fs/btrfs/volumes.h

   1 /* SPDX-License-Identifier: GPL-2.0 */
   2 /*
   3  * Copyright (C) 2007 Oracle.  All rights reserved.
   4  */
   5
   6 #ifndef BTRFS_VOLUMES_H
   7 #define BTRFS_VOLUMES_H
   8
   9 #include <linux/bio.h>
  10 #include <linux/sort.h>
  11 #include <linux/btrfs.h>
  12 #include "async-thread.h"
  13
  14 #define BTRFS_MAX_DATA_CHUNK_SIZE       (10ULL * SZ_1G)
  15
  16 extern struct mutex uuid_mutex;
  17
  18 #define BTRFS_STRIPE_LEN        SZ_64K
  19
  20 struct buffer_head;
  21 struct btrfs_pending_bios {
  22         struct bio *head;
  23         struct bio *tail;
  24 };
  25
  26 /*
  27  * Use sequence counter to get consistent device stat data on
  28  * 32-bit processors.
  29  */
  30 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
  31 #include <linux/seqlock.h>
  32 #define __BTRFS_NEED_DEVICE_DATA_ORDERED
  33 #define btrfs_device_data_ordered_init(device)  \
  34         seqcount_init(&device->data_seqcount)
  35 #else
  36 #define btrfs_device_data_ordered_init(device) do { } while (0)
  37 #endif
  38
  39 #define BTRFS_DEV_STATE_WRITEABLE       (0)
  40 #define BTRFS_DEV_STATE_IN_FS_METADATA  (1)
  41 #define BTRFS_DEV_STATE_MISSING         (2)
  42 #define BTRFS_DEV_STATE_REPLACE_TGT     (3)
  43 #define BTRFS_DEV_STATE_FLUSH_SENT      (4)
  44
  45 struct btrfs_device {
  46         struct list_head dev_list; /* device_list_mutex */
  47         struct list_head dev_alloc_list; /* chunk mutex */
  48         struct list_head post_commit_list; /* chunk mutex */
  49         struct btrfs_fs_devices *fs_devices;
  50         struct btrfs_fs_info *fs_info;
  51
  52         struct rcu_string *name;
  53
  54         u64 generation;
  55
  56         spinlock_t io_lock ____cacheline_aligned;
  57         int running_pending;
  58         /* regular prio bios */
  59         struct btrfs_pending_bios pending_bios;
  60         /* sync bios */
  61         struct btrfs_pending_bios pending_sync_bios;
  62
  63         struct block_device *bdev;
  64
  65         /* the mode sent to blkdev_get */
  66         fmode_t mode;
  67
  68         unsigned long dev_state;
  69         blk_status_t last_flush_error;
  70         int flush_bio_sent;
  71
  72 #ifdef __BTRFS_NEED_DEVICE_DATA_ORDERED
  73         seqcount_t data_seqcount;
  74 #endif
  75
  76         /* the internal btrfs device id */
  77         u64 devid;
  78
  79         /* size of the device in memory */
  80         u64 total_bytes;
  81
  82         /* size of the device on disk */
  83         u64 disk_total_bytes;
  84
  85         /* bytes used */
  86         u64 bytes_used;
  87
  88         /* optimal io alignment for this device */
  89         u32 io_align;
  90
  91         /* optimal io width for this device */
  92         u32 io_width;
  93         /* type and info about this device */
  94         u64 type;
  95
  96         /* minimal io size for this device */
  97         u32 sector_size;
  98
  99         /* physical drive uuid (or lvm uuid) */
 100         u8 uuid[BTRFS_UUID_SIZE];
 101
 102         /*
 103          * size of the device on the current transaction
 104          *
 105          * This variant is update when committing the transaction,
 106          * and protected by chunk mutex
 107          */
 108         u64 commit_total_bytes;
 109
 110         /* bytes used on the current transaction */
 111         u64 commit_bytes_used;
 112
 113         /* for sending down flush barriers */
 114         struct bio *flush_bio;
 115         struct completion flush_wait;
 116
 117         /* per-device scrub information */
 118         struct scrub_ctx *scrub_ctx;
 119
 120         struct btrfs_work work;
 121
 122         /* readahead state */
 123         atomic_t reada_in_flight;
 124         u64 reada_next;
 125         struct reada_zone *reada_curr_zone;
 126         struct radix_tree_root reada_zones;
 127         struct radix_tree_root reada_extents;
 128
 129         /* disk I/O failure stats. For detailed description refer to
 130          * enum btrfs_dev_stat_values in ioctl.h */
 131         int dev_stats_valid;
 132
 133         /* Counter to record the change of device stats */
 134         atomic_t dev_stats_ccnt;
 135         atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
 136
 137         struct extent_io_tree alloc_state;
 138 };
 139
 140 /*
 141  * If we read those variants at the context of their own lock, we needn't
 142  * use the following helpers, reading them directly is safe.
