Manage_subdevs(): Use a dev_t
[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
7071320a 92static struct dlm_hooks *dlm_hooks = NULL;
d15a1f72
GJ
93struct dlm_lock_resource *dlm_lock_res = NULL;
94static int ast_called = 0;
95
96struct dlm_lock_resource {
97 dlm_lshandle_t *ls;
98 struct dlm_lksb lksb;
99};
100
d15a1f72
GJ
101/* Using poll(2) to wait for and dispatch ASTs */
102static int poll_for_ast(dlm_lshandle_t ls)
103{
104 struct pollfd pfd;
105
106 pfd.fd = dlm_hooks->ls_get_fd(ls);
107 pfd.events = POLLIN;
108
109 while (!ast_called)
110 {
111 if (poll(&pfd, 1, 0) < 0)
112 {
113 perror("poll");
114 return -1;
115 }
116 dlm_hooks->dispatch(dlm_hooks->ls_get_fd(ls));
117 }
118 ast_called = 0;
119
120 return 0;
121}
122
123static void dlm_ast(void *arg)
124{
125 ast_called = 1;
126}
127
81a8a694 128static char *cluster_name = NULL;
d15a1f72 129/* Create the lockspace, take bitmapXXX locks on all the bitmaps. */
81a8a694 130int cluster_get_dlmlock(int *lockid)
d15a1f72
GJ
131{
132 int ret = -1;
133 char str[64];
134 int flags = LKF_NOQUEUE;
135
81a8a694
GJ
136 ret = get_cluster_name(&cluster_name);
137 if (ret) {
138 pr_err("The md can't get cluster name\n");
139 return -1;
140 }
141
d15a1f72 142 dlm_lock_res = xmalloc(sizeof(struct dlm_lock_resource));
81a8a694 143 dlm_lock_res->ls = dlm_hooks->create_lockspace(cluster_name, O_RDWR);
d15a1f72 144 if (!dlm_lock_res->ls) {
81a8a694 145 pr_err("%s failed to create lockspace\n", cluster_name);
32539f74 146 return -ENOMEM;
d15a1f72
GJ
147 }
148
81a8a694 149 snprintf(str, 64, "bitmap%s", cluster_name);
d15a1f72
GJ
150 ret = dlm_hooks->ls_lock(dlm_lock_res->ls, LKM_PWMODE, &dlm_lock_res->lksb,
151 flags, str, strlen(str), 0, dlm_ast,
152 dlm_lock_res, NULL, NULL);
153 if (ret) {
154 pr_err("error %d when get PW mode on lock %s\n", errno, str);
32539f74
GJ
155 dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1);
156 return ret;
d15a1f72
GJ
157 }
158
159 /* Wait for it to complete */
160 poll_for_ast(dlm_lock_res->ls);
161 *lockid = dlm_lock_res->lksb.sb_lkid;
162
32539f74 163 return dlm_lock_res->lksb.sb_status;
d15a1f72
GJ
164}
165
81a8a694 166int cluster_release_dlmlock(int lockid)
d15a1f72
GJ
167{
168 int ret = -1;
169
81a8a694
GJ
170 if (!cluster_name)
171 return -1;
172
d15a1f72
GJ
173 ret = dlm_hooks->ls_unlock(dlm_lock_res->ls, lockid, 0,
174 &dlm_lock_res->lksb, dlm_lock_res);
175 if (ret) {
176 pr_err("error %d happened when unlock\n", errno);
177 /* XXX make sure the lock is unlocked eventually */
178 goto out;
179 }
180
181 /* Wait for it to complete */
182 poll_for_ast(dlm_lock_res->ls);
183
184 errno = dlm_lock_res->lksb.sb_status;
185 if (errno != EUNLOCK) {
186 pr_err("error %d happened in ast when unlock lockspace\n", errno);
187 /* XXX make sure the lockspace is unlocked eventually */
188 goto out;
189 }
190
81a8a694 191 ret = dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1);
d15a1f72
GJ
192 if (ret) {
193 pr_err("error %d happened when release lockspace\n", errno);
194 /* XXX make sure the lockspace is released eventually */
195 goto out;
196 }
197 free(dlm_lock_res);
198
199out:
200 return ret;
201}
202
9db2ab4e
JS
203int md_array_valid(int fd)
204{
205 struct mdinfo *sra;
206 int ret;
207
208 sra = sysfs_read(fd, NULL, GET_ARRAY_STATE);
209 if (sra) {
210 if (sra->array_state != ARRAY_UNKNOWN_STATE)
211 ret = 0;
212 else
213 ret = -ENODEV;
214
215 free(sra);
216 } else {
217 /*
218 * GET_ARRAY_INFO doesn't provide access to the proper state
219 * information, so fallback to a basic check for raid_disks != 0
220 */
221 ret = ioctl(fd, RAID_VERSION);
222 }
223
224 return !ret;
225}
226
3ab8f4bf
JS
227int md_array_active(int fd)
228{
229 struct mdinfo *sra;
230 struct mdu_array_info_s array;
a822017f 231 int ret = 0;
3ab8f4bf
JS
232
233 sra = sysfs_read(fd, NULL, GET_ARRAY_STATE);
234 if (sra) {
a822017f 235 if (!md_array_is_active(sra))
3ab8f4bf
JS
236 ret = -ENODEV;
237
238 free(sra);
239 } else {
240 /*
241 * GET_ARRAY_INFO doesn't provide access to the proper state
242 * information, so fallback to a basic check for raid_disks != 0
243 */
244 ret = ioctl(fd, GET_ARRAY_INFO, &array);
245 }
246
247 return !ret;
248}
249
a822017f
MT
250int md_array_is_active(struct mdinfo *info)
251{
252 return (info->array_state != ARRAY_CLEAR &&
253 info->array_state != ARRAY_INACTIVE &&
254 info->array_state != ARRAY_UNKNOWN_STATE);
255}
256
9cd39f01
JS
257/*
258 * Get array info from the kernel. Longer term we want to deprecate the
259 * ioctl and get it from sysfs.
260 */
261int md_get_array_info(int fd, struct mdu_array_info_s *array)
262{
263 return ioctl(fd, GET_ARRAY_INFO, array);
264}
265
018a4882
JS
266/*
267 * Set array info
268 */
269int md_set_array_info(int fd, struct mdu_array_info_s *array)
270{
271 return ioctl(fd, SET_ARRAY_INFO, array);
272}
273
d97572f5
JS
274/*
275 * Get disk info from the kernel.
276 */
277int md_get_disk_info(int fd, struct mdu_disk_info_s *disk)
278{
279 return ioctl(fd, GET_DISK_INFO, disk);
280}
281
64c4757e
NB
282/*
283 * Parse a 128 bit uuid in 4 integers
284 * format is 32 hexx nibbles with options :.<space> separator
285 * If not exactly 32 hex digits are found, return 0
286 * else return 1
287 */
288int parse_uuid(char *str, int uuid[4])
289{
aba69144
NB
290 int hit = 0; /* number of Hex digIT */
291 int i;
292 char c;
ca3b6696
N
293 for (i = 0; i < 4; i++)
294 uuid[i] = 0;
aba69144 295
ca3b6696 296 while ((c = *str++) != 0) {
aba69144 297 int n;
ca3b6696 298 if (c >= '0' && c <= '9')
aba69144 299 n = c-'0';
ca3b6696 300 else if (c >= 'a' && c <= 'f')
aba69144 301 n = 10 + c - 'a';
ca3b6696 302 else if (c >= 'A' && c <= 'F')
aba69144
NB
303 n = 10 + c - 'A';
304 else if (strchr(":. -", c))
305 continue;
306 else return 0;
307
308 if (hit<32) {
309 uuid[hit/8] <<= 4;
310 uuid[hit/8] += n;
311 }
312 hit++;
82b27616 313 }
aba69144
NB
314 if (hit == 32)
315 return 1;
316 return 0;
64c4757e
NB
317}
318
64c4757e
NB
319int get_linux_version()
320{
321 struct utsname name;
98c6faba 322 char *cp;
f161d047 323 int a = 0, b = 0,c = 0;
64c4757e
NB
324 if (uname(&name) <0)
325 return -1;
326
98c6faba
NB
327 cp = name.release;
328 a = strtoul(cp, &cp, 10);
f161d047
N
329 if (*cp == '.')
330 b = strtoul(cp+1, &cp, 10);
331 if (*cp == '.')
332 c = strtoul(cp+1, &cp, 10);
98c6faba 333
682c7051 334 return (a*1000000)+(b*1000)+c;
64c4757e
NB
335}
336
bc77ed53
DW
337int mdadm_version(char *version)
338{
339 int a, b, c;
340 char *cp;
341
342 if (!version)
343 version = Version;
344
345 cp = strchr(version, '-');
346 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
347 return -1;
348 cp += 3;
349 a = strtoul(cp, &cp, 10);
350 if (*cp != '.')
351 return -1;
352 b = strtoul(cp+1, &cp, 10);
353 if (*cp == '.')
354 c = strtoul(cp+1, &cp, 10);
355 else
356 c = 0;
357 if (*cp != ' ' && *cp != '-')
358 return -1;
359 return (a*1000000)+(b*1000)+c;
360}
361
51d4261c 362unsigned long long parse_size(char *size)
84e11361
N
363{
364 /* parse 'size' which should be a number optionally
365 * followed by 'K', 'M', or 'G'.
