]>
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 | 32 | #include <sys/vfs.h> |
b2514242 | 33 | #include <sys/mman.h> |
a9c15847 | 34 | #include <linux/magic.h> |
986b8688 | 35 | #include <poll.h> |
98c6faba | 36 | #include <ctype.h> |
a322f70c | 37 | #include <dirent.h> |
7716570e | 38 | #include <dlfcn.h> |
7716570e | 39 | |
0a816ef9 NB |
40 | |
41 | /* | |
42 | * following taken from linux/blkpg.h because they aren't | |
43 | * anywhere else and it isn't safe to #include linux/ * stuff. | |
44 | */ | |
45 | ||
46 | #define BLKPG _IO(0x12,105) | |
47 | ||
48 | /* The argument structure */ | |
49 | struct blkpg_ioctl_arg { | |
5d500228 N |
50 | int op; |
51 | int flags; | |
52 | int datalen; | |
53 | void *data; | |
0a816ef9 NB |
54 | }; |
55 | ||
56 | /* The subfunctions (for the op field) */ | |
57 | #define BLKPG_ADD_PARTITION 1 | |
58 | #define BLKPG_DEL_PARTITION 2 | |
59 | ||
60 | /* Sizes of name fields. Unused at present. */ | |
61 | #define BLKPG_DEVNAMELTH 64 | |
62 | #define BLKPG_VOLNAMELTH 64 | |
63 | ||
64 | /* The data structure for ADD_PARTITION and DEL_PARTITION */ | |
65 | struct blkpg_partition { | |
66 | long long start; /* starting offset in bytes */ | |
67 | long long length; /* length in bytes */ | |
68 | int pno; /* partition number */ | |
69 | char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2, | |
70 | to be used in kernel messages */ | |
71 | char volname[BLKPG_VOLNAMELTH]; /* volume label */ | |
72 | }; | |
64c4757e | 73 | |
0f22b998 | 74 | #include "part.h" |
056b331e N |
75 | |
76 | /* Force a compilation error if condition is true */ | |
77 | #define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition)) | |
78 | ||
79 | /* Force a compilation error if condition is true, but also produce a | |
80 | result (of value 0 and type size_t), so the expression can be used | |
81 | e.g. in a structure initializer (or where-ever else comma expressions | |
82 | aren't permitted). */ | |
83 | #define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); })) | |
84 | ||
d15a1f72 | 85 | static int is_dlm_hooks_ready = 0; |
7071320a N |
86 | |
87 | int dlm_funs_ready(void) | |
88 | { | |
89 | return is_dlm_hooks_ready ? 1 : 0; | |
90 | } | |
91 | ||
7071320a | 92 | static struct dlm_hooks *dlm_hooks = NULL; |
d15a1f72 GJ |
93 | struct dlm_lock_resource *dlm_lock_res = NULL; |
94 | static int ast_called = 0; | |
95 | ||
96 | struct dlm_lock_resource { | |
97 | dlm_lshandle_t *ls; | |
98 | struct dlm_lksb lksb; | |
99 | }; | |
100 | ||
d15a1f72 GJ |
101 | /* Using poll(2) to wait for and dispatch ASTs */ |
102 | static int poll_for_ast(dlm_lshandle_t ls) | |
103 | { | |
104 | struct pollfd pfd; | |
105 | ||
106 | pfd.fd = dlm_hooks->ls_get_fd(ls); | |
107 | pfd.events = POLLIN; | |
108 | ||
109 | while (!ast_called) | |
110 | { | |
111 | if (poll(&pfd, 1, 0) < 0) | |
112 | { | |
113 | perror("poll"); | |
114 | return -1; | |
115 | } | |
116 | dlm_hooks->dispatch(dlm_hooks->ls_get_fd(ls)); | |
117 | } | |
118 | ast_called = 0; | |
119 | ||
120 | return 0; | |
121 | } | |
122 | ||
123 | static void dlm_ast(void *arg) | |
124 | { | |
125 | ast_called = 1; | |
126 | } | |
127 | ||
81a8a694 | 128 | static char *cluster_name = NULL; |
d15a1f72 | 129 | /* Create the lockspace, take bitmapXXX locks on all the bitmaps. */ |
1b7eb962 | 130 | int cluster_get_dlmlock(void) |
d15a1f72 GJ |
131 | { |
132 | int ret = -1; | |
133 | char str[64]; | |
134 | int flags = LKF_NOQUEUE; | |
1b7eb962 GJ |
135 | int retry_count = 0; |
136 | ||
137 | if (!dlm_funs_ready()) { | |
138 | pr_err("Something wrong with dlm library\n"); | |
139 | return -1; | |
140 | } | |
d15a1f72 | 141 | |
81a8a694 GJ |
142 | ret = get_cluster_name(&cluster_name); |
143 | if (ret) { | |
144 | pr_err("The md can't get cluster name\n"); | |
145 | return -1; | |
146 | } | |
147 | ||
d15a1f72 | 148 | dlm_lock_res = xmalloc(sizeof(struct dlm_lock_resource)); |
1b7eb962 | 149 | dlm_lock_res->ls = dlm_hooks->open_lockspace(cluster_name); |
d15a1f72 | 150 | if (!dlm_lock_res->ls) { |
1b7eb962 GJ |
151 | dlm_lock_res->ls = dlm_hooks->create_lockspace(cluster_name, O_RDWR); |
152 | if (!dlm_lock_res->ls) { | |
153 | pr_err("%s failed to create lockspace\n", cluster_name); | |
154 | return -ENOMEM; | |
155 | } | |
156 | } else { | |
157 | pr_err("open existed %s lockspace\n", cluster_name); | |
d15a1f72 GJ |
158 | } |
159 | ||
81a8a694 | 160 | snprintf(str, 64, "bitmap%s", cluster_name); |
1b7eb962 | 161 | retry: |
b7a462e5 JS |
162 | ret = dlm_hooks->ls_lock(dlm_lock_res->ls, LKM_PWMODE, |
163 | &dlm_lock_res->lksb, flags, str, strlen(str), | |
164 | 0, dlm_ast, dlm_lock_res, NULL, NULL); | |
d15a1f72 GJ |
165 | if (ret) { |
166 | pr_err("error %d when get PW mode on lock %s\n", errno, str); | |
1b7eb962 GJ |
167 | /* let's try several times if EAGAIN happened */ |
168 | if (dlm_lock_res->lksb.sb_status == EAGAIN && retry_count < 10) { | |
239b3cc0 | 169 | sleep_for(10, 0, true); |
1b7eb962 GJ |
170 | retry_count++; |
171 | goto retry; | |
172 | } | |
32539f74 GJ |
173 | dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1); |
174 | return ret; | |
d15a1f72 GJ |
175 | } |
176 | ||
177 | /* Wait for it to complete */ | |
178 | poll_for_ast(dlm_lock_res->ls); | |
d15a1f72 | 179 | |
1b7eb962 GJ |
180 | if (dlm_lock_res->lksb.sb_status) { |
181 | pr_err("failed to lock cluster\n"); | |
182 | return -1; | |
183 | } | |
184 | return 1; | |
d15a1f72 GJ |
185 | } |
186 | ||
1b7eb962 | 187 | int cluster_release_dlmlock(void) |
d15a1f72 GJ |
188 | { |
189 | int ret = -1; | |
190 | ||
81a8a694 | 191 | if (!cluster_name) |
1b7eb962 GJ |
192 | goto out; |
193 | ||
194 | if (!dlm_lock_res->lksb.sb_lkid) | |
195 | goto out; | |
81a8a694 | 196 | |
1b7eb962 GJ |
197 | ret = dlm_hooks->ls_unlock_wait(dlm_lock_res->ls, |
198 | dlm_lock_res->lksb.sb_lkid, 0, | |
199 | &dlm_lock_res->lksb); | |
d15a1f72 GJ |
200 | if (ret) { |
201 | pr_err("error %d happened when unlock\n", errno); | |
202 | /* XXX make sure the lock is unlocked eventually */ | |
203 | goto out; | |
204 | } | |
205 | ||
206 | /* Wait for it to complete */ | |
207 | poll_for_ast(dlm_lock_res->ls); | |
208 | ||
209 | errno = dlm_lock_res->lksb.sb_status; | |
210 | if (errno != EUNLOCK) { | |
b7a462e5 JS |
211 | pr_err("error %d happened in ast when unlock lockspace\n", |
212 | errno); | |
d15a1f72 GJ |
213 | /* XXX make sure the lockspace is unlocked eventually */ |
214 | goto out; | |
215 | } | |
216 | ||
81a8a694 | 217 | ret = dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1); |
d15a1f72 GJ |
218 | if (ret) { |
219 | pr_err("error %d happened when release lockspace\n", errno); | |
220 | /* XXX make sure the lockspace is released eventually */ | |
221 | goto out; | |
222 | } | |
223 | free(dlm_lock_res); | |
224 | ||
225 | out: | |
226 | return ret; | |
227 | } | |
228 | ||
9db2ab4e JS |
229 | int md_array_valid(int fd) |
230 | { | |
231 | struct mdinfo *sra; | |
232 | int ret; | |
233 | ||
234 | sra = sysfs_read(fd, NULL, GET_ARRAY_STATE); | |
235 | if (sra) { | |
236 | if (sra->array_state != ARRAY_UNKNOWN_STATE) | |
237 | ret = 0; | |
238 | else | |
239 | ret = -ENODEV; | |
240 | ||
241 | free(sra); | |
242 | } else { | |
243 | /* | |
244 | * GET_ARRAY_INFO doesn't provide access to the proper state | |
245 | * information, so fallback to a basic check for raid_disks != 0 | |
246 | */ | |
247 | ret = ioctl(fd, RAID_VERSION); | |
248 | } | |
249 | ||
250 | return !ret; | |
251 | } | |
252 | ||
3ab8f4bf JS |
253 | int md_array_active(int fd) |
254 | { | |
255 | struct mdinfo *sra; | |
256 | struct mdu_array_info_s array; | |
a822017f | 257 | int ret = 0; |
3ab8f4bf JS |
258 | |
259 | sra = sysfs_read(fd, NULL, GET_ARRAY_STATE); | |
260 | if (sra) { | |
a822017f | 261 | if (!md_array_is_active(sra)) |
3ab8f4bf JS |
262 | ret = -ENODEV; |
263 | ||
264 | free(sra); | |
265 | } else { | |
266 | /* | |
267 | * GET_ARRAY_INFO doesn't provide access to the proper state | |
268 | * information, so fallback to a basic check for raid_disks != 0 | |
269 | */ | |
953cc7e5 | 270 | ret = md_get_array_info(fd, &array); |
3ab8f4bf JS |
271 | } |
272 | ||
273 | return !ret; | |
274 | } | |
275 | ||
a822017f MT |
276 | int md_array_is_active(struct mdinfo *info) |
277 | { | |
278 | return (info->array_state != ARRAY_CLEAR && | |
279 | info->array_state != ARRAY_INACTIVE && | |
280 | info->array_state != ARRAY_UNKNOWN_STATE); | |
281 | } | |
282 | ||
9cd39f01 JS |
283 | /* |
284 | * Get array info from the kernel. Longer term we want to deprecate the | |
285 | * ioctl and get it from sysfs. | |
286 | */ | |
287 | int md_get_array_info(int fd, struct mdu_array_info_s *array) | |
288 | { | |
289 | return ioctl(fd, GET_ARRAY_INFO, array); | |
290 | } | |
291 | ||
018a4882 JS |
292 | /* |
293 | * Set array info | |
294 | */ | |
295 | int md_set_array_info(int fd, struct mdu_array_info_s *array) | |
296 | { | |
297 | return ioctl(fd, SET_ARRAY_INFO, array); | |
298 | } | |
299 | ||
d97572f5 JS |
300 | /* |
301 | * Get disk info from the kernel. | |
302 | */ | |
303 | int md_get_disk_info(int fd, struct mdu_disk_info_s *disk) | |
304 | { | |
305 | return ioctl(fd, GET_DISK_INFO, disk); | |
306 | } | |
307 | ||
64c4757e NB |
308 | int get_linux_version() |
309 | { | |
310 | struct utsname name; | |
98c6faba | 311 | char *cp; |
f161d047 | 312 | int a = 0, b = 0,c = 0; |
64c4757e NB |
313 | if (uname(&name) <0) |
314 | return -1; | |
315 | ||
98c6faba NB |
316 | cp = name.release; |
317 | a = strtoul(cp, &cp, 10); | |
f161d047 N |
318 | if (*cp == '.') |
319 | b = strtoul(cp+1, &cp, 10); | |
320 | if (*cp == '.') | |
321 | c = strtoul(cp+1, &cp, 10); | |
98c6faba | 322 | |
682c7051 | 323 | return (a*1000000)+(b*1000)+c; |
64c4757e NB |
324 | } |
325 | ||
bc77ed53 DW |
326 | int mdadm_version(char *version) |
327 | { | |
328 | int a, b, c; | |
329 | char *cp; | |
330 | ||
331 | if (!version) | |
332 | version = Version; | |
333 | ||
334 | cp = strchr(version, '-'); | |
335 | if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v') | |
336 | return -1; | |
337 | cp += 3; | |
338 | a = strtoul(cp, &cp, 10); | |
339 | if (*cp != '.') | |
340 | return -1; | |
341 | b = strtoul(cp+1, &cp, 10); | |
342 | if (*cp == '.') | |
343 | c = strtoul(cp+1, &cp, 10); | |
344 | else | |
345 | c = 0; | |
346 | if (*cp != ' ' && *cp != '-') | |
347 | return -1; | |
348 | return (a*1000000)+(b*1000)+c; | |
349 | } | |
350 | ||
