]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super0.c
Don't close fds in write_init_super
[thirdparty/mdadm.git] / super0.c
1 /*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
22 * Email: <neilb@suse.de>
23 */
24
25 #define HAVE_STDINT_H 1
26 #include "mdadm.h"
27 #include "sha1.h"
28 /*
29 * All handling for the 0.90.0 version superblock is in
30 * this file.
31 * This includes:
32 * - finding, loading, and writing the superblock.
33 * - initialising a new superblock
34 * - printing the superblock for --examine
35 * - printing part of the superblock for --detail
36 * .. other stuff
37 */
38
39
40 static unsigned long calc_sb0_csum(mdp_super_t *super)
41 {
42 unsigned long csum = super->sb_csum;
43 unsigned long newcsum;
44 super->sb_csum= 0 ;
45 newcsum = calc_csum(super, MD_SB_BYTES);
46 super->sb_csum = csum;
47 return newcsum;
48 }
49
50
51 static void super0_swap_endian(struct mdp_superblock_s *sb)
52 {
53 /* as super0 superblocks are host-endian, it is sometimes
54 * useful to be able to swap the endianness
55 * as (almost) everything is u32's we byte-swap every 4byte
56 * number.
57 * We then also have to swap the events_hi and events_lo
58 */
59 char *sbc = (char *)sb;
60 __u32 t32;
61 int i;
62
63 for (i=0; i < MD_SB_BYTES ; i+=4) {
64 char t = sbc[i];
65 sbc[i] = sbc[i+3];
66 sbc[i+3] = t;
67 t=sbc[i+1];
68 sbc[i+1]=sbc[i+2];
69 sbc[i+2]=t;
70 }
71 t32 = sb->events_hi;
72 sb->events_hi = sb->events_lo;
73 sb->events_lo = t32;
74
75 t32 = sb->cp_events_hi;
76 sb->cp_events_hi = sb->cp_events_lo;
77 sb->cp_events_lo = t32;
78
79 }
80
81 #ifndef MDASSEMBLE
82
83 static void examine_super0(struct supertype *st, char *homehost)
84 {
85 mdp_super_t *sb = st->sb;
86 time_t atime;
87 int d;
88 int delta_extra = 0;
89 char *c;
90
91 printf(" Magic : %08x\n", sb->md_magic);
92 printf(" Version : %d.%02d.%02d\n", sb->major_version, sb->minor_version,
93 sb->patch_version);
94 if (sb->minor_version >= 90) {
95 printf(" UUID : %08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
96 sb->set_uuid2, sb->set_uuid3);
97 if (homehost) {
98 char buf[20];
99 void *hash = sha1_buffer(homehost,
100 strlen(homehost),
101 buf);
102 if (memcmp(&sb->set_uuid2, hash, 8)==0)
103 printf(" (local to host %s)", homehost);
104 }
105 printf("\n");
106 } else
107 printf(" UUID : %08x\n", sb->set_uuid0);
108
109 if (sb->not_persistent)
110 printf(" Eedk : not persistent\n");
111
112 atime = sb->ctime;
113 printf(" Creation Time : %.24s\n", ctime(&atime));
114 c=map_num(pers, sb->level);
115 printf(" Raid Level : %s\n", c?c:"-unknown-");
116 if ((int)sb->level > 0) {
117 int ddsks=0;
118 printf(" Used Dev Size : %d%s\n", sb->size,
119 human_size((long long)sb->size<<10));
120 switch(sb->level) {
121 case 1: ddsks=1;break;
122 case 4:
123 case 5: ddsks = sb->raid_disks-1; break;
124 case 6: ddsks = sb->raid_disks-2; break;
125 case 10: ddsks = sb->raid_disks / (sb->layout&255) / ((sb->layout>>8)&255);
126 }
127 if (ddsks)
128 printf(" Array Size : %llu%s\n", (unsigned long long)ddsks * sb->size,
129 human_size(ddsks*(long long)sb->size<<10));
130 }
131 printf(" Raid Devices : %d\n", sb->raid_disks);
132 printf(" Total Devices : %d\n", sb->nr_disks);
133 printf("Preferred Minor : %d\n", sb->md_minor);
134 printf("\n");
135 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
136 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)sb->reshape_position/2, human_size((long long)sb->reshape_position<<9));
137 if (sb->delta_disks) {
138 printf(" Delta Devices : %d", sb->delta_disks);
139 printf(" (%d->%d)\n", sb->raid_disks-sb->delta_disks, sb->raid_disks);
140 if (((int)sb->delta_disks) < 0)
141 delta_extra = - sb->delta_disks;
142 }
143 if (sb->new_level != sb->level) {
144 c = map_num(pers, sb->new_level);
145 printf(" New Level : %s\n", c?c:"-unknown-");
146 }
147 if (sb->new_layout != sb->layout) {
148 if (sb->level == 5) {
149 c = map_num(r5layout, sb->new_layout);
150 printf(" New Layout : %s\n", c?c:"-unknown-");
151 }
152 if (sb->level == 6) {
153 c = map_num(r6layout, sb->new_layout);
154 printf(" New Layout : %s\n", c?c:"-unknown-");
155 }
