1/* SPDX-License-Identifier: GPL-2.0 */
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
9#include <linux/types.h>
10#include <linux/bvec.h>
11#include <linux/device.h>
12#include <linux/ktime.h>
13#include <linux/rw_hint.h>
14
15struct bio_set;
16struct bio;
17struct bio_integrity_payload;
18struct page;
19struct io_context;
20struct cgroup_subsys_state;
21typedef void (bio_end_io_t) (struct bio *);
22struct bio_crypt_ctx;
23
24/*
25 * The basic unit of block I/O is a sector. It is used in a number of contexts
26 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9
27 * bytes. Variables of type sector_t represent an offset or size that is a
28 * multiple of 512 bytes. Hence these two constants.
29 */
30#ifndef SECTOR_SHIFT
31#define SECTOR_SHIFT 9
32#endif
33#ifndef SECTOR_SIZE
34#define SECTOR_SIZE (1 << SECTOR_SHIFT)
35#endif
36
37#define PAGE_SECTORS_SHIFT (PAGE_SHIFT - SECTOR_SHIFT)
38#define PAGE_SECTORS (1 << PAGE_SECTORS_SHIFT)
39#define SECTOR_MASK (PAGE_SECTORS - 1)
40
41struct block_device {
42 sector_t bd_start_sect;
43 sector_t bd_nr_sectors;
44 struct gendisk * bd_disk;
45 struct request_queue * bd_queue;
46 struct disk_stats __percpu *bd_stats;
47 unsigned long bd_stamp;
48 atomic_t __bd_flags; // partition number + flags
49#define BD_PARTNO 255 // lower 8 bits; assign-once
50#define BD_READ_ONLY (1u<<8) // read-only policy
51#define BD_WRITE_HOLDER (1u<<9)
52#define BD_HAS_SUBMIT_BIO (1u<<10)
53#define BD_RO_WARNED (1u<<11)
54#ifdef CONFIG_FAIL_MAKE_REQUEST
55#define BD_MAKE_IT_FAIL (1u<<12)
56#endif
57 dev_t bd_dev;
58 struct address_space *bd_mapping; /* page cache */
59
60 atomic_t bd_openers;
61 spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
62 void * bd_claiming;
63 void * bd_holder;
64 const struct blk_holder_ops *bd_holder_ops;
65 struct mutex bd_holder_lock;
66 int bd_holders;
67 struct kobject *bd_holder_dir;
68
69 atomic_t bd_fsfreeze_count; /* number of freeze requests */
70 struct mutex bd_fsfreeze_mutex; /* serialize freeze/thaw */
71
72 struct partition_meta_info *bd_meta_info;
73 int bd_writers;
74#ifdef CONFIG_SECURITY
75 void *bd_security;
76#endif
77 /*
78 * keep this out-of-line as it's both big and not needed in the fast
79 * path
80 */
81 struct device bd_device;
82} __randomize_layout;
83
84#define bdev_whole(_bdev) \
85 ((_bdev)->bd_disk->part0)
86
87#define dev_to_bdev(device) \
88 container_of((device), struct block_device, bd_device)
89
90#define bdev_kobj(_bdev) \
91 (&((_bdev)->bd_device.kobj))
92
93/*
94 * Block error status values. See block/blk-core:blk_errors for the details.
95 */
96typedef u8 __bitwise blk_status_t;
97typedef u16 blk_short_t;
98#define BLK_STS_OK 0
99#define BLK_STS_NOTSUPP ((__force blk_status_t)1)
100#define BLK_STS_TIMEOUT ((__force blk_status_t)2)
101#define BLK_STS_NOSPC ((__force blk_status_t)3)
102#define BLK_STS_TRANSPORT ((__force blk_status_t)4)
103#define BLK_STS_TARGET ((__force blk_status_t)5)
104#define BLK_STS_RESV_CONFLICT ((__force blk_status_t)6)
105#define BLK_STS_MEDIUM ((__force blk_status_t)7)
106#define BLK_STS_PROTECTION ((__force blk_status_t)8)
107#define BLK_STS_RESOURCE ((__force blk_status_t)9)
108#define BLK_STS_IOERR ((__force blk_status_t)10)
109
110/* hack for device mapper, don't use elsewhere: */
111#define BLK_STS_DM_REQUEUE ((__force blk_status_t)11)
112
113/*
114 * BLK_STS_AGAIN should only be returned if RQF_NOWAIT is set
115 * and the bio would block (cf bio_wouldblock_error())
116 */
117#define BLK_STS_AGAIN ((__force blk_status_t)12)
118
119/*
120 * BLK_STS_DEV_RESOURCE is returned from the driver to the block layer if
121 * device related resources are unavailable, but the driver can guarantee
122 * that the queue will be rerun in the future once resources become
123 * available again. This is typically the case for device specific
124 * resources that are consumed for IO. If the driver fails allocating these
125 * resources, we know that inflight (or pending) IO will free these
126 * resource upon completion.
