xref: /openbmc/linux/fs/xfs/xfs_mount.h (revision 239480ab)
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 /*
61  * Although retry_timeout is in jiffies which is normally an unsigned long,
62  * we limit the retry timeout to 86400 seconds, or one day.  So even a
63  * signed 32-bit long is sufficient for a HZ value up to 24855.  Making it
64  * signed lets us store the special "-1" value, meaning retry forever.
65  */
66 struct xfs_error_cfg {
67 	struct xfs_kobj	kobj;
68 	int		max_retries;
69 	long		retry_timeout;	/* in jiffies, -1 = infinite */
70 };
71 
72 typedef struct xfs_mount {
73 	struct super_block	*m_super;
74 	xfs_tid_t		m_tid;		/* next unused tid for fs */
75 	struct xfs_ail		*m_ail;		/* fs active log item list */
76 
77 	struct xfs_sb		m_sb;		/* copy of fs superblock */
78 	spinlock_t		m_sb_lock;	/* sb counter lock */
79 	struct percpu_counter	m_icount;	/* allocated inodes counter */
80 	struct percpu_counter	m_ifree;	/* free inodes counter */
81 	struct percpu_counter	m_fdblocks;	/* free block counter */
82 
83 	struct xfs_buf		*m_sb_bp;	/* buffer for superblock */
84 	char			*m_fsname;	/* filesystem name */
85 	int			m_fsname_len;	/* strlen of fs name */
86 	char			*m_rtname;	/* realtime device name */
87 	char			*m_logname;	/* external log device name */
88 	int			m_bsize;	/* fs logical block size */
89 	xfs_agnumber_t		m_agfrotor;	/* last ag where space found */
90 	xfs_agnumber_t		m_agirotor;	/* last ag dir inode alloced */
91 	spinlock_t		m_agirotor_lock;/* .. and lock protecting it */
92 	xfs_agnumber_t		m_maxagi;	/* highest inode alloc group */
93 	uint			m_readio_log;	/* min read size log bytes */
94 	uint			m_readio_blocks; /* min read size blocks */
95 	uint			m_writeio_log;	/* min write size log bytes */
96 	uint			m_writeio_blocks; /* min write size blocks */
97 	struct xfs_da_geometry	*m_dir_geo;	/* directory block geometry */
98 	struct xfs_da_geometry	*m_attr_geo;	/* attribute block geometry */
99 	struct xlog		*m_log;		/* log specific stuff */
100 	int			m_logbufs;	/* number of log buffers */
101 	int			m_logbsize;	/* size of each log buffer */
102 	uint			m_rsumlevels;	/* rt summary levels */
103 	uint			m_rsumsize;	/* size of rt summary, bytes */
104 	struct xfs_inode	*m_rbmip;	/* pointer to bitmap inode */
105 	struct xfs_inode	*m_rsumip;	/* pointer to summary inode */
106 	struct xfs_inode	*m_rootip;	/* pointer to root directory */
107 	struct xfs_quotainfo	*m_quotainfo;	/* disk quota information */
108 	xfs_buftarg_t		*m_ddev_targp;	/* saves taking the address */
109 	xfs_buftarg_t		*m_logdev_targp;/* ptr to log device */
110 	xfs_buftarg_t		*m_rtdev_targp;	/* ptr to rt device */
111 	__uint8_t		m_blkbit_log;	/* blocklog + NBBY */
112 	__uint8_t		m_blkbb_log;	/* blocklog - BBSHIFT */
113 	__uint8_t		m_agno_log;	/* log #ag's */
114 	__uint8_t		m_agino_log;	/* #bits for agino in inum */
115 	uint			m_inode_cluster_size;/* min inode buf size */
116 	uint			m_blockmask;	/* sb_blocksize-1 */
117 	uint			m_blockwsize;	/* sb_blocksize in words */
118 	uint			m_blockwmask;	/* blockwsize-1 */
119 	uint			m_alloc_mxr[2];	/* max alloc btree records */
120 	uint			m_alloc_mnr[2];	/* min alloc btree records */
121 	uint			m_bmap_dmxr[2];	/* max bmap btree records */
122 	uint			m_bmap_dmnr[2];	/* min bmap btree records */
123 	uint			m_inobt_mxr[2];	/* max inobt btree records */
124 	uint			m_inobt_mnr[2];	/* min inobt btree records */
125 	uint			m_rmap_mxr[2];	/* max rmap btree records */
126 	uint			m_rmap_mnr[2];	/* min rmap btree records */
127 	uint			m_refc_mxr[2];	/* max refc btree records */
128 	uint			m_refc_mnr[2];	/* min refc btree records */
