util: Introduce md_get_disk_info()
[thirdparty/mdadm.git] / util.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
6f02172d 4 * Copyright (C) 2001-2013 Neil Brown <neilb@suse.de>
64c4757e
NB
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
e736b623 22 * Email: <neilb@suse.de>
64c4757e
NB
23 */
24
9a9dab36 25#include "mdadm.h"
64c4757e 26#include "md_p.h"
edd8d13c 27#include <sys/socket.h>
64c4757e 28#include <sys/utsname.h>
9fe32043 29#include <sys/wait.h>
edd8d13c 30#include <sys/un.h>
a7dec3fd 31#include <sys/resource.h>
a9c15847
N
32#include <sys/vfs.h>
33#include <linux/magic.h>
986b8688 34#include <poll.h>
98c6faba 35#include <ctype.h>
a322f70c 36#include <dirent.h>
a931db9e 37#include <signal.h>
7716570e 38#include <dlfcn.h>
7716570e 39
0a816ef9
NB
40
41/*
42 * following taken from linux/blkpg.h because they aren't
43 * anywhere else and it isn't safe to #include linux/ * stuff.
44 */
45
46#define BLKPG _IO(0x12,105)
47
48/* The argument structure */
49struct blkpg_ioctl_arg {
5d500228
N
50 int op;
51 int flags;
52 int datalen;
53 void *data;
0a816ef9
NB
54};
55
56/* The subfunctions (for the op field) */
57#define BLKPG_ADD_PARTITION 1
58#define BLKPG_DEL_PARTITION 2
59
60/* Sizes of name fields. Unused at present. */
61#define BLKPG_DEVNAMELTH 64
62#define BLKPG_VOLNAMELTH 64
63
64/* The data structure for ADD_PARTITION and DEL_PARTITION */
65struct blkpg_partition {
66 long long start; /* starting offset in bytes */
67 long long length; /* length in bytes */
68 int pno; /* partition number */
69 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
70 to be used in kernel messages */
71 char volname[BLKPG_VOLNAMELTH]; /* volume label */
72};
64c4757e 73
0f22b998 74#include "part.h"
056b331e
N
75
76/* Force a compilation error if condition is true */
77#define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
78
79/* Force a compilation error if condition is true, but also produce a
80 result (of value 0 and type size_t), so the expression can be used
81 e.g. in a structure initializer (or where-ever else comma expressions
82 aren't permitted). */
83#define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
84
d15a1f72 85static int is_dlm_hooks_ready = 0;
7071320a
N
86
87int dlm_funs_ready(void)
88{
89 return is_dlm_hooks_ready ? 1 : 0;
90}
91
92#ifndef MDASSEMBLE
93static struct dlm_hooks *dlm_hooks = NULL;
d15a1f72
GJ
94struct dlm_lock_resource *dlm_lock_res = NULL;
95static int ast_called = 0;
96
97struct dlm_lock_resource {
98 dlm_lshandle_t *ls;
99 struct dlm_lksb lksb;
100};
101
d15a1f72
GJ
102/* Using poll(2) to wait for and dispatch ASTs */
103static int poll_for_ast(dlm_lshandle_t ls)
104{
105 struct pollfd pfd;
106
107 pfd.fd = dlm_hooks->ls_get_fd(ls);
108 pfd.events = POLLIN;
109
110 while (!ast_called)
111 {
112 if (poll(&pfd, 1, 0) < 0)
113 {
114 perror("poll");
115 return -1;
116 }
117 dlm_hooks->dispatch(dlm_hooks->ls_get_fd(ls));
118 }
119 ast_called = 0;
120
121 return 0;
122}
123
124static void dlm_ast(void *arg)
125{
126 ast_called = 1;
127}
128
81a8a694 129static char *cluster_name = NULL;
d15a1f72 130/* Create the lockspace, take bitmapXXX locks on all the bitmaps. */
81a8a694 131int cluster_get_dlmlock(int *lockid)
d15a1f72
GJ
132{
133 int ret = -1;
134 char str[64];
135 int flags = LKF_NOQUEUE;
136
81a8a694
GJ
137 ret = get_cluster_name(&cluster_name);
138 if (ret) {
139 pr_err("The md can't get cluster name\n");
140 return -1;
141 }
142
d15a1f72 143 dlm_lock_res = xmalloc(sizeof(struct dlm_lock_resource));
81a8a694 144 dlm_lock_res->ls = dlm_hooks->create_lockspace(cluster_name, O_RDWR);
d15a1f72 145 if (!dlm_lock_res->ls) {
81a8a694 146 pr_err("%s failed to create lockspace\n", cluster_name);
32539f74 147 return -ENOMEM;
d15a1f72
GJ
148 }
149
81a8a694 150 snprintf(str, 64, "bitmap%s", cluster_name);
d15a1f72
GJ
151 ret = dlm_hooks->ls_lock(dlm_lock_res->ls, LKM_PWMODE, &dlm_lock_res->lksb,
152 flags, str, strlen(str), 0, dlm_ast,
153 dlm_lock_res, NULL, NULL);
154 if (ret) {
155 pr_err("error %d when get PW mode on lock %s\n", errno, str);
32539f74
GJ
156 dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1);
157 return ret;
d15a1f72
GJ
158 }
159
160 /* Wait for it to complete */
161 poll_for_ast(dlm_lock_res->ls);
162 *lockid = dlm_lock_res->lksb.sb_lkid;
163
32539f74 164 return dlm_lock_res->lksb.sb_status;
d15a1f72
GJ
165}
166
81a8a694 167int cluster_release_dlmlock(int lockid)
d15a1f72
GJ
168{
169 int ret = -1;
170
81a8a694
GJ
171 if (!cluster_name)
172 return -1;
173
d15a1f72
GJ
174 ret = dlm_hooks->ls_unlock(dlm_lock_res->ls, lockid, 0,
175 &dlm_lock_res->lksb, dlm_lock_res);
176 if (ret) {
177 pr_err("error %d happened when unlock\n", errno);
178 /* XXX make sure the lock is unlocked eventually */
179 goto out;
180 }
181
182 /* Wait for it to complete */
183 poll_for_ast(dlm_lock_res->ls);
184
185 errno = dlm_lock_res->lksb.sb_status;
186 if (errno != EUNLOCK) {
187 pr_err("error %d happened in ast when unlock lockspace\n", errno);
188 /* XXX make sure the lockspace is unlocked eventually */
189 goto out;
190 }
191
81a8a694 192 ret = dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1);
d15a1f72
GJ
193 if (ret) {
194 pr_err("error %d happened when release lockspace\n", errno);
195 /* XXX make sure the lockspace is released eventually */
196 goto out;
197 }
198 free(dlm_lock_res);
199
200out:
201 return ret;
202}
7071320a
N
203#else
204int cluster_get_dlmlock(int *lockid)
205{
206 return -1;
207}
208int cluster_release_dlmlock(int lockid)
209{
210 return -1;
211}
212#endif
d15a1f72 213
9cd39f01
JS
214/*
215 * Get array info from the kernel. Longer term we want to deprecate the
216 * ioctl and get it from sysfs.
217 */
218int md_get_array_info(int fd, struct mdu_array_info_s *array)
219{
220 return ioctl(fd, GET_ARRAY_INFO, array);
221}
222
d97572f5
JS
223/*
224 * Get disk info from the kernel.
225 */
226int md_get_disk_info(int fd, struct mdu_disk_info_s *disk)
227{
228 return ioctl(fd, GET_DISK_INFO, disk);
229}
230
64c4757e
NB
231/*
232 * Parse a 128 bit uuid in 4 integers
233 * format is 32 hexx nibbles with options :.<space> separator
234 * If not exactly 32 hex digits are found, return 0
235 * else return 1
236 */
237int parse_uuid(char *str, int uuid[4])
238{
aba69144
NB
239 int hit = 0; /* number of Hex digIT */
240 int i;
241 char c;
ca3b6696
N
242 for (i = 0; i < 4; i++)
243 uuid[i] = 0;
aba69144 244
ca3b6696 245 while ((c = *str++) != 0) {
aba69144 246 int n;
ca3b6696 247 if (c >= '0' && c <= '9')
aba69144 248 n = c-'0';
ca3b6696 249 else if (c >= 'a' && c <= 'f')
aba69144 250 n = 10 + c - 'a';
ca3b6696 251 else if (c >= 'A' && c <= 'F')
aba69144
NB
252 n = 10 + c - 'A';
253 else if (strchr(":. -", c))
254 continue;
255 else return 0;
256
257 if (hit<32) {
258 uuid[hit/8] <<= 4;
259 uuid[hit/8] += n;
260 }
261 hit++;
82b27616 262 }
aba69144
NB
263 if (hit == 32)
264 return 1;
265 return 0;
64c4757e
NB
266}
267
64c4757e
NB
268/*
269 * Get the md version number.
