]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super0.c
Document PPL in man md
[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 static unsigned long calc_sb0_csum(mdp_super_t *super)
40 {
41 unsigned long csum = super->sb_csum;
42 unsigned long newcsum;
43 super->sb_csum= 0 ;
44 newcsum = calc_csum(super, MD_SB_BYTES);
45 super->sb_csum = csum;
46 return newcsum;
47 }
48
49 static void super0_swap_endian(struct mdp_superblock_s *sb)
50 {
51 /* as super0 superblocks are host-endian, it is sometimes
52 * useful to be able to swap the endianness
53 * as (almost) everything is u32's we byte-swap every 4byte
54 * number.
55 * We then also have to swap the events_hi and events_lo
56 */
57 char *sbc = (char *)sb;
58 __u32 t32;
59 int i;
60
61 for (i=0; i < MD_SB_BYTES ; i+=4) {
62 char t = sbc[i];
63 sbc[i] = sbc[i+3];
64 sbc[i+3] = t;
65 t=sbc[i+1];
66 sbc[i+1]=sbc[i+2];
67 sbc[i+2]=t;
68 }
69 t32 = sb->events_hi;
70 sb->events_hi = sb->events_lo;
71 sb->events_lo = t32;
72
73 t32 = sb->cp_events_hi;
74 sb->cp_events_hi = sb->cp_events_lo;
75 sb->cp_events_lo = t32;
76
77 }
78
79 static void examine_super0(struct supertype *st, char *homehost)
80 {
81 mdp_super_t *sb = st->sb;
82 time_t atime;
83 int d;
84 int delta_extra = 0;
85 char *c;
86
87 printf(" Magic : %08x\n", sb->md_magic);
88 printf(" Version : %d.%02d.%02d\n",
89 sb->major_version, sb->minor_version, sb->patch_version);
90 if (sb->minor_version >= 90) {
91 printf(" UUID : %08x:%08x:%08x:%08x", sb->set_uuid0,
92 sb->set_uuid1, sb->set_uuid2, sb->set_uuid3);
93 if (homehost) {
94 char buf[20];
95 void *hash;
96
97 hash = sha1_buffer(homehost, strlen(homehost), buf);
98 if (memcmp(&sb->set_uuid2, hash, 8) == 0)
99 printf(" (local to host %s)", homehost);
100 }
101 printf("\n");
102 } else
103 printf(" UUID : %08x\n", sb->set_uuid0);
104
105 if (sb->not_persistent)
106 printf(" Eedk : not persistent\n");
107
108 atime = sb->ctime;
109 printf(" Creation Time : %.24s\n", ctime(&atime));
110 c = map_num(pers, sb->level);
111 printf(" Raid Level : %s\n", c?c:"-unknown-");
112 if ((int)sb->level > 0) {
113 int ddsks = 0, ddsks_denom = 1;
114 printf(" Used Dev Size : %d%s\n", sb->size,
115 human_size((long long)sb->size<<10));
116 switch(sb->level) {
117 case 1:
118 ddsks=1;
119 break;
120 case 4:
121 case 5:
122 ddsks = sb->raid_disks - 1;
123 break;
124 case 6:
125 ddsks = sb->raid_disks - 2;
126 break;
127 case 10:
128 ddsks = sb->raid_disks;
129 ddsks_denom =
130 (sb->layout & 255) * ((sb->layout >> 8) & 255);
131 }
132 if (ddsks) {
133 long long asize = sb->size;
134 asize = (asize << 10) * ddsks / ddsks_denom;
135 printf(" Array Size : %llu%s\n",
136 asize >> 10, human_size(asize));
137 }
138 }
139 printf(" Raid Devices : %d\n", sb->raid_disks);
140 printf(" Total Devices : %d\n", sb->nr_disks);
141 printf("Preferred Minor : %d\n", sb->md_minor);
142 printf("\n");
143 if (sb->minor_version > 90 && (sb->reshape_position + 1) != 0) {
144 printf(" Reshape pos'n : %llu%s\n",
145 (unsigned long long)sb->reshape_position / 2,
146 human_size((long long)sb->reshape_position << 9));
147 if (sb->delta_disks) {
148 printf(" Delta Devices : %d", sb->delta_disks);
149 printf(" (%d->%d)\n", sb->raid_disks-sb->delta_disks,
150 sb->raid_disks);
151 if (((int)sb->delta_disks) < 0)
152 delta_extra = - sb->delta_disks;
153 }
154 if (sb->new_level != sb->level) {
155 c = map_num(pers, sb->new_level);
156 printf(" New Level : %s\n", c?c:"-unknown-");
157 }
158 if (sb->new_layout != sb->layout) {
159 if (sb->level == 5) {
160 c = map_num(r5layout, sb->new_layout);
161 printf(" New Layout : %s\n",
162 c?c:"-unknown-");
163 }
164 if (sb->level == 6) {
165 c = map_num(r6layout, sb->new_layout);
166 printf(" New Layout : %s\n",
167 c?c:"-unknown-");
168 }
169 if (sb->level == 10) {
170 printf(" New Layout : near=%d, %s=%d\n",
171 sb->new_layout&255,
172 (sb->new_layout&0x10000)?"offset":"far",
173 (sb->new_layout>>8)&255);
174 }
175 }
176 if (sb->new_chunk != sb->chunk_size)
177 printf(" New Chunksize : %d\n", sb->new_chunk);
178 printf("\n");
179 }
180 atime = sb->utime;
181 printf(" Update Time : %.24s\n", ctime(&atime));
182 printf(" State : %s\n",
183 (sb->state&(1 << MD_SB_CLEAN)) ? "clean":"active");
184 if (sb->state & (1 << MD_SB_BITMAP_PRESENT))
185 printf("Internal Bitmap : present\n");
186 printf(" Active Devices : %d\n", sb->active_disks);
187 printf("Working Devices : %d\n", sb->working_disks);
188 printf(" Failed Devices : %d\n", sb->failed_disks);
189 printf(" Spare Devices : %d\n", sb->spare_disks);
