Add one sanity check for missing device
[thirdparty/mdadm.git] / util.c
CommitLineData
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>
a931db9e 38#include <signal.h>
7716570e 39#include <dlfcn.h>
7716570e 40
0a816ef9
NB
41
42/*
43 * following taken from linux/blkpg.h because they aren't
44 * anywhere else and it isn't safe to #include linux/ * stuff.
45 */
46
47#define BLKPG _IO(0x12,105)
48
49/* The argument structure */
50struct blkpg_ioctl_arg {
5d500228
N
51 int op;
52 int flags;
53 int datalen;
54 void *data;
0a816ef9
NB
55};
56
57/* The subfunctions (for the op field) */
58#define BLKPG_ADD_PARTITION 1
59#define BLKPG_DEL_PARTITION 2
60
61/* Sizes of name fields. Unused at present. */
62#define BLKPG_DEVNAMELTH 64
63#define BLKPG_VOLNAMELTH 64
64
65/* The data structure for ADD_PARTITION and DEL_PARTITION */
66struct blkpg_partition {
67 long long start; /* starting offset in bytes */
68 long long length; /* length in bytes */
69 int pno; /* partition number */
70 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
71 to be used in kernel messages */
72 char volname[BLKPG_VOLNAMELTH]; /* volume label */
73};
64c4757e 74
0f22b998 75#include "part.h"
056b331e
N
76
77/* Force a compilation error if condition is true */
78#define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
79
80/* Force a compilation error if condition is true, but also produce a
81 result (of value 0 and type size_t), so the expression can be used
82 e.g. in a structure initializer (or where-ever else comma expressions
83 aren't permitted). */
84#define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
85
d15a1f72 86static int is_dlm_hooks_ready = 0;
7071320a
N
87
88int dlm_funs_ready(void)
89{
90 return is_dlm_hooks_ready ? 1 : 0;
91}
92
7071320a 93static struct dlm_hooks *dlm_hooks = NULL;
d15a1f72
GJ
94struct dlm_lock_resource *dlm_lock_res = NULL;
95static int ast_called = 0;
96
97struct dlm_lock_resource {
98 dlm_lshandle_t *ls;
99 struct dlm_lksb lksb;
100};
101
d15a1f72
GJ
102/* Using poll(2) to wait for and dispatch ASTs */
103static int poll_for_ast(dlm_lshandle_t ls)
104{
105 struct pollfd pfd;
106
107 pfd.fd = dlm_hooks->ls_get_fd(ls);
108 pfd.events = POLLIN;
109
110 while (!ast_called)
111 {
112 if (poll(&pfd, 1, 0) < 0)
113 {
114 perror("poll");
115 return -1;
116 }
117 dlm_hooks->dispatch(dlm_hooks->ls_get_fd(ls));
118 }
119 ast_called = 0;
120
121 return 0;
122}
123
124static void dlm_ast(void *arg)
125{
126 ast_called = 1;
127}
128
81a8a694 129static char *cluster_name = NULL;
d15a1f72 130/* Create the lockspace, take bitmapXXX locks on all the bitmaps. */
81a8a694 131int cluster_get_dlmlock(int *lockid)
d15a1f72
GJ
132{
133 int ret = -1;
134 char str[64];
135 int flags = LKF_NOQUEUE;
136
81a8a694
GJ
137 ret = get_cluster_name(&cluster_name);
138 if (ret) {
139 pr_err("The md can't get cluster name\n");
140 return -1;
141 }
142
d15a1f72 143 dlm_lock_res = xmalloc(sizeof(struct dlm_lock_resource));
81a8a694 144 dlm_lock_res->ls = dlm_hooks->create_lockspace(cluster_name, O_RDWR);
d15a1f72 145 if (!dlm_lock_res->ls) {
81a8a694 146 pr_err("%s failed to create lockspace\n", cluster_name);
32539f74 147 return -ENOMEM;
d15a1f72
GJ
148 }
149
81a8a694 150 snprintf(str, 64, "bitmap%s", cluster_name);
b7a462e5
JS
151 ret = dlm_hooks->ls_lock(dlm_lock_res->ls, LKM_PWMODE,
152 &dlm_lock_res->lksb, flags, str, strlen(str),
153 0, dlm_ast, dlm_lock_res, NULL, NULL);
d15a1f72
GJ
154 if (ret) {
155 pr_err("error %d when get PW mode on lock %s\n", errno, str);
32539f74
GJ
156 dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1);
157 return ret;
d15a1f72
GJ
158 }
159
160 /* Wait for it to complete */
161 poll_for_ast(dlm_lock_res->ls);
162 *lockid = dlm_lock_res->lksb.sb_lkid;
163
32539f74 164 return dlm_lock_res->lksb.sb_status;
d15a1f72
GJ
165}
166
81a8a694 167int cluster_release_dlmlock(int lockid)
d15a1f72
GJ
168{
169 int ret = -1;
170
81a8a694
GJ
171 if (!cluster_name)
172 return -1;
173
d15a1f72
GJ
174 ret = dlm_hooks->ls_unlock(dlm_lock_res->ls, lockid, 0,
175 &dlm_lock_res->lksb, dlm_lock_res);
176 if (ret) {
177 pr_err("error %d happened when unlock\n", errno);
178 /* XXX make sure the lock is unlocked eventually */
179 goto out;
180 }
181
182 /* Wait for it to complete */
183 poll_for_ast(dlm_lock_res->ls);
184
185 errno = dlm_lock_res->lksb.sb_status;
186 if (errno != EUNLOCK) {
b7a462e5
JS
187 pr_err("error %d happened in ast when unlock lockspace\n",
188 errno);
d15a1f72
GJ
189 /* XXX make sure the lockspace is unlocked eventually */
190 goto out;
191 }
192
81a8a694 193 ret = dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1);
d15a1f72
GJ
194 if (ret) {
195 pr_err("error %d happened when release lockspace\n", errno);
196 /* XXX make sure the lockspace is released eventually */
197 goto out;
198 }
199 free(dlm_lock_res);
200
201out:
202 return ret;
203}
204
9db2ab4e
JS
205int md_array_valid(int fd)
206{
207 struct mdinfo *sra;
208 int ret;
209
210 sra = sysfs_read(fd, NULL, GET_ARRAY_STATE);
211 if (sra) {
212 if (sra->array_state != ARRAY_UNKNOWN_STATE)
213 ret = 0;
214 else
215 ret = -ENODEV;
216
217 free(sra);
218 } else {
219 /*
220 * GET_ARRAY_INFO doesn't provide access to the proper state
221 * information, so fallback to a basic check for raid_disks != 0
222 */
223 ret = ioctl(fd, RAID_VERSION);
224 }
225
226 return !ret;
227}
228
3ab8f4bf
JS
229int md_array_active(int fd)
230{
231 struct mdinfo *sra;
232 struct mdu_array_info_s array;
a822017f 233 int ret = 0;
3ab8f4bf
JS
234
235 sra = sysfs_read(fd, NULL, GET_ARRAY_STATE);
236 if (sra) {
a822017f 237 if (!md_array_is_active(sra))
3ab8f4bf
JS
238 ret = -ENODEV;
239
240 free(sra);
241 } else {
242 /*
243 * GET_ARRAY_INFO doesn't provide access to the proper state
244 * information, so fallback to a basic check for raid_disks != 0
245 */
246 ret = ioctl(fd, GET_ARRAY_INFO, &array);
247 }
248
249 return !ret;
250}
251
a822017f
MT
252int md_array_is_active(struct mdinfo *info)
253{
254 return (info->array_state != ARRAY_CLEAR &&
255 info->array_state != ARRAY_INACTIVE &&
256 info->array_state != ARRAY_UNKNOWN_STATE);
257}
258
9cd39f01
JS
259/*
260 * Get array info from the kernel. Longer term we want to deprecate the
261 * ioctl and get it from sysfs.
262 */
263int md_get_array_info(int fd, struct mdu_array_info_s *array)
264{
265 return ioctl(fd, GET_ARRAY_INFO, array);
266}
267
018a4882
JS
268/*
269 * Set array info
270 */
271int md_set_array_info(int fd, struct mdu_array_info_s *array)
272{
273 return ioctl(fd, SET_ARRAY_INFO, array);
274}
275
d97572f5
JS
276/*
277 * Get disk info from the kernel.
278 */
279int md_get_disk_info(int fd, struct mdu_disk_info_s *disk)
280{
281 return ioctl(fd, GET_DISK_INFO, disk);
282}
283
64c4757e
NB
284/*
285 * Parse a 128 bit uuid in 4 integers
286 * format is 32 hexx nibbles with options :.<space> separator
287 * If not exactly 32 hex digits are found, return 0
288 * else return 1
289 */
290int parse_uuid(char *str, int uuid[4])
291{
aba69144
NB
292 int hit = 0; /* number of Hex digIT */
293 int i;
294 char c;
ca3b6696
N
295 for (i = 0; i < 4; i++)
296 uuid[i] = 0;
aba69144 297
ca3b6696 298 while ((c = *str++) != 0) {
aba69144 299 int n;
ca3b6696 300 if (c >= '0' && c <= '9')
aba69144 301 n = c-'0';
ca3b6696 302 else if (c >= 'a' && c <= 'f')
aba69144 303 n = 10 + c - 'a';
ca3b6696 304 else if (c >= 'A' && c <= 'F')
aba69144
NB
305 n = 10 + c - 'A';
306 else if (strchr(":. -", c))
307 continue;
308 else return 0;
309
310 if (hit<32) {
311 uuid[hit/8] <<= 4;
312 uuid[hit/8] += n;
313 }
314 hit++;
82b27616 315 }
aba69144
NB
316 if (hit == 32)
317 return 1;
318 return 0;
64c4757e
NB
319}
320
64c4757e
NB
321int get_linux_version()
322{
323 struct utsname name;
98c6faba 324 char *cp;
f161d047 325 int a = 0, b = 0,c = 0;
64c4757e
NB
326 if (uname(&name) <0)
327 return -1;
328
98c6faba
NB
329 cp = name.release;
330 a = strtoul(cp, &cp, 10);
f161d047
N
331 if (*cp == '.')
