1/* 2 md_p.h : physical layout of Linux RAID devices 3 Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman 4 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) 8 any later version. 9 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. 13*/ 14 15#ifndef _MD_P_H 16#define _MD_P_H 17 18/* 19 * RAID superblock. 20 * 21 * The RAID superblock maintains some statistics on each RAID configuration. 22 * Each real device in the RAID set contains it near the end of the device. 23 * Some of the ideas are copied from the ext2fs implementation. 24 * 25 * We currently use 4096 bytes as follows: 26 * 27 * word offset function 28 * 29 * 0 - 31 Constant generic RAID device information. 30 * 32 - 63 Generic state information. 31 * 64 - 127 Personality specific information. 32 * 128 - 511 12 32-words descriptors of the disks in the raid set. 33 * 512 - 911 Reserved. 34 * 912 - 1023 Disk specific descriptor. 35 */ 36 37/* 38 * If x is the real device size in bytes, we return an apparent size of: 39 * 40 * y = (x & ~(MD_RESERVED_BYTES - 1)) - MD_RESERVED_BYTES 41 * 42 * and place the 4kB superblock at offset y. 43 */ 44#define MD_RESERVED_BYTES (64 * 1024) 45#define MD_RESERVED_SECTORS (MD_RESERVED_BYTES / 512) 46#define MD_RESERVED_BLOCKS (MD_RESERVED_BYTES / BLOCK_SIZE) 47 48#define MD_NEW_SIZE_SECTORS(x) ((x & ~(MD_RESERVED_SECTORS - 1)) - MD_RESERVED_SECTORS) 49#define MD_NEW_SIZE_BLOCKS(x) ((x & ~(MD_RESERVED_BLOCKS - 1)) - MD_RESERVED_BLOCKS) 50 51#define MD_SB_BYTES 4096 52#define MD_SB_WORDS (MD_SB_BYTES / 4) 53#define MD_SB_BLOCKS (MD_SB_BYTES / BLOCK_SIZE) 54#define MD_SB_SECTORS (MD_SB_BYTES / 512) 55 56/* 57 * The following are counted in 32-bit words 58 */ 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 63 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) 73 74/* 75 * Device "operational" state bits 76 */ 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 82#define MD_DISK_WRITEMOSTLY 9 /* disk is "write-mostly" is RAID1 config. 83 * read requests will only be sent here in 84 * dire need 85 */ 86 87typedef struct mdp_device_descriptor_s { 88 __u32 number; /* 0 Device number in the entire set */ 89 __u32 major; /* 1 Device major number */ 90 __u32 minor; /* 2 Device minor number */ 91 __u32 raid_disk; /* 3 The role of the device in the raid set */ 92 __u32 state; /* 4 Operational state */ 93 __u32 reserved[MD_SB_DESCRIPTOR_WORDS - 5]; 94} mdp_disk_t; 95 96#define MD_SB_MAGIC 0xa92b4efc 97 98/* 99 * Superblock state bits 100 */ 101#define MD_SB_CLEAN 0 102#define MD_SB_ERRORS 1 103 104#define MD_SB_BITMAP_PRESENT 8 /* bitmap may be present nearby */ 105 106/* 107 * Notes: 108 * - if an array is being reshaped (restriped) in order to change the 109 * the number of active devices in the array, 'raid_disks' will be 110 * the larger of the old and new numbers. 'delta_disks' will 111 * be the "new - old". So if +ve, raid_disks is the new value, and 112 * "raid_disks-delta_disks" is the old. If -ve, raid_disks is the 113 * old value and "raid_disks+delta_disks" is the new (smaller) value. 114 */ 115 116 117typedef struct mdp_superblock_s { 118 /* 119 * Constant generic information 120 */ 121 __u32 md_magic; /* 0 MD identifier */ 122 __u32 major_version; /* 1 major version to which the set conforms */ 123 __u32 minor_version; /* 2 minor version ... */ 124 __u32 patch_version; /* 3 patchlevel version ... */ 125 __u32 gvalid_words; /* 4 Number of used words in this section */ 126 __u32 set_uuid0; /* 5 Raid set identifier */ 127 __u32 ctime; /* 6 Creation time */ 128 __u32 level; /* 7 Raid personality */ 129 __u32 size; /* 8 Apparent size of each individual disk */ 130 __u32 nr_disks; /* 9 total disks in the raid set */ 131 __u32 raid_disks; /* 10 disks in a fully functional raid set */ 132 __u32 md_minor; /* 11 preferred MD minor device number */ 133 __u32 not_persistent; /* 12 does it have a persistent superblock */ 134 __u32 set_uuid1; /* 13 Raid set identifier #2 */ 135 __u32 set_uuid2; /* 14 Raid set identifier #3 */ 136 __u32 set_uuid3; /* 15 Raid set identifier #4 */ 137 __u32 gstate_creserved[MD_SB_GENERIC_CONSTANT_WORDS - 16]; 138 139 /* 140 * Generic state information 141 */ 142 __u32 utime; /* 0 Superblock update time */ 143 __u32 state; /* 1 State bits (clean, ...) */ 144 __u32 active_disks; /* 2 Number of currently active disks */ 145 __u32 working_disks; /* 3 Number of working disks */ 146 __u32 failed_disks; /* 4 Number of failed disks */ 147 __u32 spare_disks; /* 5 Number of spare disks */ 148 __u32 sb_csum; /* 6 checksum of the whole superblock */ 149#ifdef __BIG_ENDIAN 150 __u32 events_hi; /* 7 high-order of superblock update count */ 151 __u32 events_lo; /* 8 low-order of superblock update count */ 152 __u32 cp_events_hi; /* 9 high-order of checkpoint update count */ 