]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
4 | * Copyright (C) 2001-2012 Neil Brown <neilb@suse.de> | |
5 | * | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License as published by | |
9 | * the Free Software Foundation; either version 2 of the License, or | |
10 | * (at your option) any later version. | |
11 | * | |
12 | * This program is distributed in the hope that it will be useful, | |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | * GNU General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with this program; if not, write to the Free Software | |
19 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
20 | * | |
21 | * Author: Neil Brown | |
22 | * Email: <neilb@suse.de> | |
23 | */ | |
24 | ||
25 | #include "mdadm.h" | |
26 | #include "md_p.h" | |
27 | #include <sys/socket.h> | |
28 | #include <sys/utsname.h> | |
29 | #include <sys/wait.h> | |
30 | #include <sys/un.h> | |
31 | #include <ctype.h> | |
32 | #include <dirent.h> | |
33 | #include <signal.h> | |
34 | ||
35 | int __offroot; | |
36 | ||
37 | /* | |
38 | * following taken from linux/blkpg.h because they aren't | |
39 | * anywhere else and it isn't safe to #include linux/ * stuff. | |
40 | */ | |
41 | ||
42 | #define BLKPG _IO(0x12,105) | |
43 | ||
44 | /* The argument structure */ | |
45 | struct blkpg_ioctl_arg { | |
46 | int op; | |
47 | int flags; | |
48 | int datalen; | |
49 | void *data; | |
50 | }; | |
51 | ||
52 | /* The subfunctions (for the op field) */ | |
53 | #define BLKPG_ADD_PARTITION 1 | |
54 | #define BLKPG_DEL_PARTITION 2 | |
55 | ||
56 | /* Sizes of name fields. Unused at present. */ | |
57 | #define BLKPG_DEVNAMELTH 64 | |
58 | #define BLKPG_VOLNAMELTH 64 | |
59 | ||
60 | /* The data structure for ADD_PARTITION and DEL_PARTITION */ | |
61 | struct blkpg_partition { | |
62 | long long start; /* starting offset in bytes */ | |
63 | long long length; /* length in bytes */ | |
64 | int pno; /* partition number */ | |
65 | char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2, | |
66 | to be used in kernel messages */ | |
67 | char volname[BLKPG_VOLNAMELTH]; /* volume label */ | |
68 | }; | |
69 | ||
70 | #include "part.h" | |
71 | ||
72 | /* Force a compilation error if condition is true */ | |
73 | #define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition)) | |
74 | ||
75 | /* Force a compilation error if condition is true, but also produce a | |
76 | result (of value 0 and type size_t), so the expression can be used | |
77 | e.g. in a structure initializer (or where-ever else comma expressions | |
78 | aren't permitted). */ | |
79 | #define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); })) | |
80 | ||
81 | /* | |
82 | * Parse a 128 bit uuid in 4 integers | |
83 | * format is 32 hexx nibbles with options :.<space> separator | |
84 | * If not exactly 32 hex digits are found, return 0 | |
85 | * else return 1 | |
86 | */ | |
87 | int parse_uuid(char *str, int uuid[4]) | |
88 | { | |
89 | int hit = 0; /* number of Hex digIT */ | |
90 | int i; | |
91 | char c; | |
92 | for (i = 0; i < 4; i++) | |
93 | uuid[i] = 0; | |
94 | ||
95 | while ((c = *str++) != 0) { | |
96 | int n; | |
97 | if (c >= '0' && c <= '9') | |
98 | n = c-'0'; | |
99 | else if (c >= 'a' && c <= 'f') | |
100 | n = 10 + c - 'a'; | |
101 | else if (c >= 'A' && c <= 'F') | |
102 | n = 10 + c - 'A'; | |
103 | else if (strchr(":. -", c)) | |
104 | continue; | |
105 | else return 0; | |
106 | ||
107 | if (hit<32) { | |
108 | uuid[hit/8] <<= 4; | |
109 | uuid[hit/8] += n; | |
110 | } | |
111 | hit++; | |
112 | } | |
113 | if (hit == 32) | |
114 | return 1; | |
115 | return 0; | |
116 | } | |
117 | ||
118 | /* | |
119 | * Get the md version number. | |
120 | * We use the RAID_VERSION ioctl if it is supported | |
121 | * If not, but we have a block device with major '9', we assume | |
122 | * 0.36.0 | |
123 | * | |
124 | * Return version number as 24 but number - assume version parts | |
125 | * always < 255 | |
126 | */ | |
127 | ||
128 | int md_get_version(int fd) | |
129 | { | |
130 | struct stat stb; | |
131 | mdu_version_t vers; | |
132 | ||
133 | if (fstat(fd, &stb)<0) | |
134 | return -1; | |
135 | if ((S_IFMT&stb.st_mode) != S_IFBLK) | |
136 | return -1; | |
137 | ||
138 | if (ioctl(fd, RAID_VERSION, &vers) == 0) | |
139 | return (vers.major*10000) + (vers.minor*100) + vers.patchlevel; | |
140 | if (errno == EACCES) | |
141 | return -1; | |
142 | if (major(stb.st_rdev) == MD_MAJOR) | |
143 | return (3600); | |
144 | return -1; | |
145 | } | |
146 | ||
147 | int get_linux_version() | |
148 | { | |
149 | struct utsname name; | |
150 | char *cp; | |
151 | int a = 0, b = 0,c = 0; | |
152 | if (uname(&name) <0) | |
153 | return -1; | |
154 | ||
155 | cp = name.release; | |
156 | a = strtoul(cp, &cp, 10); | |
157 | if (*cp == '.') | |
158 | b = strtoul(cp+1, &cp, 10); | |
159 | if (*cp == '.') | |
160 | c = strtoul(cp+1, &cp, 10); | |
161 | ||
162 | return (a*1000000)+(b*1000)+c; | |
163 | } | |
164 | ||
165 | #ifndef MDASSEMBLE | |
166 | int mdadm_version(char *version) | |
167 | { | |
168 | int a, b, c; | |
169 | char *cp; | |
170 | ||
171 | if (!version) | |
172 | version = Version; | |
173 | ||
174 | cp = strchr(version, '-'); | |
175 | if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v') | |
176 | return -1; | |
177 | cp += 3; | |
178 | a = strtoul(cp, &cp, 10); | |
179 | if (*cp != '.') | |
180 | return -1; | |
181 | b = strtoul(cp+1, &cp, 10); | |
182 | if (*cp == '.') | |
183 | c = strtoul(cp+1, &cp, 10); | |
184 | else | |
185 | c = 0; | |
186 | if (*cp != ' ' && *cp != '-') | |
187 | return -1; | |
188 | return (a*1000000)+(b*1000)+c; | |
189 | } | |
190 | ||
191 | unsigned long long parse_size(char *size) | |
192 | { | |
193 | /* parse 'size' which should be a number optionally | |
194 | * followed by 'K', 'M', or 'G'. | |
195 | * Without a suffix, K is assumed. | |
196 | * Number returned is in sectors (half-K) | |
197 | * INVALID_SECTORS returned on error. | |
198 | */ | |
199 | char *c; | |
200 | long long s = strtoll(size, &c, 10); | |
201 | if (s > 0) { | |
202 | switch (*c) { | |
203 | case 'K': | |
204 | c++; | |
205 | default: | |
206 | s *= 2; | |
207 | break; | |
208 | case 'M': | |
209 | c++; | |
210 | s *= 1024 * 2; | |
211 | break; | |
212 | case 'G': | |
213 | c++; | |
214 | s *= 1024 * 1024 * 2; | |
215 | break; | |
216 | case 's': /* sectors */ | |
217 | c++; | |
218 | break; | |
219 | } | |
220 | } else | |
221 | s = INVALID_SECTORS; | |
222 | if (*c) | |
223 | s = INVALID_SECTORS; | |
224 | return s; | |
225 | } | |
226 | ||
227 | int parse_layout_10(char *layout) | |
228 | { | |
229 | int copies, rv; | |
230 | char *cp; | |
231 | /* Parse the layout string for raid10 */ | |
232 | /* 'f', 'o' or 'n' followed by a number <= raid_disks */ | |
233 | if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') || | |
234 | (copies = strtoul(layout+1, &cp, 10)) < 1 || | |
235 | copies > 200 || | |
236 | *cp) | |
237 | return -1; | |
238 | if (layout[0] == 'n') | |
239 | rv = 256 + copies; | |
240 | else if (layout[0] == 'o') | |
241 | rv = 0x10000 + (copies<<8) + 1; | |
242 | else | |
243 | rv = 1 + (copies<<8); | |
244 | return rv; | |
