]>
Commit | Line | Data |
---|---|---|
c82f047c NB |
1 | /* |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
4 | * Copyright (C) 2004 Paul Clements, SteelEye Technology, Inc. | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License as published by | |
8 | * the Free Software Foundation; either version 2 of the License, or | |
9 | * (at your option) any later version. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License | |
17 | * along with this program; if not, write to the Free Software | |
18 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
19 | */ | |
20 | ||
c82f047c | 21 | #include "mdadm.h" |
c82f047c | 22 | |
d8f82d1d | 23 | static inline void sb_le_to_cpu(bitmap_super_t *sb) |
c82f047c NB |
24 | { |
25 | sb->magic = __le32_to_cpu(sb->magic); | |
26 | sb->version = __le32_to_cpu(sb->version); | |
27 | /* uuid gets no translation */ | |
28 | sb->events = __le64_to_cpu(sb->events); | |
29 | sb->events_cleared = __le64_to_cpu(sb->events_cleared); | |
30 | sb->state = __le32_to_cpu(sb->state); | |
31 | sb->chunksize = __le32_to_cpu(sb->chunksize); | |
32 | sb->daemon_sleep = __le32_to_cpu(sb->daemon_sleep); | |
33 | sb->sync_size = __le64_to_cpu(sb->sync_size); | |
10ae45c1 | 34 | sb->write_behind = __le32_to_cpu(sb->write_behind); |
95a05b37 GJ |
35 | sb->nodes = __le32_to_cpu(sb->nodes); |
36 | sb->sectors_reserved = __le32_to_cpu(sb->sectors_reserved); | |
c82f047c NB |
37 | } |
38 | ||
d8f82d1d | 39 | static inline void sb_cpu_to_le(bitmap_super_t *sb) |
c82f047c NB |
40 | { |
41 | sb_le_to_cpu(sb); /* these are really the same thing */ | |
42 | } | |
43 | ||
44 | mapping_t bitmap_states[] = { | |
45 | { "OK", 0 }, | |
46 | { "Out of date", 2 }, | |
47 | { NULL, -1 } | |
48 | }; | |
49 | ||
6e88b3b3 | 50 | static const char *bitmap_state(int state_num) |
c82f047c NB |
51 | { |
52 | char *state = map_num(bitmap_states, state_num); | |
53 | return state ? state : "Unknown"; | |
54 | } | |
55 | ||
6e88b3b3 | 56 | static const char *human_chunksize(unsigned long bytes) |
c82f047c NB |
57 | { |
58 | static char buf[16]; | |
59 | char *suffixes[] = { "B", "KB", "MB", "GB", "TB", NULL }; | |
60 | int i = 0; | |
61 | ||
62 | while (bytes >> 10) { | |
63 | bytes >>= 10; | |
64 | i++; | |
65 | } | |
66 | ||
8f23b0b3 | 67 | snprintf(buf, sizeof(buf), "%lu %s", bytes, suffixes[i]); |
c82f047c NB |
68 | |
69 | return buf; | |
70 | } | |
71 | ||
72 | typedef struct bitmap_info_s { | |
73 | bitmap_super_t sb; | |
74 | unsigned long long total_bits; | |
75 | unsigned long long dirty_bits; | |
76 | } bitmap_info_t; | |
77 | ||
78 | /* count the dirty bits in the first num_bits of byte */ | |
d8f82d1d | 79 | static inline int count_dirty_bits_byte(char byte, int num_bits) |
c82f047c NB |
80 | { |
81 | int num = 0; | |
82 | ||
83 | switch (num_bits) { /* fall through... */ | |
84 | case 8: if (byte & 128) num++; | |
