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 struct request_pm_state;
30 struct blk_trace;
31 struct request;
32 struct sg_io_hdr;
33
34 #define BLKDEV_MIN_RQ 4
35 #define BLKDEV_MAX_RQ 128 /* Default maximum */
36
37 struct request;
38 typedef void (rq_end_io_fn)(struct request *, int);
39
40 struct request_list {
41 int count[2];
42 int starved[2];
43 int elvpriv;
44 mempool_t *rq_pool;
45 wait_queue_head_t wait[2];
46 };
47
48 /*
49 * request command types
50 */
51 enum rq_cmd_type_bits {
52 REQ_TYPE_FS = 1, /* fs request */
53 REQ_TYPE_BLOCK_PC, /* scsi command */
54 REQ_TYPE_SENSE, /* sense request */
55 REQ_TYPE_PM_SUSPEND, /* suspend request */
56 REQ_TYPE_PM_RESUME, /* resume request */
57 REQ_TYPE_PM_SHUTDOWN, /* shutdown 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 REQ_LB_OP_EJECT = 0x40, /* eject request */
80 REQ_LB_OP_FLUSH = 0x41, /* flush request */
81 REQ_LB_OP_DISCARD = 0x42, /* discard sectors */
82 };
83
84 /*
85 * request type modified bits. first two bits match BIO_RW* bits, important
86 */
87 enum rq_flag_bits {
88 __REQ_RW, /* not set, read. set, write */
89 __REQ_FAILFAST_DEV, /* no driver retries of device errors */
90 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
91 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
92 __REQ_DISCARD, /* request to discard sectors */
93 __REQ_SORTED, /* elevator knows about this request */
94 __REQ_SOFTBARRIER, /* may not be passed by ioscheduler */
95 __REQ_HARDBARRIER, /* may not be passed by drive either */
96 __REQ_FUA, /* forced unit access */
97 __REQ_NOMERGE, /* don't touch this for merging */
98 __REQ_STARTED, /* drive already may have started this one */
99 __REQ_DONTPREP, /* don't call prep for this one */
100 __REQ_QUEUED, /* uses queueing */
101 __REQ_ELVPRIV, /* elevator private data attached */
102 __REQ_FAILED, /* set if the request failed */
103 __REQ_QUIET, /* don't worry about errors */
104 __REQ_PREEMPT, /* set for "ide_preempt" requests */
105 __REQ_ORDERED_COLOR, /* is before or after barrier */
106 __REQ_RW_SYNC, /* request is sync (O_DIRECT) */
107 __REQ_ALLOCED, /* request came from our alloc pool */
108 __REQ_RW_META, /* metadata io request */
109 __REQ_COPY_USER, /* contains copies of user pages */
110 __REQ_INTEGRITY, /* integrity metadata has been remapped */
111 __REQ_UNPLUG, /* unplug queue on submission */
112 __REQ_NR_BITS, /* stops here */
113 };
114
115 #define REQ_RW (1 << __REQ_RW)
116 #define REQ_FAILFAST_DEV (1 << __REQ_FAILFAST_DEV)
117 #define REQ_FAILFAST_TRANSPORT (1 << __REQ_FAILFAST_TRANSPORT)
118 #define REQ_FAILFAST_DRIVER (1 << __REQ_FAILFAST_DRIVER)
119 #define REQ_DISCARD (1 << __REQ_DISCARD)
120 #define REQ_SORTED (1 << __REQ_SORTED)
121 #define REQ_SOFTBARRIER (1 << __REQ_SOFTBARRIER)
122 #define REQ_HARDBARRIER (1 << __REQ_HARDBARRIER)
123 #define REQ_FUA (1 << __REQ_FUA)
124 #define REQ_NOMERGE (1 << __REQ_NOMERGE)
125 #define REQ_STARTED (1 << __REQ_STARTED)
126 #define REQ_DONTPREP (1 << __REQ_DONTPREP)
127 #define REQ_QUEUED (1 << __REQ_QUEUED)
128 #define REQ_ELVPRIV (1 << __REQ_ELVPRIV)
129 #define REQ_FAILED (1 << __REQ_FAILED)
130 #define REQ_QUIET (1 << __REQ_QUIET)
131 #define REQ_PREEMPT (1 << __REQ_PREEMPT)
132 #define REQ_ORDERED_COLOR (1 << __REQ_ORDERED_COLOR)
133 #define REQ_RW_SYNC (1 << __REQ_RW_SYNC)
134 #define REQ_ALLOCED (1 << __REQ_ALLOCED)
135 #define REQ_RW_META (1 << __REQ_RW_META)
136 #define REQ_COPY_USER (1 << __REQ_COPY_USER)
137 #define REQ_INTEGRITY (1 << __REQ_INTEGRITY)
138 #define REQ_UNPLUG (1 << __REQ_UNPLUG)
139
140 #define BLK_MAX_CDB 16
141
142 /*
143 * try to put the fields that are referenced together in the same cacheline.
144 * if you modify this structure, be sure to check block/blk-core.c:rq_init()
145 * as well!
146 */
147 struct request {
148 struct list_head queuelist;
149 struct call_single_data csd;
150 int cpu;
151
152 struct request_queue *q;
153
154 unsigned int cmd_flags;
155 enum rq_cmd_type_bits cmd_type;
156 unsigned long atomic_flags;
157
158 /* Maintain bio traversal state for part by part I/O submission.
