xref: /openbmc/linux/fs/xfs/xfs_mount.h (revision 048dc3ab)
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_nameops;
13 struct xfs_ail;
14 struct xfs_quotainfo;
15 struct xfs_dir_ops;
16 struct xfs_da_geometry;
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 typedef struct xfs_mount {
61 	struct super_block	*m_super;
62 	xfs_tid_t		m_tid;		/* next unused tid for fs */
63 
64 	/*
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.
67 	 */
68 	uint8_t			m_fs_checked;
69 	uint8_t			m_fs_sick;
70 	/*
71 	 * Bitsets of rt metadata that have been checked and/or are sick.
72 	 * Callers must hold m_sb_lock to access this field.
73 	 */
74 	uint8_t			m_rt_checked;
75 	uint8_t			m_rt_sick;
76 
77 	struct xfs_ail		*m_ail;		/* fs active log item list */
78 
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 */
84 	/*
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.
88 	 */
89 	struct percpu_counter	m_delalloc_blks;
90 
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 	struct xfs_ino_geometry	m_ino_geo;	/* inode geometry */
109 	int			m_logbufs;	/* number of log buffers */
110 	int			m_logbsize;	/* size of each log buffer */
111 	uint			m_rsumlevels;	/* rt summary levels */
112 	uint			m_rsumsize;	/* size of rt summary, bytes */
113 	/*
114 	 * Optional cache of rt summary level per bitmap block with the
115 	 * invariant that m_rsum_cache[bbno] <= the minimum i for which
116 	 * rsum[i][bbno] != 0. Reads and writes are serialized by the rsumip
117 	 * inode lock.
118 	 */
119 	uint8_t			*m_rsum_cache;
120 	struct xfs_inode	*m_rbmip;	/* pointer to bitmap inode */
121 	struct xfs_inode	*m_rsumip;	/* pointer to summary inode */
122 	struct xfs_inode	*m_rootip;	/* pointer to root directory */
123 	struct xfs_quotainfo	*m_quotainfo;	/* disk quota information */
124 	xfs_buftarg_t		*m_ddev_targp;	/* saves taking the address */
125 	xfs_buftarg_t		*m_logdev_targp;/* ptr to log device */
126 	xfs_buftarg_t		*m_rtdev_targp;	/* ptr to rt device */
127 	uint8_t			m_blkbit_log;	/* blocklog + NBBY */
128 	uint8_t			m_blkbb_log;	/* blocklog - BBSHIFT */
129 	uint8_t			m_agno_log;	/* log #ag's */
130 	uint			m_blockmask;	/* sb_blocksize-1 */
131 	uint			m_blockwsize;	/* sb_blocksize in words */
132 	uint			m_blockwmask;	/* blockwsize-1 */
133 	uint			m_alloc_mxr[2];	/* max alloc btree records */
134 	uint			m_alloc_mnr[2];	/* min alloc btree records */
135 	uint			m_bmap_dmxr[2];	/* max bmap btree records */
136 	uint			m_bmap_dmnr[2];	/* min bmap btree records */
137 	uint			m_rmap_mxr[2];	/* max rmap btree records */
138 	uint			m_rmap_mnr[2];	/* min rmap btree records */
139 	uint			m_refc_mxr[2];	/* max refc btree records */
140 	uint			m_refc_mnr[2];	/* min refc btree records */
141 	uint			m_ag_maxlevels;	/* XFS_AG_MAXLEVELS */
142 	uint			m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
143 	uint			m_rmap_maxlevels; /* max rmap btree levels */
144 	uint			m_refc_maxlevels; /* max refcount btree level */
145 	xfs_extlen_t		m_ag_prealloc_blocks; /* reserved ag blocks */
146 	uint			m_alloc_set_aside; /* space we can't use */
147 	uint			m_ag_max_usable; /* max space per AG */
