]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
4 | * Copyright (C) 2001-2009 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@suse.de> | |
23 | */ | |
24 | ||
25 | #include "mdadm.h" | |
26 | /* | |
27 | * The version-1 superblock : | |
28 | * All numeric fields are little-endian. | |
29 | * | |
30 | * total size: 256 bytes plus 2 per device. | |
31 | * 1K allows 384 devices. | |
32 | */ | |
33 | struct mdp_superblock_1 { | |
34 | /* constant array information - 128 bytes */ | |
35 | __u32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */ | |
36 | __u32 major_version; /* 1 */ | |
37 | __u32 feature_map; /* 0 for now */ | |
38 | __u32 pad0; /* always set to 0 when writing */ | |
39 | ||
40 | __u8 set_uuid[16]; /* user-space generated. */ | |
41 | char set_name[32]; /* set and interpreted by user-space */ | |
42 | ||
43 | __u64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/ | |
44 | __u32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */ | |
45 | __u32 layout; /* only for raid5 currently */ | |
46 | __u64 size; /* used size of component devices, in 512byte sectors */ | |
47 | ||
48 | __u32 chunksize; /* in 512byte sectors */ | |
49 | __u32 raid_disks; | |
50 | __u32 bitmap_offset; /* sectors after start of superblock that bitmap starts | |
51 | * NOTE: signed, so bitmap can be before superblock | |
52 | * only meaningful of feature_map[0] is set. | |
53 | */ | |
54 | ||
55 | /* These are only valid with feature bit '4' */ | |
56 | __u32 new_level; /* new level we are reshaping to */ | |
57 | __u64 reshape_position; /* next address in array-space for reshape */ | |
58 | __u32 delta_disks; /* change in number of raid_disks */ | |
59 | __u32 new_layout; /* new layout */ | |
60 | __u32 new_chunk; /* new chunk size (bytes) */ | |
61 | __u32 new_offset; /* signed number to add to data_offset in new | |
62 | * layout. 0 == no-change. This can be | |
63 | * different on each device in the array. | |
64 | */ | |
65 | ||
66 | /* constant this-device information - 64 bytes */ | |
67 | __u64 data_offset; /* sector start of data, often 0 */ | |
68 | __u64 data_size; /* sectors in this device that can be used for data */ | |
69 | __u64 super_offset; /* sector start of this superblock */ | |
70 | __u64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */ | |
71 | __u32 dev_number; /* permanent identifier of this device - not role in raid */ | |
72 | __u32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */ | |
73 | __u8 device_uuid[16]; /* user-space setable, ignored by kernel */ | |
74 | __u8 devflags; /* per-device flags. Only one defined...*/ | |
75 | #define WriteMostly1 1 /* mask for writemostly flag in above */ | |
76 | /* bad block log. If there are any bad blocks the feature flag is set. | |
77 | * if offset and size are non-zero, that space is reserved and available. | |
78 | */ | |
79 | __u8 bblog_shift; /* shift from sectors to block size for badblocklist */ | |
80 | __u16 bblog_size; /* number of sectors reserved for badblocklist */ | |
81 | __u32 bblog_offset; /* sector offset from superblock to bblog, signed */ | |
82 | ||
83 | /* array state information - 64 bytes */ | |
84 | __u64 utime; /* 40 bits second, 24 btes microseconds */ | |
85 | __u64 events; /* incremented when superblock updated */ | |
86 | __u64 resync_offset; /* data before this offset (from data_offset) known to be in sync */ | |
87 | __u32 sb_csum; /* checksum upto dev_roles[max_dev] */ | |
88 | __u32 max_dev; /* size of dev_roles[] array to consider */ | |
89 | __u8 pad3[64-32]; /* set to 0 when writing */ | |
90 | ||
91 | /* device state information. Indexed by dev_number. | |
92 | * 2 bytes per device | |
93 | * Note there are no per-device state flags. State information is rolled | |
94 | * into the 'roles' value. If a device is spare or faulty, then it doesn't | |
95 | * have a meaningful role. | |
96 | */ | |
97 | __u16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */ | |
98 | }; | |
99 | ||
100 | #define MAX_SB_SIZE 4096 | |
101 | /* bitmap super size is 256, but we round up to a sector for alignment */ | |
102 | #define BM_SUPER_SIZE 512 | |
103 | #define MAX_DEVS ((int)(MAX_SB_SIZE - sizeof(struct mdp_superblock_1)) / 2) | |
104 | #define SUPER1_SIZE (MAX_SB_SIZE + BM_SUPER_SIZE \ | |
105 | + sizeof(struct misc_dev_info)) | |
106 | ||
107 | struct misc_dev_info { | |
108 | __u64 device_size; | |
109 | }; | |
110 | ||
111 | /* feature_map bits */ | |
112 | #define MD_FEATURE_BITMAP_OFFSET 1 | |
113 | #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and | |
114 | * must be honoured | |
115 | */ | |
116 | #define MD_FEATURE_RESHAPE_ACTIVE 4 | |
117 | #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */ | |
118 | #define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an | |
119 | * active device with same 'role'. | |
120 | * 'recovery_offset' is also set. | |
121 | */ | |
122 | #define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number | |
123 | * of devices, but is going | |
124 | * backwards anyway. | |
125 | */ | |
126 | #define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */ | |
127 | #define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \ | |
128 | |MD_FEATURE_RECOVERY_OFFSET \ | |
129 | |MD_FEATURE_RESHAPE_ACTIVE \ | |
130 | |MD_FEATURE_BAD_BLOCKS \ | |
131 | |MD_FEATURE_REPLACEMENT \ | |
132 | |MD_FEATURE_RESHAPE_BACKWARDS \ | |
133 | |MD_FEATURE_NEW_OFFSET \ | |
134 | ) | |
135 | ||
136 | #ifndef offsetof | |
137 | #define offsetof(t,f) ((size_t)&(((t*)0)->f)) | |
138 | #endif | |
139 | static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb) | |
140 | { | |
141 | unsigned int disk_csum, csum; | |
142 | unsigned long long newcsum; | |
143 | int size = sizeof(*sb) + __le32_to_cpu(sb->max_dev)*2; | |
144 | unsigned int *isuper = (unsigned int*)sb; | |
145 | ||
146 | /* make sure I can count... */ | |
147 | if (offsetof(struct mdp_superblock_1,data_offset) != 128 || | |
148 | offsetof(struct mdp_superblock_1, utime) != 192 || | |
149 | sizeof(struct mdp_superblock_1) != 256) { | |
150 | fprintf(stderr, "WARNING - superblock isn't sized correctly\n"); | |
151 | } | |
152 | ||
153 | disk_csum = sb->sb_csum; | |
154 | sb->sb_csum = 0; | |
155 | newcsum = 0; | |
156 | for (; size>=4; size -= 4 ) { | |
157 | newcsum += __le32_to_cpu(*isuper); | |
158 | isuper++; | |
159 | } | |
160 | ||
161 | if (size == 2) | |
162 | newcsum += __le16_to_cpu(*(unsigned short*) isuper); | |
163 | ||
164 | csum = (newcsum & 0xffffffff) + (newcsum >> 32); | |
165 | sb->sb_csum = disk_csum; | |
166 | return __cpu_to_le32(csum); | |
167 | } | |
168 | ||
169 | /* | |
170 | * Information related to file descriptor used for aligned reads/writes. | |
171 | * Cache the block size. | |
172 | */ | |
173 | struct align_fd { | |
174 | int fd; | |
175 | int blk_sz; | |
176 | }; | |
177 | ||
178 | static void init_afd(struct align_fd *afd, int fd) | |
179 | { | |
180 | afd->fd = fd; | |
181 | ||
182 | if (ioctl(afd->fd, BLKSSZGET, &afd->blk_sz) != 0) | |
183 | afd->blk_sz = 512; | |
184 | } | |
185 | ||
186 | static char abuf[4096+4096]; | |
187 | static int aread(struct align_fd *afd, void *buf, int len) | |
188 | { | |
189 | /* aligned read. | |
190 | * On devices with a 4K sector size, we need to read | |
191 | * the full sector and copy relevant bits into | |
192 | * the buffer | |
193 | */ | |
194 | int bsize, iosize; | |
195 | char *b; | |
196 | int n; | |
197 | ||
198 | bsize = afd->blk_sz; | |
199 | ||
200 | if (!bsize || bsize > 4096 || len > 4096) { | |
201 | if (!bsize) | |
202 | fprintf(stderr, "WARNING - aread() called with " | |
203 | "invalid block size\n"); | |
204 | return -1; | |
205 | } | |
206 | b = ROUND_UP_PTR((char *)abuf, 4096); | |
207 | ||
208 | for (iosize = 0; iosize < len; iosize += bsize) | |
209 | ; | |
210 | n = read(afd->fd, b, iosize); | |
211 | if (n <= 0) | |
212 | return n; | |
213 | lseek(afd->fd, len - n, 1); | |
214 | if (n > len) | |
215 | n = len; | |
216 | memcpy(buf, b, n); | |
217 | return n; | |
218 | } | |
219 | ||
220 | static int awrite(struct align_fd *afd, void *buf, int len) | |
221 | { | |
222 | /* aligned write. | |
223 | * On devices with a 4K sector size, we need to write | |
224 | * the full sector. We pre-read if the sector is larger | |
225 | * than the write. | |
226 | * The address must be sector-aligned. | |
227 | */ | |
228 | int bsize, iosize; | |
229 | char *b; | |
230 | int n; | |
231 | ||
232 | bsize = afd->blk_sz; | |
233 | if (!bsize || bsize > 4096 || len > 4096) { | |
234 | if (!bsize) | |
235 | fprintf(stderr, "WARNING - awrite() called with " | |
236 | "invalid block size\n"); | |
237 | return -1; | |
238 | } | |
239 | b = ROUND_UP_PTR((char *)abuf, 4096); | |
240 | ||
241 | for (iosize = 0; iosize < len ; iosize += bsize) | |
242 | ; | |
243 | ||
244 | if (len != iosize) { | |
245 | n = read(afd->fd, b, iosize); | |
246 | if (n <= 0) | |
247 | return n; | |
248 | lseek(afd->fd, -n, 1); | |
249 | } | |
250 | ||
251 | memcpy(b, buf, len); | |
252 | n = write(afd->fd, b, iosize); | |
253 | if (n <= 0) | |
254 | return n; | |
255 | lseek(afd->fd, len - n, 1); | |
256 | return len; | |
