sch_fq_codel.c 18.7 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-or-later
E
Eric Dumazet 已提交
2 3 4
/*
 * Fair Queue CoDel discipline
 *
E
Eric Dumazet 已提交
5
 *  Copyright (C) 2012,2015 Eric Dumazet <edumazet@google.com>
E
Eric Dumazet 已提交
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
 */

#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/slab.h>
#include <linux/vmalloc.h>
#include <net/netlink.h>
#include <net/pkt_sched.h>
21
#include <net/pkt_cls.h>
E
Eric Dumazet 已提交
22
#include <net/codel.h>
M
Michal Kazior 已提交
23 24
#include <net/codel_impl.h>
#include <net/codel_qdisc.h>
E
Eric Dumazet 已提交
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

/*	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;
	struct codel_vars cvars;
}; /* please try to keep this structure <= 64 bytes */

struct fq_codel_sched_data {
J
John Fastabend 已提交
51
	struct tcf_proto __rcu *filter_list; /* optional external classifier */
52
	struct tcf_block *block;
E
Eric Dumazet 已提交
53 54 55 56
	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)); */
57
	u32		drop_batch_size;
58
	u32		memory_limit;
E
Eric Dumazet 已提交
59 60
	struct codel_params cparams;
	struct codel_stats cstats;
61 62
	u32		memory_usage;
	u32		drop_overmemory;
E
Eric Dumazet 已提交
63 64 65 66 67 68 69 70
	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,
71
				  struct sk_buff *skb)
E
Eric Dumazet 已提交
72
{
73
	return reciprocal_scale(skb_get_hash(skb), q->flows_cnt);
E
Eric Dumazet 已提交
74 75 76 77 78 79
}

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
80
	struct tcf_proto *filter;
E
Eric Dumazet 已提交
81 82 83 84 85 86 87 88
	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);

89
	filter = rcu_dereference_bh(q->filter_list);
J
John Fastabend 已提交
90
	if (!filter)
E
Eric Dumazet 已提交
91 92 93
		return fq_codel_hash(q, skb) + 1;

	*qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
94
	result = tcf_classify(skb, filter, &res, false);
E
Eric Dumazet 已提交
95 96 97 98 99
	if (result >= 0) {
#ifdef CONFIG_NET_CLS_ACT
		switch (result) {
		case TC_ACT_STOLEN:
		case TC_ACT_QUEUED:
100
		case TC_ACT_TRAP:
E
Eric Dumazet 已提交
101
			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
102
			/* fall through */
E
Eric Dumazet 已提交
103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120
		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;
121
	skb_mark_not_on_list(skb);
E
Eric Dumazet 已提交
122 123 124 125 126 127 128 129 130 131 132 133 134 135 136
	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;
}

137 138
static unsigned int fq_codel_drop(struct Qdisc *sch, unsigned int max_packets,
				  struct sk_buff **to_free)
E
Eric Dumazet 已提交
139 140 141 142 143
{
	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;
144
	unsigned int threshold;
145
	unsigned int mem = 0;
E
Eric Dumazet 已提交
146

147
	/* Queue is full! Find the fat flow and drop packet(s) from it.
E
Eric Dumazet 已提交
148 149 150
	 * 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.
151 152
	 * 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
153 154 155 156 157 158 159
	 */
	for (i = 0; i < q->flows_cnt; i++) {
		if (q->backlogs[i] > maxbacklog) {
			maxbacklog = q->backlogs[i];
			idx = i;
		}
	}
160 161 162 163

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

E
Eric Dumazet 已提交
164
	flow = &q->flows[idx];
165 166 167 168 169
	len = 0;
	i = 0;
	do {
		skb = dequeue_head(flow);
		len += qdisc_pkt_len(skb);
170
		mem += get_codel_cb(skb)->mem_usage;
171
		__qdisc_drop(skb, to_free);
172 173
	} while (++i < max_packets && len < threshold);

174 175
	/* Tell codel to increase its signal strength also */
	flow->cvars.count += i;
E
Eric Dumazet 已提交
176
	q->backlogs[idx] -= len;
177
	q->memory_usage -= mem;
178 179 180
	sch->qstats.drops += i;
	sch->qstats.backlog -= len;
	sch->q.qlen -= i;
E
Eric Dumazet 已提交
181 182 183
	return idx;
}