366 * Without a suffix, K is assumed.
367 * Number returned is in sectors (half-K)
822e393a 368 * INVALID_SECTORS returned on error.
84e11361
N
369 */
370 char *c;
371 long long s = strtoll(size, &c, 10);
372 if (s > 0) {
373 switch (*c) {
374 case 'K':
375 c++;
376 default:
377 s *= 2;
378 break;
379 case 'M':
380 c++;
381 s *= 1024 * 2;
382 break;
383 case 'G':
384 c++;
385 s *= 1024 * 1024 * 2;
386 break;
72ca9bcf
N
387 case 's': /* sectors */
388 c++;
389 break;
84e11361 390 }
51d4261c 391 } else
822e393a 392 s = INVALID_SECTORS;
84e11361 393 if (*c)
822e393a 394 s = INVALID_SECTORS;
84e11361
N
395 return s;
396}
397
4a06e2c2
N
398int parse_layout_10(char *layout)
399{
400 int copies, rv;
401 char *cp;
402 /* Parse the layout string for raid10 */
403 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
404 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
405 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
406 copies > 200 ||
407 *cp)
408 return -1;
409 if (layout[0] == 'n')
410 rv = 256 + copies;
411 else if (layout[0] == 'o')
412 rv = 0x10000 + (copies<<8) + 1;
413 else
414 rv = 1 + (copies<<8);
415 return rv;
416}
417
418int parse_layout_faulty(char *layout)
419{
420 /* Parse the layout string for 'faulty' */
421 int ln = strcspn(layout, "0123456789");
503975b9 422 char *m = xstrdup(layout);
4a06e2c2
N
423 int mode;
424 m[ln] = 0;
425 mode = map_name(faultylayout, m);
426 if (mode == UnSet)
427 return -1;
428
429 return mode | (atoi(layout+ln)<< ModeShift);
430}
79868890
N
431
432long parse_num(char *num)
433{
434 /* Either return a valid number, or -1 */
435 char *c;
436 long rv = strtol(num, &c, 10);
437 if (rv < 0 || *c || !num[0])
438 return -1;
439 else
440 return rv;
441}
4a06e2c2 442
4de90913
GJ
443int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
444{
445 char *dev;
446 *slot = strtoul(input, &dev, 10);
447 if (dev == input || dev[0] != ':')
448 return -1;
449 *devname = dev+1;
450 return 0;
451}
452
0430ed48
NB
453void remove_partitions(int fd)
454{
455 /* remove partitions from this block devices.
456 * This is used for components added to an array
457 */
458#ifdef BLKPG_DEL_PARTITION
459 struct blkpg_ioctl_arg a;
460 struct blkpg_partition p;
461
462 a.op = BLKPG_DEL_PARTITION;
463 a.data = (void*)&p;
464 a.datalen = sizeof(p);
465 a.flags = 0;
466 memset(a.data, 0, a.datalen);
ca3b6696 467 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
468 ioctl(fd, BLKPG, &a);
469#endif
470}
471
691c6ee1
N
472int test_partition(int fd)
473{
474 /* Check if fd is a whole-disk or a partition.
475 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
476 * will return ENXIO on an invalid partition number.
477 */
478 struct blkpg_ioctl_arg a;
479 struct blkpg_partition p;
480 a.op = BLKPG_DEL_PARTITION;
481 a.data = (void*)&p;
482 a.datalen = sizeof(p);
483 a.flags = 0;
484 memset(a.data, 0, a.datalen);
485 p.pno = 1<<30;
486 if (ioctl(fd, BLKPG, &a) == 0)
487 /* Very unlikely, but not a partition */
488 return 0;
357ac106 489 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
490 /* not a partition */
491 return 0;
492
493 return 1;
494}
495
bfd76b93
CA
496int test_partition_from_id(dev_t id)
497{
498 char buf[20];
499 int fd, rv;
500
501 sprintf(buf, "%d:%d", major(id), minor(id));
502 fd = dev_open(buf, O_RDONLY);
503 if (fd < 0)
504 return -1;
505 rv = test_partition(fd);
506 close(fd);
507 return rv;
508}
691c6ee1 509
de5a472e 510int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 511{
265e0f17 512 int copies, first;
de5a472e
N
513 int i;
514 int avail_disks = 0;
515
516 for (i = 0; i < raid_disks; i++)
517 avail_disks += !!avail[i];
518
64c4757e 519 switch (level) {
265e0f17
NB
520 case 10:
521 /* This is the tricky one - we need to check
522 * which actual disks are present.
523 */
702b557b 524 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 525 first = 0;
265e0f17
NB
526 do {
527 /* there must be one of the 'copies' form 'first' */
528 int n = copies;
ca3b6696 529 int cnt = 0;
b3ec716d 530 int this = first;
265e0f17 531 while (n--) {
b3ec716d 532 if (avail[this])
265e0f17 533 cnt++;
b3ec716d 534 this = (this+1) % raid_disks;
265e0f17
NB
535 }
536 if (cnt == 0)
537 return 0;
b3ec716d 538 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
539 } while (first != 0);
540 return 1;
e5329c37 541
df0d4ea0 542 case LEVEL_MULTIPATH:
e0d19036 543 return avail_disks>= 1;
df0d4ea0 544 case LEVEL_LINEAR:
64c4757e
NB
545 case 0:
546 return avail_disks == raid_disks;
547 case 1:
548 return avail_disks >= 1;
549 case 4:
330d6900
N
550 if (avail_disks == raid_disks - 1 &&
551 !avail[raid_disks - 1])
552 /* If just the parity device is missing, then we
553 * have enough, even if not clean
554 */
555 return 1;
556 /* FALL THROUGH */
64c4757e 557 case 5:
583315d9
NB
558 if (clean)
559 return avail_disks >= raid_disks-1;
560 else
561 return avail_disks >= raid_disks;
98c6faba 562 case 6:
583315d9
NB
563 if (clean)
564 return avail_disks >= raid_disks-2;
565 else
566 return avail_disks >= raid_disks;
64c4757e
NB
567 default:
568 return 0;
569 }
570}
571
22e263f6
AC
572const int uuid_zero[4] = { 0, 0, 0, 0 };
573
f277ce36 574int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 575{
f277ce36
NB
576 if (swapuuid) {
577 /* parse uuids are hostendian.
578 * uuid's from some superblocks are big-ending
aba69144 579 * if there is a difference, we need to swap..
f277ce36
NB
580 */
581 unsigned char *ac = (unsigned char *)a;
582 unsigned char *bc = (unsigned char *)b;
583 int i;
ca3b6696 584 for (i = 0; i < 16; i += 4) {
f277ce36
NB
585 if (ac[i+0] != bc[i+3] ||
586 ac[i+1] != bc[i+2] ||
587 ac[i+2] != bc[i+1] ||
588 ac[i+3] != bc[i+0])
589 return 0;
590 }
591 return 1;
592 } else {
593 if (a[0]==b[0] &&
594 a[1]==b[1] &&
595 a[2]==b[2] &&
596 a[3]==b[3])
597 return 1;
598 return 0;
599 }
64c4757e 600}
ca3b6696 601
350f29f9
NB
602void copy_uuid(void *a, int b[4], int swapuuid)
603{
604 if (swapuuid) {
605 /* parse uuids are hostendian.
606 * uuid's from some superblocks are big-ending
607 * if there is a difference, we need to swap..
608 */
609 unsigned char *ac = (unsigned char *)a;
610 unsigned char *bc = (unsigned char *)b;
611 int i;
ca3b6696 612 for (i = 0; i < 16; i += 4) {
350f29f9
NB
613 ac[i+0] = bc[i+3];
614 ac[i+1] = bc[i+2];
615 ac[i+2] = bc[i+1];
616 ac[i+3] = bc[i+0];
617 }
618 } else
619 memcpy(a, b, 16);
620}
64c4757e 621
aae5a112 622char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 623{
9968e376 624 int i, j;
d7288ddc
N
625 char uuid[16];
626 char *c = buf;
627 strcpy(c, "UUID-");
628 c += strlen(c);
aae5a112 629 copy_uuid(uuid, id, swap);
9968e376 630 for (i = 0; i < 4; i++) {
9968e376 631 if (i)
ff54de6e 632 *c++ = sep;
9968e376
DW
633 for (j = 3; j >= 0; j--) {
634 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
635 c+= 2;
636 }
d7288ddc
N
637 }
638 return buf;
aae5a112
DW
639
640}
641
642char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
643{
86983cce
N
644 // dirty hack to work around an issue with super1 superblocks...
645 // super1 superblocks need swapuuid set in order for assembly to
646 // work, but can't have it set if we want this printout to match
647 // all the other uuid printouts in super1.c, so we force swapuuid
648 // to 1 to make our printout match the rest of super1
649 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
650}
651
682c7051
NB
652int check_ext2(int fd, char *name)
653{
654 /*
655 * Check for an ext2fs file system.
