blob: d370d2cfe13803ed79626e90194e270cc0c6c169 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _LINUX_BLKDEV_H
2#define _LINUX_BLKDEV_H
3
Andrew Mortonbcfd8d32006-08-31 12:56:06 +02004#include <linux/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07005#include <linux/major.h>
6#include <linux/genhd.h>
7#include <linux/list.h>
8#include <linux/timer.h>
9#include <linux/workqueue.h>
10#include <linux/pagemap.h>
11#include <linux/backing-dev.h>
12#include <linux/wait.h>
13#include <linux/mempool.h>
14#include <linux/bio.h>
15#include <linux/module.h>
16#include <linux/stringify.h>
17
18#include <asm/scatterlist.h>
19
David Howells93614012006-09-30 20:45:40 +020020#ifdef CONFIG_LBD
21# include <asm/div64.h>
22# define sector_div(a, b) do_div(a, b)
23#else
24# define sector_div(n, b)( \
25{ \
26 int _res; \
27 _res = (n) % (b); \
28 (n) /= (b); \
29 _res; \
30} \
31)
32#endif
33
34#ifdef CONFIG_BLOCK
35
Christoph Hellwig21b2f0c2006-03-22 17:52:04 +010036struct scsi_ioctl_command;
37
Linus Torvalds1da177e2005-04-16 15:20:36 -070038struct request_queue;
39typedef struct request_queue request_queue_t;
40struct elevator_queue;
41typedef struct elevator_queue elevator_t;
42struct request_pm_state;
Jens Axboe2056a782006-03-23 20:00:26 +010043struct blk_trace;
Linus Torvalds1da177e2005-04-16 15:20:36 -070044
45#define BLKDEV_MIN_RQ 4
46#define BLKDEV_MAX_RQ 128 /* Default maximum */
47
48/*
49 * This is the per-process anticipatory I/O scheduler state.
50 */
51struct as_io_context {
52 spinlock_t lock;
53
54 void (*dtor)(struct as_io_context *aic); /* destructor */
55 void (*exit)(struct as_io_context *aic); /* called on task exit */
56
57 unsigned long state;
58 atomic_t nr_queued; /* queued reads & sync writes */
59 atomic_t nr_dispatched; /* number of requests gone to the drivers */
60
61 /* IO History tracking */
62 /* Thinktime */
63 unsigned long last_end_request;
64 unsigned long ttime_total;
65 unsigned long ttime_samples;
66 unsigned long ttime_mean;
67 /* Layout pattern */
68 unsigned int seek_samples;
69 sector_t last_request_pos;
70 u64 seek_total;
71 sector_t seek_mean;
72};
73
74struct cfq_queue;
75struct cfq_io_context {
Jens Axboee2d74ac2006-03-28 08:59:01 +020076 struct rb_node rb_node;
Jens Axboe22e2c502005-06-27 10:55:12 +020077 void *key;
78
Jens Axboee2d74ac2006-03-28 08:59:01 +020079 struct cfq_queue *cfqq[2];
80
Jens Axboe22e2c502005-06-27 10:55:12 +020081 struct io_context *ioc;
82
83 unsigned long last_end_request;
Jens Axboe206dc692006-03-28 13:03:44 +020084 sector_t last_request_pos;
85 unsigned long last_queue;
86
Jens Axboe22e2c502005-06-27 10:55:12 +020087 unsigned long ttime_total;
88 unsigned long ttime_samples;
89 unsigned long ttime_mean;
90
Jens Axboe206dc692006-03-28 13:03:44 +020091 unsigned int seek_samples;
92 u64 seek_total;
93 sector_t seek_mean;
94
Al Virod9ff4182006-03-18 13:51:22 -050095 struct list_head queue_list;
96
Jens Axboee2d74ac2006-03-28 08:59:01 +020097 void (*dtor)(struct io_context *); /* destructor */
98 void (*exit)(struct io_context *); /* called on task exit */
Linus Torvalds1da177e2005-04-16 15:20:36 -070099};
100
101/*
102 * This is the per-process I/O subsystem state. It is refcounted and
103 * kmalloc'ed. Currently all fields are modified in process io context
104 * (apart from the atomic refcount), so require no locking.