159 * hard_* are block layer internals, no driver should touch them!
160 */
161
162 sector_t sector; /* next sector to submit */
163 sector_t hard_sector; /* next sector to complete */
164 unsigned long nr_sectors; /* no. of sectors left to submit */
165 unsigned long hard_nr_sectors; /* no. of sectors left to complete */
166 /* no. of sectors left to submit in the current segment */
167 unsigned int current_nr_sectors;
168
169 /* no. of sectors left to complete in the current segment */
170 unsigned int hard_cur_sectors;
171
172 struct bio *bio;
173 struct bio *biotail;
174
175 struct hlist_node hash; /* merge hash */
176 /*
177 * The rb_node is only used inside the io scheduler, requests
178 * are pruned when moved to the dispatch queue. So let the
179 * completion_data share space with the rb_node.
180 */
181 union {
182 struct rb_node rb_node; /* sort/lookup */
183 void *completion_data;
184 };
185
186 /*
187 * two pointers are available for the IO schedulers, if they need
188 * more they have to dynamically allocate it.
189 */
190 void *elevator_private;
191 void *elevator_private2;
192
193 struct gendisk *rq_disk;
194 unsigned long start_time;
195
196 /* Number of scatter-gather DMA addr+len pairs after
197 * physical address coalescing is performed.
198 */
199 unsigned short nr_phys_segments;
200
201 unsigned short ioprio;
202
203 void *special;
204 char *buffer;
205
206 int tag;
207 int errors;
208
209 int ref_count;
210
211 /*
212 * when request is used as a packet command carrier
213 */
214 unsigned short cmd_len;
215 unsigned char __cmd[BLK_MAX_CDB];
216 unsigned char *cmd;
217
218 unsigned int data_len;
219 unsigned int extra_len; /* length of alignment and padding */
220 unsigned int sense_len;
221 void *data;
222 void *sense;
223
224 unsigned long deadline;
225 struct list_head timeout_list;
226 unsigned int timeout;
227 int retries;
228
229 /*
230 * completion callback.
231 */
232 rq_end_io_fn *end_io;
233 void *end_io_data;
234
235 /* for bidi */
236 struct request *next_rq;
237 };
238
req_get_ioprio(struct request * req)239 static inline unsigned short req_get_ioprio(struct request *req)
240 {
241 return req->ioprio;
242 }
243
244 /*
245 * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
246 * requests. Some step values could eventually be made generic.
247 */
248 struct request_pm_state
249 {
250 /* PM state machine step value, currently driver specific */
251 int pm_step;
252 /* requested PM state value (S1, S2, S3, S4, ...) */
253 u32 pm_state;
254 void* data; /* for driver use */
255 };
256
257 #include <linux/elevator.h>
258
259 typedef void (request_fn_proc) (struct request_queue *q);
260 typedef int (make_request_fn) (struct request_queue *q, struct bio *bio);
261 typedef int (prep_rq_fn) (struct request_queue *, struct request *);
262 typedef void (unplug_fn) (struct request_queue *);
263 typedef int (prepare_discard_fn) (struct request_queue *, struct request *);
264
265 struct bio_vec;
266 struct bvec_merge_data {
267 struct block_device *bi_bdev;
268 sector_t bi_sector;
269 unsigned bi_size;
270 unsigned long bi_rw;
271 };
272 typedef int (merge_bvec_fn) (struct request_queue *, struct bvec_merge_data *,
273 struct bio_vec *);
274 typedef void (prepare_flush_fn) (struct request_queue *, struct request *);
275 typedef void (softirq_done_fn)(struct request *);
276 typedef int (dma_drain_needed_fn)(struct request *);
277 typedef int (lld_busy_fn) (struct request_queue *q);
278
279 enum blk_eh_timer_return {
280 BLK_EH_NOT_HANDLED,
281 BLK_EH_HANDLED,
282 BLK_EH_RESET_TIMER,
283 };
284
285 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *);
286
287 enum blk_queue_state {
288 Queue_down,
289 Queue_up,
290 };
291
292 struct blk_queue_tag {
293 struct request **tag_index; /* map of busy tags */
294 unsigned long *tag_map; /* bit map of free/busy tags */
295 int busy; /* current depth */
296 int max_depth; /* what we will send to device */
297 int real_max_depth; /* what the array can hold */
298 atomic_t refcnt; /* map can be shared */
299 };
300
301 #define BLK_SCSI_MAX_CMDS (256)
302 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8))
303
304 struct blk_cmd_filter {
305 unsigned long read_ok[BLK_SCSI_CMD_PER_LONG];
306 unsigned long write_ok[BLK_SCSI_CMD_PER_LONG];
307 struct kobject kobj;
308 };
309
310 struct request_queue
311 {
312 /*
313 * Together with queue_head for cacheline sharing
314 */
315 struct list_head queue_head;
316 struct request *last_merge;
317 struct elevator_queue *elevator;
318
319 /*
320 * the queue request freelist, one for reads and one for writes
321 */
322 struct request_list rq;
323
324 request_fn_proc *request_fn;
325 make_request_fn *make_request_fn;
326 prep_rq_fn *prep_rq_fn;
327 unplug_fn *unplug_fn;
328 prepare_discard_fn *prepare_discard_fn;
329 merge_bvec_fn *merge_bvec_fn;
330 prepare_flush_fn *prepare_flush_fn;
331 softirq_done_fn *softirq_done_fn;
332 rq_timed_out_fn *rq_timed_out_fn;
333 dma_drain_needed_fn *dma_drain_needed;
334 lld_busy_fn *lld_busy_fn;
335
336 /*
337 * Dispatch queue sorting
338 */
339 sector_t end_sector;
340 struct request *boundary_rq;
341
342 /*
343 * Auto-unplugging state
344 */
345 struct timer_list unplug_timer;
346 int unplug_thresh; /* After this many requests */
347 unsigned long unplug_delay; /* After this many jiffies */
348 struct work_struct unplug_work;
349
350 struct backing_dev_info backing_dev_info;
351
352 /*
353 * The queue owner gets to use this for whatever they like.
