]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super0.c
projects / thirdparty / mdadm.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Grow: support restart of new migrations.
[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 : %d.%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 > 0)
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 == 6) {
158 c = map_num(r6layout, sb->new_layout);
159 printf(" New Layout : %s\n", c?c:"-unknown-");
160 }
161 if (sb->level == 10) {
162 printf(" New Layout : near=%d, %s=%d\n",
163 sb->new_layout&255,
164 (sb->new_layout&0x10000)?"offset":"far",
165 (sb->new_layout>>8)&255);
166 }
167 }
168 if (sb->new_chunk != sb->chunk_size)
169 printf(" New Chunksize : %d\n", sb->new_chunk);
170 printf("\n");
171 }
172 atime = sb->utime;
173 printf(" Update Time : %.24s\n", ctime(&atime));
174 printf(" State : %s\n",
175 (sb->state&(1<<MD_SB_CLEAN))?"clean":"active");
176 if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
177 printf("Internal Bitmap : present\n");
178 printf(" Active Devices : %d\n", sb->active_disks);
179 printf("Working Devices : %d\n", sb->working_disks);
180 printf(" Failed Devices : %d\n", sb->failed_disks);
181 printf(" Spare Devices : %d\n", sb->spare_disks);
182 if (calc_sb0_csum(sb) == sb->sb_csum)
183 printf(" Checksum : %x - correct\n", sb->sb_csum);
184 else
185 printf(" Checksum : %x - expected %lx\n", sb->sb_csum, calc_sb0_csum(sb));
186 printf(" Events : %llu\n",
187 ((unsigned long long)sb->events_hi << 32)
188 + sb->events_lo);
189 printf("\n");
190 if (sb->level == 5) {
191 c = map_num(r5layout, sb->layout);
192 printf(" Layout : %s\n", c?c:"-unknown-");
193 }
194 if (sb->level == 6) {
195 c = map_num(r6layout, sb->layout);
196 printf(" Layout : %s\n", c?c:"-unknown-");
197 }
198 if (sb->level == 10) {
199 printf(" Layout :");
200 print_r10_layout(sb->layout);
201 printf("\n");
202 }
203 switch(sb->level) {
204 case 0:
205 case 4:
206 case 5:
207 case 6:
208 case 10:
209 printf(" Chunk Size : %dK\n", sb->chunk_size/1024);
210 break;
211 case -1:
212 printf(" Rounding : %dK\n", sb->chunk_size/1024);
213 break;
214 default: break;
215 }
216 printf("\n");
217 printf(" Number Major Minor RaidDevice State\n");
218 for (d= -1; d<(signed int)(sb->raid_disks+sb->spare_disks); d++) {
219 mdp_disk_t *dp;
220 char *dv;
221 char nb[5];
222 int wonly;
223 if (d>=0) dp = &sb->disks[d];
224 else dp = &sb->this_disk;
225 snprintf(nb, sizeof(nb), "%4d", d);
226 printf("%4s %5d %5d %5d %5d ", d < 0 ? "this" : nb,
227 dp->number, dp->major, dp->minor, dp->raid_disk);
228 wonly = dp->state & (1<<MD_DISK_WRITEMOSTLY);
229 dp->state &= ~(1<<MD_DISK_WRITEMOSTLY);
230 if (dp->state & (1<<MD_DISK_FAULTY)) printf(" faulty");
231 if (dp->state & (1<<MD_DISK_ACTIVE)) printf(" active");
232 if (dp->state & (1<<MD_DISK_SYNC)) printf(" sync");
233 if (dp->state & (1<<MD_DISK_REMOVED)) printf(" removed");
234 if (wonly) printf(" write-mostly");
235 if (dp->state == 0) printf(" spare");
236 if ((dv=map_dev(dp->major, dp->minor, 0)))
237 printf(" %s", dv);
238 printf("\n");
239 if (d == -1) printf("\n");
240 }
241 }
242
243 static void brief_examine_super0(struct supertype *st, int verbose)
244 {
245 mdp_super_t *sb = st->sb;
246 char *c=map_num(pers, sb->level);
247 char devname[20];
248
249 sprintf(devname, "/dev/md%d", sb->md_minor);
250
251 if (verbose) {
252 printf("ARRAY %s level=%s num-devices=%d",
253 devname,
254 c?c:"-unknown-", sb->raid_disks);
255 } else
256 printf("ARRAY %s", devname);
257
258 if (sb->minor_version >= 90)
259 printf(" UUID=%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
260 sb->set_uuid2, sb->set_uuid3);
261 else
262 printf(" UUID=%08x", sb->set_uuid0);
263 printf("\n");
264 }
265
266 static void export_examine_super0(struct supertype *st)
