IMSM: retry reading sync_completed during reshape
[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;
3da92f27 713 st->ss->load_super(st, fd, name);
82d9eba6 714 /* Looks like a raid array .. */
e7b84f9d 715 pr_err("%s appears to be part of a raid array:\n",
82d9eba6 716 name);
a5d85af7 717 st->ss->getinfo_super(st, &info, NULL);
3da92f27 718 st->ss->free_super(st);
82d9eba6 719 crtime = info.array.ctime;
d078d77c
NB
720 level = map_num(pers, info.array.level);
721 if (!level) level = "-unknown-";
d33f1518
N
722 cont_err("level=%s devices=%d ctime=%s",
723 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 724 return 1;
682c7051
NB
725}
726
682c7051
NB
727int ask(char *mesg)
728{
729 char *add = "";
730 int i;
ca3b6696 731 for (i = 0; i < 5; i++) {
682c7051
NB
732 char buf[100];
733 fprintf(stderr, "%s%s", mesg, add);
734 fflush(stderr);
735 if (fgets(buf, 100, stdin)==NULL)
736 return 0;
737 if (buf[0]=='y' || buf[0]=='Y')
738 return 1;
739 if (buf[0]=='n' || buf[0]=='N')
740 return 0;
741 add = "(y/n) ";
742 }
e7b84f9d 743 pr_err("assuming 'no'\n");
682c7051
NB
744 return 0;
745}
435d4ebb 746#endif /* MDASSEMBLE */
682c7051 747
8d80900b 748int is_standard(char *dev, int *nump)
e5329c37
NB
749{
750 /* tests if dev is a "standard" md dev name.
751 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 752 * where NN is a string of digits
598f0d58
NB
753 * Returns 1 if a partitionable standard,
754 * -1 if non-partitonable,
755 * 0 if not a standard name.
e5329c37 756 */
8d80900b 757 char *d = strrchr(dev, '/');
ca3b6696 758 int type = 0;
8d80900b
NB
759 int num;
760 if (!d)
e5329c37 761 return 0;
ca3b6696
N
762 if (strncmp(d, "/d",2) == 0)
763 d += 2, type = 1; /* /dev/md/dN{pM} */
764 else if (strncmp(d, "/md_d", 5) == 0)
765 d += 5, type = 1; /* /dev/md_dN{pM} */
766 else if (strncmp(d, "/md", 3) == 0)
767 d += 3, type = -1; /* /dev/mdN */
768 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
769 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
770 else
771 return 0;
8d80900b 772 if (!*d)
e5329c37 773 return 0;
8d80900b
NB
774 num = atoi(d);
775 while (isdigit(*d))
776 d++;
777 if (*d)
e5329c37 778 return 0;
8d80900b
NB
779 if (nump) *nump = num;
780
781 return type;
e5329c37
NB
782}
783
4b1ac34b 784unsigned long calc_csum(void *super, int bytes)
82b27616 785{
56eb10c0 786 unsigned long long newcsum = 0;
82b27616 787 int i;
4b1ac34b
NB
788 unsigned int csum;
789 unsigned int *superc = (unsigned int*) super;
82b27616 790
ca3b6696
N
791 for(i = 0; i < bytes/4; i++)
792 newcsum += superc[i];
82b27616 793 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 794#ifdef __alpha__
aba69144 795/* The in-kernel checksum calculation is always 16bit on
570c0542 796 * the alpha, though it is 32 bit on i386...
ca3b6696 797 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
798 * a way that it shouldn't).
799 */
800 csum = (csum & 0xffff) + (csum >> 16);
801 csum = (csum & 0xffff) + (csum >> 16);
802#endif
82b27616
NB
803 return csum;
804}
cd29a5c8 805
435d4ebb 806#ifndef MDASSEMBLE
56eb10c0 807char *human_size(long long bytes)
cd29a5c8
NB
808{
809 static char buf[30];
d5d3721e
NB
810
811 /* We convert bytes to either centi-M{ega,ibi}bytes or
812 * centi-G{igi,ibi}bytes, with appropriate rounding,
813 * and then print 1/100th of those as a decimal.
814 * We allow upto 2048Megabytes before converting to
815 * gigabytes, as that shows more precision and isn't
816 * too large a number.
01619b48 817 * Terabytes are not yet handled.
d5d3721e 818 */
cd29a5c8 819
56eb10c0 820 if (bytes < 5000*1024)
ca3b6696 821 buf[0] = 0;
d5d3721e 822 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 823 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 824 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 825 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
826 cMiB/100 , cMiB % 100,
827 cMB/100, cMB % 100);
828 } else {
93d3bd3b 829 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 830 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 831 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
832 cGiB/100 , cGiB % 100,
833 cGB/100, cGB % 100);
834 }
cd29a5c8
NB
835 return buf;
836}
e0d19036 837
f0ec6710 838char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
839{
840 static char buf[30];
e0d19036 841
570abc6f
MN
842 /* We convert bytes to either centi-M{ega,ibi}bytes or
843 * centi-G{igi,ibi}bytes, with appropriate rounding,
844 * and then print 1/100th of those as a decimal.
845 * We allow upto 2048Megabytes before converting to
846 * gigabytes, as that shows more precision and isn't
847 * too large a number.