656 * Superblock is always 1K at 1K offset
657 *
658 * s_magic is le16 at 56 == 0xEF53
659 * report mtime - le32 at 44
660 * blocks - le32 at 4
661 * logblksize - le32 at 24
662 */
663 unsigned char sb[1024];
664 time_t mtime;
2538ba2a
N
665 unsigned long long size;
666 int bsize;
682c7051
NB
667 if (lseek(fd, 1024,0)!= 1024)
668 return 0;
669 if (read(fd, sb, 1024)!= 1024)
670 return 0;
671 if (sb[56] != 0x53 || sb[57] != 0xef)
672 return 0;
673
674 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
675 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
676 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 677 size <<= bsize;
e7b84f9d 678 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 679 name);
2538ba2a 680 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
681 return 1;
682}
683
684int check_reiser(int fd, char *name)
685{
686 /*
687 * superblock is at 64K
688 * size is 1024;
689 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
690 *
691 */
692 unsigned char sb[1024];
2538ba2a 693 unsigned long long size;
682c7051
NB
694 if (lseek(fd, 64*1024, 0) != 64*1024)
695 return 0;
696 if (read(fd, sb, 1024) != 1024)
697 return 0;
ca3b6696
N
698 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
699 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 700 return 0;
e7b84f9d 701 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 702 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 703 cont_err("size = %lluK\n", size*4);
aba69144 704
682c7051
NB
705 return 1;
706}
707
708int check_raid(int fd, char *name)
709{
4b1ac34b 710 struct mdinfo info;
682c7051 711 time_t crtime;
d078d77c 712 char *level;
82d9eba6 713 struct supertype *st = guess_super(fd);
f9ce90ba 714
cb8f6859
N
715 if (!st)
716 return 0;
fa219dd2
MD
717 if (st->ss->add_to_super != NULL) {
718 st->ss->load_super(st, fd, name);
719 /* Looks like a raid array .. */
720 pr_err("%s appears to be part of a raid array:\n", name);
721 st->ss->getinfo_super(st, &info, NULL);
722 st->ss->free_super(st);
723 crtime = info.array.ctime;
724 level = map_num(pers, info.array.level);
725 if (!level)
726 level = "-unknown-";
727 cont_err("level=%s devices=%d ctime=%s",
728 level, info.array.raid_disks, ctime(&crtime));
729 } else {
730 /* Looks like GPT or MBR */
731 pr_err("partition table exists on %s\n", name);
732 }
82d9eba6 733 return 1;
682c7051
NB
734}
735
0a6bff09
ZL
736int fstat_is_blkdev(int fd, char *devname, dev_t *rdev)
737{
738 struct stat stb;
739
740 if (fstat(fd, &stb) != 0) {
741 pr_err("fstat failed for %s: %s\n", devname, strerror(errno));
742 return 0;
743 }
744 if ((S_IFMT & stb.st_mode) != S_IFBLK) {
745 pr_err("%s is not a block device.\n", devname);
9e04ac1c
ZL
746 return 0;
747 }
748 if (rdev)
749 *rdev = stb.st_rdev;
750 return 1;
751}
752
753int stat_is_blkdev(char *devname, dev_t *rdev)
754{
755 struct stat stb;
756
757 if (stat(devname, &stb) != 0) {
758 pr_err("stat failed for %s: %s\n", devname, strerror(errno));
759 return 0;
760 }
761 if ((S_IFMT & stb.st_mode) != S_IFBLK) {
762 pr_err("%s is not a block device.\n", devname);
0a6bff09
ZL
763 return 0;
764 }
765 if (rdev)
766 *rdev = stb.st_rdev;
767 return 1;
768}
769
682c7051
NB
770int ask(char *mesg)
771{
772 char *add = "";
773 int i;
ca3b6696 774 for (i = 0; i < 5; i++) {
682c7051
NB
775 char buf[100];
776 fprintf(stderr, "%s%s", mesg, add);
777 fflush(stderr);
778 if (fgets(buf, 100, stdin)==NULL)
779 return 0;
780 if (buf[0]=='y' || buf[0]=='Y')
781 return 1;
782 if (buf[0]=='n' || buf[0]=='N')
783 return 0;
784 add = "(y/n) ";
785 }
e7b84f9d 786 pr_err("assuming 'no'\n");
682c7051
NB
787 return 0;
788}
789
8d80900b 790int is_standard(char *dev, int *nump)
e5329c37
NB
791{
792 /* tests if dev is a "standard" md dev name.
793 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 794 * where NN is a string of digits
598f0d58
NB
795 * Returns 1 if a partitionable standard,
796 * -1 if non-partitonable,
797 * 0 if not a standard name.
e5329c37 798 */
8d80900b 799 char *d = strrchr(dev, '/');
ca3b6696 800 int type = 0;
8d80900b
NB
801 int num;
802 if (!d)
e5329c37 803 return 0;
ca3b6696
N
804 if (strncmp(d, "/d",2) == 0)
805 d += 2, type = 1; /* /dev/md/dN{pM} */
806 else if (strncmp(d, "/md_d", 5) == 0)
807 d += 5, type = 1; /* /dev/md_dN{pM} */
808 else if (strncmp(d, "/md", 3) == 0)
809 d += 3, type = -1; /* /dev/mdN */
810 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
811 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
812 else
813 return 0;
8d80900b 814 if (!*d)
e5329c37 815 return 0;
8d80900b
NB
816 num = atoi(d);
817 while (isdigit(*d))
818 d++;
819 if (*d)
e5329c37 820 return 0;
8d80900b
NB
821 if (nump) *nump = num;
822
823 return type;
e5329c37
NB
824}
825
4b1ac34b 826unsigned long calc_csum(void *super, int bytes)
82b27616 827{
56eb10c0 828 unsigned long long newcsum = 0;
82b27616 829 int i;
4b1ac34b
NB
830 unsigned int csum;
831 unsigned int *superc = (unsigned int*) super;
82b27616 832
ca3b6696
N
833 for(i = 0; i < bytes/4; i++)
834 newcsum += superc[i];
82b27616 835 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 836#ifdef __alpha__
aba69144 837/* The in-kernel checksum calculation is always 16bit on
570c0542 838 * the alpha, though it is 32 bit on i386...
ca3b6696 839 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
840 * a way that it shouldn't).
841 */
842 csum = (csum & 0xffff) + (csum >> 16);
843 csum = (csum & 0xffff) + (csum >> 16);
844#endif
82b27616
NB
845 return csum;
846}
cd29a5c8 847
56eb10c0 848char *human_size(long long bytes)
cd29a5c8 849{
ff9239ee 850 static char buf[47];
d5d3721e
NB
851
852 /* We convert bytes to either centi-M{ega,ibi}bytes or
853 * centi-G{igi,ibi}bytes, with appropriate rounding,
854 * and then print 1/100th of those as a decimal.
855 * We allow upto 2048Megabytes before converting to
856 * gigabytes, as that shows more precision and isn't
857 * too large a number.
01619b48 858 * Terabytes are not yet handled.
d5d3721e 859 */
cd29a5c8 860
56eb10c0 861 if (bytes < 5000*1024)
ca3b6696 862 buf[0] = 0;
d5d3721e 863 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 864 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 865 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 866 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
efa29530 867 cMiB/100, cMiB % 100, cMB/100, cMB % 100);
d5d3721e 868 } else {
93d3bd3b 869 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 870 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 871 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
efa29530 872 cGiB/100, cGiB % 100, cGB/100, cGB % 100);
d5d3721e 873 }
cd29a5c8
NB
874 return buf;
875}
e0d19036 876
f0ec6710 877char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
878{
879 static char buf[30];
e0d19036 880
570abc6f
MN
881 /* We convert bytes to either centi-M{ega,ibi}bytes or
882 * centi-G{igi,ibi}bytes, with appropriate rounding,
883 * and then print 1/100th of those as a decimal.
884 * We allow upto 2048Megabytes before converting to
885 * gigabytes, as that shows more precision and isn't
886 * too large a number.
887 * Terabytes are not yet handled.
