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#include <linux/types.h> 19 20/* 21 * RAID superblock. 22 * 23 * The RAID superblock maintains some statistics on each RAID configuration. 24 * Each real device in the RAID set contains it near the end of the device. 25 * Some of the ideas are copied from the ext2fs implementation. 26 * 27 * We currently use 4096 bytes as follows: 28 * 29 * word offset function 30 * 31 * 0 - 31 Constant generic RAID device information. 32 * 32 - 63 Generic state information. 33 * 64 - 127 Personality specific information. 34 * 128 - 511 12 32-words descriptors of the disks in the raid set. 35 * 512 - 911 Reserved. 36 * 912 - 1023 Disk specific descriptor. 37 */ 38 39/* 40 * If x is the real device size in bytes, we return an apparent size of: 41 * 42 * y = (x & ~(MD_RESERVED_BYTES - 1)) - MD_RESERVED_BYTES 43 * 44 * and place the 4kB superblock at offset y. 45 */ 46#define MD_RESERVED_BYTES (64 * 1024) 47#define MD_RESERVED_SECTORS (MD_RESERVED_BYTES / 512) 48 49#define MD_NEW_SIZE_SECTORS(x) ((x & ~(MD_RESERVED_SECTORS - 1)) - MD_RESERVED_SECTORS) 50 51#define MD_SB_BYTES 4096 52#define MD_SB_WORDS (MD_SB_BYTES / 4) 53#define MD_SB_SECTORS (MD_SB_BYTES / 512) 54 55/* 56 * The following are counted in 32-bit words 57 */ 58#define MD_SB_GENERIC_OFFSET 0 59#define MD_SB_PERSONALITY_OFFSET 64 60#define MD_SB_DISKS_OFFSET 128 61#define MD_SB_DESCRIPTOR_OFFSET 992 62 63#define MD_SB_GENERIC_CONSTANT_WORDS 32 64#define MD_SB_GENERIC_STATE_WORDS 32 65#define MD_SB_GENERIC_WORDS (MD_SB_GENERIC_CONSTANT_WORDS + MD_SB_GENERIC_STATE_WORDS) 66#define MD_SB_PERSONALITY_WORDS 64 67#define MD_SB_DESCRIPTOR_WORDS 32 68#define MD_SB_DISKS 27 69#define MD_SB_DISKS_WORDS (MD_SB_DISKS*MD_SB_DESCRIPTOR_WORDS) 70#define MD_SB_RESERVED_WORDS (1024 - MD_SB_GENERIC_WORDS - MD_SB_PERSONALITY_WORDS - MD_SB_DISKS_WORDS - MD_SB_DESCRIPTOR_WORDS) 71#define MD_SB_EQUAL_WORDS (MD_SB_GENERIC_WORDS + MD_SB_PERSONALITY_WORDS + MD_SB_DISKS_WORDS) 72 73/* 74 * Device "operational" state bits 75 */ 76#define MD_DISK_FAULTY 0 /* disk is faulty / operational */ 77#define MD_DISK_ACTIVE 1 /* disk is running or spare disk */ 78#define MD_DISK_SYNC 2 /* disk is in sync with the raid set */ 79#define MD_DISK_REMOVED 3 /* disk is in sync with the raid set */ 80#define MD_DISK_FAILFAST 10 /* Send REQ_FAILFAST if there are multiple 81 * devices available - and don't try to 82 * correct read errors. 83 */ 84 85#define MD_DISK_WRITEMOSTLY 9 /* disk is "write-mostly" is RAID1 config. 86 * read requests will only be sent here in 87 * dire need 88 */ 89#define MD_DISK_JOURNAL 18 /* disk is used as the write journal in RAID-5/6 */ 90 91#define MD_DISK_ROLE_SPARE 0xffff 92#define MD_DISK_ROLE_FAULTY 0xfffe 93#define MD_DISK_ROLE_JOURNAL 0xfffd 94#define MD_DISK_ROLE_MAX 0xff00 /* max value of regular disk role */ 95 96typedef struct mdp_device_descriptor_s { 97 __u32 number; /* 0 Device number in the entire set */ 98 __u32 major; /* 1 Device major number */ 99 __u32 minor; /* 2 Device minor number */ 100 __u32 raid_disk; /* 3 The role of the device in the raid set */ 101 __u32 state; /* 4 Operational state */ 102 __u32 reserved[MD_SB_DESCRIPTOR_WORDS - 5]; 103} mdp_disk_t; 104 105#define MD_SB_MAGIC 0xa92b4efc 106 107/* 108 * Superblock state bits 109 */ 110#define MD_SB_CLEAN 0 111#define MD_SB_ERRORS 1 112 113#define MD_SB_BITMAP_PRESENT 8 /* bitmap may be present nearby */ 114 115/* 116 * Notes: 117 * - if an array is being reshaped (restriped) in order to change the 118 * the number of active devices in the array, 'raid_disks' will be 119 * the larger of the old and new numbers. 'delta_disks' will 120 * be the "new - old". So if +ve, raid_disks is the new value, and 121 * "raid_disks-delta_disks" is the old. If -ve, raid_disks is the 122 * old value and "raid_disks+delta_disks" is the new (smaller) value. 