257 | } | |
258 | ||
259 | #ifndef MDASSEMBLE | |
260 | static void examine_super1(struct supertype *st, char *homehost) | |
261 | { | |
262 | struct mdp_superblock_1 *sb = st->sb; | |
263 | time_t atime; | |
264 | unsigned int d; | |
265 | int role; | |
266 | int delta_extra = 0; | |
267 | int i; | |
268 | char *c; | |
269 | int l = homehost ? strlen(homehost) : 0; | |
270 | int layout; | |
271 | unsigned long long sb_offset; | |
272 | ||
273 | printf(" Magic : %08x\n", __le32_to_cpu(sb->magic)); | |
274 | printf(" Version : 1"); | |
275 | sb_offset = __le64_to_cpu(sb->super_offset); | |
276 | if (sb_offset <= 4) | |
277 | printf(".1\n"); | |
278 | else if (sb_offset <= 8) | |
279 | printf(".2\n"); | |
280 | else | |
281 | printf(".0\n"); | |
282 | printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb->feature_map)); | |
283 | printf(" Array UUID : "); | |
284 | for (i=0; i<16; i++) { | |
285 | if ((i&3)==0 && i != 0) printf(":"); | |
286 | printf("%02x", sb->set_uuid[i]); | |
287 | } | |
288 | printf("\n"); | |
289 | printf(" Name : %.32s", sb->set_name); | |
290 | if (l > 0 && l < 32 && | |
291 | sb->set_name[l] == ':' && | |
292 | strncmp(sb->set_name, homehost, l) == 0) | |
293 | printf(" (local to host %s)", homehost); | |
294 | printf("\n"); | |
295 | atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL; | |
296 | printf(" Creation Time : %.24s\n", ctime(&atime)); | |
297 | c=map_num(pers, __le32_to_cpu(sb->level)); | |
298 | printf(" Raid Level : %s\n", c?c:"-unknown-"); | |
299 | printf(" Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks)); | |
300 | printf("\n"); | |
301 | printf(" Avail Dev Size : %llu%s\n", | |
302 | (unsigned long long)__le64_to_cpu(sb->data_size), | |
303 | human_size(__le64_to_cpu(sb->data_size)<<9)); | |
304 | if (__le32_to_cpu(sb->level) > 0) { | |
305 | int ddsks = 0, ddsks_denom = 1; | |
306 | switch(__le32_to_cpu(sb->level)) { | |
307 | case 1: ddsks=1;break; | |
308 | case 4: | |
309 | case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break; | |
310 | case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break; | |
311 | case 10: | |
312 | layout = __le32_to_cpu(sb->layout); | |
313 | ddsks = __le32_to_cpu(sb->raid_disks); | |
314 | ddsks_denom = (layout&255) * ((layout>>8)&255); | |
315 | } | |
316 | if (ddsks) { | |
317 | long long asize = __le64_to_cpu(sb->size); | |
318 | asize = (asize << 9) * ddsks / ddsks_denom; | |
319 | printf(" Array Size : %llu%s\n", | |
320 | asize >> 10, human_size(asize)); | |
321 | } | |
322 | if (sb->size != sb->data_size) | |
323 | printf(" Used Dev Size : %llu%s\n", | |
324 | (unsigned long long)__le64_to_cpu(sb->size), | |
325 | human_size(__le64_to_cpu(sb->size)<<9)); | |
326 | } | |
327 | if (sb->data_offset) | |
328 | printf(" Data Offset : %llu sectors\n", | |
329 | (unsigned long long)__le64_to_cpu(sb->data_offset)); | |
330 | if (sb->new_offset) { | |
331 | unsigned long long offset = __le64_to_cpu(sb->data_offset); | |
332 | offset += (signed)(int32_t)__le32_to_cpu(sb->new_offset); | |
333 | printf(" New Offset : %llu sectors\n", offset); | |
334 | } | |
335 | printf(" Super Offset : %llu sectors\n", | |
336 | (unsigned long long)__le64_to_cpu(sb->super_offset)); | |
337 | if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET) | |
338 | printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->recovery_offset)); | |
339 | printf(" State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean"); | |
340 | printf(" Device UUID : "); | |
341 | for (i=0; i<16; i++) { | |
342 | if ((i&3)==0 && i != 0) printf(":"); | |
343 | printf("%02x", sb->device_uuid[i]); | |
344 | } | |
345 | printf("\n"); | |
346 | printf("\n"); | |
347 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
348 | printf("Internal Bitmap : %ld sectors from superblock\n", | |
349 | (long)(int32_t)__le32_to_cpu(sb->bitmap_offset)); | |
350 | } | |
351 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE)) { | |
352 | printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb->reshape_position)/2, | |
353 | human_size(__le64_to_cpu(sb->reshape_position)<<9)); | |
354 | if (__le32_to_cpu(sb->delta_disks)) { | |
355 | printf(" Delta Devices : %d", __le32_to_cpu(sb->delta_disks)); | |
356 | printf(" (%d->%d)\n", | |
357 | __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks), | |
358 | __le32_to_cpu(sb->raid_disks)); | |
359 | if ((int)__le32_to_cpu(sb->delta_disks) < 0) | |
360 | delta_extra = -__le32_to_cpu(sb->delta_disks); | |
361 | } | |
362 | if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) { | |
363 | c = map_num(pers, __le32_to_cpu(sb->new_level)); | |
364 | printf(" New Level : %s\n", c?c:"-unknown-"); | |
365 | } | |
366 | if (__le32_to_cpu(sb->new_layout) != __le32_to_cpu(sb->layout)) { | |
367 | if (__le32_to_cpu(sb->level) == 5) { | |
368 | c = map_num(r5layout, __le32_to_cpu(sb->new_layout)); | |
369 | printf(" New Layout : %s\n", c?c:"-unknown-"); | |
370 | } | |
371 | if (__le32_to_cpu(sb->level) == 6) { | |
372 | c = map_num(r6layout, __le32_to_cpu(sb->new_layout)); | |
373 | printf(" New Layout : %s\n", c?c:"-unknown-"); | |
374 | } | |
375 | if (__le32_to_cpu(sb->level) == 10) { | |
376 | printf(" New Layout :"); | |
377 | print_r10_layout(__le32_to_cpu(sb->new_layout)); | |
378 | printf("\n"); | |
379 | } | |
380 | } | |
381 | if (__le32_to_cpu(sb->new_chunk) != __le32_to_cpu(sb->chunksize)) | |
382 | printf(" New Chunksize : %dK\n", __le32_to_cpu(sb->new_chunk)/2); | |
383 | printf("\n"); | |
384 | } | |
385 | if (sb->devflags) { | |
386 | printf(" Flags :"); | |
387 | if (sb->devflags & WriteMostly1) | |
388 | printf(" write-mostly"); | |
389 | printf("\n"); | |
390 | } | |
391 | ||
392 | atime = __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL; | |
393 | printf(" Update Time : %.24s\n", ctime(&atime)); | |
394 | ||
395 | if (sb->bblog_size && sb->bblog_offset) { | |
396 | printf(" Bad Block Log : %d entries available at offset %ld sectors", | |
397 | __le16_to_cpu(sb->bblog_size)*512/8, | |
398 | (long)__le32_to_cpu(sb->bblog_offset)); | |
399 | if (sb->feature_map & | |
400 | __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
401 | printf(" - bad blocks present."); | |
402 | printf("\n"); | |
403 | } | |
404 | ||
405 | ||
406 | if (calc_sb_1_csum(sb) == sb->sb_csum) | |
407 | printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum)); | |
408 | else | |
409 | printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb->sb_csum), | |
410 | __le32_to_cpu(calc_sb_1_csum(sb))); | |
411 | printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb->events)); | |
412 | printf("\n"); | |
413 | if (__le32_to_cpu(sb->level) == 5) { | |
414 | c = map_num(r5layout, __le32_to_cpu(sb->layout)); | |
415 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
416 | } | |
417 | if (__le32_to_cpu(sb->level) == 6) { | |
418 | c = map_num(r6layout, __le32_to_cpu(sb->layout)); | |
419 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
420 | } | |
421 | if (__le32_to_cpu(sb->level) == 10) { | |
422 | int lo = __le32_to_cpu(sb->layout); | |
423 | printf(" Layout :"); | |
424 | print_r10_layout(lo); | |
425 | printf("\n"); | |
426 | } | |
427 | switch(__le32_to_cpu(sb->level)) { | |
428 | case 0: | |
429 | case 4: | |
430 | case 5: | |
431 | case 6: | |
432 | case 10: | |
433 | printf(" Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize)/2); | |
434 | break; | |
435 | case -1: | |
436 | printf(" Rounding : %dK\n", __le32_to_cpu(sb->chunksize)/2); | |
437 | break; | |
438 | default: break; | |
439 | } | |
440 | printf("\n"); | |
441 | #if 0 | |
442 | /* This turns out to just be confusing */ | |
443 | printf(" Array Slot : %d (", __le32_to_cpu(sb->dev_number)); | |
444 | for (i= __le32_to_cpu(sb->max_dev); i> 0 ; i--) | |
445 | if (__le16_to_cpu(sb->dev_roles[i-1]) != 0xffff) | |
446 | break; | |
447 | for (d=0; d < i; d++) { | |
448 | int role = __le16_to_cpu(sb->dev_roles[d]); | |
449 | if (d) printf(", "); | |
450 | if (role == 0xffff) printf("empty"); | |
451 | else if(role == 0xfffe) printf("failed"); | |
452 | else printf("%d", role); | |
453 | } | |
454 | printf(")\n"); | |
455 | #endif | |
456 | printf(" Device Role : "); | |
457 | d = __le32_to_cpu(sb->dev_number); | |
458 | if (d < __le32_to_cpu(sb->max_dev)) | |
459 | role = __le16_to_cpu(sb->dev_roles[d]); | |
460 | else | |
461 | role = 0xFFFF; | |
462 | if (role >= 0xFFFE) | |
463 | printf("spare\n"); | |
464 | else | |
465 | printf("Active device %d\n", role); | |
466 | ||
467 | printf(" Array State : "); | |
468 | for (d=0; d<__le32_to_cpu(sb->raid_disks) + delta_extra; d++) { | |
469 | int cnt = 0; | |
470 | unsigned int i; | |
471 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { | |
472 | unsigned int role = __le16_to_cpu(sb->dev_roles[i]); | |
473 | if (role == d) | |
474 | cnt++; | |
475 | } | |
476 | if (cnt > 1) printf("?"); | |
477 | else if (cnt == 1) printf("A"); | |
478 | else printf ("."); | |
479 | } | |
480 | #if 0 | |
481 | /* This is confusing too */ | |
482 | faulty = 0; | |
483 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { | |
484 | int role = __le16_to_cpu(sb->dev_roles[i]); | |
