blk-mq.h 7.53 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
struct blk_mq_ctxmap {
16
	unsigned int size;
17 18 19 20
	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
	struct blk_mq_cpu_notifier	cpu_notifier;
	struct kobject		kobj;
62 63 64

	unsigned long		poll_invoked;
	unsigned long		poll_success;
65 66
};

67
struct blk_mq_tag_set {
68 69
	struct blk_mq_ops	*ops;
	unsigned int		nr_hw_queues;
70
	unsigned int		queue_depth;	/* max hw supported */
71 72 73 74 75
	unsigned int		reserved_tags;
	unsigned int		cmd_size;	/* per-request extra data */
	int			numa_node;
	unsigned int		timeout;
	unsigned int		flags;		/* BLK_MQ_F_* */
76 77 78
	void			*driver_data;

	struct blk_mq_tags	**tags;
79 80 81

	struct mutex		tag_list_lock;
	struct list_head	tag_list;
82 83
};

84 85 86 87 88 89 90
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 *);
91
typedef struct blk_mq_hw_ctx *(map_queue_fn)(struct request_queue *, const int);
92
typedef enum blk_eh_timer_return (timeout_fn)(struct request *, bool);
93 94
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);
95 96 97 98
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);
99

100 101
typedef void (busy_iter_fn)(struct blk_mq_hw_ctx *, struct request *, void *,
		bool);
102
typedef void (busy_tag_iter_fn)(struct request *, void *, bool);
103 104
typedef int (poll_fn)(struct blk_mq_hw_ctx *, unsigned int);

105

106 107 108 109 110 111 112 113 114 115 116 117 118 119
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
	 */
120
	timeout_fn		*timeout;
121

122 123 124 125 126
	/*
	 * Called to poll for completion of a specific tag.
	 */
	poll_fn			*poll;

127 128
	softirq_done_fn		*complete;

129 130 131 132 133 134 135
	/*
	 * 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;
136 137 138 139

	/*
	 * Called for every command allocated by the block layer to allow
	 * the driver to set up driver specific data.
140 141 142 143
	 *
	 * Tag greater than or equal to queue_depth is for setting up
	 * flush request.
	 *
144 145 146 147
	 * Ditto for exit/teardown.
	 */
	init_request_fn		*init_request;
	exit_request_fn		*exit_request;
148 149 150 151 152 153 154 155
};

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,
156 157
	BLK_MQ_F_TAG_SHARED	= 1 << 1,
	BLK_MQ_F_SG_MERGE	= 1 << 2,
158
	BLK_MQ_F_DEFER_ISSUE	= 1 << 4,
159 160
	BLK_MQ_F_ALLOC_POLICY_START_BIT = 8,
	BLK_MQ_F_ALLOC_POLICY_BITS = 1,
161

162
	BLK_MQ_S_STOPPED	= 0,
163
	BLK_MQ_S_TAG_ACTIVE	= 1,
164

165
	BLK_MQ_MAX_DEPTH	= 10240,
166 167

	BLK_MQ_CPU_WORK_BATCH	= 8,
168
};
169 170 171 172 173 174
#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)
175

176
struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *);
177 178
struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
						  struct request_queue *q);
179 180 181
int blk_mq_register_disk(struct gendisk *);
void blk_mq_unregister_disk(struct gendisk *);

182 183 184
int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set);
void blk_mq_free_tag_set(struct blk_mq_tag_set *set);

185 186
void blk_mq_flush_plug_list(struct blk_plug *plug, bool from_schedule);

187
void blk_mq_insert_request(struct request *, bool, bool, bool);
188
void blk_mq_free_request(struct request *rq);
189
void blk_mq_free_hctx_request(struct blk_mq_hw_ctx *, struct request *rq);
190
bool blk_mq_can_queue(struct blk_mq_hw_ctx *);
191 192 193 194 195 196

enum {
	BLK_MQ_REQ_NOWAIT	= (1 << 0), /* return when out of requests */
	BLK_MQ_REQ_RESERVED	= (1 << 1), /* allocate from reserved pool */
};

197
struct request *blk_mq_alloc_request(struct request_queue *q, int rw,
198
		unsigned int flags);
199
struct request *blk_mq_tag_to_rq(struct blk_mq_tags *tags, unsigned int tag);
200
struct cpumask *blk_mq_tags_cpumask(struct blk_mq_tags *tags);
201

202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218
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;
}

219
struct blk_mq_hw_ctx *blk_mq_map_queue(struct request_queue *, const int ctx_index);
220
struct blk_mq_hw_ctx *blk_mq_alloc_single_hw_queue(struct blk_mq_tag_set *, unsigned int, int);
221

222
int blk_mq_request_started(struct request *rq);
223
void blk_mq_start_request(struct request *rq);
224 225
void blk_mq_end_request(struct request *rq, int error);
void __blk_mq_end_request(struct request *rq, int error);
226

227
void blk_mq_requeue_request(struct request *rq);
228
void blk_mq_add_to_requeue_list(struct request *rq, bool at_head);
229
void blk_mq_cancel_requeue_work(struct request_queue *q);
230
void blk_mq_kick_requeue_list(struct request_queue *q);
231
void blk_mq_abort_requeue_list(struct request_queue *q);
232
void blk_mq_complete_request(struct request *rq, int error);
233

234 235
void blk_mq_stop_hw_queue(struct blk_mq_hw_ctx *hctx);
void blk_mq_start_hw_queue(struct blk_mq_hw_ctx *hctx);
236
void blk_mq_stop_hw_queues(struct request_queue *q);
237
void blk_mq_start_hw_queues(struct request_queue *q);
238
void blk_mq_start_stopped_hw_queues(struct request_queue *q, bool async);
239
void blk_mq_run_hw_queues(struct request_queue *q, bool async);
240
void blk_mq_delay_queue(struct blk_mq_hw_ctx *hctx, unsigned long msecs);
241 242
void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset,
		busy_tag_iter_fn *fn, void *priv);
243
void blk_mq_freeze_queue(struct request_queue *q);
244 245
void blk_mq_unfreeze_queue(struct request_queue *q);
void blk_mq_freeze_queue_start(struct request_queue *q);
246

247 248
void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues);

249 250
/*
 * Driver command data is immediately after the request. So subtract request
251
 * size to get back to the original request, add request size to get the PDU.
252 253 254 255 256 257 258
 */
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)
{
259
	return rq + 1;
260 261 262
}

#define queue_for_each_hw_ctx(q, hctx, i)				\
263 264
	for ((i) = 0; (i) < (q)->nr_hw_queues &&			\
	     ({ hctx = (q)->queue_hw_ctx[i]; 1; }); (i)++)
265 266

#define hctx_for_each_ctx(hctx, ctx, i)					\
267 268
	for ((i) = 0; (i) < (hctx)->nr_ctx &&				\
	     ({ ctx = (hctx)->ctxs[(i)]; 1; }); (i)++)
269 270

#endif