]> git.ipfire.org Git - thirdparty/mdadm.git/blob - super1.c
projects / thirdparty / mdadm.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
super1 - do metadata IO in sector_size units.
[thirdparty/mdadm.git] / super1.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 #include "mdadm.h"
31 /*
32 * The version-1 superblock :
33 * All numeric fields are little-endian.
34 *
35 * total size: 256 bytes plus 2 per device.
36 * 1K allows 384 devices.
37 */
38 struct mdp_superblock_1 {
39 /* constant array information - 128 bytes */
40 __u32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */
41 __u32 major_version; /* 1 */
42 __u32 feature_map; /* 0 for now */
43 __u32 pad0; /* always set to 0 when writing */
44
45 __u8 set_uuid[16]; /* user-space generated. */
46 char set_name[32]; /* set and interpreted by user-space */
47
48 __u64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/
49 __u32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */
50 __u32 layout; /* only for raid5 currently */
51 __u64 size; /* used size of component devices, in 512byte sectors */
52
53 __u32 chunksize; /* in 512byte sectors */
54 __u32 raid_disks;
55 __u32 bitmap_offset; /* sectors after start of superblock that bitmap starts
56 * NOTE: signed, so bitmap can be before superblock
57 * only meaningful of feature_map[0] is set.
58 */
59
60 /* These are only valid with feature bit '4' */
61 __u32 new_level; /* new level we are reshaping to */
62 __u64 reshape_position; /* next address in array-space for reshape */
63 __u32 delta_disks; /* change in number of raid_disks */
64 __u32 new_layout; /* new layout */
65 __u32 new_chunk; /* new chunk size (bytes) */
66 __u8 pad1[128-124]; /* set to 0 when written */
67
68 /* constant this-device information - 64 bytes */
69 __u64 data_offset; /* sector start of data, often 0 */
70 __u64 data_size; /* sectors in this device that can be used for data */
71 __u64 super_offset; /* sector start of this superblock */
72 __u64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */
73 __u32 dev_number; /* permanent identifier of this device - not role in raid */
74 __u32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */
75 __u8 device_uuid[16]; /* user-space setable, ignored by kernel */
76 __u8 devflags; /* per-device flags. Only one defined...*/
77 #define WriteMostly1 1 /* mask for writemostly flag in above */
78 __u8 pad2[64-57]; /* set to 0 when writing */
79
80 /* array state information - 64 bytes */
81 __u64 utime; /* 40 bits second, 24 btes microseconds */
82 __u64 events; /* incremented when superblock updated */
83 __u64 resync_offset; /* data before this offset (from data_offset) known to be in sync */
84 __u32 sb_csum; /* checksum upto devs[max_dev] */
85 __u32 max_dev; /* size of devs[] array to consider */
86 __u8 pad3[64-32]; /* set to 0 when writing */
87
88 /* device state information. Indexed by dev_number.
89 * 2 bytes per device
90 * Note there are no per-device state flags. State information is rolled
91 * into the 'roles' value. If a device is spare or faulty, then it doesn't
92 * have a meaningful role.
93 */
94 __u16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
95 };
96
97 struct misc_dev_info {
98 __u64 device_size;
99 };
100
101 /* feature_map bits */
102 #define MD_FEATURE_BITMAP_OFFSET 1
103 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
104 * must be honoured
105 */
106 #define MD_FEATURE_RESHAPE_ACTIVE 4
107
108 #define MD_FEATURE_ALL (1|2|4)
109
110 #ifndef offsetof
111 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
112 #endif
113 static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb)
114 {
115 unsigned int disk_csum, csum;
116 unsigned long long newcsum;
117 int size = sizeof(*sb) + __le32_to_cpu(sb->max_dev)*2;
118 unsigned int *isuper = (unsigned int*)sb;
119 int i;
120
121 /* make sure I can count... */
122 if (offsetof(struct mdp_superblock_1,data_offset) != 128 ||
123 offsetof(struct mdp_superblock_1, utime) != 192 ||
124 sizeof(struct mdp_superblock_1) != 256) {
125 fprintf(stderr, "WARNING - superblock isn't sized correctly\n");
126 }
127
128 disk_csum = sb->sb_csum;
129 sb->sb_csum = 0;
130 newcsum = 0;
131 for (i=0; size>=4; size -= 4 ) {
132 newcsum += __le32_to_cpu(*isuper);
133 isuper++;
134 }
135
136 if (size == 2)
137 newcsum += __le16_to_cpu(*(unsigned short*) isuper);
138
139 csum = (newcsum & 0xffffffff) + (newcsum >> 32);
140 sb->sb_csum = disk_csum;
141 return __cpu_to_le32(csum);
142 }
143
144 static char abuf[4096+4096];
145 static int aread(int fd, void *buf, int len)
146 {
147 /* aligned read.
148 * On devices with a 4K sector size, we need to read
149 * the full sector and copy relevant bits into
150 * the buffer
151 */
152 int bsize;
153 char *b;
154 int n;
155 if (ioctl(fd, BLKSSZGET, &bsize) != 0 ||
156 bsize <= len)
157 return read(fd, buf, len);
158 if (bsize > 4096)
159 return -1;
160 b = (char*)(((long)(abuf+4096))&~4095UL);
161
162 n = read(fd, b, bsize);
163 if (n <= 0)
164 return n;
165 lseek(fd, len - n, 1);
166 if (n > len)
167 n = len;
168 memcpy(buf, b, n);
169 return n;
170 }
171
172 static int awrite(int fd, void *buf, int len)
173 {
174 /* aligned write.
175 * On devices with a 4K sector size, we need to write
176 * the full sector. We pre-read if the sector is larger
177 * than the write.
178 * The address must be sector-aligned.