332 b = strtoul(cp+1, &cp, 10);
333 if (*cp == '.')
334 c = strtoul(cp+1, &cp, 10);
98c6faba 335
682c7051 336 return (a*1000000)+(b*1000)+c;
64c4757e
NB
337}
338
bc77ed53
DW
339int mdadm_version(char *version)
340{
341 int a, b, c;
342 char *cp;
343
344 if (!version)
345 version = Version;
346
347 cp = strchr(version, '-');
348 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
349 return -1;
350 cp += 3;
351 a = strtoul(cp, &cp, 10);
352 if (*cp != '.')
353 return -1;
354 b = strtoul(cp+1, &cp, 10);
355 if (*cp == '.')
356 c = strtoul(cp+1, &cp, 10);
357 else
358 c = 0;
359 if (*cp != ' ' && *cp != '-')
360 return -1;
361 return (a*1000000)+(b*1000)+c;
362}
363
51d4261c 364unsigned long long parse_size(char *size)
84e11361
N
365{
366 /* parse 'size' which should be a number optionally
367 * followed by 'K', 'M', or 'G'.
368 * Without a suffix, K is assumed.
369 * Number returned is in sectors (half-K)
822e393a 370 * INVALID_SECTORS returned on error.
84e11361
N
371 */
372 char *c;
373 long long s = strtoll(size, &c, 10);
374 if (s > 0) {
375 switch (*c) {
376 case 'K':
377 c++;
378 default:
379 s *= 2;
380 break;
381 case 'M':
382 c++;
383 s *= 1024 * 2;
384 break;
385 case 'G':
386 c++;
387 s *= 1024 * 1024 * 2;
388 break;
72ca9bcf
N
389 case 's': /* sectors */
390 c++;
391 break;
84e11361 392 }
51d4261c 393 } else
822e393a 394 s = INVALID_SECTORS;
84e11361 395 if (*c)
822e393a 396 s = INVALID_SECTORS;
84e11361
N
397 return s;
398}
399
5339f996
GJ
400int is_near_layout_10(int layout)
401{
402 int fc, fo;
403
404 fc = (layout >> 8) & 255;
405 fo = layout & (1 << 16);
406 if (fc > 1 || fo > 0)
407 return 0;
408 return 1;
409}
410
4a06e2c2
N
411int parse_layout_10(char *layout)
412{
413 int copies, rv;
414 char *cp;
415 /* Parse the layout string for raid10 */
416 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
417 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
418 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
419 copies > 200 ||
420 *cp)
421 return -1;
422 if (layout[0] == 'n')
423 rv = 256 + copies;
424 else if (layout[0] == 'o')
425 rv = 0x10000 + (copies<<8) + 1;
426 else
427 rv = 1 + (copies<<8);
428 return rv;
429}
430
431int parse_layout_faulty(char *layout)
432{
433 /* Parse the layout string for 'faulty' */
434 int ln = strcspn(layout, "0123456789");
503975b9 435 char *m = xstrdup(layout);
4a06e2c2
N
436 int mode;
437 m[ln] = 0;
438 mode = map_name(faultylayout, m);
439 if (mode == UnSet)
440 return -1;
441
442 return mode | (atoi(layout+ln)<< ModeShift);
443}
79868890
N
444
445long parse_num(char *num)
446{
447 /* Either return a valid number, or -1 */
448 char *c;
449 long rv = strtol(num, &c, 10);
450 if (rv < 0 || *c || !num[0])
451 return -1;
452 else
453 return rv;
454}
4a06e2c2 455
4de90913
GJ
456int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
457{
458 char *dev;
459 *slot = strtoul(input, &dev, 10);
460 if (dev == input || dev[0] != ':')
461 return -1;
462 *devname = dev+1;
463 return 0;
464}
465
0430ed48
NB
466void remove_partitions(int fd)
467{
468 /* remove partitions from this block devices.
469 * This is used for components added to an array
470 */
471#ifdef BLKPG_DEL_PARTITION
472 struct blkpg_ioctl_arg a;
473 struct blkpg_partition p;
474
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);
ca3b6696 480 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
481 ioctl(fd, BLKPG, &a);
482#endif
483}
484
691c6ee1
N
485int test_partition(int fd)
486{
487 /* Check if fd is a whole-disk or a partition.
488 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
489 * will return ENXIO on an invalid partition number.
490 */
491 struct blkpg_ioctl_arg a;
492 struct blkpg_partition p;
493 a.op = BLKPG_DEL_PARTITION;
494 a.data = (void*)&p;
495 a.datalen = sizeof(p);
496 a.flags = 0;
497 memset(a.data, 0, a.datalen);
498 p.pno = 1<<30;
499 if (ioctl(fd, BLKPG, &a) == 0)
500 /* Very unlikely, but not a partition */
501 return 0;
357ac106 502 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
503 /* not a partition */
504 return 0;
505
506 return 1;
507}
508
bfd76b93
CA
509int test_partition_from_id(dev_t id)
510{
511 char buf[20];
512 int fd, rv;
513
514 sprintf(buf, "%d:%d", major(id), minor(id));
515 fd = dev_open(buf, O_RDONLY);
516 if (fd < 0)
517 return -1;
518 rv = test_partition(fd);
519 close(fd);
520 return rv;
521}
691c6ee1 522
de5a472e 523int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 524{
265e0f17 525 int copies, first;
de5a472e
N
526 int i;
527 int avail_disks = 0;
528
529 for (i = 0; i < raid_disks; i++)
530 avail_disks += !!avail[i];
531
64c4757e 532 switch (level) {
265e0f17
NB
533 case 10:
534 /* This is the tricky one - we need to check
535 * which actual disks are present.
536 */
702b557b 537 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 538 first = 0;
265e0f17
NB
539 do {
540 /* there must be one of the 'copies' form 'first' */
541 int n = copies;
ca3b6696 542 int cnt = 0;
b3ec716d 543 int this = first;
265e0f17 544 while (n--) {
b3ec716d 545 if (avail[this])
265e0f17 546 cnt++;
b3ec716d 547 this = (this+1) % raid_disks;
265e0f17
NB
548 }
549 if (cnt == 0)
550 return 0;
b3ec716d 551 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
552 } while (first != 0);
553 return 1;
e5329c37 554
df0d4ea0 555 case LEVEL_MULTIPATH:
e0d19036 556 return avail_disks>= 1;
df0d4ea0 557 case LEVEL_LINEAR:
64c4757e
NB
558 case 0:
559 return avail_disks == raid_disks;
560 case 1:
561 return avail_disks >= 1;
562 case 4:
330d6900
N
563 if (avail_disks == raid_disks - 1 &&
564 !avail[raid_disks - 1])
565 /* If just the parity device is missing, then we
566 * have enough, even if not clean
567 */
568 return 1;
569 /* FALL THROUGH */
64c4757e 570 case 5:
583315d9
NB
571 if (clean)
572 return avail_disks >= raid_disks-1;
573 else
574 return avail_disks >= raid_disks;
98c6faba 575 case 6:
583315d9
NB
576 if (clean)
577 return avail_disks >= raid_disks-2;
578 else
579 return avail_disks >= raid_disks;
64c4757e
NB
580 default:
581 return 0;
582 }
583}
584
22e263f6
AC
585const int uuid_zero[4] = { 0, 0, 0, 0 };
586
f277ce36 587int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 588{
f277ce36
NB
589 if (swapuuid) {
590 /* parse uuids are hostendian.
591 * uuid's from some superblocks are big-ending
aba69144 592 * if there is a difference, we need to swap..
f277ce36
NB
593 */
594 unsigned char *ac = (unsigned char *)a;
595 unsigned char *bc = (unsigned char *)b;
596 int i;
ca3b6696 597 for (i = 0; i < 16; i += 4) {
f277ce36
NB
598 if (ac[i+0] != bc[i+3] ||
599 ac[i+1] != bc[i+2] ||
600 ac[i+2] != bc[i+1] ||
601 ac[i+3] != bc[i+0])
602 return 0;
603 }
604 return 1;
605 } else {
606 if (a[0]==b[0] &&
607 a[1]==b[1] &&
608 a[2]==b[2] &&
609 a[3]==b[3])
610 return 1;
611 return 0;
612 }
64c4757e 613}
ca3b6696 614
350f29f9
NB
615void copy_uuid(void *a, int b[4], int swapuuid)
616{
617 if (swapuuid) {
618 /* parse uuids are hostendian.
619 * uuid's from some superblocks are big-ending
620 * if there is a difference, we need to swap..
621 */
622 unsigned char *ac = (unsigned char *)a;
623 unsigned char *bc = (unsigned char *)b;
624 int i;
ca3b6696 625 for (i = 0; i < 16; i += 4) {
350f29f9
NB
626 ac[i+0] = bc[i+3];
627 ac[i+1] = bc[i+2];
628 ac[i+2] = bc[i+1];
629 ac[i+3] = bc[i+0];
630 }
631 } else
632 memcpy(a, b, 16);
633}
64c4757e 634
aae5a112 635char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 636{
9968e376 637 int i, j;
d7288ddc
N
638 char uuid[16];
639 char *c = buf;
640 strcpy(c, "UUID-");
641 c += strlen(c);
aae5a112 642 copy_uuid(uuid, id, swap);
9968e376 643 for (i = 0; i < 4; i++) {
9968e376 644 if (i)
ff54de6e 645 *c++ = sep;
9968e376
DW
646 for (j = 3; j >= 0; j--) {
647 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
648 c+= 2;
649 }
d7288ddc
N
650 }
651 return buf;
aae5a112
DW
652
653}
654
b7a462e5
JS
655char *fname_from_uuid(struct supertype *st, struct mdinfo *info,
656 char *buf, char sep)
aae5a112 657{
86983cce
N
658 // dirty hack to work around an issue with super1 superblocks...
