DDF: increase default value for safe_mode_delay to 4000ms
[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;
2538ba2a
N
514 unsigned long long size;
515 int bsize;
682c7051
NB
516 if (lseek(fd, 1024,0)!= 1024)
517 return 0;
518 if (read(fd, sb, 1024)!= 1024)
519 return 0;
520 if (sb[56] != 0x53 || sb[57] != 0xef)
521 return 0;
522
523 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
524 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
525 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 526 size <<= bsize;
e7b84f9d 527 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 528 name);
2538ba2a 529 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
530 return 1;
531}
532
533int check_reiser(int fd, char *name)
534{
535 /*
536 * superblock is at 64K
537 * size is 1024;
538 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
539 *
540 */
541 unsigned char sb[1024];
2538ba2a 542 unsigned long long size;
682c7051
NB
543 if (lseek(fd, 64*1024, 0) != 64*1024)
544 return 0;
545 if (read(fd, sb, 1024) != 1024)
546 return 0;
ca3b6696
N
547 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
548 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 549 return 0;
e7b84f9d 550 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 551 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 552 cont_err("size = %lluK\n", size*4);
aba69144 553
682c7051
NB
554 return 1;
555}
556
557int check_raid(int fd, char *name)
558{
4b1ac34b 559 struct mdinfo info;
682c7051 560 time_t crtime;
d078d77c 561 char *level;
82d9eba6 562 struct supertype *st = guess_super(fd);
f9ce90ba 563
cb8f6859
N
564 if (!st)
565 return 0;
3da92f27 566 st->ss->load_super(st, fd, name);
82d9eba6 567 /* Looks like a raid array .. */
e7b84f9d 568 pr_err("%s appears to be part of a raid array:\n",
82d9eba6 569 name);
a5d85af7 570 st->ss->getinfo_super(st, &info, NULL);
3da92f27 571 st->ss->free_super(st);
82d9eba6 572 crtime = info.array.ctime;
d078d77c
NB
573 level = map_num(pers, info.array.level);
574 if (!level) level = "-unknown-";
d33f1518
N
575 cont_err("level=%s devices=%d ctime=%s",
576 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 577 return 1;
682c7051
NB
578}
579
682c7051
NB
580int ask(char *mesg)
581{
582 char *add = "";
583 int i;
ca3b6696 584 for (i = 0; i < 5; i++) {
682c7051
NB
585 char buf[100];
586 fprintf(stderr, "%s%s", mesg, add);
587 fflush(stderr);
588 if (fgets(buf, 100, stdin)==NULL)
589 return 0;
590 if (buf[0]=='y' || buf[0]=='Y')
591 return 1;
592 if (buf[0]=='n' || buf[0]=='N')
593 return 0;
594 add = "(y/n) ";
595 }
e7b84f9d 596 pr_err("assuming 'no'\n");
682c7051
NB
597 return 0;
598}
435d4ebb 599#endif /* MDASSEMBLE */
682c7051 600
8d80900b 601int is_standard(char *dev, int *nump)
e5329c37
NB
602{
603 /* tests if dev is a "standard" md dev name.
604 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 605 * where NN is a string of digits
598f0d58
NB
606 * Returns 1 if a partitionable standard,
607 * -1 if non-partitonable,
608 * 0 if not a standard name.
e5329c37 609 */
8d80900b 610 char *d = strrchr(dev, '/');
ca3b6696 611 int type = 0;
8d80900b
NB
612 int num;
613 if (!d)
e5329c37 614 return 0;
ca3b6696
N
615 if (strncmp(d, "/d",2) == 0)
616 d += 2, type = 1; /* /dev/md/dN{pM} */
617 else if (strncmp(d, "/md_d", 5) == 0)
618 d += 5, type = 1; /* /dev/md_dN{pM} */
619 else if (strncmp(d, "/md", 3) == 0)
620 d += 3, type = -1; /* /dev/mdN */
621 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
622 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
623 else
624 return 0;
8d80900b 625 if (!*d)
e5329c37 626 return 0;
8d80900b
NB
627 num = atoi(d);
628 while (isdigit(*d))
629 d++;
630 if (*d)
e5329c37 631 return 0;
8d80900b
NB
632 if (nump) *nump = num;
633
634 return type;
e5329c37
NB
635}
636
4b1ac34b 637unsigned long calc_csum(void *super, int bytes)
82b27616 638{
56eb10c0 639 unsigned long long newcsum = 0;
82b27616 640 int i;
4b1ac34b
NB
641 unsigned int csum;
642 unsigned int *superc = (unsigned int*) super;
82b27616 643
ca3b6696
N
644 for(i = 0; i < bytes/4; i++)
645 newcsum += superc[i];
82b27616 646 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 647#ifdef __alpha__
aba69144 648/* The in-kernel checksum calculation is always 16bit on
570c0542 649 * the alpha, though it is 32 bit on i386...
ca3b6696 650 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
651 * a way that it shouldn't).
652 */
653 csum = (csum & 0xffff) + (csum >> 16);
654 csum = (csum & 0xffff) + (csum >> 16);
655#endif
82b27616
NB
656 return csum;
657}
cd29a5c8 658
435d4ebb 659#ifndef MDASSEMBLE
56eb10c0 660char *human_size(long long bytes)
cd29a5c8
NB
661{
662 static char buf[30];
d5d3721e
NB
663
664 /* We convert bytes to either centi-M{ega,ibi}bytes or
665 * centi-G{igi,ibi}bytes, with appropriate rounding,
666 * and then print 1/100th of those as a decimal.
667 * We allow upto 2048Megabytes before converting to
668 * gigabytes, as that shows more precision and isn't
669 * too large a number.
