]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super0.c
Discard devnum in favour of devnm
[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, ddsks_denom = 1;
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;
126 ddsks_denom = (sb->layout&255) * ((sb->layout>>8)&255);
127 }
128 if (ddsks) {
129 long long asize = sb->size;
130 asize = (asize << 10) * ddsks / ddsks_denom;
131 printf(" Array Size : %llu%s\n",
132 asize >> 10, human_size(asize));
133 }
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 printf(" (%d->%d)\n", sb->raid_disks-sb->delta_disks, sb->raid_disks);
144 if (((int)sb->delta_disks) < 0)
145 delta_extra = - sb->delta_disks;
146 }
147 if (sb->new_level != sb->level) {
148 c = map_num(pers, sb->new_level);
149 printf(" New Level : %s\n", c?c:"-unknown-");
150 }
151 if (sb->new_layout != sb->layout) {
152 if (sb->level == 5) {
153 c = map_num(r5layout, sb->new_layout);
154 printf(" New Layout : %s\n", c?c:"-unknown-");
155 }
156 if (sb->level == 6) {
157 c = map_num(r6layout, sb->new_layout);
158 printf(" New Layout : %s\n", c?c:"-unknown-");
159 }
160 if (sb->level == 10) {
161 printf(" New Layout : near=%d, %s=%d\n",
162 sb->new_layout&255,
163 (sb->new_layout&0x10000)?"offset":"far",
164 (sb->new_layout>>8)&255);
165 }
166 }
167 if (sb->new_chunk != sb->chunk_size)
168 printf(" New Chunksize : %d\n", sb->new_chunk);
169 printf("\n");
170 }
171 atime = sb->utime;
172 printf(" Update Time : %.24s\n", ctime(&atime));
173 printf(" State : %s\n",
174 (sb->state&(1<<MD_SB_CLEAN))?"clean":"active");
175 if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
176 printf("Internal Bitmap : present\n");
177 printf(" Active Devices : %d\n", sb->active_disks);
178 printf("Working Devices : %d\n", sb->working_disks);
179 printf(" Failed Devices : %d\n", sb->failed_disks);
180 printf(" Spare Devices : %d\n", sb->spare_disks);
181 if (calc_sb0_csum(sb) == sb->sb_csum)
182 printf(" Checksum : %x - correct\n", sb->sb_csum);
183 else
184 printf(" Checksum : %x - expected %lx\n", sb->sb_csum, calc_sb0_csum(sb));
185 printf(" Events : %llu\n",
186 ((unsigned long long)sb->events_hi << 32)
187 + sb->events_lo);
188 printf("\n");
189 if (sb->level == 5) {
190 c = map_num(r5layout, sb->layout);
191 printf(" Layout : %s\n", c?c:"-unknown-");
192 }
193 if (sb->level == 6) {
194 c = map_num(r6layout, sb->layout);
195 printf(" Layout : %s\n", c?c:"-unknown-");
196 }
197 if (sb->level == 10) {
198 printf(" Layout :");
199 print_r10_layout(sb->layout);
200 printf("\n");
201 }
202 switch(sb->level) {
203 case 0:
204 case 4:
205 case 5:
206 case 6:
207 case 10:
208 printf(" Chunk Size : %dK\n", sb->chunk_size/1024);
209 break;
210 case -1:
211 printf(" Rounding : %dK\n", sb->chunk_size/1024);
212 break;
213 default: break;
214 }
215 printf("\n");
216 printf(" Number Major Minor RaidDevice State\n");
217 for (d= -1; d<(signed int)(sb->raid_disks+delta_extra + sb->spare_disks); d++) {
218 mdp_disk_t *dp;
219 char *dv;
220 char nb[5];
221 int wonly;
222 if (d>=0) dp = &sb->disks[d];
223 else dp = &sb->this_disk;
224 snprintf(nb, sizeof(nb), "%4d", d);
225 printf("%4s %5d %5d %5d %5d ", d < 0 ? "this" : nb,
226 dp->number, dp->major, dp->minor, dp->raid_disk);
227 wonly = dp->state & (1<<MD_DISK_WRITEMOSTLY);
228 dp->state &= ~(1<<MD_DISK_WRITEMOSTLY);
229 if (dp->state & (1<<MD_DISK_FAULTY)) printf(" faulty");
230 if (dp->state & (1<<MD_DISK_ACTIVE)) printf(" active");
231 if (dp->state & (1<<MD_DISK_SYNC)) printf(" sync");
232 if (dp->state & (1<<MD_DISK_REMOVED)) printf(" removed");
233 if (wonly) printf(" write-mostly");
234 if (dp->state == 0) printf(" spare");
235 if ((dv=map_dev(dp->major, dp->minor, 0)))
236 printf(" %s", dv);
237 printf("\n");
238 if (d == -1) printf("\n");
239 }
240 }
241
242 static void brief_examine_super0(struct supertype *st, int verbose)
