]>
Commit | Line | Data |
---|---|---|
82d9eba6 NB |
1 | /* |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
4 | * Copyright (C) 2001-2004 Neil Brown <neilb@cse.unsw.edu.au> | |
5 | * | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License as published by | |
9 | * the Free Software Foundation; either version 2 of the License, or | |
10 | * (at your option) any later version. | |
11 | * | |
12 | * This program is distributed in the hope that it will be useful, | |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | * GNU General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with this program; if not, write to the Free Software | |
19 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
20 | * | |
21 | * Author: Neil Brown | |
22 | * Email: <neilb@cse.unsw.edu.au> | |
23 | * Paper: Neil Brown | |
24 | * School of Computer Science and Engineering | |
25 | * The University of New South Wales | |
26 | * Sydney, 2052 | |
27 | * Australia | |
28 | */ | |
29 | ||
30 | #include "mdadm.h" | |
f277ce36 | 31 | #include <endian.h> |
82d9eba6 NB |
32 | #include "asm/byteorder.h" |
33 | /* | |
34 | * The version-1 superblock : | |
35 | * All numeric fields are little-endian. | |
36 | * | |
37 | * total size: 256 bytes plus 2 per device. | |
38 | * 1K allows 384 devices. | |
39 | */ | |
40 | struct mdp_superblock_1 { | |
41 | /* constant array information - 128 bytes */ | |
42 | __u32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */ | |
43 | __u32 major_version; /* 1 */ | |
44 | __u32 feature_map; /* 0 for now */ | |
45 | __u32 pad0; /* always set to 0 when writing */ | |
46 | ||
47 | __u8 set_uuid[16]; /* user-space generated. */ | |
48 | char set_name[32]; /* set and interpreted by user-space */ | |
49 | ||
50 | __u64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/ | |
51 | __u32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */ | |
52 | __u32 layout; /* only for raid5 currently */ | |
53 | __u64 size; /* used size of component devices, in 512byte sectors */ | |
54 | ||
55 | __u32 chunksize; /* in 512byte sectors */ | |
56 | __u32 raid_disks; | |
34163fc7 NB |
57 | __u32 bitmap_offset; /* sectors after start of superblock that bitmap starts |
58 | * NOTE: signed, so bitmap can be before superblock | |
59 | * only meaningful of feature_map[0] is set. | |
60 | */ | |
61 | __u8 pad1[128-100]; /* set to 0 when written */ | |
82d9eba6 NB |
62 | |
63 | /* constant this-device information - 64 bytes */ | |
64 | __u64 data_offset; /* sector start of data, often 0 */ | |
65 | __u64 data_size; /* sectors in this device that can be used for data */ | |
66 | __u64 super_offset; /* sector start of this superblock */ | |
67 | __u64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */ | |
68 | __u32 dev_number; /* permanent identifier of this device - not role in raid */ | |
69 | __u32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */ | |
70 | __u8 device_uuid[16]; /* user-space setable, ignored by kernel */ | |
dfd4d8ee NB |
71 | __u8 devflags; /* per-device flags. Only one defined...*/ |
72 | #define WriteMostly1 1 /* mask for writemostly flag in above */ | |
73 | __u8 pad2[64-57]; /* set to 0 when writing */ | |
82d9eba6 NB |
74 | |
75 | /* array state information - 64 bytes */ | |
76 | __u64 utime; /* 40 bits second, 24 btes microseconds */ | |
77 | __u64 events; /* incremented when superblock updated */ | |
78 | __u64 resync_offset; /* data before this offset (from data_offset) known to be in sync */ | |
79 | __u32 sb_csum; /* checksum upto devs[max_dev] */ | |
80 | __u32 max_dev; /* size of devs[] array to consider */ | |
81 | __u8 pad3[64-32]; /* set to 0 when writing */ | |
82 | ||
83 | /* device state information. Indexed by dev_number. | |
84 | * 2 bytes per device | |
85 | * Note there are no per-device state flags. State information is rolled | |
86 | * into the 'roles' value. If a device is spare or faulty, then it doesn't | |
87 | * have a meaningful role. | |
88 | */ | |
89 | __u16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */ | |
90 | }; | |
91 | ||
570c0542 | 92 | #ifndef offsetof |
82d9eba6 | 93 | #define offsetof(t,f) ((int)&(((t*)0)->f)) |
570c0542 | 94 | #endif |
82d9eba6 NB |
95 | static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb) |
96 | { | |
97 | unsigned int disk_csum, csum; | |
98 | unsigned long long newcsum; | |
99 | int size = sizeof(*sb) + __le32_to_cpu(sb->max_dev)*2; | |
100 | unsigned int *isuper = (unsigned int*)sb; | |
