Two small fixes related to enough()
[thirdparty/mdadm.git] / util.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
6f02172d 4 * Copyright (C) 2001-2013 Neil Brown <neilb@suse.de>
64c4757e
NB
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
e736b623 22 * Email: <neilb@suse.de>
64c4757e
NB
23 */
24
9a9dab36 25#include "mdadm.h"
64c4757e 26#include "md_p.h"
edd8d13c 27#include <sys/socket.h>
64c4757e 28#include <sys/utsname.h>
9fe32043 29#include <sys/wait.h>
edd8d13c 30#include <sys/un.h>
a7dec3fd 31#include <sys/resource.h>
a9c15847
N
32#include <sys/vfs.h>
33#include <linux/magic.h>
98c6faba 34#include <ctype.h>
a322f70c 35#include <dirent.h>
a931db9e 36#include <signal.h>
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;
357ac106 310 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
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;
aabe020d 390 int i, rv;
8453e704
N
391 char *avail;
392
393 if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 ||
394 array.raid_disks <= 0)
395 return 0;
503975b9 396 avail = xcalloc(array.raid_disks, 1);
ca3b6696 397 for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704
N
398 disk.number = i;
399 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
400 continue;
9e6d9291
N
401 if (disk.major == 0 && disk.minor == 0)
402 continue;
403 array.nr_disks--;
404
8453e704
N
405 if (! (disk.state & (1<<MD_DISK_SYNC)))
406 continue;
407 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
408 continue;
8453e704
N
409 avail[disk.raid_disk] = 1;
410 }
411 /* This is used on an active array, so assume it is clean */
aabe020d 412 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 413 1, avail);
aabe020d
JS
414 free(avail);
415 return rv;
8453e704
N
416}
417
22e263f6
AC
418const int uuid_zero[4] = { 0, 0, 0, 0 };
419
f277ce36 420int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 421{
f277ce36
NB
422 if (swapuuid) {
423 /* parse uuids are hostendian.
424 * uuid's from some superblocks are big-ending
aba69144 425 * if there is a difference, we need to swap..
f277ce36
NB
426 */
427 unsigned char *ac = (unsigned char *)a;
428 unsigned char *bc = (unsigned char *)b;
429 int i;
ca3b6696 430 for (i = 0; i < 16; i += 4) {
f277ce36
NB
431 if (ac[i+0] != bc[i+3] ||
432 ac[i+1] != bc[i+2] ||
433 ac[i+2] != bc[i+1] ||
434 ac[i+3] != bc[i+0])
435 return 0;
436 }
437 return 1;
438 } else {
439 if (a[0]==b[0] &&
440 a[1]==b[1] &&
441 a[2]==b[2] &&
442 a[3]==b[3])
443 return 1;
444 return 0;
445 }
64c4757e 446}
ca3b6696 447
350f29f9
NB
448void copy_uuid(void *a, int b[4], int swapuuid)
449{
450 if (swapuuid) {
451 /* parse uuids are hostendian.
452 * uuid's from some superblocks are big-ending
453 * if there is a difference, we need to swap..
454 */
455 unsigned char *ac = (unsigned char *)a;
456 unsigned char *bc = (unsigned char *)b;
457 int i;
ca3b6696 458 for (i = 0; i < 16; i += 4) {
350f29f9
NB
459 ac[i+0] = bc[i+3];
460 ac[i+1] = bc[i+2];
461 ac[i+2] = bc[i+1];
462 ac[i+3] = bc[i+0];
463 }
464 } else
465 memcpy(a, b, 16);
466}
64c4757e 467
aae5a112 468char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 469{
9968e376 470 int i, j;
d7288ddc
N
471 char uuid[16];
472 char *c = buf;
473 strcpy(c, "UUID-");
474 c += strlen(c);
aae5a112 475 copy_uuid(uuid, id, swap);
9968e376 476 for (i = 0; i < 4; i++) {
9968e376 477 if (i)
ff54de6e 478 *c++ = sep;
9968e376
DW
479 for (j = 3; j >= 0; j--) {
480 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
481 c+= 2;
482 }
d7288ddc
N
483 }
484 return buf;
aae5a112
DW
485
486}
487
488char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
489{
86983cce
N
490 // dirty hack to work around an issue with super1 superblocks...
491 // super1 superblocks need swapuuid set in order for assembly to
492 // work, but can't have it set if we want this printout to match
493 // all the other uuid printouts in super1.c, so we force swapuuid
494 // to 1 to make our printout match the rest of super1
495 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
496}
497
435d4ebb 498#ifndef MDASSEMBLE
682c7051
NB
499int check_ext2(int fd, char *name)
500{
501 /*
502 * Check for an ext2fs file system.
503 * Superblock is always 1K at 1K offset
504 *
505 * s_magic is le16 at 56 == 0xEF53
506 * report mtime - le32 at 44
507 * blocks - le32 at 4
508 * logblksize - le32 at 24
509 */
510 unsigned char sb[1024];
511 time_t mtime;
2538ba2a
N
512 unsigned long long size;
513 int bsize;
682c7051
NB
514 if (lseek(fd, 1024,0)!= 1024)
515 return 0;
516 if (read(fd, sb, 1024)!= 1024)
517 return 0;
518 if (sb[56] != 0x53 || sb[57] != 0xef)
519 return 0;
520
521 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
522 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
523 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 524 size <<= bsize;
e7b84f9d 525 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 526 name);
2538ba2a 527 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
528 return 1;
529}
530
531int check_reiser(int fd, char *name)
532{
533 /*
534 * superblock is at 64K
535 * size is 1024;
536 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
537 *
538 */
539 unsigned char sb[1024];
2538ba2a 540 unsigned long long size;
682c7051
NB
541 if (lseek(fd, 64*1024, 0) != 64*1024)
542 return 0;
543 if (read(fd, sb, 1024) != 1024)
544 return 0;
ca3b6696
N
545 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
546 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 547 return 0;
e7b84f9d 548 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 549 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 550 cont_err("size = %lluK\n", size*4);
aba69144 551
682c7051
NB
552 return 1;
553}
554
555int check_raid(int fd, char *name)
556{
4b1ac34b 557 struct mdinfo info;
682c7051 558 time_t crtime;
d078d77c 559 char *level;
82d9eba6 560 struct supertype *st = guess_super(fd);
f9ce90ba 561
cb8f6859
N
562 if (!st)
563 return 0;
3da92f27 564 st->ss->load_super(st, fd, name);
82d9eba6 565 /* Looks like a raid array .. */
e7b84f9d 566 pr_err("%s appears to be part of a raid array:\n",
82d9eba6 567 name);
a5d85af7 568 st->ss->getinfo_super(st, &info, NULL);
3da92f27 569 st->ss->free_super(st);
82d9eba6 570 crtime = info.array.ctime;
d078d77c
NB
571 level = map_num(pers, info.array.level);
572 if (!level) level = "-unknown-";
d33f1518
N
573 cont_err("level=%s devices=%d ctime=%s",
574 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 575 return 1;
682c7051
NB
576}
577
682c7051
NB
578int ask(char *mesg)
579{
580 char *add = "";
581 int i;
ca3b6696 582 for (i = 0; i < 5; i++) {
682c7051
NB
583 char buf[100];
584 fprintf(stderr, "%s%s", mesg, add);
585 fflush(stderr);
586 if (fgets(buf, 100, stdin)==NULL)
587 return 0;
588 if (buf[0]=='y' || buf[0]=='Y')
589 return 1;
590 if (buf[0]=='n' || buf[0]=='N')
591 return 0;
592 add = "(y/n) ";
593 }
e7b84f9d 594 pr_err("assuming 'no'\n");
682c7051
NB
595 return 0;
596}
435d4ebb 597#endif /* MDASSEMBLE */
682c7051 598
8d80900b 599int is_standard(char *dev, int *nump)
e5329c37
NB
600{
601 /* tests if dev is a "standard" md dev name.