127 *
128 * This is different from BLK_STS_RESOURCE in that it explicitly references
129 * a device specific resource. For resources of wider scope, allocation
130 * failure can happen without having pending IO. This means that we can't
131 * rely on request completions freeing these resources, as IO may not be in
132 * flight. Examples of that are kernel memory allocations, DMA mappings, or
133 * any other system wide resources.
134 */
135#define BLK_STS_DEV_RESOURCE ((__force blk_status_t)13)
136
137/*
138 * BLK_STS_ZONE_OPEN_RESOURCE is returned from the driver in the completion
139 * path if the device returns a status indicating that too many zone resources
140 * are currently open. The same command should be successful if resubmitted
141 * after the number of open zones decreases below the device's limits, which is
142 * reported in the request_queue's max_open_zones.
143 */
144#define BLK_STS_ZONE_OPEN_RESOURCE ((__force blk_status_t)14)
145
146/*
147 * BLK_STS_ZONE_ACTIVE_RESOURCE is returned from the driver in the completion
148 * path if the device returns a status indicating that too many zone resources
149 * are currently active. The same command should be successful if resubmitted
150 * after the number of active zones decreases below the device's limits, which
151 * is reported in the request_queue's max_active_zones.
152 */
153#define BLK_STS_ZONE_ACTIVE_RESOURCE ((__force blk_status_t)15)
154
155/*
156 * BLK_STS_OFFLINE is returned from the driver when the target device is offline
157 * or is being taken offline. This could help differentiate the case where a
158 * device is intentionally being shut down from a real I/O error.
159 */
160#define BLK_STS_OFFLINE ((__force blk_status_t)16)
161
162/*
163 * BLK_STS_DURATION_LIMIT is returned from the driver when the target device
164 * aborted the command because it exceeded one of its Command Duration Limits.
165 */
166#define BLK_STS_DURATION_LIMIT ((__force blk_status_t)17)
167
168/*
169 * Invalid size or alignment.
170 */
171#define BLK_STS_INVAL ((__force blk_status_t)19)
172
173/**
174 * blk_path_error - returns true if error may be path related
175 * @error: status the request was completed with
176 *
177 * Description:
178 * This classifies block error status into non-retryable errors and ones
179 * that may be successful if retried on a failover path.
180 *
181 * Return:
182 * %false - retrying failover path will not help
183 * %true - may succeed if retried
184 */
185static inline bool blk_path_error(blk_status_t error)
186{
187 switch (error) {
188 case BLK_STS_NOTSUPP:
189 case BLK_STS_NOSPC:
190 case BLK_STS_TARGET:
191 case BLK_STS_RESV_CONFLICT:
192 case BLK_STS_MEDIUM:
193 case BLK_STS_PROTECTION:
194 return false;
195 }
196
197 /* Anything else could be a path failure, so should be retried */
198 return true;
199}
200
201typedef __u32 __bitwise blk_opf_t;
202
203typedef unsigned int blk_qc_t;
204#define BLK_QC_T_NONE -1U
205
206/*
207 * main unit of I/O for the block layer and lower layers (ie drivers and
208 * stacking drivers)
209 */
210struct bio {
211 struct bio *bi_next; /* request queue link */
212 struct block_device *bi_bdev;
213 blk_opf_t bi_opf; /* bottom bits REQ_OP, top bits
214 * req_flags.
