super-ddf: sysfs_read() takes a pointer as device name argument
[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;
231 int ret;
232
233 sra = sysfs_read(fd, NULL, GET_ARRAY_STATE);
234 if (sra) {
235 if (sra->array_state != ARRAY_CLEAR &&
236 sra->array_state != ARRAY_INACTIVE &&
237 sra->array_state != ARRAY_UNKNOWN_STATE)
238 ret = 0;
239 else
240 ret = -ENODEV;
241
242 free(sra);
243 } else {
244 /*
245 * GET_ARRAY_INFO doesn't provide access to the proper state
246 * information, so fallback to a basic check for raid_disks != 0
247 */
248 ret = ioctl(fd, GET_ARRAY_INFO, &array);
249 }
250
251 return !ret;
252}
253
9cd39f01
JS
254/*
255 * Get array info from the kernel. Longer term we want to deprecate the
256 * ioctl and get it from sysfs.
257 */
258int md_get_array_info(int fd, struct mdu_array_info_s *array)
259{
260 return ioctl(fd, GET_ARRAY_INFO, array);
261}
262
018a4882
JS
263/*
264 * Set array info
265 */
266int md_set_array_info(int fd, struct mdu_array_info_s *array)
267{
268 return ioctl(fd, SET_ARRAY_INFO, array);
269}
270
d97572f5
JS
271/*
272 * Get disk info from the kernel.
273 */
274int md_get_disk_info(int fd, struct mdu_disk_info_s *disk)
275{
276 return ioctl(fd, GET_DISK_INFO, disk);
277}
278
64c4757e
NB
279/*
280 * Parse a 128 bit uuid in 4 integers
281 * format is 32 hexx nibbles with options :.<space> separator
282 * If not exactly 32 hex digits are found, return 0
283 * else return 1
284 */
285int parse_uuid(char *str, int uuid[4])
286{
aba69144
NB
287 int hit = 0; /* number of Hex digIT */
288 int i;
289 char c;
ca3b6696
N
290 for (i = 0; i < 4; i++)
291 uuid[i] = 0;
aba69144 292
ca3b6696 293 while ((c = *str++) != 0) {
aba69144 294 int n;
ca3b6696 295 if (c >= '0' && c <= '9')
aba69144 296 n = c-'0';
ca3b6696 297 else if (c >= 'a' && c <= 'f')
aba69144 298 n = 10 + c - 'a';
ca3b6696 299 else if (c >= 'A' && c <= 'F')
aba69144
NB
300 n = 10 + c - 'A';
301 else if (strchr(":. -", c))
302 continue;
303 else return 0;
304
305 if (hit<32) {
306 uuid[hit/8] <<= 4;
307 uuid[hit/8] += n;
308 }
309 hit++;
82b27616 310 }
aba69144
NB
311 if (hit == 32)
312 return 1;
313 return 0;
64c4757e
NB
314}
315
64c4757e
NB
316int get_linux_version()
317{
318 struct utsname name;
98c6faba 319 char *cp;
f161d047 320 int a = 0, b = 0,c = 0;
64c4757e
NB
321 if (uname(&name) <0)
322 return -1;
323
98c6faba
NB
324 cp = name.release;
325 a = strtoul(cp, &cp, 10);
f161d047
N
326 if (*cp == '.')
327 b = strtoul(cp+1, &cp, 10);
328 if (*cp == '.')
329 c = strtoul(cp+1, &cp, 10);
98c6faba 330
682c7051 331 return (a*1000000)+(b*1000)+c;
64c4757e
NB
332}
333
bc77ed53
DW
334int mdadm_version(char *version)
335{
336 int a, b, c;
337 char *cp;
338
339 if (!version)
340 version = Version;
341
342 cp = strchr(version, '-');
343 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
344 return -1;
345 cp += 3;
346 a = strtoul(cp, &cp, 10);
347 if (*cp != '.')
348 return -1;
349 b = strtoul(cp+1, &cp, 10);
350 if (*cp == '.')
351 c = strtoul(cp+1, &cp, 10);
352 else
353 c = 0;
354 if (*cp != ' ' && *cp != '-')
355 return -1;
356 return (a*1000000)+(b*1000)+c;
357}
358
51d4261c 359unsigned long long parse_size(char *size)
84e11361
N
360{
361 /* parse 'size' which should be a number optionally
362 * followed by 'K', 'M', or 'G'.
363 * Without a suffix, K is assumed.
364 * Number returned is in sectors (half-K)
822e393a 365 * INVALID_SECTORS returned on error.
84e11361
N
366 */
367 char *c;
368 long long s = strtoll(size, &c, 10);
369 if (s > 0) {
370 switch (*c) {
371 case 'K':
372 c++;
373 default:
374 s *= 2;
375 break;
376 case 'M':
377 c++;
378 s *= 1024 * 2;
379 break;
380 case 'G':
381 c++;
382 s *= 1024 * 1024 * 2;
383 break;
72ca9bcf
N
384 case 's': /* sectors */
385 c++;
386 break;
84e11361 387 }
51d4261c 388 } else
822e393a 389 s = INVALID_SECTORS;
84e11361 390 if (*c)
822e393a 391 s = INVALID_SECTORS;
84e11361
N
392 return s;
393}
394
4a06e2c2
N
395int parse_layout_10(char *layout)
396{
397 int copies, rv;
398 char *cp;
399 /* Parse the layout string for raid10 */
400 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
401 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
402 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
403 copies > 200 ||
404 *cp)
405 return -1;
406 if (layout[0] == 'n')
407 rv = 256 + copies;
408 else if (layout[0] == 'o')
409 rv = 0x10000 + (copies<<8) + 1;
410 else
411 rv = 1 + (copies<<8);
412 return rv;
413}
414
415int parse_layout_faulty(char *layout)
416{
417 /* Parse the layout string for 'faulty' */
418 int ln = strcspn(layout, "0123456789");
503975b9 419 char *m = xstrdup(layout);
4a06e2c2
N
420 int mode;
421 m[ln] = 0;
422 mode = map_name(faultylayout, m);
423 if (mode == UnSet)
424 return -1;
425
426 return mode | (atoi(layout+ln)<< ModeShift);
427}
79868890
N
428
429long parse_num(char *num)
430{
431 /* Either return a valid number, or -1 */
432 char *c;
433 long rv = strtol(num, &c, 10);
434 if (rv < 0 || *c || !num[0])
435 return -1;
436 else
437 return rv;
438}
4a06e2c2 439
4de90913
GJ
440int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
441{
442 char *dev;
443 *slot = strtoul(input, &dev, 10);
444 if (dev == input || dev[0] != ':')
445 return -1;
446 *devname = dev+1;
447 return 0;
448}
449
0430ed48
NB
450void remove_partitions(int fd)
451{
452 /* remove partitions from this block devices.
453 * This is used for components added to an array
454 */
455#ifdef BLKPG_DEL_PARTITION
456 struct blkpg_ioctl_arg a;
457 struct blkpg_partition p;
458
459 a.op = BLKPG_DEL_PARTITION;
460 a.data = (void*)&p;
461 a.datalen = sizeof(p);
462 a.flags = 0;
463 memset(a.data, 0, a.datalen);
ca3b6696 464 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
465 ioctl(fd, BLKPG, &a);
466#endif
467}
468
691c6ee1
N
469int test_partition(int fd)
470{
471 /* Check if fd is a whole-disk or a partition.
472 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
473 * will return ENXIO on an invalid partition number.
474 */
475 struct blkpg_ioctl_arg a;
476 struct blkpg_partition p;
477 a.op = BLKPG_DEL_PARTITION;
478 a.data = (void*)&p;
479 a.datalen = sizeof(p);
480 a.flags = 0;
481 memset(a.data, 0, a.datalen);
482 p.pno = 1<<30;
483 if (ioctl(fd, BLKPG, &a) == 0)
484 /* Very unlikely, but not a partition */
485 return 0;
357ac106 486 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
487 /* not a partition */
488 return 0;
489
490 return 1;
491}
492
bfd76b93
CA
493int test_partition_from_id(dev_t id)
494{
495 char buf[20];
496 int fd, rv;
497
498 sprintf(buf, "%d:%d", major(id), minor(id));
499 fd = dev_open(buf, O_RDONLY);
500 if (fd < 0)
501 return -1;
502 rv = test_partition(fd);
503 close(fd);
504 return rv;
505}
691c6ee1 506
de5a472e 507int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 508{
265e0f17 509 int copies, first;
de5a472e
N
510 int i;
511 int avail_disks = 0;
512
513 for (i = 0; i < raid_disks; i++)
514 avail_disks += !!avail[i];
515
64c4757e 516 switch (level) {
265e0f17
NB
517 case 10:
518 /* This is the tricky one - we need to check
519 * which actual disks are present.
