]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super1.c
projects / thirdparty / mdadm.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
imsm: FIX: Use function to obtain array layout
[thirdparty/mdadm.git] / super1.c
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 __u8 pad2[64-57]; /* set to 0 when writing */
74
75 /* array state information - 64 bytes */
76 __u64 utime; /* 40 bits second, 24 btes microseconds */
77 __u64 events; /* incremented when superblock updated */
78 __u64 resync_offset; /* data before this offset (from data_offset) known to be in sync */
79 __u32 sb_csum; /* checksum upto dev_roles[max_dev] */
80 __u32 max_dev; /* size of dev_roles[] array to consider */
81 __u8 pad3[64-32]; /* set to 0 when writing */
82
83 /* device state information. Indexed by dev_number.
84 * 2 bytes per device
85 * Note there are no per-device state flags. State information is rolled
86 * into the 'roles' value. If a device is spare or faulty, then it doesn't
87 * have a meaningful role.
88 */
89 __u16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
90 };
91
92 struct misc_dev_info {
93 __u64 device_size;
94 };
95
96 /* feature_map bits */
97 #define MD_FEATURE_BITMAP_OFFSET 1
98 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
99 * must be honoured
100 */
101 #define MD_FEATURE_RESHAPE_ACTIVE 4
102
103 #define MD_FEATURE_ALL (1|2|4)
104
105 #ifndef offsetof
106 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
107 #endif
108 static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb)
109 {
110 unsigned int disk_csum, csum;
111 unsigned long long newcsum;
112 int size = sizeof(*sb) + __le32_to_cpu(sb->max_dev)*2;
113 unsigned int *isuper = (unsigned int*)sb;
114 int i;
115
116 /* make sure I can count... */
117 if (offsetof(struct mdp_superblock_1,data_offset) != 128 ||
118 offsetof(struct mdp_superblock_1, utime) != 192 ||
119 sizeof(struct mdp_superblock_1) != 256) {
120 fprintf(stderr, "WARNING - superblock isn't sized correctly\n");
121 }
122
123 disk_csum = sb->sb_csum;
124 sb->sb_csum = 0;
125 newcsum = 0;
126 for (i=0; size>=4; size -= 4 ) {
127 newcsum += __le32_to_cpu(*isuper);
128 isuper++;
129 }
130
131 if (size == 2)
132 newcsum += __le16_to_cpu(*(unsigned short*) isuper);
133
134 csum = (newcsum & 0xffffffff) + (newcsum >> 32);
135 sb->sb_csum = disk_csum;
136 return __cpu_to_le32(csum);
137 }
138
139 static char abuf[4096+4096];
140 static int aread(int fd, void *buf, int len)
141 {
142 /* aligned read.
143 * On devices with a 4K sector size, we need to read
144 * the full sector and copy relevant bits into
145 * the buffer
146 */
147 int bsize;
148 char *b;
149 int n;
150 if (ioctl(fd, BLKSSZGET, &bsize) != 0 ||
151 bsize <= len)
152 return read(fd, buf, len);
153 if (bsize > 4096)
154 return -1;
155 b = (char*)(((long)(abuf+4096))&~4095UL);
156
157 n = read(fd, b, bsize);
158 if (n <= 0)
159 return n;
160 lseek(fd, len - n, 1);
161 if (n > len)
162 n = len;
163 memcpy(buf, b, n);
164 return n;
165 }
166
167 static int awrite(int fd, void *buf, int len)
168 {
169 /* aligned write.
170 * On devices with a 4K sector size, we need to write
171 * the full sector. We pre-read if the sector is larger
172 * than the write.
173 * The address must be sector-aligned.
174 */
175 int bsize;
176 char *b;
177 int n;
178 if (ioctl(fd, BLKSSZGET, &bsize) != 0 ||
179 bsize <= len)
180 return write(fd, buf, len);
181 if (bsize > 4096)
182 return -1;
183 b = (char*)(((long)(abuf+4096))&~4095UL);
184
185 n = read(fd, b, bsize);
186 if (n <= 0)
187 return n;
188 lseek(fd, -n, 1);
189 memcpy(b, buf, len);
190 n = write(fd, b, bsize);
191 if (n <= 0)
192 return n;
193 lseek(fd, len - n, 1);
194 return len;
195 }
196
197 #ifndef MDASSEMBLE
198 static void examine_super1(struct supertype *st, char *homehost)
199 {
200 struct mdp_superblock_1 *sb = st->sb;
201 time_t atime;
202 unsigned int d;
203 int role;
204 int delta_extra = 0;
205 int i;
206 char *c;
207 int l = homehost ? strlen(homehost) : 0;
208 int layout;
209 unsigned long long sb_offset;
210
211 printf(" Magic : %08x\n", __le32_to_cpu(sb->magic));
212 printf(" Version : 1");
213 sb_offset = __le64_to_cpu(sb->super_offset);
214 if (sb_offset <= 4)
215 printf(".1\n");
216 else if (sb_offset <= 8)
217 printf(".2\n");
218 else
219 printf(".0\n");
220 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb->feature_map));
221 printf(" Array UUID : ");
222 for (i=0; i<16; i++) {
223 if ((i&3)==0 && i != 0) printf(":");
224 printf("%02x", sb->set_uuid[i]);
225 }
226 printf("\n");
227 printf(" Name : %.32s", sb->set_name);
228 if (l > 0 && l < 32 &&
229 sb->set_name[l] == ':' &&
230 strncmp(sb->set_name, homehost, l) == 0)
231 printf(" (local to host %s)", homehost);
232 printf("\n");
233 atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL;
234 printf(" Creation Time : %.24s\n", ctime(&atime));
235 c=map_num(pers, __le32_to_cpu(sb->level));
236 printf(" Raid Level : %s\n", c?c:"-unknown-");
237 printf(" Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks));
238 printf("\n");
239 printf(" Avail Dev Size : %llu%s\n",
240 (unsigned long long)__le64_to_cpu(sb->data_size),
241 human_size(__le64_to_cpu(sb->data_size)<<9));
242 if (__le32_to_cpu(sb->level) > 0) {
243 int ddsks=0;
244 switch(__le32_to_cpu(sb->level)) {
245 case 1: ddsks=1;break;
246 case 4:
247 case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break;
248 case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break;
249 case 10:
250 layout = __le32_to_cpu(sb->layout);
251 ddsks = __le32_to_cpu(sb->raid_disks)
252 / (layout&255) / ((layout>>8)&255);
253 }
254 if (ddsks)
255 printf(" Array Size : %llu%s\n",
256 ddsks*(unsigned long long)__le64_to_cpu(sb->size),
257 human_size(ddsks*__le64_to_cpu(sb->size)<<9));
258 if (sb->size != sb->data_size)
259 printf(" Used Dev Size : %llu%s\n",
260 (unsigned long long)__le64_to_cpu(sb->size),
261 human_size(__le64_to_cpu(sb->size)<<9));
262 }
263 if (sb->data_offset)
264 printf(" Data Offset : %llu sectors\n",
265 (unsigned long long)__le64_to_cpu(sb->data_offset));
266 printf(" Super Offset : %llu sectors\n",
267 (unsigned long long)__le64_to_cpu(sb->super_offset));
268 if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET)
