]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super0.c
2a2938dfdd8c4aa0253059e25e8849660745d3e4
[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 #define HAVE_STDINT_H 1
31 #include "mdadm.h"
32 #include "sha1.h"
33 /*
34 * All handling for the 0.90.0 version superblock is in
35 * this file.
36 * This includes:
37 * - finding, loading, and writing the superblock.
38 * - initialising a new superblock
39 * - printing the superblock for --examine
40 * - printing part of the superblock for --detail
41 * .. other stuff
42 */
43
44
45 static unsigned long calc_sb0_csum(mdp_super_t *super)
46 {
47 unsigned long csum = super->sb_csum;
48 unsigned long newcsum;
49 super->sb_csum= 0 ;
50 newcsum = calc_csum(super, MD_SB_BYTES);
51 super->sb_csum = csum;
52 return newcsum;
53 }
54
55
56 void super0_swap_endian(struct mdp_superblock_s *sb)
57 {
58 /* as super0 superblocks are host-endian, it is sometimes
59 * useful to be able to swap the endianness
60 * as (almost) everything is u32's we byte-swap every 4byte
61 * number.
62 * We then also have to swap the events_hi and events_lo
63 */
64 char *sbc = (char *)sb;
65 __u32 t32;
66 int i;
67
68 for (i=0; i < MD_SB_BYTES ; i+=4) {
69 char t = sbc[i];
70 sbc[i] = sbc[i+3];
71 sbc[i+3] = t;
72 t=sbc[i+1];
73 sbc[i+1]=sbc[i+2];
74 sbc[i+2]=t;
75 }
76 t32 = sb->events_hi;
77 sb->events_hi = sb->events_lo;
78 sb->events_lo = t32;
79
80 t32 = sb->cp_events_hi;
81 sb->cp_events_hi = sb->cp_events_lo;
82 sb->cp_events_lo = t32;
83
84 }
85
86 #ifndef MDASSEMBLE
87
88 static void examine_super0(void *sbv, char *homehost)
89 {
90 mdp_super_t *sb = sbv;
91 time_t atime;
92 int d;
93 char *c;
94
95 printf(" Magic : %08x\n", sb->md_magic);
96 printf(" Version : %02d.%02d.%02d\n", sb->major_version, sb->minor_version,
97 sb->patch_version);
98 if (sb->minor_version >= 90) {
99 printf(" UUID : %08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
100 sb->set_uuid2, sb->set_uuid3);
101 if (homehost) {
102 char buf[20];
103 void *hash = sha1_buffer(homehost,
104 strlen(homehost),
105 buf);
106 if (memcmp(&sb->set_uuid2, hash, 8)==0)
107 printf(" (local to host %s)", homehost);
108 }
109 printf("\n");
110 } else
111 printf(" UUID : %08x\n", sb->set_uuid0);
112
113 if (sb->not_persistent)
114 printf(" Eedk : not persistent\n");
115
116 atime = sb->ctime;
117 printf(" Creation Time : %.24s\n", ctime(&atime));
118 c=map_num(pers, sb->level);
119 printf(" Raid Level : %s\n", c?c:"-unknown-");
120 if ((int)sb->level >= 0) {
121 int ddsks=0;
122 printf(" Used Dev Size : %d%s\n", sb->size,
123 human_size((long long)sb->size<<10));
124 switch(sb->level) {
125 case 1: ddsks=1;break;
126 case 4:
127 case 5: ddsks = sb->raid_disks-1; break;
128 case 6: ddsks = sb->raid_disks-2; break;
129 case 10: ddsks = sb->raid_disks / (sb->layout&255) / ((sb->layout>>8)&255);
130 }
131 if (ddsks)
132 printf(" Array Size : %llu%s\n", (unsigned long long)ddsks * sb->size,
133 human_size(ddsks*(long long)sb->size<<10));
134 }
135 printf(" Raid Devices : %d\n", sb->raid_disks);
136 printf(" Total Devices : %d\n", sb->nr_disks);
137 printf("Preferred Minor : %d\n", sb->md_minor);
138 printf("\n");
139 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
140 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)sb->reshape_position/2, human_size((long long)sb->reshape_position<<9));
141 if (sb->delta_disks) {
142 printf(" Delta Devices : %d", sb->delta_disks);
143 if (sb->delta_disks)
144 printf(" (%d->%d)\n", sb->raid_disks-sb->delta_disks, sb->raid_disks);
145 else
