xref: /linux-6.15/include/linux/blkdev.h (revision 384740dc)
1 #ifndef _LINUX_BLKDEV_H
2 #define _LINUX_BLKDEV_H
3 
4 #ifdef CONFIG_BLOCK
5 
6 #include <linux/sched.h>
7 #include <linux/major.h>
8 #include <linux/genhd.h>
9 #include <linux/list.h>
10 #include <linux/timer.h>
11 #include <linux/workqueue.h>
12 #include <linux/pagemap.h>
13 #include <linux/backing-dev.h>
14 #include <linux/wait.h>
15 #include <linux/mempool.h>
16 #include <linux/bio.h>
17 #include <linux/module.h>
18 #include <linux/stringify.h>
19 #include <linux/gfp.h>
20 #include <linux/bsg.h>
21 #include <linux/smp.h>
22 
23 #include <asm/scatterlist.h>
24 
25 struct scsi_ioctl_command;
26 
27 struct request_queue;
28 struct elevator_queue;
29 typedef struct elevator_queue elevator_t;
30 struct request_pm_state;
31 struct blk_trace;
32 struct request;
33 struct sg_io_hdr;
34 
35 #define BLKDEV_MIN_RQ	4
36 #define BLKDEV_MAX_RQ	128	/* Default maximum */
37 
38 struct request;
39 typedef void (rq_end_io_fn)(struct request *, int);
40 
41 struct request_list {
42 	int count[2];
43 	int starved[2];
44 	int elvpriv;
45 	mempool_t *rq_pool;
46 	wait_queue_head_t wait[2];
47 };
48 
49 /*
50  * request command types
51  */
52 enum rq_cmd_type_bits {
53 	REQ_TYPE_FS		= 1,	/* fs request */
54 	REQ_TYPE_BLOCK_PC,		/* scsi command */
55 	REQ_TYPE_SENSE,			/* sense request */
56 	REQ_TYPE_PM_SUSPEND,		/* suspend request */
57 	REQ_TYPE_PM_RESUME,		/* resume request */
58 	REQ_TYPE_PM_SHUTDOWN,		/* shutdown request */
59 	REQ_TYPE_SPECIAL,		/* driver defined type */
60 	REQ_TYPE_LINUX_BLOCK,		/* generic block layer message */
61 	/*
62 	 * for ATA/ATAPI devices. this really doesn't belong here, ide should
63 	 * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
64 	 * private REQ_LB opcodes to differentiate what type of request this is
65 	 */
66 	REQ_TYPE_ATA_TASKFILE,
67 	REQ_TYPE_ATA_PC,
68 };
69 
70 /*
71  * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
72  * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a
73  * SCSI cdb.
74  *
75  * 0x00 -> 0x3f are driver private, to be used for whatever purpose they need,
76  * typically to differentiate REQ_TYPE_SPECIAL requests.
77  *
78  */
79 enum {
80 	REQ_LB_OP_EJECT	= 0x40,		/* eject request */
81 	REQ_LB_OP_FLUSH = 0x41,		/* flush request */
82 	REQ_LB_OP_DISCARD = 0x42,	/* discard sectors */
83 };
84 
85 /*
86  * request type modified bits. first two bits match BIO_RW* bits, important
87  */
88 enum rq_flag_bits {
89 	__REQ_RW,		/* not set, read. set, write */
90 	__REQ_FAILFAST,		/* no low level driver retries */
91 	__REQ_DISCARD,		/* request to discard sectors */
92 	__REQ_SORTED,		/* elevator knows about this request */
93 	__REQ_SOFTBARRIER,	/* may not be passed by ioscheduler */
94 	__REQ_HARDBARRIER,	/* may not be passed by drive either */
95 	__REQ_FUA,		/* forced unit access */
96 	__REQ_NOMERGE,		/* don't touch this for merging */
97 	__REQ_STARTED,		/* drive already may have started this one */
98 	__REQ_DONTPREP,		/* don't call prep for this one */
99 	__REQ_QUEUED,		/* uses queueing */
100 	__REQ_ELVPRIV,		/* elevator private data attached */
101 	__REQ_FAILED,		/* set if the request failed */
102 	__REQ_QUIET,		/* don't worry about errors */
103 	__REQ_PREEMPT,		/* set for "ide_preempt" requests */
104 	__REQ_ORDERED_COLOR,	/* is before or after barrier */
105 	__REQ_RW_SYNC,		/* request is sync (O_DIRECT) */
106 	__REQ_ALLOCED,		/* request came from our alloc pool */
107 	__REQ_RW_META,		/* metadata io request */
108 	__REQ_COPY_USER,	/* contains copies of user pages */
109 	__REQ_INTEGRITY,	/* integrity metadata has been remapped */
110 	__REQ_NR_BITS,		/* stops here */
111 };
112 
113 #define REQ_RW		(1 << __REQ_RW)
114 #define REQ_DISCARD	(1 << __REQ_DISCARD)
115 #define REQ_FAILFAST	(1 << __REQ_FAILFAST)
116 #define REQ_SORTED	(1 << __REQ_SORTED)
117 #define REQ_SOFTBARRIER	(1 << __REQ_SOFTBARRIER)
118 #define REQ_HARDBARRIER	(1 << __REQ_HARDBARRIER)
119 #define REQ_FUA		(1 << __REQ_FUA)
120 #define REQ_NOMERGE	(1 << __REQ_NOMERGE)
121 #define REQ_STARTED	(1 << __REQ_STARTED)
122 #define REQ_DONTPREP	(1 << __REQ_DONTPREP)
123 #define REQ_QUEUED	(1 << __REQ_QUEUED)
124 #define REQ_ELVPRIV	(1 << __REQ_ELVPRIV)
125 #define REQ_FAILED	(1 << __REQ_FAILED)
126 #define REQ_QUIET	(1 << __REQ_QUIET)
127 #define REQ_PREEMPT	(1 << __REQ_PREEMPT)
128 #define REQ_ORDERED_COLOR	(1 << __REQ_ORDERED_COLOR)
129 #define REQ_RW_SYNC	(1 << __REQ_RW_SYNC)
130 #define REQ_ALLOCED	(1 << __REQ_ALLOCED)
131 #define REQ_RW_META	(1 << __REQ_RW_META)
132 #define REQ_COPY_USER	(1 << __REQ_COPY_USER)
133 #define REQ_INTEGRITY	(1 << __REQ_INTEGRITY)
134 
135 #define BLK_MAX_CDB	16
136 
137 /*
138  * try to put the fields that are referenced together in the same cacheline.