179 */
180 int bsize;
181 char *b;
182 int n;
183 if (ioctl(fd, BLKSSZGET, &bsize) != 0 ||
184 bsize <= len)
185 return write(fd, buf, len);
186 if (bsize > 4096)
187 return -1;
188 b = (char*)(((long)(abuf+4096))&~4095UL);
189
190 n = read(fd, b, bsize);
191 if (n <= 0)
192 return n;
193 lseek(fd, -n, 1);
194 memcpy(b, buf, len);
195 n = write(fd, b, bsize);
196 if (n <= 0)
197 return n;
198 lseek(fd, len - n, 1);
199 return len;
200 }
201
202 #ifndef MDASSEMBLE
203 static void examine_super1(struct supertype *st, char *homehost)
204 {
205 struct mdp_superblock_1 *sb = st->sb;
206 time_t atime;
207 int d;
208 int faulty;
209 int i;
210 char *c;
211 int l = homehost ? strlen(homehost) : 0;
212 int layout;
213 unsigned long long sb_offset;
214
215 printf(" Magic : %08x\n", __le32_to_cpu(sb->magic));
216 printf(" Version : 1");
217 sb_offset = __le64_to_cpu(sb->super_offset);
218 if (sb_offset <= 4)
219 printf(".1\n");
220 else if (sb_offset <= 8)
221 printf(".2\n");
222 else
223 printf(".0\n");
224 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb->feature_map));
225 printf(" Array UUID : ");
226 for (i=0; i<16; i++) {
227 if ((i&3)==0 && i != 0) printf(":");
228 printf("%02x", sb->set_uuid[i]);
229 }
230 printf("\n");
231 printf(" Name : %.32s", sb->set_name);
232 if (l > 0 && l < 32 &&
233 sb->set_name[l] == ':' &&
234 strncmp(sb->set_name, homehost, l) == 0)
235 printf(" (local to host %s)", homehost);
236 printf("\n");
237 atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL;
238 printf(" Creation Time : %.24s\n", ctime(&atime));
239 c=map_num(pers, __le32_to_cpu(sb->level));
240 printf(" Raid Level : %s\n", c?c:"-unknown-");
241 printf(" Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks));
242 printf("\n");
243 printf(" Avail Dev Size : %llu%s\n",
244 (unsigned long long)__le64_to_cpu(sb->data_size),
245 human_size(__le64_to_cpu(sb->data_size)<<9));
246 if (__le32_to_cpu(sb->level) >= 0) {
247 int ddsks=0;
248 switch(__le32_to_cpu(sb->level)) {
249 case 1: ddsks=1;break;
250 case 4:
251 case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break;
252 case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break;
253 case 10:
254 layout = __le32_to_cpu(sb->layout);
255 ddsks = __le32_to_cpu(sb->raid_disks)
256 / (layout&255) / ((layout>>8)&255);
257 }
258 if (ddsks)
259 printf(" Array Size : %llu%s\n",
260 ddsks*(unsigned long long)__le64_to_cpu(sb->size),
261 human_size(ddsks*__le64_to_cpu(sb->size)<<9));
262 if (sb->size != sb->data_size)
263 printf(" Used Dev Size : %llu%s\n",
264 (unsigned long long)__le64_to_cpu(sb->size),
265 human_size(__le64_to_cpu(sb->size)<<9));
266 }
267 if (sb->data_offset)
268 printf(" Data Offset : %llu sectors\n",
269 (unsigned long long)__le64_to_cpu(sb->data_offset));
270 printf(" Super Offset : %llu sectors\n",
271 (unsigned long long)__le64_to_cpu(sb->super_offset));
272 if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET)
273 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->recovery_offset));
274 printf(" State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean");
275 printf(" Device UUID : ");
276 for (i=0; i<16; i++) {
277 if ((i&3)==0 && i != 0) printf(":");
278 printf("%02x", sb->device_uuid[i]);
279 }
280 printf("\n");
281 printf("\n");
282 if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) {
283 printf("Internal Bitmap : %ld sectors from superblock\n",
284 (long)(int32_t)__le32_to_cpu(sb->bitmap_offset));
285 }
286 if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) {
287 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb->reshape_position)/2,
288 human_size(__le64_to_cpu(sb->reshape_position)<<9));
289 if (__le32_to_cpu(sb->delta_disks)) {
290 printf(" Delta Devices : %d", __le32_to_cpu(sb->delta_disks));
291 if (__le32_to_cpu(sb->delta_disks))
292 printf(" (%d->%d)\n",
293 __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks),
294 __le32_to_cpu(sb->raid_disks));
295 else
296 printf(" (%d->%d)\n", __le32_to_cpu(sb->raid_disks),
297 __le32_to_cpu(sb->raid_disks)+__le32_to_cpu(sb->delta_disks));
298 }
299 if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) {
300 c = map_num(pers, __le32_to_cpu(sb->new_level));
301 printf(" New Level : %s\n", c?c:"-unknown-");
302 }
303 if (__le32_to_cpu(sb->new_layout) != __le32_to_cpu(sb->layout)) {
304 if (__le32_to_cpu(sb->level) == 5) {
305 c = map_num(r5layout, __le32_to_cpu(sb->new_layout));
306 printf(" New Layout : %s\n", c?c:"-unknown-");
307 }
308 if (__le32_to_cpu(sb->level) == 10) {
309 printf(" New Layout :");
310 print_r10_layout(__le32_to_cpu(sb->new_layout));
311 printf("\n");
312 }
313 }
314 if (__le32_to_cpu(sb->new_chunk) != __le32_to_cpu(sb->chunksize))
315 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb->new_chunk)/2);
316 printf("\n");
317 }
318 if (sb->devflags) {
319 printf(" Flags :");
320 if (sb->devflags & WriteMostly1)
321 printf(" write-mostly");
322 printf("\n");
323 }
324
325 atime = __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL;
326 printf(" Update Time : %.24s\n", ctime(&atime));
327
328 if (calc_sb_1_csum(sb) == sb->sb_csum)
329 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum));
330 else
331 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb->sb_csum),
332 __le32_to_cpu(calc_sb_1_csum(sb)));
333 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb->events));
334 printf("\n");
335 if (__le32_to_cpu(sb->level) == 5) {
336 c = map_num(r5layout, __le32_to_cpu(sb->layout));
337 printf(" Layout : %s\n", c?c:"-unknown-");
338 }
339 if (__le32_to_cpu(sb->level) == 10) {
340 int lo = __le32_to_cpu(sb->layout);
341 printf(" Layout :");
342 print_r10_layout(lo);
343 printf("\n");
344 }
345 switch(__le32_to_cpu(sb->level)) {
346 case 0:
347 case 4:
348 case 5:
349 case 6:
350 case 10:
351 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize)/2);
352 break;
353 case -1:
354 printf(" Rounding : %dK\n", __le32_to_cpu(sb->chunksize)/2);
355 break;
356 default: break;
357 }
358 printf("\n");
359 printf(" Array Slot : %d (", __le32_to_cpu(sb->dev_number));
360 for (i= __le32_to_cpu(sb->max_dev); i> 0 ; i--)
361 if (__le16_to_cpu(sb->dev_roles[i-1]) != 0xffff)
362 break;
363 for (d=0; d < i; d++) {
364 int role = __le16_to_cpu(sb->dev_roles[d]);
365 if (d) printf(", ");
366 if (role == 0xffff) printf("empty");
367 else if(role == 0xfffe) printf("failed");
368 else printf("%d", role);
369 }
370 printf(")\n");
371 printf(" Array State : ");
372 for (d=0; d<__le32_to_cpu(sb->raid_disks); d++) {
373 int cnt = 0;
374 int me = 0;
375 int i;
376 for (i=0; i< __le32_to_cpu(sb->max_dev); i++) {
377 int role = __le16_to_cpu(sb->dev_roles[i]);
378 if (role == d) {
379 if (i == __le32_to_cpu(sb->dev_number))
380 me = 1;
381 cnt++;
