blk-mq: add optional request->alloc_time_ns
[linux-2.6-block.git] / include / linux / blk_types.h
CommitLineData
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
13struct bio_set;
14struct bio;
15struct bio_integrity_payload;
16struct page;
17struct block_device;
852c788f
TH
18struct io_context;
19struct cgroup_subsys_state;
4246a0b6 20typedef 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__)
27typedef u32 __bitwise blk_status_t;
28#else
2a842aca 29typedef 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 */
78static 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
5238dcf4
OS
94/*
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
99 */
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))
108
109/* Reserved bit for blk-throtl */
110#define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
111
112struct bio_issue {
113 u64 value;
114};
115
116static inline u64 __bio_issue_time(u64 time)
117{
118 return time & BIO_ISSUE_TIME_MASK;
119}
120
121static inline u64 bio_issue_time(struct bio_issue *issue)
122{
123 return __bio_issue_time(issue->value);
124}
125
126static inline sector_t bio_issue_size(struct bio_issue *issue)
127{
128 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
129}
130
131static inline void bio_issue_init(struct bio_issue *issue,
132 sector_t size)
133{
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));
138}
139
7cc01581
TH
140/*
141 * main unit of I/O for the block layer and lower layers (ie drivers and
142 * stacking drivers)
143 */
144struct bio {
7cc01581 145 struct bio *bi_next; /* request queue link */
74d46992 146 struct gendisk *bi_disk;
1eff9d32
JA
147 unsigned int bi_opf; /* bottom bits req flags,
148 * top bits REQ_OP. Use
149 * accessors.
4e1b2d52 150 */
dbde775c 151 unsigned short bi_flags; /* status, etc and bvec pool number */
43b62ce3 152 unsigned short bi_ioprio;
cb6934f8 153 unsigned short bi_write_hint;
111be883
SL
154 blk_status_t bi_status;
155 u8 bi_partno;
7cc01581 156
111be883 157 struct bvec_iter bi_iter;
196d38bc 158
111be883 159 atomic_t __bi_remaining;
7cc01581
TH
160 bio_end_io_t *bi_end_io;
161
162 void *bi_private;
852c788f
TH
163#ifdef CONFIG_BLK_CGROUP
164 /*
db6638d7
DZ
165 * Represents the association of the css and request_queue for the bio.
166 * If a bio goes direct to device, it will not have a blkg as it will
167 * not have a request_queue associated with it. The reference is put
168 * on release of the bio.
852c788f 169 */
08e18eab 170 struct blkcg_gq *bi_blkg;
5238dcf4 171 struct bio_issue bi_issue;
852c788f 172#endif
180b2f95 173 union {
7cc01581 174#if defined(CONFIG_BLK_DEV_INTEGRITY)
180b2f95 175 struct bio_integrity_payload *bi_integrity; /* data integrity */
7cc01581 176#endif
180b2f95 177 };
7cc01581 178
4f024f37
KO
179 unsigned short bi_vcnt; /* how many bio_vec's */
180
f44b48c7
KO
181 /*
182 * Everything starting with bi_max_vecs will be preserved by bio_reset()
183 */
184
4f024f37 185 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
f44b48c7 186
dac56212 187 atomic_t __bi_cnt; /* pin count */
f44b48c7
KO
188
189 struct bio_vec *bi_io_vec; /* the actual vec list */
190
395c72a7
KO
191 struct bio_set *bi_pool;
192
7cc01581
TH
193 /*
194 * We can inline a number of vecs at the end of the bio, to avoid
195 * double allocations for a small number of bio_vecs. This member
196 * MUST obviously be kept at the very end of the bio.
197 */
198 struct bio_vec bi_inline_vecs[0];
199};
200
f44b48c7
KO
201#define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
202
7cc01581
TH
203/*
204 * bio flags
205 */
2b24e6f6
JT
206enum {
207 BIO_NO_PAGE_REF, /* don't put release vec pages */
2b24e6f6
JT
208 BIO_CLONED, /* doesn't own data */
209 BIO_BOUNCED, /* bio is a bounce bio */
210 BIO_USER_MAPPED, /* contains user pages */
211 BIO_NULL_MAPPED, /* contains invalid user pages */
b8e24a93 212 BIO_WORKINGSET, /* contains userspace workingset pages */
2b24e6f6
JT
213 BIO_QUIET, /* Make BIO Quiet */
214 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
215 BIO_REFFED, /* bio has elevated ->bi_cnt */
216 BIO_THROTTLED, /* This bio has already been subjected to
8d2bbd4c 217 * throttling rules. Don't do it again. */
2b24e6f6 218 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
fbbaf700 219 * of this bio. */
2b24e6f6
JT
220 BIO_QUEUE_ENTERED, /* can use blk_queue_enter_live() */
221 BIO_TRACKED, /* set if bio goes through the rq_qos path */
222 BIO_FLAG_LAST
223};
cd4a4ae4 224
dbde775c 225/* See BVEC_POOL_OFFSET below before adding new flags */
f44b48c7 226
7cc01581 227/*
ed996a52
CH
228 * We support 6 different bvec pools, the last one is magic in that it
229 * is backed by a mempool.
