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>
16 struct bio_integrity_payload;
19 struct cgroup_subsys_state;
20 typedef void (bio_end_io_t) (struct bio *);
24 sector_t bd_start_sect;
25 struct disk_stats __percpu *bd_stats;
26 unsigned long bd_stamp;
27 bool bd_read_only; /* read-only policy */
30 struct inode * bd_inode; /* will die */
31 struct super_block * bd_super;
32 struct mutex bd_mutex; /* open/close mutex */
34 struct device bd_device;
39 struct list_head bd_holder_disks;
41 struct kobject *bd_holder_dir;
43 /* number of times partitions within this device have been opened. */
44 unsigned bd_part_count;
46 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
47 struct gendisk * bd_disk;
48 struct backing_dev_info *bd_bdi;
50 /* The counter of freeze processes */
51 int bd_fsfreeze_count;
52 /* Mutex for freeze */
53 struct mutex bd_fsfreeze_mutex;
54 struct super_block *bd_fsfreeze_sb;
56 struct partition_meta_info *bd_meta_info;
57 #ifdef CONFIG_FAIL_MAKE_REQUEST
62 #define bdev_whole(_bdev) \
63 ((_bdev)->bd_disk->part0)
65 #define dev_to_bdev(device) \
66 container_of((device), struct block_device, bd_device)
68 #define bdev_kobj(_bdev) \
69 (&((_bdev)->bd_device.kobj))
72 * Block error status values. See block/blk-core:blk_errors for the details.
73 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
75 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
76 typedef u32 __bitwise blk_status_t;
78 typedef u8 __bitwise blk_status_t;
81 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
82 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
83 #define BLK_STS_NOSPC ((__force blk_status_t)3)
84 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
85 #define BLK_STS_TARGET ((__force blk_status_t)5)
86 #define BLK_STS_NEXUS ((__force blk_status_t)6)
87 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
88 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
89 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
90 #define BLK_STS_IOERR ((__force blk_status_t)10)
92 /* hack for device mapper, don't use elsewhere: */
93 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
95 #define BLK_STS_AGAIN ((__force blk_status_t)12)
98 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
99 * device related resources are unavailable, but the driver can guarantee
100 * that the queue will be rerun in the future once resources become
101 * available again. This is typically the case for device specific
102 * resources that are consumed for IO. If the driver fails allocating these
103 * resources, we know that inflight (or pending) IO will free these
104 * resource upon completion.
106 * This is different from BLK_STS_RESOURCE in that it explicitly references
107 * a device specific resource. For resources of wider scope, allocation
108 * failure can happen without having pending IO. This means that we can't
109 * rely on request completions freeing these resources, as IO may not be in
110 * flight. Examples of that are kernel memory allocations, DMA mappings, or
111 * any other system wide resources.
113 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
116 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
117 * related resources are unavailable, but the driver can guarantee the queue
118 * will be rerun in the future once the resources become available again.
120 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
121 * a zone specific resource and IO to a different zone on the same device could
122 * still be served. Examples of that are zones that are write-locked, but a read
123 * to the same zone could be served.
125 #define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
128 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion
129 * path if the device returns a status indicating that too many zone resources
130 * are currently open. The same command should be successful if resubmitted
131 * after the number of open zones decreases below the device's limits, which is
132 * reported in the request_queue's max_open_zones.
134 #define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)15)
137 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion
138 * path if the device returns a status indicating that too many zone resources
139 * are currently active. The same command should be successful if resubmitted
140 * after the number of active zones decreases below the device's limits, which
141 * is reported in the request_queue's max_active_zones.
143 #define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)16)
146 * blk_path_error - returns true if error may be path related
147 * @error: status the request was completed with
150 * This classifies block error status into non-retryable errors and ones
151 * that may be successful if retried on a failover path.
