md-cluster: use %-64s to print cluster_name
[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>
98c6faba 34#include <ctype.h>
a322f70c 35#include <dirent.h>
a931db9e 36#include <signal.h>
7716570e
GJ
37#include <dlfcn.h>
38#include <stdint.h>
39#ifdef NO_COROSYNC
40 typedef uint64_t cmap_handle_t;
41 #define CS_OK 1
42#else
43 #include <corosync/cmap.h>
44#endif
45
0a816ef9
NB
46
47/*
48 * following taken from linux/blkpg.h because they aren't
49 * anywhere else and it isn't safe to #include linux/ * stuff.
50 */
51
52#define BLKPG _IO(0x12,105)
53
54/* The argument structure */
55struct blkpg_ioctl_arg {
5d500228
N
56 int op;
57 int flags;
58 int datalen;
59 void *data;
0a816ef9
NB
60};
61
62/* The subfunctions (for the op field) */
63#define BLKPG_ADD_PARTITION 1
64#define BLKPG_DEL_PARTITION 2
65
66/* Sizes of name fields. Unused at present. */
67#define BLKPG_DEVNAMELTH 64
68#define BLKPG_VOLNAMELTH 64
69
70/* The data structure for ADD_PARTITION and DEL_PARTITION */
71struct blkpg_partition {
72 long long start; /* starting offset in bytes */
73 long long length; /* length in bytes */
74 int pno; /* partition number */
75 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
76 to be used in kernel messages */
77 char volname[BLKPG_VOLNAMELTH]; /* volume label */
78};
64c4757e 79
0f22b998 80#include "part.h"
056b331e
N
81
82/* Force a compilation error if condition is true */
83#define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
84
85/* Force a compilation error if condition is true, but also produce a
86 result (of value 0 and type size_t), so the expression can be used
87 e.g. in a structure initializer (or where-ever else comma expressions
88 aren't permitted). */
89#define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
90
64c4757e
NB
91/*
92 * Parse a 128 bit uuid in 4 integers
93 * format is 32 hexx nibbles with options :.<space> separator
94 * If not exactly 32 hex digits are found, return 0
95 * else return 1
96 */
97int parse_uuid(char *str, int uuid[4])
98{
aba69144
NB
99 int hit = 0; /* number of Hex digIT */
100 int i;
101 char c;
ca3b6696
N
102 for (i = 0; i < 4; i++)
103 uuid[i] = 0;
aba69144 104
ca3b6696 105 while ((c = *str++) != 0) {
aba69144 106 int n;
ca3b6696 107 if (c >= '0' && c <= '9')
aba69144 108 n = c-'0';
ca3b6696 109 else if (c >= 'a' && c <= 'f')
aba69144 110 n = 10 + c - 'a';
ca3b6696 111 else if (c >= 'A' && c <= 'F')
aba69144
NB
112 n = 10 + c - 'A';
113 else if (strchr(":. -", c))
114 continue;
115 else return 0;
116
117 if (hit<32) {
118 uuid[hit/8] <<= 4;
119 uuid[hit/8] += n;
120 }
121 hit++;
82b27616 122 }
aba69144
NB
123 if (hit == 32)
124 return 1;
125 return 0;
64c4757e
NB
126}
127
64c4757e
NB
128/*
129 * Get the md version number.
130 * We use the RAID_VERSION ioctl if it is supported
131 * If not, but we have a block device with major '9', we assume
132 * 0.36.0
133 *
134 * Return version number as 24 but number - assume version parts
135 * always < 255
136 */
137
138int md_get_version(int fd)
139{
5d500228
N
140 struct stat stb;
141 mdu_version_t vers;
142
143 if (fstat(fd, &stb)<0)
144 return -1;
145 if ((S_IFMT&stb.st_mode) != S_IFBLK)
146 return -1;
147
148 if (ioctl(fd, RAID_VERSION, &vers) == 0)
149 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
150 if (errno == EACCES)
151 return -1;
152 if (major(stb.st_rdev) == MD_MAJOR)
153 return (3600);
64c4757e 154 return -1;
64c4757e
NB
155}
156
64c4757e
NB
157int get_linux_version()
158{
159 struct utsname name;
98c6faba 160 char *cp;
f161d047 161 int a = 0, b = 0,c = 0;
64c4757e
NB
162 if (uname(&name) <0)
163 return -1;
164
98c6faba
NB
165 cp = name.release;
166 a = strtoul(cp, &cp, 10);
f161d047
N
167 if (*cp == '.')
168 b = strtoul(cp+1, &cp, 10);
169 if (*cp == '.')
170 c = strtoul(cp+1, &cp, 10);
98c6faba 171
682c7051 172 return (a*1000000)+(b*1000)+c;
64c4757e
NB
173}
174
71204a50 175#ifndef MDASSEMBLE
bc77ed53
DW
176int mdadm_version(char *version)
177{
178 int a, b, c;
179 char *cp;
180
181 if (!version)
182 version = Version;
183
184 cp = strchr(version, '-');
185 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
186 return -1;
187 cp += 3;
188 a = strtoul(cp, &cp, 10);
189 if (*cp != '.')
190 return -1;
191 b = strtoul(cp+1, &cp, 10);
192 if (*cp == '.')
193 c = strtoul(cp+1, &cp, 10);
194 else
195 c = 0;
196 if (*cp != ' ' && *cp != '-')
197 return -1;
198 return (a*1000000)+(b*1000)+c;
199}
200
51d4261c 201unsigned long long parse_size(char *size)
84e11361
N
202{
203 /* parse 'size' which should be a number optionally
204 * followed by 'K', 'M', or 'G'.
205 * Without a suffix, K is assumed.
206 * Number returned is in sectors (half-K)
822e393a 207 * INVALID_SECTORS returned on error.
84e11361
N
208 */
209 char *c;
210 long long s = strtoll(size, &c, 10);
211 if (s > 0) {
212 switch (*c) {
213 case 'K':
214 c++;
215 default:
216 s *= 2;
217 break;
218 case 'M':
219 c++;
220 s *= 1024 * 2;
221 break;
222 case 'G':
223 c++;
224 s *= 1024 * 1024 * 2;
225 break;
72ca9bcf
N
226 case 's': /* sectors */
227 c++;
228 break;
84e11361 229 }
51d4261c 230 } else
822e393a 231 s = INVALID_SECTORS;
84e11361 232 if (*c)
822e393a 233 s = INVALID_SECTORS;
84e11361
N
234 return s;
235}
236
4a06e2c2
N
237int parse_layout_10(char *layout)
238{
239 int copies, rv;
240 char *cp;
241 /* Parse the layout string for raid10 */
242 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
243 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
244 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
245 copies > 200 ||
246 *cp)
247 return -1;
248 if (layout[0] == 'n')
249 rv = 256 + copies;
250 else if (layout[0] == 'o')
251 rv = 0x10000 + (copies<<8) + 1;
252 else
253 rv = 1 + (copies<<8);
254 return rv;
255}
256
257int parse_layout_faulty(char *layout)
258{
259 /* Parse the layout string for 'faulty' */
260 int ln = strcspn(layout, "0123456789");
503975b9 261 char *m = xstrdup(layout);
4a06e2c2
N
262 int mode;
263 m[ln] = 0;
264 mode = map_name(faultylayout, m);
265 if (mode == UnSet)
266 return -1;
267
268 return mode | (atoi(layout+ln)<< ModeShift);
269}
79868890
N
270
271long parse_num(char *num)
272{
273 /* Either return a valid number, or -1 */
274 char *c;
275 long rv = strtol(num, &c, 10);
276 if (rv < 0 || *c || !num[0])
277 return -1;
278 else
279 return rv;
280}
eb3929a4 281#endif
4a06e2c2 282
4de90913
GJ
283int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
284{
285 char *dev;
286 *slot = strtoul(input, &dev, 10);
287 if (dev == input || dev[0] != ':')
288 return -1;
289 *devname = dev+1;
290 return 0;
291}
292
0430ed48
NB
293void remove_partitions(int fd)
294{
295 /* remove partitions from this block devices.
