Revert "mdadm/grow: reshape would be stuck from raid1 to raid5"
[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
018a4882
JS
223/*
224 * Set array info
225 */
226int md_set_array_info(int fd, struct mdu_array_info_s *array)
227{
228 return ioctl(fd, SET_ARRAY_INFO, array);
229}
230
d97572f5
JS
231/*
232 * Get disk info from the kernel.
233 */
234int md_get_disk_info(int fd, struct mdu_disk_info_s *disk)
235{
236 return ioctl(fd, GET_DISK_INFO, disk);
237}
238
64c4757e
NB
239/*
240 * Parse a 128 bit uuid in 4 integers
241 * format is 32 hexx nibbles with options :.<space> separator
242 * If not exactly 32 hex digits are found, return 0
243 * else return 1
244 */
245int parse_uuid(char *str, int uuid[4])
246{
aba69144
NB
247 int hit = 0; /* number of Hex digIT */
248 int i;
249 char c;
ca3b6696
N
250 for (i = 0; i < 4; i++)
251 uuid[i] = 0;
aba69144 252
ca3b6696 253 while ((c = *str++) != 0) {
aba69144 254 int n;
ca3b6696 255 if (c >= '0' && c <= '9')
aba69144 256 n = c-'0';
ca3b6696 257 else if (c >= 'a' && c <= 'f')
aba69144 258 n = 10 + c - 'a';
ca3b6696 259 else if (c >= 'A' && c <= 'F')
aba69144
NB
260 n = 10 + c - 'A';
261 else if (strchr(":. -", c))
262 continue;
263 else return 0;
264
265 if (hit<32) {
266 uuid[hit/8] <<= 4;
267 uuid[hit/8] += n;
268 }
269 hit++;
82b27616 270 }
aba69144
NB
271 if (hit == 32)
272 return 1;
273 return 0;
64c4757e
NB
274}
275
64c4757e
NB
276int get_linux_version()
277{
278 struct utsname name;
98c6faba 279 char *cp;
f161d047 280 int a = 0, b = 0,c = 0;
64c4757e
NB
281 if (uname(&name) <0)
282 return -1;
283
98c6faba
NB
284 cp = name.release;
285 a = strtoul(cp, &cp, 10);
f161d047
N
286 if (*cp == '.')
287 b = strtoul(cp+1, &cp, 10);
288 if (*cp == '.')
289 c = strtoul(cp+1, &cp, 10);
98c6faba 290
682c7051 291 return (a*1000000)+(b*1000)+c;
64c4757e
NB
292}
293
71204a50 294#ifndef MDASSEMBLE
bc77ed53
DW
295int mdadm_version(char *version)
296{
297 int a, b, c;
298 char *cp;
299
300 if (!version)
301 version = Version;
302
303 cp = strchr(version, '-');
304 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
305 return -1;
306 cp += 3;
307 a = strtoul(cp, &cp, 10);
308 if (*cp != '.')
309 return -1;
310 b = strtoul(cp+1, &cp, 10);
311 if (*cp == '.')
312 c = strtoul(cp+1, &cp, 10);
313 else
314 c = 0;
315 if (*cp != ' ' && *cp != '-')
316 return -1;
317 return (a*1000000)+(b*1000)+c;
318}
319
51d4261c 320unsigned long long parse_size(char *size)
84e11361
N
321{
322 /* parse 'size' which should be a number optionally
323 * followed by 'K', 'M', or 'G'.
324 * Without a suffix, K is assumed.
325 * Number returned is in sectors (half-K)
822e393a 326 * INVALID_SECTORS returned on error.
84e11361
N
327 */
328 char *c;
329 long long s = strtoll(size, &c, 10);
330 if (s > 0) {
331 switch (*c) {
332 case 'K':
333 c++;
334 default:
335 s *= 2;
336 break;
337 case 'M':
338 c++;
339 s *= 1024 * 2;
340 break;
341 case 'G':
342 c++;
343 s *= 1024 * 1024 * 2;
344 break;
72ca9bcf
N
345 case 's': /* sectors */
346 c++;
347 break;
84e11361 348 }
51d4261c 349 } else
822e393a 350 s = INVALID_SECTORS;
84e11361 351 if (*c)
822e393a 352 s = INVALID_SECTORS;
84e11361
N
353 return s;
354}
355
4a06e2c2
N
356int parse_layout_10(char *layout)
357{
358 int copies, rv;
359 char *cp;
360 /* Parse the layout string for raid10 */
361 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
362 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
363 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
364 copies > 200 ||
365 *cp)
366 return -1;
367 if (layout[0] == 'n')
368 rv = 256 + copies;
369 else if (layout[0] == 'o')
370 rv = 0x10000 + (copies<<8) + 1;
371 else
372 rv = 1 + (copies<<8);
373 return rv;
374}
375
376int parse_layout_faulty(char *layout)
377{
378 /* Parse the layout string for 'faulty' */
379 int ln = strcspn(layout, "0123456789");
503975b9 380 char *m = xstrdup(layout);
4a06e2c2
N
381 int mode;
382 m[ln] = 0;
383 mode = map_name(faultylayout, m);
384 if (mode == UnSet)
385 return -1;
386
387 return mode | (atoi(layout+ln)<< ModeShift);
388}
79868890
N
389
390long parse_num(char *num)
391{
392 /* Either return a valid number, or -1 */
393 char *c;
394 long rv = strtol(num, &c, 10);
395 if (rv < 0 || *c || !num[0])
396 return -1;
397 else
398 return rv;
399}
eb3929a4 400#endif
4a06e2c2 401
4de90913
GJ
402int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
403{
404 char *dev;
405 *slot = strtoul(input, &dev, 10);
406 if (dev == input || dev[0] != ':')
407 return -1;
408 *devname = dev+1;
409 return 0;
410}
411
0430ed48
NB
412void remove_partitions(int fd)
413{
414 /* remove partitions from this block devices.
415 * This is used for components added to an array
416 */
417#ifdef BLKPG_DEL_PARTITION
418 struct blkpg_ioctl_arg a;
419 struct blkpg_partition p;
420
421 a.op = BLKPG_DEL_PARTITION;
422 a.data = (void*)&p;
423 a.datalen = sizeof(p);
424 a.flags = 0;
425 memset(a.data, 0, a.datalen);
ca3b6696 426 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
427 ioctl(fd, BLKPG, &a);
428#endif
429}
430
691c6ee1
N
431int test_partition(int fd)
432{
433 /* Check if fd is a whole-disk or a partition.
434 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
435 * will return ENXIO on an invalid partition number.
436 */
437 struct blkpg_ioctl_arg a;
438 struct blkpg_partition p;
439 a.op = BLKPG_DEL_PARTITION;
440 a.data = (void*)&p;
441 a.datalen = sizeof(p);
442 a.flags = 0;
443 memset(a.data, 0, a.datalen);
444 p.pno = 1<<30;
445 if (ioctl(fd, BLKPG, &a) == 0)
446 /* Very unlikely, but not a partition */
447 return 0;
357ac106 448 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
449 /* not a partition */
450 return 0;
451
452 return 1;
453}
454
bfd76b93
CA
455int test_partition_from_id(dev_t id)
456{
457 char buf[20];
458 int fd, rv;
459
460 sprintf(buf, "%d:%d", major(id), minor(id));
461 fd = dev_open(buf, O_RDONLY);
462 if (fd < 0)
463 return -1;
464 rv = test_partition(fd);
465 close(fd);
466 return rv;
467}
691c6ee1 468
de5a472e 469int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 470{
265e0f17 471 int copies, first;
de5a472e
N
472 int i;
473 int avail_disks = 0;
474
475 for (i = 0; i < raid_disks; i++)
476 avail_disks += !!avail[i];
477
64c4757e 478 switch (level) {
265e0f17
NB
479 case 10:
480 /* This is the tricky one - we need to check
481 * which actual disks are present.