85 | case 7: if (byte & 64) num++; | |
86 | case 6: if (byte & 32) num++; | |
87 | case 5: if (byte & 16) num++; | |
88 | case 4: if (byte & 8) num++; | |
89 | case 3: if (byte & 4) num++; | |
90 | case 2: if (byte & 2) num++; | |
91 | case 1: if (byte & 1) num++; | |
92 | default: break; | |
93 | } | |
94 | ||
95 | return num; | |
96 | } | |
97 | ||
6e88b3b3 | 98 | static int count_dirty_bits(char *buf, int num_bits) |
c82f047c NB |
99 | { |
100 | int i, num = 0; | |
101 | ||
ca3b6696 | 102 | for (i = 0; i < num_bits / 8; i++) |
c82f047c NB |
103 | num += count_dirty_bits_byte(buf[i], 8); |
104 | ||
105 | if (num_bits % 8) /* not an even byte boundary */ | |
106 | num += count_dirty_bits_byte(buf[i], num_bits % 8); | |
107 | ||
108 | return num; | |
109 | } | |
110 | ||
6e88b3b3 | 111 | static bitmap_info_t *bitmap_fd_read(int fd, int brief) |
c82f047c | 112 | { |
4ccdb956 NB |
113 | /* Note: fd might be open O_DIRECT, so we must be |
114 | * careful to align reads properly | |
115 | */ | |
c82f047c NB |
116 | unsigned long long total_bits = 0, read_bits = 0, dirty_bits = 0; |
117 | bitmap_info_t *info; | |
6416d527 | 118 | void *buf; |
f21e18ca | 119 | unsigned int n, skip; |
4ccdb956 | 120 | |
a7322ae1 | 121 | if (posix_memalign(&buf, 4096, 8192) != 0) { |
e7b84f9d | 122 | pr_err("failed to allocate 8192 bytes\n"); |
3d2c4fc7 DW |
123 | return NULL; |
124 | } | |
4ccdb956 | 125 | n = read(fd, buf, 8192); |
c82f047c | 126 | |
503975b9 | 127 | info = xmalloc(sizeof(*info)); |
c82f047c | 128 | |
4ccdb956 | 129 | if (n < sizeof(info->sb)) { |
7a862a02 | 130 | pr_err("failed to read superblock of bitmap file: %s\n", strerror(errno)); |
c82f047c | 131 | free(info); |
39c74d5e | 132 | free(buf); |
c82f047c NB |
133 | return NULL; |
134 | } | |
4ccdb956 NB |
135 | memcpy(&info->sb, buf, sizeof(info->sb)); |
136 | skip = sizeof(info->sb); | |
c82f047c NB |
137 | |
138 | sb_le_to_cpu(&info->sb); /* convert superblock to CPU byte ordering */ | |
aba69144 | 139 | |
b47dff66 | 140 | if (brief || info->sb.sync_size == 0 || info->sb.chunksize == 0) |
c82f047c NB |
141 | goto out; |
142 | ||
143 | /* read the rest of the file counting total bits and dirty bits -- | |
144 | * we stop when either: | |
145 | * 1) we hit EOF, in which case we assume the rest of the bits (if any) | |
146 | * are dirty | |
147 | * 2) we've read the full bitmap, in which case we ignore any trailing | |
148 | * data in the file | |
149 | */ | |
150 | total_bits = bitmap_bits(info->sb.sync_size, info->sb.chunksize); | |
151 | ||
4ccdb956 | 152 | while(read_bits < total_bits) { |
c82f047c NB |
153 | unsigned long long remaining = total_bits - read_bits; |
154 | ||
4ccdb956 NB |
155 | if (n == 0) { |
156 | n = read(fd, buf, 8192); | |
157 | skip = 0; | |
158 | if (n <= 0) | |
159 | break; | |
160 | } | |
161 | if (remaining > (n-skip) * 8) /* we want the full buffer */ | |
