tests/03r5assemV1: reduce sync speed further.
[thirdparty/mdadm.git] / util.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
e736b623 4 * Copyright (C) 2001-2009 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>
98c6faba 31#include <ctype.h>
a322f70c 32#include <dirent.h>
a931db9e 33#include <signal.h>
0a816ef9 34
a0963a86
JS
35int __offroot;
36
0a816ef9
NB
37/*
38 * following taken from linux/blkpg.h because they aren't
39 * anywhere else and it isn't safe to #include linux/ * stuff.
40 */
41
42#define BLKPG _IO(0x12,105)
43
44/* The argument structure */
45struct blkpg_ioctl_arg {
46 int op;
47 int flags;
48 int datalen;
49 void *data;
50};
51
52/* The subfunctions (for the op field) */
53#define BLKPG_ADD_PARTITION 1
54#define BLKPG_DEL_PARTITION 2
55
56/* Sizes of name fields. Unused at present. */
57#define BLKPG_DEVNAMELTH 64
58#define BLKPG_VOLNAMELTH 64
59
60/* The data structure for ADD_PARTITION and DEL_PARTITION */
61struct blkpg_partition {
62 long long start; /* starting offset in bytes */
63 long long length; /* length in bytes */
64 int pno; /* partition number */
65 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
66 to be used in kernel messages */
67 char volname[BLKPG_VOLNAMELTH]; /* volume label */
68};
64c4757e 69
0f22b998 70#include "part.h"
056b331e
N
71
72/* Force a compilation error if condition is true */
73#define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
74
75/* Force a compilation error if condition is true, but also produce a
76 result (of value 0 and type size_t), so the expression can be used
77 e.g. in a structure initializer (or where-ever else comma expressions
78 aren't permitted). */
79#define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
80
64c4757e
NB
81/*
82 * Parse a 128 bit uuid in 4 integers
83 * format is 32 hexx nibbles with options :.<space> separator
84 * If not exactly 32 hex digits are found, return 0
85 * else return 1
86 */
87int parse_uuid(char *str, int uuid[4])
88{
aba69144
NB
89 int hit = 0; /* number of Hex digIT */
90 int i;
91 char c;
92 for (i=0; i<4; i++) uuid[i]=0;
93
94 while ((c= *str++)) {
95 int n;
96 if (c>='0' && c<='9')
97 n = c-'0';
98 else if (c>='a' && c <= 'f')
99 n = 10 + c - 'a';
100 else if (c>='A' && c <= 'F')
101 n = 10 + c - 'A';
102 else if (strchr(":. -", c))
103 continue;
104 else return 0;
105
106 if (hit<32) {
107 uuid[hit/8] <<= 4;
108 uuid[hit/8] += n;
109 }
110 hit++;
82b27616 111 }
aba69144
NB
112 if (hit == 32)
113 return 1;
114 return 0;
64c4757e
NB
115}
116
117
118/*
119 * Get the md version number.
120 * We use the RAID_VERSION ioctl if it is supported
121 * If not, but we have a block device with major '9', we assume
122 * 0.36.0
123 *
124 * Return version number as 24 but number - assume version parts
125 * always < 255
126 */
127
128int md_get_version(int fd)
129{
130 struct stat stb;
131 mdu_version_t vers;
132
133 if (fstat(fd, &stb)<0)
134 return -1;
135 if ((S_IFMT&stb.st_mode) != S_IFBLK)
136 return -1;
137
138 if (ioctl(fd, RAID_VERSION, &vers) == 0)
682c7051 139 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
5787fa49
NB
140 if (errno == EACCES)
141 return -1;
0df46c2a 142 if (major(stb.st_rdev) == MD_MAJOR)
682c7051 143 return (3600);
64c4757e
NB
144 return -1;
145}
146
64c4757e
NB
147int get_linux_version()
148{
149 struct utsname name;
98c6faba 150 char *cp;
f161d047 151 int a = 0, b = 0,c = 0;
64c4757e
NB
152 if (uname(&name) <0)
153 return -1;
154
98c6faba
NB
155 cp = name.release;
156 a = strtoul(cp, &cp, 10);
f161d047
N
157 if (*cp == '.')
158 b = strtoul(cp+1, &cp, 10);
159 if (*cp == '.')
160 c = strtoul(cp+1, &cp, 10);
98c6faba 161
682c7051 162 return (a*1000000)+(b*1000)+c;
64c4757e
NB
163}
164
71204a50 165#ifndef MDASSEMBLE
bc77ed53
DW
166int mdadm_version(char *version)
167{
168 int a, b, c;
169 char *cp;
170
171 if (!version)
172 version = Version;
173
174 cp = strchr(version, '-');
175 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
176 return -1;
177 cp += 3;
178 a = strtoul(cp, &cp, 10);
179 if (*cp != '.')