156 if (sb->level == 10) {
157 printf(" New Layout : near=%d, %s=%d\n",
158 sb->new_layout&255,
159 (sb->new_layout&0x10000)?"offset":"far",
160 (sb->new_layout>>8)&255);
161 }
162 }
163 if (sb->new_chunk != sb->chunk_size)
164 printf(" New Chunksize : %d\n", sb->new_chunk);
165 printf("\n");
166 }
167 atime = sb->utime;
168 printf(" Update Time : %.24s\n", ctime(&atime));
169 printf(" State : %s\n",
170 (sb->state&(1<<MD_SB_CLEAN))?"clean":"active");
171 if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
172 printf("Internal Bitmap : present\n");
173 printf(" Active Devices : %d\n", sb->active_disks);
174 printf("Working Devices : %d\n", sb->working_disks);
175 printf(" Failed Devices : %d\n", sb->failed_disks);
176 printf(" Spare Devices : %d\n", sb->spare_disks);
177 if (calc_sb0_csum(sb) == sb->sb_csum)
178 printf(" Checksum : %x - correct\n", sb->sb_csum);
179 else
180 printf(" Checksum : %x - expected %lx\n", sb->sb_csum, calc_sb0_csum(sb));
181 printf(" Events : %llu\n",
182 ((unsigned long long)sb->events_hi << 32)
183 + sb->events_lo);
184 printf("\n");
185 if (sb->level == 5) {
186 c = map_num(r5layout, sb->layout);
187 printf(" Layout : %s\n", c?c:"-unknown-");
188 }
189 if (sb->level == 6) {
190 c = map_num(r6layout, sb->layout);
191 printf(" Layout : %s\n", c?c:"-unknown-");
192 }
193 if (sb->level == 10) {
194 printf(" Layout :");
195 print_r10_layout(sb->layout);
196 printf("\n");
197 }
198 switch(sb->level) {
199 case 0:
200 case 4:
201 case 5:
202 case 6:
203 case 10:
204 printf(" Chunk Size : %dK\n", sb->chunk_size/1024);
205 break;
206 case -1:
207 printf(" Rounding : %dK\n", sb->chunk_size/1024);
208 break;
209 default: break;
210 }
211 printf("\n");
212 printf(" Number Major Minor RaidDevice State\n");
213 for (d= -1; d<(signed int)(sb->raid_disks+delta_extra + sb->spare_disks); d++) {
214 mdp_disk_t *dp;
215 char *dv;
216 char nb[5];
217 int wonly;
218 if (d>=0) dp = &sb->disks[d];
219 else dp = &sb->this_disk;
220 snprintf(nb, sizeof(nb), "%4d", d);
221 printf("%4s %5d %5d %5d %5d ", d < 0 ? "this" : nb,
222 dp->number, dp->major, dp->minor, dp->raid_disk);
223 wonly = dp->state & (1<<MD_DISK_WRITEMOSTLY);
224 dp->state &= ~(1<<MD_DISK_WRITEMOSTLY);
225 if (dp->state & (1<<MD_DISK_FAULTY)) printf(" faulty");
226 if (dp->state & (1<<MD_DISK_ACTIVE)) printf(" active");
227 if (dp->state & (1<<MD_DISK_SYNC)) printf(" sync");
228 if (dp->state & (1<<MD_DISK_REMOVED)) printf(" removed");
229 if (wonly) printf(" write-mostly");
230 if (dp->state == 0) printf(" spare");
231 if ((dv=map_dev(dp->major, dp->minor, 0)))
232 printf(" %s", dv);
233 printf("\n");
234 if (d == -1) printf("\n");
235 }
236 }
237
238 static void brief_examine_super0(struct supertype *st, int verbose)
239 {
240 mdp_super_t *sb = st->sb;
241 char *c=map_num(pers, sb->level);
242 char devname[20];
243
244 sprintf(devname, "/dev/md%d", sb->md_minor);
245
246 if (verbose) {
247 printf("ARRAY %s level=%s num-devices=%d",
248 devname,
249 c?c:"-unknown-", sb->raid_disks);
250 } else
251 printf("ARRAY %s", devname);
252
253 if (sb->minor_version >= 90)
254 printf(" UUID=%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
255 sb->set_uuid2, sb->set_uuid3);
256 else
257 printf(" UUID=%08x", sb->set_uuid0);
258 printf("\n");
259 }
260
261 static void export_examine_super0(struct supertype *st)
262 {
263 mdp_super_t *sb = st->sb;
264
265 printf("MD_LEVEL=%s\n", map_num(pers, sb->level));
266 printf("MD_DEVICES=%d\n", sb->raid_disks);
267 if (sb->minor_version >= 90)
268 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
269 sb->set_uuid0, sb->set_uuid1,
270 sb->set_uuid2, sb->set_uuid3);
271 else
272 printf("MD_UUID=%08x\n", sb->set_uuid0);
273 printf("MD_UPDATE_TIME=%llu\n",
274 __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL);
275 printf("MD_EVENTS=%llu\n",
276 ((unsigned long long)sb->events_hi << 32)
277 + sb->events_lo);
278 }
279
280 static void detail_super0(struct supertype *st, char *homehost)
281 {
282 mdp_super_t *sb = st->sb;
283 printf(" UUID : ");
284 if (sb->minor_version >= 90)
285 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
