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/ktime.h>
15 struct bio_integrity_payload;
19 struct cgroup_subsys_state;
20 typedef void (bio_end_io_t) (struct bio *);
24 * Block error status values. See block/blk-core:blk_errors for the details.
25 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
27 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
28 typedef u32 __bitwise blk_status_t;
30 typedef u8 __bitwise blk_status_t;
33 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
34 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
35 #define BLK_STS_NOSPC ((__force blk_status_t)3)
36 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
37 #define BLK_STS_TARGET ((__force blk_status_t)5)
38 #define BLK_STS_NEXUS ((__force blk_status_t)6)
39 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
40 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
41 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
42 #define BLK_STS_IOERR ((__force blk_status_t)10)
44 /* hack for device mapper, don't use elsewhere: */
45 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
47 #define BLK_STS_AGAIN ((__force blk_status_t)12)
50 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
51 * device related resources are unavailable, but the driver can guarantee
52 * that the queue will be rerun in the future once resources become
53 * available again. This is typically the case for device specific
54 * resources that are consumed for IO. If the driver fails allocating these
55 * resources, we know that inflight (or pending) IO will free these
56 * resource upon completion.
58 * This is different from BLK_STS_RESOURCE in that it explicitly references
59 * a device specific resource. For resources of wider scope, allocation
60 * failure can happen without having pending IO. This means that we can't
61 * rely on request completions freeing these resources, as IO may not be in
62 * flight. Examples of that are kernel memory allocations, DMA mappings, or
63 * any other system wide resources.
65 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
68 * BLK_STS_ZONE_RESOURCE is returned from the driver to the block layer if zone
69 * related resources are unavailable, but the driver can guarantee the queue
70 * will be rerun in the future once the resources become available again.
72 * This is different from BLK_STS_DEV_RESOURCE in that it explicitly references
73 * a zone specific resource and IO to a different zone on the same device could
74 * still be served. Examples of that are zones that are write-locked, but a read
75 * to the same zone could be served.
77 #define BLK_STS_ZONE_RESOURCE ((__force blk_status_t)14)
80 * blk_path_error - returns true if error may be path related
81 * @error: status the request was completed with
84 * This classifies block error status into non-retryable errors and ones
85 * that may be successful if retried on a failover path.
88 * %false - retrying failover path will not help
89 * %true - may succeed if retried
91 static inline bool blk_path_error(blk_status_t error)
99 case BLK_STS_PROTECTION:
103 /* Anything else could be a path failure, so should be retried */
108 * From most significant bit:
109 * 1 bit: reserved for other usage, see below
110 * 12 bits: original size of bio
111 * 51 bits: issue time of bio
113 #define BIO_ISSUE_RES_BITS 1
114 #define BIO_ISSUE_SIZE_BITS 12
115 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
116 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
117 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
118 #define BIO_ISSUE_SIZE_MASK \
119 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
120 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
122 /* Reserved bit for blk-throtl */
123 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
129 static inline u64 __bio_issue_time(u64 time)
131 return time & BIO_ISSUE_TIME_MASK;
134 static inline u64 bio_issue_time(struct bio_issue *issue)
136 return __bio_issue_time(issue->value);
139 static inline sector_t bio_issue_size(struct bio_issue *issue)
141 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
144 static inline void bio_issue_init(struct bio_issue *issue,
147 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
148 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
149 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
150 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
154 * main unit of I/O for the block layer and lower layers (ie drivers and
158 struct bio *bi_next; /* request queue link */
159 struct gendisk *bi_disk;
160 unsigned int bi_opf; /* bottom bits req flags,
161 * top bits REQ_OP. Use
164 unsigned short bi_flags; /* status, etc and bvec pool number */
165 unsigned short bi_ioprio;
166 unsigned short bi_write_hint;
167 blk_status_t bi_status;
169 atomic_t __bi_remaining;
171 struct bvec_iter bi_iter;
173 bio_end_io_t *bi_end_io;
176 #ifdef CONFIG_BLK_CGROUP
178 * Represents the association of the css and request_queue for the bio.
179 * If a bio goes direct to device, it will not have a blkg as it will
180 * not have a request_queue associated with it. The reference is put
181 * on release of the bio.
