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_ail;
13 struct xfs_quotainfo;
14 struct xfs_da_geometry;
15 struct xfs_perag;
16 struct xfs_healthmon;
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 /*
61 * Per-cpu deferred inode inactivation GC lists.
62 */
63 struct xfs_inodegc {
64 struct xfs_mount *mp;
65 struct llist_head list;
66 struct delayed_work work;
67 int error;
68
69 /* approximate count of inodes in the list */
70 unsigned int items;
71 unsigned int shrinker_hits;
72 unsigned int cpu;
73 };
74
75 /*
76 * Container for each type of groups, used to look up individual groups and
77 * describes the geometry.
78 */
79 struct xfs_groups {
80 struct xarray xa;
81
82 /*
83 * Maximum capacity of the group in FSBs.
84 *
85 * Each group is laid out densely in the daddr space. For the
86 * degenerate case of a pre-rtgroups filesystem, the incore rtgroup
87 * pretends to have a zero-block and zero-blklog rtgroup.
88 */
89 uint32_t blocks;
90
91 /*
92 * Log(2) of the logical size of each group.
93 *
94 * Compared to the blocks field above this is rounded up to the next
95 * power of two, and thus lays out the xfs_fsblock_t/xfs_rtblock_t
96 * space sparsely with a hole from blocks to (1 << blklog) at the end
97 * of each group.
98 */
99 uint8_t blklog;
100
101 /*
102 * Zoned devices can have gaps beyond the usable capacity of a zone and
103 * the end in the LBA/daddr address space. In other words, the hardware
104 * equivalent to the RT groups already takes care of the power of 2
105 * alignment for us. In this case the sparse FSB/RTB address space maps
106 * 1:1 to the device address space.
107 */
108 bool has_daddr_gaps;
109
110 /*
111 * Mask to extract the group-relative block number from a FSB.
112 * For a pre-rtgroups filesystem we pretend to have one very large
113 * rtgroup, so this mask must be 64-bit.
114 */
115 uint64_t blkmask;
116
117 /*
118 * Start of the first group in the device. This is used to support a
119 * RT device following the data device on the same block device for
120 * SMR hard drives.
121 */
122 xfs_fsblock_t start_fsb;
123
124 /*
125 * Maximum length of an atomic write for files stored in this
126 * collection of allocation groups, in fsblocks.
127 */
128 xfs_extlen_t awu_max;
129 };
130
131 struct xfs_freecounter {
132 /* free blocks for general use: */
133 struct percpu_counter count;
134
135 /* total reserved blocks: */
136 uint64_t res_total;
137
138 /* available reserved blocks: */
139 uint64_t res_avail;
140
141 /* reserved blks @ remount,ro: */
142 uint64_t res_saved;
143 };
144
145 /*
146 * The struct xfsmount layout is optimised to separate read-mostly variables
147 * from variables that are frequently modified. We put the read-mostly variables
148 * first, then place all the other variables at the end.
149 *
150 * Typically, read-mostly variables are those that are set at mount time and
151 * never changed again, or only change rarely as a result of things like sysfs
152 * knobs being tweaked.
153 */
154 typedef struct xfs_mount {
155 struct xfs_sb m_sb; /* copy of fs superblock */
156 struct super_block *m_super;
157 struct xfs_ail *m_ail; /* fs active log item list */
158 struct xfs_buf *m_sb_bp; /* buffer for superblock */
159 struct xfs_buf *m_rtsb_bp; /* realtime superblock */
160 char *m_rtname; /* realtime device name */
161 char *m_logname; /* external log device name */
162 struct xfs_da_geometry *m_dir_geo; /* directory block geometry */
163 struct xfs_da_geometry *m_attr_geo; /* attribute block geometry */
164 struct xlog *m_log; /* log specific stuff */
165 struct xfs_inode *m_rootip; /* pointer to root directory */
166 struct xfs_inode *m_metadirip; /* ptr to metadata directory */
167 struct xfs_inode *m_rtdirip; /* ptr to realtime metadir */
168 struct xfs_quotainfo *m_quotainfo; /* disk quota information */
169 struct xfs_buftarg *m_ddev_targp; /* data device */
170 struct xfs_buftarg *m_logdev_targp;/* log device */
171 struct xfs_buftarg *m_rtdev_targp; /* rt device */
172 void __percpu *m_inodegc; /* percpu inodegc structures */
173 struct xfs_mru_cache *m_filestream; /* per-mount filestream data */
174 struct workqueue_struct *m_buf_workqueue;
175 struct workqueue_struct *m_unwritten_workqueue;
176 struct workqueue_struct *m_reclaim_workqueue;
177 struct workqueue_struct *m_sync_workqueue;
