[linux-block.git] / block / blk-ioprio.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Block rq-qos policy for assigning an I/O priority class to requests.
 *
 * Using an rq-qos policy for assigning I/O priority class has two advantages
 * over using the ioprio_set() system call:
 *
 * - This policy is cgroup based so it has all the advantages of cgroups.
 * - While ioprio_set() does not affect page cache writeback I/O, this rq-qos
 *   controller affects page cache writeback I/O for filesystems that support
 *   assiociating a cgroup with writeback I/O. See also
 *   Documentation/admin-guide/cgroup-v2.rst.
 */

#include <linux/blk-mq.h>
#include <linux/blk_types.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include "blk-cgroup.h"
#include "blk-ioprio.h"
#include "blk-rq-qos.h"

/**
 * enum prio_policy - I/O priority class policy.
 * @POLICY_NO_CHANGE: (default) do not modify the I/O priority class.
 * @POLICY_PROMOTE_TO_RT: modify no-IOPRIO_CLASS_RT to IOPRIO_CLASS_RT.
 * @POLICY_RESTRICT_TO_BE: modify IOPRIO_CLASS_NONE and IOPRIO_CLASS_RT into
 *		IOPRIO_CLASS_BE.
 * @POLICY_ALL_TO_IDLE: change the I/O priority class into IOPRIO_CLASS_IDLE.
 * @POLICY_NONE_TO_RT: an alias for POLICY_PROMOTE_TO_RT.
 *
 * See also <linux/ioprio.h>.
 */
enum prio_policy {
	POLICY_NO_CHANGE	= 0,
	POLICY_PROMOTE_TO_RT	= 1,
	POLICY_RESTRICT_TO_BE	= 2,
	POLICY_ALL_TO_IDLE	= 3,
	POLICY_NONE_TO_RT	= 4,
};

static const char *policy_name[] = {
	[POLICY_NO_CHANGE]	= "no-change",
	[POLICY_PROMOTE_TO_RT]	= "promote-to-rt",
	[POLICY_RESTRICT_TO_BE]	= "restrict-to-be",
	[POLICY_ALL_TO_IDLE]	= "idle",
	[POLICY_NONE_TO_RT]	= "none-to-rt",
};

static struct blkcg_policy ioprio_policy;

/**
 * struct ioprio_blkg - Per (cgroup, request queue) data.
 * @pd: blkg_policy_data structure.
 */
struct ioprio_blkg {
	struct blkg_policy_data pd;
};

/**
 * struct ioprio_blkcg - Per cgroup data.
 * @cpd: blkcg_policy_data structure.
 * @prio_policy: One of the IOPRIO_CLASS_* values. See also <linux/ioprio.h>.
 */
struct ioprio_blkcg {
	struct blkcg_policy_data cpd;
	enum prio_policy	 prio_policy;
};

static inline struct ioprio_blkg *pd_to_ioprio(struct blkg_policy_data *pd)
{
	return pd ? container_of(pd, struct ioprio_blkg, pd) : NULL;
}

static struct ioprio_blkcg *blkcg_to_ioprio_blkcg(struct blkcg *blkcg)
{
	return container_of(blkcg_to_cpd(blkcg, &ioprio_policy),
			    struct ioprio_blkcg, cpd);
}

static struct ioprio_blkcg *
ioprio_blkcg_from_css(struct cgroup_subsys_state *css)
{
	return blkcg_to_ioprio_blkcg(css_to_blkcg(css));
}

static struct ioprio_blkcg *ioprio_blkcg_from_bio(struct bio *bio)
{
	struct blkg_policy_data *pd = blkg_to_pd(bio->bi_blkg, &ioprio_policy);

	if (!pd)
		return NULL;

	return blkcg_to_ioprio_blkcg(pd->blkg->blkcg);
}

static int ioprio_show_prio_policy(struct seq_file *sf, void *v)
{
	struct ioprio_blkcg *blkcg = ioprio_blkcg_from_css(seq_css(sf));