482 */
702b557b 483 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 484 first = 0;
265e0f17
NB
485 do {
486 /* there must be one of the 'copies' form 'first' */
487 int n = copies;
ca3b6696 488 int cnt = 0;
b3ec716d 489 int this = first;
265e0f17 490 while (n--) {
b3ec716d 491 if (avail[this])
265e0f17 492 cnt++;
b3ec716d 493 this = (this+1) % raid_disks;
265e0f17
NB
494 }
495 if (cnt == 0)
496 return 0;
b3ec716d 497 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
498 } while (first != 0);
499 return 1;
e5329c37 500
df0d4ea0 501 case LEVEL_MULTIPATH:
e0d19036 502 return avail_disks>= 1;
df0d4ea0 503 case LEVEL_LINEAR:
64c4757e
NB
504 case 0:
505 return avail_disks == raid_disks;
506 case 1:
507 return avail_disks >= 1;
508 case 4:
330d6900
N
509 if (avail_disks == raid_disks - 1 &&
510 !avail[raid_disks - 1])
511 /* If just the parity device is missing, then we
512 * have enough, even if not clean
513 */
514 return 1;
515 /* FALL THROUGH */
64c4757e 516 case 5:
583315d9
NB
517 if (clean)
518 return avail_disks >= raid_disks-1;
519 else
520 return avail_disks >= raid_disks;
98c6faba 521 case 6:
583315d9
NB
522 if (clean)
523 return avail_disks >= raid_disks-2;
524 else
525 return avail_disks >= raid_disks;
64c4757e
NB
526 default:
527 return 0;
528 }
529}
530
8453e704
N
531int enough_fd(int fd)
532{
533 struct mdu_array_info_s array;
534 struct mdu_disk_info_s disk;
aabe020d 535 int i, rv;
8453e704
N
536 char *avail;
537
9cd39f01 538 if (md_get_array_info(fd, &array) != 0 || array.raid_disks <= 0)
8453e704 539 return 0;
503975b9 540 avail = xcalloc(array.raid_disks, 1);
ca3b6696 541 for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704 542 disk.number = i;
d97572f5 543 if (md_get_disk_info(fd, &disk) != 0)
8453e704 544 continue;
9e6d9291
N
545 if (disk.major == 0 && disk.minor == 0)
546 continue;
547 array.nr_disks--;
548
8453e704
N
549 if (! (disk.state & (1<<MD_DISK_SYNC)))
550 continue;
551 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
552 continue;
8453e704
N
553 avail[disk.raid_disk] = 1;
554 }
555 /* This is used on an active array, so assume it is clean */
aabe020d 556 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 557 1, avail);
aabe020d
JS
558 free(avail);
559 return rv;
8453e704
N
560}
561
22e263f6
AC
562const int uuid_zero[4] = { 0, 0, 0, 0 };
563
f277ce36 564int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 565{
f277ce36
NB
566 if (swapuuid) {
567 /* parse uuids are hostendian.
568 * uuid's from some superblocks are big-ending
aba69144 569 * if there is a difference, we need to swap..
f277ce36
NB
570 */
571 unsigned char *ac = (unsigned char *)a;
572 unsigned char *bc = (unsigned char *)b;
573 int i;
ca3b6696 574 for (i = 0; i < 16; i += 4) {
f277ce36
NB
575 if (ac[i+0] != bc[i+3] ||
576 ac[i+1] != bc[i+2] ||
577 ac[i+2] != bc[i+1] ||
578 ac[i+3] != bc[i+0])
579 return 0;
580 }
581 return 1;
582 } else {
583 if (a[0]==b[0] &&
584 a[1]==b[1] &&
585 a[2]==b[2] &&
586 a[3]==b[3])
587 return 1;
588 return 0;
589 }
64c4757e 590}
ca3b6696 591
350f29f9
NB
592void copy_uuid(void *a, int b[4], int swapuuid)
593{
594 if (swapuuid) {
595 /* parse uuids are hostendian.
596 * uuid's from some superblocks are big-ending
597 * if there is a difference, we need to swap..
598 */
599 unsigned char *ac = (unsigned char *)a;
600 unsigned char *bc = (unsigned char *)b;
601 int i;
ca3b6696 602 for (i = 0; i < 16; i += 4) {
350f29f9
NB
603 ac[i+0] = bc[i+3];
604 ac[i+1] = bc[i+2];
605 ac[i+2] = bc[i+1];
606 ac[i+3] = bc[i+0];
607 }
608 } else
609 memcpy(a, b, 16);
610}
64c4757e 611
aae5a112 612char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 613{
9968e376 614 int i, j;
d7288ddc
N
615 char uuid[16];
616 char *c = buf;
617 strcpy(c, "UUID-");
618 c += strlen(c);
aae5a112 619 copy_uuid(uuid, id, swap);
9968e376 620 for (i = 0; i < 4; i++) {
9968e376 621 if (i)
ff54de6e 622 *c++ = sep;
9968e376
DW
623 for (j = 3; j >= 0; j--) {
624 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
625 c+= 2;
626 }
d7288ddc
N
627 }
628 return buf;
aae5a112
DW
629
630}
631
632char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
633{
86983cce
N
634 // dirty hack to work around an issue with super1 superblocks...
635 // super1 superblocks need swapuuid set in order for assembly to
636 // work, but can't have it set if we want this printout to match
637 // all the other uuid printouts in super1.c, so we force swapuuid
638 // to 1 to make our printout match the rest of super1
639 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
640}
641
435d4ebb 642#ifndef MDASSEMBLE
682c7051
NB
643int check_ext2(int fd, char *name)
644{
645 /*
646 * Check for an ext2fs file system.
647 * Superblock is always 1K at 1K offset
648 *
649 * s_magic is le16 at 56 == 0xEF53
650 * report mtime - le32 at 44
651 * blocks - le32 at 4
652 * logblksize - le32 at 24
653 */
654 unsigned char sb[1024];
655 time_t mtime;
2538ba2a
N
656 unsigned long long size;
657 int bsize;
682c7051
NB
658 if (lseek(fd, 1024,0)!= 1024)
659 return 0;
660 if (read(fd, sb, 1024)!= 1024)
661 return 0;
662 if (sb[56] != 0x53 || sb[57] != 0xef)
663 return 0;
664
665 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
666 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
667 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 668 size <<= bsize;
e7b84f9d 669 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 670 name);
2538ba2a 671 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
672 return 1;
673}
674
675int check_reiser(int fd, char *name)
676{
677 /*
678 * superblock is at 64K
679 * size is 1024;
680 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
681 *
682 */
683 unsigned char sb[1024];
2538ba2a 684 unsigned long long size;
682c7051
NB
685 if (lseek(fd, 64*1024, 0) != 64*1024)
686 return 0;
687 if (read(fd, sb, 1024) != 1024)
688 return 0;
ca3b6696
N
689 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
690 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 691 return 0;
e7b84f9d 692 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 693 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 694 cont_err("size = %lluK\n", size*4);
aba69144 695
682c7051
NB
696 return 1;
697}
698
699int check_raid(int fd, char *name)
700{
4b1ac34b 701 struct mdinfo info;
682c7051 702 time_t crtime;
d078d77c 703 char *level;
82d9eba6 704 struct supertype *st = guess_super(fd);
f9ce90ba 705
cb8f6859
N
706 if (!st)
707 return 0;
fa219dd2
MD
708 if (st->ss->add_to_super != NULL) {
709 st->ss->load_super(st, fd, name);
710 /* Looks like a raid array .. */
711 pr_err("%s appears to be part of a raid array:\n", name);
712 st->ss->getinfo_super(st, &info, NULL);
713 st->ss->free_super(st);
714 crtime = info.array.ctime;
715 level = map_num(pers, info.array.level);
716 if (!level)
717 level = "-unknown-";
718 cont_err("level=%s devices=%d ctime=%s",
719 level, info.array.raid_disks, ctime(&crtime));
720 } else {
721 /* Looks like GPT or MBR */
722 pr_err("partition table exists on %s\n", name);
723 }
82d9eba6 724 return 1;
682c7051
NB
725}
726
682c7051
NB
727int ask(char *mesg)
728{
729 char *add = "";
730 int i;
ca3b6696 731 for (i = 0; i < 5; i++) {
682c7051
NB
732 char buf[100];
733 fprintf(stderr, "%s%s", mesg, add);
734 fflush(stderr);
735 if (fgets(buf, 100, stdin)==NULL)
736 return 0;
737 if (buf[0]=='y' || buf[0]=='Y')
738 return 1;
739 if (buf[0]=='n' || buf[0]=='N')
740 return 0;
741 add = "(y/n) ";
742 }
e7b84f9d 743 pr_err("assuming 'no'\n");
682c7051
NB
744 return 0;
745}
435d4ebb 746#endif /* MDASSEMBLE */
682c7051 747
8d80900b 748int is_standard(char *dev, int *nump)
e5329c37
NB
749{
750 /* tests if dev is a "standard" md dev name.
