blk-mq.h 7.4 KB
Newer Older
1 2 3 4 5 6
#ifndef BLK_MQ_H
#define BLK_MQ_H

#include <linux/blkdev.h>

struct blk_mq_tags;
7
struct blk_flush_queue;
8 9 10 11

struct blk_mq_cpu_notifier {
	struct list_head list;
	void *data;
12
	int (*notify)(void *data, unsigned long action, unsigned int cpu);
13 14
};

15 16 17 18 19 20
struct blk_mq_ctxmap {
	unsigned int map_size;
	unsigned int bits_per_word;
	struct blk_align_bitmap *map;
};

21 22 23 24 25 26 27
struct blk_mq_hw_ctx {
	struct {
		spinlock_t		lock;
		struct list_head	dispatch;
	} ____cacheline_aligned_in_smp;

	unsigned long		state;		/* BLK_MQ_S_* flags */
28 29
	struct delayed_work	run_work;
	struct delayed_work	delay_work;
30
	cpumask_var_t		cpumask;
31 32
	int			next_cpu;
	int			next_cpu_batch;
33 34 35 36

	unsigned long		flags;		/* BLK_MQ_F_* flags */

	struct request_queue	*queue;
37
	struct blk_flush_queue	*fq;
38 39 40

	void			*driver_data;

41 42
	struct blk_mq_ctxmap	ctx_map;

43 44 45
	unsigned int		nr_ctx;
	struct blk_mq_ctx	**ctxs;

46
	atomic_t		wait_index;
47 48 49 50 51 52 53 54 55

	struct blk_mq_tags	*tags;

	unsigned long		queued;
	unsigned long		run;
#define BLK_MQ_MAX_DISPATCH_ORDER	10
	unsigned long		dispatched[BLK_MQ_MAX_DISPATCH_ORDER];

	unsigned int		numa_node;
56
	unsigned int		queue_num;
57

58 59
	atomic_t		nr_active;

60 61 62 63
	struct blk_mq_cpu_notifier	cpu_notifier;
	struct kobject		kobj;
};

64
struct blk_mq_tag_set {
65 66
	struct blk_mq_ops	*ops;
	unsigned int		nr_hw_queues;
67
	unsigned int		queue_depth;	/* max hw supported */
68 69 70 71 72
	unsigned int		reserved_tags;
	unsigned int		cmd_size;	/* per-request extra data */
	int			numa_node;
	unsigned int		timeout;
	unsigned int		flags;		/* BLK_MQ_F_* */
73 74 75
	void			*driver_data;

	struct blk_mq_tags	**tags;
76 77 78

	struct mutex		tag_list_lock;
	struct list_head	tag_list;
79 80
};

81 82 83 84 85 86 87
struct blk_mq_queue_data {
	struct request *rq;
	struct list_head *list;
	bool last;
};

typedef int (queue_rq_fn)(struct blk_mq_hw_ctx *, const struct blk_mq_queue_data *);
88
typedef struct blk_mq_hw_ctx *(map_queue_fn)(struct request_queue *, const int);
89
typedef enum blk_eh_timer_return (timeout_fn)(struct request *, bool);
90 91
typedef int (init_hctx_fn)(struct blk_mq_hw_ctx *, void *, unsigned int);
typedef void (exit_hctx_fn)(struct blk_mq_hw_ctx *, unsigned int);
92 93 94 95
typedef int (init_request_fn)(void *, struct request *, unsigned int,
		unsigned int, unsigned int);
typedef void (exit_request_fn)(void *, struct request *, unsigned int,
		unsigned int);
96

97 98 99
typedef void (busy_iter_fn)(struct blk_mq_hw_ctx *, struct request *, void *,
		bool);

100 101 102 103 104 105 106 107 108 109 110 111 112 113
struct blk_mq_ops {
	/*
	 * Queue request
	 */
	queue_rq_fn		*queue_rq;

	/*
	 * Map to specific hardware queue
	 */
	map_queue_fn		*map_queue;