245 | } | |
246 | ||
247 | int parse_layout_faulty(char *layout) | |
248 | { | |
249 | /* Parse the layout string for 'faulty' */ | |
250 | int ln = strcspn(layout, "0123456789"); | |
251 | char *m = xstrdup(layout); | |
252 | int mode; | |
253 | m[ln] = 0; | |
254 | mode = map_name(faultylayout, m); | |
255 | if (mode == UnSet) | |
256 | return -1; | |
257 | ||
258 | return mode | (atoi(layout+ln)<< ModeShift); | |
259 | } | |
260 | ||
261 | long parse_num(char *num) | |
262 | { | |
263 | /* Either return a valid number, or -1 */ | |
264 | char *c; | |
265 | long rv = strtol(num, &c, 10); | |
266 | if (rv < 0 || *c || !num[0]) | |
267 | return -1; | |
268 | else | |
269 | return rv; | |
270 | } | |
271 | #endif | |
272 | ||
273 | void remove_partitions(int fd) | |
274 | { | |
275 | /* remove partitions from this block devices. | |
276 | * This is used for components added to an array | |
277 | */ | |
278 | #ifdef BLKPG_DEL_PARTITION | |
279 | struct blkpg_ioctl_arg a; | |
280 | struct blkpg_partition p; | |
281 | ||
282 | a.op = BLKPG_DEL_PARTITION; | |
283 | a.data = (void*)&p; | |
284 | a.datalen = sizeof(p); | |
285 | a.flags = 0; | |
286 | memset(a.data, 0, a.datalen); | |
287 | for (p.pno = 0; p.pno < 16; p.pno++) | |
288 | ioctl(fd, BLKPG, &a); | |
289 | #endif | |
290 | } | |
291 | ||
292 | int test_partition(int fd) | |
293 | { | |
294 | /* Check if fd is a whole-disk or a partition. | |
295 | * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION | |
296 | * will return ENXIO on an invalid partition number. | |
297 | */ | |
298 | struct blkpg_ioctl_arg a; | |
299 | struct blkpg_partition p; | |
300 | a.op = BLKPG_DEL_PARTITION; | |
301 | a.data = (void*)&p; | |
302 | a.datalen = sizeof(p); | |
303 | a.flags = 0; | |
304 | memset(a.data, 0, a.datalen); | |
305 | p.pno = 1<<30; | |
306 | if (ioctl(fd, BLKPG, &a) == 0) | |
307 | /* Very unlikely, but not a partition */ | |
308 | return 0; | |
309 | if (errno == ENXIO) | |
310 | /* not a partition */ | |
311 | return 0; | |
312 | ||
313 | return 1; | |
314 | } | |
315 | ||
316 | int test_partition_from_id(dev_t id) | |
317 | { | |
318 | char buf[20]; | |
319 | int fd, rv; | |
320 | ||
321 | sprintf(buf, "%d:%d", major(id), minor(id)); | |
322 | fd = dev_open(buf, O_RDONLY); | |
323 | if (fd < 0) | |
324 | return -1; | |
325 | rv = test_partition(fd); | |
326 | close(fd); | |
327 | return rv; | |
328 | } | |
329 | ||
330 | int enough(int level, int raid_disks, int layout, int clean, char *avail) | |
331 | { | |
332 | int copies, first; | |
333 | int i; | |
334 | int avail_disks = 0; | |
335 | ||
336 | for (i = 0; i < raid_disks; i++) | |
337 | avail_disks += !!avail[i]; | |
338 | ||
339 | switch (level) { | |
340 | case 10: | |
341 | /* This is the tricky one - we need to check | |
342 | * which actual disks are present. | |
343 | */ | |
344 | copies = (layout&255)* ((layout>>8) & 255); | |
345 | first = 0; | |
346 | do { | |
347 | /* there must be one of the 'copies' form 'first' */ | |
348 | int n = copies; | |
349 | int cnt = 0; | |
350 | int this = first; | |
351 | while (n--) { | |
352 | if (avail[this]) | |
353 | cnt++; | |
354 | this = (this+1) % raid_disks; | |
355 | } | |
356 | if (cnt == 0) | |
357 | return 0; | |
358 | first = (first+(layout&255)) % raid_disks; | |
359 | } while (first != 0); | |
360 | return 1; | |
361 | ||
362 | case LEVEL_MULTIPATH: | |
363 | return avail_disks>= 1; | |
364 | case LEVEL_LINEAR: | |
365 | case 0: | |
366 | return avail_disks == raid_disks; | |
367 | case 1: | |
368 | return avail_disks >= 1; | |
369 | case 4: | |
370 | case 5: | |
371 | if (clean) | |
372 | return avail_disks >= raid_disks-1; | |
373 | else | |
374 | return avail_disks >= raid_disks; | |
375 | case 6: | |
376 | if (clean) | |
377 | return avail_disks >= raid_disks-2; | |
378 | else | |
379 | return avail_disks >= raid_disks; | |
380 | default: | |
381 | return 0; | |
382 | } | |
383 | } | |
384 | ||
385 | int enough_fd(int fd) | |
386 | { | |
387 | struct mdu_array_info_s array; | |
388 | struct mdu_disk_info_s disk; | |
389 | int avail_disks = 0; | |
390 | int i, rv; | |
391 | char *avail; | |
392 | ||
393 | if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 || | |
394 | array.raid_disks <= 0) | |
395 | return 0; | |
396 | avail = xcalloc(array.raid_disks, 1); | |
397 | for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) { | |
398 | disk.number = i; | |
399 | if (ioctl(fd, GET_DISK_INFO, &disk) != 0) | |
400 | continue; | |
401 | if (disk.major == 0 && disk.minor == 0) | |
402 | continue; | |
403 | array.nr_disks--; | |
404 | ||
405 | if (! (disk.state & (1<<MD_DISK_SYNC))) | |
406 | continue; | |
407 | if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks) | |
408 | continue; | |
409 | avail_disks++; | |
410 | avail[disk.raid_disk] = 1; | |
411 | } | |
412 | /* This is used on an active array, so assume it is clean */ | |
413 | rv = enough(array.level, array.raid_disks, array.layout, | |
414 | 1, avail); | |
415 | free(avail); | |
416 | return rv; | |
417 | } | |
418 | ||
419 | const int uuid_zero[4] = { 0, 0, 0, 0 }; | |
420 | ||
421 | int same_uuid(int a[4], int b[4], int swapuuid) | |
422 | { | |
423 | if (swapuuid) { | |
424 | /* parse uuids are hostendian. | |
425 | * uuid's from some superblocks are big-ending | |
426 | * if there is a difference, we need to swap.. | |
427 | */ | |
428 | unsigned char *ac = (unsigned char *)a; | |
429 | unsigned char *bc = (unsigned char *)b; | |
430 | int i; | |
431 | for (i = 0; i < 16; i += 4) { | |
432 | if (ac[i+0] != bc[i+3] || | |
433 | ac[i+1] != bc[i+2] || | |
434 | ac[i+2] != bc[i+1] || | |
435 | ac[i+3] != bc[i+0]) | |
436 | return 0; | |
437 | } | |
438 | return 1; | |
439 | } else { | |
440 | if (a[0]==b[0] && | |
441 | a[1]==b[1] && | |
442 | a[2]==b[2] && | |
443 | a[3]==b[3]) | |
444 | return 1; | |
445 | return 0; | |
446 | } | |
447 | } | |
448 | ||
449 | void copy_uuid(void *a, int b[4], int swapuuid) | |
450 | { | |
451 | if (swapuuid) { | |
452 | /* parse uuids are hostendian. | |
453 | * uuid's from some superblocks are big-ending | |
454 | * if there is a difference, we need to swap.. | |
455 | */ | |
456 | unsigned char *ac = (unsigned char *)a; | |
457 | unsigned char *bc = (unsigned char *)b; | |
458 | int i; | |
459 | for (i = 0; i < 16; i += 4) { | |
460 | ac[i+0] = bc[i+3]; | |
461 | ac[i+1] = bc[i+2]; | |
462 | ac[i+2] = bc[i+1]; | |
463 | ac[i+3] = bc[i+0]; | |
464 | } | |
465 | } else | |
466 | memcpy(a, b, 16); | |
467 | } | |
468 | ||
469 | char *__fname_from_uuid(int id[4], int swap, char *buf, char sep) | |
470 | { | |
471 | int i, j; | |
472 | char uuid[16]; | |
473 | char *c = buf; | |
474 | strcpy(c, "UUID-"); | |
475 | c += strlen(c); | |
476 | copy_uuid(uuid, id, swap); | |
477 | for (i = 0; i < 4; i++) { | |
478 | if (i) | |
479 | *c++ = sep; | |
480 | for (j = 3; j >= 0; j--) { | |
481 | sprintf(c,"%02x", (unsigned char) uuid[j+4*i]); | |
482 | c+= 2; | |
483 | } | |
484 | } | |
485 | return buf; | |
486 | ||
487 | } | |
488 | ||
489 | char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep) | |
490 | { | |
491 | // dirty hack to work around an issue with super1 superblocks... | |
