[linux.git] / include / net / fq_impl.h

/*
 * Copyright (c) 2016 Qualcomm Atheros, Inc
 *
 * GPL v2
 *
 * Based on net/sched/sch_fq_codel.c
 */
#ifndef __NET_SCHED_FQ_IMPL_H
#define __NET_SCHED_FQ_IMPL_H

#include <net/fq.h>

/* functions that are embedded into includer */

static void fq_adjust_removal(struct fq *fq,
			      struct fq_flow *flow,
			      struct sk_buff *skb)
{
	struct fq_tin *tin = flow->tin;

	tin->backlog_bytes -= skb->len;
	tin->backlog_packets--;
	flow->backlog -= skb->len;
	fq->backlog--;
	fq->memory_usage -= skb->truesize;
}

static void fq_rejigger_backlog(struct fq *fq, struct fq_flow *flow)
{
	struct fq_flow *i;

	if (flow->backlog == 0) {
		list_del_init(&flow->backlogchain);
	} else {
		i = flow;

		list_for_each_entry_continue(i, &fq->backlogs, backlogchain)
			if (i->backlog < flow->backlog)
				break;

		list_move_tail(&flow->backlogchain,
			       &i->backlogchain);
	}
}

static struct sk_buff *fq_flow_dequeue(struct fq *fq,
				       struct fq_flow *flow)
{
	struct sk_buff *skb;

	lockdep_assert_held(&fq->lock);

	skb = __skb_dequeue(&flow->queue);
	if (!skb)
		return NULL;

	fq_adjust_removal(fq, flow, skb);
	fq_rejigger_backlog(fq, flow);

	return skb;
}

static struct sk_buff *fq_tin_dequeue(struct fq *fq,
				      struct fq_tin *tin,
				      fq_tin_dequeue_t dequeue_func)
{
	struct fq_flow *flow;
	struct list_head *head;
	struct sk_buff *skb;

	lockdep_assert_held(&fq->lock);

begin:
	head = &tin->new_flows;
	if (list_empty(head)) {
		head = &tin->old_flows;
		if (list_empty(head))
			return NULL;
	}

	flow = list_first_entry(head, struct fq_flow, flowchain);

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

	skb = dequeue_func(fq, tin, flow);
	if (!skb) {
		/* force a pass through old_flows to prevent starvation */
		if ((head == &tin->new_flows) &&
		    !list_empty(&tin->old_flows)) {
			list_move_tail(&flow->flowchain, &tin->old_flows);
		} else {
			list_del_init(&flow->flowchain);
			flow->tin = NULL;
		}
		goto begin;
	}

	flow->deficit -= skb->len;
	tin->tx_bytes += skb->len;
	tin->tx_packets++;

	return skb;
}

static struct fq_flow *fq_flow_classify(struct fq *fq,
					struct fq_tin *tin,
					struct sk_buff *skb,
					fq_flow_get_default_t get_default_func)
{
	struct fq_flow *flow;
	u32 hash;
	u32 idx;

	lockdep_assert_held(&fq->lock);

	hash = skb_get_hash_perturb(skb, fq->perturbation);
	idx = reciprocal_scale(hash, fq->flows_cnt);
	flow = &fq->flows[idx];

	if (flow->tin && flow->tin != tin) {
		flow = get_default_func(fq, tin, idx, skb);
		tin->collisions++;
		fq->collisions++;
	}

	if (!flow->tin)
		tin->flows++;

	return flow;
}

static void fq_recalc_backlog(struct fq *fq,
			      struct fq_tin *tin,
			      struct fq_flow *flow)
{
	struct fq_flow *i;

	if (list_empty(&flow->backlogchain))
		list_add_tail(&flow->backlogchain, &fq->backlogs);

	i = flow;
	list_for_each_entry_continue_reverse(i, &fq->backlogs,
					     backlogchain)
		if (i->backlog > flow->backlog)
			break;

	list_move(&flow->backlogchain, &i->backlogchain);
}

static void fq_tin_enqueue(struct fq *fq,
			   struct fq_tin *tin,
			   struct sk_buff *skb,
			   fq_skb_free_t free_func,
			   fq_flow_get_default_t get_default_func)
{
	struct fq_flow *flow;
	bool oom;

	lockdep_assert_held(&fq->lock);

	flow = fq_flow_classify(fq, tin, skb, get_default_func);

	flow->tin = tin;
	flow->backlog += skb->len;
	tin->backlog_bytes += skb->len;
	tin->backlog_packets++;
	fq->memory_usage += skb->truesize;
	fq->backlog++;

	fq_recalc_backlog(fq, tin, flow);

	if (list_empty(&flow->flowchain)) {
		flow->deficit = fq->quantum;
		list_add_tail(&flow->flowchain,
			      &tin->new_flows);
	}

