1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Block data types and constants. Directly include this file only to
4 * break include dependency loop.
6 #ifndef __LINUX_BLK_TYPES_H
7 #define __LINUX_BLK_TYPES_H
9 #include <linux/types.h>
10 #include <linux/bvec.h>
11 #include <linux/device.h>
12 #include <linux/ktime.h>
13 #include <linux/rw_hint.h>
17 struct bio_integrity_payload;
20 struct cgroup_subsys_state;
21 typedef void (bio_end_io_t) (struct bio *);
25 * The basic unit of block I/O is a sector. It is used in a number of contexts
26 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
27 * bytes. Variables of type sector_t represent an offset or size that is a
28 * multiple of 512 bytes. Hence these two constants.
31 #define SECTOR_SHIFT 9
34 #define SECTOR_SIZE (1 << SECTOR_SHIFT)
37 #define PAGE_SECTORS_SHIFT (PAGE_SHIFT - SECTOR_SHIFT)
38 #define PAGE_SECTORS (1 << PAGE_SECTORS_SHIFT)
39 #define SECTOR_MASK (PAGE_SECTORS - 1)
42 sector_t bd_start_sect;
43 sector_t bd_nr_sectors;
44 struct gendisk * bd_disk;
45 struct request_queue * bd_queue;
46 struct disk_stats __percpu *bd_stats;
47 unsigned long bd_stamp;
48 bool bd_read_only; /* read-only policy */
51 bool bd_has_submit_bio;
53 struct inode *bd_inode; /* will die */
56 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
59 const struct blk_holder_ops *bd_holder_ops;
60 struct mutex bd_holder_lock;
62 struct kobject *bd_holder_dir;
64 atomic_t bd_fsfreeze_count; /* number of freeze requests */
65 struct mutex bd_fsfreeze_mutex; /* serialize freeze/thaw */
67 struct partition_meta_info *bd_meta_info;
68 #ifdef CONFIG_FAIL_MAKE_REQUEST
74 * keep this out-of-line as it's both big and not needed in the fast
77 struct device bd_device;
80 #define bdev_whole(_bdev) \
81 ((_bdev)->bd_disk->part0)
83 #define dev_to_bdev(device) \
84 container_of((device), struct block_device, bd_device)
86 #define bdev_kobj(_bdev) \
87 (&((_bdev)->bd_device.kobj))
90 * Block error status values. See block/blk-core:blk_errors for the details.
92 typedef u8 __bitwise blk_status_t;
93 typedef u16 blk_short_t;
95 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
96 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
97 #define BLK_STS_NOSPC ((__force blk_status_t)3)
98 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
99 #define BLK_STS_TARGET ((__force blk_status_t)5)
100 #define BLK_STS_RESV_CONFLICT ((__force blk_status_t)6)
101 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
102 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
103 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
104 #define BLK_STS_IOERR ((__force blk_status_t)10)
106 /* hack for device mapper, don't use elsewhere: */
107 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
110 * BLK_STS_AGAIN should only be returned if RQF_NOWAIT is set
111 * and the bio would block (cf bio_wouldblock_error())
113 #define BLK_STS_AGAIN ((__force blk_status_t)12)
116 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
117 * device related resources are unavailable, but the driver can guarantee
118 * that the queue will be rerun in the future once resources become
119 * available again. This is typically the case for device specific
120 * resources that are consumed for IO. If the driver fails allocating these
121 * resources, we know that inflight (or pending) IO will free these
122 * resource upon completion.
124 * This is different from BLK_STS_RESOURCE in that it explicitly references
125 * a device specific resource. For resources of wider scope, allocation
126 * failure can happen without having pending IO. This means that we can't
127 * rely on request completions freeing these resources, as IO may not be in
128 * flight. Examples of that are kernel memory allocations, DMA mappings, or
129 * any other system wide resources.
