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.
87 #define MD_DISK_FAILFAST 10 /* Send REQ_FAILFAST if there are multiple
88 * devices available - and don't try to
89 * correct read errors.
92 #define MD_DISK_WRITEMOSTLY 9 /* disk is "write-mostly" is RAID1 config.
93 * read requests will only be sent here in
96 #define MD_DISK_JOURNAL 18 /* disk is used as the write journal in RAID-5/6 */
98 #define MD_DISK_ROLE_SPARE 0xffff
99 #define MD_DISK_ROLE_FAULTY 0xfffe
100 #define MD_DISK_ROLE_JOURNAL 0xfffd
101 #define MD_DISK_ROLE_MAX 0xff00 /* max value of regular disk role */
103 typedef struct mdp_device_descriptor_s {
104 __u32 number; /* 0 Device number in the entire set */
105 __u32 major; /* 1 Device major number */
106 __u32 minor; /* 2 Device minor number */
107 __u32 raid_disk; /* 3 The role of the device in the raid set */
108 __u32 state; /* 4 Operational state */
109 __u32 reserved[MD_SB_DESCRIPTOR_WORDS - 5];
112 #define MD_SB_MAGIC 0xa92b4efc
115 * Superblock state bits
117 #define MD_SB_CLEAN 0
118 #define MD_SB_ERRORS 1
120 #define MD_SB_CLUSTERED 5 /* MD is clustered */
121 #define MD_SB_BITMAP_PRESENT 8 /* bitmap may be present nearby */
125 * - if an array is being reshaped (restriped) in order to change the
126 * the number of active devices in the array, 'raid_disks' will be
127 * the larger of the old and new numbers. 'delta_disks' will
128 * be the "new - old". So if +ve, raid_disks is the new value, and
129 * "raid_disks-delta_disks" is the old. If -ve, raid_disks is the
130 * old value and "raid_disks+delta_disks" is the new (smaller) value.
134 typedef struct mdp_superblock_s {
136 * Constant generic information
138 __u32 md_magic; /* 0 MD identifier */
139 __u32 major_version; /* 1 major version to which the set conforms */
140 __u32 minor_version; /* 2 minor version ... */
141 __u32 patch_version; /* 3 patchlevel version ... */
142 __u32 gvalid_words; /* 4 Number of used words in this section */
143 __u32 set_uuid0; /* 5 Raid set identifier */
144 __u32 ctime; /* 6 Creation time */
145 __u32 level; /* 7 Raid personality */
146 __u32 size; /* 8 Apparent size of each individual disk */
147 __u32 nr_disks; /* 9 total disks in the raid set */
148 __u32 raid_disks; /* 10 disks in a fully functional raid set */
149 __u32 md_minor; /* 11 preferred MD minor device number */
150 __u32 not_persistent; /* 12 does it have a persistent superblock */
151 __u32 set_uuid1; /* 13 Raid set identifier #2 */
152 __u32 set_uuid2; /* 14 Raid set identifier #3 */
153 __u32 set_uuid3; /* 15 Raid set identifier #4 */
154 __u32 gstate_creserved[MD_SB_GENERIC_CONSTANT_WORDS - 16];
157 * Generic state information
159 __u32 utime; /* 0 Superblock update time */
160 __u32 state; /* 1 State bits (clean, ...) */
161 __u32 active_disks; /* 2 Number of currently active disks */
162 __u32 working_disks; /* 3 Number of working disks */
163 __u32 failed_disks; /* 4 Number of failed disks */
164 __u32 spare_disks; /* 5 Number of spare disks */
165 __u32 sb_csum; /* 6 checksum of the whole superblock */
166 #if defined(__BYTE_ORDER) ? __BYTE_ORDER == __BIG_ENDIAN : defined(__BIG_ENDIAN)
167 __u32 events_hi; /* 7 high-order of superblock update count */
168 __u32 events_lo; /* 8 low-order of superblock update count */
169 __u32 cp_events_hi; /* 9 high-order of checkpoint update count */
170 __u32 cp_events_lo; /* 10 low-order of checkpoint update count */
171 #elif defined(__BYTE_ORDER) ? __BYTE_ORDER == __LITTLE_ENDIAN : defined(__LITTLE_ENDIAN)
172 __u32 events_lo; /* 7 low-order of superblock update count */
173 __u32 events_hi; /* 8 high-order of superblock update count */
174 __u32 cp_events_lo; /* 9 low-order of checkpoint update count */
175 __u32 cp_events_hi; /* 10 high-order of checkpoint update count */
177 #error unspecified endianness
179 __u32 recovery_cp; /* 11 recovery checkpoint sector count */
180 /* There are only valid for minor_version > 90 */
181 __u64 reshape_position; /* 12,13 next address in array-space for reshape */
182 __u32 new_level; /* 14 new level we are reshaping to */
183 __u32 delta_disks; /* 15 change in number of raid_disks */
184 __u32 new_layout; /* 16 new layout */
185 __u32 new_chunk; /* 17 new chunk size (bytes) */
186 __u32 gstate_sreserved[MD_SB_GENERIC_STATE_WORDS - 18];
189 * Personality information
191 __u32 layout; /* 0 the array's physical layout */
192 __u32 chunk_size; /* 1 chunk size in bytes */
193 __u32 root_pv; /* 2 LV root PV */
194 __u32 root_block; /* 3 LV root block */
195 __u32 pstate_reserved[MD_SB_PERSONALITY_WORDS - 4];
200 mdp_disk_t disks[MD_SB_DISKS];
205 __u32 reserved[MD_SB_RESERVED_WORDS];
210 mdp_disk_t this_disk;
214 static inline __u64 md_event(mdp_super_t *sb) {
215 __u64 ev = sb->events_hi;
216 return (ev<<32)| sb->events_lo;
219 #define MD_SUPERBLOCK_1_TIME_SEC_MASK ((1ULL<<40) - 1)
222 * The version-1 superblock :
223 * All numeric fields are little-endian.