270 * We use the RAID_VERSION ioctl if it is supported
271 * If not, but we have a block device with major '9', we assume
272 * 0.36.0
273 *
274 * Return version number as 24 but number - assume version parts
275 * always < 255
276 */
277
278int md_get_version(int fd)
279{
5d500228
N
280 struct stat stb;
281 mdu_version_t vers;
282
283 if (fstat(fd, &stb)<0)
284 return -1;
285 if ((S_IFMT&stb.st_mode) != S_IFBLK)
286 return -1;
287
288 if (ioctl(fd, RAID_VERSION, &vers) == 0)
289 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
290 if (errno == EACCES)
291 return -1;
292 if (major(stb.st_rdev) == MD_MAJOR)
293 return (3600);
64c4757e 294 return -1;
64c4757e
NB
295}
296
64c4757e
NB
297int get_linux_version()
298{
299 struct utsname name;
98c6faba 300 char *cp;
f161d047 301 int a = 0, b = 0,c = 0;
64c4757e
NB
302 if (uname(&name) <0)
303 return -1;
304
98c6faba
NB
305 cp = name.release;
306 a = strtoul(cp, &cp, 10);
f161d047
N
307 if (*cp == '.')
308 b = strtoul(cp+1, &cp, 10);
309 if (*cp == '.')
310 c = strtoul(cp+1, &cp, 10);
98c6faba 311
682c7051 312 return (a*1000000)+(b*1000)+c;
64c4757e
NB
313}
314
71204a50 315#ifndef MDASSEMBLE
bc77ed53
DW
316int mdadm_version(char *version)
317{
318 int a, b, c;
319 char *cp;
320
321 if (!version)
322 version = Version;
323
324 cp = strchr(version, '-');
325 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
326 return -1;
327 cp += 3;
328 a = strtoul(cp, &cp, 10);
329 if (*cp != '.')
330 return -1;
331 b = strtoul(cp+1, &cp, 10);
332 if (*cp == '.')
333 c = strtoul(cp+1, &cp, 10);
334 else
335 c = 0;
336 if (*cp != ' ' && *cp != '-')
337 return -1;
338 return (a*1000000)+(b*1000)+c;
339}
340
51d4261c 341unsigned long long parse_size(char *size)
84e11361
N
342{
343 /* parse 'size' which should be a number optionally
344 * followed by 'K', 'M', or 'G'.
345 * Without a suffix, K is assumed.
346 * Number returned is in sectors (half-K)
822e393a 347 * INVALID_SECTORS returned on error.
84e11361
N
348 */
349 char *c;
350 long long s = strtoll(size, &c, 10);
351 if (s > 0) {
352 switch (*c) {
353 case 'K':
354 c++;
355 default:
356 s *= 2;
357 break;
358 case 'M':
359 c++;
360 s *= 1024 * 2;
361 break;
362 case 'G':
363 c++;
364 s *= 1024 * 1024 * 2;
365 break;
72ca9bcf
N
366 case 's': /* sectors */
367 c++;
368 break;
84e11361 369 }
51d4261c 370 } else
822e393a 371 s = INVALID_SECTORS;
84e11361 372 if (*c)
822e393a 373 s = INVALID_SECTORS;
84e11361
N
374 return s;
375}
376
4a06e2c2
N
377int parse_layout_10(char *layout)
378{
379 int copies, rv;
380 char *cp;
381 /* Parse the layout string for raid10 */
382 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
383 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
384 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
385 copies > 200 ||
386 *cp)
387 return -1;
388 if (layout[0] == 'n')
389 rv = 256 + copies;
390 else if (layout[0] == 'o')
391 rv = 0x10000 + (copies<<8) + 1;
392 else
393 rv = 1 + (copies<<8);
394 return rv;
395}
396
397int parse_layout_faulty(char *layout)
398{
399 /* Parse the layout string for 'faulty' */
400 int ln = strcspn(layout, "0123456789");
503975b9 401 char *m = xstrdup(layout);
4a06e2c2
N
402 int mode;
403 m[ln] = 0;
404 mode = map_name(faultylayout, m);
405 if (mode == UnSet)
406 return -1;
407
408 return mode | (atoi(layout+ln)<< ModeShift);
409}
79868890
N
410
411long parse_num(char *num)
412{
413 /* Either return a valid number, or -1 */
414 char *c;
415 long rv = strtol(num, &c, 10);
416 if (rv < 0 || *c || !num[0])
417 return -1;
418 else
419 return rv;
420}
eb3929a4 421#endif
4a06e2c2 422
4de90913
GJ
423int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
424{
425 char *dev;
426 *slot = strtoul(input, &dev, 10);
427 if (dev == input || dev[0] != ':')
428 return -1;
429 *devname = dev+1;
430 return 0;
431}
432
0430ed48
NB
433void remove_partitions(int fd)
434{
435 /* remove partitions from this block devices.
436 * This is used for components added to an array
437 */
438#ifdef BLKPG_DEL_PARTITION
439 struct blkpg_ioctl_arg a;
440 struct blkpg_partition p;
441
442 a.op = BLKPG_DEL_PARTITION;
443 a.data = (void*)&p;
444 a.datalen = sizeof(p);
445 a.flags = 0;
446 memset(a.data, 0, a.datalen);
ca3b6696 447 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
448 ioctl(fd, BLKPG, &a);
449#endif
450}
451
691c6ee1
N
452int test_partition(int fd)
453{
454 /* Check if fd is a whole-disk or a partition.
455 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
456 * will return ENXIO on an invalid partition number.
457 */
458 struct blkpg_ioctl_arg a;
459 struct blkpg_partition p;
460 a.op = BLKPG_DEL_PARTITION;
461 a.data = (void*)&p;
462 a.datalen = sizeof(p);
463 a.flags = 0;
464 memset(a.data, 0, a.datalen);
465 p.pno = 1<<30;
466 if (ioctl(fd, BLKPG, &a) == 0)
467 /* Very unlikely, but not a partition */
468 return 0;
357ac106 469 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
470 /* not a partition */
471 return 0;
472
473 return 1;
474}
475
bfd76b93
CA
476int test_partition_from_id(dev_t id)
477{
478 char buf[20];
479 int fd, rv;
480
481 sprintf(buf, "%d:%d", major(id), minor(id));
482 fd = dev_open(buf, O_RDONLY);
483 if (fd < 0)
484 return -1;
485 rv = test_partition(fd);
486 close(fd);
487 return rv;
488}
691c6ee1 489
de5a472e 490int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 491{
265e0f17 492 int copies, first;
de5a472e
N
493 int i;
494 int avail_disks = 0;
495
496 for (i = 0; i < raid_disks; i++)
497 avail_disks += !!avail[i];
498
64c4757e 499 switch (level) {
265e0f17
NB
500 case 10:
501 /* This is the tricky one - we need to check
502 * which actual disks are present.
503 */
702b557b 504 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 505 first = 0;
265e0f17
NB
506 do {
507 /* there must be one of the 'copies' form 'first' */
508 int n = copies;
ca3b6696 509 int cnt = 0;
b3ec716d 510 int this = first;
265e0f17 511 while (n--) {
b3ec716d 512 if (avail[this])
265e0f17 513 cnt++;
b3ec716d 514 this = (this+1) % raid_disks;
265e0f17
NB
515 }
516 if (cnt == 0)
517 return 0;
b3ec716d 518 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
519 } while (first != 0);
520 return 1;
e5329c37 521
df0d4ea0 522 case LEVEL_MULTIPATH:
e0d19036 523 return avail_disks>= 1;
df0d4ea0 524 case LEVEL_LINEAR:
64c4757e
NB
525 case 0:
526 return avail_disks == raid_disks;
527 case 1:
528 return avail_disks >= 1;
529 case 4:
330d6900
N
530 if (avail_disks == raid_disks - 1 &&
531 !avail[raid_disks - 1])
532 /* If just the parity device is missing, then we
533 * have enough, even if not clean
534 */
535 return 1;
536 /* FALL THROUGH */
64c4757e 537 case 5:
583315d9
NB
538 if (clean)
539 return avail_disks >= raid_disks-1;
540 else
541 return avail_disks >= raid_disks;
98c6faba 542 case 6:
583315d9
NB
543 if (clean)
544 return avail_disks >= raid_disks-2;
545 else
546 return avail_disks >= raid_disks;
64c4757e
NB
547 default:
548 return 0;
549 }
550}
551
8453e704
N
552int enough_fd(int fd)
553{
554 struct mdu_array_info_s array;
555 struct mdu_disk_info_s disk;
aabe020d 556 int i, rv;
8453e704
N
557 char *avail;
558
9cd39f01 559 if (md_get_array_info(fd, &array) != 0 || array.raid_disks <= 0)
8453e704 560 return 0;
503975b9 561 avail = xcalloc(array.raid_disks, 1);
ca3b6696 562 for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704 563 disk.number = i;
d97572f5 564 if (md_get_disk_info(fd, &disk) != 0)
8453e704 565 continue;
9e6d9291
N
566 if (disk.major == 0 && disk.minor == 0)
567 continue;
568 array.nr_disks--;
569
8453e704
N
570 if (! (disk.state & (1<<MD_DISK_SYNC)))
571 continue;
572 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
573 continue;
8453e704
N
574 avail[disk.raid_disk] = 1;
575 }
576 /* This is used on an active array, so assume it is clean */
aabe020d 577 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 578 1, avail);
aabe020d
JS
579 free(avail);
580 return rv;
8453e704
N
581}
582
22e263f6
AC
583const int uuid_zero[4] = { 0, 0, 0, 0 };
584
f277ce36 585int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 586{
f277ce36
NB
587 if (swapuuid) {
588 /* parse uuids are hostendian.
