]>
Commit | Line | Data |
---|---|---|
64c4757e | 1 | /* |
9a9dab36 | 2 | * mdadm - manage Linux "md" devices aka RAID arrays. |
64c4757e | 3 | * |
e736b623 | 4 | * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de> |
64c4757e 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> |
64c4757e NB |
23 | */ |
24 | ||
9a9dab36 | 25 | #include "mdadm.h" |
64c4757e | 26 | #include "md_p.h" |
edd8d13c | 27 | #include <sys/socket.h> |
64c4757e | 28 | #include <sys/utsname.h> |
9fe32043 | 29 | #include <sys/wait.h> |
edd8d13c | 30 | #include <sys/un.h> |
98c6faba | 31 | #include <ctype.h> |
a322f70c | 32 | #include <dirent.h> |
a931db9e | 33 | #include <signal.h> |
0a816ef9 | 34 | |
a0963a86 JS |
35 | int __offroot; |
36 | ||
0a816ef9 NB |
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 | }; | |
64c4757e | 69 | |
0f22b998 | 70 | #include "part.h" |
056b331e N |
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 | ||
64c4757e NB |
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 | { | |
aba69144 NB |
89 | int hit = 0; /* number of Hex digIT */ |
90 | int i; | |
91 | char c; | |
92 | for (i=0; i<4; i++) uuid[i]=0; | |
93 | ||
94 | while ((c= *str++)) { | |
95 | int n; | |
96 | if (c>='0' && c<='9') | |
97 | n = c-'0'; | |
98 | else if (c>='a' && c <= 'f') | |
99 | n = 10 + c - 'a'; | |
100 | else if (c>='A' && c <= 'F') | |
101 | n = 10 + c - 'A'; | |
102 | else if (strchr(":. -", c)) | |
103 | continue; | |
104 | else return 0; | |
105 | ||
106 | if (hit<32) { | |
107 | uuid[hit/8] <<= 4; | |
108 | uuid[hit/8] += n; | |
109 | } | |
110 | hit++; | |
82b27616 | 111 | } |
aba69144 NB |
112 | if (hit == 32) |
113 | return 1; | |
114 | return 0; | |
64c4757e NB |
115 | } |
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) | |
682c7051 | 139 | return (vers.major*10000) + (vers.minor*100) + vers.patchlevel; |
5787fa49 NB |
140 | if (errno == EACCES) |
141 | return -1; | |
0df46c2a | 142 | if (major(stb.st_rdev) == MD_MAJOR) |
682c7051 | 143 | return (3600); |
64c4757e NB |
144 | return -1; |
145 | } | |
146 | ||
64c4757e NB |
147 | int get_linux_version() |
148 | { | |
149 | struct utsname name; | |
98c6faba | 150 | char *cp; |
f161d047 | 151 | int a = 0, b = 0,c = 0; |
64c4757e NB |
152 | if (uname(&name) <0) |
153 | return -1; | |
154 | ||
98c6faba NB |
155 | cp = name.release; |
156 | a = strtoul(cp, &cp, 10); | |
f161d047 N |
157 | if (*cp == '.') |
158 | b = strtoul(cp+1, &cp, 10); | |
159 | if (*cp == '.') | |
160 | c = strtoul(cp+1, &cp, 10); | |
98c6faba | 161 | |
682c7051 | 162 | return (a*1000000)+(b*1000)+c; |
64c4757e NB |
163 | } |
164 | ||
71204a50 | 165 | #ifndef MDASSEMBLE |
bc77ed53 DW |
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 | ||
84e11361 N |
191 | 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) | |
15632a96 | 197 | * -1 returned on error. |
84e11361 N |
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 | } | |
217 | } | |
218 | if (*c) | |
15632a96 | 219 | s = -1; |
84e11361 N |
220 | return s; |
221 | } | |
222 | ||
4a06e2c2 N |
223 | int parse_layout_10(char *layout) |
224 | { | |
225 | int copies, rv; | |
226 | char *cp; | |
227 | /* Parse the layout string for raid10 */ | |
228 | /* 'f', 'o' or 'n' followed by a number <= raid_disks */ | |
229 | if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') || | |
230 | (copies = strtoul(layout+1, &cp, 10)) < 1 || | |
231 | copies > 200 || | |
232 | *cp) | |
233 | return -1; | |
234 | if (layout[0] == 'n') | |
235 | rv = 256 + copies; | |
236 | else if (layout[0] == 'o') | |
237 | rv = 0x10000 + (copies<<8) + 1; | |
238 | else | |
239 | rv = 1 + (copies<<8); | |
240 | return rv; | |
241 | } | |
242 | ||
243 | int parse_layout_faulty(char *layout) | |
244 | { | |
245 | /* Parse the layout string for 'faulty' */ | |
246 | int ln = strcspn(layout, "0123456789"); | |
503975b9 | 247 | char *m = xstrdup(layout); |
4a06e2c2 N |
248 | int mode; |
249 | m[ln] = 0; | |
250 | mode = map_name(faultylayout, m); | |
251 | if (mode == UnSet) | |
252 | return -1; | |
253 | ||
254 | return mode | (atoi(layout+ln)<< ModeShift); | |
255 | } | |
eb3929a4 | 256 | #endif |
4a06e2c2 | 257 | |
0430ed48 NB |
258 | void remove_partitions(int fd) |
259 | { | |
260 | /* remove partitions from this block devices. | |
261 | * This is used for components added to an array | |
262 | */ | |
263 | #ifdef BLKPG_DEL_PARTITION | |
264 | struct blkpg_ioctl_arg a; | |
265 | struct blkpg_partition p; | |
266 | ||
267 | a.op = BLKPG_DEL_PARTITION; | |
268 | a.data = (void*)&p; | |
269 | a.datalen = sizeof(p); | |
270 | a.flags = 0; | |
271 | memset(a.data, 0, a.datalen); | |
272 | for (p.pno=0; p.pno < 16; p.pno++) | |
273 | ioctl(fd, BLKPG, &a); | |
274 | #endif | |
275 | } | |
276 | ||
691c6ee1 N |
277 | int test_partition(int fd) |
278 | { | |
279 | /* Check if fd is a whole-disk or a partition. | |
280 | * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION | |
281 | * will return ENXIO on an invalid partition number. | |
282 | */ | |
283 | struct blkpg_ioctl_arg a; | |
284 | struct blkpg_partition p; | |
285 | a.op = BLKPG_DEL_PARTITION; | |
286 | a.data = (void*)&p; | |
287 | a.datalen = sizeof(p); | |
288 | a.flags = 0; | |
289 | memset(a.data, 0, a.datalen); | |
290 | p.pno = 1<<30; | |
291 | if (ioctl(fd, BLKPG, &a) == 0) | |
292 | /* Very unlikely, but not a partition */ | |
293 | return 0; | |
294 | if (errno == ENXIO) | |
295 | /* not a partition */ | |
296 | return 0; | |
297 | ||
298 | return 1; | |
299 | } | |
300 | ||
bfd76b93 CA |
301 | int test_partition_from_id(dev_t id) |
302 | { | |
303 | char buf[20]; | |
304 | int fd, rv; | |
305 | ||
306 | sprintf(buf, "%d:%d", major(id), minor(id)); | |
307 | fd = dev_open(buf, O_RDONLY); | |
308 | if (fd < 0) | |
309 | return -1; | |
310 | rv = test_partition(fd); | |
311 | close(fd); | |
312 | return rv; | |
313 | } | |
691c6ee1 | 314 | |
de5a472e | 315 | int enough(int level, int raid_disks, int layout, int clean, char *avail) |
64c4757e | 316 | { |
265e0f17 | 317 | int copies, first; |
de5a472e N |
318 | int i; |
319 | int avail_disks = 0; | |
320 | ||
321 | for (i = 0; i < raid_disks; i++) | |
322 | avail_disks += !!avail[i]; | |
323 | ||
64c4757e | 324 | switch (level) { |
265e0f17 NB |
325 | case 10: |
326 | /* This is the tricky one - we need to check | |
327 | * which actual disks are present. | |
328 | */ | |
702b557b | 329 | copies = (layout&255)* ((layout>>8) & 255); |
265e0f17 NB |
330 | first=0; |
331 | do { | |
332 | /* there must be one of the 'copies' form 'first' */ | |
333 | int n = copies; | |
334 | int cnt=0; | |
335 | while (n--) { | |
336 | if (avail[first]) | |
337 | cnt++; | |
338 | first = (first+1) % raid_disks; | |
339 | } | |
340 | if (cnt == 0) | |
341 | return 0; | |
342 | ||
343 | } while (first != 0); | |
344 | return 1; | |
e5329c37 | 345 | |
df0d4ea0 | 346 | case LEVEL_MULTIPATH: |
e0d19036 | 347 | return avail_disks>= 1; |
df0d4ea0 | 348 | case LEVEL_LINEAR: |
64c4757e NB |
349 | case 0: |
350 | return avail_disks == raid_disks; | |
351 | case 1: | |
352 | return avail_disks >= 1; | |
353 | case 4: | |
354 | case 5: | |
583315d9 NB |
355 | if (clean) |
356 | return avail_disks >= raid_disks-1; | |
357 | else | |
358 | return avail_disks >= raid_disks; | |