848 * Terabytes are not yet handled.
f0ec6710
MN
849 *
850 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
851 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
852 */
853
e0d19036 854 if (bytes < 5000*1024)
570abc6f 855 buf[0] = 0;
f0ec6710
MN
856 else if (prefix == IEC) {
857 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 858 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710
MN
859 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
860 cMiB/100 , cMiB % 100);
861 } else {
93d3bd3b 862 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710
MN
863 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
864 cGiB/100 , cGiB % 100);
865 }
866 }
867 else if (prefix == JEDEC) {
868 if (bytes < 2*1024LL*1024LL*1024LL) {
869 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
870 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
871 cMB/100, cMB % 100);
872 } else {
873 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
874 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
875 cGB/100 , cGB % 100);
876 }
570abc6f 877 }
f0ec6710
MN
878 else
879 buf[0] = 0;
570abc6f 880
e0d19036
NB
881 return buf;
882}
e4965ef8
N
883
884void print_r10_layout(int layout)
885{
886 int near = layout & 255;
887 int far = (layout >> 8) & 255;
888 int offset = (layout&0x10000);
889 char *sep = "";
890
891 if (near != 1) {
892 printf("%s near=%d", sep, near);
893 sep = ",";
894 }
895 if (far != 1)
896 printf("%s %s=%d", sep, offset?"offset":"far", far);
897 if (near*far == 1)
898 printf("NO REDUNDANCY");
899}
435d4ebb 900#endif
e0d19036 901
5f8097be
NB
902unsigned long long calc_array_size(int level, int raid_disks, int layout,
903 int chunksize, unsigned long long devsize)
577e8448 904{
e03640bd
CA
905 if (level == 1)
906 return devsize;
577e8448
AK
907 devsize &= ~(unsigned long long)((chunksize>>9)-1);
908 return get_data_disks(level, layout, raid_disks) * devsize;
909}
910
911int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
912{
913 int data_disks = 0;
914 switch (level) {
ca3b6696
N
915 case 0: data_disks = raid_disks;
916 break;
917 case 1: data_disks = 1;
918 break;
5f8097be 919 case 4:
ca3b6696
N
920 case 5: data_disks = raid_disks - 1;
921 break;
922 case 6: data_disks = raid_disks - 2;
923 break;
5f8097be
NB
924 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
925 break;
926 }
577e8448
AK
927
928 return data_disks;
5f8097be
NB
929}
930
4dd2df09
N
931int devnm2devid(char *devnm)
932{
933 /* First look in /sys/block/$DEVNM/dev for %d:%d
934 * If that fails, try parsing out a number
935 */
936 char path[100];
937 char *ep;
938 int fd;
939 int mjr,mnr;
940
941 sprintf(path, "/sys/block/%s/dev", devnm);
942 fd = open(path, O_RDONLY);
943 if (fd >= 0) {
944 char buf[20];
945 int n = read(fd, buf, sizeof(buf));
946 close(fd);
947 if (n > 0)
948 buf[n] = 0;
949 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
950 return makedev(mjr, mnr);
951 }
952 if (strncmp(devnm, "md_d", 4) == 0 &&
953 isdigit(devnm[4]) &&
954 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
955 ep > devnm && *ep == 0)
956 return makedev(get_mdp_major(), mnr << MdpMinorShift);
957
958 if (strncmp(devnm, "md", 2) == 0 &&
959 isdigit(devnm[2]) &&
960 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
961 ep > devnm && *ep == 0)
962 return makedev(MD_MAJOR, mnr);
963
964 return 0;
965}
966
5a23a06e 967#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 968char *get_md_name(char *devnm)
e0d19036
NB
969{
970 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 971 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 972
e0d19036
NB
973 static char devname[50];
974 struct stat stb;
4dd2df09 975 dev_t rdev = devnm2devid(devnm);
dd0781e5 976 char *dn;
98c6faba 977
4dd2df09
N
978 if (rdev == 0)
979 return 0;
980 if (strncmp(devnm, "md_", 3) == 0) {
981 snprintf(devname, sizeof(devname), "/dev/md/%s",
982 devnm + 3);
98c6faba
NB
983 if (stat(devname, &stb) == 0
984 && (S_IFMT&stb.st_mode) == S_IFBLK
985 && (stb.st_rdev == rdev))
986 return devname;
987 }
4dd2df09
N
988 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
989 if (stat(devname, &stb) == 0
990 && (S_IFMT&stb.st_mode) == S_IFBLK
991 && (stb.st_rdev == rdev))
992 return devname;
993
994 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
995 if (stat(devname, &stb) == 0
996 && (S_IFMT&stb.st_mode) == S_IFBLK
997 && (stb.st_rdev == rdev))
998 return devname;
999
16c6fa80 1000 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
1001 if (dn)
1002 return dn;
4dd2df09 1003 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 1004 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
1005 if (errno != EEXIST)
1006 return NULL;
e0d19036
NB
1007
1008 if (stat(devname, &stb) == 0
1009 && (S_IFMT&stb.st_mode) == S_IFBLK
1010 && (stb.st_rdev == rdev))
1011 return devname;
1012 unlink(devname);
1013 return NULL;
1014}
1015
1016void put_md_name(char *name)
1017{
ca3b6696 1018 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
1019 unlink(name);
1020}
435d4ebb 1021#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 1022
cc742d38
N
1023int get_maj_min(char *dev, int *major, int *minor)
1024{
1025 char *e;
1026 *major = strtoul(dev, &e, 0);
1027 return (e > dev && *e == ':' && e[1] &&
1028 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1029 *e == 0);
1030}
1031
8b0dabea
NB
1032int dev_open(char *dev, int flags)
1033{
1034 /* like 'open', but if 'dev' matches %d:%d, create a temp
1035 * block device and open that