602 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 603 * where NN is a string of digits
598f0d58
NB
604 * Returns 1 if a partitionable standard,
605 * -1 if non-partitonable,
606 * 0 if not a standard name.
e5329c37 607 */
8d80900b 608 char *d = strrchr(dev, '/');
ca3b6696 609 int type = 0;
8d80900b
NB
610 int num;
611 if (!d)
e5329c37 612 return 0;
ca3b6696
N
613 if (strncmp(d, "/d",2) == 0)
614 d += 2, type = 1; /* /dev/md/dN{pM} */
615 else if (strncmp(d, "/md_d", 5) == 0)
616 d += 5, type = 1; /* /dev/md_dN{pM} */
617 else if (strncmp(d, "/md", 3) == 0)
618 d += 3, type = -1; /* /dev/mdN */
619 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
620 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
621 else
622 return 0;
8d80900b 623 if (!*d)
e5329c37 624 return 0;
8d80900b
NB
625 num = atoi(d);
626 while (isdigit(*d))
627 d++;
628 if (*d)
e5329c37 629 return 0;
8d80900b
NB
630 if (nump) *nump = num;
631
632 return type;
e5329c37
NB
633}
634
4b1ac34b 635unsigned long calc_csum(void *super, int bytes)
82b27616 636{
56eb10c0 637 unsigned long long newcsum = 0;
82b27616 638 int i;
4b1ac34b
NB
639 unsigned int csum;
640 unsigned int *superc = (unsigned int*) super;
82b27616 641
ca3b6696
N
642 for(i = 0; i < bytes/4; i++)
643 newcsum += superc[i];
82b27616 644 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 645#ifdef __alpha__
aba69144 646/* The in-kernel checksum calculation is always 16bit on
570c0542 647 * the alpha, though it is 32 bit on i386...
ca3b6696 648 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
649 * a way that it shouldn't).
650 */
651 csum = (csum & 0xffff) + (csum >> 16);
652 csum = (csum & 0xffff) + (csum >> 16);
653#endif
82b27616
NB
654 return csum;
655}
cd29a5c8 656
435d4ebb 657#ifndef MDASSEMBLE
56eb10c0 658char *human_size(long long bytes)
cd29a5c8
NB
659{
660 static char buf[30];
d5d3721e
NB
661
662 /* We convert bytes to either centi-M{ega,ibi}bytes or
663 * centi-G{igi,ibi}bytes, with appropriate rounding,
664 * and then print 1/100th of those as a decimal.
665 * We allow upto 2048Megabytes before converting to
666 * gigabytes, as that shows more precision and isn't
667 * too large a number.
01619b48 668 * Terabytes are not yet handled.
d5d3721e 669 */
cd29a5c8 670
56eb10c0 671 if (bytes < 5000*1024)
ca3b6696 672 buf[0] = 0;
d5d3721e
NB
673 else if (bytes < 2*1024LL*1024LL*1024LL) {
674 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
675 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 676 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
677 cMiB/100 , cMiB % 100,
678 cMB/100, cMB % 100);
679 } else {
680 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
681 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 682 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
683 cGiB/100 , cGiB % 100,
684 cGB/100, cGB % 100);
685 }
cd29a5c8
NB
686 return buf;
687}
e0d19036 688
f0ec6710 689char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
690{
691 static char buf[30];
e0d19036 692
570abc6f
MN
693 /* We convert bytes to either centi-M{ega,ibi}bytes or
694 * centi-G{igi,ibi}bytes, with appropriate rounding,
695 * and then print 1/100th of those as a decimal.
696 * We allow upto 2048Megabytes before converting to
697 * gigabytes, as that shows more precision and isn't
698 * too large a number.
699 * Terabytes are not yet handled.