01619b48 670 * Terabytes are not yet handled.
d5d3721e 671 */
cd29a5c8 672
56eb10c0 673 if (bytes < 5000*1024)
ca3b6696 674 buf[0] = 0;
d5d3721e
NB
675 else if (bytes < 2*1024LL*1024LL*1024LL) {
676 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
677 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 678 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
679 cMiB/100 , cMiB % 100,
680 cMB/100, cMB % 100);
681 } else {
682 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
683 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 684 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
685 cGiB/100 , cGiB % 100,
686 cGB/100, cGB % 100);
687 }
cd29a5c8
NB
688 return buf;
689}
e0d19036 690
f0ec6710 691char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
692{
693 static char buf[30];
e0d19036 694
570abc6f
MN
695 /* We convert bytes to either centi-M{ega,ibi}bytes or
696 * centi-G{igi,ibi}bytes, with appropriate rounding,
697 * and then print 1/100th of those as a decimal.
698 * We allow upto 2048Megabytes before converting to
699 * gigabytes, as that shows more precision and isn't
700 * too large a number.
701 * Terabytes are not yet handled.
f0ec6710
MN
702 *
703 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
704 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
705 */
706
e0d19036 707 if (bytes < 5000*1024)
570abc6f 708 buf[0] = 0;
f0ec6710
MN
709 else if (prefix == IEC) {
710 if (bytes < 2*1024LL*1024LL*1024LL) {
711 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
712 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
713 cMiB/100 , cMiB % 100);
714 } else {
715 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
716 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
717 cGiB/100 , cGiB % 100);
718 }
719 }
720 else if (prefix == JEDEC) {
721 if (bytes < 2*1024LL*1024LL*1024LL) {
722 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
723 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
724 cMB/100, cMB % 100);
725 } else {
726 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
727 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
728 cGB/100 , cGB % 100);
729 }
570abc6f 730 }
f0ec6710
MN
731 else
732 buf[0] = 0;
570abc6f 733
e0d19036
NB
734 return buf;
735}
e4965ef8
N
736
737void print_r10_layout(int layout)
738{
739 int near = layout & 255;
740 int far = (layout >> 8) & 255;
741 int offset = (layout&0x10000);
742 char *sep = "";
743
744 if (near != 1) {
745 printf("%s near=%d", sep, near);
746 sep = ",";
747 }
748 if (far != 1)
749 printf("%s %s=%d", sep, offset?"offset":"far", far);
750 if (near*far == 1)
751 printf("NO REDUNDANCY");
752}
435d4ebb 753#endif
e0d19036 754
5f8097be
NB
755unsigned long long calc_array_size(int level, int raid_disks, int layout,
756 int chunksize, unsigned long long devsize)
577e8448 757{
e03640bd
CA
758 if (level == 1)
759 return devsize;
577e8448
AK
760 devsize &= ~(unsigned long long)((chunksize>>9)-1);
761 return get_data_disks(level, layout, raid_disks) * devsize;
762}
763
764int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
765{
766 int data_disks = 0;
767 switch (level) {
ca3b6696
N
768 case 0: data_disks = raid_disks;
769 break;
770 case 1: data_disks = 1;
771 break;
5f8097be 772 case 4:
ca3b6696
N
773 case 5: data_disks = raid_disks - 1;
774 break;
775 case 6: data_disks = raid_disks - 2;
776 break;
5f8097be
NB
777 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
778 break;
779 }
577e8448
AK
780
781 return data_disks;
5f8097be
NB
782}
783
4dd2df09
N
784int devnm2devid(char *devnm)
785{
786 /* First look in /sys/block/$DEVNM/dev for %d:%d
787 * If that fails, try parsing out a number
788 */
789 char path[100];
790 char *ep;
791 int fd;
792 int mjr,mnr;
793
794 sprintf(path, "/sys/block/%s/dev", devnm);
795 fd = open(path, O_RDONLY);
796 if (fd >= 0) {
797 char buf[20];
798 int n = read(fd, buf, sizeof(buf));
799 close(fd);
800 if (n > 0)
801 buf[n] = 0;
802 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
803 return makedev(mjr, mnr);
804 }
805 if (strncmp(devnm, "md_d", 4) == 0 &&
806 isdigit(devnm[4]) &&
807 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
808 ep > devnm && *ep == 0)
809 return makedev(get_mdp_major(), mnr << MdpMinorShift);
810
811 if (strncmp(devnm, "md", 2) == 0 &&
812 isdigit(devnm[2]) &&
813 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
814 ep > devnm && *ep == 0)
815 return makedev(MD_MAJOR, mnr);
816
817 return 0;
818}
819
5a23a06e 820#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 821char *get_md_name(char *devnm)
e0d19036
NB
822{
823 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 824 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 825
e0d19036
NB
826 static char devname[50];
827 struct stat stb;
4dd2df09 828 dev_t rdev = devnm2devid(devnm);
dd0781e5 829 char *dn;
98c6faba 830
4dd2df09
N
831 if (rdev == 0)
832 return 0;
833 if (strncmp(devnm, "md_", 3) == 0) {
834 snprintf(devname, sizeof(devname), "/dev/md/%s",
835 devnm + 3);
98c6faba
NB
836 if (stat(devname, &stb) == 0
837 && (S_IFMT&stb.st_mode) == S_IFBLK
838 && (stb.st_rdev == rdev))
839 return devname;
840 }
4dd2df09
N
841 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
842 if (stat(devname, &stb) == 0
843 && (S_IFMT&stb.st_mode) == S_IFBLK
844 && (stb.st_rdev == rdev))
845 return devname;
846
847 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
848 if (stat(devname, &stb) == 0
849 && (S_IFMT&stb.st_mode) == S_IFBLK
850 && (stb.st_rdev == rdev))
851 return devname;
852
16c6fa80 853 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
854 if (dn)
855 return dn;
4dd2df09 856 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 857 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