267 {
268 mdp_super_t *sb = st->sb;
269
270 printf("MD_LEVEL=%s\n", map_num(pers, sb->level));
271 printf("MD_DEVICES=%d\n", sb->raid_disks);
272 if (sb->minor_version >= 90)
273 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
274 sb->set_uuid0, sb->set_uuid1,
275 sb->set_uuid2, sb->set_uuid3);
276 else
277 printf("MD_UUID=%08x\n", sb->set_uuid0);
278 printf("MD_UPDATE_TIME=%llu\n",
279 __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL);
280 printf("MD_EVENTS=%llu\n",
281 ((unsigned long long)sb->events_hi << 32)
282 + sb->events_lo);
283 }
284
285 static void detail_super0(struct supertype *st, char *homehost)
286 {
287 mdp_super_t *sb = st->sb;
288 printf(" UUID : ");
289 if (sb->minor_version >= 90)
290 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
291 sb->set_uuid2, sb->set_uuid3);
292 else
293 printf("%08x", sb->set_uuid0);
294 if (homehost) {
295 char buf[20];
296 void *hash = sha1_buffer(homehost,
297 strlen(homehost),
298 buf);
299 if (memcmp(&sb->set_uuid2, hash, 8)==0)
300 printf(" (local to host %s)", homehost);
301 }
302 printf("\n Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
303 }
304
305 static void brief_detail_super0(struct supertype *st)
306 {
307 mdp_super_t *sb = st->sb;
308 printf(" 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 }
315 #endif
316
317 static int match_home0(struct supertype *st, char *homehost)
318 {
319 mdp_super_t *sb = st->sb;
320 char buf[20];
321 char *hash = sha1_buffer(homehost,
322 strlen(homehost),
323 buf);
324
325 return (memcmp(&sb->set_uuid2, hash, 8)==0);
326 }
327
328 static void uuid_from_super0(struct supertype *st, int uuid[4])
329 {
330 mdp_super_t *super = st->sb;
331 uuid[0] = super->set_uuid0;
332 if (super->minor_version >= 90) {
333 uuid[1] = super->set_uuid1;
334 uuid[2] = super->set_uuid2;
335 uuid[3] = super->set_uuid3;
336 } else {
337 uuid[1] = 0;
338 uuid[2] = 0;
339 uuid[3] = 0;
340 }
341 }
342
343 static void getinfo_super0(struct supertype *st, struct mdinfo *info)
344 {
345 mdp_super_t *sb = st->sb;
346 int working = 0;
347 int i;
348
349 info->array.major_version = sb->major_version;
350 info->array.minor_version = sb->minor_version;
351 info->array.patch_version = sb->patch_version;
352 info->array.raid_disks = sb->raid_disks;
353 info->array.level = sb->level;
354 info->array.layout = sb->layout;
355 info->array.md_minor = sb->md_minor;
356 info->array.ctime = sb->ctime;
357 info->array.utime = sb->utime;
358 info->array.chunk_size = sb->chunk_size;
359 info->array.state = sb->state;
360 info->component_size = sb->size*2;
361
362 info->disk.state = sb->this_disk.state;
363 info->disk.major = sb->this_disk.major;
364 info->disk.minor = sb->this_disk.minor;
365 info->disk.raid_disk = sb->this_disk.raid_disk;
366 info->disk.number = sb->this_disk.number;
367
368 info->events = md_event(sb);
369 info->data_offset = 0;
370
371 sprintf(info->text_version, "0.%d", sb->minor_version);
372 info->safe_mode_delay = 200;
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 if (posix_memalign((void**)&sb, 4096,
561 MD_SB_BYTES + ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
562 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
563 return 0;
564 }
565 memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
566
567 st->sb = sb;
568 if (info == NULL) {
569 /* zeroing the superblock */
570 return 0;
571 }
572
573 spares = info->working_disks - info->active_disks;
574 if (info->raid_disks + spares > MD_SB_DISKS) {
575 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
576 info->raid_disks , spares, MD_SB_DISKS);
577 return 0;
578 }
579
580 sb->md_magic = MD_SB_MAGIC;
581 sb->major_version = 0;
582 sb->minor_version = 90;
583 sb->patch_version = 0;