492 | // super1 superblocks need swapuuid set in order for assembly to | |
493 | // work, but can't have it set if we want this printout to match | |
494 | // all the other uuid printouts in super1.c, so we force swapuuid | |
495 | // to 1 to make our printout match the rest of super1 | |
496 | return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep); | |
497 | } | |
498 | ||
499 | #ifndef MDASSEMBLE | |
500 | int check_ext2(int fd, char *name) | |
501 | { | |
502 | /* | |
503 | * Check for an ext2fs file system. | |
504 | * Superblock is always 1K at 1K offset | |
505 | * | |
506 | * s_magic is le16 at 56 == 0xEF53 | |
507 | * report mtime - le32 at 44 | |
508 | * blocks - le32 at 4 | |
509 | * logblksize - le32 at 24 | |
510 | */ | |
511 | unsigned char sb[1024]; | |
512 | time_t mtime; | |
513 | int size, bsize; | |
514 | if (lseek(fd, 1024,0)!= 1024) | |
515 | return 0; | |
516 | if (read(fd, sb, 1024)!= 1024) | |
517 | return 0; | |
518 | if (sb[56] != 0x53 || sb[57] != 0xef) | |
519 | return 0; | |
520 | ||
521 | mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8; | |
522 | bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8; | |
523 | size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8; | |
524 | pr_err("%s appears to contain an ext2fs file system\n", | |
525 | name); | |
526 | fprintf(stderr," size=%dK mtime=%s", | |
527 | size*(1<<bsize), ctime(&mtime)); | |
528 | return 1; | |
529 | } | |
530 | ||
531 | int check_reiser(int fd, char *name) | |
532 | { | |
533 | /* | |
534 | * superblock is at 64K | |
535 | * size is 1024; | |
536 | * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52 | |
537 | * | |
538 | */ | |
539 | unsigned char sb[1024]; | |
540 | unsigned long size; | |
541 | if (lseek(fd, 64*1024, 0) != 64*1024) | |
542 | return 0; | |
543 | if (read(fd, sb, 1024) != 1024) | |
544 | return 0; | |
545 | if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 && | |
546 | strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0) | |
547 | return 0; | |
548 | pr_err("%s appears to contain a reiserfs file system\n",name); | |
549 | size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8; | |
550 | fprintf(stderr, " size = %luK\n", size*4); | |
551 | ||
552 | return 1; | |
553 | } | |
554 | ||
555 | int check_raid(int fd, char *name) | |
556 | { | |
557 | struct mdinfo info; | |
558 | time_t crtime; | |
559 | char *level; | |
560 | struct supertype *st = guess_super(fd); | |
561 | ||
562 | if (!st) return 0; | |
563 | st->ignore_hw_compat = 1; | |
564 | st->ss->load_super(st, fd, name); | |
565 | /* Looks like a raid array .. */ | |
566 | pr_err("%s appears to be part of a raid array:\n", | |
567 | name); | |
568 | st->ss->getinfo_super(st, &info, NULL); | |
569 | st->ss->free_super(st); | |
570 | crtime = info.array.ctime; | |
571 | level = map_num(pers, info.array.level); | |
572 | if (!level) level = "-unknown-"; | |
573 | fprintf(stderr, " level=%s devices=%d ctime=%s", | |
574 | level, info.array.raid_disks, ctime(&crtime)); | |
575 | return 1; | |
576 | } | |
577 | ||
578 | int ask(char *mesg) | |
579 | { | |
580 | char *add = ""; | |
581 | int i; | |
582 | for (i = 0; i < 5; i++) { | |
583 | char buf[100]; | |
584 | fprintf(stderr, "%s%s", mesg, add); | |
585 | fflush(stderr); | |
586 | if (fgets(buf, 100, stdin)==NULL) | |
587 | return 0; | |
588 | if (buf[0]=='y' || buf[0]=='Y') | |
589 | return 1; | |
590 | if (buf[0]=='n' || buf[0]=='N') | |
591 | return 0; | |
592 | add = "(y/n) "; | |
593 | } | |
594 | pr_err("assuming 'no'\n"); | |
595 | return 0; | |
596 | } | |
597 | #endif /* MDASSEMBLE */ | |
598 | ||
599 | int is_standard(char *dev, int *nump) | |
600 | { | |
601 | /* tests if dev is a "standard" md dev name. | |
602 | * i.e if the last component is "/dNN" or "/mdNN", | |
603 | * where NN is a string of digits | |
604 | * Returns 1 if a partitionable standard, | |
605 | * -1 if non-partitonable, | |
606 | * 0 if not a standard name. | |
607 | */ | |
608 | char *d = strrchr(dev, '/'); | |
609 | int type = 0; | |
610 | int num; | |
611 | if (!d) | |
612 | return 0; | |
613 | if (strncmp(d, "/d",2) == 0) | |
614 | d += 2, type = 1; /* /dev/md/dN{pM} */ | |
615 | else if (strncmp(d, "/md_d", 5) == 0) | |
616 | d += 5, type = 1; /* /dev/md_dN{pM} */ | |
617 | else if (strncmp(d, "/md", 3) == 0) | |
618 | d += 3, type = -1; /* /dev/mdN */ | |
619 | else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0) | |
620 | d += 1, type = -1; /* /dev/md/N */ | |
621 | else | |
622 | return 0; | |
623 | if (!*d) | |
624 | return 0; | |
625 | num = atoi(d); | |
626 | while (isdigit(*d)) | |
627 | d++; | |
628 | if (*d) | |
629 | return 0; | |
630 | if (nump) *nump = num; | |
631 | ||
632 | return type; | |
633 | } | |
634 | ||
635 | unsigned long calc_csum(void *super, int bytes) | |
636 | { | |
637 | unsigned long long newcsum = 0; | |
638 | int i; | |
639 | unsigned int csum; | |
640 | unsigned int *superc = (unsigned int*) super; | |
641 | ||
642 | for(i = 0; i < bytes/4; i++) | |
643 | newcsum += superc[i]; | |
644 | csum = (newcsum& 0xffffffff) + (newcsum>>32); | |
645 | #ifdef __alpha__ | |
646 | /* The in-kernel checksum calculation is always 16bit on | |
647 | * the alpha, though it is 32 bit on i386... | |
648 | * I wonder what it is elsewhere... (it uses an API in | |
649 | * a way that it shouldn't). | |
650 | */ | |
651 | csum = (csum & 0xffff) + (csum >> 16); | |
652 | csum = (csum & 0xffff) + (csum >> 16); | |
653 | #endif | |
654 | return csum; | |
655 | } | |
656 | ||
657 | #ifndef MDASSEMBLE | |
658 | char *human_size(long long bytes) | |
659 | { | |
660 | static char buf[30]; | |
661 | ||
662 | /* We convert bytes to either centi-M{ega,ibi}bytes or | |
663 | * centi-G{igi,ibi}bytes, with appropriate rounding, | |
664 | * and then print 1/100th of those as a decimal. | |
665 | * We allow upto 2048Megabytes before converting to | |
666 | * gigabytes, as that shows more precision and isn't | |
667 | * too large a number. | |
668 | * Terabytes are not yet handled. | |
669 | */ | |
670 | ||
671 | if (bytes < 5000*1024) | |
672 | buf[0] = 0; | |
673 | else if (bytes < 2*1024LL*1024LL*1024LL) { | |
674 | long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2; | |
675 | long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2; | |
676 | snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)", | |
677 | cMiB/100 , cMiB % 100, | |
678 | cMB/100, cMB % 100); | |
679 | } else { | |
680 | long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2; | |
681 | long cGB = (bytes / (1000000000LL/200LL ) +1) /2; | |
682 | snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)", | |
683 | cGiB/100 , cGiB % 100, | |
684 | cGB/100, cGB % 100); | |
685 | } | |
686 | return buf; | |
687 | } | |
688 | ||
689 | char *human_size_brief(long long bytes, int prefix) | |
690 | { | |
691 | static char buf[30]; | |
692 | ||
693 | /* We convert bytes to either centi-M{ega,ibi}bytes or | |
694 | * centi-G{igi,ibi}bytes, with appropriate rounding, | |
695 | * and then print 1/100th of those as a decimal. | |
696 | * We allow upto 2048Megabytes before converting to | |
697 | * gigabytes, as that shows more precision and isn't | |
