2 md_p.h : physical layout of Linux RAID devices
3 Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman
5 This program is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation; either version 2, or (at your option)
10 You should have received a copy of the GNU General Public License
11 (for example /usr/src/linux/COPYING); if not, write to the Free
12 Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
18 #include <linux/types.h>
19 #include <asm/byteorder.h>
24 * The RAID superblock maintains some statistics on each RAID configuration.
25 * Each real device in the RAID set contains it near the end of the device.
26 * Some of the ideas are copied from the ext2fs implementation.
28 * We currently use 4096 bytes as follows:
30 * word offset function
32 * 0 - 31 Constant generic RAID device information.
33 * 32 - 63 Generic state information.
34 * 64 - 127 Personality specific information.
35 * 128 - 511 12 32-words descriptors of the disks in the raid set.
37 * 912 - 1023 Disk specific descriptor.
41 * If x is the real device size in bytes, we return an apparent size of:
43 * y = (x & ~(MD_RESERVED_BYTES - 1)) - MD_RESERVED_BYTES
45 * and place the 4kB superblock at offset y.
47 #define MD_RESERVED_BYTES (64 * 1024)
48 #define MD_RESERVED_SECTORS (MD_RESERVED_BYTES / 512)
50 #define MD_NEW_SIZE_SECTORS(x) ((x & ~(MD_RESERVED_SECTORS - 1)) - MD_RESERVED_SECTORS)
52 #define MD_SB_BYTES 4096
53 #define MD_SB_WORDS (MD_SB_BYTES / 4)
54 #define MD_SB_SECTORS (MD_SB_BYTES / 512)
57 * The following are counted in 32-bit words
59 #define MD_SB_GENERIC_OFFSET 0
60 #define MD_SB_PERSONALITY_OFFSET 64
61 #define MD_SB_DISKS_OFFSET 128
62 #define MD_SB_DESCRIPTOR_OFFSET 992
64 #define MD_SB_GENERIC_CONSTANT_WORDS 32
65 #define MD_SB_GENERIC_STATE_WORDS 32
66 #define MD_SB_GENERIC_WORDS (MD_SB_GENERIC_CONSTANT_WORDS + MD_SB_GENERIC_STATE_WORDS)
67 #define MD_SB_PERSONALITY_WORDS 64
68 #define MD_SB_DESCRIPTOR_WORDS 32
69 #define MD_SB_DISKS 27
70 #define MD_SB_DISKS_WORDS (MD_SB_DISKS*MD_SB_DESCRIPTOR_WORDS)
71 #define MD_SB_RESERVED_WORDS (1024 - MD_SB_GENERIC_WORDS - MD_SB_PERSONALITY_WORDS - MD_SB_DISKS_WORDS - MD_SB_DESCRIPTOR_WORDS)
72 #define MD_SB_EQUAL_WORDS (MD_SB_GENERIC_WORDS + MD_SB_PERSONALITY_WORDS + MD_SB_DISKS_WORDS)
75 * Device "operational" state bits
77 #define MD_DISK_FAULTY 0 /* disk is faulty / operational */
78 #define MD_DISK_ACTIVE 1 /* disk is running or spare disk */
79 #define MD_DISK_SYNC 2 /* disk is in sync with the raid set */
80 #define MD_DISK_REMOVED 3 /* disk is in sync with the raid set */
81 #define MD_DISK_CLUSTER_ADD 4 /* Initiate a disk add across the cluster
82 * For clustered enviroments only.
84 #define MD_DISK_CANDIDATE 5 /* disk is added as spare (local) until confirmed
85 * For clustered enviroments only.
88 #define MD_DISK_WRITEMOSTLY 9 /* disk is "write-mostly" is RAID1 config.
89 * read requests will only be sent here in
93 typedef struct mdp_device_descriptor_s {
94 __u32 number; /* 0 Device number in the entire set */
95 __u32 major; /* 1 Device major number */
96 __u32 minor; /* 2 Device minor number */
97 __u32 raid_disk; /* 3 The role of the device in the raid set */
98 __u32 state; /* 4 Operational state */
99 __u32 reserved[MD_SB_DESCRIPTOR_WORDS - 5];
102 #define MD_SB_MAGIC 0xa92b4efc
105 * Superblock state bits
107 #define MD_SB_CLEAN 0
108 #define MD_SB_ERRORS 1
110 #define MD_SB_CLUSTERED 5 /* MD is clustered */
111 #define MD_SB_BITMAP_PRESENT 8 /* bitmap may be present nearby */
115 * - if an array is being reshaped (restriped) in order to change the
116 * the number of active devices in the array, 'raid_disks' will be
117 * the larger of the old and new numbers. 'delta_disks' will
118 * be the "new - old". So if +ve, raid_disks is the new value, and
119 * "raid_disks-delta_disks" is the old. If -ve, raid_disks is the
120 * old value and "raid_disks+delta_disks" is the new (smaller) value.
