xref: /linux/fs/xfs/xfs_mount.h (revision ca55b2fef3a9373fcfc30f82fd26bc7fccbda732)
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 
21 struct xlog;
22 struct xfs_inode;
23 struct xfs_mru_cache;
24 struct xfs_nameops;
25 struct xfs_ail;
26 struct xfs_quotainfo;
27 struct xfs_dir_ops;
28 struct xfs_da_geometry;
29 
30 /* dynamic preallocation free space thresholds, 5% down to 1% */
31 enum {
32 	XFS_LOWSP_1_PCNT = 0,
33 	XFS_LOWSP_2_PCNT,
34 	XFS_LOWSP_3_PCNT,
35 	XFS_LOWSP_4_PCNT,
36 	XFS_LOWSP_5_PCNT,
37 	XFS_LOWSP_MAX,
38 };
39 
40 typedef struct xfs_mount {
41 	struct super_block	*m_super;
42 	xfs_tid_t		m_tid;		/* next unused tid for fs */
43 	struct xfs_ail		*m_ail;		/* fs active log item list */
44 
45 	struct xfs_sb		m_sb;		/* copy of fs superblock */
46 	spinlock_t		m_sb_lock;	/* sb counter lock */
47 	struct percpu_counter	m_icount;	/* allocated inodes counter */
48 	struct percpu_counter	m_ifree;	/* free inodes counter */
49 	struct percpu_counter	m_fdblocks;	/* free block counter */
50 
51 	struct xfs_buf		*m_sb_bp;	/* buffer for superblock */
52 	char			*m_fsname;	/* filesystem name */
53 	int			m_fsname_len;	/* strlen of fs name */
54 	char			*m_rtname;	/* realtime device name */
55 	char			*m_logname;	/* external log device name */
56 	int			m_bsize;	/* fs logical block size */
57 	xfs_agnumber_t		m_agfrotor;	/* last ag where space found */
58 	xfs_agnumber_t		m_agirotor;	/* last ag dir inode alloced */
59 	spinlock_t		m_agirotor_lock;/* .. and lock protecting it */
60 	xfs_agnumber_t		m_maxagi;	/* highest inode alloc group */
61 	uint			m_readio_log;	/* min read size log bytes */
62 	uint			m_readio_blocks; /* min read size blocks */
63 	uint			m_writeio_log;	/* min write size log bytes */
64 	uint			m_writeio_blocks; /* min write size blocks */
65 	struct xfs_da_geometry	*m_dir_geo;	/* directory block geometry */
66 	struct xfs_da_geometry	*m_attr_geo;	/* attribute block geometry */
67 	struct xlog		*m_log;		/* log specific stuff */
68 	int			m_logbufs;	/* number of log buffers */
69 	int			m_logbsize;	/* size of each log buffer */
70 	uint			m_rsumlevels;	/* rt summary levels */
71 	uint			m_rsumsize;	/* size of rt summary, bytes */
72 	struct xfs_inode	*m_rbmip;	/* pointer to bitmap inode */
73 	struct xfs_inode	*m_rsumip;	/* pointer to summary inode */
74 	struct xfs_inode	*m_rootip;	/* pointer to root directory */
75 	struct xfs_quotainfo	*m_quotainfo;	/* disk quota information */
76 	xfs_buftarg_t		*m_ddev_targp;	/* saves taking the address */
77 	xfs_buftarg_t		*m_logdev_targp;/* ptr to log device */
78 	xfs_buftarg_t		*m_rtdev_targp;	/* ptr to rt device */
79 	__uint8_t		m_blkbit_log;	/* blocklog + NBBY */
80 	__uint8_t		m_blkbb_log;	/* blocklog - BBSHIFT */
81 	__uint8_t		m_agno_log;	/* log #ag's */
82 	__uint8_t		m_agino_log;	/* #bits for agino in inum */
83 	uint			m_inode_cluster_size;/* min inode buf size */
84 	uint			m_blockmask;	/* sb_blocksize-1 */
85 	uint			m_blockwsize;	/* sb_blocksize in words */
86 	uint			m_blockwmask;	/* blockwsize-1 */
87 	uint			m_alloc_mxr[2];	/* max alloc btree records */
88 	uint			m_alloc_mnr[2];	/* min alloc btree records */
89 	uint			m_bmap_dmxr[2];	/* max bmap btree records */
