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_nameops; 13 struct xfs_ail; 14 struct xfs_quotainfo; 15 struct xfs_dir_ops; 16 struct xfs_da_geometry; 17 18 /* dynamic preallocation free space thresholds, 5% down to 1% */ 19 enum { 20 XFS_LOWSP_1_PCNT = 0, 21 XFS_LOWSP_2_PCNT, 22 XFS_LOWSP_3_PCNT, 23 XFS_LOWSP_4_PCNT, 24 XFS_LOWSP_5_PCNT, 25 XFS_LOWSP_MAX, 26 }; 27 28 /* 29 * Error Configuration 30 * 31 * Error classes define the subsystem the configuration belongs to. 32 * Error numbers define the errors that are configurable. 33 */ 34 enum { 35 XFS_ERR_METADATA, 36 XFS_ERR_CLASS_MAX, 37 }; 38 enum { 39 XFS_ERR_DEFAULT, 40 XFS_ERR_EIO, 41 XFS_ERR_ENOSPC, 42 XFS_ERR_ENODEV, 43 XFS_ERR_ERRNO_MAX, 44 }; 45 46 #define XFS_ERR_RETRY_FOREVER -1 47 48 /* 49 * Although retry_timeout is in jiffies which is normally an unsigned long, 50 * we limit the retry timeout to 86400 seconds, or one day. So even a 51 * signed 32-bit long is sufficient for a HZ value up to 24855. Making it 52 * signed lets us store the special "-1" value, meaning retry forever. 53 */ 54 struct xfs_error_cfg { 55 struct xfs_kobj kobj; 56 int max_retries; 57 long retry_timeout; /* in jiffies, -1 = infinite */ 58 }; 59 60 typedef struct xfs_mount { 61 struct super_block *m_super; 62 xfs_tid_t m_tid; /* next unused tid for fs */ 63 struct xfs_ail *m_ail; /* fs active log item list */ 64 65 struct xfs_sb m_sb; /* copy of fs superblock */ 66 spinlock_t m_sb_lock; /* sb counter lock */ 67 struct percpu_counter m_icount; /* allocated inodes counter */ 68 struct percpu_counter m_ifree; /* free inodes counter */ 69 struct percpu_counter m_fdblocks; /* free block counter */ 70 71 struct xfs_buf *m_sb_bp; /* buffer for superblock */ 72 char *m_fsname; /* filesystem name */ 73 int m_fsname_len; /* strlen of fs name */ 74 char *m_rtname; /* realtime device name */ 75 char *m_logname; /* external log device name */ 76 int m_bsize; /* fs logical block size */ 77 xfs_agnumber_t m_agfrotor; /* last ag where space found */ 78 xfs_agnumber_t m_agirotor; /* last ag dir inode alloced */ 79 spinlock_t m_agirotor_lock;/* .. and lock protecting it */ 80 xfs_agnumber_t m_maxagi; /* highest inode alloc group */ 81 uint m_readio_log; /* min read size log bytes */ 82 uint m_readio_blocks; /* min read size blocks */ 83 uint m_writeio_log; /* min write size log bytes */ 84 uint m_writeio_blocks; /* min write size blocks */ 85 struct xfs_da_geometry *m_dir_geo; /* directory block geometry */ 86 struct xfs_da_geometry *m_attr_geo; /* attribute block geometry */ 87 struct xlog *m_log; /* log specific stuff */ 88 int m_logbufs; /* number of log buffers */ 89 int m_logbsize; /* size of each log buffer */ 90 uint m_rsumlevels; /* rt summary levels */ 91 uint m_rsumsize; /* size of rt summary, bytes */ 92 /* 93 * Optional cache of rt summary level per bitmap block with the 94 * invariant that m_rsum_cache[bbno] <= the minimum i for which 95 * rsum[i][bbno] != 0. Reads and writes are serialized by the rsumip 96 * inode lock. 97 */ 98 uint8_t *m_rsum_cache; 99 struct xfs_inode *m_rbmip; /* pointer to bitmap inode */ 100 struct xfs_inode *m_rsumip; /* pointer to summary inode */ 101 struct xfs_inode *m_rootip; /* pointer to root directory */ 102 struct xfs_quotainfo *m_quotainfo; /* disk quota information */ 103 xfs_buftarg_t *m_ddev_targp; /* saves taking the address */ 104 xfs_buftarg_t *m_logdev_targp;/* ptr to log device */ 105 xfs_buftarg_t *m_rtdev_targp; /* ptr to rt device */ 106 uint8_t m_blkbit_log; /* blocklog + NBBY */ 107 uint8_t m_blkbb_log; /* blocklog - BBSHIFT */ 108 uint8_t m_agno_log; /* log #ag's */ 109 uint8_t