1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
6 #ifndef __XFS_MOUNT_H__
7 #define __XFS_MOUNT_H__
14 struct xfs_da_geometry;
17 /* dynamic preallocation free space thresholds, 5% down to 1% */
30 * Error classes define the subsystem the configuration belongs to.
31 * Error numbers define the errors that are configurable.
45 #define XFS_ERR_RETRY_FOREVER -1
48 * Although retry_timeout is in jiffies which is normally an unsigned long,
49 * we limit the retry timeout to 86400 seconds, or one day. So even a
50 * signed 32-bit long is sufficient for a HZ value up to 24855. Making it
51 * signed lets us store the special "-1" value, meaning retry forever.
53 struct xfs_error_cfg {
56 long retry_timeout; /* in jiffies, -1 = infinite */
60 * Per-cpu deferred inode inactivation GC lists.
63 struct llist_head list;
64 struct work_struct work;
66 /* approximate count of inodes in the list */
68 unsigned int shrinker_hits;
72 * The struct xfsmount layout is optimised to separate read-mostly variables
73 * from variables that are frequently modified. We put the read-mostly variables
74 * first, then place all the other variables at the end.
76 * Typically, read-mostly variables are those that are set at mount time and
77 * never changed again, or only change rarely as a result of things like sysfs
78 * knobs being tweaked.
80 typedef struct xfs_mount {
81 struct xfs_sb m_sb; /* copy of fs superblock */
82 struct super_block *m_super;
83 struct xfs_ail *m_ail; /* fs active log item list */
84 struct xfs_buf *m_sb_bp; /* buffer for superblock */
85 char *m_rtname; /* realtime device name */
86 char *m_logname; /* external log device name */
87 struct xfs_da_geometry *m_dir_geo; /* directory block geometry */
88 struct xfs_da_geometry *m_attr_geo; /* attribute block geometry */
89 struct xlog *m_log; /* log specific stuff */
90 struct xfs_inode *m_rbmip; /* pointer to bitmap inode */
91 struct xfs_inode *m_rsumip; /* pointer to summary inode */
92 struct xfs_inode *m_rootip; /* pointer to root directory */
93 struct xfs_quotainfo *m_quotainfo; /* disk quota information */
94 xfs_buftarg_t *m_ddev_targp; /* saves taking the address */
95 xfs_buftarg_t *m_logdev_targp;/* ptr to log device */
96 xfs_buftarg_t *m_rtdev_targp; /* ptr to rt device */
97 struct list_head m_mount_list; /* global mount list */
98 void __percpu *m_inodegc; /* percpu inodegc structures */
101 * Optional cache of rt summary level per bitmap block with the
102 * invariant that m_rsum_cache[bbno] <= the minimum i for which
103 * rsum[i][bbno] != 0. Reads and writes are serialized by the rsumip
106 uint8_t *m_rsum_cache;
107 struct xfs_mru_cache *m_filestream; /* per-mount filestream data */
108 struct workqueue_struct *m_buf_workqueue;
109 struct workqueue_struct *m_unwritten_workqueue;
110 struct workqueue_struct *m_reclaim_workqueue;
111 struct workqueue_struct *m_sync_workqueue;
112 struct workqueue_struct *m_blockgc_wq;
113 struct workqueue_struct *m_inodegc_wq;
115 int m_bsize; /* fs logical block size */
116 uint8_t m_blkbit_log; /* blocklog + NBBY */
117 uint8_t m_blkbb_log; /* blocklog - BBSHIFT */
118 uint8_t m_agno_log; /* log #ag's */
119 uint8_t m_sectbb_log; /* sectlog - BBSHIFT */
120 uint m_blockmask; /* sb_blocksize-1 */