146 printf(" (%d->%d)\n", sb->raid_disks, sb->raid_disks+sb->delta_disks);
147 }
148 if (sb->new_level != sb->level) {
149 c = map_num(pers, sb->new_level);
150 printf(" New Level : %s\n", c?c:"-unknown-");
151 }
152 if (sb->new_layout != sb->layout) {
153 if (sb->level == 5) {
154 c = map_num(r5layout, sb->new_layout);
155 printf(" New Layout : %s\n", c?c:"-unknown-");
156 }
157 if (sb->level == 10) {
158 printf(" New Layout : near=%d, %s=%d\n",
159 sb->new_layout&255,
160 (sb->new_layout&0x10000)?"offset":"far",
161 (sb->new_layout>>8)&255);
162 }
163 }
164 if (sb->new_chunk != sb->chunk_size)
165 printf(" New Chunksize : %d\n", sb->new_chunk);
166 printf("\n");
167 }
168 atime = sb->utime;
169 printf(" Update Time : %.24s\n", ctime(&atime));
170 printf(" State : %s\n",
171 (sb->state&(1<<MD_SB_CLEAN))?"clean":"active");
172 if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
173 printf("Internal Bitmap : present\n");
174 printf(" Active Devices : %d\n", sb->active_disks);
175 printf("Working Devices : %d\n", sb->working_disks);
176 printf(" Failed Devices : %d\n", sb->failed_disks);
177 printf(" Spare Devices : %d\n", sb->spare_disks);
178 if (calc_sb0_csum(sb) == sb->sb_csum)
179 printf(" Checksum : %x - correct\n", sb->sb_csum);
180 else
181 printf(" Checksum : %x - expected %lx\n", sb->sb_csum, calc_sb0_csum(sb));
182 printf(" Events : %d.%d\n", sb->events_hi, sb->events_lo);
183 printf("\n");
184 if (sb->level == 5) {
185 c = map_num(r5layout, sb->layout);
186 printf(" Layout : %s\n", c?c:"-unknown-");
187 }
188 if (sb->level == 10) {
189 printf(" Layout : near=%d, %s=%d\n",
190 sb->layout&255,
191 (sb->layout&0x10000)?"offset":"far",
192 (sb->layout>>8)&255);
193 }
194 switch(sb->level) {
195 case 0:
196 case 4:
197 case 5:
198 case 6:
199 case 10:
200 printf(" Chunk Size : %dK\n", sb->chunk_size/1024);
201 break;
202 case -1:
203 printf(" Rounding : %dK\n", sb->chunk_size/1024);
204 break;
205 default: break;
206 }
207 printf("\n");
208 printf(" Number Major Minor RaidDevice State\n");
209 for (d= -1; d<(signed int)(sb->raid_disks+sb->spare_disks); d++) {
210 mdp_disk_t *dp;
211 char *dv;
212 char nb[5];
213 int wonly;
214 if (d>=0) dp = &sb->disks[d];
215 else dp = &sb->this_disk;
216 snprintf(nb, sizeof(nb), "%4d", d);
217 printf("%4s %5d %5d %5d %5d ", d < 0 ? "this" : nb,
218 dp->number, dp->major, dp->minor, dp->raid_disk);
219 wonly = dp->state & (1<<MD_DISK_WRITEMOSTLY);
220 dp->state &= ~(1<<MD_DISK_WRITEMOSTLY);
221 if (dp->state & (1<<MD_DISK_FAULTY)) printf(" faulty");
222 if (dp->state & (1<<MD_DISK_ACTIVE)) printf(" active");
223 if (dp->state & (1<<MD_DISK_SYNC)) printf(" sync");
224 if (dp->state & (1<<MD_DISK_REMOVED)) printf(" removed");
225 if (wonly) printf(" write-mostly");
226 if (dp->state == 0) printf(" spare");
227 if ((dv=map_dev(dp->major, dp->minor, 0)))
228 printf(" %s", dv);
229 printf("\n");
230 if (d == -1) printf("\n");
231 }
232 }
233
234 static void brief_examine_super0(void *sbv)
235 {
236 mdp_super_t *sb = sbv;
237 char *c=map_num(pers, sb->level);
238 char devname[20];
239
240 sprintf(devname, "/dev/md%d", sb->md_minor);
241
242 printf("ARRAY %s level=%s num-devices=%d UUID=",
243 devname,
244 c?c:"-unknown-", sb->raid_disks);
245 if (sb->minor_version >= 90)
246 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
247 sb->set_uuid2, sb->set_uuid3);
248 else
249 printf("%08x", sb->set_uuid0);
250 printf("\n");
251 }
252
253 static void detail_super0(void *sbv, char *homehost)
254 {
255 mdp_super_t *sb = sbv;
256 printf(" UUID : ");
257 if (sb->minor_version >= 90)