286 sb->set_uuid2, sb->set_uuid3);
287 else
288 printf("%08x", sb->set_uuid0);
289 if (homehost) {
290 char buf[20];
291 void *hash = sha1_buffer(homehost,
292 strlen(homehost),
293 buf);
294 if (memcmp(&sb->set_uuid2, hash, 8)==0)
295 printf(" (local to host %s)", homehost);
296 }
297 printf("\n Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
298 }
299
300 static void brief_detail_super0(struct supertype *st)
301 {
302 mdp_super_t *sb = st->sb;
303 printf(" UUID=");
304 if (sb->minor_version >= 90)
305 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
306 sb->set_uuid2, sb->set_uuid3);
307 else
308 printf("%08x", sb->set_uuid0);
309 }
310 #endif
311
312 static int match_home0(struct supertype *st, char *homehost)
313 {
314 mdp_super_t *sb = st->sb;
315 char buf[20];
316 char *hash;
317
318 if (!homehost)
319 return 0;
320 hash = sha1_buffer(homehost,
321 strlen(homehost),
322 buf);
323
324 return (memcmp(&sb->set_uuid2, hash, 8)==0);
325 }
326
327 static void uuid_from_super0(struct supertype *st, int uuid[4])
328 {
329 mdp_super_t *super = st->sb;
330 uuid[0] = super->set_uuid0;
331 if (super->minor_version >= 90) {
332 uuid[1] = super->set_uuid1;
333 uuid[2] = super->set_uuid2;
334 uuid[3] = super->set_uuid3;
335 } else {
336 uuid[1] = 0;
337 uuid[2] = 0;
338 uuid[3] = 0;
339 }
340 }
341
342 static void getinfo_super0(struct supertype *st, struct mdinfo *info, char *map)
343 {
344 mdp_super_t *sb = st->sb;
345 int working = 0;
346 int i;
347 int map_disks = info->array.raid_disks;
348
349 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 info->recovery_start = MaxSector;
377 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
378 info->reshape_active = 1;
379 info->reshape_progress = sb->reshape_position;
380 info->new_level = sb->new_level;
381 info->delta_disks = sb->delta_disks;
382 info->new_layout = sb->new_layout;
383 info->new_chunk = sb->new_chunk;
384 if (info->delta_disks < 0)
385 info->array.raid_disks -= info->delta_disks;
386 } else
387 info->reshape_active = 0;
388
389 sprintf(info->name, "%d", sb->md_minor);
390 /* work_disks is calculated rather than read directly */
391 for (i=0; i < MD_SB_DISKS; i++)
392 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
393 (sb->disks[i].raid_disk < (unsigned)info->array.raid_disks) &&
394 (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
395 !(sb->disks[i].state & (1<<MD_DISK_FAULTY))) {
396 working ++;
397 if (map && i < map_disks)
398 map[i] = 1;
399 } else if (map && i < map_disks)
400 map[i] = 0;
401 info->array.working_disks = working;
402 }
403
404 static struct mdinfo *container_content0(struct supertype *st, char *subarray)
405 {
406 struct mdinfo *info;
407
408 if (subarray)
409 return NULL;
410
411 info = malloc(sizeof(*info));
412 getinfo_super0(st, info, NULL);
413 return info;
414 }
415
416 static int update_super0(struct supertype *st, struct mdinfo *info,
417 char *update,
418 char *devname, int verbose,
419 int uuid_set, char *homehost)
420 {
421 /* NOTE: for 'assemble' and 'force' we need to return non-zero
422 * if any change was made. For others, the return value is
423 * ignored.
424 */
425 int rv = 0;
426 mdp_super_t *sb = st->sb;
427 if (strcmp(update, "sparc2.2")==0 ) {
428 /* 2.2 sparc put the events in the wrong place
429 * So we copy the tail of the superblock
430 * up 4 bytes before continuing
431 */
432 __u32 *sb32 = (__u32*)sb;
433 memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
434 sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
435 (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
436 if (verbose >= 0)
437 fprintf (stderr, Name ": adjusting superblock of %s for 2.2/sparc compatability.\n",
438 devname);
439 } else if (strcmp(update, "super-minor") ==0) {
440 sb->md_minor = info->array.md_minor;
441 if (verbose > 0)
442 fprintf(stderr, Name ": updating superblock of %s with minor number %d\n",
443 devname, info->array.md_minor);
444 } else if (strcmp(update, "summaries") == 0) {
445 unsigned int i;
446 /* set nr_disks, active_disks, working_disks,
447 * failed_disks, spare_disks based on disks[]
448 * array in superblock.
