Loading...
1/*
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18#ifndef __XFS_MOUNT_H__
19#define __XFS_MOUNT_H__
20
21typedef struct xfs_trans_reservations {
22 uint tr_write; /* extent alloc trans */
23 uint tr_itruncate; /* truncate trans */
24 uint tr_rename; /* rename trans */
25 uint tr_link; /* link trans */
26 uint tr_remove; /* unlink trans */
27 uint tr_symlink; /* symlink trans */
28 uint tr_create; /* create trans */
29 uint tr_mkdir; /* mkdir trans */
30 uint tr_ifree; /* inode free trans */
31 uint tr_ichange; /* inode update trans */
32 uint tr_growdata; /* fs data section grow trans */
33 uint tr_swrite; /* sync write inode trans */
34 uint tr_addafork; /* cvt inode to attributed trans */
35 uint tr_writeid; /* write setuid/setgid file */
36 uint tr_attrinval; /* attr fork buffer invalidation */
37 uint tr_attrset; /* set/create an attribute */
38 uint tr_attrrm; /* remove an attribute */
39 uint tr_clearagi; /* clear bad agi unlinked ino bucket */
40 uint tr_growrtalloc; /* grow realtime allocations */
41 uint tr_growrtzero; /* grow realtime zeroing */
42 uint tr_growrtfree; /* grow realtime freeing */
43} xfs_trans_reservations_t;
44
45#ifndef __KERNEL__
46
47#define xfs_daddr_to_agno(mp,d) \
48 ((xfs_agnumber_t)(XFS_BB_TO_FSBT(mp, d) / (mp)->m_sb.sb_agblocks))
49#define xfs_daddr_to_agbno(mp,d) \
50 ((xfs_agblock_t)(XFS_BB_TO_FSBT(mp, d) % (mp)->m_sb.sb_agblocks))
51
52#else /* __KERNEL__ */
53
54#include "xfs_sync.h"
55
56struct log;
57struct xfs_mount_args;
58struct xfs_inode;
59struct xfs_bmbt_irec;
60struct xfs_bmap_free;
61struct xfs_extdelta;
62struct xfs_swapext;
63struct xfs_mru_cache;
64struct xfs_nameops;
65struct xfs_ail;
66struct xfs_quotainfo;
67
68#ifdef HAVE_PERCPU_SB
69
70/*
71 * Valid per-cpu incore superblock counters. Note that if you add new counters,
72 * you may need to define new counter disabled bit field descriptors as there
73 * are more possible fields in the superblock that can fit in a bitfield on a
74 * 32 bit platform. The XFS_SBS_* values for the current current counters just
75 * fit.
76 */
77typedef struct xfs_icsb_cnts {
78 uint64_t icsb_fdblocks;
79 uint64_t icsb_ifree;
80 uint64_t icsb_icount;
81 unsigned long icsb_flags;
82} xfs_icsb_cnts_t;
83
84#define XFS_ICSB_FLAG_LOCK (1 << 0) /* counter lock bit */
85
86#define XFS_ICSB_LAZY_COUNT (1 << 1) /* accuracy not needed */
87
88extern int xfs_icsb_init_counters(struct xfs_mount *);
89extern void xfs_icsb_reinit_counters(struct xfs_mount *);
90extern void xfs_icsb_destroy_counters(struct xfs_mount *);
91extern void xfs_icsb_sync_counters(struct xfs_mount *, int);
92extern void xfs_icsb_sync_counters_locked(struct xfs_mount *, int);
93extern int xfs_icsb_modify_counters(struct xfs_mount *, xfs_sb_field_t,
94 int64_t, int);
95
96#else
97#define xfs_icsb_init_counters(mp) (0)
98#define xfs_icsb_destroy_counters(mp) do { } while (0)
99#define xfs_icsb_reinit_counters(mp) do { } while (0)
100#define xfs_icsb_sync_counters(mp, flags) do { } while (0)
101#define xfs_icsb_sync_counters_locked(mp, flags) do { } while (0)
102#define xfs_icsb_modify_counters(mp, field, delta, rsvd) \
103 xfs_mod_incore_sb(mp, field, delta, rsvd)
104#endif
105
106/* dynamic preallocation free space thresholds, 5% down to 1% */
107enum {
108 XFS_LOWSP_1_PCNT = 0,
109 XFS_LOWSP_2_PCNT,
110 XFS_LOWSP_3_PCNT,
111 XFS_LOWSP_4_PCNT,
112 XFS_LOWSP_5_PCNT,
113 XFS_LOWSP_MAX,
114};
115
116typedef struct xfs_mount {
117 struct super_block *m_super;
118 xfs_tid_t m_tid; /* next unused tid for fs */
119 struct xfs_ail *m_ail; /* fs active log item list */
120 xfs_sb_t m_sb; /* copy of fs superblock */
121 spinlock_t m_sb_lock; /* sb counter lock */
122 struct xfs_buf *m_sb_bp; /* buffer for superblock */
123 char *m_fsname; /* filesystem name */
124 int m_fsname_len; /* strlen of fs name */
125 char *m_rtname; /* realtime device name */
126 char *m_logname; /* external log device name */
127 int m_bsize; /* fs logical block size */
128 xfs_agnumber_t m_agfrotor; /* last ag where space found */
129 xfs_agnumber_t m_agirotor; /* last ag dir inode alloced */
130 spinlock_t m_agirotor_lock;/* .. and lock protecting it */
131 xfs_agnumber_t m_maxagi; /* highest inode alloc group */
132 uint m_readio_log; /* min read size log bytes */
133 uint m_readio_blocks; /* min read size blocks */
134 uint m_writeio_log; /* min write size log bytes */
