managemon: fix typo affecting incrmental assembly.
[thirdparty/mdadm.git] / util.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
ca3b6696 4 * Copyright (C) 2001-2012 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>
0a816ef9
NB
37
38/*
39 * following taken from linux/blkpg.h because they aren't
40 * anywhere else and it isn't safe to #include linux/ * stuff.
41 */
42
43#define BLKPG _IO(0x12,105)
44
45/* The argument structure */
46struct blkpg_ioctl_arg {
5d500228
N
47 int op;
48 int flags;
49 int datalen;
50 void *data;
0a816ef9
NB
51};
52
53/* The subfunctions (for the op field) */
54#define BLKPG_ADD_PARTITION 1
55#define BLKPG_DEL_PARTITION 2
56
57/* Sizes of name fields. Unused at present. */
58#define BLKPG_DEVNAMELTH 64
59#define BLKPG_VOLNAMELTH 64
60
61/* The data structure for ADD_PARTITION and DEL_PARTITION */
62struct blkpg_partition {
63 long long start; /* starting offset in bytes */
64 long long length; /* length in bytes */
65 int pno; /* partition number */
66 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
67 to be used in kernel messages */
68 char volname[BLKPG_VOLNAMELTH]; /* volume label */
69};
64c4757e 70
0f22b998 71#include "part.h"
056b331e
N
72
73/* Force a compilation error if condition is true */
74#define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
75
76/* Force a compilation error if condition is true, but also produce a
77 result (of value 0 and type size_t), so the expression can be used
78 e.g. in a structure initializer (or where-ever else comma expressions
79 aren't permitted). */
80#define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
81
64c4757e
NB
82/*
83 * Parse a 128 bit uuid in 4 integers
84 * format is 32 hexx nibbles with options :.<space> separator
85 * If not exactly 32 hex digits are found, return 0
86 * else return 1
87 */
88int parse_uuid(char *str, int uuid[4])
89{
aba69144
NB
90 int hit = 0; /* number of Hex digIT */
91 int i;
92 char c;
ca3b6696
N
93 for (i = 0; i < 4; i++)
94 uuid[i] = 0;
aba69144 95
ca3b6696 96 while ((c = *str++) != 0) {
aba69144 97 int n;
ca3b6696 98 if (c >= '0' && c <= '9')
aba69144 99 n = c-'0';
ca3b6696 100 else if (c >= 'a' && c <= 'f')
aba69144 101 n = 10 + c - 'a';
ca3b6696 102 else if (c >= 'A' && c <= 'F')
aba69144
NB
103 n = 10 + c - 'A';
104 else if (strchr(":. -", c))
105 continue;
106 else return 0;
107
108 if (hit<32) {
109 uuid[hit/8] <<= 4;
110 uuid[hit/8] += n;
111 }
112 hit++;
82b27616 113 }
aba69144
NB
114 if (hit == 32)
115 return 1;
116 return 0;
64c4757e
NB
117}
118
64c4757e
NB
119/*
120 * Get the md version number.
121 * We use the RAID_VERSION ioctl if it is supported
122 * If not, but we have a block device with major '9', we assume
123 * 0.36.0
124 *
125 * Return version number as 24 but number - assume version parts
126 * always < 255
127 */
128
129int md_get_version(int fd)
130{
5d500228
N
131 struct stat stb;
132 mdu_version_t vers;
133
134 if (fstat(fd, &stb)<0)
135 return -1;
136 if ((S_IFMT&stb.st_mode) != S_IFBLK)
137 return -1;
138
139 if (ioctl(fd, RAID_VERSION, &vers) == 0)
140 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
141 if (errno == EACCES)
142 return -1;
143 if (major(stb.st_rdev) == MD_MAJOR)
144 return (3600);
64c4757e 145 return -1;
64c4757e
NB
146}
147
64c4757e
NB
148int get_linux_version()
149{
150 struct utsname name;
98c6faba 151 char *cp;
f161d047 152 int a = 0, b = 0,c = 0;
64c4757e
NB
153 if (uname(&name) <0)
154 return -1;
155
98c6faba
NB
156 cp = name.release;
157 a = strtoul(cp, &cp, 10);
f161d047
N
158 if (*cp == '.')
159 b = strtoul(cp+1, &cp, 10);
160 if (*cp == '.')
161 c = strtoul(cp+1, &cp, 10);
98c6faba 162
682c7051 163 return (a*1000000)+(b*1000)+c;
64c4757e
NB
164}
165
71204a50 166#ifndef MDASSEMBLE
bc77ed53
DW
167int mdadm_version(char *version)
168{
169 int a, b, c;
170 char *cp;
171
172 if (!version)
173 version = Version;
174
175 cp = strchr(version, '-');
176 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
177 return -1;
178 cp += 3;
179 a = strtoul(cp, &cp, 10);
180 if (*cp != '.')
181 return -1;
182 b = strtoul(cp+1, &cp, 10);
183 if (*cp == '.')
184 c = strtoul(cp+1, &cp, 10);
185 else
186 c = 0;
187 if (*cp != ' ' && *cp != '-')
188 return -1;
189 return (a*1000000)+(b*1000)+c;
190}
191
51d4261c 192unsigned long long parse_size(char *size)
84e11361
N
193{
194 /* parse 'size' which should be a number optionally
195 * followed by 'K', 'M', or 'G'.
196 * Without a suffix, K is assumed.
197 * Number returned is in sectors (half-K)
822e393a 198 * INVALID_SECTORS returned on error.
84e11361
N
199 */
200 char *c;
201 long long s = strtoll(size, &c, 10);
202 if (s > 0) {
203 switch (*c) {
204 case 'K':
205 c++;
206 default:
207 s *= 2;
208 break;
209 case 'M':
210 c++;
211 s *= 1024 * 2;
212 break;
213 case 'G':
214 c++;
215 s *= 1024 * 1024 * 2;
216 break;
72ca9bcf
N
217 case 's': /* sectors */
218 c++;
219 break;
84e11361 220 }
51d4261c 221 } else
822e393a 222 s = INVALID_SECTORS;
84e11361 223 if (*c)
822e393a 224 s = INVALID_SECTORS;
84e11361
N
225 return s;
226}
227
4a06e2c2
N
228int parse_layout_10(char *layout)
229{
230 int copies, rv;
231 char *cp;
232 /* Parse the layout string for raid10 */
233 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
234 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
235 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
236 copies > 200 ||
237 *cp)
238 return -1;
239 if (layout[0] == 'n')
240 rv = 256 + copies;
241 else if (layout[0] == 'o')
242 rv = 0x10000 + (copies<<8) + 1;
243 else
244 rv = 1 + (copies<<8);
245 return rv;
246}
247
248int parse_layout_faulty(char *layout)
249{
250 /* Parse the layout string for 'faulty' */
251 int ln = strcspn(layout, "0123456789");
503975b9 252 char *m = xstrdup(layout);
4a06e2c2
N
253 int mode;
254 m[ln] = 0;
255 mode = map_name(faultylayout, m);
256 if (mode == UnSet)
257 return -1;
258
259 return mode | (atoi(layout+ln)<< ModeShift);
260}
79868890
N
261
262long parse_num(char *num)
263{
264 /* Either return a valid number, or -1 */
265 char *c;
266 long rv = strtol(num, &c, 10);
267 if (rv < 0 || *c || !num[0])
268 return -1;
269 else
270 return rv;
271}
eb3929a4 272#endif
4a06e2c2 273
0430ed48
NB
274void remove_partitions(int fd)
275{
276 /* remove partitions from this block devices.