m_agino_log; /* #bits for agino in inum */ 110 uint m_inode_cluster_size;/* min inode buf size */ 111 unsigned int m_inodes_per_cluster; 112 unsigned int m_blocks_per_cluster; 113 unsigned int m_cluster_align; 114 unsigned int m_cluster_align_inodes; 115 uint m_blockmask; /* sb_blocksize-1 */ 116 uint m_blockwsize; /* sb_blocksize in words */ 117 uint m_blockwmask; /* blockwsize-1 */ 118 uint m_alloc_mxr[2]; /* max alloc btree records */ 119 uint m_alloc_mnr[2]; /* min alloc btree records */ 120 uint m_bmap_dmxr[2]; /* max bmap btree records */ 121 uint m_bmap_dmnr[2]; /* min bmap btree records */ 122 uint m_inobt_mxr[2]; /* max inobt btree records */ 123 uint m_inobt_mnr[2]; /* min inobt btree records */ 124 uint m_rmap_mxr[2]; /* max rmap btree records */ 125 uint m_rmap_mnr[2]; /* min rmap btree records */ 126 uint m_refc_mxr[2]; /* max refc btree records */ 127 uint m_refc_mnr[2]; /* min refc btree records */ 128 uint m_ag_maxlevels; /* XFS_AG_MAXLEVELS */ 129 uint m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */ 130 uint m_in_maxlevels; /* max inobt btree levels. */ 131 uint m_rmap_maxlevels; /* max rmap btree levels */ 132 uint m_refc_maxlevels; /* max refcount btree level */ 133 xfs_extlen_t m_ag_prealloc_blocks; /* reserved ag blocks */ 134 uint m_alloc_set_aside; /* space we can't use */ 135 uint m_ag_max_usable; /* max space per AG */ 136 struct radix_tree_root m_perag_tree; /* per-ag accounting info */ 137 spinlock_t m_perag_lock; /* lock for m_perag_tree */ 138 struct mutex m_growlock; /* growfs mutex */ 139 int m_fixedfsid[2]; /* unchanged for life of FS */ 140 uint64_t m_flags; /* global mount flags */ 141 bool m_finobt_nores; /* no per-AG finobt resv. */ 142 int m_ialloc_inos; /* inodes in inode allocation */ 143 int m_ialloc_blks; /* blocks in inode allocation */ 144 int m_ialloc_min_blks;/* min blocks in sparse inode 145 * allocation */ 146 int m_inoalign_mask;/* mask sb_inoalignmt if used */ 147 uint m_qflags; /* quota status flags */ 148 struct xfs_trans_resv m_resv; /* precomputed res values */ 149 uint64_t m_maxicount; /* maximum inode count */ 150 uint64_t m_resblks; /* total reserved blocks */ 151 uint64_t m_resblks_avail;/* available reserved blocks */ 152 uint64_t m_resblks_save; /* reserved blks @ remount,ro */ 153 int m_dalign; /* stripe unit */ 154 int m_swidth; /* stripe width */ 155 int m_sinoalign; /* stripe unit inode alignment */ 156 uint8_t m_sectbb_log; /* sectlog - BBSHIFT */ 157 const struct xfs_nameops *m_dirnameops; /* vector of dir name ops */ 158 const struct xfs_dir_ops *m_dir_inode_ops; /* vector of dir inode ops */ 159 const struct xfs_dir_ops *m_nondir_inode_ops; /* !dir inode ops */ 160 uint m_chsize; /* size of next field */ 161 atomic_t m_active_trans; /* number trans frozen */ 162 struct xfs_mru_cache *m_filestream; /* per-mount filestream data */ 163 struct delayed_work m_reclaim_work; /* background inode reclaim */ 164 struct delayed_work m_eofblocks_work; /* background eof blocks 165 trimming */ 166 struct delayed_work m_cowblocks_work; /* background cow blocks 167 trimming */ 168 bool m_update_sb; /* sb needs update in mount */ 169 int64_t m_low_space[XFS_LOWSP_MAX]; 170 /* low free space thresholds */ 171 struct xfs_kobj m_kobj; 172 struct xfs_kobj m_error_kobj; 173 struct xfs_kobj m_error_meta_kobj; 174 struct xfs_error_cfg m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX]; 175 struct xstats m_stats; /* per-fs stats */ 176 177 struct workqueue_struct *m_buf_workqueue; 178 struct workqueue_struct *m_data_workqueue; 179 struct workqueue_struct *m_unwritten_workqueue; 180 struct workqueue_struct *m_cil_workqueue; 181 struct workqueue_struct *m_reclaim_workqueue; 182 struct workqueue_struct *m_log_workqueue; 183 struct workqueue_struct *m_eofblocks_workqueue; 184 struct workqueue_struct *m_sync_workqueue; 185 186 /* 187 * Generation of the filesysyem layout. This is incremented by each 188 * growfs, and used by the pNFS server to ensure the client updates 189 * its view of the block device once it gets a layout that might 190 * reference the newly added blocks. Does not need to be persistent 191 * as long as we only allow file system size increments, but if we 192 * ever support shrinks it would have to be persisted in addition 193 * to various other kinds of pain inflicted on the pNFS server. 194 */ 195 uint32_t m_generation; 196 197 bool m_always_cow; 198 bool m_fail_unmount; 199 #ifdef DEBUG 200 /* 201 * Frequency with which errors are injected. Replaces xfs_etest; the 202 * value stored in here is the inverse of the frequency with which the 203 * error triggers. 1 = always, 2 = half the time, etc. 204 */ 205 unsigned int *m_errortag; 206 struct xfs_kobj m_errortag_kobj; 207 #endif 208 } xfs_mount_t; 209 210 /* 211 * Flags for m_flags. 212 */ 213 #define XFS_MOUNT_WSYNC (1ULL << 0) /* for nfs - all metadata ops 214 must be synchronous except 215 for space allocations */ 216 #define XFS_MOUNT_UNMOUNTING (1ULL << 1) /* filesystem is unmounting */ 217 #define XFS_MOUNT_BAD_SUMMARY (1ULL << 2) /* summary counters are bad */ 218 #define XFS_MOUNT_WAS_CLEAN (1ULL << 3) 219 #define XFS_MOUNT_FS_SHUTDOWN (1ULL << 4) /* atomic stop of all filesystem 220 operations, typically for 221 disk errors in metadata */ 222 #define XFS_MOUNT_DISCARD (1ULL << 5) /* discard unused blocks */ 223 #define XFS_MOUNT_NOALIGN (1ULL << 7) /* turn off stripe alignment 224 allocations */ 225 #define XFS_MOUNT_ATTR2 (1ULL << 8) /* allow use of attr2 format */ 226 #define XFS_MOUNT_GRPID (1ULL << 9) /* group-ID assigned from directory */ 227 #define XFS_MOUNT_NORECOVERY (1ULL << 10) /* no recovery - dirty fs */ 228 #define XFS_MOUNT_DFLT_IOSIZE (1ULL << 12) /* set default i/o size */ 229 #define XFS_MOUNT_SMALL_INUMS (1ULL << 14) /* user wants 32bit inodes */ 230 #define XFS_MOUNT_32BITINODES (1ULL << 15) /* inode32 allocator active */ 231 #define XFS_MOUNT_NOUUID (1ULL << 16) /* ignore uuid during mount */ 232 #define XFS_MOUNT_IKEEP (1ULL << 18) /* keep empty inode clusters*/ 233 #define XFS_MOUNT_SWALLOC (1ULL << 19) /* turn on stripe width 234 * allocation */ 235 #define XFS_MOUNT_RDONLY (1ULL << 20) /* read-only fs */ 236 #define XFS_MOUNT_DIRSYNC (1ULL << 21) /* synchronous directory ops */ 237 #define XFS_MOUNT_COMPAT_IOSIZE (1ULL << 22) /* don't report large preferred 238 * I/O size in stat() */ 239 #define XFS_MOUNT_FILESTREAMS (1ULL << 24) /* enable the filestreams 240 allocator */ 241 #define XFS_MOUNT_NOATTR2 (1ULL << 25) /* disable use of attr2 format */ 242 243 #define XFS_MOUNT_DAX (1ULL << 62) /* TEST ONLY! */ 244 245 246 /* 247 * Default minimum read and write sizes. 248 */ 249 #define XFS_READIO_LOG_LARGE 16 250 #define XFS_WRITEIO_LOG_LARGE 16 251 252 /* 253 * Max and min values for mount-option defined I/O 254 * preallocation sizes. 255 */ 256 #define XFS_MAX_IO_LOG 30 /* 1G */ 257 #define XFS_MIN_IO_LOG PAGE_SHIFT 258 259 /* 260 * Synchronous read and write sizes. This should be 261 * better for NFSv2 wsync filesystems. 262 */ 263 #define XFS_WSYNC_READIO_LOG 15 /* 32k */ 264 #define XFS_WSYNC_WRITEIO_LOG 14 /* 16k */ 265 266 /* 267 * Allow large block sizes to be reported to userspace programs if the 268 * "largeio" mount option is used. 269 * 270 * If compatibility mode is specified, simply return the basic unit of caching 271 * so that we don't get inefficient read/modify/write I/O from user apps. 272 * Otherwise.... 