1 | /* SPDX-License-Identifier: GPL-2.0+ WITH Linux-syscall-note */ |
2 | /* |
3 | md_p.h : physical layout of Linux RAID devices |
4 | Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman |
5 | |
6 | This program is free software; you can redistribute it and/or modify |
7 | it under the terms of the GNU General Public License as published by |
8 | the Free Software Foundation; either version 2, or (at your option) |
9 | any later version. |
10 | |
11 | You should have received a copy of the GNU General Public License |
12 | (for example /usr/src/linux/COPYING); if not, write to the Free |
13 | Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
14 | */ |
15 | |
16 | #ifndef _MD_P_H |
17 | #define _MD_P_H |
18 | |
19 | #include <linux/types.h> |
20 | #include <asm/byteorder.h> |
21 | |
22 | /* |
23 | * RAID superblock. |
24 | * |
25 | * The RAID superblock maintains some statistics on each RAID configuration. |
26 | * Each real device in the RAID set contains it near the end of the device. |
27 | * Some of the ideas are copied from the ext2fs implementation. |
28 | * |
29 | * We currently use 4096 bytes as follows: |
30 | * |
31 | * word offset function |
32 | * |
33 | * 0 - 31 Constant generic RAID device information. |
34 | * 32 - 63 Generic state information. |
35 | * 64 - 127 Personality specific information. |
36 | * 128 - 511 12 32-words descriptors of the disks in the raid set. |
37 | * 512 - 911 Reserved. |
38 | * 912 - 1023 Disk specific descriptor. |
39 | */ |
40 | |
41 | /* |
42 | * If x is the real device size in bytes, we return an apparent size of: |
43 | * |
44 | * y = (x & ~(MD_RESERVED_BYTES - 1)) - MD_RESERVED_BYTES |
45 | * |
46 | * and place the 4kB superblock at offset y. |
47 | */ |
48 | #define MD_RESERVED_BYTES (64 * 1024) |
49 | #define MD_RESERVED_SECTORS (MD_RESERVED_BYTES / 512) |
50 | |
51 | #define MD_NEW_SIZE_SECTORS(x) ((x & ~(MD_RESERVED_SECTORS - 1)) - MD_RESERVED_SECTORS) |
52 | |
53 | #define MD_SB_BYTES 4096 |
54 | #define MD_SB_WORDS (MD_SB_BYTES / 4) |
55 | #define MD_SB_SECTORS (MD_SB_BYTES / 512) |
56 | |
57 | /* |
58 | * The following are counted in 32-bit words |
59 | */ |
60 | #define MD_SB_GENERIC_OFFSET 0 |
61 | #define MD_SB_PERSONALITY_OFFSET 64 |
62 | #define MD_SB_DISKS_OFFSET 128 |
63 | #define MD_SB_DESCRIPTOR_OFFSET 992 |
64 | |
65 | #define MD_SB_GENERIC_CONSTANT_WORDS 32 |
66 | #define MD_SB_GENERIC_STATE_WORDS 32 |
67 | #define MD_SB_GENERIC_WORDS (MD_SB_GENERIC_CONSTANT_WORDS + MD_SB_GENERIC_STATE_WORDS) |
68 | #define MD_SB_PERSONALITY_WORDS 64 |
69 | #define MD_SB_DESCRIPTOR_WORDS 32 |
70 | #define MD_SB_DISKS 27 |
71 | #define MD_SB_DISKS_WORDS (MD_SB_DISKS*MD_SB_DESCRIPTOR_WORDS) |
72 | #define MD_SB_RESERVED_WORDS (1024 - MD_SB_GENERIC_WORDS - MD_SB_PERSONALITY_WORDS - MD_SB_DISKS_WORDS - MD_SB_DESCRIPTOR_WORDS) |
73 | #define MD_SB_EQUAL_WORDS (MD_SB_GENERIC_WORDS + MD_SB_PERSONALITY_WORDS + MD_SB_DISKS_WORDS) |
74 | |
75 | /* |
76 | * Device "operational" state bits |
77 | */ |
78 | #define MD_DISK_FAULTY 0 /* disk is faulty / operational */ |
79 | #define MD_DISK_ACTIVE 1 /* disk is running or spare disk */ |
80 | #define MD_DISK_SYNC 2 /* disk is in sync with the raid set */ |
81 | #define MD_DISK_REMOVED 3 /* disk is in sync with the raid set */ |
82 | #define MD_DISK_CLUSTER_ADD 4 /* Initiate a disk add across the cluster |
83 | * For clustered enviroments only. |
84 | */ |
85 | #define MD_DISK_CANDIDATE 5 /* disk is added as spare (local) until confirmed |
86 | * For clustered enviroments only. |
87 | */ |
88 | #define MD_DISK_FAILFAST 10 /* Send REQ_FAILFAST if there are multiple |
