]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
4 | * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de> | |
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" | |
31 | /* | |
32 | * The version-1 superblock : | |
33 | * All numeric fields are little-endian. | |
34 | * | |
35 | * total size: 256 bytes plus 2 per device. | |
36 | * 1K allows 384 devices. | |
37 | */ | |
38 | struct mdp_superblock_1 { | |
39 | /* constant array information - 128 bytes */ | |
40 | __u32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */ | |
41 | __u32 major_version; /* 1 */ | |
42 | __u32 feature_map; /* 0 for now */ | |
43 | __u32 pad0; /* always set to 0 when writing */ | |
44 | ||
45 | __u8 set_uuid[16]; /* user-space generated. */ | |
46 | char set_name[32]; /* set and interpreted by user-space */ | |
47 | ||
48 | __u64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/ | |
49 | __u32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */ | |
50 | __u32 layout; /* only for raid5 currently */ | |
51 | __u64 size; /* used size of component devices, in 512byte sectors */ | |
52 | ||
53 | __u32 chunksize; /* in 512byte sectors */ | |
54 | __u32 raid_disks; | |
55 | __u32 bitmap_offset; /* sectors after start of superblock that bitmap starts | |
56 | * NOTE: signed, so bitmap can be before superblock | |
57 | * only meaningful of feature_map[0] is set. | |
58 | */ | |
59 | ||
60 | /* These are only valid with feature bit '4' */ | |
61 | __u32 new_level; /* new level we are reshaping to */ | |
62 | __u64 reshape_position; /* next address in array-space for reshape */ | |
63 | __u32 delta_disks; /* change in number of raid_disks */ | |
64 | __u32 new_layout; /* new layout */ | |
65 | __u32 new_chunk; /* new chunk size (bytes) */ | |
66 | __u8 pad1[128-124]; /* set to 0 when written */ | |
67 | ||
68 | /* constant this-device information - 64 bytes */ | |
69 | __u64 data_offset; /* sector start of data, often 0 */ | |
70 | __u64 data_size; /* sectors in this device that can be used for data */ | |
71 | __u64 super_offset; /* sector start of this superblock */ | |
72 | __u64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */ | |
73 | __u32 dev_number; /* permanent identifier of this device - not role in raid */ | |
74 | __u32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */ | |
75 | __u8 device_uuid[16]; /* user-space setable, ignored by kernel */ | |
76 | __u8 devflags; /* per-device flags. Only one defined...*/ | |
77 | #define WriteMostly1 1 /* mask for writemostly flag in above */ | |
78 | __u8 pad2[64-57]; /* set to 0 when writing */ | |
79 | ||
80 | /* array state information - 64 bytes */ | |
81 | __u64 utime; /* 40 bits second, 24 btes microseconds */ | |
82 | __u64 events; /* incremented when superblock updated */ | |
83 | __u64 resync_offset; /* data before this offset (from data_offset) known to be in sync */ | |
84 | __u32 sb_csum; /* checksum upto devs[max_dev] */ | |
85 | __u32 max_dev; /* size of devs[] array to consider */ | |
86 | __u8 pad3[64-32]; /* set to 0 when writing */ | |
87 | ||
88 | /* device state information. Indexed by dev_number. | |
89 | * 2 bytes per device | |
90 | * Note there are no per-device state flags. State information is rolled | |
91 | * into the 'roles' value. If a device is spare or faulty, then it doesn't | |
92 | * have a meaningful role. | |
93 | */ | |
94 | __u16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */ | |
95 | }; | |
96 | ||
97 | struct misc_dev_info { | |
98 | __u64 device_size; | |
99 | }; | |
100 | ||
101 | /* feature_map bits */ | |
102 | #define MD_FEATURE_BITMAP_OFFSET 1 | |
103 | #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and | |
104 | * must be honoured | |
105 | */ | |
106 | #define MD_FEATURE_RESHAPE_ACTIVE 4 | |
107 | ||
108 | #define MD_FEATURE_ALL (1|2|4) | |
109 | ||
110 | #ifndef offsetof | |
111 | #define offsetof(t,f) ((size_t)&(((t*)0)->f)) | |
112 | #endif | |
113 | static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb) | |
114 | { | |
115 | unsigned int disk_csum, csum; | |
116 | unsigned long long newcsum; | |
117 | int size = sizeof(*sb) + __le32_to_cpu(sb->max_dev)*2; | |
118 | unsigned int *isuper = (unsigned int*)sb; | |
119 | int i; | |
120 | ||
121 | /* make sure I can count... */ | |
122 | if (offsetof(struct mdp_superblock_1,data_offset) != 128 || | |
123 | offsetof(struct mdp_superblock_1, utime) != 192 || | |
124 | sizeof(struct mdp_superblock_1) != 256) { | |
125 | fprintf(stderr, "WARNING - superblock isn't sized correctly\n"); | |
126 | } | |
127 | ||
128 | disk_csum = sb->sb_csum; | |
129 | sb->sb_csum = 0; | |
130 | newcsum = 0; | |
131 | for (i=0; size>=4; size -= 4 ) { | |
132 | newcsum += __le32_to_cpu(*isuper); | |
133 | isuper++; | |
134 | } | |
135 | ||
136 | if (size == 2) | |
137 | newcsum += __le16_to_cpu(*(unsigned short*) isuper); | |
138 | ||
139 | csum = (newcsum & 0xffffffff) + (newcsum >> 32); | |
140 | sb->sb_csum = disk_csum; | |
141 | return __cpu_to_le32(csum); | |
142 | } | |
143 | ||
144 | #ifndef MDASSEMBLE | |
145 | static void examine_super1(void *sbv, char *homehost) | |
146 | { | |
147 | struct mdp_superblock_1 *sb = sbv; | |
148 | time_t atime; | |
149 | int d; | |
150 | int faulty; | |
151 | int i; | |
152 | char *c; | |
153 | int l = homehost ? strlen(homehost) : 0; | |
154 | int layout; | |
155 | ||
156 | printf(" Magic : %08x\n", __le32_to_cpu(sb->magic)); | |
157 | printf(" Version : %02d\n", 1); | |
158 | printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb->feature_map)); | |
159 | printf(" Array UUID : "); | |
160 | for (i=0; i<16; i++) { | |
161 | if ((i&3)==0 && i != 0) printf(":"); | |
162 | printf("%02x", sb->set_uuid[i]); | |
163 | } | |
164 | printf("\n"); | |
165 | printf(" Name : %.32s", sb->set_name); | |
166 | if (l > 0 && l < 32 && | |
167 | sb->set_name[l] == ':' && | |