354 * ll_rw_blk doesn't touch it.
355 */
356 void *queuedata;
357
358 /*
359 * queue needs bounce pages for pages above this limit
360 */
361 unsigned long bounce_pfn;
362 gfp_t bounce_gfp;
363
364 /*
365 * various queue flags, see QUEUE_* below
366 */
367 unsigned long queue_flags;
368
369 /*
370 * protects queue structures from reentrancy. ->__queue_lock should
371 * _never_ be used directly, it is queue private. always use
372 * ->queue_lock.
373 */
374 spinlock_t __queue_lock;
375 spinlock_t *queue_lock;
376
377 /*
378 * queue kobject
379 */
380 struct kobject kobj;
381
382 /*
383 * queue settings
384 */
385 unsigned long nr_requests; /* Max # of requests */
386 unsigned int nr_congestion_on;
387 unsigned int nr_congestion_off;
388 unsigned int nr_batching;
389
390 unsigned int max_sectors;
391 unsigned int max_hw_sectors;
392 unsigned short max_phys_segments;
393 unsigned short max_hw_segments;
394 unsigned short hardsect_size;
395 unsigned int max_segment_size;
396
397 unsigned long seg_boundary_mask;
398 void *dma_drain_buffer;
399 unsigned int dma_drain_size;
400 unsigned int dma_pad_mask;
401 unsigned int dma_alignment;
402
403 struct blk_queue_tag *queue_tags;
404 struct list_head tag_busy_list;
405
406 unsigned int nr_sorted;
407 unsigned int in_flight;
408
409 unsigned int rq_timeout;
410 struct timer_list timeout;
411 struct list_head timeout_list;
412
413 /*
414 * sg stuff
415 */
416 unsigned int sg_timeout;
417 unsigned int sg_reserved_size;
418 int node;
419 #ifdef CONFIG_BLK_DEV_IO_TRACE
420 struct blk_trace *blk_trace;
421 #endif
422 /*
423 * reserved for flush operations
424 */
425 unsigned int ordered, next_ordered, ordseq;
426 int orderr, ordcolor;
427 struct request pre_flush_rq, bar_rq, post_flush_rq;
428 struct request *orig_bar_rq;
429
430 struct mutex sysfs_lock;
431
432 #if defined(CONFIG_BLK_DEV_BSG)
433 struct bsg_class_device bsg_dev;
434 #endif
435 struct blk_cmd_filter cmd_filter;
436 };
437
438 #define QUEUE_FLAG_CLUSTER 0 /* cluster several segments into 1 */
439 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
440 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
441 #define QUEUE_FLAG_READFULL 3 /* read queue has been filled */
442 #define QUEUE_FLAG_WRITEFULL 4 /* write queue has been filled */
443 #define QUEUE_FLAG_DEAD 5 /* queue being torn down */
444 #define QUEUE_FLAG_REENTER 6 /* Re-entrancy avoidance */
445 #define QUEUE_FLAG_PLUGGED 7 /* queue is plugged */
446 #define QUEUE_FLAG_ELVSWITCH 8 /* don't use elevator, just do FIFO */
447 #define QUEUE_FLAG_BIDI 9 /* queue supports bidi requests */
448 #define QUEUE_FLAG_NOMERGES 10 /* disable merge attempts */
449 #define QUEUE_FLAG_SAME_COMP 11 /* force complete on same CPU */
450 #define QUEUE_FLAG_FAIL_IO 12 /* fake timeout */
451 #define QUEUE_FLAG_STACKABLE 13 /* supports request stacking */
452 #define QUEUE_FLAG_NONROT 14 /* non-rotational device (SSD) */
453 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
454 #define QUEUE_FLAG_IO_STAT 15 /* do IO stats */
455
456 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
457 (1 << QUEUE_FLAG_CLUSTER) | \
458 (1 << QUEUE_FLAG_STACKABLE))
459
queue_is_locked(struct request_queue * q)460 static inline int queue_is_locked(struct request_queue *q)
461 {
462 #ifdef CONFIG_SMP
463 spinlock_t *lock = q->queue_lock;
464 return lock && spin_is_locked(lock);
465 #else
466 return 1;
467 #endif
468 }
469
queue_flag_set_unlocked(unsigned int flag,struct request_queue * q)470 static inline void queue_flag_set_unlocked(unsigned int flag,
471 struct request_queue *q)
472 {
473 __set_bit(flag, &q->queue_flags);
474 }
475
queue_flag_test_and_clear(unsigned int flag,struct request_queue * q)476 static inline int queue_flag_test_and_clear(unsigned int flag,
477 struct request_queue *q)
478 {
479 WARN_ON_ONCE(!queue_is_locked(q));
480
481 if (test_bit(flag, &q->queue_flags)) {
482 __clear_bit(flag, &q->queue_flags);
483 return 1;
484 }
485
486 return 0;
487 }
488
queue_flag_test_and_set(unsigned int flag,struct request_queue * q)489 static inline int queue_flag_test_and_set(unsigned int flag,
490 struct request_queue *q)
491 {
492 WARN_ON_ONCE(!queue_is_locked(q));
493
494 if (!test_bit(flag, &q->queue_flags)) {
495 __set_bit(flag, &q->queue_flags);
496 return 0;
497 }
498
499 return 1;
500 }
501
queue_flag_set(unsigned int flag,struct request_queue * q)502 static inline void queue_flag_set(unsigned int flag, struct request_queue *q)
503 {
504 WARN_ON_ONCE(!queue_is_locked(q));
505 __set_bit(flag, &q->queue_flags);
506 }
507
