imsm: set generation number when reading superblock
[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
N
32#include <sys/vfs.h>
33#include <linux/magic.h>
986b8688 34#include <poll.h>
98c6faba 35#include <ctype.h>
a322f70c 36#include <dirent.h>
a931db9e 37#include <signal.h>
7716570e 38#include <dlfcn.h>
7716570e 39
0a816ef9
NB
40
41/*
42 * following taken from linux/blkpg.h because they aren't
43 * anywhere else and it isn't safe to #include linux/ * stuff.
44 */
45
46#define BLKPG _IO(0x12,105)
47
48/* The argument structure */
49struct blkpg_ioctl_arg {
5d500228
N
50 int op;
51 int flags;
52 int datalen;
53 void *data;
0a816ef9
NB
54};
55
56/* The subfunctions (for the op field) */
57#define BLKPG_ADD_PARTITION 1
58#define BLKPG_DEL_PARTITION 2
59
60/* Sizes of name fields. Unused at present. */
61#define BLKPG_DEVNAMELTH 64
62#define BLKPG_VOLNAMELTH 64
63
64/* The data structure for ADD_PARTITION and DEL_PARTITION */
65struct blkpg_partition {
66 long long start; /* starting offset in bytes */
67 long long length; /* length in bytes */
68 int pno; /* partition number */
69 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
70 to be used in kernel messages */
71 char volname[BLKPG_VOLNAMELTH]; /* volume label */
72};
64c4757e 73
0f22b998 74#include "part.h"
056b331e
N
75
76/* Force a compilation error if condition is true */
77#define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
78
79/* Force a compilation error if condition is true, but also produce a
80 result (of value 0 and type size_t), so the expression can be used
81 e.g. in a structure initializer (or where-ever else comma expressions
82 aren't permitted). */
83#define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
84
d15a1f72 85static int is_dlm_hooks_ready = 0;
7071320a
N
86
87int dlm_funs_ready(void)
88{
89 return is_dlm_hooks_ready ? 1 : 0;
90}
91
92#ifndef MDASSEMBLE
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);
d15a1f72
GJ
151 ret = dlm_hooks->ls_lock(dlm_lock_res->ls, LKM_PWMODE, &dlm_lock_res->lksb,
152 flags, str, strlen(str), 0, dlm_ast,
153 dlm_lock_res, NULL, NULL);
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) {
187 pr_err("error %d happened in ast when unlock lockspace\n", errno);
188 /* XXX make sure the lockspace is unlocked eventually */
189 goto out;
190 }
191
81a8a694 192 ret = dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1);
d15a1f72
GJ
193 if (ret) {
194 pr_err("error %d happened when release lockspace\n", errno);
195 /* XXX make sure the lockspace is released eventually */
196 goto out;
197 }
198 free(dlm_lock_res);
199
200out:
201 return ret;
202}
7071320a
N
203#else
204int cluster_get_dlmlock(int *lockid)
205{
206 return -1;
207}
208int cluster_release_dlmlock(int lockid)
209{
210 return -1;
211}
212#endif
d15a1f72 213
64c4757e
NB
214/*
215 * Parse a 128 bit uuid in 4 integers
216 * format is 32 hexx nibbles with options :.<space> separator
217 * If not exactly 32 hex digits are found, return 0
218 * else return 1
219 */
220int parse_uuid(char *str, int uuid[4])
221{
aba69144
NB
222 int hit = 0; /* number of Hex digIT */
223 int i;
224 char c;
ca3b6696
N
225 for (i = 0; i < 4; i++)
226 uuid[i] = 0;
aba69144 227
ca3b6696 228 while ((c = *str++) != 0) {
aba69144 229 int n;
ca3b6696 230 if (c >= '0' && c <= '9')
aba69144 231 n = c-'0';
ca3b6696 232 else if (c >= 'a' && c <= 'f')
aba69144 233 n = 10 + c - 'a';
ca3b6696 234 else if (c >= 'A' && c <= 'F')
aba69144
NB
235 n = 10 + c - 'A';
236 else if (strchr(":. -", c))
237 continue;
238 else return 0;
239
240 if (hit<32) {
241 uuid[hit/8] <<= 4;
242 uuid[hit/8] += n;
243 }
244 hit++;
82b27616 245 }
aba69144
NB
246 if (hit == 32)
247 return 1;
248 return 0;
64c4757e
NB
249}
250
64c4757e
NB
251/*
252 * Get the md version number.
253 * We use the RAID_VERSION ioctl if it is supported
254 * If not, but we have a block device with major '9', we assume
255 * 0.36.0
256 *
257 * Return version number as 24 but number - assume version parts
258 * always < 255
259 */
260
261int md_get_version(int fd)
262{
5d500228
N
263 struct stat stb;
264 mdu_version_t vers;
265
266 if (fstat(fd, &stb)<0)
267 return -1;
268 if ((S_IFMT&stb.st_mode) != S_IFBLK)
269 return -1;
270
271 if (ioctl(fd, RAID_VERSION, &vers) == 0)
272 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
273 if (errno == EACCES)
274 return -1;
275 if (major(stb.st_rdev) == MD_MAJOR)
276 return (3600);
64c4757e 277 return -1;
64c4757e
NB
278}
279
64c4757e
NB
280int get_linux_version()
281{
282 struct utsname name;
98c6faba 283 char *cp;
f161d047 284 int a = 0, b = 0,c = 0;
64c4757e
NB
285 if (uname(&name) <0)
286 return -1;
287
98c6faba
NB
288 cp = name.release;
289 a = strtoul(cp, &cp, 10);
f161d047
N
290 if (*cp == '.')
291 b = strtoul(cp+1, &cp, 10);
292 if (*cp == '.')
293 c = strtoul(cp+1, &cp, 10);
98c6faba 294
682c7051 295 return (a*1000000)+(b*1000)+c;
64c4757e
NB
296}
297
71204a50 298#ifndef MDASSEMBLE
bc77ed53
DW
299int mdadm_version(char *version)
300{
301 int a, b, c;
302 char *cp;
303
304 if (!version)
305 version = Version;
306
307 cp = strchr(version, '-');
308 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
309 return -1;
310 cp += 3;
311 a = strtoul(cp, &cp, 10);
312 if (*cp != '.')
313 return -1;
314 b = strtoul(cp+1, &cp, 10);
315 if (*cp == '.')
316 c = strtoul(cp+1, &cp, 10);
317 else
318 c = 0;
319 if (*cp != ' ' && *cp != '-')
320 return -1;
321 return (a*1000000)+(b*1000)+c;
322}
323
51d4261c 324unsigned long long parse_size(char *size)
84e11361
N
325{
326 /* parse 'size' which should be a number optionally
327 * followed by 'K', 'M', or 'G'.
328 * Without a suffix, K is assumed.
329 * Number returned is in sectors (half-K)
822e393a 330 * INVALID_SECTORS returned on error.
84e11361
N
331 */
332 char *c;
333 long long s = strtoll(size, &c, 10);
334 if (s > 0) {
335 switch (*c) {
336 case 'K':
337 c++;
338 default:
339 s *= 2;
340 break;
341 case 'M':
342 c++;
343 s *= 1024 * 2;
344 break;
345 case 'G':
346 c++;
347 s *= 1024 * 1024 * 2;
348 break;
72ca9bcf
N
349 case 's': /* sectors */
350 c++;
351 break;
84e11361 352 }
51d4261c 353 } else
822e393a 354 s = INVALID_SECTORS;
84e11361 355 if (*c)
822e393a 356 s = INVALID_SECTORS;
84e11361
N
357 return s;
358}
359
4a06e2c2
N
360int parse_layout_10(char *layout)
361{
362 int copies, rv;
363 char *cp;
364 /* Parse the layout string for raid10 */
365 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
366 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
367 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
368 copies > 200 ||
369 *cp)
370 return -1;
371 if (layout[0] == 'n')
372 rv = 256 + copies;
373 else if (layout[0] == 'o')
374 rv = 0x10000 + (copies<<8) + 1;
375 else
376 rv = 1 + (copies<<8);
377 return rv;
378}
379
380int parse_layout_faulty(char *layout)
381{
382 /* Parse the layout string for 'faulty' */
383 int ln = strcspn(layout, "0123456789");
503975b9 384 char *m = xstrdup(layout);
4a06e2c2
N
385 int mode;
386 m[ln] = 0;
387 mode = map_name(faultylayout, m);
388 if (mode == UnSet)
389 return -1;
390
391 return mode | (atoi(layout+ln)<< ModeShift);
392}
79868890
N
393
394long parse_num(char *num)
395{
396 /* Either return a valid number, or -1 */
397 char *c;
398 long rv = strtol(num, &c, 10);
399 if (rv < 0 || *c || !num[0])
400 return -1;
401 else
402 return rv;
403}
eb3929a4 404#endif
4a06e2c2 405
4de90913
GJ
406int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
407{
408 char *dev;
409 *slot = strtoul(input, &dev, 10);
410 if (dev == input || dev[0] != ':')
411 return -1;
412 *devname = dev+1;
413 return 0;
414}
415
0430ed48
NB
416void remove_partitions(int fd)
417{
418 /* remove partitions from this block devices.