243 {
244 mdp_super_t *sb = st->sb;
245 char *c=map_num(pers, sb->level);
246 char devname[20];
247
248 sprintf(devname, "/dev/md%d", sb->md_minor);
249
250 if (verbose) {
251 printf("ARRAY %s level=%s num-devices=%d",
252 devname,
253 c?c:"-unknown-", sb->raid_disks);
254 } else
255 printf("ARRAY %s", devname);
256
257 if (sb->minor_version >= 90)
258 printf(" UUID=%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
259 sb->set_uuid2, sb->set_uuid3);
260 else
261 printf(" UUID=%08x", sb->set_uuid0);
262 printf("\n");
263 }
264
265 static void export_examine_super0(struct supertype *st)
266 {
267 mdp_super_t *sb = st->sb;
268
269 printf("MD_LEVEL=%s\n", map_num(pers, sb->level));
270 printf("MD_DEVICES=%d\n", sb->raid_disks);
271 if (sb->minor_version >= 90)
272 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
273 sb->set_uuid0, sb->set_uuid1,
274 sb->set_uuid2, sb->set_uuid3);
275 else
276 printf("MD_UUID=%08x\n", sb->set_uuid0);
277 printf("MD_UPDATE_TIME=%llu\n",
278 __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL);
279 printf("MD_EVENTS=%llu\n",
280 ((unsigned long long)sb->events_hi << 32)
281 + sb->events_lo);
282 }
283
284 static void detail_super0(struct supertype *st, char *homehost)
285 {
286 mdp_super_t *sb = st->sb;
287 printf(" UUID : ");
288 if (sb->minor_version >= 90)
289 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
290 sb->set_uuid2, sb->set_uuid3);
291 else
292 printf("%08x", sb->set_uuid0);
293 if (homehost) {
294 char buf[20];
295 void *hash = sha1_buffer(homehost,
296 strlen(homehost),
297 buf);
298 if (memcmp(&sb->set_uuid2, hash, 8)==0)
299 printf(" (local to host %s)", homehost);
300 }
301 printf("\n Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
302 }
303
304 static void brief_detail_super0(struct supertype *st)
305 {
306 mdp_super_t *sb = st->sb;
307 printf(" UUID=");
308 if (sb->minor_version >= 90)
309 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
310 sb->set_uuid2, sb->set_uuid3);
311 else
312 printf("%08x", sb->set_uuid0);
313 }
314 #endif
315
316 static int match_home0(struct supertype *st, char *homehost)
317 {
318 mdp_super_t *sb = st->sb;
319 char buf[20];
320 char *hash;
321
322 if (!homehost)
323 return 0;
324 hash = sha1_buffer(homehost,
325 strlen(homehost),
326 buf);
327
328 return (memcmp(&sb->set_uuid2, hash, 8)==0);
329 }
330
331 static void uuid_from_super0(struct supertype *st, int uuid[4])
332 {
333 mdp_super_t *super = st->sb;
334 uuid[0] = super->set_uuid0;
335 if (super->minor_version >= 90) {
336 uuid[1] = super->set_uuid1;
337 uuid[2] = super->set_uuid2;
338 uuid[3] = super->set_uuid3;
339 } else {
340 uuid[1] = 0;
341 uuid[2] = 0;
342 uuid[3] = 0;
343 }
344 }
345
346 static void getinfo_super0(struct supertype *st, struct mdinfo *info, char *map)
347 {
348 mdp_super_t *sb = st->sb;
349 int working = 0;
350 int i;
351 int map_disks = info->array.raid_disks;
352
353 memset(info, 0, sizeof(*info));
354 info->array.major_version = sb->major_version;
355 info->array.minor_version = sb->minor_version;
356 info->array.patch_version = sb->patch_version;
357 info->array.raid_disks = sb->raid_disks;
358 info->array.level = sb->level;
359 info->array.layout = sb->layout;
360 info->array.md_minor = sb->md_minor;
361 info->array.ctime = sb->ctime;
362 info->array.utime = sb->utime;
363 info->array.chunk_size = sb->chunk_size;
364 info->array.state = sb->state;
365 info->component_size = sb->size*2;
366
367 if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
368 info->bitmap_offset = 8;
369
370 info->disk.state = sb->this_disk.state;
371 info->disk.major = sb->this_disk.major;
372 info->disk.minor = sb->this_disk.minor;
373 info->disk.raid_disk = sb->this_disk.raid_disk;
374 info->disk.number = sb->this_disk.number;
375
376 info->events = md_event(sb);
