xref: /linux-6.15/include/linux/blkdev.h (revision 40d3057a)
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/bsg.h>
20 
21 #include <asm/scatterlist.h>
22 
23 struct scsi_ioctl_command;
24 
25 struct request_queue;
26 struct elevator_queue;
27 typedef struct elevator_queue elevator_t;
28 struct request_pm_state;
29 struct blk_trace;
30 struct request;
31 struct sg_io_hdr;
32 
33 #define BLKDEV_MIN_RQ	4
34 #define BLKDEV_MAX_RQ	128	/* Default maximum */
35 
36 struct request;
37 typedef void (rq_end_io_fn)(struct request *, int);
38 
39 struct request_list {
40 	int count[2];
41 	int starved[2];
42 	int elvpriv;
43 	mempool_t *rq_pool;
44 	wait_queue_head_t wait[2];
45 };
46 
47 /*
48  * request command types
49  */
50 enum rq_cmd_type_bits {
51 	REQ_TYPE_FS		= 1,	/* fs request */
52 	REQ_TYPE_BLOCK_PC,		/* scsi command */
53 	REQ_TYPE_SENSE,			/* sense request */
54 	REQ_TYPE_PM_SUSPEND,		/* suspend request */
55 	REQ_TYPE_PM_RESUME,		/* resume request */
56 	REQ_TYPE_PM_SHUTDOWN,		/* shutdown request */
57 	REQ_TYPE_FLUSH,			/* flush request */
58 	REQ_TYPE_SPECIAL,		/* driver defined type */
59 	REQ_TYPE_LINUX_BLOCK,		/* generic block layer message */
60 	/*
61 	 * for ATA/ATAPI devices. this really doesn't belong here, ide should
62 	 * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
63 	 * private REQ_LB opcodes to differentiate what type of request this is
64 	 */
65 	REQ_TYPE_ATA_TASKFILE,
66 	REQ_TYPE_ATA_PC,
67 };
68 
69 /*
70  * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
71  * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a
72  * SCSI cdb.
73  *
74  * 0x00 -> 0x3f are driver private, to be used for whatever purpose they need,
75  * typically to differentiate REQ_TYPE_SPECIAL requests.
76  *
77  */
78 enum {
79 	/*
80 	 * just examples for now
81 	 */
82 	REQ_LB_OP_EJECT	= 0x40,		/* eject request */
83 	REQ_LB_OP_FLUSH = 0x41,		/* flush device */
84 };
85 
86 /*
87  * request type modified bits. first three bits match BIO_RW* bits, important
88  */
89 enum rq_flag_bits {
90 	__REQ_RW,		/* not set, read. set, write */
91 	__REQ_FAILFAST,		/* no low level driver retries */
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_FAILFAST	(1 << __REQ_FAILFAST)
115 #define REQ_SORTED	(1 << __REQ_SORTED)
116 #define REQ_SOFTBARRIER	(1 << __REQ_SOFTBARRIER)
117 #define REQ_HARDBARRIER	(1 << __REQ_HARDBARRIER)
118 #define REQ_FUA		(1 << __REQ_FUA)
119 #define REQ_NOMERGE	(1 << __REQ_NOMERGE)
120 #define REQ_STARTED	(1 << __REQ_STARTED)
121 #define REQ_DONTPREP	(1 << __REQ_DONTPREP)
122 #define REQ_QUEUED	(1 << __REQ_QUEUED)
123 #define REQ_ELVPRIV	(1 << __REQ_ELVPRIV)
124 #define REQ_FAILED	(1 << __REQ_FAILED)
125 #define REQ_QUIET	(1 << __REQ_QUIET)
126 #define REQ_PREEMPT	(1 << __REQ_PREEMPT)
127 #define REQ_ORDERED_COLOR	(1 << __REQ_ORDERED_COLOR)
128 #define REQ_RW_SYNC	(1 << __REQ_RW_SYNC)
129 #define REQ_ALLOCED	(1 << __REQ_ALLOCED)
130 #define REQ_RW_META	(1 << __REQ_RW_META)
131 #define REQ_COPY_USER	(1 << __REQ_COPY_USER)
132 #define REQ_INTEGRITY	(1 << __REQ_INTEGRITY)
133 
134 #define BLK_MAX_CDB	16
135 
136 /*
137  * try to put the fields that are referenced together in the same cacheline.
138  * if you modify this structure, be sure to check block/blk-core.c:rq_init()
139  * as well!
140  */
141 struct request {
142 	struct list_head queuelist;
143 	struct list_head donelist;
144 
145 	struct request_queue *q;
146 
147 	unsigned int cmd_flags;
148 	enum rq_cmd_type_bits cmd_type;
149 
150 	/* Maintain bio traversal state for part by part I/O submission.