139  * if you modify this structure, be sure to check block/blk-core.c:rq_init()
140  * as well!
141  */
142 struct request {
143 	struct list_head queuelist;
144 	struct call_single_data csd;
145 	int cpu;
146 
147 	struct request_queue *q;
148 
149 	unsigned int cmd_flags;
150 	enum rq_cmd_type_bits cmd_type;
151 	unsigned long atomic_flags;
152 
153 	/* Maintain bio traversal state for part by part I/O submission.
154 	 * hard_* are block layer internals, no driver should touch them!
155 	 */
156 
157 	sector_t sector;		/* next sector to submit */
158 	sector_t hard_sector;		/* next sector to complete */
159 	unsigned long nr_sectors;	/* no. of sectors left to submit */
160 	unsigned long hard_nr_sectors;	/* no. of sectors left to complete */
161 	/* no. of sectors left to submit in the current segment */
162 	unsigned int current_nr_sectors;
163 
164 	/* no. of sectors left to complete in the current segment */
165 	unsigned int hard_cur_sectors;
166 
167 	struct bio *bio;
168 	struct bio *biotail;
169 
170 	struct hlist_node hash;	/* merge hash */
171 	/*
172 	 * The rb_node is only used inside the io scheduler, requests
173 	 * are pruned when moved to the dispatch queue. So let the
174 	 * completion_data share space with the rb_node.
175 	 */
176 	union {
177 		struct rb_node rb_node;	/* sort/lookup */
178 		void *completion_data;
179 	};
180 
181 	/*
182 	 * two pointers are available for the IO schedulers, if they need
183 	 * more they have to dynamically allocate it.
184 	 */
185 	void *elevator_private;
186 	void *elevator_private2;
187 
188 	struct gendisk *rq_disk;
189 	unsigned long start_time;
190 
191 	/* Number of scatter-gather DMA addr+len pairs after
192 	 * physical address coalescing is performed.
193 	 */
194 	unsigned short nr_phys_segments;
195 
196 	unsigned short ioprio;
197 
198 	void *special;
199 	char *buffer;
200 
201 	int tag;
202 	int errors;
203 
204 	int ref_count;
205 
206 	/*
207 	 * when request is used as a packet command carrier
208 	 */
209 	unsigned short cmd_len;
210 	unsigned char __cmd[BLK_MAX_CDB];
211 	unsigned char *cmd;
212 
213 	unsigned int data_len;
214 	unsigned int extra_len;	/* length of alignment and padding */
215 	unsigned int sense_len;
216 	void *data;
217 	void *sense;
218 
219 	unsigned long deadline;
220 	struct list_head timeout_list;
221 	unsigned int timeout;
222 	int retries;
223 
224 	/*
225 	 * completion callback.
226 	 */
227 	rq_end_io_fn *end_io;
228 	void *end_io_data;
229 
230 	/* for bidi */
231 	struct request *next_rq;
232 };
233 
234 static inline unsigned short req_get_ioprio(struct request *req)
235 {
236 	return req->ioprio;
237 }
238 
239 /*
240  * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
241  * requests. Some step values could eventually be made generic.
