util/must_be_container: Use sysfs_read(GET_VERSION) to determine valid array
[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
276/*
277 * Get the md version number.
278 * We use the RAID_VERSION ioctl if it is supported
279 * If not, but we have a block device with major '9', we assume
280 * 0.36.0
281 *
282 * Return version number as 24 but number - assume version parts
283 * always < 255
284 */
285
286int md_get_version(int fd)
287{
5d500228
N
288 struct stat stb;
289 mdu_version_t vers;
290
291 if (fstat(fd, &stb)<0)
292 return -1;
293 if ((S_IFMT&stb.st_mode) != S_IFBLK)
294 return -1;
295
296 if (ioctl(fd, RAID_VERSION, &vers) == 0)
297 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
298 if (errno == EACCES)
299 return -1;
300 if (major(stb.st_rdev) == MD_MAJOR)
301 return (3600);
64c4757e 302 return -1;
64c4757e
NB
303}
304
64c4757e
NB
305int get_linux_version()
306{
307 struct utsname name;
98c6faba 308 char *cp;
f161d047 309 int a = 0, b = 0,c = 0;
64c4757e
NB
310 if (uname(&name) <0)
311 return -1;
312
98c6faba
NB
313 cp = name.release;
314 a = strtoul(cp, &cp, 10);
f161d047
N
315 if (*cp == '.')
316 b = strtoul(cp+1, &cp, 10);
317 if (*cp == '.')
318 c = strtoul(cp+1, &cp, 10);
98c6faba 319
682c7051 320 return (a*1000000)+(b*1000)+c;
64c4757e
NB
321}
322
71204a50 323#ifndef MDASSEMBLE
bc77ed53
DW
324int mdadm_version(char *version)
325{
326 int a, b, c;
327 char *cp;
328
329 if (!version)
330 version = Version;
331
332 cp = strchr(version, '-');
333 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
334 return -1;
335 cp += 3;
336 a = strtoul(cp, &cp, 10);
337 if (*cp != '.')
338 return -1;
339 b = strtoul(cp+1, &cp, 10);
340 if (*cp == '.')
341 c = strtoul(cp+1, &cp, 10);
342 else
343 c = 0;
344 if (*cp != ' ' && *cp != '-')
345 return -1;
346 return (a*1000000)+(b*1000)+c;
347}
348
51d4261c 349unsigned long long parse_size(char *size)
84e11361
N
350{
351 /* parse 'size' which should be a number optionally
352 * followed by 'K', 'M', or 'G'.
353 * Without a suffix, K is assumed.
354 * Number returned is in sectors (half-K)
822e393a 355 * INVALID_SECTORS returned on error.
84e11361
N
356 */
357 char *c;
358 long long s = strtoll(size, &c, 10);
359 if (s > 0) {
360 switch (*c) {
361 case 'K':
362 c++;
363 default:
364 s *= 2;
365 break;
366 case 'M':
367 c++;
368 s *= 1024 * 2;
369 break;
370 case 'G':
371 c++;
372 s *= 1024 * 1024 * 2;
373 break;
72ca9bcf
N
374 case 's': /* sectors */
375 c++;
376 break;
84e11361 377 }
51d4261c 378 } else
822e393a 379 s = INVALID_SECTORS;
84e11361 380 if (*c)
822e393a 381 s = INVALID_SECTORS;
84e11361
N
382 return s;
383}
384
4a06e2c2
N
385int parse_layout_10(char *layout)
386{
387 int copies, rv;
388 char *cp;
389 /* Parse the layout string for raid10 */
390 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
391 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
392 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
393 copies > 200 ||
394 *cp)
395 return -1;
396 if (layout[0] == 'n')
397 rv = 256 + copies;
398 else if (layout[0] == 'o')
399 rv = 0x10000 + (copies<<8) + 1;
400 else
401 rv = 1 + (copies<<8);
402 return rv;
403}
404
405int parse_layout_faulty(char *layout)
406{
407 /* Parse the layout string for 'faulty' */
408 int ln = strcspn(layout, "0123456789");
503975b9 409 char *m = xstrdup(layout);
4a06e2c2
N
410 int mode;
411 m[ln] = 0;
412 mode = map_name(faultylayout, m);
413 if (mode == UnSet)
414 return -1;
415
416 return mode | (atoi(layout+ln)<< ModeShift);
417}
79868890
N
418
419long parse_num(char *num)
420{
421 /* Either return a valid number, or -1 */
422 char *c;
423 long rv = strtol(num, &c, 10);
424 if (rv < 0 || *c || !num[0])
425 return -1;
426 else
427 return rv;
428}
eb3929a4 429#endif
4a06e2c2 430
4de90913
GJ
431int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
432{
433 char *dev;
434 *slot = strtoul(input, &dev, 10);
435 if (dev == input || dev[0] != ':')
436 return -1;
437 *devname = dev+1;
438 return 0;
439}
440
0430ed48
NB
441void remove_partitions(int fd)
442{
443 /* remove partitions from this block devices.
444 * This is used for components added to an array
445 */
446#ifdef BLKPG_DEL_PARTITION
447 struct blkpg_ioctl_arg a;
448 struct blkpg_partition p;
449
450 a.op = BLKPG_DEL_PARTITION;
451 a.data = (void*)&p;
452 a.datalen = sizeof(p);
453 a.flags = 0;
454 memset(a.data, 0, a.datalen);
ca3b6696 455 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
456 ioctl(fd, BLKPG, &a);
457#endif
458}
459
691c6ee1
N
460int test_partition(int fd)
461{
462 /* Check if fd is a whole-disk or a partition.
463 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
464 * will return ENXIO on an invalid partition number.
465 */
466 struct blkpg_ioctl_arg a;
467 struct blkpg_partition p;
468 a.op = BLKPG_DEL_PARTITION;
469 a.data = (void*)&p;
470 a.datalen = sizeof(p);
471 a.flags = 0;
472 memset(a.data, 0, a.datalen);
473 p.pno = 1<<30;
474 if (ioctl(fd, BLKPG, &a) == 0)
475 /* Very unlikely, but not a partition */
476 return 0;
357ac106 477 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
478 /* not a partition */
479 return 0;
480
481 return 1;
482}
483
bfd76b93
CA
484int test_partition_from_id(dev_t id)
485{
486 char buf[20];
487 int fd, rv;
488
489 sprintf(buf, "%d:%d", major(id), minor(id));
490 fd = dev_open(buf, O_RDONLY);
491 if (fd < 0)
492 return -1;
493 rv = test_partition(fd);
494 close(fd);
495 return rv;
496}
691c6ee1 497
de5a472e 498int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 499{
265e0f17 500 int copies, first;
de5a472e
N
501 int i;
502 int avail_disks = 0;
503
504 for (i = 0; i < raid_disks; i++)
505 avail_disks += !!avail[i];
506
64c4757e 507 switch (level) {
265e0f17
NB
508 case 10:
509 /* This is the tricky one - we need to check
510 * which actual disks are present.