382 }
383 }
384 if (cnt > 1) printf("?");
385 else if (cnt == 1 && me) printf("U");
386 else if (cnt == 1) printf("u");
387 else printf ("_");
388 }
389 faulty = 0;
390 for (i=0; i< __le32_to_cpu(sb->max_dev); i++) {
391 int role = __le16_to_cpu(sb->dev_roles[i]);
392 if (role == 0xFFFE)
393 faulty++;
394 }
395 if (faulty) printf(" %d failed", faulty);
396 printf("\n");
397 }
398
399
400 static void brief_examine_super1(struct supertype *st)
401 {
402 struct mdp_superblock_1 *sb = st->sb;
403 int i;
404 unsigned long long sb_offset;
405 char *nm;
406 char *c=map_num(pers, __le32_to_cpu(sb->level));
407
408 nm = strchr(sb->set_name, ':');
409 if (nm)
410 nm++;
411 else if (sb->set_name[0])
412 nm = sb->set_name;
413 else
414 nm = NULL;
415
416 printf("ARRAY%s%s level=%s ",
417 nm ? " /dev/md/":"", nm,
418 c?c:"-unknown-");
419 sb_offset = __le64_to_cpu(sb->super_offset);
420 if (sb_offset <= 4)
421 printf("metadata=1.1 ");
422 else if (sb_offset <= 8)
423 printf("metadata=1.2 ");
424 else
425 printf("metadata=1.0 ");
426 printf("num-devices=%d UUID=", __le32_to_cpu(sb->raid_disks));
427 for (i=0; i<16; i++) {
428 if ((i&3)==0 && i != 0) printf(":");
429 printf("%02x", sb->set_uuid[i]);
430 }
431 if (sb->set_name[0])
432 printf(" name=%.32s", sb->set_name);
433 printf("\n");
434 }
435
436 static void export_examine_super1(struct supertype *st)
437 {
438 struct mdp_superblock_1 *sb = st->sb;
439 int i;
440 int len = 32;
441
442 printf("MD_LEVEL=%s\n", map_num(pers, __le32_to_cpu(sb->level)));
443 printf("MD_DEVICES=%d\n", __le32_to_cpu(sb->raid_disks));
444 for (i=0; i<32; i++)
445 if (sb->set_name[i] == '\n' ||
446 sb->set_name[i] == '\0') {
447 len = i;
448 break;
449 }
450 if (len)
451 printf("MD_NAME=%.*s\n", len, sb->set_name);
452 printf("MD_UUID=");
453 for (i=0; i<16; i++) {
454 if ((i&3)==0 && i != 0) printf(":");
455 printf("%02x", sb->set_uuid[i]);
456 }
457 printf("\n");
458 printf("MD_UPDATE_TIME=%llu\n",
459 __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL);
460 printf("MD_DEV_UUID=");
461 for (i=0; i<16; i++) {
462 if ((i&3)==0 && i != 0) printf(":");
463 printf("%02x", sb->device_uuid[i]);
464 }
465 printf("\n");
466 printf("MD_EVENTS=%llu\n",
467 (unsigned long long)__le64_to_cpu(sb->events));
468 }
469
470 static void detail_super1(struct supertype *st, char *homehost)
471 {
472 struct mdp_superblock_1 *sb = st->sb;
473 int i;
474 int l = homehost ? strlen(homehost) : 0;
475
476 printf(" Name : %.32s", sb->set_name);
477 if (l > 0 && l < 32 &&
478 sb->set_name[l] == ':' &&
479 strncmp(sb->set_name, homehost, l) == 0)
480 printf(" (local to host %s)", homehost);
481 printf("\n UUID : ");
482 for (i=0; i<16; i++) {
483 if ((i&3)==0 && i != 0) printf(":");
484 printf("%02x", sb->set_uuid[i]);
485 }
486 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb->events));
487 }
488
489 static void brief_detail_super1(struct supertype *st)
490 {
491 struct mdp_superblock_1 *sb = st->sb;
492 int i;
493
494 if (sb->set_name[0])
495 printf(" name=%.32s", sb->set_name);
496 printf(" UUID=");
497 for (i=0; i<16; i++) {
498 if ((i&3)==0 && i != 0) printf(":");
499 printf("%02x", sb->set_uuid[i]);
500 }
501 }
502
503 static void export_detail_super1(struct supertype *st)
504 {
505 struct mdp_superblock_1 *sb = st->sb;
506 int i;
507 int len = 32;
508
509 for (i=0; i<32; i++)
510 if (sb->set_name[i] == '\n' ||
511 sb->set_name[i] == '\0') {
512 len = i;
513 break;
514 }
515 if (len)
516 printf("MD_NAME=%.*s\n", len, sb->set_name);
517 }
518
519 #endif
520
521 static int match_home1(struct supertype *st, char *homehost)
522 {
523 struct mdp_superblock_1 *sb = st->sb;
524 int l = homehost ? strlen(homehost) : 0;
525
526 return (l > 0 && l < 32 &&
527 sb->set_name[l] == ':' &&
528 strncmp(sb->set_name, homehost, l) == 0);
529 }
530
531 static void uuid_from_super1(struct supertype *st, int uuid[4])
532 {
533 struct mdp_superblock_1 *super = st->sb;
534 char *cuuid = (char*)uuid;
535 int i;
536 for (i=0; i<16; i++)
537 cuuid[i] = super->set_uuid[i];
538 }
539
540 static void getinfo_super1(struct supertype *st, struct mdinfo *info)
541 {
542 struct mdp_superblock_1 *sb = st->sb;
543 int working = 0;
544 int i;
545 int role;
546
547 info->array.major_version = 1;
548 info->array.minor_version = st->minor_version;
549 info->array.patch_version = 0;
550 info->array.raid_disks = __le32_to_cpu(sb->raid_disks);
551 info->array.level = __le32_to_cpu(sb->level);
552 info->array.layout = __le32_to_cpu(sb->layout);
553 info->array.md_minor = -1;
554 info->array.ctime = __le64_to_cpu(sb->ctime);
555 info->array.utime = __le64_to_cpu(sb->utime);
556 info->array.chunk_size = __le32_to_cpu(sb->chunksize)*512;
557 info->array.state =
558 (__le64_to_cpu(sb->resync_offset) >= __le64_to_cpu(sb->size))
559 ? 1 : 0;
560
561 info->data_offset = __le64_to_cpu(sb->data_offset);
562 info->component_size = __le64_to_cpu(sb->size);
563
564 info->disk.major = 0;
565 info->disk.minor = 0;
566 info->disk.number = __le32_to_cpu(sb->dev_number);
567 if (__le32_to_cpu(sb->dev_number) >= __le32_to_cpu(sb->max_dev) ||
568 __le32_to_cpu(sb->max_dev) > 512)
569 role = 0xfffe;
570 else
571 role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]);
572
573 info->disk.raid_disk = -1;
574 switch(role) {
575 case 0xFFFF:
576 info->disk.state = 2; /* spare: ACTIVE, not sync, not faulty */
577 break;
578 case 0xFFFE:
579 info->disk.state = 1; /* faulty */
580 break;
581 default:
582 info->disk.state = 6; /* active and in sync */
583 info->disk.raid_disk = role;
584 }
585 info->events = __le64_to_cpu(sb->events);
586 sprintf(info->text_version, "1.%d", st->minor_version);
587 info->safe_mode_delay = 200;
588
589 memcpy(info->uuid, sb->set_uuid, 16);
590
591 strncpy(info->name, sb->set_name, 32);
592 info->name[32] = 0;
593
594 if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) {
595 info->reshape_active = 1;
596 info->reshape_progress = __le64_to_cpu(sb->reshape_position);
597 info->new_level = __le32_to_cpu(sb->new_level);
598 info->delta_disks = __le32_to_cpu(sb->delta_disks);
599 info->new_layout = __le32_to_cpu(sb->new_layout);
600 info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9;
601 } else
602 info->reshape_active = 0;
603
604 for (i=0; i< __le32_to_cpu(sb->max_dev); i++) {
605 role = __le16_to_cpu(sb->dev_roles[i]);
606 if (/*role == 0xFFFF || */role < info->array.raid_disks)
607 working++;
608 }
609
610 info->array.working_disks = working;
611 }
612
613 static int update_super1(struct supertype *st, struct mdinfo *info,
614 char *update,
615 char *devname, int verbose,
616 int uuid_set, char *homehost)
617 {
618 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
619 * For others, the return value is ignored.