f0ec6710
MN
700 *
701 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
702 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
703 */
704
e0d19036 705 if (bytes < 5000*1024)
570abc6f 706 buf[0] = 0;
f0ec6710
MN
707 else if (prefix == IEC) {
708 if (bytes < 2*1024LL*1024LL*1024LL) {
709 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
710 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
711 cMiB/100 , cMiB % 100);
712 } else {
713 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
714 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
715 cGiB/100 , cGiB % 100);
716 }
717 }
718 else if (prefix == JEDEC) {
719 if (bytes < 2*1024LL*1024LL*1024LL) {
720 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
721 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
722 cMB/100, cMB % 100);
723 } else {
724 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
725 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
726 cGB/100 , cGB % 100);
727 }
570abc6f 728 }
f0ec6710
MN
729 else
730 buf[0] = 0;
570abc6f 731
e0d19036
NB
732 return buf;
733}
e4965ef8
N
734
735void print_r10_layout(int layout)
736{
737 int near = layout & 255;
738 int far = (layout >> 8) & 255;
739 int offset = (layout&0x10000);
740 char *sep = "";
741
742 if (near != 1) {
743 printf("%s near=%d", sep, near);
744 sep = ",";
745 }
746 if (far != 1)
747 printf("%s %s=%d", sep, offset?"offset":"far", far);
748 if (near*far == 1)
749 printf("NO REDUNDANCY");
750}
435d4ebb 751#endif
e0d19036 752
5f8097be
NB
753unsigned long long calc_array_size(int level, int raid_disks, int layout,
754 int chunksize, unsigned long long devsize)
577e8448 755{
e03640bd
CA
756 if (level == 1)
757 return devsize;
577e8448
AK
758 devsize &= ~(unsigned long long)((chunksize>>9)-1);
759 return get_data_disks(level, layout, raid_disks) * devsize;
760}
761
762int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
763{
764 int data_disks = 0;
765 switch (level) {
ca3b6696
N
766 case 0: data_disks = raid_disks;
767 break;
768 case 1: data_disks = 1;
769 break;
5f8097be 770 case 4:
ca3b6696
N
771 case 5: data_disks = raid_disks - 1;
772 break;
773 case 6: data_disks = raid_disks - 2;
774 break;
5f8097be
NB
775 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
776 break;
777 }
577e8448
AK
778
779 return data_disks;
5f8097be
NB
780}
781
4dd2df09
N
782int devnm2devid(char *devnm)
783{
784 /* First look in /sys/block/$DEVNM/dev for %d:%d
785 * If that fails, try parsing out a number
786 */
787 char path[100];
788 char *ep;
789 int fd;
790 int mjr,mnr;
791
792 sprintf(path, "/sys/block/%s/dev", devnm);
793 fd = open(path, O_RDONLY);
794 if (fd >= 0) {
795 char buf[20];
796 int n = read(fd, buf, sizeof(buf));
797 close(fd);
798 if (n > 0)
799 buf[n] = 0;
800 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
801 return makedev(mjr, mnr);
802 }
803 if (strncmp(devnm, "md_d", 4) == 0 &&
804 isdigit(devnm[4]) &&
805 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
806 ep > devnm && *ep == 0)
807 return makedev(get_mdp_major(), mnr << MdpMinorShift);
808
809 if (strncmp(devnm, "md", 2) == 0 &&
810 isdigit(devnm[2]) &&
811 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
812 ep > devnm && *ep == 0)
813 return makedev(MD_MAJOR, mnr);
814
815 return 0;
816}
817
5a23a06e 818#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 819char *get_md_name(char *devnm)
e0d19036
NB
820{
821 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 822 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 823
e0d19036
NB
824 static char devname[50];
825 struct stat stb;
4dd2df09 826 dev_t rdev = devnm2devid(devnm);
dd0781e5 827 char *dn;
98c6faba 828
4dd2df09
N
829 if (rdev == 0)
830 return 0;
831 if (strncmp(devnm, "md_", 3) == 0) {
832 snprintf(devname, sizeof(devname), "/dev/md/%s",
833 devnm + 3);
98c6faba
NB
834 if (stat(devname, &stb) == 0
835 && (S_IFMT&stb.st_mode) == S_IFBLK
836 && (stb.st_rdev == rdev))
837 return devname;
838 }
4dd2df09
N
839 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
840 if (stat(devname, &stb) == 0
841 && (S_IFMT&stb.st_mode) == S_IFBLK
842 && (stb.st_rdev == rdev))
843 return devname;
844
845 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
846 if (stat(devname, &stb) == 0
847 && (S_IFMT&stb.st_mode) == S_IFBLK
848 && (stb.st_rdev == rdev))
849 return devname;
850
16c6fa80 851 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
852 if (dn)
853 return dn;
4dd2df09 854 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 855 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
856 if (errno != EEXIST)
857 return NULL;
e0d19036
NB
858
859 if (stat(devname, &stb) == 0
860 && (S_IFMT&stb.st_mode) == S_IFBLK
861 && (stb.st_rdev == rdev))
862 return devname;
863 unlink(devname);
864 return NULL;
865}
866
867void put_md_name(char *name)
868{
ca3b6696 869 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
870 unlink(name);
871}
435d4ebb 872#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 873
8b0dabea
NB
874int dev_open(char *dev, int flags)
875{
876 /* like 'open', but if 'dev' matches %d:%d, create a temp
877 * block device and open that
878 */
879 char *e;
880 int fd = -1;
881 char devname[32];
e81cdd9f 882 int major;
8b0dabea 883 int minor;
e81cdd9f
NB
884
885 if (!dev) return -1;
6df6a774 886 flags |= O_DIRECT;
e81cdd9f
NB
887
888 major = strtoul(dev, &e, 0);
8b0dabea
NB
889 if (e > dev && *e == ':' && e[1] &&
890 (minor = strtoul(e+1, &e, 0)) >= 0 &&
891 *e == 0) {
9dc7d357
N
892 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
893 (int)getpid(), major, minor);
894 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
895 fd = open(devname, flags);
896 unlink(devname);
6df6a774
N
897 }
898 if (fd < 0) {
9dc7d357 899 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
900 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
901 (int)getpid(), major, minor);
ca3b6696 902 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
903 fd = open(devname, flags);
904 unlink(devname);
905 }
8b0dabea
NB
906 }
907 } else
6df6a774 908 fd = open(dev, flags);
8b0dabea
NB
909 return fd;
910}
f9ce90ba 911
4dd2df09 912int open_dev_flags(char *devnm, int flags)
e8a70c89 913{
4dd2df09 914 int devid;
e8a70c89
N
915 char buf[20];
916
4dd2df09
N
917 devid = devnm2devid(devnm);
918 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
919 return dev_open(buf, flags);
920}
921
4dd2df09 922int open_dev(char *devnm)
d998b738 923{
4dd2df09 924 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
925}
926
4dd2df09 927int open_dev_excl(char *devnm)
a931db9e
NB
928{
929 char buf[20];
930 int i;
7187750e 931 int flags = O_RDWR;
4dd2df09 932 int devid = devnm2devid(devnm);
399e0b97 933 long delay = 1000;
a931db9e 934
4dd2df09 935 sprintf(buf, "%d:%d", major(devid), minor(devid));
ca3b6696 936 for (i = 0 ; i < 25 ; i++) {
7187750e 937 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