485 | if (role == 0xFFFE) | |
486 | faulty++; | |
487 | } | |
488 | if (faulty) printf(" %d failed", faulty); | |
489 | #endif | |
490 | printf(" ('A' == active, '.' == missing)"); | |
491 | printf("\n"); | |
492 | } | |
493 | ||
494 | ||
495 | static void brief_examine_super1(struct supertype *st, int verbose) | |
496 | { | |
497 | struct mdp_superblock_1 *sb = st->sb; | |
498 | int i; | |
499 | unsigned long long sb_offset; | |
500 | char *nm; | |
501 | char *c=map_num(pers, __le32_to_cpu(sb->level)); | |
502 | ||
503 | nm = strchr(sb->set_name, ':'); | |
504 | if (nm) | |
505 | nm++; | |
506 | else if (sb->set_name[0]) | |
507 | nm = sb->set_name; | |
508 | else | |
509 | nm = NULL; | |
510 | ||
511 | printf("ARRAY "); | |
512 | if (nm) { | |
513 | printf("/dev/md/"); | |
514 | print_escape(nm); | |
515 | putchar(' '); | |
516 | } | |
517 | if (verbose && c) | |
518 | printf(" level=%s", c); | |
519 | sb_offset = __le64_to_cpu(sb->super_offset); | |
520 | if (sb_offset <= 4) | |
521 | printf(" metadata=1.1 "); | |
522 | else if (sb_offset <= 8) | |
523 | printf(" metadata=1.2 "); | |
524 | else | |
525 | printf(" metadata=1.0 "); | |
526 | if (verbose) | |
527 | printf("num-devices=%d ", __le32_to_cpu(sb->raid_disks)); | |
528 | printf("UUID="); | |
529 | for (i=0; i<16; i++) { | |
530 | if ((i&3)==0 && i != 0) printf(":"); | |
531 | printf("%02x", sb->set_uuid[i]); | |
532 | } | |
533 | if (sb->set_name[0]) { | |
534 | printf(" name="); | |
535 | print_quoted(sb->set_name); | |
536 | } | |
537 | printf("\n"); | |
538 | } | |
539 | ||
540 | static void export_examine_super1(struct supertype *st) | |
541 | { | |
542 | struct mdp_superblock_1 *sb = st->sb; | |
543 | int i; | |
544 | int len = 32; | |
545 | int layout; | |
546 | ||
547 | printf("MD_LEVEL=%s\n", map_num(pers, __le32_to_cpu(sb->level))); | |
548 | printf("MD_DEVICES=%d\n", __le32_to_cpu(sb->raid_disks)); | |
549 | for (i=0; i<32; i++) | |
550 | if (sb->set_name[i] == '\n' || | |
551 | sb->set_name[i] == '\0') { | |
552 | len = i; | |
553 | break; | |
554 | } | |
555 | if (len) | |
556 | printf("MD_NAME=%.*s\n", len, sb->set_name); | |
557 | if (__le32_to_cpu(sb->level) > 0) { | |
558 | int ddsks = 0, ddsks_denom = 1; | |
559 | switch(__le32_to_cpu(sb->level)) { | |
560 | case 1: ddsks=1;break; | |
561 | case 4: | |
562 | case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break; | |
563 | case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break; | |
564 | case 10: | |
565 | layout = __le32_to_cpu(sb->layout); | |
566 | ddsks = __le32_to_cpu(sb->raid_disks); | |
567 | ddsks_denom = (layout&255) * ((layout>>8)&255); | |
568 | } | |
569 | if (ddsks) { | |
570 | long long asize = __le64_to_cpu(sb->size); | |
571 | asize = (asize << 9) * ddsks / ddsks_denom; | |
572 | printf("MD_ARRAY_SIZE=%s\n",human_size_brief(asize,JEDEC)); | |
573 | } | |
574 | } | |
575 | printf("MD_UUID="); | |
576 | for (i=0; i<16; i++) { | |
577 | if ((i&3)==0 && i != 0) printf(":"); | |
578 | printf("%02x", sb->set_uuid[i]); | |
579 | } | |
580 | printf("\n"); | |
581 | printf("MD_UPDATE_TIME=%llu\n", | |
582 | __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL); | |
583 | printf("MD_DEV_UUID="); | |
584 | for (i=0; i<16; i++) { | |
585 | if ((i&3)==0 && i != 0) printf(":"); | |
586 | printf("%02x", sb->device_uuid[i]); | |
587 | } | |
588 | printf("\n"); | |
589 | printf("MD_EVENTS=%llu\n", | |
590 | (unsigned long long)__le64_to_cpu(sb->events)); | |
591 | } | |
592 | ||
593 | static void detail_super1(struct supertype *st, char *homehost) | |
594 | { | |
595 | struct mdp_superblock_1 *sb = st->sb; | |
596 | int i; | |
597 | int l = homehost ? strlen(homehost) : 0; | |
598 | ||
599 | printf(" Name : %.32s", sb->set_name); | |
600 | if (l > 0 && l < 32 && | |
601 | sb->set_name[l] == ':' && | |
602 | strncmp(sb->set_name, homehost, l) == 0) | |
603 | printf(" (local to host %s)", homehost); | |
604 | printf("\n UUID : "); | |
605 | for (i=0; i<16; i++) { | |
606 | if ((i&3)==0 && i != 0) printf(":"); | |
607 | printf("%02x", sb->set_uuid[i]); | |
608 | } | |
609 | printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb->events)); | |
610 | } | |
611 | ||
612 | static void brief_detail_super1(struct supertype *st) | |
613 | { | |
614 | struct mdp_superblock_1 *sb = st->sb; | |
615 | int i; | |
616 | ||
617 | if (sb->set_name[0]) { | |
618 | printf(" name="); | |
619 | print_quoted(sb->set_name); | |
620 | } | |
621 | printf(" UUID="); | |
622 | for (i=0; i<16; i++) { | |
623 | if ((i&3)==0 && i != 0) printf(":"); | |
624 | printf("%02x", sb->set_uuid[i]); | |
625 | } | |
626 | } | |
627 | ||
628 | static void export_detail_super1(struct supertype *st) | |
629 | { | |
630 | struct mdp_superblock_1 *sb = st->sb; | |
631 | int i; | |
632 | int len = 32; | |
633 | ||
634 | for (i=0; i<32; i++) | |
635 | if (sb->set_name[i] == '\n' || | |
636 | sb->set_name[i] == '\0') { | |
637 | len = i; | |
638 | break; | |
639 | } | |
640 | if (len) | |
641 | printf("MD_NAME=%.*s\n", len, sb->set_name); | |
642 | } | |
643 | ||
644 | #endif | |
645 | ||
646 | static int match_home1(struct supertype *st, char *homehost) | |
647 | { | |
648 | struct mdp_superblock_1 *sb = st->sb; | |
649 | int l = homehost ? strlen(homehost) : 0; | |
650 | ||
651 | return (l > 0 && l < 32 && | |
652 | sb->set_name[l] == ':' && | |
653 | strncmp(sb->set_name, homehost, l) == 0); | |
654 | } | |
655 | ||
656 | static void uuid_from_super1(struct supertype *st, int uuid[4]) | |
657 | { | |
658 | struct mdp_superblock_1 *super = st->sb; | |
659 | char *cuuid = (char*)uuid; | |
660 | int i; | |
661 | for (i=0; i<16; i++) | |
662 | cuuid[i] = super->set_uuid[i]; | |
663 | } | |
664 | ||
665 | static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map) | |
666 | { | |
667 | struct mdp_superblock_1 *sb = st->sb; | |
668 | struct bitmap_super_s *bsb = (void*)(((char*)sb)+MAX_SB_SIZE); | |
669 | struct misc_dev_info *misc = (void*)(((char*)sb)+MAX_SB_SIZE+BM_SUPER_SIZE); | |
670 | int working = 0; | |
671 | unsigned int i; | |
672 | unsigned int role; | |
673 | unsigned int map_disks = info->array.raid_disks; | |
674 | unsigned long long super_offset; | |
675 | unsigned long long data_size; | |
676 | ||
677 | memset(info, 0, sizeof(*info)); | |
678 | info->array.major_version = 1; | |
679 | info->array.minor_version = st->minor_version; | |
680 | info->array.patch_version = 0; | |
681 | info->array.raid_disks = __le32_to_cpu(sb->raid_disks); | |
682 | info->array.level = __le32_to_cpu(sb->level); | |
683 | info->array.layout = __le32_to_cpu(sb->layout); | |
684 | info->array.md_minor = -1; | |
685 | info->array.ctime = __le64_to_cpu(sb->ctime); | |
686 | info->array.utime = __le64_to_cpu(sb->utime); | |
687 | info->array.chunk_size = __le32_to_cpu(sb->chunksize)*512; | |
688 | info->array.state = | |
689 | (__le64_to_cpu(sb->resync_offset) == MaxSector) | |
690 | ? 1 : 0; | |
691 | ||
692 | info->data_offset = __le64_to_cpu(sb->data_offset); | |
693 | info->component_size = __le64_to_cpu(sb->size); | |
694 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET)) | |
695 | info->bitmap_offset = (int32_t)__le32_to_cpu(sb->bitmap_offset); | |
696 | ||
697 | info->disk.major = 0; | |
698 | info->disk.minor = 0; | |
699 | info->disk.number = __le32_to_cpu(sb->dev_number); | |
700 | if (__le32_to_cpu(sb->dev_number) >= __le32_to_cpu(sb->max_dev) || | |
701 | __le32_to_cpu(sb->dev_number) >= MAX_DEVS) | |
702 | role = 0xfffe; | |
703 | else | |
704 | role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]); | |
705 | ||
706 | super_offset = __le64_to_cpu(sb->super_offset); | |
707 | data_size = __le64_to_cpu(sb->size); | |
708 | if (info->data_offset < super_offset) { | |
709 | unsigned long long end; | |
710 | info->space_before = info->data_offset; | |
711 | end = super_offset; | |
712 | if (info->bitmap_offset < 0) | |
713 | end += info->bitmap_offset; | |
714 | if (info->data_offset + data_size < end) | |
715 | info->space_after = end - data_size - info->data_offset; | |
716 | else | |
717 | info->space_after = 0; | |
718 | } else { | |
719 | info->space_before = (info->data_offset - | |
720 | super_offset); | |
721 | if (info->bitmap_offset > 0) { | |
722 | unsigned long long bmend = info->bitmap_offset; | |
723 | unsigned long long size = __le64_to_cpu(bsb->sync_size); | |
724 | size /= __le32_to_cpu(bsb->chunksize) >> 9; | |
725 | size = (size + 7) >> 3; | |
726 | size += sizeof(bitmap_super_t); | |
727 | size = ROUND_UP(size, 4096); | |
728 | size /= 512; | |
729 | size += bmend; | |
730 | if (size < info->space_before) | |
731 | info->space_before -= size; | |
732 | else | |
733 | info->space_before = 0; | |
734 | } else | |
735 | info->space_before -= 8; /* superblock */ | |
736 | info->space_after = misc->device_size - data_size - info->data_offset; | |
737 | } | |
738 | ||
739 | info->disk.raid_disk = -1; | |
740 | switch(role) { | |
741 | case 0xFFFF: | |
742 | info->disk.state = 0; /* spare: not active, not sync, not faulty */ | |
743 | break; | |
744 | case 0xFFFE: | |
745 | info->disk.state = 1; /* faulty */ | |
746 | break; | |