751 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 752 * where NN is a string of digits
598f0d58
NB
753 * Returns 1 if a partitionable standard,
754 * -1 if non-partitonable,
755 * 0 if not a standard name.
e5329c37 756 */
8d80900b 757 char *d = strrchr(dev, '/');
ca3b6696 758 int type = 0;
8d80900b
NB
759 int num;
760 if (!d)
e5329c37 761 return 0;
ca3b6696
N
762 if (strncmp(d, "/d",2) == 0)
763 d += 2, type = 1; /* /dev/md/dN{pM} */
764 else if (strncmp(d, "/md_d", 5) == 0)
765 d += 5, type = 1; /* /dev/md_dN{pM} */
766 else if (strncmp(d, "/md", 3) == 0)
767 d += 3, type = -1; /* /dev/mdN */
768 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
769 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
770 else
771 return 0;
8d80900b 772 if (!*d)
e5329c37 773 return 0;
8d80900b
NB
774 num = atoi(d);
775 while (isdigit(*d))
776 d++;
777 if (*d)
e5329c37 778 return 0;
8d80900b
NB
779 if (nump) *nump = num;
780
781 return type;
e5329c37
NB
782}
783
4b1ac34b 784unsigned long calc_csum(void *super, int bytes)
82b27616 785{
56eb10c0 786 unsigned long long newcsum = 0;
82b27616 787 int i;
4b1ac34b
NB
788 unsigned int csum;
789 unsigned int *superc = (unsigned int*) super;
82b27616 790
ca3b6696
N
791 for(i = 0; i < bytes/4; i++)
792 newcsum += superc[i];
82b27616 793 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 794#ifdef __alpha__
aba69144 795/* The in-kernel checksum calculation is always 16bit on
570c0542 796 * the alpha, though it is 32 bit on i386...
ca3b6696 797 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
798 * a way that it shouldn't).
799 */
800 csum = (csum & 0xffff) + (csum >> 16);
801 csum = (csum & 0xffff) + (csum >> 16);
802#endif
82b27616
NB
803 return csum;
804}
cd29a5c8 805
435d4ebb 806#ifndef MDASSEMBLE
56eb10c0 807char *human_size(long long bytes)
cd29a5c8 808{
ff9239ee 809 static char buf[47];
d5d3721e
NB
810
811 /* We convert bytes to either centi-M{ega,ibi}bytes or
812 * centi-G{igi,ibi}bytes, with appropriate rounding,
813 * and then print 1/100th of those as a decimal.
814 * We allow upto 2048Megabytes before converting to
815 * gigabytes, as that shows more precision and isn't
816 * too large a number.
01619b48 817 * Terabytes are not yet handled.
d5d3721e 818 */
cd29a5c8 819
56eb10c0 820 if (bytes < 5000*1024)
ca3b6696 821 buf[0] = 0;
d5d3721e 822 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 823 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 824 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 825 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
efa29530 826 cMiB/100, cMiB % 100, cMB/100, cMB % 100);
d5d3721e 827 } else {
93d3bd3b 828 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 829 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 830 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
efa29530 831 cGiB/100, cGiB % 100, cGB/100, cGB % 100);
d5d3721e 832 }
cd29a5c8
NB
833 return buf;
834}
e0d19036 835
f0ec6710 836char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
837{
838 static char buf[30];
e0d19036 839
570abc6f
MN
840 /* We convert bytes to either centi-M{ega,ibi}bytes or
841 * centi-G{igi,ibi}bytes, with appropriate rounding,
842 * and then print 1/100th of those as a decimal.
843 * We allow upto 2048Megabytes before converting to
844 * gigabytes, as that shows more precision and isn't
845 * too large a number.
846 * Terabytes are not yet handled.
f0ec6710
MN
847 *
848 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
849 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
850 */
851
e0d19036 852 if (bytes < 5000*1024)
570abc6f 853 buf[0] = 0;
f0ec6710
MN
854 else if (prefix == IEC) {
855 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 856 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710 857 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
efa29530 858 cMiB/100, cMiB % 100);
f0ec6710 859 } else {
93d3bd3b 860 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710 861 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
efa29530 862 cGiB/100, cGiB % 100);
f0ec6710
MN
863 }
864 }
865 else if (prefix == JEDEC) {
866 if (bytes < 2*1024LL*1024LL*1024LL) {
867 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
868 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
efa29530 869 cMB/100, cMB % 100);
f0ec6710
MN
870 } else {
871 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
872 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
efa29530 873 cGB/100, cGB % 100);
f0ec6710 874 }
570abc6f 875 }
f0ec6710
MN
876 else
877 buf[0] = 0;
570abc6f 878
e0d19036
NB
879 return buf;
880}
e4965ef8
N
881
882void print_r10_layout(int layout)
883{
884 int near = layout & 255;
885 int far = (layout >> 8) & 255;
886 int offset = (layout&0x10000);
887 char *sep = "";
888
889 if (near != 1) {
890 printf("%s near=%d", sep, near);
891 sep = ",";
892 }
893 if (far != 1)
894 printf("%s %s=%d", sep, offset?"offset":"far", far);
895 if (near*far == 1)
896 printf("NO REDUNDANCY");
897}
435d4ebb 898#endif
e0d19036 899
5f8097be
NB
900unsigned long long calc_array_size(int level, int raid_disks, int layout,
901 int chunksize, unsigned long long devsize)
577e8448 902{
e03640bd
CA
903 if (level == 1)
904 return devsize;
577e8448
AK
905 devsize &= ~(unsigned long long)((chunksize>>9)-1);
906 return get_data_disks(level, layout, raid_disks) * devsize;
907}
908
909int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
910{
911 int data_disks = 0;
912 switch (level) {
ca3b6696
N
913 case 0: data_disks = raid_disks;
914 break;
915 case 1: data_disks = 1;
916 break;
5f8097be 917 case 4:
ca3b6696
N
918 case 5: data_disks = raid_disks - 1;
919 break;
920 case 6: data_disks = raid_disks - 2;
921 break;
5f8097be
NB
922 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
923 break;
924 }
577e8448
AK
925
926 return data_disks;
5f8097be
NB
927}
928
13db17bd 929dev_t devnm2devid(char *devnm)
4dd2df09
N
930{
931 /* First look in /sys/block/$DEVNM/dev for %d:%d
932 * If that fails, try parsing out a number
933 */
934 char path[100];
935 char *ep;
936 int fd;
937 int mjr,mnr;
938
939 sprintf(path, "/sys/block/%s/dev", devnm);
940 fd = open(path, O_RDONLY);
941 if (fd >= 0) {
942 char buf[20];
943 int n = read(fd, buf, sizeof(buf));
944 close(fd);
945 if (n > 0)
946 buf[n] = 0;
947 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
948 return makedev(mjr, mnr);
949 }
950 if (strncmp(devnm, "md_d", 4) == 0 &&
951 isdigit(devnm[4]) &&
952 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
953 ep > devnm && *ep == 0)
954 return makedev(get_mdp_major(), mnr << MdpMinorShift);
955
956 if (strncmp(devnm, "md", 2) == 0 &&
957 isdigit(devnm[2]) &&
958 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
959 ep > devnm && *ep == 0)
960 return makedev(MD_MAJOR, mnr);
961
962 return 0;
963}
964
5a23a06e 965#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 966char *get_md_name(char *devnm)
e0d19036
NB
967{
968 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 969 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 970
e0d19036
NB
971 static char devname[50];
972 struct stat stb;
4dd2df09 973 dev_t rdev = devnm2devid(devnm);
dd0781e5 974 char *dn;
98c6faba 975
4dd2df09
N
976 if (rdev == 0)
977 return 0;
978 if (strncmp(devnm, "md_", 3) == 0) {
979 snprintf(devname, sizeof(devname), "/dev/md/%s",
980 devnm + 3);
98c6faba
NB
981 if (stat(devname, &stb) == 0
982 && (S_IFMT&stb.st_mode) == S_IFBLK
983 && (stb.st_rdev == rdev))
984 return devname;
985 }
4dd2df09
N
986 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
987 if (stat(devname, &stb) == 0
988 && (S_IFMT&stb.st_mode) == S_IFBLK
989 && (stb.st_rdev == rdev))
990 return devname;
991
992 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
993 if (stat(devname, &stb) == 0
994 && (S_IFMT&stb.st_mode) == S_IFBLK
995 && (stb.st_rdev == rdev))
996 return devname;
997
16c6fa80 998 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
999 if (dn)
