]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super0.c
Allow --update=name to update the name during assembly.
[thirdparty/mdadm.git] / super0.c
1 /*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
4 * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
22 * Email: <neilb@cse.unsw.edu.au>
23 * Paper: Neil Brown
24 * School of Computer Science and Engineering
25 * The University of New South Wales
26 * Sydney, 2052
27 * Australia
28 */
29
30 #include "mdadm.h"
31 #include <openssl/sha.h> /* for SHA1 */
32 /*
33 * All handling for the 0.90.0 version superblock is in
34 * this file.
35 * This includes:
36 * - finding, loading, and writing the superblock.
37 * - initialising a new superblock
38 * - printing the superblock for --examine
39 * - printing part of the superblock for --detail
40 * .. other stuff
41 */
42
43
44 static unsigned long calc_sb0_csum(mdp_super_t *super)
45 {
46 unsigned long csum = super->sb_csum;
47 unsigned long newcsum;
48 super->sb_csum= 0 ;
49 newcsum = calc_csum(super, MD_SB_BYTES);
50 super->sb_csum = csum;
51 return newcsum;
52 }
53
54
55 void super0_swap_endian(struct mdp_superblock_s *sb)
56 {
57 /* as super0 superblocks are host-endian, it is sometimes
58 * useful to be able to swap the endianness
59 * as (almost) everything is u32's we byte-swap every 4byte
60 * number.
61 * We then also have to swap the events_hi and events_lo
62 */
63 char *sbc = (char *)sb;
64 __u32 t32;
65 int i;
66
67 for (i=0; i < MD_SB_BYTES ; i+=4) {
68 char t = sbc[i];
69 sbc[i] = sbc[i+3];
70 sbc[i+3] = t;
71 t=sbc[i+1];
72 sbc[i+1]=sbc[i+2];
73 sbc[i+2]=t;
74 }
75 t32 = sb->events_hi;
76 sb->events_hi = sb->events_lo;
77 sb->events_lo = t32;
78
79 t32 = sb->cp_events_hi;
80 sb->cp_events_hi = sb->cp_events_lo;
81 sb->cp_events_lo = t32;
82
83 }
84
85 #ifndef MDASSEMBLE
86
87 static void examine_super0(void *sbv, char *homehost)
88 {
89 mdp_super_t *sb = sbv;
90 time_t atime;
91 int d;
92 char *c;
93
94 printf(" Magic : %08x\n", sb->md_magic);
95 printf(" Version : %02d.%02d.%02d\n", sb->major_version, sb->minor_version,
96 sb->patch_version);
97 if (sb->minor_version >= 90) {
98 printf(" UUID : %08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
99 sb->set_uuid2, sb->set_uuid3);
100 if (homehost) {
101 unsigned char *hash = SHA1((unsigned char *)homehost,
102 strlen(homehost),
103 NULL);
104 if (memcmp(&sb->set_uuid2, hash, 8)==0)
105 printf(" (local to host %s)", homehost);
106 }
107 printf("\n");
108 } else
109 printf(" UUID : %08x\n", sb->set_uuid0);
110
111 atime = sb->ctime;
112 printf(" Creation Time : %.24s\n", ctime(&atime));
113 c=map_num(pers, sb->level);
114 printf(" Raid Level : %s\n", c?c:"-unknown-");
115 if ((int)sb->level >= 0) {
116 int ddsks=0;
117 printf(" Device Size : %d%s\n", sb->size, human_size((long long)sb->size<<10));
118 switch(sb->level) {
119 case 1: ddsks=1;break;
120 case 4:
121 case 5: ddsks = sb->raid_disks-1; break;
122 case 6: ddsks = sb->raid_disks-2; break;
123 case 10: ddsks = sb->raid_disks / (sb->layout&255) / ((sb->layout>>8)&255);
124 }
125 if (ddsks)
126 printf(" Array Size : %llu%s\n", (unsigned long long)ddsks * sb->size,
127 human_size(ddsks*(long long)sb->size<<10));
128 }
129 printf(" Raid Devices : %d\n", sb->raid_disks);
130 printf(" Total Devices : %d\n", sb->nr_disks);
131 printf("Preferred Minor : %d\n", sb->md_minor);
132 printf("\n");
