xref: /openbmc/linux/fs/xfs/xfs_mount.h (revision 5e29a910)
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 #ifdef __KERNEL__
22 
23 struct xlog;
24 struct xfs_inode;
25 struct xfs_mru_cache;
26 struct xfs_nameops;
27 struct xfs_ail;
28 struct xfs_quotainfo;
29 struct xfs_dir_ops;
30 struct xfs_da_geometry;
31 
32 #ifdef HAVE_PERCPU_SB
33 
34 /*
35  * Valid per-cpu incore superblock counters. Note that if you add new counters,
36  * you may need to define new counter disabled bit field descriptors as there
37  * are more possible fields in the superblock that can fit in a bitfield on a
38  * 32 bit platform. The XFS_SBS_* values for the current current counters just
39  * fit.
40  */
41 typedef struct xfs_icsb_cnts {
42 	uint64_t	icsb_fdblocks;
43 	uint64_t	icsb_ifree;
44 	uint64_t	icsb_icount;
45 	unsigned long	icsb_flags;
46 } xfs_icsb_cnts_t;
47 
48 #define XFS_ICSB_FLAG_LOCK	(1 << 0)	/* counter lock bit */
49 
50 #define XFS_ICSB_LAZY_COUNT	(1 << 1)	/* accuracy not needed */
51 
52 extern int	xfs_icsb_init_counters(struct xfs_mount *);
53 extern void	xfs_icsb_reinit_counters(struct xfs_mount *);
54 extern void	xfs_icsb_destroy_counters(struct xfs_mount *);
55 extern void	xfs_icsb_sync_counters(struct xfs_mount *, int);
56 extern void	xfs_icsb_sync_counters_locked(struct xfs_mount *, int);
57 extern int	xfs_icsb_modify_counters(struct xfs_mount *, xfs_sb_field_t,
58 						int64_t, int);
59 
60 #else
61 #define xfs_icsb_init_counters(mp)		(0)
62 #define xfs_icsb_destroy_counters(mp)		do { } while (0)
63 #define xfs_icsb_reinit_counters(mp)		do { } while (0)
64 #define xfs_icsb_sync_counters(mp, flags)	do { } while (0)
65 #define xfs_icsb_sync_counters_locked(mp, flags) do { } while (0)
66 #define xfs_icsb_modify_counters(mp, field, delta, rsvd) \
67 	xfs_mod_incore_sb(mp, field, delta, rsvd)
68 #endif
69 
70 /* dynamic preallocation free space thresholds, 5% down to 1% */
71 enum {
72 	XFS_LOWSP_1_PCNT = 0,
73 	XFS_LOWSP_2_PCNT,
74 	XFS_LOWSP_3_PCNT,
75 	XFS_LOWSP_4_PCNT,
76 	XFS_LOWSP_5_PCNT,
77 	XFS_LOWSP_MAX,
78 };
79 
80 typedef struct xfs_mount {
81 	struct super_block	*m_super;
82 	xfs_tid_t		m_tid;		/* next unused tid for fs */
83 	struct xfs_ail		*m_ail;		/* fs active log item list */
84 	xfs_sb_t		m_sb;		/* copy of fs superblock */
85 	spinlock_t		m_sb_lock;	/* sb counter lock */
86 	struct xfs_buf		*m_sb_bp;	/* buffer for superblock */
87 	char			*m_fsname;	/* filesystem name */
88 	int			m_fsname_len;	/* strlen of fs name */
89 	char			*m_rtname;	/* realtime device name */
90 	char			*m_logname;	/* external log device name */
91 	int			m_bsize;	/* fs logical block size */
92 	xfs_agnumber_t		m_agfrotor;	/* last ag where space found */
93 	xfs_agnumber_t		m_agirotor;	/* last ag dir inode alloced */
94 	spinlock_t		m_agirotor_lock;/* .. and lock protecting it */
95 	xfs_agnumber_t		m_maxagi;	/* highest inode alloc group */
96 	uint			m_readio_log;	/* min read size log bytes */
97 	uint			m_readio_blocks; /* min read size blocks */
98 	uint			m_writeio_log;	/* min write size log bytes */
99 	uint			m_writeio_blocks; /* min write size blocks */
100 	struct xfs_da_geometry	*m_dir_geo;	/* directory block geometry */
101 	struct xfs_da_geometry	*m_attr_geo;	/* attribute block geometry */
