]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super0.c
projects / thirdparty / mdadm.git / blob
? search:


3f02ed403435bcaf4c5196e931539627c7ba8495
[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)
144 printf(" (%d->%d)\n", sb->raid_disks-sb->delta_disks, sb->raid_disks);
145 else
146 printf(" (%d->%d)\n", sb->raid_disks, sb->raid_disks+sb->delta_disks);
147 }
148 if (sb->new_level != sb->level) {
149 c = map_num(pers, sb->new_level);
150 printf(" New Level : %s\n", c?c:"-unknown-");
151 }
152 if (sb->new_layout != sb->layout) {
153 if (sb->level == 5) {
154 c = map_num(r5layout, sb->new_layout);
155 printf(" New Layout : %s\n", c?c:"-unknown-");
156 }
157 if (sb->level == 10) {
158 printf(" New Layout : near=%d, %s=%d\n",
159 sb->new_layout&255,
160 (sb->new_layout&0x10000)?"offset":"far",
161 (sb->new_layout>>8)&255);
162 }
163 }
164 if (sb->new_chunk != sb->chunk_size)
165 printf(" New Chunksize : %d\n", sb->new_chunk);
166 printf("\n");
167 }
168 atime = sb->utime;
169 printf(" Update Time : %.24s\n", ctime(&atime));
170 printf(" State : %s\n",
171 (sb->state&(1<<MD_SB_CLEAN))?"clean":"active");
172 if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
173 printf("Internal Bitmap : present\n");
174 printf(" Active Devices : %d\n", sb->active_disks);
175 printf("Working Devices : %d\n", sb->working_disks);
176 printf(" Failed Devices : %d\n", sb->failed_disks);
177 printf(" Spare Devices : %d\n", sb->spare_disks);
178 if (calc_sb0_csum(sb) == sb->sb_csum)
179 printf(" Checksum : %x - correct\n", sb->sb_csum);
180 else
181 printf(" Checksum : %x - expected %lx\n", sb->sb_csum, calc_sb0_csum(sb));
182 printf(" Events : %llu\n",
183 ((unsigned long long)sb->events_hi << 32)
184 + sb->events_lo);
185 printf("\n");
186 if (sb->level == 5) {
187 c = map_num(r5layout, sb->layout);
188 printf(" Layout : %s\n", c?c:"-unknown-");
189 }
190 if (sb->level == 10) {
191 printf(" Layout :");
192 print_r10_layout(sb->layout);
193 printf("\n");
194 }
195 switch(sb->level) {
196 case 0:
197 case 4:
198 case 5:
199 case 6:
200 case 10:
201 printf(" Chunk Size : %dK\n", sb->chunk_size/1024);
202 break;
203 case -1:
204 printf(" Rounding : %dK\n", sb->chunk_size/1024);
205 break;
206 default: break;
207 }
208 printf("\n");
209 printf(" Number Major Minor RaidDevice State\n");
210 for (d= -1; d<(signed int)(sb->raid_disks+sb->spare_disks); d++) {
211 mdp_disk_t *dp;
212 char *dv;
213 char nb[5];
214 int wonly;
215 if (d>=0) dp = &sb->disks[d];
216 else dp = &sb->this_disk;
217 snprintf(nb, sizeof(nb), "%4d", d);
218 printf("%4s %5d %5d %5d %5d ", d < 0 ? "this" : nb,
219 dp->number, dp->major, dp->minor, dp->raid_disk);
220 wonly = dp->state & (1<<MD_DISK_WRITEMOSTLY);
221 dp->state &= ~(1<<MD_DISK_WRITEMOSTLY);
222 if (dp->state & (1<<MD_DISK_FAULTY)) printf(" faulty");
223 if (dp->state & (1<<MD_DISK_ACTIVE)) printf(" active");
224 if (dp->state & (1<<MD_DISK_SYNC)) printf(" sync");
225 if (dp->state & (1<<MD_DISK_REMOVED)) printf(" removed");
226 if (wonly) printf(" write-mostly");
227 if (dp->state == 0) printf(" spare");
228 if ((dv=map_dev(dp->major, dp->minor, 0)))
229 printf(" %s", dv);
230 printf("\n");
231 if (d == -1) printf("\n");
232 }
233 }
234
235 static void brief_examine_super0(struct supertype *st, int verbose)
