xref: /openbmc/linux/fs/ocfs2/ocfs2_fs.h (revision b4e18b29)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /* -*- mode: c; c-basic-offset: 8; -*-
3  * vim: noexpandtab sw=8 ts=8 sts=0:
4  *
5  * ocfs2_fs.h
6  *
7  * On-disk structures for OCFS2.
8  *
9  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
10  */
11 
12 #ifndef _OCFS2_FS_H
13 #define _OCFS2_FS_H
14 
15 #include <linux/magic.h>
16 
17 /* Version */
18 #define OCFS2_MAJOR_REV_LEVEL		0
19 #define OCFS2_MINOR_REV_LEVEL          	90
20 
21 /*
22  * An OCFS2 volume starts this way:
23  * Sector 0: Valid ocfs1_vol_disk_hdr that cleanly fails to mount OCFS.
24  * Sector 1: Valid ocfs1_vol_label that cleanly fails to mount OCFS.
25  * Block OCFS2_SUPER_BLOCK_BLKNO: OCFS2 superblock.
26  *
27  * All other structures are found from the superblock information.
28  *
29  * OCFS2_SUPER_BLOCK_BLKNO is in blocks, not sectors.  eg, for a
30  * blocksize of 2K, it is 4096 bytes into disk.
31  */
32 #define OCFS2_SUPER_BLOCK_BLKNO		2
33 
34 /*
35  * Cluster size limits. The maximum is kept arbitrarily at 1 MB, and could
36  * grow if needed.
37  */
38 #define OCFS2_MIN_CLUSTERSIZE		4096
39 #define OCFS2_MAX_CLUSTERSIZE		1048576
40 
41 /*
42  * Blocks cannot be bigger than clusters, so the maximum blocksize is the
43  * minimum cluster size.
44  */
45 #define OCFS2_MIN_BLOCKSIZE		512
46 #define OCFS2_MAX_BLOCKSIZE		OCFS2_MIN_CLUSTERSIZE
47 
48 /* Object signatures */
49 #define OCFS2_SUPER_BLOCK_SIGNATURE	"OCFSV2"
50 #define OCFS2_INODE_SIGNATURE		"INODE01"
51 #define OCFS2_EXTENT_BLOCK_SIGNATURE	"EXBLK01"
52 #define OCFS2_GROUP_DESC_SIGNATURE      "GROUP01"
53 #define OCFS2_XATTR_BLOCK_SIGNATURE	"XATTR01"
54 #define OCFS2_DIR_TRAILER_SIGNATURE	"DIRTRL1"
55 #define OCFS2_DX_ROOT_SIGNATURE		"DXDIR01"
56 #define OCFS2_DX_LEAF_SIGNATURE		"DXLEAF1"
57 #define OCFS2_REFCOUNT_BLOCK_SIGNATURE	"REFCNT1"
58 
59 /* Compatibility flags */
60 #define OCFS2_HAS_COMPAT_FEATURE(sb,mask)			\
61 	( OCFS2_SB(sb)->s_feature_compat & (mask) )
62 #define OCFS2_HAS_RO_COMPAT_FEATURE(sb,mask)			\
63 	( OCFS2_SB(sb)->s_feature_ro_compat & (mask) )
64 #define OCFS2_HAS_INCOMPAT_FEATURE(sb,mask)			\
65 	( OCFS2_SB(sb)->s_feature_incompat & (mask) )
66 #define OCFS2_SET_COMPAT_FEATURE(sb,mask)			\
67 	OCFS2_SB(sb)->s_feature_compat |= (mask)
68 #define OCFS2_SET_RO_COMPAT_FEATURE(sb,mask)			\
69 	OCFS2_SB(sb)->s_feature_ro_compat |= (mask)
70 #define OCFS2_SET_INCOMPAT_FEATURE(sb,mask)			\
71 	OCFS2_SB(sb)->s_feature_incompat |= (mask)
72 #define OCFS2_CLEAR_COMPAT_FEATURE(sb,mask)			\
73 	OCFS2_SB(sb)->s_feature_compat &= ~(mask)
74 #define OCFS2_CLEAR_RO_COMPAT_FEATURE(sb,mask)			\
75 	OCFS2_SB(sb)->s_feature_ro_compat &= ~(mask)
76 #define OCFS2_CLEAR_INCOMPAT_FEATURE(sb,mask)			\
77 	OCFS2_SB(sb)->s_feature_incompat &= ~(mask)
78 
79 #define OCFS2_FEATURE_COMPAT_SUPP	(OCFS2_FEATURE_COMPAT_BACKUP_SB	\
80 					 | OCFS2_FEATURE_COMPAT_JBD2_SB)
81 #define OCFS2_FEATURE_INCOMPAT_SUPP	(OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT \
82 					 | OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC \
83 					 | OCFS2_FEATURE_INCOMPAT_INLINE_DATA \
84 					 | OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP \
85 					 | OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK \
86 					 | OCFS2_FEATURE_INCOMPAT_XATTR \
87 					 | OCFS2_FEATURE_INCOMPAT_META_ECC \
88 					 | OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS \
89 					 | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE \
90 					 | OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG	\
91 					 | OCFS2_FEATURE_INCOMPAT_CLUSTERINFO \
92 					 | OCFS2_FEATURE_INCOMPAT_APPEND_DIO)
93 #define OCFS2_FEATURE_RO_COMPAT_SUPP	(OCFS2_FEATURE_RO_COMPAT_UNWRITTEN \
94 					 | OCFS2_FEATURE_RO_COMPAT_USRQUOTA \
95 					 | OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
96 
97 /*
98  * Heartbeat-only devices are missing journals and other files.  The
99  * filesystem driver can't load them, but the library can.  Never put
100  * this in OCFS2_FEATURE_INCOMPAT_SUPP, *ever*.
101  */
102 #define OCFS2_FEATURE_INCOMPAT_HEARTBEAT_DEV	0x0002
103 
104 /*
105  * tunefs sets this incompat flag before starting the resize and clears it
106  * at the end. This flag protects users from inadvertently mounting the fs
107  * after an aborted run without fsck-ing.
108  */
109 #define OCFS2_FEATURE_INCOMPAT_RESIZE_INPROG    0x0004
110 
111 /* Used to denote a non-clustered volume */
112 #define OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT	0x0008
113 
114 /* Support for sparse allocation in b-trees */
115 #define OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC	0x0010
116 
117 /*
118  * Tunefs sets this incompat flag before starting an operation which
119  * would require cleanup on abort. This is done to protect users from
120  * inadvertently mounting the fs after an aborted run without
121  * fsck-ing.
122  *
123  * s_tunefs_flags on the super block describes precisely which
124  * operations were in progress.
125  */
126 #define OCFS2_FEATURE_INCOMPAT_TUNEFS_INPROG	0x0020
127 
128 /* Support for data packed into inode blocks */
129 #define OCFS2_FEATURE_INCOMPAT_INLINE_DATA	0x0040
130 
131 /*
132  * Support for alternate, userspace cluster stacks.  If set, the superblock
133  * field s_cluster_info contains a tag for the alternate stack in use as
134  * well as the name of the cluster being joined.
135  * mount.ocfs2 must pass in a matching stack name.
136  *
137  * If not set, the classic stack will be used.  This is compatbile with
138  * all older versions.
139  */
140 #define OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK	0x0080
141 
142 /* Support for the extended slot map */
143 #define OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP 0x100
144 
145 /* Support for extended attributes */
146 #define OCFS2_FEATURE_INCOMPAT_XATTR		0x0200
147 
148 /* Support for indexed directores */
149 #define OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS	0x0400
150 
151 /* Metadata checksum and error correction */
152 #define OCFS2_FEATURE_INCOMPAT_META_ECC		0x0800
153 
154 /* Refcount tree support */
155 #define OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE	0x1000
156 
157 /* Discontiguous block groups */
158 #define OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG	0x2000
159 
160 /*
161  * Incompat bit to indicate useable clusterinfo with stackflags for all
162  * cluster stacks (userspace adnd o2cb). If this bit is set,
163  * INCOMPAT_USERSPACE_STACK becomes superfluous and thus should not be set.
164  */
165 #define OCFS2_FEATURE_INCOMPAT_CLUSTERINFO	0x4000
166 
167 /*
168  * Append Direct IO support
169  */
170 #define OCFS2_FEATURE_INCOMPAT_APPEND_DIO	0x8000
171 
172 /*
173  * backup superblock flag is used to indicate that this volume
174  * has backup superblocks.
175  */
176 #define OCFS2_FEATURE_COMPAT_BACKUP_SB		0x0001
177 
178 /*
179  * The filesystem will correctly handle journal feature bits.
180  */
181 #define OCFS2_FEATURE_COMPAT_JBD2_SB		0x0002
182 
183 /*
184  * Unwritten extents support.
185  */
186 #define OCFS2_FEATURE_RO_COMPAT_UNWRITTEN	0x0001
187 
188 /*
189  * Maintain quota information for this filesystem
190  */
191 #define OCFS2_FEATURE_RO_COMPAT_USRQUOTA	0x0002
192 #define OCFS2_FEATURE_RO_COMPAT_GRPQUOTA	0x0004
193 
194 
195 /* The byte offset of the first backup block will be 1G.