269 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->recovery_offset));
270 printf(" State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean");
271 printf(" Device UUID : ");
272 for (i=0; i<16; i++) {
273 if ((i&3)==0 && i != 0) printf(":");
274 printf("%02x", sb->device_uuid[i]);
275 }
276 printf("\n");
277 printf("\n");
278 if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) {
279 printf("Internal Bitmap : %ld sectors from superblock\n",
280 (long)(int32_t)__le32_to_cpu(sb->bitmap_offset));
281 }
282 if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) {
283 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb->reshape_position)/2,
284 human_size(__le64_to_cpu(sb->reshape_position)<<9));
285 if (__le32_to_cpu(sb->delta_disks)) {
286 printf(" Delta Devices : %d", __le32_to_cpu(sb->delta_disks));
287 printf(" (%d->%d)\n",
288 __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks),
289 __le32_to_cpu(sb->raid_disks));
290 if ((int)__le32_to_cpu(sb->delta_disks) < 0)
291 delta_extra = -__le32_to_cpu(sb->delta_disks);
292 }
293 if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) {
294 c = map_num(pers, __le32_to_cpu(sb->new_level));
295 printf(" New Level : %s\n", c?c:"-unknown-");
296 }
297 if (__le32_to_cpu(sb->new_layout) != __le32_to_cpu(sb->layout)) {
298 if (__le32_to_cpu(sb->level) == 5) {
299 c = map_num(r5layout, __le32_to_cpu(sb->new_layout));
300 printf(" New Layout : %s\n", c?c:"-unknown-");
301 }
302 if (__le32_to_cpu(sb->level) == 6) {
303 c = map_num(r6layout, __le32_to_cpu(sb->new_layout));
304 printf(" New Layout : %s\n", c?c:"-unknown-");
305 }
306 if (__le32_to_cpu(sb->level) == 10) {
307 printf(" New Layout :");
308 print_r10_layout(__le32_to_cpu(sb->new_layout));
309 printf("\n");
310 }
311 }
312 if (__le32_to_cpu(sb->new_chunk) != __le32_to_cpu(sb->chunksize))
313 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb->new_chunk)/2);
314 printf("\n");
315 }
316 if (sb->devflags) {
317 printf(" Flags :");
318 if (sb->devflags & WriteMostly1)
319 printf(" write-mostly");
320 printf("\n");
321 }
322
323 atime = __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL;
324 printf(" Update Time : %.24s\n", ctime(&atime));
325
326 if (calc_sb_1_csum(sb) == sb->sb_csum)
327 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum));
328 else
329 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb->sb_csum),
330 __le32_to_cpu(calc_sb_1_csum(sb)));
331 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb->events));
332 printf("\n");
333 if (__le32_to_cpu(sb->level) == 5) {
334 c = map_num(r5layout, __le32_to_cpu(sb->layout));
335 printf(" Layout : %s\n", c?c:"-unknown-");
336 }
337 if (__le32_to_cpu(sb->level) == 6) {
338 c = map_num(r6layout, __le32_to_cpu(sb->layout));
339 printf(" Layout : %s\n", c?c:"-unknown-");
340 }
341 if (__le32_to_cpu(sb->level) == 10) {
342 int lo = __le32_to_cpu(sb->layout);
343 printf(" Layout :");
344 print_r10_layout(lo);
345 printf("\n");
346 }
347 switch(__le32_to_cpu(sb->level)) {
348 case 0:
349 case 4:
350 case 5:
351 case 6:
352 case 10:
353 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize)/2);
354 break;
355 case -1:
356 printf(" Rounding : %dK\n", __le32_to_cpu(sb->chunksize)/2);
357 break;
358 default: break;
359 }
360 printf("\n");
361 #if 0
362 /* This turns out to just be confusing */
363 printf(" Array Slot : %d (", __le32_to_cpu(sb->dev_number));
364 for (i= __le32_to_cpu(sb->max_dev); i> 0 ; i--)
365 if (__le16_to_cpu(sb->dev_roles[i-1]) != 0xffff)
366 break;
367 for (d=0; d < i; d++) {
368 int role = __le16_to_cpu(sb->dev_roles[d]);
369 if (d) printf(", ");
370 if (role == 0xffff) printf("empty");
371 else if(role == 0xfffe) printf("failed");
372 else printf("%d", role);
373 }
374 printf(")\n");
375 #endif
376 printf(" Device Role : ");
377 d = __le32_to_cpu(sb->dev_number);
378 if (d < __le32_to_cpu(sb->max_dev))
379 role = __le16_to_cpu(sb->dev_roles[d]);
380 else
381 role = 0xFFFF;
382 if (role >= 0xFFFE)
383 printf("spare\n");
384 else
385 printf("Active device %d\n", role);
386
387 printf(" Array State : ");
388 for (d=0; d<__le32_to_cpu(sb->raid_disks) + delta_extra; d++) {
389 int cnt = 0;
390 int me = 0;
391 unsigned int i;
392 for (i=0; i< __le32_to_cpu(sb->max_dev); i++) {
393 unsigned int role = __le16_to_cpu(sb->dev_roles[i]);
394 if (role == d) {
395 if (i == __le32_to_cpu(sb->dev_number))
396 me = 1;
397 cnt++;
398 }
399 }
400 if (cnt > 1) printf("?");
401 else if (cnt == 1) printf("A");
402 else printf (".");
403 }
404 #if 0
405 /* This is confusing too */
406 faulty = 0;
407 for (i=0; i< __le32_to_cpu(sb->max_dev); i++) {
408 int role = __le16_to_cpu(sb->dev_roles[i]);
409 if (role == 0xFFFE)
410 faulty++;
411 }
412 if (faulty) printf(" %d failed", faulty);
413 #endif
414 printf(" ('A' == active, '.' == missing)");
415 printf("\n");
416 }
417
418
419 static void brief_examine_super1(struct supertype *st, int verbose)
420 {
421 struct mdp_superblock_1 *sb = st->sb;
422 int i;
423 unsigned long long sb_offset;
424 char *nm;
425 char *c=map_num(pers, __le32_to_cpu(sb->level));
426
427 nm = strchr(sb->set_name, ':');
428 if (nm)
429 nm++;
430 else if (sb->set_name[0])
431 nm = sb->set_name;
432 else
433 nm = NULL;
434
435 printf("ARRAY%s%s", nm ? " /dev/md/":"", nm);
436 if (verbose && c)
437 printf(" level=%s", c);
438 sb_offset = __le64_to_cpu(sb->super_offset);
439 if (sb_offset <= 4)
440 printf(" metadata=1.1 ");
441 else if (sb_offset <= 8)
442 printf(" metadata=1.2 ");
443 else
444 printf(" metadata=1.0 ");
445 if (verbose)
446 printf("num-devices=%d ", __le32_to_cpu(sb->raid_disks));
447 printf("UUID=");
448 for (i=0; i<16; i++) {
449 if ((i&3)==0 && i != 0) printf(":");
450 printf("%02x", sb->set_uuid[i]);
451 }
452 if (sb->set_name[0])
453 printf(" name=%.32s", sb->set_name);
454 printf("\n");
455 }
456
457 static void export_examine_super1(struct supertype *st)
458 {
459 struct mdp_superblock_1 *sb = st->sb;
460 int i;
461 int len = 32;
462
463 printf("MD_LEVEL=%s\n", map_num(pers, __le32_to_cpu(sb->level)));
464 printf("MD_DEVICES=%d\n", __le32_to_cpu(sb->raid_disks));
465 for (i=0; i<32; i++)
466 if (sb->set_name[i] == '\n' ||
467 sb->set_name[i] == '\0') {
468 len = i;
469 break;
470 }
471 if (len)