98c6faba | 359 | case 6: |
583315d9 NB |
360 | if (clean) |
361 | return avail_disks >= raid_disks-2; | |
362 | else | |
363 | return avail_disks >= raid_disks; | |
64c4757e NB |
364 | default: |
365 | return 0; | |
366 | } | |
367 | } | |
368 | ||
8453e704 N |
369 | int enough_fd(int fd) |
370 | { | |
371 | struct mdu_array_info_s array; | |
372 | struct mdu_disk_info_s disk; | |
373 | int avail_disks = 0; | |
aabe020d | 374 | int i, rv; |
8453e704 N |
375 | char *avail; |
376 | ||
377 | if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 || | |
378 | array.raid_disks <= 0) | |
379 | return 0; | |
503975b9 | 380 | avail = xcalloc(array.raid_disks, 1); |
480f3566 | 381 | for (i=0; i < MAX_DISKS && array.nr_disks > 0; i++) { |
8453e704 N |
382 | disk.number = i; |
383 | if (ioctl(fd, GET_DISK_INFO, &disk) != 0) | |
384 | continue; | |
9e6d9291 N |
385 | if (disk.major == 0 && disk.minor == 0) |
386 | continue; | |
387 | array.nr_disks--; | |
388 | ||
8453e704 N |
389 | if (! (disk.state & (1<<MD_DISK_SYNC))) |
390 | continue; | |
391 | if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks) | |
392 | continue; | |
393 | avail_disks++; | |
394 | avail[disk.raid_disk] = 1; | |
395 | } | |
396 | /* This is used on an active array, so assume it is clean */ | |
aabe020d | 397 | rv = enough(array.level, array.raid_disks, array.layout, |
de5a472e | 398 | 1, avail); |
aabe020d JS |
399 | free(avail); |
400 | return rv; | |
8453e704 N |
401 | } |
402 | ||
403 | ||
22e263f6 AC |
404 | const int uuid_zero[4] = { 0, 0, 0, 0 }; |
405 | ||
f277ce36 | 406 | int same_uuid(int a[4], int b[4], int swapuuid) |
64c4757e | 407 | { |
f277ce36 NB |
408 | if (swapuuid) { |
409 | /* parse uuids are hostendian. | |
410 | * uuid's from some superblocks are big-ending | |
aba69144 | 411 | * if there is a difference, we need to swap.. |
f277ce36 NB |
412 | */ |
413 | unsigned char *ac = (unsigned char *)a; | |
414 | unsigned char *bc = (unsigned char *)b; | |
415 | int i; | |
416 | for (i=0; i<16; i+= 4) { | |
417 | if (ac[i+0] != bc[i+3] || | |
418 | ac[i+1] != bc[i+2] || | |
419 | ac[i+2] != bc[i+1] || | |
420 | ac[i+3] != bc[i+0]) | |
421 | return 0; | |
422 | } | |
423 | return 1; | |
424 | } else { | |
425 | if (a[0]==b[0] && | |
426 | a[1]==b[1] && | |
427 | a[2]==b[2] && | |
428 | a[3]==b[3]) | |
429 | return 1; | |
430 | return 0; | |
431 | } | |
64c4757e | 432 | } |
350f29f9 NB |
433 | void copy_uuid(void *a, int b[4], int swapuuid) |
434 | { | |
435 | if (swapuuid) { | |
436 | /* parse uuids are hostendian. | |
437 | * uuid's from some superblocks are big-ending | |
438 | * if there is a difference, we need to swap.. | |
439 | */ | |
440 | unsigned char *ac = (unsigned char *)a; | |
441 | unsigned char *bc = (unsigned char *)b; | |
442 | int i; | |
443 | for (i=0; i<16; i+= 4) { | |
444 | ac[i+0] = bc[i+3]; | |
445 | ac[i+1] = bc[i+2]; | |
446 | ac[i+2] = bc[i+1]; | |
447 | ac[i+3] = bc[i+0]; | |
448 | } | |
449 | } else | |
450 | memcpy(a, b, 16); | |
451 | } | |
64c4757e | 452 | |
aae5a112 | 453 | char *__fname_from_uuid(int id[4], int swap, char *buf, char sep) |
d7288ddc | 454 | { |
9968e376 | 455 | int i, j; |
d7288ddc N |
456 | char uuid[16]; |
457 | char *c = buf; | |
458 | strcpy(c, "UUID-"); | |
459 | c += strlen(c); | |
aae5a112 | 460 | copy_uuid(uuid, id, swap); |
9968e376 | 461 | for (i = 0; i < 4; i++) { |
9968e376 | 462 | if (i) |
ff54de6e | 463 | *c++ = sep; |
9968e376 DW |
464 | for (j = 3; j >= 0; j--) { |
465 | sprintf(c,"%02x", (unsigned char) uuid[j+4*i]); | |
466 | c+= 2; | |
467 | } | |
d7288ddc N |
468 | } |
469 | return buf; | |
aae5a112 DW |
470 | |
471 | } | |
472 | ||
473 | char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep) | |
474 | { | |
86983cce N |
475 | // dirty hack to work around an issue with super1 superblocks... |
476 | // super1 superblocks need swapuuid set in order for assembly to | |
477 | // work, but can't have it set if we want this printout to match | |
478 | // all the other uuid printouts in super1.c, so we force swapuuid | |
479 | // to 1 to make our printout match the rest of super1 | |
480 | return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep); | |
d7288ddc N |
481 | } |
482 | ||
435d4ebb | 483 | #ifndef MDASSEMBLE |
682c7051 NB |
484 | int check_ext2(int fd, char *name) |
485 | { | |
486 | /* | |
487 | * Check for an ext2fs file system. | |
488 | * Superblock is always 1K at 1K offset | |
489 | * | |
490 | * s_magic is le16 at 56 == 0xEF53 | |
491 | * report mtime - le32 at 44 | |
492 | * blocks - le32 at 4 | |
493 | * logblksize - le32 at 24 | |
494 | */ | |
495 | unsigned char sb[1024]; | |
496 | time_t mtime; | |
497 | int size, bsize; | |
498 | if (lseek(fd, 1024,0)!= 1024) | |
499 | return 0; | |
500 | if (read(fd, sb, 1024)!= 1024) | |
501 | return 0; | |
502 | if (sb[56] != 0x53 || sb[57] != 0xef) | |
503 | return 0; | |
504 | ||
505 | mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8; | |
506 | bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8; | |
507 | size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8; | |
e7b84f9d | 508 | pr_err("%s appears to contain an ext2fs file system\n", |
682c7051 NB |
509 | name); |
510 | fprintf(stderr," size=%dK mtime=%s", | |
511 | size*(1<<bsize), ctime(&mtime)); | |
512 | return 1; | |
513 | } | |
514 | ||
515 | int check_reiser(int fd, char *name) | |
516 | { | |
517 | /* | |
518 | * superblock is at 64K | |
519 | * size is 1024; | |
520 | * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52 | |
521 | * | |
522 | */ | |
523 | unsigned char sb[1024]; | |
881990a2 | 524 | unsigned long size; |
682c7051 NB |
525 | if (lseek(fd, 64*1024, 0) != 64*1024) |
526 | return 0; | |
527 | if (read(fd, sb, 1024) != 1024) | |
528 | return 0; | |
a46f4061 NB |
529 | if (strncmp((char*)sb+52, "ReIsErFs",8)!=0 && |
530 | strncmp((char*)sb+52, "ReIsEr2Fs",9)!=0) | |
682c7051 | 531 | return 0; |
e7b84f9d | 532 | pr_err("%s appears to contain a reiserfs file system\n",name); |
682c7051 | 533 | size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8; |
881990a2 | 534 | fprintf(stderr, " size = %luK\n", size*4); |
aba69144 | 535 | |
682c7051 NB |
536 | return 1; |
537 | } | |
538 | ||
539 | int check_raid(int fd, char *name) | |
540 | { | |
4b1ac34b | 541 | struct mdinfo info; |
682c7051 | 542 | time_t crtime; |
d078d77c | 543 | char *level; |
82d9eba6 | 544 | struct supertype *st = guess_super(fd); |
f9ce90ba | 545 | |
82d9eba6 | 546 | if (!st) return 0; |
4a96d9ff | 547 | st->ignore_hw_compat = 1; |
3da92f27 | 548 | st->ss->load_super(st, fd, name); |
82d9eba6 | 549 | /* Looks like a raid array .. */ |
e7b84f9d | 550 | pr_err("%s appears to be part of a raid array:\n", |
82d9eba6 | 551 | name); |
a5d85af7 | 552 | st->ss->getinfo_super(st, &info, NULL); |
3da92f27 | 553 | st->ss->free_super(st); |
82d9eba6 | 554 | crtime = info.array.ctime; |
d078d77c NB |
555 | level = map_num(pers, info.array.level); |
556 | if (!level) level = "-unknown-"; | |
557 | fprintf(stderr, " level=%s devices=%d ctime=%s", | |
558 | level, info.array.raid_disks, ctime(&crtime)); | |
82d9eba6 | 559 | return 1; |
682c7051 NB |
560 | } |
561 | ||
682c7051 NB |
562 | int ask(char *mesg) |
563 | { | |