296 * This is used for components added to an array
297 */
298#ifdef BLKPG_DEL_PARTITION
299 struct blkpg_ioctl_arg a;
300 struct blkpg_partition p;
301
302 a.op = BLKPG_DEL_PARTITION;
303 a.data = (void*)&p;
304 a.datalen = sizeof(p);
305 a.flags = 0;
306 memset(a.data, 0, a.datalen);
ca3b6696 307 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
308 ioctl(fd, BLKPG, &a);
309#endif
310}
311
691c6ee1
N
312int test_partition(int fd)
313{
314 /* Check if fd is a whole-disk or a partition.
315 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
316 * will return ENXIO on an invalid partition number.
317 */
318 struct blkpg_ioctl_arg a;
319 struct blkpg_partition p;
320 a.op = BLKPG_DEL_PARTITION;
321 a.data = (void*)&p;
322 a.datalen = sizeof(p);
323 a.flags = 0;
324 memset(a.data, 0, a.datalen);
325 p.pno = 1<<30;
326 if (ioctl(fd, BLKPG, &a) == 0)
327 /* Very unlikely, but not a partition */
328 return 0;
357ac106 329 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
330 /* not a partition */
331 return 0;
332
333 return 1;
334}
335
bfd76b93
CA
336int test_partition_from_id(dev_t id)
337{
338 char buf[20];
339 int fd, rv;
340
341 sprintf(buf, "%d:%d", major(id), minor(id));
342 fd = dev_open(buf, O_RDONLY);
343 if (fd < 0)
344 return -1;
345 rv = test_partition(fd);
346 close(fd);
347 return rv;
348}
691c6ee1 349
de5a472e 350int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 351{
265e0f17 352 int copies, first;
de5a472e
N
353 int i;
354 int avail_disks = 0;
355
356 for (i = 0; i < raid_disks; i++)
357 avail_disks += !!avail[i];
358
64c4757e 359 switch (level) {
265e0f17
NB
360 case 10:
361 /* This is the tricky one - we need to check
362 * which actual disks are present.
363 */
702b557b 364 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 365 first = 0;
265e0f17
NB
366 do {
367 /* there must be one of the 'copies' form 'first' */
368 int n = copies;
ca3b6696 369 int cnt = 0;
b3ec716d 370 int this = first;
265e0f17 371 while (n--) {
b3ec716d 372 if (avail[this])
265e0f17 373 cnt++;
b3ec716d 374 this = (this+1) % raid_disks;
265e0f17
NB
375 }
376 if (cnt == 0)
377 return 0;
b3ec716d 378 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
379 } while (first != 0);
380 return 1;
e5329c37 381
df0d4ea0 382 case LEVEL_MULTIPATH:
e0d19036 383 return avail_disks>= 1;
df0d4ea0 384 case LEVEL_LINEAR:
64c4757e
NB
385 case 0:
386 return avail_disks == raid_disks;
387 case 1:
388 return avail_disks >= 1;
389 case 4:
330d6900
N
390 if (avail_disks == raid_disks - 1 &&
391 !avail[raid_disks - 1])
392 /* If just the parity device is missing, then we
393 * have enough, even if not clean
394 */
395 return 1;
396 /* FALL THROUGH */
64c4757e 397 case 5:
583315d9
NB
398 if (clean)
399 return avail_disks >= raid_disks-1;
400 else
401 return avail_disks >= raid_disks;
98c6faba 402 case 6:
583315d9
NB
403 if (clean)
404 return avail_disks >= raid_disks-2;
405 else
406 return avail_disks >= raid_disks;
64c4757e
NB
407 default:
408 return 0;
409 }
410}
411
8453e704
N
412int enough_fd(int fd)
413{
414 struct mdu_array_info_s array;
415 struct mdu_disk_info_s disk;
aabe020d 416 int i, rv;
8453e704
N
417 char *avail;
418
419 if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 ||
420 array.raid_disks <= 0)
421 return 0;
503975b9 422 avail = xcalloc(array.raid_disks, 1);
ca3b6696 423 for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704
N
424 disk.number = i;
425 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
426 continue;
9e6d9291
N
427 if (disk.major == 0 && disk.minor == 0)
428 continue;
429 array.nr_disks--;
430
8453e704
N
431 if (! (disk.state & (1<<MD_DISK_SYNC)))
432 continue;
433 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
434 continue;
8453e704
N
435 avail[disk.raid_disk] = 1;
436 }
437 /* This is used on an active array, so assume it is clean */
aabe020d 438 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 439 1, avail);
aabe020d
JS
440 free(avail);
441 return rv;
8453e704
N
442}
443
22e263f6
AC
444const int uuid_zero[4] = { 0, 0, 0, 0 };
445
f277ce36 446int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 447{
f277ce36
NB
448 if (swapuuid) {
449 /* parse uuids are hostendian.
450 * uuid's from some superblocks are big-ending
aba69144 451 * if there is a difference, we need to swap..
f277ce36
NB
452 */
453 unsigned char *ac = (unsigned char *)a;
454 unsigned char *bc = (unsigned char *)b;
455 int i;
ca3b6696 456 for (i = 0; i < 16; i += 4) {
f277ce36
NB
457 if (ac[i+0] != bc[i+3] ||
458 ac[i+1] != bc[i+2] ||
459 ac[i+2] != bc[i+1] ||
460 ac[i+3] != bc[i+0])
461 return 0;
462 }
463 return 1;
464 } else {
465 if (a[0]==b[0] &&
466 a[1]==b[1] &&
467 a[2]==b[2] &&
468 a[3]==b[3])
469 return 1;
470 return 0;
471 }
64c4757e 472}
ca3b6696 473
350f29f9
NB
474void copy_uuid(void *a, int b[4], int swapuuid)
475{
476 if (swapuuid) {
477 /* parse uuids are hostendian.
478 * uuid's from some superblocks are big-ending
479 * if there is a difference, we need to swap..
480 */
481 unsigned char *ac = (unsigned char *)a;
482 unsigned char *bc = (unsigned char *)b;
483 int i;
ca3b6696 484 for (i = 0; i < 16; i += 4) {
350f29f9
NB
485 ac[i+0] = bc[i+3];
486 ac[i+1] = bc[i+2];
487 ac[i+2] = bc[i+1];
488 ac[i+3] = bc[i+0];
489 }
490 } else
491 memcpy(a, b, 16);
492}
64c4757e 493
aae5a112 494char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 495{
9968e376 496 int i, j;
d7288ddc
N
497 char uuid[16];
498 char *c = buf;
499 strcpy(c, "UUID-");
500 c += strlen(c);
aae5a112 501 copy_uuid(uuid, id, swap);
9968e376 502 for (i = 0; i < 4; i++) {
9968e376 503 if (i)
ff54de6e 504 *c++ = sep;
9968e376
DW
505 for (j = 3; j >= 0; j--) {
506 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
507 c+= 2;
508 }
d7288ddc
N
509 }
510 return buf;
aae5a112
DW
511
512}
513
514char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
515{
86983cce
N
516 // dirty hack to work around an issue with super1 superblocks...
517 // super1 superblocks need swapuuid set in order for assembly to
518 // work, but can't have it set if we want this printout to match
519 // all the other uuid printouts in super1.c, so we force swapuuid
520 // to 1 to make our printout match the rest of super1
521 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
522}
523
435d4ebb 524#ifndef MDASSEMBLE
682c7051
NB
525int check_ext2(int fd, char *name)
526{
527 /*
528 * Check for an ext2fs file system.