133 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
134 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)sb->reshape_position/2, human_size((long long)sb->reshape_position<<9));
135 if (sb->delta_disks) {
136 printf(" Delta Devices : %d", sb->delta_disks);
137 if (sb->delta_disks)
138 printf(" (%d->%d)\n", sb->raid_disks-sb->delta_disks, sb->raid_disks);
139 else
140 printf(" (%d->%d)\n", sb->raid_disks, sb->raid_disks+sb->delta_disks);
141 }
142 if (sb->new_level != sb->level) {
143 c = map_num(pers, sb->new_level);
144 printf(" New Level : %s\n", c?c:"-unknown-");
145 }
146 if (sb->new_layout != sb->layout) {
147 if (sb->level == 5) {
148 c = map_num(r5layout, sb->new_layout);
149 printf(" New Layout : %s\n", c?c:"-unknown-");
150 }
151 if (sb->level == 10) {
152 printf(" New Layout : near=%d, %s=%d\n",
153 sb->new_layout&255,
154 (sb->new_layout&0x10000)?"offset":"far",
155 (sb->new_layout>>8)&255);
156 }
157 }
158 if (sb->new_chunk != sb->chunk_size)
159 printf(" New Chunksize : %d\n", sb->new_chunk);
160 printf("\n");
161 }
162 atime = sb->utime;
163 printf(" Update Time : %.24s\n", ctime(&atime));
164 printf(" State : %s\n",
165 (sb->state&(1<<MD_SB_CLEAN))?"clean":"active");
166 if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
167 printf("Internal Bitmap : present\n");
168 printf(" Active Devices : %d\n", sb->active_disks);
169 printf("Working Devices : %d\n", sb->working_disks);
170 printf(" Failed Devices : %d\n", sb->failed_disks);
171 printf(" Spare Devices : %d\n", sb->spare_disks);
172 if (calc_sb0_csum(sb) == sb->sb_csum)
173 printf(" Checksum : %x - correct\n", sb->sb_csum);
174 else
175 printf(" Checksum : %x - expected %lx\n", sb->sb_csum, calc_sb0_csum(sb));
176 printf(" Events : %d.%d\n", sb->events_hi, sb->events_lo);
177 printf("\n");
178 if (sb->level == 5) {
179 c = map_num(r5layout, sb->layout);
180 printf(" Layout : %s\n", c?c:"-unknown-");
181 }
182 if (sb->level == 10) {
183 printf(" Layout : near=%d, %s=%d\n",
184 sb->layout&255,
185 (sb->layout&0x10000)?"offset":"far",
186 (sb->layout>>8)&255);
187 }
188 switch(sb->level) {
189 case 0:
190 case 4:
191 case 5:
192 printf(" Chunk Size : %dK\n", sb->chunk_size/1024);
193 break;
194 case -1:
195 printf(" Rounding : %dK\n", sb->chunk_size/1024);
196 break;
197 default: break;
198 }
199 printf("\n");
200 printf(" Number Major Minor RaidDevice State\n");
201 for (d= -1; d<(signed int)(sb->raid_disks+sb->spare_disks); d++) {
202 mdp_disk_t *dp;
203 char *dv;
204 char nb[5];
205 int wonly;
206 if (d>=0) dp = &sb->disks[d];
207 else dp = &sb->this_disk;
208 snprintf(nb, sizeof(nb), "%4d", d);
209 printf("%4s %5d %5d %5d %5d ", d < 0 ? "this" : nb,
210 dp->number, dp->major, dp->minor, dp->raid_disk);
211 wonly = dp->state & (1<<MD_DISK_WRITEMOSTLY);
212 dp->state &= ~(1<<MD_DISK_WRITEMOSTLY);
213 if (dp->state & (1<<MD_DISK_FAULTY)) printf(" faulty");
214 if (dp->state & (1<<MD_DISK_ACTIVE)) printf(" active");
215 if (dp->state & (1<<MD_DISK_SYNC)) printf(" sync");
216 if (dp->state & (1<<MD_DISK_REMOVED)) printf(" removed");
217 if (wonly) printf(" write-mostly");
218 if (dp->state == 0) printf(" spare");
219 if ((dv=map_dev(dp->major, dp->minor, 0)))
220 printf(" %s", dv);
221 printf("\n");
222 if (d == -1) printf("\n");
223 }
224 }
225
226 static void brief_examine_super0(void *sbv)
227 {
228 mdp_super_t *sb = sbv;
229 char *c=map_num(pers, sb->level);
230