277 * This is used for components added to an array
278 */
279#ifdef BLKPG_DEL_PARTITION
280 struct blkpg_ioctl_arg a;
281 struct blkpg_partition p;
282
283 a.op = BLKPG_DEL_PARTITION;
284 a.data = (void*)&p;
285 a.datalen = sizeof(p);
286 a.flags = 0;
287 memset(a.data, 0, a.datalen);
ca3b6696 288 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
289 ioctl(fd, BLKPG, &a);
290#endif
291}
292
691c6ee1
N
293int test_partition(int fd)
294{
295 /* Check if fd is a whole-disk or a partition.
296 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
297 * will return ENXIO on an invalid partition number.
298 */
299 struct blkpg_ioctl_arg a;
300 struct blkpg_partition p;
301 a.op = BLKPG_DEL_PARTITION;
302 a.data = (void*)&p;
303 a.datalen = sizeof(p);
304 a.flags = 0;
305 memset(a.data, 0, a.datalen);
306 p.pno = 1<<30;
307 if (ioctl(fd, BLKPG, &a) == 0)
308 /* Very unlikely, but not a partition */
309 return 0;
310 if (errno == ENXIO)
311 /* not a partition */
312 return 0;
313
314 return 1;
315}
316
bfd76b93
CA
317int test_partition_from_id(dev_t id)
318{
319 char buf[20];
320 int fd, rv;
321
322 sprintf(buf, "%d:%d", major(id), minor(id));
323 fd = dev_open(buf, O_RDONLY);
324 if (fd < 0)
325 return -1;
326 rv = test_partition(fd);
327 close(fd);
328 return rv;
329}
691c6ee1 330
de5a472e 331int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 332{
265e0f17 333 int copies, first;
de5a472e
N
334 int i;
335 int avail_disks = 0;
336
337 for (i = 0; i < raid_disks; i++)
338 avail_disks += !!avail[i];
339
64c4757e 340 switch (level) {
265e0f17
NB
341 case 10:
342 /* This is the tricky one - we need to check
343 * which actual disks are present.
344 */
702b557b 345 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 346 first = 0;
265e0f17
NB
347 do {
348 /* there must be one of the 'copies' form 'first' */
349 int n = copies;
ca3b6696 350 int cnt = 0;
b3ec716d 351 int this = first;
265e0f17 352 while (n--) {
b3ec716d 353 if (avail[this])
265e0f17 354 cnt++;
b3ec716d 355 this = (this+1) % raid_disks;
265e0f17
NB
356 }
357 if (cnt == 0)
358 return 0;
b3ec716d 359 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
360 } while (first != 0);
361 return 1;
e5329c37 362
df0d4ea0 363 case LEVEL_MULTIPATH:
e0d19036 364 return avail_disks>= 1;
df0d4ea0 365 case LEVEL_LINEAR:
64c4757e
NB
366 case 0:
367 return avail_disks == raid_disks;
368 case 1:
369 return avail_disks >= 1;
370 case 4:
371 case 5:
583315d9
NB
372 if (clean)
373 return avail_disks >= raid_disks-1;
374 else
375 return avail_disks >= raid_disks;
98c6faba 376 case 6:
583315d9
NB
377 if (clean)
378 return avail_disks >= raid_disks-2;
379 else
380 return avail_disks >= raid_disks;
64c4757e
NB
381 default:
382 return 0;
383 }
384}
385
8453e704
N
386int enough_fd(int fd)
387{
388 struct mdu_array_info_s array;
389 struct mdu_disk_info_s disk;
390 int avail_disks = 0;
aabe020d 391 int i, rv;
8453e704
N
392 char *avail;
393
394 if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 ||
395 array.raid_disks <= 0)
396 return 0;
503975b9 397 avail = xcalloc(array.raid_disks, 1);
ca3b6696 398 for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704
N
399 disk.number = i;
400 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
401 continue;
9e6d9291
N
402 if (disk.major == 0 && disk.minor == 0)
403 continue;
404 array.nr_disks--;
405
8453e704
N
406 if (! (disk.state & (1<<MD_DISK_SYNC)))
407 continue;
408 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
409 continue;
410 avail_disks++;
411 avail[disk.raid_disk] = 1;
412 }
413 /* This is used on an active array, so assume it is clean */
aabe020d 414 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 415 1, avail);
aabe020d
JS
416 free(avail);
417 return rv;
8453e704
N
418}
419
22e263f6
AC
420const int uuid_zero[4] = { 0, 0, 0, 0 };
421
f277ce36 422int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 423{
f277ce36
NB
424 if (swapuuid) {
425 /* parse uuids are hostendian.
426 * uuid's from some superblocks are big-ending
aba69144 427 * if there is a difference, we need to swap..
f277ce36
NB
428 */
429 unsigned char *ac = (unsigned char *)a;
430 unsigned char *bc = (unsigned char *)b;
431 int i;
ca3b6696 432 for (i = 0; i < 16; i += 4) {
f277ce36
NB
433 if (ac[i+0] != bc[i+3] ||
434 ac[i+1] != bc[i+2] ||
435 ac[i+2] != bc[i+1] ||
436 ac[i+3] != bc[i+0])
437 return 0;
438 }
439 return 1;
440 } else {
441 if (a[0]==b[0] &&
442 a[1]==b[1] &&
443 a[2]==b[2] &&
444 a[3]==b[3])
445 return 1;
446 return 0;
447 }
64c4757e 448}
ca3b6696 449
350f29f9
NB
450void copy_uuid(void *a, int b[4], int swapuuid)
451{
452 if (swapuuid) {
453 /* parse uuids are hostendian.
454 * uuid's from some superblocks are big-ending
455 * if there is a difference, we need to swap..
456 */
457 unsigned char *ac = (unsigned char *)a;
458 unsigned char *bc = (unsigned char *)b;
459 int i;
ca3b6696 460 for (i = 0; i < 16; i += 4) {
350f29f9
NB
461 ac[i+0] = bc[i+3];
462 ac[i+1] = bc[i+2];
463 ac[i+2] = bc[i+1];
464 ac[i+3] = bc[i+0];
465 }
466 } else
467 memcpy(a, b, 16);
468}
64c4757e 469
aae5a112 470char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 471{
9968e376 472 int i, j;
d7288ddc
N
473 char uuid[16];
474 char *c = buf;
475 strcpy(c, "UUID-");
476 c += strlen(c);
aae5a112 477 copy_uuid(uuid, id, swap);
9968e376 478 for (i = 0; i < 4; i++) {
9968e376 479 if (i)
ff54de6e 480 *c++ = sep;
9968e376
DW
481 for (j = 3; j >= 0; j--) {
482 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
483 c+= 2;
484 }
d7288ddc
N
485 }
486 return buf;
aae5a112
DW
487
488}
489
490char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
491{
86983cce
N
492 // dirty hack to work around an issue with super1 superblocks...
493 // super1 superblocks need swapuuid set in order for assembly to
494 // work, but can't have it set if we want this printout to match
495 // all the other uuid printouts in super1.c, so we force swapuuid
496 // to 1 to make our printout match the rest of super1
497 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
498}
499
435d4ebb 500#ifndef MDASSEMBLE
682c7051
NB
501int check_ext2(int fd, char *name)
502{
503 /*
504 * Check for an ext2fs file system.