f0ec6710
MN
888 *
889 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
890 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
891 */
892
e0d19036 893 if (bytes < 5000*1024)
570abc6f 894 buf[0] = 0;
f0ec6710
MN
895 else if (prefix == IEC) {
896 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 897 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710 898 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
efa29530 899 cMiB/100, cMiB % 100);
f0ec6710 900 } else {
93d3bd3b 901 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710 902 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
efa29530 903 cGiB/100, cGiB % 100);
f0ec6710
MN
904 }
905 }
906 else if (prefix == JEDEC) {
907 if (bytes < 2*1024LL*1024LL*1024LL) {
908 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
909 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
efa29530 910 cMB/100, cMB % 100);
f0ec6710
MN
911 } else {
912 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
913 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
efa29530 914 cGB/100, cGB % 100);
f0ec6710 915 }
570abc6f 916 }
f0ec6710
MN
917 else
918 buf[0] = 0;
570abc6f 919
e0d19036
NB
920 return buf;
921}
e4965ef8
N
922
923void print_r10_layout(int layout)
924{
925 int near = layout & 255;
926 int far = (layout >> 8) & 255;
927 int offset = (layout&0x10000);
928 char *sep = "";
929
930 if (near != 1) {
931 printf("%s near=%d", sep, near);
932 sep = ",";
933 }
934 if (far != 1)
935 printf("%s %s=%d", sep, offset?"offset":"far", far);
936 if (near*far == 1)
937 printf("NO REDUNDANCY");
938}
e0d19036 939
5f8097be
NB
940unsigned long long calc_array_size(int level, int raid_disks, int layout,
941 int chunksize, unsigned long long devsize)
577e8448 942{
e03640bd
CA
943 if (level == 1)
944 return devsize;
577e8448
AK
945 devsize &= ~(unsigned long long)((chunksize>>9)-1);
946 return get_data_disks(level, layout, raid_disks) * devsize;
947}
948
949int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
950{
951 int data_disks = 0;
952 switch (level) {
ca3b6696
N
953 case 0: data_disks = raid_disks;
954 break;
955 case 1: data_disks = 1;
956 break;
5f8097be 957 case 4:
ca3b6696
N
958 case 5: data_disks = raid_disks - 1;
959 break;
960 case 6: data_disks = raid_disks - 2;
961 break;
5f8097be
NB
962 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
963 break;
964 }
577e8448
AK
965
966 return data_disks;
5f8097be
NB
967}
968
13db17bd 969dev_t devnm2devid(char *devnm)
4dd2df09
N
970{
971 /* First look in /sys/block/$DEVNM/dev for %d:%d
972 * If that fails, try parsing out a number
973 */
974 char path[100];
975 char *ep;
976 int fd;
977 int mjr,mnr;
978
979 sprintf(path, "/sys/block/%s/dev", devnm);
980 fd = open(path, O_RDONLY);
981 if (fd >= 0) {
982 char buf[20];
983 int n = read(fd, buf, sizeof(buf));
984 close(fd);
985 if (n > 0)
986 buf[n] = 0;
987 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
988 return makedev(mjr, mnr);
989 }
990 if (strncmp(devnm, "md_d", 4) == 0 &&
991 isdigit(devnm[4]) &&
992 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
993 ep > devnm && *ep == 0)
994 return makedev(get_mdp_major(), mnr << MdpMinorShift);
995
996 if (strncmp(devnm, "md", 2) == 0 &&
997 isdigit(devnm[2]) &&
998 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
999 ep > devnm && *ep == 0)
1000 return makedev(MD_MAJOR, mnr);
1001
1002 return 0;
1003}
1004
1005char *get_md_name(char *devnm)
e0d19036
NB
1006{
1007 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 1008 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 1009
e0d19036
NB
1010 static char devname[50];
1011 struct stat stb;
4dd2df09 1012 dev_t rdev = devnm2devid(devnm);
dd0781e5 1013 char *dn;
98c6faba 1014
4dd2df09
N
1015 if (rdev == 0)
1016 return 0;
1017 if (strncmp(devnm, "md_", 3) == 0) {
1018 snprintf(devname, sizeof(devname), "/dev/md/%s",
1019 devnm + 3);
fc54fe7a
JS
1020 if (stat(devname, &stb) == 0 &&
1021 (S_IFMT&stb.st_mode) == S_IFBLK && (stb.st_rdev == rdev))
98c6faba
NB
1022 return devname;
1023 }
4dd2df09 1024 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
fc54fe7a
JS
1025 if (stat(devname, &stb) == 0 && (S_IFMT&stb.st_mode) == S_IFBLK &&
1026 (stb.st_rdev == rdev))
4dd2df09
N
1027 return devname;
1028
1029 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
fc54fe7a
JS
1030 if (stat(devname, &stb) == 0 && (S_IFMT&stb.st_mode) == S_IFBLK &&
1031 (stb.st_rdev == rdev))
4dd2df09
N
1032 return devname;
1033
16c6fa80 1034 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
1035 if (dn)
1036 return dn;
4dd2df09 1037 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 1038 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
1039 if (errno != EEXIST)
1040 return NULL;
e0d19036 1041
fc54fe7a
JS
1042 if (stat(devname, &stb) == 0 && (S_IFMT&stb.st_mode) == S_IFBLK &&
1043 (stb.st_rdev == rdev))
e0d19036
NB
1044 return devname;
1045 unlink(devname);
1046 return NULL;
1047}
1048
1049void put_md_name(char *name)
1050{
ca3b6696 1051 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
1052 unlink(name);
1053}
f9ce90ba 1054
cc742d38
N
1055int get_maj_min(char *dev, int *major, int *minor)
1056{
1057 char *e;
1058 *major = strtoul(dev, &e, 0);
1059 return (e > dev && *e == ':' && e[1] &&
1060 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1061 *e == 0);
1062}
1063
8b0dabea
NB
1064int dev_open(char *dev, int flags)
1065{
1066 /* like 'open', but if 'dev' matches %d:%d, create a temp
1067 * block device and open that
1068 */
8b0dabea
NB
1069 int fd = -1;
1070 char devname[32];
e81cdd9f 1071 int major;
8b0dabea 1072 int minor;
e81cdd9f 1073
9f0ad56b
JS
1074 if (!dev)
1075 return -1;
6df6a774 1076 flags |= O_DIRECT;
e81cdd9f 1077
cc742d38 1078 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1079 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1080 (int)getpid(), major, minor);
1081 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1082 fd = open(devname, flags);
1083 unlink(devname);
6df6a774
N
1084 }
1085 if (fd < 0) {
9dc7d357 1086 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
1087 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
1088 (int)getpid(), major, minor);
ca3b6696 1089 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
1090 fd = open(devname, flags);
1091 unlink(devname);
1092 }
8b0dabea
NB
1093 }
1094 } else
6df6a774 1095 fd = open(dev, flags);
8b0dabea
NB
1096 return fd;
1097}
f9ce90ba 1098
4dd2df09 1099int open_dev_flags(char *devnm, int flags)
e8a70c89 1100{
13db17bd 1101 dev_t devid;
e8a70c89
N
1102 char buf[20];
1103
4dd2df09
N
1104 devid = devnm2devid(devnm);
1105 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1106 return dev_open(buf, flags);
1107}
1108
4dd2df09 1109int open_dev(char *devnm)
d998b738 1110{
4dd2df09 1111 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1112}
1113
4dd2df09 1114int open_dev_excl(char *devnm)
a931db9e
NB
1115{
1116 char buf[20];
1117 int i;
7187750e 1118 int flags = O_RDWR;
13db17bd 1119 dev_t devid = devnm2devid(devnm);
399e0b97 1120 long delay = 1000;
a931db9e 1121
4dd2df09 1122 sprintf(buf, "%d:%d", major(devid), minor(devid));
efa29530 1123 for (i = 0; i < 25; i++) {
7187750e 1124 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1125 if (fd >= 0)
1126 return fd;
7187750e
N
1127 if (errno == EACCES && flags == O_RDWR) {
1128 flags = O_RDONLY;
1129 continue;
1130 }
a931db9e
NB
1131 if (errno != EBUSY)
1132 return fd;
399e0b97
N
1133 usleep(delay);
1134 if (delay < 200000)
1135 delay *= 2;
a931db9e
NB
1136 }
1137 return -1;
1138}
1139
9008ed1c
N
1140int same_dev(char *one, char *two)
1141{
1142 struct stat st1, st2;
1143 if (stat(one, &st1) != 0)
1144 return 0;
1145 if (stat(two, &st2) != 0)
1146 return 0;
1147 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1148 return 0;
1149 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1150 return 0;
1151 return st1.st_rdev == st2.st_rdev;
1152}
1153
a7c6e3fb 1154void wait_for(char *dev, int fd)
a714580e
N
1155{
1156 int i;
a7c6e3fb 1157 struct stat stb_want;
399e0b97 1158 long delay = 1000;
a7c6e3fb
N
1159
1160 if (fstat(fd, &stb_want) != 0 ||
1161 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1162 return;
a714580e 1163
efa29530 1164 for (i = 0; i < 25; i++) {
a714580e 1165 struct stat stb;
a7c6e3fb
N
1166 if (stat(dev, &stb) == 0 &&
1167 (stb.st_mode & S_IFMT) == S_IFBLK &&
1168 (stb.st_rdev == stb_want.st_rdev))
a714580e 1169 return;
399e0b97
N
1170 usleep(delay);
1171 if (delay < 200000)
1172 delay *= 2;
a714580e 1173 }
436305c6 1174 if (i == 25)
9b8fea91 1175 pr_err("timeout waiting for %s\n", dev);
a714580e
N
1176}
1177
0f22b998
N
1178struct superswitch *superlist[] =
1179{
1180 &super0, &super1,
1181 &super_ddf, &super_imsm,
0592faeb 1182 &mbr, &gpt,
32141c17