231 printf("ARRAY %s level=%s num-devices=%d UUID=",
232 get_md_name(sb->md_minor),
233 c?c:"-unknown-", sb->raid_disks);
234 if (sb->minor_version >= 90)
235 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
236 sb->set_uuid2, sb->set_uuid3);
237 else
238 printf("%08x", sb->set_uuid0);
239 printf("\n");
240 }
241
242 static void detail_super0(void *sbv, char *homehost)
243 {
244 mdp_super_t *sb = sbv;
245 printf(" UUID : ");
246 if (sb->minor_version >= 90)
247 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
248 sb->set_uuid2, sb->set_uuid3);
249 else
250 printf("%08x", sb->set_uuid0);
251 if (homehost) {
252 unsigned char *hash = SHA1((unsigned char *)homehost,
253 strlen(homehost),
254 NULL);
255 if (memcmp(&sb->set_uuid2, hash, 8)==0)
256 printf(" (local to host %s)", homehost);
257 }
258 printf("\n Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
259 }
260
261 static void brief_detail_super0(void *sbv)
262 {
263 mdp_super_t *sb = sbv;
264 printf(" UUID=");
265 if (sb->minor_version >= 90)
266 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
267 sb->set_uuid2, sb->set_uuid3);
268 else
269 printf("%08x", sb->set_uuid0);
270 }
271 #endif
272 static void uuid_from_super0(int uuid[4], void * sbv)
273 {
274 mdp_super_t *super = sbv;
275 uuid[0] = super->set_uuid0;
276 if (super->minor_version >= 90) {
277 uuid[1] = super->set_uuid1;
278 uuid[2] = super->set_uuid2;
279 uuid[3] = super->set_uuid3;
280 } else {
281 uuid[1] = 0;
282 uuid[2] = 0;
283 uuid[3] = 0;
284 }
285 }
286
287 static void getinfo_super0(struct mdinfo *info, void *sbv)
288 {
289 mdp_super_t *sb = sbv;
290 int working = 0;
291 int i;
292
293 info->array.major_version = sb->major_version;
294 info->array.minor_version = sb->minor_version;
295 info->array.patch_version = sb->patch_version;
296 info->array.raid_disks = sb->raid_disks;
297 info->array.level = sb->level;
298 info->array.layout = sb->layout;
299 info->array.md_minor = sb->md_minor;
300 info->array.ctime = sb->ctime;
301 info->array.utime = sb->utime;
302 info->array.chunk_size = sb->chunk_size;
303 info->component_size = sb->size*2;
304
305 info->disk.state = sb->this_disk.state;
306 info->disk.major = sb->this_disk.major;
307 info->disk.minor = sb->this_disk.minor;
308 info->disk.raid_disk = sb->this_disk.raid_disk;
309 info->disk.number = sb->this_disk.number;
310
311 info->events = md_event(sb);
312 info->data_offset = 0;
313
314 uuid_from_super0(info->uuid, sbv);
315
316 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
317 info->reshape_active = 1;
318 info->reshape_progress = sb->reshape_position;
319 info->new_level = sb->new_level;
320 info->delta_disks = sb->delta_disks;
321 info->new_layout = sb->new_layout;
322 info->new_chunk = sb->new_chunk;
323 } else
324 info->reshape_active = 0;
325
326 info->name[0] = 0;
327 /* work_disks is calculated rather than read directly */
328 for (i=0; i < MD_SB_DISKS; i++)
329 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
330 (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
331 !(sb->disks[i].state & (1<<MD_DISK_FAULTY)))
332 working ++;
333 info->array.working_disks = working;
334 }
335
336
337 static int update_super0(struct mdinfo *info, void *sbv, char *update,
338 char *devname, int verbose,
339 int uuid_set, char *homehost)
340 {
341 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
342 * For others, the return value is ignored.