472 printf("MD_NAME=%.*s\n", len, sb->set_name);
473 printf("MD_UUID=");
474 for (i=0; i<16; i++) {
475 if ((i&3)==0 && i != 0) printf(":");
476 printf("%02x", sb->set_uuid[i]);
477 }
478 printf("\n");
479 printf("MD_UPDATE_TIME=%llu\n",
480 __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL);
481 printf("MD_DEV_UUID=");
482 for (i=0; i<16; i++) {
483 if ((i&3)==0 && i != 0) printf(":");
484 printf("%02x", sb->device_uuid[i]);
485 }
486 printf("\n");
487 printf("MD_EVENTS=%llu\n",
488 (unsigned long long)__le64_to_cpu(sb->events));
489 }
490
491 static void detail_super1(struct supertype *st, char *homehost)
492 {
493 struct mdp_superblock_1 *sb = st->sb;
494 int i;
495 int l = homehost ? strlen(homehost) : 0;
496
497 printf(" Name : %.32s", sb->set_name);
498 if (l > 0 && l < 32 &&
499 sb->set_name[l] == ':' &&
500 strncmp(sb->set_name, homehost, l) == 0)
501 printf(" (local to host %s)", homehost);
502 printf("\n UUID : ");
503 for (i=0; i<16; i++) {
504 if ((i&3)==0 && i != 0) printf(":");
505 printf("%02x", sb->set_uuid[i]);
506 }
507 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb->events));
508 }
509
510 static void brief_detail_super1(struct supertype *st)
511 {
512 struct mdp_superblock_1 *sb = st->sb;
513 int i;
514
515 if (sb->set_name[0])
516 printf(" name=%.32s", sb->set_name);
517 printf(" UUID=");
518 for (i=0; i<16; i++) {
519 if ((i&3)==0 && i != 0) printf(":");
520 printf("%02x", sb->set_uuid[i]);
521 }
522 }
523
524 static void export_detail_super1(struct supertype *st)
525 {
526 struct mdp_superblock_1 *sb = st->sb;
527 int i;
528 int len = 32;
529
530 for (i=0; i<32; i++)
531 if (sb->set_name[i] == '\n' ||
532 sb->set_name[i] == '\0') {
533 len = i;
534 break;
535 }
536 if (len)
537 printf("MD_NAME=%.*s\n", len, sb->set_name);
538 }
539
540 #endif
541
542 static int match_home1(struct supertype *st, char *homehost)
543 {
544 struct mdp_superblock_1 *sb = st->sb;
545 int l = homehost ? strlen(homehost) : 0;
546
547 return (l > 0 && l < 32 &&
548 sb->set_name[l] == ':' &&
549 strncmp(sb->set_name, homehost, l) == 0);
550 }
551
552 static void uuid_from_super1(struct supertype *st, int uuid[4])
553 {
554 struct mdp_superblock_1 *super = st->sb;
555 char *cuuid = (char*)uuid;
556 int i;
557 for (i=0; i<16; i++)
558 cuuid[i] = super->set_uuid[i];
559 }
560
561 static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map)
562 {
563 struct mdp_superblock_1 *sb = st->sb;
564 int working = 0;
565 unsigned int i;
566 unsigned int role;
567 unsigned int map_disks = info->array.raid_disks;
568
569 memset(info, 0, sizeof(*info));
570 info->array.major_version = 1;
571 info->array.minor_version = st->minor_version;
572 info->array.patch_version = 0;
573 info->array.raid_disks = __le32_to_cpu(sb->raid_disks);
574 info->array.level = __le32_to_cpu(sb->level);
575 info->array.layout = __le32_to_cpu(sb->layout);
576 info->array.md_minor = -1;
577 info->array.ctime = __le64_to_cpu(sb->ctime);
578 info->array.utime = __le64_to_cpu(sb->utime);
579 info->array.chunk_size = __le32_to_cpu(sb->chunksize)*512;
580 info->array.state =
581 (__le64_to_cpu(sb->resync_offset) >= __le64_to_cpu(sb->size))
582 ? 1 : 0;
583
584 info->data_offset = __le64_to_cpu(sb->data_offset);
585 info->component_size = __le64_to_cpu(sb->size);
586
587 info->disk.major = 0;
588 info->disk.minor = 0;
589 info->disk.number = __le32_to_cpu(sb->dev_number);
590 if (__le32_to_cpu(sb->dev_number) >= __le32_to_cpu(sb->max_dev) ||
591 __le32_to_cpu(sb->max_dev) > 512)
592 role = 0xfffe;
593 else
594 role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]);
595
596 info->disk.raid_disk = -1;
597 switch(role) {
598 case 0xFFFF:
599 info->disk.state = 0; /* spare: not active, not sync, not faulty */
600 break;
601 case 0xFFFE:
602 info->disk.state = 1; /* faulty */
603 break;
604 default:
605 info->disk.state = 6; /* active and in sync */
606 info->disk.raid_disk = role;
607 }
608 info->events = __le64_to_cpu(sb->events);
609 sprintf(info->text_version, "1.%d", st->minor_version);
610 info->safe_mode_delay = 200;
611
612 memcpy(info->uuid, sb->set_uuid, 16);
613
614 strncpy(info->name, sb->set_name, 32);
615 info->name[32] = 0;
616
617 if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET))
618 info->recovery_start = __le32_to_cpu(sb->recovery_offset);
619 else
620 info->recovery_start = MaxSector;
621
622 if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) {
623 info->reshape_active = 1;
624 info->reshape_progress = __le64_to_cpu(sb->reshape_position);
625 info->new_level = __le32_to_cpu(sb->new_level);
626 info->delta_disks = __le32_to_cpu(sb->delta_disks);
627 info->new_layout = __le32_to_cpu(sb->new_layout);
628 info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9;
629 if (info->delta_disks < 0)
630 info->array.raid_disks -= info->delta_disks;
631 } else
632 info->reshape_active = 0;
633
634 if (map)
635 for (i=0; i<map_disks; i++)
636 map[i] = 0;
637 for (i = 0; i < __le32_to_cpu(sb->max_dev); i++) {
638 role = __le16_to_cpu(sb->dev_roles[i]);
639 if (/*role == 0xFFFF || */role < (unsigned) info->array.raid_disks) {
640 working++;
641 if (map && role < map_disks)
642 map[role] = 1;
643 }
644 }
645
646 info->array.working_disks = working;
647 }
648
649 static struct mdinfo *container_content1(struct supertype *st, char *subarray)
650 {
651 struct mdinfo *info;
652
653 if (subarray)
654 return NULL;
655
656 info = malloc(sizeof(*info));
657 getinfo_super1(st, info, NULL);
658 return info;
659 }
660
661 static int update_super1(struct supertype *st, struct mdinfo *info,
662 char *update,
663 char *devname, int verbose,
664 int uuid_set, char *homehost)
665 {
666 /* NOTE: for 'assemble' and 'force' we need to return non-zero
667 * if any change was made. For others, the return value is
668 * ignored.
669 */
670 int rv = 0;
671 struct mdp_superblock_1 *sb = st->sb;
672
673 if (strcmp(update, "force-one")==0) {
674 /* Not enough devices for a working array,
675 * so bring this one up-to-date
676 */
677 if (sb->events != __cpu_to_le64(info->events))
678 rv = 1;
679 sb->events = __cpu_to_le64(info->events);
680 } else if (strcmp(update, "force-array")==0) {
681 /* Degraded array and 'force' requests to
682 * maybe need to mark it 'clean'.