258 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
259 sb->set_uuid2, sb->set_uuid3);
260 else
261 printf("%08x", sb->set_uuid0);
262 if (homehost) {
263 char buf[20];
264 void *hash = sha1_buffer(homehost,
265 strlen(homehost),
266 buf);
267 if (memcmp(&sb->set_uuid2, hash, 8)==0)
268 printf(" (local to host %s)", homehost);
269 }
270 printf("\n Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
271 }
272
273 static void brief_detail_super0(void *sbv)
274 {
275 mdp_super_t *sb = sbv;
276 printf(" UUID=");
277 if (sb->minor_version >= 90)
278 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
279 sb->set_uuid2, sb->set_uuid3);
280 else
281 printf("%08x", sb->set_uuid0);
282 }
283
284 static void export_super0(void *sbv)
285 {
286 mdp_super_t *sb = sbv;
287 printf("MD_UUID=");
288 if (sb->minor_version >= 90)
289 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
290 sb->set_uuid2, sb->set_uuid3);
291 else
292 printf("%08x", sb->set_uuid0);
293 printf("\n");
294 }
295 #endif
296
297 static int match_home0(void *sbv, char *homehost)
298 {
299 mdp_super_t *sb = sbv;
300 char buf[20];
301 char *hash = sha1_buffer(homehost,
302 strlen(homehost),
303 buf);
304
305 return (memcmp(&sb->set_uuid2, hash, 8)==0);
306 }
307
308 static void uuid_from_super0(int uuid[4], void * sbv)
309 {
310 mdp_super_t *super = sbv;
311 uuid[0] = super->set_uuid0;
312 if (super->minor_version >= 90) {
313 uuid[1] = super->set_uuid1;
314 uuid[2] = super->set_uuid2;
315 uuid[3] = super->set_uuid3;
316 } else {
317 uuid[1] = 0;
318 uuid[2] = 0;
319 uuid[3] = 0;
320 }
321 }
322
323 static void getinfo_super0(struct mdinfo *info, void *sbv)
324 {
325 mdp_super_t *sb = sbv;
326 int working = 0;
327 int i;
328
329 info->array.major_version = sb->major_version;
330 info->array.minor_version = sb->minor_version;
331 info->array.patch_version = sb->patch_version;
332 info->array.raid_disks = sb->raid_disks;
333 info->array.level = sb->level;
334 info->array.layout = sb->layout;
335 info->array.md_minor = sb->md_minor;
336 info->array.ctime = sb->ctime;
337 info->array.utime = sb->utime;
338 info->array.chunk_size = sb->chunk_size;
339 info->array.state = sb->state;
340 info->component_size = sb->size*2;
341
342 info->disk.state = sb->this_disk.state;
343 info->disk.major = sb->this_disk.major;
344 info->disk.minor = sb->this_disk.minor;
345 info->disk.raid_disk = sb->this_disk.raid_disk;
346 info->disk.number = sb->this_disk.number;
347
348 info->events = md_event(sb);
349 info->data_offset = 0;
350
351 uuid_from_super0(info->uuid, sbv);
352
353 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
354 info->reshape_active = 1;
355 info->reshape_progress = sb->reshape_position;
356 info->new_level = sb->new_level;
357 info->delta_disks = sb->delta_disks;
358 info->new_layout = sb->new_layout;
359 info->new_chunk = sb->new_chunk;
360 } else
361 info->reshape_active = 0;
362
363 sprintf(info->name, "%d", sb->md_minor);
364 /* work_disks is calculated rather than read directly */
365 for (i=0; i < MD_SB_DISKS; i++)
366 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
367 (sb->disks[i].raid_disk < info->array.raid_disks) &&
368 (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
369 !(sb->disks[i].state & (1<<MD_DISK_FAULTY)))
370 working ++;
371 info->array.working_disks = working;
372 }
373
374
375 static int update_super0(struct mdinfo *info, void *sbv, char *update,
376 char *devname, int verbose,
377 int uuid_set, char *homehost)
378 {
379 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
380 * For others, the return value is ignored.