	__skb_queue_tail(&flow->queue, skb);
	oom = (fq->memory_usage > fq->memory_limit);
	while (fq->backlog > fq->limit || oom) {
		flow = list_first_entry_or_null(&fq->backlogs,
						struct fq_flow,
						backlogchain);
		if (!flow)
			return;

		skb = fq_flow_dequeue(fq, flow);
		if (!skb)
			return;

		free_func(fq, flow->tin, flow, skb);

		flow->tin->overlimit++;
		fq->overlimit++;
		if (oom) {
			fq->overmemory++;
			oom = (fq->memory_usage > fq->memory_limit);
		}
	}
}

static void fq_flow_filter(struct fq *fq,
			   struct fq_flow *flow,
			   fq_skb_filter_t filter_func,
			   void *filter_data,
			   fq_skb_free_t free_func)
{
	struct fq_tin *tin = flow->tin;
	struct sk_buff *skb, *tmp;

	lockdep_assert_held(&fq->lock);

	skb_queue_walk_safe(&flow->queue, skb, tmp) {
		if (!filter_func(fq, tin, flow, skb, filter_data))
			continue;

		__skb_unlink(skb, &flow->queue);
		fq_adjust_removal(fq, flow, skb);
		free_func(fq, tin, flow, skb);
	}

	fq_rejigger_backlog(fq, flow);
}

static void fq_tin_filter(struct fq *fq,
			  struct fq_tin *tin,
			  fq_skb_filter_t filter_func,
			  void *filter_data,
			  fq_skb_free_t free_func)
{
	struct fq_flow *flow;

	lockdep_assert_held(&fq->lock);

	list_for_each_entry(flow, &tin->new_flows, flowchain)
		fq_flow_filter(fq, flow, filter_func, filter_data, free_func);
	list_for_each_entry(flow, &tin->old_flows, flowchain)
		fq_flow_filter(fq, flow, filter_func, filter_data, free_func);
}

static void fq_flow_reset(struct fq *fq,
			  struct fq_flow *flow,
			  fq_skb_free_t free_func)
{
	struct sk_buff *skb;

	while ((skb = fq_flow_dequeue(fq, flow)))
		free_func(fq, flow->tin, flow, skb);

	if (!list_empty(&flow->flowchain))
		list_del_init(&flow->flowchain);

	if (!list_empty(&flow->backlogchain))
		list_del_init(&flow->backlogchain);

	flow->tin = NULL;

	WARN_ON_ONCE(flow->backlog);
}

static void fq_tin_reset(struct fq *fq,
			 struct fq_tin *tin,
			 fq_skb_free_t free_func)
{
	struct list_head *head;
	struct fq_flow *flow;

	for (;;) {
		head = &tin->new_flows;
		if (list_empty(head)) {
			head = &tin->old_flows;
			if (list_empty(head))
				break;
		}

		flow = list_first_entry(head, struct fq_flow, flowchain);
		fq_flow_reset(fq, flow, free_func);
	}

	WARN_ON_ONCE(tin->backlog_bytes);
	WARN_ON_ONCE(tin->backlog_packets);
}

static void fq_flow_init(struct fq_flow *flow)
{
	INIT_LIST_HEAD(&flow->flowchain);
	INIT_LIST_HEAD(&flow->backlogchain);
	__skb_queue_head_init(&flow->queue);
}

static void fq_tin_init(struct fq_tin *tin)
{
	INIT_LIST_HEAD(&tin->new_flows);
	INIT_LIST_HEAD(&tin->old_flows);
}

static int fq_init(struct fq *fq, int flows_cnt)
{
	int i;