135 uint m_writeio_blocks; /* min write size blocks */
136 struct log *m_log; /* log specific stuff */
137 int m_logbufs; /* number of log buffers */
138 int m_logbsize; /* size of each log buffer */
139 uint m_rsumlevels; /* rt summary levels */
140 uint m_rsumsize; /* size of rt summary, bytes */
141 struct xfs_inode *m_rbmip; /* pointer to bitmap inode */
142 struct xfs_inode *m_rsumip; /* pointer to summary inode */
143 struct xfs_inode *m_rootip; /* pointer to root directory */
144 struct xfs_quotainfo *m_quotainfo; /* disk quota information */
145 xfs_buftarg_t *m_ddev_targp; /* saves taking the address */
146 xfs_buftarg_t *m_logdev_targp;/* ptr to log device */
147 xfs_buftarg_t *m_rtdev_targp; /* ptr to rt device */
148 __uint8_t m_blkbit_log; /* blocklog + NBBY */
149 __uint8_t m_blkbb_log; /* blocklog - BBSHIFT */
150 __uint8_t m_agno_log; /* log #ag's */
151 __uint8_t m_agino_log; /* #bits for agino in inum */
152 __uint16_t m_inode_cluster_size;/* min inode buf size */
153 uint m_blockmask; /* sb_blocksize-1 */
154 uint m_blockwsize; /* sb_blocksize in words */
155 uint m_blockwmask; /* blockwsize-1 */
156 uint m_alloc_mxr[2]; /* max alloc btree records */
157 uint m_alloc_mnr[2]; /* min alloc btree records */
158 uint m_bmap_dmxr[2]; /* max bmap btree records */
159 uint m_bmap_dmnr[2]; /* min bmap btree records */
160 uint m_inobt_mxr[2]; /* max inobt btree records */
161 uint m_inobt_mnr[2]; /* min inobt btree records */
162 uint m_ag_maxlevels; /* XFS_AG_MAXLEVELS */
163 uint m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
164 uint m_in_maxlevels; /* max inobt btree levels. */
165 struct radix_tree_root m_perag_tree; /* per-ag accounting info */
166 spinlock_t m_perag_lock; /* lock for m_perag_tree */
167 struct mutex m_growlock; /* growfs mutex */
168 int m_fixedfsid[2]; /* unchanged for life of FS */
169 uint m_dmevmask; /* DMI events for this FS */
170 __uint64_t m_flags; /* global mount flags */
171 uint m_dir_node_ents; /* #entries in a dir danode */
172 uint m_attr_node_ents; /* #entries in attr danode */
173 int m_ialloc_inos; /* inodes in inode allocation */
174 int m_ialloc_blks; /* blocks in inode allocation */
175 int m_inoalign_mask;/* mask sb_inoalignmt if used */
176 uint m_qflags; /* quota status flags */
177 xfs_trans_reservations_t m_reservations;/* precomputed res values */
178 __uint64_t m_maxicount; /* maximum inode count */
179 __uint64_t m_maxioffset; /* maximum inode offset */
180 __uint64_t m_resblks; /* total reserved blocks */
181 __uint64_t m_resblks_avail;/* available reserved blocks */
182 __uint64_t m_resblks_save; /* reserved blks @ remount,ro */
183 int m_dalign; /* stripe unit */
184 int m_swidth; /* stripe width */
185 int m_sinoalign; /* stripe unit inode alignment */
186 int m_attr_magicpct;/* 37% of the blocksize */
187 int m_dir_magicpct; /* 37% of the dir blocksize */
188 __uint8_t m_sectbb_log; /* sectlog - BBSHIFT */
189 const struct xfs_nameops *m_dirnameops; /* vector of dir name ops */
190 int m_dirblksize; /* directory block sz--bytes */
191 int m_dirblkfsbs; /* directory block sz--fsbs */
192 xfs_dablk_t m_dirdatablk; /* blockno of dir data v2 */
193 xfs_dablk_t m_dirleafblk; /* blockno of dir non-data v2 */
194 xfs_dablk_t m_dirfreeblk; /* blockno of dirfreeindex v2 */
195 uint m_chsize; /* size of next field */
196 struct xfs_chash *m_chash; /* fs private inode per-cluster
197 * hash table */
198 atomic_t m_active_trans; /* number trans frozen */
199#ifdef HAVE_PERCPU_SB
200 xfs_icsb_cnts_t __percpu *m_sb_cnts; /* per-cpu superblock counters */
201 unsigned long m_icsb_counters; /* disabled per-cpu counters */
202 struct notifier_block m_icsb_notifier; /* hotplug cpu notifier */
203 struct mutex m_icsb_mutex; /* balancer sync lock */
204#endif
205 struct xfs_mru_cache *m_filestream; /* per-mount filestream data */
206 struct delayed_work m_sync_work; /* background sync work */
207 struct delayed_work m_reclaim_work; /* background inode reclaim */
208 struct work_struct m_flush_work; /* background inode flush */
209 __int64_t m_update_flags; /* sb flags we need to update
210 on the next remount,rw */
211 struct shrinker m_inode_shrink; /* inode reclaim shrinker */
212 int64_t m_low_space[XFS_LOWSP_MAX];
213 /* low free space thresholds */
214} xfs_mount_t;
215
216/*
217 * Flags for m_flags.
