forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2023-12-11 072de836f53be56a70cecf70b43ae43b7ce17376 
 kernel/fs/xfs/xfs_mount.h
....@@ -9,10 +9,8 @@
99 struct xlog;
1010 struct xfs_inode;
1111 struct xfs_mru_cache;
12
-struct xfs_nameops;
1312 struct xfs_ail;
1413 struct xfs_quotainfo;
15
-struct xfs_dir_ops;
1614 struct xfs_da_geometry;
1715 
1816 /* dynamic preallocation free space thresholds, 5% down to 1% */
....@@ -57,38 +55,25 @@
5755 	long		retry_timeout;	/* in jiffies, -1 = infinite */
5856 };
5957 
58
+/*
59
+ * The struct xfsmount layout is optimised to separate read-mostly variables
60
+ * from variables that are frequently modified. We put the read-mostly variables
61
+ * first, then place all the other variables at the end.
62
+ *
63
+ * Typically, read-mostly variables are those that are set at mount time and
64
+ * never changed again, or only change rarely as a result of things like sysfs
65
+ * knobs being tweaked.
66
+ */
6067 typedef struct xfs_mount {
61
-	struct super_block	*m_super;
62
-	xfs_tid_t		m_tid;		/* next unused tid for fs */
63
-	struct xfs_ail		*m_ail;		/* fs active log item list */
64
-
6568 	struct xfs_sb		m_sb;		/* copy of fs superblock */
66
-	spinlock_t		m_sb_lock;	/* sb counter lock */
67
-	struct percpu_counter	m_icount;	/* allocated inodes counter */
68
-	struct percpu_counter	m_ifree;	/* free inodes counter */
69
-	struct percpu_counter	m_fdblocks;	/* free block counter */
70
-
69
+	struct super_block	*m_super;
70
+	struct xfs_ail		*m_ail;		/* fs active log item list */
7171 	struct xfs_buf		*m_sb_bp;	/* buffer for superblock */
72
-	char			*m_fsname;	/* filesystem name */
73
-	int			m_fsname_len;	/* strlen of fs name */
7472 	char			*m_rtname;	/* realtime device name */
7573 	char			*m_logname;	/* external log device name */
76
-	int			m_bsize;	/* fs logical block size */
77
-	xfs_agnumber_t		m_agfrotor;	/* last ag where space found */
78
-	xfs_agnumber_t		m_agirotor;	/* last ag dir inode alloced */
79
-	spinlock_t		m_agirotor_lock;/* .. and lock protecting it */
80
-	xfs_agnumber_t		m_maxagi;	/* highest inode alloc group */
81
-	uint			m_readio_log;	/* min read size log bytes */
82
-	uint			m_readio_blocks; /* min read size blocks */
83
-	uint			m_writeio_log;	/* min write size log bytes */
84
-	uint			m_writeio_blocks; /* min write size blocks */
8574 	struct xfs_da_geometry	*m_dir_geo;	/* directory block geometry */
8675 	struct xfs_da_geometry	*m_attr_geo;	/* attribute block geometry */
8776 	struct xlog		*m_log;		/* log specific stuff */
88
-	int			m_logbufs;	/* number of log buffers */
89
-	int			m_logbsize;	/* size of each log buffer */
90
-	uint			m_rsumlevels;	/* rt summary levels */
91
-	uint			m_rsumsize;	/* size of rt summary, bytes */
9277 	struct xfs_inode	*m_rbmip;	/* pointer to bitmap inode */
9378 	struct xfs_inode	*m_rsumip;	/* pointer to summary inode */
9479 	struct xfs_inode	*m_rootip;	/* pointer to root directory */
....@@ -96,11 +81,26 @@
9681 	xfs_buftarg_t		*m_ddev_targp;	/* saves taking the address */
9782 	xfs_buftarg_t		*m_logdev_targp;/* ptr to log device */
9883 	xfs_buftarg_t		*m_rtdev_targp;	/* ptr to rt device */
84
+	/*
85
+	 * Optional cache of rt summary level per bitmap block with the
86
+	 * invariant that m_rsum_cache[bbno] <= the minimum i for which
87
+	 * rsum[i][bbno] != 0. Reads and writes are serialized by the rsumip
88
+	 * inode lock.
