]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super0.c
Merge branch 'master' into from-stable
[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 static 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(struct supertype *st, char *homehost)
89 {
90 mdp_super_t *sb = st->sb;
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 : %llu\n",
183 ((unsigned long long)sb->events_hi << 32)
184 + sb->events_lo);
185 printf("\n");
186 if (sb->level == 5) {
187 c = map_num(r5layout, sb->layout);
188 printf(" Layout : %s\n", c?c:"-unknown-");
189 }
190 if (sb->level == 10) {
191 printf(" Layout : near=%d, %s=%d\n",
192 sb->layout&255,
193 (sb->layout&0x10000)?"offset":"far",
194 (sb->layout>>8)&255);
195 }
196 switch(sb->level) {
197 case 0:
198 case 4:
199 case 5:
200 case 6:
201 case 10:
202 printf(" Chunk Size : %dK\n", sb->chunk_size/1024);
203 break;
204 case -1:
205 printf(" Rounding : %dK\n", sb->chunk_size/1024);
206 break;
207 default: break;
208 }
209 printf("\n");
210 printf(" Number Major Minor RaidDevice State\n");
211 for (d= -1; d<(signed int)(sb->raid_disks+sb->spare_disks); d++) {
212 mdp_disk_t *dp;
213 char *dv;
214 char nb[5];
215 int wonly;
216 if (d>=0) dp = &sb->disks[d];
217 else dp = &sb->this_disk;
218 snprintf(nb, sizeof(nb), "%4d", d);
219 printf("%4s %5d %5d %5d %5d ", d < 0 ? "this" : nb,
220 dp->number, dp->major, dp->minor, dp->raid_disk);
221 wonly = dp->state & (1<<MD_DISK_WRITEMOSTLY);
222 dp->state &= ~(1<<MD_DISK_WRITEMOSTLY);
223 if (dp->state & (1<<MD_DISK_FAULTY)) printf(" faulty");
224 if (dp->state & (1<<MD_DISK_ACTIVE)) printf(" active");
225 if (dp->state & (1<<MD_DISK_SYNC)) printf(" sync");
226 if (dp->state & (1<<MD_DISK_REMOVED)) printf(" removed");
227 if (wonly) printf(" write-mostly");
228 if (dp->state == 0) printf(" spare");
229 if ((dv=map_dev(dp->major, dp->minor, 0)))
230 printf(" %s", dv);
231 printf("\n");
232 if (d == -1) printf("\n");
233 }
234 }
235
236 static void brief_examine_super0(struct supertype *st)
237 {
238 mdp_super_t *sb = st->sb;
239 char *c=map_num(pers, sb->level);
240 char devname[20];
241
242 sprintf(devname, "/dev/md%d", sb->md_minor);
243
244 printf("ARRAY %s level=%s num-devices=%d UUID=",
245 devname,
246 c?c:"-unknown-", sb->raid_disks);
247 if (sb->minor_version >= 90)
248 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
249 sb->set_uuid2, sb->set_uuid3);
250 else
251 printf("%08x", sb->set_uuid0);
252 printf("\n");
253 }
254
255 static void export_examine_super0(struct supertype *st)
256 {
257 mdp_super_t *sb = st->sb;
258
259 printf("MD_LEVEL=%s\n", map_num(pers, sb->level));
260 printf("MD_DEVICES=%d\n", sb->raid_disks);
261 if (sb->minor_version >= 90)
262 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
263 sb->set_uuid0, sb->set_uuid1,
264 sb->set_uuid2, sb->set_uuid3);
265 else
266 printf("MD_UUID=%08x\n", sb->set_uuid0);
267 printf("MD_UPDATE_TIME=%llu\n",
268 __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL);
269 printf("MD_EVENTS=%llu\n",
270 ((unsigned long long)sb->events_hi << 32)
271 + sb->events_lo);
272 }
273
274 static void detail_super0(struct supertype *st, char *homehost)
275 {
276 mdp_super_t *sb = st->sb;
277 printf(" UUID : ");
278 if (sb->minor_version >= 90)
279 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
280 sb->set_uuid2, sb->set_uuid3);
281 else
282 printf("%08x", sb->set_uuid0);
283 if (homehost) {
284 char buf[20];
285 void *hash = sha1_buffer(homehost,
286 strlen(homehost),
287 buf);
288 if (memcmp(&sb->set_uuid2, hash, 8)==0)
289 printf(" (local to host %s)", homehost);
