]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
7cc01581 TH |
2 | /* |
3 | * Block data types and constants. Directly include this file only to | |
4 | * break include dependency loop. | |
5 | */ | |
6 | #ifndef __LINUX_BLK_TYPES_H | |
7 | #define __LINUX_BLK_TYPES_H | |
8 | ||
7cc01581 | 9 | #include <linux/types.h> |
0781e79e | 10 | #include <linux/bvec.h> |
5238dcf4 | 11 | #include <linux/ktime.h> |
7cc01581 TH |
12 | |
13 | struct bio_set; | |
14 | struct bio; | |
15 | struct bio_integrity_payload; | |
16 | struct page; | |
17 | struct block_device; | |
852c788f TH |
18 | struct io_context; |
19 | struct cgroup_subsys_state; | |
4246a0b6 | 20 | typedef void (bio_end_io_t) (struct bio *); |
7cc01581 | 21 | |
2a842aca CH |
22 | /* |
23 | * Block error status values. See block/blk-core:blk_errors for the details. | |
6e2fb221 | 24 | * Alpha cannot write a byte atomically, so we need to use 32-bit value. |
2a842aca | 25 | */ |
6e2fb221 MP |
26 | #if defined(CONFIG_ALPHA) && !defined(__alpha_bwx__) |
27 | typedef u32 __bitwise blk_status_t; | |
28 | #else | |
2a842aca | 29 | typedef u8 __bitwise blk_status_t; |
6e2fb221 | 30 | #endif |
2a842aca CH |
31 | #define BLK_STS_OK 0 |
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) | |
42 | ||
4e4cbee9 CH |
43 | /* hack for device mapper, don't use elsewhere: */ |
44 | #define BLK_STS_DM_REQUEUE ((__force blk_status_t)11) | |
45 | ||
03a07c92 GR |
46 | #define BLK_STS_AGAIN ((__force blk_status_t)12) |
47 | ||
86ff7c2a ML |
48 | /* |
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. | |
56 | * | |
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. | |
63 | */ | |
64 | #define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13) | |
65 | ||
9111e568 KB |
66 | /** |
67 | * blk_path_error - returns true if error may be path related | |
68 | * @error: status the request was completed with | |
69 | * | |
70 | * Description: | |
71 | * This classifies block error status into non-retryable errors and ones | |
72 | * that may be successful if retried on a failover path. | |
73 | * | |
74 | * Return: | |
75 | * %false - retrying failover path will not help | |
76 | * %true - may succeed if retried | |
77 | */ | |
78 | static inline bool blk_path_error(blk_status_t error) | |
79 | { | |
80 | switch (error) { | |
81 | case BLK_STS_NOTSUPP: | |
82 | case BLK_STS_NOSPC: | |
83 | case BLK_STS_TARGET: | |
84 | case BLK_STS_NEXUS: | |
85 | case BLK_STS_MEDIUM: | |
86 | case BLK_STS_PROTECTION: | |
87 | return false; | |
88 | } | |
89 | ||
90 | /* Anything else could be a path failure, so should be retried */ | |
91 | return true; | |
92 | } | |
93 | ||
b9147dd1 SL |
94 | struct blk_issue_stat { |
95 | u64 stat; | |
96 | }; | |
97 | ||
5238dcf4 OS |
98 | /* |
99 | * From most significant bit: | |
100 | * 1 bit: reserved for other usage, see below | |
101 | * 12 bits: original size of bio | |
102 | * 51 bits: issue time of bio | |
103 | */ | |
104 | #define BIO_ISSUE_RES_BITS 1 | |
105 | #define BIO_ISSUE_SIZE_BITS 12 | |
106 | #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS) | |
107 | #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS) | |
108 | #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1) | |
109 | #define BIO_ISSUE_SIZE_MASK \ | |
110 | (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT) | |