520 */
702b557b 521 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 522 first = 0;
265e0f17
NB
523 do {
524 /* there must be one of the 'copies' form 'first' */
525 int n = copies;
ca3b6696 526 int cnt = 0;
b3ec716d 527 int this = first;
265e0f17 528 while (n--) {
b3ec716d 529 if (avail[this])
265e0f17 530 cnt++;
b3ec716d 531 this = (this+1) % raid_disks;
265e0f17
NB
532 }
533 if (cnt == 0)
534 return 0;
b3ec716d 535 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
536 } while (first != 0);
537 return 1;
e5329c37 538
df0d4ea0 539 case LEVEL_MULTIPATH:
e0d19036 540 return avail_disks>= 1;
df0d4ea0 541 case LEVEL_LINEAR:
64c4757e
NB
542 case 0:
543 return avail_disks == raid_disks;
544 case 1:
545 return avail_disks >= 1;
546 case 4:
330d6900
N
547 if (avail_disks == raid_disks - 1 &&
548 !avail[raid_disks - 1])
549 /* If just the parity device is missing, then we
550 * have enough, even if not clean
551 */
552 return 1;
553 /* FALL THROUGH */
64c4757e 554 case 5:
583315d9
NB
555 if (clean)
556 return avail_disks >= raid_disks-1;
557 else
558 return avail_disks >= raid_disks;
98c6faba 559 case 6:
583315d9
NB
560 if (clean)
561 return avail_disks >= raid_disks-2;
562 else
563 return avail_disks >= raid_disks;
64c4757e
NB
564 default:
565 return 0;
566 }
567}
568
22e263f6
AC
569const int uuid_zero[4] = { 0, 0, 0, 0 };
570
f277ce36 571int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 572{
f277ce36
NB
573 if (swapuuid) {
574 /* parse uuids are hostendian.
575 * uuid's from some superblocks are big-ending
aba69144 576 * if there is a difference, we need to swap..
f277ce36
NB
577 */
578 unsigned char *ac = (unsigned char *)a;
579 unsigned char *bc = (unsigned char *)b;
580 int i;
ca3b6696 581 for (i = 0; i < 16; i += 4) {
f277ce36
NB
582 if (ac[i+0] != bc[i+3] ||
583 ac[i+1] != bc[i+2] ||
584 ac[i+2] != bc[i+1] ||
585 ac[i+3] != bc[i+0])
586 return 0;
587 }
588 return 1;
589 } else {
590 if (a[0]==b[0] &&
591 a[1]==b[1] &&
592 a[2]==b[2] &&
593 a[3]==b[3])
594 return 1;
595 return 0;
596 }
64c4757e 597}
ca3b6696 598
350f29f9
NB
599void copy_uuid(void *a, int b[4], int swapuuid)
600{
601 if (swapuuid) {
602 /* parse uuids are hostendian.
603 * uuid's from some superblocks are big-ending
604 * if there is a difference, we need to swap..
605 */
606 unsigned char *ac = (unsigned char *)a;
607 unsigned char *bc = (unsigned char *)b;
608 int i;
ca3b6696 609 for (i = 0; i < 16; i += 4) {
350f29f9
NB
610 ac[i+0] = bc[i+3];
611 ac[i+1] = bc[i+2];
612 ac[i+2] = bc[i+1];
613 ac[i+3] = bc[i+0];
614 }
615 } else
616 memcpy(a, b, 16);
617}
64c4757e 618
aae5a112 619char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 620{
9968e376 621 int i, j;
d7288ddc
N
622 char uuid[16];
623 char *c = buf;
624 strcpy(c, "UUID-");
625 c += strlen(c);
aae5a112 626 copy_uuid(uuid, id, swap);
9968e376 627 for (i = 0; i < 4; i++) {
9968e376 628 if (i)
ff54de6e 629 *c++ = sep;
9968e376
DW
630 for (j = 3; j >= 0; j--) {
631 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
632 c+= 2;
633 }
d7288ddc
N
634 }
635 return buf;
aae5a112
DW
636
637}
638
639char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
640{
86983cce
N
641 // dirty hack to work around an issue with super1 superblocks...
642 // super1 superblocks need swapuuid set in order for assembly to
643 // work, but can't have it set if we want this printout to match
644 // all the other uuid printouts in super1.c, so we force swapuuid
645 // to 1 to make our printout match the rest of super1
646 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
647}
648
682c7051
NB
649int check_ext2(int fd, char *name)
650{
651 /*
652 * Check for an ext2fs file system.
653 * Superblock is always 1K at 1K offset
654 *
655 * s_magic is le16 at 56 == 0xEF53
656 * report mtime - le32 at 44
657 * blocks - le32 at 4
658 * logblksize - le32 at 24
659 */
660 unsigned char sb[1024];
661 time_t mtime;
2538ba2a
N
662 unsigned long long size;
663 int bsize;
682c7051
NB
664 if (lseek(fd, 1024,0)!= 1024)
665 return 0;
666 if (read(fd, sb, 1024)!= 1024)
667 return 0;
668 if (sb[56] != 0x53 || sb[57] != 0xef)
669 return 0;
670
671 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
672 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
673 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 674 size <<= bsize;
e7b84f9d 675 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 676 name);
2538ba2a 677 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
678 return 1;
679}
680
681int check_reiser(int fd, char *name)
682{
683 /*
684 * superblock is at 64K
685 * size is 1024;
686 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
687 *
688 */
689 unsigned char sb[1024];
2538ba2a 690 unsigned long long size;
682c7051
NB
691 if (lseek(fd, 64*1024, 0) != 64*1024)
692 return 0;
693 if (read(fd, sb, 1024) != 1024)
694 return 0;
ca3b6696
N
695 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
696 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 697 return 0;
e7b84f9d 698 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 699 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 700 cont_err("size = %lluK\n", size*4);
aba69144 701
682c7051
NB
702 return 1;
703}
704
705int check_raid(int fd, char *name)
706{
4b1ac34b 707 struct mdinfo info;
682c7051 708 time_t crtime;
d078d77c 709 char *level;
82d9eba6 710 struct supertype *st = guess_super(fd);
f9ce90ba 711
cb8f6859
N
712 if (!st)
713 return 0;
fa219dd2
MD
714 if (st->ss->add_to_super != NULL) {
715 st->ss->load_super(st, fd, name);
716 /* Looks like a raid array .. */
717 pr_err("%s appears to be part of a raid array:\n", name);
718 st->ss->getinfo_super(st, &info, NULL);
719 st->ss->free_super(st);
720 crtime = info.array.ctime;
721 level = map_num(pers, info.array.level);
722 if (!level)
723 level = "-unknown-";
724 cont_err("level=%s devices=%d ctime=%s",
725 level, info.array.raid_disks, ctime(&crtime));
726 } else {
727 /* Looks like GPT or MBR */
728 pr_err("partition table exists on %s\n", name);
729 }
82d9eba6 730 return 1;
682c7051
NB
731}
732
0a6bff09
ZL
733int fstat_is_blkdev(int fd, char *devname, dev_t *rdev)
734{
735 struct stat stb;
736
737 if (fstat(fd, &stb) != 0) {
738 pr_err("fstat failed for %s: %s\n", devname, strerror(errno));
739 return 0;
740 }
741 if ((S_IFMT & stb.st_mode) != S_IFBLK) {
742 pr_err("%s is not a block device.\n", devname);
9e04ac1c
ZL
743 return 0;
744 }
745 if (rdev)
746 *rdev = stb.st_rdev;
747 return 1;
748}
749
750int stat_is_blkdev(char *devname, dev_t *rdev)
751{
752 struct stat stb;
753
754 if (stat(devname, &stb) != 0) {
755 pr_err("stat failed for %s: %s\n", devname, strerror(errno));
756 return 0;
757 }
758 if ((S_IFMT & stb.st_mode) != S_IFBLK) {
759 pr_err("%s is not a block device.\n", devname);
0a6bff09
ZL
760 return 0;
761 }
762 if (rdev)
763 *rdev = stb.st_rdev;
764 return 1;
765}
766
682c7051
NB
767int ask(char *mesg)
768{
769 char *add = "";
770 int i;
ca3b6696 771 for (i = 0; i < 5; i++) {
682c7051
NB
772 char buf[100];
773 fprintf(stderr, "%s%s", mesg, add);
774 fflush(stderr);
775 if (fgets(buf, 100, stdin)==NULL)
776 return 0;
777 if (buf[0]=='y' || buf[0]=='Y')
778 return 1;
779 if (buf[0]=='n' || buf[0]=='N')
780 return 0;
781 add = "(y/n) ";
782 }
e7b84f9d 783 pr_err("assuming 'no'\n");
682c7051
NB
784 return 0;
785}
786
8d80900b 787int is_standard(char *dev, int *nump)
e5329c37
NB
788{
789 /* tests if dev is a "standard" md dev name.
