]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super0.c
imsm: FIX: Use function to obtain array layout
[thirdparty/mdadm.git] / super0.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 #define HAVE_STDINT_H 1
26 #include "mdadm.h"
27 #include "sha1.h"
28 /*
29 * All handling for the 0.90.0 version superblock is in
30 * this file.
31 * This includes:
32 * - finding, loading, and writing the superblock.
33 * - initialising a new superblock
34 * - printing the superblock for --examine
35 * - printing part of the superblock for --detail
36 * .. other stuff
37 */
38
39
40 static unsigned long calc_sb0_csum(mdp_super_t *super)
41 {
42 unsigned long csum = super->sb_csum;
43 unsigned long newcsum;
44 super->sb_csum= 0 ;
45 newcsum = calc_csum(super, MD_SB_BYTES);
46 super->sb_csum = csum;
47 return newcsum;
48 }
49
50
51 static void super0_swap_endian(struct mdp_superblock_s *sb)
52 {
53 /* as super0 superblocks are host-endian, it is sometimes
54 * useful to be able to swap the endianness
55 * as (almost) everything is u32's we byte-swap every 4byte
56 * number.
57 * We then also have to swap the events_hi and events_lo
58 */
59 char *sbc = (char *)sb;
60 __u32 t32;
61 int i;
62
63 for (i=0; i < MD_SB_BYTES ; i+=4) {
64 char t = sbc[i];
65 sbc[i] = sbc[i+3];
66 sbc[i+3] = t;
67 t=sbc[i+1];
68 sbc[i+1]=sbc[i+2];
69 sbc[i+2]=t;
70 }
71 t32 = sb->events_hi;
72 sb->events_hi = sb->events_lo;
73 sb->events_lo = t32;
74
75 t32 = sb->cp_events_hi;
76 sb->cp_events_hi = sb->cp_events_lo;
77 sb->cp_events_lo = t32;
78
79 }
80
81 #ifndef MDASSEMBLE
82
83 static void examine_super0(struct supertype *st, char *homehost)
84 {
85 mdp_super_t *sb = st->sb;
86 time_t atime;
87 int d;
88 int delta_extra = 0;
89 char *c;
90
91 printf(" Magic : %08x\n", sb->md_magic);
92 printf(" Version : %d.%02d.%02d\n", sb->major_version, sb->minor_version,
93 sb->patch_version);
94 if (sb->minor_version >= 90) {
95 printf(" UUID : %08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
96 sb->set_uuid2, sb->set_uuid3);
97 if (homehost) {
98 char buf[20];
99 void *hash = sha1_buffer(homehost,
100 strlen(homehost),
101 buf);
102 if (memcmp(&sb->set_uuid2, hash, 8)==0)
103 printf(" (local to host %s)", homehost);
104 }
105 printf("\n");
106 } else
107 printf(" UUID : %08x\n", sb->set_uuid0);
108
109 if (sb->not_persistent)
110 printf(" Eedk : not persistent\n");
111
112 atime = sb->ctime;
113 printf(" Creation Time : %.24s\n", ctime(&atime));
114 c=map_num(pers, sb->level);
115 printf(" Raid Level : %s\n", c?c:"-unknown-");
116 if ((int)sb->level > 0) {
117 int ddsks=0;
118 printf(" Used Dev Size : %d%s\n", sb->size,
119 human_size((long long)sb->size<<10));
120 switch(sb->level) {
121 case 1: ddsks=1;break;
122 case 4:
123 case 5: ddsks = sb->raid_disks-1; break;
124 case 6: ddsks = sb->raid_disks-2; break;
125 case 10: ddsks = sb->raid_disks / (sb->layout&255) / ((sb->layout>>8)&255);
126 }
127 if (ddsks)
128 printf(" Array Size : %llu%s\n", (unsigned long long)ddsks * sb->size,
129 human_size(ddsks*(long long)sb->size<<10));
130 }
131 printf(" Raid Devices : %d\n", sb->raid_disks);
132 printf(" Total Devices : %d\n", sb->nr_disks);
133 printf("Preferred Minor : %d\n", sb->md_minor);
134 printf("\n");
135 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
136 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)sb->reshape_position/2, human_size((long long)sb->reshape_position<<9));
137 if (sb->delta_disks) {
138 printf(" Delta Devices : %d", sb->delta_disks);
139 printf(" (%d->%d)\n", sb->raid_disks-sb->delta_disks, sb->raid_disks);
140 if (((int)sb->delta_disks) < 0)
141 delta_extra = - sb->delta_disks;
142 }
143 if (sb->new_level != sb->level) {
144 c = map_num(pers, sb->new_level);
145 printf(" New Level : %s\n", c?c:"-unknown-");
146 }
147 if (sb->new_layout != sb->layout) {
148 if (sb->level == 5) {
149 c = map_num(r5layout, sb->new_layout);
150 printf(" New Layout : %s\n", c?c:"-unknown-");
151 }
152 if (sb->level == 6) {
153 c = map_num(r6layout, sb->new_layout);
154 printf(" New Layout : %s\n", c?c:"-unknown-");
155 }