	seq_printf(sf, "%s\n", policy_name[blkcg->prio_policy]);
	return 0;
}

static ssize_t ioprio_set_prio_policy(struct kernfs_open_file *of, char *buf,
				      size_t nbytes, loff_t off)
{
	struct ioprio_blkcg *blkcg = ioprio_blkcg_from_css(of_css(of));
	int ret;

	if (off != 0)
		return -EIO;
	/* kernfs_fop_write_iter() terminates 'buf' with '\0'. */
	ret = sysfs_match_string(policy_name, buf);
	if (ret < 0)
		return ret;
	blkcg->prio_policy = ret;
	return nbytes;
}

static struct blkg_policy_data *
ioprio_alloc_pd(struct gendisk *disk, struct blkcg *blkcg, gfp_t gfp)
{
	struct ioprio_blkg *ioprio_blkg;

	ioprio_blkg = kzalloc(sizeof(*ioprio_blkg), gfp);
	if (!ioprio_blkg)
		return NULL;

	return &ioprio_blkg->pd;
}

static void ioprio_free_pd(struct blkg_policy_data *pd)
{
	struct ioprio_blkg *ioprio_blkg = pd_to_ioprio(pd);

	kfree(ioprio_blkg);
}

static struct blkcg_policy_data *ioprio_alloc_cpd(gfp_t gfp)
{
	struct ioprio_blkcg *blkcg;

	blkcg = kzalloc(sizeof(*blkcg), gfp);
	if (!blkcg)
		return NULL;
	blkcg->prio_policy = POLICY_NO_CHANGE;
	return &blkcg->cpd;
}

static void ioprio_free_cpd(struct blkcg_policy_data *cpd)
{
	struct ioprio_blkcg *blkcg = container_of(cpd, typeof(*blkcg), cpd);

	kfree(blkcg);
}

#define IOPRIO_ATTRS						\
	{							\
		.name		= "prio.class",			\
		.seq_show	= ioprio_show_prio_policy,	\
		.write		= ioprio_set_prio_policy,	\
	},							\
	{ } /* sentinel */

/* cgroup v2 attributes */
static struct cftype ioprio_files[] = {
	IOPRIO_ATTRS
};

/* cgroup v1 attributes */
static struct cftype ioprio_legacy_files[] = {
	IOPRIO_ATTRS
};

static struct blkcg_policy ioprio_policy = {
	.dfl_cftypes	= ioprio_files,
	.legacy_cftypes = ioprio_legacy_files,

	.cpd_alloc_fn	= ioprio_alloc_cpd,
	.cpd_free_fn	= ioprio_free_cpd,

	.pd_alloc_fn	= ioprio_alloc_pd,
	.pd_free_fn	= ioprio_free_pd,
};

void blkcg_set_ioprio(struct bio *bio)
{
	struct ioprio_blkcg *blkcg = ioprio_blkcg_from_bio(bio);
	u16 prio;

	if (!blkcg || blkcg->prio_policy == POLICY_NO_CHANGE)
		return;

	if (blkcg->prio_policy == POLICY_PROMOTE_TO_RT ||
	    blkcg->prio_policy == POLICY_NONE_TO_RT) {
		/*
		 * For RT threads, the default priority level is 4 because
		 * task_nice is 0. By promoting non-RT io-priority to RT-class
		 * and default level 4, those requests that are already
		 * RT-class but need a higher io-priority can use ioprio_set()
		 * to achieve this.
		 */
		if (IOPRIO_PRIO_CLASS(bio->bi_ioprio) != IOPRIO_CLASS_RT)
			bio->bi_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_RT, 4);
		return;
	}

