]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
4 | * Copyright (C) 2001-2013 Neil Brown <neilb@suse.de> | |
5 | * | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License as published by | |
9 | * the Free Software Foundation; either version 2 of the License, or | |
10 | * (at your option) any later version. | |
11 | * | |
12 | * This program is distributed in the hope that it will be useful, | |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | * GNU General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with this program; if not, write to the Free Software | |
19 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
20 | * | |
21 | * Author: Neil Brown | |
22 | * Email: <neilb@suse.de> | |
23 | */ | |
24 | ||
25 | #include "mdadm.h" | |
26 | #include "md_p.h" | |
27 | #include <sys/socket.h> | |
28 | #include <sys/utsname.h> | |
29 | #include <sys/wait.h> | |
30 | #include <sys/un.h> | |
31 | #include <sys/resource.h> | |
32 | #include <sys/vfs.h> | |
33 | #include <sys/mman.h> | |
34 | #include <linux/magic.h> | |
35 | #include <poll.h> | |
36 | #include <ctype.h> | |
37 | #include <dirent.h> | |
38 | #include <dlfcn.h> | |
39 | #include <limits.h> | |
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 { | |
50 | int op; | |
51 | int flags; | |
52 | int datalen; | |
53 | void *data; | |
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 | }; | |
73 | ||
74 | #include "part.h" | |
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 | ||
85 | static int is_dlm_hooks_ready = 0; | |
86 | ||
87 | int dlm_funs_ready(void) | |
88 | { | |
89 | return is_dlm_hooks_ready ? 1 : 0; | |
90 | } | |
91 | ||
92 | static struct dlm_hooks *dlm_hooks = NULL; | |
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 | ||
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 | ||
128 | static char *cluster_name = NULL; | |
129 | /* Create the lockspace, take bitmapXXX locks on all the bitmaps. */ | |
130 | int cluster_get_dlmlock(void) | |
131 | { | |
132 | int ret = -1; | |
133 | char str[64]; | |
134 | int flags = LKF_NOQUEUE; | |
135 | int retry_count = 0; | |
136 | ||
137 | if (!dlm_funs_ready()) { | |
138 | pr_err("Something wrong with dlm library\n"); | |
139 | return -1; | |
140 | } | |
141 | ||
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 | ||
148 | dlm_lock_res = xmalloc(sizeof(struct dlm_lock_resource)); | |
149 | dlm_lock_res->ls = dlm_hooks->open_lockspace(cluster_name); | |
150 | if (!dlm_lock_res->ls) { | |
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); | |
158 | } | |
159 | ||
160 | snprintf(str, 64, "bitmap%s", cluster_name); | |
161 | retry: | |
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); | |
165 | if (ret) { | |
166 | pr_err("error %d when get PW mode on lock %s\n", errno, str); | |
167 | /* let's try several times if EAGAIN happened */ | |
168 | if (dlm_lock_res->lksb.sb_status == EAGAIN && retry_count < 10) { | |
169 | sleep_for(10, 0, true); | |
170 | retry_count++; | |
171 | goto retry; | |
172 | } | |
173 | dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1); | |
174 | return ret; | |
175 | } | |
176 | ||
177 | /* Wait for it to complete */ | |
178 | poll_for_ast(dlm_lock_res->ls); | |
179 | ||
180 | if (dlm_lock_res->lksb.sb_status) { | |
181 | pr_err("failed to lock cluster\n"); | |
182 | return -1; | |
183 | } | |
184 | return 1; | |
185 | } | |
186 | ||
187 | int cluster_release_dlmlock(void) | |
188 | { | |
189 | int ret = -1; | |
190 | ||
191 | if (!cluster_name) | |
192 | goto out; | |
193 | ||
194 | if (!dlm_lock_res->lksb.sb_lkid) | |
195 | goto out; | |
196 | ||
197 | ret = dlm_hooks->ls_unlock_wait(dlm_lock_res->ls, | |
198 | dlm_lock_res->lksb.sb_lkid, 0, | |
199 | &dlm_lock_res->lksb); | |
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) { | |
211 | pr_err("error %d happened in ast when unlock lockspace\n", | |
212 | errno); | |
213 | /* XXX make sure the lockspace is unlocked eventually */ | |
214 | goto out; | |
215 | } | |
216 | ||
217 | ret = dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1); | |
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 | ||
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 | ||
253 | int md_array_active(int fd) | |
254 | { | |
255 | struct mdinfo *sra; | |
256 | struct mdu_array_info_s array; | |
257 | int ret = 0; | |
258 | ||
259 | sra = sysfs_read(fd, NULL, GET_ARRAY_STATE); | |
260 | if (sra) { | |
261 | if (!md_array_is_active(sra)) | |
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 | */ | |
270 | ret = md_get_array_info(fd, &array); | |
271 | } | |
272 | ||
273 | return !ret; | |
274 | } | |
275 | ||
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 | ||
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 | ||
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 | ||
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 | ||
308 | int get_linux_version() | |
309 | { | |
310 | struct utsname name; | |
311 | char *cp; | |
312 | int a = 0, b = 0,c = 0; | |
313 | if (uname(&name) <0) | |
314 | return -1; | |
315 | ||
316 | cp = name.release; | |
317 | a = strtoul(cp, &cp, 10); | |
318 | if (*cp == '.') | |
319 | b = strtoul(cp+1, &cp, 10); | |
320 | if (*cp == '.') | |
321 | c = strtoul(cp+1, &cp, 10); | |
322 | ||
323 | return (a*1000000)+(b*1000)+c; | |
324 | } | |
325 | ||
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 | ||
351 | unsigned long long parse_size(char *size) | |
352 | { | |
353 | /* parse 'size' which should be a number optionally | |
354 | * followed by 'K', 'M'. 'G' or 'T'. | |
355 | * Without a suffix, K is assumed. | |
356 | * Number returned is in sectors (half-K) | |
357 | * INVALID_SECTORS returned on error. | |
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; | |
376 | case 'T': | |
377 | c++; | |
378 | s *= 1024 * 1024 * 1024 * 2LL; | |
379 | break; | |
380 | case 's': /* sectors */ | |
381 | c++; | |
382 | break; | |
383 | } | |
384 | } else | |
385 | s = INVALID_SECTORS; | |
386 | if (*c) | |
387 | s = INVALID_SECTORS; | |
388 | return s; | |
389 | } | |
390 | ||
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 | ||
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 | { | |
424 | int ln, mode; | |
425 | char *m; | |
426 | ||
427 | if (!layout) | |
428 | return -1; | |
429 | ||
430 | /* Parse the layout string for 'faulty' */ | |
431 | ln = strcspn(layout, "0123456789"); | |
432 | m = xstrdup(layout); | |
433 | m[ln] = 0; | |
434 | mode = map_name(faultylayout, m); | |
435 | free(m); | |
436 | ||
437 | if (mode == UnSet) | |
438 | return -1; | |
439 | ||
440 | return mode | (atoi(layout+ln)<< ModeShift); | |
441 | } | |
442 | ||
443 | int parse_cluster_confirm_arg(char *input, char **devname, int *slot) | |
444 | { | |
445 | char *dev; | |
446 | *slot = strtoul(input, &dev, 10); | |
447 | if (dev == input || dev[0] != ':') | |
448 | return -1; | |
449 | *devname = dev+1; | |
450 | return 0; | |
451 | } | |
452 | ||
453 | void remove_partitions(int fd) | |
454 | { | |
455 | /* remove partitions from this block devices. | |
456 | * This is used for components added to an array | |
457 | */ | |
458 | #ifdef BLKPG_DEL_PARTITION | |
459 | struct blkpg_ioctl_arg a; | |
460 | struct blkpg_partition p; | |
461 | ||
462 | a.op = BLKPG_DEL_PARTITION; | |
463 | a.data = (void*)&p; | |
464 | a.datalen = sizeof(p); | |
465 | a.flags = 0; | |
466 | memset(a.data, 0, a.datalen); | |
467 | for (p.pno = 0; p.pno < 16; p.pno++) | |
468 | ioctl(fd, BLKPG, &a); | |
469 | #endif | |
470 | } | |
471 | ||
472 | int test_partition(int fd) | |
473 | { | |
474 | /* Check if fd is a whole-disk or a partition. | |
475 | * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION | |
476 | * will return ENXIO on an invalid partition number. | |
477 | */ | |
478 | struct blkpg_ioctl_arg a; | |
479 | struct blkpg_partition p; | |
480 | a.op = BLKPG_DEL_PARTITION; | |
481 | a.data = (void*)&p; | |
482 | a.datalen = sizeof(p); | |
483 | a.flags = 0; | |
484 | memset(a.data, 0, a.datalen); | |
485 | p.pno = 1<<30; | |
486 | if (ioctl(fd, BLKPG, &a) == 0) | |
487 | /* Very unlikely, but not a partition */ | |
488 | return 0; | |
489 | if (errno == ENXIO || errno == ENOTTY) | |
490 | /* not a partition */ | |
491 | return 0; | |
492 | ||
493 | return 1; | |
494 | } | |
495 | ||
496 | int test_partition_from_id(dev_t id) | |
497 | { | |
498 | char buf[20]; | |
499 | int fd, rv; | |
500 | ||
501 | sprintf(buf, "%d:%d", major(id), minor(id)); | |
502 | fd = dev_open(buf, O_RDONLY); | |
503 | if (fd < 0) | |
504 | return -1; | |
505 | rv = test_partition(fd); | |
506 | close(fd); | |
507 | return rv; | |
508 | } | |
509 | ||
510 | int enough(int level, int raid_disks, int layout, int clean, char *avail) | |
511 | { | |
512 | int copies, first; | |
513 | int i; | |
514 | int avail_disks = 0; | |
515 | ||
516 | for (i = 0; i < raid_disks; i++) | |
517 | avail_disks += !!avail[i]; | |
518 | ||
519 | switch (level) { | |
520 | case 10: | |