505 * Superblock is always 1K at 1K offset
506 *
507 * s_magic is le16 at 56 == 0xEF53
508 * report mtime - le32 at 44
509 * blocks - le32 at 4
510 * logblksize - le32 at 24
511 */
512 unsigned char sb[1024];
513 time_t mtime;
514 int size, bsize;
515 if (lseek(fd, 1024,0)!= 1024)
516 return 0;
517 if (read(fd, sb, 1024)!= 1024)
518 return 0;
519 if (sb[56] != 0x53 || sb[57] != 0xef)
520 return 0;
521
522 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
523 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
524 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
e7b84f9d 525 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 526 name);
d33f1518 527 cont_err("size=%dK mtime=%s",
682c7051
NB
528 size*(1<<bsize), ctime(&mtime));
529 return 1;
530}
531
532int check_reiser(int fd, char *name)
533{
534 /*
535 * superblock is at 64K
536 * size is 1024;
537 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
538 *
539 */
540 unsigned char sb[1024];
881990a2 541 unsigned long size;
682c7051
NB
542 if (lseek(fd, 64*1024, 0) != 64*1024)
543 return 0;
544 if (read(fd, sb, 1024) != 1024)
545 return 0;
ca3b6696
N
546 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
547 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 548 return 0;
e7b84f9d 549 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 550 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
d33f1518 551 cont_err("size = %luK\n", size*4);
aba69144 552
682c7051
NB
553 return 1;
554}
555
556int check_raid(int fd, char *name)
557{
4b1ac34b 558 struct mdinfo info;
682c7051 559 time_t crtime;
d078d77c 560 char *level;
82d9eba6 561 struct supertype *st = guess_super(fd);
f9ce90ba 562
cb8f6859
N
563 if (!st)
564 return 0;
3da92f27 565 st->ss->load_super(st, fd, name);
82d9eba6 566 /* Looks like a raid array .. */
e7b84f9d 567 pr_err("%s appears to be part of a raid array:\n",
82d9eba6 568 name);
a5d85af7 569 st->ss->getinfo_super(st, &info, NULL);
3da92f27 570 st->ss->free_super(st);
82d9eba6 571 crtime = info.array.ctime;
d078d77c
NB
572 level = map_num(pers, info.array.level);
573 if (!level) level = "-unknown-";
d33f1518
N
574 cont_err("level=%s devices=%d ctime=%s",
575 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 576 return 1;
682c7051
NB
577}
578
682c7051
NB
579int ask(char *mesg)
580{
581 char *add = "";
582 int i;
ca3b6696 583 for (i = 0; i < 5; i++) {
682c7051
NB
584 char buf[100];
585 fprintf(stderr, "%s%s", mesg, add);
586 fflush(stderr);
587 if (fgets(buf, 100, stdin)==NULL)
588 return 0;
589 if (buf[0]=='y' || buf[0]=='Y')
590 return 1;
591 if (buf[0]=='n' || buf[0]=='N')
592 return 0;
593 add = "(y/n) ";
594 }
e7b84f9d 595 pr_err("assuming 'no'\n");
682c7051
NB
596 return 0;
597}
435d4ebb 598#endif /* MDASSEMBLE */
682c7051 599
8d80900b 600int is_standard(char *dev, int *nump)
e5329c37
NB
601{
602 /* tests if dev is a "standard" md dev name.
603 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 604 * where NN is a string of digits
598f0d58
NB
605 * Returns 1 if a partitionable standard,
606 * -1 if non-partitonable,
607 * 0 if not a standard name.
e5329c37 608 */
8d80900b 609 char *d = strrchr(dev, '/');
ca3b6696 610 int type = 0;
8d80900b
NB
611 int num;
612 if (!d)
e5329c37 613 return 0;
ca3b6696
N
614 if (strncmp(d, "/d",2) == 0)
615 d += 2, type = 1; /* /dev/md/dN{pM} */
616 else if (strncmp(d, "/md_d", 5) == 0)
617 d += 5, type = 1; /* /dev/md_dN{pM} */
618 else if (strncmp(d, "/md", 3) == 0)
619 d += 3, type = -1; /* /dev/mdN */
620 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
621 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
622 else
623 return 0;
8d80900b 624 if (!*d)
e5329c37 625 return 0;
8d80900b
NB
626 num = atoi(d);
627 while (isdigit(*d))
628 d++;
629 if (*d)
e5329c37 630 return 0;
8d80900b
NB
631 if (nump) *nump = num;
632
633 return type;
e5329c37
NB
634}
635
4b1ac34b 636unsigned long calc_csum(void *super, int bytes)
82b27616 637{
56eb10c0 638 unsigned long long newcsum = 0;
82b27616 639 int i;
4b1ac34b
NB
640 unsigned int csum;
641 unsigned int *superc = (unsigned int*) super;
82b27616 642
ca3b6696
N
643 for(i = 0; i < bytes/4; i++)
644 newcsum += superc[i];
82b27616 645 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 646#ifdef __alpha__
aba69144 647/* The in-kernel checksum calculation is always 16bit on
570c0542 648 * the alpha, though it is 32 bit on i386...
ca3b6696 649 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
650 * a way that it shouldn't).
651 */
652 csum = (csum & 0xffff) + (csum >> 16);
653 csum = (csum & 0xffff) + (csum >> 16);
654#endif
82b27616
NB
655 return csum;
656}
cd29a5c8 657
435d4ebb 658#ifndef MDASSEMBLE
56eb10c0 659char *human_size(long long bytes)
cd29a5c8
NB
660{
661 static char buf[30];
d5d3721e
NB
662
663 /* We convert bytes to either centi-M{ega,ibi}bytes or
664 * centi-G{igi,ibi}bytes, with appropriate rounding,
665 * and then print 1/100th of those as a decimal.
666 * We allow upto 2048Megabytes before converting to
667 * gigabytes, as that shows more precision and isn't
668 * too large a number.
01619b48 669 * Terabytes are not yet handled.
d5d3721e 670 */
cd29a5c8 671
56eb10c0 672 if (bytes < 5000*1024)
ca3b6696 673 buf[0] = 0;
d5d3721e
NB
674 else if (bytes < 2*1024LL*1024LL*1024LL) {
675 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
676 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 677 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
678 cMiB/100 , cMiB % 100,
679 cMB/100, cMB % 100);
680 } else {
681 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
682 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 683 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
684 cGiB/100 , cGiB % 100,
685 cGB/100, cGB % 100);
686 }
cd29a5c8
NB
687 return buf;
688}
e0d19036 689
f0ec6710 690char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
691{
692 static char buf[30];
e0d19036 693
570abc6f
MN
694 /* We convert bytes to either centi-M{ega,ibi}bytes or
695 * centi-G{igi,ibi}bytes, with appropriate rounding,
696 * and then print 1/100th of those as a decimal.
697 * We allow upto 2048Megabytes before converting to
698 * gigabytes, as that shows more precision and isn't
699 * too large a number.
700 * Terabytes are not yet handled.