218 */
219#define XFS_MOUNT_WSYNC (1ULL << 0) /* for nfs - all metadata ops
220 must be synchronous except
221 for space allocations */
222#define XFS_MOUNT_DELAYLOG (1ULL << 1) /* delayed logging is enabled */
223#define XFS_MOUNT_WAS_CLEAN (1ULL << 3)
224#define XFS_MOUNT_FS_SHUTDOWN (1ULL << 4) /* atomic stop of all filesystem
225 operations, typically for
226 disk errors in metadata */
227#define XFS_MOUNT_DISCARD (1ULL << 5) /* discard unused blocks */
228#define XFS_MOUNT_RETERR (1ULL << 6) /* return alignment errors to
229 user */
230#define XFS_MOUNT_NOALIGN (1ULL << 7) /* turn off stripe alignment
231 allocations */
232#define XFS_MOUNT_ATTR2 (1ULL << 8) /* allow use of attr2 format */
233#define XFS_MOUNT_GRPID (1ULL << 9) /* group-ID assigned from directory */
234#define XFS_MOUNT_NORECOVERY (1ULL << 10) /* no recovery - dirty fs */
235#define XFS_MOUNT_DFLT_IOSIZE (1ULL << 12) /* set default i/o size */
236#define XFS_MOUNT_32BITINODES (1ULL << 14) /* do not create inodes above
237 * 32 bits in size */
238#define XFS_MOUNT_SMALL_INUMS (1ULL << 15) /* users wants 32bit inodes */
239#define XFS_MOUNT_NOUUID (1ULL << 16) /* ignore uuid during mount */
240#define XFS_MOUNT_BARRIER (1ULL << 17)
241#define XFS_MOUNT_IKEEP (1ULL << 18) /* keep empty inode clusters*/
242#define XFS_MOUNT_SWALLOC (1ULL << 19) /* turn on stripe width
243 * allocation */
244#define XFS_MOUNT_RDONLY (1ULL << 20) /* read-only fs */
245#define XFS_MOUNT_DIRSYNC (1ULL << 21) /* synchronous directory ops */
246#define XFS_MOUNT_COMPAT_IOSIZE (1ULL << 22) /* don't report large preferred
247 * I/O size in stat() */
248#define XFS_MOUNT_FILESTREAMS (1ULL << 24) /* enable the filestreams
249 allocator */
250#define XFS_MOUNT_NOATTR2 (1ULL << 25) /* disable use of attr2 format */
251
252
253/*
254 * Default minimum read and write sizes.
255 */
256#define XFS_READIO_LOG_LARGE 16
257#define XFS_WRITEIO_LOG_LARGE 16
258
259/*
260 * Max and min values for mount-option defined I/O
261 * preallocation sizes.
262 */
263#define XFS_MAX_IO_LOG 30 /* 1G */
264#define XFS_MIN_IO_LOG PAGE_SHIFT
265
266/*
267 * Synchronous read and write sizes. This should be
268 * better for NFSv2 wsync filesystems.
269 */
270#define XFS_WSYNC_READIO_LOG 15 /* 32k */
271#define XFS_WSYNC_WRITEIO_LOG 14 /* 16k */
272
273/*
274 * Allow large block sizes to be reported to userspace programs if the
275 * "largeio" mount option is used.
276 *
277 * If compatibility mode is specified, simply return the basic unit of caching
278 * so that we don't get inefficient read/modify/write I/O from user apps.
279 * Otherwise....
280 *
281 * If the underlying volume is a stripe, then return the stripe width in bytes
282 * as the recommended I/O size. It is not a stripe and we've set a default
283 * buffered I/O size, return that, otherwise return the compat default.
284 */
285static inline unsigned long
286xfs_preferred_iosize(xfs_mount_t *mp)
287{
288 if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
289 return PAGE_CACHE_SIZE;
290 return (mp->m_swidth ?