queue_flag_clear_unlocked(unsigned int flag,struct request_queue * q)508 static inline void queue_flag_clear_unlocked(unsigned int flag,
509 struct request_queue *q)
510 {
511 __clear_bit(flag, &q->queue_flags);
512 }
513
queue_flag_clear(unsigned int flag,struct request_queue * q)514 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
515 {
516 WARN_ON_ONCE(!queue_is_locked(q));
517 __clear_bit(flag, &q->queue_flags);
518 }
519
520 enum {
521 /*
522 * Hardbarrier is supported with one of the following methods.
523 *
524 * NONE : hardbarrier unsupported
525 * DRAIN : ordering by draining is enough
526 * DRAIN_FLUSH : ordering by draining w/ pre and post flushes
527 * DRAIN_FUA : ordering by draining w/ pre flush and FUA write
528 * TAG : ordering by tag is enough
529 * TAG_FLUSH : ordering by tag w/ pre and post flushes
530 * TAG_FUA : ordering by tag w/ pre flush and FUA write
531 */
532 QUEUE_ORDERED_BY_DRAIN = 0x01,
533 QUEUE_ORDERED_BY_TAG = 0x02,
534 QUEUE_ORDERED_DO_PREFLUSH = 0x10,
535 QUEUE_ORDERED_DO_BAR = 0x20,
536 QUEUE_ORDERED_DO_POSTFLUSH = 0x40,
537 QUEUE_ORDERED_DO_FUA = 0x80,
538
539 QUEUE_ORDERED_NONE = 0x00,
540
541 QUEUE_ORDERED_DRAIN = QUEUE_ORDERED_BY_DRAIN |
542 QUEUE_ORDERED_DO_BAR,
543 QUEUE_ORDERED_DRAIN_FLUSH = QUEUE_ORDERED_DRAIN |
544 QUEUE_ORDERED_DO_PREFLUSH |
545 QUEUE_ORDERED_DO_POSTFLUSH,
546 QUEUE_ORDERED_DRAIN_FUA = QUEUE_ORDERED_DRAIN |
547 QUEUE_ORDERED_DO_PREFLUSH |
548 QUEUE_ORDERED_DO_FUA,
549
550 QUEUE_ORDERED_TAG = QUEUE_ORDERED_BY_TAG |
551 QUEUE_ORDERED_DO_BAR,
552 QUEUE_ORDERED_TAG_FLUSH = QUEUE_ORDERED_TAG |
553 QUEUE_ORDERED_DO_PREFLUSH |
554 QUEUE_ORDERED_DO_POSTFLUSH,
555 QUEUE_ORDERED_TAG_FUA = QUEUE_ORDERED_TAG |
556 QUEUE_ORDERED_DO_PREFLUSH |
557 QUEUE_ORDERED_DO_FUA,
558
559 /*
560 * Ordered operation sequence
561 */
562 QUEUE_ORDSEQ_STARTED = 0x01, /* flushing in progress */
563 QUEUE_ORDSEQ_DRAIN = 0x02, /* waiting for the queue to be drained */
564 QUEUE_ORDSEQ_PREFLUSH = 0x04, /* pre-flushing in progress */
565 QUEUE_ORDSEQ_BAR = 0x08, /* original barrier req in progress */
566 QUEUE_ORDSEQ_POSTFLUSH = 0x10, /* post-flushing in progress */
567 QUEUE_ORDSEQ_DONE = 0x20,
568 };
569
570 #define blk_queue_plugged(q) test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
571 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
572 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
573 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
574 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
575 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
576 #define blk_queue_flushing(q) ((q)->ordseq)
577 #define blk_queue_stackable(q) \
578 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags)
579
580 #define blk_fs_request(rq) ((rq)->cmd_type == REQ_TYPE_FS)
581 #define blk_pc_request(rq) ((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
582 #define blk_special_request(rq) ((rq)->cmd_type == REQ_TYPE_SPECIAL)
583 #define blk_sense_request(rq) ((rq)->cmd_type == REQ_TYPE_SENSE)
584
585 #define blk_failfast_dev(rq) ((rq)->cmd_flags & REQ_FAILFAST_DEV)
586 #define blk_failfast_transport(rq) ((rq)->cmd_flags & REQ_FAILFAST_TRANSPORT)
587 #define blk_failfast_driver(rq) ((rq)->cmd_flags & REQ_FAILFAST_DRIVER)
588 #define blk_noretry_request(rq) (blk_failfast_dev(rq) || \
589 blk_failfast_transport(rq) || \
590 blk_failfast_driver(rq))
591 #define blk_rq_started(rq) ((rq)->cmd_flags & REQ_STARTED)
592
593 #define blk_account_rq(rq) (blk_rq_started(rq) && (blk_fs_request(rq) || blk_discard_rq(rq)))
594
595 #define blk_pm_suspend_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
596 #define blk_pm_resume_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_RESUME)
597 #define blk_pm_request(rq) \
598 (blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
599
600 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1)
601 #define blk_sorted_rq(rq) ((rq)->cmd_flags & REQ_SORTED)
602 #define blk_barrier_rq(rq) ((rq)->cmd_flags & REQ_HARDBARRIER)
603 #define blk_fua_rq(rq) ((rq)->cmd_flags & REQ_FUA)
604 #define blk_discard_rq(rq) ((rq)->cmd_flags & REQ_DISCARD)
605 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL)
606 /* rq->queuelist of dequeued request must be list_empty() */
607 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist))
608
609 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
610
611 #define rq_data_dir(rq) ((rq)->cmd_flags & 1)
612
613 /*
614 * We regard a request as sync, if it's a READ or a SYNC write.
615 */
616 #define rq_is_sync(rq) (rq_data_dir((rq)) == READ || (rq)->cmd_flags & REQ_RW_SYNC)
617 #define rq_is_meta(rq) ((rq)->cmd_flags & REQ_RW_META)