	/*
	 * Called on request timeout
	 */
114
	timeout_fn		*timeout;
115

116 117
	softirq_done_fn		*complete;

118 119 120 121 122 123 124
	/*
	 * Called when the block layer side of a hardware queue has been
	 * set up, allowing the driver to allocate/init matching structures.
	 * Ditto for exit/teardown.
	 */
	init_hctx_fn		*init_hctx;
	exit_hctx_fn		*exit_hctx;
125 126 127 128

	/*
	 * Called for every command allocated by the block layer to allow
	 * the driver to set up driver specific data.
129 130 131 132
	 *
	 * Tag greater than or equal to queue_depth is for setting up
	 * flush request.
	 *
133 134 135 136
	 * Ditto for exit/teardown.
	 */
	init_request_fn		*init_request;
	exit_request_fn		*exit_request;
137 138 139 140 141 142 143 144
};

enum {
	BLK_MQ_RQ_QUEUE_OK	= 0,	/* queued fine */
	BLK_MQ_RQ_QUEUE_BUSY	= 1,	/* requeue IO for later */
	BLK_MQ_RQ_QUEUE_ERROR	= 2,	/* end IO with error */

	BLK_MQ_F_SHOULD_MERGE	= 1 << 0,
145 146 147
	BLK_MQ_F_TAG_SHARED	= 1 << 1,
	BLK_MQ_F_SG_MERGE	= 1 << 2,
	BLK_MQ_F_SYSFS_UP	= 1 << 3,
148
	BLK_MQ_F_DEFER_ISSUE	= 1 << 4,
S
Shaohua Li 已提交
149 150
	BLK_MQ_F_ALLOC_POLICY_START_BIT = 8,
	BLK_MQ_F_ALLOC_POLICY_BITS = 1,
151

152
	BLK_MQ_S_STOPPED	= 0,
153
	BLK_MQ_S_TAG_ACTIVE	= 1,
154

155
	BLK_MQ_MAX_DEPTH	= 10240,
156 157

	BLK_MQ_CPU_WORK_BATCH	= 8,
158
};
S
Shaohua Li 已提交
159 160 161 162 163 164
#define BLK_MQ_FLAG_TO_ALLOC_POLICY(flags) \
	((flags >> BLK_MQ_F_ALLOC_POLICY_START_BIT) & \
		((1 << BLK_MQ_F_ALLOC_POLICY_BITS) - 1))
#define BLK_ALLOC_POLICY_TO_MQ_FLAG(policy) \
	((policy & ((1 << BLK_MQ_F_ALLOC_POLICY_BITS) - 1)) \
		<< BLK_MQ_F_ALLOC_POLICY_START_BIT)
165

166
struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *);
167 168
struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
						  struct request_queue *q);
169
void blk_mq_finish_init(struct request_queue *q);
170 171 172
int blk_mq_register_disk(struct gendisk *);
void blk_mq_unregister_disk(struct gendisk *);

173 174 175
int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set);
void blk_mq_free_tag_set(struct blk_mq_tag_set *set);

176 177
void blk_mq_flush_plug_list(struct blk_plug *plug, bool from_schedule);

178
void blk_mq_insert_request(struct request *, bool, bool, bool);
179
void blk_mq_free_request(struct request *rq);
180
void blk_mq_free_hctx_request(struct blk_mq_hw_ctx *, struct request *rq);
181
bool blk_mq_can_queue(struct blk_mq_hw_ctx *);
182 183
struct request *blk_mq_alloc_request(struct request_queue *q, int rw,
		gfp_t gfp, bool reserved);
184
struct request *blk_mq_tag_to_rq(struct blk_mq_tags *tags, unsigned int tag);
185