f0ec6710
MN
701 *
702 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
703 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
704 */
705
e0d19036 706 if (bytes < 5000*1024)
570abc6f 707 buf[0] = 0;
f0ec6710
MN
708 else if (prefix == IEC) {
709 if (bytes < 2*1024LL*1024LL*1024LL) {
710 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
711 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
712 cMiB/100 , cMiB % 100);
713 } else {
714 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
715 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
716 cGiB/100 , cGiB % 100);
717 }
718 }
719 else if (prefix == JEDEC) {
720 if (bytes < 2*1024LL*1024LL*1024LL) {
721 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
722 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
723 cMB/100, cMB % 100);
724 } else {
725 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
726 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
727 cGB/100 , cGB % 100);
728 }
570abc6f 729 }
f0ec6710
MN
730 else
731 buf[0] = 0;
570abc6f 732
e0d19036
NB
733 return buf;
734}
e4965ef8
N
735
736void print_r10_layout(int layout)
737{
738 int near = layout & 255;
739 int far = (layout >> 8) & 255;
740 int offset = (layout&0x10000);
741 char *sep = "";
742
743 if (near != 1) {
744 printf("%s near=%d", sep, near);
745 sep = ",";
746 }
747 if (far != 1)
748 printf("%s %s=%d", sep, offset?"offset":"far", far);
749 if (near*far == 1)
750 printf("NO REDUNDANCY");
751}
435d4ebb 752#endif
e0d19036 753
5f8097be
NB
754unsigned long long calc_array_size(int level, int raid_disks, int layout,
755 int chunksize, unsigned long long devsize)
577e8448 756{
e03640bd
CA
757 if (level == 1)
758 return devsize;
577e8448
AK
759 devsize &= ~(unsigned long long)((chunksize>>9)-1);
760 return get_data_disks(level, layout, raid_disks) * devsize;
761}
762
763int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
764{
765 int data_disks = 0;
766 switch (level) {
ca3b6696
N
767 case 0: data_disks = raid_disks;
768 break;
769 case 1: data_disks = 1;
770 break;
5f8097be 771 case 4:
ca3b6696
N
772 case 5: data_disks = raid_disks - 1;
773 break;
774 case 6: data_disks = raid_disks - 2;
775 break;
5f8097be
NB
776 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
777 break;
778 }
577e8448
AK
779
780 return data_disks;
5f8097be
NB
781}
782
4dd2df09
N
783int devnm2devid(char *devnm)
784{
785 /* First look in /sys/block/$DEVNM/dev for %d:%d
786 * If that fails, try parsing out a number
787 */
788 char path[100];
789 char *ep;
790 int fd;
791 int mjr,mnr;
792
793 sprintf(path, "/sys/block/%s/dev", devnm);
794 fd = open(path, O_RDONLY);
795 if (fd >= 0) {
796 char buf[20];
797 int n = read(fd, buf, sizeof(buf));
798 close(fd);
799 if (n > 0)
800 buf[n] = 0;
801 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
802 return makedev(mjr, mnr);
803 }
804 if (strncmp(devnm, "md_d", 4) == 0 &&
805 isdigit(devnm[4]) &&
806 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
807 ep > devnm && *ep == 0)
808 return makedev(get_mdp_major(), mnr << MdpMinorShift);
809
810 if (strncmp(devnm, "md", 2) == 0 &&
811 isdigit(devnm[2]) &&
812 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
813 ep > devnm && *ep == 0)
814 return makedev(MD_MAJOR, mnr);
815
816 return 0;
817}
818
5a23a06e 819#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 820char *get_md_name(char *devnm)
e0d19036
NB
821{
822 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 823 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 824
e0d19036
NB
825 static char devname[50];
826 struct stat stb;
4dd2df09 827 dev_t rdev = devnm2devid(devnm);
dd0781e5 828 char *dn;
98c6faba 829
4dd2df09
N
830 if (rdev == 0)
831 return 0;
832 if (strncmp(devnm, "md_", 3) == 0) {
833 snprintf(devname, sizeof(devname), "/dev/md/%s",
834 devnm + 3);
98c6faba
NB
835 if (stat(devname, &stb) == 0
836 && (S_IFMT&stb.st_mode) == S_IFBLK
837 && (stb.st_rdev == rdev))
838 return devname;
839 }
4dd2df09
N
840 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
841 if (stat(devname, &stb) == 0
842 && (S_IFMT&stb.st_mode) == S_IFBLK
843 && (stb.st_rdev == rdev))
844 return devname;
845
846 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
847 if (stat(devname, &stb) == 0
848 && (S_IFMT&stb.st_mode) == S_IFBLK
849 && (stb.st_rdev == rdev))
850 return devname;
851
16c6fa80 852 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
853 if (dn)
854 return dn;
4dd2df09 855 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 856 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
857 if (errno != EEXIST)
858 return NULL;
e0d19036
NB
859
860 if (stat(devname, &stb) == 0
861 && (S_IFMT&stb.st_mode) == S_IFBLK
862 && (stb.st_rdev == rdev))
863 return devname;
864 unlink(devname);
865 return NULL;
866}
867
868void put_md_name(char *name)
869{
ca3b6696 870 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
871 unlink(name);
872}
435d4ebb 873#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 874
8b0dabea
NB
875int dev_open(char *dev, int flags)
876{
877 /* like 'open', but if 'dev' matches %d:%d, create a temp
878 * block device and open that
879 */
880 char *e;
881 int fd = -1;
882 char devname[32];
e81cdd9f 883 int major;
8b0dabea 884 int minor;
e81cdd9f
NB
885
886 if (!dev) return -1;
6df6a774 887 flags |= O_DIRECT;
e81cdd9f
NB
888
889 major = strtoul(dev, &e, 0);
8b0dabea
NB
890 if (e > dev && *e == ':' && e[1] &&
891 (minor = strtoul(e+1, &e, 0)) >= 0 &&
892 *e == 0) {
9dc7d357
N
893 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
894 (int)getpid(), major, minor);
895 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
896 fd = open(devname, flags);
897 unlink(devname);
6df6a774
N
898 }
899 if (fd < 0) {
9dc7d357 900 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
901 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
902 (int)getpid(), major, minor);
ca3b6696 903 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
904 fd = open(devname, flags);
905 unlink(devname);
906 }
8b0dabea
NB
907 }
908 } else
6df6a774 909 fd = open(dev, flags);
8b0dabea
NB
910 return fd;
911}
f9ce90ba 912
4dd2df09 913int open_dev_flags(char *devnm, int flags)
e8a70c89 914{
4dd2df09 915 int devid;
e8a70c89
N
916 char buf[20];
917
4dd2df09
N
918 devid = devnm2devid(devnm);
919 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
920 return dev_open(buf, flags);
921}
922
4dd2df09 923int open_dev(char *devnm)
d998b738 924{
4dd2df09 925 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
926}
927
4dd2df09 928int open_dev_excl(char *devnm)
a931db9e
NB
929{
930 char buf[20];
931 int i;
7187750e 932 int flags = O_RDWR;
4dd2df09 933 int devid = devnm2devid(devnm);
399e0b97 934 long delay = 1000;
a931db9e 935
4dd2df09 936 sprintf(buf, "%d:%d", major(devid), minor(devid));
ca3b6696 937 for (i = 0 ; i < 25 ; i++) {
7187750e 938 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
939 if (fd >= 0)
940 return fd;
7187750e
N
941 if (errno == EACCES && flags == O_RDWR) {
942 flags = O_RDONLY;
943 continue;
944 }
a931db9e
NB
945 if (errno != EBUSY)
946 return fd;
399e0b97
N
947 usleep(delay);
948 if (delay < 200000)
949 delay *= 2;
a931db9e
NB
950 }
951 return -1;
952}
953
9008ed1c
N
954int same_dev(char *one, char *two)
955{
956 struct stat st1, st2;
957 if (stat(one, &st1) != 0)
958 return 0;
959 if (stat(two, &st2) != 0)
960 return 0;
961 if ((st1.st_mode & S_IFMT) != S_IFBLK)
962 return 0;
963 if ((st2.st_mode & S_IFMT) != S_IFBLK)
964 return 0;
965 return st1.st_rdev == st2.st_rdev;
966}
967
a7c6e3fb 968void wait_for(char *dev, int fd)
a714580e
N
969{
970 int i;
a7c6e3fb 971 struct stat stb_want;
399e0b97 972 long delay = 1000;
a7c6e3fb
N
973
974 if (fstat(fd, &stb_want) != 0 ||
975 (stb_want.st_mode & S_IFMT) != S_IFBLK)
976 return;
a714580e 977
ca3b6696 978 for (i = 0 ; i < 25 ; i++) {
a714580e 979 struct stat stb;
a7c6e3fb
N
980 if (stat(dev, &stb) == 0 &&
981 (stb.st_mode & S_IFMT) == S_IFBLK &&
982 (stb.st_rdev == stb_want.st_rdev))
a714580e 983 return;
399e0b97
N