1000 return dn;
4dd2df09 1001 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 1002 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
1003 if (errno != EEXIST)
1004 return NULL;
e0d19036
NB
1005
1006 if (stat(devname, &stb) == 0
1007 && (S_IFMT&stb.st_mode) == S_IFBLK
1008 && (stb.st_rdev == rdev))
1009 return devname;
1010 unlink(devname);
1011 return NULL;
1012}
1013
1014void put_md_name(char *name)
1015{
ca3b6696 1016 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
1017 unlink(name);
1018}
435d4ebb 1019#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 1020
cc742d38
N
1021int get_maj_min(char *dev, int *major, int *minor)
1022{
1023 char *e;
1024 *major = strtoul(dev, &e, 0);
1025 return (e > dev && *e == ':' && e[1] &&
1026 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1027 *e == 0);
1028}
1029
8b0dabea
NB
1030int dev_open(char *dev, int flags)
1031{
1032 /* like 'open', but if 'dev' matches %d:%d, create a temp
1033 * block device and open that
1034 */
8b0dabea
NB
1035 int fd = -1;
1036 char devname[32];
e81cdd9f 1037 int major;
8b0dabea 1038 int minor;
e81cdd9f 1039
9f0ad56b
JS
1040 if (!dev)
1041 return -1;
6df6a774 1042 flags |= O_DIRECT;
e81cdd9f 1043
cc742d38 1044 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1045 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1046 (int)getpid(), major, minor);
1047 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1048 fd = open(devname, flags);
1049 unlink(devname);
6df6a774
N
1050 }
1051 if (fd < 0) {
9dc7d357 1052 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
1053 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
1054 (int)getpid(), major, minor);
ca3b6696 1055 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
1056 fd = open(devname, flags);
1057 unlink(devname);
1058 }
8b0dabea
NB
1059 }
1060 } else
6df6a774 1061 fd = open(dev, flags);
8b0dabea
NB
1062 return fd;
1063}
f9ce90ba 1064
4dd2df09 1065int open_dev_flags(char *devnm, int flags)
e8a70c89 1066{
13db17bd 1067 dev_t devid;
e8a70c89
N
1068 char buf[20];
1069
4dd2df09
N
1070 devid = devnm2devid(devnm);
1071 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1072 return dev_open(buf, flags);
1073}
1074
4dd2df09 1075int open_dev(char *devnm)
d998b738 1076{
4dd2df09 1077 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1078}
1079
4dd2df09 1080int open_dev_excl(char *devnm)
a931db9e
NB
1081{
1082 char buf[20];
1083 int i;
7187750e 1084 int flags = O_RDWR;
13db17bd 1085 dev_t devid = devnm2devid(devnm);
399e0b97 1086 long delay = 1000;
a931db9e 1087
4dd2df09 1088 sprintf(buf, "%d:%d", major(devid), minor(devid));
efa29530 1089 for (i = 0; i < 25; i++) {
7187750e 1090 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1091 if (fd >= 0)
1092 return fd;
7187750e
N
1093 if (errno == EACCES && flags == O_RDWR) {
1094 flags = O_RDONLY;
1095 continue;
1096 }
a931db9e
NB
1097 if (errno != EBUSY)
1098 return fd;
399e0b97
N
1099 usleep(delay);
1100 if (delay < 200000)
1101 delay *= 2;
a931db9e
NB
1102 }
1103 return -1;
1104}
1105
9008ed1c
N
1106int same_dev(char *one, char *two)
1107{
1108 struct stat st1, st2;
1109 if (stat(one, &st1) != 0)
1110 return 0;
1111 if (stat(two, &st2) != 0)
1112 return 0;
1113 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1114 return 0;
1115 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1116 return 0;
1117 return st1.st_rdev == st2.st_rdev;
1118}
1119
a7c6e3fb 1120void wait_for(char *dev, int fd)
a714580e
N
1121{
1122 int i;
a7c6e3fb 1123 struct stat stb_want;
399e0b97 1124 long delay = 1000;
a7c6e3fb
N
1125
1126 if (fstat(fd, &stb_want) != 0 ||
1127 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1128 return;
a714580e 1129
efa29530 1130 for (i = 0; i < 25; i++) {
a714580e 1131 struct stat stb;
a7c6e3fb
N
1132 if (stat(dev, &stb) == 0 &&
1133 (stb.st_mode & S_IFMT) == S_IFBLK &&
1134 (stb.st_rdev == stb_want.st_rdev))
a714580e 1135 return;
399e0b97
N
1136 usleep(delay);
1137 if (delay < 200000)
1138 delay *= 2;
a714580e 1139 }
436305c6 1140 if (i == 25)
1ade5cc1 1141 dprintf("timeout waiting for %s\n", dev);
a714580e
N
1142}
1143
0f22b998
N
1144struct superswitch *superlist[] =
1145{
1146 &super0, &super1,
1147 &super_ddf, &super_imsm,
0592faeb 1148 &mbr, &gpt,
0f22b998 1149 NULL };
f9ce90ba 1150
ea24acd0 1151#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1152
4725bc31 1153struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1154{
1686dc25
NB
1155 mdu_array_info_t array;
1156 int vers;
1157 int minor;
1158 struct supertype *st = NULL;
7e0f6979 1159 struct mdinfo *sra;
142cb9e1 1160 char *verstr;
1686dc25
NB
1161 char version[20];
1162 int i;
f7e7067b 1163 char *subarray = NULL;
4dd2df09 1164 char container[32] = "";
1686dc25 1165
4dd2df09 1166 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1167
1168 if (sra) {
7e0f6979
NB
1169 vers = sra->array.major_version;
1170 minor = sra->array.minor_version;
142cb9e1 1171 verstr = sra->text_version;
1686dc25 1172 } else {
9cd39f01 1173 if (md_get_array_info(fd, &array))
1686dc25
NB
1174 array.major_version = array.minor_version = 0;
1175 vers = array.major_version;
1176 minor = array.minor_version;
142cb9e1 1177 verstr = "";
6fbba4c9 1178 }
82d9eba6 1179
1686dc25
NB
1180 if (vers != -1) {
1181 sprintf(version, "%d.%d", vers, minor);
1182 verstr = version;
6fbba4c9 1183 }
3c558363 1184 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1185 char *dev = verstr+1;
5f7e44b2 1186
f7e7067b 1187 subarray = strchr(dev, '/');
92d49ecf 1188 if (subarray) {
f7e7067b 1189 *subarray++ = '\0';
503975b9 1190 subarray = xstrdup(subarray);
92d49ecf 1191 }
4dd2df09 1192 strcpy(container, dev);
15d230f7 1193 sysfs_free(sra);
5f7e44b2 1194 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1195 if (sra && sra->text_version[0])
1196 verstr = sra->text_version;
1197 else
1198 verstr = "-no-metadata-";
f7e7067b
NB
1199 }
1200
efa29530 1201 for (i = 0; st == NULL && superlist[i]; i++)
f7e7067b 1202 st = superlist[i]->match_metadata_desc(verstr);
1686dc25 1203
15d230f7 1204 sysfs_free(sra);
f7e7067b 1205 if (st) {
3b0896f8 1206 st->sb = NULL;
1f49fb3a
N
1207 if (subarrayp)
1208 *subarrayp = subarray;
4dd2df09
N
1209 strcpy(st->container_devnm, container);
1210 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1211 } else
1212 free(subarray);
5f7e44b2 1213
82d9eba6 1214 return st;
f9ce90ba 1215}
ea24acd0
NB
1216#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1217
0f0749ad 1218int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1219{
1220 char buf[20];
1221 int fd;
1222
1223 sprintf(buf, "%d:%d", major(id), minor(id));
1224 fd = dev_open(buf, O_RDONLY);
1225 if (fd < 0)
1226 return 0;
1227 if (get_dev_size(fd, NULL, size)) {
1228 close(fd);
1229 return 1;
1230 }
1231 close(fd);
1232 return 0;
1233}
f9ce90ba 1234
159c3a1a 1235struct supertype *dup_super(struct supertype *orig)
3da92f27 1236{
159c3a1a 1237 struct supertype *st;
1686dc25 1238
d2ca6449
NB
1239 if (!orig)
1240 return orig;
503975b9 1241 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1242 st->ss = orig->ss;
1243 st->max_devs = orig->max_devs;
1244 st->minor_version = orig->minor_version;
cb8f6859 1245 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1246 st->data_offset = orig->data_offset;
159c3a1a
NB
1247 st->sb = NULL;
1248 st->info = NULL;
1249 return st;
3da92f27
NB
1250}
1251
54887ad8 1252struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1253{
1254 /* try each load_super to find the best match,
1255 * and return the best superswitch
1256 */
82d9eba6
NB
1257 struct superswitch *ss;
1258 struct supertype *st;
1d13b599 1259 unsigned int besttime = 0;
570c0542 1260 int bestsuper = -1;
f9ce90ba
NB
1261 int i;
1262
503975b9 1263 st = xcalloc(1, sizeof(*st));
4dd2df09 1264 st->container_devnm[0] = 0;
d1d599ea 1265
efa29530 1266 for (i = 0; superlist[i]; i++) {
f9ce90ba
NB
1267 int rv;
1268 ss = superlist[i];
54887ad8
N
1269 if (guess_type == guess_array && ss->add_to_super == NULL)