273 * 274 * If the underlying volume is a stripe, then return the stripe width in bytes 275 * as the recommended I/O size. It is not a stripe and we've set a default 276 * buffered I/O size, return that, otherwise return the compat default. 277 */ 278 static inline unsigned long 279 xfs_preferred_iosize(xfs_mount_t *mp) 280 { 281 if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE) 282 return PAGE_SIZE; 283 return (mp->m_swidth ? 284 (mp->m_swidth << mp->m_sb.sb_blocklog) : 285 ((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ? 286 (1 << (int)max(mp->m_readio_log, mp->m_writeio_log)) : 287 PAGE_SIZE)); 288 } 289 290 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp) \ 291 ((mp)->m_flags & XFS_MOUNT_WAS_CLEAN) 292 #define XFS_FORCED_SHUTDOWN(mp) ((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN) 293 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname, 294 int lnnum); 295 #define xfs_force_shutdown(m,f) \ 296 xfs_do_force_shutdown(m, f, __FILE__, __LINE__) 297 298 #define SHUTDOWN_META_IO_ERROR 0x0001 /* write attempt to metadata failed */ 299 #define SHUTDOWN_LOG_IO_ERROR 0x0002 /* write attempt to the log failed */ 300 #define SHUTDOWN_FORCE_UMOUNT 0x0004 /* shutdown from a forced unmount */ 301 #define SHUTDOWN_CORRUPT_INCORE 0x0008 /* corrupt in-memory data structures */ 302 #define SHUTDOWN_REMOTE_REQ 0x0010 /* shutdown came from remote cell */ 303 #define SHUTDOWN_DEVICE_REQ 0x0020 /* failed all paths to the device */ 304 305 /* 306 * Flags for xfs_mountfs 307 */ 308 #define XFS_MFSI_QUIET 0x40 /* Be silent if mount errors found */ 309 310 static inline xfs_agnumber_t 311 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d) 312 { 313 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d); 314 do_div(ld, mp->m_sb.sb_agblocks); 315 return (xfs_agnumber_t) ld; 316 } 317 318 static inline xfs_agblock_t 319 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d) 320 { 321 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d); 322 return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks); 323 } 324 325 /* per-AG block reservation data structures*/ 326 enum xfs_ag_resv_type { 327 XFS_AG_RESV_NONE = 0, 328 XFS_AG_RESV_AGFL, 329 XFS_AG_RESV_METADATA, 330 XFS_AG_RESV_RMAPBT, 331 }; 332 333 struct xfs_ag_resv { 334 /* number of blocks originally reserved here */ 335 xfs_extlen_t ar_orig_reserved; 336 /* number of blocks reserved here */ 337 xfs_extlen_t ar_reserved; 338 /* number of blocks originally asked for */ 339 xfs_extlen_t ar_asked; 340 }; 341 342 /* 343 * Per-ag incore structure, copies of information in agf and agi, to improve the 344 * performance of allocation group selection. 345 */ 346 typedef struct xfs_perag { 347 struct xfs_mount *pag_mount; /* owner filesystem */ 348 xfs_agnumber_t pag_agno; /* AG this structure belongs to */ 349 atomic_t pag_ref; /* perag reference count */ 350 char pagf_init; /* this agf's entry is initialized */ 351 char pagi_init; /* this agi's entry is initialized */ 352 char pagf_metadata; /* the agf is preferred to be metadata */ 353 char pagi_inodeok; /* The agi is ok for inodes */ 354 uint8_t pagf_levels[XFS_BTNUM_AGF]; 355 /* # of levels in bno & cnt btree */ 356 bool pagf_agflreset; /* agfl requires reset before use */ 357 uint32_t pagf_flcount; /* count of blocks in freelist */ 358 xfs_extlen_t pagf_freeblks; /* total free blocks */ 359 xfs_extlen_t pagf_longest; /* longest free space */ 360 uint32_t pagf_btreeblks; /* # of blocks held in AGF btrees */ 361 xfs_agino_t pagi_freecount; /* number of free inodes */ 362 xfs_agino_t pagi_count; /* number of allocated inodes */ 363 364 /* 365 * Inode allocation search lookup optimisation. 