		__qdisc_drop(skb, to_free);
E
Eric Dumazet 已提交
203 204 205 206 207 208 209 210
		return ret;
	}
	idx--;

	codel_set_enqueue_time(skb);
	flow = &q->flows[idx];
	flow_queue_add(flow, skb);
	q->backlogs[idx] += qdisc_pkt_len(skb);
211
	qdisc_qstats_backlog_inc(sch, skb);
E
Eric Dumazet 已提交
212 213 214 215 216 217 218

	if (list_empty(&flow->flowchain)) {
		list_add_tail(&flow->flowchain, &q->new_flows);
		q->new_flow_count++;
		flow->deficit = q->quantum;
		flow->dropped = 0;
	}
219 220
	get_codel_cb(skb)->mem_usage = skb->truesize;
	q->memory_usage += get_codel_cb(skb)->mem_usage;
221 222
	memory_limited = q->memory_usage > q->memory_limit;
	if (++sch->q.qlen <= sch->limit && !memory_limited)
E
Eric Dumazet 已提交
223 224
		return NET_XMIT_SUCCESS;

225
	prev_backlog = sch->qstats.backlog;
226 227
	prev_qlen = sch->q.qlen;

228 229
	/* save this packet length as it might be dropped by fq_codel_drop() */
	pkt_len = qdisc_pkt_len(skb);
230 231 232 233
	/* fq_codel_drop() is quite expensive, as it performs a linear search
	 * in q->backlogs[] to find a fat flow.
	 * So instead of dropping a single packet, drop half of its backlog
	 * with a 64 packets limit to not add a too big cpu spike here.
E
Eric Dumazet 已提交
234
	 */
235
	ret = fq_codel_drop(sch, q->drop_batch_size, to_free);
236

237 238 239
	prev_qlen -= sch->q.qlen;
	prev_backlog -= sch->qstats.backlog;
	q->drop_overlimit += prev_qlen;
240
	if (memory_limited)
241
		q->drop_overmemory += prev_qlen;
242

243 244 245 246 247 248 249 250 251 252 253
	/* As we dropped packet(s), better let upper stack know this.
	 * If we dropped a packet for this flow, return NET_XMIT_CN,
	 * but in this case, our parents wont increase their backlogs.
	 */
	if (ret == idx) {
		qdisc_tree_reduce_backlog(sch, prev_qlen - 1,
					  prev_backlog - pkt_len);
		return NET_XMIT_CN;
	}
	qdisc_tree_reduce_backlog(sch, prev_qlen, prev_backlog);
	return NET_XMIT_SUCCESS;
E
Eric Dumazet 已提交
254 255 256 257 258 259
}

/* This is the specific function called from codel_dequeue()
 * to dequeue a packet from queue. Note: backlog is handled in
 * codel, we dont need to reduce it here.
 */
260
static struct sk_buff *dequeue_func(struct codel_vars *vars, void *ctx)
E
Eric Dumazet 已提交
261
{
262
	struct Qdisc *sch = ctx;
263
	struct fq_codel_sched_data *q = qdisc_priv(sch);
E
Eric Dumazet 已提交
264 265 266 267 268 269
	struct fq_codel_flow *flow;
	struct sk_buff *skb = NULL;

	flow = container_of(vars, struct fq_codel_flow, cvars);
	if (flow->head) {
		skb = dequeue_head(flow);
270
		q->backlogs[flow - q->flows] -= qdisc_pkt_len(skb);
271
		q->memory_usage -= get_codel_cb(skb)->mem_usage;
E
Eric Dumazet 已提交
272
		sch->q.qlen--;
273
		sch->qstats.backlog -= qdisc_pkt_len(skb);
E
Eric Dumazet 已提交
274 275 276 277
	}
	return skb;
}

278 279 280 281
static void drop_func(struct sk_buff *skb, void *ctx)
{
	struct Qdisc *sch = ctx;