B
Bart Van Assche 已提交
186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202
enum {
	BLK_MQ_UNIQUE_TAG_BITS = 16,
	BLK_MQ_UNIQUE_TAG_MASK = (1 << BLK_MQ_UNIQUE_TAG_BITS) - 1,
};

u32 blk_mq_unique_tag(struct request *rq);

static inline u16 blk_mq_unique_tag_to_hwq(u32 unique_tag)
{
	return unique_tag >> BLK_MQ_UNIQUE_TAG_BITS;
}

static inline u16 blk_mq_unique_tag_to_tag(u32 unique_tag)
{
	return unique_tag & BLK_MQ_UNIQUE_TAG_MASK;
}

203
struct blk_mq_hw_ctx *blk_mq_map_queue(struct request_queue *, const int ctx_index);
204
struct blk_mq_hw_ctx *blk_mq_alloc_single_hw_queue(struct blk_mq_tag_set *, unsigned int, int);
205

206
int blk_mq_request_started(struct request *rq);
207
void blk_mq_start_request(struct request *rq);
208 209
void blk_mq_end_request(struct request *rq, int error);
void __blk_mq_end_request(struct request *rq, int error);
210

211
void blk_mq_requeue_request(struct request *rq);
212
void blk_mq_add_to_requeue_list(struct request *rq, bool at_head);
213
void blk_mq_cancel_requeue_work(struct request_queue *q);
214
void blk_mq_kick_requeue_list(struct request_queue *q);
215
void blk_mq_abort_requeue_list(struct request_queue *q);
216 217
void blk_mq_complete_request(struct request *rq);

218 219
void blk_mq_stop_hw_queue(struct blk_mq_hw_ctx *hctx);
void blk_mq_start_hw_queue(struct blk_mq_hw_ctx *hctx);
220
void blk_mq_stop_hw_queues(struct request_queue *q);
221
void blk_mq_start_hw_queues(struct request_queue *q);
222
void blk_mq_start_stopped_hw_queues(struct request_queue *q, bool async);
223
void blk_mq_run_hw_queues(struct request_queue *q, bool async);
224
void blk_mq_delay_queue(struct blk_mq_hw_ctx *hctx, unsigned long msecs);
225 226
void blk_mq_tag_busy_iter(struct blk_mq_hw_ctx *hctx, busy_iter_fn *fn,
		void *priv);
227
void blk_mq_freeze_queue(struct request_queue *q);
228 229
void blk_mq_unfreeze_queue(struct request_queue *q);
void blk_mq_freeze_queue_start(struct request_queue *q);
230 231 232

/*
 * Driver command data is immediately after the request. So subtract request
J
Jens Axboe 已提交
233
 * size to get back to the original request, add request size to get the PDU.
234 235 236 237 238 239 240
 */
static inline struct request *blk_mq_rq_from_pdu(void *pdu)
{
	return pdu - sizeof(struct request);
}
static inline void *blk_mq_rq_to_pdu(struct request *rq)
{
J
Jens Axboe 已提交
241
	return rq + 1;
242 243 244
}

#define queue_for_each_hw_ctx(q, hctx, i)				\
245 246
	for ((i) = 0; (i) < (q)->nr_hw_queues &&			\
	     ({ hctx = (q)->queue_hw_ctx[i]; 1; }); (i)++)
247 248

#define queue_for_each_ctx(q, ctx, i)					\
249 250
	for ((i) = 0; (i) < (q)->nr_queues &&				\
	     ({ ctx = per_cpu_ptr((q)->queue_ctx, (i)); 1; }); (i)++)
251 252

#define hctx_for_each_ctx(hctx, ctx, i)					\
253 254
	for ((i) = 0; (i) < (hctx)->nr_ctx &&				\
	     ({ ctx = (hctx)->ctxs[(i)]; 1; }); (i)++)
255 256 257 258 259 260 261 262 263 264 265 266

#define blk_ctx_sum(q, sum)						\
({									\
	struct blk_mq_ctx *__x;						\
	unsigned int __ret = 0, __i;					\
									\
	queue_for_each_ctx((q), __x, __i)				\
		__ret += sum;						\
	__ret;								\
})

#endif