511 */
702b557b 512 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 513 first = 0;
265e0f17
NB
514 do {
515 /* there must be one of the 'copies' form 'first' */
516 int n = copies;
ca3b6696 517 int cnt = 0;
b3ec716d 518 int this = first;
265e0f17 519 while (n--) {
b3ec716d 520 if (avail[this])
265e0f17 521 cnt++;
b3ec716d 522 this = (this+1) % raid_disks;
265e0f17
NB
523 }
524 if (cnt == 0)
525 return 0;
b3ec716d 526 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
527 } while (first != 0);
528 return 1;
e5329c37 529
df0d4ea0 530 case LEVEL_MULTIPATH:
e0d19036 531 return avail_disks>= 1;
df0d4ea0 532 case LEVEL_LINEAR:
64c4757e
NB
533 case 0:
534 return avail_disks == raid_disks;
535 case 1:
536 return avail_disks >= 1;
537 case 4:
330d6900
N
538 if (avail_disks == raid_disks - 1 &&
539 !avail[raid_disks - 1])
540 /* If just the parity device is missing, then we
541 * have enough, even if not clean
542 */
543 return 1;
544 /* FALL THROUGH */
64c4757e 545 case 5:
583315d9
NB
546 if (clean)
547 return avail_disks >= raid_disks-1;
548 else
549 return avail_disks >= raid_disks;
98c6faba 550 case 6:
583315d9
NB
551 if (clean)
552 return avail_disks >= raid_disks-2;
553 else
554 return avail_disks >= raid_disks;
64c4757e
NB
555 default:
556 return 0;
557 }
558}
559
8453e704
N
560int enough_fd(int fd)
561{
562 struct mdu_array_info_s array;
563 struct mdu_disk_info_s disk;
aabe020d 564 int i, rv;
8453e704
N
565 char *avail;
566
9cd39f01 567 if (md_get_array_info(fd, &array) != 0 || array.raid_disks <= 0)
8453e704 568 return 0;
503975b9 569 avail = xcalloc(array.raid_disks, 1);
ca3b6696 570 for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704 571 disk.number = i;
d97572f5 572 if (md_get_disk_info(fd, &disk) != 0)
8453e704 573 continue;
9e6d9291
N
574 if (disk.major == 0 && disk.minor == 0)
575 continue;
576 array.nr_disks--;
577
8453e704
N
578 if (! (disk.state & (1<<MD_DISK_SYNC)))
579 continue;
580 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
581 continue;
8453e704
N
582 avail[disk.raid_disk] = 1;
583 }
584 /* This is used on an active array, so assume it is clean */
aabe020d 585 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 586 1, avail);
aabe020d
JS
587 free(avail);
588 return rv;
8453e704
N
589}
590
22e263f6
AC
591const int uuid_zero[4] = { 0, 0, 0, 0 };
592
f277ce36 593int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 594{
f277ce36
NB
595 if (swapuuid) {
596 /* parse uuids are hostendian.
597 * uuid's from some superblocks are big-ending
aba69144 598 * if there is a difference, we need to swap..
f277ce36
NB
599 */
600 unsigned char *ac = (unsigned char *)a;
601 unsigned char *bc = (unsigned char *)b;
602 int i;
ca3b6696 603 for (i = 0; i < 16; i += 4) {
f277ce36
NB
604 if (ac[i+0] != bc[i+3] ||
605 ac[i+1] != bc[i+2] ||
606 ac[i+2] != bc[i+1] ||
607 ac[i+3] != bc[i+0])
608 return 0;
609 }
610 return 1;
611 } else {
612 if (a[0]==b[0] &&
613 a[1]==b[1] &&
614 a[2]==b[2] &&
615 a[3]==b[3])
616 return 1;
617 return 0;
618 }
64c4757e 619}
ca3b6696 620
350f29f9
NB
621void copy_uuid(void *a, int b[4], int swapuuid)
622{
623 if (swapuuid) {
624 /* parse uuids are hostendian.
625 * uuid's from some superblocks are big-ending
626 * if there is a difference, we need to swap..
627 */
628 unsigned char *ac = (unsigned char *)a;
629 unsigned char *bc = (unsigned char *)b;
630 int i;
ca3b6696 631 for (i = 0; i < 16; i += 4) {
350f29f9
NB
632 ac[i+0] = bc[i+3];
633 ac[i+1] = bc[i+2];
634 ac[i+2] = bc[i+1];
635 ac[i+3] = bc[i+0];
636 }
637 } else
638 memcpy(a, b, 16);
639}
64c4757e 640
aae5a112 641char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 642{
9968e376 643 int i, j;
d7288ddc
N
644 char uuid[16];
645 char *c = buf;
646 strcpy(c, "UUID-");
647 c += strlen(c);
aae5a112 648 copy_uuid(uuid, id, swap);
9968e376 649 for (i = 0; i < 4; i++) {
9968e376 650 if (i)
ff54de6e 651 *c++ = sep;
9968e376
DW
652 for (j = 3; j >= 0; j--) {
653 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
654 c+= 2;
655 }
d7288ddc
N
656 }
657 return buf;
aae5a112
DW
658
659}
660
661char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
662{
86983cce
N
663 // dirty hack to work around an issue with super1 superblocks...
664 // super1 superblocks need swapuuid set in order for assembly to
665 // work, but can't have it set if we want this printout to match
666 // all the other uuid printouts in super1.c, so we force swapuuid
667 // to 1 to make our printout match the rest of super1
668 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
669}
670
435d4ebb 671#ifndef MDASSEMBLE
682c7051
NB
672int check_ext2(int fd, char *name)
673{
674 /*
675 * Check for an ext2fs file system.
676 * Superblock is always 1K at 1K offset
677 *
678 * s_magic is le16 at 56 == 0xEF53
679 * report mtime - le32 at 44
680 * blocks - le32 at 4
681 * logblksize - le32 at 24
682 */
683 unsigned char sb[1024];
684 time_t mtime;
2538ba2a
N
685 unsigned long long size;
686 int bsize;
682c7051
NB
687 if (lseek(fd, 1024,0)!= 1024)
688 return 0;
689 if (read(fd, sb, 1024)!= 1024)
690 return 0;
691 if (sb[56] != 0x53 || sb[57] != 0xef)
692 return 0;
693
694 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
695 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
696 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 697 size <<= bsize;
e7b84f9d 698 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 699 name);
2538ba2a 700 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
701 return 1;
702}
703
704int check_reiser(int fd, char *name)
705{
706 /*
707 * superblock is at 64K
708 * size is 1024;
709 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
710 *
711 */
712 unsigned char sb[1024];
2538ba2a 713 unsigned long long size;
682c7051
NB
714 if (lseek(fd, 64*1024, 0) != 64*1024)
715 return 0;
716 if (read(fd, sb, 1024) != 1024)
717 return 0;
ca3b6696
N
718 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
719 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 720 return 0;
e7b84f9d 721 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 722 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 723 cont_err("size = %lluK\n", size*4);
aba69144 724
682c7051
NB
725 return 1;
726}
727
728int check_raid(int fd, char *name)
729{
4b1ac34b 730 struct mdinfo info;
682c7051 731 time_t crtime;
d078d77c 732 char *level;
82d9eba6 733 struct supertype *st = guess_super(fd);
f9ce90ba 734
cb8f6859
N
735 if (!st)
736 return 0;
fa219dd2
MD
737 if (st->ss->add_to_super != NULL) {
738 st->ss->load_super(st, fd, name);
739 /* Looks like a raid array .. */
740 pr_err("%s appears to be part of a raid array:\n", name);
741 st->ss->getinfo_super(st, &info, NULL);
742 st->ss->free_super(st);
743 crtime = info.array.ctime;
744 level = map_num(pers, info.array.level);
745 if (!level)
746 level = "-unknown-";
747 cont_err("level=%s devices=%d ctime=%s",
748 level, info.array.raid_disks, ctime(&crtime));
749 } else {
750 /* Looks like GPT or MBR */
751 pr_err("partition table exists on %s\n", name);
752 }
82d9eba6 753 return 1;
682c7051
NB
754}
755
682c7051
NB
756int ask(char *mesg)
757{
758 char *add = "";
759 int i;
ca3b6696 760 for (i = 0; i < 5; i++) {
682c7051
NB
761 char buf[100];
762 fprintf(stderr, "%s%s", mesg, add);
763 fflush(stderr);
764 if (fgets(buf, 100, stdin)==NULL)
765 return 0;
766 if (buf[0]=='y' || buf[0]=='Y')
767 return 1;
768 if (buf[0]=='n' || buf[0]=='N')
769 return 0;
770 add = "(y/n) ";
771 }
e7b84f9d 772 pr_err("assuming 'no'\n");
682c7051
NB
773 return 0;
774}
435d4ebb 775#endif /* MDASSEMBLE */
682c7051 776
8d80900b 777int is_standard(char *dev, int *nump)
e5329c37
NB
778{
779 /* tests if dev is a "standard" md dev name.