7cc01581 230 */
ed996a52
CH
231#define BVEC_POOL_NR 6
232#define BVEC_POOL_MAX (BVEC_POOL_NR - 1)
233
234/*
dbde775c 235 * Top 3 bits of bio flags indicate the pool the bvecs came from. We add
ed996a52
CH
236 * 1 to the actual index so that 0 indicates that there are no bvecs to be
237 * freed.
238 */
dbde775c 239#define BVEC_POOL_BITS (3)
c0acf12a 240#define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS)
ed996a52 241#define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET)
dbde775c
N
242#if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1)
243# error "BVEC_POOL_BITS is too small"
244#endif
245
246/*
247 * Flags starting here get preserved by bio_reset() - this includes
248 * only BVEC_POOL_IDX()
249 */
250#define BIO_RESET_BITS BVEC_POOL_OFFSET
7cc01581 251
9a95e4ef
BVA
252typedef __u32 __bitwise blk_mq_req_flags_t;
253
7cc01581 254/*
ef295ecf
CH
255 * Operations and flags common to the bio and request structures.
256 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
87374179
CH
257 *
258 * The least significant bit of the operation number indicates the data
259 * transfer direction:
260 *
261 * - if the least significant bit is set transfers are TO the device
262 * - if the least significant bit is not set transfers are FROM the device
263 *
264 * If a operation does not transfer data the least significant bit has no
265 * meaning.
7cc01581 266 */
ef295ecf
CH
267#define REQ_OP_BITS 8
268#define REQ_OP_MASK ((1 << REQ_OP_BITS) - 1)
269#define REQ_FLAG_BITS 24
270
271enum req_opf {
87374179
CH
272 /* read sectors from the device */
273 REQ_OP_READ = 0,
274 /* write sectors to the device */
275 REQ_OP_WRITE = 1,
276 /* flush the volatile write cache */
277 REQ_OP_FLUSH = 2,
278 /* discard sectors */
279 REQ_OP_DISCARD = 3,
87374179
CH
280 /* securely erase sectors */
281 REQ_OP_SECURE_ERASE = 5,
8367de2c 282 /* reset a zone write pointer */
87374179
CH
283 REQ_OP_ZONE_RESET = 6,
284 /* write the same sector many times */
285 REQ_OP_WRITE_SAME = 7,
e84e8f06
CK
286 /* reset all the zone present on the device */
287 REQ_OP_ZONE_RESET_ALL = 8,
a6f0788e 288 /* write the zero filled sector many times */
1d62ac13 289 REQ_OP_WRITE_ZEROES = 9,
ef295ecf 290
aebf526b
CH
291 /* SCSI passthrough using struct scsi_request */
292 REQ_OP_SCSI_IN = 32,
293 REQ_OP_SCSI_OUT = 33,
294 /* Driver private requests */
295 REQ_OP_DRV_IN = 34,
296 REQ_OP_DRV_OUT = 35,
297
ef295ecf
CH
298 REQ_OP_LAST,
299};
300
301enum req_flag_bits {
302 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
303 REQ_OP_BITS,
7cc01581
TH
304 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
305 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
7cc01581
TH
306 __REQ_SYNC, /* request is sync (sync write or read) */
307 __REQ_META, /* metadata io request */
65299a3b 308 __REQ_PRIO, /* boost priority in cfq */
bd1c1c21 309 __REQ_NOMERGE, /* don't touch this for merging */
a2b80967 310 __REQ_IDLE, /* anticipate more IO after this one */
180b2f95 311 __REQ_INTEGRITY, /* I/O includes block integrity payload */
8e4bf844 312 __REQ_FUA, /* forced unit access */
28a8f0d3 313 __REQ_PREFLUSH, /* request for cache flush */
188bd2b1 314 __REQ_RAHEAD, /* read ahead, can fail anytime */
1d796d6a 315 __REQ_BACKGROUND, /* background IO */
8977f563 316 __REQ_NOWAIT, /* Don't wait if request will block */
893a1c97 317 __REQ_NOWAIT_INLINE, /* Return would-block error inline */
d3f77dfd
TH
318 /*
319 * When a shared kthread needs to issue a bio for a cgroup, doing
320 * so synchronously can lead to priority inversions as the kthread
321 * can be trapped waiting for that cgroup. CGROUP_PUNT flag makes
322 * submit_bio() punt the actual issuing to a dedicated per-blkcg
323 * work item to avoid such priority inversions.