190 if (calc_sb0_csum(sb) == sb->sb_csum)
191 printf(" Checksum : %x - correct\n", sb->sb_csum);
192 else
193 printf(" Checksum : %x - expected %lx\n",
194 sb->sb_csum, calc_sb0_csum(sb));
195 printf(" Events : %llu\n",
196 ((unsigned long long)sb->events_hi << 32) + sb->events_lo);
197 printf("\n");
198 if (sb->level == 5) {
199 c = map_num(r5layout, sb->layout);
200 printf(" Layout : %s\n", c?c:"-unknown-");
201 }
202 if (sb->level == 6) {
203 c = map_num(r6layout, sb->layout);
204 printf(" Layout : %s\n", c?c:"-unknown-");
205 }
206 if (sb->level == 10) {
207 printf(" Layout :");
208 print_r10_layout(sb->layout);
209 printf("\n");
210 }
211 switch(sb->level) {
212 case 0:
213 case 4:
214 case 5:
215 case 6:
216 case 10:
217 printf(" Chunk Size : %dK\n", sb->chunk_size / 1024);
218 break;
219 case -1:
220 printf(" Rounding : %dK\n", sb->chunk_size / 1024);
221 break;
222 default:
223 break;
224 }
225 printf("\n");
226 printf(" Number Major Minor RaidDevice State\n");
227 for (d = -1;
228 d < (signed int)(sb->raid_disks + delta_extra + sb->spare_disks);
229 d++) {
230 mdp_disk_t *dp;
231 char *dv;
232 char nb[11];
233 int wonly, failfast;
234 if (d>=0) dp = &sb->disks[d];
235 else dp = &sb->this_disk;
236 snprintf(nb, sizeof(nb), "%4d", d);
237 printf("%4s %5d %5d %5d %5d ", d < 0 ? "this" : nb,
238 dp->number, dp->major, dp->minor, dp->raid_disk);
239 wonly = dp->state & (1 << MD_DISK_WRITEMOSTLY);
240 failfast = dp->state & (1<<MD_DISK_FAILFAST);
241 dp->state &= ~(wonly | failfast);
242 if (dp->state & (1 << MD_DISK_FAULTY))
243 printf(" faulty");
244 if (dp->state & (1 << MD_DISK_ACTIVE))
245 printf(" active");
246 if (dp->state & (1 << MD_DISK_SYNC))
247 printf(" sync");
248 if (dp->state & (1 << MD_DISK_REMOVED))
249 printf(" removed");
250 if (wonly)
251 printf(" write-mostly");
252 if (failfast)
253 printf(" failfast");
254 if (dp->state == 0)
255 printf(" spare");
256 if ((dv = map_dev(dp->major, dp->minor, 0)))
257 printf(" %s", dv);
258 printf("\n");
259 if (d == -1)
260 printf("\n");
261 }
262 }
263
264 static void brief_examine_super0(struct supertype *st, int verbose)
265 {
266 mdp_super_t *sb = st->sb;
267 char *c=map_num(pers, sb->level);
268 char devname[20];
269
270 sprintf(devname, "/dev/md%d", sb->md_minor);
271
272 if (verbose) {
273 printf("ARRAY %s level=%s num-devices=%d",
274 devname,
275 c?c:"-unknown-", sb->raid_disks);
276 } else
277 printf("ARRAY %s", devname);
278
279 if (sb->minor_version >= 90)
280 printf(" UUID=%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
281 sb->set_uuid2, sb->set_uuid3);
282 else
283 printf(" UUID=%08x", sb->set_uuid0);
284 printf("\n");
285 }
286
287 static void export_examine_super0(struct supertype *st)
288 {
289 mdp_super_t *sb = st->sb;
290
291 printf("MD_LEVEL=%s\n", map_num(pers, sb->level));
292 printf("MD_DEVICES=%d\n", sb->raid_disks);
293 if (sb->minor_version >= 90)
294 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
295 sb->set_uuid0, sb->set_uuid1,
296 sb->set_uuid2, sb->set_uuid3);
297 else
298 printf("MD_UUID=%08x\n", sb->set_uuid0);
299 printf("MD_UPDATE_TIME=%llu\n",
300 __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL);
301 printf("MD_EVENTS=%llu\n",
302 ((unsigned long long)sb->events_hi << 32)
303 + sb->events_lo);
304 }
305
306 static int copy_metadata0(struct supertype *st, int from, int to)
307 {
308 /* Read 64K from the appropriate offset of 'from'
309 * and if it looks a little like a 0.90 superblock,
310 * write it to the same offset of 'to'
311 */
312 void *buf;
313 unsigned long long dsize, offset;
314 const int bufsize = 64*1024;
315 mdp_super_t *super;
316
317 if (posix_memalign(&buf, 4096, bufsize) != 0)
318 return 1;
319
320 if (!get_dev_size(from, NULL, &dsize))
321 goto err;
322
323 if (dsize < MD_RESERVED_SECTORS*512)
324 goto err;
325
326 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
327
328 offset *= 512;
329
330 if (lseek64(from, offset, 0) < 0LL)
331 goto err;
332 if (read(from, buf, bufsize) != bufsize)
333 goto err;
334
335 if (lseek64(to, offset, 0) < 0LL)
336 goto err;
337 super = buf;
338 if (super->md_magic != MD_SB_MAGIC ||
339 super->major_version != 0 ||
340 calc_sb0_csum(super) != super->sb_csum)
341 goto err;
342 if (write(to, buf, bufsize) != bufsize)
343 goto err;
344 free(buf);
345 return 0;
346 err:
347 free(buf);
348 return 1;
349 }
350
351 static void detail_super0(struct supertype *st, char *homehost, char *subarray)
352 {
353 mdp_super_t *sb = st->sb;
354 printf(" UUID : ");