129 	uint			m_ag_maxlevels;	/* XFS_AG_MAXLEVELS */
130 	uint			m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
131 	uint			m_in_maxlevels;	/* max inobt btree levels. */
132 	uint			m_rmap_maxlevels; /* max rmap btree levels */
133 	uint			m_refc_maxlevels; /* max refcount btree level */
134 	xfs_extlen_t		m_ag_prealloc_blocks; /* reserved ag blocks */
135 	uint			m_alloc_set_aside; /* space we can't use */
136 	uint			m_ag_max_usable; /* max space per AG */
137 	struct radix_tree_root	m_perag_tree;	/* per-ag accounting info */
138 	spinlock_t		m_perag_lock;	/* lock for m_perag_tree */
139 	struct mutex		m_growlock;	/* growfs mutex */
140 	int			m_fixedfsid[2];	/* unchanged for life of FS */
141 	uint			m_dmevmask;	/* DMI events for this FS */
142 	__uint64_t		m_flags;	/* global mount flags */
143 	bool			m_inotbt_nores; /* no per-AG finobt resv. */
144 	int			m_ialloc_inos;	/* inodes in inode allocation */
145 	int			m_ialloc_blks;	/* blocks in inode allocation */
146 	int			m_ialloc_min_blks;/* min blocks in sparse inode
147 						   * allocation */
148 	int			m_inoalign_mask;/* mask sb_inoalignmt if used */
149 	uint			m_qflags;	/* quota status flags */
150 	struct xfs_trans_resv	m_resv;		/* precomputed res values */
151 	__uint64_t		m_maxicount;	/* maximum inode count */
152 	__uint64_t		m_resblks;	/* total reserved blocks */
153 	__uint64_t		m_resblks_avail;/* available reserved blocks */
154 	__uint64_t		m_resblks_save;	/* reserved blks @ remount,ro */
155 	int			m_dalign;	/* stripe unit */
156 	int			m_swidth;	/* stripe width */
157 	int			m_sinoalign;	/* stripe unit inode alignment */
158 	__uint8_t		m_sectbb_log;	/* sectlog - BBSHIFT */
159 	const struct xfs_nameops *m_dirnameops;	/* vector of dir name ops */
160 	const struct xfs_dir_ops *m_dir_inode_ops; /* vector of dir inode ops */
161 	const struct xfs_dir_ops *m_nondir_inode_ops; /* !dir inode ops */
162 	uint			m_chsize;	/* size of next field */
163 	atomic_t		m_active_trans;	/* number trans frozen */
164 	struct xfs_mru_cache	*m_filestream;  /* per-mount filestream data */
165 	struct delayed_work	m_reclaim_work;	/* background inode reclaim */
166 	struct delayed_work	m_eofblocks_work; /* background eof blocks
167 						     trimming */
168 	struct delayed_work	m_cowblocks_work; /* background cow blocks
169 						     trimming */
170 	bool			m_update_sb;	/* sb needs update in mount */
171 	int64_t			m_low_space[XFS_LOWSP_MAX];
172 						/* low free space thresholds */
173 	struct xfs_kobj		m_kobj;
174 	struct xfs_kobj		m_error_kobj;
175 	struct xfs_kobj		m_error_meta_kobj;
176 	struct xfs_error_cfg	m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX];
177 	struct xstats		m_stats;	/* per-fs stats */
178 
179 	struct workqueue_struct *m_buf_workqueue;
180 	struct workqueue_struct	*m_data_workqueue;
181 	struct workqueue_struct	*m_unwritten_workqueue;
182 	struct workqueue_struct	*m_cil_workqueue;
183 	struct workqueue_struct	*m_reclaim_workqueue;
184 	struct workqueue_struct	*m_log_workqueue;
185 	struct workqueue_struct *m_eofblocks_workqueue;
186 	struct workqueue_struct	*m_sync_workqueue;
187 
188 	/*
189 	 * Generation of the filesysyem layout.  This is incremented by each
190 	 * growfs, and used by the pNFS server to ensure the client updates
191 	 * its view of the block device once it gets a layout that might
192 	 * reference the newly added blocks.  Does not need to be persistent
193 	 * as long as we only allow file system size increments, but if we
194 	 * ever support shrinks it would have to be persisted in addition
195 	 * to various other kinds of pain inflicted on the pNFS server.