123 */ 124 125 126typedef struct mdp_superblock_s { 127 /* 128 * Constant generic information 129 */ 130 __u32 md_magic; /* 0 MD identifier */ 131 __u32 major_version; /* 1 major version to which the set conforms */ 132 __u32 minor_version; /* 2 minor version ... */ 133 __u32 patch_version; /* 3 patchlevel version ... */ 134 __u32 gvalid_words; /* 4 Number of used words in this section */ 135 __u32 set_uuid0; /* 5 Raid set identifier */ 136 __u32 ctime; /* 6 Creation time */ 137 __u32 level; /* 7 Raid personality */ 138 __u32 size; /* 8 Apparent size of each individual disk */ 139 __u32 nr_disks; /* 9 total disks in the raid set */ 140 __u32 raid_disks; /* 10 disks in a fully functional raid set */ 141 __u32 md_minor; /* 11 preferred MD minor device number */ 142 __u32 not_persistent; /* 12 does it have a persistent superblock */ 143 __u32 set_uuid1; /* 13 Raid set identifier #2 */ 144 __u32 set_uuid2; /* 14 Raid set identifier #3 */ 145 __u32 set_uuid3; /* 15 Raid set identifier #4 */ 146 __u32 gstate_creserved[MD_SB_GENERIC_CONSTANT_WORDS - 16]; 147 148 /* 149 * Generic state information 150 */ 151 __u32 utime; /* 0 Superblock update time */ 152 __u32 state; /* 1 State bits (clean, ...) */ 153 __u32 active_disks; /* 2 Number of currently active disks */ 154 __u32 working_disks; /* 3 Number of working disks */ 155 __u32 failed_disks; /* 4 Number of failed disks */ 156 __u32 spare_disks; /* 5 Number of spare disks */ 157 __u32 sb_csum; /* 6 checksum of the whole superblock */ 158#if defined(__BYTE_ORDER) ? __BYTE_ORDER == __BIG_ENDIAN : defined(__BIG_ENDIAN) 159 __u32 events_hi; /* 7 high-order of superblock update count */ 160 __u32 events_lo; /* 8 low-order of superblock update count */ 161 __u32 cp_events_hi; /* 9 high-order of checkpoint update count */ 162 __u32 cp_events_lo; /* 10 low-order of checkpoint update count */ 163#elif defined(__BYTE_ORDER) ? __BYTE_ORDER == __LITTLE_ENDIAN : defined(__LITTLE_ENDIAN) 164 __u32 events_lo; /* 7 low-order of superblock update count */ 165 __u32 events_hi; /* 8 high-order of superblock update count */ 166 __u32 cp_events_lo; /* 9 low-order of checkpoint update count */ 167 __u32 cp_events_hi; /* 10 high-order of checkpoint update count */ 168#else 169#error unspecified endianness 170#endif 171 __u32 recovery_cp; /* 11 recovery checkpoint sector count */ 172 /* There are only valid for minor_version > 90 */ 173 __u64 reshape_position; /* 12,13 next address in array-space for reshape */ 174 __u32 new_level; /* 14 new level we are reshaping to */ 175 __u32 delta_disks; /* 15 change in number of raid_disks */ 176 __u32 new_layout; /* 16 new layout */ 177 __u32 new_chunk; /* 17 new chunk size (bytes) */ 178 __u32 gstate_sreserved[MD_SB_GENERIC_STATE_WORDS - 18]; 179 180 /* 181 * Personality information 182 */ 183 __u32 layout; /* 0 the array's physical layout */ 184 __u32 chunk_size; /* 1 chunk size in bytes */ 185 __u32 root_pv; /* 2 LV root PV */ 186 __u32 root_block; /* 3 LV root block */ 187 __u32 pstate_reserved[MD_SB_PERSONALITY_WORDS - 4]; 188 189 /* 190 * Disks information 191 */ 192 mdp_disk_t disks[MD_SB_DISKS]; 193 194 /* 195 * Reserved 196 */ 197 __u32 reserved[MD_SB_RESERVED_WORDS]; 198 199 /* 200 * Active descriptor 201 */ 202 mdp_disk_t this_disk; 203 204} mdp_super_t; 205 206static inline __u64 md_event(mdp_super_t *sb) { 207 __u64 ev = sb->events_hi; 208 return (ev<<32)| sb->events_lo; 209} 210 211#define MD_SUPERBLOCK_1_TIME_SEC_MASK ((1ULL<<40) - 1) 212 213/* 214 * The version-1 superblock : 215 * All numeric fields are little-endian. 216 * 217 * total size: 256 bytes plus 2 per device. 218 * 1K allows 384 devices. 