]>
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 | ||
25 | #include "mdadm.h" | |
82d9eba6 NB |
26 | /* |
27 | * The version-1 superblock : | |
28 | * All numeric fields are little-endian. | |
29 | * | |
30 | * total size: 256 bytes plus 2 per device. | |
31 | * 1K allows 384 devices. | |
32 | */ | |
33 | struct mdp_superblock_1 { | |
34 | /* constant array information - 128 bytes */ | |
35 | __u32 magic; /* MD_SB_MAGIC: 0xa92b4efc - little endian */ | |
36 | __u32 major_version; /* 1 */ | |
37 | __u32 feature_map; /* 0 for now */ | |
38 | __u32 pad0; /* always set to 0 when writing */ | |
39 | ||
40 | __u8 set_uuid[16]; /* user-space generated. */ | |
41 | char set_name[32]; /* set and interpreted by user-space */ | |
42 | ||
43 | __u64 ctime; /* lo 40 bits are seconds, top 24 are microseconds or 0*/ | |
44 | __u32 level; /* -4 (multipath), -1 (linear), 0,1,4,5 */ | |
45 | __u32 layout; /* only for raid5 currently */ | |
46 | __u64 size; /* used size of component devices, in 512byte sectors */ | |
47 | ||
48 | __u32 chunksize; /* in 512byte sectors */ | |
49 | __u32 raid_disks; | |
34163fc7 NB |
50 | __u32 bitmap_offset; /* sectors after start of superblock that bitmap starts |
51 | * NOTE: signed, so bitmap can be before superblock | |
52 | * only meaningful of feature_map[0] is set. | |
53 | */ | |
b674b5b8 NB |
54 | |
55 | /* These are only valid with feature bit '4' */ | |
b674b5b8 | 56 | __u32 new_level; /* new level we are reshaping to */ |
ac957baa | 57 | __u64 reshape_position; /* next address in array-space for reshape */ |
b674b5b8 NB |
58 | __u32 delta_disks; /* change in number of raid_disks */ |
59 | __u32 new_layout; /* new layout */ | |
60 | __u32 new_chunk; /* new chunk size (bytes) */ | |
8fe1c44f N |
61 | __u32 new_offset; /* signed number to add to data_offset in new |
62 | * layout. 0 == no-change. This can be | |
63 | * different on each device in the array. | |
64 | */ | |
82d9eba6 NB |
65 | |
66 | /* constant this-device information - 64 bytes */ | |
67 | __u64 data_offset; /* sector start of data, often 0 */ | |
68 | __u64 data_size; /* sectors in this device that can be used for data */ | |
69 | __u64 super_offset; /* sector start of this superblock */ | |
70 | __u64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */ | |
71 | __u32 dev_number; /* permanent identifier of this device - not role in raid */ | |
72 | __u32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */ | |
73 | __u8 device_uuid[16]; /* user-space setable, ignored by kernel */ | |
5d500228 | 74 | __u8 devflags; /* per-device flags. Only one defined...*/ |
dfd4d8ee | 75 | #define WriteMostly1 1 /* mask for writemostly flag in above */ |
bf95d0f3 N |
76 | /* bad block log. If there are any bad blocks the feature flag is set. |
77 | * if offset and size are non-zero, that space is reserved and available. | |
78 | */ | |
79 | __u8 bblog_shift; /* shift from sectors to block size for badblocklist */ | |
80 | __u16 bblog_size; /* number of sectors reserved for badblocklist */ | |
81 | __u32 bblog_offset; /* sector offset from superblock to bblog, signed */ | |
82d9eba6 NB |
82 | |
83 | /* array state information - 64 bytes */ | |
84 | __u64 utime; /* 40 bits second, 24 btes microseconds */ | |
85 | __u64 events; /* incremented when superblock updated */ | |
86 | __u64 resync_offset; /* data before this offset (from data_offset) known to be in sync */ | |
4180aa4d N |
87 | __u32 sb_csum; /* checksum upto dev_roles[max_dev] */ |
88 | __u32 max_dev; /* size of dev_roles[] array to consider */ | |
82d9eba6 NB |
89 | __u8 pad3[64-32]; /* set to 0 when writing */ |
90 | ||
91 | /* device state information. Indexed by dev_number. | |
92 | * 2 bytes per device | |
93 | * Note there are no per-device state flags. State information is rolled | |
94 | * into the 'roles' value. If a device is spare or faulty, then it doesn't | |
95 | * have a meaningful role. | |
96 | */ | |
97 | __u16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */ | |
98 | }; | |
99 | ||
911cead7 N |
100 | #define MAX_SB_SIZE 4096 |
101 | /* bitmap super size is 256, but we round up to a sector for alignment */ | |
102 | #define BM_SUPER_SIZE 512 | |
103 | #define MAX_DEVS ((int)(MAX_SB_SIZE - sizeof(struct mdp_superblock_1)) / 2) | |
41226756 JS |
104 | #define SUPER1_SIZE (MAX_SB_SIZE + BM_SUPER_SIZE \ |
105 | + sizeof(struct misc_dev_info)) | |
911cead7 | 106 | |
bee8ec56 NB |
107 | struct misc_dev_info { |
108 | __u64 device_size; | |
109 | }; | |
110 | ||
d2cd3ffc NB |
111 | /* feature_map bits */ |
112 | #define MD_FEATURE_BITMAP_OFFSET 1 | |
113 | #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and | |
114 | * must be honoured | |
115 | */ | |
b674b5b8 | 116 | #define MD_FEATURE_RESHAPE_ACTIVE 4 |
bf95d0f3 | 117 | #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */ |
8fe1c44f N |
118 | #define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an |
119 | * active device with same 'role'. | |
120 | * 'recovery_offset' is also set. | |
121 | */ | |
122 | #define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number | |
123 | * of devices, but is going | |
124 | * backwards anyway. | |
125 | */ | |
126 | #define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */ | |
127 | #define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \ | |
128 | |MD_FEATURE_RECOVERY_OFFSET \ | |
129 | |MD_FEATURE_RESHAPE_ACTIVE \ | |
130 | |MD_FEATURE_BAD_BLOCKS \ | |
131 | |MD_FEATURE_REPLACEMENT \ | |
132 | |MD_FEATURE_RESHAPE_BACKWARDS \ | |
133 | |MD_FEATURE_NEW_OFFSET \ | |
134 | ) | |
d2cd3ffc | 135 | |
570c0542 | 136 | #ifndef offsetof |
067db4df | 137 | #define offsetof(t,f) ((size_t)&(((t*)0)->f)) |
570c0542 | 138 | #endif |
82d9eba6 NB |
139 | static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb) |
140 | { | |
141 | unsigned int disk_csum, csum; | |
142 | unsigned long long newcsum; | |
143 | int size = sizeof(*sb) + __le32_to_cpu(sb->max_dev)*2; | |
144 | unsigned int *isuper = (unsigned int*)sb; | |
82d9eba6 NB |
145 | |
146 | /* make sure I can count... */ | |
147 | if (offsetof(struct mdp_superblock_1,data_offset) != 128 || | |
148 | offsetof(struct mdp_superblock_1, utime) != 192 || | |
149 | sizeof(struct mdp_superblock_1) != 256) { | |
150 | fprintf(stderr, "WARNING - superblock isn't sized correctly\n"); | |
151 | } | |
152 | ||
153 | disk_csum = sb->sb_csum; | |
154 | sb->sb_csum = 0; | |
155 | newcsum = 0; | |
e4c72d1d | 156 | for (; size>=4; size -= 4 ) { |
722966c6 PC |
157 | newcsum += __le32_to_cpu(*isuper); |
158 | isuper++; | |
159 | } | |
82d9eba6 NB |
160 | |
161 | if (size == 2) | |
162 | newcsum += __le16_to_cpu(*(unsigned short*) isuper); | |
163 | ||
164 | csum = (newcsum & 0xffffffff) + (newcsum >> 32); | |
165 | sb->sb_csum = disk_csum; | |
83205b64 | 166 | return __cpu_to_le32(csum); |
82d9eba6 NB |
167 | } |
168 | ||
308340aa JS |
169 | /* |
170 | * Information related to file descriptor used for aligned reads/writes. | |
171 | * Cache the block size. | |
172 | */ | |
173 | struct align_fd { | |
174 | int fd; | |
175 | int blk_sz; | |
176 | }; | |
177 | ||
178 | static void init_afd(struct align_fd *afd, int fd) | |
179 | { | |
180 | afd->fd = fd; | |
181 | ||
182 | if (ioctl(afd->fd, BLKSSZGET, &afd->blk_sz) != 0) | |
183 | afd->blk_sz = 512; | |
184 | } | |
185 | ||
5ea022a1 | 186 | static char abuf[4096+4096]; |
308340aa | 187 | static int aread(struct align_fd *afd, void *buf, int len) |
5ea022a1 N |
188 | { |
189 | /* aligned read. | |
190 | * On devices with a 4K sector size, we need to read | |
191 | * the full sector and copy relevant bits into | |
192 | * the buffer | |
193 | */ | |
6ef89052 | 194 | int bsize, iosize; |
5ea022a1 N |
195 | char *b; |
196 | int n; | |
6ef89052 | 197 | |
308340aa JS |
198 | bsize = afd->blk_sz; |
199 | ||
200 | if (!bsize || bsize > 4096 || len > 4096) { | |
201 | if (!bsize) | |
202 | fprintf(stderr, "WARNING - aread() called with " | |
203 | "invalid block size\n"); | |
5ea022a1 | 204 | return -1; |
308340aa | 205 | } |
0a2f1894 | 206 | b = ROUND_UP_PTR((char *)abuf, 4096); |
5ea022a1 | 207 | |
6ef89052 N |
208 | for (iosize = 0; iosize < len; iosize += bsize) |
209 | ; | |
308340aa | 210 | n = read(afd->fd, b, iosize); |
5ea022a1 N |
211 | if (n <= 0) |
212 | return n; | |
308340aa | 213 | lseek(afd->fd, len - n, 1); |
5ea022a1 N |
214 | if (n > len) |
215 | n = len; | |
216 | memcpy(buf, b, n); | |
217 | return n; | |
218 | } | |
219 | ||
308340aa | 220 | static int awrite(struct align_fd *afd, void *buf, int len) |
5ea022a1 N |
221 | { |
222 | /* aligned write. | |
223 | * On devices with a 4K sector size, we need to write | |
224 | * the full sector. We pre-read if the sector is larger | |
225 | * than the write. | |
226 | * The address must be sector-aligned. | |
227 | */ | |
6ef89052 | 228 | int bsize, iosize; |
5ea022a1 N |
229 | char *b; |
230 | int n; | |
308340aa JS |
231 | |
232 | bsize = afd->blk_sz; | |
233 | if (!bsize || bsize > 4096 || len > 4096) { | |
234 | if (!bsize) | |
235 | fprintf(stderr, "WARNING - awrite() called with " | |
236 | "invalid block size\n"); | |
5ea022a1 | 237 | return -1; |
308340aa | 238 | } |
0a2f1894 | 239 | b = ROUND_UP_PTR((char *)abuf, 4096); |
5ea022a1 | 240 | |
6ef89052 N |
241 | for (iosize = 0; iosize < len ; iosize += bsize) |
242 | ; | |
243 | ||
244 | if (len != iosize) { | |
308340aa | 245 | n = read(afd->fd, b, iosize); |
6ef89052 N |
246 | if (n <= 0) |
247 | return n; | |
308340aa | 248 | lseek(afd->fd, -n, 1); |
6ef89052 N |
249 | } |
250 | ||
5ea022a1 | 251 | memcpy(b, buf, len); |
308340aa | 252 | n = write(afd->fd, b, iosize); |
5ea022a1 N |
253 | if (n <= 0) |
254 | return n; | |
308340aa | 255 | lseek(afd->fd, len - n, 1); |
5ea022a1 N |
256 | return len; |
257 | } | |
258 | ||
c7654afc | 259 | #ifndef MDASSEMBLE |
3da92f27 | 260 | static void examine_super1(struct supertype *st, char *homehost) |
82d9eba6 | 261 | { |
3da92f27 | 262 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 | 263 | time_t atime; |
f21e18ca | 264 | unsigned int d; |
e8e1c176 | 265 | int role; |
4180aa4d | 266 | int delta_extra = 0; |
82d9eba6 NB |
267 | int i; |
268 | char *c; | |
a1cbd7d0 | 269 | int l = homehost ? strlen(homehost) : 0; |
83205b64 | 270 | int layout; |
a17a3de3 | 271 | unsigned long long sb_offset; |
82d9eba6 NB |
272 | |
273 | printf(" Magic : %08x\n", __le32_to_cpu(sb->magic)); | |
a17a3de3 DL |
274 | printf(" Version : 1"); |
275 | sb_offset = __le64_to_cpu(sb->super_offset); | |
276 | if (sb_offset <= 4) | |
277 | printf(".1\n"); | |
278 | else if (sb_offset <= 8) | |
279 | printf(".2\n"); | |
280 | else | |
281 | printf(".0\n"); | |
91eedefc | 282 | printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb->feature_map)); |
82d9eba6 NB |
283 | printf(" Array UUID : "); |
284 | for (i=0; i<16; i++) { | |
82d9eba6 | 285 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 286 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 NB |
287 | } |
288 | printf("\n"); | |
a1cbd7d0 NB |
289 | printf(" Name : %.32s", sb->set_name); |
290 | if (l > 0 && l < 32 && | |
291 | sb->set_name[l] == ':' && | |
292 | strncmp(sb->set_name, homehost, l) == 0) | |
293 | printf(" (local to host %s)", homehost); | |
294 | printf("\n"); | |
82d9eba6 NB |
295 | atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL; |