178 struct workqueue_struct *m_blockgc_wq;
179 struct workqueue_struct *m_inodegc_wq;
180
181 int m_bsize; /* fs logical block size */
182 uint8_t m_blkbit_log; /* blocklog + NBBY */
183 uint8_t m_blkbb_log; /* blocklog - BBSHIFT */
184 uint8_t m_agno_log; /* log #ag's */
185 uint8_t m_sectbb_log; /* sectlog - BBSHIFT */
186 int8_t m_rtxblklog; /* log2 of rextsize, if possible */
187
188 uint m_blockmask; /* sb_blocksize-1 */
189 uint m_blockwsize; /* sb_blocksize in words */
190 /* number of rt extents per rt bitmap block if rtgroups enabled */
191 unsigned int m_rtx_per_rbmblock;
192 uint m_alloc_mxr[2]; /* max alloc btree records */
193 uint m_alloc_mnr[2]; /* min alloc btree records */
194 uint m_bmap_dmxr[2]; /* max bmap btree records */
195 uint m_bmap_dmnr[2]; /* min bmap btree records */
196 uint m_rmap_mxr[2]; /* max rmap btree records */
197 uint m_rmap_mnr[2]; /* min rmap btree records */
198 uint m_rtrmap_mxr[2]; /* max rtrmap btree records */
199 uint m_rtrmap_mnr[2]; /* min rtrmap btree records */
200 uint m_refc_mxr[2]; /* max refc btree records */
201 uint m_refc_mnr[2]; /* min refc btree records */
202 uint m_rtrefc_mxr[2]; /* max rtrefc btree records */
203 uint m_rtrefc_mnr[2]; /* min rtrefc btree records */
204 uint m_alloc_maxlevels; /* max alloc btree levels */
205 uint m_bm_maxlevels[2]; /* max bmap btree levels */
206 uint m_rmap_maxlevels; /* max rmap btree levels */
207 uint m_rtrmap_maxlevels; /* max rtrmap btree level */
208 uint m_refc_maxlevels; /* max refcount btree level */
209 uint m_rtrefc_maxlevels; /* max rtrefc btree level */
210 unsigned int m_agbtree_maxlevels; /* max level of all AG btrees */
211 unsigned int m_rtbtree_maxlevels; /* max level of all rt btrees */
212 xfs_extlen_t m_ag_prealloc_blocks; /* reserved ag blocks */
213 uint m_alloc_set_aside; /* space we can't use */
214 uint m_ag_max_usable; /* max space per AG */
215 int m_dalign; /* stripe unit */
216 int m_swidth; /* stripe width */
217 xfs_agnumber_t m_maxagi; /* highest inode alloc group */
218 uint m_allocsize_log;/* min write size log bytes */
219 uint m_allocsize_blocks; /* min write size blocks */
220 int m_logbufs; /* number of log buffers */
221 int m_logbsize; /* size of each log buffer */
222 unsigned int m_rsumlevels; /* rt summary levels */
223 xfs_filblks_t m_rsumblocks; /* size of rt summary, FSBs */
224 int m_fixedfsid[2]; /* unchanged for life of FS */
225 uint m_qflags; /* quota status flags */
226 uint64_t m_features; /* active filesystem features */
227 uint64_t m_low_space[XFS_LOWSP_MAX];
228 uint64_t m_low_rtexts[XFS_LOWSP_MAX];
229 uint64_t m_rtxblkmask; /* rt extent block mask */
230 struct xfs_ino_geometry m_ino_geo; /* inode geometry */
231 struct xfs_trans_resv m_resv; /* precomputed res values */
232 /* low free space thresholds */
233 unsigned long m_opstate; /* dynamic state flags */
234 bool m_always_cow;
235 bool m_fail_unmount;
236 bool m_finobt_nores; /* no per-AG finobt resv. */
237 bool m_update_sb; /* sb needs update in mount */
238 unsigned int m_max_open_zones;
239 unsigned int m_zonegc_low_space;
240
241 /* max_atomic_write mount option value */
242 unsigned long long m_awu_max_bytes;
243
244 /*
245 * Bitsets of per-fs metadata that have been checked and/or are sick.
246 * Callers must hold m_sb_lock to access these two fields.
247 */
248 uint8_t m_fs_checked;
249 uint8_t m_fs_sick;
250 /*
251 * Bitsets of rt metadata that have been checked and/or are sick.
252 * Callers must hold m_sb_lock to access this field.
253 */
254 uint8_t m_rt_checked;
255 uint8_t m_rt_sick;
256
257 /*
258 * End of read-mostly variables. Frequently written variables and locks
259 * should be placed below this comment from now on. The first variable
260 * here is marked as cacheline aligned so they it is separated from
261 * the read-mostly variables.
262 */
263
264 spinlock_t ____cacheline_aligned m_sb_lock; /* sb counter lock */
265 struct percpu_counter m_icount; /* allocated inodes counter */
266 struct percpu_counter m_ifree; /* free inodes counter */
267
268 struct xfs_freecounter m_free[XC_FREE_NR];
269
270 /*
271 * Count of data device blocks reserved for delayed allocations,
272 * including indlen blocks. Does not include allocated CoW staging
273 * extents or anything related to the rt device.