196 	 */
197 	__uint32_t		m_generation;
198 
199 	bool			m_fail_unmount;
200 #ifdef DEBUG
201 	/*
202 	 * DEBUG mode instrumentation to test and/or trigger delayed allocation
203 	 * block killing in the event of failed writes. When enabled, all
204 	 * buffered writes are silenty dropped and handled as if they failed.
205 	 * All delalloc blocks in the range of the write (including pre-existing
206 	 * delalloc blocks!) are tossed as part of the write failure error
207 	 * handling sequence.
208 	 */
209 	bool			m_drop_writes;
210 #endif
211 } xfs_mount_t;
212 
213 /*
214  * Flags for m_flags.
215  */
216 #define XFS_MOUNT_WSYNC		(1ULL << 0)	/* for nfs - all metadata ops
217 						   must be synchronous except
218 						   for space allocations */
219 #define XFS_MOUNT_UNMOUNTING	(1ULL << 1)	/* filesystem is unmounting */
220 #define XFS_MOUNT_WAS_CLEAN	(1ULL << 3)
221 #define XFS_MOUNT_FS_SHUTDOWN	(1ULL << 4)	/* atomic stop of all filesystem
222 						   operations, typically for
223 						   disk errors in metadata */
224 #define XFS_MOUNT_DISCARD	(1ULL << 5)	/* discard unused blocks */
225 #define XFS_MOUNT_NOALIGN	(1ULL << 7)	/* turn off stripe alignment
226 						   allocations */
227 #define XFS_MOUNT_ATTR2		(1ULL << 8)	/* allow use of attr2 format */
228 #define XFS_MOUNT_GRPID		(1ULL << 9)	/* group-ID assigned from directory */
229 #define XFS_MOUNT_NORECOVERY	(1ULL << 10)	/* no recovery - dirty fs */
230 #define XFS_MOUNT_DFLT_IOSIZE	(1ULL << 12)	/* set default i/o size */
231 #define XFS_MOUNT_SMALL_INUMS	(1ULL << 14)	/* user wants 32bit inodes */
232 #define XFS_MOUNT_32BITINODES	(1ULL << 15)	/* inode32 allocator active */
233 #define XFS_MOUNT_NOUUID	(1ULL << 16)	/* ignore uuid during mount */
234 #define XFS_MOUNT_BARRIER	(1ULL << 17)
235 #define XFS_MOUNT_IKEEP		(1ULL << 18)	/* keep empty inode clusters*/
236 #define XFS_MOUNT_SWALLOC	(1ULL << 19)	/* turn on stripe width
237 						 * allocation */
238 #define XFS_MOUNT_RDONLY	(1ULL << 20)	/* read-only fs */
239 #define XFS_MOUNT_DIRSYNC	(1ULL << 21)	/* synchronous directory ops */
240 #define XFS_MOUNT_COMPAT_IOSIZE	(1ULL << 22)	/* don't report large preferred
241 						 * I/O size in stat() */
242 #define XFS_MOUNT_FILESTREAMS	(1ULL << 24)	/* enable the filestreams
243 						   allocator */
244 #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
245 
246 #define XFS_MOUNT_DAX		(1ULL << 62)	/* TEST ONLY! */
247 
248 
249 /*
250  * Default minimum read and write sizes.
251  */
252 #define XFS_READIO_LOG_LARGE	16
253 #define XFS_WRITEIO_LOG_LARGE	16
254 
255 /*
256  * Max and min values for mount-option defined I/O
257  * preallocation sizes.
258  */
259 #define XFS_MAX_IO_LOG		30	/* 1G */
260 #define XFS_MIN_IO_LOG		PAGE_SHIFT
261 
262 /*
263  * Synchronous read and write sizes.  This should be
264  * better for NFSv2 wsync filesystems.
265  */
266 #define	XFS_WSYNC_READIO_LOG	15	/* 32k */
267 #define	XFS_WSYNC_WRITEIO_LOG	14	/* 16k */
268 
269 /*
270  * Allow large block sizes to be reported to userspace programs if the
271  * "largeio" mount option is used.
