]>
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 | __u8 pad1[128-124]; /* set to 0 when written */ | |
62 | ||
63 | /* constant this-device information - 64 bytes */ | |
64 | __u64 data_offset; /* sector start of data, often 0 */ | |
65 | __u64 data_size; /* sectors in this device that can be used for data */ | |
66 | __u64 super_offset; /* sector start of this superblock */ | |
67 | __u64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */ | |
68 | __u32 dev_number; /* permanent identifier of this device - not role in raid */ | |
69 | __u32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */ | |
70 | __u8 device_uuid[16]; /* user-space setable, ignored by kernel */ | |
71 | __u8 devflags; /* per-device flags. Only one defined...*/ | |
72 | #define WriteMostly1 1 /* mask for writemostly flag in above */ | |
73 | /* bad block log. If there are any bad blocks the feature flag is set. | |
74 | * if offset and size are non-zero, that space is reserved and available. | |
75 | */ | |
76 | __u8 bblog_shift; /* shift from sectors to block size for badblocklist */ | |
77 | __u16 bblog_size; /* number of sectors reserved for badblocklist */ | |
78 | __u32 bblog_offset; /* sector offset from superblock to bblog, signed */ | |
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 dev_roles[max_dev] */ | |
85 | __u32 max_dev; /* size of dev_roles[] 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 | #define MAX_SB_SIZE 4096 | |
98 | /* bitmap super size is 256, but we round up to a sector for alignment */ | |
99 | #define BM_SUPER_SIZE 512 | |
100 | #define MAX_DEVS ((int)(MAX_SB_SIZE - sizeof(struct mdp_superblock_1)) / 2) | |
101 | #define SUPER1_SIZE (MAX_SB_SIZE + BM_SUPER_SIZE \ | |
102 | + sizeof(struct misc_dev_info)) | |
103 | ||
104 | struct misc_dev_info { | |
105 | __u64 device_size; | |
106 | }; | |
107 | ||
108 | /* feature_map bits */ | |
109 | #define MD_FEATURE_BITMAP_OFFSET 1 | |
110 | #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and | |
111 | * must be honoured | |
112 | */ | |
113 | #define MD_FEATURE_RESHAPE_ACTIVE 4 | |
114 | #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */ | |
115 | ||
116 | #define MD_FEATURE_ALL (1|2|4|8) | |
117 | ||
118 | #ifndef offsetof | |
119 | #define offsetof(t,f) ((size_t)&(((t*)0)->f)) | |
120 | #endif | |
121 | static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb) | |
122 | { | |
123 | unsigned int disk_csum, csum; | |
124 | unsigned long long newcsum; | |
125 | int size = sizeof(*sb) + __le32_to_cpu(sb->max_dev)*2; | |
126 | unsigned int *isuper = (unsigned int*)sb; | |
127 | ||
128 | /* make sure I can count... */ | |
129 | if (offsetof(struct mdp_superblock_1,data_offset) != 128 || | |
130 | offsetof(struct mdp_superblock_1, utime) != 192 || | |
131 | sizeof(struct mdp_superblock_1) != 256) { | |
132 | fprintf(stderr, "WARNING - superblock isn't sized correctly\n"); | |
133 | } | |
134 | ||
135 | disk_csum = sb->sb_csum; | |
136 | sb->sb_csum = 0; | |
137 | newcsum = 0; | |
138 | for (; size>=4; size -= 4 ) { | |
139 | newcsum += __le32_to_cpu(*isuper); | |
140 | isuper++; | |
141 | } | |
142 | ||
143 | if (size == 2) | |
144 | newcsum += __le16_to_cpu(*(unsigned short*) isuper); | |
145 | ||
146 | csum = (newcsum & 0xffffffff) + (newcsum >> 32); | |
147 | sb->sb_csum = disk_csum; | |
148 | return __cpu_to_le32(csum); | |
149 | } | |
150 | ||
151 | /* | |
152 | * Information related to file descriptor used for aligned reads/writes. | |
153 | * Cache the block size. | |
154 | */ | |
155 | struct align_fd { | |
156 | int fd; | |
157 | int blk_sz; | |
158 | }; | |
159 | ||
160 | static void init_afd(struct align_fd *afd, int fd) | |
161 | { | |
162 | afd->fd = fd; | |
163 | ||
164 | if (ioctl(afd->fd, BLKSSZGET, &afd->blk_sz) != 0) | |
165 | afd->blk_sz = 512; | |
166 | } | |
167 | ||
168 | static char abuf[4096+4096]; | |
169 | static int aread(struct align_fd *afd, void *buf, int len) | |
170 | { | |
171 | /* aligned read. | |
172 | * On devices with a 4K sector size, we need to read | |
173 | * the full sector and copy relevant bits into | |
174 | * the buffer | |
175 | */ | |
176 | int bsize, iosize; | |
177 | char *b; | |
178 | int n; | |
179 | ||
180 | bsize = afd->blk_sz; | |
181 | ||
182 | if (!bsize || bsize > 4096 || len > 4096) { | |
183 | if (!bsize) | |
184 | fprintf(stderr, "WARNING - aread() called with " | |
185 | "invalid block size\n"); | |
186 | return -1; | |
187 | } | |
188 | b = ROUND_UP_PTR((char *)abuf, 4096); | |
189 | ||
190 | for (iosize = 0; iosize < len; iosize += bsize) | |
191 | ; | |
192 | n = read(afd->fd, b, iosize); | |
193 | if (n <= 0) | |
194 | return n; | |
195 | lseek(afd->fd, len - n, 1); | |
196 | if (n > len) | |
197 | n = len; | |
198 | memcpy(buf, b, n); | |
199 | return n; | |
200 | } | |
201 | ||
202 | static int awrite(struct align_fd *afd, void *buf, int len) | |
203 | { | |
204 | /* aligned write. | |
205 | * On devices with a 4K sector size, we need to write | |
206 | * the full sector. We pre-read if the sector is larger | |
207 | * than the write. | |
208 | * The address must be sector-aligned. | |
209 | */ | |
210 | int bsize, iosize; | |
211 | char *b; | |
212 | int n; | |
213 | ||
214 | bsize = afd->blk_sz; | |
215 | if (!bsize || bsize > 4096 || len > 4096) { | |
216 | if (!bsize) | |
217 | fprintf(stderr, "WARNING - awrite() called with " | |
218 | "invalid block size\n"); | |
219 | return -1; | |
220 | } | |
221 | b = ROUND_UP_PTR((char *)abuf, 4096); | |
222 | ||
223 | for (iosize = 0; iosize < len ; iosize += bsize) | |
224 | ; | |
225 | ||
226 | if (len != iosize) { | |
227 | n = read(afd->fd, b, iosize); | |
228 | if (n <= 0) | |
229 | return n; | |
230 | lseek(afd->fd, -n, 1); | |
231 | } | |
232 | ||
233 | memcpy(b, buf, len); | |
234 | n = write(afd->fd, b, iosize); | |
235 | if (n <= 0) | |
236 | return n; | |
237 | lseek(afd->fd, len - n, 1); | |
238 | return len; | |
239 | } | |
240 | ||
241 | #ifndef MDASSEMBLE | |
242 | static void examine_super1(struct supertype *st, char *homehost) | |
243 | { | |
244 | struct mdp_superblock_1 *sb = st->sb; | |
245 | time_t atime; | |
246 | unsigned int d; | |
247 | int role; | |
248 | int delta_extra = 0; | |
249 | int i; | |
250 | char *c; | |
251 | int l = homehost ? strlen(homehost) : 0; | |
252 | int layout; | |
253 | unsigned long long sb_offset; | |
254 | ||
255 | printf(" Magic : %08x\n", __le32_to_cpu(sb->magic)); | |
256 | printf(" Version : 1"); | |
257 | sb_offset = __le64_to_cpu(sb->super_offset); | |
258 | if (sb_offset <= 4) | |
259 | printf(".1\n"); | |
260 | else if (sb_offset <= 8) | |
261 | printf(".2\n"); | |
262 | else | |
263 | printf(".0\n"); | |
264 | printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb->feature_map)); | |
265 | printf(" Array UUID : "); | |
266 | for (i=0; i<16; i++) { | |
267 | if ((i&3)==0 && i != 0) printf(":"); | |
268 | printf("%02x", sb->set_uuid[i]); | |
269 | } | |
270 | printf("\n"); | |
271 | printf(" Name : %.32s", sb->set_name); | |
272 | if (l > 0 && l < 32 && | |
273 | sb->set_name[l] == ':' && | |
274 | strncmp(sb->set_name, homehost, l) == 0) | |
275 | printf(" (local to host %s)", homehost); | |
276 | printf("\n"); | |
277 | atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL; | |
278 | printf(" Creation Time : %.24s\n", ctime(&atime)); | |
279 | c=map_num(pers, __le32_to_cpu(sb->level)); | |
280 | printf(" Raid Level : %s\n", c?c:"-unknown-"); | |
281 | printf(" Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks)); | |
282 | printf("\n"); | |
283 | printf(" Avail Dev Size : %llu%s\n", | |
284 | (unsigned long long)__le64_to_cpu(sb->data_size), | |
285 | human_size(__le64_to_cpu(sb->data_size)<<9)); | |
286 | if (__le32_to_cpu(sb->level) > 0) { | |
287 | int ddsks = 0, ddsks_denom = 1; | |
288 | switch(__le32_to_cpu(sb->level)) { | |
289 | case 1: ddsks=1;break; | |
290 | case 4: | |
291 | case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break; | |
292 | case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break; | |
293 | case 10: | |
294 | layout = __le32_to_cpu(sb->layout); | |
295 | ddsks = __le32_to_cpu(sb->raid_disks); | |
296 | ddsks_denom = (layout&255) * ((layout>>8)&255); | |
297 | } | |
298 | if (ddsks) { | |
299 | long long asize = __le64_to_cpu(sb->size); | |
