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 */
63 struct xfs_ail *m_ail; /* fs active log item list */
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 */
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 */
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
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_inotbt_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
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
166 struct delayed_work m_cowblocks_work; /* background cow blocks
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 */
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;
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.
195 uint32_t m_generation;
200 * Frequency with which errors are injected. Replaces xfs_etest; the
201 * value stored in here is the inverse of the frequency with which the
202 * error triggers. 1 = always, 2 = half the time, etc.
204 unsigned int *m_errortag;
205 struct xfs_kobj m_errortag_kobj;
212 #define XFS_MOUNT_WSYNC (1ULL << 0) /* for nfs - all metadata ops
213 must be synchronous except
214 for space allocations */
215 #define XFS_MOUNT_UNMOUNTING (1ULL << 1) /* filesystem is unmounting */
216 #define XFS_MOUNT_BAD_SUMMARY (1ULL << 2) /* summary counters are bad */
217 #define XFS_MOUNT_WAS_CLEAN (1ULL << 3)
218 #define XFS_MOUNT_FS_SHUTDOWN (1ULL << 4) /* atomic stop of all filesystem
219 operations, typically for
220 disk errors in metadata */
221 #define XFS_MOUNT_DISCARD (1ULL << 5) /* discard unused blocks */
222 #define XFS_MOUNT_NOALIGN (1ULL << 7) /* turn off stripe alignment
224 #define XFS_MOUNT_ATTR2 (1ULL << 8) /* allow use of attr2 format */
225 #define XFS_MOUNT_GRPID (1ULL << 9) /* group-ID assigned from directory */
226 #define XFS_MOUNT_NORECOVERY (1ULL << 10) /* no recovery - dirty fs */
227 #define XFS_MOUNT_DFLT_IOSIZE (1ULL << 12) /* set default i/o size */
228 #define XFS_MOUNT_SMALL_INUMS (1ULL << 14) /* user wants 32bit inodes */
229 #define XFS_MOUNT_32BITINODES (1ULL << 15) /* inode32 allocator active */
230 #define XFS_MOUNT_NOUUID (1ULL << 16) /* ignore uuid during mount */
231 #define XFS_MOUNT_IKEEP (1ULL << 18) /* keep empty inode clusters*/
232 #define XFS_MOUNT_SWALLOC (1ULL << 19) /* turn on stripe width
234 #define XFS_MOUNT_RDONLY (1ULL << 20) /* read-only fs */
235 #define XFS_MOUNT_DIRSYNC (1ULL << 21) /* synchronous directory ops */
236 #define XFS_MOUNT_COMPAT_IOSIZE (1ULL << 22) /* don't report large preferred
237 * I/O size in stat() */
238 #define XFS_MOUNT_FILESTREAMS (1ULL << 24) /* enable the filestreams
240 #define XFS_MOUNT_NOATTR2 (1ULL << 25) /* disable use of attr2 format */
242 #define XFS_MOUNT_DAX (1ULL << 62) /* TEST ONLY! */
246 * Default minimum read and write sizes.
248 #define XFS_READIO_LOG_LARGE 16
249 #define XFS_WRITEIO_LOG_LARGE 16
252 * Max and min values for mount-option defined I/O
253 * preallocation sizes.
255 #define XFS_MAX_IO_LOG 30 /* 1G */
256 #define XFS_MIN_IO_LOG PAGE_SHIFT
259 * Synchronous read and write sizes. This should be
260 * better for NFSv2 wsync filesystems.
262 #define XFS_WSYNC_READIO_LOG 15 /* 32k */
263 #define XFS_WSYNC_WRITEIO_LOG 14 /* 16k */
266 * Allow large block sizes to be reported to userspace programs if the
267 * "largeio" mount option is used.
269 * If compatibility mode is specified, simply return the basic unit of caching
270 * so that we don't get inefficient read/modify/write I/O from user apps.
273 * If the underlying volume is a stripe, then return the stripe width in bytes
274 * as the recommended I/O size. It is not a stripe and we've set a default
275 * buffered I/O size, return that, otherwise return the compat default.
277 static inline unsigned long
278 xfs_preferred_iosize(xfs_mount_t *mp)
280 if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
282 return (mp->m_swidth ?
283 (mp->m_swidth << mp->m_sb.sb_blocklog) :
284 ((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
285 (1 << (int)max(mp->m_readio_log, mp->m_writeio_log)) :
289 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp) \
290 ((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
291 #define XFS_FORCED_SHUTDOWN(mp) ((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN)
292 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
294 #define xfs_force_shutdown(m,f) \
295 xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
297 #define SHUTDOWN_META_IO_ERROR 0x0001 /* write attempt to metadata failed */
298 #define SHUTDOWN_LOG_IO_ERROR 0x0002 /* write attempt to the log failed */
299 #define SHUTDOWN_FORCE_UMOUNT 0x0004 /* shutdown from a forced unmount */
300 #define SHUTDOWN_CORRUPT_INCORE 0x0008 /* corrupt in-memory data structures */
301 #define SHUTDOWN_REMOTE_REQ 0x0010 /* shutdown came from remote cell */
302 #define SHUTDOWN_DEVICE_REQ 0x0020 /* failed all paths to the device */
305 * Flags for xfs_mountfs
307 #define XFS_MFSI_QUIET 0x40 /* Be silent if mount errors found */
309 static inline xfs_agnumber_t
310 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
312 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
313 do_div(ld, mp->m_sb.sb_agblocks);
314 return (xfs_agnumber_t) ld;
317 static inline xfs_agblock_t
318 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
320 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
321 return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
324 /* per-AG block reservation data structures*/
325 enum xfs_ag_resv_type {
326 XFS_AG_RESV_NONE = 0,
328 XFS_AG_RESV_METADATA,
333 /* number of blocks originally reserved here */
334 xfs_extlen_t ar_orig_reserved;
335 /* number of blocks reserved here */
336 xfs_extlen_t ar_reserved;
337 /* number of blocks originally asked for */
338 xfs_extlen_t ar_asked;
342 * Per-ag incore structure, copies of information in agf and agi, to improve the
343 * performance of allocation group selection.