90 	uint			m_bmap_dmnr[2];	/* min bmap btree records */
91 	uint			m_inobt_mxr[2];	/* max inobt btree records */
92 	uint			m_inobt_mnr[2];	/* min inobt btree records */
93 	uint			m_ag_maxlevels;	/* XFS_AG_MAXLEVELS */
94 	uint			m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
95 	uint			m_in_maxlevels;	/* max inobt btree levels. */
96 	struct radix_tree_root	m_perag_tree;	/* per-ag accounting info */
97 	spinlock_t		m_perag_lock;	/* lock for m_perag_tree */
98 	struct mutex		m_growlock;	/* growfs mutex */
99 	int			m_fixedfsid[2];	/* unchanged for life of FS */
100 	uint			m_dmevmask;	/* DMI events for this FS */
101 	__uint64_t		m_flags;	/* global mount flags */
102 	int			m_ialloc_inos;	/* inodes in inode allocation */
103 	int			m_ialloc_blks;	/* blocks in inode allocation */
104 	int			m_ialloc_min_blks;/* min blocks in sparse inode
105 						   * allocation */
106 	int			m_inoalign_mask;/* mask sb_inoalignmt if used */
107 	uint			m_qflags;	/* quota status flags */
108 	struct xfs_trans_resv	m_resv;		/* precomputed res values */
109 	__uint64_t		m_maxicount;	/* maximum inode count */
110 	__uint64_t		m_resblks;	/* total reserved blocks */
111 	__uint64_t		m_resblks_avail;/* available reserved blocks */
112 	__uint64_t		m_resblks_save;	/* reserved blks @ remount,ro */
113 	int			m_dalign;	/* stripe unit */
114 	int			m_swidth;	/* stripe width */
115 	int			m_sinoalign;	/* stripe unit inode alignment */
116 	__uint8_t		m_sectbb_log;	/* sectlog - BBSHIFT */
117 	const struct xfs_nameops *m_dirnameops;	/* vector of dir name ops */
118 	const struct xfs_dir_ops *m_dir_inode_ops; /* vector of dir inode ops */
119 	const struct xfs_dir_ops *m_nondir_inode_ops; /* !dir inode ops */
120 	uint			m_chsize;	/* size of next field */
121 	atomic_t		m_active_trans;	/* number trans frozen */
122 	struct xfs_mru_cache	*m_filestream;  /* per-mount filestream data */
123 	struct delayed_work	m_reclaim_work;	/* background inode reclaim */
124 	struct delayed_work	m_eofblocks_work; /* background eof blocks
125 						     trimming */
126 	bool			m_update_sb;	/* sb needs update in mount */
127 	int64_t			m_low_space[XFS_LOWSP_MAX];
128 						/* low free space thresholds */
129 	struct xfs_kobj		m_kobj;
130 
131 	struct workqueue_struct *m_buf_workqueue;
132 	struct workqueue_struct	*m_data_workqueue;
133 	struct workqueue_struct	*m_unwritten_workqueue;
134 	struct workqueue_struct	*m_cil_workqueue;
135 	struct workqueue_struct	*m_reclaim_workqueue;
136 	struct workqueue_struct	*m_log_workqueue;
137 	struct workqueue_struct *m_eofblocks_workqueue;
138 
139 	/*
140 	 * Generation of the filesysyem layout.  This is incremented by each
141 	 * growfs, and used by the pNFS server to ensure the client updates
142 	 * its view of the block device once it gets a layout that might
143 	 * reference the newly added blocks.  Does not need to be persistent
144 	 * as long as we only allow file system size increments, but if we
145 	 * ever support shrinks it would have to be persisted in addition
146 	 * to various other kinds of pain inflicted on the pNFS server.
147 	 */
148 	__uint32_t		m_generation;
149 } xfs_mount_t;
150 
151 /*
152  * Flags for m_flags.