419 * This is used for components added to an array
420 */
421#ifdef BLKPG_DEL_PARTITION
422 struct blkpg_ioctl_arg a;
423 struct blkpg_partition p;
424
425 a.op = BLKPG_DEL_PARTITION;
426 a.data = (void*)&p;
427 a.datalen = sizeof(p);
428 a.flags = 0;
429 memset(a.data, 0, a.datalen);
ca3b6696 430 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
431 ioctl(fd, BLKPG, &a);
432#endif
433}
434
691c6ee1
N
435int test_partition(int fd)
436{
437 /* Check if fd is a whole-disk or a partition.
438 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
439 * will return ENXIO on an invalid partition number.
440 */
441 struct blkpg_ioctl_arg a;
442 struct blkpg_partition p;
443 a.op = BLKPG_DEL_PARTITION;
444 a.data = (void*)&p;
445 a.datalen = sizeof(p);
446 a.flags = 0;
447 memset(a.data, 0, a.datalen);
448 p.pno = 1<<30;
449 if (ioctl(fd, BLKPG, &a) == 0)
450 /* Very unlikely, but not a partition */
451 return 0;
357ac106 452 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
453 /* not a partition */
454 return 0;
455
456 return 1;
457}
458
bfd76b93
CA
459int test_partition_from_id(dev_t id)
460{
461 char buf[20];
462 int fd, rv;
463
464 sprintf(buf, "%d:%d", major(id), minor(id));
465 fd = dev_open(buf, O_RDONLY);
466 if (fd < 0)
467 return -1;
468 rv = test_partition(fd);
469 close(fd);
470 return rv;
471}
691c6ee1 472
de5a472e 473int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 474{
265e0f17 475 int copies, first;
de5a472e
N
476 int i;
477 int avail_disks = 0;
478
479 for (i = 0; i < raid_disks; i++)
480 avail_disks += !!avail[i];
481
64c4757e 482 switch (level) {
265e0f17
NB
483 case 10:
484 /* This is the tricky one - we need to check
485 * which actual disks are present.
486 */
702b557b 487 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 488 first = 0;
265e0f17
NB
489 do {
490 /* there must be one of the 'copies' form 'first' */
491 int n = copies;
ca3b6696 492 int cnt = 0;
b3ec716d 493 int this = first;
265e0f17 494 while (n--) {
b3ec716d 495 if (avail[this])
265e0f17 496 cnt++;
b3ec716d 497 this = (this+1) % raid_disks;
265e0f17
NB
498 }
499 if (cnt == 0)
500 return 0;
b3ec716d 501 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
502 } while (first != 0);
503 return 1;
e5329c37 504
df0d4ea0 505 case LEVEL_MULTIPATH:
e0d19036 506 return avail_disks>= 1;
df0d4ea0 507 case LEVEL_LINEAR:
64c4757e
NB
508 case 0:
509 return avail_disks == raid_disks;
510 case 1:
511 return avail_disks >= 1;
512 case 4:
330d6900
N
513 if (avail_disks == raid_disks - 1 &&
514 !avail[raid_disks - 1])
515 /* If just the parity device is missing, then we
516 * have enough, even if not clean
517 */
518 return 1;
519 /* FALL THROUGH */
64c4757e 520 case 5:
583315d9
NB
521 if (clean)
522 return avail_disks >= raid_disks-1;
523 else
524 return avail_disks >= raid_disks;
98c6faba 525 case 6:
583315d9
NB
526 if (clean)
527 return avail_disks >= raid_disks-2;
528 else
529 return avail_disks >= raid_disks;
64c4757e
NB
530 default:
531 return 0;
532 }
533}
534
8453e704
N
535int enough_fd(int fd)
536{
537 struct mdu_array_info_s array;
538 struct mdu_disk_info_s disk;
aabe020d 539 int i, rv;
8453e704
N
540 char *avail;
541
542 if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 ||
543 array.raid_disks <= 0)
544 return 0;
503975b9 545 avail = xcalloc(array.raid_disks, 1);
ca3b6696 546 for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704
N
547 disk.number = i;
548 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
549 continue;
9e6d9291
N
550 if (disk.major == 0 && disk.minor == 0)
551 continue;
552 array.nr_disks--;
553
8453e704
N
554 if (! (disk.state & (1<<MD_DISK_SYNC)))
555 continue;
556 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
557 continue;
8453e704
N
558 avail[disk.raid_disk] = 1;
559 }
560 /* This is used on an active array, so assume it is clean */
aabe020d 561 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 562 1, avail);
aabe020d
JS
563 free(avail);
564 return rv;
8453e704
N
565}
566
22e263f6
AC
567const int uuid_zero[4] = { 0, 0, 0, 0 };
568
f277ce36 569int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 570{
f277ce36
NB
571 if (swapuuid) {
572 /* parse uuids are hostendian.
573 * uuid's from some superblocks are big-ending
aba69144 574 * if there is a difference, we need to swap..
f277ce36
NB
575 */
576 unsigned char *ac = (unsigned char *)a;
577 unsigned char *bc = (unsigned char *)b;
578 int i;
ca3b6696 579 for (i = 0; i < 16; i += 4) {
f277ce36
NB
580 if (ac[i+0] != bc[i+3] ||
581 ac[i+1] != bc[i+2] ||
582 ac[i+2] != bc[i+1] ||
583 ac[i+3] != bc[i+0])
584 return 0;
585 }
586 return 1;
587 } else {
588 if (a[0]==b[0] &&
589 a[1]==b[1] &&
590 a[2]==b[2] &&
591 a[3]==b[3])
592 return 1;
593 return 0;
594 }
64c4757e 595}
ca3b6696 596
350f29f9
NB
597void copy_uuid(void *a, int b[4], int swapuuid)
598{
599 if (swapuuid) {
600 /* parse uuids are hostendian.
601 * uuid's from some superblocks are big-ending
602 * if there is a difference, we need to swap..
603 */
604 unsigned char *ac = (unsigned char *)a;
605 unsigned char *bc = (unsigned char *)b;
606 int i;
ca3b6696 607 for (i = 0; i < 16; i += 4) {
350f29f9
NB
608 ac[i+0] = bc[i+3];
609 ac[i+1] = bc[i+2];
610 ac[i+2] = bc[i+1];
611 ac[i+3] = bc[i+0];
612 }
613 } else
614 memcpy(a, b, 16);
615}
64c4757e 616
aae5a112 617char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 618{
9968e376 619 int i, j;
d7288ddc
N
620 char uuid[16];
621 char *c = buf;
622 strcpy(c, "UUID-");
623 c += strlen(c);
aae5a112 624 copy_uuid(uuid, id, swap);
9968e376 625 for (i = 0; i < 4; i++) {
9968e376 626 if (i)
ff54de6e 627 *c++ = sep;
9968e376
DW
628 for (j = 3; j >= 0; j--) {
629 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
630 c+= 2;
631 }
d7288ddc
N
632 }
633 return buf;
aae5a112
DW
634
635}
636
637char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
638{
86983cce
N
639 // dirty hack to work around an issue with super1 superblocks...
640 // super1 superblocks need swapuuid set in order for assembly to
641 // work, but can't have it set if we want this printout to match
642 // all the other uuid printouts in super1.c, so we force swapuuid
643 // to 1 to make our printout match the rest of super1
644 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
645}
646
435d4ebb 647#ifndef MDASSEMBLE
682c7051
NB
648int check_ext2(int fd, char *name)
649{
650 /*
651 * Check for an ext2fs file system.