156 if (sb->level == 10) {
157 printf(" New Layout : near=%d, %s=%d\n",
158 sb->new_layout&255,
159 (sb->new_layout&0x10000)?"offset":"far",
160 (sb->new_layout>>8)&255);
161 }
162 }
163 if (sb->new_chunk != sb->chunk_size)
164 printf(" New Chunksize : %d\n", sb->new_chunk);
165 printf("\n");
166 }
167 atime = sb->utime;
168 printf(" Update Time : %.24s\n", ctime(&atime));
169 printf(" State : %s\n",
170 (sb->state&(1<<MD_SB_CLEAN))?"clean":"active");
171 if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
172 printf("Internal Bitmap : present\n");
173 printf(" Active Devices : %d\n", sb->active_disks);
174 printf("Working Devices : %d\n", sb->working_disks);
175 printf(" Failed Devices : %d\n", sb->failed_disks);
176 printf(" Spare Devices : %d\n", sb->spare_disks);
177 if (calc_sb0_csum(sb) == sb->sb_csum)
178 printf(" Checksum : %x - correct\n", sb->sb_csum);
179 else
180 printf(" Checksum : %x - expected %lx\n", sb->sb_csum, calc_sb0_csum(sb));
181 printf(" Events : %llu\n",
182 ((unsigned long long)sb->events_hi << 32)
183 + sb->events_lo);
184 printf("\n");
185 if (sb->level == 5) {
186 c = map_num(r5layout, sb->layout);
187 printf(" Layout : %s\n", c?c:"-unknown-");
188 }
189 if (sb->level == 6) {
190 c = map_num(r6layout, sb->layout);
191 printf(" Layout : %s\n", c?c:"-unknown-");
192 }
193 if (sb->level == 10) {
194 printf(" Layout :");
195 print_r10_layout(sb->layout);
196 printf("\n");
197 }
198 switch(sb->level) {
199 case 0:
200 case 4:
201 case 5:
202 case 6:
203 case 10:
204 printf(" Chunk Size : %dK\n", sb->chunk_size/1024);
205 break;
206 case -1:
207 printf(" Rounding : %dK\n", sb->chunk_size/1024);
208 break;
209 default: break;
210 }
211 printf("\n");
212 printf(" Number Major Minor RaidDevice State\n");
213 for (d= -1; d<(signed int)(sb->raid_disks+delta_extra + sb->spare_disks); d++) {
214 mdp_disk_t *dp;
215 char *dv;
216 char nb[5];
217 int wonly;
218 if (d>=0) dp = &sb->disks[d];
219 else dp = &sb->this_disk;
220 snprintf(nb, sizeof(nb), "%4d", d);
221 printf("%4s %5d %5d %5d %5d ", d < 0 ? "this" : nb,
222 dp->number, dp->major, dp->minor, dp->raid_disk);
223 wonly = dp->state & (1<<MD_DISK_WRITEMOSTLY);
224 dp->state &= ~(1<<MD_DISK_WRITEMOSTLY);
225 if (dp->state & (1<<MD_DISK_FAULTY)) printf(" faulty");
226 if (dp->state & (1<<MD_DISK_ACTIVE)) printf(" active");
227 if (dp->state & (1<<MD_DISK_SYNC)) printf(" sync");
228 if (dp->state & (1<<MD_DISK_REMOVED)) printf(" removed");
229 if (wonly) printf(" write-mostly");
230 if (dp->state == 0) printf(" spare");
231 if ((dv=map_dev(dp->major, dp->minor, 0)))
232 printf(" %s", dv);
233 printf("\n");
234 if (d == -1) printf("\n");
235 }
236 }
237
238 static void brief_examine_super0(struct supertype *st, int verbose)
239 {
240 mdp_super_t *sb = st->sb;
241 char *c=map_num(pers, sb->level);
242 char devname[20];
243
244 sprintf(devname, "/dev/md%d", sb->md_minor);
245
246 if (verbose) {
247 printf("ARRAY %s level=%s num-devices=%d",
248 devname,
249 c?c:"-unknown-", sb->raid_disks);
250 } else
251 printf("ARRAY %s", devname);
252
253 if (sb->minor_version >= 90)
254 printf(" UUID=%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
255 sb->set_uuid2, sb->set_uuid3);
256 else
257 printf(" UUID=%08x", sb->set_uuid0);
258 printf("\n");
259 }
260
261 static void export_examine_super0(struct supertype *st)
262 {
263 mdp_super_t *sb = st->sb;
264
265 printf("MD_LEVEL=%s\n", map_num(pers, sb->level));
266 printf("MD_DEVICES=%d\n", sb->raid_disks);
267 if (sb->minor_version >= 90)
268 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
269 sb->set_uuid0, sb->set_uuid1,
270 sb->set_uuid2, sb->set_uuid3);
271 else
272 printf("MD_UUID=%08x\n", sb->set_uuid0);
273 printf("MD_UPDATE_TIME=%llu\n",
274 __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL);
275 printf("MD_EVENTS=%llu\n",
276 ((unsigned long long)sb->events_hi << 32)
277 + sb->events_lo);
278 }
279
280 static void detail_super0(struct supertype *st, char *homehost)
281 {
282 mdp_super_t *sb = st->sb;
283 printf(" UUID : ");
284 if (sb->minor_version >= 90)