377 info->data_offset = 0;
378
379 sprintf(info->text_version, "0.%d", sb->minor_version);
380 info->safe_mode_delay = 200;
381
382 uuid_from_super0(st, info->uuid);
383
384 info->recovery_start = MaxSector;
385 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
386 info->reshape_active = 1;
387 info->reshape_progress = sb->reshape_position;
388 info->new_level = sb->new_level;
389 info->delta_disks = sb->delta_disks;
390 info->new_layout = sb->new_layout;
391 info->new_chunk = sb->new_chunk;
392 if (info->delta_disks < 0)
393 info->array.raid_disks -= info->delta_disks;
394 } else
395 info->reshape_active = 0;
396
397 info->recovery_blocked = info->reshape_active;
398
399 sprintf(info->name, "%d", sb->md_minor);
400 /* work_disks is calculated rather than read directly */
401 for (i=0; i < MD_SB_DISKS; i++)
402 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
403 (sb->disks[i].raid_disk < (unsigned)info->array.raid_disks) &&
404 (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
405 !(sb->disks[i].state & (1<<MD_DISK_FAULTY))) {
406 working ++;
407 if (map && i < map_disks)
408 map[i] = 1;
409 } else if (map && i < map_disks)
410 map[i] = 0;
411 info->array.working_disks = working;
412 }
413
414 static struct mdinfo *container_content0(struct supertype *st, char *subarray)
415 {
416 struct mdinfo *info;
417
418 if (subarray)
419 return NULL;
420
421 info = xmalloc(sizeof(*info));
422 getinfo_super0(st, info, NULL);
423 return info;
424 }
425
426 static int update_super0(struct supertype *st, struct mdinfo *info,
427 char *update,
428 char *devname, int verbose,
429 int uuid_set, char *homehost)
430 {
431 /* NOTE: for 'assemble' and 'force' we need to return non-zero
432 * if any change was made. For others, the return value is
433 * ignored.
434 */
435 int rv = 0;
436 int uuid[4];
437 mdp_super_t *sb = st->sb;
438
439 if (strcmp(update, "homehost") == 0 &&
440 homehost) {
441 /* note that 'homehost' is special as it is really
442 * a "uuid" update.
443 */
444 uuid_set = 0;
445 update = "uuid";
446 info->uuid[0] = sb->set_uuid0;
447 info->uuid[1] = sb->set_uuid1;
448 }
449
450 if (strcmp(update, "sparc2.2")==0 ) {
451 /* 2.2 sparc put the events in the wrong place
452 * So we copy the tail of the superblock
453 * up 4 bytes before continuing
454 */
455 __u32 *sb32 = (__u32*)sb;
456 memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
457 sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
458 (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
459 if (verbose >= 0)
460 pr_err("adjusting superblock of %s for 2.2/sparc compatibility.\n",
461 devname);
462 } else if (strcmp(update, "super-minor") ==0) {
463 sb->md_minor = info->array.md_minor;
464 if (verbose > 0)
465 pr_err("updating superblock of %s with minor number %d\n",
466 devname, info->array.md_minor);
467 } else if (strcmp(update, "summaries") == 0) {
468 unsigned int i;
469 /* set nr_disks, active_disks, working_disks,
470 * failed_disks, spare_disks based on disks[]
471 * array in superblock.
472 * Also make sure extra slots aren't 'failed'
473 */
474 sb->nr_disks = sb->active_disks =
475 sb->working_disks = sb->failed_disks =
476 sb->spare_disks = 0;
477 for (i=0; i < MD_SB_DISKS ; i++)
478 if (sb->disks[i].major ||
479 sb->disks[i].minor) {
480 int state = sb->disks[i].state;
481 if (state & (1<<MD_DISK_REMOVED))
482 continue;
483 sb->nr_disks++;
484 if (state & (1<<MD_DISK_ACTIVE))
485 sb->active_disks++;
486 if (state & (1<<MD_DISK_FAULTY))
487 sb->failed_disks++;
488 else
489 sb->working_disks++;
490 if (state == 0)
491 sb->spare_disks++;
492 } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
493 sb->disks[i].state = 0;
494 } else if (strcmp(update, "force-one")==0) {
495 /* Not enough devices for a working array, so
496 * bring this one up-to-date.