355 if (sb->minor_version >= 90)
356 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
357 sb->set_uuid2, sb->set_uuid3);
358 else
359 printf("%08x", sb->set_uuid0);
360 if (homehost) {
361 char buf[20];
362 void *hash = sha1_buffer(homehost,
363 strlen(homehost),
364 buf);
365 if (memcmp(&sb->set_uuid2, hash, 8)==0)
366 printf(" (local to host %s)", homehost);
367 }
368 printf("\n Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
369 }
370
371 static void brief_detail_super0(struct supertype *st, char *subarray)
372 {
373 mdp_super_t *sb = st->sb;
374 printf(" UUID=");
375 if (sb->minor_version >= 90)
376 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
377 sb->set_uuid2, sb->set_uuid3);
378 else
379 printf("%08x", sb->set_uuid0);
380 }
381
382 static int match_home0(struct supertype *st, char *homehost)
383 {
384 mdp_super_t *sb = st->sb;
385 char buf[20];
386 char *hash;
387
388 if (!homehost)
389 return 0;
390 hash = sha1_buffer(homehost,
391 strlen(homehost),
392 buf);
393
394 return (memcmp(&sb->set_uuid2, hash, 8)==0);
395 }
396
397 static void uuid_from_super0(struct supertype *st, int uuid[4])
398 {
399 mdp_super_t *super = st->sb;
400 uuid[0] = super->set_uuid0;
401 if (super->minor_version >= 90) {
402 uuid[1] = super->set_uuid1;
403 uuid[2] = super->set_uuid2;
404 uuid[3] = super->set_uuid3;
405 } else {
406 uuid[1] = 0;
407 uuid[2] = 0;
408 uuid[3] = 0;
409 }
410 }
411
412 static void getinfo_super0(struct supertype *st, struct mdinfo *info, char *map)
413 {
414 mdp_super_t *sb = st->sb;
415 int working = 0;
416 int i;
417 int map_disks = info->array.raid_disks;
418
419 memset(info, 0, sizeof(*info));
420 info->array.major_version = sb->major_version;
421 info->array.minor_version = sb->minor_version;
422 info->array.patch_version = sb->patch_version;
423 info->array.raid_disks = sb->raid_disks;
424 info->array.level = sb->level;
425 info->array.layout = sb->layout;
426 info->array.md_minor = sb->md_minor;
427 info->array.ctime = sb->ctime;
428 info->array.utime = sb->utime;
429 info->array.chunk_size = sb->chunk_size;
430 info->array.state = sb->state;
431 info->component_size = sb->size;
432 info->component_size *= 2;
433
434 if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
435 info->bitmap_offset = 8;
436
437 info->disk.state = sb->this_disk.state;
438 info->disk.major = sb->this_disk.major;
439 info->disk.minor = sb->this_disk.minor;
440 info->disk.raid_disk = sb->this_disk.raid_disk;
441 info->disk.number = sb->this_disk.number;
442
443 info->events = md_event(sb);
444 info->data_offset = 0;
445
446 sprintf(info->text_version, "0.%d", sb->minor_version);
447 info->safe_mode_delay = 200;
448
449 uuid_from_super0(st, info->uuid);
450
451 info->recovery_start = MaxSector;
452 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
453 info->reshape_active = 1;
454 info->reshape_progress = sb->reshape_position;
455 info->new_level = sb->new_level;
456 info->delta_disks = sb->delta_disks;
457 info->new_layout = sb->new_layout;
458 info->new_chunk = sb->new_chunk;
459 if (info->delta_disks < 0)
460 info->array.raid_disks -= info->delta_disks;
461 } else
462 info->reshape_active = 0;
463
464 info->recovery_blocked = info->reshape_active;
465
466 sprintf(info->name, "%d", sb->md_minor);
467 /* work_disks is calculated rather than read directly */
468 for (i=0; i < MD_SB_DISKS; i++)
469 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
470 (sb->disks[i].raid_disk < (unsigned)info->array.raid_disks) &&
471 (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
472 !(sb->disks[i].state & (1<<MD_DISK_FAULTY))) {
473 working ++;
474 if (map && i < map_disks)
475 map[i] = 1;
476 } else if (map && i < map_disks)
477 map[i] = 0;
478 info->array.working_disks = working;
479 }
480
481 static struct mdinfo *container_content0(struct supertype *st, char *subarray)
482 {
483 struct mdinfo *info;
484
485 if (subarray)
486 return NULL;
487
488 info = xmalloc(sizeof(*info));
489 getinfo_super0(st, info, NULL);
490 return info;
491 }
492
493 static int update_super0(struct supertype *st, struct mdinfo *info,
494 char *update,
495 char *devname, int verbose,
496 int uuid_set, char *homehost)
497 {
498 /* NOTE: for 'assemble' and 'force' we need to return non-zero
499 * if any change was made. For others, the return value is
500 * ignored.
501 */
502 int rv = 0;
503 int uuid[4];
504 mdp_super_t *sb = st->sb;
505
506 if (strcmp(update, "homehost") == 0 &&
507 homehost) {
508 /* note that 'homehost' is special as it is really
509 * a "uuid" update.