285 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
286 sb->set_uuid2, sb->set_uuid3);
287 else
288 printf("%08x", sb->set_uuid0);
289 if (homehost) {
290 char buf[20];
291 void *hash = sha1_buffer(homehost,
292 strlen(homehost),
293 buf);
294 if (memcmp(&sb->set_uuid2, hash, 8)==0)
295 printf(" (local to host %s)", homehost);
296 }
297 printf("\n Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
298 }
299
300 static void brief_detail_super0(struct supertype *st)
301 {
302 mdp_super_t *sb = st->sb;
303 printf(" UUID=");
304 if (sb->minor_version >= 90)
305 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
306 sb->set_uuid2, sb->set_uuid3);
307 else
308 printf("%08x", sb->set_uuid0);
309 }
310 #endif
311
312 static int match_home0(struct supertype *st, char *homehost)
313 {
314 mdp_super_t *sb = st->sb;
315 char buf[20];
316 char *hash;
317
318 if (!homehost)
319 return 0;
320 hash = sha1_buffer(homehost,
321 strlen(homehost),
322 buf);
323
324 return (memcmp(&sb->set_uuid2, hash, 8)==0);
325 }
326
327 static void uuid_from_super0(struct supertype *st, int uuid[4])
328 {
329 mdp_super_t *super = st->sb;
330 uuid[0] = super->set_uuid0;
331 if (super->minor_version >= 90) {
332 uuid[1] = super->set_uuid1;
333 uuid[2] = super->set_uuid2;
334 uuid[3] = super->set_uuid3;
335 } else {
336 uuid[1] = 0;
337 uuid[2] = 0;
338 uuid[3] = 0;
339 }
340 }
341
342 static void getinfo_super0(struct supertype *st, struct mdinfo *info, char *map)
343 {
344 mdp_super_t *sb = st->sb;
345 int working = 0;
346 int i;
347 int map_disks = info->array.raid_disks;
348
349 memset(info, 0, sizeof(*info));
350 info->array.major_version = sb->major_version;
351 info->array.minor_version = sb->minor_version;
352 info->array.patch_version = sb->patch_version;
353 info->array.raid_disks = sb->raid_disks;
354 info->array.level = sb->level;
355 info->array.layout = sb->layout;
356 info->array.md_minor = sb->md_minor;
357 info->array.ctime = sb->ctime;
358 info->array.utime = sb->utime;
359 info->array.chunk_size = sb->chunk_size;
360 info->array.state = sb->state;
361 info->component_size = sb->size*2;
362
363 info->disk.state = sb->this_disk.state;
364 info->disk.major = sb->this_disk.major;
365 info->disk.minor = sb->this_disk.minor;
366 info->disk.raid_disk = sb->this_disk.raid_disk;
367 info->disk.number = sb->this_disk.number;
368
369 info->events = md_event(sb);
370 info->data_offset = 0;
371
372 sprintf(info->text_version, "0.%d", sb->minor_version);
373 info->safe_mode_delay = 200;
374
375 uuid_from_super0(st, info->uuid);
376
377 info->recovery_start = MaxSector;
378 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
379 info->reshape_active = 1;
380 info->reshape_progress = sb->reshape_position;
381 info->new_level = sb->new_level;
382 info->delta_disks = sb->delta_disks;
383 info->new_layout = sb->new_layout;
384 info->new_chunk = sb->new_chunk;
385 if (info->delta_disks < 0)
386 info->array.raid_disks -= info->delta_disks;
387 } else
388 info->reshape_active = 0;
389
390 sprintf(info->name, "%d", sb->md_minor);
391 /* work_disks is calculated rather than read directly */
392 for (i=0; i < MD_SB_DISKS; i++)
393 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
394 (sb->disks[i].raid_disk < (unsigned)info->array.raid_disks) &&
395 (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
396 !(sb->disks[i].state & (1<<MD_DISK_FAULTY))) {
397 working ++;
398 if (map && i < map_disks)
399 map[i] = 1;
400 } else if (map && i < map_disks)
401 map[i] = 0;
402 info->array.working_disks = working;
403 }
404
405 static struct mdinfo *container_content0(struct supertype *st, char *subarray)
406 {
407 struct mdinfo *info;
408
409 if (subarray)
410 return NULL;
411
412 info = malloc(sizeof(*info));
413 getinfo_super0(st, info, NULL);
414 return info;
415 }
416
417 static int update_super0(struct supertype *st, struct mdinfo *info,
418 char *update,
419 char *devname, int verbose,
420 int uuid_set, char *homehost)
421 {
422 /* NOTE: for 'assemble' and 'force' we need to return non-zero
423 * if any change was made. For others, the return value is
424 * ignored.