343 */
344 int rv = 0;
345 mdp_super_t *sb = sbv;
346 if (strcmp(update, "sparc2.2")==0 ) {
347 /* 2.2 sparc put the events in the wrong place
348 * So we copy the tail of the superblock
349 * up 4 bytes before continuing
350 */
351 __u32 *sb32 = (__u32*)sb;
352 memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
353 sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
354 (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
355 if (verbose >= 0)
356 fprintf (stderr, Name ": adjusting superblock of %s for 2.2/sparc compatability.\n",
357 devname);
358 }
359 if (strcmp(update, "super-minor") ==0) {
360 sb->md_minor = info->array.md_minor;
361 if (verbose > 0)
362 fprintf(stderr, Name ": updating superblock of %s with minor number %d\n",
363 devname, info->array.md_minor);
364 }
365 if (strcmp(update, "summaries") == 0) {
366 int i;
367 /* set nr_disks, active_disks, working_disks,
368 * failed_disks, spare_disks based on disks[]
369 * array in superblock.
370 * Also make sure extra slots aren't 'failed'
371 */
372 sb->nr_disks = sb->active_disks =
373 sb->working_disks = sb->failed_disks =
374 sb->spare_disks = 0;
375 for (i=0; i < MD_SB_DISKS ; i++)
376 if (sb->disks[i].major ||
377 sb->disks[i].minor) {
378 int state = sb->disks[i].state;
379 if (state & (1<<MD_DISK_REMOVED))
380 continue;
381 sb->nr_disks++;
382 if (state & (1<<MD_DISK_ACTIVE))
383 sb->active_disks++;
384 if (state & (1<<MD_DISK_FAULTY))
385 sb->failed_disks++;
386 else
387 sb->working_disks++;
388 if (state == 0)
389 sb->spare_disks++;
390 } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
391 sb->disks[i].state = 0;
392 }
393 if (strcmp(update, "force")==0) {
394 __u32 ehi = sb->events_hi, elo = sb->events_lo;
395 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
396 sb->events_lo = (info->events) & 0xFFFFFFFF;
397 if (sb->events_hi != ehi ||
398 sb->events_lo != elo)
399 rv = 1;
400 if ((sb->level == 5 || sb->level == 4 || sb->level == 6) &&
401 (sb->state & (1 << MD_SB_CLEAN)) == 0) {
402 /* need to force clean */
403 sb->state |= (1 << MD_SB_CLEAN);
404 rv = 1;
405 }
406 }
407 if (strcmp(update, "assemble")==0) {
408 int d = info->disk.number;
409 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
410 if ((sb->disks[d].state & ~(1<<MD_DISK_WRITEMOSTLY))
411 != info->disk.state) {
412 sb->disks[d].state = info->disk.state | wonly;
413 rv = 1;
414 }
415 }
416 if (strcmp(update, "newdev") == 0) {
417 int d = info->disk.number;
418 memset(&sb->disks[d], 0, sizeof(sb->disks[d]));
419 sb->disks[d].number = d;
420 sb->disks[d].major = info->disk.major;
421 sb->disks[d].minor = info->disk.minor;
422 sb->disks[d].raid_disk = info->disk.raid_disk;
423 sb->disks[d].state = info->disk.state;
424 sb->this_disk = sb->disks[d];
425 }
426 if (strcmp(update, "grow") == 0) {
427 sb->raid_disks = info->array.raid_disks;
428 sb->nr_disks = info->array.nr_disks;
429 sb->active_disks = info->array.active_disks;
430 sb->working_disks = info->array.working_disks;
431 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
432 sb->disks[info->disk.number].number = info->disk.number;
433 sb->disks[info->disk.number].major = info->disk.major;
434 sb->disks[info->disk.number].minor = info->disk.minor;
435 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
436 sb->disks[info->disk.number].state = info->disk.state;
437 if (sb->this_disk.number == info->disk.number)