180 return -1;
181 b = strtoul(cp+1, &cp, 10);
182 if (*cp == '.')
183 c = strtoul(cp+1, &cp, 10);
184 else
185 c = 0;
186 if (*cp != ' ' && *cp != '-')
187 return -1;
188 return (a*1000000)+(b*1000)+c;
189}
190
84e11361
N
191long long parse_size(char *size)
192{
193 /* parse 'size' which should be a number optionally
194 * followed by 'K', 'M', or 'G'.
195 * Without a suffix, K is assumed.
196 * Number returned is in sectors (half-K)
15632a96 197 * -1 returned on error.
84e11361
N
198 */
199 char *c;
200 long long s = strtoll(size, &c, 10);
201 if (s > 0) {
202 switch (*c) {
203 case 'K':
204 c++;
205 default:
206 s *= 2;
207 break;
208 case 'M':
209 c++;
210 s *= 1024 * 2;
211 break;
212 case 'G':
213 c++;
214 s *= 1024 * 1024 * 2;
215 break;
216 }
217 }
218 if (*c)
15632a96 219 s = -1;
84e11361
N
220 return s;
221}
222
4a06e2c2
N
223int parse_layout_10(char *layout)
224{
225 int copies, rv;
226 char *cp;
227 /* Parse the layout string for raid10 */
228 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
229 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
230 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
231 copies > 200 ||
232 *cp)
233 return -1;
234 if (layout[0] == 'n')
235 rv = 256 + copies;
236 else if (layout[0] == 'o')
237 rv = 0x10000 + (copies<<8) + 1;
238 else
239 rv = 1 + (copies<<8);
240 return rv;
241}
242
243int parse_layout_faulty(char *layout)
244{
245 /* Parse the layout string for 'faulty' */
246 int ln = strcspn(layout, "0123456789");
503975b9 247 char *m = xstrdup(layout);
4a06e2c2
N
248 int mode;
249 m[ln] = 0;
250 mode = map_name(faultylayout, m);
251 if (mode == UnSet)
252 return -1;
253
254 return mode | (atoi(layout+ln)<< ModeShift);
255}
79868890
N
256
257long parse_num(char *num)
258{
259 /* Either return a valid number, or -1 */
260 char *c;
261 long rv = strtol(num, &c, 10);
262 if (rv < 0 || *c || !num[0])
263 return -1;
264 else
265 return rv;
266}
eb3929a4 267#endif
4a06e2c2 268
0430ed48
NB
269void remove_partitions(int fd)
270{
271 /* remove partitions from this block devices.
272 * This is used for components added to an array
273 */
274#ifdef BLKPG_DEL_PARTITION
275 struct blkpg_ioctl_arg a;
276 struct blkpg_partition p;
277
278 a.op = BLKPG_DEL_PARTITION;
279 a.data = (void*)&p;
280 a.datalen = sizeof(p);
281 a.flags = 0;
282 memset(a.data, 0, a.datalen);
283 for (p.pno=0; p.pno < 16; p.pno++)
284 ioctl(fd, BLKPG, &a);
285#endif
286}
287
691c6ee1
N
288int test_partition(int fd)
289{
290 /* Check if fd is a whole-disk or a partition.
291 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
292 * will return ENXIO on an invalid partition number.
293 */
294 struct blkpg_ioctl_arg a;
295 struct blkpg_partition p;
296 a.op = BLKPG_DEL_PARTITION;
297 a.data = (void*)&p;
298 a.datalen = sizeof(p);
299 a.flags = 0;
300 memset(a.data, 0, a.datalen);
301 p.pno = 1<<30;
302 if (ioctl(fd, BLKPG, &a) == 0)
303 /* Very unlikely, but not a partition */
304 return 0;
305 if (errno == ENXIO)
306 /* not a partition */
307 return 0;
308
309 return 1;
310}
311
bfd76b93
CA
312int test_partition_from_id(dev_t id)
313{
314 char buf[20];
315 int fd, rv;
316
317 sprintf(buf, "%d:%d", major(id), minor(id));
318 fd = dev_open(buf, O_RDONLY);
319 if (fd < 0)
320 return -1;
321 rv = test_partition(fd);
322 close(fd);
323 return rv;
324}
691c6ee1 325
de5a472e 326int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 327{
265e0f17 328 int copies, first;
de5a472e
N
329 int i;
330 int avail_disks = 0;
331
332 for (i = 0; i < raid_disks; i++)
333 avail_disks += !!avail[i];
334
64c4757e 335 switch (level) {
265e0f17
NB
336 case 10:
337 /* This is the tricky one - we need to check
338 * which actual disks are present.