345 typedef struct xfs_perag {
346 struct xfs_mount *pag_mount; /* owner filesystem */
347 xfs_agnumber_t pag_agno; /* AG this structure belongs to */
348 atomic_t pag_ref; /* perag reference count */
349 char pagf_init; /* this agf's entry is initialized */
350 char pagi_init; /* this agi's entry is initialized */
351 char pagf_metadata; /* the agf is preferred to be metadata */
352 char pagi_inodeok; /* The agi is ok for inodes */
353 uint8_t pagf_levels[XFS_BTNUM_AGF];
354 /* # of levels in bno & cnt btree */
355 bool pagf_agflreset; /* agfl requires reset before use */
356 uint32_t pagf_flcount; /* count of blocks in freelist */
357 xfs_extlen_t pagf_freeblks; /* total free blocks */
358 xfs_extlen_t pagf_longest; /* longest free space */
359 uint32_t pagf_btreeblks; /* # of blocks held in AGF btrees */
360 xfs_agino_t pagi_freecount; /* number of free inodes */
361 xfs_agino_t pagi_count; /* number of allocated inodes */
364 * Inode allocation search lookup optimisation.
365 * If the pagino matches, the search for new inodes
366 * doesn't need to search the near ones again straight away
368 xfs_agino_t pagl_pagino;
369 xfs_agino_t pagl_leftrec;
370 xfs_agino_t pagl_rightrec;
371 spinlock_t pagb_lock; /* lock for pagb_tree */
372 struct rb_root pagb_tree; /* ordered tree of busy extents */
373 unsigned int pagb_gen; /* generation count for pagb_tree */
374 wait_queue_head_t pagb_wait; /* woken when pagb_gen changes */
376 atomic_t pagf_fstrms; /* # of filestreams active in this AG */
378 spinlock_t pag_ici_lock; /* incore inode cache lock */
379 struct radix_tree_root pag_ici_root; /* incore inode cache root */
380 int pag_ici_reclaimable; /* reclaimable inodes */
381 struct mutex pag_ici_reclaim_lock; /* serialisation point */
382 unsigned long pag_ici_reclaim_cursor; /* reclaim restart point */
384 /* buffer cache index */
385 spinlock_t pag_buf_lock; /* lock for pag_buf_hash */
386 struct rhashtable pag_buf_hash;
388 /* for rcu-safe freeing */
389 struct rcu_head rcu_head;
390 int pagb_count; /* pagb slots in use */
392 /* Blocks reserved for all kinds of metadata. */
393 struct xfs_ag_resv pag_meta_resv;
394 /* Blocks reserved for the reverse mapping btree. */
395 struct xfs_ag_resv pag_rmapbt_resv;
397 /* reference count */
398 uint8_t pagf_refcount_level;
401 static inline struct xfs_ag_resv *
403 struct xfs_perag *pag,
404 enum xfs_ag_resv_type type)
407 case XFS_AG_RESV_METADATA:
408 return &pag->pag_meta_resv;
409 case XFS_AG_RESV_RMAPBT:
410 return &pag->pag_rmapbt_resv;
416 int xfs_buf_hash_init(xfs_perag_t *pag);
417 void xfs_buf_hash_destroy(xfs_perag_t *pag);
419 extern void xfs_uuid_table_free(void);
420 extern int xfs_log_sbcount(xfs_mount_t *);
421 extern uint64_t xfs_default_resblks(xfs_mount_t *mp);
422 extern int xfs_mountfs(xfs_mount_t *mp);
423 extern int xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount,
424 xfs_agnumber_t *maxagi);
425 extern void xfs_unmountfs(xfs_mount_t *);
427 extern int xfs_mod_icount(struct xfs_mount *mp, int64_t delta);
428 extern int xfs_mod_ifree(struct xfs_mount *mp, int64_t delta);
429 extern int xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
431 extern int xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
433 extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
434 extern int xfs_readsb(xfs_mount_t *, int);
435 extern void xfs_freesb(xfs_mount_t *);
436 extern bool xfs_fs_writable(struct xfs_mount *mp, int level);
437 extern int xfs_sb_validate_fsb_count(struct xfs_sb *, uint64_t);
439 extern int xfs_dev_is_read_only(struct xfs_mount *, char *);
441 extern void xfs_set_low_space_thresholds(struct xfs_mount *);
443 int xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb,
444 xfs_off_t count_fsb);
446 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp,
447 int error_class, int error);
448 void xfs_force_summary_recalc(struct xfs_mount *mp);
450 #endif /* __XFS_MOUNT_H__ */