215 */
216 unsigned short bi_flags; /* BIO_* below */
217 unsigned short bi_ioprio;
218 enum rw_hint bi_write_hint;
219 u8 bi_write_stream;
220 blk_status_t bi_status;
221 atomic_t __bi_remaining;
222
223 struct bvec_iter bi_iter;
224
225 union {
226 /* for polled bios: */
227 blk_qc_t bi_cookie;
228 /* for plugged zoned writes only: */
229 unsigned int __bi_nr_segments;
230 };
231 bio_end_io_t *bi_end_io;
232 void *bi_private;
233#ifdef CONFIG_BLK_CGROUP
234 /*
235 * Represents the association of the css and request_queue for the bio.
236 * If a bio goes direct to device, it will not have a blkg as it will
237 * not have a request_queue associated with it. The reference is put
238 * on release of the bio.
239 */
240 struct blkcg_gq *bi_blkg;
241 /* Time that this bio was issued. */
242 u64 issue_time_ns;
243#ifdef CONFIG_BLK_CGROUP_IOCOST
244 u64 bi_iocost_cost;
245#endif
246#endif
247
248#ifdef CONFIG_BLK_INLINE_ENCRYPTION
249 struct bio_crypt_ctx *bi_crypt_context;
250#endif
251
252#if defined(CONFIG_BLK_DEV_INTEGRITY)
253 struct bio_integrity_payload *bi_integrity; /* data integrity */
254#endif
255
256 unsigned short bi_vcnt; /* how many bio_vec's */
257
258 /*
259 * Everything starting with bi_max_vecs will be preserved by bio_reset()
260 */
261
262 unsigned short bi_max_vecs; /* max bvl_vecs we can hold */
263
264 atomic_t __bi_cnt; /* pin count */
265
266 struct bio_vec *bi_io_vec; /* the actual vec list */
267
268 struct bio_set *bi_pool;
269};
270
271#define BIO_RESET_BYTES offsetof(struct bio, bi_max_vecs)
272#define BIO_MAX_SECTORS (UINT_MAX >> SECTOR_SHIFT)
273
274static inline struct bio_vec *bio_inline_vecs(struct bio *bio)
275{
276 return (struct bio_vec *)(bio + 1);
277}
278
279/*
280 * bio flags
281 */
282enum {
283 BIO_PAGE_PINNED, /* Unpin pages in bio_release_pages() */
284 BIO_CLONED, /* doesn't own data */
285 BIO_QUIET, /* Make BIO Quiet */
286 BIO_CHAIN, /* chained bio, ->bi_remaining in effect */
287 BIO_REFFED, /* bio has elevated ->bi_cnt */
288 BIO_BPS_THROTTLED, /* This bio has already been subjected to
289 * throttling rules. Don't do it again. */
290 BIO_TRACE_COMPLETION, /* bio_endio() should trace the final completion
291 * of this bio. */
292 BIO_CGROUP_ACCT, /* has been accounted to a cgroup */
293 BIO_QOS_THROTTLED, /* bio went through rq_qos throttle path */
294 /*
295 * This bio has completed bps throttling at the single tg granularity,
296 * which is different from BIO_BPS_THROTTLED. When the bio is enqueued
297 * into the sq->queued of the upper tg, or is about to be dispatched,
298 * this flag needs to be cleared. Since blk-throttle and rq_qos are not
299 * on the same hierarchical level, reuse the value.
300 */
301 BIO_TG_BPS_THROTTLED = BIO_QOS_THROTTLED,
302 BIO_QOS_MERGED, /* but went through rq_qos merge path */
303 BIO_REMAPPED,
304 BIO_ZONE_WRITE_PLUGGING, /* bio handled through zone write plugging */
305 BIO_EMULATES_ZONE_APPEND, /* bio emulates a zone append operation */
306 BIO_FLAG_LAST
307};
308
309typedef __u32 __bitwise blk_mq_req_flags_t;
310
311#define REQ_OP_BITS 8
312#define REQ_OP_MASK (__force blk_opf_t)((1 << REQ_OP_BITS) - 1)
313#define REQ_FLAG_BITS 24
314
315/**
316 * enum req_op - Operations common to the bio and request structures.
317 * We use 8 bits for encoding the operation, and the remaining 24 for flags.
318 *
319 * The least significant bit of the operation number indicates the data
320 * transfer direction:
321 *
322 * - if the least significant bit is set transfers are TO the device
323 * - if the least significant bit is not set transfers are FROM the device
324 *
325 * If a operation does not transfer data the least significant bit has no
326 * meaning.
