1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (c) 2000-2005 Silicon Graphics, Inc. 4 * All Rights Reserved. 5 */ 6 #ifndef __XFS_MOUNT_H__ 7 #define __XFS_MOUNT_H__ 8 9 struct xlog; 10 struct xfs_inode; 11 struct xfs_mru_cache; 12 struct xfs_ail; 13 struct xfs_quotainfo; 14 struct xfs_da_geometry; 15 16 /* dynamic preallocation free space thresholds, 5% down to 1% */ 17 enum { 18 XFS_LOWSP_1_PCNT = 0, 19 XFS_LOWSP_2_PCNT, 20 XFS_LOWSP_3_PCNT, 21 XFS_LOWSP_4_PCNT, 22 XFS_LOWSP_5_PCNT, 23 XFS_LOWSP_MAX, 24 }; 25 26 /* 27 * Error Configuration 28 * 29 * Error classes define the subsystem the configuration belongs to. 30 * Error numbers define the errors that are configurable. 31 */ 32 enum { 33 XFS_ERR_METADATA, 34 XFS_ERR_CLASS_MAX, 35 }; 36 enum { 37 XFS_ERR_DEFAULT, 38 XFS_ERR_EIO, 39 XFS_ERR_ENOSPC, 40 XFS_ERR_ENODEV, 41 XFS_ERR_ERRNO_MAX, 42 }; 43 44 #define XFS_ERR_RETRY_FOREVER -1 45 46 /* 47 * Although retry_timeout is in jiffies which is normally an unsigned long, 48 * we limit the retry timeout to 86400 seconds, or one day. So even a 49 * signed 32-bit long is sufficient for a HZ value up to 24855. Making it 50 * signed lets us store the special "-1" value, meaning retry forever. 51 */ 52 struct xfs_error_cfg { 53 struct xfs_kobj kobj; 54 int max_retries; 55 long retry_timeout; /* in jiffies, -1 = infinite */ 56 }; 57 58 /* 59 * The struct xfsmount layout is optimised to separate read-mostly variables 60 * from variables that are frequently modified. We put the read-mostly variables 61 * first, then place all the other variables at the end. 62 * 63 * Typically, read-mostly variables are those that are set at mount time and 64 * never changed again, or only change rarely as a result of things like sysfs 65 * knobs being tweaked. 66 */ 67 typedef struct xfs_mount { 68 struct xfs_sb m_sb; /* copy of fs superblock */ 69 struct super_block *m_super; 70 struct xfs_ail *m_ail; /* fs active log item list */ 71 struct xfs_buf *m_sb_bp; /* buffer for superblock */ 72 char *m_rtname; /* realtime device name */ 73 char *m_logname; /* external log device name */ 74 struct xfs_da_geometry *m_dir_geo; /* directory block geometry */ 75 struct xfs_da_geometry *m_attr_geo; /* attribute block geometry */ 76 struct xlog *m_log; /* log specific stuff */ 77 struct xfs_inode *m_rbmip; /* pointer to bitmap inode */ 78 struct xfs_inode *m_rsumip; /* pointer to summary inode */ 79 struct xfs_inode *m_rootip; /* pointer to root directory */ 80 struct xfs_quotainfo *m_quotainfo; /* disk quota information */ 81 xfs_buftarg_t *m_ddev_targp; /* saves taking the address */ 82 xfs_buftarg_t *m_logdev_targp;/* ptr to log device */ 83 xfs_buftarg_t *m_rtdev_targp; /* ptr to rt device */ 84 /* 85 * Optional cache of rt summary level per bitmap block with the 86 * invariant that m_rsum_cache[bbno] <= the minimum i for which 87 * rsum[i][bbno] != 0. Reads and writes are serialized by the rsumip 88 * inode lock. 89 */ 90 uint8_t *m_rsum_cache; 91 struct xfs_mru_cache *m_filestream; /* per-mount filestream data */ 92 struct workqueue_struct *m_buf_workqueue; 93 struct workqueue_struct *m_unwritten_workqueue; 94 struct workqueue_struct *m_cil_workqueue; 95 struct workqueue_struct *m_reclaim_workqueue; 96 struct workqueue_struct *m_gc_workqueue; 97 struct workqueue_struct *m_sync_workqueue; 98 99 int m_bsize; /* fs logical block size */ 100 uint8_t m_blkbit_log; /* blocklog + NBBY */ 101 uint8_t m_blkbb_log; /* blocklog - BBSHIFT */ 102 uint8_t m_agno_log; /* log #ag's */ 103 uint8_t m_sectbb_log; /* sectlog - BBSHIFT */ 104 uint m_blockmask; /* sb_blocksize-1 */ 105 uint m_blockwsize; /* sb_blocksize in words */ 106 uint m_blockwmask; /* blockwsize-1 */ 107 uint m_alloc_mxr[2]; /* max alloc btree records */ 108 uint m_alloc_mnr[2]; /* min alloc btree records */ 109 uint m_bmap_dmxr[2]; /* max bmap btree records */ 110 uint m_bmap_dmnr[2]; /* min bmap btree records */ 111 uint m_rmap_mxr[2]; /* max rmap btree records */ 112 uint m_rmap_mnr[2]; /* min rmap btree records */ 113 uint m_refc_mxr[2]; /* max refc btree records */ 114 uint m_refc_mnr[2]; /* min refc btree records */ 115 uint m_ag_maxlevels; /* XFS_AG_MAXLEVELS */ 116 uint m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */ 117 uint m_rmap_maxlevels; /* max rmap btree levels */ 118 uint m_refc_maxlevels; /* max refcount btree level */ 119 xfs_extlen_t m_ag_prealloc_blocks; /* reserved ag blocks */ 120 uint m_alloc_set_aside; /* space we can't use */ 121 uint m_ag_max_usable; /* max space per AG */ 122 int m_dalign; /* stripe unit */ 123 int m_swidth; /* stripe width */ 124 xfs_agnumber_t m_maxagi; /* highest inode alloc group */ 125 uint m_allocsize_log;/* min write size log bytes */ 126 uint m_allocsize_blocks; /* min write size blocks */ 127 int m_logbufs; /* number of log buffers */ 128 int m_logbsize; /* size of each log buffer */ 129 uint m_rsumlevels; /* rt summary levels */ 130 uint m_rsumsize; /* size of rt summary, bytes */ 131 int m_fixedfsid[2]; /* unchanged for life of FS */ 132 uint m_qflags; /* quota status flags */ 133 uint64_t m_flags; /* global mount flags */ 134 int64_t m_low_space[XFS_LOWSP_MAX]; 135 struct xfs_ino_geometry m_ino_geo; /* inode geometry */ 136 struct xfs_trans_resv m_resv; /* precomputed res values */ 137 /* low free space thresholds */ 138 bool m_always_cow; 139 bool m_fail_unmount; 140 bool m_finobt_nores; /* no per-AG finobt resv. */ 141 bool m_update_sb; /* sb needs update in mount */ 142 143 /* 144 * Bitsets of per-fs metadata that have been checked and/or are sick. 145 * Callers must hold m_sb_lock to access these two fields. 146 */ 147 uint8_t m_fs_checked; 148 uint8_t m_fs_sick; 149 /* 150 * Bitsets of rt metadata that have been checked and/or are sick. 151 * Callers must hold m_sb_lock to access this field. 152 */ 153 uint8_t m_rt_checked; 154 uint8_t m_rt_sick; 155 156 /* 157 * End of read-mostly variables. Frequently written variables and locks 158 * should be placed below this comment from now on. The first variable 159 * here is marked as cacheline aligned so they it is separated from 160 * the read-mostly variables. 161 */ 162 163 spinlock_t ____cacheline_aligned m_sb_lock; /* sb counter lock */ 164 struct percpu_counter m_icount; /* allocated inodes counter */ 165 struct percpu_counter m_ifree; /* free inodes counter */ 166 struct percpu_counter m_fdblocks; /* free block counter */ 167 /* 168 * Count of data device blocks reserved for delayed allocations, 169 * including indlen blocks. Does not include allocated CoW staging 170 * extents or anything related to the rt device. 171 */ 172 struct percpu_counter m_delalloc_blks; 173 /* 174 * Global count of allocation btree blocks in use across all AGs. Only 175 * used when perag reservation is enabled. Helps prevent block 176 * reservation from attempting to reserve allocation btree blocks. 