510 */
511 uuid_set = 0;
512 update = "uuid";
513 info->uuid[0] = sb->set_uuid0;
514 info->uuid[1] = sb->set_uuid1;
515 }
516
517 if (strcmp(update, "sparc2.2")==0 ) {
518 /* 2.2 sparc put the events in the wrong place
519 * So we copy the tail of the superblock
520 * up 4 bytes before continuing
521 */
522 __u32 *sb32 = (__u32*)sb;
523
524 memmove(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
525 sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
526 (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
527 if (verbose >= 0)
528 pr_err("adjusting superblock of %s for 2.2/sparc compatibility.\n",
529 devname);
530 } else if (strcmp(update, "super-minor") ==0) {
531 sb->md_minor = info->array.md_minor;
532 if (verbose > 0)
533 pr_err("updating superblock of %s with minor number %d\n",
534 devname, info->array.md_minor);
535 } else if (strcmp(update, "summaries") == 0) {
536 unsigned int i;
537 /* set nr_disks, active_disks, working_disks,
538 * failed_disks, spare_disks based on disks[]
539 * array in superblock.
540 * Also make sure extra slots aren't 'failed'
541 */
542 sb->nr_disks = sb->active_disks =
543 sb->working_disks = sb->failed_disks =
544 sb->spare_disks = 0;
545 for (i=0; i < MD_SB_DISKS ; i++)
546 if (sb->disks[i].major ||
547 sb->disks[i].minor) {
548 int state = sb->disks[i].state;
549 if (state & (1<<MD_DISK_REMOVED))
550 continue;
551 sb->nr_disks++;
552 if (state & (1<<MD_DISK_ACTIVE))
553 sb->active_disks++;
554 if (state & (1<<MD_DISK_FAULTY))
555 sb->failed_disks++;
556 else
557 sb->working_disks++;
558 if (state == 0)
559 sb->spare_disks++;
560 } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
561 sb->disks[i].state = 0;
562 } else if (strcmp(update, "force-one")==0) {
563 /* Not enough devices for a working array, so
564 * bring this one up-to-date.
565 */
566 __u32 ehi = sb->events_hi, elo = sb->events_lo;
567 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
568 sb->events_lo = (info->events) & 0xFFFFFFFF;
569 if (sb->events_hi != ehi ||
570 sb->events_lo != elo)
571 rv = 1;
572 } else if (strcmp(update, "force-array")==0) {
573 /* degraded array and 'force' requested, so
574 * maybe need to mark it 'clean'
575 */
576 if ((sb->level == 5 || sb->level == 4 || sb->level == 6) &&
577 (sb->state & (1 << MD_SB_CLEAN)) == 0) {
578 /* need to force clean */
579 sb->state |= (1 << MD_SB_CLEAN);
580 rv = 1;
581 }
582 } else if (strcmp(update, "assemble")==0) {
583 int d = info->disk.number;
584 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
585 int failfast = sb->disks[d].state & (1<<MD_DISK_FAILFAST);
586 int mask = (1<<MD_DISK_WRITEMOSTLY)|(1<<MD_DISK_FAILFAST);
587 int add = 0;
588 if (sb->minor_version >= 91)
589 /* During reshape we don't insist on everything
590 * being marked 'sync'
591 */
592 add = (1<<MD_DISK_SYNC);
593 if (((sb->disks[d].state & ~mask) | add) !=
594 (unsigned)info->disk.state) {
595 sb->disks[d].state = info->disk.state | wonly |failfast;
596 rv = 1;
597 }
598 if (info->reshape_active &&
599 sb->minor_version > 90 && (sb->reshape_position+1) != 0 &&
600 info->delta_disks >= 0 &&
601 info->reshape_progress < sb->reshape_position) {
602 sb->reshape_position = info->reshape_progress;
603 rv = 1;
604 }
605 if (info->reshape_active &&
606 sb->minor_version > 90 && (sb->reshape_position+1) != 0 &&
607 info->delta_disks < 0 &&
608 info->reshape_progress > sb->reshape_position) {
609 sb->reshape_position = info->reshape_progress;
610 rv = 1;
611 }
612 } else if (strcmp(update, "linear-grow-new") == 0) {
613 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
614 sb->disks[info->disk.number].number = info->disk.number;
615 sb->disks[info->disk.number].major = info->disk.major;
616 sb->disks[info->disk.number].minor = info->disk.minor;
617 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
618 sb->disks[info->disk.number].state = info->disk.state;
619 sb->this_disk = sb->disks[info->disk.number];
620 } else if (strcmp(update, "linear-grow-update") == 0) {
621 sb->raid_disks = info->array.raid_disks;
622 sb->nr_disks = info->array.nr_disks;
623 sb->active_disks = info->array.active_disks;
624 sb->working_disks = info->array.working_disks;
625 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
626 sb->disks[info->disk.number].number = info->disk.number;
627 sb->disks[info->disk.number].major = info->disk.major;
628 sb->disks[info->disk.number].minor = info->disk.minor;
629 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
630 sb->disks[info->disk.number].state = info->disk.state;
631 } else if (strcmp(update, "resync") == 0) {
632 /* make sure resync happens */
633 sb->state &= ~(1<<MD_SB_CLEAN);
634 sb->recovery_cp = 0;
635 } else if (strcmp(update, "uuid") == 0) {
636 if (!uuid_set && homehost) {
637 char buf[20];
638 char *hash = sha1_buffer(homehost,
639 strlen(homehost),
640 buf);
641 memcpy(info->uuid+2, hash, 8);
642 }
643 sb->set_uuid0 = info->uuid[0];
644 sb->set_uuid1 = info->uuid[1];
645 sb->set_uuid2 = info->uuid[2];
646 sb->set_uuid3 = info->uuid[3];
647 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
648 struct bitmap_super_s *bm;
649 bm = (struct bitmap_super_s*)(sb+1);
650 uuid_from_super0(st, uuid);
651 memcpy(bm->uuid, uuid, 16);
652 }
653 } else if (strcmp(update, "metadata") == 0) {
654 /* Create some v1.0 metadata to match ours but make the
655 * ctime bigger. Also update info->array.*_version.