589 * uuid's from some superblocks are big-ending
aba69144 590 * if there is a difference, we need to swap..
f277ce36
NB
591 */
592 unsigned char *ac = (unsigned char *)a;
593 unsigned char *bc = (unsigned char *)b;
594 int i;
ca3b6696 595 for (i = 0; i < 16; i += 4) {
f277ce36
NB
596 if (ac[i+0] != bc[i+3] ||
597 ac[i+1] != bc[i+2] ||
598 ac[i+2] != bc[i+1] ||
599 ac[i+3] != bc[i+0])
600 return 0;
601 }
602 return 1;
603 } else {
604 if (a[0]==b[0] &&
605 a[1]==b[1] &&
606 a[2]==b[2] &&
607 a[3]==b[3])
608 return 1;
609 return 0;
610 }
64c4757e 611}
ca3b6696 612
350f29f9
NB
613void copy_uuid(void *a, int b[4], int swapuuid)
614{
615 if (swapuuid) {
616 /* parse uuids are hostendian.
617 * uuid's from some superblocks are big-ending
618 * if there is a difference, we need to swap..
619 */
620 unsigned char *ac = (unsigned char *)a;
621 unsigned char *bc = (unsigned char *)b;
622 int i;
ca3b6696 623 for (i = 0; i < 16; i += 4) {
350f29f9
NB
624 ac[i+0] = bc[i+3];
625 ac[i+1] = bc[i+2];
626 ac[i+2] = bc[i+1];
627 ac[i+3] = bc[i+0];
628 }
629 } else
630 memcpy(a, b, 16);
631}
64c4757e 632
aae5a112 633char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 634{
9968e376 635 int i, j;
d7288ddc
N
636 char uuid[16];
637 char *c = buf;
638 strcpy(c, "UUID-");
639 c += strlen(c);
aae5a112 640 copy_uuid(uuid, id, swap);
9968e376 641 for (i = 0; i < 4; i++) {
9968e376 642 if (i)
ff54de6e 643 *c++ = sep;
9968e376
DW
644 for (j = 3; j >= 0; j--) {
645 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
646 c+= 2;
647 }
d7288ddc
N
648 }
649 return buf;
aae5a112
DW
650
651}
652
653char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
654{
86983cce
N
655 // dirty hack to work around an issue with super1 superblocks...
656 // super1 superblocks need swapuuid set in order for assembly to
657 // work, but can't have it set if we want this printout to match
658 // all the other uuid printouts in super1.c, so we force swapuuid
659 // to 1 to make our printout match the rest of super1
660 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
661}
662
435d4ebb 663#ifndef MDASSEMBLE
682c7051
NB
664int check_ext2(int fd, char *name)
665{
666 /*
667 * Check for an ext2fs file system.
668 * Superblock is always 1K at 1K offset
669 *
670 * s_magic is le16 at 56 == 0xEF53
671 * report mtime - le32 at 44
672 * blocks - le32 at 4
673 * logblksize - le32 at 24
674 */
675 unsigned char sb[1024];
676 time_t mtime;
2538ba2a
N
677 unsigned long long size;
678 int bsize;
682c7051
NB
679 if (lseek(fd, 1024,0)!= 1024)
680 return 0;
681 if (read(fd, sb, 1024)!= 1024)
682 return 0;
683 if (sb[56] != 0x53 || sb[57] != 0xef)
684 return 0;
685
686 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
687 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
688 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 689 size <<= bsize;
e7b84f9d 690 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 691 name);
2538ba2a 692 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
693 return 1;
694}
695
696int check_reiser(int fd, char *name)
697{
698 /*
699 * superblock is at 64K
700 * size is 1024;
701 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
702 *
703 */
704 unsigned char sb[1024];
2538ba2a 705 unsigned long long size;
682c7051
NB
706 if (lseek(fd, 64*1024, 0) != 64*1024)
707 return 0;
708 if (read(fd, sb, 1024) != 1024)
709 return 0;
ca3b6696
N
710 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
711 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 712 return 0;
e7b84f9d 713 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 714 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 715 cont_err("size = %lluK\n", size*4);
aba69144 716
682c7051
NB
717 return 1;
718}
719
720int check_raid(int fd, char *name)
721{
4b1ac34b 722 struct mdinfo info;
682c7051 723 time_t crtime;
d078d77c 724 char *level;
82d9eba6 725 struct supertype *st = guess_super(fd);
f9ce90ba 726
cb8f6859
N
727 if (!st)
728 return 0;
fa219dd2
MD
729 if (st->ss->add_to_super != NULL) {
730 st->ss->load_super(st, fd, name);
731 /* Looks like a raid array .. */
732 pr_err("%s appears to be part of a raid array:\n", name);
733 st->ss->getinfo_super(st, &info, NULL);
734 st->ss->free_super(st);
735 crtime = info.array.ctime;
736 level = map_num(pers, info.array.level);
737 if (!level)
738 level = "-unknown-";
739 cont_err("level=%s devices=%d ctime=%s",
740 level, info.array.raid_disks, ctime(&crtime));
741 } else {
742 /* Looks like GPT or MBR */
743 pr_err("partition table exists on %s\n", name);
744 }
82d9eba6 745 return 1;
682c7051
NB
746}
747
682c7051
NB
748int ask(char *mesg)
749{
750 char *add = "";
751 int i;
ca3b6696 752 for (i = 0; i < 5; i++) {
682c7051
NB
753 char buf[100];
754 fprintf(stderr, "%s%s", mesg, add);
755 fflush(stderr);
756 if (fgets(buf, 100, stdin)==NULL)
757 return 0;
758 if (buf[0]=='y' || buf[0]=='Y')
759 return 1;
760 if (buf[0]=='n' || buf[0]=='N')
761 return 0;
762 add = "(y/n) ";
763 }
e7b84f9d 764 pr_err("assuming 'no'\n");
682c7051
NB
765 return 0;
766}
435d4ebb 767#endif /* MDASSEMBLE */
682c7051 768
8d80900b 769int is_standard(char *dev, int *nump)
e5329c37
NB
770{
771 /* tests if dev is a "standard" md dev name.
772 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 773 * where NN is a string of digits
598f0d58
NB
774 * Returns 1 if a partitionable standard,
775 * -1 if non-partitonable,
776 * 0 if not a standard name.
e5329c37 777 */
8d80900b 778 char *d = strrchr(dev, '/');
ca3b6696 779 int type = 0;
8d80900b
NB
780 int num;
781 if (!d)
e5329c37 782 return 0;
ca3b6696
N
783 if (strncmp(d, "/d",2) == 0)
784 d += 2, type = 1; /* /dev/md/dN{pM} */
785 else if (strncmp(d, "/md_d", 5) == 0)
786 d += 5, type = 1; /* /dev/md_dN{pM} */
787 else if (strncmp(d, "/md", 3) == 0)
788 d += 3, type = -1; /* /dev/mdN */
789 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
790 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
791 else
792 return 0;
8d80900b 793 if (!*d)
e5329c37 794 return 0;
8d80900b
NB
795 num = atoi(d);
796 while (isdigit(*d))
797 d++;
798 if (*d)
e5329c37 799 return 0;
8d80900b
NB
800 if (nump) *nump = num;
801
802 return type;
e5329c37
NB
803}
804
4b1ac34b 805unsigned long calc_csum(void *super, int bytes)
82b27616 806{
56eb10c0 807 unsigned long long newcsum = 0;
82b27616 808 int i;
4b1ac34b
NB
809 unsigned int csum;
810 unsigned int *superc = (unsigned int*) super;
82b27616 811
ca3b6696
N
812 for(i = 0; i < bytes/4; i++)
813 newcsum += superc[i];
82b27616 814 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 815#ifdef __alpha__
aba69144 816/* The in-kernel checksum calculation is always 16bit on
570c0542 817 * the alpha, though it is 32 bit on i386...
ca3b6696 818 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
819 * a way that it shouldn't).
820 */
821 csum = (csum & 0xffff) + (csum >> 16);
822 csum = (csum & 0xffff) + (csum >> 16);
823#endif
82b27616
NB
824 return csum;
825}
cd29a5c8 826
435d4ebb 827#ifndef MDASSEMBLE
56eb10c0 828char *human_size(long long bytes)
cd29a5c8 829{
ff9239ee 830 static char buf[47];
d5d3721e
NB
831
832 /* We convert bytes to either centi-M{ega,ibi}bytes or
833 * centi-G{igi,ibi}bytes, with appropriate rounding,
834 * and then print 1/100th of those as a decimal.
835 * We allow upto 2048Megabytes before converting to
836 * gigabytes, as that shows more precision and isn't
837 * too large a number.
01619b48 838 * Terabytes are not yet handled.
d5d3721e 839 */
cd29a5c8 840
56eb10c0 841 if (bytes < 5000*1024)
ca3b6696 842 buf[0] = 0;
d5d3721e 843 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 844 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 845 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 846 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
efa29530 847 cMiB/100, cMiB % 100, cMB/100, cMB % 100);
d5d3721e 848 } else {
93d3bd3b 849 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 850 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 851 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
efa29530 852 cGiB/100, cGiB % 100, cGB/100, cGB % 100);
d5d3721e 853 }
cd29a5c8
NB
854 return buf;
855}
e0d19036 856
f0ec6710 857char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
858{
859 static char buf[30];
e0d19036 860
570abc6f
MN
861 /* We convert bytes to either centi-M{ega,ibi}bytes or
862 * centi-G{igi,ibi}bytes, with appropriate rounding,
863 * and then print 1/100th of those as a decimal.