564 | char *add = ""; | |
565 | int i; | |
566 | for (i=0; i<5; i++) { | |
567 | char buf[100]; | |
568 | fprintf(stderr, "%s%s", mesg, add); | |
569 | fflush(stderr); | |
570 | if (fgets(buf, 100, stdin)==NULL) | |
571 | return 0; | |
572 | if (buf[0]=='y' || buf[0]=='Y') | |
573 | return 1; | |
574 | if (buf[0]=='n' || buf[0]=='N') | |
575 | return 0; | |
576 | add = "(y/n) "; | |
577 | } | |
e7b84f9d | 578 | pr_err("assuming 'no'\n"); |
682c7051 NB |
579 | return 0; |
580 | } | |
435d4ebb | 581 | #endif /* MDASSEMBLE */ |
682c7051 | 582 | |
8d80900b | 583 | int is_standard(char *dev, int *nump) |
e5329c37 NB |
584 | { |
585 | /* tests if dev is a "standard" md dev name. | |
586 | * i.e if the last component is "/dNN" or "/mdNN", | |
aba69144 | 587 | * where NN is a string of digits |
598f0d58 NB |
588 | * Returns 1 if a partitionable standard, |
589 | * -1 if non-partitonable, | |
590 | * 0 if not a standard name. | |
e5329c37 | 591 | */ |
8d80900b NB |
592 | char *d = strrchr(dev, '/'); |
593 | int type=0; | |
594 | int num; | |
595 | if (!d) | |
e5329c37 | 596 | return 0; |
8d80900b NB |
597 | if (strncmp(d, "/d",2)==0) |
598 | d += 2, type=1; /* /dev/md/dN{pM} */ | |
599 | else if (strncmp(d, "/md_d", 5)==0) | |
2b4ca8f0 | 600 | d += 5, type=1; /* /dev/md_dN{pM} */ |
8d80900b NB |
601 | else if (strncmp(d, "/md", 3)==0) |
602 | d += 3, type=-1; /* /dev/mdN */ | |
603 | else if (d-dev > 3 && strncmp(d-2, "md/", 3)==0) | |
5a6d1148 | 604 | d += 1, type=-1; /* /dev/md/N */ |
e5329c37 NB |
605 | else |
606 | return 0; | |
8d80900b | 607 | if (!*d) |
e5329c37 | 608 | return 0; |
8d80900b NB |
609 | num = atoi(d); |
610 | while (isdigit(*d)) | |
611 | d++; | |
612 | if (*d) | |
e5329c37 | 613 | return 0; |
8d80900b NB |
614 | if (nump) *nump = num; |
615 | ||
616 | return type; | |
e5329c37 NB |
617 | } |
618 | ||
4b1ac34b | 619 | unsigned long calc_csum(void *super, int bytes) |
82b27616 | 620 | { |
56eb10c0 | 621 | unsigned long long newcsum = 0; |
82b27616 | 622 | int i; |
4b1ac34b NB |
623 | unsigned int csum; |
624 | unsigned int *superc = (unsigned int*) super; | |
82b27616 | 625 | |
4b1ac34b | 626 | for(i=0; i<bytes/4; i++) |
82b27616 NB |
627 | newcsum+= superc[i]; |
628 | csum = (newcsum& 0xffffffff) + (newcsum>>32); | |
570c0542 | 629 | #ifdef __alpha__ |
aba69144 | 630 | /* The in-kernel checksum calculation is always 16bit on |
570c0542 NB |
631 | * the alpha, though it is 32 bit on i386... |
632 | * I wonder what it is elsewhere... (it uses and API in | |
633 | * a way that it shouldn't). | |
634 | */ | |
635 | csum = (csum & 0xffff) + (csum >> 16); | |
636 | csum = (csum & 0xffff) + (csum >> 16); | |
637 | #endif | |
82b27616 NB |
638 | return csum; |
639 | } | |
cd29a5c8 | 640 | |
435d4ebb | 641 | #ifndef MDASSEMBLE |
56eb10c0 | 642 | char *human_size(long long bytes) |
cd29a5c8 NB |
643 | { |
644 | static char buf[30]; | |
d5d3721e NB |
645 | |
646 | /* We convert bytes to either centi-M{ega,ibi}bytes or | |
647 | * centi-G{igi,ibi}bytes, with appropriate rounding, | |
648 | * and then print 1/100th of those as a decimal. | |
649 | * We allow upto 2048Megabytes before converting to | |
650 | * gigabytes, as that shows more precision and isn't | |
651 | * too large a number. | |
01619b48 | 652 | * Terabytes are not yet handled. |
d5d3721e | 653 | */ |
cd29a5c8 | 654 | |
56eb10c0 | 655 | if (bytes < 5000*1024) |
cd29a5c8 | 656 | buf[0]=0; |
d5d3721e NB |
657 | else if (bytes < 2*1024LL*1024LL*1024LL) { |
658 | long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2; | |
659 | long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2; | |
8f23b0b3 | 660 | snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)", |
d5d3721e NB |
661 | cMiB/100 , cMiB % 100, |
662 | cMB/100, cMB % 100); | |
663 | } else { | |
664 | long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2; | |
665 | long cGB = (bytes / (1000000000LL/200LL ) +1) /2; | |
8f23b0b3 | 666 | snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)", |
d5d3721e NB |
667 | cGiB/100 , cGiB % 100, |
668 | cGB/100, cGB % 100); | |
669 | } | |
cd29a5c8 NB |
670 | return buf; |
671 | } | |
e0d19036 NB |
672 | |
673 | char *human_size_brief(long long bytes) | |
674 | { | |
675 | static char buf[30]; | |
e0d19036 NB |
676 | |
677 | if (bytes < 5000*1024) | |
8f23b0b3 | 678 | snprintf(buf, sizeof(buf), "%ld.%02ldKiB", |
bd526cee | 679 | (long)(bytes>>10), (long)(((bytes&1023)*100+512)/1024) |
e0d19036 NB |
680 | ); |
681 | else if (bytes < 2*1024LL*1024LL*1024LL) | |
8f23b0b3 | 682 | snprintf(buf, sizeof(buf), "%ld.%02ldMiB", |
e0d19036 | 683 | (long)(bytes>>20), |
bd526cee | 684 | (long)((bytes&0xfffff)+0x100000/200)/(0x100000/100) |
e0d19036 NB |
685 | ); |
686 | else | |
8f23b0b3 | 687 | snprintf(buf, sizeof(buf), "%ld.%02ldGiB", |
e0d19036 | 688 | (long)(bytes>>30), |
bd526cee | 689 | (long)(((bytes>>10)&0xfffff)+0x100000/200)/(0x100000/100) |
e0d19036 NB |
690 | ); |
691 | return buf; | |
692 | } | |
e4965ef8 N |
693 | |
694 | void print_r10_layout(int layout) | |
695 | { | |
696 | int near = layout & 255; | |
697 | int far = (layout >> 8) & 255; | |
698 | int offset = (layout&0x10000); | |
699 | char *sep = ""; | |
700 | ||
701 | if (near != 1) { | |
702 | printf("%s near=%d", sep, near); | |
703 | sep = ","; | |
704 | } | |
705 | if (far != 1) | |
706 | printf("%s %s=%d", sep, offset?"offset":"far", far); | |
707 | if (near*far == 1) | |
708 | printf("NO REDUNDANCY"); | |
709 | } | |
435d4ebb | 710 | #endif |
e0d19036 | 711 | |
5f8097be NB |
712 | unsigned long long calc_array_size(int level, int raid_disks, int layout, |
713 | int chunksize, unsigned long long devsize) | |
577e8448 | 714 | { |
e03640bd CA |
715 | if (level == 1) |
716 | return devsize; | |
577e8448 AK |
717 | devsize &= ~(unsigned long long)((chunksize>>9)-1); |
718 | return get_data_disks(level, layout, raid_disks) * devsize; | |
719 | } | |
720 | ||
721 | int get_data_disks(int level, int layout, int raid_disks) | |
5f8097be NB |
722 | { |
723 | int data_disks = 0; | |
724 | switch (level) { | |
725 | case 0: data_disks = raid_disks; break; | |
726 | case 1: data_disks = 1; break; | |
727 | case 4: | |
728 | case 5: data_disks = raid_disks - 1; break; | |
729 | case 6: data_disks = raid_disks - 2; break; | |
730 | case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255); | |
731 | break; | |
732 | } | |
577e8448 AK |
733 | |
734 | return data_disks; | |
5f8097be NB |
735 | } |
736 | ||
0e600426 | 737 | #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) |
e0d19036 NB |
738 | char *get_md_name(int dev) |
739 | { | |
740 | /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */ | |
98c6faba | 741 | /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */ |
e0d19036 NB |
742 | static char devname[50]; |
743 | struct stat stb; | |
98c6faba | 744 | dev_t rdev; |
dd0781e5 | 745 | char *dn; |
98c6faba NB |
746 | |
747 | if (dev < 0) { | |
dd0781e5 NB |
748 | int mdp = get_mdp_major(); |
749 | if (mdp < 0) return NULL; | |
0df46c2a | 750 | rdev = makedev(mdp, (-1-dev)<<6); |
8f23b0b3 | 751 | snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev); |
98c6faba NB |
752 | if (stat(devname, &stb) == 0 |
753 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