105 */
106struct io_context {
107 atomic_t refcount;
Jens Axboe22e2c502005-06-27 10:55:12 +0200108 struct task_struct *task;
109
Jens Axboefc463792006-08-29 09:05:44 +0200110 unsigned int ioprio_changed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111
112 /*
113 * For request batching
114 */
115 unsigned long last_waited; /* Time last woken after wait for request */
116 int nr_batch_requests; /* Number of requests left in the batch */
117
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118 struct as_io_context *aic;
Jens Axboee2d74ac2006-03-28 08:59:01 +0200119 struct rb_root cic_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120};
121
122void put_io_context(struct io_context *ioc);
123void exit_io_context(void);
Jens Axboeb5deef92006-07-19 23:39:40 +0200124struct io_context *get_io_context(gfp_t gfp_flags, int node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125void copy_io_context(struct io_context **pdst, struct io_context **psrc);
126void swap_io_context(struct io_context **ioc1, struct io_context **ioc2);
127
128struct request;
Tejun Heo8ffdc652006-01-06 09:49:03 +0100129typedef void (rq_end_io_fn)(struct request *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130
131struct request_list {
132 int count[2];
133 int starved[2];
Tejun Heocb98fc82005-10-28 08:29:39 +0200134 int elvpriv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 mempool_t *rq_pool;
136 wait_queue_head_t wait[2];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137};
138
Jens Axboe4aff5e22006-08-10 08:44:47 +0200139/*
140 * request command types
141 */
142enum rq_cmd_type_bits {
143 REQ_TYPE_FS = 1, /* fs request */
144 REQ_TYPE_BLOCK_PC, /* scsi command */
145 REQ_TYPE_SENSE, /* sense request */
146 REQ_TYPE_PM_SUSPEND, /* suspend request */
147 REQ_TYPE_PM_RESUME, /* resume request */
148 REQ_TYPE_PM_SHUTDOWN, /* shutdown request */
149 REQ_TYPE_FLUSH, /* flush request */
150 REQ_TYPE_SPECIAL, /* driver defined type */
151 REQ_TYPE_LINUX_BLOCK, /* generic block layer message */
152 /*
153 * for ATA/ATAPI devices. this really doesn't belong here, ide should
154 * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver
155 * private REQ_LB opcodes to differentiate what type of request this is
156 */
157 REQ_TYPE_ATA_CMD,
158 REQ_TYPE_ATA_TASK,
159 REQ_TYPE_ATA_TASKFILE,
Jens Axboecea28852006-10-12 15:08:45 +0200160 REQ_TYPE_ATA_PC,
Jens Axboe4aff5e22006-08-10 08:44:47 +0200161};
162
163/*
164 * For request of type REQ_TYPE_LINUX_BLOCK, rq->cmd[0] is the opcode being
165 * sent down (similar to how REQ_TYPE_BLOCK_PC means that ->cmd[] holds a
166 * SCSI cdb.
167 *
168 * 0x00 -> 0x3f are driver private, to be used for whatever purpose they need,
169 * typically to differentiate REQ_TYPE_SPECIAL requests.
170 *
171 */
172enum {
173 /*
174 * just examples for now
175 */
176 REQ_LB_OP_EJECT = 0x40, /* eject request */
177 REQ_LB_OP_FLUSH = 0x41, /* flush device */
178};
179
180/*
181 * request type modified bits. first three bits match BIO_RW* bits, important
182 */
183enum rq_flag_bits {
184 __REQ_RW, /* not set, read. set, write */
185 __REQ_FAILFAST, /* no low level driver retries */
186 __REQ_SORTED, /* elevator knows about this request */
187 __REQ_SOFTBARRIER, /* may not be passed by ioscheduler */
188 __REQ_HARDBARRIER, /* may not be passed by drive either */
189 __REQ_FUA, /* forced unit access */
190 __REQ_NOMERGE, /* don't touch this for merging */
191 __REQ_STARTED, /* drive already may have started this one */
192 __REQ_DONTPREP, /* don't call prep for this one */
193 __REQ_QUEUED, /* uses queueing */
194 __REQ_ELVPRIV, /* elevator private data attached */
195 __REQ_FAILED, /* set if the request failed */
196 __REQ_QUIET, /* don't worry about errors */
197 __REQ_PREEMPT, /* set for "ide_preempt" requests */
198 __REQ_ORDERED_COLOR, /* is before or after barrier */
199 __REQ_RW_SYNC, /* request is sync (O_DIRECT) */
Jens Axboe49171e52006-08-10 08:59:11 +0200200 __REQ_ALLOCED, /* request came from our alloc pool */
Jens Axboe5404bc72006-08-10 09:01:02 +0200201 __REQ_RW_META, /* metadata io request */
Jens Axboe4aff5e22006-08-10 08:44:47 +0200202 __REQ_NR_BITS, /* stops here */
203};
204
205#define REQ_RW (1 << __REQ_RW)
206#define REQ_FAILFAST (1 << __REQ_FAILFAST)
207#define REQ_SORTED (1 << __REQ_SORTED)
208#define REQ_SOFTBARRIER (1 << __REQ_SOFTBARRIER)
209#define REQ_HARDBARRIER (1 << __REQ_HARDBARRIER)
210#define REQ_FUA (1 << __REQ_FUA)
211#define REQ_NOMERGE (1 << __REQ_NOMERGE)
212#define REQ_STARTED (1 << __REQ_STARTED)
213#define REQ_DONTPREP (1 << __REQ_DONTPREP)
214#define REQ_QUEUED (1 << __REQ_QUEUED)
215#define REQ_ELVPRIV (1 << __REQ_ELVPRIV)
216#define REQ_FAILED (1 << __REQ_FAILED)
217#define REQ_QUIET (1 << __REQ_QUIET)
218#define REQ_PREEMPT (1 << __REQ_PREEMPT)
219#define REQ_ORDERED_COLOR (1 << __REQ_ORDERED_COLOR)
220#define REQ_RW_SYNC (1 << __REQ_RW_SYNC)
Jens Axboe49171e52006-08-10 08:59:11 +0200221#define REQ_ALLOCED (1 << __REQ_ALLOCED)
Jens Axboe5404bc72006-08-10 09:01:02 +0200222#define REQ_RW_META (1 << __REQ_RW_META)
Jens Axboe4aff5e22006-08-10 08:44:47 +0200223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224#define BLK_MAX_CDB 16
225
226/*
227 * try to put the fields that are referenced together in the same cacheline
228 */
229struct request {
Jens Axboeff856ba2006-01-09 16:02:34 +0100230 struct list_head queuelist;
231 struct list_head donelist;
232
Jens Axboee6a1c872006-08-10 09:00:21 +0200233 request_queue_t *q;
234
Jens Axboe4aff5e22006-08-10 08:44:47 +0200235 unsigned int cmd_flags;
236 enum rq_cmd_type_bits cmd_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237
238 /* Maintain bio traversal state for part by part I/O submission.