111 | #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1)) | |
112 | ||
113 | /* Reserved bit for blk-throtl */ | |
114 | #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63) | |
115 | ||
116 | struct bio_issue { | |
117 | u64 value; | |
118 | }; | |
119 | ||
120 | static inline u64 __bio_issue_time(u64 time) | |
121 | { | |
122 | return time & BIO_ISSUE_TIME_MASK; | |
123 | } | |
124 | ||
125 | static inline u64 bio_issue_time(struct bio_issue *issue) | |
126 | { | |
127 | return __bio_issue_time(issue->value); | |
128 | } | |
129 | ||
130 | static inline sector_t bio_issue_size(struct bio_issue *issue) | |
131 | { | |
132 | return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT); | |
133 | } | |
134 | ||
135 | static inline void bio_issue_init(struct bio_issue *issue, | |
136 | sector_t size) | |
137 | { | |
138 | size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1; | |
139 | issue->value = ((issue->value & BIO_ISSUE_RES_MASK) | | |
140 | (ktime_get_ns() & BIO_ISSUE_TIME_MASK) | | |
141 | ((u64)size << BIO_ISSUE_SIZE_SHIFT)); | |
142 | } | |
143 | ||
7cc01581 TH |
144 | /* |
145 | * main unit of I/O for the block layer and lower layers (ie drivers and | |
146 | * stacking drivers) | |
147 | */ | |
148 | struct bio { | |
7cc01581 | 149 | struct bio *bi_next; /* request queue link */ |
74d46992 | 150 | struct gendisk *bi_disk; |
1eff9d32 JA |
151 | unsigned int bi_opf; /* bottom bits req flags, |
152 | * top bits REQ_OP. Use | |
153 | * accessors. | |
4e1b2d52 | 154 | */ |
dbde775c | 155 | unsigned short bi_flags; /* status, etc and bvec pool number */ |
43b62ce3 | 156 | unsigned short bi_ioprio; |
cb6934f8 | 157 | unsigned short bi_write_hint; |
111be883 SL |
158 | blk_status_t bi_status; |
159 | u8 bi_partno; | |
7cc01581 TH |
160 | |
161 | /* Number of segments in this BIO after | |
162 | * physical address coalescing is performed. | |
163 | */ | |
164 | unsigned int bi_phys_segments; | |
165 | ||
7cc01581 TH |
166 | /* |
167 | * To keep track of the max segment size, we account for the | |
168 | * sizes of the first and last mergeable segments in this bio. | |
169 | */ | |
170 | unsigned int bi_seg_front_size; | |
171 | unsigned int bi_seg_back_size; | |
172 | ||
111be883 | 173 | struct bvec_iter bi_iter; |
196d38bc | 174 | |
111be883 | 175 | atomic_t __bi_remaining; |
7cc01581 TH |
176 | bio_end_io_t *bi_end_io; |
177 | ||
178 | void *bi_private; | |
852c788f TH |
179 | #ifdef CONFIG_BLK_CGROUP |
180 | /* | |
181 | * Optional ioc and css associated with this bio. Put on bio | |
182 | * release. Read comment on top of bio_associate_current(). | |
183 | */ | |
184 | struct io_context *bi_ioc; | |
185 | struct cgroup_subsys_state *bi_css; | |
9e234eea SL |
186 | #ifdef CONFIG_BLK_DEV_THROTTLING_LOW |
187 | void *bi_cg_private; | |
5238dcf4 | 188 | struct bio_issue bi_issue; |
9e234eea | 189 | #endif |
852c788f | 190 | #endif |
180b2f95 | 191 | union { |
7cc01581 | 192 | #if defined(CONFIG_BLK_DEV_INTEGRITY) |
180b2f95 | 193 | struct bio_integrity_payload *bi_integrity; /* data integrity */ |
7cc01581 | 194 | #endif |
180b2f95 | 195 | }; |
7cc01581 | 196 | |
4f024f37 KO |
197 | unsigned short bi_vcnt; /* how many bio_vec's */ |
198 | ||
f44b48c7 KO |
199 | /* |
200 | * Everything starting with bi_max_vecs will be preserved by bio_reset() | |
201 | */ | |
202 | ||
4f024f37 | 203 | unsigned short bi_max_vecs; /* max bvl_vecs we can hold */ |