102 	struct xlog		*m_log;		/* log specific stuff */
103 	int			m_logbufs;	/* number of log buffers */
104 	int			m_logbsize;	/* size of each log buffer */
105 	uint			m_rsumlevels;	/* rt summary levels */
106 	uint			m_rsumsize;	/* size of rt summary, bytes */
107 	struct xfs_inode	*m_rbmip;	/* pointer to bitmap inode */
108 	struct xfs_inode	*m_rsumip;	/* pointer to summary inode */
109 	struct xfs_inode	*m_rootip;	/* pointer to root directory */
110 	struct xfs_quotainfo	*m_quotainfo;	/* disk quota information */
111 	xfs_buftarg_t		*m_ddev_targp;	/* saves taking the address */
112 	xfs_buftarg_t		*m_logdev_targp;/* ptr to log device */
113 	xfs_buftarg_t		*m_rtdev_targp;	/* ptr to rt device */
114 	__uint8_t		m_blkbit_log;	/* blocklog + NBBY */
115 	__uint8_t		m_blkbb_log;	/* blocklog - BBSHIFT */
116 	__uint8_t		m_agno_log;	/* log #ag's */
117 	__uint8_t		m_agino_log;	/* #bits for agino in inum */
118 	uint			m_inode_cluster_size;/* min inode buf size */
119 	uint			m_blockmask;	/* sb_blocksize-1 */
120 	uint			m_blockwsize;	/* sb_blocksize in words */
121 	uint			m_blockwmask;	/* blockwsize-1 */
122 	uint			m_alloc_mxr[2];	/* max alloc btree records */
123 	uint			m_alloc_mnr[2];	/* min alloc btree records */
124 	uint			m_bmap_dmxr[2];	/* max bmap btree records */
125 	uint			m_bmap_dmnr[2];	/* min bmap btree records */
126 	uint			m_inobt_mxr[2];	/* max inobt btree records */
127 	uint			m_inobt_mnr[2];	/* min inobt btree records */
128 	uint			m_ag_maxlevels;	/* XFS_AG_MAXLEVELS */
129 	uint			m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
130 	uint			m_in_maxlevels;	/* max inobt btree levels. */
131 	struct radix_tree_root	m_perag_tree;	/* per-ag accounting info */
132 	spinlock_t		m_perag_lock;	/* lock for m_perag_tree */
133 	struct mutex		m_growlock;	/* growfs mutex */
134 	int			m_fixedfsid[2];	/* unchanged for life of FS */
135 	uint			m_dmevmask;	/* DMI events for this FS */
136 	__uint64_t		m_flags;	/* global mount flags */
137 	int			m_ialloc_inos;	/* inodes in inode allocation */
138 	int			m_ialloc_blks;	/* blocks in inode allocation */
139 	int			m_inoalign_mask;/* mask sb_inoalignmt if used */
140 	uint			m_qflags;	/* quota status flags */
141 	struct xfs_trans_resv	m_resv;		/* precomputed res values */
142 	__uint64_t		m_maxicount;	/* maximum inode count */
143 	__uint64_t		m_resblks;	/* total reserved blocks */
144 	__uint64_t		m_resblks_avail;/* available reserved blocks */
145 	__uint64_t		m_resblks_save;	/* reserved blks @ remount,ro */
146 	int			m_dalign;	/* stripe unit */
147 	int			m_swidth;	/* stripe width */
148 	int			m_sinoalign;	/* stripe unit inode alignment */
149 	__uint8_t		m_sectbb_log;	/* sectlog - BBSHIFT */
150 	const struct xfs_nameops *m_dirnameops;	/* vector of dir name ops */
151 	const struct xfs_dir_ops *m_dir_inode_ops; /* vector of dir inode ops */
152 	const struct xfs_dir_ops *m_nondir_inode_ops; /* !dir inode ops */
153 	uint			m_chsize;	/* size of next field */
154 	atomic_t		m_active_trans;	/* number trans frozen */
155 #ifdef HAVE_PERCPU_SB
156 	xfs_icsb_cnts_t __percpu *m_sb_cnts;	/* per-cpu superblock counters */
157 	unsigned long		m_icsb_counters; /* disabled per-cpu counters */
158 	struct notifier_block	m_icsb_notifier; /* hotplug cpu notifier */