790 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 791 * where NN is a string of digits
598f0d58
NB
792 * Returns 1 if a partitionable standard,
793 * -1 if non-partitonable,
794 * 0 if not a standard name.
e5329c37 795 */
8d80900b 796 char *d = strrchr(dev, '/');
ca3b6696 797 int type = 0;
8d80900b
NB
798 int num;
799 if (!d)
e5329c37 800 return 0;
ca3b6696
N
801 if (strncmp(d, "/d",2) == 0)
802 d += 2, type = 1; /* /dev/md/dN{pM} */
803 else if (strncmp(d, "/md_d", 5) == 0)
804 d += 5, type = 1; /* /dev/md_dN{pM} */
805 else if (strncmp(d, "/md", 3) == 0)
806 d += 3, type = -1; /* /dev/mdN */
807 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
808 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
809 else
810 return 0;
8d80900b 811 if (!*d)
e5329c37 812 return 0;
8d80900b
NB
813 num = atoi(d);
814 while (isdigit(*d))
815 d++;
816 if (*d)
e5329c37 817 return 0;
8d80900b
NB
818 if (nump) *nump = num;
819
820 return type;
e5329c37
NB
821}
822
4b1ac34b 823unsigned long calc_csum(void *super, int bytes)
82b27616 824{
56eb10c0 825 unsigned long long newcsum = 0;
82b27616 826 int i;
4b1ac34b
NB
827 unsigned int csum;
828 unsigned int *superc = (unsigned int*) super;
82b27616 829
ca3b6696
N
830 for(i = 0; i < bytes/4; i++)
831 newcsum += superc[i];
82b27616 832 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 833#ifdef __alpha__
aba69144 834/* The in-kernel checksum calculation is always 16bit on
570c0542 835 * the alpha, though it is 32 bit on i386...
ca3b6696 836 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
837 * a way that it shouldn't).
838 */
839 csum = (csum & 0xffff) + (csum >> 16);
840 csum = (csum & 0xffff) + (csum >> 16);
841#endif
82b27616
NB
842 return csum;
843}
cd29a5c8 844
56eb10c0 845char *human_size(long long bytes)
cd29a5c8 846{
ff9239ee 847 static char buf[47];
d5d3721e
NB
848
849 /* We convert bytes to either centi-M{ega,ibi}bytes or
850 * centi-G{igi,ibi}bytes, with appropriate rounding,
851 * and then print 1/100th of those as a decimal.
852 * We allow upto 2048Megabytes before converting to
853 * gigabytes, as that shows more precision and isn't
854 * too large a number.
01619b48 855 * Terabytes are not yet handled.
d5d3721e 856 */
cd29a5c8 857
56eb10c0 858 if (bytes < 5000*1024)
ca3b6696 859 buf[0] = 0;
d5d3721e 860 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 861 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 862 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 863 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
efa29530 864 cMiB/100, cMiB % 100, cMB/100, cMB % 100);
d5d3721e 865 } else {
93d3bd3b 866 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 867 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 868 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
efa29530 869 cGiB/100, cGiB % 100, cGB/100, cGB % 100);
d5d3721e 870 }
cd29a5c8
NB
871 return buf;
872}
e0d19036 873
f0ec6710 874char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
875{
876 static char buf[30];
e0d19036 877
570abc6f
MN
878 /* We convert bytes to either centi-M{ega,ibi}bytes or
879 * centi-G{igi,ibi}bytes, with appropriate rounding,
880 * and then print 1/100th of those as a decimal.
881 * We allow upto 2048Megabytes before converting to
882 * gigabytes, as that shows more precision and isn't
883 * too large a number.
884 * Terabytes are not yet handled.
f0ec6710
MN
885 *
886 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
887 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
888 */
889
e0d19036 890 if (bytes < 5000*1024)
570abc6f 891 buf[0] = 0;
f0ec6710
MN
892 else if (prefix == IEC) {
893 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 894 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710 895 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
efa29530 896 cMiB/100, cMiB % 100);
f0ec6710 897 } else {
93d3bd3b 898 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710 899 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
efa29530 900 cGiB/100, cGiB % 100);
f0ec6710
MN
901 }
902 }
903 else if (prefix == JEDEC) {
904 if (bytes < 2*1024LL*1024LL*1024LL) {
905 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
906 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
efa29530 907 cMB/100, cMB % 100);
f0ec6710
MN
908 } else {
909 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
910 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
efa29530 911 cGB/100, cGB % 100);
f0ec6710 912 }
570abc6f 913 }
f0ec6710
MN
914 else
915 buf[0] = 0;
570abc6f 916
e0d19036
NB
917 return buf;
918}
e4965ef8
N
919
920void print_r10_layout(int layout)
921{
922 int near = layout & 255;
923 int far = (layout >> 8) & 255;
924 int offset = (layout&0x10000);
925 char *sep = "";
926
927 if (near != 1) {
928 printf("%s near=%d", sep, near);
929 sep = ",";
930 }
931 if (far != 1)
932 printf("%s %s=%d", sep, offset?"offset":"far", far);
933 if (near*far == 1)
934 printf("NO REDUNDANCY");
935}
e0d19036 936
5f8097be
NB
937unsigned long long calc_array_size(int level, int raid_disks, int layout,
938 int chunksize, unsigned long long devsize)
577e8448 939{
e03640bd
CA
940 if (level == 1)
941 return devsize;
577e8448
AK
942 devsize &= ~(unsigned long long)((chunksize>>9)-1);
943 return get_data_disks(level, layout, raid_disks) * devsize;
944}
945
946int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
947{
948 int data_disks = 0;
949 switch (level) {
ca3b6696
N
950 case 0: data_disks = raid_disks;
951 break;
952 case 1: data_disks = 1;
953 break;
5f8097be 954 case 4:
ca3b6696
N
955 case 5: data_disks = raid_disks - 1;
956 break;
957 case 6: data_disks = raid_disks - 2;
958 break;
5f8097be
NB
959 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
960 break;
961 }
577e8448
AK
962
963 return data_disks;
5f8097be
NB
964}
965
13db17bd 966dev_t devnm2devid(char *devnm)
4dd2df09
N
967{
968 /* First look in /sys/block/$DEVNM/dev for %d:%d
969 * If that fails, try parsing out a number
970 */
971 char path[100];
972 char *ep;
973 int fd;
974 int mjr,mnr;
975
976 sprintf(path, "/sys/block/%s/dev", devnm);
977 fd = open(path, O_RDONLY);
978 if (fd >= 0) {
979 char buf[20];
980 int n = read(fd, buf, sizeof(buf));
981 close(fd);
982 if (n > 0)
983 buf[n] = 0;
984 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
985 return makedev(mjr, mnr);
986 }
987 if (strncmp(devnm, "md_d", 4) == 0 &&
988 isdigit(devnm[4]) &&
989 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
990 ep > devnm && *ep == 0)
991 return makedev(get_mdp_major(), mnr << MdpMinorShift);
992
993 if (strncmp(devnm, "md", 2) == 0 &&
994 isdigit(devnm[2]) &&
995 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
996 ep > devnm && *ep == 0)
997 return makedev(MD_MAJOR, mnr);
998
999 return 0;
1000}
1001
1002char *get_md_name(char *devnm)
e0d19036
NB
1003{
1004 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 1005 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 1006
e0d19036
NB
1007 static char devname[50];
1008 struct stat stb;
4dd2df09 1009 dev_t rdev = devnm2devid(devnm);
dd0781e5 1010 char *dn;
98c6faba 1011
4dd2df09
N
1012 if (rdev == 0)
1013 return 0;
1014 if (strncmp(devnm, "md_", 3) == 0) {