858 if (errno != EEXIST)
859 return NULL;
e0d19036
NB
860
861 if (stat(devname, &stb) == 0
862 && (S_IFMT&stb.st_mode) == S_IFBLK
863 && (stb.st_rdev == rdev))
864 return devname;
865 unlink(devname);
866 return NULL;
867}
868
869void put_md_name(char *name)
870{
ca3b6696 871 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
872 unlink(name);
873}
435d4ebb 874#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 875
8b0dabea
NB
876int dev_open(char *dev, int flags)
877{
878 /* like 'open', but if 'dev' matches %d:%d, create a temp
879 * block device and open that
880 */
881 char *e;
882 int fd = -1;
883 char devname[32];
e81cdd9f 884 int major;
8b0dabea 885 int minor;
e81cdd9f
NB
886
887 if (!dev) return -1;
6df6a774 888 flags |= O_DIRECT;
e81cdd9f
NB
889
890 major = strtoul(dev, &e, 0);
8b0dabea
NB
891 if (e > dev && *e == ':' && e[1] &&
892 (minor = strtoul(e+1, &e, 0)) >= 0 &&
893 *e == 0) {
9dc7d357
N
894 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
895 (int)getpid(), major, minor);
896 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
897 fd = open(devname, flags);
898 unlink(devname);
6df6a774
N
899 }
900 if (fd < 0) {
9dc7d357 901 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
902 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
903 (int)getpid(), major, minor);
ca3b6696 904 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
905 fd = open(devname, flags);
906 unlink(devname);
907 }
8b0dabea
NB
908 }
909 } else
6df6a774 910 fd = open(dev, flags);
8b0dabea
NB
911 return fd;
912}
f9ce90ba 913
4dd2df09 914int open_dev_flags(char *devnm, int flags)
e8a70c89 915{
4dd2df09 916 int devid;
e8a70c89
N
917 char buf[20];
918
4dd2df09
N
919 devid = devnm2devid(devnm);
920 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
921 return dev_open(buf, flags);
922}
923
4dd2df09 924int open_dev(char *devnm)
d998b738 925{
4dd2df09 926 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
927}
928
4dd2df09 929int open_dev_excl(char *devnm)
a931db9e
NB
930{
931 char buf[20];
932 int i;
7187750e 933 int flags = O_RDWR;
4dd2df09 934 int devid = devnm2devid(devnm);
399e0b97 935 long delay = 1000;
a931db9e 936
4dd2df09 937 sprintf(buf, "%d:%d", major(devid), minor(devid));
ca3b6696 938 for (i = 0 ; i < 25 ; i++) {
7187750e 939 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
940 if (fd >= 0)
941 return fd;
7187750e
N
942 if (errno == EACCES && flags == O_RDWR) {
943 flags = O_RDONLY;
944 continue;
945 }
a931db9e
NB
946 if (errno != EBUSY)
947 return fd;
399e0b97
N
948 usleep(delay);
949 if (delay < 200000)
950 delay *= 2;
a931db9e
NB
951 }
952 return -1;
953}
954
9008ed1c
N
955int same_dev(char *one, char *two)
956{
957 struct stat st1, st2;
958 if (stat(one, &st1) != 0)
959 return 0;
960 if (stat(two, &st2) != 0)
961 return 0;
962 if ((st1.st_mode & S_IFMT) != S_IFBLK)
963 return 0;
964 if ((st2.st_mode & S_IFMT) != S_IFBLK)
965 return 0;
966 return st1.st_rdev == st2.st_rdev;
967}
968
a7c6e3fb 969void wait_for(char *dev, int fd)
a714580e
N
970{
971 int i;
a7c6e3fb 972 struct stat stb_want;
399e0b97 973 long delay = 1000;
a7c6e3fb
N
974
975 if (fstat(fd, &stb_want) != 0 ||
976 (stb_want.st_mode & S_IFMT) != S_IFBLK)
977 return;
a714580e 978
ca3b6696 979 for (i = 0 ; i < 25 ; i++) {
a714580e 980 struct stat stb;
a7c6e3fb
N
981 if (stat(dev, &stb) == 0 &&
982 (stb.st_mode & S_IFMT) == S_IFBLK &&
983 (stb.st_rdev == stb_want.st_rdev))
a714580e 984 return;
399e0b97
N
985 usleep(delay);
986 if (delay < 200000)
987 delay *= 2;
a714580e 988 }
436305c6
DW
989 if (i == 25)
990 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
991}
992
0f22b998
N
993struct superswitch *superlist[] =
994{
995 &super0, &super1,
996 &super_ddf, &super_imsm,
0592faeb 997 &mbr, &gpt,
0f22b998 998 NULL };
f9ce90ba 999
ea24acd0 1000#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1001
4725bc31 1002struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1003{
1686dc25
NB
1004 mdu_array_info_t array;
1005 int vers;
1006 int minor;
1007 struct supertype *st = NULL;
7e0f6979 1008 struct mdinfo *sra;
142cb9e1 1009 char *verstr;
1686dc25
NB
1010 char version[20];
1011 int i;
f7e7067b 1012 char *subarray = NULL;
4dd2df09 1013 char container[32] = "";
1686dc25 1014
4dd2df09 1015 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1016
1017 if (sra) {
7e0f6979
NB
1018 vers = sra->array.major_version;
1019 minor = sra->array.minor_version;
142cb9e1 1020 verstr = sra->text_version;
1686dc25
NB
1021 } else {
1022 if (ioctl(fd, GET_ARRAY_INFO, &array))
1023 array.major_version = array.minor_version = 0;
1024 vers = array.major_version;
1025 minor = array.minor_version;
142cb9e1 1026 verstr = "";
6fbba4c9 1027 }
82d9eba6 1028
1686dc25
NB
1029 if (vers != -1) {
1030 sprintf(version, "%d.%d", vers, minor);
1031 verstr = version;
6fbba4c9 1032 }
3c558363 1033 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1034 char *dev = verstr+1;
5f7e44b2 1035
f7e7067b 1036 subarray = strchr(dev, '/');
92d49ecf 1037 if (subarray) {
f7e7067b 1038 *subarray++ = '\0';
503975b9 1039 subarray = xstrdup(subarray);
92d49ecf 1040 }
4dd2df09 1041 strcpy(container, dev);
f7e7067b
NB
1042 if (sra)
1043 sysfs_free(sra);
5f7e44b2 1044 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1045 if (sra && sra->text_version[0])
1046 verstr = sra->text_version;
1047 else
1048 verstr = "-no-metadata-";
f7e7067b
NB
1049 }
1050
1051 for (i = 0; st == NULL && superlist[i] ; i++)
1052 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
1053
1054 if (sra)
1055 sysfs_free(sra);
f7e7067b 1056 if (st) {
3b0896f8 1057 st->sb = NULL;
1f49fb3a
N
1058 if (subarrayp)
1059 *subarrayp = subarray;
4dd2df09
N
1060 strcpy(st->container_devnm, container);