381 */
382 int rv = 0;
383 mdp_super_t *sb = sbv;
384 if (strcmp(update, "sparc2.2")==0 ) {
385 /* 2.2 sparc put the events in the wrong place
386 * So we copy the tail of the superblock
387 * up 4 bytes before continuing
388 */
389 __u32 *sb32 = (__u32*)sb;
390 memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
391 sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
392 (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
393 if (verbose >= 0)
394 fprintf (stderr, Name ": adjusting superblock of %s for 2.2/sparc compatability.\n",
395 devname);
396 }
397 if (strcmp(update, "super-minor") ==0) {
398 sb->md_minor = info->array.md_minor;
399 if (verbose > 0)
400 fprintf(stderr, Name ": updating superblock of %s with minor number %d\n",
401 devname, info->array.md_minor);
402 }
403 if (strcmp(update, "summaries") == 0) {
404 int i;
405 /* set nr_disks, active_disks, working_disks,
406 * failed_disks, spare_disks based on disks[]
407 * array in superblock.
408 * Also make sure extra slots aren't 'failed'
409 */
410 sb->nr_disks = sb->active_disks =
411 sb->working_disks = sb->failed_disks =
412 sb->spare_disks = 0;
413 for (i=0; i < MD_SB_DISKS ; i++)
414 if (sb->disks[i].major ||
415 sb->disks[i].minor) {
416 int state = sb->disks[i].state;
417 if (state & (1<<MD_DISK_REMOVED))
418 continue;
419 sb->nr_disks++;
420 if (state & (1<<MD_DISK_ACTIVE))
421 sb->active_disks++;
422 if (state & (1<<MD_DISK_FAULTY))
423 sb->failed_disks++;
424 else
425 sb->working_disks++;
426 if (state == 0)
427 sb->spare_disks++;
428 } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
429 sb->disks[i].state = 0;
430 }
431 if (strcmp(update, "force-one")==0) {
432 /* Not enough devices for a working array, so
433 * bring this one up-to-date.