984 usleep(delay);
985 if (delay < 200000)
986 delay *= 2;
a714580e 987 }
436305c6
DW
988 if (i == 25)
989 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
990}
991
0f22b998
N
992struct superswitch *superlist[] =
993{
994 &super0, &super1,
995 &super_ddf, &super_imsm,
0592faeb 996 &mbr, &gpt,
0f22b998 997 NULL };
f9ce90ba 998
ea24acd0 999#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1000
4725bc31 1001struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1002{
1686dc25
NB
1003 mdu_array_info_t array;
1004 int vers;
1005 int minor;
1006 struct supertype *st = NULL;
7e0f6979 1007 struct mdinfo *sra;
142cb9e1 1008 char *verstr;
1686dc25
NB
1009 char version[20];
1010 int i;
f7e7067b 1011 char *subarray = NULL;
4dd2df09 1012 char container[32] = "";
1686dc25 1013
4dd2df09 1014 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1015
1016 if (sra) {
7e0f6979
NB
1017 vers = sra->array.major_version;
1018 minor = sra->array.minor_version;
142cb9e1 1019 verstr = sra->text_version;
1686dc25
NB
1020 } else {
1021 if (ioctl(fd, GET_ARRAY_INFO, &array))
1022 array.major_version = array.minor_version = 0;
1023 vers = array.major_version;
1024 minor = array.minor_version;
142cb9e1 1025 verstr = "";
6fbba4c9 1026 }
82d9eba6 1027
1686dc25
NB
1028 if (vers != -1) {
1029 sprintf(version, "%d.%d", vers, minor);
1030 verstr = version;
6fbba4c9 1031 }
3c558363 1032 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1033 char *dev = verstr+1;
5f7e44b2 1034
f7e7067b 1035 subarray = strchr(dev, '/');
92d49ecf 1036 if (subarray) {
f7e7067b 1037 *subarray++ = '\0';
503975b9 1038 subarray = xstrdup(subarray);
92d49ecf 1039 }
4dd2df09 1040 strcpy(container, dev);
f7e7067b
NB
1041 if (sra)
1042 sysfs_free(sra);
5f7e44b2 1043 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1044 if (sra && sra->text_version[0])
1045 verstr = sra->text_version;
1046 else
1047 verstr = "-no-metadata-";
f7e7067b
NB
1048 }
1049
1050 for (i = 0; st == NULL && superlist[i] ; i++)
1051 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
1052
1053 if (sra)
1054 sysfs_free(sra);
f7e7067b 1055 if (st) {
3b0896f8 1056 st->sb = NULL;
1f49fb3a
N
1057 if (subarrayp)
1058 *subarrayp = subarray;
4dd2df09
N
1059 strcpy(st->container_devnm, container);
1060 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1061 } else
1062 free(subarray);
5f7e44b2 1063
82d9eba6 1064 return st;
f9ce90ba 1065}
ea24acd0
NB
1066#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1067
0f0749ad 1068int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1069{
1070 char buf[20];
1071 int fd;
1072
1073 sprintf(buf, "%d:%d", major(id), minor(id));
1074 fd = dev_open(buf, O_RDONLY);
1075 if (fd < 0)
1076 return 0;
1077 if (get_dev_size(fd, NULL, size)) {
1078 close(fd);
1079 return 1;
1080 }
1081 close(fd);
1082 return 0;
1083}
f9ce90ba 1084
159c3a1a 1085struct supertype *dup_super(struct supertype *orig)
3da92f27 1086{
159c3a1a 1087 struct supertype *st;
1686dc25 1088
d2ca6449
NB
1089 if (!orig)
1090 return orig;
503975b9 1091 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1092 st->ss = orig->ss;
1093 st->max_devs = orig->max_devs;
1094 st->minor_version = orig->minor_version;
cb8f6859 1095 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1096 st->data_offset = orig->data_offset;
159c3a1a
NB
1097 st->sb = NULL;
1098 st->info = NULL;
1099 return st;
3da92f27
NB
1100}
1101
54887ad8 1102struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1103{
1104 /* try each load_super to find the best match,
1105 * and return the best superswitch
1106 */
82d9eba6
NB
1107 struct superswitch *ss;
1108 struct supertype *st;
f21e18ca 1109 time_t besttime = 0;
570c0542 1110 int bestsuper = -1;
f9ce90ba
NB
1111 int i;
1112
503975b9 1113 st = xcalloc(1, sizeof(*st));
4dd2df09 1114 st->container_devnm[0] = 0;
d1d599ea 1115
ca3b6696 1116 for (i = 0 ; superlist[i]; i++) {
f9ce90ba
NB
1117 int rv;
1118 ss = superlist[i];
54887ad8
N
1119 if (guess_type == guess_array && ss->add_to_super == NULL)
1120 continue;
1121 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1122 continue;
ef609477 1123 memset(st, 0, sizeof(*st));
df3346e6 1124 st->ignore_hw_compat = 1;
3da92f27 1125 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1126 if (rv == 0) {
1127 struct mdinfo info;
a5d85af7 1128 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1129 if (bestsuper == -1 ||
1130 besttime < info.array.ctime) {
1131 bestsuper = i;
1132 besttime = info.array.ctime;
570c0542 1133 }
3da92f27 1134 ss->free_super(st);
570c0542
NB
1135 }
1136 }
1137 if (bestsuper != -1) {
1138 int rv;
ef609477 1139 memset(st, 0, sizeof(*st));
df3346e6 1140 st->ignore_hw_compat = 1;
3da92f27 1141 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1142 if (rv == 0) {
5e747af2 1143 superlist[bestsuper]->free_super(st);
82d9eba6 1144 return st;
f9ce90ba
NB
1145 }
1146 }
570c0542 1147 free(st);
f9ce90ba
NB
1148 return NULL;
1149}
fe6729fa 1150
beae1dfe
NB
1151/* Return size of device in bytes */
1152int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1153{
1154 unsigned long long ldsize;
c2c9bb6f
NB
1155 struct stat st;
1156
1157 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1158 ldsize = (unsigned long long)st.st_size;
1159 else
beae1dfe
NB
1160#ifdef BLKGETSIZE64
1161 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1162#endif
1163 {
1164 unsigned long dsize;
1165 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1166 ldsize = dsize;
1167 ldsize <<= 9;
1168 } else {
1169 if (dname)
e7b84f9d 1170 pr_err("Cannot get size of %s: %s\b",
beae1dfe
NB
1171 dname, strerror(errno));
1172 return 0;
1173 }
1174 }
1175 *sizep = ldsize;
1176 return 1;
1177}
8fac0577 1178
3a371610
N
1179/* Return true if this can only be a container, not a member device.
1180 * i.e. is and md device and size is zero
1181 */
1182int must_be_container(int fd)
1183{
1184 unsigned long long size;
1185 if (md_get_version(fd) < 0)
1186 return 0;
1187 if (get_dev_size(fd, NULL, &size) == 0)
1188 return 1;
1189 if (size == 0)
1190 return 1;
1191 return 0;
1192}
034b203a
TM
1193
1194/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1195 * Returns: 1 if successful
1196 * -1 for unknown partition type
1197 * 0 for other errors
1198 */
1199static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1200{
056b331e 1201 struct GPT gpt;
034b203a 1202 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1203 struct GPT_part_entry *part;
1204 char buf[512];
034b203a
TM
1205 unsigned long long curr_part_end;
1206 unsigned all_partitions, entry_size;
f21e18ca 1207 unsigned part_nr;
034b203a
TM
1208
1209 *endofpart = 0;
1210
056b331e 1211 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1212 /* skip protective MBR */
034b203a 1213 lseek(fd, 512, SEEK_SET);
73e658d8 1214 /* read GPT header */
056b331e 1215 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1216 return 0;
1217
1218 /* get the number of partition entries and the entry size */
056b331e
N
1219 all_partitions = __le32_to_cpu(gpt.part_cnt);
1220 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1221
1222 /* Check GPT signature*/
056b331e 1223 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1224 return -1;
1225
1226 /* sanity checks */
1227 if (all_partitions > 1024 ||
db7fdfe4 1228 entry_size > sizeof(buf))
034b203a
TM
1229 return -1;
1230
db7fdfe4
JS
1231 part = (struct GPT_part_entry *)buf;
1232
ca3b6696 1233 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1234 /* read partition entry */
db7fdfe4 1235 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1236 return 0;
1237
034b203a 1238 /* is this valid partition? */
db7fdfe4 1239 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1240 /* check the last lba for the current partition */
db7fdfe4 1241 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1242 if (curr_part_end > *endofpart)
1243 *endofpart = curr_part_end;
1244 }
1245
034b203a
TM
1246 }
1247 return 1;
1248}
1249
1250/* Sets endofpart parameter to the last block used by the last partition on the device.