1061 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1062 } else
1063 free(subarray);
5f7e44b2 1064
82d9eba6 1065 return st;
f9ce90ba 1066}
ea24acd0
NB
1067#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1068
0f0749ad 1069int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1070{
1071 char buf[20];
1072 int fd;
1073
1074 sprintf(buf, "%d:%d", major(id), minor(id));
1075 fd = dev_open(buf, O_RDONLY);
1076 if (fd < 0)
1077 return 0;
1078 if (get_dev_size(fd, NULL, size)) {
1079 close(fd);
1080 return 1;
1081 }
1082 close(fd);
1083 return 0;
1084}
f9ce90ba 1085
159c3a1a 1086struct supertype *dup_super(struct supertype *orig)
3da92f27 1087{
159c3a1a 1088 struct supertype *st;
1686dc25 1089
d2ca6449
NB
1090 if (!orig)
1091 return orig;
503975b9 1092 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1093 st->ss = orig->ss;
1094 st->max_devs = orig->max_devs;
1095 st->minor_version = orig->minor_version;
cb8f6859 1096 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1097 st->data_offset = orig->data_offset;
159c3a1a
NB
1098 st->sb = NULL;
1099 st->info = NULL;
1100 return st;
3da92f27
NB
1101}
1102
54887ad8 1103struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1104{
1105 /* try each load_super to find the best match,
1106 * and return the best superswitch
1107 */
82d9eba6
NB
1108 struct superswitch *ss;
1109 struct supertype *st;
f21e18ca 1110 time_t besttime = 0;
570c0542 1111 int bestsuper = -1;
f9ce90ba
NB
1112 int i;
1113
503975b9 1114 st = xcalloc(1, sizeof(*st));
4dd2df09 1115 st->container_devnm[0] = 0;
d1d599ea 1116
ca3b6696 1117 for (i = 0 ; superlist[i]; i++) {
f9ce90ba
NB
1118 int rv;
1119 ss = superlist[i];
54887ad8
N
1120 if (guess_type == guess_array && ss->add_to_super == NULL)
1121 continue;
1122 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1123 continue;
ef609477 1124 memset(st, 0, sizeof(*st));
df3346e6 1125 st->ignore_hw_compat = 1;
3da92f27 1126 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1127 if (rv == 0) {
1128 struct mdinfo info;
a5d85af7 1129 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1130 if (bestsuper == -1 ||
1131 besttime < info.array.ctime) {
1132 bestsuper = i;
1133 besttime = info.array.ctime;
570c0542 1134 }
3da92f27 1135 ss->free_super(st);
570c0542
NB
1136 }
1137 }
1138 if (bestsuper != -1) {
1139 int rv;
ef609477 1140 memset(st, 0, sizeof(*st));
df3346e6 1141 st->ignore_hw_compat = 1;
3da92f27 1142 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1143 if (rv == 0) {
5e747af2 1144 superlist[bestsuper]->free_super(st);
82d9eba6 1145 return st;
f9ce90ba
NB
1146 }
1147 }
570c0542 1148 free(st);
f9ce90ba
NB
1149 return NULL;
1150}
fe6729fa 1151
beae1dfe
NB
1152/* Return size of device in bytes */
1153int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1154{
1155 unsigned long long ldsize;
c2c9bb6f
NB
1156 struct stat st;
1157
1158 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1159 ldsize = (unsigned long long)st.st_size;
1160 else
beae1dfe
NB
1161#ifdef BLKGETSIZE64
1162 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1163#endif
1164 {
1165 unsigned long dsize;
1166 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1167 ldsize = dsize;
1168 ldsize <<= 9;
1169 } else {
1170 if (dname)
e7b84f9d 1171 pr_err("Cannot get size of %s: %s\b",
beae1dfe
NB
1172 dname, strerror(errno));
1173 return 0;
1174 }
1175 }
1176 *sizep = ldsize;
1177 return 1;
1178}
8fac0577 1179
3a371610
N
1180/* Return true if this can only be a container, not a member device.
1181 * i.e. is and md device and size is zero
1182 */
1183int must_be_container(int fd)
1184{
1185 unsigned long long size;
1186 if (md_get_version(fd) < 0)
1187 return 0;
1188 if (get_dev_size(fd, NULL, &size) == 0)
1189 return 1;
1190 if (size == 0)
1191 return 1;
1192 return 0;
1193}
034b203a
TM
1194
1195/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1196 * Returns: 1 if successful
1197 * -1 for unknown partition type
1198 * 0 for other errors
1199 */
1200static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1201{
056b331e 1202 struct GPT gpt;
034b203a 1203 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1204 struct GPT_part_entry *part;
1205 char buf[512];
034b203a
TM
1206 unsigned long long curr_part_end;
1207 unsigned all_partitions, entry_size;
f21e18ca 1208 unsigned part_nr;
034b203a
TM
1209
1210 *endofpart = 0;
1211
056b331e 1212 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1213 /* skip protective MBR */
034b203a 1214 lseek(fd, 512, SEEK_SET);
73e658d8 1215 /* read GPT header */
056b331e 1216 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1217 return 0;
1218
1219 /* get the number of partition entries and the entry size */
056b331e
N
1220 all_partitions = __le32_to_cpu(gpt.part_cnt);
1221 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1222
1223 /* Check GPT signature*/
056b331e 1224 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1225 return -1;
1226
1227 /* sanity checks */
1228 if (all_partitions > 1024 ||
db7fdfe4 1229 entry_size > sizeof(buf))
034b203a
TM
1230 return -1;
1231
db7fdfe4
JS
1232 part = (struct GPT_part_entry *)buf;
1233
ca3b6696 1234 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1235 /* read partition entry */
db7fdfe4 1236 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1237 return 0;
1238
034b203a 1239 /* is this valid partition? */
db7fdfe4 1240 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1241 /* check the last lba for the current partition */
db7fdfe4 1242 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1243 if (curr_part_end > *endofpart)
1244 *endofpart = curr_part_end;
1245 }
1246
034b203a
TM
1247 }
1248 return 1;
1249}
1250
1251/* Sets endofpart parameter to the last block used by the last partition on the device.