236 {
237 mdp_super_t *sb = st->sb;
238 char *c=map_num(pers, sb->level);
239 char devname[20];
240
241 sprintf(devname, "/dev/md%d", sb->md_minor);
242
243 if (verbose) {
244 printf("ARRAY %s level=%s num-devices=%d",
245 devname,
246 c?c:"-unknown-", sb->raid_disks);
247 } else
248 printf("ARRAY %s", devname);
249
250 if (sb->minor_version >= 90)
251 printf(" UUID=%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
252 sb->set_uuid2, sb->set_uuid3);
253 else
254 printf(" UUID=%08x", sb->set_uuid0);
255 printf("\n");
256 }
257
258 static void export_examine_super0(struct supertype *st)
259 {
260 mdp_super_t *sb = st->sb;
261
262 printf("MD_LEVEL=%s\n", map_num(pers, sb->level));
263 printf("MD_DEVICES=%d\n", sb->raid_disks);
264 if (sb->minor_version >= 90)
265 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
266 sb->set_uuid0, sb->set_uuid1,
267 sb->set_uuid2, sb->set_uuid3);
268 else
269 printf("MD_UUID=%08x\n", sb->set_uuid0);
270 printf("MD_UPDATE_TIME=%llu\n",
271 __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL);
272 printf("MD_EVENTS=%llu\n",
273 ((unsigned long long)sb->events_hi << 32)
274 + sb->events_lo);
275 }
276
277 static void detail_super0(struct supertype *st, char *homehost)
278 {
279 mdp_super_t *sb = st->sb;
280 printf(" UUID : ");
281 if (sb->minor_version >= 90)
282 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
283 sb->set_uuid2, sb->set_uuid3);
284 else
285 printf("%08x", sb->set_uuid0);
286 if (homehost) {
287 char buf[20];
288 void *hash = sha1_buffer(homehost,
289 strlen(homehost),
290 buf);
291 if (memcmp(&sb->set_uuid2, hash, 8)==0)
292 printf(" (local to host %s)", homehost);
293 }
294 printf("\n Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
295 }
296
297 static void brief_detail_super0(struct supertype *st)
298 {
299 mdp_super_t *sb = st->sb;
300 printf(" UUID=");
301 if (sb->minor_version >= 90)
302 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
303 sb->set_uuid2, sb->set_uuid3);
304 else
305 printf("%08x", sb->set_uuid0);
306 }
307 #endif
308
309 static int match_home0(struct supertype *st, char *homehost)
310 {
311 mdp_super_t *sb = st->sb;
312 char buf[20];
313 char *hash = sha1_buffer(homehost,
314 strlen(homehost),
315 buf);
316
317 return (memcmp(&sb->set_uuid2, hash, 8)==0);
318 }
319
320 static void uuid_from_super0(struct supertype *st, int uuid[4])
321 {
322 mdp_super_t *super = st->sb;
323 uuid[0] = super->set_uuid0;
324 if (super->minor_version >= 90) {
325 uuid[1] = super->set_uuid1;
326 uuid[2] = super->set_uuid2;
327 uuid[3] = super->set_uuid3;
328 } else {
329 uuid[1] = 0;
330 uuid[2] = 0;
331 uuid[3] = 0;
332 }
333 }
334
335 static void getinfo_super0(struct supertype *st, struct mdinfo *info)
336 {
337 mdp_super_t *sb = st->sb;
338 int working = 0;
339 int i;
340
341 info->array.major_version = sb->major_version;
342 info->array.minor_version = sb->minor_version;
343 info->array.patch_version = sb->patch_version;
344 info->array.raid_disks = sb->raid_disks;
345 info->array.level = sb->level;
346 info->array.layout = sb->layout;
347 info->array.md_minor = sb->md_minor;
348 info->array.ctime = sb->ctime;
349 info->array.utime = sb->utime;
350 info->array.chunk_size = sb->chunk_size;
351 info->array.state = sb->state;