272  *
273  * If compatibility mode is specified, simply return the basic unit of caching
274  * so that we don't get inefficient read/modify/write I/O from user apps.
275  * Otherwise....
276  *
277  * If the underlying volume is a stripe, then return the stripe width in bytes
278  * as the recommended I/O size. It is not a stripe and we've set a default
279  * buffered I/O size, return that, otherwise return the compat default.
280  */
281 static inline unsigned long
282 xfs_preferred_iosize(xfs_mount_t *mp)
283 {
284 	if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
285 		return PAGE_SIZE;
286 	return (mp->m_swidth ?
287 		(mp->m_swidth << mp->m_sb.sb_blocklog) :
288 		((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
289 			(1 << (int)MAX(mp->m_readio_log, mp->m_writeio_log)) :
290 			PAGE_SIZE));
291 }
292 
293 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp)	\
294 				((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
295 #define XFS_FORCED_SHUTDOWN(mp)	((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN)
296 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
297 		int lnnum);
298 #define xfs_force_shutdown(m,f)	\
299 	xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
300 
301 #define SHUTDOWN_META_IO_ERROR	0x0001	/* write attempt to metadata failed */
302 #define SHUTDOWN_LOG_IO_ERROR	0x0002	/* write attempt to the log failed */
303 #define SHUTDOWN_FORCE_UMOUNT	0x0004	/* shutdown from a forced unmount */
304 #define SHUTDOWN_CORRUPT_INCORE	0x0008	/* corrupt in-memory data structures */
305 #define SHUTDOWN_REMOTE_REQ	0x0010	/* shutdown came from remote cell */
306 #define SHUTDOWN_DEVICE_REQ	0x0020	/* failed all paths to the device */
307 
308 /*
309  * Flags for xfs_mountfs
310  */
311 #define XFS_MFSI_QUIET		0x40	/* Be silent if mount errors found */
312 
313 static inline xfs_agnumber_t
314 xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
315 {
316 	xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
317 	do_div(ld, mp->m_sb.sb_agblocks);
318 	return (xfs_agnumber_t) ld;
319 }
320 
321 static inline xfs_agblock_t
322 xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
323 {
324 	xfs_rfsblock_t ld = XFS_BB_TO_FSBT(mp, d);
325 	return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
326 }
327 
328 #ifdef DEBUG
329 static inline bool
330 xfs_mp_drop_writes(struct xfs_mount *mp)
331 {
332 	return mp->m_drop_writes;
333 }
334 #else
335 static inline bool
336 xfs_mp_drop_writes(struct xfs_mount *mp)
337 {
338 	return 0;
339 }
340 #endif
341 
342 /* per-AG block reservation data structures*/
343 enum xfs_ag_resv_type {
344 	XFS_AG_RESV_NONE = 0,
345 	XFS_AG_RESV_METADATA,
346 	XFS_AG_RESV_AGFL,
347 };
348 
349 struct xfs_ag_resv {
350 	/* number of blocks originally reserved here */
351 	xfs_extlen_t			ar_orig_reserved;
352 	/* number of blocks reserved here */
353 	xfs_extlen_t			ar_reserved;
354 	/* number of blocks originally asked for */
355 	xfs_extlen_t			ar_asked;
356 };
357 
358 /*
359  * Per-ag incore structure, copies of information in agf and agi, to improve the
360  * performance of allocation group selection.
361  */
362 typedef struct xfs_perag {
363 	struct xfs_mount *pag_mount;	/* owner filesystem */
364 	xfs_agnumber_t	pag_agno;	/* AG this structure belongs to */
365 	atomic_t	pag_ref;	/* perag reference count */
366 	char		pagf_init;	/* this agf's entry is initialized */
367 	char		pagi_init;	/* this agi's entry is initialized */
368 	char		pagf_metadata;	/* the agf is preferred to be metadata */
369 	char		pagi_inodeok;	/* The agi is ok for inodes */
370 	__uint8_t	pagf_levels[XFS_BTNUM_AGF];
371 					/* # of levels in bno & cnt btree */
372 	__uint32_t	pagf_flcount;	/* count of blocks in freelist */
373 	xfs_extlen_t	pagf_freeblks;	/* total free blocks */
374 	xfs_extlen_t	pagf_longest;	/* longest free space */
375 	__uint32_t	pagf_btreeblks;	/* # of blocks held in AGF btrees */
376 	xfs_agino_t	pagi_freecount;	/* number of free inodes */
377 	xfs_agino_t	pagi_count;	/* number of allocated inodes */
378 
379 	/*
380 	 * Inode allocation search lookup optimisation.