938 if (fd >= 0)
939 return fd;
7187750e
N
940 if (errno == EACCES && flags == O_RDWR) {
941 flags = O_RDONLY;
942 continue;
943 }
a931db9e
NB
944 if (errno != EBUSY)
945 return fd;
399e0b97
N
946 usleep(delay);
947 if (delay < 200000)
948 delay *= 2;
a931db9e
NB
949 }
950 return -1;
951}
952
9008ed1c
N
953int same_dev(char *one, char *two)
954{
955 struct stat st1, st2;
956 if (stat(one, &st1) != 0)
957 return 0;
958 if (stat(two, &st2) != 0)
959 return 0;
960 if ((st1.st_mode & S_IFMT) != S_IFBLK)
961 return 0;
962 if ((st2.st_mode & S_IFMT) != S_IFBLK)
963 return 0;
964 return st1.st_rdev == st2.st_rdev;
965}
966
a7c6e3fb 967void wait_for(char *dev, int fd)
a714580e
N
968{
969 int i;
a7c6e3fb 970 struct stat stb_want;
399e0b97 971 long delay = 1000;
a7c6e3fb
N
972
973 if (fstat(fd, &stb_want) != 0 ||
974 (stb_want.st_mode & S_IFMT) != S_IFBLK)
975 return;
a714580e 976
ca3b6696 977 for (i = 0 ; i < 25 ; i++) {
a714580e 978 struct stat stb;
a7c6e3fb
N
979 if (stat(dev, &stb) == 0 &&
980 (stb.st_mode & S_IFMT) == S_IFBLK &&
981 (stb.st_rdev == stb_want.st_rdev))
a714580e 982 return;
399e0b97
N
983 usleep(delay);
984 if (delay < 200000)
985 delay *= 2;
a714580e 986 }
436305c6
DW
987 if (i == 25)
988 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
989}
990
0f22b998
N
991struct superswitch *superlist[] =
992{
993 &super0, &super1,
994 &super_ddf, &super_imsm,
0592faeb 995 &mbr, &gpt,
0f22b998 996 NULL };
f9ce90ba 997
ea24acd0 998#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 999
4725bc31 1000struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1001{
1686dc25
NB
1002 mdu_array_info_t array;
1003 int vers;
1004 int minor;
1005 struct supertype *st = NULL;
7e0f6979 1006 struct mdinfo *sra;
142cb9e1 1007 char *verstr;
1686dc25
NB
1008 char version[20];
1009 int i;
f7e7067b 1010 char *subarray = NULL;
4dd2df09 1011 char container[32] = "";
1686dc25 1012
4dd2df09 1013 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1014
1015 if (sra) {
7e0f6979
NB
1016 vers = sra->array.major_version;
1017 minor = sra->array.minor_version;
142cb9e1 1018 verstr = sra->text_version;
1686dc25
NB
1019 } else {
1020 if (ioctl(fd, GET_ARRAY_INFO, &array))
1021 array.major_version = array.minor_version = 0;
1022 vers = array.major_version;
1023 minor = array.minor_version;
142cb9e1 1024 verstr = "";
6fbba4c9 1025 }
82d9eba6 1026
1686dc25
NB
1027 if (vers != -1) {
1028 sprintf(version, "%d.%d", vers, minor);
1029 verstr = version;
6fbba4c9 1030 }
3c558363 1031 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1032 char *dev = verstr+1;
5f7e44b2 1033
f7e7067b 1034 subarray = strchr(dev, '/');
92d49ecf 1035 if (subarray) {
f7e7067b 1036 *subarray++ = '\0';
503975b9 1037 subarray = xstrdup(subarray);
92d49ecf 1038 }
4dd2df09 1039 strcpy(container, dev);
f7e7067b
NB
1040 if (sra)
1041 sysfs_free(sra);
5f7e44b2 1042 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1043 if (sra && sra->text_version[0])
1044 verstr = sra->text_version;
1045 else
1046 verstr = "-no-metadata-";
f7e7067b
NB
1047 }
1048
1049 for (i = 0; st == NULL && superlist[i] ; i++)
1050 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
1051
1052 if (sra)
1053 sysfs_free(sra);
f7e7067b 1054 if (st) {
3b0896f8 1055 st->sb = NULL;
1f49fb3a
N
1056 if (subarrayp)
1057 *subarrayp = subarray;
4dd2df09
N
1058 strcpy(st->container_devnm, container);
1059 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1060 } else
1061 free(subarray);
5f7e44b2 1062
82d9eba6 1063 return st;
f9ce90ba 1064}
ea24acd0
NB
1065#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1066
0f0749ad 1067int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1068{
1069 char buf[20];
1070 int fd;
1071
1072 sprintf(buf, "%d:%d", major(id), minor(id));
1073 fd = dev_open(buf, O_RDONLY);
1074 if (fd < 0)
1075 return 0;
1076 if (get_dev_size(fd, NULL, size)) {
1077 close(fd);
1078 return 1;
1079 }
1080 close(fd);
1081 return 0;
1082}
f9ce90ba 1083
159c3a1a 1084struct supertype *dup_super(struct supertype *orig)
3da92f27 1085{
159c3a1a 1086 struct supertype *st;
1686dc25 1087
d2ca6449
NB
1088 if (!orig)
1089 return orig;
503975b9 1090 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1091 st->ss = orig->ss;
1092 st->max_devs = orig->max_devs;
1093 st->minor_version = orig->minor_version;
cb8f6859 1094 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1095 st->data_offset = orig->data_offset;
159c3a1a
NB
1096 st->sb = NULL;
1097 st->info = NULL;
1098 return st;
3da92f27
NB
1099}
1100
54887ad8 1101struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1102{
1103 /* try each load_super to find the best match,
1104 * and return the best superswitch
1105 */
82d9eba6
NB
1106 struct superswitch *ss;
1107 struct supertype *st;
f21e18ca 1108 time_t besttime = 0;
570c0542 1109 int bestsuper = -1;
f9ce90ba
NB
1110 int i;
1111
503975b9 1112 st = xcalloc(1, sizeof(*st));
4dd2df09 1113 st->container_devnm[0] = 0;
d1d599ea 1114
ca3b6696 1115 for (i = 0 ; superlist[i]; i++) {
f9ce90ba
NB
1116 int rv;
1117 ss = superlist[i];
54887ad8
N
1118 if (guess_type == guess_array && ss->add_to_super == NULL)
1119 continue;
1120 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1121 continue;
ef609477 1122 memset(st, 0, sizeof(*st));
df3346e6 1123 st->ignore_hw_compat = 1;
3da92f27 1124 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1125 if (rv == 0) {
1126 struct mdinfo info;
a5d85af7 1127 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1128 if (bestsuper == -1 ||
1129 besttime < info.array.ctime) {
1130 bestsuper = i;
1131 besttime = info.array.ctime;
570c0542 1132 }
3da92f27 1133 ss->free_super(st);
570c0542
NB
1134 }
1135 }
1136 if (bestsuper != -1) {
1137 int rv;
ef609477 1138 memset(st, 0, sizeof(*st));
df3346e6 1139 st->ignore_hw_compat = 1;
3da92f27 1140 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1141 if (rv == 0) {
5e747af2 1142 superlist[bestsuper]->free_super(st);
82d9eba6 1143 return st;
f9ce90ba
NB
1144 }
1145 }
570c0542 1146 free(st);
f9ce90ba
NB
1147 return NULL;
1148}
fe6729fa 1149
beae1dfe
NB
1150/* Return size of device in bytes */
1151int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1152{
1153 unsigned long long ldsize;
c2c9bb6f
NB
1154 struct stat st;
1155
1156 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1157 ldsize = (unsigned long long)st.st_size;
1158 else
beae1dfe
NB
1159#ifdef BLKGETSIZE64
1160 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1161#endif
1162 {
1163 unsigned long dsize;
1164 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1165 ldsize = dsize;
1166 ldsize <<= 9;
1167 } else {
1168 if (dname)
e7b84f9d 1169 pr_err("Cannot get size of %s: %s\b",
beae1dfe
NB
1170 dname, strerror(errno));
1171 return 0;
1172 }
1173 }
1174 *sizep = ldsize;
1175 return 1;
1176}
8fac0577 1177
3a371610
N
1178/* Return true if this can only be a container, not a member device.