584 sb->gvalid_words = 0; /* ignored */
585 sb->ctime = time(0);
586 sb->level = info->level;
587 if (size != info->size)
588 return 0;
589 sb->size = info->size;
590 sb->nr_disks = info->nr_disks;
591 sb->raid_disks = info->raid_disks;
592 sb->md_minor = info->md_minor;
593 sb->not_persistent = 0;
594 if (uuid) {
595 sb->set_uuid0 = uuid[0];
596 sb->set_uuid1 = uuid[1];
597 sb->set_uuid2 = uuid[2];
598 sb->set_uuid3 = uuid[3];
599 } else {
600 int rfd = open("/dev/urandom", O_RDONLY);
601 if (rfd < 0 || read(rfd, &sb->set_uuid0, 4) != 4)
602 sb->set_uuid0 = random();
603 if (rfd < 0 || read(rfd, &sb->set_uuid1, 12) != 12) {
604 sb->set_uuid1 = random();
605 sb->set_uuid2 = random();
606 sb->set_uuid3 = random();
607 }
608 if (rfd >= 0)
609 close(rfd);
610 }
611 if (homehost) {
612 char buf[20];
613 char *hash = sha1_buffer(homehost,
614 strlen(homehost),
615 buf);
616 memcpy(&sb->set_uuid2, hash, 8);
617 }
618
619 sb->utime = sb->ctime;
620 sb->state = info->state;
621 sb->active_disks = info->active_disks;
622 sb->working_disks = info->working_disks;
623 sb->failed_disks = info->failed_disks;
624 sb->spare_disks = info->spare_disks;
625 sb->events_hi = 0;
626 sb->events_lo = 1;
627
628 sb->layout = info->layout;
629 sb->chunk_size = info->chunk_size;
630
631 return 1;
632 }
633
634 struct devinfo {
635 int fd;
636 char *devname;
637 mdu_disk_info_t disk;
638 struct devinfo *next;
639 };
640
641 #ifndef MDASSEMBLE
642 /* Add a device to the superblock being created */
643 static int add_to_super0(struct supertype *st, mdu_disk_info_t *dinfo,
644 int fd, char *devname)
645 {
646 mdp_super_t *sb = st->sb;
647 mdp_disk_t *dk = &sb->disks[dinfo->number];
648 struct devinfo *di, **dip;
649
650 dk->number = dinfo->number;
651 dk->major = dinfo->major;
652 dk->minor = dinfo->minor;
653 dk->raid_disk = dinfo->raid_disk;
654 dk->state = dinfo->state;
655
656 sb->this_disk = sb->disks[dinfo->number];
657 sb->sb_csum = calc_sb0_csum(sb);
658
659 dip = (struct devinfo **)&st->info;
660 while (*dip)
661 dip = &(*dip)->next;
662 di = malloc(sizeof(struct devinfo));
663 di->fd = fd;
664 di->devname = devname;
665 di->disk = *dinfo;
666 di->next = NULL;
667 *dip = di;
668
669 return 0;
670 }
671 #endif
672
673 static int store_super0(struct supertype *st, int fd)
674 {
675 unsigned long long dsize;
676 unsigned long long offset;
677 mdp_super_t *super = st->sb;
678
679 if (!get_dev_size(fd, NULL, &dsize))
680 return 1;
681
682 if (dsize < MD_RESERVED_SECTORS*512)
683 return 2;
684
685 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
686
687 offset *= 512;
688
689 if (lseek64(fd, offset, 0)< 0LL)
690 return 3;
691
692 if (write(fd, super, sizeof(*super)) != sizeof(*super))
693 return 4;
694
695 if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
696 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
697 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
698 if (write(fd, bm, ROUND_UP(sizeof(*bm),4096)) !=
699 ROUND_UP(sizeof(*bm),4096))
700 return 5;
701 }
702
703 fsync(fd);
704 return 0;
705 }
706
707 #ifndef MDASSEMBLE
708 static int write_init_super0(struct supertype *st)
709 {
710 mdp_super_t *sb = st->sb;
711 int rv = 0;
712 struct devinfo *di;
713
714 for (di = st->info ; di && ! rv ; di = di->next) {
715
716 if (di->disk.state == 1)
717 continue;
718 if (di->fd == -1)
719 continue;
720 Kill(di->devname, 0, 1, 1);
721 Kill(di->devname, 0, 1, 1);
722
723 sb->disks[di->disk.number].state &= ~(1<<MD_DISK_FAULTY);
724
725 sb->this_disk = sb->disks[di->disk.number];
726 sb->sb_csum = calc_sb0_csum(sb);
727 rv = store_super0(st, di->fd);
728
729 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
730 rv = st->ss->write_bitmap(st, di->fd);
731
732 if (rv)
733 fprintf(stderr,
734 Name ": failed to write superblock to %s\n",