121 uint m_blockwsize; /* sb_blocksize in words */
122 uint m_blockwmask; /* blockwsize-1 */
123 uint m_alloc_mxr[2]; /* max alloc btree records */
124 uint m_alloc_mnr[2]; /* min alloc btree records */
125 uint m_bmap_dmxr[2]; /* max bmap btree records */
126 uint m_bmap_dmnr[2]; /* min bmap btree records */
127 uint m_rmap_mxr[2]; /* max rmap btree records */
128 uint m_rmap_mnr[2]; /* min rmap btree records */
129 uint m_refc_mxr[2]; /* max refc btree records */
130 uint m_refc_mnr[2]; /* min refc btree records */
131 uint m_alloc_maxlevels; /* max alloc btree levels */
132 uint m_bm_maxlevels[2]; /* max bmap btree levels */
133 uint m_rmap_maxlevels; /* max rmap btree levels */
134 uint m_refc_maxlevels; /* max refcount btree level */
135 unsigned int m_agbtree_maxlevels; /* max level of all AG btrees */
136 xfs_extlen_t m_ag_prealloc_blocks; /* reserved ag blocks */
137 uint m_alloc_set_aside; /* space we can't use */
138 uint m_ag_max_usable; /* max space per AG */
139 int m_dalign; /* stripe unit */
140 int m_swidth; /* stripe width */
141 xfs_agnumber_t m_maxagi; /* highest inode alloc group */
142 uint m_allocsize_log;/* min write size log bytes */
143 uint m_allocsize_blocks; /* min write size blocks */
144 int m_logbufs; /* number of log buffers */
145 int m_logbsize; /* size of each log buffer */
146 uint m_rsumlevels; /* rt summary levels */
147 uint m_rsumsize; /* size of rt summary, bytes */
148 int m_fixedfsid[2]; /* unchanged for life of FS */
149 uint m_qflags; /* quota status flags */
150 uint64_t m_features; /* active filesystem features */
151 uint64_t m_low_space[XFS_LOWSP_MAX];
152 uint64_t m_low_rtexts[XFS_LOWSP_MAX];
153 struct xfs_ino_geometry m_ino_geo; /* inode geometry */
154 struct xfs_trans_resv m_resv; /* precomputed res values */
155 /* low free space thresholds */
156 unsigned long m_opstate; /* dynamic state flags */
159 bool m_finobt_nores; /* no per-AG finobt resv. */
160 bool m_update_sb; /* sb needs update in mount */
163 * Bitsets of per-fs metadata that have been checked and/or are sick.
164 * Callers must hold m_sb_lock to access these two fields.
166 uint8_t m_fs_checked;
169 * Bitsets of rt metadata that have been checked and/or are sick.
170 * Callers must hold m_sb_lock to access this field.
172 uint8_t m_rt_checked;
176 * End of read-mostly variables. Frequently written variables and locks
177 * should be placed below this comment from now on. The first variable
178 * here is marked as cacheline aligned so they it is separated from
179 * the read-mostly variables.
182 spinlock_t ____cacheline_aligned m_sb_lock; /* sb counter lock */
183 struct percpu_counter m_icount; /* allocated inodes counter */
184 struct percpu_counter m_ifree; /* free inodes counter */
185 struct percpu_counter m_fdblocks; /* free block counter */
187 * Count of data device blocks reserved for delayed allocations,
188 * including indlen blocks. Does not include allocated CoW staging
189 * extents or anything related to the rt device.
191 struct percpu_counter m_delalloc_blks;
193 * Global count of allocation btree blocks in use across all AGs. Only
194 * used when perag reservation is enabled. Helps prevent block
195 * reservation from attempting to reserve allocation btree blocks.