780 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 781 * where NN is a string of digits
598f0d58
NB
782 * Returns 1 if a partitionable standard,
783 * -1 if non-partitonable,
784 * 0 if not a standard name.
e5329c37 785 */
8d80900b 786 char *d = strrchr(dev, '/');
ca3b6696 787 int type = 0;
8d80900b
NB
788 int num;
789 if (!d)
e5329c37 790 return 0;
ca3b6696
N
791 if (strncmp(d, "/d",2) == 0)
792 d += 2, type = 1; /* /dev/md/dN{pM} */
793 else if (strncmp(d, "/md_d", 5) == 0)
794 d += 5, type = 1; /* /dev/md_dN{pM} */
795 else if (strncmp(d, "/md", 3) == 0)
796 d += 3, type = -1; /* /dev/mdN */
797 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
798 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
799 else
800 return 0;
8d80900b 801 if (!*d)
e5329c37 802 return 0;
8d80900b
NB
803 num = atoi(d);
804 while (isdigit(*d))
805 d++;
806 if (*d)
e5329c37 807 return 0;
8d80900b
NB
808 if (nump) *nump = num;
809
810 return type;
e5329c37
NB
811}
812
4b1ac34b 813unsigned long calc_csum(void *super, int bytes)
82b27616 814{
56eb10c0 815 unsigned long long newcsum = 0;
82b27616 816 int i;
4b1ac34b
NB
817 unsigned int csum;
818 unsigned int *superc = (unsigned int*) super;
82b27616 819
ca3b6696
N
820 for(i = 0; i < bytes/4; i++)
821 newcsum += superc[i];
82b27616 822 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 823#ifdef __alpha__
aba69144 824/* The in-kernel checksum calculation is always 16bit on
570c0542 825 * the alpha, though it is 32 bit on i386...
ca3b6696 826 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
827 * a way that it shouldn't).
828 */
829 csum = (csum & 0xffff) + (csum >> 16);
830 csum = (csum & 0xffff) + (csum >> 16);
831#endif
82b27616
NB
832 return csum;
833}
cd29a5c8 834
435d4ebb 835#ifndef MDASSEMBLE
56eb10c0 836char *human_size(long long bytes)
cd29a5c8 837{
ff9239ee 838 static char buf[47];
d5d3721e
NB
839
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.
01619b48 846 * Terabytes are not yet handled.
d5d3721e 847 */
cd29a5c8 848
56eb10c0 849 if (bytes < 5000*1024)
ca3b6696 850 buf[0] = 0;
d5d3721e 851 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 852 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 853 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 854 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
efa29530 855 cMiB/100, cMiB % 100, cMB/100, cMB % 100);
d5d3721e 856 } else {
93d3bd3b 857 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 858 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 859 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
efa29530 860 cGiB/100, cGiB % 100, cGB/100, cGB % 100);
d5d3721e 861 }
cd29a5c8
NB
862 return buf;
863}
e0d19036 864
f0ec6710 865char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
866{
867 static char buf[30];
e0d19036 868
570abc6f
MN
869 /* We convert bytes to either centi-M{ega,ibi}bytes or
870 * centi-G{igi,ibi}bytes, with appropriate rounding,
871 * and then print 1/100th of those as a decimal.
872 * We allow upto 2048Megabytes before converting to
873 * gigabytes, as that shows more precision and isn't
874 * too large a number.
875 * Terabytes are not yet handled.
f0ec6710
MN
876 *
877 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
878 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
879 */
880
e0d19036 881 if (bytes < 5000*1024)
570abc6f 882 buf[0] = 0;
f0ec6710
MN
883 else if (prefix == IEC) {
884 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 885 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710 886 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
efa29530 887 cMiB/100, cMiB % 100);
f0ec6710 888 } else {
93d3bd3b 889 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710 890 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
efa29530 891 cGiB/100, cGiB % 100);
f0ec6710
MN
892 }
893 }
894 else if (prefix == JEDEC) {
895 if (bytes < 2*1024LL*1024LL*1024LL) {
896 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
897 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
efa29530 898 cMB/100, cMB % 100);
f0ec6710
MN
899 } else {
900 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
901 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
efa29530 902 cGB/100, cGB % 100);
f0ec6710 903 }
570abc6f 904 }
f0ec6710
MN
905 else
906 buf[0] = 0;
570abc6f 907
e0d19036
NB
908 return buf;
909}
e4965ef8
N
910
911void print_r10_layout(int layout)
912{
913 int near = layout & 255;
914 int far = (layout >> 8) & 255;
915 int offset = (layout&0x10000);
916 char *sep = "";
917
918 if (near != 1) {
919 printf("%s near=%d", sep, near);
920 sep = ",";
921 }
922 if (far != 1)
923 printf("%s %s=%d", sep, offset?"offset":"far", far);
924 if (near*far == 1)
925 printf("NO REDUNDANCY");
926}
435d4ebb 927#endif
e0d19036 928
5f8097be
NB
929unsigned long long calc_array_size(int level, int raid_disks, int layout,
930 int chunksize, unsigned long long devsize)
577e8448 931{
e03640bd
CA
932 if (level == 1)
933 return devsize;
577e8448
AK
934 devsize &= ~(unsigned long long)((chunksize>>9)-1);
935 return get_data_disks(level, layout, raid_disks) * devsize;
936}
937
938int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
939{
940 int data_disks = 0;
941 switch (level) {
ca3b6696
N
942 case 0: data_disks = raid_disks;
943 break;
944 case 1: data_disks = 1;
945 break;
5f8097be 946 case 4:
ca3b6696
N
947 case 5: data_disks = raid_disks - 1;
948 break;
949 case 6: data_disks = raid_disks - 2;
950 break;
5f8097be
NB
951 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
952 break;
953 }
577e8448
AK
954
955 return data_disks;
5f8097be
NB
956}
957
13db17bd 958dev_t devnm2devid(char *devnm)
4dd2df09
N
959{
960 /* First look in /sys/block/$DEVNM/dev for %d:%d
961 * If that fails, try parsing out a number
962 */
963 char path[100];
964 char *ep;
965 int fd;
966 int mjr,mnr;
967
968 sprintf(path, "/sys/block/%s/dev", devnm);
969 fd = open(path, O_RDONLY);
970 if (fd >= 0) {
971 char buf[20];
972 int n = read(fd, buf, sizeof(buf));
973 close(fd);
974 if (n > 0)
975 buf[n] = 0;
976 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
977 return makedev(mjr, mnr);
978 }
979 if (strncmp(devnm, "md_d", 4) == 0 &&
980 isdigit(devnm[4]) &&
981 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
982 ep > devnm && *ep == 0)
983 return makedev(get_mdp_major(), mnr << MdpMinorShift);
984
985 if (strncmp(devnm, "md", 2) == 0 &&
986 isdigit(devnm[2]) &&
987 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
988 ep > devnm && *ep == 0)
989 return makedev(MD_MAJOR, mnr);
990
991 return 0;
992}
993
5a23a06e 994#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 995char *get_md_name(char *devnm)
e0d19036
NB
996{
997 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 998 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 999
e0d19036
NB
1000 static char devname[50];
1001 struct stat stb;
4dd2df09 1002 dev_t rdev = devnm2devid(devnm);
dd0781e5 1003 char *dn;
98c6faba 1004
4dd2df09
N
1005 if (rdev == 0)
1006 return 0;
1007 if (strncmp(devnm, "md_", 3) == 0) {
1008 snprintf(devname, sizeof(devname), "/dev/md/%s",
1009 devnm + 3);
98c6faba
NB
1010 if (stat(devname, &stb) == 0
1011 && (S_IFMT&stb.st_mode) == S_IFBLK
1012 && (stb.st_rdev == rdev))
1013 return devname;
1014 }
4dd2df09
N
1015 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
1016 if (stat(devname, &stb) == 0
1017 && (S_IFMT&stb.st_mode) == S_IFBLK
1018 && (stb.st_rdev == rdev))
1019 return devname;
1020
1021 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
1022 if (stat(devname, &stb) == 0
1023 && (S_IFMT&stb.st_mode) == S_IFBLK
1024 && (stb.st_rdev == rdev))
1025 return devname;
1026
16c6fa80 1027 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
1028 if (dn)
1029 return dn;
4dd2df09 1030 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 1031 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