239 * hard_* are block layer internals, no driver should touch them!
240 */
241
242 sector_t sector; /* next sector to submit */
Jens Axboee6a1c872006-08-10 09:00:21 +0200243 sector_t hard_sector; /* next sector to complete */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 unsigned long nr_sectors; /* no. of sectors left to submit */
Jens Axboee6a1c872006-08-10 09:00:21 +0200245 unsigned long hard_nr_sectors; /* no. of sectors left to complete */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 /* no. of sectors left to submit in the current segment */
247 unsigned int current_nr_sectors;
248
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 /* no. of sectors left to complete in the current segment */
250 unsigned int hard_cur_sectors;
251
252 struct bio *bio;
253 struct bio *biotail;
254
Jens Axboe98170642006-07-28 09:23:08 +0200255 struct hlist_node hash; /* merge hash */
Jens Axboee6a1c872006-08-10 09:00:21 +0200256 /*
257 * The rb_node is only used inside the io scheduler, requests
258 * are pruned when moved to the dispatch queue. So let the
259 * completion_data share space with the rb_node.
260 */
261 union {
262 struct rb_node rb_node; /* sort/lookup */
263 void *completion_data;
264 };
Jens Axboe98170642006-07-28 09:23:08 +0200265
Jens Axboeff7d1452006-07-12 14:04:37 +0200266 /*
267 * two pointers are available for the IO schedulers, if they need
268 * more they have to dynamically allocate it.
269 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 void *elevator_private;
Jens Axboeff7d1452006-07-12 14:04:37 +0200271 void *elevator_private2;
272
Jens Axboe8f34ee72006-06-13 09:02:34 +0200273 struct gendisk *rq_disk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 unsigned long start_time;
275
276 /* Number of scatter-gather DMA addr+len pairs after
277 * physical address coalescing is performed.
278 */
279 unsigned short nr_phys_segments;
280
281 /* Number of scatter-gather addr+len pairs after
282 * physical and DMA remapping hardware coalescing is performed.
283 * This is the number of scatter-gather entries the driver
284 * will actually have to deal with after DMA mapping is done.
285 */
286 unsigned short nr_hw_segments;
287
Jens Axboe8f34ee72006-06-13 09:02:34 +0200288 unsigned short ioprio;
289
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 void *special;
Jens Axboe8f34ee72006-06-13 09:02:34 +0200291 char *buffer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292
Jens Axboecdd60262006-07-28 09:32:07 +0200293 int tag;
294 int errors;
295
296 int ref_count;
297
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 /*
299 * when request is used as a packet command carrier
300 */
301 unsigned int cmd_len;
302 unsigned char cmd[BLK_MAX_CDB];
303
304 unsigned int data_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 unsigned int sense_len;
Jens Axboe8f34ee72006-06-13 09:02:34 +0200306 void *data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 void *sense;
308
309 unsigned int timeout;
Mike Christie17e01f22005-11-11 05:31:37 -0600310 int retries;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311
312 /*
Jens Axboec00895a2006-09-30 20:29:12 +0200313 * completion callback.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 */
315 rq_end_io_fn *end_io;
316 void *end_io_data;
317};
318
319/*
Jens Axboe4aff5e22006-08-10 08:44:47 +0200320 * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 * requests. Some step values could eventually be made generic.