183 struct blkcg_gq *bi_blkg;
184 struct bio_issue bi_issue;
185 #ifdef CONFIG_BLK_CGROUP_IOCOST
190 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
191 struct bio_crypt_ctx *bi_crypt_context;
195 #if defined(CONFIG_BLK_DEV_INTEGRITY)
196 struct bio_integrity_payload *bi_integrity; /* data integrity */
200 unsigned short bi_vcnt; /* how many bio_vec's */
203 * Everything starting with bi_max_vecs will be preserved by bio_reset()
206 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
208 atomic_t __bi_cnt; /* pin count */
210 struct bio_vec *bi_io_vec; /* the actual vec list */
212 struct bio_set *bi_pool;
215 * We can inline a number of vecs at the end of the bio, to avoid
216 * double allocations for a small number of bio_vecs. This member
217 * MUST obviously be kept at the very end of the bio.
219 struct bio_vec bi_inline_vecs[];
222 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
228 BIO_NO_PAGE_REF, /* don't put release vec pages */
229 BIO_CLONED, /* doesn't own data */
230 BIO_BOUNCED, /* bio is a bounce bio */
231 BIO_USER_MAPPED, /* contains user pages */
232 BIO_NULL_MAPPED, /* contains invalid user pages */
233 BIO_WORKINGSET, /* contains userspace workingset pages */
234 BIO_QUIET, /* Make BIO Quiet */
235 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
236 BIO_REFFED, /* bio has elevated ->bi_cnt */
237 BIO_THROTTLED, /* This bio has already been subjected to
238 * throttling rules. Don't do it again. */
239 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
241 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
242 BIO_TRACKED, /* set if bio goes through the rq_qos path */
246 /* See BVEC_POOL_OFFSET below before adding new flags */
249 * We support 6 different bvec pools, the last one is magic in that it
250 * is backed by a mempool.
252 #define BVEC_POOL_NR 6
253 #define BVEC_POOL_MAX (BVEC_POOL_NR - 1)
256 * Top 3 bits of bio flags indicate the pool the bvecs came from. We add
257 * 1 to the actual index so that 0 indicates that there are no bvecs to be
260 #define BVEC_POOL_BITS (3)
261 #define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS)
262 #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET)
263 #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1)
264 # error "BVEC_POOL_BITS is too small"
268 * Flags starting here get preserved by bio_reset() - this includes
269 * only BVEC_POOL_IDX()
271 #define BIO_RESET_BITS BVEC_POOL_OFFSET
273 typedef __u32 __bitwise blk_mq_req_flags_t;
276 * Operations and flags common to the bio and request structures.
277 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
279 * The least significant bit of the operation number indicates the data
280 * transfer direction:
282 * - if the least significant bit is set transfers are TO the device
283 * - if the least significant bit is not set transfers are FROM the device
285 * If a operation does not transfer data the least significant bit has no
288 #define REQ_OP_BITS 8
289 #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
290 #define REQ_FLAG_BITS 24
293 /* read sectors from the device */
295 /* write sectors to the device */
297 /* flush the volatile write cache */
299 /* discard sectors */
301 /* securely erase sectors */
302 REQ_OP_SECURE_ERASE = 5,
303 /* reset a zone write pointer */
304 REQ_OP_ZONE_RESET = 6,
305 /* write the same sector many times */
306 REQ_OP_WRITE_SAME = 7,
307 /* reset all the zone present on the device */
308 REQ_OP_ZONE_RESET_ALL = 8,
309 /* write the zero filled sector many times */
310 REQ_OP_WRITE_ZEROES = 9,
312 REQ_OP_ZONE_OPEN = 10,
314 REQ_OP_ZONE_CLOSE = 11,
315 /* Transition a zone to full */
316 REQ_OP_ZONE_FINISH = 12,
317 /* write data at the current zone write pointer */
318 REQ_OP_ZONE_APPEND = 13,
320 /* SCSI passthrough using struct scsi_request */
322 REQ_OP_SCSI_OUT = 33,
323 /* Driver private requests */
331 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
333 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
334 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
335 __REQ_SYNC, /* request is sync (sync write or read) */
336 __REQ_META, /* metadata io request */
337 __REQ_PRIO, /* boost priority in cfq */
338 __REQ_NOMERGE, /* don't touch this for merging */
339 __REQ_IDLE, /* anticipate more IO after this one */
340 __REQ_INTEGRITY, /* I/O includes block integrity payload */
341 __REQ_FUA, /* forced unit access */
342 __REQ_PREFLUSH, /* request for cache flush */
343 __REQ_RAHEAD, /* read ahead, can fail anytime */
344 __REQ_BACKGROUND, /* background IO */
345 __REQ_NOWAIT, /* Don't wait if request will block */
347 * When a shared kthread needs to issue a bio for a cgroup, doing
348 * so synchronously can lead to priority inversions as the kthread
349 * can be trapped waiting for that cgroup. CGROUP_PUNT flag makes
350 * submit_bio() punt the actual issuing to a dedicated per-blkcg
351 * work item to avoid such priority inversions.
