Add casts for the addr arg of connect and bind
[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"
d15a1f72 27#include <sys/poll.h>
edd8d13c 28#include <sys/socket.h>
64c4757e 29#include <sys/utsname.h>
9fe32043 30#include <sys/wait.h>
edd8d13c 31#include <sys/un.h>
a7dec3fd 32#include <sys/resource.h>
a9c15847
N
33#include <sys/vfs.h>
34#include <linux/magic.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
GJ
85static struct dlm_hooks *dlm_hooks = NULL;
86static int is_dlm_hooks_ready = 0;
87struct dlm_lock_resource *dlm_lock_res = NULL;
88static int ast_called = 0;
89
90struct dlm_lock_resource {
91 dlm_lshandle_t *ls;
92 struct dlm_lksb lksb;
93};
94
81a8a694 95int dlm_funs_ready(void)
d15a1f72 96{
81a8a694 97 return is_dlm_hooks_ready ? 1 : 0;
d15a1f72
GJ
98}
99
100/* Using poll(2) to wait for and dispatch ASTs */
101static int poll_for_ast(dlm_lshandle_t ls)
102{
103 struct pollfd pfd;
104
105 pfd.fd = dlm_hooks->ls_get_fd(ls);
106 pfd.events = POLLIN;
107
108 while (!ast_called)
109 {
110 if (poll(&pfd, 1, 0) < 0)
111 {
112 perror("poll");
113 return -1;
114 }
115 dlm_hooks->dispatch(dlm_hooks->ls_get_fd(ls));
116 }
117 ast_called = 0;
118
119 return 0;
120}
121
122static void dlm_ast(void *arg)
123{
124 ast_called = 1;
125}
126
81a8a694 127static char *cluster_name = NULL;
d15a1f72 128/* Create the lockspace, take bitmapXXX locks on all the bitmaps. */
81a8a694 129int cluster_get_dlmlock(int *lockid)
d15a1f72
GJ
130{
131 int ret = -1;
132 char str[64];
133 int flags = LKF_NOQUEUE;
134
81a8a694
GJ
135 ret = get_cluster_name(&cluster_name);
136 if (ret) {
137 pr_err("The md can't get cluster name\n");
138 return -1;
139 }
140
d15a1f72 141 dlm_lock_res = xmalloc(sizeof(struct dlm_lock_resource));
81a8a694 142 dlm_lock_res->ls = dlm_hooks->create_lockspace(cluster_name, O_RDWR);
d15a1f72 143 if (!dlm_lock_res->ls) {
81a8a694 144 pr_err("%s failed to create lockspace\n", cluster_name);
d15a1f72
GJ
145 goto out;
146 }
147
148 /* Conversions need the lockid in the LKSB */
149 if (flags & LKF_CONVERT)
150 dlm_lock_res->lksb.sb_lkid = *lockid;
151
81a8a694 152 snprintf(str, 64, "bitmap%s", cluster_name);
d15a1f72
GJ
153 /* if flags with LKF_CONVERT causes below return ENOENT which means
154 * "No such file or directory" */
155 ret = dlm_hooks->ls_lock(dlm_lock_res->ls, LKM_PWMODE, &dlm_lock_res->lksb,
156 flags, str, strlen(str), 0, dlm_ast,
157 dlm_lock_res, NULL, NULL);
158 if (ret) {
159 pr_err("error %d when get PW mode on lock %s\n", errno, str);
160 goto out;
161 }
162
163 /* Wait for it to complete */
164 poll_for_ast(dlm_lock_res->ls);
165 *lockid = dlm_lock_res->lksb.sb_lkid;
166
167 errno = dlm_lock_res->lksb.sb_status;
168 if (errno) {
169 pr_err("error %d happened in ast with lock %s\n", errno, str);
170 goto out;
171 }
172
173out:
174 return ret;
175}
176
81a8a694 177int cluster_release_dlmlock(int lockid)
d15a1f72
GJ
178{
179 int ret = -1;
180
81a8a694
GJ
181 if (!cluster_name)
182 return -1;
183
d15a1f72
GJ
184 /* if flags with LKF_CONVERT causes below return EINVAL which means
185 * "Invalid argument" */
186 ret = dlm_hooks->ls_unlock(dlm_lock_res->ls, lockid, 0,
187 &dlm_lock_res->lksb, dlm_lock_res);
188 if (ret) {
189 pr_err("error %d happened when unlock\n", errno);
190 /* XXX make sure the lock is unlocked eventually */
191 goto out;
192 }
193
194 /* Wait for it to complete */
195 poll_for_ast(dlm_lock_res->ls);
196
197 errno = dlm_lock_res->lksb.sb_status;
198 if (errno != EUNLOCK) {
199 pr_err("error %d happened in ast when unlock lockspace\n", errno);
200 /* XXX make sure the lockspace is unlocked eventually */
201 goto out;
202 }
203
81a8a694 204 ret = dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1);
d15a1f72
GJ
205 if (ret) {
206 pr_err("error %d happened when release lockspace\n", errno);
207 /* XXX make sure the lockspace is released eventually */
208 goto out;
209 }
210 free(dlm_lock_res);
211
212out:
213 return ret;
214}
215
64c4757e
NB
216/*
217 * Parse a 128 bit uuid in 4 integers
218 * format is 32 hexx nibbles with options :.<space> separator
219 * If not exactly 32 hex digits are found, return 0
220 * else return 1
221 */
222int parse_uuid(char *str, int uuid[4])
223{
aba69144
NB
224 int hit = 0; /* number of Hex digIT */
225 int i;
226 char c;
ca3b6696
N
227 for (i = 0; i < 4; i++)
228 uuid[i] = 0;
aba69144 229
ca3b6696 230 while ((c = *str++) != 0) {
aba69144 231 int n;
ca3b6696 232 if (c >= '0' && c <= '9')
aba69144 233 n = c-'0';
ca3b6696 234 else if (c >= 'a' && c <= 'f')
aba69144 235 n = 10 + c - 'a';
ca3b6696 236 else if (c >= 'A' && c <= 'F')
aba69144
NB
237 n = 10 + c - 'A';
238 else if (strchr(":. -", c))
239 continue;
240 else return 0;
241
242 if (hit<32) {
243 uuid[hit/8] <<= 4;
244 uuid[hit/8] += n;
245 }
246 hit++;
82b27616 247 }
aba69144
NB
248 if (hit == 32)
249 return 1;
250 return 0;
64c4757e
NB
251}
252
64c4757e
NB
253/*
254 * Get the md version number.
255 * We use the RAID_VERSION ioctl if it is supported
256 * If not, but we have a block device with major '9', we assume
257 * 0.36.0
258 *
259 * Return version number as 24 but number - assume version parts
260 * always < 255
261 */
262
263int md_get_version(int fd)
264{
5d500228
N
265 struct stat stb;
266 mdu_version_t vers;
267
268 if (fstat(fd, &stb)<0)
269 return -1;
270 if ((S_IFMT&stb.st_mode) != S_IFBLK)
271 return -1;
272
273 if (ioctl(fd, RAID_VERSION, &vers) == 0)
274 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
275 if (errno == EACCES)
276 return -1;
277 if (major(stb.st_rdev) == MD_MAJOR)
278 return (3600);
64c4757e 279 return -1;
64c4757e
NB
280}
281
64c4757e
NB
282int get_linux_version()
283{
284 struct utsname name;
98c6faba 285 char *cp;
f161d047 286 int a = 0, b = 0,c = 0;
64c4757e
NB
287 if (uname(&name) <0)
288 return -1;
289
98c6faba
NB
290 cp = name.release;
291 a = strtoul(cp, &cp, 10);
f161d047
N
292 if (*cp == '.')
293 b = strtoul(cp+1, &cp, 10);
294 if (*cp == '.')
295 c = strtoul(cp+1, &cp, 10);
98c6faba 296
682c7051 297 return (a*1000000)+(b*1000)+c;
64c4757e
NB
298}
299
71204a50 300#ifndef MDASSEMBLE
bc77ed53
DW
301int mdadm_version(char *version)
302{
303 int a, b, c;
304 char *cp;
305
306 if (!version)
307 version = Version;
308
309 cp = strchr(version, '-');
310 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
311 return -1;
312 cp += 3;
313 a = strtoul(cp, &cp, 10);
314 if (*cp != '.')
315 return -1;
316 b = strtoul(cp+1, &cp, 10);
317 if (*cp == '.')