51d4261c | 351 | unsigned long long parse_size(char *size) |
84e11361 N |
352 | { |
353 | /* parse 'size' which should be a number optionally | |
42e641ab | 354 | * followed by 'K', 'M'. 'G' or 'T'. |
84e11361 N |
355 | * Without a suffix, K is assumed. |
356 | * Number returned is in sectors (half-K) | |
822e393a | 357 | * INVALID_SECTORS returned on error. |
84e11361 N |
358 | */ |
359 | char *c; | |
360 | long long s = strtoll(size, &c, 10); | |
361 | if (s > 0) { | |
362 | switch (*c) { | |
363 | case 'K': | |
364 | c++; | |
365 | default: | |
366 | s *= 2; | |
367 | break; | |
368 | case 'M': | |
369 | c++; | |
370 | s *= 1024 * 2; | |
371 | break; | |
372 | case 'G': | |
373 | c++; | |
374 | s *= 1024 * 1024 * 2; | |
375 | break; | |
42e641ab KT |
376 | case 'T': |
377 | c++; | |
378 | s *= 1024 * 1024 * 1024 * 2LL; | |
379 | break; | |
72ca9bcf N |
380 | case 's': /* sectors */ |
381 | c++; | |
382 | break; | |
84e11361 | 383 | } |
51d4261c | 384 | } else |
822e393a | 385 | s = INVALID_SECTORS; |
84e11361 | 386 | if (*c) |
822e393a | 387 | s = INVALID_SECTORS; |
84e11361 N |
388 | return s; |
389 | } | |
390 | ||
5339f996 GJ |
391 | int is_near_layout_10(int layout) |
392 | { | |
393 | int fc, fo; | |
394 | ||
395 | fc = (layout >> 8) & 255; | |
396 | fo = layout & (1 << 16); | |
397 | if (fc > 1 || fo > 0) | |
398 | return 0; | |
399 | return 1; | |
400 | } | |
401 | ||
4a06e2c2 N |
402 | int parse_layout_10(char *layout) |
403 | { | |
404 | int copies, rv; | |
405 | char *cp; | |
406 | /* Parse the layout string for raid10 */ | |
407 | /* 'f', 'o' or 'n' followed by a number <= raid_disks */ | |
408 | if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') || | |
409 | (copies = strtoul(layout+1, &cp, 10)) < 1 || | |
410 | copies > 200 || | |
411 | *cp) | |
412 | return -1; | |
413 | if (layout[0] == 'n') | |
414 | rv = 256 + copies; | |
415 | else if (layout[0] == 'o') | |
416 | rv = 0x10000 + (copies<<8) + 1; | |
417 | else | |
418 | rv = 1 + (copies<<8); | |
419 | return rv; | |
420 | } | |
421 | ||
422 | int parse_layout_faulty(char *layout) | |
423 | { | |
60815698 MG |
424 | if (!layout) |
425 | return -1; | |
4a06e2c2 N |
426 | /* Parse the layout string for 'faulty' */ |
427 | int ln = strcspn(layout, "0123456789"); | |
503975b9 | 428 | char *m = xstrdup(layout); |
4a06e2c2 N |
429 | int mode; |
430 | m[ln] = 0; | |
431 | mode = map_name(faultylayout, m); | |
432 | if (mode == UnSet) | |
433 | return -1; | |
434 | ||
435 | return mode | (atoi(layout+ln)<< ModeShift); | |
436 | } | |
79868890 | 437 | |
4de90913 GJ |
438 | int parse_cluster_confirm_arg(char *input, char **devname, int *slot) |
439 | { | |
440 | char *dev; | |
441 | *slot = strtoul(input, &dev, 10); | |
442 | if (dev == input || dev[0] != ':') | |
443 | return -1; | |
444 | *devname = dev+1; | |
445 | return 0; | |
446 | } | |
447 | ||
0430ed48 NB |
448 | void remove_partitions(int fd) |
449 | { | |
450 | /* remove partitions from this block devices. | |
451 | * This is used for components added to an array | |
452 | */ | |
453 | #ifdef BLKPG_DEL_PARTITION | |
454 | struct blkpg_ioctl_arg a; | |
455 | struct blkpg_partition p; | |
456 | ||
457 | a.op = BLKPG_DEL_PARTITION; | |
458 | a.data = (void*)&p; | |
459 | a.datalen = sizeof(p); | |
460 | a.flags = 0; | |
461 | memset(a.data, 0, a.datalen); | |
ca3b6696 | 462 | for (p.pno = 0; p.pno < 16; p.pno++) |
0430ed48 NB |
463 | ioctl(fd, BLKPG, &a); |
464 | #endif | |
465 | } | |
466 | ||
691c6ee1 N |
467 | int test_partition(int fd) |
468 | { | |
469 | /* Check if fd is a whole-disk or a partition. | |
470 | * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION | |
471 | * will return ENXIO on an invalid partition number. | |
472 | */ | |
473 | struct blkpg_ioctl_arg a; | |
474 | struct blkpg_partition p; | |
475 | a.op = BLKPG_DEL_PARTITION; | |
476 | a.data = (void*)&p; | |
477 | a.datalen = sizeof(p); | |
478 | a.flags = 0; | |
479 | memset(a.data, 0, a.datalen); | |
480 | p.pno = 1<<30; | |
481 | if (ioctl(fd, BLKPG, &a) == 0) | |
482 | /* Very unlikely, but not a partition */ | |
483 | return 0; | |
357ac106 | 484 | if (errno == ENXIO || errno == ENOTTY) |
691c6ee1 N |
485 | /* not a partition */ |
486 | return 0; | |
487 | ||
488 | return 1; | |
489 | } | |
490 | ||
bfd76b93 CA |
491 | int test_partition_from_id(dev_t id) |
492 | { | |
493 | char buf[20]; | |
494 | int fd, rv; | |
495 | ||
496 | sprintf(buf, "%d:%d", major(id), minor(id)); | |
497 | fd = dev_open(buf, O_RDONLY); | |
498 | if (fd < 0) | |
499 | return -1; | |
500 | rv = test_partition(fd); | |
501 | close(fd); | |
502 | return rv; | |
503 | } | |
691c6ee1 | 504 | |
de5a472e | 505 | int enough(int level, int raid_disks, int layout, int clean, char *avail) |
64c4757e | 506 | { |
265e0f17 | 507 | int copies, first; |
de5a472e N |
508 | int i; |
509 | int avail_disks = 0; | |
510 | ||
511 | for (i = 0; i < raid_disks; i++) | |
512 | avail_disks += !!avail[i]; | |
513 | ||
64c4757e | 514 | switch (level) { |
265e0f17 NB |
515 | case 10: |
516 | /* This is the tricky one - we need to check | |
517 | * which actual disks are present. | |
518 | */ | |
702b557b | 519 | copies = (layout&255)* ((layout>>8) & 255); |
ca3b6696 | 520 | first = 0; |
265e0f17 NB |
521 | do { |
522 | /* there must be one of the 'copies' form 'first' */ | |
523 | int n = copies; | |
ca3b6696 | 524 | int cnt = 0; |
b3ec716d | 525 | int this = first; |
265e0f17 | 526 | while (n--) { |
b3ec716d | 527 | if (avail[this]) |
265e0f17 | 528 | cnt++; |
b3ec716d | 529 | this = (this+1) % raid_disks; |
265e0f17 NB |
530 | } |
531 | if (cnt == 0) | |
532 | return 0; | |
b3ec716d | 533 | first = (first+(layout&255)) % raid_disks; |
265e0f17 NB |
534 | } while (first != 0); |
535 | return 1; | |
e5329c37 | 536 | |
df0d4ea0 | 537 | case LEVEL_MULTIPATH: |
e0d19036 | 538 | return avail_disks>= 1; |
df0d4ea0 | 539 | case LEVEL_LINEAR: |
64c4757e NB |
540 | case 0: |
541 | return avail_disks == raid_disks; | |
542 | case 1: | |
543 | return avail_disks >= 1; | |
544 | case 4: | |
330d6900 N |
545 | if (avail_disks == raid_disks - 1 && |
546 | !avail[raid_disks - 1]) | |
547 | /* If just the parity device is missing, then we | |
548 | * have enough, even if not clean | |
549 | */ | |
550 | return 1; | |
551 | /* FALL THROUGH */ | |
64c4757e | 552 | case 5: |
583315d9 NB |
553 | if (clean) |
554 | return avail_disks >= raid_disks-1; | |
555 | else | |
556 | return avail_disks >= raid_disks; | |
98c6faba | 557 | case 6: |
583315d9 NB |
558 | if (clean) |
559 | return avail_disks >= raid_disks-2; | |
560 | else | |
561 | return avail_disks >= raid_disks; | |
64c4757e NB |
562 | default: |
563 | return 0; | |
564 | } | |
565 | } | |
566 | ||
aae5a112 | 567 | char *__fname_from_uuid(int id[4], int swap, char *buf, char sep) |
d7288ddc | 568 | { |
9968e376 | 569 | int i, j; |
d7288ddc N |
570 | char uuid[16]; |
571 | char *c = buf; | |
572 | strcpy(c, "UUID-"); | |
573 | c += strlen(c); | |
aae5a112 | 574 | copy_uuid(uuid, id, swap); |
9968e376 | 575 | for (i = 0; i < 4; i++) { |
9968e376 | 576 | if (i) |
ff54de6e | 577 | *c++ = sep; |
9968e376 DW |
578 | for (j = 3; j >= 0; j--) { |
579 | sprintf(c,"%02x", (unsigned char) uuid[j+4*i]); | |
580 | c+= 2; | |
581 | } | |
d7288ddc N |
582 | } |
583 | return buf; | |
aae5a112 DW |
584 | |
585 | } | |
586 | ||
b7a462e5 JS |
587 | char *fname_from_uuid(struct supertype *st, struct mdinfo *info, |
588 | char *buf, char sep) | |
aae5a112 | 589 | { |
86983cce N |
590 | // dirty hack to work around an issue with super1 superblocks... |
591 | // super1 superblocks need swapuuid set in order for assembly to | |
592 | // work, but can't have it set if we want this printout to match | |
593 | // all the other uuid printouts in super1.c, so we force swapuuid | |
594 | // to 1 to make our printout match the rest of super1 | |
2c2d9c48 NC |
595 | #if __BYTE_ORDER == BIG_ENDIAN |
596 | return __fname_from_uuid(info->uuid, 1, buf, sep); | |
597 | #else | |
b7a462e5 JS |
598 | return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : |
599 | st->ss->swapuuid, buf, sep); | |
2c2d9c48 | 600 | #endif |
d7288ddc N |
601 | } |
602 | ||
682c7051 NB |
603 | int check_ext2(int fd, char *name) |
604 | { | |
605 | /* | |
606 | * Check for an ext2fs file system. | |
607 | * Superblock is always 1K at 1K offset | |
608 | * | |
609 | * s_magic is le16 at 56 == 0xEF53 | |
610 | * report mtime - le32 at 44 | |
611 | * blocks - le32 at 4 | |
612 | * logblksize - le32 at 24 | |
613 | */ | |
614 | unsigned char sb[1024]; | |
615 | time_t mtime; | |
2538ba2a N |
616 | unsigned long long size; |
617 | int bsize; | |
682c7051 NB |
618 | if (lseek(fd, 1024,0)!= 1024) |
619 | return 0; | |
620 | if (read(fd, sb, 1024)!= 1024) | |
621 | return 0; | |
622 | if (sb[56] != 0x53 || sb[57] != 0xef) | |
623 | return 0; | |
624 | ||
625 | mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8; | |
626 | bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8; | |
627 | size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8; | |
2538ba2a | 628 | size <<= bsize; |
e7b84f9d | 629 | pr_err("%s appears to contain an ext2fs file system\n", |
682c7051 | 630 | name); |
2538ba2a | 631 | cont_err("size=%lluK mtime=%s", size, ctime(&mtime)); |
682c7051 NB |
632 | return 1; |
633 | } | |
634 | ||
635 | int check_reiser(int fd, char *name) | |
636 | { | |
637 | /* | |
638 | * superblock is at 64K | |
639 | * size is 1024; | |
640 | * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52 | |
641 | * | |
642 | */ | |
643 | unsigned char sb[1024]; | |
2538ba2a | 644 | unsigned long long size; |
682c7051 NB |
645 | if (lseek(fd, 64*1024, 0) != 64*1024) |
646 | return 0; | |
647 | if (read(fd, sb, 1024) != 1024) | |
648 | return 0; | |
ca3b6696 N |
649 | if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 && |
650 | strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0) | |
682c7051 | 651 | return 0; |
e7b84f9d | 652 | pr_err("%s appears to contain a reiserfs file system\n",name); |
682c7051 | 653 | size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8; |
2538ba2a | 654 | cont_err("size = %lluK\n", size*4); |
aba69144 | 655 | |
682c7051 NB |
656 | return 1; |
657 | } | |
658 | ||
659 | int check_raid(int fd, char *name) | |
660 | { | |
4b1ac34b | 661 | struct mdinfo info; |
682c7051 | 662 | time_t crtime; |
d078d77c | 663 | char *level; |
82d9eba6 | 664 | struct supertype *st = guess_super(fd); |