 143  */
 144 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
 145 #define BTRFS_DEVICE_GETSET_FUNCS(name)                                 \
 146 static inline u64                                                       \
 147 btrfs_device_get_##name(const struct btrfs_device *dev)                 \
 148 {                                                                       \
 149         u64 size;                                                       \
 150         unsigned int seq;                                               \
 151                                                                         \
 152         do {                                                            \
 153                 seq = read_seqcount_begin(&dev->data_seqcount);         \
 154                 size = dev->name;                                       \
 155         } while (read_seqcount_retry(&dev->data_seqcount, seq));        \
 156         return size;                                                    \
 157 }                                                                       \
 158                                                                         \
 159 static inline void                                                      \
 160 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)             \
 161 {                                                                       \
 162         preempt_disable();                                              \
 163         write_seqcount_begin(&dev->data_seqcount);                      \
 164         dev->name = size;                                               \
 165         write_seqcount_end(&dev->data_seqcount);                        \
 166         preempt_enable();                                               \
 167 }
 168 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
 169 #define BTRFS_DEVICE_GETSET_FUNCS(name)                                 \
 170 static inline u64                                                       \
 171 btrfs_device_get_##name(const struct btrfs_device *dev)                 \
 172 {                                                                       \
 173         u64 size;                                                       \
 174                                                                         \
 175         preempt_disable();                                              \
 176         size = dev->name;                                               \
 177         preempt_enable();                                               \
 178         return size;                                                    \
 179 }                                                                       \
 180                                                                         \
 181 static inline void                                                      \
 182 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)             \
 183 {                                                                       \
 184         preempt_disable();                                              \
 185         dev->name = size;                                               \
 186         preempt_enable();                                               \
 187 }
 188 #else
 189 #define BTRFS_DEVICE_GETSET_FUNCS(name)                                 \
 190 static inline u64                                                       \
 191 btrfs_device_get_##name(const struct btrfs_device *dev)                 \
 192 {                                                                       \
 193         return dev->name;                                               \
 194 }                                                                       \
 195                                                                         \
 196 static inline void                                                      \
 197 btrfs_device_set_##name(struct btrfs_device *dev, u64 size)             \
 198 {                                                                       \
 199         dev->name = size;                                               \
 200 }
 201 #endif
 202
 203 BTRFS_DEVICE_GETSET_FUNCS(total_bytes);
 204 BTRFS_DEVICE_GETSET_FUNCS(disk_total_bytes);
 205 BTRFS_DEVICE_GETSET_FUNCS(bytes_used);
 206
 207 struct btrfs_fs_devices {
 208         u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
 209         u8 metadata_uuid[BTRFS_FSID_SIZE];
 210         bool fsid_change;
 211         struct list_head fs_list;
 212
 213         u64 num_devices;
 214         u64 open_devices;
 215         u64 rw_devices;
 216         u64 missing_devices;
 217         u64 total_rw_bytes;
 218         u64 total_devices;
 219
 220         /* Highest generation number of seen devices */
 221         u64 latest_generation;
 222
 223         struct block_device *latest_bdev;
 224
 225         /* all of the devices in the FS, protected by a mutex
 226          * so we can safely walk it to write out the supers without
 227          * worrying about add/remove by the multi-device code.
 228          * Scrubbing super can kick off supers writing by holding
 229          * this mutex lock.
