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__
16 struct xfs_da_geometry;
18 /* dynamic preallocation free space thresholds, 5% down to 1% */
31 * Error classes define the subsystem the configuration belongs to.
32 * Error numbers define the errors that are configurable.
46 #define XFS_ERR_RETRY_FOREVER -1
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.
54 struct xfs_error_cfg {
57 long retry_timeout; /* in jiffies, -1 = infinite */
60 typedef struct xfs_mount {
61 struct super_block *m_super;
62 xfs_tid_t m_tid; /* next unused tid for fs */
65 * Bitsets of per-fs metadata that have been checked and/or are sick.
66 * Callers must hold m_sb_lock to access these two fields.
71 * Bitsets of rt metadata that have been checked and/or are sick.
72 * Callers must hold m_sb_lock to access this field.
77 struct xfs_ail *m_ail; /* fs active log item list */
79 struct xfs_sb m_sb; /* copy of fs superblock */
80 spinlock_t m_sb_lock; /* sb counter lock */
81 struct percpu_counter m_icount; /* allocated inodes counter */
82 struct percpu_counter m_ifree; /* free inodes counter */
83 struct percpu_counter m_fdblocks; /* free block counter */
85 * Count of data device blocks reserved for delayed allocations,
86 * including indlen blocks. Does not include allocated CoW staging
87 * extents or anything related to the rt device.
89 struct percpu_counter m_delalloc_blks;
91 struct xfs_buf *m_sb_bp; /* buffer for superblock */
92 char *m_fsname; /* filesystem name */
93 int m_fsname_len; /* strlen of fs name */
94 char *m_rtname; /* realtime device name */
95 char *m_logname; /* external log device name */
96 int m_bsize; /* fs logical block size */
97 xfs_agnumber_t m_agfrotor; /* last ag where space found */
98 xfs_agnumber_t m_agirotor; /* last ag dir inode alloced */
99 spinlock_t m_agirotor_lock;/* .. and lock protecting it */
100 xfs_agnumber_t m_maxagi; /* highest inode alloc group */
101 uint m_readio_log; /* min read size log bytes */
102 uint m_readio_blocks; /* min read size blocks */
103 uint m_writeio_log; /* min write size log bytes */
104 uint m_writeio_blocks; /* min write size blocks */
105 struct xfs_da_geometry *m_dir_geo; /* directory block geometry */
106 struct xfs_da_geometry *m_attr_geo; /* attribute block geometry */
107 struct xlog *m_log; /* log specific stuff */
108 int m_logbufs; /* number of log buffers */
109 int m_logbsize; /* size of each log buffer */
110 uint m_rsumlevels; /* rt summary levels */
111 uint m_rsumsize; /* size of rt summary, bytes */
113 * Optional cache of rt summary level per bitmap block with the
114 * invariant that m_rsum_cache[bbno] <= the minimum i for which
115 * rsum[i][bbno] != 0. Reads and writes are serialized by the rsumip
118 uint8_t *m_rsum_cache;
119 struct xfs_inode *m_rbmip; /* pointer to bitmap inode */
120 struct xfs_inode *m_rsumip; /* pointer to summary inode */
121 struct xfs_inode *m_rootip; /* pointer to root directory */
122 struct xfs_quotainfo *m_quotainfo; /* disk quota information */
123 xfs_buftarg_t *m_ddev_targp; /* saves taking the address */
124 xfs_buftarg_t *m_logdev_targp;/* ptr to log device */
125 xfs_buftarg_t *m_rtdev_targp; /* ptr to rt device */
126 uint8_t m_blkbit_log; /* blocklog + NBBY */
127 uint8_t m_blkbb_log; /* blocklog - BBSHIFT */
128 uint8_t m_agno_log; /* log #ag's */
129 uint8_t m_agino_log; /* #bits for agino in inum */
130 uint m_inode_cluster_size;/* min inode buf size */
131 unsigned int m_inodes_per_cluster;
132 unsigned int m_blocks_per_cluster;
133 unsigned int m_cluster_align;
134 unsigned int m_cluster_align_inodes;
135 uint m_blockmask; /* sb_blocksize-1 */
136 uint m_blockwsize; /* sb_blocksize in words */
137 uint m_blockwmask; /* blockwsize-1 */
138 uint m_alloc_mxr[2]; /* max alloc btree records */