124 typedef struct mdp_superblock_s {
126 * Constant generic information
128 __u32 md_magic; /* 0 MD identifier */
129 __u32 major_version; /* 1 major version to which the set conforms */
130 __u32 minor_version; /* 2 minor version ... */
131 __u32 patch_version; /* 3 patchlevel version ... */
132 __u32 gvalid_words; /* 4 Number of used words in this section */
133 __u32 set_uuid0; /* 5 Raid set identifier */
134 __u32 ctime; /* 6 Creation time */
135 __u32 level; /* 7 Raid personality */
136 __u32 size; /* 8 Apparent size of each individual disk */
137 __u32 nr_disks; /* 9 total disks in the raid set */
138 __u32 raid_disks; /* 10 disks in a fully functional raid set */
139 __u32 md_minor; /* 11 preferred MD minor device number */
140 __u32 not_persistent; /* 12 does it have a persistent superblock */
141 __u32 set_uuid1; /* 13 Raid set identifier #2 */
142 __u32 set_uuid2; /* 14 Raid set identifier #3 */
143 __u32 set_uuid3; /* 15 Raid set identifier #4 */
144 __u32 gstate_creserved[MD_SB_GENERIC_CONSTANT_WORDS - 16];
147 * Generic state information
149 __u32 utime; /* 0 Superblock update time */
150 __u32 state; /* 1 State bits (clean, ...) */
151 __u32 active_disks; /* 2 Number of currently active disks */
152 __u32 working_disks; /* 3 Number of working disks */
153 __u32 failed_disks; /* 4 Number of failed disks */
154 __u32 spare_disks; /* 5 Number of spare disks */
155 __u32 sb_csum; /* 6 checksum of the whole superblock */
156 #if defined(__BYTE_ORDER) ? __BYTE_ORDER == __BIG_ENDIAN : defined(__BIG_ENDIAN)
157 __u32 events_hi; /* 7 high-order of superblock update count */
158 __u32 events_lo; /* 8 low-order of superblock update count */
159 __u32 cp_events_hi; /* 9 high-order of checkpoint update count */
160 __u32 cp_events_lo; /* 10 low-order of checkpoint update count */
161 #elif defined(__BYTE_ORDER) ? __BYTE_ORDER == __LITTLE_ENDIAN : defined(__LITTLE_ENDIAN)
162 __u32 events_lo; /* 7 low-order of superblock update count */
163 __u32 events_hi; /* 8 high-order of superblock update count */
164 __u32 cp_events_lo; /* 9 low-order of checkpoint update count */
165 __u32 cp_events_hi; /* 10 high-order of checkpoint update count */
167 #error unspecified endianness
169 __u32 recovery_cp; /* 11 recovery checkpoint sector count */
170 /* There are only valid for minor_version > 90 */
171 __u64 reshape_position; /* 12,13 next address in array-space for reshape */
172 __u32 new_level; /* 14 new level we are reshaping to */
173 __u32 delta_disks; /* 15 change in number of raid_disks */
174 __u32 new_layout; /* 16 new layout */
175 __u32 new_chunk; /* 17 new chunk size (bytes) */
176 __u32 gstate_sreserved[MD_SB_GENERIC_STATE_WORDS - 18];
179 * Personality information
181 __u32 layout; /* 0 the array's physical layout */
182 __u32 chunk_size; /* 1 chunk size in bytes */
183 __u32 root_pv; /* 2 LV root PV */
184 __u32 root_block; /* 3 LV root block */
185 __u32 pstate_reserved[MD_SB_PERSONALITY_WORDS - 4];
190 mdp_disk_t disks[MD_SB_DISKS];
195 __u32 reserved[MD_SB_RESERVED_WORDS];
200 mdp_disk_t this_disk;
204 static inline __u64 md_event(mdp_super_t *sb) {
205 __u64 ev = sb->events_hi;
206 return (ev<<32)| sb->events_lo;
209 #define MD_SUPERBLOCK_1_TIME_SEC_MASK ((1ULL<<40) - 1)
212 * The version-1 superblock :
213 * All numeric fields are little-endian.