618
blk_queue_full(struct request_queue * q,int rw)619 static inline int blk_queue_full(struct request_queue *q, int rw)
620 {
621 if (rw == READ)
622 return test_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
623 return test_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
624 }
625
blk_set_queue_full(struct request_queue * q,int rw)626 static inline void blk_set_queue_full(struct request_queue *q, int rw)
627 {
628 if (rw == READ)
629 queue_flag_set(QUEUE_FLAG_READFULL, q);
630 else
631 queue_flag_set(QUEUE_FLAG_WRITEFULL, q);
632 }
633
blk_clear_queue_full(struct request_queue * q,int rw)634 static inline void blk_clear_queue_full(struct request_queue *q, int rw)
635 {
636 if (rw == READ)
637 queue_flag_clear(QUEUE_FLAG_READFULL, q);
638 else
639 queue_flag_clear(QUEUE_FLAG_WRITEFULL, q);
640 }
641
642
643 /*
644 * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
645 * it already be started by driver.
646 */
647 #define RQ_NOMERGE_FLAGS \
648 (REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
649 #define rq_mergeable(rq) \
650 (!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && \
651 (blk_discard_rq(rq) || blk_fs_request((rq))))
652
653 /*
654 * q->prep_rq_fn return values
655 */
656 #define BLKPREP_OK 0 /* serve it */
657 #define BLKPREP_KILL 1 /* fatal error, kill */
658 #define BLKPREP_DEFER 2 /* leave on queue */
659
660 extern unsigned long blk_max_low_pfn, blk_max_pfn;
661
662 /*
663 * standard bounce addresses:
664 *
665 * BLK_BOUNCE_HIGH : bounce all highmem pages
666 * BLK_BOUNCE_ANY : don't bounce anything
667 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
668 */
669
670 #if BITS_PER_LONG == 32
671 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
672 #else
673 #define BLK_BOUNCE_HIGH -1ULL
674 #endif
675 #define BLK_BOUNCE_ANY (-1ULL)
676 #define BLK_BOUNCE_ISA (ISA_DMA_THRESHOLD)
677
678 /*
679 * default timeout for SG_IO if none specified
680 */
681 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
682 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
683
684 #ifdef CONFIG_BOUNCE
685 extern int init_emergency_isa_pool(void);
686 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio);
687 #else
init_emergency_isa_pool(void)688 static inline int init_emergency_isa_pool(void)
689 {
690 return 0;
691 }
blk_queue_bounce(struct request_queue * q,struct bio ** bio)692 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio)
693 {
694 }
695 #endif /* CONFIG_MMU */
696
697 struct rq_map_data {
698 struct page **pages;
699 int page_order;
700 int nr_entries;
701 unsigned long offset;
702 int null_mapped;
703 };
704
705 struct req_iterator {
706 int i;
707 struct bio *bio;
708 };
709
710 /* This should not be used directly - use rq_for_each_segment */
711 #define for_each_bio(_bio) \
712 for (; _bio; _bio = _bio->bi_next)
713 #define __rq_for_each_bio(_bio, rq) \
714 if ((rq->bio)) \
715 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
716
717 #define rq_for_each_segment(bvl, _rq, _iter) \
718 __rq_for_each_bio(_iter.bio, _rq) \
719 bio_for_each_segment(bvl, _iter.bio, _iter.i)
720
721 #define rq_iter_last(rq, _iter) \
722 (_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
723
724 extern int blk_register_queue(struct gendisk *disk);
725 extern void blk_unregister_queue(struct gendisk *disk);
726 extern void register_disk(struct gendisk *dev);
727 extern void generic_make_request(struct bio *bio);
728 extern void blk_rq_init(struct request_queue *q, struct request *rq);
729 extern void blk_put_request(struct request *);
730 extern void __blk_put_request(struct request_queue *, struct request *);
731 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
732 extern void blk_insert_request(struct request_queue *, struct request *, int, void *);
733 extern void blk_requeue_request(struct request_queue *, struct request *);
734 extern int blk_rq_check_limits(struct request_queue *q, struct request *rq);
735 extern int blk_lld_busy(struct request_queue *q);
736 extern int blk_insert_cloned_request(struct request_queue *q,
737 struct request *rq);
738 extern void blk_plug_device(struct request_queue *);
739 extern void blk_plug_device_unlocked(struct request_queue *);
740 extern int blk_remove_plug(struct request_queue *);
741 extern void blk_recount_segments(struct request_queue *, struct bio *);
742 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t,
743 unsigned int, void __user *);
744 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t,
745 struct scsi_ioctl_command __user *);
746
747 /*
748 * Temporary export, until SCSI gets fixed up.