659 // super1 superblocks need swapuuid set in order for assembly to
660 // work, but can't have it set if we want this printout to match
661 // all the other uuid printouts in super1.c, so we force swapuuid
662 // to 1 to make our printout match the rest of super1
b7a462e5
JS
663 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 :
664 st->ss->swapuuid, buf, sep);
d7288ddc
N
665}
666
682c7051
NB
667int check_ext2(int fd, char *name)
668{
669 /*
670 * Check for an ext2fs file system.
671 * Superblock is always 1K at 1K offset
672 *
673 * s_magic is le16 at 56 == 0xEF53
674 * report mtime - le32 at 44
675 * blocks - le32 at 4
676 * logblksize - le32 at 24
677 */
678 unsigned char sb[1024];
679 time_t mtime;
2538ba2a
N
680 unsigned long long size;
681 int bsize;
682c7051
NB
682 if (lseek(fd, 1024,0)!= 1024)
683 return 0;
684 if (read(fd, sb, 1024)!= 1024)
685 return 0;
686 if (sb[56] != 0x53 || sb[57] != 0xef)
687 return 0;
688
689 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
690 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
691 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 692 size <<= bsize;
e7b84f9d 693 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 694 name);
2538ba2a 695 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
696 return 1;
697}
698
699int check_reiser(int fd, char *name)
700{
701 /*
702 * superblock is at 64K
703 * size is 1024;
704 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
705 *
706 */
707 unsigned char sb[1024];
2538ba2a 708 unsigned long long size;
682c7051
NB
709 if (lseek(fd, 64*1024, 0) != 64*1024)
710 return 0;
711 if (read(fd, sb, 1024) != 1024)
712 return 0;
ca3b6696
N
713 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
714 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 715 return 0;
e7b84f9d 716 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 717 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 718 cont_err("size = %lluK\n", size*4);
aba69144 719
682c7051
NB
720 return 1;
721}
722
723int check_raid(int fd, char *name)
724{
4b1ac34b 725 struct mdinfo info;
682c7051 726 time_t crtime;
d078d77c 727 char *level;
82d9eba6 728 struct supertype *st = guess_super(fd);
f9ce90ba 729
cb8f6859
N
730 if (!st)
731 return 0;
fa219dd2
MD
732 if (st->ss->add_to_super != NULL) {
733 st->ss->load_super(st, fd, name);
734 /* Looks like a raid array .. */
735 pr_err("%s appears to be part of a raid array:\n", name);
736 st->ss->getinfo_super(st, &info, NULL);
737 st->ss->free_super(st);
738 crtime = info.array.ctime;
739 level = map_num(pers, info.array.level);
740 if (!level)
741 level = "-unknown-";
742 cont_err("level=%s devices=%d ctime=%s",
743 level, info.array.raid_disks, ctime(&crtime));
744 } else {
745 /* Looks like GPT or MBR */
746 pr_err("partition table exists on %s\n", name);
747 }
82d9eba6 748 return 1;
682c7051
NB
749}
750
0a6bff09
ZL
751int fstat_is_blkdev(int fd, char *devname, dev_t *rdev)
752{
753 struct stat stb;
754
755 if (fstat(fd, &stb) != 0) {
756 pr_err("fstat failed for %s: %s\n", devname, strerror(errno));
757 return 0;
758 }
759 if ((S_IFMT & stb.st_mode) != S_IFBLK) {
760 pr_err("%s is not a block device.\n", devname);
9e04ac1c
ZL
761 return 0;
762 }
763 if (rdev)
764 *rdev = stb.st_rdev;
765 return 1;
766}
767
768int stat_is_blkdev(char *devname, dev_t *rdev)
769{
770 struct stat stb;
771
772 if (stat(devname, &stb) != 0) {
773 pr_err("stat failed for %s: %s\n", devname, strerror(errno));
774 return 0;
775 }
776 if ((S_IFMT & stb.st_mode) != S_IFBLK) {
777 pr_err("%s is not a block device.\n", devname);
0a6bff09
ZL
778 return 0;
779 }
780 if (rdev)
781 *rdev = stb.st_rdev;
782 return 1;
783}
784
682c7051
NB
785int ask(char *mesg)
786{
787 char *add = "";
788 int i;
ca3b6696 789 for (i = 0; i < 5; i++) {
682c7051
NB
790 char buf[100];
791 fprintf(stderr, "%s%s", mesg, add);
792 fflush(stderr);
793 if (fgets(buf, 100, stdin)==NULL)
794 return 0;
795 if (buf[0]=='y' || buf[0]=='Y')
796 return 1;
797 if (buf[0]=='n' || buf[0]=='N')
798 return 0;
799 add = "(y/n) ";
800 }
e7b84f9d 801 pr_err("assuming 'no'\n");
682c7051
NB
802 return 0;
803}
804
8d80900b 805int is_standard(char *dev, int *nump)
e5329c37
NB
806{
807 /* tests if dev is a "standard" md dev name.
808 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 809 * where NN is a string of digits
598f0d58
NB
810 * Returns 1 if a partitionable standard,
811 * -1 if non-partitonable,
812 * 0 if not a standard name.
e5329c37 813 */
8d80900b 814 char *d = strrchr(dev, '/');
ca3b6696 815 int type = 0;
8d80900b
NB
816 int num;
817 if (!d)
e5329c37 818 return 0;
ca3b6696
N
819 if (strncmp(d, "/d",2) == 0)
820 d += 2, type = 1; /* /dev/md/dN{pM} */
821 else if (strncmp(d, "/md_d", 5) == 0)
822 d += 5, type = 1; /* /dev/md_dN{pM} */
823 else if (strncmp(d, "/md", 3) == 0)
824 d += 3, type = -1; /* /dev/mdN */
825 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
826 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
827 else
828 return 0;
8d80900b 829 if (!*d)
e5329c37 830 return 0;
8d80900b
NB
831 num = atoi(d);
832 while (isdigit(*d))
833 d++;
834 if (*d)
e5329c37 835 return 0;
8d80900b
NB
836 if (nump) *nump = num;
837
838 return type;
e5329c37
NB
839}
840
4b1ac34b 841unsigned long calc_csum(void *super, int bytes)
82b27616 842{
56eb10c0 843 unsigned long long newcsum = 0;
82b27616 844 int i;
4b1ac34b
NB
845 unsigned int csum;
846 unsigned int *superc = (unsigned int*) super;
82b27616 847
ca3b6696
N
848 for(i = 0; i < bytes/4; i++)
849 newcsum += superc[i];
82b27616 850 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 851#ifdef __alpha__
aba69144 852/* The in-kernel checksum calculation is always 16bit on
570c0542 853 * the alpha, though it is 32 bit on i386...
ca3b6696 854 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
855 * a way that it shouldn't).
856 */
857 csum = (csum & 0xffff) + (csum >> 16);
858 csum = (csum & 0xffff) + (csum >> 16);
859#endif
82b27616
NB
860 return csum;
861}
cd29a5c8 862
56eb10c0 863char *human_size(long long bytes)
cd29a5c8 864{
ff9239ee 865 static char buf[47];
d5d3721e
NB
866
867 /* We convert bytes to either centi-M{ega,ibi}bytes or
868 * centi-G{igi,ibi}bytes, with appropriate rounding,
869 * and then print 1/100th of those as a decimal.
870 * We allow upto 2048Megabytes before converting to
871 * gigabytes, as that shows more precision and isn't
872 * too large a number.
01619b48 873 * Terabytes are not yet handled.
d5d3721e 874 */
cd29a5c8 875
56eb10c0 876 if (bytes < 5000*1024)
ca3b6696 877 buf[0] = 0;
d5d3721e 878 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 879 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 880 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 881 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
efa29530 882 cMiB/100, cMiB % 100, cMB/100, cMB % 100);
d5d3721e 883 } else {
93d3bd3b 884 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 885 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 886 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
efa29530 887 cGiB/100, cGiB % 100, cGB/100, cGB % 100);
d5d3721e 888 }
cd29a5c8
NB
889 return buf;
890}
e0d19036 891
f0ec6710 892char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
893{
894 static char buf[30];
e0d19036 895
570abc6f
MN
896 /* We convert bytes to either centi-M{ega,ibi}bytes or
897 * centi-G{igi,ibi}bytes, with appropriate rounding,
898 * and then print 1/100th of those as a decimal.
899 * We allow upto 2048Megabytes before converting to
900 * gigabytes, as that shows more precision and isn't
901 * too large a number.
