sch_fq_codel.c 19.0 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
			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
108
			/* fall through */
E
Eric Dumazet 已提交
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
		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;
}

143 144
static unsigned int fq_codel_drop(struct Qdisc *sch, unsigned int max_packets,
				  struct sk_buff **to_free)
E
Eric Dumazet 已提交
145 146 147 148 149
{
	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;
150
	unsigned int threshold;
151
	unsigned int mem = 0;
E
Eric Dumazet 已提交
152

153
	/* Queue is full! Find the fat flow and drop packet(s) from it.
E
Eric Dumazet 已提交
154 155 156
	 * 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.
157 158
	 * 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
159 160 161 162 163 164 165
	 */
	for (i = 0; i < q->flows_cnt; i++) {
		if (q->backlogs[i] > maxbacklog) {
			maxbacklog = q->backlogs[i];
			idx = i;
		}
	}
166 167 168 169

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

E
Eric Dumazet 已提交
170
	flow = &q->flows[idx];
171 172 173 174 175
	len = 0;
	i = 0;
	do {
		skb = dequeue_head(flow);
		len += qdisc_pkt_len(skb);
176
		mem += get_codel_cb(skb)->mem_usage;
177
		__qdisc_drop(skb, to_free);
178 179 180
	} while (++i < max_packets && len < threshold);

	flow->dropped += i;
E
Eric Dumazet 已提交
181
	q->backlogs[idx] -= len;
182
	q->memory_usage -= mem;
183 184 185
	sch->qstats.drops += i;
	sch->qstats.backlog -= len;
	sch->q.qlen -= i;
E
Eric Dumazet 已提交
186 187 188
	return idx;
}

189 190
static int fq_codel_enqueue(struct sk_buff *skb, struct Qdisc *sch,
			    struct sk_buff **to_free)
E
Eric Dumazet 已提交
191 192
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
193
	unsigned int idx, prev_backlog, prev_qlen;
E
Eric Dumazet 已提交
194 195
	struct fq_codel_flow *flow;
	int uninitialized_var(ret);
196
	unsigned int pkt_len;
197
	bool memory_limited;
E
Eric Dumazet 已提交
198 199 200 201

	idx = fq_codel_classify(skb, sch, &ret);
	if (idx == 0) {
		if (ret & __NET_XMIT_BYPASS)
202
			qdisc_qstats_drop(sch);
203
		__qdisc_drop(skb, to_free);
E
Eric Dumazet 已提交
204 205 206 207 208 209 210 211
		return ret;
	}
	idx--;

	codel_set_enqueue_time(skb);
	flow = &q->flows[idx];
	flow_queue_add(flow, skb);
	q->backlogs[idx] += qdisc_pkt_len(skb);
212
	qdisc_qstats_backlog_inc(sch, skb);
E
Eric Dumazet 已提交
213 214 215 216 217 218 219

	if (list_empty(&flow->flowchain)) {
		list_add_tail(&flow->flowchain, &q->new_flows);
		q->new_flow_count++;
		flow->deficit = q->quantum;
		flow->dropped = 0;
	}
220 221
	get_codel_cb(skb)->mem_usage = skb->truesize;
	q->memory_usage += get_codel_cb(skb)->mem_usage;
222 223
	memory_limited = q->memory_usage > q->memory_limit;
	if (++sch->q.qlen <= sch->limit && !memory_limited)
E
Eric Dumazet 已提交
224 225
		return NET_XMIT_SUCCESS;

226
	prev_backlog = sch->qstats.backlog;
227 228
	prev_qlen = sch->q.qlen;

229 230
	/* save this packet length as it might be dropped by fq_codel_drop() */
	pkt_len = qdisc_pkt_len(skb);
231 232 233 234
	/* 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
235
	 */
236
	ret = fq_codel_drop(sch, q->drop_batch_size, to_free);
237

238 239 240
	prev_qlen -= sch->q.qlen;
	prev_backlog -= sch->qstats.backlog;
	q->drop_overlimit += prev_qlen;
241
	if (memory_limited)
242
		q->drop_overmemory += prev_qlen;
243

244 245 246 247 248 249 250 251 252 253 254
	/* 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
255 256 257 258 259 260
}