f9ce90ba | 665 | |
cb8f6859 N |
666 | if (!st) |
667 | return 0; | |
fa219dd2 MD |
668 | if (st->ss->add_to_super != NULL) { |
669 | st->ss->load_super(st, fd, name); | |
670 | /* Looks like a raid array .. */ | |
671 | pr_err("%s appears to be part of a raid array:\n", name); | |
672 | st->ss->getinfo_super(st, &info, NULL); | |
673 | st->ss->free_super(st); | |
674 | crtime = info.array.ctime; | |
675 | level = map_num(pers, info.array.level); | |
676 | if (!level) | |
677 | level = "-unknown-"; | |
678 | cont_err("level=%s devices=%d ctime=%s", | |
679 | level, info.array.raid_disks, ctime(&crtime)); | |
680 | } else { | |
681 | /* Looks like GPT or MBR */ | |
682 | pr_err("partition table exists on %s\n", name); | |
683 | } | |
82d9eba6 | 684 | return 1; |
682c7051 NB |
685 | } |
686 | ||
0a6bff09 ZL |
687 | int fstat_is_blkdev(int fd, char *devname, dev_t *rdev) |
688 | { | |
689 | struct stat stb; | |
690 | ||
691 | if (fstat(fd, &stb) != 0) { | |
692 | pr_err("fstat failed for %s: %s\n", devname, strerror(errno)); | |
693 | return 0; | |
694 | } | |
695 | if ((S_IFMT & stb.st_mode) != S_IFBLK) { | |
696 | pr_err("%s is not a block device.\n", devname); | |
9e04ac1c ZL |
697 | return 0; |
698 | } | |
699 | if (rdev) | |
700 | *rdev = stb.st_rdev; | |
701 | return 1; | |
702 | } | |
703 | ||
704 | int stat_is_blkdev(char *devname, dev_t *rdev) | |
705 | { | |
706 | struct stat stb; | |
707 | ||
708 | if (stat(devname, &stb) != 0) { | |
709 | pr_err("stat failed for %s: %s\n", devname, strerror(errno)); | |
710 | return 0; | |
711 | } | |
712 | if ((S_IFMT & stb.st_mode) != S_IFBLK) { | |
713 | pr_err("%s is not a block device.\n", devname); | |
0a6bff09 ZL |
714 | return 0; |
715 | } | |
716 | if (rdev) | |
717 | *rdev = stb.st_rdev; | |
718 | return 1; | |
719 | } | |
720 | ||
682c7051 NB |
721 | int ask(char *mesg) |
722 | { | |
723 | char *add = ""; | |
724 | int i; | |
ca3b6696 | 725 | for (i = 0; i < 5; i++) { |
682c7051 NB |
726 | char buf[100]; |
727 | fprintf(stderr, "%s%s", mesg, add); | |
728 | fflush(stderr); | |
729 | if (fgets(buf, 100, stdin)==NULL) | |
730 | return 0; | |
731 | if (buf[0]=='y' || buf[0]=='Y') | |
732 | return 1; | |
733 | if (buf[0]=='n' || buf[0]=='N') | |
734 | return 0; | |
735 | add = "(y/n) "; | |
736 | } | |
e7b84f9d | 737 | pr_err("assuming 'no'\n"); |
682c7051 NB |
738 | return 0; |
739 | } | |
740 | ||
8d80900b | 741 | int is_standard(char *dev, int *nump) |
e5329c37 NB |
742 | { |
743 | /* tests if dev is a "standard" md dev name. | |
744 | * i.e if the last component is "/dNN" or "/mdNN", | |
aba69144 | 745 | * where NN is a string of digits |
598f0d58 NB |
746 | * Returns 1 if a partitionable standard, |
747 | * -1 if non-partitonable, | |
748 | * 0 if not a standard name. | |
e5329c37 | 749 | */ |
8d80900b | 750 | char *d = strrchr(dev, '/'); |
ca3b6696 | 751 | int type = 0; |
8d80900b NB |
752 | int num; |
753 | if (!d) | |
e5329c37 | 754 | return 0; |
ca3b6696 N |
755 | if (strncmp(d, "/d",2) == 0) |
756 | d += 2, type = 1; /* /dev/md/dN{pM} */ | |
757 | else if (strncmp(d, "/md_d", 5) == 0) | |
758 | d += 5, type = 1; /* /dev/md_dN{pM} */ | |
759 | else if (strncmp(d, "/md", 3) == 0) | |
760 | d += 3, type = -1; /* /dev/mdN */ | |
761 | else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0) | |
762 | d += 1, type = -1; /* /dev/md/N */ | |
e5329c37 NB |
763 | else |
764 | return 0; | |
8d80900b | 765 | if (!*d) |
e5329c37 | 766 | return 0; |
8d80900b NB |
767 | num = atoi(d); |
768 | while (isdigit(*d)) | |
769 | d++; | |
770 | if (*d) | |
e5329c37 | 771 | return 0; |
8d80900b NB |
772 | if (nump) *nump = num; |
773 | ||
774 | return type; | |
e5329c37 NB |
775 | } |
776 | ||
4b1ac34b | 777 | unsigned long calc_csum(void *super, int bytes) |
82b27616 | 778 | { |
56eb10c0 | 779 | unsigned long long newcsum = 0; |
82b27616 | 780 | int i; |
4b1ac34b NB |
781 | unsigned int csum; |
782 | unsigned int *superc = (unsigned int*) super; | |
82b27616 | 783 | |
ca3b6696 N |
784 | for(i = 0; i < bytes/4; i++) |
785 | newcsum += superc[i]; | |
82b27616 | 786 | csum = (newcsum& 0xffffffff) + (newcsum>>32); |
570c0542 | 787 | #ifdef __alpha__ |
aba69144 | 788 | /* The in-kernel checksum calculation is always 16bit on |
570c0542 | 789 | * the alpha, though it is 32 bit on i386... |
ca3b6696 | 790 | * I wonder what it is elsewhere... (it uses an API in |
570c0542 NB |
791 | * a way that it shouldn't). |
792 | */ | |
793 | csum = (csum & 0xffff) + (csum >> 16); | |
794 | csum = (csum & 0xffff) + (csum >> 16); | |
795 | #endif | |
82b27616 NB |
796 | return csum; |
797 | } | |
cd29a5c8 | 798 | |
56eb10c0 | 799 | char *human_size(long long bytes) |
cd29a5c8 | 800 | { |
ff9239ee | 801 | static char buf[47]; |
d5d3721e | 802 | |
42e641ab KT |
803 | /* We convert bytes to either centi-M{ega,ibi}bytes, |
804 | * centi-G{igi,ibi}bytes or centi-T{era,ebi}bytes | |
805 | * with appropriate rounding, and then print | |
806 | * 1/100th of those as a decimal. | |
d5d3721e | 807 | * We allow upto 2048Megabytes before converting to |
42e641ab KT |
808 | * gigabytes and 2048Gigabytes before converting to |
809 | * terabytes, as that shows more precision and isn't | |
d5d3721e | 810 | * too large a number. |
d5d3721e | 811 | */ |
cd29a5c8 | 812 | |
56eb10c0 | 813 | if (bytes < 5000*1024) |
ca3b6696 | 814 | buf[0] = 0; |
d5d3721e | 815 | else if (bytes < 2*1024LL*1024LL*1024LL) { |
93d3bd3b | 816 | long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2; |
d5d3721e | 817 | long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2; |
8f23b0b3 | 818 | snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)", |
efa29530 | 819 | cMiB/100, cMiB % 100, cMB/100, cMB % 100); |
42e641ab | 820 | } else if (bytes < 2*1024LL*1024LL*1024LL*1024LL) { |
93d3bd3b | 821 | long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2; |
d5d3721e | 822 | long cGB = (bytes / (1000000000LL/200LL ) +1) /2; |
8f23b0b3 | 823 | snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)", |
efa29530 | 824 | cGiB/100, cGiB % 100, cGB/100, cGB % 100); |
42e641ab KT |
825 | } else { |
826 | long cTiB = (bytes * 200LL / (1LL<<40) + 1) / 2; | |
827 | long cTB = (bytes / (1000000000000LL / 200LL) + 1) / 2; | |
828 | snprintf(buf, sizeof(buf), " (%ld.%02ld TiB %ld.%02ld TB)", | |
829 | cTiB/100, cTiB % 100, cTB/100, cTB % 100); | |
d5d3721e | 830 | } |
cd29a5c8 NB |
831 | return buf; |
832 | } | |
e0d19036 | 833 | |
f0ec6710 | 834 | char *human_size_brief(long long bytes, int prefix) |
e0d19036 NB |
835 | { |
836 | static char buf[30]; | |
e0d19036 | 837 | |
42e641ab KT |
838 | /* We convert bytes to either centi-M{ega,ibi}bytes, |
839 | * centi-G{igi,ibi}bytes or centi-T{era,ebi}bytes | |
840 | * with appropriate rounding, and then print | |
841 | * 1/100th of those as a decimal. | |
570abc6f | 842 | * We allow upto 2048Megabytes before converting to |
42e641ab KT |
843 | * gigabytes and 2048Gigabytes before converting to |
844 | * terabytes, as that shows more precision and isn't | |
570abc6f | 845 | * too large a number. |
f0ec6710 MN |
846 | * |
847 | * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc. | |
848 | * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc. | |
570abc6f MN |
849 | */ |
850 | ||
e0d19036 | 851 | if (bytes < 5000*1024) |
570abc6f | 852 | buf[0] = 0; |
f0ec6710 MN |
853 | else if (prefix == IEC) { |
854 | if (bytes < 2*1024LL*1024LL*1024LL) { | |
93d3bd3b | 855 | long cMiB = (bytes * 200LL / (1LL<<20) +1) /2; |
f0ec6710 | 856 | snprintf(buf, sizeof(buf), "%ld.%02ldMiB", |
efa29530 | 857 | cMiB/100, cMiB % 100); |
42e641ab | 858 | } else if (bytes < 2*1024LL*1024LL*1024LL*1024LL) { |
93d3bd3b | 859 | long cGiB = (bytes * 200LL / (1LL<<30) +1) /2; |
f0ec6710 | 860 | snprintf(buf, sizeof(buf), "%ld.%02ldGiB", |
efa29530 | 861 | cGiB/100, cGiB % 100); |
42e641ab KT |
862 | } else { |
863 | long cTiB = (bytes * 200LL / (1LL<<40) + 1) / 2; | |
864 | snprintf(buf, sizeof(buf), "%ld.%02ldTiB", | |
865 | cTiB/100, cTiB % 100); | |
f0ec6710 MN |
866 | } |
867 | } | |
868 | else if (prefix == JEDEC) { | |
869 | if (bytes < 2*1024LL*1024LL*1024LL) { | |
870 | long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2; | |
871 | snprintf(buf, sizeof(buf), "%ld.%02ldMB", | |
efa29530 | 872 | cMB/100, cMB % 100); |
42e641ab | 873 | } else if (bytes < 2*1024LL*1024LL*1024LL*1024LL) { |
f0ec6710 MN |
874 | long cGB = (bytes / (1000000000LL/200LL ) +1) /2; |
875 | snprintf(buf, sizeof(buf), "%ld.%02ldGB", | |
efa29530 | 876 | cGB/100, cGB % 100); |
42e641ab KT |
877 | } else { |
878 | long cTB = (bytes / (1000000000000LL / 200LL) + 1) / 2; | |
879 | snprintf(buf, sizeof(buf), "%ld.%02ldTB", | |
880 | cTB/100, cTB % 100); | |
f0ec6710 | 881 | } |
570abc6f | 882 | } |
f0ec6710 MN |
883 | else |
884 | buf[0] = 0; | |
570abc6f | 885 | |
e0d19036 NB |
886 | return buf; |
887 | } | |
e4965ef8 N |
888 | |
889 | void print_r10_layout(int layout) | |
890 | { | |
891 | int near = layout & 255; | |
892 | int far = (layout >> 8) & 255; | |
893 | int offset = (layout&0x10000); | |
894 | char *sep = ""; | |
895 | ||
896 | if (near != 1) { | |
897 | printf("%s near=%d", sep, near); | |
898 | sep = ","; | |
899 | } | |
900 | if (far != 1) | |
901 | printf("%s %s=%d", sep, offset?"offset":"far", far); | |
902 | if (near*far == 1) | |
903 | printf("NO REDUNDANCY"); | |
904 | } | |
e0d19036 | 905 | |
5f8097be NB |
906 | unsigned long long calc_array_size(int level, int raid_disks, int layout, |
907 | int chunksize, unsigned long long devsize) | |
577e8448 | 908 | { |
e03640bd CA |
909 | if (level == 1) |
910 | return devsize; | |
577e8448 AK |
911 | devsize &= ~(unsigned long long)((chunksize>>9)-1); |
912 | return get_data_disks(level, layout, raid_disks) * devsize; | |
913 | } | |
914 | ||
915 | int get_data_disks(int level, int layout, int raid_disks) | |
5f8097be NB |
916 | { |
917 | int data_disks = 0; | |
918 | switch (level) { | |
ca3b6696 N |
919 | case 0: data_disks = raid_disks; |
920 | break; | |
921 | case 1: data_disks = 1; | |
922 | break; | |
5f8097be | 923 | case 4: |
ca3b6696 N |
924 | case 5: data_disks = raid_disks - 1; |
925 | break; | |
926 | case 6: data_disks = raid_disks - 2; | |
927 | break; | |
5f8097be NB |
928 | case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255); |
929 | break; | |
930 | } | |
577e8448 AK |
931 | |
932 | return data_disks; | |
5f8097be NB |
933 | } |
934 | ||
13db17bd | 935 | dev_t devnm2devid(char *devnm) |
