xref: /openbmc/linux/fs/xfs/xfs_mount.h (revision 8dda2eac)
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 
17 /* dynamic preallocation free space thresholds, 5% down to 1% */
18 enum {
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  */
33 enum {
34 	XFS_ERR_METADATA,
35 	XFS_ERR_CLASS_MAX,
36 };
37 enum {
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  */
53 struct xfs_error_cfg {
54 	struct xfs_kobj	kobj;
55 	int		max_retries;
56 	long		retry_timeout;	/* in jiffies, -1 = infinite */
57 };
58 
59 /*
60  * The struct xfsmount layout is optimised to separate read-mostly variables
61  * from variables that are frequently modified. We put the read-mostly variables
62  * first, then place all the other variables at the end.
63  *
64  * Typically, read-mostly variables are those that are set at mount time and
65  * never changed again, or only change rarely as a result of things like sysfs
66  * knobs being tweaked.
67  */
68 typedef struct xfs_mount {
69 	struct xfs_sb		m_sb;		/* copy of fs superblock */
70 	struct super_block	*m_super;
71 	struct xfs_ail		*m_ail;		/* fs active log item list */
72 	struct xfs_buf		*m_sb_bp;	/* buffer for superblock */
73 	char			*m_rtname;	/* realtime device name */
74 	char			*m_logname;	/* external log device name */
75 	struct xfs_da_geometry	*m_dir_geo;	/* directory block geometry */
76 	struct xfs_da_geometry	*m_attr_geo;	/* attribute block geometry */
77 	struct xlog		*m_log;		/* log specific stuff */
78 	struct xfs_inode	*m_rbmip;	/* pointer to bitmap inode */
79 	struct xfs_inode	*m_rsumip;	/* pointer to summary inode */
80 	struct xfs_inode	*m_rootip;	/* pointer to root directory */
81 	struct xfs_quotainfo	*m_quotainfo;	/* disk quota information */
82 	xfs_buftarg_t		*m_ddev_targp;	/* saves taking the address */
83 	xfs_buftarg_t		*m_logdev_targp;/* ptr to log device */
84 	xfs_buftarg_t		*m_rtdev_targp;	/* ptr to rt device */
85 	/*
86 	 * Optional cache of rt summary level per bitmap block with the
87 	 * invariant that m_rsum_cache[bbno] <= the minimum i for which
88 	 * rsum[i][bbno] != 0. Reads and writes are serialized by the rsumip
89 	 * inode lock.
90 	 */
91 	uint8_t			*m_rsum_cache;
92 	struct xfs_mru_cache	*m_filestream;  /* per-mount filestream data */
93 	struct workqueue_struct *m_buf_workqueue;
94 	struct workqueue_struct	*m_unwritten_workqueue;
95 	struct workqueue_struct	*m_cil_workqueue;
96 	struct workqueue_struct	*m_reclaim_workqueue;
97 	struct workqueue_struct *m_gc_workqueue;
98 	struct workqueue_struct	*m_sync_workqueue;
99 
100 	int			m_bsize;	/* fs logical block size */
101 	uint8_t			m_blkbit_log;	/* blocklog + NBBY */
102 	uint8_t			m_blkbb_log;	/* blocklog - BBSHIFT */
103 	uint8_t			m_agno_log;	/* log #ag's */
104 	uint8_t			m_sectbb_log;	/* sectlog - BBSHIFT */
105 	uint			m_blockmask;	/* sb_blocksize-1 */
106 	uint			m_blockwsize;	/* sb_blocksize in words */
107 	uint			m_blockwmask;	/* blockwsize-1 */
108 	uint			m_alloc_mxr[2];	/* max alloc btree records */
109 	uint			m_alloc_mnr[2];	/* min alloc btree records */
110 	uint			m_bmap_dmxr[2];	/* max bmap btree records */
111 	uint			m_bmap_dmnr[2];	/* min bmap btree records */
112 	uint			m_rmap_mxr[2];	/* max rmap btree records */
113 	uint			m_rmap_mnr[2];	/* min rmap btree records */
114 	uint			m_refc_mxr[2];	/* max refc btree records */
115 	uint			m_refc_mnr[2];	/* min refc btree records */
116 	uint			m_ag_maxlevels;	/* XFS_AG_MAXLEVELS */
117 	uint			m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
118 	uint			m_rmap_maxlevels; /* max rmap btree levels */
119 	uint			m_refc_maxlevels; /* max refcount btree level */
120 	xfs_extlen_t		m_ag_prealloc_blocks; /* reserved ag blocks */
121 	uint			m_alloc_set_aside; /* space we can't use */
122 	uint			m_ag_max_usable; /* max space per AG */
123 	int			m_dalign;	/* stripe unit */
124 	int			m_swidth;	/* stripe width */
125 	xfs_agnumber_t		m_maxagi;	/* highest inode alloc group */
126 	uint			m_allocsize_log;/* min write size log bytes */
127 	uint			m_allocsize_blocks; /* min write size blocks */
128 	int			m_logbufs;	/* number of log buffers */
129 	int			m_logbsize;	/* size of each log buffer */
130 	uint			m_rsumlevels;	/* rt summary levels */
131 	uint			m_rsumsize;	/* size of rt summary, bytes */
132 	int			m_fixedfsid[2];	/* unchanged for life of FS */
133 	uint			m_qflags;	/* quota status flags */
134 	uint64_t		m_flags;	/* global mount flags */
135 	int64_t			m_low_space[XFS_LOWSP_MAX];
136 	struct xfs_ino_geometry	m_ino_geo;	/* inode geometry */
137 	struct xfs_trans_resv	m_resv;		/* precomputed res values */
138 						/* low free space thresholds */
139 	bool			m_always_cow;
140 	bool			m_fail_unmount;
141 	bool			m_finobt_nores; /* no per-AG finobt resv. */
142 	bool			m_update_sb;	/* sb needs update in mount */
143 
144 	/*
145 	 * Bitsets of per-fs metadata that have been checked and/or are sick.