339 */
702b557b 340 copies = (layout&255)* ((layout>>8) & 255);
265e0f17
NB
341 first=0;
342 do {
343 /* there must be one of the 'copies' form 'first' */
344 int n = copies;
345 int cnt=0;
346 while (n--) {
347 if (avail[first])
348 cnt++;
349 first = (first+1) % raid_disks;
350 }
351 if (cnt == 0)
352 return 0;
353
354 } while (first != 0);
355 return 1;
e5329c37 356
df0d4ea0 357 case LEVEL_MULTIPATH:
e0d19036 358 return avail_disks>= 1;
df0d4ea0 359 case LEVEL_LINEAR:
64c4757e
NB
360 case 0:
361 return avail_disks == raid_disks;
362 case 1:
363 return avail_disks >= 1;
364 case 4:
365 case 5:
583315d9
NB
366 if (clean)
367 return avail_disks >= raid_disks-1;
368 else
369 return avail_disks >= raid_disks;
98c6faba 370 case 6:
583315d9
NB
371 if (clean)
372 return avail_disks >= raid_disks-2;
373 else
374 return avail_disks >= raid_disks;
64c4757e
NB
375 default:
376 return 0;
377 }
378}
379
8453e704
N
380int enough_fd(int fd)
381{
382 struct mdu_array_info_s array;
383 struct mdu_disk_info_s disk;
384 int avail_disks = 0;
aabe020d 385 int i, rv;
8453e704
N
386 char *avail;
387
388 if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 ||
389 array.raid_disks <= 0)
390 return 0;
503975b9 391 avail = xcalloc(array.raid_disks, 1);
480f3566 392 for (i=0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704
N
393 disk.number = i;
394 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
395 continue;
9e6d9291
N
396 if (disk.major == 0 && disk.minor == 0)
397 continue;
398 array.nr_disks--;
399
8453e704
N
400 if (! (disk.state & (1<<MD_DISK_SYNC)))
401 continue;
402 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
403 continue;
404 avail_disks++;
405 avail[disk.raid_disk] = 1;
406 }
407 /* This is used on an active array, so assume it is clean */
aabe020d 408 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 409 1, avail);
aabe020d
JS
410 free(avail);
411 return rv;
8453e704
N
412}
413
414
22e263f6
AC
415const int uuid_zero[4] = { 0, 0, 0, 0 };
416
f277ce36 417int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 418{
f277ce36
NB
419 if (swapuuid) {
420 /* parse uuids are hostendian.
421 * uuid's from some superblocks are big-ending
aba69144 422 * if there is a difference, we need to swap..
f277ce36
NB
423 */
424 unsigned char *ac = (unsigned char *)a;
425 unsigned char *bc = (unsigned char *)b;
426 int i;
427 for (i=0; i<16; i+= 4) {
428 if (ac[i+0] != bc[i+3] ||
429 ac[i+1] != bc[i+2] ||
430 ac[i+2] != bc[i+1] ||
431 ac[i+3] != bc[i+0])
432 return 0;
433 }
434 return 1;
435 } else {
436 if (a[0]==b[0] &&
437 a[1]==b[1] &&
438 a[2]==b[2] &&
439 a[3]==b[3])
440 return 1;
441 return 0;
442 }
64c4757e 443}
350f29f9
NB
444void copy_uuid(void *a, int b[4], int swapuuid)
445{
446 if (swapuuid) {
447 /* parse uuids are hostendian.
448 * uuid's from some superblocks are big-ending
449 * if there is a difference, we need to swap..
450 */
451 unsigned char *ac = (unsigned char *)a;
452 unsigned char *bc = (unsigned char *)b;
453 int i;
454 for (i=0; i<16; i+= 4) {
455 ac[i+0] = bc[i+3];
456 ac[i+1] = bc[i+2];
457 ac[i+2] = bc[i+1];
458 ac[i+3] = bc[i+0];
459 }
460 } else
461 memcpy(a, b, 16);
462}
64c4757e 463
aae5a112 464char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 465{
9968e376 466 int i, j;
d7288ddc
N
467 char uuid[16];
468 char *c = buf;
469 strcpy(c, "UUID-");
470 c += strlen(c);
aae5a112 471 copy_uuid(uuid, id, swap);
9968e376 472 for (i = 0; i < 4; i++) {
9968e376 473 if (i)
ff54de6e 474 *c++ = sep;
9968e376
DW
475 for (j = 3; j >= 0; j--) {
476 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
477 c+= 2;
478 }
d7288ddc
N
479 }
480 return buf;
aae5a112
DW
481
482}
483
484char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
485{
86983cce
N
486 // dirty hack to work around an issue with super1 superblocks...