355 /* command specific flags for REQ_OP_WRITE_ZEROES: */
356 __REQ_NOUNMAP, /* do not free blocks when zeroing */
362 __REQ_SWAP, /* swapping request. */
363 __REQ_NR_BITS, /* stops here */
366 #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV)
367 #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT)
368 #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER)
369 #define REQ_SYNC (1ULL << __REQ_SYNC)
370 #define REQ_META (1ULL << __REQ_META)
371 #define REQ_PRIO (1ULL << __REQ_PRIO)
372 #define REQ_NOMERGE (1ULL << __REQ_NOMERGE)
373 #define REQ_IDLE (1ULL << __REQ_IDLE)
374 #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY)
375 #define REQ_FUA (1ULL << __REQ_FUA)
376 #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH)
377 #define REQ_RAHEAD (1ULL << __REQ_RAHEAD)
378 #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND)
379 #define REQ_NOWAIT (1ULL << __REQ_NOWAIT)
380 #define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT)
382 #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
383 #define REQ_HIPRI (1ULL << __REQ_HIPRI)
385 #define REQ_DRV (1ULL << __REQ_DRV)
386 #define REQ_SWAP (1ULL << __REQ_SWAP)
388 #define REQ_FAILFAST_MASK \
389 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
391 #define REQ_NOMERGE_FLAGS \
392 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
403 #define bio_op(bio) \
404 ((bio)->bi_opf & REQ_OP_MASK)
405 #define req_op(req) \
406 ((req)->cmd_flags & REQ_OP_MASK)
408 /* obsolete, don't use in new code */
409 static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
412 bio->bi_opf = op | op_flags;
415 static inline bool op_is_write(unsigned int op)
421 * Check if the bio or request is one that needs special treatment in the
422 * flush state machine.
424 static inline bool op_is_flush(unsigned int op)
426 return op & (REQ_FUA | REQ_PREFLUSH);
430 * Reads are always treated as synchronous, as are requests with the FUA or
431 * PREFLUSH flag. Other operations may be marked as synchronous using the
434 static inline bool op_is_sync(unsigned int op)
436 return (op & REQ_OP_MASK) == REQ_OP_READ ||
437 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
440 static inline bool op_is_discard(unsigned int op)
442 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
446 * Check if a bio or request operation is a zone management operation, with
447 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
448 * due to its different handling in the block layer and device response in
449 * case of command failure.
451 static inline bool op_is_zone_mgmt(enum req_opf op)
453 switch (op & REQ_OP_MASK) {
454 case REQ_OP_ZONE_RESET:
455 case REQ_OP_ZONE_OPEN:
456 case REQ_OP_ZONE_CLOSE:
457 case REQ_OP_ZONE_FINISH:
464 static inline int op_stat_group(unsigned int op)
466 if (op_is_discard(op))
468 return op_is_write(op);
471 typedef unsigned int blk_qc_t;
472 #define BLK_QC_T_NONE -1U
473 #define BLK_QC_T_EAGAIN -2U
474 #define BLK_QC_T_SHIFT 16
475 #define BLK_QC_T_INTERNAL (1U << 31)
477 static inline bool blk_qc_t_valid(blk_qc_t cookie)
479 return cookie != BLK_QC_T_NONE && cookie != BLK_QC_T_EAGAIN;
482 static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
484 return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
487 static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
489 return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
492 static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
494 return (cookie & BLK_QC_T_INTERNAL) != 0;
505 #endif /* __LINUX_BLK_TYPES_H */