1015 snprintf(devname, sizeof(devname), "/dev/md/%s",
1016 devnm + 3);
98c6faba
NB
1017 if (stat(devname, &stb) == 0
1018 && (S_IFMT&stb.st_mode) == S_IFBLK
1019 && (stb.st_rdev == rdev))
1020 return devname;
1021 }
4dd2df09
N
1022 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
1023 if (stat(devname, &stb) == 0
1024 && (S_IFMT&stb.st_mode) == S_IFBLK
1025 && (stb.st_rdev == rdev))
1026 return devname;
1027
1028 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
1029 if (stat(devname, &stb) == 0
1030 && (S_IFMT&stb.st_mode) == S_IFBLK
1031 && (stb.st_rdev == rdev))
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
NB
1041
1042 if (stat(devname, &stb) == 0
1043 && (S_IFMT&stb.st_mode) == S_IFBLK
1044 && (stb.st_rdev == rdev))
1045 return devname;
1046 unlink(devname);
1047 return NULL;
1048}
1049
1050void put_md_name(char *name)
1051{
ca3b6696 1052 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
1053 unlink(name);
1054}
f9ce90ba 1055
cc742d38
N
1056int get_maj_min(char *dev, int *major, int *minor)
1057{
1058 char *e;
1059 *major = strtoul(dev, &e, 0);
1060 return (e > dev && *e == ':' && e[1] &&
1061 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1062 *e == 0);
1063}
1064
8b0dabea
NB
1065int dev_open(char *dev, int flags)
1066{
1067 /* like 'open', but if 'dev' matches %d:%d, create a temp
1068 * block device and open that
1069 */
8b0dabea
NB
1070 int fd = -1;
1071 char devname[32];
e81cdd9f 1072 int major;
8b0dabea 1073 int minor;
e81cdd9f 1074
9f0ad56b
JS
1075 if (!dev)
1076 return -1;
6df6a774 1077 flags |= O_DIRECT;
e81cdd9f 1078
cc742d38 1079 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1080 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1081 (int)getpid(), major, minor);
1082 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1083 fd = open(devname, flags);
1084 unlink(devname);
6df6a774
N
1085 }
1086 if (fd < 0) {
9dc7d357 1087 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
1088 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
1089 (int)getpid(), major, minor);
ca3b6696 1090 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
1091 fd = open(devname, flags);
1092 unlink(devname);
1093 }
8b0dabea
NB
1094 }
1095 } else
6df6a774 1096 fd = open(dev, flags);
8b0dabea
NB
1097 return fd;
1098}
f9ce90ba 1099
4dd2df09 1100int open_dev_flags(char *devnm, int flags)
e8a70c89 1101{
13db17bd 1102 dev_t devid;
e8a70c89
N
1103 char buf[20];
1104
4dd2df09
N
1105 devid = devnm2devid(devnm);
1106 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1107 return dev_open(buf, flags);
1108}
1109
4dd2df09 1110int open_dev(char *devnm)
d998b738 1111{
4dd2df09 1112 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1113}
1114
4dd2df09 1115int open_dev_excl(char *devnm)
a931db9e
NB
1116{
1117 char buf[20];
1118 int i;
7187750e 1119 int flags = O_RDWR;
13db17bd 1120 dev_t devid = devnm2devid(devnm);
399e0b97 1121 long delay = 1000;
a931db9e 1122
4dd2df09 1123 sprintf(buf, "%d:%d", major(devid), minor(devid));
efa29530 1124 for (i = 0; i < 25; i++) {
7187750e 1125 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1126 if (fd >= 0)
1127 return fd;
7187750e
N
1128 if (errno == EACCES && flags == O_RDWR) {
1129 flags = O_RDONLY;
1130 continue;
1131 }
a931db9e
NB
1132 if (errno != EBUSY)
1133 return fd;
399e0b97
N
1134 usleep(delay);
1135 if (delay < 200000)
1136 delay *= 2;
a931db9e
NB
1137 }
1138 return -1;
1139}
1140
9008ed1c
N
1141int same_dev(char *one, char *two)
1142{
1143 struct stat st1, st2;
1144 if (stat(one, &st1) != 0)
1145 return 0;
1146 if (stat(two, &st2) != 0)
1147 return 0;
1148 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1149 return 0;
1150 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1151 return 0;
1152 return st1.st_rdev == st2.st_rdev;
1153}
1154
a7c6e3fb 1155void wait_for(char *dev, int fd)
a714580e
N
1156{
1157 int i;
a7c6e3fb 1158 struct stat stb_want;
399e0b97 1159 long delay = 1000;
a7c6e3fb
N
1160
1161 if (fstat(fd, &stb_want) != 0 ||
1162 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1163 return;
a714580e 1164
efa29530 1165 for (i = 0; i < 25; i++) {
a714580e 1166 struct stat stb;
a7c6e3fb
N
1167 if (stat(dev, &stb) == 0 &&
1168 (stb.st_mode & S_IFMT) == S_IFBLK &&
1169 (stb.st_rdev == stb_want.st_rdev))
a714580e 1170 return;
399e0b97
N
1171 usleep(delay);
1172 if (delay < 200000)
1173 delay *= 2;
a714580e 1174 }
436305c6 1175 if (i == 25)
1ade5cc1 1176 dprintf("timeout waiting for %s\n", dev);
a714580e
N
1177}
1178
0f22b998
N
1179struct superswitch *superlist[] =
1180{
1181 &super0, &super1,
1182 &super_ddf, &super_imsm,
0592faeb 1183 &mbr, &gpt,
32141c17
JS
1184 NULL
1185};
f7dd881f 1186
4725bc31 1187struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1188{
1686dc25
NB
1189 mdu_array_info_t array;
1190 int vers;
1191 int minor;
1192 struct supertype *st = NULL;
7e0f6979 1193 struct mdinfo *sra;
142cb9e1 1194 char *verstr;
1686dc25
NB
1195 char version[20];
1196 int i;
f7e7067b 1197 char *subarray = NULL;
4dd2df09 1198 char container[32] = "";
1686dc25 1199
4dd2df09 1200 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1201
1202 if (sra) {
7e0f6979
NB
1203 vers = sra->array.major_version;
1204 minor = sra->array.minor_version;
142cb9e1 1205 verstr = sra->text_version;
1686dc25 1206 } else {
9cd39f01 1207 if (md_get_array_info(fd, &array))
1686dc25
NB
1208 array.major_version = array.minor_version = 0;
1209 vers = array.major_version;
1210 minor = array.minor_version;
142cb9e1 1211 verstr = "";
6fbba4c9 1212 }
82d9eba6 1213
1686dc25
NB
1214 if (vers != -1) {
1215 sprintf(version, "%d.%d", vers, minor);
1216 verstr = version;
6fbba4c9 1217 }
3c558363 1218 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1219 char *dev = verstr+1;
5f7e44b2 1220
f7e7067b 1221 subarray = strchr(dev, '/');
92d49ecf 1222 if (subarray) {
f7e7067b 1223 *subarray++ = '\0';
503975b9 1224 subarray = xstrdup(subarray);
92d49ecf 1225 }
4dd2df09 1226 strcpy(container, dev);
15d230f7 1227 sysfs_free(sra);
5f7e44b2 1228 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1229 if (sra && sra->text_version[0])
1230 verstr = sra->text_version;
1231 else
1232 verstr = "-no-metadata-";
f7e7067b
NB
1233 }
1234
efa29530 1235 for (i = 0; st == NULL && superlist[i]; i++)
f7e7067b 1236 st = superlist[i]->match_metadata_desc(verstr);
1686dc25 1237
15d230f7 1238 sysfs_free(sra);
f7e7067b 1239 if (st) {
3b0896f8 1240 st->sb = NULL;
1f49fb3a
N
1241 if (subarrayp)
1242 *subarrayp = subarray;
4dd2df09
N
1243 strcpy(st->container_devnm, container);
1244 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1245 } else
1246 free(subarray);
5f7e44b2 1247
82d9eba6 1248 return st;
f9ce90ba 1249}
ea24acd0 1250
0f0749ad 1251int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1252{
1253 char buf[20];
1254 int fd;
1255
1256 sprintf(buf, "%d:%d", major(id), minor(id));
1257 fd = dev_open(buf, O_RDONLY);
1258 if (fd < 0)
1259 return 0;
1260 if (get_dev_size(fd, NULL, size)) {
1261 close(fd);
1262 return 1;
1263 }
1264 close(fd);
1265 return 0;
1266}
f9ce90ba 1267
4b57ecf6
AO
1268int dev_sector_size_from_id(dev_t id, unsigned int *size)
1269{
1270 char buf[20];
1271 int fd;
1272
1273 sprintf(buf, "%d:%d", major(id), minor(id));