318 c = strtoul(cp+1, &cp, 10);
319 else
320 c = 0;
321 if (*cp != ' ' && *cp != '-')
322 return -1;
323 return (a*1000000)+(b*1000)+c;
324}
325
51d4261c 326unsigned long long parse_size(char *size)
84e11361
N
327{
328 /* parse 'size' which should be a number optionally
329 * followed by 'K', 'M', or 'G'.
330 * Without a suffix, K is assumed.
331 * Number returned is in sectors (half-K)
822e393a 332 * INVALID_SECTORS returned on error.
84e11361
N
333 */
334 char *c;
335 long long s = strtoll(size, &c, 10);
336 if (s > 0) {
337 switch (*c) {
338 case 'K':
339 c++;
340 default:
341 s *= 2;
342 break;
343 case 'M':
344 c++;
345 s *= 1024 * 2;
346 break;
347 case 'G':
348 c++;
349 s *= 1024 * 1024 * 2;
350 break;
72ca9bcf
N
351 case 's': /* sectors */
352 c++;
353 break;
84e11361 354 }
51d4261c 355 } else
822e393a 356 s = INVALID_SECTORS;
84e11361 357 if (*c)
822e393a 358 s = INVALID_SECTORS;
84e11361
N
359 return s;
360}
361
4a06e2c2
N
362int parse_layout_10(char *layout)
363{
364 int copies, rv;
365 char *cp;
366 /* Parse the layout string for raid10 */
367 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
368 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
369 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
370 copies > 200 ||
371 *cp)
372 return -1;
373 if (layout[0] == 'n')
374 rv = 256 + copies;
375 else if (layout[0] == 'o')
376 rv = 0x10000 + (copies<<8) + 1;
377 else
378 rv = 1 + (copies<<8);
379 return rv;
380}
381
382int parse_layout_faulty(char *layout)
383{
384 /* Parse the layout string for 'faulty' */
385 int ln = strcspn(layout, "0123456789");
503975b9 386 char *m = xstrdup(layout);
4a06e2c2
N
387 int mode;
388 m[ln] = 0;
389 mode = map_name(faultylayout, m);
390 if (mode == UnSet)
391 return -1;
392
393 return mode | (atoi(layout+ln)<< ModeShift);
394}
79868890
N
395
396long parse_num(char *num)
397{
398 /* Either return a valid number, or -1 */
399 char *c;
400 long rv = strtol(num, &c, 10);
401 if (rv < 0 || *c || !num[0])
402 return -1;
403 else
404 return rv;
405}
eb3929a4 406#endif
4a06e2c2 407
4de90913
GJ
408int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
409{
410 char *dev;
411 *slot = strtoul(input, &dev, 10);
412 if (dev == input || dev[0] != ':')
413 return -1;
414 *devname = dev+1;
415 return 0;
416}
417
0430ed48
NB
418void remove_partitions(int fd)
419{
420 /* remove partitions from this block devices.
421 * This is used for components added to an array
422 */
423#ifdef BLKPG_DEL_PARTITION
424 struct blkpg_ioctl_arg a;
425 struct blkpg_partition p;
426
427 a.op = BLKPG_DEL_PARTITION;
428 a.data = (void*)&p;
429 a.datalen = sizeof(p);
430 a.flags = 0;
431 memset(a.data, 0, a.datalen);
ca3b6696 432 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
433 ioctl(fd, BLKPG, &a);
434#endif
435}
436
691c6ee1
N
437int test_partition(int fd)
438{
439 /* Check if fd is a whole-disk or a partition.
440 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
441 * will return ENXIO on an invalid partition number.
442 */
443 struct blkpg_ioctl_arg a;
444 struct blkpg_partition p;
445 a.op = BLKPG_DEL_PARTITION;
446 a.data = (void*)&p;
447 a.datalen = sizeof(p);
448 a.flags = 0;
449 memset(a.data, 0, a.datalen);
450 p.pno = 1<<30;
451 if (ioctl(fd, BLKPG, &a) == 0)
452 /* Very unlikely, but not a partition */
453 return 0;
357ac106 454 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
455 /* not a partition */
456 return 0;
457
458 return 1;
459}
460
bfd76b93
CA
461int test_partition_from_id(dev_t id)
462{
463 char buf[20];
464 int fd, rv;
465
466 sprintf(buf, "%d:%d", major(id), minor(id));
467 fd = dev_open(buf, O_RDONLY);
468 if (fd < 0)
469 return -1;
470 rv = test_partition(fd);
471 close(fd);
472 return rv;
473}
691c6ee1 474
de5a472e 475int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 476{
265e0f17 477 int copies, first;
de5a472e
N
478 int i;
479 int avail_disks = 0;
480
481 for (i = 0; i < raid_disks; i++)
482 avail_disks += !!avail[i];
483
64c4757e 484 switch (level) {
265e0f17
NB
485 case 10:
486 /* This is the tricky one - we need to check
487 * which actual disks are present.
488 */
702b557b 489 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 490 first = 0;
265e0f17
NB
491 do {
492 /* there must be one of the 'copies' form 'first' */
493 int n = copies;
ca3b6696 494 int cnt = 0;
b3ec716d 495 int this = first;
265e0f17 496 while (n--) {
b3ec716d 497 if (avail[this])
265e0f17 498 cnt++;
b3ec716d 499 this = (this+1) % raid_disks;
265e0f17
NB
500 }
501 if (cnt == 0)
502 return 0;
b3ec716d 503 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
504 } while (first != 0);
505 return 1;
e5329c37 506
df0d4ea0 507 case LEVEL_MULTIPATH:
e0d19036 508 return avail_disks>= 1;
df0d4ea0 509 case LEVEL_LINEAR:
64c4757e
NB
510 case 0:
511 return avail_disks == raid_disks;
512 case 1:
513 return avail_disks >= 1;
514 case 4:
330d6900
N
515 if (avail_disks == raid_disks - 1 &&
516 !avail[raid_disks - 1])
517 /* If just the parity device is missing, then we
518 * have enough, even if not clean
519 */
520 return 1;
521 /* FALL THROUGH */
64c4757e 522 case 5:
583315d9
NB
523 if (clean)
524 return avail_disks >= raid_disks-1;
525 else
526 return avail_disks >= raid_disks;
98c6faba 527 case 6:
583315d9
NB
528 if (clean)
529 return avail_disks >= raid_disks-2;
530 else
531 return avail_disks >= raid_disks;
64c4757e
NB
532 default:
533 return 0;
534 }
535}
536
8453e704
N
537int enough_fd(int fd)
538{
539 struct mdu_array_info_s array;
540 struct mdu_disk_info_s disk;
aabe020d 541 int i, rv;
8453e704
N
542 char *avail;
543
544 if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 ||
545 array.raid_disks <= 0)
546 return 0;
503975b9 547 avail = xcalloc(array.raid_disks, 1);
ca3b6696 548 for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704
N
549 disk.number = i;
550 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
551 continue;
9e6d9291
N
552 if (disk.major == 0 && disk.minor == 0)
553 continue;
554 array.nr_disks--;
555
8453e704
N
556 if (! (disk.state & (1<<MD_DISK_SYNC)))
557 continue;
558 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
559 continue;
8453e704
N
560 avail[disk.raid_disk] = 1;
561 }
562 /* This is used on an active array, so assume it is clean */
aabe020d 563 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 564 1, avail);
aabe020d
JS
565 free(avail);
566 return rv;
8453e704
N
567}
568
22e263f6
AC
569const int uuid_zero[4] = { 0, 0, 0, 0 };
570
f277ce36 571int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 572{
f277ce36
NB
573 if (swapuuid) {
574 /* parse uuids are hostendian.
575 * uuid's from some superblocks are big-ending
aba69144 576 * if there is a difference, we need to swap..
f277ce36
NB
577 */
578 unsigned char *ac = (unsigned char *)a;
579 unsigned char *bc = (unsigned char *)b;
580 int i;
ca3b6696 581 for (i = 0; i < 16; i += 4) {
f277ce36
NB
582 if (ac[i+0] != bc[i+3] ||
583 ac[i+1] != bc[i+2] ||
584 ac[i+2] != bc[i+1] ||
585 ac[i+3] != bc[i+0])
586 return 0;
587 }
588 return 1;
589 } else {
590 if (a[0]==b[0] &&
591 a[1]==b[1] &&
592 a[2]==b[2] &&
593 a[3]==b[3])
594 return 1;
595 return 0;
596 }
64c4757e 597}
ca3b6696 598
350f29f9
NB
599void copy_uuid(void *a, int b[4], int swapuuid)
600{
601 if (swapuuid) {
602 /* parse uuids are hostendian.
603 * uuid's from some superblocks are big-ending
604 * if there is a difference, we need to swap..