89
+	 */
90
+	uint8_t			*m_rsum_cache;
91
+	struct xfs_mru_cache	*m_filestream;  /* per-mount filestream data */
92
+	struct workqueue_struct *m_buf_workqueue;
93
+	struct workqueue_struct	*m_unwritten_workqueue;
94
+	struct workqueue_struct	*m_cil_workqueue;
95
+	struct workqueue_struct	*m_reclaim_workqueue;
96
+	struct workqueue_struct *m_eofblocks_workqueue;
97
+	struct workqueue_struct	*m_sync_workqueue;
98
+
99
+	int			m_bsize;	/* fs logical block size */
99100 	uint8_t			m_blkbit_log;	/* blocklog + NBBY */
100101 	uint8_t			m_blkbb_log;	/* blocklog - BBSHIFT */
101102 	uint8_t			m_agno_log;	/* log #ag's */
102
-	uint8_t			m_agino_log;	/* #bits for agino in inum */
103
-	uint			m_inode_cluster_size;/* min inode buf size */
103
+	uint8_t			m_sectbb_log;	/* sectlog - BBSHIFT */
104104 	uint			m_blockmask;	/* sb_blocksize-1 */
105105 	uint			m_blockwsize;	/* sb_blocksize in words */
106106 	uint			m_blockwmask;	/* blockwsize-1 */
....@@ -108,69 +108,93 @@
108108 	uint			m_alloc_mnr[2];	/* min alloc btree records */
109109 	uint			m_bmap_dmxr[2];	/* max bmap btree records */
110110 	uint			m_bmap_dmnr[2];	/* min bmap btree records */
111
-	uint			m_inobt_mxr[2];	/* max inobt btree records */
112
-	uint			m_inobt_mnr[2];	/* min inobt btree records */
113111 	uint			m_rmap_mxr[2];	/* max rmap btree records */
114112 	uint			m_rmap_mnr[2];	/* min rmap btree records */
115113 	uint			m_refc_mxr[2];	/* max refc btree records */
116114 	uint			m_refc_mnr[2];	/* min refc btree records */
117115 	uint			m_ag_maxlevels;	/* XFS_AG_MAXLEVELS */
118116 	uint			m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
119
-	uint			m_in_maxlevels;	/* max inobt btree levels. */
120117 	uint			m_rmap_maxlevels; /* max rmap btree levels */
121118 	uint			m_refc_maxlevels; /* max refcount btree level */
122119 	xfs_extlen_t		m_ag_prealloc_blocks; /* reserved ag blocks */
123120 	uint			m_alloc_set_aside; /* space we can't use */
124121 	uint			m_ag_max_usable; /* max space per AG */
122
+	int			m_dalign;	/* stripe unit */
123
+	int			m_swidth;	/* stripe width */
124
+	xfs_agnumber_t		m_maxagi;	/* highest inode alloc group */
125
+	uint			m_allocsize_log;/* min write size log bytes */
126
+	uint			m_allocsize_blocks; /* min write size blocks */
127
+	int			m_logbufs;	/* number of log buffers */
128
+	int			m_logbsize;	/* size of each log buffer */
129
+	uint			m_rsumlevels;	/* rt summary levels */
130
+	uint			m_rsumsize;	/* size of rt summary, bytes */
131
+	int			m_fixedfsid[2];	/* unchanged for life of FS */
132
+	uint			m_qflags;	/* quota status flags */
133
+	uint64_t		m_flags;	/* global mount flags */
134
+	int64_t			m_low_space[XFS_LOWSP_MAX];
135
+	struct xfs_ino_geometry	m_ino_geo;	/* inode geometry */
136
+	struct xfs_trans_resv	m_resv;		/* precomputed res values */
137
+						/* low free space thresholds */
138
+	bool			m_always_cow;
139
+	bool			m_fail_unmount;
140
+	bool			m_finobt_nores; /* no per-AG finobt resv. */
141
+	bool			m_update_sb;	/* sb needs update in mount */
142
+
143
+	/*
144
+	 * Bitsets of per-fs metadata that have been checked and/or are sick.