291 (mp->m_swidth << mp->m_sb.sb_blocklog) :
292 ((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
293 (1 << (int)MAX(mp->m_readio_log, mp->m_writeio_log)) :
294 PAGE_CACHE_SIZE));
295}
296
297#define XFS_MAXIOFFSET(mp) ((mp)->m_maxioffset)
298
299#define XFS_LAST_UNMOUNT_WAS_CLEAN(mp) \
300 ((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
301#define XFS_FORCED_SHUTDOWN(mp) ((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN)
302void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
303 int lnnum);
304#define xfs_force_shutdown(m,f) \
305 xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
306
307#define SHUTDOWN_META_IO_ERROR 0x0001 /* write attempt to metadata failed */
308#define SHUTDOWN_LOG_IO_ERROR 0x0002 /* write attempt to the log failed */
309#define SHUTDOWN_FORCE_UMOUNT 0x0004 /* shutdown from a forced unmount */
310#define SHUTDOWN_CORRUPT_INCORE 0x0008 /* corrupt in-memory data structures */
311#define SHUTDOWN_REMOTE_REQ 0x0010 /* shutdown came from remote cell */
312#define SHUTDOWN_DEVICE_REQ 0x0020 /* failed all paths to the device */
313
314#define xfs_test_for_freeze(mp) ((mp)->m_super->s_frozen)
315#define xfs_wait_for_freeze(mp,l) vfs_check_frozen((mp)->m_super, (l))
316
317/*
318 * Flags for xfs_mountfs
319 */
320#define XFS_MFSI_QUIET 0x40 /* Be silent if mount errors found */
321
322static inline xfs_agnumber_t
323xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
324{
325 xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
326 do_div(ld, mp->m_sb.sb_agblocks);
327 return (xfs_agnumber_t) ld;
328}
329
330static inline xfs_agblock_t
331xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
332{
333 xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
334 return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
335}
336
337/*
338 * perag get/put wrappers for ref counting
339 */
340struct xfs_perag *xfs_perag_get(struct xfs_mount *mp, xfs_agnumber_t agno);
341struct xfs_perag *xfs_perag_get_tag(struct xfs_mount *mp, xfs_agnumber_t agno,
342 int tag);
343void xfs_perag_put(struct xfs_perag *pag);
344
345/*
346 * Per-cpu superblock locking functions
347 */
348#ifdef HAVE_PERCPU_SB
349static inline void
350xfs_icsb_lock(xfs_mount_t *mp)
351{
352 mutex_lock(&mp->m_icsb_mutex);
353}
354
355static inline void
356xfs_icsb_unlock(xfs_mount_t *mp)
357{
358 mutex_unlock(&mp->m_icsb_mutex);
359}
360#else
361#define xfs_icsb_lock(mp)
362#define xfs_icsb_unlock(mp)
363#endif
364
365/*
366 * This structure is for use by the xfs_mod_incore_sb_batch() routine.
367 * xfs_growfs can specify a few fields which are more than int limit
368 */
369typedef struct xfs_mod_sb {
370 xfs_sb_field_t msb_field; /* Field to modify, see below */
371 int64_t msb_delta; /* Change to make to specified field */
372} xfs_mod_sb_t;
373
374extern int xfs_log_sbcount(xfs_mount_t *);
375extern __uint64_t xfs_default_resblks(xfs_mount_t *mp);
376extern int xfs_mountfs(xfs_mount_t *mp);
377
378extern void xfs_unmountfs(xfs_mount_t *);
379extern int xfs_unmountfs_writesb(xfs_mount_t *);
380extern int xfs_mod_incore_sb(xfs_mount_t *, xfs_sb_field_t, int64_t, int);
381extern int xfs_mod_incore_sb_batch(xfs_mount_t *, xfs_mod_sb_t *,
382 uint, int);
383extern int xfs_mount_log_sb(xfs_mount_t *, __int64_t);
384extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
385extern int xfs_readsb(xfs_mount_t *, int);
386extern void xfs_freesb(xfs_mount_t *);
387extern int xfs_fs_writable(xfs_mount_t *);
388extern int xfs_sb_validate_fsb_count(struct xfs_sb *, __uint64_t);
389
390extern int xfs_dev_is_read_only(struct xfs_mount *, char *);
391
392extern void xfs_set_low_space_thresholds(struct xfs_mount *);
393
394#endif /* __KERNEL__ */
395
396extern void xfs_mod_sb(struct xfs_trans *, __int64_t);
397extern int xfs_initialize_perag(struct xfs_mount *, xfs_agnumber_t,
398 xfs_agnumber_t *);
399extern void xfs_sb_from_disk(struct xfs_sb *, struct xfs_dsb *);
400extern void xfs_sb_to_disk(struct xfs_dsb *, struct xfs_sb *, __int64_t);
401
402#endif /* __XFS_MOUNT_H__ */
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
9struct xlog;
10struct xfs_inode;
11struct xfs_mru_cache;
12struct xfs_ail;
13struct xfs_quotainfo;
14struct xfs_da_geometry;
15struct xfs_perag;
16
17/* dynamic preallocation free space thresholds, 5% down to 1% */
18enum {
19 XFS_LOWSP_1_PCNT = 0,
20 XFS_LOWSP_2_PCNT,
21 XFS_LOWSP_3_PCNT,
22 XFS_LOWSP_4_PCNT,
23 XFS_LOWSP_5_PCNT,
24 XFS_LOWSP_MAX,
25};
26
27/*
28 * Error Configuration
29 *
30 * Error classes define the subsystem the configuration belongs to.
31 * Error numbers define the errors that are configurable.
32 */
33enum {
34 XFS_ERR_METADATA,
35 XFS_ERR_CLASS_MAX,
36};
37enum {
38 XFS_ERR_DEFAULT,
39 XFS_ERR_EIO,
40 XFS_ERR_ENOSPC,
41 XFS_ERR_ENODEV,
42 XFS_ERR_ERRNO_MAX,
43};
44
45#define XFS_ERR_RETRY_FOREVER -1
46
47/*
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.