434 */
435 __u32 ehi = sb->events_hi, elo = sb->events_lo;
436 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
437 sb->events_lo = (info->events) & 0xFFFFFFFF;
438 if (sb->events_hi != ehi ||
439 sb->events_lo != elo)
440 rv = 1;
441 }
442 if (strcmp(update, "force-array")==0) {
443 /* degraded array and 'force' requested, so
444 * maybe need to mark it 'clean'
445 */
446 if ((sb->level == 5 || sb->level == 4 || sb->level == 6) &&
447 (sb->state & (1 << MD_SB_CLEAN)) == 0) {
448 /* need to force clean */
449 sb->state |= (1 << MD_SB_CLEAN);
450 rv = 1;
451 }
452 }
453 if (strcmp(update, "assemble")==0) {
454 int d = info->disk.number;
455 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
456 if ((sb->disks[d].state & ~(1<<MD_DISK_WRITEMOSTLY))
457 != info->disk.state) {
458 sb->disks[d].state = info->disk.state | wonly;
459 rv = 1;
460 }
461 }
462 if (strcmp(update, "linear-grow-new") == 0) {
463 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
464 sb->disks[info->disk.number].number = info->disk.number;
465 sb->disks[info->disk.number].major = info->disk.major;
466 sb->disks[info->disk.number].minor = info->disk.minor;
467 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
468 sb->disks[info->disk.number].state = info->disk.state;
469 sb->this_disk = sb->disks[info->disk.number];
470 }
471 if (strcmp(update, "linear-grow-update") == 0) {
472 sb->raid_disks = info->array.raid_disks;
473 sb->nr_disks = info->array.nr_disks;
474 sb->active_disks = info->array.active_disks;
475 sb->working_disks = info->array.working_disks;
476 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
477 sb->disks[info->disk.number].number = info->disk.number;
478 sb->disks[info->disk.number].major = info->disk.major;
479 sb->disks[info->disk.number].minor = info->disk.minor;
480 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
481 sb->disks[info->disk.number].state = info->disk.state;
482 }
483 if (strcmp(update, "resync") == 0) {
484 /* make sure resync happens */
485 sb->state &= ~(1<<MD_SB_CLEAN);
486 sb->recovery_cp = 0;
487 }
488 if (strcmp(update, "homehost") == 0 &&
489 homehost) {
490 uuid_set = 0;
491 update = "uuid";
492 info->uuid[0] = sb->set_uuid0;
493 info->uuid[1] = sb->set_uuid1;
494 }
495 if (strcmp(update, "uuid") == 0) {
496 if (!uuid_set && homehost) {
497 char buf[20];
498 char *hash = sha1_buffer(homehost,
499 strlen(homehost),
500 buf);
501 memcpy(info->uuid+2, hash, 8);
502 }
503 sb->set_uuid0 = info->uuid[0];
504 sb->set_uuid1 = info->uuid[1];
505 sb->set_uuid2 = info->uuid[2];
506 sb->set_uuid3 = info->uuid[3];
507 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
508 struct bitmap_super_s *bm;
509 bm = (struct bitmap_super_s*)(sb+1);
510 uuid_from_super0((int*)bm->uuid, sbv);
511 }
512 }
513 if (strcmp(update, "_reshape_progress")==0)
514 sb->reshape_position = info->reshape_progress;
515
516 sb->sb_csum = calc_sb0_csum(sb);
517 return rv;
518 }
519
520 /*
521 * For verion-0 superblock, the homehost is 'stored' in the
522 * uuid. 8 bytes for a hash of the host leaving 8 bytes
523 * of random material.
524 * We use the first 8 bytes (64bits) of the sha1 of the
525 * host name
526 */
527
528
529 static int init_super0(struct supertype *st, void **sbp, mdu_array_info_t *info,
530 unsigned long long size, char *ignored_name, char *homehost,
531 int *uuid)
532 {
533 mdp_super_t *sb = malloc(MD_SB_BYTES + sizeof(bitmap_super_t));
534 int spares;
535 memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
536
537 if (info->major_version == -1) {
538 /* zeroing the superblock */
539 *sbp = sb;
540 return 0;
541 }
542
543 spares = info->working_disks - info->active_disks;
544 if (info->raid_disks + spares > MD_SB_DISKS) {
545 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
546 info->raid_disks , spares, MD_SB_DISKS);
547 return 0;
548 }
549
550 sb->md_magic = MD_SB_MAGIC;
551 sb->major_version = 0;
552 sb->minor_version = 90;
553 sb->patch_version = 0;