529 * Superblock is always 1K at 1K offset
530 *
531 * s_magic is le16 at 56 == 0xEF53
532 * report mtime - le32 at 44
533 * blocks - le32 at 4
534 * logblksize - le32 at 24
535 */
536 unsigned char sb[1024];
537 time_t mtime;
2538ba2a
N
538 unsigned long long size;
539 int bsize;
682c7051
NB
540 if (lseek(fd, 1024,0)!= 1024)
541 return 0;
542 if (read(fd, sb, 1024)!= 1024)
543 return 0;
544 if (sb[56] != 0x53 || sb[57] != 0xef)
545 return 0;
546
547 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
548 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
549 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 550 size <<= bsize;
e7b84f9d 551 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 552 name);
2538ba2a 553 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
554 return 1;
555}
556
557int check_reiser(int fd, char *name)
558{
559 /*
560 * superblock is at 64K
561 * size is 1024;
562 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
563 *
564 */
565 unsigned char sb[1024];
2538ba2a 566 unsigned long long size;
682c7051
NB
567 if (lseek(fd, 64*1024, 0) != 64*1024)
568 return 0;
569 if (read(fd, sb, 1024) != 1024)
570 return 0;
ca3b6696
N
571 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
572 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 573 return 0;
e7b84f9d 574 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 575 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 576 cont_err("size = %lluK\n", size*4);
aba69144 577
682c7051
NB
578 return 1;
579}
580
581int check_raid(int fd, char *name)
582{
4b1ac34b 583 struct mdinfo info;
682c7051 584 time_t crtime;
d078d77c 585 char *level;
82d9eba6 586 struct supertype *st = guess_super(fd);
f9ce90ba 587
cb8f6859
N
588 if (!st)
589 return 0;
3da92f27 590 st->ss->load_super(st, fd, name);
82d9eba6 591 /* Looks like a raid array .. */
e7b84f9d 592 pr_err("%s appears to be part of a raid array:\n",
82d9eba6 593 name);
a5d85af7 594 st->ss->getinfo_super(st, &info, NULL);
3da92f27 595 st->ss->free_super(st);
82d9eba6 596 crtime = info.array.ctime;
d078d77c
NB
597 level = map_num(pers, info.array.level);
598 if (!level) level = "-unknown-";
d33f1518
N
599 cont_err("level=%s devices=%d ctime=%s",
600 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 601 return 1;
682c7051
NB
602}
603
682c7051
NB
604int ask(char *mesg)
605{
606 char *add = "";
607 int i;
ca3b6696 608 for (i = 0; i < 5; i++) {
682c7051
NB
609 char buf[100];
610 fprintf(stderr, "%s%s", mesg, add);
611 fflush(stderr);
612 if (fgets(buf, 100, stdin)==NULL)
613 return 0;
614 if (buf[0]=='y' || buf[0]=='Y')
615 return 1;
616 if (buf[0]=='n' || buf[0]=='N')
617 return 0;
618 add = "(y/n) ";
619 }
e7b84f9d 620 pr_err("assuming 'no'\n");
682c7051
NB
621 return 0;
622}
435d4ebb 623#endif /* MDASSEMBLE */
682c7051 624
8d80900b 625int is_standard(char *dev, int *nump)
e5329c37
NB
626{
627 /* tests if dev is a "standard" md dev name.
628 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 629 * where NN is a string of digits
598f0d58
NB
630 * Returns 1 if a partitionable standard,
631 * -1 if non-partitonable,
632 * 0 if not a standard name.
e5329c37 633 */
8d80900b 634 char *d = strrchr(dev, '/');
ca3b6696 635 int type = 0;
8d80900b
NB
636 int num;
637 if (!d)
e5329c37 638 return 0;
ca3b6696
N
639 if (strncmp(d, "/d",2) == 0)
640 d += 2, type = 1; /* /dev/md/dN{pM} */
641 else if (strncmp(d, "/md_d", 5) == 0)
642 d += 5, type = 1; /* /dev/md_dN{pM} */
643 else if (strncmp(d, "/md", 3) == 0)
644 d += 3, type = -1; /* /dev/mdN */
645 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
646 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
647 else
648 return 0;
8d80900b 649 if (!*d)
e5329c37 650 return 0;
8d80900b
NB
651 num = atoi(d);
652 while (isdigit(*d))
653 d++;
654 if (*d)
e5329c37 655 return 0;
8d80900b
NB
656 if (nump) *nump = num;
657
658 return type;
e5329c37
NB
659}
660
4b1ac34b 661unsigned long calc_csum(void *super, int bytes)
82b27616 662{
56eb10c0 663 unsigned long long newcsum = 0;
82b27616 664 int i;
4b1ac34b
NB
665 unsigned int csum;
666 unsigned int *superc = (unsigned int*) super;
82b27616 667
ca3b6696
N
668 for(i = 0; i < bytes/4; i++)
669 newcsum += superc[i];
82b27616 670 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 671#ifdef __alpha__
aba69144 672/* The in-kernel checksum calculation is always 16bit on
570c0542 673 * the alpha, though it is 32 bit on i386...
ca3b6696 674 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
675 * a way that it shouldn't).
676 */
677 csum = (csum & 0xffff) + (csum >> 16);
678 csum = (csum & 0xffff) + (csum >> 16);
679#endif
82b27616
NB
680 return csum;
681}
cd29a5c8 682
435d4ebb 683#ifndef MDASSEMBLE
56eb10c0 684char *human_size(long long bytes)
cd29a5c8
NB
685{
686 static char buf[30];
d5d3721e
NB
687
688 /* We convert bytes to either centi-M{ega,ibi}bytes or
689 * centi-G{igi,ibi}bytes, with appropriate rounding,
690 * and then print 1/100th of those as a decimal.
691 * We allow upto 2048Megabytes before converting to
692 * gigabytes, as that shows more precision and isn't
693 * too large a number.
01619b48 694 * Terabytes are not yet handled.
d5d3721e 695 */
cd29a5c8 696
56eb10c0 697 if (bytes < 5000*1024)
ca3b6696 698 buf[0] = 0;
d5d3721e 699 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 700 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 701 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 702 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
703 cMiB/100 , cMiB % 100,
704 cMB/100, cMB % 100);
705 } else {
93d3bd3b 706 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 707 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 708 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
709 cGiB/100 , cGiB % 100,
710 cGB/100, cGB % 100);
711 }
cd29a5c8
NB
712 return buf;
713}
e0d19036 714
f0ec6710 715char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
716{
717 static char buf[30];
e0d19036 718
570abc6f
MN
719 /* We convert bytes to either centi-M{ega,ibi}bytes or
720 * centi-G{igi,ibi}bytes, with appropriate rounding,
721 * and then print 1/100th of those as a decimal.
722 * We allow upto 2048Megabytes before converting to
723 * gigabytes, as that shows more precision and isn't
724 * too large a number.
725 * Terabytes are not yet handled.