300 | asize = (asize << 9) * ddsks / ddsks_denom; | |
301 | printf(" Array Size : %llu%s\n", | |
302 | asize >> 10, human_size(asize)); | |
303 | } | |
304 | if (sb->size != sb->data_size) | |
305 | printf(" Used Dev Size : %llu%s\n", | |
306 | (unsigned long long)__le64_to_cpu(sb->size), | |
307 | human_size(__le64_to_cpu(sb->size)<<9)); | |
308 | } | |
309 | if (sb->data_offset) | |
310 | printf(" Data Offset : %llu sectors\n", | |
311 | (unsigned long long)__le64_to_cpu(sb->data_offset)); | |
312 | printf(" Super Offset : %llu sectors\n", | |
313 | (unsigned long long)__le64_to_cpu(sb->super_offset)); | |
314 | if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET) | |
315 | printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->recovery_offset)); | |
316 | printf(" State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean"); | |
317 | printf(" Device UUID : "); | |
318 | for (i=0; i<16; i++) { | |
319 | if ((i&3)==0 && i != 0) printf(":"); | |
320 | printf("%02x", sb->device_uuid[i]); | |
321 | } | |
322 | printf("\n"); | |
323 | printf("\n"); | |
324 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
325 | printf("Internal Bitmap : %ld sectors from superblock\n", | |
326 | (long)(int32_t)__le32_to_cpu(sb->bitmap_offset)); | |
327 | } | |
328 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE)) { | |
329 | printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb->reshape_position)/2, | |
330 | human_size(__le64_to_cpu(sb->reshape_position)<<9)); | |
331 | if (__le32_to_cpu(sb->delta_disks)) { | |
332 | printf(" Delta Devices : %d", __le32_to_cpu(sb->delta_disks)); | |
333 | printf(" (%d->%d)\n", | |
334 | __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks), | |
335 | __le32_to_cpu(sb->raid_disks)); | |
336 | if ((int)__le32_to_cpu(sb->delta_disks) < 0) | |
337 | delta_extra = -__le32_to_cpu(sb->delta_disks); | |
338 | } | |
339 | if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) { | |
340 | c = map_num(pers, __le32_to_cpu(sb->new_level)); | |
341 | printf(" New Level : %s\n", c?c:"-unknown-"); | |
342 | } | |
343 | if (__le32_to_cpu(sb->new_layout) != __le32_to_cpu(sb->layout)) { | |
344 | if (__le32_to_cpu(sb->level) == 5) { | |
345 | c = map_num(r5layout, __le32_to_cpu(sb->new_layout)); | |
346 | printf(" New Layout : %s\n", c?c:"-unknown-"); | |
347 | } | |
348 | if (__le32_to_cpu(sb->level) == 6) { | |
349 | c = map_num(r6layout, __le32_to_cpu(sb->new_layout)); | |
350 | printf(" New Layout : %s\n", c?c:"-unknown-"); | |
351 | } | |
352 | if (__le32_to_cpu(sb->level) == 10) { | |
353 | printf(" New Layout :"); | |
354 | print_r10_layout(__le32_to_cpu(sb->new_layout)); | |
355 | printf("\n"); | |
356 | } | |
357 | } | |
358 | if (__le32_to_cpu(sb->new_chunk) != __le32_to_cpu(sb->chunksize)) | |
359 | printf(" New Chunksize : %dK\n", __le32_to_cpu(sb->new_chunk)/2); | |
360 | printf("\n"); | |
361 | } | |
362 | if (sb->devflags) { | |
363 | printf(" Flags :"); | |
364 | if (sb->devflags & WriteMostly1) | |
365 | printf(" write-mostly"); | |
366 | printf("\n"); | |
367 | } | |
368 | ||
369 | atime = __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL; | |
370 | printf(" Update Time : %.24s\n", ctime(&atime)); | |
371 | ||
372 | if (sb->bblog_size && sb->bblog_offset) { | |
373 | printf(" Bad Block Log : %d entries available at offset %ld sectors", | |
374 | __le16_to_cpu(sb->bblog_size)*512/8, | |
375 | (long)__le32_to_cpu(sb->bblog_offset)); | |
376 | if (sb->feature_map & | |
377 | __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
378 | printf(" - bad blocks present."); | |
379 | printf("\n"); | |
380 | } | |
381 | ||
382 | ||
383 | if (calc_sb_1_csum(sb) == sb->sb_csum) | |
384 | printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum)); | |
385 | else | |
386 | printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb->sb_csum), | |
387 | __le32_to_cpu(calc_sb_1_csum(sb))); | |
388 | printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb->events)); | |
389 | printf("\n"); | |
390 | if (__le32_to_cpu(sb->level) == 5) { | |
391 | c = map_num(r5layout, __le32_to_cpu(sb->layout)); | |
392 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
393 | } | |
394 | if (__le32_to_cpu(sb->level) == 6) { | |
395 | c = map_num(r6layout, __le32_to_cpu(sb->layout)); | |
396 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
397 | } | |
398 | if (__le32_to_cpu(sb->level) == 10) { | |
399 | int lo = __le32_to_cpu(sb->layout); | |
400 | printf(" Layout :"); | |
401 | print_r10_layout(lo); | |
402 | printf("\n"); | |
403 | } | |
404 | switch(__le32_to_cpu(sb->level)) { | |
405 | case 0: | |
406 | case 4: | |
407 | case 5: | |
408 | case 6: | |
409 | case 10: | |
410 | printf(" Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize)/2); | |
411 | break; | |
412 | case -1: | |
413 | printf(" Rounding : %dK\n", __le32_to_cpu(sb->chunksize)/2); | |
414 | break; | |
415 | default: break; | |
416 | } | |
417 | printf("\n"); | |
418 | #if 0 | |
419 | /* This turns out to just be confusing */ | |
420 | printf(" Array Slot : %d (", __le32_to_cpu(sb->dev_number)); | |
421 | for (i= __le32_to_cpu(sb->max_dev); i> 0 ; i--) | |
422 | if (__le16_to_cpu(sb->dev_roles[i-1]) != 0xffff) | |
423 | break; | |
424 | for (d=0; d < i; d++) { | |
425 | int role = __le16_to_cpu(sb->dev_roles[d]); | |
426 | if (d) printf(", "); | |
427 | if (role == 0xffff) printf("empty"); | |
428 | else if(role == 0xfffe) printf("failed"); | |
429 | else printf("%d", role); | |
430 | } | |
431 | printf(")\n"); | |
432 | #endif | |
433 | printf(" Device Role : "); | |
434 | d = __le32_to_cpu(sb->dev_number); | |
435 | if (d < __le32_to_cpu(sb->max_dev)) | |
436 | role = __le16_to_cpu(sb->dev_roles[d]); | |
437 | else | |
438 | role = 0xFFFF; | |
439 | if (role >= 0xFFFE) | |
440 | printf("spare\n"); | |
441 | else | |
442 | printf("Active device %d\n", role); | |
443 | ||
444 | printf(" Array State : "); | |
445 | for (d=0; d<__le32_to_cpu(sb->raid_disks) + delta_extra; d++) { | |
446 | int cnt = 0; | |
447 | unsigned int i; | |
448 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { | |
449 | unsigned int role = __le16_to_cpu(sb->dev_roles[i]); | |
450 | if (role == d) | |
451 | cnt++; | |
452 | } | |
453 | if (cnt > 1) printf("?"); | |
454 | else if (cnt == 1) printf("A"); | |
455 | else printf ("."); | |
456 | } | |
457 | #if 0 | |
458 | /* This is confusing too */ | |
459 | faulty = 0; | |
460 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { | |
461 | int role = __le16_to_cpu(sb->dev_roles[i]); | |
462 | if (role == 0xFFFE) | |
463 | faulty++; | |
464 | } | |
465 | if (faulty) printf(" %d failed", faulty); | |
466 | #endif | |
467 | printf(" ('A' == active, '.' == missing)"); | |
468 | printf("\n"); | |
469 | } | |
470 | ||
471 | ||
472 | static void brief_examine_super1(struct supertype *st, int verbose) | |
473 | { | |
474 | struct mdp_superblock_1 *sb = st->sb; | |
475 | int i; | |
476 | unsigned long long sb_offset; | |
477 | char *nm; | |
478 | char *c=map_num(pers, __le32_to_cpu(sb->level)); | |
479 | ||
480 | nm = strchr(sb->set_name, ':'); | |
481 | if (nm) | |
482 | nm++; | |
483 | else if (sb->set_name[0]) | |
484 | nm = sb->set_name; | |
485 | else | |
486 | nm = NULL; | |
487 | ||
488 | printf("ARRAY "); | |
489 | if (nm) { | |
490 | printf("/dev/md/"); | |
491 | print_escape(nm); | |
492 | putchar(' '); | |
493 | } | |
494 | if (verbose && c) | |
495 | printf(" level=%s", c); | |
496 | sb_offset = __le64_to_cpu(sb->super_offset); | |
497 | if (sb_offset <= 4) | |
498 | printf(" metadata=1.1 "); | |
499 | else if (sb_offset <= 8) | |
500 | printf(" metadata=1.2 "); | |
501 | else | |
502 | printf(" metadata=1.0 "); | |
503 | if (verbose) | |
504 | printf("num-devices=%d ", __le32_to_cpu(sb->raid_disks)); | |
505 | printf("UUID="); | |
506 | for (i=0; i<16; i++) { | |
507 | if ((i&3)==0 && i != 0) printf(":"); | |
508 | printf("%02x", sb->set_uuid[i]); | |
509 | } | |
510 | if (sb->set_name[0]) { | |
511 | printf(" name="); | |
512 | print_quoted(sb->set_name); | |
513 | } | |
514 | printf("\n"); | |
515 | } | |
516 | ||
517 | static void export_examine_super1(struct supertype *st) | |
518 | { | |
519 | struct mdp_superblock_1 *sb = st->sb; | |
520 | int i; | |
521 | int len = 32; | |
522 | int layout; | |
523 | ||