154 * %false - retrying failover path will not help
155 * %true - may succeed if retried
157 static inline bool blk_path_error(blk_status_t error)
160 case BLK_STS_NOTSUPP:
165 case BLK_STS_PROTECTION:
169 /* Anything else could be a path failure, so should be retried */
174 * From most significant bit:
175 * 1 bit: reserved for other usage, see below
176 * 12 bits: original size of bio
177 * 51 bits: issue time of bio
179 #define BIO_ISSUE_RES_BITS 1
180 #define BIO_ISSUE_SIZE_BITS 12
181 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
182 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
183 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
184 #define BIO_ISSUE_SIZE_MASK \
185 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
186 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
188 /* Reserved bit for blk-throtl */
189 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
195 static inline u64 __bio_issue_time(u64 time)
197 return time & BIO_ISSUE_TIME_MASK;
200 static inline u64 bio_issue_time(struct bio_issue *issue)
202 return __bio_issue_time(issue->value);
205 static inline sector_t bio_issue_size(struct bio_issue *issue)
207 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
210 static inline void bio_issue_init(struct bio_issue *issue,
213 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
214 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
215 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
216 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
220 * main unit of I/O for the block layer and lower layers (ie drivers and
224 struct bio *bi_next; /* request queue link */
225 struct gendisk *bi_disk;
226 unsigned int bi_opf; /* bottom bits req flags,
227 * top bits REQ_OP. Use
230 unsigned short bi_flags; /* status, etc and bvec pool number */
231 unsigned short bi_ioprio;
232 unsigned short bi_write_hint;
233 blk_status_t bi_status;
235 atomic_t __bi_remaining;
237 struct bvec_iter bi_iter;
239 bio_end_io_t *bi_end_io;
242 #ifdef CONFIG_BLK_CGROUP
244 * Represents the association of the css and request_queue for the bio.
245 * If a bio goes direct to device, it will not have a blkg as it will
246 * not have a request_queue associated with it. The reference is put
247 * on release of the bio.
249 struct blkcg_gq *bi_blkg;
250 struct bio_issue bi_issue;
251 #ifdef CONFIG_BLK_CGROUP_IOCOST
256 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
257 struct bio_crypt_ctx *bi_crypt_context;
261 #if defined(CONFIG_BLK_DEV_INTEGRITY)
262 struct bio_integrity_payload *bi_integrity; /* data integrity */
266 unsigned short bi_vcnt; /* how many bio_vec's */
269 * Everything starting with bi_max_vecs will be preserved by bio_reset()
272 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
274 atomic_t __bi_cnt; /* pin count */
276 struct bio_vec *bi_io_vec; /* the actual vec list */
278 struct bio_set *bi_pool;
281 * We can inline a number of vecs at the end of the bio, to avoid
282 * double allocations for a small number of bio_vecs. This member
283 * MUST obviously be kept at the very end of the bio.
285 struct bio_vec bi_inline_vecs[];
288 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
294 BIO_NO_PAGE_REF, /* don't put release vec pages */
295 BIO_CLONED, /* doesn't own data */
296 BIO_BOUNCED, /* bio is a bounce bio */
297 BIO_WORKINGSET, /* contains userspace workingset pages */
298 BIO_QUIET, /* Make BIO Quiet */
299 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
300 BIO_REFFED, /* bio has elevated ->bi_cnt */
301 BIO_THROTTLED, /* This bio has already been subjected to
302 * throttling rules. Don't do it again. */
303 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
305 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
306 BIO_TRACKED, /* set if bio goes through the rq_qos path */
310 /* See BVEC_POOL_OFFSET below before adding new flags */
313 * We support 6 different bvec pools, the last one is magic in that it
314 * is backed by a mempool.
316 #define BVEC_POOL_NR 6
317 #define BVEC_POOL_MAX (BVEC_POOL_NR - 1)
320 * Top 3 bits of bio flags indicate the pool the bvecs came from. We add
321 * 1 to the actual index so that 0 indicates that there are no bvecs to be
324 #define BVEC_POOL_BITS (3)
325 #define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS)
326 #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET)
327 #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1)
328 # error "BVEC_POOL_BITS is too small"
332 * Flags starting here get preserved by bio_reset() - this includes
333 * only BVEC_POOL_IDX()
335 #define BIO_RESET_BITS BVEC_POOL_OFFSET
337 typedef __u32 __bitwise blk_mq_req_flags_t;
340 * Operations and flags common to the bio and request structures.