159 	struct mutex		m_icsb_mutex;	/* balancer sync lock */
160 #endif
161 	struct xfs_mru_cache	*m_filestream;  /* per-mount filestream data */
162 	struct delayed_work	m_reclaim_work;	/* background inode reclaim */
163 	struct delayed_work	m_eofblocks_work; /* background eof blocks
164 						     trimming */
165 	bool			m_update_sb;	/* sb needs update in mount */
166 	int64_t			m_low_space[XFS_LOWSP_MAX];
167 						/* low free space thresholds */
168 	struct xfs_kobj		m_kobj;
169 
170 	struct workqueue_struct *m_buf_workqueue;
171 	struct workqueue_struct	*m_data_workqueue;
172 	struct workqueue_struct	*m_unwritten_workqueue;
173 	struct workqueue_struct	*m_cil_workqueue;
174 	struct workqueue_struct	*m_reclaim_workqueue;
175 	struct workqueue_struct	*m_log_workqueue;
176 	struct workqueue_struct *m_eofblocks_workqueue;
177 
178 	/*
179 	 * Generation of the filesysyem layout.  This is incremented by each
180 	 * growfs, and used by the pNFS server to ensure the client updates
181 	 * its view of the block device once it gets a layout that might
182 	 * reference the newly added blocks.  Does not need to be persistent
183 	 * as long as we only allow file system size increments, but if we
184 	 * ever support shrinks it would have to be persisted in addition
185 	 * to various other kinds of pain inflicted on the pNFS server.
186 	 */
187 	__uint32_t		m_generation;
188 } xfs_mount_t;
189 
190 /*
191  * Flags for m_flags.
192  */
193 #define XFS_MOUNT_WSYNC		(1ULL << 0)	/* for nfs - all metadata ops
194 						   must be synchronous except
195 						   for space allocations */
196 #define XFS_MOUNT_WAS_CLEAN	(1ULL << 3)
197 #define XFS_MOUNT_FS_SHUTDOWN	(1ULL << 4)	/* atomic stop of all filesystem
198 						   operations, typically for
199 						   disk errors in metadata */
200 #define XFS_MOUNT_DISCARD	(1ULL << 5)	/* discard unused blocks */
201 #define XFS_MOUNT_NOALIGN	(1ULL << 7)	/* turn off stripe alignment
202 						   allocations */
203 #define XFS_MOUNT_ATTR2		(1ULL << 8)	/* allow use of attr2 format */
204 #define XFS_MOUNT_GRPID		(1ULL << 9)	/* group-ID assigned from directory */
205 #define XFS_MOUNT_NORECOVERY	(1ULL << 10)	/* no recovery - dirty fs */
206 #define XFS_MOUNT_DFLT_IOSIZE	(1ULL << 12)	/* set default i/o size */
207 #define XFS_MOUNT_32BITINODES	(1ULL << 14)	/* do not create inodes above
208 						 * 32 bits in size */
209 #define XFS_MOUNT_SMALL_INUMS	(1ULL << 15)	/* users wants 32bit inodes */
210 #define XFS_MOUNT_NOUUID	(1ULL << 16)	/* ignore uuid during mount */
211 #define XFS_MOUNT_BARRIER	(1ULL << 17)
212 #define XFS_MOUNT_IKEEP		(1ULL << 18)	/* keep empty inode clusters*/
213 #define XFS_MOUNT_SWALLOC	(1ULL << 19)	/* turn on stripe width
214 						 * allocation */
215 #define XFS_MOUNT_RDONLY	(1ULL << 20)	/* read-only fs */
216 #define XFS_MOUNT_DIRSYNC	(1ULL << 21)	/* synchronous directory ops */
217 #define XFS_MOUNT_COMPAT_IOSIZE	(1ULL << 22)	/* don't report large preferred
218 						 * I/O size in stat() */
219 #define XFS_MOUNT_FILESTREAMS	(1ULL << 24)	/* enable the filestreams
220 						   allocator */
221 #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
222 
223 
224 /*
225  * Default minimum read and write sizes.