282 283
	kfree_skb(skb);
	qdisc_qstats_drop(sch);
284 285
}

E
Eric Dumazet 已提交
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
static struct sk_buff *fq_codel_dequeue(struct Qdisc *sch)
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	struct sk_buff *skb;
	struct fq_codel_flow *flow;
	struct list_head *head;
	u32 prev_drop_count, prev_ecn_mark;

begin:
	head = &q->new_flows;
	if (list_empty(head)) {
		head = &q->old_flows;
		if (list_empty(head))
			return NULL;
	}
	flow = list_first_entry(head, struct fq_codel_flow, flowchain);

	if (flow->deficit <= 0) {
		flow->deficit += q->quantum;
		list_move_tail(&flow->flowchain, &q->old_flows);
		goto begin;
	}

	prev_drop_count = q->cstats.drop_count;
	prev_ecn_mark = q->cstats.ecn_mark;

312 313 314
	skb = codel_dequeue(sch, &sch->qstats.backlog, &q->cparams,
			    &flow->cvars, &q->cstats, qdisc_pkt_len,
			    codel_get_enqueue_time, drop_func, dequeue_func);
E
Eric Dumazet 已提交
315 316 317 318 319 320 321 322 323 324 325 326 327 328

	flow->dropped += q->cstats.drop_count - prev_drop_count;
	flow->dropped += q->cstats.ecn_mark - prev_ecn_mark;

	if (!skb) {
		/* force a pass through old_flows to prevent starvation */
		if ((head == &q->new_flows) && !list_empty(&q->old_flows))
			list_move_tail(&flow->flowchain, &q->old_flows);
		else
			list_del_init(&flow->flowchain);
		goto begin;
	}
	qdisc_bstats_update(sch, skb);
	flow->deficit -= qdisc_pkt_len(skb);
329
	/* We cant call qdisc_tree_reduce_backlog() if our qlen is 0,
E
Eric Dumazet 已提交
330 331 332
	 * or HTB crashes. Defer it for next round.
	 */
	if (q->cstats.drop_count && sch->q.qlen) {
333 334
		qdisc_tree_reduce_backlog(sch, q->cstats.drop_count,
					  q->cstats.drop_len);
E
Eric Dumazet 已提交
335
		q->cstats.drop_count = 0;
336
		q->cstats.drop_len = 0;
E
Eric Dumazet 已提交
337 338 339 340
	}
	return skb;
}

341 342 343 344 345 346
static void fq_codel_flow_purge(struct fq_codel_flow *flow)
{
	rtnl_kfree_skbs(flow->head, flow->tail);
	flow->head = NULL;
}

E
Eric Dumazet 已提交
347 348
static void fq_codel_reset(struct Qdisc *sch)
{
349 350
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	int i;
E
Eric Dumazet 已提交
351

352 353 354 355 356
	INIT_LIST_HEAD(&q->new_flows);
	INIT_LIST_HEAD(&q->old_flows);
	for (i = 0; i < q->flows_cnt; i++) {
		struct fq_codel_flow *flow = q->flows + i;

357
		fq_codel_flow_purge(flow);
358 359 360 361 362
		INIT_LIST_HEAD(&flow->flowchain);
		codel_vars_init(&flow->cvars);
	}
	memset(q->backlogs, 0, q->flows_cnt * sizeof(u32));
	sch->q.qlen = 0;
363
	sch->qstats.backlog = 0;
364
	q->memory_usage = 0;
E
Eric Dumazet 已提交
365 366 367 368 369 370 371 372 373
}