197 atomic64_t m_allocbt_blks;
199 struct radix_tree_root m_perag_tree; /* per-ag accounting info */
200 spinlock_t m_perag_lock; /* lock for m_perag_tree */
201 uint64_t m_resblks; /* total reserved blocks */
202 uint64_t m_resblks_avail;/* available reserved blocks */
203 uint64_t m_resblks_save; /* reserved blks @ remount,ro */
204 struct delayed_work m_reclaim_work; /* background inode reclaim */
205 struct xfs_kobj m_kobj;
206 struct xfs_kobj m_error_kobj;
207 struct xfs_kobj m_error_meta_kobj;
208 struct xfs_error_cfg m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX];
209 struct xstats m_stats; /* per-fs stats */
210 xfs_agnumber_t m_agfrotor; /* last ag where space found */
211 xfs_agnumber_t m_agirotor; /* last ag dir inode alloced */
212 spinlock_t m_agirotor_lock;/* .. and lock protecting it */
214 /* Memory shrinker to throttle and reprioritize inodegc */
215 struct shrinker m_inodegc_shrinker;
217 * Workqueue item so that we can coalesce multiple inode flush attempts
218 * into a single flush.
220 struct work_struct m_flush_inodes_work;
223 * Generation of the filesysyem layout. This is incremented by each
224 * growfs, and used by the pNFS server to ensure the client updates
225 * its view of the block device once it gets a layout that might
226 * reference the newly added blocks. Does not need to be persistent
227 * as long as we only allow file system size increments, but if we
228 * ever support shrinks it would have to be persisted in addition
229 * to various other kinds of pain inflicted on the pNFS server.
231 uint32_t m_generation;
232 struct mutex m_growlock; /* growfs mutex */
236 * Frequency with which errors are injected. Replaces xfs_etest; the
237 * value stored in here is the inverse of the frequency with which the
238 * error triggers. 1 = always, 2 = half the time, etc.
240 unsigned int *m_errortag;
241 struct xfs_kobj m_errortag_kobj;
245 #define M_IGEO(mp) (&(mp)->m_ino_geo)
248 * Flags for m_features.
250 * These are all the active features in the filesystem, regardless of how
251 * they are configured.
253 #define XFS_FEAT_ATTR (1ULL << 0) /* xattrs present in fs */
254 #define XFS_FEAT_NLINK (1ULL << 1) /* 32 bit link counts */
255 #define XFS_FEAT_QUOTA (1ULL << 2) /* quota active */
256 #define XFS_FEAT_ALIGN (1ULL << 3) /* inode alignment */
257 #define XFS_FEAT_DALIGN (1ULL << 4) /* data alignment */
258 #define XFS_FEAT_LOGV2 (1ULL << 5) /* version 2 logs */
259 #define XFS_FEAT_SECTOR (1ULL << 6) /* sector size > 512 bytes */
260 #define XFS_FEAT_EXTFLG (1ULL << 7) /* unwritten extents */
261 #define XFS_FEAT_ASCIICI (1ULL << 8) /* ASCII only case-insens. */
262 #define XFS_FEAT_LAZYSBCOUNT (1ULL << 9) /* Superblk counters */
263 #define XFS_FEAT_ATTR2 (1ULL << 10) /* dynamic attr fork */
264 #define XFS_FEAT_PARENT (1ULL << 11) /* parent pointers */
265 #define XFS_FEAT_PROJID32 (1ULL << 12) /* 32 bit project id */
266 #define XFS_FEAT_CRC (1ULL << 13) /* metadata CRCs */
267 #define XFS_FEAT_V3INODES (1ULL << 14) /* Version 3 inodes */
268 #define XFS_FEAT_PQUOTINO (1ULL << 15) /* non-shared proj/grp quotas */