	memset(fq, 0, sizeof(fq[0]));
	INIT_LIST_HEAD(&fq->backlogs);
	spin_lock_init(&fq->lock);
	fq->flows_cnt = max_t(u32, flows_cnt, 1);
	fq->perturbation = prandom_u32();
	fq->quantum = 300;
	fq->limit = 8192;
	fq->memory_limit = 16 << 20; /* 16 MBytes */

	fq->flows = kcalloc(fq->flows_cnt, sizeof(fq->flows[0]), GFP_KERNEL);
	if (!fq->flows)
		return -ENOMEM;

	for (i = 0; i < fq->flows_cnt; i++)
		fq_flow_init(&fq->flows[i]);

	return 0;
}

static void fq_reset(struct fq *fq,
		     fq_skb_free_t free_func)
{
	int i;

	for (i = 0; i < fq->flows_cnt; i++)
		fq_flow_reset(fq, &fq->flows[i], free_func);

	kfree(fq->flows);
	fq->flows = NULL;
}

#endif
Commit	Line	Data
557fc4a0 MK	1	/*
	2	* Copyright (c) 2016 Qualcomm Atheros, Inc
	3	*
	4	* GPL v2
	5	*
	6	* Based on net/sched/sch_fq_codel.c
	7	*/
	8	#ifndef __NET_SCHED_FQ_IMPL_H
	9	#define __NET_SCHED_FQ_IMPL_H
	10
	11	#include <net/fq.h>
	12
	13	/* functions that are embedded into includer */
	14
8c418b5b JB	15	static void fq_adjust_removal(struct fq *fq,
	16	struct fq_flow *flow,
	17	struct sk_buff *skb)
557fc4a0 MK	18	{
557fc4a0 MK	19	struct fq_tin *tin = flow->tin;
557fc4a0 MK	20
	21	tin->backlog_bytes -= skb->len;
	22	tin->backlog_packets--;
	23	flow->backlog -= skb->len;
	24	fq->backlog--;
097b065b	25	fq->memory_usage -= skb->truesize;
8c418b5b JB	26	}
	27
	28	static void fq_rejigger_backlog(struct fq fq, struct fq_flow flow)
	29	{
	30	struct fq_flow *i;
557fc4a0 MK	31
	32	if (flow->backlog == 0) {
	33	list_del_init(&flow->backlogchain);
	34	} else {
	35	i = flow;
	36
	37	list_for_each_entry_continue(i, &fq->backlogs, backlogchain)
	38	if (i->backlog < flow->backlog)
	39	break;
	40
	41	list_move_tail(&flow->backlogchain,
	42	&i->backlogchain);
	43	}
8c418b5b JB	44	}
	45
	46	static struct sk_buff fq_flow_dequeue(struct fq fq,
	47	struct fq_flow *flow)
	48	{
	49	struct sk_buff *skb;
	50
	51	lockdep_assert_held(&fq->lock);
	52
	53	skb = __skb_dequeue(&flow->queue);
	54	if (!skb)
	55	return NULL;
	56
	57	fq_adjust_removal(fq, flow, skb);
	58	fq_rejigger_backlog(fq, flow);
557fc4a0 MK	59
	60	return skb;
	61	}
	62
	63	static struct sk_buff fq_tin_dequeue(struct fq fq,
	64	struct fq_tin *tin,
	65	fq_tin_dequeue_t dequeue_func)
	66	{
	67	struct fq_flow *flow;
	68	struct list_head *head;
	69	struct sk_buff *skb;
	70
	71	lockdep_assert_held(&fq->lock);
	72
	73	begin:
	74	head = &tin->new_flows;
	75	if (list_empty(head)) {
	76	head = &tin->old_flows;
	77	if (list_empty(head))
	78	return NULL;
	79	}
	80
	81	flow = list_first_entry(head, struct fq_flow, flowchain);
	82
	83	if (flow->deficit <= 0) {
	84	flow->deficit += fq->quantum;
	85	list_move_tail(&flow->flowchain,
	86	&tin->old_flows);
	87	goto begin;
	88	}
	89
	90	skb = dequeue_func(fq, tin, flow);
	91	if (!skb) {
	92	/* force a pass through old_flows to prevent starvation */
	93	if ((head == &tin->new_flows) &&
	94	!list_empty(&tin->old_flows)) {
	95	list_move_tail(&flow->flowchain, &tin->old_flows);
	96	} else {
	97	list_del_init(&flow->flowchain);