196  * The following will be 4G, 16G, 64G, 256G and 1T.
197  */
198 #define OCFS2_BACKUP_SB_START			1 << 30
199 
200 /* the max backup superblock nums */
201 #define OCFS2_MAX_BACKUP_SUPERBLOCKS	6
202 
203 /*
204  * Flags on ocfs2_super_block.s_tunefs_flags
205  */
206 #define OCFS2_TUNEFS_INPROG_REMOVE_SLOT		0x0001	/* Removing slots */
207 
208 /*
209  * Flags on ocfs2_dinode.i_flags
210  */
211 #define OCFS2_VALID_FL		(0x00000001)	/* Inode is valid */
212 #define OCFS2_UNUSED2_FL	(0x00000002)
213 #define OCFS2_ORPHANED_FL	(0x00000004)	/* On the orphan list */
214 #define OCFS2_UNUSED3_FL	(0x00000008)
215 /* System inode flags */
216 #define OCFS2_SYSTEM_FL		(0x00000010)	/* System inode */
217 #define OCFS2_SUPER_BLOCK_FL	(0x00000020)	/* Super block */
218 #define OCFS2_LOCAL_ALLOC_FL	(0x00000040)	/* Slot local alloc bitmap */
219 #define OCFS2_BITMAP_FL		(0x00000080)	/* Allocation bitmap */
220 #define OCFS2_JOURNAL_FL	(0x00000100)	/* Slot local journal */
221 #define OCFS2_HEARTBEAT_FL	(0x00000200)	/* Heartbeat area */
222 #define OCFS2_CHAIN_FL		(0x00000400)	/* Chain allocator */
223 #define OCFS2_DEALLOC_FL	(0x00000800)	/* Truncate log */
224 #define OCFS2_QUOTA_FL		(0x00001000)	/* Quota file */
225 #define OCFS2_DIO_ORPHANED_FL	(0X00002000)	/* On the orphan list especially
226 						 * for dio */
227 
228 /*
229  * Flags on ocfs2_dinode.i_dyn_features
230  *
231  * These can change much more often than i_flags. When adding flags,
232  * keep in mind that i_dyn_features is only 16 bits wide.
233  */
234 #define OCFS2_INLINE_DATA_FL	(0x0001)	/* Data stored in inode block */
235 #define OCFS2_HAS_XATTR_FL	(0x0002)
236 #define OCFS2_INLINE_XATTR_FL	(0x0004)
237 #define OCFS2_INDEXED_DIR_FL	(0x0008)
238 #define OCFS2_HAS_REFCOUNT_FL   (0x0010)
239 
240 /* Inode attributes, keep in sync with EXT2 */
241 #define OCFS2_SECRM_FL			FS_SECRM_FL	/* Secure deletion */
242 #define OCFS2_UNRM_FL			FS_UNRM_FL	/* Undelete */
243 #define OCFS2_COMPR_FL			FS_COMPR_FL	/* Compress file */
244 #define OCFS2_SYNC_FL			FS_SYNC_FL	/* Synchronous updates */
245 #define OCFS2_IMMUTABLE_FL		FS_IMMUTABLE_FL	/* Immutable file */
246 #define OCFS2_APPEND_FL			FS_APPEND_FL	/* writes to file may only append */
247 #define OCFS2_NODUMP_FL			FS_NODUMP_FL	/* do not dump file */
248 #define OCFS2_NOATIME_FL		FS_NOATIME_FL	/* do not update atime */
249 /* Reserved for compression usage... */
250 #define OCFS2_DIRTY_FL			FS_DIRTY_FL
251 #define OCFS2_COMPRBLK_FL		FS_COMPRBLK_FL	/* One or more compressed clusters */
252 #define OCFS2_NOCOMP_FL			FS_NOCOMP_FL	/* Don't compress */
253 #define OCFS2_ECOMPR_FL			FS_ECOMPR_FL	/* Compression error */
254 /* End compression flags --- maybe not all used */
255 #define OCFS2_BTREE_FL			FS_BTREE_FL	/* btree format dir */
256 #define OCFS2_INDEX_FL			FS_INDEX_FL	/* hash-indexed directory */
257 #define OCFS2_IMAGIC_FL			FS_IMAGIC_FL	/* AFS directory */
258 #define OCFS2_JOURNAL_DATA_FL		FS_JOURNAL_DATA_FL /* Reserved for ext3 */
259 #define OCFS2_NOTAIL_FL			FS_NOTAIL_FL	/* file tail should not be merged */
260 #define OCFS2_DIRSYNC_FL		FS_DIRSYNC_FL	/* dirsync behaviour (directories only) */
261 #define OCFS2_TOPDIR_FL			FS_TOPDIR_FL	/* Top of directory hierarchies*/
262 #define OCFS2_RESERVED_FL		FS_RESERVED_FL	/* reserved for ext2 lib */
263 
264 #define OCFS2_FL_VISIBLE		FS_FL_USER_VISIBLE	/* User visible flags */
265 #define OCFS2_FL_MODIFIABLE		FS_FL_USER_MODIFIABLE	/* User modifiable flags */
266 
267 /*
268  * Extent record flags (e_node.leaf.flags)
269  */
270 #define OCFS2_EXT_UNWRITTEN		(0x01)	/* Extent is allocated but
271 						 * unwritten */
272 #define OCFS2_EXT_REFCOUNTED		(0x02)  /* Extent is reference
273 						 * counted in an associated
274 						 * refcount tree */
275 
276 /*
277  * Journal Flags (ocfs2_dinode.id1.journal1.i_flags)
278  */
279 #define OCFS2_JOURNAL_DIRTY_FL	(0x00000001)	/* Journal needs recovery */
280 
281 /*
282  * superblock s_state flags
283  */
284 #define OCFS2_ERROR_FS		(0x00000001)	/* FS saw errors */
285 
286 /* Limit of space in ocfs2_dir_entry */
287 #define OCFS2_MAX_FILENAME_LEN		255
288 
289 /* Maximum slots on an ocfs2 file system */
290 #define OCFS2_MAX_SLOTS			255
291 
292 /* Slot map indicator for an empty slot */
293 #define OCFS2_INVALID_SLOT		((u16)-1)
294 
295 #define OCFS2_VOL_UUID_LEN		16
296 #define OCFS2_MAX_VOL_LABEL_LEN		64
297 
298 /* The cluster stack fields */
299 #define OCFS2_STACK_LABEL_LEN		4
300 #define OCFS2_CLUSTER_NAME_LEN		16
301 
302 /* Classic (historically speaking) cluster stack */
303 #define OCFS2_CLASSIC_CLUSTER_STACK	"o2cb"
304 
305 /* Journal limits (in bytes) */
306 #define OCFS2_MIN_JOURNAL_SIZE		(4 * 1024 * 1024)
307 
308 /*
309  * Inline extended attribute size (in bytes)
310  * The value chosen should be aligned to 16 byte boundaries.