754 | && (stb.st_rdev == rdev)) | |
755 | return devname; | |
756 | } else { | |
0df46c2a | 757 | rdev = makedev(MD_MAJOR, dev); |
8f23b0b3 | 758 | snprintf(devname, sizeof(devname), "/dev/md%d", dev); |
98c6faba NB |
759 | if (stat(devname, &stb) == 0 |
760 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
761 | && (stb.st_rdev == rdev)) | |
762 | return devname; | |
763 | ||
8f23b0b3 | 764 | snprintf(devname, sizeof(devname), "/dev/md/%d", dev); |
98c6faba NB |
765 | if (stat(devname, &stb) == 0 |
766 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
767 | && (stb.st_rdev == rdev)) | |
768 | return devname; | |
769 | } | |
16c6fa80 | 770 | dn = map_dev(major(rdev), minor(rdev), 0); |
dd0781e5 NB |
771 | if (dn) |
772 | return dn; | |
8f23b0b3 | 773 | snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev); |
e0d19036 | 774 | if (mknod(devname, S_IFBLK | 0600, rdev) == -1) |
dd0781e5 NB |
775 | if (errno != EEXIST) |
776 | return NULL; | |
e0d19036 NB |
777 | |
778 | if (stat(devname, &stb) == 0 | |
779 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
780 | && (stb.st_rdev == rdev)) | |
781 | return devname; | |
782 | unlink(devname); | |
783 | return NULL; | |
784 | } | |
785 | ||
786 | void put_md_name(char *name) | |
787 | { | |
788 | if (strncmp(name, "/dev/.tmp.md", 12)==0) | |
789 | unlink(name); | |
790 | } | |
ea24acd0 | 791 | |
ea24acd0 NB |
792 | int find_free_devnum(int use_partitions) |
793 | { | |
794 | int devnum; | |
795 | for (devnum = 127; devnum != 128; | |
a56fb7ec | 796 | devnum = devnum ? devnum-1 : (1<<20)-1) { |
ea24acd0 NB |
797 | char *dn; |
798 | int _devnum; | |
799 | ||
800 | _devnum = use_partitions ? (-1-devnum) : devnum; | |
801 | if (mddev_busy(_devnum)) | |
802 | continue; | |
803 | /* make sure it is new to /dev too, at least as a | |
804 | * non-standard */ | |
805 | dn = map_dev(dev2major(_devnum), dev2minor(_devnum), 0); | |
806 | if (dn && ! is_standard(dn, NULL)) | |
807 | continue; | |
808 | break; | |
809 | } | |
810 | if (devnum == 128) | |
811 | return NoMdDev; | |
812 | return use_partitions ? (-1-devnum) : devnum; | |
813 | } | |
435d4ebb | 814 | #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */ |
f9ce90ba | 815 | |
8b0dabea NB |
816 | int dev_open(char *dev, int flags) |
817 | { | |
818 | /* like 'open', but if 'dev' matches %d:%d, create a temp | |
819 | * block device and open that | |
820 | */ | |
821 | char *e; | |
822 | int fd = -1; | |
823 | char devname[32]; | |
e81cdd9f | 824 | int major; |
8b0dabea | 825 | int minor; |
e81cdd9f NB |
826 | |
827 | if (!dev) return -1; | |
6df6a774 | 828 | flags |= O_DIRECT; |
e81cdd9f NB |
829 | |
830 | major = strtoul(dev, &e, 0); | |
8b0dabea NB |
831 | if (e > dev && *e == ':' && e[1] && |
832 | (minor = strtoul(e+1, &e, 0)) >= 0 && | |
833 | *e == 0) { | |
6df6a774 N |
834 | char *path = map_dev(major, minor, 0); |
835 | if (path) | |
836 | fd = open(path, flags); | |
837 | if (fd < 0) { | |
838 | snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d", | |
839 | (int)getpid(), major, minor); | |
840 | if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) { | |
841 | fd = open(devname, flags); | |
842 | unlink(devname); | |
843 | } | |
844 | } | |
845 | if (fd < 0) { | |
846 | snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d", | |
847 | (int)getpid(), major, minor); | |
848 | if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) { | |
849 | fd = open(devname, flags); | |
850 | unlink(devname); | |
851 | } | |
8b0dabea NB |
852 | } |
853 | } else | |
6df6a774 | 854 | fd = open(dev, flags); |
8b0dabea NB |
855 | return fd; |
856 | } | |
f9ce90ba | 857 | |
d998b738 | 858 | int open_dev_flags(int devnum, int flags) |
e8a70c89 N |
859 | { |
860 | char buf[20]; | |
861 | ||
862 | sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum)); | |
d998b738 N |
863 | return dev_open(buf, flags); |
864 | } | |
865 | ||
866 | int open_dev(int devnum) | |
867 | { | |
868 | return open_dev_flags(devnum, O_RDONLY); | |
e8a70c89 N |
869 | } |
870 | ||
a931db9e NB |
871 | int open_dev_excl(int devnum) |
872 | { | |
873 | char buf[20]; | |
874 | int i; | |
7187750e | 875 | int flags = O_RDWR; |
a931db9e NB |
876 | |
877 | sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum)); | |
878 | for (i=0 ; i<25 ; i++) { | |
7187750e | 879 | int fd = dev_open(buf, flags|O_EXCL); |
a931db9e NB |
880 | if (fd >= 0) |
881 | return fd; | |
7187750e N |
882 | if (errno == EACCES && flags == O_RDWR) { |
883 | flags = O_RDONLY; | |
884 | continue; | |
885 | } | |
a931db9e NB |
886 | if (errno != EBUSY) |
887 | return fd; | |
888 | usleep(200000); | |
889 | } | |
890 | return -1; | |
891 | } | |
892 | ||
9008ed1c N |
893 | int same_dev(char *one, char *two) |
894 | { | |
895 | struct stat st1, st2; | |
896 | if (stat(one, &st1) != 0) | |
897 | return 0; | |
898 | if (stat(two, &st2) != 0) | |
899 | return 0; | |
900 | if ((st1.st_mode & S_IFMT) != S_IFBLK) | |
901 | return 0; | |
902 | if ((st2.st_mode & S_IFMT) != S_IFBLK) | |
903 | return 0; | |
904 | return st1.st_rdev == st2.st_rdev; | |
905 | } | |
906 | ||
a7c6e3fb | 907 | void wait_for(char *dev, int fd) |
a714580e N |
908 | { |
909 | int i; | |
a7c6e3fb N |
910 | struct stat stb_want; |
911 | ||
912 | if (fstat(fd, &stb_want) != 0 || | |
913 | (stb_want.st_mode & S_IFMT) != S_IFBLK) | |
914 | return; | |
a714580e N |
915 | |
916 | for (i=0 ; i<25 ; i++) { | |
917 | struct stat stb; | |
a7c6e3fb N |
918 | if (stat(dev, &stb) == 0 && |
919 | (stb.st_mode & S_IFMT) == S_IFBLK && | |
920 | (stb.st_rdev == stb_want.st_rdev)) | |
a714580e N |
921 | return; |
922 | usleep(200000); | |
923 | } | |
436305c6 DW |
924 | if (i == 25) |
925 | dprintf("%s: timeout waiting for %s\n", __func__, dev); | |
a714580e N |
926 | } |
927 | ||
0f22b998 N |
928 | struct superswitch *superlist[] = |
929 | { | |
930 | &super0, &super1, | |
931 | &super_ddf, &super_imsm, | |
0592faeb | 932 | &mbr, &gpt, |
0f22b998 | 933 | NULL }; |
f9ce90ba | 934 | |
ea24acd0 | 935 | #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) |
f7dd881f | 936 | |
4725bc31 | 937 | struct supertype *super_by_fd(int fd, char **subarrayp) |
f9ce90ba | 938 | { |
1686dc25 NB |
939 | mdu_array_info_t array; |
940 | int vers; | |
941 | int minor; | |
942 | struct supertype *st = NULL; | |
7e0f6979 | 943 | struct mdinfo *sra; |
142cb9e1 | 944 | char *verstr; |
1686dc25 NB |
945 | char version[20]; |
946 | int i; | |
f7e7067b | 947 | char *subarray = NULL; |
5f7e44b2 | 948 | int container = NoMdDev; |
1686dc25 NB |
949 | |
950 | sra = sysfs_read(fd, 0, GET_VERSION); | |
951 | ||
952 | if (sra) { | |
7e0f6979 NB |
953 | vers = sra->array.major_version; |
954 | minor = sra->array.minor_version; | |
142cb9e1 | 955 | verstr = sra->text_version; |
1686dc25 NB |
956 | } else { |
957 | if (ioctl(fd, GET_ARRAY_INFO, &array)) | |
958 | array.major_version = array.minor_version = 0; | |
959 | vers = array.major_version; | |
960 | minor = array.minor_version; | |
142cb9e1 | 961 | verstr = ""; |
6fbba4c9 | 962 | } |
82d9eba6 | 963 | |
1686dc25 NB |
964 | if (vers != -1) { |
965 | sprintf(version, "%d.%d", vers, minor); | |
966 | verstr = version; | |
6fbba4c9 | 967 | } |
3c558363 | 968 | if (minor == -2 && is_subarray(verstr)) { |
