xref: /linux/fs/ocfs2/ocfs2_fs.h (revision d39d0ed196aa1685bb24771e92f78633c66ac9cb)
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * ocfs2_fs.h
5  *
6  * On-disk structures for OCFS2.
7  *
8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
9  *
10  * This program is free software; you can redistribute it and/or
11  * modify it under the terms of the GNU General Public
12  * License, version 2,  as published by the Free Software Foundation.
13  *
14  * This program is distributed in the hope that it will be useful,
15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
17  * General Public License for more details.
18  *
19  * You should have received a copy of the GNU General Public
20  * License along with this program; if not, write to the
21  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
22  * Boston, MA 021110-1307, USA.
23  */
24 
25 #ifndef _OCFS2_FS_H
26 #define _OCFS2_FS_H
27 
28 /* Version */
29 #define OCFS2_MAJOR_REV_LEVEL		0
30 #define OCFS2_MINOR_REV_LEVEL          	90
31 
32 /*
33  * An OCFS2 volume starts this way:
34  * Sector 0: Valid ocfs1_vol_disk_hdr that cleanly fails to mount OCFS.
35  * Sector 1: Valid ocfs1_vol_label that cleanly fails to mount OCFS.
36  * Block OCFS2_SUPER_BLOCK_BLKNO: OCFS2 superblock.
37  *
38  * All other structures are found from the superblock information.
39  *
40  * OCFS2_SUPER_BLOCK_BLKNO is in blocks, not sectors.  eg, for a
41  * blocksize of 2K, it is 4096 bytes into disk.
42  */
43 #define OCFS2_SUPER_BLOCK_BLKNO		2
44 
45 /*
46  * Cluster size limits. The maximum is kept arbitrarily at 1 MB, and could
47  * grow if needed.
48  */
49 #define OCFS2_MIN_CLUSTERSIZE		4096
50 #define OCFS2_MAX_CLUSTERSIZE		1048576
51 
52 /*
53  * Blocks cannot be bigger than clusters, so the maximum blocksize is the
54  * minimum cluster size.
55  */
56 #define OCFS2_MIN_BLOCKSIZE		512
57 #define OCFS2_MAX_BLOCKSIZE		OCFS2_MIN_CLUSTERSIZE
58 
59 /* Filesystem magic number */
60 #define OCFS2_SUPER_MAGIC		0x7461636f
61 
62 /* Object signatures */
63 #define OCFS2_SUPER_BLOCK_SIGNATURE	"OCFSV2"
64 #define OCFS2_INODE_SIGNATURE		"INODE01"
65 #define OCFS2_EXTENT_BLOCK_SIGNATURE	"EXBLK01"
66 #define OCFS2_GROUP_DESC_SIGNATURE      "GROUP01"
67 #define OCFS2_XATTR_BLOCK_SIGNATURE	"XATTR01"
68 #define OCFS2_DIR_TRAILER_SIGNATURE	"DIRTRL1"
69 #define OCFS2_DX_ROOT_SIGNATURE		"DXDIR01"
70 #define OCFS2_DX_LEAF_SIGNATURE		"DXLEAF1"
71 #define OCFS2_REFCOUNT_BLOCK_SIGNATURE	"REFCNT1"
72 
73 /* Compatibility flags */
74 #define OCFS2_HAS_COMPAT_FEATURE(sb,mask)			\
75 	( OCFS2_SB(sb)->s_feature_compat & (mask) )
76 #define OCFS2_HAS_RO_COMPAT_FEATURE(sb,mask)			\
77 	( OCFS2_SB(sb)->s_feature_ro_compat & (mask) )
78 #define OCFS2_HAS_INCOMPAT_FEATURE(sb,mask)			\
79 	( OCFS2_SB(sb)->s_feature_incompat & (mask) )
80 #define OCFS2_SET_COMPAT_FEATURE(sb,mask)			\
81 	OCFS2_SB(sb)->s_feature_compat |= (mask)
82 #define OCFS2_SET_RO_COMPAT_FEATURE(sb,mask)			\
83 	OCFS2_SB(sb)->s_feature_ro_compat |= (mask)
84 #define OCFS2_SET_INCOMPAT_FEATURE(sb,mask)			\
85 	OCFS2_SB(sb)->s_feature_incompat |= (mask)
86 #define OCFS2_CLEAR_COMPAT_FEATURE(sb,mask)			\
87 	OCFS2_SB(sb)->s_feature_compat &= ~(mask)
88 #define OCFS2_CLEAR_RO_COMPAT_FEATURE(sb,mask)			\
89 	OCFS2_SB(sb)->s_feature_ro_compat &= ~(mask)
90 #define OCFS2_CLEAR_INCOMPAT_FEATURE(sb,mask)			\
91 	OCFS2_SB(sb)->s_feature_incompat &= ~(mask)
92 
93 #define OCFS2_FEATURE_COMPAT_SUPP	(OCFS2_FEATURE_COMPAT_BACKUP_SB	\
94 					 | OCFS2_FEATURE_COMPAT_JBD2_SB)
95 #define OCFS2_FEATURE_INCOMPAT_SUPP	(OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT \
96 					 | OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC \
97 					 | OCFS2_FEATURE_INCOMPAT_INLINE_DATA \
98 					 | OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP \
99 					 | OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK \
100 					 | OCFS2_FEATURE_INCOMPAT_XATTR \
101 					 | OCFS2_FEATURE_INCOMPAT_META_ECC \
102 					 | OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS \
103 					 | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE \
104 					 | OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
105 #define OCFS2_FEATURE_RO_COMPAT_SUPP	(OCFS2_FEATURE_RO_COMPAT_UNWRITTEN \
106 					 | OCFS2_FEATURE_RO_COMPAT_USRQUOTA \
107 					 | OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
108 
109 /*
110  * Heartbeat-only devices are missing journals and other files.  The
111  * filesystem driver can't load them, but the library can.  Never put
112  * this in OCFS2_FEATURE_INCOMPAT_SUPP, *ever*.
113  */
114 #define OCFS2_FEATURE_INCOMPAT_HEARTBEAT_DEV	0x0002
115 
116 /*
117  * tunefs sets this incompat flag before starting the resize and clears it
118  * at the end. This flag protects users from inadvertently mounting the fs
119  * after an aborted run without fsck-ing.
120  */
121 #define OCFS2_FEATURE_INCOMPAT_RESIZE_INPROG    0x0004
122 
123 /* Used to denote a non-clustered volume */
124 #define OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT	0x0008
125 
126 /* Support for sparse allocation in b-trees */
127 #define OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC	0x0010
128 
129 /*
130  * Tunefs sets this incompat flag before starting an operation which
131  * would require cleanup on abort. This is done to protect users from
132  * inadvertently mounting the fs after an aborted run without
133  * fsck-ing.
134  *
135  * s_tunefs_flags on the super block describes precisely which
136  * operations were in progress.
137  */
138 #define OCFS2_FEATURE_INCOMPAT_TUNEFS_INPROG	0x0020
139 
140 /* Support for data packed into inode blocks */
141 #define OCFS2_FEATURE_INCOMPAT_INLINE_DATA	0x0040
142 
143 /*
144  * Support for alternate, userspace cluster stacks.  If set, the superblock
145  * field s_cluster_info contains a tag for the alternate stack in use as
146  * well as the name of the cluster being joined.
147  * mount.ocfs2 must pass in a matching stack name.
148  *
149  * If not set, the classic stack will be used.  This is compatbile with
150  * all older versions.
151  */
152 #define OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK	0x0080
153 
154 /* Support for the extended slot map */
155 #define OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP 0x100
156 
157 /* Support for extended attributes */
158 #define OCFS2_FEATURE_INCOMPAT_XATTR		0x0200
159 
160 /* Support for indexed directores */
161 #define OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS	0x0400
162 
163 /* Metadata checksum and error correction */
164 #define OCFS2_FEATURE_INCOMPAT_META_ECC		0x0800
165 
166 /* Refcount tree support */
167 #define OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE	0x1000
168 
169 /* Discontigous block groups */
170 #define OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG	0x2000
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 /* The byte offset of the first backup block will be 1G.
195  * The following will be 4G, 16G, 64G, 256G and 1T.