296 | printf(" Creation Time : %.24s\n", ctime(&atime)); | |
297 | c=map_num(pers, __le32_to_cpu(sb->level)); | |
298 | printf(" Raid Level : %s\n", c?c:"-unknown-"); | |
299 | printf(" Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks)); | |
300 | printf("\n"); | |
5cda0964 | 301 | printf(" Avail Dev Size : %llu%s\n", |
2fb749d1 NB |
302 | (unsigned long long)__le64_to_cpu(sb->data_size), |
303 | human_size(__le64_to_cpu(sb->data_size)<<9)); | |
c43f7d91 | 304 | if (__le32_to_cpu(sb->level) > 0) { |
d4633e06 | 305 | int ddsks = 0, ddsks_denom = 1; |
b674b5b8 NB |
306 | switch(__le32_to_cpu(sb->level)) { |
307 | case 1: ddsks=1;break; | |
308 | case 4: | |
83205b64 NB |
309 | case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break; |
310 | case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break; | |
311 | case 10: | |
312 | layout = __le32_to_cpu(sb->layout); | |
d4633e06 N |
313 | ddsks = __le32_to_cpu(sb->raid_disks); |
314 | ddsks_denom = (layout&255) * ((layout>>8)&255); | |
b674b5b8 | 315 | } |
d4633e06 N |
316 | if (ddsks) { |
317 | long long asize = __le64_to_cpu(sb->size); | |
318 | asize = (asize << 9) * ddsks / ddsks_denom; | |
83205b64 | 319 | printf(" Array Size : %llu%s\n", |
d4633e06 N |
320 | asize >> 10, human_size(asize)); |
321 | } | |
b674b5b8 | 322 | if (sb->size != sb->data_size) |
5cda0964 | 323 | printf(" Used Dev Size : %llu%s\n", |
83205b64 NB |
324 | (unsigned long long)__le64_to_cpu(sb->size), |
325 | human_size(__le64_to_cpu(sb->size)<<9)); | |
b674b5b8 | 326 | } |
82d9eba6 | 327 | if (sb->data_offset) |
199171a2 NB |
328 | printf(" Data Offset : %llu sectors\n", |
329 | (unsigned long long)__le64_to_cpu(sb->data_offset)); | |
8772113a N |
330 | if (sb->new_offset && |
331 | (__le32_to_cpu(sb->feature_map) & MD_FEATURE_NEW_OFFSET)) { | |
8fe1c44f N |
332 | unsigned long long offset = __le64_to_cpu(sb->data_offset); |
333 | offset += (signed)(int32_t)__le32_to_cpu(sb->new_offset); | |
334 | printf(" New Offset : %llu sectors\n", offset); | |
335 | } | |
199171a2 NB |
336 | printf(" Super Offset : %llu sectors\n", |
337 | (unsigned long long)__le64_to_cpu(sb->super_offset)); | |
d2cd3ffc NB |
338 | if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET) |
339 | printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb->recovery_offset)); | |
a3fd117c | 340 | printf(" State : %s\n", (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean"); |
82d9eba6 NB |
341 | printf(" Device UUID : "); |
342 | for (i=0; i<16; i++) { | |
82d9eba6 | 343 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 344 | printf("%02x", sb->device_uuid[i]); |
82d9eba6 NB |
345 | } |
346 | printf("\n"); | |
b674b5b8 | 347 | printf("\n"); |
91eedefc NB |
348 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { |
349 | printf("Internal Bitmap : %ld sectors from superblock\n", | |
68754bd1 | 350 | (long)(int32_t)__le32_to_cpu(sb->bitmap_offset)); |
8fac0577 | 351 | } |
bf95d0f3 | 352 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE)) { |
1e0d770c | 353 | printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb->reshape_position)/2, |
b674b5b8 NB |
354 | human_size(__le64_to_cpu(sb->reshape_position)<<9)); |
355 | if (__le32_to_cpu(sb->delta_disks)) { | |
356 | printf(" Delta Devices : %d", __le32_to_cpu(sb->delta_disks)); | |
4180aa4d N |
357 | printf(" (%d->%d)\n", |
358 | __le32_to_cpu(sb->raid_disks)-__le32_to_cpu(sb->delta_disks), | |
359 | __le32_to_cpu(sb->raid_disks)); | |
360 | if ((int)__le32_to_cpu(sb->delta_disks) < 0) | |
361 | delta_extra = -__le32_to_cpu(sb->delta_disks); | |
b674b5b8 NB |
362 | } |
363 | if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) { | |
364 | c = map_num(pers, __le32_to_cpu(sb->new_level)); | |
365 | printf(" New Level : %s\n", c?c:"-unknown-"); | |
366 | } | |
367 | if (__le32_to_cpu(sb->new_layout) != __le32_to_cpu(sb->layout)) { | |
368 | if (__le32_to_cpu(sb->level) == 5) { | |
369 | c = map_num(r5layout, __le32_to_cpu(sb->new_layout)); | |
370 | printf(" New Layout : %s\n", c?c:"-unknown-"); | |
371 | } | |
fe77a154 N |
372 | if (__le32_to_cpu(sb->level) == 6) { |
373 | c = map_num(r6layout, __le32_to_cpu(sb->new_layout)); | |
374 | printf(" New Layout : %s\n", c?c:"-unknown-"); | |
375 | } | |
b674b5b8 | 376 | if (__le32_to_cpu(sb->level) == 10) { |
e4965ef8 N |
377 | printf(" New Layout :"); |
378 | print_r10_layout(__le32_to_cpu(sb->new_layout)); | |
379 | printf("\n"); | |
b674b5b8 NB |
380 | } |
381 | } | |
382 | if (__le32_to_cpu(sb->new_chunk) != __le32_to_cpu(sb->chunksize)) | |
383 | printf(" New Chunksize : %dK\n", __le32_to_cpu(sb->new_chunk)/2); | |
384 | printf("\n"); | |
385 | } | |
dfd4d8ee | 386 | if (sb->devflags) { |
62184891 | 387 | printf(" Flags :"); |
34163fc7 | 388 | if (sb->devflags & WriteMostly1) |
dfd4d8ee NB |
389 | printf(" write-mostly"); |
390 | printf("\n"); | |
391 | } | |
82d9eba6 NB |
392 | |
393 | atime = __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL; | |
394 | printf(" Update Time : %.24s\n", ctime(&atime)); | |
395 | ||
bf95d0f3 N |
396 | if (sb->bblog_size && sb->bblog_offset) { |
397 | printf(" Bad Block Log : %d entries available at offset %ld sectors", | |
398 | __le16_to_cpu(sb->bblog_size)*512/8, | |
399 | (long)__le32_to_cpu(sb->bblog_offset)); | |
400 | if (sb->feature_map & | |
401 | __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
402 | printf(" - bad blocks present."); | |
403 | printf("\n"); | |
404 | } | |
405 | ||
406 | ||
82d9eba6 NB |
407 | if (calc_sb_1_csum(sb) == sb->sb_csum) |
408 | printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum)); | |
409 | else | |
410 | printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb->sb_csum), | |
411 | __le32_to_cpu(calc_sb_1_csum(sb))); | |
570c0542 | 412 | printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb->events)); |
82d9eba6 NB |
413 | printf("\n"); |
414 | if (__le32_to_cpu(sb->level) == 5) { | |
415 | c = map_num(r5layout, __le32_to_cpu(sb->layout)); | |
416 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
417 | } | |
fe77a154 N |
418 | if (__le32_to_cpu(sb->level) == 6) { |
419 | c = map_num(r6layout, __le32_to_cpu(sb->layout)); | |
420 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
421 | } | |
265e0f17 NB |
422 | if (__le32_to_cpu(sb->level) == 10) { |
423 | int lo = __le32_to_cpu(sb->layout); | |
e4965ef8 N |
424 | printf(" Layout :"); |
425 | print_r10_layout(lo); | |
426 | printf("\n"); | |
265e0f17 | 427 | } |
82d9eba6 NB |
428 | switch(__le32_to_cpu(sb->level)) { |
429 | case 0: | |
430 | case 4: | |
431 | case 5: | |
d2cd3ffc NB |
432 | case 6: |
433 | case 10: | |
b674b5b8 | 434 | printf(" Chunk Size : %dK\n", __le32_to_cpu(sb->chunksize)/2); |
82d9eba6 NB |
435 | break; |
436 | case -1: | |
b674b5b8 | 437 | printf(" Rounding : %dK\n", __le32_to_cpu(sb->chunksize)/2); |
82d9eba6 NB |
438 | break; |
439 | default: break; | |
440 | } | |
441 | printf("\n"); | |
e8e1c176 N |
442 | #if 0 |
443 | /* This turns out to just be confusing */ | |
c3684618 NB |
444 | printf(" Array Slot : %d (", __le32_to_cpu(sb->dev_number)); |
445 | for (i= __le32_to_cpu(sb->max_dev); i> 0 ; i--) | |
446 | if (__le16_to_cpu(sb->dev_roles[i-1]) != 0xffff) | |
447 | break; | |
448 | for (d=0; d < i; d++) { | |
449 | int role = __le16_to_cpu(sb->dev_roles[d]); | |
450 | if (d) printf(", "); | |
451 | if (role == 0xffff) printf("empty"); | |
452 | else if(role == 0xfffe) printf("failed"); | |
453 | else printf("%d", role); | |
454 | } | |
455 | printf(")\n"); | |
e8e1c176 N |
456 | #endif |
457 | printf(" Device Role : "); | |
458 | d = __le32_to_cpu(sb->dev_number); | |
4180aa4d | 459 | if (d < __le32_to_cpu(sb->max_dev)) |
e8e1c176 N |
460 | role = __le16_to_cpu(sb->dev_roles[d]); |
461 | else | |
462 | role = 0xFFFF; | |
463 | if (role >= 0xFFFE) | |
464 | printf("spare\n"); | |
24c7bc84 N |
465 | else if (sb->feature_map & __cpu_to_le32(MD_FEATURE_REPLACEMENT)) |
466 | printf("Replacement device %d\n", role); | |
e8e1c176 N |
467 | else |
468 | printf("Active device %d\n", role); | |
469 | ||
82d9eba6 | 470 | printf(" Array State : "); |
4180aa4d | 471 | for (d=0; d<__le32_to_cpu(sb->raid_disks) + delta_extra; d++) { |
82d9eba6 | 472 | int cnt = 0; |
f21e18ca | 473 | unsigned int i; |
82d9eba6 | 474 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { |
f21e18ca | 475 | unsigned int role = __le16_to_cpu(sb->dev_roles[i]); |
e4c72d1d | 476 | if (role == d) |
82d9eba6 | 477 | cnt++; |
82d9eba6 | 478 | } |
24c7bc84 N |
479 | if (cnt == 2) |
480 | printf("R"); | |
481 | else if (cnt == 1) | |
482 | printf("A"); | |
483 | else if (cnt == 0) | |
484 | printf("."); | |
485 | else | |
486 | printf("?"); | |
82d9eba6 | 487 | } |
e8e1c176 N |
488 | #if 0 |
489 | /* This is confusing too */ | |
6fbba4c9 | 490 | faulty = 0; |
82d9eba6 NB |
491 | for (i=0; i< __le32_to_cpu(sb->max_dev); i++) { |
492 | int role = __le16_to_cpu(sb->dev_roles[i]); | |
6fbba4c9 NB |
493 | if (role == 0xFFFE) |
494 | faulty++; | |
82d9eba6 | 495 | } |
82d9eba6 | 496 | if (faulty) printf(" %d failed", faulty); |
e8e1c176 | 497 | #endif |
24c7bc84 | 498 | printf(" ('A' == active, '.' == missing, 'R' == replacing)"); |
82d9eba6 NB |
499 | printf("\n"); |
500 | } | |
501 | ||
502 | ||
061f2c6a | 503 | static void brief_examine_super1(struct supertype *st, int verbose) |
82d9eba6 | 504 | { |
3da92f27 | 505 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 | 506 | int i; |
a17a3de3 | 507 | unsigned long long sb_offset; |
2998bc01 | 508 | char *nm; |
b015e6c2 | 509 | char *c=map_num(pers, __le32_to_cpu(sb->level)); |
2998bc01 NB |
510 | |
511 | nm = strchr(sb->set_name, ':'); | |
512 | if (nm) | |
513 | nm++; | |
514 | else if (sb->set_name[0]) | |
515 | nm = sb->set_name; | |
516 | else | |
cf8de691 | 517 | nm = NULL; |
82d9eba6 | 518 | |
7103b9b8 N |
519 | printf("ARRAY "); |
520 | if (nm) { | |
521 | printf("/dev/md/"); | |
522 | print_escape(nm); | |
523 | putchar(' '); | |
524 | } | |
061f2c6a N |
525 | if (verbose && c) |
526 | printf(" level=%s", c); | |
a17a3de3 DL |
527 | sb_offset = __le64_to_cpu(sb->super_offset); |
528 | if (sb_offset <= 4) | |
e9a0e728 | 529 | printf(" metadata=1.1 "); |
a17a3de3 | 530 | else if (sb_offset <= 8) |
e9a0e728 | 531 | printf(" metadata=1.2 "); |
a17a3de3 | 532 | else |
e9a0e728 | 533 | printf(" metadata=1.0 "); |
061f2c6a N |
534 | if (verbose) |
535 | printf("num-devices=%d ", __le32_to_cpu(sb->raid_disks)); | |
536 | printf("UUID="); | |
82d9eba6 | 537 | for (i=0; i<16; i++) { |
82d9eba6 | 538 | if ((i&3)==0 && i != 0) printf(":"); |
b7a708af | 539 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 | 540 | } |
7103b9b8 N |
541 | if (sb->set_name[0]) { |
542 | printf(" name="); | |
543 | print_quoted(sb->set_name); | |
544 | } | |
82d9eba6 NB |
545 | printf("\n"); |
546 | } | |
547 | ||
0d726f17 KS |
548 | static void export_examine_super1(struct supertype *st) |
549 | { | |
550 | struct mdp_superblock_1 *sb = st->sb; | |
551 | int i; | |
552 | int len = 32; | |
80730bae | 553 | int layout; |
0d726f17 KS |
554 | |
555 | printf("MD_LEVEL=%s\n", map_num(pers, __le32_to_cpu(sb->level))); | |
556 | printf("MD_DEVICES=%d\n", __le32_to_cpu(sb->raid_disks)); | |
557 | for (i=0; i<32; i++) | |
558 | if (sb->set_name[i] == '\n' || | |
559 | sb->set_name[i] == '\0') { | |