327 */
328enum req_op {
329 /* read sectors from the device */
330 REQ_OP_READ = (__force blk_opf_t)0,
331 /* write sectors to the device */
332 REQ_OP_WRITE = (__force blk_opf_t)1,
333 /* flush the volatile write cache */
334 REQ_OP_FLUSH = (__force blk_opf_t)2,
335 /* discard sectors */
336 REQ_OP_DISCARD = (__force blk_opf_t)3,
337 /* securely erase sectors */
338 REQ_OP_SECURE_ERASE = (__force blk_opf_t)5,
339 /* write data at the current zone write pointer */
340 REQ_OP_ZONE_APPEND = (__force blk_opf_t)7,
341 /* write the zero filled sector many times */
342 REQ_OP_WRITE_ZEROES = (__force blk_opf_t)9,
343 /* Open a zone */
344 REQ_OP_ZONE_OPEN = (__force blk_opf_t)10,
345 /* Close a zone */
346 REQ_OP_ZONE_CLOSE = (__force blk_opf_t)11,
347 /* Transition a zone to full */
348 REQ_OP_ZONE_FINISH = (__force blk_opf_t)13,
349 /* reset a zone write pointer */
350 REQ_OP_ZONE_RESET = (__force blk_opf_t)15,
351 /* reset all the zone present on the device */
352 REQ_OP_ZONE_RESET_ALL = (__force blk_opf_t)17,
353
354 /* Driver private requests */
355 REQ_OP_DRV_IN = (__force blk_opf_t)34,
356 REQ_OP_DRV_OUT = (__force blk_opf_t)35,
357
358 REQ_OP_LAST = (__force blk_opf_t)36,
359};
360
361/* Keep cmd_flag_name[] in sync with the definitions below */
362enum req_flag_bits {
363 __REQ_FAILFAST_DEV = /* no driver retries of device errors */
364 REQ_OP_BITS,
365 __REQ_FAILFAST_TRANSPORT, /* no driver retries of transport errors */
366 __REQ_FAILFAST_DRIVER, /* no driver retries of driver errors */
367 __REQ_SYNC, /* request is sync (sync write or read) */
368 __REQ_META, /* metadata io request */
369 __REQ_PRIO, /* boost priority in cfq */
370 __REQ_NOMERGE, /* don't touch this for merging */
371 __REQ_IDLE, /* anticipate more IO after this one */
372 __REQ_INTEGRITY, /* I/O includes block integrity payload */
373 __REQ_FUA, /* forced unit access */
374 __REQ_PREFLUSH, /* request for cache flush */
375 __REQ_RAHEAD, /* read ahead, can fail anytime */
376 __REQ_BACKGROUND, /* background IO */
377 __REQ_NOWAIT, /* Don't wait if request will block */
378 __REQ_POLLED, /* caller polls for completion using bio_poll */
379 __REQ_ALLOC_CACHE, /* allocate IO from cache if available */
380 __REQ_SWAP, /* swap I/O */
381 __REQ_DRV, /* for driver use */
382 __REQ_FS_PRIVATE, /* for file system (submitter) use */
383 __REQ_ATOMIC, /* for atomic write operations */
384 __REQ_P2PDMA, /* contains P2P DMA pages */
385 /*
386 * Command specific flags, keep last:
387 */
388 /* for REQ_OP_WRITE_ZEROES: */
389 __REQ_NOUNMAP, /* do not free blocks when zeroing */
390
391 __REQ_NR_BITS, /* stops here */
392};
393
394#define REQ_FAILFAST_DEV \
395 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DEV)
396#define REQ_FAILFAST_TRANSPORT \
397 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_TRANSPORT)
398#define REQ_FAILFAST_DRIVER \
399 (__force blk_opf_t)(1ULL << __REQ_FAILFAST_DRIVER)
400#define REQ_SYNC (__force blk_opf_t)(1ULL << __REQ_SYNC)
401#define REQ_META (__force blk_opf_t)(1ULL << __REQ_META)
402#define REQ_PRIO (__force blk_opf_t)(1ULL << __REQ_PRIO)
