]>
Commit | Line | Data |
---|---|---|
64c4757e | 1 | /* |
9a9dab36 | 2 | * mdadm - manage Linux "md" devices aka RAID arrays. |
64c4757e | 3 | * |
6f02172d | 4 | * Copyright (C) 2001-2013 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> |
a7dec3fd | 31 | #include <sys/resource.h> |
a9c15847 N |
32 | #include <sys/vfs.h> |
33 | #include <linux/magic.h> | |
98c6faba | 34 | #include <ctype.h> |
a322f70c | 35 | #include <dirent.h> |
a931db9e | 36 | #include <signal.h> |
0a816ef9 NB |
37 | |
38 | /* | |
39 | * following taken from linux/blkpg.h because they aren't | |
40 | * anywhere else and it isn't safe to #include linux/ * stuff. | |
41 | */ | |
42 | ||
43 | #define BLKPG _IO(0x12,105) | |
44 | ||
45 | /* The argument structure */ | |
46 | struct blkpg_ioctl_arg { | |
5d500228 N |
47 | int op; |
48 | int flags; | |
49 | int datalen; | |
50 | void *data; | |
0a816ef9 NB |
51 | }; |
52 | ||
53 | /* The subfunctions (for the op field) */ | |
54 | #define BLKPG_ADD_PARTITION 1 | |
55 | #define BLKPG_DEL_PARTITION 2 | |
56 | ||
57 | /* Sizes of name fields. Unused at present. */ | |
58 | #define BLKPG_DEVNAMELTH 64 | |
59 | #define BLKPG_VOLNAMELTH 64 | |
60 | ||
61 | /* The data structure for ADD_PARTITION and DEL_PARTITION */ | |
62 | struct blkpg_partition { | |
63 | long long start; /* starting offset in bytes */ | |
64 | long long length; /* length in bytes */ | |
65 | int pno; /* partition number */ | |
66 | char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2, | |
67 | to be used in kernel messages */ | |
68 | char volname[BLKPG_VOLNAMELTH]; /* volume label */ | |
69 | }; | |
64c4757e | 70 | |
0f22b998 | 71 | #include "part.h" |
056b331e N |
72 | |
73 | /* Force a compilation error if condition is true */ | |
74 | #define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition)) | |
75 | ||
76 | /* Force a compilation error if condition is true, but also produce a | |
77 | result (of value 0 and type size_t), so the expression can be used | |
78 | e.g. in a structure initializer (or where-ever else comma expressions | |
79 | aren't permitted). */ | |
80 | #define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); })) | |
81 | ||
64c4757e NB |
82 | /* |
83 | * Parse a 128 bit uuid in 4 integers | |
84 | * format is 32 hexx nibbles with options :.<space> separator | |
85 | * If not exactly 32 hex digits are found, return 0 | |
86 | * else return 1 | |
87 | */ | |
88 | int parse_uuid(char *str, int uuid[4]) | |
89 | { | |
aba69144 NB |
90 | int hit = 0; /* number of Hex digIT */ |
91 | int i; | |
92 | char c; | |
ca3b6696 N |
93 | for (i = 0; i < 4; i++) |
94 | uuid[i] = 0; | |
aba69144 | 95 | |
ca3b6696 | 96 | while ((c = *str++) != 0) { |
aba69144 | 97 | int n; |
ca3b6696 | 98 | if (c >= '0' && c <= '9') |
aba69144 | 99 | n = c-'0'; |
ca3b6696 | 100 | else if (c >= 'a' && c <= 'f') |
aba69144 | 101 | n = 10 + c - 'a'; |
ca3b6696 | 102 | else if (c >= 'A' && c <= 'F') |
aba69144 NB |
103 | n = 10 + c - 'A'; |
104 | else if (strchr(":. -", c)) | |
105 | continue; | |
106 | else return 0; | |
107 | ||
108 | if (hit<32) { | |
109 | uuid[hit/8] <<= 4; | |
110 | uuid[hit/8] += n; | |
111 | } | |
112 | hit++; | |
82b27616 | 113 | } |
aba69144 NB |
114 | if (hit == 32) |
115 | return 1; | |
116 | return 0; | |
64c4757e NB |
117 | } |
118 | ||
64c4757e NB |
119 | /* |
120 | * Get the md version number. | |
121 | * We use the RAID_VERSION ioctl if it is supported | |
122 | * If not, but we have a block device with major '9', we assume | |
123 | * 0.36.0 | |
124 | * | |
125 | * Return version number as 24 but number - assume version parts | |
126 | * always < 255 | |
127 | */ | |
128 | ||
129 | int md_get_version(int fd) | |
130 | { | |
5d500228 N |
131 | struct stat stb; |
132 | mdu_version_t vers; | |
133 | ||
134 | if (fstat(fd, &stb)<0) | |
135 | return -1; | |
136 | if ((S_IFMT&stb.st_mode) != S_IFBLK) | |
137 | return -1; | |
138 | ||
139 | if (ioctl(fd, RAID_VERSION, &vers) == 0) | |
140 | return (vers.major*10000) + (vers.minor*100) + vers.patchlevel; | |
141 | if (errno == EACCES) | |
142 | return -1; | |
143 | if (major(stb.st_rdev) == MD_MAJOR) | |
144 | return (3600); | |
64c4757e | 145 | return -1; |
64c4757e NB |
146 | } |
147 | ||
64c4757e NB |
148 | int get_linux_version() |
149 | { | |
150 | struct utsname name; | |
98c6faba | 151 | char *cp; |
f161d047 | 152 | int a = 0, b = 0,c = 0; |
64c4757e NB |
153 | if (uname(&name) <0) |
154 | return -1; | |
155 | ||
98c6faba NB |
156 | cp = name.release; |
157 | a = strtoul(cp, &cp, 10); | |
f161d047 N |
158 | if (*cp == '.') |
159 | b = strtoul(cp+1, &cp, 10); | |
160 | if (*cp == '.') | |
161 | c = strtoul(cp+1, &cp, 10); | |
98c6faba | 162 | |
682c7051 | 163 | return (a*1000000)+(b*1000)+c; |
64c4757e NB |
164 | } |
165 | ||
71204a50 | 166 | #ifndef MDASSEMBLE |
bc77ed53 DW |
167 | int mdadm_version(char *version) |
168 | { | |
169 | int a, b, c; | |
170 | char *cp; | |
171 | ||
172 | if (!version) | |
173 | version = Version; | |
174 | ||
175 | cp = strchr(version, '-'); | |
176 | if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v') | |
177 | return -1; | |
178 | cp += 3; | |
179 | a = strtoul(cp, &cp, 10); | |
180 | if (*cp != '.') | |
181 | return -1; | |
182 | b = strtoul(cp+1, &cp, 10); | |
183 | if (*cp == '.') | |
184 | c = strtoul(cp+1, &cp, 10); | |
185 | else | |
186 | c = 0; | |
187 | if (*cp != ' ' && *cp != '-') | |
188 | return -1; | |
189 | return (a*1000000)+(b*1000)+c; | |
190 | } | |
191 | ||
51d4261c | 192 | unsigned long long parse_size(char *size) |
84e11361 N |
193 | { |
194 | /* parse 'size' which should be a number optionally | |
195 | * followed by 'K', 'M', or 'G'. | |
196 | * Without a suffix, K is assumed. | |
197 | * Number returned is in sectors (half-K) | |
822e393a | 198 | * INVALID_SECTORS returned on error. |
84e11361 N |
199 | */ |
200 | char *c; | |
201 | long long s = strtoll(size, &c, 10); | |
202 | if (s > 0) { | |
203 | switch (*c) { | |
204 | case 'K': | |
205 | c++; | |
206 | default: | |
207 | s *= 2; | |
208 | break; | |
209 | case 'M': | |
210 | c++; | |
211 | s *= 1024 * 2; | |
212 | break; | |
213 | case 'G': | |
214 | c++; | |
215 | s *= 1024 * 1024 * 2; | |
216 | break; | |
72ca9bcf N |
217 | case 's': /* sectors */ |
218 | c++; | |
219 | break; | |
84e11361 | 220 | } |
51d4261c | 221 | } else |
822e393a | 222 | s = INVALID_SECTORS; |
84e11361 | 223 | if (*c) |
822e393a | 224 | s = INVALID_SECTORS; |
84e11361 N |
225 | return s; |
226 | } | |
227 | ||
4a06e2c2 N |
228 | int parse_layout_10(char *layout) |
229 | { | |
230 | int copies, rv; | |
231 | char *cp; | |
232 | /* Parse the layout string for raid10 */ | |
233 | /* 'f', 'o' or 'n' followed by a number <= raid_disks */ | |
234 | if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') || | |
235 | (copies = strtoul(layout+1, &cp, 10)) < 1 || | |
236 | copies > 200 || | |
237 | *cp) | |
238 | return -1; | |
239 | if (layout[0] == 'n') | |
240 | rv = 256 + copies; | |
241 | else if (layout[0] == 'o') | |
242 | rv = 0x10000 + (copies<<8) + 1; | |
243 | else | |
244 | rv = 1 + (copies<<8); | |
245 | return rv; | |
246 | } | |
247 | ||
248 | int parse_layout_faulty(char *layout) | |
249 | { | |
250 | /* Parse the layout string for 'faulty' */ | |
251 | int ln = strcspn(layout, "0123456789"); | |
503975b9 | 252 | char *m = xstrdup(layout); |
4a06e2c2 N |
253 | int mode; |
254 | m[ln] = 0; | |
255 | mode = map_name(faultylayout, m); | |
256 | if (mode == UnSet) | |
257 | return -1; | |
258 | ||
259 | return mode | (atoi(layout+ln)<< ModeShift); | |
260 | } | |
79868890 N |
261 | |
262 | long parse_num(char *num) | |
263 | { | |
264 | /* Either return a valid number, or -1 */ | |
265 | char *c; | |
266 | long rv = strtol(num, &c, 10); | |
267 | if (rv < 0 || *c || !num[0]) | |
268 | return -1; | |
269 | else | |
270 | return rv; | |
271 | } | |
eb3929a4 | 272 | #endif |
4a06e2c2 | 273 | |
0430ed48 NB |
274 | void remove_partitions(int fd) |
275 | { | |
276 | /* remove partitions from this block devices. | |
277 | * This is used for components added to an array | |
278 | */ | |
279 | #ifdef BLKPG_DEL_PARTITION | |
280 | struct blkpg_ioctl_arg a; | |
281 | struct blkpg_partition p; | |
282 | ||
283 | a.op = BLKPG_DEL_PARTITION; | |
284 | a.data = (void*)&p; | |
285 | a.datalen = sizeof(p); | |
286 | a.flags = 0; | |
287 | memset(a.data, 0, a.datalen); | |
ca3b6696 | 288 | for (p.pno = 0; p.pno < 16; p.pno++) |