560 | len = i; | |
561 | break; | |
562 | } | |
563 | if (len) | |
564 | printf("MD_NAME=%.*s\n", len, sb->set_name); | |
80730bae MN |
565 | if (__le32_to_cpu(sb->level) > 0) { |
566 | int ddsks = 0, ddsks_denom = 1; | |
567 | switch(__le32_to_cpu(sb->level)) { | |
568 | case 1: ddsks=1;break; | |
569 | case 4: | |
570 | case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break; | |
571 | case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break; | |
572 | case 10: | |
573 | layout = __le32_to_cpu(sb->layout); | |
574 | ddsks = __le32_to_cpu(sb->raid_disks); | |
575 | ddsks_denom = (layout&255) * ((layout>>8)&255); | |
576 | } | |
577 | if (ddsks) { | |
578 | long long asize = __le64_to_cpu(sb->size); | |
579 | asize = (asize << 9) * ddsks / ddsks_denom; | |
580 | printf("MD_ARRAY_SIZE=%s\n",human_size_brief(asize,JEDEC)); | |
581 | } | |
582 | } | |
0d726f17 KS |
583 | printf("MD_UUID="); |
584 | for (i=0; i<16; i++) { | |
585 | if ((i&3)==0 && i != 0) printf(":"); | |
586 | printf("%02x", sb->set_uuid[i]); | |
587 | } | |
588 | printf("\n"); | |
589 | printf("MD_UPDATE_TIME=%llu\n", | |
590 | __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL); | |
591 | printf("MD_DEV_UUID="); | |
592 | for (i=0; i<16; i++) { | |
593 | if ((i&3)==0 && i != 0) printf(":"); | |
594 | printf("%02x", sb->device_uuid[i]); | |
595 | } | |
596 | printf("\n"); | |
597 | printf("MD_EVENTS=%llu\n", | |
598 | (unsigned long long)__le64_to_cpu(sb->events)); | |
599 | } | |
600 | ||
74db60b0 N |
601 | static int copy_metadata1(struct supertype *st, int from, int to) |
602 | { | |
603 | /* Read superblock. If it looks good, write it out. | |
604 | * Then if a bitmap is present, copy that. | |
605 | * And if a bad-block-list is present, copy that too. | |
606 | */ | |
607 | void *buf; | |
608 | unsigned long long dsize, sb_offset; | |
609 | const int bufsize = 4*1024; | |
610 | struct mdp_superblock_1 super, *sb; | |
611 | ||
612 | if (posix_memalign(&buf, 4096, bufsize) != 0) | |
613 | return 1; | |
614 | ||
615 | if (!get_dev_size(from, NULL, &dsize)) | |
616 | goto err; | |
617 | ||
618 | dsize >>= 9; | |
619 | if (dsize < 24) | |
620 | goto err; | |
621 | switch(st->minor_version) { | |
622 | case 0: | |
623 | sb_offset = dsize; | |
624 | sb_offset -= 8*2; | |
625 | sb_offset &= ~(4*2-1); | |
626 | break; | |
627 | case 1: | |
628 | sb_offset = 0; | |
629 | break; | |
630 | case 2: | |
631 | sb_offset = 4*2; | |
632 | break; | |
633 | default: | |
634 | goto err; | |
635 | } | |
636 | ||
637 | if (lseek64(from, sb_offset << 9, 0) < 0LL) | |
638 | goto err; | |
639 | if (read(from, buf, bufsize) != bufsize) | |
640 | goto err; | |
641 | ||
642 | sb = buf; | |
643 | super = *sb; // save most of sb for when we reuse buf | |
644 | ||
645 | if (__le32_to_cpu(super.magic) != MD_SB_MAGIC || | |
646 | __le32_to_cpu(super.major_version) != 1 || | |
647 | __le64_to_cpu(super.super_offset) != sb_offset || | |
648 | calc_sb_1_csum(sb) != super.sb_csum) | |
649 | goto err; | |
650 | ||
651 | if (lseek64(to, sb_offset << 9, 0) < 0LL) | |
652 | goto err; | |
653 | if (write(to, buf, bufsize) != bufsize) | |
654 | goto err; | |
655 | ||
656 | if (super.feature_map & __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET)) { | |
657 | unsigned long long bitmap_offset = sb_offset; | |
658 | int bytes = 4096; // just an estimate. | |
659 | int written = 0; | |
660 | struct align_fd afrom, ato; | |
661 | ||
662 | init_afd(&afrom, from); | |
663 | init_afd(&ato, to); | |
664 | ||
665 | bitmap_offset += (int32_t)__le32_to_cpu(super.bitmap_offset); | |
666 | ||
667 | if (lseek64(from, bitmap_offset<<9, 0) < 0) | |
668 | goto err; | |
669 | if (lseek64(to, bitmap_offset<<9, 0) < 0) | |
670 | goto err; | |
671 | ||
672 | for (written = 0; written < bytes ; ) { | |
673 | int n = bytes - written; | |
674 | if (n > 4096) | |
675 | n = 4096; | |
676 | if (aread(&afrom, buf, n) != n) | |
677 | goto err; | |
678 | if (written == 0) { | |
679 | /* have the header, can calculate | |
680 | * correct bitmap bytes */ | |
681 | bitmap_super_t *bms; | |
682 | int bits; | |
683 | bms = (void*)buf; | |
684 | bits = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9); | |
685 | bytes = (bits+7) >> 3; | |
686 | bytes += sizeof(bitmap_super_t); | |
687 | bytes = ROUND_UP(bytes, 512); | |
688 | if (n > bytes) | |
689 | n = bytes; | |
690 | } | |
691 | if (awrite(&ato, buf, n) != n) | |
692 | goto err; | |
693 | written += n; | |
694 | } | |
695 | } | |
696 | ||
697 | if (super.bblog_size != 0 && | |
698 | __le32_to_cpu(super.bblog_size) <= 100 && | |
699 | super.bblog_offset != 0 && | |
700 | (super.feature_map & __le32_to_cpu(MD_FEATURE_BAD_BLOCKS))) { | |
701 | /* There is a bad block log */ | |
702 | unsigned long long bb_offset = sb_offset; | |
703 | int bytes = __le32_to_cpu(super.bblog_size) * 512; | |
704 | int written = 0; | |
705 | struct align_fd afrom, ato; | |
706 | ||
707 | init_afd(&afrom, from); | |
708 | init_afd(&ato, to); | |
709 | ||
710 | bb_offset += (int32_t)__le32_to_cpu(super.bblog_offset); | |
711 | ||
712 | if (lseek64(from, bb_offset<<9, 0) < 0) | |
713 | goto err; | |
714 | if (lseek64(to, bb_offset<<9, 0) < 0) | |
715 | goto err; | |
716 | ||
717 | for (written = 0; written < bytes ; ) { | |
718 | int n = bytes - written; | |
719 | if (n > 4096) | |
720 | n = 4096; | |
721 | if (aread(&afrom, buf, n) != n) | |
722 | goto err; | |
723 | ||
724 | if (awrite(&ato, buf, n) != n) | |
725 | goto err; | |
726 | written += n; | |
727 | } | |
728 | } | |
729 | ||
730 | free(buf); | |
731 | return 0; | |
732 | ||
733 | err: | |
734 | free(buf); | |
735 | return 1; | |
736 | } | |
737 | ||
3da92f27 | 738 | static void detail_super1(struct supertype *st, char *homehost) |
82d9eba6 | 739 | { |
3da92f27 | 740 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 | 741 | int i; |
b6750aa8 | 742 | int l = homehost ? strlen(homehost) : 0; |
82d9eba6 | 743 | |
b6750aa8 NB |
744 | printf(" Name : %.32s", sb->set_name); |
745 | if (l > 0 && l < 32 && | |
746 | sb->set_name[l] == ':' && | |
747 | strncmp(sb->set_name, homehost, l) == 0) | |
748 | printf(" (local to host %s)", homehost); | |
749 | printf("\n UUID : "); | |
82d9eba6 | 750 | for (i=0; i<16; i++) { |
82d9eba6 | 751 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 752 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 | 753 | } |
570c0542 | 754 | printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb->events)); |
82d9eba6 NB |
755 | } |
756 | ||
3da92f27 | 757 | static void brief_detail_super1(struct supertype *st) |
82d9eba6 | 758 | { |
3da92f27 | 759 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 NB |
760 | int i; |
761 | ||
7103b9b8 N |
762 | if (sb->set_name[0]) { |
763 | printf(" name="); | |
764 | print_quoted(sb->set_name); | |
765 | } | |
82d9eba6 NB |
766 | printf(" UUID="); |
767 | for (i=0; i<16; i++) { | |
82d9eba6 | 768 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 769 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 NB |
770 | } |
771 | } | |
772 | ||
0d726f17 | 773 | static void export_detail_super1(struct supertype *st) |
54bad364 | 774 | { |
3da92f27 | 775 | struct mdp_superblock_1 *sb = st->sb; |
54bad364 KS |
776 | int i; |
777 | int len = 32; | |
778 | ||
779 | for (i=0; i<32; i++) | |
780 | if (sb->set_name[i] == '\n' || | |
781 | sb->set_name[i] == '\0') { | |
782 | len = i; | |
783 | break; | |
784 | } | |
785 | if (len) | |
786 | printf("MD_NAME=%.*s\n", len, sb->set_name); | |
54bad364 KS |
787 | } |
788 | ||
6d388a88 N |
789 | static int examine_badblocks_super1(struct supertype *st, int fd, char *devname) |
790 | { | |
791 | struct mdp_superblock_1 *sb = st->sb; | |
792 | unsigned long long offset; | |
793 | int size; | |
794 | __u64 *bbl, *bbp; | |
795 | int i; | |
796 | ||
797 | if (!sb->bblog_size || __le32_to_cpu(sb->bblog_size) > 100 | |
798 | || !sb->bblog_offset){ | |
799 | printf("No bad-blocks list configured on %s\n", devname); | |
800 | return 0; | |
801 | } | |
802 | if ((sb->feature_map & __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
803 | == 0) { | |
804 | printf("Bad-blocks list is empty in %s\n", devname); | |
805 | return 0; | |
806 | } | |
807 | ||
808 | size = __le32_to_cpu(sb->bblog_size)* 512; | |
0cf83229 N |
809 | if (posix_memalign((void**)&bbl, 4096, size) != 0) { |
810 | pr_err("%s could not allocate badblocks list\n", __func__); | |
811 | return 0; | |
812 | } | |
6d388a88 N |
813 | offset = __le64_to_cpu(sb->super_offset) + |
814 | (int)__le32_to_cpu(sb->bblog_offset); | |
815 | offset <<= 9; | |
816 | if (lseek64(fd, offset, 0) < 0) { | |
817 | pr_err("Cannot seek to bad-blocks list\n"); | |
818 | return 1; | |
819 | } | |
820 | if (read(fd, bbl, size) != size) { | |
821 | pr_err("Cannot read bad-blocks list\n"); | |
822 | return 1; | |
823 | } | |
824 | /* 64bits per entry. 10 bits is block-count, 54 bits is block | |
825 | * offset. Blocks are sectors unless bblog->shift makes them bigger | |
826 | */ | |
827 | bbp = (__u64*)bbl; | |
828 | printf("Bad-blocks on %s:\n", devname); | |
829 | for (i = 0; i < size/8; i++, bbp++) { | |
830 | __u64 bb = __le64_to_cpu(*bbp); | |
831 | int count = bb & 0x3ff; | |
832 | unsigned long long sector = bb >> 10; | |
833 | ||
834 | if (bb + 1 == 0) | |
835 | break; | |
836 | ||
837 | sector <<= sb->bblog_shift; | |
838 | count <<= sb->bblog_shift; | |
839 | ||
840 | printf("%20llu for %d sectors\n", sector, count); | |
841 | } | |
842 | return 0; | |
843 | } | |
844 | ||
c7654afc NB |
845 | #endif |
846 | ||
3da92f27 | 847 | static int match_home1(struct supertype *st, char *homehost) |
83b6208e | 848 | { |
3da92f27 | 849 | struct mdp_superblock_1 *sb = st->sb; |
83b6208e NB |
850 | int l = homehost ? strlen(homehost) : 0; |
851 | ||
852 | return (l > 0 && l < 32 && | |
853 | sb->set_name[l] == ':' && | |
854 | strncmp(sb->set_name, homehost, l) == 0); | |
855 | } | |
856 | ||
3da92f27 | 857 | static void uuid_from_super1(struct supertype *st, int uuid[4]) |
82d9eba6 | 858 | { |
3da92f27 | 859 | struct mdp_superblock_1 *super = st->sb; |
82d9eba6 NB |
860 | char *cuuid = (char*)uuid; |
861 | int i; | |
862 | for (i=0; i<16; i++) | |
863 | cuuid[i] = super->set_uuid[i]; | |
864 | } | |
865 | ||
a5d85af7 | 866 | static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map) |
82d9eba6 | 867 | { |
3da92f27 | 868 | struct mdp_superblock_1 *sb = st->sb; |
80bf9135 N |
869 | struct bitmap_super_s *bsb = (void*)(((char*)sb)+MAX_SB_SIZE); |
870 | struct misc_dev_info *misc = (void*)(((char*)sb)+MAX_SB_SIZE+BM_SUPER_SIZE); | |
82d9eba6 | 871 | int working = 0; |
f21e18ca | 872 | unsigned int i; |
a5d85af7 N |
873 | unsigned int role; |
874 | unsigned int map_disks = info->array.raid_disks; | |
80bf9135 N |
875 | unsigned long long super_offset; |
876 | unsigned long long data_size; | |
82d9eba6 | 877 | |
95eeceeb | 878 | memset(info, 0, sizeof(*info)); |
82d9eba6 | 879 | info->array.major_version = 1; |
b8ac1967 | 880 | info->array.minor_version = st->minor_version; |
82d9eba6 NB |
881 | info->array.patch_version = 0; |
882 | info->array.raid_disks = __le32_to_cpu(sb->raid_disks); | |
883 | info->array.level = __le32_to_cpu(sb->level); | |
265e0f17 | 884 | info->array.layout = __le32_to_cpu(sb->layout); |
82d9eba6 NB |
885 | info->array.md_minor = -1; |