487 // super1 superblocks need swapuuid set in order for assembly to
488 // work, but can't have it set if we want this printout to match
489 // all the other uuid printouts in super1.c, so we force swapuuid
490 // to 1 to make our printout match the rest of super1
491 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
492}
493
435d4ebb 494#ifndef MDASSEMBLE
682c7051
NB
495int check_ext2(int fd, char *name)
496{
497 /*
498 * Check for an ext2fs file system.
499 * Superblock is always 1K at 1K offset
500 *
501 * s_magic is le16 at 56 == 0xEF53
502 * report mtime - le32 at 44
503 * blocks - le32 at 4
504 * logblksize - le32 at 24
505 */
506 unsigned char sb[1024];
507 time_t mtime;
508 int size, bsize;
509 if (lseek(fd, 1024,0)!= 1024)
510 return 0;
511 if (read(fd, sb, 1024)!= 1024)
512 return 0;
513 if (sb[56] != 0x53 || sb[57] != 0xef)
514 return 0;
515
516 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
517 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
518 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
e7b84f9d 519 pr_err("%s appears to contain an ext2fs file system\n",
682c7051
NB
520 name);
521 fprintf(stderr," size=%dK mtime=%s",
522 size*(1<<bsize), ctime(&mtime));
523 return 1;
524}
525
526int check_reiser(int fd, char *name)
527{
528 /*
529 * superblock is at 64K
530 * size is 1024;
531 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
532 *
533 */
534 unsigned char sb[1024];
881990a2 535 unsigned long size;
682c7051
NB
536 if (lseek(fd, 64*1024, 0) != 64*1024)
537 return 0;
538 if (read(fd, sb, 1024) != 1024)
539 return 0;
a46f4061
NB
540 if (strncmp((char*)sb+52, "ReIsErFs",8)!=0 &&
541 strncmp((char*)sb+52, "ReIsEr2Fs",9)!=0)
682c7051 542 return 0;
e7b84f9d 543 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 544 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
881990a2 545 fprintf(stderr, " size = %luK\n", size*4);
aba69144 546
682c7051
NB
547 return 1;
548}
549
550int check_raid(int fd, char *name)
551{
4b1ac34b 552 struct mdinfo info;
682c7051 553 time_t crtime;
d078d77c 554 char *level;
82d9eba6 555 struct supertype *st = guess_super(fd);
f9ce90ba 556
82d9eba6 557 if (!st) return 0;
4a96d9ff 558 st->ignore_hw_compat = 1;
3da92f27 559 st->ss->load_super(st, fd, name);
82d9eba6 560 /* Looks like a raid array .. */
e7b84f9d 561 pr_err("%s appears to be part of a raid array:\n",
82d9eba6 562 name);
a5d85af7 563 st->ss->getinfo_super(st, &info, NULL);
3da92f27 564 st->ss->free_super(st);
82d9eba6 565 crtime = info.array.ctime;
d078d77c
NB
566 level = map_num(pers, info.array.level);
567 if (!level) level = "-unknown-";
568 fprintf(stderr, " level=%s devices=%d ctime=%s",
569 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 570 return 1;
682c7051
NB
571}
572
682c7051
NB
573int ask(char *mesg)
574{
575 char *add = "";
576 int i;
577 for (i=0; i<5; i++) {
578 char buf[100];
579 fprintf(stderr, "%s%s", mesg, add);
580 fflush(stderr);
581 if (fgets(buf, 100, stdin)==NULL)
582 return 0;
583 if (buf[0]=='y' || buf[0]=='Y')
584 return 1;
585 if (buf[0]=='n' || buf[0]=='N')
586 return 0;
587 add = "(y/n) ";
588 }
e7b84f9d 589 pr_err("assuming 'no'\n");
682c7051
NB
590 return 0;
591}
435d4ebb 592#endif /* MDASSEMBLE */
682c7051 593
8d80900b 594int is_standard(char *dev, int *nump)
e5329c37
NB
595{
596 /* tests if dev is a "standard" md dev name.