698 | * too large a number. | |
699 | * Terabytes are not yet handled. | |
700 | * | |
701 | * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc. | |
702 | * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc. | |
703 | */ | |
704 | ||
705 | if (bytes < 5000*1024) | |
706 | buf[0] = 0; | |
707 | else if (prefix == IEC) { | |
708 | if (bytes < 2*1024LL*1024LL*1024LL) { | |
709 | long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2; | |
710 | snprintf(buf, sizeof(buf), "%ld.%02ldMiB", | |
711 | cMiB/100 , cMiB % 100); | |
712 | } else { | |
713 | long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2; | |
714 | snprintf(buf, sizeof(buf), "%ld.%02ldGiB", | |
715 | cGiB/100 , cGiB % 100); | |
716 | } | |
717 | } | |
718 | else if (prefix == JEDEC) { | |
719 | if (bytes < 2*1024LL*1024LL*1024LL) { | |
720 | long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2; | |
721 | snprintf(buf, sizeof(buf), "%ld.%02ldMB", | |
722 | cMB/100, cMB % 100); | |
723 | } else { | |
724 | long cGB = (bytes / (1000000000LL/200LL ) +1) /2; | |
725 | snprintf(buf, sizeof(buf), "%ld.%02ldGB", | |
726 | cGB/100 , cGB % 100); | |
727 | } | |
728 | } | |
729 | else | |
730 | buf[0] = 0; | |
731 | ||
732 | return buf; | |
733 | } | |
734 | ||
735 | void print_r10_layout(int layout) | |
736 | { | |
737 | int near = layout & 255; | |
738 | int far = (layout >> 8) & 255; | |
739 | int offset = (layout&0x10000); | |
740 | char *sep = ""; | |
741 | ||
742 | if (near != 1) { | |
743 | printf("%s near=%d", sep, near); | |
744 | sep = ","; | |
745 | } | |
746 | if (far != 1) | |
747 | printf("%s %s=%d", sep, offset?"offset":"far", far); | |
748 | if (near*far == 1) | |
749 | printf("NO REDUNDANCY"); | |
750 | } | |
751 | #endif | |
752 | ||
753 | unsigned long long calc_array_size(int level, int raid_disks, int layout, | |
754 | int chunksize, unsigned long long devsize) | |
755 | { | |
756 | if (level == 1) | |
757 | return devsize; | |
758 | devsize &= ~(unsigned long long)((chunksize>>9)-1); | |
759 | return get_data_disks(level, layout, raid_disks) * devsize; | |
760 | } | |
761 | ||
762 | int get_data_disks(int level, int layout, int raid_disks) | |
763 | { | |
764 | int data_disks = 0; | |
765 | switch (level) { | |
766 | case 0: data_disks = raid_disks; | |
767 | break; | |
768 | case 1: data_disks = 1; | |
769 | break; | |
770 | case 4: | |
771 | case 5: data_disks = raid_disks - 1; | |
772 | break; | |
773 | case 6: data_disks = raid_disks - 2; | |
774 | break; | |
775 | case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255); | |
776 | break; | |
777 | } | |
778 | ||
779 | return data_disks; | |
780 | } | |
781 | ||
782 | #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) | |
783 | char *get_md_name(int dev) | |
784 | { | |
785 | /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */ | |
786 | /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */ | |
787 | static char devname[50]; | |
788 | struct stat stb; | |
789 | dev_t rdev; | |
790 | char *dn; | |
791 | ||
792 | if (dev < 0) { | |
793 | int mdp = get_mdp_major(); | |
794 | if (mdp < 0) return NULL; | |
795 | rdev = makedev(mdp, (-1-dev)<<6); | |
796 | snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev); | |
797 | if (stat(devname, &stb) == 0 | |
798 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
799 | && (stb.st_rdev == rdev)) | |
800 | return devname; | |
801 | } else { | |
802 | rdev = makedev(MD_MAJOR, dev); | |
803 | snprintf(devname, sizeof(devname), "/dev/md%d", dev); | |
804 | if (stat(devname, &stb) == 0 | |
805 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
806 | && (stb.st_rdev == rdev)) | |
807 | return devname; | |
808 | ||
809 | snprintf(devname, sizeof(devname), "/dev/md/%d", dev); | |
810 | if (stat(devname, &stb) == 0 | |
811 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
812 | && (stb.st_rdev == rdev)) | |
813 | return devname; | |
814 | } | |
815 | dn = map_dev(major(rdev), minor(rdev), 0); | |
816 | if (dn) | |
817 | return dn; | |
818 | snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev); | |
819 | if (mknod(devname, S_IFBLK | 0600, rdev) == -1) | |
820 | if (errno != EEXIST) | |
821 | return NULL; | |
822 | ||
823 | if (stat(devname, &stb) == 0 | |
824 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
825 | && (stb.st_rdev == rdev)) | |
826 | return devname; | |
827 | unlink(devname); | |
828 | return NULL; | |
829 | } | |
830 | ||
831 | void put_md_name(char *name) | |
832 | { | |
833 | if (strncmp(name, "/dev/.tmp.md", 12) == 0) | |
834 | unlink(name); | |
835 | } | |
836 | ||
837 | int find_free_devnum(int use_partitions) | |
838 | { | |
839 | int devnum; | |
840 | for (devnum = 127; devnum != 128; | |
841 | devnum = devnum ? devnum-1 : (1<<20)-1) { | |
842 | char *dn; | |
843 | int _devnum; | |
844 | char nbuf[50]; | |
845 | ||
846 | _devnum = use_partitions ? (-1-devnum) : devnum; | |
847 | if (mddev_busy(_devnum)) | |
848 | continue; | |
849 | sprintf(nbuf, "%s%d", use_partitions?"mdp":"md", devnum); | |
850 | if (!conf_name_is_free(nbuf)) | |
851 | continue; | |
852 | /* make sure it is new to /dev too, at least as a | |
853 | * non-standard */ | |
854 | dn = map_dev(dev2major(_devnum), dev2minor(_devnum), 0); | |
855 | if (dn && ! is_standard(dn, NULL)) | |
856 | continue; | |
857 | break; | |
858 | } | |
859 | if (devnum == 128) | |
860 | return NoMdDev; | |
861 | return use_partitions ? (-1-devnum) : devnum; | |
862 | } | |
863 | #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */ | |
864 | ||
865 | int dev_open(char *dev, int flags) | |
866 | { | |
867 | /* like 'open', but if 'dev' matches %d:%d, create a temp | |
868 | * block device and open that | |
869 | */ | |
870 | char *e; | |
871 | int fd = -1; | |
872 | char devname[32]; | |
873 | int major; | |
874 | int minor; | |
875 | ||
876 | if (!dev) return -1; | |
877 | flags |= O_DIRECT; | |
878 | ||
879 | major = strtoul(dev, &e, 0); | |
880 | if (e > dev && *e == ':' && e[1] && | |
881 | (minor = strtoul(e+1, &e, 0)) >= 0 && | |
882 | *e == 0) { | |
883 | char *path = map_dev(major, minor, 0); | |
884 | if (path) | |
885 | fd = open(path, flags); | |
886 | if (fd < 0) { | |
887 | snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d", | |
888 | (int)getpid(), major, minor); | |
889 | if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) { | |
890 | fd = open(devname, flags); | |
891 | unlink(devname); | |
892 | } | |
893 | } | |
894 | if (fd < 0) { | |
895 | snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d", | |
896 | (int)getpid(), major, minor); | |
897 | if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) { | |
898 | fd = open(devname, flags); | |
899 | unlink(devname); | |
900 | } | |
901 | } | |
902 | } else | |
903 | fd = open(dev, flags); | |
904 | return fd; | |
905 | } | |
906 | ||
907 | int open_dev_flags(int devnum, int flags) | |
908 | { | |
909 | char buf[20]; | |
910 | ||
911 | sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum)); | |
912 | return dev_open(buf, flags); | |
913 | } | |
914 | ||
915 | int open_dev(int devnum) | |
916 | { | |
917 | return open_dev_flags(devnum, O_RDONLY); | |
918 | } | |
919 | ||
