]>
Commit | Line | Data |
---|---|---|
82d9eba6 NB |
1 | /* |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
e736b623 | 4 | * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de> |
82d9eba6 NB |
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 | |
e736b623 | 22 | * Email: <neilb@suse.de> |
82d9eba6 NB |
23 | */ |
24 | ||
04f903b2 | 25 | #include <stddef.h> |
82d9eba6 | 26 | #include "mdadm.h" |
82d9eba6 NB |
27 | /* |
28 | * The version-1 superblock : | |
29 | * All numeric fields are little-endian. | |
30 | * | |
31 | * total size: 256 bytes plus 2 per device. | |
32 | * 1K allows 384 devices. | |
33 | */ | |
34 | struct mdp_superblock_1 { | |
35 | /* constant array information - 128 bytes */ | |
36 | __u32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */ | |
37 | __u32 major_version; /* 1 */ | |
38 | __u32 feature_map; /* 0 for now */ | |
39 | __u32 pad0; /* always set to 0 when writing */ | |
40 | ||
41 | __u8 set_uuid[16]; /* user-space generated. */ | |
42 | char set_name[32]; /* set and interpreted by user-space */ | |
43 | ||
44 | __u64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/ | |
45 | __u32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */ | |
46 | __u32 layout; /* only for raid5 currently */ | |
47 | __u64 size; /* used size of component devices, in 512byte sectors */ | |
48 | ||
49 | __u32 chunksize; /* in 512byte sectors */ | |
50 | __u32 raid_disks; | |
34163fc7 NB |
51 | __u32 bitmap_offset; /* sectors after start of superblock that bitmap starts |
52 | * NOTE: signed, so bitmap can be before superblock | |
53 | * only meaningful of feature_map[0] is set. | |
54 | */ | |
b674b5b8 NB |
55 | |
56 | /* These are only valid with feature bit '4' */ | |
b674b5b8 | 57 | __u32 new_level; /* new level we are reshaping to */ |
ac957baa | 58 | __u64 reshape_position; /* next address in array-space for reshape */ |
b674b5b8 NB |
59 | __u32 delta_disks; /* change in number of raid_disks */ |
60 | __u32 new_layout; /* new layout */ | |
a2836f12 | 61 | __u32 new_chunk; /* new chunk size (sectors) */ |
8fe1c44f N |
62 | __u32 new_offset; /* signed number to add to data_offset in new |
63 | * layout. 0 == no-change. This can be | |
64 | * different on each device in the array. | |
65 | */ | |
82d9eba6 NB |
66 | |
67 | /* constant this-device information - 64 bytes */ | |
68 | __u64 data_offset; /* sector start of data, often 0 */ | |
69 | __u64 data_size; /* sectors in this device that can be used for data */ | |
70 | __u64 super_offset; /* sector start of this superblock */ | |
cc1799c3 SL |
71 | union { |
72 | __u64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */ | |
73 | __u64 journal_tail;/* journal tail of journal device (from data_offset) */ | |
74 | }; | |
82d9eba6 NB |
75 | __u32 dev_number; /* permanent identifier of this device - not role in raid */ |
76 | __u32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */ | |
77 | __u8 device_uuid[16]; /* user-space setable, ignored by kernel */ | |
5d500228 | 78 | __u8 devflags; /* per-device flags. Only one defined...*/ |
dfd4d8ee | 79 | #define WriteMostly1 1 /* mask for writemostly flag in above */ |
bf95d0f3 N |
80 | /* bad block log. If there are any bad blocks the feature flag is set. |
81 | * if offset and size are non-zero, that space is reserved and available. | |
82 | */ | |
83 | __u8 bblog_shift; /* shift from sectors to block size for badblocklist */ | |
84 | __u16 bblog_size; /* number of sectors reserved for badblocklist */ | |
85 | __u32 bblog_offset; /* sector offset from superblock to bblog, signed */ | |
82d9eba6 NB |
86 | |
87 | /* array state information - 64 bytes */ | |
88 | __u64 utime; /* 40 bits second, 24 btes microseconds */ | |
89 | __u64 events; /* incremented when superblock updated */ | |
90 | __u64 resync_offset; /* data before this offset (from data_offset) known to be in sync */ | |
4180aa4d N |
91 | __u32 sb_csum; /* checksum upto dev_roles[max_dev] */ |
92 | __u32 max_dev; /* size of dev_roles[] array to consider */ | |
82d9eba6 NB |
93 | __u8 pad3[64-32]; /* set to 0 when writing */ |
94 | ||
95 | /* device state information. Indexed by dev_number. | |
96 | * 2 bytes per device | |
97 | * Note there are no per-device state flags. State information is rolled | |
98 | * into the 'roles' value. If a device is spare or faulty, then it doesn't | |
99 | * have a meaningful role. | |
100 | */ | |
101 | __u16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */ | |
102 | }; | |
103 | ||
911cead7 N |
104 | #define MAX_SB_SIZE 4096 |
105 | /* bitmap super size is 256, but we round up to a sector for alignment */ | |
106 | #define BM_SUPER_SIZE 512 | |
107 | #define MAX_DEVS ((int)(MAX_SB_SIZE - sizeof(struct mdp_superblock_1)) / 2) | |
41226756 JS |
108 | #define SUPER1_SIZE (MAX_SB_SIZE + BM_SUPER_SIZE \ |
109 | + sizeof(struct misc_dev_info)) | |
911cead7 | 110 | |
bee8ec56 NB |
111 | struct misc_dev_info { |
112 | __u64 device_size; | |
113 | }; | |
114 | ||
d2cd3ffc NB |
115 | /* feature_map bits */ |
116 | #define MD_FEATURE_BITMAP_OFFSET 1 | |
117 | #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and | |
118 | * must be honoured | |
119 | */ | |
b674b5b8 | 120 | #define MD_FEATURE_RESHAPE_ACTIVE 4 |
bf95d0f3 | 121 | #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */ |
8fe1c44f N |
122 | #define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an |
123 | * active device with same 'role'. | |
124 | * 'recovery_offset' is also set. | |
125 | */ | |
126 | #define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number | |
127 | * of devices, but is going | |
128 | * backwards anyway. | |
129 | */ | |
130 | #define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */ | |
6d9c7c25 | 131 | #define MD_FEATURE_BITMAP_VERSIONED 256 /* bitmap version number checked properly */ |
ed94976d | 132 | #define MD_FEATURE_JOURNAL 512 /* support write journal */ |
8fe1c44f N |
133 | #define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \ |
134 | |MD_FEATURE_RECOVERY_OFFSET \ | |
135 | |MD_FEATURE_RESHAPE_ACTIVE \ | |
136 | |MD_FEATURE_BAD_BLOCKS \ | |
137 | |MD_FEATURE_REPLACEMENT \ | |
138 | |MD_FEATURE_RESHAPE_BACKWARDS \ | |
139 | |MD_FEATURE_NEW_OFFSET \ | |
6d9c7c25 | 140 | |MD_FEATURE_BITMAP_VERSIONED \ |
ed94976d | 141 | |MD_FEATURE_JOURNAL \ |
8fe1c44f | 142 | ) |
69a48116 SL |
143 | |
144 | static int role_from_sb(struct mdp_superblock_1 *sb) | |
145 | { | |
146 | unsigned int d; | |
147 | int role; | |
148 | ||
149 | d = __le32_to_cpu(sb->dev_number); | |
150 | if (d < __le32_to_cpu(sb->max_dev)) | |
151 | role = __le16_to_cpu(sb->dev_roles[d]); | |
152 | else | |
153 | role = MD_DISK_ROLE_SPARE; | |
154 | return role; | |
155 | } | |
d2cd3ffc | 156 | |
7e6e839a GJ |
157 | /* return how many bytes are needed for bitmap, for cluster-md each node |
158 | * should have it's own bitmap */ | |
159 | static unsigned int calc_bitmap_size(bitmap_super_t *bms, unsigned int boundary) | |
160 | { | |
161 | unsigned long long bits, bytes; | |
162 | ||
163 | bits = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9); | |
164 | bytes = (bits+7) >> 3; | |
165 | bytes += sizeof(bitmap_super_t); | |
166 | bytes = ROUND_UP(bytes, boundary); | |
167 | ||
168 | return bytes; | |
169 | } | |
170 | ||
82d9eba6 NB |
171 | static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb) |
172 | { | |
173 | unsigned int disk_csum, csum; | |
174 | unsigned long long newcsum; | |
175 | int size = sizeof(*sb) + __le32_to_cpu(sb->max_dev)*2; | |
176 | unsigned int *isuper = (unsigned int*)sb; | |
82d9eba6 NB |
177 | |
178 | /* make sure I can count... */ | |
179 | if (offsetof(struct mdp_superblock_1,data_offset) != 128 || | |
180 | offsetof(struct mdp_superblock_1, utime) != 192 || | |
181 | sizeof(struct mdp_superblock_1) != 256) { | |
182 | fprintf(stderr, "WARNING - superblock isn't sized correctly\n"); | |
183 | } | |
184 | ||
185 | disk_csum = sb->sb_csum; | |
186 | sb->sb_csum = 0; | |
187 | newcsum = 0; | |
e4c72d1d | 188 | for (; size>=4; size -= 4 ) { |
722966c6 PC |
189 | newcsum += __le32_to_cpu(*isuper); |
190 | isuper++; | |
191 | } | |
82d9eba6 NB |
192 | |
193 | if (size == 2) | |
194 | newcsum += __le16_to_cpu(*(unsigned short*) isuper); | |
195 | ||
196 | csum = (newcsum & 0xffffffff) + (newcsum >> 32); | |
197 | sb->sb_csum = disk_csum; | |
83205b64 | 198 | return __cpu_to_le32(csum); |
82d9eba6 NB |
199 | } |
200 | ||
308340aa JS |
201 | /* |
202 | * Information related to file descriptor used for aligned reads/writes. | |
203 | * Cache the block size. | |
204 | */ | |
205 | struct align_fd { | |
206 | int fd; | |
207 | int blk_sz; | |
208 | }; | |
209 | ||
210 | static void init_afd(struct align_fd *afd, int fd) | |
211 | { | |
212 | afd->fd = fd; | |
213 | ||
214 | if (ioctl(afd->fd, BLKSSZGET, &afd->blk_sz) != 0) | |
215 | afd->blk_sz = 512; | |
216 | } | |
217 | ||
5ea022a1 | 218 | static char abuf[4096+4096]; |
308340aa | 219 | static int aread(struct align_fd *afd, void *buf, int len) |
5ea022a1 N |
220 | { |
221 | /* aligned read. | |
222 | * On devices with a 4K sector size, we need to read | |
223 | * the full sector and copy relevant bits into | |
224 | * the buffer | |
225 | */ | |
6ef89052 | 226 | int bsize, iosize; |
5ea022a1 N |
227 | char *b; |
228 | int n; | |
6ef89052 | 229 | |
308340aa JS |
230 | bsize = afd->blk_sz; |
231 | ||
232 | if (!bsize || bsize > 4096 || len > 4096) { | |
233 | if (!bsize) | |
7a862a02 | 234 | fprintf(stderr, "WARNING - aread() called with invalid block size\n"); |
5ea022a1 | 235 | return -1; |
308340aa | 236 | } |
0a2f1894 | 237 | b = ROUND_UP_PTR((char *)abuf, 4096); |
5ea022a1 | 238 | |
6ef89052 N |
239 | for (iosize = 0; iosize < len; iosize += bsize) |
240 | ; | |
308340aa | 241 | n = read(afd->fd, b, iosize); |
5ea022a1 N |
242 | if (n <= 0) |
243 | return n; | |
308340aa | 244 | lseek(afd->fd, len - n, 1); |
5ea022a1 N |
245 | if (n > len) |
246 | n = len; | |
247 | memcpy(buf, b, n); | |
248 | return n; | |
249 | } | |
250 | ||
308340aa | 251 | static int awrite(struct align_fd *afd, void *buf, int len) |
5ea022a1 N |
252 | { |
253 | /* aligned write. | |
254 | * On devices with a 4K sector size, we need to write | |
255 | * the full sector. We pre-read if the sector is larger | |
256 | * than the write. | |
257 | * The address must be sector-aligned. | |
258 | */ | |
6ef89052 | 259 | int bsize, iosize; |
5ea022a1 N |
260 | char *b; |
261 | int n; | |
308340aa JS |
262 | |
263 | bsize = afd->blk_sz; | |
264 | if (!bsize || bsize > 4096 || len > 4096) { | |
265 | if (!bsize) | |
7a862a02 | 266 | fprintf(stderr, "WARNING - awrite() called with invalid block size\n"); |
5ea022a1 | 267 | return -1; |
308340aa | 268 | } |
0a2f1894 | 269 | b = ROUND_UP_PTR((char *)abuf, 4096); |
5ea022a1 | 270 | |
6ef89052 N |
271 | for (iosize = 0; iosize < len ; iosize += bsize) |
272 | ; | |
273 | ||
274 | if (len != iosize) { | |
308340aa | 275 | n = read(afd->fd, b, iosize); |
6ef89052 N |
276 | if (n <= 0) |
277 | return n; | |
308340aa | 278 | lseek(afd->fd, -n, 1); |
6ef89052 N |
279 | } |
280 | ||
5ea022a1 | 281 | memcpy(b, buf, len); |
308340aa | 282 | n = write(afd->fd, b, iosize); |
5ea022a1 N |
283 | if (n <= 0) |
284 | return n; | |
308340aa | 285 | lseek(afd->fd, len - n, 1); |
5ea022a1 N |
286 | return len; |
287 | } | |
288 | ||
c7654afc | 289 | #ifndef MDASSEMBLE |
3da92f27 | 290 | static void examine_super1(struct supertype *st, char *homehost) |
82d9eba6 | 291 | { |
3da92f27 | 292 | struct mdp_superblock_1 *sb = st->sb; |
0aa2f15b | 293 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+MAX_SB_SIZE); |
82d9eba6 | 294 | time_t atime; |
f21e18ca | 295 | unsigned int d; |
e8e1c176 | 296 | int role; |
4180aa4d | 297 | int delta_extra = 0; |
82d9eba6 NB |
298 | int i; |
299 | char *c; | |
a1cbd7d0 | 300 | int l = homehost ? strlen(homehost) : 0; |
83205b64 | 301 | int layout; |
a17a3de3 | 302 | unsigned long long sb_offset; |
5e1863d4 | 303 | struct mdinfo info; |
82d9eba6 NB |
304 | |
305 | printf(" Magic : %08x\n", __le32_to_cpu(sb->magic)); | |
a17a3de3 DL |
306 | printf(" Version : 1"); |
307 | sb_offset = __le64_to_cpu(sb->super_offset); | |
308 | if (sb_offset <= 4) | |
309 | printf(".1\n"); | |
310 | else if (sb_offset <= 8) | |
311 | printf(".2\n"); | |
312 | else | |
313 | printf(".0\n"); | |
91eedefc | 314 | printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb->feature_map)); |
82d9eba6 NB |
315 | printf(" Array UUID : "); |
316 | for (i=0; i<16; i++) { | |
82d9eba6 | 317 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 318 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 NB |
319 | } |
320 | printf("\n"); | |
a1cbd7d0 NB |
321 | printf(" Name : %.32s", sb->set_name); |
322 | if (l > 0 && l < 32 && | |
323 | sb->set_name[l] == ':' && | |
324 | strncmp(sb->set_name, homehost, l) == 0) | |
325 | printf(" (local to host %s)", homehost); | |
326 | printf("\n"); | |
8b2202de | 327 | if (bms->nodes > 0 && (__le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET)) |
2cf42394 | 328 | printf(" Cluster Name : %-64s\n", bms->cluster_name); |
82d9eba6 NB |
329 | atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL; |