1270 continue;
1271 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1272 continue;
ef609477 1273 memset(st, 0, sizeof(*st));
df3346e6 1274 st->ignore_hw_compat = 1;
3da92f27 1275 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1276 if (rv == 0) {
1277 struct mdinfo info;
a5d85af7 1278 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1279 if (bestsuper == -1 ||
1280 besttime < info.array.ctime) {
1281 bestsuper = i;
1282 besttime = info.array.ctime;
570c0542 1283 }
3da92f27 1284 ss->free_super(st);
570c0542
NB
1285 }
1286 }
1287 if (bestsuper != -1) {
1288 int rv;
ef609477 1289 memset(st, 0, sizeof(*st));
df3346e6 1290 st->ignore_hw_compat = 1;
3da92f27 1291 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1292 if (rv == 0) {
5e747af2 1293 superlist[bestsuper]->free_super(st);
82d9eba6 1294 return st;
f9ce90ba
NB
1295 }
1296 }
570c0542 1297 free(st);
f9ce90ba
NB
1298 return NULL;
1299}
fe6729fa 1300
beae1dfe
NB
1301/* Return size of device in bytes */
1302int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1303{
1304 unsigned long long ldsize;
c2c9bb6f
NB
1305 struct stat st;
1306
1307 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1308 ldsize = (unsigned long long)st.st_size;
1309 else
beae1dfe
NB
1310#ifdef BLKGETSIZE64
1311 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1312#endif
1313 {
1314 unsigned long dsize;
1315 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1316 ldsize = dsize;
1317 ldsize <<= 9;
1318 } else {
1319 if (dname)
32971509 1320 pr_err("Cannot get size of %s: %s\n",
beae1dfe
NB
1321 dname, strerror(errno));
1322 return 0;
1323 }
1324 }
1325 *sizep = ldsize;
1326 return 1;
1327}
8fac0577 1328
32971509
PB
1329/* Return sector size of device in bytes */
1330int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep)
1331{
1332 unsigned int sectsize;
1333
1334 if (ioctl(fd, BLKSSZGET, &sectsize) != 0) {
1335 if (dname)
1336 pr_err("Cannot get sector size of %s: %s\n",
1337 dname, strerror(errno));
1338 return 0;
1339 }
1340
1341 *sectsizep = sectsize;
1342 return 1;
1343}
1344
3a371610
N
1345/* Return true if this can only be a container, not a member device.
1346 * i.e. is and md device and size is zero
1347 */
1348int must_be_container(int fd)
1349{
f5c924f4 1350 struct mdinfo *mdi;
3a371610 1351 unsigned long long size;
f5c924f4
JS
1352
1353 mdi = sysfs_read(fd, NULL, GET_VERSION);
1354 if (!mdi)
3a371610 1355 return 0;
f5c924f4
JS
1356 sysfs_free(mdi);
1357
3a371610
N
1358 if (get_dev_size(fd, NULL, &size) == 0)
1359 return 1;
1360 if (size == 0)
1361 return 1;
1362 return 0;
1363}
034b203a
TM
1364
1365/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1366 * Returns: 1 if successful
1367 * -1 for unknown partition type
1368 * 0 for other errors
1369 */
1370static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1371{
056b331e 1372 struct GPT gpt;
034b203a 1373 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1374 struct GPT_part_entry *part;
1375 char buf[512];
034b203a
TM
1376 unsigned long long curr_part_end;
1377 unsigned all_partitions, entry_size;
f21e18ca 1378 unsigned part_nr;
41b06495 1379 unsigned int sector_size = 0;
034b203a
TM
1380
1381 *endofpart = 0;
1382
056b331e 1383 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1384 /* skip protective MBR */
41b06495
MD
1385 if (!get_dev_sector_size(fd, NULL, &sector_size))
1386 return 0;
1387 lseek(fd, sector_size, SEEK_SET);
73e658d8 1388 /* read GPT header */
056b331e 1389 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1390 return 0;
1391
1392 /* get the number of partition entries and the entry size */
056b331e
N
1393 all_partitions = __le32_to_cpu(gpt.part_cnt);
1394 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1395
1396 /* Check GPT signature*/
056b331e 1397 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1398 return -1;
1399
1400 /* sanity checks */
1401 if (all_partitions > 1024 ||
db7fdfe4 1402 entry_size > sizeof(buf))
034b203a
TM
1403 return -1;
1404
db7fdfe4
JS
1405 part = (struct GPT_part_entry *)buf;
1406
41b06495
MD
1407 /* set offset to third block (GPT entries) */
1408 lseek(fd, sector_size*2, SEEK_SET);
ca3b6696 1409 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1410 /* read partition entry */
db7fdfe4 1411 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1412 return 0;
1413
034b203a 1414 /* is this valid partition? */
db7fdfe4 1415 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1416 /* check the last lba for the current partition */
db7fdfe4 1417 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1418 if (curr_part_end > *endofpart)
1419 *endofpart = curr_part_end;
1420 }
1421
034b203a
TM
1422 }
1423 return 1;
1424}
1425
1426/* Sets endofpart parameter to the last block used by the last partition on the device.
1427 * Returns: 1 if successful
1428 * -1 for unknown partition type
1429 * 0 for other errors
1430 */
1431static int get_last_partition_end(int fd, unsigned long long *endofpart)
1432{
056b331e 1433 struct MBR boot_sect;
034b203a 1434 unsigned long long curr_part_end;
f21e18ca 1435 unsigned part_nr;
31208db9 1436 unsigned int sector_size;
034b203a
TM
1437 int retval = 0;
1438
1439 *endofpart = 0;
1440
056b331e 1441 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1442 /* read MBR */
1443 lseek(fd, 0, 0);
056b331e 1444 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1445 goto abort;
1446
1447 /* check MBP signature */
056b331e 1448 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1449 retval = 1;
1450 /* found the correct signature */
034b203a 1451
ca3b6696 1452 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
8e2bca51
JC
1453 /*
1454 * Have to make every access through boot_sect rather
1455 * than using a pointer to the partition table (or an
1456 * entry), since the entries are not properly aligned.
1457 */
1458
034b203a 1459 /* check for GPT type */
8e2bca51
JC
1460 if (boot_sect.parts[part_nr].part_type ==
1461 MBR_GPT_PARTITION_TYPE) {
034b203a
TM
1462 retval = get_gpt_last_partition_end(fd, endofpart);
1463 break;
1464 }
1465 /* check the last used lba for the current partition */
8e2bca51
JC
1466 curr_part_end =
1467 __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) +
1468 __le32_to_cpu(boot_sect.parts[part_nr].blocks_num);
034b203a
TM
1469 if (curr_part_end > *endofpart)
1470 *endofpart = curr_part_end;
034b203a
TM
1471 }
1472 } else {
1473 /* Unknown partition table */
1474 retval = -1;
1475 }
31208db9
MD
1476 /* calculate number of 512-byte blocks */
1477 if (get_dev_sector_size(fd, NULL, &sector_size))
1478 *endofpart *= (sector_size / 512);
034b203a
TM
1479 abort:
1480 return retval;
1481}
1482
53ed6ac3
KW
1483int check_partitions(int fd, char *dname, unsigned long long freesize,
1484 unsigned long long size)
034b203a
TM
1485{
1486 /*
1487 * Check where the last partition ends
1488 */
1489 unsigned long long endofpart;
034b203a 1490
31208db9 1491 if (get_last_partition_end(fd, &endofpart) > 0) {
034b203a
TM
1492 /* There appears to be a partition table here */
1493 if (freesize == 0) {
1494 /* partitions will not be visible in new device */
e7b84f9d
N
1495 pr_err("partition table exists on %s but will be lost or\n"
1496 " meaningless after creating array\n",
1497 dname);
034b203a
TM
1498 return 1;
1499 } else if (endofpart > freesize) {
1500 /* last partition overlaps metadata */
e7b84f9d
N
1501 pr_err("metadata will over-write last partition on %s.\n",
1502 dname);
034b203a 1503 return 1;
53ed6ac3
KW
1504 } else if (size && endofpart > size) {
1505 /* partitions will be truncated in new device */
e7b84f9d
N
1506 pr_err("array size is too small to cover all partitions on %s.\n",
1507 dname);
53ed6ac3 1508 return 1;
034b203a
TM
1509 }
1510 }
1511 return 0;
1512}
1513
a322f70c
DW
1514int open_container(int fd)
1515{
1516 /* 'fd' is a block device. Find out if it is in use
1517 * by a container, and return an open fd on that container.