521 | /* This is the tricky one - we need to check | |
522 | * which actual disks are present. | |
523 | */ | |
524 | copies = (layout&255)* ((layout>>8) & 255); | |
525 | first = 0; | |
526 | do { | |
527 | /* there must be one of the 'copies' form 'first' */ | |
528 | int n = copies; | |
529 | int cnt = 0; | |
530 | int this = first; | |
531 | while (n--) { | |
532 | if (avail[this]) | |
533 | cnt++; | |
534 | this = (this+1) % raid_disks; | |
535 | } | |
536 | if (cnt == 0) | |
537 | return 0; | |
538 | first = (first+(layout&255)) % raid_disks; | |
539 | } while (first != 0); | |
540 | return 1; | |
541 | ||
542 | case LEVEL_MULTIPATH: | |
543 | return avail_disks>= 1; | |
544 | case LEVEL_LINEAR: | |
545 | case 0: | |
546 | return avail_disks == raid_disks; | |
547 | case 1: | |
548 | return avail_disks >= 1; | |
549 | case 4: | |
550 | if (avail_disks == raid_disks - 1 && | |
551 | !avail[raid_disks - 1]) | |
552 | /* If just the parity device is missing, then we | |
553 | * have enough, even if not clean | |
554 | */ | |
555 | return 1; | |
556 | /* FALL THROUGH */ | |
557 | case 5: | |
558 | if (clean) | |
559 | return avail_disks >= raid_disks-1; | |
560 | else | |
561 | return avail_disks >= raid_disks; | |
562 | case 6: | |
563 | if (clean) | |
564 | return avail_disks >= raid_disks-2; | |
565 | else | |
566 | return avail_disks >= raid_disks; | |
567 | default: | |
568 | return 0; | |
569 | } | |
570 | } | |
571 | ||
572 | char *__fname_from_uuid(int id[4], int swap, char *buf, char sep) | |
573 | { | |
574 | int i, j; | |
575 | char uuid[16]; | |
576 | char *c = buf; | |
577 | strcpy(c, "UUID-"); | |
578 | c += strlen(c); | |
579 | copy_uuid(uuid, id, swap); | |
580 | for (i = 0; i < 4; i++) { | |
581 | if (i) | |
582 | *c++ = sep; | |
583 | for (j = 3; j >= 0; j--) { | |
584 | sprintf(c,"%02x", (unsigned char) uuid[j+4*i]); | |
585 | c+= 2; | |
586 | } | |
587 | } | |
588 | return buf; | |
589 | ||
590 | } | |
591 | ||
592 | /** | |
593 | * fname_from_uuid() - generate uuid string. Should not be used with super1. | |
594 | * @info: info with uuid | |
595 | * @buf: buf to fill. | |
596 | * | |
597 | * This routine should not be used with super1. See detail_fname_from_uuid() for details. It does | |
598 | * not use superswitch swapuuid as it should be 0 but it has to do UUID conversion if host is big | |
599 | * endian- left for backward compatibility. | |
600 | */ | |
601 | char *fname_from_uuid(struct mdinfo *info, char *buf) | |
602 | { | |
603 | #if __BYTE_ORDER == BIG_ENDIAN | |
604 | return __fname_from_uuid(info->uuid, true, buf, ':'); | |
605 | #else | |
606 | return __fname_from_uuid(info->uuid, false, buf, ':'); | |
607 | #endif | |
608 | } | |
609 | ||
610 | int check_ext2(int fd, char *name) | |
611 | { | |
612 | /* | |
613 | * Check for an ext2fs file system. | |
614 | * Superblock is always 1K at 1K offset | |
615 | * | |
616 | * s_magic is le16 at 56 == 0xEF53 | |
617 | * report mtime - le32 at 44 | |
618 | * blocks - le32 at 4 | |
619 | * logblksize - le32 at 24 | |
620 | */ | |
621 | unsigned char sb[1024]; | |
622 | time_t mtime; | |
623 | unsigned long long size; | |
624 | int bsize; | |
625 | if (lseek(fd, 1024,0)!= 1024) | |
626 | return 0; | |
627 | if (read(fd, sb, 1024)!= 1024) | |
628 | return 0; | |
629 | if (sb[56] != 0x53 || sb[57] != 0xef) | |
630 | return 0; | |
631 | ||
632 | mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8; | |
633 | bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8; | |
634 | size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8; | |
635 | size <<= bsize; | |
636 | pr_err("%s appears to contain an ext2fs file system\n", | |
637 | name); | |
638 | cont_err("size=%lluK mtime=%s", size, ctime(&mtime)); | |
639 | return 1; | |
640 | } | |
641 | ||
642 | int check_reiser(int fd, char *name) | |
643 | { | |
644 | /* | |
645 | * superblock is at 64K | |
646 | * size is 1024; | |
647 | * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52 | |
648 | * | |
649 | */ | |
650 | unsigned char sb[1024]; | |
651 | unsigned long long size; | |
652 | if (lseek(fd, 64*1024, 0) != 64*1024) | |
653 | return 0; | |
654 | if (read(fd, sb, 1024) != 1024) | |
655 | return 0; | |
656 | if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 && | |
657 | strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0) | |
658 | return 0; | |
659 | pr_err("%s appears to contain a reiserfs file system\n",name); | |
660 | size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8; | |
661 | cont_err("size = %lluK\n", size*4); | |
662 | ||
663 | return 1; | |
664 | } | |
665 | ||
666 | int check_raid(int fd, char *name) | |
667 | { | |
668 | struct mdinfo info; | |
669 | time_t crtime; | |
670 | char *level; | |
671 | struct supertype *st = guess_super(fd); | |
672 | ||
673 | if (!st) | |
674 | return 0; | |
675 | if (st->ss->add_to_super != NULL) { | |
676 | st->ss->load_super(st, fd, name); | |
677 | /* Looks like a raid array .. */ | |
678 | pr_err("%s appears to be part of a raid array:\n", name); | |
679 | st->ss->getinfo_super(st, &info, NULL); | |
680 | st->ss->free_super(st); | |
681 | crtime = info.array.ctime; | |
682 | level = map_num(pers, info.array.level); | |
683 | if (!level) | |
684 | level = "-unknown-"; | |
685 | cont_err("level=%s devices=%d ctime=%s", | |
686 | level, info.array.raid_disks, ctime(&crtime)); | |
687 | } else { | |
688 | /* Looks like GPT or MBR */ | |
689 | pr_err("partition table exists on %s\n", name); | |
690 | } | |
691 | return 1; | |
692 | } | |
693 | ||
694 | int fstat_is_blkdev(int fd, char *devname, dev_t *rdev) | |
695 | { | |
696 | struct stat stb; | |
697 | ||
698 | if (fstat(fd, &stb) != 0) { | |
699 | pr_err("fstat failed for %s: %s\n", devname, strerror(errno)); | |
700 | return 0; | |
701 | } | |
702 | if ((S_IFMT & stb.st_mode) != S_IFBLK) { | |
703 | pr_err("%s is not a block device.\n", devname); | |
704 | return 0; | |
705 | } | |
706 | if (rdev) | |
707 | *rdev = stb.st_rdev; | |
708 | return 1; | |
709 | } | |
710 | ||
711 | int stat_is_blkdev(char *devname, dev_t *rdev) | |
712 | { | |
713 | struct stat stb; | |
714 | ||
715 | if (stat(devname, &stb) != 0) { | |
716 | pr_err("stat failed for %s: %s\n", devname, strerror(errno)); | |
717 | return 0; | |
718 | } | |
719 | if ((S_IFMT & stb.st_mode) != S_IFBLK) { | |
720 | pr_err("%s is not a block device.\n", devname); | |
721 | return 0; | |
722 | } | |
723 | if (rdev) | |
724 | *rdev = stb.st_rdev; | |
725 | return 1; | |
726 | } | |
727 | ||
728 | int ask(char *mesg) | |
729 | { | |
730 | char *add = ""; | |
731 | int i; | |
732 | for (i = 0; i < 5; i++) { | |
733 | char buf[100]; | |
734 | fprintf(stderr, "%s%s", mesg, add); | |
735 | fflush(stderr); | |
736 | if (fgets(buf, 100, stdin)==NULL) | |
737 | return 0; | |
738 | if (buf[0]=='y' || buf[0]=='Y') | |
739 | return 1; | |
740 | if (buf[0]=='n' || buf[0]=='N') | |
741 | return 0; | |
742 | add = "(y/n) "; | |
743 | } | |
744 | pr_err("assuming 'no'\n"); | |
745 | return 0; | |
746 | } | |
747 | ||
748 | int is_standard(char *dev, int *nump) | |
749 | { | |
750 | /* tests if dev is a "standard" md dev name. | |
751 | * i.e if the last component is "/dNN" or "/mdNN", | |
752 | * where NN is a string of digits | |
753 | * Returns 1 if a partitionable standard, | |
754 | * -1 if non-partitonable, | |
755 | * 0 if not a standard name. | |
756 | */ | |
757 | char *d = strrchr(dev, '/'); | |
758 | int type = 0; | |
759 | int num; | |
760 | if (!d) | |
761 | return 0; | |
762 | if (strncmp(d, "/d",2) == 0) | |
763 | d += 2, type = 1; /* /dev/md/dN{pM} */ | |
764 | else if (strncmp(d, "/md_d", 5) == 0) | |
765 | d += 5, type = 1; /* /dev/md_dN{pM} */ | |
766 | else if (strncmp(d, "/md", 3) == 0) | |
767 | d += 3, type = -1; /* /dev/mdN */ | |
768 | else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0) | |
769 | d += 1, type = -1; /* /dev/md/N */ | |
770 | else | |
771 | return 0; | |
772 | if (!*d) | |
773 | return 0; | |
774 | num = atoi(d); | |
775 | while (isdigit(*d)) | |
776 | d++; | |
777 | if (*d) | |
778 | return 0; | |
779 | if (nump) *nump = num; | |
780 | ||
781 | return type; | |
782 | } | |
783 | ||
784 | unsigned long calc_csum(void *super, int bytes) | |
785 | { | |
786 | unsigned long long newcsum = 0; | |
787 | int i; | |
788 | unsigned int csum; | |
789 | unsigned int *superc = (unsigned int*) super; | |
790 | ||
791 | for(i = 0; i < bytes/4; i++) | |
792 | newcsum += superc[i]; | |
793 | csum = (newcsum& 0xffffffff) + (newcsum>>32); | |
794 | #ifdef __alpha__ | |
795 | /* The in-kernel checksum calculation is always 16bit on | |
796 | * the alpha, though it is 32 bit on i386... | |
797 | * I wonder what it is elsewhere... (it uses an API in | |
798 | * a way that it shouldn't). | |
799 | */ | |
800 | csum = (csum & 0xffff) + (csum >> 16); | |
801 | csum = (csum & 0xffff) + (csum >> 16); | |
802 | #endif | |
803 | return csum; | |
804 | } | |
805 | ||
806 | char *human_size(long long bytes) | |