920 | int open_dev_excl(int devnum) | |
921 | { | |
922 | char buf[20]; | |
923 | int i; | |
924 | int flags = O_RDWR; | |
925 | ||
926 | sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum)); | |
927 | for (i = 0 ; i < 25 ; i++) { | |
928 | int fd = dev_open(buf, flags|O_EXCL); | |
929 | if (fd >= 0) | |
930 | return fd; | |
931 | if (errno == EACCES && flags == O_RDWR) { | |
932 | flags = O_RDONLY; | |
933 | continue; | |
934 | } | |
935 | if (errno != EBUSY) | |
936 | return fd; | |
937 | usleep(200000); | |
938 | } | |
939 | return -1; | |
940 | } | |
941 | ||
942 | int same_dev(char *one, char *two) | |
943 | { | |
944 | struct stat st1, st2; | |
945 | if (stat(one, &st1) != 0) | |
946 | return 0; | |
947 | if (stat(two, &st2) != 0) | |
948 | return 0; | |
949 | if ((st1.st_mode & S_IFMT) != S_IFBLK) | |
950 | return 0; | |
951 | if ((st2.st_mode & S_IFMT) != S_IFBLK) | |
952 | return 0; | |
953 | return st1.st_rdev == st2.st_rdev; | |
954 | } | |
955 | ||
956 | void wait_for(char *dev, int fd) | |
957 | { | |
958 | int i; | |
959 | struct stat stb_want; | |
960 | ||
961 | if (fstat(fd, &stb_want) != 0 || | |
962 | (stb_want.st_mode & S_IFMT) != S_IFBLK) | |
963 | return; | |
964 | ||
965 | for (i = 0 ; i < 25 ; i++) { | |
966 | struct stat stb; | |
967 | if (stat(dev, &stb) == 0 && | |
968 | (stb.st_mode & S_IFMT) == S_IFBLK && | |
969 | (stb.st_rdev == stb_want.st_rdev)) | |
970 | return; | |
971 | usleep(200000); | |
972 | } | |
973 | if (i == 25) | |
974 | dprintf("%s: timeout waiting for %s\n", __func__, dev); | |
975 | } | |
976 | ||
977 | struct superswitch *superlist[] = | |
978 | { | |
979 | &super0, &super1, | |
980 | &super_ddf, &super_imsm, | |
981 | &mbr, &gpt, | |
982 | NULL }; | |
983 | ||
984 | #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) | |
985 | ||
986 | struct supertype *super_by_fd(int fd, char **subarrayp) | |
987 | { | |
988 | mdu_array_info_t array; | |
989 | int vers; | |
990 | int minor; | |
991 | struct supertype *st = NULL; | |
992 | struct mdinfo *sra; | |
993 | char *verstr; | |
994 | char version[20]; | |
995 | int i; | |
996 | char *subarray = NULL; | |
997 | int container = NoMdDev; | |
998 | ||
999 | sra = sysfs_read(fd, 0, GET_VERSION); | |
1000 | ||
1001 | if (sra) { | |
1002 | vers = sra->array.major_version; | |
1003 | minor = sra->array.minor_version; | |
1004 | verstr = sra->text_version; | |
1005 | } else { | |
1006 | if (ioctl(fd, GET_ARRAY_INFO, &array)) | |
1007 | array.major_version = array.minor_version = 0; | |
1008 | vers = array.major_version; | |
1009 | minor = array.minor_version; | |
1010 | verstr = ""; | |
1011 | } | |
1012 | ||
1013 | if (vers != -1) { | |
1014 | sprintf(version, "%d.%d", vers, minor); | |
1015 | verstr = version; | |
1016 | } | |
1017 | if (minor == -2 && is_subarray(verstr)) { | |
1018 | char *dev = verstr+1; | |
1019 | ||
1020 | subarray = strchr(dev, '/'); | |
1021 | if (subarray) { | |
1022 | *subarray++ = '\0'; | |
1023 | subarray = xstrdup(subarray); | |
1024 | } | |
1025 | container = devname2devnum(dev); | |
1026 | if (sra) | |
1027 | sysfs_free(sra); | |
1028 | sra = sysfs_read(-1, container, GET_VERSION); | |
1029 | if (sra && sra->text_version[0]) | |
1030 | verstr = sra->text_version; | |
1031 | else | |
1032 | verstr = "-no-metadata-"; | |
1033 | } | |
1034 | ||
1035 | for (i = 0; st == NULL && superlist[i] ; i++) | |
1036 | st = superlist[i]->match_metadata_desc(verstr); | |
1037 | ||
1038 | if (sra) | |
1039 | sysfs_free(sra); | |
1040 | if (st) { | |
1041 | st->sb = NULL; | |
1042 | if (subarrayp) | |
1043 | *subarrayp = subarray; | |
1044 | st->container_dev = container; | |
1045 | st->devnum = fd2devnum(fd); | |
1046 | } else | |
1047 | free(subarray); | |
1048 | ||
1049 | return st; | |
1050 | } | |
1051 | #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */ | |
1052 | ||
1053 | int dev_size_from_id(dev_t id, unsigned long long *size) | |
1054 | { | |
1055 | char buf[20]; | |
1056 | int fd; | |
1057 | ||
1058 | sprintf(buf, "%d:%d", major(id), minor(id)); | |
1059 | fd = dev_open(buf, O_RDONLY); | |
1060 | if (fd < 0) | |
1061 | return 0; | |
1062 | if (get_dev_size(fd, NULL, size)) { | |
1063 | close(fd); | |
1064 | return 1; | |
1065 | } | |
1066 | close(fd); | |
1067 | return 0; | |
1068 | } | |
1069 | ||
1070 | struct supertype *dup_super(struct supertype *orig) | |
1071 | { | |
1072 | struct supertype *st; | |
1073 | ||
1074 | if (!orig) | |
1075 | return orig; | |
1076 | st = xcalloc(1, sizeof(*st)); | |
1077 | st->ss = orig->ss; | |
1078 | st->max_devs = orig->max_devs; | |
1079 | st->minor_version = orig->minor_version; | |
1080 | st->sb = NULL; | |
1081 | st->info = NULL; | |
1082 | return st; | |
1083 | } | |
1084 | ||
1085 | struct supertype *guess_super_type(int fd, enum guess_types guess_type) | |
1086 | { | |
1087 | /* try each load_super to find the best match, | |
1088 | * and return the best superswitch | |
1089 | */ | |
1090 | struct superswitch *ss; | |
1091 | struct supertype *st; | |
1092 | time_t besttime = 0; | |
1093 | int bestsuper = -1; | |
1094 | int i; | |
1095 | ||
1096 | st = xcalloc(1, sizeof(*st)); | |
1097 | st->container_dev = NoMdDev; | |
1098 | ||
1099 | for (i = 0 ; superlist[i]; i++) { | |
1100 | int rv; | |
1101 | ss = superlist[i]; | |
1102 | if (guess_type == guess_array && ss->add_to_super == NULL) | |
1103 | continue; | |
1104 | if (guess_type == guess_partitions && ss->add_to_super != NULL) | |
1105 | continue; | |
1106 | memset(st, 0, sizeof(*st)); | |
1107 | st->ignore_hw_compat = 1; | |
1108 | rv = ss->load_super(st, fd, NULL); | |
1109 | if (rv == 0) { | |
1110 | struct mdinfo info; | |
1111 | st->ss->getinfo_super(st, &info, NULL); | |
1112 | if (bestsuper == -1 || | |
1113 | besttime < info.array.ctime) { | |
1114 | bestsuper = i; | |
1115 | besttime = info.array.ctime; | |
1116 | } | |
1117 | ss->free_super(st); | |
1118 | } | |
1119 | } | |
1120 | if (bestsuper != -1) { | |
1121 | int rv; | |
1122 | memset(st, 0, sizeof(*st)); | |
1123 | st->ignore_hw_compat = 1; | |
1124 | rv = superlist[bestsuper]->load_super(st, fd, NULL); | |
1125 | if (rv == 0) { | |
1126 | superlist[bestsuper]->free_super(st); | |
1127 | st->ignore_hw_compat = 0; | |
1128 | return st; | |
1129 | } | |
1130 | } | |
1131 | free(st); | |
1132 | return NULL; | |
1133 | } | |
1134 | ||
1135 | /* Return size of device in bytes */ | |
1136 | int get_dev_size(int fd, char *dname, unsigned long long *sizep) | |
1137 | { | |
1138 | unsigned long long ldsize; | |
1139 | struct stat st; | |
1140 | ||
1141 | if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode)) | |
1142 | ldsize = (unsigned long long)st.st_size; | |
1143 | else | |
1144 | #ifdef BLKGETSIZE64 | |
1145 | if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0) | |
1146 | #endif | |
1147 | { | |
1148 | unsigned long dsize; | |
1149 | if (ioctl(fd, BLKGETSIZE, &dsize) == 0) { | |
1150 | ldsize = dsize; | |
1151 | ldsize <<= 9; | |
1152 | } else { | |
1153 | if (dname) | |
1154 | pr_err("Cannot get size of %s: %s\b", | |
1155 | dname, strerror(errno)); | |
1156 | return 0; | |
1157 | } | |
1158 | } | |
1159 | *sizep = ldsize; | |
1160 | return 1; | |
1161 | } | |
