430 lines · c
1/* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */2/*3 md_p.h : physical layout of Linux RAID devices4 Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman5 6 This program is free software; you can redistribute it and/or modify7 it under the terms of the GNU General Public License as published by8 the Free Software Foundation; either version 2, or (at your option)9 any later version.10*/11 12#ifndef _MD_P_H13#define _MD_P_H14 15#include <linux/types.h>16#include <asm/byteorder.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 function28 *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_BYTES41 *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 47#define MD_NEW_SIZE_SECTORS(x) ((x & ~(MD_RESERVED_SECTORS - 1)) - MD_RESERVED_SECTORS)48 49#define MD_SB_BYTES 409650#define MD_SB_WORDS (MD_SB_BYTES / 4)51#define MD_SB_SECTORS (MD_SB_BYTES / 512)52 53/*54 * The following are counted in 32-bit words55 */56#define MD_SB_GENERIC_OFFSET 057#define MD_SB_PERSONALITY_OFFSET 6458#define MD_SB_DISKS_OFFSET 12859#define MD_SB_DESCRIPTOR_OFFSET 99260 61#define MD_SB_GENERIC_CONSTANT_WORDS 3262#define MD_SB_GENERIC_STATE_WORDS 3263#define MD_SB_GENERIC_WORDS (MD_SB_GENERIC_CONSTANT_WORDS + MD_SB_GENERIC_STATE_WORDS)64#define MD_SB_PERSONALITY_WORDS 6465#define MD_SB_DESCRIPTOR_WORDS 3266#define MD_SB_DISKS 2767#define MD_SB_DISKS_WORDS (MD_SB_DISKS*MD_SB_DESCRIPTOR_WORDS)68#define MD_SB_RESERVED_WORDS (1024 - MD_SB_GENERIC_WORDS - MD_SB_PERSONALITY_WORDS - MD_SB_DISKS_WORDS - MD_SB_DESCRIPTOR_WORDS)69#define MD_SB_EQUAL_WORDS (MD_SB_GENERIC_WORDS + MD_SB_PERSONALITY_WORDS + MD_SB_DISKS_WORDS)70 71/*72 * Device "operational" state bits73 */74#define MD_DISK_FAULTY 0 /* disk is faulty / operational */75#define MD_DISK_ACTIVE 1 /* disk is running or spare disk */76#define MD_DISK_SYNC 2 /* disk is in sync with the raid set */77#define MD_DISK_REMOVED 3 /* disk is in sync with the raid set */78#define MD_DISK_CLUSTER_ADD 4 /* Initiate a disk add across the cluster79 * For clustered enviroments only.80 */81#define MD_DISK_CANDIDATE 5 /* disk is added as spare (local) until confirmed82 * For clustered enviroments only.83 */84#define MD_DISK_FAILFAST 10 /* Send REQ_FAILFAST if there are multiple85 * devices available - and don't try to86 * correct read errors.87 */88 89#define MD_DISK_WRITEMOSTLY 9 /* disk is "write-mostly" is RAID1 config.90 * read requests will only be sent here in91 * dire need92 */93#define MD_DISK_JOURNAL 18 /* disk is used as the write journal in RAID-5/6 */94 95#define MD_DISK_ROLE_SPARE 0xffff96#define MD_DISK_ROLE_FAULTY 0xfffe97#define MD_DISK_ROLE_JOURNAL 0xfffd98#define MD_DISK_ROLE_MAX 0xff00 /* max value of regular disk role */99 100typedef struct mdp_device_descriptor_s {101 __u32 number; /* 0 Device number in the entire set */102 __u32 major; /* 1 Device major number */103 __u32 minor; /* 2 Device minor number */104 __u32 raid_disk; /* 3 The role of the device in the raid set */105 __u32 state; /* 4 Operational state */106 __u32 reserved[MD_SB_DESCRIPTOR_WORDS - 5];107} mdp_disk_t;108 109#define MD_SB_MAGIC 0xa92b4efc110 111/*112 * Superblock state bits113 */114#define MD_SB_CLEAN 0115#define MD_SB_ERRORS 1116 117#define MD_SB_CLUSTERED 5 /* MD is clustered */118#define MD_SB_BITMAP_PRESENT 8 /* bitmap may be present nearby */119 120/*121 * Notes:122 * - if an array is being reshaped (restriped) in order to change123 * the number of active devices in the array, 'raid_disks' will be124 * the larger of the old and new numbers. 