886 | info->array.ctime = __le64_to_cpu(sb->ctime); | |
353632d9 | 887 | info->array.utime = __le64_to_cpu(sb->utime); |
bed256c2 | 888 | info->array.chunk_size = __le32_to_cpu(sb->chunksize)*512; |
4855f95c | 889 | info->array.state = |
d5977056 | 890 | (__le64_to_cpu(sb->resync_offset) == MaxSector) |
4855f95c | 891 | ? 1 : 0; |
353632d9 NB |
892 | |
893 | info->data_offset = __le64_to_cpu(sb->data_offset); | |
894 | info->component_size = __le64_to_cpu(sb->size); | |
c0c1acd6 | 895 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET)) |
34a13953 | 896 | info->bitmap_offset = (int32_t)__le32_to_cpu(sb->bitmap_offset); |
82d9eba6 NB |
897 | |
898 | info->disk.major = 0; | |
899 | info->disk.minor = 0; | |
fbf8a0b7 | 900 | info->disk.number = __le32_to_cpu(sb->dev_number); |
82d9eba6 | 901 | if (__le32_to_cpu(sb->dev_number) >= __le32_to_cpu(sb->max_dev) || |
911cead7 | 902 | __le32_to_cpu(sb->dev_number) >= MAX_DEVS) |
82d9eba6 NB |
903 | role = 0xfffe; |
904 | else | |
905 | role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]); | |
906 | ||
80bf9135 N |
907 | super_offset = __le64_to_cpu(sb->super_offset); |
908 | data_size = __le64_to_cpu(sb->size); | |
909 | if (info->data_offset < super_offset) { | |
910 | unsigned long long end; | |
911 | info->space_before = info->data_offset; | |
912 | end = super_offset; | |
cc3130a7 N |
913 | |
914 | if (sb->bblog_offset && sb->bblog_size) { | |
915 | unsigned long long bboffset = super_offset; | |
916 | bboffset += (int32_t)__le32_to_cpu(sb->bblog_offset); | |
917 | if (bboffset < end) | |
918 | end = bboffset; | |
919 | } | |
920 | ||
921 | if (super_offset + info->bitmap_offset < end) | |
922 | end = super_offset + info->bitmap_offset; | |
923 | ||
80bf9135 N |
924 | if (info->data_offset + data_size < end) |
925 | info->space_after = end - data_size - info->data_offset; | |
926 | else | |
927 | info->space_after = 0; | |
928 | } else { | |
cc3130a7 N |
929 | unsigned long long earliest; |
930 | earliest = super_offset + (32+4)*2; /* match kernel */ | |
80bf9135 N |
931 | if (info->bitmap_offset > 0) { |
932 | unsigned long long bmend = info->bitmap_offset; | |
933 | unsigned long long size = __le64_to_cpu(bsb->sync_size); | |
934 | size /= __le32_to_cpu(bsb->chunksize) >> 9; | |
935 | size = (size + 7) >> 3; | |
936 | size += sizeof(bitmap_super_t); | |
937 | size = ROUND_UP(size, 4096); | |
938 | size /= 512; | |
cc3130a7 N |
939 | bmend += size; |
940 | if (bmend > earliest) | |
941 | bmend = earliest; | |
942 | } | |
943 | if (sb->bblog_offset && sb->bblog_size) { | |
944 | unsigned long long bbend = super_offset; | |
945 | bbend += (int32_t)__le32_to_cpu(sb->bblog_offset); | |
946 | bbend += __le32_to_cpu(sb->bblog_size); | |
947 | if (bbend > earliest) | |
948 | earliest = bbend; | |
949 | } | |
950 | if (earliest < info->data_offset) | |
951 | info->space_before = info->data_offset - earliest; | |
952 | else | |
953 | info->space_before = 0; | |
80bf9135 N |
954 | info->space_after = misc->device_size - data_size - info->data_offset; |
955 | } | |
c4b26c64 N |
956 | if (info->space_before == 0 && info->space_after == 0) { |
957 | /* It will look like we don't support data_offset changes, | |
958 | * be we do - it's just that there is no room. | |
959 | * A change that reduced the number of devices should | |
960 | * still be allowed, so set the otherwise useless value of '1' | |
961 | */ | |
962 | info->space_after = 1; | |
963 | } | |
80bf9135 | 964 | |
82d9eba6 NB |
965 | info->disk.raid_disk = -1; |
966 | switch(role) { | |
967 | case 0xFFFF: | |
f22385f9 | 968 | info->disk.state = 0; /* spare: not active, not sync, not faulty */ |
82d9eba6 NB |
969 | break; |
970 | case 0xFFFE: | |
971 | info->disk.state = 1; /* faulty */ | |
972 | break; | |
973 | default: | |
974 | info->disk.state = 6; /* active and in sync */ | |
975 | info->disk.raid_disk = role; | |
976 | } | |
cb099724 N |
977 | if (sb->devflags & WriteMostly1) |
978 | info->disk.state |= (1 << MD_DISK_WRITEMOSTLY); | |
82d9eba6 | 979 | info->events = __le64_to_cpu(sb->events); |
1522c538 | 980 | sprintf(info->text_version, "1.%d", st->minor_version); |
a67dd8cc | 981 | info->safe_mode_delay = 200; |
82d9eba6 NB |
982 | |
983 | memcpy(info->uuid, sb->set_uuid, 16); | |
984 | ||
31317663 NB |
985 | strncpy(info->name, sb->set_name, 32); |
986 | info->name[32] = 0; | |
947fd4dd | 987 | |
aacb2f81 | 988 | if ((__le32_to_cpu(sb->feature_map)&MD_FEATURE_REPLACEMENT)) { |
72e7fb13 | 989 | info->disk.state &= ~(1 << MD_DISK_SYNC); |
aacb2f81 N |
990 | info->disk.state |= 1 << MD_DISK_REPLACEMENT; |
991 | } | |
992 | ||
993 | ||
921d9e16 N |
994 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET)) |
995 | info->recovery_start = __le32_to_cpu(sb->recovery_offset); | |
996 | else | |
997 | info->recovery_start = MaxSector; | |
998 | ||
353632d9 NB |
999 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) { |
1000 | info->reshape_active = 1; | |
5e88ab2e N |
1001 | if (info->array.level == 10) |
1002 | info->reshape_active |= RESHAPE_NO_BACKUP; | |
353632d9 NB |
1003 | info->reshape_progress = __le64_to_cpu(sb->reshape_position); |
1004 | info->new_level = __le32_to_cpu(sb->new_level); | |
1005 | info->delta_disks = __le32_to_cpu(sb->delta_disks); | |
1006 | info->new_layout = __le32_to_cpu(sb->new_layout); | |
bed256c2 | 1007 | info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9; |
4180aa4d N |
1008 | if (info->delta_disks < 0) |
1009 | info->array.raid_disks -= info->delta_disks; | |
353632d9 NB |
1010 | } else |
1011 | info->reshape_active = 0; | |
1012 | ||
6e75048b AK |
1013 | info->recovery_blocked = info->reshape_active; |
1014 | ||
a5d85af7 N |
1015 | if (map) |
1016 | for (i=0; i<map_disks; i++) | |
1017 | map[i] = 0; | |
f21e18ca | 1018 | for (i = 0; i < __le32_to_cpu(sb->max_dev); i++) { |
82d9eba6 | 1019 | role = __le16_to_cpu(sb->dev_roles[i]); |
a5d85af7 | 1020 | if (/*role == 0xFFFF || */role < (unsigned) info->array.raid_disks) { |
82d9eba6 | 1021 | working++; |
a5d85af7 N |
1022 | if (map && role < map_disks) |
1023 | map[role] = 1; | |
1024 | } | |
82d9eba6 NB |
1025 | } |
1026 | ||
1027 | info->array.working_disks = working; | |
1028 | } | |
1029 | ||
00bbdbda N |
1030 | static struct mdinfo *container_content1(struct supertype *st, char *subarray) |
1031 | { | |
1032 | struct mdinfo *info; | |
1033 | ||
1034 | if (subarray) | |
1035 | return NULL; | |
1036 | ||
503975b9 | 1037 | info = xmalloc(sizeof(*info)); |
00bbdbda N |
1038 | getinfo_super1(st, info, NULL); |
1039 | return info; | |
1040 | } | |
1041 | ||
68c7d6d7 | 1042 | static int update_super1(struct supertype *st, struct mdinfo *info, |
3da92f27 | 1043 | char *update, |
e5eac01f NB |
1044 | char *devname, int verbose, |
1045 | int uuid_set, char *homehost) | |
82d9eba6 | 1046 | { |
5a31170d N |
1047 | /* NOTE: for 'assemble' and 'force' we need to return non-zero |
1048 | * if any change was made. For others, the return value is | |
1049 | * ignored. | |
c4d831e1 | 1050 | */ |
82d9eba6 | 1051 | int rv = 0; |
3da92f27 | 1052 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 | 1053 | |
def11332 N |
1054 | if (strcmp(update, "homehost") == 0 && |
1055 | homehost) { | |
1056 | /* Note that 'homehost' is special as it is really | |
1057 | * a "name" update. | |
1058 | */ | |
1059 | char *c; | |
1060 | update = "name"; | |
1061 | c = strchr(sb->set_name, ':'); | |
1062 | if (c) | |
1063 | strncpy(info->name, c+1, 31 - (c-sb->set_name)); | |
1064 | else | |
1065 | strncpy(info->name, sb->set_name, 32); | |
1066 | info->name[32] = 0; | |
1067 | } | |
1068 | ||
67a8c82d NB |
1069 | if (strcmp(update, "force-one")==0) { |
1070 | /* Not enough devices for a working array, | |
1071 | * so bring this one up-to-date | |
1072 | */ | |
c4d831e1 NB |
1073 | if (sb->events != __cpu_to_le64(info->events)) |
1074 | rv = 1; | |
6d3d5804 | 1075 | sb->events = __cpu_to_le64(info->events); |
1e2b2765 | 1076 | } else if (strcmp(update, "force-array")==0) { |
67a8c82d NB |
1077 | /* Degraded array and 'force' requests to |
1078 | * maybe need to mark it 'clean'. | |
1079 | */ | |
82d9eba6 NB |
1080 | switch(__le32_to_cpu(sb->level)) { |
1081 | case 5: case 4: case 6: | |
1082 | /* need to force clean */ | |
b7528a20 | 1083 | if (sb->resync_offset != MaxSector) |
c4d831e1 | 1084 | rv = 1; |
b7528a20 | 1085 | sb->resync_offset = MaxSector; |
82d9eba6 | 1086 | } |
1e2b2765 | 1087 | } else if (strcmp(update, "assemble")==0) { |
82d9eba6 NB |
1088 | int d = info->disk.number; |
1089 | int want; | |
aacb2f81 | 1090 | if (info->disk.state & (1<<MD_DISK_ACTIVE)) |
a2ce5a1a | 1091 | want = info->disk.raid_disk; |
82d9eba6 NB |
1092 | else |
1093 | want = 0xFFFF; | |
a2ce5a1a N |
1094 | if (sb->dev_roles[d] != __cpu_to_le16(want)) { |
1095 | sb->dev_roles[d] = __cpu_to_le16(want); | |
82d9eba6 NB |
1096 | rv = 1; |
1097 | } | |
d43494fc N |
1098 | if (info->reshape_active && |
1099 | sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) && | |
1100 | info->delta_disks >= 0 && | |
1101 | info->reshape_progress < __le64_to_cpu(sb->reshape_position)) { | |
1102 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
1103 | rv = 1; | |
1104 | } | |
1105 | if (info->reshape_active && | |
1106 | sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) && | |
1107 | info->delta_disks < 0 && | |
1108 | info->reshape_progress > __le64_to_cpu(sb->reshape_position)) { | |
1109 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); | |
1110 | rv = 1; | |
1111 | } | |
1e2b2765 | 1112 | } else if (strcmp(update, "linear-grow-new") == 0) { |
f21e18ca | 1113 | unsigned int i; |
1c6cb603 | 1114 | int rfd, fd; |
f21e18ca | 1115 | unsigned int max = __le32_to_cpu(sb->max_dev); |
f752781f NB |
1116 | |
1117 | for (i=0 ; i < max ; i++) | |
1118 | if (__le16_to_cpu(sb->dev_roles[i]) >= 0xfffe) | |
1119 | break; | |
1120 | sb->dev_number = __cpu_to_le32(i); | |
1121 | info->disk.number = i; | |
1122 | if (max >= __le32_to_cpu(sb->max_dev)) | |
1123 | sb->max_dev = __cpu_to_le32(max+1); | |
1124 | ||
1125 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || | |
1126 | read(rfd, sb->device_uuid, 16) != 16) { | |
caa0f6c6 N |
1127 | __u32 r[4] = {random(), random(), random(), random()}; |
1128 | memcpy(sb->device_uuid, r, 16); | |
f752781f | 1129 | } |
7cbeb80e N |
1130 | if (rfd >= 0) |
1131 | close(rfd); | |
f752781f NB |
1132 | |
1133 | sb->dev_roles[i] = | |
1134 | __cpu_to_le16(info->disk.raid_disk); | |
1c6cb603 N |
1135 | |
1136 | fd = open(devname, O_RDONLY); | |
1137 | if (fd >= 0) { | |
1138 | unsigned long long ds; | |
1139 | get_dev_size(fd, devname, &ds); | |
1140 | close(fd); | |
1141 | ds >>= 9; | |
1142 | if (__le64_to_cpu(sb->super_offset) < | |
1143 | __le64_to_cpu(sb->data_offset)) { | |
1144 | sb->data_size = __cpu_to_le64( | |
1145 | ds - __le64_to_cpu(sb->data_offset)); | |
1146 | } else { | |
1147 | ds -= 8*2; | |
1148 | ds &= ~(unsigned long long)(4*2-1); | |
1149 | sb->super_offset = __cpu_to_le64(ds); | |
1150 | sb->data_size = __cpu_to_le64( | |
1151 | ds - __le64_to_cpu(sb->data_offset)); | |
1152 | } | |
1153 | } | |
1e2b2765 | 1154 | } else if (strcmp(update, "linear-grow-update") == 0) { |
82d9eba6 | 1155 | sb->raid_disks = __cpu_to_le32(info->array.raid_disks); |
f752781f NB |
1156 | sb->dev_roles[info->disk.number] = |