219 */ 220struct mdp_superblock_1 { 221 /* constant array information - 128 bytes */ 222 __le32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */ 223 __le32 major_version; /* 1 */ 224 __le32 feature_map; /* bit 0 set if 'bitmap_offset' is meaningful */ 225 __le32 pad0; /* always set to 0 when writing */ 226 227 __u8 set_uuid[16]; /* user-space generated. */ 228 char set_name[32]; /* set and interpreted by user-space */ 229 230 __le64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/ 231 __le32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */ 232 __le32 layout; /* only for raid5 and raid10 currently */ 233 __le64 size; /* used size of component devices, in 512byte sectors */ 234 235 __le32 chunksize; /* in 512byte sectors */ 236 __le32 raid_disks; 237 union { 238 __le32 bitmap_offset; /* sectors after start of superblock that bitmap starts 239 * NOTE: signed, so bitmap can be before superblock 240 * only meaningful of feature_map[0] is set. 241 */ 242 243 /* only meaningful when feature_map[MD_FEATURE_PPL] is set */ 244 struct { 245 __le16 offset; /* sectors from start of superblock that ppl starts (signed) */ 246 __le16 size; /* ppl size in sectors */ 247 } ppl; 248 }; 249 250 /* These are only valid with feature bit '4' */ 251 __le32 new_level; /* new level we are reshaping to */ 252 __le64 reshape_position; /* next address in array-space for reshape */ 253 __le32 delta_disks; /* change in number of raid_disks */ 254 __le32 new_layout; /* new layout */ 255 __le32 new_chunk; /* new chunk size (512byte sectors) */ 256 __le32 new_offset; /* signed number to add to data_offset in new 257 * layout. 0 == no-change. This can be 258 * different on each device in the array. 259 */ 260 261 /* constant this-device information - 64 bytes */ 262 __le64 data_offset; /* sector start of data, often 0 */ 263 __le64 data_size; /* sectors in this device that can be used for data */ 264 __le64 super_offset; /* sector start of this superblock */ 265 union { 266 __le64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */ 267 __le64 journal_tail;/* journal tail of journal device (from data_offset) */ 268 }; 269 __le32 dev_number; /* permanent identifier of this device - not role in raid */ 270 __le32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */ 271 __u8 device_uuid[16]; /* user-space setable, ignored by kernel */ 272 __u8 devflags; /* per-device flags. Only two defined...*/ 273#define WriteMostly1 1 /* mask for writemostly flag in above */ 274#define FailFast1 2 /* Should avoid retries and fixups and just fail */ 275 /* Bad block log. If there are any bad blocks the feature flag is set. 276 * If offset and size are non-zero, that space is reserved and available 277 */ 278 __u8 bblog_shift; /* shift from sectors to block size */ 279 __le16 bblog_size; /* number of sectors reserved for list */ 280 __le32 bblog_offset; /* sector offset from superblock to bblog, 281 * signed - not unsigned */ 282 283 /* array state information - 64 bytes */ 284 __le64 utime; /* 40 bits second, 24 bits microseconds */ 285 __le64 events; /* incremented when superblock updated */ 286 __le64 resync_offset; /* data before this offset (from data_offset) known to be in sync */ 287 __le32 sb_csum; /* checksum up to devs[max_dev] */ 288 __le32 max_dev; /* size of devs[] array to consider */ 289 __u8 pad3[64-32]; /* set to 0 when writing */ 290 291 /* device state information. Indexed by dev_number. 292 * 2 bytes per device 293 * Note there are no per-device state flags. State information is rolled 294 * into the 'roles' value. If a device is spare or faulty, then it doesn't 295 * have a meaningful role. 296 */ 297 __le16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */ 298}; 299 300/* feature_map bits */ 301#define MD_FEATURE_BITMAP_OFFSET 1 302#define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and 303 * must be honoured 304 */ 305#define MD_FEATURE_RESHAPE_ACTIVE 4 306#define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */ 307#define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an 308 * active device with same 'role'. 