1251 * Returns: 1 if successful
1252 * -1 for unknown partition type
1253 * 0 for other errors
1254 */
1255static int get_last_partition_end(int fd, unsigned long long *endofpart)
1256{
056b331e 1257 struct MBR boot_sect;
034b203a
TM
1258 struct MBR_part_record *part;
1259 unsigned long long curr_part_end;
f21e18ca 1260 unsigned part_nr;
034b203a
TM
1261 int retval = 0;
1262
1263 *endofpart = 0;
1264
056b331e 1265 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1266 /* read MBR */
1267 lseek(fd, 0, 0);
056b331e 1268 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1269 goto abort;
1270
1271 /* check MBP signature */
056b331e 1272 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1273 retval = 1;
1274 /* found the correct signature */
056b331e 1275 part = boot_sect.parts;
034b203a 1276
ca3b6696 1277 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
034b203a
TM
1278 /* check for GPT type */
1279 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1280 retval = get_gpt_last_partition_end(fd, endofpart);
1281 break;
1282 }
1283 /* check the last used lba for the current partition */
1284 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1285 __le32_to_cpu(part->blocks_num);
1286 if (curr_part_end > *endofpart)
1287 *endofpart = curr_part_end;
1288
1289 part++;
1290 }
1291 } else {
1292 /* Unknown partition table */
1293 retval = -1;
1294 }
1295 abort:
1296 return retval;
1297}
1298
53ed6ac3
KW
1299int check_partitions(int fd, char *dname, unsigned long long freesize,
1300 unsigned long long size)
034b203a
TM
1301{
1302 /*
1303 * Check where the last partition ends
1304 */
1305 unsigned long long endofpart;
1306 int ret;
1307
1308 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1309 /* There appears to be a partition table here */
1310 if (freesize == 0) {
1311 /* partitions will not be visible in new device */
e7b84f9d
N
1312 pr_err("partition table exists on %s but will be lost or\n"
1313 " meaningless after creating array\n",
1314 dname);
034b203a
TM
1315 return 1;
1316 } else if (endofpart > freesize) {
1317 /* last partition overlaps metadata */
e7b84f9d
N
1318 pr_err("metadata will over-write last partition on %s.\n",
1319 dname);
034b203a 1320 return 1;
53ed6ac3
KW
1321 } else if (size && endofpart > size) {
1322 /* partitions will be truncated in new device */
e7b84f9d
N
1323 pr_err("array size is too small to cover all partitions on %s.\n",
1324 dname);
53ed6ac3 1325 return 1;
034b203a
TM
1326 }
1327 }
1328 return 0;
1329}
1330
a322f70c
DW
1331int open_container(int fd)
1332{
1333 /* 'fd' is a block device. Find out if it is in use
1334 * by a container, and return an open fd on that container.
1335 */
1336 char path[256];
1337 char *e;
1338 DIR *dir;
1339 struct dirent *de;
1340 int dfd, n;
1341 char buf[200];
1342 int major, minor;
1343 struct stat st;
1344
1345 if (fstat(fd, &st) != 0)
1346 return -1;
1347 sprintf(path, "/sys/dev/block/%d:%d/holders",
1348 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1349 e = path + strlen(path);
1350
1351 dir = opendir(path);
1352 if (!dir)
1353 return -1;
1354 while ((de = readdir(dir))) {
1355 if (de->d_ino == 0)
1356 continue;
1357 if (de->d_name[0] == '.')
1358 continue;
a994592d
N
1359 /* Need to make sure it is a container and not a volume */
1360 sprintf(e, "/%s/md/metadata_version", de->d_name);
1361 dfd = open(path, O_RDONLY);
1362 if (dfd < 0)
1363 continue;
1364 n = read(dfd, buf, sizeof(buf));
1365 close(dfd);
1366 if (n <= 0 || (unsigned)n >= sizeof(buf))
1367 continue;
1368 buf[n] = 0;
1369 if (strncmp(buf, "external", 8) != 0 ||
1370 n < 10 ||
1371 buf[9] == '/')
1372 continue;
a322f70c
DW
1373 sprintf(e, "/%s/dev", de->d_name);
1374 dfd = open(path, O_RDONLY);
1375 if (dfd < 0)
1376 continue;
1377 n = read(dfd, buf, sizeof(buf));
1378 close(dfd);
f21e18ca 1379 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1380 continue;
1381 buf[n] = 0;
1382 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1383 continue;
1384 sprintf(buf, "%d:%d", major, minor);
1385 dfd = dev_open(buf, O_RDONLY);
1386 if (dfd >= 0) {
1387 closedir(dir);
1388 return dfd;
1389 }
1390 }
355726fa 1391 closedir(dir);
a322f70c
DW
1392 return -1;
1393}
1394
33414a01
DW
1395struct superswitch *version_to_superswitch(char *vers)
1396{
1397 int i;
1398
1399 for (i = 0; superlist[i]; i++) {
1400 struct superswitch *ss = superlist[i];
1401
1402 if (strcmp(vers, ss->name) == 0)
1403 return ss;
1404 }
1405
1406 return NULL;
1407}
1408
4dd2df09
N
1409int metadata_container_matches(char *metadata, char *devnm)
1410{
1411 /* Check if 'devnm' is the container named in 'metadata'
1412 * which is
1413 * /containername/componentname or
1414 * -containername/componentname
1415 */
1416 int l;
1417 if (*metadata != '/' && *metadata != '-')
1418 return 0;
1419 l = strlen(devnm);
1420 if (strncmp(metadata+1, devnm, l) != 0)
1421 return 0;
1422 if (metadata[l+1] != '/')
1423 return 0;
1424 return 1;
1425}
1426
1427int metadata_subdev_matches(char *metadata, char *devnm)
1428{
1429 /* Check if 'devnm' is the subdev named in 'metadata'
1430 * which is
1431 * /containername/subdev or
1432 * -containername/subdev
1433 */
1434 char *sl;
1435 if (*metadata != '/' && *metadata != '-')
1436 return 0;
1437 sl = strchr(metadata+1, '/');
1438 if (!sl)
1439 return 0;
1440 if (strcmp(sl+1, devnm) == 0)
1441 return 1;
1442 return 0;
1443}
1444
33414a01
DW
1445int is_container_member(struct mdstat_ent *mdstat, char *container)
1446{
1447 if (mdstat->metadata_version == NULL ||
1448 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1449 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1450 return 0;
1451
1452 return 1;
1453}
1454
1455int is_subarray_active(char *subarray, char *container)
1456{
1457 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1458 struct mdstat_ent *ent;
1459
e5408a32
DW
1460 for (ent = mdstat; ent; ent = ent->next)
1461 if (is_container_member(ent, container))
e5e5d7ce 1462 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1463 break;
33414a01
DW
1464
1465 free_mdstat(mdstat);
1466
1467 return ent != NULL;
1468}
1469
1470/* open_subarray - opens a subarray in a container
1471 * @dev: container device name
feab51f8 1472 * @st: empty supertype
33414a01
DW
1473 * @quiet: block reporting errors flag
1474 *
1475 * On success returns an fd to a container and fills in *st
1476 */
feab51f8 1477int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1478{
1479 struct mdinfo *mdi;