620 */
621 int rv = 0;
622 struct mdp_superblock_1 *sb = st->sb;
623
624 if (strcmp(update, "force-one")==0) {
625 /* Not enough devices for a working array,
626 * so bring this one up-to-date
627 */
628 if (sb->events != __cpu_to_le64(info->events))
629 rv = 1;
630 sb->events = __cpu_to_le64(info->events);
631 }
632 if (strcmp(update, "force-array")==0) {
633 /* Degraded array and 'force' requests to
634 * maybe need to mark it 'clean'.
635 */
636 switch(__le32_to_cpu(sb->level)) {
637 case 5: case 4: case 6:
638 /* need to force clean */
639 if (sb->resync_offset != ~0ULL)
640 rv = 1;
641 sb->resync_offset = ~0ULL;
642 }
643 }
644 if (strcmp(update, "assemble")==0) {
645 int d = info->disk.number;
646 int want;
647 if (info->disk.state == 6)
648 want = __cpu_to_le32(info->disk.raid_disk);
649 else
650 want = 0xFFFF;
651 if (sb->dev_roles[d] != want) {
652 sb->dev_roles[d] = want;
653 rv = 1;
654 }
655 }
656 if (strcmp(update, "linear-grow-new") == 0) {
657 int i;
658 int rfd, fd;
659 int max = __le32_to_cpu(sb->max_dev);
660
661 for (i=0 ; i < max ; i++)
662 if (__le16_to_cpu(sb->dev_roles[i]) >= 0xfffe)
663 break;
664 sb->dev_number = __cpu_to_le32(i);
665 info->disk.number = i;
666 if (max >= __le32_to_cpu(sb->max_dev))
667 sb->max_dev = __cpu_to_le32(max+1);
668
669 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
670 read(rfd, sb->device_uuid, 16) != 16) {
671 *(__u32*)(sb->device_uuid) = random();
672 *(__u32*)(sb->device_uuid+4) = random();
673 *(__u32*)(sb->device_uuid+8) = random();
674 *(__u32*)(sb->device_uuid+12) = random();
675 }
676
677 sb->dev_roles[i] =
678 __cpu_to_le16(info->disk.raid_disk);
679
680 fd = open(devname, O_RDONLY);
681 if (fd >= 0) {
682 unsigned long long ds;
683 get_dev_size(fd, devname, &ds);
684 close(fd);
685 ds >>= 9;
686 if (__le64_to_cpu(sb->super_offset) <
687 __le64_to_cpu(sb->data_offset)) {
688 sb->data_size = __cpu_to_le64(
689 ds - __le64_to_cpu(sb->data_offset));
690 } else {
691 ds -= 8*2;
692 ds &= ~(unsigned long long)(4*2-1);
693 sb->super_offset = __cpu_to_le64(ds);
694 sb->data_size = __cpu_to_le64(
695 ds - __le64_to_cpu(sb->data_offset));
696 }
697 }
698 }
699 if (strcmp(update, "linear-grow-update") == 0) {
700 sb->raid_disks = __cpu_to_le32(info->array.raid_disks);
701 sb->dev_roles[info->disk.number] =
702 __cpu_to_le16(info->disk.raid_disk);
703 }
704 if (strcmp(update, "resync") == 0) {
705 /* make sure resync happens */
706 sb->resync_offset = 0ULL;
707 }
708 if (strcmp(update, "uuid") == 0) {
709 copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid);
710
711 if (__le32_to_cpu(sb->feature_map)&MD_FEATURE_BITMAP_OFFSET) {
712 struct bitmap_super_s *bm;
713 bm = (struct bitmap_super_s*)(st->sb+1024);
714 memcpy(bm->uuid, sb->set_uuid, 16);
715 }
716 }
717 if (strcmp(update, "homehost") == 0 &&
718 homehost) {
719 char *c;
720 update = "name";
721 c = strchr(sb->set_name, ':');
722 if (c)
723 strncpy(info->name, c+1, 31 - (c-sb->set_name));
724 else
725 strncpy(info->name, sb->set_name, 32);
726 info->name[32] = 0;
727 }
728 if (strcmp(update, "name") == 0) {
729 if (info->name[0] == 0)
730 sprintf(info->name, "%d", info->array.md_minor);
731 memset(sb->set_name, 0, sizeof(sb->set_name));
732 if (homehost &&
733 strchr(info->name, ':') == NULL &&
734 strlen(homehost)+1+strlen(info->name) < 32) {
735 strcpy(sb->set_name, homehost);
736 strcat(sb->set_name, ":");
737 strcat(sb->set_name, info->name);
738 } else
739 strcpy(sb->set_name, info->name);
740 }
741 if (strcmp(update, "devicesize") == 0 &&
742 __le64_to_cpu(sb->super_offset) <
743 __le64_to_cpu(sb->data_offset)) {
744 /* set data_size to device size less data_offset */
745 struct misc_dev_info *misc = (struct misc_dev_info*)
746 (st->sb + 1024 + 512);
747 printf("Size was %llu\n", (unsigned long long)
748 __le64_to_cpu(sb->data_size));
749 sb->data_size = __cpu_to_le64(
750 misc->device_size - __le64_to_cpu(sb->data_offset));
751 printf("Size is %llu\n", (unsigned long long)
752 __le64_to_cpu(sb->data_size));
753 }
754 if (strcmp(update, "_reshape_progress")==0)
755 sb->reshape_position = __cpu_to_le64(info->reshape_progress);
756
757 sb->sb_csum = calc_sb_1_csum(sb);
758 return rv;
759 }
760
761 static int init_super1(struct supertype *st, mdu_array_info_t *info,
762 unsigned long long size, char *name, char *homehost, int *uuid)
763 {
764 struct mdp_superblock_1 *sb;
765 int spares;
766 int rfd;
767 char defname[10];
768
769 if (posix_memalign((void**)&sb, 512, (1024 + 512 +
770 sizeof(struct misc_dev_info))) != 0) {
771 fprintf(stderr, Name
772 ": %s could not allocate superblock\n", __func__);
773 return 0;
774 }
775 memset(sb, 0, 1024);
776
777 st->sb = sb;
778 if (info == NULL) {
779 /* zeroing superblock */
780 return 0;
781 }
782
783 spares = info->working_disks - info->active_disks;
784 if (info->raid_disks + spares > 384) {
785 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
786 info->raid_disks , spares, 384);
787 return 0;
788 }
789
790 sb->magic = __cpu_to_le32(MD_SB_MAGIC);
791 sb->major_version = __cpu_to_le32(1);
792 sb->feature_map = 0;
793 sb->pad0 = 0;
794
795 if (uuid)
796 copy_uuid(sb->set_uuid, uuid, super1.swapuuid);
797 else {
798 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
799 read(rfd, sb->set_uuid, 16) != 16) {
800 *(__u32*)(sb->set_uuid) = random();
801 *(__u32*)(sb->set_uuid+4) = random();
802 *(__u32*)(sb->set_uuid+8) = random();
803 *(__u32*)(sb->set_uuid+12) = random();
804 }
805 if (rfd >= 0) close(rfd);
806 }
807
808 if (name == NULL || *name == 0) {
809 sprintf(defname, "%d", info->md_minor);
810 name = defname;
811 }
812 memset(sb->set_name, 0, 32);
813 if (homehost &&
814 strchr(name, ':')== NULL &&
815 strlen(homehost)+1+strlen(name) < 32) {
816 strcpy(sb->set_name, homehost);
817 strcat(sb->set_name, ":");
818 strcat(sb->set_name, name);
819 } else