162 | remaining = (n-skip) * 8; | |
163 | ||
164 | dirty_bits += count_dirty_bits(buf+skip, remaining); | |
c82f047c NB |
165 | |
166 | read_bits += remaining; | |
4ccdb956 | 167 | n = 0; |
c82f047c NB |
168 | } |
169 | ||
170 | if (read_bits < total_bits) { /* file truncated... */ | |
7a862a02 | 171 | pr_err("WARNING: bitmap file is not large enough for array size %llu!\n\n", |
f9c25f1d | 172 | (unsigned long long)info->sb.sync_size); |
c82f047c NB |
173 | total_bits = read_bits; |
174 | } | |
175 | out: | |
39c74d5e | 176 | free(buf); |
c82f047c NB |
177 | info->total_bits = total_bits; |
178 | info->dirty_bits = dirty_bits; | |
179 | return info; | |
180 | } | |
181 | ||
6e88b3b3 JS |
182 | static int |
183 | bitmap_file_open(char *filename, struct supertype **stp, int node_num) | |
c82f047c NB |
184 | { |
185 | int fd; | |
8cc56e8b | 186 | struct stat stb; |
34163fc7 | 187 | struct supertype *st = *stp; |
c82f047c | 188 | |
9ca0de62 JS |
189 | fd = open(filename, O_RDONLY|O_DIRECT); |
190 | if (fd < 0) { | |
191 | pr_err("failed to open bitmap file %s: %s\n", | |
192 | filename, strerror(errno)); | |
1e781e07 | 193 | return -1; |
c82f047c | 194 | } |
9ca0de62 | 195 | |
8cc56e8b ZL |
196 | if (fstat(fd, &stb) < 0) { |
197 | pr_err("fstat failed for %s: %s\n", filename, strerror(errno)); | |
198 | close(fd); | |
199 | return -1; | |
200 | } | |
201 | if ((stb.st_mode & S_IFMT) == S_IFBLK) { | |
55935d51 | 202 | /* block device, so we are probably after an internal bitmap */ |
9ca0de62 JS |
203 | if (!st) |
204 | st = guess_super(fd); | |
55935d51 NB |
205 | if (!st) { |
206 | /* just look at device... */ | |
207 | lseek(fd, 0, 0); | |
ebeb3663 | 208 | } else if (!st->ss->locate_bitmap) { |
e7b84f9d | 209 | pr_err("No bitmap possible with %s metadata\n", |
ebeb3663 | 210 | st->ss->name); |
de12cdc7 | 211 | close(fd); |
1e781e07 | 212 | return -1; |
53e76b1d | 213 | } else { |
b138214f | 214 | if (st->ss->locate_bitmap(st, fd, node_num)) { |
53e76b1d | 215 | pr_err("%s doesn't have bitmap\n", filename); |
de12cdc7 | 216 | close(fd); |
53e76b1d GJ |
217 | fd = -1; |
218 | } | |
219 | } | |
34163fc7 | 220 | *stp = st; |
55935d51 | 221 | } |
c82f047c | 222 | |
1e781e07 | 223 | return fd; |
c82f047c NB |
224 | } |
225 | ||
6e88b3b3 | 226 | static __u32 swapl(__u32 l) |
34163fc7 NB |
227 | { |
228 | char *c = (char*)&l; | |
229 | char t= c[0]; | |
230 | c[0] = c[3]; | |
231 | c[3] = t; | |
232 | ||
233 | t = c[1]; | |
234 | c[1] = c[2]; | |
235 | c[2] = t; | |
236 | return l; | |
237 | } | |
55935d51 | 238 | int ExamineBitmap(char *filename, int brief, struct supertype *st) |
c82f047c NB |
239 | { |
240 | /* | |
241 | * Read the bitmap file and display its contents | |
242 | */ | |
243 | ||
244 | bitmap_super_t *sb; | |
245 | bitmap_info_t *info; | |
246 | int rv = 1; | |
dfd4d8ee | 247 | char buf[64]; |
bf4fb153 | 248 | int swap; |