a951a4f7 1480 struct mdinfo *info;
33414a01 1481 int fd, err = 1;
4dd2df09 1482 char *_devnm;
33414a01
DW
1483
1484 fd = open(dev, O_RDWR|O_EXCL);
1485 if (fd < 0) {
1486 if (!quiet)
e7b84f9d 1487 pr_err("Couldn't open %s, aborting\n",
33414a01 1488 dev);
b990032d 1489 return -1;
33414a01
DW
1490 }
1491
4dd2df09
N
1492 _devnm = fd2devnm(fd);
1493 if (_devnm == NULL) {
33414a01 1494 if (!quiet)
e7b84f9d
N
1495 pr_err("Failed to determine device number for %s\n",
1496 dev);
33414a01
DW
1497 goto close_fd;
1498 }
4dd2df09 1499 strcpy(st->devnm, _devnm);
33414a01 1500
4dd2df09 1501 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1502 if (!mdi) {
1503 if (!quiet)
e7b84f9d 1504 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1505 dev);
1506 goto close_fd;
1507 }
1508
1509 if (mdi->array.level != UnSet) {
1510 if (!quiet)
e7b84f9d 1511 pr_err("%s is not a container\n", dev);
33414a01
DW
1512 goto free_sysfs;
1513 }
1514
1515 st->ss = version_to_superswitch(mdi->text_version);
1516 if (!st->ss) {
1517 if (!quiet)
e7b84f9d
N
1518 pr_err("Operation not supported for %s metadata\n",
1519 mdi->text_version);
33414a01
DW
1520 goto free_sysfs;
1521 }
1522
4dd2df09 1523 if (st->devnm[0] == 0) {
33414a01 1524 if (!quiet)
e7b84f9d 1525 pr_err("Failed to allocate device name\n");
33414a01
DW
1526 goto free_sysfs;
1527 }
1528
db20d413 1529 if (!st->ss->load_container) {
33414a01 1530 if (!quiet)
e7b84f9d 1531 pr_err("%s is not a container\n", dev);
4dd2df09 1532 goto free_sysfs;
33414a01
DW
1533 }
1534
db20d413 1535 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1536 if (!quiet)
e7b84f9d 1537 pr_err("Failed to load metadata for %s\n",
db20d413 1538 dev);
4dd2df09 1539 goto free_sysfs;
33414a01
DW
1540 }
1541
a951a4f7
N
1542 info = st->ss->container_content(st, subarray);
1543 if (!info) {
1544 if (!quiet)
e7b84f9d 1545 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1546 subarray, dev);
1547 goto free_super;
1548 }
1549 free(info);
1550
33414a01
DW
1551 err = 0;
1552
1553 free_super:
1554 if (err)
1555 st->ss->free_super(st);
33414a01
DW
1556 free_sysfs:
1557 sysfs_free(mdi);
1558 close_fd:
1559 if (err)
1560 close(fd);
1561
1562 if (err)
1563 return -1;
1564 else
1565 return fd;
1566}
1567
7801ac20
N
1568int add_disk(int mdfd, struct supertype *st,
1569 struct mdinfo *sra, struct mdinfo *info)
1570{
1571 /* Add a device to an array, in one of 2 ways. */
1572 int rv;
1573#ifndef MDASSEMBLE
1574 if (st->ss->external) {
d23534e4
DW
1575 if (info->disk.state & (1<<MD_DISK_SYNC))
1576 info->recovery_start = MaxSector;
1577 else
1578 info->recovery_start = 0;
2904b26f 1579 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1580 if (! rv) {
1581 struct mdinfo *sd2;
f35f2525
N
1582 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1583 if (sd2 == info)
1584 break;
1585 if (sd2 == NULL) {
503975b9 1586 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1587 *sd2 = *info;
1588 sd2->next = sra->devs;
1589 sra->devs = sd2;
1590 }
7801ac20
N
1591 }
1592 } else
1593#endif
1594 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1595 return rv;
1596}
1597
de6ae750
N
1598int remove_disk(int mdfd, struct supertype *st,
1599 struct mdinfo *sra, struct mdinfo *info)
1600{
1601 int rv;
1602 /* Remove the disk given by 'info' from the array */
1603#ifndef MDASSEMBLE
1604 if (st->ss->external)
1605 rv = sysfs_set_str(sra, info, "slot", "none");
1606 else
1607#endif
1608 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1609 info->disk.minor));
1610 return rv;
1611}
1612
f35f2525
N
1613int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1614{
1615 /* Initialise kernel's knowledge of array.
1616 * This varies between externally managed arrays
1617 * and older kernels
1618 */
1619 int vers = md_get_version(mdfd);
1620 int rv;
1621
1622#ifndef MDASSEMBLE
1623 if (st->ss->external)
1624 rv = sysfs_set_array(info, vers);
1625 else
1626#endif
1627 if ((vers % 100) >= 1) { /* can use different versions */
1628 mdu_array_info_t inf;
1629 memset(&inf, 0, sizeof(inf));
1630 inf.major_version = info->array.major_version;
1631 inf.minor_version = info->array.minor_version;
1632 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1633 } else
1634 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1635 return rv;
1636}
1637
1e5c6983
DW
1638unsigned long long min_recovery_start(struct mdinfo *array)
1639{
1640 /* find the minimum recovery_start in an array for metadata
1641 * formats that only record per-array recovery progress instead
1642 * of per-device
1643 */
1644 unsigned long long recovery_start = MaxSector;
1645 struct mdinfo *d;
1646
1647 for (d = array->devs; d; d = d->next)
1648 recovery_start = min(recovery_start, d->recovery_start);
1649
1650 return recovery_start;
1651}
1652
4dd2df09 1653int mdmon_pid(char *devnm)
a931db9e
NB
1654{
1655 char path[100];
1656 char pid[10];
1657 int fd;
1658 int n;
10013317 1659
4dd2df09 1660 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1661
24f6f99b 1662 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1663
1664 if (fd < 0)
cf556303 1665 return -1;
a931db9e
NB
1666 n = read(fd, pid, 9);
1667 close(fd);
1668 if (n <= 0)
cf556303 1669 return -1;
24f6f99b 1670 return atoi(pid);
a931db9e
NB
1671}
1672
4dd2df09 1673int mdmon_running(char *devnm)
a931db9e 1674{
4dd2df09 1675 int pid = mdmon_pid(devnm);
24f6f99b 1676 if (pid <= 0)
a931db9e 1677 return 0;
24f6f99b 1678 if (kill(pid, 0) == 0)
a931db9e
NB
1679 return 1;
1680 return 0;
1681}
1682
4dd2df09 1683int start_mdmon(char *devnm)
8850ee3e 1684{
1913c325 1685 int i, skipped;
44d2e365 1686 int len;
ca3b6696 1687 pid_t pid;
9fe32043 1688 int status;
44d2e365
N
1689 char pathbuf[1024];
1690 char *paths[4] = {
1691 pathbuf,
1692 "/sbin/mdmon",
1693 "mdmon",
1694 NULL
1695 };
8850ee3e 1696
40ebbb9c 1697 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1698 return 0;
1699
9cf014ec 1700 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1701 if (len > 0) {
1702 char *sl;
1703 pathbuf[len] = 0;
1704 sl = strrchr(pathbuf, '/');
1705 if (sl)
1706 sl++;
1707 else
1708 sl = pathbuf;
1709 strcpy(sl, "mdmon");
1710 } else
1711 pathbuf[0] = '\0';
1712
0f7bdf89
JS
1713 /* First try to run systemctl */
1714 switch(fork()) {
1715 case 0:
1716 /* FIXME yuk. CLOSE_EXEC?? */
1717 skipped = 0;
1718 for (i = 3; skipped < 20; i++)
1719 if (close(i) < 0)
1720 skipped++;
1721 else
1722 skipped = 0;
1723
701d5b4a
N
1724 /* Don't want to see error messages from systemctl.
