xref: /openbmc/linux/fs/xfs/xfs_mount.h (revision 293d5b43)
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 /*
41  * Error Configuration
42  *
43  * Error classes define the subsystem the configuration belongs to.
44  * Error numbers define the errors that are configurable.
45  */
46 enum {
47 	XFS_ERR_METADATA,
48 	XFS_ERR_CLASS_MAX,
49 };
50 enum {
51 	XFS_ERR_DEFAULT,
52 	XFS_ERR_EIO,
53 	XFS_ERR_ENOSPC,
54 	XFS_ERR_ENODEV,
55 	XFS_ERR_ERRNO_MAX,
56 };
57 
58 #define XFS_ERR_RETRY_FOREVER	-1
59 
60 struct xfs_error_cfg {
61 	struct xfs_kobj	kobj;
62 	int		max_retries;
63 	unsigned long	retry_timeout;	/* in jiffies, 0 = no timeout */
64 };
65 
66 typedef struct xfs_mount {
67 	struct super_block	*m_super;
68 	xfs_tid_t		m_tid;		/* next unused tid for fs */
69 	struct xfs_ail		*m_ail;		/* fs active log item list */
70 
71 	struct xfs_sb		m_sb;		/* copy of fs superblock */
72 	spinlock_t		m_sb_lock;	/* sb counter lock */
73 	struct percpu_counter	m_icount;	/* allocated inodes counter */
74 	struct percpu_counter	m_ifree;	/* free inodes counter */
75 	struct percpu_counter	m_fdblocks;	/* free block counter */
76 
77 	struct xfs_buf		*m_sb_bp;	/* buffer for superblock */
78 	char			*m_fsname;	/* filesystem name */
79 	int			m_fsname_len;	/* strlen of fs name */
80 	char			*m_rtname;	/* realtime device name */
81 	char			*m_logname;	/* external log device name */
82 	int			m_bsize;	/* fs logical block size */
83 	xfs_agnumber_t		m_agfrotor;	/* last ag where space found */
84 	xfs_agnumber_t		m_agirotor;	/* last ag dir inode alloced */
85 	spinlock_t		m_agirotor_lock;/* .. and lock protecting it */
86 	xfs_agnumber_t		m_maxagi;	/* highest inode alloc group */
87 	uint			m_readio_log;	/* min read size log bytes */
88 	uint			m_readio_blocks; /* min read size blocks */
89 	uint			m_writeio_log;	/* min write size log bytes */
90 	uint			m_writeio_blocks; /* min write size blocks */
91 	struct xfs_da_geometry	*m_dir_geo;	/* directory block geometry */
92 	struct xfs_da_geometry	*m_attr_geo;	/* attribute block geometry */
93 	struct xlog		*m_log;		/* log specific stuff */
94 	int			m_logbufs;	/* number of log buffers */
95 	int			m_logbsize;	/* size of each log buffer */
96 	uint			m_rsumlevels;	/* rt summary levels */
97 	uint			m_rsumsize;	/* size of rt summary, bytes */
98 	struct xfs_inode	*m_rbmip;	/* pointer to bitmap inode */
99 	struct xfs_inode	*m_rsumip;	/* pointer to summary inode */
100 	struct xfs_inode	*m_rootip;	/* pointer to root directory */
101 	struct xfs_quotainfo	*m_quotainfo;	/* disk quota information */
102 	xfs_buftarg_t		*m_ddev_targp;	/* saves taking the address */
103 	xfs_buftarg_t		*m_logdev_targp;/* ptr to log device */
104 	xfs_buftarg_t		*m_rtdev_targp;	/* ptr to rt device */
105 	__uint8_t		m_blkbit_log;	/* blocklog + NBBY */
106 	__uint8_t		m_blkbb_log;	/* blocklog - BBSHIFT */
107 	__uint8_t		m_agno_log;	/* log #ag's */
108 	__uint8_t		m_agino_log;	/* #bits for agino in inum */
109 	uint			m_inode_cluster_size;/* min inode buf size */
110 	uint			m_blockmask;	/* sb_blocksize-1 */
111 	uint			m_blockwsize;	/* sb_blocksize in words */
112 	uint			m_blockwmask;	/* blockwsize-1 */
113 	uint			m_alloc_mxr[2];	/* max alloc btree records */
114 	uint			m_alloc_mnr[2];	/* min alloc btree records */
115 	uint			m_bmap_dmxr[2];	/* max bmap btree records */
116 	uint			m_bmap_dmnr[2];	/* min bmap btree records */
117 	uint			m_inobt_mxr[2];	/* max inobt btree records */
118 	uint			m_inobt_mnr[2];	/* min inobt btree records */