4dd2df09 N |
936 | { |
937 | /* First look in /sys/block/$DEVNM/dev for %d:%d | |
938 | * If that fails, try parsing out a number | |
939 | */ | |
feeb2785 | 940 | char path[PATH_MAX]; |
4dd2df09 N |
941 | char *ep; |
942 | int fd; | |
943 | int mjr,mnr; | |
944 | ||
feeb2785 | 945 | snprintf(path, sizeof(path), "/sys/block/%s/dev", devnm); |
4dd2df09 N |
946 | fd = open(path, O_RDONLY); |
947 | if (fd >= 0) { | |
948 | char buf[20]; | |
949 | int n = read(fd, buf, sizeof(buf)); | |
950 | close(fd); | |
951 | if (n > 0) | |
952 | buf[n] = 0; | |
953 | if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2) | |
954 | return makedev(mjr, mnr); | |
955 | } | |
956 | if (strncmp(devnm, "md_d", 4) == 0 && | |
957 | isdigit(devnm[4]) && | |
958 | (mnr = strtoul(devnm+4, &ep, 10)) >= 0 && | |
959 | ep > devnm && *ep == 0) | |
960 | return makedev(get_mdp_major(), mnr << MdpMinorShift); | |
961 | ||
962 | if (strncmp(devnm, "md", 2) == 0 && | |
963 | isdigit(devnm[2]) && | |
964 | (mnr = strtoul(devnm+2, &ep, 10)) >= 0 && | |
965 | ep > devnm && *ep == 0) | |
966 | return makedev(MD_MAJOR, mnr); | |
967 | ||
968 | return 0; | |
969 | } | |
970 | ||
b938519e MK |
971 | /** |
972 | * get_md_name() - Get main dev node of the md device. | |
973 | * @devnm: Md device name or path. | |
974 | * | |
975 | * Function checks if the full name was passed and returns md name | |
976 | * if it is the MD device. | |
977 | * | |
978 | * Return: Main dev node of the md device or NULL if not found. | |
979 | */ | |
4dd2df09 | 980 | char *get_md_name(char *devnm) |
e0d19036 | 981 | { |
b938519e | 982 | static char devname[NAME_MAX]; |
e0d19036 | 983 | struct stat stb; |
98c6faba | 984 | |
b938519e MK |
985 | if (strncmp(devnm, "/dev/", 5) == 0) |
986 | snprintf(devname, sizeof(devname), "%s", devnm); | |
987 | else | |
988 | snprintf(devname, sizeof(devname), "/dev/%s", devnm); | |
4dd2df09 | 989 | |
b938519e MK |
990 | if (!is_mddev(devname)) |
991 | return NULL; | |
992 | if (stat(devname, &stb) == 0 && (S_IFMT&stb.st_mode) == S_IFBLK) | |
4dd2df09 N |
993 | return devname; |
994 | ||
e0d19036 NB |
995 | return NULL; |
996 | } | |
997 | ||
998 | void put_md_name(char *name) | |
999 | { | |
ca3b6696 | 1000 | if (strncmp(name, "/dev/.tmp.md", 12) == 0) |
e0d19036 NB |
1001 | unlink(name); |
1002 | } | |
f9ce90ba | 1003 | |
cc742d38 N |
1004 | int get_maj_min(char *dev, int *major, int *minor) |
1005 | { | |
1006 | char *e; | |
1007 | *major = strtoul(dev, &e, 0); | |
1008 | return (e > dev && *e == ':' && e[1] && | |
1009 | (*minor = strtoul(e+1, &e, 0)) >= 0 && | |
1010 | *e == 0); | |
1011 | } | |
1012 | ||
7211116c MK |
1013 | /** |
1014 | * is_bit_set() - get bit value by index. | |
1015 | * @val: value. | |
1016 | * @index: index of the bit (LSB numbering). | |
1017 | * | |
1018 | * Return: bit value. | |
1019 | */ | |
1020 | bool is_bit_set(int *val, unsigned char index) | |
1021 | { | |
1022 | if ((*val) & (1 << index)) | |
1023 | return true; | |
1024 | return false; | |
1025 | } | |
1026 | ||
8b0dabea NB |
1027 | int dev_open(char *dev, int flags) |
1028 | { | |
1029 | /* like 'open', but if 'dev' matches %d:%d, create a temp | |
1030 | * block device and open that | |
1031 | */ | |
8b0dabea NB |
1032 | int fd = -1; |
1033 | char devname[32]; | |
e81cdd9f | 1034 | int major; |
8b0dabea | 1035 | int minor; |
e81cdd9f | 1036 | |
9f0ad56b JS |
1037 | if (!dev) |
1038 | return -1; | |
6df6a774 | 1039 | flags |= O_DIRECT; |
e81cdd9f | 1040 | |
cc742d38 | 1041 | if (get_maj_min(dev, &major, &minor)) { |
9dc7d357 N |
1042 | snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d", |
1043 | (int)getpid(), major, minor); | |
1044 | if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) { | |
1045 | fd = open(devname, flags); | |
1046 | unlink(devname); | |
6df6a774 N |
1047 | } |
1048 | if (fd < 0) { | |
9dc7d357 | 1049 | /* Try /tmp as /dev appear to be read-only */ |
b7a462e5 JS |
1050 | snprintf(devname, sizeof(devname), |
1051 | "/tmp/.tmp.md.%d:%d:%d", | |
6df6a774 | 1052 | (int)getpid(), major, minor); |
b7a462e5 JS |
1053 | if (mknod(devname, S_IFBLK|0600, |
1054 | makedev(major, minor)) == 0) { | |
6df6a774 N |
1055 | fd = open(devname, flags); |
1056 | unlink(devname); | |
1057 | } | |
8b0dabea NB |
1058 | } |
1059 | } else | |
6df6a774 | 1060 | fd = open(dev, flags); |
8b0dabea NB |
1061 | return fd; |
1062 | } | |
f9ce90ba | 1063 | |
4dd2df09 | 1064 | int open_dev_flags(char *devnm, int flags) |
e8a70c89 | 1065 | { |
13db17bd | 1066 | dev_t devid; |
e8a70c89 N |
1067 | char buf[20]; |
1068 | ||
4dd2df09 N |
1069 | devid = devnm2devid(devnm); |
1070 | sprintf(buf, "%d:%d", major(devid), minor(devid)); | |
d998b738 N |
1071 | return dev_open(buf, flags); |
1072 | } | |
1073 | ||
4dd2df09 | 1074 | int open_dev(char *devnm) |
d998b738 | 1075 | { |
4dd2df09 | 1076 | return open_dev_flags(devnm, O_RDONLY); |
e8a70c89 N |
1077 | } |
1078 | ||
4dd2df09 | 1079 | int open_dev_excl(char *devnm) |
a931db9e NB |
1080 | { |
1081 | char buf[20]; | |
1082 | int i; | |
7187750e | 1083 | int flags = O_RDWR; |
13db17bd | 1084 | dev_t devid = devnm2devid(devnm); |
239b3cc0 | 1085 | unsigned int delay = 1; // miliseconds |
a931db9e | 1086 | |
4dd2df09 | 1087 | sprintf(buf, "%d:%d", major(devid), minor(devid)); |
efa29530 | 1088 | for (i = 0; i < 25; i++) { |
7187750e | 1089 | int fd = dev_open(buf, flags|O_EXCL); |
a931db9e NB |
1090 | if (fd >= 0) |
1091 | return fd; | |
7187750e N |
1092 | if (errno == EACCES && flags == O_RDWR) { |
1093 | flags = O_RDONLY; | |
1094 | continue; | |
1095 | } | |
a931db9e NB |
1096 | if (errno != EBUSY) |
1097 | return fd; | |
239b3cc0 MG |
1098 | sleep_for(0, MSEC_TO_NSEC(delay), true); |
1099 | if (delay < 200) | |
399e0b97 | 1100 | delay *= 2; |
a931db9e NB |
1101 | } |
1102 | return -1; | |
1103 | } | |
1104 | ||
9008ed1c N |
1105 | int same_dev(char *one, char *two) |
1106 | { | |
1107 | struct stat st1, st2; | |
1108 | if (stat(one, &st1) != 0) | |
1109 | return 0; | |
1110 | if (stat(two, &st2) != 0) | |
1111 | return 0; | |
1112 | if ((st1.st_mode & S_IFMT) != S_IFBLK) | |
1113 | return 0; | |
1114 | if ((st2.st_mode & S_IFMT) != S_IFBLK) | |
1115 | return 0; | |
1116 | return st1.st_rdev == st2.st_rdev; | |
1117 | } | |
1118 | ||
a7c6e3fb | 1119 | void wait_for(char *dev, int fd) |
a714580e N |
1120 | { |
1121 | int i; | |
a7c6e3fb | 1122 | struct stat stb_want; |
239b3cc0 | 1123 | unsigned int delay = 1; // miliseconds |
a7c6e3fb N |
1124 | |
1125 | if (fstat(fd, &stb_want) != 0 || | |
1126 | (stb_want.st_mode & S_IFMT) != S_IFBLK) | |
1127 | return; | |
a714580e | 1128 | |
efa29530 | 1129 | for (i = 0; i < 25; i++) { |
a714580e | 1130 | struct stat stb; |
a7c6e3fb N |
1131 | if (stat(dev, &stb) == 0 && |
1132 | (stb.st_mode & S_IFMT) == S_IFBLK && | |
1133 | (stb.st_rdev == stb_want.st_rdev)) | |
a714580e | 1134 | return; |
239b3cc0 MG |
1135 | sleep_for(0, MSEC_TO_NSEC(delay), true); |
1136 | if (delay < 200) | |
399e0b97 | 1137 | delay *= 2; |
a714580e | 1138 | } |
436305c6 | 1139 | if (i == 25) |
9b8fea91 | 1140 | pr_err("timeout waiting for %s\n", dev); |
a714580e N |
1141 | } |
1142 | ||
0f22b998 N |
1143 | struct superswitch *superlist[] = |
1144 | { | |
1145 | &super0, &super1, | |
1146 | &super_ddf, &super_imsm, | |
0592faeb | 1147 | &mbr, &gpt, |
32141c17 JS |
1148 | NULL |
1149 | }; | |
f7dd881f | 1150 | |
4725bc31 | 1151 | struct supertype *super_by_fd(int fd, char **subarrayp) |
f9ce90ba | 1152 | { |
1686dc25 NB |
1153 | mdu_array_info_t array; |
1154 | int vers; | |
1155 | int minor; | |
1156 | struct supertype *st = NULL; | |
7e0f6979 | 1157 | struct mdinfo *sra; |
142cb9e1 | 1158 | char *verstr; |
1686dc25 NB |
1159 | char version[20]; |
1160 | int i; | |
f7e7067b | 1161 | char *subarray = NULL; |
4dd2df09 | 1162 | char container[32] = ""; |
f1f3ef7d LXK |
1163 | char *devnm = NULL; |
1164 | ||
1165 | devnm = fd2devnm(fd); | |
1166 | if (!devnm) | |
1167 | return NULL; | |
1686dc25 | 1168 | |
4dd2df09 | 1169 | sra = sysfs_read(fd, NULL, GET_VERSION); |
1686dc25 NB |
1170 | |
1171 | if (sra) { | |
7e0f6979 NB |
1172 | vers = sra->array.major_version; |
1173 | minor = sra->array.minor_version; | |
142cb9e1 | 1174 | verstr = sra->text_version; |
1686dc25 | 1175 | } else { |
9cd39f01 | 1176 | if (md_get_array_info(fd, &array)) |
1686dc25 NB |
1177 | array.major_version = array.minor_version = 0; |
1178 | vers = array.major_version; | |
1179 | minor = array.minor_version; | |
142cb9e1 | 1180 | verstr = ""; |
6fbba4c9 | 1181 | } |
82d9eba6 | 1182 | |
1686dc25 NB |
1183 | if (vers != -1) { |
1184 | sprintf(version, "%d.%d", vers, minor); | |
1185 | verstr = version; | |
6fbba4c9 | 1186 | } |
3c558363 | 1187 | if (minor == -2 && is_subarray(verstr)) { |
f7e7067b | 1188 | char *dev = verstr+1; |
5f7e44b2 | 1189 | |
f7e7067b | 1190 | subarray = strchr(dev, '/'); |
92d49ecf | 1191 | if (subarray) { |
f7e7067b | 1192 | *subarray++ = '\0'; |
503975b9 | 1193 | subarray = xstrdup(subarray); |
92d49ecf | 1194 | } |
4dd2df09 | 1195 | strcpy(container, dev); |
15d230f7 | 1196 | sysfs_free(sra); |
5f7e44b2 | 1197 | sra = sysfs_read(-1, container, GET_VERSION); |
603f24a0 N |
1198 | if (sra && sra->text_version[0]) |
1199 | verstr = sra->text_version; | |
1200 | else | |
1201 | verstr = "-no-metadata-"; | |
f7e7067b NB |
1202 | } |
1203 | ||
efa29530 | 1204 | for (i = 0; st == NULL && superlist[i]; i++) |
f7e7067b | 1205 | st = superlist[i]->match_metadata_desc(verstr); |
1686dc25 | 1206 | |
15d230f7 | 1207 | sysfs_free(sra); |
f7e7067b | 1208 | if (st) { |
3b0896f8 | 1209 | st->sb = NULL; |
1f49fb3a N |
1210 | if (subarrayp) |
1211 | *subarrayp = subarray; | |
4dd2df09 | 1212 | strcpy(st->container_devnm, container); |
f1f3ef7d | 1213 | strncpy(st->devnm, devnm, MD_NAME_MAX - 1); |
4725bc31 N |
1214 | } else |
1215 | free(subarray); | |
5f7e44b2 | 1216 | |
82d9eba6 | 1217 | return st; |
f9ce90ba | 1218 | } |
ea24acd0 | 1219 | |
0f0749ad | 1220 | int dev_size_from_id(dev_t id, unsigned long long *size) |
52d5d101 AC |
1221 | { |
1222 | char buf[20]; | |
1223 | int fd; | |
1224 | ||
1225 | sprintf(buf, "%d:%d", major(id), minor(id)); | |
1226 | fd = dev_open(buf, O_RDONLY); | |
1227 | if (fd < 0) | |
1228 | return 0; | |
1229 | if (get_dev_size(fd, NULL, size)) { | |
1230 | close(fd); | |
1231 | return 1; | |
1232 | } | |
1233 | close(fd); | |
1234 | return 0; | |
1235 | } | |
f9ce90ba | 1236 | |
4b57ecf6 AO |