b98043a2 | 249 | int fd, i; |
caa0f6c6 | 250 | __u32 uuid32[4]; |
c82f047c | 251 | |
b138214f | 252 | fd = bitmap_file_open(filename, &st, 0); |
1e781e07 | 253 | if (fd < 0) |
c82f047c NB |
254 | return rv; |
255 | ||
1e781e07 N |
256 | info = bitmap_fd_read(fd, brief); |
257 | if (!info) | |
258 | return rv; | |
c82f047c | 259 | sb = &info->sb; |
5d89b18d | 260 | if (sb->magic != BITMAP_MAGIC) { |
1e781e07 N |
261 | pr_err("This is an md array. To view a bitmap you need to examine\n"); |
262 | pr_err("a member device, not the array.\n"); | |
263 | pr_err("Reporting bitmap that would be used if this array were used\n"); | |
264 | pr_err("as a member of some other array\n"); | |
265 | } | |
266 | close(fd); | |
c82f047c NB |
267 | printf(" Filename : %s\n", filename); |
268 | printf(" Magic : %08x\n", sb->magic); | |
269 | if (sb->magic != BITMAP_MAGIC) { | |
1e781e07 N |
270 | pr_err("invalid bitmap magic 0x%x, the bitmap file appears\n", |
271 | sb->magic); | |
272 | pr_err("to be corrupted or missing.\n"); | |
c82f047c NB |
273 | } |
274 | printf(" Version : %d\n", sb->version); | |
dcec9ee5 | 275 | if (sb->version < BITMAP_MAJOR_LO || |
6d9c7c25 | 276 | sb->version > BITMAP_MAJOR_CLUSTERED) { |
1e781e07 N |
277 | pr_err("unknown bitmap version %d, either the bitmap file\n", |
278 | sb->version); | |
279 | pr_err("is corrupted or you need to upgrade your tools\n"); | |
c82f047c NB |
280 | goto free_info; |
281 | } | |
282 | ||
283 | rv = 0; | |
bf4fb153 NB |
284 | if (st) |
285 | swap = st->ss->swapuuid; | |
286 | else | |
287 | #if __BYTE_ORDER == BIG_ENDIAN | |
288 | swap = 0; | |
289 | #else | |
290 | swap = 1; | |
291 | #endif | |
caa0f6c6 N |
292 | memcpy(uuid32, sb->uuid, 16); |
293 | if (swap) | |
294 | printf(" UUID : %08x:%08x:%08x:%08x\n", | |
295 | swapl(uuid32[0]), | |
296 | swapl(uuid32[1]), | |
297 | swapl(uuid32[2]), | |
298 | swapl(uuid32[3])); | |
299 | else | |
300 | printf(" UUID : %08x:%08x:%08x:%08x\n", | |
301 | uuid32[0], | |
302 | uuid32[1], | |
303 | uuid32[2], | |
304 | uuid32[3]); | |
305 | ||
b98043a2 GJ |
306 | if (sb->nodes == 0) { |
307 | printf(" Events : %llu\n", (unsigned long long)sb->events); | |
308 | printf(" Events Cleared : %llu\n", (unsigned long long)sb->events_cleared); | |
309 | printf(" State : %s\n", bitmap_state(sb->state)); | |
310 | ||
311 | } | |
312 | ||
c82f047c NB |
313 | printf(" Chunksize : %s\n", human_chunksize(sb->chunksize)); |
314 | printf(" Daemon : %ds flush period\n", sb->daemon_sleep); | |
dfd4d8ee NB |
315 | if (sb->write_behind) |
316 | sprintf(buf, "Allow write behind, max %d", sb->write_behind); | |
317 | else | |
318 | sprintf(buf, "Normal"); | |
319 | printf(" Write Mode : %s\n", buf); | |
f9c25f1d | 320 | printf(" Sync Size : %llu%s\n", (unsigned long long)sb->sync_size/2, |
f277ce36 | 321 | human_size(sb->sync_size * 512)); |
b98043a2 GJ |
322 | |
323 | if (sb->nodes == 0) { | |