177 */ 178 atomic64_t m_allocbt_blks; 179 180 struct radix_tree_root m_perag_tree; /* per-ag accounting info */ 181 spinlock_t m_perag_lock; /* lock for m_perag_tree */ 182 uint64_t m_resblks; /* total reserved blocks */ 183 uint64_t m_resblks_avail;/* available reserved blocks */ 184 uint64_t m_resblks_save; /* reserved blks @ remount,ro */ 185 struct delayed_work m_reclaim_work; /* background inode reclaim */ 186 struct xfs_kobj m_kobj; 187 struct xfs_kobj m_error_kobj; 188 struct xfs_kobj m_error_meta_kobj; 189 struct xfs_error_cfg m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX]; 190 struct xstats m_stats; /* per-fs stats */ 191 xfs_agnumber_t m_agfrotor; /* last ag where space found */ 192 xfs_agnumber_t m_agirotor; /* last ag dir inode alloced */ 193 spinlock_t m_agirotor_lock;/* .. and lock protecting it */ 194 195 /* 196 * Workqueue item so that we can coalesce multiple inode flush attempts 197 * into a single flush. 198 */ 199 struct work_struct m_flush_inodes_work; 200 201 /* 202 * Generation of the filesysyem layout. This is incremented by each 203 * growfs, and used by the pNFS server to ensure the client updates 204 * its view of the block device once it gets a layout that might 205 * reference the newly added blocks. Does not need to be persistent 206 * as long as we only allow file system size increments, but if we 207 * ever support shrinks it would have to be persisted in addition 208 * to various other kinds of pain inflicted on the pNFS server. 209 */ 210 uint32_t m_generation; 211 struct mutex m_growlock; /* growfs mutex */ 212 213 #ifdef DEBUG 214 /* 215 * Frequency with which errors are injected. Replaces xfs_etest; the 216 * value stored in here is the inverse of the frequency with which the 217 * error triggers. 1 = always, 2 = half the time, etc. 218 */ 219 unsigned int *m_errortag; 220 struct xfs_kobj m_errortag_kobj; 221 #endif 222 } xfs_mount_t; 223 224 #define M_IGEO(mp) (&(mp)->m_ino_geo) 225 226 /* 227 * Flags for m_flags. 228 */ 229 #define XFS_MOUNT_WSYNC (1ULL << 0) /* for nfs - all metadata ops 230 must be synchronous except 231 for space allocations */ 232 #define XFS_MOUNT_UNMOUNTING (1ULL << 1) /* filesystem is unmounting */ 233 #define XFS_MOUNT_WAS_CLEAN (1ULL << 3) 234 #define XFS_MOUNT_FS_SHUTDOWN (1ULL << 4) /* atomic stop of all filesystem 235 operations, typically for 236 disk errors in metadata */ 237 #define XFS_MOUNT_DISCARD (1ULL << 5) /* discard unused blocks */ 238 #define XFS_MOUNT_NOALIGN (1ULL << 7) /* turn off stripe alignment 239 allocations */ 240 #define XFS_MOUNT_ATTR2 (1ULL << 8) /* allow use of attr2 format */ 241 #define XFS_MOUNT_GRPID (1ULL << 9) /* group-ID assigned from directory */ 242 #define XFS_MOUNT_NORECOVERY (1ULL << 10) /* no recovery - dirty fs */ 243 #define XFS_MOUNT_ALLOCSIZE (1ULL << 12) /* specified allocation size */ 244 #define XFS_MOUNT_SMALL_INUMS (1ULL << 14) /* user wants 32bit inodes */ 245 #define XFS_MOUNT_32BITINODES (1ULL << 15) /* inode32 allocator active */ 246 #define XFS_MOUNT_NOUUID (1ULL << 16) /* ignore uuid during mount */ 247 #define XFS_MOUNT_IKEEP (1ULL << 18) /* keep empty inode clusters*/ 248 #define XFS_MOUNT_SWALLOC (1ULL << 19) /* turn on stripe width 249 * allocation */ 250 #define XFS_MOUNT_RDONLY (1ULL << 20) /* read-only fs */ 251 #define XFS_MOUNT_DIRSYNC (1ULL << 21) /* synchronous directory ops */ 252 #define XFS_MOUNT_LARGEIO (1ULL << 22) /* report large preferred 253 * I/O size in stat() */ 254 #define XFS_MOUNT_FILESTREAMS (1ULL << 24) /* enable the filestreams 255 allocator */ 256 #define XFS_MOUNT_NOATTR2 (1ULL << 25) /* disable use of attr2 format */ 257 #define XFS_MOUNT_DAX_ALWAYS (1ULL << 26) 258 #define XFS_MOUNT_DAX_NEVER (1ULL << 27) 259 260 /* 261 * Max and min values for mount-option defined I/O 262 * preallocation sizes. 