597 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 598 * where NN is a string of digits
598f0d58
NB
599 * Returns 1 if a partitionable standard,
600 * -1 if non-partitonable,
601 * 0 if not a standard name.
e5329c37 602 */
8d80900b
NB
603 char *d = strrchr(dev, '/');
604 int type=0;
605 int num;
606 if (!d)
e5329c37 607 return 0;
8d80900b
NB
608 if (strncmp(d, "/d",2)==0)
609 d += 2, type=1; /* /dev/md/dN{pM} */
610 else if (strncmp(d, "/md_d", 5)==0)
2b4ca8f0 611 d += 5, type=1; /* /dev/md_dN{pM} */
8d80900b
NB
612 else if (strncmp(d, "/md", 3)==0)
613 d += 3, type=-1; /* /dev/mdN */
614 else if (d-dev > 3 && strncmp(d-2, "md/", 3)==0)
5a6d1148 615 d += 1, type=-1; /* /dev/md/N */
e5329c37
NB
616 else
617 return 0;
8d80900b 618 if (!*d)
e5329c37 619 return 0;
8d80900b
NB
620 num = atoi(d);
621 while (isdigit(*d))
622 d++;
623 if (*d)
e5329c37 624 return 0;
8d80900b
NB
625 if (nump) *nump = num;
626
627 return type;
e5329c37
NB
628}
629
4b1ac34b 630unsigned long calc_csum(void *super, int bytes)
82b27616 631{
56eb10c0 632 unsigned long long newcsum = 0;
82b27616 633 int i;
4b1ac34b
NB
634 unsigned int csum;
635 unsigned int *superc = (unsigned int*) super;
82b27616 636
4b1ac34b 637 for(i=0; i<bytes/4; i++)
82b27616
NB
638 newcsum+= superc[i];
639 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 640#ifdef __alpha__
aba69144 641/* The in-kernel checksum calculation is always 16bit on
570c0542
NB
642 * the alpha, though it is 32 bit on i386...
643 * I wonder what it is elsewhere... (it uses and API in
644 * a way that it shouldn't).
645 */
646 csum = (csum & 0xffff) + (csum >> 16);
647 csum = (csum & 0xffff) + (csum >> 16);
648#endif
82b27616
NB
649 return csum;
650}
cd29a5c8 651
435d4ebb 652#ifndef MDASSEMBLE
56eb10c0 653char *human_size(long long bytes)
cd29a5c8
NB
654{
655 static char buf[30];
d5d3721e
NB
656
657 /* We convert bytes to either centi-M{ega,ibi}bytes or
658 * centi-G{igi,ibi}bytes, with appropriate rounding,
659 * and then print 1/100th of those as a decimal.
660 * We allow upto 2048Megabytes before converting to
661 * gigabytes, as that shows more precision and isn't
662 * too large a number.
01619b48 663 * Terabytes are not yet handled.
d5d3721e 664 */
cd29a5c8 665
56eb10c0 666 if (bytes < 5000*1024)
cd29a5c8 667 buf[0]=0;
d5d3721e
NB
668 else if (bytes < 2*1024LL*1024LL*1024LL) {
669 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
670 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 671 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
672 cMiB/100 , cMiB % 100,
673 cMB/100, cMB % 100);
674 } else {
675 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
676 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 677 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
678 cGiB/100 , cGiB % 100,
679 cGB/100, cGB % 100);
680 }
cd29a5c8
NB
681 return buf;
682}
e0d19036
NB
683
684char *human_size_brief(long long bytes)
685{
686 static char buf[30];
e0d19036
NB
687
688 if (bytes < 5000*1024)
8f23b0b3 689 snprintf(buf, sizeof(buf), "%ld.%02ldKiB",
bd526cee 690 (long)(bytes>>10), (long)(((bytes&1023)*100+512)/1024)
e0d19036
NB
691 );
692 else if (bytes < 2*1024LL*1024LL*1024LL)
8f23b0b3 693 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
e0d19036 694 (long)(bytes>>20),
bd526cee 695 (long)((bytes&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
696 );
697 else
8f23b0b3 698 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
e0d19036 699 (long)(bytes>>30),
bd526cee 700 (long)(((bytes>>10)&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
701 );
702 return buf;
703}
e4965ef8
N
704
705void print_r10_layout(int layout)
706{
707 int near = layout & 255;
708 int far = (layout >> 8) & 255;
709 int offset = (layout&0x10000);
710 char *sep = "";
711
712 if (near != 1) {
713 printf("%s near=%d", sep, near);
714 sep = ",";
715 }
716 if (far != 1)