JS
1183 NULL
1184};
f7dd881f 1185
4725bc31 1186struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1187{
1686dc25
NB
1188 mdu_array_info_t array;
1189 int vers;
1190 int minor;
1191 struct supertype *st = NULL;
7e0f6979 1192 struct mdinfo *sra;
142cb9e1 1193 char *verstr;
1686dc25
NB
1194 char version[20];
1195 int i;
f7e7067b 1196 char *subarray = NULL;
4dd2df09 1197 char container[32] = "";
1686dc25 1198
4dd2df09 1199 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1200
1201 if (sra) {
7e0f6979
NB
1202 vers = sra->array.major_version;
1203 minor = sra->array.minor_version;
142cb9e1 1204 verstr = sra->text_version;
1686dc25 1205 } else {
9cd39f01 1206 if (md_get_array_info(fd, &array))
1686dc25
NB
1207 array.major_version = array.minor_version = 0;
1208 vers = array.major_version;
1209 minor = array.minor_version;
142cb9e1 1210 verstr = "";
6fbba4c9 1211 }
82d9eba6 1212
1686dc25
NB
1213 if (vers != -1) {
1214 sprintf(version, "%d.%d", vers, minor);
1215 verstr = version;
6fbba4c9 1216 }
3c558363 1217 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1218 char *dev = verstr+1;
5f7e44b2 1219
f7e7067b 1220 subarray = strchr(dev, '/');
92d49ecf 1221 if (subarray) {
f7e7067b 1222 *subarray++ = '\0';
503975b9 1223 subarray = xstrdup(subarray);
92d49ecf 1224 }
4dd2df09 1225 strcpy(container, dev);
15d230f7 1226 sysfs_free(sra);
5f7e44b2 1227 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1228 if (sra && sra->text_version[0])
1229 verstr = sra->text_version;
1230 else
1231 verstr = "-no-metadata-";
f7e7067b
NB
1232 }
1233
efa29530 1234 for (i = 0; st == NULL && superlist[i]; i++)
f7e7067b 1235 st = superlist[i]->match_metadata_desc(verstr);
1686dc25 1236
15d230f7 1237 sysfs_free(sra);
f7e7067b 1238 if (st) {
3b0896f8 1239 st->sb = NULL;
1f49fb3a
N
1240 if (subarrayp)
1241 *subarrayp = subarray;
4dd2df09
N
1242 strcpy(st->container_devnm, container);
1243 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1244 } else
1245 free(subarray);
5f7e44b2 1246
82d9eba6 1247 return st;
f9ce90ba 1248}
ea24acd0 1249
0f0749ad 1250int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1251{
1252 char buf[20];
1253 int fd;
1254
1255 sprintf(buf, "%d:%d", major(id), minor(id));
1256 fd = dev_open(buf, O_RDONLY);
1257 if (fd < 0)
1258 return 0;
1259 if (get_dev_size(fd, NULL, size)) {
1260 close(fd);
1261 return 1;
1262 }
1263 close(fd);
1264 return 0;
1265}
f9ce90ba 1266
4b57ecf6
AO
1267int dev_sector_size_from_id(dev_t id, unsigned int *size)
1268{
1269 char buf[20];
1270 int fd;
1271
1272 sprintf(buf, "%d:%d", major(id), minor(id));
1273 fd = dev_open(buf, O_RDONLY);
1274 if (fd < 0)
1275 return 0;
1276 if (get_dev_sector_size(fd, NULL, size)) {
1277 close(fd);
1278 return 1;
1279 }
1280 close(fd);
1281 return 0;
1282}
1283
159c3a1a 1284struct supertype *dup_super(struct supertype *orig)
3da92f27 1285{
159c3a1a 1286 struct supertype *st;
1686dc25 1287
d2ca6449
NB
1288 if (!orig)
1289 return orig;
503975b9 1290 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1291 st->ss = orig->ss;
1292 st->max_devs = orig->max_devs;
1293 st->minor_version = orig->minor_version;
cb8f6859 1294 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1295 st->data_offset = orig->data_offset;
159c3a1a
NB
1296 st->sb = NULL;
1297 st->info = NULL;
1298 return st;
3da92f27
NB
1299}
1300
54887ad8 1301struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1302{
1303 /* try each load_super to find the best match,
1304 * and return the best superswitch
1305 */
82d9eba6
NB
1306 struct superswitch *ss;
1307 struct supertype *st;
1d13b599 1308 unsigned int besttime = 0;
570c0542 1309 int bestsuper = -1;
f9ce90ba
NB
1310 int i;
1311
503975b9 1312 st = xcalloc(1, sizeof(*st));
4dd2df09 1313 st->container_devnm[0] = 0;
d1d599ea 1314
efa29530 1315 for (i = 0; superlist[i]; i++) {
f9ce90ba
NB
1316 int rv;
1317 ss = superlist[i];
54887ad8
N
1318 if (guess_type == guess_array && ss->add_to_super == NULL)
1319 continue;
1320 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1321 continue;
ef609477 1322 memset(st, 0, sizeof(*st));
df3346e6 1323 st->ignore_hw_compat = 1;
3da92f27 1324 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1325 if (rv == 0) {
1326 struct mdinfo info;
a5d85af7 1327 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1328 if (bestsuper == -1 ||
1329 besttime < info.array.ctime) {
1330 bestsuper = i;
1331 besttime = info.array.ctime;
570c0542 1332 }
3da92f27 1333 ss->free_super(st);
570c0542
NB
1334 }
1335 }
1336 if (bestsuper != -1) {
1337 int rv;
ef609477 1338 memset(st, 0, sizeof(*st));
df3346e6 1339 st->ignore_hw_compat = 1;
3da92f27 1340 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1341 if (rv == 0) {
5e747af2 1342 superlist[bestsuper]->free_super(st);
82d9eba6 1343 return st;
f9ce90ba
NB
1344 }
1345 }
570c0542 1346 free(st);
f9ce90ba
NB
1347 return NULL;
1348}
fe6729fa 1349
beae1dfe
NB
1350/* Return size of device in bytes */
1351int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1352{
1353 unsigned long long ldsize;
c2c9bb6f
NB
1354 struct stat st;
1355
1356 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1357 ldsize = (unsigned long long)st.st_size;
1358 else
beae1dfe
NB
1359#ifdef BLKGETSIZE64
1360 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1361#endif
1362 {
1363 unsigned long dsize;
1364 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1365 ldsize = dsize;
1366 ldsize <<= 9;
1367 } else {
1368 if (dname)
32971509 1369 pr_err("Cannot get size of %s: %s\n",
beae1dfe
NB
1370 dname, strerror(errno));
1371 return 0;
1372 }
1373 }
1374 *sizep = ldsize;
1375 return 1;
1376}
8fac0577 1377
32971509
PB
1378/* Return sector size of device in bytes */
1379int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep)
1380{
1381 unsigned int sectsize;
1382
1383 if (ioctl(fd, BLKSSZGET, &sectsize) != 0) {
1384 if (dname)
1385 pr_err("Cannot get sector size of %s: %s\n",
1386 dname, strerror(errno));
1387 return 0;
1388 }
1389
1390 *sectsizep = sectsize;
1391 return 1;
1392}
1393
3a371610
N
1394/* Return true if this can only be a container, not a member device.
1395 * i.e. is and md device and size is zero
1396 */
1397int must_be_container(int fd)
1398{
f5c924f4 1399 struct mdinfo *mdi;
3a371610 1400 unsigned long long size;
f5c924f4
JS
1401
1402 mdi = sysfs_read(fd, NULL, GET_VERSION);
1403 if (!mdi)
3a371610 1404 return 0;
f5c924f4
JS
1405 sysfs_free(mdi);
1406
3a371610
N
1407 if (get_dev_size(fd, NULL, &size) == 0)
1408 return 1;
1409 if (size == 0)
1410 return 1;
1411 return 0;
1412}
034b203a
TM
1413
1414/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1415 * Returns: 1 if successful
1416 * -1 for unknown partition type
1417 * 0 for other errors
1418 */
1419static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1420{
056b331e 1421 struct GPT gpt;
034b203a 1422 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1423 struct GPT_part_entry *part;
1424 char buf[512];
034b203a
TM
1425 unsigned long long curr_part_end;
1426 unsigned all_partitions, entry_size;
f21e18ca 1427 unsigned part_nr;
41b06495 1428 unsigned int sector_size = 0;
034b203a
TM
1429
1430 *endofpart = 0;
1431
056b331e 1432 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1433 /* skip protective MBR */
41b06495
MD
1434 if (!get_dev_sector_size(fd, NULL, &sector_size))
1435 return 0;
1436 lseek(fd, sector_size, SEEK_SET);
73e658d8 1437 /* read GPT header */
056b331e 1438 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1439 return 0;
1440
1441 /* get the number of partition entries and the entry size */
056b331e
N
1442 all_partitions = __le32_to_cpu(gpt.part_cnt);
1443 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1444
1445 /* Check GPT signature*/
056b331e 1446 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1447 return -1;
1448
1449 /* sanity checks */
1450 if (all_partitions > 1024 ||
db7fdfe4 1451 entry_size > sizeof(buf))
034b203a
TM
1452 return -1;
1453
db7fdfe4
JS
1454 part = (struct GPT_part_entry *)buf;
1455
41b06495
MD
1456 /* set offset to third block (GPT entries) */
1457 lseek(fd, sector_size*2, SEEK_SET);
ca3b6696 1458 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1459 /* read partition entry */
db7fdfe4 1460 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1461 return 0;
1462
034b203a 1463 /* is this valid partition? */
db7fdfe4 1464 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1465 /* check the last lba for the current partition */
db7fdfe4 1466 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1467 if (curr_part_end > *endofpart)
1468 *endofpart = curr_part_end;
1469 }
1470
034b203a
TM
1471 }
1472 return 1;
1473}
1474
1475/* Sets endofpart parameter to the last block used by the last partition on the device.