554 sb->gvalid_words = 0; /* ignored */
555 sb->ctime = time(0);
556 sb->level = info->level;
557 if (size != info->size)
558 return 0;
559 sb->size = info->size;
560 sb->nr_disks = info->nr_disks;
561 sb->raid_disks = info->raid_disks;
562 sb->md_minor = info->md_minor;
563 sb->not_persistent = 0;
564 if (uuid) {
565 sb->set_uuid0 = uuid[0];
566 sb->set_uuid1 = uuid[1];
567 sb->set_uuid2 = uuid[2];
568 sb->set_uuid3 = uuid[3];
569 } else {
570 int rfd = open("/dev/urandom", O_RDONLY);
571 if (rfd < 0 || read(rfd, &sb->set_uuid0, 4) != 4)
572 sb->set_uuid0 = random();
573 if (rfd < 0 || read(rfd, &sb->set_uuid1, 12) != 12) {
574 sb->set_uuid1 = random();
575 sb->set_uuid2 = random();
576 sb->set_uuid3 = random();
577 }
578 if (rfd >= 0)
579 close(rfd);
580 }
581 if (homehost) {
582 char buf[20];
583 char *hash = sha1_buffer(homehost,
584 strlen(homehost),
585 buf);
586 memcpy(&sb->set_uuid2, hash, 8);
587 }
588
589 sb->utime = sb->ctime;
590 sb->state = info->state;
591 sb->active_disks = info->active_disks;
592 sb->working_disks = info->working_disks;
593 sb->failed_disks = info->failed_disks;
594 sb->spare_disks = info->spare_disks;
595 sb->events_hi = 0;
596 sb->events_lo = 1;
597
598 sb->layout = info->layout;
599 sb->chunk_size = info->chunk_size;
600
601 *sbp = sb;
602 return 1;
603 }
604
605 /* Add a device to the superblock being created */
606 static void add_to_super0(void *sbv, mdu_disk_info_t *dinfo)
607 {
608 mdp_super_t *sb = sbv;
609 mdp_disk_t *dk = &sb->disks[dinfo->number];
610
611 dk->number = dinfo->number;
612 dk->major = dinfo->major;
613 dk->minor = dinfo->minor;
614 dk->raid_disk = dinfo->raid_disk;
615 dk->state = dinfo->state;
616 }
617
618 static int store_super0(struct supertype *st, int fd, void *sbv)
619 {
620 unsigned long long dsize;
621 unsigned long long offset;
622 mdp_super_t *super = sbv;
623
624 if (!get_dev_size(fd, NULL, &dsize))
625 return 1;
626
627 if (dsize < MD_RESERVED_SECTORS*512)
628 return 2;
629
630 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
631
632 offset *= 512;
633
634 if (lseek64(fd, offset, 0)< 0LL)
635 return 3;
636
637 if (write(fd, super, sizeof(*super)) != sizeof(*super))
638 return 4;
639
640 if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
641 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
642 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
643 if (write(fd, bm, sizeof(*bm)) != sizeof(*bm))
644 return 5;
645 }
646
647 fsync(fd);
648 return 0;
649 }
650
651 static int write_init_super0(struct supertype *st, void *sbv, mdu_disk_info_t *dinfo, char *devname)
652 {
653 mdp_super_t *sb = sbv;
654 int fd = open(devname, O_RDWR|O_EXCL);
655 int rv;
656
657 if (fd < 0) {
658 fprintf(stderr, Name ": Failed to open %s to write superblock\n", devname);
659 return -1;
660 }
661
662 sb->disks[dinfo->number].state &= ~(1<<MD_DISK_FAULTY);
663
664 sb->this_disk = sb->disks[dinfo->number];
665 sb->sb_csum = calc_sb0_csum(sb);
666 rv = store_super0(st, fd, sb);
667
668 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
669 rv = st->ss->write_bitmap(st, fd, sbv);
670
671 close(fd);
672 if (rv)
673 fprintf(stderr, Name ": failed to write superblock to %s\n", devname);
674 return rv;
675 }
676
677 static int compare_super0(void **firstp, void *secondv)
678 {
679 /*
680 * return:
681 * 0 same, or first was empty, and second was copied
682 * 1 second had wrong number
683 * 2 wrong uuid
684 * 3 wrong other info
685 */
686 mdp_super_t *first = *firstp;
687 mdp_super_t *second = secondv;
688
689 int uuid1[4], uuid2[4];
690 if (second->md_magic != MD_SB_MAGIC)
691 return 1;
692 if (!first) {
693 first = malloc(MD_SB_BYTES + sizeof(struct bitmap_super_s));
694 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
695 *firstp = first;
696 return 0;
697 }
698
699 uuid_from_super0(uuid1, first);
700 uuid_from_super0(uuid2, second);
701 if (!same_uuid(uuid1, uuid2, 0))
702 return 2;
703 if (first->major_version != second->major_version ||