322 */
323struct request_pm_state
324{
325 /* PM state machine step value, currently driver specific */
326 int pm_step;
327 /* requested PM state value (S1, S2, S3, S4, ...) */
328 u32 pm_state;
329 void* data; /* for driver use */
330};
331
332#include <linux/elevator.h>
333
334typedef int (merge_request_fn) (request_queue_t *, struct request *,
335 struct bio *);
336typedef int (merge_requests_fn) (request_queue_t *, struct request *,
337 struct request *);
338typedef void (request_fn_proc) (request_queue_t *q);
339typedef int (make_request_fn) (request_queue_t *q, struct bio *bio);
340typedef int (prep_rq_fn) (request_queue_t *, struct request *);
341typedef void (unplug_fn) (request_queue_t *);
342
343struct bio_vec;
344typedef int (merge_bvec_fn) (request_queue_t *, struct bio *, struct bio_vec *);
345typedef void (activity_fn) (void *data, int rw);
346typedef int (issue_flush_fn) (request_queue_t *, struct gendisk *, sector_t *);
Tejun Heo797e7db2006-01-06 09:51:03 +0100347typedef void (prepare_flush_fn) (request_queue_t *, struct request *);
Jens Axboeff856ba2006-01-09 16:02:34 +0100348typedef void (softirq_done_fn)(struct request *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349
350enum blk_queue_state {
351 Queue_down,
352 Queue_up,
353};
354
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355struct blk_queue_tag {
356 struct request **tag_index; /* map of busy tags */
357 unsigned long *tag_map; /* bit map of free/busy tags */
358 struct list_head busy_list; /* fifo list of busy tags */
359 int busy; /* current depth */
360 int max_depth; /* what we will send to device */
Tejun Heoba025082005-08-05 13:28:11 -0700361 int real_max_depth; /* what the array can hold */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 atomic_t refcnt; /* map can be shared */
363};
364
365struct request_queue
366{
367 /*
368 * Together with queue_head for cacheline sharing
369 */
370 struct list_head queue_head;
371 struct request *last_merge;
372 elevator_t *elevator;
373
374 /*
375 * the queue request freelist, one for reads and one for writes
376 */
377 struct request_list rq;
378
379 request_fn_proc *request_fn;
380 merge_request_fn *back_merge_fn;
381 merge_request_fn *front_merge_fn;
382 merge_requests_fn *merge_requests_fn;
383 make_request_fn *make_request_fn;
384 prep_rq_fn *prep_rq_fn;
385 unplug_fn *unplug_fn;
386 merge_bvec_fn *merge_bvec_fn;
387 activity_fn *activity_fn;
388 issue_flush_fn *issue_flush_fn;
389 prepare_flush_fn *prepare_flush_fn;
Jens Axboeff856ba2006-01-09 16:02:34 +0100390 softirq_done_fn *softirq_done_fn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391
392 /*
Tejun Heo8922e162005-10-20 16:23:44 +0200393 * Dispatch queue sorting
394 */
Jens Axboe1b47f532005-10-20 16:37:00 +0200395 sector_t end_sector;
Tejun Heo8922e162005-10-20 16:23:44 +0200396 struct request *boundary_rq;
Tejun Heo8922e162005-10-20 16:23:44 +0200397
398 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 * Auto-unplugging state
400 */
401 struct timer_list unplug_timer;
402 int unplug_thresh; /* After this many requests */
403 unsigned long unplug_delay; /* After this many jiffies */
404 struct work_struct unplug_work;
405
406 struct backing_dev_info backing_dev_info;
407
408 /*
409 * The queue owner gets to use this for whatever they like.
410 * ll_rw_blk doesn't touch it.
411 */
412 void *queuedata;
413
414 void *activity_data;
415
416 /*
417 * queue needs bounce pages for pages above this limit
418 */
419 unsigned long bounce_pfn;
Al Viro8267e262005-10-21 03:20:53 -0400420 gfp_t bounce_gfp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421
422 /*
423 * various queue flags, see QUEUE_* below
424 */
425 unsigned long queue_flags;
426
427 /*
152587d2005-04-12 16:22:06 -0500428 * protects queue structures from reentrancy. ->__queue_lock should
429 * _never_ be used directly, it is queue private. always use
430 * ->queue_lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 */
152587d2005-04-12 16:22:06 -0500432 spinlock_t __queue_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 spinlock_t *queue_lock;
434
435 /*
436 * queue kobject
437 */
438 struct kobject kobj;
439
440 /*
441 * queue settings
442 */
443 unsigned long nr_requests; /* Max # of requests */
444 unsigned int nr_congestion_on;
445 unsigned int nr_congestion_off;
446 unsigned int nr_batching;
447
Jens Axboe2cb2e142006-01-17 09:04:32 +0100448 unsigned int max_sectors;
449 unsigned int max_hw_sectors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 unsigned short max_phys_segments;
451 unsigned short max_hw_segments;
452 unsigned short hardsect_size;
453 unsigned int max_segment_size;
454
455 unsigned long seg_boundary_mask;
456 unsigned int dma_alignment;
457
458 struct blk_queue_tag *queue_tags;
459
Tejun Heo15853af2005-11-10 08:52:05 +0100460 unsigned int nr_sorted;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 unsigned int in_flight;
462
463 /*
464 * sg stuff
465 */
466 unsigned int sg_timeout;
467 unsigned int sg_reserved_size;
Christoph Lameter19460892005-06-23 00:08:19 -0700468 int node;
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700469#ifdef CONFIG_BLK_DEV_IO_TRACE
Jens Axboe2056a782006-03-23 20:00:26 +0100470 struct blk_trace *blk_trace;
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700471#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 /*
473 * reserved for flush operations
474 */
Tejun Heo797e7db2006-01-06 09:51:03 +0100475 unsigned int ordered, next_ordered, ordseq;
476 int orderr, ordcolor;
477 struct request pre_flush_rq, bar_rq, post_flush_rq;
478 struct request *orig_bar_rq;
479 unsigned int bi_size;
Al Viro483f4af2006-03-18 18:34:37 -0500480
481 struct mutex sysfs_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482};
483
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484#define QUEUE_FLAG_CLUSTER 0 /* cluster several segments into 1 */
485#define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */
486#define QUEUE_FLAG_STOPPED 2 /* queue is stopped */
487#define QUEUE_FLAG_READFULL 3 /* write queue has been filled */
488#define QUEUE_FLAG_WRITEFULL 4 /* read queue has been filled */
489#define QUEUE_FLAG_DEAD 5 /* queue being torn down */
490#define QUEUE_FLAG_REENTER 6 /* Re-entrancy avoidance */
491#define QUEUE_FLAG_PLUGGED 7 /* queue is plugged */
Jens Axboe64521d12005-10-28 08:30:39 +0200492#define QUEUE_FLAG_ELVSWITCH 8 /* don't use elevator, just do FIFO */
Tejun Heo797e7db2006-01-06 09:51:03 +0100493
494enum {
495 /*
496 * Hardbarrier is supported with one of the following methods.