146 	 * Callers must hold m_sb_lock to access these two fields.
147 	 */
148 	uint8_t			m_fs_checked;
149 	uint8_t			m_fs_sick;
150 	/*
151 	 * Bitsets of rt metadata that have been checked and/or are sick.
152 	 * Callers must hold m_sb_lock to access this field.
153 	 */
154 	uint8_t			m_rt_checked;
155 	uint8_t			m_rt_sick;
156 
157 	/*
158 	 * End of read-mostly variables. Frequently written variables and locks
159 	 * should be placed below this comment from now on. The first variable
160 	 * here is marked as cacheline aligned so they it is separated from
161 	 * the read-mostly variables.
162 	 */
163 
164 	spinlock_t ____cacheline_aligned m_sb_lock; /* sb counter lock */
165 	struct percpu_counter	m_icount;	/* allocated inodes counter */
166 	struct percpu_counter	m_ifree;	/* free inodes counter */
167 	struct percpu_counter	m_fdblocks;	/* free block counter */
168 	/*
169 	 * Count of data device blocks reserved for delayed allocations,
170 	 * including indlen blocks.  Does not include allocated CoW staging
171 	 * extents or anything related to the rt device.
172 	 */
173 	struct percpu_counter	m_delalloc_blks;
174 	/*
175 	 * Global count of allocation btree blocks in use across all AGs. Only
176 	 * used when perag reservation is enabled. Helps prevent block
177 	 * reservation from attempting to reserve allocation btree blocks.
178 	 */
179 	atomic64_t		m_allocbt_blks;
180 
181 	struct radix_tree_root	m_perag_tree;	/* per-ag accounting info */
182 	spinlock_t		m_perag_lock;	/* lock for m_perag_tree */
183 	uint64_t		m_resblks;	/* total reserved blocks */
184 	uint64_t		m_resblks_avail;/* available reserved blocks */
185 	uint64_t		m_resblks_save;	/* reserved blks @ remount,ro */
186 	struct delayed_work	m_reclaim_work;	/* background inode reclaim */
187 	struct xfs_kobj		m_kobj;
188 	struct xfs_kobj		m_error_kobj;
189 	struct xfs_kobj		m_error_meta_kobj;
190 	struct xfs_error_cfg	m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX];
191 	struct xstats		m_stats;	/* per-fs stats */
192 	xfs_agnumber_t		m_agfrotor;	/* last ag where space found */
193 	xfs_agnumber_t		m_agirotor;	/* last ag dir inode alloced */
194 	spinlock_t		m_agirotor_lock;/* .. and lock protecting it */
195 
196 	/*
197 	 * Workqueue item so that we can coalesce multiple inode flush attempts
198 	 * into a single flush.
199 	 */
200 	struct work_struct	m_flush_inodes_work;
201 
202 	/*
203 	 * Generation of the filesysyem layout.  This is incremented by each
204 	 * growfs, and used by the pNFS server to ensure the client updates
205 	 * its view of the block device once it gets a layout that might
206 	 * reference the newly added blocks.  Does not need to be persistent
207 	 * as long as we only allow file system size increments, but if we
208 	 * ever support shrinks it would have to be persisted in addition
209 	 * to various other kinds of pain inflicted on the pNFS server.
210 	 */
211 	uint32_t		m_generation;
212 	struct mutex		m_growlock;	/* growfs mutex */
213 
214 #ifdef DEBUG
215 	/*
216 	 * Frequency with which errors are injected.  Replaces xfs_etest; the
217 	 * value stored in here is the inverse of the frequency with which the
218 	 * error triggers.  1 = always, 2 = half the time, etc.
219 	 */
220 	unsigned int		*m_errortag;
221 	struct xfs_kobj		m_errortag_kobj;
222 #endif
223 } xfs_mount_t;
224 
225 #define M_IGEO(mp)		(&(mp)->m_ino_geo)
226 
227 /*
228  * Flags for m_flags.