683 */
684 switch(__le32_to_cpu(sb->level)) {
685 case 5: case 4: case 6:
686 /* need to force clean */
687 if (sb->resync_offset != MaxSector)
688 rv = 1;
689 sb->resync_offset = MaxSector;
690 }
691 } else if (strcmp(update, "assemble")==0) {
692 int d = info->disk.number;
693 int want;
694 if (info->disk.state == 6)
695 want = info->disk.raid_disk;
696 else
697 want = 0xFFFF;
698 if (sb->dev_roles[d] != __cpu_to_le16(want)) {
699 sb->dev_roles[d] = __cpu_to_le16(want);
700 rv = 1;
701 }
702 if (info->reshape_active &&
703 sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) &&
704 info->delta_disks >= 0 &&
705 info->reshape_progress < __le64_to_cpu(sb->reshape_position)) {
706 sb->reshape_position = __cpu_to_le64(info->reshape_progress);
707 rv = 1;
708 }
709 if (info->reshape_active &&
710 sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) &&
711 info->delta_disks < 0 &&
712 info->reshape_progress > __le64_to_cpu(sb->reshape_position)) {
713 sb->reshape_position = __cpu_to_le64(info->reshape_progress);
714 rv = 1;
715 }
716 } else if (strcmp(update, "linear-grow-new") == 0) {
717 unsigned int i;
718 int rfd, fd;
719 unsigned int max = __le32_to_cpu(sb->max_dev);
720
721 for (i=0 ; i < max ; i++)
722 if (__le16_to_cpu(sb->dev_roles[i]) >= 0xfffe)
723 break;
724 sb->dev_number = __cpu_to_le32(i);
725 info->disk.number = i;
726 if (max >= __le32_to_cpu(sb->max_dev))
727 sb->max_dev = __cpu_to_le32(max+1);
728
729 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
730 read(rfd, sb->device_uuid, 16) != 16) {
731 __u32 r[4] = {random(), random(), random(), random()};
732 memcpy(sb->device_uuid, r, 16);
733 }
734 if (rfd >= 0)
735 close(rfd);
736
737 sb->dev_roles[i] =
738 __cpu_to_le16(info->disk.raid_disk);
739
740 fd = open(devname, O_RDONLY);
741 if (fd >= 0) {
742 unsigned long long ds;
743 get_dev_size(fd, devname, &ds);
744 close(fd);
745 ds >>= 9;
746 if (__le64_to_cpu(sb->super_offset) <
747 __le64_to_cpu(sb->data_offset)) {
748 sb->data_size = __cpu_to_le64(
749 ds - __le64_to_cpu(sb->data_offset));
750 } else {
751 ds -= 8*2;
752 ds &= ~(unsigned long long)(4*2-1);
753 sb->super_offset = __cpu_to_le64(ds);
754 sb->data_size = __cpu_to_le64(
755 ds - __le64_to_cpu(sb->data_offset));
756 }
757 }
758 } else if (strcmp(update, "linear-grow-update") == 0) {
759 sb->raid_disks = __cpu_to_le32(info->array.raid_disks);
760 sb->dev_roles[info->disk.number] =
761 __cpu_to_le16(info->disk.raid_disk);
762 } else if (strcmp(update, "resync") == 0) {
763 /* make sure resync happens */
764 sb->resync_offset = 0ULL;
765 } else if (strcmp(update, "uuid") == 0) {
766 copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid);
767
768 if (__le32_to_cpu(sb->feature_map)&MD_FEATURE_BITMAP_OFFSET) {
769 struct bitmap_super_s *bm;
770 bm = (struct bitmap_super_s*)(st->sb+1024);
771 memcpy(bm->uuid, sb->set_uuid, 16);
772 }
773 } else if (strcmp(update, "no-bitmap") == 0) {
774 sb->feature_map &= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET);
775 } else if (strcmp(update, "homehost") == 0 &&
776 homehost) {
777 char *c;
778 update = "name";
779 c = strchr(sb->set_name, ':');
780 if (c)
781 strncpy(info->name, c+1, 31 - (c-sb->set_name));
782 else
783 strncpy(info->name, sb->set_name, 32);
784 info->name[32] = 0;
785 } else if (strcmp(update, "name") == 0) {
786 if (info->name[0] == 0)
787 sprintf(info->name, "%d", info->array.md_minor);
788 memset(sb->set_name, 0, sizeof(sb->set_name));
789 if (homehost &&
790 strchr(info->name, ':') == NULL &&
791 strlen(homehost)+1+strlen(info->name) < 32) {
792 strcpy(sb->set_name, homehost);
793 strcat(sb->set_name, ":");
794 strcat(sb->set_name, info->name);
795 } else
796 strcpy(sb->set_name, info->name);
797 } else if (strcmp(update, "devicesize") == 0 &&
798 __le64_to_cpu(sb->super_offset) <
799 __le64_to_cpu(sb->data_offset)) {
800 /* set data_size to device size less data_offset */
801 struct misc_dev_info *misc = (struct misc_dev_info*)
802 (st->sb + 1024 + 512);
803 printf("Size was %llu\n", (unsigned long long)
804 __le64_to_cpu(sb->data_size));
805 sb->data_size = __cpu_to_le64(
806 misc->device_size - __le64_to_cpu(sb->data_offset));
807 printf("Size is %llu\n", (unsigned long long)
808 __le64_to_cpu(sb->data_size));
809 } else if (strcmp(update, "_reshape_progress")==0)
810 sb->reshape_position = __cpu_to_le64(info->reshape_progress);
811 else
812 rv = -1;
813
814 sb->sb_csum = calc_sb_1_csum(sb);
815 return rv;
816 }
817
818 static int init_super1(struct supertype *st, mdu_array_info_t *info,
819 unsigned long long size, char *name, char *homehost, int *uuid)
820 {
821 struct mdp_superblock_1 *sb;
822 int spares;
823 int rfd;
824 char defname[10];
825
826 if (posix_memalign((void**)&sb, 512, (1024 + 512 +
827 sizeof(struct misc_dev_info))) != 0) {
828 fprintf(stderr, Name
829 ": %s could not allocate superblock\n", __func__);
830 return 0;
831 }
832 memset(sb, 0, 1024);
833
834 st->sb = sb;
835 if (info == NULL) {
836 /* zeroing superblock */
837 return 0;
838 }
839
840 spares = info->working_disks - info->active_disks;
841 if (info->raid_disks + spares > 384) {
842 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
843 info->raid_disks , spares, 384);
844 return 0;
845 }
846
847 sb->magic = __cpu_to_le32(MD_SB_MAGIC);
848 sb->major_version = __cpu_to_le32(1);
849 sb->feature_map = 0;
850 sb->pad0 = 0;
851
852 if (uuid)
853 copy_uuid(sb->set_uuid, uuid, super1.swapuuid);
854 else {
855 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
856 read(rfd, sb->set_uuid, 16) != 16) {
857 __u32 r[4] = {random(), random(), random(), random()};
858 memcpy(sb->set_uuid, r, 16);
859 }
860 if (rfd >= 0) close(rfd);
861 }
862
863 if (name == NULL || *name == 0) {
864 sprintf(defname, "%d", info->md_minor);
865 name = defname;
866 }
867 memset(sb->set_name, 0, 32);
868 if (homehost &&
869 strchr(name, ':')== NULL &&
870 strlen(homehost)+1+strlen(name) < 32) {
871 strcpy(sb->set_name, homehost);
872 strcat(sb->set_name, ":");
873 strcat(sb->set_name, name);
874 } else
875 strcpy(sb->set_name, name);
876
877 sb->ctime = __cpu_to_le64((unsigned long long)time(0));