324 */
325 __REQ_CGROUP_PUNT,
d928be9f
CH
326
327 /* command specific flags for REQ_OP_WRITE_ZEROES: */
328 __REQ_NOUNMAP, /* do not free blocks when zeroing */
329
d1e36282
JA
330 __REQ_HIPRI,
331
96222bcc
CH
332 /* for driver use */
333 __REQ_DRV,
0d1e0c7c 334 __REQ_SWAP, /* swapping request. */
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)
893a1c97 352#define REQ_NOWAIT_INLINE (1ULL << __REQ_NOWAIT_INLINE)
d3f77dfd
TH
353#define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT)
354
d928be9f 355#define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP)
d1e36282 356#define REQ_HIPRI (1ULL << __REQ_HIPRI)
d928be9f 357
96222bcc 358#define REQ_DRV (1ULL << __REQ_DRV)
0d1e0c7c 359#define REQ_SWAP (1ULL << __REQ_SWAP)
d928be9f 360
7cc01581
TH
361#define REQ_FAILFAST_MASK \
362 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
7cc01581 363
e2a60da7 364#define REQ_NOMERGE_FLAGS \
e8064021 365 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
e2a60da7 366
dbae2c55
MC
367enum stat_group {
368 STAT_READ,
369 STAT_WRITE,
bdca3c87 370 STAT_DISCARD,
dbae2c55
MC
371
372 NR_STAT_GROUPS
373};
374
ef295ecf
CH
375#define bio_op(bio) \
376 ((bio)->bi_opf & REQ_OP_MASK)
377#define req_op(req) \
378 ((req)->cmd_flags & REQ_OP_MASK)
7cc01581 379
ef295ecf 380/* obsolete, don't use in new code */
93c5bdf7
CH
381static inline void bio_set_op_attrs(struct bio *bio, unsigned op,
382 unsigned op_flags)
383{
384 bio->bi_opf = op | op_flags;
385}
c11f0c0b 386
87374179
CH
387static inline bool op_is_write(unsigned int op)
388{
389 return (op & 1);
390}
391
f73f44eb
CH
392/*
393 * Check if the bio or request is one that needs special treatment in the
394 * flush state machine.
395 */
396static inline bool op_is_flush(unsigned int op)
397{
398 return op & (REQ_FUA | REQ_PREFLUSH);
399}
400
b685d3d6
CH
401/*
402 * Reads are always treated as synchronous, as are requests with the FUA or
403 * PREFLUSH flag. Other operations may be marked as synchronous using the
404 * REQ_SYNC flag.
405 */
ef295ecf
CH
406static inline bool op_is_sync(unsigned int op)
407{
b685d3d6
CH
408 return (op & REQ_OP_MASK) == REQ_OP_READ ||
409 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
ef295ecf 410}
c11f0c0b 411
bdca3c87
MC
412static inline bool op_is_discard(unsigned int op)
413{
414 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
415}
416
ddcf35d3
MC
417static inline int op_stat_group(unsigned int op)
418{
bdca3c87
MC
419 if (op_is_discard(op))
420 return STAT_DISCARD;
ddcf35d3
MC
421 return op_is_write(op);
422}
423
dece1635 424typedef unsigned int blk_qc_t;
fd2d3326 425#define BLK_QC_T_NONE -1U
893a1c97 426#define BLK_QC_T_EAGAIN -2U
fd2d3326
JA
427#define BLK_QC_T_SHIFT 16
428#define BLK_QC_T_INTERNAL (1U << 31)
dece1635
JA
429
430static inline bool blk_qc_t_valid(blk_qc_t cookie)
431{
893a1c97 432 return cookie != BLK_QC_T_NONE && cookie != BLK_QC_T_EAGAIN;
dece1635
JA
433}
434
dece1635
JA
435static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
436{
fd2d3326 437 return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
dece1635
JA
438}
439
440static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
441{
e3a7a3bf 442 return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
dece1635
JA
443}
444
fd2d3326
JA
445static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
446{
447 return (cookie & BLK_QC_T_INTERNAL) != 0;
448}
449
cf43e6be 450struct blk_rq_stat {
eca8b53a 451 u64 mean;
cf43e6be
JA
452 u64 min;
453 u64 max;
eca8b53a 454 u32 nr_samples;
cf43e6be 455 u64 batch;
cf43e6be
JA
456};
457
7cc01581 458#endif /* __LINUX_BLK_TYPES_H */