0430ed48 NB |
289 | ioctl(fd, BLKPG, &a); |
290 | #endif | |
291 | } | |
292 | ||
691c6ee1 N |
293 | int test_partition(int fd) |
294 | { | |
295 | /* Check if fd is a whole-disk or a partition. | |
296 | * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION | |
297 | * will return ENXIO on an invalid partition number. | |
298 | */ | |
299 | struct blkpg_ioctl_arg a; | |
300 | struct blkpg_partition p; | |
301 | a.op = BLKPG_DEL_PARTITION; | |
302 | a.data = (void*)&p; | |
303 | a.datalen = sizeof(p); | |
304 | a.flags = 0; | |
305 | memset(a.data, 0, a.datalen); | |
306 | p.pno = 1<<30; | |
307 | if (ioctl(fd, BLKPG, &a) == 0) | |
308 | /* Very unlikely, but not a partition */ | |
309 | return 0; | |
357ac106 | 310 | if (errno == ENXIO || errno == ENOTTY) |
691c6ee1 N |
311 | /* not a partition */ |
312 | return 0; | |
313 | ||
314 | return 1; | |
315 | } | |
316 | ||
bfd76b93 CA |
317 | int test_partition_from_id(dev_t id) |
318 | { | |
319 | char buf[20]; | |
320 | int fd, rv; | |
321 | ||
322 | sprintf(buf, "%d:%d", major(id), minor(id)); | |
323 | fd = dev_open(buf, O_RDONLY); | |
324 | if (fd < 0) | |
325 | return -1; | |
326 | rv = test_partition(fd); | |
327 | close(fd); | |
328 | return rv; | |
329 | } | |
691c6ee1 | 330 | |
de5a472e | 331 | int enough(int level, int raid_disks, int layout, int clean, char *avail) |
64c4757e | 332 | { |
265e0f17 | 333 | int copies, first; |
de5a472e N |
334 | int i; |
335 | int avail_disks = 0; | |
336 | ||
337 | for (i = 0; i < raid_disks; i++) | |
338 | avail_disks += !!avail[i]; | |
339 | ||
64c4757e | 340 | switch (level) { |
265e0f17 NB |
341 | case 10: |
342 | /* This is the tricky one - we need to check | |
343 | * which actual disks are present. | |
344 | */ | |
702b557b | 345 | copies = (layout&255)* ((layout>>8) & 255); |
ca3b6696 | 346 | first = 0; |
265e0f17 NB |
347 | do { |
348 | /* there must be one of the 'copies' form 'first' */ | |
349 | int n = copies; | |
ca3b6696 | 350 | int cnt = 0; |
b3ec716d | 351 | int this = first; |
265e0f17 | 352 | while (n--) { |
b3ec716d | 353 | if (avail[this]) |
265e0f17 | 354 | cnt++; |
b3ec716d | 355 | this = (this+1) % raid_disks; |
265e0f17 NB |
356 | } |
357 | if (cnt == 0) | |
358 | return 0; | |
b3ec716d | 359 | first = (first+(layout&255)) % raid_disks; |
265e0f17 NB |
360 | } while (first != 0); |
361 | return 1; | |
e5329c37 | 362 | |
df0d4ea0 | 363 | case LEVEL_MULTIPATH: |
e0d19036 | 364 | return avail_disks>= 1; |
df0d4ea0 | 365 | case LEVEL_LINEAR: |
64c4757e NB |
366 | case 0: |
367 | return avail_disks == raid_disks; | |
368 | case 1: | |
369 | return avail_disks >= 1; | |
370 | case 4: | |
330d6900 N |
371 | if (avail_disks == raid_disks - 1 && |
372 | !avail[raid_disks - 1]) | |
373 | /* If just the parity device is missing, then we | |
374 | * have enough, even if not clean | |
375 | */ | |
376 | return 1; | |
377 | /* FALL THROUGH */ | |
64c4757e | 378 | case 5: |
583315d9 NB |
379 | if (clean) |
380 | return avail_disks >= raid_disks-1; | |
381 | else | |
382 | return avail_disks >= raid_disks; | |
98c6faba | 383 | case 6: |
583315d9 NB |
384 | if (clean) |
385 | return avail_disks >= raid_disks-2; | |
386 | else | |
387 | return avail_disks >= raid_disks; | |
64c4757e NB |
388 | default: |
389 | return 0; | |
390 | } | |
391 | } | |
392 | ||
8453e704 N |
393 | int enough_fd(int fd) |
394 | { | |
395 | struct mdu_array_info_s array; | |
396 | struct mdu_disk_info_s disk; | |
aabe020d | 397 | int i, rv; |
8453e704 N |
398 | char *avail; |
399 | ||
400 | if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 || | |
401 | array.raid_disks <= 0) | |
402 | return 0; | |
503975b9 | 403 | avail = xcalloc(array.raid_disks, 1); |
ca3b6696 | 404 | for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) { |
8453e704 N |
405 | disk.number = i; |
406 | if (ioctl(fd, GET_DISK_INFO, &disk) != 0) | |
407 | continue; | |
9e6d9291 N |
408 | if (disk.major == 0 && disk.minor == 0) |
409 | continue; | |
410 | array.nr_disks--; | |
411 | ||
8453e704 N |
412 | if (! (disk.state & (1<<MD_DISK_SYNC))) |
413 | continue; | |
414 | if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks) | |
415 | continue; | |
8453e704 N |
416 | avail[disk.raid_disk] = 1; |
417 | } | |
418 | /* This is used on an active array, so assume it is clean */ | |
aabe020d | 419 | rv = enough(array.level, array.raid_disks, array.layout, |
de5a472e | 420 | 1, avail); |
aabe020d JS |
421 | free(avail); |
422 | return rv; | |
8453e704 N |
423 | } |
424 | ||
22e263f6 AC |
425 | const int uuid_zero[4] = { 0, 0, 0, 0 }; |
426 | ||
f277ce36 | 427 | int same_uuid(int a[4], int b[4], int swapuuid) |
64c4757e | 428 | { |
f277ce36 NB |
429 | if (swapuuid) { |
430 | /* parse uuids are hostendian. | |
431 | * uuid's from some superblocks are big-ending | |
aba69144 | 432 | * if there is a difference, we need to swap.. |
f277ce36 NB |
433 | */ |
434 | unsigned char *ac = (unsigned char *)a; | |
435 | unsigned char *bc = (unsigned char *)b; | |
436 | int i; | |
ca3b6696 | 437 | for (i = 0; i < 16; i += 4) { |
f277ce36 NB |
438 | if (ac[i+0] != bc[i+3] || |
439 | ac[i+1] != bc[i+2] || | |
440 | ac[i+2] != bc[i+1] || | |
441 | ac[i+3] != bc[i+0]) | |
442 | return 0; | |
443 | } | |
444 | return 1; | |
445 | } else { | |
446 | if (a[0]==b[0] && | |
447 | a[1]==b[1] && | |
448 | a[2]==b[2] && | |
449 | a[3]==b[3]) | |
450 | return 1; | |
451 | return 0; | |
452 | } | |
64c4757e | 453 | } |
ca3b6696 | 454 | |
350f29f9 NB |
455 | void copy_uuid(void *a, int b[4], int swapuuid) |
456 | { | |
457 | if (swapuuid) { | |
458 | /* parse uuids are hostendian. | |
459 | * uuid's from some superblocks are big-ending | |
460 | * if there is a difference, we need to swap.. | |
461 | */ | |
462 | unsigned char *ac = (unsigned char *)a; | |
463 | unsigned char *bc = (unsigned char *)b; | |
464 | int i; | |
ca3b6696 | 465 | for (i = 0; i < 16; i += 4) { |
350f29f9 NB |
466 | ac[i+0] = bc[i+3]; |
467 | ac[i+1] = bc[i+2]; | |
468 | ac[i+2] = bc[i+1]; | |
469 | ac[i+3] = bc[i+0]; | |
470 | } | |
471 | } else | |
472 | memcpy(a, b, 16); | |
473 | } | |
64c4757e | 474 | |
aae5a112 | 475 | char *__fname_from_uuid(int id[4], int swap, char *buf, char sep) |
d7288ddc | 476 | { |
9968e376 | 477 | int i, j; |
d7288ddc N |
478 | char uuid[16]; |
479 | char *c = buf; | |
480 | strcpy(c, "UUID-"); | |
481 | c += strlen(c); | |
aae5a112 | 482 | copy_uuid(uuid, id, swap); |
9968e376 | 483 | for (i = 0; i < 4; i++) { |
9968e376 | 484 | if (i) |
ff54de6e | 485 | *c++ = sep; |
9968e376 DW |
486 | for (j = 3; j >= 0; j--) { |
487 | sprintf(c,"%02x", (unsigned char) uuid[j+4*i]); | |
488 | c+= 2; | |
489 | } | |
d7288ddc N |
490 | } |
491 | return buf; | |
aae5a112 DW |
492 | |
493 | } | |
494 | ||
495 | char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep) | |
496 | { | |
86983cce N |
497 | // dirty hack to work around an issue with super1 superblocks... |
498 | // super1 superblocks need swapuuid set in order for assembly to | |
499 | // work, but can't have it set if we want this printout to match | |
500 | // all the other uuid printouts in super1.c, so we force swapuuid | |
501 | // to 1 to make our printout match the rest of super1 | |
502 | return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep); | |
d7288ddc N |
503 | } |
504 | ||
435d4ebb | 505 | #ifndef MDASSEMBLE |
682c7051 NB |
506 | int check_ext2(int fd, char *name) |
507 | { | |
508 | /* | |
509 | * Check for an ext2fs file system. | |
510 | * Superblock is always 1K at 1K offset | |
511 | * | |
512 | * s_magic is le16 at 56 == 0xEF53 | |
513 | * report mtime - le32 at 44 | |
514 | * blocks - le32 at 4 | |
515 | * logblksize - le32 at 24 | |
516 | */ | |
517 | unsigned char sb[1024]; | |
518 | time_t mtime; | |
2538ba2a N |
519 | unsigned long long size; |
520 | int bsize; | |
682c7051 NB |
521 | if (lseek(fd, 1024,0)!= 1024) |
522 | return 0; | |
523 | if (read(fd, sb, 1024)!= 1024) | |
524 | return 0; | |
525 | if (sb[56] != 0x53 || sb[57] != 0xef) | |
526 | return 0; | |
527 | ||
528 | mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8; | |
529 | bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8; | |
530 | size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8; | |
2538ba2a | 531 | size <<= bsize; |
e7b84f9d | 532 | pr_err("%s appears to contain an ext2fs file system\n", |
682c7051 | 533 | name); |
2538ba2a | 534 | cont_err("size=%lluK mtime=%s", size, ctime(&mtime)); |