1237 | int dev_sector_size_from_id(dev_t id, unsigned int *size) |
1238 | { | |
1239 | char buf[20]; | |
1240 | int fd; | |
1241 | ||
1242 | sprintf(buf, "%d:%d", major(id), minor(id)); | |
1243 | fd = dev_open(buf, O_RDONLY); | |
1244 | if (fd < 0) | |
1245 | return 0; | |
1246 | if (get_dev_sector_size(fd, NULL, size)) { | |
1247 | close(fd); | |
1248 | return 1; | |
1249 | } | |
1250 | close(fd); | |
1251 | return 0; | |
1252 | } | |
1253 | ||
159c3a1a | 1254 | struct supertype *dup_super(struct supertype *orig) |
3da92f27 | 1255 | { |
159c3a1a | 1256 | struct supertype *st; |
1686dc25 | 1257 | |
d2ca6449 NB |
1258 | if (!orig) |
1259 | return orig; | |
503975b9 | 1260 | st = xcalloc(1, sizeof(*st)); |
159c3a1a NB |
1261 | st->ss = orig->ss; |
1262 | st->max_devs = orig->max_devs; | |
1263 | st->minor_version = orig->minor_version; | |
cb8f6859 | 1264 | st->ignore_hw_compat = orig->ignore_hw_compat; |
23bf42cc | 1265 | st->data_offset = orig->data_offset; |
159c3a1a NB |
1266 | st->sb = NULL; |
1267 | st->info = NULL; | |
1268 | return st; | |
3da92f27 NB |
1269 | } |
1270 | ||
54887ad8 | 1271 | struct supertype *guess_super_type(int fd, enum guess_types guess_type) |
f9ce90ba NB |
1272 | { |
1273 | /* try each load_super to find the best match, | |
1274 | * and return the best superswitch | |
1275 | */ | |
82d9eba6 NB |
1276 | struct superswitch *ss; |
1277 | struct supertype *st; | |
1d13b599 | 1278 | unsigned int besttime = 0; |
570c0542 | 1279 | int bestsuper = -1; |
f9ce90ba NB |
1280 | int i; |
1281 | ||
503975b9 | 1282 | st = xcalloc(1, sizeof(*st)); |
4dd2df09 | 1283 | st->container_devnm[0] = 0; |
d1d599ea | 1284 | |
efa29530 | 1285 | for (i = 0; superlist[i]; i++) { |
f9ce90ba NB |
1286 | int rv; |
1287 | ss = superlist[i]; | |
54887ad8 N |
1288 | if (guess_type == guess_array && ss->add_to_super == NULL) |
1289 | continue; | |
1290 | if (guess_type == guess_partitions && ss->add_to_super != NULL) | |
1291 | continue; | |
ef609477 | 1292 | memset(st, 0, sizeof(*st)); |
df3346e6 | 1293 | st->ignore_hw_compat = 1; |
3da92f27 | 1294 | rv = ss->load_super(st, fd, NULL); |
570c0542 NB |
1295 | if (rv == 0) { |
1296 | struct mdinfo info; | |
a5d85af7 | 1297 | st->ss->getinfo_super(st, &info, NULL); |
570c0542 NB |
1298 | if (bestsuper == -1 || |
1299 | besttime < info.array.ctime) { | |
1300 | bestsuper = i; | |
1301 | besttime = info.array.ctime; | |
570c0542 | 1302 | } |
3da92f27 | 1303 | ss->free_super(st); |
570c0542 NB |
1304 | } |
1305 | } | |
1306 | if (bestsuper != -1) { | |
1307 | int rv; | |
ef609477 | 1308 | memset(st, 0, sizeof(*st)); |
df3346e6 | 1309 | st->ignore_hw_compat = 1; |
3da92f27 | 1310 | rv = superlist[bestsuper]->load_super(st, fd, NULL); |
f9ce90ba | 1311 | if (rv == 0) { |
5e747af2 | 1312 | superlist[bestsuper]->free_super(st); |
82d9eba6 | 1313 | return st; |
f9ce90ba NB |
1314 | } |
1315 | } | |
570c0542 | 1316 | free(st); |
f9ce90ba NB |
1317 | return NULL; |
1318 | } | |
fe6729fa | 1319 | |
beae1dfe NB |
1320 | /* Return size of device in bytes */ |
1321 | int get_dev_size(int fd, char *dname, unsigned long long *sizep) | |
1322 | { | |
1323 | unsigned long long ldsize; | |
c2c9bb6f NB |
1324 | struct stat st; |
1325 | ||
1326 | if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode)) | |
1327 | ldsize = (unsigned long long)st.st_size; | |
1328 | else | |
beae1dfe NB |
1329 | #ifdef BLKGETSIZE64 |
1330 | if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0) | |
1331 | #endif | |
1332 | { | |
1333 | unsigned long dsize; | |
1334 | if (ioctl(fd, BLKGETSIZE, &dsize) == 0) { | |
1335 | ldsize = dsize; | |
1336 | ldsize <<= 9; | |
1337 | } else { | |
1338 | if (dname) | |
32971509 | 1339 | pr_err("Cannot get size of %s: %s\n", |
beae1dfe NB |
1340 | dname, strerror(errno)); |
1341 | return 0; | |
1342 | } | |
1343 | } | |
1344 | *sizep = ldsize; | |
1345 | return 1; | |
1346 | } | |
8fac0577 | 1347 | |
32971509 PB |
1348 | /* Return sector size of device in bytes */ |
1349 | int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep) | |
1350 | { | |
1351 | unsigned int sectsize; | |
1352 | ||
1353 | if (ioctl(fd, BLKSSZGET, §size) != 0) { | |
1354 | if (dname) | |
1355 | pr_err("Cannot get sector size of %s: %s\n", | |
1356 | dname, strerror(errno)); | |
1357 | return 0; | |
1358 | } | |
1359 | ||
1360 | *sectsizep = sectsize; | |
1361 | return 1; | |
1362 | } | |
1363 | ||
3a371610 N |
1364 | /* Return true if this can only be a container, not a member device. |
1365 | * i.e. is and md device and size is zero | |
1366 | */ | |
1367 | int must_be_container(int fd) | |
1368 | { | |
f5c924f4 | 1369 | struct mdinfo *mdi; |
3a371610 | 1370 | unsigned long long size; |
f5c924f4 JS |
1371 | |
1372 | mdi = sysfs_read(fd, NULL, GET_VERSION); | |
1373 | if (!mdi) | |
3a371610 | 1374 | return 0; |
f5c924f4 JS |
1375 | sysfs_free(mdi); |
1376 | ||
3a371610 N |
1377 | if (get_dev_size(fd, NULL, &size) == 0) |
1378 | return 1; | |
1379 | if (size == 0) | |
1380 | return 1; | |
1381 | return 0; | |
1382 | } | |
034b203a TM |
1383 | |
1384 | /* Sets endofpart parameter to the last block used by the last GPT partition on the device. | |
1385 | * Returns: 1 if successful | |
1386 | * -1 for unknown partition type | |
1387 | * 0 for other errors | |
1388 | */ | |
1389 | static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart) | |
1390 | { | |
056b331e | 1391 | struct GPT gpt; |
034b203a | 1392 | unsigned char empty_gpt_entry[16]= {0}; |
db7fdfe4 JS |
1393 | struct GPT_part_entry *part; |
1394 | char buf[512]; | |
034b203a TM |
1395 | unsigned long long curr_part_end; |
1396 | unsigned all_partitions, entry_size; | |
f21e18ca | 1397 | unsigned part_nr; |
41b06495 | 1398 | unsigned int sector_size = 0; |
034b203a TM |
1399 | |
1400 | *endofpart = 0; | |
1401 | ||
056b331e | 1402 | BUILD_BUG_ON(sizeof(gpt) != 512); |
73e658d8 | 1403 | /* skip protective MBR */ |
41b06495 MD |
1404 | if (!get_dev_sector_size(fd, NULL, §or_size)) |
1405 | return 0; | |
1406 | lseek(fd, sector_size, SEEK_SET); | |
73e658d8 | 1407 | /* read GPT header */ |
056b331e | 1408 | if (read(fd, &gpt, 512) != 512) |
034b203a TM |
1409 | return 0; |
1410 | ||
1411 | /* get the number of partition entries and the entry size */ | |
056b331e N |
1412 | all_partitions = __le32_to_cpu(gpt.part_cnt); |
1413 | entry_size = __le32_to_cpu(gpt.part_size); | |
034b203a TM |
1414 | |
1415 | /* Check GPT signature*/ | |
056b331e | 1416 | if (gpt.magic != GPT_SIGNATURE_MAGIC) |
034b203a TM |
1417 | return -1; |
1418 | ||
1419 | /* sanity checks */ | |
1420 | if (all_partitions > 1024 || | |
db7fdfe4 | 1421 | entry_size > sizeof(buf)) |
034b203a TM |
1422 | return -1; |
1423 | ||
db7fdfe4 JS |
1424 | part = (struct GPT_part_entry *)buf; |
1425 | ||
41b06495 MD |
1426 | /* set offset to third block (GPT entries) */ |
1427 | lseek(fd, sector_size*2, SEEK_SET); | |
ca3b6696 | 1428 | for (part_nr = 0; part_nr < all_partitions; part_nr++) { |
73e658d8 | 1429 | /* read partition entry */ |
db7fdfe4 | 1430 | if (read(fd, buf, entry_size) != (ssize_t)entry_size) |
73e658d8 LB |
1431 | return 0; |
1432 | ||
034b203a | 1433 | /* is this valid partition? */ |
db7fdfe4 | 1434 | if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) { |
034b203a | 1435 | /* check the last lba for the current partition */ |
db7fdfe4 | 1436 | curr_part_end = __le64_to_cpu(part->ending_lba); |
034b203a TM |
1437 | if (curr_part_end > *endofpart) |
1438 | *endofpart = curr_part_end; | |
1439 | } | |
1440 | ||
034b203a TM |
1441 | } |
1442 | return 1; | |
1443 | } | |
1444 | ||
1445 | /* Sets endofpart parameter to the last block used by the last partition on the device. | |
1446 | * Returns: 1 if successful | |
1447 | * -1 for unknown partition type | |
1448 | * 0 for other errors | |
1449 | */ | |
1450 | static int get_last_partition_end(int fd, unsigned long long *endofpart) | |
1451 | { | |
056b331e | 1452 | struct MBR boot_sect; |
034b203a | 1453 | unsigned long long curr_part_end; |
f21e18ca | 1454 | unsigned part_nr; |
31208db9 | 1455 | unsigned int sector_size; |
034b203a TM |
1456 | int retval = 0; |
1457 | ||
1458 | *endofpart = 0; | |
1459 | ||
056b331e | 1460 | BUILD_BUG_ON(sizeof(boot_sect) != 512); |
034b203a TM |
1461 | /* read MBR */ |
1462 | lseek(fd, 0, 0); | |
056b331e | 1463 | if (read(fd, &boot_sect, 512) != 512) |
034b203a TM |
1464 | goto abort; |
1465 | ||
1466 | /* check MBP signature */ | |
056b331e | 1467 | if (boot_sect.magic == MBR_SIGNATURE_MAGIC) { |
034b203a TM |
1468 | retval = 1; |
1469 | /* found the correct signature */ | |
034b203a | 1470 | |
ca3b6696 | 1471 | for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) { |
8e2bca51 JC |
1472 | /* |
1473 | * Have to make every access through boot_sect rather | |
1474 | * than using a pointer to the partition table (or an | |
1475 | * entry), since the entries are not properly aligned. | |
1476 | */ | |
1477 | ||
034b203a | 1478 | /* check for GPT type */ |
8e2bca51 JC |
1479 | if (boot_sect.parts[part_nr].part_type == |
1480 | MBR_GPT_PARTITION_TYPE) { | |
034b203a TM |
1481 | retval = get_gpt_last_partition_end(fd, endofpart); |
1482 | break; | |
1483 | } | |
1484 | /* check the last used lba for the current partition */ | |
8e2bca51 JC |
1485 | curr_part_end = |
1486 | __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) + | |
1487 | __le32_to_cpu(boot_sect.parts[part_nr].blocks_num); | |
034b203a TM |
1488 | if (curr_part_end > *endofpart) |
1489 | *endofpart = curr_part_end; | |
034b203a TM |
1490 | } |
1491 | } else { | |
1492 | /* Unknown partition table */ | |
1493 | retval = -1; | |
1494 | } | |
31208db9 MD |
1495 | /* calculate number of 512-byte blocks */ |
1496 | if (get_dev_sector_size(fd, NULL, §or_size)) | |
1497 | *endofpart *= (sector_size / 512); | |
034b203a TM |
1498 | abort: |
1499 | return retval; | |
1500 | } | |
1501 | ||
53ed6ac3 KW |
1502 | int check_partitions(int fd, char *dname, unsigned long long freesize, |
1503 | unsigned long long size) | |
034b203a TM |
1504 | { |
1505 | /* | |
1506 | * Check where the last partition ends | |
1507 | */ | |
1508 | unsigned long long endofpart; | |
034b203a | 1509 | |
31208db9 | 1510 | if (get_last_partition_end(fd, &endofpart) > 0) { |
034b203a TM |
1511 | /* There appears to be a partition table here */ |
1512 | if (freesize == 0) { | |
1513 | /* partitions will not be visible in new device */ | |
e7b84f9d N |