52 */
53struct xfs_error_cfg {
54 struct xfs_kobj kobj;
55 int max_retries;
56 long retry_timeout; /* in jiffies, -1 = infinite */
57};
58
59/*
60 * Per-cpu deferred inode inactivation GC lists.
61 */
62struct xfs_inodegc {
63 struct llist_head list;
64 struct delayed_work work;
65
66 /* approximate count of inodes in the list */
67 unsigned int items;
68 unsigned int shrinker_hits;
69};
70
71/*
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.
75 *
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.
79 */
80typedef 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 */
99
100 /*
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
104 * inode lock.
105 */
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;
114
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 */
157 bool m_always_cow;
158 bool m_fail_unmount;
159 bool m_finobt_nores; /* no per-AG finobt resv. */
160 bool m_update_sb; /* sb needs update in mount */
161
162 /*
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.
165 */
166 uint8_t m_fs_checked;
167 uint8_t m_fs_sick;
168 /*
169 * Bitsets of rt metadata that have been checked and/or are sick.
170 * Callers must hold m_sb_lock to access this field.
171 */
172 uint8_t m_rt_checked;
173 uint8_t m_rt_sick;
174
175 /*
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.
180 */
181
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 */
186 struct percpu_counter m_frextents; /* free rt extent counter */
187
188 /*
189 * Count of data device blocks reserved for delayed allocations,
190 * including indlen blocks. Does not include allocated CoW staging
191 * extents or anything related to the rt device.
192 */
193 struct percpu_counter m_delalloc_blks;
194 /*
195 * Global count of allocation btree blocks in use across all AGs. Only
196 * used when perag reservation is enabled. Helps prevent block
197 * reservation from attempting to reserve allocation btree blocks.
198 */
199 atomic64_t m_allocbt_blks;
200
201 struct radix_tree_root m_perag_tree; /* per-ag accounting info */
202 spinlock_t m_perag_lock; /* lock for m_perag_tree */
203 uint64_t m_resblks; /* total reserved blocks */
204 uint64_t m_resblks_avail;/* available reserved blocks */
205 uint64_t m_resblks_save; /* reserved blks @ remount,ro */
206 struct delayed_work m_reclaim_work; /* background inode reclaim */
207 struct xfs_kobj m_kobj;
208 struct xfs_kobj m_error_kobj;
209 struct xfs_kobj m_error_meta_kobj;
210 struct xfs_error_cfg m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX];
211 struct xstats m_stats; /* per-fs stats */
212 xfs_agnumber_t m_agfrotor; /* last ag where space found */
213 xfs_agnumber_t m_agirotor; /* last ag dir inode alloced */
214 spinlock_t m_agirotor_lock;/* .. and lock protecting it */
215
216 /* Memory shrinker to throttle and reprioritize inodegc */
217 struct shrinker m_inodegc_shrinker;
218 /*
219 * Workqueue item so that we can coalesce multiple inode flush attempts
220 * into a single flush.
221 */
222 struct work_struct m_flush_inodes_work;
223
224 /*
225 * Generation of the filesysyem layout. This is incremented by each
226 * growfs, and used by the pNFS server to ensure the client updates
227 * its view of the block device once it gets a layout that might
228 * reference the newly added blocks. Does not need to be persistent
229 * as long as we only allow file system size increments, but if we
230 * ever support shrinks it would have to be persisted in addition
231 * to various other kinds of pain inflicted on the pNFS server.
232 */
233 uint32_t m_generation;
234 struct mutex m_growlock; /* growfs mutex */
235
236#ifdef DEBUG
237 /*
238 * Frequency with which errors are injected. Replaces xfs_etest; the
239 * value stored in here is the inverse of the frequency with which the
240 * error triggers. 1 = always, 2 = half the time, etc.
241 */
242 unsigned int *m_errortag;
243 struct xfs_kobj m_errortag_kobj;
244#endif
245} xfs_mount_t;
246
247#define M_IGEO(mp) (&(mp)->m_ino_geo)
248
249/*
250 * Flags for m_features.
251 *
252 * These are all the active features in the filesystem, regardless of how
253 * they are configured.