682c7051 NB |
535 | return 1; |
536 | } | |
537 | ||
538 | int check_reiser(int fd, char *name) | |
539 | { | |
540 | /* | |
541 | * superblock is at 64K | |
542 | * size is 1024; | |
543 | * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52 | |
544 | * | |
545 | */ | |
546 | unsigned char sb[1024]; | |
2538ba2a | 547 | unsigned long long size; |
682c7051 NB |
548 | if (lseek(fd, 64*1024, 0) != 64*1024) |
549 | return 0; | |
550 | if (read(fd, sb, 1024) != 1024) | |
551 | return 0; | |
ca3b6696 N |
552 | if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 && |
553 | strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0) | |
682c7051 | 554 | return 0; |
e7b84f9d | 555 | pr_err("%s appears to contain a reiserfs file system\n",name); |
682c7051 | 556 | size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8; |
2538ba2a | 557 | cont_err("size = %lluK\n", size*4); |
aba69144 | 558 | |
682c7051 NB |
559 | return 1; |
560 | } | |
561 | ||
562 | int check_raid(int fd, char *name) | |
563 | { | |
4b1ac34b | 564 | struct mdinfo info; |
682c7051 | 565 | time_t crtime; |
d078d77c | 566 | char *level; |
82d9eba6 | 567 | struct supertype *st = guess_super(fd); |
f9ce90ba | 568 | |
cb8f6859 N |
569 | if (!st) |
570 | return 0; | |
3da92f27 | 571 | st->ss->load_super(st, fd, name); |
82d9eba6 | 572 | /* Looks like a raid array .. */ |
e7b84f9d | 573 | pr_err("%s appears to be part of a raid array:\n", |
82d9eba6 | 574 | name); |
a5d85af7 | 575 | st->ss->getinfo_super(st, &info, NULL); |
3da92f27 | 576 | st->ss->free_super(st); |
82d9eba6 | 577 | crtime = info.array.ctime; |
d078d77c NB |
578 | level = map_num(pers, info.array.level); |
579 | if (!level) level = "-unknown-"; | |
d33f1518 N |
580 | cont_err("level=%s devices=%d ctime=%s", |
581 | level, info.array.raid_disks, ctime(&crtime)); | |
82d9eba6 | 582 | return 1; |
682c7051 NB |
583 | } |
584 | ||
682c7051 NB |
585 | int ask(char *mesg) |
586 | { | |
587 | char *add = ""; | |
588 | int i; | |
ca3b6696 | 589 | for (i = 0; i < 5; i++) { |
682c7051 NB |
590 | char buf[100]; |
591 | fprintf(stderr, "%s%s", mesg, add); | |
592 | fflush(stderr); | |
593 | if (fgets(buf, 100, stdin)==NULL) | |
594 | return 0; | |
595 | if (buf[0]=='y' || buf[0]=='Y') | |
596 | return 1; | |
597 | if (buf[0]=='n' || buf[0]=='N') | |
598 | return 0; | |
599 | add = "(y/n) "; | |
600 | } | |
e7b84f9d | 601 | pr_err("assuming 'no'\n"); |
682c7051 NB |
602 | return 0; |
603 | } | |
435d4ebb | 604 | #endif /* MDASSEMBLE */ |
682c7051 | 605 | |
8d80900b | 606 | int is_standard(char *dev, int *nump) |
e5329c37 NB |
607 | { |
608 | /* tests if dev is a "standard" md dev name. | |
609 | * i.e if the last component is "/dNN" or "/mdNN", | |
aba69144 | 610 | * where NN is a string of digits |
598f0d58 NB |
611 | * Returns 1 if a partitionable standard, |
612 | * -1 if non-partitonable, | |
613 | * 0 if not a standard name. | |
e5329c37 | 614 | */ |
8d80900b | 615 | char *d = strrchr(dev, '/'); |
ca3b6696 | 616 | int type = 0; |
8d80900b NB |
617 | int num; |
618 | if (!d) | |
e5329c37 | 619 | return 0; |
ca3b6696 N |
620 | if (strncmp(d, "/d",2) == 0) |
621 | d += 2, type = 1; /* /dev/md/dN{pM} */ | |
622 | else if (strncmp(d, "/md_d", 5) == 0) | |
623 | d += 5, type = 1; /* /dev/md_dN{pM} */ | |
624 | else if (strncmp(d, "/md", 3) == 0) | |
625 | d += 3, type = -1; /* /dev/mdN */ | |
626 | else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0) | |
627 | d += 1, type = -1; /* /dev/md/N */ | |
e5329c37 NB |
628 | else |
629 | return 0; | |
8d80900b | 630 | if (!*d) |
e5329c37 | 631 | return 0; |
8d80900b NB |
632 | num = atoi(d); |
633 | while (isdigit(*d)) | |
634 | d++; | |
635 | if (*d) | |
e5329c37 | 636 | return 0; |
8d80900b NB |
637 | if (nump) *nump = num; |
638 | ||
639 | return type; | |
e5329c37 NB |
640 | } |
641 | ||
4b1ac34b | 642 | unsigned long calc_csum(void *super, int bytes) |
82b27616 | 643 | { |
56eb10c0 | 644 | unsigned long long newcsum = 0; |
82b27616 | 645 | int i; |
4b1ac34b NB |
646 | unsigned int csum; |
647 | unsigned int *superc = (unsigned int*) super; | |
82b27616 | 648 | |
ca3b6696 N |
649 | for(i = 0; i < bytes/4; i++) |
650 | newcsum += superc[i]; | |
82b27616 | 651 | csum = (newcsum& 0xffffffff) + (newcsum>>32); |
570c0542 | 652 | #ifdef __alpha__ |
aba69144 | 653 | /* The in-kernel checksum calculation is always 16bit on |
570c0542 | 654 | * the alpha, though it is 32 bit on i386... |
ca3b6696 | 655 | * I wonder what it is elsewhere... (it uses an API in |
570c0542 NB |
656 | * a way that it shouldn't). |
657 | */ | |
658 | csum = (csum & 0xffff) + (csum >> 16); | |
659 | csum = (csum & 0xffff) + (csum >> 16); | |
660 | #endif | |
82b27616 NB |
661 | return csum; |
662 | } | |
cd29a5c8 | 663 | |
435d4ebb | 664 | #ifndef MDASSEMBLE |
56eb10c0 | 665 | char *human_size(long long bytes) |
cd29a5c8 NB |
666 | { |
667 | static char buf[30]; | |
d5d3721e NB |
668 | |
669 | /* We convert bytes to either centi-M{ega,ibi}bytes or | |
670 | * centi-G{igi,ibi}bytes, with appropriate rounding, | |
671 | * and then print 1/100th of those as a decimal. | |
672 | * We allow upto 2048Megabytes before converting to | |
673 | * gigabytes, as that shows more precision and isn't | |
674 | * too large a number. | |
01619b48 | 675 | * Terabytes are not yet handled. |
d5d3721e | 676 | */ |
cd29a5c8 | 677 | |
56eb10c0 | 678 | if (bytes < 5000*1024) |
ca3b6696 | 679 | buf[0] = 0; |
d5d3721e | 680 | else if (bytes < 2*1024LL*1024LL*1024LL) { |
93d3bd3b | 681 | long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2; |
d5d3721e | 682 | long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2; |
8f23b0b3 | 683 | snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)", |
d5d3721e NB |
684 | cMiB/100 , cMiB % 100, |
685 | cMB/100, cMB % 100); | |
686 | } else { | |
93d3bd3b | 687 | long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2; |
d5d3721e | 688 | long cGB = (bytes / (1000000000LL/200LL ) +1) /2; |
8f23b0b3 | 689 | snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)", |
d5d3721e NB |
690 | cGiB/100 , cGiB % 100, |
691 | cGB/100, cGB % 100); | |
692 | } | |
cd29a5c8 NB |
693 | return buf; |
694 | } | |
e0d19036 | 695 | |
f0ec6710 | 696 | char *human_size_brief(long long bytes, int prefix) |
e0d19036 NB |
697 | { |
698 | static char buf[30]; | |
e0d19036 | 699 | |
570abc6f MN |
700 | /* We convert bytes to either centi-M{ega,ibi}bytes or |
701 | * centi-G{igi,ibi}bytes, with appropriate rounding, | |
702 | * and then print 1/100th of those as a decimal. | |
703 | * We allow upto 2048Megabytes before converting to | |
704 | * gigabytes, as that shows more precision and isn't | |
705 | * too large a number. | |
706 | * Terabytes are not yet handled. | |
f0ec6710 MN |
707 | * |
708 | * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc. | |
709 | * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc. | |
570abc6f MN |
710 | */ |
711 | ||
e0d19036 | 712 | if (bytes < 5000*1024) |
570abc6f | 713 | buf[0] = 0; |
f0ec6710 MN |
714 | else if (prefix == IEC) { |
715 | if (bytes < 2*1024LL*1024LL*1024LL) { | |
93d3bd3b | 716 | long cMiB = (bytes * 200LL / (1LL<<20) +1) /2; |
f0ec6710 MN |
717 | snprintf(buf, sizeof(buf), "%ld.%02ldMiB", |
718 | cMiB/100 , cMiB % 100); | |
719 | } else { | |
93d3bd3b | 720 | long cGiB = (bytes * 200LL / (1LL<<30) +1) /2; |
f0ec6710 MN |
721 | snprintf(buf, sizeof(buf), "%ld.%02ldGiB", |
722 | cGiB/100 , cGiB % 100); | |
723 | } | |
724 | } | |
725 | else if (prefix == JEDEC) { | |
726 | if (bytes < 2*1024LL*1024LL*1024LL) { | |
727 | long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2; | |
728 | snprintf(buf, sizeof(buf), "%ld.%02ldMB", | |
729 | cMB/100, cMB % 100); | |
730 | } else { | |
731 | long cGB = (bytes / (1000000000LL/200LL ) +1) /2; | |
732 | snprintf(buf, sizeof(buf), "%ld.%02ldGB", | |
733 | cGB/100 , cGB % 100); | |
734 | } | |
570abc6f | 735 | } |
f0ec6710 MN |
736 | else |
737 | buf[0] = 0; | |
570abc6f | 738 | |
e0d19036 NB |
739 | return buf; |
740 | } | |
e4965ef8 N |
741 | |
742 | void print_r10_layout(int layout) | |
743 | { | |
744 | int near = layout & 255; | |
745 | int far = (layout >> 8) & 255; | |
746 | int offset = (layout&0x10000); | |
747 | char *sep = ""; | |
748 | ||
749 | if (near != 1) { | |
750 | printf("%s near=%d", sep, near); | |
751 | sep = ","; | |
752 | } | |
753 | if (far != 1) | |
754 | printf("%s %s=%d", sep, offset?"offset":"far", far); | |