524 | printf("MD_LEVEL=%s\n", map_num(pers, __le32_to_cpu(sb->level))); | |
525 | printf("MD_DEVICES=%d\n", __le32_to_cpu(sb->raid_disks)); | |
526 | for (i=0; i<32; i++) | |
527 | if (sb->set_name[i] == '\n' || | |
528 | sb->set_name[i] == '\0') { | |
529 | len = i; | |
530 | break; | |
531 | } | |
532 | if (len) | |
533 | printf("MD_NAME=%.*s\n", len, sb->set_name); | |
534 | if (__le32_to_cpu(sb->level) > 0) { | |
535 | int ddsks = 0, ddsks_denom = 1; | |
536 | switch(__le32_to_cpu(sb->level)) { | |
537 | case 1: ddsks=1;break; | |
538 | case 4: | |
539 | case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break; | |
540 | case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break; | |
541 | case 10: | |
542 | layout = __le32_to_cpu(sb->layout); | |
543 | ddsks = __le32_to_cpu(sb->raid_disks); | |
544 | ddsks_denom = (layout&255) * ((layout>>8)&255); | |
545 | } | |
546 | if (ddsks) { | |
547 | long long asize = __le64_to_cpu(sb->size); | |
548 | asize = (asize << 9) * ddsks / ddsks_denom; | |
549 | printf("MD_ARRAY_SIZE=%s\n",human_size_brief(asize,JEDEC)); | |
550 | } | |
551 | } | |
552 | printf("MD_UUID="); | |
553 | for (i=0; i<16; i++) { | |
554 | if ((i&3)==0 && i != 0) printf(":"); | |
555 | printf("%02x", sb->set_uuid[i]); | |
556 | } | |
557 | printf("\n"); | |
558 | printf("MD_UPDATE_TIME=%llu\n", | |
559 | __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL); | |
560 | printf("MD_DEV_UUID="); | |
561 | for (i=0; i<16; i++) { | |
562 | if ((i&3)==0 && i != 0) printf(":"); | |
563 | printf("%02x", sb->device_uuid[i]); | |
564 | } | |
565 | printf("\n"); | |
566 | printf("MD_EVENTS=%llu\n", | |
567 | (unsigned long long)__le64_to_cpu(sb->events)); | |
568 | } | |
569 | ||
570 | static void detail_super1(struct supertype *st, char *homehost) | |
571 | { | |
572 | struct mdp_superblock_1 *sb = st->sb; | |
573 | int i; | |
574 | int l = homehost ? strlen(homehost) : 0; | |
575 | ||
576 | printf(" Name : %.32s", sb->set_name); | |
577 | if (l > 0 && l < 32 && | |
578 | sb->set_name[l] == ':' && | |
579 | strncmp(sb->set_name, homehost, l) == 0) | |
580 | printf(" (local to host %s)", homehost); | |
581 | printf("\n UUID : "); | |
582 | for (i=0; i<16; i++) { | |
583 | if ((i&3)==0 && i != 0) printf(":"); | |
584 | printf("%02x", sb->set_uuid[i]); | |
585 | } | |
586 | printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb->events)); | |
587 | } | |
588 | ||
589 | static void brief_detail_super1(struct supertype *st) | |
590 | { | |
591 | struct mdp_superblock_1 *sb = st->sb; | |
592 | int i; | |
593 | ||
594 | if (sb->set_name[0]) { | |
595 | printf(" name="); | |
596 | print_quoted(sb->set_name); | |
597 | } | |
598 | printf(" UUID="); | |
599 | for (i=0; i<16; i++) { | |
600 | if ((i&3)==0 && i != 0) printf(":"); | |
601 | printf("%02x", sb->set_uuid[i]); | |
602 | } | |
603 | } | |
604 | ||
605 | static void export_detail_super1(struct supertype *st) | |
606 | { | |
607 | struct mdp_superblock_1 *sb = st->sb; | |
608 | int i; | |
609 | int len = 32; | |
610 | ||
611 | for (i=0; i<32; i++) | |
612 | if (sb->set_name[i] == '\n' || | |
613 | sb->set_name[i] == '\0') { | |
614 | len = i; | |
615 | break; | |
616 | } | |
617 | if (len) | |
618 | printf("MD_NAME=%.*s\n", len, sb->set_name); | |
619 | } | |
620 | ||
621 | #endif | |
622 | ||
623 | static int match_home1(struct supertype *st, char *homehost) | |
624 | { | |
625 | struct mdp_superblock_1 *sb = st->sb; | |
626 | int l = homehost ? strlen(homehost) : 0; | |
627 | ||
628 | return (l > 0 && l < 32 && | |
629 | sb->set_name[l] == ':' && | |
630 | strncmp(sb->set_name, homehost, l) == 0); | |
631 | } | |
632 | ||
633 | static void uuid_from_super1(struct supertype *st, int uuid[4]) | |
634 | { | |
635 | struct mdp_superblock_1 *super = st->sb; | |
636 | char *cuuid = (char*)uuid; | |
637 | int i; | |
638 | for (i=0; i<16; i++) | |
639 | cuuid[i] = super->set_uuid[i]; | |
640 | } | |
641 | ||
642 | static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map) | |
643 | { | |
644 | struct mdp_superblock_1 *sb = st->sb; | |
645 | int working = 0; | |
646 | unsigned int i; | |
647 | unsigned int role; | |
648 | unsigned int map_disks = info->array.raid_disks; | |
649 | ||
650 | memset(info, 0, sizeof(*info)); | |
651 | info->array.major_version = 1; | |
652 | info->array.minor_version = st->minor_version; | |
653 | info->array.patch_version = 0; | |
654 | info->array.raid_disks = __le32_to_cpu(sb->raid_disks); | |
655 | info->array.level = __le32_to_cpu(sb->level); | |
656 | info->array.layout = __le32_to_cpu(sb->layout); | |
657 | info->array.md_minor = -1; | |
658 | info->array.ctime = __le64_to_cpu(sb->ctime); | |
659 | info->array.utime = __le64_to_cpu(sb->utime); | |
660 | info->array.chunk_size = __le32_to_cpu(sb->chunksize)*512; | |
661 | info->array.state = | |
662 | (__le64_to_cpu(sb->resync_offset) == MaxSector) | |
663 | ? 1 : 0; | |
664 | ||
665 | info->data_offset = __le64_to_cpu(sb->data_offset); | |
666 | info->component_size = __le64_to_cpu(sb->size); | |
667 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET)) | |
668 | info->bitmap_offset = (int32_t)__le32_to_cpu(sb->bitmap_offset); | |
669 | ||
670 | info->disk.major = 0; | |
671 | info->disk.minor = 0; | |
672 | info->disk.number = __le32_to_cpu(sb->dev_number); | |
673 | if (__le32_to_cpu(sb->dev_number) >= __le32_to_cpu(sb->max_dev) || | |
674 | __le32_to_cpu(sb->dev_number) >= MAX_DEVS) | |
675 | role = 0xfffe; | |
676 | else | |
677 | role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]); | |
678 | ||
679 | info->disk.raid_disk = -1; | |
680 | switch(role) { | |
681 | case 0xFFFF: | |
682 | info->disk.state = 0; /* spare: not active, not sync, not faulty */ | |
683 | break; | |
684 | case 0xFFFE: | |
685 | info->disk.state = 1; /* faulty */ | |
686 | break; | |
687 | default: | |
688 | info->disk.state = 6; /* active and in sync */ | |
689 | info->disk.raid_disk = role; | |
690 | } | |
691 | if (sb->devflags & WriteMostly1) | |
692 | info->disk.state |= (1 << MD_DISK_WRITEMOSTLY); | |
693 | info->events = __le64_to_cpu(sb->events); | |
694 | sprintf(info->text_version, "1.%d", st->minor_version); | |
695 | info->safe_mode_delay = 200; | |
696 | ||
697 | memcpy(info->uuid, sb->set_uuid, 16); | |
698 | ||
699 | strncpy(info->name, sb->set_name, 32); | |
700 | info->name[32] = 0; | |
701 | ||
702 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET)) | |
703 | info->recovery_start = __le32_to_cpu(sb->recovery_offset); | |
704 | else | |
705 | info->recovery_start = MaxSector; | |
706 | ||
707 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) { | |
708 | info->reshape_active = 1; | |
709 | info->reshape_progress = __le64_to_cpu(sb->reshape_position); | |
710 | info->new_level = __le32_to_cpu(sb->new_level); | |
711 | info->delta_disks = __le32_to_cpu(sb->delta_disks); | |
712 | info->new_layout = __le32_to_cpu(sb->new_layout); | |
713 | info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9; | |
714 | if (info->delta_disks < 0) | |
715 | info->array.raid_disks -= info->delta_disks; | |
716 | } else | |
717 | info->reshape_active = 0; | |
718 | ||
719 | info->recovery_blocked = info->reshape_active; | |
720 | ||
721 | if (map) | |
722 | for (i=0; i<map_disks; i++) | |
723 | map[i] = 0; | |
724 | for (i = 0; i < __le32_to_cpu(sb->max_dev); i++) { | |
725 | role = __le16_to_cpu(sb->dev_roles[i]); | |
726 | if (/*role == 0xFFFF || */role < (unsigned) info->array.raid_disks) { | |
727 | working++; | |
728 | if (map && role < map_disks) | |
729 | map[role] = 1; | |
730 | } | |
731 | } | |
732 | ||
733 | info->array.working_disks = working; | |
734 | } | |
735 | ||
736 | static struct mdinfo *container_content1(struct supertype *st, char *subarray) | |
737 | { | |
738 | struct mdinfo *info; | |
739 | ||
740 | if (subarray) | |
741 | return NULL; | |
742 | ||
743 | info = xmalloc(sizeof(*info)); | |
744 | getinfo_super1(st, info, NULL); | |
745 | return info; | |
746 | } | |
747 | ||
748 | static int update_super1(struct supertype *st, struct mdinfo *info, | |
749 | char *update, | |
750 | char *devname, int verbose, | |
751 | int uuid_set, char *homehost) | |
752 | { | |
753 | /* NOTE: for 'assemble' and 'force' we need to return non-zero | |
754 | * if any change was made. For others, the return value is | |
755 | * ignored. | |