1179 * i.e. is and md device and size is zero
1180 */
1181int must_be_container(int fd)
1182{
1183 unsigned long long size;
1184 if (md_get_version(fd) < 0)
1185 return 0;
1186 if (get_dev_size(fd, NULL, &size) == 0)
1187 return 1;
1188 if (size == 0)
1189 return 1;
1190 return 0;
1191}
034b203a
TM
1192
1193/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1194 * Returns: 1 if successful
1195 * -1 for unknown partition type
1196 * 0 for other errors
1197 */
1198static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1199{
056b331e 1200 struct GPT gpt;
034b203a 1201 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1202 struct GPT_part_entry *part;
1203 char buf[512];
034b203a
TM
1204 unsigned long long curr_part_end;
1205 unsigned all_partitions, entry_size;
f21e18ca 1206 unsigned part_nr;
034b203a
TM
1207
1208 *endofpart = 0;
1209
056b331e 1210 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1211 /* skip protective MBR */
034b203a 1212 lseek(fd, 512, SEEK_SET);
73e658d8 1213 /* read GPT header */
056b331e 1214 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1215 return 0;
1216
1217 /* get the number of partition entries and the entry size */
056b331e
N
1218 all_partitions = __le32_to_cpu(gpt.part_cnt);
1219 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1220
1221 /* Check GPT signature*/
056b331e 1222 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1223 return -1;
1224
1225 /* sanity checks */
1226 if (all_partitions > 1024 ||
db7fdfe4 1227 entry_size > sizeof(buf))
034b203a
TM
1228 return -1;
1229
db7fdfe4
JS
1230 part = (struct GPT_part_entry *)buf;
1231
ca3b6696 1232 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1233 /* read partition entry */
db7fdfe4 1234 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1235 return 0;
1236
034b203a 1237 /* is this valid partition? */
db7fdfe4 1238 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1239 /* check the last lba for the current partition */
db7fdfe4 1240 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1241 if (curr_part_end > *endofpart)
1242 *endofpart = curr_part_end;
1243 }
1244
034b203a
TM
1245 }
1246 return 1;
1247}
1248
1249/* Sets endofpart parameter to the last block used by the last partition on the device.
1250 * Returns: 1 if successful
1251 * -1 for unknown partition type
1252 * 0 for other errors
1253 */
1254static int get_last_partition_end(int fd, unsigned long long *endofpart)
1255{
056b331e 1256 struct MBR boot_sect;
034b203a
TM
1257 struct MBR_part_record *part;
1258 unsigned long long curr_part_end;
f21e18ca 1259 unsigned part_nr;
034b203a
TM
1260 int retval = 0;
1261
1262 *endofpart = 0;
1263
056b331e 1264 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1265 /* read MBR */
1266 lseek(fd, 0, 0);
056b331e 1267 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1268 goto abort;
1269
1270 /* check MBP signature */
056b331e 1271 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1272 retval = 1;
1273 /* found the correct signature */
056b331e 1274 part = boot_sect.parts;
034b203a 1275
ca3b6696 1276 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
034b203a
TM
1277 /* check for GPT type */
1278 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1279 retval = get_gpt_last_partition_end(fd, endofpart);
1280 break;
1281 }
1282 /* check the last used lba for the current partition */
1283 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1284 __le32_to_cpu(part->blocks_num);
1285 if (curr_part_end > *endofpart)
1286 *endofpart = curr_part_end;
1287
1288 part++;
1289 }
1290 } else {
1291 /* Unknown partition table */
1292 retval = -1;
1293 }
1294 abort:
1295 return retval;
1296}
1297
53ed6ac3
KW
1298int check_partitions(int fd, char *dname, unsigned long long freesize,
1299 unsigned long long size)
034b203a
TM
1300{
1301 /*
1302 * Check where the last partition ends
1303 */
1304 unsigned long long endofpart;
1305 int ret;
1306
1307 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1308 /* There appears to be a partition table here */
1309 if (freesize == 0) {
1310 /* partitions will not be visible in new device */
e7b84f9d
N
1311 pr_err("partition table exists on %s but will be lost or\n"
1312 " meaningless after creating array\n",
1313 dname);
034b203a
TM
1314 return 1;
1315 } else if (endofpart > freesize) {
1316 /* last partition overlaps metadata */
e7b84f9d
N
1317 pr_err("metadata will over-write last partition on %s.\n",
1318 dname);
034b203a 1319 return 1;
53ed6ac3
KW
1320 } else if (size && endofpart > size) {
1321 /* partitions will be truncated in new device */
e7b84f9d
N
1322 pr_err("array size is too small to cover all partitions on %s.\n",
1323 dname);
53ed6ac3 1324 return 1;
034b203a
TM
1325 }
1326 }
1327 return 0;
1328}
1329
a322f70c
DW
1330int open_container(int fd)
1331{
1332 /* 'fd' is a block device. Find out if it is in use
1333 * by a container, and return an open fd on that container.