497 *
498 * NONE : hardbarrier unsupported
499 * DRAIN : ordering by draining is enough
500 * DRAIN_FLUSH : ordering by draining w/ pre and post flushes
501 * DRAIN_FUA : ordering by draining w/ pre flush and FUA write
502 * TAG : ordering by tag is enough
503 * TAG_FLUSH : ordering by tag w/ pre and post flushes
504 * TAG_FUA : ordering by tag w/ pre flush and FUA write
505 */
506 QUEUE_ORDERED_NONE = 0x00,
507 QUEUE_ORDERED_DRAIN = 0x01,
508 QUEUE_ORDERED_TAG = 0x02,
509
510 QUEUE_ORDERED_PREFLUSH = 0x10,
511 QUEUE_ORDERED_POSTFLUSH = 0x20,
512 QUEUE_ORDERED_FUA = 0x40,
513
514 QUEUE_ORDERED_DRAIN_FLUSH = QUEUE_ORDERED_DRAIN |
515 QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
516 QUEUE_ORDERED_DRAIN_FUA = QUEUE_ORDERED_DRAIN |
517 QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
518 QUEUE_ORDERED_TAG_FLUSH = QUEUE_ORDERED_TAG |
519 QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_POSTFLUSH,
520 QUEUE_ORDERED_TAG_FUA = QUEUE_ORDERED_TAG |
521 QUEUE_ORDERED_PREFLUSH | QUEUE_ORDERED_FUA,
522
523 /*
524 * Ordered operation sequence
525 */
526 QUEUE_ORDSEQ_STARTED = 0x01, /* flushing in progress */
527 QUEUE_ORDSEQ_DRAIN = 0x02, /* waiting for the queue to be drained */
528 QUEUE_ORDSEQ_PREFLUSH = 0x04, /* pre-flushing in progress */
529 QUEUE_ORDSEQ_BAR = 0x08, /* original barrier req in progress */
530 QUEUE_ORDSEQ_POSTFLUSH = 0x10, /* post-flushing in progress */
531 QUEUE_ORDSEQ_DONE = 0x20,
532};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
534#define blk_queue_plugged(q) test_bit(QUEUE_FLAG_PLUGGED, &(q)->queue_flags)
535#define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
536#define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
Tejun Heo797e7db2006-01-06 09:51:03 +0100537#define blk_queue_flushing(q) ((q)->ordseq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
Jens Axboe4aff5e22006-08-10 08:44:47 +0200539#define blk_fs_request(rq) ((rq)->cmd_type == REQ_TYPE_FS)
540#define blk_pc_request(rq) ((rq)->cmd_type == REQ_TYPE_BLOCK_PC)
541#define blk_special_request(rq) ((rq)->cmd_type == REQ_TYPE_SPECIAL)
542#define blk_sense_request(rq) ((rq)->cmd_type == REQ_TYPE_SENSE)
543
544#define blk_noretry_request(rq) ((rq)->cmd_flags & REQ_FAILFAST)
545#define blk_rq_started(rq) ((rq)->cmd_flags & REQ_STARTED)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546
547#define blk_account_rq(rq) (blk_rq_started(rq) && blk_fs_request(rq))
548
Jens Axboe4aff5e22006-08-10 08:44:47 +0200549#define blk_pm_suspend_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND)
550#define blk_pm_resume_request(rq) ((rq)->cmd_type == REQ_TYPE_PM_RESUME)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551#define blk_pm_request(rq) \
Jens Axboe4aff5e22006-08-10 08:44:47 +0200552 (blk_pm_suspend_request(rq) || blk_pm_resume_request(rq))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553
Jens Axboe4aff5e22006-08-10 08:44:47 +0200554#define blk_sorted_rq(rq) ((rq)->cmd_flags & REQ_SORTED)
555#define blk_barrier_rq(rq) ((rq)->cmd_flags & REQ_HARDBARRIER)
556#define blk_fua_rq(rq) ((rq)->cmd_flags & REQ_FUA)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557
558#define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
559
Jens Axboe4aff5e22006-08-10 08:44:47 +0200560#define rq_data_dir(rq) ((rq)->cmd_flags & 1)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561
Jens Axboe9e2585a2006-07-28 09:26:13 +0200562/*
563 * We regard a request as sync, if it's a READ or a SYNC write.