	98	flow->tin = NULL;
	99	}
	100	goto begin;
	101	}
	102
	103	flow->deficit -= skb->len;
	104	tin->tx_bytes += skb->len;
	105	tin->tx_packets++;
	106
	107	return skb;
	108	}
	109
	110	static struct fq_flow fq_flow_classify(struct fq fq,
	111	struct fq_tin *tin,
	112	struct sk_buff *skb,
	113	fq_flow_get_default_t get_default_func)
	114	{
	115	struct fq_flow *flow;
	116	u32 hash;
	117	u32 idx;
	118
	119	lockdep_assert_held(&fq->lock);
	120
	121	hash = skb_get_hash_perturb(skb, fq->perturbation);
	122	idx = reciprocal_scale(hash, fq->flows_cnt);
123	flow = &fq->flows[idx];
124
125	if (flow->tin && flow->tin != tin) {
126	flow = get_default_func(fq, tin, idx, skb);
127	tin->collisions++;
128	fq->collisions++;
129	}
130
131	if (!flow->tin)
132	tin->flows++;
133
134	return flow;
135	}
136
b43e7199 MK	137	static void fq_recalc_backlog(struct fq *fq,
	138	struct fq_tin *tin,
	139	struct fq_flow *flow)
	140	{
	141	struct fq_flow *i;
	142
	143	if (list_empty(&flow->backlogchain))
	144	list_add_tail(&flow->backlogchain, &fq->backlogs);
	145
	146	i = flow;
	147	list_for_each_entry_continue_reverse(i, &fq->backlogs,
	148	backlogchain)
	149	if (i->backlog > flow->backlog)
	150	break;
	151
	152	list_move(&flow->backlogchain, &i->backlogchain);
	153	}
	154
557fc4a0 MK	155	static void fq_tin_enqueue(struct fq *fq,
	156	struct fq_tin *tin,
	157	struct sk_buff *skb,
	158	fq_skb_free_t free_func,
	159	fq_flow_get_default_t get_default_func)
	160	{
	161	struct fq_flow *flow;
0bfe649f	162	bool oom;
557fc4a0 MK	163
	164	lockdep_assert_held(&fq->lock);
	165
	166	flow = fq_flow_classify(fq, tin, skb, get_default_func);
	167
	168	flow->tin = tin;
	169	flow->backlog += skb->len;
	170	tin->backlog_bytes += skb->len;
	171	tin->backlog_packets++;
097b065b	172	fq->memory_usage += skb->truesize;
557fc4a0 MK	173	fq->backlog++;
557fc4a0 MK	174
b43e7199	175	fq_recalc_backlog(fq, tin, flow);
557fc4a0 MK	176
	177	if (list_empty(&flow->flowchain)) {
	178	flow->deficit = fq->quantum;
	179	list_add_tail(&flow->flowchain,
	180	&tin->new_flows);
	181	}
	182
	183	__skb_queue_tail(&flow->queue, skb);
0bfe649f THJ	184	oom = (fq->memory_usage > fq->memory_limit);
0bfe649f THJ	185	while (fq->backlog > fq->limit \|\| oom) {
557fc4a0 MK	186	flow = list_first_entry_or_null(&fq->backlogs,
	187	struct fq_flow,
	188	backlogchain);
	189	if (!flow)
	190	return;
	191
	192	skb = fq_flow_dequeue(fq, flow);
	193	if (!skb)
	194	return;
	195
	196	free_func(fq, flow->tin, flow, skb);
	197
	198	flow->tin->overlimit++;
	199	fq->overlimit++;
0bfe649f	200	if (oom) {
097b065b	201	fq->overmemory++;
0bfe649f THJ	202	oom = (fq->memory_usage > fq->memory_limit);
0bfe649f THJ	203	}
557fc4a0 MK	204	}
	205	}
	206
8c418b5b JB	207	static void fq_flow_filter(struct fq *fq,
	208	struct fq_flow *flow,
	209	fq_skb_filter_t filter_func,
	210	void *filter_data,
	211	fq_skb_free_t free_func)
	212	{
	213	struct fq_tin *tin = flow->tin;
	214	struct sk_buff skb, tmp;
	215
	216	lockdep_assert_held(&fq->lock);
	217
	218	skb_queue_walk_safe(&flow->queue, skb, tmp) {