311  */
312 #define OCFS2_MIN_XATTR_INLINE_SIZE     256
313 
314 /*
315  * Cluster info flags (ocfs2_cluster_info.ci_stackflags)
316  */
317 #define OCFS2_CLUSTER_O2CB_GLOBAL_HEARTBEAT	(0x01)
318 
319 struct ocfs2_system_inode_info {
320 	char	*si_name;
321 	int	si_iflags;
322 	int	si_mode;
323 };
324 
325 /* System file index */
326 enum {
327 	BAD_BLOCK_SYSTEM_INODE = 0,
328 	GLOBAL_INODE_ALLOC_SYSTEM_INODE,
329 #define OCFS2_FIRST_ONLINE_SYSTEM_INODE GLOBAL_INODE_ALLOC_SYSTEM_INODE
330 	SLOT_MAP_SYSTEM_INODE,
331 	HEARTBEAT_SYSTEM_INODE,
332 	GLOBAL_BITMAP_SYSTEM_INODE,
333 	USER_QUOTA_SYSTEM_INODE,
334 	GROUP_QUOTA_SYSTEM_INODE,
335 #define OCFS2_LAST_GLOBAL_SYSTEM_INODE GROUP_QUOTA_SYSTEM_INODE
336 #define OCFS2_FIRST_LOCAL_SYSTEM_INODE ORPHAN_DIR_SYSTEM_INODE
337 	ORPHAN_DIR_SYSTEM_INODE,
338 	EXTENT_ALLOC_SYSTEM_INODE,
339 	INODE_ALLOC_SYSTEM_INODE,
340 	JOURNAL_SYSTEM_INODE,
341 	LOCAL_ALLOC_SYSTEM_INODE,
342 	TRUNCATE_LOG_SYSTEM_INODE,
343 	LOCAL_USER_QUOTA_SYSTEM_INODE,
344 	LOCAL_GROUP_QUOTA_SYSTEM_INODE,
345 #define OCFS2_LAST_LOCAL_SYSTEM_INODE LOCAL_GROUP_QUOTA_SYSTEM_INODE
346 	NUM_SYSTEM_INODES
347 };
348 #define NUM_GLOBAL_SYSTEM_INODES OCFS2_FIRST_LOCAL_SYSTEM_INODE
349 #define NUM_LOCAL_SYSTEM_INODES	\
350 		(NUM_SYSTEM_INODES - OCFS2_FIRST_LOCAL_SYSTEM_INODE)
351 
352 static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = {
353 	/* Global system inodes (single copy) */
354 	/* The first two are only used from userspace mfks/tunefs */
355 	[BAD_BLOCK_SYSTEM_INODE]		= { "bad_blocks", 0, S_IFREG | 0644 },
356 	[GLOBAL_INODE_ALLOC_SYSTEM_INODE] 	= { "global_inode_alloc", OCFS2_BITMAP_FL | OCFS2_CHAIN_FL, S_IFREG | 0644 },
357 
358 	/* These are used by the running filesystem */
359 	[SLOT_MAP_SYSTEM_INODE]			= { "slot_map", 0, S_IFREG | 0644 },
360 	[HEARTBEAT_SYSTEM_INODE]		= { "heartbeat", OCFS2_HEARTBEAT_FL, S_IFREG | 0644 },
361 	[GLOBAL_BITMAP_SYSTEM_INODE]		= { "global_bitmap", 0, S_IFREG | 0644 },
362 	[USER_QUOTA_SYSTEM_INODE]		= { "aquota.user", OCFS2_QUOTA_FL, S_IFREG | 0644 },
363 	[GROUP_QUOTA_SYSTEM_INODE]		= { "aquota.group", OCFS2_QUOTA_FL, S_IFREG | 0644 },
364 
365 	/* Slot-specific system inodes (one copy per slot) */
366 	[ORPHAN_DIR_SYSTEM_INODE]		= { "orphan_dir:%04d", 0, S_IFDIR | 0755 },
367 	[EXTENT_ALLOC_SYSTEM_INODE]		= { "extent_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_CHAIN_FL, S_IFREG | 0644 },
368 	[INODE_ALLOC_SYSTEM_INODE]		= { "inode_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_CHAIN_FL, S_IFREG | 0644 },
369 	[JOURNAL_SYSTEM_INODE]			= { "journal:%04d", OCFS2_JOURNAL_FL, S_IFREG | 0644 },
370 	[LOCAL_ALLOC_SYSTEM_INODE]		= { "local_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_LOCAL_ALLOC_FL, S_IFREG | 0644 },
371 	[TRUNCATE_LOG_SYSTEM_INODE]		= { "truncate_log:%04d", OCFS2_DEALLOC_FL, S_IFREG | 0644 },
372 	[LOCAL_USER_QUOTA_SYSTEM_INODE]		= { "aquota.user:%04d", OCFS2_QUOTA_FL, S_IFREG | 0644 },
373 	[LOCAL_GROUP_QUOTA_SYSTEM_INODE]	= { "aquota.group:%04d", OCFS2_QUOTA_FL, S_IFREG | 0644 },
374 };
375 
376 /* Parameter passed from mount.ocfs2 to module */
377 #define OCFS2_HB_NONE			"heartbeat=none"
378 #define OCFS2_HB_LOCAL			"heartbeat=local"
379 #define OCFS2_HB_GLOBAL			"heartbeat=global"
380 
381 /*
382  * OCFS2_DIR_PAD defines the directory entries boundaries
383  *
384  * NOTE: It must be a multiple of 4
385  */
386 #define OCFS2_DIR_PAD			4
387 #define OCFS2_DIR_ROUND			(OCFS2_DIR_PAD - 1)
388 #define OCFS2_DIR_MEMBER_LEN 		offsetof(struct ocfs2_dir_entry, name)
389 #define OCFS2_DIR_REC_LEN(name_len)	(((name_len) + OCFS2_DIR_MEMBER_LEN + \
390                                           OCFS2_DIR_ROUND) & \
391 					 ~OCFS2_DIR_ROUND)
392 #define OCFS2_DIR_MIN_REC_LEN	OCFS2_DIR_REC_LEN(1)
393 
394 #define OCFS2_LINK_MAX		32000
395 #define	OCFS2_DX_LINK_MAX	((1U << 31) - 1U)
396 #define	OCFS2_LINKS_HI_SHIFT	16
397 #define	OCFS2_DX_ENTRIES_MAX	(0xffffffffU)
398 
399 
400 /*
401  * Convenience casts
402  */
403 #define OCFS2_RAW_SB(dinode)		(&((dinode)->id2.i_super))
404 
405 /*
406  * Block checking structure.  This is used in metadata to validate the
407  * contents.  If OCFS2_FEATURE_INCOMPAT_META_ECC is not set, it is all
408  * zeros.
409  */
410 struct ocfs2_block_check {
411 /*00*/	__le32 bc_crc32e;	/* 802.3 Ethernet II CRC32 */
412 	__le16 bc_ecc;		/* Single-error-correction parity vector.
413 				   This is a simple Hamming code dependent
414 				   on the blocksize.  OCFS2's maximum
415 				   blocksize, 4K, requires 16 parity bits,
416 				   so we fit in __le16. */
417 	__le16 bc_reserved1;
418 /*08*/
419 };
420 
421 /*
422  * On disk extent record for OCFS2
423  * It describes a range of clusters on disk.
424  *
425  * Length fields are divided into interior and leaf node versions.
426  * This leaves room for a flags field (OCFS2_EXT_*) in the leaf nodes.
427  */
428 struct ocfs2_extent_rec {
429 /*00*/	__le32 e_cpos;		/* Offset into the file, in clusters */
430 	union {
431 		__le32 e_int_clusters; /* Clusters covered by all children */
432 		struct {
433 			__le16 e_leaf_clusters; /* Clusters covered by this
434 						   extent */
435 			__u8 e_reserved1;
436 			__u8 e_flags; /* Extent flags */
437 		};
438 	};
439 	__le64 e_blkno;		/* Physical disk offset, in blocks */
440 /*10*/
441 };
442 
443 struct ocfs2_chain_rec {
444 	__le32 c_free;	/* Number of free bits in this chain. */
445 	__le32 c_total;	/* Number of total bits in this chain */
446 	__le64 c_blkno;	/* Physical disk offset (blocks) of 1st group */
447 };
448 
449 struct ocfs2_truncate_rec {
450 	__le32 t_start;		/* 1st cluster in this log */
451 	__le32 t_clusters;	/* Number of total clusters covered */
452 };
453 
454 /*
455  * On disk extent list for OCFS2 (node in the tree).  Note that this
456  * is contained inside ocfs2_dinode or ocfs2_extent_block, so the
457  * offsets are relative to ocfs2_dinode.id2.i_list or
458  * ocfs2_extent_block.h_list, respectively.
459  */
460 struct ocfs2_extent_list {
461 /*00*/	__le16 l_tree_depth;		/* Extent tree depth from this
462 					   point.  0 means data extents
463 					   hang directly off this
464 					   header (a leaf)
465 					   NOTE: The high 8 bits cannot be
466 					   used - tree_depth is never that big.
467 					*/
468 	__le16 l_count;			/* Number of extent records */
469 	__le16 l_next_free_rec;		/* Next unused extent slot */
470 	__le16 l_reserved1;
471 	__le64 l_reserved2;		/* Pad to
472 					   sizeof(ocfs2_extent_rec) */
473 /*10*/	struct ocfs2_extent_rec l_recs[];	/* Extent records */
474 };
475 
476 /*
477  * On disk allocation chain list for OCFS2.  Note that this is
478  * contained inside ocfs2_dinode, so the offsets are relative to
479  * ocfs2_dinode.id2.i_chain.
480  */
481 struct ocfs2_chain_list {
482 /*00*/	__le16 cl_cpg;			/* Clusters per Block Group */
483 	__le16 cl_bpc;			/* Bits per cluster */
484 	__le16 cl_count;		/* Total chains in this list */
485 	__le16 cl_next_free_rec;	/* Next unused chain slot */
486 	__le64 cl_reserved1;
487 /*10*/	struct ocfs2_chain_rec cl_recs[];	/* Chain records */
488 };
489 
490 /*
491  * On disk deallocation log for OCFS2.  Note that this is
492  * contained inside ocfs2_dinode, so the offsets are relative to
493  * ocfs2_dinode.id2.i_dealloc.
494  */
495 struct ocfs2_truncate_log {
496 /*00*/	__le16 tl_count;		/* Total records in this log */
497 	__le16 tl_used;			/* Number of records in use */
498 	__le32 tl_reserved1;
499 /*08*/	struct ocfs2_truncate_rec tl_recs[];	/* Truncate records */
500 };
501 
502 /*
503  * On disk extent block (indirect block) for OCFS2
504  */
505 struct ocfs2_extent_block
506 {
507 /*00*/	__u8 h_signature[8];		/* Signature for verification */
508 	struct ocfs2_block_check h_check;	/* Error checking */
509 /*10*/	__le16 h_suballoc_slot;		/* Slot suballocator this
510 					   extent_header belongs to */
511 	__le16 h_suballoc_bit;		/* Bit offset in suballocator
512 					   block group */
513 	__le32 h_fs_generation;		/* Must match super block */
514 	__le64 h_blkno;			/* Offset on disk, in blocks */
515 /*20*/	__le64 h_suballoc_loc;		/* Suballocator block group this
516 					   eb belongs to.  Only valid
517 					   if allocated from a
518 					   discontiguous block group */
519 	__le64 h_next_leaf_blk;		/* Offset on disk, in blocks,
520 					   of next leaf header pointing
521 					   to data */
522 /*30*/	struct ocfs2_extent_list h_list;	/* Extent record list */
523 /* Actual on-disk size is one block */
524 };
525 
526 /*
527  * On disk slot map for OCFS2.  This defines the contents of the "slot_map"
528  * system file.  A slot is valid if it contains a node number >= 0.  The
529  * value -1 (0xFFFF) is OCFS2_INVALID_SLOT.  This marks a slot empty.