605 */
606 unsigned char *ac = (unsigned char *)a;
607 unsigned char *bc = (unsigned char *)b;
608 int i;
ca3b6696 609 for (i = 0; i < 16; i += 4) {
350f29f9
NB
610 ac[i+0] = bc[i+3];
611 ac[i+1] = bc[i+2];
612 ac[i+2] = bc[i+1];
613 ac[i+3] = bc[i+0];
614 }
615 } else
616 memcpy(a, b, 16);
617}
64c4757e 618
aae5a112 619char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 620{
9968e376 621 int i, j;
d7288ddc
N
622 char uuid[16];
623 char *c = buf;
624 strcpy(c, "UUID-");
625 c += strlen(c);
aae5a112 626 copy_uuid(uuid, id, swap);
9968e376 627 for (i = 0; i < 4; i++) {
9968e376 628 if (i)
ff54de6e 629 *c++ = sep;
9968e376
DW
630 for (j = 3; j >= 0; j--) {
631 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
632 c+= 2;
633 }
d7288ddc
N
634 }
635 return buf;
aae5a112
DW
636
637}
638
639char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
640{
86983cce
N
641 // dirty hack to work around an issue with super1 superblocks...
642 // super1 superblocks need swapuuid set in order for assembly to
643 // work, but can't have it set if we want this printout to match
644 // all the other uuid printouts in super1.c, so we force swapuuid
645 // to 1 to make our printout match the rest of super1
646 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
647}
648
435d4ebb 649#ifndef MDASSEMBLE
682c7051
NB
650int check_ext2(int fd, char *name)
651{
652 /*
653 * Check for an ext2fs file system.
654 * Superblock is always 1K at 1K offset
655 *
656 * s_magic is le16 at 56 == 0xEF53
657 * report mtime - le32 at 44
658 * blocks - le32 at 4
659 * logblksize - le32 at 24
660 */
661 unsigned char sb[1024];
662 time_t mtime;
2538ba2a
N
663 unsigned long long size;
664 int bsize;
682c7051
NB
665 if (lseek(fd, 1024,0)!= 1024)
666 return 0;
667 if (read(fd, sb, 1024)!= 1024)
668 return 0;
669 if (sb[56] != 0x53 || sb[57] != 0xef)
670 return 0;
671
672 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
673 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
674 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 675 size <<= bsize;
e7b84f9d 676 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 677 name);
2538ba2a 678 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
679 return 1;
680}
681
682int check_reiser(int fd, char *name)
683{
684 /*
685 * superblock is at 64K
686 * size is 1024;
687 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
688 *
689 */
690 unsigned char sb[1024];
2538ba2a 691 unsigned long long size;
682c7051
NB
692 if (lseek(fd, 64*1024, 0) != 64*1024)
693 return 0;
694 if (read(fd, sb, 1024) != 1024)
695 return 0;
ca3b6696
N
696 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
697 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 698 return 0;
e7b84f9d 699 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 700 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 701 cont_err("size = %lluK\n", size*4);
aba69144 702
682c7051
NB
703 return 1;
704}
705
706int check_raid(int fd, char *name)
707{
4b1ac34b 708 struct mdinfo info;
682c7051 709 time_t crtime;
d078d77c 710 char *level;
82d9eba6 711 struct supertype *st = guess_super(fd);
f9ce90ba 712
cb8f6859
N
713 if (!st)
714 return 0;
3da92f27 715 st->ss->load_super(st, fd, name);
82d9eba6 716 /* Looks like a raid array .. */
e7b84f9d 717 pr_err("%s appears to be part of a raid array:\n",
82d9eba6 718 name);
a5d85af7 719 st->ss->getinfo_super(st, &info, NULL);
3da92f27 720 st->ss->free_super(st);
82d9eba6 721 crtime = info.array.ctime;
d078d77c
NB
722 level = map_num(pers, info.array.level);
723 if (!level) level = "-unknown-";
d33f1518
N
724 cont_err("level=%s devices=%d ctime=%s",
725 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 726 return 1;
682c7051
NB
727}
728
682c7051
NB
729int ask(char *mesg)
730{
731 char *add = "";
732 int i;
ca3b6696 733 for (i = 0; i < 5; i++) {
682c7051
NB
734 char buf[100];
735 fprintf(stderr, "%s%s", mesg, add);
736 fflush(stderr);
737 if (fgets(buf, 100, stdin)==NULL)
738 return 0;
739 if (buf[0]=='y' || buf[0]=='Y')
740 return 1;
741 if (buf[0]=='n' || buf[0]=='N')
742 return 0;
743 add = "(y/n) ";
744 }
e7b84f9d 745 pr_err("assuming 'no'\n");
682c7051
NB
746 return 0;
747}
435d4ebb 748#endif /* MDASSEMBLE */
682c7051 749
8d80900b 750int is_standard(char *dev, int *nump)
e5329c37
NB
751{
752 /* tests if dev is a "standard" md dev name.
753 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 754 * where NN is a string of digits
598f0d58
NB
755 * Returns 1 if a partitionable standard,
756 * -1 if non-partitonable,
757 * 0 if not a standard name.
e5329c37 758 */
8d80900b 759 char *d = strrchr(dev, '/');
ca3b6696 760 int type = 0;
8d80900b
NB
761 int num;
762 if (!d)
e5329c37 763 return 0;
ca3b6696
N
764 if (strncmp(d, "/d",2) == 0)
765 d += 2, type = 1; /* /dev/md/dN{pM} */
766 else if (strncmp(d, "/md_d", 5) == 0)
767 d += 5, type = 1; /* /dev/md_dN{pM} */
768 else if (strncmp(d, "/md", 3) == 0)
769 d += 3, type = -1; /* /dev/mdN */
770 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
771 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
772 else
773 return 0;
8d80900b 774 if (!*d)
e5329c37 775 return 0;
8d80900b
NB
776 num = atoi(d);
777 while (isdigit(*d))
778 d++;
779 if (*d)
e5329c37 780 return 0;
8d80900b
NB
781 if (nump) *nump = num;
782
783 return type;
e5329c37
NB
784}
785
4b1ac34b 786unsigned long calc_csum(void *super, int bytes)
82b27616 787{
56eb10c0 788 unsigned long long newcsum = 0;
82b27616 789 int i;
4b1ac34b
NB
790 unsigned int csum;
791 unsigned int *superc = (unsigned int*) super;
82b27616 792
ca3b6696
N
793 for(i = 0; i < bytes/4; i++)
794 newcsum += superc[i];
82b27616 795 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 796#ifdef __alpha__
aba69144 797/* The in-kernel checksum calculation is always 16bit on
570c0542 798 * the alpha, though it is 32 bit on i386...
ca3b6696 799 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
800 * a way that it shouldn't).
801 */
802 csum = (csum & 0xffff) + (csum >> 16);
803 csum = (csum & 0xffff) + (csum >> 16);
804#endif
82b27616
NB
805 return csum;
806}
cd29a5c8 807
435d4ebb 808#ifndef MDASSEMBLE
56eb10c0 809char *human_size(long long bytes)
cd29a5c8
NB
810{
811 static char buf[30];
d5d3721e
NB
812
813 /* We convert bytes to either centi-M{ega,ibi}bytes or
814 * centi-G{igi,ibi}bytes, with appropriate rounding,
815 * and then print 1/100th of those as a decimal.
816 * We allow upto 2048Megabytes before converting to
817 * gigabytes, as that shows more precision and isn't
818 * too large a number.
01619b48 819 * Terabytes are not yet handled.
d5d3721e 820 */
cd29a5c8 821
56eb10c0 822 if (bytes < 5000*1024)
ca3b6696 823 buf[0] = 0;
d5d3721e 824 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 825 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 826 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 827 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
828 cMiB/100 , cMiB % 100,
829 cMB/100, cMB % 100);
830 } else {
93d3bd3b 831 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 832 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 833 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
834 cGiB/100 , cGiB % 100,
835 cGB/100, cGB % 100);
836 }
cd29a5c8
NB
837 return buf;
838}
e0d19036 839
f0ec6710 840char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
841{
842 static char buf[30];
e0d19036 843
570abc6f
MN
844 /* We convert bytes to either centi-M{ega,ibi}bytes or
845 * centi-G{igi,ibi}bytes, with appropriate rounding,
846 * and then print 1/100th of those as a decimal.
847 * We allow upto 2048Megabytes before converting to
848 * gigabytes, as that shows more precision and isn't
849 * too large a number.