f7e7067b | 969 | char *dev = verstr+1; |
5f7e44b2 | 970 | |
f7e7067b | 971 | subarray = strchr(dev, '/'); |
92d49ecf | 972 | if (subarray) { |
f7e7067b | 973 | *subarray++ = '\0'; |
503975b9 | 974 | subarray = xstrdup(subarray); |
92d49ecf | 975 | } |
5f7e44b2 | 976 | container = devname2devnum(dev); |
f7e7067b NB |
977 | if (sra) |
978 | sysfs_free(sra); | |
5f7e44b2 | 979 | sra = sysfs_read(-1, container, GET_VERSION); |
603f24a0 N |
980 | if (sra && sra->text_version[0]) |
981 | verstr = sra->text_version; | |
982 | else | |
983 | verstr = "-no-metadata-"; | |
f7e7067b NB |
984 | } |
985 | ||
986 | for (i = 0; st == NULL && superlist[i] ; i++) | |
987 | st = superlist[i]->match_metadata_desc(verstr); | |
1686dc25 NB |
988 | |
989 | if (sra) | |
990 | sysfs_free(sra); | |
f7e7067b | 991 | if (st) { |
3b0896f8 | 992 | st->sb = NULL; |
1f49fb3a N |
993 | if (subarrayp) |
994 | *subarrayp = subarray; | |
5f7e44b2 DW |
995 | st->container_dev = container; |
996 | st->devnum = fd2devnum(fd); | |
4725bc31 N |
997 | } else |
998 | free(subarray); | |
5f7e44b2 | 999 | |
82d9eba6 | 1000 | return st; |
f9ce90ba | 1001 | } |
ea24acd0 NB |
1002 | #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */ |
1003 | ||
0f0749ad | 1004 | int dev_size_from_id(dev_t id, unsigned long long *size) |
52d5d101 AC |
1005 | { |
1006 | char buf[20]; | |
1007 | int fd; | |
1008 | ||
1009 | sprintf(buf, "%d:%d", major(id), minor(id)); | |
1010 | fd = dev_open(buf, O_RDONLY); | |
1011 | if (fd < 0) | |
1012 | return 0; | |
1013 | if (get_dev_size(fd, NULL, size)) { | |
1014 | close(fd); | |
1015 | return 1; | |
1016 | } | |
1017 | close(fd); | |
1018 | return 0; | |
1019 | } | |
f9ce90ba | 1020 | |
159c3a1a | 1021 | struct supertype *dup_super(struct supertype *orig) |
3da92f27 | 1022 | { |
159c3a1a | 1023 | struct supertype *st; |
1686dc25 | 1024 | |
d2ca6449 NB |
1025 | if (!orig) |
1026 | return orig; | |
503975b9 | 1027 | st = xcalloc(1, sizeof(*st)); |
159c3a1a NB |
1028 | st->ss = orig->ss; |
1029 | st->max_devs = orig->max_devs; | |
1030 | st->minor_version = orig->minor_version; | |
1031 | st->sb = NULL; | |
1032 | st->info = NULL; | |
1033 | return st; | |
3da92f27 NB |
1034 | } |
1035 | ||
54887ad8 | 1036 | struct supertype *guess_super_type(int fd, enum guess_types guess_type) |
f9ce90ba NB |
1037 | { |
1038 | /* try each load_super to find the best match, | |
1039 | * and return the best superswitch | |
1040 | */ | |
82d9eba6 NB |
1041 | struct superswitch *ss; |
1042 | struct supertype *st; | |
f21e18ca | 1043 | time_t besttime = 0; |
570c0542 | 1044 | int bestsuper = -1; |
f9ce90ba NB |
1045 | int i; |
1046 | ||
503975b9 | 1047 | st = xcalloc(1, sizeof(*st)); |
d1d599ea N |
1048 | st->container_dev = NoMdDev; |
1049 | ||
f9ce90ba NB |
1050 | for (i=0 ; superlist[i]; i++) { |
1051 | int rv; | |
1052 | ss = superlist[i]; | |
54887ad8 N |
1053 | if (guess_type == guess_array && ss->add_to_super == NULL) |
1054 | continue; | |
1055 | if (guess_type == guess_partitions && ss->add_to_super != NULL) | |
1056 | continue; | |
ef609477 | 1057 | memset(st, 0, sizeof(*st)); |
df3346e6 | 1058 | st->ignore_hw_compat = 1; |
3da92f27 | 1059 | rv = ss->load_super(st, fd, NULL); |
570c0542 NB |
1060 | if (rv == 0) { |
1061 | struct mdinfo info; | |
a5d85af7 | 1062 | st->ss->getinfo_super(st, &info, NULL); |
570c0542 NB |
1063 | if (bestsuper == -1 || |
1064 | besttime < info.array.ctime) { | |
1065 | bestsuper = i; | |
1066 | besttime = info.array.ctime; | |
570c0542 | 1067 | } |
3da92f27 | 1068 | ss->free_super(st); |
570c0542 NB |
1069 | } |
1070 | } | |
1071 | if (bestsuper != -1) { | |
1072 | int rv; | |
ef609477 | 1073 | memset(st, 0, sizeof(*st)); |
df3346e6 | 1074 | st->ignore_hw_compat = 1; |
3da92f27 | 1075 | rv = superlist[bestsuper]->load_super(st, fd, NULL); |
f9ce90ba | 1076 | if (rv == 0) { |
5e747af2 | 1077 | superlist[bestsuper]->free_super(st); |
df3346e6 | 1078 | st->ignore_hw_compat = 0; |
82d9eba6 | 1079 | return st; |
f9ce90ba NB |
1080 | } |
1081 | } | |
570c0542 | 1082 | free(st); |
f9ce90ba NB |
1083 | return NULL; |
1084 | } | |
fe6729fa | 1085 | |
beae1dfe NB |
1086 | /* Return size of device in bytes */ |
1087 | int get_dev_size(int fd, char *dname, unsigned long long *sizep) | |
1088 | { | |
1089 | unsigned long long ldsize; | |
c2c9bb6f NB |
1090 | struct stat st; |
1091 | ||
1092 | if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode)) | |
1093 | ldsize = (unsigned long long)st.st_size; | |
1094 | else | |
beae1dfe NB |
1095 | #ifdef BLKGETSIZE64 |
1096 | if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0) | |
1097 | #endif | |
1098 | { | |
1099 | unsigned long dsize; | |
1100 | if (ioctl(fd, BLKGETSIZE, &dsize) == 0) { | |
1101 | ldsize = dsize; | |
1102 | ldsize <<= 9; | |
1103 | } else { | |
1104 | if (dname) | |
e7b84f9d | 1105 | pr_err("Cannot get size of %s: %s\b", |
beae1dfe NB |
1106 | dname, strerror(errno)); |
1107 | return 0; | |
1108 | } | |
1109 | } | |
1110 | *sizep = ldsize; | |
1111 | return 1; | |
1112 | } | |
8fac0577 | 1113 | |
3a371610 N |
1114 | /* Return true if this can only be a container, not a member device. |
1115 | * i.e. is and md device and size is zero | |
1116 | */ | |
1117 | int must_be_container(int fd) | |
1118 | { | |
1119 | unsigned long long size; | |
1120 | if (md_get_version(fd) < 0) | |
1121 | return 0; | |
1122 | if (get_dev_size(fd, NULL, &size) == 0) | |
1123 | return 1; | |
1124 | if (size == 0) | |
1125 | return 1; | |
1126 | return 0; | |
1127 | } | |
034b203a TM |
1128 | |
1129 | /* Sets endofpart parameter to the last block used by the last GPT partition on the device. | |
1130 | * Returns: 1 if successful | |
1131 | * -1 for unknown partition type | |
1132 | * 0 for other errors | |
1133 | */ | |
1134 | static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart) | |
1135 | { | |
056b331e | 1136 | struct GPT gpt; |
034b203a | 1137 | unsigned char empty_gpt_entry[16]= {0}; |
db7fdfe4 JS |
1138 | struct GPT_part_entry *part; |
1139 | char buf[512]; | |
034b203a TM |
1140 | unsigned long long curr_part_end; |
1141 | unsigned all_partitions, entry_size; | |
f21e18ca | 1142 | unsigned part_nr; |
034b203a TM |
1143 | |
1144 | *endofpart = 0; | |
1145 | ||
056b331e | 1146 | BUILD_BUG_ON(sizeof(gpt) != 512); |
73e658d8 | 1147 | /* skip protective MBR */ |
034b203a | 1148 | lseek(fd, 512, SEEK_SET); |
73e658d8 | 1149 | /* read GPT header */ |
056b331e | 1150 | if (read(fd, &gpt, 512) != 512) |
034b203a TM |
1151 | return 0; |
1152 | ||
1153 | /* get the number of partition entries and the entry size */ | |
056b331e N |
1154 | all_partitions = __le32_to_cpu(gpt.part_cnt); |
1155 | entry_size = __le32_to_cpu(gpt.part_size); | |
034b203a TM |
1156 | |
1157 | /* Check GPT signature*/ | |
056b331e | 1158 | if (gpt.magic != GPT_SIGNATURE_MAGIC) |
034b203a TM |
1159 | return -1; |
1160 | ||
1161 | /* sanity checks */ | |
1162 | if (all_partitions > 1024 || | |
db7fdfe4 | 1163 | entry_size > sizeof(buf)) |
034b203a TM |
1164 | return -1; |
1165 | ||
db7fdfe4 JS |
1166 | part = (struct GPT_part_entry *)buf; |
1167 | ||
034b203a | 1168 | for (part_nr=0; part_nr < all_partitions; part_nr++) { |
73e658d8 | 1169 | /* read partition entry */ |