131 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
134 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
135 * related resources are unavailable, but the driver can guarantee the queue
136 * will be rerun in the future once the resources become available again.
138 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
139 * a zone specific resource and IO to a different zone on the same device could
140 * still be served. Examples of that are zones that are write-locked, but a read
141 * to the same zone could be served.
143 #define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
146 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion
147 * path if the device returns a status indicating that too many zone resources
148 * are currently open. The same command should be successful if resubmitted
149 * after the number of open zones decreases below the device's limits, which is
150 * reported in the request_queue's max_open_zones.
152 #define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)15)
155 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion
156 * path if the device returns a status indicating that too many zone resources
157 * are currently active. The same command should be successful if resubmitted
158 * after the number of active zones decreases below the device's limits, which
159 * is reported in the request_queue's max_active_zones.
161 #define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)16)
164 * BLK_STS_OFFLINE is returned from the driver when the target device is offline
165 * or is being taken offline. This could help differentiate the case where a
166 * device is intentionally being shut down from a real I/O error.
168 #define BLK_STS_OFFLINE ((__force blk_status_t)17)
171 * BLK_STS_DURATION_LIMIT is returned from the driver when the target device
172 * aborted the command because it exceeded one of its Command Duration Limits.
174 #define BLK_STS_DURATION_LIMIT ((__force blk_status_t)18)
177 * blk_path_error - returns true if error may be path related
178 * @error: status the request was completed with
181 * This classifies block error status into non-retryable errors and ones
182 * that may be successful if retried on a failover path.
185 * %false - retrying failover path will not help
186 * %true - may succeed if retried
188 static inline bool blk_path_error(blk_status_t error)
191 case BLK_STS_NOTSUPP:
194 case BLK_STS_RESV_CONFLICT:
196 case BLK_STS_PROTECTION:
200 /* Anything else could be a path failure, so should be retried */
208 typedef __u32 __bitwise blk_opf_t;
210 typedef unsigned int blk_qc_t;
211 #define BLK_QC_T_NONE -1U
214 * main unit of I/O for the block layer and lower layers (ie drivers and
218 struct bio *bi_next; /* request queue link */
219 struct block_device *bi_bdev;
220 blk_opf_t bi_opf; /* bottom bits REQ_OP, top bits
223 unsigned short bi_flags; /* BIO_* below */
224 unsigned short bi_ioprio;
225 enum rw_hint bi_write_hint;
226 blk_status_t bi_status;
227 atomic_t __bi_remaining;
229 struct bvec_iter bi_iter;
232 bio_end_io_t *bi_end_io;
234 #ifdef CONFIG_BLK_CGROUP
236 * Represents the association of the css and request_queue for the bio.
237 * If a bio goes direct to device, it will not have a blkg as it will
238 * not have a request_queue associated with it. The reference is put
239 * on release of the bio.
241 struct blkcg_gq *bi_blkg;
242 struct bio_issue bi_issue;
243 #ifdef CONFIG_BLK_CGROUP_IOCOST
248 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
249 struct bio_crypt_ctx *bi_crypt_context;
253 #if defined(CONFIG_BLK_DEV_INTEGRITY)
254 struct bio_integrity_payload *bi_integrity; /* data integrity */
258 unsigned short bi_vcnt; /* how many bio_vec's */
261 * Everything starting with bi_max_vecs will be preserved by bio_reset()
264 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
266 atomic_t __bi_cnt; /* pin count */
268 struct bio_vec *bi_io_vec; /* the actual vec list */
270 struct bio_set *bi_pool;
273 * We can inline a number of vecs at the end of the bio, to avoid
274 * double allocations for a small number of bio_vecs. This member
275 * MUST obviously be kept at the very end of the bio.
