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 *);
23 * Block error status values. See block/blk-core:blk_errors for the details.
24 * Alpha cannot write a byte atomically, so we need to use 32-bit value.
26 #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__)
27 typedef u32 __bitwise blk_status_t;
29 typedef u8 __bitwise blk_status_t;
32 #define BLK_STS_NOTSUPP ((__force blk_status_t)1)
33 #define BLK_STS_TIMEOUT ((__force blk_status_t)2)
34 #define BLK_STS_NOSPC ((__force blk_status_t)3)
35 #define BLK_STS_TRANSPORT ((__force blk_status_t)4)
36 #define BLK_STS_TARGET ((__force blk_status_t)5)
37 #define BLK_STS_NEXUS ((__force blk_status_t)6)
38 #define BLK_STS_MEDIUM ((__force blk_status_t)7)
39 #define BLK_STS_PROTECTION ((__force blk_status_t)8)
40 #define BLK_STS_RESOURCE ((__force blk_status_t)9)
41 #define BLK_STS_IOERR ((__force blk_status_t)10)
43 /* hack for device mapper, don't use elsewhere: */
44 #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
46 #define BLK_STS_AGAIN ((__force blk_status_t)12)
49 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
50 * device related resources are unavailable, but the driver can guarantee
51 * that the queue will be rerun in the future once resources become
52 * available again. This is typically the case for device specific
53 * resources that are consumed for IO. If the driver fails allocating these
54 * resources, we know that inflight (or pending) IO will free these
55 * resource upon completion.
57 * This is different from BLK_STS_RESOURCE in that it explicitly references
58 * a device specific resource. For resources of wider scope, allocation
59 * failure can happen without having pending IO. This means that we can't
60 * rely on request completions freeing these resources, as IO may not be in
61 * flight. Examples of that are kernel memory allocations, DMA mappings, or
62 * any other system wide resources.
64 #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
67 * blk_path_error - returns true if error may be path related
68 * @error: status the request was completed with
71 * This classifies block error status into non-retryable errors and ones
72 * that may be successful if retried on a failover path.
75 * %false - retrying failover path will not help
76 * %true - may succeed if retried
78 static inline bool blk_path_error(blk_status_t error)
86 case BLK_STS_PROTECTION:
90 /* Anything else could be a path failure, so should be retried */
95 * From most significant bit:
96 * 1 bit: reserved for other usage, see below
97 * 12 bits: original size of bio
98 * 51 bits: issue time of bio
100 #define BIO_ISSUE_RES_BITS 1
101 #define BIO_ISSUE_SIZE_BITS 12
102 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
103 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
104 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
105 #define BIO_ISSUE_SIZE_MASK \
106 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
107 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
109 /* Reserved bit for blk-throtl */
110 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
116 static inline u64 __bio_issue_time(u64 time)
118 return time & BIO_ISSUE_TIME_MASK;
121 static inline u64 bio_issue_time(struct bio_issue *issue)
123 return __bio_issue_time(issue->value);
126 static inline sector_t bio_issue_size(struct bio_issue *issue)
128 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
131 static inline void bio_issue_init(struct bio_issue *issue,
134 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
135 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
136 (ktime_get_ns() & BIO_ISSUE_TIME_MASK) |
137 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
141 * main unit of I/O for the block layer and lower layers (ie drivers and
145 struct bio *bi_next; /* request queue link */
146 struct gendisk *bi_disk;
147 unsigned int bi_opf; /* bottom bits req flags,
148 * top bits REQ_OP. Use
151 unsigned short bi_flags; /* status, etc and bvec pool number */
152 unsigned short bi_ioprio;
153 unsigned short bi_write_hint;
154 blk_status_t bi_status;
157 /* Number of segments in this BIO after
158 * physical address coalescing is performed.
160 unsigned int bi_phys_segments;
163 * To keep track of the max segment size, we account for the
164 * sizes of the first and last mergeable segments in this bio.
166 unsigned int bi_seg_front_size;
167 unsigned int bi_seg_back_size;
169 struct bvec_iter bi_iter;
171 atomic_t __bi_remaining;
172 bio_end_io_t *bi_end_io;
175 #ifdef CONFIG_BLK_CGROUP
177 * Optional ioc and css associated with this bio. Put on bio
178 * release. Read comment on top of bio_associate_current().
180 struct io_context *bi_ioc;
181 struct cgroup_subsys_state *bi_css;
182 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
184 struct bio_issue bi_issue;
188 #if defined(CONFIG_BLK_DEV_INTEGRITY)
189 struct bio_integrity_payload *bi_integrity; /* data integrity */
193 unsigned short bi_vcnt; /* how many bio_vec's */
196 * Everything starting with bi_max_vecs will be preserved by bio_reset()
199 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
201 atomic_t __bi_cnt; /* pin count */
203 struct bio_vec *bi_io_vec; /* the actual vec list */
205 struct bio_set *bi_pool;
208 * We can inline a number of vecs at the end of the bio, to avoid
209 * double allocations for a small number of bio_vecs. This member
210 * MUST obviously be kept at the very end of the bio.
212 struct bio_vec bi_inline_vecs[0];
215 #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
220 #define BIO_SEG_VALID 1 /* bi_phys_segments valid */
221 #define BIO_CLONED 2 /* doesn't own data */
222 #define BIO_BOUNCED 3 /* bio is a bounce bio */
223 #define BIO_USER_MAPPED 4 /* contains user pages */
224 #define BIO_NULL_MAPPED 5 /* contains invalid user pages */
225 #define BIO_QUIET 6 /* Make BIO Quiet */
226 #define BIO_CHAIN 7 /* chained bio, ->bi_remaining in effect */
227 #define BIO_REFFED 8 /* bio has elevated ->bi_cnt */
228 #define BIO_THROTTLED 9 /* This bio has already been subjected to
229 * throttling rules. Don't do it again. */
230 #define BIO_TRACE_COMPLETION 10 /* bio_endio() should trace the final completion
232 #define BIO_QUEUE_ENTERED 11 /* can use blk_queue_enter_live() */
234 /* See BVEC_POOL_OFFSET below before adding new flags */
237 * We support 6 different bvec pools, the last one is magic in that it
238 * is backed by a mempool.