101 | int i; | |
102 | ||
103 | /* make sure I can count... */ | |
104 | if (offsetof(struct mdp_superblock_1,data_offset) != 128 || | |
105 | offsetof(struct mdp_superblock_1, utime) != 192 || | |
106 | sizeof(struct mdp_superblock_1) != 256) { | |
107 | fprintf(stderr, "WARNING - superblock isn't sized correctly\n"); | |
108 | } | |
109 | ||
110 | disk_csum = sb->sb_csum; | |
111 | sb->sb_csum = 0; | |
112 | newcsum = 0; | |
113 | for (i=0; size>=4; size -= 4 ) | |
114 | newcsum += __le32_to_cpu(*isuper++); | |
115 | ||
116 | if (size == 2) | |
117 | newcsum += __le16_to_cpu(*(unsigned short*) isuper); | |
118 | ||
119 | csum = (newcsum & 0xffffffff) + (newcsum >> 32); | |
120 | sb->sb_csum = disk_csum; | |
121 | return csum; | |
122 | } | |
123 | ||
c7654afc | 124 | #ifndef MDASSEMBLE |
82d9eba6 NB |
125 | static void examine_super1(void *sbv) |
126 | { | |
127 | struct mdp_superblock_1 *sb = sbv; | |
128 | time_t atime; | |
129 | int d; | |
6fbba4c9 | 130 | int faulty; |
82d9eba6 NB |
131 | int i; |
132 | char *c; | |
133 | ||
134 | printf(" Magic : %08x\n", __le32_to_cpu(sb->magic)); | |
135 | printf(" Version : %02d.%02d\n", 1, __le32_to_cpu(sb->feature_map)); | |
136 | printf(" Array UUID : "); | |
137 | for (i=0; i<16; i++) { | |
82d9eba6 | 138 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 139 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 NB |
140 | } |
141 | printf("\n"); | |
142 | printf(" Name : %.32s\n", sb->set_name); | |
143 | ||
144 | atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL; | |
145 | printf(" Creation Time : %.24s\n", ctime(&atime)); | |
146 | c=map_num(pers, __le32_to_cpu(sb->level)); | |
147 | printf(" Raid Level : %s\n", c?c:"-unknown-"); | |
148 | printf(" Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks)); | |
149 | printf("\n"); | |
570c0542 | 150 | printf(" Device Size : %llu%s\n", (unsigned long long)sb->data_size, human_size(sb->data_size<<9)); |
82d9eba6 | 151 | if (sb->data_offset) |
570c0542 | 152 | printf(" Data Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->data_offset)); |
82d9eba6 | 153 | if (sb->super_offset) |
570c0542 | 154 | printf(" Super Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->super_offset)); |
a3fd117c | 155 | printf(" State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean"); |
82d9eba6 NB |
156 | printf(" Device UUID : "); |
157 | for (i=0; i<16; i++) { | |
82d9eba6 | 158 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 159 | printf("%02x", sb->device_uuid[i]); |
82d9eba6 NB |
160 | } |
161 | printf("\n"); | |
dfd4d8ee NB |
162 | if (sb->devflags) { |
163 | printf(" Flags :"); | |
34163fc7 | 164 | if (sb->devflags & WriteMostly1) |
dfd4d8ee NB |
165 | printf(" write-mostly"); |
166 | printf("\n"); | |
167 | } | |
82d9eba6 NB |
168 | |
169 | atime = __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL; | |
170 | printf(" Update Time : %.24s\n", ctime(&atime)); | |
171 | ||
172 | if (calc_sb_1_csum(sb) == sb->sb_csum) | |
173 | printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum)); | |
174 | else | |
175 | printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb->sb_csum), | |
176 | __le32_to_cpu(calc_sb_1_csum(sb))); | |
570c0542 | 177 | printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb->events)); |
82d9eba6 NB |
178 | printf("\n"); |
179 | if (__le32_to_cpu(sb->level) == 5) { | |
180 | c = map_num(r5layout, __le32_to_cpu(sb->layout)); | |
181 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
182 | } | |
265e0f17 NB |
183 | if (__le32_to_cpu(sb->level) == 10) { |
184 | int lo = __le32_to_cpu(sb->layout); | |
185 | printf(" Layout : near=%d, far=%d\n", | |
186 | lo&255, (lo>>8)&255); | |
187 | } | |
82d9eba6 NB |
188 | switch(__le32_to_cpu(sb->level)) { |
189 | case 0: | |
190 | case 4: | |
191 | case 5: | |
192 | printf(" Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize/2)); | |
193 | break; | |
194 | case -1: | |
195 | printf(" Rounding : %dK\n", __le32_to_cpu(sb->chunksize/2)); | |
196 | break; | |
197 | default: break; | |
198 | } | |
199 | printf("\n"); | |
200 | printf(" Array State : "); | |
201 | for (d=0; d<__le32_to_cpu(sb->raid_disks); d++) { | |
202 | int cnt = 0; | |
203 | int me = 0; | |
204 | int i; | |
205 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { | |
206 | int role = __le16_to_cpu(sb->dev_roles[i]); | |
207 | if (role == d) { | |
208 | if (i == __le32_to_cpu(sb->dev_number)) | |
209 | me = 1; | |
210 | cnt++; | |