352 info->component_size = sb->size*2;
353
354 info->disk.state = sb->this_disk.state;
355 info->disk.major = sb->this_disk.major;
356 info->disk.minor = sb->this_disk.minor;
357 info->disk.raid_disk = sb->this_disk.raid_disk;
358 info->disk.number = sb->this_disk.number;
359
360 info->events = md_event(sb);
361 info->data_offset = 0;
362
363 sprintf(info->text_version, "0.%d", sb->minor_version);
364 info->safe_mode_delay = 200;
365
366 uuid_from_super0(st, info->uuid);
367
368 if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
369 info->reshape_active = 1;
370 info->reshape_progress = sb->reshape_position;
371 info->new_level = sb->new_level;
372 info->delta_disks = sb->delta_disks;
373 info->new_layout = sb->new_layout;
374 info->new_chunk = sb->new_chunk;
375 } else
376 info->reshape_active = 0;
377
378 sprintf(info->name, "%d", sb->md_minor);
379 /* work_disks is calculated rather than read directly */
380 for (i=0; i < MD_SB_DISKS; i++)
381 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
382 (sb->disks[i].raid_disk < info->array.raid_disks) &&
383 (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
384 !(sb->disks[i].state & (1<<MD_DISK_FAULTY)))
385 working ++;
386 info->array.working_disks = working;
387 }
388
389
390 static int update_super0(struct supertype *st, struct mdinfo *info,
391 char *update,
392 char *devname, int verbose,
393 int uuid_set, char *homehost)
394 {
395 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
396 * For others, the return value is ignored.
397 */
398 int rv = 0;
399 mdp_super_t *sb = st->sb;
400 if (strcmp(update, "sparc2.2")==0 ) {
401 /* 2.2 sparc put the events in the wrong place
402 * So we copy the tail of the superblock
403 * up 4 bytes before continuing
404 */
405 __u32 *sb32 = (__u32*)sb;
406 memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
407 sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
408 (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
409 if (verbose >= 0)
410 fprintf (stderr, Name ": adjusting superblock of %s for 2.2/sparc compatability.\n",
411 devname);
412 }
413 if (strcmp(update, "super-minor") ==0) {
414 sb->md_minor = info->array.md_minor;
415 if (verbose > 0)
416 fprintf(stderr, Name ": updating superblock of %s with minor number %d\n",
417 devname, info->array.md_minor);
418 }
419 if (strcmp(update, "summaries") == 0) {
420 int i;
421 /* set nr_disks, active_disks, working_disks,
422 * failed_disks, spare_disks based on disks[]
423 * array in superblock.
424 * Also make sure extra slots aren't 'failed'
425 */
426 sb->nr_disks = sb->active_disks =
427 sb->working_disks = sb->failed_disks =
428 sb->spare_disks = 0;
429 for (i=0; i < MD_SB_DISKS ; i++)
430 if (sb->disks[i].major ||
431 sb->disks[i].minor) {
432 int state = sb->disks[i].state;
433 if (state & (1<<MD_DISK_REMOVED))
434 continue;
435 sb->nr_disks++;
436 if (state & (1<<MD_DISK_ACTIVE))
437 sb->active_disks++;
438 if (state & (1<<MD_DISK_FAULTY))
439 sb->failed_disks++;
440 else
441 sb->working_disks++;
442 if (state == 0)
443 sb->spare_disks++;
444 } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
445 sb->disks[i].state = 0;
446 }
447 if (strcmp(update, "force-one")==0) {
448 /* Not enough devices for a working array, so
449 * bring this one up-to-date.