850 * Terabytes are not yet handled.
f0ec6710
MN
851 *
852 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
853 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
854 */
855
e0d19036 856 if (bytes < 5000*1024)
570abc6f 857 buf[0] = 0;
f0ec6710
MN
858 else if (prefix == IEC) {
859 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 860 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710
MN
861 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
862 cMiB/100 , cMiB % 100);
863 } else {
93d3bd3b 864 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710
MN
865 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
866 cGiB/100 , cGiB % 100);
867 }
868 }
869 else if (prefix == JEDEC) {
870 if (bytes < 2*1024LL*1024LL*1024LL) {
871 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
872 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
873 cMB/100, cMB % 100);
874 } else {
875 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
876 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
877 cGB/100 , cGB % 100);
878 }
570abc6f 879 }
f0ec6710
MN
880 else
881 buf[0] = 0;
570abc6f 882
e0d19036
NB
883 return buf;
884}
e4965ef8
N
885
886void print_r10_layout(int layout)
887{
888 int near = layout & 255;
889 int far = (layout >> 8) & 255;
890 int offset = (layout&0x10000);
891 char *sep = "";
892
893 if (near != 1) {
894 printf("%s near=%d", sep, near);
895 sep = ",";
896 }
897 if (far != 1)
898 printf("%s %s=%d", sep, offset?"offset":"far", far);
899 if (near*far == 1)
900 printf("NO REDUNDANCY");
901}
435d4ebb 902#endif
e0d19036 903
5f8097be
NB
904unsigned long long calc_array_size(int level, int raid_disks, int layout,
905 int chunksize, unsigned long long devsize)
577e8448 906{
e03640bd
CA
907 if (level == 1)
908 return devsize;
577e8448
AK
909 devsize &= ~(unsigned long long)((chunksize>>9)-1);
910 return get_data_disks(level, layout, raid_disks) * devsize;
911}
912
913int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
914{
915 int data_disks = 0;
916 switch (level) {
ca3b6696
N
917 case 0: data_disks = raid_disks;
918 break;
919 case 1: data_disks = 1;
920 break;
5f8097be 921 case 4:
ca3b6696
N
922 case 5: data_disks = raid_disks - 1;
923 break;
924 case 6: data_disks = raid_disks - 2;
925 break;
5f8097be
NB
926 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
927 break;
928 }
577e8448
AK
929
930 return data_disks;
5f8097be
NB
931}
932
4dd2df09
N
933int devnm2devid(char *devnm)
934{
935 /* First look in /sys/block/$DEVNM/dev for %d:%d
936 * If that fails, try parsing out a number
937 */
938 char path[100];
939 char *ep;
940 int fd;
941 int mjr,mnr;
942
943 sprintf(path, "/sys/block/%s/dev", devnm);
944 fd = open(path, O_RDONLY);
945 if (fd >= 0) {
946 char buf[20];
947 int n = read(fd, buf, sizeof(buf));
948 close(fd);
949 if (n > 0)
950 buf[n] = 0;
951 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
952 return makedev(mjr, mnr);
953 }
954 if (strncmp(devnm, "md_d", 4) == 0 &&
955 isdigit(devnm[4]) &&
956 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
957 ep > devnm && *ep == 0)
958 return makedev(get_mdp_major(), mnr << MdpMinorShift);
959
960 if (strncmp(devnm, "md", 2) == 0 &&
961 isdigit(devnm[2]) &&
962 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
963 ep > devnm && *ep == 0)
964 return makedev(MD_MAJOR, mnr);
965
966 return 0;
967}
968
5a23a06e 969#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 970char *get_md_name(char *devnm)
e0d19036
NB
971{
972 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 973 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 974
e0d19036
NB
975 static char devname[50];
976 struct stat stb;
4dd2df09 977 dev_t rdev = devnm2devid(devnm);
dd0781e5 978 char *dn;
98c6faba 979
4dd2df09
N
980 if (rdev == 0)
981 return 0;
982 if (strncmp(devnm, "md_", 3) == 0) {
983 snprintf(devname, sizeof(devname), "/dev/md/%s",
984 devnm + 3);
98c6faba
NB
985 if (stat(devname, &stb) == 0
986 && (S_IFMT&stb.st_mode) == S_IFBLK
987 && (stb.st_rdev == rdev))
988 return devname;
989 }
4dd2df09
N
990 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
991 if (stat(devname, &stb) == 0
992 && (S_IFMT&stb.st_mode) == S_IFBLK
993 && (stb.st_rdev == rdev))
994 return devname;
995
996 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
997 if (stat(devname, &stb) == 0
998 && (S_IFMT&stb.st_mode) == S_IFBLK
999 && (stb.st_rdev == rdev))
1000 return devname;
1001
16c6fa80 1002 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
1003 if (dn)
1004 return dn;
4dd2df09 1005 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 1006 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
1007 if (errno != EEXIST)
1008 return NULL;
e0d19036
NB
1009
1010 if (stat(devname, &stb) == 0
1011 && (S_IFMT&stb.st_mode) == S_IFBLK
1012 && (stb.st_rdev == rdev))
1013 return devname;
1014 unlink(devname);
1015 return NULL;
1016}
1017
1018void put_md_name(char *name)
1019{
ca3b6696 1020 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
1021 unlink(name);
1022}
435d4ebb 1023#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 1024
cc742d38
N
1025int get_maj_min(char *dev, int *major, int *minor)
1026{
1027 char *e;
1028 *major = strtoul(dev, &e, 0);
1029 return (e > dev && *e == ':' && e[1] &&
1030 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1031 *e == 0);
1032}
1033
8b0dabea
NB
1034int dev_open(char *dev, int flags)
1035{
1036 /* like 'open', but if 'dev' matches %d:%d, create a temp
1037 * block device and open that
1038 */
8b0dabea
NB
1039 int fd = -1;
1040 char devname[32];
e81cdd9f 1041 int major;
8b0dabea 1042 int minor;
e81cdd9f
NB
1043
1044 if (!dev) return -1;
6df6a774 1045 flags |= O_DIRECT;
e81cdd9f 1046
cc742d38 1047 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1048 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1049 (int)getpid(), major, minor);
1050 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1051 fd = open(devname, flags);
1052 unlink(devname);
6df6a774
N
1053 }
1054 if (fd < 0) {
9dc7d357 1055 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
1056 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
1057 (int)getpid(), major, minor);
ca3b6696 1058 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
1059 fd = open(devname, flags);
1060 unlink(devname);
1061 }
8b0dabea
NB
1062 }
1063 } else
6df6a774 1064 fd = open(dev, flags);
8b0dabea
NB
1065 return fd;
1066}
f9ce90ba 1067
4dd2df09 1068int open_dev_flags(char *devnm, int flags)
e8a70c89 1069{
4dd2df09 1070 int devid;
e8a70c89
N
1071 char buf[20];
1072
4dd2df09
N
1073 devid = devnm2devid(devnm);
1074 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1075 return dev_open(buf, flags);
1076}
1077
4dd2df09 1078int open_dev(char *devnm)
d998b738 1079{
4dd2df09 1080 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1081}
1082
4dd2df09 1083int open_dev_excl(char *devnm)
a931db9e
NB
1084{
1085 char buf[20];
1086 int i;
7187750e 1087 int flags = O_RDWR;
4dd2df09 1088 int devid = devnm2devid(devnm);
399e0b97 1089 long delay = 1000;
a931db9e 1090
4dd2df09 1091 sprintf(buf, "%d:%d", major(devid), minor(devid));
ca3b6696 1092 for (i = 0 ; i < 25 ; i++) {
7187750e 1093 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1094 if (fd >= 0)
1095 return fd;
7187750e
N
1096 if (errno == EACCES && flags == O_RDWR) {
1097 flags = O_RDONLY;
1098 continue;
1099 }
a931db9e
NB
1100 if (errno != EBUSY)
1101 return fd;
399e0b97
N
1102 usleep(delay);
1103 if (delay < 200000)
1104 delay *= 2;
a931db9e
NB
1105 }
1106 return -1;
1107}
1108
9008ed1c
N
1109int same_dev(char *one, char *two)
1110{
1111 struct stat st1, st2;
1112 if (stat(one, &st1) != 0)
1113 return 0;
1114 if (stat(two, &st2) != 0)
1115 return 0;
1116 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1117 return 0;
1118 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1119 return 0;
1120 return st1.st_rdev == st2.st_rdev;
1121}
1122
a7c6e3fb 1123void wait_for(char *dev, int fd)
a714580e
N
1124{
1125 int i;
a7c6e3fb 1126 struct stat stb_want;
399e0b97 1127 long delay = 1000;
a7c6e3fb
N
1128
1129 if (fstat(fd, &stb_want) != 0 ||
1130 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1131 return;
a714580e 1132
ca3b6696 1133 for (i = 0 ; i < 25 ; i++) {