225 * total size: 256 bytes plus 2 per device.
226 * 1K allows 384 devices.
228 struct mdp_superblock_1 {
229 /* constant array information - 128 bytes */
230 __le32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */
231 __le32 major_version; /* 1 */
232 __le32 feature_map; /* bit 0 set if 'bitmap_offset' is meaningful */
233 __le32 pad0; /* always set to 0 when writing */
235 __u8 set_uuid[16]; /* user-space generated. */
236 char set_name[32]; /* set and interpreted by user-space */
238 __le64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/
239 __le32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */
240 __le32 layout; /* only for raid5 and raid10 currently */
241 __le64 size; /* used size of component devices, in 512byte sectors */
243 __le32 chunksize; /* in 512byte sectors */
246 __le32 bitmap_offset; /* sectors after start of superblock that bitmap starts
247 * NOTE: signed, so bitmap can be before superblock
248 * only meaningful of feature_map[0] is set.
251 /* only meaningful when feature_map[MD_FEATURE_PPL] is set */
253 __le16 offset; /* sectors from start of superblock that ppl starts (signed) */
254 __le16 size; /* ppl size in sectors */
258 /* These are only valid with feature bit '4' */
259 __le32 new_level; /* new level we are reshaping to */
260 __le64 reshape_position; /* next address in array-space for reshape */
261 __le32 delta_disks; /* change in number of raid_disks */
262 __le32 new_layout; /* new layout */
263 __le32 new_chunk; /* new chunk size (512byte sectors) */
264 __le32 new_offset; /* signed number to add to data_offset in new
265 * layout. 0 == no-change. This can be
266 * different on each device in the array.
269 /* constant this-device information - 64 bytes */
270 __le64 data_offset; /* sector start of data, often 0 */
271 __le64 data_size; /* sectors in this device that can be used for data */
272 __le64 super_offset; /* sector start of this superblock */
274 __le64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */
275 __le64 journal_tail;/* journal tail of journal device (from data_offset) */
277 __le32 dev_number; /* permanent identifier of this device - not role in raid */
278 __le32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */
279 __u8 device_uuid[16]; /* user-space setable, ignored by kernel */
280 __u8 devflags; /* per-device flags. Only two defined...*/
281 #define WriteMostly1 1 /* mask for writemostly flag in above */
282 #define FailFast1 2 /* Should avoid retries and fixups and just fail */
283 /* Bad block log. If there are any bad blocks the feature flag is set.
284 * If offset and size are non-zero, that space is reserved and available
286 __u8 bblog_shift; /* shift from sectors to block size */
287 __le16 bblog_size; /* number of sectors reserved for list */
288 __le32 bblog_offset; /* sector offset from superblock to bblog,
289 * signed - not unsigned */
291 /* array state information - 64 bytes */
292 __le64 utime; /* 40 bits second, 24 bits microseconds */
293 __le64 events; /* incremented when superblock updated */
294 __le64 resync_offset; /* data before this offset (from data_offset) known to be in sync */
295 __le32 sb_csum; /* checksum up to devs[max_dev] */
296 __le32 max_dev; /* size of devs[] array to consider */
297 __u8 pad3[64-32]; /* set to 0 when writing */
299 /* device state information. Indexed by dev_number.
301 * Note there are no per-device state flags. State information is rolled
302 * into the 'roles' value. If a device is spare or faulty, then it doesn't
303 * have a meaningful role.