168 | strncmp(sb->set_name, homehost, l) == 0) | |
169 | printf(" (local to host %s)", homehost); | |
170 | printf("\n"); | |
171 | atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL; | |
172 | printf(" Creation Time : %.24s\n", ctime(&atime)); | |
173 | c=map_num(pers, __le32_to_cpu(sb->level)); | |
174 | printf(" Raid Level : %s\n", c?c:"-unknown-"); | |
175 | printf(" Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks)); | |
176 | printf("\n"); | |
177 | printf(" Used Dev Size : %llu%s\n", | |
178 | (unsigned long long)sb->data_size, | |
179 | human_size(sb->data_size<<9)); | |
180 | if (__le32_to_cpu(sb->level) >= 0) { | |
181 | int ddsks=0; | |
182 | switch(__le32_to_cpu(sb->level)) { | |
183 | case 1: ddsks=1;break; | |
184 | case 4: | |
185 | case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break; | |
186 | case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break; | |
187 | case 10: | |
188 | layout = __le32_to_cpu(sb->layout); | |
189 | ddsks = __le32_to_cpu(sb->raid_disks) | |
190 | / (layout&255) / ((layout>>8)&255); | |
191 | } | |
192 | if (ddsks) | |
193 | printf(" Array Size : %llu%s\n", | |
194 | ddsks*(unsigned long long)__le64_to_cpu(sb->size), | |
195 | human_size(ddsks*__le64_to_cpu(sb->size)<<9)); | |
196 | if (sb->size != sb->data_size) | |
197 | printf(" Used Size : %llu%s\n", | |
198 | (unsigned long long)__le64_to_cpu(sb->size), | |
199 | human_size(__le64_to_cpu(sb->size)<<9)); | |
200 | } | |
201 | if (sb->data_offset) | |
202 | printf(" Data Offset : %llu sectors\n", | |
203 | (unsigned long long)__le64_to_cpu(sb->data_offset)); | |
204 | printf(" Super Offset : %llu sectors\n", | |
205 | (unsigned long long)__le64_to_cpu(sb->super_offset)); | |
206 | if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET) | |
207 | printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->recovery_offset)); | |
208 | printf(" State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean"); | |
209 | printf(" Device UUID : "); | |
210 | for (i=0; i<16; i++) { | |
211 | if ((i&3)==0 && i != 0) printf(":"); | |
212 | printf("%02x", sb->device_uuid[i]); | |
213 | } | |
214 | printf("\n"); | |
215 | printf("\n"); | |
216 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
217 | printf("Internal Bitmap : %ld sectors from superblock\n", | |
218 | (long)__le32_to_cpu(sb->bitmap_offset)); | |
219 | } | |
220 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) { | |
221 | printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb->reshape_position)/2, | |
222 | human_size(__le64_to_cpu(sb->reshape_position)<<9)); | |
223 | if (__le32_to_cpu(sb->delta_disks)) { | |
224 | printf(" Delta Devices : %d", __le32_to_cpu(sb->delta_disks)); | |
225 | if (__le32_to_cpu(sb->delta_disks)) | |
226 | printf(" (%d->%d)\n", | |
227 | __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks), | |
228 | __le32_to_cpu(sb->raid_disks)); | |
229 | else | |
230 | printf(" (%d->%d)\n", __le32_to_cpu(sb->raid_disks), | |
231 | __le32_to_cpu(sb->raid_disks)+__le32_to_cpu(sb->delta_disks)); | |
232 | } | |
233 | if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) { | |
234 | c = map_num(pers, __le32_to_cpu(sb->new_level)); | |
235 | printf(" New Level : %s\n", c?c:"-unknown-"); | |
236 | } | |
237 | if (__le32_to_cpu(sb->new_layout) != __le32_to_cpu(sb->layout)) { | |
238 | if (__le32_to_cpu(sb->level) == 5) { | |
239 | c = map_num(r5layout, __le32_to_cpu(sb->new_layout)); | |
240 | printf(" New Layout : %s\n", c?c:"-unknown-"); | |
241 | } | |
242 | if (__le32_to_cpu(sb->level) == 10) { | |
243 | printf(" New Layout : near=%d, %s=%d\n", | |
244 | __le32_to_cpu(sb->new_layout)&255, | |
245 | (__le32_to_cpu(sb->new_layout)&0x10000)?"offset":"far", | |
246 | (__le32_to_cpu(sb->new_layout)>>8)&255); | |
247 | } | |
248 | } | |
249 | if (__le32_to_cpu(sb->new_chunk) != __le32_to_cpu(sb->chunksize)) | |
250 | printf(" New Chunksize : %dK\n", __le32_to_cpu(sb->new_chunk)/2); | |
251 | printf("\n"); | |
252 | } | |
253 | if (sb->devflags) { | |
254 | printf(" Flags :"); | |
255 | if (sb->devflags & WriteMostly1) | |
256 | printf(" write-mostly"); | |
257 | printf("\n"); | |
258 | } | |
259 | ||
260 | atime = __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL; | |
261 | printf(" Update Time : %.24s\n", ctime(&atime)); | |
262 | ||
263 | if (calc_sb_1_csum(sb) == sb->sb_csum) | |
264 | printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum)); | |
265 | else | |
266 | printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb->sb_csum), | |
267 | __le32_to_cpu(calc_sb_1_csum(sb))); | |
268 | printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb->events)); | |
269 | printf("\n"); | |
270 | if (__le32_to_cpu(sb->level) == 5) { | |
271 | c = map_num(r5layout, __le32_to_cpu(sb->layout)); | |
272 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
273 | } | |
274 | if (__le32_to_cpu(sb->level) == 10) { | |
275 | int lo = __le32_to_cpu(sb->layout); | |
276 | printf(" Layout : near=%d, %s=%d\n", | |
277 | lo&255, | |
278 | (lo&0x10000)?"offset":"far", | |
279 | (lo>>8)&255); | |
280 | } | |
281 | switch(__le32_to_cpu(sb->level)) { | |
282 | case 0: | |
283 | case 4: | |
284 | case 5: | |
285 | case 6: | |
286 | case 10: | |
287 | printf(" Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize)/2); | |
288 | break; | |
289 | case -1: | |
290 | printf(" Rounding : %dK\n", __le32_to_cpu(sb->chunksize)/2); | |
291 | break; | |
292 | default: break; | |
293 | } | |
294 | printf("\n"); | |
295 | printf(" Array Slot : %d (", __le32_to_cpu(sb->dev_number)); | |
296 | for (i= __le32_to_cpu(sb->max_dev); i> 0 ; i--) | |
297 | if (__le16_to_cpu(sb->dev_roles[i-1]) != 0xffff) | |
298 | break; | |
299 | for (d=0; d < i; d++) { | |
300 | int role = __le16_to_cpu(sb->dev_roles[d]); | |
301 | if (d) printf(", "); | |