f44b48c7 | 204 | |
dac56212 | 205 | atomic_t __bi_cnt; /* pin count */ |
f44b48c7 KO |
206 | |
207 | struct bio_vec *bi_io_vec; /* the actual vec list */ | |
208 | ||
395c72a7 KO |
209 | struct bio_set *bi_pool; |
210 | ||
7cc01581 TH |
211 | /* |
212 | * We can inline a number of vecs at the end of the bio, to avoid | |
213 | * double allocations for a small number of bio_vecs. This member | |
214 | * MUST obviously be kept at the very end of the bio. | |
215 | */ | |
216 | struct bio_vec bi_inline_vecs[0]; | |
217 | }; | |
218 | ||
f44b48c7 KO |
219 | #define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs) |
220 | ||
7cc01581 TH |
221 | /* |
222 | * bio flags | |
223 | */ | |
b2dbe0a6 JA |
224 | #define BIO_SEG_VALID 1 /* bi_phys_segments valid */ |
225 | #define BIO_CLONED 2 /* doesn't own data */ | |
226 | #define BIO_BOUNCED 3 /* bio is a bounce bio */ | |
227 | #define BIO_USER_MAPPED 4 /* contains user pages */ | |
228 | #define BIO_NULL_MAPPED 5 /* contains invalid user pages */ | |
229 | #define BIO_QUIET 6 /* Make BIO Quiet */ | |
a3ad0a9d JK |
230 | #define BIO_CHAIN 7 /* chained bio, ->bi_remaining in effect */ |
231 | #define BIO_REFFED 8 /* bio has elevated ->bi_cnt */ | |
8d2bbd4c CH |
232 | #define BIO_THROTTLED 9 /* This bio has already been subjected to |
233 | * throttling rules. Don't do it again. */ | |
fbbaf700 N |
234 | #define BIO_TRACE_COMPLETION 10 /* bio_endio() should trace the final completion |
235 | * of this bio. */ | |
dbde775c | 236 | /* See BVEC_POOL_OFFSET below before adding new flags */ |
f44b48c7 | 237 | |
7cc01581 | 238 | /* |
ed996a52 CH |
239 | * We support 6 different bvec pools, the last one is magic in that it |
240 | * is backed by a mempool. | |
7cc01581 | 241 | */ |
ed996a52 CH |
242 | #define BVEC_POOL_NR 6 |
243 | #define BVEC_POOL_MAX (BVEC_POOL_NR - 1) | |
244 | ||
245 | /* | |
dbde775c | 246 | * Top 3 bits of bio flags indicate the pool the bvecs came from. We add |
ed996a52 CH |
247 | * 1 to the actual index so that 0 indicates that there are no bvecs to be |
248 | * freed. | |
249 | */ | |
dbde775c | 250 | #define BVEC_POOL_BITS (3) |
c0acf12a | 251 | #define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS) |
ed996a52 | 252 | #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET) |
dbde775c N |
253 | #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1) |
254 | # error "BVEC_POOL_BITS is too small" | |
255 | #endif | |
256 | ||
257 | /* | |
258 | * Flags starting here get preserved by bio_reset() - this includes | |
259 | * only BVEC_POOL_IDX() | |
260 | */ | |
261 | #define BIO_RESET_BITS BVEC_POOL_OFFSET | |
7cc01581 | 262 | |
9a95e4ef BVA |
263 | typedef __u32 __bitwise blk_mq_req_flags_t; |
264 | ||
7cc01581 | 265 | /* |
ef295ecf CH |
266 | * Operations and flags common to the bio and request structures. |
267 | * We use 8 bits for encoding the operation, and the remaining 24 for flags. | |
87374179 CH |
268 | * |
269 | * The least significant bit of the operation number indicates the data | |
270 | * transfer direction: | |
271 | * | |
272 | * - if the least significant bit is set transfers are TO the device | |
273 | * - if the least significant bit is not set transfers are FROM the device | |
274 | * | |
275 | * If a operation does not transfer data the least significant bit has no | |