db7fdfe4 | 1170 | if (read(fd, buf, entry_size) != (ssize_t)entry_size) |
73e658d8 LB |
1171 | return 0; |
1172 | ||
034b203a | 1173 | /* is this valid partition? */ |
db7fdfe4 | 1174 | if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) { |
034b203a | 1175 | /* check the last lba for the current partition */ |
db7fdfe4 | 1176 | curr_part_end = __le64_to_cpu(part->ending_lba); |
034b203a TM |
1177 | if (curr_part_end > *endofpart) |
1178 | *endofpart = curr_part_end; | |
1179 | } | |
1180 | ||
034b203a TM |
1181 | } |
1182 | return 1; | |
1183 | } | |
1184 | ||
1185 | /* Sets endofpart parameter to the last block used by the last partition on the device. | |
1186 | * Returns: 1 if successful | |
1187 | * -1 for unknown partition type | |
1188 | * 0 for other errors | |
1189 | */ | |
1190 | static int get_last_partition_end(int fd, unsigned long long *endofpart) | |
1191 | { | |
056b331e | 1192 | struct MBR boot_sect; |
034b203a TM |
1193 | struct MBR_part_record *part; |
1194 | unsigned long long curr_part_end; | |
f21e18ca | 1195 | unsigned part_nr; |
034b203a TM |
1196 | int retval = 0; |
1197 | ||
1198 | *endofpart = 0; | |
1199 | ||
056b331e | 1200 | BUILD_BUG_ON(sizeof(boot_sect) != 512); |
034b203a TM |
1201 | /* read MBR */ |
1202 | lseek(fd, 0, 0); | |
056b331e | 1203 | if (read(fd, &boot_sect, 512) != 512) |
034b203a TM |
1204 | goto abort; |
1205 | ||
1206 | /* check MBP signature */ | |
056b331e | 1207 | if (boot_sect.magic == MBR_SIGNATURE_MAGIC) { |
034b203a TM |
1208 | retval = 1; |
1209 | /* found the correct signature */ | |
056b331e | 1210 | part = boot_sect.parts; |
034b203a TM |
1211 | |
1212 | for (part_nr=0; part_nr < MBR_PARTITIONS; part_nr++) { | |
1213 | /* check for GPT type */ | |
1214 | if (part->part_type == MBR_GPT_PARTITION_TYPE) { | |
1215 | retval = get_gpt_last_partition_end(fd, endofpart); | |
1216 | break; | |
1217 | } | |
1218 | /* check the last used lba for the current partition */ | |
1219 | curr_part_end = __le32_to_cpu(part->first_sect_lba) + | |
1220 | __le32_to_cpu(part->blocks_num); | |
1221 | if (curr_part_end > *endofpart) | |
1222 | *endofpart = curr_part_end; | |
1223 | ||
1224 | part++; | |
1225 | } | |
1226 | } else { | |
1227 | /* Unknown partition table */ | |
1228 | retval = -1; | |
1229 | } | |
1230 | abort: | |
1231 | return retval; | |
1232 | } | |
1233 | ||
53ed6ac3 KW |
1234 | int check_partitions(int fd, char *dname, unsigned long long freesize, |
1235 | unsigned long long size) | |
034b203a TM |
1236 | { |
1237 | /* | |
1238 | * Check where the last partition ends | |
1239 | */ | |
1240 | unsigned long long endofpart; | |
1241 | int ret; | |
1242 | ||
1243 | if ((ret = get_last_partition_end(fd, &endofpart)) > 0) { | |
1244 | /* There appears to be a partition table here */ | |
1245 | if (freesize == 0) { | |
1246 | /* partitions will not be visible in new device */ | |
e7b84f9d N |
1247 | pr_err("partition table exists on %s but will be lost or\n" |
1248 | " meaningless after creating array\n", | |
1249 | dname); | |
034b203a TM |
1250 | return 1; |
1251 | } else if (endofpart > freesize) { | |
1252 | /* last partition overlaps metadata */ | |
e7b84f9d N |
1253 | pr_err("metadata will over-write last partition on %s.\n", |
1254 | dname); | |
034b203a | 1255 | return 1; |
53ed6ac3 KW |
1256 | } else if (size && endofpart > size) { |
1257 | /* partitions will be truncated in new device */ | |
e7b84f9d N |
1258 | pr_err("array size is too small to cover all partitions on %s.\n", |
1259 | dname); | |
53ed6ac3 | 1260 | return 1; |
034b203a TM |
1261 | } |
1262 | } | |
1263 | return 0; | |
1264 | } | |
1265 | ||
8382f19b NB |
1266 | void get_one_disk(int mdfd, mdu_array_info_t *ainf, mdu_disk_info_t *disk) |
1267 | { | |
1268 | int d; | |
9e6d9291 | 1269 | |
8382f19b | 1270 | ioctl(mdfd, GET_ARRAY_INFO, ainf); |
480f3566 | 1271 | for (d = 0 ; d < MAX_DISKS ; d++) { |
9e6d9291 N |
1272 | if (ioctl(mdfd, GET_DISK_INFO, disk) == 0 && |
1273 | (disk->major || disk->minor)) | |
8382f19b | 1274 | return; |
9e6d9291 | 1275 | } |
8382f19b | 1276 | } |
63152c1b | 1277 | |
a322f70c DW |
1278 | int open_container(int fd) |
1279 | { | |
1280 | /* 'fd' is a block device. Find out if it is in use | |
1281 | * by a container, and return an open fd on that container. | |
1282 | */ | |
1283 | char path[256]; | |
1284 | char *e; | |
1285 | DIR *dir; | |
1286 | struct dirent *de; | |
1287 | int dfd, n; | |
1288 | char buf[200]; | |
1289 | int major, minor; | |
1290 | struct stat st; | |
1291 | ||
1292 | if (fstat(fd, &st) != 0) | |
1293 | return -1; | |
1294 | sprintf(path, "/sys/dev/block/%d:%d/holders", | |
1295 | (int)major(st.st_rdev), (int)minor(st.st_rdev)); | |
1296 | e = path + strlen(path); | |
1297 | ||
1298 | dir = opendir(path); | |
1299 | if (!dir) | |
1300 | return -1; | |
1301 | while ((de = readdir(dir))) { | |
1302 | if (de->d_ino == 0) | |
1303 | continue; | |
1304 | if (de->d_name[0] == '.') | |
1305 | continue; | |
1306 | sprintf(e, "/%s/dev", de->d_name); | |
1307 | dfd = open(path, O_RDONLY); | |
1308 | if (dfd < 0) | |
1309 | continue; | |
1310 | n = read(dfd, buf, sizeof(buf)); | |
1311 | close(dfd); | |
f21e18ca | 1312 | if (n <= 0 || (unsigned)n >= sizeof(buf)) |
a322f70c DW |
1313 | continue; |
1314 | buf[n] = 0; | |
1315 | if (sscanf(buf, "%d:%d", &major, &minor) != 2) | |
1316 | continue; | |
1317 | sprintf(buf, "%d:%d", major, minor); | |
1318 | dfd = dev_open(buf, O_RDONLY); | |
1319 | if (dfd >= 0) { | |
1320 | closedir(dir); | |
1321 | return dfd; | |
1322 | } | |
1323 | } | |
355726fa | 1324 | closedir(dir); |
a322f70c DW |
1325 | return -1; |
1326 | } | |
1327 | ||
33414a01 DW |
1328 | struct superswitch *version_to_superswitch(char *vers) |
1329 | { | |
1330 | int i; | |
1331 | ||
1332 | for (i = 0; superlist[i]; i++) { | |
1333 | struct superswitch *ss = superlist[i]; | |
1334 | ||
1335 | if (strcmp(vers, ss->name) == 0) | |
1336 | return ss; | |
1337 | } | |
1338 | ||
1339 | return NULL; | |
1340 | } | |
1341 | ||
1342 | int is_container_member(struct mdstat_ent *mdstat, char *container) | |
1343 | { | |
1344 | if (mdstat->metadata_version == NULL || | |
1345 | strncmp(mdstat->metadata_version, "external:", 9) != 0 || | |
1346 | !is_subarray(mdstat->metadata_version+9) || | |
1347 | strncmp(mdstat->metadata_version+10, container, strlen(container)) != 0 || | |
1348 | mdstat->metadata_version[10+strlen(container)] != '/') | |
1349 | return 0; | |
1350 | ||
1351 | return 1; | |
1352 | } | |
1353 | ||
1354 | int is_subarray_active(char *subarray, char *container) | |
1355 | { | |
1356 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
1357 | struct mdstat_ent *ent; | |
1358 | ||
e5408a32 DW |
1359 | for (ent = mdstat; ent; ent = ent->next) |
1360 | if (is_container_member(ent, container)) | |
e5e5d7ce | 1361 | if (strcmp(to_subarray(ent, container), subarray) == 0) |
33414a01 | 1362 | break; |
33414a01 DW |
1363 | |
1364 | free_mdstat(mdstat); | |
1365 | ||
1366 | return ent != NULL; | |
1367 | } | |
1368 | ||
1369 | /* open_subarray - opens a subarray in a container | |
1370 | * @dev: container device name | |
feab51f8 | 1371 | * @st: empty supertype |
33414a01 DW |
1372 | * @quiet: block reporting errors flag |