269 #define XFS_FEAT_FTYPE (1ULL << 16) /* inode type in dir */
270 #define XFS_FEAT_FINOBT (1ULL << 17) /* free inode btree */
271 #define XFS_FEAT_RMAPBT (1ULL << 18) /* reverse map btree */
272 #define XFS_FEAT_REFLINK (1ULL << 19) /* reflinked files */
273 #define XFS_FEAT_SPINODES (1ULL << 20) /* sparse inode chunks */
274 #define XFS_FEAT_META_UUID (1ULL << 21) /* metadata UUID */
275 #define XFS_FEAT_REALTIME (1ULL << 22) /* realtime device present */
276 #define XFS_FEAT_INOBTCNT (1ULL << 23) /* inobt block counts */
277 #define XFS_FEAT_BIGTIME (1ULL << 24) /* large timestamps */
278 #define XFS_FEAT_NEEDSREPAIR (1ULL << 25) /* needs xfs_repair */
281 #define XFS_FEAT_NOATTR2 (1ULL << 48) /* disable attr2 creation */
282 #define XFS_FEAT_NOALIGN (1ULL << 49) /* ignore alignment */
283 #define XFS_FEAT_ALLOCSIZE (1ULL << 50) /* user specified allocation size */
284 #define XFS_FEAT_LARGE_IOSIZE (1ULL << 51) /* report large preferred
285 * I/O size in stat() */
286 #define XFS_FEAT_WSYNC (1ULL << 52) /* synchronous metadata ops */
287 #define XFS_FEAT_DIRSYNC (1ULL << 53) /* synchronous directory ops */
288 #define XFS_FEAT_DISCARD (1ULL << 54) /* discard unused blocks */
289 #define XFS_FEAT_GRPID (1ULL << 55) /* group-ID assigned from directory */
290 #define XFS_FEAT_SMALL_INUMS (1ULL << 56) /* user wants 32bit inodes */
291 #define XFS_FEAT_IKEEP (1ULL << 57) /* keep empty inode clusters*/
292 #define XFS_FEAT_SWALLOC (1ULL << 58) /* stripe width allocation */
293 #define XFS_FEAT_FILESTREAMS (1ULL << 59) /* use filestreams allocator */
294 #define XFS_FEAT_DAX_ALWAYS (1ULL << 60) /* DAX always enabled */
295 #define XFS_FEAT_DAX_NEVER (1ULL << 61) /* DAX never enabled */
296 #define XFS_FEAT_NORECOVERY (1ULL << 62) /* no recovery - dirty fs */
297 #define XFS_FEAT_NOUUID (1ULL << 63) /* ignore uuid during mount */
299 #define __XFS_HAS_FEAT(name, NAME) \
300 static inline bool xfs_has_ ## name (struct xfs_mount *mp) \
302 return mp->m_features & XFS_FEAT_ ## NAME; \
305 /* Some features can be added dynamically so they need a set wrapper, too. */
306 #define __XFS_ADD_FEAT(name, NAME) \
307 __XFS_HAS_FEAT(name, NAME); \
308 static inline void xfs_add_ ## name (struct xfs_mount *mp) \
310 mp->m_features |= XFS_FEAT_ ## NAME; \
311 xfs_sb_version_add ## name(&mp->m_sb); \
314 /* Superblock features */
315 __XFS_ADD_FEAT(attr, ATTR)
316 __XFS_HAS_FEAT(nlink, NLINK)
317 __XFS_ADD_FEAT(quota, QUOTA)
318 __XFS_HAS_FEAT(align, ALIGN)
319 __XFS_HAS_FEAT(dalign, DALIGN)
320 __XFS_HAS_FEAT(logv2, LOGV2)
321 __XFS_HAS_FEAT(sector, SECTOR)
322 __XFS_HAS_FEAT(extflg, EXTFLG)
323 __XFS_HAS_FEAT(asciici, ASCIICI)
324 __XFS_HAS_FEAT(lazysbcount, LAZYSBCOUNT)
325 __XFS_ADD_FEAT(attr2, ATTR2)
326 __XFS_HAS_FEAT(parent, PARENT)
327 __XFS_ADD_FEAT(projid32, PROJID32)
328 __XFS_HAS_FEAT(crc, CRC)
329 __XFS_HAS_FEAT(v3inodes, V3INODES)
330 __XFS_HAS_FEAT(pquotino, PQUOTINO)
331 __XFS_HAS_FEAT(ftype, FTYPE)
332 __XFS_HAS_FEAT(finobt, FINOBT)
333 __XFS_HAS_FEAT(rmapbt, RMAPBT)
334 __XFS_HAS_FEAT(reflink, REFLINK)