1162 | ||
1163 | /* Return true if this can only be a container, not a member device. | |
1164 | * i.e. is and md device and size is zero | |
1165 | */ | |
1166 | int must_be_container(int fd) | |
1167 | { | |
1168 | unsigned long long size; | |
1169 | if (md_get_version(fd) < 0) | |
1170 | return 0; | |
1171 | if (get_dev_size(fd, NULL, &size) == 0) | |
1172 | return 1; | |
1173 | if (size == 0) | |
1174 | return 1; | |
1175 | return 0; | |
1176 | } | |
1177 | ||
1178 | /* Sets endofpart parameter to the last block used by the last GPT partition on the device. | |
1179 | * Returns: 1 if successful | |
1180 | * -1 for unknown partition type | |
1181 | * 0 for other errors | |
1182 | */ | |
1183 | static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart) | |
1184 | { | |
1185 | struct GPT gpt; | |
1186 | unsigned char empty_gpt_entry[16]= {0}; | |
1187 | struct GPT_part_entry *part; | |
1188 | char buf[512]; | |
1189 | unsigned long long curr_part_end; | |
1190 | unsigned all_partitions, entry_size; | |
1191 | unsigned part_nr; | |
1192 | ||
1193 | *endofpart = 0; | |
1194 | ||
1195 | BUILD_BUG_ON(sizeof(gpt) != 512); | |
1196 | /* skip protective MBR */ | |
1197 | lseek(fd, 512, SEEK_SET); | |
1198 | /* read GPT header */ | |
1199 | if (read(fd, &gpt, 512) != 512) | |
1200 | return 0; | |
1201 | ||
1202 | /* get the number of partition entries and the entry size */ | |
1203 | all_partitions = __le32_to_cpu(gpt.part_cnt); | |
1204 | entry_size = __le32_to_cpu(gpt.part_size); | |
1205 | ||
1206 | /* Check GPT signature*/ | |
1207 | if (gpt.magic != GPT_SIGNATURE_MAGIC) | |
1208 | return -1; | |
1209 | ||
1210 | /* sanity checks */ | |
1211 | if (all_partitions > 1024 || | |
1212 | entry_size > sizeof(buf)) | |
1213 | return -1; | |
1214 | ||
1215 | part = (struct GPT_part_entry *)buf; | |
1216 | ||
1217 | for (part_nr = 0; part_nr < all_partitions; part_nr++) { | |
1218 | /* read partition entry */ | |
1219 | if (read(fd, buf, entry_size) != (ssize_t)entry_size) | |
1220 | return 0; | |
1221 | ||
1222 | /* is this valid partition? */ | |
1223 | if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) { | |
1224 | /* check the last lba for the current partition */ | |
1225 | curr_part_end = __le64_to_cpu(part->ending_lba); | |
1226 | if (curr_part_end > *endofpart) | |
1227 | *endofpart = curr_part_end; | |
1228 | } | |
1229 | ||
1230 | } | |
1231 | return 1; | |
1232 | } | |
1233 | ||
1234 | /* Sets endofpart parameter to the last block used by the last partition on the device. | |
1235 | * Returns: 1 if successful | |
1236 | * -1 for unknown partition type | |
1237 | * 0 for other errors | |
1238 | */ | |
1239 | static int get_last_partition_end(int fd, unsigned long long *endofpart) | |
1240 | { | |
1241 | struct MBR boot_sect; | |
1242 | struct MBR_part_record *part; | |
1243 | unsigned long long curr_part_end; | |
1244 | unsigned part_nr; | |
1245 | int retval = 0; | |
1246 | ||
1247 | *endofpart = 0; | |
1248 | ||
1249 | BUILD_BUG_ON(sizeof(boot_sect) != 512); | |
1250 | /* read MBR */ | |
1251 | lseek(fd, 0, 0); | |
1252 | if (read(fd, &boot_sect, 512) != 512) | |
1253 | goto abort; | |
1254 | ||
1255 | /* check MBP signature */ | |
1256 | if (boot_sect.magic == MBR_SIGNATURE_MAGIC) { | |
1257 | retval = 1; | |
1258 | /* found the correct signature */ | |
1259 | part = boot_sect.parts; | |
1260 | ||
1261 | for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) { | |
1262 | /* check for GPT type */ | |
1263 | if (part->part_type == MBR_GPT_PARTITION_TYPE) { | |
1264 | retval = get_gpt_last_partition_end(fd, endofpart); | |
1265 | break; | |
1266 | } | |
1267 | /* check the last used lba for the current partition */ | |
1268 | curr_part_end = __le32_to_cpu(part->first_sect_lba) + | |
1269 | __le32_to_cpu(part->blocks_num); | |
1270 | if (curr_part_end > *endofpart) | |
1271 | *endofpart = curr_part_end; | |
1272 | ||
1273 | part++; | |
1274 | } | |
1275 | } else { | |
1276 | /* Unknown partition table */ | |
1277 | retval = -1; | |
1278 | } | |
1279 | abort: | |
1280 | return retval; | |
1281 | } | |
1282 | ||
1283 | int check_partitions(int fd, char *dname, unsigned long long freesize, | |
1284 | unsigned long long size) | |
1285 | { | |
1286 | /* | |
1287 | * Check where the last partition ends | |
1288 | */ | |
1289 | unsigned long long endofpart; | |
1290 | int ret; | |
1291 | ||
1292 | if ((ret = get_last_partition_end(fd, &endofpart)) > 0) { | |
1293 | /* There appears to be a partition table here */ | |
1294 | if (freesize == 0) { | |
1295 | /* partitions will not be visible in new device */ | |
1296 | pr_err("partition table exists on %s but will be lost or\n" | |
1297 | " meaningless after creating array\n", | |
1298 | dname); | |
1299 | return 1; | |
1300 | } else if (endofpart > freesize) { | |
1301 | /* last partition overlaps metadata */ | |
1302 | pr_err("metadata will over-write last partition on %s.\n", | |
1303 | dname); | |
1304 | return 1; | |
1305 | } else if (size && endofpart > size) { | |
1306 | /* partitions will be truncated in new device */ | |
1307 | pr_err("array size is too small to cover all partitions on %s.\n", | |
1308 | dname); | |
1309 | return 1; | |
1310 | } | |
1311 | } | |
1312 | return 0; | |
1313 | } | |
1314 | ||
1315 | void get_one_disk(int mdfd, mdu_array_info_t *ainf, mdu_disk_info_t *disk) | |
1316 | { | |
1317 | int d; | |
1318 | ||
1319 | ioctl(mdfd, GET_ARRAY_INFO, ainf); | |
1320 | for (d = 0 ; d < MAX_DISKS ; d++) { | |
1321 | if (ioctl(mdfd, GET_DISK_INFO, disk) == 0 && | |
1322 | (disk->major || disk->minor)) | |
1323 | return; | |
1324 | } | |
1325 | } | |
1326 | ||
1327 | int open_container(int fd) | |
1328 | { | |
1329 | /* 'fd' is a block device. Find out if it is in use | |
1330 | * by a container, and return an open fd on that container. | |
1331 | */ | |
1332 | char path[256]; | |
1333 | char *e; | |
1334 | DIR *dir; | |
1335 | struct dirent *de; | |
1336 | int dfd, n; | |
1337 | char buf[200]; | |
1338 | int major, minor; | |
1339 | struct stat st; | |
1340 | ||
1341 | if (fstat(fd, &st) != 0) | |
1342 | return -1; | |
1343 | sprintf(path, "/sys/dev/block/%d:%d/holders", | |
1344 | (int)major(st.st_rdev), (int)minor(st.st_rdev)); | |
1345 | e = path + strlen(path); | |
1346 | ||
1347 | dir = opendir(path); | |
1348 | if (!dir) | |
1349 | return -1; | |
1350 | while ((de = readdir(dir))) { | |
1351 | if (de->d_ino == 0) | |
1352 | continue; | |
1353 | if (de->d_name[0] == '.') | |
1354 | continue; | |
1355 | sprintf(e, "/%s/dev", de->d_name); | |
1356 | dfd = open(path, O_RDONLY); | |
1357 | if (dfd < 0) | |
1358 | continue; | |
1359 | n = read(dfd, buf, sizeof(buf)); | |
1360 | close(dfd); | |
1361 | if (n <= 0 || (unsigned)n >= sizeof(buf)) | |
1362 | continue; | |
1363 | buf[n] = 0; | |
1364 | if (sscanf(buf, "%d:%d", &major, &minor) != 2) | |
1365 | continue; | |
1366 | sprintf(buf, "%d:%d", major, minor); | |
1367 | dfd = dev_open(buf, O_RDONLY); | |
1368 | if (dfd >= 0) { | |
1369 | closedir(dir); | |
1370 | return dfd; | |
1371 | } | |
1372 | } | |
1373 | closedir(dir); | |
1374 | return -1; | |
1375 | } | |
1376 | ||
1377 | struct superswitch *version_to_superswitch(char *vers) | |
