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