Incremental: Remove redundant call for GET_ARRAY_INFO
[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
9cd39f01
JS
214/*
215 * Get array info from the kernel. Longer term we want to deprecate the
216 * ioctl and get it from sysfs.
217 */
218int md_get_array_info(int fd, struct mdu_array_info_s *array)
219{
220 return ioctl(fd, GET_ARRAY_INFO, array);
221}
222
64c4757e
NB
223/*
224 * Parse a 128 bit uuid in 4 integers
225 * format is 32 hexx nibbles with options :.<space> separator
226 * If not exactly 32 hex digits are found, return 0
227 * else return 1
228 */
229int parse_uuid(char *str, int uuid[4])
230{
aba69144
NB
231 int hit = 0; /* number of Hex digIT */
232 int i;
233 char c;
ca3b6696
N
234 for (i = 0; i < 4; i++)
235 uuid[i] = 0;
aba69144 236
ca3b6696 237 while ((c = *str++) != 0) {
aba69144 238 int n;
ca3b6696 239 if (c >= '0' && c <= '9')
aba69144 240 n = c-'0';
ca3b6696 241 else if (c >= 'a' && c <= 'f')
aba69144 242 n = 10 + c - 'a';
ca3b6696 243 else if (c >= 'A' && c <= 'F')
aba69144
NB
244 n = 10 + c - 'A';
245 else if (strchr(":. -", c))
246 continue;
247 else return 0;
248
249 if (hit<32) {
250 uuid[hit/8] <<= 4;
251 uuid[hit/8] += n;
252 }
253 hit++;
82b27616 254 }
aba69144
NB
255 if (hit == 32)
256 return 1;
257 return 0;
64c4757e
NB
258}
259
64c4757e
NB
260/*
261 * Get the md version number.
262 * We use the RAID_VERSION ioctl if it is supported
263 * If not, but we have a block device with major '9', we assume
264 * 0.36.0
265 *
266 * Return version number as 24 but number - assume version parts
267 * always < 255
268 */
269
270int md_get_version(int fd)
271{
5d500228
N
272 struct stat stb;
273 mdu_version_t vers;
274
275 if (fstat(fd, &stb)<0)
276 return -1;
277 if ((S_IFMT&stb.st_mode) != S_IFBLK)
278 return -1;
279
280 if (ioctl(fd, RAID_VERSION, &vers) == 0)
281 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
282 if (errno == EACCES)
283 return -1;
284 if (major(stb.st_rdev) == MD_MAJOR)
285 return (3600);
64c4757e 286 return -1;
64c4757e
NB
287}
288
64c4757e
NB
289int get_linux_version()
290{
291 struct utsname name;
98c6faba 292 char *cp;
f161d047 293 int a = 0, b = 0,c = 0;
64c4757e
NB
294 if (uname(&name) <0)
295 return -1;
296
98c6faba
NB
297 cp = name.release;
298 a = strtoul(cp, &cp, 10);
f161d047
N
299 if (*cp == '.')
300 b = strtoul(cp+1, &cp, 10);
301 if (*cp == '.')
302 c = strtoul(cp+1, &cp, 10);
98c6faba 303
682c7051 304 return (a*1000000)+(b*1000)+c;
64c4757e
NB
305}
306
71204a50 307#ifndef MDASSEMBLE
bc77ed53
DW
308int mdadm_version(char *version)
309{
310 int a, b, c;
311 char *cp;
312
313 if (!version)
314 version = Version;
315
316 cp = strchr(version, '-');
317 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
318 return -1;
319 cp += 3;
320 a = strtoul(cp, &cp, 10);
321 if (*cp != '.')
322 return -1;
323 b = strtoul(cp+1, &cp, 10);
324 if (*cp == '.')
325 c = strtoul(cp+1, &cp, 10);
326 else
327 c = 0;
328 if (*cp != ' ' && *cp != '-')
329 return -1;
330 return (a*1000000)+(b*1000)+c;
331}
332
51d4261c 333unsigned long long parse_size(char *size)
84e11361
N
334{
335 /* parse 'size' which should be a number optionally
336 * followed by 'K', 'M', or 'G'.
337 * Without a suffix, K is assumed.
338 * Number returned is in sectors (half-K)
822e393a 339 * INVALID_SECTORS returned on error.
84e11361
N
340 */
341 char *c;
342 long long s = strtoll(size, &c, 10);
343 if (s > 0) {
344 switch (*c) {
345 case 'K':
346 c++;
347 default:
348 s *= 2;
349 break;
350 case 'M':
351 c++;
352 s *= 1024 * 2;
353 break;
354 case 'G':
355 c++;
356 s *= 1024 * 1024 * 2;
357 break;
72ca9bcf
N
358 case 's': /* sectors */
359 c++;
360 break;
84e11361 361 }
51d4261c 362 } else
822e393a 363 s = INVALID_SECTORS;
84e11361 364 if (*c)
822e393a 365 s = INVALID_SECTORS;
84e11361
N
366 return s;
367}
368
4a06e2c2
N
369int parse_layout_10(char *layout)
370{
371 int copies, rv;
372 char *cp;
373 /* Parse the layout string for raid10 */
374 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
375 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
376 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
377 copies > 200 ||
378 *cp)
379 return -1;
380 if (layout[0] == 'n')
381 rv = 256 + copies;
382 else if (layout[0] == 'o')
383 rv = 0x10000 + (copies<<8) + 1;
384 else
385 rv = 1 + (copies<<8);
386 return rv;
387}
388
389int parse_layout_faulty(char *layout)
390{
391 /* Parse the layout string for 'faulty' */
392 int ln = strcspn(layout, "0123456789");
503975b9 393 char *m = xstrdup(layout);
4a06e2c2
N
394 int mode;
395 m[ln] = 0;
396 mode = map_name(faultylayout, m);
397 if (mode == UnSet)
398 return -1;
399
400 return mode | (atoi(layout+ln)<< ModeShift);
401}
79868890
N
402
403long parse_num(char *num)
404{
405 /* Either return a valid number, or -1 */
406 char *c;
407 long rv = strtol(num, &c, 10);
408 if (rv < 0 || *c || !num[0])
409 return -1;
410 else
411 return rv;
412}
eb3929a4 413#endif
4a06e2c2 414
4de90913
GJ
415int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
416{
417 char *dev;
418 *slot = strtoul(input, &dev, 10);
419 if (dev == input || dev[0] != ':')
420 return -1;
421 *devname = dev+1;
422 return 0;
423}
424
0430ed48
NB
425void remove_partitions(int fd)
426{
427 /* remove partitions from this block devices.
428 * This is used for components added to an array
429 */
430#ifdef BLKPG_DEL_PARTITION
431 struct blkpg_ioctl_arg a;
432 struct blkpg_partition p;
433
434 a.op = BLKPG_DEL_PARTITION;
435 a.data = (void*)&p;
436 a.datalen = sizeof(p);
437 a.flags = 0;
438 memset(a.data, 0, a.datalen);
ca3b6696 439 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
440 ioctl(fd, BLKPG, &a);
441#endif
442}
443
691c6ee1
N
444int test_partition(int fd)
445{
446 /* Check if fd is a whole-disk or a partition.
447 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
448 * will return ENXIO on an invalid partition number.
449 */
450 struct blkpg_ioctl_arg a;
451 struct blkpg_partition p;
452 a.op = BLKPG_DEL_PARTITION;
453 a.data = (void*)&p;
454 a.datalen = sizeof(p);
455 a.flags = 0;
456 memset(a.data, 0, a.datalen);
457 p.pno = 1<<30;
458 if (ioctl(fd, BLKPG, &a) == 0)
459 /* Very unlikely, but not a partition */
460 return 0;
357ac106 461 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
462 /* not a partition */
463 return 0;
464
465 return 1;
466}
467
bfd76b93
CA
468int test_partition_from_id(dev_t id)
469{
470 char buf[20];
471 int fd, rv;
472
473 sprintf(buf, "%d:%d", major(id), minor(id));
474 fd = dev_open(buf, O_RDONLY);
475 if (fd < 0)
476 return -1;
477 rv = test_partition(fd);
478 close(fd);
479 return rv;
480}
691c6ee1 481
de5a472e 482int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 483{
265e0f17 484 int copies, first;
de5a472e
N
485 int i;
486 int avail_disks = 0;
487
488 for (i = 0; i < raid_disks; i++)
489 avail_disks += !!avail[i];
490
64c4757e 491 switch (level) {
265e0f17
NB
492 case 10:
493 /* This is the tricky one - we need to check
494 * which actual disks are present.