438 sb->this_disk = sb->disks[info->disk.number];
439 }
440 if (strcmp(update, "resync") == 0) {
441 /* make sure resync happens */
442 sb->state &= ~(1<<MD_SB_CLEAN);
443 sb->recovery_cp = 0;
444 }
445 if (strcmp(update, "uuid") == 0) {
446 if (!uuid_set && homehost) {
447 unsigned char *hash = SHA1((unsigned char*)homehost,
448 strlen(homehost),
449 NULL);
450 memcpy(info->uuid+2, hash, 8);
451 }
452 sb->set_uuid0 = info->uuid[0];
453 sb->set_uuid1 = info->uuid[1];
454 sb->set_uuid2 = info->uuid[2];
455 sb->set_uuid3 = info->uuid[3];
456 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
457 struct bitmap_super_s *bm;
458 bm = (struct bitmap_super_s*)(sb+1);
459 uuid_from_super0((int*)bm->uuid, sbv);
460 }
461 }
462 if (strcmp(update, "_reshape_progress")==0)
463 sb->reshape_position = info->reshape_progress;
464
465 sb->sb_csum = calc_sb0_csum(sb);
466 return rv;
467 }
468
469 static __u64 event_super0(void *sbv)
470 {
471 mdp_super_t *sb = sbv;
472 return md_event(sb);
473 }
474
475 /*
476 * For verion-0 superblock, the homehost is 'stored' in the
477 * uuid. 8 bytes for a hash of the host leaving 8 bytes
478 * of random material.
479 * We use the first 8 bytes (64bits) of the sha1 of the
480 * host name
481 */
482
483
484 static int init_super0(struct supertype *st, void **sbp, mdu_array_info_t *info,
485 unsigned long long size, char *ignored_name, char *homehost)
486 {
487 mdp_super_t *sb = malloc(MD_SB_BYTES + sizeof(bitmap_super_t));
488 int spares;
489 int rfd;
490 memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
491
492 if (info->major_version == -1) {
493 /* zeroing the superblock */
494 *sbp = sb;
495 return 0;
496 }
497
498 spares = info->working_disks - info->active_disks;
499 if (info->raid_disks + spares > MD_SB_DISKS) {
500 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
501 info->raid_disks , spares, MD_SB_DISKS);
502 return 0;
503 }
504
505 rfd = open("/dev/urandom", O_RDONLY);
506 sb->md_magic = MD_SB_MAGIC;
507 sb->major_version = 0;
508 sb->minor_version = 90;
509 sb->patch_version = 0;
510 sb->gvalid_words = 0; /* ignored */
511 if (rfd < 0 || read(rfd, &sb->set_uuid0, 4) != 4)
512 sb->set_uuid0 = random();
513 sb->ctime = time(0);
514 sb->level = info->level;
515 if (size != info->size)
516 return 0;
517 sb->size = info->size;
518 sb->nr_disks = info->nr_disks;
519 sb->raid_disks = info->raid_disks;
520 sb->md_minor = info->md_minor;
521 sb->not_persistent = 0;
522 if (rfd < 0 || read(rfd, &sb->set_uuid1, 12) != 12) {
523 sb->set_uuid1 = random();
524 sb->set_uuid2 = random();
525 sb->set_uuid3 = random();
526 }
527 if (rfd >= 0)
528 close(rfd);
529 if (homehost) {
530 unsigned char *hash = SHA1((unsigned char*)homehost,
531 strlen(homehost),
532 NULL);
533 memcpy(&sb->set_uuid2, hash, 8);
534 }
535
536 sb->utime = sb->ctime;
537 sb->state = info->state;
538 sb->active_disks = info->active_disks;
539 sb->working_disks = info->working_disks;
540 sb->failed_disks = info->failed_disks;
541 sb->spare_disks = info->spare_disks;
542 sb->events_hi = 0;
543 sb->events_lo = 1;
544
545 sb->layout = info->layout;
546 sb->chunk_size = info->chunk_size;
547
548 *sbp = sb;
549 return 1;
550 }
551
552 /* Add a device to the superblock being created */
553 static void add_to_super0(void *sbv, mdu_disk_info_t *dinfo)
554 {
555 mdp_super_t *sb = sbv;
556 mdp_disk_t *dk = &sb->disks[dinfo->number];
557
558 dk->number = dinfo->number;