	219	if (!filter_func(fq, tin, flow, skb, filter_data))
	220	continue;
	221
	222	__skb_unlink(skb, &flow->queue);
	223	fq_adjust_removal(fq, flow, skb);
	224	free_func(fq, tin, flow, skb);
	225	}
	226
	227	fq_rejigger_backlog(fq, flow);
	228	}
	229
	230	static void fq_tin_filter(struct fq *fq,
	231	struct fq_tin *tin,
	232	fq_skb_filter_t filter_func,
	233	void *filter_data,
	234	fq_skb_free_t free_func)
	235	{
	236	struct fq_flow *flow;
	237
	238	lockdep_assert_held(&fq->lock);
	239
	240	list_for_each_entry(flow, &tin->new_flows, flowchain)
	241	fq_flow_filter(fq, flow, filter_func, filter_data, free_func);
	242	list_for_each_entry(flow, &tin->old_flows, flowchain)
	243	fq_flow_filter(fq, flow, filter_func, filter_data, free_func);
	244	}
	245
557fc4a0 MK	246	static void fq_flow_reset(struct fq *fq,
	247	struct fq_flow *flow,
	248	fq_skb_free_t free_func)
	249	{
	250	struct sk_buff *skb;
	251
	252	while ((skb = fq_flow_dequeue(fq, flow)))
	253	free_func(fq, flow->tin, flow, skb);
	254
	255	if (!list_empty(&flow->flowchain))
	256	list_del_init(&flow->flowchain);
	257
	258	if (!list_empty(&flow->backlogchain))
	259	list_del_init(&flow->backlogchain);
	260
	261	flow->tin = NULL;
	262
	263	WARN_ON_ONCE(flow->backlog);
	264	}
	265
	266	static void fq_tin_reset(struct fq *fq,
	267	struct fq_tin *tin,
	268	fq_skb_free_t free_func)
	269	{
	270	struct list_head *head;
	271	struct fq_flow *flow;
	272
	273	for (;;) {
	274	head = &tin->new_flows;
	275	if (list_empty(head)) {
	276	head = &tin->old_flows;
	277	if (list_empty(head))
	278	break;
	279	}
	280
	281	flow = list_first_entry(head, struct fq_flow, flowchain);
	282	fq_flow_reset(fq, flow, free_func);
	283	}
	284
	285	WARN_ON_ONCE(tin->backlog_bytes);
	286	WARN_ON_ONCE(tin->backlog_packets);
	287	}
	288
	289	static void fq_flow_init(struct fq_flow *flow)
	290	{
	291	INIT_LIST_HEAD(&flow->flowchain);
	292	INIT_LIST_HEAD(&flow->backlogchain);
	293	__skb_queue_head_init(&flow->queue);
	294	}
	295
	296	static void fq_tin_init(struct fq_tin *tin)
	297	{
	298	INIT_LIST_HEAD(&tin->new_flows);
	299	INIT_LIST_HEAD(&tin->old_flows);
	300	}
	301
	302	static int fq_init(struct fq *fq, int flows_cnt)
	303	{
	304	int i;
	305
	306	memset(fq, 0, sizeof(fq[0]));
	307	INIT_LIST_HEAD(&fq->backlogs);
	308	spin_lock_init(&fq->lock);
	309	fq->flows_cnt = max_t(u32, flows_cnt, 1);
310	fq->perturbation = prandom_u32();
311	fq->quantum = 300;
312	fq->limit = 8192;
097b065b	313	fq->memory_limit = 16 << 20; /* 16 MBytes */
557fc4a0 MK	314
	315	fq->flows = kcalloc(fq->flows_cnt, sizeof(fq->flows[0]), GFP_KERNEL);
	316	if (!fq->flows)
	317	return -ENOMEM;
	318
	319	for (i = 0; i < fq->flows_cnt; i++)
	320	fq_flow_init(&fq->flows[i]);
	321
	322	return 0;
	323	}
	324
	325	static void fq_reset(struct fq *fq,
	326	fq_skb_free_t free_func)
	327	{
	328	int i;
	329
	330	for (i = 0; i < fq->flows_cnt; i++)
	331	fq_flow_reset(fq, &fq->flows[i], free_func);
	332
	333	kfree(fq->flows);
	334	fq->flows = NULL;
	335	}
	336
	337	#endif