755 | if (near*far == 1) | |
756 | printf("NO REDUNDANCY"); | |
757 | } | |
435d4ebb | 758 | #endif |
e0d19036 | 759 | |
5f8097be NB |
760 | unsigned long long calc_array_size(int level, int raid_disks, int layout, |
761 | int chunksize, unsigned long long devsize) | |
577e8448 | 762 | { |
e03640bd CA |
763 | if (level == 1) |
764 | return devsize; | |
577e8448 AK |
765 | devsize &= ~(unsigned long long)((chunksize>>9)-1); |
766 | return get_data_disks(level, layout, raid_disks) * devsize; | |
767 | } | |
768 | ||
769 | int get_data_disks(int level, int layout, int raid_disks) | |
5f8097be NB |
770 | { |
771 | int data_disks = 0; | |
772 | switch (level) { | |
ca3b6696 N |
773 | case 0: data_disks = raid_disks; |
774 | break; | |
775 | case 1: data_disks = 1; | |
776 | break; | |
5f8097be | 777 | case 4: |
ca3b6696 N |
778 | case 5: data_disks = raid_disks - 1; |
779 | break; | |
780 | case 6: data_disks = raid_disks - 2; | |
781 | break; | |
5f8097be NB |
782 | case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255); |
783 | break; | |
784 | } | |
577e8448 AK |
785 | |
786 | return data_disks; | |
5f8097be NB |
787 | } |
788 | ||
4dd2df09 N |
789 | int devnm2devid(char *devnm) |
790 | { | |
791 | /* First look in /sys/block/$DEVNM/dev for %d:%d | |
792 | * If that fails, try parsing out a number | |
793 | */ | |
794 | char path[100]; | |
795 | char *ep; | |
796 | int fd; | |
797 | int mjr,mnr; | |
798 | ||
799 | sprintf(path, "/sys/block/%s/dev", devnm); | |
800 | fd = open(path, O_RDONLY); | |
801 | if (fd >= 0) { | |
802 | char buf[20]; | |
803 | int n = read(fd, buf, sizeof(buf)); | |
804 | close(fd); | |
805 | if (n > 0) | |
806 | buf[n] = 0; | |
807 | if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2) | |
808 | return makedev(mjr, mnr); | |
809 | } | |
810 | if (strncmp(devnm, "md_d", 4) == 0 && | |
811 | isdigit(devnm[4]) && | |
812 | (mnr = strtoul(devnm+4, &ep, 10)) >= 0 && | |
813 | ep > devnm && *ep == 0) | |
814 | return makedev(get_mdp_major(), mnr << MdpMinorShift); | |
815 | ||
816 | if (strncmp(devnm, "md", 2) == 0 && | |
817 | isdigit(devnm[2]) && | |
818 | (mnr = strtoul(devnm+2, &ep, 10)) >= 0 && | |
819 | ep > devnm && *ep == 0) | |
820 | return makedev(MD_MAJOR, mnr); | |
821 | ||
822 | return 0; | |
823 | } | |
824 | ||
5a23a06e | 825 | #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) |
4dd2df09 | 826 | char *get_md_name(char *devnm) |
e0d19036 NB |
827 | { |
828 | /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */ | |
98c6faba | 829 | /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */ |
4dd2df09 | 830 | |
e0d19036 NB |
831 | static char devname[50]; |
832 | struct stat stb; | |
4dd2df09 | 833 | dev_t rdev = devnm2devid(devnm); |
dd0781e5 | 834 | char *dn; |
98c6faba | 835 | |
4dd2df09 N |
836 | if (rdev == 0) |
837 | return 0; | |
838 | if (strncmp(devnm, "md_", 3) == 0) { | |
839 | snprintf(devname, sizeof(devname), "/dev/md/%s", | |
840 | devnm + 3); | |
98c6faba NB |
841 | if (stat(devname, &stb) == 0 |
842 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
843 | && (stb.st_rdev == rdev)) | |
844 | return devname; | |
845 | } | |
4dd2df09 N |
846 | snprintf(devname, sizeof(devname), "/dev/%s", devnm); |
847 | if (stat(devname, &stb) == 0 | |
848 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
849 | && (stb.st_rdev == rdev)) | |
850 | return devname; | |
851 | ||
852 | snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2); | |
853 | if (stat(devname, &stb) == 0 | |
854 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
855 | && (stb.st_rdev == rdev)) | |
856 | return devname; | |
857 | ||
16c6fa80 | 858 | dn = map_dev(major(rdev), minor(rdev), 0); |
dd0781e5 NB |
859 | if (dn) |
860 | return dn; | |
4dd2df09 | 861 | snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm); |
e0d19036 | 862 | if (mknod(devname, S_IFBLK | 0600, rdev) == -1) |
dd0781e5 NB |
863 | if (errno != EEXIST) |
864 | return NULL; | |
e0d19036 NB |
865 | |
866 | if (stat(devname, &stb) == 0 | |
867 | && (S_IFMT&stb.st_mode) == S_IFBLK | |
868 | && (stb.st_rdev == rdev)) | |
869 | return devname; | |
870 | unlink(devname); | |
871 | return NULL; | |
872 | } | |
873 | ||
874 | void put_md_name(char *name) | |
875 | { | |
ca3b6696 | 876 | if (strncmp(name, "/dev/.tmp.md", 12) == 0) |
e0d19036 NB |
877 | unlink(name); |
878 | } | |
435d4ebb | 879 | #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */ |
f9ce90ba | 880 | |
cc742d38 N |
881 | int get_maj_min(char *dev, int *major, int *minor) |
882 | { | |
883 | char *e; | |
884 | *major = strtoul(dev, &e, 0); | |
885 | return (e > dev && *e == ':' && e[1] && | |
886 | (*minor = strtoul(e+1, &e, 0)) >= 0 && | |
887 | *e == 0); | |
888 | } | |
889 | ||
8b0dabea NB |
890 | int dev_open(char *dev, int flags) |
891 | { | |
892 | /* like 'open', but if 'dev' matches %d:%d, create a temp | |
893 | * block device and open that | |
894 | */ | |
8b0dabea NB |
895 | int fd = -1; |
896 | char devname[32]; | |
e81cdd9f | 897 | int major; |
8b0dabea | 898 | int minor; |
e81cdd9f NB |
899 | |
900 | if (!dev) return -1; | |
6df6a774 | 901 | flags |= O_DIRECT; |
e81cdd9f | 902 | |
cc742d38 | 903 | if (get_maj_min(dev, &major, &minor)) { |
9dc7d357 N |
904 | snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d", |
905 | (int)getpid(), major, minor); | |
906 | if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) { | |
907 | fd = open(devname, flags); | |
908 | unlink(devname); | |
6df6a774 N |
909 | } |
910 | if (fd < 0) { | |
9dc7d357 | 911 | /* Try /tmp as /dev appear to be read-only */ |
6df6a774 N |
912 | snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d", |
913 | (int)getpid(), major, minor); | |
ca3b6696 | 914 | if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) { |
6df6a774 N |
915 | fd = open(devname, flags); |
916 | unlink(devname); | |
917 | } | |
8b0dabea NB |
918 | } |
919 | } else | |
6df6a774 | 920 | fd = open(dev, flags); |
8b0dabea NB |
921 | return fd; |
922 | } | |
f9ce90ba | 923 | |
4dd2df09 | 924 | int open_dev_flags(char *devnm, int flags) |
e8a70c89 | 925 | { |
4dd2df09 | 926 | int devid; |
e8a70c89 N |
927 | char buf[20]; |
928 | ||
4dd2df09 N |
929 | devid = devnm2devid(devnm); |
930 | sprintf(buf, "%d:%d", major(devid), minor(devid)); | |
d998b738 N |
931 | return dev_open(buf, flags); |
932 | } | |
933 | ||
4dd2df09 | 934 | int open_dev(char *devnm) |
d998b738 | 935 | { |
4dd2df09 | 936 | return open_dev_flags(devnm, O_RDONLY); |
e8a70c89 N |
937 | } |
938 | ||
4dd2df09 | 939 | int open_dev_excl(char *devnm) |
a931db9e NB |
940 | { |
941 | char buf[20]; | |
942 | int i; | |
7187750e | 943 | int flags = O_RDWR; |
4dd2df09 | 944 | int devid = devnm2devid(devnm); |
399e0b97 | 945 | long delay = 1000; |
a931db9e | 946 | |
4dd2df09 | 947 | sprintf(buf, "%d:%d", major(devid), minor(devid)); |
ca3b6696 | 948 | for (i = 0 ; i < 25 ; i++) { |
7187750e | 949 | int fd = dev_open(buf, flags|O_EXCL); |
a931db9e NB |
950 | if (fd >= 0) |
951 | return fd; | |
7187750e N |
952 | if (errno == EACCES && flags == O_RDWR) { |
953 | flags = O_RDONLY; | |
954 | continue; | |
955 | } | |
a931db9e NB |
956 | if (errno != EBUSY) |
957 | return fd; | |
399e0b97 N |
958 | usleep(delay); |
959 | if (delay < 200000) | |
960 | delay *= 2; | |
a931db9e NB |
961 | } |
962 | return -1; | |
963 | } | |
964 | ||
9008ed1c N |
965 | int same_dev(char *one, char *two) |
966 | { | |
967 | struct stat st1, st2; | |
968 | if (stat(one, &st1) != 0) | |
969 | return 0; | |
970 | if (stat(two, &st2) != 0) | |
971 | return 0; | |
972 | if ((st1.st_mode & S_IFMT) != S_IFBLK) | |
973 | return 0; | |
974 | if ((st2.st_mode & S_IFMT) != S_IFBLK) | |
975 | return 0; | |
976 | return st1.st_rdev == st2.st_rdev; | |
977 | } | |
978 | ||
a7c6e3fb | 979 | void wait_for(char *dev, int fd) |
a714580e N |
980 | { |
981 | int i; | |
a7c6e3fb | 982 | struct stat stb_want; |
399e0b97 | 983 | long delay = 1000; |
a7c6e3fb N |
984 | |
985 | if (fstat(fd, &stb_want) != 0 || | |
986 | (stb_want.st_mode & S_IFMT) != S_IFBLK) | |
987 | return; | |
a714580e | 988 | |
ca3b6696 | 989 | for (i = 0 ; i < 25 ; i++) { |
a714580e | 990 | struct stat stb; |
a7c6e3fb N |
991 | if (stat(dev, &stb) == 0 && |
992 | (stb.st_mode & S_IFMT) == S_IFBLK && | |
993 | (stb.st_rdev == stb_want.st_rdev)) | |
a714580e | 994 | return; |
399e0b97 N |
995 | usleep(delay); |
996 | if (delay < 200000) | |
997 | delay *= 2; | |
a714580e | 998 | } |