1032 if (errno != EEXIST)
1033 return NULL;
e0d19036
NB
1034
1035 if (stat(devname, &stb) == 0
1036 && (S_IFMT&stb.st_mode) == S_IFBLK
1037 && (stb.st_rdev == rdev))
1038 return devname;
1039 unlink(devname);
1040 return NULL;
1041}
1042
1043void put_md_name(char *name)
1044{
ca3b6696 1045 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
1046 unlink(name);
1047}
435d4ebb 1048#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 1049
cc742d38
N
1050int get_maj_min(char *dev, int *major, int *minor)
1051{
1052 char *e;
1053 *major = strtoul(dev, &e, 0);
1054 return (e > dev && *e == ':' && e[1] &&
1055 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1056 *e == 0);
1057}
1058
8b0dabea
NB
1059int dev_open(char *dev, int flags)
1060{
1061 /* like 'open', but if 'dev' matches %d:%d, create a temp
1062 * block device and open that
1063 */
8b0dabea
NB
1064 int fd = -1;
1065 char devname[32];
e81cdd9f 1066 int major;
8b0dabea 1067 int minor;
e81cdd9f 1068
9f0ad56b
JS
1069 if (!dev)
1070 return -1;
6df6a774 1071 flags |= O_DIRECT;
e81cdd9f 1072
cc742d38 1073 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1074 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1075 (int)getpid(), major, minor);
1076 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1077 fd = open(devname, flags);
1078 unlink(devname);
6df6a774
N
1079 }
1080 if (fd < 0) {
9dc7d357 1081 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
1082 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
1083 (int)getpid(), major, minor);
ca3b6696 1084 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
1085 fd = open(devname, flags);
1086 unlink(devname);
1087 }
8b0dabea
NB
1088 }
1089 } else
6df6a774 1090 fd = open(dev, flags);
8b0dabea
NB
1091 return fd;
1092}
f9ce90ba 1093
4dd2df09 1094int open_dev_flags(char *devnm, int flags)
e8a70c89 1095{
13db17bd 1096 dev_t devid;
e8a70c89
N
1097 char buf[20];
1098
4dd2df09
N
1099 devid = devnm2devid(devnm);
1100 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1101 return dev_open(buf, flags);
1102}
1103
4dd2df09 1104int open_dev(char *devnm)
d998b738 1105{
4dd2df09 1106 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1107}
1108
4dd2df09 1109int open_dev_excl(char *devnm)
a931db9e
NB
1110{
1111 char buf[20];
1112 int i;
7187750e 1113 int flags = O_RDWR;
13db17bd 1114 dev_t devid = devnm2devid(devnm);
399e0b97 1115 long delay = 1000;
a931db9e 1116
4dd2df09 1117 sprintf(buf, "%d:%d", major(devid), minor(devid));
efa29530 1118 for (i = 0; i < 25; i++) {
7187750e 1119 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1120 if (fd >= 0)
1121 return fd;
7187750e
N
1122 if (errno == EACCES && flags == O_RDWR) {
1123 flags = O_RDONLY;
1124 continue;
1125 }
a931db9e
NB
1126 if (errno != EBUSY)
1127 return fd;
399e0b97
N
1128 usleep(delay);
1129 if (delay < 200000)
1130 delay *= 2;
a931db9e
NB
1131 }
1132 return -1;
1133}
1134
9008ed1c
N
1135int same_dev(char *one, char *two)
1136{
1137 struct stat st1, st2;
1138 if (stat(one, &st1) != 0)
1139 return 0;
1140 if (stat(two, &st2) != 0)
1141 return 0;
1142 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1143 return 0;
1144 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1145 return 0;
1146 return st1.st_rdev == st2.st_rdev;
1147}
1148
a7c6e3fb 1149void wait_for(char *dev, int fd)
a714580e
N
1150{
1151 int i;
a7c6e3fb 1152 struct stat stb_want;
399e0b97 1153 long delay = 1000;
a7c6e3fb
N
1154
1155 if (fstat(fd, &stb_want) != 0 ||
1156 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1157 return;
a714580e 1158
efa29530 1159 for (i = 0; i < 25; i++) {
a714580e 1160 struct stat stb;
a7c6e3fb
N
1161 if (stat(dev, &stb) == 0 &&
1162 (stb.st_mode & S_IFMT) == S_IFBLK &&
1163 (stb.st_rdev == stb_want.st_rdev))
a714580e 1164 return;
399e0b97
N
1165 usleep(delay);
1166 if (delay < 200000)
1167 delay *= 2;
a714580e 1168 }
436305c6 1169 if (i == 25)
1ade5cc1 1170 dprintf("timeout waiting for %s\n", dev);
a714580e
N
1171}
1172
0f22b998
N
1173struct superswitch *superlist[] =
1174{
1175 &super0, &super1,
1176 &super_ddf, &super_imsm,
0592faeb 1177 &mbr, &gpt,
0f22b998 1178 NULL };
f9ce90ba 1179
ea24acd0 1180#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1181
4725bc31 1182struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1183{
1686dc25
NB
1184 mdu_array_info_t array;
1185 int vers;
1186 int minor;
1187 struct supertype *st = NULL;
7e0f6979 1188 struct mdinfo *sra;
142cb9e1 1189 char *verstr;
1686dc25
NB
1190 char version[20];
1191 int i;
f7e7067b 1192 char *subarray = NULL;
4dd2df09 1193 char container[32] = "";
1686dc25 1194
4dd2df09 1195 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1196
1197 if (sra) {
7e0f6979
NB
1198 vers = sra->array.major_version;
1199 minor = sra->array.minor_version;
142cb9e1 1200 verstr = sra->text_version;
1686dc25 1201 } else {
9cd39f01 1202 if (md_get_array_info(fd, &array))
1686dc25
NB
1203 array.major_version = array.minor_version = 0;
1204 vers = array.major_version;
1205 minor = array.minor_version;
142cb9e1 1206 verstr = "";
6fbba4c9 1207 }
82d9eba6 1208
1686dc25
NB
1209 if (vers != -1) {
1210 sprintf(version, "%d.%d", vers, minor);
1211 verstr = version;
6fbba4c9 1212 }
3c558363 1213 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1214 char *dev = verstr+1;
5f7e44b2 1215
f7e7067b 1216 subarray = strchr(dev, '/');
92d49ecf 1217 if (subarray) {
f7e7067b 1218 *subarray++ = '\0';
503975b9 1219 subarray = xstrdup(subarray);
92d49ecf 1220 }
4dd2df09 1221 strcpy(container, dev);
15d230f7 1222 sysfs_free(sra);
5f7e44b2 1223 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1224 if (sra && sra->text_version[0])
1225 verstr = sra->text_version;
1226 else
1227 verstr = "-no-metadata-";
f7e7067b
NB
1228 }
1229
efa29530 1230 for (i = 0; st == NULL && superlist[i]; i++)
f7e7067b 1231 st = superlist[i]->match_metadata_desc(verstr);
1686dc25 1232
15d230f7 1233 sysfs_free(sra);
f7e7067b 1234 if (st) {
3b0896f8 1235 st->sb = NULL;
1f49fb3a
N
1236 if (subarrayp)
1237 *subarrayp = subarray;
4dd2df09
N
1238 strcpy(st->container_devnm, container);
1239 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1240 } else
1241 free(subarray);
5f7e44b2 1242
82d9eba6 1243 return st;
f9ce90ba 1244}
ea24acd0
NB
1245#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1246
0f0749ad 1247int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1248{
1249 char buf[20];
1250 int fd;
1251
1252 sprintf(buf, "%d:%d", major(id), minor(id));
1253 fd = dev_open(buf, O_RDONLY);
1254 if (fd < 0)
1255 return 0;
1256 if (get_dev_size(fd, NULL, size)) {
1257 close(fd);
1258 return 1;
1259 }
1260 close(fd);
1261 return 0;
1262}
f9ce90ba 1263
159c3a1a 1264struct supertype *dup_super(struct supertype *orig)
3da92f27 1265{
159c3a1a 1266 struct supertype *st;
1686dc25 1267
d2ca6449
NB
1268 if (!orig)
1269 return orig;
503975b9 1270 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1271 st->ss = orig->ss;
1272 st->max_devs = orig->max_devs;
1273 st->minor_version = orig->minor_version;
cb8f6859 1274 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1275 st->data_offset = orig->data_offset;
159c3a1a
NB
1276 st->sb = NULL;
1277 st->info = NULL;
1278 return st;
3da92f27
NB
1279}
1280
54887ad8 1281struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1282{
1283 /* try each load_super to find the best match,
1284 * and return the best superswitch
1285 */
82d9eba6
NB
1286 struct superswitch *ss;
1287 struct supertype *st;
1d13b599 1288 unsigned int besttime = 0;
570c0542 1289 int bestsuper = -1;
f9ce90ba
NB
1290 int i;
1291
503975b9 1292 st = xcalloc(1, sizeof(*st));
4dd2df09 1293 st->container_devnm[0] = 0;
d1d599ea 1294
efa29530 1295 for (i = 0; superlist[i]; i++) {
f9ce90ba
NB
1296 int rv;
1297 ss = superlist[i];
54887ad8
N
1298 if (guess_type == guess_array && ss->add_to_super == NULL)
1299 continue;