242  */
243 struct request_pm_state
244 {
245 	/* PM state machine step value, currently driver specific */
246 	int	pm_step;
247 	/* requested PM state value (S1, S2, S3, S4, ...) */
248 	u32	pm_state;
249 	void*	data;		/* for driver use */
250 };
251 
252 #include <linux/elevator.h>
253 
254 typedef void (request_fn_proc) (struct request_queue *q);
255 typedef int (make_request_fn) (struct request_queue *q, struct bio *bio);
256 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
257 typedef void (unplug_fn) (struct request_queue *);
258 typedef int (prepare_discard_fn) (struct request_queue *, struct request *);
259 
260 struct bio_vec;
261 struct bvec_merge_data {
262 	struct block_device *bi_bdev;
263 	sector_t bi_sector;
264 	unsigned bi_size;
265 	unsigned long bi_rw;
266 };
267 typedef int (merge_bvec_fn) (struct request_queue *, struct bvec_merge_data *,
268 			     struct bio_vec *);
269 typedef void (prepare_flush_fn) (struct request_queue *, struct request *);
270 typedef void (softirq_done_fn)(struct request *);
271 typedef int (dma_drain_needed_fn)(struct request *);
272 typedef int (lld_busy_fn) (struct request_queue *q);
273 
274 enum blk_eh_timer_return {
275 	BLK_EH_NOT_HANDLED,
276 	BLK_EH_HANDLED,
277 	BLK_EH_RESET_TIMER,
278 };
279 
280 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
281 
282 enum blk_queue_state {
283 	Queue_down,
284 	Queue_up,
285 };
286 
287 struct blk_queue_tag {
288 	struct request **tag_index;	/* map of busy tags */
289 	unsigned long *tag_map;		/* bit map of free/busy tags */
290 	int busy;			/* current depth */
291 	int max_depth;			/* what we will send to device */
292 	int real_max_depth;		/* what the array can hold */
293 	atomic_t refcnt;		/* map can be shared */
294 };
295 
296 #define BLK_SCSI_MAX_CMDS	(256)
297 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
298 
299 struct blk_cmd_filter {
300 	unsigned long read_ok[BLK_SCSI_CMD_PER_LONG];
301 	unsigned long write_ok[BLK_SCSI_CMD_PER_LONG];
302 	struct kobject kobj;
303 };
304 
305 struct request_queue
306 {
307 	/*
308 	 * Together with queue_head for cacheline sharing
309 	 */
310 	struct list_head	queue_head;
311 	struct request		*last_merge;
312 	elevator_t		*elevator;
313 
314 	/*
315 	 * the queue request freelist, one for reads and one for writes
316 	 */
317 	struct request_list	rq;
318 
319 	request_fn_proc		*request_fn;
320 	make_request_fn		*make_request_fn;
321 	prep_rq_fn		*prep_rq_fn;
322 	unplug_fn		*unplug_fn;
323 	prepare_discard_fn	*prepare_discard_fn;
324 	merge_bvec_fn		*merge_bvec_fn;
325 	prepare_flush_fn	*prepare_flush_fn;
326 	softirq_done_fn		*softirq_done_fn;
327 	rq_timed_out_fn		*rq_timed_out_fn;
328 	dma_drain_needed_fn	*dma_drain_needed;
329 	lld_busy_fn		*lld_busy_fn;
330 
331 	/*
332 	 * Dispatch queue sorting
333 	 */
334 	sector_t		end_sector;
335 	struct request		*boundary_rq;
336 
337 	/*
338 	 * Auto-unplugging state
339 	 */
340 	struct timer_list	unplug_timer;
341 	int			unplug_thresh;	/* After this many requests */
342 	unsigned long		unplug_delay;	/* After this many jiffies */
343 	struct work_struct	unplug_work;
344 
345 	struct backing_dev_info	backing_dev_info;
346 
347 	/*
348 	 * The queue owner gets to use this for whatever they like.
349 	 * ll_rw_blk doesn't touch it.
350 	 */
351 	void			*queuedata;
352 
353 	/*
354 	 * queue needs bounce pages for pages above this limit
355 	 */
356 	unsigned long		bounce_pfn;
357 	gfp_t			bounce_gfp;
358 
359 	/*
360 	 * various queue flags, see QUEUE_* below
361 	 */
362 	unsigned long		queue_flags;
363 
364 	/*
365 	 * protects queue structures from reentrancy. ->__queue_lock should
366 	 * _never_ be used directly, it is queue private. always use
367 	 * ->queue_lock.