1157 | __cpu_to_le16(info->disk.raid_disk); | |
1e2b2765 | 1158 | } else if (strcmp(update, "resync") == 0) { |
82d9eba6 | 1159 | /* make sure resync happens */ |
434b7755 | 1160 | sb->resync_offset = 0ULL; |
1e2b2765 | 1161 | } else if (strcmp(update, "uuid") == 0) { |
350f29f9 | 1162 | copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid); |
bf4fb153 | 1163 | |
14263cf1 NB |
1164 | if (__le32_to_cpu(sb->feature_map)&MD_FEATURE_BITMAP_OFFSET) { |
1165 | struct bitmap_super_s *bm; | |
911cead7 | 1166 | bm = (struct bitmap_super_s*)(st->sb+MAX_SB_SIZE); |
bf4fb153 | 1167 | memcpy(bm->uuid, sb->set_uuid, 16); |
14263cf1 | 1168 | } |
5a31170d N |
1169 | } else if (strcmp(update, "no-bitmap") == 0) { |
1170 | sb->feature_map &= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET); | |
688e99a7 N |
1171 | } else if (strcmp(update, "bbl") == 0) { |
1172 | /* only possible if there is room after the bitmap, or if | |
1173 | * there is no bitmap | |
1174 | */ | |
1175 | unsigned long long sb_offset = __le64_to_cpu(sb->super_offset); | |
1176 | unsigned long long data_offset = __le64_to_cpu(sb->data_offset); | |
1177 | long bitmap_offset = (long)__le64_to_cpu(sb->bitmap_offset); | |
1178 | long bm_sectors = 0; | |
1179 | long space; | |
1180 | ||
5a23a06e | 1181 | #ifndef MDASSEMBLE |
688e99a7 N |
1182 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { |
1183 | struct bitmap_super_s *bsb; | |
1184 | bsb = (struct bitmap_super_s *)(((char*)sb)+MAX_SB_SIZE); | |
1185 | bm_sectors = bitmap_sectors(bsb); | |
1186 | } | |
5a23a06e | 1187 | #endif |
688e99a7 N |
1188 | if (sb_offset < data_offset) { |
1189 | /* 1.1 or 1.2. Put bbl just before data | |
1190 | */ | |
1191 | long bb_offset; | |
1192 | space = data_offset - sb_offset; | |
1193 | bb_offset = space - 8; | |
1194 | if (bm_sectors && bitmap_offset > 0) | |
1195 | space -= (bitmap_offset + bm_sectors); | |
1196 | else | |
1197 | space -= 8; /* The superblock */ | |
1198 | if (space >= 8) { | |
1199 | sb->bblog_size = __cpu_to_le16(8); | |
1200 | sb->bblog_offset = __cpu_to_le32(bb_offset); | |
1201 | } | |
1202 | } else { | |
1203 | /* 1.0 - Put bbl just before super block */ | |
1204 | if (bm_sectors && bitmap_offset < 0) | |
1205 | space = -bitmap_offset - bm_sectors; | |
1206 | else | |
1207 | space = sb_offset - data_offset - | |
1208 | __le64_to_cpu(sb->data_size); | |
1209 | if (space >= 8) { | |
1210 | sb->bblog_size = __cpu_to_le16(8); | |
1211 | sb->bblog_offset = __cpu_to_le32((unsigned)-8); | |
1212 | } | |
1213 | } | |
1214 | } else if (strcmp(update, "no-bbl") == 0) { | |
1215 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
1216 | pr_err("Cannot remove active bbl from %s\n",devname); | |
1217 | else { | |
1218 | sb->bblog_size = 0; | |
1219 | sb->bblog_shift = 0; | |
1220 | sb->bblog_offset = 0; | |
1221 | } | |
1e2b2765 | 1222 | } else if (strcmp(update, "name") == 0) { |
c4f12c13 NB |
1223 | if (info->name[0] == 0) |
1224 | sprintf(info->name, "%d", info->array.md_minor); | |
1225 | memset(sb->set_name, 0, sizeof(sb->set_name)); | |
1226 | if (homehost && | |
1227 | strchr(info->name, ':') == NULL && | |
1228 | strlen(homehost)+1+strlen(info->name) < 32) { | |
1229 | strcpy(sb->set_name, homehost); | |
1230 | strcat(sb->set_name, ":"); | |
1231 | strcat(sb->set_name, info->name); | |
1232 | } else | |
1233 | strcpy(sb->set_name, info->name); | |
1e2b2765 | 1234 | } else if (strcmp(update, "devicesize") == 0 && |
bee8ec56 NB |
1235 | __le64_to_cpu(sb->super_offset) < |
1236 | __le64_to_cpu(sb->data_offset)) { | |
1237 | /* set data_size to device size less data_offset */ | |
1238 | struct misc_dev_info *misc = (struct misc_dev_info*) | |
911cead7 | 1239 | (st->sb + MAX_SB_SIZE + BM_SUPER_SIZE); |
eb9199fb NB |
1240 | printf("Size was %llu\n", (unsigned long long) |
1241 | __le64_to_cpu(sb->data_size)); | |
bee8ec56 NB |
1242 | sb->data_size = __cpu_to_le64( |
1243 | misc->device_size - __le64_to_cpu(sb->data_offset)); | |
eb9199fb NB |
1244 | printf("Size is %llu\n", (unsigned long long) |
1245 | __le64_to_cpu(sb->data_size)); | |
1e2b2765 | 1246 | } else if (strcmp(update, "_reshape_progress")==0) |
353632d9 | 1247 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); |
16715c01 DL |
1248 | else if (strcmp(update, "writemostly")==0) |
1249 | sb->devflags |= WriteMostly1; | |
1250 | else if (strcmp(update, "readwrite")==0) | |
1251 | sb->devflags &= ~WriteMostly1; | |
1e2b2765 N |
1252 | else |
1253 | rv = -1; | |
82d9eba6 NB |
1254 | |
1255 | sb->sb_csum = calc_sb_1_csum(sb); | |
1256 | return rv; | |
1257 | } | |
1258 | ||
3da92f27 | 1259 | static int init_super1(struct supertype *st, mdu_array_info_t *info, |
83cd1e97 N |
1260 | unsigned long long size, char *name, char *homehost, |
1261 | int *uuid, unsigned long long data_offset) | |
82d9eba6 | 1262 | { |
6416d527 | 1263 | struct mdp_superblock_1 *sb; |
82d9eba6 | 1264 | int spares; |
34163fc7 | 1265 | int rfd; |
05697ec1 | 1266 | char defname[10]; |
911cead7 | 1267 | int sbsize; |
6416d527 | 1268 | |
3c0bcd46 | 1269 | if (posix_memalign((void**)&sb, 4096, SUPER1_SIZE) != 0) { |
e7b84f9d | 1270 | pr_err("%s could not allocate superblock\n", __func__); |
3d2c4fc7 DW |
1271 | return 0; |
1272 | } | |
1afa9308 | 1273 | memset(sb, 0, SUPER1_SIZE); |
82d9eba6 | 1274 | |
64557c33 | 1275 | st->sb = sb; |
ba7eb04f | 1276 | if (info == NULL) { |
82d9eba6 NB |
1277 | /* zeroing superblock */ |
1278 | return 0; | |
05697ec1 | 1279 | } |
82d9eba6 NB |
1280 | |
1281 | spares = info->working_disks - info->active_disks; | |
911cead7 | 1282 | if (info->raid_disks + spares > MAX_DEVS) { |
e7b84f9d | 1283 | pr_err("too many devices requested: %d+%d > %d\n", |
911cead7 | 1284 | info->raid_disks , spares, MAX_DEVS); |
82d9eba6 NB |
1285 | return 0; |
1286 | } | |
1287 | ||
82d9eba6 NB |
1288 | sb->magic = __cpu_to_le32(MD_SB_MAGIC); |
1289 | sb->major_version = __cpu_to_le32(1); | |
1290 | sb->feature_map = 0; | |
1291 | sb->pad0 = 0; | |
34163fc7 | 1292 | |
350f29f9 NB |
1293 | if (uuid) |
1294 | copy_uuid(sb->set_uuid, uuid, super1.swapuuid); | |
1295 | else { | |
3d3dd91e NB |
1296 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || |
1297 | read(rfd, sb->set_uuid, 16) != 16) { | |
caa0f6c6 N |
1298 | __u32 r[4] = {random(), random(), random(), random()}; |
1299 | memcpy(sb->set_uuid, r, 16); | |
3d3dd91e NB |
1300 | } |
1301 | if (rfd >= 0) close(rfd); | |
34163fc7 | 1302 | } |
82d9eba6 | 1303 | |
05697ec1 NB |
1304 | if (name == NULL || *name == 0) { |
1305 | sprintf(defname, "%d", info->md_minor); | |
1306 | name = defname; | |
1307 | } | |
05697ec1 NB |
1308 | if (homehost && |
1309 | strchr(name, ':')== NULL && | |
1310 | strlen(homehost)+1+strlen(name) < 32) { | |
1311 | strcpy(sb->set_name, homehost); | |
1312 | strcat(sb->set_name, ":"); | |
1313 | strcat(sb->set_name, name); | |
1314 | } else | |
1315 | strcpy(sb->set_name, name); | |
82d9eba6 NB |
1316 | |
1317 | sb->ctime = __cpu_to_le64((unsigned long long)time(0)); | |
1318 | sb->level = __cpu_to_le32(info->level); | |
ea329559 | 1319 | sb->layout = __cpu_to_le32(info->layout); |
5dd497ee | 1320 | sb->size = __cpu_to_le64(size*2ULL); |
82d9eba6 NB |
1321 | sb->chunksize = __cpu_to_le32(info->chunk_size>>9); |
1322 | sb->raid_disks = __cpu_to_le32(info->raid_disks); | |
1323 | ||
83cd1e97 | 1324 | sb->data_offset = __cpu_to_le64(data_offset); |
82d9eba6 NB |
1325 | sb->data_size = __cpu_to_le64(0); |
1326 | sb->super_offset = __cpu_to_le64(0); | |
1327 | sb->recovery_offset = __cpu_to_le64(0); | |
1328 | ||
1329 | sb->utime = sb->ctime; | |
1330 | sb->events = __cpu_to_le64(1); | |
34321279 | 1331 | if (info->state & (1<<MD_SB_CLEAN)) |
b7528a20 | 1332 | sb->resync_offset = MaxSector; |
82d9eba6 NB |
1333 | else |
1334 | sb->resync_offset = 0; | |
911cead7 N |
1335 | sbsize = sizeof(struct mdp_superblock_1) + 2 * (info->raid_disks + spares); |
1336 | sbsize = ROUND_UP(sbsize, 512); | |
1337 | sb->max_dev = __cpu_to_le32((sbsize - sizeof(struct mdp_superblock_1)) / 2); | |
82d9eba6 | 1338 | |
911cead7 | 1339 | memset(sb->dev_roles, 0xff, MAX_SB_SIZE - sizeof(struct mdp_superblock_1)); |
82d9eba6 | 1340 | |
82d9eba6 NB |
1341 | return 1; |
1342 | } | |
1343 | ||
111d01fc NB |
1344 | struct devinfo { |
1345 | int fd; | |
1346 | char *devname; | |
72ca9bcf | 1347 | long long data_offset; |
111d01fc NB |
1348 | mdu_disk_info_t disk; |
1349 | struct devinfo *next; | |
1350 | }; | |
0e600426 | 1351 | #ifndef MDASSEMBLE |
82d9eba6 | 1352 | /* Add a device to the superblock being created */ |
f20c3968 | 1353 | static int add_to_super1(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf | 1354 | int fd, char *devname, unsigned long long data_offset) |
82d9eba6 | 1355 | { |
3da92f27 | 1356 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 | 1357 | __u16 *rp = sb->dev_roles + dk->number; |
111d01fc NB |
1358 | struct devinfo *di, **dip; |
1359 | ||
dfd4d8ee | 1360 | if ((dk->state & 6) == 6) /* active, sync */ |
82d9eba6 | 1361 | *rp = __cpu_to_le16(dk->raid_disk); |
412ca2e5 | 1362 | else if ((dk->state & ~2) == 0) /* active or idle -> spare */ |
82d9eba6 | 1363 | *rp = 0xffff; |
34163fc7 | 1364 | else |
82d9eba6 | 1365 | *rp = 0xfffe; |
111d01fc | 1366 | |
f21e18ca | 1367 | if (dk->number >= (int)__le32_to_cpu(sb->max_dev) && |
911cead7 | 1368 | __le32_to_cpu(sb->max_dev) < MAX_DEVS) |
9df04a4f | 1369 | sb->max_dev = __cpu_to_le32(dk->number+1); |
8844e291 | 1370 | |
d2ca6449 | 1371 | sb->dev_number = __cpu_to_le32(dk->number); |
16715c01 | 1372 | sb->devflags = 0; /* don't copy another disks flags */ |
d2ca6449 NB |
1373 | sb->sb_csum = calc_sb_1_csum(sb); |
1374 | ||
111d01fc NB |
1375 | dip = (struct devinfo **)&st->info; |
1376 | while (*dip) | |
1377 | dip = &(*dip)->next; | |
503975b9 | 1378 | di = xmalloc(sizeof(struct devinfo)); |
111d01fc NB |
1379 | di->fd = fd; |
1380 | di->devname = devname; | |
1381 | di->disk = *dk; | |
72ca9bcf | 1382 | di->data_offset = data_offset; |
111d01fc NB |
1383 | di->next = NULL; |
1384 | *dip = di; | |
f20c3968 DW |
1385 | |
1386 | return 0; | |
82d9eba6 | 1387 | } |
0e600426 | 1388 | #endif |
82d9eba6 | 1389 | |
3da92f27 | 1390 | static void locate_bitmap1(struct supertype *st, int fd); |
14263cf1 | 1391 | |
3da92f27 | 1392 | static int store_super1(struct supertype *st, int fd) |
82d9eba6 | 1393 | { |
3da92f27 | 1394 | struct mdp_superblock_1 *sb = st->sb; |
6fbba4c9 | 1395 | unsigned long long sb_offset; |
308340aa | 1396 | struct align_fd afd; |
82d9eba6 | 1397 | int sbsize; |
5dd497ee | 1398 | unsigned long long dsize; |
96395475 | 1399 | |
beae1dfe NB |
1400 | if (!get_dev_size(fd, NULL, &dsize)) |
1401 | return 1; | |
1402 | ||
1403 | dsize >>= 9; | |
96395475 | 1404 | |
5dd497ee | 1405 | if (dsize < 24) |
96395475 NB |
1406 | return 2; |
1407 | ||
308340aa JS |
1408 | init_afd(&afd, fd); |
1409 | ||
96395475 NB |
1410 | /* |
1411 | * Calculate the position of the superblock. | |
1412 | * It is always aligned to a 4K boundary and | |
1413 | * depending on minor_version, it can be: | |
1414 | * 0: At least 8K, but less than 12K, from end of device | |
1415 | * 1: At start of device | |
1416 | * 2: 4K from start of device. | |
1417 | */ | |
1418 | switch(st->minor_version) { | |
1419 | case 0: | |
5dd497ee | 1420 | sb_offset = dsize; |
96395475 NB |
1421 | sb_offset -= 8*2; |