878 sb->level = __cpu_to_le32(info->level);
879 sb->layout = __cpu_to_le32(info->layout);
880 sb->size = __cpu_to_le64(size*2ULL);
881 sb->chunksize = __cpu_to_le32(info->chunk_size>>9);
882 sb->raid_disks = __cpu_to_le32(info->raid_disks);
883
884 sb->data_offset = __cpu_to_le64(0);
885 sb->data_size = __cpu_to_le64(0);
886 sb->super_offset = __cpu_to_le64(0);
887 sb->recovery_offset = __cpu_to_le64(0);
888
889 sb->utime = sb->ctime;
890 sb->events = __cpu_to_le64(1);
891 if (info->state & (1<<MD_SB_CLEAN))
892 sb->resync_offset = MaxSector;
893 else
894 sb->resync_offset = 0;
895 sb->max_dev = __cpu_to_le32((1024- sizeof(struct mdp_superblock_1))/
896 sizeof(sb->dev_roles[0]));
897 memset(sb->pad3, 0, sizeof(sb->pad3));
898
899 memset(sb->dev_roles, 0xff, 1024 - sizeof(struct mdp_superblock_1));
900
901 return 1;
902 }
903
904 struct devinfo {
905 int fd;
906 char *devname;
907 mdu_disk_info_t disk;
908 struct devinfo *next;
909 };
910 #ifndef MDASSEMBLE
911 /* Add a device to the superblock being created */
912 static int add_to_super1(struct supertype *st, mdu_disk_info_t *dk,
913 int fd, char *devname)
914 {
915 struct mdp_superblock_1 *sb = st->sb;
916 __u16 *rp = sb->dev_roles + dk->number;
917 struct devinfo *di, **dip;
918
919 if ((dk->state & 6) == 6) /* active, sync */
920 *rp = __cpu_to_le16(dk->raid_disk);
921 else if ((dk->state & ~2) == 0) /* active or idle -> spare */
922 *rp = 0xffff;
923 else
924 *rp = 0xfffe;
925
926 if (dk->number >= (int)__le32_to_cpu(sb->max_dev) &&
927 __le32_to_cpu(sb->max_dev) < 384)
928 sb->max_dev = __cpu_to_le32(dk->number+1);
929
930 sb->dev_number = __cpu_to_le32(dk->number);
931 sb->sb_csum = calc_sb_1_csum(sb);
932
933 dip = (struct devinfo **)&st->info;
934 while (*dip)
935 dip = &(*dip)->next;
936 di = malloc(sizeof(struct devinfo));
937 di->fd = fd;
938 di->devname = devname;
939 di->disk = *dk;
940 di->next = NULL;
941 *dip = di;
942
943 return 0;
944 }
945 #endif
946
947 static void locate_bitmap1(struct supertype *st, int fd);
948
949 static int store_super1(struct supertype *st, int fd)
950 {
951 struct mdp_superblock_1 *sb = st->sb;
952 unsigned long long sb_offset;
953 int sbsize;
954 unsigned long long dsize;
955
956 if (!get_dev_size(fd, NULL, &dsize))
957 return 1;
958
959 dsize >>= 9;
960
961 if (dsize < 24)
962 return 2;
963
964 /*
965 * Calculate the position of the superblock.
966 * It is always aligned to a 4K boundary and
967 * depending on minor_version, it can be:
968 * 0: At least 8K, but less than 12K, from end of device
969 * 1: At start of device
970 * 2: 4K from start of device.
971 */
972 switch(st->minor_version) {
973 case 0:
974 sb_offset = dsize;
975 sb_offset -= 8*2;
976 sb_offset &= ~(4*2-1);
977 break;
978 case 1:
979 sb_offset = 0;
980 break;
981 case 2:
982 sb_offset = 4*2;
983 break;
984 default:
985 return -EINVAL;
986 }
987
988
989
990 if (sb_offset != __le64_to_cpu(sb->super_offset) &&
991 0 != __le64_to_cpu(sb->super_offset)
992 ) {
993 fprintf(stderr, Name ": internal error - sb_offset is wrong\n");
994 abort();
995 }
996
997 if (lseek64(fd, sb_offset << 9, 0)< 0LL)
998 return 3;
999
1000 sbsize = sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev);
1001 sbsize = (sbsize+511)&(~511UL);
1002
1003 if (awrite(fd, sb, sbsize) != sbsize)
1004 return 4;
1005
1006 if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) {
1007 struct bitmap_super_s *bm = (struct bitmap_super_s*)
1008 (((char*)sb)+1024);
1009 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) {
1010 locate_bitmap1(st, fd);
1011 if (awrite(fd, bm, sizeof(*bm)) !=
1012 sizeof(*bm))
1013 return 5;
1014 }
1015 }
1016 fsync(fd);
1017 return 0;
1018 }
1019
1020 static int load_super1(struct supertype *st, int fd, char *devname);
1021
1022 static unsigned long choose_bm_space(unsigned long devsize)
1023 {
1024 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
1025 * if bigger than 200Gig, save 128k
1026 * NOTE: result must be multiple of 4K else bad things happen
1027 * on 4K-sector devices.
1028 */
1029 if (devsize < 64*2) return 0;
1030 if (devsize - 64*2 >= 200*1024*1024*2)
1031 return 128*2;
1032 if (devsize - 4*2 > 8*1024*1024*2)
1033 return 64*2;
1034 return 4*2;
1035 }
1036
1037 static void free_super1(struct supertype *st);
1038
1039 #ifndef MDASSEMBLE
1040 static int write_init_super1(struct supertype *st)
1041 {
1042 struct mdp_superblock_1 *sb = st->sb;
1043 struct supertype *refst;
1044 int rfd;
1045 int rv = 0;
1046 unsigned long long bm_space;
1047 unsigned long long reserved;
1048 struct devinfo *di;
1049 unsigned long long dsize, array_size;
1050 unsigned long long sb_offset;
1051
1052 for (di = st->info; di && ! rv ; di = di->next) {
1053 if (di->disk.state == 1)
1054 continue;
1055 if (di->fd < 0)
1056 continue;
1057
1058 while (Kill(di->devname, NULL, 0, 1, 1) == 0)
1059 ;
1060
1061 sb->dev_number = __cpu_to_le32(di->disk.number);
1062 if (di->disk.state & (1<<MD_DISK_WRITEMOSTLY))
1063 sb->devflags |= __cpu_to_le32(WriteMostly1);
1064
1065 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
1066 read(rfd, sb->device_uuid, 16) != 16) {
1067 __u32 r[4] = {random(), random(), random(), random()};
1068 memcpy(sb->device_uuid, r, 16);
1069 }
1070 if (rfd >= 0)
1071 close(rfd);
1072
1073 sb->events = 0;
1074
1075 refst = dup_super(st);
1076 if (load_super1(refst, di->fd, NULL)==0) {
1077 struct mdp_superblock_1 *refsb = refst->sb;
1078
1079 memcpy(sb->device_uuid, refsb->device_uuid, 16);
1080 if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) {
1081 /* same array, so preserve events and
1082 * dev_number */
1083 sb->events = refsb->events;
1084 /* bugs in 2.6.17 and earlier mean the
1085 * dev_number chosen in Manage must be preserved
1086 */
1087 if (get_linux_version() >= 2006018)
1088 sb->dev_number = refsb->dev_number;
1089 }
1090 free_super1(refst);
1091 }
1092 free(refst);
1093
1094 if (!get_dev_size(di->fd, NULL, &dsize))
1095 return 1;
1096 dsize >>= 9;
1097
1098 if (dsize < 24) {
1099 close(di->fd);
1100 return 2;
1101 }
1102
1103
1104 /*
1105 * Calculate the position of the superblock.