449 * Also make sure extra slots aren't 'failed'
450 */
451 sb->nr_disks = sb->active_disks =
452 sb->working_disks = sb->failed_disks =
453 sb->spare_disks = 0;
454 for (i=0; i < MD_SB_DISKS ; i++)
455 if (sb->disks[i].major ||
456 sb->disks[i].minor) {
457 int state = sb->disks[i].state;
458 if (state & (1<<MD_DISK_REMOVED))
459 continue;
460 sb->nr_disks++;
461 if (state & (1<<MD_DISK_ACTIVE))
462 sb->active_disks++;
463 if (state & (1<<MD_DISK_FAULTY))
464 sb->failed_disks++;
465 else
466 sb->working_disks++;
467 if (state == 0)
468 sb->spare_disks++;
469 } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
470 sb->disks[i].state = 0;
471 } else if (strcmp(update, "force-one")==0) {
472 /* Not enough devices for a working array, so
473 * bring this one up-to-date.
474 */
475 __u32 ehi = sb->events_hi, elo = sb->events_lo;
476 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
477 sb->events_lo = (info->events) & 0xFFFFFFFF;
478 if (sb->events_hi != ehi ||
479 sb->events_lo != elo)
480 rv = 1;
481 } else if (strcmp(update, "force-array")==0) {
482 /* degraded array and 'force' requested, so
483 * maybe need to mark it 'clean'
484 */
485 if ((sb->level == 5 || sb->level == 4 || sb->level == 6) &&
486 (sb->state & (1 << MD_SB_CLEAN)) == 0) {
487 /* need to force clean */
488 sb->state |= (1 << MD_SB_CLEAN);
489 rv = 1;
490 }
491 } else if (strcmp(update, "assemble")==0) {
492 int d = info->disk.number;
493 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
494 int mask = (1<<MD_DISK_WRITEMOSTLY);
495 int add = 0;
496 if (sb->minor_version >= 91)
497 /* During reshape we don't insist on everything
498 * being marked 'sync'
499 */
500 add = (1<<MD_DISK_SYNC);
501 if (((sb->disks[d].state & ~mask) | add)
502 != (unsigned)info->disk.state) {
503 sb->disks[d].state = info->disk.state | wonly;
504 rv = 1;
505 }
506 } else if (strcmp(update, "linear-grow-new") == 0) {
507 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
508 sb->disks[info->disk.number].number = info->disk.number;
509 sb->disks[info->disk.number].major = info->disk.major;
510 sb->disks[info->disk.number].minor = info->disk.minor;
511 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
512 sb->disks[info->disk.number].state = info->disk.state;
513 sb->this_disk = sb->disks[info->disk.number];
514 } else if (strcmp(update, "linear-grow-update") == 0) {
515 sb->raid_disks = info->array.raid_disks;
516 sb->nr_disks = info->array.nr_disks;
517 sb->active_disks = info->array.active_disks;
518 sb->working_disks = info->array.working_disks;
519 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
520 sb->disks[info->disk.number].number = info->disk.number;
521 sb->disks[info->disk.number].major = info->disk.major;
522 sb->disks[info->disk.number].minor = info->disk.minor;
523 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
524 sb->disks[info->disk.number].state = info->disk.state;
525 } else if (strcmp(update, "resync") == 0) {
526 /* make sure resync happens */
527 sb->state &= ~(1<<MD_SB_CLEAN);
528 sb->recovery_cp = 0;
529 } else if (strcmp(update, "homehost") == 0 &&
530 homehost) {
531 uuid_set = 0;
532 update = "uuid";
533 info->uuid[0] = sb->set_uuid0;
534 info->uuid[1] = sb->set_uuid1;
535 } else if (strcmp(update, "uuid") == 0) {
536 if (!uuid_set && homehost) {
537 char buf[20];
538 char *hash = sha1_buffer(homehost,
539 strlen(homehost),
540 buf);
541 memcpy(info->uuid+2, hash, 8);
542 }
543 sb->set_uuid0 = info->uuid[0];
544 sb->set_uuid1 = info->uuid[1];
545 sb->set_uuid2 = info->uuid[2];
546 sb->set_uuid3 = info->uuid[3];
547 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
548 struct bitmap_super_s *bm;
549 bm = (struct bitmap_super_s*)(sb+1);
550 uuid_from_super0(st, (int*)bm->uuid);
551 }
552 } else if (strcmp(update, "no-bitmap") == 0) {
553 sb->state &= ~(1<<MD_SB_BITMAP_PRESENT);
554 } else if (strcmp(update, "_reshape_progress")==0)
555 sb->reshape_position = info->reshape_progress;
556 else
557 rv = -1;
558
559 sb->sb_csum = calc_sb0_csum(sb);
560 return rv;
561 }
562
563 /*
564 * For verion-0 superblock, the homehost is 'stored' in the
565 * uuid. 8 bytes for a hash of the host leaving 8 bytes
566 * of random material.