330 | printf(" Creation Time : %.24s\n", ctime(&atime)); | |
331 | c=map_num(pers, __le32_to_cpu(sb->level)); | |
332 | printf(" Raid Level : %s\n", c?c:"-unknown-"); | |
333 | printf(" Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks)); | |
334 | printf("\n"); | |
5cda0964 | 335 | printf(" Avail Dev Size : %llu%s\n", |
2fb749d1 NB |
336 | (unsigned long long)__le64_to_cpu(sb->data_size), |
337 | human_size(__le64_to_cpu(sb->data_size)<<9)); | |
c43f7d91 | 338 | if (__le32_to_cpu(sb->level) > 0) { |
d4633e06 | 339 | int ddsks = 0, ddsks_denom = 1; |
b674b5b8 NB |
340 | switch(__le32_to_cpu(sb->level)) { |
341 | case 1: ddsks=1;break; | |
342 | case 4: | |
83205b64 NB |
343 | case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break; |
344 | case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break; | |
345 | case 10: | |
346 | layout = __le32_to_cpu(sb->layout); | |
d4633e06 N |
347 | ddsks = __le32_to_cpu(sb->raid_disks); |
348 | ddsks_denom = (layout&255) * ((layout>>8)&255); | |
b674b5b8 | 349 | } |
d4633e06 N |
350 | if (ddsks) { |
351 | long long asize = __le64_to_cpu(sb->size); | |
352 | asize = (asize << 9) * ddsks / ddsks_denom; | |
83205b64 | 353 | printf(" Array Size : %llu%s\n", |
d4633e06 N |
354 | asize >> 10, human_size(asize)); |
355 | } | |
b674b5b8 | 356 | if (sb->size != sb->data_size) |
5cda0964 | 357 | printf(" Used Dev Size : %llu%s\n", |
83205b64 NB |
358 | (unsigned long long)__le64_to_cpu(sb->size), |
359 | human_size(__le64_to_cpu(sb->size)<<9)); | |
b674b5b8 | 360 | } |
82d9eba6 | 361 | if (sb->data_offset) |
199171a2 NB |
362 | printf(" Data Offset : %llu sectors\n", |
363 | (unsigned long long)__le64_to_cpu(sb->data_offset)); | |
8772113a N |
364 | if (sb->new_offset && |
365 | (__le32_to_cpu(sb->feature_map) & MD_FEATURE_NEW_OFFSET)) { | |
8fe1c44f N |
366 | unsigned long long offset = __le64_to_cpu(sb->data_offset); |
367 | offset += (signed)(int32_t)__le32_to_cpu(sb->new_offset); | |
368 | printf(" New Offset : %llu sectors\n", offset); | |
369 | } | |
199171a2 NB |
370 | printf(" Super Offset : %llu sectors\n", |
371 | (unsigned long long)__le64_to_cpu(sb->super_offset)); | |
d2cd3ffc NB |
372 | if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET) |
373 | printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->recovery_offset)); | |
5e1863d4 N |
374 | |
375 | st->ss->getinfo_super(st, &info, NULL); | |
376 | if (info.space_after != 1 && | |
377 | !(__le32_to_cpu(sb->feature_map) & MD_FEATURE_NEW_OFFSET)) | |
378 | printf(" Unused Space : before=%llu sectors, after=%llu sectors\n", | |
379 | info.space_before, info.space_after); | |
380 | ||
a3fd117c | 381 | printf(" State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean"); |
82d9eba6 NB |
382 | printf(" Device UUID : "); |
383 | for (i=0; i<16; i++) { | |
82d9eba6 | 384 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 385 | printf("%02x", sb->device_uuid[i]); |
82d9eba6 NB |
386 | } |
387 | printf("\n"); | |
b674b5b8 | 388 | printf("\n"); |
91eedefc NB |
389 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { |
390 | printf("Internal Bitmap : %ld sectors from superblock\n", | |
68754bd1 | 391 | (long)(int32_t)__le32_to_cpu(sb->bitmap_offset)); |
8fac0577 | 392 | } |
bf95d0f3 | 393 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE)) { |
1e0d770c | 394 | printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb->reshape_position)/2, |
b674b5b8 NB |
395 | human_size(__le64_to_cpu(sb->reshape_position)<<9)); |
396 | if (__le32_to_cpu(sb->delta_disks)) { | |
397 | printf(" Delta Devices : %d", __le32_to_cpu(sb->delta_disks)); | |
4180aa4d N |
398 | printf(" (%d->%d)\n", |
399 | __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks), | |
400 | __le32_to_cpu(sb->raid_disks)); | |
401 | if ((int)__le32_to_cpu(sb->delta_disks) < 0) | |
402 | delta_extra = -__le32_to_cpu(sb->delta_disks); | |
b674b5b8 NB |
403 | } |
404 | if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) { | |
405 | c = map_num(pers, __le32_to_cpu(sb->new_level)); | |
406 | printf(" New Level : %s\n", c?c:"-unknown-"); | |
407 | } | |
408 | if (__le32_to_cpu(sb->new_layout) != __le32_to_cpu(sb->layout)) { | |
409 | if (__le32_to_cpu(sb->level) == 5) { | |
410 | c = map_num(r5layout, __le32_to_cpu(sb->new_layout)); | |
411 | printf(" New Layout : %s\n", c?c:"-unknown-"); | |
412 | } | |
fe77a154 N |
413 | if (__le32_to_cpu(sb->level) == 6) { |
414 | c = map_num(r6layout, __le32_to_cpu(sb->new_layout)); | |
415 | printf(" New Layout : %s\n", c?c:"-unknown-"); | |
416 | } | |
b674b5b8 | 417 | if (__le32_to_cpu(sb->level) == 10) { |
e4965ef8 N |
418 | printf(" New Layout :"); |
419 | print_r10_layout(__le32_to_cpu(sb->new_layout)); | |
420 | printf("\n"); | |
b674b5b8 NB |
421 | } |
422 | } | |
423 | if (__le32_to_cpu(sb->new_chunk) != __le32_to_cpu(sb->chunksize)) | |
424 | printf(" New Chunksize : %dK\n", __le32_to_cpu(sb->new_chunk)/2); | |
425 | printf("\n"); | |
426 | } | |
dfd4d8ee | 427 | if (sb->devflags) { |
62184891 | 428 | printf(" Flags :"); |
34163fc7 | 429 | if (sb->devflags & WriteMostly1) |
dfd4d8ee NB |
430 | printf(" write-mostly"); |
431 | printf("\n"); | |
432 | } | |
82d9eba6 NB |
433 | |
434 | atime = __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL; | |
435 | printf(" Update Time : %.24s\n", ctime(&atime)); | |
436 | ||
bf95d0f3 N |
437 | if (sb->bblog_size && sb->bblog_offset) { |
438 | printf(" Bad Block Log : %d entries available at offset %ld sectors", | |
439 | __le16_to_cpu(sb->bblog_size)*512/8, | |
d6e4b44f | 440 | (long)(int32_t)__le32_to_cpu(sb->bblog_offset)); |
bf95d0f3 N |
441 | if (sb->feature_map & |
442 | __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
443 | printf(" - bad blocks present."); | |
444 | printf("\n"); | |
445 | } | |
446 | ||
82d9eba6 NB |
447 | if (calc_sb_1_csum(sb) == sb->sb_csum) |
448 | printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum)); | |
449 | else | |
450 | printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb->sb_csum), | |
451 | __le32_to_cpu(calc_sb_1_csum(sb))); | |
570c0542 | 452 | printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb->events)); |
82d9eba6 NB |
453 | printf("\n"); |
454 | if (__le32_to_cpu(sb->level) == 5) { | |
455 | c = map_num(r5layout, __le32_to_cpu(sb->layout)); | |
456 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
457 | } | |
fe77a154 N |
458 | if (__le32_to_cpu(sb->level) == 6) { |
459 | c = map_num(r6layout, __le32_to_cpu(sb->layout)); | |
460 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
461 | } | |
265e0f17 NB |
462 | if (__le32_to_cpu(sb->level) == 10) { |
463 | int lo = __le32_to_cpu(sb->layout); | |
e4965ef8 N |
464 | printf(" Layout :"); |
465 | print_r10_layout(lo); | |
466 | printf("\n"); | |
265e0f17 | 467 | } |
82d9eba6 NB |
468 | switch(__le32_to_cpu(sb->level)) { |
469 | case 0: | |
470 | case 4: | |
471 | case 5: | |
d2cd3ffc NB |
472 | case 6: |
473 | case 10: | |
b674b5b8 | 474 | printf(" Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize)/2); |
82d9eba6 NB |
475 | break; |
476 | case -1: | |
b674b5b8 | 477 | printf(" Rounding : %dK\n", __le32_to_cpu(sb->chunksize)/2); |
82d9eba6 NB |
478 | break; |
479 | default: break; | |
480 | } | |
481 | printf("\n"); | |
e8e1c176 N |
482 | #if 0 |
483 | /* This turns out to just be confusing */ | |
c3684618 NB |
484 | printf(" Array Slot : %d (", __le32_to_cpu(sb->dev_number)); |
485 | for (i= __le32_to_cpu(sb->max_dev); i> 0 ; i--) | |
fa7574f6 | 486 | if (__le16_to_cpu(sb->dev_roles[i-1]) != MD_DISK_ROLE_SPARE) |
c3684618 NB |
487 | break; |
488 | for (d=0; d < i; d++) { | |
489 | int role = __le16_to_cpu(sb->dev_roles[d]); | |
490 | if (d) printf(", "); | |
fa7574f6 SL |
491 | if (role == MD_DISK_ROLE_SPARE) printf("empty"); |
492 | else if(role == MD_DISK_ROLE_FAULTY) printf("failed"); | |
c3684618 NB |
493 | else printf("%d", role); |
494 | } | |
495 | printf(")\n"); | |
e8e1c176 N |
496 | #endif |
497 | printf(" Device Role : "); | |
69a48116 | 498 | role = role_from_sb(sb); |
fa7574f6 | 499 | if (role >= MD_DISK_ROLE_FAULTY) |
e8e1c176 | 500 | printf("spare\n"); |
ed94976d SL |
501 | else if (role == MD_DISK_ROLE_JOURNAL) |
502 | printf("Journal\n"); | |
24c7bc84 N |
503 | else if (sb->feature_map & __cpu_to_le32(MD_FEATURE_REPLACEMENT)) |
504 | printf("Replacement device %d\n", role); | |
e8e1c176 N |
505 | else |
506 | printf("Active device %d\n", role); | |
507 | ||
82d9eba6 | 508 | printf(" Array State : "); |
4180aa4d | 509 | for (d=0; d<__le32_to_cpu(sb->raid_disks) + delta_extra; d++) { |
82d9eba6 | 510 | int cnt = 0; |
f21e18ca | 511 | unsigned int i; |
82d9eba6 | 512 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { |
f21e18ca | 513 | unsigned int role = __le16_to_cpu(sb->dev_roles[i]); |
e4c72d1d | 514 | if (role == d) |
82d9eba6 | 515 | cnt++; |
82d9eba6 | 516 | } |
24c7bc84 N |
517 | if (cnt == 2) |
518 | printf("R"); | |
519 | else if (cnt == 1) | |
520 | printf("A"); | |
521 | else if (cnt == 0) | |
522 | printf("."); | |
523 | else | |
524 | printf("?"); | |
82d9eba6 | 525 | } |
e8e1c176 N |
526 | #if 0 |
527 | /* This is confusing too */ | |
6fbba4c9 | 528 | faulty = 0; |
82d9eba6 NB |
529 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { |
530 | int role = __le16_to_cpu(sb->dev_roles[i]); | |
fa7574f6 | 531 | if (role == MD_DISK_ROLE_FAULTY) |
6fbba4c9 | 532 | faulty++; |
82d9eba6 | 533 | } |
82d9eba6 | 534 | if (faulty) printf(" %d failed", faulty); |
e8e1c176 | 535 | #endif |
24c7bc84 | 536 | printf(" ('A' == active, '.' == missing, 'R' == replacing)"); |
82d9eba6 NB |
537 | printf("\n"); |
538 | } | |
539 | ||
061f2c6a | 540 | static void brief_examine_super1(struct supertype *st, int verbose) |
82d9eba6 | 541 | { |
3da92f27 | 542 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 | 543 | int i; |
a17a3de3 | 544 | unsigned long long sb_offset; |
2998bc01 | 545 | char *nm; |
b015e6c2 | 546 | char *c=map_num(pers, __le32_to_cpu(sb->level)); |
2998bc01 NB |
547 | |
548 | nm = strchr(sb->set_name, ':'); | |
549 | if (nm) | |
550 | nm++; | |
551 | else if (sb->set_name[0]) | |
552 | nm = sb->set_name; | |
553 | else | |
cf8de691 | 554 | nm = NULL; |
82d9eba6 | 555 | |
7103b9b8 N |
556 | printf("ARRAY "); |
557 | if (nm) { | |
558 | printf("/dev/md/"); | |
559 | print_escape(nm); | |
560 | putchar(' '); | |
561 | } | |
061f2c6a N |
562 | if (verbose && c) |
563 | printf(" level=%s", c); | |
a17a3de3 DL |
564 | sb_offset = __le64_to_cpu(sb->super_offset); |
565 | if (sb_offset <= 4) | |
e9a0e728 | 566 | printf(" metadata=1.1 "); |
a17a3de3 | 567 | else if (sb_offset <= 8) |
e9a0e728 | 568 | printf(" metadata=1.2 "); |
a17a3de3 | 569 | else |
e9a0e728 | 570 | printf(" metadata=1.0 "); |
061f2c6a N |
571 | if (verbose) |
572 | printf("num-devices=%d ", __le32_to_cpu(sb->raid_disks)); | |
573 | printf("UUID="); | |
82d9eba6 | 574 | for (i=0; i<16; i++) { |
82d9eba6 | 575 | if ((i&3)==0 && i != 0) printf(":"); |
b7a708af | 576 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 | 577 | } |
7103b9b8 N |
578 | if (sb->set_name[0]) { |
579 | printf(" name="); | |
580 | print_quoted(sb->set_name); | |
581 | } | |
82d9eba6 NB |
582 | printf("\n"); |
583 | } | |
584 | ||
0d726f17 KS |
585 | static void export_examine_super1(struct supertype *st) |
586 | { | |
587 | struct mdp_superblock_1 *sb = st->sb; | |
588 | int i; | |
589 | int len = 32; | |
80730bae | 590 | int layout; |
0d726f17 KS |
591 | |
592 | printf("MD_LEVEL=%s\n", map_num(pers, __le32_to_cpu(sb->level))); | |
593 | printf("MD_DEVICES=%d\n", __le32_to_cpu(sb->raid_disks)); | |
594 | for (i=0; i<32; i++) | |
595 | if (sb->set_name[i] == '\n' || | |
596 | sb->set_name[i] == '\0') { | |
597 | len = i; | |
598 | break; | |
599 | } | |
600 | if (len) | |
601 | printf("MD_NAME=%.*s\n", len, sb->set_name); | |
80730bae MN |
602 | if (__le32_to_cpu(sb->level) > 0) { |
603 | int ddsks = 0, ddsks_denom = 1; | |
604 | switch(__le32_to_cpu(sb->level)) { | |
605 | case 1: ddsks=1;break; | |
606 | case 4: | |
607 | case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break; | |
608 | case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break; | |
609 | case 10: | |
610 | layout = __le32_to_cpu(sb->layout); | |
611 | ddsks = __le32_to_cpu(sb->raid_disks); | |
612 | ddsks_denom = (layout&255) * ((layout>>8)&255); | |
613 | } | |
614 | if (ddsks) { | |
615 | long long asize = __le64_to_cpu(sb->size); | |
616 | asize = (asize << 9) * ddsks / ddsks_denom; | |
617 | printf("MD_ARRAY_SIZE=%s\n",human_size_brief(asize,JEDEC)); | |
618 | } | |
619 | } | |
0d726f17 KS |
620 | printf("MD_UUID="); |
621 | for (i=0; i<16; i++) { | |
622 | if ((i&3)==0 && i != 0) printf(":"); | |
623 | printf("%02x", sb->set_uuid[i]); | |
624 | } | |
625 | printf("\n"); | |
626 | printf("MD_UPDATE_TIME=%llu\n", | |
627 | __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL); | |
628 | printf("MD_DEV_UUID="); | |
629 | for (i=0; i<16; i++) { | |
630 | if ((i&3)==0 && i != 0) printf(":"); | |
631 | printf("%02x", sb->device_uuid[i]); | |
632 | } | |
633 | printf("\n"); | |
634 | printf("MD_EVENTS=%llu\n", | |
635 | (unsigned long long)__le64_to_cpu(sb->events)); | |
636 | } | |
637 | ||
74db60b0 N |