324 | if (brief) | |
325 | goto free_info; | |
326 | printf(" Bitmap : %llu bits (chunks), %llu dirty (%2.1f%%)\n", | |
327 | info->total_bits, info->dirty_bits, | |
328 | 100.0 * info->dirty_bits / (info->total_bits?:1)); | |
329 | } else { | |
330 | printf(" Cluster nodes : %d\n", sb->nodes); | |
2cf42394 | 331 | printf(" Cluster name : %-64s\n", sb->cluster_name); |
b98043a2 | 332 | for (i = 0; i < (int)sb->nodes; i++) { |
b138214f GJ |
333 | st = NULL; |
334 | free(info); | |
335 | fd = bitmap_file_open(filename, &st, i); | |
34996a5f JS |
336 | if (fd < 0) { |
337 | printf(" Unable to open bitmap file on node: %i\n", i); | |
338 | ||
339 | continue; | |
340 | } | |
b138214f | 341 | info = bitmap_fd_read(fd, brief); |
34996a5f JS |
342 | if (!info) { |
343 | close(fd); | |
344 | printf(" Unable to read bitmap on node: %i\n", i); | |
345 | continue; | |
346 | } | |
b138214f | 347 | sb = &info->sb; |
b98043a2 GJ |
348 | if (sb->magic != BITMAP_MAGIC) |
349 | pr_err("invalid bitmap magic 0x%x, the bitmap file appears to be corrupted\n", sb->magic); | |
350 | ||
351 | printf(" Node Slot : %d\n", i); | |
352 | printf(" Events : %llu\n", | |
353 | (unsigned long long)sb->events); | |
354 | printf(" Events Cleared : %llu\n", | |
355 | (unsigned long long)sb->events_cleared); | |
356 | printf(" State : %s\n", bitmap_state(sb->state)); | |
357 | if (brief) | |
358 | continue; | |
359 | printf(" Bitmap : %llu bits (chunks), %llu dirty (%2.1f%%)\n", | |
360 | info->total_bits, info->dirty_bits, | |
361 | 100.0 * info->dirty_bits / (info->total_bits?:1)); | |
b138214f | 362 | close(fd); |
b98043a2 GJ |
363 | } |
364 | } | |
365 | ||
c82f047c NB |
366 | free_info: |
367 | free(info); | |
368 | return rv; | |
369 | } | |
370 | ||
371 | int CreateBitmap(char *filename, int force, char uuid[16], | |
1bf4e2d9 NB |
372 | unsigned long chunksize, unsigned long daemon_sleep, |
373 | unsigned long write_behind, | |
dcec9ee5 NB |
374 | unsigned long long array_size /* sectors */, |
375 | int major) | |
c82f047c NB |
376 | { |
377 | /* | |
378 | * Create a bitmap file with a superblock and (optionally) a full bitmap | |
379 | */ | |
380 | ||
381 | FILE *fp; | |
382 | int rv = 1; | |
383 | char block[512]; | |
384 | bitmap_super_t sb; | |
385 | long long bytes, filesize; | |
386 | ||
387 | if (!force && access(filename, F_OK) == 0) { | |
e7b84f9d | 388 | pr_err("bitmap file %s already exists, use --force to overwrite\n", filename); |
c82f047c NB |
389 | return rv; |
390 | } | |
391 | ||
392 | fp = fopen(filename, "w"); | |
393 | if (fp == NULL) { | |
e7b84f9d | 394 | pr_err("failed to open bitmap file %s: %s\n", |
c82f047c NB |
395 | filename, strerror(errno)); |
396 | return rv; | |
397 | } | |
398 | ||
1bfdbe01 NB |
399 | if (chunksize == UnSet) { |
400 | /* We don't want more than 2^21 chunks, as 2^11 fill up one | |