196  */
197 #define OCFS2_BACKUP_SB_START			1 << 30
198 
199 /* the max backup superblock nums */
200 #define OCFS2_MAX_BACKUP_SUPERBLOCKS	6
201 
202 /*
203  * Flags on ocfs2_super_block.s_tunefs_flags
204  */
205 #define OCFS2_TUNEFS_INPROG_REMOVE_SLOT		0x0001	/* Removing slots */
206 
207 /*
208  * Flags on ocfs2_dinode.i_flags
209  */
210 #define OCFS2_VALID_FL		(0x00000001)	/* Inode is valid */
211 #define OCFS2_UNUSED2_FL	(0x00000002)
212 #define OCFS2_ORPHANED_FL	(0x00000004)	/* On the orphan list */
213 #define OCFS2_UNUSED3_FL	(0x00000008)
214 /* System inode flags */
215 #define OCFS2_SYSTEM_FL		(0x00000010)	/* System inode */
216 #define OCFS2_SUPER_BLOCK_FL	(0x00000020)	/* Super block */
217 #define OCFS2_LOCAL_ALLOC_FL	(0x00000040)	/* Slot local alloc bitmap */
218 #define OCFS2_BITMAP_FL		(0x00000080)	/* Allocation bitmap */
219 #define OCFS2_JOURNAL_FL	(0x00000100)	/* Slot local journal */
220 #define OCFS2_HEARTBEAT_FL	(0x00000200)	/* Heartbeat area */
221 #define OCFS2_CHAIN_FL		(0x00000400)	/* Chain allocator */
222 #define OCFS2_DEALLOC_FL	(0x00000800)	/* Truncate log */
223 #define OCFS2_QUOTA_FL		(0x00001000)	/* Quota file */
224 
225 /*
226  * Flags on ocfs2_dinode.i_dyn_features
227  *
228  * These can change much more often than i_flags. When adding flags,
229  * keep in mind that i_dyn_features is only 16 bits wide.
230  */
231 #define OCFS2_INLINE_DATA_FL	(0x0001)	/* Data stored in inode block */
232 #define OCFS2_HAS_XATTR_FL	(0x0002)
233 #define OCFS2_INLINE_XATTR_FL	(0x0004)
234 #define OCFS2_INDEXED_DIR_FL	(0x0008)
235 #define OCFS2_HAS_REFCOUNT_FL   (0x0010)
236 
237 /* Inode attributes, keep in sync with EXT2 */
238 #define OCFS2_SECRM_FL		(0x00000001)	/* Secure deletion */
239 #define OCFS2_UNRM_FL		(0x00000002)	/* Undelete */
240 #define OCFS2_COMPR_FL		(0x00000004)	/* Compress file */
241 #define OCFS2_SYNC_FL		(0x00000008)	/* Synchronous updates */
242 #define OCFS2_IMMUTABLE_FL	(0x00000010)	/* Immutable file */
243 #define OCFS2_APPEND_FL		(0x00000020)	/* writes to file may only append */
244 #define OCFS2_NODUMP_FL		(0x00000040)	/* do not dump file */
245 #define OCFS2_NOATIME_FL	(0x00000080)	/* do not update atime */
246 #define OCFS2_DIRSYNC_FL	(0x00010000)	/* dirsync behaviour (directories only) */
247 
248 #define OCFS2_FL_VISIBLE	(0x000100FF)	/* User visible flags */
249 #define OCFS2_FL_MODIFIABLE	(0x000100FF)	/* User modifiable flags */
250 
251 /*
252  * Extent record flags (e_node.leaf.flags)
253  */
254 #define OCFS2_EXT_UNWRITTEN		(0x01)	/* Extent is allocated but
255 						 * unwritten */
256 #define OCFS2_EXT_REFCOUNTED		(0x02)  /* Extent is reference
257 						 * counted in an associated
258 						 * refcount tree */
259 
260 /*
261  * Journal Flags (ocfs2_dinode.id1.journal1.i_flags)
262  */
263 #define OCFS2_JOURNAL_DIRTY_FL	(0x00000001)	/* Journal needs recovery */
264 
265 /*
266  * superblock s_state flags
267  */
268 #define OCFS2_ERROR_FS		(0x00000001)	/* FS saw errors */
269 
270 /* Limit of space in ocfs2_dir_entry */
271 #define OCFS2_MAX_FILENAME_LEN		255
272 
273 /* Maximum slots on an ocfs2 file system */
274 #define OCFS2_MAX_SLOTS			255
275 
276 /* Slot map indicator for an empty slot */
277 #define OCFS2_INVALID_SLOT		-1
278 
279 #define OCFS2_VOL_UUID_LEN		16
280 #define OCFS2_MAX_VOL_LABEL_LEN		64
281 
282 /* The alternate, userspace stack fields */
283 #define OCFS2_STACK_LABEL_LEN		4
284 #define OCFS2_CLUSTER_NAME_LEN		16
285 
286 /* Journal limits (in bytes) */
287 #define OCFS2_MIN_JOURNAL_SIZE		(4 * 1024 * 1024)
288 
289 /*
290  * Inline extended attribute size (in bytes)
291  * The value chosen should be aligned to 16 byte boundaries.
292  */
293 #define OCFS2_MIN_XATTR_INLINE_SIZE     256
294 
295 struct ocfs2_system_inode_info {
296 	char	*si_name;
297 	int	si_iflags;
298 	int	si_mode;
299 };
300 
301 /* System file index */
302 enum {
303 	BAD_BLOCK_SYSTEM_INODE = 0,
304 	GLOBAL_INODE_ALLOC_SYSTEM_INODE,
305 	SLOT_MAP_SYSTEM_INODE,
306 #define OCFS2_FIRST_ONLINE_SYSTEM_INODE SLOT_MAP_SYSTEM_INODE
307 	HEARTBEAT_SYSTEM_INODE,
308 	GLOBAL_BITMAP_SYSTEM_INODE,
309 	USER_QUOTA_SYSTEM_INODE,
310 	GROUP_QUOTA_SYSTEM_INODE,
311 #define OCFS2_LAST_GLOBAL_SYSTEM_INODE GROUP_QUOTA_SYSTEM_INODE
312 	ORPHAN_DIR_SYSTEM_INODE,
313 	EXTENT_ALLOC_SYSTEM_INODE,
314 	INODE_ALLOC_SYSTEM_INODE,
315 	JOURNAL_SYSTEM_INODE,
316 	LOCAL_ALLOC_SYSTEM_INODE,
317 	TRUNCATE_LOG_SYSTEM_INODE,
318 	LOCAL_USER_QUOTA_SYSTEM_INODE,
319 	LOCAL_GROUP_QUOTA_SYSTEM_INODE,
320 	NUM_SYSTEM_INODES
321 };
322 
323 static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = {
324 	/* Global system inodes (single copy) */
325 	/* The first two are only used from userspace mfks/tunefs */
326 	[BAD_BLOCK_SYSTEM_INODE]		= { "bad_blocks", 0, S_IFREG | 0644 },
327 	[GLOBAL_INODE_ALLOC_SYSTEM_INODE] 	= { "global_inode_alloc", OCFS2_BITMAP_FL | OCFS2_CHAIN_FL, S_IFREG | 0644 },
328 
329 	/* These are used by the running filesystem */
330 	[SLOT_MAP_SYSTEM_INODE]			= { "slot_map", 0, S_IFREG | 0644 },
331 	[HEARTBEAT_SYSTEM_INODE]		= { "heartbeat", OCFS2_HEARTBEAT_FL, S_IFREG | 0644 },
332 	[GLOBAL_BITMAP_SYSTEM_INODE]		= { "global_bitmap", 0, S_IFREG | 0644 },
333 	[USER_QUOTA_SYSTEM_INODE]		= { "aquota.user", OCFS2_QUOTA_FL, S_IFREG | 0644 },
334 	[GROUP_QUOTA_SYSTEM_INODE]		= { "aquota.group", OCFS2_QUOTA_FL, S_IFREG | 0644 },
335 
336 	/* Slot-specific system inodes (one copy per slot) */
337 	[ORPHAN_DIR_SYSTEM_INODE]		= { "orphan_dir:%04d", 0, S_IFDIR | 0755 },
338 	[EXTENT_ALLOC_SYSTEM_INODE]		= { "extent_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_CHAIN_FL, S_IFREG | 0644 },
339 	[INODE_ALLOC_SYSTEM_INODE]		= { "inode_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_CHAIN_FL, S_IFREG | 0644 },
340 	[JOURNAL_SYSTEM_INODE]			= { "journal:%04d", OCFS2_JOURNAL_FL, S_IFREG | 0644 },
341 	[LOCAL_ALLOC_SYSTEM_INODE]		= { "local_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_LOCAL_ALLOC_FL, S_IFREG | 0644 },
342 	[TRUNCATE_LOG_SYSTEM_INODE]		= { "truncate_log:%04d", OCFS2_DEALLOC_FL, S_IFREG | 0644 },
343 	[LOCAL_USER_QUOTA_SYSTEM_INODE]		= { "aquota.user:%04d", OCFS2_QUOTA_FL, S_IFREG | 0644 },
344 	[LOCAL_GROUP_QUOTA_SYSTEM_INODE]	= { "aquota.group:%04d", OCFS2_QUOTA_FL, S_IFREG | 0644 },
345 };
346 
347 /* Parameter passed from mount.ocfs2 to module */
348 #define OCFS2_HB_NONE			"heartbeat=none"
349 #define OCFS2_HB_LOCAL			"heartbeat=local"
350 
351 /*
352  * OCFS2 directory file types.  Only the low 3 bits are used.  The
353  * other bits are reserved for now.