902 * Terabytes are not yet handled.
f0ec6710
MN
903 *
904 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
905 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
906 */
907
e0d19036 908 if (bytes < 5000*1024)
570abc6f 909 buf[0] = 0;
f0ec6710
MN
910 else if (prefix == IEC) {
911 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 912 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710 913 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
efa29530 914 cMiB/100, cMiB % 100);
f0ec6710 915 } else {
93d3bd3b 916 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710 917 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
efa29530 918 cGiB/100, cGiB % 100);
f0ec6710
MN
919 }
920 }
921 else if (prefix == JEDEC) {
922 if (bytes < 2*1024LL*1024LL*1024LL) {
923 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
924 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
efa29530 925 cMB/100, cMB % 100);
f0ec6710
MN
926 } else {
927 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
928 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
efa29530 929 cGB/100, cGB % 100);
f0ec6710 930 }
570abc6f 931 }
f0ec6710
MN
932 else
933 buf[0] = 0;
570abc6f 934
e0d19036
NB
935 return buf;
936}
e4965ef8
N
937
938void print_r10_layout(int layout)
939{
940 int near = layout & 255;
941 int far = (layout >> 8) & 255;
942 int offset = (layout&0x10000);
943 char *sep = "";
944
945 if (near != 1) {
946 printf("%s near=%d", sep, near);
947 sep = ",";
948 }
949 if (far != 1)
950 printf("%s %s=%d", sep, offset?"offset":"far", far);
951 if (near*far == 1)
952 printf("NO REDUNDANCY");
953}
e0d19036 954
5f8097be
NB
955unsigned long long calc_array_size(int level, int raid_disks, int layout,
956 int chunksize, unsigned long long devsize)
577e8448 957{
e03640bd
CA
958 if (level == 1)
959 return devsize;
577e8448
AK
960 devsize &= ~(unsigned long long)((chunksize>>9)-1);
961 return get_data_disks(level, layout, raid_disks) * devsize;
962}
963
964int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
965{
966 int data_disks = 0;
967 switch (level) {
ca3b6696
N
968 case 0: data_disks = raid_disks;
969 break;
970 case 1: data_disks = 1;
971 break;
5f8097be 972 case 4:
ca3b6696
N
973 case 5: data_disks = raid_disks - 1;
974 break;
975 case 6: data_disks = raid_disks - 2;
976 break;
5f8097be
NB
977 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
978 break;
979 }
577e8448
AK
980
981 return data_disks;
5f8097be
NB
982}
983
13db17bd 984dev_t devnm2devid(char *devnm)
4dd2df09
N
985{
986 /* First look in /sys/block/$DEVNM/dev for %d:%d
987 * If that fails, try parsing out a number
988 */
989 char path[100];
990 char *ep;
991 int fd;
992 int mjr,mnr;
993
994 sprintf(path, "/sys/block/%s/dev", devnm);
995 fd = open(path, O_RDONLY);
996 if (fd >= 0) {
997 char buf[20];
998 int n = read(fd, buf, sizeof(buf));
999 close(fd);
1000 if (n > 0)
1001 buf[n] = 0;
1002 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
1003 return makedev(mjr, mnr);
1004 }
1005 if (strncmp(devnm, "md_d", 4) == 0 &&
1006 isdigit(devnm[4]) &&
1007 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
1008 ep > devnm && *ep == 0)
1009 return makedev(get_mdp_major(), mnr << MdpMinorShift);
1010
1011 if (strncmp(devnm, "md", 2) == 0 &&
1012 isdigit(devnm[2]) &&
1013 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
1014 ep > devnm && *ep == 0)
1015 return makedev(MD_MAJOR, mnr);
1016
1017 return 0;
1018}
1019
1020char *get_md_name(char *devnm)
e0d19036
NB
1021{
1022 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 1023 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 1024
e0d19036
NB
1025 static char devname[50];
1026 struct stat stb;
4dd2df09 1027 dev_t rdev = devnm2devid(devnm);
dd0781e5 1028 char *dn;
98c6faba 1029
4dd2df09
N
1030 if (rdev == 0)
1031 return 0;
1032 if (strncmp(devnm, "md_", 3) == 0) {
1033 snprintf(devname, sizeof(devname), "/dev/md/%s",
1034 devnm + 3);
fc54fe7a
JS
1035 if (stat(devname, &stb) == 0 &&
1036 (S_IFMT&stb.st_mode) == S_IFBLK && (stb.st_rdev == rdev))
98c6faba
NB
1037 return devname;
1038 }
4dd2df09 1039 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
fc54fe7a
JS
1040 if (stat(devname, &stb) == 0 && (S_IFMT&stb.st_mode) == S_IFBLK &&
1041 (stb.st_rdev == rdev))
4dd2df09
N
1042 return devname;
1043
1044 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
fc54fe7a
JS
1045 if (stat(devname, &stb) == 0 && (S_IFMT&stb.st_mode) == S_IFBLK &&
1046 (stb.st_rdev == rdev))
4dd2df09
N
1047 return devname;
1048
16c6fa80 1049 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
1050 if (dn)
1051 return dn;
4dd2df09 1052 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 1053 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
1054 if (errno != EEXIST)
1055 return NULL;
e0d19036 1056
fc54fe7a
JS
1057 if (stat(devname, &stb) == 0 && (S_IFMT&stb.st_mode) == S_IFBLK &&
1058 (stb.st_rdev == rdev))
e0d19036
NB
1059 return devname;
1060 unlink(devname);
1061 return NULL;
1062}
1063
1064void put_md_name(char *name)
1065{
ca3b6696 1066 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
1067 unlink(name);
1068}
f9ce90ba 1069
cc742d38
N
1070int get_maj_min(char *dev, int *major, int *minor)
1071{
1072 char *e;
1073 *major = strtoul(dev, &e, 0);
1074 return (e > dev && *e == ':' && e[1] &&
1075 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1076 *e == 0);
1077}
1078
8b0dabea
NB
1079int dev_open(char *dev, int flags)
1080{
1081 /* like 'open', but if 'dev' matches %d:%d, create a temp
1082 * block device and open that
1083 */
8b0dabea
NB
1084 int fd = -1;
1085 char devname[32];
e81cdd9f 1086 int major;
8b0dabea 1087 int minor;
e81cdd9f 1088
9f0ad56b
JS
1089 if (!dev)
1090 return -1;
6df6a774 1091 flags |= O_DIRECT;
e81cdd9f 1092
cc742d38 1093 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1094 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1095 (int)getpid(), major, minor);
1096 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1097 fd = open(devname, flags);
1098 unlink(devname);
6df6a774
N
1099 }
1100 if (fd < 0) {
9dc7d357 1101 /* Try /tmp as /dev appear to be read-only */
b7a462e5
JS
1102 snprintf(devname, sizeof(devname),
1103 "/tmp/.tmp.md.%d:%d:%d",
6df6a774 1104 (int)getpid(), major, minor);
b7a462e5
JS
1105 if (mknod(devname, S_IFBLK|0600,
1106 makedev(major, minor)) == 0) {
6df6a774
N
1107 fd = open(devname, flags);
1108 unlink(devname);
1109 }
8b0dabea
NB
1110 }
1111 } else
6df6a774 1112 fd = open(dev, flags);
8b0dabea
NB
1113 return fd;
1114}
f9ce90ba 1115
4dd2df09 1116int open_dev_flags(char *devnm, int flags)
e8a70c89 1117{
13db17bd 1118 dev_t devid;
e8a70c89
N
1119 char buf[20];
1120
4dd2df09
N
1121 devid = devnm2devid(devnm);
1122 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1123 return dev_open(buf, flags);
1124}
1125
4dd2df09 1126int open_dev(char *devnm)
d998b738 1127{
4dd2df09 1128 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1129}
1130
4dd2df09 1131int open_dev_excl(char *devnm)
a931db9e
NB
1132{
1133 char buf[20];
1134 int i;
7187750e 1135 int flags = O_RDWR;
13db17bd 1136 dev_t devid = devnm2devid(devnm);
399e0b97 1137 long delay = 1000;
a931db9e 1138
4dd2df09 1139 sprintf(buf, "%d:%d", major(devid), minor(devid));
efa29530 1140 for (i = 0; i < 25; i++) {
7187750e 1141 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1142 if (fd >= 0)
1143 return fd;
7187750e
N
1144 if (errno == EACCES && flags == O_RDWR) {
1145 flags = O_RDONLY;
1146 continue;
1147 }
a931db9e
NB
1148 if (errno != EBUSY)
1149 return fd;
399e0b97
N
1150 usleep(delay);
1151 if (delay < 200000)
1152 delay *= 2;
a931db9e
NB
1153 }
1154 return -1;
1155}
1156
9008ed1c
N
1157int same_dev(char *one, char *two)
1158{
1159 struct stat st1, st2;
1160 if (stat(one, &st1) != 0)
1161 return 0;
1162 if (stat(two, &st2) != 0)
1163 return 0;
1164 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1165 return 0;
1166 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1167 return 0;
1168 return st1.st_rdev == st2.st_rdev;
1169}
1170
a7c6e3fb 1171void wait_for(char *dev, int fd)
a714580e
N
1172{
1173 int i;
a7c6e3fb 1174 struct stat stb_want;
399e0b97 1175 long delay = 1000;
a7c6e3fb
N
1176
1177 if (fstat(fd, &stb_want) != 0 ||
1178 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1179 return;
a714580e 1180
efa29530 1181 for (i = 0; i < 25; i++) {
a714580e 1182 struct stat stb;
a7c6e3fb
N
1183 if (stat(dev, &stb) == 0 &&
1184 (stb.st_mode & S_IFMT) == S_IFBLK &&
1185 (stb.st_rdev == stb_want.st_rdev))
a714580e 1186 return;
399e0b97
N
1187 usleep(delay);
1188 if (delay < 200000)
1189 delay *= 2;
a714580e 1190 }
436305c6 1191 if (i == 25)
9b8fea91 1192 pr_err("timeout waiting for %s\n", dev);
a714580e
N
1193}
1194
0f22b998
N
1195struct superswitch *superlist[] =
1196{
1197 &super0, &super1,