559 dk->major = dinfo->major;
560 dk->minor = dinfo->minor;
561 dk->raid_disk = dinfo->raid_disk;
562 dk->state = dinfo->state;
563 }
564
565 static int store_super0(struct supertype *st, int fd, void *sbv)
566 {
567 unsigned long size;
568 unsigned long long dsize;
569 unsigned long long offset;
570 mdp_super_t *super = sbv;
571
572 #ifdef BLKGETSIZE64
573 if (ioctl(fd, BLKGETSIZE64, &dsize) != 0)
574 #endif
575 {
576 if (ioctl(fd, BLKGETSIZE, &size))
577 return 1;
578 else
579 dsize = ((unsigned long long)size)<<9;
580 }
581
582 if (dsize < MD_RESERVED_SECTORS*2*512)
583 return 2;
584
585 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
586
587 offset *= 512;
588
589 if (lseek64(fd, offset, 0)< 0LL)
590 return 3;
591
592 if (write(fd, super, sizeof(*super)) != sizeof(*super))
593 return 4;
594
595 if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
596 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
597 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
598 write(fd, bm, sizeof(*bm));
599 }
600
601 fsync(fd);
602 return 0;
603 }
604
605 static int write_init_super0(struct supertype *st, void *sbv, mdu_disk_info_t *dinfo, char *devname)
606 {
607 mdp_super_t *sb = sbv;
608 int fd = open(devname, O_RDWR|O_EXCL);
609 int rv;
610
611 if (fd < 0) {
612 fprintf(stderr, Name ": Failed to open %s to write superblock\n", devname);
613 return -1;
614 }
615
616 sb->disks[dinfo->number].state &= ~(1<<MD_DISK_FAULTY);
617 sb->disks[dinfo->number].state |= (1<<MD_DISK_SYNC);
618
619 sb->this_disk = sb->disks[dinfo->number];
620 sb->sb_csum = calc_sb0_csum(sb);
621 rv = store_super0(st, fd, sb);
622
623 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
624 rv = st->ss->write_bitmap(st, fd, sbv);
625
626 close(fd);
627 if (rv)
628 fprintf(stderr, Name ": failed to write superblock to %s\n", devname);
629 return rv;
630 }
631
632 static int compare_super0(void **firstp, void *secondv)
633 {
634 /*
635 * return:
636 * 0 same, or first was empty, and second was copied
637 * 1 second had wrong number
638 * 2 wrong uuid
639 * 3 wrong other info
640 */
641 mdp_super_t *first = *firstp;
642 mdp_super_t *second = secondv;
643
644 int uuid1[4], uuid2[4];
645 if (second->md_magic != MD_SB_MAGIC)
646 return 1;
647 if (!first) {
648 first = malloc(MD_SB_BYTES + sizeof(struct bitmap_super_s));
649 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
650 *firstp = first;
651 return 0;
652 }
653
654 uuid_from_super0(uuid1, first);
655 uuid_from_super0(uuid2, second);
656 if (!same_uuid(uuid1, uuid2, 0))
657 return 2;
658 if (first->major_version != second->major_version ||
659 first->minor_version != second->minor_version ||
660 first->patch_version != second->patch_version ||
661 first->gvalid_words != second->gvalid_words ||
662 first->ctime != second->ctime ||
663 first->level != second->level ||
664 first->size != second->size ||
665 first->raid_disks != second->raid_disks )
666 return 3;
667
668 return 0;
669 }
670
671
672 static int load_super0(struct supertype *st, int fd, void **sbp, char *devname)
673 {
674 /* try to read in the superblock
675 * Return:
676 * 0 on success
677 * 1 on cannot get superblock
678 * 2 on superblock meaningless
679 */
680 unsigned long size;
681 unsigned long long dsize;
682 unsigned long long offset;
683 mdp_super_t *super;
684 int uuid[4];
685 struct bitmap_super_s *bsb;
686
687 #ifdef BLKGETSIZE64