	/*
	 * Except for IOPRIO_CLASS_NONE, higher I/O priority numbers
	 * correspond to a lower priority. Hence, the max_t() below selects
	 * the lower priority of bi_ioprio and the cgroup I/O priority class.
	 * If the bio I/O priority equals IOPRIO_CLASS_NONE, the cgroup I/O
	 * priority is assigned to the bio.
	 */
	prio = max_t(u16, bio->bi_ioprio,
			IOPRIO_PRIO_VALUE(blkcg->prio_policy, 0));
	if (prio > bio->bi_ioprio)
		bio->bi_ioprio = prio;
}

void blk_ioprio_exit(struct gendisk *disk)
{
	blkcg_deactivate_policy(disk, &ioprio_policy);
}

int blk_ioprio_init(struct gendisk *disk)
{
	return blkcg_activate_policy(disk, &ioprio_policy);
}

static int __init ioprio_init(void)
{
	return blkcg_policy_register(&ioprio_policy);
}

static void __exit ioprio_exit(void)
{
	blkcg_policy_unregister(&ioprio_policy);
}

module_init(ioprio_init);
module_exit(ioprio_exit);
Commit	Line	Data
556910e3 BVA	1	// SPDX-License-Identifier: GPL-2.0
	2	/*
	3	* Block rq-qos policy for assigning an I/O priority class to requests.
	4	*
	5	* Using an rq-qos policy for assigning I/O priority class has two advantages
	6	* over using the ioprio_set() system call:
	7	*
	8	* - This policy is cgroup based so it has all the advantages of cgroups.
	9	* - While ioprio_set() does not affect page cache writeback I/O, this rq-qos
	10	* controller affects page cache writeback I/O for filesystems that support
	11	* assiociating a cgroup with writeback I/O. See also
	12	* Documentation/admin-guide/cgroup-v2.rst.
	13	*/
	14
556910e3 BVA	15	#include <linux/blk-mq.h>
	16	#include <linux/blk_types.h>
	17	#include <linux/kernel.h>
	18	#include <linux/module.h>
672fdcf0	19	#include "blk-cgroup.h"
556910e3 BVA	20	#include "blk-ioprio.h"
	21	#include "blk-rq-qos.h"
	22
	23	/**
	24	* enum prio_policy - I/O priority class policy.
	25	* @POLICY_NO_CHANGE: (default) do not modify the I/O priority class.
ddf63516	26	* @POLICY_PROMOTE_TO_RT: modify no-IOPRIO_CLASS_RT to IOPRIO_CLASS_RT.
556910e3 BVA	27	* @POLICY_RESTRICT_TO_BE: modify IOPRIO_CLASS_NONE and IOPRIO_CLASS_RT into
	28	* IOPRIO_CLASS_BE.
	29	* @POLICY_ALL_TO_IDLE: change the I/O priority class into IOPRIO_CLASS_IDLE.
ddf63516	30	* @POLICY_NONE_TO_RT: an alias for POLICY_PROMOTE_TO_RT.
556910e3 BVA	31	*
	32	* See also <linux/ioprio.h>.
	33	*/
	34	enum prio_policy {
	35	POLICY_NO_CHANGE = 0,
ddf63516	36	POLICY_PROMOTE_TO_RT = 1,
556910e3 BVA	37	POLICY_RESTRICT_TO_BE = 2,
556910e3 BVA	38	POLICY_ALL_TO_IDLE = 3,
ddf63516	39	POLICY_NONE_TO_RT = 4,
556910e3 BVA	40	};
	41
	42	static const char *policy_name[] = {
	43	[POLICY_NO_CHANGE] = "no-change",
ddf63516	44	[POLICY_PROMOTE_TO_RT] = "promote-to-rt",
556910e3 BVA	45	[POLICY_RESTRICT_TO_BE] = "restrict-to-be",
556910e3 BVA	46	[POLICY_ALL_TO_IDLE] = "idle",
ddf63516	47	[POLICY_NONE_TO_RT] = "none-to-rt",
556910e3 BVA	48	};
	49