277 struct bio_vec bi_inline_vecs[];
280 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
281 #define BIO_MAX_SECTORS (UINT_MAX >> SECTOR_SHIFT)
287 BIO_PAGE_PINNED, /* Unpin pages in bio_release_pages() */
288 BIO_CLONED, /* doesn't own data */
289 BIO_BOUNCED, /* bio is a bounce bio */
290 BIO_QUIET, /* Make BIO Quiet */
291 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
292 BIO_REFFED, /* bio has elevated ->bi_cnt */
293 BIO_BPS_THROTTLED, /* This bio has already been subjected to
294 * throttling rules. Don't do it again. */
295 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
297 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
298 BIO_QOS_THROTTLED, /* bio went through rq_qos throttle path */
299 BIO_QOS_MERGED, /* but went through rq_qos merge path */
301 BIO_ZONE_WRITE_LOCKED, /* Owns a zoned device zone write lock */
305 typedef __u32 __bitwise blk_mq_req_flags_t;
307 #define REQ_OP_BITS 8
308 #define REQ_OP_MASK (__force blk_opf_t)((1 << REQ_OP_BITS) - 1)
309 #define REQ_FLAG_BITS 24
312 * enum req_op - Operations common to the bio and request structures.
313 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
315 * The least significant bit of the operation number indicates the data
316 * transfer direction:
318 * - if the least significant bit is set transfers are TO the device
319 * - if the least significant bit is not set transfers are FROM the device
321 * If a operation does not transfer data the least significant bit has no
325 /* read sectors from the device */
326 REQ_OP_READ = (__force blk_opf_t)0,
327 /* write sectors to the device */
328 REQ_OP_WRITE = (__force blk_opf_t)1,
329 /* flush the volatile write cache */
330 REQ_OP_FLUSH = (__force blk_opf_t)2,
331 /* discard sectors */
332 REQ_OP_DISCARD = (__force blk_opf_t)3,
333 /* securely erase sectors */
334 REQ_OP_SECURE_ERASE = (__force blk_opf_t)5,
335 /* write data at the current zone write pointer */
336 REQ_OP_ZONE_APPEND = (__force blk_opf_t)7,
337 /* write the zero filled sector many times */
338 REQ_OP_WRITE_ZEROES = (__force blk_opf_t)9,
340 REQ_OP_ZONE_OPEN = (__force blk_opf_t)10,
342 REQ_OP_ZONE_CLOSE = (__force blk_opf_t)11,
343 /* Transition a zone to full */
344 REQ_OP_ZONE_FINISH = (__force blk_opf_t)12,
345 /* reset a zone write pointer */
346 REQ_OP_ZONE_RESET = (__force blk_opf_t)13,
347 /* reset all the zone present on the device */
348 REQ_OP_ZONE_RESET_ALL = (__force blk_opf_t)15,
350 /* Driver private requests */
351 REQ_OP_DRV_IN = (__force blk_opf_t)34,
352 REQ_OP_DRV_OUT = (__force blk_opf_t)35,
354 REQ_OP_LAST = (__force blk_opf_t)36,
358 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
360 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
361 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
362 __REQ_SYNC, /* request is sync (sync write or read) */
363 __REQ_META, /* metadata io request */
364 __REQ_PRIO, /* boost priority in cfq */
365 __REQ_NOMERGE, /* don't touch this for merging */
366 __REQ_IDLE, /* anticipate more IO after this one */
367 __REQ_INTEGRITY, /* I/O includes block integrity payload */
368 __REQ_FUA, /* forced unit access */
369 __REQ_PREFLUSH, /* request for cache flush */
370 __REQ_RAHEAD, /* read ahead, can fail anytime */
371 __REQ_BACKGROUND, /* background IO */
372 __REQ_NOWAIT, /* Don't wait if request will block */
373 __REQ_POLLED, /* caller polls for completion using bio_poll */