756 | */ | |
757 | int rv = 0; | |
758 | struct mdp_superblock_1 *sb = st->sb; | |
759 | ||
760 | if (strcmp(update, "force-one")==0) { | |
761 | /* Not enough devices for a working array, | |
762 | * so bring this one up-to-date | |
763 | */ | |
764 | if (sb->events != __cpu_to_le64(info->events)) | |
765 | rv = 1; | |
766 | sb->events = __cpu_to_le64(info->events); | |
767 | } else if (strcmp(update, "force-array")==0) { | |
768 | /* Degraded array and 'force' requests to | |
769 | * maybe need to mark it 'clean'. | |
770 | */ | |
771 | switch(__le32_to_cpu(sb->level)) { | |
772 | case 5: case 4: case 6: | |
773 | /* need to force clean */ | |
774 | if (sb->resync_offset != MaxSector) | |
775 | rv = 1; | |
776 | sb->resync_offset = MaxSector; | |
777 | } | |
778 | } else if (strcmp(update, "assemble")==0) { | |
779 | int d = info->disk.number; | |
780 | int want; | |
781 | if (info->disk.state == 6) | |
782 | want = info->disk.raid_disk; | |
783 | else | |
784 | want = 0xFFFF; | |
785 | if (sb->dev_roles[d] != __cpu_to_le16(want)) { | |
786 | sb->dev_roles[d] = __cpu_to_le16(want); | |
787 | rv = 1; | |
788 | } | |
789 | if (info->reshape_active && | |
790 | sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) && | |
791 | info->delta_disks >= 0 && | |
792 | info->reshape_progress < __le64_to_cpu(sb->reshape_position)) { | |
793 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
794 | rv = 1; | |
795 | } | |
796 | if (info->reshape_active && | |
797 | sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) && | |
798 | info->delta_disks < 0 && | |
799 | info->reshape_progress > __le64_to_cpu(sb->reshape_position)) { | |
800 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
801 | rv = 1; | |
802 | } | |
803 | } else if (strcmp(update, "linear-grow-new") == 0) { | |
804 | unsigned int i; | |
805 | int rfd, fd; | |
806 | unsigned int max = __le32_to_cpu(sb->max_dev); | |
807 | ||
808 | for (i=0 ; i < max ; i++) | |
809 | if (__le16_to_cpu(sb->dev_roles[i]) >= 0xfffe) | |
810 | break; | |
811 | sb->dev_number = __cpu_to_le32(i); | |
812 | info->disk.number = i; | |
813 | if (max >= __le32_to_cpu(sb->max_dev)) | |
814 | sb->max_dev = __cpu_to_le32(max+1); | |
815 | ||
816 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
817 | read(rfd, sb->device_uuid, 16) != 16) { | |
818 | __u32 r[4] = {random(), random(), random(), random()}; | |
819 | memcpy(sb->device_uuid, r, 16); | |
820 | } | |
821 | if (rfd >= 0) | |
822 | close(rfd); | |
823 | ||
824 | sb->dev_roles[i] = | |
825 | __cpu_to_le16(info->disk.raid_disk); | |
826 | ||
827 | fd = open(devname, O_RDONLY); | |
828 | if (fd >= 0) { | |
829 | unsigned long long ds; | |
830 | get_dev_size(fd, devname, &ds); | |
831 | close(fd); | |
832 | ds >>= 9; | |
833 | if (__le64_to_cpu(sb->super_offset) < | |
834 | __le64_to_cpu(sb->data_offset)) { | |
835 | sb->data_size = __cpu_to_le64( | |
836 | ds - __le64_to_cpu(sb->data_offset)); | |
837 | } else { | |
838 | ds -= 8*2; | |
839 | ds &= ~(unsigned long long)(4*2-1); | |
840 | sb->super_offset = __cpu_to_le64(ds); | |
841 | sb->data_size = __cpu_to_le64( | |
842 | ds - __le64_to_cpu(sb->data_offset)); | |
843 | } | |
844 | } | |
845 | } else if (strcmp(update, "linear-grow-update") == 0) { | |
846 | sb->raid_disks = __cpu_to_le32(info->array.raid_disks); | |
847 | sb->dev_roles[info->disk.number] = | |
848 | __cpu_to_le16(info->disk.raid_disk); | |
849 | } else if (strcmp(update, "resync") == 0) { | |
850 | /* make sure resync happens */ | |
851 | sb->resync_offset = 0ULL; | |
852 | } else if (strcmp(update, "uuid") == 0) { | |
853 | copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid); | |
854 | ||
855 | if (__le32_to_cpu(sb->feature_map)&MD_FEATURE_BITMAP_OFFSET) { | |
856 | struct bitmap_super_s *bm; | |
857 | bm = (struct bitmap_super_s*)(st->sb+MAX_SB_SIZE); | |
858 | memcpy(bm->uuid, sb->set_uuid, 16); | |
859 | } | |
860 | } else if (strcmp(update, "no-bitmap") == 0) { | |
861 | sb->feature_map &= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET); | |
862 | } else if (strcmp(update, "bbl") == 0) { | |
863 | /* only possible if there is room after the bitmap, or if | |
864 | * there is no bitmap | |
865 | */ | |
866 | unsigned long long sb_offset = __le64_to_cpu(sb->super_offset); | |
867 | unsigned long long data_offset = __le64_to_cpu(sb->data_offset); | |
868 | long bitmap_offset = (long)__le64_to_cpu(sb->bitmap_offset); | |
869 | long bm_sectors = 0; | |
870 | long space; | |
871 | ||
872 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
873 | struct bitmap_super_s *bsb; | |
874 | bsb = (struct bitmap_super_s *)(((char*)sb)+MAX_SB_SIZE); | |
875 | bm_sectors = bitmap_sectors(bsb); | |
876 | } | |
877 | ||
878 | if (sb_offset < data_offset) { | |
879 | /* 1.1 or 1.2. Put bbl just before data | |
880 | */ | |
881 | long bb_offset; | |
882 | space = data_offset - sb_offset; | |
883 | bb_offset = space - 8; | |
884 | if (bm_sectors && bitmap_offset > 0) | |
885 | space -= (bitmap_offset + bm_sectors); | |
886 | else | |
887 | space -= 8; /* The superblock */ | |
888 | if (space >= 8) { | |
889 | sb->bblog_size = __cpu_to_le16(8); | |
890 | sb->bblog_offset = __cpu_to_le32(bb_offset); | |
891 | } | |
892 | } else { | |
893 | /* 1.0 - Put bbl just before super block */ | |
894 | if (bm_sectors && bitmap_offset < 0) | |
895 | space = -bitmap_offset - bm_sectors; | |
896 | else | |
897 | space = sb_offset - data_offset - | |
898 | __le64_to_cpu(sb->data_size); | |
899 | if (space >= 8) { | |
900 | sb->bblog_size = __cpu_to_le16(8); | |
901 | sb->bblog_offset = __cpu_to_le32((unsigned)-8); | |
902 | } | |
903 | } | |
904 | } else if (strcmp(update, "no-bbl") == 0) { | |
905 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
906 | pr_err("Cannot remove active bbl from %s\n",devname); | |
907 | else { | |
908 | sb->bblog_size = 0; | |
909 | sb->bblog_shift = 0; | |
910 | sb->bblog_offset = 0; | |
911 | } | |
912 | } else if (strcmp(update, "homehost") == 0 && | |
913 | homehost) { | |
914 | char *c; | |
915 | update = "name"; | |
916 | c = strchr(sb->set_name, ':'); | |
917 | if (c) | |
918 | strncpy(info->name, c+1, 31 - (c-sb->set_name)); | |
919 | else | |
920 | strncpy(info->name, sb->set_name, 32); | |
921 | info->name[32] = 0; | |
922 | } else if (strcmp(update, "name") == 0) { | |
923 | if (info->name[0] == 0) | |
924 | sprintf(info->name, "%d", info->array.md_minor); | |
925 | memset(sb->set_name, 0, sizeof(sb->set_name)); | |
926 | if (homehost && | |
927 | strchr(info->name, ':') == NULL && | |
928 | strlen(homehost)+1+strlen(info->name) < 32) { | |
929 | strcpy(sb->set_name, homehost); | |
930 | strcat(sb->set_name, ":"); | |
931 | strcat(sb->set_name, info->name); | |
932 | } else | |
933 | strcpy(sb->set_name, info->name); | |
934 | } else if (strcmp(update, "devicesize") == 0 && | |
935 | __le64_to_cpu(sb->super_offset) < | |
936 | __le64_to_cpu(sb->data_offset)) { | |
937 | /* set data_size to device size less data_offset */ | |
938 | struct misc_dev_info *misc = (struct misc_dev_info*) | |
939 | (st->sb + MAX_SB_SIZE + BM_SUPER_SIZE); | |
940 | printf("Size was %llu\n", (unsigned long long) | |
941 | __le64_to_cpu(sb->data_size)); | |
942 | sb->data_size = __cpu_to_le64( | |
943 | misc->device_size - __le64_to_cpu(sb->data_offset)); | |
944 | printf("Size is %llu\n", (unsigned long long) | |
945 | __le64_to_cpu(sb->data_size)); | |
946 | } else if (strcmp(update, "_reshape_progress")==0) | |
947 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
948 | else if (strcmp(update, "writemostly")==0) | |
949 | sb->devflags |= WriteMostly1; | |
950 | else if (strcmp(update, "readwrite")==0) | |
951 | sb->devflags &= ~WriteMostly1; | |
952 | else | |
953 | rv = -1; | |
954 | ||
955 | sb->sb_csum = calc_sb_1_csum(sb); | |
956 | return rv; | |
957 | } | |
958 | ||
959 | static int init_super1(struct supertype *st, mdu_array_info_t *info, | |
960 | unsigned long long size, char *name, char *homehost, int *uuid) | |
961 | { | |
962 | struct mdp_superblock_1 *sb; | |
963 | int spares; | |
964 | int rfd; | |
965 | char defname[10]; | |
966 | int sbsize; | |
967 | ||
968 | if (posix_memalign((void**)&sb, 4096, SUPER1_SIZE) != 0) { | |
969 | pr_err("%s could not allocate superblock\n", __func__); | |
970 | return 0; | |
971 | } | |