302 | if (role == 0xffff) printf("empty"); | |
303 | else if(role == 0xfffe) printf("failed"); | |
304 | else printf("%d", role); | |
305 | } | |
306 | printf(")\n"); | |
307 | printf(" Array State : "); | |
308 | for (d=0; d<__le32_to_cpu(sb->raid_disks); d++) { | |
309 | int cnt = 0; | |
310 | int me = 0; | |
311 | int i; | |
312 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { | |
313 | int role = __le16_to_cpu(sb->dev_roles[i]); | |
314 | if (role == d) { | |
315 | if (i == __le32_to_cpu(sb->dev_number)) | |
316 | me = 1; | |
317 | cnt++; | |
318 | } | |
319 | } | |
320 | if (cnt > 1) printf("?"); | |
321 | else if (cnt == 1 && me) printf("U"); | |
322 | else if (cnt == 1) printf("u"); | |
323 | else printf ("_"); | |
324 | } | |
325 | faulty = 0; | |
326 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { | |
327 | int role = __le16_to_cpu(sb->dev_roles[i]); | |
328 | if (role == 0xFFFE) | |
329 | faulty++; | |
330 | } | |
331 | if (faulty) printf(" %d failed", faulty); | |
332 | printf("\n"); | |
333 | } | |
334 | ||
335 | ||
336 | static void brief_examine_super1(void *sbv) | |
337 | { | |
338 | struct mdp_superblock_1 *sb = sbv; | |
339 | int i; | |
340 | char *nm; | |
341 | char *c=map_num(pers, __le32_to_cpu(sb->level)); | |
342 | ||
343 | nm = strchr(sb->set_name, ':'); | |
344 | if (nm) | |
345 | nm++; | |
346 | else if (sb->set_name[0]) | |
347 | nm = sb->set_name; | |
348 | else | |
349 | nm = "??"; | |
350 | ||
351 | printf("ARRAY /dev/md/%s level=%s metadata=1 num-devices=%d UUID=", | |
352 | nm, | |
353 | c?c:"-unknown-", __le32_to_cpu(sb->raid_disks)); | |
354 | for (i=0; i<16; i++) { | |
355 | if ((i&3)==0 && i != 0) printf(":"); | |
356 | printf("%02x", sb->set_uuid[i]); | |
357 | } | |
358 | if (sb->set_name[0]) | |
359 | printf(" name=%.32s", sb->set_name); | |
360 | printf("\n"); | |
361 | } | |
362 | ||
363 | static void detail_super1(void *sbv, char *homehost) | |
364 | { | |
365 | struct mdp_superblock_1 *sb = sbv; | |
366 | int i; | |
367 | int l = homehost ? strlen(homehost) : 0; | |
368 | ||
369 | printf(" Name : %.32s", sb->set_name); | |
370 | if (l > 0 && l < 32 && | |
371 | sb->set_name[l] == ':' && | |
372 | strncmp(sb->set_name, homehost, l) == 0) | |
373 | printf(" (local to host %s)", homehost); | |
374 | printf("\n UUID : "); | |
375 | for (i=0; i<16; i++) { | |
376 | if ((i&3)==0 && i != 0) printf(":"); | |
377 | printf("%02x", sb->set_uuid[i]); | |
378 | } | |
379 | printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb->events)); | |
380 | } | |
381 | ||
382 | static void brief_detail_super1(void *sbv) | |
383 | { | |
384 | struct mdp_superblock_1 *sb = sbv; | |
385 | int i; | |
386 | ||
387 | if (sb->set_name[0]) | |
388 | printf(" name=%.32s", sb->set_name); | |
389 | printf(" UUID="); | |
390 | for (i=0; i<16; i++) { | |
391 | if ((i&3)==0 && i != 0) printf(":"); | |
392 | printf("%02x", sb->set_uuid[i]); | |
393 | } | |
394 | } | |
395 | ||
396 | #endif | |
397 | ||
398 | static int match_home1(void *sbv, char *homehost) | |
399 | { | |
400 | struct mdp_superblock_1 *sb = sbv; | |
401 | int l = homehost ? strlen(homehost) : 0; | |
402 | ||
403 | return (l > 0 && l < 32 && | |
404 | sb->set_name[l] == ':' && | |
405 | strncmp(sb->set_name, homehost, l) == 0); | |
406 | } | |
407 | ||
408 | static void uuid_from_super1(int uuid[4], void * sbv) | |
409 | { | |
410 | struct mdp_superblock_1 *super = sbv; | |
411 | char *cuuid = (char*)uuid; | |
412 | int i; | |
413 | for (i=0; i<16; i++) | |
414 | cuuid[i] = super->set_uuid[i]; | |
415 | } | |
416 | ||
417 | static void getinfo_super1(struct mdinfo *info, void *sbv) | |
418 | { | |
419 | struct mdp_superblock_1 *sb = sbv; | |
420 | int working = 0; | |
421 | int i; | |
422 | int role; | |
423 | ||
424 | info->array.major_version = 1; | |
425 | info->array.minor_version = __le32_to_cpu(sb->feature_map); | |
426 | info->array.patch_version = 0; | |
427 | info->array.raid_disks = __le32_to_cpu(sb->raid_disks); | |
428 | info->array.level = __le32_to_cpu(sb->level); | |
429 | info->array.layout = __le32_to_cpu(sb->layout); | |
430 | info->array.md_minor = -1; | |
431 | info->array.ctime = __le64_to_cpu(sb->ctime); | |
432 | info->array.utime = __le64_to_cpu(sb->utime); | |
433 | info->array.chunk_size = __le32_to_cpu(sb->chunksize)*512; | |
434 | info->array.state = (__le64_to_cpu(sb->resync_offset)+1) ? 0 : 1; | |
435 | ||
436 | info->data_offset = __le64_to_cpu(sb->data_offset); | |
437 | info->component_size = __le64_to_cpu(sb->size); | |
438 | ||
439 | info->disk.major = 0; | |
440 | info->disk.minor = 0; | |
441 | info->disk.number = __le32_to_cpu(sb->dev_number); | |
442 | if (__le32_to_cpu(sb->dev_number) >= __le32_to_cpu(sb->max_dev) || | |
443 | __le32_to_cpu(sb->max_dev) > 512) | |
444 | role = 0xfffe; | |
445 | else | |
446 | role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]); | |
447 | ||
448 | info->disk.raid_disk = -1; | |
449 | switch(role) { | |
450 | case 0xFFFF: | |
451 | info->disk.state = 2; /* spare: ACTIVE, not sync, not faulty */ | |
452 | break; | |
453 | case 0xFFFE: | |
454 | info->disk.state = 1; /* faulty */ | |
455 | break; | |
456 | default: | |
457 | info->disk.state = 6; /* active and in sync */ | |
458 | info->disk.raid_disk = role; | |
459 | } | |
460 | info->events = __le64_to_cpu(sb->events); | |
461 | ||
462 | memcpy(info->uuid, sb->set_uuid, 16); | |
463 | ||
464 | strncpy(info->name, sb->set_name, 32); | |
465 | info->name[32] = 0; | |
466 | ||
467 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) { | |
468 | info->reshape_active = 1; | |
469 | info->reshape_progress = __le64_to_cpu(sb->reshape_position); | |
470 | info->new_level = __le32_to_cpu(sb->new_level); | |
471 | info->delta_disks = __le32_to_cpu(sb->delta_disks); | |
472 | info->new_layout = __le32_to_cpu(sb->new_layout); | |