 230          */
 231         struct mutex device_list_mutex;
 232
 233         /* List of all devices, protected by device_list_mutex */
 234         struct list_head devices;
 235
 236         /*
 237          * Devices which can satisfy space allocation. Protected by
 238          * chunk_mutex
 239          */
 240         struct list_head alloc_list;
 241
 242         struct btrfs_fs_devices *seed;
 243         int seeding;
 244
 245         int opened;
 246
 247         /* set when we find or add a device that doesn't have the
 248          * nonrot flag set
 249          */
 250         int rotating;
 251
 252         struct btrfs_fs_info *fs_info;
 253         /* sysfs kobjects */
 254         struct kobject fsid_kobj;
 255         struct kobject *device_dir_kobj;
 256         struct completion kobj_unregister;
 257 };
 258
 259 #define BTRFS_BIO_INLINE_CSUM_SIZE      64
 260
 261 #define BTRFS_MAX_DEVS(info) ((BTRFS_MAX_ITEM_SIZE(info)        \
 262                         - sizeof(struct btrfs_chunk))           \
 263                         / sizeof(struct btrfs_stripe) + 1)
 264
 265 #define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE        \
 266                                 - 2 * sizeof(struct btrfs_disk_key)     \
 267                                 - 2 * sizeof(struct btrfs_chunk))       \
 268                                 / sizeof(struct btrfs_stripe) + 1)
 269
 270 /*
 271  * we need the mirror number and stripe index to be passed around
 272  * the call chain while we are processing end_io (especially errors).
 273  * Really, what we need is a btrfs_bio structure that has this info
 274  * and is properly sized with its stripe array, but we're not there
 275  * quite yet.  We have our own btrfs bioset, and all of the bios
 276  * we allocate are actually btrfs_io_bios.  We'll cram as much of
 277  * struct btrfs_bio as we can into this over time.
 278  */
 279 struct btrfs_io_bio {
 280         unsigned int mirror_num;
 281         unsigned int stripe_index;
 282         u64 logical;
 283         u8 *csum;
 284         u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
 285         struct bvec_iter iter;
 286         /*
 287          * This member must come last, bio_alloc_bioset will allocate enough
 288          * bytes for entire btrfs_io_bio but relies on bio being last.
 289          */
 290         struct bio bio;
 291 };
 292
 293 static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
 294 {
 295         return container_of(bio, struct btrfs_io_bio, bio);
 296 }
 297
 298 static inline void btrfs_io_bio_free_csum(struct btrfs_io_bio *io_bio)
 299 {
 300         if (io_bio->csum != io_bio->csum_inline) {
 301                 kfree(io_bio->csum);
 302                 io_bio->csum = NULL;
 303         }
 304 }
 305
 306 struct btrfs_bio_stripe {
 307         struct btrfs_device *dev;
 308         u64 physical;
 309         u64 length; /* only used for discard mappings */
 310 };
 311
 312 struct btrfs_bio {
 313         refcount_t refs;
 314         atomic_t stripes_pending;
 315         struct btrfs_fs_info *fs_info;
 316         u64 map_type; /* get from map_lookup->type */
 317         bio_end_io_t *end_io;
 318         struct bio *orig_bio;
 319         unsigned long flags;
 320         void *private;
 321         atomic_t error;
 322         int max_errors;
 323         int num_stripes;
 324         int mirror_num;
 325         int num_tgtdevs;
 326         int *tgtdev_map;
 327         /*
 328          * logical block numbers for the start of each stripe
 329          * The last one or two are p/q.  These are sorted,
 330          * so raid_map[0] is the start of our full stripe
 331          */
 332         u64 *raid_map;
 333         struct btrfs_bio_stripe stripes[];
 334 };
 335
 336 struct btrfs_device_info {
 337         struct btrfs_device *dev;
 338         u64 dev_offset;
 339         u64 max_avail;
 340         u64 total_avail;
 341 };
 342
 343 struct btrfs_raid_attr {
 344         u8 sub_stripes;         /* sub_stripes info for map */
 345         u8 dev_stripes;         /* stripes per dev */
 346         u8 devs_max;            /* max devs to use */
 347         u8 devs_min;            /* min devs needed */
 348         u8 tolerated_failures;  /* max tolerated fail devs */
 349         u8 devs_increment;      /* ndevs has to be a multiple of this */
 350         u8 ncopies;             /* how many copies to data has */
 351         u8 nparity;             /* number of stripes worth of bytes to store
 352                                  * parity information */
 353         u8 mindev_error;        /* error code if min devs requisite is unmet */
 354         const char raid_name[8]; /* name of the raid */
 355         u64 bg_flag;            /* block group flag of the raid */
 356 };
 357
 358 extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];
 359
 360 struct map_lookup {
 361         u64 type;
 362         int io_align;
 363         int io_width;
 364         u64 stripe_len;
 365         int num_stripes;
 366         int sub_stripes;
 367         int verified_stripes; /* For mount time dev extent verification */
 368         struct btrfs_bio_stripe stripes[];
 369 };
 370
 371 #define map_lookup_size(n) (sizeof(struct map_lookup) + \
 372                             (sizeof(struct btrfs_bio_stripe) * (n)))
 373
 374 struct btrfs_balance_args;
 375 struct btrfs_balance_progress;
 376 struct btrfs_balance_control {
 377         struct btrfs_balance_args data;
 378         struct btrfs_balance_args meta;
 379         struct btrfs_balance_args sys;
 380
 381         u64 flags;
 382
 383         struct btrfs_balance_progress stat;
 384 };
 385
 386 enum btrfs_map_op {
 387         BTRFS_MAP_READ,
 388         BTRFS_MAP_WRITE,
 389         BTRFS_MAP_DISCARD,