564 */
565#define rq_is_sync(rq) (rq_data_dir((rq)) == READ || (rq)->cmd_flags & REQ_RW_SYNC)
Jens Axboe5404bc72006-08-10 09:01:02 +0200566#define rq_is_meta(rq) ((rq)->cmd_flags & REQ_RW_META)
Jens Axboe9e2585a2006-07-28 09:26:13 +0200567
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568static inline int blk_queue_full(struct request_queue *q, int rw)
569{
570 if (rw == READ)
571 return test_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
572 return test_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
573}
574
575static inline void blk_set_queue_full(struct request_queue *q, int rw)
576{
577 if (rw == READ)
578 set_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
579 else
580 set_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
581}
582
583static inline void blk_clear_queue_full(struct request_queue *q, int rw)
584{
585 if (rw == READ)
586 clear_bit(QUEUE_FLAG_READFULL, &q->queue_flags);
587 else
588 clear_bit(QUEUE_FLAG_WRITEFULL, &q->queue_flags);
589}
590
591
592/*
593 * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may
594 * it already be started by driver.
595 */
596#define RQ_NOMERGE_FLAGS \
597 (REQ_NOMERGE | REQ_STARTED | REQ_HARDBARRIER | REQ_SOFTBARRIER)
598#define rq_mergeable(rq) \
Jens Axboe4aff5e22006-08-10 08:44:47 +0200599 (!((rq)->cmd_flags & RQ_NOMERGE_FLAGS) && blk_fs_request((rq)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
601/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 * q->prep_rq_fn return values
603 */
604#define BLKPREP_OK 0 /* serve it */
605#define BLKPREP_KILL 1 /* fatal error, kill */
606#define BLKPREP_DEFER 2 /* leave on queue */
607
608extern unsigned long blk_max_low_pfn, blk_max_pfn;
609
610/*
611 * standard bounce addresses:
612 *
613 * BLK_BOUNCE_HIGH : bounce all highmem pages
614 * BLK_BOUNCE_ANY : don't bounce anything
615 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary
616 */
617#define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT)
618#define BLK_BOUNCE_ANY ((u64)blk_max_pfn << PAGE_SHIFT)
619#define BLK_BOUNCE_ISA (ISA_DMA_THRESHOLD)
620
621#ifdef CONFIG_MMU
622extern int init_emergency_isa_pool(void);
623extern void blk_queue_bounce(request_queue_t *q, struct bio **bio);
624#else
625static inline int init_emergency_isa_pool(void)
626{
627 return 0;
628}
629static inline void blk_queue_bounce(request_queue_t *q, struct bio **bio)
630{
631}
632#endif /* CONFIG_MMU */
633
634#define rq_for_each_bio(_bio, rq) \
635 if ((rq->bio)) \
636 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next)
637
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638extern int blk_register_queue(struct gendisk *disk);
639extern void blk_unregister_queue(struct gendisk *disk);
640extern void register_disk(struct gendisk *dev);
641extern void generic_make_request(struct bio *bio);
642extern void blk_put_request(struct request *);
Mike Christie6e39b69e2005-11-11 05:30:24 -0600643extern void __blk_put_request(request_queue_t *, struct request *);
Tejun Heo8ffdc652006-01-06 09:49:03 +0100644extern void blk_end_sync_rq(struct request *rq, int error);
Al Viro8267e262005-10-21 03:20:53 -0400645extern struct request *blk_get_request(request_queue_t *, int, gfp_t);
Tejun Heo 867d1192005-04-24 02:06:05 -0500646extern void blk_insert_request(request_queue_t *, struct request *, int, void *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647extern void blk_requeue_request(request_queue_t *, struct request *);
648extern void blk_plug_device(request_queue_t *);
649extern int blk_remove_plug(request_queue_t *);
650extern void blk_recount_segments(request_queue_t *, struct bio *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651extern int scsi_cmd_ioctl(struct file *, struct gendisk *, unsigned int, void __user *);
Christoph Hellwig21b2f0c2006-03-22 17:52:04 +0100652extern int sg_scsi_ioctl(struct file *, struct request_queue *,
653 struct gendisk *, struct scsi_ioctl_command __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654extern void blk_start_queue(request_queue_t *q);
655extern void blk_stop_queue(request_queue_t *q);
656extern void blk_sync_queue(struct request_queue *q);
657extern void __blk_stop_queue(request_queue_t *q);
658extern void blk_run_queue(request_queue_t *);
Jens Axboedc72ef42006-07-20 14:54:05 +0200659extern void blk_start_queueing(request_queue_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660extern void blk_queue_activity_fn(request_queue_t *, activity_fn *, void *);
Jens Axboedd1cab92005-06-20 14:06:01 +0200661extern int blk_rq_map_user(request_queue_t *, struct request *, void __user *, unsigned int);
662extern int blk_rq_unmap_user(struct bio *, unsigned int);
Al Viro8267e262005-10-21 03:20:53 -0400663extern int blk_rq_map_kern(request_queue_t *, struct request *, void *, unsigned int, gfp_t);