473 | info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9; | |
474 | } else | |
475 | info->reshape_active = 0; | |
476 | ||
477 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { | |
478 | role = __le16_to_cpu(sb->dev_roles[i]); | |
479 | if (/*role == 0xFFFF || */role < info->array.raid_disks) | |
480 | working++; | |
481 | } | |
482 | ||
483 | info->array.working_disks = working; | |
484 | } | |
485 | ||
486 | static int update_super1(struct mdinfo *info, void *sbv, char *update, | |
487 | char *devname, int verbose, | |
488 | int uuid_set, char *homehost) | |
489 | { | |
490 | /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made. | |
491 | * For others, the return value is ignored. | |
492 | */ | |
493 | int rv = 0; | |
494 | struct mdp_superblock_1 *sb = sbv; | |
495 | ||
496 | if (strcmp(update, "force-one")==0) { | |
497 | /* Not enough devices for a working array, | |
498 | * so bring this one up-to-date | |
499 | */ | |
500 | if (sb->events != __cpu_to_le64(info->events)) | |
501 | rv = 1; | |
502 | sb->events = __cpu_to_le64(info->events); | |
503 | } | |
504 | if (strcmp(update, "force-array")==0) { | |
505 | /* Degraded array and 'force' requests to | |
506 | * maybe need to mark it 'clean'. | |
507 | */ | |
508 | switch(__le32_to_cpu(sb->level)) { | |
509 | case 5: case 4: case 6: | |
510 | /* need to force clean */ | |
511 | if (sb->resync_offset != ~0ULL) | |
512 | rv = 1; | |
513 | sb->resync_offset = ~0ULL; | |
514 | } | |
515 | } | |
516 | if (strcmp(update, "assemble")==0) { | |
517 | int d = info->disk.number; | |
518 | int want; | |
519 | if (info->disk.state == 6) | |
520 | want = __cpu_to_le32(info->disk.raid_disk); | |
521 | else | |
522 | want = 0xFFFF; | |
523 | if (sb->dev_roles[d] != want) { | |
524 | sb->dev_roles[d] = want; | |
525 | rv = 1; | |
526 | } | |
527 | } | |
528 | if (strcmp(update, "grow") == 0) { | |
529 | sb->raid_disks = __cpu_to_le32(info->array.raid_disks); | |
530 | /* As we are just adding a spare, there is no need to | |
531 | * make any change to the dev_roles array | |
532 | */ | |
533 | } | |
534 | if (strcmp(update, "resync") == 0) { | |
535 | /* make sure resync happens */ | |
536 | sb->resync_offset = 0ULL; | |
537 | } | |
538 | if (strcmp(update, "uuid") == 0) { | |
539 | if (super1.swapuuid) { | |
540 | unsigned char *ac = (unsigned char *)sb->set_uuid; | |
541 | unsigned char *bc = (unsigned char *)info->uuid; | |
542 | int i; | |
543 | for (i=0; i<16; i+= 4) { | |
544 | ac[i+0] = bc[i+3]; | |
545 | ac[i+1] = bc[i+2]; | |
546 | ac[i+2] = bc[i+1]; | |
547 | ac[i+3] = bc[i+0]; | |
548 | } | |
549 | } else | |
550 | memcpy(sb->set_uuid, info->uuid, 16); | |
551 | ||
552 | if (__le32_to_cpu(sb->feature_map)&MD_FEATURE_BITMAP_OFFSET) { | |
553 | struct bitmap_super_s *bm; | |
554 | bm = (struct bitmap_super_s*)(sbv+1024); | |
555 | memcpy(bm->uuid, sb->set_uuid, 16); | |
556 | } | |
557 | } | |
558 | if (strcmp(update, "homehost") == 0 && | |
559 | homehost) { | |
560 | char *c; | |
561 | update = "name"; | |
562 | c = strchr(sb->set_name, ':'); | |
563 | if (c) | |
564 | strncpy(info->name, c+1, 31 - (c-sb->set_name)); | |
565 | else | |
566 | strncpy(info->name, sb->set_name, 32); | |
567 | info->name[32] = 0; | |
568 | } | |
569 | if (strcmp(update, "name") == 0) { | |
570 | if (info->name[0] == 0) | |
571 | sprintf(info->name, "%d", info->array.md_minor); | |
572 | memset(sb->set_name, 0, sizeof(sb->set_name)); | |
573 | if (homehost && | |
574 | strchr(info->name, ':') == NULL && | |
575 | strlen(homehost)+1+strlen(info->name) < 32) { | |
576 | strcpy(sb->set_name, homehost); | |
577 | strcat(sb->set_name, ":"); | |
578 | strcat(sb->set_name, info->name); | |
579 | } else | |
580 | strcpy(sb->set_name, info->name); | |
581 | } | |
582 | if (strcmp(update, "devicesize") == 0 && | |
583 | __le64_to_cpu(sb->super_offset) < | |
584 | __le64_to_cpu(sb->data_offset)) { | |
585 | /* set data_size to device size less data_offset */ | |
586 | struct misc_dev_info *misc = (struct misc_dev_info*) | |
587 | (sbv + 1024 + sizeof(struct bitmap_super_s)); | |
588 | printf("Size was %llu\n", __le64_to_cpu(sb->data_size)); | |
589 | sb->data_size = __cpu_to_le64( | |
590 | misc->device_size - __le64_to_cpu(sb->data_offset)); | |
591 | printf("Size is %llu\n", __le64_to_cpu(sb->data_size)); | |
592 | } | |
593 | if (strcmp(update, "_reshape_progress")==0) | |
594 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
595 | ||
596 | sb->sb_csum = calc_sb_1_csum(sb); | |
597 | return rv; | |
598 | } | |
599 | ||
600 | static int init_super1(struct supertype *st, void **sbp, mdu_array_info_t *info, | |
601 | unsigned long long size, char *name, char *homehost) | |
602 | { | |
603 | struct mdp_superblock_1 *sb = malloc(1024 + sizeof(bitmap_super_t) + | |
604 | sizeof(struct misc_dev_info)); | |
605 | int spares; | |
606 | int rfd; | |
607 | char defname[10]; | |
608 | memset(sb, 0, 1024); | |
609 | ||
610 | if (info->major_version == -1) { | |
611 | /* zeroing superblock */ | |
612 | *sbp = sb; | |
613 | return 0; | |
614 | } | |
615 | ||
616 | spares = info->working_disks - info->active_disks; | |
617 | if (info->raid_disks + spares > 384) { | |
618 | fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n", | |
619 | info->raid_disks , spares, 384); | |
620 | return 0; | |
621 | } | |
622 | ||
623 | sb->magic = __cpu_to_le32(MD_SB_MAGIC); | |
624 | sb->major_version = __cpu_to_le32(1); | |
625 | sb->feature_map = 0; | |
626 | sb->pad0 = 0; | |
627 | ||
628 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
629 | read(rfd, sb->set_uuid, 16) != 16) { | |
630 | *(__u32*)(sb->set_uuid) = random(); | |
631 | *(__u32*)(sb->set_uuid+4) = random(); | |
632 | *(__u32*)(sb->set_uuid+8) = random(); | |
633 | *(__u32*)(sb->set_uuid+12) = random(); | |
634 | } | |