495 */
702b557b 496 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 497 first = 0;
265e0f17
NB
498 do {
499 /* there must be one of the 'copies' form 'first' */
500 int n = copies;
ca3b6696 501 int cnt = 0;
b3ec716d 502 int this = first;
265e0f17 503 while (n--) {
b3ec716d 504 if (avail[this])
265e0f17 505 cnt++;
b3ec716d 506 this = (this+1) % raid_disks;
265e0f17
NB
507 }
508 if (cnt == 0)
509 return 0;
b3ec716d 510 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
511 } while (first != 0);
512 return 1;
e5329c37 513
df0d4ea0 514 case LEVEL_MULTIPATH:
e0d19036 515 return avail_disks>= 1;
df0d4ea0 516 case LEVEL_LINEAR:
64c4757e
NB
517 case 0:
518 return avail_disks == raid_disks;
519 case 1:
520 return avail_disks >= 1;
521 case 4:
330d6900
N
522 if (avail_disks == raid_disks - 1 &&
523 !avail[raid_disks - 1])
524 /* If just the parity device is missing, then we
525 * have enough, even if not clean
526 */
527 return 1;
528 /* FALL THROUGH */
64c4757e 529 case 5:
583315d9
NB
530 if (clean)
531 return avail_disks >= raid_disks-1;
532 else
533 return avail_disks >= raid_disks;
98c6faba 534 case 6:
583315d9
NB
535 if (clean)
536 return avail_disks >= raid_disks-2;
537 else
538 return avail_disks >= raid_disks;
64c4757e
NB
539 default:
540 return 0;
541 }
542}
543
8453e704
N
544int enough_fd(int fd)
545{
546 struct mdu_array_info_s array;
547 struct mdu_disk_info_s disk;
aabe020d 548 int i, rv;
8453e704
N
549 char *avail;
550
9cd39f01 551 if (md_get_array_info(fd, &array) != 0 || array.raid_disks <= 0)
8453e704 552 return 0;
503975b9 553 avail = xcalloc(array.raid_disks, 1);
ca3b6696 554 for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704
N
555 disk.number = i;
556 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
557 continue;
9e6d9291
N
558 if (disk.major == 0 && disk.minor == 0)
559 continue;
560 array.nr_disks--;
561
8453e704
N
562 if (! (disk.state & (1<<MD_DISK_SYNC)))
563 continue;
564 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
565 continue;
8453e704
N
566 avail[disk.raid_disk] = 1;
567 }
568 /* This is used on an active array, so assume it is clean */
aabe020d 569 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 570 1, avail);
aabe020d
JS
571 free(avail);
572 return rv;
8453e704
N
573}
574
22e263f6
AC
575const int uuid_zero[4] = { 0, 0, 0, 0 };
576
f277ce36 577int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 578{
f277ce36
NB
579 if (swapuuid) {
580 /* parse uuids are hostendian.
581 * uuid's from some superblocks are big-ending
aba69144 582 * if there is a difference, we need to swap..
f277ce36
NB
583 */
584 unsigned char *ac = (unsigned char *)a;
585 unsigned char *bc = (unsigned char *)b;
586 int i;
ca3b6696 587 for (i = 0; i < 16; i += 4) {
f277ce36
NB
588 if (ac[i+0] != bc[i+3] ||
589 ac[i+1] != bc[i+2] ||
590 ac[i+2] != bc[i+1] ||
591 ac[i+3] != bc[i+0])
592 return 0;
593 }
594 return 1;
595 } else {
596 if (a[0]==b[0] &&
597 a[1]==b[1] &&
598 a[2]==b[2] &&
599 a[3]==b[3])
600 return 1;
601 return 0;
602 }
64c4757e 603}
ca3b6696 604
350f29f9
NB
605void copy_uuid(void *a, int b[4], int swapuuid)
606{
607 if (swapuuid) {
608 /* parse uuids are hostendian.
609 * uuid's from some superblocks are big-ending
610 * if there is a difference, we need to swap..
611 */
612 unsigned char *ac = (unsigned char *)a;
613 unsigned char *bc = (unsigned char *)b;
614 int i;
ca3b6696 615 for (i = 0; i < 16; i += 4) {
350f29f9
NB
616 ac[i+0] = bc[i+3];
617 ac[i+1] = bc[i+2];
618 ac[i+2] = bc[i+1];
619 ac[i+3] = bc[i+0];
620 }
621 } else
622 memcpy(a, b, 16);
623}
64c4757e 624
aae5a112 625char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 626{
9968e376 627 int i, j;
d7288ddc
N
628 char uuid[16];
629 char *c = buf;
630 strcpy(c, "UUID-");
631 c += strlen(c);
aae5a112 632 copy_uuid(uuid, id, swap);
9968e376 633 for (i = 0; i < 4; i++) {
9968e376 634 if (i)
ff54de6e 635 *c++ = sep;
9968e376
DW
636 for (j = 3; j >= 0; j--) {
637 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
638 c+= 2;
639 }
d7288ddc
N
640 }
641 return buf;
aae5a112
DW
642
643}
644
645char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
646{
86983cce
N
647 // dirty hack to work around an issue with super1 superblocks...
648 // super1 superblocks need swapuuid set in order for assembly to
649 // work, but can't have it set if we want this printout to match
650 // all the other uuid printouts in super1.c, so we force swapuuid
651 // to 1 to make our printout match the rest of super1
652 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
653}
654
435d4ebb 655#ifndef MDASSEMBLE
682c7051
NB
656int check_ext2(int fd, char *name)
657{
658 /*
659 * Check for an ext2fs file system.
660 * Superblock is always 1K at 1K offset
661 *
662 * s_magic is le16 at 56 == 0xEF53
663 * report mtime - le32 at 44
664 * blocks - le32 at 4
665 * logblksize - le32 at 24
666 */
667 unsigned char sb[1024];
668 time_t mtime;
2538ba2a
N
669 unsigned long long size;
670 int bsize;
682c7051
NB
671 if (lseek(fd, 1024,0)!= 1024)
672 return 0;
673 if (read(fd, sb, 1024)!= 1024)
674 return 0;
675 if (sb[56] != 0x53 || sb[57] != 0xef)
676 return 0;
677
678 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
679 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
680 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 681 size <<= bsize;
e7b84f9d 682 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 683 name);
2538ba2a 684 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
685 return 1;
686}
687
688int check_reiser(int fd, char *name)
689{
690 /*
691 * superblock is at 64K
692 * size is 1024;
693 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
694 *
695 */
696 unsigned char sb[1024];
2538ba2a 697 unsigned long long size;
682c7051
NB
698 if (lseek(fd, 64*1024, 0) != 64*1024)
699 return 0;
700 if (read(fd, sb, 1024) != 1024)
701 return 0;
ca3b6696
N
702 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
703 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 704 return 0;
e7b84f9d 705 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 706 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 707 cont_err("size = %lluK\n", size*4);
aba69144 708
682c7051
NB
709 return 1;
710}
711
712int check_raid(int fd, char *name)
713{
4b1ac34b 714 struct mdinfo info;
682c7051 715 time_t crtime;
d078d77c 716 char *level;
82d9eba6 717 struct supertype *st = guess_super(fd);
f9ce90ba 718
cb8f6859
N
719 if (!st)
720 return 0;
fa219dd2
MD
721 if (st->ss->add_to_super != NULL) {
722 st->ss->load_super(st, fd, name);
723 /* Looks like a raid array .. */
724 pr_err("%s appears to be part of a raid array:\n", name);
725 st->ss->getinfo_super(st, &info, NULL);
726 st->ss->free_super(st);
727 crtime = info.array.ctime;
728 level = map_num(pers, info.array.level);
729 if (!level)
730 level = "-unknown-";
731 cont_err("level=%s devices=%d ctime=%s",
732 level, info.array.raid_disks, ctime(&crtime));
733 } else {
734 /* Looks like GPT or MBR */
735 pr_err("partition table exists on %s\n", name);
736 }
82d9eba6 737 return 1;
682c7051
NB
738}
739
682c7051
NB
740int ask(char *mesg)
741{
742 char *add = "";
743 int i;
ca3b6696 744 for (i = 0; i < 5; i++) {
682c7051
NB
745 char buf[100];
746 fprintf(stderr, "%s%s", mesg, add);
747 fflush(stderr);
748 if (fgets(buf, 100, stdin)==NULL)
749 return 0;
750 if (buf[0]=='y' || buf[0]=='Y')
751 return 1;
752 if (buf[0]=='n' || buf[0]=='N')
753 return 0;
754 add = "(y/n) ";
755 }
e7b84f9d 756 pr_err("assuming 'no'\n");
682c7051
NB
757 return 0;
758}
435d4ebb 759#endif /* MDASSEMBLE */
682c7051 760
8d80900b 761int is_standard(char *dev, int *nump)
e5329c37
NB
762{
763 /* tests if dev is a "standard" md dev name.