290 }
291 printf("\n Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
292 }
293
294 static void brief_detail_super0(struct supertype *st)
295 {
296 mdp_super_t *sb = st->sb;
297 printf(" UUID=");
298 if (sb->minor_version >= 90)
299 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
300 sb->set_uuid2, sb->set_uuid3);
301 else
302 printf("%08x", sb->set_uuid0);
303 }
304
305 static void export_detail_super0(struct supertype *st)
306 {
307 mdp_super_t *sb = st->sb;
308 printf("MD_UUID=");
309 if (sb->minor_version >= 90)
310 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
311 sb->set_uuid2, sb->set_uuid3);
312 else
313 printf("%08x", sb->set_uuid0);
314 printf("\n");
315 }
316 #endif
317
318 static int match_home0(struct supertype *st, char *homehost)
319 {
320 mdp_super_t *sb = st->sb;
321 char buf[20];
322 char *hash = sha1_buffer(homehost,
323 strlen(homehost),
324 buf);
325
326 return (memcmp(&sb->set_uuid2, hash, 8)==0);
327 }
328
329 static void uuid_from_super0(struct supertype *st, int uuid[4])
330 {
331 mdp_super_t *super = st->sb;
332 uuid[0] = super->set_uuid0;
333 if (super->minor_version >= 90) {
334 uuid[1] = super->set_uuid1;
335 uuid[2] = super->set_uuid2;
336 uuid[3] = super->set_uuid3;
337 } else {
338 uuid[1] = 0;
339 uuid[2] = 0;
340 uuid[3] = 0;
341 }
342 }
343
344 static void getinfo_super0(struct supertype *st, struct mdinfo *info)
345 {
346 mdp_super_t *sb = st->sb;
347 int working = 0;
348 int i;
349
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
374 uuid_from_super0(st, info->uuid);
375
376 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
377 info->reshape_active = 1;
378 info->reshape_progress = sb->reshape_position;
379 info->new_level = sb->new_level;
380 info->delta_disks = sb->delta_disks;
381 info->new_layout = sb->new_layout;
382 info->new_chunk = sb->new_chunk;
383 } else
384 info->reshape_active = 0;
385
386 sprintf(info->name, "%d", sb->md_minor);
387 /* work_disks is calculated rather than read directly */
388 for (i=0; i < MD_SB_DISKS; i++)
389 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
390 (sb->disks[i].raid_disk < info->array.raid_disks) &&
391 (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
392 !(sb->disks[i].state & (1<<MD_DISK_FAULTY)))
393 working ++;
394 info->array.working_disks = working;
395 }
396
397
398 static int update_super0(struct supertype *st, struct mdinfo *info,
399 char *update,
400 char *devname, int verbose,
401 int uuid_set, char *homehost)
402 {
403 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
404 * For others, the return value is ignored.
405 */
406 int rv = 0;
407 mdp_super_t *sb = st->sb;
408 if (strcmp(update, "sparc2.2")==0 ) {
409 /* 2.2 sparc put the events in the wrong place
410 * So we copy the tail of the superblock
411 * up 4 bytes before continuing
412 */
413 __u32 *sb32 = (__u32*)sb;
414 memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
415 sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
416 (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
417 if (verbose >= 0)
418 fprintf (stderr, Name ": adjusting superblock of %s for 2.2/sparc compatability.\n",
419 devname);
420 }
421 if (strcmp(update, "super-minor") ==0) {
422 sb->md_minor = info->array.md_minor;
423 if (verbose > 0)
424 fprintf(stderr, Name ": updating superblock of %s with minor number %d\n",
425 devname, info->array.md_minor);
426 }
427 if (strcmp(update, "summaries") == 0) {
428 int i;
429 /* set nr_disks, active_disks, working_disks,
430 * failed_disks, spare_disks based on disks[]
431 * array in superblock.