89 | * devices available - and don't try to |
90 | * correct read errors. |
91 | */ |
92 | |
93 | #define MD_DISK_WRITEMOSTLY 9 /* disk is "write-mostly" is RAID1 config. |
94 | * read requests will only be sent here in |
95 | * dire need |
96 | */ |
97 | #define MD_DISK_JOURNAL 18 /* disk is used as the write journal in RAID-5/6 */ |
98 | |
99 | #define MD_DISK_ROLE_SPARE 0xffff |
100 | #define MD_DISK_ROLE_FAULTY 0xfffe |
101 | #define MD_DISK_ROLE_JOURNAL 0xfffd |
102 | #define MD_DISK_ROLE_MAX 0xff00 /* max value of regular disk role */ |
103 | |
104 | typedef struct mdp_device_descriptor_s { |
105 | __u32 number; /* 0 Device number in the entire set */ |
106 | __u32 major; /* 1 Device major number */ |
107 | __u32 minor; /* 2 Device minor number */ |
108 | __u32 raid_disk; /* 3 The role of the device in the raid set */ |
109 | __u32 state; /* 4 Operational state */ |
110 | __u32 reserved[MD_SB_DESCRIPTOR_WORDS - 5]; |
111 | } mdp_disk_t; |
112 | |
113 | #define MD_SB_MAGIC 0xa92b4efc |
114 | |
115 | /* |
116 | * Superblock state bits |
117 | */ |
118 | #define MD_SB_CLEAN 0 |
119 | #define MD_SB_ERRORS 1 |
120 | |
121 | #define MD_SB_CLUSTERED 5 /* MD is clustered */ |
122 | #define MD_SB_BITMAP_PRESENT 8 /* bitmap may be present nearby */ |
123 | |
124 | /* |
125 | * Notes: |
126 | * - if an array is being reshaped (restriped) in order to change |
127 | * the number of active devices in the array, 'raid_disks' will be |
128 | * the larger of the old and new numbers. 'delta_disks' will |
129 | * be the "new - old". So if +ve, raid_disks is the new value, and |
130 | * "raid_disks-delta_disks" is the old. If -ve, raid_disks is the |
131 | * old value and "raid_disks+delta_disks" is the new (smaller) value. |
132 | */ |
133 | |
134 | |
135 | typedef struct mdp_superblock_s { |
136 | /* |
137 | * Constant generic information |
138 | */ |
139 | __u32 md_magic; /* 0 MD identifier */ |
140 | __u32 major_version; /* 1 major version to which the set conforms */ |
141 | __u32 minor_version; /* 2 minor version ... */ |
142 | __u32 patch_version; /* 3 patchlevel version ... */ |
143 | __u32 gvalid_words; /* 4 Number of used words in this section */ |
144 | __u32 set_uuid0; /* 5 Raid set identifier */ |
145 | __u32 ctime; /* 6 Creation time */ |
146 | __u32 level; /* 7 Raid personality */ |
147 | __u32 size; /* 8 Apparent size of each individual disk */ |
148 | __u32 nr_disks; /* 9 total disks in the raid set */ |
149 | __u32 raid_disks; /* 10 disks in a fully functional raid set */ |
150 | __u32 md_minor; /* 11 preferred MD minor device number */ |
151 | __u32 not_persistent; /* 12 does it have a persistent superblock */ |
152 | __u32 set_uuid1; /* 13 Raid set identifier #2 */ |
153 | __u32 set_uuid2; /* 14 Raid set identifier #3 */ |
154 | __u32 set_uuid3; /* 15 Raid set identifier #4 */ |
155 | __u32 gstate_creserved[MD_SB_GENERIC_CONSTANT_WORDS - 16]; |
156 | |
157 | /* |
158 | * Generic state information |
159 | */ |
160 | __u32 utime; /* 0 Superblock update time */ |
161 | __u32 state; /* 1 State bits (clean, ...) */ |
162 | __u32 active_disks; /* 2 Number of currently active disks */ |
163 | __u32 working_disks; /* 3 Number of working disks */ |
164 | __u32 failed_disks; /* 4 Number of failed disks */ |
165 | __u32 spare_disks; /* 5 Number of spare disks */ |
166 | __u32 sb_csum; /* 6 checksum of the whole superblock */ |
167 | #if defined(__BYTE_ORDER) ? __BYTE_ORDER == __BIG_ENDIAN : defined(__BIG_ENDIAN) |
168 | __u32 events_hi; /* 7 high-order of superblock update count */ |
169 | __u32 events_lo; /* 8 low-order of superblock update count */ |
170 | __u32 cp_events_hi; /* 9 high-order of checkpoint update count */ |