436305c6 | 999 | if (i == 25) |
1ade5cc1 | 1000 | dprintf("timeout waiting for %s\n", dev); |
a714580e N |
1001 | } |
1002 | ||
0f22b998 N |
1003 | struct superswitch *superlist[] = |
1004 | { | |
1005 | &super0, &super1, | |
1006 | &super_ddf, &super_imsm, | |
0592faeb | 1007 | &mbr, &gpt, |
0f22b998 | 1008 | NULL }; |
f9ce90ba | 1009 | |
ea24acd0 | 1010 | #if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) |
f7dd881f | 1011 | |
4725bc31 | 1012 | struct supertype *super_by_fd(int fd, char **subarrayp) |
f9ce90ba | 1013 | { |
1686dc25 NB |
1014 | mdu_array_info_t array; |
1015 | int vers; | |
1016 | int minor; | |
1017 | struct supertype *st = NULL; | |
7e0f6979 | 1018 | struct mdinfo *sra; |
142cb9e1 | 1019 | char *verstr; |
1686dc25 NB |
1020 | char version[20]; |
1021 | int i; | |
f7e7067b | 1022 | char *subarray = NULL; |
4dd2df09 | 1023 | char container[32] = ""; |
1686dc25 | 1024 | |
4dd2df09 | 1025 | sra = sysfs_read(fd, NULL, GET_VERSION); |
1686dc25 NB |
1026 | |
1027 | if (sra) { | |
7e0f6979 NB |
1028 | vers = sra->array.major_version; |
1029 | minor = sra->array.minor_version; | |
142cb9e1 | 1030 | verstr = sra->text_version; |
1686dc25 NB |
1031 | } else { |
1032 | if (ioctl(fd, GET_ARRAY_INFO, &array)) | |
1033 | array.major_version = array.minor_version = 0; | |
1034 | vers = array.major_version; | |
1035 | minor = array.minor_version; | |
142cb9e1 | 1036 | verstr = ""; |
6fbba4c9 | 1037 | } |
82d9eba6 | 1038 | |
1686dc25 NB |
1039 | if (vers != -1) { |
1040 | sprintf(version, "%d.%d", vers, minor); | |
1041 | verstr = version; | |
6fbba4c9 | 1042 | } |
3c558363 | 1043 | if (minor == -2 && is_subarray(verstr)) { |
f7e7067b | 1044 | char *dev = verstr+1; |
5f7e44b2 | 1045 | |
f7e7067b | 1046 | subarray = strchr(dev, '/'); |
92d49ecf | 1047 | if (subarray) { |
f7e7067b | 1048 | *subarray++ = '\0'; |
503975b9 | 1049 | subarray = xstrdup(subarray); |
92d49ecf | 1050 | } |
4dd2df09 | 1051 | strcpy(container, dev); |
f7e7067b NB |
1052 | if (sra) |
1053 | sysfs_free(sra); | |
5f7e44b2 | 1054 | sra = sysfs_read(-1, container, GET_VERSION); |
603f24a0 N |
1055 | if (sra && sra->text_version[0]) |
1056 | verstr = sra->text_version; | |
1057 | else | |
1058 | verstr = "-no-metadata-"; | |
f7e7067b NB |
1059 | } |
1060 | ||
1061 | for (i = 0; st == NULL && superlist[i] ; i++) | |
1062 | st = superlist[i]->match_metadata_desc(verstr); | |
1686dc25 NB |
1063 | |
1064 | if (sra) | |
1065 | sysfs_free(sra); | |
f7e7067b | 1066 | if (st) { |
3b0896f8 | 1067 | st->sb = NULL; |
1f49fb3a N |
1068 | if (subarrayp) |
1069 | *subarrayp = subarray; | |
4dd2df09 N |
1070 | strcpy(st->container_devnm, container); |
1071 | strcpy(st->devnm, fd2devnm(fd)); | |
4725bc31 N |
1072 | } else |
1073 | free(subarray); | |
5f7e44b2 | 1074 | |
82d9eba6 | 1075 | return st; |
f9ce90ba | 1076 | } |
ea24acd0 NB |
1077 | #endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */ |
1078 | ||
0f0749ad | 1079 | int dev_size_from_id(dev_t id, unsigned long long *size) |
52d5d101 AC |
1080 | { |
1081 | char buf[20]; | |
1082 | int fd; | |
1083 | ||
1084 | sprintf(buf, "%d:%d", major(id), minor(id)); | |
1085 | fd = dev_open(buf, O_RDONLY); | |
1086 | if (fd < 0) | |
1087 | return 0; | |
1088 | if (get_dev_size(fd, NULL, size)) { | |
1089 | close(fd); | |
1090 | return 1; | |
1091 | } | |
1092 | close(fd); | |
1093 | return 0; | |
1094 | } | |
f9ce90ba | 1095 | |
159c3a1a | 1096 | struct supertype *dup_super(struct supertype *orig) |
3da92f27 | 1097 | { |
159c3a1a | 1098 | struct supertype *st; |
1686dc25 | 1099 | |
d2ca6449 NB |
1100 | if (!orig) |
1101 | return orig; | |
503975b9 | 1102 | st = xcalloc(1, sizeof(*st)); |
159c3a1a NB |
1103 | st->ss = orig->ss; |
1104 | st->max_devs = orig->max_devs; | |
1105 | st->minor_version = orig->minor_version; | |
cb8f6859 | 1106 | st->ignore_hw_compat = orig->ignore_hw_compat; |
23bf42cc | 1107 | st->data_offset = orig->data_offset; |
159c3a1a NB |
1108 | st->sb = NULL; |
1109 | st->info = NULL; | |
1110 | return st; | |
3da92f27 NB |
1111 | } |
1112 | ||
54887ad8 | 1113 | struct supertype *guess_super_type(int fd, enum guess_types guess_type) |
f9ce90ba NB |
1114 | { |
1115 | /* try each load_super to find the best match, | |
1116 | * and return the best superswitch | |
1117 | */ | |
82d9eba6 NB |
1118 | struct superswitch *ss; |
1119 | struct supertype *st; | |
f21e18ca | 1120 | time_t besttime = 0; |
570c0542 | 1121 | int bestsuper = -1; |
f9ce90ba NB |
1122 | int i; |
1123 | ||
503975b9 | 1124 | st = xcalloc(1, sizeof(*st)); |
4dd2df09 | 1125 | st->container_devnm[0] = 0; |
d1d599ea | 1126 | |
ca3b6696 | 1127 | for (i = 0 ; superlist[i]; i++) { |
f9ce90ba NB |
1128 | int rv; |
1129 | ss = superlist[i]; | |
54887ad8 N |
1130 | if (guess_type == guess_array && ss->add_to_super == NULL) |
1131 | continue; | |
1132 | if (guess_type == guess_partitions && ss->add_to_super != NULL) | |
1133 | continue; | |
ef609477 | 1134 | memset(st, 0, sizeof(*st)); |
df3346e6 | 1135 | st->ignore_hw_compat = 1; |
3da92f27 | 1136 | rv = ss->load_super(st, fd, NULL); |
570c0542 NB |
1137 | if (rv == 0) { |
1138 | struct mdinfo info; | |
a5d85af7 | 1139 | st->ss->getinfo_super(st, &info, NULL); |
570c0542 NB |
1140 | if (bestsuper == -1 || |
1141 | besttime < info.array.ctime) { | |
1142 | bestsuper = i; | |
1143 | besttime = info.array.ctime; | |
570c0542 | 1144 | } |
3da92f27 | 1145 | ss->free_super(st); |
570c0542 NB |
1146 | } |
1147 | } | |
1148 | if (bestsuper != -1) { | |
1149 | int rv; | |
ef609477 | 1150 | memset(st, 0, sizeof(*st)); |
df3346e6 | 1151 | st->ignore_hw_compat = 1; |
3da92f27 | 1152 | rv = superlist[bestsuper]->load_super(st, fd, NULL); |
f9ce90ba | 1153 | if (rv == 0) { |
5e747af2 | 1154 | superlist[bestsuper]->free_super(st); |
82d9eba6 | 1155 | return st; |
f9ce90ba NB |
1156 | } |
1157 | } | |
570c0542 | 1158 | free(st); |
f9ce90ba NB |
1159 | return NULL; |
1160 | } | |
fe6729fa | 1161 | |
beae1dfe NB |
1162 | /* Return size of device in bytes */ |
1163 | int get_dev_size(int fd, char *dname, unsigned long long *sizep) | |
1164 | { | |
1165 | unsigned long long ldsize; | |
c2c9bb6f NB |
1166 | struct stat st; |
1167 | ||
1168 | if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode)) | |
1169 | ldsize = (unsigned long long)st.st_size; | |
1170 | else | |
beae1dfe NB |
1171 | #ifdef BLKGETSIZE64 |
1172 | if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0) | |
1173 | #endif | |
1174 | { | |
1175 | unsigned long dsize; | |
1176 | if (ioctl(fd, BLKGETSIZE, &dsize) == 0) { | |
1177 | ldsize = dsize; | |
1178 | ldsize <<= 9; | |
1179 | } else { | |
1180 | if (dname) | |
e7b84f9d | 1181 | pr_err("Cannot get size of %s: %s\b", |
beae1dfe NB |
1182 | dname, strerror(errno)); |
1183 | return 0; | |
1184 | } | |
1185 | } | |
1186 | *sizep = ldsize; | |
1187 | return 1; | |
1188 | } | |
8fac0577 | 1189 | |
3a371610 N |
1190 | /* Return true if this can only be a container, not a member device. |
1191 | * i.e. is and md device and size is zero | |
1192 | */ | |
1193 | int must_be_container(int fd) | |
1194 | { | |
1195 | unsigned long long size; | |
1196 | if (md_get_version(fd) < 0) | |
1197 | return 0; | |
1198 | if (get_dev_size(fd, NULL, &size) == 0) | |
1199 | return 1; | |
1200 | if (size == 0) | |
1201 | return 1; | |
1202 | return 0; | |
1203 | } | |
034b203a TM |
1204 | |
1205 | /* Sets endofpart parameter to the last block used by the last GPT partition on the device. | |
1206 | * Returns: 1 if successful | |
1207 | * -1 for unknown partition type | |
1208 | * 0 for other errors | |
1209 | */ | |
1210 | static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart) | |
1211 | { | |
056b331e | 1212 | struct GPT gpt; |
034b203a | 1213 | unsigned char empty_gpt_entry[16]= {0}; |
db7fdfe4 JS |
1214 | struct GPT_part_entry *part; |
1215 | char buf[512]; | |
034b203a TM |
1216 | unsigned long long curr_part_end; |
1217 | unsigned all_partitions, entry_size; | |
f21e18ca | 1218 | unsigned part_nr; |
034b203a TM |
1219 | |
1220 | *endofpart = 0; | |
1221 | ||
056b331e | 1222 | BUILD_BUG_ON(sizeof(gpt) != 512); |
73e658d8 | 1223 | /* skip protective MBR */ |
034b203a | 1224 | lseek(fd, 512, SEEK_SET); |
73e658d8 | 1225 | /* read GPT header */ |
056b331e | 1226 | if (read(fd, &gpt, 512) != 512) |
034b203a TM |
1227 | return 0; |
1228 | ||
1229 | /* get the number of partition entries and the entry size */ | |
056b331e N |
1230 | all_partitions = __le32_to_cpu(gpt.part_cnt); |