226  */
227 #define XFS_READIO_LOG_LARGE	16
228 #define XFS_WRITEIO_LOG_LARGE	16
229 
230 /*
231  * Max and min values for mount-option defined I/O
232  * preallocation sizes.
233  */
234 #define XFS_MAX_IO_LOG		30	/* 1G */
235 #define XFS_MIN_IO_LOG		PAGE_SHIFT
236 
237 /*
238  * Synchronous read and write sizes.  This should be
239  * better for NFSv2 wsync filesystems.
240  */
241 #define	XFS_WSYNC_READIO_LOG	15	/* 32k */
242 #define	XFS_WSYNC_WRITEIO_LOG	14	/* 16k */
243 
244 /*
245  * Allow large block sizes to be reported to userspace programs if the
246  * "largeio" mount option is used.
247  *
248  * If compatibility mode is specified, simply return the basic unit of caching
249  * so that we don't get inefficient read/modify/write I/O from user apps.
250  * Otherwise....
251  *
252  * If the underlying volume is a stripe, then return the stripe width in bytes
253  * as the recommended I/O size. It is not a stripe and we've set a default
254  * buffered I/O size, return that, otherwise return the compat default.
255  */
256 static inline unsigned long
257 xfs_preferred_iosize(xfs_mount_t *mp)
258 {
259 	if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
260 		return PAGE_CACHE_SIZE;
261 	return (mp->m_swidth ?
262 		(mp->m_swidth << mp->m_sb.sb_blocklog) :
263 		((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
264 			(1 << (int)MAX(mp->m_readio_log, mp->m_writeio_log)) :
265 			PAGE_CACHE_SIZE));
266 }
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 #define SHUTDOWN_REMOTE_REQ	0x0010	/* shutdown came from remote cell */
281 #define SHUTDOWN_DEVICE_REQ	0x0020	/* failed all paths to the device */
282 
283 /*
284  * Flags for xfs_mountfs
285  */
286 #define XFS_MFSI_QUIET		0x40	/* Be silent if mount errors found */
287 
288 static inline xfs_agnumber_t
289 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
290 {
291 	xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
292 	do_div(ld, mp->m_sb.sb_agblocks);
293 	return (xfs_agnumber_t) ld;
294 }
295 
296 static inline xfs_agblock_t
297 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
298 {
299 	xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
300 	return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
301 }
302 
303 /*
304  * Per-cpu superblock locking functions
305  */
306 #ifdef HAVE_PERCPU_SB
307 static inline void
308 xfs_icsb_lock(xfs_mount_t *mp)
309 {
310 	mutex_lock(&mp->m_icsb_mutex);
311 }
312 
313 static inline void
314 xfs_icsb_unlock(xfs_mount_t *mp)
315 {
316 	mutex_unlock(&mp->m_icsb_mutex);
317 }
318 #else
319 #define xfs_icsb_lock(mp)
320 #define xfs_icsb_unlock(mp)
321 #endif
322 
323 /*
324  * This structure is for use by the xfs_mod_incore_sb_batch() routine.
325  * xfs_growfs can specify a few fields which are more than int limit
326  */
327 typedef struct xfs_mod_sb {
328 	xfs_sb_field_t	msb_field;	/* Field to modify, see below */
329 	int64_t		msb_delta;	/* Change to make to specified field */
330 } xfs_mod_sb_t;
331 
332 /*
333  * Per-ag incore structure, copies of information in agf and agi, to improve the
334  * performance of allocation group selection.