425 */
426 int rv = 0;
427 mdp_super_t *sb = st->sb;
428 if (strcmp(update, "sparc2.2")==0 ) {
429 /* 2.2 sparc put the events in the wrong place
430 * So we copy the tail of the superblock
431 * up 4 bytes before continuing
432 */
433 __u32 *sb32 = (__u32*)sb;
434 memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
435 sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
436 (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
437 if (verbose >= 0)
438 fprintf (stderr, Name ": adjusting superblock of %s for 2.2/sparc compatability.\n",
439 devname);
440 } else if (strcmp(update, "super-minor") ==0) {
441 sb->md_minor = info->array.md_minor;
442 if (verbose > 0)
443 fprintf(stderr, Name ": updating superblock of %s with minor number %d\n",
444 devname, info->array.md_minor);
445 } else if (strcmp(update, "summaries") == 0) {
446 unsigned int i;
447 /* set nr_disks, active_disks, working_disks,
448 * failed_disks, spare_disks based on disks[]
449 * array in superblock.
450 * Also make sure extra slots aren't 'failed'
451 */
452 sb->nr_disks = sb->active_disks =
453 sb->working_disks = sb->failed_disks =
454 sb->spare_disks = 0;
455 for (i=0; i < MD_SB_DISKS ; i++)
456 if (sb->disks[i].major ||
457 sb->disks[i].minor) {
458 int state = sb->disks[i].state;
459 if (state & (1<<MD_DISK_REMOVED))
460 continue;
461 sb->nr_disks++;
462 if (state & (1<<MD_DISK_ACTIVE))
463 sb->active_disks++;
464 if (state & (1<<MD_DISK_FAULTY))
465 sb->failed_disks++;
466 else
467 sb->working_disks++;
468 if (state == 0)
469 sb->spare_disks++;
470 } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
471 sb->disks[i].state = 0;
472 } else if (strcmp(update, "force-one")==0) {
473 /* Not enough devices for a working array, so
474 * bring this one up-to-date.
475 */
476 __u32 ehi = sb->events_hi, elo = sb->events_lo;
477 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
478 sb->events_lo = (info->events) & 0xFFFFFFFF;
479 if (sb->events_hi != ehi ||
480 sb->events_lo != elo)
481 rv = 1;
482 } else if (strcmp(update, "force-array")==0) {
483 /* degraded array and 'force' requested, so
484 * maybe need to mark it 'clean'
485 */
486 if ((sb->level == 5 || sb->level == 4 || sb->level == 6) &&
487 (sb->state & (1 << MD_SB_CLEAN)) == 0) {
488 /* need to force clean */
489 sb->state |= (1 << MD_SB_CLEAN);
490 rv = 1;
491 }
492 } else if (strcmp(update, "assemble")==0) {
493 int d = info->disk.number;
494 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
495 int mask = (1<<MD_DISK_WRITEMOSTLY);
496 int add = 0;
497 if (sb->minor_version >= 91)
498 /* During reshape we don't insist on everything
499 * being marked 'sync'
500 */
501 add = (1<<MD_DISK_SYNC);
502 if (((sb->disks[d].state & ~mask) | add)
503 != (unsigned)info->disk.state) {
504 sb->disks[d].state = info->disk.state | wonly;
505 rv = 1;
506 }
507 if (info->reshape_active &&
508 sb->minor_version > 90 && (sb->reshape_position+1) != 0 &&
509 info->delta_disks >= 0 &&
510 info->reshape_progress < sb->reshape_position) {
511 sb->reshape_position = info->reshape_progress;
512 rv = 1;
513 }
514 if (info->reshape_active &&
515 sb->minor_version > 90 && (sb->reshape_position+1) != 0 &&
516 info->delta_disks < 0 &&
517 info->reshape_progress > sb->reshape_position) {
518 sb->reshape_position = info->reshape_progress;
519 rv = 1;
520 }
521 } else if (strcmp(update, "linear-grow-new") == 0) {
522 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
523 sb->disks[info->disk.number].number = info->disk.number;
524 sb->disks[info->disk.number].major = info->disk.major;
525 sb->disks[info->disk.number].minor = info->disk.minor;
526 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
527 sb->disks[info->disk.number].state = info->disk.state;
528 sb->this_disk = sb->disks[info->disk.number];
529 } else if (strcmp(update, "linear-grow-update") == 0) {
530 sb->raid_disks = info->array.raid_disks;
531 sb->nr_disks = info->array.nr_disks;
532 sb->active_disks = info->array.active_disks;
533 sb->working_disks = info->array.working_disks;
534 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
535 sb->disks[info->disk.number].number = info->disk.number;
536 sb->disks[info->disk.number].major = info->disk.major;
537 sb->disks[info->disk.number].minor = info->disk.minor;
538 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
539 sb->disks[info->disk.number].state = info->disk.state;
540 } else if (strcmp(update, "resync") == 0) {
541 /* make sure resync happens */
542 sb->state &= ~(1<<MD_SB_CLEAN);
543 sb->recovery_cp = 0;
544 } else if (strcmp(update, "homehost") == 0 &&
545 homehost) {
546 uuid_set = 0;
547 update = "uuid";
548 info->uuid[0] = sb->set_uuid0;
549 info->uuid[1] = sb->set_uuid1;
550 } else if (strcmp(update, "uuid") == 0) {
551 if (!uuid_set && homehost) {
552 char buf[20];
553 char *hash = sha1_buffer(homehost,
554 strlen(homehost),
555 buf);
556 memcpy(info->uuid+2, hash, 8);
557 }
558 sb->set_uuid0 = info->uuid[0];
559 sb->set_uuid1 = info->uuid[1];
560 sb->set_uuid2 = info->uuid[2];
561 sb->set_uuid3 = info->uuid[3];
562 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
563 struct bitmap_super_s *bm;
564 bm = (struct bitmap_super_s*)(sb+1);
565 uuid_from_super0(st, (int*)bm->uuid);
566 }
567 } else if (strcmp(update, "no-bitmap") == 0) {
568 sb->state &= ~(1<<MD_SB_BITMAP_PRESENT);
569 } else if (strcmp(update, "_reshape_progress")==0)
570 sb->reshape_position = info->reshape_progress;
571 else
572 rv = -1;
573
574 sb->sb_csum = calc_sb0_csum(sb);
575 return rv;
576 }
577
578 /*
579 * For verion-0 superblock, the homehost is 'stored' in the
580 * uuid. 8 bytes for a hash of the host leaving 8 bytes
581 * of random material.
