]> 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
Add some comments to explain some of the bits of superswitch.
[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 #ifndef MDASSEMBLE
145 static void examine_super1(struct supertype *st, char *homehost)
146 {
147 struct mdp_superblock_1 *sb = st->sb;
148 time_t atime;
149 int d;
150 int faulty;
151 int i;
152 char *c;
153 int l = homehost ? strlen(homehost) : 0;
154 int layout;
155 unsigned long long sb_offset;
156
157 printf(" Magic : %08x\n", __le32_to_cpu(sb->magic));
158 printf(" Version : 1");
159 sb_offset = __le64_to_cpu(sb->super_offset);
160 if (sb_offset <= 4)
161 printf(".1\n");
162 else if (sb_offset <= 8)
163 printf(".2\n");
164 else
165 printf(".0\n");
166 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb->feature_map));
167 printf(" Array UUID : ");
168 for (i=0; i<16; i++) {
169 if ((i&3)==0 && i != 0) printf(":");
170 printf("%02x", sb->set_uuid[i]);
171 }
172 printf("\n");
173 printf(" Name : %.32s", sb->set_name);
174 if (l > 0 && l < 32 &&
175 sb->set_name[l] == ':' &&
176 strncmp(sb->set_name, homehost, l) == 0)
177 printf(" (local to host %s)", homehost);
178 printf("\n");
179 atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL;
180 printf(" Creation Time : %.24s\n", ctime(&atime));
181 c=map_num(pers, __le32_to_cpu(sb->level));
182 printf(" Raid Level : %s\n", c?c:"-unknown-");
183 printf(" Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks));
184 printf("\n");
185 printf(" Avail Dev Size : %llu%s\n",
186 (unsigned long long)__le64_to_cpu(sb->data_size),
187 human_size(__le64_to_cpu(sb->data_size)<<9));
188 if (__le32_to_cpu(sb->level) >= 0) {
189 int ddsks=0;
190 switch(__le32_to_cpu(sb->level)) {
191 case 1: ddsks=1;break;
192 case 4:
193 case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break;
194 case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break;
195 case 10:
196 layout = __le32_to_cpu(sb->layout);
197 ddsks = __le32_to_cpu(sb->raid_disks)
198 / (layout&255) / ((layout>>8)&255);
199 }
200 if (ddsks)
201 printf(" Array Size : %llu%s\n",
202 ddsks*(unsigned long long)__le64_to_cpu(sb->size),
203 human_size(ddsks*__le64_to_cpu(sb->size)<<9));
204 if (sb->size != sb->data_size)
205 printf(" Used Dev Size : %llu%s\n",
206 (unsigned long long)__le64_to_cpu(sb->size),
207 human_size(__le64_to_cpu(sb->size)<<9));
208 }
209 if (sb->data_offset)
210 printf(" Data Offset : %llu sectors\n",
211 (unsigned long long)__le64_to_cpu(sb->data_offset));
212 printf(" Super Offset : %llu sectors\n",
213 (unsigned long long)__le64_to_cpu(sb->super_offset));
214 if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET)
215 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->recovery_offset));
216 printf(" State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean");
217 printf(" Device UUID : ");
218 for (i=0; i<16; i++) {
219 if ((i&3)==0 && i != 0) printf(":");
220 printf("%02x", sb->device_uuid[i]);
221 }
222 printf("\n");
223 printf("\n");
224 if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) {
225 printf("Internal Bitmap : %ld sectors from superblock\n",
226 (long)(int32_t)__le32_to_cpu(sb->bitmap_offset));
227 }
228 if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) {
229 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb->reshape_position)/2,
230 human_size(__le64_to_cpu(sb->reshape_position)<<9));
231 if (__le32_to_cpu(sb->delta_disks)) {
232 printf(" Delta Devices : %d", __le32_to_cpu(sb->delta_disks));
233 if (__le32_to_cpu(sb->delta_disks))
234 printf(" (%d->%d)\n",
235 __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks),
236 __le32_to_cpu(sb->raid_disks));
237 else
238 printf(" (%d->%d)\n", __le32_to_cpu(sb->raid_disks),
239 __le32_to_cpu(sb->raid_disks)+__le32_to_cpu(sb->delta_disks));
240 }
241 if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) {
242 c = map_num(pers, __le32_to_cpu(sb->new_level));
243 printf(" New Level : %s\n", c?c:"-unknown-");
244 }
245 if (__le32_to_cpu(sb->new_layout) != __le32_to_cpu(sb->layout)) {
246 if (__le32_to_cpu(sb->level) == 5) {
247 c = map_num(r5layout, __le32_to_cpu(sb->new_layout));
248 printf(" New Layout : %s\n", c?c:"-unknown-");
249 }
250 if (__le32_to_cpu(sb->level) == 10) {
251 printf(" New Layout : near=%d, %s=%d\n",
252 __le32_to_cpu(sb->new_layout)&255,
253 (__le32_to_cpu(sb->new_layout)&0x10000)?"offset":"far",
254 (__le32_to_cpu(sb->new_layout)>>8)&255);
255 }
256 }
257 if (__le32_to_cpu(sb->new_chunk) != __le32_to_cpu(sb->chunksize))
258 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb->new_chunk)/2);
259 printf("\n");
260 }