148 	struct radix_tree_root	m_perag_tree;	/* per-ag accounting info */
149 	spinlock_t		m_perag_lock;	/* lock for m_perag_tree */
150 	struct mutex		m_growlock;	/* growfs mutex */
151 	int			m_fixedfsid[2];	/* unchanged for life of FS */
152 	uint64_t		m_flags;	/* global mount flags */
153 	bool			m_finobt_nores; /* no per-AG finobt resv. */
154 	uint			m_qflags;	/* quota status flags */
155 	struct xfs_trans_resv	m_resv;		/* precomputed res values */
156 	uint64_t		m_resblks;	/* total reserved blocks */
157 	uint64_t		m_resblks_avail;/* available reserved blocks */
158 	uint64_t		m_resblks_save;	/* reserved blks @ remount,ro */
159 	int			m_dalign;	/* stripe unit */
160 	int			m_swidth;	/* stripe width */
161 	uint8_t			m_sectbb_log;	/* sectlog - BBSHIFT */
162 	const struct xfs_nameops *m_dirnameops;	/* vector of dir name ops */
163 	const struct xfs_dir_ops *m_dir_inode_ops; /* vector of dir inode ops */
164 	const struct xfs_dir_ops *m_nondir_inode_ops; /* !dir inode ops */
165 	uint			m_chsize;	/* size of next field */
166 	atomic_t		m_active_trans;	/* number trans frozen */
167 	struct xfs_mru_cache	*m_filestream;  /* per-mount filestream data */
168 	struct delayed_work	m_reclaim_work;	/* background inode reclaim */
169 	struct delayed_work	m_eofblocks_work; /* background eof blocks
170 						     trimming */
171 	struct delayed_work	m_cowblocks_work; /* background cow blocks
172 						     trimming */
173 	bool			m_update_sb;	/* sb needs update in mount */
174 	int64_t			m_low_space[XFS_LOWSP_MAX];
175 						/* low free space thresholds */
176 	struct xfs_kobj		m_kobj;
177 	struct xfs_kobj		m_error_kobj;
178 	struct xfs_kobj		m_error_meta_kobj;
179 	struct xfs_error_cfg	m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX];
180 	struct xstats		m_stats;	/* per-fs stats */
181 
182 	struct workqueue_struct *m_buf_workqueue;
183 	struct workqueue_struct	*m_unwritten_workqueue;
184 	struct workqueue_struct	*m_cil_workqueue;
185 	struct workqueue_struct	*m_reclaim_workqueue;
186 	struct workqueue_struct *m_eofblocks_workqueue;
187 	struct workqueue_struct	*m_sync_workqueue;
188 
189 	/*
190 	 * Generation of the filesysyem layout.  This is incremented by each
191 	 * growfs, and used by the pNFS server to ensure the client updates
192 	 * its view of the block device once it gets a layout that might
193 	 * reference the newly added blocks.  Does not need to be persistent
194 	 * as long as we only allow file system size increments, but if we
195 	 * ever support shrinks it would have to be persisted in addition
196 	 * to various other kinds of pain inflicted on the pNFS server.
197 	 */
198 	uint32_t		m_generation;
199 
200 	bool			m_always_cow;
201 	bool			m_fail_unmount;
202 #ifdef DEBUG
203 	/*
204 	 * Frequency with which errors are injected.  Replaces xfs_etest; the
205 	 * value stored in here is the inverse of the frequency with which the
206 	 * error triggers.  1 = always, 2 = half the time, etc.
207 	 */
208 	unsigned int		*m_errortag;
209 	struct xfs_kobj		m_errortag_kobj;
210 #endif
211 } xfs_mount_t;
212 
213 #define M_IGEO(mp)		(&(mp)->m_ino_geo)
214 
215 /*
216  * Flags for m_flags.