366 * If the pagino matches, the search for new inodes 367 * doesn't need to search the near ones again straight away 368 */ 369 xfs_agino_t pagl_pagino; 370 xfs_agino_t pagl_leftrec; 371 xfs_agino_t pagl_rightrec; 372 spinlock_t pagb_lock; /* lock for pagb_tree */ 373 struct rb_root pagb_tree; /* ordered tree of busy extents */ 374 unsigned int pagb_gen; /* generation count for pagb_tree */ 375 wait_queue_head_t pagb_wait; /* woken when pagb_gen changes */ 376 377 atomic_t pagf_fstrms; /* # of filestreams active in this AG */ 378 379 spinlock_t pag_ici_lock; /* incore inode cache lock */ 380 struct radix_tree_root pag_ici_root; /* incore inode cache root */ 381 int pag_ici_reclaimable; /* reclaimable inodes */ 382 struct mutex pag_ici_reclaim_lock; /* serialisation point */ 383 unsigned long pag_ici_reclaim_cursor; /* reclaim restart point */ 384 385 /* buffer cache index */ 386 spinlock_t pag_buf_lock; /* lock for pag_buf_hash */ 387 struct rhashtable pag_buf_hash; 388 389 /* for rcu-safe freeing */ 390 struct rcu_head rcu_head; 391 int pagb_count; /* pagb slots in use */ 392 393 /* Blocks reserved for all kinds of metadata. */ 394 struct xfs_ag_resv pag_meta_resv; 395 /* Blocks reserved for the reverse mapping btree. */ 396 struct xfs_ag_resv pag_rmapbt_resv; 397 398 /* reference count */ 399 uint8_t pagf_refcount_level; 400 401 /* 402 * Unlinked inode information. This incore information reflects 403 * data stored in the AGI, so callers must hold the AGI buffer lock 404 * or have some other means to control concurrency. 405 */ 406 struct rhashtable pagi_unlinked_hash; 407 } xfs_perag_t; 408 409 static inline struct xfs_ag_resv * 410 xfs_perag_resv( 411 struct xfs_perag *pag, 412 enum xfs_ag_resv_type type) 413 { 414 switch (type) { 415 case XFS_AG_RESV_METADATA: 416 return &pag->pag_meta_resv; 417 case XFS_AG_RESV_RMAPBT: 418 return &pag->pag_rmapbt_resv; 419 default: 420 return NULL; 421 } 422 } 423 424 int xfs_buf_hash_init(xfs_perag_t *pag); 425 void xfs_buf_hash_destroy(xfs_perag_t *pag); 426 427 extern void xfs_uuid_table_free(void); 428 extern int xfs_log_sbcount(xfs_mount_t *); 429 extern uint64_t xfs_default_resblks(xfs_mount_t *mp); 430 extern int xfs_mountfs(xfs_mount_t *mp); 431 extern int xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount, 432 xfs_agnumber_t *maxagi); 433 extern void xfs_unmountfs(xfs_mount_t *); 434 435 extern int xfs_mod_icount(struct xfs_mount *mp, int64_t delta); 436 extern int xfs_mod_ifree(struct xfs_mount *mp, int64_t delta); 437 extern int xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta, 438 bool reserved); 439 extern int xfs_mod_frextents(struct xfs_mount *mp, int64_t delta); 440 441 extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int); 442 extern int xfs_readsb(xfs_mount_t *, int); 443 extern void xfs_freesb(xfs_mount_t *); 444 extern bool xfs_fs_writable(struct xfs_mount *mp, int level); 445 extern int xfs_sb_validate_fsb_count(struct xfs_sb *, uint64_t); 446 447 extern int xfs_dev_is_read_only(struct xfs_mount *, char *); 448 449 extern void xfs_set_low_space_thresholds(struct xfs_mount *); 450 451 int xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb, 452 xfs_off_t count_fsb); 453 454 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp, 455 int error_class, int error); 456 void xfs_force_summary_recalc(struct xfs_mount *mp); 457 458 #endif /* __XFS_MOUNT_H__ */ 459