807 | { | |
808 | static char buf[47]; | |
809 | ||
810 | /* We convert bytes to either centi-M{ega,ibi}bytes, | |
811 | * centi-G{igi,ibi}bytes or centi-T{era,ebi}bytes | |
812 | * with appropriate rounding, and then print | |
813 | * 1/100th of those as a decimal. | |
814 | * We allow upto 2048Megabytes before converting to | |
815 | * gigabytes and 2048Gigabytes before converting to | |
816 | * terabytes, as that shows more precision and isn't | |
817 | * too large a number. | |
818 | */ | |
819 | ||
820 | if (bytes < 5000*1024) | |
821 | buf[0] = 0; | |
822 | else if (bytes < 2*1024LL*1024LL*1024LL) { | |
823 | long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2; | |
824 | long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2; | |
825 | snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)", | |
826 | cMiB/100, cMiB % 100, cMB/100, cMB % 100); | |
827 | } else if (bytes < 2*1024LL*1024LL*1024LL*1024LL) { | |
828 | long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2; | |
829 | long cGB = (bytes / (1000000000LL/200LL ) +1) /2; | |
830 | snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)", | |
831 | cGiB/100, cGiB % 100, cGB/100, cGB % 100); | |
832 | } else { | |
833 | long cTiB = (bytes * 200LL / (1LL<<40) + 1) / 2; | |
834 | long cTB = (bytes / (1000000000000LL / 200LL) + 1) / 2; | |
835 | snprintf(buf, sizeof(buf), " (%ld.%02ld TiB %ld.%02ld TB)", | |
836 | cTiB/100, cTiB % 100, cTB/100, cTB % 100); | |
837 | } | |
838 | return buf; | |
839 | } | |
840 | ||
841 | char *human_size_brief(long long bytes, int prefix) | |
842 | { | |
843 | static char buf[30]; | |
844 | ||
845 | /* We convert bytes to either centi-M{ega,ibi}bytes, | |
846 | * centi-G{igi,ibi}bytes or centi-T{era,ebi}bytes | |
847 | * with appropriate rounding, and then print | |
848 | * 1/100th of those as a decimal. | |
849 | * We allow upto 2048Megabytes before converting to | |
850 | * gigabytes and 2048Gigabytes before converting to | |
851 | * terabytes, as that shows more precision and isn't | |
852 | * too large a number. | |
853 | * | |
854 | * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc. | |
855 | * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc. | |
856 | */ | |
857 | ||
858 | if (bytes < 5000*1024) | |
859 | buf[0] = 0; | |
860 | else if (prefix == IEC) { | |
861 | if (bytes < 2*1024LL*1024LL*1024LL) { | |
862 | long cMiB = (bytes * 200LL / (1LL<<20) +1) /2; | |
863 | snprintf(buf, sizeof(buf), "%ld.%02ldMiB", | |
864 | cMiB/100, cMiB % 100); | |
865 | } else if (bytes < 2*1024LL*1024LL*1024LL*1024LL) { | |
866 | long cGiB = (bytes * 200LL / (1LL<<30) +1) /2; | |
867 | snprintf(buf, sizeof(buf), "%ld.%02ldGiB", | |
868 | cGiB/100, cGiB % 100); | |
869 | } else { | |
870 | long cTiB = (bytes * 200LL / (1LL<<40) + 1) / 2; | |
871 | snprintf(buf, sizeof(buf), "%ld.%02ldTiB", | |
872 | cTiB/100, cTiB % 100); | |
873 | } | |
874 | } | |
875 | else if (prefix == JEDEC) { | |
876 | if (bytes < 2*1024LL*1024LL*1024LL) { | |
877 | long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2; | |
878 | snprintf(buf, sizeof(buf), "%ld.%02ldMB", | |
879 | cMB/100, cMB % 100); | |
880 | } else if (bytes < 2*1024LL*1024LL*1024LL*1024LL) { | |
881 | long cGB = (bytes / (1000000000LL/200LL ) +1) /2; | |
882 | snprintf(buf, sizeof(buf), "%ld.%02ldGB", | |
883 | cGB/100, cGB % 100); | |
884 | } else { | |
885 | long cTB = (bytes / (1000000000000LL / 200LL) + 1) / 2; | |
886 | snprintf(buf, sizeof(buf), "%ld.%02ldTB", | |
887 | cTB/100, cTB % 100); | |
888 | } | |
889 | } | |
890 | else | |
891 | buf[0] = 0; | |
892 | ||
893 | return buf; | |
894 | } | |
895 | ||
896 | void print_r10_layout(int layout) | |
897 | { | |
898 | int near = layout & 255; | |
899 | int far = (layout >> 8) & 255; | |
900 | int offset = (layout&0x10000); | |
901 | char *sep = ""; | |
902 | ||
903 | if (near != 1) { | |
904 | printf("%s near=%d", sep, near); | |
905 | sep = ","; | |
906 | } | |
907 | if (far != 1) | |
908 | printf("%s %s=%d", sep, offset?"offset":"far", far); | |
909 | if (near*far == 1) | |
910 | printf("NO REDUNDANCY"); | |
911 | } | |
912 | ||
913 | unsigned long long calc_array_size(int level, int raid_disks, int layout, | |
914 | int chunksize, unsigned long long devsize) | |
915 | { | |
916 | if (level == 1) | |
917 | return devsize; | |
918 | devsize &= ~(unsigned long long)((chunksize>>9)-1); | |
919 | return get_data_disks(level, layout, raid_disks) * devsize; | |
920 | } | |
921 | ||
922 | int get_data_disks(int level, int layout, int raid_disks) | |
923 | { | |
924 | int data_disks = 0; | |
925 | switch (level) { | |
926 | case 0: data_disks = raid_disks; | |
927 | break; | |
928 | case 1: data_disks = 1; | |
929 | break; | |
930 | case 4: | |
931 | case 5: data_disks = raid_disks - 1; | |
932 | break; | |
933 | case 6: data_disks = raid_disks - 2; | |
934 | break; | |
935 | case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255); | |
936 | break; | |
937 | } | |
938 | ||
939 | return data_disks; | |
940 | } | |
941 | ||
942 | dev_t devnm2devid(char *devnm) | |
943 | { | |
944 | /* First look in /sys/block/$DEVNM/dev for %d:%d | |
945 | * If that fails, try parsing out a number | |
946 | */ | |
947 | char path[PATH_MAX]; | |
948 | char *ep; | |
949 | int fd; | |
950 | int mjr,mnr; | |
951 | ||
952 | snprintf(path, sizeof(path), "/sys/block/%s/dev", devnm); | |
953 | fd = open(path, O_RDONLY); | |
954 | if (fd >= 0) { | |
955 | char buf[20]; | |
956 | int n = read(fd, buf, sizeof(buf)); | |
957 | close(fd); | |
958 | if (n > 0) | |
959 | buf[n] = 0; | |
960 | if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2) | |
961 | return makedev(mjr, mnr); | |
962 | } | |
963 | if (strncmp(devnm, "md_d", 4) == 0 && | |
964 | isdigit(devnm[4]) && | |
965 | (mnr = strtoul(devnm+4, &ep, 10)) >= 0 && | |
966 | ep > devnm && *ep == 0) | |
967 | return makedev(get_mdp_major(), mnr << MdpMinorShift); | |
968 | ||
969 | if (strncmp(devnm, "md", 2) == 0 && | |
970 | isdigit(devnm[2]) && | |
971 | (mnr = strtoul(devnm+2, &ep, 10)) >= 0 && | |
972 | ep > devnm && *ep == 0) | |
973 | return makedev(MD_MAJOR, mnr); | |
974 | ||
975 | return 0; | |
976 | } | |
977 | ||
978 | /** | |
979 | * is_devname_numbered() - helper for numbered devname verification. | |
980 | * @devname: path or name to check. | |
981 | * @pref: expected devname prefix. | |
982 | * @pref_len: prefix len. | |
983 | */ | |
984 | static bool is_devname_numbered(const char *devname, const char *pref, const int pref_len) | |
985 | { | |
986 | int val; | |
987 | ||
988 | assert(devname && pref); | |
989 | ||
990 | if (strncmp(devname, pref, pref_len) != 0) | |
991 | return false; | |
992 | ||
993 | if (parse_num(&val, devname + pref_len) != 0) | |
994 | return false; | |
995 | ||
996 | if (val > 127) | |
997 | return false; | |
998 | ||
999 | return true; | |
1000 | } | |
1001 | ||
1002 | /** | |
1003 | * is_devname_md_numbered() - check if &devname is numbered MD device (md). | |
1004 | * @devname: path or name to check. | |
1005 | */ | |
1006 | bool is_devname_md_numbered(const char *devname) | |
1007 | { | |
1008 | return is_devname_numbered(devname, DEV_NUM_PREF, DEV_NUM_PREF_LEN); | |
1009 | } | |
1010 | ||
1011 | /** | |
1012 | * is_devname_md_d_numbered() - check if &devname is secondary numbered MD device (md_d). | |
1013 | * @devname: path or name to check. | |
1014 | */ | |
1015 | bool is_devname_md_d_numbered(const char *devname) | |
1016 | { | |
1017 | static const char d_dev[] = DEV_NUM_PREF "_d"; | |
1018 | ||
1019 | return is_devname_numbered(devname, d_dev, sizeof(d_dev) - 1); | |
1020 | } | |
1021 | ||
1022 | /** | |
1023 | * get_md_name() - Get main dev node of the md device. | |
1024 | * @devnm: Md device name or path. | |
1025 | * | |
1026 | * Function checks if the full name was passed and returns md name | |
1027 | * if it is the MD device. | |
1028 | * | |
1029 | * Return: Main dev node of the md device or NULL if not found. | |
1030 | */ | |
1031 | char *get_md_name(char *devnm) | |
1032 | { | |
1033 | static char devname[NAME_MAX]; | |
1034 | struct stat stb; | |
1035 | ||
1036 | if (strncmp(devnm, "/dev/", 5) == 0) | |
1037 | snprintf(devname, sizeof(devname), "%s", devnm); | |
1038 | else | |
1039 | snprintf(devname, sizeof(devname), "/dev/%s", devnm); | |
1040 | ||
1041 | if (!is_mddev(devname)) | |
1042 | return NULL; | |
1043 | if (stat(devname, &stb) == 0 && (S_IFMT&stb.st_mode) == S_IFBLK) | |
1044 | return devname; | |
1045 | ||
1046 | return NULL; | |
1047 | } | |
1048 | ||
1049 | void put_md_name(char *name) | |
1050 | { | |
1051 | if (strncmp(name, "/dev/.tmp.md", 12) == 0) | |
1052 | unlink(name); | |
1053 | } | |
1054 | ||
1055 | int get_maj_min(char *dev, int *major, int *minor) | |
1056 | { | |
1057 | char *e; | |
1058 | *major = strtoul(dev, &e, 0); | |
1059 | return (e > dev && *e == ':' && e[1] && | |
1060 | (*minor = strtoul(e+1, &e, 0)) >= 0 && | |
1061 | *e == 0); | |
1062 | } | |
1063 | ||
1064 | /** | |