261 if (sb->devflags) {
262 printf(" Flags :");
263 if (sb->devflags & WriteMostly1)
264 printf(" write-mostly");
265 printf("\n");
266 }
267
268 atime = __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL;
269 printf(" Update Time : %.24s\n", ctime(&atime));
270
271 if (calc_sb_1_csum(sb) == sb->sb_csum)
272 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum));
273 else
274 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb->sb_csum),
275 __le32_to_cpu(calc_sb_1_csum(sb)));
276 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb->events));
277 printf("\n");
278 if (__le32_to_cpu(sb->level) == 5) {
279 c = map_num(r5layout, __le32_to_cpu(sb->layout));
280 printf(" Layout : %s\n", c?c:"-unknown-");
281 }
282 if (__le32_to_cpu(sb->level) == 10) {
283 int lo = __le32_to_cpu(sb->layout);
284 printf(" Layout : near=%d, %s=%d\n",
285 lo&255,
286 (lo&0x10000)?"offset":"far",
287 (lo>>8)&255);
288 }
289 switch(__le32_to_cpu(sb->level)) {
290 case 0:
291 case 4:
292 case 5:
293 case 6:
294 case 10:
295 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize)/2);
296 break;
297 case -1:
298 printf(" Rounding : %dK\n", __le32_to_cpu(sb->chunksize)/2);
299 break;
300 default: break;
301 }
302 printf("\n");
303 printf(" Array Slot : %d (", __le32_to_cpu(sb->dev_number));
304 for (i= __le32_to_cpu(sb->max_dev); i> 0 ; i--)
305 if (__le16_to_cpu(sb->dev_roles[i-1]) != 0xffff)
306 break;
307 for (d=0; d < i; d++) {
308 int role = __le16_to_cpu(sb->dev_roles[d]);
309 if (d) printf(", ");
310 if (role == 0xffff) printf("empty");
311 else if(role == 0xfffe) printf("failed");
312 else printf("%d", role);
313 }
314 printf(")\n");
315 printf(" Array State : ");
316 for (d=0; d<__le32_to_cpu(sb->raid_disks); d++) {
317 int cnt = 0;
318 int me = 0;
319 int i;
320 for (i=0; i< __le32_to_cpu(sb->max_dev); i++) {
321 int role = __le16_to_cpu(sb->dev_roles[i]);
322 if (role == d) {
323 if (i == __le32_to_cpu(sb->dev_number))
324 me = 1;
325 cnt++;
326 }
327 }
328 if (cnt > 1) printf("?");
329 else if (cnt == 1 && me) printf("U");
330 else if (cnt == 1) printf("u");
331 else printf ("_");
332 }
333 faulty = 0;
334 for (i=0; i< __le32_to_cpu(sb->max_dev); i++) {
335 int role = __le16_to_cpu(sb->dev_roles[i]);
336 if (role == 0xFFFE)
337 faulty++;
338 }
339 if (faulty) printf(" %d failed", faulty);
340 printf("\n");
341 }
342
343
344 static void brief_examine_super1(struct supertype *st)
345 {
346 struct mdp_superblock_1 *sb = st->sb;
347 int i;
348 unsigned long long sb_offset;
349 char *nm;
350 char *c=map_num(pers, __le32_to_cpu(sb->level));
351
352 nm = strchr(sb->set_name, ':');
353 if (nm)
354 nm++;
355 else if (sb->set_name[0])
356 nm = sb->set_name;
357 else
358 nm = "??";
359
360 printf("ARRAY /dev/md/%s level=%s ", nm, c?c:"-unknown-");
361 sb_offset = __le64_to_cpu(sb->super_offset);
362 if (sb_offset <= 4)
363 printf("metadata=1.1 ");
364 else if (sb_offset <= 8)
365 printf("metadata=1.2 ");
366 else
367 printf("metadata=1.0 ");
368 printf("num-devices=%d UUID=", __le32_to_cpu(sb->raid_disks));
369 for (i=0; i<16; i++) {
370 if ((i&3)==0 && i != 0) printf(":");
371 printf("%02x", sb->set_uuid[i]);
372 }
373 if (sb->set_name[0])
374 printf(" name=%.32s", sb->set_name);
375 printf("\n");
376 }
377
378 static void export_examine_super1(struct supertype *st)
379 {
380 struct mdp_superblock_1 *sb = st->sb;
381 int i;
382 int len = 32;
383
384 printf("MD_LEVEL=%s\n", map_num(pers, __le32_to_cpu(sb->level)));
385 printf("MD_DEVICES=%d\n", __le32_to_cpu(sb->raid_disks));
386 for (i=0; i<32; i++)
387 if (sb->set_name[i] == '\n' ||
388 sb->set_name[i] == '\0') {
389 len = i;
390 break;
391 }
392 if (len)
393 printf("MD_NAME=%.*s\n", len, sb->set_name);
394 printf("MD_UUID=");
395 for (i=0; i<16; i++) {
396 if ((i&3)==0 && i != 0) printf(":");
397 printf("%02x", sb->set_uuid[i]);
398 }
399 printf("\n");
400 printf("MD_UPDATE_TIME=%llu\n",
401 __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL);
402 printf("MD_DEV_UUID=");
403 for (i=0; i<16; i++) {
404 if ((i&3)==0 && i != 0) printf(":");
405 printf("%02x", sb->device_uuid[i]);
406 }
407 printf("\n");
408 printf("MD_EVENTS=%llu\n",
409 (unsigned long long)__le64_to_cpu(sb->events));
410 }
411
412 static void detail_super1(struct supertype *st, char *homehost)
413 {
414 struct mdp_superblock_1 *sb = st->sb;
415 int i;
416 int l = homehost ? strlen(homehost) : 0;
417
418 printf(" Name : %.32s", sb->set_name);
419 if (l > 0 && l < 32 &&
420 sb->set_name[l] == ':' &&
421 strncmp(sb->set_name, homehost, l) == 0)
422 printf(" (local to host %s)", homehost);
423 printf("\n UUID : ");