735 di->devname);
736 close(di->fd);
737 di->fd = -1;
738 }
739 return rv;
740 }
741 #endif
742
743 static int compare_super0(struct supertype *st, struct supertype *tst)
744 {
745 /*
746 * return:
747 * 0 same, or first was empty, and second was copied
748 * 1 second had wrong number
749 * 2 wrong uuid
750 * 3 wrong other info
751 */
752 mdp_super_t *first = st->sb;
753 mdp_super_t *second = tst->sb;
754 int uuid1[4], uuid2[4];
755
756 if (second->md_magic != MD_SB_MAGIC)
757 return 1;
758 if (!first) {
759 if (posix_memalign((void**)&first, 4096,
760 MD_SB_BYTES +
761 ROUND_UP(sizeof(struct bitmap_super_s), 4096)) != 0) {
762 fprintf(stderr, Name
763 ": %s could not allocate superblock\n", __func__);
764 return 1;
765 }
766 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
767 st->sb = first;
768 return 0;
769 }
770
771 uuid_from_super0(st, uuid1);
772 uuid_from_super0(tst, uuid2);
773 if (!same_uuid(uuid1, uuid2, 0))
774 return 2;
775 if (first->major_version != second->major_version ||
776 first->minor_version != second->minor_version ||
777 first->patch_version != second->patch_version ||
778 first->gvalid_words != second->gvalid_words ||
779 first->ctime != second->ctime ||
780 first->level != second->level ||
781 first->size != second->size ||
782 first->raid_disks != second->raid_disks )
783 return 3;
784
785 return 0;
786 }
787
788
789 static void free_super0(struct supertype *st);
790
791 static int load_super0(struct supertype *st, int fd, char *devname)
792 {
793 /* try to read in the superblock
794 * Return:
795 * 0 on success
796 * 1 on cannot get superblock
797 * 2 on superblock meaningless
798 */
799 unsigned long long dsize;
800 unsigned long long offset;
801 mdp_super_t *super;
802 int uuid[4];
803 struct bitmap_super_s *bsb;
804
805 free_super0(st);
806
807 if (st->subarray[0])
808 return 1;
809
810 if (!get_dev_size(fd, devname, &dsize))
811 return 1;
812
813 if (dsize < MD_RESERVED_SECTORS*512) {
814 if (devname)
815 fprintf(stderr, Name
816 ": %s is too small for md: size is %llu sectors.\n",
817 devname, dsize);
818 return 1;
819 }
820
821 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
822
823 offset *= 512;
824
825 ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
826
827 if (lseek64(fd, offset, 0)< 0LL) {
828 if (devname)
829 fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
830 devname, strerror(errno));
831 return 1;
832 }
833
834 if (posix_memalign((void**)&super, 4096,
835 MD_SB_BYTES +
836 ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
837 fprintf(stderr, Name
838 ": %s could not allocate superblock\n", __func__);
839 return 1;
840 }
841
842 if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
843 if (devname)
844 fprintf(stderr, Name ": Cannot read superblock on %s\n",
845 devname);
846 free(super);
847 return 1;
848 }
849
850 if (st->ss && st->minor_version == 9)
851 super0_swap_endian(super);
852
853 if (super->md_magic != MD_SB_MAGIC) {
854 if (devname)
855 fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
856 devname, MD_SB_MAGIC, super->md_magic);
857 free(super);
858 return 2;
859 }
860
861 if (super->major_version != 0) {
862 if (devname)
863 fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
864 devname, super->major_version);
865 free(super);
866 return 2;
867 }
868 st->sb = super;
869
870 if (st->ss == NULL) {
871 st->ss = &super0;
872 st->minor_version = super->minor_version;
873 st->max_devs = MD_SB_DISKS;
874 st->info = NULL;
875 }
876
877 /* Now check on the bitmap superblock */
878 if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
879 return 0;
880 /* Read the bitmap superblock and make sure it looks
881 * valid. If it doesn't clear the bit. An --assemble --force
882 * should get that written out.