'delta_disks' will125 * be the "new - old". So if +ve, raid_disks is the new value, and126 * "raid_disks-delta_disks" is the old. If -ve, raid_disks is the127 * old value and "raid_disks+delta_disks" is the new (smaller) value.128 */129 130 131typedef struct mdp_superblock_s {132 /*133 * Constant generic information134 */135 __u32 md_magic; /* 0 MD identifier */136 __u32 major_version; /* 1 major version to which the set conforms */137 __u32 minor_version; /* 2 minor version ... */138 __u32 patch_version; /* 3 patchlevel version ... */139 __u32 gvalid_words; /* 4 Number of used words in this section */140 __u32 set_uuid0; /* 5 Raid set identifier */141 __u32 ctime; /* 6 Creation time */142 __u32 level; /* 7 Raid personality */143 __u32 size; /* 8 Apparent size of each individual disk */144 __u32 nr_disks; /* 9 total disks in the raid set */145 __u32 raid_disks; /* 10 disks in a fully functional raid set */146 __u32 md_minor; /* 11 preferred MD minor device number */147 __u32 not_persistent; /* 12 does it have a persistent superblock */148 __u32 set_uuid1; /* 13 Raid set identifier #2 */149 __u32 set_uuid2; /* 14 Raid set identifier #3 */150 __u32 set_uuid3; /* 15 Raid set identifier #4 */151 __u32 gstate_creserved[MD_SB_GENERIC_CONSTANT_WORDS - 16];152 153 /*154 * Generic state information155 */156 __u32 utime; /* 0 Superblock update time */157 __u32 state; /* 1 State bits (clean, ...) */158 __u32 active_disks; /* 2 Number of currently active disks */159 __u32 working_disks; /* 3 Number of working disks */160 __u32 failed_disks; /* 4 Number of failed disks */161 __u32 spare_disks; /* 5 Number of spare disks */162 __u32 sb_csum; /* 6 checksum of the whole superblock */163#if defined(__BYTE_ORDER) ? __BYTE_ORDER == __BIG_ENDIAN : defined(__BIG_ENDIAN)164 __u32 events_hi; /* 7 high-order of superblock update count */165 __u32 events_lo; /* 8 low-order of superblock update count */166 __u32 cp_events_hi; /* 9 high-order of checkpoint update count */167 __u32 cp_events_lo; /* 10 low-order of checkpoint update count */168#elif defined(__BYTE_ORDER) ? __BYTE_ORDER == __LITTLE_ENDIAN : defined(__LITTLE_ENDIAN)169 __u32 events_lo; /* 7 low-order of superblock update count */170 __u32 events_hi; /* 8 high-order of superblock update count */171 __u32 cp_events_lo; /* 9 low-order of checkpoint update count */172 __u32 cp_events_hi; /* 10 high-order of checkpoint update count */173#else174#error unspecified endianness175#endif176 __u32 recovery_cp; /* 11 recovery checkpoint sector count */177 /* There are only valid for minor_version > 90 */178 __u64 reshape_position; /* 12,13 next address in array-space for reshape */179 __u32 new_level; /* 14 new level we are reshaping to */180 __u32 delta_disks; /* 15 change in number of raid_disks */181 __u32 new_layout; /* 16 new layout */182 __u32 new_chunk; /* 17 new chunk size (bytes) */183 __u32 gstate_sreserved[MD_SB_GENERIC_STATE_WORDS - 18];184 185 /*186 * Personality information187 */188 __u32 layout; /* 0 the array's physical layout */189 __u32 chunk_size; /* 1 chunk size in bytes */190 __u32 root_pv; /* 2 LV root PV */191 __u32 root_block; /* 3 LV root block */192 __u32 pstate_reserved[MD_SB_PERSONALITY_WORDS - 4];193 194 /*195 * Disks information196 */197 mdp_disk_t disks[MD_SB_DISKS];198 199 /*200 * Reserved201 */202 __u32 reserved[MD_SB_RESERVED_WORDS];203 204 /*205 * Active descriptor206 */207 mdp_disk_t this_disk;208 209} mdp_super_t;210 211static inline __u64 md_event(mdp_super_t *sb) {212 __u64 ev = sb->events_hi;213 return (ev<<32)| sb->events_lo;214}215 216#define MD_SUPERBLOCK_1_TIME_SEC_MASK ((1ULL<<40) - 1)217 218/*219 * The version-1 superblock :220 * All numeric fields are little-endian.221 *222 * total size: 256 bytes plus 2 per device.223 * 1K allows 384 devices.224 */225struct mdp_superblock_1 {226 /* constant array information - 128 bytes */227 __le32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */228 __le32 major_version; /* 1 */229 __le32 feature_map; /* bit 0 set if 'bitmap_offset' is meaningful */230 __le32 pad0; /* always set to 0 