211 | } | |
212 | } | |
213 | if (cnt > 1) printf("?"); | |
214 | else if (cnt == 1 && me) printf("U"); | |
215 | else if (cnt == 1) printf("u"); | |
216 | else printf ("_"); | |
217 | } | |
6fbba4c9 | 218 | faulty = 0; |
82d9eba6 NB |
219 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { |
220 | int role = __le16_to_cpu(sb->dev_roles[i]); | |
6fbba4c9 NB |
221 | if (role == 0xFFFE) |
222 | faulty++; | |
82d9eba6 | 223 | } |
82d9eba6 NB |
224 | if (faulty) printf(" %d failed", faulty); |
225 | printf("\n"); | |
226 | } | |
227 | ||
228 | ||
229 | static void brief_examine_super1(void *sbv) | |
230 | { | |
231 | struct mdp_superblock_1 *sb = sbv; | |
232 | int i; | |
233 | ||
234 | char *c=map_num(pers, __le32_to_cpu(sb->level)); | |
235 | ||
236 | printf("ARRAY /dev/?? level=%s metadata=1 num-devices=%d UUID=", | |
237 | c?c:"-unknown-", sb->raid_disks); | |
238 | for (i=0; i<16; i++) { | |
239 | printf("%02x", sb->set_uuid[i]); | |
240 | if ((i&3)==0 && i != 0) printf(":"); | |
241 | } | |
947fd4dd NB |
242 | if (sb->set_name[0]) |
243 | printf(" name=%.32s", sb->set_name); | |
82d9eba6 NB |
244 | printf("\n"); |
245 | } | |
246 | ||
247 | static void detail_super1(void *sbv) | |
248 | { | |
249 | struct mdp_superblock_1 *sb = sbv; | |
250 | int i; | |
251 | ||
947fd4dd | 252 | printf(" Name : %.32s\n", sb->set_name); |
82d9eba6 NB |
253 | printf(" UUID : "); |
254 | for (i=0; i<16; i++) { | |
82d9eba6 | 255 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 256 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 | 257 | } |
570c0542 | 258 | printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb->events)); |
82d9eba6 NB |
259 | } |
260 | ||
261 | static void brief_detail_super1(void *sbv) | |
262 | { | |
263 | struct mdp_superblock_1 *sb = sbv; | |
264 | int i; | |
265 | ||
947fd4dd NB |
266 | if (sb->set_name[0]) |
267 | printf(" name=%.32s", sb->set_name); | |
82d9eba6 NB |
268 | printf(" UUID="); |
269 | for (i=0; i<16; i++) { | |
82d9eba6 | 270 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 271 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 NB |
272 | } |
273 | } | |
274 | ||
c7654afc NB |
275 | #endif |
276 | ||
82d9eba6 NB |
277 | static void uuid_from_super1(int uuid[4], void * sbv) |
278 | { | |
279 | struct mdp_superblock_1 *super = sbv; | |
280 | char *cuuid = (char*)uuid; | |
281 | int i; | |
282 | for (i=0; i<16; i++) | |
283 | cuuid[i] = super->set_uuid[i]; | |
284 | } | |
285 | ||
947fd4dd | 286 | static void getinfo_super1(struct mdinfo *info, mddev_ident_t ident, void *sbv) |
82d9eba6 NB |
287 | { |
288 | struct mdp_superblock_1 *sb = sbv; | |
289 | int working = 0; | |
290 | int i; | |
291 | int role; | |
292 | ||
293 | info->array.major_version = 1; | |
294 | info->array.minor_version = __le32_to_cpu(sb->feature_map); | |
295 | info->array.patch_version = 0; | |
296 | info->array.raid_disks = __le32_to_cpu(sb->raid_disks); | |
297 | info->array.level = __le32_to_cpu(sb->level); | |
265e0f17 | 298 | info->array.layout = __le32_to_cpu(sb->layout); |
82d9eba6 NB |
299 | info->array.md_minor = -1; |
300 | info->array.ctime = __le64_to_cpu(sb->ctime); | |
301 | ||
302 | info->disk.major = 0; | |
303 | info->disk.minor = 0; | |
fbf8a0b7 | 304 | info->disk.number = __le32_to_cpu(sb->dev_number); |
82d9eba6 NB |
305 | if (__le32_to_cpu(sb->dev_number) >= __le32_to_cpu(sb->max_dev) || |
306 | __le32_to_cpu(sb->max_dev) > 512) | |
307 | role = 0xfffe; | |
308 | else | |
309 | role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]); | |
310 | ||
311 | info->disk.raid_disk = -1; | |
312 | switch(role) { | |
313 | case 0xFFFF: | |
314 | info->disk.state = 2; /* spare: ACTIVE, not sync, not faulty */ | |
315 | break; | |
316 | case 0xFFFE: | |
317 | info->disk.state = 1; /* faulty */ | |
318 | break; | |
319 | default: | |
320 | info->disk.state = 6; /* active and in sync */ | |
321 | info->disk.raid_disk = role; | |
322 | } | |
323 | info->events = __le64_to_cpu(sb->events); | |
324 | ||
325 | memcpy(info->uuid, sb->set_uuid, 16); | |
326 | ||
947fd4dd NB |
327 | strncpy(ident->name, sb->set_name, 32); |
328 | ident->name[32] = 0; | |
329 | ||
82d9eba6 NB |
330 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { |
331 | role = __le16_to_cpu(sb->dev_roles[i]); | |
fbf8a0b7 | 332 | if (/*role == 0xFFFF || */role < info->array.raid_disks) |
82d9eba6 NB |
333 | working++; |
334 | } | |
335 | ||
336 | info->array.working_disks = working; | |