638 | static int copy_metadata1(struct supertype *st, int from, int to) |
639 | { | |
640 | /* Read superblock. If it looks good, write it out. | |
641 | * Then if a bitmap is present, copy that. | |
642 | * And if a bad-block-list is present, copy that too. | |
643 | */ | |
644 | void *buf; | |
645 | unsigned long long dsize, sb_offset; | |
646 | const int bufsize = 4*1024; | |
647 | struct mdp_superblock_1 super, *sb; | |
648 | ||
649 | if (posix_memalign(&buf, 4096, bufsize) != 0) | |
650 | return 1; | |
651 | ||
652 | if (!get_dev_size(from, NULL, &dsize)) | |
653 | goto err; | |
654 | ||
655 | dsize >>= 9; | |
656 | if (dsize < 24) | |
657 | goto err; | |
658 | switch(st->minor_version) { | |
659 | case 0: | |
660 | sb_offset = dsize; | |
661 | sb_offset -= 8*2; | |
662 | sb_offset &= ~(4*2-1); | |
663 | break; | |
664 | case 1: | |
665 | sb_offset = 0; | |
666 | break; | |
667 | case 2: | |
668 | sb_offset = 4*2; | |
669 | break; | |
670 | default: | |
671 | goto err; | |
672 | } | |
673 | ||
674 | if (lseek64(from, sb_offset << 9, 0) < 0LL) | |
675 | goto err; | |
676 | if (read(from, buf, bufsize) != bufsize) | |
677 | goto err; | |
678 | ||
679 | sb = buf; | |
680 | super = *sb; // save most of sb for when we reuse buf | |
681 | ||
682 | if (__le32_to_cpu(super.magic) != MD_SB_MAGIC || | |
683 | __le32_to_cpu(super.major_version) != 1 || | |
684 | __le64_to_cpu(super.super_offset) != sb_offset || | |
685 | calc_sb_1_csum(sb) != super.sb_csum) | |
686 | goto err; | |
687 | ||
688 | if (lseek64(to, sb_offset << 9, 0) < 0LL) | |
689 | goto err; | |
690 | if (write(to, buf, bufsize) != bufsize) | |
691 | goto err; | |
692 | ||
693 | if (super.feature_map & __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET)) { | |
694 | unsigned long long bitmap_offset = sb_offset; | |
695 | int bytes = 4096; // just an estimate. | |
696 | int written = 0; | |
697 | struct align_fd afrom, ato; | |
698 | ||
699 | init_afd(&afrom, from); | |
700 | init_afd(&ato, to); | |
701 | ||
702 | bitmap_offset += (int32_t)__le32_to_cpu(super.bitmap_offset); | |
703 | ||
704 | if (lseek64(from, bitmap_offset<<9, 0) < 0) | |
705 | goto err; | |
706 | if (lseek64(to, bitmap_offset<<9, 0) < 0) | |
707 | goto err; | |
708 | ||
709 | for (written = 0; written < bytes ; ) { | |
710 | int n = bytes - written; | |
711 | if (n > 4096) | |
712 | n = 4096; | |
713 | if (aread(&afrom, buf, n) != n) | |
714 | goto err; | |
715 | if (written == 0) { | |
716 | /* have the header, can calculate | |
717 | * correct bitmap bytes */ | |
718 | bitmap_super_t *bms; | |
74db60b0 | 719 | bms = (void*)buf; |
4a3d29ed | 720 | bytes = calc_bitmap_size(bms, 512); |
74db60b0 N |
721 | if (n > bytes) |
722 | n = bytes; | |
723 | } | |
724 | if (awrite(&ato, buf, n) != n) | |
725 | goto err; | |
726 | written += n; | |
727 | } | |
728 | } | |
729 | ||
730 | if (super.bblog_size != 0 && | |
731 | __le32_to_cpu(super.bblog_size) <= 100 && | |
732 | super.bblog_offset != 0 && | |
733 | (super.feature_map & __le32_to_cpu(MD_FEATURE_BAD_BLOCKS))) { | |
734 | /* There is a bad block log */ | |
735 | unsigned long long bb_offset = sb_offset; | |
736 | int bytes = __le32_to_cpu(super.bblog_size) * 512; | |
737 | int written = 0; | |
738 | struct align_fd afrom, ato; | |
739 | ||
740 | init_afd(&afrom, from); | |
741 | init_afd(&ato, to); | |
742 | ||
743 | bb_offset += (int32_t)__le32_to_cpu(super.bblog_offset); | |
744 | ||
745 | if (lseek64(from, bb_offset<<9, 0) < 0) | |
746 | goto err; | |
747 | if (lseek64(to, bb_offset<<9, 0) < 0) | |
748 | goto err; | |
749 | ||
750 | for (written = 0; written < bytes ; ) { | |
751 | int n = bytes - written; | |
752 | if (n > 4096) | |
753 | n = 4096; | |
754 | if (aread(&afrom, buf, n) != n) | |
755 | goto err; | |
756 | ||
757 | if (awrite(&ato, buf, n) != n) | |
758 | goto err; | |
759 | written += n; | |
760 | } | |
761 | } | |
762 | ||
763 | free(buf); | |
764 | return 0; | |
765 | ||
766 | err: | |
767 | free(buf); | |
768 | return 1; | |
769 | } | |
770 | ||
3da92f27 | 771 | static void detail_super1(struct supertype *st, char *homehost) |
82d9eba6 | 772 | { |
3da92f27 | 773 | struct mdp_superblock_1 *sb = st->sb; |
0aa2f15b | 774 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); |
82d9eba6 | 775 | int i; |
b6750aa8 | 776 | int l = homehost ? strlen(homehost) : 0; |
82d9eba6 | 777 | |
b6750aa8 NB |
778 | printf(" Name : %.32s", sb->set_name); |
779 | if (l > 0 && l < 32 && | |
780 | sb->set_name[l] == ':' && | |
781 | strncmp(sb->set_name, homehost, l) == 0) | |
782 | printf(" (local to host %s)", homehost); | |
8b2202de | 783 | if (bms->nodes > 0 && (__le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET)) |
2cf42394 | 784 | printf("\n Cluster Name : %-64s", bms->cluster_name); |
b6750aa8 | 785 | printf("\n UUID : "); |
82d9eba6 | 786 | for (i=0; i<16; i++) { |
82d9eba6 | 787 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 788 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 | 789 | } |
570c0542 | 790 | printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb->events)); |
82d9eba6 NB |
791 | } |
792 | ||
3da92f27 | 793 | static void brief_detail_super1(struct supertype *st) |
82d9eba6 | 794 | { |
3da92f27 | 795 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 NB |
796 | int i; |
797 | ||
7103b9b8 N |
798 | if (sb->set_name[0]) { |
799 | printf(" name="); | |
800 | print_quoted(sb->set_name); | |
801 | } | |
82d9eba6 NB |
802 | printf(" UUID="); |
803 | for (i=0; i<16; i++) { | |
82d9eba6 | 804 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 805 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 NB |
806 | } |
807 | } | |
808 | ||
0d726f17 | 809 | static void export_detail_super1(struct supertype *st) |
54bad364 | 810 | { |
3da92f27 | 811 | struct mdp_superblock_1 *sb = st->sb; |
54bad364 KS |
812 | int i; |
813 | int len = 32; | |
814 | ||
815 | for (i=0; i<32; i++) | |
816 | if (sb->set_name[i] == '\n' || | |
817 | sb->set_name[i] == '\0') { | |
818 | len = i; | |
819 | break; | |
820 | } | |
821 | if (len) | |
822 | printf("MD_NAME=%.*s\n", len, sb->set_name); | |
54bad364 KS |
823 | } |
824 | ||
6d388a88 N |
825 | static int examine_badblocks_super1(struct supertype *st, int fd, char *devname) |
826 | { | |
827 | struct mdp_superblock_1 *sb = st->sb; | |
828 | unsigned long long offset; | |
829 | int size; | |
830 | __u64 *bbl, *bbp; | |
831 | int i; | |
832 | ||
833 | if (!sb->bblog_size || __le32_to_cpu(sb->bblog_size) > 100 | |
834 | || !sb->bblog_offset){ | |
835 | printf("No bad-blocks list configured on %s\n", devname); | |
836 | return 0; | |
837 | } | |
838 | if ((sb->feature_map & __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
839 | == 0) { | |
840 | printf("Bad-blocks list is empty in %s\n", devname); | |
841 | return 0; | |
842 | } | |
843 | ||
844 | size = __le32_to_cpu(sb->bblog_size)* 512; | |
0cf83229 | 845 | if (posix_memalign((void**)&bbl, 4096, size) != 0) { |
1ade5cc1 | 846 | pr_err("could not allocate badblocks list\n"); |
0cf83229 N |
847 | return 0; |
848 | } | |
6d388a88 N |
849 | offset = __le64_to_cpu(sb->super_offset) + |
850 | (int)__le32_to_cpu(sb->bblog_offset); | |
851 | offset <<= 9; | |
852 | if (lseek64(fd, offset, 0) < 0) { | |
853 | pr_err("Cannot seek to bad-blocks list\n"); | |
854 | return 1; | |
855 | } | |
856 | if (read(fd, bbl, size) != size) { | |
857 | pr_err("Cannot read bad-blocks list\n"); | |
858 | return 1; | |
859 | } | |
860 | /* 64bits per entry. 10 bits is block-count, 54 bits is block | |
861 | * offset. Blocks are sectors unless bblog->shift makes them bigger | |
862 | */ | |
863 | bbp = (__u64*)bbl; | |
864 | printf("Bad-blocks on %s:\n", devname); | |
865 | for (i = 0; i < size/8; i++, bbp++) { | |
866 | __u64 bb = __le64_to_cpu(*bbp); | |
867 | int count = bb & 0x3ff; | |
868 | unsigned long long sector = bb >> 10; | |
869 | ||
870 | if (bb + 1 == 0) | |
871 | break; | |
872 | ||
873 | sector <<= sb->bblog_shift; | |
874 | count <<= sb->bblog_shift; | |
875 | ||
876 | printf("%20llu for %d sectors\n", sector, count); | |
877 | } | |
878 | return 0; | |
879 | } | |
880 | ||
c7654afc NB |
881 | #endif |
882 | ||
3da92f27 | 883 | static int match_home1(struct supertype *st, char *homehost) |
83b6208e | 884 | { |
3da92f27 | 885 | struct mdp_superblock_1 *sb = st->sb; |
83b6208e NB |
886 | int l = homehost ? strlen(homehost) : 0; |
887 | ||
888 | return (l > 0 && l < 32 && | |
889 | sb->set_name[l] == ':' && | |
890 | strncmp(sb->set_name, homehost, l) == 0); | |
891 | } | |
892 | ||
3da92f27 | 893 | static void uuid_from_super1(struct supertype *st, int uuid[4]) |
82d9eba6 | 894 | { |
3da92f27 | 895 | struct mdp_superblock_1 *super = st->sb; |
82d9eba6 NB |
896 | char *cuuid = (char*)uuid; |
897 | int i; | |
898 | for (i=0; i<16; i++) | |
899 | cuuid[i] = super->set_uuid[i]; | |
900 | } | |
901 | ||
a5d85af7 | 902 | static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map) |
82d9eba6 | 903 | { |
3da92f27 | 904 | struct mdp_superblock_1 *sb = st->sb; |
80bf9135 N |
905 | struct bitmap_super_s *bsb = (void*)(((char*)sb)+MAX_SB_SIZE); |
906 | struct misc_dev_info *misc = (void*)(((char*)sb)+MAX_SB_SIZE+BM_SUPER_SIZE); | |
82d9eba6 | 907 | int working = 0; |
f21e18ca | 908 | unsigned int i; |
a5d85af7 N |
909 | unsigned int role; |
910 | unsigned int map_disks = info->array.raid_disks; | |
80bf9135 N |
911 | unsigned long long super_offset; |
912 | unsigned long long data_size; | |
82d9eba6 | 913 | |
95eeceeb | 914 | memset(info, 0, sizeof(*info)); |
82d9eba6 | 915 | info->array.major_version = 1; |
b8ac1967 | 916 | info->array.minor_version = st->minor_version; |
82d9eba6 NB |
917 | info->array.patch_version = 0; |
918 | info->array.raid_disks = __le32_to_cpu(sb->raid_disks); | |
919 | info->array.level = __le32_to_cpu(sb->level); | |
265e0f17 | 920 | info->array.layout = __le32_to_cpu(sb->layout); |
82d9eba6 NB |
921 | info->array.md_minor = -1; |
922 | info->array.ctime = __le64_to_cpu(sb->ctime); | |
353632d9 | 923 | info->array.utime = __le64_to_cpu(sb->utime); |
bed256c2 | 924 | info->array.chunk_size = __le32_to_cpu(sb->chunksize)*512; |
4855f95c | 925 | info->array.state = |
d5977056 | 926 | (__le64_to_cpu(sb->resync_offset) == MaxSector) |
4855f95c | 927 | ? 1 : 0; |
06bd6793 GJ |
928 | if (__le32_to_cpu(bsb->nodes) > 1) |
929 | info->array.state |= (1 << MD_SB_CLUSTERED); | |
353632d9 NB |
930 | |
931 | info->data_offset = __le64_to_cpu(sb->data_offset); | |
932 | info->component_size = __le64_to_cpu(sb->size); | |
c0c1acd6 | 933 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET)) |
34a13953 | 934 | info->bitmap_offset = (int32_t)__le32_to_cpu(sb->bitmap_offset); |
82d9eba6 NB |
935 | |
936 | info->disk.major = 0; | |
937 | info->disk.minor = 0; | |
fbf8a0b7 | 938 | info->disk.number = __le32_to_cpu(sb->dev_number); |
82d9eba6 | 939 | if (__le32_to_cpu(sb->dev_number) >= __le32_to_cpu(sb->max_dev) || |
911cead7 | 940 | __le32_to_cpu(sb->dev_number) >= MAX_DEVS) |
fa7574f6 | 941 | role = MD_DISK_ROLE_FAULTY; |
82d9eba6 NB |
942 | else |
943 | role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]); | |
944 | ||
80bf9135 | 945 | super_offset = __le64_to_cpu(sb->super_offset); |
0ddc35be N |
946 | if (info->array.level <= 0) |
947 | data_size = __le64_to_cpu(sb->data_size); | |
948 | else | |
949 | data_size = __le64_to_cpu(sb->size); | |
80bf9135 N |
950 | if (info->data_offset < super_offset) { |
951 | unsigned long long end; | |
952 | info->space_before = info->data_offset; | |
953 | end = super_offset; | |
cc3130a7 N |
954 | |
955 | if (sb->bblog_offset && sb->bblog_size) { | |
956 | unsigned long long bboffset = super_offset; | |
957 | bboffset += (int32_t)__le32_to_cpu(sb->bblog_offset); | |
958 | if (bboffset < end) | |
959 | end = bboffset; | |
960 | } | |
961 | ||
962 | if (super_offset + info->bitmap_offset < end) | |
963 | end = super_offset + info->bitmap_offset; | |
964 | ||
80bf9135 N |
965 | if (info->data_offset + data_size < end) |
966 | info->space_after = end - data_size - info->data_offset; | |
967 | else | |
968 | info->space_after = 0; | |
969 | } else { | |
cc3130a7 N |
970 | unsigned long long earliest; |
971 | earliest = super_offset + (32+4)*2; /* match kernel */ | |
80bf9135 N |
972 | if (info->bitmap_offset > 0) { |
973 | unsigned long long bmend = info->bitmap_offset; | |
974 | unsigned long long size = __le64_to_cpu(bsb->sync_size); | |
975 | size /= __le32_to_cpu(bsb->chunksize) >> 9; | |
976 | size = (size + 7) >> 3; | |
977 | size += sizeof(bitmap_super_t); | |
978 | size = ROUND_UP(size, 4096); | |
979 | size /= 512; | |
cc3130a7 N |
980 | bmend += size; |
981 | if (bmend > earliest) | |
982 | bmend = earliest; | |
983 | } | |
984 | if (sb->bblog_offset && sb->bblog_size) { | |
985 | unsigned long long bbend = super_offset; | |
986 | bbend += (int32_t)__le32_to_cpu(sb->bblog_offset); | |
987 | bbend += __le32_to_cpu(sb->bblog_size); | |
988 | if (bbend > earliest) | |
989 | earliest = bbend; | |
990 | } | |
991 | if (earliest < info->data_offset) | |
992 | info->space_before = info->data_offset - earliest; | |
993 | else | |
994 | info->space_before = 0; | |
80bf9135 N |
995 | info->space_after = misc->device_size - data_size - info->data_offset; |