1106 * It is always aligned to a 4K boundary and
1107 * depending on minor_version, it can be:
1108 * 0: At least 8K, but less than 12K, from end of device
1109 * 1: At start of device
1110 * 2: 4K from start of device.
1111 * Depending on the array size, we might leave extra space
1112 * for a bitmap.
1113 */
1114 array_size = __le64_to_cpu(sb->size);
1115 /* work out how much space we left for a bitmap */
1116 bm_space = choose_bm_space(array_size);
1117
1118 switch(st->minor_version) {
1119 case 0:
1120 sb_offset = dsize;
1121 sb_offset -= 8*2;
1122 sb_offset &= ~(4*2-1);
1123 sb->super_offset = __cpu_to_le64(sb_offset);
1124 sb->data_offset = __cpu_to_le64(0);
1125 if (sb_offset < array_size + bm_space)
1126 bm_space = sb_offset - array_size;
1127 sb->data_size = __cpu_to_le64(sb_offset - bm_space);
1128 break;
1129 case 1:
1130 sb->super_offset = __cpu_to_le64(0);
1131 reserved = bm_space + 4*2;
1132 /* Try for multiple of 1Meg so it is nicely aligned */
1133 #define ONE_MEG (2*1024)
1134 reserved = ((reserved + ONE_MEG-1)/ONE_MEG) * ONE_MEG;
1135 if (reserved + __le64_to_cpu(sb->size) > dsize)
1136 reserved = dsize - __le64_to_cpu(sb->size);
1137 /* force 4K alignment */
1138 reserved &= ~7ULL;
1139
1140 sb->data_offset = __cpu_to_le64(reserved);
1141 sb->data_size = __cpu_to_le64(dsize - reserved);
1142 break;
1143 case 2:
1144 sb_offset = 4*2;
1145 sb->super_offset = __cpu_to_le64(4*2);
1146 if (4*2 + 4*2 + bm_space + __le64_to_cpu(sb->size)
1147 > dsize)
1148 bm_space = dsize - __le64_to_cpu(sb->size)
1149 - 4*2 - 4*2;
1150
1151 reserved = bm_space + 4*2 + 4*2;
1152 /* Try for multiple of 1Meg so it is nicely aligned */
1153 #define ONE_MEG (2*1024)
1154 reserved = ((reserved + ONE_MEG-1)/ONE_MEG) * ONE_MEG;
1155 if (reserved + __le64_to_cpu(sb->size) > dsize)
1156 reserved = dsize - __le64_to_cpu(sb->size);
1157 /* force 4K alignment */
1158 reserved &= ~7ULL;
1159
1160 sb->data_offset = __cpu_to_le64(reserved);
1161 sb->data_size = __cpu_to_le64(dsize - reserved);
1162 break;
1163 default:
1164 return -EINVAL;
1165 }
1166
1167
1168 sb->sb_csum = calc_sb_1_csum(sb);
1169 rv = store_super1(st, di->fd);
1170 if (rv)
1171 fprintf(stderr,
1172 Name ": failed to write superblock to %s\n",
1173 di->devname);
1174
1175 if (rv == 0 && (__le32_to_cpu(sb->feature_map) & 1))
1176 rv = st->ss->write_bitmap(st, di->fd);
1177 close(di->fd);
1178 di->fd = -1;
1179 }
1180 return rv;
1181 }
1182 #endif
1183
1184 static int compare_super1(struct supertype *st, struct supertype *tst)
1185 {
1186 /*
1187 * return:
1188 * 0 same, or first was empty, and second was copied
1189 * 1 second had wrong number
1190 * 2 wrong uuid
1191 * 3 wrong other info
1192 */
1193 struct mdp_superblock_1 *first = st->sb;
1194 struct mdp_superblock_1 *second = tst->sb;
1195
1196 if (second->magic != __cpu_to_le32(MD_SB_MAGIC))
1197 return 1;
1198 if (second->major_version != __cpu_to_le32(1))
1199 return 1;
1200
1201 if (!first) {
1202 if (posix_memalign((void**)&first, 512,
1203 1024 + 512 +
1204 sizeof(struct misc_dev_info)) != 0) {
1205 fprintf(stderr, Name
1206 ": %s could not allocate superblock\n", __func__);
1207 return 1;
1208 }
1209 memcpy(first, second, 1024 + 512 +
1210 sizeof(struct misc_dev_info));
1211 st->sb = first;
1212 return 0;
1213 }
1214 if (memcmp(first->set_uuid, second->set_uuid, 16)!= 0)
1215 return 2;
1216
1217 if (first->ctime != second->ctime ||
1218 first->level != second->level ||
1219 first->layout != second->layout ||
1220 first->size != second->size ||
1221 first->chunksize != second->chunksize ||
1222 first->raid_disks != second->raid_disks)
1223 return 3;
1224 return 0;
1225 }
1226
1227 static int load_super1(struct supertype *st, int fd, char *devname)
1228 {
1229 unsigned long long dsize;
1230 unsigned long long sb_offset;
1231 struct mdp_superblock_1 *super;
1232 int uuid[4];
1233 struct bitmap_super_s *bsb;
1234 struct misc_dev_info *misc;
1235
1236 free_super1(st);
1237
1238 if (st->ss == NULL || st->minor_version == -1) {
1239 int bestvers = -1;
1240 struct supertype tst;
1241 __u64 bestctime = 0;
1242 /* guess... choose latest ctime */
1243 memset(&tst, 0, sizeof(tst));
1244 tst.ss = &super1;
1245 for (tst.minor_version = 0; tst.minor_version <= 2 ; tst.minor_version++) {
1246 switch(load_super1(&tst, fd, devname)) {
1247 case 0: super = tst.sb;
1248 if (bestvers == -1 ||
1249 bestctime < __le64_to_cpu(super->ctime)) {
1250 bestvers = tst.minor_version;
1251 bestctime = __le64_to_cpu(super->ctime);
1252 }
1253 free(super);
1254 tst.sb = NULL;
1255 break;
1256 case 1: return 1; /*bad device */
1257 case 2: break; /* bad, try next */
1258 }
1259 }
1260 if (bestvers != -1) {
1261 int rv;
1262 tst.minor_version = bestvers;
1263 tst.ss = &super1;
1264 tst.max_devs = 384;
1265 rv = load_super1(&tst, fd, devname);
1266 if (rv == 0)
1267 *st = tst;
1268 return rv;
1269 }
1270 return 2;
1271 }
1272 if (!get_dev_size(fd, devname, &dsize))
1273 return 1;
1274 dsize >>= 9;
1275
1276 if (dsize < 24) {
1277 if (devname)
1278 fprintf(stderr, Name ": %s is too small for md: size is %llu sectors.\n",
1279 devname, dsize);
1280 return 1;
1281 }
1282
1283 /*
1284 * Calculate the position of the superblock.
1285 * It is always aligned to a 4K boundary and
1286 * depending on minor_version, it can be:
1287 * 0: At least 8K, but less than 12K, from end of device
1288 * 1: At start of device
1289 * 2: 4K from start of device.