151 	 * hard_* are block layer internals, no driver should touch them!
152 	 */
153 
154 	sector_t sector;		/* next sector to submit */
155 	sector_t hard_sector;		/* next sector to complete */
156 	unsigned long nr_sectors;	/* no. of sectors left to submit */
157 	unsigned long hard_nr_sectors;	/* no. of sectors left to complete */
158 	/* no. of sectors left to submit in the current segment */
159 	unsigned int current_nr_sectors;
160 
161 	/* no. of sectors left to complete in the current segment */
162 	unsigned int hard_cur_sectors;
163 
164 	struct bio *bio;
165 	struct bio *biotail;
166 
167 	struct hlist_node hash;	/* merge hash */
168 	/*
169 	 * The rb_node is only used inside the io scheduler, requests
170 	 * are pruned when moved to the dispatch queue. So let the
171 	 * completion_data share space with the rb_node.
172 	 */
173 	union {
174 		struct rb_node rb_node;	/* sort/lookup */
175 		void *completion_data;
176 	};
177 
178 	/*
179 	 * two pointers are available for the IO schedulers, if they need
180 	 * more they have to dynamically allocate it.
181 	 */
182 	void *elevator_private;
183 	void *elevator_private2;
184 
185 	struct gendisk *rq_disk;
186 	unsigned long start_time;
187 
188 	/* Number of scatter-gather DMA addr+len pairs after
189 	 * physical address coalescing is performed.
190 	 */
191 	unsigned short nr_phys_segments;
192 
193 	/* Number of scatter-gather addr+len pairs after
194 	 * physical and DMA remapping hardware coalescing is performed.
195 	 * This is the number of scatter-gather entries the driver
196 	 * will actually have to deal with after DMA mapping is done.
197 	 */
198 	unsigned short nr_hw_segments;
199 
200 	unsigned short ioprio;
201 
202 	void *special;
203 	char *buffer;
204 
205 	int tag;
206 	int errors;
207 
208 	int ref_count;
209 
210 	/*
211 	 * when request is used as a packet command carrier
212 	 */
213 	unsigned short cmd_len;
214 	unsigned char __cmd[BLK_MAX_CDB];
215 	unsigned char *cmd;
216 
217 	unsigned int data_len;
218 	unsigned int extra_len;	/* length of alignment and padding */
219 	unsigned int sense_len;
220 	void *data;
221 	void *sense;
222 
223 	unsigned int timeout;
224 	int retries;
225 
226 	/*
227 	 * completion callback.
228 	 */
229 	rq_end_io_fn *end_io;
230 	void *end_io_data;
231 
232 	/* for bidi */
233 	struct request *next_rq;
234 };
235 
236 /*
237  * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
238  * requests. Some step values could eventually be made generic.
239  */
240 struct request_pm_state
241 {
242 	/* PM state machine step value, currently driver specific */
243 	int	pm_step;
244 	/* requested PM state value (S1, S2, S3, S4, ...) */
245 	u32	pm_state;
246 	void*	data;		/* for driver use */
247 };
248 
249 #include <linux/elevator.h>
250 
251 typedef void (request_fn_proc) (struct request_queue *q);
252 typedef int (make_request_fn) (struct request_queue *q, struct bio *bio);
253 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
254 typedef void (unplug_fn) (struct request_queue *);
255 
256 struct bio_vec;
257 struct bvec_merge_data {
258 	struct block_device *bi_bdev;
259 	sector_t bi_sector;
260 	unsigned bi_size;
261 	unsigned long bi_rw;
262 };
263 typedef int (merge_bvec_fn) (struct request_queue *, struct bvec_merge_data *,
264 			     struct bio_vec *);
265 typedef void (prepare_flush_fn) (struct request_queue *, struct request *);
266 typedef void (softirq_done_fn)(struct request *);
267 typedef int (dma_drain_needed_fn)(struct request *);
268 
269 enum blk_queue_state {
270 	Queue_down,
271 	Queue_up,
272 };
273 
274 struct blk_queue_tag {
275 	struct request **tag_index;	/* map of busy tags */
276 	unsigned long *tag_map;		/* bit map of free/busy tags */
277 	int busy;			/* current depth */
278 	int max_depth;			/* what we will send to device */
279 	int real_max_depth;		/* what the array can hold */
280 	atomic_t refcnt;		/* map can be shared */
281 };
282 
283 #define BLK_SCSI_MAX_CMDS	(256)
284 #define BLK_SCSI_CMD_PER_LONG	(BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
