]>
Commit | Line | Data |
---|---|---|
82d9eba6 NB |
1 | /* |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
7071320a | 4 | * Copyright (C) 2001-2016 Neil Brown <neilb@suse.com> |
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; | |
e97a7cd0 AP |
51 | union { |
52 | __u32 bitmap_offset; /* sectors after start of superblock that bitmap starts | |
53 | * NOTE: signed, so bitmap can be before superblock | |
54 | * only meaningful of feature_map[0] is set. | |
55 | */ | |
56 | ||
57 | /* only meaningful when feature_map[MD_FEATURE_PPL] is set */ | |
58 | struct { | |
59 | __s16 offset; /* sectors from start of superblock that ppl starts */ | |
60 | __u16 size; /* ppl size in sectors */ | |
61 | } ppl; | |
62 | }; | |
b674b5b8 NB |
63 | |
64 | /* These are only valid with feature bit '4' */ | |
b674b5b8 | 65 | __u32 new_level; /* new level we are reshaping to */ |
ac957baa | 66 | __u64 reshape_position; /* next address in array-space for reshape */ |
b674b5b8 NB |
67 | __u32 delta_disks; /* change in number of raid_disks */ |
68 | __u32 new_layout; /* new layout */ | |
a2836f12 | 69 | __u32 new_chunk; /* new chunk size (sectors) */ |
8fe1c44f N |
70 | __u32 new_offset; /* signed number to add to data_offset in new |
71 | * layout. 0 == no-change. This can be | |
72 | * different on each device in the array. | |
73 | */ | |
82d9eba6 NB |
74 | |
75 | /* constant this-device information - 64 bytes */ | |
76 | __u64 data_offset; /* sector start of data, often 0 */ | |
77 | __u64 data_size; /* sectors in this device that can be used for data */ | |
78 | __u64 super_offset; /* sector start of this superblock */ | |
cc1799c3 SL |
79 | union { |
80 | __u64 recovery_offset;/* sectors before this offset (from data_offset) have been recovered */ | |
81 | __u64 journal_tail;/* journal tail of journal device (from data_offset) */ | |
82 | }; | |
82d9eba6 NB |
83 | __u32 dev_number; /* permanent identifier of this device - not role in raid */ |
84 | __u32 cnt_corrected_read; /* number of read errors that were corrected by re-writing */ | |
85 | __u8 device_uuid[16]; /* user-space setable, ignored by kernel */ | |
5d500228 | 86 | __u8 devflags; /* per-device flags. Only one defined...*/ |
dfd4d8ee | 87 | #define WriteMostly1 1 /* mask for writemostly flag in above */ |
71574efb | 88 | #define FailFast1 2 /* Device should get FailFast requests */ |
bf95d0f3 N |
89 | /* bad block log. If there are any bad blocks the feature flag is set. |
90 | * if offset and size are non-zero, that space is reserved and available. | |
91 | */ | |
5c1dec3e ZL |
92 | __u8 bblog_shift; /* shift from sectors to block size for badblock list */ |
93 | __u16 bblog_size; /* number of sectors reserved for badblock list */ | |
bf95d0f3 | 94 | __u32 bblog_offset; /* sector offset from superblock to bblog, signed */ |
82d9eba6 NB |
95 | |
96 | /* array state information - 64 bytes */ | |
5c1dec3e | 97 | __u64 utime; /* 40 bits second, 24 bits microseconds */ |
82d9eba6 NB |
98 | __u64 events; /* incremented when superblock updated */ |
99 | __u64 resync_offset; /* data before this offset (from data_offset) known to be in sync */ | |
4180aa4d N |
100 | __u32 sb_csum; /* checksum upto dev_roles[max_dev] */ |
101 | __u32 max_dev; /* size of dev_roles[] array to consider */ | |
82d9eba6 NB |
102 | __u8 pad3[64-32]; /* set to 0 when writing */ |
103 | ||
104 | /* device state information. Indexed by dev_number. | |
105 | * 2 bytes per device | |
106 | * Note there are no per-device state flags. State information is rolled | |
107 | * into the 'roles' value. If a device is spare or faulty, then it doesn't | |
108 | * have a meaningful role. | |
109 | */ | |
110 | __u16 dev_roles[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */ | |
111 | }; | |
112 | ||
911cead7 N |
113 | #define MAX_SB_SIZE 4096 |
114 | /* bitmap super size is 256, but we round up to a sector for alignment */ | |
115 | #define BM_SUPER_SIZE 512 | |
116 | #define MAX_DEVS ((int)(MAX_SB_SIZE - sizeof(struct mdp_superblock_1)) / 2) | |
41226756 JS |
117 | #define SUPER1_SIZE (MAX_SB_SIZE + BM_SUPER_SIZE \ |
118 | + sizeof(struct misc_dev_info)) | |
911cead7 | 119 | |
bee8ec56 NB |
120 | struct misc_dev_info { |
121 | __u64 device_size; | |
122 | }; | |
123 | ||
fa601c2e PB |
124 | #define MULTIPLE_PPL_AREA_SIZE_SUPER1 (1024 * 1024) /* Size of the whole |
125 | * mutliple PPL area | |
126 | */ | |
d2cd3ffc NB |
127 | /* feature_map bits */ |
128 | #define MD_FEATURE_BITMAP_OFFSET 1 | |
129 | #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and | |
130 | * must be honoured | |
131 | */ | |
b674b5b8 | 132 | #define MD_FEATURE_RESHAPE_ACTIVE 4 |
bf95d0f3 | 133 | #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */ |
8fe1c44f N |
134 | #define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an |
135 | * active device with same 'role'. | |
136 | * 'recovery_offset' is also set. | |
137 | */ | |
138 | #define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number | |
139 | * of devices, but is going | |
140 | * backwards anyway. | |
141 | */ | |
142 | #define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */ | |
6d9c7c25 | 143 | #define MD_FEATURE_BITMAP_VERSIONED 256 /* bitmap version number checked properly */ |
ed94976d | 144 | #define MD_FEATURE_JOURNAL 512 /* support write journal */ |
e97a7cd0 | 145 | #define MD_FEATURE_PPL 1024 /* support PPL */ |
fa601c2e | 146 | #define MD_FEATURE_MUTLIPLE_PPLS 2048 /* support for multiple PPLs */ |
8fe1c44f N |
147 | #define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \ |
148 | |MD_FEATURE_RECOVERY_OFFSET \ | |
149 | |MD_FEATURE_RESHAPE_ACTIVE \ | |
150 | |MD_FEATURE_BAD_BLOCKS \ | |
151 | |MD_FEATURE_REPLACEMENT \ | |
152 | |MD_FEATURE_RESHAPE_BACKWARDS \ | |
153 | |MD_FEATURE_NEW_OFFSET \ | |
6d9c7c25 | 154 | |MD_FEATURE_BITMAP_VERSIONED \ |
ed94976d | 155 | |MD_FEATURE_JOURNAL \ |
e97a7cd0 | 156 | |MD_FEATURE_PPL \ |
fa601c2e | 157 | |MD_FEATURE_MULTIPLE_PPLS \ |
8fe1c44f | 158 | ) |
69a48116 SL |
159 | |
160 | static int role_from_sb(struct mdp_superblock_1 *sb) | |
161 | { | |
162 | unsigned int d; | |
163 | int role; | |
164 | ||
165 | d = __le32_to_cpu(sb->dev_number); | |
166 | if (d < __le32_to_cpu(sb->max_dev)) | |
167 | role = __le16_to_cpu(sb->dev_roles[d]); | |
168 | else | |
169 | role = MD_DISK_ROLE_SPARE; | |
170 | return role; | |
171 | } | |
d2cd3ffc | 172 | |
7e6e839a GJ |
173 | /* return how many bytes are needed for bitmap, for cluster-md each node |
174 | * should have it's own bitmap */ | |
175 | static unsigned int calc_bitmap_size(bitmap_super_t *bms, unsigned int boundary) | |
176 | { | |
177 | unsigned long long bits, bytes; | |
178 | ||
561ad559 AP |
179 | bits = bitmap_bits(__le64_to_cpu(bms->sync_size), |
180 | __le32_to_cpu(bms->chunksize)); | |
7e6e839a GJ |
181 | bytes = (bits+7) >> 3; |
182 | bytes += sizeof(bitmap_super_t); | |
183 | bytes = ROUND_UP(bytes, boundary); | |
184 | ||
185 | return bytes; | |
186 | } | |
187 | ||
82d9eba6 NB |
188 | static unsigned int calc_sb_1_csum(struct mdp_superblock_1 * sb) |
189 | { | |
190 | unsigned int disk_csum, csum; | |
191 | unsigned long long newcsum; | |
192 | int size = sizeof(*sb) + __le32_to_cpu(sb->max_dev)*2; | |
193 | unsigned int *isuper = (unsigned int*)sb; | |
82d9eba6 NB |
194 | |
195 | /* make sure I can count... */ | |
196 | if (offsetof(struct mdp_superblock_1,data_offset) != 128 || | |
197 | offsetof(struct mdp_superblock_1, utime) != 192 || | |
198 | sizeof(struct mdp_superblock_1) != 256) { | |
199 | fprintf(stderr, "WARNING - superblock isn't sized correctly\n"); | |
200 | } | |
201 | ||
202 | disk_csum = sb->sb_csum; | |
203 | sb->sb_csum = 0; | |
204 | newcsum = 0; | |
e4c72d1d | 205 | for (; size>=4; size -= 4 ) { |
722966c6 PC |
206 | newcsum += __le32_to_cpu(*isuper); |
207 | isuper++; | |
208 | } | |
82d9eba6 NB |
209 | |
210 | if (size == 2) | |
211 | newcsum += __le16_to_cpu(*(unsigned short*) isuper); | |
212 | ||
213 | csum = (newcsum & 0xffffffff) + (newcsum >> 32); | |
214 | sb->sb_csum = disk_csum; | |
83205b64 | 215 | return __cpu_to_le32(csum); |
82d9eba6 NB |
216 | } |
217 | ||
308340aa JS |
218 | /* |
219 | * Information related to file descriptor used for aligned reads/writes. | |
220 | * Cache the block size. | |
221 | */ | |
222 | struct align_fd { | |
223 | int fd; | |
224 | int blk_sz; | |
225 | }; | |
226 | ||
227 | static void init_afd(struct align_fd *afd, int fd) | |
228 | { | |
229 | afd->fd = fd; | |
32971509 | 230 | if (!get_dev_sector_size(afd->fd, NULL, (unsigned int *)&afd->blk_sz)) |
308340aa JS |
231 | afd->blk_sz = 512; |
232 | } | |
233 | ||
5ea022a1 | 234 | static char abuf[4096+4096]; |
46a533a9 | 235 | |
308340aa | 236 | static int aread(struct align_fd *afd, void *buf, int len) |
5ea022a1 N |
237 | { |
238 | /* aligned read. | |
239 | * On devices with a 4K sector size, we need to read | |
240 | * the full sector and copy relevant bits into | |
241 | * the buffer | |
242 | */ | |
6ef89052 | 243 | int bsize, iosize; |
5ea022a1 N |
244 | char *b; |
245 | int n; | |
6ef89052 | 246 | |
308340aa JS |
247 | bsize = afd->blk_sz; |
248 | ||
249 | if (!bsize || bsize > 4096 || len > 4096) { | |
250 | if (!bsize) | |
7a862a02 | 251 | fprintf(stderr, "WARNING - aread() called with invalid block size\n"); |
5ea022a1 | 252 | return -1; |
308340aa | 253 | } |
0a2f1894 | 254 | b = ROUND_UP_PTR((char *)abuf, 4096); |
5ea022a1 | 255 | |
6ef89052 N |
256 | for (iosize = 0; iosize < len; iosize += bsize) |
257 | ; | |
308340aa | 258 | n = read(afd->fd, b, iosize); |
5ea022a1 N |
259 | if (n <= 0) |
260 | return n; | |
308340aa | 261 | lseek(afd->fd, len - n, 1); |
5ea022a1 N |
262 | if (n > len) |
263 | n = len; | |
264 | memcpy(buf, b, n); | |
265 | return n; | |
266 | } | |
267 | ||
308340aa | 268 | static int awrite(struct align_fd *afd, void *buf, int len) |
5ea022a1 N |
269 | { |
270 | /* aligned write. | |
271 | * On devices with a 4K sector size, we need to write | |
272 | * the full sector. We pre-read if the sector is larger | |
273 | * than the write. | |
274 | * The address must be sector-aligned. | |
275 | */ | |
6ef89052 | 276 | int bsize, iosize; |
5ea022a1 N |
277 | char *b; |
278 | int n; | |
308340aa JS |
279 | |
280 | bsize = afd->blk_sz; | |
281 | if (!bsize || bsize > 4096 || len > 4096) { | |
282 | if (!bsize) | |
7a862a02 | 283 | fprintf(stderr, "WARNING - awrite() called with invalid block size\n"); |
5ea022a1 | 284 | return -1; |
308340aa | 285 | } |
0a2f1894 | 286 | b = ROUND_UP_PTR((char *)abuf, 4096); |
5ea022a1 | 287 | |
6ef89052 N |
288 | for (iosize = 0; iosize < len ; iosize += bsize) |
289 | ; | |
290 | ||
291 | if (len != iosize) { | |
308340aa | 292 | n = read(afd->fd, b, iosize); |
6ef89052 N |
293 | if (n <= 0) |
294 | return n; | |
308340aa | 295 | lseek(afd->fd, -n, 1); |
6ef89052 N |
296 | } |
297 | ||
5ea022a1 | 298 | memcpy(b, buf, len); |
308340aa | 299 | n = write(afd->fd, b, iosize); |
5ea022a1 N |
300 | if (n <= 0) |
301 | return n; | |
308340aa | 302 | lseek(afd->fd, len - n, 1); |
5ea022a1 N |
303 | return len; |
304 | } | |
305 | ||
fa601c2e PB |
306 | static inline unsigned int md_feature_any_ppl_on(__u32 feature_map) |
307 | { | |
308 | return ((__cpu_to_le32(feature_map) & | |
309 | (MD_FEATURE_PPL | MD_FEATURE_MUTLIPLE_PPLS))); | |
310 | } | |
311 | ||
e97a7cd0 AP |
312 | static inline unsigned int choose_ppl_space(int chunk) |
313 | { | |
314 | return (PPL_HEADER_SIZE >> 9) + (chunk > 128*2 ? chunk : 128*2); | |
315 | } | |
316 | ||
3da92f27 | 317 | static void examine_super1(struct supertype *st, char *homehost) |
82d9eba6 | 318 | { |
3da92f27 | 319 | struct mdp_superblock_1 *sb = st->sb; |
0aa2f15b | 320 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+MAX_SB_SIZE); |
82d9eba6 | 321 | time_t atime; |
f21e18ca | 322 | unsigned int d; |
e8e1c176 | 323 | int role; |
4180aa4d | 324 | int delta_extra = 0; |
82d9eba6 NB |
325 | int i; |
326 | char *c; | |
a1cbd7d0 | 327 | int l = homehost ? strlen(homehost) : 0; |
83205b64 | 328 | int layout; |
a17a3de3 | 329 | unsigned long long sb_offset; |
5e1863d4 | 330 | struct mdinfo info; |
82d9eba6 NB |
331 | |
332 | printf(" Magic : %08x\n", __le32_to_cpu(sb->magic)); | |
a17a3de3 DL |
333 | printf(" Version : 1"); |
334 | sb_offset = __le64_to_cpu(sb->super_offset); | |
335 | if (sb_offset <= 4) | |
336 | printf(".1\n"); | |
337 | else if (sb_offset <= 8) | |
338 | printf(".2\n"); | |
339 | else | |
340 | printf(".0\n"); | |
91eedefc | 341 | printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb->feature_map)); |
82d9eba6 NB |
342 | printf(" Array UUID : "); |
343 | for (i=0; i<16; i++) { | |
82d9eba6 | 344 | if ((i&3)==0 && i != 0) printf(":"); |
34163fc7 | 345 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 NB |
346 | } |
347 | printf("\n"); | |
a1cbd7d0 NB |
348 | printf(" Name : %.32s", sb->set_name); |
349 | if (l > 0 && l < 32 && | |
350 | sb->set_name[l] == ':' && | |
351 | strncmp(sb->set_name, homehost, l) == 0) | |
352 | printf(" (local to host %s)", homehost); | |
353 | printf("\n"); | |
46a533a9 JS |
354 | if (bms->nodes > 0 && |
355 | (__le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET)) | |
2cf42394 | 356 | printf(" Cluster Name : %-64s\n", bms->cluster_name); |
82d9eba6 NB |
357 | atime = __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL; |
358 | printf(" Creation Time : %.24s\n", ctime(&atime)); | |
359 | c=map_num(pers, __le32_to_cpu(sb->level)); | |
360 | printf(" Raid Level : %s\n", c?c:"-unknown-"); | |
361 | printf(" Raid Devices : %d\n", __le32_to_cpu(sb->raid_disks)); | |
362 | printf("\n"); | |
5cda0964 | 363 | printf(" Avail Dev Size : %llu%s\n", |
2fb749d1 NB |
364 | (unsigned long long)__le64_to_cpu(sb->data_size), |
365 | human_size(__le64_to_cpu(sb->data_size)<<9)); | |
c43f7d91 | 366 | if (__le32_to_cpu(sb->level) > 0) { |
d4633e06 | 367 | int ddsks = 0, ddsks_denom = 1; |
b674b5b8 NB |
368 | switch(__le32_to_cpu(sb->level)) { |
369 | case 1: ddsks=1;break; | |
370 | case 4: | |
83205b64 NB |
371 | case 5: ddsks = __le32_to_cpu(sb->raid_disks)-1; break; |
372 | case 6: ddsks = __le32_to_cpu(sb->raid_disks)-2; break; | |
373 | case 10: | |
374 | layout = __le32_to_cpu(sb->layout); | |
d4633e06 N |
375 | ddsks = __le32_to_cpu(sb->raid_disks); |
376 | ddsks_denom = (layout&255) * ((layout>>8)&255); | |
b674b5b8 | 377 | } |
d4633e06 N |
378 | if (ddsks) { |
379 | long long asize = __le64_to_cpu(sb->size); | |
380 | asize = (asize << 9) * ddsks / ddsks_denom; | |
83205b64 | 381 | printf(" Array Size : %llu%s\n", |
d4633e06 N |
382 | asize >> 10, human_size(asize)); |