1036 */
8b0dabea
NB
1037 int fd = -1;
1038 char devname[32];
e81cdd9f 1039 int major;
8b0dabea 1040 int minor;
e81cdd9f
NB
1041
1042 if (!dev) return -1;
6df6a774 1043 flags |= O_DIRECT;
e81cdd9f 1044
cc742d38 1045 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1046 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1047 (int)getpid(), major, minor);
1048 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1049 fd = open(devname, flags);
1050 unlink(devname);
6df6a774
N
1051 }
1052 if (fd < 0) {
9dc7d357 1053 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
1054 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
1055 (int)getpid(), major, minor);
ca3b6696 1056 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
1057 fd = open(devname, flags);
1058 unlink(devname);
1059 }
8b0dabea
NB
1060 }
1061 } else
6df6a774 1062 fd = open(dev, flags);
8b0dabea
NB
1063 return fd;
1064}
f9ce90ba 1065
4dd2df09 1066int open_dev_flags(char *devnm, int flags)
e8a70c89 1067{
4dd2df09 1068 int devid;
e8a70c89
N
1069 char buf[20];
1070
4dd2df09
N
1071 devid = devnm2devid(devnm);
1072 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1073 return dev_open(buf, flags);
1074}
1075
4dd2df09 1076int open_dev(char *devnm)
d998b738 1077{
4dd2df09 1078 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1079}
1080
4dd2df09 1081int open_dev_excl(char *devnm)
a931db9e
NB
1082{
1083 char buf[20];
1084 int i;
7187750e 1085 int flags = O_RDWR;
4dd2df09 1086 int devid = devnm2devid(devnm);
399e0b97 1087 long delay = 1000;
a931db9e 1088
4dd2df09 1089 sprintf(buf, "%d:%d", major(devid), minor(devid));
ca3b6696 1090 for (i = 0 ; i < 25 ; i++) {
7187750e 1091 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1092 if (fd >= 0)
1093 return fd;
7187750e
N
1094 if (errno == EACCES && flags == O_RDWR) {
1095 flags = O_RDONLY;
1096 continue;
1097 }
a931db9e
NB
1098 if (errno != EBUSY)
1099 return fd;
399e0b97
N
1100 usleep(delay);
1101 if (delay < 200000)
1102 delay *= 2;
a931db9e
NB
1103 }
1104 return -1;
1105}
1106
9008ed1c
N
1107int same_dev(char *one, char *two)
1108{
1109 struct stat st1, st2;
1110 if (stat(one, &st1) != 0)
1111 return 0;
1112 if (stat(two, &st2) != 0)
1113 return 0;
1114 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1115 return 0;
1116 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1117 return 0;
1118 return st1.st_rdev == st2.st_rdev;
1119}
1120
a7c6e3fb 1121void wait_for(char *dev, int fd)
a714580e
N
1122{
1123 int i;
a7c6e3fb 1124 struct stat stb_want;
399e0b97 1125 long delay = 1000;
a7c6e3fb
N
1126
1127 if (fstat(fd, &stb_want) != 0 ||
1128 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1129 return;
a714580e 1130
ca3b6696 1131 for (i = 0 ; i < 25 ; i++) {
a714580e 1132 struct stat stb;
a7c6e3fb
N
1133 if (stat(dev, &stb) == 0 &&
1134 (stb.st_mode & S_IFMT) == S_IFBLK &&
1135 (stb.st_rdev == stb_want.st_rdev))
a714580e 1136 return;
399e0b97
N
1137 usleep(delay);
1138 if (delay < 200000)
1139 delay *= 2;
a714580e 1140 }
436305c6 1141 if (i == 25)
1ade5cc1 1142 dprintf("timeout waiting for %s\n", dev);
a714580e
N
1143}
1144
0f22b998
N
1145struct superswitch *superlist[] =
1146{
1147 &super0, &super1,
1148 &super_ddf, &super_imsm,
0592faeb 1149 &mbr, &gpt,
0f22b998 1150 NULL };
f9ce90ba 1151
ea24acd0 1152#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1153
4725bc31 1154struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1155{
1686dc25
NB
1156 mdu_array_info_t array;
1157 int vers;
1158 int minor;
1159 struct supertype *st = NULL;
7e0f6979 1160 struct mdinfo *sra;
142cb9e1 1161 char *verstr;
1686dc25
NB
1162 char version[20];
1163 int i;
f7e7067b 1164 char *subarray = NULL;
4dd2df09 1165 char container[32] = "";
1686dc25 1166
4dd2df09 1167 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1168
1169 if (sra) {
7e0f6979
NB
1170 vers = sra->array.major_version;
1171 minor = sra->array.minor_version;
142cb9e1 1172 verstr = sra->text_version;
1686dc25
NB
1173 } else {
1174 if (ioctl(fd, GET_ARRAY_INFO, &array))
1175 array.major_version = array.minor_version = 0;
1176 vers = array.major_version;
1177 minor = array.minor_version;
142cb9e1 1178 verstr = "";
6fbba4c9 1179 }
82d9eba6 1180
1686dc25
NB
1181 if (vers != -1) {
1182 sprintf(version, "%d.%d", vers, minor);
1183 verstr = version;
6fbba4c9 1184 }
3c558363 1185 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1186 char *dev = verstr+1;
5f7e44b2 1187
f7e7067b 1188 subarray = strchr(dev, '/');
92d49ecf 1189 if (subarray) {
f7e7067b 1190 *subarray++ = '\0';
503975b9 1191 subarray = xstrdup(subarray);
92d49ecf 1192 }
4dd2df09 1193 strcpy(container, dev);
15d230f7 1194 sysfs_free(sra);
5f7e44b2 1195 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1196 if (sra && sra->text_version[0])
1197 verstr = sra->text_version;
1198 else
1199 verstr = "-no-metadata-";
f7e7067b
NB
1200 }
1201
1202 for (i = 0; st == NULL && superlist[i] ; i++)
1203 st = superlist[i]->match_metadata_desc(verstr);
1686dc25 1204
15d230f7 1205 sysfs_free(sra);
f7e7067b 1206 if (st) {
3b0896f8 1207 st->sb = NULL;
1f49fb3a
N
1208 if (subarrayp)
1209 *subarrayp = subarray;
4dd2df09
N
1210 strcpy(st->container_devnm, container);
1211 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1212 } else
1213 free(subarray);
5f7e44b2 1214
82d9eba6 1215 return st;
f9ce90ba 1216}
ea24acd0
NB
1217#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1218
0f0749ad 1219int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1220{