when writing */231 232 __u8 set_uuid[16]; /* user-space generated. */233 char set_name[32]; /* set and interpreted by user-space */234 235 __le64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/236 __le32 level; /* 0,1,4,5 */237 __le32 layout; /* only for raid5 and raid10 currently */238 __le64 size; /* used size of component devices, in 512byte sectors */239 240 __le32 chunksize; /* in 512byte sectors */241 __le32 raid_disks;242 union {243 __le32 bitmap_offset; /* sectors after start of superblock that bitmap starts244 * NOTE: signed, so bitmap can be before superblock245 * only meaningful of feature_map[0] is set.246 */247 248 /* only meaningful when feature_map[MD_FEATURE_PPL] is set */249 struct {250 __le16 offset; /* sectors from start of superblock that ppl starts (signed) */251 __le16 size; /* ppl size in sectors */252 } ppl;253 };254 255 /* These are only valid with feature bit '4' */256 __le32 new_level; /* new level we are reshaping to */257 __le64 reshape_position; /* next address in array-space for reshape */258 __le32 delta_disks; /* change in number of raid_disks */259 __le32 new_layout; /* new layout */260 __le32 new_chunk; /* new chunk size (512byte sectors) */261 __le32 new_offset; /* signed number to add to data_offset in new262 * layout. 0 == no-change. This can be263 * different on each device in the array.264 */265 266 /* constant this-device information - 64 bytes */267 __le64 data_offset; /* sector start of data, often 0 */268 __le64 data_size; /* sectors in this device that can be used for data */269 __le64 super_offset; /* sector start of this superblock */270 union {271 __le64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */272 __le64 journal_tail;/* journal tail of journal device (from data_offset) */273 };274 __le32 dev_number; /* permanent identifier of this device - not role in raid */275 __le32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */276 __u8 device_uuid[16]; /* user-space setable, ignored by kernel */277 __u8 devflags; /* per-device flags. Only two defined...*/278#define WriteMostly1 1 /* mask for writemostly flag in above */279#define FailFast1 2 /* Should avoid retries and fixups and just fail */280 /* Bad block log. If there are any bad blocks the feature flag is set.281 * If offset and size are non-zero, that space is reserved and available282 */283 __u8 bblog_shift; /* shift from sectors to block size */284 __le16 bblog_size; /* number of sectors reserved for list */285 __le32 bblog_offset; /* sector offset from superblock to bblog,286 * signed - not unsigned */287 288 /* array state information - 64 bytes */289 __le64 utime; /* 40 bits second, 24 bits microseconds */290 __le64 events; /* incremented when superblock updated */291 __le64 resync_offset; /* data before this offset (from data_offset) known to be in sync */292 __le32 sb_csum; /* checksum up to devs[max_dev] */293 __le32 max_dev; /* size of devs[] array to consider */294 __u8 pad3[64-32]; /* set to 0 when writing */295 296 /* device state information. Indexed by dev_number.297 * 2 bytes per device298 * Note there are no per-device state flags. State information is rolled299 * into the 'roles' value. If a device is spare or faulty, then it doesn't300 * have a meaningful role.301 */302 __le16 dev_roles[]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */303};304 305/* feature_map bits */306#define MD_FEATURE_BITMAP_OFFSET 1307#define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and308 * must be honoured309 */310#define MD_FEATURE_RESHAPE_ACTIVE 4311#define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */312#define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an313 * active device with same 'role'.314 * 'recovery_offset' is also set.315 */316#define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number317 * of devices, but is going318 * backwards anyway.319 */320#define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */321#define MD_FEATURE_RECOVERY_BITMAP 128 /* recovery that is happening322 * is guided by bitmap.323 */324#define MD_FEATURE_CLUSTERED 256 /* clustered MD */325#define MD_FEATURE_JOURNAL 512 /* support write cache */326#define MD_FEATURE_PPL 1024 /* support PPL */327#define MD_FEATURE_MULTIPLE_PPLS 2048 /* support for multiple PPLs */328#define MD_FEATURE_RAID0_LAYOUT 4096 /* layout is meaningful for RAID0 */329#define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \330 |MD_FEATURE_RECOVERY_OFFSET \331 |MD_FEATURE_RESHAPE_ACTIVE \332 |MD_FEATURE_BAD_BLOCKS \333 |MD_FEATURE_REPLACEMENT \334 |MD_FEATURE_RESHAPE_BACKWARDS \335 |MD_FEATURE_NEW_OFFSET \336 |MD_FEATURE_RECOVERY_BITMAP \337 |MD_FEATURE_CLUSTERED \338 |MD_FEATURE_JOURNAL \339 |MD_FEATURE_PPL \340 |MD_FEATURE_MULTIPLE_PPLS \341 |MD_FEATURE_RAID0_LAYOUT \342 )343 344struct r5l_payload_header {345 __le16 type;346 __le16 flags;347} __attribute__ ((__packed__));348 349enum r5l_payload_type {350 R5LOG_PAYLOAD_DATA = 0,351 R5LOG_PAYLOAD_PARITY = 1,352 R5LOG_PAYLOAD_FLUSH = 2,353};354 355struct r5l_payload_data_parity {356 struct r5l_payload_header header;357 __le32 size; /* sector. data/parity size. each 4k358 * has a checksum */359 __le64 location; /* sector. For data, it's raid sector. For360 * parity, it's stripe sector */361 __le32 checksum[];362} __attribute__ ((__packed__));363 364enum r5l_payload_data_parity_flag {365 R5LOG_PAYLOAD_FLAG_DISCARD = 1, /* payload is discard */366 /*367 * RESHAPED/RESHAPING is only set when there is reshape activity. Note,368 * both data/parity of a stripe should have the same flag set369 *370 * RESHAPED: reshape is running, and this stripe finished reshape371 * RESHAPING: reshape is running, and this stripe isn't reshaped372 */373 R5LOG_PAYLOAD_FLAG_RESHAPED = 2,374 R5LOG_PAYLOAD_FLAG_RESHAPING = 3,375};376 377struct r5l_payload_flush {378 struct r5l_payload_header header;379 __le32 size; /* flush_stripes size, bytes */380 __le64 flush_stripes[];381} __attribute__ ((__packed__));382 383enum r5l_payload_flush_flag {384 R5LOG_PAYLOAD_FLAG_FLUSH_STRIPE = 1, /* data represents whole stripe */385};386 387struct r5l_meta_block {388 __le32 magic;389 __le32 checksum;390 __u8 version;391 __u8 __zero_pading_1;392 __le16 __zero_pading_2;393 __le32 meta_size; /* whole size of the block */394 395 __le64 seq;396 __le64 position; /* sector, start from rdev->data_offset, current position */397 struct r5l_payload_header payloads[];398} __attribute__ ((__packed__));399 400#define R5LOG_VERSION 0x1401#define R5LOG_MAGIC 0x6433c509402 403struct ppl_header_entry {404 __le64 data_sector; /* raid sector of the new data */405 __le32 pp_size; /* length of partial parity */406 __le32 data_size; /* length of data */407 __le32 parity_disk; /* member disk containing parity */408 __le32 checksum; /* checksum of partial parity data for this409 * entry (~crc32c) */410} __attribute__ ((__packed__));411 412#define PPL_HEADER_SIZE 4096413#define PPL_HDR_RESERVED 512414#define PPL_HDR_ENTRY_SPACE \415 (PPL_HEADER_SIZE - PPL_HDR_RESERVED - 4 * sizeof(__le32) - sizeof(__le64))416#define PPL_HDR_MAX_ENTRIES \417 (PPL_HDR_ENTRY_SPACE / sizeof(struct ppl_header_entry))418 419struct ppl_header {420 __u8 reserved[PPL_HDR_RESERVED];/* reserved space, fill with 0xff */421 __le32 signature; /* signature (family number of volume) */422 __le32 padding; /* zero pad */423 __le64 generation; /* generation number of the header */424 __le32 entries_count; /* number of entries in entry array */425 __le32 checksum; /* checksum of the header (~crc32c) */426 struct ppl_header_entry entries[PPL_HDR_MAX_ENTRIES];427} __attribute__ ((__packed__));428 429#endif430