820 strcpy(sb->set_name, name);
821
822 sb->ctime = __cpu_to_le64((unsigned long long)time(0));
823 sb->level = __cpu_to_le32(info->level);
824 sb->layout = __cpu_to_le32(info->layout);
825 sb->size = __cpu_to_le64(size*2ULL);
826 sb->chunksize = __cpu_to_le32(info->chunk_size>>9);
827 sb->raid_disks = __cpu_to_le32(info->raid_disks);
828
829 sb->data_offset = __cpu_to_le64(0);
830 sb->data_size = __cpu_to_le64(0);
831 sb->super_offset = __cpu_to_le64(0);
832 sb->recovery_offset = __cpu_to_le64(0);
833
834 sb->utime = sb->ctime;
835 sb->events = __cpu_to_le64(1);
836 if (info->state & (1<<MD_SB_CLEAN))
837 sb->resync_offset = ~0ULL;
838 else
839 sb->resync_offset = 0;
840 sb->max_dev = __cpu_to_le32((1024- sizeof(struct mdp_superblock_1))/
841 sizeof(sb->dev_roles[0]));
842 memset(sb->pad3, 0, sizeof(sb->pad3));
843
844 memset(sb->dev_roles, 0xff, 1024 - sizeof(struct mdp_superblock_1));
845
846 return 1;
847 }
848
849 struct devinfo {
850 int fd;
851 char *devname;
852 mdu_disk_info_t disk;
853 struct devinfo *next;
854 };
855 #ifndef MDASSEMBLE
856 /* Add a device to the superblock being created */
857 static int add_to_super1(struct supertype *st, mdu_disk_info_t *dk,
858 int fd, char *devname)
859 {
860 struct mdp_superblock_1 *sb = st->sb;
861 __u16 *rp = sb->dev_roles + dk->number;
862 struct devinfo *di, **dip;
863
864 if ((dk->state & 6) == 6) /* active, sync */
865 *rp = __cpu_to_le16(dk->raid_disk);
866 else if ((dk->state & ~2) == 0) /* active or idle -> spare */
867 *rp = 0xffff;
868 else
869 *rp = 0xfffe;
870
871 sb->dev_number = __cpu_to_le32(dk->number);
872 sb->sb_csum = calc_sb_1_csum(sb);
873
874 dip = (struct devinfo **)&st->info;
875 while (*dip)
876 dip = &(*dip)->next;
877 di = malloc(sizeof(struct devinfo));
878 di->fd = fd;
879 di->devname = devname;
880 di->disk = *dk;
881 di->next = NULL;
882 *dip = di;
883
884 return 0;
885 }
886 #endif
887
888 static void locate_bitmap1(struct supertype *st, int fd);
889
890 static int store_super1(struct supertype *st, int fd)
891 {
892 struct mdp_superblock_1 *sb = st->sb;
893 unsigned long long sb_offset;
894 int sbsize;
895 unsigned long long dsize;
896
897 if (!get_dev_size(fd, NULL, &dsize))
898 return 1;
899
900 dsize >>= 9;
901
902 if (dsize < 24)
903 return 2;
904
905 /*
906 * Calculate the position of the superblock.
907 * It is always aligned to a 4K boundary and
908 * depending on minor_version, it can be:
909 * 0: At least 8K, but less than 12K, from end of device
910 * 1: At start of device
911 * 2: 4K from start of device.
912 */
913 switch(st->minor_version) {
914 case 0:
915 sb_offset = dsize;
916 sb_offset -= 8*2;
917 sb_offset &= ~(4*2-1);
918 break;
919 case 1:
920 sb_offset = 0;
921 break;
922 case 2:
923 sb_offset = 4*2;
924 break;
925 default:
926 return -EINVAL;
927 }
928
929
930
931 if (sb_offset != __le64_to_cpu(sb->super_offset) &&
932 0 != __le64_to_cpu(sb->super_offset)
933 ) {
934 fprintf(stderr, Name ": internal error - sb_offset is wrong\n");
935 abort();
936 }
937
938 if (lseek64(fd, sb_offset << 9, 0)< 0LL)
939 return 3;
940
941 sbsize = sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev);
942 sbsize = (sbsize+511)&(~511UL);
943
944 if (awrite(fd, sb, sbsize) != sbsize)
945 return 4;
946
947 if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) {
948 struct bitmap_super_s *bm = (struct bitmap_super_s*)
949 (((char*)sb)+1024);
950 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) {
951 locate_bitmap1(st, fd);
952 if (awrite(fd, bm, sizeof(*bm)) !=
953 sizeof(*bm))
954 return 5;
955 }
956 }
957 fsync(fd);
958 return 0;
959 }
960
961 static int load_super1(struct supertype *st, int fd, char *devname);
962
963 static unsigned long choose_bm_space(unsigned long devsize)
964 {
965 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
966 * if bigger than 200Gig, save 128k
967 */
968 if (devsize < 64*2) return 0;
969 if (devsize - 64*2 >= 200*1024*1024*2)
970 return 128*2;
971 if (devsize - 4*2 > 8*1024*1024*2)
972 return 64*2;
973 return 4*2;
974 }
975
976 #ifndef MDASSEMBLE
977 static int write_init_super1(struct supertype *st)
978 {
979 struct mdp_superblock_1 *sb = st->sb;
980 struct supertype refst;
981 int rfd;
982 int rv = 0;
983 int bm_space;
984 struct devinfo *di;
985 unsigned long long dsize, array_size;
986 long long sb_offset;
987
988 for (di = st->info; di && ! rv ; di = di->next) {
989 if (di->disk.state == 1)
990 continue;
991 if (di->fd < 0)
992 continue;
993
994 Kill(di->devname, 0, 1, 1);
995 Kill(di->devname, 0, 1, 1);
996
997 sb->dev_number = __cpu_to_le32(di->disk.number);
998 if (di->disk.state & (1<<MD_DISK_WRITEMOSTLY))
999 sb->devflags |= __cpu_to_le32(WriteMostly1);
1000
1001 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
1002 read(rfd, sb->device_uuid, 16) != 16) {
1003 *(__u32*)(sb->device_uuid) = random();
1004 *(__u32*)(sb->device_uuid+4) = random();
1005 *(__u32*)(sb->device_uuid+8) = random();
1006 *(__u32*)(sb->device_uuid+12) = random();
1007 }
1008 if (rfd >= 0) close(rfd);
1009 sb->events = 0;
1010
1011 refst =*st;
1012 refst.sb = NULL;
1013 if (load_super1(&refst, di->fd, NULL)==0) {
1014 struct mdp_superblock_1 *refsb = refst.sb;
1015
1016 memcpy(sb->device_uuid, refsb->device_uuid, 16);
1017 if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) {
1018 /* same array, so preserve events and
1019 * dev_number */
1020 sb->events = refsb->events;
1021 /* bugs in 2.6.17 and earlier mean the
1022 * dev_number chosen in Manage must be preserved
1023 */
1024 if (get_linux_version() >= 2006018)
1025 sb->dev_number = refsb->dev_number;
1026 }
1027 free(refsb);
1028 }
1029
1030 if (!get_dev_size(di->fd, NULL, &dsize))
1031 return 1;
1032 dsize >>= 9;
1033
1034 if (dsize < 24) {
1035 close(di->fd);
1036 return 2;
1037 }
1038
1039
1040 /*
1041 * Calculate the position of the superblock.