354  */
355 #define OCFS2_FT_UNKNOWN	0
356 #define OCFS2_FT_REG_FILE	1
357 #define OCFS2_FT_DIR		2
358 #define OCFS2_FT_CHRDEV		3
359 #define OCFS2_FT_BLKDEV		4
360 #define OCFS2_FT_FIFO		5
361 #define OCFS2_FT_SOCK		6
362 #define OCFS2_FT_SYMLINK	7
363 
364 #define OCFS2_FT_MAX		8
365 
366 /*
367  * OCFS2_DIR_PAD defines the directory entries boundaries
368  *
369  * NOTE: It must be a multiple of 4
370  */
371 #define OCFS2_DIR_PAD			4
372 #define OCFS2_DIR_ROUND			(OCFS2_DIR_PAD - 1)
373 #define OCFS2_DIR_MEMBER_LEN 		offsetof(struct ocfs2_dir_entry, name)
374 #define OCFS2_DIR_REC_LEN(name_len)	(((name_len) + OCFS2_DIR_MEMBER_LEN + \
375                                           OCFS2_DIR_ROUND) & \
376 					 ~OCFS2_DIR_ROUND)
377 #define OCFS2_DIR_MIN_REC_LEN	OCFS2_DIR_REC_LEN(1)
378 
379 #define OCFS2_LINK_MAX		32000
380 #define	OCFS2_DX_LINK_MAX	((1U << 31) - 1U)
381 #define	OCFS2_LINKS_HI_SHIFT	16
382 #define	OCFS2_DX_ENTRIES_MAX	(0xffffffffU)
383 
384 #define S_SHIFT			12
385 static unsigned char ocfs2_type_by_mode[S_IFMT >> S_SHIFT] = {
386 	[S_IFREG >> S_SHIFT]  = OCFS2_FT_REG_FILE,
387 	[S_IFDIR >> S_SHIFT]  = OCFS2_FT_DIR,
388 	[S_IFCHR >> S_SHIFT]  = OCFS2_FT_CHRDEV,
389 	[S_IFBLK >> S_SHIFT]  = OCFS2_FT_BLKDEV,
390 	[S_IFIFO >> S_SHIFT]  = OCFS2_FT_FIFO,
391 	[S_IFSOCK >> S_SHIFT] = OCFS2_FT_SOCK,
392 	[S_IFLNK >> S_SHIFT]  = OCFS2_FT_SYMLINK,
393 };
394 
395 
396 /*
397  * Convenience casts
398  */
399 #define OCFS2_RAW_SB(dinode)		(&((dinode)->id2.i_super))
400 
401 /*
402  * Block checking structure.  This is used in metadata to validate the
403  * contents.  If OCFS2_FEATURE_INCOMPAT_META_ECC is not set, it is all
404  * zeros.
405  */
406 struct ocfs2_block_check {
407 /*00*/	__le32 bc_crc32e;	/* 802.3 Ethernet II CRC32 */
408 	__le16 bc_ecc;		/* Single-error-correction parity vector.
409 				   This is a simple Hamming code dependant
410 				   on the blocksize.  OCFS2's maximum
411 				   blocksize, 4K, requires 16 parity bits,
412 				   so we fit in __le16. */
413 	__le16 bc_reserved1;
414 /*08*/
415 };
416 
417 /*
418  * On disk extent record for OCFS2
419  * It describes a range of clusters on disk.
420  *
421  * Length fields are divided into interior and leaf node versions.
422  * This leaves room for a flags field (OCFS2_EXT_*) in the leaf nodes.
423  */
424 struct ocfs2_extent_rec {
425 /*00*/	__le32 e_cpos;		/* Offset into the file, in clusters */
426 	union {
427 		__le32 e_int_clusters; /* Clusters covered by all children */
428 		struct {
429 			__le16 e_leaf_clusters; /* Clusters covered by this
430 						   extent */
431 			__u8 e_reserved1;
432 			__u8 e_flags; /* Extent flags */
433 		};
434 	};
435 	__le64 e_blkno;		/* Physical disk offset, in blocks */
436 /*10*/
437 };
438 
439 struct ocfs2_chain_rec {
440 	__le32 c_free;	/* Number of free bits in this chain. */
441 	__le32 c_total;	/* Number of total bits in this chain */
442 	__le64 c_blkno;	/* Physical disk offset (blocks) of 1st group */
443 };
444 
445 struct ocfs2_truncate_rec {
446 	__le32 t_start;		/* 1st cluster in this log */
447 	__le32 t_clusters;	/* Number of total clusters covered */
448 };
449 
450 /*
451  * On disk extent list for OCFS2 (node in the tree).  Note that this
452  * is contained inside ocfs2_dinode or ocfs2_extent_block, so the
453  * offsets are relative to ocfs2_dinode.id2.i_list or
454  * ocfs2_extent_block.h_list, respectively.
455  */
456 struct ocfs2_extent_list {
457 /*00*/	__le16 l_tree_depth;		/* Extent tree depth from this
458 					   point.  0 means data extents
459 					   hang directly off this
460 					   header (a leaf)
461 					   NOTE: The high 8 bits cannot be
462 					   used - tree_depth is never that big.
463 					*/
464 	__le16 l_count;			/* Number of extent records */
465 	__le16 l_next_free_rec;		/* Next unused extent slot */
466 	__le16 l_reserved1;
467 	__le64 l_reserved2;		/* Pad to
468 					   sizeof(ocfs2_extent_rec) */
469 /*10*/	struct ocfs2_extent_rec l_recs[0];	/* Extent records */
470 };
471 
472 /*
473  * On disk allocation chain list for OCFS2.  Note that this is
474  * contained inside ocfs2_dinode, so the offsets are relative to
475  * ocfs2_dinode.id2.i_chain.
476  */
477 struct ocfs2_chain_list {
478 /*00*/	__le16 cl_cpg;			/* Clusters per Block Group */
479 	__le16 cl_bpc;			/* Bits per cluster */
480 	__le16 cl_count;		/* Total chains in this list */
481 	__le16 cl_next_free_rec;	/* Next unused chain slot */
482 	__le64 cl_reserved1;
483 /*10*/	struct ocfs2_chain_rec cl_recs[0];	/* Chain records */
484 };
485 
486 /*
487  * On disk deallocation log for OCFS2.  Note that this is
488  * contained inside ocfs2_dinode, so the offsets are relative to
489  * ocfs2_dinode.id2.i_dealloc.
490  */
491 struct ocfs2_truncate_log {
492 /*00*/	__le16 tl_count;		/* Total records in this log */
493 	__le16 tl_used;			/* Number of records in use */
494 	__le32 tl_reserved1;
495 /*08*/	struct ocfs2_truncate_rec tl_recs[0];	/* Truncate records */
496 };
497 
498 /*
499  * On disk extent block (indirect block) for OCFS2
500  */
501 struct ocfs2_extent_block
502 {
503 /*00*/	__u8 h_signature[8];		/* Signature for verification */
504 	struct ocfs2_block_check h_check;	/* Error checking */
505 /*10*/	__le16 h_suballoc_slot;		/* Slot suballocator this
506 					   extent_header belongs to */
507 	__le16 h_suballoc_bit;		/* Bit offset in suballocator
508 					   block group */
509 	__le32 h_fs_generation;		/* Must match super block */
510 	__le64 h_blkno;			/* Offset on disk, in blocks */
511 /*20*/	__le64 h_suballoc_loc;		/* Suballocator block group this
512 					   eb belongs to.  Only valid
513 					   if allocated from a
514 					   discontiguous block group */
515 	__le64 h_next_leaf_blk;		/* Offset on disk, in blocks,
516 					   of next leaf header pointing
517 					   to data */
518 /*30*/	struct ocfs2_extent_list h_list;	/* Extent record list */
519 /* Actual on-disk size is one block */
520 };
521 
522 /*
523  * On disk slot map for OCFS2.  This defines the contents of the "slot_map"
524  * system file.  A slot is valid if it contains a node number >= 0.  The
525  * value -1 (0xFFFF) is OCFS2_INVALID_SLOT.  This marks a slot empty.
526  */
527 struct ocfs2_slot_map {
528 /*00*/	__le16 sm_slots[0];
529 /*
530  * Actual on-disk size is one block.  OCFS2_MAX_SLOTS is 255,
531  * 255 * sizeof(__le16) == 512B, within the 512B block minimum blocksize.
532  */
533 };
534 
535 struct ocfs2_extended_slot {
536 /*00*/	__u8	es_valid;
537 	__u8	es_reserved1[3];
538 	__le32	es_node_num;
539 /*10*/
540 };
541 
542 /*
543  * The extended slot map, used when OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP
544  * is set.  It separates out the valid marker from the node number, and
545  * has room to grow.  Unlike the old slot map, this format is defined by
546  * i_size.