497 */
498 __u32 ehi = sb->events_hi, elo = sb->events_lo;
499 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
500 sb->events_lo = (info->events) & 0xFFFFFFFF;
501 if (sb->events_hi != ehi ||
502 sb->events_lo != elo)
503 rv = 1;
504 } else if (strcmp(update, "force-array")==0) {
505 /* degraded array and 'force' requested, so
506 * maybe need to mark it 'clean'
507 */
508 if ((sb->level == 5 || sb->level == 4 || sb->level == 6) &&
509 (sb->state & (1 << MD_SB_CLEAN)) == 0) {
510 /* need to force clean */
511 sb->state |= (1 << MD_SB_CLEAN);
512 rv = 1;
513 }
514 } else if (strcmp(update, "assemble")==0) {
515 int d = info->disk.number;
516 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
517 int mask = (1<<MD_DISK_WRITEMOSTLY);
518 int add = 0;
519 if (sb->minor_version >= 91)
520 /* During reshape we don't insist on everything
521 * being marked 'sync'
522 */
523 add = (1<<MD_DISK_SYNC);
524 if (((sb->disks[d].state & ~mask) | add)
525 != (unsigned)info->disk.state) {
526 sb->disks[d].state = info->disk.state | wonly;
527 rv = 1;
528 }
529 if (info->reshape_active &&
530 sb->minor_version > 90 && (sb->reshape_position+1) != 0 &&
531 info->delta_disks >= 0 &&
532 info->reshape_progress < sb->reshape_position) {
533 sb->reshape_position = info->reshape_progress;
534 rv = 1;
535 }
536 if (info->reshape_active &&
537 sb->minor_version > 90 && (sb->reshape_position+1) != 0 &&
538 info->delta_disks < 0 &&
539 info->reshape_progress > sb->reshape_position) {
540 sb->reshape_position = info->reshape_progress;
541 rv = 1;
542 }
543 } else if (strcmp(update, "linear-grow-new") == 0) {
544 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
545 sb->disks[info->disk.number].number = info->disk.number;
546 sb->disks[info->disk.number].major = info->disk.major;
547 sb->disks[info->disk.number].minor = info->disk.minor;
548 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
549 sb->disks[info->disk.number].state = info->disk.state;
550 sb->this_disk = sb->disks[info->disk.number];
551 } else if (strcmp(update, "linear-grow-update") == 0) {
552 sb->raid_disks = info->array.raid_disks;
553 sb->nr_disks = info->array.nr_disks;
554 sb->active_disks = info->array.active_disks;
555 sb->working_disks = info->array.working_disks;
556 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
557 sb->disks[info->disk.number].number = info->disk.number;
558 sb->disks[info->disk.number].major = info->disk.major;
559 sb->disks[info->disk.number].minor = info->disk.minor;
560 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
561 sb->disks[info->disk.number].state = info->disk.state;
562 } else if (strcmp(update, "resync") == 0) {
563 /* make sure resync happens */
564 sb->state &= ~(1<<MD_SB_CLEAN);
565 sb->recovery_cp = 0;
566 } else if (strcmp(update, "uuid") == 0) {
567 if (!uuid_set && homehost) {
568 char buf[20];
569 char *hash = sha1_buffer(homehost,
570 strlen(homehost),
571 buf);
572 memcpy(info->uuid+2, hash, 8);
573 }
574 sb->set_uuid0 = info->uuid[0];
575 sb->set_uuid1 = info->uuid[1];
576 sb->set_uuid2 = info->uuid[2];
577 sb->set_uuid3 = info->uuid[3];
578 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
579 struct bitmap_super_s *bm;
580 bm = (struct bitmap_super_s*)(sb+1);
581 uuid_from_super0(st, uuid);
582 memcpy(bm->uuid, uuid, 16);
583 }
584 } else if (strcmp(update, "no-bitmap") == 0) {
585 sb->state &= ~(1<<MD_SB_BITMAP_PRESENT);
586 } else if (strcmp(update, "_reshape_progress")==0)
587 sb->reshape_position = info->reshape_progress;
588 else if (strcmp(update, "writemostly")==0)
589 sb->state |= (1<<MD_DISK_WRITEMOSTLY);
590 else if (strcmp(update, "readwrite")==0)
591 sb->state &= ~(1<<MD_DISK_WRITEMOSTLY);
592 else
593 rv = -1;
594
595 sb->sb_csum = calc_sb0_csum(sb);
596 return rv;
597 }
598
599 /*
600 * For verion-0 superblock, the homehost is 'stored' in the
601 * uuid. 8 bytes for a hash of the host leaving 8 bytes
602 * of random material.