374 __REQ_ALLOC_CACHE, /* allocate IO from cache if available */
375 __REQ_SWAP, /* swap I/O */
376 __REQ_DRV, /* for driver use */
377 __REQ_FS_PRIVATE, /* for file system (submitter) use */
380 * Command specific flags, keep last:
382 /* for REQ_OP_WRITE_ZEROES: */
383 __REQ_NOUNMAP, /* do not free blocks when zeroing */
385 __REQ_NR_BITS, /* stops here */
388 #define REQ_FAILFAST_DEV \
389 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DEV)
390 #define REQ_FAILFAST_TRANSPORT \
391 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_TRANSPORT)
392 #define REQ_FAILFAST_DRIVER \
393 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DRIVER)
394 #define REQ_SYNC (__force blk_opf_t)(1ULL << __REQ_SYNC)
395 #define REQ_META (__force blk_opf_t)(1ULL << __REQ_META)
396 #define REQ_PRIO (__force blk_opf_t)(1ULL << __REQ_PRIO)
397 #define REQ_NOMERGE (__force blk_opf_t)(1ULL << __REQ_NOMERGE)
398 #define REQ_IDLE (__force blk_opf_t)(1ULL << __REQ_IDLE)
399 #define REQ_INTEGRITY (__force blk_opf_t)(1ULL << __REQ_INTEGRITY)
400 #define REQ_FUA (__force blk_opf_t)(1ULL << __REQ_FUA)
401 #define REQ_PREFLUSH (__force blk_opf_t)(1ULL << __REQ_PREFLUSH)
402 #define REQ_RAHEAD (__force blk_opf_t)(1ULL << __REQ_RAHEAD)
403 #define REQ_BACKGROUND (__force blk_opf_t)(1ULL << __REQ_BACKGROUND)
404 #define REQ_NOWAIT (__force blk_opf_t)(1ULL << __REQ_NOWAIT)
405 #define REQ_POLLED (__force blk_opf_t)(1ULL << __REQ_POLLED)
406 #define REQ_ALLOC_CACHE (__force blk_opf_t)(1ULL << __REQ_ALLOC_CACHE)
407 #define REQ_SWAP (__force blk_opf_t)(1ULL << __REQ_SWAP)
408 #define REQ_DRV (__force blk_opf_t)(1ULL << __REQ_DRV)
409 #define REQ_FS_PRIVATE (__force blk_opf_t)(1ULL << __REQ_FS_PRIVATE)
411 #define REQ_NOUNMAP (__force blk_opf_t)(1ULL << __REQ_NOUNMAP)
413 #define REQ_FAILFAST_MASK \
414 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
416 #define REQ_NOMERGE_FLAGS \
417 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
428 static inline enum req_op bio_op(const struct bio *bio)
430 return bio->bi_opf & REQ_OP_MASK;
433 static inline bool op_is_write(blk_opf_t op)
435 return !!(op & (__force blk_opf_t)1);
439 * Check if the bio or request is one that needs special treatment in the
440 * flush state machine.
442 static inline bool op_is_flush(blk_opf_t op)
444 return op & (REQ_FUA | REQ_PREFLUSH);
448 * Reads are always treated as synchronous, as are requests with the FUA or
449 * PREFLUSH flag. Other operations may be marked as synchronous using the
452 static inline bool op_is_sync(blk_opf_t op)
454 return (op & REQ_OP_MASK) == REQ_OP_READ ||
455 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
458 static inline bool op_is_discard(blk_opf_t op)
460 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
464 * Check if a bio or request operation is a zone management operation, with
465 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
466 * due to its different handling in the block layer and device response in
467 * case of command failure.
469 static inline bool op_is_zone_mgmt(enum req_op op)
471 switch (op & REQ_OP_MASK) {
472 case REQ_OP_ZONE_RESET:
473 case REQ_OP_ZONE_OPEN:
474 case REQ_OP_ZONE_CLOSE:
475 case REQ_OP_ZONE_FINISH:
482 static inline int op_stat_group(enum req_op op)
484 if (op_is_discard(op))
486 return op_is_write(op);
497 #endif /* __LINUX_BLK_TYPES_H */