972 | memset(sb, 0, SUPER1_SIZE); | |
973 | ||
974 | st->sb = sb; | |
975 | if (info == NULL) { | |
976 | /* zeroing superblock */ | |
977 | return 0; | |
978 | } | |
979 | ||
980 | spares = info->working_disks - info->active_disks; | |
981 | if (info->raid_disks + spares > MAX_DEVS) { | |
982 | pr_err("too many devices requested: %d+%d > %d\n", | |
983 | info->raid_disks , spares, MAX_DEVS); | |
984 | return 0; | |
985 | } | |
986 | ||
987 | sb->magic = __cpu_to_le32(MD_SB_MAGIC); | |
988 | sb->major_version = __cpu_to_le32(1); | |
989 | sb->feature_map = 0; | |
990 | sb->pad0 = 0; | |
991 | ||
992 | if (uuid) | |
993 | copy_uuid(sb->set_uuid, uuid, super1.swapuuid); | |
994 | else { | |
995 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
996 | read(rfd, sb->set_uuid, 16) != 16) { | |
997 | __u32 r[4] = {random(), random(), random(), random()}; | |
998 | memcpy(sb->set_uuid, r, 16); | |
999 | } | |
1000 | if (rfd >= 0) close(rfd); | |
1001 | } | |
1002 | ||
1003 | if (name == NULL || *name == 0) { | |
1004 | sprintf(defname, "%d", info->md_minor); | |
1005 | name = defname; | |
1006 | } | |
1007 | if (homehost && | |
1008 | strchr(name, ':')== NULL && | |
1009 | strlen(homehost)+1+strlen(name) < 32) { | |
1010 | strcpy(sb->set_name, homehost); | |
1011 | strcat(sb->set_name, ":"); | |
1012 | strcat(sb->set_name, name); | |
1013 | } else | |
1014 | strcpy(sb->set_name, name); | |
1015 | ||
1016 | sb->ctime = __cpu_to_le64((unsigned long long)time(0)); | |
1017 | sb->level = __cpu_to_le32(info->level); | |
1018 | sb->layout = __cpu_to_le32(info->layout); | |
1019 | sb->size = __cpu_to_le64(size*2ULL); | |
1020 | sb->chunksize = __cpu_to_le32(info->chunk_size>>9); | |
1021 | sb->raid_disks = __cpu_to_le32(info->raid_disks); | |
1022 | ||
1023 | sb->data_offset = __cpu_to_le64(0); | |
1024 | sb->data_size = __cpu_to_le64(0); | |
1025 | sb->super_offset = __cpu_to_le64(0); | |
1026 | sb->recovery_offset = __cpu_to_le64(0); | |
1027 | ||
1028 | sb->utime = sb->ctime; | |
1029 | sb->events = __cpu_to_le64(1); | |
1030 | if (info->state & (1<<MD_SB_CLEAN)) | |
1031 | sb->resync_offset = MaxSector; | |
1032 | else | |
1033 | sb->resync_offset = 0; | |
1034 | sbsize = sizeof(struct mdp_superblock_1) + 2 * (info->raid_disks + spares); | |
1035 | sbsize = ROUND_UP(sbsize, 512); | |
1036 | sb->max_dev = __cpu_to_le32((sbsize - sizeof(struct mdp_superblock_1)) / 2); | |
1037 | ||
1038 | memset(sb->dev_roles, 0xff, MAX_SB_SIZE - sizeof(struct mdp_superblock_1)); | |
1039 | ||
1040 | return 1; | |
1041 | } | |
1042 | ||
1043 | struct devinfo { | |
1044 | int fd; | |
1045 | char *devname; | |
1046 | mdu_disk_info_t disk; | |
1047 | struct devinfo *next; | |
1048 | }; | |
1049 | #ifndef MDASSEMBLE | |
1050 | /* Add a device to the superblock being created */ | |
1051 | static int add_to_super1(struct supertype *st, mdu_disk_info_t *dk, | |
1052 | int fd, char *devname) | |
1053 | { | |
1054 | struct mdp_superblock_1 *sb = st->sb; | |
1055 | __u16 *rp = sb->dev_roles + dk->number; | |
1056 | struct devinfo *di, **dip; | |
1057 | ||
1058 | if ((dk->state & 6) == 6) /* active, sync */ | |
1059 | *rp = __cpu_to_le16(dk->raid_disk); | |
1060 | else if ((dk->state & ~2) == 0) /* active or idle -> spare */ | |
1061 | *rp = 0xffff; | |
1062 | else | |
1063 | *rp = 0xfffe; | |
1064 | ||
1065 | if (dk->number >= (int)__le32_to_cpu(sb->max_dev) && | |
1066 | __le32_to_cpu(sb->max_dev) < MAX_DEVS) | |
1067 | sb->max_dev = __cpu_to_le32(dk->number+1); | |
1068 | ||
1069 | sb->dev_number = __cpu_to_le32(dk->number); | |
1070 | sb->devflags = 0; /* don't copy another disks flags */ | |
1071 | sb->sb_csum = calc_sb_1_csum(sb); | |
1072 | ||
1073 | dip = (struct devinfo **)&st->info; | |
1074 | while (*dip) | |
1075 | dip = &(*dip)->next; | |
1076 | di = xmalloc(sizeof(struct devinfo)); | |
1077 | di->fd = fd; | |
1078 | di->devname = devname; | |
1079 | di->disk = *dk; | |
1080 | di->next = NULL; | |
1081 | *dip = di; | |
1082 | ||
1083 | return 0; | |
1084 | } | |
1085 | #endif | |
1086 | ||
1087 | static void locate_bitmap1(struct supertype *st, int fd); | |
1088 | ||
1089 | static int store_super1(struct supertype *st, int fd) | |
1090 | { | |
1091 | struct mdp_superblock_1 *sb = st->sb; | |
1092 | unsigned long long sb_offset; | |
1093 | struct align_fd afd; | |
1094 | int sbsize; | |
1095 | unsigned long long dsize; | |
1096 | ||
1097 | if (!get_dev_size(fd, NULL, &dsize)) | |
1098 | return 1; | |
1099 | ||
1100 | dsize >>= 9; | |
1101 | ||
1102 | if (dsize < 24) | |
1103 | return 2; | |
1104 | ||
1105 | init_afd(&afd, fd); | |
1106 | ||
1107 | /* | |
1108 | * Calculate the position of the superblock. | |
1109 | * It is always aligned to a 4K boundary and | |
1110 | * depending on minor_version, it can be: | |
1111 | * 0: At least 8K, but less than 12K, from end of device | |
1112 | * 1: At start of device | |
1113 | * 2: 4K from start of device. | |
1114 | */ | |
1115 | switch(st->minor_version) { | |
1116 | case 0: | |
1117 | sb_offset = dsize; | |
1118 | sb_offset -= 8*2; | |
1119 | sb_offset &= ~(4*2-1); | |
1120 | break; | |
1121 | case 1: | |
1122 | sb_offset = 0; | |
1123 | break; | |
1124 | case 2: | |
1125 | sb_offset = 4*2; | |
1126 | break; | |
1127 | default: | |
1128 | return -EINVAL; | |
1129 | } | |
1130 | ||
1131 | ||
1132 | ||
1133 | if (sb_offset != __le64_to_cpu(sb->super_offset) && | |
1134 | 0 != __le64_to_cpu(sb->super_offset) | |
1135 | ) { | |
1136 | pr_err("internal error - sb_offset is wrong\n"); | |
1137 | abort(); | |
1138 | } | |
1139 | ||
1140 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) | |
1141 | return 3; | |
1142 | ||
1143 | sbsize = ROUND_UP(sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev), 512); | |
1144 | ||
1145 | if (awrite(&afd, sb, sbsize) != sbsize) | |
1146 | return 4; | |
1147 | ||
1148 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
1149 | struct bitmap_super_s *bm = (struct bitmap_super_s*) | |
1150 | (((char*)sb)+MAX_SB_SIZE); | |
1151 | if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) { | |
1152 | locate_bitmap1(st, fd); | |
1153 | if (awrite(&afd, bm, sizeof(*bm)) != sizeof(*bm)) | |
1154 | return 5; | |
1155 | } | |
1156 | } | |
1157 | fsync(fd); | |
1158 | return 0; | |
1159 | } | |
1160 | ||
1161 | static int load_super1(struct supertype *st, int fd, char *devname); | |
1162 | ||
1163 | static unsigned long choose_bm_space(unsigned long devsize) | |
1164 | { | |
1165 | /* if the device is bigger than 8Gig, save 64k for bitmap usage, | |
1166 | * if bigger than 200Gig, save 128k | |
1167 | * NOTE: result must be multiple of 4K else bad things happen | |
1168 | * on 4K-sector devices. | |
1169 | */ | |
1170 | if (devsize < 64*2) return 0; | |
1171 | if (devsize - 64*2 >= 200*1024*1024*2) | |
1172 | return 128*2; | |
1173 | if (devsize - 4*2 > 8*1024*1024*2) | |
1174 | return 64*2; | |
1175 | return 4*2; | |
1176 | } | |
1177 | ||
1178 | static void free_super1(struct supertype *st); | |
1179 | ||
1180 | #ifndef MDASSEMBLE | |
1181 | static int write_init_super1(struct supertype *st) | |
1182 | { | |
1183 | struct mdp_superblock_1 *sb = st->sb; | |
1184 | struct supertype *refst; | |
1185 | int rfd; | |
1186 | int rv = 0; | |
1187 | unsigned long long bm_space; | |
1188 | unsigned long long reserved; | |
1189 | struct devinfo *di; | |
1190 | unsigned long long dsize, array_size; | |
1191 | unsigned long long sb_offset, headroom; | |
1192 | ||
1193 | for (di = st->info; di; di = di->next) { | |
1194 | if (di->disk.state == 1) | |
1195 | continue; | |
1196 | if (di->fd < 0) | |
1197 | continue; | |
1198 | ||
1199 | while (Kill(di->devname, NULL, 0, -1, 1) == 0) | |
1200 | ; | |
1201 | ||
1202 | sb->dev_number = __cpu_to_le32(di->disk.number); | |
1203 | if (di->disk.state & (1<<MD_DISK_WRITEMOSTLY)) | |
1204 | sb->devflags |= WriteMostly1; | |
1205 | else | |
1206 | sb->devflags &= ~WriteMostly1; | |
1207 | ||
1208 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
1209 | read(rfd, sb->device_uuid, 16) != 16) { | |
1210 | __u32 r[4] = {random(), random(), random(), random()}; | |
1211 | memcpy(sb->device_uuid, r, 16); | |
1212 | } | |
1213 | if (rfd >= 0) | |
1214 | close(rfd); | |
1215 | ||
1216 | sb->events = 0; | |
1217 | ||
1218 | refst = dup_super(st); | |
1219 | if (load_super1(refst, di->fd, NULL)==0) { | |
