[mirror_ubuntu-bionic-kernel.git] / kernel / irq / affinity.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Copyright (C) 2016 Thomas Gleixner.
 * Copyright (C) 2016-2017 Christoph Hellwig.
 */
#include <linux/interrupt.h>
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/cpu.h>

static void irq_spread_init_one(struct cpumask *irqmsk, struct cpumask *nmsk,
				int cpus_per_vec)
{
	const struct cpumask *siblmsk;
	int cpu, sibl;

	for ( ; cpus_per_vec > 0; ) {
		cpu = cpumask_first(nmsk);

		/* Should not happen, but I'm too lazy to think about it */
		if (cpu >= nr_cpu_ids)
			return;

		cpumask_clear_cpu(cpu, nmsk);
		cpumask_set_cpu(cpu, irqmsk);
		cpus_per_vec--;

		/* If the cpu has siblings, use them first */
		siblmsk = topology_sibling_cpumask(cpu);
		for (sibl = -1; cpus_per_vec > 0; ) {
			sibl = cpumask_next(sibl, siblmsk);
			if (sibl >= nr_cpu_ids)
				break;
			if (!cpumask_test_and_clear_cpu(sibl, nmsk))
				continue;
			cpumask_set_cpu(sibl, irqmsk);
			cpus_per_vec--;
		}
	}
}

static cpumask_var_t *alloc_node_to_cpumask(void)
{
	cpumask_var_t *masks;
	int node;

	masks = kcalloc(nr_node_ids, sizeof(cpumask_var_t), GFP_KERNEL);
	if (!masks)
		return NULL;

	for (node = 0; node < nr_node_ids; node++) {
		if (!zalloc_cpumask_var(&masks[node], GFP_KERNEL))
			goto out_unwind;
	}

	return masks;

out_unwind:
	while (--node >= 0)
		free_cpumask_var(masks[node]);
	kfree(masks);
	return NULL;
}

static void free_node_to_cpumask(cpumask_var_t *masks)
{
	int node;

	for (node = 0; node < nr_node_ids; node++)
		free_cpumask_var(masks[node]);
	kfree(masks);
}

static void build_node_to_cpumask(cpumask_var_t *masks)
{
	int cpu;

	for_each_possible_cpu(cpu)
		cpumask_set_cpu(cpu, masks[cpu_to_node(cpu)]);
}

static int get_nodes_in_cpumask(cpumask_var_t *node_to_cpumask,
				const struct cpumask *mask, nodemask_t *nodemsk)
{
	int n, nodes = 0;

	/* Calculate the number of nodes in the supplied affinity mask */
	for_each_node(n) {
		if (cpumask_intersects(mask, node_to_cpumask[n])) {
			node_set(n, *nodemsk);
			nodes++;
		}
	}
	return nodes;
}

static int irq_build_affinity_masks(const struct irq_affinity *affd,
				    int startvec, int numvecs,
				    cpumask_var_t *node_to_cpumask,
				    const struct cpumask *cpu_mask,
				    struct cpumask *nmsk,
				    struct cpumask *masks)
{
	int n, nodes, cpus_per_vec, extra_vecs, done = 0;
	int last_affv = affd->pre_vectors + numvecs;
	int curvec = startvec;
	nodemask_t nodemsk = NODE_MASK_NONE;

	if (!cpumask_weight(cpu_mask))
		return 0;

	nodes = get_nodes_in_cpumask(node_to_cpumask, cpu_mask, &nodemsk);

	/*
	 * If the number of nodes in the mask is greater than or equal the
	 * number of vectors we just spread the vectors across the nodes.
	 */
	if (numvecs <= nodes) {
		for_each_node_mask(n, nodemsk) {
			cpumask_or(masks + curvec, masks + curvec, node_to_cpumask[n]);
			if (++curvec == last_affv)
				curvec = affd->pre_vectors;
		}
		done = numvecs;
		goto out;
	}

	for_each_node_mask(n, nodemsk) {
		int ncpus, v, vecs_to_assign, vecs_per_node;

		/* Spread the vectors per node */
		vecs_per_node = (numvecs - (curvec - affd->pre_vectors)) / nodes;

		/* Get the cpus on this node which are in the mask */
		cpumask_and(nmsk, cpu_mask, node_to_cpumask[n]);