764 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 765 * where NN is a string of digits
598f0d58
NB
766 * Returns 1 if a partitionable standard,
767 * -1 if non-partitonable,
768 * 0 if not a standard name.
e5329c37 769 */
8d80900b 770 char *d = strrchr(dev, '/');
ca3b6696 771 int type = 0;
8d80900b
NB
772 int num;
773 if (!d)
e5329c37 774 return 0;
ca3b6696
N
775 if (strncmp(d, "/d",2) == 0)
776 d += 2, type = 1; /* /dev/md/dN{pM} */
777 else if (strncmp(d, "/md_d", 5) == 0)
778 d += 5, type = 1; /* /dev/md_dN{pM} */
779 else if (strncmp(d, "/md", 3) == 0)
780 d += 3, type = -1; /* /dev/mdN */
781 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
782 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
783 else
784 return 0;
8d80900b 785 if (!*d)
e5329c37 786 return 0;
8d80900b
NB
787 num = atoi(d);
788 while (isdigit(*d))
789 d++;
790 if (*d)
e5329c37 791 return 0;
8d80900b
NB
792 if (nump) *nump = num;
793
794 return type;
e5329c37
NB
795}
796
4b1ac34b 797unsigned long calc_csum(void *super, int bytes)
82b27616 798{
56eb10c0 799 unsigned long long newcsum = 0;
82b27616 800 int i;
4b1ac34b
NB
801 unsigned int csum;
802 unsigned int *superc = (unsigned int*) super;
82b27616 803
ca3b6696
N
804 for(i = 0; i < bytes/4; i++)
805 newcsum += superc[i];
82b27616 806 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 807#ifdef __alpha__
aba69144 808/* The in-kernel checksum calculation is always 16bit on
570c0542 809 * the alpha, though it is 32 bit on i386...
ca3b6696 810 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
811 * a way that it shouldn't).
812 */
813 csum = (csum & 0xffff) + (csum >> 16);
814 csum = (csum & 0xffff) + (csum >> 16);
815#endif
82b27616
NB
816 return csum;
817}
cd29a5c8 818
435d4ebb 819#ifndef MDASSEMBLE
56eb10c0 820char *human_size(long long bytes)
cd29a5c8 821{
ff9239ee 822 static char buf[47];
d5d3721e
NB
823
824 /* We convert bytes to either centi-M{ega,ibi}bytes or
825 * centi-G{igi,ibi}bytes, with appropriate rounding,
826 * and then print 1/100th of those as a decimal.
827 * We allow upto 2048Megabytes before converting to
828 * gigabytes, as that shows more precision and isn't
829 * too large a number.
01619b48 830 * Terabytes are not yet handled.
d5d3721e 831 */
cd29a5c8 832
56eb10c0 833 if (bytes < 5000*1024)
ca3b6696 834 buf[0] = 0;
d5d3721e 835 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 836 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 837 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 838 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
efa29530 839 cMiB/100, cMiB % 100, cMB/100, cMB % 100);
d5d3721e 840 } else {
93d3bd3b 841 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 842 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 843 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
efa29530 844 cGiB/100, cGiB % 100, cGB/100, cGB % 100);
d5d3721e 845 }
cd29a5c8
NB
846 return buf;
847}
e0d19036 848
f0ec6710 849char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
850{
851 static char buf[30];
e0d19036 852
570abc6f
MN
853 /* We convert bytes to either centi-M{ega,ibi}bytes or
854 * centi-G{igi,ibi}bytes, with appropriate rounding,
855 * and then print 1/100th of those as a decimal.
856 * We allow upto 2048Megabytes before converting to
857 * gigabytes, as that shows more precision and isn't
858 * too large a number.
859 * Terabytes are not yet handled.
f0ec6710
MN
860 *
861 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
862 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
863 */
864
e0d19036 865 if (bytes < 5000*1024)
570abc6f 866 buf[0] = 0;
f0ec6710
MN
867 else if (prefix == IEC) {
868 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 869 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710 870 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
efa29530 871 cMiB/100, cMiB % 100);
f0ec6710 872 } else {
93d3bd3b 873 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710 874 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
efa29530 875 cGiB/100, cGiB % 100);
f0ec6710
MN
876 }
877 }
878 else if (prefix == JEDEC) {
879 if (bytes < 2*1024LL*1024LL*1024LL) {
880 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
881 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
efa29530 882 cMB/100, cMB % 100);
f0ec6710
MN
883 } else {
884 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
885 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
efa29530 886 cGB/100, cGB % 100);
f0ec6710 887 }
570abc6f 888 }
f0ec6710
MN
889 else
890 buf[0] = 0;
570abc6f 891
e0d19036
NB
892 return buf;
893}
e4965ef8
N
894
895void print_r10_layout(int layout)
896{
897 int near = layout & 255;
898 int far = (layout >> 8) & 255;
899 int offset = (layout&0x10000);
900 char *sep = "";
901
902 if (near != 1) {
903 printf("%s near=%d", sep, near);
904 sep = ",";
905 }
906 if (far != 1)
907 printf("%s %s=%d", sep, offset?"offset":"far", far);
908 if (near*far == 1)
909 printf("NO REDUNDANCY");
910}
435d4ebb 911#endif
e0d19036 912
5f8097be
NB
913unsigned long long calc_array_size(int level, int raid_disks, int layout,
914 int chunksize, unsigned long long devsize)
577e8448 915{
e03640bd
CA
916 if (level == 1)
917 return devsize;
577e8448
AK
918 devsize &= ~(unsigned long long)((chunksize>>9)-1);
919 return get_data_disks(level, layout, raid_disks) * devsize;
920}
921
922int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
923{
924 int data_disks = 0;
925 switch (level) {
ca3b6696
N
926 case 0: data_disks = raid_disks;
927 break;
928 case 1: data_disks = 1;
929 break;
5f8097be 930 case 4:
ca3b6696
N
931 case 5: data_disks = raid_disks - 1;
932 break;
933 case 6: data_disks = raid_disks - 2;
934 break;
5f8097be
NB
935 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
936 break;
937 }
577e8448
AK
938
939 return data_disks;
5f8097be
NB
940}
941
13db17bd 942dev_t devnm2devid(char *devnm)
4dd2df09
N
943{
944 /* First look in /sys/block/$DEVNM/dev for %d:%d
945 * If that fails, try parsing out a number
946 */
947 char path[100];
948 char *ep;
949 int fd;
950 int mjr,mnr;
951
952 sprintf(path, "/sys/block/%s/dev", devnm);
953 fd = open(path, O_RDONLY);
954 if (fd >= 0) {
955 char buf[20];
956 int n = read(fd, buf, sizeof(buf));
957 close(fd);
958 if (n > 0)
959 buf[n] = 0;
960 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
961 return makedev(mjr, mnr);
962 }
963 if (strncmp(devnm, "md_d", 4) == 0 &&
964 isdigit(devnm[4]) &&
965 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
966 ep > devnm && *ep == 0)
967 return makedev(get_mdp_major(), mnr << MdpMinorShift);
968
969 if (strncmp(devnm, "md", 2) == 0 &&
970 isdigit(devnm[2]) &&
971 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
972 ep > devnm && *ep == 0)
973 return makedev(MD_MAJOR, mnr);
974
975 return 0;
976}
977
5a23a06e 978#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 979char *get_md_name(char *devnm)
e0d19036
NB
980{
981 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 982 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 983
e0d19036
NB
984 static char devname[50];
985 struct stat stb;
4dd2df09 986 dev_t rdev = devnm2devid(devnm);
dd0781e5 987 char *dn;
98c6faba 988
4dd2df09