f0ec6710
MN
726 *
727 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
728 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
729 */
730
e0d19036 731 if (bytes < 5000*1024)
570abc6f 732 buf[0] = 0;
f0ec6710
MN
733 else if (prefix == IEC) {
734 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 735 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710
MN
736 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
737 cMiB/100 , cMiB % 100);
738 } else {
93d3bd3b 739 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710
MN
740 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
741 cGiB/100 , cGiB % 100);
742 }
743 }
744 else if (prefix == JEDEC) {
745 if (bytes < 2*1024LL*1024LL*1024LL) {
746 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
747 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
748 cMB/100, cMB % 100);
749 } else {
750 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
751 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
752 cGB/100 , cGB % 100);
753 }
570abc6f 754 }
f0ec6710
MN
755 else
756 buf[0] = 0;
570abc6f 757
e0d19036
NB
758 return buf;
759}
e4965ef8
N
760
761void print_r10_layout(int layout)
762{
763 int near = layout & 255;
764 int far = (layout >> 8) & 255;
765 int offset = (layout&0x10000);
766 char *sep = "";
767
768 if (near != 1) {
769 printf("%s near=%d", sep, near);
770 sep = ",";
771 }
772 if (far != 1)
773 printf("%s %s=%d", sep, offset?"offset":"far", far);
774 if (near*far == 1)
775 printf("NO REDUNDANCY");
776}
435d4ebb 777#endif
e0d19036 778
5f8097be
NB
779unsigned long long calc_array_size(int level, int raid_disks, int layout,
780 int chunksize, unsigned long long devsize)
577e8448 781{
e03640bd
CA
782 if (level == 1)
783 return devsize;
577e8448
AK
784 devsize &= ~(unsigned long long)((chunksize>>9)-1);
785 return get_data_disks(level, layout, raid_disks) * devsize;
786}
787
788int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
789{
790 int data_disks = 0;
791 switch (level) {
ca3b6696
N
792 case 0: data_disks = raid_disks;
793 break;
794 case 1: data_disks = 1;
795 break;
5f8097be 796 case 4:
ca3b6696
N
797 case 5: data_disks = raid_disks - 1;
798 break;
799 case 6: data_disks = raid_disks - 2;
800 break;
5f8097be
NB
801 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
802 break;
803 }
577e8448
AK
804
805 return data_disks;
5f8097be
NB
806}
807
4dd2df09
N
808int devnm2devid(char *devnm)
809{
810 /* First look in /sys/block/$DEVNM/dev for %d:%d
811 * If that fails, try parsing out a number
812 */
813 char path[100];
814 char *ep;
815 int fd;
816 int mjr,mnr;
817
818 sprintf(path, "/sys/block/%s/dev", devnm);
819 fd = open(path, O_RDONLY);
820 if (fd >= 0) {
821 char buf[20];
822 int n = read(fd, buf, sizeof(buf));
823 close(fd);
824 if (n > 0)
825 buf[n] = 0;
826 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
827 return makedev(mjr, mnr);
828 }
829 if (strncmp(devnm, "md_d", 4) == 0 &&
830 isdigit(devnm[4]) &&
831 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
832 ep > devnm && *ep == 0)
833 return makedev(get_mdp_major(), mnr << MdpMinorShift);
834
835 if (strncmp(devnm, "md", 2) == 0 &&
836 isdigit(devnm[2]) &&
837 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
838 ep > devnm && *ep == 0)
839 return makedev(MD_MAJOR, mnr);
840
841 return 0;
842}
843
5a23a06e 844#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 845char *get_md_name(char *devnm)
e0d19036
NB
846{
847 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 848 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 849
e0d19036
NB
850 static char devname[50];
851 struct stat stb;
4dd2df09 852 dev_t rdev = devnm2devid(devnm);
dd0781e5 853 char *dn;
98c6faba 854
4dd2df09
N
855 if (rdev == 0)
856 return 0;
857 if (strncmp(devnm, "md_", 3) == 0) {
858 snprintf(devname, sizeof(devname), "/dev/md/%s",
859 devnm + 3);
98c6faba
NB
860 if (stat(devname, &stb) == 0
861 && (S_IFMT&stb.st_mode) == S_IFBLK
862 && (stb.st_rdev == rdev))
863 return devname;
864 }
4dd2df09
N
865 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
866 if (stat(devname, &stb) == 0
867 && (S_IFMT&stb.st_mode) == S_IFBLK
868 && (stb.st_rdev == rdev))
869 return devname;
870
871 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
872 if (stat(devname, &stb) == 0
873 && (S_IFMT&stb.st_mode) == S_IFBLK
874 && (stb.st_rdev == rdev))
875 return devname;
876
16c6fa80 877 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
878 if (dn)
879 return dn;
4dd2df09 880 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 881 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
882 if (errno != EEXIST)
883 return NULL;
e0d19036
NB
884
885 if (stat(devname, &stb) == 0
886 && (S_IFMT&stb.st_mode) == S_IFBLK
887 && (stb.st_rdev == rdev))
888 return devname;
889 unlink(devname);
890 return NULL;
891}
892
893void put_md_name(char *name)
894{
ca3b6696 895 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
896 unlink(name);
897}
435d4ebb 898#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 899
cc742d38
N
900int get_maj_min(char *dev, int *major, int *minor)
901{
902 char *e;
903 *major = strtoul(dev, &e, 0);
904 return (e > dev && *e == ':' && e[1] &&
905 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
906 *e == 0);
907}
908
8b0dabea
NB
909int dev_open(char *dev, int flags)
910{
911 /* like 'open', but if 'dev' matches %d:%d, create a temp
912 * block device and open that
913 */
8b0dabea
NB
914 int fd = -1;
915 char devname[32];
e81cdd9f 916 int major;
8b0dabea 917 int minor;
e81cdd9f
NB
918
919 if (!dev) return -1;
6df6a774 920 flags |= O_DIRECT;
e81cdd9f 921
cc742d38 922 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
923 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
924 (int)getpid(), major, minor);
925 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
926 fd = open(devname, flags);
927 unlink(devname);
6df6a774
N
928 }
929 if (fd < 0) {
9dc7d357 930 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
931 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
932 (int)getpid(), major, minor);
ca3b6696 933 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
934 fd = open(devname, flags);
935 unlink(devname);
936 }
8b0dabea
NB
937 }
938 } else
6df6a774 939 fd = open(dev, flags);
8b0dabea
NB
940 return fd;
941}
f9ce90ba 942
4dd2df09 943int open_dev_flags(char *devnm, int flags)
e8a70c89 944{
4dd2df09 945 int devid;
e8a70c89
N
946 char buf[20];
947
4dd2df09
N
948 devid = devnm2devid(devnm);
949 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
950 return dev_open(buf, flags);
951}
952
4dd2df09 953int open_dev(char *devnm)
d998b738 954{
4dd2df09 955 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
956}
957
4dd2df09 958int open_dev_excl(char *devnm)
a931db9e
NB
959{
960 char buf[20];
961 int i;
7187750e 962 int flags = O_RDWR;
4dd2df09 963 int devid = devnm2devid(devnm);
399e0b97 964 long delay = 1000;
a931db9e 965
4dd2df09 966 sprintf(buf, "%d:%d", major(devid), minor(devid));
ca3b6696 967 for (i = 0 ; i < 25 ; i++) {
7187750e 968 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
969 if (fd >= 0)
970 return fd;
7187750e
N
971 if (errno == EACCES && flags == O_RDWR) {
972 flags = O_RDONLY;
973 continue;
974 }
a931db9e
NB
975 if (errno != EBUSY)
976 return fd;
399e0b97
N
977 usleep(delay);
978 if (delay < 200000)
979 delay *= 2;
a931db9e
NB
980 }
981 return -1;
982}
983
9008ed1c
N
984int same_dev(char *one, char *two)
985{
986 struct stat st1, st2;
987 if (stat(one, &st1) != 0)
988 return 0;
989 if (stat(two, &st2) != 0)
990 return 0;
991 if ((st1.st_mode & S_IFMT) != S_IFBLK)
992 return 0;
993 if ((st2.st_mode & S_IFMT) != S_IFBLK)
994 return 0;
995 return st1.st_rdev == st2.st_rdev;
996}
997
a7c6e3fb 998void wait_for(char *dev, int fd)
a714580e
N
999{
1000 int i;
a7c6e3fb 1001 struct stat stb_want;
399e0b97 1002 long delay = 1000;
a7c6e3fb
N
1003
1004 if (fstat(fd, &stb_want) != 0 ||
1005 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1006 return;
a714580e 1007
ca3b6696 1008 for (i = 0 ; i < 25 ; i++) {
a714580e 1009 struct stat stb;
a7c6e3fb
N
1010 if (stat(dev, &stb) == 0 &&
1011 (stb.st_mode & S_IFMT) == S_IFBLK &&