119 	uint			m_rmap_mxr[2];	/* max rmap btree records */
120 	uint			m_rmap_mnr[2];	/* min rmap btree records */
121 	uint			m_ag_maxlevels;	/* XFS_AG_MAXLEVELS */
122 	uint			m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
123 	uint			m_in_maxlevels;	/* max inobt btree levels. */
124 	uint			m_rmap_maxlevels; /* max rmap btree levels */
125 	xfs_extlen_t		m_ag_prealloc_blocks; /* reserved ag blocks */
126 	uint			m_alloc_set_aside; /* space we can't use */
127 	uint			m_ag_max_usable; /* max space per AG */
128 	struct radix_tree_root	m_perag_tree;	/* per-ag accounting info */
129 	spinlock_t		m_perag_lock;	/* lock for m_perag_tree */
130 	struct mutex		m_growlock;	/* growfs mutex */
131 	int			m_fixedfsid[2];	/* unchanged for life of FS */
132 	uint			m_dmevmask;	/* DMI events for this FS */
133 	__uint64_t		m_flags;	/* global mount flags */
134 	int			m_ialloc_inos;	/* inodes in inode allocation */
135 	int			m_ialloc_blks;	/* blocks in inode allocation */
136 	int			m_ialloc_min_blks;/* min blocks in sparse inode
137 						   * allocation */
138 	int			m_inoalign_mask;/* mask sb_inoalignmt if used */
139 	uint			m_qflags;	/* quota status flags */
140 	struct xfs_trans_resv	m_resv;		/* precomputed res values */
141 	__uint64_t		m_maxicount;	/* maximum inode count */
142 	__uint64_t		m_resblks;	/* total reserved blocks */
143 	__uint64_t		m_resblks_avail;/* available reserved blocks */
144 	__uint64_t		m_resblks_save;	/* reserved blks @ remount,ro */
145 	int			m_dalign;	/* stripe unit */
146 	int			m_swidth;	/* stripe width */
147 	int			m_sinoalign;	/* stripe unit inode alignment */
148 	__uint8_t		m_sectbb_log;	/* sectlog - BBSHIFT */
149 	const struct xfs_nameops *m_dirnameops;	/* vector of dir name ops */
150 	const struct xfs_dir_ops *m_dir_inode_ops; /* vector of dir inode ops */
151 	const struct xfs_dir_ops *m_nondir_inode_ops; /* !dir inode ops */
152 	uint			m_chsize;	/* size of next field */
153 	atomic_t		m_active_trans;	/* number trans frozen */
154 	struct xfs_mru_cache	*m_filestream;  /* per-mount filestream data */
155 	struct delayed_work	m_reclaim_work;	/* background inode reclaim */
156 	struct delayed_work	m_eofblocks_work; /* background eof blocks
157 						     trimming */
158 	bool			m_update_sb;	/* sb needs update in mount */
159 	int64_t			m_low_space[XFS_LOWSP_MAX];
160 						/* low free space thresholds */
161 	struct xfs_kobj		m_kobj;
162 	struct xfs_kobj		m_error_kobj;
163 	struct xfs_kobj		m_error_meta_kobj;
164 	struct xfs_error_cfg	m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX];
165 	struct xstats		m_stats;	/* per-fs stats */
166 
167 	struct workqueue_struct *m_buf_workqueue;
168 	struct workqueue_struct	*m_data_workqueue;
169 	struct workqueue_struct	*m_unwritten_workqueue;
170 	struct workqueue_struct	*m_cil_workqueue;
171 	struct workqueue_struct	*m_reclaim_workqueue;
172 	struct workqueue_struct	*m_log_workqueue;
173 	struct workqueue_struct *m_eofblocks_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 	 * DEBUG mode instrumentation to test and/or trigger delayed allocation
190 	 * block killing in the event of failed writes. When enabled, all
191 	 * buffered writes are forced to fail. All delalloc blocks in the range
192 	 * of the write (including pre-existing delalloc blocks!) are tossed as
193 	 * part of the write failure error handling sequence.
194 	 */
195 	bool			m_fail_writes;
196 #endif
197 } xfs_mount_t;
198 
199 /*
200  * Flags for m_flags.