383 | } | |
b674b5b8 | 384 | if (sb->size != sb->data_size) |
5cda0964 | 385 | printf(" Used Dev Size : %llu%s\n", |
83205b64 NB |
386 | (unsigned long long)__le64_to_cpu(sb->size), |
387 | human_size(__le64_to_cpu(sb->size)<<9)); | |
b674b5b8 | 388 | } |
82d9eba6 | 389 | if (sb->data_offset) |
199171a2 NB |
390 | printf(" Data Offset : %llu sectors\n", |
391 | (unsigned long long)__le64_to_cpu(sb->data_offset)); | |
8772113a N |
392 | if (sb->new_offset && |
393 | (__le32_to_cpu(sb->feature_map) & MD_FEATURE_NEW_OFFSET)) { | |
8fe1c44f N |
394 | unsigned long long offset = __le64_to_cpu(sb->data_offset); |
395 | offset += (signed)(int32_t)__le32_to_cpu(sb->new_offset); | |
396 | printf(" New Offset : %llu sectors\n", offset); | |
397 | } | |
199171a2 NB |
398 | printf(" Super Offset : %llu sectors\n", |
399 | (unsigned long long)__le64_to_cpu(sb->super_offset)); | |
d2cd3ffc | 400 | if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_RECOVERY_OFFSET) |
46a533a9 JS |
401 | printf("Recovery Offset : %llu sectors\n", |
402 | (unsigned long long)__le64_to_cpu(sb->recovery_offset)); | |
5e1863d4 N |
403 | |
404 | st->ss->getinfo_super(st, &info, NULL); | |
405 | if (info.space_after != 1 && | |
406 | !(__le32_to_cpu(sb->feature_map) & MD_FEATURE_NEW_OFFSET)) | |
407 | printf(" Unused Space : before=%llu sectors, after=%llu sectors\n", | |
408 | info.space_before, info.space_after); | |
409 | ||
46a533a9 JS |
410 | printf(" State : %s\n", |
411 | (__le64_to_cpu(sb->resync_offset)+1)? "active":"clean"); | |
82d9eba6 NB |
412 | printf(" Device UUID : "); |
413 | for (i=0; i<16; i++) { | |
46a533a9 JS |
414 | if ((i&3)==0 && i != 0) |
415 | printf(":"); | |
34163fc7 | 416 | printf("%02x", sb->device_uuid[i]); |
82d9eba6 NB |
417 | } |
418 | printf("\n"); | |
b674b5b8 | 419 | printf("\n"); |
91eedefc NB |
420 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { |
421 | printf("Internal Bitmap : %ld sectors from superblock\n", | |
68754bd1 | 422 | (long)(int32_t)__le32_to_cpu(sb->bitmap_offset)); |
fa601c2e | 423 | } else if (md_feature_any_ppl_on(sb->feature_map)) { |
e97a7cd0 AP |
424 | printf(" PPL : %u sectors at offset %d sectors from superblock\n", |
425 | __le16_to_cpu(sb->ppl.size), | |
426 | __le16_to_cpu(sb->ppl.offset)); | |
8fac0577 | 427 | } |
bf95d0f3 | 428 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE)) { |
46a533a9 JS |
429 | printf(" Reshape pos'n : %llu%s\n", (unsigned long long) |
430 | __le64_to_cpu(sb->reshape_position)/2, | |
b674b5b8 NB |
431 | human_size(__le64_to_cpu(sb->reshape_position)<<9)); |
432 | if (__le32_to_cpu(sb->delta_disks)) { | |
46a533a9 JS |
433 | printf(" Delta Devices : %d", |
434 | __le32_to_cpu(sb->delta_disks)); | |
4180aa4d | 435 | printf(" (%d->%d)\n", |
46a533a9 JS |
436 | __le32_to_cpu(sb->raid_disks) - |
437 | __le32_to_cpu(sb->delta_disks), | |
4180aa4d N |
438 | __le32_to_cpu(sb->raid_disks)); |
439 | if ((int)__le32_to_cpu(sb->delta_disks) < 0) | |
440 | delta_extra = -__le32_to_cpu(sb->delta_disks); | |
b674b5b8 NB |
441 | } |
442 | if (__le32_to_cpu(sb->new_level) != __le32_to_cpu(sb->level)) { | |
443 | c = map_num(pers, __le32_to_cpu(sb->new_level)); | |
444 | printf(" New Level : %s\n", c?c:"-unknown-"); | |
445 | } | |
46a533a9 JS |
446 | if (__le32_to_cpu(sb->new_layout) != |
447 | __le32_to_cpu(sb->layout)) { | |
b674b5b8 | 448 | if (__le32_to_cpu(sb->level) == 5) { |
46a533a9 JS |
449 | c = map_num(r5layout, |
450 | __le32_to_cpu(sb->new_layout)); | |
b674b5b8 NB |
451 | printf(" New Layout : %s\n", c?c:"-unknown-"); |
452 | } | |
fe77a154 | 453 | if (__le32_to_cpu(sb->level) == 6) { |
46a533a9 JS |
454 | c = map_num(r6layout, |
455 | __le32_to_cpu(sb->new_layout)); | |
fe77a154 N |
456 | printf(" New Layout : %s\n", c?c:"-unknown-"); |
457 | } | |
b674b5b8 | 458 | if (__le32_to_cpu(sb->level) == 10) { |
e4965ef8 N |
459 | printf(" New Layout :"); |
460 | print_r10_layout(__le32_to_cpu(sb->new_layout)); | |
461 | printf("\n"); | |
b674b5b8 NB |
462 | } |
463 | } | |
46a533a9 JS |
464 | if (__le32_to_cpu(sb->new_chunk) != |
465 | __le32_to_cpu(sb->chunksize)) | |
466 | printf(" New Chunksize : %dK\n", | |
467 | __le32_to_cpu(sb->new_chunk)/2); | |
b674b5b8 NB |
468 | printf("\n"); |
469 | } | |
dfd4d8ee | 470 | if (sb->devflags) { |
62184891 | 471 | printf(" Flags :"); |
34163fc7 | 472 | if (sb->devflags & WriteMostly1) |
dfd4d8ee | 473 | printf(" write-mostly"); |
71574efb N |
474 | if (sb->devflags & FailFast1) |
475 | printf(" failfast"); | |
dfd4d8ee NB |
476 | printf("\n"); |
477 | } | |
82d9eba6 NB |
478 | |
479 | atime = __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL; | |
480 | printf(" Update Time : %.24s\n", ctime(&atime)); | |
481 | ||
bf95d0f3 N |
482 | if (sb->bblog_size && sb->bblog_offset) { |
483 | printf(" Bad Block Log : %d entries available at offset %ld sectors", | |
484 | __le16_to_cpu(sb->bblog_size)*512/8, | |
d6e4b44f | 485 | (long)(int32_t)__le32_to_cpu(sb->bblog_offset)); |
46a533a9 | 486 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) |
bf95d0f3 N |
487 | printf(" - bad blocks present."); |
488 | printf("\n"); | |
489 | } | |
490 | ||
82d9eba6 | 491 | if (calc_sb_1_csum(sb) == sb->sb_csum) |
46a533a9 JS |
492 | printf(" Checksum : %x - correct\n", |
493 | __le32_to_cpu(sb->sb_csum)); | |
82d9eba6 | 494 | else |
46a533a9 JS |
495 | printf(" Checksum : %x - expected %x\n", |
496 | __le32_to_cpu(sb->sb_csum), | |
82d9eba6 | 497 | __le32_to_cpu(calc_sb_1_csum(sb))); |
46a533a9 JS |
498 | printf(" Events : %llu\n", |
499 | (unsigned long long)__le64_to_cpu(sb->events)); | |
82d9eba6 NB |
500 | printf("\n"); |
501 | if (__le32_to_cpu(sb->level) == 5) { | |
502 | c = map_num(r5layout, __le32_to_cpu(sb->layout)); | |
503 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
504 | } | |
fe77a154 N |
505 | if (__le32_to_cpu(sb->level) == 6) { |
506 | c = map_num(r6layout, __le32_to_cpu(sb->layout)); | |
507 | printf(" Layout : %s\n", c?c:"-unknown-"); | |
508 | } | |
265e0f17 NB |
509 | if (__le32_to_cpu(sb->level) == 10) { |
510 | int lo = __le32_to_cpu(sb->layout); | |
e4965ef8 N |
511 | printf(" Layout :"); |
512 | print_r10_layout(lo); | |
513 | printf("\n"); | |
265e0f17 | 514 | } |
82d9eba6 NB |
515 | switch(__le32_to_cpu(sb->level)) { |
516 | case 0: | |
517 | case 4: | |
518 | case 5: | |
d2cd3ffc NB |
519 | case 6: |
520 | case 10: | |
46a533a9 JS |
521 | printf(" Chunk Size : %dK\n", |
522 | __le32_to_cpu(sb->chunksize)/2); | |
82d9eba6 NB |
523 | break; |
524 | case -1: | |
46a533a9 JS |
525 | printf(" Rounding : %dK\n", |
526 | __le32_to_cpu(sb->chunksize)/2); | |
527 | break; | |
528 | default: | |
82d9eba6 | 529 | break; |
82d9eba6 NB |
530 | } |
531 | printf("\n"); | |
e8e1c176 N |
532 | #if 0 |
533 | /* This turns out to just be confusing */ | |
c3684618 | 534 | printf(" Array Slot : %d (", __le32_to_cpu(sb->dev_number)); |
46a533a9 | 535 | for (i = __le32_to_cpu(sb->max_dev); i> 0 ; i--) |
fa7574f6 | 536 | if (__le16_to_cpu(sb->dev_roles[i-1]) != MD_DISK_ROLE_SPARE) |
c3684618 | 537 | break; |
46a533a9 | 538 | for (d = 0; d < i; d++) { |
c3684618 | 539 | int role = __le16_to_cpu(sb->dev_roles[d]); |
46a533a9 JS |
540 | if (d) |
541 | printf(", "); | |
542 | if (role == MD_DISK_ROLE_SPARE) | |
543 | printf("empty"); | |
544 | else | |
545 | if(role == MD_DISK_ROLE_FAULTY) | |
546 | printf("failed"); | |
547 | else | |
548 | printf("%d", role); | |
c3684618 NB |
549 | } |
550 | printf(")\n"); | |
e8e1c176 N |
551 | #endif |
552 | printf(" Device Role : "); | |
69a48116 | 553 | role = role_from_sb(sb); |
fa7574f6 | 554 | if (role >= MD_DISK_ROLE_FAULTY) |
e8e1c176 | 555 | printf("spare\n"); |
ed94976d SL |
556 | else if (role == MD_DISK_ROLE_JOURNAL) |
557 | printf("Journal\n"); | |
24c7bc84 N |
558 | else if (sb->feature_map & __cpu_to_le32(MD_FEATURE_REPLACEMENT)) |
559 | printf("Replacement device %d\n", role); | |
e8e1c176 N |
560 | else |
561 | printf("Active device %d\n", role); | |
562 | ||
82d9eba6 | 563 | printf(" Array State : "); |
46a533a9 | 564 | for (d = 0; d < __le32_to_cpu(sb->raid_disks) + delta_extra; d++) { |
82d9eba6 | 565 | int cnt = 0; |
f21e18ca | 566 | unsigned int i; |
46a533a9 | 567 | for (i = 0; i < __le32_to_cpu(sb->max_dev); i++) { |
f21e18ca | 568 | unsigned int role = __le16_to_cpu(sb->dev_roles[i]); |
e4c72d1d | 569 | if (role == d) |
82d9eba6 | 570 | cnt++; |
82d9eba6 | 571 | } |
24c7bc84 N |
572 | if (cnt == 2) |
573 | printf("R"); | |
574 | else if (cnt == 1) | |
575 | printf("A"); | |
576 | else if (cnt == 0) | |
577 | printf("."); | |
578 | else | |
579 | printf("?"); | |
82d9eba6 | 580 | } |
e8e1c176 N |
581 | #if 0 |
582 | /* This is confusing too */ | |
6fbba4c9 | 583 | faulty = 0; |
46a533a9 | 584 | for (i = 0; i< __le32_to_cpu(sb->max_dev); i++) { |
82d9eba6 | 585 | int role = __le16_to_cpu(sb->dev_roles[i]); |
fa7574f6 | 586 | if (role == MD_DISK_ROLE_FAULTY) |
6fbba4c9 | 587 | faulty++; |
82d9eba6 | 588 | } |
46a533a9 JS |
589 | if (faulty) |
590 | printf(" %d failed", faulty); | |
e8e1c176 | 591 | #endif |
24c7bc84 | 592 | printf(" ('A' == active, '.' == missing, 'R' == replacing)"); |
82d9eba6 NB |
593 | printf("\n"); |
594 | } | |
595 | ||
061f2c6a | 596 | static void brief_examine_super1(struct supertype *st, int verbose) |
82d9eba6 | 597 | { |
3da92f27 | 598 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 | 599 | int i; |
a17a3de3 | 600 | unsigned long long sb_offset; |
2998bc01 | 601 | char *nm; |
46a533a9 | 602 | char *c = map_num(pers, __le32_to_cpu(sb->level)); |
2998bc01 NB |
603 | |
604 | nm = strchr(sb->set_name, ':'); | |
605 | if (nm) | |
606 | nm++; | |
607 | else if (sb->set_name[0]) | |
608 | nm = sb->set_name; | |
609 | else | |
cf8de691 | 610 | nm = NULL; |
82d9eba6 | 611 | |
7103b9b8 N |
612 | printf("ARRAY "); |
613 | if (nm) { | |
614 | printf("/dev/md/"); | |
615 | print_escape(nm); | |
616 | putchar(' '); | |
617 | } | |
061f2c6a N |
618 | if (verbose && c) |
619 | printf(" level=%s", c); | |
a17a3de3 DL |
620 | sb_offset = __le64_to_cpu(sb->super_offset); |
621 | if (sb_offset <= 4) | |
e9a0e728 | 622 | printf(" metadata=1.1 "); |
a17a3de3 | 623 | else if (sb_offset <= 8) |
e9a0e728 | 624 | printf(" metadata=1.2 "); |
a17a3de3 | 625 | else |
e9a0e728 | 626 | printf(" metadata=1.0 "); |
061f2c6a N |
627 | if (verbose) |
628 | printf("num-devices=%d ", __le32_to_cpu(sb->raid_disks)); | |
629 | printf("UUID="); | |
46a533a9 JS |
630 | for (i = 0; i < 16; i++) { |
631 | if ((i&3)==0 && i != 0) | |
632 | printf(":"); | |
b7a708af | 633 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 | 634 | } |
7103b9b8 N |
635 | if (sb->set_name[0]) { |
636 | printf(" name="); | |
637 | print_quoted(sb->set_name); | |
638 | } | |
82d9eba6 NB |
639 | printf("\n"); |
640 | } | |
641 | ||
0d726f17 KS |
642 | static void export_examine_super1(struct supertype *st) |
643 | { | |
644 | struct mdp_superblock_1 *sb = st->sb; | |
645 | int i; | |
646 | int len = 32; | |
80730bae | 647 | int layout; |
0d726f17 KS |
648 | |
649 | printf("MD_LEVEL=%s\n", map_num(pers, __le32_to_cpu(sb->level))); | |
650 | printf("MD_DEVICES=%d\n", __le32_to_cpu(sb->raid_disks)); | |
46a533a9 JS |
651 | for (i = 0; i < 32; i++) |
652 | if (sb->set_name[i] == '\n' || sb->set_name[i] == '\0') { | |
0d726f17 KS |
653 | len = i; |
654 | break; | |
655 | } | |
656 | if (len) | |
657 | printf("MD_NAME=%.*s\n", len, sb->set_name); | |
80730bae MN |
658 | if (__le32_to_cpu(sb->level) > 0) { |
659 | int ddsks = 0, ddsks_denom = 1; | |
660 | switch(__le32_to_cpu(sb->level)) { | |
46a533a9 JS |
661 | case 1: |
662 | ddsks = 1; | |
663 | break; | |
80730bae | 664 | case 4: |
46a533a9 JS |
665 | case 5: |
666 | ddsks = __le32_to_cpu(sb->raid_disks)-1; | |
667 | break; | |
668 | case 6: | |
669 | ddsks = __le32_to_cpu(sb->raid_disks)-2; | |
670 | break; | |
80730bae MN |
671 | case 10: |
672 | layout = __le32_to_cpu(sb->layout); | |
673 | ddsks = __le32_to_cpu(sb->raid_disks); | |
674 | ddsks_denom = (layout&255) * ((layout>>8)&255); | |
675 | } | |
676 | if (ddsks) { | |
677 | long long asize = __le64_to_cpu(sb->size); | |
678 | asize = (asize << 9) * ddsks / ddsks_denom; | |
46a533a9 JS |
679 | printf("MD_ARRAY_SIZE=%s\n", |
680 | human_size_brief(asize, JEDEC)); | |
80730bae MN |
681 | } |
682 | } | |
0d726f17 | 683 | printf("MD_UUID="); |
46a533a9 JS |
684 | for (i = 0; i < 16; i++) { |
685 | if ((i&3) == 0 && i != 0) | |
686 | printf(":"); | |
0d726f17 KS |
687 | printf("%02x", sb->set_uuid[i]); |
688 | } | |
689 | printf("\n"); | |
690 | printf("MD_UPDATE_TIME=%llu\n", | |
691 | __le64_to_cpu(sb->utime) & 0xFFFFFFFFFFULL); | |
692 | printf("MD_DEV_UUID="); | |
46a533a9 JS |
693 | for (i = 0; i < 16; i++) { |
694 | if ((i&3) == 0 && i != 0) | |
695 | printf(":"); | |
0d726f17 KS |
696 | printf("%02x", sb->device_uuid[i]); |
697 | } | |
698 | printf("\n"); | |
699 | printf("MD_EVENTS=%llu\n", | |
700 | (unsigned long long)__le64_to_cpu(sb->events)); | |
701 | } | |
702 | ||
74db60b0 N |
703 | static int copy_metadata1(struct supertype *st, int from, int to) |
704 | { | |
705 | /* Read superblock. If it looks good, write it out. | |
706 | * Then if a bitmap is present, copy that. | |
707 | * And if a bad-block-list is present, copy that too. | |
708 | */ | |
709 | void *buf; | |
710 | unsigned long long dsize, sb_offset; | |
711 | const int bufsize = 4*1024; | |
712 | struct mdp_superblock_1 super, *sb; | |
713 | ||
714 | if (posix_memalign(&buf, 4096, bufsize) != 0) | |
715 | return 1; | |
716 | ||
717 | if (!get_dev_size(from, NULL, &dsize)) | |
718 | goto err; | |
719 | ||
720 | dsize >>= 9; | |
721 | if (dsize < 24) | |
722 | goto err; | |
723 | switch(st->minor_version) { | |
724 | case 0: | |
725 | sb_offset = dsize; | |
726 | sb_offset -= 8*2; | |
727 | sb_offset &= ~(4*2-1); | |
728 | break; | |
729 | case 1: | |
730 | sb_offset = 0; | |
731 | break; | |
732 | case 2: | |
733 | sb_offset = 4*2; | |
734 | break; | |
735 | default: | |
736 | goto err; | |
737 | } | |
738 | ||
739 | if (lseek64(from, sb_offset << 9, 0) < 0LL) | |
740 | goto err; | |
741 | if (read(from, buf, bufsize) != bufsize) | |
742 | goto err; | |
743 | ||
744 | sb = buf; | |
745 | super = *sb; // save most of sb for when we reuse buf | |
746 | ||
747 | if (__le32_to_cpu(super.magic) != MD_SB_MAGIC || | |
748 | __le32_to_cpu(super.major_version) != 1 || | |
749 | __le64_to_cpu(super.super_offset) != sb_offset || | |
750 | calc_sb_1_csum(sb) != super.sb_csum) | |
751 | goto err; | |
752 | ||
753 | if (lseek64(to, sb_offset << 9, 0) < 0LL) | |
754 | goto err; | |
755 | if (write(to, buf, bufsize) != bufsize) | |