1221 char buf[20];
1222 int fd;
1223
1224 sprintf(buf, "%d:%d", major(id), minor(id));
1225 fd = dev_open(buf, O_RDONLY);
1226 if (fd < 0)
1227 return 0;
1228 if (get_dev_size(fd, NULL, size)) {
1229 close(fd);
1230 return 1;
1231 }
1232 close(fd);
1233 return 0;
1234}
f9ce90ba 1235
159c3a1a 1236struct supertype *dup_super(struct supertype *orig)
3da92f27 1237{
159c3a1a 1238 struct supertype *st;
1686dc25 1239
d2ca6449
NB
1240 if (!orig)
1241 return orig;
503975b9 1242 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1243 st->ss = orig->ss;
1244 st->max_devs = orig->max_devs;
1245 st->minor_version = orig->minor_version;
cb8f6859 1246 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1247 st->data_offset = orig->data_offset;
159c3a1a
NB
1248 st->sb = NULL;
1249 st->info = NULL;
1250 return st;
3da92f27
NB
1251}
1252
54887ad8 1253struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1254{
1255 /* try each load_super to find the best match,
1256 * and return the best superswitch
1257 */
82d9eba6
NB
1258 struct superswitch *ss;
1259 struct supertype *st;
1d13b599 1260 unsigned int besttime = 0;
570c0542 1261 int bestsuper = -1;
f9ce90ba
NB
1262 int i;
1263
503975b9 1264 st = xcalloc(1, sizeof(*st));
4dd2df09 1265 st->container_devnm[0] = 0;
d1d599ea 1266
ca3b6696 1267 for (i = 0 ; superlist[i]; i++) {
f9ce90ba
NB
1268 int rv;
1269 ss = superlist[i];
54887ad8
N
1270 if (guess_type == guess_array && ss->add_to_super == NULL)
1271 continue;
1272 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1273 continue;
ef609477 1274 memset(st, 0, sizeof(*st));
df3346e6 1275 st->ignore_hw_compat = 1;
3da92f27 1276 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1277 if (rv == 0) {
1278 struct mdinfo info;
a5d85af7 1279 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1280 if (bestsuper == -1 ||
1281 besttime < info.array.ctime) {
1282 bestsuper = i;
1283 besttime = info.array.ctime;
570c0542 1284 }
3da92f27 1285 ss->free_super(st);
570c0542
NB
1286 }
1287 }
1288 if (bestsuper != -1) {
1289 int rv;
ef609477 1290 memset(st, 0, sizeof(*st));
df3346e6 1291 st->ignore_hw_compat = 1;
3da92f27 1292 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1293 if (rv == 0) {
5e747af2 1294 superlist[bestsuper]->free_super(st);
82d9eba6 1295 return st;
f9ce90ba
NB
1296 }
1297 }
570c0542 1298 free(st);
f9ce90ba
NB
1299 return NULL;
1300}
fe6729fa 1301
beae1dfe
NB
1302/* Return size of device in bytes */
1303int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1304{
1305 unsigned long long ldsize;
c2c9bb6f
NB
1306 struct stat st;
1307
1308 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1309 ldsize = (unsigned long long)st.st_size;
1310 else
beae1dfe
NB
1311#ifdef BLKGETSIZE64
1312 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1313#endif
1314 {
1315 unsigned long dsize;
1316 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1317 ldsize = dsize;
1318 ldsize <<= 9;
1319 } else {
1320 if (dname)
e7b84f9d 1321 pr_err("Cannot get size of %s: %s\b",
beae1dfe
NB
1322 dname, strerror(errno));
1323 return 0;
1324 }
1325 }
1326 *sizep = ldsize;
1327 return 1;
1328}
8fac0577 1329
3a371610
N
1330/* Return true if this can only be a container, not a member device.
1331 * i.e. is and md device and size is zero
1332 */
1333int must_be_container(int fd)
1334{
1335 unsigned long long size;
1336 if (md_get_version(fd) < 0)
1337 return 0;
1338 if (get_dev_size(fd, NULL, &size) == 0)
1339 return 1;
1340 if (size == 0)
1341 return 1;
1342 return 0;
1343}
034b203a
TM
1344
1345/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1346 * Returns: 1 if successful
1347 * -1 for unknown partition type
1348 * 0 for other errors
1349 */
1350static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1351{
056b331e 1352 struct GPT gpt;
034b203a 1353 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1354 struct GPT_part_entry *part;
1355 char buf[512];
034b203a
TM
1356 unsigned long long curr_part_end;
1357 unsigned all_partitions, entry_size;
f21e18ca 1358 unsigned part_nr;
034b203a
TM
1359
1360 *endofpart = 0;
1361
056b331e 1362 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1363 /* skip protective MBR */
034b203a 1364 lseek(fd, 512, SEEK_SET);
73e658d8 1365 /* read GPT header */
056b331e 1366 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1367 return 0;
1368
1369 /* get the number of partition entries and the entry size */
056b331e
N
1370 all_partitions = __le32_to_cpu(gpt.part_cnt);
1371 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1372
1373 /* Check GPT signature*/
056b331e 1374 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1375 return -1;
1376
1377 /* sanity checks */
1378 if (all_partitions > 1024 ||
db7fdfe4 1379 entry_size > sizeof(buf))
034b203a
TM
1380 return -1;
1381
db7fdfe4
JS
1382 part = (struct GPT_part_entry *)buf;
1383
ca3b6696 1384 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1385 /* read partition entry */
db7fdfe4 1386 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1387 return 0;
1388
034b203a 1389 /* is this valid partition? */
db7fdfe4 1390 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1391 /* check the last lba for the current partition */
db7fdfe4 1392 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1393 if (curr_part_end > *endofpart)
1394 *endofpart = curr_part_end;
1395 }
1396
034b203a
TM
1397 }
1398 return 1;
1399}
1400
1401/* Sets endofpart parameter to the last block used by the last partition on the device.