215 * total size: 256 bytes plus 2 per device.
216 * 1K allows 384 devices.
218 struct mdp_superblock_1 {
219 /* constant array information - 128 bytes */
220 __le32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */
221 __le32 major_version; /* 1 */
222 __le32 feature_map; /* bit 0 set if 'bitmap_offset' is meaningful */
223 __le32 pad0; /* always set to 0 when writing */
225 __u8 set_uuid[16]; /* user-space generated. */
226 char set_name[32]; /* set and interpreted by user-space */
228 __le64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/
229 __le32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */
230 __le32 layout; /* only for raid5 and raid10 currently */
231 __le64 size; /* used size of component devices, in 512byte sectors */
233 __le32 chunksize; /* in 512byte sectors */
235 __le32 bitmap_offset; /* sectors after start of superblock that bitmap starts
236 * NOTE: signed, so bitmap can be before superblock
237 * only meaningful of feature_map[0] is set.
240 /* These are only valid with feature bit '4' */
241 __le32 new_level; /* new level we are reshaping to */
242 __le64 reshape_position; /* next address in array-space for reshape */
243 __le32 delta_disks; /* change in number of raid_disks */
244 __le32 new_layout; /* new layout */
245 __le32 new_chunk; /* new chunk size (512byte sectors) */
246 __le32 new_offset; /* signed number to add to data_offset in new
247 * layout. 0 == no-change. This can be
248 * different on each device in the array.
251 /* constant this-device information - 64 bytes */
252 __le64 data_offset; /* sector start of data, often 0 */
253 __le64 data_size; /* sectors in this device that can be used for data */
254 __le64 super_offset; /* sector start of this superblock */
255 __le64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */
256 __le32 dev_number; /* permanent identifier of this device - not role in raid */
257 __le32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */
258 __u8 device_uuid[16]; /* user-space setable, ignored by kernel */
259 __u8 devflags; /* per-device flags. Only one defined...*/
260 #define WriteMostly1 1 /* mask for writemostly flag in above */
261 /* Bad block log. If there are any bad blocks the feature flag is set.
262 * If offset and size are non-zero, that space is reserved and available
264 __u8 bblog_shift; /* shift from sectors to block size */
265 __le16 bblog_size; /* number of sectors reserved for list */
266 __le32 bblog_offset; /* sector offset from superblock to bblog,
267 * signed - not unsigned */
269 /* array state information - 64 bytes */
270 __le64 utime; /* 40 bits second, 24 bits microseconds */
271 __le64 events; /* incremented when superblock updated */
272 __le64 resync_offset; /* data before this offset (from data_offset) known to be in sync */
273 __le32 sb_csum; /* checksum up to devs[max_dev] */
274 __le32 max_dev; /* size of devs[] array to consider */
275 __u8 pad3[64-32]; /* set to 0 when writing */
277 /* device state information. Indexed by dev_number.
279 * Note there are no per-device state flags. State information is rolled
280 * into the 'roles' value. If a device is spare or faulty, then it doesn't
281 * have a meaningful role.
283 __le16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
286 /* feature_map bits */
287 #define MD_FEATURE_BITMAP_OFFSET 1
288 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
291 #define MD_FEATURE_RESHAPE_ACTIVE 4
292 #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */
293 #define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an
294 * active device with same 'role'.
295 * 'recovery_offset' is also set.
297 #define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number
298 * of devices, but is going
301 #define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */
302 #define MD_FEATURE_RECOVERY_BITMAP 128 /* recovery that is happening
303 * is guided by bitmap.
305 #define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \
306 |MD_FEATURE_RECOVERY_OFFSET \
307 |MD_FEATURE_RESHAPE_ACTIVE \
308 |MD_FEATURE_BAD_BLOCKS \
309 |MD_FEATURE_REPLACEMENT \
310 |MD_FEATURE_RESHAPE_BACKWARDS \
311 |MD_FEATURE_NEW_OFFSET \
312 |MD_FEATURE_RECOVERY_BITMAP \