1220 | struct mdp_superblock_1 *refsb = refst->sb; | |
1221 | ||
1222 | memcpy(sb->device_uuid, refsb->device_uuid, 16); | |
1223 | if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) { | |
1224 | /* same array, so preserve events and | |
1225 | * dev_number */ | |
1226 | sb->events = refsb->events; | |
1227 | /* bugs in 2.6.17 and earlier mean the | |
1228 | * dev_number chosen in Manage must be preserved | |
1229 | */ | |
1230 | if (get_linux_version() >= 2006018) | |
1231 | sb->dev_number = refsb->dev_number; | |
1232 | } | |
1233 | free_super1(refst); | |
1234 | } | |
1235 | free(refst); | |
1236 | ||
1237 | if (!get_dev_size(di->fd, NULL, &dsize)) { | |
1238 | rv = 1; | |
1239 | goto error_out; | |
1240 | } | |
1241 | dsize >>= 9; | |
1242 | ||
1243 | if (dsize < 24) { | |
1244 | close(di->fd); | |
1245 | rv = 2; | |
1246 | goto error_out; | |
1247 | } | |
1248 | ||
1249 | ||
1250 | /* | |
1251 | * Calculate the position of the superblock. | |
1252 | * It is always aligned to a 4K boundary and | |
1253 | * depending on minor_version, it can be: | |
1254 | * 0: At least 8K, but less than 12K, from end of device | |
1255 | * 1: At start of device | |
1256 | * 2: 4K from start of device. | |
1257 | * Depending on the array size, we might leave extra space | |
1258 | * for a bitmap. | |
1259 | * Also leave 4K for bad-block log. | |
1260 | */ | |
1261 | array_size = __le64_to_cpu(sb->size); | |
1262 | /* work out how much space we left for a bitmap, | |
1263 | * Add 8 sectors for bad block log */ | |
1264 | bm_space = choose_bm_space(array_size) + 8; | |
1265 | ||
1266 | /* We try to leave 0.1% at the start for reshape | |
1267 | * operations, but limit this to 128Meg (0.1% of 10Gig) | |
1268 | * which is plenty for efficient reshapes | |
1269 | */ | |
1270 | headroom = 128 * 1024 * 2; | |
1271 | while (headroom << 10 > array_size) | |
1272 | headroom >>= 1; | |
1273 | ||
1274 | switch(st->minor_version) { | |
1275 | case 0: | |
1276 | sb_offset = dsize; | |
1277 | sb_offset -= 8*2; | |
1278 | sb_offset &= ~(4*2-1); | |
1279 | sb->super_offset = __cpu_to_le64(sb_offset); | |
1280 | sb->data_offset = __cpu_to_le64(0); | |
1281 | if (sb_offset < array_size + bm_space) | |
1282 | bm_space = sb_offset - array_size; | |
1283 | sb->data_size = __cpu_to_le64(sb_offset - bm_space); | |
1284 | if (bm_space >= 8) { | |
1285 | sb->bblog_size = __cpu_to_le16(8); | |
1286 | sb->bblog_offset = __cpu_to_le32((unsigned)-8); | |
1287 | } | |
1288 | break; | |
1289 | case 1: | |
1290 | sb->super_offset = __cpu_to_le64(0); | |
1291 | reserved = bm_space + 4*2; | |
1292 | if (reserved < headroom) | |
1293 | reserved = headroom; | |
1294 | if (reserved + array_size > dsize) | |
1295 | reserved = dsize - array_size; | |
1296 | /* Try for multiple of 1Meg so it is nicely aligned */ | |
1297 | #define ONE_MEG (2*1024) | |
1298 | if (reserved > ONE_MEG) | |
1299 | reserved = (reserved/ONE_MEG) * ONE_MEG; | |
1300 | ||
1301 | /* force 4K alignment */ | |
1302 | reserved &= ~7ULL; | |
1303 | ||
1304 | sb->data_offset = __cpu_to_le64(reserved); | |
1305 | sb->data_size = __cpu_to_le64(dsize - reserved); | |
1306 | if (reserved >= 16) { | |
1307 | sb->bblog_size = __cpu_to_le16(8); | |
1308 | sb->bblog_offset = __cpu_to_le32(reserved-8); | |
1309 | } | |
1310 | break; | |
1311 | case 2: | |
1312 | sb_offset = 4*2; | |
1313 | sb->super_offset = __cpu_to_le64(4*2); | |
1314 | if (4*2 + 4*2 + bm_space + array_size | |
1315 | > dsize) | |
1316 | bm_space = dsize - array_size | |
1317 | - 4*2 - 4*2; | |
1318 | ||
1319 | reserved = bm_space + 4*2 + 4*2; | |
1320 | if (reserved < headroom) | |
1321 | reserved = headroom; | |
1322 | if (reserved + array_size > dsize) | |
1323 | reserved = dsize - array_size; | |
1324 | /* Try for multiple of 1Meg so it is nicely aligned */ | |
1325 | #define ONE_MEG (2*1024) | |
1326 | if (reserved > ONE_MEG) | |
1327 | reserved = (reserved/ONE_MEG) * ONE_MEG; | |
1328 | ||
1329 | /* force 4K alignment */ | |
1330 | reserved &= ~7ULL; | |
1331 | ||
1332 | sb->data_offset = __cpu_to_le64(reserved); | |
1333 | sb->data_size = __cpu_to_le64(dsize - reserved); | |
1334 | if (reserved >= 16+16) { | |
1335 | sb->bblog_size = __cpu_to_le16(8); | |
1336 | /* '8' sectors for the bblog, and another '8' | |
1337 | * because we want offset from superblock, not | |
1338 | * start of device. | |
1339 | */ | |
1340 | sb->bblog_offset = __cpu_to_le32(reserved-8-8); | |
1341 | } | |
1342 | break; | |
1343 | default: | |
1344 | pr_err("Failed to write invalid " | |
1345 | "metadata format 1.%i to %s\n", | |
1346 | st->minor_version, di->devname); | |
1347 | rv = -EINVAL; | |
1348 | goto out; | |
1349 | } | |
1350 | ||
1351 | sb->sb_csum = calc_sb_1_csum(sb); | |
1352 | rv = store_super1(st, di->fd); | |
1353 | if (rv == 0 && (__le32_to_cpu(sb->feature_map) & 1)) | |
1354 | rv = st->ss->write_bitmap(st, di->fd); | |
1355 | close(di->fd); | |
1356 | di->fd = -1; | |
1357 | if (rv) | |
1358 | goto error_out; | |
1359 | } | |
1360 | error_out: | |
1361 | if (rv) | |
1362 | pr_err("Failed to write metadata to %s\n", | |
1363 | di->devname); | |
1364 | out: | |
1365 | return rv; | |
1366 | } | |
1367 | #endif | |
1368 | ||
1369 | static int compare_super1(struct supertype *st, struct supertype *tst) | |
1370 | { | |
1371 | /* | |
1372 | * return: | |
1373 | * 0 same, or first was empty, and second was copied | |
1374 | * 1 second had wrong number | |
1375 | * 2 wrong uuid | |
1376 | * 3 wrong other info | |
1377 | */ | |
1378 | struct mdp_superblock_1 *first = st->sb; | |
1379 | struct mdp_superblock_1 *second = tst->sb; | |
1380 | ||
1381 | if (second->magic != __cpu_to_le32(MD_SB_MAGIC)) | |
1382 | return 1; | |
1383 | if (second->major_version != __cpu_to_le32(1)) | |
1384 | return 1; | |
1385 | ||
1386 | if (!first) { | |
1387 | if (posix_memalign((void**)&first, 4096, SUPER1_SIZE) != 0) { | |
1388 | pr_err("%s could not allocate superblock\n", __func__); | |
1389 | return 1; | |
1390 | } | |
1391 | memcpy(first, second, SUPER1_SIZE); | |
1392 | st->sb = first; | |
1393 | return 0; | |
1394 | } | |
1395 | if (memcmp(first->set_uuid, second->set_uuid, 16)!= 0) | |
1396 | return 2; | |
1397 | ||
1398 | if (first->ctime != second->ctime || | |
1399 | first->level != second->level || | |
1400 | first->layout != second->layout || | |
1401 | first->size != second->size || | |
1402 | first->chunksize != second->chunksize || | |
1403 | first->raid_disks != second->raid_disks) | |
1404 | return 3; | |
1405 | return 0; | |
1406 | } | |
1407 | ||
1408 | static int load_super1(struct supertype *st, int fd, char *devname) | |
1409 | { | |
1410 | unsigned long long dsize; | |
1411 | unsigned long long sb_offset; | |
1412 | struct mdp_superblock_1 *super; | |
1413 | int uuid[4]; | |
1414 | struct bitmap_super_s *bsb; | |
1415 | struct misc_dev_info *misc; | |
1416 | struct align_fd afd; | |
1417 | ||
1418 | free_super1(st); | |
1419 | ||
1420 | init_afd(&afd, fd); | |
1421 | ||
1422 | if (st->ss == NULL || st->minor_version == -1) { | |
1423 | int bestvers = -1; | |
1424 | struct supertype tst; | |
1425 | __u64 bestctime = 0; | |
1426 | /* guess... choose latest ctime */ | |
1427 | memset(&tst, 0, sizeof(tst)); | |
1428 | tst.ss = &super1; | |
1429 | for (tst.minor_version = 0; tst.minor_version <= 2 ; tst.minor_version++) { | |
1430 | switch(load_super1(&tst, fd, devname)) { | |
1431 | case 0: super = tst.sb; | |
1432 | if (bestvers == -1 || | |
1433 | bestctime < __le64_to_cpu(super->ctime)) { | |
1434 | bestvers = tst.minor_version; | |
1435 | bestctime = __le64_to_cpu(super->ctime); | |
1436 | } | |
1437 | free(super); | |
1438 | tst.sb = NULL; | |
1439 | break; | |
1440 | case 1: return 1; /*bad device */ | |
1441 | case 2: break; /* bad, try next */ | |
1442 | } | |
1443 | } | |
1444 | if (bestvers != -1) { | |
1445 | int rv; | |
1446 | tst.minor_version = bestvers; | |
1447 | tst.ss = &super1; | |
1448 | tst.max_devs = MAX_DEVS; | |
1449 | rv = load_super1(&tst, fd, devname); | |
1450 | if (rv == 0) | |
1451 | *st = tst; | |
1452 | return rv; | |
1453 | } | |
1454 | return 2; | |
1455 | } | |
1456 | if (!get_dev_size(fd, devname, &dsize)) | |
1457 | return 1; | |
1458 | dsize >>= 9; | |
1459 | ||
1460 | if (dsize < 24) { | |
1461 | if (devname) | |