/* 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.
 */
261
static struct sk_buff *dequeue_func(struct codel_vars *vars, void *ctx)
E
Eric Dumazet 已提交
262
{
263
	struct Qdisc *sch = ctx;
264
	struct fq_codel_sched_data *q = qdisc_priv(sch);
E
Eric Dumazet 已提交
265 266 267 268 269 270
	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);
271
		q->backlogs[flow - q->flows] -= qdisc_pkt_len(skb);
272
		q->memory_usage -= get_codel_cb(skb)->mem_usage;
E
Eric Dumazet 已提交
273
		sch->q.qlen--;
274
		sch->qstats.backlog -= qdisc_pkt_len(skb);
E
Eric Dumazet 已提交
275 276 277 278
	}
	return skb;
}

279 280 281 282
static void drop_func(struct sk_buff *skb, void *ctx)
{
	struct Qdisc *sch = ctx;

283 284
	kfree_skb(skb);
	qdisc_qstats_drop(sch);
285 286
}

E
Eric Dumazet 已提交
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
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;

313 314 315
	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
316 317 318 319 320 321 322 323 324 325 326 327 328 329

	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);
330
	/* We cant call qdisc_tree_reduce_backlog() if our qlen is 0,
E
Eric Dumazet 已提交
331 332 333
	 * or HTB crashes. Defer it for next round.
	 */
	if (q->cstats.drop_count && sch->q.qlen) {
334 335
		qdisc_tree_reduce_backlog(sch, q->cstats.drop_count,
					  q->cstats.drop_len);
E
Eric Dumazet 已提交
336
		q->cstats.drop_count = 0;
337
		q->cstats.drop_len = 0;
E
Eric Dumazet 已提交
338 339 340 341
	}
	return skb;
}

342 343 344 345 346 347
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
348 349
static void fq_codel_reset(struct Qdisc *sch)
{
350 351
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	int i;
E
Eric Dumazet 已提交
352

353 354 355 356 357
	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;

358
		fq_codel_flow_purge(flow);
359 360 361 362 363
		INIT_LIST_HEAD(&flow->flowchain);
		codel_vars_init(&flow->cvars);
	}
	memset(q->backlogs, 0, q->flows_cnt * sizeof(u32));
	sch->q.qlen = 0;
364
	sch->qstats.backlog = 0;
365
	q->memory_usage = 0;
E
Eric Dumazet 已提交
366 367 368 369 370 371 372 373 374
}

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
375
	[TCA_FQ_CODEL_CE_THRESHOLD] = { .type = NLA_U32 },
376
	[TCA_FQ_CODEL_DROP_BATCH_SIZE] = { .type = NLA_U32 },
377
	[TCA_FQ_CODEL_MEMORY_LIMIT] = { .type = NLA_U32 },
E
Eric Dumazet 已提交
378 379
};

380 381
static int fq_codel_change(struct Qdisc *sch, struct nlattr *opt,
			   struct netlink_ext_ack *extack)
E
Eric Dumazet 已提交
382 383 384 385 386 387 388 389
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	struct nlattr *tb[TCA_FQ_CODEL_MAX + 1];
	int err;

	if (!opt)
		return -EINVAL;

390 391
	err = nla_parse_nested(tb, TCA_FQ_CODEL_MAX, opt, fq_codel_policy,
			       NULL);
E
Eric Dumazet 已提交
392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409
	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
410 411 412 413 414 415
	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
416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
	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]));

431 432 433
	if (tb[TCA_FQ_CODEL_DROP_BATCH_SIZE])
		q->drop_batch_size = min(1U, nla_get_u32(tb[TCA_FQ_CODEL_DROP_BATCH_SIZE]));

434 435 436 437 438
	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
439 440
		struct sk_buff *skb = fq_codel_dequeue(sch);