756 | goto err; | |
757 | ||
758 | if (super.feature_map & __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET)) { | |
759 | unsigned long long bitmap_offset = sb_offset; | |
760 | int bytes = 4096; // just an estimate. | |
761 | int written = 0; | |
762 | struct align_fd afrom, ato; | |
763 | ||
764 | init_afd(&afrom, from); | |
765 | init_afd(&ato, to); | |
766 | ||
767 | bitmap_offset += (int32_t)__le32_to_cpu(super.bitmap_offset); | |
768 | ||
769 | if (lseek64(from, bitmap_offset<<9, 0) < 0) | |
770 | goto err; | |
771 | if (lseek64(to, bitmap_offset<<9, 0) < 0) | |
772 | goto err; | |
773 | ||
774 | for (written = 0; written < bytes ; ) { | |
775 | int n = bytes - written; | |
776 | if (n > 4096) | |
777 | n = 4096; | |
778 | if (aread(&afrom, buf, n) != n) | |
779 | goto err; | |
780 | if (written == 0) { | |
781 | /* have the header, can calculate | |
782 | * correct bitmap bytes */ | |
783 | bitmap_super_t *bms; | |
74db60b0 | 784 | bms = (void*)buf; |
4a3d29ed | 785 | bytes = calc_bitmap_size(bms, 512); |
74db60b0 N |
786 | if (n > bytes) |
787 | n = bytes; | |
788 | } | |
789 | if (awrite(&ato, buf, n) != n) | |
790 | goto err; | |
791 | written += n; | |
792 | } | |
793 | } | |
794 | ||
795 | if (super.bblog_size != 0 && | |
27c7c87a | 796 | __le16_to_cpu(super.bblog_size) <= 100 && |
74db60b0 N |
797 | super.bblog_offset != 0 && |
798 | (super.feature_map & __le32_to_cpu(MD_FEATURE_BAD_BLOCKS))) { | |
799 | /* There is a bad block log */ | |
800 | unsigned long long bb_offset = sb_offset; | |
27c7c87a | 801 | int bytes = __le16_to_cpu(super.bblog_size) * 512; |
74db60b0 N |
802 | int written = 0; |
803 | struct align_fd afrom, ato; | |
804 | ||
805 | init_afd(&afrom, from); | |
806 | init_afd(&ato, to); | |
807 | ||
808 | bb_offset += (int32_t)__le32_to_cpu(super.bblog_offset); | |
809 | ||
810 | if (lseek64(from, bb_offset<<9, 0) < 0) | |
811 | goto err; | |
812 | if (lseek64(to, bb_offset<<9, 0) < 0) | |
813 | goto err; | |
814 | ||
815 | for (written = 0; written < bytes ; ) { | |
816 | int n = bytes - written; | |
817 | if (n > 4096) | |
818 | n = 4096; | |
819 | if (aread(&afrom, buf, n) != n) | |
820 | goto err; | |
821 | ||
822 | if (awrite(&ato, buf, n) != n) | |
823 | goto err; | |
824 | written += n; | |
825 | } | |
826 | } | |
827 | ||
828 | free(buf); | |
829 | return 0; | |
830 | ||
831 | err: | |
832 | free(buf); | |
833 | return 1; | |
834 | } | |
835 | ||
3da92f27 | 836 | static void detail_super1(struct supertype *st, char *homehost) |
82d9eba6 | 837 | { |
3da92f27 | 838 | struct mdp_superblock_1 *sb = st->sb; |
0aa2f15b | 839 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); |
82d9eba6 | 840 | int i; |
b6750aa8 | 841 | int l = homehost ? strlen(homehost) : 0; |
82d9eba6 | 842 | |
65884368 | 843 | printf(" Name : %.32s", sb->set_name); |
46a533a9 | 844 | if (l > 0 && l < 32 && sb->set_name[l] == ':' && |
b6750aa8 NB |
845 | strncmp(sb->set_name, homehost, l) == 0) |
846 | printf(" (local to host %s)", homehost); | |
46a533a9 JS |
847 | if (bms->nodes > 0 && |
848 | (__le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET)) | |
849 | printf("\n Cluster Name : %-64s", bms->cluster_name); | |
65884368 | 850 | printf("\n UUID : "); |
46a533a9 JS |
851 | for (i = 0; i < 16; i++) { |
852 | if ((i&3) == 0 && i != 0) | |
853 | printf(":"); | |
34163fc7 | 854 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 | 855 | } |
65884368 AP |
856 | printf("\n Events : %llu\n\n", |
857 | (unsigned long long)__le64_to_cpu(sb->events)); | |
82d9eba6 NB |
858 | } |
859 | ||
3da92f27 | 860 | static void brief_detail_super1(struct supertype *st) |
82d9eba6 | 861 | { |
3da92f27 | 862 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 NB |
863 | int i; |
864 | ||
7103b9b8 N |
865 | if (sb->set_name[0]) { |
866 | printf(" name="); | |
867 | print_quoted(sb->set_name); | |
868 | } | |
82d9eba6 | 869 | printf(" UUID="); |
46a533a9 JS |
870 | for (i = 0; i < 16; i++) { |
871 | if ((i & 3) == 0 && i != 0) | |
872 | printf(":"); | |
34163fc7 | 873 | printf("%02x", sb->set_uuid[i]); |
82d9eba6 NB |
874 | } |
875 | } | |
876 | ||
0d726f17 | 877 | static void export_detail_super1(struct supertype *st) |
54bad364 | 878 | { |
3da92f27 | 879 | struct mdp_superblock_1 *sb = st->sb; |
54bad364 KS |
880 | int i; |
881 | int len = 32; | |
882 | ||
46a533a9 JS |
883 | for (i = 0; i < 32; i++) |
884 | if (sb->set_name[i] == '\n' || sb->set_name[i] == '\0') { | |
54bad364 KS |
885 | len = i; |
886 | break; | |
887 | } | |
888 | if (len) | |
889 | printf("MD_NAME=%.*s\n", len, sb->set_name); | |
54bad364 KS |
890 | } |
891 | ||
6d388a88 N |
892 | static int examine_badblocks_super1(struct supertype *st, int fd, char *devname) |
893 | { | |
894 | struct mdp_superblock_1 *sb = st->sb; | |
895 | unsigned long long offset; | |
896 | int size; | |
897 | __u64 *bbl, *bbp; | |
898 | int i; | |
899 | ||
46a533a9 JS |
900 | if (!sb->bblog_size || __le16_to_cpu(sb->bblog_size) > 100 || |
901 | !sb->bblog_offset){ | |
6d388a88 N |
902 | printf("No bad-blocks list configured on %s\n", devname); |
903 | return 0; | |
904 | } | |
46a533a9 | 905 | if ((sb->feature_map & __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) == 0) { |
6d388a88 N |
906 | printf("Bad-blocks list is empty in %s\n", devname); |
907 | return 0; | |
908 | } | |
909 | ||
27c7c87a | 910 | size = __le16_to_cpu(sb->bblog_size)* 512; |
0cf83229 | 911 | if (posix_memalign((void**)&bbl, 4096, size) != 0) { |
1ade5cc1 | 912 | pr_err("could not allocate badblocks list\n"); |
0cf83229 N |
913 | return 0; |
914 | } | |
6d388a88 N |
915 | offset = __le64_to_cpu(sb->super_offset) + |
916 | (int)__le32_to_cpu(sb->bblog_offset); | |
917 | offset <<= 9; | |
918 | if (lseek64(fd, offset, 0) < 0) { | |
919 | pr_err("Cannot seek to bad-blocks list\n"); | |
920 | return 1; | |
921 | } | |
922 | if (read(fd, bbl, size) != size) { | |
923 | pr_err("Cannot read bad-blocks list\n"); | |
924 | return 1; | |
925 | } | |
926 | /* 64bits per entry. 10 bits is block-count, 54 bits is block | |
927 | * offset. Blocks are sectors unless bblog->shift makes them bigger | |
928 | */ | |
929 | bbp = (__u64*)bbl; | |
930 | printf("Bad-blocks on %s:\n", devname); | |
931 | for (i = 0; i < size/8; i++, bbp++) { | |
932 | __u64 bb = __le64_to_cpu(*bbp); | |
933 | int count = bb & 0x3ff; | |
934 | unsigned long long sector = bb >> 10; | |
935 | ||
936 | if (bb + 1 == 0) | |
937 | break; | |
938 | ||
939 | sector <<= sb->bblog_shift; | |
940 | count <<= sb->bblog_shift; | |
941 | ||
942 | printf("%20llu for %d sectors\n", sector, count); | |
943 | } | |
944 | return 0; | |
945 | } | |
946 | ||
3da92f27 | 947 | static int match_home1(struct supertype *st, char *homehost) |
83b6208e | 948 | { |
3da92f27 | 949 | struct mdp_superblock_1 *sb = st->sb; |
83b6208e NB |
950 | int l = homehost ? strlen(homehost) : 0; |
951 | ||
46a533a9 | 952 | return (l > 0 && l < 32 && sb->set_name[l] == ':' && |
83b6208e NB |
953 | strncmp(sb->set_name, homehost, l) == 0); |
954 | } | |
955 | ||
3da92f27 | 956 | static void uuid_from_super1(struct supertype *st, int uuid[4]) |
82d9eba6 | 957 | { |
3da92f27 | 958 | struct mdp_superblock_1 *super = st->sb; |
82d9eba6 NB |
959 | char *cuuid = (char*)uuid; |
960 | int i; | |
46a533a9 | 961 | for (i = 0; i < 16; i++) |
82d9eba6 NB |
962 | cuuid[i] = super->set_uuid[i]; |
963 | } | |
964 | ||
a5d85af7 | 965 | static void getinfo_super1(struct supertype *st, struct mdinfo *info, char *map) |
82d9eba6 | 966 | { |
3da92f27 | 967 | struct mdp_superblock_1 *sb = st->sb; |
80bf9135 | 968 | struct bitmap_super_s *bsb = (void*)(((char*)sb)+MAX_SB_SIZE); |
46a533a9 JS |
969 | struct misc_dev_info *misc = |
970 | (void*)(((char*)sb)+MAX_SB_SIZE+BM_SUPER_SIZE); | |
82d9eba6 | 971 | int working = 0; |
f21e18ca | 972 | unsigned int i; |
a5d85af7 N |
973 | unsigned int role; |
974 | unsigned int map_disks = info->array.raid_disks; | |
80bf9135 N |
975 | unsigned long long super_offset; |
976 | unsigned long long data_size; | |
82d9eba6 | 977 | |
95eeceeb | 978 | memset(info, 0, sizeof(*info)); |
82d9eba6 | 979 | info->array.major_version = 1; |
b8ac1967 | 980 | info->array.minor_version = st->minor_version; |
82d9eba6 NB |
981 | info->array.patch_version = 0; |
982 | info->array.raid_disks = __le32_to_cpu(sb->raid_disks); | |
983 | info->array.level = __le32_to_cpu(sb->level); | |
265e0f17 | 984 | info->array.layout = __le32_to_cpu(sb->layout); |
82d9eba6 NB |
985 | info->array.md_minor = -1; |
986 | info->array.ctime = __le64_to_cpu(sb->ctime); | |
353632d9 | 987 | info->array.utime = __le64_to_cpu(sb->utime); |
bed256c2 | 988 | info->array.chunk_size = __le32_to_cpu(sb->chunksize)*512; |
4855f95c | 989 | info->array.state = |
46a533a9 | 990 | (__le64_to_cpu(sb->resync_offset) == MaxSector) ? 1 : 0; |
353632d9 | 991 | |
e97a7cd0 | 992 | super_offset = __le64_to_cpu(sb->super_offset); |
353632d9 NB |
993 | info->data_offset = __le64_to_cpu(sb->data_offset); |
994 | info->component_size = __le64_to_cpu(sb->size); | |
e97a7cd0 | 995 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET)) { |
34a13953 | 996 | info->bitmap_offset = (int32_t)__le32_to_cpu(sb->bitmap_offset); |
6438c249 N |
997 | if (__le32_to_cpu(bsb->nodes) > 1) |
998 | info->array.state |= (1 << MD_SB_CLUSTERED); | |
fa601c2e | 999 | } else if (md_feature_any_ppl_on(sb->feature_map)) { |
e97a7cd0 AP |
1000 | info->ppl_offset = __le16_to_cpu(sb->ppl.offset); |
1001 | info->ppl_size = __le16_to_cpu(sb->ppl.size); | |
1002 | info->ppl_sector = super_offset + info->ppl_offset; | |
1003 | } | |
82d9eba6 NB |
1004 | |
1005 | info->disk.major = 0; | |
1006 | info->disk.minor = 0; | |
fbf8a0b7 | 1007 | info->disk.number = __le32_to_cpu(sb->dev_number); |
82d9eba6 | 1008 | if (__le32_to_cpu(sb->dev_number) >= __le32_to_cpu(sb->max_dev) || |
911cead7 | 1009 | __le32_to_cpu(sb->dev_number) >= MAX_DEVS) |
fa7574f6 | 1010 | role = MD_DISK_ROLE_FAULTY; |
82d9eba6 NB |
1011 | else |
1012 | role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]); | |
1013 | ||
0ddc35be N |
1014 | if (info->array.level <= 0) |
1015 | data_size = __le64_to_cpu(sb->data_size); | |
1016 | else | |
1017 | data_size = __le64_to_cpu(sb->size); | |
80bf9135 N |
1018 | if (info->data_offset < super_offset) { |
1019 | unsigned long long end; | |
1020 | info->space_before = info->data_offset; | |
1021 | end = super_offset; | |
cc3130a7 N |
1022 | |
1023 | if (sb->bblog_offset && sb->bblog_size) { | |
1024 | unsigned long long bboffset = super_offset; | |
1025 | bboffset += (int32_t)__le32_to_cpu(sb->bblog_offset); | |
1026 | if (bboffset < end) | |
1027 | end = bboffset; | |
1028 | } | |
1029 | ||
e97a7cd0 | 1030 | if (super_offset + info->bitmap_offset + info->ppl_offset < end) |
46a533a9 JS |
1031 | end = super_offset + info->bitmap_offset + |
1032 | info->ppl_offset; | |
cc3130a7 | 1033 | |
80bf9135 N |
1034 | if (info->data_offset + data_size < end) |
1035 | info->space_after = end - data_size - info->data_offset; | |
1036 | else | |
1037 | info->space_after = 0; | |
1038 | } else { | |
cc3130a7 N |
1039 | unsigned long long earliest; |
1040 | earliest = super_offset + (32+4)*2; /* match kernel */ | |
80bf9135 N |
1041 | if (info->bitmap_offset > 0) { |
1042 | unsigned long long bmend = info->bitmap_offset; | |
561ad559 | 1043 | unsigned long long size = calc_bitmap_size(bsb, 4096); |
80bf9135 | 1044 | size /= 512; |
cc3130a7 N |
1045 | bmend += size; |
1046 | if (bmend > earliest) | |
ac92b44a | 1047 | earliest = bmend; |
e97a7cd0 | 1048 | } else if (info->ppl_offset > 0) { |
46a533a9 JS |
1049 | unsigned long long pplend; |
1050 | ||
1051 | pplend = info->ppl_offset + info->ppl_size; | |
e97a7cd0 AP |
1052 | if (pplend > earliest) |
1053 | earliest = pplend; | |
cc3130a7 N |
1054 | } |
1055 | if (sb->bblog_offset && sb->bblog_size) { | |
1056 | unsigned long long bbend = super_offset; | |
1057 | bbend += (int32_t)__le32_to_cpu(sb->bblog_offset); | |
27c7c87a | 1058 | bbend += __le16_to_cpu(sb->bblog_size); |
cc3130a7 N |
1059 | if (bbend > earliest) |
1060 | earliest = bbend; | |
1061 | } | |
1062 | if (earliest < info->data_offset) | |
1063 | info->space_before = info->data_offset - earliest; | |
1064 | else | |
1065 | info->space_before = 0; | |
46a533a9 JS |
1066 | info->space_after = misc->device_size - data_size - |
1067 | info->data_offset; | |
80bf9135 | 1068 | } |
c4b26c64 N |
1069 | if (info->space_before == 0 && info->space_after == 0) { |
1070 | /* It will look like we don't support data_offset changes, | |
1071 | * be we do - it's just that there is no room. | |
1072 | * A change that reduced the number of devices should | |
1073 | * still be allowed, so set the otherwise useless value of '1' | |
1074 | */ | |
1075 | info->space_after = 1; | |
1076 | } | |
80bf9135 | 1077 | |
82d9eba6 NB |
1078 | info->disk.raid_disk = -1; |
1079 | switch(role) { | |
fa7574f6 | 1080 | case MD_DISK_ROLE_SPARE: |
46a533a9 JS |
1081 | /* spare: not active, not sync, not faulty */ |
1082 | info->disk.state = 0; | |
82d9eba6 | 1083 | break; |
fa7574f6 | 1084 | case MD_DISK_ROLE_FAULTY: |
aa313118 | 1085 | info->disk.state = (1 << MD_DISK_FAULTY); /* faulty */ |
82d9eba6 | 1086 | break; |
ed94976d SL |
1087 | case MD_DISK_ROLE_JOURNAL: |
1088 | info->disk.state = (1 << MD_DISK_JOURNAL); | |
1089 | info->disk.raid_disk = role; | |
46a533a9 JS |
1090 | /* journal uses all 4kB blocks*/ |
1091 | info->space_after = (misc->device_size - info->data_offset) % 8; | |
ed94976d | 1092 | break; |
82d9eba6 NB |
1093 | default: |
1094 | info->disk.state = 6; /* active and in sync */ | |
1095 | info->disk.raid_disk = role; | |
1096 | } | |
cb099724 N |
1097 | if (sb->devflags & WriteMostly1) |
1098 | info->disk.state |= (1 << MD_DISK_WRITEMOSTLY); | |
71574efb N |
1099 | if (sb->devflags & FailFast1) |
1100 | info->disk.state |= (1 << MD_DISK_FAILFAST); | |
82d9eba6 | 1101 | info->events = __le64_to_cpu(sb->events); |
1522c538 | 1102 | sprintf(info->text_version, "1.%d", st->minor_version); |
a67dd8cc | 1103 | info->safe_mode_delay = 200; |
82d9eba6 NB |
1104 | |
1105 | memcpy(info->uuid, sb->set_uuid, 16); | |
1106 | ||
31317663 NB |
1107 | strncpy(info->name, sb->set_name, 32); |
1108 | info->name[32] = 0; | |
947fd4dd | 1109 | |
aacb2f81 | 1110 | if ((__le32_to_cpu(sb->feature_map)&MD_FEATURE_REPLACEMENT)) { |
72e7fb13 | 1111 | info->disk.state &= ~(1 << MD_DISK_SYNC); |
aacb2f81 N |
1112 | info->disk.state |= 1 << MD_DISK_REPLACEMENT; |
1113 | } | |
1114 | ||
921d9e16 N |
1115 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET)) |
1116 | info->recovery_start = __le32_to_cpu(sb->recovery_offset); | |