274 */
275 struct percpu_counter m_delalloc_blks;
276
277 /*
278 * RT version of the above.
279 */
280 struct percpu_counter m_delalloc_rtextents;
281
282 /*
283 * Global count of allocation btree blocks in use across all AGs. Only
284 * used when perag reservation is enabled. Helps prevent block
285 * reservation from attempting to reserve allocation btree blocks.
286 */
287 atomic64_t m_allocbt_blks;
288
289 struct xfs_groups m_groups[XG_TYPE_MAX];
290 struct delayed_work m_reclaim_work; /* background inode reclaim */
291 struct xfs_zone_info *m_zone_info; /* zone allocator information */
292 struct dentry *m_debugfs; /* debugfs parent */
293 struct xfs_kobj m_kobj;
294 struct xfs_kobj m_error_kobj;
295 struct xfs_kobj m_error_meta_kobj;
296 struct xfs_error_cfg m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX];
297 struct xstats m_stats; /* per-fs stats */
298 #ifdef CONFIG_XFS_ONLINE_SCRUB_STATS
299 struct xchk_stats *m_scrub_stats;
300 #endif
301 struct xfs_kobj m_zoned_kobj;
302 xfs_agnumber_t m_agfrotor; /* last ag where space found */
303 atomic_t m_agirotor; /* last ag dir inode alloced */
304 atomic_t m_rtgrotor; /* last rtgroup rtpicked */
305
306 struct mutex m_metafile_resv_lock;
307 uint64_t m_metafile_resv_target;
308 uint64_t m_metafile_resv_used;
309 uint64_t m_metafile_resv_avail;
310
311 /* Memory shrinker to throttle and reprioritize inodegc */
312 struct shrinker *m_inodegc_shrinker;
313 /*
314 * Workqueue item so that we can coalesce multiple inode flush attempts
315 * into a single flush.
316 */
317 struct work_struct m_flush_inodes_work;
318
319 /*
320 * Generation of the filesysyem layout. This is incremented by each
321 * growfs, and used by the pNFS server to ensure the client updates
322 * its view of the block device once it gets a layout that might
323 * reference the newly added blocks. Does not need to be persistent
324 * as long as we only allow file system size increments, but if we
325 * ever support shrinks it would have to be persisted in addition
326 * to various other kinds of pain inflicted on the pNFS server.
327 */
328 uint32_t m_generation;
329 struct mutex m_growlock; /* growfs mutex */
330
331 #ifdef DEBUG
332 /*
333 * Frequency with which errors are injected. Replaces xfs_etest; the
334 * value stored in here is the inverse of the frequency with which the
335 * error triggers. 1 = always, 2 = half the time, etc.
336 */
337 unsigned int *m_errortag;
338 struct xfs_kobj m_errortag_kobj;
339 #endif
340
341 /* cpus that have inodes queued for inactivation */
342 struct cpumask m_inodegc_cpumask;
343
344 /* Hook to feed dirent updates to an active online repair. */
345 struct xfs_hooks m_dir_update_hooks;
346
347 /* Private data referring to a health monitor object. */
348 struct xfs_healthmon __rcu *m_healthmon;
349 } xfs_mount_t;
350
351 #define M_IGEO(mp) (&(mp)->m_ino_geo)
352
353 /*
354 * Flags for m_features.
355 *
356 * These are all the active features in the filesystem, regardless of how
357 * they are configured.
358 */
359 #define XFS_FEAT_ATTR (1ULL << 0) /* xattrs present in fs */
360 #define XFS_FEAT_NLINK (1ULL << 1) /* 32 bit link counts */
361 #define XFS_FEAT_QUOTA (1ULL << 2) /* quota active */
362 #define XFS_FEAT_ALIGN (1ULL << 3) /* inode alignment */
363 #define XFS_FEAT_DALIGN (1ULL << 4) /* data alignment */
364 #define XFS_FEAT_LOGV2 (1ULL << 5) /* version 2 logs */
365 #define XFS_FEAT_SECTOR (1ULL << 6) /* sector size > 512 bytes */
366 #define XFS_FEAT_EXTFLG (1ULL << 7) /* unwritten extents */
367 #define XFS_FEAT_ASCIICI (1ULL << 8) /* ASCII only case-insens. */