285 
286 struct blk_cmd_filter {
287 	unsigned long read_ok[BLK_SCSI_CMD_PER_LONG];
288 	unsigned long write_ok[BLK_SCSI_CMD_PER_LONG];
289 	struct kobject kobj;
290 };
291 
292 struct request_queue
293 {
294 	/*
295 	 * Together with queue_head for cacheline sharing
296 	 */
297 	struct list_head	queue_head;
298 	struct request		*last_merge;
299 	elevator_t		*elevator;
300 
301 	/*
302 	 * the queue request freelist, one for reads and one for writes
303 	 */
304 	struct request_list	rq;
305 
306 	request_fn_proc		*request_fn;
307 	make_request_fn		*make_request_fn;
308 	prep_rq_fn		*prep_rq_fn;
309 	unplug_fn		*unplug_fn;
310 	merge_bvec_fn		*merge_bvec_fn;
311 	prepare_flush_fn	*prepare_flush_fn;
312 	softirq_done_fn		*softirq_done_fn;
313 	dma_drain_needed_fn	*dma_drain_needed;
314 
315 	/*
316 	 * Dispatch queue sorting
317 	 */
318 	sector_t		end_sector;
319 	struct request		*boundary_rq;
320 
321 	/*
322 	 * Auto-unplugging state
323 	 */
324 	struct timer_list	unplug_timer;
325 	int			unplug_thresh;	/* After this many requests */
326 	unsigned long		unplug_delay;	/* After this many jiffies */
327 	struct work_struct	unplug_work;
328 
329 	struct backing_dev_info	backing_dev_info;
330 
331 	/*
332 	 * The queue owner gets to use this for whatever they like.
333 	 * ll_rw_blk doesn't touch it.
334 	 */
335 	void			*queuedata;
336 
337 	/*
338 	 * queue needs bounce pages for pages above this limit
339 	 */
340 	unsigned long		bounce_pfn;
341 	gfp_t			bounce_gfp;
342 
343 	/*
344 	 * various queue flags, see QUEUE_* below
345 	 */
346 	unsigned long		queue_flags;
347 
348 	/*
349 	 * protects queue structures from reentrancy. ->__queue_lock should
350 	 * _never_ be used directly, it is queue private. always use
351 	 * ->queue_lock.
352 	 */
353 	spinlock_t		__queue_lock;
354 	spinlock_t		*queue_lock;
355 
356 	/*
357 	 * queue kobject
358 	 */
359 	struct kobject kobj;
360 
361 	/*
362 	 * queue settings
363 	 */
364 	unsigned long		nr_requests;	/* Max # of requests */
365 	unsigned int		nr_congestion_on;
366 	unsigned int		nr_congestion_off;
367 	unsigned int		nr_batching;
368 
369 	unsigned int		max_sectors;
370 	unsigned int		max_hw_sectors;
371 	unsigned short		max_phys_segments;
372 	unsigned short		max_hw_segments;
373 	unsigned short		hardsect_size;
374 	unsigned int		max_segment_size;
375 
376 	unsigned long		seg_boundary_mask;
377 	void			*dma_drain_buffer;
378 	unsigned int		dma_drain_size;
379 	unsigned int		dma_pad_mask;
380 	unsigned int		dma_alignment;
381 
382 	struct blk_queue_tag	*queue_tags;
383 	struct list_head	tag_busy_list;
384 
385 	unsigned int		nr_sorted;
386 	unsigned int		in_flight;
387 
388 	/*
389 	 * sg stuff
390 	 */
391 	unsigned int		sg_timeout;
392 	unsigned int		sg_reserved_size;
393 	int			node;
394 #ifdef CONFIG_BLK_DEV_IO_TRACE
395 	struct blk_trace	*blk_trace;
396 #endif
397 	/*
398 	 * reserved for flush operations
399 	 */
400 	unsigned int		ordered, next_ordered, ordseq;
401 	int			orderr, ordcolor;
402 	struct request		pre_flush_rq, bar_rq, post_flush_rq;
403 	struct request		*orig_bar_rq;
404 
405 	struct mutex		sysfs_lock;
406 
407 #if defined(CONFIG_BLK_DEV_BSG)
408 	struct bsg_class_device bsg_dev;
409 #endif
410 	struct blk_cmd_filter cmd_filter;
411 };
412 
413 #define QUEUE_FLAG_CLUSTER	0	/* cluster several segments into 1 */
414 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */
415 #define QUEUE_FLAG_STOPPED	2	/* queue is stopped */
416 #define	QUEUE_FLAG_READFULL	3	/* read queue has been filled */
417 #define QUEUE_FLAG_WRITEFULL	4	/* write queue has been filled */
418 #define QUEUE_FLAG_DEAD		5	/* queue being torn down */
419 #define QUEUE_FLAG_REENTER	6	/* Re-entrancy avoidance */
420 #define QUEUE_FLAG_PLUGGED	7	/* queue is plugged */
421 #define QUEUE_FLAG_ELVSWITCH	8	/* don't use elevator, just do FIFO */
422 #define QUEUE_FLAG_BIDI		9	/* queue supports bidi requests */
423 #define QUEUE_FLAG_NOMERGES    10	/* disable merge attempts */
424 
425 static inline int queue_is_locked(struct request_queue *q)
426 {