1231 | entry_size = __le32_to_cpu(gpt.part_size); | |
034b203a TM |
1232 | |
1233 | /* Check GPT signature*/ | |
056b331e | 1234 | if (gpt.magic != GPT_SIGNATURE_MAGIC) |
034b203a TM |
1235 | return -1; |
1236 | ||
1237 | /* sanity checks */ | |
1238 | if (all_partitions > 1024 || | |
db7fdfe4 | 1239 | entry_size > sizeof(buf)) |
034b203a TM |
1240 | return -1; |
1241 | ||
db7fdfe4 JS |
1242 | part = (struct GPT_part_entry *)buf; |
1243 | ||
ca3b6696 | 1244 | for (part_nr = 0; part_nr < all_partitions; part_nr++) { |
73e658d8 | 1245 | /* read partition entry */ |
db7fdfe4 | 1246 | if (read(fd, buf, entry_size) != (ssize_t)entry_size) |
73e658d8 LB |
1247 | return 0; |
1248 | ||
034b203a | 1249 | /* is this valid partition? */ |
db7fdfe4 | 1250 | if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) { |
034b203a | 1251 | /* check the last lba for the current partition */ |
db7fdfe4 | 1252 | curr_part_end = __le64_to_cpu(part->ending_lba); |
034b203a TM |
1253 | if (curr_part_end > *endofpart) |
1254 | *endofpart = curr_part_end; | |
1255 | } | |
1256 | ||
034b203a TM |
1257 | } |
1258 | return 1; | |
1259 | } | |
1260 | ||
1261 | /* Sets endofpart parameter to the last block used by the last partition on the device. | |
1262 | * Returns: 1 if successful | |
1263 | * -1 for unknown partition type | |
1264 | * 0 for other errors | |
1265 | */ | |
1266 | static int get_last_partition_end(int fd, unsigned long long *endofpart) | |
1267 | { | |
056b331e | 1268 | struct MBR boot_sect; |
034b203a TM |
1269 | struct MBR_part_record *part; |
1270 | unsigned long long curr_part_end; | |
f21e18ca | 1271 | unsigned part_nr; |
034b203a TM |
1272 | int retval = 0; |
1273 | ||
1274 | *endofpart = 0; | |
1275 | ||
056b331e | 1276 | BUILD_BUG_ON(sizeof(boot_sect) != 512); |
034b203a TM |
1277 | /* read MBR */ |
1278 | lseek(fd, 0, 0); | |
056b331e | 1279 | if (read(fd, &boot_sect, 512) != 512) |
034b203a TM |
1280 | goto abort; |
1281 | ||
1282 | /* check MBP signature */ | |
056b331e | 1283 | if (boot_sect.magic == MBR_SIGNATURE_MAGIC) { |
034b203a TM |
1284 | retval = 1; |
1285 | /* found the correct signature */ | |
056b331e | 1286 | part = boot_sect.parts; |
034b203a | 1287 | |
ca3b6696 | 1288 | for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) { |
034b203a TM |
1289 | /* check for GPT type */ |
1290 | if (part->part_type == MBR_GPT_PARTITION_TYPE) { | |
1291 | retval = get_gpt_last_partition_end(fd, endofpart); | |
1292 | break; | |
1293 | } | |
1294 | /* check the last used lba for the current partition */ | |
1295 | curr_part_end = __le32_to_cpu(part->first_sect_lba) + | |
1296 | __le32_to_cpu(part->blocks_num); | |
1297 | if (curr_part_end > *endofpart) | |
1298 | *endofpart = curr_part_end; | |
1299 | ||
1300 | part++; | |
1301 | } | |
1302 | } else { | |
1303 | /* Unknown partition table */ | |
1304 | retval = -1; | |
1305 | } | |
1306 | abort: | |
1307 | return retval; | |
1308 | } | |
1309 | ||
53ed6ac3 KW |
1310 | int check_partitions(int fd, char *dname, unsigned long long freesize, |
1311 | unsigned long long size) | |
034b203a TM |
1312 | { |
1313 | /* | |
1314 | * Check where the last partition ends | |
1315 | */ | |
1316 | unsigned long long endofpart; | |
1317 | int ret; | |
1318 | ||
1319 | if ((ret = get_last_partition_end(fd, &endofpart)) > 0) { | |
1320 | /* There appears to be a partition table here */ | |
1321 | if (freesize == 0) { | |
1322 | /* partitions will not be visible in new device */ | |
e7b84f9d N |
1323 | pr_err("partition table exists on %s but will be lost or\n" |
1324 | " meaningless after creating array\n", | |
1325 | dname); | |
034b203a TM |
1326 | return 1; |
1327 | } else if (endofpart > freesize) { | |
1328 | /* last partition overlaps metadata */ | |
e7b84f9d N |
1329 | pr_err("metadata will over-write last partition on %s.\n", |
1330 | dname); | |
034b203a | 1331 | return 1; |
53ed6ac3 KW |
1332 | } else if (size && endofpart > size) { |
1333 | /* partitions will be truncated in new device */ | |
e7b84f9d N |
1334 | pr_err("array size is too small to cover all partitions on %s.\n", |
1335 | dname); | |
53ed6ac3 | 1336 | return 1; |
034b203a TM |
1337 | } |
1338 | } | |
1339 | return 0; | |
1340 | } | |
1341 | ||
a322f70c DW |
1342 | int open_container(int fd) |
1343 | { | |
1344 | /* 'fd' is a block device. Find out if it is in use | |
1345 | * by a container, and return an open fd on that container. | |
1346 | */ | |
1347 | char path[256]; | |
1348 | char *e; | |
1349 | DIR *dir; | |
1350 | struct dirent *de; | |
1351 | int dfd, n; | |
1352 | char buf[200]; | |
1353 | int major, minor; | |
1354 | struct stat st; | |
1355 | ||
1356 | if (fstat(fd, &st) != 0) | |
1357 | return -1; | |
1358 | sprintf(path, "/sys/dev/block/%d:%d/holders", | |
1359 | (int)major(st.st_rdev), (int)minor(st.st_rdev)); | |
1360 | e = path + strlen(path); | |
1361 | ||
1362 | dir = opendir(path); | |
1363 | if (!dir) | |
1364 | return -1; | |
1365 | while ((de = readdir(dir))) { | |
1366 | if (de->d_ino == 0) | |
1367 | continue; | |
1368 | if (de->d_name[0] == '.') | |
1369 | continue; | |
a994592d N |
1370 | /* Need to make sure it is a container and not a volume */ |
1371 | sprintf(e, "/%s/md/metadata_version", de->d_name); | |
1372 | dfd = open(path, O_RDONLY); | |
1373 | if (dfd < 0) | |
1374 | continue; | |
1375 | n = read(dfd, buf, sizeof(buf)); | |
1376 | close(dfd); | |
1377 | if (n <= 0 || (unsigned)n >= sizeof(buf)) | |
1378 | continue; | |
1379 | buf[n] = 0; | |
1380 | if (strncmp(buf, "external", 8) != 0 || | |
1381 | n < 10 || | |
1382 | buf[9] == '/') | |
1383 | continue; | |
a322f70c DW |
1384 | sprintf(e, "/%s/dev", de->d_name); |
1385 | dfd = open(path, O_RDONLY); | |
1386 | if (dfd < 0) | |
1387 | continue; | |
1388 | n = read(dfd, buf, sizeof(buf)); | |
1389 | close(dfd); | |
f21e18ca | 1390 | if (n <= 0 || (unsigned)n >= sizeof(buf)) |
a322f70c DW |
1391 | continue; |
1392 | buf[n] = 0; | |
1393 | if (sscanf(buf, "%d:%d", &major, &minor) != 2) | |
1394 | continue; | |
1395 | sprintf(buf, "%d:%d", major, minor); | |
1396 | dfd = dev_open(buf, O_RDONLY); | |
1397 | if (dfd >= 0) { | |
1398 | closedir(dir); | |
1399 | return dfd; | |
1400 | } | |
1401 | } | |
355726fa | 1402 | closedir(dir); |
a322f70c DW |
1403 | return -1; |
1404 | } | |
1405 | ||
33414a01 DW |
1406 | struct superswitch *version_to_superswitch(char *vers) |
1407 | { | |
1408 | int i; | |
1409 | ||
1410 | for (i = 0; superlist[i]; i++) { | |
1411 | struct superswitch *ss = superlist[i]; | |
1412 | ||
1413 | if (strcmp(vers, ss->name) == 0) | |
1414 | return ss; | |
1415 | } | |
1416 | ||
1417 | return NULL; | |
1418 | } | |
1419 | ||
4dd2df09 N |
1420 | int metadata_container_matches(char *metadata, char *devnm) |
1421 | { | |
1422 | /* Check if 'devnm' is the container named in 'metadata' | |
1423 | * which is | |
1424 | * /containername/componentname or | |
1425 | * -containername/componentname | |
1426 | */ | |
1427 | int l; | |
1428 | if (*metadata != '/' && *metadata != '-') | |
1429 | return 0; | |
1430 | l = strlen(devnm); | |
1431 | if (strncmp(metadata+1, devnm, l) != 0) | |
1432 | return 0; | |
1433 | if (metadata[l+1] != '/') | |
1434 | return 0; | |
1435 | return 1; | |
1436 | } | |
1437 | ||
1438 | int metadata_subdev_matches(char *metadata, char *devnm) | |
1439 | { | |
1440 | /* Check if 'devnm' is the subdev named in 'metadata' | |
1441 | * which is | |
1442 | * /containername/subdev or | |
1443 | * -containername/subdev | |
1444 | */ | |
1445 | char *sl; | |
1446 | if (*metadata != '/' && *metadata != '-') | |
1447 | return 0; | |
1448 | sl = strchr(metadata+1, '/'); | |
1449 | if (!sl) | |
1450 | return 0; | |
1451 | if (strcmp(sl+1, devnm) == 0) | |
1452 | return 1; | |
1453 | return 0; | |
1454 | } | |
1455 | ||
33414a01 DW |
1456 | int is_container_member(struct mdstat_ent *mdstat, char *container) |
1457 | { | |
1458 | if (mdstat->metadata_version == NULL || | |
1459 | strncmp(mdstat->metadata_version, "external:", 9) != 0 || | |
4dd2df09 | 1460 | !metadata_container_matches(mdstat->metadata_version+9, container)) |
33414a01 DW |
1461 | return 0; |
1462 | ||
1463 | return 1; | |
1464 | } | |
1465 | ||
1466 | int is_subarray_active(char *subarray, char *container) | |
1467 | { | |
1468 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
1469 | struct mdstat_ent *ent; | |
1470 | ||
e5408a32 DW |
1471 | for (ent = mdstat; ent; ent = ent->next) |