1252 * Returns: 1 if successful
1253 * -1 for unknown partition type
1254 * 0 for other errors
1255 */
1256static int get_last_partition_end(int fd, unsigned long long *endofpart)
1257{
056b331e 1258 struct MBR boot_sect;
034b203a
TM
1259 struct MBR_part_record *part;
1260 unsigned long long curr_part_end;
f21e18ca 1261 unsigned part_nr;
034b203a
TM
1262 int retval = 0;
1263
1264 *endofpart = 0;
1265
056b331e 1266 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1267 /* read MBR */
1268 lseek(fd, 0, 0);
056b331e 1269 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1270 goto abort;
1271
1272 /* check MBP signature */
056b331e 1273 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1274 retval = 1;
1275 /* found the correct signature */
056b331e 1276 part = boot_sect.parts;
034b203a 1277
ca3b6696 1278 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
034b203a
TM
1279 /* check for GPT type */
1280 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1281 retval = get_gpt_last_partition_end(fd, endofpart);
1282 break;
1283 }
1284 /* check the last used lba for the current partition */
1285 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1286 __le32_to_cpu(part->blocks_num);
1287 if (curr_part_end > *endofpart)
1288 *endofpart = curr_part_end;
1289
1290 part++;
1291 }
1292 } else {
1293 /* Unknown partition table */
1294 retval = -1;
1295 }
1296 abort:
1297 return retval;
1298}
1299
53ed6ac3
KW
1300int check_partitions(int fd, char *dname, unsigned long long freesize,
1301 unsigned long long size)
034b203a
TM
1302{
1303 /*
1304 * Check where the last partition ends
1305 */
1306 unsigned long long endofpart;
1307 int ret;
1308
1309 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1310 /* There appears to be a partition table here */
1311 if (freesize == 0) {
1312 /* partitions will not be visible in new device */
e7b84f9d
N
1313 pr_err("partition table exists on %s but will be lost or\n"
1314 " meaningless after creating array\n",
1315 dname);
034b203a
TM
1316 return 1;
1317 } else if (endofpart > freesize) {
1318 /* last partition overlaps metadata */
e7b84f9d
N
1319 pr_err("metadata will over-write last partition on %s.\n",
1320 dname);
034b203a 1321 return 1;
53ed6ac3
KW
1322 } else if (size && endofpart > size) {
1323 /* partitions will be truncated in new device */
e7b84f9d
N
1324 pr_err("array size is too small to cover all partitions on %s.\n",
1325 dname);
53ed6ac3 1326 return 1;
034b203a
TM
1327 }
1328 }
1329 return 0;
1330}
1331
a322f70c
DW
1332int open_container(int fd)
1333{
1334 /* 'fd' is a block device. Find out if it is in use
1335 * by a container, and return an open fd on that container.
1336 */
1337 char path[256];
1338 char *e;
1339 DIR *dir;
1340 struct dirent *de;
1341 int dfd, n;
1342 char buf[200];
1343 int major, minor;
1344 struct stat st;
1345
1346 if (fstat(fd, &st) != 0)
1347 return -1;
1348 sprintf(path, "/sys/dev/block/%d:%d/holders",
1349 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1350 e = path + strlen(path);
1351
1352 dir = opendir(path);
1353 if (!dir)
1354 return -1;
1355 while ((de = readdir(dir))) {
1356 if (de->d_ino == 0)
1357 continue;
1358 if (de->d_name[0] == '.')