1514 | pr_err("partition table exists on %s but will be lost or\n" |
1515 | " meaningless after creating array\n", | |
1516 | dname); | |
034b203a TM |
1517 | return 1; |
1518 | } else if (endofpart > freesize) { | |
1519 | /* last partition overlaps metadata */ | |
e7b84f9d N |
1520 | pr_err("metadata will over-write last partition on %s.\n", |
1521 | dname); | |
034b203a | 1522 | return 1; |
53ed6ac3 KW |
1523 | } else if (size && endofpart > size) { |
1524 | /* partitions will be truncated in new device */ | |
e7b84f9d N |
1525 | pr_err("array size is too small to cover all partitions on %s.\n", |
1526 | dname); | |
53ed6ac3 | 1527 | return 1; |
034b203a TM |
1528 | } |
1529 | } | |
1530 | return 0; | |
1531 | } | |
1532 | ||
a322f70c DW |
1533 | int open_container(int fd) |
1534 | { | |
1535 | /* 'fd' is a block device. Find out if it is in use | |
1536 | * by a container, and return an open fd on that container. | |
1537 | */ | |
ccd61ebf | 1538 | char path[288]; |
a322f70c DW |
1539 | char *e; |
1540 | DIR *dir; | |
1541 | struct dirent *de; | |
1542 | int dfd, n; | |
1543 | char buf[200]; | |
1544 | int major, minor; | |
1545 | struct stat st; | |
1546 | ||
1547 | if (fstat(fd, &st) != 0) | |
1548 | return -1; | |
1549 | sprintf(path, "/sys/dev/block/%d:%d/holders", | |
1550 | (int)major(st.st_rdev), (int)minor(st.st_rdev)); | |
1551 | e = path + strlen(path); | |
1552 | ||
1553 | dir = opendir(path); | |
1554 | if (!dir) | |
1555 | return -1; | |
1556 | while ((de = readdir(dir))) { | |
1557 | if (de->d_ino == 0) | |
1558 | continue; | |
1559 | if (de->d_name[0] == '.') | |
1560 | continue; | |
a994592d N |
1561 | /* Need to make sure it is a container and not a volume */ |
1562 | sprintf(e, "/%s/md/metadata_version", de->d_name); | |
1563 | dfd = open(path, O_RDONLY); | |
1564 | if (dfd < 0) | |
1565 | continue; | |
1566 | n = read(dfd, buf, sizeof(buf)); | |
1567 | close(dfd); | |
1568 | if (n <= 0 || (unsigned)n >= sizeof(buf)) | |
1569 | continue; | |
1570 | buf[n] = 0; | |
1571 | if (strncmp(buf, "external", 8) != 0 || | |
1572 | n < 10 || | |
1573 | buf[9] == '/') | |
1574 | continue; | |
a322f70c DW |
1575 | sprintf(e, "/%s/dev", de->d_name); |
1576 | dfd = open(path, O_RDONLY); | |
1577 | if (dfd < 0) | |
1578 | continue; | |
1579 | n = read(dfd, buf, sizeof(buf)); | |
1580 | close(dfd); | |
f21e18ca | 1581 | if (n <= 0 || (unsigned)n >= sizeof(buf)) |
a322f70c DW |
1582 | continue; |
1583 | buf[n] = 0; | |
1584 | if (sscanf(buf, "%d:%d", &major, &minor) != 2) | |
1585 | continue; | |
1586 | sprintf(buf, "%d:%d", major, minor); | |
1587 | dfd = dev_open(buf, O_RDONLY); | |
1588 | if (dfd >= 0) { | |
1589 | closedir(dir); | |
1590 | return dfd; | |
1591 | } | |
1592 | } | |
355726fa | 1593 | closedir(dir); |
a322f70c DW |
1594 | return -1; |
1595 | } | |
1596 | ||
33414a01 DW |
1597 | struct superswitch *version_to_superswitch(char *vers) |
1598 | { | |
1599 | int i; | |
1600 | ||
1601 | for (i = 0; superlist[i]; i++) { | |
1602 | struct superswitch *ss = superlist[i]; | |
1603 | ||
1604 | if (strcmp(vers, ss->name) == 0) | |
1605 | return ss; | |
1606 | } | |
1607 | ||
1608 | return NULL; | |
1609 | } | |
1610 | ||
4dd2df09 N |
1611 | int metadata_container_matches(char *metadata, char *devnm) |
1612 | { | |
1613 | /* Check if 'devnm' is the container named in 'metadata' | |
1614 | * which is | |
1615 | * /containername/componentname or | |
1616 | * -containername/componentname | |
1617 | */ | |
1618 | int l; | |
1619 | if (*metadata != '/' && *metadata != '-') | |
1620 | return 0; | |
1621 | l = strlen(devnm); | |
1622 | if (strncmp(metadata+1, devnm, l) != 0) | |
1623 | return 0; | |
1624 | if (metadata[l+1] != '/') | |
1625 | return 0; | |
1626 | return 1; | |
1627 | } | |
1628 | ||
1629 | int metadata_subdev_matches(char *metadata, char *devnm) | |
1630 | { | |
1631 | /* Check if 'devnm' is the subdev named in 'metadata' | |
1632 | * which is | |
1633 | * /containername/subdev or | |
1634 | * -containername/subdev | |
1635 | */ | |
1636 | char *sl; | |
1637 | if (*metadata != '/' && *metadata != '-') | |
1638 | return 0; | |
1639 | sl = strchr(metadata+1, '/'); | |
1640 | if (!sl) | |
1641 | return 0; | |
1642 | if (strcmp(sl+1, devnm) == 0) | |
1643 | return 1; | |
1644 | return 0; | |
1645 | } | |
1646 | ||
33414a01 DW |
1647 | int is_container_member(struct mdstat_ent *mdstat, char *container) |
1648 | { | |
1649 | if (mdstat->metadata_version == NULL || | |
1650 | strncmp(mdstat->metadata_version, "external:", 9) != 0 || | |
4dd2df09 | 1651 | !metadata_container_matches(mdstat->metadata_version+9, container)) |
33414a01 DW |
1652 | return 0; |
1653 | ||
1654 | return 1; | |
1655 | } | |
1656 | ||
1657 | int is_subarray_active(char *subarray, char *container) | |
1658 | { | |
1659 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
1660 | struct mdstat_ent *ent; | |
1661 | ||
e5408a32 DW |
1662 | for (ent = mdstat; ent; ent = ent->next) |
1663 | if (is_container_member(ent, container)) | |
e5e5d7ce | 1664 | if (strcmp(to_subarray(ent, container), subarray) == 0) |
33414a01 | 1665 | break; |
33414a01 DW |
1666 | |
1667 | free_mdstat(mdstat); | |
1668 | ||
1669 | return ent != NULL; | |
1670 | } | |
1671 | ||
1672 | /* open_subarray - opens a subarray in a container | |
1673 | * @dev: container device name | |
feab51f8 | 1674 | * @st: empty supertype |
33414a01 DW |
1675 | * @quiet: block reporting errors flag |
1676 | * | |
1677 | * On success returns an fd to a container and fills in *st | |
1678 | */ | |
feab51f8 | 1679 | int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet) |
33414a01 DW |
1680 | { |
1681 | struct mdinfo *mdi; | |
a951a4f7 | 1682 | struct mdinfo *info; |
33414a01 | 1683 | int fd, err = 1; |
4dd2df09 | 1684 | char *_devnm; |
33414a01 DW |
1685 | |
1686 | fd = open(dev, O_RDWR|O_EXCL); | |
1687 | if (fd < 0) { | |
1688 | if (!quiet) | |
e7b84f9d | 1689 | pr_err("Couldn't open %s, aborting\n", |
33414a01 | 1690 | dev); |
b990032d | 1691 | return -1; |
33414a01 DW |
1692 | } |
1693 | ||
4dd2df09 N |
1694 | _devnm = fd2devnm(fd); |
1695 | if (_devnm == NULL) { | |
33414a01 | 1696 | if (!quiet) |
e7b84f9d N |
1697 | pr_err("Failed to determine device number for %s\n", |
1698 | dev); | |
33414a01 DW |
1699 | goto close_fd; |
1700 | } | |
4dd2df09 | 1701 | strcpy(st->devnm, _devnm); |
33414a01 | 1702 | |
4dd2df09 | 1703 | mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL); |
33414a01 DW |
1704 | if (!mdi) { |
1705 | if (!quiet) | |
e7b84f9d | 1706 | pr_err("Failed to read sysfs for %s\n", |
33414a01 DW |
1707 | dev); |
1708 | goto close_fd; | |
1709 | } | |
1710 | ||
1711 | if (mdi->array.level != UnSet) { | |
1712 | if (!quiet) | |
e7b84f9d | 1713 | pr_err("%s is not a container\n", dev); |
33414a01 DW |
1714 | goto free_sysfs; |
1715 | } | |
1716 | ||
1717 | st->ss = version_to_superswitch(mdi->text_version); | |
1718 | if (!st->ss) { | |
1719 | if (!quiet) | |
e7b84f9d N |
1720 | pr_err("Operation not supported for %s metadata\n", |
1721 | mdi->text_version); | |
33414a01 DW |
1722 | goto free_sysfs; |
1723 | } | |
1724 | ||
4dd2df09 | 1725 | if (st->devnm[0] == 0) { |
33414a01 | 1726 | if (!quiet) |
e7b84f9d | 1727 | pr_err("Failed to allocate device name\n"); |
33414a01 DW |
1728 | goto free_sysfs; |
1729 | } | |
1730 | ||
db20d413 | 1731 | if (!st->ss->load_container) { |
33414a01 | 1732 | if (!quiet) |
e7b84f9d | 1733 | pr_err("%s is not a container\n", dev); |
4dd2df09 | 1734 | goto free_sysfs; |
33414a01 DW |
1735 | } |
1736 | ||
db20d413 | 1737 | if (st->ss->load_container(st, fd, NULL)) { |
33414a01 | 1738 | if (!quiet) |
e7b84f9d | 1739 | pr_err("Failed to load metadata for %s\n", |
db20d413 | 1740 | dev); |
4dd2df09 | 1741 | goto free_sysfs; |
33414a01 DW |
1742 | } |
1743 | ||
a951a4f7 N |
1744 | info = st->ss->container_content(st, subarray); |
1745 | if (!info) { | |
1746 | if (!quiet) | |
e7b84f9d | 1747 | pr_err("Failed to find subarray-%s in %s\n", |
a951a4f7 N |
1748 | subarray, dev); |
1749 | goto free_super; | |
1750 | } | |
1751 | free(info); | |
1752 | ||
33414a01 DW |
1753 | err = 0; |
1754 | ||
1755 | free_super: | |
1756 | if (err) | |
1757 | st->ss->free_super(st); | |
33414a01 DW |
1758 | free_sysfs: |
1759 | sysfs_free(mdi); | |
1760 | close_fd: | |
1761 | if (err) | |
1762 | close(fd); | |
1763 | ||
1764 | if (err) | |
1765 | return -1; | |
1766 | else | |
1767 | return fd; | |
1768 | } | |
1769 | ||
7801ac20 N |
1770 | int add_disk(int mdfd, struct supertype *st, |
1771 | struct mdinfo *sra, struct mdinfo *info) | |
1772 | { | |
1773 | /* Add a device to an array, in one of 2 ways. */ | |
1774 | int rv; | |
32141c17 | 1775 | |
7801ac20 | 1776 | if (st->ss->external) { |
d23534e4 DW |
1777 | if (info->disk.state & (1<<MD_DISK_SYNC)) |
1778 | info->recovery_start = MaxSector; | |
1779 | else | |
1780 | info->recovery_start = 0; | |
2904b26f | 1781 | rv = sysfs_add_disk(sra, info, 0); |
7801ac20 N |
1782 | if (! rv) { |
1783 | struct mdinfo *sd2; | |
f35f2525 N |
1784 | for (sd2 = sra->devs; sd2; sd2=sd2->next) |
1785 | if (sd2 == info) | |
1786 | break; | |
1787 | if (sd2 == NULL) { | |
503975b9 | 1788 | sd2 = xmalloc(sizeof(*sd2)); |
f35f2525 N |
1789 | *sd2 = *info; |
1790 | sd2->next = sra->devs; | |
1791 | sra->devs = sd2; | |
1792 | } | |
7801ac20 N |
1793 | } |
1794 | } else | |
7801ac20 N |
1795 | rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk); |
1796 | return rv; | |
1797 | } | |
1798 | ||
de6ae750 N |
1799 | int remove_disk(int mdfd, struct supertype *st, |
1800 | struct mdinfo *sra, struct mdinfo *info) | |
1801 | { | |
1802 | int rv; | |
32141c17 | 1803 | |
de6ae750 | 1804 | /* Remove the disk given by 'info' from the array */ |
de6ae750 N |
1805 | if (st->ss->external) |
1806 | rv = sysfs_set_str(sra, info, "slot", "none"); | |
1807 | else | |
de6ae750 N |
1808 | rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major, |
1809 | info->disk.minor)); | |
1810 | return rv; | |
1811 | } | |
1812 | ||
1ab9ed2a | 1813 | int hot_remove_disk(int mdfd, unsigned long dev, int force) |
2dd271fe | 1814 | { |
1ab9ed2a | 1815 | int cnt = force ? 500 : 5; |
2dd271fe N |
1816 | int ret; |
1817 | ||
1818 | /* HOT_REMOVE_DISK can fail with EBUSY if there are | |
1819 | * outstanding IO requests to the device. | |
1820 | * In this case, it can be helpful to wait a little while, | |
1ab9ed2a | 1821 | * up to 5 seconds if 'force' is set, or 50 msec if not. |
2dd271fe N |
1822 | */ |