424 for (i=0; i<16; i++) {
425 if ((i&3)==0 && i != 0) printf(":");
426 printf("%02x", sb->set_uuid[i]);
427 }
428 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb->events));
429 }
430
431 static void brief_detail_super1(struct supertype *st)
432 {
433 struct mdp_superblock_1 *sb = st->sb;
434 int i;
435
436 if (sb->set_name[0])
437 printf(" name=%.32s", sb->set_name);
438 printf(" UUID=");
439 for (i=0; i<16; i++) {
440 if ((i&3)==0 && i != 0) printf(":");
441 printf("%02x", sb->set_uuid[i]);
442 }
443 }
444
445 static void export_detail_super1(struct supertype *st)
446 {
447 struct mdp_superblock_1 *sb = st->sb;
448 int i;
449 int len = 32;
450
451 for (i=0; i<32; i++)
452 if (sb->set_name[i] == '\n' ||
453 sb->set_name[i] == '\0') {
454 len = i;
455 break;
456 }
457 if (len)
458 printf("MD_NAME=%.*s\n", len, sb->set_name);
459 printf("MD_UUID=");
460 for (i=0; i<16; i++) {
461 if ((i&3)==0 && i != 0) printf(":");
462 printf("%02x", sb->set_uuid[i]);
463 }
464 printf("\n");
465 }
466
467 #endif
468
469 static int match_home1(struct supertype *st, char *homehost)
470 {
471 struct mdp_superblock_1 *sb = st->sb;
472 int l = homehost ? strlen(homehost) : 0;
473
474 return (l > 0 && l < 32 &&
475 sb->set_name[l] == ':' &&
476 strncmp(sb->set_name, homehost, l) == 0);
477 }
478
479 static void uuid_from_super1(struct supertype *st, int uuid[4])
480 {
481 struct mdp_superblock_1 *super = st->sb;
482 char *cuuid = (char*)uuid;
483 int i;
484 for (i=0; i<16; i++)
485 cuuid[i] = super->set_uuid[i];
486 }
487
488 static void getinfo_super1(struct supertype *st, struct mdinfo *info)
489 {
490 struct mdp_superblock_1 *sb = st->sb;
491 int working = 0;
492 int i;
493 int role;
494
495 info->array.major_version = 1;
496 info->array.minor_version = st->minor_version;
497 info->array.patch_version = 0;
498 info->array.raid_disks = __le32_to_cpu(sb->raid_disks);
499 info->array.level = __le32_to_cpu(sb->level);
500 info->array.layout = __le32_to_cpu(sb->layout);
501 info->array.md_minor = -1;
502 info->array.ctime = __le64_to_cpu(sb->ctime);
503 info->array.utime = __le64_to_cpu(sb->utime);
504 info->array.chunk_size = __le32_to_cpu(sb->chunksize)*512;
505 info->array.state =
506 (__le64_to_cpu(sb->resync_offset) >= __le64_to_cpu(sb->size))
507 ? 1 : 0;
508
509 info->data_offset = __le64_to_cpu(sb->data_offset);
510 info->component_size = __le64_to_cpu(sb->size);
511
512 info->disk.major = 0;
513 info->disk.minor = 0;
514 info->disk.number = __le32_to_cpu(sb->dev_number);
515 if (__le32_to_cpu(sb->dev_number) >= __le32_to_cpu(sb->max_dev) ||
516 __le32_to_cpu(sb->max_dev) > 512)
517 role = 0xfffe;
518 else
519 role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]);
520
521 info->disk.raid_disk = -1;
522 switch(role) {
523 case 0xFFFF:
524 info->disk.state = 2; /* spare: ACTIVE, not sync, not faulty */
525 break;
526 case 0xFFFE:
527 info->disk.state = 1; /* faulty */
528 break;
529 default:
530 info->disk.state = 6; /* active and in sync */
531 info->disk.raid_disk = role;
532 }
533 info->events = __le64_to_cpu(sb->events);
534 sprintf(info->text_version, "1.%d", st->minor_version);
535
536 memcpy(info->uuid, sb->set_uuid, 16);
537
538 strncpy(info->name, sb->set_name, 32);
539 info->name[32] = 0;
540
541 if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) {
542 info->reshape_active = 1;
543 info->reshape_progress = __le64_to_cpu(sb->reshape_position);
544 info->new_level = __le32_to_cpu(sb->new_level);
545 info->delta_disks = __le32_to_cpu(sb->delta_disks);
546 info->new_layout = __le32_to_cpu(sb->new_layout);
547 info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9;
548 } else
549 info->reshape_active = 0;
550
551 for (i=0; i< __le32_to_cpu(sb->max_dev); i++) {
552 role = __le16_to_cpu(sb->dev_roles[i]);
553 if (/*role == 0xFFFF || */role < info->array.raid_disks)
554 working++;
555 }
556
557 info->array.working_disks = working;
558 }
559
560 static int update_super1(struct supertype *st, struct mdinfo *info,
561 char *update,
562 char *devname, int verbose,
563 int uuid_set, char *homehost)
564 {
565 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
566 * For others, the return value is ignored.
567 */
568 int rv = 0;
569 struct mdp_superblock_1 *sb = st->sb;
570
571 if (strcmp(update, "force-one")==0) {
572 /* Not enough devices for a working array,
573 * so bring this one up-to-date
574 */
575 if (sb->events != __cpu_to_le64(info->events))
576 rv = 1;
577 sb->events = __cpu_to_le64(info->events);
578 }
579 if (strcmp(update, "force-array")==0) {
580 /* Degraded array and 'force' requests to
581 * maybe need to mark it 'clean'.