635 | if (rfd >= 0) close(rfd); | |
636 | ||
637 | if (name == NULL || *name == 0) { | |
638 | sprintf(defname, "%d", info->md_minor); | |
639 | name = defname; | |
640 | } | |
641 | memset(sb->set_name, 0, 32); | |
642 | if (homehost && | |
643 | strchr(name, ':')== NULL && | |
644 | strlen(homehost)+1+strlen(name) < 32) { | |
645 | strcpy(sb->set_name, homehost); | |
646 | strcat(sb->set_name, ":"); | |
647 | strcat(sb->set_name, name); | |
648 | } else | |
649 | strcpy(sb->set_name, name); | |
650 | ||
651 | sb->ctime = __cpu_to_le64((unsigned long long)time(0)); | |
652 | sb->level = __cpu_to_le32(info->level); | |
653 | sb->layout = __cpu_to_le32(info->layout); | |
654 | sb->size = __cpu_to_le64(size*2ULL); | |
655 | sb->chunksize = __cpu_to_le32(info->chunk_size>>9); | |
656 | sb->raid_disks = __cpu_to_le32(info->raid_disks); | |
657 | ||
658 | sb->data_offset = __cpu_to_le64(0); | |
659 | sb->data_size = __cpu_to_le64(0); | |
660 | sb->super_offset = __cpu_to_le64(0); | |
661 | sb->recovery_offset = __cpu_to_le64(0); | |
662 | ||
663 | sb->utime = sb->ctime; | |
664 | sb->events = __cpu_to_le64(1); | |
665 | if (info->state & (1<<MD_SB_CLEAN)) | |
666 | sb->resync_offset = ~0ULL; | |
667 | else | |
668 | sb->resync_offset = 0; | |
669 | sb->max_dev = __cpu_to_le32((1024- sizeof(struct mdp_superblock_1))/ | |
670 | sizeof(sb->dev_roles[0])); | |
671 | memset(sb->pad3, 0, sizeof(sb->pad3)); | |
672 | ||
673 | memset(sb->dev_roles, 0xff, 1024 - sizeof(struct mdp_superblock_1)); | |
674 | ||
675 | *sbp = sb; | |
676 | return 1; | |
677 | } | |
678 | ||
679 | /* Add a device to the superblock being created */ | |
680 | static void add_to_super1(void *sbv, mdu_disk_info_t *dk) | |
681 | { | |
682 | struct mdp_superblock_1 *sb = sbv; | |
683 | __u16 *rp = sb->dev_roles + dk->number; | |
684 | if ((dk->state & 6) == 6) /* active, sync */ | |
685 | *rp = __cpu_to_le16(dk->raid_disk); | |
686 | else if ((dk->state & ~2) == 0) /* active or idle -> spare */ | |
687 | *rp = 0xffff; | |
688 | else | |
689 | *rp = 0xfffe; | |
690 | } | |
691 | ||
692 | static void locate_bitmap1(struct supertype *st, int fd, void *sbv); | |
693 | ||
694 | static int store_super1(struct supertype *st, int fd, void *sbv) | |
695 | { | |
696 | struct mdp_superblock_1 *sb = sbv; | |
697 | unsigned long long sb_offset; | |
698 | int sbsize; | |
699 | unsigned long long dsize; | |
700 | ||
701 | if (!get_dev_size(fd, NULL, &dsize)) | |
702 | return 1; | |
703 | ||
704 | dsize >>= 9; | |
705 | ||
706 | if (dsize < 24) | |
707 | return 2; | |
708 | ||
709 | /* | |
710 | * Calculate the position of the superblock. | |
711 | * It is always aligned to a 4K boundary and | |
712 | * depending on minor_version, it can be: | |
713 | * 0: At least 8K, but less than 12K, from end of device | |
714 | * 1: At start of device | |
715 | * 2: 4K from start of device. | |
716 | */ | |
717 | switch(st->minor_version) { | |
718 | case 0: | |
719 | sb_offset = dsize; | |
720 | sb_offset -= 8*2; | |
721 | sb_offset &= ~(4*2-1); | |
722 | break; | |
723 | case 1: | |
724 | sb_offset = 0; | |
725 | break; | |
726 | case 2: | |
727 | sb_offset = 4*2; | |
728 | break; | |
729 | default: | |
730 | return -EINVAL; | |
731 | } | |
732 | ||
733 | ||
734 | ||
735 | if (sb_offset != __le64_to_cpu(sb->super_offset) && | |
736 | 0 != __le64_to_cpu(sb->super_offset) | |
737 | ) { | |
738 | fprintf(stderr, Name ": internal error - sb_offset is wrong\n"); | |
739 | abort(); | |
740 | } | |
741 | ||
742 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) | |
743 | return 3; | |
744 | ||
745 | sbsize = sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev); | |
746 | ||
747 | if (write(fd, sb, sbsize) != sbsize) | |
748 | return 4; | |
749 | ||
750 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
751 | struct bitmap_super_s *bm = (struct bitmap_super_s*) | |
752 | (((char*)sb)+1024); | |
753 | if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) { | |
754 | locate_bitmap1(st, fd, sbv); | |
755 | if (write(fd, bm, sizeof(*bm)) != sizeof(*bm)) | |
756 | return 5; | |
757 | } | |
758 | } | |
759 | fsync(fd); | |
760 | return 0; | |
761 | } | |
762 | ||
763 | static int load_super1(struct supertype *st, int fd, void **sbp, char *devname); | |
764 | ||
765 | static int write_init_super1(struct supertype *st, void *sbv, | |
766 | mdu_disk_info_t *dinfo, char *devname) | |
767 | { | |
768 | struct mdp_superblock_1 *sb = sbv; | |
769 | void *refsbv = NULL; | |
770 | int fd = open(devname, O_RDWR | O_EXCL); | |
771 | int rfd; | |
772 | int rv; | |
773 | int bm_space; | |
774 | ||
775 | unsigned long space; | |
776 | unsigned long long dsize, array_size; | |
777 | long long sb_offset; | |
778 | ||
779 | ||
780 | if (fd < 0) { | |
781 | fprintf(stderr, Name ": Failed to open %s to write superblock\n", | |
782 | devname); | |
783 | return -1; | |
784 | } | |
785 | ||
786 | sb->dev_number = __cpu_to_le32(dinfo->number); | |
787 | if (dinfo->state & (1<<MD_DISK_WRITEMOSTLY)) | |
788 | sb->devflags |= __cpu_to_le32(WriteMostly1); | |
789 | ||
790 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
791 | read(rfd, sb->device_uuid, 16) != 16) { | |
792 | *(__u32*)(sb->device_uuid) = random(); | |
793 | *(__u32*)(sb->device_uuid+4) = random(); | |
794 | *(__u32*)(sb->device_uuid+8) = random(); | |
795 | *(__u32*)(sb->device_uuid+12) = random(); | |
796 | } | |
797 | if (rfd >= 0) close(rfd); | |
798 | sb->events = 0; | |
799 | ||
800 | if (load_super1(st, fd, &refsbv, NULL)==0) { | |
801 | struct mdp_superblock_1 *refsb = refsbv; | |
802 | ||
803 | memcpy(sb->device_uuid, refsb->device_uuid, 16); | |
804 | if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) { | |
805 | /* same array, so preserve events and dev_number */ | |
806 | sb->events = refsb->events; | |
807 | /* bugs in 2.6.17 and earlier mean the dev_number | |