263 */ 264 #define XFS_MAX_IO_LOG 30 /* 1G */ 265 #define XFS_MIN_IO_LOG PAGE_SHIFT 266 267 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp) \ 268 ((mp)->m_flags & XFS_MOUNT_WAS_CLEAN) 269 #define XFS_FORCED_SHUTDOWN(mp) ((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN) 270 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname, 271 int lnnum); 272 #define xfs_force_shutdown(m,f) \ 273 xfs_do_force_shutdown(m, f, __FILE__, __LINE__) 274 275 #define SHUTDOWN_META_IO_ERROR 0x0001 /* write attempt to metadata failed */ 276 #define SHUTDOWN_LOG_IO_ERROR 0x0002 /* write attempt to the log failed */ 277 #define SHUTDOWN_FORCE_UMOUNT 0x0004 /* shutdown from a forced unmount */ 278 #define SHUTDOWN_CORRUPT_INCORE 0x0008 /* corrupt in-memory data structures */ 279 280 /* 281 * Flags for xfs_mountfs 282 */ 283 #define XFS_MFSI_QUIET 0x40 /* Be silent if mount errors found */ 284 285 static inline xfs_agnumber_t 286 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d) 287 { 288 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d); 289 do_div(ld, mp->m_sb.sb_agblocks); 290 return (xfs_agnumber_t) ld; 291 } 292 293 static inline xfs_agblock_t 294 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d) 295 { 296 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d); 297 return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks); 298 } 299 300 /* per-AG block reservation data structures*/ 301 struct xfs_ag_resv { 302 /* number of blocks originally reserved here */ 303 xfs_extlen_t ar_orig_reserved; 304 /* number of blocks reserved here */ 305 xfs_extlen_t ar_reserved; 306 /* number of blocks originally asked for */ 307 xfs_extlen_t ar_asked; 308 }; 309 310 /* 311 * Per-ag incore structure, copies of information in agf and agi, to improve the 312 * performance of allocation group selection. 313 */ 314 typedef struct xfs_perag { 315 struct xfs_mount *pag_mount; /* owner filesystem */ 316 xfs_agnumber_t pag_agno; /* AG this structure belongs to */ 317 atomic_t pag_ref; /* perag reference count */ 318 char pagf_init; /* this agf's entry is initialized */ 319 char pagi_init; /* this agi's entry is initialized */ 320 char pagf_metadata; /* the agf is preferred to be metadata */ 321 char pagi_inodeok; /* The agi is ok for inodes */ 322 uint8_t pagf_levels[XFS_BTNUM_AGF]; 323 /* # of levels in bno & cnt btree */ 324 bool pagf_agflreset; /* agfl requires reset before use */ 325 uint32_t pagf_flcount; /* count of blocks in freelist */ 326 xfs_extlen_t pagf_freeblks; /* total free blocks */ 327 xfs_extlen_t pagf_longest; /* longest free space */ 328 uint32_t pagf_btreeblks; /* # of blocks held in AGF btrees */ 329 xfs_agino_t pagi_freecount; /* number of free inodes */ 330 xfs_agino_t pagi_count; /* number of allocated inodes */ 331 332 /* 333 * Inode allocation search lookup optimisation. 