James Bottomley f1970ba2005-06-20 14:06:52 +0200664extern int blk_rq_map_user_iov(request_queue_t *, struct request *, struct sg_iovec *, int);
James Bottomley 994ca9a2005-06-20 14:11:09 +0200665extern int blk_execute_rq(request_queue_t *, struct gendisk *,
666 struct request *, int);
Mike Christie6e39b69e2005-11-11 05:30:24 -0600667extern void blk_execute_rq_nowait(request_queue_t *, struct gendisk *,
Jens Axboe15fc8582006-01-06 10:00:50 +0100668 struct request *, int, rq_end_io_fn *);
Mike Christie6e39b69e2005-11-11 05:30:24 -0600669
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670static inline request_queue_t *bdev_get_queue(struct block_device *bdev)
671{
672 return bdev->bd_disk->queue;
673}
674
675static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
676 struct page *page)
677{
678 if (bdi && bdi->unplug_io_fn)
679 bdi->unplug_io_fn(bdi, page);
680}
681
682static inline void blk_run_address_space(struct address_space *mapping)
683{
684 if (mapping)
685 blk_run_backing_dev(mapping->backing_dev_info, NULL);
686}
687
688/*
689 * end_request() and friends. Must be called with the request queue spinlock
690 * acquired. All functions called within end_request() _must_be_ atomic.
691 *
692 * Several drivers define their own end_request and call
693 * end_that_request_first() and end_that_request_last()
694 * for parts of the original function. This prevents
695 * code duplication in drivers.
696 */
697extern int end_that_request_first(struct request *, int, int);
698extern int end_that_request_chunk(struct request *, int, int);
Tejun Heo8ffdc652006-01-06 09:49:03 +0100699extern void end_that_request_last(struct request *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700extern void end_request(struct request *req, int uptodate);
Jens Axboeff856ba2006-01-09 16:02:34 +0100701extern void blk_complete_request(struct request *);
702
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703/*
704 * end_that_request_first/chunk() takes an uptodate argument. we account
705 * any value <= as an io error. 0 means -EIO for compatability reasons,
706 * any other < 0 value is the direct error type. An uptodate value of
707 * 1 indicates successful io completion
708 */
709#define end_io_error(uptodate) (unlikely((uptodate) <= 0))
710
711static inline void blkdev_dequeue_request(struct request *req)
712{
Tejun Heo8922e162005-10-20 16:23:44 +0200713 elv_dequeue_request(req->q, req);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716/*
717 * Access functions for manipulating queue properties
718 */
Christoph Lameter19460892005-06-23 00:08:19 -0700719extern request_queue_t *blk_init_queue_node(request_fn_proc *rfn,
720 spinlock_t *lock, int node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721extern request_queue_t *blk_init_queue(request_fn_proc *, spinlock_t *);
722extern void blk_cleanup_queue(request_queue_t *);
723extern void blk_queue_make_request(request_queue_t *, make_request_fn *);
724extern void blk_queue_bounce_limit(request_queue_t *, u64);
Jens Axboe2cb2e142006-01-17 09:04:32 +0100725extern void blk_queue_max_sectors(request_queue_t *, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726extern void blk_queue_max_phys_segments(request_queue_t *, unsigned short);
727extern void blk_queue_max_hw_segments(request_queue_t *, unsigned short);
728extern void blk_queue_max_segment_size(request_queue_t *, unsigned int);
729extern void blk_queue_hardsect_size(request_queue_t *, unsigned short);
730extern void blk_queue_stack_limits(request_queue_t *t, request_queue_t *b);
731extern void blk_queue_segment_boundary(request_queue_t *, unsigned long);
732extern void blk_queue_prep_rq(request_queue_t *, prep_rq_fn *pfn);
733extern void blk_queue_merge_bvec(request_queue_t *, merge_bvec_fn *);
734extern void blk_queue_dma_alignment(request_queue_t *, int);
Jens Axboeff856ba2006-01-09 16:02:34 +0100735extern void blk_queue_softirq_done(request_queue_t *, softirq_done_fn *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
Tejun Heo797e7db2006-01-06 09:51:03 +0100737extern int blk_queue_ordered(request_queue_t *, unsigned, prepare_flush_fn *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738extern void blk_queue_issue_flush_fn(request_queue_t *, issue_flush_fn *);
Tejun Heo797e7db2006-01-06 09:51:03 +0100739extern int blk_do_ordered(request_queue_t *, struct request **);
740extern unsigned blk_ordered_cur_seq(request_queue_t *);
741extern unsigned blk_ordered_req_seq(struct request *);
742extern void blk_ordered_complete_seq(request_queue_t *, unsigned, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
744extern int blk_rq_map_sg(request_queue_t *, struct request *, struct scatterlist *);
745extern void blk_dump_rq_flags(struct request *, char *);