1476 * Returns: 1 if successful
1477 * -1 for unknown partition type
1478 * 0 for other errors
1479 */
1480static int get_last_partition_end(int fd, unsigned long long *endofpart)
1481{
056b331e 1482 struct MBR boot_sect;
034b203a 1483 unsigned long long curr_part_end;
f21e18ca 1484 unsigned part_nr;
31208db9 1485 unsigned int sector_size;
034b203a
TM
1486 int retval = 0;
1487
1488 *endofpart = 0;
1489
056b331e 1490 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1491 /* read MBR */
1492 lseek(fd, 0, 0);
056b331e 1493 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1494 goto abort;
1495
1496 /* check MBP signature */
056b331e 1497 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1498 retval = 1;
1499 /* found the correct signature */
034b203a 1500
ca3b6696 1501 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
8e2bca51
JC
1502 /*
1503 * Have to make every access through boot_sect rather
1504 * than using a pointer to the partition table (or an
1505 * entry), since the entries are not properly aligned.
1506 */
1507
034b203a 1508 /* check for GPT type */
8e2bca51
JC
1509 if (boot_sect.parts[part_nr].part_type ==
1510 MBR_GPT_PARTITION_TYPE) {
034b203a
TM
1511 retval = get_gpt_last_partition_end(fd, endofpart);
1512 break;
1513 }
1514 /* check the last used lba for the current partition */
8e2bca51
JC
1515 curr_part_end =
1516 __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) +
1517 __le32_to_cpu(boot_sect.parts[part_nr].blocks_num);
034b203a
TM
1518 if (curr_part_end > *endofpart)
1519 *endofpart = curr_part_end;
034b203a
TM
1520 }
1521 } else {
1522 /* Unknown partition table */
1523 retval = -1;
1524 }
31208db9
MD
1525 /* calculate number of 512-byte blocks */
1526 if (get_dev_sector_size(fd, NULL, &sector_size))
1527 *endofpart *= (sector_size / 512);
034b203a
TM
1528 abort:
1529 return retval;
1530}
1531
53ed6ac3
KW
1532int check_partitions(int fd, char *dname, unsigned long long freesize,
1533 unsigned long long size)
034b203a
TM
1534{
1535 /*
1536 * Check where the last partition ends
1537 */
1538 unsigned long long endofpart;
034b203a 1539
31208db9 1540 if (get_last_partition_end(fd, &endofpart) > 0) {
034b203a
TM
1541 /* There appears to be a partition table here */
1542 if (freesize == 0) {
1543 /* partitions will not be visible in new device */
e7b84f9d
N
1544 pr_err("partition table exists on %s but will be lost or\n"
1545 " meaningless after creating array\n",
1546 dname);
034b203a
TM
1547 return 1;
1548 } else if (endofpart > freesize) {
1549 /* last partition overlaps metadata */
e7b84f9d
N
1550 pr_err("metadata will over-write last partition on %s.\n",
1551 dname);
034b203a 1552 return 1;
53ed6ac3
KW
1553 } else if (size && endofpart > size) {
1554 /* partitions will be truncated in new device */
e7b84f9d
N
1555 pr_err("array size is too small to cover all partitions on %s.\n",
1556 dname);
53ed6ac3 1557 return 1;
034b203a
TM
1558 }
1559 }
1560 return 0;
1561}
1562
a322f70c
DW
1563int open_container(int fd)
1564{
1565 /* 'fd' is a block device. Find out if it is in use
1566 * by a container, and return an open fd on that container.
1567 */
1568 char path[256];
1569 char *e;
1570 DIR *dir;
1571 struct dirent *de;
1572 int dfd, n;
1573 char buf[200];
1574 int major, minor;
1575 struct stat st;
1576
1577 if (fstat(fd, &st) != 0)
1578 return -1;
1579 sprintf(path, "/sys/dev/block/%d:%d/holders",
1580 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1581 e = path + strlen(path);
1582
1583 dir = opendir(path);
1584 if (!dir)
1585 return -1;
1586 while ((de = readdir(dir))) {
1587 if (de->d_ino == 0)
1588 continue;
1589 if (de->d_name[0] == '.')
1590 continue;
a994592d
N
1591 /* Need to make sure it is a container and not a volume */
1592 sprintf(e, "/%s/md/metadata_version", de->d_name);
1593 dfd = open(path, O_RDONLY);
1594 if (dfd < 0)
1595 continue;
1596 n = read(dfd, buf, sizeof(buf));
1597 close(dfd);
1598 if (n <= 0 || (unsigned)n >= sizeof(buf))
1599 continue;
1600 buf[n] = 0;
1601 if (strncmp(buf, "external", 8) != 0 ||
1602 n < 10 ||
1603 buf[9] == '/')
1604 continue;
a322f70c
DW
1605 sprintf(e, "/%s/dev", de->d_name);
1606 dfd = open(path, O_RDONLY);
1607 if (dfd < 0)
1608 continue;
1609 n = read(dfd, buf, sizeof(buf));
1610 close(dfd);
f21e18ca 1611 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1612 continue;
1613 buf[n] = 0;
1614 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1615 continue;
1616 sprintf(buf, "%d:%d", major, minor);
1617 dfd = dev_open(buf, O_RDONLY);
1618 if (dfd >= 0) {
1619 closedir(dir);
1620 return dfd;
1621 }
1622 }
355726fa 1623 closedir(dir);
a322f70c
DW
1624 return -1;
1625}
1626
33414a01
DW
1627struct superswitch *version_to_superswitch(char *vers)
1628{
1629 int i;
1630
1631 for (i = 0; superlist[i]; i++) {
1632 struct superswitch *ss = superlist[i];
1633
1634 if (strcmp(vers, ss->name) == 0)
1635 return ss;
1636 }
1637
1638 return NULL;
1639}
1640
4dd2df09
N
1641int metadata_container_matches(char *metadata, char *devnm)
1642{
1643 /* Check if 'devnm' is the container named in 'metadata'
1644 * which is
1645 * /containername/componentname or
1646 * -containername/componentname
1647 */
1648 int l;
1649 if (*metadata != '/' && *metadata != '-')
1650 return 0;
1651 l = strlen(devnm);
1652 if (strncmp(metadata+1, devnm, l) != 0)
1653 return 0;
1654 if (metadata[l+1] != '/')
1655 return 0;
1656 return 1;
1657}
1658
1659int metadata_subdev_matches(char *metadata, char *devnm)
1660{
1661 /* Check if 'devnm' is the subdev named in 'metadata'
1662 * which is
1663 * /containername/subdev or
1664 * -containername/subdev
1665 */
1666 char *sl;
1667 if (*metadata != '/' && *metadata != '-')
1668 return 0;
1669 sl = strchr(metadata+1, '/');
1670 if (!sl)
1671 return 0;
1672 if (strcmp(sl+1, devnm) == 0)
1673 return 1;
1674 return 0;
1675}
1676
33414a01
DW
1677int is_container_member(struct mdstat_ent *mdstat, char *container)
1678{
1679 if (mdstat->metadata_version == NULL ||
1680 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1681 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1682 return 0;
1683
1684 return 1;
1685}
1686
1687int is_subarray_active(char *subarray, char *container)
1688{
1689 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1690 struct mdstat_ent *ent;
1691
e5408a32
DW
1692 for (ent = mdstat; ent; ent = ent->next)
1693 if (is_container_member(ent, container))
e5e5d7ce 1694 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1695 break;
33414a01
DW
1696
1697 free_mdstat(mdstat);
1698
1699 return ent != NULL;
1700}
1701
1702/* open_subarray - opens a subarray in a container
1703 * @dev: container device name
feab51f8 1704 * @st: empty supertype
33414a01
DW
1705 * @quiet: block reporting errors flag
1706 *
1707 * On success returns an fd to a container and fills in *st
1708 */
feab51f8 1709int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1710{
1711 struct mdinfo *mdi;
a951a4f7 1712 struct mdinfo *info;
33414a01 1713 int fd, err = 1;
4dd2df09 1714 char *_devnm;
33414a01
DW
1715
1716 fd = open(dev, O_RDWR|O_EXCL);
1717 if (fd < 0) {
1718 if (!quiet)
e7b84f9d 1719 pr_err("Couldn't open %s, aborting\n",