335 __XFS_HAS_FEAT(sparseinodes, SPINODES)
336 __XFS_HAS_FEAT(metauuid, META_UUID)
337 __XFS_HAS_FEAT(realtime, REALTIME)
338 __XFS_HAS_FEAT(inobtcounts, INOBTCNT)
339 __XFS_HAS_FEAT(bigtime, BIGTIME)
340 __XFS_HAS_FEAT(needsrepair, NEEDSREPAIR)
345 * These do not change dynamically - features that can come and go, such as 32
346 * bit inodes and read-only state, are kept as operational state rather than
349 __XFS_HAS_FEAT(noattr2, NOATTR2)
350 __XFS_HAS_FEAT(noalign, NOALIGN)
351 __XFS_HAS_FEAT(allocsize, ALLOCSIZE)
352 __XFS_HAS_FEAT(large_iosize, LARGE_IOSIZE)
353 __XFS_HAS_FEAT(wsync, WSYNC)
354 __XFS_HAS_FEAT(dirsync, DIRSYNC)
355 __XFS_HAS_FEAT(discard, DISCARD)
356 __XFS_HAS_FEAT(grpid, GRPID)
357 __XFS_HAS_FEAT(small_inums, SMALL_INUMS)
358 __XFS_HAS_FEAT(ikeep, IKEEP)
359 __XFS_HAS_FEAT(swalloc, SWALLOC)
360 __XFS_HAS_FEAT(filestreams, FILESTREAMS)
361 __XFS_HAS_FEAT(dax_always, DAX_ALWAYS)
362 __XFS_HAS_FEAT(dax_never, DAX_NEVER)
363 __XFS_HAS_FEAT(norecovery, NORECOVERY)
364 __XFS_HAS_FEAT(nouuid, NOUUID)
367 * Operational mount state flags
369 * Use these with atomic bit ops only!
371 #define XFS_OPSTATE_UNMOUNTING 0 /* filesystem is unmounting */
372 #define XFS_OPSTATE_CLEAN 1 /* mount was clean */
373 #define XFS_OPSTATE_SHUTDOWN 2 /* stop all fs operations */
374 #define XFS_OPSTATE_INODE32 3 /* inode32 allocator active */
375 #define XFS_OPSTATE_READONLY 4 /* read-only fs */
378 * If set, inactivation worker threads will be scheduled to process queued
379 * inodegc work. If not, queued inodes remain in memory waiting to be
382 #define XFS_OPSTATE_INODEGC_ENABLED 5
384 * If set, background speculative prealloc gc worker threads will be scheduled
385 * to process queued blockgc work. If not, inodes retain their preallocations
386 * until explicitly deleted.
388 #define XFS_OPSTATE_BLOCKGC_ENABLED 6
390 #define __XFS_IS_OPSTATE(name, NAME) \
391 static inline bool xfs_is_ ## name (struct xfs_mount *mp) \
393 return test_bit(XFS_OPSTATE_ ## NAME, &mp->m_opstate); \
395 static inline bool xfs_clear_ ## name (struct xfs_mount *mp) \
397 return test_and_clear_bit(XFS_OPSTATE_ ## NAME, &mp->m_opstate); \
399 static inline bool xfs_set_ ## name (struct xfs_mount *mp) \
401 return test_and_set_bit(XFS_OPSTATE_ ## NAME, &mp->m_opstate); \
404 __XFS_IS_OPSTATE(unmounting, UNMOUNTING)
405 __XFS_IS_OPSTATE(clean, CLEAN)
406 __XFS_IS_OPSTATE(shutdown, SHUTDOWN)
407 __XFS_IS_OPSTATE(inode32, INODE32)
408 __XFS_IS_OPSTATE(readonly, READONLY)
409 __XFS_IS_OPSTATE(inodegc_enabled, INODEGC_ENABLED)
410 __XFS_IS_OPSTATE(blockgc_enabled, BLOCKGC_ENABLED)
412 #define XFS_OPSTATE_STRINGS \
413 { (1UL << XFS_OPSTATE_UNMOUNTING), "unmounting" }, \
414 { (1UL << XFS_OPSTATE_CLEAN), "clean" }, \
415 { (1UL << XFS_OPSTATE_SHUTDOWN), "shutdown" }, \
416 { (1UL << XFS_OPSTATE_INODE32), "inode32" }, \
417 { (1UL << XFS_OPSTATE_READONLY), "read_only" }, \
418 { (1UL << XFS_OPSTATE_INODEGC_ENABLED), "inodegc" }, \
419 { (1UL << XFS_OPSTATE_BLOCKGC_ENABLED), "blockgc" }
422 * Max and min values for mount-option defined I/O
423 * preallocation sizes.