652 * Superblock is always 1K at 1K offset
653 *
654 * s_magic is le16 at 56 == 0xEF53
655 * report mtime - le32 at 44
656 * blocks - le32 at 4
657 * logblksize - le32 at 24
658 */
659 unsigned char sb[1024];
660 time_t mtime;
2538ba2a
N
661 unsigned long long size;
662 int bsize;
682c7051
NB
663 if (lseek(fd, 1024,0)!= 1024)
664 return 0;
665 if (read(fd, sb, 1024)!= 1024)
666 return 0;
667 if (sb[56] != 0x53 || sb[57] != 0xef)
668 return 0;
669
670 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
671 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
672 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 673 size <<= bsize;
e7b84f9d 674 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 675 name);
2538ba2a 676 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
677 return 1;
678}
679
680int check_reiser(int fd, char *name)
681{
682 /*
683 * superblock is at 64K
684 * size is 1024;
685 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
686 *
687 */
688 unsigned char sb[1024];
2538ba2a 689 unsigned long long size;
682c7051
NB
690 if (lseek(fd, 64*1024, 0) != 64*1024)
691 return 0;
692 if (read(fd, sb, 1024) != 1024)
693 return 0;
ca3b6696
N
694 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
695 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 696 return 0;
e7b84f9d 697 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 698 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 699 cont_err("size = %lluK\n", size*4);
aba69144 700
682c7051
NB
701 return 1;
702}
703
704int check_raid(int fd, char *name)
705{
4b1ac34b 706 struct mdinfo info;
682c7051 707 time_t crtime;
d078d77c 708 char *level;
82d9eba6 709 struct supertype *st = guess_super(fd);
f9ce90ba 710
cb8f6859
N
711 if (!st)
712 return 0;
fa219dd2
MD
713 if (st->ss->add_to_super != NULL) {
714 st->ss->load_super(st, fd, name);
715 /* Looks like a raid array .. */
716 pr_err("%s appears to be part of a raid array:\n", name);
717 st->ss->getinfo_super(st, &info, NULL);
718 st->ss->free_super(st);
719 crtime = info.array.ctime;
720 level = map_num(pers, info.array.level);
721 if (!level)
722 level = "-unknown-";
723 cont_err("level=%s devices=%d ctime=%s",
724 level, info.array.raid_disks, ctime(&crtime));
725 } else {
726 /* Looks like GPT or MBR */
727 pr_err("partition table exists on %s\n", name);
728 }
82d9eba6 729 return 1;
682c7051
NB
730}
731
682c7051
NB
732int ask(char *mesg)
733{
734 char *add = "";
735 int i;
ca3b6696 736 for (i = 0; i < 5; i++) {
682c7051
NB
737 char buf[100];
738 fprintf(stderr, "%s%s", mesg, add);
739 fflush(stderr);
740 if (fgets(buf, 100, stdin)==NULL)
741 return 0;
742 if (buf[0]=='y' || buf[0]=='Y')
743 return 1;
744 if (buf[0]=='n' || buf[0]=='N')
745 return 0;
746 add = "(y/n) ";
747 }
e7b84f9d 748 pr_err("assuming 'no'\n");
682c7051
NB
749 return 0;
750}
435d4ebb 751#endif /* MDASSEMBLE */
682c7051 752
8d80900b 753int is_standard(char *dev, int *nump)
e5329c37
NB
754{
755 /* tests if dev is a "standard" md dev name.
756 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 757 * where NN is a string of digits
598f0d58
NB
758 * Returns 1 if a partitionable standard,
759 * -1 if non-partitonable,
760 * 0 if not a standard name.
e5329c37 761 */
8d80900b 762 char *d = strrchr(dev, '/');
ca3b6696 763 int type = 0;
8d80900b
NB
764 int num;
765 if (!d)
e5329c37 766 return 0;
ca3b6696
N
767 if (strncmp(d, "/d",2) == 0)
768 d += 2, type = 1; /* /dev/md/dN{pM} */
769 else if (strncmp(d, "/md_d", 5) == 0)
770 d += 5, type = 1; /* /dev/md_dN{pM} */
771 else if (strncmp(d, "/md", 3) == 0)
772 d += 3, type = -1; /* /dev/mdN */
773 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
774 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
775 else
776 return 0;
8d80900b 777 if (!*d)
e5329c37 778 return 0;
8d80900b
NB
779 num = atoi(d);
780 while (isdigit(*d))
781 d++;
782 if (*d)
e5329c37 783 return 0;
8d80900b
NB
784 if (nump) *nump = num;
785
786 return type;
e5329c37
NB
787}
788
4b1ac34b 789unsigned long calc_csum(void *super, int bytes)
82b27616 790{
56eb10c0 791 unsigned long long newcsum = 0;
82b27616 792 int i;
4b1ac34b
NB
793 unsigned int csum;
794 unsigned int *superc = (unsigned int*) super;
82b27616 795
ca3b6696
N
796 for(i = 0; i < bytes/4; i++)
797 newcsum += superc[i];
82b27616 798 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 799#ifdef __alpha__
aba69144 800/* The in-kernel checksum calculation is always 16bit on
570c0542 801 * the alpha, though it is 32 bit on i386...
ca3b6696 802 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
803 * a way that it shouldn't).
804 */
805 csum = (csum & 0xffff) + (csum >> 16);
806 csum = (csum & 0xffff) + (csum >> 16);
807#endif
82b27616
NB
808 return csum;
809}
cd29a5c8 810
435d4ebb 811#ifndef MDASSEMBLE
56eb10c0 812char *human_size(long long bytes)
cd29a5c8
NB
813{
814 static char buf[30];
d5d3721e
NB
815
816 /* We convert bytes to either centi-M{ega,ibi}bytes or
817 * centi-G{igi,ibi}bytes, with appropriate rounding,
818 * and then print 1/100th of those as a decimal.
819 * We allow upto 2048Megabytes before converting to
820 * gigabytes, as that shows more precision and isn't
821 * too large a number.
01619b48 822 * Terabytes are not yet handled.
d5d3721e 823 */
cd29a5c8 824
56eb10c0 825 if (bytes < 5000*1024)
ca3b6696 826 buf[0] = 0;
d5d3721e 827 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 828 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 829 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 830 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
831 cMiB/100 , cMiB % 100,
832 cMB/100, cMB % 100);
833 } else {
93d3bd3b 834 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 835 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 836 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
837 cGiB/100 , cGiB % 100,
838 cGB/100, cGB % 100);
839 }
cd29a5c8
NB
840 return buf;
841}
e0d19036 842
f0ec6710 843char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
844{
845 static char buf[30];
e0d19036 846
570abc6f
MN
847 /* We convert bytes to either centi-M{ega,ibi}bytes or
848 * centi-G{igi,ibi}bytes, with appropriate rounding,
849 * and then print 1/100th of those as a decimal.
850 * We allow upto 2048Megabytes before converting to
851 * gigabytes, as that shows more precision and isn't
852 * too large a number.
853 * Terabytes are not yet handled.