201  */
202 #define XFS_MOUNT_WSYNC		(1ULL << 0)	/* for nfs - all metadata ops
203 						   must be synchronous except
204 						   for space allocations */
205 #define XFS_MOUNT_UNMOUNTING	(1ULL << 1)	/* filesystem is unmounting */
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_BARRIER	(1ULL << 17)
221 #define XFS_MOUNT_IKEEP		(1ULL << 18)	/* keep empty inode clusters*/
222 #define XFS_MOUNT_SWALLOC	(1ULL << 19)	/* turn on stripe width
223 						 * allocation */
224 #define XFS_MOUNT_RDONLY	(1ULL << 20)	/* read-only fs */
225 #define XFS_MOUNT_DIRSYNC	(1ULL << 21)	/* synchronous directory ops */
226 #define XFS_MOUNT_COMPAT_IOSIZE	(1ULL << 22)	/* don't report large preferred
227 						 * I/O size in stat() */
228 #define XFS_MOUNT_FILESTREAMS	(1ULL << 24)	/* enable the filestreams
229 						   allocator */
230 #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
231 
232 #define XFS_MOUNT_DAX		(1ULL << 62)	/* TEST ONLY! */
233 
234 
235 /*
236  * Default minimum read and write sizes.
237  */
238 #define XFS_READIO_LOG_LARGE	16
239 #define XFS_WRITEIO_LOG_LARGE	16
240 
241 /*
242  * Max and min values for mount-option defined I/O
243  * preallocation sizes.
244  */
245 #define XFS_MAX_IO_LOG		30	/* 1G */
246 #define XFS_MIN_IO_LOG		PAGE_SHIFT
247 
248 /*
249  * Synchronous read and write sizes.  This should be
250  * better for NFSv2 wsync filesystems.
251  */
252 #define	XFS_WSYNC_READIO_LOG	15	/* 32k */
253 #define	XFS_WSYNC_WRITEIO_LOG	14	/* 16k */
254 
255 /*
256  * Allow large block sizes to be reported to userspace programs if the
257  * "largeio" mount option is used.
258  *
259  * If compatibility mode is specified, simply return the basic unit of caching
260  * so that we don't get inefficient read/modify/write I/O from user apps.
261  * Otherwise....
262  *
263  * If the underlying volume is a stripe, then return the stripe width in bytes
264  * as the recommended I/O size. It is not a stripe and we've set a default
265  * buffered I/O size, return that, otherwise return the compat default.
266  */
267 static inline unsigned long
268 xfs_preferred_iosize(xfs_mount_t *mp)
269 {
270 	if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
271 		return PAGE_SIZE;
272 	return (mp->m_swidth ?
273 		(mp->m_swidth << mp->m_sb.sb_blocklog) :
274 		((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
275 			(1 << (int)MAX(mp->m_readio_log, mp->m_writeio_log)) :
276 			PAGE_SIZE));
277 }
278 
279 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp)	\
280 				((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
281 #define XFS_FORCED_SHUTDOWN(mp)	((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN)
282 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
283 		int lnnum);
284 #define xfs_force_shutdown(m,f)	\
285 	xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
286 
287 #define SHUTDOWN_META_IO_ERROR	0x0001	/* write attempt to metadata failed */
288 #define SHUTDOWN_LOG_IO_ERROR	0x0002	/* write attempt to the log failed */
289 #define SHUTDOWN_FORCE_UMOUNT	0x0004	/* shutdown from a forced unmount */
290 #define SHUTDOWN_CORRUPT_INCORE	0x0008	/* corrupt in-memory data structures */
291 #define SHUTDOWN_REMOTE_REQ	0x0010	/* shutdown came from remote cell */
292 #define SHUTDOWN_DEVICE_REQ	0x0020	/* failed all paths to the device */
293 
294 /*
295  * Flags for xfs_mountfs
296  */
297 #define XFS_MFSI_QUIET		0x40	/* Be silent if mount errors found */
298 
299 static inline xfs_agnumber_t
300 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
301 {
302 	xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
303 	do_div(ld, mp->m_sb.sb_agblocks);
304 	return (xfs_agnumber_t) ld;
305 }
306 
307 static inline xfs_agblock_t
308 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
309 {
310 	xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
311 	return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
312 }
313 
314 #ifdef DEBUG
315 static inline bool
316 xfs_mp_fail_writes(struct xfs_mount *mp)
317 {
318 	return mp->m_fail_writes;
319 }
320 #else
321 static inline bool
322 xfs_mp_fail_writes(struct xfs_mount *mp)
323 {
324 	return 0;
325 }
326 #endif
327 
328 /*
329  * Per-ag incore structure, copies of information in agf and agi, to improve the
330  * performance of allocation group selection.