1422 | sb_offset &= ~(4*2-1); | |
1423 | break; | |
1424 | case 1: | |
6fbba4c9 | 1425 | sb_offset = 0; |
96395475 NB |
1426 | break; |
1427 | case 2: | |
1428 | sb_offset = 4*2; | |
1429 | break; | |
1430 | default: | |
1431 | return -EINVAL; | |
1432 | } | |
1433 | ||
82d9eba6 | 1434 | |
34163fc7 | 1435 | |
6fbba4c9 NB |
1436 | if (sb_offset != __le64_to_cpu(sb->super_offset) && |
1437 | 0 != __le64_to_cpu(sb->super_offset) | |
96395475 | 1438 | ) { |
e7b84f9d | 1439 | pr_err("internal error - sb_offset is wrong\n"); |
96395475 NB |
1440 | abort(); |
1441 | } | |
82d9eba6 | 1442 | |
6fbba4c9 | 1443 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) |
82d9eba6 NB |
1444 | return 3; |
1445 | ||
0a2f1894 | 1446 | sbsize = ROUND_UP(sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev), 512); |
82d9eba6 | 1447 | |
308340aa | 1448 | if (awrite(&afd, sb, sbsize) != sbsize) |
82d9eba6 NB |
1449 | return 4; |
1450 | ||
14263cf1 NB |
1451 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { |
1452 | struct bitmap_super_s *bm = (struct bitmap_super_s*) | |
911cead7 | 1453 | (((char*)sb)+MAX_SB_SIZE); |
14263cf1 | 1454 | if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) { |
3da92f27 | 1455 | locate_bitmap1(st, fd); |
308340aa JS |
1456 | if (awrite(&afd, bm, sizeof(*bm)) != sizeof(*bm)) |
1457 | return 5; | |
14263cf1 NB |
1458 | } |
1459 | } | |
570c0542 | 1460 | fsync(fd); |
82d9eba6 NB |
1461 | return 0; |
1462 | } | |
1463 | ||
3da92f27 | 1464 | static int load_super1(struct supertype *st, int fd, char *devname); |
892debc8 | 1465 | |
2fb749d1 NB |
1466 | static unsigned long choose_bm_space(unsigned long devsize) |
1467 | { | |
1468 | /* if the device is bigger than 8Gig, save 64k for bitmap usage, | |
1469 | * if bigger than 200Gig, save 128k | |
a380e275 N |
1470 | * NOTE: result must be multiple of 4K else bad things happen |
1471 | * on 4K-sector devices. | |
2fb749d1 | 1472 | */ |
69646c14 | 1473 | if (devsize < 64*2) return 0; |
2fb749d1 NB |
1474 | if (devsize - 64*2 >= 200*1024*1024*2) |
1475 | return 128*2; | |
1476 | if (devsize - 4*2 > 8*1024*1024*2) | |
1477 | return 64*2; | |
1478 | return 4*2; | |
1479 | } | |
1480 | ||
e8090005 N |
1481 | static void free_super1(struct supertype *st); |
1482 | ||
111d01fc NB |
1483 | #ifndef MDASSEMBLE |
1484 | static int write_init_super1(struct supertype *st) | |
82d9eba6 | 1485 | { |
3da92f27 | 1486 | struct mdp_superblock_1 *sb = st->sb; |
e8090005 | 1487 | struct supertype *refst; |
dfe47e00 | 1488 | int rfd; |
111d01fc | 1489 | int rv = 0; |
f21e18ca | 1490 | unsigned long long bm_space; |
a380e275 | 1491 | unsigned long long reserved; |
111d01fc | 1492 | struct devinfo *di; |
8fac0577 | 1493 | unsigned long long dsize, array_size; |
508a7f16 | 1494 | unsigned long long sb_offset, headroom; |
83cd1e97 | 1495 | unsigned long long data_offset; |
82d9eba6 | 1496 | |
4687f160 | 1497 | for (di = st->info; di; di = di->next) { |
111d01fc NB |
1498 | if (di->disk.state == 1) |
1499 | continue; | |
a0c8a17f NB |
1500 | if (di->fd < 0) |
1501 | continue; | |
82d9eba6 | 1502 | |
ba728be7 | 1503 | while (Kill(di->devname, NULL, 0, -1, 1) == 0) |
9277cc77 | 1504 | ; |
82d9eba6 | 1505 | |
111d01fc NB |
1506 | sb->dev_number = __cpu_to_le32(di->disk.number); |
1507 | if (di->disk.state & (1<<MD_DISK_WRITEMOSTLY)) | |
adbb382b | 1508 | sb->devflags |= WriteMostly1; |
9a88e7b6 | 1509 | else |
adbb382b | 1510 | sb->devflags &= ~WriteMostly1; |
892debc8 | 1511 | |
111d01fc NB |
1512 | if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 || |
1513 | read(rfd, sb->device_uuid, 16) != 16) { | |
83208785 N |
1514 | __u32 r[4] = {random(), random(), random(), random()}; |
1515 | memcpy(sb->device_uuid, r, 16); | |
111d01fc | 1516 | } |
83208785 N |
1517 | if (rfd >= 0) |
1518 | close(rfd); | |
1519 | ||
111d01fc NB |
1520 | sb->events = 0; |
1521 | ||
e8090005 N |
1522 | refst = dup_super(st); |
1523 | if (load_super1(refst, di->fd, NULL)==0) { | |
1524 | struct mdp_superblock_1 *refsb = refst->sb; | |
111d01fc NB |
1525 | |
1526 | memcpy(sb->device_uuid, refsb->device_uuid, 16); | |
1527 | if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) { | |
1528 | /* same array, so preserve events and | |
1529 | * dev_number */ | |
1530 | sb->events = refsb->events; | |
1531 | /* bugs in 2.6.17 and earlier mean the | |
1532 | * dev_number chosen in Manage must be preserved | |
1533 | */ | |
1534 | if (get_linux_version() >= 2006018) | |
1535 | sb->dev_number = refsb->dev_number; | |
1536 | } | |
e8090005 | 1537 | free_super1(refst); |
892debc8 | 1538 | } |
e8090005 | 1539 | free(refst); |
34163fc7 | 1540 | |
40114213 JS |
1541 | if (!get_dev_size(di->fd, NULL, &dsize)) { |
1542 | rv = 1; | |
1543 | goto error_out; | |
1544 | } | |
111d01fc | 1545 | dsize >>= 9; |
82d9eba6 | 1546 | |
111d01fc NB |
1547 | if (dsize < 24) { |
1548 | close(di->fd); | |
40114213 JS |
1549 | rv = 2; |
1550 | goto error_out; | |
111d01fc | 1551 | } |
82d9eba6 NB |
1552 | |
1553 | ||
111d01fc NB |
1554 | /* |
1555 | * Calculate the position of the superblock. | |
1556 | * It is always aligned to a 4K boundary and | |
1557 | * depending on minor_version, it can be: | |
1558 | * 0: At least 8K, but less than 12K, from end of device | |
1559 | * 1: At start of device | |
1560 | * 2: 4K from start of device. | |
1561 | * Depending on the array size, we might leave extra space | |
1562 | * for a bitmap. | |
bf95d0f3 | 1563 | * Also leave 4K for bad-block log. |
111d01fc NB |
1564 | */ |
1565 | array_size = __le64_to_cpu(sb->size); | |
bf95d0f3 N |
1566 | /* work out how much space we left for a bitmap, |
1567 | * Add 8 sectors for bad block log */ | |
1568 | bm_space = choose_bm_space(array_size) + 8; | |
111d01fc | 1569 | |
508a7f16 N |
1570 | /* We try to leave 0.1% at the start for reshape |
1571 | * operations, but limit this to 128Meg (0.1% of 10Gig) | |
1572 | * which is plenty for efficient reshapes | |
cb19a251 N |
1573 | * However we make it at least 2 chunks as one chunk |
1574 | * is minimum needed for reshape. | |
508a7f16 N |
1575 | */ |
1576 | headroom = 128 * 1024 * 2; | |
cb19a251 N |
1577 | while (headroom << 10 > array_size && |
1578 | headroom/2 >= __le32_to_cpu(sb->chunksize) * 2) | |
508a7f16 N |
1579 | headroom >>= 1; |
1580 | ||
72ca9bcf | 1581 | data_offset = di->data_offset; |
111d01fc NB |
1582 | switch(st->minor_version) { |
1583 | case 0: | |
1584 | sb_offset = dsize; | |
1585 | sb_offset -= 8*2; | |
1586 | sb_offset &= ~(4*2-1); | |
1587 | sb->super_offset = __cpu_to_le64(sb_offset); | |
83cd1e97 N |
1588 | if (data_offset == INVALID_SECTORS) |
1589 | sb->data_offset = 0; | |
f21e18ca | 1590 | if (sb_offset < array_size + bm_space) |
a380e275 | 1591 | bm_space = sb_offset - array_size; |
111d01fc | 1592 | sb->data_size = __cpu_to_le64(sb_offset - bm_space); |
bf95d0f3 N |
1593 | if (bm_space >= 8) { |
1594 | sb->bblog_size = __cpu_to_le16(8); | |
1595 | sb->bblog_offset = __cpu_to_le32((unsigned)-8); | |
1596 | } | |
111d01fc NB |
1597 | break; |
1598 | case 1: | |
1599 | sb->super_offset = __cpu_to_le64(0); | |
83cd1e97 N |
1600 | if (data_offset == INVALID_SECTORS) { |
1601 | reserved = bm_space + 4*2; | |
1602 | if (reserved < headroom) | |
1603 | reserved = headroom; | |
1604 | if (reserved + array_size > dsize) | |
1605 | reserved = dsize - array_size; | |
1606 | /* Try for multiple of 1Meg so it is nicely aligned */ | |
1607 | #define ONE_MEG (2*1024) | |
1608 | if (reserved > ONE_MEG) | |
1609 | reserved = (reserved/ONE_MEG) * ONE_MEG; | |
1610 | ||
1611 | /* force 4K alignment */ | |
1612 | reserved &= ~7ULL; | |
1613 | ||
1614 | } else | |
1615 | reserved = data_offset; | |
a380e275 N |
1616 | |
1617 | sb->data_offset = __cpu_to_le64(reserved); | |
1618 | sb->data_size = __cpu_to_le64(dsize - reserved); | |
bf95d0f3 N |
1619 | if (reserved >= 16) { |
1620 | sb->bblog_size = __cpu_to_le16(8); | |
1621 | sb->bblog_offset = __cpu_to_le32(reserved-8); | |
1622 | } | |
111d01fc NB |
1623 | break; |
1624 | case 2: | |
1625 | sb_offset = 4*2; | |
1626 | sb->super_offset = __cpu_to_le64(4*2); | |
83cd1e97 N |
1627 | if (data_offset == INVALID_SECTORS) { |
1628 | if (4*2 + 4*2 + bm_space + array_size | |
1629 | > dsize) | |
1630 | bm_space = dsize - array_size | |
1631 | - 4*2 - 4*2; | |
1632 | ||
1633 | reserved = bm_space + 4*2 + 4*2; | |
1634 | if (reserved < headroom) | |
1635 | reserved = headroom; | |
1636 | if (reserved + array_size > dsize) | |
1637 | reserved = dsize - array_size; | |
1638 | /* Try for multiple of 1Meg so it is nicely aligned */ | |
1639 | #define ONE_MEG (2*1024) | |
1640 | if (reserved > ONE_MEG) | |
1641 | reserved = (reserved/ONE_MEG) * ONE_MEG; | |
1642 | ||
1643 | /* force 4K alignment */ | |
1644 | reserved &= ~7ULL; | |
1645 | ||
1646 | } else | |
1647 | reserved = data_offset; | |
a380e275 N |
1648 | |
1649 | sb->data_offset = __cpu_to_le64(reserved); | |
1650 | sb->data_size = __cpu_to_le64(dsize - reserved); | |
bf95d0f3 N |
1651 | if (reserved >= 16+16) { |
1652 | sb->bblog_size = __cpu_to_le16(8); | |
1653 | /* '8' sectors for the bblog, and another '8' | |
1654 | * because we want offset from superblock, not | |
1655 | * start of device. | |
1656 | */ | |
1657 | sb->bblog_offset = __cpu_to_le32(reserved-8-8); | |
1658 | } | |
111d01fc NB |
1659 | break; |
1660 | default: | |
e7b84f9d N |
1661 | pr_err("Failed to write invalid " |
1662 | "metadata format 1.%i to %s\n", | |
1663 | st->minor_version, di->devname); | |
40114213 JS |
1664 | rv = -EINVAL; |
1665 | goto out; | |
111d01fc | 1666 | } |
82d9eba6 | 1667 | |
111d01fc NB |
1668 | sb->sb_csum = calc_sb_1_csum(sb); |
1669 | rv = store_super1(st, di->fd); | |
111d01fc NB |
1670 | if (rv == 0 && (__le32_to_cpu(sb->feature_map) & 1)) |
1671 | rv = st->ss->write_bitmap(st, di->fd); | |
1672 | close(di->fd); | |
1673 | di->fd = -1; | |
4687f160 | 1674 | if (rv) |
1675 | goto error_out; | |
111d01fc | 1676 | } |
40114213 JS |
1677 | error_out: |
1678 | if (rv) | |
e7b84f9d N |
1679 | pr_err("Failed to write metadata to %s\n", |
1680 | di->devname); | |
40114213 | 1681 | out: |
82d9eba6 NB |
1682 | return rv; |
1683 | } | |
111d01fc | 1684 | #endif |
82d9eba6 | 1685 | |
64557c33 | 1686 | static int compare_super1(struct supertype *st, struct supertype *tst) |
82d9eba6 NB |
1687 | { |
1688 | /* | |
1689 | * return: | |
1690 | * 0 same, or first was empty, and second was copied | |
1691 | * 1 second had wrong number | |
1692 | * 2 wrong uuid | |
1693 | * 3 wrong other info | |
1694 | */ | |
64557c33 NB |
1695 | struct mdp_superblock_1 *first = st->sb; |
1696 | struct mdp_superblock_1 *second = tst->sb; | |
82d9eba6 NB |
1697 | |
1698 | if (second->magic != __cpu_to_le32(MD_SB_MAGIC)) | |
1699 | return 1; | |
1700 | if (second->major_version != __cpu_to_le32(1)) | |
1701 | return 1; | |
1702 | ||
1703 | if (!first) { | |
3c0bcd46 | 1704 | if (posix_memalign((void**)&first, 4096, SUPER1_SIZE) != 0) { |
e7b84f9d | 1705 | pr_err("%s could not allocate superblock\n", __func__); |
3d2c4fc7 DW |
1706 | return 1; |
1707 | } | |
41226756 | 1708 | memcpy(first, second, SUPER1_SIZE); |
64557c33 | 1709 | st->sb = first; |
82d9eba6 NB |
1710 | return 0; |
1711 | } | |
1712 | if (memcmp(first->set_uuid, second->set_uuid, 16)!= 0) | |
1713 | return 2; | |
1714 | ||
1715 | if (first->ctime != second->ctime || | |