1472 | if (is_container_member(ent, container)) | |
e5e5d7ce | 1473 | if (strcmp(to_subarray(ent, container), subarray) == 0) |
33414a01 | 1474 | break; |
33414a01 DW |
1475 | |
1476 | free_mdstat(mdstat); | |
1477 | ||
1478 | return ent != NULL; | |
1479 | } | |
1480 | ||
1481 | /* open_subarray - opens a subarray in a container | |
1482 | * @dev: container device name | |
feab51f8 | 1483 | * @st: empty supertype |
33414a01 DW |
1484 | * @quiet: block reporting errors flag |
1485 | * | |
1486 | * On success returns an fd to a container and fills in *st | |
1487 | */ | |
feab51f8 | 1488 | int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet) |
33414a01 DW |
1489 | { |
1490 | struct mdinfo *mdi; | |
a951a4f7 | 1491 | struct mdinfo *info; |
33414a01 | 1492 | int fd, err = 1; |
4dd2df09 | 1493 | char *_devnm; |
33414a01 DW |
1494 | |
1495 | fd = open(dev, O_RDWR|O_EXCL); | |
1496 | if (fd < 0) { | |
1497 | if (!quiet) | |
e7b84f9d | 1498 | pr_err("Couldn't open %s, aborting\n", |
33414a01 | 1499 | dev); |
b990032d | 1500 | return -1; |
33414a01 DW |
1501 | } |
1502 | ||
4dd2df09 N |
1503 | _devnm = fd2devnm(fd); |
1504 | if (_devnm == NULL) { | |
33414a01 | 1505 | if (!quiet) |
e7b84f9d N |
1506 | pr_err("Failed to determine device number for %s\n", |
1507 | dev); | |
33414a01 DW |
1508 | goto close_fd; |
1509 | } | |
4dd2df09 | 1510 | strcpy(st->devnm, _devnm); |
33414a01 | 1511 | |
4dd2df09 | 1512 | mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL); |
33414a01 DW |
1513 | if (!mdi) { |
1514 | if (!quiet) | |
e7b84f9d | 1515 | pr_err("Failed to read sysfs for %s\n", |
33414a01 DW |
1516 | dev); |
1517 | goto close_fd; | |
1518 | } | |
1519 | ||
1520 | if (mdi->array.level != UnSet) { | |
1521 | if (!quiet) | |
e7b84f9d | 1522 | pr_err("%s is not a container\n", dev); |
33414a01 DW |
1523 | goto free_sysfs; |
1524 | } | |
1525 | ||
1526 | st->ss = version_to_superswitch(mdi->text_version); | |
1527 | if (!st->ss) { | |
1528 | if (!quiet) | |
e7b84f9d N |
1529 | pr_err("Operation not supported for %s metadata\n", |
1530 | mdi->text_version); | |
33414a01 DW |
1531 | goto free_sysfs; |
1532 | } | |
1533 | ||
4dd2df09 | 1534 | if (st->devnm[0] == 0) { |
33414a01 | 1535 | if (!quiet) |
e7b84f9d | 1536 | pr_err("Failed to allocate device name\n"); |
33414a01 DW |
1537 | goto free_sysfs; |
1538 | } | |
1539 | ||
db20d413 | 1540 | if (!st->ss->load_container) { |
33414a01 | 1541 | if (!quiet) |
e7b84f9d | 1542 | pr_err("%s is not a container\n", dev); |
4dd2df09 | 1543 | goto free_sysfs; |
33414a01 DW |
1544 | } |
1545 | ||
db20d413 | 1546 | if (st->ss->load_container(st, fd, NULL)) { |
33414a01 | 1547 | if (!quiet) |
e7b84f9d | 1548 | pr_err("Failed to load metadata for %s\n", |
db20d413 | 1549 | dev); |
4dd2df09 | 1550 | goto free_sysfs; |
33414a01 DW |
1551 | } |
1552 | ||
a951a4f7 N |
1553 | info = st->ss->container_content(st, subarray); |
1554 | if (!info) { | |
1555 | if (!quiet) | |
e7b84f9d | 1556 | pr_err("Failed to find subarray-%s in %s\n", |
a951a4f7 N |
1557 | subarray, dev); |
1558 | goto free_super; | |
1559 | } | |
1560 | free(info); | |
1561 | ||
33414a01 DW |
1562 | err = 0; |
1563 | ||
1564 | free_super: | |
1565 | if (err) | |
1566 | st->ss->free_super(st); | |
33414a01 DW |
1567 | free_sysfs: |
1568 | sysfs_free(mdi); | |
1569 | close_fd: | |
1570 | if (err) | |
1571 | close(fd); | |
1572 | ||
1573 | if (err) | |
1574 | return -1; | |
1575 | else | |
1576 | return fd; | |
1577 | } | |
1578 | ||
7801ac20 N |
1579 | int add_disk(int mdfd, struct supertype *st, |
1580 | struct mdinfo *sra, struct mdinfo *info) | |
1581 | { | |
1582 | /* Add a device to an array, in one of 2 ways. */ | |
1583 | int rv; | |
1584 | #ifndef MDASSEMBLE | |
1585 | if (st->ss->external) { | |
d23534e4 DW |
1586 | if (info->disk.state & (1<<MD_DISK_SYNC)) |
1587 | info->recovery_start = MaxSector; | |
1588 | else | |
1589 | info->recovery_start = 0; | |
2904b26f | 1590 | rv = sysfs_add_disk(sra, info, 0); |
7801ac20 N |
1591 | if (! rv) { |
1592 | struct mdinfo *sd2; | |
f35f2525 N |
1593 | for (sd2 = sra->devs; sd2; sd2=sd2->next) |
1594 | if (sd2 == info) | |
1595 | break; | |
1596 | if (sd2 == NULL) { | |
503975b9 | 1597 | sd2 = xmalloc(sizeof(*sd2)); |
f35f2525 N |
1598 | *sd2 = *info; |
1599 | sd2->next = sra->devs; | |
1600 | sra->devs = sd2; | |
1601 | } | |
7801ac20 N |
1602 | } |
1603 | } else | |
1604 | #endif | |
1605 | rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk); | |
1606 | return rv; | |
1607 | } | |
1608 | ||
de6ae750 N |
1609 | int remove_disk(int mdfd, struct supertype *st, |
1610 | struct mdinfo *sra, struct mdinfo *info) | |
1611 | { | |
1612 | int rv; | |
1613 | /* Remove the disk given by 'info' from the array */ | |
1614 | #ifndef MDASSEMBLE | |
1615 | if (st->ss->external) | |
1616 | rv = sysfs_set_str(sra, info, "slot", "none"); | |
1617 | else | |
1618 | #endif | |
1619 | rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major, | |
1620 | info->disk.minor)); | |
1621 | return rv; | |
1622 | } | |
1623 | ||
f35f2525 N |
1624 | int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info) |
1625 | { | |
1626 | /* Initialise kernel's knowledge of array. | |
1627 | * This varies between externally managed arrays | |
1628 | * and older kernels | |
1629 | */ | |
1630 | int vers = md_get_version(mdfd); | |
1631 | int rv; | |
1632 | ||
1633 | #ifndef MDASSEMBLE | |
1634 | if (st->ss->external) | |
1635 | rv = sysfs_set_array(info, vers); | |
1636 | else | |
1637 | #endif | |
1638 | if ((vers % 100) >= 1) { /* can use different versions */ | |
1639 | mdu_array_info_t inf; | |
1640 | memset(&inf, 0, sizeof(inf)); | |
1641 | inf.major_version = info->array.major_version; | |
1642 | inf.minor_version = info->array.minor_version; | |
1643 | rv = ioctl(mdfd, SET_ARRAY_INFO, &inf); | |
1644 | } else | |
1645 | rv = ioctl(mdfd, SET_ARRAY_INFO, NULL); | |
1646 | return rv; | |
1647 | } | |
1648 | ||
1e5c6983 DW |
1649 | unsigned long long min_recovery_start(struct mdinfo *array) |
1650 | { | |
1651 | /* find the minimum recovery_start in an array for metadata | |
1652 | * formats that only record per-array recovery progress instead | |
1653 | * of per-device | |
1654 | */ | |
1655 | unsigned long long recovery_start = MaxSector; | |
1656 | struct mdinfo *d; | |
1657 | ||
1658 | for (d = array->devs; d; d = d->next) | |
1659 | recovery_start = min(recovery_start, d->recovery_start); | |
1660 | ||
1661 | return recovery_start; | |
1662 | } | |
1663 | ||
4dd2df09 | 1664 | int mdmon_pid(char *devnm) |
a931db9e NB |
1665 | { |
1666 | char path[100]; | |
1667 | char pid[10]; | |
1668 | int fd; | |
1669 | int n; | |
10013317 | 1670 | |
4dd2df09 | 1671 | sprintf(path, "%s/%s.pid", MDMON_DIR, devnm); |
10013317 | 1672 | |
24f6f99b | 1673 | fd = open(path, O_RDONLY | O_NOATIME, 0); |
a931db9e NB |
1674 | |
1675 | if (fd < 0) | |
cf556303 | 1676 | return -1; |
a931db9e NB |
1677 | n = read(fd, pid, 9); |
1678 | close(fd); | |
1679 | if (n <= 0) | |
cf556303 | 1680 | return -1; |
24f6f99b | 1681 | return atoi(pid); |
a931db9e NB |
1682 | } |
1683 | ||
4dd2df09 | 1684 | int mdmon_running(char *devnm) |
a931db9e | 1685 | { |
4dd2df09 | 1686 | int pid = mdmon_pid(devnm); |
24f6f99b | 1687 | if (pid <= 0) |
a931db9e | 1688 | return 0; |
24f6f99b | 1689 | if (kill(pid, 0) == 0) |
a931db9e NB |
1690 | return 1; |
1691 | return 0; | |
1692 | } | |
1693 | ||
4dd2df09 | 1694 | int start_mdmon(char *devnm) |
8850ee3e | 1695 | { |
1913c325 | 1696 | int i, skipped; |
44d2e365 | 1697 | int len; |
ca3b6696 | 1698 | pid_t pid; |
9fe32043 | 1699 | int status; |
44d2e365 N |
1700 | char pathbuf[1024]; |
1701 | char *paths[4] = { | |
1702 | pathbuf, | |
85945e19 | 1703 | BINDIR "/mdmon", |
2f1bcf43 | 1704 | "./mdmon", |
44d2e365 N |
1705 | NULL |
1706 | }; | |
8850ee3e | 1707 | |
40ebbb9c | 1708 | if (check_env("MDADM_NO_MDMON")) |
8850ee3e N |
1709 | return 0; |
1710 | ||
9cf014ec | 1711 | len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1); |
44d2e365 N |
1712 | if (len > 0) { |
1713 | char *sl; | |
1714 | pathbuf[len] = 0; | |
1715 | sl = strrchr(pathbuf, '/'); | |
1716 | if (sl) | |
1717 | sl++; | |
1718 | else | |
1719 | sl = pathbuf; | |
1720 | strcpy(sl, "mdmon"); | |
1721 | } else | |
1722 | pathbuf[0] = '\0'; | |
1723 | ||
0f7bdf89 | 1724 | /* First try to run systemctl */ |
9540cc24 N |
1725 | if (!check_env("MDADM_NO_SYSTEMCTL")) |
1726 | switch(fork()) { | |
1727 | case 0: | |