1823 | while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 && | |
1824 | errno == EBUSY && | |
1825 | cnt-- > 0) | |
239b3cc0 | 1826 | sleep_for(0, MSEC_TO_NSEC(10), true); |
2dd271fe N |
1827 | |
1828 | return ret; | |
1829 | } | |
1830 | ||
1ab9ed2a | 1831 | int sys_hot_remove_disk(int statefd, int force) |
fdd01569 | 1832 | { |
1ab9ed2a | 1833 | int cnt = force ? 500 : 5; |
fdd01569 N |
1834 | int ret; |
1835 | ||
1836 | while ((ret = write(statefd, "remove", 6)) == -1 && | |
1837 | errno == EBUSY && | |
1838 | cnt-- > 0) | |
239b3cc0 | 1839 | sleep_for(0, MSEC_TO_NSEC(10), true); |
fdd01569 N |
1840 | return ret == 6 ? 0 : -1; |
1841 | } | |
1842 | ||
f35f2525 N |
1843 | int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info) |
1844 | { | |
1845 | /* Initialise kernel's knowledge of array. | |
1846 | * This varies between externally managed arrays | |
1847 | * and older kernels | |
1848 | */ | |
700483a2 | 1849 | mdu_array_info_t inf; |
f35f2525 N |
1850 | int rv; |
1851 | ||
f35f2525 | 1852 | if (st->ss->external) |
700483a2 | 1853 | return sysfs_set_array(info, 9003); |
700483a2 JS |
1854 | |
1855 | memset(&inf, 0, sizeof(inf)); | |
1856 | inf.major_version = info->array.major_version; | |
1857 | inf.minor_version = info->array.minor_version; | |
1858 | rv = md_set_array_info(mdfd, &inf); | |
1859 | ||
f35f2525 N |
1860 | return rv; |
1861 | } | |
1862 | ||
1e5c6983 DW |
1863 | unsigned long long min_recovery_start(struct mdinfo *array) |
1864 | { | |
1865 | /* find the minimum recovery_start in an array for metadata | |
1866 | * formats that only record per-array recovery progress instead | |
1867 | * of per-device | |
1868 | */ | |
1869 | unsigned long long recovery_start = MaxSector; | |
1870 | struct mdinfo *d; | |
1871 | ||
1872 | for (d = array->devs; d; d = d->next) | |
1873 | recovery_start = min(recovery_start, d->recovery_start); | |
1874 | ||
1875 | return recovery_start; | |
1876 | } | |
1877 | ||
4dd2df09 | 1878 | int mdmon_pid(char *devnm) |
a931db9e NB |
1879 | { |
1880 | char path[100]; | |
1881 | char pid[10]; | |
1882 | int fd; | |
1883 | int n; | |
10013317 | 1884 | |
4dd2df09 | 1885 | sprintf(path, "%s/%s.pid", MDMON_DIR, devnm); |
10013317 | 1886 | |
24f6f99b | 1887 | fd = open(path, O_RDONLY | O_NOATIME, 0); |
a931db9e NB |
1888 | |
1889 | if (fd < 0) | |
cf556303 | 1890 | return -1; |
a931db9e NB |
1891 | n = read(fd, pid, 9); |
1892 | close(fd); | |
1893 | if (n <= 0) | |
cf556303 | 1894 | return -1; |
24f6f99b | 1895 | return atoi(pid); |
a931db9e NB |
1896 | } |
1897 | ||
4dd2df09 | 1898 | int mdmon_running(char *devnm) |
a931db9e | 1899 | { |
4dd2df09 | 1900 | int pid = mdmon_pid(devnm); |
24f6f99b | 1901 | if (pid <= 0) |
a931db9e | 1902 | return 0; |
24f6f99b | 1903 | if (kill(pid, 0) == 0) |
a931db9e NB |
1904 | return 1; |
1905 | return 0; | |
1906 | } | |
1907 | ||
4dd2df09 | 1908 | int start_mdmon(char *devnm) |
8850ee3e | 1909 | { |
ff6bb131 | 1910 | int i; |
44d2e365 | 1911 | int len; |
ca3b6696 | 1912 | pid_t pid; |
9fe32043 | 1913 | int status; |
44d2e365 N |
1914 | char pathbuf[1024]; |
1915 | char *paths[4] = { | |
1916 | pathbuf, | |
85945e19 | 1917 | BINDIR "/mdmon", |
2f1bcf43 | 1918 | "./mdmon", |
44d2e365 N |
1919 | NULL |
1920 | }; | |
8850ee3e | 1921 | |
40ebbb9c | 1922 | if (check_env("MDADM_NO_MDMON")) |
8850ee3e | 1923 | return 0; |
ff6bb131 MT |
1924 | if (continue_via_systemd(devnm, MDMON_SERVICE)) |
1925 | return 0; | |
8850ee3e | 1926 | |
ff6bb131 | 1927 | /* That failed, try running mdmon directly */ |
9cf014ec | 1928 | len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1); |
44d2e365 N |
1929 | if (len > 0) { |
1930 | char *sl; | |
1931 | pathbuf[len] = 0; | |
1932 | sl = strrchr(pathbuf, '/'); | |
1933 | if (sl) | |
1934 | sl++; | |
1935 | else | |
1936 | sl = pathbuf; | |
1937 | strcpy(sl, "mdmon"); | |
1938 | } else | |
1939 | pathbuf[0] = '\0'; | |
1940 | ||
8850ee3e N |
1941 | switch(fork()) { |
1942 | case 0: | |
ff6bb131 | 1943 | manage_fork_fds(1); |
ca3b6696 | 1944 | for (i = 0; paths[i]; i++) |
a0963a86 | 1945 | if (paths[i][0]) { |
2f1bcf43 | 1946 | execl(paths[i], paths[i], |
4dd2df09 | 1947 | devnm, NULL); |
a0963a86 | 1948 | } |
8850ee3e | 1949 | exit(1); |
7a862a02 | 1950 | case -1: pr_err("cannot run mdmon. Array remains readonly\n"); |
8850ee3e | 1951 | return -1; |
9fe32043 N |
1952 | default: /* parent - good */ |
1953 | pid = wait(&status); | |
15c10423 | 1954 | if (pid < 0 || status != 0) { |
7a862a02 | 1955 | pr_err("failed to launch mdmon. Array remains readonly\n"); |
9fe32043 | 1956 | return -1; |
15c10423 | 1957 | } |
8850ee3e N |
1958 | } |
1959 | return 0; | |
1960 | } | |
1961 | ||
148acb7b DW |
1962 | __u32 random32(void) |
1963 | { | |
1964 | __u32 rv; | |
1965 | int rfd = open("/dev/urandom", O_RDONLY); | |
1966 | if (rfd < 0 || read(rfd, &rv, 4) != 4) | |
1967 | rv = random(); | |
1968 | if (rfd >= 0) | |
1969 | close(rfd); | |
1970 | return rv; | |
1971 | } | |
1972 | ||
c5f71c24 JS |
1973 | void random_uuid(__u8 *buf) |
1974 | { | |
1975 | int fd, i, len; | |
1976 | __u32 r[4]; | |
1977 | ||
1978 | fd = open("/dev/urandom", O_RDONLY); | |
1979 | if (fd < 0) | |
1980 | goto use_random; | |
1981 | len = read(fd, buf, 16); | |
1982 | close(fd); | |
1983 | if (len != 16) | |
1984 | goto use_random; | |
1985 | ||
1986 | return; | |
1987 | ||
1988 | use_random: | |
1989 | for (i = 0; i < 4; i++) | |
1990 | r[i] = random(); | |
1991 | memcpy(buf, r, 16); | |
1992 | } | |
1993 | ||
edd8d13c NB |
1994 | int flush_metadata_updates(struct supertype *st) |
1995 | { | |
1996 | int sfd; | |
1997 | if (!st->updates) { | |
1998 | st->update_tail = NULL; | |
1999 | return -1; | |
2000 | } | |
2001 | ||
4dd2df09 | 2002 | sfd = connect_monitor(st->container_devnm); |
edd8d13c NB |
2003 | if (sfd < 0) |
2004 | return -1; | |
2005 | ||
2006 | while (st->updates) { | |
2007 | struct metadata_update *mu = st->updates; | |
2008 | st->updates = mu->next; | |
2009 | ||
2010 | send_message(sfd, mu, 0); | |
2011 | wait_reply(sfd, 0); | |
2012 | free(mu->buf); | |
2013 | free(mu); | |
2014 | } | |
2015 | ack(sfd, 0); | |
2016 | wait_reply(sfd, 0); | |
2017 | close(sfd); | |
2018 | st->update_tail = NULL; | |
2019 | return 0; | |
2020 | } | |
2021 | ||
2022 | void append_metadata_update(struct supertype *st, void *buf, int len) | |
2023 | { | |
2024 | ||
503975b9 | 2025 | struct metadata_update *mu = xmalloc(sizeof(*mu)); |
edd8d13c NB |
2026 | |
2027 | mu->buf = buf; | |
2028 | mu->len = len; | |
2029 | mu->space = NULL; | |
cb23f1f4 | 2030 | mu->space_list = NULL; |
edd8d13c NB |
2031 | mu->next = NULL; |
2032 | *st->update_tail = mu; | |
2033 | st->update_tail = &mu->next; | |
2034 | } | |
a931db9e | 2035 | |
fe6729fa NB |
2036 | #ifdef __TINYC__ |
2037 | /* tinyc doesn't optimize this check in ioctl.h out ... */ | |
2038 | unsigned int __invalid_size_argument_for_IOC = 0; | |
2039 | #endif | |
2040 | ||
326727d9 AC |
2041 | /* Pick all spares matching given criteria from a container |
2042 | * if min_size == 0 do not check size | |
2043 | * if domlist == NULL do not check domains | |
2044 | * if spare_group given add it to domains of each spare | |
2045 | * metadata allows to test domains using metadata of destination array */ | |
2046 | struct mdinfo *container_choose_spares(struct supertype *st, | |
fbfdcb06 | 2047 | struct spare_criteria *criteria, |
326727d9 AC |
2048 | struct domainlist *domlist, |
2049 | char *spare_group, | |
2050 | const char *metadata, int get_one) | |
2051 | { | |
2052 | struct mdinfo *d, **dp, *disks = NULL; | |
2053 | ||
2054 | /* get list of all disks in container */ | |
2055 | if (st->ss->getinfo_super_disks) | |
2056 | disks = st->ss->getinfo_super_disks(st); | |
2057 | ||
2058 | if (!disks) | |
2059 | return disks; | |
2060 | /* find spare devices on the list */ | |
2061 | dp = &disks->devs; | |
2062 | disks->array.spare_disks = 0; | |
2063 | while (*dp) { | |
2064 | int found = 0; | |
2065 | d = *dp; | |
2066 | if (d->disk.state == 0) { | |
2067 | /* check if size is acceptable */ | |
2068 | unsigned long long dev_size; | |
4b57ecf6 AO |
2069 | unsigned int dev_sector_size; |
2070 | int size_valid = 0; | |
2071 | int sector_size_valid = 0; | |
2072 | ||
326727d9 AC |
2073 | dev_t dev = makedev(d->disk.major,d->disk.minor); |
2074 | ||
fbfdcb06 | 2075 | if (!criteria->min_size || |
326727d9 | 2076 | (dev_size_from_id(dev, &dev_size) && |
fbfdcb06 | 2077 | dev_size >= criteria->min_size)) |
4b57ecf6 AO |
2078 | size_valid = 1; |
2079 | ||
2080 | if (!criteria->sector_size || | |
2081 | (dev_sector_size_from_id(dev, &dev_sector_size) && | |
2082 | criteria->sector_size == dev_sector_size)) | |
2083 | sector_size_valid = 1; | |
2084 | ||
2085 | found = size_valid && sector_size_valid; | |
2086 | ||
326727d9 AC |
2087 | /* check if domain matches */ |
2088 | if (found && domlist) { | |
4dd2df09 | 2089 | struct dev_policy *pol = devid_policy(dev); |
326727d9 AC |
2090 | if (spare_group) |
2091 | pol_add(&pol, pol_domain, | |
2092 | spare_group, NULL); | |
e5508b36 | 2093 | if (domain_test(domlist, pol, metadata) != 1) |
326727d9 AC |
2094 | found = 0; |
2095 | dev_policy_free(pol); | |
2096 | } | |
2097 | } | |
2098 | if (found) { | |
2099 | dp = &d->next; | |
2100 | disks->array.spare_disks++; | |
2101 | if (get_one) { | |
2102 | sysfs_free(*dp); | |
2103 | d->next = NULL; | |
2104 | } | |
2105 | } else { | |
2106 | *dp = d->next; | |
2107 | d->next = NULL; | |
2108 | sysfs_free(d); | |
2109 | } | |
2110 | } | |
2111 | return disks; | |
2112 | } | |
9eafa1de MN |
2113 | |
2114 | /* Checks if paths point to the same device | |
2115 | * Returns 0 if they do. | |
2116 | * Returns 1 if they don't. | |
2117 | * Returns -1 if something went wrong, | |
2118 | * e.g. paths are empty or the files | |
2119 | * they point to don't exist */ | |
2120 | int compare_paths (char* path1, char* path2) | |
2121 | { | |
2122 | struct stat st1,st2; | |
2123 | ||
2124 | if (path1 == NULL || path2 == NULL) | |
2125 | return -1; | |
2126 | if (stat(path1,&st1) != 0) | |
2127 | return -1; | |
2128 | if (stat(path2,&st2) != 0) | |
2129 | return -1; | |
2130 | if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev)) | |
2131 | return 0; | |
2132 | return 1; | |
2133 | } | |
a7dec3fd N |
2134 | |
2135 | /* Make sure we can open as many devices as needed */ | |
2136 | void enable_fds(int devices) | |
2137 | { | |
2138 | unsigned int fds = 20 + devices; | |