656 * We need to arrange that store_super writes out
657 * the v1.0 metadata.
658 * Not permitted for unclean array, or array with
659 * bitmap.
660 */
661 if (info->bitmap_offset) {
662 pr_err("Cannot update metadata when bitmap is present\n");
663 rv = -2;
664 } else if (info->array.state != 1) {
665 pr_err("Cannot update metadata on unclean array\n");
666 rv = -2;
667 } else {
668 info->array.major_version = 1;
669 info->array.minor_version = 0;
670 uuid_from_super0(st, info->uuid);
671 st->other = super1_make_v0(st, info, st->sb);
672 }
673 } else if (strcmp(update, "revert-reshape") == 0) {
674 rv = -2;
675 if (sb->minor_version <= 90)
676 pr_err("No active reshape to revert on %s\n",
677 devname);
678 else if (sb->delta_disks == 0)
679 pr_err("%s: Can only revert reshape which changes number of devices\n",
680 devname);
681 else {
682 int tmp;
683 int parity = sb->level == 6 ? 2 : 1;
684 rv = 0;
685
686 if (sb->level >= 4 && sb->level <= 6 &&
687 sb->reshape_position % (
688 sb->new_chunk/512 *
689 (sb->raid_disks - sb->delta_disks - parity))) {
690 pr_err("Reshape position is not suitably aligned.\n");
691 pr_err("Try normal assembly and stop again\n");
692 return -2;
693 }
694 sb->raid_disks -= sb->delta_disks;
695 sb->delta_disks = -sb->delta_disks;
696
697 tmp = sb->new_layout;
698 sb->new_layout = sb->layout;
699 sb->layout = tmp;
700
701 tmp = sb->new_chunk;
702 sb->new_chunk = sb->chunk_size;
703 sb->chunk_size = tmp;
704 }
705 } else if (strcmp(update, "no-bitmap") == 0) {
706 sb->state &= ~(1<<MD_SB_BITMAP_PRESENT);
707 } else if (strcmp(update, "_reshape_progress")==0)
708 sb->reshape_position = info->reshape_progress;
709 else if (strcmp(update, "writemostly")==0)
710 sb->state |= (1<<MD_DISK_WRITEMOSTLY);
711 else if (strcmp(update, "readwrite")==0)
712 sb->state &= ~(1<<MD_DISK_WRITEMOSTLY);
713 else
714 rv = -1;
715
716 sb->sb_csum = calc_sb0_csum(sb);
717 return rv;
718 }
719
720 /*
721 * For version-0 superblock, the homehost is 'stored' in the uuid.
722 * 8 bytes for a hash of the host leaving 8 bytes of random material.
723 * We use the first 8 bytes (64bits) of the sha1 of the host name
724 */
725 static int init_super0(struct supertype *st, mdu_array_info_t *info,
726 struct shape *s, char *ignored_name,
727 char *homehost, int *uuid,
728 unsigned long long data_offset)
729 {
730 mdp_super_t *sb;
731 int spares;
732
733 if (data_offset != INVALID_SECTORS) {
734 pr_err("data-offset not support for 0.90\n");
735 return 0;
736 }
737
738 if (posix_memalign((void**)&sb, 4096,
739 MD_SB_BYTES + ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
740 pr_err("could not allocate superblock\n");
741 return 0;
742 }
743 memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
744
745 st->sb = sb;
746 if (info == NULL) {
747 /* zeroing the superblock */
748 return 0;
749 }
750
751 spares = info->working_disks - info->active_disks;
752 if (info->raid_disks + spares > MD_SB_DISKS) {
753 pr_err("too many devices requested: %d+%d > %d\n",
754 info->raid_disks , spares, MD_SB_DISKS);
755 return 0;
756 }
757
758 sb->md_magic = MD_SB_MAGIC;
759 sb->major_version = 0;
760 sb->minor_version = 90;
761 sb->patch_version = 0;
762 sb->gvalid_words = 0; /* ignored */
763 sb->ctime = time(0);
764 sb->level = info->level;
765 sb->size = s->size;
766 if (s->size != (unsigned long long)sb->size)
767 return 0;
768 sb->nr_disks = info->nr_disks;
769 sb->raid_disks = info->raid_disks;
770 sb->md_minor = info->md_minor;
771 sb->not_persistent = 0;
772 if (uuid) {
773 sb->set_uuid0 = uuid[0];
774 sb->set_uuid1 = uuid[1];
775 sb->set_uuid2 = uuid[2];
776 sb->set_uuid3 = uuid[3];
777 } else {
778 __u32 r[4];
779 random_uuid((__u8 *)r);
780 sb->set_uuid0 = r[0];
781 sb->set_uuid1 = r[1];
782 sb->set_uuid2 = r[2];
783 sb->set_uuid3 = r[3];
784 }
785 if (homehost && !uuid) {
786 char buf[20];
787 char *hash = sha1_buffer(homehost,
788 strlen(homehost),
789 buf);
790 memcpy(&sb->set_uuid2, hash, 8);
791 }
792
793 sb->utime = sb->ctime;
794 sb->state = info->state;
795 sb->active_disks = info->active_disks;
796 sb->working_disks = info->working_disks;
797 sb->failed_disks = info->failed_disks;
798 sb->spare_disks = info->spare_disks;
799 sb->events_hi = 0;
800 sb->events_lo = 1;
801
802 sb->layout = info->layout;
803 sb->chunk_size = info->chunk_size;
804
805 return 1;
806 }
807
808 struct devinfo {
809 int fd;
810 char *devname;
811 mdu_disk_info_t disk;
812 struct devinfo *next;
813 };
814
815 /* Add a device to the superblock being created */
816 static int add_to_super0(struct supertype *st, mdu_disk_info_t *dinfo,
817 int fd, char *devname, unsigned long long data_offset)
818 {
819 mdp_super_t *sb = st->sb;
820 mdp_disk_t *dk = &sb->disks[dinfo->number];
821 struct devinfo *di, **dip;
822
823 dk->number = dinfo->number;
824 dk->major = dinfo->major;
825 dk->minor = dinfo->minor;
826 dk->raid_disk = dinfo->raid_disk;
827 dk->state = dinfo->state & ((1<<MD_DISK_ACTIVE) |
828 (1<<MD_DISK_SYNC));
829
830 sb->this_disk = sb->disks[dinfo->number];
831 sb->sb_csum = calc_sb0_csum(sb);
832
833 dip = (struct devinfo **)&st->info;
834 while (*dip)
835 dip = &(*dip)->next;
836 di = xmalloc(sizeof(struct devinfo));
837 di->fd = fd;
838 di->devname = devname;
839 di->disk = *dinfo;
840 di->next = NULL;
841 *dip = di;
842
843 return 0;
844 }
845
846 static int store_super0(struct supertype *st, int fd)
847 {
848 unsigned long long dsize;