567 * We use the first 8 bytes (64bits) of the sha1 of the
568 * host name
569 */
570
571
572 static int init_super0(struct supertype *st, mdu_array_info_t *info,
573 unsigned long long size, char *ignored_name, char *homehost,
574 int *uuid)
575 {
576 mdp_super_t *sb;
577 int spares;
578
579 if (posix_memalign((void**)&sb, 4096,
580 MD_SB_BYTES + ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
581 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
582 return 0;
583 }
584 memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
585
586 st->sb = sb;
587 if (info == NULL) {
588 /* zeroing the superblock */
589 return 0;
590 }
591
592 spares = info->working_disks - info->active_disks;
593 if (info->raid_disks + spares > MD_SB_DISKS) {
594 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
595 info->raid_disks , spares, MD_SB_DISKS);
596 return 0;
597 }
598
599 sb->md_magic = MD_SB_MAGIC;
600 sb->major_version = 0;
601 sb->minor_version = 90;
602 sb->patch_version = 0;
603 sb->gvalid_words = 0; /* ignored */
604 sb->ctime = time(0);
605 sb->level = info->level;
606 if (size != (unsigned long long)info->size)
607 return 0;
608 sb->size = info->size;
609 sb->nr_disks = info->nr_disks;
610 sb->raid_disks = info->raid_disks;
611 sb->md_minor = info->md_minor;
612 sb->not_persistent = 0;
613 if (uuid) {
614 sb->set_uuid0 = uuid[0];
615 sb->set_uuid1 = uuid[1];
616 sb->set_uuid2 = uuid[2];
617 sb->set_uuid3 = uuid[3];
618 } else {
619 int rfd = open("/dev/urandom", O_RDONLY);
620 if (rfd < 0 || read(rfd, &sb->set_uuid0, 4) != 4)
621 sb->set_uuid0 = random();
622 if (rfd < 0 || read(rfd, &sb->set_uuid1, 12) != 12) {
623 sb->set_uuid1 = random();
624 sb->set_uuid2 = random();
625 sb->set_uuid3 = random();
626 }
627 if (rfd >= 0)
628 close(rfd);
629 }
630 if (homehost) {
631 char buf[20];
632 char *hash = sha1_buffer(homehost,
633 strlen(homehost),
634 buf);
635 memcpy(&sb->set_uuid2, hash, 8);
636 }
637
638 sb->utime = sb->ctime;
639 sb->state = info->state;
640 sb->active_disks = info->active_disks;
641 sb->working_disks = info->working_disks;
642 sb->failed_disks = info->failed_disks;
643 sb->spare_disks = info->spare_disks;
644 sb->events_hi = 0;
645 sb->events_lo = 1;
646
647 sb->layout = info->layout;
648 sb->chunk_size = info->chunk_size;
649
650 return 1;
651 }
652
653 struct devinfo {
654 int fd;
655 char *devname;
656 mdu_disk_info_t disk;
657 struct devinfo *next;
658 };
659
660 #ifndef MDASSEMBLE
661 /* Add a device to the superblock being created */
662 static int add_to_super0(struct supertype *st, mdu_disk_info_t *dinfo,
663 int fd, char *devname)
664 {
665 mdp_super_t *sb = st->sb;
666 mdp_disk_t *dk = &sb->disks[dinfo->number];
667 struct devinfo *di, **dip;
668
669 dk->number = dinfo->number;
670 dk->major = dinfo->major;
671 dk->minor = dinfo->minor;
672 dk->raid_disk = dinfo->raid_disk;
673 dk->state = dinfo->state;
674
675 sb->this_disk = sb->disks[dinfo->number];
676 sb->sb_csum = calc_sb0_csum(sb);
677
678 dip = (struct devinfo **)&st->info;
679 while (*dip)
680 dip = &(*dip)->next;
681 di = malloc(sizeof(struct devinfo));
682 di->fd = fd;
683 di->devname = devname;
684 di->disk = *dinfo;
685 di->next = NULL;
686 *dip = di;
687
688 return 0;
689 }
690 #endif
691
692 static int store_super0(struct supertype *st, int fd)
693 {
694 unsigned long long dsize;
695 unsigned long long offset;
696 mdp_super_t *super = st->sb;
697
698 if (!get_dev_size(fd, NULL, &dsize))
699 return 1;
700
701 if (dsize < MD_RESERVED_SECTORS*512)
702 return 2;
703
704 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
705
706 offset *= 512;
707
708 if (lseek64(fd, offset, 0)< 0LL)
709 return 3;
710
711 if (write(fd, super, sizeof(*super)) != sizeof(*super))
712 return 4;
713
714 if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
715 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
716 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
717 if (write(fd, bm, ROUND_UP(sizeof(*bm),4096)) !=
718 ROUND_UP(sizeof(*bm),4096))
719 return 5;
720 }
721
722 fsync(fd);
723 return 0;
724 }
725
726 #ifndef MDASSEMBLE
727 static int write_init_super0(struct supertype *st)
728 {
729 mdp_super_t *sb = st->sb;
730 int rv = 0;
731 struct devinfo *di;
732