547  */
548 struct ocfs2_slot_map_extended {
549 /*00*/	struct ocfs2_extended_slot se_slots[0];
550 /*
551  * Actual size is i_size of the slot_map system file.  It should
552  * match s_max_slots * sizeof(struct ocfs2_extended_slot)
553  */
554 };
555 
556 struct ocfs2_cluster_info {
557 /*00*/	__u8   ci_stack[OCFS2_STACK_LABEL_LEN];
558 	__le32 ci_reserved;
559 /*08*/	__u8   ci_cluster[OCFS2_CLUSTER_NAME_LEN];
560 /*18*/
561 };
562 
563 /*
564  * On disk superblock for OCFS2
565  * Note that it is contained inside an ocfs2_dinode, so all offsets
566  * are relative to the start of ocfs2_dinode.id2.
567  */
568 struct ocfs2_super_block {
569 /*00*/	__le16 s_major_rev_level;
570 	__le16 s_minor_rev_level;
571 	__le16 s_mnt_count;
572 	__le16 s_max_mnt_count;
573 	__le16 s_state;			/* File system state */
574 	__le16 s_errors;			/* Behaviour when detecting errors */
575 	__le32 s_checkinterval;		/* Max time between checks */
576 /*10*/	__le64 s_lastcheck;		/* Time of last check */
577 	__le32 s_creator_os;		/* OS */
578 	__le32 s_feature_compat;		/* Compatible feature set */
579 /*20*/	__le32 s_feature_incompat;	/* Incompatible feature set */
580 	__le32 s_feature_ro_compat;	/* Readonly-compatible feature set */
581 	__le64 s_root_blkno;		/* Offset, in blocks, of root directory
582 					   dinode */
583 /*30*/	__le64 s_system_dir_blkno;	/* Offset, in blocks, of system
584 					   directory dinode */
585 	__le32 s_blocksize_bits;		/* Blocksize for this fs */
586 	__le32 s_clustersize_bits;	/* Clustersize for this fs */
587 /*40*/	__le16 s_max_slots;		/* Max number of simultaneous mounts
588 					   before tunefs required */
589 	__le16 s_tunefs_flag;
590 	__le32 s_uuid_hash;		/* hash value of uuid */
591 	__le64 s_first_cluster_group;	/* Block offset of 1st cluster
592 					 * group header */
593 /*50*/	__u8  s_label[OCFS2_MAX_VOL_LABEL_LEN];	/* Label for mounting, etc. */
594 /*90*/	__u8  s_uuid[OCFS2_VOL_UUID_LEN];	/* 128-bit uuid */
595 /*A0*/  struct ocfs2_cluster_info s_cluster_info; /* Selected userspace
596 						     stack.  Only valid
597 						     with INCOMPAT flag. */
598 /*B8*/	__le16 s_xattr_inline_size;	/* extended attribute inline size
599 					   for this fs*/
600 	__le16 s_reserved0;
601 	__le32 s_dx_seed[3];		/* seed[0-2] for dx dir hash.
602 					 * s_uuid_hash serves as seed[3]. */
603 /*C0*/  __le64 s_reserved2[15];		/* Fill out superblock */
604 /*140*/
605 
606 	/*
607 	 * NOTE: As stated above, all offsets are relative to
608 	 * ocfs2_dinode.id2, which is at 0xC0 in the inode.
609 	 * 0xC0 + 0x140 = 0x200 or 512 bytes.  A superblock must fit within
610 	 * our smallest blocksize, which is 512 bytes.  To ensure this,
611 	 * we reserve the space in s_reserved2.  Anything past s_reserved2
612 	 * will not be available on the smallest blocksize.
613 	 */
614 };
615 
616 /*
617  * Local allocation bitmap for OCFS2 slots
618  * Note that it exists inside an ocfs2_dinode, so all offsets are
619  * relative to the start of ocfs2_dinode.id2.
620  */
621 struct ocfs2_local_alloc
622 {
623 /*00*/	__le32 la_bm_off;	/* Starting bit offset in main bitmap */
624 	__le16 la_size;		/* Size of included bitmap, in bytes */
625 	__le16 la_reserved1;
626 	__le64 la_reserved2;
627 /*10*/	__u8   la_bitmap[0];
628 };
629 
630 /*
631  * Data-in-inode header. This is only used if i_dyn_features has
632  * OCFS2_INLINE_DATA_FL set.
633  */
634 struct ocfs2_inline_data
635 {
636 /*00*/	__le16	id_count;	/* Number of bytes that can be used
637 				 * for data, starting at id_data */
638 	__le16	id_reserved0;
639 	__le32	id_reserved1;
640 	__u8	id_data[0];	/* Start of user data */
641 };
642 
643 /*
644  * On disk inode for OCFS2
645  */
646 struct ocfs2_dinode {
647 /*00*/	__u8 i_signature[8];		/* Signature for validation */
648 	__le32 i_generation;		/* Generation number */
649 	__le16 i_suballoc_slot;		/* Slot suballocator this inode
650 					   belongs to */
651 	__le16 i_suballoc_bit;		/* Bit offset in suballocator
652 					   block group */
653 /*10*/	__le16 i_links_count_hi;	/* High 16 bits of links count */
654 	__le16 i_xattr_inline_size;
655 	__le32 i_clusters;		/* Cluster count */
656 	__le32 i_uid;			/* Owner UID */
657 	__le32 i_gid;			/* Owning GID */
658 /*20*/	__le64 i_size;			/* Size in bytes */
659 	__le16 i_mode;			/* File mode */
660 	__le16 i_links_count;		/* Links count */
661 	__le32 i_flags;			/* File flags */
662 /*30*/	__le64 i_atime;			/* Access time */
663 	__le64 i_ctime;			/* Creation time */
664 /*40*/	__le64 i_mtime;			/* Modification time */
665 	__le64 i_dtime;			/* Deletion time */
666 /*50*/	__le64 i_blkno;			/* Offset on disk, in blocks */
667 	__le64 i_last_eb_blk;		/* Pointer to last extent
668 					   block */
669 /*60*/	__le32 i_fs_generation;		/* Generation per fs-instance */
670 	__le32 i_atime_nsec;
671 	__le32 i_ctime_nsec;
672 	__le32 i_mtime_nsec;
673 /*70*/	__le32 i_attr;
674 	__le16 i_orphaned_slot;		/* Only valid when OCFS2_ORPHANED_FL
675 					   was set in i_flags */
676 	__le16 i_dyn_features;
677 	__le64 i_xattr_loc;
678 /*80*/	struct ocfs2_block_check i_check;	/* Error checking */
679 /*88*/	__le64 i_dx_root;		/* Pointer to dir index root block */
680 /*90*/	__le64 i_refcount_loc;
681 	__le64 i_suballoc_loc;		/* Suballocator block group this
682 					   inode belongs to.  Only valid
683 					   if allocated from a
684 					   discontiguous block group */
685 /*A0*/	__le64 i_reserved2[3];
686 /*B8*/	union {
687 		__le64 i_pad1;		/* Generic way to refer to this
688 					   64bit union */
689 		struct {
690 			__le64 i_rdev;	/* Device number */
691 		} dev1;
692 		struct {		/* Info for bitmap system
693 					   inodes */
694 			__le32 i_used;	/* Bits (ie, clusters) used  */
695 			__le32 i_total;	/* Total bits (clusters)
696 					   available */
697 		} bitmap1;
698 		struct {		/* Info for journal system
699 					   inodes */
700 			__le32 ij_flags;	/* Mounted, version, etc. */
701 			__le32 ij_recovery_generation; /* Incremented when the
702 							  journal is recovered
703 							  after an unclean
704 							  shutdown */
705 		} journal1;
706 	} id1;				/* Inode type dependant 1 */
707 /*C0*/	union {
708 		struct ocfs2_super_block	i_super;
709 		struct ocfs2_local_alloc	i_lab;
710 		struct ocfs2_chain_list		i_chain;
711 		struct ocfs2_extent_list	i_list;
712 		struct ocfs2_truncate_log	i_dealloc;
713 		struct ocfs2_inline_data	i_data;
714 		__u8               		i_symlink[0];
715 	} id2;
716 /* Actual on-disk size is one block */
717 };
718 
719 /*
720  * On-disk directory entry structure for OCFS2
721  *
722  * Packed as this structure could be accessed unaligned on 64-bit platforms
723  */
724 struct ocfs2_dir_entry {
725 /*00*/	__le64   inode;                  /* Inode number */
726 	__le16   rec_len;                /* Directory entry length */
727 	__u8    name_len;               /* Name length */
728 	__u8    file_type;
729 /*0C*/	char    name[OCFS2_MAX_FILENAME_LEN];   /* File name */
730 /* Actual on-disk length specified by rec_len */
731 } __attribute__ ((packed));
732 
733 /*
734  * Per-block record for the unindexed directory btree. This is carefully
735  * crafted so that the rec_len and name_len records of an ocfs2_dir_entry are
736  * mirrored. That way, the directory manipulation code needs a minimal amount
737  * of update.