184 185
static int fq_codel_enqueue(struct sk_buff *skb, struct Qdisc *sch,
			    struct sk_buff **to_free)
E
Eric Dumazet 已提交
186 187
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
188
	unsigned int idx, prev_backlog, prev_qlen;
E
Eric Dumazet 已提交
189 190
	struct fq_codel_flow *flow;
	int uninitialized_var(ret);
191
	unsigned int pkt_len;
192
	bool memory_limited;
E
Eric Dumazet 已提交
193 194 195 196

	idx = fq_codel_classify(skb, sch, &ret);
	if (idx == 0) {
		if (ret & __NET_XMIT_BYPASS)
197
			qdisc_qstats_drop(sch);
198
		__qdisc_drop(skb, to_free);
E
Eric Dumazet 已提交
199 200 201 202 203 204 205 206
		return ret;
	}
	idx--;

	codel_set_enqueue_time(skb);
	flow = &q->flows[idx];
	flow_queue_add(flow, skb);
	q->backlogs[idx] += qdisc_pkt_len(skb);
207
	qdisc_qstats_backlog_inc(sch, skb);
E
Eric Dumazet 已提交
208 209 210 211 212 213

	if (list_empty(&flow->flowchain)) {
		list_add_tail(&flow->flowchain, &q->new_flows);
		q->new_flow_count++;
		flow->deficit = q->quantum;
	}
214 215
	get_codel_cb(skb)->mem_usage = skb->truesize;
	q->memory_usage += get_codel_cb(skb)->mem_usage;
216 217
	memory_limited = q->memory_usage > q->memory_limit;
	if (++sch->q.qlen <= sch->limit && !memory_limited)
E
Eric Dumazet 已提交
218 219
		return NET_XMIT_SUCCESS;

220
	prev_backlog = sch->qstats.backlog;
221 222
	prev_qlen = sch->q.qlen;

223 224
	/* save this packet length as it might be dropped by fq_codel_drop() */
	pkt_len = qdisc_pkt_len(skb);
225 226 227 228
	/* 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
229
	 */
230
	ret = fq_codel_drop(sch, q->drop_batch_size, to_free);
231

232 233 234
	prev_qlen -= sch->q.qlen;
	prev_backlog -= sch->qstats.backlog;
	q->drop_overlimit += prev_qlen;
235
	if (memory_limited)
236
		q->drop_overmemory += prev_qlen;
237

238 239 240 241 242 243 244 245 246 247 248
	/* 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
249 250 251 252 253 254
}

/* 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.
 */
255
static struct sk_buff *dequeue_func(struct codel_vars *vars, void *ctx)
E
Eric Dumazet 已提交
256
{
257
	struct Qdisc *sch = ctx;
258
	struct fq_codel_sched_data *q = qdisc_priv(sch);
E
Eric Dumazet 已提交
259 260 261 262 263 264
	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);
265
		q->backlogs[flow - q->flows] -= qdisc_pkt_len(skb);
266
		q->memory_usage -= get_codel_cb(skb)->mem_usage;
E
Eric Dumazet 已提交
267
		sch->q.qlen--;
268
		sch->qstats.backlog -= qdisc_pkt_len(skb);
E
Eric Dumazet 已提交
269 270 271 272
	}
	return skb;
}

273 274 275 276
static void drop_func(struct sk_buff *skb, void *ctx)
{
	struct Qdisc *sch = ctx;

277 278
	kfree_skb(skb);
	qdisc_qstats_drop(sch);
279 280
}

E
Eric Dumazet 已提交
281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302
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;

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;
	}

303 304 305
	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
306 307 308 309 310 311 312 313 314 315 316

	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);
317
	/* We cant call qdisc_tree_reduce_backlog() if our qlen is 0,
E
Eric Dumazet 已提交
318 319 320
	 * or HTB crashes. Defer it for next round.
	 */
	if (q->cstats.drop_count && sch->q.qlen) {
321 322
		qdisc_tree_reduce_backlog(sch, q->cstats.drop_count,
					  q->cstats.drop_len);
E
Eric Dumazet 已提交
323
		q->cstats.drop_count = 0;
324
		q->cstats.drop_len = 0;
E
Eric Dumazet 已提交
325 326 327 328
	}
	return skb;
}