368 	 */
369 	spinlock_t		__queue_lock;
370 	spinlock_t		*queue_lock;
371 
372 	/*
373 	 * queue kobject
374 	 */
375 	struct kobject kobj;
376 
377 	/*
378 	 * queue settings
379 	 */
380 	unsigned long		nr_requests;	/* Max # of requests */
381 	unsigned int		nr_congestion_on;
382 	unsigned int		nr_congestion_off;
383 	unsigned int		nr_batching;
384 
385 	unsigned int		max_sectors;
386 	unsigned int		max_hw_sectors;
387 	unsigned short		max_phys_segments;
388 	unsigned short		max_hw_segments;
389 	unsigned short		hardsect_size;
390 	unsigned int		max_segment_size;
391 
392 	unsigned long		seg_boundary_mask;
393 	void			*dma_drain_buffer;
394 	unsigned int		dma_drain_size;
395 	unsigned int		dma_pad_mask;
396 	unsigned int		dma_alignment;
397 
398 	struct blk_queue_tag	*queue_tags;
399 	struct list_head	tag_busy_list;
400 
401 	unsigned int		nr_sorted;
402 	unsigned int		in_flight;
403 
404 	unsigned int		rq_timeout;
405 	struct timer_list	timeout;
406 	struct list_head	timeout_list;
407 
408 	/*
409 	 * sg stuff
410 	 */
411 	unsigned int		sg_timeout;
412 	unsigned int		sg_reserved_size;
413 	int			node;
414 #ifdef CONFIG_BLK_DEV_IO_TRACE
415 	struct blk_trace	*blk_trace;
416 #endif
417 	/*
418 	 * reserved for flush operations
419 	 */
420 	unsigned int		ordered, next_ordered, ordseq;
421 	int			orderr, ordcolor;
422 	struct request		pre_flush_rq, bar_rq, post_flush_rq;
423 	struct request		*orig_bar_rq;
424 
425 	struct mutex		sysfs_lock;
426 
427 #if defined(CONFIG_BLK_DEV_BSG)
428 	struct bsg_class_device bsg_dev;
429 #endif
430 	struct blk_cmd_filter cmd_filter;
431 };
432 
433 #define QUEUE_FLAG_CLUSTER	0	/* cluster several segments into 1 */
434 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */
435 #define QUEUE_FLAG_STOPPED	2	/* queue is stopped */
436 #define	QUEUE_FLAG_READFULL	3	/* read queue has been filled */
437 #define QUEUE_FLAG_WRITEFULL	4	/* write queue has been filled */
438 #define QUEUE_FLAG_DEAD		5	/* queue being torn down */
439 #define QUEUE_FLAG_REENTER	6	/* Re-entrancy avoidance */
440 #define QUEUE_FLAG_PLUGGED	7	/* queue is plugged */
441 #define QUEUE_FLAG_ELVSWITCH	8	/* don't use elevator, just do FIFO */
442 #define QUEUE_FLAG_BIDI		9	/* queue supports bidi requests */
443 #define QUEUE_FLAG_NOMERGES    10	/* disable merge attempts */
444 #define QUEUE_FLAG_SAME_COMP   11	/* force complete on same CPU */
445 #define QUEUE_FLAG_FAIL_IO     12	/* fake timeout */
446 #define QUEUE_FLAG_STACKABLE   13	/* supports request stacking */
447 #define QUEUE_FLAG_NONROT      14	/* non-rotational device (SSD) */
448 
449 static inline int queue_is_locked(struct request_queue *q)
450 {
451 #ifdef CONFIG_SMP
452 	spinlock_t *lock = q->queue_lock;
453 	return lock && spin_is_locked(lock);
454 #else
455 	return 1;
456 #endif
457 }
458 
459 static inline void queue_flag_set_unlocked(unsigned int flag,
460 					   struct request_queue *q)
461 {
462 	__set_bit(flag, &q->queue_flags);
463 }
464 
465 static inline int queue_flag_test_and_clear(unsigned int flag,
466 					    struct request_queue *q)
467 {
468 	WARN_ON_ONCE(!queue_is_locked(q));
469 
470 	if (test_bit(flag, &q->queue_flags)) {
471 		__clear_bit(flag, &q->queue_flags);
472 		return 1;
473 	}
474 
475 	return 0;
476 }
477 
478 static inline int queue_flag_test_and_set(unsigned int flag,
479 					  struct request_queue *q)
480 {
481 	WARN_ON_ONCE(!queue_is_locked(q));
482 
483 	if (!test_bit(flag, &q->queue_flags)) {
484 		__set_bit(flag, &q->queue_flags);
485 		return 0;
486 	}
487 
488 	return 1;
489 }
490 
491 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
492 {
493 	WARN_ON_ONCE(!queue_is_locked(q));
494 	__set_bit(flag, &q->queue_flags);
495 }
496 
497 static inline void queue_flag_clear_unlocked(unsigned int flag,
498 					     struct request_queue *q)
499 {
500 	__clear_bit(flag, &q->queue_flags);
501 }
502 
503 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
504 {
505 	WARN_ON_ONCE(!queue_is_locked(q));
506 	__clear_bit(flag, &q->queue_flags);
507 }
508 
509 enum {
510 	/*
511 	 * Hardbarrier is supported with one of the following methods.