1117 | else | |
1118 | info->recovery_start = MaxSector; | |
1119 | ||
353632d9 NB |
1120 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) { |
1121 | info->reshape_active = 1; | |
26bf5587 N |
1122 | if ((sb->feature_map & __le32_to_cpu(MD_FEATURE_NEW_OFFSET)) && |
1123 | sb->new_offset != 0) | |
5e88ab2e | 1124 | info->reshape_active |= RESHAPE_NO_BACKUP; |
353632d9 NB |
1125 | info->reshape_progress = __le64_to_cpu(sb->reshape_position); |
1126 | info->new_level = __le32_to_cpu(sb->new_level); | |
1127 | info->delta_disks = __le32_to_cpu(sb->delta_disks); | |
1128 | info->new_layout = __le32_to_cpu(sb->new_layout); | |
bed256c2 | 1129 | info->new_chunk = __le32_to_cpu(sb->new_chunk)<<9; |
4180aa4d N |
1130 | if (info->delta_disks < 0) |
1131 | info->array.raid_disks -= info->delta_disks; | |
353632d9 NB |
1132 | } else |
1133 | info->reshape_active = 0; | |
1134 | ||
6e75048b AK |
1135 | info->recovery_blocked = info->reshape_active; |
1136 | ||
a5d85af7 N |
1137 | if (map) |
1138 | for (i=0; i<map_disks; i++) | |
1139 | map[i] = 0; | |
f21e18ca | 1140 | for (i = 0; i < __le32_to_cpu(sb->max_dev); i++) { |
82d9eba6 | 1141 | role = __le16_to_cpu(sb->dev_roles[i]); |
fa7574f6 | 1142 | if (/*role == MD_DISK_ROLE_SPARE || */role < (unsigned) info->array.raid_disks) { |
82d9eba6 | 1143 | working++; |
a5d85af7 N |
1144 | if (map && role < map_disks) |
1145 | map[role] = 1; | |
1146 | } | |
82d9eba6 NB |
1147 | } |
1148 | ||
1149 | info->array.working_disks = working; | |
e97a7cd0 AP |
1150 | |
1151 | if (sb->feature_map & __le32_to_cpu(MD_FEATURE_JOURNAL)) { | |
051f3265 | 1152 | info->journal_device_required = 1; |
e97a7cd0 | 1153 | info->consistency_policy = CONSISTENCY_POLICY_JOURNAL; |
fa601c2e | 1154 | } else if (md_feature_any_ppl_on(sb->feature_map)) { |
e97a7cd0 AP |
1155 | info->consistency_policy = CONSISTENCY_POLICY_PPL; |
1156 | } else if (sb->feature_map & __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET)) { | |
1157 | info->consistency_policy = CONSISTENCY_POLICY_BITMAP; | |
1158 | } else if (info->array.level <= 0) { | |
1159 | info->consistency_policy = CONSISTENCY_POLICY_NONE; | |
1160 | } else { | |
1161 | info->consistency_policy = CONSISTENCY_POLICY_RESYNC; | |
1162 | } | |
1163 | ||
051f3265 | 1164 | info->journal_clean = 0; |
82d9eba6 NB |
1165 | } |
1166 | ||
00bbdbda N |
1167 | static struct mdinfo *container_content1(struct supertype *st, char *subarray) |
1168 | { | |
1169 | struct mdinfo *info; | |
1170 | ||
1171 | if (subarray) | |
1172 | return NULL; | |
1173 | ||
503975b9 | 1174 | info = xmalloc(sizeof(*info)); |
00bbdbda N |
1175 | getinfo_super1(st, info, NULL); |
1176 | return info; | |
1177 | } | |
1178 | ||
68c7d6d7 | 1179 | static int update_super1(struct supertype *st, struct mdinfo *info, |
46a533a9 | 1180 | char *update, char *devname, int verbose, |
e5eac01f | 1181 | int uuid_set, char *homehost) |
82d9eba6 | 1182 | { |
5a31170d N |
1183 | /* NOTE: for 'assemble' and 'force' we need to return non-zero |
1184 | * if any change was made. For others, the return value is | |
1185 | * ignored. | |
c4d831e1 | 1186 | */ |
82d9eba6 | 1187 | int rv = 0; |
d15a1f72 | 1188 | int lockid; |
3da92f27 | 1189 | struct mdp_superblock_1 *sb = st->sb; |
81a8a694 | 1190 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); |
82d9eba6 | 1191 | |
81a8a694 GJ |
1192 | if (bms->version == BITMAP_MAJOR_CLUSTERED && dlm_funs_ready()) { |
1193 | rv = cluster_get_dlmlock(&lockid); | |
d15a1f72 | 1194 | if (rv) { |
46a533a9 JS |
1195 | pr_err("Cannot get dlmlock in %s return %d\n", |
1196 | __func__, rv); | |
81a8a694 | 1197 | cluster_release_dlmlock(lockid); |
d15a1f72 GJ |
1198 | return rv; |
1199 | } | |
1200 | } | |
1201 | ||
def11332 N |
1202 | if (strcmp(update, "homehost") == 0 && |
1203 | homehost) { | |
1204 | /* Note that 'homehost' is special as it is really | |
1205 | * a "name" update. | |
1206 | */ | |
1207 | char *c; | |
1208 | update = "name"; | |
1209 | c = strchr(sb->set_name, ':'); | |
1210 | if (c) | |
1211 | strncpy(info->name, c+1, 31 - (c-sb->set_name)); | |
1212 | else | |
1213 | strncpy(info->name, sb->set_name, 32); | |
1214 | info->name[32] = 0; | |
1215 | } | |
1216 | ||
67a8c82d NB |
1217 | if (strcmp(update, "force-one")==0) { |
1218 | /* Not enough devices for a working array, | |
1219 | * so bring this one up-to-date | |
1220 | */ | |
c4d831e1 NB |
1221 | if (sb->events != __cpu_to_le64(info->events)) |
1222 | rv = 1; | |
6d3d5804 | 1223 | sb->events = __cpu_to_le64(info->events); |
1e2b2765 | 1224 | } else if (strcmp(update, "force-array")==0) { |
67a8c82d NB |
1225 | /* Degraded array and 'force' requests to |
1226 | * maybe need to mark it 'clean'. | |
1227 | */ | |
82d9eba6 | 1228 | switch(__le32_to_cpu(sb->level)) { |
46a533a9 JS |
1229 | case 4: |
1230 | case 5: | |
1231 | case 6: | |
82d9eba6 | 1232 | /* need to force clean */ |
b7528a20 | 1233 | if (sb->resync_offset != MaxSector) |
c4d831e1 | 1234 | rv = 1; |
b7528a20 | 1235 | sb->resync_offset = MaxSector; |
82d9eba6 | 1236 | } |
1e2b2765 | 1237 | } else if (strcmp(update, "assemble")==0) { |
82d9eba6 NB |
1238 | int d = info->disk.number; |
1239 | int want; | |
aacb2f81 | 1240 | if (info->disk.state & (1<<MD_DISK_ACTIVE)) |
a2ce5a1a | 1241 | want = info->disk.raid_disk; |
69a48116 SL |
1242 | else if (info->disk.state & (1<<MD_DISK_JOURNAL)) |
1243 | want = MD_DISK_ROLE_JOURNAL; | |
82d9eba6 | 1244 | else |
fa7574f6 | 1245 | want = MD_DISK_ROLE_SPARE; |
a2ce5a1a N |
1246 | if (sb->dev_roles[d] != __cpu_to_le16(want)) { |
1247 | sb->dev_roles[d] = __cpu_to_le16(want); | |
82d9eba6 NB |
1248 | rv = 1; |
1249 | } | |
d43494fc | 1250 | if (info->reshape_active && |
46a533a9 JS |
1251 | sb->feature_map & |
1252 | __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) && | |
d43494fc | 1253 | info->delta_disks >= 0 && |
46a533a9 JS |
1254 | info->reshape_progress < |
1255 | __le64_to_cpu(sb->reshape_position)) { | |
1256 | sb->reshape_position = | |
1257 | __cpu_to_le64(info->reshape_progress); | |
d43494fc N |
1258 | rv = 1; |
1259 | } | |
1260 | if (info->reshape_active && | |
46a533a9 JS |
1261 | sb->feature_map & |
1262 | __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE) && | |
d43494fc | 1263 | info->delta_disks < 0 && |
46a533a9 JS |
1264 | info->reshape_progress > |
1265 | __le64_to_cpu(sb->reshape_position)) { | |
1266 | sb->reshape_position = | |
1267 | __cpu_to_le64(info->reshape_progress); | |
d43494fc N |
1268 | rv = 1; |
1269 | } | |
1e2b2765 | 1270 | } else if (strcmp(update, "linear-grow-new") == 0) { |
f21e18ca | 1271 | unsigned int i; |
c5f71c24 | 1272 | int fd; |
f21e18ca | 1273 | unsigned int max = __le32_to_cpu(sb->max_dev); |
f752781f | 1274 | |
46a533a9 JS |
1275 | for (i = 0; i < max; i++) |
1276 | if (__le16_to_cpu(sb->dev_roles[i]) >= | |
1277 | MD_DISK_ROLE_FAULTY) | |
f752781f NB |
1278 | break; |
1279 | sb->dev_number = __cpu_to_le32(i); | |
1280 | info->disk.number = i; | |
68fee4af | 1281 | if (i >= max) { |
f752781f | 1282 | sb->max_dev = __cpu_to_le32(max+1); |
68fee4af | 1283 | } |
f752781f | 1284 | |
c5f71c24 | 1285 | random_uuid(sb->device_uuid); |
f752781f | 1286 | |
46a533a9 | 1287 | sb->dev_roles[i] = __cpu_to_le16(info->disk.raid_disk); |
1c6cb603 N |
1288 | |
1289 | fd = open(devname, O_RDONLY); | |
1290 | if (fd >= 0) { | |
1291 | unsigned long long ds; | |
1292 | get_dev_size(fd, devname, &ds); | |
1293 | close(fd); | |
1294 | ds >>= 9; | |
1295 | if (__le64_to_cpu(sb->super_offset) < | |
1296 | __le64_to_cpu(sb->data_offset)) { | |
1297 | sb->data_size = __cpu_to_le64( | |
1298 | ds - __le64_to_cpu(sb->data_offset)); | |
1299 | } else { | |
1300 | ds -= 8*2; | |
1301 | ds &= ~(unsigned long long)(4*2-1); | |
1302 | sb->super_offset = __cpu_to_le64(ds); | |
1303 | sb->data_size = __cpu_to_le64( | |
1304 | ds - __le64_to_cpu(sb->data_offset)); | |
1305 | } | |
1306 | } | |
1e2b2765 | 1307 | } else if (strcmp(update, "linear-grow-update") == 0) { |
68fee4af | 1308 | int max = __le32_to_cpu(sb->max_dev); |
82d9eba6 | 1309 | sb->raid_disks = __cpu_to_le32(info->array.raid_disks); |
68fee4af LZ |
1310 | if (info->array.raid_disks > max) { |
1311 | sb->max_dev = __cpu_to_le32(max+1); | |
1312 | } | |
f752781f NB |
1313 | sb->dev_roles[info->disk.number] = |
1314 | __cpu_to_le16(info->disk.raid_disk); | |
1e2b2765 | 1315 | } else if (strcmp(update, "resync") == 0) { |
82d9eba6 | 1316 | /* make sure resync happens */ |
434b7755 | 1317 | sb->resync_offset = 0ULL; |
1e2b2765 | 1318 | } else if (strcmp(update, "uuid") == 0) { |
350f29f9 | 1319 | copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid); |
bf4fb153 | 1320 | |
561ad559 AP |
1321 | if (__le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET) |
1322 | memcpy(bms->uuid, sb->set_uuid, 16); | |
5a31170d N |
1323 | } else if (strcmp(update, "no-bitmap") == 0) { |
1324 | sb->feature_map &= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET); | |
688e99a7 N |
1325 | } else if (strcmp(update, "bbl") == 0) { |
1326 | /* only possible if there is room after the bitmap, or if | |
1327 | * there is no bitmap | |
1328 | */ | |
1329 | unsigned long long sb_offset = __le64_to_cpu(sb->super_offset); | |
1330 | unsigned long long data_offset = __le64_to_cpu(sb->data_offset); | |
561ad559 | 1331 | long bitmap_offset = 0; |
688e99a7 N |
1332 | long bm_sectors = 0; |
1333 | long space; | |
1334 | ||
1335 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
561ad559 AP |
1336 | bitmap_offset = (long)__le32_to_cpu(sb->bitmap_offset); |
1337 | bm_sectors = calc_bitmap_size(bms, 4096) >> 9; | |
fa601c2e | 1338 | } else if (md_feature_any_ppl_on(sb->feature_map)) { |
e97a7cd0 AP |
1339 | bitmap_offset = (long)__le16_to_cpu(sb->ppl.offset); |
1340 | bm_sectors = (long)__le16_to_cpu(sb->ppl.size); | |
688e99a7 | 1341 | } |
32141c17 | 1342 | |
688e99a7 | 1343 | if (sb_offset < data_offset) { |
46a533a9 JS |
1344 | /* |
1345 | * 1.1 or 1.2. Put bbl after bitmap leaving | |
1346 | * at least 32K | |
688e99a7 N |
1347 | */ |
1348 | long bb_offset; | |
f79bbf4f | 1349 | bb_offset = sb_offset + 8; |
688e99a7 | 1350 | if (bm_sectors && bitmap_offset > 0) |
f79bbf4f | 1351 | bb_offset = bitmap_offset + bm_sectors; |
46a533a9 JS |
1352 | while (bb_offset < (long)sb_offset + 8 + 32*2 && |
1353 | bb_offset + 8+8 <= (long)data_offset) | |
f79bbf4f N |
1354 | /* too close to bitmap, and room to grow */ |
1355 | bb_offset += 8; | |
1356 | if (bb_offset + 8 <= (long)data_offset) { | |
688e99a7 N |
1357 | sb->bblog_size = __cpu_to_le16(8); |
1358 | sb->bblog_offset = __cpu_to_le32(bb_offset); | |
1359 | } | |
1360 | } else { | |
1361 | /* 1.0 - Put bbl just before super block */ | |
1362 | if (bm_sectors && bitmap_offset < 0) | |
1363 | space = -bitmap_offset - bm_sectors; | |
1364 | else | |
1365 | space = sb_offset - data_offset - | |
1366 | __le64_to_cpu(sb->data_size); | |
1367 | if (space >= 8) { | |
1368 | sb->bblog_size = __cpu_to_le16(8); | |
1369 | sb->bblog_offset = __cpu_to_le32((unsigned)-8); | |
1370 | } | |
1371 | } | |
1372 | } else if (strcmp(update, "no-bbl") == 0) { | |
1373 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BAD_BLOCKS)) | |
1374 | pr_err("Cannot remove active bbl from %s\n",devname); | |
1375 | else { | |
1376 | sb->bblog_size = 0; | |
1377 | sb->bblog_shift = 0; | |
1378 | sb->bblog_offset = 0; | |
1379 | } | |
6dd16dac N |
1380 | } else if (strcmp(update, "force-no-bbl") == 0) { |
1381 | sb->feature_map &= ~ __cpu_to_le32(MD_FEATURE_BAD_BLOCKS); | |
1382 | sb->bblog_size = 0; | |
1383 | sb->bblog_shift = 0; | |
1384 | sb->bblog_offset = 0; | |
e6e9dd3f AP |
1385 | } else if (strcmp(update, "ppl") == 0) { |
1386 | unsigned long long sb_offset = __le64_to_cpu(sb->super_offset); | |
1387 | unsigned long long data_offset = __le64_to_cpu(sb->data_offset); | |
1388 | unsigned long long data_size = __le64_to_cpu(sb->data_size); | |
1389 | long bb_offset = __le32_to_cpu(sb->bblog_offset); | |
1390 | int space; | |
e6e9dd3f AP |
1391 | int offset; |
1392 | ||
1393 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
1394 | pr_err("Cannot add PPL to array with bitmap\n"); | |
1395 | return -2; | |
1396 | } | |
1397 | ||
1398 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_JOURNAL)) { | |
1399 | pr_err("Cannot add PPL to array with journal\n"); | |
1400 | return -2; | |
1401 | } | |
1402 | ||
1403 | if (sb_offset < data_offset) { | |
1404 | if (bb_offset) | |
1405 | space = bb_offset - 8; | |
1406 | else | |
1407 | space = data_offset - sb_offset - 8; | |
1408 | offset = 8; | |
1409 | } else { | |
1410 | offset = -(sb_offset - data_offset - data_size); | |
1411 | if (offset < INT16_MIN) | |
1412 | offset = INT16_MIN; | |
1413 | space = -(offset - bb_offset); | |
1414 | } | |
1415 | ||
1416 | if (space < (PPL_HEADER_SIZE >> 9) + 8) { | |
1417 | pr_err("Not enough space to add ppl\n"); | |
1418 | return -2; | |
1419 | } | |
1420 | ||
fa601c2e PB |
1421 | if (space >= (MULTIPLE_PPL_AREA_SIZE_SUPER1 >> 9)) { |
1422 | space = (MULTIPLE_PPL_AREA_SIZE_SUPER1 >> 9); | |
1423 | } else { | |
1424 | int optimal_space = choose_ppl_space( | |
1425 | __le32_to_cpu(sb->chunksize)); | |
1426 | if (space > optimal_space) | |
1427 | space = optimal_space; | |
1428 | if (space > UINT16_MAX) | |
1429 | space = UINT16_MAX; | |
1430 | } | |
e6e9dd3f AP |
1431 | |
1432 | sb->ppl.offset = __cpu_to_le16(offset); | |
1433 | sb->ppl.size = __cpu_to_le16(space); | |
1434 | sb->feature_map |= __cpu_to_le32(MD_FEATURE_PPL); | |
1435 | } else if (strcmp(update, "no-ppl") == 0) { | |
fa601c2e PB |
1436 | sb->feature_map &= ~__cpu_to_le32(MD_FEATURE_PPL | |
1437 | MD_FEATURE_MUTLIPLE_PPLS); | |
1e2b2765 | 1438 | } else if (strcmp(update, "name") == 0) { |
c4f12c13 NB |
1439 | if (info->name[0] == 0) |
1440 | sprintf(info->name, "%d", info->array.md_minor); | |
1441 | memset(sb->set_name, 0, sizeof(sb->set_name)); | |
1442 | if (homehost && | |
1443 | strchr(info->name, ':') == NULL && | |
1444 | strlen(homehost)+1+strlen(info->name) < 32) { | |
1445 | strcpy(sb->set_name, homehost); | |
1446 | strcat(sb->set_name, ":"); | |
1447 | strcat(sb->set_name, info->name); | |
1448 | } else | |
47426701 | 1449 | strncpy(sb->set_name, info->name, sizeof(sb->set_name)); |
1e2b2765 | 1450 | } else if (strcmp(update, "devicesize") == 0 && |
46a533a9 JS |
1451 | __le64_to_cpu(sb->super_offset) < |
1452 | __le64_to_cpu(sb->data_offset)) { | |
bee8ec56 NB |
1453 | /* set data_size to device size less data_offset */ |
1454 | struct misc_dev_info *misc = (struct misc_dev_info*) | |
911cead7 | 1455 | (st->sb + MAX_SB_SIZE + BM_SUPER_SIZE); |