603 * We use the first 8 bytes (64bits) of the sha1 of the
604 * host name
605 */
606
607
608 static int init_super0(struct supertype *st, mdu_array_info_t *info,
609 unsigned long long size, char *ignored_name, char *homehost,
610 int *uuid, unsigned long long data_offset)
611 {
612 mdp_super_t *sb;
613 int spares;
614
615 if (data_offset != INVALID_SECTORS) {
616 fprintf(stderr, Name ": data-offset not support for 0.90\n");
617 return 0;
618 }
619
620 if (posix_memalign((void**)&sb, 4096,
621 MD_SB_BYTES + ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
622 pr_err("%s could not allocate superblock\n", __func__);
623 return 0;
624 }
625 memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
626
627 st->sb = sb;
628 if (info == NULL) {
629 /* zeroing the superblock */
630 return 0;
631 }
632
633 spares = info->working_disks - info->active_disks;
634 if (info->raid_disks + spares > MD_SB_DISKS) {
635 pr_err("too many devices requested: %d+%d > %d\n",
636 info->raid_disks , spares, MD_SB_DISKS);
637 return 0;
638 }
639
640 sb->md_magic = MD_SB_MAGIC;
641 sb->major_version = 0;
642 sb->minor_version = 90;
643 sb->patch_version = 0;
644 sb->gvalid_words = 0; /* ignored */
645 sb->ctime = time(0);
646 sb->level = info->level;
647 sb->size = size;
648 if (size != (unsigned long long)sb->size)
649 return 0;
650 sb->nr_disks = info->nr_disks;
651 sb->raid_disks = info->raid_disks;
652 sb->md_minor = info->md_minor;
653 sb->not_persistent = 0;
654 if (uuid) {
655 sb->set_uuid0 = uuid[0];
656 sb->set_uuid1 = uuid[1];
657 sb->set_uuid2 = uuid[2];
658 sb->set_uuid3 = uuid[3];
659 } else {
660 int rfd = open("/dev/urandom", O_RDONLY);
661 if (rfd < 0 || read(rfd, &sb->set_uuid0, 4) != 4)
662 sb->set_uuid0 = random();
663 if (rfd < 0 || read(rfd, &sb->set_uuid1, 12) != 12) {
664 sb->set_uuid1 = random();
665 sb->set_uuid2 = random();
666 sb->set_uuid3 = random();
667 }
668 if (rfd >= 0)
669 close(rfd);
670 }
671 if (homehost && !uuid) {
672 char buf[20];
673 char *hash = sha1_buffer(homehost,
674 strlen(homehost),
675 buf);
676 memcpy(&sb->set_uuid2, hash, 8);
677 }
678
679 sb->utime = sb->ctime;
680 sb->state = info->state;
681 sb->active_disks = info->active_disks;
682 sb->working_disks = info->working_disks;
683 sb->failed_disks = info->failed_disks;
684 sb->spare_disks = info->spare_disks;
685 sb->events_hi = 0;
686 sb->events_lo = 1;
687
688 sb->layout = info->layout;
689 sb->chunk_size = info->chunk_size;
690
691 return 1;
692 }
693
694 struct devinfo {
695 int fd;
696 char *devname;
697 mdu_disk_info_t disk;
698 struct devinfo *next;
699 };
700
701 #ifndef MDASSEMBLE
702 /* Add a device to the superblock being created */
703 static int add_to_super0(struct supertype *st, mdu_disk_info_t *dinfo,
704 int fd, char *devname, unsigned long long data_offset)
705 {
706 mdp_super_t *sb = st->sb;
707 mdp_disk_t *dk = &sb->disks[dinfo->number];
708 struct devinfo *di, **dip;
709
710 dk->number = dinfo->number;
711 dk->major = dinfo->major;
712 dk->minor = dinfo->minor;
713 dk->raid_disk = dinfo->raid_disk;
714 dk->state = dinfo->state;
715 /* In case our source disk was writemostly, don't copy that bit */
716 dk->state &= ~(1<<MD_DISK_WRITEMOSTLY);
717
718 sb->this_disk = sb->disks[dinfo->number];
719 sb->sb_csum = calc_sb0_csum(sb);
720
721 dip = (struct devinfo **)&st->info;
722 while (*dip)
723 dip = &(*dip)->next;
724 di = xmalloc(sizeof(struct devinfo));
725 di->fd = fd;
726 di->devname = devname;
727 di->disk = *dinfo;
728 di->next = NULL;
729 *dip = di;
730
731 return 0;
732 }
733 #endif
734
735 static int store_super0(struct supertype *st, int fd)
736 {
737 unsigned long long dsize;
738 unsigned long long offset;
739 mdp_super_t *super = st->sb;
740
741 if (!get_dev_size(fd, NULL, &dsize))
742 return 1;
743
744 if (dsize < MD_RESERVED_SECTORS*512)
745 return 2;
746
747 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
748
749 offset *= 512;
750
751 if (lseek64(fd, offset, 0)< 0LL)
752 return 3;
753
754 if (write(fd, super, sizeof(*super)) != sizeof(*super))
755 return 4;
756
757 if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
758 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
759 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
760 if (write(fd, bm, ROUND_UP(sizeof(*bm),4096)) !=
761 ROUND_UP(sizeof(*bm),4096))
762 return 5;