171 | __u32 cp_events_lo; /* 10 low-order of checkpoint update count */ |
172 | #elif defined(__BYTE_ORDER) ? __BYTE_ORDER == __LITTLE_ENDIAN : defined(__LITTLE_ENDIAN) |
173 | __u32 events_lo; /* 7 low-order of superblock update count */ |
174 | __u32 events_hi; /* 8 high-order of superblock update count */ |
175 | __u32 cp_events_lo; /* 9 low-order of checkpoint update count */ |
176 | __u32 cp_events_hi; /* 10 high-order of checkpoint update count */ |
177 | #else |
178 | #error unspecified endianness |
179 | #endif |
180 | __u32 recovery_cp; /* 11 recovery checkpoint sector count */ |
181 | /* There are only valid for minor_version > 90 */ |
182 | __u64 reshape_position; /* 12,13 next address in array-space for reshape */ |
183 | __u32 new_level; /* 14 new level we are reshaping to */ |
184 | __u32 delta_disks; /* 15 change in number of raid_disks */ |
185 | __u32 new_layout; /* 16 new layout */ |
186 | __u32 new_chunk; /* 17 new chunk size (bytes) */ |
187 | __u32 gstate_sreserved[MD_SB_GENERIC_STATE_WORDS - 18]; |
188 | |
189 | /* |
190 | * Personality information |
191 | */ |
192 | __u32 layout; /* 0 the array's physical layout */ |
193 | __u32 chunk_size; /* 1 chunk size in bytes */ |
194 | __u32 root_pv; /* 2 LV root PV */ |
195 | __u32 root_block; /* 3 LV root block */ |
196 | __u32 pstate_reserved[MD_SB_PERSONALITY_WORDS - 4]; |
197 | |
198 | /* |
199 | * Disks information |
200 | */ |
201 | mdp_disk_t disks[MD_SB_DISKS]; |
202 | |
203 | /* |
204 | * Reserved |
205 | */ |
206 | __u32 reserved[MD_SB_RESERVED_WORDS]; |
207 | |
208 | /* |
209 | * Active descriptor |
210 | */ |
211 | mdp_disk_t this_disk; |
212 | |
213 | } mdp_super_t; |
214 | |
215 | static inline __u64 md_event(mdp_super_t *sb) { |
216 | __u64 ev = sb->events_hi; |
217 | return (ev<<32)| sb->events_lo; |
218 | } |
219 | |
220 | #define MD_SUPERBLOCK_1_TIME_SEC_MASK ((1ULL<<40) - 1) |
221 | |
222 | /* |
223 | * The version-1 superblock : |
224 | * All numeric fields are little-endian. |
225 | * |
226 | * total size: 256 bytes plus 2 per device. |
227 | * 1K allows 384 devices. |
228 | */ |
229 | struct mdp_superblock_1 { |
230 | /* constant array information - 128 bytes */ |
231 | __le32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */ |
232 | __le32 major_version; /* 1 */ |
233 | __le32 feature_map; /* bit 0 set if 'bitmap_offset' is meaningful */ |
234 | __le32 pad0; /* always set to 0 when writing */ |
235 | |
236 | __u8 set_uuid[16]; /* user-space generated. */ |
237 | char set_name[32]; /* set and interpreted by user-space */ |
238 | |
239 | __le64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/ |
240 | __le32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */ |
241 | __le32 layout; /* only for raid5 and raid10 currently */ |
242 | __le64 size; /* used size of component devices, in 512byte sectors */ |
243 | |
244 | __le32 chunksize; /* in 512byte sectors */ |
245 | __le32 raid_disks; |
246 | union { |
247 | __le32 bitmap_offset; /* sectors after start of superblock that bitmap starts |
248 | * NOTE: signed, so bitmap can be before superblock |
249 | * only meaningful of feature_map[0] is set. |
250 | */ |
251 | |
252 | /* only meaningful when feature_map[MD_FEATURE_PPL] is set */ |
253 | struct { |
254 | __le16 offset; /* sectors from start of superblock that ppl starts (signed) */ |
255 | __le16 size; /* ppl size in sectors */ |
256 | } ppl; |
257 | }; |
258 | |
259 | /* These are only valid with feature bit '4' */ |
260 | __le32 new_level; /* new level we are reshaping to */ |
261 | __le64 reshape_position; /* next address in array-space for reshape */ |