747 | default: | |
748 | info->disk.state = 6; /* active and in sync */ | |
749 | info->disk.raid_disk = role; | |
750 | } | |
751 | if (sb->devflags & WriteMostly1) | |
752 | info->disk.state |= (1 << MD_DISK_WRITEMOSTLY); | |
753 | info->events = __le64_to_cpu(sb->events); | |
754 | sprintf(info->text_version, "1.%d", st->minor_version); | |
755 | info->safe_mode_delay = 200; | |
756 | ||
757 | memcpy(info->uuid, sb->set_uuid, 16); | |
758 | ||
759 | strncpy(info->name, sb->set_name, 32); | |
760 | info->name[32] = 0; | |
761 | ||
762 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET)) | |
763 | info->recovery_start = __le32_to_cpu(sb->recovery_offset); | |
764 | else | |
765 | info->recovery_start = MaxSector; | |
766 | ||
767 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) { | |
768 | info->reshape_active = 1; | |
769 | if (info->array.level == 10) | |
770 | info->reshape_active |= RESHAPE_NO_BACKUP; | |
771 | info->reshape_progress = __le64_to_cpu(sb->reshape_position); | |
772 | info->new_level = __le32_to_cpu(sb->new_level); | |
773 | info->delta_disks = __le32_to_cpu(sb->delta_disks); | |
774 | info->new_layout = __le32_to_cpu(sb->new_layout); | |
775 | info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9; | |
776 | if (info->delta_disks < 0) | |
777 | info->array.raid_disks -= info->delta_disks; | |
778 | } else | |
779 | info->reshape_active = 0; | |
780 | ||
781 | info->recovery_blocked = info->reshape_active; | |
782 | ||
783 | if (map) | |
784 | for (i=0; i<map_disks; i++) | |
785 | map[i] = 0; | |
786 | for (i = 0; i < __le32_to_cpu(sb->max_dev); i++) { | |
787 | role = __le16_to_cpu(sb->dev_roles[i]); | |
788 | if (/*role == 0xFFFF || */role < (unsigned) info->array.raid_disks) { | |
789 | working++; | |
790 | if (map && role < map_disks) | |
791 | map[role] = 1; | |
792 | } | |
793 | } | |
794 | ||
795 | info->array.working_disks = working; | |
796 | } | |
797 | ||
798 | static struct mdinfo *container_content1(struct supertype *st, char *subarray) | |
799 | { | |
800 | struct mdinfo *info; | |
801 | ||
802 | if (subarray) | |
803 | return NULL; | |
804 | ||
805 | info = xmalloc(sizeof(*info)); | |
806 | getinfo_super1(st, info, NULL); | |
807 | return info; | |
808 | } | |
809 | ||
810 | static int update_super1(struct supertype *st, struct mdinfo *info, | |
811 | char *update, | |
812 | char *devname, int verbose, | |
813 | int uuid_set, char *homehost) | |
814 | { | |
815 | /* NOTE: for 'assemble' and 'force' we need to return non-zero | |
816 | * if any change was made. For others, the return value is | |
817 | * ignored. | |
818 | */ | |
819 | int rv = 0; | |
820 | struct mdp_superblock_1 *sb = st->sb; | |
821 | ||
822 | if (strcmp(update, "force-one")==0) { | |
823 | /* Not enough devices for a working array, | |
824 | * so bring this one up-to-date | |
825 | */ | |
826 | if (sb->events != __cpu_to_le64(info->events)) | |
827 | rv = 1; | |
828 | sb->events = __cpu_to_le64(info->events); | |
829 | } else if (strcmp(update, "force-array")==0) { | |
830 | /* Degraded array and 'force' requests to | |
831 | * maybe need to mark it 'clean'. | |
832 | */ | |
833 | switch(__le32_to_cpu(sb->level)) { | |
834 | case 5: case 4: case 6: | |
835 | /* need to force clean */ | |
836 | if (sb->resync_offset != MaxSector) | |
837 | rv = 1; | |
838 | sb->resync_offset = MaxSector; | |
839 | } | |
840 | } else if (strcmp(update, "assemble")==0) { | |
841 | int d = info->disk.number; | |
842 | int want; | |
843 | if (info->disk.state == 6) | |
844 | want = info->disk.raid_disk; | |
845 | else | |
846 | want = 0xFFFF; | |
847 | if (sb->dev_roles[d] != __cpu_to_le16(want)) { | |
848 | sb->dev_roles[d] = __cpu_to_le16(want); | |
849 | rv = 1; | |
850 | } | |
851 | if (info->reshape_active && | |
852 | sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) && | |
853 | info->delta_disks >= 0 && | |
854 | info->reshape_progress < __le64_to_cpu(sb->reshape_position)) { | |
855 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
856 | rv = 1; | |
857 | } | |
858 | if (info->reshape_active && | |
859 | sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) && | |
860 | info->delta_disks < 0 && | |
861 | info->reshape_progress > __le64_to_cpu(sb->reshape_position)) { | |
862 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
863 | rv = 1; | |
864 | } | |
865 | } else if (strcmp(update, "linear-grow-new") == 0) { | |
866 | unsigned int i; | |
867 | int rfd, fd; | |
868 | unsigned int max = __le32_to_cpu(sb->max_dev); | |
869 | ||
870 | for (i=0 ; i < max ; i++) | |
871 | if (__le16_to_cpu(sb->dev_roles[i]) >= 0xfffe) | |
872 | break; | |
873 | sb->dev_number = __cpu_to_le32(i); | |
874 | info->disk.number = i; | |
875 | if (max >= __le32_to_cpu(sb->max_dev)) | |
876 | sb->max_dev = __cpu_to_le32(max+1); | |
877 | ||
878 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
879 | read(rfd, sb->device_uuid, 16) != 16) { | |
880 | __u32 r[4] = {random(), random(), random(), random()}; | |
881 | memcpy(sb->device_uuid, r, 16); | |
882 | } | |
883 | if (rfd >= 0) | |
884 | close(rfd); | |
885 | ||
886 | sb->dev_roles[i] = | |
887 | __cpu_to_le16(info->disk.raid_disk); | |
888 | ||
889 | fd = open(devname, O_RDONLY); | |
890 | if (fd >= 0) { | |
891 | unsigned long long ds; | |
892 | get_dev_size(fd, devname, &ds); | |
893 | close(fd); | |
894 | ds >>= 9; | |
895 | if (__le64_to_cpu(sb->super_offset) < | |
896 | __le64_to_cpu(sb->data_offset)) { | |
897 | sb->data_size = __cpu_to_le64( | |
898 | ds - __le64_to_cpu(sb->data_offset)); | |
899 | } else { | |
900 | ds -= 8*2; | |
901 | ds &= ~(unsigned long long)(4*2-1); | |
902 | sb->super_offset = __cpu_to_le64(ds); | |
903 | sb->data_size = __cpu_to_le64( | |
904 | ds - __le64_to_cpu(sb->data_offset)); | |
905 | } | |
906 | } | |
907 | } else if (strcmp(update, "linear-grow-update") == 0) { | |
908 | sb->raid_disks = __cpu_to_le32(info->array.raid_disks); | |
909 | sb->dev_roles[info->disk.number] = | |
910 | __cpu_to_le16(info->disk.raid_disk); | |
911 | } else if (strcmp(update, "resync") == 0) { | |
912 | /* make sure resync happens */ | |
913 | sb->resync_offset = 0ULL; | |
914 | } else if (strcmp(update, "uuid") == 0) { | |
915 | copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid); | |
916 | ||
917 | if (__le32_to_cpu(sb->feature_map)&MD_FEATURE_BITMAP_OFFSET) { | |
918 | struct bitmap_super_s *bm; | |
919 | bm = (struct bitmap_super_s*)(st->sb+MAX_SB_SIZE); | |
920 | memcpy(bm->uuid, sb->set_uuid, 16); | |
921 | } | |
922 | } else if (strcmp(update, "no-bitmap") == 0) { | |
923 | sb->feature_map &= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET); | |
924 | } else if (strcmp(update, "bbl") == 0) { | |
925 | /* only possible if there is room after the bitmap, or if | |
926 | * there is no bitmap | |
927 | */ | |
928 | unsigned long long sb_offset = __le64_to_cpu(sb->super_offset); | |
929 | unsigned long long data_offset = __le64_to_cpu(sb->data_offset); | |
930 | long bitmap_offset = (long)__le64_to_cpu(sb->bitmap_offset); | |
931 | long bm_sectors = 0; | |
932 | long space; | |
933 | ||
934 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
935 | struct bitmap_super_s *bsb; | |
936 | bsb = (struct bitmap_super_s *)(((char*)sb)+MAX_SB_SIZE); | |
937 | bm_sectors = bitmap_sectors(bsb); | |
938 | } | |
939 | ||
940 | if (sb_offset < data_offset) { | |
941 | /* 1.1 or 1.2. Put bbl just before data | |
942 | */ | |
943 | long bb_offset; | |
944 | space = data_offset - sb_offset; | |
945 | bb_offset = space - 8; | |
946 | if (bm_sectors && bitmap_offset > 0) | |
947 | space -= (bitmap_offset + bm_sectors); | |
948 | else | |
949 | space -= 8; /* The superblock */ | |
950 | if (space >= 8) { | |
951 | sb->bblog_size = __cpu_to_le16(8); | |
952 | sb->bblog_offset = __cpu_to_le32(bb_offset); | |
953 | } | |
954 | } else { | |
955 | /* 1.0 - Put bbl just before super block */ | |
956 | if (bm_sectors && bitmap_offset < 0) | |
957 | space = -bitmap_offset - bm_sectors; | |
958 | else | |
959 | space = sb_offset - data_offset - | |
960 | __le64_to_cpu(sb->data_size); | |
961 | if (space >= 8) { | |
962 | sb->bblog_size = __cpu_to_le16(8); | |
963 | sb->bblog_offset = __cpu_to_le32((unsigned)-8); | |
964 | } | |
965 | } | |
966 | } else if (strcmp(update, "no-bbl") == 0) { | |
967 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
968 | pr_err("Cannot remove active bbl from %s\n",devname); | |
969 | else { | |
970 | sb->bblog_size = 0; | |
971 | sb->bblog_shift = 0; | |
972 | sb->bblog_offset = 0; | |
973 | } | |
974 | } else if (strcmp(update, "homehost") == 0 && | |
975 | homehost) { | |
976 | char *c; | |
977 | update = "name"; | |
978 | c = strchr(sb->set_name, ':'); | |
979 | if (c) | |
980 | strncpy(info->name, c+1, 31 - (c-sb->set_name)); | |
981 | else | |
982 | strncpy(info->name, sb->set_name, 32); | |
983 | info->name[32] = 0; | |
984 | } else if (strcmp(update, "name") == 0) { | |
985 | if (info->name[0] == 0) | |
986 | sprintf(info->name, "%d", info->array.md_minor); | |
987 | memset(sb->set_name, 0, sizeof(sb->set_name)); | |
988 | if (homehost && | |
989 | strchr(info->name, ':') == NULL && | |