381 	 * If the pagino matches, the search for new inodes
382 	 * doesn't need to search the near ones again straight away
383 	 */
384 	xfs_agino_t	pagl_pagino;
385 	xfs_agino_t	pagl_leftrec;
386 	xfs_agino_t	pagl_rightrec;
387 	spinlock_t	pagb_lock;	/* lock for pagb_tree */
388 	struct rb_root	pagb_tree;	/* ordered tree of busy extents */
389 	unsigned int	pagb_gen;	/* generation count for pagb_tree */
390 	wait_queue_head_t pagb_wait;	/* woken when pagb_gen changes */
391 
392 	atomic_t        pagf_fstrms;    /* # of filestreams active in this AG */
393 
394 	spinlock_t	pag_ici_lock;	/* incore inode cache lock */
395 	struct radix_tree_root pag_ici_root;	/* incore inode cache root */
396 	int		pag_ici_reclaimable;	/* reclaimable inodes */
397 	struct mutex	pag_ici_reclaim_lock;	/* serialisation point */
398 	unsigned long	pag_ici_reclaim_cursor;	/* reclaim restart point */
399 
400 	/* buffer cache index */
401 	spinlock_t	pag_buf_lock;	/* lock for pag_buf_hash */
402 	struct rhashtable pag_buf_hash;
403 
404 	/* for rcu-safe freeing */
405 	struct rcu_head	rcu_head;
406 	int		pagb_count;	/* pagb slots in use */
407 
408 	/* Blocks reserved for all kinds of metadata. */
409 	struct xfs_ag_resv	pag_meta_resv;
410 	/* Blocks reserved for just AGFL-based metadata. */
411 	struct xfs_ag_resv	pag_agfl_resv;
412 
413 	/* reference count */
414 	__uint8_t		pagf_refcount_level;
415 } xfs_perag_t;
416 
417 static inline struct xfs_ag_resv *
418 xfs_perag_resv(
419 	struct xfs_perag	*pag,
420 	enum xfs_ag_resv_type	type)
421 {
422 	switch (type) {
423 	case XFS_AG_RESV_METADATA:
424 		return &pag->pag_meta_resv;
425 	case XFS_AG_RESV_AGFL:
426 		return &pag->pag_agfl_resv;
427 	default:
428 		return NULL;
429 	}
430 }
431 
432 int xfs_buf_hash_init(xfs_perag_t *pag);
433 void xfs_buf_hash_destroy(xfs_perag_t *pag);
434 
435 extern void	xfs_uuid_table_free(void);
436 extern int	xfs_log_sbcount(xfs_mount_t *);
437 extern __uint64_t xfs_default_resblks(xfs_mount_t *mp);
438 extern int	xfs_mountfs(xfs_mount_t *mp);
439 extern int	xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount,
440 				     xfs_agnumber_t *maxagi);
441 extern void	xfs_unmountfs(xfs_mount_t *);
442 
443 extern int	xfs_mod_icount(struct xfs_mount *mp, int64_t delta);
444 extern int	xfs_mod_ifree(struct xfs_mount *mp, int64_t delta);
445 extern int	xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
446 				 bool reserved);
447 extern int	xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
448 
449 extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
450 extern int	xfs_readsb(xfs_mount_t *, int);
451 extern void	xfs_freesb(xfs_mount_t *);
452 extern bool	xfs_fs_writable(struct xfs_mount *mp, int level);
453 extern int	xfs_sb_validate_fsb_count(struct xfs_sb *, __uint64_t);
454 
455 extern int	xfs_dev_is_read_only(struct xfs_mount *, char *);
456 
457 extern void	xfs_set_low_space_thresholds(struct xfs_mount *);
458 
459 int	xfs_zero_extent(struct xfs_inode *ip, xfs_fsblock_t start_fsb,
460 			xfs_off_t count_fsb);
461 
462 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp,
463 		int error_class, int error);
464 
465 #endif	/* __XFS_MOUNT_H__ */
466