309 * 'recovery_offset' is also set. 310 */ 311#define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number 312 * of devices, but is going 313 * backwards anyway. 314 */ 315#define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */ 316#define MD_FEATURE_RECOVERY_BITMAP 128 /* recovery that is happening 317 * is guided by bitmap. 318 */ 319#define MD_FEATURE_JOURNAL 512 /* support write cache */ 320#define MD_FEATURE_PPL 1024 /* support PPL */ 321#define MD_FEATURE_MULTIPLE_PPLS 2048 /* support for multiple PPLs */ 322#define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \ 323 |MD_FEATURE_RECOVERY_OFFSET \ 324 |MD_FEATURE_RESHAPE_ACTIVE \ 325 |MD_FEATURE_BAD_BLOCKS \ 326 |MD_FEATURE_REPLACEMENT \ 327 |MD_FEATURE_RESHAPE_BACKWARDS \ 328 |MD_FEATURE_NEW_OFFSET \ 329 |MD_FEATURE_RECOVERY_BITMAP \ 330 |MD_FEATURE_JOURNAL \ 331 |MD_FEATURE_PPL \ 332 |MD_FEATURE_MULTIPLE_PPLS \ 333 ) 334 335struct r5l_payload_header { 336 __le16 type; 337 __le16 flags; 338} __attribute__ ((__packed__)); 339 340enum r5l_payload_type { 341 R5LOG_PAYLOAD_DATA = 0, 342 R5LOG_PAYLOAD_PARITY = 1, 343 R5LOG_PAYLOAD_FLUSH = 2, 344}; 345 346struct r5l_payload_data_parity { 347 struct r5l_payload_header header; 348 __le32 size; /* sector. data/parity size. each 4k 349 * has a checksum */ 350 __le64 location; /* sector. For data, it's raid sector. For 351 * parity, it's stripe sector */ 352 __le32 checksum[]; 353} __attribute__ ((__packed__)); 354 355enum r5l_payload_data_parity_flag { 356 R5LOG_PAYLOAD_FLAG_DISCARD = 1, /* payload is discard */ 357 /* 358 * RESHAPED/RESHAPING is only set when there is reshape activity. Note, 359 * both data/parity of a stripe should have the same flag set 360 * 361 * RESHAPED: reshape is running, and this stripe finished reshape 362 * RESHAPING: reshape is running, and this stripe isn't reshaped 363 */ 364 R5LOG_PAYLOAD_FLAG_RESHAPED = 2, 365 R5LOG_PAYLOAD_FLAG_RESHAPING = 3, 366}; 367 368struct r5l_payload_flush { 369 struct r5l_payload_header header; 370 __le32 size; /* flush_stripes size, bytes */ 371 __le64 flush_stripes[]; 372} __attribute__ ((__packed__)); 373 374enum r5l_payload_flush_flag { 375 R5LOG_PAYLOAD_FLAG_FLUSH_STRIPE = 1, /* data represents whole stripe */ 376}; 377 378struct r5l_meta_block { 379 __le32 magic; 380 __le32 checksum; 381 __u8 version; 382 __u8 __zero_pading_1; 383 __le16 __zero_pading_2; 384 __le32 meta_size; /* whole size of the block */ 385 386 __le64 seq; 387 __le64 position; /* sector, start from rdev->data_offset, current position */ 388 struct r5l_payload_header payloads[]; 389} __attribute__ ((__packed__)); 390 391#define R5LOG_VERSION 0x1 392#define R5LOG_MAGIC 0x6433c509 393 394struct ppl_header_entry { 395 __le64 data_sector; /* raid sector of the new data */ 396 __le32 pp_size; /* length of partial parity */ 397 __le32 data_size; /* length of data */ 398 __le32 parity_disk; /* member disk containing parity */ 399 __le32 checksum; /* checksum of partial parity data for this 400 * entry (~crc32c) */ 401} __attribute__ ((__packed__)); 402 403#define PPL_HEADER_SIZE 4096 404#define PPL_HDR_RESERVED 512 405#define PPL_HDR_ENTRY_SPACE \ 406 (PPL_HEADER_SIZE - PPL_HDR_RESERVED - 4 * sizeof(__le32) - sizeof(__le64)) 407#define PPL_HDR_MAX_ENTRIES \ 408 (PPL_HDR_ENTRY_SPACE / sizeof(struct ppl_header_entry)) 409 410struct ppl_header { 411 __u8 reserved[PPL_HDR_RESERVED];/* reserved space, fill with 0xff */ 412 __le32 signature; /* signature (family number of volume) */ 413 __le32 padding; /* zero pad */ 414 __le64 generation; /* generation number of the header */ 415 __le32 entries_count; /* number of entries in entry array */ 416 __le32 checksum; /* checksum of the header (~crc32c) */ 417 struct ppl_header_entry entries[PPL_HDR_MAX_ENTRIES]; 418} __attribute__ ((__packed__)); 419 420#endif 421