864 * We allow upto 2048Megabytes before converting to
865 * gigabytes, as that shows more precision and isn't
866 * too large a number.
867 * Terabytes are not yet handled.
f0ec6710
MN
868 *
869 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
870 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
871 */
872
e0d19036 873 if (bytes < 5000*1024)
570abc6f 874 buf[0] = 0;
f0ec6710
MN
875 else if (prefix == IEC) {
876 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 877 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710 878 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
efa29530 879 cMiB/100, cMiB % 100);
f0ec6710 880 } else {
93d3bd3b 881 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710 882 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
efa29530 883 cGiB/100, cGiB % 100);
f0ec6710
MN
884 }
885 }
886 else if (prefix == JEDEC) {
887 if (bytes < 2*1024LL*1024LL*1024LL) {
888 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
889 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
efa29530 890 cMB/100, cMB % 100);
f0ec6710
MN
891 } else {
892 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
893 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
efa29530 894 cGB/100, cGB % 100);
f0ec6710 895 }
570abc6f 896 }
f0ec6710
MN
897 else
898 buf[0] = 0;
570abc6f 899
e0d19036
NB
900 return buf;
901}
e4965ef8
N
902
903void print_r10_layout(int layout)
904{
905 int near = layout & 255;
906 int far = (layout >> 8) & 255;
907 int offset = (layout&0x10000);
908 char *sep = "";
909
910 if (near != 1) {
911 printf("%s near=%d", sep, near);
912 sep = ",";
913 }
914 if (far != 1)
915 printf("%s %s=%d", sep, offset?"offset":"far", far);
916 if (near*far == 1)
917 printf("NO REDUNDANCY");
918}
435d4ebb 919#endif
e0d19036 920
5f8097be
NB
921unsigned long long calc_array_size(int level, int raid_disks, int layout,
922 int chunksize, unsigned long long devsize)
577e8448 923{
e03640bd
CA
924 if (level == 1)
925 return devsize;
577e8448
AK
926 devsize &= ~(unsigned long long)((chunksize>>9)-1);
927 return get_data_disks(level, layout, raid_disks) * devsize;
928}
929
930int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
931{
932 int data_disks = 0;
933 switch (level) {
ca3b6696
N
934 case 0: data_disks = raid_disks;
935 break;
936 case 1: data_disks = 1;
937 break;
5f8097be 938 case 4:
ca3b6696
N
939 case 5: data_disks = raid_disks - 1;
940 break;
941 case 6: data_disks = raid_disks - 2;
942 break;
5f8097be
NB
943 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
944 break;
945 }
577e8448
AK
946
947 return data_disks;
5f8097be
NB
948}
949
13db17bd 950dev_t devnm2devid(char *devnm)
4dd2df09
N
951{
952 /* First look in /sys/block/$DEVNM/dev for %d:%d
953 * If that fails, try parsing out a number
954 */
955 char path[100];
956 char *ep;
957 int fd;
958 int mjr,mnr;
959
960 sprintf(path, "/sys/block/%s/dev", devnm);
961 fd = open(path, O_RDONLY);
962 if (fd >= 0) {
963 char buf[20];
964 int n = read(fd, buf, sizeof(buf));
965 close(fd);
966 if (n > 0)
967 buf[n] = 0;
968 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
969 return makedev(mjr, mnr);
970 }
971 if (strncmp(devnm, "md_d", 4) == 0 &&
972 isdigit(devnm[4]) &&
973 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
974 ep > devnm && *ep == 0)
975 return makedev(get_mdp_major(), mnr << MdpMinorShift);
976
977 if (strncmp(devnm, "md", 2) == 0 &&
978 isdigit(devnm[2]) &&
979 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
980 ep > devnm && *ep == 0)
981 return makedev(MD_MAJOR, mnr);
982
983 return 0;
984}
985
5a23a06e 986#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 987char *get_md_name(char *devnm)
e0d19036
NB
988{
989 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 990 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 991
e0d19036
NB
992 static char devname[50];
993 struct stat stb;
4dd2df09 994 dev_t rdev = devnm2devid(devnm);
dd0781e5 995 char *dn;
98c6faba 996
4dd2df09
N
997 if (rdev == 0)
998 return 0;
999 if (strncmp(devnm, "md_", 3) == 0) {
1000 snprintf(devname, sizeof(devname), "/dev/md/%s",
1001 devnm + 3);
98c6faba
NB
1002 if (stat(devname, &stb) == 0
1003 && (S_IFMT&stb.st_mode) == S_IFBLK
1004 && (stb.st_rdev == rdev))
1005 return devname;
1006 }
4dd2df09
N
1007 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
1008 if (stat(devname, &stb) == 0
1009 && (S_IFMT&stb.st_mode) == S_IFBLK
1010 && (stb.st_rdev == rdev))
1011 return devname;
1012
1013 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
1014 if (stat(devname, &stb) == 0
1015 && (S_IFMT&stb.st_mode) == S_IFBLK
1016 && (stb.st_rdev == rdev))
1017 return devname;
1018
16c6fa80 1019 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
1020 if (dn)
1021 return dn;
4dd2df09 1022 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 1023 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
1024 if (errno != EEXIST)
1025 return NULL;
e0d19036
NB
1026
1027 if (stat(devname, &stb) == 0
1028 && (S_IFMT&stb.st_mode) == S_IFBLK
1029 && (stb.st_rdev == rdev))
1030 return devname;
1031 unlink(devname);
1032 return NULL;
1033}
1034
1035void put_md_name(char *name)
1036{
ca3b6696 1037 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
1038 unlink(name);
1039}
435d4ebb 1040#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 1041
cc742d38
N
1042int get_maj_min(char *dev, int *major, int *minor)
1043{
1044 char *e;
1045 *major = strtoul(dev, &e, 0);
1046 return (e > dev && *e == ':' && e[1] &&
1047 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1048 *e == 0);
1049}
1050
8b0dabea
NB
1051int dev_open(char *dev, int flags)
1052{
1053 /* like 'open', but if 'dev' matches %d:%d, create a temp
1054 * block device and open that
1055 */
8b0dabea
NB
1056 int fd = -1;
1057 char devname[32];
e81cdd9f 1058 int major;
8b0dabea 1059 int minor;
e81cdd9f 1060
9f0ad56b
JS
1061 if (!dev)
1062 return -1;
6df6a774 1063 flags |= O_DIRECT;
e81cdd9f 1064
cc742d38 1065 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1066 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1067 (int)getpid(), major, minor);
1068 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1069 fd = open(devname, flags);
1070 unlink(devname);
6df6a774
N
1071 }
1072 if (fd < 0) {
9dc7d357 1073 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
1074 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
1075 (int)getpid(), major, minor);
ca3b6696 1076 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
1077 fd = open(devname, flags);
1078 unlink(devname);
1079 }
8b0dabea
NB
1080 }
1081 } else
6df6a774 1082 fd = open(dev, flags);
8b0dabea
NB
1083 return fd;
1084}
f9ce90ba 1085
4dd2df09 1086int open_dev_flags(char *devnm, int flags)
e8a70c89 1087{
13db17bd 1088 dev_t devid;
e8a70c89
N
1089 char buf[20];
1090
4dd2df09
N
1091 devid = devnm2devid(devnm);
1092 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1093 return dev_open(buf, flags);
1094}
1095
4dd2df09 1096int open_dev(char *devnm)
d998b738 1097{
4dd2df09 1098 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1099}
1100
4dd2df09 1101int open_dev_excl(char *devnm)
a931db9e
NB
1102{
1103 char buf[20];
1104 int i;
7187750e 1105 int flags = O_RDWR;
13db17bd 1106 dev_t devid = devnm2devid(devnm);
399e0b97 1107 long delay = 1000;
a931db9e 1108
4dd2df09 1109 sprintf(buf, "%d:%d", major(devid), minor(devid));
efa29530 1110 for (i = 0; i < 25; i++) {
7187750e 1111 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1112 if (fd >= 0)
1113 return fd;
7187750e
N
1114 if (errno == EACCES && flags == O_RDWR) {
1115 flags = O_RDONLY;
1116 continue;
1117 }
a931db9e
NB
1118 if (errno != EBUSY)
1119 return fd;
399e0b97
N
1120 usleep(delay);
1121 if (delay < 200000)
1122 delay *= 2;
a931db9e
NB
1123 }
1124 return -1;
1125}
1126
9008ed1c
N
1127int same_dev(char *one, char *two)
1128{
1129 struct stat st1, st2;
1130 if (stat(one, &st1) != 0)
1131 return 0;
1132 if (stat(two, &st2) != 0)
1133 return 0;
1134 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1135 return 0;
1136 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1137 return 0;
1138 return st1.st_rdev == st2.st_rdev;
1139}
1140
a7c6e3fb 1141void wait_for(char *dev, int fd)
a714580e
N
1142{
1143 int i;
a7c6e3fb 1144 struct stat stb_want;
399e0b97 1145 long delay = 1000;
a7c6e3fb
N
1146
1147 if (fstat(fd, &stb_want) != 0 ||
1148 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1149 return;
a714580e 1150
efa29530 1151 for (i = 0; i < 25; i++) {
a714580e 1152 struct stat stb;
a7c6e3fb
N
1153 if (stat(dev, &stb) == 0 &&
1154 (stb.st_mode & S_IFMT) == S_IFBLK &&
1155 (stb.st_rdev == stb_want.st_rdev))
a714580e 1156 return;
399e0b97
N
1157 usleep(delay);