849 unsigned long long offset;
850 mdp_super_t *super = st->sb;
851
852 if (!get_dev_size(fd, NULL, &dsize))
853 return 1;
854
855 if (dsize < MD_RESERVED_SECTORS*512)
856 return 2;
857
858 if (st->other) {
859 /* Writing out v1.0 metadata for --update=metadata */
860 int ret = 0;
861
862 offset = dsize/512 - 8*2;
863 offset &= ~(4*2-1);
864 offset *= 512;
865 if (lseek64(fd, offset, 0)< 0LL)
866 ret = 3;
867 else if (write(fd, st->other, 1024) != 1024)
868 ret = 4;
869 else
870 fsync(fd);
871 free(st->other);
872 st->other = NULL;
873 return ret;
874 }
875
876 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
877
878 offset *= 512;
879
880 if (lseek64(fd, offset, 0)< 0LL)
881 return 3;
882
883 if (write(fd, super, sizeof(*super)) != sizeof(*super))
884 return 4;
885
886 if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
887 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
888 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
889 if (write(fd, bm, ROUND_UP(sizeof(*bm),4096)) !=
890 ROUND_UP(sizeof(*bm),4096))
891 return 5;
892 }
893
894 fsync(fd);
895 return 0;
896 }
897
898 static int write_init_super0(struct supertype *st)
899 {
900 mdp_super_t *sb = st->sb;
901 int rv = 0;
902 struct devinfo *di;
903
904 for (di = st->info ; di && ! rv ; di = di->next) {
905
906 if (di->disk.state & (1 << MD_DISK_FAULTY))
907 continue;
908 if (di->fd == -1)
909 continue;
910 while (Kill(di->devname, NULL, 0, -1, 1) == 0)
911 ;
912
913 sb->disks[di->disk.number].state &= ~(1<<MD_DISK_FAULTY);
914
915 sb->this_disk = sb->disks[di->disk.number];
916 sb->sb_csum = calc_sb0_csum(sb);
917 rv = store_super0(st, di->fd);
918
919 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
920 rv = st->ss->write_bitmap(st, di->fd, NoUpdate);
921
922 if (rv)
923 pr_err("failed to write superblock to %s\n",
924 di->devname);
925 }
926 return rv;
927 }
928
929 static int compare_super0(struct supertype *st, struct supertype *tst)
930 {
931 /*
932 * return:
933 * 0 same, or first was empty, and second was copied
934 * 1 second had wrong number
935 * 2 wrong uuid
936 * 3 wrong other info
937 */
938 mdp_super_t *first = st->sb;
939 mdp_super_t *second = tst->sb;
940 int uuid1[4], uuid2[4];
941
942 if (second->md_magic != MD_SB_MAGIC)
943 return 1;
944 if (!first) {
945 if (posix_memalign((void**)&first, 4096,
946 MD_SB_BYTES +
947 ROUND_UP(sizeof(struct bitmap_super_s), 4096)) != 0) {
948 pr_err("could not allocate superblock\n");
949 return 1;
950 }
951 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
952 st->sb = first;
953 return 0;
954 }
955
956 uuid_from_super0(st, uuid1);
957 uuid_from_super0(tst, uuid2);
958 if (!same_uuid(uuid1, uuid2, 0))
959 return 2;
960 if (first->major_version != second->major_version ||
961 first->minor_version != second->minor_version ||
962 first->patch_version != second->patch_version ||
963 first->gvalid_words != second->gvalid_words ||
964 first->ctime != second->ctime ||
965 first->level != second->level ||
966 first->size != second->size ||
967 first->raid_disks != second->raid_disks )
968 return 3;
969
970 return 0;
971 }
972
973 static void free_super0(struct supertype *st);
974
975 static int load_super0(struct supertype *st, int fd, char *devname)
976 {
977 /* try to read in the superblock
978 * Return:
979 * 0 on success
980 * 1 on cannot get superblock
981 * 2 on superblock meaningless
982 */
983 unsigned long long dsize;
984 unsigned long long offset;
985 mdp_super_t *super;
986 int uuid[4];
987 struct bitmap_super_s *bsb;
988
989 free_super0(st);
990
991 if (!get_dev_size(fd, devname, &dsize))
992 return 1;
993
994 if (dsize < MD_RESERVED_SECTORS*512) {
995 if (devname)
996 pr_err("%s is too small for md: size is %llu sectors.\n",
997 devname, dsize);
998 return 1;
999 }
1000 st->devsize = dsize;
1001
1002 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1003
1004 offset *= 512;
1005
1006 if (lseek64(fd, offset, 0)< 0LL) {
1007 if (devname)
1008 pr_err("Cannot seek to superblock on %s: %s\n",
1009 devname, strerror(errno));
1010 return 1;
1011 }
1012
1013 if (posix_memalign((void**)&super, 4096,
1014 MD_SB_BYTES +
1015 ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
1016 pr_err("could not allocate superblock\n");
1017 return 1;
1018 }
1019
1020 if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
1021 if (devname)
1022 pr_err("Cannot read superblock on %s\n",
1023 devname);
1024 free(super);
1025 return 1;
1026 }
1027
1028 if (st->ss && st->minor_version == 9)
1029 super0_swap_endian(super);
1030
1031 if (super->md_magic != MD_SB_MAGIC) {
1032 if (devname)
1033 pr_err("No super block found on %s (Expected magic %08x, got %08x)\n",
1034 devname, MD_SB_MAGIC, super->md_magic);
1035 free(super);
1036 return 2;
1037 }
1038
1039 if (super->major_version != 0) {
1040 if (devname)
1041 pr_err("Cannot interpret superblock on %s - version is %d\n",
1042 devname, super->major_version);
1043 free(super);
1044 return 2;
1045 }
1046 st->sb = super;
1047
1048 if (st->ss == NULL) {
1049 st->ss = &super0;
1050 st->minor_version = super->minor_version;
1051 st->max_devs = MD_SB_DISKS;
1052 st->info = NULL;
1053 }
1054
1055 /* Now check on the bitmap superblock */
1056 if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
1057 return 0;
1058 /* Read the bitmap superblock and make sure it looks
1059 * valid. If it doesn't clear the bit. An --assemble --force
1060 * should get that written out.