static const struct nla_policy fq_codel_policy[TCA_FQ_CODEL_MAX + 1] = {
	[TCA_FQ_CODEL_TARGET]	= { .type = NLA_U32 },
	[TCA_FQ_CODEL_LIMIT]	= { .type = NLA_U32 },
	[TCA_FQ_CODEL_INTERVAL]	= { .type = NLA_U32 },
	[TCA_FQ_CODEL_ECN]	= { .type = NLA_U32 },
	[TCA_FQ_CODEL_FLOWS]	= { .type = NLA_U32 },
	[TCA_FQ_CODEL_QUANTUM]	= { .type = NLA_U32 },
E
Eric Dumazet 已提交
374
	[TCA_FQ_CODEL_CE_THRESHOLD] = { .type = NLA_U32 },
375
	[TCA_FQ_CODEL_DROP_BATCH_SIZE] = { .type = NLA_U32 },
376
	[TCA_FQ_CODEL_MEMORY_LIMIT] = { .type = NLA_U32 },
E
Eric Dumazet 已提交
377 378 379 380 381 382 383 384 385 386 387
};

static int fq_codel_change(struct Qdisc *sch, struct nlattr *opt)
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	struct nlattr *tb[TCA_FQ_CODEL_MAX + 1];
	int err;

	if (!opt)
		return -EINVAL;

388 389
	err = nla_parse_nested(tb, TCA_FQ_CODEL_MAX, opt, fq_codel_policy,
			       NULL);
E
Eric Dumazet 已提交
390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407
	if (err < 0)
		return err;
	if (tb[TCA_FQ_CODEL_FLOWS]) {
		if (q->flows)
			return -EINVAL;
		q->flows_cnt = nla_get_u32(tb[TCA_FQ_CODEL_FLOWS]);
		if (!q->flows_cnt ||
		    q->flows_cnt > 65536)
			return -EINVAL;
	}
	sch_tree_lock(sch);

	if (tb[TCA_FQ_CODEL_TARGET]) {
		u64 target = nla_get_u32(tb[TCA_FQ_CODEL_TARGET]);

		q->cparams.target = (target * NSEC_PER_USEC) >> CODEL_SHIFT;
	}

E
Eric Dumazet 已提交
408 409 410 411 412 413
	if (tb[TCA_FQ_CODEL_CE_THRESHOLD]) {
		u64 val = nla_get_u32(tb[TCA_FQ_CODEL_CE_THRESHOLD]);

		q->cparams.ce_threshold = (val * NSEC_PER_USEC) >> CODEL_SHIFT;
	}

E
Eric Dumazet 已提交
414 415 416 417 418 419 420 421 422 423 424 425 426 427 428
	if (tb[TCA_FQ_CODEL_INTERVAL]) {
		u64 interval = nla_get_u32(tb[TCA_FQ_CODEL_INTERVAL]);

		q->cparams.interval = (interval * NSEC_PER_USEC) >> CODEL_SHIFT;
	}

	if (tb[TCA_FQ_CODEL_LIMIT])
		sch->limit = nla_get_u32(tb[TCA_FQ_CODEL_LIMIT]);

	if (tb[TCA_FQ_CODEL_ECN])
		q->cparams.ecn = !!nla_get_u32(tb[TCA_FQ_CODEL_ECN]);

	if (tb[TCA_FQ_CODEL_QUANTUM])
		q->quantum = max(256U, nla_get_u32(tb[TCA_FQ_CODEL_QUANTUM]));

429 430 431
	if (tb[TCA_FQ_CODEL_DROP_BATCH_SIZE])
		q->drop_batch_size = min(1U, nla_get_u32(tb[TCA_FQ_CODEL_DROP_BATCH_SIZE]));

432 433 434 435 436
	if (tb[TCA_FQ_CODEL_MEMORY_LIMIT])
		q->memory_limit = min(1U << 31, nla_get_u32(tb[TCA_FQ_CODEL_MEMORY_LIMIT]));

	while (sch->q.qlen > sch->limit ||
	       q->memory_usage > q->memory_limit) {
E
Eric Dumazet 已提交
437 438
		struct sk_buff *skb = fq_codel_dequeue(sch);