bee8ec56 NB |
1456 | sb->data_size = __cpu_to_le64( |
1457 | misc->device_size - __le64_to_cpu(sb->data_offset)); | |
d5ff855d | 1458 | } else if (strncmp(update, "revert-reshape", 14) == 0) { |
199f1a1f | 1459 | rv = -2; |
46a533a9 JS |
1460 | if (!(sb->feature_map & |
1461 | __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE))) | |
199f1a1f N |
1462 | pr_err("No active reshape to revert on %s\n", |
1463 | devname); | |
1464 | else { | |
199f1a1f | 1465 | __u32 temp; |
a2836f12 N |
1466 | unsigned long long reshape_sectors; |
1467 | long reshape_chunk; | |
1468 | rv = 0; | |
d5ff855d N |
1469 | /* If the reshape hasn't started, just stop it. |
1470 | * It is conceivable that a stripe was modified but | |
1471 | * the metadata not updated. In that case the backup | |
1472 | * should have been used to get passed the critical stage. | |
1473 | * If that couldn't happen, the "-nobackup" version | |
1474 | * will be used. | |
1475 | */ | |
1476 | if (strcmp(update, "revert-reshape-nobackup") == 0 && | |
1477 | sb->reshape_position == 0 && | |
1478 | (__le32_to_cpu(sb->delta_disks) > 0 || | |
1479 | (__le32_to_cpu(sb->delta_disks) == 0 && | |
1480 | !(sb->feature_map & __cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS))))) { | |
1481 | sb->feature_map &= ~__cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE); | |
1482 | sb->raid_disks = __cpu_to_le32(__le32_to_cpu(sb->raid_disks) - | |
1483 | __le32_to_cpu(sb->delta_disks)); | |
1484 | sb->delta_disks = 0; | |
1485 | goto done; | |
1486 | } | |
a2836f12 N |
1487 | /* reshape_position is a little messy. |
1488 | * Its value must be a multiple of the larger | |
1489 | * chunk size, and of the "after" data disks. | |
1490 | * So when reverting we need to change it to | |
1491 | * be a multiple of the new "after" data disks, | |
1492 | * which is the old "before". | |
1493 | * If it isn't already a multiple of 'before', | |
1494 | * the only thing we could do would be | |
1495 | * copy some block around on the disks, which | |
1496 | * is easy to get wrong. | |
1497 | * So we reject a revert-reshape unless the | |
1498 | * alignment is good. | |
1499 | */ | |
efb3994e N |
1500 | if (__le32_to_cpu(sb->level) >= 4 && |
1501 | __le32_to_cpu(sb->level) <= 6) { | |
46a533a9 JS |
1502 | reshape_sectors = |
1503 | __le64_to_cpu(sb->reshape_position); | |
efb3994e | 1504 | reshape_chunk = __le32_to_cpu(sb->new_chunk); |
46a533a9 JS |
1505 | reshape_chunk *= __le32_to_cpu(sb->raid_disks) - |
1506 | __le32_to_cpu(sb->delta_disks) - | |
efb3994e N |
1507 | (__le32_to_cpu(sb->level)==6 ? 2 : 1); |
1508 | if (reshape_sectors % reshape_chunk) { | |
1509 | pr_err("Reshape position is not suitably aligned.\n"); | |
2bf62891 | 1510 | pr_err("Try normal assembly and stop again\n"); |
efb3994e N |
1511 | return -2; |
1512 | } | |
a2836f12 | 1513 | } |
46a533a9 JS |
1514 | sb->raid_disks = |
1515 | __cpu_to_le32(__le32_to_cpu(sb->raid_disks) - | |
1516 | __le32_to_cpu(sb->delta_disks)); | |
199f1a1f N |
1517 | if (sb->delta_disks == 0) |
1518 | sb->feature_map ^= __cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS); | |
1519 | else | |
1520 | sb->delta_disks = __cpu_to_le32(-__le32_to_cpu(sb->delta_disks)); | |
1521 | ||
1522 | temp = sb->new_layout; | |
1523 | sb->new_layout = sb->layout; | |
1524 | sb->layout = temp; | |
1525 | ||
1526 | temp = sb->new_chunk; | |
1527 | sb->new_chunk = sb->chunksize; | |
1528 | sb->chunksize = temp; | |
1529 | ||
46a533a9 JS |
1530 | if (sb->feature_map & |
1531 | __cpu_to_le32(MD_FEATURE_NEW_OFFSET)) { | |
1532 | long offset_delta = | |
1533 | (int32_t)__le32_to_cpu(sb->new_offset); | |
419e0182 N |
1534 | sb->data_offset = __cpu_to_le64(__le64_to_cpu(sb->data_offset) + offset_delta); |
1535 | sb->new_offset = __cpu_to_le32(-offset_delta); | |
1536 | sb->data_size = __cpu_to_le64(__le64_to_cpu(sb->data_size) - offset_delta); | |
199f1a1f | 1537 | } |
d5ff855d | 1538 | done:; |
199f1a1f | 1539 | } |
46a533a9 | 1540 | } else if (strcmp(update, "_reshape_progress") == 0) |
353632d9 | 1541 | sb->reshape_position = __cpu_to_le64(info->reshape_progress); |
46a533a9 | 1542 | else if (strcmp(update, "writemostly") == 0) |
16715c01 | 1543 | sb->devflags |= WriteMostly1; |
46a533a9 | 1544 | else if (strcmp(update, "readwrite") == 0) |
16715c01 | 1545 | sb->devflags &= ~WriteMostly1; |
71574efb N |
1546 | else if (strcmp(update, "failfast") == 0) |
1547 | sb->devflags |= FailFast1; | |
1548 | else if (strcmp(update, "nofailfast") == 0) | |
1549 | sb->devflags &= ~FailFast1; | |
1e2b2765 N |
1550 | else |
1551 | rv = -1; | |
82d9eba6 NB |
1552 | |
1553 | sb->sb_csum = calc_sb_1_csum(sb); | |
81a8a694 GJ |
1554 | if (bms->version == BITMAP_MAJOR_CLUSTERED && dlm_funs_ready()) |
1555 | cluster_release_dlmlock(lockid); | |
d15a1f72 | 1556 | |
82d9eba6 NB |
1557 | return rv; |
1558 | } | |
1559 | ||
3da92f27 | 1560 | static int init_super1(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 1561 | struct shape *s, char *name, char *homehost, |
83cd1e97 | 1562 | int *uuid, unsigned long long data_offset) |
82d9eba6 | 1563 | { |
6416d527 | 1564 | struct mdp_superblock_1 *sb; |
82d9eba6 | 1565 | int spares; |
05697ec1 | 1566 | char defname[10]; |
911cead7 | 1567 | int sbsize; |
6416d527 | 1568 | |
3c0bcd46 | 1569 | if (posix_memalign((void**)&sb, 4096, SUPER1_SIZE) != 0) { |
1ade5cc1 | 1570 | pr_err("could not allocate superblock\n"); |
3d2c4fc7 DW |
1571 | return 0; |
1572 | } | |
1afa9308 | 1573 | memset(sb, 0, SUPER1_SIZE); |
82d9eba6 | 1574 | |
64557c33 | 1575 | st->sb = sb; |
ba7eb04f | 1576 | if (info == NULL) { |
82d9eba6 NB |
1577 | /* zeroing superblock */ |
1578 | return 0; | |
05697ec1 | 1579 | } |
82d9eba6 NB |
1580 | |
1581 | spares = info->working_disks - info->active_disks; | |
911cead7 | 1582 | if (info->raid_disks + spares > MAX_DEVS) { |
e7b84f9d | 1583 | pr_err("too many devices requested: %d+%d > %d\n", |
911cead7 | 1584 | info->raid_disks , spares, MAX_DEVS); |
82d9eba6 NB |
1585 | return 0; |
1586 | } | |
1587 | ||
82d9eba6 NB |
1588 | sb->magic = __cpu_to_le32(MD_SB_MAGIC); |
1589 | sb->major_version = __cpu_to_le32(1); | |
1590 | sb->feature_map = 0; | |
1591 | sb->pad0 = 0; | |
34163fc7 | 1592 | |
350f29f9 NB |
1593 | if (uuid) |
1594 | copy_uuid(sb->set_uuid, uuid, super1.swapuuid); | |
c5f71c24 JS |
1595 | else |
1596 | random_uuid(sb->set_uuid);; | |
82d9eba6 | 1597 | |
05697ec1 NB |
1598 | if (name == NULL || *name == 0) { |
1599 | sprintf(defname, "%d", info->md_minor); | |
1600 | name = defname; | |
1601 | } | |
05697ec1 NB |
1602 | if (homehost && |
1603 | strchr(name, ':')== NULL && | |
1604 | strlen(homehost)+1+strlen(name) < 32) { | |
1605 | strcpy(sb->set_name, homehost); | |
1606 | strcat(sb->set_name, ":"); | |
1607 | strcat(sb->set_name, name); | |
1608 | } else | |
47426701 | 1609 | strncpy(sb->set_name, name, sizeof(sb->set_name)); |
82d9eba6 NB |
1610 | |
1611 | sb->ctime = __cpu_to_le64((unsigned long long)time(0)); | |
1612 | sb->level = __cpu_to_le32(info->level); | |
ea329559 | 1613 | sb->layout = __cpu_to_le32(info->layout); |
5308f117 | 1614 | sb->size = __cpu_to_le64(s->size*2ULL); |
82d9eba6 NB |
1615 | sb->chunksize = __cpu_to_le32(info->chunk_size>>9); |
1616 | sb->raid_disks = __cpu_to_le32(info->raid_disks); | |
1617 | ||
83cd1e97 | 1618 | sb->data_offset = __cpu_to_le64(data_offset); |
82d9eba6 NB |
1619 | sb->data_size = __cpu_to_le64(0); |
1620 | sb->super_offset = __cpu_to_le64(0); | |
1621 | sb->recovery_offset = __cpu_to_le64(0); | |
1622 | ||
1623 | sb->utime = sb->ctime; | |
1624 | sb->events = __cpu_to_le64(1); | |
34321279 | 1625 | if (info->state & (1<<MD_SB_CLEAN)) |
b7528a20 | 1626 | sb->resync_offset = MaxSector; |
82d9eba6 NB |
1627 | else |
1628 | sb->resync_offset = 0; | |
46a533a9 JS |
1629 | sbsize = sizeof(struct mdp_superblock_1) + |
1630 | 2 * (info->raid_disks + spares); | |
911cead7 | 1631 | sbsize = ROUND_UP(sbsize, 512); |
46a533a9 JS |
1632 | sb->max_dev = |
1633 | __cpu_to_le32((sbsize - sizeof(struct mdp_superblock_1)) / 2); | |
82d9eba6 | 1634 | |
46a533a9 JS |
1635 | memset(sb->dev_roles, 0xff, |
1636 | MAX_SB_SIZE - sizeof(struct mdp_superblock_1)); | |
82d9eba6 | 1637 | |
e97a7cd0 AP |
1638 | if (s->consistency_policy == CONSISTENCY_POLICY_PPL) |
1639 | sb->feature_map |= __cpu_to_le32(MD_FEATURE_PPL); | |
1640 | ||
82d9eba6 NB |
1641 | return 1; |
1642 | } | |
1643 | ||
111d01fc NB |
1644 | struct devinfo { |
1645 | int fd; | |
1646 | char *devname; | |
72ca9bcf | 1647 | long long data_offset; |
111d01fc NB |
1648 | mdu_disk_info_t disk; |
1649 | struct devinfo *next; | |
1650 | }; | |
32141c17 | 1651 | |
82d9eba6 | 1652 | /* Add a device to the superblock being created */ |
f20c3968 | 1653 | static int add_to_super1(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf | 1654 | int fd, char *devname, unsigned long long data_offset) |
82d9eba6 | 1655 | { |
3da92f27 | 1656 | struct mdp_superblock_1 *sb = st->sb; |
82d9eba6 | 1657 | __u16 *rp = sb->dev_roles + dk->number; |
111d01fc | 1658 | struct devinfo *di, **dip; |
81a8a694 | 1659 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); |
d15a1f72 | 1660 | int rv, lockid; |
e23c2663 | 1661 | int dk_state; |
d15a1f72 | 1662 | |
81a8a694 GJ |
1663 | if (bms->version == BITMAP_MAJOR_CLUSTERED && dlm_funs_ready()) { |
1664 | rv = cluster_get_dlmlock(&lockid); | |
d15a1f72 | 1665 | if (rv) { |
46a533a9 JS |
1666 | pr_err("Cannot get dlmlock in %s return %d\n", |
1667 | __func__, rv); | |
81a8a694 | 1668 | cluster_release_dlmlock(lockid); |
d15a1f72 GJ |
1669 | return rv; |
1670 | } | |
1671 | } | |
111d01fc | 1672 | |
e23c2663 | 1673 | dk_state = dk->state & ~(1<<MD_DISK_FAILFAST); |
aa313118 GK |
1674 | if ((dk_state & (1<<MD_DISK_ACTIVE)) && |
1675 | (dk_state & (1<<MD_DISK_SYNC)))/* active, sync */ | |
82d9eba6 | 1676 | *rp = __cpu_to_le16(dk->raid_disk); |
e23c2663 | 1677 | else if (dk_state & (1<<MD_DISK_JOURNAL)) |
cc1799c3 | 1678 | *rp = MD_DISK_ROLE_JOURNAL; |
46a533a9 JS |
1679 | else if ((dk_state & ~(1<<MD_DISK_ACTIVE)) == 0) |
1680 | /* active or idle -> spare */ | |
fa7574f6 | 1681 | *rp = MD_DISK_ROLE_SPARE; |
34163fc7 | 1682 | else |
fa7574f6 | 1683 | *rp = MD_DISK_ROLE_FAULTY; |
111d01fc | 1684 | |
f21e18ca | 1685 | if (dk->number >= (int)__le32_to_cpu(sb->max_dev) && |
911cead7 | 1686 | __le32_to_cpu(sb->max_dev) < MAX_DEVS) |
9df04a4f | 1687 | sb->max_dev = __cpu_to_le32(dk->number+1); |
8844e291 | 1688 | |
d2ca6449 | 1689 | sb->dev_number = __cpu_to_le32(dk->number); |
16715c01 | 1690 | sb->devflags = 0; /* don't copy another disks flags */ |
d2ca6449 NB |
1691 | sb->sb_csum = calc_sb_1_csum(sb); |
1692 | ||
111d01fc NB |
1693 | dip = (struct devinfo **)&st->info; |
1694 | while (*dip) | |
1695 | dip = &(*dip)->next; | |
503975b9 | 1696 | di = xmalloc(sizeof(struct devinfo)); |
111d01fc NB |
1697 | di->fd = fd; |
1698 | di->devname = devname; | |
1699 | di->disk = *dk; | |
72ca9bcf | 1700 | di->data_offset = data_offset; |
111d01fc NB |
1701 | di->next = NULL; |
1702 | *dip = di; | |
f20c3968 | 1703 | |
81a8a694 GJ |
1704 | if (bms->version == BITMAP_MAJOR_CLUSTERED && dlm_funs_ready()) |
1705 | cluster_release_dlmlock(lockid); | |
d15a1f72 | 1706 | |
f20c3968 | 1707 | return 0; |
82d9eba6 NB |
1708 | } |
1709 | ||
b138214f | 1710 | static int locate_bitmap1(struct supertype *st, int fd, int node_num); |
14263cf1 | 1711 | |
3da92f27 | 1712 | static int store_super1(struct supertype *st, int fd) |
82d9eba6 | 1713 | { |
3da92f27 | 1714 | struct mdp_superblock_1 *sb = st->sb; |
6fbba4c9 | 1715 | unsigned long long sb_offset; |
308340aa | 1716 | struct align_fd afd; |
82d9eba6 | 1717 | int sbsize; |
5dd497ee | 1718 | unsigned long long dsize; |
81a8a694 | 1719 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); |
d15a1f72 GJ |
1720 | int rv, lockid; |
1721 | ||
81a8a694 GJ |
1722 | if (bms->version == BITMAP_MAJOR_CLUSTERED && dlm_funs_ready()) { |
1723 | rv = cluster_get_dlmlock(&lockid); | |
d15a1f72 | 1724 | if (rv) { |
46a533a9 JS |
1725 | pr_err("Cannot get dlmlock in %s return %d\n", |
1726 | __func__, rv); | |
81a8a694 | 1727 | cluster_release_dlmlock(lockid); |
d15a1f72 GJ |
1728 | return rv; |
1729 | } | |
1730 | } | |
96395475 | 1731 | |
beae1dfe NB |
1732 | if (!get_dev_size(fd, NULL, &dsize)) |
1733 | return 1; | |
1734 | ||
1735 | dsize >>= 9; | |
96395475 | 1736 | |
5dd497ee | 1737 | if (dsize < 24) |
96395475 NB |
1738 | return 2; |
1739 | ||
308340aa JS |
1740 | init_afd(&afd, fd); |
1741 | ||
96395475 NB |
1742 | /* |
1743 | * Calculate the position of the superblock. | |
1744 | * It is always aligned to a 4K boundary and | |
1745 | * depending on minor_version, it can be: | |
1746 | * 0: At least 8K, but less than 12K, from end of device | |
1747 | * 1: At start of device | |
1748 | * 2: 4K from start of device. | |
1749 | */ | |
1750 | switch(st->minor_version) { | |
1751 | case 0: | |
5dd497ee | 1752 | sb_offset = dsize; |
96395475 NB |
1753 | sb_offset -= 8*2; |
1754 | sb_offset &= ~(4*2-1); | |
1755 | break; | |
1756 | case 1: | |
6fbba4c9 | 1757 | sb_offset = 0; |
96395475 NB |
1758 | break; |
1759 | case 2: | |
1760 | sb_offset = 4*2; | |
1761 | break; | |
1762 | default: | |
1763 | return -EINVAL; | |
1764 | } | |
1765 | ||
6fbba4c9 NB |
1766 | if (sb_offset != __le64_to_cpu(sb->super_offset) && |
1767 | 0 != __le64_to_cpu(sb->super_offset) | |
96395475 | 1768 | ) { |
e7b84f9d | 1769 | pr_err("internal error - sb_offset is wrong\n"); |
96395475 NB |
1770 | abort(); |
1771 | } | |
82d9eba6 | 1772 | |
6fbba4c9 | 1773 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) |
82d9eba6 NB |
1774 | return 3; |
1775 | ||
0a2f1894 | 1776 | sbsize = ROUND_UP(sizeof(*sb) + 2 * __le32_to_cpu(sb->max_dev), 512); |
82d9eba6 | 1777 | |
308340aa | 1778 | if (awrite(&afd, sb, sbsize) != sbsize) |
82d9eba6 NB |
1779 | return 4; |
1780 | ||
14263cf1 NB |
1781 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { |
1782 | struct bitmap_super_s *bm = (struct bitmap_super_s*) | |
911cead7 | 1783 | (((char*)sb)+MAX_SB_SIZE); |
14263cf1 | 1784 | if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC) { |
b138214f | 1785 | locate_bitmap1(st, fd, 0); |
308340aa JS |
1786 | if (awrite(&afd, bm, sizeof(*bm)) != sizeof(*bm)) |
1787 | return 5; | |
14263cf1 NB |
1788 | } |
1789 | } | |
570c0542 | 1790 | fsync(fd); |
81a8a694 GJ |
1791 | if (bms->version == BITMAP_MAJOR_CLUSTERED && dlm_funs_ready()) |
1792 | cluster_release_dlmlock(lockid); | |
d15a1f72 | 1793 | |
82d9eba6 NB |
1794 | return 0; |
1795 | } | |
1796 | ||
3da92f27 | 1797 | static int load_super1(struct supertype *st, int fd, char *devname); |