512 	 *
513 	 * NONE		: hardbarrier unsupported
514 	 * DRAIN	: ordering by draining is enough
515 	 * DRAIN_FLUSH	: ordering by draining w/ pre and post flushes
516 	 * DRAIN_FUA	: ordering by draining w/ pre flush and FUA write
517 	 * TAG		: ordering by tag is enough
518 	 * TAG_FLUSH	: ordering by tag w/ pre and post flushes
519 	 * TAG_FUA	: ordering by tag w/ pre flush and FUA write
520 	 */
521 	QUEUE_ORDERED_NONE	= 0x00,
522 	QUEUE_ORDERED_DRAIN	= 0x01,
523 	QUEUE_ORDERED_TAG	= 0x02,
524 
525 	QUEUE_ORDERED_PREFLUSH	= 0x10,
526 	QUEUE_ORDERED_POSTFLUSH	= 0x20,
527 	QUEUE_ORDERED_FUA	= 0x40,
528 
529 	QUEUE_ORDERED_DRAIN_FLUSH = QUEUE_ORDERED_DRAIN |
530 			QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
531 	QUEUE_ORDERED_DRAIN_FUA	= QUEUE_ORDERED_DRAIN |
532 			QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
533 	QUEUE_ORDERED_TAG_FLUSH	= QUEUE_ORDERED_TAG |
534 			QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
535 	QUEUE_ORDERED_TAG_FUA	= QUEUE_ORDERED_TAG |
536 			QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
537 
538 	/*
539 	 * Ordered operation sequence
540 	 */
541 	QUEUE_ORDSEQ_STARTED	= 0x01,	/* flushing in progress */
542 	QUEUE_ORDSEQ_DRAIN	= 0x02,	/* waiting for the queue to be drained */
543 	QUEUE_ORDSEQ_PREFLUSH	= 0x04,	/* pre-flushing in progress */
544 	QUEUE_ORDSEQ_BAR	= 0x08,	/* original barrier req in progress */
545 	QUEUE_ORDSEQ_POSTFLUSH	= 0x10,	/* post-flushing in progress */
546 	QUEUE_ORDSEQ_DONE	= 0x20,
547 };
548 
549 #define blk_queue_plugged(q)	test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
550 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
551 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
552 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
553 #define blk_queue_nonrot(q)	test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
554 #define blk_queue_flushing(q)	((q)->ordseq)
555 #define blk_queue_stackable(q)	\
556 	test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
557 
558 #define blk_fs_request(rq)	((rq)->cmd_type == REQ_TYPE_FS)
559 #define blk_pc_request(rq)	((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
560 #define blk_special_request(rq)	((rq)->cmd_type == REQ_TYPE_SPECIAL)
561 #define blk_sense_request(rq)	((rq)->cmd_type == REQ_TYPE_SENSE)
562 
563 #define blk_noretry_request(rq)	((rq)->cmd_flags & REQ_FAILFAST)
564 #define blk_rq_started(rq)	((rq)->cmd_flags & REQ_STARTED)
565 
566 #define blk_account_rq(rq)	(blk_rq_started(rq) && (blk_fs_request(rq) || blk_discard_rq(rq)))
567 
568 #define blk_pm_suspend_request(rq)	((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
569 #define blk_pm_resume_request(rq)	((rq)->cmd_type == REQ_TYPE_PM_RESUME)
570 #define blk_pm_request(rq)	\
571 	(blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
572 
573 #define blk_rq_cpu_valid(rq)	((rq)->cpu != -1)
574 #define blk_sorted_rq(rq)	((rq)->cmd_flags & REQ_SORTED)
575 #define blk_barrier_rq(rq)	((rq)->cmd_flags & REQ_HARDBARRIER)
576 #define blk_fua_rq(rq)		((rq)->cmd_flags & REQ_FUA)
577 #define blk_discard_rq(rq)	((rq)->cmd_flags & REQ_DISCARD)
578 #define blk_bidi_rq(rq)		((rq)->next_rq != NULL)
579 #define blk_empty_barrier(rq)	(blk_barrier_rq(rq) && blk_fs_request(rq) && !(rq)->hard_nr_sectors)
580 /* rq->queuelist of dequeued request must be list_empty() */
581 #define blk_queued_rq(rq)	(!list_empty(&(rq)->queuelist))
582 
583 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
584 
585 #define rq_data_dir(rq)		((rq)->cmd_flags & 1)
586 
587 /*
588  * We regard a request as sync, if it's a READ or a SYNC write.
589  */
590 #define rq_is_sync(rq)		(rq_data_dir((rq)) == READ || (rq)->cmd_flags & REQ_RW_SYNC)
591 #define rq_is_meta(rq)		((rq)->cmd_flags & REQ_RW_META)
592 
593 static inline int blk_queue_full(struct request_queue *q, int rw)
594 {
595 	if (rw == READ)
596 		return test_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
597 	return test_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
598 }
599 
600 static inline void blk_set_queue_full(struct request_queue *q, int rw)
601 {
602 	if (rw == READ)
603 		queue_flag_set(QUEUE_FLAG_READFULL, q);
604 	else
605 		queue_flag_set(QUEUE_FLAG_WRITEFULL, q);
606 }
607 
608 static inline void blk_clear_queue_full(struct request_queue *q, int rw)
609 {
610 	if (rw == READ)
611 		queue_flag_clear(QUEUE_FLAG_READFULL, q);
612 	else
613 		queue_flag_clear(QUEUE_FLAG_WRITEFULL, q);
614 }
615 
616 
617 /*
618  * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
619  * it already be started by driver.