704 first->minor_version != second->minor_version ||
705 first->patch_version != second->patch_version ||
706 first->gvalid_words != second->gvalid_words ||
707 first->ctime != second->ctime ||
708 first->level != second->level ||
709 first->size != second->size ||
710 first->raid_disks != second->raid_disks )
711 return 3;
712
713 return 0;
714 }
715
716
717 static int load_super0(struct supertype *st, int fd, void **sbp, char *devname)
718 {
719 /* try to read in the superblock
720 * Return:
721 * 0 on success
722 * 1 on cannot get superblock
723 * 2 on superblock meaningless
724 */
725 unsigned long long dsize;
726 unsigned long long offset;
727 mdp_super_t *super;
728 int uuid[4];
729 struct bitmap_super_s *bsb;
730
731 if (!get_dev_size(fd, devname, &dsize))
732 return 1;
733
734 if (dsize < MD_RESERVED_SECTORS*512) {
735 if (devname)
736 fprintf(stderr, Name
737 ": %s is too small for md: size is %llu sectors.\n",
738 devname, dsize);
739 return 1;
740 }
741
742 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
743
744 offset *= 512;
745
746 ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
747
748 if (lseek64(fd, offset, 0)< 0LL) {
749 if (devname)
750 fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
751 devname, strerror(errno));
752 return 1;
753 }
754
755 super = malloc(MD_SB_BYTES + sizeof(bitmap_super_t));
756
757 if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
758 if (devname)
759 fprintf(stderr, Name ": Cannot read superblock on %s\n",
760 devname);
761 free(super);
762 return 1;
763 }
764
765 if (st->ss && st->minor_version == 9)
766 super0_swap_endian(super);
767
768 if (super->md_magic != MD_SB_MAGIC) {
769 if (devname)
770 fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
771 devname, MD_SB_MAGIC, super->md_magic);
772 free(super);
773 return 2;
774 }
775
776 if (super->major_version != 0) {
777 if (devname)
778 fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
779 devname, super->major_version);
780 free(super);
781 return 2;
782 }
783 *sbp = super;
784 if (st->ss == NULL) {
785 st->ss = &super0;
786 st->minor_version = 90;
787 st->max_devs = MD_SB_DISKS;
788 }
789
790 /* Now check on the bitmap superblock */
791 if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
792 return 0;
793 /* Read the bitmap superblock and make sure it looks
794 * valid. If it doesn't clear the bit. An --assemble --force
795 * should get that written out.
796 */
797 if (read(fd, super+1, sizeof(struct bitmap_super_s))
798 != sizeof(struct bitmap_super_s))
799 goto no_bitmap;
800
801 uuid_from_super0(uuid, super);
802 bsb = (struct bitmap_super_s *)(super+1);
803 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
804 memcmp(bsb->uuid, uuid, 16) != 0)
805 goto no_bitmap;
806 return 0;
807
808 no_bitmap:
809 super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
810
811 return 0;
812 }
813
814 static struct supertype *match_metadata_desc0(char *arg)
815 {
816 struct supertype *st = malloc(sizeof(*st));
817 if (!st) return st;
818
819 st->ss = &super0;
820 st->minor_version = 90;
821 st->max_devs = MD_SB_DISKS;
822 if (strcmp(arg, "0") == 0 ||
823 strcmp(arg, "0.90") == 0 ||
824 strcmp(arg, "default") == 0
825 )
826 return st;
827
828 st->minor_version = 9; /* flag for 'byte-swapped' */
829 if (strcmp(arg, "0.swap")==0)
830 return st;
831
832 free(st);
833 return NULL;
834 }
835
836 static __u64 avail_size0(struct supertype *st, __u64 devsize)
837 {
838 if (devsize < MD_RESERVED_SECTORS)
839 return 0ULL;
840 return MD_NEW_SIZE_SECTORS(devsize);
841 }
842
843 static int add_internal_bitmap0(struct supertype *st, void *sbv, int *chunkp,
844 int delay, int write_behind,
845 unsigned long long size, int may_change,
846 int major)
847 {
848 /*
849 * The bitmap comes immediately after the superblock and must be 60K in size
850 * at most. The default size is between 30K and 60K
851 *
852 * size is in sectors, chunk is in bytes !!!