530  */
531 struct ocfs2_slot_map {
532 /*00*/	__le16 sm_slots[0];
533 /*
534  * Actual on-disk size is one block.  OCFS2_MAX_SLOTS is 255,
535  * 255 * sizeof(__le16) == 512B, within the 512B block minimum blocksize.
536  */
537 };
538 
539 struct ocfs2_extended_slot {
540 /*00*/	__u8	es_valid;
541 	__u8	es_reserved1[3];
542 	__le32	es_node_num;
543 /*08*/
544 };
545 
546 /*
547  * The extended slot map, used when OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP
548  * is set.  It separates out the valid marker from the node number, and
549  * has room to grow.  Unlike the old slot map, this format is defined by
550  * i_size.
551  */
552 struct ocfs2_slot_map_extended {
553 /*00*/	struct ocfs2_extended_slot se_slots[0];
554 /*
555  * Actual size is i_size of the slot_map system file.  It should
556  * match s_max_slots * sizeof(struct ocfs2_extended_slot)
557  */
558 };
559 
560 /*
561  * ci_stackflags is only valid if the incompat bit
562  * OCFS2_FEATURE_INCOMPAT_CLUSTERINFO is set.
563  */
564 struct ocfs2_cluster_info {
565 /*00*/	__u8   ci_stack[OCFS2_STACK_LABEL_LEN];
566 	union {
567 		__le32 ci_reserved;
568 		struct {
569 			__u8 ci_stackflags;
570 			__u8 ci_reserved1;
571 			__u8 ci_reserved2;
572 			__u8 ci_reserved3;
573 		};
574 	};
575 /*08*/	__u8   ci_cluster[OCFS2_CLUSTER_NAME_LEN];
576 /*18*/
577 };
578 
579 /*
580  * On disk superblock for OCFS2
581  * Note that it is contained inside an ocfs2_dinode, so all offsets
582  * are relative to the start of ocfs2_dinode.id2.
583  */
584 struct ocfs2_super_block {
585 /*00*/	__le16 s_major_rev_level;
586 	__le16 s_minor_rev_level;
587 	__le16 s_mnt_count;
588 	__le16 s_max_mnt_count;
589 	__le16 s_state;			/* File system state */
590 	__le16 s_errors;			/* Behaviour when detecting errors */
591 	__le32 s_checkinterval;		/* Max time between checks */
592 /*10*/	__le64 s_lastcheck;		/* Time of last check */
593 	__le32 s_creator_os;		/* OS */
594 	__le32 s_feature_compat;		/* Compatible feature set */
595 /*20*/	__le32 s_feature_incompat;	/* Incompatible feature set */
596 	__le32 s_feature_ro_compat;	/* Readonly-compatible feature set */
597 	__le64 s_root_blkno;		/* Offset, in blocks, of root directory
598 					   dinode */
599 /*30*/	__le64 s_system_dir_blkno;	/* Offset, in blocks, of system
600 					   directory dinode */
601 	__le32 s_blocksize_bits;		/* Blocksize for this fs */
602 	__le32 s_clustersize_bits;	/* Clustersize for this fs */
603 /*40*/	__le16 s_max_slots;		/* Max number of simultaneous mounts
604 					   before tunefs required */
605 	__le16 s_tunefs_flag;
606 	__le32 s_uuid_hash;		/* hash value of uuid */
607 	__le64 s_first_cluster_group;	/* Block offset of 1st cluster
608 					 * group header */
609 /*50*/	__u8  s_label[OCFS2_MAX_VOL_LABEL_LEN];	/* Label for mounting, etc. */
610 /*90*/	__u8  s_uuid[OCFS2_VOL_UUID_LEN];	/* 128-bit uuid */
611 /*A0*/  struct ocfs2_cluster_info s_cluster_info; /* Only valid if either
612 						     userspace or clusterinfo
613 						     INCOMPAT flag set. */
614 /*B8*/	__le16 s_xattr_inline_size;	/* extended attribute inline size
615 					   for this fs*/
616 	__le16 s_reserved0;
617 	__le32 s_dx_seed[3];		/* seed[0-2] for dx dir hash.
618 					 * s_uuid_hash serves as seed[3]. */
619 /*C0*/  __le64 s_reserved2[15];		/* Fill out superblock */
620 /*140*/
621 
622 	/*
623 	 * NOTE: As stated above, all offsets are relative to
624 	 * ocfs2_dinode.id2, which is at 0xC0 in the inode.
625 	 * 0xC0 + 0x140 = 0x200 or 512 bytes.  A superblock must fit within
626 	 * our smallest blocksize, which is 512 bytes.  To ensure this,
627 	 * we reserve the space in s_reserved2.  Anything past s_reserved2
628 	 * will not be available on the smallest blocksize.
629 	 */
630 };
631 
632 /*
633  * Local allocation bitmap for OCFS2 slots
634  * Note that it exists inside an ocfs2_dinode, so all offsets are
635  * relative to the start of ocfs2_dinode.id2.
636  */
637 struct ocfs2_local_alloc
638 {
639 /*00*/	__le32 la_bm_off;	/* Starting bit offset in main bitmap */
640 	__le16 la_size;		/* Size of included bitmap, in bytes */
641 	__le16 la_reserved1;
642 	__le64 la_reserved2;
643 /*10*/	__u8   la_bitmap[];
644 };
645 
646 /*
647  * Data-in-inode header. This is only used if i_dyn_features has
648  * OCFS2_INLINE_DATA_FL set.
649  */
650 struct ocfs2_inline_data
651 {
652 /*00*/	__le16	id_count;	/* Number of bytes that can be used
653 				 * for data, starting at id_data */
654 	__le16	id_reserved0;
655 	__le32	id_reserved1;
656 	__u8	id_data[];	/* Start of user data */
657 };
658 
659 /*
660  * On disk inode for OCFS2
661  */
662 struct ocfs2_dinode {
663 /*00*/	__u8 i_signature[8];		/* Signature for validation */
664 	__le32 i_generation;		/* Generation number */
665 	__le16 i_suballoc_slot;		/* Slot suballocator this inode
666 					   belongs to */
667 	__le16 i_suballoc_bit;		/* Bit offset in suballocator
668 					   block group */
669 /*10*/	__le16 i_links_count_hi;	/* High 16 bits of links count */
670 	__le16 i_xattr_inline_size;
671 	__le32 i_clusters;		/* Cluster count */
672 	__le32 i_uid;			/* Owner UID */
673 	__le32 i_gid;			/* Owning GID */
674 /*20*/	__le64 i_size;			/* Size in bytes */
675 	__le16 i_mode;			/* File mode */
676 	__le16 i_links_count;		/* Links count */
677 	__le32 i_flags;			/* File flags */
678 /*30*/	__le64 i_atime;			/* Access time */
679 	__le64 i_ctime;			/* Creation time */
680 /*40*/	__le64 i_mtime;			/* Modification time */
681 	__le64 i_dtime;			/* Deletion time */
682 /*50*/	__le64 i_blkno;			/* Offset on disk, in blocks */
683 	__le64 i_last_eb_blk;		/* Pointer to last extent
684 					   block */
685 /*60*/	__le32 i_fs_generation;		/* Generation per fs-instance */
686 	__le32 i_atime_nsec;
687 	__le32 i_ctime_nsec;
688 	__le32 i_mtime_nsec;
689 /*70*/	__le32 i_attr;
690 	__le16 i_orphaned_slot;		/* Only valid when OCFS2_ORPHANED_FL
691 					   was set in i_flags */
692 	__le16 i_dyn_features;
693 	__le64 i_xattr_loc;
694 /*80*/	struct ocfs2_block_check i_check;	/* Error checking */
695 /*88*/	__le64 i_dx_root;		/* Pointer to dir index root block */
696 /*90*/	__le64 i_refcount_loc;
697 	__le64 i_suballoc_loc;		/* Suballocator block group this
698 					   inode belongs to.  Only valid
699 					   if allocated from a
700 					   discontiguous block group */
701 /*A0*/	__le16 i_dio_orphaned_slot;	/* only used for append dio write */
702 	__le16 i_reserved1[3];
703 	__le64 i_reserved2[2];
704 /*B8*/	union {
705 		__le64 i_pad1;		/* Generic way to refer to this
706 					   64bit union */
707 		struct {
708 			__le64 i_rdev;	/* Device number */
709 		} dev1;
710 		struct {		/* Info for bitmap system
711 					   inodes */
712 			__le32 i_used;	/* Bits (ie, clusters) used  */
713 			__le32 i_total;	/* Total bits (clusters)
714 					   available */
715 		} bitmap1;
716 		struct {		/* Info for journal system
717 					   inodes */
718 			__le32 ij_flags;	/* Mounted, version, etc. */
719 			__le32 ij_recovery_generation; /* Incremented when the
720 							  journal is recovered
721 							  after an unclean
722 							  shutdown */
723 		} journal1;
724 	} id1;				/* Inode type dependent 1 */
725 /*C0*/	union {
726 		struct ocfs2_super_block	i_super;
727 		struct ocfs2_local_alloc	i_lab;
728 		struct ocfs2_chain_list		i_chain;
729 		struct ocfs2_extent_list	i_list;
730 		struct ocfs2_truncate_log	i_dealloc;
731 		struct ocfs2_inline_data	i_data;
732 		__u8               		i_symlink[0];
733 	} id2;
734 /* Actual on-disk size is one block */
735 };
736 
737 /*
738  * On-disk directory entry structure for OCFS2
739  *
740  * Packed as this structure could be accessed unaligned on 64-bit platforms
741  */
742 struct ocfs2_dir_entry {
743 /*00*/	__le64   inode;                  /* Inode number */
744 	__le16   rec_len;                /* Directory entry length */
745 	__u8    name_len;               /* Name length */
746 	__u8    file_type;
747 /*0C*/	char    name[OCFS2_MAX_FILENAME_LEN];   /* File name */
748 /* Actual on-disk length specified by rec_len */
749 } __attribute__ ((packed));
750 
751 /*
752  * Per-block record for the unindexed directory btree. This is carefully
753  * crafted so that the rec_len and name_len records of an ocfs2_dir_entry are
754  * mirrored. That way, the directory manipulation code needs a minimal amount
755  * of update.