1198 &super_ddf, &super_imsm,
0592faeb 1199 &mbr, &gpt,
32141c17
JS
1200 NULL
1201};
f7dd881f 1202
4725bc31 1203struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1204{
1686dc25
NB
1205 mdu_array_info_t array;
1206 int vers;
1207 int minor;
1208 struct supertype *st = NULL;
7e0f6979 1209 struct mdinfo *sra;
142cb9e1 1210 char *verstr;
1686dc25
NB
1211 char version[20];
1212 int i;
f7e7067b 1213 char *subarray = NULL;
4dd2df09 1214 char container[32] = "";
1686dc25 1215
4dd2df09 1216 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1217
1218 if (sra) {
7e0f6979
NB
1219 vers = sra->array.major_version;
1220 minor = sra->array.minor_version;
142cb9e1 1221 verstr = sra->text_version;
1686dc25 1222 } else {
9cd39f01 1223 if (md_get_array_info(fd, &array))
1686dc25
NB
1224 array.major_version = array.minor_version = 0;
1225 vers = array.major_version;
1226 minor = array.minor_version;
142cb9e1 1227 verstr = "";
6fbba4c9 1228 }
82d9eba6 1229
1686dc25
NB
1230 if (vers != -1) {
1231 sprintf(version, "%d.%d", vers, minor);
1232 verstr = version;
6fbba4c9 1233 }
3c558363 1234 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1235 char *dev = verstr+1;
5f7e44b2 1236
f7e7067b 1237 subarray = strchr(dev, '/');
92d49ecf 1238 if (subarray) {
f7e7067b 1239 *subarray++ = '\0';
503975b9 1240 subarray = xstrdup(subarray);
92d49ecf 1241 }
4dd2df09 1242 strcpy(container, dev);
15d230f7 1243 sysfs_free(sra);
5f7e44b2 1244 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1245 if (sra && sra->text_version[0])
1246 verstr = sra->text_version;
1247 else
1248 verstr = "-no-metadata-";
f7e7067b
NB
1249 }
1250
efa29530 1251 for (i = 0; st == NULL && superlist[i]; i++)
f7e7067b 1252 st = superlist[i]->match_metadata_desc(verstr);
1686dc25 1253
15d230f7 1254 sysfs_free(sra);
f7e7067b 1255 if (st) {
3b0896f8 1256 st->sb = NULL;
1f49fb3a
N
1257 if (subarrayp)
1258 *subarrayp = subarray;
4dd2df09
N
1259 strcpy(st->container_devnm, container);
1260 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1261 } else
1262 free(subarray);
5f7e44b2 1263
82d9eba6 1264 return st;
f9ce90ba 1265}
ea24acd0 1266
0f0749ad 1267int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1268{
1269 char buf[20];
1270 int fd;
1271
1272 sprintf(buf, "%d:%d", major(id), minor(id));
1273 fd = dev_open(buf, O_RDONLY);
1274 if (fd < 0)
1275 return 0;
1276 if (get_dev_size(fd, NULL, size)) {
1277 close(fd);
1278 return 1;
1279 }
1280 close(fd);
1281 return 0;
1282}
f9ce90ba 1283
4b57ecf6
AO
1284int dev_sector_size_from_id(dev_t id, unsigned int *size)
1285{
1286 char buf[20];
1287 int fd;
1288
1289 sprintf(buf, "%d:%d", major(id), minor(id));
1290 fd = dev_open(buf, O_RDONLY);
1291 if (fd < 0)
1292 return 0;
1293 if (get_dev_sector_size(fd, NULL, size)) {
1294 close(fd);
1295 return 1;
1296 }
1297 close(fd);
1298 return 0;
1299}
1300
159c3a1a 1301struct supertype *dup_super(struct supertype *orig)
3da92f27 1302{
159c3a1a 1303 struct supertype *st;
1686dc25 1304
d2ca6449
NB
1305 if (!orig)
1306 return orig;
503975b9 1307 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1308 st->ss = orig->ss;
1309 st->max_devs = orig->max_devs;
1310 st->minor_version = orig->minor_version;
cb8f6859 1311 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1312 st->data_offset = orig->data_offset;
159c3a1a
NB
1313 st->sb = NULL;
1314 st->info = NULL;
1315 return st;
3da92f27
NB
1316}
1317
54887ad8 1318struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1319{
1320 /* try each load_super to find the best match,
1321 * and return the best superswitch
1322 */
82d9eba6
NB
1323 struct superswitch *ss;
1324 struct supertype *st;
1d13b599 1325 unsigned int besttime = 0;
570c0542 1326 int bestsuper = -1;
f9ce90ba
NB
1327 int i;
1328
503975b9 1329 st = xcalloc(1, sizeof(*st));
4dd2df09 1330 st->container_devnm[0] = 0;
d1d599ea 1331
efa29530 1332 for (i = 0; superlist[i]; i++) {
f9ce90ba
NB
1333 int rv;
1334 ss = superlist[i];
54887ad8
N
1335 if (guess_type == guess_array && ss->add_to_super == NULL)
1336 continue;
1337 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1338 continue;
ef609477 1339 memset(st, 0, sizeof(*st));
df3346e6 1340 st->ignore_hw_compat = 1;
3da92f27 1341 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1342 if (rv == 0) {
1343 struct mdinfo info;
a5d85af7 1344 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1345 if (bestsuper == -1 ||
1346 besttime < info.array.ctime) {
1347 bestsuper = i;
1348 besttime = info.array.ctime;
570c0542 1349 }
3da92f27 1350 ss->free_super(st);
570c0542
NB
1351 }
1352 }
1353 if (bestsuper != -1) {
1354 int rv;
ef609477 1355 memset(st, 0, sizeof(*st));
df3346e6 1356 st->ignore_hw_compat = 1;
3da92f27 1357 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1358 if (rv == 0) {
5e747af2 1359 superlist[bestsuper]->free_super(st);
82d9eba6 1360 return st;
f9ce90ba
NB
1361 }
1362 }
570c0542 1363 free(st);
f9ce90ba
NB
1364 return NULL;
1365}
fe6729fa 1366
beae1dfe
NB
1367/* Return size of device in bytes */
1368int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1369{
1370 unsigned long long ldsize;
c2c9bb6f
NB
1371 struct stat st;
1372
1373 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1374 ldsize = (unsigned long long)st.st_size;
1375 else
beae1dfe
NB
1376#ifdef BLKGETSIZE64
1377 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1378#endif
1379 {
1380 unsigned long dsize;
1381 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1382 ldsize = dsize;
1383 ldsize <<= 9;
1384 } else {
1385 if (dname)
32971509 1386 pr_err("Cannot get size of %s: %s\n",
beae1dfe
NB
1387 dname, strerror(errno));
1388 return 0;
1389 }
1390 }
1391 *sizep = ldsize;
1392 return 1;
1393}
8fac0577 1394
32971509
PB
1395/* Return sector size of device in bytes */
1396int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep)
1397{
1398 unsigned int sectsize;
1399
1400 if (ioctl(fd, BLKSSZGET, &sectsize) != 0) {
1401 if (dname)
1402 pr_err("Cannot get sector size of %s: %s\n",
1403 dname, strerror(errno));
1404 return 0;
1405 }
1406
1407 *sectsizep = sectsize;
1408 return 1;
1409}
1410
3a371610
N
1411/* Return true if this can only be a container, not a member device.
1412 * i.e. is and md device and size is zero
1413 */
1414int must_be_container(int fd)
1415{
f5c924f4 1416 struct mdinfo *mdi;
3a371610 1417 unsigned long long size;
f5c924f4
JS
1418
1419 mdi = sysfs_read(fd, NULL, GET_VERSION);
1420 if (!mdi)
3a371610 1421 return 0;
f5c924f4
JS
1422 sysfs_free(mdi);
1423
3a371610
N
1424 if (get_dev_size(fd, NULL, &size) == 0)
1425 return 1;
1426 if (size == 0)
1427 return 1;
1428 return 0;
1429}
034b203a
TM
1430
1431/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1432 * Returns: 1 if successful
1433 * -1 for unknown partition type
1434 * 0 for other errors
1435 */
1436static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1437{
056b331e 1438 struct GPT gpt;
034b203a 1439 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1440 struct GPT_part_entry *part;
1441 char buf[512];
034b203a
TM
1442 unsigned long long curr_part_end;
1443 unsigned all_partitions, entry_size;
f21e18ca 1444 unsigned part_nr;
41b06495 1445 unsigned int sector_size = 0;
034b203a
TM
1446
1447 *endofpart = 0;
1448
056b331e 1449 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1450 /* skip protective MBR */
41b06495
MD
1451 if (!get_dev_sector_size(fd, NULL, &sector_size))
1452 return 0;
1453 lseek(fd, sector_size, SEEK_SET);
73e658d8 1454 /* read GPT header */
056b331e 1455 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1456 return 0;
1457
1458 /* get the number of partition entries and the entry size */
056b331e
N
1459 all_partitions = __le32_to_cpu(gpt.part_cnt);
1460 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1461
1462 /* Check GPT signature*/
056b331e 1463 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1464 return -1;
1465
1466 /* sanity checks */
1467 if (all_partitions > 1024 ||
db7fdfe4 1468 entry_size > sizeof(buf))
034b203a
TM
1469 return -1;
1470
db7fdfe4
JS
1471 part = (struct GPT_part_entry *)buf;
1472
41b06495
MD
1473 /* set offset to third block (GPT entries) */
1474 lseek(fd, sector_size*2, SEEK_SET);
ca3b6696 1475 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1476 /* read partition entry */
db7fdfe4 1477 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1478 return 0;
1479
034b203a 1480 /* is this valid partition? */
db7fdfe4 1481 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1482 /* check the last lba for the current partition */
db7fdfe4 1483 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1484 if (curr_part_end > *endofpart)
1485 *endofpart = curr_part_end;
1486 }
1487
034b203a
TM
1488 }
1489 return 1;
1490}
1491
1492/* Sets endofpart parameter to the last block used by the last partition on the device.