763 }
764
765 fsync(fd);
766 return 0;
767 }
768
769 #ifndef MDASSEMBLE
770 static int write_init_super0(struct supertype *st)
771 {
772 mdp_super_t *sb = st->sb;
773 int rv = 0;
774 struct devinfo *di;
775
776 for (di = st->info ; di && ! rv ; di = di->next) {
777
778 if (di->disk.state == 1)
779 continue;
780 if (di->fd == -1)
781 continue;
782 while (Kill(di->devname, NULL, 0, -1, 1) == 0)
783 ;
784
785 sb->disks[di->disk.number].state &= ~(1<<MD_DISK_FAULTY);
786
787 sb->this_disk = sb->disks[di->disk.number];
788 sb->sb_csum = calc_sb0_csum(sb);
789 rv = store_super0(st, di->fd);
790
791 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
792 rv = st->ss->write_bitmap(st, di->fd);
793
794 if (rv)
795 pr_err("failed to write superblock to %s\n",
796 di->devname);
797 }
798 return rv;
799 }
800 #endif
801
802 static int compare_super0(struct supertype *st, struct supertype *tst)
803 {
804 /*
805 * return:
806 * 0 same, or first was empty, and second was copied
807 * 1 second had wrong number
808 * 2 wrong uuid
809 * 3 wrong other info
810 */
811 mdp_super_t *first = st->sb;
812 mdp_super_t *second = tst->sb;
813 int uuid1[4], uuid2[4];
814
815 if (second->md_magic != MD_SB_MAGIC)
816 return 1;
817 if (!first) {
818 if (posix_memalign((void**)&first, 4096,
819 MD_SB_BYTES +
820 ROUND_UP(sizeof(struct bitmap_super_s), 4096)) != 0) {
821 pr_err("%s could not allocate superblock\n", __func__);
822 return 1;
823 }
824 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
825 st->sb = first;
826 return 0;
827 }
828
829 uuid_from_super0(st, uuid1);
830 uuid_from_super0(tst, uuid2);
831 if (!same_uuid(uuid1, uuid2, 0))
832 return 2;
833 if (first->major_version != second->major_version ||
834 first->minor_version != second->minor_version ||
835 first->patch_version != second->patch_version ||
836 first->gvalid_words != second->gvalid_words ||
837 first->ctime != second->ctime ||
838 first->level != second->level ||
839 first->size != second->size ||
840 first->raid_disks != second->raid_disks )
841 return 3;
842
843 return 0;
844 }
845
846
847 static void free_super0(struct supertype *st);
848
849 static int load_super0(struct supertype *st, int fd, char *devname)
850 {
851 /* try to read in the superblock
852 * Return:
853 * 0 on success
854 * 1 on cannot get superblock
855 * 2 on superblock meaningless
856 */
857 unsigned long long dsize;
858 unsigned long long offset;
859 mdp_super_t *super;
860 int uuid[4];
861 struct bitmap_super_s *bsb;
862
863 free_super0(st);
864
865 if (!get_dev_size(fd, devname, &dsize))
866 return 1;
867
868 if (dsize < MD_RESERVED_SECTORS*512) {
869 if (devname)
870 pr_err("%s is too small for md: size is %llu sectors.\n",
871 devname, dsize);
872 return 1;
873 }
874
875 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
876
877 offset *= 512;
878
879 if (lseek64(fd, offset, 0)< 0LL) {
880 if (devname)
881 pr_err("Cannot seek to superblock on %s: %s\n",
882 devname, strerror(errno));
883 return 1;
884 }
885
886 if (posix_memalign((void**)&super, 4096,
887 MD_SB_BYTES +
888 ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
889 pr_err("%s could not allocate superblock\n", __func__);
890 return 1;
891 }
892
893 if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
894 if (devname)
895 pr_err("Cannot read superblock on %s\n",
896 devname);
897 free(super);
898 return 1;
899 }
900
901 if (st->ss && st->minor_version == 9)
902 super0_swap_endian(super);
903
904 if (super->md_magic != MD_SB_MAGIC) {
905 if (devname)
906 pr_err("No super block found on %s (Expected magic %08x, got %08x)\n",
907 devname, MD_SB_MAGIC, super->md_magic);
908 free(super);
909 return 2;
910 }
911
912 if (super->major_version != 0) {
913 if (devname)
914 pr_err("Cannot interpret superblock on %s - version is %d\n",
915 devname, super->major_version);
916 free(super);
917 return 2;
918 }
919 st->sb = super;
920
921 if (st->ss == NULL) {
922 st->ss = &super0;
923 st->minor_version = super->minor_version;
924 st->max_devs = MD_SB_DISKS;
925 st->info = NULL;
926 }
927
928 /* Now check on the bitmap superblock */
929 if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
930 return 0;
931 /* Read the bitmap superblock and make sure it looks
932 * valid. If it doesn't clear the bit. An --assemble --force
933 * should get that written out.