1042 * It is always aligned to a 4K boundary and
1043 * depending on minor_version, it can be:
1044 * 0: At least 8K, but less than 12K, from end of device
1045 * 1: At start of device
1046 * 2: 4K from start of device.
1047 * Depending on the array size, we might leave extra space
1048 * for a bitmap.
1049 */
1050 array_size = __le64_to_cpu(sb->size);
1051 /* work out how much space we left for a bitmap */
1052 bm_space = choose_bm_space(array_size);
1053
1054 switch(st->minor_version) {
1055 case 0:
1056 sb_offset = dsize;
1057 sb_offset -= 8*2;
1058 sb_offset &= ~(4*2-1);
1059 sb->super_offset = __cpu_to_le64(sb_offset);
1060 sb->data_offset = __cpu_to_le64(0);
1061 if (sb_offset - bm_space < array_size)
1062 bm_space = sb_offset - array_size;
1063 sb->data_size = __cpu_to_le64(sb_offset - bm_space);
1064 break;
1065 case 1:
1066 sb->super_offset = __cpu_to_le64(0);
1067 if (4*2 + bm_space + __le64_to_cpu(sb->size) > dsize)
1068 bm_space = dsize - __le64_to_cpu(sb->size) -4*2;
1069 sb->data_offset = __cpu_to_le64(bm_space + 4*2);
1070 sb->data_size = __cpu_to_le64(dsize - bm_space - 4*2);
1071 break;
1072 case 2:
1073 sb_offset = 4*2;
1074 sb->super_offset = __cpu_to_le64(4*2);
1075 if (4*2 + 4*2 + bm_space + __le64_to_cpu(sb->size)
1076 > dsize)
1077 bm_space = dsize - __le64_to_cpu(sb->size)
1078 - 4*2 - 4*2;
1079 sb->data_offset = __cpu_to_le64(4*2 + 4*2 + bm_space);
1080 sb->data_size = __cpu_to_le64(dsize - 4*2 - 4*2
1081 - bm_space );
1082 break;
1083 default:
1084 return -EINVAL;
1085 }
1086
1087
1088 sb->sb_csum = calc_sb_1_csum(sb);
1089 rv = store_super1(st, di->fd);
1090 if (rv)
1091 fprintf(stderr,
1092 Name ": failed to write superblock to %s\n",
1093 di->devname);
1094
1095 if (rv == 0 && (__le32_to_cpu(sb->feature_map) & 1))
1096 rv = st->ss->write_bitmap(st, di->fd);
1097 close(di->fd);
1098 di->fd = -1;
1099 }
1100 return rv;
1101 }
1102 #endif
1103
1104 static int compare_super1(struct supertype *st, struct supertype *tst)
1105 {
1106 /*
1107 * return:
1108 * 0 same, or first was empty, and second was copied
1109 * 1 second had wrong number
1110 * 2 wrong uuid
1111 * 3 wrong other info
1112 */
1113 struct mdp_superblock_1 *first = st->sb;
1114 struct mdp_superblock_1 *second = tst->sb;
1115
1116 if (second->magic != __cpu_to_le32(MD_SB_MAGIC))
1117 return 1;
1118 if (second->major_version != __cpu_to_le32(1))
1119 return 1;
1120
1121 if (!first) {
1122 if (posix_memalign((void**)&first, 512,
1123 1024 + 512 +
1124 sizeof(struct misc_dev_info)) != 0) {
1125 fprintf(stderr, Name
1126 ": %s could not allocate superblock\n", __func__);
1127 return 1;
1128 }
1129 memcpy(first, second, 1024 + 512 +
1130 sizeof(struct misc_dev_info));
1131 st->sb = first;
1132 return 0;
1133 }
1134 if (memcmp(first->set_uuid, second->set_uuid, 16)!= 0)
1135 return 2;
1136
1137 if (first->ctime != second->ctime ||
1138 first->level != second->level ||
1139 first->layout != second->layout ||
1140 first->size != second->size ||
1141 first->chunksize != second->chunksize ||
1142 first->raid_disks != second->raid_disks)
1143 return 3;
1144 return 0;
1145 }
1146
1147 static void free_super1(struct supertype *st);
1148
1149 static int load_super1(struct supertype *st, int fd, char *devname)
1150 {
1151 unsigned long long dsize;
1152 unsigned long long sb_offset;
1153 struct mdp_superblock_1 *super;
1154 int uuid[4];
1155 struct bitmap_super_s *bsb;
1156 struct misc_dev_info *misc;
1157
1158 free_super1(st);
1159
1160 if (st->subarray[0])
1161 return 1;
1162
1163 if (st->ss == NULL || st->minor_version == -1) {
1164 int bestvers = -1;
1165 struct supertype tst;
1166 __u64 bestctime = 0;
1167 /* guess... choose latest ctime */
1168 memset(&tst, 0, sizeof(tst));
1169 tst.ss = &super1;
1170 for (tst.minor_version = 0; tst.minor_version <= 2 ; tst.minor_version++) {
1171 switch(load_super1(&tst, fd, devname)) {
1172 case 0: super = tst.sb;
1173 if (bestvers == -1 ||
1174 bestctime < __le64_to_cpu(super->ctime)) {
1175 bestvers = tst.minor_version;
1176 bestctime = __le64_to_cpu(super->ctime);
1177 }
1178 free(super);
1179 tst.sb = NULL;
1180 break;
1181 case 1: return 1; /*bad device */
1182 case 2: break; /* bad, try next */
1183 }
1184 }
1185 if (bestvers != -1) {
1186 int rv;
1187 tst.minor_version = bestvers;
1188 tst.ss = &super1;
1189 tst.max_devs = 384;
1190 rv = load_super1(&tst, fd, devname);
1191 if (rv == 0)
1192 *st = tst;
1193 return rv;
1194 }
1195 return 2;
1196 }
1197 if (!get_dev_size(fd, devname, &dsize))
1198 return 1;
1199 dsize >>= 9;
1200
1201 if (dsize < 24) {
1202 if (devname)
1203 fprintf(stderr, Name ": %s is too small for md: size is %llu sectors.\n",
1204 devname, dsize);
1205 return 1;
1206 }
1207
1208 /*
1209 * Calculate the position of the superblock.