1065 | * is_bit_set() - get bit value by index. | |
1066 | * @val: value. | |
1067 | * @index: index of the bit (LSB numbering). | |
1068 | * | |
1069 | * Return: bit value. | |
1070 | */ | |
1071 | bool is_bit_set(int *val, unsigned char index) | |
1072 | { | |
1073 | if ((*val) & (1 << index)) | |
1074 | return true; | |
1075 | return false; | |
1076 | } | |
1077 | ||
1078 | int dev_open(char *dev, int flags) | |
1079 | { | |
1080 | /* like 'open', but if 'dev' matches %d:%d, create a temp | |
1081 | * block device and open that | |
1082 | */ | |
1083 | int fd = -1; | |
1084 | char devname[32]; | |
1085 | int major; | |
1086 | int minor; | |
1087 | ||
1088 | if (!dev) | |
1089 | return -1; | |
1090 | flags |= O_DIRECT; | |
1091 | ||
1092 | if (get_maj_min(dev, &major, &minor)) { | |
1093 | snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d", | |
1094 | (int)getpid(), major, minor); | |
1095 | if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) { | |
1096 | fd = open(devname, flags); | |
1097 | unlink(devname); | |
1098 | } | |
1099 | if (fd < 0) { | |
1100 | /* Try /tmp as /dev appear to be read-only */ | |
1101 | snprintf(devname, sizeof(devname), | |
1102 | "/tmp/.tmp.md.%d:%d:%d", | |
1103 | (int)getpid(), major, minor); | |
1104 | if (mknod(devname, S_IFBLK|0600, | |
1105 | makedev(major, minor)) == 0) { | |
1106 | fd = open(devname, flags); | |
1107 | unlink(devname); | |
1108 | } | |
1109 | } | |
1110 | } else | |
1111 | fd = open(dev, flags); | |
1112 | return fd; | |
1113 | } | |
1114 | ||
1115 | int open_dev_flags(char *devnm, int flags) | |
1116 | { | |
1117 | dev_t devid; | |
1118 | char buf[20]; | |
1119 | ||
1120 | devid = devnm2devid(devnm); | |
1121 | sprintf(buf, "%d:%d", major(devid), minor(devid)); | |
1122 | return dev_open(buf, flags); | |
1123 | } | |
1124 | ||
1125 | int open_dev(char *devnm) | |
1126 | { | |
1127 | return open_dev_flags(devnm, O_RDONLY); | |
1128 | } | |
1129 | ||
1130 | int open_dev_excl(char *devnm) | |
1131 | { | |
1132 | char buf[20]; | |
1133 | int i; | |
1134 | int flags = O_RDWR; | |
1135 | dev_t devid = devnm2devid(devnm); | |
1136 | unsigned int delay = 1; // miliseconds | |
1137 | ||
1138 | sprintf(buf, "%d:%d", major(devid), minor(devid)); | |
1139 | for (i = 0; i < 25; i++) { | |
1140 | int fd = dev_open(buf, flags|O_EXCL); | |
1141 | if (fd >= 0) | |
1142 | return fd; | |
1143 | if (errno == EACCES && flags == O_RDWR) { | |
1144 | flags = O_RDONLY; | |
1145 | continue; | |
1146 | } | |
1147 | if (errno != EBUSY) | |
1148 | return fd; | |
1149 | sleep_for(0, MSEC_TO_NSEC(delay), true); | |
1150 | if (delay < 200) | |
1151 | delay *= 2; | |
1152 | } | |
1153 | return -1; | |
1154 | } | |
1155 | ||
1156 | int same_dev(char *one, char *two) | |
1157 | { | |
1158 | struct stat st1, st2; | |
1159 | if (stat(one, &st1) != 0) | |
1160 | return 0; | |
1161 | if (stat(two, &st2) != 0) | |
1162 | return 0; | |
1163 | if ((st1.st_mode & S_IFMT) != S_IFBLK) | |
1164 | return 0; | |
1165 | if ((st2.st_mode & S_IFMT) != S_IFBLK) | |
1166 | return 0; | |
1167 | return st1.st_rdev == st2.st_rdev; | |
1168 | } | |
1169 | ||
1170 | void wait_for(char *dev, int fd) | |
1171 | { | |
1172 | int i; | |
1173 | struct stat stb_want; | |
1174 | unsigned int delay = 1; // miliseconds | |
1175 | ||
1176 | if (fstat(fd, &stb_want) != 0 || | |
1177 | (stb_want.st_mode & S_IFMT) != S_IFBLK) | |
1178 | return; | |
1179 | ||
1180 | for (i = 0; i < 25; i++) { | |
1181 | struct stat stb; | |
1182 | if (stat(dev, &stb) == 0 && | |
1183 | (stb.st_mode & S_IFMT) == S_IFBLK && | |
1184 | (stb.st_rdev == stb_want.st_rdev)) | |
1185 | return; | |
1186 | sleep_for(0, MSEC_TO_NSEC(delay), true); | |
1187 | if (delay < 200) | |
1188 | delay *= 2; | |
1189 | } | |
1190 | if (i == 25) | |
1191 | pr_err("timeout waiting for %s\n", dev); | |
1192 | } | |
1193 | ||
1194 | struct superswitch *superlist[] = | |
1195 | { | |
1196 | &super0, &super1, | |
1197 | &super_ddf, &super_imsm, | |
1198 | &mbr, &gpt, | |
1199 | NULL | |
1200 | }; | |
1201 | ||
1202 | struct supertype *super_by_fd(int fd, char **subarrayp) | |
1203 | { | |
1204 | mdu_array_info_t array; | |
1205 | int vers; | |
1206 | int minor; | |
1207 | struct supertype *st = NULL; | |
1208 | struct mdinfo *sra; | |
1209 | char *verstr; | |
1210 | char version[20]; | |
1211 | int i; | |
1212 | char *subarray = NULL; | |
1213 | char container[32] = ""; | |
1214 | char *devnm = NULL; | |
1215 | ||
1216 | devnm = fd2devnm(fd); | |
1217 | if (!devnm) | |
1218 | return NULL; | |
1219 | ||
1220 | sra = sysfs_read(fd, NULL, GET_VERSION); | |
1221 | ||
1222 | if (sra) { | |
1223 | vers = sra->array.major_version; | |
1224 | minor = sra->array.minor_version; | |
1225 | verstr = sra->text_version; | |
1226 | } else { | |
1227 | if (md_get_array_info(fd, &array)) | |
1228 | array.major_version = array.minor_version = 0; | |
1229 | vers = array.major_version; | |
1230 | minor = array.minor_version; | |
1231 | verstr = ""; | |
1232 | } | |
1233 | ||
1234 | if (vers != -1) { | |
1235 | sprintf(version, "%d.%d", vers, minor); | |
1236 | verstr = version; | |
1237 | } | |
1238 | if (minor == -2 && is_subarray(verstr)) { | |
1239 | char *dev = verstr+1; | |
1240 | ||
1241 | subarray = strchr(dev, '/'); | |
1242 | if (subarray) { | |
1243 | *subarray++ = '\0'; | |
1244 | subarray = xstrdup(subarray); | |
1245 | } | |
1246 | strcpy(container, dev); | |
1247 | sysfs_free(sra); | |
1248 | sra = sysfs_read(-1, container, GET_VERSION); | |
1249 | if (sra && sra->text_version[0]) | |
1250 | verstr = sra->text_version; | |
1251 | else | |
1252 | verstr = "-no-metadata-"; | |
1253 | } | |
1254 | ||
1255 | for (i = 0; st == NULL && superlist[i]; i++) | |
1256 | st = superlist[i]->match_metadata_desc(verstr); | |
1257 | ||
1258 | sysfs_free(sra); | |
1259 | if (st) { | |
1260 | st->sb = NULL; | |
1261 | if (subarrayp) | |
1262 | *subarrayp = subarray; | |
1263 | strcpy(st->container_devnm, container); | |
1264 | strncpy(st->devnm, devnm, MD_NAME_MAX - 1); | |
1265 | } else | |
1266 | free(subarray); | |
1267 | ||
1268 | return st; | |
1269 | } | |
1270 | ||
1271 | struct supertype *dup_super(struct supertype *orig) | |
1272 | { | |
1273 | struct supertype *st; | |
1274 | ||
1275 | if (!orig) | |
1276 | return orig; | |
1277 | st = xcalloc(1, sizeof(*st)); | |
1278 | st->ss = orig->ss; | |
1279 | st->max_devs = orig->max_devs; | |
1280 | st->minor_version = orig->minor_version; | |
1281 | st->ignore_hw_compat = orig->ignore_hw_compat; | |
1282 | st->data_offset = orig->data_offset; | |
1283 | st->sb = NULL; | |
1284 | st->info = NULL; | |
1285 | return st; | |
1286 | } | |
1287 | ||
1288 | struct supertype *guess_super_type(int fd, enum guess_types guess_type) | |
1289 | { | |
1290 | /* try each load_super to find the best match, | |
1291 | * and return the best superswitch | |
1292 | */ | |
1293 | struct superswitch *ss; | |
1294 | struct supertype *st; | |
1295 | unsigned int besttime = 0; | |
1296 | int bestsuper = -1; | |
1297 | int i; | |
1298 | ||
1299 | st = xcalloc(1, sizeof(*st)); | |
1300 | st->container_devnm[0] = 0; | |
1301 | ||
1302 | for (i = 0; superlist[i]; i++) { | |
1303 | int rv; | |
1304 | ss = superlist[i]; | |
1305 | if (guess_type == guess_array && ss->add_to_super == NULL) | |
1306 | continue; | |
1307 | if (guess_type == guess_partitions && ss->add_to_super != NULL) | |
1308 | continue; | |
1309 | memset(st, 0, sizeof(*st)); | |
1310 | st->ignore_hw_compat = 1; | |
1311 | rv = ss->load_super(st, fd, NULL); | |
1312 | if (rv == 0) { | |
1313 | struct mdinfo info; | |
1314 | st->ss->getinfo_super(st, &info, NULL); | |
1315 | if (bestsuper == -1 || | |
1316 | besttime < info.array.ctime) { | |
1317 | bestsuper = i; | |
1318 | besttime = info.array.ctime; | |
1319 | } | |
1320 | ss->free_super(st); | |
1321 | } | |
1322 | } | |
1323 | if (bestsuper != -1) { | |
1324 | int rv; | |
1325 | memset(st, 0, sizeof(*st)); | |
1326 | st->ignore_hw_compat = 1; | |
1327 | rv = superlist[bestsuper]->load_super(st, fd, NULL); | |
1328 | if (rv == 0) { | |
1329 | superlist[bestsuper]->free_super(st); | |
1330 | return st; | |
1331 | } | |
1332 | } | |
1333 | free(st); | |
1334 | return NULL; | |
1335 | } | |
1336 | ||
1337 | /* Return size of device in bytes */ | |
1338 | int get_dev_size(int fd, char *dname, unsigned long long *sizep) | |
1339 | { | |
1340 | unsigned long long ldsize; | |
1341 | struct stat st; | |
1342 | ||
1343 | if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode)) | |
1344 | ldsize = (unsigned long long)st.st_size; | |
1345 | else | |
1346 | #ifdef BLKGETSIZE64 | |
1347 | if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0) | |
1348 | #endif | |
1349 | { | |
1350 | unsigned long dsize; | |
1351 | if (ioctl(fd, BLKGETSIZE, &dsize) == 0) { | |
1352 | ldsize = dsize; | |
1353 | ldsize <<= 9; | |
1354 | } else { | |
1355 | if (dname) | |
1356 | pr_err("Cannot get size of %s: %s\n", | |
1357 | dname, strerror(errno)); | |