145
+	 * Callers must hold m_sb_lock to access these two fields.
146
+	 */
147
+	uint8_t			m_fs_checked;
148
+	uint8_t			m_fs_sick;
149
+	/*
150
+	 * Bitsets of rt metadata that have been checked and/or are sick.
151
+	 * Callers must hold m_sb_lock to access this field.
152
+	 */
153
+	uint8_t			m_rt_checked;
154
+	uint8_t			m_rt_sick;
155
+
156
+	/*
157
+	 * End of read-mostly variables. Frequently written variables and locks
158
+	 * should be placed below this comment from now on. The first variable
159
+	 * here is marked as cacheline aligned so they it is separated from
160
+	 * the read-mostly variables.
161
+	 */
162
+
163
+	spinlock_t ____cacheline_aligned m_sb_lock; /* sb counter lock */
164
+	struct percpu_counter	m_icount;	/* allocated inodes counter */
165
+	struct percpu_counter	m_ifree;	/* free inodes counter */
166
+	struct percpu_counter	m_fdblocks;	/* free block counter */
167
+	/*
168
+	 * Count of data device blocks reserved for delayed allocations,
169
+	 * including indlen blocks.  Does not include allocated CoW staging
170
+	 * extents or anything related to the rt device.
171
+	 */
172
+	struct percpu_counter	m_delalloc_blks;
173
+
125174 	struct radix_tree_root	m_perag_tree;	/* per-ag accounting info */
126175 	spinlock_t		m_perag_lock;	/* lock for m_perag_tree */
127
-	struct mutex		m_growlock;	/* growfs mutex */
128
-	int			m_fixedfsid[2];	/* unchanged for life of FS */
129
-	uint64_t		m_flags;	/* global mount flags */
130
-	bool			m_finobt_nores; /* no per-AG finobt resv. */
131
-	int			m_ialloc_inos;	/* inodes in inode allocation */
132
-	int			m_ialloc_blks;	/* blocks in inode allocation */
133
-	int			m_ialloc_min_blks;/* min blocks in sparse inode
134
-						   * allocation */
135
-	int			m_inoalign_mask;/* mask sb_inoalignmt if used */
136
-	uint			m_qflags;	/* quota status flags */
137
-	struct xfs_trans_resv	m_resv;		/* precomputed res values */
138
-	uint64_t		m_maxicount;	/* maximum inode count */
139176 	uint64_t		m_resblks;	/* total reserved blocks */
140177 	uint64_t		m_resblks_avail;/* available reserved blocks */
141178 	uint64_t		m_resblks_save;	/* reserved blks @ remount,ro */
142
-	int			m_dalign;	/* stripe unit */
143
-	int			m_swidth;	/* stripe width */
144
-	int			m_sinoalign;	/* stripe unit inode alignment */
145
-	uint8_t			m_sectbb_log;	/* sectlog - BBSHIFT */
146
-	const struct xfs_nameops *m_dirnameops;	/* vector of dir name ops */
147
-	const struct xfs_dir_ops *m_dir_inode_ops; /* vector of dir inode ops */
148
-	const struct xfs_dir_ops *m_nondir_inode_ops; /* !dir inode ops */
149
-	uint			m_chsize;	/* size of next field */
150
-	atomic_t		m_active_trans;	/* number trans frozen */