717 printf("%s %s=%d", sep, offset?"offset":"far", far);
718 if (near*far == 1)
719 printf("NO REDUNDANCY");
720}
435d4ebb 721#endif
e0d19036 722
5f8097be
NB
723unsigned long long calc_array_size(int level, int raid_disks, int layout,
724 int chunksize, unsigned long long devsize)
577e8448 725{
e03640bd
CA
726 if (level == 1)
727 return devsize;
577e8448
AK
728 devsize &= ~(unsigned long long)((chunksize>>9)-1);
729 return get_data_disks(level, layout, raid_disks) * devsize;
730}
731
732int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
733{
734 int data_disks = 0;
735 switch (level) {
736 case 0: data_disks = raid_disks; break;
737 case 1: data_disks = 1; break;
738 case 4:
739 case 5: data_disks = raid_disks - 1; break;
740 case 6: data_disks = raid_disks - 2; break;
741 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
742 break;
743 }
577e8448
AK
744
745 return data_disks;
5f8097be
NB
746}
747
0e600426 748#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
e0d19036
NB
749char *get_md_name(int dev)
750{
751 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 752 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
e0d19036
NB
753 static char devname[50];
754 struct stat stb;
98c6faba 755 dev_t rdev;
dd0781e5 756 char *dn;
98c6faba
NB
757
758 if (dev < 0) {
dd0781e5
NB
759 int mdp = get_mdp_major();
760 if (mdp < 0) return NULL;
0df46c2a 761 rdev = makedev(mdp, (-1-dev)<<6);
8f23b0b3 762 snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev);
98c6faba
NB
763 if (stat(devname, &stb) == 0
764 && (S_IFMT&stb.st_mode) == S_IFBLK
765 && (stb.st_rdev == rdev))
766 return devname;
767 } else {
0df46c2a 768 rdev = makedev(MD_MAJOR, dev);
8f23b0b3 769 snprintf(devname, sizeof(devname), "/dev/md%d", dev);
98c6faba
NB
770 if (stat(devname, &stb) == 0
771 && (S_IFMT&stb.st_mode) == S_IFBLK
772 && (stb.st_rdev == rdev))
773 return devname;
774
8f23b0b3 775 snprintf(devname, sizeof(devname), "/dev/md/%d", dev);
98c6faba
NB
776 if (stat(devname, &stb) == 0
777 && (S_IFMT&stb.st_mode) == S_IFBLK
778 && (stb.st_rdev == rdev))
779 return devname;
780 }
16c6fa80 781 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
782 if (dn)
783 return dn;
8f23b0b3 784 snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev);
e0d19036 785 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
786 if (errno != EEXIST)
787 return NULL;
e0d19036
NB
788
789 if (stat(devname, &stb) == 0
790 && (S_IFMT&stb.st_mode) == S_IFBLK
791 && (stb.st_rdev == rdev))
792 return devname;
793 unlink(devname);
794 return NULL;
795}
796
797void put_md_name(char *name)
798{
799 if (strncmp(name, "/dev/.tmp.md", 12)==0)
800 unlink(name);
801}
ea24acd0 802
ea24acd0
NB
803int find_free_devnum(int use_partitions)
804{
805 int devnum;
806 for (devnum = 127; devnum != 128;
a56fb7ec 807 devnum = devnum ? devnum-1 : (1<<20)-1) {
ea24acd0
NB
808 char *dn;
809 int _devnum;
810
811 _devnum = use_partitions ? (-1-devnum) : devnum;
812 if (mddev_busy(_devnum))
813 continue;
814 /* make sure it is new to /dev too, at least as a
815 * non-standard */
816 dn = map_dev(dev2major(_devnum), dev2minor(_devnum), 0);
817 if (dn && ! is_standard(dn, NULL))
818 continue;
819 break;
820 }
821 if (devnum == 128)
822 return NoMdDev;
823 return use_partitions ? (-1-devnum) : devnum;
824}
435d4ebb 825#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 826
8b0dabea
NB
827int dev_open(char *dev, int flags)
828{
829 /* like 'open', but if 'dev' matches %d:%d, create a temp
830 * block device and open that
831 */
832 char *e;
833 int fd = -1;
834 char devname[32];
e81cdd9f 835 int major;
8b0dabea 836 int minor;
e81cdd9f
NB
837
838 if (!dev) return -1;
6df6a774 839 flags |= O_DIRECT;
e81cdd9f
NB
840
841 major = strtoul(dev, &e, 0);
8b0dabea
NB
842 if (e > dev && *e == ':' && e[1] &&
843 (minor = strtoul(e+1, &e, 0)) >= 0 &&
844 *e == 0) {
6df6a774
N
845 char *path = map_dev(major, minor, 0);