1358 | return 0; | |
1359 | } | |
1360 | } | |
1361 | *sizep = ldsize; | |
1362 | return 1; | |
1363 | } | |
1364 | ||
1365 | /* Return sector size of device in bytes */ | |
1366 | int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep) | |
1367 | { | |
1368 | unsigned int sectsize; | |
1369 | ||
1370 | if (ioctl(fd, BLKSSZGET, §size) != 0) { | |
1371 | if (dname) | |
1372 | pr_err("Cannot get sector size of %s: %s\n", | |
1373 | dname, strerror(errno)); | |
1374 | return 0; | |
1375 | } | |
1376 | ||
1377 | *sectsizep = sectsize; | |
1378 | return 1; | |
1379 | } | |
1380 | ||
1381 | /* Return true if this can only be a container, not a member device. | |
1382 | * i.e. is and md device and size is zero | |
1383 | */ | |
1384 | int must_be_container(int fd) | |
1385 | { | |
1386 | struct mdinfo *mdi; | |
1387 | unsigned long long size; | |
1388 | ||
1389 | mdi = sysfs_read(fd, NULL, GET_VERSION); | |
1390 | if (!mdi) | |
1391 | return 0; | |
1392 | sysfs_free(mdi); | |
1393 | ||
1394 | if (get_dev_size(fd, NULL, &size) == 0) | |
1395 | return 1; | |
1396 | if (size == 0) | |
1397 | return 1; | |
1398 | return 0; | |
1399 | } | |
1400 | ||
1401 | /* Sets endofpart parameter to the last block used by the last GPT partition on the device. | |
1402 | * Returns: 1 if successful | |
1403 | * -1 for unknown partition type | |
1404 | * 0 for other errors | |
1405 | */ | |
1406 | static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart) | |
1407 | { | |
1408 | struct GPT gpt; | |
1409 | unsigned char empty_gpt_entry[16]= {0}; | |
1410 | struct GPT_part_entry *part; | |
1411 | char buf[512]; | |
1412 | unsigned long long curr_part_end; | |
1413 | unsigned all_partitions, entry_size; | |
1414 | unsigned part_nr; | |
1415 | unsigned int sector_size = 0; | |
1416 | ||
1417 | *endofpart = 0; | |
1418 | ||
1419 | BUILD_BUG_ON(sizeof(gpt) != 512); | |
1420 | /* skip protective MBR */ | |
1421 | if (!get_dev_sector_size(fd, NULL, §or_size)) | |
1422 | return 0; | |
1423 | lseek(fd, sector_size, SEEK_SET); | |
1424 | /* read GPT header */ | |
1425 | if (read(fd, &gpt, 512) != 512) | |
1426 | return 0; | |
1427 | ||
1428 | /* get the number of partition entries and the entry size */ | |
1429 | all_partitions = __le32_to_cpu(gpt.part_cnt); | |
1430 | entry_size = __le32_to_cpu(gpt.part_size); | |
1431 | ||
1432 | /* Check GPT signature*/ | |
1433 | if (gpt.magic != GPT_SIGNATURE_MAGIC) | |
1434 | return -1; | |
1435 | ||
1436 | /* sanity checks */ | |
1437 | if (all_partitions > 1024 || | |
1438 | entry_size > sizeof(buf)) | |
1439 | return -1; | |
1440 | ||
1441 | part = (struct GPT_part_entry *)buf; | |
1442 | ||
1443 | /* set offset to third block (GPT entries) */ | |
1444 | lseek(fd, sector_size*2, SEEK_SET); | |
1445 | for (part_nr = 0; part_nr < all_partitions; part_nr++) { | |
1446 | /* read partition entry */ | |
1447 | if (read(fd, buf, entry_size) != (ssize_t)entry_size) | |
1448 | return 0; | |
1449 | ||
1450 | /* is this valid partition? */ | |
1451 | if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) { | |
1452 | /* check the last lba for the current partition */ | |
1453 | curr_part_end = __le64_to_cpu(part->ending_lba); | |
1454 | if (curr_part_end > *endofpart) | |
1455 | *endofpart = curr_part_end; | |
1456 | } | |
1457 | ||
1458 | } | |
1459 | return 1; | |
1460 | } | |
1461 | ||
1462 | /* Sets endofpart parameter to the last block used by the last partition on the device. | |
1463 | * Returns: 1 if successful | |
1464 | * -1 for unknown partition type | |
1465 | * 0 for other errors | |
1466 | */ | |
1467 | static int get_last_partition_end(int fd, unsigned long long *endofpart) | |
1468 | { | |
1469 | struct MBR boot_sect; | |
1470 | unsigned long long curr_part_end; | |
1471 | unsigned part_nr; | |
1472 | unsigned int sector_size; | |
1473 | int retval = 0; | |
1474 | ||
1475 | *endofpart = 0; | |
1476 | ||
1477 | BUILD_BUG_ON(sizeof(boot_sect) != 512); | |
1478 | /* read MBR */ | |
1479 | lseek(fd, 0, 0); | |
1480 | if (read(fd, &boot_sect, 512) != 512) | |
1481 | goto abort; | |
1482 | ||
1483 | /* check MBP signature */ | |
1484 | if (boot_sect.magic == MBR_SIGNATURE_MAGIC) { | |
1485 | retval = 1; | |
1486 | /* found the correct signature */ | |
1487 | ||
1488 | for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) { | |
1489 | /* | |
1490 | * Have to make every access through boot_sect rather | |
1491 | * than using a pointer to the partition table (or an | |
1492 | * entry), since the entries are not properly aligned. | |
1493 | */ | |
1494 | ||
1495 | /* check for GPT type */ | |
1496 | if (boot_sect.parts[part_nr].part_type == | |
1497 | MBR_GPT_PARTITION_TYPE) { | |
1498 | retval = get_gpt_last_partition_end(fd, endofpart); | |
1499 | break; | |
1500 | } | |
1501 | /* check the last used lba for the current partition */ | |
1502 | curr_part_end = | |
1503 | __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) + | |
1504 | __le32_to_cpu(boot_sect.parts[part_nr].blocks_num); | |
1505 | if (curr_part_end > *endofpart) | |
1506 | *endofpart = curr_part_end; | |
1507 | } | |
1508 | } else { | |
1509 | /* Unknown partition table */ | |
1510 | retval = -1; | |
1511 | } | |
1512 | /* calculate number of 512-byte blocks */ | |
1513 | if (get_dev_sector_size(fd, NULL, §or_size)) | |
1514 | *endofpart *= (sector_size / 512); | |
1515 | abort: | |
1516 | return retval; | |
1517 | } | |
1518 | ||
1519 | int check_partitions(int fd, char *dname, unsigned long long freesize, | |
1520 | unsigned long long size) | |
1521 | { | |
1522 | /* | |
1523 | * Check where the last partition ends | |
1524 | */ | |
1525 | unsigned long long endofpart; | |
1526 | ||
1527 | if (get_last_partition_end(fd, &endofpart) > 0) { | |
1528 | /* There appears to be a partition table here */ | |
1529 | if (freesize == 0) { | |
1530 | /* partitions will not be visible in new device */ | |
1531 | pr_err("partition table exists on %s but will be lost or\n" | |
1532 | " meaningless after creating array\n", | |
1533 | dname); | |
1534 | return 1; | |
1535 | } else if (endofpart > freesize) { | |
1536 | /* last partition overlaps metadata */ | |
1537 | pr_err("metadata will over-write last partition on %s.\n", | |
1538 | dname); | |
1539 | return 1; | |
1540 | } else if (size && endofpart > size) { | |
1541 | /* partitions will be truncated in new device */ | |
1542 | pr_err("array size is too small to cover all partitions on %s.\n", | |
1543 | dname); | |
1544 | return 1; | |
1545 | } | |
1546 | } | |
1547 | return 0; | |
1548 | } | |
1549 | ||
1550 | int open_container(int fd) | |
1551 | { | |
1552 | /* 'fd' is a block device. Find out if it is in use | |
1553 | * by a container, and return an open fd on that container. | |
1554 | */ | |
1555 | char path[288]; | |
1556 | char *e; | |
1557 | DIR *dir; | |
1558 | struct dirent *de; | |
1559 | int dfd, n; | |
1560 | char buf[200]; | |
1561 | int major, minor; | |
1562 | struct stat st; | |
1563 | ||
1564 | if (fstat(fd, &st) != 0) | |
1565 | return -1; | |
1566 | sprintf(path, "/sys/dev/block/%d:%d/holders", | |
1567 | (int)major(st.st_rdev), (int)minor(st.st_rdev)); | |
1568 | e = path + strlen(path); | |
1569 | ||
1570 | dir = opendir(path); | |
1571 | if (!dir) | |
1572 | return -1; | |
1573 | while ((de = readdir(dir))) { | |
1574 | if (de->d_ino == 0) | |
1575 | continue; | |
1576 | if (de->d_name[0] == '.') | |
1577 | continue; | |
1578 | /* Need to make sure it is a container and not a volume */ | |
1579 | sprintf(e, "/%s/md/metadata_version", de->d_name); | |
1580 | dfd = open(path, O_RDONLY); | |
1581 | if (dfd < 0) | |
1582 | continue; | |
1583 | n = read(dfd, buf, sizeof(buf)); | |
1584 | close(dfd); | |
1585 | if (n <= 0 || (unsigned)n >= sizeof(buf)) | |
1586 | continue; | |
1587 | buf[n] = 0; | |
1588 | if (strncmp(buf, "external", 8) != 0 || | |
1589 | n < 10 || | |
1590 | buf[9] == '/') | |
1591 | continue; | |
1592 | sprintf(e, "/%s/dev", de->d_name); | |
1593 | dfd = open(path, O_RDONLY); | |
1594 | if (dfd < 0) | |
1595 | continue; | |
1596 | n = read(dfd, buf, sizeof(buf)); | |
1597 | close(dfd); | |
1598 | if (n <= 0 || (unsigned)n >= sizeof(buf)) | |
1599 | continue; | |
1600 | buf[n] = 0; | |
1601 | if (sscanf(buf, "%d:%d", &major, &minor) != 2) | |
1602 | continue; | |
1603 | sprintf(buf, "%d:%d", major, minor); | |
1604 | dfd = dev_open(buf, O_RDONLY); | |
1605 | if (dfd >= 0) { | |
1606 | closedir(dir); | |
1607 | return dfd; | |
1608 | } | |
1609 | } | |
1610 | closedir(dir); | |
1611 | return -1; | |
1612 | } | |
1613 | ||
1614 | struct superswitch *version_to_superswitch(char *vers) | |
1615 | { | |
1616 | int i; | |
1617 | ||
1618 | for (i = 0; superlist[i]; i++) { | |
1619 | struct superswitch *ss = superlist[i]; | |
1620 | ||
1621 | if (strcmp(vers, ss->name) == 0) | |
1622 | return ss; | |
1623 | } | |
1624 | ||
1625 | return NULL; | |
1626 | } | |
1627 | ||
1628 | int metadata_container_matches(char *metadata, char *devnm) | |
1629 | { | |