582 */
583 switch(__le32_to_cpu(sb->level)) {
584 case 5: case 4: case 6:
585 /* need to force clean */
586 if (sb->resync_offset != ~0ULL)
587 rv = 1;
588 sb->resync_offset = ~0ULL;
589 }
590 }
591 if (strcmp(update, "assemble")==0) {
592 int d = info->disk.number;
593 int want;
594 if (info->disk.state == 6)
595 want = __cpu_to_le32(info->disk.raid_disk);
596 else
597 want = 0xFFFF;
598 if (sb->dev_roles[d] != want) {
599 sb->dev_roles[d] = want;
600 rv = 1;
601 }
602 }
603 if (strcmp(update, "linear-grow-new") == 0) {
604 int i;
605 int rfd;
606 int max = __le32_to_cpu(sb->max_dev);
607
608 for (i=0 ; i < max ; i++)
609 if (__le16_to_cpu(sb->dev_roles[i]) >= 0xfffe)
610 break;
611 sb->dev_number = __cpu_to_le32(i);
612 info->disk.number = i;
613 if (max >= __le32_to_cpu(sb->max_dev))
614 sb->max_dev = __cpu_to_le32(max+1);
615
616 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
617 read(rfd, sb->device_uuid, 16) != 16) {
618 *(__u32*)(sb->device_uuid) = random();
619 *(__u32*)(sb->device_uuid+4) = random();
620 *(__u32*)(sb->device_uuid+8) = random();
621 *(__u32*)(sb->device_uuid+12) = random();
622 }
623
624 sb->dev_roles[i] =
625 __cpu_to_le16(info->disk.raid_disk);
626 }
627 if (strcmp(update, "linear-grow-update") == 0) {
628 sb->raid_disks = __cpu_to_le32(info->array.raid_disks);
629 sb->dev_roles[info->disk.number] =
630 __cpu_to_le16(info->disk.raid_disk);
631 }
632 if (strcmp(update, "resync") == 0) {
633 /* make sure resync happens */
634 sb->resync_offset = 0ULL;
635 }
636 if (strcmp(update, "uuid") == 0) {
637 copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid);
638
639 if (__le32_to_cpu(sb->feature_map)&MD_FEATURE_BITMAP_OFFSET) {
640 struct bitmap_super_s *bm;
641 bm = (struct bitmap_super_s*)(st->sb+1024);
642 memcpy(bm->uuid, sb->set_uuid, 16);
643 }
644 }
645 if (strcmp(update, "homehost") == 0 &&
646 homehost) {
647 char *c;
648 update = "name";
649 c = strchr(sb->set_name, ':');
650 if (c)
651 strncpy(info->name, c+1, 31 - (c-sb->set_name));
652 else
653 strncpy(info->name, sb->set_name, 32);
654 info->name[32] = 0;
655 }
656 if (strcmp(update, "name") == 0) {
657 if (info->name[0] == 0)
658 sprintf(info->name, "%d", info->array.md_minor);
659 memset(sb->set_name, 0, sizeof(sb->set_name));
660 if (homehost &&
661 strchr(info->name, ':') == NULL &&
662 strlen(homehost)+1+strlen(info->name) < 32) {
663 strcpy(sb->set_name, homehost);
664 strcat(sb->set_name, ":");
665 strcat(sb->set_name, info->name);
666 } else
667 strcpy(sb->set_name, info->name);
668 }
669 if (strcmp(update, "devicesize") == 0 &&
670 __le64_to_cpu(sb->super_offset) <
671 __le64_to_cpu(sb->data_offset)) {
672 /* set data_size to device size less data_offset */
673 struct misc_dev_info *misc = (struct misc_dev_info*)
674 (st->sb + 1024 + sizeof(struct bitmap_super_s));
675 printf("Size was %llu\n", (unsigned long long)
676 __le64_to_cpu(sb->data_size));
677 sb->data_size = __cpu_to_le64(
678 misc->device_size - __le64_to_cpu(sb->data_offset));
679 printf("Size is %llu\n", (unsigned long long)
680 __le64_to_cpu(sb->data_size));
681 }
682 if (strcmp(update, "_reshape_progress")==0)
683 sb->reshape_position = __cpu_to_le64(info->reshape_progress);
684
685 sb->sb_csum = calc_sb_1_csum(sb);
686 return rv;
687 }
688
689 static int init_super1(struct supertype *st, mdu_array_info_t *info,
690 unsigned long long size, char *name, char *homehost, int *uuid)
691 {
692 struct mdp_superblock_1 *sb = malloc(1024 + sizeof(bitmap_super_t) +
693 sizeof(struct misc_dev_info));
694 int spares;
695 int rfd;
696 char defname[10];
697 memset(sb, 0, 1024);
698
699 st->sb = sb;
700 if (info->major_version == -1) {
701 /* zeroing superblock */
702 return 0;
703 }
704
705 spares = info->working_disks - info->active_disks;
706 if (info->raid_disks + spares > 384) {
707 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
708 info->raid_disks , spares, 384);
709 return 0;
710 }
711
712 sb->magic = __cpu_to_le32(MD_SB_MAGIC);
713 sb->major_version = __cpu_to_le32(1);
714 sb->feature_map = 0;
715 sb->pad0 = 0;
716
717 if (uuid)
718 copy_uuid(sb->set_uuid, uuid, super1.swapuuid);
719 else {
720 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
721 read(rfd, sb->set_uuid, 16) != 16) {
722 *(__u32*)(sb->set_uuid) = random();
723 *(__u32*)(sb->set_uuid+4) = random();
724 *(__u32*)(sb->set_uuid+8) = random();
725 *(__u32*)(sb->set_uuid+12) = random();
726 }
727 if (rfd >= 0) close(rfd);
728 }
729
730 if (name == NULL || *name == 0) {
731 sprintf(defname, "%d", info->md_minor);
732 name = defname;
733 }
734 memset(sb->set_name, 0, 32);
735 if (homehost &&
736 strchr(name, ':')== NULL &&
737 strlen(homehost)+1+strlen(name) < 32) {
738 strcpy(sb->set_name, homehost);
739 strcat(sb->set_name, ":");
740 strcat(sb->set_name, name);
741 } else
742 strcpy(sb->set_name, name);
743
744 sb->ctime = __cpu_to_le64((unsigned long long)time(0));
745 sb->level = __cpu_to_le32(info->level);
746 sb->layout = __cpu_to_le32(info->layout);
747 sb->size = __cpu_to_le64(size*2ULL);
748 sb->chunksize = __cpu_to_le32(info->chunk_size>>9);
749 sb->raid_disks = __cpu_to_le32(info->raid_disks);
750
751 sb->data_offset = __cpu_to_le64(0);
752 sb->data_size = __cpu_to_le64(0);
753 sb->super_offset = __cpu_to_le64(0);
754 sb->recovery_offset = __cpu_to_le64(0);
755
756 sb->utime = sb->ctime;
757 sb->events = __cpu_to_le64(1);
758 if (info->state & (1<<MD_SB_CLEAN))
759 sb->resync_offset = ~0ULL;
760 else
761 sb->resync_offset = 0;
762 sb->max_dev = __cpu_to_le32((1024- sizeof(struct mdp_superblock_1))/
763 sizeof(sb->dev_roles[0]));
764 memset(sb->pad3, 0, sizeof(sb->pad3));
765
766 memset(sb->dev_roles, 0xff, 1024 - sizeof(struct mdp_superblock_1));
767
768 return 1;
769 }
770
771 struct devinfo {
772 int fd;
773 char *devname;
774 mdu_disk_info_t disk;
775 struct devinfo *next;
776 };
777 /* Add a device to the superblock being created */
778 static void add_to_super1(struct supertype *st, mdu_disk_info_t *dk,
779 int fd, char *devname)
780 {
781 struct mdp_superblock_1 *sb = st->sb;
782 __u16 *rp = sb->dev_roles + dk->number;
783 struct devinfo *di, **dip;
784
785 if ((dk->state & 6) == 6) /* active, sync */
786 *rp = __cpu_to_le16(dk->raid_disk);
787 else if ((dk->state & ~2) == 0) /* active or idle -> spare */
788 *rp = 0xffff;
789 else
790 *rp = 0xfffe;
791
792 dip = (struct devinfo **)&st->info;
793 while (*dip)
794 dip = &(*dip)->next;
795 di = malloc(sizeof(struct devinfo));
796 di->fd = fd;
797 di->devname = devname;
798 di->disk = *dk;
799 di->next = NULL;
800 *dip = di;
801 }
802
803 static void locate_bitmap1(struct supertype *st, int fd);
804
805 static int store_super1(struct supertype *st, int fd)
806 {
807 struct mdp_superblock_1 *sb = st->sb;
808 unsigned long long sb_offset;
809 int sbsize;
810 unsigned long long dsize;
811
812 if (!get_dev_size(fd, NULL, &dsize))
813 return 1;
814
815 dsize >>= 9;
816
817 if (dsize < 24)
818 return 2;
819
820 /*
821 * Calculate the position of the superblock.