450 */
451 __u32 ehi = sb->events_hi, elo = sb->events_lo;
452 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
453 sb->events_lo = (info->events) & 0xFFFFFFFF;
454 if (sb->events_hi != ehi ||
455 sb->events_lo != elo)
456 rv = 1;
457 }
458 if (strcmp(update, "force-array")==0) {
459 /* degraded array and 'force' requested, so
460 * maybe need to mark it 'clean'
461 */
462 if ((sb->level == 5 || sb->level == 4 || sb->level == 6) &&
463 (sb->state & (1 << MD_SB_CLEAN)) == 0) {
464 /* need to force clean */
465 sb->state |= (1 << MD_SB_CLEAN);
466 rv = 1;
467 }
468 }
469 if (strcmp(update, "assemble")==0) {
470 int d = info->disk.number;
471 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
472 if ((sb->disks[d].state & ~(1<<MD_DISK_WRITEMOSTLY))
473 != info->disk.state) {
474 sb->disks[d].state = info->disk.state | wonly;
475 rv = 1;
476 }
477 }
478 if (strcmp(update, "linear-grow-new") == 0) {
479 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
480 sb->disks[info->disk.number].number = info->disk.number;
481 sb->disks[info->disk.number].major = info->disk.major;
482 sb->disks[info->disk.number].minor = info->disk.minor;
483 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
484 sb->disks[info->disk.number].state = info->disk.state;
485 sb->this_disk = sb->disks[info->disk.number];
486 }
487 if (strcmp(update, "linear-grow-update") == 0) {
488 sb->raid_disks = info->array.raid_disks;
489 sb->nr_disks = info->array.nr_disks;
490 sb->active_disks = info->array.active_disks;
491 sb->working_disks = info->array.working_disks;
492 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
493 sb->disks[info->disk.number].number = info->disk.number;
494 sb->disks[info->disk.number].major = info->disk.major;
495 sb->disks[info->disk.number].minor = info->disk.minor;
496 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
497 sb->disks[info->disk.number].state = info->disk.state;
498 }
499 if (strcmp(update, "resync") == 0) {
500 /* make sure resync happens */
501 sb->state &= ~(1<<MD_SB_CLEAN);
502 sb->recovery_cp = 0;
503 }
504 if (strcmp(update, "homehost") == 0 &&
505 homehost) {
506 uuid_set = 0;
507 update = "uuid";
508 info->uuid[0] = sb->set_uuid0;
509 info->uuid[1] = sb->set_uuid1;
510 }
511 if (strcmp(update, "uuid") == 0) {
512 if (!uuid_set && homehost) {
513 char buf[20];
514 char *hash = sha1_buffer(homehost,
515 strlen(homehost),
516 buf);
517 memcpy(info->uuid+2, hash, 8);
518 }
519 sb->set_uuid0 = info->uuid[0];
520 sb->set_uuid1 = info->uuid[1];
521 sb->set_uuid2 = info->uuid[2];
522 sb->set_uuid3 = info->uuid[3];
523 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
524 struct bitmap_super_s *bm;
525 bm = (struct bitmap_super_s*)(sb+1);
526 uuid_from_super0(st, (int*)bm->uuid);
527 }
528 }
529 if (strcmp(update, "_reshape_progress")==0)
530 sb->reshape_position = info->reshape_progress;
531
532 sb->sb_csum = calc_sb0_csum(sb);
533 return rv;
534 }
535
536 /*
537 * For verion-0 superblock, the homehost is 'stored' in the
538 * uuid. 8 bytes for a hash of the host leaving 8 bytes
539 * of random material.
540 * We use the first 8 bytes (64bits) of the sha1 of the
541 * host name
542 */
543
544
545 static int init_super0(struct supertype *st, mdu_array_info_t *info,
546 unsigned long long size, char *ignored_name, char *homehost,
547 int *uuid)
548 {
549 mdp_super_t *sb;
550 int spares;
551
552 if (posix_memalign((void**)&sb, 4096,
553 MD_SB_BYTES + ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
554 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
555 return 0;
556 }
557 memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
558
559 st->sb = sb;
560 if (info == NULL) {
561 /* zeroing the superblock */
562 return 0;
563 }
564
565 spares = info->working_disks - info->active_disks;
566 if (info->raid_disks + spares > MD_SB_DISKS) {
567 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