 390         BTRFS_MAP_GET_READ_MIRRORS,
 391 };
 392
 393 static inline enum btrfs_map_op btrfs_op(struct bio *bio)
 394 {
 395         switch (bio_op(bio)) {
 396         case REQ_OP_DISCARD:
 397                 return BTRFS_MAP_DISCARD;
 398         case REQ_OP_WRITE:
 399                 return BTRFS_MAP_WRITE;
 400         default:
 401                 WARN_ON_ONCE(1);
 402                 /* fall through */
 403         case REQ_OP_READ:
 404                 return BTRFS_MAP_READ;
 405         }
 406 }
 407
 408 void btrfs_get_bbio(struct btrfs_bio *bbio);
 409 void btrfs_put_bbio(struct btrfs_bio *bbio);
 410 int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
 411                     u64 logical, u64 *length,
 412                     struct btrfs_bio **bbio_ret, int mirror_num);
 413 int btrfs_map_sblock(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
 414                      u64 logical, u64 *length,
 415                      struct btrfs_bio **bbio_ret);
 416 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
 417                      u64 physical, u64 **logical, int *naddrs, int *stripe_len);
 418 int btrfs_read_sys_array(struct btrfs_fs_info *fs_info);
 419 int btrfs_read_chunk_tree(struct btrfs_fs_info *fs_info);
 420 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, u64 type);
 421 void btrfs_mapping_tree_free(struct extent_map_tree *tree);
 422 blk_status_t btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
 423                            int mirror_num, int async_submit);
 424 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 425                        fmode_t flags, void *holder);
 426 struct btrfs_device *btrfs_scan_one_device(const char *path,
 427                                            fmode_t flags, void *holder);
 428 int btrfs_forget_devices(const char *path);
 429 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
 430 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices, int step);
 431 void btrfs_assign_next_active_device(struct btrfs_device *device,
 432                                      struct btrfs_device *this_dev);
 433 struct btrfs_device *btrfs_find_device_by_devspec(struct btrfs_fs_info *fs_info,
 434                                                   u64 devid,
 435                                                   const char *devpath);
 436 struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
 437                                         const u64 *devid,
 438                                         const u8 *uuid);
 439 void btrfs_free_device(struct btrfs_device *device);
 440 int btrfs_rm_device(struct btrfs_fs_info *fs_info,
 441                     const char *device_path, u64 devid);
 442 void __exit btrfs_cleanup_fs_uuids(void);
 443 int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
 444 int btrfs_grow_device(struct btrfs_trans_handle *trans,
 445                       struct btrfs_device *device, u64 new_size);
 446 struct btrfs_device *btrfs_find_device(struct btrfs_fs_devices *fs_devices,
 447                                        u64 devid, u8 *uuid, u8 *fsid, bool seed);
 448 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
 449 int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *path);
 450 int btrfs_balance(struct btrfs_fs_info *fs_info,
 451                   struct btrfs_balance_control *bctl,
 452                   struct btrfs_ioctl_balance_args *bargs);
 453 void btrfs_describe_block_groups(u64 flags, char *buf, u32 size_buf);
 454 int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
 455 int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
 456 int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
 457 int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
 458 int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
 459 int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
 460 int btrfs_chunk_readonly(struct btrfs_fs_info *fs_info, u64 chunk_offset);
 461 int find_free_dev_extent_start(struct btrfs_device *device, u64 num_bytes,
 462                                u64 search_start, u64 *start, u64 *max_avail);
 463 int find_free_dev_extent(struct btrfs_device *device, u64 num_bytes,
 464                          u64 *start, u64 *max_avail);
 465 void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
 466 int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info,
 467                         struct btrfs_ioctl_get_dev_stats *stats);
 468 void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
 469 int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
 470 int btrfs_run_dev_stats(struct btrfs_trans_handle *trans);
 471 void btrfs_rm_dev_replace_remove_srcdev(struct btrfs_device *srcdev);
 472 void btrfs_rm_dev_replace_free_srcdev(struct btrfs_device *srcdev);
 473 void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev);
 474 void btrfs_scratch_superblocks(struct block_device *bdev, const char *device_path);
 475 int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
 476                            u64 logical, u64 len);
 477 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
 478                                     u64 logical);
 479 int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
 480                              u64 chunk_offset, u64 chunk_size);
 481 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
 482 struct extent_map *btrfs_get_chunk_map(struct btrfs_fs_info *fs_info,
 483                                        u64 logical, u64 length);
 484
 485 static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
 486                                       int index)
 487 {
 488         atomic_inc(dev->dev_stat_values + index);
 489         /*
 490          * This memory barrier orders stores updating statistics before stores
 491          * updating dev_stats_ccnt.