33414a01 1720 dev);
b990032d 1721 return -1;
33414a01
DW
1722 }
1723
4dd2df09
N
1724 _devnm = fd2devnm(fd);
1725 if (_devnm == NULL) {
33414a01 1726 if (!quiet)
e7b84f9d
N
1727 pr_err("Failed to determine device number for %s\n",
1728 dev);
33414a01
DW
1729 goto close_fd;
1730 }
4dd2df09 1731 strcpy(st->devnm, _devnm);
33414a01 1732
4dd2df09 1733 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1734 if (!mdi) {
1735 if (!quiet)
e7b84f9d 1736 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1737 dev);
1738 goto close_fd;
1739 }
1740
1741 if (mdi->array.level != UnSet) {
1742 if (!quiet)
e7b84f9d 1743 pr_err("%s is not a container\n", dev);
33414a01
DW
1744 goto free_sysfs;
1745 }
1746
1747 st->ss = version_to_superswitch(mdi->text_version);
1748 if (!st->ss) {
1749 if (!quiet)
e7b84f9d
N
1750 pr_err("Operation not supported for %s metadata\n",
1751 mdi->text_version);
33414a01
DW
1752 goto free_sysfs;
1753 }
1754
4dd2df09 1755 if (st->devnm[0] == 0) {
33414a01 1756 if (!quiet)
e7b84f9d 1757 pr_err("Failed to allocate device name\n");
33414a01
DW
1758 goto free_sysfs;
1759 }
1760
db20d413 1761 if (!st->ss->load_container) {
33414a01 1762 if (!quiet)
e7b84f9d 1763 pr_err("%s is not a container\n", dev);
4dd2df09 1764 goto free_sysfs;
33414a01
DW
1765 }
1766
db20d413 1767 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1768 if (!quiet)
e7b84f9d 1769 pr_err("Failed to load metadata for %s\n",
db20d413 1770 dev);
4dd2df09 1771 goto free_sysfs;
33414a01
DW
1772 }
1773
a951a4f7
N
1774 info = st->ss->container_content(st, subarray);
1775 if (!info) {
1776 if (!quiet)
e7b84f9d 1777 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1778 subarray, dev);
1779 goto free_super;
1780 }
1781 free(info);
1782
33414a01
DW
1783 err = 0;
1784
1785 free_super:
1786 if (err)
1787 st->ss->free_super(st);
33414a01
DW
1788 free_sysfs:
1789 sysfs_free(mdi);
1790 close_fd:
1791 if (err)
1792 close(fd);
1793
1794 if (err)
1795 return -1;
1796 else
1797 return fd;
1798}
1799
7801ac20
N
1800int add_disk(int mdfd, struct supertype *st,
1801 struct mdinfo *sra, struct mdinfo *info)
1802{
1803 /* Add a device to an array, in one of 2 ways. */
1804 int rv;
32141c17 1805
7801ac20 1806 if (st->ss->external) {
d23534e4
DW
1807 if (info->disk.state & (1<<MD_DISK_SYNC))
1808 info->recovery_start = MaxSector;
1809 else
1810 info->recovery_start = 0;
2904b26f 1811 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1812 if (! rv) {
1813 struct mdinfo *sd2;
f35f2525
N
1814 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1815 if (sd2 == info)
1816 break;
1817 if (sd2 == NULL) {
503975b9 1818 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1819 *sd2 = *info;
1820 sd2->next = sra->devs;
1821 sra->devs = sd2;
1822 }
7801ac20
N
1823 }
1824 } else
7801ac20
N
1825 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1826 return rv;
1827}
1828
de6ae750
N
1829int remove_disk(int mdfd, struct supertype *st,
1830 struct mdinfo *sra, struct mdinfo *info)
1831{
1832 int rv;
32141c17 1833
de6ae750 1834 /* Remove the disk given by 'info' from the array */
de6ae750
N
1835 if (st->ss->external)
1836 rv = sysfs_set_str(sra, info, "slot", "none");
1837 else
de6ae750
N
1838 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1839 info->disk.minor));
1840 return rv;
1841}
1842
1ab9ed2a 1843int hot_remove_disk(int mdfd, unsigned long dev, int force)
2dd271fe 1844{
1ab9ed2a 1845 int cnt = force ? 500 : 5;
2dd271fe
N
1846 int ret;
1847
1848 /* HOT_REMOVE_DISK can fail with EBUSY if there are
1849 * outstanding IO requests to the device.
1850 * In this case, it can be helpful to wait a little while,
1ab9ed2a 1851 * up to 5 seconds if 'force' is set, or 50 msec if not.
2dd271fe
N
1852 */
1853 while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 &&
1854 errno == EBUSY &&
1855 cnt-- > 0)
1856 usleep(10000);
1857
1858 return ret;
1859}
1860
1ab9ed2a 1861int sys_hot_remove_disk(int statefd, int force)
fdd01569 1862{
1ab9ed2a 1863 int cnt = force ? 500 : 5;
fdd01569
N
1864 int ret;
1865
1866 while ((ret = write(statefd, "remove", 6)) == -1 &&
1867 errno == EBUSY &&
1868 cnt-- > 0)
1869 usleep(10000);
1870 return ret == 6 ? 0 : -1;
1871}
1872
f35f2525
N
1873int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1874{
1875 /* Initialise kernel's knowledge of array.
1876 * This varies between externally managed arrays
1877 * and older kernels
1878 */
700483a2 1879 mdu_array_info_t inf;
f35f2525
N
1880 int rv;
1881
f35f2525 1882 if (st->ss->external)
700483a2 1883 return sysfs_set_array(info, 9003);
700483a2
JS
1884
1885 memset(&inf, 0, sizeof(inf));
1886 inf.major_version = info->array.major_version;
1887 inf.minor_version = info->array.minor_version;
1888 rv = md_set_array_info(mdfd, &inf);
1889
f35f2525
N
1890 return rv;
1891}
1892
1e5c6983
DW
1893unsigned long long min_recovery_start(struct mdinfo *array)
1894{
1895 /* find the minimum recovery_start in an array for metadata
1896 * formats that only record per-array recovery progress instead
1897 * of per-device
1898 */
1899 unsigned long long recovery_start = MaxSector;
1900 struct mdinfo *d;
1901
1902 for (d = array->devs; d; d = d->next)
1903 recovery_start = min(recovery_start, d->recovery_start);
1904
1905 return recovery_start;
1906}
1907
4dd2df09 1908int mdmon_pid(char *devnm)
a931db9e
NB
1909{
1910 char path[100];
1911 char pid[10];
1912 int fd;
1913 int n;
10013317 1914
4dd2df09 1915 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1916
24f6f99b 1917 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1918
1919 if (fd < 0)
cf556303 1920 return -1;
a931db9e
NB
1921 n = read(fd, pid, 9);
1922 close(fd);
1923 if (n <= 0)
cf556303 1924 return -1;
24f6f99b 1925 return atoi(pid);
a931db9e
NB
1926}
1927
4dd2df09 1928int mdmon_running(char *devnm)
a931db9e 1929{
4dd2df09 1930 int pid = mdmon_pid(devnm);
24f6f99b 1931 if (pid <= 0)
a931db9e 1932 return 0;
24f6f99b 1933 if (kill(pid, 0) == 0)
a931db9e
NB
1934 return 1;
1935 return 0;
1936}
1937
4dd2df09 1938int start_mdmon(char *devnm)
8850ee3e 1939{
1913c325 1940 int i, skipped;
44d2e365 1941 int len;
ca3b6696 1942 pid_t pid;
9fe32043 1943 int status;
44d2e365
N
1944 char pathbuf[1024];
1945 char *paths[4] = {
1946 pathbuf,
85945e19 1947 BINDIR "/mdmon",
2f1bcf43 1948 "./mdmon",
44d2e365
N
1949 NULL
1950 };
8850ee3e 1951
40ebbb9c 1952 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1953 return 0;
1954
9cf014ec 1955 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1956 if (len > 0) {
1957 char *sl;
1958 pathbuf[len] = 0;
1959 sl = strrchr(pathbuf, '/');
1960 if (sl)
1961 sl++;
1962 else
1963 sl = pathbuf;
1964 strcpy(sl, "mdmon");
1965 } else
1966 pathbuf[0] = '\0';
1967
0f7bdf89 1968 /* First try to run systemctl */
9540cc24
N
1969 if (!check_env("MDADM_NO_SYSTEMCTL"))
1970 switch(fork()) {
1971 case 0:
1972 /* FIXME yuk. CLOSE_EXEC?? */
1973 skipped = 0;
1974 for (i = 3; skipped < 20; i++)
1975 if (close(i) < 0)
1976 skipped++;
1977 else
1978 skipped = 0;
1979
1980 /* Don't want to see error messages from
1981 * systemctl. If the service doesn't exist,
1982 * we start mdmon ourselves.
