[linux-block.git] / block / blk-throttle.h

#ifndef BLK_THROTTLE_H
#define BLK_THROTTLE_H

#include "blk-cgroup-rwstat.h"

/*
 * To implement hierarchical throttling, throtl_grps form a tree and bios
 * are dispatched upwards level by level until they reach the top and get
 * issued.  When dispatching bios from the children and local group at each
 * level, if the bios are dispatched into a single bio_list, there's a risk
 * of a local or child group which can queue many bios at once filling up
 * the list starving others.
 *
 * To avoid such starvation, dispatched bios are queued separately
 * according to where they came from.  When they are again dispatched to
 * the parent, they're popped in round-robin order so that no single source
 * hogs the dispatch window.
 *
 * throtl_qnode is used to keep the queued bios separated by their sources.
 * Bios are queued to throtl_qnode which in turn is queued to
 * throtl_service_queue and then dispatched in round-robin order.
 *
 * It's also used to track the reference counts on blkg's.  A qnode always
 * belongs to a throtl_grp and gets queued on itself or the parent, so
 * incrementing the reference of the associated throtl_grp when a qnode is
 * queued and decrementing when dequeued is enough to keep the whole blkg
 * tree pinned while bios are in flight.
 */
struct throtl_qnode {
	struct list_head	node;		/* service_queue->queued[] */
	struct bio_list		bios;		/* queued bios */
	struct throtl_grp	*tg;		/* tg this qnode belongs to */
};

struct throtl_service_queue {
	struct throtl_service_queue *parent_sq;	/* the parent service_queue */

	/*
	 * Bios queued directly to this service_queue or dispatched from
	 * children throtl_grp's.
	 */
	struct list_head	queued[2];	/* throtl_qnode [READ/WRITE] */
	unsigned int		nr_queued[2];	/* number of queued bios */

	/*
	 * RB tree of active children throtl_grp's, which are sorted by
	 * their ->disptime.
	 */
	struct rb_root_cached	pending_tree;	/* RB tree of active tgs */
	unsigned int		nr_pending;	/* # queued in the tree */
	unsigned long		first_pending_disptime;	/* disptime of the first tg */
	struct timer_list	pending_timer;	/* fires on first_pending_disptime */
};

enum tg_state_flags {
	THROTL_TG_PENDING	= 1 << 0,	/* on parent's pending tree */
	THROTL_TG_WAS_EMPTY	= 1 << 1,	/* bio_lists[] became non-empty */
	THROTL_TG_CANCELING	= 1 << 2,	/* starts to cancel bio */
};

struct throtl_grp {
	/* must be the first member */
	struct blkg_policy_data pd;

	/* active throtl group service_queue member */
	struct rb_node rb_node;

	/* throtl_data this group belongs to */
	struct throtl_data *td;

	/* this group's service queue */
	struct throtl_service_queue service_queue;

	/*
	 * qnode_on_self is used when bios are directly queued to this
	 * throtl_grp so that local bios compete fairly with bios
	 * dispatched from children.  qnode_on_parent is used when bios are
	 * dispatched from this throtl_grp into its parent and will compete
	 * with the sibling qnode_on_parents and the parent's
	 * qnode_on_self.
	 */
	struct throtl_qnode qnode_on_self[2];
	struct throtl_qnode qnode_on_parent[2];

	/*
	 * Dispatch time in jiffies. This is the estimated time when group
	 * will unthrottle and is ready to dispatch more bio. It is used as
	 * key to sort active groups in service tree.
	 */
	unsigned long disptime;

	unsigned int flags;

	/* are there any throtl rules between this group and td? */
	bool has_rules_bps[2];
	bool has_rules_iops[2];

	/* internally used bytes per second rate limits */
	uint64_t bps[2];
	/* user configured bps limits */
	uint64_t bps_conf[2];

	/* internally used IOPS limits */
	unsigned int iops[2];
	/* user configured IOPS limits */
	unsigned int iops_conf[2];