1300 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1301 continue;
ef609477 1302 memset(st, 0, sizeof(*st));
df3346e6 1303 st->ignore_hw_compat = 1;
3da92f27 1304 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1305 if (rv == 0) {
1306 struct mdinfo info;
a5d85af7 1307 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1308 if (bestsuper == -1 ||
1309 besttime < info.array.ctime) {
1310 bestsuper = i;
1311 besttime = info.array.ctime;
570c0542 1312 }
3da92f27 1313 ss->free_super(st);
570c0542
NB
1314 }
1315 }
1316 if (bestsuper != -1) {
1317 int rv;
ef609477 1318 memset(st, 0, sizeof(*st));
df3346e6 1319 st->ignore_hw_compat = 1;
3da92f27 1320 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1321 if (rv == 0) {
5e747af2 1322 superlist[bestsuper]->free_super(st);
82d9eba6 1323 return st;
f9ce90ba
NB
1324 }
1325 }
570c0542 1326 free(st);
f9ce90ba
NB
1327 return NULL;
1328}
fe6729fa 1329
beae1dfe
NB
1330/* Return size of device in bytes */
1331int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1332{
1333 unsigned long long ldsize;
c2c9bb6f
NB
1334 struct stat st;
1335
1336 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1337 ldsize = (unsigned long long)st.st_size;
1338 else
beae1dfe
NB
1339#ifdef BLKGETSIZE64
1340 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1341#endif
1342 {
1343 unsigned long dsize;
1344 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1345 ldsize = dsize;
1346 ldsize <<= 9;
1347 } else {
1348 if (dname)
32971509 1349 pr_err("Cannot get size of %s: %s\n",
beae1dfe
NB
1350 dname, strerror(errno));
1351 return 0;
1352 }
1353 }
1354 *sizep = ldsize;
1355 return 1;
1356}
8fac0577 1357
32971509
PB
1358/* Return sector size of device in bytes */
1359int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep)
1360{
1361 unsigned int sectsize;
1362
1363 if (ioctl(fd, BLKSSZGET, &sectsize) != 0) {
1364 if (dname)
1365 pr_err("Cannot get sector size of %s: %s\n",
1366 dname, strerror(errno));
1367 return 0;
1368 }
1369
1370 *sectsizep = sectsize;
1371 return 1;
1372}
1373
3a371610
N
1374/* Return true if this can only be a container, not a member device.
1375 * i.e. is and md device and size is zero
1376 */
1377int must_be_container(int fd)
1378{
f5c924f4 1379 struct mdinfo *mdi;
3a371610 1380 unsigned long long size;
f5c924f4
JS
1381
1382 mdi = sysfs_read(fd, NULL, GET_VERSION);
1383 if (!mdi)
3a371610 1384 return 0;
f5c924f4
JS
1385 sysfs_free(mdi);
1386
3a371610
N
1387 if (get_dev_size(fd, NULL, &size) == 0)
1388 return 1;
1389 if (size == 0)
1390 return 1;
1391 return 0;
1392}
034b203a
TM
1393
1394/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1395 * Returns: 1 if successful
1396 * -1 for unknown partition type
1397 * 0 for other errors
1398 */
1399static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1400{
056b331e 1401 struct GPT gpt;
034b203a 1402 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1403 struct GPT_part_entry *part;
1404 char buf[512];
034b203a
TM
1405 unsigned long long curr_part_end;
1406 unsigned all_partitions, entry_size;
f21e18ca 1407 unsigned part_nr;
41b06495 1408 unsigned int sector_size = 0;
034b203a
TM
1409
1410 *endofpart = 0;
1411
056b331e 1412 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1413 /* skip protective MBR */
41b06495
MD
1414 if (!get_dev_sector_size(fd, NULL, &sector_size))
1415 return 0;
1416 lseek(fd, sector_size, SEEK_SET);
73e658d8 1417 /* read GPT header */
056b331e 1418 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1419 return 0;
1420
1421 /* get the number of partition entries and the entry size */
056b331e
N
1422 all_partitions = __le32_to_cpu(gpt.part_cnt);
1423 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1424
1425 /* Check GPT signature*/
056b331e 1426 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1427 return -1;
1428
1429 /* sanity checks */
1430 if (all_partitions > 1024 ||
db7fdfe4 1431 entry_size > sizeof(buf))
034b203a
TM
1432 return -1;
1433
db7fdfe4
JS
1434 part = (struct GPT_part_entry *)buf;
1435
41b06495
MD
1436 /* set offset to third block (GPT entries) */
1437 lseek(fd, sector_size*2, SEEK_SET);
ca3b6696 1438 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1439 /* read partition entry */
db7fdfe4 1440 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1441 return 0;
1442
034b203a 1443 /* is this valid partition? */
db7fdfe4 1444 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1445 /* check the last lba for the current partition */
db7fdfe4 1446 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1447 if (curr_part_end > *endofpart)
1448 *endofpart = curr_part_end;
1449 }
1450
034b203a
TM
1451 }
1452 return 1;
1453}
1454
1455/* Sets endofpart parameter to the last block used by the last partition on the device.
1456 * Returns: 1 if successful
1457 * -1 for unknown partition type
1458 * 0 for other errors
1459 */
1460static int get_last_partition_end(int fd, unsigned long long *endofpart)
1461{
056b331e 1462 struct MBR boot_sect;
034b203a 1463 unsigned long long curr_part_end;
f21e18ca 1464 unsigned part_nr;
31208db9 1465 unsigned int sector_size;
034b203a
TM
1466 int retval = 0;
1467
1468 *endofpart = 0;
1469
056b331e 1470 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1471 /* read MBR */
1472 lseek(fd, 0, 0);
056b331e 1473 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1474 goto abort;
1475
1476 /* check MBP signature */
056b331e 1477 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1478 retval = 1;
1479 /* found the correct signature */
034b203a 1480
ca3b6696 1481 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
8e2bca51
JC
1482 /*
1483 * Have to make every access through boot_sect rather
1484 * than using a pointer to the partition table (or an
1485 * entry), since the entries are not properly aligned.
1486 */
1487
034b203a 1488 /* check for GPT type */
8e2bca51
JC
1489 if (boot_sect.parts[part_nr].part_type ==
1490 MBR_GPT_PARTITION_TYPE) {
034b203a
TM
1491 retval = get_gpt_last_partition_end(fd, endofpart);
1492 break;
1493 }
1494 /* check the last used lba for the current partition */
8e2bca51
JC
1495 curr_part_end =
1496 __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) +
1497 __le32_to_cpu(boot_sect.parts[part_nr].blocks_num);
034b203a
TM
1498 if (curr_part_end > *endofpart)
1499 *endofpart = curr_part_end;
034b203a
TM
1500 }
1501 } else {
1502 /* Unknown partition table */
1503 retval = -1;
1504 }
31208db9
MD
1505 /* calculate number of 512-byte blocks */
1506 if (get_dev_sector_size(fd, NULL, &sector_size))
1507 *endofpart *= (sector_size / 512);
034b203a
TM
1508 abort:
1509 return retval;
1510}
1511
53ed6ac3
KW
1512int check_partitions(int fd, char *dname, unsigned long long freesize,
1513 unsigned long long size)
034b203a
TM
1514{
1515 /*
1516 * Check where the last partition ends
1517 */
1518 unsigned long long endofpart;
034b203a 1519
31208db9 1520 if (get_last_partition_end(fd, &endofpart) > 0) {
034b203a
TM
1521 /* There appears to be a partition table here */
1522 if (freesize == 0) {
1523 /* partitions will not be visible in new device */
e7b84f9d
N
1524 pr_err("partition table exists on %s but will be lost or\n"
1525 " meaningless after creating array\n",
1526 dname);
034b203a
TM
1527 return 1;
1528 } else if (endofpart > freesize) {
1529 /* last partition overlaps metadata */
e7b84f9d
N
1530 pr_err("metadata will over-write last partition on %s.\n",
1531 dname);
034b203a 1532 return 1;
53ed6ac3
KW
1533 } else if (size && endofpart > size) {
1534 /* partitions will be truncated in new device */
e7b84f9d
N
1535 pr_err("array size is too small to cover all partitions on %s.\n",
1536 dname);
53ed6ac3 1537 return 1;
034b203a
TM
1538 }
1539 }
1540 return 0;
1541}
1542
a322f70c
DW
1543int open_container(int fd)
1544{
1545 /* 'fd' is a block device. Find out if it is in use
1546 * by a container, and return an open fd on that container.