892debc8 | 1798 | |
2fb749d1 NB |
1799 | static unsigned long choose_bm_space(unsigned long devsize) |
1800 | { | |
1801 | /* if the device is bigger than 8Gig, save 64k for bitmap usage, | |
1802 | * if bigger than 200Gig, save 128k | |
a380e275 N |
1803 | * NOTE: result must be multiple of 4K else bad things happen |
1804 | * on 4K-sector devices. | |
2fb749d1 | 1805 | */ |
7eef9be2 JS |
1806 | if (devsize < 64*2) |
1807 | return 0; | |
2fb749d1 NB |
1808 | if (devsize - 64*2 >= 200*1024*1024*2) |
1809 | return 128*2; | |
1810 | if (devsize - 4*2 > 8*1024*1024*2) | |
1811 | return 64*2; | |
1812 | return 4*2; | |
1813 | } | |
1814 | ||
e8090005 N |
1815 | static void free_super1(struct supertype *st); |
1816 | ||
198d5478 | 1817 | __u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len); |
cc1799c3 | 1818 | |
e97a7cd0 AP |
1819 | static int write_init_ppl1(struct supertype *st, struct mdinfo *info, int fd) |
1820 | { | |
1821 | struct mdp_superblock_1 *sb = st->sb; | |
1822 | void *buf; | |
1823 | struct ppl_header *ppl_hdr; | |
1824 | int ret; | |
1825 | ||
b2514242 PB |
1826 | /* first clear entire ppl space */ |
1827 | ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size); | |
1828 | if (ret) | |
1829 | return ret; | |
1830 | ||
e97a7cd0 AP |
1831 | ret = posix_memalign(&buf, 4096, PPL_HEADER_SIZE); |
1832 | if (ret) { | |
1833 | pr_err("Failed to allocate PPL header buffer\n"); | |
1834 | return ret; | |
1835 | } | |
1836 | ||
1837 | memset(buf, 0, PPL_HEADER_SIZE); | |
1838 | ppl_hdr = buf; | |
1839 | memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED); | |
1840 | ppl_hdr->signature = __cpu_to_le32(~crc32c_le(~0, sb->set_uuid, | |
1841 | sizeof(sb->set_uuid))); | |
1842 | ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE)); | |
1843 | ||
1844 | if (lseek64(fd, info->ppl_sector * 512, SEEK_SET) < 0) { | |
1845 | ret = errno; | |
1846 | perror("Failed to seek to PPL header location"); | |
1847 | } | |
1848 | ||
1849 | if (!ret && write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { | |
1850 | ret = errno; | |
1851 | perror("Write PPL header failed"); | |
1852 | } | |
1853 | ||
1854 | if (!ret) | |
1855 | fsync(fd); | |
1856 | ||
1857 | free(buf); | |
1858 | return ret; | |
1859 | } | |
1860 | ||
1861 | #define META_BLOCK_SIZE 4096 | |
1862 | ||
cc1799c3 SL |
1863 | static int write_empty_r5l_meta_block(struct supertype *st, int fd) |
1864 | { | |
1865 | struct r5l_meta_block *mb; | |
1866 | struct mdp_superblock_1 *sb = st->sb; | |
1867 | struct align_fd afd; | |
1868 | __u32 crc; | |
1869 | ||
1870 | init_afd(&afd, fd); | |
1871 | ||
1872 | if (posix_memalign((void**)&mb, 4096, META_BLOCK_SIZE) != 0) { | |
1873 | pr_err("Could not allocate memory for the meta block.\n"); | |
1874 | return 1; | |
1875 | } | |
1876 | ||
1877 | memset(mb, 0, META_BLOCK_SIZE); | |
1878 | ||
1879 | mb->magic = __cpu_to_le32(R5LOG_MAGIC); | |
1880 | mb->version = R5LOG_VERSION; | |
1881 | mb->meta_size = __cpu_to_le32(sizeof(struct r5l_meta_block)); | |
1882 | mb->seq = __cpu_to_le64(random32()); | |
1883 | mb->position = __cpu_to_le64(0); | |
1884 | ||
198d5478 SL |
1885 | crc = crc32c_le(0xffffffff, sb->set_uuid, sizeof(sb->set_uuid)); |
1886 | crc = crc32c_le(crc, (void *)mb, META_BLOCK_SIZE); | |
1887 | mb->checksum = crc; | |
cc1799c3 | 1888 | |
e97a7cd0 | 1889 | if (lseek64(fd, __le64_to_cpu(sb->data_offset) * 512, 0) < 0LL) { |
cc1799c3 SL |
1890 | pr_err("cannot seek to offset of the meta block\n"); |
1891 | goto fail_to_write; | |
1892 | } | |
1893 | ||
1894 | if (awrite(&afd, mb, META_BLOCK_SIZE) != META_BLOCK_SIZE) { | |
1895 | pr_err("failed to store write the meta block \n"); | |
1896 | goto fail_to_write; | |
1897 | } | |
1898 | fsync(fd); | |
1899 | ||
1900 | free(mb); | |
1901 | return 0; | |
1902 | ||
1903 | fail_to_write: | |
1904 | free(mb); | |
1905 | return 1; | |
1906 | } | |
1907 | ||
111d01fc | 1908 | static int write_init_super1(struct supertype *st) |
82d9eba6 | 1909 | { |
3da92f27 | 1910 | struct mdp_superblock_1 *sb = st->sb; |
e8090005 | 1911 | struct supertype *refst; |
111d01fc | 1912 | int rv = 0; |
f21e18ca | 1913 | unsigned long long bm_space; |
111d01fc | 1914 | struct devinfo *di; |
8fac0577 | 1915 | unsigned long long dsize, array_size; |
23bf42cc | 1916 | unsigned long long sb_offset; |
83cd1e97 | 1917 | unsigned long long data_offset; |
1b7eb672 | 1918 | long bm_offset; |
82d9eba6 | 1919 | |
cc1799c3 SL |
1920 | for (di = st->info; di; di = di->next) { |
1921 | if (di->disk.state & (1 << MD_DISK_JOURNAL)) | |
e97a7cd0 | 1922 | sb->feature_map |= __cpu_to_le32(MD_FEATURE_JOURNAL); |
cc1799c3 SL |
1923 | } |
1924 | ||
4687f160 | 1925 | for (di = st->info; di; di = di->next) { |
f4dc5e9b | 1926 | if (di->disk.state & (1 << MD_DISK_FAULTY)) |
111d01fc | 1927 | continue; |
a0c8a17f NB |
1928 | if (di->fd < 0) |
1929 | continue; | |
82d9eba6 | 1930 | |
ba728be7 | 1931 | while (Kill(di->devname, NULL, 0, -1, 1) == 0) |
9277cc77 | 1932 | ; |
82d9eba6 | 1933 | |
111d01fc NB |
1934 | sb->dev_number = __cpu_to_le32(di->disk.number); |
1935 | if (di->disk.state & (1<<MD_DISK_WRITEMOSTLY)) | |
adbb382b | 1936 | sb->devflags |= WriteMostly1; |
9a88e7b6 | 1937 | else |
adbb382b | 1938 | sb->devflags &= ~WriteMostly1; |
71574efb N |
1939 | if (di->disk.state & (1<<MD_DISK_FAILFAST)) |
1940 | sb->devflags |= FailFast1; | |
1941 | else | |
1942 | sb->devflags &= ~FailFast1; | |
892debc8 | 1943 | |
c5f71c24 | 1944 | random_uuid(sb->device_uuid); |
83208785 | 1945 | |
01290056 SL |
1946 | if (!(di->disk.state & (1<<MD_DISK_JOURNAL))) |
1947 | sb->events = 0; | |
111d01fc | 1948 | |
e8090005 | 1949 | refst = dup_super(st); |
1011e834 | 1950 | if (load_super1(refst, di->fd, NULL)==0) { |
e8090005 | 1951 | struct mdp_superblock_1 *refsb = refst->sb; |
111d01fc NB |
1952 | |
1953 | memcpy(sb->device_uuid, refsb->device_uuid, 16); | |
1954 | if (memcmp(sb->set_uuid, refsb->set_uuid, 16)==0) { | |
1955 | /* same array, so preserve events and | |
1956 | * dev_number */ | |
1957 | sb->events = refsb->events; | |
1958 | /* bugs in 2.6.17 and earlier mean the | |
1959 | * dev_number chosen in Manage must be preserved | |
1960 | */ | |
1961 | if (get_linux_version() >= 2006018) | |
1962 | sb->dev_number = refsb->dev_number; | |
1963 | } | |
e8090005 | 1964 | free_super1(refst); |
892debc8 | 1965 | } |
e8090005 | 1966 | free(refst); |
34163fc7 | 1967 | |
40114213 JS |
1968 | if (!get_dev_size(di->fd, NULL, &dsize)) { |
1969 | rv = 1; | |
1970 | goto error_out; | |
1971 | } | |
111d01fc | 1972 | dsize >>= 9; |
82d9eba6 | 1973 | |
111d01fc NB |
1974 | if (dsize < 24) { |
1975 | close(di->fd); | |
40114213 JS |
1976 | rv = 2; |
1977 | goto error_out; | |
111d01fc | 1978 | } |
82d9eba6 | 1979 | |
111d01fc NB |
1980 | /* |
1981 | * Calculate the position of the superblock. | |
1982 | * It is always aligned to a 4K boundary and | |
1983 | * depending on minor_version, it can be: | |
1984 | * 0: At least 8K, but less than 12K, from end of device | |
1985 | * 1: At start of device | |
1986 | * 2: 4K from start of device. | |
23bf42cc | 1987 | * data_offset has already been set. |
111d01fc NB |
1988 | */ |
1989 | array_size = __le64_to_cpu(sb->size); | |
1b7eb672 AP |
1990 | |
1991 | /* work out how much space we left for a bitmap */ | |
1992 | if (sb->feature_map & __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET)) { | |
1993 | bitmap_super_t *bms = (bitmap_super_t *) | |
1994 | (((char *)sb) + MAX_SB_SIZE); | |
1995 | bm_space = calc_bitmap_size(bms, 4096) >> 9; | |
1996 | bm_offset = (long)__le32_to_cpu(sb->bitmap_offset); | |
fa601c2e PB |
1997 | } else if (md_feature_any_ppl_on(sb->feature_map)) { |
1998 | bm_space = MULTIPLE_PPL_AREA_SIZE_SUPER1 >> 9; | |
1999 | if (st->minor_version == 0) | |
e97a7cd0 | 2000 | bm_offset = -bm_space - 8; |
fa601c2e | 2001 | else |
e97a7cd0 | 2002 | bm_offset = 8; |
e97a7cd0 AP |
2003 | sb->ppl.offset = __cpu_to_le16(bm_offset); |
2004 | sb->ppl.size = __cpu_to_le16(bm_space); | |
1b7eb672 AP |
2005 | } else { |
2006 | bm_space = choose_bm_space(array_size); | |
2007 | bm_offset = 8; | |
2008 | } | |
111d01fc | 2009 | |
72ca9bcf | 2010 | data_offset = di->data_offset; |
23bf42cc N |
2011 | if (data_offset == INVALID_SECTORS) |
2012 | data_offset = st->data_offset; | |
111d01fc NB |
2013 | switch(st->minor_version) { |
2014 | case 0: | |
1b7eb672 AP |
2015 | /* Add 8 sectors for bad block log */ |
2016 | bm_space += 8; | |
23bf42cc N |
2017 | if (data_offset == INVALID_SECTORS) |
2018 | data_offset = 0; | |
111d01fc NB |
2019 | sb_offset = dsize; |
2020 | sb_offset -= 8*2; | |
2021 | sb_offset &= ~(4*2-1); | |
4c0ea7b0 | 2022 | sb->data_offset = __cpu_to_le64(data_offset); |
111d01fc | 2023 | sb->super_offset = __cpu_to_le64(sb_offset); |
f21e18ca | 2024 | if (sb_offset < array_size + bm_space) |
a380e275 | 2025 | bm_space = sb_offset - array_size; |
111d01fc | 2026 | sb->data_size = __cpu_to_le64(sb_offset - bm_space); |
bf95d0f3 N |
2027 | if (bm_space >= 8) { |
2028 | sb->bblog_size = __cpu_to_le16(8); | |
2029 | sb->bblog_offset = __cpu_to_le32((unsigned)-8); | |
2030 | } | |
111d01fc NB |
2031 | break; |
2032 | case 1: | |
111d01fc | 2033 | case 2: |
1b7eb672 | 2034 | sb_offset = st->minor_version == 2 ? 8 : 0; |
23bf42cc N |
2035 | sb->super_offset = __cpu_to_le64(sb_offset); |
2036 | if (data_offset == INVALID_SECTORS) | |
1b7eb672 | 2037 | data_offset = sb_offset + 16; |
23bf42cc N |
2038 | |
2039 | sb->data_offset = __cpu_to_le64(data_offset); | |
2040 | sb->data_size = __cpu_to_le64(dsize - data_offset); | |
1b7eb672 | 2041 | if (data_offset >= sb_offset+bm_offset+bm_space+8) { |
f79bbf4f | 2042 | sb->bblog_size = __cpu_to_le16(8); |
1b7eb672 AP |
2043 | sb->bblog_offset = __cpu_to_le32(bm_offset + |
2044 | bm_space); | |
2045 | } else if (data_offset >= sb_offset + 16) { | |
bf95d0f3 | 2046 | sb->bblog_size = __cpu_to_le16(8); |
1b7eb672 | 2047 | /* '8' sectors for the bblog, and 'sb_offset' |
bf95d0f3 N |
2048 | * because we want offset from superblock, not |
2049 | * start of device. | |
2050 | */ | |
1b7eb672 AP |
2051 | sb->bblog_offset = __cpu_to_le32(data_offset - |
2052 | 8 - sb_offset); | |
bf95d0f3 | 2053 | } |
111d01fc NB |
2054 | break; |
2055 | default: | |
7a862a02 | 2056 | pr_err("Failed to write invalid metadata format 1.%i to %s\n", |
e7b84f9d | 2057 | st->minor_version, di->devname); |
40114213 JS |
2058 | rv = -EINVAL; |
2059 | goto out; | |
111d01fc | 2060 | } |
46a533a9 JS |
2061 | /* |
2062 | * Disable badblock log on clusters, or when | |
2063 | * explicitly requested | |
2064 | */ | |
86a406c2 | 2065 | if (st->nodes > 0 || conf_get_create_info()->bblist == 0) { |
e2efe9e7 N |
2066 | sb->bblog_size = 0; |
2067 | sb->bblog_offset = 0; | |
2068 | } | |
82d9eba6 | 2069 | |
111d01fc NB |
2070 | sb->sb_csum = calc_sb_1_csum(sb); |
2071 | rv = store_super1(st, di->fd); | |
cc1799c3 SL |
2072 | |
2073 | if (rv == 0 && (di->disk.state & (1 << MD_DISK_JOURNAL))) { | |
2074 | rv = write_empty_r5l_meta_block(st, di->fd); | |
2075 | if (rv) | |
2076 | goto error_out; | |
2077 | } | |
2078 | ||
e97a7cd0 | 2079 | if (rv == 0 && |
46a533a9 JS |
2080 | (__le32_to_cpu(sb->feature_map) & |
2081 | MD_FEATURE_BITMAP_OFFSET)) { | |
81306e02 | 2082 | rv = st->ss->write_bitmap(st, di->fd, NodeNumUpdate); |
e97a7cd0 | 2083 | } else if (rv == 0 && |
fa601c2e | 2084 | md_feature_any_ppl_on(sb->feature_map)) { |
e97a7cd0 AP |
2085 | struct mdinfo info; |
2086 | ||
2087 | st->ss->getinfo_super(st, &info, NULL); | |
2088 | rv = st->ss->write_init_ppl(st, &info, di->fd); | |
2089 | } | |
2090 | ||
111d01fc NB |
2091 | close(di->fd); |
2092 | di->fd = -1; | |
4687f160 | 2093 | if (rv) |
2094 | goto error_out; | |
111d01fc | 2095 | } |
40114213 JS |
2096 | error_out: |
2097 | if (rv) | |
46a533a9 | 2098 | pr_err("Failed to write metadata to %s\n", di->devname); |
40114213 | 2099 | out: |
82d9eba6 NB |
2100 | return rv; |
2101 | } | |
2102 | ||
64557c33 | 2103 | static int compare_super1(struct supertype *st, struct supertype *tst) |
82d9eba6 NB |
2104 | { |
2105 | /* | |
2106 | * return: | |
2107 | * 0 same, or first was empty, and second was copied | |
2108 | * 1 second had wrong number | |
2109 | * 2 wrong uuid | |
2110 | * 3 wrong other info | |
2111 | */ | |
64557c33 NB |
2112 | struct mdp_superblock_1 *first = st->sb; |
2113 | struct mdp_superblock_1 *second = tst->sb; | |
82d9eba6 NB |
2114 | |
2115 | if (second->magic != __cpu_to_le32(MD_SB_MAGIC)) | |
2116 | return 1; | |
2117 | if (second->major_version != __cpu_to_le32(1)) | |
2118 | return 1; | |
2119 | ||
2120 | if (!first) { | |
3c0bcd46 | 2121 | if (posix_memalign((void**)&first, 4096, SUPER1_SIZE) != 0) { |
1ade5cc1 | 2122 | pr_err("could not allocate superblock\n"); |
3d2c4fc7 DW |
2123 | return 1; |
2124 | } | |
41226756 | 2125 | memcpy(first, second, SUPER1_SIZE); |
64557c33 | 2126 | st->sb = first; |
82d9eba6 NB |
2127 | return 0; |
2128 | } | |
2129 | if (memcmp(first->set_uuid, second->set_uuid, 16)!= 0) | |
2130 | return 2; | |
2131 | ||
2132 | if (first->ctime != second->ctime || | |
2133 | first->level != second->level || | |
2134 | first->layout != second->layout || | |
2135 | first->size != second->size || | |
2136 | first->chunksize != second->chunksize || | |
2137 | first->raid_disks != second->raid_disks) | |
2138 | return 3; | |
2139 | return 0; | |
2140 | } | |
2141 | ||
3da92f27 | 2142 | static int load_super1(struct supertype *st, int fd, char *devname) |
82d9eba6 | 2143 | { |
5dd497ee | 2144 | unsigned long long dsize; |
82d9eba6 NB |
2145 | unsigned long long sb_offset; |
2146 | struct mdp_superblock_1 *super; | |
14263cf1 NB |
2147 | int uuid[4]; |
2148 | struct bitmap_super_s *bsb; | |
bee8ec56 | 2149 | struct misc_dev_info *misc; |
308340aa | 2150 | struct align_fd afd; |
82d9eba6 | 2151 | |
3da92f27 | 2152 | free_super1(st); |
82d9eba6 | 2153 | |
308340aa JS |
2154 | init_afd(&afd, fd); |
2155 | ||
a17a3de3 | 2156 | if (st->ss == NULL || st->minor_version == -1) { |
570c0542 | 2157 | int bestvers = -1; |
23dc1ae8 | 2158 | struct supertype tst; |
570c0542 NB |
2159 | __u64 bestctime = 0; |
2160 | /* guess... choose latest ctime */ | |
ef609477 | 2161 | memset(&tst, 0, sizeof(tst)); |
23dc1ae8 | 2162 | tst.ss = &super1; |
46a533a9 JS |
2163 | for (tst.minor_version = 0; tst.minor_version <= 2; |
2164 | tst.minor_version++) { | |
3da92f27 NB |
2165 | switch(load_super1(&tst, fd, devname)) { |
2166 | case 0: super = tst.sb; | |
570c0542 NB |
2167 | if (bestvers == -1 || |
2168 | bestctime < __le64_to_cpu(super->ctime)) { | |
23dc1ae8 | 2169 | bestvers = tst.minor_version; |