1158 if (delay < 200000)
1159 delay *= 2;
a714580e 1160 }
436305c6 1161 if (i == 25)
1ade5cc1 1162 dprintf("timeout waiting for %s\n", dev);
a714580e
N
1163}
1164
0f22b998
N
1165struct superswitch *superlist[] =
1166{
1167 &super0, &super1,
1168 &super_ddf, &super_imsm,
0592faeb 1169 &mbr, &gpt,
0f22b998 1170 NULL };
f9ce90ba 1171
ea24acd0 1172#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1173
4725bc31 1174struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1175{
1686dc25
NB
1176 mdu_array_info_t array;
1177 int vers;
1178 int minor;
1179 struct supertype *st = NULL;
7e0f6979 1180 struct mdinfo *sra;
142cb9e1 1181 char *verstr;
1686dc25
NB
1182 char version[20];
1183 int i;
f7e7067b 1184 char *subarray = NULL;
4dd2df09 1185 char container[32] = "";
1686dc25 1186
4dd2df09 1187 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1188
1189 if (sra) {
7e0f6979
NB
1190 vers = sra->array.major_version;
1191 minor = sra->array.minor_version;
142cb9e1 1192 verstr = sra->text_version;
1686dc25 1193 } else {
9cd39f01 1194 if (md_get_array_info(fd, &array))
1686dc25
NB
1195 array.major_version = array.minor_version = 0;
1196 vers = array.major_version;
1197 minor = array.minor_version;
142cb9e1 1198 verstr = "";
6fbba4c9 1199 }
82d9eba6 1200
1686dc25
NB
1201 if (vers != -1) {
1202 sprintf(version, "%d.%d", vers, minor);
1203 verstr = version;
6fbba4c9 1204 }
3c558363 1205 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1206 char *dev = verstr+1;
5f7e44b2 1207
f7e7067b 1208 subarray = strchr(dev, '/');
92d49ecf 1209 if (subarray) {
f7e7067b 1210 *subarray++ = '\0';
503975b9 1211 subarray = xstrdup(subarray);
92d49ecf 1212 }
4dd2df09 1213 strcpy(container, dev);
15d230f7 1214 sysfs_free(sra);
5f7e44b2 1215 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1216 if (sra && sra->text_version[0])
1217 verstr = sra->text_version;
1218 else
1219 verstr = "-no-metadata-";
f7e7067b
NB
1220 }
1221
efa29530 1222 for (i = 0; st == NULL && superlist[i]; i++)
f7e7067b 1223 st = superlist[i]->match_metadata_desc(verstr);
1686dc25 1224
15d230f7 1225 sysfs_free(sra);
f7e7067b 1226 if (st) {
3b0896f8 1227 st->sb = NULL;
1f49fb3a
N
1228 if (subarrayp)
1229 *subarrayp = subarray;
4dd2df09
N
1230 strcpy(st->container_devnm, container);
1231 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1232 } else
1233 free(subarray);
5f7e44b2 1234
82d9eba6 1235 return st;
f9ce90ba 1236}
ea24acd0
NB
1237#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1238
0f0749ad 1239int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1240{
1241 char buf[20];
1242 int fd;
1243
1244 sprintf(buf, "%d:%d", major(id), minor(id));
1245 fd = dev_open(buf, O_RDONLY);
1246 if (fd < 0)
1247 return 0;
1248 if (get_dev_size(fd, NULL, size)) {
1249 close(fd);
1250 return 1;
1251 }
1252 close(fd);
1253 return 0;
1254}
f9ce90ba 1255
159c3a1a 1256struct supertype *dup_super(struct supertype *orig)
3da92f27 1257{
159c3a1a 1258 struct supertype *st;
1686dc25 1259
d2ca6449
NB
1260 if (!orig)
1261 return orig;
503975b9 1262 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1263 st->ss = orig->ss;
1264 st->max_devs = orig->max_devs;
1265 st->minor_version = orig->minor_version;
cb8f6859 1266 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1267 st->data_offset = orig->data_offset;
159c3a1a
NB
1268 st->sb = NULL;
1269 st->info = NULL;
1270 return st;
3da92f27
NB
1271}
1272
54887ad8 1273struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1274{
1275 /* try each load_super to find the best match,
1276 * and return the best superswitch
1277 */
82d9eba6
NB
1278 struct superswitch *ss;
1279 struct supertype *st;
1d13b599 1280 unsigned int besttime = 0;
570c0542 1281 int bestsuper = -1;
f9ce90ba
NB
1282 int i;
1283
503975b9 1284 st = xcalloc(1, sizeof(*st));
4dd2df09 1285 st->container_devnm[0] = 0;
d1d599ea 1286
efa29530 1287 for (i = 0; superlist[i]; i++) {
f9ce90ba
NB
1288 int rv;
1289 ss = superlist[i];
54887ad8
N
1290 if (guess_type == guess_array && ss->add_to_super == NULL)
1291 continue;
1292 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1293 continue;
ef609477 1294 memset(st, 0, sizeof(*st));
df3346e6 1295 st->ignore_hw_compat = 1;
3da92f27 1296 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1297 if (rv == 0) {
1298 struct mdinfo info;
a5d85af7 1299 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1300 if (bestsuper == -1 ||
1301 besttime < info.array.ctime) {
1302 bestsuper = i;
1303 besttime = info.array.ctime;
570c0542 1304 }
3da92f27 1305 ss->free_super(st);
570c0542
NB
1306 }
1307 }
1308 if (bestsuper != -1) {
1309 int rv;
ef609477 1310 memset(st, 0, sizeof(*st));
df3346e6 1311 st->ignore_hw_compat = 1;
3da92f27 1312 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1313 if (rv == 0) {
5e747af2 1314 superlist[bestsuper]->free_super(st);
82d9eba6 1315 return st;
f9ce90ba
NB
1316 }
1317 }
570c0542 1318 free(st);
f9ce90ba
NB
1319 return NULL;
1320}
fe6729fa 1321
beae1dfe
NB
1322/* Return size of device in bytes */
1323int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1324{
1325 unsigned long long ldsize;
c2c9bb6f
NB
1326 struct stat st;
1327
1328 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1329 ldsize = (unsigned long long)st.st_size;
1330 else
beae1dfe
NB
1331#ifdef BLKGETSIZE64
1332 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1333#endif
1334 {
1335 unsigned long dsize;
1336 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1337 ldsize = dsize;
1338 ldsize <<= 9;
1339 } else {
1340 if (dname)
32971509 1341 pr_err("Cannot get size of %s: %s\n",
beae1dfe
NB
1342 dname, strerror(errno));
1343 return 0;
1344 }
1345 }
1346 *sizep = ldsize;
1347 return 1;
1348}
8fac0577 1349
32971509
PB
1350/* Return sector size of device in bytes */
1351int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep)
1352{
1353 unsigned int sectsize;
1354
1355 if (ioctl(fd, BLKSSZGET, &sectsize) != 0) {
1356 if (dname)
1357 pr_err("Cannot get sector size of %s: %s\n",
1358 dname, strerror(errno));
1359 return 0;
1360 }
1361
1362 *sectsizep = sectsize;
1363 return 1;
1364}
1365
3a371610
N
1366/* Return true if this can only be a container, not a member device.
1367 * i.e. is and md device and size is zero
1368 */
1369int must_be_container(int fd)
1370{
1371 unsigned long long size;
1372 if (md_get_version(fd) < 0)
1373 return 0;
1374 if (get_dev_size(fd, NULL, &size) == 0)
1375 return 1;
1376 if (size == 0)
1377 return 1;
1378 return 0;
1379}
034b203a
TM
1380
1381/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1382 * Returns: 1 if successful
1383 * -1 for unknown partition type
1384 * 0 for other errors
1385 */
1386static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1387{
056b331e 1388 struct GPT gpt;
034b203a 1389 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1390 struct GPT_part_entry *part;
1391 char buf[512];
034b203a
TM
1392 unsigned long long curr_part_end;
1393 unsigned all_partitions, entry_size;
f21e18ca 1394 unsigned part_nr;
41b06495 1395 unsigned int sector_size = 0;
034b203a
TM
1396
1397 *endofpart = 0;
1398
056b331e 1399 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1400 /* skip protective MBR */
41b06495
MD
1401 if (!get_dev_sector_size(fd, NULL, &sector_size))
1402 return 0;
1403 lseek(fd, sector_size, SEEK_SET);
73e658d8 1404 /* read GPT header */
056b331e 1405 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1406 return 0;
1407
1408 /* get the number of partition entries and the entry size */
056b331e
N
1409 all_partitions = __le32_to_cpu(gpt.part_cnt);
1410 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1411
1412 /* Check GPT signature*/
056b331e 1413 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1414 return -1;
1415
1416 /* sanity checks */
1417 if (all_partitions > 1024 ||
db7fdfe4 1418 entry_size > sizeof(buf))
034b203a
TM
1419 return -1;
1420
db7fdfe4
JS
1421 part = (struct GPT_part_entry *)buf;
1422
41b06495
MD
1423 /* set offset to third block (GPT entries) */
1424 lseek(fd, sector_size*2, SEEK_SET);
ca3b6696 1425 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1426 /* read partition entry */
db7fdfe4 1427 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1428 return 0;
1429
034b203a 1430 /* is this valid partition? */
db7fdfe4 1431 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1432 /* check the last lba for the current partition */
db7fdfe4 1433 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1434 if (curr_part_end > *endofpart)
1435 *endofpart = curr_part_end;
1436 }
1437
034b203a
TM
1438 }
1439 return 1;
1440}
1441
1442/* Sets endofpart parameter to the last block used by the last partition on the device.