139 uint m_alloc_mnr[2]; /* min alloc btree records */
140 uint m_bmap_dmxr[2]; /* max bmap btree records */
141 uint m_bmap_dmnr[2]; /* min bmap btree records */
142 uint m_inobt_mxr[2]; /* max inobt btree records */
143 uint m_inobt_mnr[2]; /* min inobt btree records */
144 uint m_rmap_mxr[2]; /* max rmap btree records */
145 uint m_rmap_mnr[2]; /* min rmap btree records */
146 uint m_refc_mxr[2]; /* max refc btree records */
147 uint m_refc_mnr[2]; /* min refc btree records */
148 uint m_ag_maxlevels; /* XFS_AG_MAXLEVELS */
149 uint m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
150 uint m_in_maxlevels; /* max inobt btree levels. */
151 uint m_rmap_maxlevels; /* max rmap btree levels */
152 uint m_refc_maxlevels; /* max refcount btree level */
153 xfs_extlen_t m_ag_prealloc_blocks; /* reserved ag blocks */
154 uint m_alloc_set_aside; /* space we can't use */
155 uint m_ag_max_usable; /* max space per AG */
156 struct radix_tree_root m_perag_tree; /* per-ag accounting info */
157 spinlock_t m_perag_lock; /* lock for m_perag_tree */
158 struct mutex m_growlock; /* growfs mutex */
159 int m_fixedfsid[2]; /* unchanged for life of FS */
160 uint64_t m_flags; /* global mount flags */
161 bool m_finobt_nores; /* no per-AG finobt resv. */
162 int m_ialloc_inos; /* inodes in inode allocation */
163 int m_ialloc_blks; /* blocks in inode allocation */
164 int m_ialloc_min_blks;/* min blocks in sparse inode
166 int m_inoalign_mask;/* mask sb_inoalignmt if used */
167 uint m_qflags; /* quota status flags */
168 struct xfs_trans_resv m_resv; /* precomputed res values */
169 uint64_t m_maxicount; /* maximum inode count */
170 uint64_t m_resblks; /* total reserved blocks */
171 uint64_t m_resblks_avail;/* available reserved blocks */
172 uint64_t m_resblks_save; /* reserved blks @ remount,ro */
173 int m_dalign; /* stripe unit */
174 int m_swidth; /* stripe width */
175 int m_sinoalign; /* stripe unit inode alignment */
176 uint8_t m_sectbb_log; /* sectlog - BBSHIFT */
177 const struct xfs_nameops *m_dirnameops; /* vector of dir name ops */
178 const struct xfs_dir_ops *m_dir_inode_ops; /* vector of dir inode ops */
179 const struct xfs_dir_ops *m_nondir_inode_ops; /* !dir inode ops */
180 uint m_chsize; /* size of next field */
181 atomic_t m_active_trans; /* number trans frozen */
182 struct xfs_mru_cache *m_filestream; /* per-mount filestream data */
183 struct delayed_work m_reclaim_work; /* background inode reclaim */
184 struct delayed_work m_eofblocks_work; /* background eof blocks
186 struct delayed_work m_cowblocks_work; /* background cow blocks
188 bool m_update_sb; /* sb needs update in mount */
189 int64_t m_low_space[XFS_LOWSP_MAX];
190 /* low free space thresholds */
191 struct xfs_kobj m_kobj;
192 struct xfs_kobj m_error_kobj;
193 struct xfs_kobj m_error_meta_kobj;
194 struct xfs_error_cfg m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX];
195 struct xstats m_stats; /* per-fs stats */
197 struct workqueue_struct *m_buf_workqueue;
198 struct workqueue_struct *m_unwritten_workqueue;
199 struct workqueue_struct *m_cil_workqueue;
200 struct workqueue_struct *m_reclaim_workqueue;
201 struct workqueue_struct *m_log_workqueue;
202 struct workqueue_struct *m_eofblocks_workqueue;
203 struct workqueue_struct *m_sync_workqueue;
206 * Generation of the filesysyem layout. This is incremented by each
207 * growfs, and used by the pNFS server to ensure the client updates
208 * its view of the block device once it gets a layout that might
209 * reference the newly added blocks. Does not need to be persistent
210 * as long as we only allow file system size increments, but if we
211 * ever support shrinks it would have to be persisted in addition
212 * to various other kinds of pain inflicted on the pNFS server.