996 | } | |
c4b26c64 N |
997 | if (info->space_before == 0 && info->space_after == 0) { |
998 | /* It will look like we don't support data_offset changes, | |
999 | * be we do - it's just that there is no room. | |
1000 | * A change that reduced the number of devices should | |
1001 | * still be allowed, so set the otherwise useless value of '1' | |
1002 | */ | |
1003 | info->space_after = 1; | |
1004 | } | |
80bf9135 | 1005 | |
82d9eba6 NB |
1006 | info->disk.raid_disk = -1; |
1007 | switch(role) { | |
fa7574f6 | 1008 | case MD_DISK_ROLE_SPARE: |
f22385f9 | 1009 | info->disk.state = 0; /* spare: not active, not sync, not faulty */ |
82d9eba6 | 1010 | break; |
fa7574f6 | 1011 | case MD_DISK_ROLE_FAULTY: |
82d9eba6 NB |
1012 | info->disk.state = 1; /* faulty */ |
1013 | break; | |
ed94976d SL |
1014 | case MD_DISK_ROLE_JOURNAL: |
1015 | info->disk.state = (1 << MD_DISK_JOURNAL); | |
1016 | info->disk.raid_disk = role; | |
1017 | info->space_after = (misc->device_size - info->data_offset) % 8; /* journal uses all 4kB blocks*/ | |
1018 | break; | |
82d9eba6 NB |
1019 | default: |
1020 | info->disk.state = 6; /* active and in sync */ | |
1021 | info->disk.raid_disk = role; | |
1022 | } | |
cb099724 N |
1023 | if (sb->devflags & WriteMostly1) |
1024 | info->disk.state |= (1 << MD_DISK_WRITEMOSTLY); | |
82d9eba6 | 1025 | info->events = __le64_to_cpu(sb->events); |
1522c538 | 1026 | sprintf(info->text_version, "1.%d", st->minor_version); |
a67dd8cc | 1027 | info->safe_mode_delay = 200; |
82d9eba6 NB |
1028 | |
1029 | memcpy(info->uuid, sb->set_uuid, 16); | |
1030 | ||
31317663 NB |
1031 | strncpy(info->name, sb->set_name, 32); |
1032 | info->name[32] = 0; | |
947fd4dd | 1033 | |
aacb2f81 | 1034 | if ((__le32_to_cpu(sb->feature_map)&MD_FEATURE_REPLACEMENT)) { |
72e7fb13 | 1035 | info->disk.state &= ~(1 << MD_DISK_SYNC); |
aacb2f81 N |
1036 | info->disk.state |= 1 << MD_DISK_REPLACEMENT; |
1037 | } | |
1038 | ||
921d9e16 N |
1039 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET)) |
1040 | info->recovery_start = __le32_to_cpu(sb->recovery_offset); | |
1041 | else | |
1042 | info->recovery_start = MaxSector; | |
1043 | ||
353632d9 NB |
1044 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) { |
1045 | info->reshape_active = 1; | |
26bf5587 N |
1046 | if ((sb->feature_map & __le32_to_cpu(MD_FEATURE_NEW_OFFSET)) && |
1047 | sb->new_offset != 0) | |
5e88ab2e | 1048 | info->reshape_active |= RESHAPE_NO_BACKUP; |
353632d9 NB |
1049 | info->reshape_progress = __le64_to_cpu(sb->reshape_position); |
1050 | info->new_level = __le32_to_cpu(sb->new_level); | |
1051 | info->delta_disks = __le32_to_cpu(sb->delta_disks); | |
1052 | info->new_layout = __le32_to_cpu(sb->new_layout); | |
bed256c2 | 1053 | info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9; |
4180aa4d N |
1054 | if (info->delta_disks < 0) |
1055 | info->array.raid_disks -= info->delta_disks; | |
353632d9 NB |
1056 | } else |
1057 | info->reshape_active = 0; | |
1058 | ||
6e75048b AK |
1059 | info->recovery_blocked = info->reshape_active; |
1060 | ||
a5d85af7 N |
1061 | if (map) |
1062 | for (i=0; i<map_disks; i++) | |
1063 | map[i] = 0; | |
f21e18ca | 1064 | for (i = 0; i < __le32_to_cpu(sb->max_dev); i++) { |
82d9eba6 | 1065 | role = __le16_to_cpu(sb->dev_roles[i]); |
fa7574f6 | 1066 | if (/*role == MD_DISK_ROLE_SPARE || */role < (unsigned) info->array.raid_disks) { |
82d9eba6 | 1067 | working++; |
a5d85af7 N |
1068 | if (map && role < map_disks) |
1069 | map[role] = 1; | |
1070 | } | |
82d9eba6 NB |
1071 | } |
1072 | ||
1073 | info->array.working_disks = working; | |
051f3265 SL |
1074 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_JOURNAL)) |
1075 | info->journal_device_required = 1; | |
1076 | info->journal_clean = 0; | |
82d9eba6 NB |
1077 | } |
1078 | ||
00bbdbda N |
1079 | static struct mdinfo *container_content1(struct supertype *st, char *subarray) |
1080 | { | |
1081 | struct mdinfo *info; | |
1082 | ||
1083 | if (subarray) | |
1084 | return NULL; | |
1085 | ||
503975b9 | 1086 | info = xmalloc(sizeof(*info)); |
00bbdbda N |
1087 | getinfo_super1(st, info, NULL); |
1088 | return info; | |
1089 | } | |
1090 | ||
68c7d6d7 | 1091 | static int update_super1(struct supertype *st, struct mdinfo *info, |
3da92f27 | 1092 | char *update, |
e5eac01f NB |
1093 | char *devname, int verbose, |
1094 | int uuid_set, char *homehost) | |
82d9eba6 | 1095 | { |
5a31170d N |
1096 | /* NOTE: for 'assemble' and 'force' we need to return non-zero |
1097 | * if any change was made. For others, the return value is | |
1098 | * ignored. | |
c4d831e1 | 1099 | */ |
82d9eba6 | 1100 | int rv = 0; |
d15a1f72 | 1101 | int lockid; |
3da92f27 | 1102 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 | 1103 | |
d15a1f72 GJ |
1104 | if (is_clustered(st)) { |
1105 | rv = cluster_get_dlmlock(st, &lockid); | |
1106 | if (rv) { | |
1107 | pr_err("Cannot get dlmlock in %s return %d\n", __func__, rv); | |
1108 | cluster_release_dlmlock(st, lockid); | |
1109 | return rv; | |
1110 | } | |
1111 | } | |
1112 | ||
def11332 N |
1113 | if (strcmp(update, "homehost") == 0 && |
1114 | homehost) { | |
1115 | /* Note that 'homehost' is special as it is really | |
1116 | * a "name" update. | |
1117 | */ | |
1118 | char *c; | |
1119 | update = "name"; | |
1120 | c = strchr(sb->set_name, ':'); | |
1121 | if (c) | |
1122 | strncpy(info->name, c+1, 31 - (c-sb->set_name)); | |
1123 | else | |
1124 | strncpy(info->name, sb->set_name, 32); | |
1125 | info->name[32] = 0; | |
1126 | } | |
1127 | ||
67a8c82d NB |
1128 | if (strcmp(update, "force-one")==0) { |
1129 | /* Not enough devices for a working array, | |
1130 | * so bring this one up-to-date | |
1131 | */ | |
c4d831e1 NB |
1132 | if (sb->events != __cpu_to_le64(info->events)) |
1133 | rv = 1; | |
6d3d5804 | 1134 | sb->events = __cpu_to_le64(info->events); |
1e2b2765 | 1135 | } else if (strcmp(update, "force-array")==0) { |
67a8c82d NB |
1136 | /* Degraded array and 'force' requests to |
1137 | * maybe need to mark it 'clean'. | |
1138 | */ | |
82d9eba6 NB |
1139 | switch(__le32_to_cpu(sb->level)) { |
1140 | case 5: case 4: case 6: | |
1141 | /* need to force clean */ | |
b7528a20 | 1142 | if (sb->resync_offset != MaxSector) |
c4d831e1 | 1143 | rv = 1; |
b7528a20 | 1144 | sb->resync_offset = MaxSector; |
82d9eba6 | 1145 | } |
1e2b2765 | 1146 | } else if (strcmp(update, "assemble")==0) { |
82d9eba6 NB |
1147 | int d = info->disk.number; |
1148 | int want; | |
aacb2f81 | 1149 | if (info->disk.state & (1<<MD_DISK_ACTIVE)) |
a2ce5a1a | 1150 | want = info->disk.raid_disk; |
69a48116 SL |
1151 | else if (info->disk.state & (1<<MD_DISK_JOURNAL)) |
1152 | want = MD_DISK_ROLE_JOURNAL; | |
82d9eba6 | 1153 | else |
fa7574f6 | 1154 | want = MD_DISK_ROLE_SPARE; |
a2ce5a1a N |
1155 | if (sb->dev_roles[d] != __cpu_to_le16(want)) { |
1156 | sb->dev_roles[d] = __cpu_to_le16(want); | |
82d9eba6 NB |
1157 | rv = 1; |
1158 | } | |
d43494fc N |
1159 | if (info->reshape_active && |
1160 | sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) && | |
1161 | info->delta_disks >= 0 && | |
1162 | info->reshape_progress < __le64_to_cpu(sb->reshape_position)) { | |
1163 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
1164 | rv = 1; | |
1165 | } | |
1166 | if (info->reshape_active && | |
1167 | sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) && | |
1168 | info->delta_disks < 0 && | |
1169 | info->reshape_progress > __le64_to_cpu(sb->reshape_position)) { | |
1170 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
1171 | rv = 1; | |
1172 | } | |
1e2b2765 | 1173 | } else if (strcmp(update, "linear-grow-new") == 0) { |
f21e18ca | 1174 | unsigned int i; |
1c6cb603 | 1175 | int rfd, fd; |
f21e18ca | 1176 | unsigned int max = __le32_to_cpu(sb->max_dev); |
f752781f NB |
1177 | |
1178 | for (i=0 ; i < max ; i++) | |
fa7574f6 | 1179 | if (__le16_to_cpu(sb->dev_roles[i]) >= MD_DISK_ROLE_FAULTY) |
f752781f NB |
1180 | break; |
1181 | sb->dev_number = __cpu_to_le32(i); | |
1182 | info->disk.number = i; | |
1183 | if (max >= __le32_to_cpu(sb->max_dev)) | |
1184 | sb->max_dev = __cpu_to_le32(max+1); | |
1185 | ||
1186 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
1187 | read(rfd, sb->device_uuid, 16) != 16) { | |
caa0f6c6 N |
1188 | __u32 r[4] = {random(), random(), random(), random()}; |
1189 | memcpy(sb->device_uuid, r, 16); | |
f752781f | 1190 | } |
7cbeb80e N |
1191 | if (rfd >= 0) |
1192 | close(rfd); | |
f752781f NB |
1193 | |
1194 | sb->dev_roles[i] = | |
1195 | __cpu_to_le16(info->disk.raid_disk); | |
1c6cb603 N |
1196 | |
1197 | fd = open(devname, O_RDONLY); | |
1198 | if (fd >= 0) { | |
1199 | unsigned long long ds; | |
1200 | get_dev_size(fd, devname, &ds); | |
1201 | close(fd); | |
1202 | ds >>= 9; | |
1203 | if (__le64_to_cpu(sb->super_offset) < | |
1204 | __le64_to_cpu(sb->data_offset)) { | |
1205 | sb->data_size = __cpu_to_le64( | |
1206 | ds - __le64_to_cpu(sb->data_offset)); | |
1207 | } else { | |
1208 | ds -= 8*2; | |
1209 | ds &= ~(unsigned long long)(4*2-1); | |
1210 | sb->super_offset = __cpu_to_le64(ds); | |
1211 | sb->data_size = __cpu_to_le64( | |
1212 | ds - __le64_to_cpu(sb->data_offset)); | |
1213 | } | |
1214 | } | |
1e2b2765 | 1215 | } else if (strcmp(update, "linear-grow-update") == 0) { |
82d9eba6 | 1216 | sb->raid_disks = __cpu_to_le32(info->array.raid_disks); |
f752781f NB |
1217 | sb->dev_roles[info->disk.number] = |
1218 | __cpu_to_le16(info->disk.raid_disk); | |
1e2b2765 | 1219 | } else if (strcmp(update, "resync") == 0) { |
82d9eba6 | 1220 | /* make sure resync happens */ |
434b7755 | 1221 | sb->resync_offset = 0ULL; |
1e2b2765 | 1222 | } else if (strcmp(update, "uuid") == 0) { |
350f29f9 | 1223 | copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid); |
bf4fb153 | 1224 | |
14263cf1 NB |
1225 | if (__le32_to_cpu(sb->feature_map)&MD_FEATURE_BITMAP_OFFSET) { |
1226 | struct bitmap_super_s *bm; | |
911cead7 | 1227 | bm = (struct bitmap_super_s*)(st->sb+MAX_SB_SIZE); |
bf4fb153 | 1228 | memcpy(bm->uuid, sb->set_uuid, 16); |
14263cf1 | 1229 | } |
5a31170d N |
1230 | } else if (strcmp(update, "no-bitmap") == 0) { |
1231 | sb->feature_map &= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET); | |
688e99a7 N |
1232 | } else if (strcmp(update, "bbl") == 0) { |
1233 | /* only possible if there is room after the bitmap, or if | |
1234 | * there is no bitmap | |
1235 | */ | |
1236 | unsigned long long sb_offset = __le64_to_cpu(sb->super_offset); | |
1237 | unsigned long long data_offset = __le64_to_cpu(sb->data_offset); | |
d6e4b44f | 1238 | long bitmap_offset = (long)(int32_t)__le32_to_cpu(sb->bitmap_offset); |
688e99a7 N |
1239 | long bm_sectors = 0; |
1240 | long space; | |
1241 | ||
5a23a06e | 1242 | #ifndef MDASSEMBLE |
688e99a7 N |
1243 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { |
1244 | struct bitmap_super_s *bsb; | |
1245 | bsb = (struct bitmap_super_s *)(((char*)sb)+MAX_SB_SIZE); | |
1246 | bm_sectors = bitmap_sectors(bsb); | |
1247 | } | |
5a23a06e | 1248 | #endif |
688e99a7 | 1249 | if (sb_offset < data_offset) { |
f79bbf4f | 1250 | /* 1.1 or 1.2. Put bbl after bitmap leaving at least 32K |
688e99a7 N |
1251 | */ |
1252 | long bb_offset; | |
f79bbf4f | 1253 | bb_offset = sb_offset + 8; |
688e99a7 | 1254 | if (bm_sectors && bitmap_offset > 0) |
f79bbf4f N |
1255 | bb_offset = bitmap_offset + bm_sectors; |
1256 | while (bb_offset < (long)sb_offset + 8 + 32*2 | |
1257 | && bb_offset + 8+8 <= (long)data_offset) | |
1258 | /* too close to bitmap, and room to grow */ | |
1259 | bb_offset += 8; | |
1260 | if (bb_offset + 8 <= (long)data_offset) { | |
688e99a7 N |
1261 | sb->bblog_size = __cpu_to_le16(8); |
1262 | sb->bblog_offset = __cpu_to_le32(bb_offset); | |
1263 | } | |
1264 | } else { | |
1265 | /* 1.0 - Put bbl just before super block */ | |
1266 | if (bm_sectors && bitmap_offset < 0) | |
1267 | space = -bitmap_offset - bm_sectors; | |
1268 | else | |
1269 | space = sb_offset - data_offset - | |
1270 | __le64_to_cpu(sb->data_size); | |
1271 | if (space >= 8) { | |
1272 | sb->bblog_size = __cpu_to_le16(8); | |
1273 | sb->bblog_offset = __cpu_to_le32((unsigned)-8); | |
1274 | } | |
1275 | } | |
1276 | } else if (strcmp(update, "no-bbl") == 0) { | |
1277 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
1278 | pr_err("Cannot remove active bbl from %s\n",devname); | |
1279 | else { | |
1280 | sb->bblog_size = 0; | |
1281 | sb->bblog_shift = 0; | |
1282 | sb->bblog_offset = 0; | |
1283 | } | |
1e2b2765 | 1284 | } else if (strcmp(update, "name") == 0) { |
c4f12c13 NB |
1285 | if (info->name[0] == 0) |
1286 | sprintf(info->name, "%d", info->array.md_minor); | |
1287 | memset(sb->set_name, 0, sizeof(sb->set_name)); | |
1288 | if (homehost && | |
1289 | strchr(info->name, ':') == NULL && | |
1290 | strlen(homehost)+1+strlen(info->name) < 32) { | |
1291 | strcpy(sb->set_name, homehost); | |
1292 | strcat(sb->set_name, ":"); | |
1293 | strcat(sb->set_name, info->name); | |
1294 | } else | |
1295 | strcpy(sb->set_name, info->name); | |
1e2b2765 | 1296 | } else if (strcmp(update, "devicesize") == 0 && |
bee8ec56 NB |
1297 | __le64_to_cpu(sb->super_offset) < |
1298 | __le64_to_cpu(sb->data_offset)) { | |