883 */
884 if (read(fd, super+1, ROUND_UP(sizeof(struct bitmap_super_s),4096))
885 != ROUND_UP(sizeof(struct bitmap_super_s),4096))
886 goto no_bitmap;
887
888 uuid_from_super0(st, uuid);
889 bsb = (struct bitmap_super_s *)(super+1);
890 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
891 memcmp(bsb->uuid, uuid, 16) != 0)
892 goto no_bitmap;
893 return 0;
894
895 no_bitmap:
896 super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
897
898 return 0;
899 }
900
901 static struct supertype *match_metadata_desc0(char *arg)
902 {
903 struct supertype *st = malloc(sizeof(*st));
904 if (!st) return st;
905
906 memset(st, 0, sizeof(*st));
907 st->ss = &super0;
908 st->info = NULL;
909 st->minor_version = 90;
910 st->max_devs = MD_SB_DISKS;
911 st->sb = NULL;
912 /* we sometimes get 00.90 */
913 while (arg[0] == '0' && arg[1] == '0')
914 arg++;
915 if (strcmp(arg, "0") == 0 ||
916 strcmp(arg, "0.90") == 0 ||
917 strcmp(arg, "default") == 0 ||
918 strcmp(arg, "") == 0 /* no metadata */
919 )
920 return st;
921
922 st->minor_version = 91; /* reshape in progress */
923 if (strcmp(arg, "0.91") == 0) /* For dup_super support */
924 return st;
925
926 st->minor_version = 9; /* flag for 'byte-swapped' */
927 if (strcmp(arg, "0.swap")==0 ||
928 strcmp(arg, "0.9") == 0) /* For dup_super support */
929 return st;
930
931 free(st);
932 return NULL;
933 }
934
935 static __u64 avail_size0(struct supertype *st, __u64 devsize)
936 {
937 if (devsize < MD_RESERVED_SECTORS)
938 return 0ULL;
939 return MD_NEW_SIZE_SECTORS(devsize);
940 }
941
942 static int add_internal_bitmap0(struct supertype *st, int *chunkp,
943 int delay, int write_behind,
944 unsigned long long size, int may_change,
945 int major)
946 {
947 /*
948 * The bitmap comes immediately after the superblock and must be 60K in size
949 * at most. The default size is between 30K and 60K
950 *
951 * size is in sectors, chunk is in bytes !!!
952 */
953 unsigned long long bits;
954 unsigned long long max_bits = 60*1024*8;
955 unsigned long long min_chunk;
956 int chunk = *chunkp;
957 mdp_super_t *sb = st->sb;
958 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
959
960
961 min_chunk = 4096; /* sub-page chunks don't work yet.. */
962 bits = (size * 512) / min_chunk + 1;
963 while (bits > max_bits) {
964 min_chunk *= 2;
965 bits = (bits+1)/2;
966 }
967 if (chunk == UnSet)
968 chunk = min_chunk;
969 else if (chunk < min_chunk)
970 return 0; /* chunk size too small */
971
972 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
973
974 memset(bms, 0, sizeof(*bms));
975 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
976 bms->version = __cpu_to_le32(major);
977 uuid_from_super0(st, (int*)bms->uuid);
978 bms->chunksize = __cpu_to_le32(chunk);
979 bms->daemon_sleep = __cpu_to_le32(delay);
980 bms->sync_size = __cpu_to_le64(size);
981 bms->write_behind = __cpu_to_le32(write_behind);
982 *chunkp = chunk;
983 return 1;
984 }
985
986
987 static void locate_bitmap0(struct supertype *st, int fd)
988 {
989 unsigned long long dsize;
990 unsigned long long offset;
991
992 if (!get_dev_size(fd, NULL, &dsize))
993 return;
994
995 if (dsize < MD_RESERVED_SECTORS*512)
996 return;
997
998 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
999