1334 */
1335 char path[256];
1336 char *e;
1337 DIR *dir;
1338 struct dirent *de;
1339 int dfd, n;
1340 char buf[200];
1341 int major, minor;
1342 struct stat st;
1343
1344 if (fstat(fd, &st) != 0)
1345 return -1;
1346 sprintf(path, "/sys/dev/block/%d:%d/holders",
1347 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1348 e = path + strlen(path);
1349
1350 dir = opendir(path);
1351 if (!dir)
1352 return -1;
1353 while ((de = readdir(dir))) {
1354 if (de->d_ino == 0)
1355 continue;
1356 if (de->d_name[0] == '.')
1357 continue;
a994592d
N
1358 /* Need to make sure it is a container and not a volume */
1359 sprintf(e, "/%s/md/metadata_version", de->d_name);
1360 dfd = open(path, O_RDONLY);
1361 if (dfd < 0)
1362 continue;
1363 n = read(dfd, buf, sizeof(buf));
1364 close(dfd);
1365 if (n <= 0 || (unsigned)n >= sizeof(buf))
1366 continue;
1367 buf[n] = 0;
1368 if (strncmp(buf, "external", 8) != 0 ||
1369 n < 10 ||
1370 buf[9] == '/')
1371 continue;
a322f70c
DW
1372 sprintf(e, "/%s/dev", de->d_name);
1373 dfd = open(path, O_RDONLY);
1374 if (dfd < 0)
1375 continue;
1376 n = read(dfd, buf, sizeof(buf));
1377 close(dfd);
f21e18ca 1378 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1379 continue;
1380 buf[n] = 0;
1381 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1382 continue;
1383 sprintf(buf, "%d:%d", major, minor);
1384 dfd = dev_open(buf, O_RDONLY);
1385 if (dfd >= 0) {
1386 closedir(dir);
1387 return dfd;
1388 }
1389 }
355726fa 1390 closedir(dir);
a322f70c
DW
1391 return -1;
1392}
1393
33414a01
DW
1394struct superswitch *version_to_superswitch(char *vers)
1395{
1396 int i;
1397
1398 for (i = 0; superlist[i]; i++) {
1399 struct superswitch *ss = superlist[i];
1400
1401 if (strcmp(vers, ss->name) == 0)
1402 return ss;
1403 }
1404
1405 return NULL;
1406}
1407
4dd2df09
N
1408int metadata_container_matches(char *metadata, char *devnm)
1409{
1410 /* Check if 'devnm' is the container named in 'metadata'
1411 * which is
1412 * /containername/componentname or
1413 * -containername/componentname
1414 */
1415 int l;
1416 if (*metadata != '/' && *metadata != '-')
1417 return 0;
1418 l = strlen(devnm);
1419 if (strncmp(metadata+1, devnm, l) != 0)
1420 return 0;
1421 if (metadata[l+1] != '/')
1422 return 0;
1423 return 1;
1424}
1425
1426int metadata_subdev_matches(char *metadata, char *devnm)
1427{
1428 /* Check if 'devnm' is the subdev named in 'metadata'
1429 * which is
1430 * /containername/subdev or
1431 * -containername/subdev
1432 */
1433 char *sl;
1434 if (*metadata != '/' && *metadata != '-')
1435 return 0;
1436 sl = strchr(metadata+1, '/');
1437 if (!sl)
1438 return 0;
1439 if (strcmp(sl+1, devnm) == 0)
1440 return 1;
1441 return 0;
1442}
1443
33414a01
DW
1444int is_container_member(struct mdstat_ent *mdstat, char *container)
1445{
1446 if (mdstat->metadata_version == NULL ||
1447 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1448 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1449 return 0;
1450
1451 return 1;
1452}
1453
1454int is_subarray_active(char *subarray, char *container)
1455{
1456 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1457 struct mdstat_ent *ent;
1458
e5408a32
DW
1459 for (ent = mdstat; ent; ent = ent->next)
1460 if (is_container_member(ent, container))
e5e5d7ce 1461 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1462 break;
33414a01
DW
1463
1464 free_mdstat(mdstat);
1465
1466 return ent != NULL;
1467}
1468
1469/* open_subarray - opens a subarray in a container
1470 * @dev: container device name
feab51f8 1471 * @st: empty supertype
33414a01
DW
1472 * @quiet: block reporting errors flag
1473 *
1474 * On success returns an fd to a container and fills in *st
1475 */
feab51f8 1476int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1477{
1478 struct mdinfo *mdi;
a951a4f7 1479 struct mdinfo *info;
33414a01 1480 int fd, err = 1;
4dd2df09 1481 char *_devnm;
33414a01
DW
1482
1483 fd = open(dev, O_RDWR|O_EXCL);
1484 if (fd < 0) {
1485 if (!quiet)
e7b84f9d 1486 pr_err("Couldn't open %s, aborting\n",
33414a01 1487 dev);
b990032d 1488 return -1;
33414a01
DW
1489 }
1490
4dd2df09
N
1491 _devnm = fd2devnm(fd);
1492 if (_devnm == NULL) {
33414a01 1493 if (!quiet)
e7b84f9d
N
1494 pr_err("Failed to determine device number for %s\n",
1495 dev);
33414a01
DW
1496 goto close_fd;
1497 }
4dd2df09 1498 strcpy(st->devnm, _devnm);
33414a01 1499
4dd2df09 1500 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1501 if (!mdi) {
1502 if (!quiet)
e7b84f9d 1503 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1504 dev);
1505 goto close_fd;
1506 }
1507
1508 if (mdi->array.level != UnSet) {
1509 if (!quiet)
e7b84f9d 1510 pr_err("%s is not a container\n", dev);
33414a01
DW
1511 goto free_sysfs;
1512 }
1513
1514 st->ss = version_to_superswitch(mdi->text_version);
1515 if (!st->ss) {
1516 if (!quiet)
e7b84f9d
N
1517 pr_err("Operation not supported for %s metadata\n",
1518 mdi->text_version);
33414a01
DW
1519 goto free_sysfs;
1520 }
1521
4dd2df09 1522 if (st->devnm[0] == 0) {
33414a01 1523 if (!quiet)
e7b84f9d 1524 pr_err("Failed to allocate device name\n");
33414a01
DW
1525 goto free_sysfs;