1402 * Returns: 1 if successful
1403 * -1 for unknown partition type
1404 * 0 for other errors
1405 */
1406static int get_last_partition_end(int fd, unsigned long long *endofpart)
1407{
056b331e 1408 struct MBR boot_sect;
034b203a
TM
1409 struct MBR_part_record *part;
1410 unsigned long long curr_part_end;
f21e18ca 1411 unsigned part_nr;
034b203a
TM
1412 int retval = 0;
1413
1414 *endofpart = 0;
1415
056b331e 1416 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1417 /* read MBR */
1418 lseek(fd, 0, 0);
056b331e 1419 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1420 goto abort;
1421
1422 /* check MBP signature */
056b331e 1423 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1424 retval = 1;
1425 /* found the correct signature */
056b331e 1426 part = boot_sect.parts;
034b203a 1427
ca3b6696 1428 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
034b203a
TM
1429 /* check for GPT type */
1430 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1431 retval = get_gpt_last_partition_end(fd, endofpart);
1432 break;
1433 }
1434 /* check the last used lba for the current partition */
1435 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1436 __le32_to_cpu(part->blocks_num);
1437 if (curr_part_end > *endofpart)
1438 *endofpart = curr_part_end;
1439
1440 part++;
1441 }
1442 } else {
1443 /* Unknown partition table */
1444 retval = -1;
1445 }
1446 abort:
1447 return retval;
1448}
1449
53ed6ac3
KW
1450int check_partitions(int fd, char *dname, unsigned long long freesize,
1451 unsigned long long size)
034b203a
TM
1452{
1453 /*
1454 * Check where the last partition ends
1455 */
1456 unsigned long long endofpart;
1457 int ret;
1458
1459 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1460 /* There appears to be a partition table here */
1461 if (freesize == 0) {
1462 /* partitions will not be visible in new device */
e7b84f9d
N
1463 pr_err("partition table exists on %s but will be lost or\n"
1464 " meaningless after creating array\n",
1465 dname);
034b203a
TM
1466 return 1;
1467 } else if (endofpart > freesize) {
1468 /* last partition overlaps metadata */
e7b84f9d
N
1469 pr_err("metadata will over-write last partition on %s.\n",
1470 dname);
034b203a 1471 return 1;
53ed6ac3
KW
1472 } else if (size && endofpart > size) {
1473 /* partitions will be truncated in new device */
e7b84f9d
N
1474 pr_err("array size is too small to cover all partitions on %s.\n",
1475 dname);
53ed6ac3 1476 return 1;
034b203a
TM
1477 }
1478 }
1479 return 0;
1480}
1481
a322f70c
DW
1482int open_container(int fd)
1483{
1484 /* 'fd' is a block device. Find out if it is in use
1485 * by a container, and return an open fd on that container.
1486 */
1487 char path[256];
1488 char *e;
1489 DIR *dir;
1490 struct dirent *de;
1491 int dfd, n;
1492 char buf[200];
1493 int major, minor;
1494 struct stat st;
1495
1496 if (fstat(fd, &st) != 0)
1497 return -1;
1498 sprintf(path, "/sys/dev/block/%d:%d/holders",
1499 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1500 e = path + strlen(path);
1501
1502 dir = opendir(path);
1503 if (!dir)
1504 return -1;
1505 while ((de = readdir(dir))) {
1506 if (de->d_ino == 0)
1507 continue;
1508 if (de->d_name[0] == '.')
1509 continue;
a994592d
N
1510 /* Need to make sure it is a container and not a volume */
1511 sprintf(e, "/%s/md/metadata_version", de->d_name);
1512 dfd = open(path, O_RDONLY);
1513 if (dfd < 0)
1514 continue;
1515 n = read(dfd, buf, sizeof(buf));
1516 close(dfd);
1517 if (n <= 0 || (unsigned)n >= sizeof(buf))
1518 continue;
1519 buf[n] = 0;
1520 if (strncmp(buf, "external", 8) != 0 ||
1521 n < 10 ||
1522 buf[9] == '/')
1523 continue;
a322f70c
DW
1524 sprintf(e, "/%s/dev", de->d_name);
1525 dfd = open(path, O_RDONLY);
1526 if (dfd < 0)
1527 continue;
1528 n = read(dfd, buf, sizeof(buf));
1529 close(dfd);
f21e18ca 1530 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1531 continue;
1532 buf[n] = 0;
1533 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1534 continue;
1535 sprintf(buf, "%d:%d", major, minor);
1536 dfd = dev_open(buf, O_RDONLY);
1537 if (dfd >= 0) {
1538 closedir(dir);
1539 return dfd;
1540 }
1541 }
355726fa 1542 closedir(dir);
a322f70c
DW
1543 return -1;
1544}
1545
33414a01
DW
1546struct superswitch *version_to_superswitch(char *vers)
1547{
1548 int i;
1549
1550 for (i = 0; superlist[i]; i++) {
1551 struct superswitch *ss = superlist[i];
1552
1553 if (strcmp(vers, ss->name) == 0)
1554 return ss;
1555 }
1556
1557 return NULL;
1558}
1559
4dd2df09
N
1560int metadata_container_matches(char *metadata, char *devnm)
1561{
1562 /* Check if 'devnm' is the container named in 'metadata'
1563 * which is
1564 * /containername/componentname or
1565 * -containername/componentname
1566 */
1567 int l;
1568 if (*metadata != '/' && *metadata != '-')
1569 return 0;
1570 l = strlen(devnm);
1571 if (strncmp(metadata+1, devnm, l) != 0)
1572 return 0;
1573 if (metadata[l+1] != '/')
1574 return 0;
1575 return 1;
1576}
1577
1578int metadata_subdev_matches(char *metadata, char *devnm)
1579{
1580 /* Check if 'devnm' is the subdev named in 'metadata'