756  *
757  * NOTE: Keep this structure aligned to a multiple of 4 bytes.
758  */
759 struct ocfs2_dir_block_trailer {
760 /*00*/	__le64		db_compat_inode;	/* Always zero. Was inode */
761 
762 	__le16		db_compat_rec_len;	/* Backwards compatible with
763 						 * ocfs2_dir_entry. */
764 	__u8		db_compat_name_len;	/* Always zero. Was name_len */
765 	__u8		db_reserved0;
766 	__le16		db_reserved1;
767 	__le16		db_free_rec_len;	/* Size of largest empty hole
768 						 * in this block. (unused) */
769 /*10*/	__u8		db_signature[8];	/* Signature for verification */
770 	__le64		db_reserved2;
771 /*20*/	__le64		db_free_next;		/* Next block in list (unused) */
772 	__le64		db_blkno;		/* Offset on disk, in blocks */
773 /*30*/	__le64		db_parent_dinode;	/* dinode which owns me, in
774 						   blocks */
775 	struct ocfs2_block_check db_check;	/* Error checking */
776 /*40*/
777 };
778 
779  /*
780  * A directory entry in the indexed tree. We don't store the full name here,
781  * but instead provide a pointer to the full dirent in the unindexed tree.
782  *
783  * We also store name_len here so as to reduce the number of leaf blocks we
784  * need to search in case of collisions.
785  */
786 struct ocfs2_dx_entry {
787 	__le32		dx_major_hash;	/* Used to find logical
788 					 * cluster in index */
789 	__le32		dx_minor_hash;	/* Lower bits used to find
790 					 * block in cluster */
791 	__le64		dx_dirent_blk;	/* Physical block in unindexed
792 					 * tree holding this dirent. */
793 };
794 
795 struct ocfs2_dx_entry_list {
796 	__le32		de_reserved;
797 	__le16		de_count;	/* Maximum number of entries
798 					 * possible in de_entries */
799 	__le16		de_num_used;	/* Current number of
800 					 * de_entries entries */
801 	struct	ocfs2_dx_entry		de_entries[];	/* Indexed dir entries
802 							 * in a packed array of
803 							 * length de_num_used */
804 };
805 
806 #define OCFS2_DX_FLAG_INLINE	0x01
807 
808 /*
809  * A directory indexing block. Each indexed directory has one of these,
810  * pointed to by ocfs2_dinode.
811  *
812  * This block stores an indexed btree root, and a set of free space
813  * start-of-list pointers.
814  */
815 struct ocfs2_dx_root_block {
816 	__u8		dr_signature[8];	/* Signature for verification */
817 	struct ocfs2_block_check dr_check;	/* Error checking */
818 	__le16		dr_suballoc_slot;	/* Slot suballocator this
819 						 * block belongs to. */
820 	__le16		dr_suballoc_bit;	/* Bit offset in suballocator
821 						 * block group */
822 	__le32		dr_fs_generation;	/* Must match super block */
823 	__le64		dr_blkno;		/* Offset on disk, in blocks */
824 	__le64		dr_last_eb_blk;		/* Pointer to last
825 						 * extent block */
826 	__le32		dr_clusters;		/* Clusters allocated
827 						 * to the indexed tree. */
828 	__u8		dr_flags;		/* OCFS2_DX_FLAG_* flags */
829 	__u8		dr_reserved0;
830 	__le16		dr_reserved1;
831 	__le64		dr_dir_blkno;		/* Pointer to parent inode */
832 	__le32		dr_num_entries;		/* Total number of
833 						 * names stored in
834 						 * this directory.*/
835 	__le32		dr_reserved2;
836 	__le64		dr_free_blk;		/* Pointer to head of free
837 						 * unindexed block list. */
838 	__le64		dr_suballoc_loc;	/* Suballocator block group
839 						   this root belongs to.
840 						   Only valid if allocated
841 						   from a discontiguous
842 						   block group */
843 	__le64		dr_reserved3[14];
844 	union {
845 		struct ocfs2_extent_list dr_list; /* Keep this aligned to 128
846 						   * bits for maximum space
847 						   * efficiency. */
848 		struct ocfs2_dx_entry_list dr_entries; /* In-root-block list of
849 							* entries. We grow out
850 							* to extents if this
851 							* gets too big. */
852 	};
853 };
854 
855 /*
856  * The header of a leaf block in the indexed tree.
857  */
858 struct ocfs2_dx_leaf {
859 	__u8		dl_signature[8];/* Signature for verification */
860 	struct ocfs2_block_check dl_check;	/* Error checking */
861 	__le64		dl_blkno;	/* Offset on disk, in blocks */
862 	__le32		dl_fs_generation;/* Must match super block */
863 	__le32		dl_reserved0;
864 	__le64		dl_reserved1;
865 	struct ocfs2_dx_entry_list	dl_list;
866 };
867 
868 /*
869  * Largest bitmap for a block (suballocator) group in bytes.  This limit
870  * does not affect cluster groups (global allocator).  Cluster group
871  * bitmaps run to the end of the block.
872  */
873 #define OCFS2_MAX_BG_BITMAP_SIZE	256
874 
875 /*
876  * On disk allocator group structure for OCFS2
877  */
878 struct ocfs2_group_desc
879 {
880 /*00*/	__u8    bg_signature[8];        /* Signature for validation */
881 	__le16   bg_size;                /* Size of included bitmap in
882 					   bytes. */
883 	__le16   bg_bits;                /* Bits represented by this
884 					   group. */
885 	__le16	bg_free_bits_count;     /* Free bits count */
886 	__le16   bg_chain;               /* What chain I am in. */
887 /*10*/	__le32   bg_generation;
888 	__le32	bg_reserved1;
889 	__le64   bg_next_group;          /* Next group in my list, in
890 					   blocks */
891 /*20*/	__le64   bg_parent_dinode;       /* dinode which owns me, in
892 					   blocks */
893 	__le64   bg_blkno;               /* Offset on disk, in blocks */
894 /*30*/	struct ocfs2_block_check bg_check;	/* Error checking */
895 	__le64   bg_reserved2;
896 /*40*/	union {
897 		__u8    bg_bitmap[0];
898 		struct {
899 			/*
900 			 * Block groups may be discontiguous when
901 			 * OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG is set.
902 			 * The extents of a discontiguous block group are
903 			 * stored in bg_list.  It is a flat list.
904 			 * l_tree_depth must always be zero.  A
905 			 * discontiguous group is signified by a non-zero
906 			 * bg_list->l_next_free_rec.  Only block groups
907 			 * can be discontiguous; Cluster groups cannot.
908 			 * We've never made a block group with more than
909 			 * 2048 blocks (256 bytes of bg_bitmap).  This
910 			 * codifies that limit so that we can fit bg_list.
911 			 * bg_size of a discontiguous block group will
912 			 * be 256 to match bg_bitmap_filler.