427 #ifdef CONFIG_SMP
428 	spinlock_t *lock = q->queue_lock;
429 	return lock && spin_is_locked(lock);
430 #else
431 	return 1;
432 #endif
433 }
434 
435 static inline void queue_flag_set_unlocked(unsigned int flag,
436 					   struct request_queue *q)
437 {
438 	__set_bit(flag, &q->queue_flags);
439 }
440 
441 static inline int queue_flag_test_and_clear(unsigned int flag,
442 					    struct request_queue *q)
443 {
444 	WARN_ON_ONCE(!queue_is_locked(q));
445 
446 	if (test_bit(flag, &q->queue_flags)) {
447 		__clear_bit(flag, &q->queue_flags);
448 		return 1;
449 	}
450 
451 	return 0;
452 }
453 
454 static inline int queue_flag_test_and_set(unsigned int flag,
455 					  struct request_queue *q)
456 {
457 	WARN_ON_ONCE(!queue_is_locked(q));
458 
459 	if (!test_bit(flag, &q->queue_flags)) {
460 		__set_bit(flag, &q->queue_flags);
461 		return 0;
462 	}
463 
464 	return 1;
465 }
466 
467 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
468 {
469 	WARN_ON_ONCE(!queue_is_locked(q));
470 	__set_bit(flag, &q->queue_flags);
471 }
472 
473 static inline void queue_flag_clear_unlocked(unsigned int flag,
474 					     struct request_queue *q)
475 {
476 	__clear_bit(flag, &q->queue_flags);
477 }
478 
479 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
480 {
481 	WARN_ON_ONCE(!queue_is_locked(q));
482 	__clear_bit(flag, &q->queue_flags);
483 }
484 
485 enum {
486 	/*
487 	 * Hardbarrier is supported with one of the following methods.
488 	 *
489 	 * NONE		: hardbarrier unsupported
490 	 * DRAIN	: ordering by draining is enough
491 	 * DRAIN_FLUSH	: ordering by draining w/ pre and post flushes
492 	 * DRAIN_FUA	: ordering by draining w/ pre flush and FUA write
493 	 * TAG		: ordering by tag is enough
494 	 * TAG_FLUSH	: ordering by tag w/ pre and post flushes
495 	 * TAG_FUA	: ordering by tag w/ pre flush and FUA write
496 	 */
497 	QUEUE_ORDERED_NONE	= 0x00,
498 	QUEUE_ORDERED_DRAIN	= 0x01,
499 	QUEUE_ORDERED_TAG	= 0x02,
500 
501 	QUEUE_ORDERED_PREFLUSH	= 0x10,
502 	QUEUE_ORDERED_POSTFLUSH	= 0x20,
503 	QUEUE_ORDERED_FUA	= 0x40,
504 
505 	QUEUE_ORDERED_DRAIN_FLUSH = QUEUE_ORDERED_DRAIN |
506 			QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
507 	QUEUE_ORDERED_DRAIN_FUA	= QUEUE_ORDERED_DRAIN |
508 			QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
509 	QUEUE_ORDERED_TAG_FLUSH	= QUEUE_ORDERED_TAG |
510 			QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
511 	QUEUE_ORDERED_TAG_FUA	= QUEUE_ORDERED_TAG |
512 			QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
513 
514 	/*
515 	 * Ordered operation sequence
516 	 */
517 	QUEUE_ORDSEQ_STARTED	= 0x01,	/* flushing in progress */
518 	QUEUE_ORDSEQ_DRAIN	= 0x02,	/* waiting for the queue to be drained */
519 	QUEUE_ORDSEQ_PREFLUSH	= 0x04,	/* pre-flushing in progress */
520 	QUEUE_ORDSEQ_BAR	= 0x08,	/* original barrier req in progress */
521 	QUEUE_ORDSEQ_POSTFLUSH	= 0x10,	/* post-flushing in progress */
522 	QUEUE_ORDSEQ_DONE	= 0x20,
523 };
524 
525 #define blk_queue_plugged(q)	test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
526 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
527 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
528 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
529 #define blk_queue_flushing(q)	((q)->ordseq)
530 
531 #define blk_fs_request(rq)	((rq)->cmd_type == REQ_TYPE_FS)
532 #define blk_pc_request(rq)	((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
533 #define blk_special_request(rq)	((rq)->cmd_type == REQ_TYPE_SPECIAL)
534 #define blk_sense_request(rq)	((rq)->cmd_type == REQ_TYPE_SENSE)
535 
536 #define blk_noretry_request(rq)	((rq)->cmd_flags & REQ_FAILFAST)
537 #define blk_rq_started(rq)	((rq)->cmd_flags & REQ_STARTED)
538 
539 #define blk_account_rq(rq)	(blk_rq_started(rq) && blk_fs_request(rq))
540 
541 #define blk_pm_suspend_request(rq)	((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
542 #define blk_pm_resume_request(rq)	((rq)->cmd_type == REQ_TYPE_PM_RESUME)
543 #define blk_pm_request(rq)	\
544 	(blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
545 