738  *
739  * NOTE: Keep this structure aligned to a multiple of 4 bytes.
740  */
741 struct ocfs2_dir_block_trailer {
742 /*00*/	__le64		db_compat_inode;	/* Always zero. Was inode */
743 
744 	__le16		db_compat_rec_len;	/* Backwards compatible with
745 						 * ocfs2_dir_entry. */
746 	__u8		db_compat_name_len;	/* Always zero. Was name_len */
747 	__u8		db_reserved0;
748 	__le16		db_reserved1;
749 	__le16		db_free_rec_len;	/* Size of largest empty hole
750 						 * in this block. (unused) */
751 /*10*/	__u8		db_signature[8];	/* Signature for verification */
752 	__le64		db_reserved2;
753 	__le64		db_free_next;		/* Next block in list (unused) */
754 /*20*/	__le64		db_blkno;		/* Offset on disk, in blocks */
755 	__le64		db_parent_dinode;	/* dinode which owns me, in
756 						   blocks */
757 /*30*/	struct ocfs2_block_check db_check;	/* Error checking */
758 /*40*/
759 };
760 
761  /*
762  * A directory entry in the indexed tree. We don't store the full name here,
763  * but instead provide a pointer to the full dirent in the unindexed tree.
764  *
765  * We also store name_len here so as to reduce the number of leaf blocks we
766  * need to search in case of collisions.
767  */
768 struct ocfs2_dx_entry {
769 	__le32		dx_major_hash;	/* Used to find logical
770 					 * cluster in index */
771 	__le32		dx_minor_hash;	/* Lower bits used to find
772 					 * block in cluster */
773 	__le64		dx_dirent_blk;	/* Physical block in unindexed
774 					 * tree holding this dirent. */
775 };
776 
777 struct ocfs2_dx_entry_list {
778 	__le32		de_reserved;
779 	__le16		de_count;	/* Maximum number of entries
780 					 * possible in de_entries */
781 	__le16		de_num_used;	/* Current number of
782 					 * de_entries entries */
783 	struct	ocfs2_dx_entry		de_entries[0];	/* Indexed dir entries
784 							 * in a packed array of
785 							 * length de_num_used */
786 };
787 
788 #define OCFS2_DX_FLAG_INLINE	0x01
789 
790 /*
791  * A directory indexing block. Each indexed directory has one of these,
792  * pointed to by ocfs2_dinode.
793  *
794  * This block stores an indexed btree root, and a set of free space
795  * start-of-list pointers.
796  */
797 struct ocfs2_dx_root_block {
798 	__u8		dr_signature[8];	/* Signature for verification */
799 	struct ocfs2_block_check dr_check;	/* Error checking */
800 	__le16		dr_suballoc_slot;	/* Slot suballocator this
801 						 * block belongs to. */
802 	__le16		dr_suballoc_bit;	/* Bit offset in suballocator
803 						 * block group */
804 	__le32		dr_fs_generation;	/* Must match super block */
805 	__le64		dr_blkno;		/* Offset on disk, in blocks */
806 	__le64		dr_last_eb_blk;		/* Pointer to last
807 						 * extent block */
808 	__le32		dr_clusters;		/* Clusters allocated
809 						 * to the indexed tree. */
810 	__u8		dr_flags;		/* OCFS2_DX_FLAG_* flags */
811 	__u8		dr_reserved0;
812 	__le16		dr_reserved1;
813 	__le64		dr_dir_blkno;		/* Pointer to parent inode */
814 	__le32		dr_num_entries;		/* Total number of
815 						 * names stored in
816 						 * this directory.*/
817 	__le32		dr_reserved2;
818 	__le64		dr_free_blk;		/* Pointer to head of free
819 						 * unindexed block list. */
820 	__le64		dr_suballoc_loc;	/* Suballocator block group
821 						   this root belongs to.
822 						   Only valid if allocated
823 						   from a discontiguous
824 						   block group */
825 	__le64		dr_reserved3[14];
826 	union {
827 		struct ocfs2_extent_list dr_list; /* Keep this aligned to 128
828 						   * bits for maximum space
829 						   * efficiency. */
830 		struct ocfs2_dx_entry_list dr_entries; /* In-root-block list of
831 							* entries. We grow out
832 							* to extents if this
833 							* gets too big. */
834 	};
835 };
836 
837 /*
838  * The header of a leaf block in the indexed tree.
839  */
840 struct ocfs2_dx_leaf {
841 	__u8		dl_signature[8];/* Signature for verification */
842 	struct ocfs2_block_check dl_check;	/* Error checking */
843 	__le64		dl_blkno;	/* Offset on disk, in blocks */
844 	__le32		dl_fs_generation;/* Must match super block */
845 	__le32		dl_reserved0;
846 	__le64		dl_reserved1;
847 	struct ocfs2_dx_entry_list	dl_list;
848 };
849 
850 /*
851  * Largest bitmap for a block (suballocator) group in bytes.  This limit
852  * does not affect cluster groups (global allocator).  Cluster group
853  * bitmaps run to the end of the block.
854  */
855 #define OCFS2_MAX_BG_BITMAP_SIZE	256
856 
857 /*
858  * On disk allocator group structure for OCFS2
859  */
860 struct ocfs2_group_desc
861 {
862 /*00*/	__u8    bg_signature[8];        /* Signature for validation */
863 	__le16   bg_size;                /* Size of included bitmap in
864 					   bytes. */
865 	__le16   bg_bits;                /* Bits represented by this
866 					   group. */
867 	__le16	bg_free_bits_count;     /* Free bits count */
868 	__le16   bg_chain;               /* What chain I am in. */
869 /*10*/	__le32   bg_generation;
870 	__le32	bg_reserved1;
871 	__le64   bg_next_group;          /* Next group in my list, in
872 					   blocks */
873 /*20*/	__le64   bg_parent_dinode;       /* dinode which owns me, in
874 					   blocks */
875 	__le64   bg_blkno;               /* Offset on disk, in blocks */
876 /*30*/	struct ocfs2_block_check bg_check;	/* Error checking */
877 	__le64   bg_reserved2;
878 /*40*/	union {
879 		__u8    bg_bitmap[0];
880 		struct {
881 			/*
882 			 * Block groups may be discontiguous when
883 			 * OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG is set.
884 			 * The extents of a discontigous block group are
885 			 * stored in bg_list.  It is a flat list.
886 			 * l_tree_depth must always be zero.  A
887 			 * discontiguous group is signified by a non-zero
888 			 * bg_list->l_next_free_rec.  Only block groups
889 			 * can be discontiguous; Cluster groups cannot.
890 			 * We've never made a block group with more than
891 			 * 2048 blocks (256 bytes of bg_bitmap).  This
892 			 * codifies that limit so that we can fit bg_list.
893 			 * bg_size of a discontiguous block group will
894 			 * be 256 to match bg_bitmap_filler.
895 			 */
896 			__u8 bg_bitmap_filler[OCFS2_MAX_BG_BITMAP_SIZE];
897 /*140*/			struct ocfs2_extent_list bg_list;
898 		};
899 	};
900 /* Actual on-disk size is one block */
901 };
902 
903 struct ocfs2_refcount_rec {
904 /*00*/	__le64 r_cpos;		/* Physical offset, in clusters */
905 	__le32 r_clusters;	/* Clusters covered by this extent */
906 	__le32 r_refcount;	/* Reference count of this extent */
907 /*10*/
908 };
909 #define OCFS2_32BIT_POS_MASK		(0xffffffffULL)
910 
911 #define OCFS2_REFCOUNT_LEAF_FL          (0x00000001)
912 #define OCFS2_REFCOUNT_TREE_FL          (0x00000002)
913 
914 struct ocfs2_refcount_list {
915 /*00*/	__le16 rl_count;	/* Maximum number of entries possible
916 				   in rl_records */
917 	__le16 rl_used;		/* Current number of used records */
918 	__le32 rl_reserved2;
919 	__le64 rl_reserved1;	/* Pad to sizeof(ocfs2_refcount_record) */
920 /*10*/	struct ocfs2_refcount_rec rl_recs[0];	/* Refcount records */
921 };
922 
923 
924 struct ocfs2_refcount_block {
925 /*00*/	__u8 rf_signature[8];		/* Signature for verification */
926 	__le16 rf_suballoc_slot;	/* Slot suballocator this block
927 					   belongs to */
928 	__le16 rf_suballoc_bit;		/* Bit offset in suballocator
929 					   block group */
930 	__le32 rf_fs_generation;	/* Must match superblock */
931 /*10*/	__le64 rf_blkno;		/* Offset on disk, in blocks */
932 	__le64 rf_parent;		/* Parent block, only valid if
933 					   OCFS2_REFCOUNT_LEAF_FL is set in
934 					   rf_flags */
935 /*20*/	struct ocfs2_block_check rf_check;	/* Error checking */
936 	__le64 rf_last_eb_blk;		/* Pointer to last extent block */
937 /*30*/	__le32 rf_count;		/* Number of inodes sharing this
938 					   refcount tree */
939 	__le32 rf_flags;		/* See the flags above */
940 	__le32 rf_clusters;		/* clusters covered by refcount tree. */
941 	__le32 rf_cpos;			/* cluster offset in refcount tree.*/
942 /*40*/	__le32 rf_generation;		/* generation number. all be the same
943 					 * for the same refcount tree. */
944 	__le32 rf_reserved0;
945 	__le64 rf_suballoc_loc;		/* Suballocator block group this
946 					   refcount block belongs to. Only
947 					   valid if allocated from a
948 					   discontiguous block group */
949 /*50*/	__le64 rf_reserved1[6];
950 /*80*/	union {
951 		struct ocfs2_refcount_list rf_records;  /* List of refcount
952 							  records */
953 		struct ocfs2_extent_list rf_list;	/* Extent record list,
954 							only valid if
955 							OCFS2_REFCOUNT_TREE_FL
956 							is set in rf_flags */
957 	};
958 /* Actual on-disk size is one block */
959 };
960 
961 /*
962  * On disk extended attribute structure for OCFS2.