1547 */
1548 char path[256];
1549 char *e;
1550 DIR *dir;
1551 struct dirent *de;
1552 int dfd, n;
1553 char buf[200];
1554 int major, minor;
1555 struct stat st;
1556
1557 if (fstat(fd, &st) != 0)
1558 return -1;
1559 sprintf(path, "/sys/dev/block/%d:%d/holders",
1560 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1561 e = path + strlen(path);
1562
1563 dir = opendir(path);
1564 if (!dir)
1565 return -1;
1566 while ((de = readdir(dir))) {
1567 if (de->d_ino == 0)
1568 continue;
1569 if (de->d_name[0] == '.')
1570 continue;
a994592d
N
1571 /* Need to make sure it is a container and not a volume */
1572 sprintf(e, "/%s/md/metadata_version", de->d_name);
1573 dfd = open(path, O_RDONLY);
1574 if (dfd < 0)
1575 continue;
1576 n = read(dfd, buf, sizeof(buf));
1577 close(dfd);
1578 if (n <= 0 || (unsigned)n >= sizeof(buf))
1579 continue;
1580 buf[n] = 0;
1581 if (strncmp(buf, "external", 8) != 0 ||
1582 n < 10 ||
1583 buf[9] == '/')
1584 continue;
a322f70c
DW
1585 sprintf(e, "/%s/dev", de->d_name);
1586 dfd = open(path, O_RDONLY);
1587 if (dfd < 0)
1588 continue;
1589 n = read(dfd, buf, sizeof(buf));
1590 close(dfd);
f21e18ca 1591 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1592 continue;
1593 buf[n] = 0;
1594 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1595 continue;
1596 sprintf(buf, "%d:%d", major, minor);
1597 dfd = dev_open(buf, O_RDONLY);
1598 if (dfd >= 0) {
1599 closedir(dir);
1600 return dfd;
1601 }
1602 }
355726fa 1603 closedir(dir);
a322f70c
DW
1604 return -1;
1605}
1606
33414a01
DW
1607struct superswitch *version_to_superswitch(char *vers)
1608{
1609 int i;
1610
1611 for (i = 0; superlist[i]; i++) {
1612 struct superswitch *ss = superlist[i];
1613
1614 if (strcmp(vers, ss->name) == 0)
1615 return ss;
1616 }
1617
1618 return NULL;
1619}
1620
4dd2df09
N
1621int metadata_container_matches(char *metadata, char *devnm)
1622{
1623 /* Check if 'devnm' is the container named in 'metadata'
1624 * which is
1625 * /containername/componentname or
1626 * -containername/componentname
1627 */
1628 int l;
1629 if (*metadata != '/' && *metadata != '-')
1630 return 0;
1631 l = strlen(devnm);
1632 if (strncmp(metadata+1, devnm, l) != 0)
1633 return 0;
1634 if (metadata[l+1] != '/')
1635 return 0;
1636 return 1;
1637}
1638
1639int metadata_subdev_matches(char *metadata, char *devnm)
1640{
1641 /* Check if 'devnm' is the subdev named in 'metadata'
1642 * which is
1643 * /containername/subdev or
1644 * -containername/subdev
1645 */
1646 char *sl;
1647 if (*metadata != '/' && *metadata != '-')
1648 return 0;
1649 sl = strchr(metadata+1, '/');
1650 if (!sl)
1651 return 0;
1652 if (strcmp(sl+1, devnm) == 0)
1653 return 1;
1654 return 0;
1655}
1656
33414a01
DW
1657int is_container_member(struct mdstat_ent *mdstat, char *container)
1658{
1659 if (mdstat->metadata_version == NULL ||
1660 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1661 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1662 return 0;
1663
1664 return 1;
1665}
1666
1667int is_subarray_active(char *subarray, char *container)
1668{
1669 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1670 struct mdstat_ent *ent;
1671
e5408a32
DW
1672 for (ent = mdstat; ent; ent = ent->next)
1673 if (is_container_member(ent, container))
e5e5d7ce 1674 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1675 break;
33414a01
DW
1676
1677 free_mdstat(mdstat);
1678
1679 return ent != NULL;
1680}
1681
1682/* open_subarray - opens a subarray in a container
1683 * @dev: container device name
feab51f8 1684 * @st: empty supertype
33414a01
DW
1685 * @quiet: block reporting errors flag
1686 *
1687 * On success returns an fd to a container and fills in *st
1688 */
feab51f8 1689int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1690{
1691 struct mdinfo *mdi;
a951a4f7 1692 struct mdinfo *info;
33414a01 1693 int fd, err = 1;
4dd2df09 1694 char *_devnm;
33414a01
DW
1695
1696 fd = open(dev, O_RDWR|O_EXCL);
1697 if (fd < 0) {
1698 if (!quiet)
e7b84f9d 1699 pr_err("Couldn't open %s, aborting\n",
33414a01 1700 dev);
b990032d 1701 return -1;
33414a01
DW
1702 }
1703
4dd2df09
N
1704 _devnm = fd2devnm(fd);
1705 if (_devnm == NULL) {
33414a01 1706 if (!quiet)
e7b84f9d
N
1707 pr_err("Failed to determine device number for %s\n",
1708 dev);
33414a01
DW
1709 goto close_fd;
1710 }
4dd2df09 1711 strcpy(st->devnm, _devnm);
33414a01 1712
4dd2df09 1713 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1714 if (!mdi) {
1715 if (!quiet)
e7b84f9d 1716 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1717 dev);
1718 goto close_fd;
1719 }
1720
1721 if (mdi->array.level != UnSet) {
1722 if (!quiet)
e7b84f9d 1723 pr_err("%s is not a container\n", dev);
33414a01
DW
1724 goto free_sysfs;
1725 }
1726
1727 st->ss = version_to_superswitch(mdi->text_version);
1728 if (!st->ss) {
1729 if (!quiet)
e7b84f9d
N
1730 pr_err("Operation not supported for %s metadata\n",
1731 mdi->text_version);
33414a01
DW
1732 goto free_sysfs;
1733 }
1734
4dd2df09 1735 if (st->devnm[0] == 0) {
33414a01 1736 if (!quiet)
e7b84f9d 1737 pr_err("Failed to allocate device name\n");
33414a01
DW
1738 goto free_sysfs;
1739 }
1740
db20d413 1741 if (!st->ss->load_container) {
33414a01 1742 if (!quiet)
e7b84f9d 1743 pr_err("%s is not a container\n", dev);
4dd2df09 1744 goto free_sysfs;
33414a01
DW
1745 }
1746
db20d413 1747 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1748 if (!quiet)
e7b84f9d 1749 pr_err("Failed to load metadata for %s\n",
db20d413 1750 dev);
4dd2df09 1751 goto free_sysfs;
33414a01
DW
1752 }
1753
a951a4f7
N
1754 info = st->ss->container_content(st, subarray);
1755 if (!info) {
1756 if (!quiet)
e7b84f9d 1757 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1758 subarray, dev);
1759 goto free_super;
1760 }
1761 free(info);
1762
33414a01
DW
1763 err = 0;
1764
1765 free_super:
1766 if (err)
1767 st->ss->free_super(st);
33414a01
DW
1768 free_sysfs:
1769 sysfs_free(mdi);
1770 close_fd:
1771 if (err)
1772 close(fd);
1773
1774 if (err)
1775 return -1;
1776 else
1777 return fd;
1778}
1779
7801ac20
N
1780int add_disk(int mdfd, struct supertype *st,
1781 struct mdinfo *sra, struct mdinfo *info)
1782{
1783 /* Add a device to an array, in one of 2 ways. */
1784 int rv;
1785#ifndef MDASSEMBLE
1786 if (st->ss->external) {
d23534e4
DW
1787 if (info->disk.state & (1<<MD_DISK_SYNC))
1788 info->recovery_start = MaxSector;
1789 else
1790 info->recovery_start = 0;
2904b26f 1791 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1792 if (! rv) {
1793 struct mdinfo *sd2;
f35f2525
N
1794 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1795 if (sd2 == info)
1796 break;
1797 if (sd2 == NULL) {
503975b9 1798 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1799 *sd2 = *info;
1800 sd2->next = sra->devs;
1801 sra->devs = sd2;
1802 }
7801ac20
N
1803 }
1804 } else
1805#endif
1806 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1807 return rv;
1808}
1809
de6ae750
N
1810int remove_disk(int mdfd, struct supertype *st,
1811 struct mdinfo *sra, struct mdinfo *info)
1812{
1813 int rv;
1814 /* Remove the disk given by 'info' from the array */
1815#ifndef MDASSEMBLE
1816 if (st->ss->external)
1817 rv = sysfs_set_str(sra, info, "slot", "none");
1818 else
1819#endif
1820 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1821 info->disk.minor));
1822 return rv;
1823}
1824
1ab9ed2a 1825int hot_remove_disk(int mdfd, unsigned long dev, int force)
2dd271fe 1826{
1ab9ed2a 1827 int cnt = force ? 500 : 5;
2dd271fe
N
1828 int ret;
1829
1830 /* HOT_REMOVE_DISK can fail with EBUSY if there are
1831 * outstanding IO requests to the device.