1210 * It is always aligned to a 4K boundary and
1211 * depending on minor_version, it can be:
1212 * 0: At least 8K, but less than 12K, from end of device
1213 * 1: At start of device
1214 * 2: 4K from start of device.
1215 */
1216 switch(st->minor_version) {
1217 case 0:
1218 sb_offset = dsize;
1219 sb_offset -= 8*2;
1220 sb_offset &= ~(4*2-1);
1221 break;
1222 case 1:
1223 sb_offset = 0;
1224 break;
1225 case 2:
1226 sb_offset = 4*2;
1227 break;
1228 default:
1229 return -EINVAL;
1230 }
1231
1232 ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
1233
1234
1235 if (lseek64(fd, sb_offset << 9, 0)< 0LL) {
1236 if (devname)
1237 fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
1238 devname, strerror(errno));
1239 return 1;
1240 }
1241
1242 if (posix_memalign((void**)&super, 512,
1243 1024 + 512 +
1244 sizeof(struct misc_dev_info)) != 0) {
1245 fprintf(stderr, Name ": %s could not allocate superblock\n",
1246 __func__);
1247 return 1;
1248 }
1249
1250 if (aread(fd, super, 1024) != 1024) {
1251 if (devname)
1252 fprintf(stderr, Name ": Cannot read superblock on %s\n",
1253 devname);
1254 free(super);
1255 return 1;
1256 }
1257
1258 if (__le32_to_cpu(super->magic) != MD_SB_MAGIC) {
1259 if (devname)
1260 fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
1261 devname, MD_SB_MAGIC, __le32_to_cpu(super->magic));
1262 free(super);
1263 return 2;
1264 }
1265
1266 if (__le32_to_cpu(super->major_version) != 1) {
1267 if (devname)
1268 fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
1269 devname, __le32_to_cpu(super->major_version));
1270 free(super);
1271 return 2;
1272 }
1273 if (__le64_to_cpu(super->super_offset) != sb_offset) {
1274 if (devname)
1275 fprintf(stderr, Name ": No superblock found on %s (super_offset is wrong)\n",
1276 devname);
1277 free(super);
1278 return 2;
1279 }
1280 st->sb = super;
1281
1282 bsb = (struct bitmap_super_s *)(((char*)super)+1024);
1283
1284 misc = (struct misc_dev_info*) (((char*)super)+1024+512);
1285 misc->device_size = dsize;
1286
1287 /* Now check on the bitmap superblock */
1288 if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0)
1289 return 0;
1290 /* Read the bitmap superblock and make sure it looks
1291 * valid. If it doesn't clear the bit. An --assemble --force
1292 * should get that written out.
1293 */
1294 locate_bitmap1(st, fd);
1295 if (aread(fd, ((char*)super)+1024, 512)
1296 != 512)
1297 goto no_bitmap;
1298
1299 uuid_from_super1(st, uuid);
1300 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
1301 memcmp(bsb->uuid, uuid, 16) != 0)
1302 goto no_bitmap;
1303 return 0;
1304
1305 no_bitmap:
1306 super->feature_map = __cpu_to_le32(__le32_to_cpu(super->feature_map) & ~1);
1307 return 0;
1308 }
1309
1310
1311 static struct supertype *match_metadata_desc1(char *arg)
1312 {
1313 struct supertype *st = malloc(sizeof(*st));
1314 if (!st) return st;
1315
1316 memset(st, 0, sizeof(*st));
1317 st->ss = &super1;
1318 st->max_devs = 384;
1319 st->sb = NULL;
1320 /* leading zeros can be safely ignored. --detail generates them. */
1321 while (*arg == '0')
1322 arg++;
1323 if (strcmp(arg, "1.0") == 0 ||
1324 strcmp(arg, "1.00") == 0) {
1325 st->minor_version = 0;
1326 return st;
1327 }
1328 if (strcmp(arg, "1.1") == 0 ||
1329 strcmp(arg, "1.01") == 0) {
1330 st->minor_version = 1;
1331 return st;
1332 }
1333 if (strcmp(arg, "1.2") == 0 ||
1334 strcmp(arg, "1.02") == 0) {
1335 st->minor_version = 2;
1336 return st;
1337 }
1338 if (strcmp(arg, "1") == 0 ||
1339 strcmp(arg, "default") == 0) {
1340 st->minor_version = -1;
1341 return st;
1342 }
1343
1344 free(st);
1345 return NULL;
1346 }
1347
1348 /* find available size on device with this devsize, using
1349 * superblock type st, and reserving 'reserve' sectors for
1350 * a possible bitmap
1351 */
1352 static __u64 avail_size1(struct supertype *st, __u64 devsize)
1353 {
1354 struct mdp_superblock_1 *super = st->sb;
1355 if (devsize < 24)
1356 return 0;
1357
1358 if (super == NULL)
1359 /* creating: allow suitable space for bitmap */
1360 devsize -= choose_bm_space(devsize);
1361 #ifndef MDASSEMBLE
1362 else if (__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) {
1363 /* hot-add. allow for actual size of bitmap */
1364 struct bitmap_super_s *bsb;
1365 bsb = (struct bitmap_super_s *)(((char*)super)+1024);
1366 devsize -= bitmap_sectors(bsb);
1367 }
1368 #endif
1369
1370 switch(st->minor_version) {
1371 case -1: /* no specified. Now time to set default */
1372 st->minor_version = 0;
1373 /* FALL THROUGH */
1374 case 0:
1375 /* at end */
1376 return ((devsize - 8*2 ) & ~(4*2-1));
1377 case 1:
1378 /* at start, 4K for superblock and possible bitmap */
1379 return devsize - 4*2;
1380 case 2:
1381 /* 4k from start, 4K for superblock and possible bitmap */
1382 return devsize - (4+4)*2;
1383 }
1384 return 0;
1385 }
1386
1387 static int
1388 add_internal_bitmap1(struct supertype *st,
1389 int *chunkp, int delay, int write_behind,
1390 unsigned long long size,
1391 int may_change, int major)
1392 {
1393 /*
1394 * If not may_change, then this is a 'Grow', and the bitmap
1395 * must fit after the superblock.
1396 * If may_change, then this is create, and we can put the bitmap
1397 * before the superblock if we like, or may move the start.
1398 * If !may_change, the bitmap MUST live at offset of 1K, until
1399 * we get a sysfs interface.
1400 *
1401 * size is in sectors, chunk is in bytes !!!
1402 */
1403
1404 unsigned long long bits;
1405 unsigned long long max_bits;
1406 unsigned long long min_chunk;
1407 long offset;
1408 int chunk = *chunkp;
1409 int room = 0;
1410 struct mdp_superblock_1 *sb = st->sb;
1411 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + 1024);
1412
1413 switch(st->minor_version) {
1414 case 0:
1415 /* either 3K after the superblock, or some amount of space
1416 * before.
1417 */
1418 if (may_change) {
1419 /* We are creating array, so we *know* how much room has
1420 * been left.