620  */
621 #define RQ_NOMERGE_FLAGS	\
622 	(REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
623 #define rq_mergeable(rq)	\
624 	(!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && \
625 	 (blk_discard_rq(rq) || blk_fs_request((rq))))
626 
627 /*
628  * q->prep_rq_fn return values
629  */
630 #define BLKPREP_OK		0	/* serve it */
631 #define BLKPREP_KILL		1	/* fatal error, kill */
632 #define BLKPREP_DEFER		2	/* leave on queue */
633 
634 extern unsigned long blk_max_low_pfn, blk_max_pfn;
635 
636 /*
637  * standard bounce addresses:
638  *
639  * BLK_BOUNCE_HIGH	: bounce all highmem pages
640  * BLK_BOUNCE_ANY	: don't bounce anything
641  * BLK_BOUNCE_ISA	: bounce pages above ISA DMA boundary
642  */
643 
644 #if BITS_PER_LONG == 32
645 #define BLK_BOUNCE_HIGH		((u64)blk_max_low_pfn << PAGE_SHIFT)
646 #else
647 #define BLK_BOUNCE_HIGH		-1ULL
648 #endif
649 #define BLK_BOUNCE_ANY		(-1ULL)
650 #define BLK_BOUNCE_ISA		(ISA_DMA_THRESHOLD)
651 
652 /*
653  * default timeout for SG_IO if none specified
654  */
655 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
656 
657 #ifdef CONFIG_BOUNCE
658 extern int init_emergency_isa_pool(void);
659 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
660 #else
661 static inline int init_emergency_isa_pool(void)
662 {
663 	return 0;
664 }
665 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
666 {
667 }
668 #endif /* CONFIG_MMU */
669 
670 struct rq_map_data {
671 	struct page **pages;
672 	int page_order;
673 	int nr_entries;
674 };
675 
676 struct req_iterator {
677 	int i;
678 	struct bio *bio;
679 };
680 
681 /* This should not be used directly - use rq_for_each_segment */
682 #define __rq_for_each_bio(_bio, rq)	\
683 	if ((rq->bio))			\
684 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
685 
686 #define rq_for_each_segment(bvl, _rq, _iter)			\
687 	__rq_for_each_bio(_iter.bio, _rq)			\
688 		bio_for_each_segment(bvl, _iter.bio, _iter.i)
689 
690 #define rq_iter_last(rq, _iter)					\
691 		(_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
692 
693 extern int blk_register_queue(struct gendisk *disk);
694 extern void blk_unregister_queue(struct gendisk *disk);
695 extern void register_disk(struct gendisk *dev);
696 extern void generic_make_request(struct bio *bio);
697 extern void blk_rq_init(struct request_queue *q, struct request *rq);
698 extern void blk_put_request(struct request *);
699 extern void __blk_put_request(struct request_queue *, struct request *);
700 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
701 extern void blk_insert_request(struct request_queue *, struct request *, int, void *);
702 extern void blk_requeue_request(struct request_queue *, struct request *);
703 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
704 extern int blk_lld_busy(struct request_queue *q);
705 extern int blk_insert_cloned_request(struct request_queue *q,
706 				     struct request *rq);
707 extern void blk_plug_device(struct request_queue *);
708 extern void blk_plug_device_unlocked(struct request_queue *);
709 extern int blk_remove_plug(struct request_queue *);
710 extern void blk_recount_segments(struct request_queue *, struct bio *);
711 extern int scsi_cmd_ioctl(struct file *, struct request_queue *,
712 			  struct gendisk *, unsigned int, void __user *);
713 extern int sg_scsi_ioctl(struct file *, struct request_queue *,
714 		struct gendisk *, struct scsi_ioctl_command __user *);
715 
716 /*
717  * Temporary export, until SCSI gets fixed up.
718  */
719 extern int blk_rq_append_bio(struct request_queue *q, struct request *rq,
720 			     struct bio *bio);
721 
722 /*
723  * A queue has just exitted congestion.  Note this in the global counter of
724  * congested queues, and wake up anyone who was waiting for requests to be
725  * put back.
726  */
727 static inline void blk_clear_queue_congested(struct request_queue *q, int rw)
728 {
729 	clear_bdi_congested(&q->backing_dev_info, rw);
730 }
731 
732 /*
733  * A queue has just entered congestion.  Flag that in the queue's VM-visible
734  * state flags and increment the global gounter of congested queues.
735  */
736 static inline void blk_set_queue_congested(struct request_queue *q, int rw)
737 {
738 	set_bdi_congested(&q->backing_dev_info, rw);
739 }
740 
741 extern void blk_start_queue(struct request_queue *q);
742 extern void blk_stop_queue(struct request_queue *q);
743 extern void blk_sync_queue(struct request_queue *q);
744 extern void __blk_stop_queue(struct request_queue *q);
745 extern void __blk_run_queue(struct request_queue *);
746 extern void blk_run_queue(struct request_queue *);
747 extern void blk_start_queueing(struct request_queue *);
748 extern int blk_rq_map_user(struct request_queue *, struct request *,
749 			   struct rq_map_data *, void __user *, unsigned long,
750 			   gfp_t);
751 extern int blk_rq_unmap_user(struct bio *);
752 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
753 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
754 			       struct rq_map_data *, struct sg_iovec *, int,
755 			       unsigned int, gfp_t);
756 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
757 			  struct request *, int);
758 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
759 				  struct request *, int, rq_end_io_fn *);
760 extern void blk_unplug(struct request_queue *q);
761 
762 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
763 {
764 	return bdev->bd_disk->queue;
765 }
766 
767 static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
768 				       struct page *page)
769 {
770 	if (bdi && bdi->unplug_io_fn)
771 		bdi->unplug_io_fn(bdi, page);
772 }
773 
774 static inline void blk_run_address_space(struct address_space *mapping)
775 {
776 	if (mapping)
777 		blk_run_backing_dev(mapping->backing_dev_info, NULL);
778 }
779 
780 /*
781  * blk_end_request() and friends.