1359 continue;
a994592d
N
1360 /* Need to make sure it is a container and not a volume */
1361 sprintf(e, "/%s/md/metadata_version", de->d_name);
1362 dfd = open(path, O_RDONLY);
1363 if (dfd < 0)
1364 continue;
1365 n = read(dfd, buf, sizeof(buf));
1366 close(dfd);
1367 if (n <= 0 || (unsigned)n >= sizeof(buf))
1368 continue;
1369 buf[n] = 0;
1370 if (strncmp(buf, "external", 8) != 0 ||
1371 n < 10 ||
1372 buf[9] == '/')
1373 continue;
a322f70c
DW
1374 sprintf(e, "/%s/dev", de->d_name);
1375 dfd = open(path, O_RDONLY);
1376 if (dfd < 0)
1377 continue;
1378 n = read(dfd, buf, sizeof(buf));
1379 close(dfd);
f21e18ca 1380 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1381 continue;
1382 buf[n] = 0;
1383 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1384 continue;
1385 sprintf(buf, "%d:%d", major, minor);
1386 dfd = dev_open(buf, O_RDONLY);
1387 if (dfd >= 0) {
1388 closedir(dir);
1389 return dfd;
1390 }
1391 }
355726fa 1392 closedir(dir);
a322f70c
DW
1393 return -1;
1394}
1395
33414a01
DW
1396struct superswitch *version_to_superswitch(char *vers)
1397{
1398 int i;
1399
1400 for (i = 0; superlist[i]; i++) {
1401 struct superswitch *ss = superlist[i];
1402
1403 if (strcmp(vers, ss->name) == 0)
1404 return ss;
1405 }
1406
1407 return NULL;
1408}
1409
4dd2df09
N
1410int metadata_container_matches(char *metadata, char *devnm)
1411{
1412 /* Check if 'devnm' is the container named in 'metadata'
1413 * which is
1414 * /containername/componentname or
1415 * -containername/componentname
1416 */
1417 int l;
1418 if (*metadata != '/' && *metadata != '-')
1419 return 0;
1420 l = strlen(devnm);
1421 if (strncmp(metadata+1, devnm, l) != 0)
1422 return 0;
1423 if (metadata[l+1] != '/')
1424 return 0;
1425 return 1;
1426}
1427
1428int metadata_subdev_matches(char *metadata, char *devnm)
1429{
1430 /* Check if 'devnm' is the subdev named in 'metadata'
1431 * which is
1432 * /containername/subdev or
1433 * -containername/subdev
1434 */
1435 char *sl;
1436 if (*metadata != '/' && *metadata != '-')
1437 return 0;
1438 sl = strchr(metadata+1, '/');
1439 if (!sl)
1440 return 0;
1441 if (strcmp(sl+1, devnm) == 0)
1442 return 1;
1443 return 0;
1444}
1445
33414a01
DW
1446int is_container_member(struct mdstat_ent *mdstat, char *container)
1447{
1448 if (mdstat->metadata_version == NULL ||
1449 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1450 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1451 return 0;
1452
1453 return 1;
1454}
1455
1456int is_subarray_active(char *subarray, char *container)
1457{
1458 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1459 struct mdstat_ent *ent;
1460
e5408a32
DW
1461 for (ent = mdstat; ent; ent = ent->next)
1462 if (is_container_member(ent, container))
e5e5d7ce 1463 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1464 break;
33414a01
DW
1465
1466 free_mdstat(mdstat);
1467
1468 return ent != NULL;
1469}
1470
1471/* open_subarray - opens a subarray in a container
1472 * @dev: container device name
feab51f8 1473 * @st: empty supertype
33414a01
DW
1474 * @quiet: block reporting errors flag
1475 *
1476 * On success returns an fd to a container and fills in *st
1477 */
feab51f8 1478int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1479{
1480 struct mdinfo *mdi;
a951a4f7 1481 struct mdinfo *info;
33414a01 1482 int fd, err = 1;
4dd2df09 1483 char *_devnm;
33414a01
DW
1484
1485 fd = open(dev, O_RDWR|O_EXCL);
1486 if (fd < 0) {
1487 if (!quiet)
e7b84f9d 1488 pr_err("Couldn't open %s, aborting\n",
33414a01 1489 dev);
b990032d 1490 return -1;
33414a01
DW
1491 }
1492
4dd2df09
N
1493 _devnm = fd2devnm(fd);
1494 if (_devnm == NULL) {
33414a01 1495 if (!quiet)
e7b84f9d
N
1496 pr_err("Failed to determine device number for %s\n",
1497 dev);
33414a01
DW
1498 goto close_fd;
1499 }
4dd2df09 1500 strcpy(st->devnm, _devnm);
33414a01 1501
4dd2df09 1502 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1503 if (!mdi) {
1504 if (!quiet)
e7b84f9d 1505 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1506 dev);
1507 goto close_fd;
1508 }
1509
1510 if (mdi->array.level != UnSet) {
1511 if (!quiet)
e7b84f9d 1512 pr_err("%s is not a container\n", dev);
33414a01
DW
1513 goto free_sysfs;
1514 }
1515
1516 st->ss = version_to_superswitch(mdi->text_version);
1517 if (!st->ss) {
1518 if (!quiet)
e7b84f9d
N
1519 pr_err("Operation not supported for %s metadata\n",
1520 mdi->text_version);
33414a01
DW
1521 goto free_sysfs;
1522 }
1523
4dd2df09 1524 if (st->devnm[0] == 0) {
33414a01 1525 if (!quiet)
e7b84f9d 1526 pr_err("Failed to allocate device name\n");
33414a01
DW
1527 goto free_sysfs;
1528 }
1529
db20d413 1530 if (!st->ss->load_container) {
33414a01 1531 if (!quiet)
e7b84f9d 1532 pr_err("%s is not a container\n", dev);
4dd2df09 1533 goto free_sysfs;
33414a01
DW
1534 }
1535
db20d413 1536 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1537 if (!quiet)
e7b84f9d 1538 pr_err("Failed to load metadata for %s\n",
db20d413 1539 dev);
4dd2df09 1540 goto free_sysfs;
33414a01
DW
1541 }
1542
a951a4f7
N
1543 info = st->ss->container_content(st, subarray);
1544 if (!info) {
1545 if (!quiet)
e7b84f9d 1546 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1547 subarray, dev);
1548 goto free_super;
1549 }
1550 free(info);
1551
33414a01
DW
1552 err = 0;
1553
1554 free_super:
1555 if (err)
1556 st->ss->free_super(st);
33414a01
DW
1557 free_sysfs:
1558 sysfs_free(mdi);
1559 close_fd:
1560 if (err)
1561 close(fd);
1562
1563 if (err)
1564 return -1;
1565 else
1566 return fd;
1567}
1568
7801ac20
N
1569int add_disk(int mdfd, struct supertype *st,
1570 struct mdinfo *sra, struct mdinfo *info)
1571{
1572 /* Add a device to an array, in one of 2 ways. */
1573 int rv;
1574#ifndef MDASSEMBLE
1575 if (st->ss->external) {
d23534e4
DW
1576 if (info->disk.state & (1<<MD_DISK_SYNC))
1577 info->recovery_start = MaxSector;
1578 else
1579 info->recovery_start = 0;
2904b26f 1580 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1581 if (! rv) {
1582 struct mdinfo *sd2;
f35f2525
N
1583 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1584 if (sd2 == info)
1585 break;
1586 if (sd2 == NULL) {
503975b9 1587 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1588 *sd2 = *info;
1589 sd2->next = sra->devs;
1590 sra->devs = sd2;
1591 }
7801ac20
N
1592 }
1593 } else
1594#endif
1595 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1596 return rv;
1597}
1598
de6ae750
N
1599int remove_disk(int mdfd, struct supertype *st,
1600 struct mdinfo *sra, struct mdinfo *info)
1601{
1602 int rv;
1603 /* Remove the disk given by 'info' from the array */
1604#ifndef MDASSEMBLE
1605 if (st->ss->external)
1606 rv = sysfs_set_str(sra, info, "slot", "none");
1607 else
1608#endif
1609 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1610 info->disk.minor));
1611 return rv;
1612}
1613
f35f2525
N
1614int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1615{
1616 /* Initialise kernel's knowledge of array.