1373 | * | |
1374 | * On success returns an fd to a container and fills in *st | |
1375 | */ | |
feab51f8 | 1376 | int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet) |
33414a01 DW |
1377 | { |
1378 | struct mdinfo *mdi; | |
a951a4f7 | 1379 | struct mdinfo *info; |
33414a01 DW |
1380 | int fd, err = 1; |
1381 | ||
1382 | fd = open(dev, O_RDWR|O_EXCL); | |
1383 | if (fd < 0) { | |
1384 | if (!quiet) | |
e7b84f9d | 1385 | pr_err("Couldn't open %s, aborting\n", |
33414a01 | 1386 | dev); |
b990032d | 1387 | return -1; |
33414a01 DW |
1388 | } |
1389 | ||
1390 | st->devnum = fd2devnum(fd); | |
1391 | if (st->devnum == NoMdDev) { | |
1392 | if (!quiet) | |
e7b84f9d N |
1393 | pr_err("Failed to determine device number for %s\n", |
1394 | dev); | |
33414a01 DW |
1395 | goto close_fd; |
1396 | } | |
1397 | ||
1398 | mdi = sysfs_read(fd, st->devnum, GET_VERSION|GET_LEVEL); | |
1399 | if (!mdi) { | |
1400 | if (!quiet) | |
e7b84f9d | 1401 | pr_err("Failed to read sysfs for %s\n", |
33414a01 DW |
1402 | dev); |
1403 | goto close_fd; | |
1404 | } | |
1405 | ||
1406 | if (mdi->array.level != UnSet) { | |
1407 | if (!quiet) | |
e7b84f9d | 1408 | pr_err("%s is not a container\n", dev); |
33414a01 DW |
1409 | goto free_sysfs; |
1410 | } | |
1411 | ||
1412 | st->ss = version_to_superswitch(mdi->text_version); | |
1413 | if (!st->ss) { | |
1414 | if (!quiet) | |
e7b84f9d N |
1415 | pr_err("Operation not supported for %s metadata\n", |
1416 | mdi->text_version); | |
33414a01 DW |
1417 | goto free_sysfs; |
1418 | } | |
1419 | ||
1420 | st->devname = devnum2devname(st->devnum); | |
1421 | if (!st->devname) { | |
1422 | if (!quiet) | |
e7b84f9d | 1423 | pr_err("Failed to allocate device name\n"); |
33414a01 DW |
1424 | goto free_sysfs; |
1425 | } | |
1426 | ||
db20d413 | 1427 | if (!st->ss->load_container) { |
33414a01 | 1428 | if (!quiet) |
e7b84f9d | 1429 | pr_err("%s is not a container\n", dev); |
33414a01 DW |
1430 | goto free_name; |
1431 | } | |
1432 | ||
db20d413 | 1433 | if (st->ss->load_container(st, fd, NULL)) { |
33414a01 | 1434 | if (!quiet) |
e7b84f9d | 1435 | pr_err("Failed to load metadata for %s\n", |
db20d413 N |
1436 | dev); |
1437 | goto free_name; | |
33414a01 DW |
1438 | } |
1439 | ||
a951a4f7 N |
1440 | info = st->ss->container_content(st, subarray); |
1441 | if (!info) { | |
1442 | if (!quiet) | |
e7b84f9d | 1443 | pr_err("Failed to find subarray-%s in %s\n", |
a951a4f7 N |
1444 | subarray, dev); |
1445 | goto free_super; | |
1446 | } | |
1447 | free(info); | |
1448 | ||
33414a01 DW |
1449 | err = 0; |
1450 | ||
1451 | free_super: | |
1452 | if (err) | |
1453 | st->ss->free_super(st); | |
1454 | free_name: | |
1455 | if (err) | |
1456 | free(st->devname); | |
1457 | free_sysfs: | |
1458 | sysfs_free(mdi); | |
1459 | close_fd: | |
1460 | if (err) | |
1461 | close(fd); | |
1462 | ||
1463 | if (err) | |
1464 | return -1; | |
1465 | else | |
1466 | return fd; | |
1467 | } | |
1468 | ||
7801ac20 N |
1469 | int add_disk(int mdfd, struct supertype *st, |
1470 | struct mdinfo *sra, struct mdinfo *info) | |
1471 | { | |
1472 | /* Add a device to an array, in one of 2 ways. */ | |
1473 | int rv; | |
1474 | #ifndef MDASSEMBLE | |
1475 | if (st->ss->external) { | |
d23534e4 DW |
1476 | if (info->disk.state & (1<<MD_DISK_SYNC)) |
1477 | info->recovery_start = MaxSector; | |
1478 | else | |
1479 | info->recovery_start = 0; | |
2904b26f | 1480 | rv = sysfs_add_disk(sra, info, 0); |
7801ac20 N |
1481 | if (! rv) { |
1482 | struct mdinfo *sd2; | |
f35f2525 N |
1483 | for (sd2 = sra->devs; sd2; sd2=sd2->next) |
1484 | if (sd2 == info) | |
1485 | break; | |
1486 | if (sd2 == NULL) { | |
503975b9 | 1487 | sd2 = xmalloc(sizeof(*sd2)); |
f35f2525 N |
1488 | *sd2 = *info; |
1489 | sd2->next = sra->devs; | |
1490 | sra->devs = sd2; | |
1491 | } | |
7801ac20 N |
1492 | } |
1493 | } else | |
1494 | #endif | |
1495 | rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk); | |
1496 | return rv; | |
1497 | } | |
1498 | ||
de6ae750 N |
1499 | int remove_disk(int mdfd, struct supertype *st, |
1500 | struct mdinfo *sra, struct mdinfo *info) | |
1501 | { | |
1502 | int rv; | |
1503 | /* Remove the disk given by 'info' from the array */ | |
1504 | #ifndef MDASSEMBLE | |
1505 | if (st->ss->external) | |
1506 | rv = sysfs_set_str(sra, info, "slot", "none"); | |
1507 | else | |
1508 | #endif | |
1509 | rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major, | |
1510 | info->disk.minor)); | |
1511 | return rv; | |
1512 | } | |
1513 | ||
f35f2525 N |
1514 | int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info) |
1515 | { | |
1516 | /* Initialise kernel's knowledge of array. | |
1517 | * This varies between externally managed arrays | |
1518 | * and older kernels | |
1519 | */ | |
1520 | int vers = md_get_version(mdfd); | |
1521 | int rv; | |
1522 | ||
1523 | #ifndef MDASSEMBLE | |
1524 | if (st->ss->external) | |
1525 | rv = sysfs_set_array(info, vers); | |
1526 | else | |
1527 | #endif | |
1528 | if ((vers % 100) >= 1) { /* can use different versions */ | |
1529 | mdu_array_info_t inf; | |
1530 | memset(&inf, 0, sizeof(inf)); | |
1531 | inf.major_version = info->array.major_version; | |
1532 | inf.minor_version = info->array.minor_version; | |
1533 | rv = ioctl(mdfd, SET_ARRAY_INFO, &inf); | |
1534 | } else | |
1535 | rv = ioctl(mdfd, SET_ARRAY_INFO, NULL); | |
1536 | return rv; | |
1537 | } | |
1538 | ||
1e5c6983 DW |
1539 | unsigned long long min_recovery_start(struct mdinfo *array) |
1540 | { | |
1541 | /* find the minimum recovery_start in an array for metadata | |
1542 | * formats that only record per-array recovery progress instead | |
1543 | * of per-device | |
1544 | */ | |
1545 | unsigned long long recovery_start = MaxSector; | |
1546 | struct mdinfo *d; | |
1547 | ||
1548 | for (d = array->devs; d; d = d->next) | |
1549 | recovery_start = min(recovery_start, d->recovery_start); | |
1550 | ||
1551 | return recovery_start; | |
1552 | } | |
1553 | ||
24f6f99b | 1554 | int mdmon_pid(int devnum) |
a931db9e NB |
1555 | { |
1556 | char path[100]; | |
1557 | char pid[10]; | |
1558 | int fd; | |
1559 | int n; | |
10013317 PHC |
1560 | char *devname = devnum2devname(devnum); |
1561 | ||
753cf905 | 1562 | sprintf(path, "%s/%s.pid", MDMON_DIR, devname); |
10013317 PHC |
1563 | free(devname); |
1564 | ||
24f6f99b | 1565 | fd = open(path, O_RDONLY | O_NOATIME, 0); |
a931db9e NB |
1566 | |
1567 | if (fd < 0) | |
cf556303 | 1568 | return -1; |
a931db9e NB |
1569 | n = read(fd, pid, 9); |
1570 | close(fd); | |
1571 | if (n <= 0) | |
cf556303 | 1572 | return -1; |
24f6f99b | 1573 | return atoi(pid); |
a931db9e NB |
1574 | } |
1575 | ||
24f6f99b | 1576 | int mdmon_running(int devnum) |
a931db9e | 1577 | { |
24f6f99b N |
1578 | int pid = mdmon_pid(devnum); |
1579 | if (pid <= 0) | |
a931db9e | 1580 | return 0; |
24f6f99b | 1581 | if (kill(pid, 0) == 0) |
a931db9e NB |
1582 | return 1; |
1583 | return 0; | |
1584 | } | |
1585 | ||
8850ee3e N |
1586 | int start_mdmon(int devnum) |
1587 | { | |
1913c325 | 1588 | int i, skipped; |
44d2e365 | 1589 | int len; |
9fe32043 N |
1590 | pid_t pid; |
1591 | int status; | |
44d2e365 N |
1592 | char pathbuf[1024]; |
1593 | char *paths[4] = { | |
1594 | pathbuf, | |
1595 | "/sbin/mdmon", | |