153  */
154 #define XFS_MOUNT_WSYNC		(1ULL << 0)	/* for nfs - all metadata ops
155 						   must be synchronous except
156 						   for space allocations */
157 #define XFS_MOUNT_WAS_CLEAN	(1ULL << 3)
158 #define XFS_MOUNT_FS_SHUTDOWN	(1ULL << 4)	/* atomic stop of all filesystem
159 						   operations, typically for
160 						   disk errors in metadata */
161 #define XFS_MOUNT_DISCARD	(1ULL << 5)	/* discard unused blocks */
162 #define XFS_MOUNT_NOALIGN	(1ULL << 7)	/* turn off stripe alignment
163 						   allocations */
164 #define XFS_MOUNT_ATTR2		(1ULL << 8)	/* allow use of attr2 format */
165 #define XFS_MOUNT_GRPID		(1ULL << 9)	/* group-ID assigned from directory */
166 #define XFS_MOUNT_NORECOVERY	(1ULL << 10)	/* no recovery - dirty fs */
167 #define XFS_MOUNT_DFLT_IOSIZE	(1ULL << 12)	/* set default i/o size */
168 #define XFS_MOUNT_32BITINODES	(1ULL << 14)	/* do not create inodes above
169 						 * 32 bits in size */
170 #define XFS_MOUNT_SMALL_INUMS	(1ULL << 15)	/* users wants 32bit inodes */
171 #define XFS_MOUNT_NOUUID	(1ULL << 16)	/* ignore uuid during mount */
172 #define XFS_MOUNT_BARRIER	(1ULL << 17)
173 #define XFS_MOUNT_IKEEP		(1ULL << 18)	/* keep empty inode clusters*/
174 #define XFS_MOUNT_SWALLOC	(1ULL << 19)	/* turn on stripe width
175 						 * allocation */
176 #define XFS_MOUNT_RDONLY	(1ULL << 20)	/* read-only fs */
177 #define XFS_MOUNT_DIRSYNC	(1ULL << 21)	/* synchronous directory ops */
178 #define XFS_MOUNT_COMPAT_IOSIZE	(1ULL << 22)	/* don't report large preferred
179 						 * I/O size in stat() */
180 #define XFS_MOUNT_FILESTREAMS	(1ULL << 24)	/* enable the filestreams
181 						   allocator */
182 #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
183 
184 #define XFS_MOUNT_DAX		(1ULL << 62)	/* TEST ONLY! */
185 
186 
187 /*
188  * Default minimum read and write sizes.
189  */
190 #define XFS_READIO_LOG_LARGE	16
191 #define XFS_WRITEIO_LOG_LARGE	16
192 
193 /*
194  * Max and min values for mount-option defined I/O
195  * preallocation sizes.
196  */
197 #define XFS_MAX_IO_LOG		30	/* 1G */
198 #define XFS_MIN_IO_LOG		PAGE_SHIFT
199 
200 /*
201  * Synchronous read and write sizes.  This should be
202  * better for NFSv2 wsync filesystems.
203  */
204 #define	XFS_WSYNC_READIO_LOG	15	/* 32k */
205 #define	XFS_WSYNC_WRITEIO_LOG	14	/* 16k */
206 
207 /*
208  * Allow large block sizes to be reported to userspace programs if the
209  * "largeio" mount option is used.
210  *
211  * If compatibility mode is specified, simply return the basic unit of caching
212  * so that we don't get inefficient read/modify/write I/O from user apps.
213  * Otherwise....
214  *
215  * If the underlying volume is a stripe, then return the stripe width in bytes
216  * as the recommended I/O size. It is not a stripe and we've set a default
217  * buffered I/O size, return that, otherwise return the compat default.
218  */
219 static inline unsigned long
220 xfs_preferred_iosize(xfs_mount_t *mp)
221 {
222 	if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
223 		return PAGE_CACHE_SIZE;
224 	return (mp->m_swidth ?
225 		(mp->m_swidth << mp->m_sb.sb_blocklog) :
226 		((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
227 			(1 << (int)MAX(mp->m_readio_log, mp->m_writeio_log)) :
228 			PAGE_CACHE_SIZE));
229 }
230 
231 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp)	\
232 				((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
233 #define XFS_FORCED_SHUTDOWN(mp)	((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN)
234 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
235 		int lnnum);
236 #define xfs_force_shutdown(m,f)	\
237 	xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
238 
239 #define SHUTDOWN_META_IO_ERROR	0x0001	/* write attempt to metadata failed */
240 #define SHUTDOWN_LOG_IO_ERROR	0x0002	/* write attempt to the log failed */
241 #define SHUTDOWN_FORCE_UMOUNT	0x0004	/* shutdown from a forced unmount */
242 #define SHUTDOWN_CORRUPT_INCORE	0x0008	/* corrupt in-memory data structures */
243 #define SHUTDOWN_REMOTE_REQ	0x0010	/* shutdown came from remote cell */
244 #define SHUTDOWN_DEVICE_REQ	0x0020	/* failed all paths to the device */
245 
246 /*
247  * Flags for xfs_mountfs
248  */
249 #define XFS_MFSI_QUIET		0x40	/* Be silent if mount errors found */
250 
251 static inline xfs_agnumber_t
252 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
253 {
254 	xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
255 	do_div(ld, mp->m_sb.sb_agblocks);
256 	return (xfs_agnumber_t) ld;
257 }
258 
259 static inline xfs_agblock_t
260 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
261 {
262 	xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
263 	return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
264 }
265 
266 /*
267  * Per-ag incore structure, copies of information in agf and agi, to improve the
268  * performance of allocation group selection.