368 #define XFS_FEAT_LAZYSBCOUNT (1ULL << 9) /* Superblk counters */
369 #define XFS_FEAT_PARENT (1ULL << 11) /* parent pointers */
370 #define XFS_FEAT_PROJID32 (1ULL << 12) /* 32 bit project id */
371 #define XFS_FEAT_CRC (1ULL << 13) /* metadata CRCs */
372 #define XFS_FEAT_V3INODES (1ULL << 14) /* Version 3 inodes */
373 #define XFS_FEAT_PQUOTINO (1ULL << 15) /* non-shared proj/grp quotas */
374 #define XFS_FEAT_FTYPE (1ULL << 16) /* inode type in dir */
375 #define XFS_FEAT_FINOBT (1ULL << 17) /* free inode btree */
376 #define XFS_FEAT_RMAPBT (1ULL << 18) /* reverse map btree */
377 #define XFS_FEAT_REFLINK (1ULL << 19) /* reflinked files */
378 #define XFS_FEAT_SPINODES (1ULL << 20) /* sparse inode chunks */
379 #define XFS_FEAT_META_UUID (1ULL << 21) /* metadata UUID */
380 #define XFS_FEAT_REALTIME (1ULL << 22) /* realtime device present */
381 #define XFS_FEAT_INOBTCNT (1ULL << 23) /* inobt block counts */
382 #define XFS_FEAT_BIGTIME (1ULL << 24) /* large timestamps */
383 #define XFS_FEAT_NEEDSREPAIR (1ULL << 25) /* needs xfs_repair */
384 #define XFS_FEAT_NREXT64 (1ULL << 26) /* large extent counters */
385 #define XFS_FEAT_EXCHANGE_RANGE (1ULL << 27) /* exchange range */
386 #define XFS_FEAT_METADIR (1ULL << 28) /* metadata directory tree */
387 #define XFS_FEAT_ZONED (1ULL << 29) /* zoned RT device */
388
389 /* Mount features */
390 #define XFS_FEAT_NOLIFETIME (1ULL << 47) /* disable lifetime hints */
391 #define XFS_FEAT_NOALIGN (1ULL << 49) /* ignore alignment */
392 #define XFS_FEAT_ALLOCSIZE (1ULL << 50) /* user specified allocation size */
393 #define XFS_FEAT_LARGE_IOSIZE (1ULL << 51) /* report large preferred
394 * I/O size in stat() */
395 #define XFS_FEAT_WSYNC (1ULL << 52) /* synchronous metadata ops */
396 #define XFS_FEAT_DIRSYNC (1ULL << 53) /* synchronous directory ops */
397 #define XFS_FEAT_DISCARD (1ULL << 54) /* discard unused blocks */
398 #define XFS_FEAT_GRPID (1ULL << 55) /* group-ID assigned from directory */
399 #define XFS_FEAT_SMALL_INUMS (1ULL << 56) /* user wants 32bit inodes */
400 #define XFS_FEAT_SWALLOC (1ULL << 58) /* stripe width allocation */
401 #define XFS_FEAT_FILESTREAMS (1ULL << 59) /* use filestreams allocator */
402 #define XFS_FEAT_DAX_ALWAYS (1ULL << 60) /* DAX always enabled */
403 #define XFS_FEAT_DAX_NEVER (1ULL << 61) /* DAX never enabled */
404 #define XFS_FEAT_NORECOVERY (1ULL << 62) /* no recovery - dirty fs */
405 #define XFS_FEAT_NOUUID (1ULL << 63) /* ignore uuid during mount */
406
407 #define __XFS_HAS_FEAT(name, NAME) \
408 static inline bool xfs_has_ ## name (const struct xfs_mount *mp) \
409 { \
410 return mp->m_features & XFS_FEAT_ ## NAME; \
411 }
412
413 /* Some features can be added dynamically so they need a set wrapper, too. */
414 #define __XFS_ADD_FEAT(name, NAME) \
415 __XFS_HAS_FEAT(name, NAME); \
416 static inline void xfs_add_ ## name (struct xfs_mount *mp) \
417 { \
418 mp->m_features |= XFS_FEAT_ ## NAME; \
419 xfs_sb_version_add ## name(&mp->m_sb); \
420 }
421
422 /* Superblock features */
__XFS_ADD_FEAT(attr,ATTR)423 __XFS_ADD_FEAT(attr, ATTR)
424 __XFS_HAS_FEAT(nlink, NLINK)
425 __XFS_ADD_FEAT(quota, QUOTA)
426 __XFS_HAS_FEAT(dalign, DALIGN)
427 __XFS_HAS_FEAT(sector, SECTOR)
428 __XFS_HAS_FEAT(asciici, ASCIICI)
429 __XFS_HAS_FEAT(parent, PARENT)
430 __XFS_HAS_FEAT(ftype, FTYPE)
431 __XFS_HAS_FEAT(finobt, FINOBT)
432 __XFS_HAS_FEAT(rmapbt, RMAPBT)
433 __XFS_HAS_FEAT(reflink, REFLINK)
434 __XFS_HAS_FEAT(sparseinodes, SPINODES)
435 __XFS_HAS_FEAT(metauuid, META_UUID)
436 __XFS_HAS_FEAT(realtime, REALTIME)
437 __XFS_HAS_FEAT(inobtcounts, INOBTCNT)