1832 * In this case, it can be helpful to wait a little while,
1ab9ed2a 1833 * up to 5 seconds if 'force' is set, or 50 msec if not.
2dd271fe
N
1834 */
1835 while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 &&
1836 errno == EBUSY &&
1837 cnt-- > 0)
1838 usleep(10000);
1839
1840 return ret;
1841}
1842
1ab9ed2a 1843int sys_hot_remove_disk(int statefd, int force)
fdd01569 1844{
1ab9ed2a 1845 int cnt = force ? 500 : 5;
fdd01569
N
1846 int ret;
1847
1848 while ((ret = write(statefd, "remove", 6)) == -1 &&
1849 errno == EBUSY &&
1850 cnt-- > 0)
1851 usleep(10000);
1852 return ret == 6 ? 0 : -1;
1853}
1854
f35f2525
N
1855int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1856{
1857 /* Initialise kernel's knowledge of array.
1858 * This varies between externally managed arrays
1859 * and older kernels
1860 */
1861 int vers = md_get_version(mdfd);
1862 int rv;
1863
1864#ifndef MDASSEMBLE
1865 if (st->ss->external)
1866 rv = sysfs_set_array(info, vers);
1867 else
1868#endif
1869 if ((vers % 100) >= 1) { /* can use different versions */
1870 mdu_array_info_t inf;
1871 memset(&inf, 0, sizeof(inf));
1872 inf.major_version = info->array.major_version;
1873 inf.minor_version = info->array.minor_version;
018a4882 1874 rv = md_set_array_info(mdfd, &inf);
f35f2525 1875 } else
018a4882 1876 rv = md_set_array_info(mdfd, NULL);
f35f2525
N
1877 return rv;
1878}
1879
1e5c6983
DW
1880unsigned long long min_recovery_start(struct mdinfo *array)
1881{
1882 /* find the minimum recovery_start in an array for metadata
1883 * formats that only record per-array recovery progress instead
1884 * of per-device
1885 */
1886 unsigned long long recovery_start = MaxSector;
1887 struct mdinfo *d;
1888
1889 for (d = array->devs; d; d = d->next)
1890 recovery_start = min(recovery_start, d->recovery_start);
1891
1892 return recovery_start;
1893}
1894
4dd2df09 1895int mdmon_pid(char *devnm)
a931db9e
NB
1896{
1897 char path[100];
1898 char pid[10];
1899 int fd;
1900 int n;
10013317 1901
4dd2df09 1902 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1903
24f6f99b 1904 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1905
1906 if (fd < 0)
cf556303 1907 return -1;
a931db9e
NB
1908 n = read(fd, pid, 9);
1909 close(fd);
1910 if (n <= 0)
cf556303 1911 return -1;
24f6f99b 1912 return atoi(pid);
a931db9e
NB
1913}
1914
4dd2df09 1915int mdmon_running(char *devnm)
a931db9e 1916{
4dd2df09 1917 int pid = mdmon_pid(devnm);
24f6f99b 1918 if (pid <= 0)
a931db9e 1919 return 0;
24f6f99b 1920 if (kill(pid, 0) == 0)
a931db9e
NB
1921 return 1;
1922 return 0;
1923}
1924
4dd2df09 1925int start_mdmon(char *devnm)
8850ee3e 1926{
1913c325 1927 int i, skipped;
44d2e365 1928 int len;
ca3b6696 1929 pid_t pid;
9fe32043 1930 int status;
44d2e365
N
1931 char pathbuf[1024];
1932 char *paths[4] = {
1933 pathbuf,
85945e19 1934 BINDIR "/mdmon",
2f1bcf43 1935 "./mdmon",
44d2e365
N
1936 NULL
1937 };
8850ee3e 1938
40ebbb9c 1939 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1940 return 0;
1941
9cf014ec 1942 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1943 if (len > 0) {
1944 char *sl;
1945 pathbuf[len] = 0;
1946 sl = strrchr(pathbuf, '/');
1947 if (sl)
1948 sl++;
1949 else
1950 sl = pathbuf;
1951 strcpy(sl, "mdmon");
1952 } else
1953 pathbuf[0] = '\0';
1954
0f7bdf89 1955 /* First try to run systemctl */
9540cc24
N
1956 if (!check_env("MDADM_NO_SYSTEMCTL"))
1957 switch(fork()) {
1958 case 0:
1959 /* FIXME yuk. CLOSE_EXEC?? */
1960 skipped = 0;
1961 for (i = 3; skipped < 20; i++)
1962 if (close(i) < 0)
1963 skipped++;
1964 else
1965 skipped = 0;
1966
1967 /* Don't want to see error messages from
1968 * systemctl. If the service doesn't exist,
1969 * we start mdmon ourselves.