329 330 331 332 333 334
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
335 336
static void fq_codel_reset(struct Qdisc *sch)
{
337 338
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	int i;
E
Eric Dumazet 已提交
339

340 341 342 343 344
	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;

345
		fq_codel_flow_purge(flow);
346 347 348 349 350
		INIT_LIST_HEAD(&flow->flowchain);
		codel_vars_init(&flow->cvars);
	}
	memset(q->backlogs, 0, q->flows_cnt * sizeof(u32));
	sch->q.qlen = 0;
351
	sch->qstats.backlog = 0;
352
	q->memory_usage = 0;
E
Eric Dumazet 已提交
353 354 355 356 357 358 359 360 361
}

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
362
	[TCA_FQ_CODEL_CE_THRESHOLD] = { .type = NLA_U32 },
363
	[TCA_FQ_CODEL_DROP_BATCH_SIZE] = { .type = NLA_U32 },
364
	[TCA_FQ_CODEL_MEMORY_LIMIT] = { .type = NLA_U32 },
E
Eric Dumazet 已提交
365 366
};

367 368
static int fq_codel_change(struct Qdisc *sch, struct nlattr *opt,
			   struct netlink_ext_ack *extack)
E
Eric Dumazet 已提交
369 370 371 372 373 374 375 376
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	struct nlattr *tb[TCA_FQ_CODEL_MAX + 1];
	int err;

	if (!opt)
		return -EINVAL;

377 378
	err = nla_parse_nested_deprecated(tb, TCA_FQ_CODEL_MAX, opt,
					  fq_codel_policy, NULL);
E
Eric Dumazet 已提交
379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
	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
397 398 399 400 401 402
	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
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417
	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]));

418 419 420
	if (tb[TCA_FQ_CODEL_DROP_BATCH_SIZE])
		q->drop_batch_size = min(1U, nla_get_u32(tb[TCA_FQ_CODEL_DROP_BATCH_SIZE]));

421 422 423 424 425
	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
426 427
		struct sk_buff *skb = fq_codel_dequeue(sch);

428
		q->cstats.drop_len += qdisc_pkt_len(skb);
429
		rtnl_kfree_skbs(skb, skb);
E
Eric Dumazet 已提交
430 431
		q->cstats.drop_count++;
	}
432
	qdisc_tree_reduce_backlog(sch, q->cstats.drop_count, q->cstats.drop_len);
E
Eric Dumazet 已提交
433
	q->cstats.drop_count = 0;
434
	q->cstats.drop_len = 0;
E
Eric Dumazet 已提交
435 436 437 438 439 440 441 442 443

	sch_tree_unlock(sch);
	return 0;
}

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

444
	tcf_block_put(q->block);
445 446
	kvfree(q->backlogs);
	kvfree(q->flows);
E
Eric Dumazet 已提交
447 448
}

449 450
static int fq_codel_init(struct Qdisc *sch, struct nlattr *opt,
			 struct netlink_ext_ack *extack)
E
Eric Dumazet 已提交
451 452 453
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);
	int i;
454
	int err;
E
Eric Dumazet 已提交
455 456 457

	sch->limit = 10*1024;
	q->flows_cnt = 1024;
458
	q->memory_limit = 32 << 20; /* 32 MBytes */
459
	q->drop_batch_size = 64;
E
Eric Dumazet 已提交
460 461 462
	q->quantum = psched_mtu(qdisc_dev(sch));
	INIT_LIST_HEAD(&q->new_flows);
	INIT_LIST_HEAD(&q->old_flows);
463
	codel_params_init(&q->cparams);
E
Eric Dumazet 已提交
464 465
	codel_stats_init(&q->cstats);
	q->cparams.ecn = true;
466
	q->cparams.mtu = psched_mtu(qdisc_dev(sch));
E
Eric Dumazet 已提交
467 468

	if (opt) {
469
		err = fq_codel_change(sch, opt, extack);
E
Eric Dumazet 已提交
470
		if (err)
471
			goto init_failure;
E
Eric Dumazet 已提交
472 473
	}