432 * Also make sure extra slots aren't 'failed'
433 */
434 sb->nr_disks = sb->active_disks =
435 sb->working_disks = sb->failed_disks =
436 sb->spare_disks = 0;
437 for (i=0; i < MD_SB_DISKS ; i++)
438 if (sb->disks[i].major ||
439 sb->disks[i].minor) {
440 int state = sb->disks[i].state;
441 if (state & (1<<MD_DISK_REMOVED))
442 continue;
443 sb->nr_disks++;
444 if (state & (1<<MD_DISK_ACTIVE))
445 sb->active_disks++;
446 if (state & (1<<MD_DISK_FAULTY))
447 sb->failed_disks++;
448 else
449 sb->working_disks++;
450 if (state == 0)
451 sb->spare_disks++;
452 } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
453 sb->disks[i].state = 0;
454 }
455 if (strcmp(update, "force-one")==0) {
456 /* Not enough devices for a working array, so
457 * bring this one up-to-date.
458 */
459 __u32 ehi = sb->events_hi, elo = sb->events_lo;
460 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
461 sb->events_lo = (info->events) & 0xFFFFFFFF;
462 if (sb->events_hi != ehi ||
463 sb->events_lo != elo)
464 rv = 1;
465 }
466 if (strcmp(update, "force-array")==0) {
467 /* degraded array and 'force' requested, so
468 * maybe need to mark it 'clean'
469 */
470 if ((sb->level == 5 || sb->level == 4 || sb->level == 6) &&
471 (sb->state & (1 << MD_SB_CLEAN)) == 0) {
472 /* need to force clean */
473 sb->state |= (1 << MD_SB_CLEAN);
474 rv = 1;
475 }
476 }
477 if (strcmp(update, "assemble")==0) {
478 int d = info->disk.number;
479 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
480 if ((sb->disks[d].state & ~(1<<MD_DISK_WRITEMOSTLY))
481 != info->disk.state) {
482 sb->disks[d].state = info->disk.state | wonly;
483 rv = 1;
484 }
485 }
486 if (strcmp(update, "linear-grow-new") == 0) {
487 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
488 sb->disks[info->disk.number].number = info->disk.number;
489 sb->disks[info->disk.number].major = info->disk.major;
490 sb->disks[info->disk.number].minor = info->disk.minor;
491 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
492 sb->disks[info->disk.number].state = info->disk.state;
493 sb->this_disk = sb->disks[info->disk.number];
494 }
495 if (strcmp(update, "linear-grow-update") == 0) {
496 sb->raid_disks = info->array.raid_disks;
497 sb->nr_disks = info->array.nr_disks;
498 sb->active_disks = info->array.active_disks;
499 sb->working_disks = info->array.working_disks;
500 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
501 sb->disks[info->disk.number].number = info->disk.number;
502 sb->disks[info->disk.number].major = info->disk.major;
503 sb->disks[info->disk.number].minor = info->disk.minor;
504 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
505 sb->disks[info->disk.number].state = info->disk.state;
506 }
507 if (strcmp(update, "resync") == 0) {
508 /* make sure resync happens */
509 sb->state &= ~(1<<MD_SB_CLEAN);
510 sb->recovery_cp = 0;
511 }
512 if (strcmp(update, "homehost") == 0 &&
513 homehost) {
514 uuid_set = 0;
515 update = "uuid";
516 info->uuid[0] = sb->set_uuid0;
517 info->uuid[1] = sb->set_uuid1;
518 }
519 if (strcmp(update, "uuid") == 0) {
520 if (!uuid_set && homehost) {
521 char buf[20];
522 char *hash = sha1_buffer(homehost,
523 strlen(homehost),
524 buf);
525 memcpy(info->uuid+2, hash, 8);
526 }
527 sb->set_uuid0 = info->uuid[0];
528 sb->set_uuid1 = info->uuid[1];
529 sb->set_uuid2 = info->uuid[2];
530 sb->set_uuid3 = info->uuid[3];
531 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
532 struct bitmap_super_s *bm;
533 bm = (struct bitmap_super_s*)(sb+1);
534 uuid_from_super0(st, (int*)bm->uuid);
535 }
536 }
537 if (strcmp(update, "_reshape_progress")==0)
538 sb->reshape_position = info->reshape_progress;
539
540 sb->sb_csum = calc_sb0_csum(sb);
541 return rv;
542 }
543
544 /*
545 * For verion-0 superblock, the homehost is 'stored' in the
546 * uuid. 8 bytes for a hash of the host leaving 8 bytes
547 * of random material.