1630 | /* Check if 'devnm' is the container named in 'metadata' | |
1631 | * which is | |
1632 | * /containername/componentname or | |
1633 | * -containername/componentname | |
1634 | */ | |
1635 | int l; | |
1636 | if (*metadata != '/' && *metadata != '-') | |
1637 | return 0; | |
1638 | l = strlen(devnm); | |
1639 | if (strncmp(metadata+1, devnm, l) != 0) | |
1640 | return 0; | |
1641 | if (metadata[l+1] != '/') | |
1642 | return 0; | |
1643 | return 1; | |
1644 | } | |
1645 | ||
1646 | int metadata_subdev_matches(char *metadata, char *devnm) | |
1647 | { | |
1648 | /* Check if 'devnm' is the subdev named in 'metadata' | |
1649 | * which is | |
1650 | * /containername/subdev or | |
1651 | * -containername/subdev | |
1652 | */ | |
1653 | char *sl; | |
1654 | if (*metadata != '/' && *metadata != '-') | |
1655 | return 0; | |
1656 | sl = strchr(metadata+1, '/'); | |
1657 | if (!sl) | |
1658 | return 0; | |
1659 | if (strcmp(sl+1, devnm) == 0) | |
1660 | return 1; | |
1661 | return 0; | |
1662 | } | |
1663 | ||
1664 | int is_container_member(struct mdstat_ent *mdstat, char *container) | |
1665 | { | |
1666 | if (mdstat->metadata_version == NULL || | |
1667 | strncmp(mdstat->metadata_version, "external:", 9) != 0 || | |
1668 | !metadata_container_matches(mdstat->metadata_version+9, container)) | |
1669 | return 0; | |
1670 | ||
1671 | return 1; | |
1672 | } | |
1673 | ||
1674 | int is_subarray_active(char *subarray, char *container) | |
1675 | { | |
1676 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
1677 | struct mdstat_ent *ent; | |
1678 | ||
1679 | for (ent = mdstat; ent; ent = ent->next) | |
1680 | if (is_container_member(ent, container)) | |
1681 | if (strcmp(to_subarray(ent, container), subarray) == 0) | |
1682 | break; | |
1683 | ||
1684 | free_mdstat(mdstat); | |
1685 | ||
1686 | return ent != NULL; | |
1687 | } | |
1688 | ||
1689 | /* open_subarray - opens a subarray in a container | |
1690 | * @dev: container device name | |
1691 | * @st: empty supertype | |
1692 | * @quiet: block reporting errors flag | |
1693 | * | |
1694 | * On success returns an fd to a container and fills in *st | |
1695 | */ | |
1696 | int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet) | |
1697 | { | |
1698 | struct mdinfo *mdi; | |
1699 | struct mdinfo *info; | |
1700 | int fd, err = 1; | |
1701 | char *_devnm; | |
1702 | ||
1703 | fd = open(dev, O_RDWR|O_EXCL); | |
1704 | if (fd < 0) { | |
1705 | if (!quiet) | |
1706 | pr_err("Couldn't open %s, aborting\n", | |
1707 | dev); | |
1708 | return -1; | |
1709 | } | |
1710 | ||
1711 | _devnm = fd2devnm(fd); | |
1712 | if (_devnm == NULL) { | |
1713 | if (!quiet) | |
1714 | pr_err("Failed to determine device number for %s\n", | |
1715 | dev); | |
1716 | goto close_fd; | |
1717 | } | |
1718 | strcpy(st->devnm, _devnm); | |
1719 | ||
1720 | mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL); | |
1721 | if (!mdi) { | |
1722 | if (!quiet) | |
1723 | pr_err("Failed to read sysfs for %s\n", | |
1724 | dev); | |
1725 | goto close_fd; | |
1726 | } | |
1727 | ||
1728 | if (mdi->array.level != UnSet) { | |
1729 | if (!quiet) | |
1730 | pr_err("%s is not a container\n", dev); | |
1731 | goto free_sysfs; | |
1732 | } | |
1733 | ||
1734 | st->ss = version_to_superswitch(mdi->text_version); | |
1735 | if (!st->ss) { | |
1736 | if (!quiet) | |
1737 | pr_err("Operation not supported for %s metadata\n", | |
1738 | mdi->text_version); | |
1739 | goto free_sysfs; | |
1740 | } | |
1741 | ||
1742 | if (st->devnm[0] == 0) { | |
1743 | if (!quiet) | |
1744 | pr_err("Failed to allocate device name\n"); | |
1745 | goto free_sysfs; | |
1746 | } | |
1747 | ||
1748 | if (!st->ss->load_container) { | |
1749 | if (!quiet) | |
1750 | pr_err("%s is not a container\n", dev); | |
1751 | goto free_sysfs; | |
1752 | } | |
1753 | ||
1754 | if (st->ss->load_container(st, fd, NULL)) { | |
1755 | if (!quiet) | |
1756 | pr_err("Failed to load metadata for %s\n", | |
1757 | dev); | |
1758 | goto free_sysfs; | |
1759 | } | |
1760 | ||
1761 | info = st->ss->container_content(st, subarray); | |
1762 | if (!info) { | |
1763 | if (!quiet) | |
1764 | pr_err("Failed to find subarray-%s in %s\n", | |
1765 | subarray, dev); | |
1766 | goto free_super; | |
1767 | } | |
1768 | free(info); | |
1769 | ||
1770 | err = 0; | |
1771 | ||
1772 | free_super: | |
1773 | if (err) | |
1774 | st->ss->free_super(st); | |
1775 | free_sysfs: | |
1776 | sysfs_free(mdi); | |
1777 | close_fd: | |
1778 | if (err) | |
1779 | close(fd); | |
1780 | ||
1781 | if (err) | |
1782 | return -1; | |
1783 | else | |
1784 | return fd; | |
1785 | } | |
1786 | ||
1787 | int add_disk(int mdfd, struct supertype *st, | |
1788 | struct mdinfo *sra, struct mdinfo *info) | |
1789 | { | |
1790 | /* Add a device to an array, in one of 2 ways. */ | |
1791 | int rv; | |
1792 | ||
1793 | if (st->ss->external) { | |
1794 | if (info->disk.state & (1<<MD_DISK_SYNC)) | |
1795 | info->recovery_start = MaxSector; | |
1796 | else | |
1797 | info->recovery_start = 0; | |
1798 | rv = sysfs_add_disk(sra, info, 0); | |
1799 | if (! rv) { | |
1800 | struct mdinfo *sd2; | |
1801 | for (sd2 = sra->devs; sd2; sd2=sd2->next) | |
1802 | if (sd2 == info) | |
1803 | break; | |
1804 | if (sd2 == NULL) { | |
1805 | sd2 = xmalloc(sizeof(*sd2)); | |
1806 | *sd2 = *info; | |
1807 | sd2->next = sra->devs; | |
1808 | sra->devs = sd2; | |
1809 | } | |
1810 | } | |
1811 | } else | |
1812 | rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk); | |
1813 | return rv; | |
1814 | } | |
1815 | ||
1816 | int remove_disk(int mdfd, struct supertype *st, | |
1817 | struct mdinfo *sra, struct mdinfo *info) | |
1818 | { | |
1819 | int rv; | |
1820 | ||
1821 | /* Remove the disk given by 'info' from the array */ | |
1822 | if (st->ss->external) | |
1823 | rv = sysfs_set_str(sra, info, "slot", STR_COMMON_NONE); | |
1824 | else | |
1825 | rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major, | |
1826 | info->disk.minor)); | |
1827 | return rv; | |
1828 | } | |
1829 | ||
1830 | int hot_remove_disk(int mdfd, unsigned long dev, int force) | |
1831 | { | |
1832 | int cnt = force ? 500 : 5; | |
1833 | int ret; | |
1834 | ||
1835 | /* HOT_REMOVE_DISK can fail with EBUSY if there are | |
1836 | * outstanding IO requests to the device. | |
1837 | * In this case, it can be helpful to wait a little while, | |
1838 | * up to 5 seconds if 'force' is set, or 50 msec if not. | |
1839 | */ | |
1840 | while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 && | |
1841 | errno == EBUSY && | |
1842 | cnt-- > 0) | |
1843 | sleep_for(0, MSEC_TO_NSEC(10), true); | |
1844 | ||
1845 | return ret; | |
1846 | } | |
1847 | ||
1848 | int sys_hot_remove_disk(int statefd, int force) | |
1849 | { | |
1850 | int cnt = force ? 500 : 5; | |
1851 | int ret; | |
1852 | ||
1853 | while ((ret = write(statefd, "remove", 6)) == -1 && | |
1854 | errno == EBUSY && | |
1855 | cnt-- > 0) | |
1856 | sleep_for(0, MSEC_TO_NSEC(10), true); | |
1857 | return ret == 6 ? 0 : -1; | |
1858 | } | |
1859 | ||
1860 | int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info) | |
1861 | { | |
1862 | /* Initialise kernel's knowledge of array. | |
1863 | * This varies between externally managed arrays | |
1864 | * and older kernels | |
1865 | */ | |
1866 | mdu_array_info_t inf; | |
1867 | int rv; | |
1868 | ||
1869 | if (st->ss->external) | |
1870 | return sysfs_set_array(info); | |
1871 | memset(&inf, 0, sizeof(inf)); | |
1872 | inf.major_version = info->array.major_version; | |
1873 | inf.minor_version = info->array.minor_version; | |
1874 | rv = md_set_array_info(mdfd, &inf); | |
1875 | ||
1876 | return rv; | |
1877 | } | |
1878 | ||
1879 | unsigned long long min_recovery_start(struct mdinfo *array) | |
1880 | { | |
1881 | /* find the minimum recovery_start in an array for metadata | |
1882 | * formats that only record per-array recovery progress instead | |
1883 | * of per-device | |
1884 | */ | |
1885 | unsigned long long recovery_start = MaxSector; | |
1886 | struct mdinfo *d; | |
1887 | ||
1888 | for (d = array->devs; d; d = d->next) | |
1889 | recovery_start = min(recovery_start, d->recovery_start); | |
1890 | ||
1891 | return recovery_start; | |
1892 | } | |
1893 | ||
1894 | int mdmon_pid(char *devnm) | |
1895 | { | |
1896 | char path[100]; | |
1897 | char pid[10]; | |
1898 | int fd; | |
1899 | int n; | |
1900 | ||
1901 | sprintf(path, "%s/%s.pid", MDMON_DIR, devnm); | |
1902 | ||
1903 | fd = open(path, O_RDONLY | O_NOATIME, 0); | |
1904 | ||
1905 | if (fd < 0) | |
1906 | return -1; | |
1907 | n = read(fd, pid, 9); | |
1908 | close(fd); | |
1909 | if (n <= 0) | |
1910 | return -1; | |
1911 | return atoi(pid); | |
1912 | } | |
1913 | ||
1914 | int mdmon_running(char *devnm) | |
1915 | { | |
1916 | int pid = mdmon_pid(devnm); | |
1917 | if (pid <= 0) | |
1918 | return 0; | |
1919 | if (kill(pid, 0) == 0) | |
1920 | return 1; | |
1921 | return 0; | |
1922 | } | |
1923 | ||
1924 | int start_mdmon(char *devnm) | |
1925 | { | |