214 uint32_t m_generation;
220 * Frequency with which errors are injected. Replaces xfs_etest; the
221 * value stored in here is the inverse of the frequency with which the
222 * error triggers. 1 = always, 2 = half the time, etc.
224 unsigned int *m_errortag;
225 struct xfs_kobj m_errortag_kobj;
232 #define XFS_MOUNT_WSYNC (1ULL << 0) /* for nfs - all metadata ops
233 must be synchronous except
234 for space allocations */
235 #define XFS_MOUNT_UNMOUNTING (1ULL << 1) /* filesystem is unmounting */
236 #define XFS_MOUNT_WAS_CLEAN (1ULL << 3)
237 #define XFS_MOUNT_FS_SHUTDOWN (1ULL << 4) /* atomic stop of all filesystem
238 operations, typically for
239 disk errors in metadata */
240 #define XFS_MOUNT_DISCARD (1ULL << 5) /* discard unused blocks */
241 #define XFS_MOUNT_NOALIGN (1ULL << 7) /* turn off stripe alignment
243 #define XFS_MOUNT_ATTR2 (1ULL << 8) /* allow use of attr2 format */
244 #define XFS_MOUNT_GRPID (1ULL << 9) /* group-ID assigned from directory */
245 #define XFS_MOUNT_NORECOVERY (1ULL << 10) /* no recovery - dirty fs */
246 #define XFS_MOUNT_DFLT_IOSIZE (1ULL << 12) /* set default i/o size */
247 #define XFS_MOUNT_SMALL_INUMS (1ULL << 14) /* user wants 32bit inodes */
248 #define XFS_MOUNT_32BITINODES (1ULL << 15) /* inode32 allocator active */
249 #define XFS_MOUNT_NOUUID (1ULL << 16) /* ignore uuid during mount */
250 #define XFS_MOUNT_IKEEP (1ULL << 18) /* keep empty inode clusters*/
251 #define XFS_MOUNT_SWALLOC (1ULL << 19) /* turn on stripe width
253 #define XFS_MOUNT_RDONLY (1ULL << 20) /* read-only fs */
254 #define XFS_MOUNT_DIRSYNC (1ULL << 21) /* synchronous directory ops */
255 #define XFS_MOUNT_COMPAT_IOSIZE (1ULL << 22) /* don't report large preferred
256 * I/O size in stat() */
257 #define XFS_MOUNT_FILESTREAMS (1ULL << 24) /* enable the filestreams
259 #define XFS_MOUNT_NOATTR2 (1ULL << 25) /* disable use of attr2 format */
261 #define XFS_MOUNT_DAX (1ULL << 62) /* TEST ONLY! */
265 * Default minimum read and write sizes.
267 #define XFS_READIO_LOG_LARGE 16
268 #define XFS_WRITEIO_LOG_LARGE 16
271 * Max and min values for mount-option defined I/O
272 * preallocation sizes.
274 #define XFS_MAX_IO_LOG 30 /* 1G */
275 #define XFS_MIN_IO_LOG PAGE_SHIFT
278 * Synchronous read and write sizes. This should be
279 * better for NFSv2 wsync filesystems.