913 			 */
914 			__u8 bg_bitmap_filler[OCFS2_MAX_BG_BITMAP_SIZE];
915 /*140*/			struct ocfs2_extent_list bg_list;
916 		};
917 	};
918 /* Actual on-disk size is one block */
919 };
920 
921 struct ocfs2_refcount_rec {
922 /*00*/	__le64 r_cpos;		/* Physical offset, in clusters */
923 	__le32 r_clusters;	/* Clusters covered by this extent */
924 	__le32 r_refcount;	/* Reference count of this extent */
925 /*10*/
926 };
927 #define OCFS2_32BIT_POS_MASK		(0xffffffffULL)
928 
929 #define OCFS2_REFCOUNT_LEAF_FL          (0x00000001)
930 #define OCFS2_REFCOUNT_TREE_FL          (0x00000002)
931 
932 struct ocfs2_refcount_list {
933 /*00*/	__le16 rl_count;	/* Maximum number of entries possible
934 				   in rl_records */
935 	__le16 rl_used;		/* Current number of used records */
936 	__le32 rl_reserved2;
937 	__le64 rl_reserved1;	/* Pad to sizeof(ocfs2_refcount_record) */
938 /*10*/	struct ocfs2_refcount_rec rl_recs[];	/* Refcount records */
939 };
940 
941 
942 struct ocfs2_refcount_block {
943 /*00*/	__u8 rf_signature[8];		/* Signature for verification */
944 	__le16 rf_suballoc_slot;	/* Slot suballocator this block
945 					   belongs to */
946 	__le16 rf_suballoc_bit;		/* Bit offset in suballocator
947 					   block group */
948 	__le32 rf_fs_generation;	/* Must match superblock */
949 /*10*/	__le64 rf_blkno;		/* Offset on disk, in blocks */
950 	__le64 rf_parent;		/* Parent block, only valid if
951 					   OCFS2_REFCOUNT_LEAF_FL is set in
952 					   rf_flags */
953 /*20*/	struct ocfs2_block_check rf_check;	/* Error checking */
954 	__le64 rf_last_eb_blk;		/* Pointer to last extent block */
955 /*30*/	__le32 rf_count;		/* Number of inodes sharing this
956 					   refcount tree */
957 	__le32 rf_flags;		/* See the flags above */
958 	__le32 rf_clusters;		/* clusters covered by refcount tree. */
959 	__le32 rf_cpos;			/* cluster offset in refcount tree.*/
960 /*40*/	__le32 rf_generation;		/* generation number. all be the same
961 					 * for the same refcount tree. */
962 	__le32 rf_reserved0;
963 	__le64 rf_suballoc_loc;		/* Suballocator block group this
964 					   refcount block belongs to. Only
965 					   valid if allocated from a
966 					   discontiguous block group */
967 /*50*/	__le64 rf_reserved1[6];
968 /*80*/	union {
969 		struct ocfs2_refcount_list rf_records;  /* List of refcount
970 							  records */
971 		struct ocfs2_extent_list rf_list;	/* Extent record list,
972 							only valid if
973 							OCFS2_REFCOUNT_TREE_FL
974 							is set in rf_flags */
975 	};
976 /* Actual on-disk size is one block */
977 };
978 
979 /*
980  * On disk extended attribute structure for OCFS2.
981  */
982 
983 /*
984  * ocfs2_xattr_entry indicates one extend attribute.
985  *
986  * Note that it can be stored in inode, one block or one xattr bucket.
987  */
988 struct ocfs2_xattr_entry {
989 	__le32	xe_name_hash;    /* hash value of xattr prefix+suffix. */
990 	__le16	xe_name_offset;  /* byte offset from the 1st entry in the
991 				    local xattr storage(inode, xattr block or
992 				    xattr bucket). */
993 	__u8	xe_name_len;	 /* xattr name len, doesn't include prefix. */
994 	__u8	xe_type;         /* the low 7 bits indicate the name prefix
995 				  * type and the highest bit indicates whether
996 				  * the EA is stored in the local storage. */
997 	__le64	xe_value_size;	 /* real xattr value length. */
998 };
999 
1000 /*
1001  * On disk structure for xattr header.
1002  *
1003  * One ocfs2_xattr_header describes how many ocfs2_xattr_entry records in
1004  * the local xattr storage.
1005  */
1006 struct ocfs2_xattr_header {
1007 	__le16	xh_count;                       /* contains the count of how
1008 						   many records are in the
1009 						   local xattr storage. */
1010 	__le16	xh_free_start;                  /* current offset for storing
1011 						   xattr. */
1012 	__le16	xh_name_value_len;              /* total length of name/value
1013 						   length in this bucket. */
1014 	__le16	xh_num_buckets;                 /* Number of xattr buckets
1015 						   in this extent record,
1016 						   only valid in the first
1017 						   bucket. */
1018 	struct ocfs2_block_check xh_check;	/* Error checking
1019 						   (Note, this is only
1020 						    used for xattr
1021 						    buckets.  A block uses
1022 						    xb_check and sets
1023 						    this field to zero.) */
1024 	struct ocfs2_xattr_entry xh_entries[]; /* xattr entry list. */
1025 };
1026 
1027 /*
1028  * On disk structure for xattr value root.
1029  *
1030  * When an xattr's value is large enough, it is stored in an external
1031  * b-tree like file data.  The xattr value root points to this structure.
1032  */
1033 struct ocfs2_xattr_value_root {
1034 /*00*/	__le32	xr_clusters;              /* clusters covered by xattr value. */
1035 	__le32	xr_reserved0;
1036 	__le64	xr_last_eb_blk;           /* Pointer to last extent block */
1037 /*10*/	struct ocfs2_extent_list xr_list; /* Extent record list */
1038 };
1039 
1040 /*
1041  * On disk structure for xattr tree root.
1042  *
1043  * It is used when there are too many extended attributes for one file. These
1044  * attributes will be organized and stored in an indexed-btree.
1045  */
1046 struct ocfs2_xattr_tree_root {
1047 /*00*/	__le32	xt_clusters;              /* clusters covered by xattr. */
1048 	__le32	xt_reserved0;
1049 	__le64	xt_last_eb_blk;           /* Pointer to last extent block */
1050 /*10*/	struct ocfs2_extent_list xt_list; /* Extent record list */
1051 };
1052 
1053 #define OCFS2_XATTR_INDEXED	0x1
1054 #define OCFS2_HASH_SHIFT	5
1055 #define OCFS2_XATTR_ROUND	3
1056 #define OCFS2_XATTR_SIZE(size)	(((size) + OCFS2_XATTR_ROUND) & \
1057 				~(OCFS2_XATTR_ROUND))
1058 
1059 #define OCFS2_XATTR_BUCKET_SIZE			4096
1060 #define OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET 	(OCFS2_XATTR_BUCKET_SIZE \
1061 						 / OCFS2_MIN_BLOCKSIZE)
1062 
1063 /*
1064  * On disk structure for xattr block.
1065  */
1066 struct ocfs2_xattr_block {
1067 /*00*/	__u8	xb_signature[8];     /* Signature for verification */
1068 	__le16	xb_suballoc_slot;    /* Slot suballocator this
1069 					block belongs to. */
1070 	__le16	xb_suballoc_bit;     /* Bit offset in suballocator
1071 					block group */
1072 	__le32	xb_fs_generation;    /* Must match super block */
1073 /*10*/	__le64	xb_blkno;            /* Offset on disk, in blocks */
1074 	struct ocfs2_block_check xb_check;	/* Error checking */
1075 /*20*/	__le16	xb_flags;            /* Indicates whether this block contains
1076 					real xattr or a xattr tree. */
1077 	__le16	xb_reserved0;
1078 	__le32  xb_reserved1;
1079 	__le64	xb_suballoc_loc;	/* Suballocator block group this
1080 					   xattr block belongs to. Only
1081 					   valid if allocated from a
1082 					   discontiguous block group */
1083 /*30*/	union {
1084 		struct ocfs2_xattr_header xb_header; /* xattr header if this
1085 							block contains xattr */
1086 		struct ocfs2_xattr_tree_root xb_root;/* xattr tree root if this
1087 							block cotains xattr
1088 							tree. */
1089 	} xb_attrs;
1090 };
1091 
1092 #define OCFS2_XATTR_ENTRY_LOCAL		0x80
1093 #define OCFS2_XATTR_TYPE_MASK		0x7F
1094 static inline void ocfs2_xattr_set_local(struct ocfs2_xattr_entry *xe,
1095 					 int local)
1096 {
1097 	if (local)
1098 		xe->xe_type |= OCFS2_XATTR_ENTRY_LOCAL;
1099 	else
1100 		xe->xe_type &= ~OCFS2_XATTR_ENTRY_LOCAL;
1101 }
1102 
1103 static inline int ocfs2_xattr_is_local(struct ocfs2_xattr_entry *xe)
1104 {
1105 	return xe->xe_type & OCFS2_XATTR_ENTRY_LOCAL;
1106 }
1107 
1108 static inline void ocfs2_xattr_set_type(struct ocfs2_xattr_entry *xe, int type)
1109 {
1110 	xe->xe_type |= type & OCFS2_XATTR_TYPE_MASK;
1111 }
1112 
1113 static inline int ocfs2_xattr_get_type(struct ocfs2_xattr_entry *xe)
1114 {
1115 	return xe->xe_type & OCFS2_XATTR_TYPE_MASK;
1116 }
1117 
1118 /*
1119  *  On disk structures for global quota file
1120  */
1121 
1122 /* Magic numbers and known versions for global quota files */
1123 #define OCFS2_GLOBAL_QMAGICS {\
1124 	0x0cf52470, /* USRQUOTA */ \
1125 	0x0cf52471  /* GRPQUOTA */ \
1126 }