439
		q->cstats.drop_len += qdisc_pkt_len(skb);
440
		rtnl_kfree_skbs(skb, skb);
E
Eric Dumazet 已提交
441 442
		q->cstats.drop_count++;
	}
443
	qdisc_tree_reduce_backlog(sch, q->cstats.drop_count, q->cstats.drop_len);
E
Eric Dumazet 已提交
444
	q->cstats.drop_count = 0;
445
	q->cstats.drop_len = 0;
E
Eric Dumazet 已提交
446 447 448 449 450 451 452 453 454

	sch_tree_unlock(sch);
	return 0;
}

static void fq_codel_destroy(struct Qdisc *sch)
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);

455
	tcf_block_put(q->block);
456 457
	kvfree(q->backlogs);
	kvfree(q->flows);
E
Eric Dumazet 已提交
458 459 460 461 462 463
}

static int fq_codel_init(struct Qdisc *sch, struct nlattr *opt)
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	int i;
464
	int err;
E
Eric Dumazet 已提交
465 466 467

	sch->limit = 10*1024;
	q->flows_cnt = 1024;
468
	q->memory_limit = 32 << 20; /* 32 MBytes */
469
	q->drop_batch_size = 64;
E
Eric Dumazet 已提交
470 471 472
	q->quantum = psched_mtu(qdisc_dev(sch));
	INIT_LIST_HEAD(&q->new_flows);
	INIT_LIST_HEAD(&q->old_flows);
473
	codel_params_init(&q->cparams);
E
Eric Dumazet 已提交
474 475
	codel_stats_init(&q->cstats);
	q->cparams.ecn = true;
476
	q->cparams.mtu = psched_mtu(qdisc_dev(sch));
E
Eric Dumazet 已提交
477 478 479 480 481 482 483

	if (opt) {
		int err = fq_codel_change(sch, opt);
		if (err)
			return err;
	}

484
	err = tcf_block_get(&q->block, &q->filter_list, sch);
485 486 487
	if (err)
		return err;

E
Eric Dumazet 已提交
488
	if (!q->flows) {
489 490
		q->flows = kvzalloc(q->flows_cnt *
					   sizeof(struct fq_codel_flow), GFP_KERNEL);
E
Eric Dumazet 已提交
491 492
		if (!q->flows)
			return -ENOMEM;
493
		q->backlogs = kvzalloc(q->flows_cnt * sizeof(u32), GFP_KERNEL);
494
		if (!q->backlogs)
E
Eric Dumazet 已提交
495 496 497 498 499
			return -ENOMEM;
		for (i = 0; i < q->flows_cnt; i++) {
			struct fq_codel_flow *flow = q->flows + i;

			INIT_LIST_HEAD(&flow->flowchain);
E
Eric Dumazet 已提交
500
			codel_vars_init(&flow->cvars);
E
Eric Dumazet 已提交
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
		}
	}
	if (sch->limit >= 1)
		sch->flags |= TCQ_F_CAN_BYPASS;
	else
		sch->flags &= ~TCQ_F_CAN_BYPASS;
	return 0;
}

static int fq_codel_dump(struct Qdisc *sch, struct sk_buff *skb)
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	struct nlattr *opts;

	opts = nla_nest_start(skb, TCA_OPTIONS);
	if (opts == NULL)
		goto nla_put_failure;

	if (nla_put_u32(skb, TCA_FQ_CODEL_TARGET,
			codel_time_to_us(q->cparams.target)) ||
	    nla_put_u32(skb, TCA_FQ_CODEL_LIMIT,
			sch->limit) ||
	    nla_put_u32(skb, TCA_FQ_CODEL_INTERVAL,
			codel_time_to_us(q->cparams.interval)) ||
	    nla_put_u32(skb, TCA_FQ_CODEL_ECN,
			q->cparams.ecn) ||
	    nla_put_u32(skb, TCA_FQ_CODEL_QUANTUM,
			q->quantum) ||
529 530
	    nla_put_u32(skb, TCA_FQ_CODEL_DROP_BATCH_SIZE,
			q->drop_batch_size) ||
531 532
	    nla_put_u32(skb, TCA_FQ_CODEL_MEMORY_LIMIT,
			q->memory_limit) ||
E
Eric Dumazet 已提交
533 534 535 536
	    nla_put_u32(skb, TCA_FQ_CODEL_FLOWS,
			q->flows_cnt))
		goto nla_put_failure;