546 #define blk_sorted_rq(rq)	((rq)->cmd_flags & REQ_SORTED)
547 #define blk_barrier_rq(rq)	((rq)->cmd_flags & REQ_HARDBARRIER)
548 #define blk_fua_rq(rq)		((rq)->cmd_flags & REQ_FUA)
549 #define blk_bidi_rq(rq)		((rq)->next_rq != NULL)
550 #define blk_empty_barrier(rq)	(blk_barrier_rq(rq) && blk_fs_request(rq) && !(rq)->hard_nr_sectors)
551 /* rq->queuelist of dequeued request must be list_empty() */
552 #define blk_queued_rq(rq)	(!list_empty(&(rq)->queuelist))
553 
554 #define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
555 
556 #define rq_data_dir(rq)		((rq)->cmd_flags & 1)
557 
558 /*
559  * We regard a request as sync, if it's a READ or a SYNC write.
560  */
561 #define rq_is_sync(rq)		(rq_data_dir((rq)) == READ || (rq)->cmd_flags & REQ_RW_SYNC)
562 #define rq_is_meta(rq)		((rq)->cmd_flags & REQ_RW_META)
563 
564 static inline int blk_queue_full(struct request_queue *q, int rw)
565 {
566 	if (rw == READ)
567 		return test_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
568 	return test_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
569 }
570 
571 static inline void blk_set_queue_full(struct request_queue *q, int rw)
572 {
573 	if (rw == READ)
574 		queue_flag_set(QUEUE_FLAG_READFULL, q);
575 	else
576 		queue_flag_set(QUEUE_FLAG_WRITEFULL, q);
577 }
578 
579 static inline void blk_clear_queue_full(struct request_queue *q, int rw)
580 {
581 	if (rw == READ)
582 		queue_flag_clear(QUEUE_FLAG_READFULL, q);
583 	else
584 		queue_flag_clear(QUEUE_FLAG_WRITEFULL, q);
585 }
586 
587 
588 /*
589  * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
590  * it already be started by driver.
591  */
592 #define RQ_NOMERGE_FLAGS	\
593 	(REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
594 #define rq_mergeable(rq)	\
595 	(!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && blk_fs_request((rq)))
596 
597 /*
598  * q->prep_rq_fn return values
599  */
600 #define BLKPREP_OK		0	/* serve it */
601 #define BLKPREP_KILL		1	/* fatal error, kill */
602 #define BLKPREP_DEFER		2	/* leave on queue */
603 
604 extern unsigned long blk_max_low_pfn, blk_max_pfn;
605 
606 /*
607  * standard bounce addresses:
608  *
609  * BLK_BOUNCE_HIGH	: bounce all highmem pages
610  * BLK_BOUNCE_ANY	: don't bounce anything
611  * BLK_BOUNCE_ISA	: bounce pages above ISA DMA boundary
612  */
613 
614 #if BITS_PER_LONG == 32
615 #define BLK_BOUNCE_HIGH		((u64)blk_max_low_pfn << PAGE_SHIFT)
616 #else
617 #define BLK_BOUNCE_HIGH		-1ULL
618 #endif
619 #define BLK_BOUNCE_ANY		(-1ULL)
620 #define BLK_BOUNCE_ISA		(ISA_DMA_THRESHOLD)
621 
622 /*
623  * default timeout for SG_IO if none specified
624  */
625 #define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
626 
627 #ifdef CONFIG_BOUNCE
628 extern int init_emergency_isa_pool(void);
629 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
630 #else
631 static inline int init_emergency_isa_pool(void)
632 {
633 	return 0;
634 }
635 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
636 {
637 }
638 #endif /* CONFIG_MMU */
639 
640 struct req_iterator {
641 	int i;
642 	struct bio *bio;
643 };
644 
645 /* This should not be used directly - use rq_for_each_segment */
646 #define __rq_for_each_bio(_bio, rq)	\
647 	if ((rq->bio))			\
648 		for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
649 
650 #define rq_for_each_segment(bvl, _rq, _iter)			\
651 	__rq_for_each_bio(_iter.bio, _rq)			\
652 		bio_for_each_segment(bvl, _iter.bio, _iter.i)
653 
654 #define rq_iter_last(rq, _iter)					\
655 		(_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
656 
657 extern int blk_register_queue(struct gendisk *disk);
658 extern void blk_unregister_queue(struct gendisk *disk);
659 extern void register_disk(struct gendisk *dev);
660 extern void generic_make_request(struct bio *bio);
661 extern void blk_rq_init(struct request_queue *q, struct request *rq);
662 extern void blk_put_request(struct request *);
663 extern void __blk_put_request(struct request_queue *, struct request *);
664 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