1518 */
1519 char path[256];
1520 char *e;
1521 DIR *dir;
1522 struct dirent *de;
1523 int dfd, n;
1524 char buf[200];
1525 int major, minor;
1526 struct stat st;
1527
1528 if (fstat(fd, &st) != 0)
1529 return -1;
1530 sprintf(path, "/sys/dev/block/%d:%d/holders",
1531 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1532 e = path + strlen(path);
1533
1534 dir = opendir(path);
1535 if (!dir)
1536 return -1;
1537 while ((de = readdir(dir))) {
1538 if (de->d_ino == 0)
1539 continue;
1540 if (de->d_name[0] == '.')
1541 continue;
a994592d
N
1542 /* Need to make sure it is a container and not a volume */
1543 sprintf(e, "/%s/md/metadata_version", de->d_name);
1544 dfd = open(path, O_RDONLY);
1545 if (dfd < 0)
1546 continue;
1547 n = read(dfd, buf, sizeof(buf));
1548 close(dfd);
1549 if (n <= 0 || (unsigned)n >= sizeof(buf))
1550 continue;
1551 buf[n] = 0;
1552 if (strncmp(buf, "external", 8) != 0 ||
1553 n < 10 ||
1554 buf[9] == '/')
1555 continue;
a322f70c
DW
1556 sprintf(e, "/%s/dev", de->d_name);
1557 dfd = open(path, O_RDONLY);
1558 if (dfd < 0)
1559 continue;
1560 n = read(dfd, buf, sizeof(buf));
1561 close(dfd);
f21e18ca 1562 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1563 continue;
1564 buf[n] = 0;
1565 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1566 continue;
1567 sprintf(buf, "%d:%d", major, minor);
1568 dfd = dev_open(buf, O_RDONLY);
1569 if (dfd >= 0) {
1570 closedir(dir);
1571 return dfd;
1572 }
1573 }
355726fa 1574 closedir(dir);
a322f70c
DW
1575 return -1;
1576}
1577
33414a01
DW
1578struct superswitch *version_to_superswitch(char *vers)
1579{
1580 int i;
1581
1582 for (i = 0; superlist[i]; i++) {
1583 struct superswitch *ss = superlist[i];
1584
1585 if (strcmp(vers, ss->name) == 0)
1586 return ss;
1587 }
1588
1589 return NULL;
1590}
1591
4dd2df09
N
1592int metadata_container_matches(char *metadata, char *devnm)
1593{
1594 /* Check if 'devnm' is the container named in 'metadata'
1595 * which is
1596 * /containername/componentname or
1597 * -containername/componentname
1598 */
1599 int l;
1600 if (*metadata != '/' && *metadata != '-')
1601 return 0;
1602 l = strlen(devnm);
1603 if (strncmp(metadata+1, devnm, l) != 0)
1604 return 0;
1605 if (metadata[l+1] != '/')
1606 return 0;
1607 return 1;
1608}
1609
1610int metadata_subdev_matches(char *metadata, char *devnm)
1611{
1612 /* Check if 'devnm' is the subdev named in 'metadata'
1613 * which is
1614 * /containername/subdev or
1615 * -containername/subdev
1616 */
1617 char *sl;
1618 if (*metadata != '/' && *metadata != '-')
1619 return 0;
1620 sl = strchr(metadata+1, '/');
1621 if (!sl)
1622 return 0;
1623 if (strcmp(sl+1, devnm) == 0)
1624 return 1;
1625 return 0;
1626}
1627
33414a01
DW
1628int is_container_member(struct mdstat_ent *mdstat, char *container)
1629{
1630 if (mdstat->metadata_version == NULL ||
1631 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1632 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1633 return 0;
1634
1635 return 1;
1636}
1637
1638int is_subarray_active(char *subarray, char *container)
1639{
1640 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1641 struct mdstat_ent *ent;
1642
e5408a32
DW
1643 for (ent = mdstat; ent; ent = ent->next)
1644 if (is_container_member(ent, container))
e5e5d7ce 1645 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1646 break;
33414a01
DW
1647
1648 free_mdstat(mdstat);
1649
1650 return ent != NULL;
1651}
1652
1653/* open_subarray - opens a subarray in a container
1654 * @dev: container device name
feab51f8 1655 * @st: empty supertype
33414a01
DW
1656 * @quiet: block reporting errors flag
1657 *
1658 * On success returns an fd to a container and fills in *st
1659 */
feab51f8 1660int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1661{
1662 struct mdinfo *mdi;
a951a4f7 1663 struct mdinfo *info;
33414a01 1664 int fd, err = 1;
4dd2df09 1665 char *_devnm;
33414a01
DW
1666
1667 fd = open(dev, O_RDWR|O_EXCL);
1668 if (fd < 0) {
1669 if (!quiet)
e7b84f9d 1670 pr_err("Couldn't open %s, aborting\n",
33414a01 1671 dev);
b990032d 1672 return -1;
33414a01
DW
1673 }
1674
4dd2df09
N
1675 _devnm = fd2devnm(fd);
1676 if (_devnm == NULL) {
33414a01 1677 if (!quiet)
e7b84f9d
N
1678 pr_err("Failed to determine device number for %s\n",
1679 dev);
33414a01
DW
1680 goto close_fd;
1681 }
4dd2df09 1682 strcpy(st->devnm, _devnm);
33414a01 1683
4dd2df09 1684 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1685 if (!mdi) {
1686 if (!quiet)
e7b84f9d 1687 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1688 dev);
1689 goto close_fd;
1690 }
1691
1692 if (mdi->array.level != UnSet) {
1693 if (!quiet)
e7b84f9d 1694 pr_err("%s is not a container\n", dev);
33414a01
DW
1695 goto free_sysfs;
1696 }
1697
1698 st->ss = version_to_superswitch(mdi->text_version);
1699 if (!st->ss) {
1700 if (!quiet)
e7b84f9d
N
1701 pr_err("Operation not supported for %s metadata\n",
1702 mdi->text_version);
33414a01
DW
1703 goto free_sysfs;
1704 }
1705
4dd2df09 1706 if (st->devnm[0] == 0) {
33414a01 1707 if (!quiet)
e7b84f9d 1708 pr_err("Failed to allocate device name\n");
33414a01
DW
1709 goto free_sysfs;
1710 }
1711
db20d413 1712 if (!st->ss->load_container) {
33414a01 1713 if (!quiet)
e7b84f9d 1714 pr_err("%s is not a container\n", dev);
4dd2df09 1715 goto free_sysfs;
33414a01
DW
1716 }
1717
db20d413 1718 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1719 if (!quiet)
e7b84f9d 1720 pr_err("Failed to load metadata for %s\n",
db20d413 1721 dev);
4dd2df09 1722 goto free_sysfs;
33414a01
DW
1723 }
1724
a951a4f7
N
1725 info = st->ss->container_content(st, subarray);
1726 if (!info) {
1727 if (!quiet)
e7b84f9d 1728 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1729 subarray, dev);
1730 goto free_super;
1731 }
1732 free(info);
1733
33414a01
DW
1734 err = 0;
1735
1736 free_super:
1737 if (err)
1738 st->ss->free_super(st);
33414a01
DW
1739 free_sysfs:
1740 sysfs_free(mdi);
1741 close_fd:
1742 if (err)
1743 close(fd);
1744
1745 if (err)
1746 return -1;
1747 else
1748 return fd;
1749}
1750
7801ac20
N
1751int add_disk(int mdfd, struct supertype *st,
1752 struct mdinfo *sra, struct mdinfo *info)
1753{
1754 /* Add a device to an array, in one of 2 ways. */
1755 int rv;
1756#ifndef MDASSEMBLE
1757 if (st->ss->external) {
d23534e4
DW
1758 if (info->disk.state & (1<<MD_DISK_SYNC))
1759 info->recovery_start = MaxSector;
1760 else
1761 info->recovery_start = 0;
2904b26f 1762 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1763 if (! rv) {
1764 struct mdinfo *sd2;
f35f2525
N
1765 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1766 if (sd2 == info)
1767 break;
1768 if (sd2 == NULL) {
503975b9 1769 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1770 *sd2 = *info;
1771 sd2->next = sra->devs;
1772 sra->devs = sd2;
1773 }
7801ac20
N
1774 }
1775 } else
1776#endif
1777 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1778 return rv;
1779}
1780
de6ae750
N
1781int remove_disk(int mdfd, struct supertype *st,
1782 struct mdinfo *sra, struct mdinfo *info)
1783{
1784 int rv;
1785 /* Remove the disk given by 'info' from the array */
1786#ifndef MDASSEMBLE
1787 if (st->ss->external)
1788 rv = sysfs_set_str(sra, info, "slot", "none");
1789 else
1790#endif
1791 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1792 info->disk.minor));
1793 return rv;
1794}
1795
1ab9ed2a 1796int hot_remove_disk(int mdfd, unsigned long dev, int force)
2dd271fe 1797{
1ab9ed2a 1798 int cnt = force ? 500 : 5;
2dd271fe
N
1799 int ret;
1800
1801 /* HOT_REMOVE_DISK can fail with EBUSY if there are
1802 * outstanding IO requests to the device.