276 | * meaning. | |
7cc01581 | 277 | */ |
ef295ecf CH |
278 | #define REQ_OP_BITS 8 |
279 | #define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1) | |
280 | #define REQ_FLAG_BITS 24 | |
281 | ||
282 | enum req_opf { | |
87374179 CH |
283 | /* read sectors from the device */ |
284 | REQ_OP_READ = 0, | |
285 | /* write sectors to the device */ | |
286 | REQ_OP_WRITE = 1, | |
287 | /* flush the volatile write cache */ | |
288 | REQ_OP_FLUSH = 2, | |
289 | /* discard sectors */ | |
290 | REQ_OP_DISCARD = 3, | |
291 | /* get zone information */ | |
292 | REQ_OP_ZONE_REPORT = 4, | |
293 | /* securely erase sectors */ | |
294 | REQ_OP_SECURE_ERASE = 5, | |
295 | /* seset a zone write pointer */ | |
296 | REQ_OP_ZONE_RESET = 6, | |
297 | /* write the same sector many times */ | |
298 | REQ_OP_WRITE_SAME = 7, | |
a6f0788e | 299 | /* write the zero filled sector many times */ |
1d62ac13 | 300 | REQ_OP_WRITE_ZEROES = 9, |
ef295ecf | 301 | |
aebf526b CH |
302 | /* SCSI passthrough using struct scsi_request */ |
303 | REQ_OP_SCSI_IN = 32, | |
304 | REQ_OP_SCSI_OUT = 33, | |
305 | /* Driver private requests */ | |
306 | REQ_OP_DRV_IN = 34, | |
307 | REQ_OP_DRV_OUT = 35, | |
308 | ||
ef295ecf CH |
309 | REQ_OP_LAST, |
310 | }; | |
311 | ||
312 | enum req_flag_bits { | |
313 | __REQ_FAILFAST_DEV = /* no driver retries of device errors */ | |
314 | REQ_OP_BITS, | |
7cc01581 TH |
315 | __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */ |
316 | __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */ | |
7cc01581 TH |
317 | __REQ_SYNC, /* request is sync (sync write or read) */ |
318 | __REQ_META, /* metadata io request */ | |
65299a3b | 319 | __REQ_PRIO, /* boost priority in cfq */ |
bd1c1c21 | 320 | __REQ_NOMERGE, /* don't touch this for merging */ |
a2b80967 | 321 | __REQ_IDLE, /* anticipate more IO after this one */ |
180b2f95 | 322 | __REQ_INTEGRITY, /* I/O includes block integrity payload */ |
8e4bf844 | 323 | __REQ_FUA, /* forced unit access */ |
28a8f0d3 | 324 | __REQ_PREFLUSH, /* request for cache flush */ |
188bd2b1 | 325 | __REQ_RAHEAD, /* read ahead, can fail anytime */ |
1d796d6a | 326 | __REQ_BACKGROUND, /* background IO */ |
8977f563 | 327 | __REQ_NOWAIT, /* Don't wait if request will block */ |
d928be9f CH |
328 | |
329 | /* command specific flags for REQ_OP_WRITE_ZEROES: */ | |
330 | __REQ_NOUNMAP, /* do not free blocks when zeroing */ | |
331 | ||
96222bcc CH |
332 | /* for driver use */ |
333 | __REQ_DRV, | |
334 | ||
7cc01581 TH |
335 | __REQ_NR_BITS, /* stops here */ |
336 | }; | |
337 | ||
5953316d JA |
338 | #define REQ_FAILFAST_DEV (1ULL << __REQ_FAILFAST_DEV) |
339 | #define REQ_FAILFAST_TRANSPORT (1ULL << __REQ_FAILFAST_TRANSPORT) | |
340 | #define REQ_FAILFAST_DRIVER (1ULL << __REQ_FAILFAST_DRIVER) | |
341 | #define REQ_SYNC (1ULL << __REQ_SYNC) | |
342 | #define REQ_META (1ULL << __REQ_META) | |
343 | #define REQ_PRIO (1ULL << __REQ_PRIO) | |
ef295ecf | 344 | #define REQ_NOMERGE (1ULL << __REQ_NOMERGE) |
a2b80967 | 345 | #define REQ_IDLE (1ULL << __REQ_IDLE) |
180b2f95 | 346 | #define REQ_INTEGRITY (1ULL << __REQ_INTEGRITY) |
ef295ecf CH |
347 | #define REQ_FUA (1ULL << __REQ_FUA) |
348 | #define REQ_PREFLUSH (1ULL << __REQ_PREFLUSH) | |
349 | #define REQ_RAHEAD (1ULL << __REQ_RAHEAD) | |
1d796d6a | 350 | #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND) |