217  */
218 #define XFS_MOUNT_WSYNC		(1ULL << 0)	/* for nfs - all metadata ops
219 						   must be synchronous except
220 						   for space allocations */
221 #define XFS_MOUNT_UNMOUNTING	(1ULL << 1)	/* filesystem is unmounting */
222 #define XFS_MOUNT_WAS_CLEAN	(1ULL << 3)
223 #define XFS_MOUNT_FS_SHUTDOWN	(1ULL << 4)	/* atomic stop of all filesystem
224 						   operations, typically for
225 						   disk errors in metadata */
226 #define XFS_MOUNT_DISCARD	(1ULL << 5)	/* discard unused blocks */
227 #define XFS_MOUNT_NOALIGN	(1ULL << 7)	/* turn off stripe alignment
228 						   allocations */
229 #define XFS_MOUNT_ATTR2		(1ULL << 8)	/* allow use of attr2 format */
230 #define XFS_MOUNT_GRPID		(1ULL << 9)	/* group-ID assigned from directory */
231 #define XFS_MOUNT_NORECOVERY	(1ULL << 10)	/* no recovery - dirty fs */
232 #define XFS_MOUNT_DFLT_IOSIZE	(1ULL << 12)	/* set default i/o size */
233 #define XFS_MOUNT_SMALL_INUMS	(1ULL << 14)	/* user wants 32bit inodes */
234 #define XFS_MOUNT_32BITINODES	(1ULL << 15)	/* inode32 allocator active */
235 #define XFS_MOUNT_NOUUID	(1ULL << 16)	/* ignore uuid during mount */
236 #define XFS_MOUNT_IKEEP		(1ULL << 18)	/* keep empty inode clusters*/
237 #define XFS_MOUNT_SWALLOC	(1ULL << 19)	/* turn on stripe width
238 						 * allocation */
239 #define XFS_MOUNT_RDONLY	(1ULL << 20)	/* read-only fs */
240 #define XFS_MOUNT_DIRSYNC	(1ULL << 21)	/* synchronous directory ops */
241 #define XFS_MOUNT_COMPAT_IOSIZE	(1ULL << 22)	/* don't report large preferred
242 						 * I/O size in stat() */
243 #define XFS_MOUNT_FILESTREAMS	(1ULL << 24)	/* enable the filestreams
244 						   allocator */
245 #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
246 
247 #define XFS_MOUNT_DAX		(1ULL << 62)	/* TEST ONLY! */
248 
249 
250 /*
251  * Default minimum read and write sizes.
252  */
253 #define XFS_READIO_LOG_LARGE	16
254 #define XFS_WRITEIO_LOG_LARGE	16
255 
256 /*
257  * Max and min values for mount-option defined I/O
258  * preallocation sizes.
259  */
260 #define XFS_MAX_IO_LOG		30	/* 1G */
261 #define XFS_MIN_IO_LOG		PAGE_SHIFT
262 
263 /*
264  * Synchronous read and write sizes.  This should be
265  * better for NFSv2 wsync filesystems.
266  */
267 #define	XFS_WSYNC_READIO_LOG	15	/* 32k */
268 #define	XFS_WSYNC_WRITEIO_LOG	14	/* 16k */
269 
270 /*
271  * Allow large block sizes to be reported to userspace programs if the
272  * "largeio" mount option is used.
273  *
274  * If compatibility mode is specified, simply return the basic unit of caching
275  * so that we don't get inefficient read/modify/write I/O from user apps.
276  * Otherwise....
277  *
278  * If the underlying volume is a stripe, then return the stripe width in bytes
279  * as the recommended I/O size. It is not a stripe and we've set a default
280  * buffered I/O size, return that, otherwise return the compat default.
281  */
282 static inline unsigned long
283 xfs_preferred_iosize(xfs_mount_t *mp)
284 {
285 	if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
286 		return PAGE_SIZE;
287 	return (mp->m_swidth ?