1127 
1128 #define OCFS2_GLOBAL_QVERSIONS {\
1129 	0, \
1130 	0, \
1131 }
1132 
1133 
1134 /* Each block of each quota file has a certain fixed number of bytes reserved
1135  * for OCFS2 internal use at its end. OCFS2 can use it for things like
1136  * checksums, etc. */
1137 #define OCFS2_QBLK_RESERVED_SPACE 8
1138 
1139 /* Generic header of all quota files */
1140 struct ocfs2_disk_dqheader {
1141 	__le32 dqh_magic;	/* Magic number identifying file */
1142 	__le32 dqh_version;	/* Quota format version */
1143 };
1144 
1145 #define OCFS2_GLOBAL_INFO_OFF (sizeof(struct ocfs2_disk_dqheader))
1146 
1147 /* Information header of global quota file (immediately follows the generic
1148  * header) */
1149 struct ocfs2_global_disk_dqinfo {
1150 /*00*/	__le32 dqi_bgrace;	/* Grace time for space softlimit excess */
1151 	__le32 dqi_igrace;	/* Grace time for inode softlimit excess */
1152 	__le32 dqi_syncms;	/* Time after which we sync local changes to
1153 				 * global quota file */
1154 	__le32 dqi_blocks;	/* Number of blocks in quota file */
1155 /*10*/	__le32 dqi_free_blk;	/* First free block in quota file */
1156 	__le32 dqi_free_entry;	/* First block with free dquot entry in quota
1157 				 * file */
1158 };
1159 
1160 /* Structure with global user / group information. We reserve some space
1161  * for future use. */
1162 struct ocfs2_global_disk_dqblk {
1163 /*00*/	__le32 dqb_id;          /* ID the structure belongs to */
1164 	__le32 dqb_use_count;   /* Number of nodes having reference to this structure */
1165 	__le64 dqb_ihardlimit;  /* absolute limit on allocated inodes */
1166 /*10*/	__le64 dqb_isoftlimit;  /* preferred inode limit */
1167 	__le64 dqb_curinodes;   /* current # allocated inodes */
1168 /*20*/	__le64 dqb_bhardlimit;  /* absolute limit on disk space */
1169 	__le64 dqb_bsoftlimit;  /* preferred limit on disk space */
1170 /*30*/	__le64 dqb_curspace;    /* current space occupied */
1171 	__le64 dqb_btime;       /* time limit for excessive disk use */
1172 /*40*/	__le64 dqb_itime;       /* time limit for excessive inode use */
1173 	__le64 dqb_pad1;
1174 /*50*/	__le64 dqb_pad2;
1175 };
1176 
1177 /*
1178  *  On-disk structures for local quota file
1179  */
1180 
1181 /* Magic numbers and known versions for local quota files */
1182 #define OCFS2_LOCAL_QMAGICS {\
1183 	0x0cf524c0, /* USRQUOTA */ \
1184 	0x0cf524c1  /* GRPQUOTA */ \
1185 }
1186 
1187 #define OCFS2_LOCAL_QVERSIONS {\
1188 	0, \
1189 	0, \
1190 }
1191 
1192 /* Quota flags in dqinfo header */
1193 #define OLQF_CLEAN	0x0001	/* Quota file is empty (this should be after\
1194 				 * quota has been cleanly turned off) */
1195 
1196 #define OCFS2_LOCAL_INFO_OFF (sizeof(struct ocfs2_disk_dqheader))
1197 
1198 /* Information header of local quota file (immediately follows the generic
1199  * header) */
1200 struct ocfs2_local_disk_dqinfo {
1201 	__le32 dqi_flags;	/* Flags for quota file */
1202 	__le32 dqi_chunks;	/* Number of chunks of quota structures
1203 				 * with a bitmap */
1204 	__le32 dqi_blocks;	/* Number of blocks allocated for quota file */
1205 };
1206 
1207 /* Header of one chunk of a quota file */
1208 struct ocfs2_local_disk_chunk {
1209 	__le32 dqc_free;	/* Number of free entries in the bitmap */
1210 	__u8 dqc_bitmap[];	/* Bitmap of entries in the corresponding
1211 				 * chunk of quota file */
1212 };
1213 
1214 /* One entry in local quota file */
1215 struct ocfs2_local_disk_dqblk {
1216 /*00*/	__le64 dqb_id;		/* id this quota applies to */
1217 	__le64 dqb_spacemod;	/* Change in the amount of used space */
1218 /*10*/	__le64 dqb_inodemod;	/* Change in the amount of used inodes */
1219 };
1220 
1221 
1222 /*
1223  * The quota trailer lives at the end of each quota block.
1224  */
1225 
1226 struct ocfs2_disk_dqtrailer {
1227 /*00*/	struct ocfs2_block_check dq_check;	/* Error checking */
1228 /*08*/	/* Cannot be larger than OCFS2_QBLK_RESERVED_SPACE */
1229 };
1230 
1231 static inline struct ocfs2_disk_dqtrailer *ocfs2_block_dqtrailer(int blocksize,
1232 								 void *buf)
1233 {
1234 	char *ptr = buf;
1235 	ptr += blocksize - OCFS2_QBLK_RESERVED_SPACE;
1236 
1237 	return (struct ocfs2_disk_dqtrailer *)ptr;
1238 }
1239 
1240 #ifdef __KERNEL__
1241 static inline int ocfs2_fast_symlink_chars(struct super_block *sb)
1242 {
1243 	return  sb->s_blocksize -
1244 		 offsetof(struct ocfs2_dinode, id2.i_symlink);
1245 }
1246 
1247 static inline int ocfs2_max_inline_data_with_xattr(struct super_block *sb,
1248 						   struct ocfs2_dinode *di)
1249 {
1250 	unsigned int xattrsize = le16_to_cpu(di->i_xattr_inline_size);
1251 
1252 	if (le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_XATTR_FL)
1253 		return sb->s_blocksize -
1254 			offsetof(struct ocfs2_dinode, id2.i_data.id_data) -
1255 			xattrsize;
1256 	else
1257 		return sb->s_blocksize -
1258 			offsetof(struct ocfs2_dinode, id2.i_data.id_data);
1259 }
1260 
1261 static inline int ocfs2_extent_recs_per_inode(struct super_block *sb)
1262 {
1263 	int size;
1264 
1265 	size = sb->s_blocksize -
1266 		offsetof(struct ocfs2_dinode, id2.i_list.l_recs);
1267 
1268 	return size / sizeof(struct ocfs2_extent_rec);
1269 }
1270 
1271 static inline int ocfs2_extent_recs_per_inode_with_xattr(
1272 						struct super_block *sb,
1273 						struct ocfs2_dinode *di)
1274 {
1275 	int size;
1276 	unsigned int xattrsize = le16_to_cpu(di->i_xattr_inline_size);
1277 
1278 	if (le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_XATTR_FL)
1279 		size = sb->s_blocksize -
1280 			offsetof(struct ocfs2_dinode, id2.i_list.l_recs) -
1281 			xattrsize;
1282 	else
1283 		size = sb->s_blocksize -
1284 			offsetof(struct ocfs2_dinode, id2.i_list.l_recs);
1285 
1286 	return size / sizeof(struct ocfs2_extent_rec);
1287 }
1288 
1289 static inline int ocfs2_extent_recs_per_dx_root(struct super_block *sb)
1290 {
1291 	int size;
1292 
1293 	size = sb->s_blocksize -
1294 		offsetof(struct ocfs2_dx_root_block, dr_list.l_recs);
1295 
1296 	return size / sizeof(struct ocfs2_extent_rec);
1297 }
1298 
1299 static inline int ocfs2_chain_recs_per_inode(struct super_block *sb)
1300 {
1301 	int size;
1302 
1303 	size = sb->s_blocksize -
1304 		offsetof(struct ocfs2_dinode, id2.i_chain.cl_recs);
1305 
1306 	return size / sizeof(struct ocfs2_chain_rec);
1307 }
1308 
1309 static inline u16 ocfs2_extent_recs_per_eb(struct super_block *sb)
1310 {
1311 	int size;
1312 
1313 	size = sb->s_blocksize -
1314 		offsetof(struct ocfs2_extent_block, h_list.l_recs);
1315 
1316 	return size / sizeof(struct ocfs2_extent_rec);
1317 }
1318 
1319 static inline u16 ocfs2_extent_recs_per_gd(struct super_block *sb)
1320 {
1321 	int size;
1322 
1323 	size = sb->s_blocksize -
1324 		offsetof(struct ocfs2_group_desc, bg_list.l_recs);
1325 
1326 	return size / sizeof(struct ocfs2_extent_rec);
1327 }
1328 
1329 static inline int ocfs2_dx_entries_per_leaf(struct super_block *sb)
1330 {
1331 	int size;
1332 
1333 	size = sb->s_blocksize -
1334 		offsetof(struct ocfs2_dx_leaf, dl_list.de_entries);
1335 
1336 	return size / sizeof(struct ocfs2_dx_entry);
1337 }
1338 
1339 static inline int ocfs2_dx_entries_per_root(struct super_block *sb)
1340 {
1341 	int size;
1342 
1343 	size = sb->s_blocksize -
1344 		offsetof(struct ocfs2_dx_root_block, dr_entries.de_entries);
1345 
1346 	return size / sizeof(struct ocfs2_dx_entry);
1347 }
1348 
1349 static inline u16 ocfs2_local_alloc_size(struct super_block *sb)
1350 {
1351 	u16 size;
1352 
1353 	size = sb->s_blocksize -
1354 		offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap);
1355 
1356 	return size;
1357 }
1358 
1359 static inline int ocfs2_group_bitmap_size(struct super_block *sb,
1360 					  int suballocator,
1361 					  u32 feature_incompat)
1362 {
1363 	int size = sb->s_blocksize -
1364 		offsetof(struct ocfs2_group_desc, bg_bitmap);
1365 
1366 	/*
1367 	 * The cluster allocator uses the entire block.  Suballocators have
1368 	 * never used more than OCFS2_MAX_BG_BITMAP_SIZE.  Unfortunately, older
1369 	 * code expects bg_size set to the maximum.  Thus we must keep
1370 	 * bg_size as-is unless discontig_bg is enabled.