570c0542 NB |
2170 | bestctime = __le64_to_cpu(super->ctime); |
2171 | } | |
2172 | free(super); | |
3da92f27 | 2173 | tst.sb = NULL; |
570c0542 | 2174 | break; |
23dc1ae8 | 2175 | case 1: return 1; /*bad device */ |
82d9eba6 NB |
2176 | case 2: break; /* bad, try next */ |
2177 | } | |
2178 | } | |
570c0542 NB |
2179 | if (bestvers != -1) { |
2180 | int rv; | |
23dc1ae8 NB |
2181 | tst.minor_version = bestvers; |
2182 | tst.ss = &super1; | |
911cead7 | 2183 | tst.max_devs = MAX_DEVS; |
3da92f27 | 2184 | rv = load_super1(&tst, fd, devname); |
23dc1ae8 NB |
2185 | if (rv == 0) |
2186 | *st = tst; | |
570c0542 NB |
2187 | return rv; |
2188 | } | |
82d9eba6 NB |
2189 | return 2; |
2190 | } | |
beae1dfe NB |
2191 | if (!get_dev_size(fd, devname, &dsize)) |
2192 | return 1; | |
2193 | dsize >>= 9; | |
82d9eba6 | 2194 | |
5dd497ee | 2195 | if (dsize < 24) { |
82d9eba6 | 2196 | if (devname) |
e7b84f9d | 2197 | pr_err("%s is too small for md: size is %llu sectors.\n", |
5dd497ee | 2198 | devname, dsize); |
82d9eba6 NB |
2199 | return 1; |
2200 | } | |
2201 | ||
2202 | /* | |
2203 | * Calculate the position of the superblock. | |
2204 | * It is always aligned to a 4K boundary and | |
14263cf1 | 2205 | * depending on minor_version, it can be: |
82d9eba6 NB |
2206 | * 0: At least 8K, but less than 12K, from end of device |
2207 | * 1: At start of device | |
2208 | * 2: 4K from start of device. | |
2209 | */ | |
2210 | switch(st->minor_version) { | |
2211 | case 0: | |
5dd497ee | 2212 | sb_offset = dsize; |
82d9eba6 NB |
2213 | sb_offset -= 8*2; |
2214 | sb_offset &= ~(4*2-1); | |
2215 | break; | |
2216 | case 1: | |
2217 | sb_offset = 0; | |
2218 | break; | |
2219 | case 2: | |
2220 | sb_offset = 4*2; | |
2221 | break; | |
2222 | default: | |
2223 | return -EINVAL; | |
2224 | } | |
2225 | ||
82d9eba6 NB |
2226 | if (lseek64(fd, sb_offset << 9, 0)< 0LL) { |
2227 | if (devname) | |
e7b84f9d | 2228 | pr_err("Cannot seek to superblock on %s: %s\n", |
82d9eba6 NB |
2229 | devname, strerror(errno)); |
2230 | return 1; | |
2231 | } | |
2232 | ||
3c0bcd46 | 2233 | if (posix_memalign((void**)&super, 4096, SUPER1_SIZE) != 0) { |
1ade5cc1 | 2234 | pr_err("could not allocate superblock\n"); |
3d2c4fc7 DW |
2235 | return 1; |
2236 | } | |
82d9eba6 | 2237 | |
1dcee1c9 JS |
2238 | memset(super, 0, SUPER1_SIZE); |
2239 | ||
308340aa | 2240 | if (aread(&afd, super, MAX_SB_SIZE) != MAX_SB_SIZE) { |
82d9eba6 | 2241 | if (devname) |
e7b84f9d | 2242 | pr_err("Cannot read superblock on %s\n", |
82d9eba6 NB |
2243 | devname); |
2244 | free(super); | |
2245 | return 1; | |
2246 | } | |
2247 | ||
2248 | if (__le32_to_cpu(super->magic) != MD_SB_MAGIC) { | |
2249 | if (devname) | |
e7b84f9d | 2250 | pr_err("No super block found on %s (Expected magic %08x, got %08x)\n", |
46a533a9 JS |
2251 | devname, MD_SB_MAGIC, |
2252 | __le32_to_cpu(super->magic)); | |
82d9eba6 NB |
2253 | free(super); |
2254 | return 2; | |
2255 | } | |
2256 | ||
2257 | if (__le32_to_cpu(super->major_version) != 1) { | |
2258 | if (devname) | |
e7b84f9d | 2259 | pr_err("Cannot interpret superblock on %s - version is %d\n", |
82d9eba6 NB |
2260 | devname, __le32_to_cpu(super->major_version)); |
2261 | free(super); | |
2262 | return 2; | |
2263 | } | |
2264 | if (__le64_to_cpu(super->super_offset) != sb_offset) { | |
2265 | if (devname) | |
e7b84f9d | 2266 | pr_err("No superblock found on %s (super_offset is wrong)\n", |
82d9eba6 NB |
2267 | devname); |
2268 | free(super); | |
2269 | return 2; | |
2270 | } | |
64557c33 | 2271 | st->sb = super; |
14263cf1 | 2272 | |
911cead7 | 2273 | bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); |
bee8ec56 | 2274 | |
46a533a9 JS |
2275 | misc = (struct misc_dev_info*) |
2276 | (((char*)super)+MAX_SB_SIZE+BM_SUPER_SIZE); | |
bee8ec56 | 2277 | misc->device_size = dsize; |
23bf42cc N |
2278 | if (st->data_offset == INVALID_SECTORS) |
2279 | st->data_offset = __le64_to_cpu(super->data_offset); | |
bee8ec56 | 2280 | |
14263cf1 NB |
2281 | /* Now check on the bitmap superblock */ |
2282 | if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0) | |
2283 | return 0; | |
2284 | /* Read the bitmap superblock and make sure it looks | |
2285 | * valid. If it doesn't clear the bit. An --assemble --force | |
2286 | * should get that written out. | |
2287 | */ | |
b138214f | 2288 | locate_bitmap1(st, fd, 0); |
308340aa | 2289 | if (aread(&afd, bsb, 512) != 512) |
14263cf1 NB |
2290 | goto no_bitmap; |
2291 | ||
3da92f27 | 2292 | uuid_from_super1(st, uuid); |
14263cf1 NB |
2293 | if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC || |
2294 | memcmp(bsb->uuid, uuid, 16) != 0) | |
2295 | goto no_bitmap; | |
2296 | return 0; | |
2297 | ||
2298 | no_bitmap: | |
c0c1acd6 N |
2299 | super->feature_map = __cpu_to_le32(__le32_to_cpu(super->feature_map) |
2300 | & ~MD_FEATURE_BITMAP_OFFSET); | |
82d9eba6 NB |
2301 | return 0; |
2302 | } | |
2303 | ||
82d9eba6 NB |
2304 | static struct supertype *match_metadata_desc1(char *arg) |
2305 | { | |
503975b9 | 2306 | struct supertype *st = xcalloc(1, sizeof(*st)); |
82d9eba6 | 2307 | |
4dd2df09 | 2308 | st->container_devnm[0] = 0; |
82d9eba6 | 2309 | st->ss = &super1; |
911cead7 | 2310 | st->max_devs = MAX_DEVS; |
64557c33 | 2311 | st->sb = NULL; |
23bf42cc | 2312 | st->data_offset = INVALID_SECTORS; |
9b2a22d3 N |
2313 | /* leading zeros can be safely ignored. --detail generates them. */ |
2314 | while (*arg == '0') | |
2315 | arg++; | |
46a533a9 | 2316 | if (strcmp(arg, "1.0") == 0 || strcmp(arg, "1.00") == 0) { |
82d9eba6 NB |
2317 | st->minor_version = 0; |
2318 | return st; | |
2319 | } | |
46a533a9 | 2320 | if (strcmp(arg, "1.1") == 0 || strcmp(arg, "1.01") == 0 |
7d5c3964 | 2321 | ) { |
82d9eba6 NB |
2322 | st->minor_version = 1; |
2323 | return st; | |
2324 | } | |
9b2a22d3 | 2325 | if (strcmp(arg, "1.2") == 0 || |
26f467a9 | 2326 | #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */ |
ad90adb6 | 2327 | strcmp(arg, "default") == 0 || |
26f467a9 | 2328 | #endif /* DEFAULT_OLD_METADATA */ |
9b2a22d3 | 2329 | strcmp(arg, "1.02") == 0) { |
82d9eba6 NB |
2330 | st->minor_version = 2; |
2331 | return st; | |
2332 | } | |
46a533a9 | 2333 | if (strcmp(arg, "1") == 0 || strcmp(arg, "default") == 0) { |
a17a3de3 DL |
2334 | st->minor_version = -1; |
2335 | return st; | |
2336 | } | |
82d9eba6 NB |
2337 | |
2338 | free(st); | |
2339 | return NULL; | |
2340 | } | |
2341 | ||
34163fc7 NB |
2342 | /* find available size on device with this devsize, using |
2343 | * superblock type st, and reserving 'reserve' sectors for | |
2344 | * a possible bitmap | |
2345 | */ | |
641da745 N |
2346 | static __u64 avail_size1(struct supertype *st, __u64 devsize, |
2347 | unsigned long long data_offset) | |
82d9eba6 | 2348 | { |
2a528478 | 2349 | struct mdp_superblock_1 *super = st->sb; |
387fcd59 | 2350 | int bmspace = 0; |
23bf42cc | 2351 | int bbspace = 0; |
82d9eba6 NB |
2352 | if (devsize < 24) |
2353 | return 0; | |
2354 | ||
e97a7cd0 | 2355 | if (__le32_to_cpu(super->feature_map) & MD_FEATURE_BITMAP_OFFSET) { |
2a528478 N |
2356 | /* hot-add. allow for actual size of bitmap */ |
2357 | struct bitmap_super_s *bsb; | |
911cead7 | 2358 | bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE); |
561ad559 | 2359 | bmspace = calc_bitmap_size(bsb, 4096) >> 9; |
fa601c2e | 2360 | } else if (md_feature_any_ppl_on(super->feature_map)) { |
e97a7cd0 | 2361 | bmspace = __le16_to_cpu(super->ppl.size); |
2a528478 | 2362 | } |
32141c17 | 2363 | |
88af981f | 2364 | /* Allow space for bad block log */ |
641da745 | 2365 | if (super->bblog_size) |
23bf42cc | 2366 | bbspace = __le16_to_cpu(super->bblog_size); |
21e92547 | 2367 | |
a380e275 N |
2368 | if (st->minor_version < 0) |
2369 | /* not specified, so time to set default */ | |
2370 | st->minor_version = 2; | |
387fcd59 | 2371 | |
23bf42cc N |
2372 | if (data_offset == INVALID_SECTORS) |
2373 | data_offset = st->data_offset; | |
2374 | ||
387fcd59 N |
2375 | if (data_offset != INVALID_SECTORS) |
2376 | switch(st->minor_version) { | |
2377 | case 0: | |
23bf42cc | 2378 | return devsize - data_offset - 8*2 - bbspace; |
387fcd59 N |
2379 | case 1: |
2380 | case 2: | |
2381 | return devsize - data_offset; | |
2382 | default: | |
2383 | return 0; | |
2384 | } | |
2385 | ||
2386 | devsize -= bmspace; | |
2387 | ||
34163fc7 NB |
2388 | switch(st->minor_version) { |
2389 | case 0: | |
1bf4e2d9 | 2390 | /* at end */ |
23bf42cc | 2391 | return ((devsize - 8*2 - bbspace ) & ~(4*2-1)); |
34163fc7 | 2392 | case 1: |
1bf4e2d9 | 2393 | /* at start, 4K for superblock and possible bitmap */ |
23bf42cc | 2394 | return devsize - 4*2 - bbspace; |
34163fc7 | 2395 | case 2: |
1bf4e2d9 | 2396 | /* 4k from start, 4K for superblock and possible bitmap */ |
23bf42cc | 2397 | return devsize - (4+4)*2 - bbspace; |
34163fc7 NB |
2398 | } |
2399 | return 0; | |
2400 | } | |
2401 | ||
1bf4e2d9 | 2402 | static int |
3da92f27 | 2403 | add_internal_bitmap1(struct supertype *st, |
199171a2 NB |
2404 | int *chunkp, int delay, int write_behind, |
2405 | unsigned long long size, | |
21e92547 | 2406 | int may_change, int major) |
34163fc7 NB |
2407 | { |
2408 | /* | |
c0c1acd6 N |
2409 | * If not may_change, then this is a 'Grow' without sysfs support for |
2410 | * bitmaps, and the bitmap must fit after the superblock at 1K offset. | |
5c1dec3e | 2411 | * If may_change, then this is create or a Grow with sysfs support, |
c0c1acd6 | 2412 | * and we can put the bitmap wherever we like. |
34163fc7 | 2413 | * |
f9c25f1d | 2414 | * size is in sectors, chunk is in bytes !!! |
34163fc7 NB |
2415 | */ |
2416 | ||
1bf4e2d9 | 2417 | unsigned long long bits; |
199171a2 | 2418 | unsigned long long max_bits; |
34163fc7 | 2419 | unsigned long long min_chunk; |
199171a2 | 2420 | long offset; |
88af981f | 2421 | long bbl_offset, bbl_size; |
f21e18ca | 2422 | unsigned long long chunk = *chunkp; |
37dfc3d6 | 2423 | int room = 0; |
c0c1acd6 | 2424 | int creating = 0; |
4b3eb4d2 | 2425 | int len; |
3da92f27 | 2426 | struct mdp_superblock_1 *sb = st->sb; |
911cead7 | 2427 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE); |
3b7e9d0c | 2428 | int uuid[4]; |
34163fc7 | 2429 | |
c0c1acd6 | 2430 | if (__le64_to_cpu(sb->data_size) == 0) |
46a533a9 JS |
2431 | /* |
2432 | * Must be creating the array, else data_size | |
2433 | * would be non-zero | |
2434 | */ | |
c0c1acd6 | 2435 | creating = 1; |
199171a2 NB |
2436 | switch(st->minor_version) { |
2437 | case 0: | |
46a533a9 JS |
2438 | /* |
2439 | * either 3K after the superblock (when hot-add), | |
a31128d2 | 2440 | * or some amount of space before. |
199171a2 | 2441 | */ |
c0c1acd6 | 2442 | if (creating) { |
46a533a9 JS |
2443 | /* |
2444 | * We are creating array, so we *know* how much room has | |
199171a2 NB |
2445 | * been left. |
2446 | */ | |
2447 | offset = 0; | |
88af981f | 2448 | bbl_size = 8; |
46a533a9 JS |
2449 | room = |
2450 | choose_bm_space(__le64_to_cpu(sb->size)) + bbl_size; | |
199171a2 NB |
2451 | } else { |
2452 | room = __le64_to_cpu(sb->super_offset) | |
2453 | - __le64_to_cpu(sb->data_offset) | |
2454 | - __le64_to_cpu(sb->data_size); | |
88af981f N |
2455 | bbl_size = __le16_to_cpu(sb->bblog_size); |
2456 | if (bbl_size < 8) | |
2457 | bbl_size = 8; | |
2458 | bbl_offset = (__s32)__le32_to_cpu(sb->bblog_offset); | |
2459 | if (bbl_size < -bbl_offset) | |
2460 | bbl_size = -bbl_offset; | |
c0c1acd6 | 2461 | |
46a533a9 JS |
2462 | if (!may_change || |
2463 | (room < 3*2 && __le32_to_cpu(sb->max_dev) <= 384)) { | |
199171a2 NB |
2464 | room = 3*2; |
2465 | offset = 1*2; | |
88af981f | 2466 | bbl_size = 0; |
199171a2 NB |
2467 | } else { |
2468 | offset = 0; /* means movable offset */ | |
2469 | } | |
2470 | } | |
2471 | break; | |
2472 | case 1: | |
2473 | case 2: /* between superblock and data */ | |
c0c1acd6 | 2474 | if (creating) { |
199171a2 | 2475 | offset = 4*2; |
88af981f | 2476 | bbl_size = 8; |
46a533a9 JS |
2477 | room = |
2478 | choose_bm_space(__le64_to_cpu(sb->size)) + bbl_size; | |
199171a2 NB |
2479 | } else { |
2480 | room = __le64_to_cpu(sb->data_offset) | |
2481 | - __le64_to_cpu(sb->super_offset); | |
88af981f N |
2482 | bbl_size = __le16_to_cpu(sb->bblog_size); |
2483 | if (bbl_size) | |
46a533a9 JS |
2484 | room = |
2485 | __le32_to_cpu(sb->bblog_offset) + bbl_size; | |
88af981f N |
2486 | else |
2487 | bbl_size = 8; | |
2488 | ||
c0c1acd6 N |
2489 | if (!may_change) { |
2490 | room -= 2; /* Leave 1K for superblock */ | |
199171a2 | 2491 | offset = 2; |
88af981f | 2492 | bbl_size = 0; |
199171a2 | 2493 | } else { |
c0c1acd6 | 2494 | room -= 4*2; /* leave 4K for superblock */ |
199171a2 NB |
2495 | offset = 4*2; |
2496 | } | |
2497 | } | |
2498 | break; | |
ae491d1e | 2499 | default: |
2ec2b7e9 | 2500 | return -ENOSPC; |
199171a2 | 2501 | } |
1bf4e2d9 | 2502 | |
88af981f | 2503 | room -= bbl_size; |
199171a2 NB |
2504 | if (chunk == UnSet && room > 128*2) |
2505 | /* Limit to 128K of bitmap when chunk size not requested */ | |
2506 | room = 128*2; | |
1bf4e2d9 | 2507 | |
268cccac N |
2508 | if (room <= 1) |
2509 | /* No room for a bitmap */ | |
2ec2b7e9 | 2510 | return -ENOSPC; |
268cccac | 2511 | |
199171a2 | 2512 | max_bits = (room * 512 - sizeof(bitmap_super_t)) * 8; |
34163fc7 NB |
2513 | |
2514 | min_chunk = 4096; /* sub-page chunks don't work yet.. */ | |
f9c25f1d | 2515 | bits = (size*512)/min_chunk +1; |
34163fc7 NB |
2516 | while (bits > max_bits) { |
2517 | min_chunk *= 2; | |
2518 | bits = (bits+1)/2; | |
2519 | } | |
b8ab2a50 N |
2520 | if (chunk == UnSet) { |
2521 | /* For practical purpose, 64Meg is a good | |
2522 | * default chunk size for internal bitmaps. | |
2523 | */ | |
34163fc7 | 2524 | chunk = min_chunk; |
b8ab2a50 N |
2525 | if (chunk < 64*1024*1024) |
2526 | chunk = 64*1024*1024; | |
2527 | } else if (chunk < min_chunk) | |
2ec2b7e9 | 2528 | return -EINVAL; /* chunk size too small */ |
21e92547 | 2529 | if (chunk == 0) /* rounding problem */ |
2ec2b7e9 | 2530 | return -EINVAL; |
34163fc7 | 2531 | |
199171a2 | 2532 | if (offset == 0) { |
a31128d2 N |
2533 | /* start bitmap on a 4K boundary with enough space for |
2534 | * the bitmap | |
2535 | */ | |
199171a2 | 2536 | bits = (size*512) / chunk + 1; |
a31128d2 N |
2537 | room = ((bits+7)/8 + sizeof(bitmap_super_t) +4095)/4096; |
2538 | room *= 8; /* convert 4K blocks to sectors */ | |
88af981f | 2539 | offset = -room - bbl_size; |
199171a2 NB |
2540 | } |
2541 | ||
34a13953 | 2542 | sb->bitmap_offset = (int32_t)__cpu_to_le32(offset); |
34163fc7 | 2543 | |
46a533a9 JS |
2544 | sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | |
2545 | MD_FEATURE_BITMAP_OFFSET); | |