1526 }
1527
db20d413 1528 if (!st->ss->load_container) {
33414a01 1529 if (!quiet)
e7b84f9d 1530 pr_err("%s is not a container\n", dev);
4dd2df09 1531 goto free_sysfs;
33414a01
DW
1532 }
1533
db20d413 1534 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1535 if (!quiet)
e7b84f9d 1536 pr_err("Failed to load metadata for %s\n",
db20d413 1537 dev);
4dd2df09 1538 goto free_sysfs;
33414a01
DW
1539 }
1540
a951a4f7
N
1541 info = st->ss->container_content(st, subarray);
1542 if (!info) {
1543 if (!quiet)
e7b84f9d 1544 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1545 subarray, dev);
1546 goto free_super;
1547 }
1548 free(info);
1549
33414a01
DW
1550 err = 0;
1551
1552 free_super:
1553 if (err)
1554 st->ss->free_super(st);
33414a01
DW
1555 free_sysfs:
1556 sysfs_free(mdi);
1557 close_fd:
1558 if (err)
1559 close(fd);
1560
1561 if (err)
1562 return -1;
1563 else
1564 return fd;
1565}
1566
7801ac20
N
1567int add_disk(int mdfd, struct supertype *st,
1568 struct mdinfo *sra, struct mdinfo *info)
1569{
1570 /* Add a device to an array, in one of 2 ways. */
1571 int rv;
1572#ifndef MDASSEMBLE
1573 if (st->ss->external) {
d23534e4
DW
1574 if (info->disk.state & (1<<MD_DISK_SYNC))
1575 info->recovery_start = MaxSector;
1576 else
1577 info->recovery_start = 0;
2904b26f 1578 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1579 if (! rv) {
1580 struct mdinfo *sd2;
f35f2525
N
1581 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1582 if (sd2 == info)
1583 break;
1584 if (sd2 == NULL) {
503975b9 1585 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1586 *sd2 = *info;
1587 sd2->next = sra->devs;
1588 sra->devs = sd2;
1589 }
7801ac20
N
1590 }
1591 } else
1592#endif
1593 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1594 return rv;
1595}
1596
de6ae750
N
1597int remove_disk(int mdfd, struct supertype *st,
1598 struct mdinfo *sra, struct mdinfo *info)
1599{
1600 int rv;
1601 /* Remove the disk given by 'info' from the array */
1602#ifndef MDASSEMBLE
1603 if (st->ss->external)
1604 rv = sysfs_set_str(sra, info, "slot", "none");
1605 else
1606#endif
1607 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1608 info->disk.minor));
1609 return rv;
1610}
1611
f35f2525
N
1612int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1613{
1614 /* Initialise kernel's knowledge of array.
1615 * This varies between externally managed arrays
1616 * and older kernels
1617 */
1618 int vers = md_get_version(mdfd);
1619 int rv;
1620
1621#ifndef MDASSEMBLE
1622 if (st->ss->external)
1623 rv = sysfs_set_array(info, vers);
1624 else
1625#endif
1626 if ((vers % 100) >= 1) { /* can use different versions */
1627 mdu_array_info_t inf;
1628 memset(&inf, 0, sizeof(inf));
1629 inf.major_version = info->array.major_version;
1630 inf.minor_version = info->array.minor_version;
1631 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1632 } else
1633 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1634 return rv;
1635}
1636
1e5c6983
DW
1637unsigned long long min_recovery_start(struct mdinfo *array)
1638{
1639 /* find the minimum recovery_start in an array for metadata
1640 * formats that only record per-array recovery progress instead
1641 * of per-device
1642 */
1643 unsigned long long recovery_start = MaxSector;
1644 struct mdinfo *d;
1645
1646 for (d = array->devs; d; d = d->next)
1647 recovery_start = min(recovery_start, d->recovery_start);
1648
1649 return recovery_start;
1650}
1651
4dd2df09 1652int mdmon_pid(char *devnm)
a931db9e
NB
1653{
1654 char path[100];
1655 char pid[10];
1656 int fd;
1657 int n;
10013317 1658
4dd2df09 1659 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1660
24f6f99b 1661 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1662
1663 if (fd < 0)
cf556303 1664 return -1;
a931db9e
NB
1665 n = read(fd, pid, 9);
1666 close(fd);
1667 if (n <= 0)
cf556303 1668 return -1;
24f6f99b 1669 return atoi(pid);
a931db9e
NB
1670}
1671
4dd2df09 1672int mdmon_running(char *devnm)
a931db9e 1673{
4dd2df09 1674 int pid = mdmon_pid(devnm);
24f6f99b 1675 if (pid <= 0)
a931db9e 1676 return 0;
24f6f99b 1677 if (kill(pid, 0) == 0)
a931db9e
NB
1678 return 1;
1679 return 0;
1680}
1681
4dd2df09 1682int start_mdmon(char *devnm)
8850ee3e 1683{
1913c325 1684 int i, skipped;
44d2e365 1685 int len;
ca3b6696 1686 pid_t pid;
9fe32043 1687 int status;
44d2e365
N
1688 char pathbuf[1024];
1689 char *paths[4] = {
1690 pathbuf,
1691 "/sbin/mdmon",
2f1bcf43 1692 "./mdmon",
44d2e365
N
1693 NULL
1694 };
8850ee3e 1695
40ebbb9c 1696 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1697 return 0;
1698
9cf014ec 1699 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1700 if (len > 0) {
1701 char *sl;
1702 pathbuf[len] = 0;
1703 sl = strrchr(pathbuf, '/');
1704 if (sl)
1705 sl++;
1706 else
1707 sl = pathbuf;
1708 strcpy(sl, "mdmon");
1709 } else
1710 pathbuf[0] = '\0';
1711
0f7bdf89 1712 /* First try to run systemctl */
9540cc24
N
1713 if (!check_env("MDADM_NO_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
1724 /* Don't want to see error messages from
1725 * systemctl. If the service doesn't exist,
1726 * we start mdmon ourselves.