262 | __le32 delta_disks; /* change in number of raid_disks */ |
263 | __le32 new_layout; /* new layout */ |
264 | __le32 new_chunk; /* new chunk size (512byte sectors) */ |
265 | __le32 new_offset; /* signed number to add to data_offset in new |
266 | * layout. 0 == no-change. This can be |
267 | * different on each device in the array. |
268 | */ |
269 | |
270 | /* constant this-device information - 64 bytes */ |
271 | __le64 data_offset; /* sector start of data, often 0 */ |
272 | __le64 data_size; /* sectors in this device that can be used for data */ |
273 | __le64 super_offset; /* sector start of this superblock */ |
274 | union { |
275 | __le64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */ |
276 | __le64 journal_tail;/* journal tail of journal device (from data_offset) */ |
277 | }; |
278 | __le32 dev_number; /* permanent identifier of this device - not role in raid */ |
279 | __le32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */ |
280 | __u8 device_uuid[16]; /* user-space setable, ignored by kernel */ |
281 | __u8 devflags; /* per-device flags. Only two defined...*/ |
282 | #define WriteMostly1 1 /* mask for writemostly flag in above */ |
283 | #define FailFast1 2 /* Should avoid retries and fixups and just fail */ |
284 | /* Bad block log. If there are any bad blocks the feature flag is set. |
285 | * If offset and size are non-zero, that space is reserved and available |
286 | */ |
287 | __u8 bblog_shift; /* shift from sectors to block size */ |
288 | __le16 bblog_size; /* number of sectors reserved for list */ |
289 | __le32 bblog_offset; /* sector offset from superblock to bblog, |
290 | * signed - not unsigned */ |
291 | |
292 | /* array state information - 64 bytes */ |
293 | __le64 utime; /* 40 bits second, 24 bits microseconds */ |
294 | __le64 events; /* incremented when superblock updated */ |
295 | __le64 resync_offset; /* data before this offset (from data_offset) known to be in sync */ |
296 | __le32 sb_csum; /* checksum up to devs[max_dev] */ |
297 | __le32 max_dev; /* size of devs[] array to consider */ |
298 | __u8 pad3[64-32]; /* set to 0 when writing */ |
299 | |
300 | /* device state information. Indexed by dev_number. |
301 | * 2 bytes per device |
302 | * Note there are no per-device state flags. State information is rolled |
303 | * into the 'roles' value. If a device is spare or faulty, then it doesn't |
304 | * have a meaningful role. |
305 | */ |
306 | __le16 dev_roles[]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */ |
307 | }; |
308 | |
309 | /* feature_map bits */ |
310 | #define MD_FEATURE_BITMAP_OFFSET 1 |
311 | #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and |
312 | * must be honoured |
313 | */ |
314 | #define MD_FEATURE_RESHAPE_ACTIVE 4 |
315 | #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */ |
316 | #define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an |
317 | * active device with same 'role'. |
318 | * 'recovery_offset' is also set. |
319 | */ |
320 | #define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number |
321 | * of devices, but is going |
322 | * backwards anyway. |
323 | */ |
324 | #define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */ |
325 | #define MD_FEATURE_RECOVERY_BITMAP 128 /* recovery that is happening |
326 | * is guided by bitmap. |
327 | */ |
328 | #define MD_FEATURE_CLUSTERED 256 /* clustered MD */ |
329 | #define MD_FEATURE_JOURNAL 512 /* support write cache */ |
330 | #define MD_FEATURE_PPL 1024 /* support PPL */ |
331 | #define MD_FEATURE_MULTIPLE_PPLS 2048 /* support for multiple PPLs */ |