1299 | /* set data_size to device size less data_offset */ | |
1300 | struct misc_dev_info *misc = (struct misc_dev_info*) | |
911cead7 | 1301 | (st->sb + MAX_SB_SIZE + BM_SUPER_SIZE); |
bee8ec56 NB |
1302 | sb->data_size = __cpu_to_le64( |
1303 | misc->device_size - __le64_to_cpu(sb->data_offset)); | |
199f1a1f N |
1304 | } else if (strcmp(update, "revert-reshape") == 0) { |
1305 | rv = -2; | |
1306 | if (!(sb->feature_map & __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE))) | |
1307 | pr_err("No active reshape to revert on %s\n", | |
1308 | devname); | |
1309 | else { | |
199f1a1f | 1310 | __u32 temp; |
a2836f12 N |
1311 | unsigned long long reshape_sectors; |
1312 | long reshape_chunk; | |
1313 | rv = 0; | |
1314 | /* reshape_position is a little messy. | |
1315 | * Its value must be a multiple of the larger | |
1316 | * chunk size, and of the "after" data disks. | |
1317 | * So when reverting we need to change it to | |
1318 | * be a multiple of the new "after" data disks, | |
1319 | * which is the old "before". | |
1320 | * If it isn't already a multiple of 'before', | |
1321 | * the only thing we could do would be | |
1322 | * copy some block around on the disks, which | |
1323 | * is easy to get wrong. | |
1324 | * So we reject a revert-reshape unless the | |
1325 | * alignment is good. | |
1326 | */ | |
efb3994e N |
1327 | if (__le32_to_cpu(sb->level) >= 4 && |
1328 | __le32_to_cpu(sb->level) <= 6) { | |
1329 | reshape_sectors = __le64_to_cpu(sb->reshape_position); | |
1330 | reshape_chunk = __le32_to_cpu(sb->new_chunk); | |
1331 | reshape_chunk *= __le32_to_cpu(sb->raid_disks) - __le32_to_cpu(sb->delta_disks) - | |
1332 | (__le32_to_cpu(sb->level)==6 ? 2 : 1); | |
1333 | if (reshape_sectors % reshape_chunk) { | |
1334 | pr_err("Reshape position is not suitably aligned.\n"); | |
2bf62891 | 1335 | pr_err("Try normal assembly and stop again\n"); |
efb3994e N |
1336 | return -2; |
1337 | } | |
a2836f12 N |
1338 | } |
1339 | sb->raid_disks = __cpu_to_le32(__le32_to_cpu(sb->raid_disks) - | |
199f1a1f N |
1340 | __le32_to_cpu(sb->delta_disks)); |
1341 | if (sb->delta_disks == 0) | |
1342 | sb->feature_map ^= __cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS); | |
1343 | else | |
1344 | sb->delta_disks = __cpu_to_le32(-__le32_to_cpu(sb->delta_disks)); | |
1345 | ||
1346 | temp = sb->new_layout; | |
1347 | sb->new_layout = sb->layout; | |
1348 | sb->layout = temp; | |
1349 | ||
1350 | temp = sb->new_chunk; | |
1351 | sb->new_chunk = sb->chunksize; | |
1352 | sb->chunksize = temp; | |
1353 | ||
1354 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_NEW_OFFSET)) { | |
419e0182 N |
1355 | long offset_delta = (int32_t)__le32_to_cpu(sb->new_offset); |
1356 | sb->data_offset = __cpu_to_le64(__le64_to_cpu(sb->data_offset) + offset_delta); | |
1357 | sb->new_offset = __cpu_to_le32(-offset_delta); | |
1358 | sb->data_size = __cpu_to_le64(__le64_to_cpu(sb->data_size) - offset_delta); | |
199f1a1f N |
1359 | } |
1360 | } | |
1e2b2765 | 1361 | } else if (strcmp(update, "_reshape_progress")==0) |
353632d9 | 1362 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); |
16715c01 DL |
1363 | else if (strcmp(update, "writemostly")==0) |
1364 | sb->devflags |= WriteMostly1; | |
1365 | else if (strcmp(update, "readwrite")==0) | |
1366 | sb->devflags &= ~WriteMostly1; | |
1e2b2765 N |
1367 | else |
1368 | rv = -1; | |
82d9eba6 NB |
1369 | |
1370 | sb->sb_csum = calc_sb_1_csum(sb); | |
d15a1f72 GJ |
1371 | if (is_clustered(st)) |
1372 | cluster_release_dlmlock(st, lockid); | |
1373 | ||
82d9eba6 NB |
1374 | return rv; |
1375 | } | |
1376 | ||
3da92f27 | 1377 | static int init_super1(struct supertype *st, mdu_array_info_t *info, |
83cd1e97 N |
1378 | unsigned long long size, char *name, char *homehost, |
1379 | int *uuid, unsigned long long data_offset) | |
82d9eba6 | 1380 | { |
6416d527 | 1381 | struct mdp_superblock_1 *sb; |
82d9eba6 | 1382 | int spares; |
34163fc7 | 1383 | int rfd; |
05697ec1 | 1384 | char defname[10]; |
911cead7 | 1385 | int sbsize; |
6416d527 | 1386 | |
3c0bcd46 | 1387 | if (posix_memalign((void**)&sb, 4096, SUPER1_SIZE) != 0) { |
1ade5cc1 | 1388 | pr_err("could not allocate superblock\n"); |
3d2c4fc7 DW |
1389 | return 0; |
1390 | } | |
1afa9308 | 1391 | memset(sb, 0, SUPER1_SIZE); |
82d9eba6 | 1392 | |
64557c33 | 1393 | st->sb = sb; |
ba7eb04f | 1394 | if (info == NULL) { |
82d9eba6 NB |
1395 | /* zeroing superblock */ |
1396 | return 0; | |
05697ec1 | 1397 | } |
82d9eba6 NB |
1398 | |
1399 | spares = info->working_disks - info->active_disks; | |
911cead7 | 1400 | if (info->raid_disks + spares > MAX_DEVS) { |
e7b84f9d | 1401 | pr_err("too many devices requested: %d+%d > %d\n", |
911cead7 | 1402 | info->raid_disks , spares, MAX_DEVS); |
82d9eba6 NB |
1403 | return 0; |
1404 | } | |
1405 | ||
82d9eba6 NB |
1406 | sb->magic = __cpu_to_le32(MD_SB_MAGIC); |
1407 | sb->major_version = __cpu_to_le32(1); | |
1408 | sb->feature_map = 0; | |
1409 | sb->pad0 = 0; | |
34163fc7 | 1410 | |
350f29f9 NB |
1411 | if (uuid) |
1412 | copy_uuid(sb->set_uuid, uuid, super1.swapuuid); | |
1413 | else { | |
3d3dd91e NB |
1414 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || |
1415 | read(rfd, sb->set_uuid, 16) != 16) { | |
caa0f6c6 N |
1416 | __u32 r[4] = {random(), random(), random(), random()}; |
1417 | memcpy(sb->set_uuid, r, 16); | |
3d3dd91e NB |
1418 | } |
1419 | if (rfd >= 0) close(rfd); | |
34163fc7 | 1420 | } |
82d9eba6 | 1421 | |
05697ec1 NB |
1422 | if (name == NULL || *name == 0) { |
1423 | sprintf(defname, "%d", info->md_minor); | |
1424 | name = defname; | |
1425 | } | |
05697ec1 NB |
1426 | if (homehost && |
1427 | strchr(name, ':')== NULL && | |
1428 | strlen(homehost)+1+strlen(name) < 32) { | |
1429 | strcpy(sb->set_name, homehost); | |
1430 | strcat(sb->set_name, ":"); | |
1431 | strcat(sb->set_name, name); | |
1432 | } else | |
1433 | strcpy(sb->set_name, name); | |
82d9eba6 NB |
1434 | |
1435 | sb->ctime = __cpu_to_le64((unsigned long long)time(0)); | |
1436 | sb->level = __cpu_to_le32(info->level); | |
ea329559 | 1437 | sb->layout = __cpu_to_le32(info->layout); |
5dd497ee | 1438 | sb->size = __cpu_to_le64(size*2ULL); |
82d9eba6 NB |
1439 | sb->chunksize = __cpu_to_le32(info->chunk_size>>9); |
1440 | sb->raid_disks = __cpu_to_le32(info->raid_disks); | |
1441 | ||
83cd1e97 | 1442 | sb->data_offset = __cpu_to_le64(data_offset); |
82d9eba6 NB |
1443 | sb->data_size = __cpu_to_le64(0); |
1444 | sb->super_offset = __cpu_to_le64(0); | |
1445 | sb->recovery_offset = __cpu_to_le64(0); | |
1446 | ||
1447 | sb->utime = sb->ctime; | |
1448 | sb->events = __cpu_to_le64(1); | |
34321279 | 1449 | if (info->state & (1<<MD_SB_CLEAN)) |
b7528a20 | 1450 | sb->resync_offset = MaxSector; |
82d9eba6 NB |
1451 | else |
1452 | sb->resync_offset = 0; | |
911cead7 N |
1453 | sbsize = sizeof(struct mdp_superblock_1) + 2 * (info->raid_disks + spares); |
1454 | sbsize = ROUND_UP(sbsize, 512); | |
1455 | sb->max_dev = __cpu_to_le32((sbsize - sizeof(struct mdp_superblock_1)) / 2); | |
82d9eba6 | 1456 | |
911cead7 | 1457 | memset(sb->dev_roles, 0xff, MAX_SB_SIZE - sizeof(struct mdp_superblock_1)); |
82d9eba6 | 1458 | |
82d9eba6 NB |
1459 | return 1; |
1460 | } | |
1461 | ||
111d01fc NB |
1462 | struct devinfo { |
1463 | int fd; | |
1464 | char *devname; | |
72ca9bcf | 1465 | long long data_offset; |
111d01fc NB |
1466 | mdu_disk_info_t disk; |
1467 | struct devinfo *next; | |
1468 | }; | |
0e600426 | 1469 | #ifndef MDASSEMBLE |
82d9eba6 | 1470 | /* Add a device to the superblock being created */ |
f20c3968 | 1471 | static int add_to_super1(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf | 1472 | int fd, char *devname, unsigned long long data_offset) |
82d9eba6 | 1473 | { |
3da92f27 | 1474 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 | 1475 | __u16 *rp = sb->dev_roles + dk->number; |
111d01fc | 1476 | struct devinfo *di, **dip; |
d15a1f72 GJ |
1477 | int rv, lockid; |
1478 | ||
1479 | if (is_clustered(st)) { | |
1480 | rv = cluster_get_dlmlock(st, &lockid); | |
1481 | if (rv) { | |
1482 | pr_err("Cannot get dlmlock in %s return %d\n", __func__, rv); | |
1483 | cluster_release_dlmlock(st, lockid); | |
1484 | return rv; | |
1485 | } | |
1486 | } | |
111d01fc | 1487 | |
dfd4d8ee | 1488 | if ((dk->state & 6) == 6) /* active, sync */ |
82d9eba6 | 1489 | *rp = __cpu_to_le16(dk->raid_disk); |
cc1799c3 SL |
1490 | else if (dk->state & (1<<MD_DISK_JOURNAL)) |
1491 | *rp = MD_DISK_ROLE_JOURNAL; | |
412ca2e5 | 1492 | else if ((dk->state & ~2) == 0) /* active or idle -> spare */ |
fa7574f6 | 1493 | *rp = MD_DISK_ROLE_SPARE; |
34163fc7 | 1494 | else |
fa7574f6 | 1495 | *rp = MD_DISK_ROLE_FAULTY; |
111d01fc | 1496 | |
f21e18ca | 1497 | if (dk->number >= (int)__le32_to_cpu(sb->max_dev) && |
911cead7 | 1498 | __le32_to_cpu(sb->max_dev) < MAX_DEVS) |
9df04a4f | 1499 | sb->max_dev = __cpu_to_le32(dk->number+1); |
8844e291 | 1500 | |
d2ca6449 | 1501 | sb->dev_number = __cpu_to_le32(dk->number); |
16715c01 | 1502 | sb->devflags = 0; /* don't copy another disks flags */ |
d2ca6449 NB |
1503 | sb->sb_csum = calc_sb_1_csum(sb); |
1504 | ||
111d01fc NB |
1505 | dip = (struct devinfo **)&st->info; |
1506 | while (*dip) | |
1507 | dip = &(*dip)->next; | |
503975b9 | 1508 | di = xmalloc(sizeof(struct devinfo)); |
111d01fc NB |
1509 | di->fd = fd; |
1510 | di->devname = devname; | |
1511 | di->disk = *dk; | |
72ca9bcf | 1512 | di->data_offset = data_offset; |
111d01fc NB |
1513 | di->next = NULL; |
1514 | *dip = di; | |
f20c3968 | 1515 | |
d15a1f72 GJ |
1516 | if (is_clustered(st)) |
1517 | cluster_release_dlmlock(st, lockid); | |
1518 | ||
f20c3968 | 1519 | return 0; |
82d9eba6 | 1520 | } |
0e600426 | 1521 | #endif |
82d9eba6 | 1522 | |
53e76b1d | 1523 | static int locate_bitmap1(struct supertype *st, int fd); |
14263cf1 | 1524 | |
3da92f27 | 1525 | static int store_super1(struct supertype *st, int fd) |
82d9eba6 | 1526 | { |
3da92f27 | 1527 | struct mdp_superblock_1 *sb = st->sb; |
6fbba4c9 | 1528 | unsigned long long sb_offset; |
308340aa | 1529 | struct align_fd afd; |
82d9eba6 | 1530 | int sbsize; |
5dd497ee | 1531 | unsigned long long dsize; |
d15a1f72 GJ |
1532 | int rv, lockid; |
1533 | ||
1534 | if (is_clustered(st)) { | |
1535 | rv = cluster_get_dlmlock(st, &lockid); | |
1536 | if (rv) { | |
1537 | pr_err("Cannot get dlmlock in %s return %d\n", __func__, rv); | |
1538 | cluster_release_dlmlock(st, lockid); | |
1539 | return rv; | |
1540 | } | |
1541 | } | |
96395475 | 1542 | |
beae1dfe NB |
1543 | if (!get_dev_size(fd, NULL, &dsize)) |
1544 | return 1; | |
1545 | ||
1546 | dsize >>= 9; | |
96395475 | 1547 | |
5dd497ee | 1548 | if (dsize < 24) |
96395475 NB |
1549 | return 2; |
1550 | ||
308340aa JS |
1551 | init_afd(&afd, fd); |
1552 | ||
96395475 NB |
1553 | /* |
1554 | * Calculate the position of the superblock. | |
1555 | * It is always aligned to a 4K boundary and | |
1556 | * depending on minor_version, it can be: | |
1557 | * 0: At least 8K, but less than 12K, from end of device | |
1558 | * 1: At start of device | |
1559 | * 2: 4K from start of device. | |
1560 | */ | |
1561 | switch(st->minor_version) { | |
1562 | case 0: | |
5dd497ee | 1563 | sb_offset = dsize; |
96395475 NB |
1564 | sb_offset -= 8*2; |
1565 | sb_offset &= ~(4*2-1); | |
1566 | break; | |
1567 | case 1: | |
6fbba4c9 | 1568 | sb_offset = 0; |
96395475 NB |
1569 | break; |
1570 | case 2: | |
1571 | sb_offset = 4*2; | |
1572 | break; | |
1573 | default: | |
1574 | return -EINVAL; | |
1575 | } | |
1576 | ||
6fbba4c9 NB |
1577 | if (sb_offset != __le64_to_cpu(sb->super_offset) && |
1578 | 0 != __le64_to_cpu(sb->super_offset) | |
96395475 | 1579 | ) { |
e7b84f9d | 1580 | pr_err("internal error - sb_offset is wrong\n"); |
96395475 NB |
1581 | abort(); |
1582 | } | |
82d9eba6 | 1583 | |
6fbba4c9 | 1584 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) |
82d9eba6 NB |
1585 | return 3; |
1586 | ||
0a2f1894 | 1587 | sbsize = ROUND_UP(sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev), 512); |
82d9eba6 | 1588 | |
308340aa | 1589 | if (awrite(&afd, sb, sbsize) != sbsize) |
82d9eba6 NB |
1590 | return 4; |
1591 | ||
14263cf1 NB |
1592 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { |
1593 | struct bitmap_super_s *bm = (struct bitmap_super_s*) | |
911cead7 | 1594 | (((char*)sb)+MAX_SB_SIZE); |
14263cf1 | 1595 | if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) { |
3da92f27 | 1596 | locate_bitmap1(st, fd); |
308340aa JS |
1597 | if (awrite(&afd, bm, sizeof(*bm)) != sizeof(*bm)) |
1598 | return 5; | |
14263cf1 NB |
1599 | } |
1600 | } | |
570c0542 | 1601 | fsync(fd); |
d15a1f72 GJ |
1602 | if (is_clustered(st)) |
1603 | cluster_release_dlmlock(st, lockid); | |
1604 | ||
82d9eba6 NB |
1605 | return 0; |
1606 | } | |
1607 | ||
3da92f27 | 1608 | static int load_super1(struct supertype *st, int fd, char *devname); |
892debc8 | 1609 | |
2fb749d1 NB |
1610 | static unsigned long choose_bm_space(unsigned long devsize) |
1611 | { | |
1612 | /* if the device is bigger than 8Gig, save 64k for bitmap usage, | |
1613 | * if bigger than 200Gig, save 128k | |
a380e275 N |
1614 | * NOTE: result must be multiple of 4K else bad things happen |
1615 | * on 4K-sector devices. | |
2fb749d1 | 1616 | */ |