582 * We use the first 8 bytes (64bits) of the sha1 of the
583 * host name
584 */
585
586
587 static int init_super0(struct supertype *st, mdu_array_info_t *info,
588 unsigned long long size, char *ignored_name, char *homehost,
589 int *uuid)
590 {
591 mdp_super_t *sb;
592 int spares;
593
594 if (posix_memalign((void**)&sb, 4096,
595 MD_SB_BYTES + ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
596 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
597 return 0;
598 }
599 memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
600
601 st->sb = sb;
602 if (info == NULL) {
603 /* zeroing the superblock */
604 return 0;
605 }
606
607 spares = info->working_disks - info->active_disks;
608 if (info->raid_disks + spares > MD_SB_DISKS) {
609 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
610 info->raid_disks , spares, MD_SB_DISKS);
611 return 0;
612 }
613
614 sb->md_magic = MD_SB_MAGIC;
615 sb->major_version = 0;
616 sb->minor_version = 90;
617 sb->patch_version = 0;
618 sb->gvalid_words = 0; /* ignored */
619 sb->ctime = time(0);
620 sb->level = info->level;
621 if (size != (unsigned long long)info->size)
622 return 0;
623 sb->size = info->size;
624 sb->nr_disks = info->nr_disks;
625 sb->raid_disks = info->raid_disks;
626 sb->md_minor = info->md_minor;
627 sb->not_persistent = 0;
628 if (uuid) {
629 sb->set_uuid0 = uuid[0];
630 sb->set_uuid1 = uuid[1];
631 sb->set_uuid2 = uuid[2];
632 sb->set_uuid3 = uuid[3];
633 } else {
634 int rfd = open("/dev/urandom", O_RDONLY);
635 if (rfd < 0 || read(rfd, &sb->set_uuid0, 4) != 4)
636 sb->set_uuid0 = random();
637 if (rfd < 0 || read(rfd, &sb->set_uuid1, 12) != 12) {
638 sb->set_uuid1 = random();
639 sb->set_uuid2 = random();
640 sb->set_uuid3 = random();
641 }
642 if (rfd >= 0)
643 close(rfd);
644 }
645 if (homehost) {
646 char buf[20];
647 char *hash = sha1_buffer(homehost,
648 strlen(homehost),
649 buf);
650 memcpy(&sb->set_uuid2, hash, 8);
651 }
652
653 sb->utime = sb->ctime;
654 sb->state = info->state;
655 sb->active_disks = info->active_disks;
656 sb->working_disks = info->working_disks;
657 sb->failed_disks = info->failed_disks;
658 sb->spare_disks = info->spare_disks;
659 sb->events_hi = 0;
660 sb->events_lo = 1;
661
662 sb->layout = info->layout;
663 sb->chunk_size = info->chunk_size;
664
665 return 1;
666 }
667
668 struct devinfo {
669 int fd;
670 char *devname;
671 mdu_disk_info_t disk;
672 struct devinfo *next;
673 };
674
675 #ifndef MDASSEMBLE
676 /* Add a device to the superblock being created */
677 static int add_to_super0(struct supertype *st, mdu_disk_info_t *dinfo,
678 int fd, char *devname)
679 {
680 mdp_super_t *sb = st->sb;
681 mdp_disk_t *dk = &sb->disks[dinfo->number];
682 struct devinfo *di, **dip;
683
684 dk->number = dinfo->number;
685 dk->major = dinfo->major;
686 dk->minor = dinfo->minor;
687 dk->raid_disk = dinfo->raid_disk;
688 dk->state = dinfo->state;
689
690 sb->this_disk = sb->disks[dinfo->number];
691 sb->sb_csum = calc_sb0_csum(sb);
692
693 dip = (struct devinfo **)&st->info;
694 while (*dip)
695 dip = &(*dip)->next;
696 di = malloc(sizeof(struct devinfo));
697 di->fd = fd;
698 di->devname = devname;
699 di->disk = *dinfo;
700 di->next = NULL;
701 *dip = di;
702
703 return 0;
704 }
705 #endif
706
707 static int store_super0(struct supertype *st, int fd)
708 {
709 unsigned long long dsize;
710 unsigned long long offset;
711 mdp_super_t *super = st->sb;
712
713 if (!get_dev_size(fd, NULL, &dsize))
714 return 1;
715
716 if (dsize < MD_RESERVED_SECTORS*512)
717 return 2;
718
719 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
720
721 offset *= 512;
722
723 if (lseek64(fd, offset, 0)< 0LL)
724 return 3;
725
726 if (write(fd, super, sizeof(*super)) != sizeof(*super))
727 return 4;
728
729 if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
730 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
731 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
732 if (write(fd, bm, ROUND_UP(sizeof(*bm),4096)) !=
733 ROUND_UP(sizeof(*bm),4096))
734 return 5;
735 }
736
737 fsync(fd);
738 return 0;
739 }
740
741 #ifndef MDASSEMBLE
742 static int write_init_super0(struct supertype *st)
743 {
744 mdp_super_t *sb = st->sb;
745 int rv = 0;
746 struct devinfo *di;
747