963  */
964 
965 /*
966  * ocfs2_xattr_entry indicates one extend attribute.
967  *
968  * Note that it can be stored in inode, one block or one xattr bucket.
969  */
970 struct ocfs2_xattr_entry {
971 	__le32	xe_name_hash;    /* hash value of xattr prefix+suffix. */
972 	__le16	xe_name_offset;  /* byte offset from the 1st entry in the
973 				    local xattr storage(inode, xattr block or
974 				    xattr bucket). */
975 	__u8	xe_name_len;	 /* xattr name len, does't include prefix. */
976 	__u8	xe_type;         /* the low 7 bits indicate the name prefix
977 				  * type and the highest bit indicates whether
978 				  * the EA is stored in the local storage. */
979 	__le64	xe_value_size;	 /* real xattr value length. */
980 };
981 
982 /*
983  * On disk structure for xattr header.
984  *
985  * One ocfs2_xattr_header describes how many ocfs2_xattr_entry records in
986  * the local xattr storage.
987  */
988 struct ocfs2_xattr_header {
989 	__le16	xh_count;                       /* contains the count of how
990 						   many records are in the
991 						   local xattr storage. */
992 	__le16	xh_free_start;                  /* current offset for storing
993 						   xattr. */
994 	__le16	xh_name_value_len;              /* total length of name/value
995 						   length in this bucket. */
996 	__le16	xh_num_buckets;                 /* Number of xattr buckets
997 						   in this extent record,
998 						   only valid in the first
999 						   bucket. */
1000 	struct ocfs2_block_check xh_check;	/* Error checking
1001 						   (Note, this is only
1002 						    used for xattr
1003 						    buckets.  A block uses
1004 						    xb_check and sets
1005 						    this field to zero.) */
1006 	struct ocfs2_xattr_entry xh_entries[0]; /* xattr entry list. */
1007 };
1008 
1009 /*
1010  * On disk structure for xattr value root.
1011  *
1012  * When an xattr's value is large enough, it is stored in an external
1013  * b-tree like file data.  The xattr value root points to this structure.
1014  */
1015 struct ocfs2_xattr_value_root {
1016 /*00*/	__le32	xr_clusters;              /* clusters covered by xattr value. */
1017 	__le32	xr_reserved0;
1018 	__le64	xr_last_eb_blk;           /* Pointer to last extent block */
1019 /*10*/	struct ocfs2_extent_list xr_list; /* Extent record list */
1020 };
1021 
1022 /*
1023  * On disk structure for xattr tree root.
1024  *
1025  * It is used when there are too many extended attributes for one file. These
1026  * attributes will be organized and stored in an indexed-btree.
1027  */
1028 struct ocfs2_xattr_tree_root {
1029 /*00*/	__le32	xt_clusters;              /* clusters covered by xattr. */
1030 	__le32	xt_reserved0;
1031 	__le64	xt_last_eb_blk;           /* Pointer to last extent block */
1032 /*10*/	struct ocfs2_extent_list xt_list; /* Extent record list */
1033 };
1034 
1035 #define OCFS2_XATTR_INDEXED	0x1
1036 #define OCFS2_HASH_SHIFT	5
1037 #define OCFS2_XATTR_ROUND	3
1038 #define OCFS2_XATTR_SIZE(size)	(((size) + OCFS2_XATTR_ROUND) & \
1039 				~(OCFS2_XATTR_ROUND))
1040 
1041 #define OCFS2_XATTR_BUCKET_SIZE			4096
1042 #define OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET 	(OCFS2_XATTR_BUCKET_SIZE \
1043 						 / OCFS2_MIN_BLOCKSIZE)
1044 
1045 /*
1046  * On disk structure for xattr block.
1047  */
1048 struct ocfs2_xattr_block {
1049 /*00*/	__u8	xb_signature[8];     /* Signature for verification */
1050 	__le16	xb_suballoc_slot;    /* Slot suballocator this
1051 					block belongs to. */
1052 	__le16	xb_suballoc_bit;     /* Bit offset in suballocator
1053 					block group */
1054 	__le32	xb_fs_generation;    /* Must match super block */
1055 /*10*/	__le64	xb_blkno;            /* Offset on disk, in blocks */
1056 	struct ocfs2_block_check xb_check;	/* Error checking */
1057 /*20*/	__le16	xb_flags;            /* Indicates whether this block contains
1058 					real xattr or a xattr tree. */
1059 	__le16	xb_reserved0;
1060 	__le32  xb_reserved1;
1061 	__le64	xb_suballoc_loc;	/* Suballocator block group this
1062 					   xattr block belongs to. Only
1063 					   valid if allocated from a
1064 					   discontiguous block group */
1065 /*30*/	union {
1066 		struct ocfs2_xattr_header xb_header; /* xattr header if this
1067 							block contains xattr */
1068 		struct ocfs2_xattr_tree_root xb_root;/* xattr tree root if this
1069 							block cotains xattr
1070 							tree. */
1071 	} xb_attrs;
1072 };
1073 
1074 #define OCFS2_XATTR_ENTRY_LOCAL		0x80
1075 #define OCFS2_XATTR_TYPE_MASK		0x7F
1076 static inline void ocfs2_xattr_set_local(struct ocfs2_xattr_entry *xe,
1077 					 int local)
1078 {
1079 	if (local)
1080 		xe->xe_type |= OCFS2_XATTR_ENTRY_LOCAL;
1081 	else
1082 		xe->xe_type &= ~OCFS2_XATTR_ENTRY_LOCAL;
1083 }
1084 
1085 static inline int ocfs2_xattr_is_local(struct ocfs2_xattr_entry *xe)
1086 {
1087 	return xe->xe_type & OCFS2_XATTR_ENTRY_LOCAL;
1088 }
1089 
1090 static inline void ocfs2_xattr_set_type(struct ocfs2_xattr_entry *xe, int type)
1091 {
1092 	xe->xe_type |= type & OCFS2_XATTR_TYPE_MASK;
1093 }
1094 
1095 static inline int ocfs2_xattr_get_type(struct ocfs2_xattr_entry *xe)
1096 {
1097 	return xe->xe_type & OCFS2_XATTR_TYPE_MASK;
1098 }
1099 
1100 /*
1101  *  On disk structures for global quota file
1102  */
1103 
1104 /* Magic numbers and known versions for global quota files */
1105 #define OCFS2_GLOBAL_QMAGICS {\
1106 	0x0cf52470, /* USRQUOTA */ \
1107 	0x0cf52471  /* GRPQUOTA */ \
1108 }
1109 
1110 #define OCFS2_GLOBAL_QVERSIONS {\
1111 	0, \
1112 	0, \
1113 }
1114 
1115 
1116 /* Each block of each quota file has a certain fixed number of bytes reserved
1117  * for OCFS2 internal use at its end. OCFS2 can use it for things like
1118  * checksums, etc. */
1119 #define OCFS2_QBLK_RESERVED_SPACE 8
1120 
1121 /* Generic header of all quota files */
1122 struct ocfs2_disk_dqheader {
1123 	__le32 dqh_magic;	/* Magic number identifying file */
1124 	__le32 dqh_version;	/* Quota format version */
1125 };
1126 
1127 #define OCFS2_GLOBAL_INFO_OFF (sizeof(struct ocfs2_disk_dqheader))
1128 
1129 /* Information header of global quota file (immediately follows the generic
1130  * header) */
1131 struct ocfs2_global_disk_dqinfo {
1132 /*00*/	__le32 dqi_bgrace;	/* Grace time for space softlimit excess */
1133 	__le32 dqi_igrace;	/* Grace time for inode softlimit excess */
1134 	__le32 dqi_syncms;	/* Time after which we sync local changes to
1135 				 * global quota file */
1136 	__le32 dqi_blocks;	/* Number of blocks in quota file */
1137 /*10*/	__le32 dqi_free_blk;	/* First free block in quota file */
1138 	__le32 dqi_free_entry;	/* First block with free dquot entry in quota
1139 				 * file */
1140 };
1141 