1617 * This varies between externally managed arrays
1618 * and older kernels
1619 */
1620 int vers = md_get_version(mdfd);
1621 int rv;
1622
1623#ifndef MDASSEMBLE
1624 if (st->ss->external)
1625 rv = sysfs_set_array(info, vers);
1626 else
1627#endif
1628 if ((vers % 100) >= 1) { /* can use different versions */
1629 mdu_array_info_t inf;
1630 memset(&inf, 0, sizeof(inf));
1631 inf.major_version = info->array.major_version;
1632 inf.minor_version = info->array.minor_version;
1633 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1634 } else
1635 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1636 return rv;
1637}
1638
1e5c6983
DW
1639unsigned long long min_recovery_start(struct mdinfo *array)
1640{
1641 /* find the minimum recovery_start in an array for metadata
1642 * formats that only record per-array recovery progress instead
1643 * of per-device
1644 */
1645 unsigned long long recovery_start = MaxSector;
1646 struct mdinfo *d;
1647
1648 for (d = array->devs; d; d = d->next)
1649 recovery_start = min(recovery_start, d->recovery_start);
1650
1651 return recovery_start;
1652}
1653
4dd2df09 1654int mdmon_pid(char *devnm)
a931db9e
NB
1655{
1656 char path[100];
1657 char pid[10];
1658 int fd;
1659 int n;
10013317 1660
4dd2df09 1661 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1662
24f6f99b 1663 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1664
1665 if (fd < 0)
cf556303 1666 return -1;
a931db9e
NB
1667 n = read(fd, pid, 9);
1668 close(fd);
1669 if (n <= 0)
cf556303 1670 return -1;
24f6f99b 1671 return atoi(pid);
a931db9e
NB
1672}
1673
4dd2df09 1674int mdmon_running(char *devnm)
a931db9e 1675{
4dd2df09 1676 int pid = mdmon_pid(devnm);
24f6f99b 1677 if (pid <= 0)
a931db9e 1678 return 0;
24f6f99b 1679 if (kill(pid, 0) == 0)
a931db9e
NB
1680 return 1;
1681 return 0;
1682}
1683
4dd2df09 1684int start_mdmon(char *devnm)
8850ee3e 1685{
1913c325 1686 int i, skipped;
44d2e365 1687 int len;
ca3b6696 1688 pid_t pid;
9fe32043 1689 int status;
44d2e365
N
1690 char pathbuf[1024];
1691 char *paths[4] = {
1692 pathbuf,
1693 "/sbin/mdmon",
1694 "mdmon",
1695 NULL
1696 };
8850ee3e 1697
40ebbb9c 1698 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1699 return 0;
1700
9cf014ec 1701 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1702 if (len > 0) {
1703 char *sl;
1704 pathbuf[len] = 0;
1705 sl = strrchr(pathbuf, '/');
1706 if (sl)
1707 sl++;
1708 else
1709 sl = pathbuf;
1710 strcpy(sl, "mdmon");
1711 } else
1712 pathbuf[0] = '\0';
1713
0f7bdf89 1714 /* First try to run systemctl */
9540cc24
N
1715 if (!check_env("MDADM_NO_SYSTEMCTL"))
1716 switch(fork()) {
1717 case 0:
1718 /* FIXME yuk. CLOSE_EXEC?? */
1719 skipped = 0;
1720 for (i = 3; skipped < 20; i++)
1721 if (close(i) < 0)
1722 skipped++;
1723 else
1724 skipped = 0;
1725
1726 /* Don't want to see error messages from
1727 * systemctl. If the service doesn't exist,
1728 * we start mdmon ourselves.