1728 | /* FIXME yuk. CLOSE_EXEC?? */ | |
1729 | skipped = 0; | |
1730 | for (i = 3; skipped < 20; i++) | |
1731 | if (close(i) < 0) | |
1732 | skipped++; | |
1733 | else | |
1734 | skipped = 0; | |
1735 | ||
1736 | /* Don't want to see error messages from | |
1737 | * systemctl. If the service doesn't exist, | |
1738 | * we start mdmon ourselves. | |
1739 | */ | |
1740 | close(2); | |
1741 | open("/dev/null", O_WRONLY); | |
1742 | snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service", | |
1743 | devnm); | |
1744 | status = execl("/usr/bin/systemctl", "systemctl", | |
1745 | "start", | |
1746 | pathbuf, NULL); | |
1747 | status = execl("/bin/systemctl", "systemctl", "start", | |
1748 | pathbuf, NULL); | |
1749 | exit(1); | |
7a862a02 | 1750 | case -1: pr_err("cannot run mdmon. Array remains readonly\n"); |
9540cc24 N |
1751 | return -1; |
1752 | default: /* parent - good */ | |
1753 | pid = wait(&status); | |
1754 | if (pid >= 0 && status == 0) | |
1755 | return 0; | |
1756 | } | |
0f7bdf89 JS |
1757 | |
1758 | /* That failed, try running mdmon directly */ | |
8850ee3e N |
1759 | switch(fork()) { |
1760 | case 0: | |
1761 | /* FIXME yuk. CLOSE_EXEC?? */ | |
1913c325 | 1762 | skipped = 0; |
ca3b6696 | 1763 | for (i = 3; skipped < 20; i++) |
1913c325 N |
1764 | if (close(i) < 0) |
1765 | skipped++; | |
1766 | else | |
1767 | skipped = 0; | |
1768 | ||
ca3b6696 | 1769 | for (i = 0; paths[i]; i++) |
a0963a86 | 1770 | if (paths[i][0]) { |
2f1bcf43 | 1771 | execl(paths[i], paths[i], |
4dd2df09 | 1772 | devnm, NULL); |
a0963a86 | 1773 | } |
8850ee3e | 1774 | exit(1); |
7a862a02 | 1775 | case -1: pr_err("cannot run mdmon. Array remains readonly\n"); |
8850ee3e | 1776 | return -1; |
9fe32043 N |
1777 | default: /* parent - good */ |
1778 | pid = wait(&status); | |
15c10423 | 1779 | if (pid < 0 || status != 0) { |
7a862a02 | 1780 | pr_err("failed to launch mdmon. Array remains readonly\n"); |
9fe32043 | 1781 | return -1; |
15c10423 | 1782 | } |
8850ee3e N |
1783 | } |
1784 | return 0; | |
1785 | } | |
1786 | ||
148acb7b DW |
1787 | __u32 random32(void) |
1788 | { | |
1789 | __u32 rv; | |
1790 | int rfd = open("/dev/urandom", O_RDONLY); | |
1791 | if (rfd < 0 || read(rfd, &rv, 4) != 4) | |
1792 | rv = random(); | |
1793 | if (rfd >= 0) | |
1794 | close(rfd); | |
1795 | return rv; | |
1796 | } | |
1797 | ||
0e600426 | 1798 | #ifndef MDASSEMBLE |
edd8d13c NB |
1799 | int flush_metadata_updates(struct supertype *st) |
1800 | { | |
1801 | int sfd; | |
1802 | if (!st->updates) { | |
1803 | st->update_tail = NULL; | |
1804 | return -1; | |
1805 | } | |
1806 | ||
4dd2df09 | 1807 | sfd = connect_monitor(st->container_devnm); |
edd8d13c NB |
1808 | if (sfd < 0) |
1809 | return -1; | |
1810 | ||
1811 | while (st->updates) { | |
1812 | struct metadata_update *mu = st->updates; | |
1813 | st->updates = mu->next; | |
1814 | ||
1815 | send_message(sfd, mu, 0); | |
1816 | wait_reply(sfd, 0); | |
1817 | free(mu->buf); | |
1818 | free(mu); | |
1819 | } | |
1820 | ack(sfd, 0); | |
1821 | wait_reply(sfd, 0); | |
1822 | close(sfd); | |
1823 | st->update_tail = NULL; | |
1824 | return 0; | |
1825 | } | |
1826 | ||
1827 | void append_metadata_update(struct supertype *st, void *buf, int len) | |
1828 | { | |
1829 | ||
503975b9 | 1830 | struct metadata_update *mu = xmalloc(sizeof(*mu)); |
edd8d13c NB |
1831 | |
1832 | mu->buf = buf; | |
1833 | mu->len = len; | |
1834 | mu->space = NULL; | |
cb23f1f4 | 1835 | mu->space_list = NULL; |
edd8d13c NB |
1836 | mu->next = NULL; |
1837 | *st->update_tail = mu; | |
1838 | st->update_tail = &mu->next; | |
1839 | } | |
0e600426 | 1840 | #endif /* MDASSEMBLE */ |
a931db9e | 1841 | |
fe6729fa NB |
1842 | #ifdef __TINYC__ |
1843 | /* tinyc doesn't optimize this check in ioctl.h out ... */ | |
1844 | unsigned int __invalid_size_argument_for_IOC = 0; | |
1845 | #endif | |
1846 | ||
6d11ec6f AK |
1847 | int experimental(void) |
1848 | { | |
1849 | if (check_env("MDADM_EXPERIMENTAL")) | |
1850 | return 1; | |
1851 | else { | |
7a862a02 | 1852 | pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n"); |
6d11ec6f AK |
1853 | return 0; |
1854 | } | |
1855 | } | |
1856 | ||
326727d9 AC |
1857 | /* Pick all spares matching given criteria from a container |
1858 | * if min_size == 0 do not check size | |
1859 | * if domlist == NULL do not check domains | |
1860 | * if spare_group given add it to domains of each spare | |
1861 | * metadata allows to test domains using metadata of destination array */ | |
1862 | struct mdinfo *container_choose_spares(struct supertype *st, | |
1863 | unsigned long long min_size, | |
1864 | struct domainlist *domlist, | |
1865 | char *spare_group, | |
1866 | const char *metadata, int get_one) | |
1867 | { | |
1868 | struct mdinfo *d, **dp, *disks = NULL; | |
1869 | ||
1870 | /* get list of all disks in container */ | |
1871 | if (st->ss->getinfo_super_disks) | |
1872 | disks = st->ss->getinfo_super_disks(st); | |
1873 | ||
1874 | if (!disks) | |
1875 | return disks; | |
1876 | /* find spare devices on the list */ | |
1877 | dp = &disks->devs; | |
1878 | disks->array.spare_disks = 0; | |
1879 | while (*dp) { | |
1880 | int found = 0; | |
1881 | d = *dp; | |
1882 | if (d->disk.state == 0) { | |
1883 | /* check if size is acceptable */ | |
1884 | unsigned long long dev_size; | |
1885 | dev_t dev = makedev(d->disk.major,d->disk.minor); | |
1886 | ||
1887 | if (!min_size || | |
1888 | (dev_size_from_id(dev, &dev_size) && | |
1889 | dev_size >= min_size)) | |
1890 | found = 1; | |
1891 | /* check if domain matches */ | |
1892 | if (found && domlist) { | |
4dd2df09 | 1893 | struct dev_policy *pol = devid_policy(dev); |
326727d9 AC |
1894 | if (spare_group) |
1895 | pol_add(&pol, pol_domain, | |
1896 | spare_group, NULL); | |
e5508b36 | 1897 | if (domain_test(domlist, pol, metadata) != 1) |
326727d9 AC |
1898 | found = 0; |
1899 | dev_policy_free(pol); | |
1900 | } | |
1901 | } | |
1902 | if (found) { | |
1903 | dp = &d->next; | |
1904 | disks->array.spare_disks++; | |
1905 | if (get_one) { | |
1906 | sysfs_free(*dp); | |
1907 | d->next = NULL; | |
1908 | } | |
1909 | } else { | |
1910 | *dp = d->next; | |
1911 | d->next = NULL; | |
1912 | sysfs_free(d); | |
1913 | } | |
1914 | } | |
1915 | return disks; | |
1916 | } | |
9eafa1de MN |
1917 | |
1918 | /* Checks if paths point to the same device | |
1919 | * Returns 0 if they do. | |
1920 | * Returns 1 if they don't. | |
1921 | * Returns -1 if something went wrong, | |
1922 | * e.g. paths are empty or the files | |
1923 | * they point to don't exist */ | |
1924 | int compare_paths (char* path1, char* path2) | |
1925 | { | |
1926 | struct stat st1,st2; | |
1927 | ||
1928 | if (path1 == NULL || path2 == NULL) | |
1929 | return -1; | |
1930 | if (stat(path1,&st1) != 0) | |
1931 | return -1; | |
1932 | if (stat(path2,&st2) != 0) | |
1933 | return -1; | |
1934 | if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev)) | |
1935 | return 0; | |
1936 | return 1; | |
1937 | } | |
a7dec3fd N |
1938 | |
1939 | /* Make sure we can open as many devices as needed */ | |
1940 | void enable_fds(int devices) | |
1941 | { | |
1942 | unsigned int fds = 20 + devices; | |
1943 | struct rlimit lim; | |
1944 | if (getrlimit(RLIMIT_NOFILE, &lim) != 0 | |
1945 | || lim.rlim_cur >= fds) | |
1946 | return; | |
1947 | if (lim.rlim_max < fds) | |
1948 | lim.rlim_max = fds; | |
1949 | lim.rlim_cur = fds; | |
1950 | setrlimit(RLIMIT_NOFILE, &lim); | |
1951 | } | |
a9c15847 N |
1952 | |
1953 | int in_initrd(void) | |
1954 | { | |
1955 | /* This is based on similar function in systemd. */ | |
1956 | struct statfs s; | |
76d0f188 JS |
1957 | /* statfs.f_type is signed long on s390x and MIPS, causing all |
1958 | sorts of sign extension problems with RAMFS_MAGIC being | |
1959 | defined as 0x858458f6 */ | |
a9c15847 | 1960 | return statfs("/", &s) >= 0 && |
7ac5d47e | 1961 | ((unsigned long)s.f_type == TMPFS_MAGIC || |
76d0f188 JS |
1962 | ((unsigned long)s.f_type & 0xFFFFFFFFUL) == |
1963 | ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL)); | |
a9c15847 | 1964 | } |
8832342d N |
1965 | |
1966 | void reopen_mddev(int mdfd) | |
1967 | { | |
1968 | /* Re-open without any O_EXCL, but keep | |
1969 | * the same fd | |
1970 | */ | |
1971 | char *devnm; | |
1972 | int fd; | |
1973 | devnm = fd2devnm(mdfd); | |
1974 | close(mdfd); | |
1975 | fd = open_dev(devnm); | |
1976 | if (fd >= 0 && fd != mdfd) | |
1977 | dup2(fd, mdfd); | |
1978 | } |