846 if (path)
847 fd = open(path, flags);
848 if (fd < 0) {
849 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
850 (int)getpid(), major, minor);
851 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
852 fd = open(devname, flags);
853 unlink(devname);
854 }
855 }
856 if (fd < 0) {
857 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
858 (int)getpid(), major, minor);
859 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
860 fd = open(devname, flags);
861 unlink(devname);
862 }
8b0dabea
NB
863 }
864 } else
6df6a774 865 fd = open(dev, flags);
8b0dabea
NB
866 return fd;
867}
f9ce90ba 868
d998b738 869int open_dev_flags(int devnum, int flags)
e8a70c89
N
870{
871 char buf[20];
872
873 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
d998b738
N
874 return dev_open(buf, flags);
875}
876
877int open_dev(int devnum)
878{
879 return open_dev_flags(devnum, O_RDONLY);
e8a70c89
N
880}
881
a931db9e
NB
882int open_dev_excl(int devnum)
883{
884 char buf[20];
885 int i;
7187750e 886 int flags = O_RDWR;
a931db9e
NB
887
888 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
889 for (i=0 ; i<25 ; i++) {
7187750e 890 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
891 if (fd >= 0)
892 return fd;
7187750e
N
893 if (errno == EACCES && flags == O_RDWR) {
894 flags = O_RDONLY;
895 continue;
896 }
a931db9e
NB
897 if (errno != EBUSY)
898 return fd;
899 usleep(200000);
900 }
901 return -1;
902}
903
9008ed1c
N
904int same_dev(char *one, char *two)
905{
906 struct stat st1, st2;
907 if (stat(one, &st1) != 0)
908 return 0;
909 if (stat(two, &st2) != 0)
910 return 0;
911 if ((st1.st_mode & S_IFMT) != S_IFBLK)
912 return 0;
913 if ((st2.st_mode & S_IFMT) != S_IFBLK)
914 return 0;
915 return st1.st_rdev == st2.st_rdev;
916}
917
a7c6e3fb 918void wait_for(char *dev, int fd)
a714580e
N
919{
920 int i;
a7c6e3fb
N
921 struct stat stb_want;
922
923 if (fstat(fd, &stb_want) != 0 ||
924 (stb_want.st_mode & S_IFMT) != S_IFBLK)
925 return;
a714580e
N
926
927 for (i=0 ; i<25 ; i++) {
928 struct stat stb;
a7c6e3fb
N
929 if (stat(dev, &stb) == 0 &&
930 (stb.st_mode & S_IFMT) == S_IFBLK &&
931 (stb.st_rdev == stb_want.st_rdev))
a714580e
N
932 return;
933 usleep(200000);
934 }
436305c6
DW
935 if (i == 25)
936 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
937}
938
0f22b998
N
939struct superswitch *superlist[] =
940{
941 &super0, &super1,
942 &super_ddf, &super_imsm,
0592faeb 943 &mbr, &gpt,
0f22b998 944 NULL };
f9ce90ba 945
ea24acd0 946#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 947
4725bc31 948struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 949{
1686dc25
NB
950 mdu_array_info_t array;
951 int vers;
952 int minor;
953 struct supertype *st = NULL;
7e0f6979 954 struct mdinfo *sra;
142cb9e1 955 char *verstr;
1686dc25
NB
956 char version[20];
957 int i;
f7e7067b 958 char *subarray = NULL;
5f7e44b2 959 int container = NoMdDev;
1686dc25
NB
960
961 sra = sysfs_read(fd, 0, GET_VERSION);
962
963 if (sra) {
7e0f6979
NB
964 vers = sra->array.major_version;
965 minor = sra->array.minor_version;
142cb9e1 966 verstr = sra->text_version;
1686dc25
NB
967 } else {
968 if (ioctl(fd, GET_ARRAY_INFO, &array))
969 array.major_version = array.minor_version = 0;
970 vers = array.major_version;
971 minor = array.minor_version;
142cb9e1 972 verstr = "";
6fbba4c9 973 }
82d9eba6 974
1686dc25
NB
975 if (vers != -1) {
976 sprintf(version, "%d.%d", vers, minor);
977 verstr = version;
6fbba4c9 978 }
3c558363 979 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 980 char *dev = verstr+1;
5f7e44b2 981
f7e7067b 982 subarray = strchr(dev, '/');
92d49ecf 983 if (subarray) {
f7e7067b 984 *subarray++ = '\0';
503975b9 985 subarray = xstrdup(subarray);
92d49ecf 986 }
5f7e44b2 987 container = devname2devnum(dev);
f7e7067b
NB
988 if (sra)
989 sysfs_free(sra);
5f7e44b2 990 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