1142 /* Structure with global user / group information. We reserve some space
1143  * for future use. */
1144 struct ocfs2_global_disk_dqblk {
1145 /*00*/	__le32 dqb_id;          /* ID the structure belongs to */
1146 	__le32 dqb_use_count;   /* Number of nodes having reference to this structure */
1147 	__le64 dqb_ihardlimit;  /* absolute limit on allocated inodes */
1148 /*10*/	__le64 dqb_isoftlimit;  /* preferred inode limit */
1149 	__le64 dqb_curinodes;   /* current # allocated inodes */
1150 /*20*/	__le64 dqb_bhardlimit;  /* absolute limit on disk space */
1151 	__le64 dqb_bsoftlimit;  /* preferred limit on disk space */
1152 /*30*/	__le64 dqb_curspace;    /* current space occupied */
1153 	__le64 dqb_btime;       /* time limit for excessive disk use */
1154 /*40*/	__le64 dqb_itime;       /* time limit for excessive inode use */
1155 	__le64 dqb_pad1;
1156 /*50*/	__le64 dqb_pad2;
1157 };
1158 
1159 /*
1160  *  On-disk structures for local quota file
1161  */
1162 
1163 /* Magic numbers and known versions for local quota files */
1164 #define OCFS2_LOCAL_QMAGICS {\
1165 	0x0cf524c0, /* USRQUOTA */ \
1166 	0x0cf524c1  /* GRPQUOTA */ \
1167 }
1168 
1169 #define OCFS2_LOCAL_QVERSIONS {\
1170 	0, \
1171 	0, \
1172 }
1173 
1174 /* Quota flags in dqinfo header */
1175 #define OLQF_CLEAN	0x0001	/* Quota file is empty (this should be after\
1176 				 * quota has been cleanly turned off) */
1177 
1178 #define OCFS2_LOCAL_INFO_OFF (sizeof(struct ocfs2_disk_dqheader))
1179 
1180 /* Information header of local quota file (immediately follows the generic
1181  * header) */
1182 struct ocfs2_local_disk_dqinfo {
1183 	__le32 dqi_flags;	/* Flags for quota file */
1184 	__le32 dqi_chunks;	/* Number of chunks of quota structures
1185 				 * with a bitmap */
1186 	__le32 dqi_blocks;	/* Number of blocks allocated for quota file */
1187 };
1188 
1189 /* Header of one chunk of a quota file */
1190 struct ocfs2_local_disk_chunk {
1191 	__le32 dqc_free;	/* Number of free entries in the bitmap */
1192 	__u8 dqc_bitmap[0];	/* Bitmap of entries in the corresponding
1193 				 * chunk of quota file */
1194 };
1195 
1196 /* One entry in local quota file */
1197 struct ocfs2_local_disk_dqblk {
1198 /*00*/	__le64 dqb_id;		/* id this quota applies to */
1199 	__le64 dqb_spacemod;	/* Change in the amount of used space */
1200 /*10*/	__le64 dqb_inodemod;	/* Change in the amount of used inodes */
1201 };
1202 
1203 
1204 /*
1205  * The quota trailer lives at the end of each quota block.
1206  */
1207 
1208 struct ocfs2_disk_dqtrailer {
1209 /*00*/	struct ocfs2_block_check dq_check;	/* Error checking */
1210 /*08*/	/* Cannot be larger than OCFS2_QBLK_RESERVED_SPACE */
1211 };
1212 
1213 static inline struct ocfs2_disk_dqtrailer *ocfs2_block_dqtrailer(int blocksize,
1214 								 void *buf)
1215 {
1216 	char *ptr = buf;
1217 	ptr += blocksize - OCFS2_QBLK_RESERVED_SPACE;
1218 
1219 	return (struct ocfs2_disk_dqtrailer *)ptr;
1220 }
1221 
1222 #ifdef __KERNEL__
1223 static inline int ocfs2_fast_symlink_chars(struct super_block *sb)
1224 {
1225 	return  sb->s_blocksize -
1226 		 offsetof(struct ocfs2_dinode, id2.i_symlink);
1227 }
1228 
1229 static inline int ocfs2_max_inline_data_with_xattr(struct super_block *sb,
1230 						   struct ocfs2_dinode *di)
1231 {
1232 	unsigned int xattrsize = le16_to_cpu(di->i_xattr_inline_size);
1233 
1234 	if (le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_XATTR_FL)
1235 		return sb->s_blocksize -
1236 			offsetof(struct ocfs2_dinode, id2.i_data.id_data) -
1237 			xattrsize;
1238 	else
1239 		return sb->s_blocksize -
1240 			offsetof(struct ocfs2_dinode, id2.i_data.id_data);
1241 }
1242 
1243 static inline int ocfs2_extent_recs_per_inode(struct super_block *sb)
1244 {
1245 	int size;
1246 
1247 	size = sb->s_blocksize -
1248 		offsetof(struct ocfs2_dinode, id2.i_list.l_recs);
1249 
1250 	return size / sizeof(struct ocfs2_extent_rec);
1251 }
1252 
1253 static inline int ocfs2_extent_recs_per_inode_with_xattr(
1254 						struct super_block *sb,
1255 						struct ocfs2_dinode *di)
1256 {
1257 	int size;
1258 	unsigned int xattrsize = le16_to_cpu(di->i_xattr_inline_size);
1259 
1260 	if (le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_XATTR_FL)
1261 		size = sb->s_blocksize -
1262 			offsetof(struct ocfs2_dinode, id2.i_list.l_recs) -
1263 			xattrsize;
1264 	else
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_dx_root(struct super_block *sb)
1272 {
1273 	int size;
1274 
1275 	size = sb->s_blocksize -
1276 		offsetof(struct ocfs2_dx_root_block, dr_list.l_recs);
1277 
1278 	return size / sizeof(struct ocfs2_extent_rec);
1279 }
1280 
1281 static inline int ocfs2_chain_recs_per_inode(struct super_block *sb)
1282 {
1283 	int size;
1284 
1285 	size = sb->s_blocksize -
1286 		offsetof(struct ocfs2_dinode, id2.i_chain.cl_recs);
1287 
1288 	return size / sizeof(struct ocfs2_chain_rec);
1289 }
1290 
1291 static inline u16 ocfs2_extent_recs_per_eb(struct super_block *sb)
1292 {
1293 	int size;
1294 
1295 	size = sb->s_blocksize -
1296 		offsetof(struct ocfs2_extent_block, h_list.l_recs);
1297 
1298 	return size / sizeof(struct ocfs2_extent_rec);
1299 }
1300 
1301 static inline u16 ocfs2_extent_recs_per_gd(struct super_block *sb)
1302 {
1303 	int size;
1304 
1305 	size = sb->s_blocksize -
1306 		offsetof(struct ocfs2_group_desc, bg_list.l_recs);
1307 
1308 	return size / sizeof(struct ocfs2_extent_rec);
1309 }
1310 
1311 static inline int ocfs2_dx_entries_per_leaf(struct super_block *sb)
1312 {
1313 	int size;
1314 
1315 	size = sb->s_blocksize -
1316 		offsetof(struct ocfs2_dx_leaf, dl_list.de_entries);
1317 
1318 	return size / sizeof(struct ocfs2_dx_entry);
1319 }
1320 
1321 static inline int ocfs2_dx_entries_per_root(struct super_block *sb)
1322 {
1323 	int size;
1324 
1325 	size = sb->s_blocksize -
1326 		offsetof(struct ocfs2_dx_root_block, dr_entries.de_entries);
1327 
1328 	return size / sizeof(struct ocfs2_dx_entry);
1329 }
1330 
1331 static inline u16 ocfs2_local_alloc_size(struct super_block *sb)
1332 {
1333 	u16 size;
1334 
1335 	size = sb->s_blocksize -
1336 		offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap);
1337 
1338 	return size;
1339 }
1340 
1341 static inline int ocfs2_group_bitmap_size(struct super_block *sb,
1342 					  int suballocator,
1343 					  u32 feature_incompat)
1344 {
1345 	int size = sb->s_blocksize -
1346 		offsetof(struct ocfs2_group_desc, bg_bitmap);
1347 
1348 	/*
1349 	 * The cluster allocator uses the entire block.  Suballocators have
1350 	 * never used more than OCFS2_MAX_BG_BITMAP_SIZE.  Unfortunately, older
1351 	 * code expects bg_size set to the maximum.  Thus we must keep
1352 	 * bg_size as-is unless discontig_bg is enabled.