438 __XFS_HAS_FEAT(bigtime, BIGTIME)
439 __XFS_HAS_FEAT(needsrepair, NEEDSREPAIR)
440 __XFS_HAS_FEAT(large_extent_counts, NREXT64)
441 __XFS_HAS_FEAT(exchange_range, EXCHANGE_RANGE)
442 __XFS_HAS_FEAT(metadir, METADIR)
443 __XFS_HAS_FEAT(zoned, ZONED)
444 __XFS_HAS_FEAT(nolifetime, NOLIFETIME)
445
446 static inline bool xfs_has_rtgroups(const struct xfs_mount *mp)
447 {
448 /* all metadir file systems also allow rtgroups */
449 return xfs_has_metadir(mp);
450 }
451
xfs_has_rtsb(const struct xfs_mount * mp)452 static inline bool xfs_has_rtsb(const struct xfs_mount *mp)
453 {
454 /* all rtgroups filesystems with an rt section have an rtsb */
455 return xfs_has_rtgroups(mp) &&
456 xfs_has_realtime(mp) &&
457 !xfs_has_zoned(mp);
458 }
459
xfs_has_rtrmapbt(const struct xfs_mount * mp)460 static inline bool xfs_has_rtrmapbt(const struct xfs_mount *mp)
461 {
462 return xfs_has_rtgroups(mp) && xfs_has_realtime(mp) &&
463 xfs_has_rmapbt(mp);
464 }
465
xfs_has_rtreflink(const struct xfs_mount * mp)466 static inline bool xfs_has_rtreflink(const struct xfs_mount *mp)
467 {
468 return xfs_has_metadir(mp) && xfs_has_realtime(mp) &&
469 xfs_has_reflink(mp);
470 }
471
xfs_has_nonzoned(const struct xfs_mount * mp)472 static inline bool xfs_has_nonzoned(const struct xfs_mount *mp)
473 {
474 return !xfs_has_zoned(mp);
475 }
476
xfs_can_sw_atomic_write(struct xfs_mount * mp)477 static inline bool xfs_can_sw_atomic_write(struct xfs_mount *mp)
478 {
479 return xfs_has_reflink(mp);
480 }
481
482 /*
483 * Some features are always on for v5 file systems, allow the compiler to
484 * eliminiate dead code when building without v4 support.
485 */
486 #define __XFS_HAS_V4_FEAT(name, NAME) \
487 static inline bool xfs_has_ ## name (struct xfs_mount *mp) \
488 { \
489 return !IS_ENABLED(CONFIG_XFS_SUPPORT_V4) || \
490 (mp->m_features & XFS_FEAT_ ## NAME); \
491 }
492
493 #define __XFS_ADD_V4_FEAT(name, NAME) \
494 __XFS_HAS_V4_FEAT(name, NAME); \
495 static inline void xfs_add_ ## name (struct xfs_mount *mp) \
496 { \
497 if (IS_ENABLED(CONFIG_XFS_SUPPORT_V4)) { \
498 mp->m_features |= XFS_FEAT_ ## NAME; \
499 xfs_sb_version_add ## name(&mp->m_sb); \
500 } \
501 }
502
__XFS_HAS_V4_FEAT(align,ALIGN)503 __XFS_HAS_V4_FEAT(align, ALIGN)
504 __XFS_HAS_V4_FEAT(logv2, LOGV2)
505 __XFS_HAS_V4_FEAT(extflg, EXTFLG)
506 __XFS_HAS_V4_FEAT(lazysbcount, LAZYSBCOUNT)
507 __XFS_ADD_V4_FEAT(projid32, PROJID32)
508 __XFS_HAS_V4_FEAT(v3inodes, V3INODES)
509 __XFS_HAS_V4_FEAT(crc, CRC)
510 __XFS_HAS_V4_FEAT(pquotino, PQUOTINO)
511
512 static inline void xfs_add_attr2(struct xfs_mount *mp)
513 {
514 if (IS_ENABLED(CONFIG_XFS_SUPPORT_V4))
515 xfs_sb_version_addattr2(&mp->m_sb);
516 }
517
518 /*
519 * Mount features
520 *
521 * These do not change dynamically - features that can come and go, such as 32
522 * bit inodes and read-only state, are kept as operational state rather than
523 * features.
524 */
__XFS_HAS_FEAT(noalign,NOALIGN)525 __XFS_HAS_FEAT(noalign, NOALIGN)
526 __XFS_HAS_FEAT(allocsize, ALLOCSIZE)
527 __XFS_HAS_FEAT(large_iosize, LARGE_IOSIZE)
528 __XFS_HAS_FEAT(wsync, WSYNC)
529 __XFS_HAS_FEAT(dirsync, DIRSYNC)
530 __XFS_HAS_FEAT(discard, DISCARD)
531 __XFS_HAS_FEAT(grpid, GRPID)
532 __XFS_HAS_FEAT(small_inums, SMALL_INUMS)
533 __XFS_HAS_FEAT(swalloc, SWALLOC)
534 __XFS_HAS_FEAT(filestreams, FILESTREAMS)
535 __XFS_HAS_FEAT(dax_always, DAX_ALWAYS)
536 __XFS_HAS_FEAT(dax_never, DAX_NEVER)
537 __XFS_HAS_FEAT(norecovery, NORECOVERY)
538 __XFS_HAS_FEAT(nouuid, NOUUID)
539
540 /*
541 * Operational mount state flags
542 *
543 * Use these with atomic bit ops only!