N
989 if (rdev == 0)
990 return 0;
991 if (strncmp(devnm, "md_", 3) == 0) {
992 snprintf(devname, sizeof(devname), "/dev/md/%s",
993 devnm + 3);
98c6faba
NB
994 if (stat(devname, &stb) == 0
995 && (S_IFMT&stb.st_mode) == S_IFBLK
996 && (stb.st_rdev == rdev))
997 return devname;
998 }
4dd2df09
N
999 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
1000 if (stat(devname, &stb) == 0
1001 && (S_IFMT&stb.st_mode) == S_IFBLK
1002 && (stb.st_rdev == rdev))
1003 return devname;
1004
1005 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
1006 if (stat(devname, &stb) == 0
1007 && (S_IFMT&stb.st_mode) == S_IFBLK
1008 && (stb.st_rdev == rdev))
1009 return devname;
1010
16c6fa80 1011 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
1012 if (dn)
1013 return dn;
4dd2df09 1014 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 1015 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
1016 if (errno != EEXIST)
1017 return NULL;
e0d19036
NB
1018
1019 if (stat(devname, &stb) == 0
1020 && (S_IFMT&stb.st_mode) == S_IFBLK
1021 && (stb.st_rdev == rdev))
1022 return devname;
1023 unlink(devname);
1024 return NULL;
1025}
1026
1027void put_md_name(char *name)
1028{
ca3b6696 1029 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
1030 unlink(name);
1031}
435d4ebb 1032#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 1033
cc742d38
N
1034int get_maj_min(char *dev, int *major, int *minor)
1035{
1036 char *e;
1037 *major = strtoul(dev, &e, 0);
1038 return (e > dev && *e == ':' && e[1] &&
1039 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1040 *e == 0);
1041}
1042
8b0dabea
NB
1043int dev_open(char *dev, int flags)
1044{
1045 /* like 'open', but if 'dev' matches %d:%d, create a temp
1046 * block device and open that
1047 */
8b0dabea
NB
1048 int fd = -1;
1049 char devname[32];
e81cdd9f 1050 int major;
8b0dabea 1051 int minor;
e81cdd9f 1052
9f0ad56b
JS
1053 if (!dev)
1054 return -1;
6df6a774 1055 flags |= O_DIRECT;
e81cdd9f 1056
cc742d38 1057 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1058 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1059 (int)getpid(), major, minor);
1060 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1061 fd = open(devname, flags);
1062 unlink(devname);
6df6a774
N
1063 }
1064 if (fd < 0) {
9dc7d357 1065 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
1066 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
1067 (int)getpid(), major, minor);
ca3b6696 1068 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
1069 fd = open(devname, flags);
1070 unlink(devname);
1071 }
8b0dabea
NB
1072 }
1073 } else
6df6a774 1074 fd = open(dev, flags);
8b0dabea
NB
1075 return fd;
1076}
f9ce90ba 1077
4dd2df09 1078int open_dev_flags(char *devnm, int flags)
e8a70c89 1079{
13db17bd 1080 dev_t devid;
e8a70c89
N
1081 char buf[20];
1082
4dd2df09
N
1083 devid = devnm2devid(devnm);
1084 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1085 return dev_open(buf, flags);
1086}
1087
4dd2df09 1088int open_dev(char *devnm)
d998b738 1089{
4dd2df09 1090 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1091}
1092
4dd2df09 1093int open_dev_excl(char *devnm)
a931db9e
NB
1094{
1095 char buf[20];
1096 int i;
7187750e 1097 int flags = O_RDWR;
13db17bd 1098 dev_t devid = devnm2devid(devnm);
399e0b97 1099 long delay = 1000;
a931db9e 1100
4dd2df09 1101 sprintf(buf, "%d:%d", major(devid), minor(devid));
efa29530 1102 for (i = 0; i < 25; i++) {
7187750e 1103 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1104 if (fd >= 0)
1105 return fd;
7187750e
N
1106 if (errno == EACCES && flags == O_RDWR) {
1107 flags = O_RDONLY;
1108 continue;
1109 }
a931db9e
NB
1110 if (errno != EBUSY)
1111 return fd;
399e0b97
N
1112 usleep(delay);
1113 if (delay < 200000)
1114 delay *= 2;
a931db9e
NB
1115 }
1116 return -1;
1117}
1118
9008ed1c
N
1119int same_dev(char *one, char *two)
1120{
1121 struct stat st1, st2;
1122 if (stat(one, &st1) != 0)
1123 return 0;
1124 if (stat(two, &st2) != 0)
1125 return 0;
1126 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1127 return 0;
1128 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1129 return 0;
1130 return st1.st_rdev == st2.st_rdev;
1131}
1132
a7c6e3fb 1133void wait_for(char *dev, int fd)
a714580e
N
1134{
1135 int i;
a7c6e3fb 1136 struct stat stb_want;
399e0b97 1137 long delay = 1000;
a7c6e3fb
N
1138
1139 if (fstat(fd, &stb_want) != 0 ||
1140 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1141 return;
a714580e 1142
efa29530 1143 for (i = 0; i < 25; i++) {
a714580e 1144 struct stat stb;
a7c6e3fb
N
1145 if (stat(dev, &stb) == 0 &&
1146 (stb.st_mode & S_IFMT) == S_IFBLK &&
1147 (stb.st_rdev == stb_want.st_rdev))
a714580e 1148 return;
399e0b97
N
1149 usleep(delay);
1150 if (delay < 200000)
1151 delay *= 2;
a714580e 1152 }
436305c6 1153 if (i == 25)
1ade5cc1 1154 dprintf("timeout waiting for %s\n", dev);
a714580e
N
1155}
1156
0f22b998
N
1157struct superswitch *superlist[] =
1158{
1159 &super0, &super1,
1160 &super_ddf, &super_imsm,
0592faeb 1161 &mbr, &gpt,
0f22b998 1162 NULL };
f9ce90ba 1163
ea24acd0 1164#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1165
4725bc31 1166struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1167{
1686dc25
NB
1168 mdu_array_info_t array;
1169 int vers;
1170 int minor;
1171 struct supertype *st = NULL;
7e0f6979 1172 struct mdinfo *sra;
142cb9e1 1173 char *verstr;
1686dc25
NB
1174 char version[20];
1175 int i;
f7e7067b 1176 char *subarray = NULL;
4dd2df09 1177 char container[32] = "";
1686dc25 1178
4dd2df09 1179 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1180
1181 if (sra) {
7e0f6979
NB
1182 vers = sra->array.major_version;
1183 minor = sra->array.minor_version;
142cb9e1 1184 verstr = sra->text_version;
1686dc25 1185 } else {
9cd39f01 1186 if (md_get_array_info(fd, &array))
1686dc25
NB
1187 array.major_version = array.minor_version = 0;
1188 vers = array.major_version;
1189 minor = array.minor_version;
142cb9e1 1190 verstr = "";
6fbba4c9 1191 }
82d9eba6 1192
1686dc25
NB
1193 if (vers != -1) {
1194 sprintf(version, "%d.%d", vers, minor);
1195 verstr = version;
6fbba4c9 1196 }
3c558363 1197 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1198 char *dev = verstr+1;
5f7e44b2 1199
f7e7067b 1200 subarray = strchr(dev, '/');
92d49ecf 1201 if (subarray) {
f7e7067b 1202 *subarray++ = '\0';
503975b9 1203 subarray = xstrdup(subarray);
92d49ecf 1204 }
4dd2df09 1205 strcpy(container, dev);
15d230f7 1206 sysfs_free(sra);
5f7e44b2 1207 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1208 if (sra && sra->text_version[0])
1209 verstr = sra->text_version;
1210 else
1211 verstr = "-no-metadata-";
f7e7067b
NB
1212 }
1213
efa29530 1214 for (i = 0; st == NULL && superlist[i]; i++)
f7e7067b 1215 st = superlist[i]->match_metadata_desc(verstr);
1686dc25 1216
15d230f7 1217 sysfs_free(sra);
f7e7067b 1218 if (st) {
3b0896f8 1219 st->sb = NULL;
1f49fb3a
N
1220 if (subarrayp)
1221 *subarrayp = subarray;
4dd2df09
N
1222 strcpy(st->container_devnm, container);
1223 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1224 } else
1225 free(subarray);
5f7e44b2 1226
82d9eba6 1227 return st;
f9ce90ba 1228}
ea24acd0
NB
1229#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1230