153 __u32 cp_events_lo; /* 10 low-order of checkpoint update count */ 154#else 155 __u32 events_lo; /* 7 low-order of superblock update count */ 156 __u32 events_hi; /* 8 high-order of superblock update count */ 157 __u32 cp_events_lo; /* 9 low-order of checkpoint update count */ 158 __u32 cp_events_hi; /* 10 high-order of checkpoint update count */ 159#endif 160 __u32 recovery_cp; /* 11 recovery checkpoint sector count */ 161 /* There are only valid for minor_version > 90 */ 162 __u64 reshape_position; /* 12,13 next address in array-space for reshape */ 163 __u32 new_level; /* 14 new level we are reshaping to */ 164 __u32 delta_disks; /* 15 change in number of raid_disks */ 165 __u32 new_layout; /* 16 new layout */ 166 __u32 new_chunk; /* 17 new chunk size (bytes) */ 167 __u32 gstate_sreserved[MD_SB_GENERIC_STATE_WORDS - 18]; 168 169 /* 170 * Personality information 171 */ 172 __u32 layout; /* 0 the array's physical layout */ 173 __u32 chunk_size; /* 1 chunk size in bytes */ 174 __u32 root_pv; /* 2 LV root PV */ 175 __u32 root_block; /* 3 LV root block */ 176 __u32 pstate_reserved[MD_SB_PERSONALITY_WORDS - 4]; 177 178 /* 179 * Disks information 180 */ 181 mdp_disk_t disks[MD_SB_DISKS]; 182 183 /* 184 * Reserved 185 */ 186 __u32 reserved[MD_SB_RESERVED_WORDS]; 187 188 /* 189 * Active descriptor 190 */ 191 mdp_disk_t this_disk; 192 193} mdp_super_t; 194 195static inline __u64 md_event(mdp_super_t *sb) { 196 __u64 ev = sb->events_hi; 197 return (ev<<32)| sb->events_lo; 198} 199 200/* 201 * The version-1 superblock : 202 * All numeric fields are little-endian. 203 * 204 * total size: 256 bytes plus 2 per device. 205 * 1K allows 384 devices. 206 */ 207struct mdp_superblock_1 { 208 /* constant array information - 128 bytes */ 209 __le32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */ 210 __le32 major_version; /* 1 */ 211 __le32 feature_map; /* bit 0 set if 'bitmap_offset' is meaningful */ 212 __le32 pad0; /* always set to 0 when writing */ 213 214 __u8 set_uuid[16]; /* user-space generated. */ 215 char set_name[32]; /* set and interpreted by user-space */ 216 217 __le64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/ 218 __le32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */ 219 __le32 layout; /* only for raid5 and raid10 currently */ 220 __le64 size; /* used size of component devices, in 512byte sectors */ 221 222 __le32 chunksize; /* in 512byte sectors */ 223 __le32 raid_disks; 224 __le32 bitmap_offset; /* sectors after start of superblock that bitmap starts 225 * NOTE: signed, so bitmap can be before superblock 226 * only meaningful of feature_map[0] is set. 227 */ 228 229 /* These are only valid with feature bit '4' */ 230 __le32 new_level; /* new level we are reshaping to */ 231 __le64 reshape_position; /* next address in array-space for reshape */ 232 __le32 delta_disks; /* change in number of raid_disks */ 233 __le32 new_layout; /* new layout */ 234 __le32 new_chunk; /* new chunk size (bytes) */ 235 __u8 pad1[128-124]; /* set to 0 when written */ 236 237 /* constant this-device information - 64 bytes */ 238 __le64 data_offset; /* sector start of data, often 0 */ 239 __le64 data_size; /* sectors in this device that can be used for data */ 240 __le64 super_offset; /* sector start of this superblock */ 241 __le64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */ 242 __le32 dev_number; /* permanent identifier of this device - not role in raid */ 243 __le32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */ 244 __u8 device_uuid[16]; /* user-space setable, ignored by kernel */ 245 __u8 devflags; /* per-device flags. Only one defined...*/ 246#define WriteMostly1 1 /* mask for writemostly flag in above */ 247 __u8 pad2[64-57]; /* set to 0 when writing */ 248 249 /* array state information - 64 bytes */ 250 __le64 utime; /* 40 bits second, 24 btes microseconds */ 251 __le64 events; /* incremented when superblock updated */ 252 __le64 resync_offset; /* data before this offset (from data_offset) known to be in sync */ 253 __le32 sb_csum; /* checksum upto devs[max_dev] */ 254 __le32 max_dev; /* size of devs[] array to consider */ 255 __u8 pad3[64-32]; /* set to 0 when writing */ 256 257 /* device state information. Indexed by dev_number. 258 * 2 bytes per device 259 * Note there are no per-device state flags. State information is rolled 260 * into the 'roles' value. If a device is spare or faulty, then it doesn't 261 * have a meaningful role. 262 */ 263 __le16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */ 264}; 265 266/* feature_map bits */ 267#define MD_FEATURE_BITMAP_OFFSET 1 268#define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and 269 * must be honoured 270 */ 271#define MD_FEATURE_RESHAPE_ACTIVE 4 272 273#define MD_FEATURE_ALL (1|2|4) 274 275#endif 276