548 * We use the first 8 bytes (64bits) of the sha1 of the
549 * host name
550 */
551
552
553 static int init_super0(struct supertype *st, mdu_array_info_t *info,
554 unsigned long long size, char *ignored_name, char *homehost,
555 int *uuid)
556 {
557 mdp_super_t *sb;
558 int spares;
559
560 posix_memalign((void**)&sb, 512, MD_SB_BYTES + sizeof(bitmap_super_t));
561 memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
562
563 st->sb = sb;
564 if (info == NULL) {
565 /* zeroing the superblock */
566 return 0;
567 }
568
569 spares = info->working_disks - info->active_disks;
570 if (info->raid_disks + spares > MD_SB_DISKS) {
571 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
572 info->raid_disks , spares, MD_SB_DISKS);
573 return 0;
574 }
575
576 sb->md_magic = MD_SB_MAGIC;
577 sb->major_version = 0;
578 sb->minor_version = 90;
579 sb->patch_version = 0;
580 sb->gvalid_words = 0; /* ignored */
581 sb->ctime = time(0);
582 sb->level = info->level;
583 if (size != info->size)
584 return 0;
585 sb->size = info->size;
586 sb->nr_disks = info->nr_disks;
587 sb->raid_disks = info->raid_disks;
588 sb->md_minor = info->md_minor;
589 sb->not_persistent = 0;
590 if (uuid) {
591 sb->set_uuid0 = uuid[0];
592 sb->set_uuid1 = uuid[1];
593 sb->set_uuid2 = uuid[2];
594 sb->set_uuid3 = uuid[3];
595 } else {
596 int rfd = open("/dev/urandom", O_RDONLY);
597 if (rfd < 0 || read(rfd, &sb->set_uuid0, 4) != 4)
598 sb->set_uuid0 = random();
599 if (rfd < 0 || read(rfd, &sb->set_uuid1, 12) != 12) {
600 sb->set_uuid1 = random();
601 sb->set_uuid2 = random();
602 sb->set_uuid3 = random();
603 }
604 if (rfd >= 0)
605 close(rfd);
606 }
607 if (homehost) {
608 char buf[20];
609 char *hash = sha1_buffer(homehost,
610 strlen(homehost),
611 buf);
612 memcpy(&sb->set_uuid2, hash, 8);
613 }
614
615 sb->utime = sb->ctime;
616 sb->state = info->state;
617 sb->active_disks = info->active_disks;
618 sb->working_disks = info->working_disks;
619 sb->failed_disks = info->failed_disks;
620 sb->spare_disks = info->spare_disks;
621 sb->events_hi = 0;
622 sb->events_lo = 1;
623
624 sb->layout = info->layout;
625 sb->chunk_size = info->chunk_size;
626
627 return 1;
628 }
629
630 struct devinfo {
631 int fd;
632 char *devname;
633 mdu_disk_info_t disk;
634 struct devinfo *next;
635 };
636 /* Add a device to the superblock being created */
637 static void add_to_super0(struct supertype *st, mdu_disk_info_t *dinfo,
638 int fd, char *devname)
639 {
640 mdp_super_t *sb = st->sb;
641 mdp_disk_t *dk = &sb->disks[dinfo->number];
642 struct devinfo *di, **dip;
643
644 dk->number = dinfo->number;
645 dk->major = dinfo->major;
646 dk->minor = dinfo->minor;
647 dk->raid_disk = dinfo->raid_disk;
648 dk->state = dinfo->state;
649
650 sb->this_disk = sb->disks[dinfo->number];
651 sb->sb_csum = calc_sb0_csum(sb);
652
653 dip = (struct devinfo **)&st->info;
654 while (*dip)
655 dip = &(*dip)->next;
656 di = malloc(sizeof(struct devinfo));
657 di->fd = fd;
658 di->devname = devname;
659 di->disk = *dinfo;
660 di->next = NULL;
661 *dip = di;
662 }
663
664 static int store_super0(struct supertype *st, int fd)
665 {
666 unsigned long long dsize;
667 unsigned long long offset;
668 mdp_super_t *super = st->sb;
669
670 if (!get_dev_size(fd, NULL, &dsize))
671 return 1;
672
673 if (dsize < MD_RESERVED_SECTORS*512)
674 return 2;
675
676 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
677
678 offset *= 512;
679
680 if (lseek64(fd, offset, 0)< 0LL)
681 return 3;
682
683 if (write(fd, super, sizeof(*super)) != sizeof(*super))
684 return 4;
685
686 if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
687 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
688 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
689 if (write(fd, bm, ROUND_UP(sizeof(*bm),512)) !=
690 ROUND_UP(sizeof(*bm),512))
691 return 5;
692 }
693
694 fsync(fd);
695 return 0;
696 }
697
698 #ifndef MDASSEMBLE
699 static int write_init_super0(struct supertype *st)
700 {
701 mdp_super_t *sb = st->sb;
702 int rv = 0;
703 struct devinfo *di;
704