1581 * which is
1582 * /containername/subdev or
1583 * -containername/subdev
1584 */
1585 char *sl;
1586 if (*metadata != '/' && *metadata != '-')
1587 return 0;
1588 sl = strchr(metadata+1, '/');
1589 if (!sl)
1590 return 0;
1591 if (strcmp(sl+1, devnm) == 0)
1592 return 1;
1593 return 0;
1594}
1595
33414a01
DW
1596int is_container_member(struct mdstat_ent *mdstat, char *container)
1597{
1598 if (mdstat->metadata_version == NULL ||
1599 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1600 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1601 return 0;
1602
1603 return 1;
1604}
1605
1606int is_subarray_active(char *subarray, char *container)
1607{
1608 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1609 struct mdstat_ent *ent;
1610
e5408a32
DW
1611 for (ent = mdstat; ent; ent = ent->next)
1612 if (is_container_member(ent, container))
e5e5d7ce 1613 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1614 break;
33414a01
DW
1615
1616 free_mdstat(mdstat);
1617
1618 return ent != NULL;
1619}
1620
1621/* open_subarray - opens a subarray in a container
1622 * @dev: container device name
feab51f8 1623 * @st: empty supertype
33414a01
DW
1624 * @quiet: block reporting errors flag
1625 *
1626 * On success returns an fd to a container and fills in *st
1627 */
feab51f8 1628int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1629{
1630 struct mdinfo *mdi;
a951a4f7 1631 struct mdinfo *info;
33414a01 1632 int fd, err = 1;
4dd2df09 1633 char *_devnm;
33414a01
DW
1634
1635 fd = open(dev, O_RDWR|O_EXCL);
1636 if (fd < 0) {
1637 if (!quiet)
e7b84f9d 1638 pr_err("Couldn't open %s, aborting\n",
33414a01 1639 dev);
b990032d 1640 return -1;
33414a01
DW
1641 }
1642
4dd2df09
N
1643 _devnm = fd2devnm(fd);
1644 if (_devnm == NULL) {
33414a01 1645 if (!quiet)
e7b84f9d
N
1646 pr_err("Failed to determine device number for %s\n",
1647 dev);
33414a01
DW
1648 goto close_fd;
1649 }
4dd2df09 1650 strcpy(st->devnm, _devnm);
33414a01 1651
4dd2df09 1652 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1653 if (!mdi) {
1654 if (!quiet)
e7b84f9d 1655 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1656 dev);
1657 goto close_fd;
1658 }
1659
1660 if (mdi->array.level != UnSet) {
1661 if (!quiet)
e7b84f9d 1662 pr_err("%s is not a container\n", dev);
33414a01
DW
1663 goto free_sysfs;
1664 }
1665
1666 st->ss = version_to_superswitch(mdi->text_version);
1667 if (!st->ss) {
1668 if (!quiet)
e7b84f9d
N
1669 pr_err("Operation not supported for %s metadata\n",
1670 mdi->text_version);
33414a01
DW
1671 goto free_sysfs;
1672 }
1673
4dd2df09 1674 if (st->devnm[0] == 0) {
33414a01 1675 if (!quiet)
e7b84f9d 1676 pr_err("Failed to allocate device name\n");
33414a01
DW
1677 goto free_sysfs;
1678 }
1679
db20d413 1680 if (!st->ss->load_container) {
33414a01 1681 if (!quiet)
e7b84f9d 1682 pr_err("%s is not a container\n", dev);
4dd2df09 1683 goto free_sysfs;
33414a01
DW
1684 }
1685
db20d413 1686 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1687 if (!quiet)
e7b84f9d 1688 pr_err("Failed to load metadata for %s\n",
db20d413 1689 dev);
4dd2df09 1690 goto free_sysfs;
33414a01
DW
1691 }
1692
a951a4f7
N
1693 info = st->ss->container_content(st, subarray);
1694 if (!info) {
1695 if (!quiet)
e7b84f9d 1696 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1697 subarray, dev);
1698 goto free_super;
1699 }
1700 free(info);
1701
33414a01
DW
1702 err = 0;
1703
1704 free_super:
1705 if (err)
1706 st->ss->free_super(st);
33414a01
DW
1707 free_sysfs:
1708 sysfs_free(mdi);
1709 close_fd:
1710 if (err)
1711 close(fd);
1712
1713 if (err)
1714 return -1;
1715 else
1716 return fd;
1717}
1718
7801ac20
N
1719int add_disk(int mdfd, struct supertype *st,
1720 struct mdinfo *sra, struct mdinfo *info)
1721{
1722 /* Add a device to an array, in one of 2 ways. */
1723 int rv;
1724#ifndef MDASSEMBLE
1725 if (st->ss->external) {
d23534e4
DW
1726 if (info->disk.state & (1<<MD_DISK_SYNC))
1727 info->recovery_start = MaxSector;
1728 else
1729 info->recovery_start = 0;
2904b26f 1730 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1731 if (! rv) {
1732 struct mdinfo *sd2;
f35f2525
N
1733 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1734 if (sd2 == info)
1735 break;
1736 if (sd2 == NULL) {
503975b9 1737 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1738 *sd2 = *info;
1739 sd2->next = sra->devs;
1740 sra->devs = sd2;
1741 }
7801ac20
N
1742 }
1743 } else
1744#endif
1745 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1746 return rv;
1747}
1748
de6ae750
N
1749int remove_disk(int mdfd, struct supertype *st,
1750 struct mdinfo *sra, struct mdinfo *info)
1751{
1752 int rv;
1753 /* Remove the disk given by 'info' from the array */
1754#ifndef MDASSEMBLE
1755 if (st->ss->external)
1756 rv = sysfs_set_str(sra, info, "slot", "none");
1757 else
1758#endif
1759 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1760 info->disk.minor));
1761 return rv;
1762}
1763
f35f2525
N
1764int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1765{
1766 /* Initialise kernel's knowledge of array.