a714580e 1134 struct stat stb;
a7c6e3fb
N
1135 if (stat(dev, &stb) == 0 &&
1136 (stb.st_mode & S_IFMT) == S_IFBLK &&
1137 (stb.st_rdev == stb_want.st_rdev))
a714580e 1138 return;
399e0b97
N
1139 usleep(delay);
1140 if (delay < 200000)
1141 delay *= 2;
a714580e 1142 }
436305c6 1143 if (i == 25)
1ade5cc1 1144 dprintf("timeout waiting for %s\n", dev);
a714580e
N
1145}
1146
0f22b998
N
1147struct superswitch *superlist[] =
1148{
1149 &super0, &super1,
1150 &super_ddf, &super_imsm,
0592faeb 1151 &mbr, &gpt,
0f22b998 1152 NULL };
f9ce90ba 1153
ea24acd0 1154#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1155
4725bc31 1156struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1157{
1686dc25
NB
1158 mdu_array_info_t array;
1159 int vers;
1160 int minor;
1161 struct supertype *st = NULL;
7e0f6979 1162 struct mdinfo *sra;
142cb9e1 1163 char *verstr;
1686dc25
NB
1164 char version[20];
1165 int i;
f7e7067b 1166 char *subarray = NULL;
4dd2df09 1167 char container[32] = "";
1686dc25 1168
4dd2df09 1169 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1170
1171 if (sra) {
7e0f6979
NB
1172 vers = sra->array.major_version;
1173 minor = sra->array.minor_version;
142cb9e1 1174 verstr = sra->text_version;
1686dc25
NB
1175 } else {
1176 if (ioctl(fd, GET_ARRAY_INFO, &array))
1177 array.major_version = array.minor_version = 0;
1178 vers = array.major_version;
1179 minor = array.minor_version;
142cb9e1 1180 verstr = "";
6fbba4c9 1181 }
82d9eba6 1182
1686dc25
NB
1183 if (vers != -1) {
1184 sprintf(version, "%d.%d", vers, minor);
1185 verstr = version;
6fbba4c9 1186 }
3c558363 1187 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1188 char *dev = verstr+1;
5f7e44b2 1189
f7e7067b 1190 subarray = strchr(dev, '/');
92d49ecf 1191 if (subarray) {
f7e7067b 1192 *subarray++ = '\0';
503975b9 1193 subarray = xstrdup(subarray);
92d49ecf 1194 }
4dd2df09 1195 strcpy(container, dev);
f7e7067b
NB
1196 if (sra)
1197 sysfs_free(sra);
5f7e44b2 1198 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1199 if (sra && sra->text_version[0])
1200 verstr = sra->text_version;
1201 else
1202 verstr = "-no-metadata-";
f7e7067b
NB
1203 }
1204
1205 for (i = 0; st == NULL && superlist[i] ; i++)
1206 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
1207
1208 if (sra)
1209 sysfs_free(sra);
f7e7067b 1210 if (st) {
3b0896f8 1211 st->sb = NULL;
1f49fb3a
N
1212 if (subarrayp)
1213 *subarrayp = subarray;
4dd2df09
N
1214 strcpy(st->container_devnm, container);
1215 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1216 } else
1217 free(subarray);
5f7e44b2 1218
82d9eba6 1219 return st;
f9ce90ba 1220}
ea24acd0
NB
1221#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1222
0f0749ad 1223int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1224{
1225 char buf[20];
1226 int fd;
1227
1228 sprintf(buf, "%d:%d", major(id), minor(id));
1229 fd = dev_open(buf, O_RDONLY);
1230 if (fd < 0)
1231 return 0;
1232 if (get_dev_size(fd, NULL, size)) {
1233 close(fd);
1234 return 1;
1235 }
1236 close(fd);
1237 return 0;
1238}
f9ce90ba 1239
159c3a1a 1240struct supertype *dup_super(struct supertype *orig)
3da92f27 1241{
159c3a1a 1242 struct supertype *st;
1686dc25 1243
d2ca6449
NB
1244 if (!orig)
1245 return orig;
503975b9 1246 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1247 st->ss = orig->ss;
1248 st->max_devs = orig->max_devs;
1249 st->minor_version = orig->minor_version;
cb8f6859 1250 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1251 st->data_offset = orig->data_offset;
159c3a1a
NB
1252 st->sb = NULL;
1253 st->info = NULL;
1254 return st;
3da92f27
NB
1255}
1256
54887ad8 1257struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1258{
1259 /* try each load_super to find the best match,
1260 * and return the best superswitch
1261 */
82d9eba6
NB
1262 struct superswitch *ss;
1263 struct supertype *st;
f21e18ca 1264 time_t besttime = 0;
570c0542 1265 int bestsuper = -1;
f9ce90ba
NB
1266 int i;
1267
503975b9 1268 st = xcalloc(1, sizeof(*st));
4dd2df09 1269 st->container_devnm[0] = 0;
d1d599ea 1270
ca3b6696 1271 for (i = 0 ; superlist[i]; i++) {
f9ce90ba
NB
1272 int rv;
1273 ss = superlist[i];
54887ad8
N
1274 if (guess_type == guess_array && ss->add_to_super == NULL)
1275 continue;
1276 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1277 continue;
ef609477 1278 memset(st, 0, sizeof(*st));
df3346e6 1279 st->ignore_hw_compat = 1;
3da92f27 1280 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1281 if (rv == 0) {
1282 struct mdinfo info;
a5d85af7 1283 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1284 if (bestsuper == -1 ||
1285 besttime < info.array.ctime) {
1286 bestsuper = i;
1287 besttime = info.array.ctime;
570c0542 1288 }
3da92f27 1289 ss->free_super(st);
570c0542
NB
1290 }
1291 }
1292 if (bestsuper != -1) {
1293 int rv;
ef609477 1294 memset(st, 0, sizeof(*st));
df3346e6 1295 st->ignore_hw_compat = 1;
3da92f27 1296 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1297 if (rv == 0) {
5e747af2 1298 superlist[bestsuper]->free_super(st);
82d9eba6 1299 return st;
f9ce90ba
NB
1300 }
1301 }
570c0542 1302 free(st);
f9ce90ba
NB
1303 return NULL;
1304}
fe6729fa 1305
beae1dfe
NB
1306/* Return size of device in bytes */
1307int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1308{
1309 unsigned long long ldsize;
c2c9bb6f
NB
1310 struct stat st;
1311
1312 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1313 ldsize = (unsigned long long)st.st_size;
1314 else
beae1dfe
NB
1315#ifdef BLKGETSIZE64
1316 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1317#endif
1318 {
1319 unsigned long dsize;
1320 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1321 ldsize = dsize;
1322 ldsize <<= 9;
1323 } else {
1324 if (dname)
e7b84f9d 1325 pr_err("Cannot get size of %s: %s\b",
beae1dfe
NB
1326 dname, strerror(errno));
1327 return 0;
1328 }
1329 }
1330 *sizep = ldsize;
1331 return 1;
1332}
8fac0577 1333
3a371610
N
1334/* Return true if this can only be a container, not a member device.
1335 * i.e. is and md device and size is zero
1336 */
1337int must_be_container(int fd)
1338{
1339 unsigned long long size;
1340 if (md_get_version(fd) < 0)
1341 return 0;
1342 if (get_dev_size(fd, NULL, &size) == 0)
1343 return 1;
1344 if (size == 0)
1345 return 1;
1346 return 0;
1347}
034b203a
TM
1348
1349/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1350 * Returns: 1 if successful
1351 * -1 for unknown partition type
1352 * 0 for other errors
1353 */
1354static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1355{
056b331e 1356 struct GPT gpt;
034b203a 1357 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1358 struct GPT_part_entry *part;
1359 char buf[512];
034b203a
TM
1360 unsigned long long curr_part_end;
1361 unsigned all_partitions, entry_size;
f21e18ca 1362 unsigned part_nr;
034b203a
TM
1363
1364 *endofpart = 0;
1365
056b331e 1366 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1367 /* skip protective MBR */
034b203a 1368 lseek(fd, 512, SEEK_SET);
73e658d8 1369 /* read GPT header */
056b331e 1370 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1371 return 0;
1372
1373 /* get the number of partition entries and the entry size */
056b331e
N
1374 all_partitions = __le32_to_cpu(gpt.part_cnt);
1375 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1376
1377 /* Check GPT signature*/
056b331e 1378 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1379 return -1;
1380
1381 /* sanity checks */
1382 if (all_partitions > 1024 ||
db7fdfe4 1383 entry_size > sizeof(buf))
034b203a
TM
1384 return -1;
1385
db7fdfe4
JS
1386 part = (struct GPT_part_entry *)buf;
1387
ca3b6696 1388 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1389 /* read partition entry */
db7fdfe4 1390 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1391 return 0;
1392
034b203a 1393 /* is this valid partition? */
db7fdfe4 1394 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1395 /* check the last lba for the current partition */
db7fdfe4 1396 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1397 if (curr_part_end > *endofpart)
1398 *endofpart = curr_part_end;
1399 }
1400
034b203a
TM
1401 }
1402 return 1;
1403}
1404
1405/* Sets endofpart parameter to the last block used by the last partition on the device.