281 #define XFS_WSYNC_READIO_LOG 15 /* 32k */
282 #define XFS_WSYNC_WRITEIO_LOG 14 /* 16k */
285 * Allow large block sizes to be reported to userspace programs if the
286 * "largeio" mount option is used.
288 * If compatibility mode is specified, simply return the basic unit of caching
289 * so that we don't get inefficient read/modify/write I/O from user apps.
292 * If the underlying volume is a stripe, then return the stripe width in bytes
293 * as the recommended I/O size. It is not a stripe and we've set a default
294 * buffered I/O size, return that, otherwise return the compat default.
296 static inline unsigned long
297 xfs_preferred_iosize(xfs_mount_t *mp)
299 if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
301 return (mp->m_swidth ?
302 (mp->m_swidth << mp->m_sb.sb_blocklog) :
303 ((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
304 (1 << (int)max(mp->m_readio_log, mp->m_writeio_log)) :
308 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp) \
309 ((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
310 #define XFS_FORCED_SHUTDOWN(mp) ((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN)
311 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
313 #define xfs_force_shutdown(m,f) \
314 xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
316 #define SHUTDOWN_META_IO_ERROR 0x0001 /* write attempt to metadata failed */
317 #define SHUTDOWN_LOG_IO_ERROR 0x0002 /* write attempt to the log failed */
318 #define SHUTDOWN_FORCE_UMOUNT 0x0004 /* shutdown from a forced unmount */
319 #define SHUTDOWN_CORRUPT_INCORE 0x0008 /* corrupt in-memory data structures */
320 #define SHUTDOWN_REMOTE_REQ 0x0010 /* shutdown came from remote cell */
321 #define SHUTDOWN_DEVICE_REQ 0x0020 /* failed all paths to the device */
324 * Flags for xfs_mountfs
326 #define XFS_MFSI_QUIET 0x40 /* Be silent if mount errors found */
328 static inline xfs_agnumber_t
329 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
331 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
332 do_div(ld, mp->m_sb.sb_agblocks);
333 return (xfs_agnumber_t) ld;
336 static inline xfs_agblock_t
337 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
339 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
340 return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
343 /* per-AG block reservation data structures*/
344 enum xfs_ag_resv_type {
345 XFS_AG_RESV_NONE = 0,
347 XFS_AG_RESV_METADATA,
352 /* number of blocks originally reserved here */
353 xfs_extlen_t ar_orig_reserved;
354 /* number of blocks reserved here */
355 xfs_extlen_t ar_reserved;
356 /* number of blocks originally asked for */
357 xfs_extlen_t ar_asked;
361 * Per-ag incore structure, copies of information in agf and agi, to improve the
362 * performance of allocation group selection.
364 typedef struct xfs_perag {
365 struct xfs_mount *pag_mount; /* owner filesystem */
366 xfs_agnumber_t pag_agno; /* AG this structure belongs to */
367 atomic_t pag_ref; /* perag reference count */
368 char pagf_init; /* this agf's entry is initialized */
369 char pagi_init; /* this agi's entry is initialized */
370 char pagf_metadata; /* the agf is preferred to be metadata */
371 char pagi_inodeok; /* The agi is ok for inodes */
372 uint8_t pagf_levels[XFS_BTNUM_AGF];
373 /* # of levels in bno & cnt btree */
374 bool pagf_agflreset; /* agfl requires reset before use */
375 uint32_t pagf_flcount; /* count of blocks in freelist */
376 xfs_extlen_t pagf_freeblks; /* total free blocks */
377 xfs_extlen_t pagf_longest; /* longest free space */
378 uint32_t pagf_btreeblks; /* # of blocks held in AGF btrees */
379 xfs_agino_t pagi_freecount; /* number of free inodes */
380 xfs_agino_t pagi_count; /* number of allocated inodes */
383 * Inode allocation search lookup optimisation.