782  * __blk_end_request() and end_request() must be called with
783  * the request queue spinlock acquired.
784  *
785  * Several drivers define their own end_request and call
786  * blk_end_request() for parts of the original function.
787  * This prevents code duplication in drivers.
788  */
789 extern int blk_end_request(struct request *rq, int error,
790 				unsigned int nr_bytes);
791 extern int __blk_end_request(struct request *rq, int error,
792 				unsigned int nr_bytes);
793 extern int blk_end_bidi_request(struct request *rq, int error,
794 				unsigned int nr_bytes, unsigned int bidi_bytes);
795 extern void end_request(struct request *, int);
796 extern int blk_end_request_callback(struct request *rq, int error,
797 				unsigned int nr_bytes,
798 				int (drv_callback)(struct request *));
799 extern void blk_complete_request(struct request *);
800 extern void __blk_complete_request(struct request *);
801 extern void blk_abort_request(struct request *);
802 extern void blk_abort_queue(struct request_queue *);
803 extern void blk_update_request(struct request *rq, int error,
804 			       unsigned int nr_bytes);
805 
806 /*
807  * blk_end_request() takes bytes instead of sectors as a complete size.
808  * blk_rq_bytes() returns bytes left to complete in the entire request.
809  * blk_rq_cur_bytes() returns bytes left to complete in the current segment.
810  */
811 extern unsigned int blk_rq_bytes(struct request *rq);
812 extern unsigned int blk_rq_cur_bytes(struct request *rq);
813 
814 static inline void blkdev_dequeue_request(struct request *req)
815 {
816 	elv_dequeue_request(req->q, req);
817 }
818 
819 /*
820  * Access functions for manipulating queue properties
821  */
822 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
823 					spinlock_t *lock, int node_id);
824 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
825 extern void blk_cleanup_queue(struct request_queue *);
826 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
827 extern void blk_queue_bounce_limit(struct request_queue *, u64);
828 extern void blk_queue_max_sectors(struct request_queue *, unsigned int);
829 extern void blk_queue_max_phys_segments(struct request_queue *, unsigned short);
830 extern void blk_queue_max_hw_segments(struct request_queue *, unsigned short);
831 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
832 extern void blk_queue_hardsect_size(struct request_queue *, unsigned short);
833 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
834 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
835 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
836 extern int blk_queue_dma_drain(struct request_queue *q,
837 			       dma_drain_needed_fn *dma_drain_needed,
838 			       void *buf, unsigned int size);
839 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
840 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
841 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
842 extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *);
843 extern void blk_queue_dma_alignment(struct request_queue *, int);
844 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
845 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
846 extern void blk_queue_set_discard(struct request_queue *, prepare_discard_fn *);
847 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
848 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
849 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
850 extern int blk_queue_ordered(struct request_queue *, unsigned, prepare_flush_fn *);
851 extern int blk_do_ordered(struct request_queue *, struct request **);
852 extern unsigned blk_ordered_cur_seq(struct request_queue *);
853 extern unsigned blk_ordered_req_seq(struct request *);
854 extern void blk_ordered_complete_seq(struct request_queue *, unsigned, int);
855 
856 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
857 extern void blk_dump_rq_flags(struct request *, char *);
858 extern void generic_unplug_device(struct request_queue *);
859 extern void __generic_unplug_device(struct request_queue *);
860 extern long nr_blockdev_pages(void);
861 
862 int blk_get_queue(struct request_queue *);
863 struct request_queue *blk_alloc_queue(gfp_t);
864 struct request_queue *blk_alloc_queue_node(gfp_t, int);
865 extern void blk_put_queue(struct request_queue *);
866 
867 /*
868  * tag stuff
869  */
870 #define blk_rq_tagged(rq)		((rq)->cmd_flags & REQ_QUEUED)
871 extern int blk_queue_start_tag(struct request_queue *, struct request *);
872 extern struct request *blk_queue_find_tag(struct request_queue *, int);
873 extern void blk_queue_end_tag(struct request_queue *, struct request *);
874 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *);
875 extern void blk_queue_free_tags(struct request_queue *);
876 extern int blk_queue_resize_tags(struct request_queue *, int);
877 extern void blk_queue_invalidate_tags(struct request_queue *);
878 extern struct blk_queue_tag *blk_init_tags(int);
879 extern void blk_free_tags(struct blk_queue_tag *);
880 
881 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
882 						int tag)
883 {
884 	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
885 		return NULL;
886 	return bqt->tag_index[tag];
887 }