1274 fd = dev_open(buf, O_RDONLY);
1275 if (fd < 0)
1276 return 0;
1277 if (get_dev_sector_size(fd, NULL, size)) {
1278 close(fd);
1279 return 1;
1280 }
1281 close(fd);
1282 return 0;
1283}
1284
159c3a1a 1285struct supertype *dup_super(struct supertype *orig)
3da92f27 1286{
159c3a1a 1287 struct supertype *st;
1686dc25 1288
d2ca6449
NB
1289 if (!orig)
1290 return orig;
503975b9 1291 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1292 st->ss = orig->ss;
1293 st->max_devs = orig->max_devs;
1294 st->minor_version = orig->minor_version;
cb8f6859 1295 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1296 st->data_offset = orig->data_offset;
159c3a1a
NB
1297 st->sb = NULL;
1298 st->info = NULL;
1299 return st;
3da92f27
NB
1300}
1301
54887ad8 1302struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1303{
1304 /* try each load_super to find the best match,
1305 * and return the best superswitch
1306 */
82d9eba6
NB
1307 struct superswitch *ss;
1308 struct supertype *st;
1d13b599 1309 unsigned int besttime = 0;
570c0542 1310 int bestsuper = -1;
f9ce90ba
NB
1311 int i;
1312
503975b9 1313 st = xcalloc(1, sizeof(*st));
4dd2df09 1314 st->container_devnm[0] = 0;
d1d599ea 1315
efa29530 1316 for (i = 0; superlist[i]; i++) {
f9ce90ba
NB
1317 int rv;
1318 ss = superlist[i];
54887ad8
N
1319 if (guess_type == guess_array && ss->add_to_super == NULL)
1320 continue;
1321 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1322 continue;
ef609477 1323 memset(st, 0, sizeof(*st));
df3346e6 1324 st->ignore_hw_compat = 1;
3da92f27 1325 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1326 if (rv == 0) {
1327 struct mdinfo info;
a5d85af7 1328 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1329 if (bestsuper == -1 ||
1330 besttime < info.array.ctime) {
1331 bestsuper = i;
1332 besttime = info.array.ctime;
570c0542 1333 }
3da92f27 1334 ss->free_super(st);
570c0542
NB
1335 }
1336 }
1337 if (bestsuper != -1) {
1338 int rv;
ef609477 1339 memset(st, 0, sizeof(*st));
df3346e6 1340 st->ignore_hw_compat = 1;
3da92f27 1341 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1342 if (rv == 0) {
5e747af2 1343 superlist[bestsuper]->free_super(st);
82d9eba6 1344 return st;
f9ce90ba
NB
1345 }
1346 }
570c0542 1347 free(st);
f9ce90ba
NB
1348 return NULL;
1349}
fe6729fa 1350
beae1dfe
NB
1351/* Return size of device in bytes */
1352int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1353{
1354 unsigned long long ldsize;
c2c9bb6f
NB
1355 struct stat st;
1356
1357 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1358 ldsize = (unsigned long long)st.st_size;
1359 else
beae1dfe
NB
1360#ifdef BLKGETSIZE64
1361 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1362#endif
1363 {
1364 unsigned long dsize;
1365 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1366 ldsize = dsize;
1367 ldsize <<= 9;
1368 } else {
1369 if (dname)
32971509 1370 pr_err("Cannot get size of %s: %s\n",
beae1dfe
NB
1371 dname, strerror(errno));
1372 return 0;
1373 }
1374 }
1375 *sizep = ldsize;
1376 return 1;
1377}
8fac0577 1378
32971509
PB
1379/* Return sector size of device in bytes */
1380int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep)
1381{
1382 unsigned int sectsize;
1383
1384 if (ioctl(fd, BLKSSZGET, &sectsize) != 0) {
1385 if (dname)
1386 pr_err("Cannot get sector size of %s: %s\n",
1387 dname, strerror(errno));
1388 return 0;
1389 }
1390
1391 *sectsizep = sectsize;
1392 return 1;
1393}
1394
3a371610
N
1395/* Return true if this can only be a container, not a member device.
1396 * i.e. is and md device and size is zero
1397 */
1398int must_be_container(int fd)
1399{
f5c924f4 1400 struct mdinfo *mdi;
3a371610 1401 unsigned long long size;
f5c924f4
JS
1402
1403 mdi = sysfs_read(fd, NULL, GET_VERSION);
1404 if (!mdi)
3a371610 1405 return 0;
f5c924f4
JS
1406 sysfs_free(mdi);
1407
3a371610
N
1408 if (get_dev_size(fd, NULL, &size) == 0)
1409 return 1;
1410 if (size == 0)
1411 return 1;
1412 return 0;
1413}
034b203a
TM
1414
1415/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1416 * Returns: 1 if successful
1417 * -1 for unknown partition type
1418 * 0 for other errors
1419 */
1420static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1421{
056b331e 1422 struct GPT gpt;
034b203a 1423 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1424 struct GPT_part_entry *part;
1425 char buf[512];
034b203a
TM
1426 unsigned long long curr_part_end;
1427 unsigned all_partitions, entry_size;
f21e18ca 1428 unsigned part_nr;
41b06495 1429 unsigned int sector_size = 0;
034b203a
TM
1430
1431 *endofpart = 0;
1432
056b331e 1433 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1434 /* skip protective MBR */
41b06495
MD
1435 if (!get_dev_sector_size(fd, NULL, &sector_size))
1436 return 0;
1437 lseek(fd, sector_size, SEEK_SET);
73e658d8 1438 /* read GPT header */
056b331e 1439 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1440 return 0;
1441
1442 /* get the number of partition entries and the entry size */
056b331e
N
1443 all_partitions = __le32_to_cpu(gpt.part_cnt);
1444 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1445
1446 /* Check GPT signature*/
056b331e 1447 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1448 return -1;
1449
1450 /* sanity checks */
1451 if (all_partitions > 1024 ||
db7fdfe4 1452 entry_size > sizeof(buf))
034b203a
TM
1453 return -1;
1454
db7fdfe4
JS
1455 part = (struct GPT_part_entry *)buf;
1456
41b06495
MD
1457 /* set offset to third block (GPT entries) */
1458 lseek(fd, sector_size*2, SEEK_SET);
ca3b6696 1459 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1460 /* read partition entry */
db7fdfe4 1461 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1462 return 0;
1463
034b203a 1464 /* is this valid partition? */
db7fdfe4 1465 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1466 /* check the last lba for the current partition */
db7fdfe4 1467 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1468 if (curr_part_end > *endofpart)
1469 *endofpart = curr_part_end;
1470 }
1471
034b203a
TM
1472 }
1473 return 1;
1474}
1475
1476/* Sets endofpart parameter to the last block used by the last partition on the device.
1477 * Returns: 1 if successful
1478 * -1 for unknown partition type
1479 * 0 for other errors
1480 */
1481static int get_last_partition_end(int fd, unsigned long long *endofpart)
1482{
056b331e 1483 struct MBR boot_sect;
034b203a 1484 unsigned long long curr_part_end;
f21e18ca 1485 unsigned part_nr;
31208db9 1486 unsigned int sector_size;
034b203a
TM
1487 int retval = 0;
1488
1489 *endofpart = 0;
1490
056b331e 1491 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1492 /* read MBR */
1493 lseek(fd, 0, 0);
056b331e 1494 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1495 goto abort;
1496
1497 /* check MBP signature */
056b331e 1498 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1499 retval = 1;
1500 /* found the correct signature */
034b203a 1501
ca3b6696 1502 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
8e2bca51
JC
1503 /*
1504 * Have to make every access through boot_sect rather
1505 * than using a pointer to the partition table (or an
1506 * entry), since the entries are not properly aligned.