1596 | "mdmon", | |
1597 | NULL | |
1598 | }; | |
8850ee3e | 1599 | |
40ebbb9c | 1600 | if (check_env("MDADM_NO_MDMON")) |
8850ee3e N |
1601 | return 0; |
1602 | ||
9cf014ec | 1603 | len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1); |
44d2e365 N |
1604 | if (len > 0) { |
1605 | char *sl; | |
1606 | pathbuf[len] = 0; | |
1607 | sl = strrchr(pathbuf, '/'); | |
1608 | if (sl) | |
1609 | sl++; | |
1610 | else | |
1611 | sl = pathbuf; | |
1612 | strcpy(sl, "mdmon"); | |
1613 | } else | |
1614 | pathbuf[0] = '\0'; | |
1615 | ||
8850ee3e N |
1616 | switch(fork()) { |
1617 | case 0: | |
1618 | /* FIXME yuk. CLOSE_EXEC?? */ | |
1913c325 N |
1619 | skipped = 0; |
1620 | for (i=3; skipped < 20; i++) | |
1621 | if (close(i) < 0) | |
1622 | skipped++; | |
1623 | else | |
1624 | skipped = 0; | |
1625 | ||
44d2e365 | 1626 | for (i=0; paths[i]; i++) |
a0963a86 JS |
1627 | if (paths[i][0]) { |
1628 | if (__offroot) { | |
1629 | execl(paths[i], "mdmon", "--offroot", | |
1630 | devnum2devname(devnum), | |
1631 | NULL); | |
1632 | } else { | |
1633 | execl(paths[i], "mdmon", | |
1634 | devnum2devname(devnum), | |
1635 | NULL); | |
1636 | } | |
1637 | } | |
8850ee3e | 1638 | exit(1); |
e7b84f9d | 1639 | case -1: pr_err("cannot run mdmon. " |
8850ee3e N |
1640 | "Array remains readonly\n"); |
1641 | return -1; | |
9fe32043 N |
1642 | default: /* parent - good */ |
1643 | pid = wait(&status); | |
1644 | if (pid < 0 || status != 0) | |
1645 | return -1; | |
8850ee3e N |
1646 | } |
1647 | return 0; | |
1648 | } | |
1649 | ||
40ebbb9c | 1650 | int check_env(char *name) |
5dcfcb71 | 1651 | { |
40ebbb9c | 1652 | char *val = getenv(name); |
5dcfcb71 DW |
1653 | |
1654 | if (val && atoi(val) == 1) | |
1655 | return 1; | |
1656 | ||
1657 | return 0; | |
1658 | } | |
1659 | ||
148acb7b DW |
1660 | __u32 random32(void) |
1661 | { | |
1662 | __u32 rv; | |
1663 | int rfd = open("/dev/urandom", O_RDONLY); | |
1664 | if (rfd < 0 || read(rfd, &rv, 4) != 4) | |
1665 | rv = random(); | |
1666 | if (rfd >= 0) | |
1667 | close(rfd); | |
1668 | return rv; | |
1669 | } | |
1670 | ||
0e600426 | 1671 | #ifndef MDASSEMBLE |
edd8d13c NB |
1672 | int flush_metadata_updates(struct supertype *st) |
1673 | { | |
1674 | int sfd; | |
1675 | if (!st->updates) { | |
1676 | st->update_tail = NULL; | |
1677 | return -1; | |
1678 | } | |
1679 | ||
1680 | sfd = connect_monitor(devnum2devname(st->container_dev)); | |
1681 | if (sfd < 0) | |
1682 | return -1; | |
1683 | ||
1684 | while (st->updates) { | |
1685 | struct metadata_update *mu = st->updates; | |
1686 | st->updates = mu->next; | |
1687 | ||
1688 | send_message(sfd, mu, 0); | |
1689 | wait_reply(sfd, 0); | |
1690 | free(mu->buf); | |
1691 | free(mu); | |
1692 | } | |
1693 | ack(sfd, 0); | |
1694 | wait_reply(sfd, 0); | |
1695 | close(sfd); | |
1696 | st->update_tail = NULL; | |
1697 | return 0; | |
1698 | } | |
1699 | ||
1700 | void append_metadata_update(struct supertype *st, void *buf, int len) | |
1701 | { | |
1702 | ||
503975b9 | 1703 | struct metadata_update *mu = xmalloc(sizeof(*mu)); |
edd8d13c NB |
1704 | |
1705 | mu->buf = buf; | |
1706 | mu->len = len; | |
1707 | mu->space = NULL; | |
cb23f1f4 | 1708 | mu->space_list = NULL; |
edd8d13c NB |
1709 | mu->next = NULL; |
1710 | *st->update_tail = mu; | |
1711 | st->update_tail = &mu->next; | |
1712 | } | |
0e600426 | 1713 | #endif /* MDASSEMBLE */ |
a931db9e | 1714 | |
fe6729fa NB |
1715 | #ifdef __TINYC__ |
1716 | /* tinyc doesn't optimize this check in ioctl.h out ... */ | |
1717 | unsigned int __invalid_size_argument_for_IOC = 0; | |
1718 | #endif | |
1719 | ||
6d11ec6f AK |
1720 | int experimental(void) |
1721 | { | |
1722 | if (check_env("MDADM_EXPERIMENTAL")) | |
1723 | return 1; | |
1724 | else { | |
e7b84f9d | 1725 | pr_err("To use this feature MDADM_EXPERIMENTAL" |
65c83a80 | 1726 | " environment variable has to be defined.\n"); |
6d11ec6f AK |
1727 | return 0; |
1728 | } | |
1729 | } | |
1730 | ||
326727d9 AC |
1731 | /* Pick all spares matching given criteria from a container |
1732 | * if min_size == 0 do not check size | |
1733 | * if domlist == NULL do not check domains | |
1734 | * if spare_group given add it to domains of each spare | |
1735 | * metadata allows to test domains using metadata of destination array */ | |
1736 | struct mdinfo *container_choose_spares(struct supertype *st, | |
1737 | unsigned long long min_size, | |
1738 | struct domainlist *domlist, | |
1739 | char *spare_group, | |
1740 | const char *metadata, int get_one) | |
1741 | { | |
1742 | struct mdinfo *d, **dp, *disks = NULL; | |
1743 | ||
1744 | /* get list of all disks in container */ | |
1745 | if (st->ss->getinfo_super_disks) | |
1746 | disks = st->ss->getinfo_super_disks(st); | |
1747 | ||
1748 | if (!disks) | |
1749 | return disks; | |
1750 | /* find spare devices on the list */ | |
1751 | dp = &disks->devs; | |
1752 | disks->array.spare_disks = 0; | |
1753 | while (*dp) { | |
1754 | int found = 0; | |
1755 | d = *dp; | |
1756 | if (d->disk.state == 0) { | |
1757 | /* check if size is acceptable */ | |
1758 | unsigned long long dev_size; | |
1759 | dev_t dev = makedev(d->disk.major,d->disk.minor); | |
1760 | ||
1761 | if (!min_size || | |
1762 | (dev_size_from_id(dev, &dev_size) && | |
1763 | dev_size >= min_size)) | |
1764 | found = 1; | |
1765 | /* check if domain matches */ | |
1766 | if (found && domlist) { | |
1767 | struct dev_policy *pol = devnum_policy(dev); | |
1768 | if (spare_group) | |
1769 | pol_add(&pol, pol_domain, | |
1770 | spare_group, NULL); | |
e5508b36 | 1771 | if (domain_test(domlist, pol, metadata) != 1) |
326727d9 AC |
1772 | found = 0; |
1773 | dev_policy_free(pol); | |
1774 | } | |
1775 | } | |
1776 | if (found) { | |
1777 | dp = &d->next; | |
1778 | disks->array.spare_disks++; | |
1779 | if (get_one) { | |
1780 | sysfs_free(*dp); | |
1781 | d->next = NULL; | |
1782 | } | |
1783 | } else { | |
1784 | *dp = d->next; | |
1785 | d->next = NULL; | |
1786 | sysfs_free(d); | |
1787 | } | |
1788 | } | |
1789 | return disks; | |
1790 | } | |
503975b9 N |
1791 | |
1792 | void *xmalloc(size_t len) | |
1793 | { | |
1794 | void *rv = malloc(len); | |
1795 | char *msg; | |
1796 | if (rv) | |
1797 | return rv; | |
1798 | msg = Name ": memory allocation failure - aborting\n"; | |
1799 | exit(4+!!write(2, msg, strlen(msg))); | |
1800 | } | |
1801 | ||
1802 | void *xrealloc(void *ptr, size_t len) | |
1803 | { | |
1804 | void *rv = realloc(ptr, len); | |
1805 | char *msg; | |
1806 | if (rv) | |
1807 | return rv; | |
1808 | msg = Name ": memory allocation failure - aborting\n"; | |
1809 | exit(4+!!write(2, msg, strlen(msg))); | |
1810 | } | |
1811 | ||
1812 | void *xcalloc(size_t num, size_t size) | |
1813 | { | |
1814 | void *rv = calloc(num, size); | |
1815 | char *msg; | |
1816 | if (rv) | |
1817 | return rv; | |
1818 | msg = Name ": memory allocation failure - aborting\n"; | |
1819 | exit(4+!!write(2, msg, strlen(msg))); | |
1820 | } | |
1821 | ||
1822 | char *xstrdup(const char *str) | |
1823 | { | |
1824 | char *rv = strdup(str); | |
1825 | char *msg; | |
1826 | if (rv) | |
1827 | return rv; | |
1828 | msg = Name ": memory allocation failure - aborting\n"; | |
1829 | exit(4+!!write(2, msg, strlen(msg))); | |
1830 | } |