748 for (di = st->info ; di && ! rv ; di = di->next) {
749
750 if (di->disk.state == 1)
751 continue;
752 if (di->fd == -1)
753 continue;
754 while (Kill(di->devname, NULL, 0, 1, 1) == 0)
755 ;
756
757 sb->disks[di->disk.number].state &= ~(1<<MD_DISK_FAULTY);
758
759 sb->this_disk = sb->disks[di->disk.number];
760 sb->sb_csum = calc_sb0_csum(sb);
761 rv = store_super0(st, di->fd);
762
763 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
764 rv = st->ss->write_bitmap(st, di->fd);
765
766 if (rv)
767 fprintf(stderr,
768 Name ": failed to write superblock to %s\n",
769 di->devname);
770 }
771 return rv;
772 }
773 #endif
774
775 static int compare_super0(struct supertype *st, struct supertype *tst)
776 {
777 /*
778 * return:
779 * 0 same, or first was empty, and second was copied
780 * 1 second had wrong number
781 * 2 wrong uuid
782 * 3 wrong other info
783 */
784 mdp_super_t *first = st->sb;
785 mdp_super_t *second = tst->sb;
786 int uuid1[4], uuid2[4];
787
788 if (second->md_magic != MD_SB_MAGIC)
789 return 1;
790 if (!first) {
791 if (posix_memalign((void**)&first, 4096,
792 MD_SB_BYTES +
793 ROUND_UP(sizeof(struct bitmap_super_s), 4096)) != 0) {
794 fprintf(stderr, Name
795 ": %s could not allocate superblock\n", __func__);
796 return 1;
797 }
798 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
799 st->sb = first;
800 return 0;
801 }
802
803 uuid_from_super0(st, uuid1);
804 uuid_from_super0(tst, uuid2);
805 if (!same_uuid(uuid1, uuid2, 0))
806 return 2;
807 if (first->major_version != second->major_version ||
808 first->minor_version != second->minor_version ||
809 first->patch_version != second->patch_version ||
810 first->gvalid_words != second->gvalid_words ||
811 first->ctime != second->ctime ||
812 first->level != second->level ||
813 first->size != second->size ||
814 first->raid_disks != second->raid_disks )
815 return 3;
816
817 return 0;
818 }
819
820
821 static void free_super0(struct supertype *st);
822
823 static int load_super0(struct supertype *st, int fd, char *devname)
824 {
825 /* try to read in the superblock
826 * Return:
827 * 0 on success
828 * 1 on cannot get superblock
829 * 2 on superblock meaningless
830 */
831 unsigned long long dsize;
832 unsigned long long offset;
833 mdp_super_t *super;
834 int uuid[4];
835 struct bitmap_super_s *bsb;
836
837 free_super0(st);
838
839 if (!get_dev_size(fd, devname, &dsize))
840 return 1;
841
842 if (dsize < MD_RESERVED_SECTORS*512) {
843 if (devname)
844 fprintf(stderr, Name
845 ": %s is too small for md: size is %llu sectors.\n",
846 devname, dsize);
847 return 1;
848 }
849
850 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
851
852 offset *= 512;
853
854 ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
855
856 if (lseek64(fd, offset, 0)< 0LL) {
857 if (devname)
858 fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
859 devname, strerror(errno));
860 return 1;
861 }
862
863 if (posix_memalign((void**)&super, 4096,
864 MD_SB_BYTES +
865 ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
866 fprintf(stderr, Name
867 ": %s could not allocate superblock\n", __func__);
868 return 1;
869 }
870
871 if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
872 if (devname)
873 fprintf(stderr, Name ": Cannot read superblock on %s\n",
874 devname);
875 free(super);
876 return 1;
877 }
878
879 if (st->ss && st->minor_version == 9)
880 super0_swap_endian(super);
881
882 if (super->md_magic != MD_SB_MAGIC) {
883 if (devname)
884 fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
885 devname, MD_SB_MAGIC, super->md_magic);
886 free(super);
887 return 2;
888 }
889
890 if (super->major_version != 0) {
891 if (devname)
892 fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
893 devname, super->major_version);
894 free(super);
895 return 2;
896 }
897 st->sb = super;
898
899 if (st->ss == NULL) {
900 st->ss = &super0;
901 st->minor_version = super->minor_version;
902 st->max_devs = MD_SB_DISKS;
903 st->info = NULL;
904 }
905
906 /* Now check on the bitmap superblock */
907 if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
908 return 0;
909 /* Read the bitmap superblock and make sure it looks
910 * valid. If it doesn't clear the bit. An --assemble --force
911 * should get that written out.