1371 	 */
1372 	if (suballocator &&
1373 	    (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG))
1374 		size = OCFS2_MAX_BG_BITMAP_SIZE;
1375 
1376 	return size;
1377 }
1378 
1379 static inline int ocfs2_truncate_recs_per_inode(struct super_block *sb)
1380 {
1381 	int size;
1382 
1383 	size = sb->s_blocksize -
1384 		offsetof(struct ocfs2_dinode, id2.i_dealloc.tl_recs);
1385 
1386 	return size / sizeof(struct ocfs2_truncate_rec);
1387 }
1388 
1389 static inline u64 ocfs2_backup_super_blkno(struct super_block *sb, int index)
1390 {
1391 	u64 offset = OCFS2_BACKUP_SB_START;
1392 
1393 	if (index >= 0 && index < OCFS2_MAX_BACKUP_SUPERBLOCKS) {
1394 		offset <<= (2 * index);
1395 		offset >>= sb->s_blocksize_bits;
1396 		return offset;
1397 	}
1398 
1399 	return 0;
1400 
1401 }
1402 
1403 static inline u16 ocfs2_xattr_recs_per_xb(struct super_block *sb)
1404 {
1405 	int size;
1406 
1407 	size = sb->s_blocksize -
1408 		offsetof(struct ocfs2_xattr_block,
1409 			 xb_attrs.xb_root.xt_list.l_recs);
1410 
1411 	return size / sizeof(struct ocfs2_extent_rec);
1412 }
1413 
1414 static inline u16 ocfs2_extent_recs_per_rb(struct super_block *sb)
1415 {
1416 	int size;
1417 
1418 	size = sb->s_blocksize -
1419 		offsetof(struct ocfs2_refcount_block, rf_list.l_recs);
1420 
1421 	return size / sizeof(struct ocfs2_extent_rec);
1422 }
1423 
1424 static inline u16 ocfs2_refcount_recs_per_rb(struct super_block *sb)
1425 {
1426 	int size;
1427 
1428 	size = sb->s_blocksize -
1429 		offsetof(struct ocfs2_refcount_block, rf_records.rl_recs);
1430 
1431 	return size / sizeof(struct ocfs2_refcount_rec);
1432 }
1433 
1434 static inline u32
1435 ocfs2_get_ref_rec_low_cpos(const struct ocfs2_refcount_rec *rec)
1436 {
1437 	return le64_to_cpu(rec->r_cpos) & OCFS2_32BIT_POS_MASK;
1438 }
1439 #else
1440 static inline int ocfs2_fast_symlink_chars(int blocksize)
1441 {
1442 	return blocksize - offsetof(struct ocfs2_dinode, id2.i_symlink);
1443 }
1444 
1445 static inline int ocfs2_max_inline_data_with_xattr(int blocksize,
1446 						   struct ocfs2_dinode *di)
1447 {
1448 	if (di && (di->i_dyn_features & OCFS2_INLINE_XATTR_FL))
1449 		return blocksize -
1450 			offsetof(struct ocfs2_dinode, id2.i_data.id_data) -
1451 			di->i_xattr_inline_size;
1452 	else
1453 		return blocksize -
1454 			offsetof(struct ocfs2_dinode, id2.i_data.id_data);
1455 }
1456 
1457 static inline int ocfs2_extent_recs_per_inode(int blocksize)
1458 {
1459 	int size;
1460 
1461 	size = blocksize -
1462 		offsetof(struct ocfs2_dinode, id2.i_list.l_recs);
1463 
1464 	return size / sizeof(struct ocfs2_extent_rec);
1465 }
1466 
1467 static inline int ocfs2_chain_recs_per_inode(int blocksize)
1468 {
1469 	int size;
1470 
1471 	size = blocksize -
1472 		offsetof(struct ocfs2_dinode, id2.i_chain.cl_recs);
1473 
1474 	return size / sizeof(struct ocfs2_chain_rec);
1475 }
1476 
1477 static inline int ocfs2_extent_recs_per_eb(int blocksize)
1478 {
1479 	int size;
1480 
1481 	size = blocksize -
1482 		offsetof(struct ocfs2_extent_block, h_list.l_recs);
1483 
1484 	return size / sizeof(struct ocfs2_extent_rec);
1485 }
1486 
1487 static inline int ocfs2_extent_recs_per_gd(int blocksize)
1488 {
1489 	int size;
1490 
1491 	size = blocksize -
1492 		offsetof(struct ocfs2_group_desc, bg_list.l_recs);
1493 
1494 	return size / sizeof(struct ocfs2_extent_rec);
1495 }
1496 
1497 static inline int ocfs2_local_alloc_size(int blocksize)
1498 {
1499 	int size;
1500 
1501 	size = blocksize -
1502 		offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap);
1503 
1504 	return size;
1505 }
1506 
1507 static inline int ocfs2_group_bitmap_size(int blocksize,
1508 					  int suballocator,
1509 					  uint32_t feature_incompat)
1510 {
1511 	int size = sb->s_blocksize -
1512 		offsetof(struct ocfs2_group_desc, bg_bitmap);
1513 
1514 	/*
1515 	 * The cluster allocator uses the entire block.  Suballocators have
1516 	 * never used more than OCFS2_MAX_BG_BITMAP_SIZE.  Unfortunately, older
1517 	 * code expects bg_size set to the maximum.  Thus we must keep
1518 	 * bg_size as-is unless discontig_bg is enabled.
1519 	 */
1520 	if (suballocator &&
1521 	    (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG))
1522 		size = OCFS2_MAX_BG_BITMAP_SIZE;
1523 
1524 	return size;
1525 }
1526 
1527 static inline int ocfs2_truncate_recs_per_inode(int blocksize)
1528 {
1529 	int size;
1530 
1531 	size = blocksize -
1532 		offsetof(struct ocfs2_dinode, id2.i_dealloc.tl_recs);
1533 
1534 	return size / sizeof(struct ocfs2_truncate_rec);
1535 }
1536 
1537 static inline uint64_t ocfs2_backup_super_blkno(int blocksize, int index)
1538 {
1539 	uint64_t offset = OCFS2_BACKUP_SB_START;
1540 
1541 	if (index >= 0 && index < OCFS2_MAX_BACKUP_SUPERBLOCKS) {
1542 		offset <<= (2 * index);
1543 		offset /= blocksize;
1544 		return offset;
1545 	}
1546 
1547 	return 0;
1548 }
1549 
1550 static inline int ocfs2_xattr_recs_per_xb(int blocksize)
1551 {
1552 	int size;
1553 
1554 	size = blocksize -
1555 		offsetof(struct ocfs2_xattr_block,
1556 			 xb_attrs.xb_root.xt_list.l_recs);
1557 
1558 	return size / sizeof(struct ocfs2_extent_rec);
1559 }
1560 #endif  /* __KERNEL__ */
1561 
1562 
1563 static inline int ocfs2_system_inode_is_global(int type)
1564 {
1565 	return ((type >= 0) &&
1566 		(type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE));
1567 }
1568 
1569 static inline int ocfs2_sprintf_system_inode_name(char *buf, int len,
1570 						  int type, int slot)
1571 {
1572 	int chars;
1573 
1574         /*
1575          * Global system inodes can only have one copy.  Everything
1576          * after OCFS2_LAST_GLOBAL_SYSTEM_INODE in the system inode
1577          * list has a copy per slot.
1578          */
1579 	if (type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE)
1580 		chars = snprintf(buf, len, "%s",
1581 				 ocfs2_system_inodes[type].si_name);
1582 	else
1583 		chars = snprintf(buf, len,
1584 				 ocfs2_system_inodes[type].si_name,
1585 				 slot);
1586 
1587 	return chars;
1588 }
1589 
1590 static inline void ocfs2_set_de_type(struct ocfs2_dir_entry *de,
1591 				    umode_t mode)
1592 {
1593 	de->file_type = fs_umode_to_ftype(mode);
1594 }
1595 
1596 static inline int ocfs2_gd_is_discontig(struct ocfs2_group_desc *gd)
1597 {
1598 	if ((offsetof(struct ocfs2_group_desc, bg_bitmap) +
1599 	     le16_to_cpu(gd->bg_size)) !=
1600 	    offsetof(struct ocfs2_group_desc, bg_list))
1601 		return 0;
1602 	/*
1603 	 * Only valid to check l_next_free_rec if
1604 	 * bg_bitmap + bg_size == bg_list.
1605 	 */
1606 	if (!gd->bg_list.l_next_free_rec)
1607 		return 0;
1608 	return 1;
1609 }
1610 #endif  /* _OCFS2_FS_H */
1611 
1612