254 */
255#define XFS_FEAT_ATTR (1ULL << 0) /* xattrs present in fs */
256#define XFS_FEAT_NLINK (1ULL << 1) /* 32 bit link counts */
257#define XFS_FEAT_QUOTA (1ULL << 2) /* quota active */
258#define XFS_FEAT_ALIGN (1ULL << 3) /* inode alignment */
259#define XFS_FEAT_DALIGN (1ULL << 4) /* data alignment */
260#define XFS_FEAT_LOGV2 (1ULL << 5) /* version 2 logs */
261#define XFS_FEAT_SECTOR (1ULL << 6) /* sector size > 512 bytes */
262#define XFS_FEAT_EXTFLG (1ULL << 7) /* unwritten extents */
263#define XFS_FEAT_ASCIICI (1ULL << 8) /* ASCII only case-insens. */
264#define XFS_FEAT_LAZYSBCOUNT (1ULL << 9) /* Superblk counters */
265#define XFS_FEAT_ATTR2 (1ULL << 10) /* dynamic attr fork */
266#define XFS_FEAT_PARENT (1ULL << 11) /* parent pointers */
267#define XFS_FEAT_PROJID32 (1ULL << 12) /* 32 bit project id */
268#define XFS_FEAT_CRC (1ULL << 13) /* metadata CRCs */
269#define XFS_FEAT_V3INODES (1ULL << 14) /* Version 3 inodes */
270#define XFS_FEAT_PQUOTINO (1ULL << 15) /* non-shared proj/grp quotas */
271#define XFS_FEAT_FTYPE (1ULL << 16) /* inode type in dir */
272#define XFS_FEAT_FINOBT (1ULL << 17) /* free inode btree */
273#define XFS_FEAT_RMAPBT (1ULL << 18) /* reverse map btree */
274#define XFS_FEAT_REFLINK (1ULL << 19) /* reflinked files */
275#define XFS_FEAT_SPINODES (1ULL << 20) /* sparse inode chunks */
276#define XFS_FEAT_META_UUID (1ULL << 21) /* metadata UUID */
277#define XFS_FEAT_REALTIME (1ULL << 22) /* realtime device present */
278#define XFS_FEAT_INOBTCNT (1ULL << 23) /* inobt block counts */
279#define XFS_FEAT_BIGTIME (1ULL << 24) /* large timestamps */
280#define XFS_FEAT_NEEDSREPAIR (1ULL << 25) /* needs xfs_repair */
281#define XFS_FEAT_NREXT64 (1ULL << 26) /* large extent counters */
282
283/* Mount features */
284#define XFS_FEAT_NOATTR2 (1ULL << 48) /* disable attr2 creation */
285#define XFS_FEAT_NOALIGN (1ULL << 49) /* ignore alignment */
286#define XFS_FEAT_ALLOCSIZE (1ULL << 50) /* user specified allocation size */
287#define XFS_FEAT_LARGE_IOSIZE (1ULL << 51) /* report large preferred
288 * I/O size in stat() */
289#define XFS_FEAT_WSYNC (1ULL << 52) /* synchronous metadata ops */
290#define XFS_FEAT_DIRSYNC (1ULL << 53) /* synchronous directory ops */
291#define XFS_FEAT_DISCARD (1ULL << 54) /* discard unused blocks */
292#define XFS_FEAT_GRPID (1ULL << 55) /* group-ID assigned from directory */
293#define XFS_FEAT_SMALL_INUMS (1ULL << 56) /* user wants 32bit inodes */
294#define XFS_FEAT_IKEEP (1ULL << 57) /* keep empty inode clusters*/
295#define XFS_FEAT_SWALLOC (1ULL << 58) /* stripe width allocation */
296#define XFS_FEAT_FILESTREAMS (1ULL << 59) /* use filestreams allocator */
297#define XFS_FEAT_DAX_ALWAYS (1ULL << 60) /* DAX always enabled */
298#define XFS_FEAT_DAX_NEVER (1ULL << 61) /* DAX never enabled */
299#define XFS_FEAT_NORECOVERY (1ULL << 62) /* no recovery - dirty fs */
300#define XFS_FEAT_NOUUID (1ULL << 63) /* ignore uuid during mount */
301
302#define __XFS_HAS_FEAT(name, NAME) \
303static inline bool xfs_has_ ## name (struct xfs_mount *mp) \
304{ \
305 return mp->m_features & XFS_FEAT_ ## NAME; \
306}
307
308/* Some features can be added dynamically so they need a set wrapper, too. */
309#define __XFS_ADD_FEAT(name, NAME) \
310 __XFS_HAS_FEAT(name, NAME); \
311static inline void xfs_add_ ## name (struct xfs_mount *mp) \
312{ \
313 mp->m_features |= XFS_FEAT_ ## NAME; \
314 xfs_sb_version_add ## name(&mp->m_sb); \
315}
316
317/* Superblock features */
318__XFS_ADD_FEAT(attr, ATTR)
319__XFS_HAS_FEAT(nlink, NLINK)
320__XFS_ADD_FEAT(quota, QUOTA)
321__XFS_HAS_FEAT(align, ALIGN)
322__XFS_HAS_FEAT(dalign, DALIGN)
323__XFS_HAS_FEAT(logv2, LOGV2)
324__XFS_HAS_FEAT(sector, SECTOR)
325__XFS_HAS_FEAT(extflg, EXTFLG)
326__XFS_HAS_FEAT(asciici, ASCIICI)
327__XFS_HAS_FEAT(lazysbcount, LAZYSBCOUNT)
328__XFS_ADD_FEAT(attr2, ATTR2)
329__XFS_HAS_FEAT(parent, PARENT)
330__XFS_ADD_FEAT(projid32, PROJID32)
331__XFS_HAS_FEAT(crc, CRC)
332__XFS_HAS_FEAT(v3inodes, V3INODES)
333__XFS_HAS_FEAT(pquotino, PQUOTINO)
334__XFS_HAS_FEAT(ftype, FTYPE)
335__XFS_HAS_FEAT(finobt, FINOBT)
336__XFS_HAS_FEAT(rmapbt, RMAPBT)
337__XFS_HAS_FEAT(reflink, REFLINK)
338__XFS_HAS_FEAT(sparseinodes, SPINODES)
339__XFS_HAS_FEAT(metauuid, META_UUID)
340__XFS_HAS_FEAT(realtime, REALTIME)
341__XFS_HAS_FEAT(inobtcounts, INOBTCNT)
342__XFS_HAS_FEAT(bigtime, BIGTIME)
343__XFS_HAS_FEAT(needsrepair, NEEDSREPAIR)
344__XFS_HAS_FEAT(large_extent_counts, NREXT64)
345
346/*
347 * Mount features
348 *
349 * These do not change dynamically - features that can come and go, such as 32
350 * bit inodes and read-only state, are kept as operational state rather than
351 * features.