808 | * chosen in Manage must be preserved | |
809 | */ | |
810 | if (get_linux_version() >= 2006018) | |
811 | sb->dev_number = refsb->dev_number; | |
812 | } | |
813 | free(refsb); | |
814 | } | |
815 | ||
816 | if (!get_dev_size(fd, NULL, &dsize)) | |
817 | return 1; | |
818 | dsize >>= 9; | |
819 | ||
820 | if (dsize < 24) { | |
821 | close(fd); | |
822 | return 2; | |
823 | } | |
824 | ||
825 | ||
826 | /* | |
827 | * Calculate the position of the superblock. | |
828 | * It is always aligned to a 4K boundary and | |
829 | * depending on minor_version, it can be: | |
830 | * 0: At least 8K, but less than 12K, from end of device | |
831 | * 1: At start of device | |
832 | * 2: 4K from start of device. | |
833 | * Depending on the array size, we might leave extra space | |
834 | * for a bitmap. | |
835 | */ | |
836 | array_size = __le64_to_cpu(sb->size); | |
837 | /* work out how much space we left of a bitmap */ | |
838 | if (array_size >= 200*1024*1024*2) | |
839 | bm_space = 128*2; | |
840 | else if (array_size > 8*1024*1024*2) | |
841 | bm_space = 64*2; | |
842 | else | |
843 | bm_space = 0; | |
844 | ||
845 | switch(st->minor_version) { | |
846 | case 0: | |
847 | sb_offset = dsize; | |
848 | sb_offset -= 8*2; | |
849 | sb_offset &= ~(4*2-1); | |
850 | sb->super_offset = __cpu_to_le64(sb_offset); | |
851 | sb->data_offset = __cpu_to_le64(0); | |
852 | sb->data_size = __cpu_to_le64(sb_offset - bm_space); | |
853 | break; | |
854 | case 1: | |
855 | sb->super_offset = __cpu_to_le64(0); | |
856 | sb->data_offset = __cpu_to_le64(bm_space + 4*2); | |
857 | sb->data_size = __cpu_to_le64(dsize - bm_space - 4*2); | |
858 | break; | |
859 | case 2: | |
860 | sb_offset = 4*2; | |
861 | if (dsize - 4*2 - 64*2 >= array_size && array_size > 8*1024*1024*2) | |
862 | space = 64*2; | |
863 | else | |
864 | space = 4*2; | |
865 | sb->super_offset = __cpu_to_le64(4*2); | |
866 | sb->data_offset = __cpu_to_le64(4*2 + 4*2 + bm_space); | |
867 | sb->data_size = __cpu_to_le64(dsize - 4*2 - 4*2 - bm_space ); | |
868 | break; | |
869 | default: | |
870 | return -EINVAL; | |
871 | } | |
872 | ||
873 | ||
874 | sb->sb_csum = calc_sb_1_csum(sb); | |
875 | rv = store_super1(st, fd, sb); | |
876 | if (rv) | |
877 | fprintf(stderr, Name ": failed to write superblock to %s\n", devname); | |
878 | ||
879 | if (rv == 0 && (__le32_to_cpu(sb->feature_map) & 1)) | |
880 | rv = st->ss->write_bitmap(st, fd, sbv); | |
881 | close(fd); | |
882 | return rv; | |
883 | } | |
884 | ||
885 | static int compare_super1(void **firstp, void *secondv) | |
886 | { | |
887 | /* | |
888 | * return: | |
889 | * 0 same, or first was empty, and second was copied | |
890 | * 1 second had wrong number | |
891 | * 2 wrong uuid | |
892 | * 3 wrong other info | |
893 | */ | |
894 | struct mdp_superblock_1 *first = *firstp; | |
895 | struct mdp_superblock_1 *second = secondv; | |
896 | ||
897 | if (second->magic != __cpu_to_le32(MD_SB_MAGIC)) | |
898 | return 1; | |
899 | if (second->major_version != __cpu_to_le32(1)) | |
900 | return 1; | |
901 | ||
902 | if (!first) { | |
903 | first = malloc(1024+sizeof(bitmap_super_t) + | |
904 | sizeof(struct misc_dev_info)); | |
905 | memcpy(first, second, 1024+sizeof(bitmap_super_t) + | |
906 | sizeof(struct misc_dev_info)); | |
907 | *firstp = first; | |
908 | return 0; | |
909 | } | |
910 | if (memcmp(first->set_uuid, second->set_uuid, 16)!= 0) | |
911 | return 2; | |
912 | ||
913 | if (first->ctime != second->ctime || | |
914 | first->level != second->level || | |
915 | first->layout != second->layout || | |
916 | first->size != second->size || | |
917 | first->chunksize != second->chunksize || | |
918 | first->raid_disks != second->raid_disks) | |
919 | return 3; | |
920 | return 0; | |
921 | } | |
922 | ||
923 | static int load_super1(struct supertype *st, int fd, void **sbp, char *devname) | |
924 | { | |
925 | unsigned long long dsize; | |
926 | unsigned long long sb_offset; | |
927 | struct mdp_superblock_1 *super; | |
928 | int uuid[4]; | |
929 | struct bitmap_super_s *bsb; | |
930 | struct misc_dev_info *misc; | |
931 | ||
932 | ||
933 | if (st->ss == NULL) { | |
934 | int bestvers = -1; | |
935 | __u64 bestctime = 0; | |
936 | /* guess... choose latest ctime */ | |
937 | st->ss = &super1; | |
938 | for (st->minor_version = 0; st->minor_version <= 2 ; st->minor_version++) { | |
939 | switch(load_super1(st, fd, sbp, devname)) { | |
940 | case 0: super = *sbp; | |
941 | if (bestvers == -1 || | |
942 | bestctime < __le64_to_cpu(super->ctime)) { | |
943 | bestvers = st->minor_version; | |
944 | bestctime = __le64_to_cpu(super->ctime); | |
945 | } | |
946 | free(super); | |
947 | *sbp = NULL; | |
948 | break; | |
949 | case 1: st->ss = NULL; return 1; /*bad device */ | |
950 | case 2: break; /* bad, try next */ | |
951 | } | |
952 | } | |
953 | if (bestvers != -1) { | |
954 | int rv; | |
955 | st->minor_version = bestvers; | |
956 | st->ss = &super1; | |
957 | st->max_devs = 384; | |
958 | rv = load_super1(st, fd, sbp, devname); | |
959 | if (rv) st->ss = NULL; | |
960 | return rv; | |
961 | } | |
962 | st->ss = NULL; | |
963 | return 2; | |
964 | } | |
965 | if (!get_dev_size(fd, devname, &dsize)) | |
966 | return 1; | |
967 | dsize >>= 9; | |
968 | ||
969 | if (dsize < 24) { | |
970 | if (devname) | |
971 | fprintf(stderr, Name ": %s is too small for md: size is %llu sectors.\n", | |
972 | devname, dsize); | |
973 | return 1; | |
974 | } | |
975 | ||
976 | /* | |
977 | * Calculate the position of the superblock. | |
978 | * It is always aligned to a 4K boundary and | |
979 | * depending on minor_version, it can be: | |
980 | * 0: At least 8K, but less than 12K, from end of device | |
981 | * 1: At start of device | |
982 | * 2: 4K from start of device. | |
983 | */ | |
984 | switch(st->minor_version) { | |
985 | case 0: | |