1353 	 */
1354 	if (suballocator &&
1355 	    (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG))
1356 		size = OCFS2_MAX_BG_BITMAP_SIZE;
1357 
1358 	return size;
1359 }
1360 
1361 static inline int ocfs2_truncate_recs_per_inode(struct super_block *sb)
1362 {
1363 	int size;
1364 
1365 	size = sb->s_blocksize -
1366 		offsetof(struct ocfs2_dinode, id2.i_dealloc.tl_recs);
1367 
1368 	return size / sizeof(struct ocfs2_truncate_rec);
1369 }
1370 
1371 static inline u64 ocfs2_backup_super_blkno(struct super_block *sb, int index)
1372 {
1373 	u64 offset = OCFS2_BACKUP_SB_START;
1374 
1375 	if (index >= 0 && index < OCFS2_MAX_BACKUP_SUPERBLOCKS) {
1376 		offset <<= (2 * index);
1377 		offset >>= sb->s_blocksize_bits;
1378 		return offset;
1379 	}
1380 
1381 	return 0;
1382 
1383 }
1384 
1385 static inline u16 ocfs2_xattr_recs_per_xb(struct super_block *sb)
1386 {
1387 	int size;
1388 
1389 	size = sb->s_blocksize -
1390 		offsetof(struct ocfs2_xattr_block,
1391 			 xb_attrs.xb_root.xt_list.l_recs);
1392 
1393 	return size / sizeof(struct ocfs2_extent_rec);
1394 }
1395 
1396 static inline u16 ocfs2_extent_recs_per_rb(struct super_block *sb)
1397 {
1398 	int size;
1399 
1400 	size = sb->s_blocksize -
1401 		offsetof(struct ocfs2_refcount_block, rf_list.l_recs);
1402 
1403 	return size / sizeof(struct ocfs2_extent_rec);
1404 }
1405 
1406 static inline u16 ocfs2_refcount_recs_per_rb(struct super_block *sb)
1407 {
1408 	int size;
1409 
1410 	size = sb->s_blocksize -
1411 		offsetof(struct ocfs2_refcount_block, rf_records.rl_recs);
1412 
1413 	return size / sizeof(struct ocfs2_refcount_rec);
1414 }
1415 
1416 static inline u32
1417 ocfs2_get_ref_rec_low_cpos(const struct ocfs2_refcount_rec *rec)
1418 {
1419 	return le64_to_cpu(rec->r_cpos) & OCFS2_32BIT_POS_MASK;
1420 }
1421 #else
1422 static inline int ocfs2_fast_symlink_chars(int blocksize)
1423 {
1424 	return blocksize - offsetof(struct ocfs2_dinode, id2.i_symlink);
1425 }
1426 
1427 static inline int ocfs2_max_inline_data_with_xattr(int blocksize,
1428 						   struct ocfs2_dinode *di)
1429 {
1430 	if (di && (di->i_dyn_features & OCFS2_INLINE_XATTR_FL))
1431 		return blocksize -
1432 			offsetof(struct ocfs2_dinode, id2.i_data.id_data) -
1433 			di->i_xattr_inline_size;
1434 	else
1435 		return blocksize -
1436 			offsetof(struct ocfs2_dinode, id2.i_data.id_data);
1437 }
1438 
1439 static inline int ocfs2_extent_recs_per_inode(int blocksize)
1440 {
1441 	int size;
1442 
1443 	size = blocksize -
1444 		offsetof(struct ocfs2_dinode, id2.i_list.l_recs);
1445 
1446 	return size / sizeof(struct ocfs2_extent_rec);
1447 }
1448 
1449 static inline int ocfs2_chain_recs_per_inode(int blocksize)
1450 {
1451 	int size;
1452 
1453 	size = blocksize -
1454 		offsetof(struct ocfs2_dinode, id2.i_chain.cl_recs);
1455 
1456 	return size / sizeof(struct ocfs2_chain_rec);
1457 }
1458 
1459 static inline int ocfs2_extent_recs_per_eb(int blocksize)
1460 {
1461 	int size;
1462 
1463 	size = blocksize -
1464 		offsetof(struct ocfs2_extent_block, h_list.l_recs);
1465 
1466 	return size / sizeof(struct ocfs2_extent_rec);
1467 }
1468 
1469 static inline int ocfs2_extent_recs_per_gd(int blocksize)
1470 {
1471 	int size;
1472 
1473 	size = blocksize -
1474 		offsetof(struct ocfs2_group_desc, bg_list.l_recs);
1475 
1476 	return size / sizeof(struct ocfs2_extent_rec);
1477 }
1478 
1479 static inline int ocfs2_local_alloc_size(int blocksize)
1480 {
1481 	int size;
1482 
1483 	size = blocksize -
1484 		offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap);
1485 
1486 	return size;
1487 }
1488 
1489 static inline int ocfs2_group_bitmap_size(int blocksize,
1490 					  int suballocator,
1491 					  uint32_t feature_incompat)
1492 {
1493 	int size = sb->s_blocksize -
1494 		offsetof(struct ocfs2_group_desc, bg_bitmap);
1495 
1496 	/*
1497 	 * The cluster allocator uses the entire block.  Suballocators have
1498 	 * never used more than OCFS2_MAX_BG_BITMAP_SIZE.  Unfortunately, older
1499 	 * code expects bg_size set to the maximum.  Thus we must keep
1500 	 * bg_size as-is unless discontig_bg is enabled.
1501 	 */
1502 	if (suballocator &&
1503 	    (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG))
1504 		size = OCFS2_MAX_BG_BITMAP_SIZE;
1505 
1506 	return size;
1507 }
1508 
1509 static inline int ocfs2_truncate_recs_per_inode(int blocksize)
1510 {
1511 	int size;
1512 
1513 	size = blocksize -
1514 		offsetof(struct ocfs2_dinode, id2.i_dealloc.tl_recs);
1515 
1516 	return size / sizeof(struct ocfs2_truncate_rec);
1517 }
1518 
1519 static inline uint64_t ocfs2_backup_super_blkno(int blocksize, int index)
1520 {
1521 	uint64_t offset = OCFS2_BACKUP_SB_START;
1522 
1523 	if (index >= 0 && index < OCFS2_MAX_BACKUP_SUPERBLOCKS) {
1524 		offset <<= (2 * index);
1525 		offset /= blocksize;
1526 		return offset;
1527 	}
1528 
1529 	return 0;
1530 }
1531 
1532 static inline int ocfs2_xattr_recs_per_xb(int blocksize)
1533 {
1534 	int size;
1535 
1536 	size = blocksize -
1537 		offsetof(struct ocfs2_xattr_block,
1538 			 xb_attrs.xb_root.xt_list.l_recs);
1539 
1540 	return size / sizeof(struct ocfs2_extent_rec);
1541 }
1542 #endif  /* __KERNEL__ */
1543 
1544 
1545 static inline int ocfs2_system_inode_is_global(int type)
1546 {
1547 	return ((type >= 0) &&
1548 		(type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE));
1549 }
1550 
1551 static inline int ocfs2_sprintf_system_inode_name(char *buf, int len,
1552 						  int type, int slot)
1553 {
1554 	int chars;
1555 
1556         /*
1557          * Global system inodes can only have one copy.  Everything
1558          * after OCFS2_LAST_GLOBAL_SYSTEM_INODE in the system inode
1559          * list has a copy per slot.
1560          */
1561 	if (type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE)
1562 		chars = snprintf(buf, len, "%s",
1563 				 ocfs2_system_inodes[type].si_name);
1564 	else
1565 		chars = snprintf(buf, len,
1566 				 ocfs2_system_inodes[type].si_name,
1567 				 slot);
1568 
1569 	return chars;
1570 }
1571 
1572 static inline void ocfs2_set_de_type(struct ocfs2_dir_entry *de,
1573 				    umode_t mode)
1574 {
1575 	de->file_type = ocfs2_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
1576 }
1577 
1578 static inline int ocfs2_gd_is_discontig(struct ocfs2_group_desc *gd)
1579 {
1580 	if ((offsetof(struct ocfs2_group_desc, bg_bitmap) +
1581 	     le16_to_cpu(gd->bg_size)) !=
1582 	    offsetof(struct ocfs2_group_desc, bg_list))
1583 		return 0;
1584 	/*
1585 	 * Only valid to check l_next_free_rec if
1586 	 * bg_bitmap + bg_size == bg_list.
1587 	 */
1588 	if (!gd->bg_list.l_next_free_rec)
1589 		return 0;
1590 	return 1;
1591 }
1592 #endif  /* _OCFS2_FS_H */
1593 
1594