568 info->raid_disks , spares, MD_SB_DISKS);
569 return 0;
570 }
571
572 sb->md_magic = MD_SB_MAGIC;
573 sb->major_version = 0;
574 sb->minor_version = 90;
575 sb->patch_version = 0;
576 sb->gvalid_words = 0; /* ignored */
577 sb->ctime = time(0);
578 sb->level = info->level;
579 if (size != info->size)
580 return 0;
581 sb->size = info->size;
582 sb->nr_disks = info->nr_disks;
583 sb->raid_disks = info->raid_disks;
584 sb->md_minor = info->md_minor;
585 sb->not_persistent = 0;
586 if (uuid) {
587 sb->set_uuid0 = uuid[0];
588 sb->set_uuid1 = uuid[1];
589 sb->set_uuid2 = uuid[2];
590 sb->set_uuid3 = uuid[3];
591 } else {
592 int rfd = open("/dev/urandom", O_RDONLY);
593 if (rfd < 0 || read(rfd, &sb->set_uuid0, 4) != 4)
594 sb->set_uuid0 = random();
595 if (rfd < 0 || read(rfd, &sb->set_uuid1, 12) != 12) {
596 sb->set_uuid1 = random();
597 sb->set_uuid2 = random();
598 sb->set_uuid3 = random();
599 }
600 if (rfd >= 0)
601 close(rfd);
602 }
603 if (homehost) {
604 char buf[20];
605 char *hash = sha1_buffer(homehost,
606 strlen(homehost),
607 buf);
608 memcpy(&sb->set_uuid2, hash, 8);
609 }
610
611 sb->utime = sb->ctime;
612 sb->state = info->state;
613 sb->active_disks = info->active_disks;
614 sb->working_disks = info->working_disks;
615 sb->failed_disks = info->failed_disks;
616 sb->spare_disks = info->spare_disks;
617 sb->events_hi = 0;
618 sb->events_lo = 1;
619
620 sb->layout = info->layout;
621 sb->chunk_size = info->chunk_size;
622
623 return 1;
624 }
625
626 struct devinfo {
627 int fd;
628 char *devname;
629 mdu_disk_info_t disk;
630 struct devinfo *next;
631 };
632
633 #ifndef MDASSEMBLE
634 /* Add a device to the superblock being created */
635 static int add_to_super0(struct supertype *st, mdu_disk_info_t *dinfo,
636 int fd, char *devname)
637 {
638 mdp_super_t *sb = st->sb;
639 mdp_disk_t *dk = &sb->disks[dinfo->number];
640 struct devinfo *di, **dip;
641
642 dk->number = dinfo->number;
643 dk->major = dinfo->major;
644 dk->minor = dinfo->minor;
645 dk->raid_disk = dinfo->raid_disk;
646 dk->state = dinfo->state;
647
648 sb->this_disk = sb->disks[dinfo->number];
649 sb->sb_csum = calc_sb0_csum(sb);
650
651 dip = (struct devinfo **)&st->info;
652 while (*dip)
653 dip = &(*dip)->next;
654 di = malloc(sizeof(struct devinfo));
655 di->fd = fd;
656 di->devname = devname;
657 di->disk = *dinfo;
658 di->next = NULL;
659 *dip = di;
660
661 return 0;
662 }
663 #endif
664
665 static int store_super0(struct supertype *st, int fd)
666 {
667 unsigned long long dsize;
668 unsigned long long offset;
669 mdp_super_t *super = st->sb;
670
671 if (!get_dev_size(fd, NULL, &dsize))
672 return 1;
673
674 if (dsize < MD_RESERVED_SECTORS*512)
675 return 2;
676
677 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
678
679 offset *= 512;
680
681 if (lseek64(fd, offset, 0)< 0LL)
682 return 3;
683
684 if (write(fd, super, sizeof(*super)) != sizeof(*super))
685 return 4;
686
687 if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
688 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
689 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
690 if (write(fd, bm, ROUND_UP(sizeof(*bm),4096)) !=
691 ROUND_UP(sizeof(*bm),4096))
692 return 5;
693 }
694
695 fsync(fd);
696 return 0;
697 }
698
699 #ifndef MDASSEMBLE
700 static int write_init_super0(struct supertype *st)
701 {
702 mdp_super_t *sb = st->sb;
703 int rv = 0;
704 struct devinfo *di;
705
706 for (di = st->info ; di && ! rv ; di = di->next) {
707
708 if (di->disk.state == 1)
709 continue;
710 if (di->fd == -1)
711 continue;
712 Kill(di->devname, 0, 1, 1);
713 Kill(di->devname, 0, 1, 1);
714
715 sb->disks[di->disk.number].state &= ~(1<<MD_DISK_FAULTY);
716
717 sb->this_disk = sb->disks[di->disk.number];
718 sb->sb_csum = calc_sb0_csum(sb);
719 rv = store_super0(st, di->fd);
720
721 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