749 */
750 extern int blk_rq_append_bio(struct request_queue *q, struct request *rq,
751 struct bio *bio);
752
753 /*
754 * A queue has just exitted congestion. Note this in the global counter of
755 * congested queues, and wake up anyone who was waiting for requests to be
756 * put back.
757 */
blk_clear_queue_congested(struct request_queue * q,int rw)758 static inline void blk_clear_queue_congested(struct request_queue *q, int rw)
759 {
760 clear_bdi_congested(&q->backing_dev_info, rw);
761 }
762
763 /*
764 * A queue has just entered congestion. Flag that in the queue's VM-visible
765 * state flags and increment the global gounter of congested queues.
766 */
blk_set_queue_congested(struct request_queue * q,int rw)767 static inline void blk_set_queue_congested(struct request_queue *q, int rw)
768 {
769 set_bdi_congested(&q->backing_dev_info, rw);
770 }
771
772 extern void blk_start_queue(struct request_queue *q);
773 extern void blk_stop_queue(struct request_queue *q);
774 extern void blk_sync_queue(struct request_queue *q);
775 extern void __blk_stop_queue(struct request_queue *q);
776 extern void __blk_run_queue(struct request_queue *);
777 extern void blk_run_queue(struct request_queue *);
778 extern void blk_start_queueing(struct request_queue *);
779 extern int blk_rq_map_user(struct request_queue *, struct request *,
780 struct rq_map_data *, void __user *, unsigned long,
781 gfp_t);
782 extern int blk_rq_unmap_user(struct bio *);
783 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t);
784 extern int blk_rq_map_user_iov(struct request_queue *, struct request *,
785 struct rq_map_data *, struct sg_iovec *, int,
786 unsigned int, gfp_t);
787 extern int blk_execute_rq(struct request_queue *, struct gendisk *,
788 struct request *, int);
789 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *,
790 struct request *, int, rq_end_io_fn *);
791 extern void blk_unplug(struct request_queue *q);
792
bdev_get_queue(struct block_device * bdev)793 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
794 {
795 return bdev->bd_disk->queue;
796 }
797
blk_run_backing_dev(struct backing_dev_info * bdi,struct page * page)798 static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
799 struct page *page)
800 {
801 if (bdi && bdi->unplug_io_fn)
802 bdi->unplug_io_fn(bdi, page);
803 }
804
blk_run_address_space(struct address_space * mapping)805 static inline void blk_run_address_space(struct address_space *mapping)
806 {
807 if (mapping)
808 blk_run_backing_dev(mapping->backing_dev_info, NULL);
809 }
810
811 extern void blkdev_dequeue_request(struct request *req);
812
813 /*
814 * blk_end_request() and friends.
815 * __blk_end_request() and end_request() must be called with
816 * the request queue spinlock acquired.
817 *
818 * Several drivers define their own end_request and call
819 * blk_end_request() for parts of the original function.
820 * This prevents code duplication in drivers.
821 */
822 extern int blk_end_request(struct request *rq, int error,
823 unsigned int nr_bytes);
824 extern int __blk_end_request(struct request *rq, int error,
825 unsigned int nr_bytes);
826 extern int blk_end_bidi_request(struct request *rq, int error,
827 unsigned int nr_bytes, unsigned int bidi_bytes);
828 extern void end_request(struct request *, int);
829 extern int blk_end_request_callback(struct request *rq, int error,
830 unsigned int nr_bytes,
831 int (drv_callback)(struct request *));
832 extern void blk_complete_request(struct request *);
833 extern void __blk_complete_request(struct request *);
834 extern void blk_abort_request(struct request *);
835 extern void blk_abort_queue(struct request_queue *);
836 extern void blk_update_request(struct request *rq, int error,
837 unsigned int nr_bytes);
838
839 /*
840 * blk_end_request() takes bytes instead of sectors as a complete size.
841 * blk_rq_bytes() returns bytes left to complete in the entire request.