		/* Calculate the number of cpus per vector */
		ncpus = cpumask_weight(nmsk);
		vecs_to_assign = min(vecs_per_node, ncpus);

		/* Account for rounding errors */
		extra_vecs = ncpus - vecs_to_assign * (ncpus / vecs_to_assign);

		for (v = 0; curvec < last_affv && v < vecs_to_assign;
		     curvec++, v++) {
			cpus_per_vec = ncpus / vecs_to_assign;

			/* Account for extra vectors to compensate rounding errors */
			if (extra_vecs) {
				cpus_per_vec++;
				--extra_vecs;
			}
			irq_spread_init_one(masks + curvec, nmsk, cpus_per_vec);
		}

		done += v;
		if (done >= numvecs)
			break;
		if (curvec >= last_affv)
			curvec = affd->pre_vectors;
		--nodes;
	}

out:
	return done;
}

/**
 * irq_create_affinity_masks - Create affinity masks for multiqueue spreading
 * @nvecs:	The total number of vectors
 * @affd:	Description of the affinity requirements
 *
 * Returns the masks pointer or NULL if allocation failed.
 */
struct cpumask *
irq_create_affinity_masks(int nvecs, const struct irq_affinity *affd)
{
	int affvecs = nvecs - affd->pre_vectors - affd->post_vectors;
	int curvec, usedvecs;
	cpumask_var_t nmsk, npresmsk, *node_to_cpumask;
	struct cpumask *masks = NULL;

	/*
	 * If there aren't any vectors left after applying the pre/post
	 * vectors don't bother with assigning affinity.
	 */
	if (nvecs == affd->pre_vectors + affd->post_vectors)
		return NULL;

	if (!zalloc_cpumask_var(&nmsk, GFP_KERNEL))
		return NULL;

	if (!zalloc_cpumask_var(&npresmsk, GFP_KERNEL))
		goto outcpumsk;

	node_to_cpumask = alloc_node_to_cpumask();
	if (!node_to_cpumask)
		goto outnpresmsk;

	masks = kcalloc(nvecs, sizeof(*masks), GFP_KERNEL);
	if (!masks)
		goto outnodemsk;

	/* Fill out vectors at the beginning that don't need affinity */
	for (curvec = 0; curvec < affd->pre_vectors; curvec++)
		cpumask_copy(masks + curvec, irq_default_affinity);

	/* Stabilize the cpumasks */
	get_online_cpus();
	build_node_to_cpumask(node_to_cpumask);

	/* Spread on present CPUs starting from affd->pre_vectors */
	usedvecs = irq_build_affinity_masks(affd, curvec, affvecs,
					    node_to_cpumask, cpu_present_mask,
					    nmsk, masks);

	/*
	 * Spread on non present CPUs starting from the next vector to be
	 * handled. If the spreading of present CPUs already exhausted the
	 * vector space, assign the non present CPUs to the already spread
	 * out vectors.
	 */
	if (usedvecs >= affvecs)
		curvec = affd->pre_vectors;
	else
		curvec = affd->pre_vectors + usedvecs;
	cpumask_andnot(npresmsk, cpu_possible_mask, cpu_present_mask);
	usedvecs += irq_build_affinity_masks(affd, curvec, affvecs,
					     node_to_cpumask, npresmsk,
					     nmsk, masks);
	put_online_cpus();

	/* Fill out vectors at the end that don't need affinity */
	if (usedvecs >= affvecs)
		curvec = affd->pre_vectors + affvecs;
	else
		curvec = affd->pre_vectors + usedvecs;
	for (; curvec < nvecs; curvec++)
		cpumask_copy(masks + curvec, irq_default_affinity);

outnodemsk:
	free_node_to_cpumask(node_to_cpumask);
outnpresmsk:
	free_cpumask_var(npresmsk);
outcpumsk:
	free_cpumask_var(nmsk);
	return masks;
}