29e766a5 | 2546 | memset(bms, 0, sizeof(*bms)); |
34163fc7 | 2547 | bms->magic = __cpu_to_le32(BITMAP_MAGIC); |
dcec9ee5 | 2548 | bms->version = __cpu_to_le32(major); |
3b7e9d0c LB |
2549 | uuid_from_super1(st, uuid); |
2550 | memcpy(bms->uuid, uuid, 16); | |
34163fc7 NB |
2551 | bms->chunksize = __cpu_to_le32(chunk); |
2552 | bms->daemon_sleep = __cpu_to_le32(delay); | |
f9c25f1d | 2553 | bms->sync_size = __cpu_to_le64(size); |
34163fc7 | 2554 | bms->write_behind = __cpu_to_le32(write_behind); |
529e2aa5 | 2555 | bms->nodes = __cpu_to_le32(st->nodes); |
6d9c7c25 | 2556 | if (st->nodes) |
46a533a9 JS |
2557 | sb->feature_map = __cpu_to_le32(__le32_to_cpu(sb->feature_map) | |
2558 | MD_FEATURE_BITMAP_VERSIONED); | |
4b3eb4d2 JS |
2559 | if (st->cluster_name) { |
2560 | len = sizeof(bms->cluster_name); | |
2561 | strncpy((char *)bms->cluster_name, st->cluster_name, len); | |
2562 | bms->cluster_name[len - 1] = '\0'; | |
2563 | } | |
34163fc7 | 2564 | |
199171a2 | 2565 | *chunkp = chunk; |
2ec2b7e9 | 2566 | return 0; |
34163fc7 NB |
2567 | } |
2568 | ||
b138214f | 2569 | static int locate_bitmap1(struct supertype *st, int fd, int node_num) |
34163fc7 | 2570 | { |
34163fc7 | 2571 | unsigned long long offset; |
1e0d770c NB |
2572 | struct mdp_superblock_1 *sb; |
2573 | int mustfree = 0; | |
53e76b1d | 2574 | int ret; |
34163fc7 | 2575 | |
3da92f27 NB |
2576 | if (!st->sb) { |
2577 | if (st->ss->load_super(st, fd, NULL)) | |
53e76b1d | 2578 | return -1; /* no error I hope... */ |
1e0d770c NB |
2579 | mustfree = 1; |
2580 | } | |
3da92f27 | 2581 | sb = st->sb; |
e478dc86 | 2582 | |
53e76b1d GJ |
2583 | if ((__le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET)) |
2584 | ret = 0; | |
2585 | else | |
2586 | ret = -1; | |
1bf4e2d9 | 2587 | offset = __le64_to_cpu(sb->super_offset); |
b138214f | 2588 | offset += (int32_t) __le32_to_cpu(sb->bitmap_offset) * (node_num + 1); |
1e0d770c | 2589 | if (mustfree) |
f6d75de8 | 2590 | free(sb); |
1bf4e2d9 | 2591 | lseek64(fd, offset<<9, 0); |
53e76b1d | 2592 | return ret; |
34163fc7 NB |
2593 | } |
2594 | ||
0aa2f15b | 2595 | static int write_bitmap1(struct supertype *st, int fd, enum bitmap_update update) |
34163fc7 | 2596 | { |
3da92f27 | 2597 | struct mdp_superblock_1 *sb = st->sb; |
911cead7 | 2598 | bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb)+MAX_SB_SIZE); |
34163fc7 | 2599 | int rv = 0; |
d669228f | 2600 | void *buf; |
4b3eb4d2 | 2601 | int towrite, n, len; |
308340aa | 2602 | struct align_fd afd; |
95a05b37 | 2603 | unsigned int i = 0; |
7e6e839a | 2604 | unsigned long long total_bm_space, bm_space_per_node; |
308340aa | 2605 | |
0aa2f15b GJ |
2606 | switch (update) { |
2607 | case NameUpdate: | |
2608 | /* update cluster name */ | |
2609 | if (st->cluster_name) { | |
4b3eb4d2 JS |
2610 | len = sizeof(bms->cluster_name); |
2611 | memset((char *)bms->cluster_name, 0, len); | |
2612 | strncpy((char *)bms->cluster_name, | |
2613 | st->cluster_name, len); | |
2614 | bms->cluster_name[len - 1] = '\0'; | |
0aa2f15b GJ |
2615 | } |
2616 | break; | |
7e6e839a GJ |
2617 | case NodeNumUpdate: |
2618 | /* cluster md only supports superblock 1.2 now */ | |
46a533a9 JS |
2619 | if (st->minor_version != 2 && |
2620 | bms->version == BITMAP_MAJOR_CLUSTERED) { | |
7e6e839a GJ |
2621 | pr_err("Warning: cluster md only works with superblock 1.2\n"); |
2622 | return -EINVAL; | |
2623 | } | |
2624 | ||
45a87c2f | 2625 | if (bms->version == BITMAP_MAJOR_CLUSTERED) { |
9d67f649 | 2626 | if (__cpu_to_le32(st->nodes) < bms->nodes) { |
46a533a9 JS |
2627 | /* |
2628 | * Since the nodes num is not increased, no | |
2629 | * need to check the space enough or not, | |
2630 | * just update bms->nodes | |
2631 | */ | |
45a87c2f GJ |
2632 | bms->nodes = __cpu_to_le32(st->nodes); |
2633 | break; | |
2634 | } | |
2635 | } else { | |
46a533a9 JS |
2636 | /* |
2637 | * no need to change bms->nodes for other | |
2638 | * bitmap types | |
2639 | */ | |
45a87c2f GJ |
2640 | if (st->nodes) |
2641 | pr_err("Warning: --nodes option is only suitable for clustered bitmap\n"); | |
2642 | break; | |
261b57fe GJ |
2643 | } |
2644 | ||
46a533a9 JS |
2645 | /* |
2646 | * Each node has an independent bitmap, it is necessary to | |
2647 | * calculate the space is enough or not, first get how many | |
2648 | * bytes for the total bitmap | |
2649 | */ | |
7e6e839a GJ |
2650 | bm_space_per_node = calc_bitmap_size(bms, 4096); |
2651 | ||
46a533a9 JS |
2652 | total_bm_space = 512 * (__le64_to_cpu(sb->data_offset) - |
2653 | __le64_to_cpu(sb->super_offset)); | |
2654 | /* leave another 4k for superblock */ | |
2655 | total_bm_space = total_bm_space - 4096; | |
7e6e839a GJ |
2656 | |
2657 | if (bm_space_per_node * st->nodes > total_bm_space) { | |
2658 | pr_err("Warning: The max num of nodes can't exceed %llu\n", | |
2659 | total_bm_space / bm_space_per_node); | |
2660 | return -ENOMEM; | |
2661 | } | |
2662 | ||
2663 | bms->nodes = __cpu_to_le32(st->nodes); | |
2664 | break; | |
0aa2f15b GJ |
2665 | case NoUpdate: |
2666 | default: | |
2667 | break; | |
2668 | } | |
2669 | ||
308340aa | 2670 | init_afd(&afd, fd); |
34163fc7 | 2671 | |
b138214f | 2672 | locate_bitmap1(st, fd, 0); |
34163fc7 | 2673 | |
d669228f JS |
2674 | if (posix_memalign(&buf, 4096, 4096)) |
2675 | return -ENOMEM; | |
2676 | ||
95a05b37 GJ |
2677 | do { |
2678 | /* Only the bitmap[0] should resync | |
2679 | * whole device on initial assembly | |
2680 | */ | |
2681 | if (i) | |
2682 | memset(buf, 0x00, 4096); | |
34163fc7 | 2683 | else |
95a05b37 GJ |
2684 | memset(buf, 0xff, 4096); |
2685 | memcpy(buf, (char *)bms, sizeof(bitmap_super_t)); | |
2686 | ||
119b66a4 GJ |
2687 | /* |
2688 | * use 4096 boundary if bitmap_offset is aligned | |
2689 | * with 8 sectors, then it should compatible with | |
2690 | * older mdadm. | |
2691 | */ | |
2692 | if (__le32_to_cpu(sb->bitmap_offset) & 7) | |
2693 | towrite = calc_bitmap_size(bms, 512); | |
2694 | else | |
2695 | towrite = calc_bitmap_size(bms, 4096); | |
95a05b37 GJ |
2696 | while (towrite > 0) { |
2697 | n = towrite; | |
2698 | if (n > 4096) | |
2699 | n = 4096; | |
2700 | n = awrite(&afd, buf, n); | |
2701 | if (n > 0) | |
2702 | towrite -= n; | |
2703 | else | |
2704 | break; | |
2705 | if (i) | |
2706 | memset(buf, 0x00, 4096); | |
2707 | else | |
2708 | memset(buf, 0xff, 4096); | |
2709 | } | |
2710 | fsync(fd); | |
2711 | if (towrite) { | |
2712 | rv = -2; | |
34163fc7 | 2713 | break; |
95a05b37 GJ |
2714 | } |
2715 | } while (++i < __le32_to_cpu(bms->nodes)); | |
34163fc7 | 2716 | |
d669228f | 2717 | free(buf); |
34163fc7 | 2718 | return rv; |
82d9eba6 NB |
2719 | } |
2720 | ||
3da92f27 | 2721 | static void free_super1(struct supertype *st) |
df37ffc0 | 2722 | { |
d15a1f72 | 2723 | |
3da92f27 NB |
2724 | if (st->sb) |
2725 | free(st->sb); | |
1cc7f4fe N |
2726 | while (st->info) { |
2727 | struct devinfo *di = st->info; | |
2728 | st->info = di->next; | |
2729 | if (di->fd >= 0) | |
2730 | close(di->fd); | |
2731 | free(di); | |
2732 | } | |
3da92f27 | 2733 | st->sb = NULL; |
df37ffc0 NB |
2734 | } |
2735 | ||
17f25ca6 NB |
2736 | static int validate_geometry1(struct supertype *st, int level, |
2737 | int layout, int raiddisks, | |
c21e737b | 2738 | int *chunk, unsigned long long size, |
af4348dd | 2739 | unsigned long long data_offset, |
2c514b71 | 2740 | char *subdev, unsigned long long *freesize, |
5308f117 | 2741 | int consistency_policy, int verbose) |
17f25ca6 | 2742 | { |
641da745 N |
2743 | unsigned long long ldsize, devsize; |
2744 | int bmspace; | |
23bf42cc | 2745 | unsigned long long headroom; |
17f25ca6 NB |
2746 | int fd; |
2747 | ||
b42f577a N |
2748 | if (level == LEVEL_CONTAINER) { |
2749 | if (verbose) | |
e7b84f9d | 2750 | pr_err("1.x metadata does not support containers\n"); |
17f25ca6 | 2751 | return 0; |
b42f577a | 2752 | } |
7ccc4cc4 | 2753 | if (*chunk == UnSet) |
bb7295f1 N |
2754 | *chunk = DEFAULT_CHUNK; |
2755 | ||
17f25ca6 NB |
2756 | if (!subdev) |
2757 | return 1; | |
2758 | ||
23bf42cc N |
2759 | if (st->minor_version < 0) |
2760 | /* not specified, so time to set default */ | |
2761 | st->minor_version = 2; | |
2762 | ||
17f25ca6 NB |
2763 | fd = open(subdev, O_RDONLY|O_EXCL, 0); |
2764 | if (fd < 0) { | |
2c514b71 | 2765 | if (verbose) |
e7b84f9d | 2766 | pr_err("super1.x cannot open %s: %s\n", |
2c514b71 | 2767 | subdev, strerror(errno)); |
17f25ca6 NB |
2768 | return 0; |
2769 | } | |
2c514b71 | 2770 | |
17f25ca6 NB |
2771 | if (!get_dev_size(fd, subdev, &ldsize)) { |
2772 | close(fd); | |
2773 | return 0; | |
2774 | } | |
2775 | close(fd); | |
2776 | ||
641da745 N |
2777 | devsize = ldsize >> 9; |
2778 | if (devsize < 24) { | |
2779 | *freesize = 0; | |
2780 | return 0; | |
2781 | } | |
2782 | ||
e97a7cd0 | 2783 | /* creating: allow suitable space for bitmap or PPL */ |
fa601c2e PB |
2784 | if (consistency_policy == CONSISTENCY_POLICY_PPL) |
2785 | bmspace = MULTIPLE_PPL_AREA_SIZE_SUPER1 >> 9; | |
2786 | else | |
2787 | bmspace = choose_bm_space(devsize); | |
641da745 | 2788 | |
23bf42cc N |
2789 | if (data_offset == INVALID_SECTORS) |
2790 | data_offset = st->data_offset; | |
2791 | if (data_offset == INVALID_SECTORS) | |
2792 | switch (st->minor_version) { | |
641da745 | 2793 | case 0: |
23bf42cc N |
2794 | data_offset = 0; |
2795 | break; | |
641da745 N |
2796 | case 1: |
2797 | case 2: | |
23bf42cc N |
2798 | /* Choose data offset appropriate for this device |
2799 | * and use as default for whole array. | |
2800 | * The data_offset must allow for bitmap space | |
2801 | * and base metadata, should allow for some headroom | |
2802 | * for reshape, and should be rounded to multiple | |
2803 | * of 1M. | |
2804 | * Headroom is limited to 128M, but aim for about 0.1% | |
641da745 | 2805 | */ |
23bf42cc | 2806 | headroom = 128*1024*2; |
641da745 N |
2807 | while ((headroom << 10) > devsize && |
2808 | (*chunk == 0 || | |
2809 | headroom / 2 >= ((unsigned)(*chunk)*2)*2)) | |
2810 | headroom >>= 1; | |
23bf42cc N |
2811 | data_offset = 12*2 + bmspace + headroom; |
2812 | #define ONE_MEG (2*1024) | |
795bd44e | 2813 | data_offset = ROUND_UP(data_offset, ONE_MEG); |
641da745 N |
2814 | break; |
2815 | } | |
23bf42cc N |
2816 | if (st->data_offset == INVALID_SECTORS) |
2817 | st->data_offset = data_offset; | |
2818 | switch(st->minor_version) { | |
2819 | case 0: /* metadata at end. Round down and subtract space to reserve */ | |
2820 | devsize = (devsize & ~(4ULL*2-1)); | |
e97a7cd0 | 2821 | /* space for metadata, bblog, bitmap/ppl */ |
4c0ea7b0 | 2822 | devsize -= 8*2 + 8 + bmspace; |
23bf42cc N |
2823 | break; |
2824 | case 1: | |
2825 | case 2: | |
2826 | devsize -= data_offset; | |
2827 | break; | |
641da745 | 2828 | } |
23bf42cc | 2829 | *freesize = devsize; |
17f25ca6 NB |
2830 | return 1; |
2831 | } | |
2832 | ||
afa368f4 N |
2833 | void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0) |
2834 | { | |
2835 | /* Create a v1.0 superblock based on 'info'*/ | |
2836 | void *ret; | |
2837 | struct mdp_superblock_1 *sb; | |
2838 | int i; | |
afa368f4 N |
2839 | unsigned long long offset; |
2840 | ||
2841 | if (posix_memalign(&ret, 4096, 1024) != 0) | |
2842 | return NULL; | |
2843 | sb = ret; | |
2844 | memset(ret, 0, 1024); | |
2845 | sb->magic = __cpu_to_le32(MD_SB_MAGIC); | |
2846 | sb->major_version = __cpu_to_le32(1); | |
2847 | ||
2848 | copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid); | |
2849 | sprintf(sb->set_name, "%d", sb0->md_minor); | |
2850 | sb->ctime = __cpu_to_le32(info->array.ctime+1); | |
2851 | sb->level = __cpu_to_le32(info->array.level); | |
2852 | sb->layout = __cpu_to_le32(info->array.layout); | |
2853 | sb->size = __cpu_to_le64(info->component_size); | |
2854 | sb->chunksize = __cpu_to_le32(info->array.chunk_size/512); | |
2855 | sb->raid_disks = __cpu_to_le32(info->array.raid_disks); | |
ccec2685 N |
2856 | if (info->array.level > 0) |
2857 | sb->data_size = sb->size; | |
2858 | else | |
2859 | sb->data_size = st->ss->avail_size(st, st->devsize/512, 0); | |
afa368f4 N |
2860 | sb->resync_offset = MaxSector; |
2861 | sb->max_dev = __cpu_to_le32(MD_SB_DISKS); | |
2862 | sb->dev_number = __cpu_to_le32(info->disk.number); | |
2863 | sb->utime = __cpu_to_le64(info->array.utime); | |
2864 | ||
2865 | offset = st->devsize/512 - 8*2; | |
2866 | offset &= ~(4*2-1); | |
2867 | sb->super_offset = __cpu_to_le64(offset); | |
2868 | //*(__u64*)(st->other + 128 + 8 + 8) = __cpu_to_le64(offset); | |
2869 | ||
c5f71c24 | 2870 | random_uuid(sb->device_uuid); |
afa368f4 N |
2871 | |
2872 | for (i = 0; i < MD_SB_DISKS; i++) { | |
2873 | int state = sb0->disks[i].state; | |
fa7574f6 | 2874 | sb->dev_roles[i] = MD_DISK_ROLE_SPARE; |
afa368f4 N |
2875 | if ((state & (1<<MD_DISK_SYNC)) && |
2876 | !(state & (1<<MD_DISK_FAULTY))) | |
2877 | sb->dev_roles[i] = __cpu_to_le16(sb0->disks[i].raid_disk); | |
2878 | } | |
2879 | sb->sb_csum = calc_sb_1_csum(sb); | |
2880 | return ret; | |
2881 | } | |
2882 | ||
82d9eba6 NB |
2883 | struct superswitch super1 = { |
2884 | .examine_super = examine_super1, | |
2885 | .brief_examine_super = brief_examine_super1, | |
0d726f17 | 2886 | .export_examine_super = export_examine_super1, |
82d9eba6 NB |
2887 | .detail_super = detail_super1, |
2888 | .brief_detail_super = brief_detail_super1, | |
0d726f17 | 2889 | .export_detail_super = export_detail_super1, |
111d01fc | 2890 | .write_init_super = write_init_super1, |
0e600426 N |
2891 | .validate_geometry = validate_geometry1, |
2892 | .add_to_super = add_to_super1, | |
6d388a88 | 2893 | .examine_badblocks = examine_badblocks_super1, |
74db60b0 | 2894 | .copy_metadata = copy_metadata1, |
e97a7cd0 | 2895 | .write_init_ppl = write_init_ppl1, |
83b6208e | 2896 | .match_home = match_home1, |
82d9eba6 NB |
2897 | .uuid_from_super = uuid_from_super1, |
2898 | .getinfo_super = getinfo_super1, | |
00bbdbda | 2899 | .container_content = container_content1, |
82d9eba6 | 2900 | .update_super = update_super1, |
82d9eba6 | 2901 | .init_super = init_super1, |
82d9eba6 | 2902 | .store_super = store_super1, |
82d9eba6 NB |
2903 | .compare_super = compare_super1, |
2904 | .load_super = load_super1, | |
2905 | .match_metadata_desc = match_metadata_desc1, | |
2906 | .avail_size = avail_size1, | |
34163fc7 NB |
2907 | .add_internal_bitmap = add_internal_bitmap1, |
2908 | .locate_bitmap = locate_bitmap1, | |
2909 | .write_bitmap = write_bitmap1, | |
df37ffc0 | 2910 | .free_super = free_super1, |
f277ce36 NB |
2911 | #if __BYTE_ORDER == BIG_ENDIAN |
2912 | .swapuuid = 0, | |
2913 | #else | |
2914 | .swapuuid = 1, | |
2915 | #endif | |
31015d57 | 2916 | .name = "1.x", |
82d9eba6 | 2917 | }; |