733 for (di = st->info ; di && ! rv ; di = di->next) {
734
735 if (di->disk.state == 1)
736 continue;
737 if (di->fd == -1)
738 continue;
739 while (Kill(di->devname, NULL, 0, 1, 1) == 0)
740 ;
741
742 sb->disks[di->disk.number].state &= ~(1<<MD_DISK_FAULTY);
743
744 sb->this_disk = sb->disks[di->disk.number];
745 sb->sb_csum = calc_sb0_csum(sb);
746 rv = store_super0(st, di->fd);
747
748 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
749 rv = st->ss->write_bitmap(st, di->fd);
750
751 if (rv)
752 fprintf(stderr,
753 Name ": failed to write superblock to %s\n",
754 di->devname);
755 }
756 return rv;
757 }
758 #endif
759
760 static int compare_super0(struct supertype *st, struct supertype *tst)
761 {
762 /*
763 * return:
764 * 0 same, or first was empty, and second was copied
765 * 1 second had wrong number
766 * 2 wrong uuid
767 * 3 wrong other info
768 */
769 mdp_super_t *first = st->sb;
770 mdp_super_t *second = tst->sb;
771 int uuid1[4], uuid2[4];
772
773 if (second->md_magic != MD_SB_MAGIC)
774 return 1;
775 if (!first) {
776 if (posix_memalign((void**)&first, 4096,
777 MD_SB_BYTES +
778 ROUND_UP(sizeof(struct bitmap_super_s), 4096)) != 0) {
779 fprintf(stderr, Name
780 ": %s could not allocate superblock\n", __func__);
781 return 1;
782 }
783 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
784 st->sb = first;
785 return 0;
786 }
787
788 uuid_from_super0(st, uuid1);
789 uuid_from_super0(tst, uuid2);
790 if (!same_uuid(uuid1, uuid2, 0))
791 return 2;
792 if (first->major_version != second->major_version ||
793 first->minor_version != second->minor_version ||
794 first->patch_version != second->patch_version ||
795 first->gvalid_words != second->gvalid_words ||
796 first->ctime != second->ctime ||
797 first->level != second->level ||
798 first->size != second->size ||
799 first->raid_disks != second->raid_disks )
800 return 3;
801
802 return 0;
803 }
804
805
806 static void free_super0(struct supertype *st);
807
808 static int load_super0(struct supertype *st, int fd, char *devname)
809 {
810 /* try to read in the superblock
811 * Return:
812 * 0 on success
813 * 1 on cannot get superblock
814 * 2 on superblock meaningless
815 */
816 unsigned long long dsize;
817 unsigned long long offset;
818 mdp_super_t *super;
819 int uuid[4];
820 struct bitmap_super_s *bsb;
821
822 free_super0(st);
823
824 if (!get_dev_size(fd, devname, &dsize))
825 return 1;
826
827 if (dsize < MD_RESERVED_SECTORS*512) {
828 if (devname)
829 fprintf(stderr, Name
830 ": %s is too small for md: size is %llu sectors.\n",
831 devname, dsize);
832 return 1;
833 }
834
835 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
836
837 offset *= 512;
838
839 ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
840
841 if (lseek64(fd, offset, 0)< 0LL) {
842 if (devname)
843 fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
844 devname, strerror(errno));
845 return 1;
846 }
847
848 if (posix_memalign((void**)&super, 4096,
849 MD_SB_BYTES +
850 ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
851 fprintf(stderr, Name
852 ": %s could not allocate superblock\n", __func__);
853 return 1;
854 }
855
856 if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
857 if (devname)
858 fprintf(stderr, Name ": Cannot read superblock on %s\n",
859 devname);
860 free(super);
861 return 1;
862 }
863
864 if (st->ss && st->minor_version == 9)
865 super0_swap_endian(super);
866
867 if (super->md_magic != MD_SB_MAGIC) {
868 if (devname)
869 fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
870 devname, MD_SB_MAGIC, super->md_magic);
871 free(super);
872 return 2;
873 }
874
875 if (super->major_version != 0) {
876 if (devname)
877 fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
878 devname, super->major_version);
879 free(super);
880 return 2;
881 }
882 st->sb = super;
883
884 if (st->ss == NULL) {
885 st->ss = &super0;
886 st->minor_version = super->minor_version;
887 st->max_devs = MD_SB_DISKS;
888 st->info = NULL;
889 }
890
891 /* Now check on the bitmap superblock */
892 if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
893 return 0;
894 /* Read the bitmap superblock and make sure it looks
895 * valid. If it doesn't clear the bit. An --assemble --force
896 * should get that written out.