822 * It is always aligned to a 4K boundary and
823 * depending on minor_version, it can be:
824 * 0: At least 8K, but less than 12K, from end of device
825 * 1: At start of device
826 * 2: 4K from start of device.
827 */
828 switch(st->minor_version) {
829 case 0:
830 sb_offset = dsize;
831 sb_offset -= 8*2;
832 sb_offset &= ~(4*2-1);
833 break;
834 case 1:
835 sb_offset = 0;
836 break;
837 case 2:
838 sb_offset = 4*2;
839 break;
840 default:
841 return -EINVAL;
842 }
843
844
845
846 if (sb_offset != __le64_to_cpu(sb->super_offset) &&
847 0 != __le64_to_cpu(sb->super_offset)
848 ) {
849 fprintf(stderr, Name ": internal error - sb_offset is wrong\n");
850 abort();
851 }
852
853 if (lseek64(fd, sb_offset << 9, 0)< 0LL)
854 return 3;
855
856 sbsize = sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev);
857
858 if (write(fd, sb, sbsize) != sbsize)
859 return 4;
860
861 if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) {
862 struct bitmap_super_s *bm = (struct bitmap_super_s*)
863 (((char*)sb)+1024);
864 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) {
865 locate_bitmap1(st, fd);
866 if (write(fd, bm, sizeof(*bm)) != sizeof(*bm))
867 return 5;
868 }
869 }
870 fsync(fd);
871 return 0;
872 }
873
874 static int load_super1(struct supertype *st, int fd, char *devname);
875
876 static unsigned long choose_bm_space(unsigned long devsize)
877 {
878 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
879 * if bigger than 200Gig, save 128k
880 */
881 if (devsize < 64*2) return 0;
882 if (devsize - 64*2 >= 200*1024*1024*2)
883 return 128*2;
884 if (devsize - 4*2 > 8*1024*1024*2)
885 return 64*2;
886 return 4*2;
887 }
888
889 #ifndef MDASSEMBLE
890 static int write_init_super1(struct supertype *st)
891 {
892 struct mdp_superblock_1 *sb = st->sb;
893 struct supertype refst;
894 int rfd;
895 int rv = 0;
896 int bm_space;
897 struct devinfo *di;
898 unsigned long long dsize, array_size;
899 long long sb_offset;
900
901 for (di = st->info; di && ! rv ; di = di->next) {
902 if (di->disk.state == 1)
903 continue;
904 if (di->fd < 0)
905 continue;
906
907 Kill(di->devname, 0, 1, 1);
908 Kill(di->devname, 0, 1, 1);
909
910 sb->dev_number = __cpu_to_le32(di->disk.number);
911 if (di->disk.state & (1<<MD_DISK_WRITEMOSTLY))
912 sb->devflags |= __cpu_to_le32(WriteMostly1);
913
914 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
915 read(rfd, sb->device_uuid, 16) != 16) {
916 *(__u32*)(sb->device_uuid) = random();
917 *(__u32*)(sb->device_uuid+4) = random();
918 *(__u32*)(sb->device_uuid+8) = random();
919 *(__u32*)(sb->device_uuid+12) = random();
920 }
921 if (rfd >= 0) close(rfd);
922 sb->events = 0;
923
924 refst =*st;
925 refst.sb = NULL;
926 if (load_super1(&refst, di->fd, NULL)==0) {
927 struct mdp_superblock_1 *refsb = refst.sb;
928
929 memcpy(sb->device_uuid, refsb->device_uuid, 16);
930 if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) {
931 /* same array, so preserve events and
932 * dev_number */
933 sb->events = refsb->events;
934 /* bugs in 2.6.17 and earlier mean the
935 * dev_number chosen in Manage must be preserved
936 */
937 if (get_linux_version() >= 2006018)
938 sb->dev_number = refsb->dev_number;
939 }
940 free(refsb);
941 }
942
943 if (!get_dev_size(di->fd, NULL, &dsize))
944 return 1;
945 dsize >>= 9;
946
947 if (dsize < 24) {
948 close(di->fd);
949 return 2;
950 }
951
952
953 /*
954 * Calculate the position of the superblock.
955 * It is always aligned to a 4K boundary and
956 * depending on minor_version, it can be:
957 * 0: At least 8K, but less than 12K, from end of device
958 * 1: At start of device
959 * 2: 4K from start of device.
960 * Depending on the array size, we might leave extra space
961 * for a bitmap.