1378 | { | |
1379 | int i; | |
1380 | ||
1381 | for (i = 0; superlist[i]; i++) { | |
1382 | struct superswitch *ss = superlist[i]; | |
1383 | ||
1384 | if (strcmp(vers, ss->name) == 0) | |
1385 | return ss; | |
1386 | } | |
1387 | ||
1388 | return NULL; | |
1389 | } | |
1390 | ||
1391 | int is_container_member(struct mdstat_ent *mdstat, char *container) | |
1392 | { | |
1393 | if (mdstat->metadata_version == NULL || | |
1394 | strncmp(mdstat->metadata_version, "external:", 9) != 0 || | |
1395 | !is_subarray(mdstat->metadata_version+9) || | |
1396 | strncmp(mdstat->metadata_version+10, container, strlen(container)) != 0 || | |
1397 | mdstat->metadata_version[10+strlen(container)] != '/') | |
1398 | return 0; | |
1399 | ||
1400 | return 1; | |
1401 | } | |
1402 | ||
1403 | int is_subarray_active(char *subarray, char *container) | |
1404 | { | |
1405 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
1406 | struct mdstat_ent *ent; | |
1407 | ||
1408 | for (ent = mdstat; ent; ent = ent->next) | |
1409 | if (is_container_member(ent, container)) | |
1410 | if (strcmp(to_subarray(ent, container), subarray) == 0) | |
1411 | break; | |
1412 | ||
1413 | free_mdstat(mdstat); | |
1414 | ||
1415 | return ent != NULL; | |
1416 | } | |
1417 | ||
1418 | /* open_subarray - opens a subarray in a container | |
1419 | * @dev: container device name | |
1420 | * @st: empty supertype | |
1421 | * @quiet: block reporting errors flag | |
1422 | * | |
1423 | * On success returns an fd to a container and fills in *st | |
1424 | */ | |
1425 | int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet) | |
1426 | { | |
1427 | struct mdinfo *mdi; | |
1428 | struct mdinfo *info; | |
1429 | int fd, err = 1; | |
1430 | ||
1431 | fd = open(dev, O_RDWR|O_EXCL); | |
1432 | if (fd < 0) { | |
1433 | if (!quiet) | |
1434 | pr_err("Couldn't open %s, aborting\n", | |
1435 | dev); | |
1436 | return -1; | |
1437 | } | |
1438 | ||
1439 | st->devnum = fd2devnum(fd); | |
1440 | if (st->devnum == NoMdDev) { | |
1441 | if (!quiet) | |
1442 | pr_err("Failed to determine device number for %s\n", | |
1443 | dev); | |
1444 | goto close_fd; | |
1445 | } | |
1446 | ||
1447 | mdi = sysfs_read(fd, st->devnum, GET_VERSION|GET_LEVEL); | |
1448 | if (!mdi) { | |
1449 | if (!quiet) | |
1450 | pr_err("Failed to read sysfs for %s\n", | |
1451 | dev); | |
1452 | goto close_fd; | |
1453 | } | |
1454 | ||
1455 | if (mdi->array.level != UnSet) { | |
1456 | if (!quiet) | |
1457 | pr_err("%s is not a container\n", dev); | |
1458 | goto free_sysfs; | |
1459 | } | |
1460 | ||
1461 | st->ss = version_to_superswitch(mdi->text_version); | |
1462 | if (!st->ss) { | |
1463 | if (!quiet) | |
1464 | pr_err("Operation not supported for %s metadata\n", | |
1465 | mdi->text_version); | |
1466 | goto free_sysfs; | |
1467 | } | |
1468 | ||
1469 | st->devname = devnum2devname(st->devnum); | |
1470 | if (!st->devname) { | |
1471 | if (!quiet) | |
1472 | pr_err("Failed to allocate device name\n"); | |
1473 | goto free_sysfs; | |
1474 | } | |
1475 | ||
1476 | if (!st->ss->load_container) { | |
1477 | if (!quiet) | |
1478 | pr_err("%s is not a container\n", dev); | |
1479 | goto free_name; | |
1480 | } | |
1481 | ||
1482 | if (st->ss->load_container(st, fd, NULL)) { | |
1483 | if (!quiet) | |
1484 | pr_err("Failed to load metadata for %s\n", | |
1485 | dev); | |
1486 | goto free_name; | |
1487 | } | |
1488 | ||
1489 | info = st->ss->container_content(st, subarray); | |
1490 | if (!info) { | |
1491 | if (!quiet) | |
1492 | pr_err("Failed to find subarray-%s in %s\n", | |
1493 | subarray, dev); | |
1494 | goto free_super; | |
1495 | } | |
1496 | free(info); | |
1497 | ||
1498 | err = 0; | |
1499 | ||
1500 | free_super: | |
1501 | if (err) | |
1502 | st->ss->free_super(st); | |
1503 | free_name: | |
1504 | if (err) | |
1505 | free(st->devname); | |
1506 | free_sysfs: | |
1507 | sysfs_free(mdi); | |
1508 | close_fd: | |
1509 | if (err) | |
1510 | close(fd); | |
1511 | ||
1512 | if (err) | |
1513 | return -1; | |
1514 | else | |
1515 | return fd; | |
1516 | } | |
1517 | ||
1518 | int add_disk(int mdfd, struct supertype *st, | |
1519 | struct mdinfo *sra, struct mdinfo *info) | |
1520 | { | |
1521 | /* Add a device to an array, in one of 2 ways. */ | |
1522 | int rv; | |
1523 | #ifndef MDASSEMBLE | |
1524 | if (st->ss->external) { | |
1525 | if (info->disk.state & (1<<MD_DISK_SYNC)) | |
1526 | info->recovery_start = MaxSector; | |
1527 | else | |
1528 | info->recovery_start = 0; | |
1529 | rv = sysfs_add_disk(sra, info, 0); | |
1530 | if (! rv) { | |
1531 | struct mdinfo *sd2; | |
1532 | for (sd2 = sra->devs; sd2; sd2=sd2->next) | |
1533 | if (sd2 == info) | |
1534 | break; | |
1535 | if (sd2 == NULL) { | |
1536 | sd2 = xmalloc(sizeof(*sd2)); | |
1537 | *sd2 = *info; | |
1538 | sd2->next = sra->devs; | |
1539 | sra->devs = sd2; | |
1540 | } | |
1541 | } | |
1542 | } else | |
1543 | #endif | |
1544 | rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk); | |
1545 | return rv; | |
1546 | } | |
1547 | ||
1548 | int remove_disk(int mdfd, struct supertype *st, | |
1549 | struct mdinfo *sra, struct mdinfo *info) | |
1550 | { | |
1551 | int rv; | |
1552 | /* Remove the disk given by 'info' from the array */ | |
1553 | #ifndef MDASSEMBLE | |
1554 | if (st->ss->external) | |
1555 | rv = sysfs_set_str(sra, info, "slot", "none"); | |
1556 | else | |
1557 | #endif | |
1558 | rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major, | |
1559 | info->disk.minor)); | |
1560 | return rv; | |
1561 | } | |
1562 | ||
1563 | int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info) | |
1564 | { | |
1565 | /* Initialise kernel's knowledge of array. | |
1566 | * This varies between externally managed arrays | |
1567 | * and older kernels | |
1568 | */ | |
1569 | int vers = md_get_version(mdfd); | |
1570 | int rv; | |
1571 | ||
1572 | #ifndef MDASSEMBLE | |
1573 | if (st->ss->external) | |
1574 | rv = sysfs_set_array(info, vers); | |
1575 | else | |
1576 | #endif | |
1577 | if ((vers % 100) >= 1) { /* can use different versions */ | |
1578 | mdu_array_info_t inf; | |
1579 | memset(&inf, 0, sizeof(inf)); | |
1580 | inf.major_version = info->array.major_version; | |
1581 | inf.minor_version = info->array.minor_version; | |
1582 | rv = ioctl(mdfd, SET_ARRAY_INFO, &inf); | |
1583 | } else | |
1584 | rv = ioctl(mdfd, SET_ARRAY_INFO, NULL); | |
1585 | return rv; | |
1586 | } | |
1587 | ||
1588 | unsigned long long min_recovery_start(struct mdinfo *array) | |
1589 | { | |
1590 | /* find the minimum recovery_start in an array for metadata | |
1591 | * formats that only record per-array recovery progress instead | |
1592 | * of per-device | |
1593 | */ | |
1594 | unsigned long long recovery_start = MaxSector; | |
1595 | struct mdinfo *d; | |
1596 | ||
1597 | for (d = array->devs; d; d = d->next) | |
1598 | recovery_start = min(recovery_start, d->recovery_start); | |
1599 | ||
1600 | return recovery_start; | |
1601 | } | |
1602 | ||
1603 | int mdmon_pid(int devnum) | |
1604 | { | |
1605 | char path[100]; | |
1606 | char pid[10]; | |
1607 | int fd; | |
1608 | int n; | |
1609 | char *devname = devnum2devname(devnum); | |
1610 | ||