f0ec6710
MN
854 *
855 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
856 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
857 */
858
e0d19036 859 if (bytes < 5000*1024)
570abc6f 860 buf[0] = 0;
f0ec6710
MN
861 else if (prefix == IEC) {
862 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 863 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710
MN
864 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
865 cMiB/100 , cMiB % 100);
866 } else {
93d3bd3b 867 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710
MN
868 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
869 cGiB/100 , cGiB % 100);
870 }
871 }
872 else if (prefix == JEDEC) {
873 if (bytes < 2*1024LL*1024LL*1024LL) {
874 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
875 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
876 cMB/100, cMB % 100);
877 } else {
878 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
879 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
880 cGB/100 , cGB % 100);
881 }
570abc6f 882 }
f0ec6710
MN
883 else
884 buf[0] = 0;
570abc6f 885
e0d19036
NB
886 return buf;
887}
e4965ef8
N
888
889void print_r10_layout(int layout)
890{
891 int near = layout & 255;
892 int far = (layout >> 8) & 255;
893 int offset = (layout&0x10000);
894 char *sep = "";
895
896 if (near != 1) {
897 printf("%s near=%d", sep, near);
898 sep = ",";
899 }
900 if (far != 1)
901 printf("%s %s=%d", sep, offset?"offset":"far", far);
902 if (near*far == 1)
903 printf("NO REDUNDANCY");
904}
435d4ebb 905#endif
e0d19036 906
5f8097be
NB
907unsigned long long calc_array_size(int level, int raid_disks, int layout,
908 int chunksize, unsigned long long devsize)
577e8448 909{
e03640bd
CA
910 if (level == 1)
911 return devsize;
577e8448
AK
912 devsize &= ~(unsigned long long)((chunksize>>9)-1);
913 return get_data_disks(level, layout, raid_disks) * devsize;
914}
915
916int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
917{
918 int data_disks = 0;
919 switch (level) {
ca3b6696
N
920 case 0: data_disks = raid_disks;
921 break;
922 case 1: data_disks = 1;
923 break;
5f8097be 924 case 4:
ca3b6696
N
925 case 5: data_disks = raid_disks - 1;
926 break;
927 case 6: data_disks = raid_disks - 2;
928 break;
5f8097be
NB
929 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
930 break;
931 }
577e8448
AK
932
933 return data_disks;
5f8097be
NB
934}
935
13db17bd 936dev_t devnm2devid(char *devnm)
4dd2df09
N
937{
938 /* First look in /sys/block/$DEVNM/dev for %d:%d
939 * If that fails, try parsing out a number
940 */
941 char path[100];
942 char *ep;
943 int fd;
944 int mjr,mnr;
945
946 sprintf(path, "/sys/block/%s/dev", devnm);
947 fd = open(path, O_RDONLY);
948 if (fd >= 0) {
949 char buf[20];
950 int n = read(fd, buf, sizeof(buf));
951 close(fd);
952 if (n > 0)
953 buf[n] = 0;
954 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
955 return makedev(mjr, mnr);
956 }
957 if (strncmp(devnm, "md_d", 4) == 0 &&
958 isdigit(devnm[4]) &&
959 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
960 ep > devnm && *ep == 0)
961 return makedev(get_mdp_major(), mnr << MdpMinorShift);
962
963 if (strncmp(devnm, "md", 2) == 0 &&
964 isdigit(devnm[2]) &&
965 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
966 ep > devnm && *ep == 0)
967 return makedev(MD_MAJOR, mnr);
968
969 return 0;
970}
971
5a23a06e 972#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 973char *get_md_name(char *devnm)
e0d19036
NB
974{
975 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 976 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 977
e0d19036
NB
978 static char devname[50];
979 struct stat stb;
4dd2df09 980 dev_t rdev = devnm2devid(devnm);
dd0781e5 981 char *dn;
98c6faba 982
4dd2df09
N
983 if (rdev == 0)
984 return 0;
985 if (strncmp(devnm, "md_", 3) == 0) {
986 snprintf(devname, sizeof(devname), "/dev/md/%s",
987 devnm + 3);
98c6faba
NB
988 if (stat(devname, &stb) == 0
989 && (S_IFMT&stb.st_mode) == S_IFBLK
990 && (stb.st_rdev == rdev))
991 return devname;
992 }
4dd2df09
N
993 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
994 if (stat(devname, &stb) == 0
995 && (S_IFMT&stb.st_mode) == S_IFBLK
996 && (stb.st_rdev == rdev))
997 return devname;
998
999 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
1000 if (stat(devname, &stb) == 0
1001 && (S_IFMT&stb.st_mode) == S_IFBLK
1002 && (stb.st_rdev == rdev))
1003 return devname;
1004
16c6fa80 1005 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
1006 if (dn)
1007 return dn;
4dd2df09 1008 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 1009 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
1010 if (errno != EEXIST)
1011 return NULL;
e0d19036
NB
1012
1013 if (stat(devname, &stb) == 0
1014 && (S_IFMT&stb.st_mode) == S_IFBLK
1015 && (stb.st_rdev == rdev))
1016 return devname;
1017 unlink(devname);
1018 return NULL;
1019}
1020
1021void put_md_name(char *name)
1022{
ca3b6696 1023 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
1024 unlink(name);
1025}
435d4ebb 1026#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 1027
cc742d38
N
1028int get_maj_min(char *dev, int *major, int *minor)
1029{
1030 char *e;
1031 *major = strtoul(dev, &e, 0);
1032 return (e > dev && *e == ':' && e[1] &&
1033 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1034 *e == 0);
1035}
1036
8b0dabea
NB
1037int dev_open(char *dev, int flags)
1038{
1039 /* like 'open', but if 'dev' matches %d:%d, create a temp
1040 * block device and open that
1041 */
8b0dabea
NB
1042 int fd = -1;
1043 char devname[32];
e81cdd9f 1044 int major;
8b0dabea 1045 int minor;
e81cdd9f 1046
9f0ad56b
JS
1047 if (!dev)
1048 return -1;
6df6a774 1049 flags |= O_DIRECT;
e81cdd9f 1050
cc742d38 1051 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1052 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1053 (int)getpid(), major, minor);
1054 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1055 fd = open(devname, flags);
1056 unlink(devname);
6df6a774
N
1057 }
1058 if (fd < 0) {
9dc7d357 1059 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
1060 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
1061 (int)getpid(), major, minor);
ca3b6696 1062 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
1063 fd = open(devname, flags);
1064 unlink(devname);
1065 }
8b0dabea
NB
1066 }
1067 } else
6df6a774 1068 fd = open(dev, flags);
8b0dabea
NB
1069 return fd;
1070}
f9ce90ba 1071
4dd2df09 1072int open_dev_flags(char *devnm, int flags)
e8a70c89 1073{
13db17bd 1074 dev_t devid;
e8a70c89
N
1075 char buf[20];
1076
4dd2df09
N
1077 devid = devnm2devid(devnm);
1078 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1079 return dev_open(buf, flags);
1080}
1081
4dd2df09 1082int open_dev(char *devnm)
d998b738 1083{
4dd2df09 1084 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1085}
1086
4dd2df09 1087int open_dev_excl(char *devnm)
a931db9e
NB
1088{
1089 char buf[20];
1090 int i;
7187750e 1091 int flags = O_RDWR;
13db17bd 1092 dev_t devid = devnm2devid(devnm);
399e0b97 1093 long delay = 1000;
a931db9e 1094
4dd2df09 1095 sprintf(buf, "%d:%d", major(devid), minor(devid));
ca3b6696 1096 for (i = 0 ; i < 25 ; i++) {
7187750e 1097 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1098 if (fd >= 0)
1099 return fd;
7187750e
N
1100 if (errno == EACCES && flags == O_RDWR) {
1101 flags = O_RDONLY;
1102 continue;
1103 }
a931db9e
NB
1104 if (errno != EBUSY)
1105 return fd;
399e0b97
N
1106 usleep(delay);
1107 if (delay < 200000)
1108 delay *= 2;
a931db9e
NB
1109 }
1110 return -1;
1111}
1112
9008ed1c
N
1113int same_dev(char *one, char *two)
1114{
1115 struct stat st1, st2;
1116 if (stat(one, &st1) != 0)
1117 return 0;
1118 if (stat(two, &st2) != 0)
1119 return 0;
1120 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1121 return 0;
1122 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1123 return 0;
1124 return st1.st_rdev == st2.st_rdev;
1125}
1126
a7c6e3fb 1127void wait_for(char *dev, int fd)
a714580e
N
1128{
1129 int i;
a7c6e3fb 1130 struct stat stb_want;
399e0b97 1131 long delay = 1000;
a7c6e3fb
N
1132
1133 if (fstat(fd, &stb_want) != 0 ||
1134 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1135 return;
a714580e 1136
ca3b6696 1137 for (i = 0 ; i < 25 ; i++) {
a714580e 1138 struct stat stb;
a7c6e3fb
N
1139 if (stat(dev, &stb) == 0 &&
1140 (stb.st_mode & S_IFMT) == S_IFBLK &&
1141 (stb.st_rdev == stb_want.st_rdev))
a714580e 1142 return;
399e0b97
N