240 #define BVEC_POOL_NR 6
241 #define BVEC_POOL_MAX (BVEC_POOL_NR - 1)
244 * Top 3 bits of bio flags indicate the pool the bvecs came from. We add
245 * 1 to the actual index so that 0 indicates that there are no bvecs to be
248 #define BVEC_POOL_BITS (3)
249 #define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS)
250 #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET)
251 #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1)
252 # error "BVEC_POOL_BITS is too small"
256 * Flags starting here get preserved by bio_reset() - this includes
257 * only BVEC_POOL_IDX()
259 #define BIO_RESET_BITS BVEC_POOL_OFFSET
261 typedef __u32 __bitwise blk_mq_req_flags_t;
264 * Operations and flags common to the bio and request structures.
265 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
267 * The least significant bit of the operation number indicates the data
268 * transfer direction:
270 * - if the least significant bit is set transfers are TO the device
271 * - if the least significant bit is not set transfers are FROM the device
273 * If a operation does not transfer data the least significant bit has no
276 #define REQ_OP_BITS 8
277 #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
278 #define REQ_FLAG_BITS 24
281 /* read sectors from the device */
283 /* write sectors to the device */
285 /* flush the volatile write cache */
287 /* discard sectors */
289 /* get zone information */
290 REQ_OP_ZONE_REPORT = 4,
291 /* securely erase sectors */
292 REQ_OP_SECURE_ERASE = 5,
293 /* seset a zone write pointer */
294 REQ_OP_ZONE_RESET = 6,
295 /* write the same sector many times */
296 REQ_OP_WRITE_SAME = 7,
297 /* write the zero filled sector many times */
298 REQ_OP_WRITE_ZEROES = 9,
300 /* SCSI passthrough using struct scsi_request */
302 REQ_OP_SCSI_OUT = 33,
303 /* Driver private requests */
311 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
313 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
314 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
315 __REQ_SYNC, /* request is sync (sync write or read) */
316 __REQ_META, /* metadata io request */
317 __REQ_PRIO, /* boost priority in cfq */
318 __REQ_NOMERGE, /* don't touch this for merging */
319 __REQ_IDLE, /* anticipate more IO after this one */
320 __REQ_INTEGRITY, /* I/O includes block integrity payload */
321 __REQ_FUA, /* forced unit access */
322 __REQ_PREFLUSH, /* request for cache flush */
323 __REQ_RAHEAD, /* read ahead, can fail anytime */
324 __REQ_BACKGROUND, /* background IO */
325 __REQ_NOWAIT, /* Don't wait if request will block */
327 /* command specific flags for REQ_OP_WRITE_ZEROES: */
328 __REQ_NOUNMAP, /* do not free blocks when zeroing */
333 __REQ_NR_BITS, /* stops here */
336 #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV)
337 #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT)
338 #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER)
339 #define REQ_SYNC (1ULL << __REQ_SYNC)
340 #define REQ_META (1ULL << __REQ_META)
341 #define REQ_PRIO (1ULL << __REQ_PRIO)
342 #define REQ_NOMERGE (1ULL << __REQ_NOMERGE)
343 #define REQ_IDLE (1ULL << __REQ_IDLE)
344 #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY)
345 #define REQ_FUA (1ULL << __REQ_FUA)
346 #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH)
347 #define REQ_RAHEAD (1ULL << __REQ_RAHEAD)
348 #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND)
349 #define REQ_NOWAIT (1ULL << __REQ_NOWAIT)
351 #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
353 #define REQ_DRV (1ULL << __REQ_DRV)
355 #define REQ_FAILFAST_MASK \
356 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
358 #define REQ_NOMERGE_FLAGS \
359 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
361 #define bio_op(bio) \
362 ((bio)->bi_opf & REQ_OP_MASK)
363 #define req_op(req) \
364 ((req)->cmd_flags & REQ_OP_MASK)
366 /* obsolete, don't use in new code */
367 static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
370 bio->bi_opf = op | op_flags;
373 static inline bool op_is_write(unsigned int op)
379 * Check if the bio or request is one that needs special treatment in the
380 * flush state machine.
382 static inline bool op_is_flush(unsigned int op)
384 return op & (REQ_FUA | REQ_PREFLUSH);
388 * Reads are always treated as synchronous, as are requests with the FUA or
389 * PREFLUSH flag. Other operations may be marked as synchronous using the
392 static inline bool op_is_sync(unsigned int op)
394 return (op & REQ_OP_MASK) == REQ_OP_READ ||
395 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
398 typedef unsigned int blk_qc_t;
399 #define BLK_QC_T_NONE -1U
400 #define BLK_QC_T_SHIFT 16
401 #define BLK_QC_T_INTERNAL (1U << 31)
403 static inline bool blk_qc_t_valid(blk_qc_t cookie)
405 return cookie != BLK_QC_T_NONE;
408 static inline blk_qc_t blk_tag_to_qc_t(unsigned int tag, unsigned int queue_num,
411 blk_qc_t ret = tag | (queue_num << BLK_QC_T_SHIFT);
414 ret |= BLK_QC_T_INTERNAL;
419 static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
421 return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
424 static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
426 return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
429 static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
431 return (cookie & BLK_QC_T_INTERNAL) != 0;
442 #endif /* __LINUX_BLK_TYPES_H */