1421 */
1422 offset = 0;
1423 room = choose_bm_space(__le64_to_cpu(sb->size));
1424 if (room == 4*2) {
1425 /* make it 3K after the superblock */
1426 room = 3*2;
1427 offset = 2;
1428 }
1429 } else {
1430 room = __le64_to_cpu(sb->super_offset)
1431 - __le64_to_cpu(sb->data_offset)
1432 - __le64_to_cpu(sb->data_size);
1433 /* remove '1 ||' when we can set offset via sysfs */
1434 if (1 || (room < 3*2 &&
1435 __le32_to_cpu(sb->max_dev) <= 384)) {
1436 room = 3*2;
1437 offset = 1*2;
1438 } else {
1439 offset = 0; /* means movable offset */
1440 }
1441 }
1442 break;
1443 case 1:
1444 case 2: /* between superblock and data */
1445 if (may_change) {
1446 offset = 4*2;
1447 room = choose_bm_space(__le64_to_cpu(sb->size));
1448 } else {
1449 room = __le64_to_cpu(sb->data_offset)
1450 - __le64_to_cpu(sb->super_offset);
1451 if (1 || __le32_to_cpu(sb->max_dev) <= 384) {
1452 room -= 2;
1453 offset = 2;
1454 } else {
1455 room -= 4*2;
1456 offset = 4*2;
1457 }
1458 }
1459 break;
1460 default:
1461 return 0;
1462 }
1463
1464 if (chunk == UnSet && room > 128*2)
1465 /* Limit to 128K of bitmap when chunk size not requested */
1466 room = 128*2;
1467
1468 max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8;
1469
1470 min_chunk = 4096; /* sub-page chunks don't work yet.. */
1471 bits = (size*512)/min_chunk +1;
1472 while (bits > max_bits) {
1473 min_chunk *= 2;
1474 bits = (bits+1)/2;
1475 }
1476 if (chunk == UnSet)
1477 chunk = min_chunk;
1478 else if (chunk < min_chunk)
1479 return 0; /* chunk size too small */
1480 if (chunk == 0) /* rounding problem */
1481 return 0;
1482
1483 if (offset == 0) {
1484 bits = (size*512) / chunk + 1;
1485 room = ((bits+7)/8 + sizeof(bitmap_super_t) +511)/512;
1486 offset = -room;
1487 }
1488
1489 sb->bitmap_offset = __cpu_to_le32(offset);
1490
1491 sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | 1);
1492 memset(bms, 0, sizeof(*bms));
1493 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
1494 bms->version = __cpu_to_le32(major);
1495 uuid_from_super1(st, (int*)bms->uuid);
1496 bms->chunksize = __cpu_to_le32(chunk);
1497 bms->daemon_sleep = __cpu_to_le32(delay);
1498 bms->sync_size = __cpu_to_le64(size);
1499 bms->write_behind = __cpu_to_le32(write_behind);
1500
1501 *chunkp = chunk;
1502 return 1;
1503 }
1504
1505
1506 static void locate_bitmap1(struct supertype *st, int fd)
1507 {
1508 unsigned long long offset;
1509 struct mdp_superblock_1 *sb;
1510 int mustfree = 0;
1511
1512 if (!st->sb) {
1513 if (st->ss->load_super(st, fd, NULL))
1514 return; /* no error I hope... */
1515 mustfree = 1;
1516 }
1517 sb = st->sb;
1518
1519 offset = __le64_to_cpu(sb->super_offset);
1520 offset += (int32_t) __le32_to_cpu(sb->bitmap_offset);
1521 if (mustfree)
1522 free(sb);
1523 lseek64(fd, offset<<9, 0);
1524 }
1525
1526 static int write_bitmap1(struct supertype *st, int fd)
1527 {
1528 struct mdp_superblock_1 *sb = st->sb;
1529 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+1024);
1530 int rv = 0;
1531
1532 int towrite, n;
1533 char *buf = (char*)(((long)(abuf+4096))&~4095UL);
1534
1535 locate_bitmap1(st, fd);
1536
1537 memset(buf, 0xff, 4096);
1538 memcpy(buf, ((char*)sb)+1024, sizeof(bitmap_super_t));
1539
1540 towrite = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9);
1541 towrite = (towrite+7) >> 3; /* bits to bytes */
1542 towrite += sizeof(bitmap_super_t);
1543 towrite = ROUND_UP(towrite, 512);
1544 while (towrite > 0) {
1545 n = towrite;
1546 if (n > 4096)
1547 n = 4096;
1548 n = write(fd, buf, n);
1549 if (n > 0)
1550 towrite -= n;
1551 else
1552 break;
1553 memset(buf, 0xff, 4096);
1554 }
1555 fsync(fd);
1556 if (towrite)
1557 rv = -2;
1558
1559 return rv;
1560 }
1561
1562 static void free_super1(struct supertype *st)
1563 {
1564 if (st->sb)
1565 free(st->sb);
1566 st->sb = NULL;
1567 }
1568
1569 #ifndef MDASSEMBLE
1570 static int validate_geometry1(struct supertype *st, int level,
1571 int layout, int raiddisks,
1572 int chunk, unsigned long long size,
1573 char *subdev, unsigned long long *freesize,
1574 int verbose)
1575 {
1576 unsigned long long ldsize;
1577 int fd;
1578
1579 if (level == LEVEL_CONTAINER)
1580 return 0;
1581 if (!subdev)
1582 return 1;
1583
1584 fd = open(subdev, O_RDONLY|O_EXCL, 0);
1585 if (fd < 0) {
1586 if (verbose)
1587 fprintf(stderr, Name ": super1.x cannot open %s: %s\n",
1588 subdev, strerror(errno));
1589 return 0;
1590 }
1591
1592 if (!get_dev_size(fd, subdev, &ldsize)) {
1593 close(fd);
1594 return 0;
1595 }
1596 close(fd);
1597
1598 *freesize = avail_size1(st, ldsize >> 9);
1599 return 1;
1600 }
1601 #endif /* MDASSEMBLE */
1602
1603 struct superswitch super1 = {
1604 #ifndef MDASSEMBLE
1605 .examine_super = examine_super1,
1606 .brief_examine_super = brief_examine_super1,
1607 .export_examine_super = export_examine_super1,
1608 .detail_super = detail_super1,
1609 .brief_detail_super = brief_detail_super1,
1610 .export_detail_super = export_detail_super1,
1611 .write_init_super = write_init_super1,
1612 .validate_geometry = validate_geometry1,
1613 .add_to_super = add_to_super1,
1614 #endif
1615 .match_home = match_home1,
1616 .uuid_from_super = uuid_from_super1,
1617 .getinfo_super = getinfo_super1,
1618 .update_super = update_super1,
1619 .init_super = init_super1,
1620 .store_super = store_super1,
1621 .compare_super = compare_super1,
1622 .load_super = load_super1,
1623 .match_metadata_desc = match_metadata_desc1,
1624 .avail_size = avail_size1,
1625 .add_internal_bitmap = add_internal_bitmap1,
1626 .locate_bitmap = locate_bitmap1,
1627 .write_bitmap = write_bitmap1,
1628 .free_super = free_super1,
1629 #if __BYTE_ORDER == BIG_ENDIAN
1630 .swapuuid = 0,
1631 #else
1632 .swapuuid = 1,
1633 #endif
1634 .name = "1.0",
1635 };
Unnamed repository; edit this file 'description' to name the repository.