0f0749ad 1231int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1232{
1233 char buf[20];
1234 int fd;
1235
1236 sprintf(buf, "%d:%d", major(id), minor(id));
1237 fd = dev_open(buf, O_RDONLY);
1238 if (fd < 0)
1239 return 0;
1240 if (get_dev_size(fd, NULL, size)) {
1241 close(fd);
1242 return 1;
1243 }
1244 close(fd);
1245 return 0;
1246}
f9ce90ba 1247
159c3a1a 1248struct supertype *dup_super(struct supertype *orig)
3da92f27 1249{
159c3a1a 1250 struct supertype *st;
1686dc25 1251
d2ca6449
NB
1252 if (!orig)
1253 return orig;
503975b9 1254 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1255 st->ss = orig->ss;
1256 st->max_devs = orig->max_devs;
1257 st->minor_version = orig->minor_version;
cb8f6859 1258 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1259 st->data_offset = orig->data_offset;
159c3a1a
NB
1260 st->sb = NULL;
1261 st->info = NULL;
1262 return st;
3da92f27
NB
1263}
1264
54887ad8 1265struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1266{
1267 /* try each load_super to find the best match,
1268 * and return the best superswitch
1269 */
82d9eba6
NB
1270 struct superswitch *ss;
1271 struct supertype *st;
1d13b599 1272 unsigned int besttime = 0;
570c0542 1273 int bestsuper = -1;
f9ce90ba
NB
1274 int i;
1275
503975b9 1276 st = xcalloc(1, sizeof(*st));
4dd2df09 1277 st->container_devnm[0] = 0;
d1d599ea 1278
efa29530 1279 for (i = 0; superlist[i]; i++) {
f9ce90ba
NB
1280 int rv;
1281 ss = superlist[i];
54887ad8
N
1282 if (guess_type == guess_array && ss->add_to_super == NULL)
1283 continue;
1284 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1285 continue;
ef609477 1286 memset(st, 0, sizeof(*st));
df3346e6 1287 st->ignore_hw_compat = 1;
3da92f27 1288 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1289 if (rv == 0) {
1290 struct mdinfo info;
a5d85af7 1291 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1292 if (bestsuper == -1 ||
1293 besttime < info.array.ctime) {
1294 bestsuper = i;
1295 besttime = info.array.ctime;
570c0542 1296 }
3da92f27 1297 ss->free_super(st);
570c0542
NB
1298 }
1299 }
1300 if (bestsuper != -1) {
1301 int rv;
ef609477 1302 memset(st, 0, sizeof(*st));
df3346e6 1303 st->ignore_hw_compat = 1;
3da92f27 1304 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1305 if (rv == 0) {
5e747af2 1306 superlist[bestsuper]->free_super(st);
82d9eba6 1307 return st;
f9ce90ba
NB
1308 }
1309 }
570c0542 1310 free(st);
f9ce90ba
NB
1311 return NULL;
1312}
fe6729fa 1313
beae1dfe
NB
1314/* Return size of device in bytes */
1315int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1316{
1317 unsigned long long ldsize;
c2c9bb6f
NB
1318 struct stat st;
1319
1320 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1321 ldsize = (unsigned long long)st.st_size;
1322 else
beae1dfe
NB
1323#ifdef BLKGETSIZE64
1324 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1325#endif
1326 {
1327 unsigned long dsize;
1328 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1329 ldsize = dsize;
1330 ldsize <<= 9;
1331 } else {
1332 if (dname)
32971509 1333 pr_err("Cannot get size of %s: %s\n",
beae1dfe
NB
1334 dname, strerror(errno));
1335 return 0;
1336 }
1337 }
1338 *sizep = ldsize;
1339 return 1;
1340}
8fac0577 1341
32971509
PB
1342/* Return sector size of device in bytes */
1343int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep)
1344{
1345 unsigned int sectsize;
1346
1347 if (ioctl(fd, BLKSSZGET, &sectsize) != 0) {
1348 if (dname)
1349 pr_err("Cannot get sector size of %s: %s\n",
1350 dname, strerror(errno));
1351 return 0;
1352 }
1353
1354 *sectsizep = sectsize;
1355 return 1;
1356}
1357
3a371610
N
1358/* Return true if this can only be a container, not a member device.
1359 * i.e. is and md device and size is zero
1360 */
1361int must_be_container(int fd)
1362{
1363 unsigned long long size;
1364 if (md_get_version(fd) < 0)
1365 return 0;
1366 if (get_dev_size(fd, NULL, &size) == 0)
1367 return 1;
1368 if (size == 0)
1369 return 1;
1370 return 0;
1371}
034b203a
TM
1372
1373/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1374 * Returns: 1 if successful
1375 * -1 for unknown partition type
1376 * 0 for other errors
1377 */
1378static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1379{
056b331e 1380 struct GPT gpt;
034b203a 1381 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1382 struct GPT_part_entry *part;
1383 char buf[512];
034b203a
TM
1384 unsigned long long curr_part_end;
1385 unsigned all_partitions, entry_size;
f21e18ca 1386 unsigned part_nr;
41b06495 1387 unsigned int sector_size = 0;
034b203a
TM
1388
1389 *endofpart = 0;
1390
056b331e 1391 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1392 /* skip protective MBR */
41b06495
MD
1393 if (!get_dev_sector_size(fd, NULL, &sector_size))
1394 return 0;
1395 lseek(fd, sector_size, SEEK_SET);
73e658d8 1396 /* read GPT header */
056b331e 1397 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1398 return 0;
1399
1400 /* get the number of partition entries and the entry size */
056b331e
N
1401 all_partitions = __le32_to_cpu(gpt.part_cnt);
1402 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1403
1404 /* Check GPT signature*/
056b331e 1405 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1406 return -1;
1407
1408 /* sanity checks */
1409 if (all_partitions > 1024 ||
db7fdfe4 1410 entry_size > sizeof(buf))
034b203a
TM
1411 return -1;
1412
db7fdfe4
JS
1413 part = (struct GPT_part_entry *)buf;
1414
41b06495
MD
1415 /* set offset to third block (GPT entries) */
1416 lseek(fd, sector_size*2, SEEK_SET);
ca3b6696 1417 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1418 /* read partition entry */
db7fdfe4 1419 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1420 return 0;
1421
034b203a 1422 /* is this valid partition? */
db7fdfe4 1423 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1424 /* check the last lba for the current partition */
db7fdfe4 1425 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1426 if (curr_part_end > *endofpart)
1427 *endofpart = curr_part_end;
1428 }
1429
034b203a
TM
1430 }
1431 return 1;
1432}
1433
1434/* Sets endofpart parameter to the last block used by the last partition on the device.
1435 * Returns: 1 if successful
1436 * -1 for unknown partition type
1437 * 0 for other errors
1438 */
1439static int get_last_partition_end(int fd, unsigned long long *endofpart)
1440{
056b331e 1441 struct MBR boot_sect;
034b203a 1442 unsigned long long curr_part_end;
f21e18ca 1443 unsigned part_nr;
31208db9 1444 unsigned int sector_size;
034b203a
TM
1445 int retval = 0;
1446
1447 *endofpart = 0;
1448
056b331e 1449 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1450 /* read MBR */
1451 lseek(fd, 0, 0);
056b331e 1452 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1453 goto abort;
1454
1455 /* check MBP signature */
056b331e 1456 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1457 retval = 1;
1458 /* found the correct signature */
034b203a 1459
ca3b6696 1460 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
8e2bca51
JC
1461 /*
1462 * Have to make every access through boot_sect rather
1463 * than using a pointer to the partition table (or an
1464 * entry), since the entries are not properly aligned.