305 __le16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
308 /* feature_map bits */
309 #define MD_FEATURE_BITMAP_OFFSET 1
310 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
313 #define MD_FEATURE_RESHAPE_ACTIVE 4
314 #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */
315 #define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an
316 * active device with same 'role'.
317 * 'recovery_offset' is also set.
319 #define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number
320 * of devices, but is going
323 #define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */
324 #define MD_FEATURE_RECOVERY_BITMAP 128 /* recovery that is happening
325 * is guided by bitmap.
327 #define MD_FEATURE_CLUSTERED 256 /* clustered MD */
328 #define MD_FEATURE_JOURNAL 512 /* support write cache */
329 #define MD_FEATURE_PPL 1024 /* support PPL */
330 #define MD_FEATURE_MULTIPLE_PPLS 2048 /* support for multiple PPLs */
331 #define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \
332 |MD_FEATURE_RECOVERY_OFFSET \
333 |MD_FEATURE_RESHAPE_ACTIVE \
334 |MD_FEATURE_BAD_BLOCKS \
335 |MD_FEATURE_REPLACEMENT \
336 |MD_FEATURE_RESHAPE_BACKWARDS \
337 |MD_FEATURE_NEW_OFFSET \
338 |MD_FEATURE_RECOVERY_BITMAP \
339 |MD_FEATURE_CLUSTERED \
340 |MD_FEATURE_JOURNAL \
342 |MD_FEATURE_MULTIPLE_PPLS \
345 struct r5l_payload_header {
348 } __attribute__ ((__packed__));
350 enum r5l_payload_type {
351 R5LOG_PAYLOAD_DATA = 0,
352 R5LOG_PAYLOAD_PARITY = 1,
353 R5LOG_PAYLOAD_FLUSH = 2,
356 struct r5l_payload_data_parity {
357 struct r5l_payload_header header;
358 __le32 size; /* sector. data/parity size. each 4k
360 __le64 location; /* sector. For data, it's raid sector. For
361 * parity, it's stripe sector */
363 } __attribute__ ((__packed__));
365 enum r5l_payload_data_parity_flag {
366 R5LOG_PAYLOAD_FLAG_DISCARD = 1, /* payload is discard */
368 * RESHAPED/RESHAPING is only set when there is reshape activity. Note,
369 * both data/parity of a stripe should have the same flag set
371 * RESHAPED: reshape is running, and this stripe finished reshape
372 * RESHAPING: reshape is running, and this stripe isn't reshaped
374 R5LOG_PAYLOAD_FLAG_RESHAPED = 2,
375 R5LOG_PAYLOAD_FLAG_RESHAPING = 3,
378 struct r5l_payload_flush {
379 struct r5l_payload_header header;
380 __le32 size; /* flush_stripes size, bytes */
381 __le64 flush_stripes[];
382 } __attribute__ ((__packed__));
384 enum r5l_payload_flush_flag {
385 R5LOG_PAYLOAD_FLAG_FLUSH_STRIPE = 1, /* data represents whole stripe */
388 struct r5l_meta_block {
392 __u8 __zero_pading_1;
393 __le16 __zero_pading_2;
394 __le32 meta_size; /* whole size of the block */
397 __le64 position; /* sector, start from rdev->data_offset, current position */
398 struct r5l_payload_header payloads[];
399 } __attribute__ ((__packed__));
401 #define R5LOG_VERSION 0x1
402 #define R5LOG_MAGIC 0x6433c509
404 struct ppl_header_entry {
405 __le64 data_sector; /* raid sector of the new data */
406 __le32 pp_size; /* length of partial parity */
407 __le32 data_size; /* length of data */
408 __le32 parity_disk; /* member disk containing parity */
409 __le32 checksum; /* checksum of partial parity data for this
411 } __attribute__ ((__packed__));
413 #define PPL_HEADER_SIZE 4096
414 #define PPL_HDR_RESERVED 512
415 #define PPL_HDR_ENTRY_SPACE \
416 (PPL_HEADER_SIZE - PPL_HDR_RESERVED - 4 * sizeof(__le32) - sizeof(__le64))
417 #define PPL_HDR_MAX_ENTRIES \
418 (PPL_HDR_ENTRY_SPACE / sizeof(struct ppl_header_entry))
421 __u8 reserved[PPL_HDR_RESERVED];/* reserved space, fill with 0xff */
422 __le32 signature; /* signature (family number of volume) */
423 __le32 padding; /* zero pad */
424 __le64 generation; /* generation number of the header */
425 __le32 entries_count; /* number of entries in entry array */
426 __le32 checksum; /* checksum of the header (~crc32c) */
427 struct ppl_header_entry entries[PPL_HDR_MAX_ENTRIES];
428 } __attribute__ ((__packed__));