544 */
545 #define XFS_OPSTATE_UNMOUNTING 0 /* filesystem is unmounting */
546 #define XFS_OPSTATE_CLEAN 1 /* mount was clean */
547 #define XFS_OPSTATE_SHUTDOWN 2 /* stop all fs operations */
548 #define XFS_OPSTATE_INODE32 3 /* inode32 allocator active */
549 #define XFS_OPSTATE_READONLY 4 /* read-only fs */
550
551 /*
552 * If set, inactivation worker threads will be scheduled to process queued
553 * inodegc work. If not, queued inodes remain in memory waiting to be
554 * processed.
555 */
556 #define XFS_OPSTATE_INODEGC_ENABLED 5
557 /*
558 * If set, background speculative prealloc gc worker threads will be scheduled
559 * to process queued blockgc work. If not, inodes retain their preallocations
560 * until explicitly deleted.
561 */
562 #define XFS_OPSTATE_BLOCKGC_ENABLED 6
563
564 /* Kernel has logged a warning about shrink being used on this fs. */
565 #define XFS_OPSTATE_WARNED_SHRINK 9
566 /* Kernel has logged a warning about logged xattr updates being used. */
567 #define XFS_OPSTATE_WARNED_LARP 10
568 /* Mount time quotacheck is running */
569 #define XFS_OPSTATE_QUOTACHECK_RUNNING 11
570 /* Do we want to clear log incompat flags? */
571 #define XFS_OPSTATE_UNSET_LOG_INCOMPAT 12
572 /* Filesystem can use logged extended attributes */
573 #define XFS_OPSTATE_USE_LARP 13
574 /* Kernel has logged a warning about blocksize > pagesize on this fs. */
575 #define XFS_OPSTATE_WARNED_LBS 14
576 /* Kernel has logged a warning about metadata dirs being used on this fs. */
577 #define XFS_OPSTATE_WARNED_METADIR 17
578 /* Filesystem should use qflags to determine quotaon status */
579 #define XFS_OPSTATE_RESUMING_QUOTAON 18
580 /* Kernel has logged a warning about zoned RT device being used on this fs. */
581 #define XFS_OPSTATE_WARNED_ZONED 19
582 /* (Zoned) GC is in progress */
583 #define XFS_OPSTATE_ZONEGC_RUNNING 20
584
585 #define __XFS_IS_OPSTATE(name, NAME) \
586 static inline bool xfs_is_ ## name (struct xfs_mount *mp) \
587 { \
588 return test_bit(XFS_OPSTATE_ ## NAME, &mp->m_opstate); \
589 } \
590 static inline bool xfs_clear_ ## name (struct xfs_mount *mp) \
591 { \
592 return test_and_clear_bit(XFS_OPSTATE_ ## NAME, &mp->m_opstate); \
593 } \
594 static inline bool xfs_set_ ## name (struct xfs_mount *mp) \
595 { \
596 return test_and_set_bit(XFS_OPSTATE_ ## NAME, &mp->m_opstate); \
597 }
598
599 __XFS_IS_OPSTATE(unmounting, UNMOUNTING)
600 __XFS_IS_OPSTATE(clean, CLEAN)
601 __XFS_IS_OPSTATE(shutdown, SHUTDOWN)
602 __XFS_IS_OPSTATE(inode32, INODE32)
603 __XFS_IS_OPSTATE(readonly, READONLY)
604 __XFS_IS_OPSTATE(inodegc_enabled, INODEGC_ENABLED)
605 __XFS_IS_OPSTATE(blockgc_enabled, BLOCKGC_ENABLED)
606 #ifdef CONFIG_XFS_QUOTA
607 __XFS_IS_OPSTATE(quotacheck_running, QUOTACHECK_RUNNING)
608 __XFS_IS_OPSTATE(resuming_quotaon, RESUMING_QUOTAON)
609 #else
610 static inline bool xfs_is_quotacheck_running(struct xfs_mount *mp)
611 {
612 return false;
613 }
614 static inline bool xfs_is_resuming_quotaon(struct xfs_mount *mp)
615 {
616 return false;
617 }
618 static inline void xfs_set_resuming_quotaon(struct xfs_mount *m)
619 {
620 }
621 static inline bool xfs_clear_resuming_quotaon(struct xfs_mount *mp)
622 {
623 return false;
624 }
625 #endif /* CONFIG_XFS_QUOTA */
626 __XFS_IS_OPSTATE(done_with_log_incompat, UNSET_LOG_INCOMPAT)
627 __XFS_IS_OPSTATE(using_logged_xattrs, USE_LARP)
628 __XFS_IS_OPSTATE(zonegc_running, ZONEGC_RUNNING)
629
630 static inline bool
631 xfs_should_warn(struct xfs_mount *mp, long nr)
632 {
633 return !test_and_set_bit(nr, &mp->m_opstate);
634 }
635
636 #define XFS_OPSTATE_STRINGS \
637 { (1UL << XFS_OPSTATE_UNMOUNTING), "unmounting" }, \
638 { (1UL << XFS_OPSTATE_CLEAN), "clean" }, \
639 { (1UL << XFS_OPSTATE_SHUTDOWN), "shutdown" }, \
640 { (1UL << XFS_OPSTATE_INODE32), "inode32" }, \
641 { (1UL << XFS_OPSTATE_READONLY), "read_only" }, \
642 { (1UL << XFS_OPSTATE_INODEGC_ENABLED), "inodegc" }, \
643 { (1UL << XFS_OPSTATE_BLOCKGC_ENABLED), "blockgc" }, \
644 { (1UL << XFS_OPSTATE_WARNED_SHRINK), "wshrink" }, \
645 { (1UL << XFS_OPSTATE_WARNED_LARP), "wlarp" }, \
646 { (1UL << XFS_OPSTATE_QUOTACHECK_RUNNING), "quotacheck" }, \
647 { (1UL << XFS_OPSTATE_UNSET_LOG_INCOMPAT), "unset_log_incompat" }, \
648 { (1UL << XFS_OPSTATE_USE_LARP), "logged_xattrs" }
649
650 /*
651 * Max and min values for mount-option defined I/O
652 * preallocation sizes.