334 * If the pagino matches, the search for new inodes 335 * doesn't need to search the near ones again straight away 336 */ 337 xfs_agino_t pagl_pagino; 338 xfs_agino_t pagl_leftrec; 339 xfs_agino_t pagl_rightrec; 340 341 /* 342 * Bitsets of per-ag metadata that have been checked and/or are sick. 343 * Callers should hold pag_state_lock before accessing this field. 344 */ 345 uint16_t pag_checked; 346 uint16_t pag_sick; 347 spinlock_t pag_state_lock; 348 349 spinlock_t pagb_lock; /* lock for pagb_tree */ 350 struct rb_root pagb_tree; /* ordered tree of busy extents */ 351 unsigned int pagb_gen; /* generation count for pagb_tree */ 352 wait_queue_head_t pagb_wait; /* woken when pagb_gen changes */ 353 354 atomic_t pagf_fstrms; /* # of filestreams active in this AG */ 355 356 spinlock_t pag_ici_lock; /* incore inode cache lock */ 357 struct radix_tree_root pag_ici_root; /* incore inode cache root */ 358 int pag_ici_reclaimable; /* reclaimable inodes */ 359 unsigned long pag_ici_reclaim_cursor; /* reclaim restart point */ 360 361 /* buffer cache index */ 362 spinlock_t pag_buf_lock; /* lock for pag_buf_hash */ 363 struct rhashtable pag_buf_hash; 364 365 /* for rcu-safe freeing */ 366 struct rcu_head rcu_head; 367 int pagb_count; /* pagb slots in use */ 368 369 /* Blocks reserved for all kinds of metadata. */ 370 struct xfs_ag_resv pag_meta_resv; 371 /* Blocks reserved for the reverse mapping btree. */ 372 struct xfs_ag_resv pag_rmapbt_resv; 373 374 /* background prealloc block trimming */ 375 struct delayed_work pag_blockgc_work; 376 377 /* reference count */ 378 uint8_t pagf_refcount_level; 379 380 /* 381 * Unlinked inode information. This incore information reflects 382 * data stored in the AGI, so callers must hold the AGI buffer lock 383 * or have some other means to control concurrency. 384 */ 385 struct rhashtable pagi_unlinked_hash; 386 } xfs_perag_t; 387 388 static inline struct xfs_ag_resv * 389 xfs_perag_resv( 390 struct xfs_perag *pag, 391 enum xfs_ag_resv_type type) 392 { 393 switch (type) { 394 case XFS_AG_RESV_METADATA: 395 return &pag->pag_meta_resv; 396 case XFS_AG_RESV_RMAPBT: 397 return &pag->pag_rmapbt_resv; 398 default: 399 return NULL; 400 } 401 } 402 403 int xfs_buf_hash_init(xfs_perag_t *pag); 404 void xfs_buf_hash_destroy(xfs_perag_t *pag); 405 406 extern void xfs_uuid_table_free(void); 407 extern uint64_t xfs_default_resblks(xfs_mount_t *mp); 408 extern int xfs_mountfs(xfs_mount_t *mp); 409 extern int xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount, 410 xfs_agnumber_t *maxagi); 411 extern void xfs_unmountfs(xfs_mount_t *); 412 413 extern int xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta, 414 bool reserved); 415 extern int xfs_mod_frextents(struct xfs_mount *mp, int64_t delta); 416 417 extern int xfs_readsb(xfs_mount_t *, int); 418 extern void xfs_freesb(xfs_mount_t *); 419 extern bool xfs_fs_writable(struct xfs_mount *mp, int level); 420 extern int xfs_sb_validate_fsb_count(struct xfs_sb *, uint64_t); 421 422 extern int xfs_dev_is_read_only(struct xfs_mount *, char *); 423 424 extern void xfs_set_low_space_thresholds(struct xfs_mount *); 425 426 int xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb, 427 xfs_off_t count_fsb); 428 429 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp, 430 int error_class, int error); 431 void xfs_force_summary_recalc(struct xfs_mount *mp); 432 void xfs_mod_delalloc(struct xfs_mount *mp, int64_t delta); 433 434 #endif /* __XFS_MOUNT_H__ */ 435