1143 usleep(delay);
1144 if (delay < 200000)
1145 delay *= 2;
a714580e 1146 }
436305c6 1147 if (i == 25)
1ade5cc1 1148 dprintf("timeout waiting for %s\n", dev);
a714580e
N
1149}
1150
0f22b998
N
1151struct superswitch *superlist[] =
1152{
1153 &super0, &super1,
1154 &super_ddf, &super_imsm,
0592faeb 1155 &mbr, &gpt,
0f22b998 1156 NULL };
f9ce90ba 1157
ea24acd0 1158#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1159
4725bc31 1160struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1161{
1686dc25
NB
1162 mdu_array_info_t array;
1163 int vers;
1164 int minor;
1165 struct supertype *st = NULL;
7e0f6979 1166 struct mdinfo *sra;
142cb9e1 1167 char *verstr;
1686dc25
NB
1168 char version[20];
1169 int i;
f7e7067b 1170 char *subarray = NULL;
4dd2df09 1171 char container[32] = "";
1686dc25 1172
4dd2df09 1173 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1174
1175 if (sra) {
7e0f6979
NB
1176 vers = sra->array.major_version;
1177 minor = sra->array.minor_version;
142cb9e1 1178 verstr = sra->text_version;
1686dc25
NB
1179 } else {
1180 if (ioctl(fd, GET_ARRAY_INFO, &array))
1181 array.major_version = array.minor_version = 0;
1182 vers = array.major_version;
1183 minor = array.minor_version;
142cb9e1 1184 verstr = "";
6fbba4c9 1185 }
82d9eba6 1186
1686dc25
NB
1187 if (vers != -1) {
1188 sprintf(version, "%d.%d", vers, minor);
1189 verstr = version;
6fbba4c9 1190 }
3c558363 1191 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1192 char *dev = verstr+1;
5f7e44b2 1193
f7e7067b 1194 subarray = strchr(dev, '/');
92d49ecf 1195 if (subarray) {
f7e7067b 1196 *subarray++ = '\0';
503975b9 1197 subarray = xstrdup(subarray);
92d49ecf 1198 }
4dd2df09 1199 strcpy(container, dev);
15d230f7 1200 sysfs_free(sra);
5f7e44b2 1201 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1202 if (sra && sra->text_version[0])
1203 verstr = sra->text_version;
1204 else
1205 verstr = "-no-metadata-";
f7e7067b
NB
1206 }
1207
1208 for (i = 0; st == NULL && superlist[i] ; i++)
1209 st = superlist[i]->match_metadata_desc(verstr);
1686dc25 1210
15d230f7 1211 sysfs_free(sra);
f7e7067b 1212 if (st) {
3b0896f8 1213 st->sb = NULL;
1f49fb3a
N
1214 if (subarrayp)
1215 *subarrayp = subarray;
4dd2df09
N
1216 strcpy(st->container_devnm, container);
1217 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1218 } else
1219 free(subarray);
5f7e44b2 1220
82d9eba6 1221 return st;
f9ce90ba 1222}
ea24acd0
NB
1223#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1224
0f0749ad 1225int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1226{
1227 char buf[20];
1228 int fd;
1229
1230 sprintf(buf, "%d:%d", major(id), minor(id));
1231 fd = dev_open(buf, O_RDONLY);
1232 if (fd < 0)
1233 return 0;
1234 if (get_dev_size(fd, NULL, size)) {
1235 close(fd);
1236 return 1;
1237 }
1238 close(fd);
1239 return 0;
1240}
f9ce90ba 1241
159c3a1a 1242struct supertype *dup_super(struct supertype *orig)
3da92f27 1243{
159c3a1a 1244 struct supertype *st;
1686dc25 1245
d2ca6449
NB
1246 if (!orig)
1247 return orig;
503975b9 1248 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1249 st->ss = orig->ss;
1250 st->max_devs = orig->max_devs;
1251 st->minor_version = orig->minor_version;
cb8f6859 1252 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1253 st->data_offset = orig->data_offset;
159c3a1a
NB
1254 st->sb = NULL;
1255 st->info = NULL;
1256 return st;
3da92f27
NB
1257}
1258
54887ad8 1259struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1260{
1261 /* try each load_super to find the best match,
1262 * and return the best superswitch
1263 */
82d9eba6
NB
1264 struct superswitch *ss;
1265 struct supertype *st;
1d13b599 1266 unsigned int besttime = 0;
570c0542 1267 int bestsuper = -1;
f9ce90ba
NB
1268 int i;
1269
503975b9 1270 st = xcalloc(1, sizeof(*st));
4dd2df09 1271 st->container_devnm[0] = 0;
d1d599ea 1272
ca3b6696 1273 for (i = 0 ; superlist[i]; i++) {
f9ce90ba
NB
1274 int rv;
1275 ss = superlist[i];
54887ad8
N
1276 if (guess_type == guess_array && ss->add_to_super == NULL)
1277 continue;
1278 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1279 continue;
ef609477 1280 memset(st, 0, sizeof(*st));
df3346e6 1281 st->ignore_hw_compat = 1;
3da92f27 1282 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1283 if (rv == 0) {
1284 struct mdinfo info;
a5d85af7 1285 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1286 if (bestsuper == -1 ||
1287 besttime < info.array.ctime) {
1288 bestsuper = i;
1289 besttime = info.array.ctime;
570c0542 1290 }
3da92f27 1291 ss->free_super(st);
570c0542
NB
1292 }
1293 }
1294 if (bestsuper != -1) {
1295 int rv;
ef609477 1296 memset(st, 0, sizeof(*st));
df3346e6 1297 st->ignore_hw_compat = 1;
3da92f27 1298 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1299 if (rv == 0) {
5e747af2 1300 superlist[bestsuper]->free_super(st);
82d9eba6 1301 return st;
f9ce90ba
NB
1302 }
1303 }
570c0542 1304 free(st);
f9ce90ba
NB
1305 return NULL;
1306}
fe6729fa 1307
beae1dfe
NB
1308/* Return size of device in bytes */
1309int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1310{
1311 unsigned long long ldsize;
c2c9bb6f
NB
1312 struct stat st;
1313
1314 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1315 ldsize = (unsigned long long)st.st_size;
1316 else
beae1dfe
NB
1317#ifdef BLKGETSIZE64
1318 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1319#endif
1320 {
1321 unsigned long dsize;
1322 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1323 ldsize = dsize;
1324 ldsize <<= 9;
1325 } else {
1326 if (dname)
32971509 1327 pr_err("Cannot get size of %s: %s\n",
beae1dfe
NB
1328 dname, strerror(errno));
1329 return 0;
1330 }
1331 }
1332 *sizep = ldsize;
1333 return 1;
1334}
8fac0577 1335
32971509
PB
1336/* Return sector size of device in bytes */
1337int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep)
1338{
1339 unsigned int sectsize;
1340
1341 if (ioctl(fd, BLKSSZGET, &sectsize) != 0) {
1342 if (dname)
1343 pr_err("Cannot get sector size of %s: %s\n",
1344 dname, strerror(errno));
1345 return 0;
1346 }
1347
1348 *sectsizep = sectsize;
1349 return 1;
1350}
1351
3a371610
N
1352/* Return true if this can only be a container, not a member device.
1353 * i.e. is and md device and size is zero
1354 */
1355int must_be_container(int fd)
1356{
1357 unsigned long long size;
1358 if (md_get_version(fd) < 0)
1359 return 0;
1360 if (get_dev_size(fd, NULL, &size) == 0)
1361 return 1;
1362 if (size == 0)
1363 return 1;
1364 return 0;
1365}
034b203a
TM
1366
1367/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1368 * Returns: 1 if successful
1369 * -1 for unknown partition type
1370 * 0 for other errors
1371 */
1372static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1373{
056b331e 1374 struct GPT gpt;
034b203a 1375 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1376 struct GPT_part_entry *part;
1377 char buf[512];
034b203a
TM
1378 unsigned long long curr_part_end;
1379 unsigned all_partitions, entry_size;
f21e18ca 1380 unsigned part_nr;
034b203a
TM
1381
1382 *endofpart = 0;
1383
056b331e 1384 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1385 /* skip protective MBR */
034b203a 1386 lseek(fd, 512, SEEK_SET);
73e658d8 1387 /* read GPT header */
056b331e 1388 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1389 return 0;
1390
1391 /* get the number of partition entries and the entry size */
056b331e
N
1392 all_partitions = __le32_to_cpu(gpt.part_cnt);
1393 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1394
1395 /* Check GPT signature*/
056b331e 1396 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1397 return -1;
1398
1399 /* sanity checks */
1400 if (all_partitions > 1024 ||
db7fdfe4 1401 entry_size > sizeof(buf))
034b203a
TM
1402 return -1;
1403
db7fdfe4
JS
1404 part = (struct GPT_part_entry *)buf;
1405
ca3b6696 1406 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1407 /* read partition entry */
db7fdfe4 1408 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1409 return 0;
1410
034b203a 1411 /* is this valid partition? */
db7fdfe4 1412 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1413 /* check the last lba for the current partition */
db7fdfe4 1414 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1415 if (curr_part_end > *endofpart)
1416 *endofpart = curr_part_end;
1417 }
1418
034b203a
TM
1419 }
1420 return 1;
1421}
1422
1423/* Sets endofpart parameter to the last block used by the last partition on the device.