1443 * Returns: 1 if successful
1444 * -1 for unknown partition type
1445 * 0 for other errors
1446 */
1447static int get_last_partition_end(int fd, unsigned long long *endofpart)
1448{
056b331e 1449 struct MBR boot_sect;
034b203a 1450 unsigned long long curr_part_end;
f21e18ca 1451 unsigned part_nr;
31208db9 1452 unsigned int sector_size;
034b203a
TM
1453 int retval = 0;
1454
1455 *endofpart = 0;
1456
056b331e 1457 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1458 /* read MBR */
1459 lseek(fd, 0, 0);
056b331e 1460 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1461 goto abort;
1462
1463 /* check MBP signature */
056b331e 1464 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1465 retval = 1;
1466 /* found the correct signature */
034b203a 1467
ca3b6696 1468 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
8e2bca51
JC
1469 /*
1470 * Have to make every access through boot_sect rather
1471 * than using a pointer to the partition table (or an
1472 * entry), since the entries are not properly aligned.
1473 */
1474
034b203a 1475 /* check for GPT type */
8e2bca51
JC
1476 if (boot_sect.parts[part_nr].part_type ==
1477 MBR_GPT_PARTITION_TYPE) {
034b203a
TM
1478 retval = get_gpt_last_partition_end(fd, endofpart);
1479 break;
1480 }
1481 /* check the last used lba for the current partition */
8e2bca51
JC
1482 curr_part_end =
1483 __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) +
1484 __le32_to_cpu(boot_sect.parts[part_nr].blocks_num);
034b203a
TM
1485 if (curr_part_end > *endofpart)
1486 *endofpart = curr_part_end;
034b203a
TM
1487 }
1488 } else {
1489 /* Unknown partition table */
1490 retval = -1;
1491 }
31208db9
MD
1492 /* calculate number of 512-byte blocks */
1493 if (get_dev_sector_size(fd, NULL, &sector_size))
1494 *endofpart *= (sector_size / 512);
034b203a
TM
1495 abort:
1496 return retval;
1497}
1498
53ed6ac3
KW
1499int check_partitions(int fd, char *dname, unsigned long long freesize,
1500 unsigned long long size)
034b203a
TM
1501{
1502 /*
1503 * Check where the last partition ends
1504 */
1505 unsigned long long endofpart;
034b203a 1506
31208db9 1507 if (get_last_partition_end(fd, &endofpart) > 0) {
034b203a
TM
1508 /* There appears to be a partition table here */
1509 if (freesize == 0) {
1510 /* partitions will not be visible in new device */
e7b84f9d
N
1511 pr_err("partition table exists on %s but will be lost or\n"
1512 " meaningless after creating array\n",
1513 dname);
034b203a
TM
1514 return 1;
1515 } else if (endofpart > freesize) {
1516 /* last partition overlaps metadata */
e7b84f9d
N
1517 pr_err("metadata will over-write last partition on %s.\n",
1518 dname);
034b203a 1519 return 1;
53ed6ac3
KW
1520 } else if (size && endofpart > size) {
1521 /* partitions will be truncated in new device */
e7b84f9d
N
1522 pr_err("array size is too small to cover all partitions on %s.\n",
1523 dname);
53ed6ac3 1524 return 1;
034b203a
TM
1525 }
1526 }
1527 return 0;
1528}
1529
a322f70c
DW
1530int open_container(int fd)
1531{
1532 /* 'fd' is a block device. Find out if it is in use
1533 * by a container, and return an open fd on that container.
1534 */
1535 char path[256];
1536 char *e;
1537 DIR *dir;
1538 struct dirent *de;
1539 int dfd, n;
1540 char buf[200];
1541 int major, minor;
1542 struct stat st;
1543
1544 if (fstat(fd, &st) != 0)
1545 return -1;
1546 sprintf(path, "/sys/dev/block/%d:%d/holders",
1547 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1548 e = path + strlen(path);
1549
1550 dir = opendir(path);
1551 if (!dir)
1552 return -1;
1553 while ((de = readdir(dir))) {
1554 if (de->d_ino == 0)
1555 continue;
1556 if (de->d_name[0] == '.')
1557 continue;
a994592d
N
1558 /* Need to make sure it is a container and not a volume */
1559 sprintf(e, "/%s/md/metadata_version", de->d_name);
1560 dfd = open(path, O_RDONLY);
1561 if (dfd < 0)
1562 continue;
1563 n = read(dfd, buf, sizeof(buf));
1564 close(dfd);
1565 if (n <= 0 || (unsigned)n >= sizeof(buf))
1566 continue;
1567 buf[n] = 0;
1568 if (strncmp(buf, "external", 8) != 0 ||
1569 n < 10 ||
1570 buf[9] == '/')
1571 continue;
a322f70c
DW
1572 sprintf(e, "/%s/dev", de->d_name);
1573 dfd = open(path, O_RDONLY);
1574 if (dfd < 0)
1575 continue;
1576 n = read(dfd, buf, sizeof(buf));
1577 close(dfd);
f21e18ca 1578 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1579 continue;
1580 buf[n] = 0;
1581 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1582 continue;
1583 sprintf(buf, "%d:%d", major, minor);
1584 dfd = dev_open(buf, O_RDONLY);
1585 if (dfd >= 0) {
1586 closedir(dir);
1587 return dfd;
1588 }
1589 }
355726fa 1590 closedir(dir);
a322f70c
DW
1591 return -1;
1592}
1593
33414a01
DW
1594struct superswitch *version_to_superswitch(char *vers)
1595{
1596 int i;
1597
1598 for (i = 0; superlist[i]; i++) {
1599 struct superswitch *ss = superlist[i];
1600
1601 if (strcmp(vers, ss->name) == 0)
1602 return ss;
1603 }
1604
1605 return NULL;
1606}
1607
4dd2df09
N
1608int metadata_container_matches(char *metadata, char *devnm)
1609{
1610 /* Check if 'devnm' is the container named in 'metadata'
1611 * which is
1612 * /containername/componentname or
1613 * -containername/componentname
1614 */
1615 int l;
1616 if (*metadata != '/' && *metadata != '-')
1617 return 0;
1618 l = strlen(devnm);
1619 if (strncmp(metadata+1, devnm, l) != 0)
1620 return 0;
1621 if (metadata[l+1] != '/')
1622 return 0;
1623 return 1;
1624}
1625
1626int metadata_subdev_matches(char *metadata, char *devnm)
1627{
1628 /* Check if 'devnm' is the subdev named in 'metadata'
1629 * which is
1630 * /containername/subdev or
1631 * -containername/subdev
1632 */
1633 char *sl;
1634 if (*metadata != '/' && *metadata != '-')
1635 return 0;
1636 sl = strchr(metadata+1, '/');
1637 if (!sl)
1638 return 0;
1639 if (strcmp(sl+1, devnm) == 0)
1640 return 1;
1641 return 0;
1642}
1643
33414a01
DW
1644int is_container_member(struct mdstat_ent *mdstat, char *container)
1645{
1646 if (mdstat->metadata_version == NULL ||
1647 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1648 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1649 return 0;
1650
1651 return 1;
1652}
1653
1654int is_subarray_active(char *subarray, char *container)
1655{
1656 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1657 struct mdstat_ent *ent;
1658
e5408a32
DW
1659 for (ent = mdstat; ent; ent = ent->next)
1660 if (is_container_member(ent, container))
e5e5d7ce 1661 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1662 break;
33414a01
DW
1663
1664 free_mdstat(mdstat);
1665
1666 return ent != NULL;
1667}
1668
1669/* open_subarray - opens a subarray in a container
1670 * @dev: container device name
feab51f8 1671 * @st: empty supertype
33414a01
DW
1672 * @quiet: block reporting errors flag
1673 *
1674 * On success returns an fd to a container and fills in *st
1675 */
feab51f8 1676int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1677{
1678 struct mdinfo *mdi;
a951a4f7 1679 struct mdinfo *info;
33414a01 1680 int fd, err = 1;
4dd2df09 1681 char *_devnm;
33414a01
DW
1682
1683 fd = open(dev, O_RDWR|O_EXCL);
1684 if (fd < 0) {
1685 if (!quiet)
e7b84f9d 1686 pr_err("Couldn't open %s, aborting\n",
33414a01 1687 dev);
b990032d 1688 return -1;