1803 * In this case, it can be helpful to wait a little while,
1ab9ed2a 1804 * up to 5 seconds if 'force' is set, or 50 msec if not.
2dd271fe
N
1805 */
1806 while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 &&
1807 errno == EBUSY &&
1808 cnt-- > 0)
1809 usleep(10000);
1810
1811 return ret;
1812}
1813
1ab9ed2a 1814int sys_hot_remove_disk(int statefd, int force)
fdd01569 1815{
1ab9ed2a 1816 int cnt = force ? 500 : 5;
fdd01569
N
1817 int ret;
1818
1819 while ((ret = write(statefd, "remove", 6)) == -1 &&
1820 errno == EBUSY &&
1821 cnt-- > 0)
1822 usleep(10000);
1823 return ret == 6 ? 0 : -1;
1824}
1825
f35f2525
N
1826int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1827{
1828 /* Initialise kernel's knowledge of array.
1829 * This varies between externally managed arrays
1830 * and older kernels
1831 */
700483a2 1832 mdu_array_info_t inf;
f35f2525
N
1833 int rv;
1834
1835#ifndef MDASSEMBLE
1836 if (st->ss->external)
700483a2 1837 return sysfs_set_array(info, 9003);
f35f2525 1838#endif
700483a2
JS
1839
1840 memset(&inf, 0, sizeof(inf));
1841 inf.major_version = info->array.major_version;
1842 inf.minor_version = info->array.minor_version;
1843 rv = md_set_array_info(mdfd, &inf);
1844
f35f2525
N
1845 return rv;
1846}
1847
1e5c6983
DW
1848unsigned long long min_recovery_start(struct mdinfo *array)
1849{
1850 /* find the minimum recovery_start in an array for metadata
1851 * formats that only record per-array recovery progress instead
1852 * of per-device
1853 */
1854 unsigned long long recovery_start = MaxSector;
1855 struct mdinfo *d;
1856
1857 for (d = array->devs; d; d = d->next)
1858 recovery_start = min(recovery_start, d->recovery_start);
1859
1860 return recovery_start;
1861}
1862
4dd2df09 1863int mdmon_pid(char *devnm)
a931db9e
NB
1864{
1865 char path[100];
1866 char pid[10];
1867 int fd;
1868 int n;
10013317 1869
4dd2df09 1870 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1871
24f6f99b 1872 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1873
1874 if (fd < 0)
cf556303 1875 return -1;
a931db9e
NB
1876 n = read(fd, pid, 9);
1877 close(fd);
1878 if (n <= 0)
cf556303 1879 return -1;
24f6f99b 1880 return atoi(pid);
a931db9e
NB
1881}
1882
4dd2df09 1883int mdmon_running(char *devnm)
a931db9e 1884{
4dd2df09 1885 int pid = mdmon_pid(devnm);
24f6f99b 1886 if (pid <= 0)
a931db9e 1887 return 0;
24f6f99b 1888 if (kill(pid, 0) == 0)
a931db9e
NB
1889 return 1;
1890 return 0;
1891}
1892
4dd2df09 1893int start_mdmon(char *devnm)
8850ee3e 1894{
1913c325 1895 int i, skipped;
44d2e365 1896 int len;
ca3b6696 1897 pid_t pid;
9fe32043 1898 int status;
44d2e365
N
1899 char pathbuf[1024];
1900 char *paths[4] = {
1901 pathbuf,
85945e19 1902 BINDIR "/mdmon",
2f1bcf43 1903 "./mdmon",
44d2e365
N
1904 NULL
1905 };
8850ee3e 1906
40ebbb9c 1907 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1908 return 0;
1909
9cf014ec 1910 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1911 if (len > 0) {
1912 char *sl;
1913 pathbuf[len] = 0;
1914 sl = strrchr(pathbuf, '/');
1915 if (sl)
1916 sl++;
1917 else
1918 sl = pathbuf;
1919 strcpy(sl, "mdmon");
1920 } else
1921 pathbuf[0] = '\0';
1922
0f7bdf89 1923 /* First try to run systemctl */
9540cc24
N
1924 if (!check_env("MDADM_NO_SYSTEMCTL"))
1925 switch(fork()) {
1926 case 0:
1927 /* FIXME yuk. CLOSE_EXEC?? */
1928 skipped = 0;
1929 for (i = 3; skipped < 20; i++)
1930 if (close(i) < 0)
1931 skipped++;
1932 else
1933 skipped = 0;
1934
1935 /* Don't want to see error messages from
1936 * systemctl. If the service doesn't exist,
1937 * we start mdmon ourselves.