8977f563 | 351 | #define REQ_NOWAIT (1ULL << __REQ_NOWAIT) |
7cc01581 | 352 | |
d928be9f CH |
353 | #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP) |
354 | ||
96222bcc | 355 | #define REQ_DRV (1ULL << __REQ_DRV) |
d928be9f | 356 | |
7cc01581 TH |
357 | #define REQ_FAILFAST_MASK \ |
358 | (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER) | |
7cc01581 | 359 | |
e2a60da7 | 360 | #define REQ_NOMERGE_FLAGS \ |
e8064021 | 361 | (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA) |
e2a60da7 | 362 | |
ef295ecf CH |
363 | #define bio_op(bio) \ |
364 | ((bio)->bi_opf & REQ_OP_MASK) | |
365 | #define req_op(req) \ | |
366 | ((req)->cmd_flags & REQ_OP_MASK) | |
7cc01581 | 367 | |
ef295ecf | 368 | /* obsolete, don't use in new code */ |
93c5bdf7 CH |
369 | static inline void bio_set_op_attrs(struct bio *bio, unsigned op, |
370 | unsigned op_flags) | |
371 | { | |
372 | bio->bi_opf = op | op_flags; | |
373 | } | |
c11f0c0b | 374 | |
87374179 CH |
375 | static inline bool op_is_write(unsigned int op) |
376 | { | |
377 | return (op & 1); | |
378 | } | |
379 | ||
f73f44eb CH |
380 | /* |
381 | * Check if the bio or request is one that needs special treatment in the | |
382 | * flush state machine. | |
383 | */ | |
384 | static inline bool op_is_flush(unsigned int op) | |
385 | { | |
386 | return op & (REQ_FUA | REQ_PREFLUSH); | |
387 | } | |
388 | ||
b685d3d6 CH |
389 | /* |
390 | * Reads are always treated as synchronous, as are requests with the FUA or | |
391 | * PREFLUSH flag. Other operations may be marked as synchronous using the | |
392 | * REQ_SYNC flag. | |
393 | */ | |
ef295ecf CH |
394 | static inline bool op_is_sync(unsigned int op) |
395 | { | |
b685d3d6 CH |
396 | return (op & REQ_OP_MASK) == REQ_OP_READ || |
397 | (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH)); | |
ef295ecf | 398 | } |
c11f0c0b | 399 | |
dece1635 | 400 | typedef unsigned int blk_qc_t; |
fd2d3326 JA |
401 | #define BLK_QC_T_NONE -1U |
402 | #define BLK_QC_T_SHIFT 16 | |
403 | #define BLK_QC_T_INTERNAL (1U << 31) | |
dece1635 JA |
404 | |
405 | static inline bool blk_qc_t_valid(blk_qc_t cookie) | |
406 | { | |
407 | return cookie != BLK_QC_T_NONE; | |
408 | } | |
409 | ||
fd2d3326 JA |
410 | static inline blk_qc_t blk_tag_to_qc_t(unsigned int tag, unsigned int queue_num, |
411 | bool internal) | |
dece1635 | 412 | { |
fd2d3326 JA |
413 | blk_qc_t ret = tag | (queue_num << BLK_QC_T_SHIFT); |
414 | ||
415 | if (internal) | |
416 | ret |= BLK_QC_T_INTERNAL; | |
417 | ||
418 | return ret; | |
dece1635 JA |
419 | } |
420 | ||
421 | static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie) | |
422 | { | |
fd2d3326 | 423 | return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT; |
dece1635 JA |
424 | } |
425 | ||
426 | static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie) | |
427 | { | |
e3a7a3bf | 428 | return cookie & ((1u << BLK_QC_T_SHIFT) - 1); |
dece1635 JA |
429 | } |
430 | ||
fd2d3326 JA |
431 | static inline bool blk_qc_t_is_internal(blk_qc_t cookie) |
432 | { | |
433 | return (cookie & BLK_QC_T_INTERNAL) != 0; | |
434 | } | |
435 | ||
cf43e6be | 436 | struct blk_rq_stat { |
eca8b53a | 437 | u64 mean; |
cf43e6be JA |
438 | u64 min; |
439 | u64 max; | |
eca8b53a | 440 | u32 nr_samples; |
cf43e6be | 441 | u64 batch; |
cf43e6be JA |
442 | }; |
443 | ||
7cc01581 | 444 | #endif /* __LINUX_BLK_TYPES_H */ |