288 		(mp->m_swidth << mp->m_sb.sb_blocklog) :
289 		((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
290 			(1 << (int)max(mp->m_readio_log, mp->m_writeio_log)) :
291 			PAGE_SIZE));
292 }
293 
294 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp)	\
295 				((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
296 #define XFS_FORCED_SHUTDOWN(mp)	((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN)
297 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
298 		int lnnum);
299 #define xfs_force_shutdown(m,f)	\
300 	xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
301 
302 #define SHUTDOWN_META_IO_ERROR	0x0001	/* write attempt to metadata failed */
303 #define SHUTDOWN_LOG_IO_ERROR	0x0002	/* write attempt to the log failed */
304 #define SHUTDOWN_FORCE_UMOUNT	0x0004	/* shutdown from a forced unmount */
305 #define SHUTDOWN_CORRUPT_INCORE	0x0008	/* corrupt in-memory data structures */
306 #define SHUTDOWN_REMOTE_REQ	0x0010	/* shutdown came from remote cell */
307 #define SHUTDOWN_DEVICE_REQ	0x0020	/* failed all paths to the device */
308 
309 /*
310  * Flags for xfs_mountfs
311  */
312 #define XFS_MFSI_QUIET		0x40	/* Be silent if mount errors found */
313 
314 static inline xfs_agnumber_t
315 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
316 {
317 	xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
318 	do_div(ld, mp->m_sb.sb_agblocks);
319 	return (xfs_agnumber_t) ld;
320 }
321 
322 static inline xfs_agblock_t
323 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
324 {
325 	xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
326 	return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
327 }
328 
329 /* per-AG block reservation data structures*/
330 enum xfs_ag_resv_type {
331 	XFS_AG_RESV_NONE = 0,
332 	XFS_AG_RESV_AGFL,
333 	XFS_AG_RESV_METADATA,
334 	XFS_AG_RESV_RMAPBT,
335 };
336 
337 struct xfs_ag_resv {
338 	/* number of blocks originally reserved here */
339 	xfs_extlen_t			ar_orig_reserved;
340 	/* number of blocks reserved here */
341 	xfs_extlen_t			ar_reserved;
342 	/* number of blocks originally asked for */
343 	xfs_extlen_t			ar_asked;
344 };
345 
346 /*
347  * Per-ag incore structure, copies of information in agf and agi, to improve the
348  * performance of allocation group selection.
349  */
350 typedef struct xfs_perag {
351 	struct xfs_mount *pag_mount;	/* owner filesystem */
352 	xfs_agnumber_t	pag_agno;	/* AG this structure belongs to */
353 	atomic_t	pag_ref;	/* perag reference count */
354 	char		pagf_init;	/* this agf's entry is initialized */
355 	char		pagi_init;	/* this agi's entry is initialized */
356 	char		pagf_metadata;	/* the agf is preferred to be metadata */
357 	char		pagi_inodeok;	/* The agi is ok for inodes */
358 	uint8_t		pagf_levels[XFS_BTNUM_AGF];
359 					/* # of levels in bno & cnt btree */
360 	bool		pagf_agflreset; /* agfl requires reset before use */
361 	uint32_t	pagf_flcount;	/* count of blocks in freelist */
362 	xfs_extlen_t	pagf_freeblks;	/* total free blocks */
363 	xfs_extlen_t	pagf_longest;	/* longest free space */
364 	uint32_t	pagf_btreeblks;	/* # of blocks held in AGF btrees */
365 	xfs_agino_t	pagi_freecount;	/* number of free inodes */
366 	xfs_agino_t	pagi_count;	/* number of allocated inodes */
367 
368 	/*
369 	 * Inode allocation search lookup optimisation.
370 	 * If the pagino matches, the search for new inodes
371 	 * doesn't need to search the near ones again straight away
372 	 */
373 	xfs_agino_t	pagl_pagino;
374 	xfs_agino_t	pagl_leftrec;
375 	xfs_agino_t	pagl_rightrec;
376 
377 	/*
378 	 * Bitsets of per-ag metadata that have been checked and/or are sick.
379 	 * Callers should hold pag_state_lock before accessing this field.