665 extern void blk_insert_request(struct request_queue *, struct request *, int, void *);
666 extern void blk_requeue_request(struct request_queue *, struct request *);
667 extern void blk_plug_device(struct request_queue *);
668 extern void blk_plug_device_unlocked(struct request_queue *);
669 extern int blk_remove_plug(struct request_queue *);
670 extern void blk_recount_segments(struct request_queue *, struct bio *);
671 extern int scsi_cmd_ioctl(struct file *, struct request_queue *,
672 			  struct gendisk *, unsigned int, void __user *);
673 extern int sg_scsi_ioctl(struct file *, struct request_queue *,
674 		struct gendisk *, struct scsi_ioctl_command __user *);
675 
676 /*
677  * Temporary export, until SCSI gets fixed up.
678  */
679 extern int blk_rq_append_bio(struct request_queue *q, struct request *rq,
680 			     struct bio *bio);
681 
682 /*
683  * A queue has just exitted congestion.  Note this in the global counter of
684  * congested queues, and wake up anyone who was waiting for requests to be
685  * put back.
686  */
687 static inline void blk_clear_queue_congested(struct request_queue *q, int rw)
688 {
689 	clear_bdi_congested(&q->backing_dev_info, rw);
690 }
691 
692 /*
693  * A queue has just entered congestion.  Flag that in the queue's VM-visible
694  * state flags and increment the global gounter of congested queues.
695  */
696 static inline void blk_set_queue_congested(struct request_queue *q, int rw)
697 {
698 	set_bdi_congested(&q->backing_dev_info, rw);
699 }
700 
701 extern void blk_start_queue(struct request_queue *q);
702 extern void blk_stop_queue(struct request_queue *q);
703 extern void blk_sync_queue(struct request_queue *q);
704 extern void __blk_stop_queue(struct request_queue *q);
705 extern void __blk_run_queue(struct request_queue *);
706 extern void blk_run_queue(struct request_queue *);
707 extern void blk_start_queueing(struct request_queue *);
708 extern int blk_rq_map_user(struct request_queue *, struct request *, void __user *, unsigned long);
709 extern int blk_rq_unmap_user(struct bio *);
710 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
711 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
712 			       struct sg_iovec *, int, unsigned int);
713 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
714 			  struct request *, int);
715 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
716 				  struct request *, int, rq_end_io_fn *);
717 extern void blk_unplug(struct request_queue *q);
718 
719 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
720 {
721 	return bdev->bd_disk->queue;
722 }
723 
724 static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
725 				       struct page *page)
726 {
727 	if (bdi && bdi->unplug_io_fn)
728 		bdi->unplug_io_fn(bdi, page);
729 }
730 
731 static inline void blk_run_address_space(struct address_space *mapping)
732 {
733 	if (mapping)
734 		blk_run_backing_dev(mapping->backing_dev_info, NULL);
735 }
736 
737 /*
738  * blk_end_request() and friends.
739  * __blk_end_request() and end_request() must be called with
740  * the request queue spinlock acquired.
741  *
742  * Several drivers define their own end_request and call
743  * blk_end_request() for parts of the original function.
744  * This prevents code duplication in drivers.
745  */
746 extern int blk_end_request(struct request *rq, int error,
747 				unsigned int nr_bytes);
748 extern int __blk_end_request(struct request *rq, int error,
749 				unsigned int nr_bytes);
750 extern int blk_end_bidi_request(struct request *rq, int error,
751 				unsigned int nr_bytes, unsigned int bidi_bytes);
752 extern void end_request(struct request *, int);
753 extern void end_queued_request(struct request *, int);
754 extern void end_dequeued_request(struct request *, int);
755 extern int blk_end_request_callback(struct request *rq, int error,
756 				unsigned int nr_bytes,
757 				int (drv_callback)(struct request *));
758 extern void blk_complete_request(struct request *);
759 
760 /*
761  * blk_end_request() takes bytes instead of sectors as a complete size.