1465 */
1466
034b203a 1467 /* check for GPT type */
8e2bca51
JC
1468 if (boot_sect.parts[part_nr].part_type ==
1469 MBR_GPT_PARTITION_TYPE) {
034b203a
TM
1470 retval = get_gpt_last_partition_end(fd, endofpart);
1471 break;
1472 }
1473 /* check the last used lba for the current partition */
8e2bca51
JC
1474 curr_part_end =
1475 __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) +
1476 __le32_to_cpu(boot_sect.parts[part_nr].blocks_num);
034b203a
TM
1477 if (curr_part_end > *endofpart)
1478 *endofpart = curr_part_end;
034b203a
TM
1479 }
1480 } else {
1481 /* Unknown partition table */
1482 retval = -1;
1483 }
31208db9
MD
1484 /* calculate number of 512-byte blocks */
1485 if (get_dev_sector_size(fd, NULL, &sector_size))
1486 *endofpart *= (sector_size / 512);
034b203a
TM
1487 abort:
1488 return retval;
1489}
1490
53ed6ac3
KW
1491int check_partitions(int fd, char *dname, unsigned long long freesize,
1492 unsigned long long size)
034b203a
TM
1493{
1494 /*
1495 * Check where the last partition ends
1496 */
1497 unsigned long long endofpart;
034b203a 1498
31208db9 1499 if (get_last_partition_end(fd, &endofpart) > 0) {
034b203a
TM
1500 /* There appears to be a partition table here */
1501 if (freesize == 0) {
1502 /* partitions will not be visible in new device */
e7b84f9d
N
1503 pr_err("partition table exists on %s but will be lost or\n"
1504 " meaningless after creating array\n",
1505 dname);
034b203a
TM
1506 return 1;
1507 } else if (endofpart > freesize) {
1508 /* last partition overlaps metadata */
e7b84f9d
N
1509 pr_err("metadata will over-write last partition on %s.\n",
1510 dname);
034b203a 1511 return 1;
53ed6ac3
KW
1512 } else if (size && endofpart > size) {
1513 /* partitions will be truncated in new device */
e7b84f9d
N
1514 pr_err("array size is too small to cover all partitions on %s.\n",
1515 dname);
53ed6ac3 1516 return 1;
034b203a
TM
1517 }
1518 }
1519 return 0;
1520}
1521
a322f70c
DW
1522int open_container(int fd)
1523{
1524 /* 'fd' is a block device. Find out if it is in use
1525 * by a container, and return an open fd on that container.
1526 */
1527 char path[256];
1528 char *e;
1529 DIR *dir;
1530 struct dirent *de;
1531 int dfd, n;
1532 char buf[200];
1533 int major, minor;
1534 struct stat st;
1535
1536 if (fstat(fd, &st) != 0)
1537 return -1;
1538 sprintf(path, "/sys/dev/block/%d:%d/holders",
1539 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1540 e = path + strlen(path);
1541
1542 dir = opendir(path);
1543 if (!dir)
1544 return -1;
1545 while ((de = readdir(dir))) {
1546 if (de->d_ino == 0)
1547 continue;
1548 if (de->d_name[0] == '.')
1549 continue;
a994592d
N
1550 /* Need to make sure it is a container and not a volume */
1551 sprintf(e, "/%s/md/metadata_version", de->d_name);
1552 dfd = open(path, O_RDONLY);
1553 if (dfd < 0)
1554 continue;
1555 n = read(dfd, buf, sizeof(buf));
1556 close(dfd);
1557 if (n <= 0 || (unsigned)n >= sizeof(buf))
1558 continue;
1559 buf[n] = 0;
1560 if (strncmp(buf, "external", 8) != 0 ||
1561 n < 10 ||
1562 buf[9] == '/')
1563 continue;
a322f70c
DW
1564 sprintf(e, "/%s/dev", de->d_name);
1565 dfd = open(path, O_RDONLY);
1566 if (dfd < 0)
1567 continue;
1568 n = read(dfd, buf, sizeof(buf));
1569 close(dfd);
f21e18ca 1570 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1571 continue;
1572 buf[n] = 0;
1573 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1574 continue;
1575 sprintf(buf, "%d:%d", major, minor);
1576 dfd = dev_open(buf, O_RDONLY);
1577 if (dfd >= 0) {
1578 closedir(dir);
1579 return dfd;
1580 }
1581 }
355726fa 1582 closedir(dir);
a322f70c
DW
1583 return -1;
1584}
1585
33414a01
DW
1586struct superswitch *version_to_superswitch(char *vers)
1587{
1588 int i;
1589
1590 for (i = 0; superlist[i]; i++) {
1591 struct superswitch *ss = superlist[i];
1592
1593 if (strcmp(vers, ss->name) == 0)
1594 return ss;
1595 }
1596
1597 return NULL;
1598}
1599
4dd2df09
N
1600int metadata_container_matches(char *metadata, char *devnm)
1601{
1602 /* Check if 'devnm' is the container named in 'metadata'
1603 * which is
1604 * /containername/componentname or
1605 * -containername/componentname
1606 */
1607 int l;
1608 if (*metadata != '/' && *metadata != '-')
1609 return 0;
1610 l = strlen(devnm);
1611 if (strncmp(metadata+1, devnm, l) != 0)
1612 return 0;
1613 if (metadata[l+1] != '/')
1614 return 0;
1615 return 1;
1616}
1617
1618int metadata_subdev_matches(char *metadata, char *devnm)
1619{
1620 /* Check if 'devnm' is the subdev named in 'metadata'
1621 * which is
1622 * /containername/subdev or
1623 * -containername/subdev
1624 */
1625 char *sl;
1626 if (*metadata != '/' && *metadata != '-')
1627 return 0;
1628 sl = strchr(metadata+1, '/');
1629 if (!sl)
1630 return 0;
1631 if (strcmp(sl+1, devnm) == 0)
1632 return 1;
1633 return 0;
1634}
1635
33414a01
DW
1636int is_container_member(struct mdstat_ent *mdstat, char *container)
1637{
1638 if (mdstat->metadata_version == NULL ||
1639 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1640 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1641 return 0;
1642
1643 return 1;
1644}
1645
1646int is_subarray_active(char *subarray, char *container)
1647{
1648 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1649 struct mdstat_ent *ent;
1650
e5408a32
DW
1651 for (ent = mdstat; ent; ent = ent->next)
1652 if (is_container_member(ent, container))
e5e5d7ce 1653 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1654 break;
33414a01
DW
1655
1656 free_mdstat(mdstat);
1657
1658 return ent != NULL;
1659}
1660
1661/* open_subarray - opens a subarray in a container
1662 * @dev: container device name
feab51f8 1663 * @st: empty supertype
33414a01
DW
1664 * @quiet: block reporting errors flag
1665 *
1666 * On success returns an fd to a container and fills in *st
1667 */
feab51f8 1668int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1669{
1670 struct mdinfo *mdi;
a951a4f7 1671 struct mdinfo *info;
33414a01 1672 int fd, err = 1;
4dd2df09 1673 char *_devnm;
33414a01
DW
1674
1675 fd = open(dev, O_RDWR|O_EXCL);
1676 if (fd < 0) {
1677 if (!quiet)
e7b84f9d 1678 pr_err("Couldn't open %s, aborting\n",
33414a01 1679 dev);
b990032d 1680 return -1;
33414a01
DW
1681 }
1682
4dd2df09
N
1683 _devnm = fd2devnm(fd);
1684 if (_devnm == NULL) {
33414a01 1685 if (!quiet)
e7b84f9d
N
1686 pr_err("Failed to determine device number for %s\n",
1687 dev);
33414a01
DW
1688 goto close_fd;
1689 }
4dd2df09 1690 strcpy(st->devnm, _devnm);
33414a01 1691
4dd2df09 1692 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1693 if (!mdi) {
1694 if (!quiet)
e7b84f9d 1695 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1696 dev);
1697 goto close_fd;
1698 }
1699
1700 if (mdi->array.level != UnSet) {
1701 if (!quiet)
e7b84f9d 1702 pr_err("%s is not a container\n", dev);
33414a01
DW
1703 goto free_sysfs;
1704 }
1705
1706 st->ss = version_to_superswitch(mdi->text_version);
1707 if (!st->ss) {
1708 if (!quiet)
e7b84f9d
N
1709 pr_err("Operation not supported for %s metadata\n",
1710 mdi->text_version);
33414a01
DW
1711 goto free_sysfs;
1712 }
1713
4dd2df09 1714 if (st->devnm[0] == 0) {
33414a01 1715 if (!quiet)
e7b84f9d 1716 pr_err("Failed to allocate device name\n");
33414a01
DW
1717 goto free_sysfs;
1718 }
1719
db20d413 1720 if (!st->ss->load_container) {
33414a01 1721 if (!quiet)
e7b84f9d 1722 pr_err("%s is not a container\n", dev);
4dd2df09 1723 goto free_sysfs;
33414a01
DW
1724 }
1725
db20d413 1726 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1727 if (!quiet)
e7b84f9d 1728 pr_err("Failed to load metadata for %s\n",
db20d413 1729 dev);
4dd2df09 1730 goto free_sysfs;
33414a01
DW
1731 }
1732
a951a4f7
N
1733 info = st->ss->container_content(st, subarray);
1734 if (!info) {
1735 if (!quiet)
e7b84f9d 1736 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1737 subarray, dev);
1738 goto free_super;
1739 }
1740 free(info);
1741
33414a01
DW
1742 err = 0;
1743
1744 free_super:
1745 if (err)
1746 st->ss->free_super(st);
33414a01
DW
1747 free_sysfs:
1748 sysfs_free(mdi);
1749 close_fd:
1750 if (err)
1751 close(fd);
1752
1753 if (err)
1754 return -1;
1755 else
1756 return fd;
1757}
1758
7801ac20
N
1759int add_disk(int mdfd, struct supertype *st,
1760 struct mdinfo *sra, struct mdinfo *info)
1761{
1762 /* Add a device to an array, in one of 2 ways. */
1763 int rv;
1764#ifndef MDASSEMBLE
1765 if (st->ss->external) {
d23534e4
DW
1766 if (info->disk.state & (1<<MD_DISK_SYNC))
1767 info->recovery_start = MaxSector;
1768 else
1769 info->recovery_start = 0;
2904b26f 1770 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1771 if (! rv) {
1772 struct mdinfo *sd2;
f35f2525
N
1773 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1774 if (sd2 == info)
1775 break;
1776 if (sd2 == NULL) {
503975b9 1777 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1778 *sd2 = *info;
1779 sd2->next = sra->devs;
1780 sra->devs = sd2;
1781 }
7801ac20
N
1782 }
1783 } else
1784#endif
1785 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1786 return rv;
1787}
1788
de6ae750
N
1789int remove_disk(int mdfd, struct supertype *st,
1790 struct mdinfo *sra, struct mdinfo *info)
1791{
1792 int rv;
1793 /* Remove the disk given by 'info' from the array */
1794#ifndef MDASSEMBLE
1795 if (st->ss->external)
1796 rv = sysfs_set_str(sra, info, "slot", "none");
1797 else
1798#endif
1799 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1800 info->disk.minor));
1801 return rv;
1802}
1803
1ab9ed2a 1804int hot_remove_disk(int mdfd, unsigned long dev, int force)
2dd271fe 1805{
1ab9ed2a 1806 int cnt = force ? 500 : 5;
2dd271fe
N
1807 int ret;
1808
1809 /* HOT_REMOVE_DISK can fail with EBUSY if there are
1810 * outstanding IO requests to the device.