1926 | int i; | |
1927 | int len; | |
1928 | pid_t pid; | |
1929 | int status; | |
1930 | char *prefix = in_initrd() ? "initrd-" : ""; | |
1931 | char pathbuf[1024]; | |
1932 | char *paths[4] = { | |
1933 | pathbuf, | |
1934 | BINDIR "/mdmon", | |
1935 | "./mdmon", | |
1936 | NULL | |
1937 | }; | |
1938 | ||
1939 | if (check_env("MDADM_NO_MDMON")) | |
1940 | return 0; | |
1941 | if (continue_via_systemd(devnm, MDMON_SERVICE, prefix)) | |
1942 | return 0; | |
1943 | ||
1944 | /* That failed, try running mdmon directly */ | |
1945 | len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1); | |
1946 | if (len > 0) { | |
1947 | char *sl; | |
1948 | pathbuf[len] = 0; | |
1949 | sl = strrchr(pathbuf, '/'); | |
1950 | if (sl) | |
1951 | sl++; | |
1952 | else | |
1953 | sl = pathbuf; | |
1954 | strcpy(sl, "mdmon"); | |
1955 | } else | |
1956 | pathbuf[0] = '\0'; | |
1957 | ||
1958 | switch(fork()) { | |
1959 | case 0: | |
1960 | manage_fork_fds(1); | |
1961 | for (i = 0; paths[i]; i++) | |
1962 | if (paths[i][0]) { | |
1963 | execl(paths[i], paths[i], | |
1964 | devnm, NULL); | |
1965 | } | |
1966 | exit(1); | |
1967 | case -1: pr_err("cannot run mdmon. Array remains readonly\n"); | |
1968 | return -1; | |
1969 | default: /* parent - good */ | |
1970 | pid = wait(&status); | |
1971 | if (pid < 0 || status != 0) { | |
1972 | pr_err("failed to launch mdmon. Array remains readonly\n"); | |
1973 | return -1; | |
1974 | } | |
1975 | } | |
1976 | return 0; | |
1977 | } | |
1978 | ||
1979 | __u32 random32(void) | |
1980 | { | |
1981 | __u32 rv; | |
1982 | int rfd = open("/dev/urandom", O_RDONLY); | |
1983 | if (rfd < 0 || read(rfd, &rv, 4) != 4) | |
1984 | rv = random(); | |
1985 | if (rfd >= 0) | |
1986 | close(rfd); | |
1987 | return rv; | |
1988 | } | |
1989 | ||
1990 | void random_uuid(__u8 *buf) | |
1991 | { | |
1992 | int fd, i, len; | |
1993 | __u32 r[4]; | |
1994 | ||
1995 | fd = open("/dev/urandom", O_RDONLY); | |
1996 | if (fd < 0) | |
1997 | goto use_random; | |
1998 | len = read(fd, buf, 16); | |
1999 | close(fd); | |
2000 | if (len != 16) | |
2001 | goto use_random; | |
2002 | ||
2003 | return; | |
2004 | ||
2005 | use_random: | |
2006 | for (i = 0; i < 4; i++) | |
2007 | r[i] = random(); | |
2008 | memcpy(buf, r, 16); | |
2009 | } | |
2010 | ||
2011 | int flush_metadata_updates(struct supertype *st) | |
2012 | { | |
2013 | int sfd; | |
2014 | if (!st->updates) { | |
2015 | st->update_tail = NULL; | |
2016 | return -1; | |
2017 | } | |
2018 | ||
2019 | sfd = connect_monitor(st->container_devnm); | |
2020 | if (sfd < 0) | |
2021 | return -1; | |
2022 | ||
2023 | while (st->updates) { | |
2024 | struct metadata_update *mu = st->updates; | |
2025 | st->updates = mu->next; | |
2026 | ||
2027 | send_message(sfd, mu, 0); | |
2028 | wait_reply(sfd, 0); | |
2029 | free(mu->buf); | |
2030 | free(mu); | |
2031 | } | |
2032 | ack(sfd, 0); | |
2033 | wait_reply(sfd, 0); | |
2034 | close(sfd); | |
2035 | st->update_tail = NULL; | |
2036 | return 0; | |
2037 | } | |
2038 | ||
2039 | void append_metadata_update(struct supertype *st, void *buf, int len) | |
2040 | { | |
2041 | ||
2042 | struct metadata_update *mu = xmalloc(sizeof(*mu)); | |
2043 | ||
2044 | mu->buf = buf; | |
2045 | mu->len = len; | |
2046 | mu->space = NULL; | |
2047 | mu->space_list = NULL; | |
2048 | mu->next = NULL; | |
2049 | *st->update_tail = mu; | |
2050 | st->update_tail = &mu->next; | |
2051 | } | |
2052 | ||
2053 | #ifdef __TINYC__ | |
2054 | /* tinyc doesn't optimize this check in ioctl.h out ... */ | |
2055 | unsigned int __invalid_size_argument_for_IOC = 0; | |
2056 | #endif | |
2057 | ||
2058 | /** | |
2059 | * disk_fd_matches_criteria() - check if device matches spare criteria. | |
2060 | * @st: supertype, not NULL. | |
2061 | * @disk_fd: file descriptor of the disk. | |
2062 | * @sc: criteria to test. | |
2063 | * | |
2064 | * Return: true if disk matches criteria, false otherwise. | |
2065 | */ | |
2066 | bool disk_fd_matches_criteria(struct supertype *st, int disk_fd, struct spare_criteria *sc) | |
2067 | { | |
2068 | unsigned int dev_sector_size = 0; | |
2069 | unsigned long long dev_size = 0; | |
2070 | ||
2071 | if (!sc->criteria_set) | |
2072 | return true; | |
2073 | ||
2074 | if (!get_dev_size(disk_fd, NULL, &dev_size) || dev_size < sc->min_size) | |
2075 | return false; | |
2076 | ||
2077 | if (!get_dev_sector_size(disk_fd, NULL, &dev_sector_size) || | |
2078 | sc->sector_size != dev_sector_size) | |
2079 | return false; | |
2080 | ||
2081 | if (drive_test_and_add_policies(st, &sc->pols, disk_fd, 0)) | |
2082 | return false; | |
2083 | ||
2084 | return true; | |
2085 | } | |
2086 | ||
2087 | /** | |
2088 | * devid_matches_criteria() - check if device referenced by devid matches spare criteria. | |
2089 | * @st: supertype, not NULL. | |
2090 | * @devid: devid of the device to check. | |
2091 | * @sc: criteria to test. | |
2092 | * | |
2093 | * Return: true if disk matches criteria, false otherwise. | |
2094 | */ | |
2095 | bool devid_matches_criteria(struct supertype *st, dev_t devid, struct spare_criteria *sc) | |
2096 | { | |
2097 | char buf[NAME_MAX]; | |
2098 | bool ret; | |
2099 | int fd; | |
2100 | ||
2101 | if (!sc->criteria_set) | |
2102 | return true; | |
2103 | ||
2104 | snprintf(buf, NAME_MAX, "%d:%d", major(devid), minor(devid)); | |
2105 | ||
2106 | fd = dev_open(buf, O_RDONLY); | |
2107 | if (!is_fd_valid(fd)) | |
2108 | return false; | |
2109 | ||
2110 | /* Error code inherited */ | |
2111 | ret = disk_fd_matches_criteria(st, fd, sc); | |
2112 | ||
2113 | close(fd); | |
2114 | return ret; | |
2115 | } | |
2116 | ||
2117 | /* Pick all spares matching given criteria from a container | |
2118 | * if min_size == 0 do not check size | |
2119 | * if domlist == NULL do not check domains | |
2120 | * if spare_group given add it to domains of each spare | |
2121 | * metadata allows to test domains using metadata of destination array */ | |
2122 | struct mdinfo *container_choose_spares(struct supertype *st, | |
2123 | struct spare_criteria *criteria, | |
2124 | struct domainlist *domlist, | |
2125 | char *spare_group, | |
2126 | const char *metadata, int get_one) | |
2127 | { | |
2128 | struct mdinfo *d, **dp, *disks = NULL; | |
2129 | ||
2130 | /* get list of all disks in container */ | |
2131 | if (st->ss->getinfo_super_disks) | |
2132 | disks = st->ss->getinfo_super_disks(st); | |
2133 | ||
2134 | if (!disks) | |
2135 | return disks; | |
2136 | /* find spare devices on the list */ | |
2137 | dp = &disks->devs; | |
2138 | disks->array.spare_disks = 0; | |
2139 | while (*dp) { | |
2140 | bool found = false; | |
2141 | ||
2142 | d = *dp; | |
2143 | if (d->disk.state == 0) { | |
2144 | dev_t dev = makedev(d->disk.major,d->disk.minor); | |
2145 | ||
2146 | found = devid_matches_criteria(st, dev, criteria); | |
2147 | ||
2148 | /* check if domain matches */ | |
2149 | if (found && domlist) { | |
2150 | struct dev_policy *pol = devid_policy(dev); | |
2151 | if (spare_group) | |
2152 | pol_add(&pol, pol_domain, | |
2153 | spare_group, NULL); | |
2154 | if (domain_test(domlist, pol, metadata) != 1) | |
2155 | found = false; | |
2156 | ||
2157 | dev_policy_free(pol); | |
2158 | } | |
2159 | } | |
2160 | if (found) { | |
2161 | dp = &d->next; | |
2162 | disks->array.spare_disks++; | |
2163 | if (get_one) { | |
2164 | sysfs_free(*dp); | |
2165 | d->next = NULL; | |
2166 | } | |
2167 | } else { | |
2168 | *dp = d->next; | |
2169 | d->next = NULL; | |
2170 | sysfs_free(d); | |
2171 | } | |
2172 | } | |
2173 | return disks; | |
2174 | } | |
2175 | ||
2176 | /* Checks if paths point to the same device | |
2177 | * Returns 0 if they do. | |
2178 | * Returns 1 if they don't. | |
2179 | * Returns -1 if something went wrong, | |
2180 | * e.g. paths are empty or the files | |
2181 | * they point to don't exist */ | |
2182 | int compare_paths (char* path1, char* path2) | |
2183 | { | |
2184 | struct stat st1,st2; | |
2185 | ||
2186 | if (path1 == NULL || path2 == NULL) | |
2187 | return -1; | |
2188 | if (stat(path1,&st1) != 0) | |
2189 | return -1; | |
2190 | if (stat(path2,&st2) != 0) | |
2191 | return -1; | |
2192 | if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev)) | |
2193 | return 0; | |
2194 | return 1; | |
2195 | } | |
2196 | ||
2197 | /* Make sure we can open as many devices as needed */ | |
2198 | void enable_fds(int devices) | |
2199 | { | |
2200 | unsigned int fds = 20 + devices; | |
2201 | struct rlimit lim; | |
2202 | if (getrlimit(RLIMIT_NOFILE, &lim) != 0 || lim.rlim_cur >= fds) | |
2203 | return; | |
2204 | if (lim.rlim_max < fds) | |
2205 | lim.rlim_max = fds; | |
2206 | lim.rlim_cur = fds; | |
2207 | setrlimit(RLIMIT_NOFILE, &lim); | |
2208 | } | |
2209 | ||
2210 | /* Close all opened descriptors if needed and redirect | |
2211 | * streams to /dev/null. | |
2212 | * For debug purposed, leave STDOUT and STDERR untouched | |
2213 | * Returns: | |
2214 | * 1- if any error occurred | |
2215 | * 0- otherwise | |
2216 | */ | |