986 | sb_offset = dsize; | |
987 | sb_offset -= 8*2; | |
988 | sb_offset &= ~(4*2-1); | |
989 | break; | |
990 | case 1: | |
991 | sb_offset = 0; | |
992 | break; | |
993 | case 2: | |
994 | sb_offset = 4*2; | |
995 | break; | |
996 | default: | |
997 | return -EINVAL; | |
998 | } | |
999 | ||
1000 | ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */ | |
1001 | ||
1002 | ||
1003 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) { | |
1004 | if (devname) | |
1005 | fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n", | |
1006 | devname, strerror(errno)); | |
1007 | return 1; | |
1008 | } | |
1009 | ||
1010 | super = malloc(1024 + sizeof(bitmap_super_t) + | |
1011 | sizeof(struct misc_dev_info)); | |
1012 | ||
1013 | if (read(fd, super, 1024) != 1024) { | |
1014 | if (devname) | |
1015 | fprintf(stderr, Name ": Cannot read superblock on %s\n", | |
1016 | devname); | |
1017 | free(super); | |
1018 | return 1; | |
1019 | } | |
1020 | ||
1021 | if (__le32_to_cpu(super->magic) != MD_SB_MAGIC) { | |
1022 | if (devname) | |
1023 | fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n", | |
1024 | devname, MD_SB_MAGIC, __le32_to_cpu(super->magic)); | |
1025 | free(super); | |
1026 | return 2; | |
1027 | } | |
1028 | ||
1029 | if (__le32_to_cpu(super->major_version) != 1) { | |
1030 | if (devname) | |
1031 | fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n", | |
1032 | devname, __le32_to_cpu(super->major_version)); | |
1033 | free(super); | |
1034 | return 2; | |
1035 | } | |
1036 | if (__le64_to_cpu(super->super_offset) != sb_offset) { | |
1037 | if (devname) | |
1038 | fprintf(stderr, Name ": No superblock found on %s (super_offset is wrong)\n", | |
1039 | devname); | |
1040 | free(super); | |
1041 | return 2; | |
1042 | } | |
1043 | *sbp = super; | |
1044 | ||
1045 | bsb = (struct bitmap_super_s *)(((char*)super)+1024); | |
1046 | ||
1047 | misc = (struct misc_dev_info*) (bsb+1); | |
1048 | misc->device_size = dsize; | |
1049 | ||
1050 | /* Now check on the bitmap superblock */ | |
1051 | if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0) | |
1052 | return 0; | |
1053 | /* Read the bitmap superblock and make sure it looks | |
1054 | * valid. If it doesn't clear the bit. An --assemble --force | |
1055 | * should get that written out. | |
1056 | */ | |
1057 | locate_bitmap1(st, fd, super); | |
1058 | if (read(fd, ((char*)super)+1024, sizeof(struct bitmap_super_s)) | |
1059 | != sizeof(struct bitmap_super_s)) | |
1060 | goto no_bitmap; | |
1061 | ||
1062 | uuid_from_super1(uuid, super); | |
1063 | if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC || | |
1064 | memcmp(bsb->uuid, uuid, 16) != 0) | |
1065 | goto no_bitmap; | |
1066 | return 0; | |
1067 | ||
1068 | no_bitmap: | |
1069 | super->feature_map = __cpu_to_le32(__le32_to_cpu(super->feature_map) & ~1); | |
1070 | return 0; | |
1071 | } | |
1072 | ||
1073 | ||
1074 | static struct supertype *match_metadata_desc1(char *arg) | |
1075 | { | |
1076 | struct supertype *st = malloc(sizeof(*st)); | |
1077 | if (!st) return st; | |
1078 | ||
1079 | st->ss = &super1; | |
1080 | st->max_devs = 384; | |
1081 | if (strcmp(arg, "1") == 0 || | |
1082 | strcmp(arg, "1.0") == 0 || | |
1083 | strcmp(arg, "default/large") == 0) { | |
1084 | st->minor_version = 0; | |
1085 | return st; | |
1086 | } | |
1087 | if (strcmp(arg, "1.1") == 0) { | |
1088 | st->minor_version = 1; | |
1089 | return st; | |
1090 | } | |
1091 | if (strcmp(arg, "1.2") == 0) { | |
1092 | st->minor_version = 2; | |
1093 | return st; | |
1094 | } | |
1095 | ||
1096 | free(st); | |
1097 | return NULL; | |
1098 | } | |
1099 | ||
1100 | /* find available size on device with this devsize, using | |
1101 | * superblock type st, and reserving 'reserve' sectors for | |
1102 | * a possible bitmap | |
1103 | */ | |
1104 | static __u64 avail_size1(struct supertype *st, __u64 devsize) | |
1105 | { | |
1106 | if (devsize < 24) | |
1107 | return 0; | |
1108 | ||
1109 | /* if the device is bigger than 8Gig, save 64k for bitmap usage, | |
1110 | * if bigger than 200Gig, save 128k | |
1111 | */ | |
1112 | if (devsize-64*2 >= 200*1024*1024*2) | |
1113 | devsize -= 128*2; | |
1114 | else if (devsize >= 8*1024*1024*2) | |
1115 | devsize -= 64*2; | |
1116 | ||
1117 | switch(st->minor_version) { | |
1118 | case 0: | |
1119 | /* at end */ | |
1120 | return ((devsize - 8*2 ) & ~(4*2-1)); | |
1121 | case 1: | |
1122 | /* at start, 4K for superblock and possible bitmap */ | |
1123 | return devsize - 4*2; | |
1124 | case 2: | |
1125 | /* 4k from start, 4K for superblock and possible bitmap */ | |
1126 | return devsize - (4+4)*2; | |
1127 | } | |
1128 | return 0; | |
1129 | } | |
1130 | ||
1131 | static int | |
1132 | add_internal_bitmap1(struct supertype *st, void *sbv, | |
1133 | int *chunkp, int delay, int write_behind, | |
1134 | unsigned long long size, | |
1135 | int may_change, int major) | |
1136 | { | |
1137 | /* | |
1138 | * If not may_change, then this is a 'Grow', and the bitmap | |
1139 | * must fit after the superblock. | |
1140 | * If may_change, then this is create, and we can put the bitmap | |
1141 | * before the superblock if we like, or may move the start. | |
1142 | * If !may_change, the bitmap MUST live at offset of 1K, until | |
1143 | * we get a sysfs interface. | |
1144 | * | |
1145 | * size is in sectors, chunk is in bytes !!! | |
1146 | */ | |
1147 | ||
1148 | unsigned long long bits; | |
1149 | unsigned long long max_bits; | |
1150 | unsigned long long min_chunk; | |
1151 | long offset; | |
1152 | int chunk = *chunkp; | |
1153 | int room = 0; | |
1154 | struct mdp_superblock_1 *sb = sbv; | |
1155 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + 1024); | |
1156 | ||
1157 | switch(st->minor_version) { | |
1158 | case 0: | |
1159 | /* either 3K after the superblock, or some amount of space | |