403#define REQ_NOMERGE (__force blk_opf_t)(1ULL << __REQ_NOMERGE)
404#define REQ_IDLE (__force blk_opf_t)(1ULL << __REQ_IDLE)
405#define REQ_INTEGRITY (__force blk_opf_t)(1ULL << __REQ_INTEGRITY)
406#define REQ_FUA (__force blk_opf_t)(1ULL << __REQ_FUA)
407#define REQ_PREFLUSH (__force blk_opf_t)(1ULL << __REQ_PREFLUSH)
408#define REQ_RAHEAD (__force blk_opf_t)(1ULL << __REQ_RAHEAD)
409#define REQ_BACKGROUND (__force blk_opf_t)(1ULL << __REQ_BACKGROUND)
410#define REQ_NOWAIT (__force blk_opf_t)(1ULL << __REQ_NOWAIT)
411#define REQ_POLLED (__force blk_opf_t)(1ULL << __REQ_POLLED)
412#define REQ_ALLOC_CACHE (__force blk_opf_t)(1ULL << __REQ_ALLOC_CACHE)
413#define REQ_SWAP (__force blk_opf_t)(1ULL << __REQ_SWAP)
414#define REQ_DRV (__force blk_opf_t)(1ULL << __REQ_DRV)
415#define REQ_FS_PRIVATE (__force blk_opf_t)(1ULL << __REQ_FS_PRIVATE)
416#define REQ_ATOMIC (__force blk_opf_t)(1ULL << __REQ_ATOMIC)
417#define REQ_P2PDMA (__force blk_opf_t)(1ULL << __REQ_P2PDMA)
418
419#define REQ_NOUNMAP (__force blk_opf_t)(1ULL << __REQ_NOUNMAP)
420
421#define REQ_FAILFAST_MASK \
422 (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT | REQ_FAILFAST_DRIVER)
423
424#define REQ_NOMERGE_FLAGS \
425 (REQ_NOMERGE | REQ_PREFLUSH | REQ_FUA)
426
427enum stat_group {
428 STAT_READ,
429 STAT_WRITE,
430 STAT_DISCARD,
431 STAT_FLUSH,
432
433 NR_STAT_GROUPS
434};
435
436static inline enum req_op bio_op(const struct bio *bio)
437{
438 return bio->bi_opf & REQ_OP_MASK;
439}
440
441static inline bool op_is_write(blk_opf_t op)
442{
443 return !!(op & (__force blk_opf_t)1);
444}
445
446/*
447 * Check if the bio or request is one that needs special treatment in the
448 * flush state machine.
449 */
450static inline bool op_is_flush(blk_opf_t op)
451{
452 return op & (REQ_FUA | REQ_PREFLUSH);
453}
454
455/*
456 * Reads are always treated as synchronous, as are requests with the FUA or
457 * PREFLUSH flag. Other operations may be marked as synchronous using the
458 * REQ_SYNC flag.
459 */
460static inline bool op_is_sync(blk_opf_t op)
461{
462 return (op & REQ_OP_MASK) == REQ_OP_READ ||
463 (op & (REQ_SYNC | REQ_FUA | REQ_PREFLUSH));
464}
465
466static inline bool op_is_discard(blk_opf_t op)
467{
468 return (op & REQ_OP_MASK) == REQ_OP_DISCARD;
469}
470
471/*
472 * Check if a bio or request operation is a zone management operation, with
473 * the exception of REQ_OP_ZONE_RESET_ALL which is treated as a special case
474 * due to its different handling in the block layer and device response in
475 * case of command failure.
476 */
477static inline bool op_is_zone_mgmt(enum req_op op)
478{
479 switch (op & REQ_OP_MASK) {
480 case REQ_OP_ZONE_RESET:
481 case REQ_OP_ZONE_OPEN:
482 case REQ_OP_ZONE_CLOSE:
483 case REQ_OP_ZONE_FINISH:
484 return true;
485 default:
486 return false;
487 }
488}
489
490static inline int op_stat_group(enum req_op op)
491{
492 if (op_is_discard(op))
493 return STAT_DISCARD;
494 return op_is_write(op);
495}
496
497struct blk_rq_stat {
498 u64 mean;
499 u64 min;
500 u64 max;
501 u32 nr_samples;
502 u64 batch;
503};
504
505#endif /* __LINUX_BLK_TYPES_H */
506