441
		q->cstats.drop_len += qdisc_pkt_len(skb);
442
		rtnl_kfree_skbs(skb, skb);
E
Eric Dumazet 已提交
443 444
		q->cstats.drop_count++;
	}
445
	qdisc_tree_reduce_backlog(sch, q->cstats.drop_count, q->cstats.drop_len);
E
Eric Dumazet 已提交
446
	q->cstats.drop_count = 0;
447
	q->cstats.drop_len = 0;
E
Eric Dumazet 已提交
448 449 450 451 452 453 454 455 456

	sch_tree_unlock(sch);
	return 0;
}

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

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

462 463
static int fq_codel_init(struct Qdisc *sch, struct nlattr *opt,
			 struct netlink_ext_ack *extack)
E
Eric Dumazet 已提交
464 465 466
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	int i;
467
	int err;
E
Eric Dumazet 已提交
468 469 470

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

	if (opt) {
482
		int err = fq_codel_change(sch, opt, extack);
E
Eric Dumazet 已提交
483 484 485 486
		if (err)
			return err;
	}

487
	err = tcf_block_get(&q->block, &q->filter_list, sch);
488 489 490
	if (err)
		return err;

E
Eric Dumazet 已提交
491
	if (!q->flows) {
492 493
		q->flows = kvzalloc(q->flows_cnt *
					   sizeof(struct fq_codel_flow), GFP_KERNEL);
E
Eric Dumazet 已提交
494 495
		if (!q->flows)
			return -ENOMEM;
496
		q->backlogs = kvzalloc(q->flows_cnt * sizeof(u32), GFP_KERNEL);
497
		if (!q->backlogs)
E
Eric Dumazet 已提交
498 499 500 501 502
			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
503
			codel_vars_init(&flow->cvars);
E
Eric Dumazet 已提交
504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531
		}
	}
	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) ||
532 533
	    nla_put_u32(skb, TCA_FQ_CODEL_DROP_BATCH_SIZE,
			q->drop_batch_size) ||
534 535
	    nla_put_u32(skb, TCA_FQ_CODEL_MEMORY_LIMIT,
			q->memory_limit) ||
E
Eric Dumazet 已提交
536 537 538 539
	    nla_put_u32(skb, TCA_FQ_CODEL_FLOWS,
			q->flows_cnt))
		goto nla_put_failure;

E
Eric Dumazet 已提交
540 541 542 543 544
	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;

545
	return nla_nest_end(skb, opts);
E
Eric Dumazet 已提交
546 547 548 549 550 551 552 553 554 555 556 557 558

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
559 560 561 562
	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
563
	st.qdisc_stats.ce_mark = q->cstats.ce_mark;
564 565
	st.qdisc_stats.memory_usage  = q->memory_usage;
	st.qdisc_stats.drop_overmemory = q->drop_overmemory;
S
Sasha Levin 已提交
566

567
	sch_tree_lock(sch);
E
Eric Dumazet 已提交
568 569 570 571 572
	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++;
573
	sch_tree_unlock(sch);
E
Eric Dumazet 已提交
574 575 576 577 578 579 580 581 582

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

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

583
static unsigned long fq_codel_find(struct Qdisc *sch, u32 classid)
E
Eric Dumazet 已提交
584 585 586 587 588 589 590 591 592 593 594 595
{
	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;
}

596
static void fq_codel_unbind(struct Qdisc *q, unsigned long cl)
E
Eric Dumazet 已提交
597 598 599
{
}

600 601
static struct tcf_block *fq_codel_tcf_block(struct Qdisc *sch, unsigned long cl,
					    struct netlink_ext_ack *extack)
E
Eric Dumazet 已提交
602 603 604 605 606
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);

	if (cl)
		return NULL;
607
	return q->block;
E
Eric Dumazet 已提交
608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626
}

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];
627
		const struct sk_buff *skb;
E
Eric Dumazet 已提交
628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644

		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);
		}
645 646 647 648 649 650 651 652
		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
653 654 655 656
		}
		qs.backlog = q->backlogs[idx];
		qs.drops = flow->dropped;
	}
657
	if (gnet_stats_copy_queue(d, NULL, &qs, qs.qlen) < 0)
E
Eric Dumazet 已提交
658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687
		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,
688
	.find		=	fq_codel_find,
689
	.tcf_block	=	fq_codel_tcf_block,
E
Eric Dumazet 已提交
690
	.bind_tcf	=	fq_codel_bind,
691
	.unbind_tcf	=	fq_codel_unbind,
E
Eric Dumazet 已提交
692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726
	.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");