962 */
963 array_size = __le64_to_cpu(sb->size);
964 /* work out how much space we left for a bitmap */
965 bm_space = choose_bm_space(array_size);
966
967 switch(st->minor_version) {
968 case 0:
969 sb_offset = dsize;
970 sb_offset -= 8*2;
971 sb_offset &= ~(4*2-1);
972 sb->super_offset = __cpu_to_le64(sb_offset);
973 sb->data_offset = __cpu_to_le64(0);
974 if (sb_offset - bm_space < array_size)
975 bm_space = sb_offset - array_size;
976 sb->data_size = __cpu_to_le64(sb_offset - bm_space);
977 break;
978 case 1:
979 sb->super_offset = __cpu_to_le64(0);
980 if (4*2 + bm_space + __le64_to_cpu(sb->size) > dsize)
981 bm_space = dsize - __le64_to_cpu(sb->size) -4*2;
982 sb->data_offset = __cpu_to_le64(bm_space + 4*2);
983 sb->data_size = __cpu_to_le64(dsize - bm_space - 4*2);
984 break;
985 case 2:
986 sb_offset = 4*2;
987 sb->super_offset = __cpu_to_le64(4*2);
988 if (4*2 + 4*2 + bm_space + __le64_to_cpu(sb->size)
989 > dsize)
990 bm_space = dsize - __le64_to_cpu(sb->size)
991 - 4*2 - 4*2;
992 sb->data_offset = __cpu_to_le64(4*2 + 4*2 + bm_space);
993 sb->data_size = __cpu_to_le64(dsize - 4*2 - 4*2
994 - bm_space );
995 break;
996 default:
997 return -EINVAL;
998 }
999
1000
1001 sb->sb_csum = calc_sb_1_csum(sb);
1002 rv = store_super1(st, di->fd);
1003 if (rv)
1004 fprintf(stderr,
1005 Name ": failed to write superblock to %s\n",
1006 di->devname);
1007
1008 if (rv == 0 && (__le32_to_cpu(sb->feature_map) & 1))
1009 rv = st->ss->write_bitmap(st, di->fd);
1010 close(di->fd);
1011 di->fd = -1;
1012 }
1013 return rv;
1014 }
1015 #endif
1016
1017 static int compare_super1(struct supertype *st, struct supertype *tst)
1018 {
1019 /*
1020 * return:
1021 * 0 same, or first was empty, and second was copied
1022 * 1 second had wrong number
1023 * 2 wrong uuid
1024 * 3 wrong other info
1025 */
1026 struct mdp_superblock_1 *first = st->sb;
1027 struct mdp_superblock_1 *second = tst->sb;
1028
1029 if (second->magic != __cpu_to_le32(MD_SB_MAGIC))
1030 return 1;
1031 if (second->major_version != __cpu_to_le32(1))
1032 return 1;
1033
1034 if (!first) {
1035 first = malloc(1024+sizeof(bitmap_super_t) +
1036 sizeof(struct misc_dev_info));
1037 memcpy(first, second, 1024+sizeof(bitmap_super_t) +
1038 sizeof(struct misc_dev_info));
1039 st->sb = first;
1040 return 0;
1041 }
1042 if (memcmp(first->set_uuid, second->set_uuid, 16)!= 0)
1043 return 2;
1044
1045 if (first->ctime != second->ctime ||
1046 first->level != second->level ||
1047 first->layout != second->layout ||
1048 first->size != second->size ||
1049 first->chunksize != second->chunksize ||
1050 first->raid_disks != second->raid_disks)
1051 return 3;
1052 return 0;
1053 }
1054
1055 static void free_super1(struct supertype *st);
1056
1057 static int load_super1(struct supertype *st, int fd, char *devname)
1058 {
1059 unsigned long long dsize;
1060 unsigned long long sb_offset;
1061 struct mdp_superblock_1 *super;
1062 int uuid[4];
1063 struct bitmap_super_s *bsb;
1064 struct misc_dev_info *misc;
1065
1066 free_super1(st);
1067
1068 if (st->ss == NULL || st->minor_version == -1) {
1069 int bestvers = -1;
1070 struct supertype tst;
1071 __u64 bestctime = 0;
1072 /* guess... choose latest ctime */
1073 memset(&tst, 0, sizeof(tst));
1074 tst.ss = &super1;
1075 for (tst.minor_version = 0; tst.minor_version <= 2 ; tst.minor_version++) {
1076 switch(load_super1(&tst, fd, devname)) {
1077 case 0: super = tst.sb;
1078 if (bestvers == -1 ||
1079 bestctime < __le64_to_cpu(super->ctime)) {
1080 bestvers = tst.minor_version;
1081 bestctime = __le64_to_cpu(super->ctime);
1082 }
1083 free(super);
1084 tst.sb = NULL;
1085 break;
1086 case 1: return 1; /*bad device */
1087 case 2: break; /* bad, try next */
1088 }
1089 }
1090 if (bestvers != -1) {
1091 int rv;
1092 tst.minor_version = bestvers;
1093 tst.ss = &super1;
1094 tst.max_devs = 384;
1095 rv = load_super1(&tst, fd, devname);
1096 if (rv == 0)
1097 *st = tst;
1098 return rv;
1099 }
1100 return 2;
1101 }
1102 if (!get_dev_size(fd, devname, &dsize))
1103 return 1;
1104 dsize >>= 9;
1105
1106 if (dsize < 24) {
1107 if (devname)
1108 fprintf(stderr, Name ": %s is too small for md: size is %llu sectors.\n",
1109 devname, dsize);
1110 return 1;
1111 }
1112
1113 /*
1114 * Calculate the position of the superblock.
1115 * It is always aligned to a 4K boundary and
1116 * depending on minor_version, it can be:
1117 * 0: At least 8K, but less than 12K, from end of device
1118 * 1: At start of device
1119 * 2: 4K from start of device.