352 */
353__XFS_HAS_FEAT(noattr2, NOATTR2)
354__XFS_HAS_FEAT(noalign, NOALIGN)
355__XFS_HAS_FEAT(allocsize, ALLOCSIZE)
356__XFS_HAS_FEAT(large_iosize, LARGE_IOSIZE)
357__XFS_HAS_FEAT(wsync, WSYNC)
358__XFS_HAS_FEAT(dirsync, DIRSYNC)
359__XFS_HAS_FEAT(discard, DISCARD)
360__XFS_HAS_FEAT(grpid, GRPID)
361__XFS_HAS_FEAT(small_inums, SMALL_INUMS)
362__XFS_HAS_FEAT(ikeep, IKEEP)
363__XFS_HAS_FEAT(swalloc, SWALLOC)
364__XFS_HAS_FEAT(filestreams, FILESTREAMS)
365__XFS_HAS_FEAT(dax_always, DAX_ALWAYS)
366__XFS_HAS_FEAT(dax_never, DAX_NEVER)
367__XFS_HAS_FEAT(norecovery, NORECOVERY)
368__XFS_HAS_FEAT(nouuid, NOUUID)
369
370/*
371 * Operational mount state flags
372 *
373 * Use these with atomic bit ops only!
374 */
375#define XFS_OPSTATE_UNMOUNTING 0 /* filesystem is unmounting */
376#define XFS_OPSTATE_CLEAN 1 /* mount was clean */
377#define XFS_OPSTATE_SHUTDOWN 2 /* stop all fs operations */
378#define XFS_OPSTATE_INODE32 3 /* inode32 allocator active */
379#define XFS_OPSTATE_READONLY 4 /* read-only fs */
380
381/*
382 * If set, inactivation worker threads will be scheduled to process queued
383 * inodegc work. If not, queued inodes remain in memory waiting to be
384 * processed.
385 */
386#define XFS_OPSTATE_INODEGC_ENABLED 5
387/*
388 * If set, background speculative prealloc gc worker threads will be scheduled
389 * to process queued blockgc work. If not, inodes retain their preallocations
390 * until explicitly deleted.
391 */
392#define XFS_OPSTATE_BLOCKGC_ENABLED 6
393
394/* Kernel has logged a warning about online fsck being used on this fs. */
395#define XFS_OPSTATE_WARNED_SCRUB 7
396/* Kernel has logged a warning about shrink being used on this fs. */
397#define XFS_OPSTATE_WARNED_SHRINK 8
398/* Kernel has logged a warning about logged xattr updates being used. */
399#define XFS_OPSTATE_WARNED_LARP 9
400
401#define __XFS_IS_OPSTATE(name, NAME) \
402static inline bool xfs_is_ ## name (struct xfs_mount *mp) \
403{ \
404 return test_bit(XFS_OPSTATE_ ## NAME, &mp->m_opstate); \
405} \
406static inline bool xfs_clear_ ## name (struct xfs_mount *mp) \
407{ \
408 return test_and_clear_bit(XFS_OPSTATE_ ## NAME, &mp->m_opstate); \
409} \
410static inline bool xfs_set_ ## name (struct xfs_mount *mp) \
411{ \
412 return test_and_set_bit(XFS_OPSTATE_ ## NAME, &mp->m_opstate); \
413}
414
415__XFS_IS_OPSTATE(unmounting, UNMOUNTING)
416__XFS_IS_OPSTATE(clean, CLEAN)
417__XFS_IS_OPSTATE(shutdown, SHUTDOWN)
418__XFS_IS_OPSTATE(inode32, INODE32)
419__XFS_IS_OPSTATE(readonly, READONLY)
420__XFS_IS_OPSTATE(inodegc_enabled, INODEGC_ENABLED)
421__XFS_IS_OPSTATE(blockgc_enabled, BLOCKGC_ENABLED)
422
423static inline bool
424xfs_should_warn(struct xfs_mount *mp, long nr)
425{
426 return !test_and_set_bit(nr, &mp->m_opstate);
427}
428
429#define XFS_OPSTATE_STRINGS \
430 { (1UL << XFS_OPSTATE_UNMOUNTING), "unmounting" }, \
431 { (1UL << XFS_OPSTATE_CLEAN), "clean" }, \
432 { (1UL << XFS_OPSTATE_SHUTDOWN), "shutdown" }, \
433 { (1UL << XFS_OPSTATE_INODE32), "inode32" }, \
434 { (1UL << XFS_OPSTATE_READONLY), "read_only" }, \
435 { (1UL << XFS_OPSTATE_INODEGC_ENABLED), "inodegc" }, \
436 { (1UL << XFS_OPSTATE_BLOCKGC_ENABLED), "blockgc" }, \
437 { (1UL << XFS_OPSTATE_WARNED_SCRUB), "wscrub" }, \
438 { (1UL << XFS_OPSTATE_WARNED_SHRINK), "wshrink" }, \
439 { (1UL << XFS_OPSTATE_WARNED_LARP), "wlarp" }
440
441/*
442 * Max and min values for mount-option defined I/O
443 * preallocation sizes.