1424 * Returns: 1 if successful
1425 * -1 for unknown partition type
1426 * 0 for other errors
1427 */
1428static int get_last_partition_end(int fd, unsigned long long *endofpart)
1429{
056b331e 1430 struct MBR boot_sect;
034b203a 1431 unsigned long long curr_part_end;
f21e18ca 1432 unsigned part_nr;
034b203a
TM
1433 int retval = 0;
1434
1435 *endofpart = 0;
1436
056b331e 1437 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1438 /* read MBR */
1439 lseek(fd, 0, 0);
056b331e 1440 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1441 goto abort;
1442
1443 /* check MBP signature */
056b331e 1444 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1445 retval = 1;
1446 /* found the correct signature */
034b203a 1447
ca3b6696 1448 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
8e2bca51
JC
1449 /*
1450 * Have to make every access through boot_sect rather
1451 * than using a pointer to the partition table (or an
1452 * entry), since the entries are not properly aligned.
1453 */
1454
034b203a 1455 /* check for GPT type */
8e2bca51
JC
1456 if (boot_sect.parts[part_nr].part_type ==
1457 MBR_GPT_PARTITION_TYPE) {
034b203a
TM
1458 retval = get_gpt_last_partition_end(fd, endofpart);
1459 break;
1460 }
1461 /* check the last used lba for the current partition */
8e2bca51
JC
1462 curr_part_end =
1463 __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) +
1464 __le32_to_cpu(boot_sect.parts[part_nr].blocks_num);
034b203a
TM
1465 if (curr_part_end > *endofpart)
1466 *endofpart = curr_part_end;
034b203a
TM
1467 }
1468 } else {
1469 /* Unknown partition table */
1470 retval = -1;
1471 }
1472 abort:
1473 return retval;
1474}
1475
53ed6ac3
KW
1476int check_partitions(int fd, char *dname, unsigned long long freesize,
1477 unsigned long long size)
034b203a
TM
1478{
1479 /*
1480 * Check where the last partition ends
1481 */
1482 unsigned long long endofpart;
1483 int ret;
1484
1485 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1486 /* There appears to be a partition table here */
1487 if (freesize == 0) {
1488 /* partitions will not be visible in new device */
e7b84f9d
N
1489 pr_err("partition table exists on %s but will be lost or\n"
1490 " meaningless after creating array\n",
1491 dname);
034b203a
TM
1492 return 1;
1493 } else if (endofpart > freesize) {
1494 /* last partition overlaps metadata */
e7b84f9d
N
1495 pr_err("metadata will over-write last partition on %s.\n",
1496 dname);
034b203a 1497 return 1;
53ed6ac3
KW
1498 } else if (size && endofpart > size) {
1499 /* partitions will be truncated in new device */
e7b84f9d
N
1500 pr_err("array size is too small to cover all partitions on %s.\n",
1501 dname);
53ed6ac3 1502 return 1;
034b203a
TM
1503 }
1504 }
1505 return 0;
1506}
1507
a322f70c
DW
1508int open_container(int fd)
1509{
1510 /* 'fd' is a block device. Find out if it is in use
1511 * by a container, and return an open fd on that container.
1512 */
1513 char path[256];
1514 char *e;
1515 DIR *dir;
1516 struct dirent *de;
1517 int dfd, n;
1518 char buf[200];
1519 int major, minor;
1520 struct stat st;
1521
1522 if (fstat(fd, &st) != 0)
1523 return -1;
1524 sprintf(path, "/sys/dev/block/%d:%d/holders",
1525 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1526 e = path + strlen(path);
1527
1528 dir = opendir(path);
1529 if (!dir)
1530 return -1;
1531 while ((de = readdir(dir))) {
1532 if (de->d_ino == 0)
1533 continue;
1534 if (de->d_name[0] == '.')
1535 continue;
a994592d
N
1536 /* Need to make sure it is a container and not a volume */
1537 sprintf(e, "/%s/md/metadata_version", de->d_name);
1538 dfd = open(path, O_RDONLY);
1539 if (dfd < 0)
1540 continue;
1541 n = read(dfd, buf, sizeof(buf));
1542 close(dfd);
1543 if (n <= 0 || (unsigned)n >= sizeof(buf))
1544 continue;
1545 buf[n] = 0;
1546 if (strncmp(buf, "external", 8) != 0 ||
1547 n < 10 ||
1548 buf[9] == '/')
1549 continue;
a322f70c
DW
1550 sprintf(e, "/%s/dev", de->d_name);
1551 dfd = open(path, O_RDONLY);
1552 if (dfd < 0)
1553 continue;
1554 n = read(dfd, buf, sizeof(buf));
1555 close(dfd);
f21e18ca 1556 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1557 continue;
1558 buf[n] = 0;
1559 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1560 continue;
1561 sprintf(buf, "%d:%d", major, minor);
1562 dfd = dev_open(buf, O_RDONLY);
1563 if (dfd >= 0) {
1564 closedir(dir);
1565 return dfd;
1566 }
1567 }
355726fa 1568 closedir(dir);
a322f70c
DW
1569 return -1;
1570}
1571
33414a01
DW
1572struct superswitch *version_to_superswitch(char *vers)
1573{
1574 int i;
1575
1576 for (i = 0; superlist[i]; i++) {
1577 struct superswitch *ss = superlist[i];
1578
1579 if (strcmp(vers, ss->name) == 0)
1580 return ss;
1581 }
1582
1583 return NULL;
1584}
1585
4dd2df09
N
1586int metadata_container_matches(char *metadata, char *devnm)
1587{
1588 /* Check if 'devnm' is the container named in 'metadata'
1589 * which is
1590 * /containername/componentname or
1591 * -containername/componentname
1592 */
1593 int l;
1594 if (*metadata != '/' && *metadata != '-')
1595 return 0;
1596 l = strlen(devnm);
1597 if (strncmp(metadata+1, devnm, l) != 0)
1598 return 0;
1599 if (metadata[l+1] != '/')
1600 return 0;
1601 return 1;
1602}
1603
1604int metadata_subdev_matches(char *metadata, char *devnm)
1605{
1606 /* Check if 'devnm' is the subdev named in 'metadata'
1607 * which is
1608 * /containername/subdev or
1609 * -containername/subdev
1610 */
1611 char *sl;
1612 if (*metadata != '/' && *metadata != '-')
1613 return 0;
1614 sl = strchr(metadata+1, '/');
1615 if (!sl)
1616 return 0;
1617 if (strcmp(sl+1, devnm) == 0)
1618 return 1;
1619 return 0;
1620}
1621
33414a01
DW
1622int is_container_member(struct mdstat_ent *mdstat, char *container)
1623{
1624 if (mdstat->metadata_version == NULL ||
1625 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1626 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1627 return 0;
1628
1629 return 1;
1630}
1631
1632int is_subarray_active(char *subarray, char *container)
1633{
1634 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1635 struct mdstat_ent *ent;
1636
e5408a32
DW
1637 for (ent = mdstat; ent; ent = ent->next)
1638 if (is_container_member(ent, container))
e5e5d7ce 1639 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1640 break;
33414a01
DW
1641
1642 free_mdstat(mdstat);
1643
1644 return ent != NULL;
1645}
1646
1647/* open_subarray - opens a subarray in a container
1648 * @dev: container device name
feab51f8 1649 * @st: empty supertype
33414a01
DW
1650 * @quiet: block reporting errors flag
1651 *
1652 * On success returns an fd to a container and fills in *st
1653 */
feab51f8 1654int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1655{