1811 * In this case, it can be helpful to wait a little while,
1ab9ed2a 1812 * up to 5 seconds if 'force' is set, or 50 msec if not.
2dd271fe
N
1813 */
1814 while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 &&
1815 errno == EBUSY &&
1816 cnt-- > 0)
1817 usleep(10000);
1818
1819 return ret;
1820}
1821
1ab9ed2a 1822int sys_hot_remove_disk(int statefd, int force)
fdd01569 1823{
1ab9ed2a 1824 int cnt = force ? 500 : 5;
fdd01569
N
1825 int ret;
1826
1827 while ((ret = write(statefd, "remove", 6)) == -1 &&
1828 errno == EBUSY &&
1829 cnt-- > 0)
1830 usleep(10000);
1831 return ret == 6 ? 0 : -1;
1832}
1833
f35f2525
N
1834int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1835{
1836 /* Initialise kernel's knowledge of array.
1837 * This varies between externally managed arrays
1838 * and older kernels
1839 */
1840 int vers = md_get_version(mdfd);
1841 int rv;
1842
1843#ifndef MDASSEMBLE
1844 if (st->ss->external)
1845 rv = sysfs_set_array(info, vers);
1846 else
1847#endif
1848 if ((vers % 100) >= 1) { /* can use different versions */
1849 mdu_array_info_t inf;
1850 memset(&inf, 0, sizeof(inf));
1851 inf.major_version = info->array.major_version;
1852 inf.minor_version = info->array.minor_version;
1853 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1854 } else
1855 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1856 return rv;
1857}
1858
1e5c6983
DW
1859unsigned long long min_recovery_start(struct mdinfo *array)
1860{
1861 /* find the minimum recovery_start in an array for metadata
1862 * formats that only record per-array recovery progress instead
1863 * of per-device
1864 */
1865 unsigned long long recovery_start = MaxSector;
1866 struct mdinfo *d;
1867
1868 for (d = array->devs; d; d = d->next)
1869 recovery_start = min(recovery_start, d->recovery_start);
1870
1871 return recovery_start;
1872}
1873
4dd2df09 1874int mdmon_pid(char *devnm)
a931db9e
NB
1875{
1876 char path[100];
1877 char pid[10];
1878 int fd;
1879 int n;
10013317 1880
4dd2df09 1881 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1882
24f6f99b 1883 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1884
1885 if (fd < 0)
cf556303 1886 return -1;
a931db9e
NB
1887 n = read(fd, pid, 9);
1888 close(fd);
1889 if (n <= 0)
cf556303 1890 return -1;
24f6f99b 1891 return atoi(pid);
a931db9e
NB
1892}
1893
4dd2df09 1894int mdmon_running(char *devnm)
a931db9e 1895{
4dd2df09 1896 int pid = mdmon_pid(devnm);
24f6f99b 1897 if (pid <= 0)
a931db9e 1898 return 0;
24f6f99b 1899 if (kill(pid, 0) == 0)
a931db9e
NB
1900 return 1;
1901 return 0;
1902}
1903
4dd2df09 1904int start_mdmon(char *devnm)
8850ee3e 1905{
1913c325 1906 int i, skipped;
44d2e365 1907 int len;
ca3b6696 1908 pid_t pid;
9fe32043 1909 int status;
44d2e365
N
1910 char pathbuf[1024];
1911 char *paths[4] = {
1912 pathbuf,
85945e19 1913 BINDIR "/mdmon",
2f1bcf43 1914 "./mdmon",
44d2e365
N
1915 NULL
1916 };
8850ee3e 1917
40ebbb9c 1918 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1919 return 0;
1920
9cf014ec 1921 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1922 if (len > 0) {
1923 char *sl;
1924 pathbuf[len] = 0;
1925 sl = strrchr(pathbuf, '/');
1926 if (sl)
1927 sl++;
1928 else
1929 sl = pathbuf;
1930 strcpy(sl, "mdmon");
1931 } else
1932 pathbuf[0] = '\0';
1933
0f7bdf89 1934 /* First try to run systemctl */
9540cc24
N
1935 if (!check_env("MDADM_NO_SYSTEMCTL"))
1936 switch(fork()) {
1937 case 0:
1938 /* FIXME yuk. CLOSE_EXEC?? */
1939 skipped = 0;
1940 for (i = 3; skipped < 20; i++)
1941 if (close(i) < 0)
1942 skipped++;
1943 else
1944 skipped = 0;
1945
1946 /* Don't want to see error messages from
1947 * systemctl. If the service doesn't exist,
1948 * we start mdmon ourselves.