653 */
654 #define XFS_MAX_IO_LOG 30 /* 1G */
655 #define XFS_MIN_IO_LOG PAGE_SHIFT
656
657 void xfs_do_force_shutdown(struct xfs_mount *mp, uint32_t flags, char *fname,
658 int lnnum);
659 #define xfs_force_shutdown(m,f) \
660 xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
661
662 #define SHUTDOWN_META_IO_ERROR (1u << 0) /* write attempt to metadata failed */
663 #define SHUTDOWN_LOG_IO_ERROR (1u << 1) /* write attempt to the log failed */
664 #define SHUTDOWN_FORCE_UMOUNT (1u << 2) /* shutdown from a forced unmount */
665 #define SHUTDOWN_CORRUPT_INCORE (1u << 3) /* corrupt in-memory structures */
666 #define SHUTDOWN_CORRUPT_ONDISK (1u << 4) /* corrupt metadata on device */
667 #define SHUTDOWN_DEVICE_REMOVED (1u << 5) /* device removed underneath us */
668
669 #define XFS_SHUTDOWN_STRINGS \
670 { SHUTDOWN_META_IO_ERROR, "metadata_io" }, \
671 { SHUTDOWN_LOG_IO_ERROR, "log_io" }, \
672 { SHUTDOWN_FORCE_UMOUNT, "force_umount" }, \
673 { SHUTDOWN_CORRUPT_INCORE, "corruption" }, \
674 { SHUTDOWN_DEVICE_REMOVED, "device_removed" }
675
676 /*
677 * Flags for xfs_mountfs
678 */
679 #define XFS_MFSI_QUIET 0x40 /* Be silent if mount errors found */
680
681 static inline xfs_agnumber_t
xfs_daddr_to_agno(struct xfs_mount * mp,xfs_daddr_t d)682 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
683 {
684 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
685 do_div(ld, mp->m_sb.sb_agblocks);
686 return (xfs_agnumber_t) ld;
687 }
688
689 static inline xfs_agblock_t
xfs_daddr_to_agbno(struct xfs_mount * mp,xfs_daddr_t d)690 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
691 {
692 xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
693 return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
694 }
695
696 extern void xfs_uuid_table_free(void);
697 uint64_t xfs_default_resblks(struct xfs_mount *mp,
698 enum xfs_free_counter ctr);
699 extern int xfs_mountfs(xfs_mount_t *mp);
700 extern void xfs_unmountfs(xfs_mount_t *);
701
702 /*
703 * Deltas for the block count can vary from 1 to very large, but lock contention
704 * only occurs on frequent small block count updates such as in the delayed
705 * allocation path for buffered writes (page a time updates). Hence we set
706 * a large batch count (1024) to minimise global counter updates except when
707 * we get near to ENOSPC and we have to be very accurate with our updates.
708 */
709 #define XFS_FDBLOCKS_BATCH 1024
710
711 uint64_t xfs_freecounter_unavailable(struct xfs_mount *mp,
712 enum xfs_free_counter ctr);
713
714 /*
715 * Sum up the freecount, but never return negative values.
716 */
xfs_sum_freecounter(struct xfs_mount * mp,enum xfs_free_counter ctr)717 static inline s64 xfs_sum_freecounter(struct xfs_mount *mp,
718 enum xfs_free_counter ctr)
719 {
720 return percpu_counter_sum_positive(&mp->m_free[ctr].count);
721 }
722
723 /*
724 * Same as above, but does return negative values. Mostly useful for
725 * special cases like repair and tracing.