337 | } | |
338 | ||
339 | static int update_super1(struct mdinfo *info, void *sbv, char *update, char *devname, int verbose) | |
340 | { | |
341 | int rv = 0; | |
342 | struct mdp_superblock_1 *sb = sbv; | |
343 | ||
344 | if (strcmp(update, "force")==0) { | |
6d3d5804 | 345 | sb->events = __cpu_to_le64(info->events); |
82d9eba6 NB |
346 | switch(__le32_to_cpu(sb->level)) { |
347 | case 5: case 4: case 6: | |
348 | /* need to force clean */ | |
349 | sb->resync_offset = ~0ULL; | |
350 | } | |
351 | } | |
352 | if (strcmp(update, "assemble")==0) { | |
353 | int d = info->disk.number; | |
354 | int want; | |
355 | if (info->disk.state == 6) | |
356 | want = __cpu_to_le32(info->disk.raid_disk); | |
357 | else | |
358 | want = 0xFFFF; | |
359 | if (sb->dev_roles[d] != want) { | |
360 | sb->dev_roles[d] = want; | |
361 | rv = 1; | |
362 | } | |
363 | } | |
364 | #if 0 | |
365 | if (strcmp(update, "newdev") == 0) { | |
366 | int d = info->disk.number; | |
367 | memset(&sb->disks[d], 0, sizeof(sb->disks[d])); | |
368 | sb->disks[d].number = d; | |
369 | sb->disks[d].major = info->disk.major; | |
370 | sb->disks[d].minor = info->disk.minor; | |
371 | sb->disks[d].raid_disk = info->disk.raid_disk; | |
372 | sb->disks[d].state = info->disk.state; | |
373 | sb->this_disk = sb->disks[d]; | |
374 | } | |
375 | #endif | |
376 | if (strcmp(update, "grow") == 0) { | |
377 | sb->raid_disks = __cpu_to_le32(info->array.raid_disks); | |
378 | /* FIXME */ | |
379 | } | |
380 | if (strcmp(update, "resync") == 0) { | |
381 | /* make sure resync happens */ | |
382 | sb->resync_offset = ~0ULL; | |
383 | } | |
7d99579f NB |
384 | if (strcmp(update, "uuid") == 0) |
385 | memcmp(sb->set_uuid, info->uuid, 16); | |
82d9eba6 NB |
386 | |
387 | sb->sb_csum = calc_sb_1_csum(sb); | |
388 | return rv; | |
389 | } | |
390 | ||
391 | ||
392 | static __u64 event_super1(void *sbv) | |
393 | { | |
394 | struct mdp_superblock_1 *sb = sbv; | |
395 | return __le64_to_cpu(sb->events); | |
396 | } | |
397 | ||
947fd4dd | 398 | static int init_super1(struct supertype *st, void **sbp, mdu_array_info_t *info, char *name) |
82d9eba6 | 399 | { |
34163fc7 | 400 | struct mdp_superblock_1 *sb = malloc(1024 + sizeof(bitmap_super_t)); |
82d9eba6 | 401 | int spares; |
34163fc7 | 402 | int rfd; |
82d9eba6 NB |
403 | memset(sb, 0, 1024); |
404 | ||
405 | if (info->major_version == -1) | |
406 | /* zeroing superblock */ | |
407 | return 0; | |
408 | ||
409 | spares = info->working_disks - info->active_disks; | |
410 | if (info->raid_disks + spares > 384) { | |
411 | fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n", | |
412 | info->raid_disks , spares, 384); | |
413 | return 0; | |
414 | } | |
415 | ||
416 | ||
417 | sb->magic = __cpu_to_le32(MD_SB_MAGIC); | |
418 | sb->major_version = __cpu_to_le32(1); | |
419 | sb->feature_map = 0; | |
420 | sb->pad0 = 0; | |
34163fc7 NB |
421 | |
422 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
423 | read(rfd, sb->set_uuid, 16) != 16) { | |
424 | *(__u32*)(sb->set_uuid) = random(); | |
425 | *(__u32*)(sb->set_uuid+4) = random(); | |
426 | *(__u32*)(sb->set_uuid+8) = random(); | |
427 | *(__u32*)(sb->set_uuid+12) = random(); | |
428 | } | |
429 | if (rfd >= 0) close(rfd); | |
82d9eba6 | 430 | |
947fd4dd NB |
431 | memset(sb->set_name, 0, 32); |
432 | strcpy(sb->set_name, name); | |
82d9eba6 NB |
433 | |
434 | sb->ctime = __cpu_to_le64((unsigned long long)time(0)); | |
435 | sb->level = __cpu_to_le32(info->level); | |
ea329559 | 436 | sb->layout = __cpu_to_le32(info->layout); |
82d9eba6 NB |
437 | sb->size = __cpu_to_le64(info->size*2ULL); |
438 | sb->chunksize = __cpu_to_le32(info->chunk_size>>9); | |
439 | sb->raid_disks = __cpu_to_le32(info->raid_disks); | |
440 | ||
441 | sb->data_offset = __cpu_to_le64(0); | |
442 | sb->data_size = __cpu_to_le64(0); | |
443 | sb->super_offset = __cpu_to_le64(0); | |
444 | sb->recovery_offset = __cpu_to_le64(0); | |
445 | ||
446 | sb->utime = sb->ctime; | |
447 | sb->events = __cpu_to_le64(1); | |
34321279 | 448 | if (info->state & (1<<MD_SB_CLEAN)) |
82d9eba6 NB |
449 | sb->resync_offset = ~0ULL; |
450 | else | |
451 | sb->resync_offset = 0; | |
34163fc7 | 452 | sb->max_dev = __cpu_to_le32((1024- sizeof(struct mdp_superblock_1))/ |
82d9eba6 NB |
453 | sizeof(sb->dev_roles[0])); |
454 | memset(sb->pad3, 0, sizeof(sb->pad3)); | |
455 | ||
456 | memset(sb->dev_roles, 0xff, 1024 - sizeof(struct mdp_superblock_1)); | |
457 | ||
458 | *sbp = sb; | |
459 | return 1; | |