2217 | void manage_fork_fds(int close_all) | |
2218 | { | |
2219 | DIR *dir; | |
2220 | struct dirent *dirent; | |
2221 | ||
2222 | close(0); | |
2223 | open("/dev/null", O_RDWR); | |
2224 | ||
2225 | #ifndef DEBUG | |
2226 | dup2(0, 1); | |
2227 | dup2(0, 2); | |
2228 | #endif | |
2229 | ||
2230 | if (close_all == 0) | |
2231 | return; | |
2232 | ||
2233 | dir = opendir("/proc/self/fd"); | |
2234 | if (!dir) { | |
2235 | pr_err("Cannot open /proc/self/fd directory.\n"); | |
2236 | return; | |
2237 | } | |
2238 | for (dirent = readdir(dir); dirent; dirent = readdir(dir)) { | |
2239 | int fd = -1; | |
2240 | ||
2241 | if ((strcmp(dirent->d_name, ".") == 0) || | |
2242 | (strcmp(dirent->d_name, "..")) == 0) | |
2243 | continue; | |
2244 | ||
2245 | fd = strtol(dirent->d_name, NULL, 10); | |
2246 | if (fd > 2) | |
2247 | close(fd); | |
2248 | } | |
2249 | } | |
2250 | ||
2251 | /* In a systemd/udev world, it is best to get systemd to | |
2252 | * run daemon rather than running in the background. | |
2253 | * Returns: | |
2254 | * 1- if systemd service has been started | |
2255 | * 0- otherwise | |
2256 | */ | |
2257 | int continue_via_systemd(char *devnm, char *service_name, char *prefix) | |
2258 | { | |
2259 | int pid, status; | |
2260 | char pathbuf[1024]; | |
2261 | ||
2262 | /* Simply return that service cannot be started */ | |
2263 | if (check_env("MDADM_NO_SYSTEMCTL")) | |
2264 | return 0; | |
2265 | switch (fork()) { | |
2266 | case 0: | |
2267 | manage_fork_fds(1); | |
2268 | snprintf(pathbuf, sizeof(pathbuf), | |
2269 | "%s@%s%s.service", service_name, prefix ?: "", devnm); | |
2270 | status = execl("/usr/bin/systemctl", "systemctl", "restart", | |
2271 | pathbuf, NULL); | |
2272 | status = execl("/bin/systemctl", "systemctl", "restart", | |
2273 | pathbuf, NULL); | |
2274 | exit(1); | |
2275 | case -1: /* Just do it ourselves. */ | |
2276 | break; | |
2277 | default: /* parent - good */ | |
2278 | pid = wait(&status); | |
2279 | if (pid >= 0 && status == 0) | |
2280 | return 1; | |
2281 | } | |
2282 | return 0; | |
2283 | } | |
2284 | ||
2285 | int in_initrd(void) | |
2286 | { | |
2287 | return access("/etc/initrd-release", F_OK) >= 0; | |
2288 | } | |
2289 | ||
2290 | void reopen_mddev(int mdfd) | |
2291 | { | |
2292 | /* Re-open without any O_EXCL, but keep | |
2293 | * the same fd | |
2294 | */ | |
2295 | char *devnm; | |
2296 | int fd; | |
2297 | devnm = fd2devnm(mdfd); | |
2298 | close(mdfd); | |
2299 | fd = open_dev(devnm); | |
2300 | if (fd >= 0 && fd != mdfd) | |
2301 | dup2(fd, mdfd); | |
2302 | } | |
2303 | ||
2304 | static struct cmap_hooks *cmap_hooks = NULL; | |
2305 | static int is_cmap_hooks_ready = 0; | |
2306 | ||
2307 | void set_cmap_hooks(void) | |
2308 | { | |
2309 | cmap_hooks = xmalloc(sizeof(struct cmap_hooks)); | |
2310 | cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL); | |
2311 | if (!cmap_hooks->cmap_handle) | |
2312 | return; | |
2313 | ||
2314 | cmap_hooks->initialize = | |
2315 | dlsym(cmap_hooks->cmap_handle, "cmap_initialize"); | |
2316 | cmap_hooks->get_string = | |
2317 | dlsym(cmap_hooks->cmap_handle, "cmap_get_string"); | |
2318 | cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize"); | |
2319 | ||
2320 | if (!cmap_hooks->initialize || !cmap_hooks->get_string || | |
2321 | !cmap_hooks->finalize) | |
2322 | dlclose(cmap_hooks->cmap_handle); | |
2323 | else | |
2324 | is_cmap_hooks_ready = 1; | |
2325 | } | |
2326 | ||
2327 | int get_cluster_name(char **cluster_name) | |
2328 | { | |
2329 | int rv = -1; | |
2330 | cmap_handle_t handle; | |
2331 | ||
2332 | if (!is_cmap_hooks_ready) | |
2333 | return rv; | |
2334 | ||
2335 | rv = cmap_hooks->initialize(&handle); | |
2336 | if (rv != CS_OK) | |
2337 | goto out; | |
2338 | ||
2339 | rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name); | |
2340 | if (rv != CS_OK) { | |
2341 | free(*cluster_name); | |
2342 | rv = -1; | |
2343 | goto name_err; | |
2344 | } | |
2345 | ||
2346 | rv = 0; | |
2347 | name_err: | |
2348 | cmap_hooks->finalize(handle); | |
2349 | out: | |
2350 | return rv; | |
2351 | } | |
2352 | ||
2353 | void set_dlm_hooks(void) | |
2354 | { | |
2355 | dlm_hooks = xmalloc(sizeof(struct dlm_hooks)); | |
2356 | dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL); | |
2357 | if (!dlm_hooks->dlm_handle) | |
2358 | return; | |
2359 | ||
2360 | dlm_hooks->open_lockspace = | |
2361 | dlsym(dlm_hooks->dlm_handle, "dlm_open_lockspace"); | |
2362 | dlm_hooks->create_lockspace = | |
2363 | dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace"); | |
2364 | dlm_hooks->release_lockspace = | |
2365 | dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace"); | |
2366 | dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock"); | |
2367 | dlm_hooks->ls_unlock_wait = | |
2368 | dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock_wait"); | |
2369 | dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd"); | |
2370 | dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch"); | |
2371 | ||
2372 | if (!dlm_hooks->open_lockspace || !dlm_hooks->create_lockspace || | |
2373 | !dlm_hooks->ls_lock || !dlm_hooks->ls_unlock_wait || | |
2374 | !dlm_hooks->release_lockspace || !dlm_hooks->ls_get_fd || | |
2375 | !dlm_hooks->dispatch) | |
2376 | dlclose(dlm_hooks->dlm_handle); | |
2377 | else | |
2378 | is_dlm_hooks_ready = 1; | |
2379 | } | |
2380 | ||
2381 | void set_hooks(void) | |
2382 | { | |
2383 | set_dlm_hooks(); | |
2384 | set_cmap_hooks(); | |
2385 | } | |
2386 | ||
2387 | int zero_disk_range(int fd, unsigned long long sector, size_t count) | |
2388 | { | |
2389 | int ret = 0; | |
2390 | int fd_zero; | |
2391 | void *addr = NULL; | |
2392 | size_t written = 0; | |
2393 | size_t len = count * 512; | |
2394 | ssize_t n; | |
2395 | ||
2396 | fd_zero = open("/dev/zero", O_RDONLY); | |
2397 | if (fd_zero < 0) { | |
2398 | pr_err("Cannot open /dev/zero\n"); | |
2399 | return -1; | |
2400 | } | |
2401 | ||
2402 | if (lseek64(fd, sector * 512, SEEK_SET) < 0) { | |
2403 | ret = -errno; | |
2404 | pr_err("Failed to seek offset for zeroing\n"); | |
2405 | goto out; | |
2406 | } | |
2407 | ||
2408 | addr = mmap(NULL, len, PROT_READ, MAP_PRIVATE, fd_zero, 0); | |
2409 | ||
2410 | if (addr == MAP_FAILED) { | |
2411 | ret = -errno; | |
2412 | pr_err("Mapping /dev/zero failed\n"); | |
2413 | goto out; | |
2414 | } | |
2415 | ||
2416 | do { | |
2417 | n = write(fd, addr + written, len - written); | |
2418 | if (n < 0) { | |
2419 | if (errno == EINTR) | |
2420 | continue; | |
2421 | ret = -errno; | |
2422 | pr_err("Zeroing disk range failed\n"); | |
2423 | break; | |
2424 | } | |
2425 | written += n; | |
2426 | } while (written != len); | |
2427 | ||
2428 | munmap(addr, len); | |
2429 | ||
2430 | out: | |
2431 | close(fd_zero); | |
2432 | return ret; | |
2433 | } | |
2434 | ||
2435 | /** | |
2436 | * sleep_for() - Sleeps for specified time. | |
2437 | * @sec: Seconds to sleep for. | |
2438 | * @nsec: Nanoseconds to sleep for, has to be less than one second. | |
2439 | * @wake_after_interrupt: If set, wake up if interrupted. | |
2440 | * | |
2441 | * Function immediately returns if error different than EINTR occurs. | |
2442 | */ | |
2443 | void sleep_for(unsigned int sec, long nsec, bool wake_after_interrupt) | |
2444 | { | |
2445 | struct timespec delay = {.tv_sec = sec, .tv_nsec = nsec}; | |
2446 | ||
2447 | assert(nsec < MSEC_TO_NSEC(1000)); | |
2448 | ||
2449 | do { | |
2450 | errno = 0; | |
2451 | nanosleep(&delay, &delay); | |
2452 | if (errno != 0 && errno != EINTR) { | |
2453 | pr_err("Error sleeping for %us %ldns: %s\n", sec, nsec, strerror(errno)); | |
2454 | return; | |
2455 | } | |
2456 | } while (!wake_after_interrupt && errno == EINTR); | |
2457 | } | |
2458 | ||
2459 | /* is_directory() - Checks if directory provided by path is indeed a regular directory. | |
2460 | * @path: directory path to be checked | |
2461 | * | |
2462 | * Doesn't accept symlinks. | |
2463 | * | |
2464 | * Return: true if is a directory, false if not | |
2465 | */ | |
2466 | bool is_directory(const char *path) | |
2467 | { | |
2468 | struct stat st; | |
2469 | ||
2470 | if (lstat(path, &st) != 0) { | |
2471 | pr_err("%s: %s\n", strerror(errno), path); | |
2472 | return false; | |
2473 | } | |
2474 | ||
2475 | if (!S_ISDIR(st.st_mode)) | |
2476 | return false; | |
2477 | ||
2478 | return true; | |
2479 | } | |
2480 | ||
2481 | /* | |
2482 | * is_file() - Checks if file provided by path is indeed a regular file. | |
2483 | * @path: file path to be checked | |
2484 | * | |
2485 | * Doesn't accept symlinks. | |
2486 | * | |
2487 | * Return: true if is a file, false if not | |
2488 | */ | |
2489 | bool is_file(const char *path) | |
2490 | { | |
2491 | struct stat st; | |
2492 | ||
2493 | if (lstat(path, &st) != 0) { | |
2494 | pr_err("%s: %s\n", strerror(errno), path); | |
2495 | return false; | |
2496 | } | |
2497 | ||
2498 | if (!S_ISREG(st.st_mode)) | |
2499 | return false; | |
2500 | ||
2501 | return true; | |
2502 | } |