229  */
230 #define XFS_MOUNT_WSYNC		(1ULL << 0)	/* for nfs - all metadata ops
231 						   must be synchronous except
232 						   for space allocations */
233 #define XFS_MOUNT_UNMOUNTING	(1ULL << 1)	/* filesystem is unmounting */
234 #define XFS_MOUNT_WAS_CLEAN	(1ULL << 3)
235 #define XFS_MOUNT_FS_SHUTDOWN	(1ULL << 4)	/* atomic stop of all filesystem
236 						   operations, typically for
237 						   disk errors in metadata */
238 #define XFS_MOUNT_DISCARD	(1ULL << 5)	/* discard unused blocks */
239 #define XFS_MOUNT_NOALIGN	(1ULL << 7)	/* turn off stripe alignment
240 						   allocations */
241 #define XFS_MOUNT_ATTR2		(1ULL << 8)	/* allow use of attr2 format */
242 #define XFS_MOUNT_GRPID		(1ULL << 9)	/* group-ID assigned from directory */
243 #define XFS_MOUNT_NORECOVERY	(1ULL << 10)	/* no recovery - dirty fs */
244 #define XFS_MOUNT_ALLOCSIZE	(1ULL << 12)	/* specified allocation size */
245 #define XFS_MOUNT_SMALL_INUMS	(1ULL << 14)	/* user wants 32bit inodes */
246 #define XFS_MOUNT_32BITINODES	(1ULL << 15)	/* inode32 allocator active */
247 #define XFS_MOUNT_NOUUID	(1ULL << 16)	/* ignore uuid during mount */
248 #define XFS_MOUNT_IKEEP		(1ULL << 18)	/* keep empty inode clusters*/
249 #define XFS_MOUNT_SWALLOC	(1ULL << 19)	/* turn on stripe width
250 						 * allocation */
251 #define XFS_MOUNT_RDONLY	(1ULL << 20)	/* read-only fs */
252 #define XFS_MOUNT_DIRSYNC	(1ULL << 21)	/* synchronous directory ops */
253 #define XFS_MOUNT_LARGEIO	(1ULL << 22)	/* report large preferred
254 						 * I/O size in stat() */
255 #define XFS_MOUNT_FILESTREAMS	(1ULL << 24)	/* enable the filestreams
256 						   allocator */
257 #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
258 #define XFS_MOUNT_DAX_ALWAYS	(1ULL << 26)
259 #define XFS_MOUNT_DAX_NEVER	(1ULL << 27)
260 
261 /*
262  * Max and min values for mount-option defined I/O
263  * preallocation sizes.
264  */
265 #define XFS_MAX_IO_LOG		30	/* 1G */
266 #define XFS_MIN_IO_LOG		PAGE_SHIFT
267 
268 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp)	\
269 				((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
270 #define XFS_FORCED_SHUTDOWN(mp)	((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN)
271 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
272 		int lnnum);
273 #define xfs_force_shutdown(m,f)	\
274 	xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
275 
276 #define SHUTDOWN_META_IO_ERROR	0x0001	/* write attempt to metadata failed */
277 #define SHUTDOWN_LOG_IO_ERROR	0x0002	/* write attempt to the log failed */
278 #define SHUTDOWN_FORCE_UMOUNT	0x0004	/* shutdown from a forced unmount */
279 #define SHUTDOWN_CORRUPT_INCORE	0x0008	/* corrupt in-memory data structures */
280 
281 /*
282  * Flags for xfs_mountfs
283  */
284 #define XFS_MFSI_QUIET		0x40	/* Be silent if mount errors found */
285 
286 static inline xfs_agnumber_t
287 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
288 {
289 	xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
290 	do_div(ld, mp->m_sb.sb_agblocks);
291 	return (xfs_agnumber_t) ld;
292 }
293 
294 static inline xfs_agblock_t
295 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
296 {
297 	xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
298 	return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
299 }
300 
301 int xfs_buf_hash_init(struct xfs_perag *pag);
302 void xfs_buf_hash_destroy(struct xfs_perag *pag);
303 
304 extern void	xfs_uuid_table_free(void);
305 extern uint64_t xfs_default_resblks(xfs_mount_t *mp);
306 extern int	xfs_mountfs(xfs_mount_t *mp);
307 extern void	xfs_unmountfs(xfs_mount_t *);
308 
309 extern int	xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
310 				 bool reserved);
311 extern int	xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
312 
313 extern int	xfs_readsb(xfs_mount_t *, int);
314 extern void	xfs_freesb(xfs_mount_t *);
315 extern bool	xfs_fs_writable(struct xfs_mount *mp, int level);
316 extern int	xfs_sb_validate_fsb_count(struct xfs_sb *, uint64_t);
317 
318 extern int	xfs_dev_is_read_only(struct xfs_mount *, char *);
319 
320 extern void	xfs_set_low_space_thresholds(struct xfs_mount *);
321 
322 int	xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb,
323 			xfs_off_t count_fsb);
324 
325 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp,
326 		int error_class, int error);
327 void xfs_force_summary_recalc(struct xfs_mount *mp);
328 void xfs_mod_delalloc(struct xfs_mount *mp, int64_t delta);
329 
330 #endif	/* __XFS_MOUNT_H__ */
331