335  */
336 typedef struct xfs_perag {
337 	struct xfs_mount *pag_mount;	/* owner filesystem */
338 	xfs_agnumber_t	pag_agno;	/* AG this structure belongs to */
339 	atomic_t	pag_ref;	/* perag reference count */
340 	char		pagf_init;	/* this agf's entry is initialized */
341 	char		pagi_init;	/* this agi's entry is initialized */
342 	char		pagf_metadata;	/* the agf is preferred to be metadata */
343 	char		pagi_inodeok;	/* The agi is ok for inodes */
344 	__uint8_t	pagf_levels[XFS_BTNUM_AGF];
345 					/* # of levels in bno & cnt btree */
346 	__uint32_t	pagf_flcount;	/* count of blocks in freelist */
347 	xfs_extlen_t	pagf_freeblks;	/* total free blocks */
348 	xfs_extlen_t	pagf_longest;	/* longest free space */
349 	__uint32_t	pagf_btreeblks;	/* # of blocks held in AGF btrees */
350 	xfs_agino_t	pagi_freecount;	/* number of free inodes */
351 	xfs_agino_t	pagi_count;	/* number of allocated inodes */
352 
353 	/*
354 	 * Inode allocation search lookup optimisation.
355 	 * If the pagino matches, the search for new inodes
356 	 * doesn't need to search the near ones again straight away
357 	 */
358 	xfs_agino_t	pagl_pagino;
359 	xfs_agino_t	pagl_leftrec;
360 	xfs_agino_t	pagl_rightrec;
361 	spinlock_t	pagb_lock;	/* lock for pagb_tree */
362 	struct rb_root	pagb_tree;	/* ordered tree of busy extents */
363 
364 	atomic_t        pagf_fstrms;    /* # of filestreams active in this AG */
365 
366 	spinlock_t	pag_ici_lock;	/* incore inode cache lock */
367 	struct radix_tree_root pag_ici_root;	/* incore inode cache root */
368 	int		pag_ici_reclaimable;	/* reclaimable inodes */
369 	struct mutex	pag_ici_reclaim_lock;	/* serialisation point */
370 	unsigned long	pag_ici_reclaim_cursor;	/* reclaim restart point */
371 
372 	/* buffer cache index */
373 	spinlock_t	pag_buf_lock;	/* lock for pag_buf_tree */
374 	struct rb_root	pag_buf_tree;	/* ordered tree of active buffers */
375 
376 	/* for rcu-safe freeing */
377 	struct rcu_head	rcu_head;
378 	int		pagb_count;	/* pagb slots in use */
379 } xfs_perag_t;
380 
381 extern int	xfs_log_sbcount(xfs_mount_t *);
382 extern __uint64_t xfs_default_resblks(xfs_mount_t *mp);
383 extern int	xfs_mountfs(xfs_mount_t *mp);
384 extern int	xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount,
385 				     xfs_agnumber_t *maxagi);
386 
387 extern void	xfs_unmountfs(xfs_mount_t *);
388 extern int	xfs_mod_incore_sb(xfs_mount_t *, xfs_sb_field_t, int64_t, int);
389 extern int	xfs_mod_incore_sb_batch(xfs_mount_t *, xfs_mod_sb_t *,
390 			uint, int);
391 extern int	xfs_mount_log_sb(xfs_mount_t *);
392 extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
393 extern int	xfs_readsb(xfs_mount_t *, int);
394 extern void	xfs_freesb(xfs_mount_t *);
395 extern bool	xfs_fs_writable(struct xfs_mount *mp, int level);
396 extern int	xfs_sb_validate_fsb_count(struct xfs_sb *, __uint64_t);
397 
398 extern int	xfs_dev_is_read_only(struct xfs_mount *, char *);
399 
400 extern void	xfs_set_low_space_thresholds(struct xfs_mount *);
401 
402 #endif	/* __KERNEL__ */
403 
404 #endif	/* __XFS_MOUNT_H__ */
405