722 rv = st->ss->write_bitmap(st, di->fd);
723
724 if (rv)
725 fprintf(stderr,
726 Name ": failed to write superblock to %s\n",
727 di->devname);
728 close(di->fd);
729 di->fd = -1;
730 }
731 return rv;
732 }
733 #endif
734
735 static int compare_super0(struct supertype *st, struct supertype *tst)
736 {
737 /*
738 * return:
739 * 0 same, or first was empty, and second was copied
740 * 1 second had wrong number
741 * 2 wrong uuid
742 * 3 wrong other info
743 */
744 mdp_super_t *first = st->sb;
745 mdp_super_t *second = tst->sb;
746 int uuid1[4], uuid2[4];
747
748 if (second->md_magic != MD_SB_MAGIC)
749 return 1;
750 if (!first) {
751 if (posix_memalign((void**)&first, 4096,
752 MD_SB_BYTES +
753 ROUND_UP(sizeof(struct bitmap_super_s), 4096)) != 0) {
754 fprintf(stderr, Name
755 ": %s could not allocate superblock\n", __func__);
756 return 1;
757 }
758 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
759 st->sb = first;
760 return 0;
761 }
762
763 uuid_from_super0(st, uuid1);
764 uuid_from_super0(tst, uuid2);
765 if (!same_uuid(uuid1, uuid2, 0))
766 return 2;
767 if (first->major_version != second->major_version ||
768 first->minor_version != second->minor_version ||
769 first->patch_version != second->patch_version ||
770 first->gvalid_words != second->gvalid_words ||
771 first->ctime != second->ctime ||
772 first->level != second->level ||
773 first->size != second->size ||
774 first->raid_disks != second->raid_disks )
775 return 3;
776
777 return 0;
778 }
779
780
781 static void free_super0(struct supertype *st);
782
783 static int load_super0(struct supertype *st, int fd, char *devname)
784 {
785 /* try to read in the superblock
786 * Return:
787 * 0 on success
788 * 1 on cannot get superblock
789 * 2 on superblock meaningless
790 */
791 unsigned long long dsize;
792 unsigned long long offset;
793 mdp_super_t *super;
794 int uuid[4];
795 struct bitmap_super_s *bsb;
796
797 free_super0(st);
798
799 if (st->subarray[0])
800 return 1;
801
802 if (!get_dev_size(fd, devname, &dsize))
803 return 1;
804
805 if (dsize < MD_RESERVED_SECTORS*512) {
806 if (devname)
807 fprintf(stderr, Name
808 ": %s is too small for md: size is %llu sectors.\n",
809 devname, dsize);
810 return 1;
811 }
812
813 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
814
815 offset *= 512;
816
817 ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
818
819 if (lseek64(fd, offset, 0)< 0LL) {
820 if (devname)
821 fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
822 devname, strerror(errno));
823 return 1;
824 }
825
826 if (posix_memalign((void**)&super, 4096,
827 MD_SB_BYTES +
828 ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
829 fprintf(stderr, Name
830 ": %s could not allocate superblock\n", __func__);
831 return 1;
832 }
833
834 if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
835 if (devname)
836 fprintf(stderr, Name ": Cannot read superblock on %s\n",
837 devname);
838 free(super);
839 return 1;
840 }
841
842 if (st->ss && st->minor_version == 9)
843 super0_swap_endian(super);
844
845 if (super->md_magic != MD_SB_MAGIC) {
846 if (devname)
847 fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
848 devname, MD_SB_MAGIC, super->md_magic);
849 free(super);
850 return 2;
851 }
852
853 if (super->major_version != 0) {
854 if (devname)
855 fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
856 devname, super->major_version);
857 free(super);
858 return 2;
859 }
860 st->sb = super;
861
862 if (st->ss == NULL) {
863 st->ss = &super0;
864 st->minor_version = super->minor_version;
865 st->max_devs = MD_SB_DISKS;
866 st->info = NULL;
867 }
868
869 /* Now check on the bitmap superblock */
870 if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
871 return 0;
872 /* Read the bitmap superblock and make sure it looks
873 * valid. If it doesn't clear the bit. An --assemble --force
874 * should get that written out.