1120 */
1121 switch(st->minor_version) {
1122 case 0:
1123 sb_offset = dsize;
1124 sb_offset -= 8*2;
1125 sb_offset &= ~(4*2-1);
1126 break;
1127 case 1:
1128 sb_offset = 0;
1129 break;
1130 case 2:
1131 sb_offset = 4*2;
1132 break;
1133 default:
1134 return -EINVAL;
1135 }
1136
1137 ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
1138
1139
1140 if (lseek64(fd, sb_offset << 9, 0)< 0LL) {
1141 if (devname)
1142 fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
1143 devname, strerror(errno));
1144 return 1;
1145 }
1146
1147 super = malloc(1024 + sizeof(bitmap_super_t) +
1148 sizeof(struct misc_dev_info));
1149
1150 if (read(fd, super, 1024) != 1024) {
1151 if (devname)
1152 fprintf(stderr, Name ": Cannot read superblock on %s\n",
1153 devname);
1154 free(super);
1155 return 1;
1156 }
1157
1158 if (__le32_to_cpu(super->magic) != MD_SB_MAGIC) {
1159 if (devname)
1160 fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
1161 devname, MD_SB_MAGIC, __le32_to_cpu(super->magic));
1162 free(super);
1163 return 2;
1164 }
1165
1166 if (__le32_to_cpu(super->major_version) != 1) {
1167 if (devname)
1168 fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
1169 devname, __le32_to_cpu(super->major_version));
1170 free(super);
1171 return 2;
1172 }
1173 if (__le64_to_cpu(super->super_offset) != sb_offset) {
1174 if (devname)
1175 fprintf(stderr, Name ": No superblock found on %s (super_offset is wrong)\n",
1176 devname);
1177 free(super);
1178 return 2;
1179 }
1180 st->sb = super;
1181
1182 bsb = (struct bitmap_super_s *)(((char*)super)+1024);
1183
1184 misc = (struct misc_dev_info*) (bsb+1);
1185 misc->device_size = dsize;
1186
1187 /* Now check on the bitmap superblock */
1188 if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0)
1189 return 0;
1190 /* Read the bitmap superblock and make sure it looks
1191 * valid. If it doesn't clear the bit. An --assemble --force
1192 * should get that written out.
1193 */
1194 locate_bitmap1(st, fd);
1195 if (read(fd, ((char*)super)+1024, sizeof(struct bitmap_super_s))
1196 != sizeof(struct bitmap_super_s))
1197 goto no_bitmap;
1198
1199 uuid_from_super1(st, uuid);
1200 if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
1201 memcmp(bsb->uuid, uuid, 16) != 0)
1202 goto no_bitmap;
1203 return 0;
1204
1205 no_bitmap:
1206 super->feature_map = __cpu_to_le32(__le32_to_cpu(super->feature_map) & ~1);
1207 return 0;
1208 }
1209
1210
1211 static struct supertype *match_metadata_desc1(char *arg)
1212 {
1213 struct supertype *st = malloc(sizeof(*st));
1214 if (!st) return st;
1215
1216 memset(st, 0, sizeof(*st));
1217 st->ss = &super1;
1218 st->max_devs = 384;
1219 st->sb = NULL;
1220 if (strcmp(arg, "1.0") == 0) {
1221 st->minor_version = 0;
1222 return st;
1223 }
1224 if (strcmp(arg, "1.1") == 0) {
1225 st->minor_version = 1;
1226 return st;
1227 }
1228 if (strcmp(arg, "1.2") == 0) {
1229 st->minor_version = 2;
1230 return st;
1231 }
1232 if (strcmp(arg, "1") == 0 ||
1233 strcmp(arg, "default") == 0) {
1234 st->minor_version = -1;
1235 return st;
1236 }
1237
1238 free(st);
1239 return NULL;
1240 }
1241
1242 /* find available size on device with this devsize, using
1243 * superblock type st, and reserving 'reserve' sectors for
1244 * a possible bitmap
1245 */
1246 static __u64 avail_size1(struct supertype *st, __u64 devsize)
1247 {
1248 if (devsize < 24)
1249 return 0;
1250
1251 devsize -= choose_bm_space(devsize);
1252
1253 switch(st->minor_version) {
1254 case -1: /* no specified. Now time to set default */
1255 st->minor_version = 0;
1256 /* FALL THROUGH */
1257 case 0:
1258 /* at end */
1259 return ((devsize - 8*2 ) & ~(4*2-1));
1260 case 1:
1261 /* at start, 4K for superblock and possible bitmap */
1262 return devsize - 4*2;
1263 case 2:
1264 /* 4k from start, 4K for superblock and possible bitmap */
1265 return devsize - (4+4)*2;
1266 }
1267 return 0;
1268 }
1269
1270 static int
1271 add_internal_bitmap1(struct supertype *st,
1272 int *chunkp, int delay, int write_behind,
1273 unsigned long long size,
1274 int may_change, int major)
1275 {
1276 /*
1277 * If not may_change, then this is a 'Grow', and the bitmap
1278 * must fit after the superblock.
1279 * If may_change, then this is create, and we can put the bitmap
1280 * before the superblock if we like, or may move the start.
1281 * If !may_change, the bitmap MUST live at offset of 1K, until
1282 * we get a sysfs interface.
1283 *
1284 * size is in sectors, chunk is in bytes !!!
1285 */
1286
1287 unsigned long long bits;
1288 unsigned long long max_bits;
1289 unsigned long long min_chunk;
1290 long offset;
1291 int chunk = *chunkp;
1292 int room = 0;
1293 struct mdp_superblock_1 *sb = st->sb;
1294 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + 1024);
1295
1296 switch(st->minor_version) {
1297 case 0:
1298 /* either 3K after the superblock, or some amount of space
1299 * before.
1300 */
1301 if (may_change) {
1302 /* We are creating array, so we *know* how much room has
1303 * been left.