341 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
343 * The least significant bit of the operation number indicates the data
344 * transfer direction:
346 * - if the least significant bit is set transfers are TO the device
347 * - if the least significant bit is not set transfers are FROM the device
349 * If a operation does not transfer data the least significant bit has no
352 #define REQ_OP_BITS 8
353 #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
354 #define REQ_FLAG_BITS 24
357 /* read sectors from the device */
359 /* write sectors to the device */
361 /* flush the volatile write cache */
363 /* discard sectors */
365 /* securely erase sectors */
366 REQ_OP_SECURE_ERASE = 5,
367 /* write the same sector many times */
368 REQ_OP_WRITE_SAME = 7,
369 /* write the zero filled sector many times */
370 REQ_OP_WRITE_ZEROES = 9,
372 REQ_OP_ZONE_OPEN = 10,
374 REQ_OP_ZONE_CLOSE = 11,
375 /* Transition a zone to full */
376 REQ_OP_ZONE_FINISH = 12,
377 /* write data at the current zone write pointer */
378 REQ_OP_ZONE_APPEND = 13,
379 /* reset a zone write pointer */
380 REQ_OP_ZONE_RESET = 15,
381 /* reset all the zone present on the device */
382 REQ_OP_ZONE_RESET_ALL = 17,
384 /* SCSI passthrough using struct scsi_request */
386 REQ_OP_SCSI_OUT = 33,
387 /* Driver private requests */
395 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
397 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
398 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
399 __REQ_SYNC, /* request is sync (sync write or read) */
400 __REQ_META, /* metadata io request */
401 __REQ_PRIO, /* boost priority in cfq */
402 __REQ_NOMERGE, /* don't touch this for merging */
403 __REQ_IDLE, /* anticipate more IO after this one */
404 __REQ_INTEGRITY, /* I/O includes block integrity payload */
405 __REQ_FUA, /* forced unit access */
406 __REQ_PREFLUSH, /* request for cache flush */
407 __REQ_RAHEAD, /* read ahead, can fail anytime */
408 __REQ_BACKGROUND, /* background IO */
409 __REQ_NOWAIT, /* Don't wait if request will block */
411 * When a shared kthread needs to issue a bio for a cgroup, doing
412 * so synchronously can lead to priority inversions as the kthread
413 * can be trapped waiting for that cgroup. CGROUP_PUNT flag makes
414 * submit_bio() punt the actual issuing to a dedicated per-blkcg
415 * work item to avoid such priority inversions.
419 /* command specific flags for REQ_OP_WRITE_ZEROES: */
420 __REQ_NOUNMAP, /* do not free blocks when zeroing */
426 __REQ_SWAP, /* swapping request. */
427 __REQ_NR_BITS, /* stops here */
430 #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV)
431 #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT)
432 #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER)
433 #define REQ_SYNC (1ULL << __REQ_SYNC)
434 #define REQ_META (1ULL << __REQ_META)
435 #define REQ_PRIO (1ULL << __REQ_PRIO)
436 #define REQ_NOMERGE (1ULL << __REQ_NOMERGE)
437 #define REQ_IDLE (1ULL << __REQ_IDLE)
438 #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY)
439 #define REQ_FUA (1ULL << __REQ_FUA)
440 #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH)
441 #define REQ_RAHEAD (1ULL << __REQ_RAHEAD)
442 #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND)
443 #define REQ_NOWAIT (1ULL << __REQ_NOWAIT)
444 #define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT)
446 #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
447 #define REQ_HIPRI (1ULL << __REQ_HIPRI)
449 #define REQ_DRV (1ULL << __REQ_DRV)
450 #define REQ_SWAP (1ULL << __REQ_SWAP)
452 #define REQ_FAILFAST_MASK \
453 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
455 #define REQ_NOMERGE_FLAGS \
456 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
467 #define bio_op(bio) \
468 ((bio)->bi_opf & REQ_OP_MASK)
469 #define req_op(req) \
470 ((req)->cmd_flags & REQ_OP_MASK)
472 /* obsolete, don't use in new code */
473 static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
476 bio->bi_opf = op | op_flags;
479 static inline bool op_is_write(unsigned int op)
485 * Check if the bio or request is one that needs special treatment in the
486 * flush state machine.
488 static inline bool op_is_flush(unsigned int op)
490 return op & (REQ_FUA | REQ_PREFLUSH);
494 * Reads are always treated as synchronous, as are requests with the FUA or
495 * PREFLUSH flag. Other operations may be marked as synchronous using the
498 static inline bool op_is_sync(unsigned int op)
500 return (op & REQ_OP_MASK) == REQ_OP_READ ||
501 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
504 static inline bool op_is_discard(unsigned int op)
506 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
510 * Check if a bio or request operation is a zone management operation, with
511 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
512 * due to its different handling in the block layer and device response in
513 * case of command failure.
515 static inline bool op_is_zone_mgmt(enum req_opf op)
517 switch (op & REQ_OP_MASK) {
518 case REQ_OP_ZONE_RESET:
519 case REQ_OP_ZONE_OPEN:
520 case REQ_OP_ZONE_CLOSE:
521 case REQ_OP_ZONE_FINISH:
528 static inline int op_stat_group(unsigned int op)
530 if (op_is_discard(op))
532 return op_is_write(op);
535 typedef unsigned int blk_qc_t;
536 #define BLK_QC_T_NONE -1U
537 #define BLK_QC_T_SHIFT 16
538 #define BLK_QC_T_INTERNAL (1U << 31)
540 static inline bool blk_qc_t_valid(blk_qc_t cookie)
542 return cookie != BLK_QC_T_NONE;
545 static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
547 return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
550 static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
552 return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
555 static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
557 return (cookie & BLK_QC_T_INTERNAL) != 0;
568 #endif /* __LINUX_BLK_TYPES_H */