460 | } | |
461 | ||
462 | /* Add a device to the superblock being created */ | |
463 | static void add_to_super1(void *sbv, mdu_disk_info_t *dk) | |
464 | { | |
465 | struct mdp_superblock_1 *sb = sbv; | |
466 | __u16 *rp = sb->dev_roles + dk->number; | |
dfd4d8ee | 467 | if ((dk->state & 6) == 6) /* active, sync */ |
82d9eba6 | 468 | *rp = __cpu_to_le16(dk->raid_disk); |
412ca2e5 | 469 | else if ((dk->state & ~2) == 0) /* active or idle -> spare */ |
82d9eba6 | 470 | *rp = 0xffff; |
34163fc7 | 471 | else |
82d9eba6 NB |
472 | *rp = 0xfffe; |
473 | } | |
474 | ||
96395475 | 475 | static int store_super1(struct supertype *st, int fd, void *sbv) |
82d9eba6 NB |
476 | { |
477 | struct mdp_superblock_1 *sb = sbv; | |
6fbba4c9 | 478 | unsigned long long sb_offset; |
82d9eba6 | 479 | int sbsize; |
96395475 NB |
480 | long size; |
481 | ||
482 | if (ioctl(fd, BLKGETSIZE, &size)) | |
483 | return 1; | |
484 | ||
485 | ||
486 | if (size < 24) | |
487 | return 2; | |
488 | ||
489 | /* | |
490 | * Calculate the position of the superblock. | |
491 | * It is always aligned to a 4K boundary and | |
492 | * depending on minor_version, it can be: | |
493 | * 0: At least 8K, but less than 12K, from end of device | |
494 | * 1: At start of device | |
495 | * 2: 4K from start of device. | |
496 | */ | |
497 | switch(st->minor_version) { | |
498 | case 0: | |
499 | sb_offset = size; | |
500 | sb_offset -= 8*2; | |
501 | sb_offset &= ~(4*2-1); | |
502 | break; | |
503 | case 1: | |
6fbba4c9 | 504 | sb_offset = 0; |
96395475 NB |
505 | break; |
506 | case 2: | |
507 | sb_offset = 4*2; | |
508 | break; | |
509 | default: | |
510 | return -EINVAL; | |
511 | } | |
512 | ||
82d9eba6 | 513 | |
34163fc7 | 514 | |
6fbba4c9 NB |
515 | if (sb_offset != __le64_to_cpu(sb->super_offset) && |
516 | 0 != __le64_to_cpu(sb->super_offset) | |
96395475 NB |
517 | ) { |
518 | fprintf(stderr, Name ": internal error - sb_offset is wrong\n"); | |
519 | abort(); | |
520 | } | |
82d9eba6 | 521 | |
6fbba4c9 | 522 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) |
82d9eba6 NB |
523 | return 3; |
524 | ||
525 | sbsize = sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev); | |
526 | ||
527 | if (write(fd, sb, sbsize) != sbsize) | |
528 | return 4; | |
529 | ||
570c0542 | 530 | fsync(fd); |
82d9eba6 NB |
531 | return 0; |
532 | } | |
533 | ||
892debc8 NB |
534 | static int load_super1(struct supertype *st, int fd, void **sbp, char *devname); |
535 | ||
34163fc7 | 536 | static int write_init_super1(struct supertype *st, void *sbv, |
1bf4e2d9 | 537 | mdu_disk_info_t *dinfo, char *devname) |
82d9eba6 NB |
538 | { |
539 | struct mdp_superblock_1 *sb = sbv; | |
e478dc86 | 540 | void *refsbv = NULL; |
892debc8 | 541 | int fd = open(devname, O_RDWR | O_EXCL); |
dfe47e00 | 542 | int rfd; |
82d9eba6 NB |
543 | int rv; |
544 | ||
545 | long size; | |
546 | long long sb_offset; | |
547 | ||
548 | ||
549 | if (fd < 0) { | |
550 | fprintf(stderr, Name ": Failed to open %s to write superblock\n", | |
551 | devname); | |
552 | return -1; | |
553 | } | |
554 | ||
555 | sb->dev_number = __cpu_to_le32(dinfo->number); | |
34163fc7 | 556 | if (dinfo->state & (1<<MD_DISK_WRITEMOSTLY)) |
dfd4d8ee | 557 | sb->devflags |= WriteMostly1; |
892debc8 | 558 | |
dfe47e00 NB |
559 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || |
560 | read(rfd, sb->device_uuid, 16) != 16) { | |
561 | *(__u32*)(sb->device_uuid) = random(); | |
562 | *(__u32*)(sb->device_uuid+4) = random(); | |
563 | *(__u32*)(sb->device_uuid+8) = random(); | |
564 | *(__u32*)(sb->device_uuid+12) = random(); | |
565 | } | |
566 | if (rfd >= 0) close(rfd); | |
892debc8 NB |
567 | sb->events = 0; |
568 | ||
e478dc86 NB |
569 | if (load_super1(st, fd, &refsbv, NULL)==0) { |
570 | struct mdp_superblock_1 *refsb = refsbv; | |
571 | ||
892debc8 NB |
572 | memcpy(sb->device_uuid, refsb->device_uuid, 16); |
573 | if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) { | |
574 | /* same array, so preserve events and dev_number */ | |
575 | sb->events = refsb->events; | |
576 | sb->dev_number = refsb->dev_number; | |
577 | } | |
578 | free(refsb); | |
579 | } | |
34163fc7 | 580 | |
024ce7fe NB |
581 | if (ioctl(fd, BLKGETSIZE, &size)) { |
582 | close(fd); | |
82d9eba6 | 583 | return 1; |
024ce7fe | 584 | } |
82d9eba6 | 585 | |
024ce7fe NB |
586 | if (size < 24) { |
587 | close(fd); | |
82d9eba6 | 588 | return 2; |
024ce7fe | 589 | } |
82d9eba6 NB |
590 | |
591 | ||
592 | /* | |