1729 */
1730 close(2);
1731 open("/dev/null", O_WRONLY);
1732 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1733 devnm);
1734 status = execl("/usr/bin/systemctl", "systemctl",
1735 "start",
1736 pathbuf, NULL);
1737 status = execl("/bin/systemctl", "systemctl", "start",
1738 pathbuf, NULL);
1739 exit(1);
1740 case -1: pr_err("cannot run mdmon. "
1741 "Array remains readonly\n");
1742 return -1;
1743 default: /* parent - good */
1744 pid = wait(&status);
1745 if (pid >= 0 && status == 0)
1746 return 0;
1747 }
0f7bdf89
JS
1748
1749 /* That failed, try running mdmon directly */
8850ee3e
N
1750 switch(fork()) {
1751 case 0:
1752 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1753 skipped = 0;
ca3b6696 1754 for (i = 3; skipped < 20; i++)
1913c325
N
1755 if (close(i) < 0)
1756 skipped++;
1757 else
1758 skipped = 0;
1759
ca3b6696 1760 for (i = 0; paths[i]; i++)
a0963a86 1761 if (paths[i][0]) {
3e23ba9d 1762 execl(paths[i], "mdmon",
4dd2df09 1763 devnm, NULL);
a0963a86 1764 }
8850ee3e 1765 exit(1);
e7b84f9d 1766 case -1: pr_err("cannot run mdmon. "
8850ee3e
N
1767 "Array remains readonly\n");
1768 return -1;
9fe32043
N
1769 default: /* parent - good */
1770 pid = wait(&status);
15c10423
JS
1771 if (pid < 0 || status != 0) {
1772 pr_err("failed to launch mdmon. "
1773 "Array remains readonly\n");
9fe32043 1774 return -1;
15c10423 1775 }
8850ee3e
N
1776 }
1777 return 0;
1778}
1779
148acb7b
DW
1780__u32 random32(void)
1781{
1782 __u32 rv;
1783 int rfd = open("/dev/urandom", O_RDONLY);
1784 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1785 rv = random();
1786 if (rfd >= 0)
1787 close(rfd);
1788 return rv;
1789}
1790
0e600426 1791#ifndef MDASSEMBLE
edd8d13c
NB
1792int flush_metadata_updates(struct supertype *st)
1793{
1794 int sfd;
1795 if (!st->updates) {
1796 st->update_tail = NULL;
1797 return -1;
1798 }
1799
4dd2df09 1800 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
1801 if (sfd < 0)
1802 return -1;
1803
1804 while (st->updates) {
1805 struct metadata_update *mu = st->updates;
1806 st->updates = mu->next;
1807
1808 send_message(sfd, mu, 0);
1809 wait_reply(sfd, 0);
1810 free(mu->buf);
1811 free(mu);
1812 }
1813 ack(sfd, 0);
1814 wait_reply(sfd, 0);
1815 close(sfd);
1816 st->update_tail = NULL;
1817 return 0;
1818}
1819
1820void append_metadata_update(struct supertype *st, void *buf, int len)
1821{
1822
503975b9 1823 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
1824
1825 mu->buf = buf;
1826 mu->len = len;
1827 mu->space = NULL;
cb23f1f4 1828 mu->space_list = NULL;
edd8d13c
NB
1829 mu->next = NULL;
1830 *st->update_tail = mu;
1831 st->update_tail = &mu->next;
1832}
0e600426 1833#endif /* MDASSEMBLE */
a931db9e 1834
fe6729fa
NB
1835#ifdef __TINYC__
1836/* tinyc doesn't optimize this check in ioctl.h out ... */
1837unsigned int __invalid_size_argument_for_IOC = 0;
1838#endif
1839
6d11ec6f
AK
1840int experimental(void)
1841{
1842 if (check_env("MDADM_EXPERIMENTAL"))
1843 return 1;
1844 else {
e7b84f9d 1845 pr_err("To use this feature MDADM_EXPERIMENTAL"
65c83a80 1846 " environment variable has to be defined.\n");
6d11ec6f
AK
1847 return 0;
1848 }
1849}
1850
326727d9
AC
1851/* Pick all spares matching given criteria from a container
1852 * if min_size == 0 do not check size
1853 * if domlist == NULL do not check domains
1854 * if spare_group given add it to domains of each spare
1855 * metadata allows to test domains using metadata of destination array */
1856struct mdinfo *container_choose_spares(struct supertype *st,
1857 unsigned long long min_size,
1858 struct domainlist *domlist,
1859 char *spare_group,
1860 const char *metadata, int get_one)
1861{
1862 struct mdinfo *d, **dp, *disks = NULL;
1863
1864 /* get list of all disks in container */
1865 if (st->ss->getinfo_super_disks)
1866 disks = st->ss->getinfo_super_disks(st);
1867
1868 if (!disks)
1869 return disks;
1870 /* find spare devices on the list */
1871 dp = &disks->devs;
1872 disks->array.spare_disks = 0;
1873 while (*dp) {
1874 int found = 0;
1875 d = *dp;
1876 if (d->disk.state == 0) {
1877 /* check if size is acceptable */
1878 unsigned long long dev_size;
1879 dev_t dev = makedev(d->disk.major,d->disk.minor);
1880
1881 if (!min_size ||
1882 (dev_size_from_id(dev, &dev_size) &&
1883 dev_size >= min_size))
1884 found = 1;
1885 /* check if domain matches */
1886 if (found && domlist) {
4dd2df09 1887 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
1888 if (spare_group)
1889 pol_add(&pol, pol_domain,
1890 spare_group, NULL);
e5508b36 1891 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1892 found = 0;
1893 dev_policy_free(pol);
1894 }
1895 }
1896 if (found) {
1897 dp = &d->next;
1898 disks->array.spare_disks++;
1899 if (get_one) {
1900 sysfs_free(*dp);
1901 d->next = NULL;
1902 }
1903 } else {
1904 *dp = d->next;
1905 d->next = NULL;
1906 sysfs_free(d);
1907 }
1908 }
1909 return disks;
1910}
9eafa1de
MN
1911
1912/* Checks if paths point to the same device
1913 * Returns 0 if they do.
1914 * Returns 1 if they don't.
1915 * Returns -1 if something went wrong,
1916 * e.g. paths are empty or the files
1917 * they point to don't exist */
1918int compare_paths (char* path1, char* path2)
1919{
1920 struct stat st1,st2;
1921
1922 if (path1 == NULL || path2 == NULL)
1923 return -1;
1924 if (stat(path1,&st1) != 0)
1925 return -1;
1926 if (stat(path2,&st2) != 0)
1927 return -1;
1928 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
1929 return 0;
1930 return 1;
1931}
a7dec3fd
N
1932
1933/* Make sure we can open as many devices as needed */
1934void enable_fds(int devices)
1935{
1936 unsigned int fds = 20 + devices;
1937 struct rlimit lim;
1938 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
1939 || lim.rlim_cur >= fds)
1940 return;
1941 if (lim.rlim_max < fds)
1942 lim.rlim_max = fds;
1943 lim.rlim_cur = fds;
1944 setrlimit(RLIMIT_NOFILE, &lim);
1945}
a9c15847
N
1946
1947int in_initrd(void)
1948{
1949 /* This is based on similar function in systemd. */
1950 struct statfs s;
1951 return statfs("/", &s) >= 0 &&
1952 (s.f_type == TMPFS_MAGIC ||
1953 s.f_type == RAMFS_MAGIC);
1954}