1727 */
1728 close(2);
1729 open("/dev/null", O_WRONLY);
1730 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1731 devnm);
1732 status = execl("/usr/bin/systemctl", "systemctl",
1733 "start",
1734 pathbuf, NULL);
1735 status = execl("/bin/systemctl", "systemctl", "start",
1736 pathbuf, NULL);
1737 exit(1);
1738 case -1: pr_err("cannot run mdmon. "
1739 "Array remains readonly\n");
1740 return -1;
1741 default: /* parent - good */
1742 pid = wait(&status);
1743 if (pid >= 0 && status == 0)
1744 return 0;
1745 }
0f7bdf89
JS
1746
1747 /* That failed, try running mdmon directly */
8850ee3e
N
1748 switch(fork()) {
1749 case 0:
1750 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1751 skipped = 0;
ca3b6696 1752 for (i = 3; skipped < 20; i++)
1913c325
N
1753 if (close(i) < 0)
1754 skipped++;
1755 else
1756 skipped = 0;
1757
ca3b6696 1758 for (i = 0; paths[i]; i++)
a0963a86 1759 if (paths[i][0]) {
2f1bcf43 1760 execl(paths[i], paths[i],
4dd2df09 1761 devnm, NULL);
a0963a86 1762 }
8850ee3e 1763 exit(1);
e7b84f9d 1764 case -1: pr_err("cannot run mdmon. "
8850ee3e
N
1765 "Array remains readonly\n");
1766 return -1;
9fe32043
N
1767 default: /* parent - good */
1768 pid = wait(&status);
15c10423
JS
1769 if (pid < 0 || status != 0) {
1770 pr_err("failed to launch mdmon. "
1771 "Array remains readonly\n");
9fe32043 1772 return -1;
15c10423 1773 }
8850ee3e
N
1774 }
1775 return 0;
1776}
1777
148acb7b
DW
1778__u32 random32(void)
1779{
1780 __u32 rv;
1781 int rfd = open("/dev/urandom", O_RDONLY);
1782 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1783 rv = random();
1784 if (rfd >= 0)
1785 close(rfd);
1786 return rv;
1787}
1788
0e600426 1789#ifndef MDASSEMBLE
edd8d13c
NB
1790int flush_metadata_updates(struct supertype *st)
1791{
1792 int sfd;
1793 if (!st->updates) {
1794 st->update_tail = NULL;
1795 return -1;
1796 }
1797
4dd2df09 1798 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
1799 if (sfd < 0)
1800 return -1;
1801
1802 while (st->updates) {
1803 struct metadata_update *mu = st->updates;
1804 st->updates = mu->next;
1805
1806 send_message(sfd, mu, 0);
1807 wait_reply(sfd, 0);
1808 free(mu->buf);
1809 free(mu);
1810 }
1811 ack(sfd, 0);
1812 wait_reply(sfd, 0);
1813 close(sfd);
1814 st->update_tail = NULL;
1815 return 0;
1816}
1817
1818void append_metadata_update(struct supertype *st, void *buf, int len)
1819{
1820
503975b9 1821 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
1822
1823 mu->buf = buf;
1824 mu->len = len;
1825 mu->space = NULL;
cb23f1f4 1826 mu->space_list = NULL;
edd8d13c
NB
1827 mu->next = NULL;
1828 *st->update_tail = mu;
1829 st->update_tail = &mu->next;
1830}
0e600426 1831#endif /* MDASSEMBLE */
a931db9e 1832
fe6729fa
NB
1833#ifdef __TINYC__
1834/* tinyc doesn't optimize this check in ioctl.h out ... */
1835unsigned int __invalid_size_argument_for_IOC = 0;
1836#endif
1837
6d11ec6f
AK
1838int experimental(void)
1839{
1840 if (check_env("MDADM_EXPERIMENTAL"))
1841 return 1;
1842 else {
e7b84f9d 1843 pr_err("To use this feature MDADM_EXPERIMENTAL"
65c83a80 1844 " environment variable has to be defined.\n");
6d11ec6f
AK
1845 return 0;
1846 }
1847}
1848
326727d9
AC
1849/* Pick all spares matching given criteria from a container
1850 * if min_size == 0 do not check size
1851 * if domlist == NULL do not check domains
1852 * if spare_group given add it to domains of each spare
1853 * metadata allows to test domains using metadata of destination array */
1854struct mdinfo *container_choose_spares(struct supertype *st,
1855 unsigned long long min_size,
1856 struct domainlist *domlist,
1857 char *spare_group,
1858 const char *metadata, int get_one)
1859{
1860 struct mdinfo *d, **dp, *disks = NULL;
1861
1862 /* get list of all disks in container */
1863 if (st->ss->getinfo_super_disks)
1864 disks = st->ss->getinfo_super_disks(st);
1865
1866 if (!disks)
1867 return disks;
1868 /* find spare devices on the list */
1869 dp = &disks->devs;
1870 disks->array.spare_disks = 0;
1871 while (*dp) {
1872 int found = 0;
1873 d = *dp;
1874 if (d->disk.state == 0) {
1875 /* check if size is acceptable */
1876 unsigned long long dev_size;
1877 dev_t dev = makedev(d->disk.major,d->disk.minor);
1878
1879 if (!min_size ||
1880 (dev_size_from_id(dev, &dev_size) &&
1881 dev_size >= min_size))
1882 found = 1;
1883 /* check if domain matches */
1884 if (found && domlist) {
4dd2df09 1885 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
1886 if (spare_group)
1887 pol_add(&pol, pol_domain,
1888 spare_group, NULL);
e5508b36 1889 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1890 found = 0;
1891 dev_policy_free(pol);
1892 }
1893 }
1894 if (found) {
1895 dp = &d->next;
1896 disks->array.spare_disks++;
1897 if (get_one) {
1898 sysfs_free(*dp);
1899 d->next = NULL;
1900 }
1901 } else {
1902 *dp = d->next;
1903 d->next = NULL;
1904 sysfs_free(d);
1905 }
1906 }
1907 return disks;
1908}
9eafa1de
MN
1909
1910/* Checks if paths point to the same device
1911 * Returns 0 if they do.
1912 * Returns 1 if they don't.
1913 * Returns -1 if something went wrong,
1914 * e.g. paths are empty or the files
1915 * they point to don't exist */
1916int compare_paths (char* path1, char* path2)
1917{
1918 struct stat st1,st2;
1919
1920 if (path1 == NULL || path2 == NULL)
1921 return -1;
1922 if (stat(path1,&st1) != 0)
1923 return -1;
1924 if (stat(path2,&st2) != 0)
1925 return -1;
1926 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
1927 return 0;
1928 return 1;
1929}
a7dec3fd
N
1930
1931/* Make sure we can open as many devices as needed */
1932void enable_fds(int devices)
1933{
1934 unsigned int fds = 20 + devices;
1935 struct rlimit lim;
1936 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
1937 || lim.rlim_cur >= fds)
1938 return;
1939 if (lim.rlim_max < fds)
1940 lim.rlim_max = fds;
1941 lim.rlim_cur = fds;
1942 setrlimit(RLIMIT_NOFILE, &lim);
1943}
a9c15847
N
1944
1945int in_initrd(void)
1946{
1947 /* This is based on similar function in systemd. */
1948 struct statfs s;
1949 return statfs("/", &s) >= 0 &&
7ac5d47e 1950 ((unsigned long)s.f_type == TMPFS_MAGIC ||
1951 (unsigned long)s.f_type == RAMFS_MAGIC);
a9c15847 1952}