	/* Number of bytes dispatched in current slice */
	uint64_t bytes_disp[2];
	/* Number of bio's dispatched in current slice */
	unsigned int io_disp[2];

	unsigned long last_low_overflow_time[2];

	uint64_t last_bytes_disp[2];
	unsigned int last_io_disp[2];

	/*
	 * The following two fields are updated when new configuration is
	 * submitted while some bios are still throttled, they record how many
	 * bytes/ios are waited already in previous configuration, and they will
	 * be used to calculate wait time under new configuration.
	 */
	long long carryover_bytes[2];
	int carryover_ios[2];

	unsigned long last_check_time;

	/* When did we start a new slice */
	unsigned long slice_start[2];
	unsigned long slice_end[2];

	struct blkg_rwstat stat_bytes;
	struct blkg_rwstat stat_ios;
};

extern struct blkcg_policy blkcg_policy_throtl;

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

static inline struct throtl_grp *blkg_to_tg(struct blkcg_gq *blkg)
{
	return pd_to_tg(blkg_to_pd(blkg, &blkcg_policy_throtl));
}

/*
 * Internal throttling interface
 */
#ifndef CONFIG_BLK_DEV_THROTTLING
static inline void blk_throtl_exit(struct gendisk *disk) { }
static inline bool blk_throtl_bio(struct bio *bio) { return false; }
static inline void blk_throtl_cancel_bios(struct gendisk *disk) { }
#else /* CONFIG_BLK_DEV_THROTTLING */
void blk_throtl_exit(struct gendisk *disk);
bool __blk_throtl_bio(struct bio *bio);
void blk_throtl_cancel_bios(struct gendisk *disk);

static inline bool blk_throtl_activated(struct request_queue *q)
{
	return q->td != NULL;
}

static inline bool blk_should_throtl(struct bio *bio)
{
	struct throtl_grp *tg;
	int rw = bio_data_dir(bio);

	/*
	 * This is called under bio_queue_enter(), and it's synchronized with
	 * the activation of blk-throtl, which is protected by
	 * blk_mq_freeze_queue().
	 */
	if (!blk_throtl_activated(bio->bi_bdev->bd_queue))
		return false;

	tg = blkg_to_tg(bio->bi_blkg);
	if (!cgroup_subsys_on_dfl(io_cgrp_subsys)) {
		if (!bio_flagged(bio, BIO_CGROUP_ACCT)) {
			bio_set_flag(bio, BIO_CGROUP_ACCT);
			blkg_rwstat_add(&tg->stat_bytes, bio->bi_opf,
					bio->bi_iter.bi_size);
		}
		blkg_rwstat_add(&tg->stat_ios, bio->bi_opf, 1);
	}

	/* iops limit is always counted */
	if (tg->has_rules_iops[rw])
		return true;

	if (tg->has_rules_bps[rw] && !bio_flagged(bio, BIO_BPS_THROTTLED))
		return true;

	return false;
}

static inline bool blk_throtl_bio(struct bio *bio)
{

	if (!blk_should_throtl(bio))
		return false;