853 */
854 unsigned long long bits;
855 unsigned long long max_bits = 60*1024*8;
856 unsigned long long min_chunk;
857 int chunk = *chunkp;
858 mdp_super_t *sb = sbv;
859 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
860
861
862 min_chunk = 4096; /* sub-page chunks don't work yet.. */
863 bits = (size * 512) / min_chunk + 1;
864 while (bits > max_bits) {
865 min_chunk *= 2;
866 bits = (bits+1)/2;
867 }
868 if (chunk == UnSet)
869 chunk = min_chunk;
870 else if (chunk < min_chunk)
871 return 0; /* chunk size too small */
872
873 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
874
875 memset(bms, 0, sizeof(*bms));
876 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
877 bms->version = __cpu_to_le32(major);
878 uuid_from_super0((int*)bms->uuid, sb);
879 bms->chunksize = __cpu_to_le32(chunk);
880 bms->daemon_sleep = __cpu_to_le32(delay);
881 bms->sync_size = __cpu_to_le64(size);
882 bms->write_behind = __cpu_to_le32(write_behind);
883 *chunkp = chunk;
884 return 1;
885 }
886
887
888 void locate_bitmap0(struct supertype *st, int fd, void *sbv)
889 {
890 unsigned long long dsize;
891 unsigned long long offset;
892
893 if (!get_dev_size(fd, NULL, &dsize))
894 return;
895
896 if (dsize < MD_RESERVED_SECTORS*512)
897 return;
898
899 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
900
901 offset *= 512;
902
903 offset += MD_SB_BYTES;
904
905 lseek64(fd, offset, 0);
906 }
907
908 int write_bitmap0(struct supertype *st, int fd, void *sbv)
909 {
910 unsigned long long dsize;
911 unsigned long long offset;
912 mdp_super_t *sb = sbv;
913
914 int rv = 0;
915
916 int towrite, n;
917 char buf[4096];
918
919 if (!get_dev_size(fd, NULL, &dsize))
920 return 1;
921
922
923 if (dsize < MD_RESERVED_SECTORS*512)
924 return -1;
925
926 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
927
928 offset *= 512;
929
930 if (lseek64(fd, offset + 4096, 0)< 0LL)
931 return 3;
932
933
934 if (write(fd, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t)) !=
935 sizeof(bitmap_super_t))
936 return -2;
937 towrite = 64*1024 - MD_SB_BYTES - sizeof(bitmap_super_t);
938 memset(buf, 0xff, sizeof(buf));
939 while (towrite > 0) {
940 n = towrite;
941 if (n > sizeof(buf))
942 n = sizeof(buf);
943 n = write(fd, buf, n);
944 if (n > 0)
945 towrite -= n;
946 else
947 break;
948 }
949 fsync(fd);
950 if (towrite)
951 rv = -2;
952
953 return rv;
954 }
955
956 struct superswitch super0 = {
957 #ifndef MDASSEMBLE
958 .examine_super = examine_super0,
959 .brief_examine_super = brief_examine_super0,
960 .detail_super = detail_super0,
961 .brief_detail_super = brief_detail_super0,
962 .export_super = export_super0,
963 #endif
964 .match_home = match_home0,
965 .uuid_from_super = uuid_from_super0,
966 .getinfo_super = getinfo_super0,
967 .update_super = update_super0,
968 .init_super = init_super0,
969 .add_to_super = add_to_super0,
970 .store_super = store_super0,
971 .write_init_super = write_init_super0,
972 .compare_super = compare_super0,
973 .load_super = load_super0,
974 .match_metadata_desc = match_metadata_desc0,
975 .avail_size = avail_size0,
976 .add_internal_bitmap = add_internal_bitmap0,
977 .locate_bitmap = locate_bitmap0,
978 .write_bitmap = write_bitmap0,
979 .major = 0,
980 .swapuuid = 0,
981 };