1656 struct mdinfo *mdi;
a951a4f7 1657 struct mdinfo *info;
33414a01 1658 int fd, err = 1;
4dd2df09 1659 char *_devnm;
33414a01
DW
1660
1661 fd = open(dev, O_RDWR|O_EXCL);
1662 if (fd < 0) {
1663 if (!quiet)
e7b84f9d 1664 pr_err("Couldn't open %s, aborting\n",
33414a01 1665 dev);
b990032d 1666 return -1;
33414a01
DW
1667 }
1668
4dd2df09
N
1669 _devnm = fd2devnm(fd);
1670 if (_devnm == NULL) {
33414a01 1671 if (!quiet)
e7b84f9d
N
1672 pr_err("Failed to determine device number for %s\n",
1673 dev);
33414a01
DW
1674 goto close_fd;
1675 }
4dd2df09 1676 strcpy(st->devnm, _devnm);
33414a01 1677
4dd2df09 1678 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1679 if (!mdi) {
1680 if (!quiet)
e7b84f9d 1681 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1682 dev);
1683 goto close_fd;
1684 }
1685
1686 if (mdi->array.level != UnSet) {
1687 if (!quiet)
e7b84f9d 1688 pr_err("%s is not a container\n", dev);
33414a01
DW
1689 goto free_sysfs;
1690 }
1691
1692 st->ss = version_to_superswitch(mdi->text_version);
1693 if (!st->ss) {
1694 if (!quiet)
e7b84f9d
N
1695 pr_err("Operation not supported for %s metadata\n",
1696 mdi->text_version);
33414a01
DW
1697 goto free_sysfs;
1698 }
1699
4dd2df09 1700 if (st->devnm[0] == 0) {
33414a01 1701 if (!quiet)
e7b84f9d 1702 pr_err("Failed to allocate device name\n");
33414a01
DW
1703 goto free_sysfs;
1704 }
1705
db20d413 1706 if (!st->ss->load_container) {
33414a01 1707 if (!quiet)
e7b84f9d 1708 pr_err("%s is not a container\n", dev);
4dd2df09 1709 goto free_sysfs;
33414a01
DW
1710 }
1711
db20d413 1712 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1713 if (!quiet)
e7b84f9d 1714 pr_err("Failed to load metadata for %s\n",
db20d413 1715 dev);
4dd2df09 1716 goto free_sysfs;
33414a01
DW
1717 }
1718
a951a4f7
N
1719 info = st->ss->container_content(st, subarray);
1720 if (!info) {
1721 if (!quiet)
e7b84f9d 1722 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1723 subarray, dev);
1724 goto free_super;
1725 }
1726 free(info);
1727
33414a01
DW
1728 err = 0;
1729
1730 free_super:
1731 if (err)
1732 st->ss->free_super(st);
33414a01
DW
1733 free_sysfs:
1734 sysfs_free(mdi);
1735 close_fd:
1736 if (err)
1737 close(fd);
1738
1739 if (err)
1740 return -1;
1741 else
1742 return fd;
1743}
1744
7801ac20
N
1745int add_disk(int mdfd, struct supertype *st,
1746 struct mdinfo *sra, struct mdinfo *info)
1747{
1748 /* Add a device to an array, in one of 2 ways. */
1749 int rv;
1750#ifndef MDASSEMBLE
1751 if (st->ss->external) {
d23534e4
DW
1752 if (info->disk.state & (1<<MD_DISK_SYNC))
1753 info->recovery_start = MaxSector;
1754 else
1755 info->recovery_start = 0;
2904b26f 1756 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1757 if (! rv) {
1758 struct mdinfo *sd2;
f35f2525
N
1759 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1760 if (sd2 == info)
1761 break;
1762 if (sd2 == NULL) {
503975b9 1763 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1764 *sd2 = *info;
1765 sd2->next = sra->devs;
1766 sra->devs = sd2;
1767 }
7801ac20
N
1768 }
1769 } else
1770#endif
1771 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1772 return rv;
1773}
1774
de6ae750
N
1775int remove_disk(int mdfd, struct supertype *st,
1776 struct mdinfo *sra, struct mdinfo *info)
1777{
1778 int rv;
1779 /* Remove the disk given by 'info' from the array */
1780#ifndef MDASSEMBLE
1781 if (st->ss->external)
1782 rv = sysfs_set_str(sra, info, "slot", "none");
1783 else
1784#endif
1785 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1786 info->disk.minor));
1787 return rv;
1788}
1789
f35f2525
N
1790int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1791{
1792 /* Initialise kernel's knowledge of array.
1793 * This varies between externally managed arrays
1794 * and older kernels
1795 */
1796 int vers = md_get_version(mdfd);
1797 int rv;
1798
1799#ifndef MDASSEMBLE
1800 if (st->ss->external)
1801 rv = sysfs_set_array(info, vers);
1802 else
1803#endif
1804 if ((vers % 100) >= 1) { /* can use different versions */
1805 mdu_array_info_t inf;
1806 memset(&inf, 0, sizeof(inf));
1807 inf.major_version = info->array.major_version;
1808 inf.minor_version = info->array.minor_version;
1809 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1810 } else
1811 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1812 return rv;
1813}
1814
1e5c6983
DW
1815unsigned long long min_recovery_start(struct mdinfo *array)
1816{
1817 /* find the minimum recovery_start in an array for metadata
1818 * formats that only record per-array recovery progress instead
1819 * of per-device
1820 */
1821 unsigned long long recovery_start = MaxSector;
1822 struct mdinfo *d;
1823
1824 for (d = array->devs; d; d = d->next)
1825 recovery_start = min(recovery_start, d->recovery_start);
1826
1827 return recovery_start;
1828}
1829
4dd2df09 1830int mdmon_pid(char *devnm)
a931db9e
NB
1831{
1832 char path[100];
1833 char pid[10];
1834 int fd;
1835 int n;
10013317 1836
4dd2df09 1837 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1838
24f6f99b 1839 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1840
1841 if (fd < 0)
cf556303 1842 return -1;
a931db9e
NB
1843 n = read(fd, pid, 9);
1844 close(fd);
1845 if (n <= 0)
cf556303 1846 return -1;
24f6f99b 1847 return atoi(pid);
a931db9e
NB
1848}
1849
4dd2df09 1850int mdmon_running(char *devnm)
a931db9e 1851{
4dd2df09 1852 int pid = mdmon_pid(devnm);
24f6f99b 1853 if (pid <= 0)
a931db9e 1854 return 0;
24f6f99b 1855 if (kill(pid, 0) == 0)
a931db9e
NB
1856 return 1;
1857 return 0;
1858}
1859
4dd2df09 1860int start_mdmon(char *devnm)
8850ee3e 1861{
1913c325 1862 int i, skipped;
44d2e365 1863 int len;
ca3b6696 1864 pid_t pid;
9fe32043 1865 int status;
44d2e365
N
1866 char pathbuf[1024];
1867 char *paths[4] = {
1868 pathbuf,
85945e19 1869 BINDIR "/mdmon",
2f1bcf43 1870 "./mdmon",
44d2e365
N
1871 NULL
1872 };
8850ee3e 1873
40ebbb9c 1874 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1875 return 0;
1876
9cf014ec 1877 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1878 if (len > 0) {
1879 char *sl;
1880 pathbuf[len] = 0;
1881 sl = strrchr(pathbuf, '/');
1882 if (sl)
1883 sl++;
1884 else
1885 sl = pathbuf;
1886 strcpy(sl, "mdmon");
1887 } else
1888 pathbuf[0] = '\0';
1889
0f7bdf89 1890 /* First try to run systemctl */
9540cc24
N
1891 if (!check_env("MDADM_NO_SYSTEMCTL"))
1892 switch(fork()) {
1893 case 0:
1894 /* FIXME yuk. CLOSE_EXEC?? */
1895 skipped = 0;
1896 for (i = 3; skipped < 20; i++)
1897 if (close(i) < 0)
1898 skipped++;
1899 else
1900 skipped = 0;
1901
1902 /* Don't want to see error messages from
1903 * systemctl. If the service doesn't exist,
1904 * we start mdmon ourselves.