1970 */
1971 close(2);
1972 open("/dev/null", O_WRONLY);
1973 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1974 devnm);
1975 status = execl("/usr/bin/systemctl", "systemctl",
1976 "start",
1977 pathbuf, NULL);
1978 status = execl("/bin/systemctl", "systemctl", "start",
1979 pathbuf, NULL);
1980 exit(1);
7a862a02 1981 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1982 return -1;
1983 default: /* parent - good */
1984 pid = wait(&status);
1985 if (pid >= 0 && status == 0)
1986 return 0;
1987 }
0f7bdf89
JS
1988
1989 /* That failed, try running mdmon directly */
8850ee3e
N
1990 switch(fork()) {
1991 case 0:
1992 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1993 skipped = 0;
ca3b6696 1994 for (i = 3; skipped < 20; i++)
1913c325
N
1995 if (close(i) < 0)
1996 skipped++;
1997 else
1998 skipped = 0;
1999
ca3b6696 2000 for (i = 0; paths[i]; i++)
a0963a86 2001 if (paths[i][0]) {
2f1bcf43 2002 execl(paths[i], paths[i],
4dd2df09 2003 devnm, NULL);
a0963a86 2004 }
8850ee3e 2005 exit(1);
7a862a02 2006 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 2007 return -1;
9fe32043
N
2008 default: /* parent - good */
2009 pid = wait(&status);
15c10423 2010 if (pid < 0 || status != 0) {
7a862a02 2011 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 2012 return -1;
15c10423 2013 }
8850ee3e
N
2014 }
2015 return 0;
2016}
2017
148acb7b
DW
2018__u32 random32(void)
2019{
2020 __u32 rv;
2021 int rfd = open("/dev/urandom", O_RDONLY);
2022 if (rfd < 0 || read(rfd, &rv, 4) != 4)
2023 rv = random();
2024 if (rfd >= 0)
2025 close(rfd);
2026 return rv;
2027}
2028
c5f71c24
JS
2029void random_uuid(__u8 *buf)
2030{
2031 int fd, i, len;
2032 __u32 r[4];
2033
2034 fd = open("/dev/urandom", O_RDONLY);
2035 if (fd < 0)
2036 goto use_random;
2037 len = read(fd, buf, 16);
2038 close(fd);
2039 if (len != 16)
2040 goto use_random;
2041
2042 return;
2043
2044use_random:
2045 for (i = 0; i < 4; i++)
2046 r[i] = random();
2047 memcpy(buf, r, 16);
2048}
2049
0e600426 2050#ifndef MDASSEMBLE
edd8d13c
NB
2051int flush_metadata_updates(struct supertype *st)
2052{
2053 int sfd;
2054 if (!st->updates) {
2055 st->update_tail = NULL;
2056 return -1;
2057 }
2058
4dd2df09 2059 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
2060 if (sfd < 0)
2061 return -1;
2062
2063 while (st->updates) {
2064 struct metadata_update *mu = st->updates;
2065 st->updates = mu->next;
2066
2067 send_message(sfd, mu, 0);
2068 wait_reply(sfd, 0);
2069 free(mu->buf);
2070 free(mu);
2071 }
2072 ack(sfd, 0);
2073 wait_reply(sfd, 0);
2074 close(sfd);
2075 st->update_tail = NULL;
2076 return 0;
2077}
2078
2079void append_metadata_update(struct supertype *st, void *buf, int len)
2080{
2081
503975b9 2082 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
2083
2084 mu->buf = buf;
2085 mu->len = len;
2086 mu->space = NULL;
cb23f1f4 2087 mu->space_list = NULL;
edd8d13c
NB
2088 mu->next = NULL;
2089 *st->update_tail = mu;
2090 st->update_tail = &mu->next;
2091}
0e600426 2092#endif /* MDASSEMBLE */
a931db9e 2093
fe6729fa
NB
2094#ifdef __TINYC__
2095/* tinyc doesn't optimize this check in ioctl.h out ... */
2096unsigned int __invalid_size_argument_for_IOC = 0;
2097#endif
2098
6d11ec6f
AK
2099int experimental(void)
2100{
2101 if (check_env("MDADM_EXPERIMENTAL"))
2102 return 1;
2103 else {
7a862a02 2104 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
2105 return 0;
2106 }
2107}
2108
326727d9
AC
2109/* Pick all spares matching given criteria from a container
2110 * if min_size == 0 do not check size
2111 * if domlist == NULL do not check domains
2112 * if spare_group given add it to domains of each spare
2113 * metadata allows to test domains using metadata of destination array */
2114struct mdinfo *container_choose_spares(struct supertype *st,
2115 unsigned long long min_size,
2116 struct domainlist *domlist,
2117 char *spare_group,
2118 const char *metadata, int get_one)
2119{
2120 struct mdinfo *d, **dp, *disks = NULL;
2121
2122 /* get list of all disks in container */
2123 if (st->ss->getinfo_super_disks)
2124 disks = st->ss->getinfo_super_disks(st);
2125
2126 if (!disks)
2127 return disks;
2128 /* find spare devices on the list */
2129 dp = &disks->devs;
2130 disks->array.spare_disks = 0;
2131 while (*dp) {
2132 int found = 0;
2133 d = *dp;
2134 if (d->disk.state == 0) {
2135 /* check if size is acceptable */
2136 unsigned long long dev_size;
2137 dev_t dev = makedev(d->disk.major,d->disk.minor);
2138
2139 if (!min_size ||
2140 (dev_size_from_id(dev, &dev_size) &&
2141 dev_size >= min_size))
2142 found = 1;
2143 /* check if domain matches */
2144 if (found && domlist) {
4dd2df09 2145 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2146 if (spare_group)
2147 pol_add(&pol, pol_domain,
2148 spare_group, NULL);
e5508b36 2149 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2150 found = 0;
2151 dev_policy_free(pol);
2152 }
2153 }
2154 if (found) {
2155 dp = &d->next;
2156 disks->array.spare_disks++;
2157 if (get_one) {
2158 sysfs_free(*dp);
2159 d->next = NULL;
2160 }
2161 } else {
2162 *dp = d->next;
2163 d->next = NULL;
2164 sysfs_free(d);
2165 }
2166 }
2167 return disks;
2168}
9eafa1de
MN
2169
2170/* Checks if paths point to the same device
2171 * Returns 0 if they do.
2172 * Returns 1 if they don't.
2173 * Returns -1 if something went wrong,
2174 * e.g. paths are empty or the files
2175 * they point to don't exist */
2176int compare_paths (char* path1, char* path2)
2177{
2178 struct stat st1,st2;
2179
2180 if (path1 == NULL || path2 == NULL)
2181 return -1;
2182 if (stat(path1,&st1) != 0)
2183 return -1;
2184 if (stat(path2,&st2) != 0)
2185 return -1;
2186 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2187 return 0;
2188 return 1;
2189}
a7dec3fd
N
2190
2191/* Make sure we can open as many devices as needed */
2192void enable_fds(int devices)
2193{
2194 unsigned int fds = 20 + devices;
2195 struct rlimit lim;
2196 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
2197 || lim.rlim_cur >= fds)
2198 return;
2199 if (lim.rlim_max < fds)
2200 lim.rlim_max = fds;
2201 lim.rlim_cur = fds;
2202 setrlimit(RLIMIT_NOFILE, &lim);
2203}
a9c15847
N
2204
2205int in_initrd(void)
2206{
2207 /* This is based on similar function in systemd. */
2208 struct statfs s;
76d0f188
JS
2209 /* statfs.f_type is signed long on s390x and MIPS, causing all
2210 sorts of sign extension problems with RAMFS_MAGIC being
2211 defined as 0x858458f6 */
a9c15847 2212 return statfs("/", &s) >= 0 &&
7ac5d47e 2213 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2214 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2215 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2216}
8832342d
N
2217
2218void reopen_mddev(int mdfd)
2219{
2220 /* Re-open without any O_EXCL, but keep
2221 * the same fd
2222 */
2223 char *devnm;
2224 int fd;
2225 devnm = fd2devnm(mdfd);
2226 close(mdfd);
2227 fd = open_dev(devnm);
2228 if (fd >= 0 && fd != mdfd)
2229 dup2(fd, mdfd);
2230}
7716570e 2231
7071320a 2232#ifndef MDASSEMBLE
e80357f8
GJ
2233static struct cmap_hooks *cmap_hooks = NULL;
2234static int is_cmap_hooks_ready = 0;
7716570e 2235
e80357f8
GJ
2236void set_cmap_hooks(void)
2237{
2238 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2239 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2240 if (!cmap_hooks->cmap_handle)
2241 return;
7716570e 2242
e80357f8
GJ
2243 cmap_hooks->initialize = dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2244 cmap_hooks->get_string = dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
2245 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2246
e80357f8
GJ
2247 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2248 !cmap_hooks->finalize)
2249 dlclose(cmap_hooks->cmap_handle);
2250 else
2251 is_cmap_hooks_ready = 1;
2252}
7716570e 2253
e80357f8
GJ
2254int get_cluster_name(char **cluster_name)
2255{
2256 int rv = -1;
2257 cmap_handle_t handle;
7716570e 2258
e80357f8
GJ
2259 if (!is_cmap_hooks_ready)
2260 return rv;
7716570e 2261
e80357f8 2262 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2263 if (rv != CS_OK)
2264 goto out;
2265
e80357f8 2266 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2267 if (rv != CS_OK) {
2268 free(*cluster_name);
2269 rv = -1;
2270 goto name_err;
2271 }
2272
2273 rv = 0;
2274name_err:
e80357f8 2275 cmap_hooks->finalize(handle);
7716570e 2276out:
7716570e
GJ
2277 return rv;
2278}
d15a1f72
GJ
2279
2280void set_dlm_hooks(void)
2281{
2282 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2283 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2284 if (!dlm_hooks->dlm_handle)
2285 return;
2286
2287 dlm_hooks->create_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2288 dlm_hooks->release_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
2289 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2290 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2291 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2292 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2293
2294 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2295 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2296 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2297 dlclose(dlm_hooks->dlm_handle);
2298 else
2299 is_dlm_hooks_ready = 1;
2300}
e80357f8
GJ
2301
2302void set_hooks(void)
2303{
2304 set_dlm_hooks();
2305 set_cmap_hooks();
2306}
7d55dca2 2307#endif