1493 * Returns: 1 if successful
1494 * -1 for unknown partition type
1495 * 0 for other errors
1496 */
1497static int get_last_partition_end(int fd, unsigned long long *endofpart)
1498{
056b331e 1499 struct MBR boot_sect;
034b203a 1500 unsigned long long curr_part_end;
f21e18ca 1501 unsigned part_nr;
31208db9 1502 unsigned int sector_size;
034b203a
TM
1503 int retval = 0;
1504
1505 *endofpart = 0;
1506
056b331e 1507 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1508 /* read MBR */
1509 lseek(fd, 0, 0);
056b331e 1510 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1511 goto abort;
1512
1513 /* check MBP signature */
056b331e 1514 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1515 retval = 1;
1516 /* found the correct signature */
034b203a 1517
ca3b6696 1518 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
8e2bca51
JC
1519 /*
1520 * Have to make every access through boot_sect rather
1521 * than using a pointer to the partition table (or an
1522 * entry), since the entries are not properly aligned.
1523 */
1524
034b203a 1525 /* check for GPT type */
8e2bca51
JC
1526 if (boot_sect.parts[part_nr].part_type ==
1527 MBR_GPT_PARTITION_TYPE) {
034b203a
TM
1528 retval = get_gpt_last_partition_end(fd, endofpart);
1529 break;
1530 }
1531 /* check the last used lba for the current partition */
8e2bca51
JC
1532 curr_part_end =
1533 __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) +
1534 __le32_to_cpu(boot_sect.parts[part_nr].blocks_num);
034b203a
TM
1535 if (curr_part_end > *endofpart)
1536 *endofpart = curr_part_end;
034b203a
TM
1537 }
1538 } else {
1539 /* Unknown partition table */
1540 retval = -1;
1541 }
31208db9
MD
1542 /* calculate number of 512-byte blocks */
1543 if (get_dev_sector_size(fd, NULL, &sector_size))
1544 *endofpart *= (sector_size / 512);
034b203a
TM
1545 abort:
1546 return retval;
1547}
1548
53ed6ac3
KW
1549int check_partitions(int fd, char *dname, unsigned long long freesize,
1550 unsigned long long size)
034b203a
TM
1551{
1552 /*
1553 * Check where the last partition ends
1554 */
1555 unsigned long long endofpart;
034b203a 1556
31208db9 1557 if (get_last_partition_end(fd, &endofpart) > 0) {
034b203a
TM
1558 /* There appears to be a partition table here */
1559 if (freesize == 0) {
1560 /* partitions will not be visible in new device */
e7b84f9d
N
1561 pr_err("partition table exists on %s but will be lost or\n"
1562 " meaningless after creating array\n",
1563 dname);
034b203a
TM
1564 return 1;
1565 } else if (endofpart > freesize) {
1566 /* last partition overlaps metadata */
e7b84f9d
N
1567 pr_err("metadata will over-write last partition on %s.\n",
1568 dname);
034b203a 1569 return 1;
53ed6ac3
KW
1570 } else if (size && endofpart > size) {
1571 /* partitions will be truncated in new device */
e7b84f9d
N
1572 pr_err("array size is too small to cover all partitions on %s.\n",
1573 dname);
53ed6ac3 1574 return 1;
034b203a
TM
1575 }
1576 }
1577 return 0;
1578}
1579
a322f70c
DW
1580int open_container(int fd)
1581{
1582 /* 'fd' is a block device. Find out if it is in use
1583 * by a container, and return an open fd on that container.
1584 */
1585 char path[256];
1586 char *e;
1587 DIR *dir;
1588 struct dirent *de;
1589 int dfd, n;
1590 char buf[200];
1591 int major, minor;
1592 struct stat st;
1593
1594 if (fstat(fd, &st) != 0)
1595 return -1;
1596 sprintf(path, "/sys/dev/block/%d:%d/holders",
1597 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1598 e = path + strlen(path);
1599
1600 dir = opendir(path);
1601 if (!dir)
1602 return -1;
1603 while ((de = readdir(dir))) {
1604 if (de->d_ino == 0)
1605 continue;
1606 if (de->d_name[0] == '.')
1607 continue;
a994592d
N
1608 /* Need to make sure it is a container and not a volume */
1609 sprintf(e, "/%s/md/metadata_version", de->d_name);
1610 dfd = open(path, O_RDONLY);
1611 if (dfd < 0)
1612 continue;
1613 n = read(dfd, buf, sizeof(buf));
1614 close(dfd);
1615 if (n <= 0 || (unsigned)n >= sizeof(buf))
1616 continue;
1617 buf[n] = 0;
1618 if (strncmp(buf, "external", 8) != 0 ||
1619 n < 10 ||
1620 buf[9] == '/')
1621 continue;
a322f70c
DW
1622 sprintf(e, "/%s/dev", de->d_name);
1623 dfd = open(path, O_RDONLY);
1624 if (dfd < 0)
1625 continue;
1626 n = read(dfd, buf, sizeof(buf));
1627 close(dfd);
f21e18ca 1628 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1629 continue;
1630 buf[n] = 0;
1631 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1632 continue;
1633 sprintf(buf, "%d:%d", major, minor);
1634 dfd = dev_open(buf, O_RDONLY);
1635 if (dfd >= 0) {
1636 closedir(dir);
1637 return dfd;
1638 }
1639 }
355726fa 1640 closedir(dir);
a322f70c
DW
1641 return -1;
1642}
1643
33414a01
DW
1644struct superswitch *version_to_superswitch(char *vers)
1645{
1646 int i;
1647
1648 for (i = 0; superlist[i]; i++) {
1649 struct superswitch *ss = superlist[i];
1650
1651 if (strcmp(vers, ss->name) == 0)
1652 return ss;
1653 }
1654
1655 return NULL;
1656}
1657
4dd2df09
N
1658int metadata_container_matches(char *metadata, char *devnm)
1659{
1660 /* Check if 'devnm' is the container named in 'metadata'
1661 * which is
1662 * /containername/componentname or
1663 * -containername/componentname
1664 */
1665 int l;
1666 if (*metadata != '/' && *metadata != '-')
1667 return 0;
1668 l = strlen(devnm);
1669 if (strncmp(metadata+1, devnm, l) != 0)
1670 return 0;
1671 if (metadata[l+1] != '/')
1672 return 0;
1673 return 1;
1674}
1675
1676int metadata_subdev_matches(char *metadata, char *devnm)
1677{
1678 /* Check if 'devnm' is the subdev named in 'metadata'
1679 * which is
1680 * /containername/subdev or
1681 * -containername/subdev
1682 */
1683 char *sl;
1684 if (*metadata != '/' && *metadata != '-')
1685 return 0;
1686 sl = strchr(metadata+1, '/');
1687 if (!sl)
1688 return 0;
1689 if (strcmp(sl+1, devnm) == 0)
1690 return 1;
1691 return 0;
1692}
1693
33414a01
DW
1694int is_container_member(struct mdstat_ent *mdstat, char *container)
1695{
1696 if (mdstat->metadata_version == NULL ||
1697 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1698 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1699 return 0;
1700
1701 return 1;
1702}
1703
1704int is_subarray_active(char *subarray, char *container)
1705{
1706 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1707 struct mdstat_ent *ent;
1708
e5408a32
DW
1709 for (ent = mdstat; ent; ent = ent->next)
1710 if (is_container_member(ent, container))
e5e5d7ce 1711 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1712 break;
33414a01
DW
1713
1714 free_mdstat(mdstat);
1715
1716 return ent != NULL;
1717}
1718
1719/* open_subarray - opens a subarray in a container
1720 * @dev: container device name
feab51f8 1721 * @st: empty supertype
33414a01
DW
1722 * @quiet: block reporting errors flag
1723 *
1724 * On success returns an fd to a container and fills in *st
1725 */
feab51f8 1726int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1727{
1728 struct mdinfo *mdi;
a951a4f7 1729 struct mdinfo *info;
33414a01 1730 int fd, err = 1;
4dd2df09 1731 char *_devnm;
33414a01
DW
1732
1733 fd = open(dev, O_RDWR|O_EXCL);
1734 if (fd < 0) {
1735 if (!quiet)
e7b84f9d 1736 pr_err("Couldn't open %s, aborting\n",
33414a01 1737 dev);
b990032d 1738 return -1;
33414a01
DW
1739 }
1740
4dd2df09
N
1741 _devnm = fd2devnm(fd);
1742 if (_devnm == NULL) {
33414a01 1743 if (!quiet)
e7b84f9d
N
1744 pr_err("Failed to determine device number for %s\n",
1745 dev);
33414a01
DW
1746 goto close_fd;
1747 }
4dd2df09 1748 strcpy(st->devnm, _devnm);
33414a01 1749
4dd2df09 1750 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1751 if (!mdi) {
1752 if (!quiet)
e7b84f9d 1753 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1754 dev);
1755 goto close_fd;
1756 }
1757
1758 if (mdi->array.level != UnSet) {
1759 if (!quiet)
e7b84f9d 1760 pr_err("%s is not a container\n", dev);
33414a01
DW
1761 goto free_sysfs;
1762 }
1763
1764 st->ss = version_to_superswitch(mdi->text_version);
1765 if (!st->ss) {
1766 if (!quiet)
e7b84f9d
N
1767 pr_err("Operation not supported for %s metadata\n",
1768 mdi->text_version);
33414a01
DW
1769 goto free_sysfs;
1770 }
1771
4dd2df09 1772 if (st->devnm[0] == 0) {
33414a01 1773 if (!quiet)
e7b84f9d 1774 pr_err("Failed to allocate device name\n");
33414a01
DW
1775 goto free_sysfs;
1776 }
1777
db20d413 1778 if (!st->ss->load_container) {
33414a01 1779 if (!quiet)
e7b84f9d 1780 pr_err("%s is not a container\n", dev);
4dd2df09 1781 goto free_sysfs;
33414a01
DW
1782 }
1783
db20d413 1784 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1785 if (!quiet)
e7b84f9d 1786 pr_err("Failed to load metadata for %s\n",
db20d413 1787 dev);
4dd2df09 1788 goto free_sysfs;
33414a01
DW
1789 }
1790
a951a4f7
N
1791 info = st->ss->container_content(st, subarray);
1792 if (!info) {
1793 if (!quiet)
e7b84f9d 1794 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1795 subarray, dev);
1796 goto free_super;
1797 }
1798 free(info);
1799
33414a01
DW
1800 err = 0;
1801
1802 free_super:
1803 if (err)
1804 st->ss->free_super(st);
33414a01
DW
1805 free_sysfs:
1806 sysfs_free(mdi);
1807 close_fd:
1808 if (err)
1809 close(fd);
1810
1811 if (err)
1812 return -1;
1813 else
1814 return fd;
1815}
1816
7801ac20
N
1817int add_disk(int mdfd, struct supertype *st,
1818 struct mdinfo *sra, struct mdinfo *info)
1819{
1820 /* Add a device to an array, in one of 2 ways. */
1821 int rv;
32141c17 1822
7801ac20 1823 if (st->ss->external) {
d23534e4
DW
1824 if (info->disk.state & (1<<MD_DISK_SYNC))
1825 info->recovery_start = MaxSector;
1826 else
1827 info->recovery_start = 0;
2904b26f 1828 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1829 if (! rv) {
1830 struct mdinfo *sd2;
f35f2525
N
1831 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1832 if (sd2 == info)
1833 break;
1834 if (sd2 == NULL) {
503975b9 1835 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1836 *sd2 = *info;
1837 sd2->next = sra->devs;
1838 sra->devs = sd2;
1839 }
7801ac20
N
1840 }
1841 } else
7801ac20
N
1842 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1843 return rv;
1844}
1845
de6ae750
N
1846int remove_disk(int mdfd, struct supertype *st,
1847 struct mdinfo *sra, struct mdinfo *info)
1848{
1849 int rv;
32141c17 1850
de6ae750 1851 /* Remove the disk given by 'info' from the array */
de6ae750
N
1852 if (st->ss->external)
1853 rv = sysfs_set_str(sra, info, "slot", "none");
1854 else
de6ae750
N
1855 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1856 info->disk.minor));
1857 return rv;
1858}
1859
1ab9ed2a 1860int hot_remove_disk(int mdfd, unsigned long dev, int force)
2dd271fe 1861{
1ab9ed2a 1862 int cnt = force ? 500 : 5;
2dd271fe
N
1863 int ret;
1864
1865 /* HOT_REMOVE_DISK can fail with EBUSY if there are
1866 * outstanding IO requests to the device.