990 | strlen(homehost)+1+strlen(info->name) < 32) { | |
991 | strcpy(sb->set_name, homehost); | |
992 | strcat(sb->set_name, ":"); | |
993 | strcat(sb->set_name, info->name); | |
994 | } else | |
995 | strcpy(sb->set_name, info->name); | |
996 | } else if (strcmp(update, "devicesize") == 0 && | |
997 | __le64_to_cpu(sb->super_offset) < | |
998 | __le64_to_cpu(sb->data_offset)) { | |
999 | /* set data_size to device size less data_offset */ | |
1000 | struct misc_dev_info *misc = (struct misc_dev_info*) | |
1001 | (st->sb + MAX_SB_SIZE + BM_SUPER_SIZE); | |
1002 | printf("Size was %llu\n", (unsigned long long) | |
1003 | __le64_to_cpu(sb->data_size)); | |
1004 | sb->data_size = __cpu_to_le64( | |
1005 | misc->device_size - __le64_to_cpu(sb->data_offset)); | |
1006 | printf("Size is %llu\n", (unsigned long long) | |
1007 | __le64_to_cpu(sb->data_size)); | |
1008 | } else if (strcmp(update, "_reshape_progress")==0) | |
1009 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
1010 | else if (strcmp(update, "writemostly")==0) | |
1011 | sb->devflags |= WriteMostly1; | |
1012 | else if (strcmp(update, "readwrite")==0) | |
1013 | sb->devflags &= ~WriteMostly1; | |
1014 | else | |
1015 | rv = -1; | |
1016 | ||
1017 | sb->sb_csum = calc_sb_1_csum(sb); | |
1018 | return rv; | |
1019 | } | |
1020 | ||
1021 | static int init_super1(struct supertype *st, mdu_array_info_t *info, | |
1022 | unsigned long long size, char *name, char *homehost, | |
1023 | int *uuid, unsigned long long data_offset) | |
1024 | { | |
1025 | struct mdp_superblock_1 *sb; | |
1026 | int spares; | |
1027 | int rfd; | |
1028 | char defname[10]; | |
1029 | int sbsize; | |
1030 | ||
1031 | if (posix_memalign((void**)&sb, 4096, SUPER1_SIZE) != 0) { | |
1032 | pr_err("%s could not allocate superblock\n", __func__); | |
1033 | return 0; | |
1034 | } | |
1035 | memset(sb, 0, SUPER1_SIZE); | |
1036 | ||
1037 | st->sb = sb; | |
1038 | if (info == NULL) { | |
1039 | /* zeroing superblock */ | |
1040 | return 0; | |
1041 | } | |
1042 | ||
1043 | spares = info->working_disks - info->active_disks; | |
1044 | if (info->raid_disks + spares > MAX_DEVS) { | |
1045 | pr_err("too many devices requested: %d+%d > %d\n", | |
1046 | info->raid_disks , spares, MAX_DEVS); | |
1047 | return 0; | |
1048 | } | |
1049 | ||
1050 | sb->magic = __cpu_to_le32(MD_SB_MAGIC); | |
1051 | sb->major_version = __cpu_to_le32(1); | |
1052 | sb->feature_map = 0; | |
1053 | sb->pad0 = 0; | |
1054 | ||
1055 | if (uuid) | |
1056 | copy_uuid(sb->set_uuid, uuid, super1.swapuuid); | |
1057 | else { | |
1058 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
1059 | read(rfd, sb->set_uuid, 16) != 16) { | |
1060 | __u32 r[4] = {random(), random(), random(), random()}; | |
1061 | memcpy(sb->set_uuid, r, 16); | |
1062 | } | |
1063 | if (rfd >= 0) close(rfd); | |
1064 | } | |
1065 | ||
1066 | if (name == NULL || *name == 0) { | |
1067 | sprintf(defname, "%d", info->md_minor); | |
1068 | name = defname; | |
1069 | } | |
1070 | if (homehost && | |
1071 | strchr(name, ':')== NULL && | |
1072 | strlen(homehost)+1+strlen(name) < 32) { | |
1073 | strcpy(sb->set_name, homehost); | |
1074 | strcat(sb->set_name, ":"); | |
1075 | strcat(sb->set_name, name); | |
1076 | } else | |
1077 | strcpy(sb->set_name, name); | |
1078 | ||
1079 | sb->ctime = __cpu_to_le64((unsigned long long)time(0)); | |
1080 | sb->level = __cpu_to_le32(info->level); | |
1081 | sb->layout = __cpu_to_le32(info->layout); | |
1082 | sb->size = __cpu_to_le64(size*2ULL); | |
1083 | sb->chunksize = __cpu_to_le32(info->chunk_size>>9); | |
1084 | sb->raid_disks = __cpu_to_le32(info->raid_disks); | |
1085 | ||
1086 | sb->data_offset = __cpu_to_le64(data_offset); | |
1087 | sb->data_size = __cpu_to_le64(0); | |
1088 | sb->super_offset = __cpu_to_le64(0); | |
1089 | sb->recovery_offset = __cpu_to_le64(0); | |
1090 | ||
1091 | sb->utime = sb->ctime; | |
1092 | sb->events = __cpu_to_le64(1); | |
1093 | if (info->state & (1<<MD_SB_CLEAN)) | |
1094 | sb->resync_offset = MaxSector; | |
1095 | else | |
1096 | sb->resync_offset = 0; | |
1097 | sbsize = sizeof(struct mdp_superblock_1) + 2 * (info->raid_disks + spares); | |
1098 | sbsize = ROUND_UP(sbsize, 512); | |
1099 | sb->max_dev = __cpu_to_le32((sbsize - sizeof(struct mdp_superblock_1)) / 2); | |
1100 | ||
1101 | memset(sb->dev_roles, 0xff, MAX_SB_SIZE - sizeof(struct mdp_superblock_1)); | |
1102 | ||
1103 | return 1; | |
1104 | } | |
1105 | ||
1106 | struct devinfo { | |
1107 | int fd; | |
1108 | char *devname; | |
1109 | long long data_offset; | |
1110 | mdu_disk_info_t disk; | |
1111 | struct devinfo *next; | |
1112 | }; | |
1113 | #ifndef MDASSEMBLE | |
1114 | /* Add a device to the superblock being created */ | |
1115 | static int add_to_super1(struct supertype *st, mdu_disk_info_t *dk, | |
1116 | int fd, char *devname, unsigned long long data_offset) | |
1117 | { | |
1118 | struct mdp_superblock_1 *sb = st->sb; | |
1119 | __u16 *rp = sb->dev_roles + dk->number; | |
1120 | struct devinfo *di, **dip; | |
1121 | ||
1122 | if ((dk->state & 6) == 6) /* active, sync */ | |
1123 | *rp = __cpu_to_le16(dk->raid_disk); | |
1124 | else if ((dk->state & ~2) == 0) /* active or idle -> spare */ | |
1125 | *rp = 0xffff; | |
1126 | else | |
1127 | *rp = 0xfffe; | |
1128 | ||
1129 | if (dk->number >= (int)__le32_to_cpu(sb->max_dev) && | |
1130 | __le32_to_cpu(sb->max_dev) < MAX_DEVS) | |
1131 | sb->max_dev = __cpu_to_le32(dk->number+1); | |
1132 | ||
1133 | sb->dev_number = __cpu_to_le32(dk->number); | |
1134 | sb->devflags = 0; /* don't copy another disks flags */ | |
1135 | sb->sb_csum = calc_sb_1_csum(sb); | |
1136 | ||
1137 | dip = (struct devinfo **)&st->info; | |
1138 | while (*dip) | |
1139 | dip = &(*dip)->next; | |
1140 | di = xmalloc(sizeof(struct devinfo)); | |
1141 | di->fd = fd; | |
1142 | di->devname = devname; | |
1143 | di->disk = *dk; | |
1144 | di->data_offset = data_offset; | |
1145 | di->next = NULL; | |
1146 | *dip = di; | |
1147 | ||
1148 | return 0; | |
1149 | } | |
1150 | #endif | |
1151 | ||
1152 | static void locate_bitmap1(struct supertype *st, int fd); | |
1153 | ||
1154 | static int store_super1(struct supertype *st, int fd) | |
1155 | { | |
1156 | struct mdp_superblock_1 *sb = st->sb; | |
1157 | unsigned long long sb_offset; | |
1158 | struct align_fd afd; | |
1159 | int sbsize; | |
1160 | unsigned long long dsize; | |
1161 | ||
1162 | if (!get_dev_size(fd, NULL, &dsize)) | |
1163 | return 1; | |
1164 | ||
1165 | dsize >>= 9; | |
1166 | ||
1167 | if (dsize < 24) | |
1168 | return 2; | |
1169 | ||
1170 | init_afd(&afd, fd); | |
1171 | ||
1172 | /* | |
1173 | * Calculate the position of the superblock. | |
1174 | * It is always aligned to a 4K boundary and | |
1175 | * depending on minor_version, it can be: | |
1176 | * 0: At least 8K, but less than 12K, from end of device | |
1177 | * 1: At start of device | |
1178 | * 2: 4K from start of device. | |
1179 | */ | |
1180 | switch(st->minor_version) { | |
1181 | case 0: | |
1182 | sb_offset = dsize; | |
1183 | sb_offset -= 8*2; | |
1184 | sb_offset &= ~(4*2-1); | |
1185 | break; | |
1186 | case 1: | |
1187 | sb_offset = 0; | |
1188 | break; | |
1189 | case 2: | |
1190 | sb_offset = 4*2; | |
1191 | break; | |
1192 | default: | |
1193 | return -EINVAL; | |
1194 | } | |
1195 | ||
1196 | ||
1197 | ||
1198 | if (sb_offset != __le64_to_cpu(sb->super_offset) && | |
1199 | 0 != __le64_to_cpu(sb->super_offset) | |
1200 | ) { | |
1201 | pr_err("internal error - sb_offset is wrong\n"); | |
1202 | abort(); | |
1203 | } | |
1204 | ||
1205 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) | |
1206 | return 3; | |
1207 | ||
1208 | sbsize = ROUND_UP(sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev), 512); | |
1209 | ||
1210 | if (awrite(&afd, sb, sbsize) != sbsize) | |
1211 | return 4; | |
1212 | ||
1213 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
1214 | struct bitmap_super_s *bm = (struct bitmap_super_s*) | |
1215 | (((char*)sb)+MAX_SB_SIZE); | |
1216 | if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) { | |
1217 | locate_bitmap1(st, fd); | |
1218 | if (awrite(&afd, bm, sizeof(*bm)) != sizeof(*bm)) | |
1219 | return 5; | |
1220 | } | |
1221 | } | |
1222 | fsync(fd); | |
1223 | return 0; | |
1224 | } | |
1225 | ||
1226 | static int load_super1(struct supertype *st, int fd, char *devname); | |
1227 | ||
1228 | static unsigned long choose_bm_space(unsigned long devsize) | |
1229 | { | |
1230 | /* if the device is bigger than 8Gig, save 64k for bitmap usage, | |
1231 | * if bigger than 200Gig, save 128k | |
1232 | * NOTE: result must be multiple of 4K else bad things happen | |
1233 | * on 4K-sector devices. | |
1234 | */ | |
1235 | if (devsize < 64*2) return 0; | |
1236 | if (devsize - 64*2 >= 200*1024*1024*2) | |
1237 | return 128*2; | |
1238 | if (devsize - 4*2 > 8*1024*1024*2) | |
1239 | return 64*2; | |
1240 | return 4*2; | |
1241 | } | |
1242 | ||
1243 | static void free_super1(struct supertype *st); | |
1244 | ||
1245 | #ifndef MDASSEMBLE | |
1246 | static int write_init_super1(struct supertype *st) | |
1247 | { | |
1248 | struct mdp_superblock_1 *sb = st->sb; | |