1012 (stb.st_rdev == stb_want.st_rdev))
a714580e 1013 return;
399e0b97
N
1014 usleep(delay);
1015 if (delay < 200000)
1016 delay *= 2;
a714580e 1017 }
436305c6 1018 if (i == 25)
1ade5cc1 1019 dprintf("timeout waiting for %s\n", dev);
a714580e
N
1020}
1021
0f22b998
N
1022struct superswitch *superlist[] =
1023{
1024 &super0, &super1,
1025 &super_ddf, &super_imsm,
0592faeb 1026 &mbr, &gpt,
0f22b998 1027 NULL };
f9ce90ba 1028
ea24acd0 1029#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1030
4725bc31 1031struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1032{
1686dc25
NB
1033 mdu_array_info_t array;
1034 int vers;
1035 int minor;
1036 struct supertype *st = NULL;
7e0f6979 1037 struct mdinfo *sra;
142cb9e1 1038 char *verstr;
1686dc25
NB
1039 char version[20];
1040 int i;
f7e7067b 1041 char *subarray = NULL;
4dd2df09 1042 char container[32] = "";
1686dc25 1043
4dd2df09 1044 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1045
1046 if (sra) {
7e0f6979
NB
1047 vers = sra->array.major_version;
1048 minor = sra->array.minor_version;
142cb9e1 1049 verstr = sra->text_version;
1686dc25
NB
1050 } else {
1051 if (ioctl(fd, GET_ARRAY_INFO, &array))
1052 array.major_version = array.minor_version = 0;
1053 vers = array.major_version;
1054 minor = array.minor_version;
142cb9e1 1055 verstr = "";
6fbba4c9 1056 }
82d9eba6 1057
1686dc25
NB
1058 if (vers != -1) {
1059 sprintf(version, "%d.%d", vers, minor);
1060 verstr = version;
6fbba4c9 1061 }
3c558363 1062 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1063 char *dev = verstr+1;
5f7e44b2 1064
f7e7067b 1065 subarray = strchr(dev, '/');
92d49ecf 1066 if (subarray) {
f7e7067b 1067 *subarray++ = '\0';
503975b9 1068 subarray = xstrdup(subarray);
92d49ecf 1069 }
4dd2df09 1070 strcpy(container, dev);
f7e7067b
NB
1071 if (sra)
1072 sysfs_free(sra);
5f7e44b2 1073 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1074 if (sra && sra->text_version[0])
1075 verstr = sra->text_version;
1076 else
1077 verstr = "-no-metadata-";
f7e7067b
NB
1078 }
1079
1080 for (i = 0; st == NULL && superlist[i] ; i++)
1081 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
1082
1083 if (sra)
1084 sysfs_free(sra);
f7e7067b 1085 if (st) {
3b0896f8 1086 st->sb = NULL;
1f49fb3a
N
1087 if (subarrayp)
1088 *subarrayp = subarray;
4dd2df09
N
1089 strcpy(st->container_devnm, container);
1090 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1091 } else
1092 free(subarray);
5f7e44b2 1093
82d9eba6 1094 return st;
f9ce90ba 1095}
ea24acd0
NB
1096#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1097
0f0749ad 1098int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1099{
1100 char buf[20];
1101 int fd;
1102
1103 sprintf(buf, "%d:%d", major(id), minor(id));
1104 fd = dev_open(buf, O_RDONLY);
1105 if (fd < 0)
1106 return 0;
1107 if (get_dev_size(fd, NULL, size)) {
1108 close(fd);
1109 return 1;
1110 }
1111 close(fd);
1112 return 0;
1113}
f9ce90ba 1114
159c3a1a 1115struct supertype *dup_super(struct supertype *orig)
3da92f27 1116{
159c3a1a 1117 struct supertype *st;
1686dc25 1118
d2ca6449
NB
1119 if (!orig)
1120 return orig;
503975b9 1121 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1122 st->ss = orig->ss;
1123 st->max_devs = orig->max_devs;
1124 st->minor_version = orig->minor_version;
cb8f6859 1125 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1126 st->data_offset = orig->data_offset;
159c3a1a
NB
1127 st->sb = NULL;
1128 st->info = NULL;
1129 return st;
3da92f27
NB
1130}
1131
54887ad8 1132struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1133{
1134 /* try each load_super to find the best match,
1135 * and return the best superswitch
1136 */
82d9eba6
NB
1137 struct superswitch *ss;
1138 struct supertype *st;
f21e18ca 1139 time_t besttime = 0;
570c0542 1140 int bestsuper = -1;
f9ce90ba
NB
1141 int i;
1142
503975b9 1143 st = xcalloc(1, sizeof(*st));
4dd2df09 1144 st->container_devnm[0] = 0;
d1d599ea 1145
ca3b6696 1146 for (i = 0 ; superlist[i]; i++) {
f9ce90ba
NB
1147 int rv;
1148 ss = superlist[i];
54887ad8
N
1149 if (guess_type == guess_array && ss->add_to_super == NULL)
1150 continue;
1151 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1152 continue;
ef609477 1153 memset(st, 0, sizeof(*st));
df3346e6 1154 st->ignore_hw_compat = 1;
3da92f27 1155 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1156 if (rv == 0) {
1157 struct mdinfo info;
a5d85af7 1158 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1159 if (bestsuper == -1 ||
1160 besttime < info.array.ctime) {
1161 bestsuper = i;
1162 besttime = info.array.ctime;
570c0542 1163 }
3da92f27 1164 ss->free_super(st);
570c0542
NB
1165 }
1166 }
1167 if (bestsuper != -1) {
1168 int rv;
ef609477 1169 memset(st, 0, sizeof(*st));
df3346e6 1170 st->ignore_hw_compat = 1;
3da92f27 1171 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1172 if (rv == 0) {
5e747af2 1173 superlist[bestsuper]->free_super(st);
82d9eba6 1174 return st;
f9ce90ba
NB
1175 }
1176 }
570c0542 1177 free(st);
f9ce90ba
NB
1178 return NULL;
1179}
fe6729fa 1180
beae1dfe
NB
1181/* Return size of device in bytes */
1182int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1183{
1184 unsigned long long ldsize;
c2c9bb6f
NB
1185 struct stat st;
1186
1187 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1188 ldsize = (unsigned long long)st.st_size;
1189 else
beae1dfe
NB
1190#ifdef BLKGETSIZE64
1191 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1192#endif
1193 {
1194 unsigned long dsize;
1195 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1196 ldsize = dsize;
1197 ldsize <<= 9;
1198 } else {
1199 if (dname)
e7b84f9d 1200 pr_err("Cannot get size of %s: %s\b",
beae1dfe
NB
1201 dname, strerror(errno));
1202 return 0;
1203 }
1204 }
1205 *sizep = ldsize;
1206 return 1;
1207}
8fac0577 1208
3a371610
N
1209/* Return true if this can only be a container, not a member device.
1210 * i.e. is and md device and size is zero
1211 */
1212int must_be_container(int fd)
1213{
1214 unsigned long long size;
1215 if (md_get_version(fd) < 0)
1216 return 0;
1217 if (get_dev_size(fd, NULL, &size) == 0)
1218 return 1;
1219 if (size == 0)
1220 return 1;
1221 return 0;
1222}
034b203a
TM
1223
1224/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1225 * Returns: 1 if successful
1226 * -1 for unknown partition type
1227 * 0 for other errors
1228 */
1229static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1230{
056b331e 1231 struct GPT gpt;
034b203a 1232 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1233 struct GPT_part_entry *part;
1234 char buf[512];
034b203a
TM
1235 unsigned long long curr_part_end;
1236 unsigned all_partitions, entry_size;
f21e18ca 1237 unsigned part_nr;
034b203a
TM
1238
1239 *endofpart = 0;
1240
056b331e 1241 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1242 /* skip protective MBR */
034b203a 1243 lseek(fd, 512, SEEK_SET);
73e658d8 1244 /* read GPT header */
056b331e 1245 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1246 return 0;
1247
1248 /* get the number of partition entries and the entry size */
056b331e
N
1249 all_partitions = __le32_to_cpu(gpt.part_cnt);
1250 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1251
1252 /* Check GPT signature*/
056b331e 1253 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1254 return -1;
1255
1256 /* sanity checks */
1257 if (all_partitions > 1024 ||
db7fdfe4 1258 entry_size > sizeof(buf))
034b203a
TM
1259 return -1;
1260
db7fdfe4
JS
1261 part = (struct GPT_part_entry *)buf;
1262
ca3b6696 1263 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1264 /* read partition entry */
db7fdfe4 1265 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1266 return 0;
1267
034b203a 1268 /* is this valid partition? */
db7fdfe4 1269 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1270 /* check the last lba for the current partition */
db7fdfe4 1271 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1272 if (curr_part_end > *endofpart)
1273 *endofpart = curr_part_end;
1274 }
1275
034b203a
TM
1276 }
1277 return 1;
1278}
1279
1280/* Sets endofpart parameter to the last block used by the last partition on the device.