1061 */
1062 if (read(fd, super+1, ROUND_UP(sizeof(struct bitmap_super_s),4096)) !=
1063 ROUND_UP(sizeof(struct bitmap_super_s), 4096))
1064 goto no_bitmap;
1065
1066 uuid_from_super0(st, uuid);
1067 bsb = (struct bitmap_super_s *)(super+1);
1068 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
1069 memcmp(bsb->uuid, uuid, 16) != 0)
1070 goto no_bitmap;
1071 return 0;
1072
1073 no_bitmap:
1074 super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
1075
1076 return 0;
1077 }
1078
1079 static struct supertype *match_metadata_desc0(char *arg)
1080 {
1081 struct supertype *st = xcalloc(1, sizeof(*st));
1082
1083 st->container_devnm[0] = 0;
1084 st->ss = &super0;
1085 st->info = NULL;
1086 st->minor_version = 90;
1087 st->max_devs = MD_SB_DISKS;
1088 st->sb = NULL;
1089 /* we sometimes get 00.90 */
1090 while (arg[0] == '0' && arg[1] == '0')
1091 arg++;
1092 if (strcmp(arg, "0") == 0 ||
1093 #ifdef DEFAULT_OLD_METADATA /* ifndef in super1.c */
1094 strcmp(arg, "default") == 0 ||
1095 #endif /* DEFAULT_OLD_METADATA */
1096 strcmp(arg, "0.90") == 0 ||
1097 strcmp(arg, "") == 0 /* no metadata - i.e. non_persistent */
1098 )
1099 return st;
1100
1101 st->minor_version = 91; /* reshape in progress */
1102 if (strcmp(arg, "0.91") == 0) /* For dup_super support */
1103 return st;
1104
1105 st->minor_version = 9; /* flag for 'byte-swapped' */
1106 if (strcmp(arg, "0.swap")==0 ||
1107 strcmp(arg, "0.9") == 0) /* For dup_super support */
1108 return st;
1109
1110 free(st);
1111 return NULL;
1112 }
1113
1114 static __u64 avail_size0(struct supertype *st, __u64 devsize,
1115 unsigned long long data_offset)
1116 {
1117 if (data_offset != 0 && data_offset != INVALID_SECTORS)
1118 return 0ULL;
1119 if (devsize < MD_RESERVED_SECTORS)
1120 return 0ULL;
1121 return MD_NEW_SIZE_SECTORS(devsize);
1122 }
1123
1124 static int add_internal_bitmap0(struct supertype *st, int *chunkp,
1125 int delay, int write_behind,
1126 unsigned long long size, int may_change,
1127 int major)
1128 {
1129 /*
1130 * The bitmap comes immediately after the superblock and must be 60K in size
1131 * at most. The default size is between 30K and 60K
1132 *
1133 * size is in sectors, chunk is in bytes !!!