1160 | * before. | |
1161 | */ | |
1162 | if (may_change) { | |
1163 | /* We are creating array, so we *know* how much room has | |
1164 | * been left. | |
1165 | */ | |
1166 | offset = 0; | |
1167 | if (__le64_to_cpu(sb->size) >= 200*1024*1024*2) | |
1168 | room = 128*2; | |
1169 | else if (__le64_to_cpu(sb->size) > 8*1024*1024*2) | |
1170 | room = 64*2; | |
1171 | else { | |
1172 | room = 3*2; | |
1173 | offset = 2; | |
1174 | } | |
1175 | } else { | |
1176 | room = __le64_to_cpu(sb->super_offset) | |
1177 | - __le64_to_cpu(sb->data_offset) | |
1178 | - __le64_to_cpu(sb->data_size); | |
1179 | /* remove '1 ||' when we can set offset via sysfs */ | |
1180 | if (1 || (room < 3*2 && | |
1181 | __le32_to_cpu(sb->max_dev) <= 384)) { | |
1182 | room = 3*2; | |
1183 | offset = 1*2; | |
1184 | } else { | |
1185 | offset = 0; /* means movable offset */ | |
1186 | } | |
1187 | } | |
1188 | break; | |
1189 | case 1: | |
1190 | case 2: /* between superblock and data */ | |
1191 | if (may_change) { | |
1192 | offset = 4*2; | |
1193 | if (__le64_to_cpu(sb->size) >= 200*1024*1024*2) | |
1194 | room = 128*2; | |
1195 | else if (__le64_to_cpu(sb->size) > 8*1024*1024*2) | |
1196 | room = 64*2; | |
1197 | else | |
1198 | room = 3*2; | |
1199 | } else { | |
1200 | room = __le64_to_cpu(sb->data_offset) | |
1201 | - __le64_to_cpu(sb->super_offset); | |
1202 | if (1 || __le32_to_cpu(sb->max_dev) <= 384) { | |
1203 | room -= 2; | |
1204 | offset = 2; | |
1205 | } else { | |
1206 | room -= 4*2; | |
1207 | offset = 4*2; | |
1208 | } | |
1209 | } | |
1210 | break; | |
1211 | } | |
1212 | ||
1213 | if (chunk == UnSet && room > 128*2) | |
1214 | /* Limit to 128K of bitmap when chunk size not requested */ | |
1215 | room = 128*2; | |
1216 | ||
1217 | max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8; | |
1218 | ||
1219 | min_chunk = 4096; /* sub-page chunks don't work yet.. */ | |
1220 | bits = (size*512)/min_chunk +1; | |
1221 | while (bits > max_bits) { | |
1222 | min_chunk *= 2; | |
1223 | bits = (bits+1)/2; | |
1224 | } | |
1225 | if (chunk == UnSet) | |
1226 | chunk = min_chunk; | |
1227 | else if (chunk < min_chunk) | |
1228 | return 0; /* chunk size too small */ | |
1229 | if (chunk == 0) /* rounding problem */ | |
1230 | return 0; | |
1231 | ||
1232 | if (offset == 0) { | |
1233 | bits = (size*512) / chunk + 1; | |
1234 | room = ((bits+7)/8 + sizeof(bitmap_super_t) +511)/512; | |
1235 | offset = -room; | |
1236 | } | |
1237 | ||
1238 | sb->bitmap_offset = __cpu_to_le32(offset); | |
1239 | ||
1240 | sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | 1); | |
1241 | memset(bms, 0, sizeof(*bms)); | |
1242 | bms->magic = __cpu_to_le32(BITMAP_MAGIC); | |
1243 | bms->version = __cpu_to_le32(major); | |
1244 | uuid_from_super1((int*)bms->uuid, sb); | |
1245 | bms->chunksize = __cpu_to_le32(chunk); | |
1246 | bms->daemon_sleep = __cpu_to_le32(delay); | |
1247 | bms->sync_size = __cpu_to_le64(size); | |
1248 | bms->write_behind = __cpu_to_le32(write_behind); | |
1249 | ||
1250 | *chunkp = chunk; | |
1251 | return 1; | |
1252 | } | |
1253 | ||
1254 | ||
1255 | static void locate_bitmap1(struct supertype *st, int fd, void *sbv) | |
1256 | { | |
1257 | unsigned long long offset; | |
1258 | struct mdp_superblock_1 *sb; | |
1259 | int mustfree = 0; | |
1260 | ||
1261 | if (!sbv) { | |
1262 | if (st->ss->load_super(st, fd, &sbv, NULL)) | |
1263 | return; /* no error I hope... */ | |
1264 | mustfree = 1; | |
1265 | } | |
1266 | sb = sbv; | |
1267 | ||
1268 | offset = __le64_to_cpu(sb->super_offset); | |
1269 | offset += (long) __le32_to_cpu(sb->bitmap_offset); | |
1270 | if (mustfree) | |
1271 | free(sb); | |
1272 | lseek64(fd, offset<<9, 0); | |
1273 | } | |
1274 | ||
1275 | static int write_bitmap1(struct supertype *st, int fd, void *sbv) | |
1276 | { | |
1277 | struct mdp_superblock_1 *sb = sbv; | |
1278 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+1024); | |
1279 | int rv = 0; | |
1280 | ||
1281 | int towrite, n; | |
1282 | char buf[4096]; | |
1283 | ||
1284 | locate_bitmap1(st, fd, sbv); | |
1285 | ||
1286 | if (write(fd, ((char*)sb)+1024, sizeof(bitmap_super_t)) != | |
1287 | sizeof(bitmap_super_t)) | |
1288 | return -2; | |
1289 | towrite = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9); | |
1290 | towrite = (towrite+7) >> 3; /* bits to bytes */ | |
1291 | memset(buf, 0xff, sizeof(buf)); | |
1292 | while (towrite > 0) { | |
1293 | n = towrite; | |
1294 | if (n > sizeof(buf)) | |
1295 | n = sizeof(buf); | |
1296 | n = write(fd, buf, n); | |
1297 | if (n > 0) | |
1298 | towrite -= n; | |
1299 | else | |
1300 | break; | |
1301 | } | |
1302 | fsync(fd); | |
1303 | if (towrite) | |
1304 | rv = -2; | |
1305 | ||
1306 | return rv; | |
1307 | } | |
1308 | ||
1309 | struct superswitch super1 = { | |
1310 | #ifndef MDASSEMBLE | |
1311 | .examine_super = examine_super1, | |
1312 | .brief_examine_super = brief_examine_super1, | |
1313 | .detail_super = detail_super1, | |
1314 | .brief_detail_super = brief_detail_super1, | |
1315 | #endif | |
1316 | .match_home = match_home1, | |
1317 | .uuid_from_super = uuid_from_super1, | |
1318 | .getinfo_super = getinfo_super1, | |
1319 | .update_super = update_super1, | |
1320 | .init_super = init_super1, | |
1321 | .add_to_super = add_to_super1, | |
1322 | .store_super = store_super1, | |
1323 | .write_init_super = write_init_super1, | |
1324 | .compare_super = compare_super1, | |
1325 | .load_super = load_super1, | |
1326 | .match_metadata_desc = match_metadata_desc1, | |
1327 | .avail_size = avail_size1, | |
1328 | .add_internal_bitmap = add_internal_bitmap1, | |
1329 | .locate_bitmap = locate_bitmap1, | |
1330 | .write_bitmap = write_bitmap1, | |
1331 | .major = 1, | |
1332 | #if __BYTE_ORDER == BIG_ENDIAN | |
1333 | .swapuuid = 0, | |
1334 | #else | |
1335 | .swapuuid = 1, | |
1336 | #endif | |
1337 | }; |