1281 * Returns: 1 if successful
1282 * -1 for unknown partition type
1283 * 0 for other errors
1284 */
1285static int get_last_partition_end(int fd, unsigned long long *endofpart)
1286{
056b331e 1287 struct MBR boot_sect;
034b203a
TM
1288 struct MBR_part_record *part;
1289 unsigned long long curr_part_end;
f21e18ca 1290 unsigned part_nr;
034b203a
TM
1291 int retval = 0;
1292
1293 *endofpart = 0;
1294
056b331e 1295 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1296 /* read MBR */
1297 lseek(fd, 0, 0);
056b331e 1298 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1299 goto abort;
1300
1301 /* check MBP signature */
056b331e 1302 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1303 retval = 1;
1304 /* found the correct signature */
056b331e 1305 part = boot_sect.parts;
034b203a 1306
ca3b6696 1307 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
034b203a
TM
1308 /* check for GPT type */
1309 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1310 retval = get_gpt_last_partition_end(fd, endofpart);
1311 break;
1312 }
1313 /* check the last used lba for the current partition */
1314 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1315 __le32_to_cpu(part->blocks_num);
1316 if (curr_part_end > *endofpart)
1317 *endofpart = curr_part_end;
1318
1319 part++;
1320 }
1321 } else {
1322 /* Unknown partition table */
1323 retval = -1;
1324 }
1325 abort:
1326 return retval;
1327}
1328
53ed6ac3
KW
1329int check_partitions(int fd, char *dname, unsigned long long freesize,
1330 unsigned long long size)
034b203a
TM
1331{
1332 /*
1333 * Check where the last partition ends
1334 */
1335 unsigned long long endofpart;
1336 int ret;
1337
1338 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1339 /* There appears to be a partition table here */
1340 if (freesize == 0) {
1341 /* partitions will not be visible in new device */
e7b84f9d
N
1342 pr_err("partition table exists on %s but will be lost or\n"
1343 " meaningless after creating array\n",
1344 dname);
034b203a
TM
1345 return 1;
1346 } else if (endofpart > freesize) {
1347 /* last partition overlaps metadata */
e7b84f9d
N
1348 pr_err("metadata will over-write last partition on %s.\n",
1349 dname);
034b203a 1350 return 1;
53ed6ac3
KW
1351 } else if (size && endofpart > size) {
1352 /* partitions will be truncated in new device */
e7b84f9d
N
1353 pr_err("array size is too small to cover all partitions on %s.\n",
1354 dname);
53ed6ac3 1355 return 1;
034b203a
TM
1356 }
1357 }
1358 return 0;
1359}
1360
a322f70c
DW
1361int open_container(int fd)
1362{
1363 /* 'fd' is a block device. Find out if it is in use
1364 * by a container, and return an open fd on that container.
1365 */
1366 char path[256];
1367 char *e;
1368 DIR *dir;
1369 struct dirent *de;
1370 int dfd, n;
1371 char buf[200];
1372 int major, minor;
1373 struct stat st;
1374
1375 if (fstat(fd, &st) != 0)
1376 return -1;
1377 sprintf(path, "/sys/dev/block/%d:%d/holders",
1378 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1379 e = path + strlen(path);
1380
1381 dir = opendir(path);
1382 if (!dir)
1383 return -1;
1384 while ((de = readdir(dir))) {
1385 if (de->d_ino == 0)
1386 continue;
1387 if (de->d_name[0] == '.')
1388 continue;
a994592d
N
1389 /* Need to make sure it is a container and not a volume */
1390 sprintf(e, "/%s/md/metadata_version", de->d_name);
1391 dfd = open(path, O_RDONLY);
1392 if (dfd < 0)
1393 continue;
1394 n = read(dfd, buf, sizeof(buf));
1395 close(dfd);
1396 if (n <= 0 || (unsigned)n >= sizeof(buf))
1397 continue;
1398 buf[n] = 0;
1399 if (strncmp(buf, "external", 8) != 0 ||
1400 n < 10 ||
1401 buf[9] == '/')
1402 continue;
a322f70c
DW
1403 sprintf(e, "/%s/dev", de->d_name);
1404 dfd = open(path, O_RDONLY);
1405 if (dfd < 0)
1406 continue;
1407 n = read(dfd, buf, sizeof(buf));
1408 close(dfd);
f21e18ca 1409 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1410 continue;
1411 buf[n] = 0;
1412 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1413 continue;
1414 sprintf(buf, "%d:%d", major, minor);
1415 dfd = dev_open(buf, O_RDONLY);
1416 if (dfd >= 0) {
1417 closedir(dir);
1418 return dfd;
1419 }
1420 }
355726fa 1421 closedir(dir);
a322f70c
DW
1422 return -1;
1423}
1424
33414a01
DW
1425struct superswitch *version_to_superswitch(char *vers)
1426{
1427 int i;
1428
1429 for (i = 0; superlist[i]; i++) {
1430 struct superswitch *ss = superlist[i];
1431
1432 if (strcmp(vers, ss->name) == 0)
1433 return ss;
1434 }
1435
1436 return NULL;
1437}
1438
4dd2df09
N
1439int metadata_container_matches(char *metadata, char *devnm)
1440{
1441 /* Check if 'devnm' is the container named in 'metadata'
1442 * which is
1443 * /containername/componentname or
1444 * -containername/componentname
1445 */
1446 int l;
1447 if (*metadata != '/' && *metadata != '-')
1448 return 0;
1449 l = strlen(devnm);
1450 if (strncmp(metadata+1, devnm, l) != 0)
1451 return 0;
1452 if (metadata[l+1] != '/')
1453 return 0;
1454 return 1;
1455}
1456
1457int metadata_subdev_matches(char *metadata, char *devnm)
1458{
1459 /* Check if 'devnm' is the subdev named in 'metadata'
1460 * which is
1461 * /containername/subdev or
1462 * -containername/subdev
1463 */
1464 char *sl;
1465 if (*metadata != '/' && *metadata != '-')
1466 return 0;
1467 sl = strchr(metadata+1, '/');
1468 if (!sl)
1469 return 0;
1470 if (strcmp(sl+1, devnm) == 0)
1471 return 1;
1472 return 0;
1473}
1474
33414a01
DW
1475int is_container_member(struct mdstat_ent *mdstat, char *container)
1476{
1477 if (mdstat->metadata_version == NULL ||
1478 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1479 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1480 return 0;
1481
1482 return 1;
1483}
1484
1485int is_subarray_active(char *subarray, char *container)
1486{
1487 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1488 struct mdstat_ent *ent;
1489
e5408a32
DW
1490 for (ent = mdstat; ent; ent = ent->next)
1491 if (is_container_member(ent, container))
e5e5d7ce 1492 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1493 break;
33414a01
DW
1494
1495 free_mdstat(mdstat);
1496
1497 return ent != NULL;
1498}
1499
1500/* open_subarray - opens a subarray in a container
1501 * @dev: container device name
feab51f8 1502 * @st: empty supertype
33414a01
DW
1503 * @quiet: block reporting errors flag
1504 *
1505 * On success returns an fd to a container and fills in *st
1506 */
feab51f8 1507int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1508{
1509 struct mdinfo *mdi;
a951a4f7 1510 struct mdinfo *info;
33414a01 1511 int fd, err = 1;