1406 * Returns: 1 if successful
1407 * -1 for unknown partition type
1408 * 0 for other errors
1409 */
1410static int get_last_partition_end(int fd, unsigned long long *endofpart)
1411{
056b331e 1412 struct MBR boot_sect;
034b203a
TM
1413 struct MBR_part_record *part;
1414 unsigned long long curr_part_end;
f21e18ca 1415 unsigned part_nr;
034b203a
TM
1416 int retval = 0;
1417
1418 *endofpart = 0;
1419
056b331e 1420 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1421 /* read MBR */
1422 lseek(fd, 0, 0);
056b331e 1423 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1424 goto abort;
1425
1426 /* check MBP signature */
056b331e 1427 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1428 retval = 1;
1429 /* found the correct signature */
056b331e 1430 part = boot_sect.parts;
034b203a 1431
ca3b6696 1432 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
034b203a
TM
1433 /* check for GPT type */
1434 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1435 retval = get_gpt_last_partition_end(fd, endofpart);
1436 break;
1437 }
1438 /* check the last used lba for the current partition */
1439 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1440 __le32_to_cpu(part->blocks_num);
1441 if (curr_part_end > *endofpart)
1442 *endofpart = curr_part_end;
1443
1444 part++;
1445 }
1446 } else {
1447 /* Unknown partition table */
1448 retval = -1;
1449 }
1450 abort:
1451 return retval;
1452}
1453
53ed6ac3
KW
1454int check_partitions(int fd, char *dname, unsigned long long freesize,
1455 unsigned long long size)
034b203a
TM
1456{
1457 /*
1458 * Check where the last partition ends
1459 */
1460 unsigned long long endofpart;
1461 int ret;
1462
1463 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1464 /* There appears to be a partition table here */
1465 if (freesize == 0) {
1466 /* partitions will not be visible in new device */
e7b84f9d
N
1467 pr_err("partition table exists on %s but will be lost or\n"
1468 " meaningless after creating array\n",
1469 dname);
034b203a
TM
1470 return 1;
1471 } else if (endofpart > freesize) {
1472 /* last partition overlaps metadata */
e7b84f9d
N
1473 pr_err("metadata will over-write last partition on %s.\n",
1474 dname);
034b203a 1475 return 1;
53ed6ac3
KW
1476 } else if (size && endofpart > size) {
1477 /* partitions will be truncated in new device */
e7b84f9d
N
1478 pr_err("array size is too small to cover all partitions on %s.\n",
1479 dname);
53ed6ac3 1480 return 1;
034b203a
TM
1481 }
1482 }
1483 return 0;
1484}
1485
a322f70c
DW
1486int open_container(int fd)
1487{
1488 /* 'fd' is a block device. Find out if it is in use
1489 * by a container, and return an open fd on that container.
1490 */
1491 char path[256];
1492 char *e;
1493 DIR *dir;
1494 struct dirent *de;
1495 int dfd, n;
1496 char buf[200];
1497 int major, minor;
1498 struct stat st;
1499
1500 if (fstat(fd, &st) != 0)
1501 return -1;
1502 sprintf(path, "/sys/dev/block/%d:%d/holders",
1503 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1504 e = path + strlen(path);
1505
1506 dir = opendir(path);
1507 if (!dir)
1508 return -1;
1509 while ((de = readdir(dir))) {
1510 if (de->d_ino == 0)
1511 continue;
1512 if (de->d_name[0] == '.')
1513 continue;
a994592d
N
1514 /* Need to make sure it is a container and not a volume */
1515 sprintf(e, "/%s/md/metadata_version", de->d_name);
1516 dfd = open(path, O_RDONLY);
1517 if (dfd < 0)
1518 continue;
1519 n = read(dfd, buf, sizeof(buf));
1520 close(dfd);
1521 if (n <= 0 || (unsigned)n >= sizeof(buf))
1522 continue;
1523 buf[n] = 0;
1524 if (strncmp(buf, "external", 8) != 0 ||
1525 n < 10 ||
1526 buf[9] == '/')
1527 continue;
a322f70c
DW
1528 sprintf(e, "/%s/dev", de->d_name);
1529 dfd = open(path, O_RDONLY);
1530 if (dfd < 0)
1531 continue;
1532 n = read(dfd, buf, sizeof(buf));
1533 close(dfd);
f21e18ca 1534 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1535 continue;
1536 buf[n] = 0;
1537 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1538 continue;
1539 sprintf(buf, "%d:%d", major, minor);
1540 dfd = dev_open(buf, O_RDONLY);
1541 if (dfd >= 0) {
1542 closedir(dir);
1543 return dfd;
1544 }
1545 }
355726fa 1546 closedir(dir);
a322f70c
DW
1547 return -1;
1548}
1549
33414a01
DW
1550struct superswitch *version_to_superswitch(char *vers)
1551{
1552 int i;
1553
1554 for (i = 0; superlist[i]; i++) {
1555 struct superswitch *ss = superlist[i];
1556
1557 if (strcmp(vers, ss->name) == 0)
1558 return ss;
1559 }
1560
1561 return NULL;
1562}
1563
4dd2df09
N
1564int metadata_container_matches(char *metadata, char *devnm)
1565{
1566 /* Check if 'devnm' is the container named in 'metadata'
1567 * which is
1568 * /containername/componentname or
1569 * -containername/componentname
1570 */
1571 int l;
1572 if (*metadata != '/' && *metadata != '-')
1573 return 0;
1574 l = strlen(devnm);
1575 if (strncmp(metadata+1, devnm, l) != 0)
1576 return 0;
1577 if (metadata[l+1] != '/')
1578 return 0;
1579 return 1;
1580}
1581
1582int metadata_subdev_matches(char *metadata, char *devnm)
1583{
1584 /* Check if 'devnm' is the subdev named in 'metadata'
1585 * which is
1586 * /containername/subdev or
1587 * -containername/subdev
1588 */
1589 char *sl;
1590 if (*metadata != '/' && *metadata != '-')
1591 return 0;
1592 sl = strchr(metadata+1, '/');
1593 if (!sl)
1594 return 0;
1595 if (strcmp(sl+1, devnm) == 0)
1596 return 1;
1597 return 0;
1598}
1599
33414a01
DW
1600int is_container_member(struct mdstat_ent *mdstat, char *container)
1601{
1602 if (mdstat->metadata_version == NULL ||
1603 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1604 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1605 return 0;
1606
1607 return 1;
1608}
1609
1610int is_subarray_active(char *subarray, char *container)
1611{
1612 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1613 struct mdstat_ent *ent;
1614
e5408a32
DW
1615 for (ent = mdstat; ent; ent = ent->next)
1616 if (is_container_member(ent, container))
e5e5d7ce 1617 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1618 break;
33414a01
DW
1619
1620 free_mdstat(mdstat);
1621
1622 return ent != NULL;
1623}
1624
1625/* open_subarray - opens a subarray in a container
1626 * @dev: container device name
feab51f8 1627 * @st: empty supertype
33414a01
DW
1628 * @quiet: block reporting errors flag
1629 *
1630 * On success returns an fd to a container and fills in *st
1631 */
feab51f8 1632int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1633{
1634 struct mdinfo *mdi;
a951a4f7 1635 struct mdinfo *info;
33414a01 1636 int fd, err = 1;