934 */
935 if (read(fd, super+1, ROUND_UP(sizeof(struct bitmap_super_s),4096))
936 != ROUND_UP(sizeof(struct bitmap_super_s),4096))
937 goto no_bitmap;
938
939 uuid_from_super0(st, uuid);
940 bsb = (struct bitmap_super_s *)(super+1);
941 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
942 memcmp(bsb->uuid, uuid, 16) != 0)
943 goto no_bitmap;
944 return 0;
945
946 no_bitmap:
947 super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
948
949 return 0;
950 }
951
952 static struct supertype *match_metadata_desc0(char *arg)
953 {
954 struct supertype *st = xcalloc(1, sizeof(*st));
955
956 st->container_devnm[0] = 0;
957 st->ss = &super0;
958 st->info = NULL;
959 st->minor_version = 90;
960 st->max_devs = MD_SB_DISKS;
961 st->sb = NULL;
962 /* we sometimes get 00.90 */
963 while (arg[0] == '0' && arg[1] == '0')
964 arg++;
965 if (strcmp(arg, "0") == 0 ||
966 #ifdef DEFAULT_OLD_METADATA /* ifndef in super1.c */
967 strcmp(arg, "default") == 0 ||
968 #endif /* DEFAULT_OLD_METADATA */
969 strcmp(arg, "0.90") == 0 ||
970 strcmp(arg, "") == 0 /* no metadata - i.e. non_persistent */
971 )
972 return st;
973
974 st->minor_version = 91; /* reshape in progress */
975 if (strcmp(arg, "0.91") == 0) /* For dup_super support */
976 return st;
977
978 st->minor_version = 9; /* flag for 'byte-swapped' */
979 if (strcmp(arg, "0.swap")==0 ||
980 strcmp(arg, "0.9") == 0) /* For dup_super support */
981 return st;
982
983 free(st);
984 return NULL;
985 }
986
987 static __u64 avail_size0(struct supertype *st, __u64 devsize,
988 unsigned long long data_offset)
989 {
990 if (data_offset != 0 && data_offset != INVALID_SECTORS)
991 return 0ULL;
992 if (devsize < MD_RESERVED_SECTORS)
993 return 0ULL;
994 return MD_NEW_SIZE_SECTORS(devsize);
995 }
996
997 static int add_internal_bitmap0(struct supertype *st, int *chunkp,
998 int delay, int write_behind,
999 unsigned long long size, int may_change,
1000 int major)
1001 {
1002 /*
1003 * The bitmap comes immediately after the superblock and must be 60K in size
1004 * at most. The default size is between 30K and 60K
1005 *
1006 * size is in sectors, chunk is in bytes !!!
1007 */
1008 unsigned long long bits;
1009 unsigned long long max_bits = (60*1024 - sizeof(bitmap_super_t))*8;
1010 unsigned long long min_chunk;
1011 int chunk = *chunkp;
1012 mdp_super_t *sb = st->sb;
1013 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
1014 int uuid[4];
1015
1016
1017 min_chunk = 4096; /* sub-page chunks don't work yet.. */
1018 bits = (size * 512) / min_chunk + 1;
1019 while (bits > max_bits) {
1020 min_chunk *= 2;
1021 bits = (bits+1)/2;
1022 }
1023 if (chunk == UnSet) {
1024 /* A chunk size less than a few Megabytes gives poor
1025 * performance without increasing resync noticeably
1026 */
1027 chunk = min_chunk;
1028 if (chunk < 64*1024*1024)
1029 chunk = 64*1024*1024;
1030 } else if ((unsigned long long)chunk < min_chunk)
1031 return 0; /* chunk size too small */
1032
1033 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
1034
1035 memset(bms, 0, sizeof(*bms));
1036 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
1037 bms->version = __cpu_to_le32(major);
1038 uuid_from_super0(st, uuid);
1039 memcpy(bms->uuid, uuid, 16);
1040 bms->chunksize = __cpu_to_le32(chunk);
1041 bms->daemon_sleep = __cpu_to_le32(delay);
1042 bms->sync_size = __cpu_to_le64(size);
1043 bms->write_behind = __cpu_to_le32(write_behind);
1044 *chunkp = chunk;
1045 return 1;
1046 }
1047
1048
1049 static void locate_bitmap0(struct supertype *st, int fd)
1050 {
1051 unsigned long long dsize;
1052 unsigned long long offset;
1053
1054 if (!get_dev_size(fd, NULL, &dsize))
1055 return;
1056
1057 if (dsize < MD_RESERVED_SECTORS*512)
1058 return;
1059
1060 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1061
1062 offset *= 512;
1063
1064 offset += MD_SB_BYTES;
1065
1066 lseek64(fd, offset, 0);
1067 }
1068