746extern void generic_unplug_device(request_queue_t *);
747extern void __generic_unplug_device(request_queue_t *);
748extern long nr_blockdev_pages(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749
750int blk_get_queue(request_queue_t *);
Al Viro8267e262005-10-21 03:20:53 -0400751request_queue_t *blk_alloc_queue(gfp_t);
752request_queue_t *blk_alloc_queue_node(gfp_t, int);
Al Viro483f4af2006-03-18 18:34:37 -0500753extern void blk_put_queue(request_queue_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
755/*
756 * tag stuff
757 */
758#define blk_queue_tag_depth(q) ((q)->queue_tags->busy)
759#define blk_queue_tag_queue(q) ((q)->queue_tags->busy < (q)->queue_tags->max_depth)
Jens Axboe4aff5e22006-08-10 08:44:47 +0200760#define blk_rq_tagged(rq) ((rq)->cmd_flags & REQ_QUEUED)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761extern int blk_queue_start_tag(request_queue_t *, struct request *);
762extern struct request *blk_queue_find_tag(request_queue_t *, int);
763extern void blk_queue_end_tag(request_queue_t *, struct request *);
764extern int blk_queue_init_tags(request_queue_t *, int, struct blk_queue_tag *);
765extern void blk_queue_free_tags(request_queue_t *);
766extern int blk_queue_resize_tags(request_queue_t *, int);
767extern void blk_queue_invalidate_tags(request_queue_t *);
768extern long blk_congestion_wait(int rw, long timeout);
James Bottomley492dfb42006-08-30 15:48:45 -0400769extern struct blk_queue_tag *blk_init_tags(int);
770extern void blk_free_tags(struct blk_queue_tag *);
Trond Myklebust275a0822006-08-22 20:06:24 -0400771extern void blk_congestion_end(int rw);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772
David C Somayajuluf583f492006-10-04 08:27:25 +0200773static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt,
774 int tag)
775{
776 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
777 return NULL;
778 return bqt->tag_index[tag];
779}
780
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781extern void blk_rq_bio_prep(request_queue_t *, struct request *, struct bio *);
782extern int blkdev_issue_flush(struct block_device *, sector_t *);
783
784#define MAX_PHYS_SEGMENTS 128
785#define MAX_HW_SEGMENTS 128
Mike Christiedefd94b2005-12-05 02:37:06 -0600786#define SAFE_MAX_SECTORS 255
787#define BLK_DEF_MAX_SECTORS 1024
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
789#define MAX_SEGMENT_SIZE 65536
790
791#define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist)
792
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793static inline int queue_hardsect_size(request_queue_t *q)
794{
795 int retval = 512;
796
797 if (q && q->hardsect_size)
798 retval = q->hardsect_size;
799
800 return retval;
801}
802
803static inline int bdev_hardsect_size(struct block_device *bdev)
804{
805 return queue_hardsect_size(bdev_get_queue(bdev));
806}
807
808static inline int queue_dma_alignment(request_queue_t *q)
809{
810 int retval = 511;
811
812 if (q && q->dma_alignment)
813 retval = q->dma_alignment;
814
815 return retval;
816}
817
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818/* assumes size > 256 */
819static inline unsigned int blksize_bits(unsigned int size)
820{
821 unsigned int bits = 8;
822 do {
823 bits++;
824 size >>= 1;
825 } while (size > 256);
826 return bits;
827}
828
Adrian Bunk2befb9e2005-09-10 00:27:17 -0700829static inline unsigned int block_size(struct block_device *bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830{
831 return bdev->bd_block_size;
832}
833
834typedef struct {struct page *v;} Sector;
835
836unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *);
837
838static inline void put_dev_sector(Sector p)
839{
840 page_cache_release(p.v);
841}
842
843struct work_struct;
844int kblockd_schedule_work(struct work_struct *work);
845void kblockd_flush(void);
846
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847#define MODULE_ALIAS_BLOCKDEV(major,minor) \
848 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
849#define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
850 MODULE_ALIAS("block-major-" __stringify(major) "-*")
851
852
David Howells93614012006-09-30 20:45:40 +0200853#else /* CONFIG_BLOCK */
854/*
855 * stubs for when the block layer is configured out
856 */
857#define buffer_heads_over_limit 0
858
859static inline long blk_congestion_wait(int rw, long timeout)
860{
Andrew Mortonbcfd8d32006-08-31 12:56:06 +0200861 return io_schedule_timeout(timeout);
David Howells93614012006-09-30 20:45:40 +0200862}
863
864static inline long nr_blockdev_pages(void)
865{
866 return 0;
867}
868
Andrew Mortonbcfd8d32006-08-31 12:56:06 +0200869static inline void exit_io_context(void)
870{
871}
David Howells93614012006-09-30 20:45:40 +0200872
873#endif /* CONFIG_BLOCK */
874
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875#endif