4dd2df09 1512 char *_devnm;
33414a01
DW
1513
1514 fd = open(dev, O_RDWR|O_EXCL);
1515 if (fd < 0) {
1516 if (!quiet)
e7b84f9d 1517 pr_err("Couldn't open %s, aborting\n",
33414a01 1518 dev);
b990032d 1519 return -1;
33414a01
DW
1520 }
1521
4dd2df09
N
1522 _devnm = fd2devnm(fd);
1523 if (_devnm == NULL) {
33414a01 1524 if (!quiet)
e7b84f9d
N
1525 pr_err("Failed to determine device number for %s\n",
1526 dev);
33414a01
DW
1527 goto close_fd;
1528 }
4dd2df09 1529 strcpy(st->devnm, _devnm);
33414a01 1530
4dd2df09 1531 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1532 if (!mdi) {
1533 if (!quiet)
e7b84f9d 1534 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1535 dev);
1536 goto close_fd;
1537 }
1538
1539 if (mdi->array.level != UnSet) {
1540 if (!quiet)
e7b84f9d 1541 pr_err("%s is not a container\n", dev);
33414a01
DW
1542 goto free_sysfs;
1543 }
1544
1545 st->ss = version_to_superswitch(mdi->text_version);
1546 if (!st->ss) {
1547 if (!quiet)
e7b84f9d
N
1548 pr_err("Operation not supported for %s metadata\n",
1549 mdi->text_version);
33414a01
DW
1550 goto free_sysfs;
1551 }
1552
4dd2df09 1553 if (st->devnm[0] == 0) {
33414a01 1554 if (!quiet)
e7b84f9d 1555 pr_err("Failed to allocate device name\n");
33414a01
DW
1556 goto free_sysfs;
1557 }
1558
db20d413 1559 if (!st->ss->load_container) {
33414a01 1560 if (!quiet)
e7b84f9d 1561 pr_err("%s is not a container\n", dev);
4dd2df09 1562 goto free_sysfs;
33414a01
DW
1563 }
1564
db20d413 1565 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1566 if (!quiet)
e7b84f9d 1567 pr_err("Failed to load metadata for %s\n",
db20d413 1568 dev);
4dd2df09 1569 goto free_sysfs;
33414a01
DW
1570 }
1571
a951a4f7
N
1572 info = st->ss->container_content(st, subarray);
1573 if (!info) {
1574 if (!quiet)
e7b84f9d 1575 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1576 subarray, dev);
1577 goto free_super;
1578 }
1579 free(info);
1580
33414a01
DW
1581 err = 0;
1582
1583 free_super:
1584 if (err)
1585 st->ss->free_super(st);
33414a01
DW
1586 free_sysfs:
1587 sysfs_free(mdi);
1588 close_fd:
1589 if (err)
1590 close(fd);
1591
1592 if (err)
1593 return -1;
1594 else
1595 return fd;
1596}
1597
7801ac20
N
1598int add_disk(int mdfd, struct supertype *st,
1599 struct mdinfo *sra, struct mdinfo *info)
1600{
1601 /* Add a device to an array, in one of 2 ways. */
1602 int rv;
1603#ifndef MDASSEMBLE
1604 if (st->ss->external) {
d23534e4
DW
1605 if (info->disk.state & (1<<MD_DISK_SYNC))
1606 info->recovery_start = MaxSector;
1607 else
1608 info->recovery_start = 0;
2904b26f 1609 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1610 if (! rv) {
1611 struct mdinfo *sd2;
f35f2525
N
1612 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1613 if (sd2 == info)
1614 break;
1615 if (sd2 == NULL) {
503975b9 1616 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1617 *sd2 = *info;
1618 sd2->next = sra->devs;
1619 sra->devs = sd2;
1620 }
7801ac20
N
1621 }
1622 } else
1623#endif
1624 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1625 return rv;
1626}
1627
de6ae750
N
1628int remove_disk(int mdfd, struct supertype *st,
1629 struct mdinfo *sra, struct mdinfo *info)
1630{
1631 int rv;
1632 /* Remove the disk given by 'info' from the array */
1633#ifndef MDASSEMBLE
1634 if (st->ss->external)
1635 rv = sysfs_set_str(sra, info, "slot", "none");
1636 else
1637#endif
1638 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1639 info->disk.minor));
1640 return rv;
1641}
1642
f35f2525
N
1643int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1644{
1645 /* Initialise kernel's knowledge of array.
1646 * This varies between externally managed arrays
1647 * and older kernels
1648 */
1649 int vers = md_get_version(mdfd);
1650 int rv;
1651
1652#ifndef MDASSEMBLE
1653 if (st->ss->external)
1654 rv = sysfs_set_array(info, vers);
1655 else
1656#endif
1657 if ((vers % 100) >= 1) { /* can use different versions */
1658 mdu_array_info_t inf;
1659 memset(&inf, 0, sizeof(inf));
1660 inf.major_version = info->array.major_version;
1661 inf.minor_version = info->array.minor_version;
1662 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1663 } else
1664 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1665 return rv;
1666}
1667
1e5c6983
DW
1668unsigned long long min_recovery_start(struct mdinfo *array)
1669{
1670 /* find the minimum recovery_start in an array for metadata
1671 * formats that only record per-array recovery progress instead
1672 * of per-device
1673 */
1674 unsigned long long recovery_start = MaxSector;
1675 struct mdinfo *d;
1676
1677 for (d = array->devs; d; d = d->next)
1678 recovery_start = min(recovery_start, d->recovery_start);
1679
1680 return recovery_start;
1681}
1682
4dd2df09 1683int mdmon_pid(char *devnm)
a931db9e
NB
1684{
1685 char path[100];
1686 char pid[10];
1687 int fd;
1688 int n;
10013317 1689
4dd2df09 1690 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1691
24f6f99b 1692 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1693
1694 if (fd < 0)
cf556303 1695 return -1;
a931db9e
NB
1696 n = read(fd, pid, 9);
1697 close(fd);
1698 if (n <= 0)
cf556303 1699 return -1;
24f6f99b 1700 return atoi(pid);
a931db9e
NB
1701}
1702
4dd2df09 1703int mdmon_running(char *devnm)
a931db9e 1704{
4dd2df09 1705 int pid = mdmon_pid(devnm);
24f6f99b 1706 if (pid <= 0)
a931db9e 1707 return 0;
24f6f99b 1708 if (kill(pid, 0) == 0)
a931db9e
NB
1709 return 1;
1710 return 0;
1711}
1712
4dd2df09 1713int start_mdmon(char *devnm)
8850ee3e 1714{
1913c325 1715 int i, skipped;
44d2e365 1716 int len;
ca3b6696 1717 pid_t pid;
9fe32043 1718 int status;
44d2e365
N
1719 char pathbuf[1024];
1720 char *paths[4] = {
1721 pathbuf,
85945e19 1722 BINDIR "/mdmon",
2f1bcf43 1723 "./mdmon",
44d2e365
N
1724 NULL
1725 };
8850ee3e 1726
40ebbb9c 1727 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1728 return 0;
1729
9cf014ec 1730 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1731 if (len > 0) {
1732 char *sl;
1733 pathbuf[len] = 0;
1734 sl = strrchr(pathbuf, '/');
1735 if (sl)
1736 sl++;
1737 else
1738 sl = pathbuf;
1739 strcpy(sl, "mdmon");
1740 } else
1741 pathbuf[0] = '\0';
1742
0f7bdf89 1743 /* First try to run systemctl */
9540cc24
N
1744 if (!check_env("MDADM_NO_SYSTEMCTL"))
1745 switch(fork()) {
1746 case 0:
1747 /* FIXME yuk. CLOSE_EXEC?? */
1748 skipped = 0;
1749 for (i = 3; skipped < 20; i++)
1750 if (close(i) < 0)
1751 skipped++;
1752 else
1753 skipped = 0;
1754
1755 /* Don't want to see error messages from
1756 * systemctl. If the service doesn't exist,
1757 * we start mdmon ourselves.