1462 | pr_err("%s is too small for md: size is %llu sectors.\n", | |
1463 | devname, dsize); | |
1464 | return 1; | |
1465 | } | |
1466 | ||
1467 | /* | |
1468 | * Calculate the position of the superblock. | |
1469 | * It is always aligned to a 4K boundary and | |
1470 | * depending on minor_version, it can be: | |
1471 | * 0: At least 8K, but less than 12K, from end of device | |
1472 | * 1: At start of device | |
1473 | * 2: 4K from start of device. | |
1474 | */ | |
1475 | switch(st->minor_version) { | |
1476 | case 0: | |
1477 | sb_offset = dsize; | |
1478 | sb_offset -= 8*2; | |
1479 | sb_offset &= ~(4*2-1); | |
1480 | break; | |
1481 | case 1: | |
1482 | sb_offset = 0; | |
1483 | break; | |
1484 | case 2: | |
1485 | sb_offset = 4*2; | |
1486 | break; | |
1487 | default: | |
1488 | return -EINVAL; | |
1489 | } | |
1490 | ||
1491 | ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */ | |
1492 | ||
1493 | ||
1494 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) { | |
1495 | if (devname) | |
1496 | pr_err("Cannot seek to superblock on %s: %s\n", | |
1497 | devname, strerror(errno)); | |
1498 | return 1; | |
1499 | } | |
1500 | ||
1501 | if (posix_memalign((void**)&super, 4096, SUPER1_SIZE) != 0) { | |
1502 | pr_err("%s could not allocate superblock\n", | |
1503 | __func__); | |
1504 | return 1; | |
1505 | } | |
1506 | ||
1507 | if (aread(&afd, super, MAX_SB_SIZE) != MAX_SB_SIZE) { | |
1508 | if (devname) | |
1509 | pr_err("Cannot read superblock on %s\n", | |
1510 | devname); | |
1511 | free(super); | |
1512 | return 1; | |
1513 | } | |
1514 | ||
1515 | if (__le32_to_cpu(super->magic) != MD_SB_MAGIC) { | |
1516 | if (devname) | |
1517 | pr_err("No super block found on %s (Expected magic %08x, got %08x)\n", | |
1518 | devname, MD_SB_MAGIC, __le32_to_cpu(super->magic)); | |
1519 | free(super); | |
1520 | return 2; | |
1521 | } | |
1522 | ||
1523 | if (__le32_to_cpu(super->major_version) != 1) { | |
1524 | if (devname) | |
1525 | pr_err("Cannot interpret superblock on %s - version is %d\n", | |
1526 | devname, __le32_to_cpu(super->major_version)); | |
1527 | free(super); | |
1528 | return 2; | |
1529 | } | |
1530 | if (__le64_to_cpu(super->super_offset) != sb_offset) { | |
1531 | if (devname) | |
1532 | pr_err("No superblock found on %s (super_offset is wrong)\n", | |
1533 | devname); | |
1534 | free(super); | |
1535 | return 2; | |
1536 | } | |
1537 | st->sb = super; | |
1538 | ||
1539 | bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); | |
1540 | ||
1541 | misc = (struct misc_dev_info*) (((char*)super)+MAX_SB_SIZE+BM_SUPER_SIZE); | |
1542 | misc->device_size = dsize; | |
1543 | ||
1544 | /* Now check on the bitmap superblock */ | |
1545 | if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0) | |
1546 | return 0; | |
1547 | /* Read the bitmap superblock and make sure it looks | |
1548 | * valid. If it doesn't clear the bit. An --assemble --force | |
1549 | * should get that written out. | |
1550 | */ | |
1551 | locate_bitmap1(st, fd); | |
1552 | if (aread(&afd, bsb, 512) != 512) | |
1553 | goto no_bitmap; | |
1554 | ||
1555 | uuid_from_super1(st, uuid); | |
1556 | if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC || | |
1557 | memcmp(bsb->uuid, uuid, 16) != 0) | |
1558 | goto no_bitmap; | |
1559 | return 0; | |
1560 | ||
1561 | no_bitmap: | |
1562 | super->feature_map = __cpu_to_le32(__le32_to_cpu(super->feature_map) | |
1563 | & ~MD_FEATURE_BITMAP_OFFSET); | |
1564 | return 0; | |
1565 | } | |
1566 | ||
1567 | ||
1568 | static struct supertype *match_metadata_desc1(char *arg) | |
1569 | { | |
1570 | struct supertype *st = xcalloc(1, sizeof(*st)); | |
1571 | ||
1572 | st->container_dev = NoMdDev; | |
1573 | st->ss = &super1; | |
1574 | st->max_devs = MAX_DEVS; | |
1575 | st->sb = NULL; | |
1576 | /* leading zeros can be safely ignored. --detail generates them. */ | |
1577 | while (*arg == '0') | |
1578 | arg++; | |
1579 | if (strcmp(arg, "1.0") == 0 || | |
1580 | strcmp(arg, "1.00") == 0) { | |
1581 | st->minor_version = 0; | |
1582 | return st; | |
1583 | } | |
1584 | if (strcmp(arg, "1.1") == 0 || | |
1585 | strcmp(arg, "1.01") == 0 | |
1586 | ) { | |
1587 | st->minor_version = 1; | |
1588 | return st; | |
1589 | } | |
1590 | if (strcmp(arg, "1.2") == 0 || | |
1591 | #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */ | |
1592 | strcmp(arg, "default") == 0 || | |
1593 | #endif /* DEFAULT_OLD_METADATA */ | |
1594 | strcmp(arg, "1.02") == 0) { | |
1595 | st->minor_version = 2; | |
1596 | return st; | |
1597 | } | |
1598 | if (strcmp(arg, "1") == 0 || | |
1599 | strcmp(arg, "default") == 0) { | |
1600 | st->minor_version = -1; | |
1601 | return st; | |
1602 | } | |
1603 | ||
1604 | free(st); | |
1605 | return NULL; | |
1606 | } | |
1607 | ||
1608 | /* find available size on device with this devsize, using | |
1609 | * superblock type st, and reserving 'reserve' sectors for | |
1610 | * a possible bitmap | |
1611 | */ | |
1612 | static __u64 avail_size1(struct supertype *st, __u64 devsize) | |
1613 | { | |
1614 | struct mdp_superblock_1 *super = st->sb; | |
1615 | if (devsize < 24) | |
1616 | return 0; | |
1617 | ||
1618 | if (super == NULL) | |
1619 | /* creating: allow suitable space for bitmap */ | |
1620 | devsize -= choose_bm_space(devsize); | |
1621 | #ifndef MDASSEMBLE | |
1622 | else if (__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) { | |
1623 | /* hot-add. allow for actual size of bitmap */ | |
1624 | struct bitmap_super_s *bsb; | |
1625 | bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); | |
1626 | devsize -= bitmap_sectors(bsb); | |
1627 | } | |
1628 | #endif | |
1629 | /* Allow space for bad block log */ | |
1630 | if (super && super->bblog_size) | |
1631 | devsize -= __le16_to_cpu(super->bblog_size); | |
1632 | else | |
1633 | devsize -= 8; | |
1634 | ||
1635 | if (st->minor_version < 0) | |
1636 | /* not specified, so time to set default */ | |
1637 | st->minor_version = 2; | |
1638 | if (super == NULL && st->minor_version > 0) { | |
1639 | /* haven't committed to a size yet, so allow some | |
1640 | * slack for space for reshape. | |
1641 | * Limit slack to 128M, but aim for about 0.1% | |
1642 | */ | |
1643 | unsigned long long headroom = 128*1024*2; | |
1644 | while ((headroom << 10) > devsize) | |
1645 | headroom >>= 1; | |
1646 | devsize -= headroom; | |
1647 | } | |
1648 | switch(st->minor_version) { | |
1649 | case 0: | |
1650 | /* at end */ | |
1651 | return ((devsize - 8*2 ) & ~(4*2-1)); | |
1652 | case 1: | |
1653 | /* at start, 4K for superblock and possible bitmap */ | |
1654 | return devsize - 4*2; | |
1655 | case 2: | |
1656 | /* 4k from start, 4K for superblock and possible bitmap */ | |
1657 | return devsize - (4+4)*2; | |
1658 | } | |
1659 | return 0; | |
1660 | } | |
1661 | ||
1662 | static int | |
1663 | add_internal_bitmap1(struct supertype *st, | |
1664 | int *chunkp, int delay, int write_behind, | |
1665 | unsigned long long size, | |
1666 | int may_change, int major) | |
1667 | { | |
1668 | /* | |
1669 | * If not may_change, then this is a 'Grow' without sysfs support for | |
1670 | * bitmaps, and the bitmap must fit after the superblock at 1K offset. | |
1671 | * If may_change, then this is create or a Grow with sysfs syupport, | |
1672 | * and we can put the bitmap wherever we like. | |
1673 | * | |
1674 | * size is in sectors, chunk is in bytes !!! | |
1675 | */ | |
1676 | ||
1677 | unsigned long long bits; | |
1678 | unsigned long long max_bits; | |
1679 | unsigned long long min_chunk; | |
1680 | long offset; | |
1681 | long bbl_offset, bbl_size; | |
1682 | unsigned long long chunk = *chunkp; | |
1683 | int room = 0; | |
1684 | int creating = 0; | |
1685 | struct mdp_superblock_1 *sb = st->sb; | |
1686 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); | |
1687 | int uuid[4]; | |
1688 | ||
1689 | ||
1690 | if (__le64_to_cpu(sb->data_size) == 0) | |
1691 | /* Must be creating the array, else data_size would be non-zero */ | |
1692 | creating = 1; | |
1693 | switch(st->minor_version) { | |
1694 | case 0: | |
1695 | /* either 3K after the superblock (when hot-add), | |
1696 | * or some amount of space before. | |
1697 | */ | |
1698 | if (creating) { | |
1699 | /* We are creating array, so we *know* how much room has | |
1700 | * been left. | |
1701 | */ | |
1702 | offset = 0; | |