2139 | struct rlimit lim; | |
d7be7d87 | 2140 | if (getrlimit(RLIMIT_NOFILE, &lim) != 0 || lim.rlim_cur >= fds) |
a7dec3fd N |
2141 | return; |
2142 | if (lim.rlim_max < fds) | |
2143 | lim.rlim_max = fds; | |
2144 | lim.rlim_cur = fds; | |
2145 | setrlimit(RLIMIT_NOFILE, &lim); | |
2146 | } | |
a9c15847 | 2147 | |
ff6bb131 MT |
2148 | /* Close all opened descriptors if needed and redirect |
2149 | * streams to /dev/null. | |
2150 | * For debug purposed, leave STDOUT and STDERR untouched | |
2151 | * Returns: | |
2152 | * 1- if any error occurred | |
2153 | * 0- otherwise | |
2154 | */ | |
2155 | void manage_fork_fds(int close_all) | |
2156 | { | |
2157 | DIR *dir; | |
2158 | struct dirent *dirent; | |
2159 | ||
2160 | close(0); | |
2161 | open("/dev/null", O_RDWR); | |
2162 | ||
2163 | #ifndef DEBUG | |
2164 | dup2(0, 1); | |
2165 | dup2(0, 2); | |
2166 | #endif | |
2167 | ||
2168 | if (close_all == 0) | |
2169 | return; | |
2170 | ||
2171 | dir = opendir("/proc/self/fd"); | |
2172 | if (!dir) { | |
2173 | pr_err("Cannot open /proc/self/fd directory.\n"); | |
2174 | return; | |
2175 | } | |
2176 | for (dirent = readdir(dir); dirent; dirent = readdir(dir)) { | |
2177 | int fd = -1; | |
2178 | ||
2179 | if ((strcmp(dirent->d_name, ".") == 0) || | |
2180 | (strcmp(dirent->d_name, "..")) == 0) | |
2181 | continue; | |
2182 | ||
2183 | fd = strtol(dirent->d_name, NULL, 10); | |
2184 | if (fd > 2) | |
2185 | close(fd); | |
2186 | } | |
2187 | } | |
2188 | ||
2189 | /* In a systemd/udev world, it is best to get systemd to | |
2190 | * run daemon rather than running in the background. | |
2191 | * Returns: | |
2192 | * 1- if systemd service has been started | |
2193 | * 0- otherwise | |
2194 | */ | |
2195 | int continue_via_systemd(char *devnm, char *service_name) | |
2196 | { | |
2197 | int pid, status; | |
2198 | char pathbuf[1024]; | |
2199 | ||
2200 | /* Simply return that service cannot be started */ | |
2201 | if (check_env("MDADM_NO_SYSTEMCTL")) | |
2202 | return 0; | |
2203 | switch (fork()) { | |
2204 | case 0: | |
2205 | manage_fork_fds(1); | |
2206 | snprintf(pathbuf, sizeof(pathbuf), | |
2207 | "%s@%s.service", service_name, devnm); | |
2208 | status = execl("/usr/bin/systemctl", "systemctl", "restart", | |
2209 | pathbuf, NULL); | |
2210 | status = execl("/bin/systemctl", "systemctl", "restart", | |
2211 | pathbuf, NULL); | |
2212 | exit(1); | |
2213 | case -1: /* Just do it ourselves. */ | |
2214 | break; | |
2215 | default: /* parent - good */ | |
2216 | pid = wait(&status); | |
2217 | if (pid >= 0 && status == 0) | |
2218 | return 1; | |
2219 | } | |
2220 | return 0; | |
2221 | } | |
2222 | ||
a9c15847 N |
2223 | int in_initrd(void) |
2224 | { | |
2225 | /* This is based on similar function in systemd. */ | |
2226 | struct statfs s; | |
76d0f188 JS |
2227 | /* statfs.f_type is signed long on s390x and MIPS, causing all |
2228 | sorts of sign extension problems with RAMFS_MAGIC being | |
2229 | defined as 0x858458f6 */ | |
a9c15847 | 2230 | return statfs("/", &s) >= 0 && |
7ac5d47e | 2231 | ((unsigned long)s.f_type == TMPFS_MAGIC || |
76d0f188 JS |
2232 | ((unsigned long)s.f_type & 0xFFFFFFFFUL) == |
2233 | ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL)); | |
a9c15847 | 2234 | } |
8832342d N |
2235 | |
2236 | void reopen_mddev(int mdfd) | |
2237 | { | |
2238 | /* Re-open without any O_EXCL, but keep | |
2239 | * the same fd | |
2240 | */ | |
2241 | char *devnm; | |
2242 | int fd; | |
2243 | devnm = fd2devnm(mdfd); | |
2244 | close(mdfd); | |
2245 | fd = open_dev(devnm); | |
2246 | if (fd >= 0 && fd != mdfd) | |
2247 | dup2(fd, mdfd); | |
2248 | } | |
7716570e | 2249 | |
e80357f8 GJ |
2250 | static struct cmap_hooks *cmap_hooks = NULL; |
2251 | static int is_cmap_hooks_ready = 0; | |
7716570e | 2252 | |
e80357f8 GJ |
2253 | void set_cmap_hooks(void) |
2254 | { | |
2255 | cmap_hooks = xmalloc(sizeof(struct cmap_hooks)); | |
2256 | cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL); | |
2257 | if (!cmap_hooks->cmap_handle) | |
2258 | return; | |
7716570e | 2259 | |
b7a462e5 JS |
2260 | cmap_hooks->initialize = |
2261 | dlsym(cmap_hooks->cmap_handle, "cmap_initialize"); | |
2262 | cmap_hooks->get_string = | |
2263 | dlsym(cmap_hooks->cmap_handle, "cmap_get_string"); | |
e80357f8 | 2264 | cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize"); |
7716570e | 2265 | |
e80357f8 GJ |
2266 | if (!cmap_hooks->initialize || !cmap_hooks->get_string || |
2267 | !cmap_hooks->finalize) | |
2268 | dlclose(cmap_hooks->cmap_handle); | |
2269 | else | |
2270 | is_cmap_hooks_ready = 1; | |
2271 | } | |
7716570e | 2272 | |
e80357f8 GJ |
2273 | int get_cluster_name(char **cluster_name) |
2274 | { | |
2275 | int rv = -1; | |
2276 | cmap_handle_t handle; | |
7716570e | 2277 | |
e80357f8 GJ |
2278 | if (!is_cmap_hooks_ready) |
2279 | return rv; | |
7716570e | 2280 | |
e80357f8 | 2281 | rv = cmap_hooks->initialize(&handle); |
7716570e GJ |
2282 | if (rv != CS_OK) |
2283 | goto out; | |
2284 | ||
e80357f8 | 2285 | rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name); |
7716570e GJ |
2286 | if (rv != CS_OK) { |
2287 | free(*cluster_name); | |
2288 | rv = -1; | |
2289 | goto name_err; | |
2290 | } | |
2291 | ||
2292 | rv = 0; | |
2293 | name_err: | |
e80357f8 | 2294 | cmap_hooks->finalize(handle); |
7716570e | 2295 | out: |
7716570e GJ |
2296 | return rv; |
2297 | } | |
d15a1f72 GJ |
2298 | |
2299 | void set_dlm_hooks(void) | |
2300 | { | |
2301 | dlm_hooks = xmalloc(sizeof(struct dlm_hooks)); | |
2302 | dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL); | |
2303 | if (!dlm_hooks->dlm_handle) | |
2304 | return; | |
2305 | ||
1b7eb962 GJ |
2306 | dlm_hooks->open_lockspace = |
2307 | dlsym(dlm_hooks->dlm_handle, "dlm_open_lockspace"); | |
b7a462e5 JS |
2308 | dlm_hooks->create_lockspace = |
2309 | dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace"); | |
2310 | dlm_hooks->release_lockspace = | |
2311 | dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace"); | |
d15a1f72 | 2312 | dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock"); |
1b7eb962 GJ |
2313 | dlm_hooks->ls_unlock_wait = |
2314 | dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock_wait"); | |
d15a1f72 GJ |
2315 | dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd"); |
2316 | dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch"); | |
2317 | ||
1b7eb962 GJ |
2318 | if (!dlm_hooks->open_lockspace || !dlm_hooks->create_lockspace || |
2319 | !dlm_hooks->ls_lock || !dlm_hooks->ls_unlock_wait || | |
2320 | !dlm_hooks->release_lockspace || !dlm_hooks->ls_get_fd || | |
2321 | !dlm_hooks->dispatch) | |
d15a1f72 GJ |
2322 | dlclose(dlm_hooks->dlm_handle); |
2323 | else | |
2324 | is_dlm_hooks_ready = 1; | |
2325 | } | |
e80357f8 GJ |
2326 | |
2327 | void set_hooks(void) | |
2328 | { | |
2329 | set_dlm_hooks(); | |
2330 | set_cmap_hooks(); | |
2331 | } | |
b2514242 PB |
2332 | |
2333 | int zero_disk_range(int fd, unsigned long long sector, size_t count) | |
2334 | { | |
2335 | int ret = 0; | |
2336 | int fd_zero; | |
2337 | void *addr = NULL; | |
2338 | size_t written = 0; | |
2339 | size_t len = count * 512; | |
2340 | ssize_t n; | |
2341 | ||
2342 | fd_zero = open("/dev/zero", O_RDONLY); | |
2343 | if (fd_zero < 0) { | |
2344 | pr_err("Cannot open /dev/zero\n"); | |
2345 | return -1; | |
2346 | } | |
2347 | ||
2348 | if (lseek64(fd, sector * 512, SEEK_SET) < 0) { | |
2349 | ret = -errno; | |
2350 | pr_err("Failed to seek offset for zeroing\n"); | |
2351 | goto out; | |
2352 | } | |
2353 | ||
2354 | addr = mmap(NULL, len, PROT_READ, MAP_PRIVATE, fd_zero, 0); | |
2355 | ||
2356 | if (addr == MAP_FAILED) { | |
2357 | ret = -errno; | |
2358 | pr_err("Mapping /dev/zero failed\n"); | |
2359 | goto out; | |
2360 | } | |
2361 | ||
2362 | do { | |
2363 | n = write(fd, addr + written, len - written); | |
2364 | if (n < 0) { | |
2365 | if (errno == EINTR) | |
2366 | continue; | |
2367 | ret = -errno; | |
2368 | pr_err("Zeroing disk range failed\n"); | |
2369 | break; | |
2370 | } | |
2371 | written += n; | |
2372 | } while (written != len); | |
2373 | ||
2374 | munmap(addr, len); | |
2375 | ||
2376 | out: | |
2377 | close(fd_zero); | |
2378 | return ret; | |
2379 | } | |
239b3cc0 MG |
2380 | |
2381 | /** | |
2382 | * sleep_for() - Sleeps for specified time. | |
2383 | * @sec: Seconds to sleep for. | |
2384 | * @nsec: Nanoseconds to sleep for, has to be less than one second. | |
2385 | * @wake_after_interrupt: If set, wake up if interrupted. | |
2386 | * | |
2387 | * Function immediately returns if error different than EINTR occurs. | |
2388 | */ | |
2389 | void sleep_for(unsigned int sec, long nsec, bool wake_after_interrupt) | |
2390 | { | |
2391 | struct timespec delay = {.tv_sec = sec, .tv_nsec = nsec}; | |
2392 | ||
2393 | assert(nsec < MSEC_TO_NSEC(1000)); | |
2394 | ||
2395 | do { | |
2396 | errno = 0; | |
2397 | nanosleep(&delay, &delay); | |
2398 | if (errno != 0 && errno != EINTR) { | |
2399 | pr_err("Error sleeping for %us %ldns: %s\n", sec, nsec, strerror(errno)); | |
2400 | return; | |
2401 | } | |
2402 | } while (!wake_after_interrupt && errno == EINTR); | |
2403 | } | |
ee9dcf95 MG |
2404 | |
2405 | /* is_directory() - Checks if directory provided by path is indeed a regular directory. | |
2406 | * @path: directory path to be checked | |
2407 | * | |
2408 | * Doesn't accept symlinks. | |
2409 | * | |
2410 | * Return: true if is a directory, false if not | |
2411 | */ | |
2412 | bool is_directory(const char *path) | |
2413 | { | |
2414 | struct stat st; | |
2415 | ||
2416 | if (lstat(path, &st) != 0) { | |
2417 | pr_err("%s: %s\n", strerror(errno), path); | |
2418 | return false; | |
2419 | } | |
2420 | ||
2421 | if (!S_ISDIR(st.st_mode)) | |
2422 | return false; | |
2423 | ||
2424 | return true; | |
2425 | } | |
2426 | ||
2427 | /* | |
2428 | * is_file() - Checks if file provided by path is indeed a regular file. | |
2429 | * @path: file path to be checked | |
2430 | * | |
2431 | * Doesn't accept symlinks. | |
2432 | * | |
2433 | * Return: true if is a file, false if not | |
2434 | */ | |
2435 | bool is_file(const char *path) | |
2436 | { | |
2437 | struct stat st; | |
2438 | ||
2439 | if (lstat(path, &st) != 0) { | |
2440 | pr_err("%s: %s\n", strerror(errno), path); | |
2441 | return false; | |
2442 | } | |
2443 | ||
2444 | if (!S_ISREG(st.st_mode)) | |
2445 | return false; | |
2446 | ||
2447 | return true; | |
2448 | } |