401 | * 4K page (2 bytes per chunk), and 2^10 address of those | |
b39827de NB |
402 | * fill up a 4K indexing page. 2^20 might be safer, especially |
403 | * on 64bit hosts, so use that. | |
1bfdbe01 NB |
404 | */ |
405 | chunksize = DEFAULT_BITMAP_CHUNK; | |
b39827de | 406 | /* <<20 for 2^20 chunks, >>9 to convert bytes to sectors */ |
3e416561 | 407 | while (array_size > ((unsigned long long)chunksize << (20-9))) |
1bfdbe01 NB |
408 | chunksize <<= 1; |
409 | } | |
410 | ||
c82f047c NB |
411 | memset(&sb, 0, sizeof(sb)); |
412 | sb.magic = BITMAP_MAGIC; | |
dcec9ee5 | 413 | sb.version = major; |
c82f047c NB |
414 | if (uuid != NULL) |
415 | memcpy(sb.uuid, uuid, 16); | |
416 | sb.chunksize = chunksize; | |
417 | sb.daemon_sleep = daemon_sleep; | |
dfd4d8ee | 418 | sb.write_behind = write_behind; |
c82f047c NB |
419 | sb.sync_size = array_size; |
420 | ||
421 | sb_cpu_to_le(&sb); /* convert to on-disk byte ordering */ | |
422 | ||
423 | if (fwrite(&sb, sizeof(sb), 1, fp) != 1) { | |
e7b84f9d | 424 | pr_err("failed to write superblock to bitmap file %s: %s\n", filename, strerror(errno)); |
c82f047c NB |
425 | goto out; |
426 | } | |
427 | ||
428 | /* calculate the size of the bitmap and write it to disk */ | |
429 | bytes = (bitmap_bits(array_size, chunksize) + 7) / 8; | |
430 | if (!bytes) { | |
431 | rv = 0; | |
432 | goto out; | |
433 | } | |
434 | ||
435 | filesize = bytes + sizeof(sb); | |
436 | ||
437 | memset(block, 0xff, sizeof(block)); | |
438 | ||
439 | while (bytes > 0) { | |
440 | if (fwrite(block, sizeof(block), 1, fp) != 1) { | |
e7b84f9d | 441 | pr_err("failed to write bitmap file %s: %s\n", filename, strerror(errno)); |
c82f047c NB |
442 | goto out; |
443 | } | |
444 | bytes -= sizeof(block); | |
445 | } | |
aba69144 | 446 | |
c82f047c | 447 | rv = 0; |
97a6748c | 448 | fflush(fp); |
c82f047c | 449 | /* make the file be the right size (well, to the nearest byte) */ |
1e0d770c NB |
450 | if (ftruncate(fileno(fp), filesize)) |
451 | perror("ftrunace"); | |
c82f047c NB |
452 | out: |
453 | fclose(fp); | |
454 | if (rv) | |
455 | unlink(filename); /* possibly corrupted, better get rid of it */ | |
456 | return rv; | |
457 | } | |
8131b493 | 458 | |
bf4fb153 | 459 | int bitmap_update_uuid(int fd, int *uuid, int swap) |
8131b493 NB |
460 | { |
461 | struct bitmap_super_s bm; | |
9fca7d62 NB |
462 | if (lseek(fd, 0, 0) != 0) |
463 | return 1; | |
8131b493 | 464 | if (read(fd, &bm, sizeof(bm)) != sizeof(bm)) |
9fca7d62 | 465 | return 1; |
8131b493 | 466 | if (bm.magic != __cpu_to_le32(BITMAP_MAGIC)) |
9fca7d62 | 467 | return 1; |
350f29f9 | 468 | copy_uuid(bm.uuid, uuid, swap); |
9fca7d62 NB |
469 | if (lseek(fd, 0, 0) != 0) |
470 | return 2; | |
471 | if (write(fd, &bm, sizeof(bm)) != sizeof(bm)) { | |
472 | lseek(fd, 0, 0); | |
473 | return 2; | |
474 | } | |
8131b493 | 475 | lseek(fd, 0, 0); |
9fca7d62 | 476 | return 0; |
8131b493 | 477 | } |