875 */
876 if (read(fd, super+1, ROUND_UP(sizeof(struct bitmap_super_s),4096))
877 != ROUND_UP(sizeof(struct bitmap_super_s),4096))
878 goto no_bitmap;
879
880 uuid_from_super0(st, uuid);
881 bsb = (struct bitmap_super_s *)(super+1);
882 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
883 memcmp(bsb->uuid, uuid, 16) != 0)
884 goto no_bitmap;
885 return 0;
886
887 no_bitmap:
888 super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
889
890 return 0;
891 }
892
893 static struct supertype *match_metadata_desc0(char *arg)
894 {
895 struct supertype *st = malloc(sizeof(*st));
896 if (!st) return st;
897
898 memset(st, 0, sizeof(*st));
899 st->ss = &super0;
900 st->info = NULL;
901 st->minor_version = 90;
902 st->max_devs = MD_SB_DISKS;
903 st->sb = NULL;
904 /* we sometimes get 00.90 */
905 while (arg[0] == '0' && arg[1] == '0')
906 arg++;
907 if (strcmp(arg, "0") == 0 ||
908 strcmp(arg, "0.90") == 0 ||
909 strcmp(arg, "default") == 0 ||
910 strcmp(arg, "") == 0 /* no metadata */
911 )
912 return st;
913
914 st->minor_version = 91; /* reshape in progress */
915 if (strcmp(arg, "0.91") == 0) /* For dup_super support */
916 return st;
917
918 st->minor_version = 9; /* flag for 'byte-swapped' */
919 if (strcmp(arg, "0.swap")==0 ||
920 strcmp(arg, "0.9") == 0) /* For dup_super support */
921 return st;
922
923 free(st);
924 return NULL;
925 }
926
927 static __u64 avail_size0(struct supertype *st, __u64 devsize)
928 {
929 if (devsize < MD_RESERVED_SECTORS)
930 return 0ULL;
931 return MD_NEW_SIZE_SECTORS(devsize);
932 }
933
934 static int add_internal_bitmap0(struct supertype *st, int *chunkp,
935 int delay, int write_behind,
936 unsigned long long size, int may_change,
937 int major)
938 {
939 /*
940 * The bitmap comes immediately after the superblock and must be 60K in size
941 * at most. The default size is between 30K and 60K
942 *
943 * size is in sectors, chunk is in bytes !!!
944 */
945 unsigned long long bits;
946 unsigned long long max_bits = 60*1024*8;
947 unsigned long long min_chunk;
948 int chunk = *chunkp;
949 mdp_super_t *sb = st->sb;
950 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
951
952
953 min_chunk = 4096; /* sub-page chunks don't work yet.. */
954 bits = (size * 512) / min_chunk + 1;
955 while (bits > max_bits) {
956 min_chunk *= 2;
957 bits = (bits+1)/2;
958 }
959 if (chunk == UnSet)
960 chunk = min_chunk;
961 else if (chunk < min_chunk)
962 return 0; /* chunk size too small */
963
964 sb->state |= (1<<MD_SB_BITMAP_PRESENT);
965
966 memset(bms, 0, sizeof(*bms));
967 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
968 bms->version = __cpu_to_le32(major);
969 uuid_from_super0(st, (int*)bms->uuid);
970 bms->chunksize = __cpu_to_le32(chunk);
971 bms->daemon_sleep = __cpu_to_le32(delay);
972 bms->sync_size = __cpu_to_le64(size);
973 bms->write_behind = __cpu_to_le32(write_behind);
974 *chunkp = chunk;
975 return 1;
976 }
977
978
979 static void locate_bitmap0(struct supertype *st, int fd)
980 {
981 unsigned long long dsize;
982 unsigned long long offset;
983
984 if (!get_dev_size(fd, NULL, &dsize))
985 return;
986
987 if (dsize < MD_RESERVED_SECTORS*512)
988 return;
989
990 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
991
992 offset *= 512;