705 for (di = st->info ; di && ! rv ; di = di->next) {
706
707 if (di->disk.state == 1)
708 continue;
709 if (di->fd == -1)
710 continue;
711 Kill(di->devname, 0, 1, 1);
712 Kill(di->devname, 0, 1, 1);
713
714 sb->disks[di->disk.number].state &= ~(1<<MD_DISK_FAULTY);
715
716 sb->this_disk = sb->disks[di->disk.number];
717 sb->sb_csum = calc_sb0_csum(sb);
718 rv = store_super0(st, di->fd);
719
720 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
721 rv = st->ss->write_bitmap(st, di->fd);
722
723 if (rv)
724 fprintf(stderr,
725 Name ": failed to write superblock to %s\n",
726 di->devname);
727 close(di->fd);
728 di->fd = -1;
729 }
730 return rv;
731 }
732 #endif
733
734 static int compare_super0(struct supertype *st, struct supertype *tst)
735 {
736 /*
737 * return:
738 * 0 same, or first was empty, and second was copied
739 * 1 second had wrong number
740 * 2 wrong uuid
741 * 3 wrong other info
742 */
743 mdp_super_t *first = st->sb;
744 mdp_super_t *second = tst->sb;
745 int uuid1[4], uuid2[4];
746
747 if (second->md_magic != MD_SB_MAGIC)
748 return 1;
749 if (!first) {
750 posix_memalign((void**)&first, 512,
751 MD_SB_BYTES + sizeof(struct bitmap_super_s));
752 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
753 st->sb = first;
754 return 0;
755 }
756
757 uuid_from_super0(st, uuid1);
758 uuid_from_super0(tst, uuid2);
759 if (!same_uuid(uuid1, uuid2, 0))
760 return 2;
761 if (first->major_version != second->major_version ||
762 first->minor_version != second->minor_version ||
763 first->patch_version != second->patch_version ||
764 first->gvalid_words != second->gvalid_words ||
765 first->ctime != second->ctime ||
766 first->level != second->level ||
767 first->size != second->size ||
768 first->raid_disks != second->raid_disks )
769 return 3;
770
771 return 0;
772 }
773
774
775 static void free_super0(struct supertype *st);
776
777 static int load_super0(struct supertype *st, int fd, char *devname)
778 {
779 /* try to read in the superblock
780 * Return:
781 * 0 on success
782 * 1 on cannot get superblock
783 * 2 on superblock meaningless
784 */
785 unsigned long long dsize;
786 unsigned long long offset;
787 mdp_super_t *super;
788 int uuid[4];
789 struct bitmap_super_s *bsb;
790
791 free_super0(st);
792
793 if (st->subarray[0])
794 return 1;
795
796 if (!get_dev_size(fd, devname, &dsize))
797 return 1;
798
799 if (dsize < MD_RESERVED_SECTORS*512) {
800 if (devname)
801 fprintf(stderr, Name
802 ": %s is too small for md: size is %llu sectors.\n",
803 devname, dsize);
804 return 1;
805 }
806
807 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
808
809 offset *= 512;
810
811 ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
812
813 if (lseek64(fd, offset, 0)< 0LL) {
814 if (devname)
815 fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
816 devname, strerror(errno));
817 return 1;
818 }
819
820 posix_memalign((void**)&super, 512, MD_SB_BYTES + sizeof(bitmap_super_t)+512);
821
822 if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
823 if (devname)
824 fprintf(stderr, Name ": Cannot read superblock on %s\n",
825 devname);
826 free(super);
827 return 1;
828 }
829
830 if (st->ss && st->minor_version == 9)
831 super0_swap_endian(super);
832
833 if (super->md_magic != MD_SB_MAGIC) {
834 if (devname)
835 fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
836 devname, MD_SB_MAGIC, super->md_magic);
837 free(super);
838 return 2;
839 }
840
841 if (super->major_version != 0) {
842 if (devname)
843 fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
844 devname, super->major_version);
845 free(super);
846 return 2;
847 }
848 st->sb = super;
849
850 if (st->ss == NULL) {
851 st->ss = &super0;
852 st->minor_version = super->minor_version;
853 st->max_devs = MD_SB_DISKS;
854 st->info = NULL;
855 }
856
857 /* Now check on the bitmap superblock */
858 if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
859 return 0;
860 /* Read the bitmap superblock and make sure it looks
861 * valid. If it doesn't clear the bit. An --assemble --force
862 * should get that written out.