1507 */
1508
034b203a 1509 /* check for GPT type */
8e2bca51
JC
1510 if (boot_sect.parts[part_nr].part_type ==
1511 MBR_GPT_PARTITION_TYPE) {
034b203a
TM
1512 retval = get_gpt_last_partition_end(fd, endofpart);
1513 break;
1514 }
1515 /* check the last used lba for the current partition */
8e2bca51
JC
1516 curr_part_end =
1517 __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) +
1518 __le32_to_cpu(boot_sect.parts[part_nr].blocks_num);
034b203a
TM
1519 if (curr_part_end > *endofpart)
1520 *endofpart = curr_part_end;
034b203a
TM
1521 }
1522 } else {
1523 /* Unknown partition table */
1524 retval = -1;
1525 }
31208db9
MD
1526 /* calculate number of 512-byte blocks */
1527 if (get_dev_sector_size(fd, NULL, &sector_size))
1528 *endofpart *= (sector_size / 512);
034b203a
TM
1529 abort:
1530 return retval;
1531}
1532
53ed6ac3
KW
1533int check_partitions(int fd, char *dname, unsigned long long freesize,
1534 unsigned long long size)
034b203a
TM
1535{
1536 /*
1537 * Check where the last partition ends
1538 */
1539 unsigned long long endofpart;
034b203a 1540
31208db9 1541 if (get_last_partition_end(fd, &endofpart) > 0) {
034b203a
TM
1542 /* There appears to be a partition table here */
1543 if (freesize == 0) {
1544 /* partitions will not be visible in new device */
e7b84f9d
N
1545 pr_err("partition table exists on %s but will be lost or\n"
1546 " meaningless after creating array\n",
1547 dname);
034b203a
TM
1548 return 1;
1549 } else if (endofpart > freesize) {
1550 /* last partition overlaps metadata */
e7b84f9d
N
1551 pr_err("metadata will over-write last partition on %s.\n",
1552 dname);
034b203a 1553 return 1;
53ed6ac3
KW
1554 } else if (size && endofpart > size) {
1555 /* partitions will be truncated in new device */
e7b84f9d
N
1556 pr_err("array size is too small to cover all partitions on %s.\n",
1557 dname);
53ed6ac3 1558 return 1;
034b203a
TM
1559 }
1560 }
1561 return 0;
1562}
1563
a322f70c
DW
1564int open_container(int fd)
1565{
1566 /* 'fd' is a block device. Find out if it is in use
1567 * by a container, and return an open fd on that container.
1568 */
1569 char path[256];
1570 char *e;
1571 DIR *dir;
1572 struct dirent *de;
1573 int dfd, n;
1574 char buf[200];
1575 int major, minor;
1576 struct stat st;
1577
1578 if (fstat(fd, &st) != 0)
1579 return -1;
1580 sprintf(path, "/sys/dev/block/%d:%d/holders",
1581 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1582 e = path + strlen(path);
1583
1584 dir = opendir(path);
1585 if (!dir)
1586 return -1;
1587 while ((de = readdir(dir))) {
1588 if (de->d_ino == 0)
1589 continue;
1590 if (de->d_name[0] == '.')
1591 continue;
a994592d
N
1592 /* Need to make sure it is a container and not a volume */
1593 sprintf(e, "/%s/md/metadata_version", de->d_name);
1594 dfd = open(path, O_RDONLY);
1595 if (dfd < 0)
1596 continue;
1597 n = read(dfd, buf, sizeof(buf));
1598 close(dfd);
1599 if (n <= 0 || (unsigned)n >= sizeof(buf))
1600 continue;
1601 buf[n] = 0;
1602 if (strncmp(buf, "external", 8) != 0 ||
1603 n < 10 ||
1604 buf[9] == '/')
1605 continue;
a322f70c
DW
1606 sprintf(e, "/%s/dev", de->d_name);
1607 dfd = open(path, O_RDONLY);
1608 if (dfd < 0)
1609 continue;
1610 n = read(dfd, buf, sizeof(buf));
1611 close(dfd);
f21e18ca 1612 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1613 continue;
1614 buf[n] = 0;
1615 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1616 continue;
1617 sprintf(buf, "%d:%d", major, minor);
1618 dfd = dev_open(buf, O_RDONLY);
1619 if (dfd >= 0) {
1620 closedir(dir);
1621 return dfd;
1622 }
1623 }
355726fa 1624 closedir(dir);
a322f70c
DW
1625 return -1;
1626}
1627
33414a01
DW
1628struct superswitch *version_to_superswitch(char *vers)
1629{
1630 int i;
1631
1632 for (i = 0; superlist[i]; i++) {
1633 struct superswitch *ss = superlist[i];
1634
1635 if (strcmp(vers, ss->name) == 0)
1636 return ss;
1637 }
1638
1639 return NULL;
1640}
1641
4dd2df09
N
1642int metadata_container_matches(char *metadata, char *devnm)
1643{
1644 /* Check if 'devnm' is the container named in 'metadata'
1645 * which is
1646 * /containername/componentname or
1647 * -containername/componentname
1648 */
1649 int l;
1650 if (*metadata != '/' && *metadata != '-')
1651 return 0;
1652 l = strlen(devnm);
1653 if (strncmp(metadata+1, devnm, l) != 0)
1654 return 0;
1655 if (metadata[l+1] != '/')
1656 return 0;
1657 return 1;
1658}
1659
1660int metadata_subdev_matches(char *metadata, char *devnm)
1661{
1662 /* Check if 'devnm' is the subdev named in 'metadata'
1663 * which is
1664 * /containername/subdev or
1665 * -containername/subdev
1666 */
1667 char *sl;
1668 if (*metadata != '/' && *metadata != '-')
1669 return 0;
1670 sl = strchr(metadata+1, '/');
1671 if (!sl)
1672 return 0;
1673 if (strcmp(sl+1, devnm) == 0)
1674 return 1;
1675 return 0;
1676}
1677
33414a01
DW
1678int is_container_member(struct mdstat_ent *mdstat, char *container)
1679{
1680 if (mdstat->metadata_version == NULL ||
1681 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1682 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1683 return 0;
1684
1685 return 1;
1686}
1687
1688int is_subarray_active(char *subarray, char *container)
1689{
1690 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1691 struct mdstat_ent *ent;
1692
e5408a32
DW
1693 for (ent = mdstat; ent; ent = ent->next)
1694 if (is_container_member(ent, container))
e5e5d7ce 1695 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1696 break;
33414a01
DW
1697
1698 free_mdstat(mdstat);
1699
1700 return ent != NULL;
1701}
1702
1703/* open_subarray - opens a subarray in a container
1704 * @dev: container device name
feab51f8 1705 * @st: empty supertype
33414a01
DW
1706 * @quiet: block reporting errors flag
1707 *
1708 * On success returns an fd to a container and fills in *st
1709 */
feab51f8 1710int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1711{
1712 struct mdinfo *mdi;
a951a4f7 1713 struct mdinfo *info;
33414a01 1714 int fd, err = 1;
4dd2df09 1715 char *_devnm;
33414a01
DW
1716
1717 fd = open(dev, O_RDWR|O_EXCL);
1718 if (fd < 0) {
1719 if (!quiet)
e7b84f9d 1720 pr_err("Couldn't open %s, aborting\n",
33414a01 1721 dev);
b990032d 1722 return -1;
33414a01
DW
1723 }
1724
4dd2df09
N
1725 _devnm = fd2devnm(fd);
1726 if (_devnm == NULL) {
33414a01 1727 if (!quiet)
e7b84f9d
N
1728 pr_err("Failed to determine device number for %s\n",
1729 dev);
33414a01
DW
1730 goto close_fd;
1731 }
4dd2df09 1732 strcpy(st->devnm, _devnm);
33414a01 1733
4dd2df09 1734 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1735 if (!mdi) {
1736 if (!quiet)
e7b84f9d 1737 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1738 dev);
1739 goto close_fd;
1740 }
1741
1742 if (mdi->array.level != UnSet) {
1743 if (!quiet)
e7b84f9d 1744 pr_err("%s is not a container\n", dev);
33414a01
DW
1745 goto free_sysfs;
1746 }
1747
1748 st->ss = version_to_superswitch(mdi->text_version);
1749 if (!st->ss) {
1750 if (!quiet)
e7b84f9d
N
1751 pr_err("Operation not supported for %s metadata\n",
1752 mdi->text_version);
33414a01
DW
1753 goto free_sysfs;
1754 }
1755
4dd2df09 1756 if (st->devnm[0] == 0) {
33414a01 1757 if (!quiet)
e7b84f9d 1758 pr_err("Failed to allocate device name\n");
33414a01
DW
1759 goto free_sysfs;
1760 }
1761
db20d413 1762 if (!st->ss->load_container) {
33414a01 1763 if (!quiet)
e7b84f9d 1764 pr_err("%s is not a container\n", dev);
4dd2df09 1765 goto free_sysfs;
33414a01
DW
1766 }
1767
db20d413 1768 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1769 if (!quiet)
e7b84f9d 1770 pr_err("Failed to load metadata for %s\n",
db20d413 1771 dev);
4dd2df09 1772 goto free_sysfs;
33414a01
DW
1773 }
1774
a951a4f7
N
1775 info = st->ss->container_content(st, subarray);
1776 if (!info) {
1777 if (!quiet)
e7b84f9d 1778 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1779 subarray, dev);
1780 goto free_super;
1781 }
1782 free(info);
1783
33414a01
DW
1784 err = 0;
1785
1786 free_super:
1787 if (err)
1788 st->ss->free_super(st);
33414a01
DW
1789 free_sysfs:
1790 sysfs_free(mdi);
1791 close_fd:
1792 if (err)
1793 close(fd);
1794
1795 if (err)
1796 return -1;
1797 else
1798 return fd;
1799}
1800
7801ac20
N
1801int add_disk(int mdfd, struct supertype *st,
1802 struct mdinfo *sra, struct mdinfo *info)
1803{
1804 /* Add a device to an array, in one of 2 ways. */
1805 int rv;
32141c17 1806
7801ac20 1807 if (st->ss->external) {
d23534e4
DW
1808 if (info->disk.state & (1<<MD_DISK_SYNC))
1809 info->recovery_start = MaxSector;
1810 else
1811 info->recovery_start = 0;
2904b26f 1812 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1813 if (! rv) {
1814 struct mdinfo *sd2;
f35f2525
N
1815 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1816 if (sd2 == info)
1817 break;
1818 if (sd2 == NULL) {
503975b9 1819 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1820 *sd2 = *info;
1821 sd2->next = sra->devs;
1822 sra->devs = sd2;
1823 }
7801ac20
N
1824 }
1825 } else
7801ac20
N
1826 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1827 return rv;
1828}
1829
de6ae750
N
1830int remove_disk(int mdfd, struct supertype *st,
1831 struct mdinfo *sra, struct mdinfo *info)
1832{
1833 int rv;
32141c17 1834
de6ae750 1835 /* Remove the disk given by 'info' from the array */
de6ae750
N
1836 if (st->ss->external)
1837 rv = sysfs_set_str(sra, info, "slot", "none");
1838 else
de6ae750
N
1839 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1840 info->disk.minor));
1841 return rv;
1842}
1843
1ab9ed2a 1844int hot_remove_disk(int mdfd, unsigned long dev, int force)
2dd271fe 1845{
1ab9ed2a 1846 int cnt = force ? 500 : 5;
2dd271fe
N
1847 int ret;
1848
1849 /* HOT_REMOVE_DISK can fail with EBUSY if there are
1850 * outstanding IO requests to the device.