E
Eric Dumazet 已提交
537 538 539 540 541
	if (q->cparams.ce_threshold != CODEL_DISABLED_THRESHOLD &&
	    nla_put_u32(skb, TCA_FQ_CODEL_CE_THRESHOLD,
			codel_time_to_us(q->cparams.ce_threshold)))
		goto nla_put_failure;

542
	return nla_nest_end(skb, opts);
E
Eric Dumazet 已提交
543 544 545 546 547 548 549 550 551 552 553 554 555

nla_put_failure:
	return -1;
}

static int fq_codel_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	struct tc_fq_codel_xstats st = {
		.type				= TCA_FQ_CODEL_XSTATS_QDISC,
	};
	struct list_head *pos;

S
Sasha Levin 已提交
556 557 558 559
	st.qdisc_stats.maxpacket = q->cstats.maxpacket;
	st.qdisc_stats.drop_overlimit = q->drop_overlimit;
	st.qdisc_stats.ecn_mark = q->cstats.ecn_mark;
	st.qdisc_stats.new_flow_count = q->new_flow_count;
E
Eric Dumazet 已提交
560
	st.qdisc_stats.ce_mark = q->cstats.ce_mark;
561 562
	st.qdisc_stats.memory_usage  = q->memory_usage;
	st.qdisc_stats.drop_overmemory = q->drop_overmemory;
S
Sasha Levin 已提交
563

564
	sch_tree_lock(sch);
E
Eric Dumazet 已提交
565 566 567 568 569
	list_for_each(pos, &q->new_flows)
		st.qdisc_stats.new_flows_len++;

	list_for_each(pos, &q->old_flows)
		st.qdisc_stats.old_flows_len++;
570
	sch_tree_unlock(sch);
E
Eric Dumazet 已提交
571 572 573 574 575 576 577 578 579

	return gnet_stats_copy_app(d, &st, sizeof(st));
}

static struct Qdisc *fq_codel_leaf(struct Qdisc *sch, unsigned long arg)
{
	return NULL;
}

580
static unsigned long fq_codel_find(struct Qdisc *sch, u32 classid)
E
Eric Dumazet 已提交
581 582 583 584 585 586 587 588 589 590 591 592
{
	return 0;
}

static unsigned long fq_codel_bind(struct Qdisc *sch, unsigned long parent,
			      u32 classid)
{
	/* we cannot bypass queue discipline anymore */
	sch->flags &= ~TCQ_F_CAN_BYPASS;
	return 0;
}

593
static void fq_codel_unbind(struct Qdisc *q, unsigned long cl)
E
Eric Dumazet 已提交
594 595 596
{
}

597
static struct tcf_block *fq_codel_tcf_block(struct Qdisc *sch, unsigned long cl)
E
Eric Dumazet 已提交
598 599 600 601 602
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);

	if (cl)
		return NULL;
603
	return q->block;
E
Eric Dumazet 已提交
604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622
}

static int fq_codel_dump_class(struct Qdisc *sch, unsigned long cl,
			  struct sk_buff *skb, struct tcmsg *tcm)
{
	tcm->tcm_handle |= TC_H_MIN(cl);
	return 0;
}

static int fq_codel_dump_class_stats(struct Qdisc *sch, unsigned long cl,
				     struct gnet_dump *d)
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	u32 idx = cl - 1;
	struct gnet_stats_queue qs = { 0 };
	struct tc_fq_codel_xstats xstats;