151
-	struct xfs_mru_cache	*m_filestream;  /* per-mount filestream data */
152179 	struct delayed_work	m_reclaim_work;	/* background inode reclaim */
153180 	struct delayed_work	m_eofblocks_work; /* background eof blocks
154181 						     trimming */
155182 	struct delayed_work	m_cowblocks_work; /* background cow blocks
156183 						     trimming */
157
-	bool			m_update_sb;	/* sb needs update in mount */
158
-	int64_t			m_low_space[XFS_LOWSP_MAX];
159
-						/* low free space thresholds */
160184 	struct xfs_kobj		m_kobj;
161185 	struct xfs_kobj		m_error_kobj;
162186 	struct xfs_kobj		m_error_meta_kobj;
163187 	struct xfs_error_cfg	m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX];
164188 	struct xstats		m_stats;	/* per-fs stats */
189
+	xfs_agnumber_t		m_agfrotor;	/* last ag where space found */
190
+	xfs_agnumber_t		m_agirotor;	/* last ag dir inode alloced */
191
+	spinlock_t		m_agirotor_lock;/* .. and lock protecting it */
165192 
166
-	struct workqueue_struct *m_buf_workqueue;
167
-	struct workqueue_struct	*m_data_workqueue;
168
-	struct workqueue_struct	*m_unwritten_workqueue;
169
-	struct workqueue_struct	*m_cil_workqueue;
170
-	struct workqueue_struct	*m_reclaim_workqueue;
171
-	struct workqueue_struct	*m_log_workqueue;
172
-	struct workqueue_struct *m_eofblocks_workqueue;
173
-	struct workqueue_struct	*m_sync_workqueue;
193
+	/*
194
+	 * Workqueue item so that we can coalesce multiple inode flush attempts
195
+	 * into a single flush.
196
+	 */
197
+	struct work_struct	m_flush_inodes_work;
174198 
175199 	/*
176200 	 * Generation of the filesysyem layout.  This is incremented by each
....@@ -182,8 +206,8 @@
182206 	 * to various other kinds of pain inflicted on the pNFS server.
183207 	 */
184208 	uint32_t		m_generation;
209
+	struct mutex		m_growlock;	/* growfs mutex */
185210 
186
-	bool			m_fail_unmount;
187211 #ifdef DEBUG
188212 	/*
189213 	 * Frequency with which errors are injected.  Replaces xfs_etest; the
....@@ -195,6 +219,8 @@
195219 #endif
196220 } xfs_mount_t;
197221 
222
+#define M_IGEO(mp)		(&(mp)->m_ino_geo)
223
+
198224 /*
199225  * Flags for m_flags.
200226  */
....@@ -202,7 +228,6 @@
202228 						   must be synchronous except
203229 						   for space allocations */
204230 #define XFS_MOUNT_UNMOUNTING	(1ULL << 1)	/* filesystem is unmounting */
205
-#define XFS_MOUNT_BAD_SUMMARY	(1ULL << 2)	/* summary counters are bad */
206231 #define XFS_MOUNT_WAS_CLEAN	(1ULL << 3)
207232 #define XFS_MOUNT_FS_SHUTDOWN	(1ULL << 4)	/* atomic stop of all filesystem
208233 						   operations, typically for
....@@ -213,7 +238,7 @@
213238 #define XFS_MOUNT_ATTR2		(1ULL << 8)	/* allow use of attr2 format */