912 */
913 if (read(fd, super+1, ROUND_UP(sizeof(struct bitmap_super_s),4096))
914 != ROUND_UP(sizeof(struct bitmap_super_s),4096))
915 goto no_bitmap;
916
917 uuid_from_super0(st, uuid);
918 bsb = (struct bitmap_super_s *)(super+1);
919 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
920 memcmp(bsb->uuid, uuid, 16) != 0)
921 goto no_bitmap;
922 return 0;
923
924 no_bitmap:
925 super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
926
927 return 0;
928 }
929
930 static struct supertype *match_metadata_desc0(char *arg)
931 {
932 struct supertype *st = malloc(sizeof(*st));
933 if (!st) return st;
934
935 memset(st, 0, sizeof(*st));
936 st->container_dev = NoMdDev;
937 st->ss = &super0;
938 st->info = NULL;
939 st->minor_version = 90;
940 st->max_devs = MD_SB_DISKS;
941 st->sb = NULL;
942 /* we sometimes get 00.90 */
943 while (arg[0] == '0' && arg[1] == '0')
944 arg++;
945 if (strcmp(arg, "0") == 0 ||
946 #ifdef DEFAULT_OLD_METADATA /* ifndef in super1.c */
947 strcmp(arg, "default") == 0 ||
948 #endif /* DEFAULT_OLD_METADATA */
949 strcmp(arg, "0.90") == 0 ||
950 strcmp(arg, "") == 0 /* no metadata - i.e. non_persistent */
951 )
952 return st;
953
954 st->minor_version = 91; /* reshape in progress */
955 if (strcmp(arg, "0.91") == 0) /* For dup_super support */
956 return st;
957
958 st->minor_version = 9; /* flag for 'byte-swapped' */
959 if (strcmp(arg, "0.swap")==0 ||
960 strcmp(arg, "0.9") == 0) /* For dup_super support */
961 return st;
962
963 free(st);
964 return NULL;
965 }
966
967 static __u64 avail_size0(struct supertype *st, __u64 devsize)
968 {
969 if (devsize < MD_RESERVED_SECTORS)
970 return 0ULL;
971 return MD_NEW_SIZE_SECTORS(devsize);
972 }
973
974 static int add_internal_bitmap0(struct supertype *st, int *chunkp,
975 int delay, int write_behind,
976 unsigned long long size, int may_change,
977 int major)
978 {
979 /*
980 * The bitmap comes immediately after the superblock and must be 60K in size
981 * at most. The default size is between 30K and 60K
982 *
983 * size is in sectors, chunk is in bytes !!!
984 */
985 unsigned long long bits;
986 unsigned long long max_bits = (60*1024 - sizeof(bitmap_super_t))*8;
987 unsigned long long min_chunk;
988 int chunk = *chunkp;
989 mdp_super_t *sb = st->sb;
990 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
991
992
993 min_chunk = 4096; /* sub-page chunks don't work yet.. */
994 bits = (size * 512) / min_chunk + 1;
995 while (bits > max_bits) {
996 min_chunk *= 2;
997 bits = (bits+1)/2;
998 }
999 if (chunk == UnSet) {
1000 /* A chunk size less than a few Megabytes gives poor
1001 * performance without increasing resync noticeably
1002 */
1003 chunk = min_chunk;
1004 if (chunk < 64*1024*1024)
1005 chunk = 64*1024*1024;
1006 } else if ((unsigned long long)chunk < min_chunk)
1007 return 0; /* chunk size too small */
1008
1009 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
1010
1011 memset(bms, 0, sizeof(*bms));
1012 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
1013 bms->version = __cpu_to_le32(major);
1014 uuid_from_super0(st, (int*)bms->uuid);
1015 bms->chunksize = __cpu_to_le32(chunk);
1016 bms->daemon_sleep = __cpu_to_le32(delay);
1017 bms->sync_size = __cpu_to_le64(size);
1018 bms->write_behind = __cpu_to_le32(write_behind);
1019 *chunkp = chunk;
1020 return 1;
1021 }
1022
1023
1024 static void locate_bitmap0(struct supertype *st, int fd)
1025 {
1026 unsigned long long dsize;
1027 unsigned long long offset;
1028
1029 if (!get_dev_size(fd, NULL, &dsize))
1030 return;
1031
1032 if (dsize < MD_RESERVED_SECTORS*512)
1033 return;
1034
1035 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1036
1037 offset *= 512;
1038
1039 offset += MD_SB_BYTES;
1040
1041 lseek64(fd, offset, 0);