1134 */
1135 unsigned long long bits;
1136 unsigned long long max_bits = (60*1024 - sizeof(bitmap_super_t))*8;
1137 unsigned long long min_chunk;
1138 int chunk = *chunkp;
1139 mdp_super_t *sb = st->sb;
1140 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
1141 int uuid[4];
1142
1143 min_chunk = 4096; /* sub-page chunks don't work yet.. */
1144 bits = (size * 512) / min_chunk + 1;
1145 while (bits > max_bits) {
1146 min_chunk *= 2;
1147 bits = (bits+1)/2;
1148 }
1149 if (chunk == UnSet) {
1150 /* A chunk size less than a few Megabytes gives poor
1151 * performance without increasing resync noticeably
1152 */
1153 chunk = min_chunk;
1154 if (chunk < 64*1024*1024)
1155 chunk = 64*1024*1024;
1156 } else if ((unsigned long long)chunk < min_chunk)
1157 return -EINVAL; /* chunk size too small */
1158
1159 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
1160
1161 memset(bms, 0, sizeof(*bms));
1162 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
1163 bms->version = __cpu_to_le32(major);
1164 uuid_from_super0(st, uuid);
1165 memcpy(bms->uuid, uuid, 16);
1166 bms->chunksize = __cpu_to_le32(chunk);
1167 bms->daemon_sleep = __cpu_to_le32(delay);
1168 bms->sync_size = __cpu_to_le64(size);
1169 bms->write_behind = __cpu_to_le32(write_behind);
1170 *chunkp = chunk;
1171 return 0;
1172 }
1173
1174 static int locate_bitmap0(struct supertype *st, int fd, int node_num)
1175 {
1176 unsigned long long dsize;
1177 unsigned long long offset;
1178
1179 if (!get_dev_size(fd, NULL, &dsize))
1180 return -1;
1181
1182 if (dsize < MD_RESERVED_SECTORS*512)
1183 return -1;
1184
1185 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1186
1187 offset *= 512;
1188
1189 offset += MD_SB_BYTES;
1190
1191 lseek64(fd, offset, 0);
1192 return 0;
1193 }
1194
1195 static int write_bitmap0(struct supertype *st, int fd, enum bitmap_update update)
1196 {
1197 unsigned long long dsize;
1198 unsigned long long offset;
1199 mdp_super_t *sb = st->sb;
1200
1201 int rv = 0;
1202
1203 int towrite, n;
1204 void *buf;
1205
1206 if (!get_dev_size(fd, NULL, &dsize))
1207 return 1;
1208
1209 if (dsize < MD_RESERVED_SECTORS*512)
1210 return -1;
1211
1212 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1213
1214 offset *= 512;
1215
1216 if (lseek64(fd, offset + 4096, 0)< 0LL)
1217 return 3;
1218
1219 if (posix_memalign(&buf, 4096, 4096))
1220 return -ENOMEM;
1221
1222 memset(buf, 0xff, 4096);
1223 memcpy(buf, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
1224 towrite = 60*1024;
1225 while (towrite > 0) {
1226 n = towrite;
1227 if (n > 4096)
1228 n = 4096;
1229 n = write(fd, buf, n);
1230 if (n > 0)
1231 towrite -= n;
1232 else
1233 break;
1234 memset(buf, 0xff, 4096);
1235 }
1236 fsync(fd);
1237 if (towrite)
1238 rv = -2;
1239
1240 free(buf);
1241 return rv;
1242 }
1243
1244 static void free_super0(struct supertype *st)
1245 {
1246 if (st->sb)
1247 free(st->sb);
1248 while (st->info) {
1249 struct devinfo *di = st->info;
1250 st->info = di->next;
1251 if (di->fd >= 0)
1252 close(di->fd);
1253 free(di);
1254 }
1255 st->sb = NULL;
1256 }
1257
1258 static int validate_geometry0(struct supertype *st, int level,
1259 int layout, int raiddisks,
1260 int *chunk, unsigned long long size,
1261 unsigned long long data_offset,
1262 char *subdev, unsigned long long *freesize,
1263 int consistency_policy, int verbose)
1264 {
1265 unsigned long long ldsize;
1266 int fd;
1267 unsigned int tbmax = 4;
1268
1269 /* prior to linux 3.1, a but limits usable device size to 2TB.
1270 * It was introduced in 2.6.29, but we won't worry about that detail
1271 */
1272 if (get_linux_version() < 3001000)
1273 tbmax = 2;
1274
1275 if (level == LEVEL_CONTAINER) {
1276 if (verbose)
1277 pr_err("0.90 metadata does not support containers\n");
1278 return 0;
1279 }
1280 if (raiddisks > MD_SB_DISKS) {
1281 if (verbose)
1282 pr_err("0.90 metadata supports at most %d devices per array\n",
1283 MD_SB_DISKS);
1284 return 0;
1285 }
1286 if (size >= tbmax * 2ULL*1024*1024*1024) {
1287 if (verbose)
1288 pr_err("0.90 metadata supports at most %d terabytes per device\n", tbmax);
1289 return 0;
1290 }
1291 if (*chunk == UnSet)
1292 *chunk = DEFAULT_CHUNK;
1293
1294 if (level == 0 && layout != UnSet) {
1295 if (verbose)
1296 pr_err("0.90 metadata does not support layouts for RAID0\n");
1297 return 0;
1298 }
1299
1300 if (!subdev)
1301 return 1;
1302
1303 fd = open(subdev, O_RDONLY|O_EXCL, 0);
1304 if (fd < 0) {
1305 if (verbose)
1306 pr_err("super0.90 cannot open %s: %s\n",
1307 subdev, strerror(errno));
1308 return 0;
1309 }
1310
1311 if (!get_dev_size(fd, subdev, &ldsize)) {
1312 close(fd);
1313 return 0;
1314 }
1315 close(fd);
1316
1317 if (ldsize < MD_RESERVED_SECTORS * 512)
1318 return 0;
1319 *freesize = MD_NEW_SIZE_SECTORS(ldsize >> 9);
1320 return 1;
1321 }
1322
1323 struct superswitch super0 = {
1324 .examine_super = examine_super0,
1325 .brief_examine_super = brief_examine_super0,
1326 .export_examine_super = export_examine_super0,
1327 .detail_super = detail_super0,
1328 .brief_detail_super = brief_detail_super0,
1329 .write_init_super = write_init_super0,
1330 .validate_geometry = validate_geometry0,
1331 .add_to_super = add_to_super0,
1332 .copy_metadata = copy_metadata0,
1333 .match_home = match_home0,
1334 .uuid_from_super = uuid_from_super0,
1335 .getinfo_super = getinfo_super0,
1336 .container_content = container_content0,
1337 .update_super = update_super0,
1338 .init_super = init_super0,
1339 .store_super = store_super0,
1340 .compare_super = compare_super0,
1341 .load_super = load_super0,
1342 .match_metadata_desc = match_metadata_desc0,
1343 .avail_size = avail_size0,
1344 .add_internal_bitmap = add_internal_bitmap0,
1345 .locate_bitmap = locate_bitmap0,
1346 .write_bitmap = write_bitmap0,
1347 .free_super = free_super0,
1348 .name = "0.90",
1349 };