4dd2df09 1637 char *_devnm;
33414a01
DW
1638
1639 fd = open(dev, O_RDWR|O_EXCL);
1640 if (fd < 0) {
1641 if (!quiet)
e7b84f9d 1642 pr_err("Couldn't open %s, aborting\n",
33414a01 1643 dev);
b990032d 1644 return -1;
33414a01
DW
1645 }
1646
4dd2df09
N
1647 _devnm = fd2devnm(fd);
1648 if (_devnm == NULL) {
33414a01 1649 if (!quiet)
e7b84f9d
N
1650 pr_err("Failed to determine device number for %s\n",
1651 dev);
33414a01
DW
1652 goto close_fd;
1653 }
4dd2df09 1654 strcpy(st->devnm, _devnm);
33414a01 1655
4dd2df09 1656 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1657 if (!mdi) {
1658 if (!quiet)
e7b84f9d 1659 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1660 dev);
1661 goto close_fd;
1662 }
1663
1664 if (mdi->array.level != UnSet) {
1665 if (!quiet)
e7b84f9d 1666 pr_err("%s is not a container\n", dev);
33414a01
DW
1667 goto free_sysfs;
1668 }
1669
1670 st->ss = version_to_superswitch(mdi->text_version);
1671 if (!st->ss) {
1672 if (!quiet)
e7b84f9d
N
1673 pr_err("Operation not supported for %s metadata\n",
1674 mdi->text_version);
33414a01
DW
1675 goto free_sysfs;
1676 }
1677
4dd2df09 1678 if (st->devnm[0] == 0) {
33414a01 1679 if (!quiet)
e7b84f9d 1680 pr_err("Failed to allocate device name\n");
33414a01
DW
1681 goto free_sysfs;
1682 }
1683
db20d413 1684 if (!st->ss->load_container) {
33414a01 1685 if (!quiet)
e7b84f9d 1686 pr_err("%s is not a container\n", dev);
4dd2df09 1687 goto free_sysfs;
33414a01
DW
1688 }
1689
db20d413 1690 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1691 if (!quiet)
e7b84f9d 1692 pr_err("Failed to load metadata for %s\n",
db20d413 1693 dev);
4dd2df09 1694 goto free_sysfs;
33414a01
DW
1695 }
1696
a951a4f7
N
1697 info = st->ss->container_content(st, subarray);
1698 if (!info) {
1699 if (!quiet)
e7b84f9d 1700 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1701 subarray, dev);
1702 goto free_super;
1703 }
1704 free(info);
1705
33414a01
DW
1706 err = 0;
1707
1708 free_super:
1709 if (err)
1710 st->ss->free_super(st);
33414a01
DW
1711 free_sysfs:
1712 sysfs_free(mdi);
1713 close_fd:
1714 if (err)
1715 close(fd);
1716
1717 if (err)
1718 return -1;
1719 else
1720 return fd;
1721}
1722
7801ac20
N
1723int add_disk(int mdfd, struct supertype *st,
1724 struct mdinfo *sra, struct mdinfo *info)
1725{
1726 /* Add a device to an array, in one of 2 ways. */
1727 int rv;
1728#ifndef MDASSEMBLE
1729 if (st->ss->external) {
d23534e4
DW
1730 if (info->disk.state & (1<<MD_DISK_SYNC))
1731 info->recovery_start = MaxSector;
1732 else
1733 info->recovery_start = 0;
2904b26f 1734 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1735 if (! rv) {
1736 struct mdinfo *sd2;
f35f2525
N
1737 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1738 if (sd2 == info)
1739 break;
1740 if (sd2 == NULL) {
503975b9 1741 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1742 *sd2 = *info;
1743 sd2->next = sra->devs;
1744 sra->devs = sd2;
1745 }
7801ac20
N
1746 }
1747 } else
1748#endif
1749 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1750 return rv;
1751}
1752
de6ae750
N
1753int remove_disk(int mdfd, struct supertype *st,
1754 struct mdinfo *sra, struct mdinfo *info)
1755{
1756 int rv;
1757 /* Remove the disk given by 'info' from the array */
1758#ifndef MDASSEMBLE
1759 if (st->ss->external)
1760 rv = sysfs_set_str(sra, info, "slot", "none");
1761 else
1762#endif
1763 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1764 info->disk.minor));
1765 return rv;
1766}
1767
f35f2525
N
1768int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1769{
1770 /* Initialise kernel's knowledge of array.
1771 * This varies between externally managed arrays
1772 * and older kernels
1773 */
1774 int vers = md_get_version(mdfd);
1775 int rv;
1776
1777#ifndef MDASSEMBLE
1778 if (st->ss->external)
1779 rv = sysfs_set_array(info, vers);
1780 else
1781#endif
1782 if ((vers % 100) >= 1) { /* can use different versions */
1783 mdu_array_info_t inf;
1784 memset(&inf, 0, sizeof(inf));
1785 inf.major_version = info->array.major_version;
1786 inf.minor_version = info->array.minor_version;
1787 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1788 } else
1789 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1790 return rv;
1791}
1792
1e5c6983
DW
1793unsigned long long min_recovery_start(struct mdinfo *array)
1794{
1795 /* find the minimum recovery_start in an array for metadata
1796 * formats that only record per-array recovery progress instead
1797 * of per-device
1798 */
1799 unsigned long long recovery_start = MaxSector;
1800 struct mdinfo *d;
1801
1802 for (d = array->devs; d; d = d->next)
1803 recovery_start = min(recovery_start, d->recovery_start);
1804
1805 return recovery_start;
1806}
1807
4dd2df09 1808int mdmon_pid(char *devnm)
a931db9e
NB
1809{
1810 char path[100];
1811 char pid[10];
1812 int fd;
1813 int n;
10013317 1814
4dd2df09 1815 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1816
24f6f99b 1817 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1818
1819 if (fd < 0)
cf556303 1820 return -1;
a931db9e
NB
1821 n = read(fd, pid, 9);
1822 close(fd);
1823 if (n <= 0)
cf556303 1824 return -1;
24f6f99b 1825 return atoi(pid);
a931db9e
NB
1826}
1827
4dd2df09 1828int mdmon_running(char *devnm)
a931db9e 1829{
4dd2df09 1830 int pid = mdmon_pid(devnm);
24f6f99b 1831 if (pid <= 0)
a931db9e 1832 return 0;
24f6f99b 1833 if (kill(pid, 0) == 0)
a931db9e
NB
1834 return 1;
1835 return 0;
1836}
1837
4dd2df09 1838int start_mdmon(char *devnm)
8850ee3e 1839{
1913c325 1840 int i, skipped;
44d2e365 1841 int len;
ca3b6696 1842 pid_t pid;
9fe32043 1843 int status;
44d2e365
N
1844 char pathbuf[1024];
1845 char *paths[4] = {
1846 pathbuf,
85945e19 1847 BINDIR "/mdmon",
2f1bcf43 1848 "./mdmon",
44d2e365
N
1849 NULL
1850 };
8850ee3e 1851
40ebbb9c 1852 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1853 return 0;
1854
9cf014ec 1855 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1856 if (len > 0) {
1857 char *sl;
1858 pathbuf[len] = 0;
1859 sl = strrchr(pathbuf, '/');
1860 if (sl)
1861 sl++;
1862 else
1863 sl = pathbuf;
1864 strcpy(sl, "mdmon");
1865 } else
1866 pathbuf[0] = '\0';
1867
0f7bdf89 1868 /* First try to run systemctl */
9540cc24
N
1869 if (!check_env("MDADM_NO_SYSTEMCTL"))
1870 switch(fork()) {
1871 case 0:
1872 /* FIXME yuk. CLOSE_EXEC?? */
1873 skipped = 0;
1874 for (i = 3; skipped < 20; i++)
1875 if (close(i) < 0)
1876 skipped++;
1877 else
1878 skipped = 0;
1879
1880 /* Don't want to see error messages from
1881 * systemctl. If the service doesn't exist,
1882 * we start mdmon ourselves.