1000 offset *= 512;
1001
1002 offset += MD_SB_BYTES;
1003
1004 lseek64(fd, offset, 0);
1005 }
1006
1007 static int write_bitmap0(struct supertype *st, int fd)
1008 {
1009 unsigned long long dsize;
1010 unsigned long long offset;
1011 mdp_super_t *sb = st->sb;
1012
1013 int rv = 0;
1014
1015 int towrite, n;
1016 char abuf[4096+4096];
1017 char *buf = (char*)(((long)(abuf+4096))&~4095L);
1018
1019 if (!get_dev_size(fd, NULL, &dsize))
1020 return 1;
1021
1022
1023 if (dsize < MD_RESERVED_SECTORS*512)
1024 return -1;
1025
1026 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1027
1028 offset *= 512;
1029
1030 if (lseek64(fd, offset + 4096, 0)< 0LL)
1031 return 3;
1032
1033 memset(buf, 0xff, 4096);
1034 memcpy(buf, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
1035 towrite = 64*1024;
1036 while (towrite > 0) {
1037 n = towrite;
1038 if (n > 4096)
1039 n = 4096;
1040 n = write(fd, buf, n);
1041 if (n > 0)
1042 towrite -= n;
1043 else
1044 break;
1045 memset(buf, 0xff, 4096);
1046 }
1047 fsync(fd);
1048 if (towrite)
1049 rv = -2;
1050
1051 return rv;
1052 }
1053
1054 static void free_super0(struct supertype *st)
1055 {
1056 if (st->sb)
1057 free(st->sb);
1058 st->sb = NULL;
1059 }
1060
1061 #ifndef MDASSEMBLE
1062 static int validate_geometry0(struct supertype *st, int level,
1063 int layout, int raiddisks,
1064 int chunk, unsigned long long size,
1065 char *subdev, unsigned long long *freesize,
1066 int verbose)
1067 {
1068 unsigned long long ldsize;
1069 int fd;
1070
1071 if (level == LEVEL_CONTAINER)
1072 return 0;
1073 if (raiddisks > MD_SB_DISKS)
1074 return 0;
1075 if (size > (0x7fffffffULL<<9))
1076 return 0;
1077 if (!subdev)
1078 return 1;
1079
1080 fd = open(subdev, O_RDONLY|O_EXCL, 0);
1081 if (fd < 0) {
1082 if (verbose)
1083 fprintf(stderr, Name ": super0.90 cannot open %s: %s\n",
1084 subdev, strerror(errno));
1085 return 0;
1086 }
1087
1088 if (!get_dev_size(fd, subdev, &ldsize)) {
1089 close(fd);
1090 return 0;
1091 }
1092 close(fd);
1093
1094 if (ldsize < MD_RESERVED_SECTORS * 512)
1095 return 0;
1096 if (size > (0x7fffffffULL<<9))
1097 return 0;
1098 *freesize = MD_NEW_SIZE_SECTORS(ldsize >> 9);
1099 return 1;
1100 }
1101 #endif /* MDASSEMBLE */
1102
1103 struct superswitch super0 = {
1104 #ifndef MDASSEMBLE
1105 .examine_super = examine_super0,
1106 .brief_examine_super = brief_examine_super0,
1107 .export_examine_super = export_examine_super0,
1108 .detail_super = detail_super0,
1109 .brief_detail_super = brief_detail_super0,
1110 .write_init_super = write_init_super0,
1111 .validate_geometry = validate_geometry0,
1112 .add_to_super = add_to_super0,
1113 #endif
1114 .match_home = match_home0,
1115 .uuid_from_super = uuid_from_super0,
1116 .getinfo_super = getinfo_super0,
1117 .update_super = update_super0,
1118 .init_super = init_super0,
1119 .store_super = store_super0,
1120 .compare_super = compare_super0,
1121 .load_super = load_super0,
1122 .match_metadata_desc = match_metadata_desc0,
1123 .avail_size = avail_size0,
1124 .add_internal_bitmap = add_internal_bitmap0,
1125 .locate_bitmap = locate_bitmap0,
1126 .write_bitmap = write_bitmap0,
1127 .free_super = free_super0,
1128 .name = "0.90",
1129 };
Unnamed repository; edit this file 'description' to name the repository.