69646c14 | 1617 | if (devsize < 64*2) return 0; |
2fb749d1 NB |
1618 | if (devsize - 64*2 >= 200*1024*1024*2) |
1619 | return 128*2; | |
1620 | if (devsize - 4*2 > 8*1024*1024*2) | |
1621 | return 64*2; | |
1622 | return 4*2; | |
1623 | } | |
1624 | ||
e8090005 N |
1625 | static void free_super1(struct supertype *st); |
1626 | ||
cc1799c3 | 1627 | #define META_BLOCK_SIZE 4096 |
198d5478 | 1628 | __u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len); |
cc1799c3 SL |
1629 | |
1630 | static int write_empty_r5l_meta_block(struct supertype *st, int fd) | |
1631 | { | |
1632 | struct r5l_meta_block *mb; | |
1633 | struct mdp_superblock_1 *sb = st->sb; | |
1634 | struct align_fd afd; | |
1635 | __u32 crc; | |
1636 | ||
1637 | init_afd(&afd, fd); | |
1638 | ||
1639 | if (posix_memalign((void**)&mb, 4096, META_BLOCK_SIZE) != 0) { | |
1640 | pr_err("Could not allocate memory for the meta block.\n"); | |
1641 | return 1; | |
1642 | } | |
1643 | ||
1644 | memset(mb, 0, META_BLOCK_SIZE); | |
1645 | ||
1646 | mb->magic = __cpu_to_le32(R5LOG_MAGIC); | |
1647 | mb->version = R5LOG_VERSION; | |
1648 | mb->meta_size = __cpu_to_le32(sizeof(struct r5l_meta_block)); | |
1649 | mb->seq = __cpu_to_le64(random32()); | |
1650 | mb->position = __cpu_to_le64(0); | |
1651 | ||
198d5478 SL |
1652 | crc = crc32c_le(0xffffffff, sb->set_uuid, sizeof(sb->set_uuid)); |
1653 | crc = crc32c_le(crc, (void *)mb, META_BLOCK_SIZE); | |
1654 | mb->checksum = crc; | |
cc1799c3 SL |
1655 | |
1656 | if (lseek64(fd, (sb->data_offset) * 512, 0) < 0LL) { | |
1657 | pr_err("cannot seek to offset of the meta block\n"); | |
1658 | goto fail_to_write; | |
1659 | } | |
1660 | ||
1661 | if (awrite(&afd, mb, META_BLOCK_SIZE) != META_BLOCK_SIZE) { | |
1662 | pr_err("failed to store write the meta block \n"); | |
1663 | goto fail_to_write; | |
1664 | } | |
1665 | fsync(fd); | |
1666 | ||
1667 | free(mb); | |
1668 | return 0; | |
1669 | ||
1670 | fail_to_write: | |
1671 | free(mb); | |
1672 | return 1; | |
1673 | } | |
1674 | ||
111d01fc NB |
1675 | #ifndef MDASSEMBLE |
1676 | static int write_init_super1(struct supertype *st) | |
82d9eba6 | 1677 | { |
3da92f27 | 1678 | struct mdp_superblock_1 *sb = st->sb; |
e8090005 | 1679 | struct supertype *refst; |
dfe47e00 | 1680 | int rfd; |
111d01fc | 1681 | int rv = 0; |
f21e18ca | 1682 | unsigned long long bm_space; |
111d01fc | 1683 | struct devinfo *di; |
8fac0577 | 1684 | unsigned long long dsize, array_size; |
23bf42cc | 1685 | unsigned long long sb_offset; |
83cd1e97 | 1686 | unsigned long long data_offset; |
82d9eba6 | 1687 | |
cc1799c3 SL |
1688 | for (di = st->info; di; di = di->next) { |
1689 | if (di->disk.state & (1 << MD_DISK_JOURNAL)) | |
1690 | sb->feature_map |= MD_FEATURE_JOURNAL; | |
1691 | } | |
1692 | ||
4687f160 | 1693 | for (di = st->info; di; di = di->next) { |
f4dc5e9b | 1694 | if (di->disk.state & (1 << MD_DISK_FAULTY)) |
111d01fc | 1695 | continue; |
a0c8a17f NB |
1696 | if (di->fd < 0) |
1697 | continue; | |
82d9eba6 | 1698 | |
ba728be7 | 1699 | while (Kill(di->devname, NULL, 0, -1, 1) == 0) |
9277cc77 | 1700 | ; |
82d9eba6 | 1701 | |
111d01fc NB |
1702 | sb->dev_number = __cpu_to_le32(di->disk.number); |
1703 | if (di->disk.state & (1<<MD_DISK_WRITEMOSTLY)) | |
adbb382b | 1704 | sb->devflags |= WriteMostly1; |
9a88e7b6 | 1705 | else |
adbb382b | 1706 | sb->devflags &= ~WriteMostly1; |
892debc8 | 1707 | |
111d01fc NB |
1708 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || |
1709 | read(rfd, sb->device_uuid, 16) != 16) { | |
83208785 N |
1710 | __u32 r[4] = {random(), random(), random(), random()}; |
1711 | memcpy(sb->device_uuid, r, 16); | |
111d01fc | 1712 | } |
83208785 N |
1713 | if (rfd >= 0) |
1714 | close(rfd); | |
1715 | ||
01290056 SL |
1716 | if (!(di->disk.state & (1<<MD_DISK_JOURNAL))) |
1717 | sb->events = 0; | |
111d01fc | 1718 | |
e8090005 | 1719 | refst = dup_super(st); |
1011e834 | 1720 | if (load_super1(refst, di->fd, NULL)==0) { |
e8090005 | 1721 | struct mdp_superblock_1 *refsb = refst->sb; |
111d01fc NB |
1722 | |
1723 | memcpy(sb->device_uuid, refsb->device_uuid, 16); | |
1724 | if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) { | |
1725 | /* same array, so preserve events and | |
1726 | * dev_number */ | |
1727 | sb->events = refsb->events; | |
1728 | /* bugs in 2.6.17 and earlier mean the | |
1729 | * dev_number chosen in Manage must be preserved | |
1730 | */ | |
1731 | if (get_linux_version() >= 2006018) | |
1732 | sb->dev_number = refsb->dev_number; | |
1733 | } | |
e8090005 | 1734 | free_super1(refst); |
892debc8 | 1735 | } |
e8090005 | 1736 | free(refst); |
34163fc7 | 1737 | |
40114213 JS |
1738 | if (!get_dev_size(di->fd, NULL, &dsize)) { |
1739 | rv = 1; | |
1740 | goto error_out; | |
1741 | } | |
111d01fc | 1742 | dsize >>= 9; |
82d9eba6 | 1743 | |
111d01fc NB |
1744 | if (dsize < 24) { |
1745 | close(di->fd); | |
40114213 JS |
1746 | rv = 2; |
1747 | goto error_out; | |
111d01fc | 1748 | } |
82d9eba6 | 1749 | |
111d01fc NB |
1750 | /* |
1751 | * Calculate the position of the superblock. | |
1752 | * It is always aligned to a 4K boundary and | |
1753 | * depending on minor_version, it can be: | |
1754 | * 0: At least 8K, but less than 12K, from end of device | |
1755 | * 1: At start of device | |
1756 | * 2: 4K from start of device. | |
23bf42cc | 1757 | * data_offset has already been set. |
111d01fc NB |
1758 | */ |
1759 | array_size = __le64_to_cpu(sb->size); | |
bf95d0f3 N |
1760 | /* work out how much space we left for a bitmap, |
1761 | * Add 8 sectors for bad block log */ | |
1762 | bm_space = choose_bm_space(array_size) + 8; | |
111d01fc | 1763 | |
72ca9bcf | 1764 | data_offset = di->data_offset; |
23bf42cc N |
1765 | if (data_offset == INVALID_SECTORS) |
1766 | data_offset = st->data_offset; | |
111d01fc NB |
1767 | switch(st->minor_version) { |
1768 | case 0: | |
23bf42cc N |
1769 | if (data_offset == INVALID_SECTORS) |
1770 | data_offset = 0; | |
111d01fc NB |
1771 | sb_offset = dsize; |
1772 | sb_offset -= 8*2; | |
1773 | sb_offset &= ~(4*2-1); | |
4c0ea7b0 | 1774 | sb->data_offset = __cpu_to_le64(data_offset); |
111d01fc | 1775 | sb->super_offset = __cpu_to_le64(sb_offset); |
f21e18ca | 1776 | if (sb_offset < array_size + bm_space) |
a380e275 | 1777 | bm_space = sb_offset - array_size; |
111d01fc | 1778 | sb->data_size = __cpu_to_le64(sb_offset - bm_space); |
bf95d0f3 N |
1779 | if (bm_space >= 8) { |
1780 | sb->bblog_size = __cpu_to_le16(8); | |
1781 | sb->bblog_offset = __cpu_to_le32((unsigned)-8); | |
1782 | } | |
111d01fc NB |
1783 | break; |
1784 | case 1: | |
1785 | sb->super_offset = __cpu_to_le64(0); | |
23bf42cc N |
1786 | if (data_offset == INVALID_SECTORS) |
1787 | data_offset = 16; | |
1788 | ||
1789 | sb->data_offset = __cpu_to_le64(data_offset); | |
1790 | sb->data_size = __cpu_to_le64(dsize - data_offset); | |
1791 | if (data_offset >= 8 + 32*2 + 8) { | |
f79bbf4f N |
1792 | sb->bblog_size = __cpu_to_le16(8); |
1793 | sb->bblog_offset = __cpu_to_le32(8 + 32*2); | |
23bf42cc | 1794 | } else if (data_offset >= 16) { |
bf95d0f3 | 1795 | sb->bblog_size = __cpu_to_le16(8); |
23bf42cc | 1796 | sb->bblog_offset = __cpu_to_le32(data_offset-8); |
bf95d0f3 | 1797 | } |
111d01fc NB |
1798 | break; |
1799 | case 2: | |
1800 | sb_offset = 4*2; | |
23bf42cc N |
1801 | sb->super_offset = __cpu_to_le64(sb_offset); |
1802 | if (data_offset == INVALID_SECTORS) | |
1803 | data_offset = 24; | |
1804 | ||
1805 | sb->data_offset = __cpu_to_le64(data_offset); | |
1806 | sb->data_size = __cpu_to_le64(dsize - data_offset); | |
1807 | if (data_offset >= 16 + 32*2 + 8) { | |
f79bbf4f N |
1808 | sb->bblog_size = __cpu_to_le16(8); |
1809 | sb->bblog_offset = __cpu_to_le32(8 + 32*2); | |
23bf42cc | 1810 | } else if (data_offset >= 16+16) { |
bf95d0f3 N |
1811 | sb->bblog_size = __cpu_to_le16(8); |
1812 | /* '8' sectors for the bblog, and another '8' | |
1813 | * because we want offset from superblock, not | |
1814 | * start of device. | |
1815 | */ | |
23bf42cc | 1816 | sb->bblog_offset = __cpu_to_le32(data_offset-8-8); |
bf95d0f3 | 1817 | } |
111d01fc NB |
1818 | break; |
1819 | default: | |
7a862a02 | 1820 | pr_err("Failed to write invalid metadata format 1.%i to %s\n", |
e7b84f9d | 1821 | st->minor_version, di->devname); |
40114213 JS |
1822 | rv = -EINVAL; |
1823 | goto out; | |
111d01fc | 1824 | } |
86a406c2 N |
1825 | /* Disable badblock log on clusters, or when explicitly requested */ |
1826 | if (st->nodes > 0 || conf_get_create_info()->bblist == 0) { | |
e2efe9e7 N |
1827 | sb->bblog_size = 0; |
1828 | sb->bblog_offset = 0; | |
1829 | } | |
82d9eba6 | 1830 | |
111d01fc NB |
1831 | sb->sb_csum = calc_sb_1_csum(sb); |
1832 | rv = store_super1(st, di->fd); | |
cc1799c3 SL |
1833 | |
1834 | if (rv == 0 && (di->disk.state & (1 << MD_DISK_JOURNAL))) { | |
1835 | rv = write_empty_r5l_meta_block(st, di->fd); | |
1836 | if (rv) | |
1837 | goto error_out; | |
1838 | } | |
1839 | ||
111d01fc | 1840 | if (rv == 0 && (__le32_to_cpu(sb->feature_map) & 1)) |
0aa2f15b | 1841 | rv = st->ss->write_bitmap(st, di->fd, NoUpdate); |
111d01fc NB |
1842 | close(di->fd); |
1843 | di->fd = -1; | |
4687f160 | 1844 | if (rv) |
1845 | goto error_out; | |
111d01fc | 1846 | } |
40114213 JS |
1847 | error_out: |
1848 | if (rv) | |
e7b84f9d N |
1849 | pr_err("Failed to write metadata to %s\n", |
1850 | di->devname); | |
40114213 | 1851 | out: |
82d9eba6 NB |
1852 | return rv; |
1853 | } | |
111d01fc | 1854 | #endif |
82d9eba6 | 1855 | |
64557c33 | 1856 | static int compare_super1(struct supertype *st, struct supertype *tst) |
82d9eba6 NB |
1857 | { |
1858 | /* | |
1859 | * return: | |
1860 | * 0 same, or first was empty, and second was copied | |
1861 | * 1 second had wrong number | |
1862 | * 2 wrong uuid | |
1863 | * 3 wrong other info | |
1864 | */ | |
64557c33 NB |
1865 | struct mdp_superblock_1 *first = st->sb; |
1866 | struct mdp_superblock_1 *second = tst->sb; | |
82d9eba6 NB |
1867 | |
1868 | if (second->magic != __cpu_to_le32(MD_SB_MAGIC)) | |
1869 | return 1; | |
1870 | if (second->major_version != __cpu_to_le32(1)) | |
1871 | return 1; | |
1872 | ||
1873 | if (!first) { | |
3c0bcd46 | 1874 | if (posix_memalign((void**)&first, 4096, SUPER1_SIZE) != 0) { |
1ade5cc1 | 1875 | pr_err("could not allocate superblock\n"); |
3d2c4fc7 DW |
1876 | return 1; |
1877 | } | |
41226756 | 1878 | memcpy(first, second, SUPER1_SIZE); |
64557c33 | 1879 | st->sb = first; |
82d9eba6 NB |
1880 | return 0; |
1881 | } | |
1882 | if (memcmp(first->set_uuid, second->set_uuid, 16)!= 0) | |
1883 | return 2; | |
1884 | ||
1885 | if (first->ctime != second->ctime || | |
1886 | first->level != second->level || | |
1887 | first->layout != second->layout || | |
1888 | first->size != second->size || | |
1889 | first->chunksize != second->chunksize || | |
1890 | first->raid_disks != second->raid_disks) | |
1891 | return 3; | |
1892 | return 0; | |
1893 | } | |
1894 | ||
3da92f27 | 1895 | static int load_super1(struct supertype *st, int fd, char *devname) |
82d9eba6 | 1896 | { |
5dd497ee | 1897 | unsigned long long dsize; |
82d9eba6 NB |
1898 | unsigned long long sb_offset; |
1899 | struct mdp_superblock_1 *super; | |
14263cf1 NB |
1900 | int uuid[4]; |
1901 | struct bitmap_super_s *bsb; | |
bee8ec56 | 1902 | struct misc_dev_info *misc; |
308340aa | 1903 | struct align_fd afd; |
82d9eba6 | 1904 | |
3da92f27 | 1905 | free_super1(st); |
82d9eba6 | 1906 | |
308340aa JS |
1907 | init_afd(&afd, fd); |
1908 | ||
a17a3de3 | 1909 | if (st->ss == NULL || st->minor_version == -1) { |
570c0542 | 1910 | int bestvers = -1; |
23dc1ae8 | 1911 | struct supertype tst; |
570c0542 NB |
1912 | __u64 bestctime = 0; |
1913 | /* guess... choose latest ctime */ | |
ef609477 | 1914 | memset(&tst, 0, sizeof(tst)); |
23dc1ae8 NB |
1915 | tst.ss = &super1; |
1916 | for (tst.minor_version = 0; tst.minor_version <= 2 ; tst.minor_version++) { | |
3da92f27 NB |
1917 | switch(load_super1(&tst, fd, devname)) { |
1918 | case 0: super = tst.sb; | |
570c0542 NB |
1919 | if (bestvers == -1 || |
1920 | bestctime < __le64_to_cpu(super->ctime)) { | |
23dc1ae8 | 1921 | bestvers = tst.minor_version; |
570c0542 NB |
1922 | bestctime = __le64_to_cpu(super->ctime); |
1923 | } | |
1924 | free(super); | |
3da92f27 | 1925 | tst.sb = NULL; |
570c0542 | 1926 | break; |
23dc1ae8 | 1927 | case 1: return 1; /*bad device */ |
82d9eba6 NB |
1928 | case 2: break; /* bad, try next */ |
1929 | } | |
1930 | } | |
570c0542 NB |
1931 | if (bestvers != -1) { |
1932 | int rv; | |
23dc1ae8 NB |
1933 | tst.minor_version = bestvers; |
1934 | tst.ss = &super1; | |
911cead7 | 1935 | tst.max_devs = MAX_DEVS; |
3da92f27 | 1936 | rv = load_super1(&tst, fd, devname); |
23dc1ae8 NB |
1937 | if (rv == 0) |
1938 | *st = tst; | |
570c0542 NB |
1939 | return rv; |
1940 | } | |
82d9eba6 NB |
1941 | return 2; |
1942 | } | |
beae1dfe NB |
1943 | if (!get_dev_size(fd, devname, &dsize)) |
1944 | return 1; | |
1945 | dsize >>= 9; | |
82d9eba6 | 1946 | |
5dd497ee | 1947 | if (dsize < 24) { |
82d9eba6 | 1948 | if (devname) |
e7b84f9d | 1949 | pr_err("%s is too small for md: size is %llu sectors.\n", |
5dd497ee | 1950 | devname, dsize); |
82d9eba6 NB |
1951 | return 1; |
1952 | } | |
1953 | ||
1954 | /* | |