1883 */
1884 close(2);
1885 open("/dev/null", O_WRONLY);
1886 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1887 devnm);
1888 status = execl("/usr/bin/systemctl", "systemctl",
1889 "start",
1890 pathbuf, NULL);
1891 status = execl("/bin/systemctl", "systemctl", "start",
1892 pathbuf, NULL);
1893 exit(1);
7a862a02 1894 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1895 return -1;
1896 default: /* parent - good */
1897 pid = wait(&status);
1898 if (pid >= 0 && status == 0)
1899 return 0;
1900 }
0f7bdf89
JS
1901
1902 /* That failed, try running mdmon directly */
8850ee3e
N
1903 switch(fork()) {
1904 case 0:
1905 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1906 skipped = 0;
ca3b6696 1907 for (i = 3; skipped < 20; i++)
1913c325
N
1908 if (close(i) < 0)
1909 skipped++;
1910 else
1911 skipped = 0;
1912
ca3b6696 1913 for (i = 0; paths[i]; i++)
a0963a86 1914 if (paths[i][0]) {
2f1bcf43 1915 execl(paths[i], paths[i],
4dd2df09 1916 devnm, NULL);
a0963a86 1917 }
8850ee3e 1918 exit(1);
7a862a02 1919 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 1920 return -1;
9fe32043
N
1921 default: /* parent - good */
1922 pid = wait(&status);
15c10423 1923 if (pid < 0 || status != 0) {
7a862a02 1924 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 1925 return -1;
15c10423 1926 }
8850ee3e
N
1927 }
1928 return 0;
1929}
1930
148acb7b
DW
1931__u32 random32(void)
1932{
1933 __u32 rv;
1934 int rfd = open("/dev/urandom", O_RDONLY);
1935 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1936 rv = random();
1937 if (rfd >= 0)
1938 close(rfd);
1939 return rv;
1940}
1941
0e600426 1942#ifndef MDASSEMBLE
edd8d13c
NB
1943int flush_metadata_updates(struct supertype *st)
1944{
1945 int sfd;
1946 if (!st->updates) {
1947 st->update_tail = NULL;
1948 return -1;
1949 }
1950
4dd2df09 1951 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
1952 if (sfd < 0)
1953 return -1;
1954
1955 while (st->updates) {
1956 struct metadata_update *mu = st->updates;
1957 st->updates = mu->next;
1958
1959 send_message(sfd, mu, 0);
1960 wait_reply(sfd, 0);
1961 free(mu->buf);
1962 free(mu);
1963 }
1964 ack(sfd, 0);
1965 wait_reply(sfd, 0);
1966 close(sfd);
1967 st->update_tail = NULL;
1968 return 0;
1969}
1970
1971void append_metadata_update(struct supertype *st, void *buf, int len)
1972{
1973
503975b9 1974 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
1975
1976 mu->buf = buf;
1977 mu->len = len;
1978 mu->space = NULL;
cb23f1f4 1979 mu->space_list = NULL;
edd8d13c
NB
1980 mu->next = NULL;
1981 *st->update_tail = mu;
1982 st->update_tail = &mu->next;
1983}
0e600426 1984#endif /* MDASSEMBLE */
a931db9e 1985
fe6729fa
NB
1986#ifdef __TINYC__
1987/* tinyc doesn't optimize this check in ioctl.h out ... */
1988unsigned int __invalid_size_argument_for_IOC = 0;
1989#endif
1990
6d11ec6f
AK
1991int experimental(void)
1992{
1993 if (check_env("MDADM_EXPERIMENTAL"))
1994 return 1;
1995 else {
7a862a02 1996 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
1997 return 0;
1998 }
1999}
2000
326727d9
AC
2001/* Pick all spares matching given criteria from a container
2002 * if min_size == 0 do not check size
2003 * if domlist == NULL do not check domains
2004 * if spare_group given add it to domains of each spare
2005 * metadata allows to test domains using metadata of destination array */
2006struct mdinfo *container_choose_spares(struct supertype *st,
2007 unsigned long long min_size,
2008 struct domainlist *domlist,
2009 char *spare_group,
2010 const char *metadata, int get_one)
2011{
2012 struct mdinfo *d, **dp, *disks = NULL;
2013
2014 /* get list of all disks in container */
2015 if (st->ss->getinfo_super_disks)
2016 disks = st->ss->getinfo_super_disks(st);
2017
2018 if (!disks)
2019 return disks;
2020 /* find spare devices on the list */
2021 dp = &disks->devs;
2022 disks->array.spare_disks = 0;
2023 while (*dp) {
2024 int found = 0;
2025 d = *dp;
2026 if (d->disk.state == 0) {
2027 /* check if size is acceptable */
2028 unsigned long long dev_size;
2029 dev_t dev = makedev(d->disk.major,d->disk.minor);
2030
2031 if (!min_size ||
2032 (dev_size_from_id(dev, &dev_size) &&
2033 dev_size >= min_size))
2034 found = 1;
2035 /* check if domain matches */
2036 if (found && domlist) {
4dd2df09 2037 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2038 if (spare_group)
2039 pol_add(&pol, pol_domain,
2040 spare_group, NULL);
e5508b36 2041 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2042 found = 0;
2043 dev_policy_free(pol);
2044 }
2045 }
2046 if (found) {
2047 dp = &d->next;
2048 disks->array.spare_disks++;
2049 if (get_one) {
2050 sysfs_free(*dp);
2051 d->next = NULL;
2052 }
2053 } else {
2054 *dp = d->next;
2055 d->next = NULL;
2056 sysfs_free(d);
2057 }
2058 }
2059 return disks;
2060}
9eafa1de
MN
2061
2062/* Checks if paths point to the same device
2063 * Returns 0 if they do.
2064 * Returns 1 if they don't.
2065 * Returns -1 if something went wrong,
2066 * e.g. paths are empty or the files
2067 * they point to don't exist */
2068int compare_paths (char* path1, char* path2)
2069{
2070 struct stat st1,st2;
2071
2072 if (path1 == NULL || path2 == NULL)
2073 return -1;
2074 if (stat(path1,&st1) != 0)
2075 return -1;
2076 if (stat(path2,&st2) != 0)
2077 return -1;
2078 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2079 return 0;
2080 return 1;
2081}
a7dec3fd
N
2082
2083/* Make sure we can open as many devices as needed */
2084void enable_fds(int devices)
2085{
2086 unsigned int fds = 20 + devices;
2087 struct rlimit lim;
2088 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
2089 || lim.rlim_cur >= fds)
2090 return;
2091 if (lim.rlim_max < fds)
2092 lim.rlim_max = fds;
2093 lim.rlim_cur = fds;
2094 setrlimit(RLIMIT_NOFILE, &lim);
2095}
a9c15847
N
2096
2097int in_initrd(void)
2098{
2099 /* This is based on similar function in systemd. */
2100 struct statfs s;
76d0f188
JS
2101 /* statfs.f_type is signed long on s390x and MIPS, causing all
2102 sorts of sign extension problems with RAMFS_MAGIC being
2103 defined as 0x858458f6 */
a9c15847 2104 return statfs("/", &s) >= 0 &&
7ac5d47e 2105 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2106 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2107 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2108}
8832342d
N
2109
2110void reopen_mddev(int mdfd)
2111{
2112 /* Re-open without any O_EXCL, but keep
2113 * the same fd
2114 */
2115 char *devnm;
2116 int fd;
2117 devnm = fd2devnm(mdfd);
2118 close(mdfd);
2119 fd = open_dev(devnm);
2120 if (fd >= 0 && fd != mdfd)
2121 dup2(fd, mdfd);
2122}
7716570e 2123
e80357f8
GJ
2124static struct cmap_hooks *cmap_hooks = NULL;
2125static int is_cmap_hooks_ready = 0;
7716570e 2126
e80357f8
GJ
2127#ifndef MDASSEMBLE
2128void set_cmap_hooks(void)
2129{
2130 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2131 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2132 if (!cmap_hooks->cmap_handle)
2133 return;
7716570e 2134
e80357f8
GJ
2135 cmap_hooks->initialize = dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2136 cmap_hooks->get_string = dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
2137 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2138
e80357f8
GJ
2139 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2140 !cmap_hooks->finalize)
2141 dlclose(cmap_hooks->cmap_handle);
2142 else
2143 is_cmap_hooks_ready = 1;
2144}
7716570e 2145
e80357f8
GJ
2146int get_cluster_name(char **cluster_name)
2147{
2148 int rv = -1;
2149 cmap_handle_t handle;
7716570e 2150
e80357f8
GJ
2151 if (!is_cmap_hooks_ready)
2152 return rv;
7716570e 2153
e80357f8 2154 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2155 if (rv != CS_OK)
2156 goto out;
2157
e80357f8 2158 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2159 if (rv != CS_OK) {
2160 free(*cluster_name);
2161 rv = -1;
2162 goto name_err;
2163 }
2164
2165 rv = 0;
2166name_err:
e80357f8 2167 cmap_hooks->finalize(handle);
7716570e 2168out:
7716570e
GJ
2169 return rv;
2170}
d15a1f72
GJ
2171
2172void set_dlm_hooks(void)
2173{
2174 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2175 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2176 if (!dlm_hooks->dlm_handle)
2177 return;
2178
2179 dlm_hooks->create_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2180 dlm_hooks->release_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
2181 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2182 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2183 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2184 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2185
2186 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2187 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2188 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2189 dlclose(dlm_hooks->dlm_handle);
2190 else
2191 is_dlm_hooks_ready = 1;
2192}
e80357f8
GJ
2193
2194void set_hooks(void)
2195{
2196 set_dlm_hooks();
2197 set_cmap_hooks();
2198}
7d55dca2 2199#endif