1905 */
1906 close(2);
1907 open("/dev/null", O_WRONLY);
1908 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1909 devnm);
1910 status = execl("/usr/bin/systemctl", "systemctl",
1911 "start",
1912 pathbuf, NULL);
1913 status = execl("/bin/systemctl", "systemctl", "start",
1914 pathbuf, NULL);
1915 exit(1);
7a862a02 1916 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1917 return -1;
1918 default: /* parent - good */
1919 pid = wait(&status);
1920 if (pid >= 0 && status == 0)
1921 return 0;
1922 }
0f7bdf89
JS
1923
1924 /* That failed, try running mdmon directly */
8850ee3e
N
1925 switch(fork()) {
1926 case 0:
1927 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1928 skipped = 0;
ca3b6696 1929 for (i = 3; skipped < 20; i++)
1913c325
N
1930 if (close(i) < 0)
1931 skipped++;
1932 else
1933 skipped = 0;
1934
ca3b6696 1935 for (i = 0; paths[i]; i++)
a0963a86 1936 if (paths[i][0]) {
2f1bcf43 1937 execl(paths[i], paths[i],
4dd2df09 1938 devnm, NULL);
a0963a86 1939 }
8850ee3e 1940 exit(1);
7a862a02 1941 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 1942 return -1;
9fe32043
N
1943 default: /* parent - good */
1944 pid = wait(&status);
15c10423 1945 if (pid < 0 || status != 0) {
7a862a02 1946 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 1947 return -1;
15c10423 1948 }
8850ee3e
N
1949 }
1950 return 0;
1951}
1952
148acb7b
DW
1953__u32 random32(void)
1954{
1955 __u32 rv;
1956 int rfd = open("/dev/urandom", O_RDONLY);
1957 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1958 rv = random();
1959 if (rfd >= 0)
1960 close(rfd);
1961 return rv;
1962}
1963
c5f71c24
JS
1964void random_uuid(__u8 *buf)
1965{
1966 int fd, i, len;
1967 __u32 r[4];
1968
1969 fd = open("/dev/urandom", O_RDONLY);
1970 if (fd < 0)
1971 goto use_random;
1972 len = read(fd, buf, 16);
1973 close(fd);
1974 if (len != 16)
1975 goto use_random;
1976
1977 return;
1978
1979use_random:
1980 for (i = 0; i < 4; i++)
1981 r[i] = random();
1982 memcpy(buf, r, 16);
1983}
1984
0e600426 1985#ifndef MDASSEMBLE
edd8d13c
NB
1986int flush_metadata_updates(struct supertype *st)
1987{
1988 int sfd;
1989 if (!st->updates) {
1990 st->update_tail = NULL;
1991 return -1;
1992 }
1993
4dd2df09 1994 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
1995 if (sfd < 0)
1996 return -1;
1997
1998 while (st->updates) {
1999 struct metadata_update *mu = st->updates;
2000 st->updates = mu->next;
2001
2002 send_message(sfd, mu, 0);
2003 wait_reply(sfd, 0);
2004 free(mu->buf);
2005 free(mu);
2006 }
2007 ack(sfd, 0);
2008 wait_reply(sfd, 0);
2009 close(sfd);
2010 st->update_tail = NULL;
2011 return 0;
2012}
2013
2014void append_metadata_update(struct supertype *st, void *buf, int len)
2015{
2016
503975b9 2017 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
2018
2019 mu->buf = buf;
2020 mu->len = len;
2021 mu->space = NULL;
cb23f1f4 2022 mu->space_list = NULL;
edd8d13c
NB
2023 mu->next = NULL;
2024 *st->update_tail = mu;
2025 st->update_tail = &mu->next;
2026}
0e600426 2027#endif /* MDASSEMBLE */
a931db9e 2028
fe6729fa
NB
2029#ifdef __TINYC__
2030/* tinyc doesn't optimize this check in ioctl.h out ... */
2031unsigned int __invalid_size_argument_for_IOC = 0;
2032#endif
2033
6d11ec6f
AK
2034int experimental(void)
2035{
2036 if (check_env("MDADM_EXPERIMENTAL"))
2037 return 1;
2038 else {
7a862a02 2039 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
2040 return 0;
2041 }
2042}
2043
326727d9
AC
2044/* Pick all spares matching given criteria from a container
2045 * if min_size == 0 do not check size
2046 * if domlist == NULL do not check domains
2047 * if spare_group given add it to domains of each spare
2048 * metadata allows to test domains using metadata of destination array */
2049struct mdinfo *container_choose_spares(struct supertype *st,
2050 unsigned long long min_size,
2051 struct domainlist *domlist,
2052 char *spare_group,
2053 const char *metadata, int get_one)
2054{
2055 struct mdinfo *d, **dp, *disks = NULL;
2056
2057 /* get list of all disks in container */
2058 if (st->ss->getinfo_super_disks)
2059 disks = st->ss->getinfo_super_disks(st);
2060
2061 if (!disks)
2062 return disks;
2063 /* find spare devices on the list */
2064 dp = &disks->devs;
2065 disks->array.spare_disks = 0;
2066 while (*dp) {
2067 int found = 0;
2068 d = *dp;
2069 if (d->disk.state == 0) {
2070 /* check if size is acceptable */
2071 unsigned long long dev_size;
2072 dev_t dev = makedev(d->disk.major,d->disk.minor);
2073
2074 if (!min_size ||
2075 (dev_size_from_id(dev, &dev_size) &&
2076 dev_size >= min_size))
2077 found = 1;
2078 /* check if domain matches */
2079 if (found && domlist) {
4dd2df09 2080 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2081 if (spare_group)
2082 pol_add(&pol, pol_domain,
2083 spare_group, NULL);
e5508b36 2084 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2085 found = 0;
2086 dev_policy_free(pol);
2087 }
2088 }
2089 if (found) {
2090 dp = &d->next;
2091 disks->array.spare_disks++;
2092 if (get_one) {
2093 sysfs_free(*dp);
2094 d->next = NULL;
2095 }
2096 } else {
2097 *dp = d->next;
2098 d->next = NULL;
2099 sysfs_free(d);
2100 }
2101 }
2102 return disks;
2103}
9eafa1de
MN
2104
2105/* Checks if paths point to the same device
2106 * Returns 0 if they do.
2107 * Returns 1 if they don't.
2108 * Returns -1 if something went wrong,
2109 * e.g. paths are empty or the files
2110 * they point to don't exist */
2111int compare_paths (char* path1, char* path2)
2112{
2113 struct stat st1,st2;
2114
2115 if (path1 == NULL || path2 == NULL)
2116 return -1;
2117 if (stat(path1,&st1) != 0)
2118 return -1;
2119 if (stat(path2,&st2) != 0)
2120 return -1;
2121 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2122 return 0;
2123 return 1;
2124}
a7dec3fd
N
2125
2126/* Make sure we can open as many devices as needed */
2127void enable_fds(int devices)
2128{
2129 unsigned int fds = 20 + devices;
2130 struct rlimit lim;
2131 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
2132 || lim.rlim_cur >= fds)
2133 return;
2134 if (lim.rlim_max < fds)
2135 lim.rlim_max = fds;
2136 lim.rlim_cur = fds;
2137 setrlimit(RLIMIT_NOFILE, &lim);
2138}
a9c15847
N
2139
2140int in_initrd(void)
2141{
2142 /* This is based on similar function in systemd. */
2143 struct statfs s;
76d0f188
JS
2144 /* statfs.f_type is signed long on s390x and MIPS, causing all
2145 sorts of sign extension problems with RAMFS_MAGIC being
2146 defined as 0x858458f6 */
a9c15847 2147 return statfs("/", &s) >= 0 &&
7ac5d47e 2148 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2149 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2150 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2151}
8832342d
N
2152
2153void reopen_mddev(int mdfd)
2154{
2155 /* Re-open without any O_EXCL, but keep
2156 * the same fd
2157 */
2158 char *devnm;
2159 int fd;
2160 devnm = fd2devnm(mdfd);
2161 close(mdfd);
2162 fd = open_dev(devnm);
2163 if (fd >= 0 && fd != mdfd)
2164 dup2(fd, mdfd);
2165}
7716570e 2166
7071320a 2167#ifndef MDASSEMBLE
e80357f8
GJ
2168static struct cmap_hooks *cmap_hooks = NULL;
2169static int is_cmap_hooks_ready = 0;
7716570e 2170
e80357f8
GJ
2171void set_cmap_hooks(void)
2172{
2173 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2174 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2175 if (!cmap_hooks->cmap_handle)
2176 return;
7716570e 2177
e80357f8
GJ
2178 cmap_hooks->initialize = dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2179 cmap_hooks->get_string = dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
2180 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2181
e80357f8
GJ
2182 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2183 !cmap_hooks->finalize)
2184 dlclose(cmap_hooks->cmap_handle);
2185 else
2186 is_cmap_hooks_ready = 1;
2187}
7716570e 2188
e80357f8
GJ
2189int get_cluster_name(char **cluster_name)
2190{
2191 int rv = -1;
2192 cmap_handle_t handle;
7716570e 2193
e80357f8
GJ
2194 if (!is_cmap_hooks_ready)
2195 return rv;
7716570e 2196
e80357f8 2197 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2198 if (rv != CS_OK)
2199 goto out;
2200
e80357f8 2201 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2202 if (rv != CS_OK) {
2203 free(*cluster_name);
2204 rv = -1;
2205 goto name_err;
2206 }
2207
2208 rv = 0;
2209name_err:
e80357f8 2210 cmap_hooks->finalize(handle);
7716570e 2211out:
7716570e
GJ
2212 return rv;
2213}
d15a1f72
GJ
2214
2215void set_dlm_hooks(void)
2216{
2217 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2218 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2219 if (!dlm_hooks->dlm_handle)
2220 return;
2221
2222 dlm_hooks->create_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2223 dlm_hooks->release_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
2224 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2225 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2226 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2227 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2228
2229 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2230 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2231 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2232 dlclose(dlm_hooks->dlm_handle);
2233 else
2234 is_dlm_hooks_ready = 1;
2235}
e80357f8
GJ
2236
2237void set_hooks(void)
2238{
2239 set_dlm_hooks();
2240 set_cmap_hooks();
2241}
7d55dca2 2242#endif