1767 * This varies between externally managed arrays
1768 * and older kernels
1769 */
1770 int vers = md_get_version(mdfd);
1771 int rv;
1772
1773#ifndef MDASSEMBLE
1774 if (st->ss->external)
1775 rv = sysfs_set_array(info, vers);
1776 else
1777#endif
1778 if ((vers % 100) >= 1) { /* can use different versions */
1779 mdu_array_info_t inf;
1780 memset(&inf, 0, sizeof(inf));
1781 inf.major_version = info->array.major_version;
1782 inf.minor_version = info->array.minor_version;
1783 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1784 } else
1785 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1786 return rv;
1787}
1788
1e5c6983
DW
1789unsigned long long min_recovery_start(struct mdinfo *array)
1790{
1791 /* find the minimum recovery_start in an array for metadata
1792 * formats that only record per-array recovery progress instead
1793 * of per-device
1794 */
1795 unsigned long long recovery_start = MaxSector;
1796 struct mdinfo *d;
1797
1798 for (d = array->devs; d; d = d->next)
1799 recovery_start = min(recovery_start, d->recovery_start);
1800
1801 return recovery_start;
1802}
1803
4dd2df09 1804int mdmon_pid(char *devnm)
a931db9e
NB
1805{
1806 char path[100];
1807 char pid[10];
1808 int fd;
1809 int n;
10013317 1810
4dd2df09 1811 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1812
24f6f99b 1813 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1814
1815 if (fd < 0)
cf556303 1816 return -1;
a931db9e
NB
1817 n = read(fd, pid, 9);
1818 close(fd);
1819 if (n <= 0)
cf556303 1820 return -1;
24f6f99b 1821 return atoi(pid);
a931db9e
NB
1822}
1823
4dd2df09 1824int mdmon_running(char *devnm)
a931db9e 1825{
4dd2df09 1826 int pid = mdmon_pid(devnm);
24f6f99b 1827 if (pid <= 0)
a931db9e 1828 return 0;
24f6f99b 1829 if (kill(pid, 0) == 0)
a931db9e
NB
1830 return 1;
1831 return 0;
1832}
1833
4dd2df09 1834int start_mdmon(char *devnm)
8850ee3e 1835{
1913c325 1836 int i, skipped;
44d2e365 1837 int len;
ca3b6696 1838 pid_t pid;
9fe32043 1839 int status;
44d2e365
N
1840 char pathbuf[1024];
1841 char *paths[4] = {
1842 pathbuf,
85945e19 1843 BINDIR "/mdmon",
2f1bcf43 1844 "./mdmon",
44d2e365
N
1845 NULL
1846 };
8850ee3e 1847
40ebbb9c 1848 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1849 return 0;
1850
9cf014ec 1851 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1852 if (len > 0) {
1853 char *sl;
1854 pathbuf[len] = 0;
1855 sl = strrchr(pathbuf, '/');
1856 if (sl)
1857 sl++;
1858 else
1859 sl = pathbuf;
1860 strcpy(sl, "mdmon");
1861 } else
1862 pathbuf[0] = '\0';
1863
0f7bdf89 1864 /* First try to run systemctl */
9540cc24
N
1865 if (!check_env("MDADM_NO_SYSTEMCTL"))
1866 switch(fork()) {
1867 case 0:
1868 /* FIXME yuk. CLOSE_EXEC?? */
1869 skipped = 0;
1870 for (i = 3; skipped < 20; i++)
1871 if (close(i) < 0)
1872 skipped++;
1873 else
1874 skipped = 0;
1875
1876 /* Don't want to see error messages from
1877 * systemctl. If the service doesn't exist,
1878 * we start mdmon ourselves.
1879 */
1880 close(2);
1881 open("/dev/null", O_WRONLY);
1882 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1883 devnm);
1884 status = execl("/usr/bin/systemctl", "systemctl",
1885 "start",
1886 pathbuf, NULL);
1887 status = execl("/bin/systemctl", "systemctl", "start",
1888 pathbuf, NULL);
1889 exit(1);
7a862a02 1890 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1891 return -1;
1892 default: /* parent - good */
1893 pid = wait(&status);
1894 if (pid >= 0 && status == 0)
1895 return 0;
1896 }
0f7bdf89
JS
1897
1898 /* That failed, try running mdmon directly */
8850ee3e
N
1899 switch(fork()) {
1900 case 0:
1901 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1902 skipped = 0;
ca3b6696 1903 for (i = 3; skipped < 20; i++)
1913c325
N
1904 if (close(i) < 0)
1905 skipped++;
1906 else
1907 skipped = 0;
1908
ca3b6696 1909 for (i = 0; paths[i]; i++)
a0963a86 1910 if (paths[i][0]) {
2f1bcf43 1911 execl(paths[i], paths[i],
4dd2df09 1912 devnm, NULL);
a0963a86 1913 }
8850ee3e 1914 exit(1);
7a862a02 1915 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 1916 return -1;
9fe32043
N
1917 default: /* parent - good */
1918 pid = wait(&status);
15c10423 1919 if (pid < 0 || status != 0) {
7a862a02 1920 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 1921 return -1;
15c10423 1922 }
8850ee3e
N
1923 }
1924 return 0;
1925}
1926
148acb7b
DW
1927__u32 random32(void)
1928{
1929 __u32 rv;
1930 int rfd = open("/dev/urandom", O_RDONLY);
1931 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1932 rv = random();
1933 if (rfd >= 0)
1934 close(rfd);
1935 return rv;
1936}
1937
0e600426 1938#ifndef MDASSEMBLE
edd8d13c
NB
1939int flush_metadata_updates(struct supertype *st)
1940{
1941 int sfd;
1942 if (!st->updates) {
1943 st->update_tail = NULL;
1944 return -1;
1945 }
1946
4dd2df09 1947 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
1948 if (sfd < 0)
1949 return -1;
1950
1951 while (st->updates) {
1952 struct metadata_update *mu = st->updates;
1953 st->updates = mu->next;
1954
1955 send_message(sfd, mu, 0);
1956 wait_reply(sfd, 0);
1957 free(mu->buf);
1958 free(mu);
1959 }
1960 ack(sfd, 0);
1961 wait_reply(sfd, 0);
1962 close(sfd);
1963 st->update_tail = NULL;
1964 return 0;
1965}
1966
1967void append_metadata_update(struct supertype *st, void *buf, int len)
1968{
1969
503975b9 1970 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
1971
1972 mu->buf = buf;
1973 mu->len = len;
1974 mu->space = NULL;
cb23f1f4 1975 mu->space_list = NULL;
edd8d13c
NB
1976 mu->next = NULL;
1977 *st->update_tail = mu;
1978 st->update_tail = &mu->next;
1979}
0e600426 1980#endif /* MDASSEMBLE */
a931db9e 1981
fe6729fa
NB
1982#ifdef __TINYC__
1983/* tinyc doesn't optimize this check in ioctl.h out ... */
1984unsigned int __invalid_size_argument_for_IOC = 0;
1985#endif
1986
6d11ec6f
AK
1987int experimental(void)
1988{
1989 if (check_env("MDADM_EXPERIMENTAL"))
1990 return 1;
1991 else {
7a862a02 1992 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
1993 return 0;