 492          *
 493          * It pairs with smp_rmb() in btrfs_run_dev_stats().
 494          */
 495         smp_mb__before_atomic();
 496         atomic_inc(&dev->dev_stats_ccnt);
 497 }
 498
 499 static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
 500                                       int index)
 501 {
 502         return atomic_read(dev->dev_stat_values + index);
 503 }
 504
 505 static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
 506                                                 int index)
 507 {
 508         int ret;
 509
 510         ret = atomic_xchg(dev->dev_stat_values + index, 0);
 511         /*
 512          * atomic_xchg implies a full memory barriers as per atomic_t.txt:
 513          * - RMW operations that have a return value are fully ordered;
 514          *
 515          * This implicit memory barriers is paired with the smp_rmb in
 516          * btrfs_run_dev_stats
 517          */
 518         atomic_inc(&dev->dev_stats_ccnt);
 519         return ret;
 520 }
 521
 522 static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
 523                                       int index, unsigned long val)
 524 {
 525         atomic_set(dev->dev_stat_values + index, val);
 526         /*
 527          * This memory barrier orders stores updating statistics before stores
 528          * updating dev_stats_ccnt.
 529          *
 530          * It pairs with smp_rmb() in btrfs_run_dev_stats().
 531          */
 532         smp_mb__before_atomic();
 533         atomic_inc(&dev->dev_stats_ccnt);
 534 }
 535
 536 static inline void btrfs_dev_stat_reset(struct btrfs_device *dev,
 537                                         int index)
 538 {
 539         btrfs_dev_stat_set(dev, index, 0);
 540 }
 541
 542 /*
 543  * Convert block group flags (BTRFS_BLOCK_GROUP_*) to btrfs_raid_types, which
 544  * can be used as index to access btrfs_raid_array[].
 545  */
 546 static inline enum btrfs_raid_types btrfs_bg_flags_to_raid_index(u64 flags)
 547 {
 548         if (flags & BTRFS_BLOCK_GROUP_RAID10)
 549                 return BTRFS_RAID_RAID10;
 550         else if (flags & BTRFS_BLOCK_GROUP_RAID1)
 551                 return BTRFS_RAID_RAID1;
 552         else if (flags & BTRFS_BLOCK_GROUP_DUP)
 553                 return BTRFS_RAID_DUP;
 554         else if (flags & BTRFS_BLOCK_GROUP_RAID0)
 555                 return BTRFS_RAID_RAID0;
 556         else if (flags & BTRFS_BLOCK_GROUP_RAID5)
 557                 return BTRFS_RAID_RAID5;
 558         else if (flags & BTRFS_BLOCK_GROUP_RAID6)
 559                 return BTRFS_RAID_RAID6;
 560
 561         return BTRFS_RAID_SINGLE; /* BTRFS_BLOCK_GROUP_SINGLE */
 562 }
 563
 564 void btrfs_commit_device_sizes(struct btrfs_transaction *trans);
 565
 566 struct list_head *btrfs_get_fs_uuids(void);
 567 void btrfs_set_fs_info_ptr(struct btrfs_fs_info *fs_info);
 568 void btrfs_reset_fs_info_ptr(struct btrfs_fs_info *fs_info);
 569 bool btrfs_check_rw_degradable(struct btrfs_fs_info *fs_info,
 570                                         struct btrfs_device *failing_dev);
 571
 572 int btrfs_bg_type_to_factor(u64 flags);
 573 const char *btrfs_bg_type_to_raid_name(u64 flags);
 574 int btrfs_verify_dev_extents(struct btrfs_fs_info *fs_info);
 575
 576 #endif