/**
 * irq_calc_affinity_vectors - Calculate the optimal number of vectors
 * @minvec:	The minimum number of vectors available
 * @maxvec:	The maximum number of vectors available
 * @affd:	Description of the affinity requirements
 */
int irq_calc_affinity_vectors(int minvec, int maxvec, const struct irq_affinity *affd)
{
	int resv = affd->pre_vectors + affd->post_vectors;
	int vecs = maxvec - resv;
	int ret;

	if (resv > minvec)
		return 0;

	get_online_cpus();
	ret = min_t(int, cpumask_weight(cpu_possible_mask), vecs) + resv;
	put_online_cpus();
	return ret;
}
Commit	Line	Data
b2441318	1	// SPDX-License-Identifier: GPL-2.0
9a0ef98e CH	2	/*
	3	* Copyright (C) 2016 Thomas Gleixner.
	4	* Copyright (C) 2016-2017 Christoph Hellwig.
	5	*/
5e385a6e CH	6	#include <linux/interrupt.h>
	7	#include <linux/kernel.h>
	8	#include <linux/slab.h>
	9	#include <linux/cpu.h>
	10
34c3d981 TG	11	static void irq_spread_init_one(struct cpumask irqmsk, struct cpumask nmsk,
	12	int cpus_per_vec)
	13	{
	14	const struct cpumask *siblmsk;
	15	int cpu, sibl;
	16
	17	for ( ; cpus_per_vec > 0; ) {
	18	cpu = cpumask_first(nmsk);
	19
	20	/* Should not happen, but I'm too lazy to think about it */
	21	if (cpu >= nr_cpu_ids)
	22	return;
	23
	24	cpumask_clear_cpu(cpu, nmsk);
	25	cpumask_set_cpu(cpu, irqmsk);
	26	cpus_per_vec--;
	27
	28	/* If the cpu has siblings, use them first */
	29	siblmsk = topology_sibling_cpumask(cpu);
	30	for (sibl = -1; cpus_per_vec > 0; ) {
	31	sibl = cpumask_next(sibl, siblmsk);
	32	if (sibl >= nr_cpu_ids)
	33	break;
	34	if (!cpumask_test_and_clear_cpu(sibl, nmsk))
	35	continue;
	36	cpumask_set_cpu(sibl, irqmsk);
	37	cpus_per_vec--;
	38	}
	39	}
	40	}
	41
c0936d9d	42	static cpumask_var_t *alloc_node_to_cpumask(void)
9a0ef98e CH	43	{
	44	cpumask_var_t *masks;
	45	int node;
	46
	47	masks = kcalloc(nr_node_ids, sizeof(cpumask_var_t), GFP_KERNEL);
	48	if (!masks)
	49	return NULL;
	50
	51	for (node = 0; node < nr_node_ids; node++) {
	52	if (!zalloc_cpumask_var(&masks[node], GFP_KERNEL))
	53	goto out_unwind;
	54	}
	55
	56	return masks;
	57
	58	out_unwind:
	59	while (--node >= 0)
	60	free_cpumask_var(masks[node]);
	61	kfree(masks);
	62	return NULL;
	63	}
	64
c0936d9d	65	static void free_node_to_cpumask(cpumask_var_t *masks)
9a0ef98e CH	66	{
	67	int node;
	68
	69	for (node = 0; node < nr_node_ids; node++)
	70	free_cpumask_var(masks[node]);
	71	kfree(masks);
	72	}
	73
c0936d9d	74	static void build_node_to_cpumask(cpumask_var_t *masks)
9a0ef98e CH	75	{
	76	int cpu;
	77
c189f43b	78	for_each_possible_cpu(cpu)
9a0ef98e CH	79	cpumask_set_cpu(cpu, masks[cpu_to_node(cpu)]);
	80	}
	81
c0936d9d	82	static int get_nodes_in_cpumask(cpumask_var_t *node_to_cpumask,
9a0ef98e	83	const struct cpumask mask, nodemask_t nodemsk)
34c3d981	84	{
c0af5243	85	int n, nodes = 0;
34c3d981 TG	86
34c3d981 TG	87	/* Calculate the number of nodes in the supplied affinity mask */
9a0ef98e	88	for_each_node(n) {
c0936d9d	89	if (cpumask_intersects(mask, node_to_cpumask[n])) {
34c3d981 TG	90	node_set(n, *nodemsk);
	91	nodes++;
	92	}
	93	}
	94	return nodes;
	95	}
	96