1867 * In this case, it can be helpful to wait a little while,
1ab9ed2a 1868 * up to 5 seconds if 'force' is set, or 50 msec if not.
2dd271fe
N
1869 */
1870 while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 &&
1871 errno == EBUSY &&
1872 cnt-- > 0)
1873 usleep(10000);
1874
1875 return ret;
1876}
1877
1ab9ed2a 1878int sys_hot_remove_disk(int statefd, int force)
fdd01569 1879{
1ab9ed2a 1880 int cnt = force ? 500 : 5;
fdd01569
N
1881 int ret;
1882
1883 while ((ret = write(statefd, "remove", 6)) == -1 &&
1884 errno == EBUSY &&
1885 cnt-- > 0)
1886 usleep(10000);
1887 return ret == 6 ? 0 : -1;
1888}
1889
f35f2525
N
1890int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1891{
1892 /* Initialise kernel's knowledge of array.
1893 * This varies between externally managed arrays
1894 * and older kernels
1895 */
700483a2 1896 mdu_array_info_t inf;
f35f2525
N
1897 int rv;
1898
f35f2525 1899 if (st->ss->external)
700483a2 1900 return sysfs_set_array(info, 9003);
700483a2
JS
1901
1902 memset(&inf, 0, sizeof(inf));
1903 inf.major_version = info->array.major_version;
1904 inf.minor_version = info->array.minor_version;
1905 rv = md_set_array_info(mdfd, &inf);
1906
f35f2525
N
1907 return rv;
1908}
1909
1e5c6983
DW
1910unsigned long long min_recovery_start(struct mdinfo *array)
1911{
1912 /* find the minimum recovery_start in an array for metadata
1913 * formats that only record per-array recovery progress instead
1914 * of per-device
1915 */
1916 unsigned long long recovery_start = MaxSector;
1917 struct mdinfo *d;
1918
1919 for (d = array->devs; d; d = d->next)
1920 recovery_start = min(recovery_start, d->recovery_start);
1921
1922 return recovery_start;
1923}
1924
4dd2df09 1925int mdmon_pid(char *devnm)
a931db9e
NB
1926{
1927 char path[100];
1928 char pid[10];
1929 int fd;
1930 int n;
10013317 1931
4dd2df09 1932 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1933
24f6f99b 1934 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1935
1936 if (fd < 0)
cf556303 1937 return -1;
a931db9e
NB
1938 n = read(fd, pid, 9);
1939 close(fd);
1940 if (n <= 0)
cf556303 1941 return -1;
24f6f99b 1942 return atoi(pid);
a931db9e
NB
1943}
1944
4dd2df09 1945int mdmon_running(char *devnm)
a931db9e 1946{
4dd2df09 1947 int pid = mdmon_pid(devnm);
24f6f99b 1948 if (pid <= 0)
a931db9e 1949 return 0;
24f6f99b 1950 if (kill(pid, 0) == 0)
a931db9e
NB
1951 return 1;
1952 return 0;
1953}
1954
4dd2df09 1955int start_mdmon(char *devnm)
8850ee3e 1956{
1913c325 1957 int i, skipped;
44d2e365 1958 int len;
ca3b6696 1959 pid_t pid;
9fe32043 1960 int status;
44d2e365
N
1961 char pathbuf[1024];
1962 char *paths[4] = {
1963 pathbuf,
85945e19 1964 BINDIR "/mdmon",
2f1bcf43 1965 "./mdmon",
44d2e365
N
1966 NULL
1967 };
8850ee3e 1968
40ebbb9c 1969 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1970 return 0;
1971
9cf014ec 1972 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1973 if (len > 0) {
1974 char *sl;
1975 pathbuf[len] = 0;
1976 sl = strrchr(pathbuf, '/');
1977 if (sl)
1978 sl++;
1979 else
1980 sl = pathbuf;
1981 strcpy(sl, "mdmon");
1982 } else
1983 pathbuf[0] = '\0';
1984
0f7bdf89 1985 /* First try to run systemctl */
9540cc24
N
1986 if (!check_env("MDADM_NO_SYSTEMCTL"))
1987 switch(fork()) {
1988 case 0:
1989 /* FIXME yuk. CLOSE_EXEC?? */
1990 skipped = 0;
1991 for (i = 3; skipped < 20; i++)
1992 if (close(i) < 0)
1993 skipped++;
1994 else
1995 skipped = 0;
1996
1997 /* Don't want to see error messages from
1998 * systemctl. If the service doesn't exist,
1999 * we start mdmon ourselves.