1716 | first->level != second->level || | |
1717 | first->layout != second->layout || | |
1718 | first->size != second->size || | |
1719 | first->chunksize != second->chunksize || | |
1720 | first->raid_disks != second->raid_disks) | |
1721 | return 3; | |
1722 | return 0; | |
1723 | } | |
1724 | ||
3da92f27 | 1725 | static int load_super1(struct supertype *st, int fd, char *devname) |
82d9eba6 | 1726 | { |
5dd497ee | 1727 | unsigned long long dsize; |
82d9eba6 NB |
1728 | unsigned long long sb_offset; |
1729 | struct mdp_superblock_1 *super; | |
14263cf1 NB |
1730 | int uuid[4]; |
1731 | struct bitmap_super_s *bsb; | |
bee8ec56 | 1732 | struct misc_dev_info *misc; |
308340aa | 1733 | struct align_fd afd; |
82d9eba6 | 1734 | |
3da92f27 | 1735 | free_super1(st); |
82d9eba6 | 1736 | |
308340aa JS |
1737 | init_afd(&afd, fd); |
1738 | ||
a17a3de3 | 1739 | if (st->ss == NULL || st->minor_version == -1) { |
570c0542 | 1740 | int bestvers = -1; |
23dc1ae8 | 1741 | struct supertype tst; |
570c0542 NB |
1742 | __u64 bestctime = 0; |
1743 | /* guess... choose latest ctime */ | |
ef609477 | 1744 | memset(&tst, 0, sizeof(tst)); |
23dc1ae8 NB |
1745 | tst.ss = &super1; |
1746 | for (tst.minor_version = 0; tst.minor_version <= 2 ; tst.minor_version++) { | |
3da92f27 NB |
1747 | switch(load_super1(&tst, fd, devname)) { |
1748 | case 0: super = tst.sb; | |
570c0542 NB |
1749 | if (bestvers == -1 || |
1750 | bestctime < __le64_to_cpu(super->ctime)) { | |
23dc1ae8 | 1751 | bestvers = tst.minor_version; |
570c0542 NB |
1752 | bestctime = __le64_to_cpu(super->ctime); |
1753 | } | |
1754 | free(super); | |
3da92f27 | 1755 | tst.sb = NULL; |
570c0542 | 1756 | break; |
23dc1ae8 | 1757 | case 1: return 1; /*bad device */ |
82d9eba6 NB |
1758 | case 2: break; /* bad, try next */ |
1759 | } | |
1760 | } | |
570c0542 NB |
1761 | if (bestvers != -1) { |
1762 | int rv; | |
23dc1ae8 NB |
1763 | tst.minor_version = bestvers; |
1764 | tst.ss = &super1; | |
911cead7 | 1765 | tst.max_devs = MAX_DEVS; |
3da92f27 | 1766 | rv = load_super1(&tst, fd, devname); |
23dc1ae8 NB |
1767 | if (rv == 0) |
1768 | *st = tst; | |
570c0542 NB |
1769 | return rv; |
1770 | } | |
82d9eba6 NB |
1771 | return 2; |
1772 | } | |
beae1dfe NB |
1773 | if (!get_dev_size(fd, devname, &dsize)) |
1774 | return 1; | |
1775 | dsize >>= 9; | |
82d9eba6 | 1776 | |
5dd497ee | 1777 | if (dsize < 24) { |
82d9eba6 | 1778 | if (devname) |
e7b84f9d | 1779 | pr_err("%s is too small for md: size is %llu sectors.\n", |
5dd497ee | 1780 | devname, dsize); |
82d9eba6 NB |
1781 | return 1; |
1782 | } | |
1783 | ||
1784 | /* | |
1785 | * Calculate the position of the superblock. | |
1786 | * It is always aligned to a 4K boundary and | |
14263cf1 | 1787 | * depending on minor_version, it can be: |
82d9eba6 NB |
1788 | * 0: At least 8K, but less than 12K, from end of device |
1789 | * 1: At start of device | |
1790 | * 2: 4K from start of device. | |
1791 | */ | |
1792 | switch(st->minor_version) { | |
1793 | case 0: | |
5dd497ee | 1794 | sb_offset = dsize; |
82d9eba6 NB |
1795 | sb_offset -= 8*2; |
1796 | sb_offset &= ~(4*2-1); | |
1797 | break; | |
1798 | case 1: | |
1799 | sb_offset = 0; | |
1800 | break; | |
1801 | case 2: | |
1802 | sb_offset = 4*2; | |
1803 | break; | |
1804 | default: | |
1805 | return -EINVAL; | |
1806 | } | |
1807 | ||
82d9eba6 NB |
1808 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) { |
1809 | if (devname) | |
e7b84f9d | 1810 | pr_err("Cannot seek to superblock on %s: %s\n", |
82d9eba6 NB |
1811 | devname, strerror(errno)); |
1812 | return 1; | |
1813 | } | |
1814 | ||
3c0bcd46 | 1815 | if (posix_memalign((void**)&super, 4096, SUPER1_SIZE) != 0) { |
e7b84f9d | 1816 | pr_err("%s could not allocate superblock\n", |
3d2c4fc7 DW |
1817 | __func__); |
1818 | return 1; | |
1819 | } | |
82d9eba6 | 1820 | |
308340aa | 1821 | if (aread(&afd, super, MAX_SB_SIZE) != MAX_SB_SIZE) { |
82d9eba6 | 1822 | if (devname) |
e7b84f9d | 1823 | pr_err("Cannot read superblock on %s\n", |
82d9eba6 NB |
1824 | devname); |
1825 | free(super); | |
1826 | return 1; | |
1827 | } | |
1828 | ||
1829 | if (__le32_to_cpu(super->magic) != MD_SB_MAGIC) { | |
1830 | if (devname) | |
e7b84f9d | 1831 | pr_err("No super block found on %s (Expected magic %08x, got %08x)\n", |
82d9eba6 NB |
1832 | devname, MD_SB_MAGIC, __le32_to_cpu(super->magic)); |
1833 | free(super); | |
1834 | return 2; | |
1835 | } | |
1836 | ||
1837 | if (__le32_to_cpu(super->major_version) != 1) { | |
1838 | if (devname) | |
e7b84f9d | 1839 | pr_err("Cannot interpret superblock on %s - version is %d\n", |
82d9eba6 NB |
1840 | devname, __le32_to_cpu(super->major_version)); |
1841 | free(super); | |
1842 | return 2; | |
1843 | } | |
1844 | if (__le64_to_cpu(super->super_offset) != sb_offset) { | |
1845 | if (devname) | |
e7b84f9d | 1846 | pr_err("No superblock found on %s (super_offset is wrong)\n", |
82d9eba6 NB |
1847 | devname); |
1848 | free(super); | |
1849 | return 2; | |
1850 | } | |
64557c33 | 1851 | st->sb = super; |
14263cf1 | 1852 | |
911cead7 | 1853 | bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); |
bee8ec56 | 1854 | |
911cead7 | 1855 | misc = (struct misc_dev_info*) (((char*)super)+MAX_SB_SIZE+BM_SUPER_SIZE); |
bee8ec56 NB |
1856 | misc->device_size = dsize; |
1857 | ||
14263cf1 NB |
1858 | /* Now check on the bitmap superblock */ |
1859 | if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0) | |
1860 | return 0; | |
1861 | /* Read the bitmap superblock and make sure it looks | |
1862 | * valid. If it doesn't clear the bit. An --assemble --force | |
1863 | * should get that written out. | |
1864 | */ | |
3da92f27 | 1865 | locate_bitmap1(st, fd); |
308340aa | 1866 | if (aread(&afd, bsb, 512) != 512) |
14263cf1 NB |
1867 | goto no_bitmap; |
1868 | ||
3da92f27 | 1869 | uuid_from_super1(st, uuid); |
14263cf1 NB |
1870 | if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC || |
1871 | memcmp(bsb->uuid, uuid, 16) != 0) | |
1872 | goto no_bitmap; | |
1873 | return 0; | |
1874 | ||
1875 | no_bitmap: | |
c0c1acd6 N |
1876 | super->feature_map = __cpu_to_le32(__le32_to_cpu(super->feature_map) |
1877 | & ~MD_FEATURE_BITMAP_OFFSET); | |
82d9eba6 NB |
1878 | return 0; |
1879 | } | |
1880 | ||
1881 | ||
1882 | static struct supertype *match_metadata_desc1(char *arg) | |
1883 | { | |
503975b9 | 1884 | struct supertype *st = xcalloc(1, sizeof(*st)); |
82d9eba6 | 1885 | |
4dd2df09 | 1886 | st->container_devnm[0] = 0; |
82d9eba6 | 1887 | st->ss = &super1; |
911cead7 | 1888 | st->max_devs = MAX_DEVS; |
64557c33 | 1889 | st->sb = NULL; |
9b2a22d3 N |
1890 | /* leading zeros can be safely ignored. --detail generates them. */ |
1891 | while (*arg == '0') | |
1892 | arg++; | |
1893 | if (strcmp(arg, "1.0") == 0 || | |
1894 | strcmp(arg, "1.00") == 0) { | |
82d9eba6 NB |
1895 | st->minor_version = 0; |
1896 | return st; | |
1897 | } | |
9b2a22d3 | 1898 | if (strcmp(arg, "1.1") == 0 || |
50827504 | 1899 | strcmp(arg, "1.01") == 0 |
7d5c3964 | 1900 | ) { |
82d9eba6 NB |
1901 | st->minor_version = 1; |
1902 | return st; | |
1903 | } | |
9b2a22d3 | 1904 | if (strcmp(arg, "1.2") == 0 || |
26f467a9 | 1905 | #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */ |
ad90adb6 | 1906 | strcmp(arg, "default") == 0 || |
26f467a9 | 1907 | #endif /* DEFAULT_OLD_METADATA */ |
9b2a22d3 | 1908 | strcmp(arg, "1.02") == 0) { |
82d9eba6 NB |
1909 | st->minor_version = 2; |
1910 | return st; | |
1911 | } | |
a17a3de3 | 1912 | if (strcmp(arg, "1") == 0 || |
17f25ca6 | 1913 | strcmp(arg, "default") == 0) { |
a17a3de3 DL |
1914 | st->minor_version = -1; |
1915 | return st; | |
1916 | } | |
82d9eba6 NB |
1917 | |
1918 | free(st); | |
1919 | return NULL; | |
1920 | } | |
1921 | ||
34163fc7 NB |
1922 | /* find available size on device with this devsize, using |
1923 | * superblock type st, and reserving 'reserve' sectors for | |
1924 | * a possible bitmap | |
1925 | */ | |
cb19a251 N |
1926 | static __u64 _avail_size1(struct supertype *st, __u64 devsize, |
1927 | unsigned long long data_offset, int chunksize) | |
82d9eba6 | 1928 | { |
2a528478 | 1929 | struct mdp_superblock_1 *super = st->sb; |
387fcd59 | 1930 | int bmspace = 0; |
82d9eba6 NB |
1931 | if (devsize < 24) |
1932 | return 0; | |
1933 | ||
2a528478 N |
1934 | if (super == NULL) |
1935 | /* creating: allow suitable space for bitmap */ | |
387fcd59 | 1936 | bmspace = choose_bm_space(devsize); |
2a528478 N |
1937 | #ifndef MDASSEMBLE |
1938 | else if (__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) { | |
1939 | /* hot-add. allow for actual size of bitmap */ | |
1940 | struct bitmap_super_s *bsb; | |
911cead7 | 1941 | bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); |
387fcd59 | 1942 | bmspace = bitmap_sectors(bsb); |
2a528478 N |
1943 | } |
1944 | #endif | |
88af981f N |
1945 | /* Allow space for bad block log */ |
1946 | if (super && super->bblog_size) | |
1947 | devsize -= __le16_to_cpu(super->bblog_size); | |
1948 | else | |
1949 | devsize -= 8; | |
21e92547 | 1950 | |
387fcd59 | 1951 | |
a380e275 N |
1952 | if (st->minor_version < 0) |
1953 | /* not specified, so time to set default */ | |
1954 | st->minor_version = 2; | |
387fcd59 N |
1955 | |
1956 | if (data_offset != INVALID_SECTORS) | |
1957 | switch(st->minor_version) { | |
1958 | case 0: | |
1959 | return devsize - data_offset - 8*2; | |
1960 | case 1: | |
1961 | case 2: | |
1962 | return devsize - data_offset; | |
1963 | default: | |
1964 | return 0; | |
1965 | } | |
1966 | ||
1967 | devsize -= bmspace; | |
1968 | ||
a380e275 N |
1969 | if (super == NULL && st->minor_version > 0) { |
1970 | /* haven't committed to a size yet, so allow some | |
508a7f16 N |
1971 | * slack for space for reshape. |
1972 | * Limit slack to 128M, but aim for about 0.1% | |
a380e275 | 1973 | */ |
508a7f16 | 1974 | unsigned long long headroom = 128*1024*2; |
cb19a251 N |
1975 | while ((headroom << 10) > devsize && |
1976 | (chunksize == 0 || | |
1977 | headroom / 2 >= ((unsigned)chunksize*2)*2)) | |
508a7f16 N |
1978 | headroom >>= 1; |
1979 | devsize -= headroom; | |
a380e275 | 1980 | } |
34163fc7 NB |
1981 | switch(st->minor_version) { |
1982 | case 0: | |
1bf4e2d9 NB |
1983 | /* at end */ |
1984 | return ((devsize - 8*2 ) & ~(4*2-1)); | |
34163fc7 | 1985 | case 1: |
1bf4e2d9 NB |
1986 | /* at start, 4K for superblock and possible bitmap */ |
1987 | return devsize - 4*2; | |
34163fc7 | 1988 | case 2: |
1bf4e2d9 NB |
1989 | /* 4k from start, 4K for superblock and possible bitmap */ |
1990 | return devsize - (4+4)*2; | |
34163fc7 NB |
1991 | } |
1992 | return 0; | |
1993 | } | |
cb19a251 N |
1994 | static __u64 avail_size1(struct supertype *st, __u64 devsize, |
1995 | unsigned long long data_offset) | |
1996 | { | |
1997 | return _avail_size1(st, devsize, data_offset, 0); | |
1998 | } | |
34163fc7 | 1999 | |
1bf4e2d9 | 2000 | static int |
3da92f27 | 2001 | add_internal_bitmap1(struct supertype *st, |
199171a2 NB |
2002 | int *chunkp, int delay, int write_behind, |
2003 | unsigned long long size, | |
21e92547 | 2004 | int may_change, int major) |
34163fc7 NB |
2005 | { |
2006 | /* | |
c0c1acd6 N |
2007 | * If not may_change, then this is a 'Grow' without sysfs support for |
2008 | * bitmaps, and the bitmap must fit after the superblock at 1K offset. | |
2009 | * If may_change, then this is create or a Grow with sysfs syupport, | |