993
994 offset += MD_SB_BYTES;
995
996 lseek64(fd, offset, 0);
997 }
998
999 static int write_bitmap0(struct supertype *st, int fd)
1000 {
1001 unsigned long long dsize;
1002 unsigned long long offset;
1003 mdp_super_t *sb = st->sb;
1004
1005 int rv = 0;
1006
1007 int towrite, n;
1008 char abuf[4096+4096];
1009 char *buf = (char*)(((long)(abuf+4096))&~4095L);
1010
1011 if (!get_dev_size(fd, NULL, &dsize))
1012 return 1;
1013
1014
1015 if (dsize < MD_RESERVED_SECTORS*512)
1016 return -1;
1017
1018 offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1019
1020 offset *= 512;
1021
1022 if (lseek64(fd, offset + 4096, 0)< 0LL)
1023 return 3;
1024
1025 memset(buf, 0xff, 4096);
1026 memcpy(buf, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
1027 towrite = 64*1024;
1028 while (towrite > 0) {
1029 n = towrite;
1030 if (n > 4096)
1031 n = 4096;
1032 n = write(fd, buf, n);
1033 if (n > 0)
1034 towrite -= n;
1035 else
1036 break;
1037 memset(buf, 0xff, 4096);
1038 }
1039 fsync(fd);
1040 if (towrite)
1041 rv = -2;
1042
1043 return rv;
1044 }
1045
1046 static void free_super0(struct supertype *st)
1047 {
1048 if (st->sb)
1049 free(st->sb);
1050 st->sb = NULL;
1051 }
1052
1053 #ifndef MDASSEMBLE
1054 static int validate_geometry0(struct supertype *st, int level,
1055 int layout, int raiddisks,
1056 int chunk, unsigned long long size,
1057 char *subdev, unsigned long long *freesize,
1058 int verbose)
1059 {
1060 unsigned long long ldsize;
1061 int fd;
1062
1063 if (level == LEVEL_CONTAINER)
1064 return 0;
1065 if (raiddisks > MD_SB_DISKS)
1066 return 0;
1067 if (size > (0x7fffffffULL<<9))
1068 return 0;
1069 if (!subdev)
1070 return 1;
1071
1072 fd = open(subdev, O_RDONLY|O_EXCL, 0);
1073 if (fd < 0) {
1074 if (verbose)
1075 fprintf(stderr, Name ": super0.90 cannot open %s: %s\n",
1076 subdev, strerror(errno));
1077 return 0;
1078 }
1079
1080 if (!get_dev_size(fd, subdev, &ldsize)) {
1081 close(fd);
1082 return 0;
1083 }
1084 close(fd);
1085
1086 if (ldsize < MD_RESERVED_SECTORS * 512)
1087 return 0;
1088 if (size > (0x7fffffffULL<<9))
1089 return 0;
1090 *freesize = MD_NEW_SIZE_SECTORS(ldsize >> 9);
1091 return 1;
1092 }
1093 #endif /* MDASSEMBLE */
1094
1095 struct superswitch super0 = {
1096 #ifndef MDASSEMBLE
1097 .examine_super = examine_super0,
1098 .brief_examine_super = brief_examine_super0,
1099 .export_examine_super = export_examine_super0,
1100 .detail_super = detail_super0,
1101 .brief_detail_super = brief_detail_super0,
1102 .write_init_super = write_init_super0,
1103 .validate_geometry = validate_geometry0,
1104 .add_to_super = add_to_super0,
1105 #endif
1106 .match_home = match_home0,
1107 .uuid_from_super = uuid_from_super0,
1108 .getinfo_super = getinfo_super0,
1109 .update_super = update_super0,
1110 .init_super = init_super0,
1111 .store_super = store_super0,
1112 .compare_super = compare_super0,
1113 .load_super = load_super0,
1114 .match_metadata_desc = match_metadata_desc0,
1115 .avail_size = avail_size0,
1116 .add_internal_bitmap = add_internal_bitmap0,
1117 .locate_bitmap = locate_bitmap0,
1118 .write_bitmap = write_bitmap0,
1119 .free_super = free_super0,
1120 .name = "0.90",
1121 };
Unnamed repository; edit this file 'description' to name the repository.