332 | #define MD_FEATURE_RAID0_LAYOUT 4096 /* layout is meaningful for RAID0 */ |
333 | #define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \ |
334 | |MD_FEATURE_RECOVERY_OFFSET \ |
335 | |MD_FEATURE_RESHAPE_ACTIVE \ |
336 | |MD_FEATURE_BAD_BLOCKS \ |
337 | |MD_FEATURE_REPLACEMENT \ |
338 | |MD_FEATURE_RESHAPE_BACKWARDS \ |
339 | |MD_FEATURE_NEW_OFFSET \ |
340 | |MD_FEATURE_RECOVERY_BITMAP \ |
341 | |MD_FEATURE_CLUSTERED \ |
342 | |MD_FEATURE_JOURNAL \ |
343 | |MD_FEATURE_PPL \ |
344 | |MD_FEATURE_MULTIPLE_PPLS \ |
345 | |MD_FEATURE_RAID0_LAYOUT \ |
346 | ) |
347 | |
348 | struct { |
349 | __le16 ; |
350 | __le16 ; |
351 | } __attribute__ ((__packed__)); |
352 | |
353 | enum r5l_payload_type { |
354 | R5LOG_PAYLOAD_DATA = 0, |
355 | R5LOG_PAYLOAD_PARITY = 1, |
356 | R5LOG_PAYLOAD_FLUSH = 2, |
357 | }; |
358 | |
359 | struct r5l_payload_data_parity { |
360 | struct r5l_payload_header ; |
361 | __le32 size; /* sector. data/parity size. each 4k |
362 | * has a checksum */ |
363 | __le64 location; /* sector. For data, it's raid sector. For |
364 | * parity, it's stripe sector */ |
365 | __le32 checksum[]; |
366 | } __attribute__ ((__packed__)); |
367 | |
368 | enum r5l_payload_data_parity_flag { |
369 | R5LOG_PAYLOAD_FLAG_DISCARD = 1, /* payload is discard */ |
370 | /* |
371 | * RESHAPED/RESHAPING is only set when there is reshape activity. Note, |
372 | * both data/parity of a stripe should have the same flag set |
373 | * |
374 | * RESHAPED: reshape is running, and this stripe finished reshape |
375 | * RESHAPING: reshape is running, and this stripe isn't reshaped |
376 | */ |
377 | R5LOG_PAYLOAD_FLAG_RESHAPED = 2, |
378 | R5LOG_PAYLOAD_FLAG_RESHAPING = 3, |
379 | }; |
380 | |
381 | struct r5l_payload_flush { |
382 | struct r5l_payload_header ; |
383 | __le32 size; /* flush_stripes size, bytes */ |
384 | __le64 flush_stripes[]; |
385 | } __attribute__ ((__packed__)); |
386 | |
387 | enum r5l_payload_flush_flag { |
388 | R5LOG_PAYLOAD_FLAG_FLUSH_STRIPE = 1, /* data represents whole stripe */ |
389 | }; |
390 | |
391 | struct r5l_meta_block { |
392 | __le32 magic; |
393 | __le32 checksum; |
394 | __u8 version; |
395 | __u8 __zero_pading_1; |
396 | __le16 __zero_pading_2; |
397 | __le32 meta_size; /* whole size of the block */ |
398 | |
399 | __le64 seq; |
400 | __le64 position; /* sector, start from rdev->data_offset, current position */ |
401 | struct r5l_payload_header payloads[]; |
402 | } __attribute__ ((__packed__)); |
403 | |
404 | #define R5LOG_VERSION 0x1 |
405 | #define R5LOG_MAGIC 0x6433c509 |
406 | |
407 | struct { |
408 | __le64 ; /* raid sector of the new data */ |
409 | __le32 ; /* length of partial parity */ |
410 | __le32 ; /* length of data */ |
411 | __le32 ; /* member disk containing parity */ |
412 | __le32 ; /* checksum of partial parity data for this |
413 | * entry (~crc32c) */ |
414 | } __attribute__ ((__packed__)); |
415 | |
416 | #define 4096 |
417 | #define PPL_HDR_RESERVED 512 |
418 | #define PPL_HDR_ENTRY_SPACE \ |
419 | (PPL_HEADER_SIZE - PPL_HDR_RESERVED - 4 * sizeof(__le32) - sizeof(__le64)) |
420 | #define PPL_HDR_MAX_ENTRIES \ |
421 | (PPL_HDR_ENTRY_SPACE / sizeof(struct ppl_header_entry)) |
422 | |
423 | struct { |
424 | __u8 [PPL_HDR_RESERVED];/* reserved space, fill with 0xff */ |
425 | __le32 ; /* signature (family number of volume) */ |
426 | __le32 ; /* zero pad */ |
427 | __le64 ; /* generation number of the header */ |
428 | __le32 ; /* number of entries in entry array */ |
429 | __le32 ; /* checksum of the header (~crc32c) */ |
430 | struct ppl_header_entry [PPL_HDR_MAX_ENTRIES]; |
431 | } __attribute__ ((__packed__)); |
432 | |
433 | #endif |
434 | |