331  */
332 typedef struct xfs_perag {
333 	struct xfs_mount *pag_mount;	/* owner filesystem */
334 	xfs_agnumber_t	pag_agno;	/* AG this structure belongs to */
335 	atomic_t	pag_ref;	/* perag reference count */
336 	char		pagf_init;	/* this agf's entry is initialized */
337 	char		pagi_init;	/* this agi's entry is initialized */
338 	char		pagf_metadata;	/* the agf is preferred to be metadata */
339 	char		pagi_inodeok;	/* The agi is ok for inodes */
340 	__uint8_t	pagf_levels[XFS_BTNUM_AGF];
341 					/* # of levels in bno & cnt btree */
342 	__uint32_t	pagf_flcount;	/* count of blocks in freelist */
343 	xfs_extlen_t	pagf_freeblks;	/* total free blocks */
344 	xfs_extlen_t	pagf_longest;	/* longest free space */
345 	__uint32_t	pagf_btreeblks;	/* # of blocks held in AGF btrees */
346 	xfs_agino_t	pagi_freecount;	/* number of free inodes */
347 	xfs_agino_t	pagi_count;	/* number of allocated inodes */
348 
349 	/*
350 	 * Inode allocation search lookup optimisation.
351 	 * If the pagino matches, the search for new inodes
352 	 * doesn't need to search the near ones again straight away
353 	 */
354 	xfs_agino_t	pagl_pagino;
355 	xfs_agino_t	pagl_leftrec;
356 	xfs_agino_t	pagl_rightrec;
357 	spinlock_t	pagb_lock;	/* lock for pagb_tree */
358 	struct rb_root	pagb_tree;	/* ordered tree of busy extents */
359 
360 	atomic_t        pagf_fstrms;    /* # of filestreams active in this AG */
361 
362 	spinlock_t	pag_ici_lock;	/* incore inode cache lock */
363 	struct radix_tree_root pag_ici_root;	/* incore inode cache root */
364 	int		pag_ici_reclaimable;	/* reclaimable inodes */
365 	struct mutex	pag_ici_reclaim_lock;	/* serialisation point */
366 	unsigned long	pag_ici_reclaim_cursor;	/* reclaim restart point */
367 
368 	/* buffer cache index */
369 	spinlock_t	pag_buf_lock;	/* lock for pag_buf_tree */
370 	struct rb_root	pag_buf_tree;	/* ordered tree of active buffers */
371 
372 	/* for rcu-safe freeing */
373 	struct rcu_head	rcu_head;
374 	int		pagb_count;	/* pagb slots in use */
375 } xfs_perag_t;
376 
377 extern void	xfs_uuid_table_free(void);
378 extern int	xfs_log_sbcount(xfs_mount_t *);
379 extern __uint64_t xfs_default_resblks(xfs_mount_t *mp);
380 extern int	xfs_mountfs(xfs_mount_t *mp);
381 extern int	xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount,
382 				     xfs_agnumber_t *maxagi);
383 extern void	xfs_unmountfs(xfs_mount_t *);
384 
385 extern int	xfs_mod_icount(struct xfs_mount *mp, int64_t delta);
386 extern int	xfs_mod_ifree(struct xfs_mount *mp, int64_t delta);
387 extern int	xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
388 				 bool reserved);
389 extern int	xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
390 
391 extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
392 extern int	xfs_readsb(xfs_mount_t *, int);
393 extern void	xfs_freesb(xfs_mount_t *);
394 extern bool	xfs_fs_writable(struct xfs_mount *mp, int level);
395 extern int	xfs_sb_validate_fsb_count(struct xfs_sb *, __uint64_t);
396 
397 extern int	xfs_dev_is_read_only(struct xfs_mount *, char *);
398 
399 extern void	xfs_set_low_space_thresholds(struct xfs_mount *);
400 
401 int	xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb,
402 			xfs_off_t count_fsb);
403 
404 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp,
405 		int error_class, int error);
406 
407 #endif	/* __XFS_MOUNT_H__ */
408