1290 */
1291 switch(st->minor_version) {
1292 case 0:
1293 sb_offset = dsize;
1294 sb_offset -= 8*2;
1295 sb_offset &= ~(4*2-1);
1296 break;
1297 case 1:
1298 sb_offset = 0;
1299 break;
1300 case 2:
1301 sb_offset = 4*2;
1302 break;
1303 default:
1304 return -EINVAL;
1305 }
1306
1307 ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
1308
1309
1310 if (lseek64(fd, sb_offset << 9, 0)< 0LL) {
1311 if (devname)
1312 fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
1313 devname, strerror(errno));
1314 return 1;
1315 }
1316
1317 if (posix_memalign((void**)&super, 512,
1318 1024 + 512 +
1319 sizeof(struct misc_dev_info)) != 0) {
1320 fprintf(stderr, Name ": %s could not allocate superblock\n",
1321 __func__);
1322 return 1;
1323 }
1324
1325 if (aread(fd, super, 1024) != 1024) {
1326 if (devname)
1327 fprintf(stderr, Name ": Cannot read superblock on %s\n",
1328 devname);
1329 free(super);
1330 return 1;
1331 }
1332
1333 if (__le32_to_cpu(super->magic) != MD_SB_MAGIC) {
1334 if (devname)
1335 fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
1336 devname, MD_SB_MAGIC, __le32_to_cpu(super->magic));
1337 free(super);
1338 return 2;
1339 }
1340
1341 if (__le32_to_cpu(super->major_version) != 1) {
1342 if (devname)
1343 fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
1344 devname, __le32_to_cpu(super->major_version));
1345 free(super);
1346 return 2;
1347 }
1348 if (__le64_to_cpu(super->super_offset) != sb_offset) {
1349 if (devname)
1350 fprintf(stderr, Name ": No superblock found on %s (super_offset is wrong)\n",
1351 devname);
1352 free(super);
1353 return 2;
1354 }
1355 st->sb = super;
1356
1357 bsb = (struct bitmap_super_s *)(((char*)super)+1024);
1358
1359 misc = (struct misc_dev_info*) (((char*)super)+1024+512);
1360 misc->device_size = dsize;
1361
1362 /* Now check on the bitmap superblock */
1363 if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0)
1364 return 0;
1365 /* Read the bitmap superblock and make sure it looks
1366 * valid. If it doesn't clear the bit. An --assemble --force
1367 * should get that written out.
1368 */
1369 locate_bitmap1(st, fd);
1370 if (aread(fd, ((char*)super)+1024, 512)
1371 != 512)
1372 goto no_bitmap;
1373
1374 uuid_from_super1(st, uuid);
1375 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
1376 memcmp(bsb->uuid, uuid, 16) != 0)
1377 goto no_bitmap;
1378 return 0;
1379
1380 no_bitmap:
1381 super->feature_map = __cpu_to_le32(__le32_to_cpu(super->feature_map) & ~1);
1382 return 0;
1383 }
1384
1385
1386 static struct supertype *match_metadata_desc1(char *arg)
1387 {
1388 struct supertype *st = malloc(sizeof(*st));
1389 if (!st) return st;
1390
1391 memset(st, 0, sizeof(*st));
1392 st->container_dev = NoMdDev;
1393 st->ss = &super1;
1394 st->max_devs = 384;
1395 st->sb = NULL;
1396 /* leading zeros can be safely ignored. --detail generates them. */
1397 while (*arg == '0')
1398 arg++;
1399 if (strcmp(arg, "1.0") == 0 ||
1400 strcmp(arg, "1.00") == 0) {
1401 st->minor_version = 0;
1402 return st;
1403 }
1404 if (strcmp(arg, "1.1") == 0 ||
1405 strcmp(arg, "1.01") == 0
1406 ) {
1407 st->minor_version = 1;
1408 return st;
1409 }
1410 if (strcmp(arg, "1.2") == 0 ||
1411 #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */
1412 strcmp(arg, "default") == 0 ||
1413 #endif /* DEFAULT_OLD_METADATA */
1414 strcmp(arg, "1.02") == 0) {
1415 st->minor_version = 2;
1416 return st;
1417 }
1418 if (strcmp(arg, "1") == 0 ||
1419 strcmp(arg, "default") == 0) {
1420 st->minor_version = -1;
1421 return st;
1422 }
1423
1424 free(st);
1425 return NULL;
1426 }
1427
1428 /* find available size on device with this devsize, using
1429 * superblock type st, and reserving 'reserve' sectors for
1430 * a possible bitmap
1431 */
1432 static __u64 avail_size1(struct supertype *st, __u64 devsize)
1433 {
1434 struct mdp_superblock_1 *super = st->sb;
1435 if (devsize < 24)
1436 return 0;
1437
1438 if (super == NULL)
1439 /* creating: allow suitable space for bitmap */
1440 devsize -= choose_bm_space(devsize);
1441 #ifndef MDASSEMBLE
1442 else if (__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) {
1443 /* hot-add. allow for actual size of bitmap */
1444 struct bitmap_super_s *bsb;
1445 bsb = (struct bitmap_super_s *)(((char*)super)+1024);
1446 devsize -= bitmap_sectors(bsb);
1447 }
1448 #endif
1449
1450 if (st->minor_version < 0)
1451 /* not specified, so time to set default */
1452 st->minor_version = 2;
1453 if (super == NULL && st->minor_version > 0) {
1454 /* haven't committed to a size yet, so allow some
1455 * slack for alignment of data_offset.
1456 * We haven't access to device details so allow
1457 * 1 Meg if bigger than 1Gig
1458 */
1459 if (devsize > 1024*1024*2)
1460 devsize -= 1024*2;
1461 }
1462 switch(st->minor_version) {
1463 case 0:
1464 /* at end */
1465 return ((devsize - 8*2 ) & ~(4*2-1));
1466 case 1:
1467 /* at start, 4K for superblock and possible bitmap */
1468 return devsize - 4*2;
1469 case 2:
1470 /* 4k from start, 4K for superblock and possible bitmap */
1471 return devsize - (4+4)*2;
1472 }
1473 return 0;
1474 }
1475
1476 static int
1477 add_internal_bitmap1(struct supertype *st,
1478 int *chunkp, int delay, int write_behind,
1479 unsigned long long size,
1480 int may_change, int major)
1481 {
1482 /*
1483 * If not may_change, then this is a 'Grow', and the bitmap
1484 * must fit after the superblock.
1485 * If may_change, then this is create, and we can put the bitmap
1486 * before the superblock if we like, or may move the start.
1487 * If !may_change, the bitmap MUST live at offset of 1K, until
1488 * we get a sysfs interface.
1489 *
1490 * size is in sectors, chunk is in bytes !!!
1491 */
1492
1493 unsigned long long bits;
1494 unsigned long long max_bits;
1495 unsigned long long min_chunk;
1496 long offset;
1497 unsigned long long chunk = *chunkp;
1498 int room = 0;
1499 struct mdp_superblock_1 *sb = st->sb;
1500 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + 1024);
1501
1502 switch(st->minor_version) {
1503 case 0:
1504 /* either 3K after the superblock (when hot-add),
1505 * or some amount of space before.
1506 */
1507 if (may_change) {
1508 /* We are creating array, so we *know* how much room has
1509 * been left.