83a604e6 ML	97	static int irq_build_affinity_masks(const struct irq_affinity *affd,
83a604e6 ML	98	int startvec, int numvecs,
c192c97f ML	99	cpumask_var_t *node_to_cpumask,
	100	const struct cpumask *cpu_mask,
	101	struct cpumask *nmsk,
	102	struct cpumask *masks)
34c3d981	103	{
83a604e6 ML	104	int n, nodes, cpus_per_vec, extra_vecs, done = 0;
	105	int last_affv = affd->pre_vectors + numvecs;
	106	int curvec = startvec;
34c3d981	107	nodemask_t nodemsk = NODE_MASK_NONE;
34c3d981	108
2a4c050c ML	109	if (!cpumask_weight(cpu_mask))
	110	return 0;
	111
c192c97f	112	nodes = get_nodes_in_cpumask(node_to_cpumask, cpu_mask, &nodemsk);
34c3d981 TG	113
34c3d981 TG	114	/*
c0af5243	115	* If the number of nodes in the mask is greater than or equal the
34c3d981 TG	116	* number of vectors we just spread the vectors across the nodes.
34c3d981 TG	117	*/
83a604e6	118	if (numvecs <= nodes) {
34c3d981	119	for_each_node_mask(n, nodemsk) {
ea8a829d	120	cpumask_or(masks + curvec, masks + curvec, node_to_cpumask[n]);
83a604e6 ML	121	if (++curvec == last_affv)
83a604e6 ML	122	curvec = affd->pre_vectors;
34c3d981	123	}
ea8a829d	124	done = numvecs;
c192c97f	125	goto out;
34c3d981 TG	126	}
34c3d981 TG	127
34c3d981	128	for_each_node_mask(n, nodemsk) {
7bf8222b KB	129	int ncpus, v, vecs_to_assign, vecs_per_node;
	130
	131	/* Spread the vectors per node */
83a604e6	132	vecs_per_node = (numvecs - (curvec - affd->pre_vectors)) / nodes;
34c3d981 TG	133
34c3d981 TG	134	/* Get the cpus on this node which are in the mask */
c192c97f	135	cpumask_and(nmsk, cpu_mask, node_to_cpumask[n]);
34c3d981 TG	136
	137	/* Calculate the number of cpus per vector */
	138	ncpus = cpumask_weight(nmsk);
7bf8222b KB	139	vecs_to_assign = min(vecs_per_node, ncpus);
	140
	141	/* Account for rounding errors */
3412386b	142	extra_vecs = ncpus - vecs_to_assign * (ncpus / vecs_to_assign);
34c3d981	143
bfe13077 CH	144	for (v = 0; curvec < last_affv && v < vecs_to_assign;
bfe13077 CH	145	curvec++, v++) {
34c3d981 TG	146	cpus_per_vec = ncpus / vecs_to_assign;
	147
	148	/* Account for extra vectors to compensate rounding errors */
	149	if (extra_vecs) {
	150	cpus_per_vec++;
7bf8222b	151	--extra_vecs;
34c3d981 TG	152	}
	153	irq_spread_init_one(masks + curvec, nmsk, cpus_per_vec);
	154	}
	155
83a604e6 ML	156	done += v;
83a604e6 ML	157	if (done >= numvecs)
34c3d981	158	break;
83a604e6 ML	159	if (curvec >= last_affv)
83a604e6 ML	160	curvec = affd->pre_vectors;
7bf8222b	161	--nodes;
34c3d981 TG	162	}
34c3d981 TG	163
c192c97f	164	out:
83a604e6	165	return done;
c192c97f ML	166	}
	167
	168	/**
	169	* irq_create_affinity_masks - Create affinity masks for multiqueue spreading
	170	* @nvecs: The total number of vectors
	171	* @affd: Description of the affinity requirements
	172	*
	173	* Returns the masks pointer or NULL if allocation failed.
	174	*/
	175	struct cpumask *
	176	irq_create_affinity_masks(int nvecs, const struct irq_affinity *affd)
	177	{
2a4c050c ML	178	int affvecs = nvecs - affd->pre_vectors - affd->post_vectors;
	179	int curvec, usedvecs;
	180	cpumask_var_t nmsk, npresmsk, *node_to_cpumask;
c192c97f	181	struct cpumask *masks = NULL;
c192c97f ML	182