214239 #define XFS_MOUNT_GRPID		(1ULL << 9)	/* group-ID assigned from directory */
215240 #define XFS_MOUNT_NORECOVERY	(1ULL << 10)	/* no recovery - dirty fs */
216
-#define XFS_MOUNT_DFLT_IOSIZE	(1ULL << 12)	/* set default i/o size */
241
+#define XFS_MOUNT_ALLOCSIZE	(1ULL << 12)	/* specified allocation size */
217242 #define XFS_MOUNT_SMALL_INUMS	(1ULL << 14)	/* user wants 32bit inodes */
218243 #define XFS_MOUNT_32BITINODES	(1ULL << 15)	/* inode32 allocator active */
219244 #define XFS_MOUNT_NOUUID	(1ULL << 16)	/* ignore uuid during mount */
....@@ -222,20 +247,13 @@
222247 						 * allocation */
223248 #define XFS_MOUNT_RDONLY	(1ULL << 20)	/* read-only fs */
224249 #define XFS_MOUNT_DIRSYNC	(1ULL << 21)	/* synchronous directory ops */
225
-#define XFS_MOUNT_COMPAT_IOSIZE	(1ULL << 22)	/* don't report large preferred
250
+#define XFS_MOUNT_LARGEIO	(1ULL << 22)	/* report large preferred
226251 						 * I/O size in stat() */
227252 #define XFS_MOUNT_FILESTREAMS	(1ULL << 24)	/* enable the filestreams
228253 						   allocator */
229254 #define XFS_MOUNT_NOATTR2	(1ULL << 25)	/* disable use of attr2 format */
230
-
231
-#define XFS_MOUNT_DAX		(1ULL << 62)	/* TEST ONLY! */
232
-
233
-
234
-/*
235
- * Default minimum read and write sizes.
236
- */
237
-#define XFS_READIO_LOG_LARGE	16
238
-#define XFS_WRITEIO_LOG_LARGE	16
255
+#define XFS_MOUNT_DAX_ALWAYS	(1ULL << 26)
256
+#define XFS_MOUNT_DAX_NEVER	(1ULL << 27)
239257 
240258 /*
241259  * Max and min values for mount-option defined I/O
....@@ -243,37 +261,6 @@
243261  */
244262 #define XFS_MAX_IO_LOG		30	/* 1G */
245263 #define XFS_MIN_IO_LOG		PAGE_SHIFT
246
-
247
-/*
248
- * Synchronous read and write sizes.  This should be
249
- * better for NFSv2 wsync filesystems.
250
- */
251
-#define	XFS_WSYNC_READIO_LOG	15	/* 32k */
252
-#define	XFS_WSYNC_WRITEIO_LOG	14	/* 16k */
253
-
254
-/*
255
- * Allow large block sizes to be reported to userspace programs if the
256
- * "largeio" mount option is used.
257
- *
258
- * If compatibility mode is specified, simply return the basic unit of caching
259
- * so that we don't get inefficient read/modify/write I/O from user apps.
260
- * Otherwise....
261
- *
262
- * If the underlying volume is a stripe, then return the stripe width in bytes
263
- * as the recommended I/O size. It is not a stripe and we've set a default
264
- * buffered I/O size, return that, otherwise return the compat default.
265
- */
266
-static inline unsigned long
267
-xfs_preferred_iosize(xfs_mount_t *mp)
268
-{
269
-	if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
270
-		return PAGE_SIZE;
271
-	return (mp->m_swidth ?