	if (idx < q->flows_cnt) {
		const struct fq_codel_flow *flow = &q->flows[idx];
623
		const struct sk_buff *skb;
E
Eric Dumazet 已提交
624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640

		memset(&xstats, 0, sizeof(xstats));
		xstats.type = TCA_FQ_CODEL_XSTATS_CLASS;
		xstats.class_stats.deficit = flow->deficit;
		xstats.class_stats.ldelay =
			codel_time_to_us(flow->cvars.ldelay);
		xstats.class_stats.count = flow->cvars.count;
		xstats.class_stats.lastcount = flow->cvars.lastcount;
		xstats.class_stats.dropping = flow->cvars.dropping;
		if (flow->cvars.dropping) {
			codel_tdiff_t delta = flow->cvars.drop_next -
					      codel_get_time();

			xstats.class_stats.drop_next = (delta >= 0) ?
				codel_time_to_us(delta) :
				-codel_time_to_us(-delta);
		}
641 642 643 644 645 646 647 648
		if (flow->head) {
			sch_tree_lock(sch);
			skb = flow->head;
			while (skb) {
				qs.qlen++;
				skb = skb->next;
			}
			sch_tree_unlock(sch);
E
Eric Dumazet 已提交
649 650 651 652
		}
		qs.backlog = q->backlogs[idx];
		qs.drops = flow->dropped;
	}
653
	if (gnet_stats_copy_queue(d, NULL, &qs, qs.qlen) < 0)
E
Eric Dumazet 已提交
654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683
		return -1;
	if (idx < q->flows_cnt)
		return gnet_stats_copy_app(d, &xstats, sizeof(xstats));
	return 0;
}

static void fq_codel_walk(struct Qdisc *sch, struct qdisc_walker *arg)
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	unsigned int i;

	if (arg->stop)
		return;

	for (i = 0; i < q->flows_cnt; i++) {
		if (list_empty(&q->flows[i].flowchain) ||
		    arg->count < arg->skip) {
			arg->count++;
			continue;
		}
		if (arg->fn(sch, i + 1, arg) < 0) {
			arg->stop = 1;
			break;
		}
		arg->count++;
	}
}

static const struct Qdisc_class_ops fq_codel_class_ops = {
	.leaf		=	fq_codel_leaf,
684
	.find		=	fq_codel_find,
685
	.tcf_block	=	fq_codel_tcf_block,
E
Eric Dumazet 已提交
686
	.bind_tcf	=	fq_codel_bind,
687
	.unbind_tcf	=	fq_codel_unbind,
E
Eric Dumazet 已提交
688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722
	.dump		=	fq_codel_dump_class,
	.dump_stats	=	fq_codel_dump_class_stats,
	.walk		=	fq_codel_walk,
};

static struct Qdisc_ops fq_codel_qdisc_ops __read_mostly = {
	.cl_ops		=	&fq_codel_class_ops,
	.id		=	"fq_codel",
	.priv_size	=	sizeof(struct fq_codel_sched_data),
	.enqueue	=	fq_codel_enqueue,
	.dequeue	=	fq_codel_dequeue,
	.peek		=	qdisc_peek_dequeued,
	.init		=	fq_codel_init,
	.reset		=	fq_codel_reset,
	.destroy	=	fq_codel_destroy,
	.change		=	fq_codel_change,
	.dump		=	fq_codel_dump,
	.dump_stats =	fq_codel_dump_stats,
	.owner		=	THIS_MODULE,
};

static int __init fq_codel_module_init(void)
{
	return register_qdisc(&fq_codel_qdisc_ops);
}

static void __exit fq_codel_module_exit(void)
{
	unregister_qdisc(&fq_codel_qdisc_ops);
}

module_init(fq_codel_module_init)
module_exit(fq_codel_module_exit)
MODULE_AUTHOR("Eric Dumazet");
MODULE_LICENSE("GPL");