888 
889 extern int blkdev_issue_flush(struct block_device *, sector_t *);
890 extern int blkdev_issue_discard(struct block_device *,
891 				sector_t sector, sector_t nr_sects, gfp_t);
892 
893 static inline int sb_issue_discard(struct super_block *sb,
894 				   sector_t block, sector_t nr_blocks)
895 {
896 	block <<= (sb->s_blocksize_bits - 9);
897 	nr_blocks <<= (sb->s_blocksize_bits - 9);
898 	return blkdev_issue_discard(sb->s_bdev, block, nr_blocks, GFP_KERNEL);
899 }
900 
901 /*
902 * command filter functions
903 */
904 extern int blk_verify_command(struct blk_cmd_filter *filter,
905 			      unsigned char *cmd, int has_write_perm);
906 extern void blk_set_cmd_filter_defaults(struct blk_cmd_filter *filter);
907 
908 #define MAX_PHYS_SEGMENTS 128
909 #define MAX_HW_SEGMENTS 128
910 #define SAFE_MAX_SECTORS 255
911 #define BLK_DEF_MAX_SECTORS 1024
912 
913 #define MAX_SEGMENT_SIZE	65536
914 
915 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
916 
917 static inline int queue_hardsect_size(struct request_queue *q)
918 {
919 	int retval = 512;
920 
921 	if (q && q->hardsect_size)
922 		retval = q->hardsect_size;
923 
924 	return retval;
925 }
926 
927 static inline int bdev_hardsect_size(struct block_device *bdev)
928 {
929 	return queue_hardsect_size(bdev_get_queue(bdev));
930 }
931 
932 static inline int queue_dma_alignment(struct request_queue *q)
933 {
934 	return q ? q->dma_alignment : 511;
935 }
936 
937 static inline int blk_rq_aligned(struct request_queue *q, void *addr,
938 				 unsigned int len)
939 {
940 	unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
941 	return !((unsigned long)addr & alignment) && !(len & alignment);
942 }
943 
944 /* assumes size > 256 */
945 static inline unsigned int blksize_bits(unsigned int size)
946 {
947 	unsigned int bits = 8;
948 	do {
949 		bits++;
950 		size >>= 1;
951 	} while (size > 256);
952 	return bits;
953 }
954 
955 static inline unsigned int block_size(struct block_device *bdev)
956 {
957 	return bdev->bd_block_size;
958 }
959 
960 typedef struct {struct page *v;} Sector;
961 
962 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
963 
964 static inline void put_dev_sector(Sector p)
965 {
966 	page_cache_release(p.v);
967 }
968 
969 struct work_struct;
970 int kblockd_schedule_work(struct request_queue *q, struct work_struct *work);
971 void kblockd_flush_work(struct work_struct *work);
972 
973 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
974 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
975 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
976 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
977 
978 #if defined(CONFIG_BLK_DEV_INTEGRITY)
979 
980 #define INTEGRITY_FLAG_READ	2	/* verify data integrity on read */
981 #define INTEGRITY_FLAG_WRITE	4	/* generate data integrity on write */
982 
983 struct blk_integrity_exchg {
984 	void			*prot_buf;
985 	void			*data_buf;
986 	sector_t		sector;
987 	unsigned int		data_size;
988 	unsigned short		sector_size;
989 	const char		*disk_name;
990 };
991 
992 typedef void (integrity_gen_fn) (struct blk_integrity_exchg *);
993 typedef int (integrity_vrfy_fn) (struct blk_integrity_exchg *);
994 typedef void (integrity_set_tag_fn) (void *, void *, unsigned int);
995 typedef void (integrity_get_tag_fn) (void *, void *, unsigned int);
996 
997 struct blk_integrity {
998 	integrity_gen_fn	*generate_fn;
999 	integrity_vrfy_fn	*verify_fn;
1000 	integrity_set_tag_fn	*set_tag_fn;
1001 	integrity_get_tag_fn	*get_tag_fn;
1002 
1003 	unsigned short		flags;
1004 	unsigned short		tuple_size;
1005 	unsigned short		sector_size;
1006 	unsigned short		tag_size;
1007 
1008 	const char		*name;
1009 
1010 	struct kobject		kobj;
1011 };
1012 
1013 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
1014 extern void blk_integrity_unregister(struct gendisk *);
1015 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1016 extern int blk_rq_map_integrity_sg(struct request *, struct scatterlist *);
1017 extern int blk_rq_count_integrity_sg(struct request *);
1018 
1019 static inline
1020 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1021 {
1022 	return bdev->bd_disk->integrity;
1023 }
1024 
1025 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1026 {
1027 	return disk->integrity;
1028 }
1029 
1030 static inline int blk_integrity_rq(struct request *rq)
1031 {
1032 	if (rq->bio == NULL)
1033 		return 0;
1034 
1035 	return bio_integrity(rq->bio);
1036 }
1037 
1038 #else /* CONFIG_BLK_DEV_INTEGRITY */
1039 
1040 #define blk_integrity_rq(rq)			(0)
1041 #define blk_rq_count_integrity_sg(a)		(0)
1042 #define blk_rq_map_integrity_sg(a, b)		(0)
1043 #define bdev_get_integrity(a)			(0)
1044 #define blk_get_integrity(a)			(0)
1045 #define blk_integrity_compare(a, b)		(0)
1046 #define blk_integrity_register(a, b)		(0)
1047 #define blk_integrity_unregister(a)		do { } while (0);
1048 
1049 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1050 
1051 #else /* CONFIG_BLOCK */
1052 /*
1053  * stubs for when the block layer is configured out
1054  */
1055 #define buffer_heads_over_limit 0
1056 
1057 static inline long nr_blockdev_pages(void)
1058 {
1059 	return 0;
1060 }
1061 
1062 #endif /* CONFIG_BLOCK */
1063 
1064 #endif
1065