1042 }
1043
1044 static int write_bitmap0(struct supertype *st, int fd)
1045 {
1046 unsigned long long dsize;
1047 unsigned long long offset;
1048 mdp_super_t *sb = st->sb;
1049
1050 int rv = 0;
1051
1052 int towrite, n;
1053 char abuf[4096+4096];
1054 char *buf = (char*)(((long)(abuf+4096))&~4095L);
1055
1056 if (!get_dev_size(fd, NULL, &dsize))
1057 return 1;
1058
1059
1060 if (dsize < MD_RESERVED_SECTORS*512)
1061 return -1;
1062
1063 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1064
1065 offset *= 512;
1066
1067 if (lseek64(fd, offset + 4096, 0)< 0LL)
1068 return 3;
1069
1070 memset(buf, 0xff, 4096);
1071 memcpy(buf, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
1072 towrite = 60*1024;
1073 while (towrite > 0) {
1074 n = towrite;
1075 if (n > 4096)
1076 n = 4096;
1077 n = write(fd, buf, n);
1078 if (n > 0)
1079 towrite -= n;
1080 else
1081 break;
1082 memset(buf, 0xff, 4096);
1083 }
1084 fsync(fd);
1085 if (towrite)
1086 rv = -2;
1087
1088 return rv;
1089 }
1090
1091 static void free_super0(struct supertype *st)
1092 {
1093 if (st->sb)
1094 free(st->sb);
1095 while (st->info) {
1096 struct devinfo *di = st->info;
1097 st->info = di->next;
1098 if (di->fd >= 0)
1099 close(di->fd);
1100 free(di);
1101 }
1102 st->sb = NULL;
1103 }
1104
1105 #ifndef MDASSEMBLE
1106 static int validate_geometry0(struct supertype *st, int level,
1107 int layout, int raiddisks,
1108 int *chunk, unsigned long long size,
1109 char *subdev, unsigned long long *freesize,
1110 int verbose)
1111 {
1112 unsigned long long ldsize;
1113 int fd;
1114
1115 if (level == LEVEL_CONTAINER) {
1116 if (verbose)
1117 fprintf(stderr, Name ": 0.90 metadata does not support containers\n");
1118 return 0;
1119 }
1120 if (raiddisks > MD_SB_DISKS) {
1121 if (verbose)
1122 fprintf(stderr, Name ": 0.90 metadata supports at most %d devices per array\n",
1123 MD_SB_DISKS);
1124 return 0;
1125 }
1126 if (size > (0x7fffffffULL<<9)) {
1127 if (verbose)
1128 fprintf(stderr, Name ": 0.90 metadata supports at most 2 terrabytes per device\n");
1129 return 0;
1130 }
1131 if (chunk && *chunk == UnSet)
1132 *chunk = DEFAULT_CHUNK;
1133
1134 if (!subdev)
1135 return 1;
1136
1137 fd = open(subdev, O_RDONLY|O_EXCL, 0);
1138 if (fd < 0) {
1139 if (verbose)
1140 fprintf(stderr, Name ": super0.90 cannot open %s: %s\n",
1141 subdev, strerror(errno));
1142 return 0;
1143 }
1144
1145 if (!get_dev_size(fd, subdev, &ldsize)) {
1146 close(fd);
1147 return 0;
1148 }
1149 close(fd);
1150
1151 if (ldsize < MD_RESERVED_SECTORS * 512)
1152 return 0;
1153 if (size > (0x7fffffffULL<<9))
1154 return 0;
1155 *freesize = MD_NEW_SIZE_SECTORS(ldsize >> 9);
1156 return 1;
1157 }
1158 #endif /* MDASSEMBLE */
1159
1160 struct superswitch super0 = {
1161 #ifndef MDASSEMBLE
1162 .examine_super = examine_super0,
1163 .brief_examine_super = brief_examine_super0,
1164 .export_examine_super = export_examine_super0,
1165 .detail_super = detail_super0,
1166 .brief_detail_super = brief_detail_super0,
1167 .write_init_super = write_init_super0,
1168 .validate_geometry = validate_geometry0,
1169 .add_to_super = add_to_super0,
1170 #endif
1171 .match_home = match_home0,
1172 .uuid_from_super = uuid_from_super0,
1173 .getinfo_super = getinfo_super0,
1174 .container_content = container_content0,
1175 .update_super = update_super0,
1176 .init_super = init_super0,
1177 .store_super = store_super0,
1178 .compare_super = compare_super0,
1179 .load_super = load_super0,
1180 .match_metadata_desc = match_metadata_desc0,
1181 .avail_size = avail_size0,
1182 .add_internal_bitmap = add_internal_bitmap0,
1183 .locate_bitmap = locate_bitmap0,
1184 .write_bitmap = write_bitmap0,
1185 .free_super = free_super0,
1186 .name = "0.90",
1187 };