1955 | * Calculate the position of the superblock. | |
1956 | * It is always aligned to a 4K boundary and | |
14263cf1 | 1957 | * depending on minor_version, it can be: |
82d9eba6 NB |
1958 | * 0: At least 8K, but less than 12K, from end of device |
1959 | * 1: At start of device | |
1960 | * 2: 4K from start of device. | |
1961 | */ | |
1962 | switch(st->minor_version) { | |
1963 | case 0: | |
5dd497ee | 1964 | sb_offset = dsize; |
82d9eba6 NB |
1965 | sb_offset -= 8*2; |
1966 | sb_offset &= ~(4*2-1); | |
1967 | break; | |
1968 | case 1: | |
1969 | sb_offset = 0; | |
1970 | break; | |
1971 | case 2: | |
1972 | sb_offset = 4*2; | |
1973 | break; | |
1974 | default: | |
1975 | return -EINVAL; | |
1976 | } | |
1977 | ||
82d9eba6 NB |
1978 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) { |
1979 | if (devname) | |
e7b84f9d | 1980 | pr_err("Cannot seek to superblock on %s: %s\n", |
82d9eba6 NB |
1981 | devname, strerror(errno)); |
1982 | return 1; | |
1983 | } | |
1984 | ||
3c0bcd46 | 1985 | if (posix_memalign((void**)&super, 4096, SUPER1_SIZE) != 0) { |
1ade5cc1 | 1986 | pr_err("could not allocate superblock\n"); |
3d2c4fc7 DW |
1987 | return 1; |
1988 | } | |
82d9eba6 | 1989 | |
308340aa | 1990 | if (aread(&afd, super, MAX_SB_SIZE) != MAX_SB_SIZE) { |
82d9eba6 | 1991 | if (devname) |
e7b84f9d | 1992 | pr_err("Cannot read superblock on %s\n", |
82d9eba6 NB |
1993 | devname); |
1994 | free(super); | |
1995 | return 1; | |
1996 | } | |
1997 | ||
1998 | if (__le32_to_cpu(super->magic) != MD_SB_MAGIC) { | |
1999 | if (devname) | |
e7b84f9d | 2000 | pr_err("No super block found on %s (Expected magic %08x, got %08x)\n", |
82d9eba6 NB |
2001 | devname, MD_SB_MAGIC, __le32_to_cpu(super->magic)); |
2002 | free(super); | |
2003 | return 2; | |
2004 | } | |
2005 | ||
2006 | if (__le32_to_cpu(super->major_version) != 1) { | |
2007 | if (devname) | |
e7b84f9d | 2008 | pr_err("Cannot interpret superblock on %s - version is %d\n", |
82d9eba6 NB |
2009 | devname, __le32_to_cpu(super->major_version)); |
2010 | free(super); | |
2011 | return 2; | |
2012 | } | |
2013 | if (__le64_to_cpu(super->super_offset) != sb_offset) { | |
2014 | if (devname) | |
e7b84f9d | 2015 | pr_err("No superblock found on %s (super_offset is wrong)\n", |
82d9eba6 NB |
2016 | devname); |
2017 | free(super); | |
2018 | return 2; | |
2019 | } | |
64557c33 | 2020 | st->sb = super; |
14263cf1 | 2021 | |
911cead7 | 2022 | bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); |
bee8ec56 | 2023 | |
911cead7 | 2024 | misc = (struct misc_dev_info*) (((char*)super)+MAX_SB_SIZE+BM_SUPER_SIZE); |
bee8ec56 | 2025 | misc->device_size = dsize; |
23bf42cc N |
2026 | if (st->data_offset == INVALID_SECTORS) |
2027 | st->data_offset = __le64_to_cpu(super->data_offset); | |
bee8ec56 | 2028 | |
14263cf1 NB |
2029 | /* Now check on the bitmap superblock */ |
2030 | if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0) | |
2031 | return 0; | |
2032 | /* Read the bitmap superblock and make sure it looks | |
2033 | * valid. If it doesn't clear the bit. An --assemble --force | |
2034 | * should get that written out. | |
2035 | */ | |
3da92f27 | 2036 | locate_bitmap1(st, fd); |
308340aa | 2037 | if (aread(&afd, bsb, 512) != 512) |
14263cf1 NB |
2038 | goto no_bitmap; |
2039 | ||
3da92f27 | 2040 | uuid_from_super1(st, uuid); |
14263cf1 NB |
2041 | if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC || |
2042 | memcmp(bsb->uuid, uuid, 16) != 0) | |
2043 | goto no_bitmap; | |
2044 | return 0; | |
2045 | ||
2046 | no_bitmap: | |
c0c1acd6 N |
2047 | super->feature_map = __cpu_to_le32(__le32_to_cpu(super->feature_map) |
2048 | & ~MD_FEATURE_BITMAP_OFFSET); | |
82d9eba6 NB |
2049 | return 0; |
2050 | } | |
2051 | ||
82d9eba6 NB |
2052 | static struct supertype *match_metadata_desc1(char *arg) |
2053 | { | |
503975b9 | 2054 | struct supertype *st = xcalloc(1, sizeof(*st)); |
82d9eba6 | 2055 | |
4dd2df09 | 2056 | st->container_devnm[0] = 0; |
82d9eba6 | 2057 | st->ss = &super1; |
911cead7 | 2058 | st->max_devs = MAX_DEVS; |
64557c33 | 2059 | st->sb = NULL; |
23bf42cc | 2060 | st->data_offset = INVALID_SECTORS; |
9b2a22d3 N |
2061 | /* leading zeros can be safely ignored. --detail generates them. */ |
2062 | while (*arg == '0') | |
2063 | arg++; | |
2064 | if (strcmp(arg, "1.0") == 0 || | |
2065 | strcmp(arg, "1.00") == 0) { | |
82d9eba6 NB |
2066 | st->minor_version = 0; |
2067 | return st; | |
2068 | } | |
9b2a22d3 | 2069 | if (strcmp(arg, "1.1") == 0 || |
50827504 | 2070 | strcmp(arg, "1.01") == 0 |
7d5c3964 | 2071 | ) { |
82d9eba6 NB |
2072 | st->minor_version = 1; |
2073 | return st; | |
2074 | } | |
9b2a22d3 | 2075 | if (strcmp(arg, "1.2") == 0 || |
26f467a9 | 2076 | #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */ |
ad90adb6 | 2077 | strcmp(arg, "default") == 0 || |
26f467a9 | 2078 | #endif /* DEFAULT_OLD_METADATA */ |
9b2a22d3 | 2079 | strcmp(arg, "1.02") == 0) { |
82d9eba6 NB |
2080 | st->minor_version = 2; |
2081 | return st; | |
2082 | } | |
a17a3de3 | 2083 | if (strcmp(arg, "1") == 0 || |
17f25ca6 | 2084 | strcmp(arg, "default") == 0) { |
a17a3de3 DL |
2085 | st->minor_version = -1; |
2086 | return st; | |
2087 | } | |
82d9eba6 NB |
2088 | |
2089 | free(st); | |
2090 | return NULL; | |
2091 | } | |
2092 | ||
34163fc7 NB |
2093 | /* find available size on device with this devsize, using |
2094 | * superblock type st, and reserving 'reserve' sectors for | |
2095 | * a possible bitmap | |
2096 | */ | |
641da745 N |
2097 | static __u64 avail_size1(struct supertype *st, __u64 devsize, |
2098 | unsigned long long data_offset) | |
82d9eba6 | 2099 | { |
2a528478 | 2100 | struct mdp_superblock_1 *super = st->sb; |
387fcd59 | 2101 | int bmspace = 0; |
23bf42cc | 2102 | int bbspace = 0; |
82d9eba6 NB |
2103 | if (devsize < 24) |
2104 | return 0; | |
2105 | ||
2a528478 | 2106 | #ifndef MDASSEMBLE |
641da745 | 2107 | if (__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) { |
2a528478 N |
2108 | /* hot-add. allow for actual size of bitmap */ |
2109 | struct bitmap_super_s *bsb; | |
911cead7 | 2110 | bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); |
387fcd59 | 2111 | bmspace = bitmap_sectors(bsb); |
2a528478 N |
2112 | } |
2113 | #endif | |
88af981f | 2114 | /* Allow space for bad block log */ |
641da745 | 2115 | if (super->bblog_size) |
23bf42cc | 2116 | bbspace = __le16_to_cpu(super->bblog_size); |
21e92547 | 2117 | |
a380e275 N |
2118 | if (st->minor_version < 0) |
2119 | /* not specified, so time to set default */ | |
2120 | st->minor_version = 2; | |
387fcd59 | 2121 | |
23bf42cc N |
2122 | if (data_offset == INVALID_SECTORS) |
2123 | data_offset = st->data_offset; | |
2124 | ||
387fcd59 N |
2125 | if (data_offset != INVALID_SECTORS) |
2126 | switch(st->minor_version) { | |
2127 | case 0: | |
23bf42cc | 2128 | return devsize - data_offset - 8*2 - bbspace; |
387fcd59 N |
2129 | case 1: |
2130 | case 2: | |
2131 | return devsize - data_offset; | |
2132 | default: | |
2133 | return 0; | |
2134 | } | |
2135 | ||
2136 | devsize -= bmspace; | |
2137 | ||
34163fc7 NB |
2138 | switch(st->minor_version) { |
2139 | case 0: | |
1bf4e2d9 | 2140 | /* at end */ |
23bf42cc | 2141 | return ((devsize - 8*2 - bbspace ) & ~(4*2-1)); |
34163fc7 | 2142 | case 1: |
1bf4e2d9 | 2143 | /* at start, 4K for superblock and possible bitmap */ |
23bf42cc | 2144 | return devsize - 4*2 - bbspace; |
34163fc7 | 2145 | case 2: |
1bf4e2d9 | 2146 | /* 4k from start, 4K for superblock and possible bitmap */ |
23bf42cc | 2147 | return devsize - (4+4)*2 - bbspace; |
34163fc7 NB |
2148 | } |
2149 | return 0; | |
2150 | } | |
2151 | ||
1bf4e2d9 | 2152 | static int |
3da92f27 | 2153 | add_internal_bitmap1(struct supertype *st, |
199171a2 NB |
2154 | int *chunkp, int delay, int write_behind, |
2155 | unsigned long long size, | |
21e92547 | 2156 | int may_change, int major) |
34163fc7 NB |
2157 | { |
2158 | /* | |
c0c1acd6 N |
2159 | * If not may_change, then this is a 'Grow' without sysfs support for |
2160 | * bitmaps, and the bitmap must fit after the superblock at 1K offset. | |
2161 | * If may_change, then this is create or a Grow with sysfs syupport, | |
2162 | * and we can put the bitmap wherever we like. | |
34163fc7 | 2163 | * |
f9c25f1d | 2164 | * size is in sectors, chunk is in bytes !!! |
34163fc7 NB |
2165 | */ |
2166 | ||
1bf4e2d9 | 2167 | unsigned long long bits; |
199171a2 | 2168 | unsigned long long max_bits; |
34163fc7 | 2169 | unsigned long long min_chunk; |
199171a2 | 2170 | long offset; |
88af981f | 2171 | long bbl_offset, bbl_size; |
f21e18ca | 2172 | unsigned long long chunk = *chunkp; |
37dfc3d6 | 2173 | int room = 0; |
c0c1acd6 | 2174 | int creating = 0; |
3da92f27 | 2175 | struct mdp_superblock_1 *sb = st->sb; |
911cead7 | 2176 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); |
3b7e9d0c | 2177 | int uuid[4]; |
34163fc7 | 2178 | |
c0c1acd6 N |
2179 | if (__le64_to_cpu(sb->data_size) == 0) |
2180 | /* Must be creating the array, else data_size would be non-zero */ | |
2181 | creating = 1; | |
199171a2 NB |
2182 | switch(st->minor_version) { |
2183 | case 0: | |
a31128d2 N |
2184 | /* either 3K after the superblock (when hot-add), |
2185 | * or some amount of space before. | |
199171a2 | 2186 | */ |
c0c1acd6 | 2187 | if (creating) { |
199171a2 NB |
2188 | /* We are creating array, so we *know* how much room has |
2189 | * been left. | |
2190 | */ | |
2191 | offset = 0; | |
88af981f | 2192 | bbl_size = 8; |
6ac17e73 | 2193 | room = choose_bm_space(__le64_to_cpu(sb->size)) + bbl_size; |
199171a2 NB |
2194 | } else { |
2195 | room = __le64_to_cpu(sb->super_offset) | |
2196 | - __le64_to_cpu(sb->data_offset) | |
2197 | - __le64_to_cpu(sb->data_size); | |
88af981f N |
2198 | bbl_size = __le16_to_cpu(sb->bblog_size); |
2199 | if (bbl_size < 8) | |
2200 | bbl_size = 8; | |
2201 | bbl_offset = (__s32)__le32_to_cpu(sb->bblog_offset); | |
2202 | if (bbl_size < -bbl_offset) | |
2203 | bbl_size = -bbl_offset; | |
c0c1acd6 N |
2204 | |
2205 | if (!may_change || (room < 3*2 && | |
7716570e | 2206 | __le32_to_cpu(sb->max_dev) <= 384)) { |
199171a2 NB |
2207 | room = 3*2; |
2208 | offset = 1*2; | |
88af981f | 2209 | bbl_size = 0; |
199171a2 NB |
2210 | } else { |
2211 | offset = 0; /* means movable offset */ | |
2212 | } | |
2213 | } | |
2214 | break; | |
2215 | case 1: | |
2216 | case 2: /* between superblock and data */ | |
c0c1acd6 | 2217 | if (creating) { |
199171a2 | 2218 | offset = 4*2; |
88af981f | 2219 | bbl_size = 8; |
6ac17e73 | 2220 | room = choose_bm_space(__le64_to_cpu(sb->size)) + bbl_size; |
199171a2 NB |
2221 | } else { |
2222 | room = __le64_to_cpu(sb->data_offset) | |
2223 | - __le64_to_cpu(sb->super_offset); | |
88af981f N |
2224 | bbl_size = __le16_to_cpu(sb->bblog_size); |
2225 | if (bbl_size) | |
2226 | room = __le32_to_cpu(sb->bblog_offset) + bbl_size; | |
2227 | else | |
2228 | bbl_size = 8; | |
2229 | ||
c0c1acd6 N |
2230 | if (!may_change) { |
2231 | room -= 2; /* Leave 1K for superblock */ | |
199171a2 | 2232 | offset = 2; |
88af981f | 2233 | bbl_size = 0; |
199171a2 | 2234 | } else { |
c0c1acd6 | 2235 | room -= 4*2; /* leave 4K for superblock */ |
199171a2 NB |
2236 | offset = 4*2; |
2237 | } | |
2238 | } | |
2239 | break; | |
ae491d1e NB |
2240 | default: |
2241 | return 0; | |
199171a2 | 2242 | } |
1bf4e2d9 | 2243 | |
88af981f | 2244 | room -= bbl_size; |
199171a2 NB |
2245 | if (chunk == UnSet && room > 128*2) |
2246 | /* Limit to 128K of bitmap when chunk size not requested */ | |
2247 | room = 128*2; | |
1bf4e2d9 | 2248 | |
268cccac N |
2249 | if (room <= 1) |
2250 | /* No room for a bitmap */ | |
2251 | return 0; | |
2252 | ||
199171a2 | 2253 | max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8; |
34163fc7 NB |
2254 | |
2255 | min_chunk = 4096; /* sub-page chunks don't work yet.. */ | |
f9c25f1d | 2256 | bits = (size*512)/min_chunk +1; |
34163fc7 NB |
2257 | while (bits > max_bits) { |
2258 | min_chunk *= 2; | |
2259 | bits = (bits+1)/2; | |
2260 | } | |
b8ab2a50 N |
2261 | if (chunk == UnSet) { |
2262 | /* For practical purpose, 64Meg is a good | |
2263 | * default chunk size for internal bitmaps. | |
2264 | */ | |
34163fc7 | 2265 | chunk = min_chunk; |
b8ab2a50 N |
2266 | if (chunk < 64*1024*1024) |
2267 | chunk = 64*1024*1024; | |
2268 | } else if (chunk < min_chunk) | |
34163fc7 | 2269 | return 0; /* chunk size too small */ |
21e92547 NB |
2270 | if (chunk == 0) /* rounding problem */ |
2271 | return 0; | |
34163fc7 | 2272 | |
199171a2 | 2273 | if (offset == 0) { |
a31128d2 N |
2274 | /* start bitmap on a 4K boundary with enough space for |
2275 | * the bitmap | |
2276 | */ | |
199171a2 | 2277 | bits = (size*512) / chunk + 1; |
a31128d2 N |
2278 | room = ((bits+7)/8 + sizeof(bitmap_super_t) +4095)/4096; |
2279 | room *= 8; /* convert 4K blocks to sectors */ | |
88af981f | 2280 | offset = -room - bbl_size; |
199171a2 NB |
2281 | } |
2282 | ||
34a13953 | 2283 | sb->bitmap_offset = (int32_t)__cpu_to_le32(offset); |
34163fc7 | 2284 | |
c0c1acd6 N |
2285 | sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) |
2286 | | MD_FEATURE_BITMAP_OFFSET); | |
29e766a5 | 2287 | memset(bms, 0, sizeof(*bms)); |
34163fc7 | 2288 | bms->magic = __cpu_to_le32(BITMAP_MAGIC); |