1304 */
1305 offset = 0;
1306 room = choose_bm_space(__le64_to_cpu(sb->size));
1307 if (room == 4*2) {
1308 /* make it 3K after the superblock */
1309 room = 3*2;
1310 offset = 2;
1311 }
1312 } else {
1313 room = __le64_to_cpu(sb->super_offset)
1314 - __le64_to_cpu(sb->data_offset)
1315 - __le64_to_cpu(sb->data_size);
1316 /* remove '1 ||' when we can set offset via sysfs */
1317 if (1 || (room < 3*2 &&
1318 __le32_to_cpu(sb->max_dev) <= 384)) {
1319 room = 3*2;
1320 offset = 1*2;
1321 } else {
1322 offset = 0; /* means movable offset */
1323 }
1324 }
1325 break;
1326 case 1:
1327 case 2: /* between superblock and data */
1328 if (may_change) {
1329 offset = 4*2;
1330 room = choose_bm_space(__le64_to_cpu(sb->size));
1331 } else {
1332 room = __le64_to_cpu(sb->data_offset)
1333 - __le64_to_cpu(sb->super_offset);
1334 if (1 || __le32_to_cpu(sb->max_dev) <= 384) {
1335 room -= 2;
1336 offset = 2;
1337 } else {
1338 room -= 4*2;
1339 offset = 4*2;
1340 }
1341 }
1342 break;
1343 default:
1344 return 0;
1345 }
1346
1347 if (chunk == UnSet && room > 128*2)
1348 /* Limit to 128K of bitmap when chunk size not requested */
1349 room = 128*2;
1350
1351 max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8;
1352
1353 min_chunk = 4096; /* sub-page chunks don't work yet.. */
1354 bits = (size*512)/min_chunk +1;
1355 while (bits > max_bits) {
1356 min_chunk *= 2;
1357 bits = (bits+1)/2;
1358 }
1359 if (chunk == UnSet)
1360 chunk = min_chunk;
1361 else if (chunk < min_chunk)
1362 return 0; /* chunk size too small */
1363 if (chunk == 0) /* rounding problem */
1364 return 0;
1365
1366 if (offset == 0) {
1367 bits = (size*512) / chunk + 1;
1368 room = ((bits+7)/8 + sizeof(bitmap_super_t) +511)/512;
1369 offset = -room;
1370 }
1371
1372 sb->bitmap_offset = __cpu_to_le32(offset);
1373
1374 sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | 1);
1375 memset(bms, 0, sizeof(*bms));
1376 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
1377 bms->version = __cpu_to_le32(major);
1378 uuid_from_super1(st, (int*)bms->uuid);
1379 bms->chunksize = __cpu_to_le32(chunk);
1380 bms->daemon_sleep = __cpu_to_le32(delay);
1381 bms->sync_size = __cpu_to_le64(size);
1382 bms->write_behind = __cpu_to_le32(write_behind);
1383
1384 *chunkp = chunk;
1385 return 1;
1386 }
1387
1388
1389 static void locate_bitmap1(struct supertype *st, int fd)
1390 {
1391 unsigned long long offset;
1392 struct mdp_superblock_1 *sb;
1393 int mustfree = 0;
1394
1395 if (!st->sb) {
1396 if (st->ss->load_super(st, fd, NULL))
1397 return; /* no error I hope... */
1398 mustfree = 1;
1399 }
1400 sb = st->sb;
1401
1402 offset = __le64_to_cpu(sb->super_offset);
1403 offset += (int32_t) __le32_to_cpu(sb->bitmap_offset);
1404 if (mustfree)
1405 free(sb);
1406 lseek64(fd, offset<<9, 0);
1407 }
1408
1409 static int write_bitmap1(struct supertype *st, int fd)
1410 {
1411 struct mdp_superblock_1 *sb = st->sb;
1412 bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+1024);
1413 int rv = 0;
1414
1415 int towrite, n;
1416 char buf[4096];
1417
1418 locate_bitmap1(st, fd);
1419
1420 if (write(fd, ((char*)sb)+1024, sizeof(bitmap_super_t)) !=
1421 sizeof(bitmap_super_t))
1422 return -2;
1423 towrite = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9);
1424 towrite = (towrite+7) >> 3; /* bits to bytes */
1425 memset(buf, 0xff, sizeof(buf));
1426 while (towrite > 0) {
1427 n = towrite;
1428 if (n > sizeof(buf))
1429 n = sizeof(buf);
1430 n = write(fd, buf, n);
1431 if (n > 0)
1432 towrite -= n;
1433 else
1434 break;
1435 }
1436 fsync(fd);
1437 if (towrite)
1438 rv = -2;
1439
1440 return rv;
1441 }
1442
1443 static void free_super1(struct supertype *st)
1444 {
1445 if (st->sb)
1446 free(st->sb);
1447 st->sb = NULL;
1448 }
1449
1450 static int validate_geometry1(struct supertype *st, int level,
1451 int layout, int raiddisks,
1452 int chunk, unsigned long long size,
1453 char *subdev, unsigned long long *freesize)
1454 {
1455 unsigned long long ldsize;
1456 int fd;
1457
1458 if (level == LEVEL_CONTAINER)
1459 return 0;
1460 if (!subdev)
1461 return 1;
1462
1463 fd = open(subdev, O_RDONLY|O_EXCL, 0);
1464 if (fd < 0) {
1465 fprintf(stderr, Name ": Cannot open %s: %s\n",
1466 subdev, strerror(errno));
1467 return 0;
1468 }
1469 if (!get_dev_size(fd, subdev, &ldsize)) {
1470 close(fd);
1471 return 0;
1472 }
1473 close(fd);
1474
1475 *freesize = avail_size1(st, ldsize >> 9);
1476 return 1;
1477 }
1478
1479 struct superswitch super1 = {
1480 #ifndef MDASSEMBLE
1481 .examine_super = examine_super1,
1482 .brief_examine_super = brief_examine_super1,
1483 .export_examine_super = export_examine_super1,
1484 .detail_super = detail_super1,
1485 .brief_detail_super = brief_detail_super1,
1486 .export_detail_super = export_detail_super1,
1487 .write_init_super = write_init_super1,
1488 #endif
1489 .match_home = match_home1,
1490 .uuid_from_super = uuid_from_super1,
1491 .getinfo_super = getinfo_super1,
1492 .update_super = update_super1,
1493 .init_super = init_super1,
1494 .add_to_super = add_to_super1,
1495 .store_super = store_super1,
1496 .compare_super = compare_super1,
1497 .load_super = load_super1,
1498 .match_metadata_desc = match_metadata_desc1,
1499 .avail_size = avail_size1,
1500 .add_internal_bitmap = add_internal_bitmap1,
1501 .locate_bitmap = locate_bitmap1,
1502 .write_bitmap = write_bitmap1,
1503 .free_super = free_super1,
1504 .validate_geometry = validate_geometry1,
1505 #if __BYTE_ORDER == BIG_ENDIAN
1506 .swapuuid = 0,
1507 #else
1508 .swapuuid = 1,
1509 #endif
1510 };
Unnamed repository; edit this file 'description' to name the repository.