	50	static struct blkcg_policy ioprio_policy;
	51
	52	/**
	53	* struct ioprio_blkg - Per (cgroup, request queue) data.
	54	* @pd: blkg_policy_data structure.
	55	*/
	56	struct ioprio_blkg {
	57	struct blkg_policy_data pd;
	58	};
	59
	60	/**
	61	* struct ioprio_blkcg - Per cgroup data.
	62	* @cpd: blkcg_policy_data structure.
	63	* @prio_policy: One of the IOPRIO_CLASS_* values. See also <linux/ioprio.h>.
	64	*/
	65	struct ioprio_blkcg {
	66	struct blkcg_policy_data cpd;
	67	enum prio_policy prio_policy;
	68	};
	69
	70	static inline struct ioprio_blkg pd_to_ioprio(struct blkg_policy_data pd)
	71	{
	72	return pd ? container_of(pd, struct ioprio_blkg, pd) : NULL;
	73	}
	74
	75	static struct ioprio_blkcg blkcg_to_ioprio_blkcg(struct blkcg blkcg)
	76	{
	77	return container_of(blkcg_to_cpd(blkcg, &ioprio_policy),
	78	struct ioprio_blkcg, cpd);
	79	}
	80
	81	static struct ioprio_blkcg *
	82	ioprio_blkcg_from_css(struct cgroup_subsys_state *css)
	83	{
	84	return blkcg_to_ioprio_blkcg(css_to_blkcg(css));
	85	}
	86
	87	static struct ioprio_blkcg ioprio_blkcg_from_bio(struct bio bio)
	88	{
	89	struct blkg_policy_data *pd = blkg_to_pd(bio->bi_blkg, &ioprio_policy);
	90
	91	if (!pd)
	92	return NULL;
	93
	94	return blkcg_to_ioprio_blkcg(pd->blkg->blkcg);
	95	}
	96
	97	static int ioprio_show_prio_policy(struct seq_file sf, void v)
	98	{
	99	struct ioprio_blkcg *blkcg = ioprio_blkcg_from_css(seq_css(sf));
	100
	101	seq_printf(sf, "%s\n", policy_name[blkcg->prio_policy]);
	102	return 0;
	103	}
	104
	105	static ssize_t ioprio_set_prio_policy(struct kernfs_open_file of, char buf,
	106	size_t nbytes, loff_t off)
	107	{
	108	struct ioprio_blkcg *blkcg = ioprio_blkcg_from_css(of_css(of));
	109	int ret;
	110
	111	if (off != 0)
112	return -EIO;
113	/* kernfs_fop_write_iter() terminates 'buf' with '\0'. */
114	ret = sysfs_match_string(policy_name, buf);
115	if (ret < 0)
116	return ret;
117	blkcg->prio_policy = ret;
556910e3 BVA	118	return nbytes;
	119	}
	120
	121	static struct blkg_policy_data *
0a0b4f79	122	ioprio_alloc_pd(struct gendisk disk, struct blkcg blkcg, gfp_t gfp)
556910e3 BVA	123	{
	124	struct ioprio_blkg *ioprio_blkg;
	125
	126	ioprio_blkg = kzalloc(sizeof(*ioprio_blkg), gfp);
	127	if (!ioprio_blkg)
	128	return NULL;
	129
	130	return &ioprio_blkg->pd;
	131	}
	132
	133	static void ioprio_free_pd(struct blkg_policy_data *pd)
	134	{
	135	struct ioprio_blkg *ioprio_blkg = pd_to_ioprio(pd);
	136
	137	kfree(ioprio_blkg);
	138	}
	139
	140	static struct blkcg_policy_data *ioprio_alloc_cpd(gfp_t gfp)
	141	{
	142	struct ioprio_blkcg *blkcg;
	143
	144	blkcg = kzalloc(sizeof(*blkcg), gfp);
	145	if (!blkcg)
	146	return NULL;
	147	blkcg->prio_policy = POLICY_NO_CHANGE;
	148	return &blkcg->cpd;
	149	}
	150
	151	static void ioprio_free_cpd(struct blkcg_policy_data *cpd)
	152	{