593 | * Calculate the position of the superblock. | |
594 | * It is always aligned to a 4K boundary and | |
595 | * depending on minor_version, it can be: | |
596 | * 0: At least 8K, but less than 12K, from end of device | |
597 | * 1: At start of device | |
598 | * 2: 4K from start of device. | |
599 | */ | |
600 | switch(st->minor_version) { | |
601 | case 0: | |
602 | sb_offset = size; | |
603 | sb_offset -= 8*2; | |
604 | sb_offset &= ~(4*2-1); | |
605 | sb->super_offset = __cpu_to_le64(sb_offset); | |
606 | sb->data_offset = __cpu_to_le64(0); | |
1bf4e2d9 | 607 | sb->data_size = __cpu_to_le64(sb_offset); |
82d9eba6 NB |
608 | break; |
609 | case 1: | |
610 | sb->super_offset = __cpu_to_le64(0); | |
1bf4e2d9 NB |
611 | sb->data_offset = __cpu_to_le64(4*2); /* leave 4k for super and bitmap */ |
612 | sb->data_size = __cpu_to_le64(size - 4*2); | |
82d9eba6 NB |
613 | break; |
614 | case 2: | |
615 | sb_offset = 4*2; | |
616 | sb->super_offset = __cpu_to_le64(sb_offset); | |
1bf4e2d9 NB |
617 | sb->data_offset = __cpu_to_le64(sb_offset+4*2); |
618 | sb->data_size = __cpu_to_le64(size - 4*2 - 4*2); | |
82d9eba6 NB |
619 | break; |
620 | default: | |
621 | return -EINVAL; | |
622 | } | |
623 | ||
624 | ||
625 | sb->sb_csum = calc_sb_1_csum(sb); | |
96395475 | 626 | rv = store_super1(st, fd, sb); |
82d9eba6 NB |
627 | if (rv) |
628 | fprintf(stderr, Name ": failed to write superblock to %s\n", devname); | |
34163fc7 | 629 | |
1bf4e2d9 NB |
630 | if (rv == 0 && (__le32_to_cpu(sb->feature_map) & 1)) |
631 | rv = st->ss->write_bitmap(st, fd, sbv); | |
024ce7fe | 632 | close(fd); |
82d9eba6 NB |
633 | return rv; |
634 | } | |
635 | ||
636 | static int compare_super1(void **firstp, void *secondv) | |
637 | { | |
638 | /* | |
639 | * return: | |
640 | * 0 same, or first was empty, and second was copied | |
641 | * 1 second had wrong number | |
642 | * 2 wrong uuid | |
643 | * 3 wrong other info | |
644 | */ | |
645 | struct mdp_superblock_1 *first = *firstp; | |
646 | struct mdp_superblock_1 *second = secondv; | |
647 | ||
648 | if (second->magic != __cpu_to_le32(MD_SB_MAGIC)) | |
649 | return 1; | |
650 | if (second->major_version != __cpu_to_le32(1)) | |
651 | return 1; | |
652 | ||
653 | if (!first) { | |
654 | first = malloc(1024); | |
655 | memcpy(first, second, 1024); | |
656 | *firstp = first; | |
657 | return 0; | |
658 | } | |
659 | if (memcmp(first->set_uuid, second->set_uuid, 16)!= 0) | |
660 | return 2; | |
661 | ||
662 | if (first->ctime != second->ctime || | |
663 | first->level != second->level || | |
664 | first->layout != second->layout || | |
665 | first->size != second->size || | |
666 | first->chunksize != second->chunksize || | |
667 | first->raid_disks != second->raid_disks) | |
668 | return 3; | |
669 | return 0; | |
670 | } | |
671 | ||
672 | static int load_super1(struct supertype *st, int fd, void **sbp, char *devname) | |
673 | { | |
674 | unsigned long size; | |
675 | unsigned long long sb_offset; | |
676 | struct mdp_superblock_1 *super; | |
677 | ||
678 | ||
679 | ||
680 | if (st->ss == NULL) { | |
570c0542 NB |
681 | int bestvers = -1; |
682 | __u64 bestctime = 0; | |
683 | /* guess... choose latest ctime */ | |
82d9eba6 NB |
684 | st->ss = &super1; |
685 | for (st->minor_version = 0; st->minor_version <= 2 ; st->minor_version++) { | |
686 | switch(load_super1(st, fd, sbp, devname)) { | |
570c0542 NB |
687 | case 0: super = *sbp; |
688 | if (bestvers == -1 || | |
689 | bestctime < __le64_to_cpu(super->ctime)) { | |
690 | bestvers = st->minor_version; | |
691 | bestctime = __le64_to_cpu(super->ctime); | |
692 | } | |
693 | free(super); | |
694 | *sbp = NULL; | |
695 | break; | |
82d9eba6 NB |
696 | case 1: st->ss = NULL; return 1; /*bad device */ |
697 | case 2: break; /* bad, try next */ | |
698 | } | |
699 | } | |
570c0542 NB |
700 | if (bestvers != -1) { |
701 | int rv; | |
702 | st->minor_version = bestvers; | |
703 | st->ss = &super1; | |
ea329559 | 704 | st->max_devs = 384; |
570c0542 NB |
705 | rv = load_super1(st, fd, sbp, devname); |
706 | if (rv) st->ss = NULL; | |
707 | return rv; | |
708 | } | |
82d9eba6 NB |
709 | st->ss = NULL; |
710 | return 2; | |
711 | } | |
712 | if (ioctl(fd, BLKGETSIZE, &size)) { | |
34163fc7 | 713 | if (devname) |
82d9eba6 NB |
714 | fprintf(stderr, Name ": cannot find device size for %s: %s\n", |
715 | devname, strerror(errno)); | |
716 | return 1; | |
717 | } | |
718 | ||
719 | if (size < 24) { | |
720 | if (devname) | |