1510 */
1511 offset = 0;
1512 room = choose_bm_space(__le64_to_cpu(sb->size));
1513 } else {
1514 room = __le64_to_cpu(sb->super_offset)
1515 - __le64_to_cpu(sb->data_offset)
1516 - __le64_to_cpu(sb->data_size);
1517 /* remove '1 ||' when we can set offset via sysfs */
1518 if (1 || (room < 3*2 &&
1519 __le32_to_cpu(sb->max_dev) <= 384)) {
1520 room = 3*2;
1521 offset = 1*2;
1522 } else {
1523 offset = 0; /* means movable offset */
1524 }
1525 }
1526 break;
1527 case 1:
1528 case 2: /* between superblock and data */
1529 if (may_change) {
1530 offset = 4*2;
1531 room = choose_bm_space(__le64_to_cpu(sb->size));
1532 } else {
1533 room = __le64_to_cpu(sb->data_offset)
1534 - __le64_to_cpu(sb->super_offset);
1535 if (1 || __le32_to_cpu(sb->max_dev) <= 384) {
1536 room -= 2;
1537 offset = 2;
1538 } else {
1539 room -= 4*2;
1540 offset = 4*2;
1541 }
1542 }
1543 break;
1544 default:
1545 return 0;
1546 }
1547
1548 if (chunk == UnSet && room > 128*2)
1549 /* Limit to 128K of bitmap when chunk size not requested */
1550 room = 128*2;
1551
1552 max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8;
1553
1554 min_chunk = 4096; /* sub-page chunks don't work yet.. */
1555 bits = (size*512)/min_chunk +1;
1556 while (bits > max_bits) {
1557 min_chunk *= 2;
1558 bits = (bits+1)/2;
1559 }
1560 if (chunk == UnSet) {
1561 /* For practical purpose, 64Meg is a good
1562 * default chunk size for internal bitmaps.
1563 */
1564 chunk = min_chunk;
1565 if (chunk < 64*1024*1024)
1566 chunk = 64*1024*1024;
1567 } else if (chunk < min_chunk)
1568 return 0; /* chunk size too small */
1569 if (chunk == 0) /* rounding problem */
1570 return 0;
1571
1572 if (offset == 0) {
1573 /* start bitmap on a 4K boundary with enough space for
1574 * the bitmap
1575 */
1576 bits = (size*512) / chunk + 1;
1577 room = ((bits+7)/8 + sizeof(bitmap_super_t) +4095)/4096;
1578 room *= 8; /* convert 4K blocks to sectors */
1579 offset = -room;
1580 }
1581
1582 sb->bitmap_offset = __cpu_to_le32(offset);
1583
1584 sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | 1);
1585 memset(bms, 0, sizeof(*bms));
1586 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
1587 bms->version = __cpu_to_le32(major);
1588 uuid_from_super1(st, (int*)bms->uuid);
1589 bms->chunksize = __cpu_to_le32(chunk);
1590 bms->daemon_sleep = __cpu_to_le32(delay);
1591 bms->sync_size = __cpu_to_le64(size);
1592 bms->write_behind = __cpu_to_le32(write_behind);
1593
1594 *chunkp = chunk;
1595 return 1;
1596 }
1597
1598
1599 static void locate_bitmap1(struct supertype *st, int fd)
1600 {
1601 unsigned long long offset;
1602 struct mdp_superblock_1 *sb;
1603 int mustfree = 0;
1604
1605 if (!st->sb) {
1606 if (st->ss->load_super(st, fd, NULL))
1607 return; /* no error I hope... */
1608 mustfree = 1;
1609 }
1610 sb = st->sb;
1611
1612 offset = __le64_to_cpu(sb->super_offset);
1613 offset += (int32_t) __le32_to_cpu(sb->bitmap_offset);
1614 if (mustfree)
1615 free(sb);
1616 lseek64(fd, offset<<9, 0);
1617 }
1618
1619 static int write_bitmap1(struct supertype *st, int fd)
1620 {
1621 struct mdp_superblock_1 *sb = st->sb;
1622 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+1024);
1623 int rv = 0;
1624
1625 int towrite, n;
1626 char *buf = (char*)(((long)(abuf+4096))&~4095UL);
1627
1628 locate_bitmap1(st, fd);
1629
1630 memset(buf, 0xff, 4096);
1631 memcpy(buf, ((char*)sb)+1024, sizeof(bitmap_super_t));
1632
1633 towrite = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9);
1634 towrite = (towrite+7) >> 3; /* bits to bytes */
1635 towrite += sizeof(bitmap_super_t);
1636 towrite = ROUND_UP(towrite, 512);
1637 while (towrite > 0) {
1638 n = towrite;
1639 if (n > 4096)
1640 n = 4096;
1641 n = write(fd, buf, n);
1642 if (n > 0)
1643 towrite -= n;
1644 else
1645 break;
1646 memset(buf, 0xff, 4096);
1647 }
1648 fsync(fd);
1649 if (towrite)
1650 rv = -2;
1651
1652 return rv;
1653 }
1654
1655 static void free_super1(struct supertype *st)
1656 {
1657 if (st->sb)
1658 free(st->sb);
1659 while (st->info) {
1660 struct devinfo *di = st->info;
1661 st->info = di->next;
1662 if (di->fd >= 0)
1663 close(di->fd);
1664 free(di);
1665 }
1666 st->sb = NULL;
1667 }
1668
1669 #ifndef MDASSEMBLE
1670 static int validate_geometry1(struct supertype *st, int level,
1671 int layout, int raiddisks,
1672 int *chunk, unsigned long long size,
1673 char *subdev, unsigned long long *freesize,
1674 int verbose)
1675 {
1676 unsigned long long ldsize;
1677 int fd;
1678
1679 if (level == LEVEL_CONTAINER) {
1680 if (verbose)
1681 fprintf(stderr, Name ": 1.x metadata does not support containers\n");
1682 return 0;
1683 }
1684 if (chunk && *chunk == UnSet)
1685 *chunk = DEFAULT_CHUNK;
1686
1687 if (!subdev)
1688 return 1;
1689
1690 fd = open(subdev, O_RDONLY|O_EXCL, 0);
1691 if (fd < 0) {
1692 if (verbose)
1693 fprintf(stderr, Name ": super1.x cannot open %s: %s\n",
1694 subdev, strerror(errno));
1695 return 0;
1696 }
1697
1698 if (!get_dev_size(fd, subdev, &ldsize)) {
1699 close(fd);
1700 return 0;
1701 }
1702 close(fd);
1703
1704 *freesize = avail_size1(st, ldsize >> 9);
1705 return 1;
1706 }
1707 #endif /* MDASSEMBLE */
1708
1709 struct superswitch super1 = {
1710 #ifndef MDASSEMBLE
1711 .examine_super = examine_super1,
1712 .brief_examine_super = brief_examine_super1,
1713 .export_examine_super = export_examine_super1,
1714 .detail_super = detail_super1,
1715 .brief_detail_super = brief_detail_super1,
1716 .export_detail_super = export_detail_super1,
1717 .write_init_super = write_init_super1,
1718 .validate_geometry = validate_geometry1,
1719 .add_to_super = add_to_super1,
1720 #endif
1721 .match_home = match_home1,
1722 .uuid_from_super = uuid_from_super1,
1723 .getinfo_super = getinfo_super1,
1724 .container_content = container_content1,
1725 .update_super = update_super1,
1726 .init_super = init_super1,
1727 .store_super = store_super1,
1728 .compare_super = compare_super1,
1729 .load_super = load_super1,
1730 .match_metadata_desc = match_metadata_desc1,
1731 .avail_size = avail_size1,
1732 .add_internal_bitmap = add_internal_bitmap1,
1733 .locate_bitmap = locate_bitmap1,
1734 .write_bitmap = write_bitmap1,
1735 .free_super = free_super1,
1736 #if __BYTE_ORDER == BIG_ENDIAN
1737 .swapuuid = 0,
1738 #else
1739 .swapuuid = 1,
1740 #endif
1741 .name = "1.x",
1742 };
Unnamed repository; edit this file 'description' to name the repository.