688 if (ioctl(fd, BLKGETSIZE64, &dsize) != 0)
689 #endif
690 {
691 if (ioctl(fd, BLKGETSIZE, &size)) {
692 if (devname)
693 fprintf(stderr, Name ": cannot find device size for %s: %s\n",
694 devname, strerror(errno));
695 return 1;
696 } else
697 dsize = size << 9;
698 }
699
700 if (dsize < MD_RESERVED_SECTORS*2) {
701 if (devname)
702 fprintf(stderr, Name ": %s is too small for md: size is %ld sectors.\n",
703 devname, size);
704 return 1;
705 }
706
707 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
708
709 offset *= 512;
710
711 ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
712
713 if (lseek64(fd, offset, 0)< 0LL) {
714 if (devname)
715 fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
716 devname, strerror(errno));
717 return 1;
718 }
719
720 super = malloc(MD_SB_BYTES + sizeof(bitmap_super_t));
721
722 if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
723 if (devname)
724 fprintf(stderr, Name ": Cannot read superblock on %s\n",
725 devname);
726 free(super);
727 return 1;
728 }
729
730 if (st->ss && st->minor_version == 9)
731 super0_swap_endian(super);
732
733 if (super->md_magic != MD_SB_MAGIC) {
734 if (devname)
735 fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
736 devname, MD_SB_MAGIC, super->md_magic);
737 free(super);
738 return 2;
739 }
740
741 if (super->major_version != 0) {
742 if (devname)
743 fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
744 devname, super->major_version);
745 free(super);
746 return 2;
747 }
748 *sbp = super;
749 if (st->ss == NULL) {
750 st->ss = &super0;
751 st->minor_version = 90;
752 st->max_devs = MD_SB_DISKS;
753 }
754
755 /* Now check on the bitmap superblock */
756 if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
757 return 0;
758 /* Read the bitmap superblock and make sure it looks
759 * valid. If it doesn't clear the bit. An --assemble --force
760 * should get that written out.
761 */
762 if (read(fd, super+1, sizeof(struct bitmap_super_s))
763 != sizeof(struct bitmap_super_s))
764 goto no_bitmap;
765
766 uuid_from_super0(uuid, super);
767 bsb = (struct bitmap_super_s *)(super+1);
768 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
769 memcmp(bsb->uuid, uuid, 16) != 0)
770 goto no_bitmap;
771 return 0;
772
773 no_bitmap:
774 super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
775
776 return 0;
777 }
778
779 static struct supertype *match_metadata_desc0(char *arg)
780 {
781 struct supertype *st = malloc(sizeof(*st));
782 if (!st) return st;
783
784 st->ss = &super0;
785 st->minor_version = 90;
786 st->max_devs = MD_SB_DISKS;
787 if (strcmp(arg, "0") == 0 ||
788 strcmp(arg, "0.90") == 0 ||
789 strcmp(arg, "default") == 0
790 )
791 return st;
792
793 st->minor_version = 9; /* flag for 'byte-swapped' */
794 if (strcmp(arg, "0.swap")==0)
795 return st;
796
797 free(st);
798 return NULL;
799 }
800
801 static __u64 avail_size0(struct supertype *st, __u64 devsize)
802 {
803 if (devsize < MD_RESERVED_SECTORS*2)
804 return 0ULL;
805 return MD_NEW_SIZE_SECTORS(devsize);
806 }
807
808 static int add_internal_bitmap0(struct supertype *st, void *sbv, int chunk, int delay, int write_behind, unsigned long long size, int may_change, int major)
809 {
810 /*
811 * The bitmap comes immediately after the superblock and must be 60K in size
812 * at most. The default size is between 30K and 60K
813 *
814 * size is in sectors, chunk is in bytes !!!