726 */
xfs_sum_freecounter_raw(struct xfs_mount * mp,enum xfs_free_counter ctr)727 static inline s64 xfs_sum_freecounter_raw(struct xfs_mount *mp,
728 enum xfs_free_counter ctr)
729 {
730 return percpu_counter_sum(&mp->m_free[ctr].count);
731 }
732
733 /*
734 * This just provides and estimate without the cpu-local updates, use
735 * xfs_sum_freecounter for the exact value.
736 */
xfs_estimate_freecounter(struct xfs_mount * mp,enum xfs_free_counter ctr)737 static inline s64 xfs_estimate_freecounter(struct xfs_mount *mp,
738 enum xfs_free_counter ctr)
739 {
740 return percpu_counter_read_positive(&mp->m_free[ctr].count);
741 }
742
xfs_compare_freecounter(struct xfs_mount * mp,enum xfs_free_counter ctr,s64 rhs,s32 batch)743 static inline int xfs_compare_freecounter(struct xfs_mount *mp,
744 enum xfs_free_counter ctr, s64 rhs, s32 batch)
745 {
746 return __percpu_counter_compare(&mp->m_free[ctr].count, rhs, batch);
747 }
748
xfs_set_freecounter(struct xfs_mount * mp,enum xfs_free_counter ctr,uint64_t val)749 static inline void xfs_set_freecounter(struct xfs_mount *mp,
750 enum xfs_free_counter ctr, uint64_t val)
751 {
752 percpu_counter_set(&mp->m_free[ctr].count, val);
753 }
754
755 int xfs_dec_freecounter(struct xfs_mount *mp, enum xfs_free_counter ctr,
756 uint64_t delta, bool rsvd);
757 void xfs_add_freecounter(struct xfs_mount *mp, enum xfs_free_counter ctr,
758 uint64_t delta);
759
xfs_dec_fdblocks(struct xfs_mount * mp,uint64_t delta,bool reserved)760 static inline int xfs_dec_fdblocks(struct xfs_mount *mp, uint64_t delta,
761 bool reserved)
762 {
763 return xfs_dec_freecounter(mp, XC_FREE_BLOCKS, delta, reserved);
764 }
765
xfs_add_fdblocks(struct xfs_mount * mp,uint64_t delta)766 static inline void xfs_add_fdblocks(struct xfs_mount *mp, uint64_t delta)
767 {
768 xfs_add_freecounter(mp, XC_FREE_BLOCKS, delta);
769 }
770
xfs_dec_frextents(struct xfs_mount * mp,uint64_t delta)771 static inline int xfs_dec_frextents(struct xfs_mount *mp, uint64_t delta)
772 {
773 return xfs_dec_freecounter(mp, XC_FREE_RTEXTENTS, delta, false);
774 }
775
xfs_add_frextents(struct xfs_mount * mp,uint64_t delta)776 static inline void xfs_add_frextents(struct xfs_mount *mp, uint64_t delta)
777 {
778 xfs_add_freecounter(mp, XC_FREE_RTEXTENTS, delta);
779 }
780
781 extern int xfs_readsb(xfs_mount_t *, int);
782 extern void xfs_freesb(xfs_mount_t *);
783 extern bool xfs_fs_writable(struct xfs_mount *mp, int level);
784 extern int xfs_sb_validate_fsb_count(struct xfs_sb *, uint64_t);
785
786 extern int xfs_dev_is_read_only(struct xfs_mount *, char *);
787
788 extern void xfs_set_low_space_thresholds(struct xfs_mount *);
789
790 int xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb,
791 xfs_off_t count_fsb);
792
793 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp,
794 int error_class, int error);
795 void xfs_force_summary_recalc(struct xfs_mount *mp);
796 int xfs_add_incompat_log_feature(struct xfs_mount *mp, uint32_t feature);
797 bool xfs_clear_incompat_log_features(struct xfs_mount *mp);
798 void xfs_mod_delalloc(struct xfs_inode *ip, int64_t data_delta,
799 int64_t ind_delta);
xfs_mod_sb_delalloc(struct xfs_mount * mp,int64_t delta)800 static inline void xfs_mod_sb_delalloc(struct xfs_mount *mp, int64_t delta)
801 {
802 percpu_counter_add(&mp->m_delalloc_blks, delta);
803 }
804
805 int xfs_set_max_atomic_write_opt(struct xfs_mount *mp,
806 unsigned long long new_max_bytes);
807
808 static inline struct xfs_buftarg *
xfs_group_type_buftarg(struct xfs_mount * mp,enum xfs_group_type type)809 xfs_group_type_buftarg(
810 struct xfs_mount *mp,
811 enum xfs_group_type type)
812 {
813 switch (type) {
814 case XG_TYPE_AG:
815 return mp->m_ddev_targp;
816 case XG_TYPE_RTG:
817 return mp->m_rtdev_targp;
818 default:
819 ASSERT(0);
820 break;
821 }
822 return NULL;
823 }
824
825 #endif /* __XFS_MOUNT_H__ */
826