2000 */
2001 close(2);
2002 open("/dev/null", O_WRONLY);
2003 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
2004 devnm);
2005 status = execl("/usr/bin/systemctl", "systemctl",
2006 "start",
2007 pathbuf, NULL);
2008 status = execl("/bin/systemctl", "systemctl", "start",
2009 pathbuf, NULL);
2010 exit(1);
7a862a02 2011 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
2012 return -1;
2013 default: /* parent - good */
2014 pid = wait(&status);
2015 if (pid >= 0 && status == 0)
2016 return 0;
2017 }
0f7bdf89
JS
2018
2019 /* That failed, try running mdmon directly */
8850ee3e
N
2020 switch(fork()) {
2021 case 0:
2022 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 2023 skipped = 0;
ca3b6696 2024 for (i = 3; skipped < 20; i++)
1913c325
N
2025 if (close(i) < 0)
2026 skipped++;
2027 else
2028 skipped = 0;
2029
ca3b6696 2030 for (i = 0; paths[i]; i++)
a0963a86 2031 if (paths[i][0]) {
2f1bcf43 2032 execl(paths[i], paths[i],
4dd2df09 2033 devnm, NULL);
a0963a86 2034 }
8850ee3e 2035 exit(1);
7a862a02 2036 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 2037 return -1;
9fe32043
N
2038 default: /* parent - good */
2039 pid = wait(&status);
15c10423 2040 if (pid < 0 || status != 0) {
7a862a02 2041 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 2042 return -1;
15c10423 2043 }
8850ee3e
N
2044 }
2045 return 0;
2046}
2047
148acb7b
DW
2048__u32 random32(void)
2049{
2050 __u32 rv;
2051 int rfd = open("/dev/urandom", O_RDONLY);
2052 if (rfd < 0 || read(rfd, &rv, 4) != 4)
2053 rv = random();
2054 if (rfd >= 0)
2055 close(rfd);
2056 return rv;
2057}
2058
c5f71c24
JS
2059void random_uuid(__u8 *buf)
2060{
2061 int fd, i, len;
2062 __u32 r[4];
2063
2064 fd = open("/dev/urandom", O_RDONLY);
2065 if (fd < 0)
2066 goto use_random;
2067 len = read(fd, buf, 16);
2068 close(fd);
2069 if (len != 16)
2070 goto use_random;
2071
2072 return;
2073
2074use_random:
2075 for (i = 0; i < 4; i++)
2076 r[i] = random();
2077 memcpy(buf, r, 16);
2078}
2079
edd8d13c
NB
2080int flush_metadata_updates(struct supertype *st)
2081{
2082 int sfd;
2083 if (!st->updates) {
2084 st->update_tail = NULL;
2085 return -1;
2086 }
2087
4dd2df09 2088 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
2089 if (sfd < 0)
2090 return -1;
2091
2092 while (st->updates) {
2093 struct metadata_update *mu = st->updates;
2094 st->updates = mu->next;
2095
2096 send_message(sfd, mu, 0);
2097 wait_reply(sfd, 0);
2098 free(mu->buf);
2099 free(mu);
2100 }
2101 ack(sfd, 0);
2102 wait_reply(sfd, 0);
2103 close(sfd);
2104 st->update_tail = NULL;
2105 return 0;
2106}
2107
2108void append_metadata_update(struct supertype *st, void *buf, int len)
2109{
2110
503975b9 2111 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
2112
2113 mu->buf = buf;
2114 mu->len = len;
2115 mu->space = NULL;
cb23f1f4 2116 mu->space_list = NULL;
edd8d13c
NB
2117 mu->next = NULL;
2118 *st->update_tail = mu;
2119 st->update_tail = &mu->next;
2120}
a931db9e 2121
fe6729fa
NB
2122#ifdef __TINYC__
2123/* tinyc doesn't optimize this check in ioctl.h out ... */
2124unsigned int __invalid_size_argument_for_IOC = 0;
2125#endif
2126
6d11ec6f
AK
2127int experimental(void)
2128{
2129 if (check_env("MDADM_EXPERIMENTAL"))
2130 return 1;
2131 else {
7a862a02 2132 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
2133 return 0;
2134 }
2135}
2136
326727d9
AC
2137/* Pick all spares matching given criteria from a container
2138 * if min_size == 0 do not check size
2139 * if domlist == NULL do not check domains
2140 * if spare_group given add it to domains of each spare
2141 * metadata allows to test domains using metadata of destination array */
2142struct mdinfo *container_choose_spares(struct supertype *st,
fbfdcb06 2143 struct spare_criteria *criteria,
326727d9
AC
2144 struct domainlist *domlist,
2145 char *spare_group,
2146 const char *metadata, int get_one)
2147{
2148 struct mdinfo *d, **dp, *disks = NULL;
2149
2150 /* get list of all disks in container */
2151 if (st->ss->getinfo_super_disks)
2152 disks = st->ss->getinfo_super_disks(st);
2153
2154 if (!disks)
2155 return disks;
2156 /* find spare devices on the list */
2157 dp = &disks->devs;
2158 disks->array.spare_disks = 0;
2159 while (*dp) {
2160 int found = 0;
2161 d = *dp;
2162 if (d->disk.state == 0) {
2163 /* check if size is acceptable */
2164 unsigned long long dev_size;
4b57ecf6
AO
2165 unsigned int dev_sector_size;
2166 int size_valid = 0;
2167 int sector_size_valid = 0;
2168
326727d9
AC
2169 dev_t dev = makedev(d->disk.major,d->disk.minor);
2170
fbfdcb06 2171 if (!criteria->min_size ||
326727d9 2172 (dev_size_from_id(dev, &dev_size) &&
fbfdcb06 2173 dev_size >= criteria->min_size))
4b57ecf6
AO
2174 size_valid = 1;
2175
2176 if (!criteria->sector_size ||
2177 (dev_sector_size_from_id(dev, &dev_sector_size) &&
2178 criteria->sector_size == dev_sector_size))
2179 sector_size_valid = 1;
2180
2181 found = size_valid && sector_size_valid;
2182
326727d9
AC
2183 /* check if domain matches */
2184 if (found && domlist) {
4dd2df09 2185 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2186 if (spare_group)
2187 pol_add(&pol, pol_domain,
2188 spare_group, NULL);
e5508b36 2189 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2190 found = 0;
2191 dev_policy_free(pol);
2192 }
2193 }
2194 if (found) {
2195 dp = &d->next;
2196 disks->array.spare_disks++;
2197 if (get_one) {
2198 sysfs_free(*dp);
2199 d->next = NULL;
2200 }
2201 } else {
2202 *dp = d->next;
2203 d->next = NULL;
2204 sysfs_free(d);
2205 }
2206 }
2207 return disks;
2208}
9eafa1de
MN
2209
2210/* Checks if paths point to the same device
2211 * Returns 0 if they do.
2212 * Returns 1 if they don't.
2213 * Returns -1 if something went wrong,
2214 * e.g. paths are empty or the files
2215 * they point to don't exist */
2216int compare_paths (char* path1, char* path2)
2217{
2218 struct stat st1,st2;
2219
2220 if (path1 == NULL || path2 == NULL)
2221 return -1;
2222 if (stat(path1,&st1) != 0)
2223 return -1;
2224 if (stat(path2,&st2) != 0)
2225 return -1;
2226 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2227 return 0;
2228 return 1;
2229}
a7dec3fd
N
2230
2231/* Make sure we can open as many devices as needed */
2232void enable_fds(int devices)
2233{
2234 unsigned int fds = 20 + devices;
2235 struct rlimit lim;
d7be7d87 2236 if (getrlimit(RLIMIT_NOFILE, &lim) != 0 || lim.rlim_cur >= fds)
a7dec3fd
N
2237 return;
2238 if (lim.rlim_max < fds)
2239 lim.rlim_max = fds;
2240 lim.rlim_cur = fds;
2241 setrlimit(RLIMIT_NOFILE, &lim);
2242}
a9c15847
N
2243
2244int in_initrd(void)
2245{
2246 /* This is based on similar function in systemd. */
2247 struct statfs s;
76d0f188
JS
2248 /* statfs.f_type is signed long on s390x and MIPS, causing all
2249 sorts of sign extension problems with RAMFS_MAGIC being
2250 defined as 0x858458f6 */
a9c15847 2251 return statfs("/", &s) >= 0 &&
7ac5d47e 2252 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2253 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2254 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2255}
8832342d
N
2256
2257void reopen_mddev(int mdfd)
2258{
2259 /* Re-open without any O_EXCL, but keep
2260 * the same fd
2261 */
2262 char *devnm;
2263 int fd;
2264 devnm = fd2devnm(mdfd);
2265 close(mdfd);
2266 fd = open_dev(devnm);
2267 if (fd >= 0 && fd != mdfd)
2268 dup2(fd, mdfd);
2269}
7716570e 2270
e80357f8
GJ
2271static struct cmap_hooks *cmap_hooks = NULL;
2272static int is_cmap_hooks_ready = 0;
7716570e 2273
e80357f8
GJ
2274void set_cmap_hooks(void)
2275{
2276 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2277 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2278 if (!cmap_hooks->cmap_handle)
2279 return;
7716570e 2280
b7a462e5
JS
2281 cmap_hooks->initialize =
2282 dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2283 cmap_hooks->get_string =
2284 dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
e80357f8 2285 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2286
e80357f8
GJ
2287 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2288 !cmap_hooks->finalize)
2289 dlclose(cmap_hooks->cmap_handle);
2290 else
2291 is_cmap_hooks_ready = 1;
2292}
7716570e 2293
e80357f8
GJ
2294int get_cluster_name(char **cluster_name)
2295{
2296 int rv = -1;
2297 cmap_handle_t handle;
7716570e 2298
e80357f8
GJ
2299 if (!is_cmap_hooks_ready)
2300 return rv;
7716570e 2301
e80357f8 2302 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2303 if (rv != CS_OK)
2304 goto out;
2305
e80357f8 2306 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2307 if (rv != CS_OK) {
2308 free(*cluster_name);
2309 rv = -1;
2310 goto name_err;
2311 }
2312
2313 rv = 0;
2314name_err:
e80357f8 2315 cmap_hooks->finalize(handle);
7716570e 2316out:
7716570e
GJ
2317 return rv;
2318}
d15a1f72
GJ
2319
2320void set_dlm_hooks(void)
2321{
2322 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2323 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2324 if (!dlm_hooks->dlm_handle)
2325 return;
2326
b7a462e5
JS
2327 dlm_hooks->create_lockspace =
2328 dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2329 dlm_hooks->release_lockspace =
2330 dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
d15a1f72
GJ
2331 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2332 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2333 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2334 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2335
2336 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2337 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2338 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2339 dlclose(dlm_hooks->dlm_handle);
2340 else
2341 is_dlm_hooks_ready = 1;
2342}
e80357f8
GJ
2343
2344void set_hooks(void)
2345{
2346 set_dlm_hooks();
2347 set_cmap_hooks();
2348}
b2514242
PB
2349
2350int zero_disk_range(int fd, unsigned long long sector, size_t count)
2351{
2352 int ret = 0;
2353 int fd_zero;
2354 void *addr = NULL;
2355 size_t written = 0;
2356 size_t len = count * 512;
2357 ssize_t n;
2358
2359 fd_zero = open("/dev/zero", O_RDONLY);
2360 if (fd_zero < 0) {
2361 pr_err("Cannot open /dev/zero\n");
2362 return -1;
2363 }
2364
2365 if (lseek64(fd, sector * 512, SEEK_SET) < 0) {
2366 ret = -errno;
2367 pr_err("Failed to seek offset for zeroing\n");
2368 goto out;
2369 }
2370
2371 addr = mmap(NULL, len, PROT_READ, MAP_PRIVATE, fd_zero, 0);
2372
2373 if (addr == MAP_FAILED) {
2374 ret = -errno;
2375 pr_err("Mapping /dev/zero failed\n");
2376 goto out;
2377 }
2378
2379 do {
2380 n = write(fd, addr + written, len - written);
2381 if (n < 0) {
2382 if (errno == EINTR)
2383 continue;
2384 ret = -errno;
2385 pr_err("Zeroing disk range failed\n");
2386 break;
2387 }
2388 written += n;
2389 } while (written != len);
2390
2391 munmap(addr, len);
2392
2393out:
2394 close(fd_zero);
2395 return ret;
2396}