1994 }
1995}
1996
326727d9
AC
1997/* Pick all spares matching given criteria from a container
1998 * if min_size == 0 do not check size
1999 * if domlist == NULL do not check domains
2000 * if spare_group given add it to domains of each spare
2001 * metadata allows to test domains using metadata of destination array */
2002struct mdinfo *container_choose_spares(struct supertype *st,
2003 unsigned long long min_size,
2004 struct domainlist *domlist,
2005 char *spare_group,
2006 const char *metadata, int get_one)
2007{
2008 struct mdinfo *d, **dp, *disks = NULL;
2009
2010 /* get list of all disks in container */
2011 if (st->ss->getinfo_super_disks)
2012 disks = st->ss->getinfo_super_disks(st);
2013
2014 if (!disks)
2015 return disks;
2016 /* find spare devices on the list */
2017 dp = &disks->devs;
2018 disks->array.spare_disks = 0;
2019 while (*dp) {
2020 int found = 0;
2021 d = *dp;
2022 if (d->disk.state == 0) {
2023 /* check if size is acceptable */
2024 unsigned long long dev_size;
2025 dev_t dev = makedev(d->disk.major,d->disk.minor);
2026
2027 if (!min_size ||
2028 (dev_size_from_id(dev, &dev_size) &&
2029 dev_size >= min_size))
2030 found = 1;
2031 /* check if domain matches */
2032 if (found && domlist) {
4dd2df09 2033 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2034 if (spare_group)
2035 pol_add(&pol, pol_domain,
2036 spare_group, NULL);
e5508b36 2037 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2038 found = 0;
2039 dev_policy_free(pol);
2040 }
2041 }
2042 if (found) {
2043 dp = &d->next;
2044 disks->array.spare_disks++;
2045 if (get_one) {
2046 sysfs_free(*dp);
2047 d->next = NULL;
2048 }
2049 } else {
2050 *dp = d->next;
2051 d->next = NULL;
2052 sysfs_free(d);
2053 }
2054 }
2055 return disks;
2056}
9eafa1de
MN
2057
2058/* Checks if paths point to the same device
2059 * Returns 0 if they do.
2060 * Returns 1 if they don't.
2061 * Returns -1 if something went wrong,
2062 * e.g. paths are empty or the files
2063 * they point to don't exist */
2064int compare_paths (char* path1, char* path2)
2065{
2066 struct stat st1,st2;
2067
2068 if (path1 == NULL || path2 == NULL)
2069 return -1;
2070 if (stat(path1,&st1) != 0)
2071 return -1;
2072 if (stat(path2,&st2) != 0)
2073 return -1;
2074 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2075 return 0;
2076 return 1;
2077}
a7dec3fd
N
2078
2079/* Make sure we can open as many devices as needed */
2080void enable_fds(int devices)
2081{
2082 unsigned int fds = 20 + devices;
2083 struct rlimit lim;
2084 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
2085 || lim.rlim_cur >= fds)
2086 return;
2087 if (lim.rlim_max < fds)
2088 lim.rlim_max = fds;
2089 lim.rlim_cur = fds;
2090 setrlimit(RLIMIT_NOFILE, &lim);
2091}
a9c15847
N
2092
2093int in_initrd(void)
2094{
2095 /* This is based on similar function in systemd. */
2096 struct statfs s;
76d0f188
JS
2097 /* statfs.f_type is signed long on s390x and MIPS, causing all
2098 sorts of sign extension problems with RAMFS_MAGIC being
2099 defined as 0x858458f6 */
a9c15847 2100 return statfs("/", &s) >= 0 &&
7ac5d47e 2101 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2102 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2103 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2104}
8832342d
N
2105
2106void reopen_mddev(int mdfd)
2107{
2108 /* Re-open without any O_EXCL, but keep
2109 * the same fd
2110 */
2111 char *devnm;
2112 int fd;
2113 devnm = fd2devnm(mdfd);
2114 close(mdfd);
2115 fd = open_dev(devnm);
2116 if (fd >= 0 && fd != mdfd)
2117 dup2(fd, mdfd);
2118}
7716570e 2119
7071320a 2120#ifndef MDASSEMBLE
e80357f8
GJ
2121static struct cmap_hooks *cmap_hooks = NULL;
2122static int is_cmap_hooks_ready = 0;
7716570e 2123
e80357f8
GJ
2124void set_cmap_hooks(void)
2125{
2126 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2127 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2128 if (!cmap_hooks->cmap_handle)
2129 return;
7716570e 2130
e80357f8
GJ
2131 cmap_hooks->initialize = dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2132 cmap_hooks->get_string = dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
2133 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2134
e80357f8
GJ
2135 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2136 !cmap_hooks->finalize)
2137 dlclose(cmap_hooks->cmap_handle);
2138 else
2139 is_cmap_hooks_ready = 1;
2140}
7716570e 2141
e80357f8
GJ
2142int get_cluster_name(char **cluster_name)
2143{
2144 int rv = -1;
2145 cmap_handle_t handle;
7716570e 2146
e80357f8
GJ
2147 if (!is_cmap_hooks_ready)
2148 return rv;
7716570e 2149
e80357f8 2150 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2151 if (rv != CS_OK)
2152 goto out;
2153
e80357f8 2154 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2155 if (rv != CS_OK) {
2156 free(*cluster_name);
2157 rv = -1;
2158 goto name_err;
2159 }
2160
2161 rv = 0;
2162name_err:
e80357f8 2163 cmap_hooks->finalize(handle);
7716570e 2164out:
7716570e
GJ
2165 return rv;
2166}
d15a1f72
GJ
2167
2168void set_dlm_hooks(void)
2169{
2170 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2171 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2172 if (!dlm_hooks->dlm_handle)
2173 return;
2174
2175 dlm_hooks->create_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2176 dlm_hooks->release_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
2177 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2178 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2179 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2180 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2181
2182 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2183 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2184 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2185 dlclose(dlm_hooks->dlm_handle);
2186 else
2187 is_dlm_hooks_ready = 1;
2188}
e80357f8
GJ
2189
2190void set_hooks(void)
2191{
2192 set_dlm_hooks();
2193 set_cmap_hooks();
2194}
7d55dca2 2195#endif