474
	err = tcf_block_get(&q->block, &q->filter_list, sch, extack);
475
	if (err)
476
		goto init_failure;
477

E
Eric Dumazet 已提交
478
	if (!q->flows) {
K
Kees Cook 已提交
479 480 481
		q->flows = kvcalloc(q->flows_cnt,
				    sizeof(struct fq_codel_flow),
				    GFP_KERNEL);
482 483 484 485
		if (!q->flows) {
			err = -ENOMEM;
			goto init_failure;
		}
K
Kees Cook 已提交
486
		q->backlogs = kvcalloc(q->flows_cnt, sizeof(u32), GFP_KERNEL);
487 488 489 490
		if (!q->backlogs) {
			err = -ENOMEM;
			goto alloc_failure;
		}
E
Eric Dumazet 已提交
491 492 493 494
		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
495
			codel_vars_init(&flow->cvars);
E
Eric Dumazet 已提交
496 497 498 499 500 501 502
		}
	}
	if (sch->limit >= 1)
		sch->flags |= TCQ_F_CAN_BYPASS;
	else
		sch->flags &= ~TCQ_F_CAN_BYPASS;
	return 0;
503 504 505 506 507 508 509

alloc_failure:
	kvfree(q->flows);
	q->flows = NULL;
init_failure:
	q->flows_cnt = 0;
	return err;
E
Eric Dumazet 已提交
510 511 512 513 514 515 516
}

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

517
	opts = nla_nest_start_noflag(skb, TCA_OPTIONS);
E
Eric Dumazet 已提交
518 519 520 521 522 523 524 525 526 527 528 529 530
	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) ||
531 532
	    nla_put_u32(skb, TCA_FQ_CODEL_DROP_BATCH_SIZE,
			q->drop_batch_size) ||
533 534
	    nla_put_u32(skb, TCA_FQ_CODEL_MEMORY_LIMIT,
			q->memory_limit) ||
E
Eric Dumazet 已提交
535 536 537 538
	    nla_put_u32(skb, TCA_FQ_CODEL_FLOWS,
			q->flows_cnt))
		goto nla_put_failure;

E
Eric Dumazet 已提交
539 540 541 542 543
	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;

544
	return nla_nest_end(skb, opts);
E
Eric Dumazet 已提交
545 546 547 548 549 550 551 552 553 554 555 556 557

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
558 559 560 561
	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
562
	st.qdisc_stats.ce_mark = q->cstats.ce_mark;
563 564
	st.qdisc_stats.memory_usage  = q->memory_usage;
	st.qdisc_stats.drop_overmemory = q->drop_overmemory;
S
Sasha Levin 已提交
565

566
	sch_tree_lock(sch);
E
Eric Dumazet 已提交
567 568 569 570 571
	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++;
572
	sch_tree_unlock(sch);
E
Eric Dumazet 已提交
573 574 575 576 577 578 579 580 581

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

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

582
static unsigned long fq_codel_find(struct Qdisc *sch, u32 classid)
E
Eric Dumazet 已提交
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)
{
	return 0;
}

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

597 598
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
599 600 601 602 603
{
	struct fq_codel_sched_data *q = qdisc_priv(sch);

	if (cl)
		return NULL;
604
	return q->block;
E
Eric Dumazet 已提交
605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623
}

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];
624
		const struct sk_buff *skb;
E
Eric Dumazet 已提交
625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641

		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);
		}
642 643 644 645 646 647 648 649
		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
650 651
		}
		qs.backlog = q->backlogs[idx];
652
		qs.drops = 0;
E
Eric Dumazet 已提交
653
	}
654
	if (gnet_stats_copy_queue(d, NULL, &qs, qs.qlen) < 0)
E
Eric Dumazet 已提交
655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684
		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,
685
	.find		=	fq_codel_find,
686
	.tcf_block	=	fq_codel_tcf_block,
E
Eric Dumazet 已提交
687
	.bind_tcf	=	fq_codel_bind,
688
	.unbind_tcf	=	fq_codel_unbind,
E
Eric Dumazet 已提交
689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723
	.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");