897 */
898 if (read(fd, super+1, ROUND_UP(sizeof(struct bitmap_super_s),4096))
899 != ROUND_UP(sizeof(struct bitmap_super_s),4096))
900 goto no_bitmap;
901
902 uuid_from_super0(st, uuid);
903 bsb = (struct bitmap_super_s *)(super+1);
904 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
905 memcmp(bsb->uuid, uuid, 16) != 0)
906 goto no_bitmap;
907 return 0;
908
909 no_bitmap:
910 super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
911
912 return 0;
913 }
914
915 static struct supertype *match_metadata_desc0(char *arg)
916 {
917 struct supertype *st = malloc(sizeof(*st));
918 if (!st) return st;
919
920 memset(st, 0, sizeof(*st));
921 st->container_dev = NoMdDev;
922 st->ss = &super0;
923 st->info = NULL;
924 st->minor_version = 90;
925 st->max_devs = MD_SB_DISKS;
926 st->sb = NULL;
927 /* we sometimes get 00.90 */
928 while (arg[0] == '0' && arg[1] == '0')
929 arg++;
930 if (strcmp(arg, "0") == 0 ||
931 #ifdef DEFAULT_OLD_METADATA /* ifndef in super1.c */
932 strcmp(arg, "default") == 0 ||
933 #endif /* DEFAULT_OLD_METADATA */
934 strcmp(arg, "0.90") == 0 ||
935 strcmp(arg, "") == 0 /* no metadata - i.e. non_persistent */
936 )
937 return st;
938
939 st->minor_version = 91; /* reshape in progress */
940 if (strcmp(arg, "0.91") == 0) /* For dup_super support */
941 return st;
942
943 st->minor_version = 9; /* flag for 'byte-swapped' */
944 if (strcmp(arg, "0.swap")==0 ||
945 strcmp(arg, "0.9") == 0) /* For dup_super support */
946 return st;
947
948 free(st);
949 return NULL;
950 }
951
952 static __u64 avail_size0(struct supertype *st, __u64 devsize)
953 {
954 if (devsize < MD_RESERVED_SECTORS)
955 return 0ULL;
956 return MD_NEW_SIZE_SECTORS(devsize);
957 }
958
959 static int add_internal_bitmap0(struct supertype *st, int *chunkp,
960 int delay, int write_behind,
961 unsigned long long size, int may_change,
962 int major)
963 {
964 /*
965 * The bitmap comes immediately after the superblock and must be 60K in size
966 * at most. The default size is between 30K and 60K
967 *
968 * size is in sectors, chunk is in bytes !!!
969 */
970 unsigned long long bits;
971 unsigned long long max_bits = (60*1024 - sizeof(bitmap_super_t))*8;
972 unsigned long long min_chunk;
973 int chunk = *chunkp;
974 mdp_super_t *sb = st->sb;
975 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
976
977
978 min_chunk = 4096; /* sub-page chunks don't work yet.. */
979 bits = (size * 512) / min_chunk + 1;
980 while (bits > max_bits) {
981 min_chunk *= 2;
982 bits = (bits+1)/2;
983 }
984 if (chunk == UnSet) {
985 /* A chunk size less than a few Megabytes gives poor
986 * performance without increasing resync noticeably
987 */
988 chunk = min_chunk;
989 if (chunk < 64*1024*1024)
990 chunk = 64*1024*1024;
991 } else if ((unsigned long long)chunk < min_chunk)
992 return 0; /* chunk size too small */
993
994 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
995
996 memset(bms, 0, sizeof(*bms));
997 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
998 bms->version = __cpu_to_le32(major);
999 uuid_from_super0(st, (int*)bms->uuid);
1000 bms->chunksize = __cpu_to_le32(chunk);
1001 bms->daemon_sleep = __cpu_to_le32(delay);
1002 bms->sync_size = __cpu_to_le64(size);
1003 bms->write_behind = __cpu_to_le32(write_behind);
1004 *chunkp = chunk;
1005 return 1;
1006 }
1007
1008
1009 static void locate_bitmap0(struct supertype *st, int fd)
1010 {
1011 unsigned long long dsize;
1012 unsigned long long offset;
1013
1014 if (!get_dev_size(fd, NULL, &dsize))
1015 return;
1016
1017 if (dsize < MD_RESERVED_SECTORS*512)
1018 return;
1019
1020 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1021
1022 offset *= 512;
1023
1024 offset += MD_SB_BYTES;