1851 * In this case, it can be helpful to wait a little while,
1ab9ed2a 1852 * up to 5 seconds if 'force' is set, or 50 msec if not.
2dd271fe
N
1853 */
1854 while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 &&
1855 errno == EBUSY &&
1856 cnt-- > 0)
1857 usleep(10000);
1858
1859 return ret;
1860}
1861
1ab9ed2a 1862int sys_hot_remove_disk(int statefd, int force)
fdd01569 1863{
1ab9ed2a 1864 int cnt = force ? 500 : 5;
fdd01569
N
1865 int ret;
1866
1867 while ((ret = write(statefd, "remove", 6)) == -1 &&
1868 errno == EBUSY &&
1869 cnt-- > 0)
1870 usleep(10000);
1871 return ret == 6 ? 0 : -1;
1872}
1873
f35f2525
N
1874int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1875{
1876 /* Initialise kernel's knowledge of array.
1877 * This varies between externally managed arrays
1878 * and older kernels
1879 */
700483a2 1880 mdu_array_info_t inf;
f35f2525
N
1881 int rv;
1882
f35f2525 1883 if (st->ss->external)
700483a2 1884 return sysfs_set_array(info, 9003);
700483a2
JS
1885
1886 memset(&inf, 0, sizeof(inf));
1887 inf.major_version = info->array.major_version;
1888 inf.minor_version = info->array.minor_version;
1889 rv = md_set_array_info(mdfd, &inf);
1890
f35f2525
N
1891 return rv;
1892}
1893
1e5c6983
DW
1894unsigned long long min_recovery_start(struct mdinfo *array)
1895{
1896 /* find the minimum recovery_start in an array for metadata
1897 * formats that only record per-array recovery progress instead
1898 * of per-device
1899 */
1900 unsigned long long recovery_start = MaxSector;
1901 struct mdinfo *d;
1902
1903 for (d = array->devs; d; d = d->next)
1904 recovery_start = min(recovery_start, d->recovery_start);
1905
1906 return recovery_start;
1907}
1908
4dd2df09 1909int mdmon_pid(char *devnm)
a931db9e
NB
1910{
1911 char path[100];
1912 char pid[10];
1913 int fd;
1914 int n;
10013317 1915
4dd2df09 1916 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1917
24f6f99b 1918 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1919
1920 if (fd < 0)
cf556303 1921 return -1;
a931db9e
NB
1922 n = read(fd, pid, 9);
1923 close(fd);
1924 if (n <= 0)
cf556303 1925 return -1;
24f6f99b 1926 return atoi(pid);
a931db9e
NB
1927}
1928
4dd2df09 1929int mdmon_running(char *devnm)
a931db9e 1930{
4dd2df09 1931 int pid = mdmon_pid(devnm);
24f6f99b 1932 if (pid <= 0)
a931db9e 1933 return 0;
24f6f99b 1934 if (kill(pid, 0) == 0)
a931db9e
NB
1935 return 1;
1936 return 0;
1937}
1938
4dd2df09 1939int start_mdmon(char *devnm)
8850ee3e 1940{
1913c325 1941 int i, skipped;
44d2e365 1942 int len;
ca3b6696 1943 pid_t pid;
9fe32043 1944 int status;
44d2e365
N
1945 char pathbuf[1024];
1946 char *paths[4] = {
1947 pathbuf,
85945e19 1948 BINDIR "/mdmon",
2f1bcf43 1949 "./mdmon",
44d2e365
N
1950 NULL
1951 };
8850ee3e 1952
40ebbb9c 1953 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1954 return 0;
1955
9cf014ec 1956 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1957 if (len > 0) {
1958 char *sl;
1959 pathbuf[len] = 0;
1960 sl = strrchr(pathbuf, '/');
1961 if (sl)
1962 sl++;
1963 else
1964 sl = pathbuf;
1965 strcpy(sl, "mdmon");
1966 } else
1967 pathbuf[0] = '\0';
1968
0f7bdf89 1969 /* First try to run systemctl */
9540cc24
N
1970 if (!check_env("MDADM_NO_SYSTEMCTL"))
1971 switch(fork()) {
1972 case 0:
1973 /* FIXME yuk. CLOSE_EXEC?? */
1974 skipped = 0;
1975 for (i = 3; skipped < 20; i++)
1976 if (close(i) < 0)
1977 skipped++;
1978 else
1979 skipped = 0;
1980
1981 /* Don't want to see error messages from
1982 * systemctl. If the service doesn't exist,
1983 * we start mdmon ourselves.