33414a01
DW
1689 }
1690
4dd2df09
N
1691 _devnm = fd2devnm(fd);
1692 if (_devnm == NULL) {
33414a01 1693 if (!quiet)
e7b84f9d
N
1694 pr_err("Failed to determine device number for %s\n",
1695 dev);
33414a01
DW
1696 goto close_fd;
1697 }
4dd2df09 1698 strcpy(st->devnm, _devnm);
33414a01 1699
4dd2df09 1700 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1701 if (!mdi) {
1702 if (!quiet)
e7b84f9d 1703 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1704 dev);
1705 goto close_fd;
1706 }
1707
1708 if (mdi->array.level != UnSet) {
1709 if (!quiet)
e7b84f9d 1710 pr_err("%s is not a container\n", dev);
33414a01
DW
1711 goto free_sysfs;
1712 }
1713
1714 st->ss = version_to_superswitch(mdi->text_version);
1715 if (!st->ss) {
1716 if (!quiet)
e7b84f9d
N
1717 pr_err("Operation not supported for %s metadata\n",
1718 mdi->text_version);
33414a01
DW
1719 goto free_sysfs;
1720 }
1721
4dd2df09 1722 if (st->devnm[0] == 0) {
33414a01 1723 if (!quiet)
e7b84f9d 1724 pr_err("Failed to allocate device name\n");
33414a01
DW
1725 goto free_sysfs;
1726 }
1727
db20d413 1728 if (!st->ss->load_container) {
33414a01 1729 if (!quiet)
e7b84f9d 1730 pr_err("%s is not a container\n", dev);
4dd2df09 1731 goto free_sysfs;
33414a01
DW
1732 }
1733
db20d413 1734 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1735 if (!quiet)
e7b84f9d 1736 pr_err("Failed to load metadata for %s\n",
db20d413 1737 dev);
4dd2df09 1738 goto free_sysfs;
33414a01
DW
1739 }
1740
a951a4f7
N
1741 info = st->ss->container_content(st, subarray);
1742 if (!info) {
1743 if (!quiet)
e7b84f9d 1744 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1745 subarray, dev);
1746 goto free_super;
1747 }
1748 free(info);
1749
33414a01
DW
1750 err = 0;
1751
1752 free_super:
1753 if (err)
1754 st->ss->free_super(st);
33414a01
DW
1755 free_sysfs:
1756 sysfs_free(mdi);
1757 close_fd:
1758 if (err)
1759 close(fd);
1760
1761 if (err)
1762 return -1;
1763 else
1764 return fd;
1765}
1766
7801ac20
N
1767int add_disk(int mdfd, struct supertype *st,
1768 struct mdinfo *sra, struct mdinfo *info)
1769{
1770 /* Add a device to an array, in one of 2 ways. */
1771 int rv;
1772#ifndef MDASSEMBLE
1773 if (st->ss->external) {
d23534e4
DW
1774 if (info->disk.state & (1<<MD_DISK_SYNC))
1775 info->recovery_start = MaxSector;
1776 else
1777 info->recovery_start = 0;
2904b26f 1778 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1779 if (! rv) {
1780 struct mdinfo *sd2;
f35f2525
N
1781 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1782 if (sd2 == info)
1783 break;
1784 if (sd2 == NULL) {
503975b9 1785 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1786 *sd2 = *info;
1787 sd2->next = sra->devs;
1788 sra->devs = sd2;
1789 }
7801ac20
N
1790 }
1791 } else
1792#endif
1793 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1794 return rv;
1795}
1796
de6ae750
N
1797int remove_disk(int mdfd, struct supertype *st,
1798 struct mdinfo *sra, struct mdinfo *info)
1799{
1800 int rv;
1801 /* Remove the disk given by 'info' from the array */
1802#ifndef MDASSEMBLE
1803 if (st->ss->external)
1804 rv = sysfs_set_str(sra, info, "slot", "none");
1805 else
1806#endif
1807 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1808 info->disk.minor));
1809 return rv;
1810}
1811
1ab9ed2a 1812int hot_remove_disk(int mdfd, unsigned long dev, int force)
2dd271fe 1813{
1ab9ed2a 1814 int cnt = force ? 500 : 5;
2dd271fe
N
1815 int ret;
1816
1817 /* HOT_REMOVE_DISK can fail with EBUSY if there are
1818 * outstanding IO requests to the device.
1819 * In this case, it can be helpful to wait a little while,
1ab9ed2a 1820 * up to 5 seconds if 'force' is set, or 50 msec if not.
2dd271fe
N
1821 */
1822 while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 &&
1823 errno == EBUSY &&
1824 cnt-- > 0)
1825 usleep(10000);
1826
1827 return ret;
1828}
1829
1ab9ed2a 1830int sys_hot_remove_disk(int statefd, int force)
fdd01569 1831{
1ab9ed2a 1832 int cnt = force ? 500 : 5;
fdd01569
N
1833 int ret;
1834
1835 while ((ret = write(statefd, "remove", 6)) == -1 &&
1836 errno == EBUSY &&
1837 cnt-- > 0)
1838 usleep(10000);
1839 return ret == 6 ? 0 : -1;
1840}
1841
f35f2525
N
1842int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1843{
1844 /* Initialise kernel's knowledge of array.
1845 * This varies between externally managed arrays
1846 * and older kernels
1847 */
1848 int vers = md_get_version(mdfd);
1849 int rv;
1850
1851#ifndef MDASSEMBLE
1852 if (st->ss->external)
1853 rv = sysfs_set_array(info, vers);
1854 else
1855#endif
1856 if ((vers % 100) >= 1) { /* can use different versions */
1857 mdu_array_info_t inf;
1858 memset(&inf, 0, sizeof(inf));
1859 inf.major_version = info->array.major_version;
1860 inf.minor_version = info->array.minor_version;
1861 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1862 } else
1863 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1864 return rv;
1865}
1866
1e5c6983
DW
1867unsigned long long min_recovery_start(struct mdinfo *array)
1868{
1869 /* find the minimum recovery_start in an array for metadata
1870 * formats that only record per-array recovery progress instead
1871 * of per-device
1872 */
1873 unsigned long long recovery_start = MaxSector;
1874 struct mdinfo *d;
1875
1876 for (d = array->devs; d; d = d->next)
1877 recovery_start = min(recovery_start, d->recovery_start);
1878
1879 return recovery_start;
1880}
1881
4dd2df09 1882int mdmon_pid(char *devnm)
a931db9e
NB
1883{
1884 char path[100];
1885 char pid[10];
1886 int fd;
1887 int n;
10013317 1888
4dd2df09 1889 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1890
24f6f99b 1891 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1892
1893 if (fd < 0)
cf556303 1894 return -1;
a931db9e
NB
1895 n = read(fd, pid, 9);
1896 close(fd);
1897 if (n <= 0)
cf556303 1898 return -1;
24f6f99b 1899 return atoi(pid);
a931db9e
NB
1900}
1901
4dd2df09 1902int mdmon_running(char *devnm)
a931db9e 1903{
4dd2df09 1904 int pid = mdmon_pid(devnm);
24f6f99b 1905 if (pid <= 0)
a931db9e 1906 return 0;
24f6f99b 1907 if (kill(pid, 0) == 0)
a931db9e
NB
1908 return 1;
1909 return 0;
1910}
1911
4dd2df09 1912int start_mdmon(char *devnm)
8850ee3e 1913{
1913c325 1914 int i, skipped;
44d2e365 1915 int len;
ca3b6696 1916 pid_t pid;
9fe32043 1917 int status;
44d2e365
N
1918 char pathbuf[1024];
1919 char *paths[4] = {
1920 pathbuf,
85945e19 1921 BINDIR "/mdmon",
2f1bcf43 1922 "./mdmon",
44d2e365
N
1923 NULL
1924 };
8850ee3e 1925
40ebbb9c 1926 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1927 return 0;
1928
9cf014ec 1929 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1930 if (len > 0) {
1931 char *sl;
1932 pathbuf[len] = 0;
1933 sl = strrchr(pathbuf, '/');
1934 if (sl)
1935 sl++;
1936 else
1937 sl = pathbuf;
1938 strcpy(sl, "mdmon");
1939 } else
1940 pathbuf[0] = '\0';
1941
0f7bdf89 1942 /* First try to run systemctl */
9540cc24
N
1943 if (!check_env("MDADM_NO_SYSTEMCTL"))
1944 switch(fork()) {
1945 case 0:
1946 /* FIXME yuk. CLOSE_EXEC?? */
1947 skipped = 0;
1948 for (i = 3; skipped < 20; i++)
1949 if (close(i) < 0)
1950 skipped++;
1951 else
1952 skipped = 0;
1953
1954 /* Don't want to see error messages from
1955 * systemctl. If the service doesn't exist,
1956 * we start mdmon ourselves.