721 | fprintf(stderr, Name ": %s is too small for md: size is %lu sectors.\n", | |
722 | devname, size); | |
723 | return 1; | |
724 | } | |
725 | ||
726 | /* | |
727 | * Calculate the position of the superblock. | |
728 | * It is always aligned to a 4K boundary and | |
729 | * depeding on minor_version, it can be: | |
730 | * 0: At least 8K, but less than 12K, from end of device | |
731 | * 1: At start of device | |
732 | * 2: 4K from start of device. | |
733 | */ | |
734 | switch(st->minor_version) { | |
735 | case 0: | |
736 | sb_offset = size; | |
737 | sb_offset -= 8*2; | |
738 | sb_offset &= ~(4*2-1); | |
739 | break; | |
740 | case 1: | |
741 | sb_offset = 0; | |
742 | break; | |
743 | case 2: | |
744 | sb_offset = 4*2; | |
745 | break; | |
746 | default: | |
747 | return -EINVAL; | |
748 | } | |
749 | ||
750 | ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */ | |
751 | ||
752 | ||
753 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) { | |
754 | if (devname) | |
755 | fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n", | |
756 | devname, strerror(errno)); | |
757 | return 1; | |
758 | } | |
759 | ||
34163fc7 | 760 | super = malloc(1024 + sizeof(bitmap_super_t)); |
82d9eba6 NB |
761 | |
762 | if (read(fd, super, 1024) != 1024) { | |
763 | if (devname) | |
764 | fprintf(stderr, Name ": Cannot read superblock on %s\n", | |
765 | devname); | |
766 | free(super); | |
767 | return 1; | |
768 | } | |
769 | ||
770 | if (__le32_to_cpu(super->magic) != MD_SB_MAGIC) { | |
771 | if (devname) | |
772 | fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n", | |
773 | devname, MD_SB_MAGIC, __le32_to_cpu(super->magic)); | |
774 | free(super); | |
775 | return 2; | |
776 | } | |
777 | ||
778 | if (__le32_to_cpu(super->major_version) != 1) { | |
779 | if (devname) | |
780 | fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n", | |
781 | devname, __le32_to_cpu(super->major_version)); | |
782 | free(super); | |
783 | return 2; | |
784 | } | |
785 | if (__le64_to_cpu(super->super_offset) != sb_offset) { | |
786 | if (devname) | |
787 | fprintf(stderr, Name ": No superblock found on %s (super_offset is wrong)\n", | |
788 | devname); | |
789 | free(super); | |
790 | return 2; | |
791 | } | |
792 | *sbp = super; | |
793 | return 0; | |
794 | } | |
795 | ||
796 | ||
797 | static struct supertype *match_metadata_desc1(char *arg) | |
798 | { | |
799 | struct supertype *st = malloc(sizeof(*st)); | |
800 | if (!st) return st; | |
801 | ||
802 | st->ss = &super1; | |
ea329559 | 803 | st->max_devs = 384; |
82d9eba6 NB |
804 | if (strcmp(arg, "1") == 0 || |
805 | strcmp(arg, "1.0") == 0) { | |
806 | st->minor_version = 0; | |
807 | return st; | |
808 | } | |
809 | if (strcmp(arg, "1.1") == 0) { | |
810 | st->minor_version = 1; | |
811 | return st; | |
812 | } | |
813 | if (strcmp(arg, "1.2") == 0) { | |
814 | st->minor_version = 2; | |
815 | return st; | |
816 | } | |
817 | ||
818 | free(st); | |
819 | return NULL; | |
820 | } | |
821 | ||
34163fc7 NB |
822 | /* find available size on device with this devsize, using |
823 | * superblock type st, and reserving 'reserve' sectors for | |
824 | * a possible bitmap | |
825 | */ | |
1bf4e2d9 | 826 | static __u64 avail_size1(struct supertype *st, __u64 devsize) |
82d9eba6 NB |
827 | { |
828 | if (devsize < 24) | |
829 | return 0; | |
830 | ||
34163fc7 NB |
831 | switch(st->minor_version) { |
832 | case 0: | |
1bf4e2d9 NB |
833 | /* at end */ |
834 | return ((devsize - 8*2 ) & ~(4*2-1)); | |
34163fc7 | 835 | case 1: |
1bf4e2d9 NB |
836 | /* at start, 4K for superblock and possible bitmap */ |
837 | return devsize - 4*2; | |
34163fc7 | 838 | case 2: |
1bf4e2d9 NB |
839 | /* 4k from start, 4K for superblock and possible bitmap */ |
840 | return devsize - (4+4)*2; | |
34163fc7 NB |
841 | } |
842 | return 0; | |
843 | } | |
844 | ||
1bf4e2d9 NB |
845 | static int |
846 | add_internal_bitmap1(struct supertype *st, void *sbv, | |
f9c25f1d | 847 | int chunk, int delay, int write_behind, unsigned long long size, int may_change, int major) |
34163fc7 NB |
848 | { |
849 | /* | |
1bf4e2d9 NB |
850 | * If not may_change, then this is a 'Grow', and the bitmap |
851 | * must fit after the superblock. | |
852 | * If may_change, then this is create, and we can put the bitmap | |
853 | * before the superblock if we like, or may move the start. | |
854 | * For now, just squeeze the bitmap into 3k and don't change anything. | |
34163fc7 | 855 | * |