1949 */
1950 close(2);
1951 open("/dev/null", O_WRONLY);
1952 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1953 devnm);
1954 status = execl("/usr/bin/systemctl", "systemctl",
1955 "start",
1956 pathbuf, NULL);
1957 status = execl("/bin/systemctl", "systemctl", "start",
1958 pathbuf, NULL);
1959 exit(1);
7a862a02 1960 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1961 return -1;
1962 default: /* parent - good */
1963 pid = wait(&status);
1964 if (pid >= 0 && status == 0)
1965 return 0;
1966 }
0f7bdf89
JS
1967
1968 /* That failed, try running mdmon directly */
8850ee3e
N
1969 switch(fork()) {
1970 case 0:
1971 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1972 skipped = 0;
ca3b6696 1973 for (i = 3; skipped < 20; i++)
1913c325
N
1974 if (close(i) < 0)
1975 skipped++;
1976 else
1977 skipped = 0;
1978
ca3b6696 1979 for (i = 0; paths[i]; i++)
a0963a86 1980 if (paths[i][0]) {
2f1bcf43 1981 execl(paths[i], paths[i],
4dd2df09 1982 devnm, NULL);
a0963a86 1983 }
8850ee3e 1984 exit(1);
7a862a02 1985 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 1986 return -1;
9fe32043
N
1987 default: /* parent - good */
1988 pid = wait(&status);
15c10423 1989 if (pid < 0 || status != 0) {
7a862a02 1990 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 1991 return -1;
15c10423 1992 }
8850ee3e
N
1993 }
1994 return 0;
1995}
1996
148acb7b
DW
1997__u32 random32(void)
1998{
1999 __u32 rv;
2000 int rfd = open("/dev/urandom", O_RDONLY);
2001 if (rfd < 0 || read(rfd, &rv, 4) != 4)
2002 rv = random();
2003 if (rfd >= 0)
2004 close(rfd);
2005 return rv;
2006}
2007
c5f71c24
JS
2008void random_uuid(__u8 *buf)
2009{
2010 int fd, i, len;
2011 __u32 r[4];
2012
2013 fd = open("/dev/urandom", O_RDONLY);
2014 if (fd < 0)
2015 goto use_random;
2016 len = read(fd, buf, 16);
2017 close(fd);
2018 if (len != 16)
2019 goto use_random;
2020
2021 return;
2022
2023use_random:
2024 for (i = 0; i < 4; i++)
2025 r[i] = random();
2026 memcpy(buf, r, 16);
2027}
2028
0e600426 2029#ifndef MDASSEMBLE
edd8d13c
NB
2030int flush_metadata_updates(struct supertype *st)
2031{
2032 int sfd;
2033 if (!st->updates) {
2034 st->update_tail = NULL;
2035 return -1;
2036 }
2037
4dd2df09 2038 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
2039 if (sfd < 0)
2040 return -1;
2041
2042 while (st->updates) {
2043 struct metadata_update *mu = st->updates;
2044 st->updates = mu->next;
2045
2046 send_message(sfd, mu, 0);
2047 wait_reply(sfd, 0);
2048 free(mu->buf);
2049 free(mu);
2050 }
2051 ack(sfd, 0);
2052 wait_reply(sfd, 0);
2053 close(sfd);
2054 st->update_tail = NULL;
2055 return 0;
2056}
2057
2058void append_metadata_update(struct supertype *st, void *buf, int len)
2059{
2060
503975b9 2061 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
2062
2063 mu->buf = buf;
2064 mu->len = len;
2065 mu->space = NULL;
cb23f1f4 2066 mu->space_list = NULL;
edd8d13c
NB
2067 mu->next = NULL;
2068 *st->update_tail = mu;
2069 st->update_tail = &mu->next;
2070}
0e600426 2071#endif /* MDASSEMBLE */
a931db9e 2072
fe6729fa
NB
2073#ifdef __TINYC__
2074/* tinyc doesn't optimize this check in ioctl.h out ... */
2075unsigned int __invalid_size_argument_for_IOC = 0;
2076#endif
2077
6d11ec6f
AK
2078int experimental(void)
2079{
2080 if (check_env("MDADM_EXPERIMENTAL"))
2081 return 1;
2082 else {
7a862a02 2083 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
2084 return 0;
2085 }
2086}
2087
326727d9
AC
2088/* Pick all spares matching given criteria from a container
2089 * if min_size == 0 do not check size
2090 * if domlist == NULL do not check domains
2091 * if spare_group given add it to domains of each spare
2092 * metadata allows to test domains using metadata of destination array */
2093struct mdinfo *container_choose_spares(struct supertype *st,
2094 unsigned long long min_size,
2095 struct domainlist *domlist,
2096 char *spare_group,
2097 const char *metadata, int get_one)
2098{
2099 struct mdinfo *d, **dp, *disks = NULL;
2100
2101 /* get list of all disks in container */
2102 if (st->ss->getinfo_super_disks)
2103 disks = st->ss->getinfo_super_disks(st);
2104
2105 if (!disks)
2106 return disks;
2107 /* find spare devices on the list */
2108 dp = &disks->devs;
2109 disks->array.spare_disks = 0;
2110 while (*dp) {
2111 int found = 0;
2112 d = *dp;
2113 if (d->disk.state == 0) {
2114 /* check if size is acceptable */
2115 unsigned long long dev_size;
2116 dev_t dev = makedev(d->disk.major,d->disk.minor);
2117
2118 if (!min_size ||
2119 (dev_size_from_id(dev, &dev_size) &&
2120 dev_size >= min_size))
2121 found = 1;
2122 /* check if domain matches */
2123 if (found && domlist) {
4dd2df09 2124 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2125 if (spare_group)
2126 pol_add(&pol, pol_domain,
2127 spare_group, NULL);
e5508b36 2128 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2129 found = 0;
2130 dev_policy_free(pol);
2131 }
2132 }
2133 if (found) {
2134 dp = &d->next;
2135 disks->array.spare_disks++;
2136 if (get_one) {
2137 sysfs_free(*dp);
2138 d->next = NULL;
2139 }
2140 } else {
2141 *dp = d->next;
2142 d->next = NULL;
2143 sysfs_free(d);
2144 }
2145 }
2146 return disks;
2147}
9eafa1de
MN
2148
2149/* Checks if paths point to the same device
2150 * Returns 0 if they do.
2151 * Returns 1 if they don't.
2152 * Returns -1 if something went wrong,
2153 * e.g. paths are empty or the files
2154 * they point to don't exist */
2155int compare_paths (char* path1, char* path2)
2156{
2157 struct stat st1,st2;
2158
2159 if (path1 == NULL || path2 == NULL)
2160 return -1;
2161 if (stat(path1,&st1) != 0)
2162 return -1;
2163 if (stat(path2,&st2) != 0)
2164 return -1;
2165 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2166 return 0;
2167 return 1;
2168}
a7dec3fd
N
2169
2170/* Make sure we can open as many devices as needed */
2171void enable_fds(int devices)
2172{
2173 unsigned int fds = 20 + devices;
2174 struct rlimit lim;
2175 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
2176 || lim.rlim_cur >= fds)
2177 return;
2178 if (lim.rlim_max < fds)
2179 lim.rlim_max = fds;
2180 lim.rlim_cur = fds;
2181 setrlimit(RLIMIT_NOFILE, &lim);
2182}
a9c15847
N
2183
2184int in_initrd(void)
2185{
2186 /* This is based on similar function in systemd. */
2187 struct statfs s;
76d0f188
JS
2188 /* statfs.f_type is signed long on s390x and MIPS, causing all
2189 sorts of sign extension problems with RAMFS_MAGIC being
2190 defined as 0x858458f6 */
a9c15847 2191 return statfs("/", &s) >= 0 &&
7ac5d47e 2192 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2193 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2194 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2195}
8832342d
N
2196
2197void reopen_mddev(int mdfd)
2198{
2199 /* Re-open without any O_EXCL, but keep
2200 * the same fd
2201 */
2202 char *devnm;
2203 int fd;
2204 devnm = fd2devnm(mdfd);
2205 close(mdfd);
2206 fd = open_dev(devnm);
2207 if (fd >= 0 && fd != mdfd)
2208 dup2(fd, mdfd);
2209}
7716570e 2210
7071320a 2211#ifndef MDASSEMBLE
e80357f8
GJ
2212static struct cmap_hooks *cmap_hooks = NULL;
2213static int is_cmap_hooks_ready = 0;
7716570e 2214
e80357f8
GJ
2215void set_cmap_hooks(void)
2216{
2217 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2218 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2219 if (!cmap_hooks->cmap_handle)
2220 return;
7716570e 2221
e80357f8
GJ
2222 cmap_hooks->initialize = dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2223 cmap_hooks->get_string = dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
2224 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2225
e80357f8
GJ
2226 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2227 !cmap_hooks->finalize)
2228 dlclose(cmap_hooks->cmap_handle);
2229 else
2230 is_cmap_hooks_ready = 1;
2231}
7716570e 2232
e80357f8
GJ
2233int get_cluster_name(char **cluster_name)
2234{
2235 int rv = -1;
2236 cmap_handle_t handle;
7716570e 2237
e80357f8
GJ
2238 if (!is_cmap_hooks_ready)
2239 return rv;
7716570e 2240
e80357f8 2241 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2242 if (rv != CS_OK)
2243 goto out;
2244
e80357f8 2245 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2246 if (rv != CS_OK) {
2247 free(*cluster_name);
2248 rv = -1;
2249 goto name_err;
2250 }
2251
2252 rv = 0;
2253name_err:
e80357f8 2254 cmap_hooks->finalize(handle);
7716570e 2255out:
7716570e
GJ
2256 return rv;
2257}
d15a1f72
GJ
2258
2259void set_dlm_hooks(void)
2260{
2261 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2262 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2263 if (!dlm_hooks->dlm_handle)
2264 return;
2265
2266 dlm_hooks->create_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2267 dlm_hooks->release_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
2268 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2269 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2270 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2271 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2272
2273 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2274 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2275 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2276 dlclose(dlm_hooks->dlm_handle);
2277 else
2278 is_dlm_hooks_ready = 1;
2279}
e80357f8
GJ
2280
2281void set_hooks(void)
2282{
2283 set_dlm_hooks();
2284 set_cmap_hooks();
2285}
7d55dca2 2286#endif