1249 | struct supertype *refst; | |
1250 | int rfd; | |
1251 | int rv = 0; | |
1252 | unsigned long long bm_space; | |
1253 | unsigned long long reserved; | |
1254 | struct devinfo *di; | |
1255 | unsigned long long dsize, array_size; | |
1256 | unsigned long long sb_offset, headroom; | |
1257 | unsigned long long data_offset; | |
1258 | ||
1259 | for (di = st->info; di; di = di->next) { | |
1260 | if (di->disk.state == 1) | |
1261 | continue; | |
1262 | if (di->fd < 0) | |
1263 | continue; | |
1264 | ||
1265 | while (Kill(di->devname, NULL, 0, -1, 1) == 0) | |
1266 | ; | |
1267 | ||
1268 | sb->dev_number = __cpu_to_le32(di->disk.number); | |
1269 | if (di->disk.state & (1<<MD_DISK_WRITEMOSTLY)) | |
1270 | sb->devflags |= WriteMostly1; | |
1271 | else | |
1272 | sb->devflags &= ~WriteMostly1; | |
1273 | ||
1274 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
1275 | read(rfd, sb->device_uuid, 16) != 16) { | |
1276 | __u32 r[4] = {random(), random(), random(), random()}; | |
1277 | memcpy(sb->device_uuid, r, 16); | |
1278 | } | |
1279 | if (rfd >= 0) | |
1280 | close(rfd); | |
1281 | ||
1282 | sb->events = 0; | |
1283 | ||
1284 | refst = dup_super(st); | |
1285 | if (load_super1(refst, di->fd, NULL)==0) { | |
1286 | struct mdp_superblock_1 *refsb = refst->sb; | |
1287 | ||
1288 | memcpy(sb->device_uuid, refsb->device_uuid, 16); | |
1289 | if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) { | |
1290 | /* same array, so preserve events and | |
1291 | * dev_number */ | |
1292 | sb->events = refsb->events; | |
1293 | /* bugs in 2.6.17 and earlier mean the | |
1294 | * dev_number chosen in Manage must be preserved | |
1295 | */ | |
1296 | if (get_linux_version() >= 2006018) | |
1297 | sb->dev_number = refsb->dev_number; | |
1298 | } | |
1299 | free_super1(refst); | |
1300 | } | |
1301 | free(refst); | |
1302 | ||
1303 | if (!get_dev_size(di->fd, NULL, &dsize)) { | |
1304 | rv = 1; | |
1305 | goto error_out; | |
1306 | } | |
1307 | dsize >>= 9; | |
1308 | ||
1309 | if (dsize < 24) { | |
1310 | close(di->fd); | |
1311 | rv = 2; | |
1312 | goto error_out; | |
1313 | } | |
1314 | ||
1315 | ||
1316 | /* | |
1317 | * Calculate the position of the superblock. | |
1318 | * It is always aligned to a 4K boundary and | |
1319 | * depending on minor_version, it can be: | |
1320 | * 0: At least 8K, but less than 12K, from end of device | |
1321 | * 1: At start of device | |
1322 | * 2: 4K from start of device. | |
1323 | * Depending on the array size, we might leave extra space | |
1324 | * for a bitmap. | |
1325 | * Also leave 4K for bad-block log. | |
1326 | */ | |
1327 | array_size = __le64_to_cpu(sb->size); | |
1328 | /* work out how much space we left for a bitmap, | |
1329 | * Add 8 sectors for bad block log */ | |
1330 | bm_space = choose_bm_space(array_size) + 8; | |
1331 | ||
1332 | /* We try to leave 0.1% at the start for reshape | |
1333 | * operations, but limit this to 128Meg (0.1% of 10Gig) | |
1334 | * which is plenty for efficient reshapes | |
1335 | * However we make it at least 2 chunks as one chunk | |
1336 | * is minimum needed for reshape. | |
1337 | */ | |
1338 | headroom = 128 * 1024 * 2; | |
1339 | while (headroom << 10 > array_size && | |
1340 | headroom/2 >= __le32_to_cpu(sb->chunksize) * 2) | |
1341 | headroom >>= 1; | |
1342 | ||
1343 | data_offset = di->data_offset; | |
1344 | switch(st->minor_version) { | |
1345 | case 0: | |
1346 | sb_offset = dsize; | |
1347 | sb_offset -= 8*2; | |
1348 | sb_offset &= ~(4*2-1); | |
1349 | sb->super_offset = __cpu_to_le64(sb_offset); | |
1350 | if (data_offset == INVALID_SECTORS) | |
1351 | sb->data_offset = 0; | |
1352 | if (sb_offset < array_size + bm_space) | |
1353 | bm_space = sb_offset - array_size; | |
1354 | sb->data_size = __cpu_to_le64(sb_offset - bm_space); | |
1355 | if (bm_space >= 8) { | |
1356 | sb->bblog_size = __cpu_to_le16(8); | |
1357 | sb->bblog_offset = __cpu_to_le32((unsigned)-8); | |
1358 | } | |
1359 | break; | |
1360 | case 1: | |
1361 | sb->super_offset = __cpu_to_le64(0); | |
1362 | if (data_offset == INVALID_SECTORS) { | |
1363 | reserved = bm_space + 4*2; | |
1364 | if (reserved < headroom) | |
1365 | reserved = headroom; | |
1366 | if (reserved + array_size > dsize) | |
1367 | reserved = dsize - array_size; | |
1368 | /* Try for multiple of 1Meg so it is nicely aligned */ | |
1369 | #define ONE_MEG (2*1024) | |
1370 | if (reserved > ONE_MEG) | |
1371 | reserved = (reserved/ONE_MEG) * ONE_MEG; | |
1372 | ||
1373 | /* force 4K alignment */ | |
1374 | reserved &= ~7ULL; | |
1375 | ||
1376 | } else | |
1377 | reserved = data_offset; | |
1378 | ||
1379 | sb->data_offset = __cpu_to_le64(reserved); | |
1380 | sb->data_size = __cpu_to_le64(dsize - reserved); | |
1381 | if (reserved >= 16) { | |
1382 | sb->bblog_size = __cpu_to_le16(8); | |
1383 | sb->bblog_offset = __cpu_to_le32(reserved-8); | |
1384 | } | |
1385 | break; | |
1386 | case 2: | |
1387 | sb_offset = 4*2; | |
1388 | sb->super_offset = __cpu_to_le64(4*2); | |
1389 | if (data_offset == INVALID_SECTORS) { | |
1390 | if (4*2 + 4*2 + bm_space + array_size | |
1391 | > dsize) | |
1392 | bm_space = dsize - array_size | |
1393 | - 4*2 - 4*2; | |
1394 | ||
1395 | reserved = bm_space + 4*2 + 4*2; | |
1396 | if (reserved < headroom) | |
1397 | reserved = headroom; | |
1398 | if (reserved + array_size > dsize) | |
1399 | reserved = dsize - array_size; | |
1400 | /* Try for multiple of 1Meg so it is nicely aligned */ | |
1401 | #define ONE_MEG (2*1024) | |
1402 | if (reserved > ONE_MEG) | |
1403 | reserved = (reserved/ONE_MEG) * ONE_MEG; | |
1404 | ||
1405 | /* force 4K alignment */ | |
1406 | reserved &= ~7ULL; | |
1407 | ||
1408 | } else | |
1409 | reserved = data_offset; | |
1410 | ||
1411 | sb->data_offset = __cpu_to_le64(reserved); | |
1412 | sb->data_size = __cpu_to_le64(dsize - reserved); | |
1413 | if (reserved >= 16+16) { | |
1414 | sb->bblog_size = __cpu_to_le16(8); | |
1415 | /* '8' sectors for the bblog, and another '8' | |
1416 | * because we want offset from superblock, not | |
1417 | * start of device. | |
1418 | */ | |
1419 | sb->bblog_offset = __cpu_to_le32(reserved-8-8); | |
1420 | } | |
1421 | break; | |
1422 | default: | |
1423 | pr_err("Failed to write invalid " | |
1424 | "metadata format 1.%i to %s\n", | |
1425 | st->minor_version, di->devname); | |
1426 | rv = -EINVAL; | |
1427 | goto out; | |
1428 | } | |
1429 | ||
1430 | sb->sb_csum = calc_sb_1_csum(sb); | |
1431 | rv = store_super1(st, di->fd); | |
1432 | if (rv == 0 && (__le32_to_cpu(sb->feature_map) & 1)) | |
1433 | rv = st->ss->write_bitmap(st, di->fd); | |
1434 | close(di->fd); | |
1435 | di->fd = -1; | |
1436 | if (rv) | |
1437 | goto error_out; | |
1438 | } | |
1439 | error_out: | |
1440 | if (rv) | |
1441 | pr_err("Failed to write metadata to %s\n", | |
1442 | di->devname); | |
1443 | out: | |
1444 | return rv; | |
1445 | } | |
1446 | #endif | |
1447 | ||
1448 | static int compare_super1(struct supertype *st, struct supertype *tst) | |
1449 | { | |
1450 | /* | |
1451 | * return: | |
1452 | * 0 same, or first was empty, and second was copied | |
1453 | * 1 second had wrong number | |
1454 | * 2 wrong uuid | |
1455 | * 3 wrong other info | |
1456 | */ | |
1457 | struct mdp_superblock_1 *first = st->sb; | |
1458 | struct mdp_superblock_1 *second = tst->sb; | |
1459 | ||
1460 | if (second->magic != __cpu_to_le32(MD_SB_MAGIC)) | |
1461 | return 1; | |
1462 | if (second->major_version != __cpu_to_le32(1)) | |
1463 | return 1; | |
1464 | ||
1465 | if (!first) { | |
1466 | if (posix_memalign((void**)&first, 4096, SUPER1_SIZE) != 0) { | |
1467 | pr_err("%s could not allocate superblock\n", __func__); | |
1468 | return 1; | |
1469 | } | |
1470 | memcpy(first, second, SUPER1_SIZE); | |
1471 | st->sb = first; | |
1472 | return 0; | |
1473 | } | |
1474 | if (memcmp(first->set_uuid, second->set_uuid, 16)!= 0) | |
1475 | return 2; | |
1476 | ||
1477 | if (first->ctime != second->ctime || | |
1478 | first->level != second->level || | |
1479 | first->layout != second->layout || | |
1480 | first->size != second->size || | |
1481 | first->chunksize != second->chunksize || | |
1482 | first->raid_disks != second->raid_disks) | |
1483 | return 3; | |
1484 | return 0; | |
1485 | } | |
1486 | ||
1487 | static int load_super1(struct supertype *st, int fd, char *devname) | |
1488 | { | |
1489 | unsigned long long dsize; | |
1490 | unsigned long long sb_offset; | |
1491 | struct mdp_superblock_1 *super; | |
1492 | int uuid[4]; | |
1493 | struct bitmap_super_s *bsb; | |
1494 | struct misc_dev_info *misc; | |
1495 | struct align_fd afd; | |
1496 | ||
1497 | free_super1(st); | |
1498 | ||
1499 | init_afd(&afd, fd); | |
1500 | ||
1501 | if (st->ss == NULL || st->minor_version == -1) { | |
1502 | int bestvers = -1; | |
1503 | struct supertype tst; | |
1504 | __u64 bestctime = 0; | |
1505 | /* guess... choose latest ctime */ | |
1506 | memset(&tst, 0, sizeof(tst)); | |
1507 | tst.ss = &super1; | |
1508 | for (tst.minor_version = 0; tst.minor_version <= 2 ; tst.minor_version++) { | |
1509 | switch(load_super1(&tst, fd, devname)) { | |
1510 | case 0: super = tst.sb; | |
1511 | if (bestvers == -1 || | |