1983 */
1984 close(2);
1985 open("/dev/null", O_WRONLY);
1986 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1987 devnm);
1988 status = execl("/usr/bin/systemctl", "systemctl",
1989 "start",
1990 pathbuf, NULL);
1991 status = execl("/bin/systemctl", "systemctl", "start",
1992 pathbuf, NULL);
1993 exit(1);
7a862a02 1994 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1995 return -1;
1996 default: /* parent - good */
1997 pid = wait(&status);
1998 if (pid >= 0 && status == 0)
1999 return 0;
2000 }
0f7bdf89
JS
2001
2002 /* That failed, try running mdmon directly */
8850ee3e
N
2003 switch(fork()) {
2004 case 0:
2005 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 2006 skipped = 0;
ca3b6696 2007 for (i = 3; skipped < 20; i++)
1913c325
N
2008 if (close(i) < 0)
2009 skipped++;
2010 else
2011 skipped = 0;
2012
ca3b6696 2013 for (i = 0; paths[i]; i++)
a0963a86 2014 if (paths[i][0]) {
2f1bcf43 2015 execl(paths[i], paths[i],
4dd2df09 2016 devnm, NULL);
a0963a86 2017 }
8850ee3e 2018 exit(1);
7a862a02 2019 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 2020 return -1;
9fe32043
N
2021 default: /* parent - good */
2022 pid = wait(&status);
15c10423 2023 if (pid < 0 || status != 0) {
7a862a02 2024 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 2025 return -1;
15c10423 2026 }
8850ee3e
N
2027 }
2028 return 0;
2029}
2030
148acb7b
DW
2031__u32 random32(void)
2032{
2033 __u32 rv;
2034 int rfd = open("/dev/urandom", O_RDONLY);
2035 if (rfd < 0 || read(rfd, &rv, 4) != 4)
2036 rv = random();
2037 if (rfd >= 0)
2038 close(rfd);
2039 return rv;
2040}
2041
c5f71c24
JS
2042void random_uuid(__u8 *buf)
2043{
2044 int fd, i, len;
2045 __u32 r[4];
2046
2047 fd = open("/dev/urandom", O_RDONLY);
2048 if (fd < 0)
2049 goto use_random;
2050 len = read(fd, buf, 16);
2051 close(fd);
2052 if (len != 16)
2053 goto use_random;
2054
2055 return;
2056
2057use_random:
2058 for (i = 0; i < 4; i++)
2059 r[i] = random();
2060 memcpy(buf, r, 16);
2061}
2062
edd8d13c
NB
2063int flush_metadata_updates(struct supertype *st)
2064{
2065 int sfd;
2066 if (!st->updates) {
2067 st->update_tail = NULL;
2068 return -1;
2069 }
2070
4dd2df09 2071 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
2072 if (sfd < 0)
2073 return -1;
2074
2075 while (st->updates) {
2076 struct metadata_update *mu = st->updates;
2077 st->updates = mu->next;
2078
2079 send_message(sfd, mu, 0);
2080 wait_reply(sfd, 0);
2081 free(mu->buf);
2082 free(mu);
2083 }
2084 ack(sfd, 0);
2085 wait_reply(sfd, 0);
2086 close(sfd);
2087 st->update_tail = NULL;
2088 return 0;
2089}
2090
2091void append_metadata_update(struct supertype *st, void *buf, int len)
2092{
2093
503975b9 2094 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
2095
2096 mu->buf = buf;
2097 mu->len = len;
2098 mu->space = NULL;
cb23f1f4 2099 mu->space_list = NULL;
edd8d13c
NB
2100 mu->next = NULL;
2101 *st->update_tail = mu;
2102 st->update_tail = &mu->next;
2103}
a931db9e 2104
fe6729fa
NB
2105#ifdef __TINYC__
2106/* tinyc doesn't optimize this check in ioctl.h out ... */
2107unsigned int __invalid_size_argument_for_IOC = 0;
2108#endif
2109
6d11ec6f
AK
2110int experimental(void)
2111{
2112 if (check_env("MDADM_EXPERIMENTAL"))
2113 return 1;
2114 else {
7a862a02 2115 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
2116 return 0;
2117 }
2118}
2119
326727d9
AC
2120/* Pick all spares matching given criteria from a container
2121 * if min_size == 0 do not check size
2122 * if domlist == NULL do not check domains
2123 * if spare_group given add it to domains of each spare
2124 * metadata allows to test domains using metadata of destination array */
2125struct mdinfo *container_choose_spares(struct supertype *st,
fbfdcb06 2126 struct spare_criteria *criteria,
326727d9
AC
2127 struct domainlist *domlist,
2128 char *spare_group,
2129 const char *metadata, int get_one)
2130{
2131 struct mdinfo *d, **dp, *disks = NULL;
2132
2133 /* get list of all disks in container */
2134 if (st->ss->getinfo_super_disks)
2135 disks = st->ss->getinfo_super_disks(st);
2136
2137 if (!disks)
2138 return disks;
2139 /* find spare devices on the list */
2140 dp = &disks->devs;
2141 disks->array.spare_disks = 0;
2142 while (*dp) {
2143 int found = 0;
2144 d = *dp;
2145 if (d->disk.state == 0) {
2146 /* check if size is acceptable */
2147 unsigned long long dev_size;
4b57ecf6
AO
2148 unsigned int dev_sector_size;
2149 int size_valid = 0;
2150 int sector_size_valid = 0;
2151
326727d9
AC
2152 dev_t dev = makedev(d->disk.major,d->disk.minor);
2153
fbfdcb06 2154 if (!criteria->min_size ||
326727d9 2155 (dev_size_from_id(dev, &dev_size) &&
fbfdcb06 2156 dev_size >= criteria->min_size))
4b57ecf6
AO
2157 size_valid = 1;
2158
2159 if (!criteria->sector_size ||
2160 (dev_sector_size_from_id(dev, &dev_sector_size) &&
2161 criteria->sector_size == dev_sector_size))
2162 sector_size_valid = 1;
2163
2164 found = size_valid && sector_size_valid;
2165
326727d9
AC
2166 /* check if domain matches */
2167 if (found && domlist) {
4dd2df09 2168 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2169 if (spare_group)
2170 pol_add(&pol, pol_domain,
2171 spare_group, NULL);
e5508b36 2172 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2173 found = 0;
2174 dev_policy_free(pol);
2175 }
2176 }
2177 if (found) {
2178 dp = &d->next;
2179 disks->array.spare_disks++;
2180 if (get_one) {
2181 sysfs_free(*dp);
2182 d->next = NULL;
2183 }
2184 } else {
2185 *dp = d->next;
2186 d->next = NULL;
2187 sysfs_free(d);
2188 }
2189 }
2190 return disks;
2191}
9eafa1de
MN
2192
2193/* Checks if paths point to the same device
2194 * Returns 0 if they do.
2195 * Returns 1 if they don't.
2196 * Returns -1 if something went wrong,
2197 * e.g. paths are empty or the files
2198 * they point to don't exist */
2199int compare_paths (char* path1, char* path2)
2200{
2201 struct stat st1,st2;
2202
2203 if (path1 == NULL || path2 == NULL)
2204 return -1;
2205 if (stat(path1,&st1) != 0)
2206 return -1;
2207 if (stat(path2,&st2) != 0)
2208 return -1;
2209 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2210 return 0;
2211 return 1;
2212}
a7dec3fd
N
2213
2214/* Make sure we can open as many devices as needed */
2215void enable_fds(int devices)
2216{
2217 unsigned int fds = 20 + devices;
2218 struct rlimit lim;
d7be7d87 2219 if (getrlimit(RLIMIT_NOFILE, &lim) != 0 || lim.rlim_cur >= fds)
a7dec3fd
N
2220 return;
2221 if (lim.rlim_max < fds)
2222 lim.rlim_max = fds;
2223 lim.rlim_cur = fds;
2224 setrlimit(RLIMIT_NOFILE, &lim);
2225}
a9c15847
N
2226
2227int in_initrd(void)
2228{
2229 /* This is based on similar function in systemd. */
2230 struct statfs s;
76d0f188
JS
2231 /* statfs.f_type is signed long on s390x and MIPS, causing all
2232 sorts of sign extension problems with RAMFS_MAGIC being
2233 defined as 0x858458f6 */
a9c15847 2234 return statfs("/", &s) >= 0 &&
7ac5d47e 2235 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2236 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2237 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2238}
8832342d
N
2239
2240void reopen_mddev(int mdfd)
2241{
2242 /* Re-open without any O_EXCL, but keep
2243 * the same fd
2244 */
2245 char *devnm;
2246 int fd;
2247 devnm = fd2devnm(mdfd);
2248 close(mdfd);
2249 fd = open_dev(devnm);
2250 if (fd >= 0 && fd != mdfd)
2251 dup2(fd, mdfd);
2252}
7716570e 2253
e80357f8
GJ
2254static struct cmap_hooks *cmap_hooks = NULL;
2255static int is_cmap_hooks_ready = 0;
7716570e 2256
e80357f8
GJ
2257void set_cmap_hooks(void)
2258{
2259 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2260 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2261 if (!cmap_hooks->cmap_handle)
2262 return;
7716570e 2263
e80357f8
GJ
2264 cmap_hooks->initialize = dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2265 cmap_hooks->get_string = dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
2266 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2267
e80357f8
GJ
2268 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2269 !cmap_hooks->finalize)
2270 dlclose(cmap_hooks->cmap_handle);
2271 else
2272 is_cmap_hooks_ready = 1;
2273}
7716570e 2274
e80357f8
GJ
2275int get_cluster_name(char **cluster_name)
2276{
2277 int rv = -1;
2278 cmap_handle_t handle;
7716570e 2279
e80357f8
GJ
2280 if (!is_cmap_hooks_ready)
2281 return rv;
7716570e 2282
e80357f8 2283 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2284 if (rv != CS_OK)
2285 goto out;
2286
e80357f8 2287 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2288 if (rv != CS_OK) {
2289 free(*cluster_name);
2290 rv = -1;
2291 goto name_err;
2292 }
2293
2294 rv = 0;
2295name_err:
e80357f8 2296 cmap_hooks->finalize(handle);
7716570e 2297out:
7716570e
GJ
2298 return rv;
2299}
d15a1f72
GJ
2300
2301void set_dlm_hooks(void)
2302{
2303 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2304 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2305 if (!dlm_hooks->dlm_handle)
2306 return;
2307
2308 dlm_hooks->create_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2309 dlm_hooks->release_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
2310 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2311 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2312 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2313 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2314
2315 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2316 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2317 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2318 dlclose(dlm_hooks->dlm_handle);
2319 else
2320 is_dlm_hooks_ready = 1;
2321}
e80357f8
GJ
2322
2323void set_hooks(void)
2324{
2325 set_dlm_hooks();
2326 set_cmap_hooks();
2327}