380 	 */
381 	uint16_t	pag_checked;
382 	uint16_t	pag_sick;
383 	spinlock_t	pag_state_lock;
384 
385 	spinlock_t	pagb_lock;	/* lock for pagb_tree */
386 	struct rb_root	pagb_tree;	/* ordered tree of busy extents */
387 	unsigned int	pagb_gen;	/* generation count for pagb_tree */
388 	wait_queue_head_t pagb_wait;	/* woken when pagb_gen changes */
389 
390 	atomic_t        pagf_fstrms;    /* # of filestreams active in this AG */
391 
392 	spinlock_t	pag_ici_lock;	/* incore inode cache lock */
393 	struct radix_tree_root pag_ici_root;	/* incore inode cache root */
394 	int		pag_ici_reclaimable;	/* reclaimable inodes */
395 	struct mutex	pag_ici_reclaim_lock;	/* serialisation point */
396 	unsigned long	pag_ici_reclaim_cursor;	/* reclaim restart point */
397 
398 	/* buffer cache index */
399 	spinlock_t	pag_buf_lock;	/* lock for pag_buf_hash */
400 	struct rhashtable pag_buf_hash;
401 
402 	/* for rcu-safe freeing */
403 	struct rcu_head	rcu_head;
404 	int		pagb_count;	/* pagb slots in use */
405 
406 	/* Blocks reserved for all kinds of metadata. */
407 	struct xfs_ag_resv	pag_meta_resv;
408 	/* Blocks reserved for the reverse mapping btree. */
409 	struct xfs_ag_resv	pag_rmapbt_resv;
410 
411 	/* reference count */
412 	uint8_t			pagf_refcount_level;
413 
414 	/*
415 	 * Unlinked inode information.  This incore information reflects
416 	 * data stored in the AGI, so callers must hold the AGI buffer lock
417 	 * or have some other means to control concurrency.
418 	 */
419 	struct rhashtable	pagi_unlinked_hash;
420 } xfs_perag_t;
421 
422 static inline struct xfs_ag_resv *
423 xfs_perag_resv(
424 	struct xfs_perag	*pag,
425 	enum xfs_ag_resv_type	type)
426 {
427 	switch (type) {
428 	case XFS_AG_RESV_METADATA:
429 		return &pag->pag_meta_resv;
430 	case XFS_AG_RESV_RMAPBT:
431 		return &pag->pag_rmapbt_resv;
432 	default:
433 		return NULL;
434 	}
435 }
436 
437 int xfs_buf_hash_init(xfs_perag_t *pag);
438 void xfs_buf_hash_destroy(xfs_perag_t *pag);
439 
440 extern void	xfs_uuid_table_free(void);
441 extern int	xfs_log_sbcount(xfs_mount_t *);
442 extern uint64_t xfs_default_resblks(xfs_mount_t *mp);
443 extern int	xfs_mountfs(xfs_mount_t *mp);
444 extern int	xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount,
445 				     xfs_agnumber_t *maxagi);
446 extern void	xfs_unmountfs(xfs_mount_t *);
447 
448 extern int	xfs_mod_icount(struct xfs_mount *mp, int64_t delta);
449 extern int	xfs_mod_ifree(struct xfs_mount *mp, int64_t delta);
450 extern int	xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
451 				 bool reserved);
452 extern int	xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
453 
454 extern struct xfs_buf *xfs_getsb(xfs_mount_t *);
455 extern int	xfs_readsb(xfs_mount_t *, int);
456 extern void	xfs_freesb(xfs_mount_t *);
457 extern bool	xfs_fs_writable(struct xfs_mount *mp, int level);
458 extern int	xfs_sb_validate_fsb_count(struct xfs_sb *, uint64_t);
459 
460 extern int	xfs_dev_is_read_only(struct xfs_mount *, char *);
461 
462 extern void	xfs_set_low_space_thresholds(struct xfs_mount *);
463 
464 int	xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb,
465 			xfs_off_t count_fsb);
466 
467 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp,
468 		int error_class, int error);
469 void xfs_force_summary_recalc(struct xfs_mount *mp);
470 void xfs_mod_delalloc(struct xfs_mount *mp, int64_t delta);
471 
472 #endif	/* __XFS_MOUNT_H__ */
473