	183	/*
	184	* If there aren't any vectors left after applying the pre/post
	185	* vectors don't bother with assigning affinity.
	186	*/
	187	if (nvecs == affd->pre_vectors + affd->post_vectors)
	188	return NULL;
	189
	190	if (!zalloc_cpumask_var(&nmsk, GFP_KERNEL))
	191	return NULL;
	192
2a4c050c ML	193	if (!zalloc_cpumask_var(&npresmsk, GFP_KERNEL))
	194	goto outcpumsk;
	195
c192c97f ML	196	node_to_cpumask = alloc_node_to_cpumask();
c192c97f ML	197	if (!node_to_cpumask)
2a4c050c	198	goto outnpresmsk;
c192c97f ML	199
	200	masks = kcalloc(nvecs, sizeof(*masks), GFP_KERNEL);
	201	if (!masks)
	202	goto outnodemsk;
	203
	204	/* Fill out vectors at the beginning that don't need affinity */
	205	for (curvec = 0; curvec < affd->pre_vectors; curvec++)
	206	cpumask_copy(masks + curvec, irq_default_affinity);
	207
	208	/* Stabilize the cpumasks */
	209	get_online_cpus();
	210	build_node_to_cpumask(node_to_cpumask);
2a4c050c ML	211
	212	/* Spread on present CPUs starting from affd->pre_vectors */
	213	usedvecs = irq_build_affinity_masks(affd, curvec, affvecs,
	214	node_to_cpumask, cpu_present_mask,
	215	nmsk, masks);
	216
	217	/*
	218	* Spread on non present CPUs starting from the next vector to be
	219	* handled. If the spreading of present CPUs already exhausted the
	220	* vector space, assign the non present CPUs to the already spread
	221	* out vectors.
	222	*/
	223	if (usedvecs >= affvecs)
	224	curvec = affd->pre_vectors;
	225	else
	226	curvec = affd->pre_vectors + usedvecs;
	227	cpumask_andnot(npresmsk, cpu_possible_mask, cpu_present_mask);
	228	usedvecs += irq_build_affinity_masks(affd, curvec, affvecs,
	229	node_to_cpumask, npresmsk,
	230	nmsk, masks);
34c3d981	231	put_online_cpus();
67c93c21 CH	232
67c93c21 CH	233	/* Fill out vectors at the end that don't need affinity */
2a4c050c ML	234	if (usedvecs >= affvecs)
	235	curvec = affd->pre_vectors + affvecs;
	236	else
	237	curvec = affd->pre_vectors + usedvecs;
67c93c21	238	for (; curvec < nvecs; curvec++)
b6e5d5b9	239	cpumask_copy(masks + curvec, irq_default_affinity);
2a4c050c	240
1efd89a3	241	outnodemsk:
c0936d9d	242	free_node_to_cpumask(node_to_cpumask);
2a4c050c ML	243	outnpresmsk:
2a4c050c ML	244	free_cpumask_var(npresmsk);
1efd89a3	245	outcpumsk:
34c3d981 TG	246	free_cpumask_var(nmsk);
	247	return masks;
	248	}
	249
	250	/**
212bd846	251	* irq_calc_affinity_vectors - Calculate the optimal number of vectors
6f9a22bc	252	* @minvec: The minimum number of vectors available
212bd846 CH	253	* @maxvec: The maximum number of vectors available
212bd846 CH	254	* @affd: Description of the affinity requirements
34c3d981	255	*/
6f9a22bc	256	int irq_calc_affinity_vectors(int minvec, int maxvec, const struct irq_affinity *affd)
34c3d981	257	{
212bd846 CH	258	int resv = affd->pre_vectors + affd->post_vectors;
212bd846 CH	259	int vecs = maxvec - resv;
9a0ef98e	260	int ret;
34c3d981	261
6f9a22bc MH	262	if (resv > minvec)
	263	return 0;
	264
34c3d981	265	get_online_cpus();
c189f43b	266	ret = min_t(int, cpumask_weight(cpu_possible_mask), vecs) + resv;
34c3d981	267	put_online_cpus();
9a0ef98e	268	return ret;
34c3d981	269	}