1025
1026 lseek64(fd, offset, 0);
1027 }
1028
1029 static int write_bitmap0(struct supertype *st, int fd)
1030 {
1031 unsigned long long dsize;
1032 unsigned long long offset;
1033 mdp_super_t *sb = st->sb;
1034
1035 int rv = 0;
1036
1037 int towrite, n;
1038 char abuf[4096+4096];
1039 char *buf = (char*)(((long)(abuf+4096))&~4095L);
1040
1041 if (!get_dev_size(fd, NULL, &dsize))
1042 return 1;
1043
1044
1045 if (dsize < MD_RESERVED_SECTORS*512)
1046 return -1;
1047
1048 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1049
1050 offset *= 512;
1051
1052 if (lseek64(fd, offset + 4096, 0)< 0LL)
1053 return 3;
1054
1055 memset(buf, 0xff, 4096);
1056 memcpy(buf, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
1057 towrite = 60*1024;
1058 while (towrite > 0) {
1059 n = towrite;
1060 if (n > 4096)
1061 n = 4096;
1062 n = write(fd, buf, n);
1063 if (n > 0)
1064 towrite -= n;
1065 else
1066 break;
1067 memset(buf, 0xff, 4096);
1068 }
1069 fsync(fd);
1070 if (towrite)
1071 rv = -2;
1072
1073 return rv;
1074 }
1075
1076 static void free_super0(struct supertype *st)
1077 {
1078 if (st->sb)
1079 free(st->sb);
1080 while (st->info) {
1081 struct devinfo *di = st->info;
1082 st->info = di->next;
1083 if (di->fd >= 0)
1084 close(di->fd);
1085 free(di);
1086 }
1087 st->sb = NULL;
1088 }
1089
1090 #ifndef MDASSEMBLE
1091 static int validate_geometry0(struct supertype *st, int level,
1092 int layout, int raiddisks,
1093 int chunk, unsigned long long size,
1094 char *subdev, unsigned long long *freesize,
1095 int verbose)
1096 {
1097 unsigned long long ldsize;
1098 int fd;
1099
1100 if (level == LEVEL_CONTAINER) {
1101 if (verbose)
1102 fprintf(stderr, Name ": 0.90 metadata does not support containers\n");
1103 return 0;
1104 }
1105 if (raiddisks > MD_SB_DISKS) {
1106 if (verbose)
1107 fprintf(stderr, Name ": 0.90 metadata supports at most %d devices per array\n",
1108 MD_SB_DISKS);
1109 return 0;
1110 }
1111 if (size > (0x7fffffffULL<<9)) {
1112 if (verbose)
1113 fprintf(stderr, Name ": 0.90 metadata supports at most 2 terrabytes per device\n");
1114 return 0;
1115 }
1116 if (!subdev)
1117 return 1;
1118
1119 fd = open(subdev, O_RDONLY|O_EXCL, 0);
1120 if (fd < 0) {
1121 if (verbose)
1122 fprintf(stderr, Name ": super0.90 cannot open %s: %s\n",
1123 subdev, strerror(errno));
1124 return 0;
1125 }
1126
1127 if (!get_dev_size(fd, subdev, &ldsize)) {
1128 close(fd);
1129 return 0;
1130 }
1131 close(fd);
1132
1133 if (ldsize < MD_RESERVED_SECTORS * 512)
1134 return 0;
1135 if (size > (0x7fffffffULL<<9))
1136 return 0;
1137 *freesize = MD_NEW_SIZE_SECTORS(ldsize >> 9);
1138 return 1;
1139 }
1140 #endif /* MDASSEMBLE */
1141
1142 struct superswitch super0 = {
1143 #ifndef MDASSEMBLE
1144 .examine_super = examine_super0,
1145 .brief_examine_super = brief_examine_super0,
1146 .export_examine_super = export_examine_super0,
1147 .detail_super = detail_super0,
1148 .brief_detail_super = brief_detail_super0,
1149 .write_init_super = write_init_super0,
1150 .validate_geometry = validate_geometry0,
1151 .add_to_super = add_to_super0,
1152 #endif
1153 .match_home = match_home0,
1154 .uuid_from_super = uuid_from_super0,
1155 .getinfo_super = getinfo_super0,
1156 .container_content = container_content0,
1157 .update_super = update_super0,
1158 .init_super = init_super0,
1159 .store_super = store_super0,
1160 .compare_super = compare_super0,
1161 .load_super = load_super0,
1162 .match_metadata_desc = match_metadata_desc0,
1163 .avail_size = avail_size0,
1164 .add_internal_bitmap = add_internal_bitmap0,
1165 .locate_bitmap = locate_bitmap0,
1166 .write_bitmap = write_bitmap0,
1167 .free_super = free_super0,
1168 .name = "0.90",
1169 };