425 #define XFS_MAX_IO_LOG 30 /* 1G */
426 #define XFS_MIN_IO_LOG PAGE_SHIFT
428 #define xfs_is_shutdown(mp) xfs_is_shutdown(mp)
429 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
431 #define xfs_force_shutdown(m,f) \
432 xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
434 #define SHUTDOWN_META_IO_ERROR 0x0001 /* write attempt to metadata failed */
435 #define SHUTDOWN_LOG_IO_ERROR 0x0002 /* write attempt to the log failed */
436 #define SHUTDOWN_FORCE_UMOUNT 0x0004 /* shutdown from a forced unmount */
437 #define SHUTDOWN_CORRUPT_INCORE 0x0008 /* corrupt in-memory data structures */
439 #define XFS_SHUTDOWN_STRINGS \
440 { SHUTDOWN_META_IO_ERROR, "metadata_io" }, \
441 { SHUTDOWN_LOG_IO_ERROR, "log_io" }, \
442 { SHUTDOWN_FORCE_UMOUNT, "force_umount" }, \
443 { SHUTDOWN_CORRUPT_INCORE, "corruption" }
446 * Flags for xfs_mountfs
448 #define XFS_MFSI_QUIET 0x40 /* Be silent if mount errors found */
450 static inline xfs_agnumber_t
451 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
453 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
454 do_div(ld, mp->m_sb.sb_agblocks);
455 return (xfs_agnumber_t) ld;
458 static inline xfs_agblock_t
459 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
461 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
462 return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
465 int xfs_buf_hash_init(struct xfs_perag *pag);
466 void xfs_buf_hash_destroy(struct xfs_perag *pag);
468 extern void xfs_uuid_table_free(void);
469 extern uint64_t xfs_default_resblks(xfs_mount_t *mp);
470 extern int xfs_mountfs(xfs_mount_t *mp);
471 extern void xfs_unmountfs(xfs_mount_t *);
474 * Deltas for the block count can vary from 1 to very large, but lock contention
475 * only occurs on frequent small block count updates such as in the delayed
476 * allocation path for buffered writes (page a time updates). Hence we set
477 * a large batch count (1024) to minimise global counter updates except when
478 * we get near to ENOSPC and we have to be very accurate with our updates.
480 #define XFS_FDBLOCKS_BATCH 1024
482 extern int xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
484 extern int xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
486 extern int xfs_readsb(xfs_mount_t *, int);
487 extern void xfs_freesb(xfs_mount_t *);
488 extern bool xfs_fs_writable(struct xfs_mount *mp, int level);
489 extern int xfs_sb_validate_fsb_count(struct xfs_sb *, uint64_t);
491 extern int xfs_dev_is_read_only(struct xfs_mount *, char *);
493 extern void xfs_set_low_space_thresholds(struct xfs_mount *);
495 int xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb,
496 xfs_off_t count_fsb);
498 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp,
499 int error_class, int error);
500 void xfs_force_summary_recalc(struct xfs_mount *mp);
501 int xfs_add_incompat_log_feature(struct xfs_mount *mp, uint32_t feature);
502 bool xfs_clear_incompat_log_features(struct xfs_mount *mp);
503 void xfs_mod_delalloc(struct xfs_mount *mp, int64_t delta);
505 #endif /* __XFS_MOUNT_H__ */