1725 * If the service doesn't exist, we start mdmon ourselves.
1726 */
1727 close(2);
1728 open("/dev/null", O_WRONLY);
0f7bdf89 1729 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
4dd2df09 1730 devnm);
0f7bdf89
JS
1731 status = execl("/usr/bin/systemctl", "systemctl", "start",
1732 pathbuf, NULL);
1733 status = execl("/bin/systemctl", "systemctl", "start",
1734 pathbuf, NULL);
1735 exit(1);
1736 case -1: pr_err("cannot run mdmon. "
1737 "Array remains readonly\n");
1738 return -1;
1739 default: /* parent - good */
1740 pid = wait(&status);
1741 if (pid >= 0 && status == 0)
1742 return 0;
1743 }
1744
1745 /* That failed, try running mdmon directly */
8850ee3e
N
1746 switch(fork()) {
1747 case 0:
1748 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1749 skipped = 0;
ca3b6696 1750 for (i = 3; skipped < 20; i++)
1913c325
N
1751 if (close(i) < 0)
1752 skipped++;
1753 else
1754 skipped = 0;
1755
ca3b6696 1756 for (i = 0; paths[i]; i++)
a0963a86 1757 if (paths[i][0]) {
3e23ba9d 1758 execl(paths[i], "mdmon",
4dd2df09 1759 devnm, NULL);
a0963a86 1760 }
8850ee3e 1761 exit(1);
e7b84f9d 1762 case -1: pr_err("cannot run mdmon. "
8850ee3e
N
1763 "Array remains readonly\n");
1764 return -1;
9fe32043
N
1765 default: /* parent - good */
1766 pid = wait(&status);
15c10423
JS
1767 if (pid < 0 || status != 0) {
1768 pr_err("failed to launch mdmon. "
1769 "Array remains readonly\n");
9fe32043 1770 return -1;
15c10423 1771 }
8850ee3e
N
1772 }
1773 return 0;
1774}
1775
148acb7b
DW
1776__u32 random32(void)
1777{
1778 __u32 rv;
1779 int rfd = open("/dev/urandom", O_RDONLY);
1780 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1781 rv = random();
1782 if (rfd >= 0)
1783 close(rfd);
1784 return rv;
1785}
1786
0e600426 1787#ifndef MDASSEMBLE
edd8d13c
NB
1788int flush_metadata_updates(struct supertype *st)
1789{
1790 int sfd;
1791 if (!st->updates) {
1792 st->update_tail = NULL;
1793 return -1;
1794 }
1795
4dd2df09 1796 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
1797 if (sfd < 0)
1798 return -1;
1799
1800 while (st->updates) {
1801 struct metadata_update *mu = st->updates;
1802 st->updates = mu->next;
1803
1804 send_message(sfd, mu, 0);
1805 wait_reply(sfd, 0);
1806 free(mu->buf);
1807 free(mu);
1808 }
1809 ack(sfd, 0);
1810 wait_reply(sfd, 0);
1811 close(sfd);
1812 st->update_tail = NULL;
1813 return 0;
1814}
1815
1816void append_metadata_update(struct supertype *st, void *buf, int len)
1817{
1818
503975b9 1819 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
1820
1821 mu->buf = buf;
1822 mu->len = len;
1823 mu->space = NULL;
cb23f1f4 1824 mu->space_list = NULL;
edd8d13c
NB
1825 mu->next = NULL;
1826 *st->update_tail = mu;
1827 st->update_tail = &mu->next;
1828}
0e600426 1829#endif /* MDASSEMBLE */
a931db9e 1830
fe6729fa
NB
1831#ifdef __TINYC__
1832/* tinyc doesn't optimize this check in ioctl.h out ... */
1833unsigned int __invalid_size_argument_for_IOC = 0;
1834#endif
1835
6d11ec6f
AK
1836int experimental(void)
1837{
1838 if (check_env("MDADM_EXPERIMENTAL"))
1839 return 1;
1840 else {
e7b84f9d 1841 pr_err("To use this feature MDADM_EXPERIMENTAL"
65c83a80 1842 " environment variable has to be defined.\n");
6d11ec6f
AK
1843 return 0;
1844 }
1845}
1846
326727d9
AC
1847/* Pick all spares matching given criteria from a container
1848 * if min_size == 0 do not check size
1849 * if domlist == NULL do not check domains
1850 * if spare_group given add it to domains of each spare
1851 * metadata allows to test domains using metadata of destination array */
1852struct mdinfo *container_choose_spares(struct supertype *st,
1853 unsigned long long min_size,
1854 struct domainlist *domlist,
1855 char *spare_group,
1856 const char *metadata, int get_one)
1857{
1858 struct mdinfo *d, **dp, *disks = NULL;
1859
1860 /* get list of all disks in container */
1861 if (st->ss->getinfo_super_disks)
1862 disks = st->ss->getinfo_super_disks(st);
1863
1864 if (!disks)
1865 return disks;
1866 /* find spare devices on the list */
1867 dp = &disks->devs;
1868 disks->array.spare_disks = 0;
1869 while (*dp) {
1870 int found = 0;
1871 d = *dp;
1872 if (d->disk.state == 0) {
1873 /* check if size is acceptable */
1874 unsigned long long dev_size;
1875 dev_t dev = makedev(d->disk.major,d->disk.minor);
1876
1877 if (!min_size ||
1878 (dev_size_from_id(dev, &dev_size) &&
1879 dev_size >= min_size))
1880 found = 1;
1881 /* check if domain matches */
1882 if (found && domlist) {
4dd2df09 1883 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
1884 if (spare_group)
1885 pol_add(&pol, pol_domain,
1886 spare_group, NULL);
e5508b36 1887 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1888 found = 0;
1889 dev_policy_free(pol);
1890 }
1891 }
1892 if (found) {
1893 dp = &d->next;
1894 disks->array.spare_disks++;
1895 if (get_one) {
1896 sysfs_free(*dp);
1897 d->next = NULL;
1898 }
1899 } else {
1900 *dp = d->next;
1901 d->next = NULL;
1902 sysfs_free(d);
1903 }
1904 }
1905 return disks;
1906}
9eafa1de
MN
1907
1908/* Checks if paths point to the same device
1909 * Returns 0 if they do.
1910 * Returns 1 if they don't.
1911 * Returns -1 if something went wrong,
1912 * e.g. paths are empty or the files
1913 * they point to don't exist */
1914int compare_paths (char* path1, char* path2)
1915{
1916 struct stat st1,st2;
1917
1918 if (path1 == NULL || path2 == NULL)
1919 return -1;
1920 if (stat(path1,&st1) != 0)
1921 return -1;
1922 if (stat(path2,&st2) != 0)
1923 return -1;
1924 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
1925 return 0;
1926 return 1;
1927}
a7dec3fd
N
1928
1929/* Make sure we can open as many devices as needed */
1930void enable_fds(int devices)
1931{
1932 unsigned int fds = 20 + devices;
1933 struct rlimit lim;
1934 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
1935 || lim.rlim_cur >= fds)
1936 return;
1937 if (lim.rlim_max < fds)
1938 lim.rlim_max = fds;
1939 lim.rlim_cur = fds;
1940 setrlimit(RLIMIT_NOFILE, &lim);
1941}
a9c15847
N
1942
1943int in_initrd(void)
1944{
1945 /* This is based on similar function in systemd. */
1946 struct statfs s;
1947 return statfs("/", &s) >= 0 &&
1948 (s.f_type == TMPFS_MAGIC ||
1949 s.f_type == RAMFS_MAGIC);
1950}