1938 */
1939 close(2);
1940 open("/dev/null", O_WRONLY);
1941 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1942 devnm);
1943 status = execl("/usr/bin/systemctl", "systemctl",
1944 "start",
1945 pathbuf, NULL);
1946 status = execl("/bin/systemctl", "systemctl", "start",
1947 pathbuf, NULL);
1948 exit(1);
7a862a02 1949 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1950 return -1;
1951 default: /* parent - good */
1952 pid = wait(&status);
1953 if (pid >= 0 && status == 0)
1954 return 0;
1955 }
0f7bdf89
JS
1956
1957 /* That failed, try running mdmon directly */
8850ee3e
N
1958 switch(fork()) {
1959 case 0:
1960 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1961 skipped = 0;
ca3b6696 1962 for (i = 3; skipped < 20; i++)
1913c325
N
1963 if (close(i) < 0)
1964 skipped++;
1965 else
1966 skipped = 0;
1967
ca3b6696 1968 for (i = 0; paths[i]; i++)
a0963a86 1969 if (paths[i][0]) {
2f1bcf43 1970 execl(paths[i], paths[i],
4dd2df09 1971 devnm, NULL);
a0963a86 1972 }
8850ee3e 1973 exit(1);
7a862a02 1974 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 1975 return -1;
9fe32043
N
1976 default: /* parent - good */
1977 pid = wait(&status);
15c10423 1978 if (pid < 0 || status != 0) {
7a862a02 1979 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 1980 return -1;
15c10423 1981 }
8850ee3e
N
1982 }
1983 return 0;
1984}
1985
148acb7b
DW
1986__u32 random32(void)
1987{
1988 __u32 rv;
1989 int rfd = open("/dev/urandom", O_RDONLY);
1990 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1991 rv = random();
1992 if (rfd >= 0)
1993 close(rfd);
1994 return rv;
1995}
1996
c5f71c24
JS
1997void random_uuid(__u8 *buf)
1998{
1999 int fd, i, len;
2000 __u32 r[4];
2001
2002 fd = open("/dev/urandom", O_RDONLY);
2003 if (fd < 0)
2004 goto use_random;
2005 len = read(fd, buf, 16);
2006 close(fd);
2007 if (len != 16)
2008 goto use_random;
2009
2010 return;
2011
2012use_random:
2013 for (i = 0; i < 4; i++)
2014 r[i] = random();
2015 memcpy(buf, r, 16);
2016}
2017
0e600426 2018#ifndef MDASSEMBLE
edd8d13c
NB
2019int flush_metadata_updates(struct supertype *st)
2020{
2021 int sfd;
2022 if (!st->updates) {
2023 st->update_tail = NULL;
2024 return -1;
2025 }
2026
4dd2df09 2027 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
2028 if (sfd < 0)
2029 return -1;
2030
2031 while (st->updates) {
2032 struct metadata_update *mu = st->updates;
2033 st->updates = mu->next;
2034
2035 send_message(sfd, mu, 0);
2036 wait_reply(sfd, 0);
2037 free(mu->buf);
2038 free(mu);
2039 }
2040 ack(sfd, 0);
2041 wait_reply(sfd, 0);
2042 close(sfd);
2043 st->update_tail = NULL;
2044 return 0;
2045}
2046
2047void append_metadata_update(struct supertype *st, void *buf, int len)
2048{
2049
503975b9 2050 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
2051
2052 mu->buf = buf;
2053 mu->len = len;
2054 mu->space = NULL;
cb23f1f4 2055 mu->space_list = NULL;
edd8d13c
NB
2056 mu->next = NULL;
2057 *st->update_tail = mu;
2058 st->update_tail = &mu->next;
2059}
0e600426 2060#endif /* MDASSEMBLE */
a931db9e 2061
fe6729fa
NB
2062#ifdef __TINYC__
2063/* tinyc doesn't optimize this check in ioctl.h out ... */
2064unsigned int __invalid_size_argument_for_IOC = 0;
2065#endif
2066
6d11ec6f
AK
2067int experimental(void)
2068{
2069 if (check_env("MDADM_EXPERIMENTAL"))
2070 return 1;
2071 else {
7a862a02 2072 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
2073 return 0;
2074 }
2075}
2076
326727d9
AC
2077/* Pick all spares matching given criteria from a container
2078 * if min_size == 0 do not check size
2079 * if domlist == NULL do not check domains
2080 * if spare_group given add it to domains of each spare
2081 * metadata allows to test domains using metadata of destination array */
2082struct mdinfo *container_choose_spares(struct supertype *st,
2083 unsigned long long min_size,
2084 struct domainlist *domlist,
2085 char *spare_group,
2086 const char *metadata, int get_one)
2087{
2088 struct mdinfo *d, **dp, *disks = NULL;
2089
2090 /* get list of all disks in container */
2091 if (st->ss->getinfo_super_disks)
2092 disks = st->ss->getinfo_super_disks(st);
2093
2094 if (!disks)
2095 return disks;
2096 /* find spare devices on the list */
2097 dp = &disks->devs;
2098 disks->array.spare_disks = 0;
2099 while (*dp) {
2100 int found = 0;
2101 d = *dp;
2102 if (d->disk.state == 0) {
2103 /* check if size is acceptable */
2104 unsigned long long dev_size;
2105 dev_t dev = makedev(d->disk.major,d->disk.minor);
2106
2107 if (!min_size ||
2108 (dev_size_from_id(dev, &dev_size) &&
2109 dev_size >= min_size))
2110 found = 1;
2111 /* check if domain matches */
2112 if (found && domlist) {
4dd2df09 2113 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2114 if (spare_group)
2115 pol_add(&pol, pol_domain,
2116 spare_group, NULL);
e5508b36 2117 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2118 found = 0;
2119 dev_policy_free(pol);
2120 }
2121 }
2122 if (found) {
2123 dp = &d->next;
2124 disks->array.spare_disks++;
2125 if (get_one) {
2126 sysfs_free(*dp);
2127 d->next = NULL;
2128 }
2129 } else {
2130 *dp = d->next;
2131 d->next = NULL;
2132 sysfs_free(d);
2133 }
2134 }
2135 return disks;
2136}
9eafa1de
MN
2137
2138/* Checks if paths point to the same device
2139 * Returns 0 if they do.
2140 * Returns 1 if they don't.
2141 * Returns -1 if something went wrong,
2142 * e.g. paths are empty or the files
2143 * they point to don't exist */
2144int compare_paths (char* path1, char* path2)
2145{
2146 struct stat st1,st2;
2147
2148 if (path1 == NULL || path2 == NULL)
2149 return -1;
2150 if (stat(path1,&st1) != 0)
2151 return -1;
2152 if (stat(path2,&st2) != 0)
2153 return -1;
2154 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2155 return 0;
2156 return 1;
2157}
a7dec3fd
N
2158
2159/* Make sure we can open as many devices as needed */
2160void enable_fds(int devices)
2161{
2162 unsigned int fds = 20 + devices;
2163 struct rlimit lim;
2164 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
2165 || lim.rlim_cur >= fds)
2166 return;
2167 if (lim.rlim_max < fds)
2168 lim.rlim_max = fds;
2169 lim.rlim_cur = fds;
2170 setrlimit(RLIMIT_NOFILE, &lim);
2171}
a9c15847
N
2172
2173int in_initrd(void)
2174{
2175 /* This is based on similar function in systemd. */
2176 struct statfs s;
76d0f188
JS
2177 /* statfs.f_type is signed long on s390x and MIPS, causing all
2178 sorts of sign extension problems with RAMFS_MAGIC being
2179 defined as 0x858458f6 */
a9c15847 2180 return statfs("/", &s) >= 0 &&
7ac5d47e 2181 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2182 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2183 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2184}
8832342d
N
2185
2186void reopen_mddev(int mdfd)
2187{
2188 /* Re-open without any O_EXCL, but keep
2189 * the same fd
2190 */
2191 char *devnm;
2192 int fd;
2193 devnm = fd2devnm(mdfd);
2194 close(mdfd);
2195 fd = open_dev(devnm);
2196 if (fd >= 0 && fd != mdfd)
2197 dup2(fd, mdfd);
2198}
7716570e 2199
7071320a 2200#ifndef MDASSEMBLE
e80357f8
GJ
2201static struct cmap_hooks *cmap_hooks = NULL;
2202static int is_cmap_hooks_ready = 0;
7716570e 2203
e80357f8
GJ
2204void set_cmap_hooks(void)
2205{
2206 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2207 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2208 if (!cmap_hooks->cmap_handle)
2209 return;
7716570e 2210
e80357f8
GJ
2211 cmap_hooks->initialize = dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2212 cmap_hooks->get_string = dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
2213 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2214
e80357f8
GJ
2215 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2216 !cmap_hooks->finalize)
2217 dlclose(cmap_hooks->cmap_handle);
2218 else
2219 is_cmap_hooks_ready = 1;
2220}
7716570e 2221
e80357f8
GJ
2222int get_cluster_name(char **cluster_name)
2223{
2224 int rv = -1;
2225 cmap_handle_t handle;
7716570e 2226
e80357f8
GJ
2227 if (!is_cmap_hooks_ready)
2228 return rv;
7716570e 2229
e80357f8 2230 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2231 if (rv != CS_OK)
2232 goto out;
2233
e80357f8 2234 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2235 if (rv != CS_OK) {
2236 free(*cluster_name);
2237 rv = -1;
2238 goto name_err;
2239 }
2240
2241 rv = 0;
2242name_err:
e80357f8 2243 cmap_hooks->finalize(handle);
7716570e 2244out:
7716570e
GJ
2245 return rv;
2246}
d15a1f72
GJ
2247
2248void set_dlm_hooks(void)
2249{
2250 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2251 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2252 if (!dlm_hooks->dlm_handle)
2253 return;
2254
2255 dlm_hooks->create_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2256 dlm_hooks->release_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
2257 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2258 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2259 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2260 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2261
2262 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2263 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2264 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2265 dlclose(dlm_hooks->dlm_handle);
2266 else
2267 is_dlm_hooks_ready = 1;
2268}
e80357f8
GJ
2269
2270void set_hooks(void)
2271{
2272 set_dlm_hooks();
2273 set_cmap_hooks();
2274}
7d55dca2 2275#endif