762  * blk_rq_bytes() returns bytes left to complete in the entire request.
763  * blk_rq_cur_bytes() returns bytes left to complete in the current segment.
764  */
765 extern unsigned int blk_rq_bytes(struct request *rq);
766 extern unsigned int blk_rq_cur_bytes(struct request *rq);
767 
768 static inline void blkdev_dequeue_request(struct request *req)
769 {
770 	elv_dequeue_request(req->q, req);
771 }
772 
773 /*
774  * Access functions for manipulating queue properties
775  */
776 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
777 					spinlock_t *lock, int node_id);
778 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
779 extern void blk_cleanup_queue(struct request_queue *);
780 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
781 extern void blk_queue_bounce_limit(struct request_queue *, u64);
782 extern void blk_queue_max_sectors(struct request_queue *, unsigned int);
783 extern void blk_queue_max_phys_segments(struct request_queue *, unsigned short);
784 extern void blk_queue_max_hw_segments(struct request_queue *, unsigned short);
785 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
786 extern void blk_queue_hardsect_size(struct request_queue *, unsigned short);
787 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
788 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
789 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
790 extern int blk_queue_dma_drain(struct request_queue *q,
791 			       dma_drain_needed_fn *dma_drain_needed,
792 			       void *buf, unsigned int size);
793 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
794 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
795 extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *);
796 extern void blk_queue_dma_alignment(struct request_queue *, int);
797 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
798 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
799 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
800 extern int blk_queue_ordered(struct request_queue *, unsigned, prepare_flush_fn *);
801 extern int blk_do_ordered(struct request_queue *, struct request **);
802 extern unsigned blk_ordered_cur_seq(struct request_queue *);
803 extern unsigned blk_ordered_req_seq(struct request *);
804 extern void blk_ordered_complete_seq(struct request_queue *, unsigned, int);
805 
806 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
807 extern void blk_dump_rq_flags(struct request *, char *);
808 extern void generic_unplug_device(struct request_queue *);
809 extern void __generic_unplug_device(struct request_queue *);
810 extern long nr_blockdev_pages(void);
811 
812 int blk_get_queue(struct request_queue *);
813 struct request_queue *blk_alloc_queue(gfp_t);
814 struct request_queue *blk_alloc_queue_node(gfp_t, int);
815 extern void blk_put_queue(struct request_queue *);
816 
817 /*
818  * tag stuff
819  */
820 #define blk_rq_tagged(rq)		((rq)->cmd_flags & REQ_QUEUED)
821 extern int blk_queue_start_tag(struct request_queue *, struct request *);
822 extern struct request *blk_queue_find_tag(struct request_queue *, int);
823 extern void blk_queue_end_tag(struct request_queue *, struct request *);
824 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *);
825 extern void blk_queue_free_tags(struct request_queue *);
826 extern int blk_queue_resize_tags(struct request_queue *, int);
827 extern void blk_queue_invalidate_tags(struct request_queue *);
828 extern struct blk_queue_tag *blk_init_tags(int);
829 extern void blk_free_tags(struct blk_queue_tag *);
830 
831 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
832 						int tag)
833 {
834 	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
835 		return NULL;
836 	return bqt->tag_index[tag];
837 }
838 
839 extern int blkdev_issue_flush(struct block_device *, sector_t *);
840 
841 /*
842 * command filter functions
843 */
844 extern int blk_verify_command(struct blk_cmd_filter *filter,
845 			      unsigned char *cmd, int has_write_perm);
846 extern int blk_register_filter(struct gendisk *disk);
847 extern void blk_unregister_filter(struct gendisk *disk);
848 extern void blk_set_cmd_filter_defaults(struct blk_cmd_filter *filter);
849 
850 #define MAX_PHYS_SEGMENTS 128
851 #define MAX_HW_SEGMENTS 128
852 #define SAFE_MAX_SECTORS 255
853 #define BLK_DEF_MAX_SECTORS 1024
854 
855 #define MAX_SEGMENT_SIZE	65536
856 
857 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
858 
859 static inline int queue_hardsect_size(struct request_queue *q)
860 {
861 	int retval = 512;
862 
863 	if (q && q->hardsect_size)