272
-		(mp->m_swidth << mp->m_sb.sb_blocklog) :
273
-		((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
274
-			(1 << (int)max(mp->m_readio_log, mp->m_writeio_log)) :
275
-			PAGE_SIZE));
276
-}
277264 
278265 #define XFS_LAST_UNMOUNT_WAS_CLEAN(mp)	\
279266 				((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
....@@ -287,8 +274,6 @@
287274 #define SHUTDOWN_LOG_IO_ERROR	0x0002	/* write attempt to the log failed */
288275 #define SHUTDOWN_FORCE_UMOUNT	0x0004	/* shutdown from a forced unmount */
289276 #define SHUTDOWN_CORRUPT_INCORE	0x0008	/* corrupt in-memory data structures */
290
-#define SHUTDOWN_REMOTE_REQ	0x0010	/* shutdown came from remote cell */
291
-#define SHUTDOWN_DEVICE_REQ	0x0020	/* failed all paths to the device */
292277 
293278 /*
294279  * Flags for xfs_mountfs
....@@ -311,13 +296,6 @@
311296 }
312297 
313298 /* per-AG block reservation data structures*/
314
-enum xfs_ag_resv_type {
315
-	XFS_AG_RESV_NONE = 0,
316
-	XFS_AG_RESV_AGFL,
317
-	XFS_AG_RESV_METADATA,
318
-	XFS_AG_RESV_RMAPBT,
319
-};
320
-
321299 struct xfs_ag_resv {
322300 	/* number of blocks originally reserved here */
323301 	xfs_extlen_t			ar_orig_reserved;
....@@ -357,6 +335,15 @@
357335 	xfs_agino_t	pagl_pagino;
358336 	xfs_agino_t	pagl_leftrec;
359337 	xfs_agino_t	pagl_rightrec;
338
+
339
+	/*
340
+	 * Bitsets of per-ag metadata that have been checked and/or are sick.
341
+	 * Callers should hold pag_state_lock before accessing this field.
342
+	 */
343
+	uint16_t	pag_checked;
344
+	uint16_t	pag_sick;
345
+	spinlock_t	pag_state_lock;
346
+
360347 	spinlock_t	pagb_lock;	/* lock for pagb_tree */
361348 	struct rb_root	pagb_tree;	/* ordered tree of busy extents */
362349 	unsigned int	pagb_gen;	/* generation count for pagb_tree */
....@@ -367,7 +354,6 @@
367354 	spinlock_t	pag_ici_lock;	/* incore inode cache lock */
368355 	struct radix_tree_root pag_ici_root;	/* incore inode cache root */
369356 	int		pag_ici_reclaimable;	/* reclaimable inodes */
370
-	struct mutex	pag_ici_reclaim_lock;	/* serialisation point */
371357 	unsigned long	pag_ici_reclaim_cursor;	/* reclaim restart point */
372358 
373359 	/* buffer cache index */
....@@ -385,6 +371,13 @@
385371 
386372 	/* reference count */
387373 	uint8_t			pagf_refcount_level;
374
+
375
+	/*
376
+	 * Unlinked inode information.  This incore information reflects
377
+	 * data stored in the AGI, so callers must hold the AGI buffer lock
378
+	 * or have some other means to control concurrency.
379
+	 */
380
+	struct rhashtable	pagi_unlinked_hash;
388381 } xfs_perag_t;
389382 
390383 static inline struct xfs_ag_resv *
....@@ -413,13 +406,18 @@
413406 				     xfs_agnumber_t *maxagi);
414407 extern void	xfs_unmountfs(xfs_mount_t *);
415408 
416
-extern int	xfs_mod_icount(struct xfs_mount *mp, int64_t delta);
417
-extern int	xfs_mod_ifree(struct xfs_mount *mp, int64_t delta);
409
+/* Accessor added for 5.10.y backport */
410
+static inline uint64_t
411
+xfs_fdblocks_unavailable(
412
+	struct xfs_mount	*mp)
413
+{
414
+	return mp->m_alloc_set_aside;
415
+}
416
+
418417 extern int	xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
419418 				 bool reserved);
420419 extern int	xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
421420 
422
-extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
423421 extern int	xfs_readsb(xfs_mount_t *, int);
424422 extern void	xfs_freesb(xfs_mount_t *);
425423 extern bool	xfs_fs_writable(struct xfs_mount *mp, int level);
....@@ -435,5 +433,6 @@
435433 struct xfs_error_cfg * xfs_error_get_cfg(struct xfs_mount *mp,
436434 		int error_class, int error);
437435 void xfs_force_summary_recalc(struct xfs_mount *mp);
436
+void xfs_mod_delalloc(struct xfs_mount *mp, int64_t delta);
438437 
439438 #endif	/* __XFS_MOUNT_H__ */