	153	struct ioprio_blkcg blkcg = container_of(cpd, typeof(blkcg), cpd);
	154
	155	kfree(blkcg);
	156	}
	157
	158	#define IOPRIO_ATTRS \
	159	{ \
	160	.name = "prio.class", \
	161	.seq_show = ioprio_show_prio_policy, \
	162	.write = ioprio_set_prio_policy, \
	163	}, \
	164	{ } /* sentinel */
	165
	166	/* cgroup v2 attributes */
	167	static struct cftype ioprio_files[] = {
	168	IOPRIO_ATTRS
	169	};
	170
	171	/* cgroup v1 attributes */
	172	static struct cftype ioprio_legacy_files[] = {
	173	IOPRIO_ATTRS
	174	};
	175
	176	static struct blkcg_policy ioprio_policy = {
	177	.dfl_cftypes = ioprio_files,
	178	.legacy_cftypes = ioprio_legacy_files,
	179
	180	.cpd_alloc_fn = ioprio_alloc_cpd,
	181	.cpd_free_fn = ioprio_free_cpd,
	182
	183	.pd_alloc_fn = ioprio_alloc_pd,
	184	.pd_free_fn = ioprio_free_pd,
	185	};
	186
82b74cac	187	void blkcg_set_ioprio(struct bio *bio)
556910e3 BVA	188	{
556910e3 BVA	189	struct ioprio_blkcg *blkcg = ioprio_blkcg_from_bio(bio);
25c4b5e0 JA	190	u16 prio;
25c4b5e0 JA	191
82b74cac	192	if (!blkcg \|\| blkcg->prio_policy == POLICY_NO_CHANGE)
25c4b5e0	193	return;
556910e3	194
ddf63516 HT	195	if (blkcg->prio_policy == POLICY_PROMOTE_TO_RT \|\|
	196	blkcg->prio_policy == POLICY_NONE_TO_RT) {
	197	/*
	198	* For RT threads, the default priority level is 4 because
	199	* task_nice is 0. By promoting non-RT io-priority to RT-class
	200	* and default level 4, those requests that are already
	201	* RT-class but need a higher io-priority can use ioprio_set()
	202	* to achieve this.
	203	*/
	204	if (IOPRIO_PRIO_CLASS(bio->bi_ioprio) != IOPRIO_CLASS_RT)
	205	bio->bi_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_RT, 4);
	206	return;
	207	}
	208
556910e3 BVA	209	/*
	210	* Except for IOPRIO_CLASS_NONE, higher I/O priority numbers
	211	* correspond to a lower priority. Hence, the max_t() below selects
	212	* the lower priority of bi_ioprio and the cgroup I/O priority class.
f2586544 JK	213	* If the bio I/O priority equals IOPRIO_CLASS_NONE, the cgroup I/O
f2586544 JK	214	* priority is assigned to the bio.
556910e3	215	*/
25c4b5e0 JA	216	prio = max_t(u16, bio->bi_ioprio,
	217	IOPRIO_PRIO_VALUE(blkcg->prio_policy, 0));
	218	if (prio > bio->bi_ioprio)
	219	bio->bi_ioprio = prio;
556910e3 BVA	220	}
556910e3 BVA	221
b0dde3f5	222	void blk_ioprio_exit(struct gendisk *disk)
556910e3	223	{
40e4996e	224	blkcg_deactivate_policy(disk, &ioprio_policy);
556910e3 BVA	225	}
556910e3 BVA	226
b0dde3f5	227	int blk_ioprio_init(struct gendisk *disk)
556910e3	228	{
40e4996e	229	return blkcg_activate_policy(disk, &ioprio_policy);
556910e3 BVA	230	}
	231
	232	static int __init ioprio_init(void)
	233	{
	234	return blkcg_policy_register(&ioprio_policy);
	235	}
	236
	237	static void __exit ioprio_exit(void)
	238	{
	239	blkcg_policy_unregister(&ioprio_policy);
	240	}
	241
	242	module_init(ioprio_init);
	243	module_exit(ioprio_exit);