1069 static int write_bitmap0(struct supertype *st, int fd)
1070 {
1071 unsigned long long dsize;
1072 unsigned long long offset;
1073 mdp_super_t *sb = st->sb;
1074
1075 int rv = 0;
1076
1077 int towrite, n;
1078 void *buf;
1079
1080 if (!get_dev_size(fd, NULL, &dsize))
1081 return 1;
1082
1083 if (dsize < MD_RESERVED_SECTORS*512)
1084 return -1;
1085
1086 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1087
1088 offset *= 512;
1089
1090 if (lseek64(fd, offset + 4096, 0)< 0LL)
1091 return 3;
1092
1093 if (posix_memalign(&buf, 4096, 4096))
1094 return -ENOMEM;
1095
1096 memset(buf, 0xff, 4096);
1097 memcpy(buf, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
1098 towrite = 60*1024;
1099 while (towrite > 0) {
1100 n = towrite;
1101 if (n > 4096)
1102 n = 4096;
1103 n = write(fd, buf, n);
1104 if (n > 0)
1105 towrite -= n;
1106 else
1107 break;
1108 memset(buf, 0xff, 4096);
1109 }
1110 fsync(fd);
1111 if (towrite)
1112 rv = -2;
1113
1114 free(buf);
1115 return rv;
1116 }
1117
1118 static void free_super0(struct supertype *st)
1119 {
1120 if (st->sb)
1121 free(st->sb);
1122 while (st->info) {
1123 struct devinfo *di = st->info;
1124 st->info = di->next;
1125 if (di->fd >= 0)
1126 close(di->fd);
1127 free(di);
1128 }
1129 st->sb = NULL;
1130 }
1131
1132 #ifndef MDASSEMBLE
1133 static int validate_geometry0(struct supertype *st, int level,
1134 int layout, int raiddisks,
1135 int *chunk, unsigned long long size,
1136 unsigned long long data_offset,
1137 char *subdev, unsigned long long *freesize,
1138 int verbose)
1139 {
1140 unsigned long long ldsize;
1141 int fd;
1142 unsigned int tbmax = 4;
1143
1144 /* prior to linux 3.1, a but limits usable device size to 2TB.
1145 * It was introduced in 2.6.29, but we won't worry about that detail
1146 */
1147 if (get_linux_version() < 3001000)
1148 tbmax = 2;
1149
1150 if (level == LEVEL_CONTAINER) {
1151 if (verbose)
1152 pr_err("0.90 metadata does not support containers\n");
1153 return 0;
1154 }
1155 if (raiddisks > MD_SB_DISKS) {
1156 if (verbose)
1157 pr_err("0.90 metadata supports at most %d devices per array\n",
1158 MD_SB_DISKS);
1159 return 0;
1160 }
1161 if (size >= tbmax * 2ULL*1024*1024*1024) {
1162 if (verbose)
1163 pr_err("0.90 metadata supports at most "
1164 "%d terabytes per device\n", tbmax);
1165 return 0;
1166 }
1167 if (chunk && *chunk == UnSet)
1168 *chunk = DEFAULT_CHUNK;
1169
1170 if (!subdev)
1171 return 1;
1172
1173 fd = open(subdev, O_RDONLY|O_EXCL, 0);
1174 if (fd < 0) {
1175 if (verbose)
1176 pr_err("super0.90 cannot open %s: %s\n",
1177 subdev, strerror(errno));
1178 return 0;
1179 }
1180
1181 if (!get_dev_size(fd, subdev, &ldsize)) {
1182 close(fd);
1183 return 0;
1184 }
1185 close(fd);
1186
1187 if (ldsize < MD_RESERVED_SECTORS * 512)
1188 return 0;
1189 *freesize = MD_NEW_SIZE_SECTORS(ldsize >> 9);
1190 return 1;
1191 }
1192 #endif /* MDASSEMBLE */
1193
1194 struct superswitch super0 = {
1195 #ifndef MDASSEMBLE
1196 .examine_super = examine_super0,
1197 .brief_examine_super = brief_examine_super0,
1198 .export_examine_super = export_examine_super0,
1199 .detail_super = detail_super0,
1200 .brief_detail_super = brief_detail_super0,
1201 .write_init_super = write_init_super0,
1202 .validate_geometry = validate_geometry0,
1203 .add_to_super = add_to_super0,
1204 #endif
1205 .match_home = match_home0,
1206 .uuid_from_super = uuid_from_super0,
1207 .getinfo_super = getinfo_super0,
1208 .container_content = container_content0,
1209 .update_super = update_super0,
1210 .init_super = init_super0,
1211 .store_super = store_super0,
1212 .compare_super = compare_super0,
1213 .load_super = load_super0,
1214 .match_metadata_desc = match_metadata_desc0,
1215 .avail_size = avail_size0,
1216 .add_internal_bitmap = add_internal_bitmap0,
1217 .locate_bitmap = locate_bitmap0,
1218 .write_bitmap = write_bitmap0,
1219 .free_super = free_super0,
1220 .name = "0.90",
1221 };