2010 | * and we can put the bitmap wherever we like. | |
34163fc7 | 2011 | * |
f9c25f1d | 2012 | * size is in sectors, chunk is in bytes !!! |
34163fc7 NB |
2013 | */ |
2014 | ||
1bf4e2d9 | 2015 | unsigned long long bits; |
199171a2 | 2016 | unsigned long long max_bits; |
34163fc7 | 2017 | unsigned long long min_chunk; |
199171a2 | 2018 | long offset; |
88af981f | 2019 | long bbl_offset, bbl_size; |
f21e18ca | 2020 | unsigned long long chunk = *chunkp; |
37dfc3d6 | 2021 | int room = 0; |
c0c1acd6 | 2022 | int creating = 0; |
3da92f27 | 2023 | struct mdp_superblock_1 *sb = st->sb; |
911cead7 | 2024 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); |
3b7e9d0c | 2025 | int uuid[4]; |
34163fc7 | 2026 | |
88af981f | 2027 | |
c0c1acd6 N |
2028 | if (__le64_to_cpu(sb->data_size) == 0) |
2029 | /* Must be creating the array, else data_size would be non-zero */ | |
2030 | creating = 1; | |
199171a2 NB |
2031 | switch(st->minor_version) { |
2032 | case 0: | |
a31128d2 N |
2033 | /* either 3K after the superblock (when hot-add), |
2034 | * or some amount of space before. | |
199171a2 | 2035 | */ |
c0c1acd6 | 2036 | if (creating) { |
199171a2 NB |
2037 | /* We are creating array, so we *know* how much room has |
2038 | * been left. | |
2039 | */ | |
2040 | offset = 0; | |
2fb749d1 | 2041 | room = choose_bm_space(__le64_to_cpu(sb->size)); |
88af981f | 2042 | bbl_size = 8; |
199171a2 NB |
2043 | } else { |
2044 | room = __le64_to_cpu(sb->super_offset) | |
2045 | - __le64_to_cpu(sb->data_offset) | |
2046 | - __le64_to_cpu(sb->data_size); | |
88af981f N |
2047 | bbl_size = __le16_to_cpu(sb->bblog_size); |
2048 | if (bbl_size < 8) | |
2049 | bbl_size = 8; | |
2050 | bbl_offset = (__s32)__le32_to_cpu(sb->bblog_offset); | |
2051 | if (bbl_size < -bbl_offset) | |
2052 | bbl_size = -bbl_offset; | |
c0c1acd6 N |
2053 | |
2054 | if (!may_change || (room < 3*2 && | |
199171a2 NB |
2055 | __le32_to_cpu(sb->max_dev) <= 384)) { |
2056 | room = 3*2; | |
2057 | offset = 1*2; | |
88af981f | 2058 | bbl_size = 0; |
199171a2 NB |
2059 | } else { |
2060 | offset = 0; /* means movable offset */ | |
2061 | } | |
2062 | } | |
2063 | break; | |
2064 | case 1: | |
2065 | case 2: /* between superblock and data */ | |
c0c1acd6 | 2066 | if (creating) { |
199171a2 | 2067 | offset = 4*2; |
2fb749d1 | 2068 | room = choose_bm_space(__le64_to_cpu(sb->size)); |
88af981f | 2069 | bbl_size = 8; |
199171a2 NB |
2070 | } else { |
2071 | room = __le64_to_cpu(sb->data_offset) | |
2072 | - __le64_to_cpu(sb->super_offset); | |
88af981f N |
2073 | bbl_size = __le16_to_cpu(sb->bblog_size); |
2074 | if (bbl_size) | |
2075 | room = __le32_to_cpu(sb->bblog_offset) + bbl_size; | |
2076 | else | |
2077 | bbl_size = 8; | |
2078 | ||
c0c1acd6 N |
2079 | if (!may_change) { |
2080 | room -= 2; /* Leave 1K for superblock */ | |
199171a2 | 2081 | offset = 2; |
88af981f | 2082 | bbl_size = 0; |
199171a2 | 2083 | } else { |
c0c1acd6 | 2084 | room -= 4*2; /* leave 4K for superblock */ |
199171a2 NB |
2085 | offset = 4*2; |
2086 | } | |
2087 | } | |
2088 | break; | |
ae491d1e NB |
2089 | default: |
2090 | return 0; | |
199171a2 | 2091 | } |
1bf4e2d9 | 2092 | |
88af981f | 2093 | room -= bbl_size; |
199171a2 NB |
2094 | if (chunk == UnSet && room > 128*2) |
2095 | /* Limit to 128K of bitmap when chunk size not requested */ | |
2096 | room = 128*2; | |
1bf4e2d9 | 2097 | |
199171a2 | 2098 | max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8; |
34163fc7 NB |
2099 | |
2100 | min_chunk = 4096; /* sub-page chunks don't work yet.. */ | |
f9c25f1d | 2101 | bits = (size*512)/min_chunk +1; |
34163fc7 NB |
2102 | while (bits > max_bits) { |
2103 | min_chunk *= 2; | |
2104 | bits = (bits+1)/2; | |
2105 | } | |
b8ab2a50 N |
2106 | if (chunk == UnSet) { |
2107 | /* For practical purpose, 64Meg is a good | |
2108 | * default chunk size for internal bitmaps. | |
2109 | */ | |
34163fc7 | 2110 | chunk = min_chunk; |
b8ab2a50 N |
2111 | if (chunk < 64*1024*1024) |
2112 | chunk = 64*1024*1024; | |
2113 | } else if (chunk < min_chunk) | |
34163fc7 | 2114 | return 0; /* chunk size too small */ |
21e92547 NB |
2115 | if (chunk == 0) /* rounding problem */ |
2116 | return 0; | |
34163fc7 | 2117 | |
199171a2 | 2118 | if (offset == 0) { |
a31128d2 N |
2119 | /* start bitmap on a 4K boundary with enough space for |
2120 | * the bitmap | |
2121 | */ | |
199171a2 | 2122 | bits = (size*512) / chunk + 1; |
a31128d2 N |
2123 | room = ((bits+7)/8 + sizeof(bitmap_super_t) +4095)/4096; |
2124 | room *= 8; /* convert 4K blocks to sectors */ | |
88af981f | 2125 | offset = -room - bbl_size; |
199171a2 NB |
2126 | } |
2127 | ||
34a13953 | 2128 | sb->bitmap_offset = (int32_t)__cpu_to_le32(offset); |
34163fc7 | 2129 | |
c0c1acd6 N |
2130 | sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) |
2131 | | MD_FEATURE_BITMAP_OFFSET); | |
29e766a5 | 2132 | memset(bms, 0, sizeof(*bms)); |
34163fc7 | 2133 | bms->magic = __cpu_to_le32(BITMAP_MAGIC); |
dcec9ee5 | 2134 | bms->version = __cpu_to_le32(major); |
3b7e9d0c LB |
2135 | uuid_from_super1(st, uuid); |
2136 | memcpy(bms->uuid, uuid, 16); | |
34163fc7 NB |
2137 | bms->chunksize = __cpu_to_le32(chunk); |
2138 | bms->daemon_sleep = __cpu_to_le32(delay); | |
f9c25f1d | 2139 | bms->sync_size = __cpu_to_le64(size); |
34163fc7 NB |
2140 | bms->write_behind = __cpu_to_le32(write_behind); |
2141 | ||
199171a2 | 2142 | *chunkp = chunk; |
34163fc7 NB |
2143 | return 1; |
2144 | } | |
2145 | ||
3da92f27 | 2146 | static void locate_bitmap1(struct supertype *st, int fd) |
34163fc7 | 2147 | { |
34163fc7 | 2148 | unsigned long long offset; |
1e0d770c NB |
2149 | struct mdp_superblock_1 *sb; |
2150 | int mustfree = 0; | |
34163fc7 | 2151 | |
3da92f27 NB |
2152 | if (!st->sb) { |
2153 | if (st->ss->load_super(st, fd, NULL)) | |
f6d75de8 | 2154 | return; /* no error I hope... */ |
1e0d770c NB |
2155 | mustfree = 1; |
2156 | } | |
3da92f27 | 2157 | sb = st->sb; |
e478dc86 | 2158 | |
1bf4e2d9 | 2159 | offset = __le64_to_cpu(sb->super_offset); |
68754bd1 | 2160 | offset += (int32_t) __le32_to_cpu(sb->bitmap_offset); |
1e0d770c | 2161 | if (mustfree) |
f6d75de8 | 2162 | free(sb); |
1bf4e2d9 | 2163 | lseek64(fd, offset<<9, 0); |
34163fc7 NB |
2164 | } |
2165 | ||
3da92f27 | 2166 | static int write_bitmap1(struct supertype *st, int fd) |
34163fc7 | 2167 | { |
3da92f27 | 2168 | struct mdp_superblock_1 *sb = st->sb; |
911cead7 | 2169 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+MAX_SB_SIZE); |
34163fc7 | 2170 | int rv = 0; |
d669228f | 2171 | void *buf; |
34163fc7 | 2172 | int towrite, n; |
308340aa JS |
2173 | struct align_fd afd; |
2174 | ||
2175 | init_afd(&afd, fd); | |
34163fc7 | 2176 | |
3da92f27 | 2177 | locate_bitmap1(st, fd); |
34163fc7 | 2178 | |
d669228f JS |
2179 | if (posix_memalign(&buf, 4096, 4096)) |
2180 | return -ENOMEM; | |
2181 | ||
6416d527 | 2182 | memset(buf, 0xff, 4096); |
b2bfdfa0 | 2183 | memcpy(buf, (char *)bms, sizeof(bitmap_super_t)); |
6416d527 | 2184 | |
1bf4e2d9 NB |
2185 | towrite = __le64_to_cpu(bms->sync_size) / (__le32_to_cpu(bms->chunksize)>>9); |
2186 | towrite = (towrite+7) >> 3; /* bits to bytes */ | |
6416d527 NB |
2187 | towrite += sizeof(bitmap_super_t); |
2188 | towrite = ROUND_UP(towrite, 512); | |
34163fc7 NB |
2189 | while (towrite > 0) { |
2190 | n = towrite; | |
6416d527 NB |
2191 | if (n > 4096) |
2192 | n = 4096; | |
308340aa | 2193 | n = awrite(&afd, buf, n); |
34163fc7 NB |
2194 | if (n > 0) |
2195 | towrite -= n; | |
2196 | else | |
2197 | break; | |
6416d527 | 2198 | memset(buf, 0xff, 4096); |
34163fc7 NB |
2199 | } |
2200 | fsync(fd); | |
2201 | if (towrite) | |
2202 | rv = -2; | |
2203 | ||
d669228f | 2204 | free(buf); |
34163fc7 | 2205 | return rv; |
82d9eba6 NB |
2206 | } |
2207 | ||
3da92f27 | 2208 | static void free_super1(struct supertype *st) |
df37ffc0 | 2209 | { |
3da92f27 NB |
2210 | if (st->sb) |
2211 | free(st->sb); | |
1cc7f4fe N |
2212 | while (st->info) { |
2213 | struct devinfo *di = st->info; | |
2214 | st->info = di->next; | |
2215 | if (di->fd >= 0) | |
2216 | close(di->fd); | |
2217 | free(di); | |
2218 | } | |
3da92f27 | 2219 | st->sb = NULL; |
df37ffc0 NB |
2220 | } |
2221 | ||
0e600426 | 2222 | #ifndef MDASSEMBLE |
17f25ca6 NB |
2223 | static int validate_geometry1(struct supertype *st, int level, |
2224 | int layout, int raiddisks, | |
c21e737b | 2225 | int *chunk, unsigned long long size, |
af4348dd | 2226 | unsigned long long data_offset, |
2c514b71 NB |
2227 | char *subdev, unsigned long long *freesize, |
2228 | int verbose) | |
17f25ca6 NB |
2229 | { |
2230 | unsigned long long ldsize; | |
2231 | int fd; | |
2232 | ||
b42f577a N |
2233 | if (level == LEVEL_CONTAINER) { |
2234 | if (verbose) | |
e7b84f9d | 2235 | pr_err("1.x metadata does not support containers\n"); |
17f25ca6 | 2236 | return 0; |
b42f577a | 2237 | } |
bb7295f1 N |
2238 | if (chunk && *chunk == UnSet) |
2239 | *chunk = DEFAULT_CHUNK; | |
2240 | ||
17f25ca6 NB |
2241 | if (!subdev) |
2242 | return 1; | |
2243 | ||
2244 | fd = open(subdev, O_RDONLY|O_EXCL, 0); | |
2245 | if (fd < 0) { | |
2c514b71 | 2246 | if (verbose) |
e7b84f9d | 2247 | pr_err("super1.x cannot open %s: %s\n", |
2c514b71 | 2248 | subdev, strerror(errno)); |
17f25ca6 NB |
2249 | return 0; |
2250 | } | |
2c514b71 | 2251 | |
17f25ca6 NB |
2252 | if (!get_dev_size(fd, subdev, &ldsize)) { |
2253 | close(fd); | |
2254 | return 0; | |
2255 | } | |
2256 | close(fd); | |
2257 | ||
cb19a251 | 2258 | *freesize = _avail_size1(st, ldsize >> 9, data_offset, *chunk); |
17f25ca6 NB |
2259 | return 1; |
2260 | } | |
0e600426 | 2261 | #endif /* MDASSEMBLE */ |
17f25ca6 | 2262 | |
82d9eba6 | 2263 | struct superswitch super1 = { |
c7654afc | 2264 | #ifndef MDASSEMBLE |
82d9eba6 NB |
2265 | .examine_super = examine_super1, |
2266 | .brief_examine_super = brief_examine_super1, | |
0d726f17 | 2267 | .export_examine_super = export_examine_super1, |
82d9eba6 NB |
2268 | .detail_super = detail_super1, |
2269 | .brief_detail_super = brief_detail_super1, | |
0d726f17 | 2270 | .export_detail_super = export_detail_super1, |
111d01fc | 2271 | .write_init_super = write_init_super1, |
0e600426 N |
2272 | .validate_geometry = validate_geometry1, |
2273 | .add_to_super = add_to_super1, | |
6d388a88 | 2274 | .examine_badblocks = examine_badblocks_super1, |
74db60b0 | 2275 | .copy_metadata = copy_metadata1, |
c7654afc | 2276 | #endif |
83b6208e | 2277 | .match_home = match_home1, |
82d9eba6 NB |
2278 | .uuid_from_super = uuid_from_super1, |
2279 | .getinfo_super = getinfo_super1, | |
00bbdbda | 2280 | .container_content = container_content1, |
82d9eba6 | 2281 | .update_super = update_super1, |
82d9eba6 | 2282 | .init_super = init_super1, |
82d9eba6 | 2283 | .store_super = store_super1, |
82d9eba6 NB |
2284 | .compare_super = compare_super1, |
2285 | .load_super = load_super1, | |
2286 | .match_metadata_desc = match_metadata_desc1, | |
2287 | .avail_size = avail_size1, | |
34163fc7 NB |
2288 | .add_internal_bitmap = add_internal_bitmap1, |
2289 | .locate_bitmap = locate_bitmap1, | |
2290 | .write_bitmap = write_bitmap1, | |
df37ffc0 | 2291 | .free_super = free_super1, |
f277ce36 NB |
2292 | #if __BYTE_ORDER == BIG_ENDIAN |
2293 | .swapuuid = 0, | |
2294 | #else | |
2295 | .swapuuid = 1, | |
2296 | #endif | |
31015d57 | 2297 | .name = "1.x", |
82d9eba6 | 2298 | }; |