dcec9ee5 | 2289 | bms->version = __cpu_to_le32(major); |
3b7e9d0c LB |
2290 | uuid_from_super1(st, uuid); |
2291 | memcpy(bms->uuid, uuid, 16); | |
34163fc7 NB |
2292 | bms->chunksize = __cpu_to_le32(chunk); |
2293 | bms->daemon_sleep = __cpu_to_le32(delay); | |
f9c25f1d | 2294 | bms->sync_size = __cpu_to_le64(size); |
34163fc7 | 2295 | bms->write_behind = __cpu_to_le32(write_behind); |
529e2aa5 | 2296 | bms->nodes = __cpu_to_le32(st->nodes); |
6d9c7c25 GR |
2297 | if (st->nodes) |
2298 | sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | |
2299 | | MD_FEATURE_BITMAP_VERSIONED); | |
7716570e GJ |
2300 | if (st->cluster_name) |
2301 | strncpy((char *)bms->cluster_name, | |
2302 | st->cluster_name, strlen(st->cluster_name)); | |
34163fc7 | 2303 | |
199171a2 | 2304 | *chunkp = chunk; |
34163fc7 NB |
2305 | return 1; |
2306 | } | |
2307 | ||
53e76b1d | 2308 | static int locate_bitmap1(struct supertype *st, int fd) |
34163fc7 | 2309 | { |
34163fc7 | 2310 | unsigned long long offset; |
1e0d770c NB |
2311 | struct mdp_superblock_1 *sb; |
2312 | int mustfree = 0; | |
53e76b1d | 2313 | int ret; |
34163fc7 | 2314 | |
3da92f27 NB |
2315 | if (!st->sb) { |
2316 | if (st->ss->load_super(st, fd, NULL)) | |
53e76b1d | 2317 | return -1; /* no error I hope... */ |
1e0d770c NB |
2318 | mustfree = 1; |
2319 | } | |
3da92f27 | 2320 | sb = st->sb; |
e478dc86 | 2321 | |
53e76b1d GJ |
2322 | if ((__le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET)) |
2323 | ret = 0; | |
2324 | else | |
2325 | ret = -1; | |
1bf4e2d9 | 2326 | offset = __le64_to_cpu(sb->super_offset); |
68754bd1 | 2327 | offset += (int32_t) __le32_to_cpu(sb->bitmap_offset); |
1e0d770c | 2328 | if (mustfree) |
f6d75de8 | 2329 | free(sb); |
1bf4e2d9 | 2330 | lseek64(fd, offset<<9, 0); |
53e76b1d | 2331 | return ret; |
34163fc7 NB |
2332 | } |
2333 | ||
0aa2f15b | 2334 | static int write_bitmap1(struct supertype *st, int fd, enum bitmap_update update) |
34163fc7 | 2335 | { |
3da92f27 | 2336 | struct mdp_superblock_1 *sb = st->sb; |
911cead7 | 2337 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+MAX_SB_SIZE); |
34163fc7 | 2338 | int rv = 0; |
d669228f | 2339 | void *buf; |
34163fc7 | 2340 | int towrite, n; |
308340aa | 2341 | struct align_fd afd; |
95a05b37 | 2342 | unsigned int i = 0; |
7e6e839a | 2343 | unsigned long long total_bm_space, bm_space_per_node; |
308340aa | 2344 | |
0aa2f15b GJ |
2345 | switch (update) { |
2346 | case NameUpdate: | |
2347 | /* update cluster name */ | |
2348 | if (st->cluster_name) { | |
2349 | memset((char *)bms->cluster_name, 0, sizeof(bms->cluster_name)); | |
2350 | strncpy((char *)bms->cluster_name, st->cluster_name, 64); | |
2351 | } | |
2352 | break; | |
7e6e839a GJ |
2353 | case NodeNumUpdate: |
2354 | /* cluster md only supports superblock 1.2 now */ | |
2355 | if (st->minor_version != 2) { | |
2356 | pr_err("Warning: cluster md only works with superblock 1.2\n"); | |
2357 | return -EINVAL; | |
2358 | } | |
2359 | ||
2360 | /* Each node has an independent bitmap, it is necessary to calculate the | |
2361 | * space is enough or not, first get how many bytes for the total bitmap */ | |
2362 | bm_space_per_node = calc_bitmap_size(bms, 4096); | |
2363 | ||
2364 | total_bm_space = 512 * (__le64_to_cpu(sb->data_offset) - __le64_to_cpu(sb->super_offset)); | |
2365 | total_bm_space = total_bm_space - 4096; /* leave another 4k for superblock */ | |
2366 | ||
2367 | if (bm_space_per_node * st->nodes > total_bm_space) { | |
2368 | pr_err("Warning: The max num of nodes can't exceed %llu\n", | |
2369 | total_bm_space / bm_space_per_node); | |
2370 | return -ENOMEM; | |
2371 | } | |
2372 | ||
2373 | bms->nodes = __cpu_to_le32(st->nodes); | |
2374 | break; | |
0aa2f15b GJ |
2375 | case NoUpdate: |
2376 | default: | |
2377 | break; | |
2378 | } | |
2379 | ||
308340aa | 2380 | init_afd(&afd, fd); |
34163fc7 | 2381 | |
3da92f27 | 2382 | locate_bitmap1(st, fd); |
34163fc7 | 2383 | |
d669228f JS |
2384 | if (posix_memalign(&buf, 4096, 4096)) |
2385 | return -ENOMEM; | |
2386 | ||
95a05b37 GJ |
2387 | do { |
2388 | /* Only the bitmap[0] should resync | |
2389 | * whole device on initial assembly | |
2390 | */ | |
2391 | if (i) | |
2392 | memset(buf, 0x00, 4096); | |
34163fc7 | 2393 | else |
95a05b37 GJ |
2394 | memset(buf, 0xff, 4096); |
2395 | memcpy(buf, (char *)bms, sizeof(bitmap_super_t)); | |
2396 | ||
4a3d29ed | 2397 | towrite = calc_bitmap_size(bms, 4096); |
95a05b37 GJ |
2398 | while (towrite > 0) { |
2399 | n = towrite; | |
2400 | if (n > 4096) | |
2401 | n = 4096; | |
2402 | n = awrite(&afd, buf, n); | |
2403 | if (n > 0) | |
2404 | towrite -= n; | |
2405 | else | |
2406 | break; | |
2407 | if (i) | |
2408 | memset(buf, 0x00, 4096); | |
2409 | else | |
2410 | memset(buf, 0xff, 4096); | |
2411 | } | |
2412 | fsync(fd); | |
2413 | if (towrite) { | |
2414 | rv = -2; | |
34163fc7 | 2415 | break; |
95a05b37 GJ |
2416 | } |
2417 | } while (++i < __le32_to_cpu(bms->nodes)); | |
34163fc7 | 2418 | |
d669228f | 2419 | free(buf); |
34163fc7 | 2420 | return rv; |
82d9eba6 NB |
2421 | } |
2422 | ||
3da92f27 | 2423 | static void free_super1(struct supertype *st) |
df37ffc0 | 2424 | { |
d15a1f72 | 2425 | |
3da92f27 NB |
2426 | if (st->sb) |
2427 | free(st->sb); | |
1cc7f4fe N |
2428 | while (st->info) { |
2429 | struct devinfo *di = st->info; | |
2430 | st->info = di->next; | |
2431 | if (di->fd >= 0) | |
2432 | close(di->fd); | |
2433 | free(di); | |
2434 | } | |
3da92f27 | 2435 | st->sb = NULL; |
df37ffc0 NB |
2436 | } |
2437 | ||
0e600426 | 2438 | #ifndef MDASSEMBLE |
17f25ca6 NB |
2439 | static int validate_geometry1(struct supertype *st, int level, |
2440 | int layout, int raiddisks, | |
c21e737b | 2441 | int *chunk, unsigned long long size, |
af4348dd | 2442 | unsigned long long data_offset, |
2c514b71 NB |
2443 | char *subdev, unsigned long long *freesize, |
2444 | int verbose) | |
17f25ca6 | 2445 | { |
641da745 N |
2446 | unsigned long long ldsize, devsize; |
2447 | int bmspace; | |
23bf42cc | 2448 | unsigned long long headroom; |
17f25ca6 NB |
2449 | int fd; |
2450 | ||
b42f577a N |
2451 | if (level == LEVEL_CONTAINER) { |
2452 | if (verbose) | |
e7b84f9d | 2453 | pr_err("1.x metadata does not support containers\n"); |
17f25ca6 | 2454 | return 0; |
b42f577a | 2455 | } |
7ccc4cc4 | 2456 | if (*chunk == UnSet) |
bb7295f1 N |
2457 | *chunk = DEFAULT_CHUNK; |
2458 | ||
17f25ca6 NB |
2459 | if (!subdev) |
2460 | return 1; | |
2461 | ||
23bf42cc N |
2462 | if (st->minor_version < 0) |
2463 | /* not specified, so time to set default */ | |
2464 | st->minor_version = 2; | |
2465 | ||
17f25ca6 NB |
2466 | fd = open(subdev, O_RDONLY|O_EXCL, 0); |
2467 | if (fd < 0) { | |
2c514b71 | 2468 | if (verbose) |
e7b84f9d | 2469 | pr_err("super1.x cannot open %s: %s\n", |
2c514b71 | 2470 | subdev, strerror(errno)); |
17f25ca6 NB |
2471 | return 0; |
2472 | } | |
2c514b71 | 2473 | |
17f25ca6 NB |
2474 | if (!get_dev_size(fd, subdev, &ldsize)) { |
2475 | close(fd); | |
2476 | return 0; | |
2477 | } | |
2478 | close(fd); | |
2479 | ||
641da745 N |
2480 | devsize = ldsize >> 9; |
2481 | if (devsize < 24) { | |
2482 | *freesize = 0; | |
2483 | return 0; | |
2484 | } | |
2485 | ||
2486 | /* creating: allow suitable space for bitmap */ | |
2487 | bmspace = choose_bm_space(devsize); | |
641da745 | 2488 | |
23bf42cc N |
2489 | if (data_offset == INVALID_SECTORS) |
2490 | data_offset = st->data_offset; | |
2491 | if (data_offset == INVALID_SECTORS) | |
2492 | switch (st->minor_version) { | |
641da745 | 2493 | case 0: |
23bf42cc N |
2494 | data_offset = 0; |
2495 | break; | |
641da745 N |
2496 | case 1: |
2497 | case 2: | |
23bf42cc N |
2498 | /* Choose data offset appropriate for this device |
2499 | * and use as default for whole array. | |
2500 | * The data_offset must allow for bitmap space | |
2501 | * and base metadata, should allow for some headroom | |
2502 | * for reshape, and should be rounded to multiple | |
2503 | * of 1M. | |
2504 | * Headroom is limited to 128M, but aim for about 0.1% | |
641da745 | 2505 | */ |
23bf42cc | 2506 | headroom = 128*1024*2; |
641da745 N |
2507 | while ((headroom << 10) > devsize && |
2508 | (*chunk == 0 || | |
2509 | headroom / 2 >= ((unsigned)(*chunk)*2)*2)) | |
2510 | headroom >>= 1; | |
23bf42cc N |
2511 | data_offset = 12*2 + bmspace + headroom; |
2512 | #define ONE_MEG (2*1024) | |
2513 | if (data_offset > ONE_MEG) | |
2514 | data_offset = (data_offset / ONE_MEG) * ONE_MEG; | |
641da745 N |
2515 | break; |
2516 | } | |
23bf42cc N |
2517 | if (st->data_offset == INVALID_SECTORS) |
2518 | st->data_offset = data_offset; | |
2519 | switch(st->minor_version) { | |
2520 | case 0: /* metadata at end. Round down and subtract space to reserve */ | |
2521 | devsize = (devsize & ~(4ULL*2-1)); | |
2522 | /* space for metadata, bblog, bitmap */ | |
4c0ea7b0 | 2523 | devsize -= 8*2 + 8 + bmspace; |
23bf42cc N |
2524 | break; |
2525 | case 1: | |
2526 | case 2: | |
2527 | devsize -= data_offset; | |
2528 | break; | |
641da745 | 2529 | } |
23bf42cc | 2530 | *freesize = devsize; |
17f25ca6 NB |
2531 | return 1; |
2532 | } | |
0e600426 | 2533 | #endif /* MDASSEMBLE */ |
17f25ca6 | 2534 | |
afa368f4 N |
2535 | void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0) |
2536 | { | |
2537 | /* Create a v1.0 superblock based on 'info'*/ | |
2538 | void *ret; | |
2539 | struct mdp_superblock_1 *sb; | |
2540 | int i; | |
2541 | int rfd; | |
2542 | unsigned long long offset; | |
2543 | ||
2544 | if (posix_memalign(&ret, 4096, 1024) != 0) | |
2545 | return NULL; | |
2546 | sb = ret; | |
2547 | memset(ret, 0, 1024); | |
2548 | sb->magic = __cpu_to_le32(MD_SB_MAGIC); | |
2549 | sb->major_version = __cpu_to_le32(1); | |
2550 | ||
2551 | copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid); | |
2552 | sprintf(sb->set_name, "%d", sb0->md_minor); | |
2553 | sb->ctime = __cpu_to_le32(info->array.ctime+1); | |
2554 | sb->level = __cpu_to_le32(info->array.level); | |
2555 | sb->layout = __cpu_to_le32(info->array.layout); | |
2556 | sb->size = __cpu_to_le64(info->component_size); | |
2557 | sb->chunksize = __cpu_to_le32(info->array.chunk_size/512); | |
2558 | sb->raid_disks = __cpu_to_le32(info->array.raid_disks); | |
ccec2685 N |
2559 | if (info->array.level > 0) |
2560 | sb->data_size = sb->size; | |
2561 | else | |
2562 | sb->data_size = st->ss->avail_size(st, st->devsize/512, 0); | |
afa368f4 N |
2563 | sb->resync_offset = MaxSector; |
2564 | sb->max_dev = __cpu_to_le32(MD_SB_DISKS); | |
2565 | sb->dev_number = __cpu_to_le32(info->disk.number); | |
2566 | sb->utime = __cpu_to_le64(info->array.utime); | |
2567 | ||
2568 | offset = st->devsize/512 - 8*2; | |
2569 | offset &= ~(4*2-1); | |
2570 | sb->super_offset = __cpu_to_le64(offset); | |
2571 | //*(__u64*)(st->other + 128 + 8 + 8) = __cpu_to_le64(offset); | |
2572 | ||
2573 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
2574 | read(rfd, sb->device_uuid, 16) != 16) { | |
2575 | __u32 r[4] = {random(), random(), random(), random()}; | |
2576 | memcpy(sb->device_uuid, r, 16); | |
2577 | } | |
2578 | if (rfd >= 0) | |
2579 | close(rfd); | |
2580 | ||
2581 | for (i = 0; i < MD_SB_DISKS; i++) { | |
2582 | int state = sb0->disks[i].state; | |
fa7574f6 | 2583 | sb->dev_roles[i] = MD_DISK_ROLE_SPARE; |
afa368f4 N |
2584 | if ((state & (1<<MD_DISK_SYNC)) && |
2585 | !(state & (1<<MD_DISK_FAULTY))) | |
2586 | sb->dev_roles[i] = __cpu_to_le16(sb0->disks[i].raid_disk); | |
2587 | } | |
2588 | sb->sb_csum = calc_sb_1_csum(sb); | |
2589 | return ret; | |
2590 | } | |
2591 | ||
82d9eba6 | 2592 | struct superswitch super1 = { |
c7654afc | 2593 | #ifndef MDASSEMBLE |
82d9eba6 NB |
2594 | .examine_super = examine_super1, |
2595 | .brief_examine_super = brief_examine_super1, | |
0d726f17 | 2596 | .export_examine_super = export_examine_super1, |
82d9eba6 NB |
2597 | .detail_super = detail_super1, |
2598 | .brief_detail_super = brief_detail_super1, | |
0d726f17 | 2599 | .export_detail_super = export_detail_super1, |
111d01fc | 2600 | .write_init_super = write_init_super1, |
0e600426 N |
2601 | .validate_geometry = validate_geometry1, |
2602 | .add_to_super = add_to_super1, | |
6d388a88 | 2603 | .examine_badblocks = examine_badblocks_super1, |
74db60b0 | 2604 | .copy_metadata = copy_metadata1, |
c7654afc | 2605 | #endif |
83b6208e | 2606 | .match_home = match_home1, |
82d9eba6 NB |
2607 | .uuid_from_super = uuid_from_super1, |
2608 | .getinfo_super = getinfo_super1, | |
00bbdbda | 2609 | .container_content = container_content1, |
82d9eba6 | 2610 | .update_super = update_super1, |
82d9eba6 | 2611 | .init_super = init_super1, |
82d9eba6 | 2612 | .store_super = store_super1, |
82d9eba6 NB |
2613 | .compare_super = compare_super1, |
2614 | .load_super = load_super1, | |
2615 | .match_metadata_desc = match_metadata_desc1, | |
2616 | .avail_size = avail_size1, | |
34163fc7 NB |
2617 | .add_internal_bitmap = add_internal_bitmap1, |
2618 | .locate_bitmap = locate_bitmap1, | |
2619 | .write_bitmap = write_bitmap1, | |
df37ffc0 | 2620 | .free_super = free_super1, |
f277ce36 NB |
2621 | #if __BYTE_ORDER == BIG_ENDIAN |
2622 | .swapuuid = 0, | |
2623 | #else | |
2624 | .swapuuid = 1, | |
2625 | #endif | |
31015d57 | 2626 | .name = "1.x", |
82d9eba6 | 2627 | }; |