864 		retval = q->hardsect_size;
865 
866 	return retval;
867 }
868 
869 static inline int bdev_hardsect_size(struct block_device *bdev)
870 {
871 	return queue_hardsect_size(bdev_get_queue(bdev));
872 }
873 
874 static inline int queue_dma_alignment(struct request_queue *q)
875 {
876 	return q ? q->dma_alignment : 511;
877 }
878 
879 /* assumes size > 256 */
880 static inline unsigned int blksize_bits(unsigned int size)
881 {
882 	unsigned int bits = 8;
883 	do {
884 		bits++;
885 		size >>= 1;
886 	} while (size > 256);
887 	return bits;
888 }
889 
890 static inline unsigned int block_size(struct block_device *bdev)
891 {
892 	return bdev->bd_block_size;
893 }
894 
895 typedef struct {struct page *v;} Sector;
896 
897 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
898 
899 static inline void put_dev_sector(Sector p)
900 {
901 	page_cache_release(p.v);
902 }
903 
904 struct work_struct;
905 int kblockd_schedule_work(struct work_struct *work);
906 void kblockd_flush_work(struct work_struct *work);
907 
908 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
909 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
910 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
911 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
912 
913 #if defined(CONFIG_BLK_DEV_INTEGRITY)
914 
915 #define INTEGRITY_FLAG_READ	2	/* verify data integrity on read */
916 #define INTEGRITY_FLAG_WRITE	4	/* generate data integrity on write */
917 
918 struct blk_integrity_exchg {
919 	void			*prot_buf;
920 	void			*data_buf;
921 	sector_t		sector;
922 	unsigned int		data_size;
923 	unsigned short		sector_size;
924 	const char		*disk_name;
925 };
926 
927 typedef void (integrity_gen_fn) (struct blk_integrity_exchg *);
928 typedef int (integrity_vrfy_fn) (struct blk_integrity_exchg *);
929 typedef void (integrity_set_tag_fn) (void *, void *, unsigned int);
930 typedef void (integrity_get_tag_fn) (void *, void *, unsigned int);
931 
932 struct blk_integrity {
933 	integrity_gen_fn	*generate_fn;
934 	integrity_vrfy_fn	*verify_fn;
935 	integrity_set_tag_fn	*set_tag_fn;
936 	integrity_get_tag_fn	*get_tag_fn;
937 
938 	unsigned short		flags;
939 	unsigned short		tuple_size;
940 	unsigned short		sector_size;
941 	unsigned short		tag_size;
942 
943 	const char		*name;
944 
945 	struct kobject		kobj;
946 };
947 
948 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
949 extern void blk_integrity_unregister(struct gendisk *);
950 extern int blk_integrity_compare(struct block_device *, struct block_device *);
951 extern int blk_rq_map_integrity_sg(struct request *, struct scatterlist *);
952 extern int blk_rq_count_integrity_sg(struct request *);
953 
954 static inline unsigned short blk_integrity_tuple_size(struct blk_integrity *bi)
955 {
956 	if (bi)
957 		return bi->tuple_size;
958 
959 	return 0;
960 }
961 
962 static inline struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
963 {
964 	return bdev->bd_disk->integrity;
965 }
966 
967 static inline unsigned int bdev_get_tag_size(struct block_device *bdev)
968 {
969 	struct blk_integrity *bi = bdev_get_integrity(bdev);
970 
971 	if (bi)
972 		return bi->tag_size;
973 
974 	return 0;
975 }
976 
977 static inline int bdev_integrity_enabled(struct block_device *bdev, int rw)
978 {
979 	struct blk_integrity *bi = bdev_get_integrity(bdev);
980 
981 	if (bi == NULL)
982 		return 0;
983 
984 	if (rw == READ && bi->verify_fn != NULL &&
985 	    (bi->flags & INTEGRITY_FLAG_READ))
986 		return 1;
987 
988 	if (rw == WRITE && bi->generate_fn != NULL &&
989 	    (bi->flags & INTEGRITY_FLAG_WRITE))
990 		return 1;
991 
992 	return 0;
993 }
994 
995 static inline int blk_integrity_rq(struct request *rq)
996 {
997 	if (rq->bio == NULL)
998 		return 0;
999 
1000 	return bio_integrity(rq->bio);
1001 }
1002 
1003 #else /* CONFIG_BLK_DEV_INTEGRITY */
1004 
1005 #define blk_integrity_rq(rq)			(0)
1006 #define blk_rq_count_integrity_sg(a)		(0)
1007 #define blk_rq_map_integrity_sg(a, b)		(0)
1008 #define bdev_get_integrity(a)			(0)
1009 #define bdev_get_tag_size(a)			(0)
1010 #define blk_integrity_compare(a, b)		(0)
1011 #define blk_integrity_register(a, b)		(0)
1012 #define blk_integrity_unregister(a)		do { } while (0);
1013 
1014 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1015 
1016 #else /* CONFIG_BLOCK */
1017 /*
1018  * stubs for when the block layer is configured out
1019  */
1020 #define buffer_heads_over_limit 0
1021 
1022 static inline long nr_blockdev_pages(void)
1023 {
1024 	return 0;
1025 }
1026 
1027 #endif /* CONFIG_BLOCK */
1028 
1029 #endif
1030