1611 | sprintf(path, "%s/%s.pid", MDMON_DIR, devname); | |
1612 | free(devname); | |
1613 | ||
1614 | fd = open(path, O_RDONLY | O_NOATIME, 0); | |
1615 | ||
1616 | if (fd < 0) | |
1617 | return -1; | |
1618 | n = read(fd, pid, 9); | |
1619 | close(fd); | |
1620 | if (n <= 0) | |
1621 | return -1; | |
1622 | return atoi(pid); | |
1623 | } | |
1624 | ||
1625 | int mdmon_running(int devnum) | |
1626 | { | |
1627 | int pid = mdmon_pid(devnum); | |
1628 | if (pid <= 0) | |
1629 | return 0; | |
1630 | if (kill(pid, 0) == 0) | |
1631 | return 1; | |
1632 | return 0; | |
1633 | } | |
1634 | ||
1635 | int start_mdmon(int devnum) | |
1636 | { | |
1637 | int i, skipped; | |
1638 | int len; | |
1639 | pid_t pid; | |
1640 | int status; | |
1641 | char pathbuf[1024]; | |
1642 | char *paths[4] = { | |
1643 | pathbuf, | |
1644 | "/sbin/mdmon", | |
1645 | "mdmon", | |
1646 | NULL | |
1647 | }; | |
1648 | ||
1649 | if (check_env("MDADM_NO_MDMON")) | |
1650 | return 0; | |
1651 | ||
1652 | len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1); | |
1653 | if (len > 0) { | |
1654 | char *sl; | |
1655 | pathbuf[len] = 0; | |
1656 | sl = strrchr(pathbuf, '/'); | |
1657 | if (sl) | |
1658 | sl++; | |
1659 | else | |
1660 | sl = pathbuf; | |
1661 | strcpy(sl, "mdmon"); | |
1662 | } else | |
1663 | pathbuf[0] = '\0'; | |
1664 | ||
1665 | switch(fork()) { | |
1666 | case 0: | |
1667 | /* FIXME yuk. CLOSE_EXEC?? */ | |
1668 | skipped = 0; | |
1669 | for (i = 3; skipped < 20; i++) | |
1670 | if (close(i) < 0) | |
1671 | skipped++; | |
1672 | else | |
1673 | skipped = 0; | |
1674 | ||
1675 | for (i = 0; paths[i]; i++) | |
1676 | if (paths[i][0]) { | |
1677 | if (__offroot) { | |
1678 | execl(paths[i], "mdmon", "--offroot", | |
1679 | devnum2devname(devnum), | |
1680 | NULL); | |
1681 | } else { | |
1682 | execl(paths[i], "mdmon", | |
1683 | devnum2devname(devnum), | |
1684 | NULL); | |
1685 | } | |
1686 | } | |
1687 | exit(1); | |
1688 | case -1: pr_err("cannot run mdmon. " | |
1689 | "Array remains readonly\n"); | |
1690 | return -1; | |
1691 | default: /* parent - good */ | |
1692 | pid = wait(&status); | |
1693 | if (pid < 0 || status != 0) | |
1694 | return -1; | |
1695 | } | |
1696 | return 0; | |
1697 | } | |
1698 | ||
1699 | int check_env(char *name) | |
1700 | { | |
1701 | char *val = getenv(name); | |
1702 | ||
1703 | if (val && atoi(val) == 1) | |
1704 | return 1; | |
1705 | ||
1706 | return 0; | |
1707 | } | |
1708 | ||
1709 | __u32 random32(void) | |
1710 | { | |
1711 | __u32 rv; | |
1712 | int rfd = open("/dev/urandom", O_RDONLY); | |
1713 | if (rfd < 0 || read(rfd, &rv, 4) != 4) | |
1714 | rv = random(); | |
1715 | if (rfd >= 0) | |
1716 | close(rfd); | |
1717 | return rv; | |
1718 | } | |
1719 | ||
1720 | #ifndef MDASSEMBLE | |
1721 | int flush_metadata_updates(struct supertype *st) | |
1722 | { | |
1723 | int sfd; | |
1724 | if (!st->updates) { | |
1725 | st->update_tail = NULL; | |
1726 | return -1; | |
1727 | } | |
1728 | ||
1729 | sfd = connect_monitor(devnum2devname(st->container_dev)); | |
1730 | if (sfd < 0) | |
1731 | return -1; | |
1732 | ||
1733 | while (st->updates) { | |
1734 | struct metadata_update *mu = st->updates; | |
1735 | st->updates = mu->next; | |
1736 | ||
1737 | send_message(sfd, mu, 0); | |
1738 | wait_reply(sfd, 0); | |
1739 | free(mu->buf); | |
1740 | free(mu); | |
1741 | } | |
1742 | ack(sfd, 0); | |
1743 | wait_reply(sfd, 0); | |
1744 | close(sfd); | |
1745 | st->update_tail = NULL; | |
1746 | return 0; | |
1747 | } | |
1748 | ||
1749 | void append_metadata_update(struct supertype *st, void *buf, int len) | |
1750 | { | |
1751 | ||
1752 | struct metadata_update *mu = xmalloc(sizeof(*mu)); | |
1753 | ||
1754 | mu->buf = buf; | |
1755 | mu->len = len; | |
1756 | mu->space = NULL; | |
1757 | mu->space_list = NULL; | |
1758 | mu->next = NULL; | |
1759 | *st->update_tail = mu; | |
1760 | st->update_tail = &mu->next; | |
1761 | } | |
1762 | #endif /* MDASSEMBLE */ | |
1763 | ||
1764 | #ifdef __TINYC__ | |
1765 | /* tinyc doesn't optimize this check in ioctl.h out ... */ | |
1766 | unsigned int __invalid_size_argument_for_IOC = 0; | |
1767 | #endif | |
1768 | ||
1769 | int experimental(void) | |
1770 | { | |
1771 | if (check_env("MDADM_EXPERIMENTAL")) | |
1772 | return 1; | |
1773 | else { | |
1774 | pr_err("To use this feature MDADM_EXPERIMENTAL" | |
1775 | " environment variable has to be defined.\n"); | |
1776 | return 0; | |
1777 | } | |
1778 | } | |
1779 | ||
1780 | /* Pick all spares matching given criteria from a container | |
1781 | * if min_size == 0 do not check size | |
1782 | * if domlist == NULL do not check domains | |
1783 | * if spare_group given add it to domains of each spare | |
1784 | * metadata allows to test domains using metadata of destination array */ | |
1785 | struct mdinfo *container_choose_spares(struct supertype *st, | |
1786 | unsigned long long min_size, | |
1787 | struct domainlist *domlist, | |
1788 | char *spare_group, | |
1789 | const char *metadata, int get_one) | |
1790 | { | |
1791 | struct mdinfo *d, **dp, *disks = NULL; | |
1792 | ||
1793 | /* get list of all disks in container */ | |
1794 | if (st->ss->getinfo_super_disks) | |
1795 | disks = st->ss->getinfo_super_disks(st); | |
1796 | ||
1797 | if (!disks) | |
1798 | return disks; | |
1799 | /* find spare devices on the list */ | |
1800 | dp = &disks->devs; | |
1801 | disks->array.spare_disks = 0; | |
1802 | while (*dp) { | |
1803 | int found = 0; | |
1804 | d = *dp; | |
1805 | if (d->disk.state == 0) { | |
1806 | /* check if size is acceptable */ | |
1807 | unsigned long long dev_size; | |
1808 | dev_t dev = makedev(d->disk.major,d->disk.minor); | |
1809 | ||
1810 | if (!min_size || | |
1811 | (dev_size_from_id(dev, &dev_size) && | |
1812 | dev_size >= min_size)) | |
1813 | found = 1; | |
1814 | /* check if domain matches */ | |
1815 | if (found && domlist) { | |
1816 | struct dev_policy *pol = devnum_policy(dev); | |
1817 | if (spare_group) | |
1818 | pol_add(&pol, pol_domain, | |
1819 | spare_group, NULL); | |
1820 | if (domain_test(domlist, pol, metadata) != 1) | |
1821 | found = 0; | |
1822 | dev_policy_free(pol); | |
1823 | } | |
1824 | } | |
1825 | if (found) { | |
1826 | dp = &d->next; | |
1827 | disks->array.spare_disks++; | |
1828 | if (get_one) { | |
1829 | sysfs_free(*dp); | |
1830 | d->next = NULL; | |
1831 | } | |
1832 | } else { | |
1833 | *dp = d->next; | |
1834 | d->next = NULL; | |
1835 | sysfs_free(d); | |
1836 | } | |
1837 | } | |
1838 | return disks; | |
1839 | } | |
1840 | ||
1841 | /* Checks if paths point to the same device | |
1842 | * Returns 0 if they do. | |
1843 | * Returns 1 if they don't. | |
1844 | * Returns -1 if something went wrong, | |
1845 | * e.g. paths are empty or the files | |
1846 | * they point to don't exist */ | |
1847 | int compare_paths (char* path1, char* path2) | |
1848 | { | |
1849 | struct stat st1,st2; | |
1850 | ||
1851 | if (path1 == NULL || path2 == NULL) | |
1852 | return -1; | |
1853 | if (stat(path1,&st1) != 0) | |
1854 | return -1; | |
1855 | if (stat(path2,&st2) != 0) | |
1856 | return -1; | |
1857 | if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev)) | |
1858 | return 0; | |
1859 | return 1; | |
1860 | } |