815 */
816 unsigned long long bits;
817 unsigned long long max_bits = 60*1024*8;
818 unsigned long long min_chunk;
819 mdp_super_t *sb = sbv;
820 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
821
822
823 min_chunk = 4096; /* sub-page chunks don't work yet.. */
824 bits = (size * 512) / min_chunk + 1;
825 while (bits > max_bits) {
826 min_chunk *= 2;
827 bits = (bits+1)/2;
828 }
829 if (chunk == UnSet)
830 chunk = min_chunk;
831 else if (chunk < min_chunk)
832 return 0; /* chunk size too small */
833
834 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
835
836 memset(bms, 0, sizeof(*bms));
837 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
838 bms->version = __cpu_to_le32(major);
839 uuid_from_super0((int*)bms->uuid, sb);
840 bms->chunksize = __cpu_to_le32(chunk);
841 bms->daemon_sleep = __cpu_to_le32(delay);
842 bms->sync_size = __cpu_to_le64(size);
843 bms->write_behind = __cpu_to_le32(write_behind);
844
845 return 1;
846 }
847
848
849 void locate_bitmap0(struct supertype *st, int fd, void *sbv)
850 {
851 unsigned long long dsize;
852 unsigned long size;
853 unsigned long long offset;
854 #ifdef BLKGETSIZE64
855 if (ioctl(fd, BLKGETSIZE64, &dsize) != 0)
856 #endif
857 {
858 if (ioctl(fd, BLKGETSIZE, &size))
859 return;
860 else
861 dsize = ((unsigned long long)size)<<9;
862 }
863
864 if (dsize < MD_RESERVED_SECTORS*2)
865 return;
866
867 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
868
869 offset *= 512;
870
871 offset += MD_SB_BYTES;
872
873 lseek64(fd, offset, 0);
874 }
875
876 int write_bitmap0(struct supertype *st, int fd, void *sbv)
877 {
878 unsigned long size;
879 unsigned long long dsize;
880 unsigned long long offset;
881 mdp_super_t *sb = sbv;
882
883 int rv = 0;
884
885 int towrite, n;
886 char buf[4096];
887
888 #ifdef BLKGETSIZE64
889 if (ioctl(fd, BLKGETSIZE64, &dsize) != 0)
890 #endif
891 {
892 if (ioctl(fd, BLKGETSIZE, &size))
893 return 1;
894 else
895 dsize = ((unsigned long long)size)<<9;
896 }
897
898 if (dsize < MD_RESERVED_SECTORS*2)
899 return -1;
900
901 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
902
903 offset *= 512;
904
905 if (lseek64(fd, offset + 4096, 0)< 0LL)
906 return 3;
907
908
909 if (write(fd, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t)) !=
910 sizeof(bitmap_super_t))
911 return -2;
912 towrite = 64*1024 - MD_SB_BYTES - sizeof(bitmap_super_t);
913 memset(buf, 0xff, sizeof(buf));
914 while (towrite > 0) {
915 n = towrite;
916 if (n > sizeof(buf))
917 n = sizeof(buf);
918 n = write(fd, buf, n);
919 if (n > 0)
920 towrite -= n;
921 else
922 break;
923 }
924 fsync(fd);
925 if (towrite)
926 rv = -2;
927
928 return rv;
929 }
930
931 struct superswitch super0 = {
932 #ifndef MDASSEMBLE
933 .examine_super = examine_super0,
934 .brief_examine_super = brief_examine_super0,
935 .detail_super = detail_super0,
936 .brief_detail_super = brief_detail_super0,
937 #endif
938 .uuid_from_super = uuid_from_super0,
939 .getinfo_super = getinfo_super0,
940 .update_super = update_super0,
941 .event_super = event_super0,
942 .init_super = init_super0,
943 .add_to_super = add_to_super0,
944 .store_super = store_super0,
945 .write_init_super = write_init_super0,
946 .compare_super = compare_super0,
947 .load_super = load_super0,
948 .match_metadata_desc = match_metadata_desc0,
949 .avail_size = avail_size0,
950 .add_internal_bitmap = add_internal_bitmap0,
951 .locate_bitmap = locate_bitmap0,
952 .write_bitmap = write_bitmap0,
953 .major = 0,
954 .swapuuid = 0,
955 };