842 * blk_rq_cur_bytes() returns bytes left to complete in the current segment.
843 */
844 extern unsigned int blk_rq_bytes(struct request *rq);
845 extern unsigned int blk_rq_cur_bytes(struct request *rq);
846
847 /*
848 * Access functions for manipulating queue properties
849 */
850 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
851 spinlock_t *lock, int node_id);
852 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
853 extern void blk_cleanup_queue(struct request_queue *);
854 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
855 extern void blk_queue_bounce_limit(struct request_queue *, u64);
856 extern void blk_queue_max_sectors(struct request_queue *, unsigned int);
857 extern void blk_queue_max_phys_segments(struct request_queue *, unsigned short);
858 extern void blk_queue_max_hw_segments(struct request_queue *, unsigned short);
859 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
860 extern void blk_queue_hardsect_size(struct request_queue *, unsigned short);
861 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b);
862 extern void blk_queue_dma_pad(struct request_queue *, unsigned int);
863 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
864 extern int blk_queue_dma_drain(struct request_queue *q,
865 dma_drain_needed_fn *dma_drain_needed,
866 void *buf, unsigned int size);
867 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn);
868 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
869 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn);
870 extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *);
871 extern void blk_queue_dma_alignment(struct request_queue *, int);
872 extern void blk_queue_update_dma_alignment(struct request_queue *, int);
873 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *);
874 extern void blk_queue_set_discard(struct request_queue *, prepare_discard_fn *);
875 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *);
876 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
877 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
878 extern int blk_queue_ordered(struct request_queue *, unsigned, prepare_flush_fn *);
879 extern bool blk_do_ordered(struct request_queue *, struct request **);
880 extern unsigned blk_ordered_cur_seq(struct request_queue *);
881 extern unsigned blk_ordered_req_seq(struct request *);
882 extern bool blk_ordered_complete_seq(struct request_queue *, unsigned, int);
883
884 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *);
885 extern void blk_dump_rq_flags(struct request *, char *);
886 extern void generic_unplug_device(struct request_queue *);
887 extern long nr_blockdev_pages(void);
888
889 int blk_get_queue(struct request_queue *);
890 struct request_queue *blk_alloc_queue(gfp_t);
891 struct request_queue *blk_alloc_queue_node(gfp_t, int);
892 extern void blk_put_queue(struct request_queue *);
893
894 /*
895 * tag stuff
896 */
897 #define blk_rq_tagged(rq) ((rq)->cmd_flags & REQ_QUEUED)
898 extern int blk_queue_start_tag(struct request_queue *, struct request *);
899 extern struct request *blk_queue_find_tag(struct request_queue *, int);
900 extern void blk_queue_end_tag(struct request_queue *, struct request *);
901 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *);
902 extern void blk_queue_free_tags(struct request_queue *);
903 extern int blk_queue_resize_tags(struct request_queue *, int);
904 extern void blk_queue_invalidate_tags(struct request_queue *);
905 extern struct blk_queue_tag *blk_init_tags(int);
906 extern void blk_free_tags(struct blk_queue_tag *);
907
blk_map_queue_find_tag(struct blk_queue_tag * bqt,int tag)908 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
909 int tag)
910 {
911 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
912 return NULL;
913 return bqt->tag_index[tag];
914 }
915
916 extern int blkdev_issue_flush(struct block_device *, sector_t *);
917 extern int blkdev_issue_discard(struct block_device *,
918 sector_t sector, sector_t nr_sects, gfp_t);
919
sb_issue_discard(struct super_block * sb,sector_t block,sector_t nr_blocks)920 static inline int sb_issue_discard(struct super_block *sb,
921 sector_t block, sector_t nr_blocks)
922 {
923 block <<= (sb->s_blocksize_bits - 9);
924 nr_blocks <<= (sb->s_blocksize_bits - 9);
925 return blkdev_issue_discard(sb->s_bdev, block, nr_blocks, GFP_KERNEL);
926 }
927
928 /*
929 * command filter functions
930 */
931 extern int blk_verify_command(struct blk_cmd_filter *filter,
932 unsigned char *cmd, fmode_t has_write_perm);
933 extern void blk_unregister_filter(struct gendisk *disk);
934 extern void blk_set_cmd_filter_defaults(struct blk_cmd_filter *filter);
935
936 #define MAX_PHYS_SEGMENTS 128
937 #define MAX_HW_SEGMENTS 128
938 #define SAFE_MAX_SECTORS 255
939 #define BLK_DEF_MAX_SECTORS 1024
940
941 #define MAX_SEGMENT_SIZE 65536
942
943 #define BLK_SEG_BOUNDARY_MASK 0xFFFFFFFFUL
944
945 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
946
queue_hardsect_size(struct request_queue * q)947 static inline int queue_hardsect_size(struct request_queue *q)
948 {
949 int retval = 512;
950
951 if (q && q->hardsect_size)