384 * If the pagino matches, the search for new inodes
385 * doesn't need to search the near ones again straight away
387 xfs_agino_t pagl_pagino;
388 xfs_agino_t pagl_leftrec;
389 xfs_agino_t pagl_rightrec;
392 * Bitsets of per-ag metadata that have been checked and/or are sick.
393 * Callers should hold pag_state_lock before accessing this field.
395 uint16_t pag_checked;
397 spinlock_t pag_state_lock;
399 spinlock_t pagb_lock; /* lock for pagb_tree */
400 struct rb_root pagb_tree; /* ordered tree of busy extents */
401 unsigned int pagb_gen; /* generation count for pagb_tree */
402 wait_queue_head_t pagb_wait; /* woken when pagb_gen changes */
404 atomic_t pagf_fstrms; /* # of filestreams active in this AG */
406 spinlock_t pag_ici_lock; /* incore inode cache lock */
407 struct radix_tree_root pag_ici_root; /* incore inode cache root */
408 int pag_ici_reclaimable; /* reclaimable inodes */
409 struct mutex pag_ici_reclaim_lock; /* serialisation point */
410 unsigned long pag_ici_reclaim_cursor; /* reclaim restart point */
412 /* buffer cache index */
413 spinlock_t pag_buf_lock; /* lock for pag_buf_hash */
414 struct rhashtable pag_buf_hash;
416 /* for rcu-safe freeing */
417 struct rcu_head rcu_head;
418 int pagb_count; /* pagb slots in use */
420 /* Blocks reserved for all kinds of metadata. */
421 struct xfs_ag_resv pag_meta_resv;
422 /* Blocks reserved for the reverse mapping btree. */
423 struct xfs_ag_resv pag_rmapbt_resv;
425 /* reference count */
426 uint8_t pagf_refcount_level;
429 * Unlinked inode information. This incore information reflects
430 * data stored in the AGI, so callers must hold the AGI buffer lock
431 * or have some other means to control concurrency.
433 struct rhashtable pagi_unlinked_hash;
436 static inline struct xfs_ag_resv *
438 struct xfs_perag *pag,
439 enum xfs_ag_resv_type type)
442 case XFS_AG_RESV_METADATA:
443 return &pag->pag_meta_resv;
444 case XFS_AG_RESV_RMAPBT:
445 return &pag->pag_rmapbt_resv;
451 int xfs_buf_hash_init(xfs_perag_t *pag);
452 void xfs_buf_hash_destroy(xfs_perag_t *pag);
454 extern void xfs_uuid_table_free(void);
455 extern int xfs_log_sbcount(xfs_mount_t *);
456 extern uint64_t xfs_default_resblks(xfs_mount_t *mp);
457 extern int xfs_mountfs(xfs_mount_t *mp);
458 extern int xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount,
459 xfs_agnumber_t *maxagi);
460 extern void xfs_unmountfs(xfs_mount_t *);
462 extern int xfs_mod_icount(struct xfs_mount *mp, int64_t delta);
463 extern int xfs_mod_ifree(struct xfs_mount *mp, int64_t delta);
464 extern int xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
466 extern int xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
468 extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
469 extern int xfs_readsb(xfs_mount_t *, int);
470 extern void xfs_freesb(xfs_mount_t *);
471 extern bool xfs_fs_writable(struct xfs_mount *mp, int level);
472 extern int xfs_sb_validate_fsb_count(struct xfs_sb *, uint64_t);
474 extern int xfs_dev_is_read_only(struct xfs_mount *, char *);
476 extern void xfs_set_low_space_thresholds(struct xfs_mount *);
478 int xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb,
479 xfs_off_t count_fsb);
481 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp,
482 int error_class, int error);
483 void xfs_force_summary_recalc(struct xfs_mount *mp);
484 void xfs_mod_delalloc(struct xfs_mount *mp, int64_t delta);
486 #endif /* __XFS_MOUNT_H__ */