863 */
864 if (read(fd, super+1, ROUND_UP(sizeof(struct bitmap_super_s),512))
865 != ROUND_UP(sizeof(struct bitmap_super_s),512))
866 goto no_bitmap;
867
868 uuid_from_super0(st, uuid);
869 bsb = (struct bitmap_super_s *)(super+1);
870 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
871 memcmp(bsb->uuid, uuid, 16) != 0)
872 goto no_bitmap;
873 return 0;
874
875 no_bitmap:
876 super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
877
878 return 0;
879 }
880
881 static struct supertype *match_metadata_desc0(char *arg)
882 {
883 struct supertype *st = malloc(sizeof(*st));
884 if (!st) return st;
885
886 memset(st, 0, sizeof(*st));
887 st->ss = &super0;
888 st->info = NULL;
889 st->minor_version = 90;
890 st->max_devs = MD_SB_DISKS;
891 st->sb = NULL;
892 if (strcmp(arg, "0") == 0 ||
893 strcmp(arg, "0.90") == 0 ||
894 strcmp(arg, "default") == 0 ||
895 strcmp(arg, "") == 0 /* no metadata */
896 )
897 return st;
898
899 st->minor_version = 91; /* reshape in progress */
900 if (strcmp(arg, "0.91") == 0) /* For dup_super support */
901 return st;
902
903 st->minor_version = 9; /* flag for 'byte-swapped' */
904 if (strcmp(arg, "0.swap")==0 ||
905 strcmp(arg, "0.9") == 0) /* For dup_super support */
906 return st;
907
908 free(st);
909 return NULL;
910 }
911
912 static __u64 avail_size0(struct supertype *st, __u64 devsize)
913 {
914 if (devsize < MD_RESERVED_SECTORS)
915 return 0ULL;
916 return MD_NEW_SIZE_SECTORS(devsize);
917 }
918
919 static int add_internal_bitmap0(struct supertype *st, int *chunkp,
920 int delay, int write_behind,
921 unsigned long long size, int may_change,
922 int major)
923 {
924 /*
925 * The bitmap comes immediately after the superblock and must be 60K in size
926 * at most. The default size is between 30K and 60K
927 *
928 * size is in sectors, chunk is in bytes !!!