952 retval = q->hardsect_size;
953
954 return retval;
955 }
956
bdev_hardsect_size(struct block_device * bdev)957 static inline int bdev_hardsect_size(struct block_device *bdev)
958 {
959 return queue_hardsect_size(bdev_get_queue(bdev));
960 }
961
queue_dma_alignment(struct request_queue * q)962 static inline int queue_dma_alignment(struct request_queue *q)
963 {
964 return q ? q->dma_alignment : 511;
965 }
966
blk_rq_aligned(struct request_queue * q,void * addr,unsigned int len)967 static inline int blk_rq_aligned(struct request_queue *q, void *addr,
968 unsigned int len)
969 {
970 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
971 return !((unsigned long)addr & alignment) && !(len & alignment);
972 }
973
974 /* assumes size > 256 */
blksize_bits(unsigned int size)975 static inline unsigned int blksize_bits(unsigned int size)
976 {
977 unsigned int bits = 8;
978 do {
979 bits++;
980 size >>= 1;
981 } while (size > 256);
982 return bits;
983 }
984
block_size(struct block_device * bdev)985 static inline unsigned int block_size(struct block_device *bdev)
986 {
987 return bdev->bd_block_size;
988 }
989
990 typedef struct {struct page *v;} Sector;
991
992 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
993
put_dev_sector(Sector p)994 static inline void put_dev_sector(Sector p)
995 {
996 page_cache_release(p.v);
997 }
998
999 struct work_struct;
1000 int kblockd_schedule_work(struct request_queue *q, struct work_struct *work);
1001
1002 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1003 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1004 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1005 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1006
1007 #if defined(CONFIG_BLK_DEV_INTEGRITY)
1008
1009 #define INTEGRITY_FLAG_READ 2 /* verify data integrity on read */
1010 #define INTEGRITY_FLAG_WRITE 4 /* generate data integrity on write */
1011
1012 struct blk_integrity_exchg {
1013 void *prot_buf;
1014 void *data_buf;
1015 sector_t sector;
1016 unsigned int data_size;
1017 unsigned short sector_size;
1018 const char *disk_name;
1019 };
1020
1021 typedef void (integrity_gen_fn) (struct blk_integrity_exchg *);
1022 typedef int (integrity_vrfy_fn) (struct blk_integrity_exchg *);
1023 typedef void (integrity_set_tag_fn) (void *, void *, unsigned int);
1024 typedef void (integrity_get_tag_fn) (void *, void *, unsigned int);
1025
1026 struct blk_integrity {
1027 integrity_gen_fn *generate_fn;
1028 integrity_vrfy_fn *verify_fn;
1029 integrity_set_tag_fn *set_tag_fn;
1030 integrity_get_tag_fn *get_tag_fn;
1031
1032 unsigned short flags;
1033 unsigned short tuple_size;
1034 unsigned short sector_size;
1035 unsigned short tag_size;
1036
1037 const char *name;
1038
1039 struct kobject kobj;
1040 };
1041
1042 extern int blk_integrity_register(struct gendisk *, struct blk_integrity *);
1043 extern void blk_integrity_unregister(struct gendisk *);
1044 extern int blk_integrity_compare(struct gendisk *, struct gendisk *);
1045 extern int blk_rq_map_integrity_sg(struct request *, struct scatterlist *);
1046 extern int blk_rq_count_integrity_sg(struct request *);
1047
1048 static inline
bdev_get_integrity(struct block_device * bdev)1049 struct blk_integrity *bdev_get_integrity(struct block_device *bdev)
1050 {
1051 return bdev->bd_disk->integrity;
1052 }
1053
blk_get_integrity(struct gendisk * disk)1054 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk)
1055 {
1056 return disk->integrity;
1057 }
1058
blk_integrity_rq(struct request * rq)1059 static inline int blk_integrity_rq(struct request *rq)
1060 {
1061 if (rq->bio == NULL)
1062 return 0;
1063
1064 return bio_integrity(rq->bio);
1065 }
1066
1067 #else /* CONFIG_BLK_DEV_INTEGRITY */
1068
1069 #define blk_integrity_rq(rq) (0)
1070 #define blk_rq_count_integrity_sg(a) (0)
1071 #define blk_rq_map_integrity_sg(a, b) (0)
1072 #define bdev_get_integrity(a) (0)
1073 #define blk_get_integrity(a) (0)
1074 #define blk_integrity_compare(a, b) (0)
1075 #define blk_integrity_register(a, b) (0)
1076 #define blk_integrity_unregister(a) do { } while (0);
1077
1078 #endif /* CONFIG_BLK_DEV_INTEGRITY */
1079
1080 struct block_device_operations {
1081 int (*open) (struct block_device *, fmode_t);
1082 int (*release) (struct gendisk *, fmode_t);
1083 int (*locked_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1084 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1085 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
1086 int (*direct_access) (struct block_device *, sector_t,
1087 void **, unsigned long *);
1088 int (*media_changed) (struct gendisk *);
1089 int (*revalidate_disk) (struct gendisk *);
1090 int (*getgeo)(struct block_device *, struct hd_geometry *);
1091 struct module *owner;
1092 };
1093
1094 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int,
1095 unsigned long);
1096 #else /* CONFIG_BLOCK */
1097 /*
1098 * stubs for when the block layer is configured out
1099 */
1100 #define buffer_heads_over_limit 0
1101
nr_blockdev_pages(void)1102 static inline long nr_blockdev_pages(void)
1103 {
1104 return 0;
1105 }
1106
1107 #endif /* CONFIG_BLOCK */
1108
1109 #endif
1110