	return __blk_throtl_bio(bio);
}
#endif /* CONFIG_BLK_DEV_THROTTLING */

#endif
Commit	Line	Data
a7b36ee6 JA	1	#ifndef BLK_THROTTLE_H
	2	#define BLK_THROTTLE_H
	3
	4	#include "blk-cgroup-rwstat.h"
	5
	6	/*
	7	* To implement hierarchical throttling, throtl_grps form a tree and bios
	8	* are dispatched upwards level by level until they reach the top and get
	9	* issued. When dispatching bios from the children and local group at each
	10	* level, if the bios are dispatched into a single bio_list, there's a risk
	11	* of a local or child group which can queue many bios at once filling up
	12	* the list starving others.
	13	*
	14	* To avoid such starvation, dispatched bios are queued separately
	15	* according to where they came from. When they are again dispatched to
	16	* the parent, they're popped in round-robin order so that no single source
	17	* hogs the dispatch window.
	18	*
	19	* throtl_qnode is used to keep the queued bios separated by their sources.
	20	* Bios are queued to throtl_qnode which in turn is queued to
	21	* throtl_service_queue and then dispatched in round-robin order.
	22	*
	23	* It's also used to track the reference counts on blkg's. A qnode always
	24	* belongs to a throtl_grp and gets queued on itself or the parent, so
	25	* incrementing the reference of the associated throtl_grp when a qnode is
	26	* queued and decrementing when dequeued is enough to keep the whole blkg
	27	* tree pinned while bios are in flight.
	28	*/
	29	struct throtl_qnode {
	30	struct list_head node; /* service_queue->queued[] */
	31	struct bio_list bios; /* queued bios */
	32	struct throtl_grp tg; / tg this qnode belongs to */
	33	};
	34
	35	struct throtl_service_queue {
	36	struct throtl_service_queue parent_sq; / the parent service_queue */
	37
	38	/*
	39	* Bios queued directly to this service_queue or dispatched from
	40	* children throtl_grp's.
	41	*/
	42	struct list_head queued[2]; /* throtl_qnode [READ/WRITE] */
	43	unsigned int nr_queued[2]; /* number of queued bios */
	44
	45	/*
	46	* RB tree of active children throtl_grp's, which are sorted by
	47	* their ->disptime.
	48	*/
	49	struct rb_root_cached pending_tree; /* RB tree of active tgs */
	50	unsigned int nr_pending; /* # queued in the tree */
	51	unsigned long first_pending_disptime; /* disptime of the first tg */
	52	struct timer_list pending_timer; /* fires on first_pending_disptime */
	53	};
	54
5a93b602 ML	55	enum tg_state_flags {
	56	THROTL_TG_PENDING = 1 << 0, /* on parent's pending tree */
	57	THROTL_TG_WAS_EMPTY = 1 << 1, /* bio_lists[] became non-empty */
85496749	58	THROTL_TG_CANCELING = 1 << 2, /* starts to cancel bio */
5a93b602 ML	59	};
5a93b602 ML	60
a7b36ee6 JA	61	struct throtl_grp {
	62	/* must be the first member */
	63	struct blkg_policy_data pd;
	64
	65	/* active throtl group service_queue member */
	66	struct rb_node rb_node;
	67
	68	/* throtl_data this group belongs to */
	69	struct throtl_data *td;
	70
	71	/* this group's service queue */
	72	struct throtl_service_queue service_queue;
	73
	74	/*
	75	* qnode_on_self is used when bios are directly queued to this
	76	* throtl_grp so that local bios compete fairly with bios
	77	* dispatched from children. qnode_on_parent is used when bios are
	78	* dispatched from this throtl_grp into its parent and will compete
	79	* with the sibling qnode_on_parents and the parent's
	80	* qnode_on_self.
	81	*/
	82	struct throtl_qnode qnode_on_self[2];
	83	struct throtl_qnode qnode_on_parent[2];
	84
	85	/*
	86	* Dispatch time in jiffies. This is the estimated time when group
	87	* will unthrottle and is ready to dispatch more bio. It is used as
	88	* key to sort active groups in service tree.
	89	*/
	90	unsigned long disptime;
	91
	92	unsigned int flags;
	93
	94	/* are there any throtl rules between this group and td? */