929 */
930 unsigned long long bits;
931 unsigned long long max_bits = 60*1024*8;
932 unsigned long long min_chunk;
933 int chunk = *chunkp;
934 mdp_super_t *sb = st->sb;
935 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
936
937
938 min_chunk = 4096; /* sub-page chunks don't work yet.. */
939 bits = (size * 512) / min_chunk + 1;
940 while (bits > max_bits) {
941 min_chunk *= 2;
942 bits = (bits+1)/2;
943 }
944 if (chunk == UnSet)
945 chunk = min_chunk;
946 else if (chunk < min_chunk)
947 return 0; /* chunk size too small */
948
949 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
950
951 memset(bms, 0, sizeof(*bms));
952 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
953 bms->version = __cpu_to_le32(major);
954 uuid_from_super0(st, (int*)bms->uuid);
955 bms->chunksize = __cpu_to_le32(chunk);
956 bms->daemon_sleep = __cpu_to_le32(delay);
957 bms->sync_size = __cpu_to_le64(size);
958 bms->write_behind = __cpu_to_le32(write_behind);
959 *chunkp = chunk;
960 return 1;
961 }
962
963
964 static void locate_bitmap0(struct supertype *st, int fd)
965 {
966 unsigned long long dsize;
967 unsigned long long offset;
968
969 if (!get_dev_size(fd, NULL, &dsize))
970 return;
971
972 if (dsize < MD_RESERVED_SECTORS*512)
973 return;
974
975 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
976
977 offset *= 512;
978
979 offset += MD_SB_BYTES;
980
981 lseek64(fd, offset, 0);
982 }
983
984 static int write_bitmap0(struct supertype *st, int fd)
985 {
986 unsigned long long dsize;
987 unsigned long long offset;
988 mdp_super_t *sb = st->sb;
989
990 int rv = 0;
991
992 int towrite, n;
993 char abuf[4096+512];
994 char *buf = (char*)(((long)(abuf+512))&~511UL);
995
996 if (!get_dev_size(fd, NULL, &dsize))
997 return 1;
998
999
1000 if (dsize < MD_RESERVED_SECTORS*512)
1001 return -1;
1002
1003 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1004
1005 offset *= 512;
1006
1007 if (lseek64(fd, offset + 4096, 0)< 0LL)
1008 return 3;
1009
1010 memset(buf, 0xff, 4096);
1011 memcpy(buf, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
1012 towrite = 64*1024;
1013 while (towrite > 0) {
1014 n = towrite;
1015 if (n > 4096)
1016 n = 4096;
1017 n = write(fd, buf, n);
1018 if (n > 0)
1019 towrite -= n;
1020 else
1021 break;
1022 memset(buf, 0xff, 4096);
1023 }
1024 fsync(fd);
1025 if (towrite)
1026 rv = -2;
1027
1028 return rv;
1029 }
1030
1031 static void free_super0(struct supertype *st)
1032 {
1033 if (st->sb)
1034 free(st->sb);
1035 st->sb = NULL;
1036 }
1037
1038 static int validate_geometry0(struct supertype *st, int level,
1039 int layout, int raiddisks,
1040 int chunk, unsigned long long size,
1041 char *subdev, unsigned long long *freesize,
1042 int verbose)
1043 {
1044 unsigned long long ldsize;
1045 int fd;
1046
1047 if (level == LEVEL_CONTAINER)
1048 return 0;
1049 if (raiddisks > MD_SB_DISKS)
1050 return 0;
1051 if (size > (0x7fffffffULL<<10))
1052 return 0;
1053 if (!subdev)
1054 return 1;
1055
1056 fd = open(subdev, O_RDONLY|O_EXCL, 0);
1057 if (fd < 0) {
1058 if (verbose)
1059 fprintf(stderr, Name ": super0.90 cannot open %s: %s\n",
1060 subdev, strerror(errno));
1061 return 0;
1062 }
1063
1064 if (!get_dev_size(fd, subdev, &ldsize)) {
1065 close(fd);
1066 return 0;
1067 }
1068 close(fd);
1069
1070 if (ldsize < MD_RESERVED_SECTORS * 512)
1071 return 0;
1072 if (size > (0x7fffffffULL<<10))
1073 return 0;
1074 *freesize = MD_NEW_SIZE_SECTORS(ldsize >> 9);
1075 return 1;
1076 }
1077
1078 struct superswitch super0 = {
1079 #ifndef MDASSEMBLE
1080 .examine_super = examine_super0,
1081 .brief_examine_super = brief_examine_super0,
1082 .export_examine_super = export_examine_super0,
1083 .detail_super = detail_super0,
1084 .brief_detail_super = brief_detail_super0,
1085 .export_detail_super = export_detail_super0,
1086 .write_init_super = write_init_super0,
1087 #endif
1088 .match_home = match_home0,
1089 .uuid_from_super = uuid_from_super0,
1090 .getinfo_super = getinfo_super0,
1091 .update_super = update_super0,
1092 .init_super = init_super0,
1093 .add_to_super = add_to_super0,
1094 .store_super = store_super0,
1095 .compare_super = compare_super0,
1096 .load_super = load_super0,
1097 .match_metadata_desc = match_metadata_desc0,
1098 .avail_size = avail_size0,
1099 .add_internal_bitmap = add_internal_bitmap0,
1100 .locate_bitmap = locate_bitmap0,
1101 .write_bitmap = write_bitmap0,
1102 .free_super = free_super0,
1103 .validate_geometry = validate_geometry0,
1104 };