1703 | room = choose_bm_space(__le64_to_cpu(sb->size)); | |
1704 | bbl_size = 8; | |
1705 | } else { | |
1706 | room = __le64_to_cpu(sb->super_offset) | |
1707 | - __le64_to_cpu(sb->data_offset) | |
1708 | - __le64_to_cpu(sb->data_size); | |
1709 | bbl_size = __le16_to_cpu(sb->bblog_size); | |
1710 | if (bbl_size < 8) | |
1711 | bbl_size = 8; | |
1712 | bbl_offset = (__s32)__le32_to_cpu(sb->bblog_offset); | |
1713 | if (bbl_size < -bbl_offset) | |
1714 | bbl_size = -bbl_offset; | |
1715 | ||
1716 | if (!may_change || (room < 3*2 && | |
1717 | __le32_to_cpu(sb->max_dev) <= 384)) { | |
1718 | room = 3*2; | |
1719 | offset = 1*2; | |
1720 | bbl_size = 0; | |
1721 | } else { | |
1722 | offset = 0; /* means movable offset */ | |
1723 | } | |
1724 | } | |
1725 | break; | |
1726 | case 1: | |
1727 | case 2: /* between superblock and data */ | |
1728 | if (creating) { | |
1729 | offset = 4*2; | |
1730 | room = choose_bm_space(__le64_to_cpu(sb->size)); | |
1731 | bbl_size = 8; | |
1732 | } else { | |
1733 | room = __le64_to_cpu(sb->data_offset) | |
1734 | - __le64_to_cpu(sb->super_offset); | |
1735 | bbl_size = __le16_to_cpu(sb->bblog_size); | |
1736 | if (bbl_size) | |
1737 | room = __le32_to_cpu(sb->bblog_offset) + bbl_size; | |
1738 | else | |
1739 | bbl_size = 8; | |
1740 | ||
1741 | if (!may_change) { | |
1742 | room -= 2; /* Leave 1K for superblock */ | |
1743 | offset = 2; | |
1744 | bbl_size = 0; | |
1745 | } else { | |
1746 | room -= 4*2; /* leave 4K for superblock */ | |
1747 | offset = 4*2; | |
1748 | } | |
1749 | } | |
1750 | break; | |
1751 | default: | |
1752 | return 0; | |
1753 | } | |
1754 | ||
1755 | room -= bbl_size; | |
1756 | if (chunk == UnSet && room > 128*2) | |
1757 | /* Limit to 128K of bitmap when chunk size not requested */ | |
1758 | room = 128*2; | |
1759 | ||
1760 | max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8; | |
1761 | ||
1762 | min_chunk = 4096; /* sub-page chunks don't work yet.. */ | |
1763 | bits = (size*512)/min_chunk +1; | |
1764 | while (bits > max_bits) { | |
1765 | min_chunk *= 2; | |
1766 | bits = (bits+1)/2; | |
1767 | } | |
1768 | if (chunk == UnSet) { | |
1769 | /* For practical purpose, 64Meg is a good | |
1770 | * default chunk size for internal bitmaps. | |
1771 | */ | |
1772 | chunk = min_chunk; | |
1773 | if (chunk < 64*1024*1024) | |
1774 | chunk = 64*1024*1024; | |
1775 | } else if (chunk < min_chunk) | |
1776 | return 0; /* chunk size too small */ | |
1777 | if (chunk == 0) /* rounding problem */ | |
1778 | return 0; | |
1779 | ||
1780 | if (offset == 0) { | |
1781 | /* start bitmap on a 4K boundary with enough space for | |
1782 | * the bitmap | |
1783 | */ | |
1784 | bits = (size*512) / chunk + 1; | |
1785 | room = ((bits+7)/8 + sizeof(bitmap_super_t) +4095)/4096; | |
1786 | room *= 8; /* convert 4K blocks to sectors */ | |
1787 | offset = -room - bbl_size; | |
1788 | } | |
1789 | ||
1790 | sb->bitmap_offset = (int32_t)__cpu_to_le32(offset); | |
1791 | ||
1792 | sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | |
1793 | | MD_FEATURE_BITMAP_OFFSET); | |
1794 | memset(bms, 0, sizeof(*bms)); | |
1795 | bms->magic = __cpu_to_le32(BITMAP_MAGIC); | |
1796 | bms->version = __cpu_to_le32(major); | |
1797 | uuid_from_super1(st, uuid); | |
1798 | memcpy(bms->uuid, uuid, 16); | |
1799 | bms->chunksize = __cpu_to_le32(chunk); | |
1800 | bms->daemon_sleep = __cpu_to_le32(delay); | |
1801 | bms->sync_size = __cpu_to_le64(size); | |
1802 | bms->write_behind = __cpu_to_le32(write_behind); | |
1803 | ||
1804 | *chunkp = chunk; | |
1805 | return 1; | |
1806 | } | |
1807 | ||
1808 | static void locate_bitmap1(struct supertype *st, int fd) | |
1809 | { | |
1810 | unsigned long long offset; | |
1811 | struct mdp_superblock_1 *sb; | |
1812 | int mustfree = 0; | |
1813 | ||
1814 | if (!st->sb) { | |
1815 | if (st->ss->load_super(st, fd, NULL)) | |
1816 | return; /* no error I hope... */ | |
1817 | mustfree = 1; | |
1818 | } | |
1819 | sb = st->sb; | |
1820 | ||
1821 | offset = __le64_to_cpu(sb->super_offset); | |
1822 | offset += (int32_t) __le32_to_cpu(sb->bitmap_offset); | |
1823 | if (mustfree) | |
1824 | free(sb); | |
1825 | lseek64(fd, offset<<9, 0); | |
1826 | } | |
1827 | ||
1828 | static int write_bitmap1(struct supertype *st, int fd) | |
1829 | { | |
1830 | struct mdp_superblock_1 *sb = st->sb; | |
1831 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+MAX_SB_SIZE); | |
1832 | int rv = 0; | |
1833 | void *buf; | |
1834 | int towrite, n; | |
1835 | struct align_fd afd; | |
1836 | ||
1837 | init_afd(&afd, fd); | |
1838 | ||
1839 | locate_bitmap1(st, fd); | |
1840 | ||
1841 | if (posix_memalign(&buf, 4096, 4096)) | |
1842 | return -ENOMEM; | |
1843 | ||
1844 | memset(buf, 0xff, 4096); | |
1845 | memcpy(buf, (char *)bms, sizeof(bitmap_super_t)); | |
1846 | ||
1847 | towrite = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9); | |
1848 | towrite = (towrite+7) >> 3; /* bits to bytes */ | |
1849 | towrite += sizeof(bitmap_super_t); | |
1850 | towrite = ROUND_UP(towrite, 512); | |
1851 | while (towrite > 0) { | |
1852 | n = towrite; | |
1853 | if (n > 4096) | |
1854 | n = 4096; | |
1855 | n = awrite(&afd, buf, n); | |
1856 | if (n > 0) | |
1857 | towrite -= n; | |
1858 | else | |
1859 | break; | |
1860 | memset(buf, 0xff, 4096); | |
1861 | } | |
1862 | fsync(fd); | |
1863 | if (towrite) | |
1864 | rv = -2; | |
1865 | ||
1866 | free(buf); | |
1867 | return rv; | |
1868 | } | |
1869 | ||
1870 | static void free_super1(struct supertype *st) | |
1871 | { | |
1872 | if (st->sb) | |
1873 | free(st->sb); | |
1874 | while (st->info) { | |
1875 | struct devinfo *di = st->info; | |
1876 | st->info = di->next; | |
1877 | if (di->fd >= 0) | |
1878 | close(di->fd); | |
1879 | free(di); | |
1880 | } | |
1881 | st->sb = NULL; | |
1882 | } | |
1883 | ||
1884 | #ifndef MDASSEMBLE | |
1885 | static int validate_geometry1(struct supertype *st, int level, | |
1886 | int layout, int raiddisks, | |
1887 | int *chunk, unsigned long long size, | |
1888 | char *subdev, unsigned long long *freesize, | |
1889 | int verbose) | |
1890 | { | |
1891 | unsigned long long ldsize; | |
1892 | int fd; | |
1893 | ||
1894 | if (level == LEVEL_CONTAINER) { | |
1895 | if (verbose) | |
1896 | pr_err("1.x metadata does not support containers\n"); | |
1897 | return 0; | |
1898 | } | |
1899 | if (chunk && *chunk == UnSet) | |
1900 | *chunk = DEFAULT_CHUNK; | |
1901 | ||
1902 | if (!subdev) | |
1903 | return 1; | |
1904 | ||
1905 | fd = open(subdev, O_RDONLY|O_EXCL, 0); | |
1906 | if (fd < 0) { | |
1907 | if (verbose) | |
1908 | pr_err("super1.x cannot open %s: %s\n", | |
1909 | subdev, strerror(errno)); | |
1910 | return 0; | |
1911 | } | |
1912 | ||
1913 | if (!get_dev_size(fd, subdev, &ldsize)) { | |
1914 | close(fd); | |
1915 | return 0; | |
1916 | } | |
1917 | close(fd); | |
1918 | ||
1919 | *freesize = avail_size1(st, ldsize >> 9); | |
1920 | return 1; | |
1921 | } | |
1922 | #endif /* MDASSEMBLE */ | |
1923 | ||
1924 | struct superswitch super1 = { | |
1925 | #ifndef MDASSEMBLE | |
1926 | .examine_super = examine_super1, | |
1927 | .brief_examine_super = brief_examine_super1, | |
1928 | .export_examine_super = export_examine_super1, | |
1929 | .detail_super = detail_super1, | |
1930 | .brief_detail_super = brief_detail_super1, | |
1931 | .export_detail_super = export_detail_super1, | |
1932 | .write_init_super = write_init_super1, | |
1933 | .validate_geometry = validate_geometry1, | |
1934 | .add_to_super = add_to_super1, | |
1935 | #endif | |
1936 | .match_home = match_home1, | |
1937 | .uuid_from_super = uuid_from_super1, | |
1938 | .getinfo_super = getinfo_super1, | |
1939 | .container_content = container_content1, | |
1940 | .update_super = update_super1, | |
1941 | .init_super = init_super1, | |
1942 | .store_super = store_super1, | |
1943 | .compare_super = compare_super1, | |
1944 | .load_super = load_super1, | |
1945 | .match_metadata_desc = match_metadata_desc1, | |
1946 | .avail_size = avail_size1, | |
1947 | .add_internal_bitmap = add_internal_bitmap1, | |
1948 | .locate_bitmap = locate_bitmap1, | |
1949 | .write_bitmap = write_bitmap1, | |
1950 | .free_super = free_super1, | |
1951 | #if __BYTE_ORDER == BIG_ENDIAN | |
1952 | .swapuuid = 0, | |
1953 | #else | |
1954 | .swapuuid = 1, | |
1955 | #endif | |
1956 | .name = "1.x", | |
1957 | }; |