1512 | bestctime < __le64_to_cpu(super->ctime)) { | |
1513 | bestvers = tst.minor_version; | |
1514 | bestctime = __le64_to_cpu(super->ctime); | |
1515 | } | |
1516 | free(super); | |
1517 | tst.sb = NULL; | |
1518 | break; | |
1519 | case 1: return 1; /*bad device */ | |
1520 | case 2: break; /* bad, try next */ | |
1521 | } | |
1522 | } | |
1523 | if (bestvers != -1) { | |
1524 | int rv; | |
1525 | tst.minor_version = bestvers; | |
1526 | tst.ss = &super1; | |
1527 | tst.max_devs = MAX_DEVS; | |
1528 | rv = load_super1(&tst, fd, devname); | |
1529 | if (rv == 0) | |
1530 | *st = tst; | |
1531 | return rv; | |
1532 | } | |
1533 | return 2; | |
1534 | } | |
1535 | if (!get_dev_size(fd, devname, &dsize)) | |
1536 | return 1; | |
1537 | dsize >>= 9; | |
1538 | ||
1539 | if (dsize < 24) { | |
1540 | if (devname) | |
1541 | pr_err("%s is too small for md: size is %llu sectors.\n", | |
1542 | devname, dsize); | |
1543 | return 1; | |
1544 | } | |
1545 | ||
1546 | /* | |
1547 | * Calculate the position of the superblock. | |
1548 | * It is always aligned to a 4K boundary and | |
1549 | * depending on minor_version, it can be: | |
1550 | * 0: At least 8K, but less than 12K, from end of device | |
1551 | * 1: At start of device | |
1552 | * 2: 4K from start of device. | |
1553 | */ | |
1554 | switch(st->minor_version) { | |
1555 | case 0: | |
1556 | sb_offset = dsize; | |
1557 | sb_offset -= 8*2; | |
1558 | sb_offset &= ~(4*2-1); | |
1559 | break; | |
1560 | case 1: | |
1561 | sb_offset = 0; | |
1562 | break; | |
1563 | case 2: | |
1564 | sb_offset = 4*2; | |
1565 | break; | |
1566 | default: | |
1567 | return -EINVAL; | |
1568 | } | |
1569 | ||
1570 | ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */ | |
1571 | ||
1572 | ||
1573 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) { | |
1574 | if (devname) | |
1575 | pr_err("Cannot seek to superblock on %s: %s\n", | |
1576 | devname, strerror(errno)); | |
1577 | return 1; | |
1578 | } | |
1579 | ||
1580 | if (posix_memalign((void**)&super, 4096, SUPER1_SIZE) != 0) { | |
1581 | pr_err("%s could not allocate superblock\n", | |
1582 | __func__); | |
1583 | return 1; | |
1584 | } | |
1585 | ||
1586 | if (aread(&afd, super, MAX_SB_SIZE) != MAX_SB_SIZE) { | |
1587 | if (devname) | |
1588 | pr_err("Cannot read superblock on %s\n", | |
1589 | devname); | |
1590 | free(super); | |
1591 | return 1; | |
1592 | } | |
1593 | ||
1594 | if (__le32_to_cpu(super->magic) != MD_SB_MAGIC) { | |
1595 | if (devname) | |
1596 | pr_err("No super block found on %s (Expected magic %08x, got %08x)\n", | |
1597 | devname, MD_SB_MAGIC, __le32_to_cpu(super->magic)); | |
1598 | free(super); | |
1599 | return 2; | |
1600 | } | |
1601 | ||
1602 | if (__le32_to_cpu(super->major_version) != 1) { | |
1603 | if (devname) | |
1604 | pr_err("Cannot interpret superblock on %s - version is %d\n", | |
1605 | devname, __le32_to_cpu(super->major_version)); | |
1606 | free(super); | |
1607 | return 2; | |
1608 | } | |
1609 | if (__le64_to_cpu(super->super_offset) != sb_offset) { | |
1610 | if (devname) | |
1611 | pr_err("No superblock found on %s (super_offset is wrong)\n", | |
1612 | devname); | |
1613 | free(super); | |
1614 | return 2; | |
1615 | } | |
1616 | st->sb = super; | |
1617 | ||
1618 | bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); | |
1619 | ||
1620 | misc = (struct misc_dev_info*) (((char*)super)+MAX_SB_SIZE+BM_SUPER_SIZE); | |
1621 | misc->device_size = dsize; | |
1622 | ||
1623 | /* Now check on the bitmap superblock */ | |
1624 | if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0) | |
1625 | return 0; | |
1626 | /* Read the bitmap superblock and make sure it looks | |
1627 | * valid. If it doesn't clear the bit. An --assemble --force | |
1628 | * should get that written out. | |
1629 | */ | |
1630 | locate_bitmap1(st, fd); | |
1631 | if (aread(&afd, bsb, 512) != 512) | |
1632 | goto no_bitmap; | |
1633 | ||
1634 | uuid_from_super1(st, uuid); | |
1635 | if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC || | |
1636 | memcmp(bsb->uuid, uuid, 16) != 0) | |
1637 | goto no_bitmap; | |
1638 | return 0; | |
1639 | ||
1640 | no_bitmap: | |
1641 | super->feature_map = __cpu_to_le32(__le32_to_cpu(super->feature_map) | |
1642 | & ~MD_FEATURE_BITMAP_OFFSET); | |
1643 | return 0; | |
1644 | } | |
1645 | ||
1646 | ||
1647 | static struct supertype *match_metadata_desc1(char *arg) | |
1648 | { | |
1649 | struct supertype *st = xcalloc(1, sizeof(*st)); | |
1650 | ||
1651 | st->container_dev = NoMdDev; | |
1652 | st->ss = &super1; | |
1653 | st->max_devs = MAX_DEVS; | |
1654 | st->sb = NULL; | |
1655 | /* leading zeros can be safely ignored. --detail generates them. */ | |
1656 | while (*arg == '0') | |
1657 | arg++; | |
1658 | if (strcmp(arg, "1.0") == 0 || | |
1659 | strcmp(arg, "1.00") == 0) { | |
1660 | st->minor_version = 0; | |
1661 | return st; | |
1662 | } | |
1663 | if (strcmp(arg, "1.1") == 0 || | |
1664 | strcmp(arg, "1.01") == 0 | |
1665 | ) { | |
1666 | st->minor_version = 1; | |
1667 | return st; | |
1668 | } | |
1669 | if (strcmp(arg, "1.2") == 0 || | |
1670 | #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */ | |
1671 | strcmp(arg, "default") == 0 || | |
1672 | #endif /* DEFAULT_OLD_METADATA */ | |
1673 | strcmp(arg, "1.02") == 0) { | |
1674 | st->minor_version = 2; | |
1675 | return st; | |
1676 | } | |
1677 | if (strcmp(arg, "1") == 0 || | |
1678 | strcmp(arg, "default") == 0) { | |
1679 | st->minor_version = -1; | |
1680 | return st; | |
1681 | } | |
1682 | ||
1683 | free(st); | |
1684 | return NULL; | |
1685 | } | |
1686 | ||
1687 | /* find available size on device with this devsize, using | |
1688 | * superblock type st, and reserving 'reserve' sectors for | |
1689 | * a possible bitmap | |
1690 | */ | |
1691 | static __u64 _avail_size1(struct supertype *st, __u64 devsize, | |
1692 | unsigned long long data_offset, int chunksize) | |
1693 | { | |
1694 | struct mdp_superblock_1 *super = st->sb; | |
1695 | int bmspace = 0; | |
1696 | if (devsize < 24) | |
1697 | return 0; | |
1698 | ||
1699 | if (super == NULL) | |
1700 | /* creating: allow suitable space for bitmap */ | |
1701 | bmspace = choose_bm_space(devsize); | |
1702 | #ifndef MDASSEMBLE | |
1703 | else if (__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) { | |
1704 | /* hot-add. allow for actual size of bitmap */ | |
1705 | struct bitmap_super_s *bsb; | |
1706 | bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); | |
1707 | bmspace = bitmap_sectors(bsb); | |
1708 | } | |
1709 | #endif | |
1710 | /* Allow space for bad block log */ | |
1711 | if (super && super->bblog_size) | |
1712 | devsize -= __le16_to_cpu(super->bblog_size); | |
1713 | else | |
1714 | devsize -= 8; | |
1715 | ||
1716 | ||
1717 | if (st->minor_version < 0) | |
1718 | /* not specified, so time to set default */ | |
1719 | st->minor_version = 2; | |
1720 | ||
1721 | if (data_offset != INVALID_SECTORS) | |
1722 | switch(st->minor_version) { | |
1723 | case 0: | |
1724 | return devsize - data_offset - 8*2; | |
1725 | case 1: | |
1726 | case 2: | |
1727 | return devsize - data_offset; | |
1728 | default: | |
1729 | return 0; | |
1730 | } | |
1731 | ||
1732 | devsize -= bmspace; | |
1733 | ||
1734 | if (super == NULL && st->minor_version > 0) { | |
1735 | /* haven't committed to a size yet, so allow some | |
1736 | * slack for space for reshape. | |
1737 | * Limit slack to 128M, but aim for about 0.1% | |
1738 | */ | |
1739 | unsigned long long headroom = 128*1024*2; | |
1740 | while ((headroom << 10) > devsize && | |
1741 | (chunksize == 0 || | |
1742 | headroom / 2 >= ((unsigned)chunksize*2)*2)) | |
1743 | headroom >>= 1; | |
1744 | devsize -= headroom; | |
1745 | } | |
1746 | switch(st->minor_version) { | |
1747 | case 0: | |
1748 | /* at end */ | |
1749 | return ((devsize - 8*2 ) & ~(4*2-1)); | |
1750 | case 1: | |
1751 | /* at start, 4K for superblock and possible bitmap */ | |
1752 | return devsize - 4*2; | |
1753 | case 2: | |
1754 | /* 4k from start, 4K for superblock and possible bitmap */ | |
1755 | return devsize - (4+4)*2; | |
1756 | } | |
1757 | return 0; | |
1758 | } | |
1759 | static __u64 avail_size1(struct supertype *st, __u64 devsize, | |
1760 | unsigned long long data_offset) | |
1761 | { | |
1762 | return _avail_size1(st, devsize, data_offset, 0); | |
1763 | } | |
1764 | ||
1765 | static int | |
1766 | add_internal_bitmap1(struct supertype *st, | |
1767 | int *chunkp, int delay, int write_behind, | |
1768 | unsigned long long size, | |
1769 | int may_change, int major) | |
1770 | { | |
1771 | /* | |
1772 | * If not may_change, then this is a 'Grow' without sysfs support for | |
1773 | * bitmaps, and the bitmap must fit after the superblock at 1K offset. | |
1774 | * If may_change, then this is create or a Grow with sysfs syupport, | |
1775 | * and we can put the bitmap wherever we like. | |
1776 | * | |
1777 | * size is in sectors, chunk is in bytes !!! | |
1778 | */ | |
1779 | ||
1780 | unsigned long long bits; | |
1781 | unsigned long long max_bits; | |
1782 | unsigned long long min_chunk; | |
1783 | long offset; | |
1784 | long bbl_offset, bbl_size; | |