1957 */
1958 close(2);
1959 open("/dev/null", O_WRONLY);
1960 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1961 devnm);
1962 status = execl("/usr/bin/systemctl", "systemctl",
1963 "start",
1964 pathbuf, NULL);
1965 status = execl("/bin/systemctl", "systemctl", "start",
1966 pathbuf, NULL);
1967 exit(1);
7a862a02 1968 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1969 return -1;
1970 default: /* parent - good */
1971 pid = wait(&status);
1972 if (pid >= 0 && status == 0)
1973 return 0;
1974 }
0f7bdf89
JS
1975
1976 /* That failed, try running mdmon directly */
8850ee3e
N
1977 switch(fork()) {
1978 case 0:
1979 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1980 skipped = 0;
ca3b6696 1981 for (i = 3; skipped < 20; i++)
1913c325
N
1982 if (close(i) < 0)
1983 skipped++;
1984 else
1985 skipped = 0;
1986
ca3b6696 1987 for (i = 0; paths[i]; i++)
a0963a86 1988 if (paths[i][0]) {
2f1bcf43 1989 execl(paths[i], paths[i],
4dd2df09 1990 devnm, NULL);
a0963a86 1991 }
8850ee3e 1992 exit(1);
7a862a02 1993 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 1994 return -1;
9fe32043
N
1995 default: /* parent - good */
1996 pid = wait(&status);
15c10423 1997 if (pid < 0 || status != 0) {
7a862a02 1998 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 1999 return -1;
15c10423 2000 }
8850ee3e
N
2001 }
2002 return 0;
2003}
2004
148acb7b
DW
2005__u32 random32(void)
2006{
2007 __u32 rv;
2008 int rfd = open("/dev/urandom", O_RDONLY);
2009 if (rfd < 0 || read(rfd, &rv, 4) != 4)
2010 rv = random();
2011 if (rfd >= 0)
2012 close(rfd);
2013 return rv;
2014}
2015
c5f71c24
JS
2016void random_uuid(__u8 *buf)
2017{
2018 int fd, i, len;
2019 __u32 r[4];
2020
2021 fd = open("/dev/urandom", O_RDONLY);
2022 if (fd < 0)
2023 goto use_random;
2024 len = read(fd, buf, 16);
2025 close(fd);
2026 if (len != 16)
2027 goto use_random;
2028
2029 return;
2030
2031use_random:
2032 for (i = 0; i < 4; i++)
2033 r[i] = random();
2034 memcpy(buf, r, 16);
2035}
2036
0e600426 2037#ifndef MDASSEMBLE
edd8d13c
NB
2038int flush_metadata_updates(struct supertype *st)
2039{
2040 int sfd;
2041 if (!st->updates) {
2042 st->update_tail = NULL;
2043 return -1;
2044 }
2045
4dd2df09 2046 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
2047 if (sfd < 0)
2048 return -1;
2049
2050 while (st->updates) {
2051 struct metadata_update *mu = st->updates;
2052 st->updates = mu->next;
2053
2054 send_message(sfd, mu, 0);
2055 wait_reply(sfd, 0);
2056 free(mu->buf);
2057 free(mu);
2058 }
2059 ack(sfd, 0);
2060 wait_reply(sfd, 0);
2061 close(sfd);
2062 st->update_tail = NULL;
2063 return 0;
2064}
2065
2066void append_metadata_update(struct supertype *st, void *buf, int len)
2067{
2068
503975b9 2069 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
2070
2071 mu->buf = buf;
2072 mu->len = len;
2073 mu->space = NULL;
cb23f1f4 2074 mu->space_list = NULL;
edd8d13c
NB
2075 mu->next = NULL;
2076 *st->update_tail = mu;
2077 st->update_tail = &mu->next;
2078}
0e600426 2079#endif /* MDASSEMBLE */
a931db9e 2080
fe6729fa
NB
2081#ifdef __TINYC__
2082/* tinyc doesn't optimize this check in ioctl.h out ... */
2083unsigned int __invalid_size_argument_for_IOC = 0;
2084#endif
2085
6d11ec6f
AK
2086int experimental(void)
2087{
2088 if (check_env("MDADM_EXPERIMENTAL"))
2089 return 1;
2090 else {
7a862a02 2091 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
2092 return 0;
2093 }
2094}
2095
326727d9
AC
2096/* Pick all spares matching given criteria from a container
2097 * if min_size == 0 do not check size
2098 * if domlist == NULL do not check domains
2099 * if spare_group given add it to domains of each spare
2100 * metadata allows to test domains using metadata of destination array */
2101struct mdinfo *container_choose_spares(struct supertype *st,
2102 unsigned long long min_size,
2103 struct domainlist *domlist,
2104 char *spare_group,
2105 const char *metadata, int get_one)
2106{
2107 struct mdinfo *d, **dp, *disks = NULL;
2108
2109 /* get list of all disks in container */
2110 if (st->ss->getinfo_super_disks)
2111 disks = st->ss->getinfo_super_disks(st);
2112
2113 if (!disks)
2114 return disks;
2115 /* find spare devices on the list */
2116 dp = &disks->devs;
2117 disks->array.spare_disks = 0;
2118 while (*dp) {
2119 int found = 0;
2120 d = *dp;
2121 if (d->disk.state == 0) {
2122 /* check if size is acceptable */
2123 unsigned long long dev_size;
2124 dev_t dev = makedev(d->disk.major,d->disk.minor);
2125
2126 if (!min_size ||
2127 (dev_size_from_id(dev, &dev_size) &&
2128 dev_size >= min_size))
2129 found = 1;
2130 /* check if domain matches */
2131 if (found && domlist) {
4dd2df09 2132 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2133 if (spare_group)
2134 pol_add(&pol, pol_domain,
2135 spare_group, NULL);
e5508b36 2136 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2137 found = 0;
2138 dev_policy_free(pol);
2139 }
2140 }
2141 if (found) {
2142 dp = &d->next;
2143 disks->array.spare_disks++;
2144 if (get_one) {
2145 sysfs_free(*dp);
2146 d->next = NULL;
2147 }
2148 } else {
2149 *dp = d->next;
2150 d->next = NULL;
2151 sysfs_free(d);
2152 }
2153 }
2154 return disks;
2155}
9eafa1de
MN
2156
2157/* Checks if paths point to the same device
2158 * Returns 0 if they do.
2159 * Returns 1 if they don't.
2160 * Returns -1 if something went wrong,
2161 * e.g. paths are empty or the files
2162 * they point to don't exist */
2163int compare_paths (char* path1, char* path2)
2164{
2165 struct stat st1,st2;
2166
2167 if (path1 == NULL || path2 == NULL)
2168 return -1;
2169 if (stat(path1,&st1) != 0)
2170 return -1;
2171 if (stat(path2,&st2) != 0)
2172 return -1;
2173 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2174 return 0;
2175 return 1;
2176}
a7dec3fd
N
2177
2178/* Make sure we can open as many devices as needed */
2179void enable_fds(int devices)
2180{
2181 unsigned int fds = 20 + devices;
2182 struct rlimit lim;
2183 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
2184 || lim.rlim_cur >= fds)
2185 return;
2186 if (lim.rlim_max < fds)
2187 lim.rlim_max = fds;
2188 lim.rlim_cur = fds;
2189 setrlimit(RLIMIT_NOFILE, &lim);
2190}
a9c15847
N
2191
2192int in_initrd(void)
2193{
2194 /* This is based on similar function in systemd. */
2195 struct statfs s;
76d0f188
JS
2196 /* statfs.f_type is signed long on s390x and MIPS, causing all
2197 sorts of sign extension problems with RAMFS_MAGIC being
2198 defined as 0x858458f6 */
a9c15847 2199 return statfs("/", &s) >= 0 &&
7ac5d47e 2200 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2201 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2202 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2203}
8832342d
N
2204
2205void reopen_mddev(int mdfd)
2206{
2207 /* Re-open without any O_EXCL, but keep
2208 * the same fd
2209 */
2210 char *devnm;
2211 int fd;
2212 devnm = fd2devnm(mdfd);
2213 close(mdfd);
2214 fd = open_dev(devnm);
2215 if (fd >= 0 && fd != mdfd)
2216 dup2(fd, mdfd);
2217}
7716570e 2218
7071320a 2219#ifndef MDASSEMBLE
e80357f8
GJ
2220static struct cmap_hooks *cmap_hooks = NULL;
2221static int is_cmap_hooks_ready = 0;
7716570e 2222
e80357f8
GJ
2223void set_cmap_hooks(void)
2224{
2225 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2226 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2227 if (!cmap_hooks->cmap_handle)
2228 return;
7716570e 2229
e80357f8
GJ
2230 cmap_hooks->initialize = dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2231 cmap_hooks->get_string = dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
2232 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2233
e80357f8
GJ
2234 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2235 !cmap_hooks->finalize)
2236 dlclose(cmap_hooks->cmap_handle);
2237 else
2238 is_cmap_hooks_ready = 1;
2239}
7716570e 2240
e80357f8
GJ
2241int get_cluster_name(char **cluster_name)
2242{
2243 int rv = -1;
2244 cmap_handle_t handle;
7716570e 2245
e80357f8
GJ
2246 if (!is_cmap_hooks_ready)
2247 return rv;
7716570e 2248
e80357f8 2249 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2250 if (rv != CS_OK)
2251 goto out;
2252
e80357f8 2253 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2254 if (rv != CS_OK) {
2255 free(*cluster_name);
2256 rv = -1;
2257 goto name_err;
2258 }
2259
2260 rv = 0;
2261name_err:
e80357f8 2262 cmap_hooks->finalize(handle);
7716570e 2263out:
7716570e
GJ
2264 return rv;
2265}
d15a1f72
GJ
2266
2267void set_dlm_hooks(void)
2268{
2269 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2270 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2271 if (!dlm_hooks->dlm_handle)
2272 return;
2273
2274 dlm_hooks->create_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2275 dlm_hooks->release_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
2276 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2277 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2278 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2279 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2280
2281 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2282 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2283 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2284 dlclose(dlm_hooks->dlm_handle);
2285 else
2286 is_dlm_hooks_ready = 1;
2287}
e80357f8
GJ
2288
2289void set_hooks(void)
2290{
2291 set_dlm_hooks();
2292 set_cmap_hooks();
2293}
7d55dca2 2294#endif