81c7a63a YK	95	bool has_rules_bps[2];
81c7a63a YK	96	bool has_rules_iops[2];
a7b36ee6 JA	97
a7b36ee6 JA	98	/* internally used bytes per second rate limits */
bf20ab53	99	uint64_t bps[2];
a7b36ee6	100	/* user configured bps limits */
bf20ab53	101	uint64_t bps_conf[2];
a7b36ee6 JA	102
a7b36ee6 JA	103	/* internally used IOPS limits */
bf20ab53	104	unsigned int iops[2];
a7b36ee6	105	/* user configured IOPS limits */
bf20ab53	106	unsigned int iops_conf[2];
a7b36ee6 JA	107
	108	/* Number of bytes dispatched in current slice */
	109	uint64_t bytes_disp[2];
	110	/* Number of bio's dispatched in current slice */
	111	unsigned int io_disp[2];
	112
	113	unsigned long last_low_overflow_time[2];
	114
	115	uint64_t last_bytes_disp[2];
	116	unsigned int last_io_disp[2];
	117
a880ae93 YK	118	/*
	119	* The following two fields are updated when new configuration is
	120	* submitted while some bios are still throttled, they record how many
	121	* bytes/ios are waited already in previous configuration, and they will
	122	* be used to calculate wait time under new configuration.
	123	*/
ef100397 YK	124	long long carryover_bytes[2];
ef100397 YK	125	int carryover_ios[2];
a880ae93	126
a7b36ee6 JA	127	unsigned long last_check_time;
a7b36ee6 JA	128
a7b36ee6 JA	129	/* When did we start a new slice */
	130	unsigned long slice_start[2];
	131	unsigned long slice_end[2];
	132
a7b36ee6 JA	133	struct blkg_rwstat stat_bytes;
	134	struct blkg_rwstat stat_ios;
	135	};
	136
	137	extern struct blkcg_policy blkcg_policy_throtl;
	138
	139	static inline struct throtl_grp pd_to_tg(struct blkg_policy_data pd)
	140	{
	141	return pd ? container_of(pd, struct throtl_grp, pd) : NULL;
	142	}
	143
	144	static inline struct throtl_grp blkg_to_tg(struct blkcg_gq blkg)
	145	{
	146	return pd_to_tg(blkg_to_pd(blkg, &blkcg_policy_throtl));
	147	}
	148
	149	/*
	150	* Internal throttling interface
	151	*/
	152	#ifndef CONFIG_BLK_DEV_THROTTLING
e13793ba	153	static inline void blk_throtl_exit(struct gendisk *disk) { }
a7b36ee6	154	static inline bool blk_throtl_bio(struct bio *bio) { return false; }
cad9266a	155	static inline void blk_throtl_cancel_bios(struct gendisk *disk) { }
a7b36ee6	156	#else /* CONFIG_BLK_DEV_THROTTLING */
e13793ba	157	void blk_throtl_exit(struct gendisk *disk);
a7b36ee6	158	bool __blk_throtl_bio(struct bio *bio);
cad9266a	159	void blk_throtl_cancel_bios(struct gendisk *disk);
81c7a63a	160
a3166c51 YK	161	static inline bool blk_throtl_activated(struct request_queue *q)
	162	{
	163	return q->td != NULL;
	164	}
	165
81c7a63a	166	static inline bool blk_should_throtl(struct bio *bio)
a7b36ee6	167	{
a3166c51	168	struct throtl_grp *tg;
81c7a63a YK	169	int rw = bio_data_dir(bio);
81c7a63a YK	170
a3166c51 YK	171	/*
	172	* This is called under bio_queue_enter(), and it's synchronized with
	173	* the activation of blk-throtl, which is protected by
	174	* blk_mq_freeze_queue().
	175	*/
	176	if (!blk_throtl_activated(bio->bi_bdev->bd_queue))
	177	return false;
	178
	179	tg = blkg_to_tg(bio->bi_blkg);
ad7c3b41 JH	180	if (!cgroup_subsys_on_dfl(io_cgrp_subsys)) {
	181	if (!bio_flagged(bio, BIO_CGROUP_ACCT)) {
	182	bio_set_flag(bio, BIO_CGROUP_ACCT);
	183	blkg_rwstat_add(&tg->stat_bytes, bio->bi_opf,
	184	bio->bi_iter.bi_size);
	185	}
	186	blkg_rwstat_add(&tg->stat_ios, bio->bi_opf, 1);
	187	}
	188
81c7a63a YK	189	/* iops limit is always counted */
	190	if (tg->has_rules_iops[rw])
	191	return true;
	192
	193	if (tg->has_rules_bps[rw] && !bio_flagged(bio, BIO_BPS_THROTTLED))
	194	return true;
	195
	196	return false;
	197	}
	198
	199	static inline bool blk_throtl_bio(struct bio *bio)
	200	{
a7b36ee6	201
81c7a63a	202	if (!blk_should_throtl(bio))
a7b36ee6 JA	203	return false;
	204
	205	return __blk_throtl_bio(bio);
	206	}
	207	#endif /* CONFIG_BLK_DEV_THROTTLING */
	208
	209	#endif