1758 */
1759 close(2);
1760 open("/dev/null", O_WRONLY);
1761 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1762 devnm);
1763 status = execl("/usr/bin/systemctl", "systemctl",
1764 "start",
1765 pathbuf, NULL);
1766 status = execl("/bin/systemctl", "systemctl", "start",
1767 pathbuf, NULL);
1768 exit(1);
7a862a02 1769 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1770 return -1;
1771 default: /* parent - good */
1772 pid = wait(&status);
1773 if (pid >= 0 && status == 0)
1774 return 0;
1775 }
0f7bdf89
JS
1776
1777 /* That failed, try running mdmon directly */
8850ee3e
N
1778 switch(fork()) {
1779 case 0:
1780 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1781 skipped = 0;
ca3b6696 1782 for (i = 3; skipped < 20; i++)
1913c325
N
1783 if (close(i) < 0)
1784 skipped++;
1785 else
1786 skipped = 0;
1787
ca3b6696 1788 for (i = 0; paths[i]; i++)
a0963a86 1789 if (paths[i][0]) {
2f1bcf43 1790 execl(paths[i], paths[i],
4dd2df09 1791 devnm, NULL);
a0963a86 1792 }
8850ee3e 1793 exit(1);
7a862a02 1794 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 1795 return -1;
9fe32043
N
1796 default: /* parent - good */
1797 pid = wait(&status);
15c10423 1798 if (pid < 0 || status != 0) {
7a862a02 1799 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 1800 return -1;
15c10423 1801 }
8850ee3e
N
1802 }
1803 return 0;
1804}
1805
148acb7b
DW
1806__u32 random32(void)
1807{
1808 __u32 rv;
1809 int rfd = open("/dev/urandom", O_RDONLY);
1810 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1811 rv = random();
1812 if (rfd >= 0)
1813 close(rfd);
1814 return rv;
1815}
1816
0e600426 1817#ifndef MDASSEMBLE
edd8d13c
NB
1818int flush_metadata_updates(struct supertype *st)
1819{
1820 int sfd;
1821 if (!st->updates) {
1822 st->update_tail = NULL;
1823 return -1;
1824 }
1825
4dd2df09 1826 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
1827 if (sfd < 0)
1828 return -1;
1829
1830 while (st->updates) {
1831 struct metadata_update *mu = st->updates;
1832 st->updates = mu->next;
1833
1834 send_message(sfd, mu, 0);
1835 wait_reply(sfd, 0);
1836 free(mu->buf);
1837 free(mu);
1838 }
1839 ack(sfd, 0);
1840 wait_reply(sfd, 0);
1841 close(sfd);
1842 st->update_tail = NULL;
1843 return 0;
1844}
1845
1846void append_metadata_update(struct supertype *st, void *buf, int len)
1847{
1848
503975b9 1849 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
1850
1851 mu->buf = buf;
1852 mu->len = len;
1853 mu->space = NULL;
cb23f1f4 1854 mu->space_list = NULL;
edd8d13c
NB
1855 mu->next = NULL;
1856 *st->update_tail = mu;
1857 st->update_tail = &mu->next;
1858}
0e600426 1859#endif /* MDASSEMBLE */
a931db9e 1860
fe6729fa
NB
1861#ifdef __TINYC__
1862/* tinyc doesn't optimize this check in ioctl.h out ... */
1863unsigned int __invalid_size_argument_for_IOC = 0;
1864#endif
1865
6d11ec6f
AK
1866int experimental(void)
1867{
1868 if (check_env("MDADM_EXPERIMENTAL"))
1869 return 1;
1870 else {
7a862a02 1871 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
1872 return 0;
1873 }
1874}
1875
326727d9
AC
1876/* Pick all spares matching given criteria from a container
1877 * if min_size == 0 do not check size
1878 * if domlist == NULL do not check domains
1879 * if spare_group given add it to domains of each spare
1880 * metadata allows to test domains using metadata of destination array */
1881struct mdinfo *container_choose_spares(struct supertype *st,
1882 unsigned long long min_size,
1883 struct domainlist *domlist,
1884 char *spare_group,
1885 const char *metadata, int get_one)
1886{
1887 struct mdinfo *d, **dp, *disks = NULL;
1888
1889 /* get list of all disks in container */
1890 if (st->ss->getinfo_super_disks)
1891 disks = st->ss->getinfo_super_disks(st);
1892
1893 if (!disks)
1894 return disks;
1895 /* find spare devices on the list */
1896 dp = &disks->devs;
1897 disks->array.spare_disks = 0;
1898 while (*dp) {
1899 int found = 0;
1900 d = *dp;
1901 if (d->disk.state == 0) {
1902 /* check if size is acceptable */
1903 unsigned long long dev_size;
1904 dev_t dev = makedev(d->disk.major,d->disk.minor);
1905
1906 if (!min_size ||
1907 (dev_size_from_id(dev, &dev_size) &&
1908 dev_size >= min_size))
1909 found = 1;
1910 /* check if domain matches */
1911 if (found && domlist) {
4dd2df09 1912 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
1913 if (spare_group)
1914 pol_add(&pol, pol_domain,
1915 spare_group, NULL);
e5508b36 1916 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1917 found = 0;
1918 dev_policy_free(pol);
1919 }
1920 }
1921 if (found) {
1922 dp = &d->next;
1923 disks->array.spare_disks++;
1924 if (get_one) {
1925 sysfs_free(*dp);
1926 d->next = NULL;
1927 }
1928 } else {
1929 *dp = d->next;
1930 d->next = NULL;
1931 sysfs_free(d);
1932 }
1933 }
1934 return disks;
1935}
9eafa1de
MN
1936
1937/* Checks if paths point to the same device
1938 * Returns 0 if they do.
1939 * Returns 1 if they don't.
1940 * Returns -1 if something went wrong,
1941 * e.g. paths are empty or the files
1942 * they point to don't exist */
1943int compare_paths (char* path1, char* path2)
1944{
1945 struct stat st1,st2;
1946
1947 if (path1 == NULL || path2 == NULL)
1948 return -1;
1949 if (stat(path1,&st1) != 0)
1950 return -1;
1951 if (stat(path2,&st2) != 0)
1952 return -1;
1953 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
1954 return 0;
1955 return 1;
1956}
a7dec3fd
N
1957
1958/* Make sure we can open as many devices as needed */
1959void enable_fds(int devices)
1960{
1961 unsigned int fds = 20 + devices;
1962 struct rlimit lim;
1963 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
1964 || lim.rlim_cur >= fds)
1965 return;
1966 if (lim.rlim_max < fds)
1967 lim.rlim_max = fds;
1968 lim.rlim_cur = fds;
1969 setrlimit(RLIMIT_NOFILE, &lim);
1970}
a9c15847
N
1971
1972int in_initrd(void)
1973{
1974 /* This is based on similar function in systemd. */
1975 struct statfs s;
76d0f188
JS
1976 /* statfs.f_type is signed long on s390x and MIPS, causing all
1977 sorts of sign extension problems with RAMFS_MAGIC being
1978 defined as 0x858458f6 */
a9c15847 1979 return statfs("/", &s) >= 0 &&
7ac5d47e 1980 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
1981 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
1982 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 1983}
8832342d
N
1984
1985void reopen_mddev(int mdfd)
1986{
1987 /* Re-open without any O_EXCL, but keep
1988 * the same fd
1989 */
1990 char *devnm;
1991 int fd;
1992 devnm = fd2devnm(mdfd);
1993 close(mdfd);
1994 fd = open_dev(devnm);
1995 if (fd >= 0 && fd != mdfd)
1996 dup2(fd, mdfd);
1997}
7716570e
GJ
1998
1999int get_cluster_name(char **cluster_name)
2000{
2001 void *lib_handle = NULL;
2002 int rv = -1;
2003
2004 cmap_handle_t handle;
2005 static int (*initialize)(cmap_handle_t *handle);
2006 static int (*get_string)(cmap_handle_t handle,
2007 const char *string,
2008 char **name);
2009 static int (*finalize)(cmap_handle_t handle);
2010
2011
2012 lib_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2013 if (!lib_handle)
2014 return rv;
2015
2016 initialize = dlsym(lib_handle, "cmap_initialize");
2017 if (!initialize)
2018 goto out;
2019
2020 get_string = dlsym(lib_handle, "cmap_get_string");
2021 if (!get_string)
2022 goto out;
2023
2024 finalize = dlsym(lib_handle, "cmap_finalize");
2025 if (!finalize)
2026 goto out;
2027
2028 rv = initialize(&handle);
2029 if (rv != CS_OK)
2030 goto out;
2031
2032 rv = get_string(handle, "totem.cluster_name", cluster_name);
2033 if (rv != CS_OK) {
2034 free(*cluster_name);
2035 rv = -1;
2036 goto name_err;
2037 }
2038
2039 rv = 0;
2040name_err:
2041 finalize(handle);
2042out:
2043 dlclose(lib_handle);
2044 return rv;
2045}