1984 */
1985 close(2);
1986 open("/dev/null", O_WRONLY);
1987 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1988 devnm);
1989 status = execl("/usr/bin/systemctl", "systemctl",
1990 "start",
1991 pathbuf, NULL);
1992 status = execl("/bin/systemctl", "systemctl", "start",
1993 pathbuf, NULL);
1994 exit(1);
7a862a02 1995 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1996 return -1;
1997 default: /* parent - good */
1998 pid = wait(&status);
1999 if (pid >= 0 && status == 0)
2000 return 0;
2001 }
0f7bdf89
JS
2002
2003 /* That failed, try running mdmon directly */
8850ee3e
N
2004 switch(fork()) {
2005 case 0:
2006 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 2007 skipped = 0;
ca3b6696 2008 for (i = 3; skipped < 20; i++)
1913c325
N
2009 if (close(i) < 0)
2010 skipped++;
2011 else
2012 skipped = 0;
2013
ca3b6696 2014 for (i = 0; paths[i]; i++)
a0963a86 2015 if (paths[i][0]) {
2f1bcf43 2016 execl(paths[i], paths[i],
4dd2df09 2017 devnm, NULL);
a0963a86 2018 }
8850ee3e 2019 exit(1);
7a862a02 2020 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 2021 return -1;
9fe32043
N
2022 default: /* parent - good */
2023 pid = wait(&status);
15c10423 2024 if (pid < 0 || status != 0) {
7a862a02 2025 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 2026 return -1;
15c10423 2027 }
8850ee3e
N
2028 }
2029 return 0;
2030}
2031
148acb7b
DW
2032__u32 random32(void)
2033{
2034 __u32 rv;
2035 int rfd = open("/dev/urandom", O_RDONLY);
2036 if (rfd < 0 || read(rfd, &rv, 4) != 4)
2037 rv = random();
2038 if (rfd >= 0)
2039 close(rfd);
2040 return rv;
2041}
2042
c5f71c24
JS
2043void random_uuid(__u8 *buf)
2044{
2045 int fd, i, len;
2046 __u32 r[4];
2047
2048 fd = open("/dev/urandom", O_RDONLY);
2049 if (fd < 0)
2050 goto use_random;
2051 len = read(fd, buf, 16);
2052 close(fd);
2053 if (len != 16)
2054 goto use_random;
2055
2056 return;
2057
2058use_random:
2059 for (i = 0; i < 4; i++)
2060 r[i] = random();
2061 memcpy(buf, r, 16);
2062}
2063
edd8d13c
NB
2064int flush_metadata_updates(struct supertype *st)
2065{
2066 int sfd;
2067 if (!st->updates) {
2068 st->update_tail = NULL;
2069 return -1;
2070 }
2071
4dd2df09 2072 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
2073 if (sfd < 0)
2074 return -1;
2075
2076 while (st->updates) {
2077 struct metadata_update *mu = st->updates;
2078 st->updates = mu->next;
2079
2080 send_message(sfd, mu, 0);
2081 wait_reply(sfd, 0);
2082 free(mu->buf);
2083 free(mu);
2084 }
2085 ack(sfd, 0);
2086 wait_reply(sfd, 0);
2087 close(sfd);
2088 st->update_tail = NULL;
2089 return 0;
2090}
2091
2092void append_metadata_update(struct supertype *st, void *buf, int len)
2093{
2094
503975b9 2095 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
2096
2097 mu->buf = buf;
2098 mu->len = len;
2099 mu->space = NULL;
cb23f1f4 2100 mu->space_list = NULL;
edd8d13c
NB
2101 mu->next = NULL;
2102 *st->update_tail = mu;
2103 st->update_tail = &mu->next;
2104}
a931db9e 2105
fe6729fa
NB
2106#ifdef __TINYC__
2107/* tinyc doesn't optimize this check in ioctl.h out ... */
2108unsigned int __invalid_size_argument_for_IOC = 0;
2109#endif
2110
6d11ec6f
AK
2111int experimental(void)
2112{
2113 if (check_env("MDADM_EXPERIMENTAL"))
2114 return 1;
2115 else {
7a862a02 2116 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
2117 return 0;
2118 }
2119}
2120
326727d9
AC
2121/* Pick all spares matching given criteria from a container
2122 * if min_size == 0 do not check size
2123 * if domlist == NULL do not check domains
2124 * if spare_group given add it to domains of each spare
2125 * metadata allows to test domains using metadata of destination array */
2126struct mdinfo *container_choose_spares(struct supertype *st,
fbfdcb06 2127 struct spare_criteria *criteria,
326727d9
AC
2128 struct domainlist *domlist,
2129 char *spare_group,
2130 const char *metadata, int get_one)
2131{
2132 struct mdinfo *d, **dp, *disks = NULL;
2133
2134 /* get list of all disks in container */
2135 if (st->ss->getinfo_super_disks)
2136 disks = st->ss->getinfo_super_disks(st);
2137
2138 if (!disks)
2139 return disks;
2140 /* find spare devices on the list */
2141 dp = &disks->devs;
2142 disks->array.spare_disks = 0;
2143 while (*dp) {
2144 int found = 0;
2145 d = *dp;
2146 if (d->disk.state == 0) {
2147 /* check if size is acceptable */
2148 unsigned long long dev_size;
4b57ecf6
AO
2149 unsigned int dev_sector_size;
2150 int size_valid = 0;
2151 int sector_size_valid = 0;
2152
326727d9
AC
2153 dev_t dev = makedev(d->disk.major,d->disk.minor);
2154
fbfdcb06 2155 if (!criteria->min_size ||
326727d9 2156 (dev_size_from_id(dev, &dev_size) &&
fbfdcb06 2157 dev_size >= criteria->min_size))
4b57ecf6
AO
2158 size_valid = 1;
2159
2160 if (!criteria->sector_size ||
2161 (dev_sector_size_from_id(dev, &dev_sector_size) &&
2162 criteria->sector_size == dev_sector_size))
2163 sector_size_valid = 1;
2164
2165 found = size_valid && sector_size_valid;
2166
326727d9
AC
2167 /* check if domain matches */
2168 if (found && domlist) {
4dd2df09 2169 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2170 if (spare_group)
2171 pol_add(&pol, pol_domain,
2172 spare_group, NULL);
e5508b36 2173 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2174 found = 0;
2175 dev_policy_free(pol);
2176 }
2177 }
2178 if (found) {
2179 dp = &d->next;
2180 disks->array.spare_disks++;
2181 if (get_one) {
2182 sysfs_free(*dp);
2183 d->next = NULL;
2184 }
2185 } else {
2186 *dp = d->next;
2187 d->next = NULL;
2188 sysfs_free(d);
2189 }
2190 }
2191 return disks;
2192}
9eafa1de
MN
2193
2194/* Checks if paths point to the same device
2195 * Returns 0 if they do.
2196 * Returns 1 if they don't.
2197 * Returns -1 if something went wrong,
2198 * e.g. paths are empty or the files
2199 * they point to don't exist */
2200int compare_paths (char* path1, char* path2)
2201{
2202 struct stat st1,st2;
2203
2204 if (path1 == NULL || path2 == NULL)
2205 return -1;
2206 if (stat(path1,&st1) != 0)
2207 return -1;
2208 if (stat(path2,&st2) != 0)
2209 return -1;
2210 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2211 return 0;
2212 return 1;
2213}
a7dec3fd
N
2214
2215/* Make sure we can open as many devices as needed */
2216void enable_fds(int devices)
2217{
2218 unsigned int fds = 20 + devices;
2219 struct rlimit lim;
2220 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
2221 || lim.rlim_cur >= fds)
2222 return;
2223 if (lim.rlim_max < fds)
2224 lim.rlim_max = fds;
2225 lim.rlim_cur = fds;
2226 setrlimit(RLIMIT_NOFILE, &lim);
2227}
a9c15847
N
2228
2229int in_initrd(void)
2230{
2231 /* This is based on similar function in systemd. */
2232 struct statfs s;
76d0f188
JS
2233 /* statfs.f_type is signed long on s390x and MIPS, causing all
2234 sorts of sign extension problems with RAMFS_MAGIC being
2235 defined as 0x858458f6 */
a9c15847 2236 return statfs("/", &s) >= 0 &&
7ac5d47e 2237 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2238 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2239 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2240}
8832342d
N
2241
2242void reopen_mddev(int mdfd)
2243{
2244 /* Re-open without any O_EXCL, but keep
2245 * the same fd
2246 */
2247 char *devnm;
2248 int fd;
2249 devnm = fd2devnm(mdfd);
2250 close(mdfd);
2251 fd = open_dev(devnm);
2252 if (fd >= 0 && fd != mdfd)
2253 dup2(fd, mdfd);
2254}
7716570e 2255
e80357f8
GJ
2256static struct cmap_hooks *cmap_hooks = NULL;
2257static int is_cmap_hooks_ready = 0;
7716570e 2258
e80357f8
GJ
2259void set_cmap_hooks(void)
2260{
2261 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2262 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2263 if (!cmap_hooks->cmap_handle)
2264 return;
7716570e 2265
e80357f8
GJ
2266 cmap_hooks->initialize = dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2267 cmap_hooks->get_string = dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
2268 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2269
e80357f8
GJ
2270 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2271 !cmap_hooks->finalize)
2272 dlclose(cmap_hooks->cmap_handle);
2273 else
2274 is_cmap_hooks_ready = 1;
2275}
7716570e 2276
e80357f8
GJ
2277int get_cluster_name(char **cluster_name)
2278{
2279 int rv = -1;
2280 cmap_handle_t handle;
7716570e 2281
e80357f8
GJ
2282 if (!is_cmap_hooks_ready)
2283 return rv;
7716570e 2284
e80357f8 2285 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2286 if (rv != CS_OK)
2287 goto out;
2288
e80357f8 2289 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2290 if (rv != CS_OK) {
2291 free(*cluster_name);
2292 rv = -1;
2293 goto name_err;
2294 }
2295
2296 rv = 0;
2297name_err:
e80357f8 2298 cmap_hooks->finalize(handle);
7716570e 2299out:
7716570e
GJ
2300 return rv;
2301}
d15a1f72
GJ
2302
2303void set_dlm_hooks(void)
2304{
2305 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2306 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2307 if (!dlm_hooks->dlm_handle)
2308 return;
2309
2310 dlm_hooks->create_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2311 dlm_hooks->release_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
2312 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2313 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2314 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2315 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2316
2317 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2318 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2319 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2320 dlclose(dlm_hooks->dlm_handle);
2321 else
2322 is_dlm_hooks_ready = 1;
2323}
e80357f8
GJ
2324
2325void set_hooks(void)
2326{
2327 set_dlm_hooks();
2328 set_cmap_hooks();
2329}