269  */
270 typedef struct xfs_perag {
271 	struct xfs_mount *pag_mount;	/* owner filesystem */
272 	xfs_agnumber_t	pag_agno;	/* AG this structure belongs to */
273 	atomic_t	pag_ref;	/* perag reference count */
274 	char		pagf_init;	/* this agf's entry is initialized */
275 	char		pagi_init;	/* this agi's entry is initialized */
276 	char		pagf_metadata;	/* the agf is preferred to be metadata */
277 	char		pagi_inodeok;	/* The agi is ok for inodes */
278 	__uint8_t	pagf_levels[XFS_BTNUM_AGF];
279 					/* # of levels in bno & cnt btree */
280 	__uint32_t	pagf_flcount;	/* count of blocks in freelist */
281 	xfs_extlen_t	pagf_freeblks;	/* total free blocks */
282 	xfs_extlen_t	pagf_longest;	/* longest free space */
283 	__uint32_t	pagf_btreeblks;	/* # of blocks held in AGF btrees */
284 	xfs_agino_t	pagi_freecount;	/* number of free inodes */
285 	xfs_agino_t	pagi_count;	/* number of allocated inodes */
286 
287 	/*
288 	 * Inode allocation search lookup optimisation.
289 	 * If the pagino matches, the search for new inodes
290 	 * doesn't need to search the near ones again straight away
291 	 */
292 	xfs_agino_t	pagl_pagino;
293 	xfs_agino_t	pagl_leftrec;
294 	xfs_agino_t	pagl_rightrec;
295 	spinlock_t	pagb_lock;	/* lock for pagb_tree */
296 	struct rb_root	pagb_tree;	/* ordered tree of busy extents */
297 
298 	atomic_t        pagf_fstrms;    /* # of filestreams active in this AG */
299 
300 	spinlock_t	pag_ici_lock;	/* incore inode cache lock */
301 	struct radix_tree_root pag_ici_root;	/* incore inode cache root */
302 	int		pag_ici_reclaimable;	/* reclaimable inodes */
303 	struct mutex	pag_ici_reclaim_lock;	/* serialisation point */
304 	unsigned long	pag_ici_reclaim_cursor;	/* reclaim restart point */
305 
306 	/* buffer cache index */
307 	spinlock_t	pag_buf_lock;	/* lock for pag_buf_tree */
308 	struct rb_root	pag_buf_tree;	/* ordered tree of active buffers */
309 
310 	/* for rcu-safe freeing */
311 	struct rcu_head	rcu_head;
312 	int		pagb_count;	/* pagb slots in use */
313 } xfs_perag_t;
314 
315 extern int	xfs_log_sbcount(xfs_mount_t *);
316 extern __uint64_t xfs_default_resblks(xfs_mount_t *mp);
317 extern int	xfs_mountfs(xfs_mount_t *mp);
318 extern int	xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount,
319 				     xfs_agnumber_t *maxagi);
320 extern void	xfs_unmountfs(xfs_mount_t *);
321 
322 extern int	xfs_mod_icount(struct xfs_mount *mp, int64_t delta);
323 extern int	xfs_mod_ifree(struct xfs_mount *mp, int64_t delta);
324 extern int	xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
325 				 bool reserved);
326 extern int	xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
327 
328 extern int	xfs_mount_log_sb(xfs_mount_t *);
329 extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
330 extern int	xfs_readsb(xfs_mount_t *, int);
331 extern void	xfs_freesb(xfs_mount_t *);
332 extern bool	xfs_fs_writable(struct xfs_mount *mp, int level);
333 extern int	xfs_sb_validate_fsb_count(struct xfs_sb *, __uint64_t);
334 
335 extern int	xfs_dev_is_read_only(struct xfs_mount *, char *);
336 
337 extern void	xfs_set_low_space_thresholds(struct xfs_mount *);
338 
339 #endif	/* __XFS_MOUNT_H__ */
340