1785 | unsigned long long chunk = *chunkp; | |
1786 | int room = 0; | |
1787 | int creating = 0; | |
1788 | struct mdp_superblock_1 *sb = st->sb; | |
1789 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); | |
1790 | int uuid[4]; | |
1791 | ||
1792 | ||
1793 | if (__le64_to_cpu(sb->data_size) == 0) | |
1794 | /* Must be creating the array, else data_size would be non-zero */ | |
1795 | creating = 1; | |
1796 | switch(st->minor_version) { | |
1797 | case 0: | |
1798 | /* either 3K after the superblock (when hot-add), | |
1799 | * or some amount of space before. | |
1800 | */ | |
1801 | if (creating) { | |
1802 | /* We are creating array, so we *know* how much room has | |
1803 | * been left. | |
1804 | */ | |
1805 | offset = 0; | |
1806 | room = choose_bm_space(__le64_to_cpu(sb->size)); | |
1807 | bbl_size = 8; | |
1808 | } else { | |
1809 | room = __le64_to_cpu(sb->super_offset) | |
1810 | - __le64_to_cpu(sb->data_offset) | |
1811 | - __le64_to_cpu(sb->data_size); | |
1812 | bbl_size = __le16_to_cpu(sb->bblog_size); | |
1813 | if (bbl_size < 8) | |
1814 | bbl_size = 8; | |
1815 | bbl_offset = (__s32)__le32_to_cpu(sb->bblog_offset); | |
1816 | if (bbl_size < -bbl_offset) | |
1817 | bbl_size = -bbl_offset; | |
1818 | ||
1819 | if (!may_change || (room < 3*2 && | |
1820 | __le32_to_cpu(sb->max_dev) <= 384)) { | |
1821 | room = 3*2; | |
1822 | offset = 1*2; | |
1823 | bbl_size = 0; | |
1824 | } else { | |
1825 | offset = 0; /* means movable offset */ | |
1826 | } | |
1827 | } | |
1828 | break; | |
1829 | case 1: | |
1830 | case 2: /* between superblock and data */ | |
1831 | if (creating) { | |
1832 | offset = 4*2; | |
1833 | room = choose_bm_space(__le64_to_cpu(sb->size)); | |
1834 | bbl_size = 8; | |
1835 | } else { | |
1836 | room = __le64_to_cpu(sb->data_offset) | |
1837 | - __le64_to_cpu(sb->super_offset); | |
1838 | bbl_size = __le16_to_cpu(sb->bblog_size); | |
1839 | if (bbl_size) | |
1840 | room = __le32_to_cpu(sb->bblog_offset) + bbl_size; | |
1841 | else | |
1842 | bbl_size = 8; | |
1843 | ||
1844 | if (!may_change) { | |
1845 | room -= 2; /* Leave 1K for superblock */ | |
1846 | offset = 2; | |
1847 | bbl_size = 0; | |
1848 | } else { | |
1849 | room -= 4*2; /* leave 4K for superblock */ | |
1850 | offset = 4*2; | |
1851 | } | |
1852 | } | |
1853 | break; | |
1854 | default: | |
1855 | return 0; | |
1856 | } | |
1857 | ||
1858 | room -= bbl_size; | |
1859 | if (chunk == UnSet && room > 128*2) | |
1860 | /* Limit to 128K of bitmap when chunk size not requested */ | |
1861 | room = 128*2; | |
1862 | ||
1863 | max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8; | |
1864 | ||
1865 | min_chunk = 4096; /* sub-page chunks don't work yet.. */ | |
1866 | bits = (size*512)/min_chunk +1; | |
1867 | while (bits > max_bits) { | |
1868 | min_chunk *= 2; | |
1869 | bits = (bits+1)/2; | |
1870 | } | |
1871 | if (chunk == UnSet) { | |
1872 | /* For practical purpose, 64Meg is a good | |
1873 | * default chunk size for internal bitmaps. | |
1874 | */ | |
1875 | chunk = min_chunk; | |
1876 | if (chunk < 64*1024*1024) | |
1877 | chunk = 64*1024*1024; | |
1878 | } else if (chunk < min_chunk) | |
1879 | return 0; /* chunk size too small */ | |
1880 | if (chunk == 0) /* rounding problem */ | |
1881 | return 0; | |
1882 | ||
1883 | if (offset == 0) { | |
1884 | /* start bitmap on a 4K boundary with enough space for | |
1885 | * the bitmap | |
1886 | */ | |
1887 | bits = (size*512) / chunk + 1; | |
1888 | room = ((bits+7)/8 + sizeof(bitmap_super_t) +4095)/4096; | |
1889 | room *= 8; /* convert 4K blocks to sectors */ | |
1890 | offset = -room - bbl_size; | |
1891 | } | |
1892 | ||
1893 | sb->bitmap_offset = (int32_t)__cpu_to_le32(offset); | |
1894 | ||
1895 | sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | |
1896 | | MD_FEATURE_BITMAP_OFFSET); | |
1897 | memset(bms, 0, sizeof(*bms)); | |
1898 | bms->magic = __cpu_to_le32(BITMAP_MAGIC); | |
1899 | bms->version = __cpu_to_le32(major); | |
1900 | uuid_from_super1(st, uuid); | |
1901 | memcpy(bms->uuid, uuid, 16); | |
1902 | bms->chunksize = __cpu_to_le32(chunk); | |
1903 | bms->daemon_sleep = __cpu_to_le32(delay); | |
1904 | bms->sync_size = __cpu_to_le64(size); | |
1905 | bms->write_behind = __cpu_to_le32(write_behind); | |
1906 | ||
1907 | *chunkp = chunk; | |
1908 | return 1; | |
1909 | } | |
1910 | ||
1911 | static void locate_bitmap1(struct supertype *st, int fd) | |
1912 | { | |
1913 | unsigned long long offset; | |
1914 | struct mdp_superblock_1 *sb; | |
1915 | int mustfree = 0; | |
1916 | ||
1917 | if (!st->sb) { | |
1918 | if (st->ss->load_super(st, fd, NULL)) | |
1919 | return; /* no error I hope... */ | |
1920 | mustfree = 1; | |
1921 | } | |
1922 | sb = st->sb; | |
1923 | ||
1924 | offset = __le64_to_cpu(sb->super_offset); | |
1925 | offset += (int32_t) __le32_to_cpu(sb->bitmap_offset); | |
1926 | if (mustfree) | |
1927 | free(sb); | |
1928 | lseek64(fd, offset<<9, 0); | |
1929 | } | |
1930 | ||
1931 | static int write_bitmap1(struct supertype *st, int fd) | |
1932 | { | |
1933 | struct mdp_superblock_1 *sb = st->sb; | |
1934 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+MAX_SB_SIZE); | |
1935 | int rv = 0; | |
1936 | void *buf; | |
1937 | int towrite, n; | |
1938 | struct align_fd afd; | |
1939 | ||
1940 | init_afd(&afd, fd); | |
1941 | ||
1942 | locate_bitmap1(st, fd); | |
1943 | ||
1944 | if (posix_memalign(&buf, 4096, 4096)) | |
1945 | return -ENOMEM; | |
1946 | ||
1947 | memset(buf, 0xff, 4096); | |
1948 | memcpy(buf, (char *)bms, sizeof(bitmap_super_t)); | |
1949 | ||
1950 | towrite = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9); | |
1951 | towrite = (towrite+7) >> 3; /* bits to bytes */ | |
1952 | towrite += sizeof(bitmap_super_t); | |
1953 | towrite = ROUND_UP(towrite, 512); | |
1954 | while (towrite > 0) { | |
1955 | n = towrite; | |
1956 | if (n > 4096) | |
1957 | n = 4096; | |
1958 | n = awrite(&afd, buf, n); | |
1959 | if (n > 0) | |
1960 | towrite -= n; | |
1961 | else | |
1962 | break; | |
1963 | memset(buf, 0xff, 4096); | |
1964 | } | |
1965 | fsync(fd); | |
1966 | if (towrite) | |
1967 | rv = -2; | |
1968 | ||
1969 | free(buf); | |
1970 | return rv; | |
1971 | } | |
1972 | ||
1973 | static void free_super1(struct supertype *st) | |
1974 | { | |
1975 | if (st->sb) | |
1976 | free(st->sb); | |
1977 | while (st->info) { | |
1978 | struct devinfo *di = st->info; | |
1979 | st->info = di->next; | |
1980 | if (di->fd >= 0) | |
1981 | close(di->fd); | |
1982 | free(di); | |
1983 | } | |
1984 | st->sb = NULL; | |
1985 | } | |
1986 | ||
1987 | #ifndef MDASSEMBLE | |
1988 | static int validate_geometry1(struct supertype *st, int level, | |
1989 | int layout, int raiddisks, | |
1990 | int *chunk, unsigned long long size, | |
1991 | unsigned long long data_offset, | |
1992 | char *subdev, unsigned long long *freesize, | |
1993 | int verbose) | |
1994 | { | |
1995 | unsigned long long ldsize; | |
1996 | int fd; | |
1997 | ||
1998 | if (level == LEVEL_CONTAINER) { | |
1999 | if (verbose) | |
2000 | pr_err("1.x metadata does not support containers\n"); | |
2001 | return 0; | |
2002 | } | |
2003 | if (chunk && *chunk == UnSet) | |
2004 | *chunk = DEFAULT_CHUNK; | |
2005 | ||
2006 | if (!subdev) | |
2007 | return 1; | |
2008 | ||
2009 | fd = open(subdev, O_RDONLY|O_EXCL, 0); | |
2010 | if (fd < 0) { | |
2011 | if (verbose) | |
2012 | pr_err("super1.x cannot open %s: %s\n", | |
2013 | subdev, strerror(errno)); | |
2014 | return 0; | |
2015 | } | |
2016 | ||
2017 | if (!get_dev_size(fd, subdev, &ldsize)) { | |
2018 | close(fd); | |
2019 | return 0; | |
2020 | } | |
2021 | close(fd); | |
2022 | ||
2023 | *freesize = _avail_size1(st, ldsize >> 9, data_offset, *chunk); | |
2024 | return 1; | |
2025 | } | |
2026 | #endif /* MDASSEMBLE */ | |
2027 | ||
2028 | struct superswitch super1 = { | |
2029 | #ifndef MDASSEMBLE | |
2030 | .examine_super = examine_super1, | |
2031 | .brief_examine_super = brief_examine_super1, | |
2032 | .export_examine_super = export_examine_super1, | |
2033 | .detail_super = detail_super1, | |
2034 | .brief_detail_super = brief_detail_super1, | |
2035 | .export_detail_super = export_detail_super1, | |
2036 | .write_init_super = write_init_super1, | |
2037 | .validate_geometry = validate_geometry1, | |
2038 | .add_to_super = add_to_super1, | |
2039 | #endif | |
2040 | .match_home = match_home1, | |
2041 | .uuid_from_super = uuid_from_super1, | |
2042 | .getinfo_super = getinfo_super1, | |
2043 | .container_content = container_content1, | |
2044 | .update_super = update_super1, | |
2045 | .init_super = init_super1, | |
2046 | .store_super = store_super1, | |
2047 | .compare_super = compare_super1, | |
2048 | .load_super = load_super1, | |
2049 | .match_metadata_desc = match_metadata_desc1, | |
2050 | .avail_size = avail_size1, | |
2051 | .add_internal_bitmap = add_internal_bitmap1, | |
2052 | .locate_bitmap = locate_bitmap1, | |
2053 | .write_bitmap = write_bitmap1, | |
2054 | .free_super = free_super1, | |
2055 | #if __BYTE_ORDER == BIG_ENDIAN | |
2056 | .swapuuid = 0, | |
2057 | #else | |
2058 | .swapuuid = 1, | |
2059 | #endif | |
2060 | .name = "1.x", | |
2061 | }; |