444 */
445#define XFS_MAX_IO_LOG 30 /* 1G */
446#define XFS_MIN_IO_LOG PAGE_SHIFT
447
448void xfs_do_force_shutdown(struct xfs_mount *mp, uint32_t flags, char *fname,
449 int lnnum);
450#define xfs_force_shutdown(m,f) \
451 xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
452
453#define SHUTDOWN_META_IO_ERROR (1u << 0) /* write attempt to metadata failed */
454#define SHUTDOWN_LOG_IO_ERROR (1u << 1) /* write attempt to the log failed */
455#define SHUTDOWN_FORCE_UMOUNT (1u << 2) /* shutdown from a forced unmount */
456#define SHUTDOWN_CORRUPT_INCORE (1u << 3) /* corrupt in-memory structures */
457#define SHUTDOWN_CORRUPT_ONDISK (1u << 4) /* corrupt metadata on device */
458
459#define XFS_SHUTDOWN_STRINGS \
460 { SHUTDOWN_META_IO_ERROR, "metadata_io" }, \
461 { SHUTDOWN_LOG_IO_ERROR, "log_io" }, \
462 { SHUTDOWN_FORCE_UMOUNT, "force_umount" }, \
463 { SHUTDOWN_CORRUPT_INCORE, "corruption" }
464
465/*
466 * Flags for xfs_mountfs
467 */
468#define XFS_MFSI_QUIET 0x40 /* Be silent if mount errors found */
469
470static inline xfs_agnumber_t
471xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
472{
473 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
474 do_div(ld, mp->m_sb.sb_agblocks);
475 return (xfs_agnumber_t) ld;
476}
477
478static inline xfs_agblock_t
479xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
480{
481 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
482 return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
483}
484
485int xfs_buf_hash_init(struct xfs_perag *pag);
486void xfs_buf_hash_destroy(struct xfs_perag *pag);
487
488extern void xfs_uuid_table_free(void);
489extern uint64_t xfs_default_resblks(xfs_mount_t *mp);
490extern int xfs_mountfs(xfs_mount_t *mp);
491extern void xfs_unmountfs(xfs_mount_t *);
492
493/*
494 * Deltas for the block count can vary from 1 to very large, but lock contention
495 * only occurs on frequent small block count updates such as in the delayed
496 * allocation path for buffered writes (page a time updates). Hence we set
497 * a large batch count (1024) to minimise global counter updates except when
498 * we get near to ENOSPC and we have to be very accurate with our updates.
499 */
500#define XFS_FDBLOCKS_BATCH 1024
501
502/*
503 * Estimate the amount of free space that is not available to userspace and is
504 * not explicitly reserved from the incore fdblocks. This includes:
505 *
506 * - The minimum number of blocks needed to support splitting a bmap btree
507 * - The blocks currently in use by the freespace btrees because they record
508 * the actual blocks that will fill per-AG metadata space reservations
509 */
510static inline uint64_t
511xfs_fdblocks_unavailable(
512 struct xfs_mount *mp)
513{
514 return mp->m_alloc_set_aside + atomic64_read(&mp->m_allocbt_blks);
515}
516
517int xfs_mod_freecounter(struct xfs_mount *mp, struct percpu_counter *counter,
518 int64_t delta, bool rsvd);
519
520static inline int
521xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta, bool reserved)
522{
523 return xfs_mod_freecounter(mp, &mp->m_fdblocks, delta, reserved);
524}
525
526static inline int
527xfs_mod_frextents(struct xfs_mount *mp, int64_t delta)
528{
529 return xfs_mod_freecounter(mp, &mp->m_frextents, delta, false);
530}
531
532extern int xfs_readsb(xfs_mount_t *, int);
533extern void xfs_freesb(xfs_mount_t *);
534extern bool xfs_fs_writable(struct xfs_mount *mp, int level);
535extern int xfs_sb_validate_fsb_count(struct xfs_sb *, uint64_t);
536
537extern int xfs_dev_is_read_only(struct xfs_mount *, char *);
538
539extern void xfs_set_low_space_thresholds(struct xfs_mount *);
540
541int xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb,
542 xfs_off_t count_fsb);
543
544struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp,
545 int error_class, int error);
546void xfs_force_summary_recalc(struct xfs_mount *mp);
547int xfs_add_incompat_log_feature(struct xfs_mount *mp, uint32_t feature);
548bool xfs_clear_incompat_log_features(struct xfs_mount *mp);
549void xfs_mod_delalloc(struct xfs_mount *mp, int64_t delta);
550
551#endif /* __XFS_MOUNT_H__ */