f9c25f1d | 856 | * size is in sectors, chunk is in bytes !!! |
34163fc7 NB |
857 | */ |
858 | ||
1bf4e2d9 NB |
859 | unsigned long long bits; |
860 | unsigned long long max_bits = (3*512 - sizeof(bitmap_super_t)) * 8; | |
34163fc7 NB |
861 | unsigned long long min_chunk; |
862 | struct mdp_superblock_1 *sb = sbv; | |
863 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + 1024); | |
864 | ||
1bf4e2d9 NB |
865 | if (st->minor_version && !may_change && |
866 | __le64_to_cpu(sb->data_offset) - __le64_to_cpu(sb->super_offset) < 8) | |
867 | return 0; /* doesn't fit */ | |
868 | ||
869 | ||
34163fc7 NB |
870 | |
871 | min_chunk = 4096; /* sub-page chunks don't work yet.. */ | |
f9c25f1d | 872 | bits = (size*512)/min_chunk +1; |
34163fc7 NB |
873 | while (bits > max_bits) { |
874 | min_chunk *= 2; | |
875 | bits = (bits+1)/2; | |
876 | } | |
877 | if (chunk == UnSet) | |
878 | chunk = min_chunk; | |
879 | else if (chunk < min_chunk) | |
880 | return 0; /* chunk size too small */ | |
881 | ||
1bf4e2d9 | 882 | sb->bitmap_offset = __cpu_to_le32(2); |
34163fc7 NB |
883 | |
884 | sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | 1); | |
885 | memset(bms, sizeof(*bms), 0); | |
886 | bms->magic = __cpu_to_le32(BITMAP_MAGIC); | |
dcec9ee5 | 887 | bms->version = __cpu_to_le32(major); |
34163fc7 NB |
888 | uuid_from_super1((int*)bms->uuid, sb); |
889 | bms->chunksize = __cpu_to_le32(chunk); | |
890 | bms->daemon_sleep = __cpu_to_le32(delay); | |
f9c25f1d | 891 | bms->sync_size = __cpu_to_le64(size); |
34163fc7 NB |
892 | bms->write_behind = __cpu_to_le32(write_behind); |
893 | ||
34163fc7 NB |
894 | return 1; |
895 | } | |
896 | ||
897 | ||
f6d75de8 | 898 | void locate_bitmap1(struct supertype *st, int fd, void *sbv) |
34163fc7 | 899 | { |
34163fc7 | 900 | unsigned long long offset; |
1bf4e2d9 | 901 | struct mdp_superblock_1 *sb; |
34163fc7 | 902 | |
e478dc86 NB |
903 | if (!sbv) |
904 | if (st->ss->load_super(st, fd, sbv, NULL)) | |
f6d75de8 | 905 | return; /* no error I hope... */ |
e478dc86 NB |
906 | |
907 | sb = sbv; | |
908 | ||
1bf4e2d9 NB |
909 | offset = __le64_to_cpu(sb->super_offset); |
910 | offset += (long) __le32_to_cpu(sb->bitmap_offset); | |
f6d75de8 NB |
911 | if (!sbv) |
912 | free(sb); | |
1bf4e2d9 | 913 | lseek64(fd, offset<<9, 0); |
34163fc7 NB |
914 | } |
915 | ||
916 | int write_bitmap1(struct supertype *st, int fd, void *sbv) | |
917 | { | |
918 | struct mdp_superblock_1 *sb = sbv; | |
1bf4e2d9 | 919 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+1024); |
34163fc7 NB |
920 | int rv = 0; |
921 | ||
922 | int towrite, n; | |
923 | char buf[4096]; | |
924 | ||
f6d75de8 | 925 | locate_bitmap1(st, fd, sbv); |
34163fc7 NB |
926 | |
927 | write(fd, ((char*)sb)+1024, sizeof(bitmap_super_t)); | |
1bf4e2d9 NB |
928 | towrite = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9); |
929 | towrite = (towrite+7) >> 3; /* bits to bytes */ | |
34163fc7 NB |
930 | memset(buf, 0xff, sizeof(buf)); |
931 | while (towrite > 0) { | |
932 | n = towrite; | |
933 | if (n > sizeof(buf)) | |
934 | n = sizeof(buf); | |
935 | n = write(fd, buf, n); | |
936 | if (n > 0) | |
937 | towrite -= n; | |
938 | else | |
939 | break; | |
940 | } | |
941 | fsync(fd); | |
942 | if (towrite) | |
943 | rv = -2; | |
944 | ||
945 | return rv; | |
82d9eba6 NB |
946 | } |
947 | ||
948 | struct superswitch super1 = { | |
c7654afc | 949 | #ifndef MDASSEMBLE |
82d9eba6 NB |
950 | .examine_super = examine_super1, |
951 | .brief_examine_super = brief_examine_super1, | |
952 | .detail_super = detail_super1, | |
953 | .brief_detail_super = brief_detail_super1, | |
c7654afc | 954 | #endif |
82d9eba6 NB |
955 | .uuid_from_super = uuid_from_super1, |
956 | .getinfo_super = getinfo_super1, | |
957 | .update_super = update_super1, | |
958 | .event_super = event_super1, | |
959 | .init_super = init_super1, | |
960 | .add_to_super = add_to_super1, | |
961 | .store_super = store_super1, | |
962 | .write_init_super = write_init_super1, | |
963 | .compare_super = compare_super1, | |
964 | .load_super = load_super1, | |
965 | .match_metadata_desc = match_metadata_desc1, | |
966 | .avail_size = avail_size1, | |
34163fc7 NB |
967 | .add_internal_bitmap = add_internal_bitmap1, |
968 | .locate_bitmap = locate_bitmap1, | |
969 | .write_bitmap = write_bitmap1, | |
82d9eba6 | 970 | .major = 1, |
f277ce36 NB |
971 | #if __BYTE_ORDER == BIG_ENDIAN |
972 | .swapuuid = 0, | |
973 | #else | |
974 | .swapuuid = 1, | |
975 | #endif | |
82d9eba6 | 976 | }; |