991 if (sra && sra->text_version[0])
992 verstr = sra->text_version;
993 else
994 verstr = "-no-metadata-";
f7e7067b
NB
995 }
996
997 for (i = 0; st == NULL && superlist[i] ; i++)
998 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
999
1000 if (sra)
1001 sysfs_free(sra);
f7e7067b 1002 if (st) {
3b0896f8 1003 st->sb = NULL;
1f49fb3a
N
1004 if (subarrayp)
1005 *subarrayp = subarray;
5f7e44b2
DW
1006 st->container_dev = container;
1007 st->devnum = fd2devnum(fd);
4725bc31
N
1008 } else
1009 free(subarray);
5f7e44b2 1010
82d9eba6 1011 return st;
f9ce90ba 1012}
ea24acd0
NB
1013#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1014
0f0749ad 1015int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1016{
1017 char buf[20];
1018 int fd;
1019
1020 sprintf(buf, "%d:%d", major(id), minor(id));
1021 fd = dev_open(buf, O_RDONLY);
1022 if (fd < 0)
1023 return 0;
1024 if (get_dev_size(fd, NULL, size)) {
1025 close(fd);
1026 return 1;
1027 }
1028 close(fd);
1029 return 0;
1030}
f9ce90ba 1031
159c3a1a 1032struct supertype *dup_super(struct supertype *orig)
3da92f27 1033{
159c3a1a 1034 struct supertype *st;
1686dc25 1035
d2ca6449
NB
1036 if (!orig)
1037 return orig;
503975b9 1038 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1039 st->ss = orig->ss;
1040 st->max_devs = orig->max_devs;
1041 st->minor_version = orig->minor_version;
1042 st->sb = NULL;
1043 st->info = NULL;
1044 return st;
3da92f27
NB
1045}
1046
54887ad8 1047struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1048{
1049 /* try each load_super to find the best match,
1050 * and return the best superswitch
1051 */
82d9eba6
NB
1052 struct superswitch *ss;
1053 struct supertype *st;
f21e18ca 1054 time_t besttime = 0;
570c0542 1055 int bestsuper = -1;
f9ce90ba
NB
1056 int i;
1057
503975b9 1058 st = xcalloc(1, sizeof(*st));
d1d599ea
N
1059 st->container_dev = NoMdDev;
1060
f9ce90ba
NB
1061 for (i=0 ; superlist[i]; i++) {
1062 int rv;
1063 ss = superlist[i];
54887ad8
N
1064 if (guess_type == guess_array && ss->add_to_super == NULL)
1065 continue;
1066 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1067 continue;
ef609477 1068 memset(st, 0, sizeof(*st));
df3346e6 1069 st->ignore_hw_compat = 1;
3da92f27 1070 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1071 if (rv == 0) {
1072 struct mdinfo info;
a5d85af7 1073 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1074 if (bestsuper == -1 ||
1075 besttime < info.array.ctime) {
1076 bestsuper = i;
1077 besttime = info.array.ctime;
570c0542 1078 }
3da92f27 1079 ss->free_super(st);
570c0542
NB
1080 }
1081 }
1082 if (bestsuper != -1) {
1083 int rv;
ef609477 1084 memset(st, 0, sizeof(*st));
df3346e6 1085 st->ignore_hw_compat = 1;
3da92f27 1086 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1087 if (rv == 0) {
5e747af2 1088 superlist[bestsuper]->free_super(st);
df3346e6 1089 st->ignore_hw_compat = 0;
82d9eba6 1090 return st;
f9ce90ba
NB
1091 }
1092 }
570c0542 1093 free(st);
f9ce90ba
NB
1094 return NULL;
1095}
fe6729fa 1096
beae1dfe
NB
1097/* Return size of device in bytes */
1098int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1099{
1100 unsigned long long ldsize;
c2c9bb6f
NB
1101 struct stat st;
1102
1103 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1104 ldsize = (unsigned long long)st.st_size;
1105 else
beae1dfe
NB
1106#ifdef BLKGETSIZE64
1107 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1108#endif
1109 {
1110 unsigned long dsize;
1111 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1112 ldsize = dsize;
1113 ldsize <<= 9;
1114 } else {
1115 if (dname)
e7b84f9d 1116 pr_err("Cannot get size of %s: %s\b",
beae1dfe
NB
1117 dname, strerror(errno));
1118 return 0;
1119 }
1120 }
1121 *sizep = ldsize;
1122 return 1;
1123}
8fac0577 1124
3a371610
N
1125/* Return true if this can only be a container, not a member device.
1126 * i.e. is and md device and size is zero
1127 */
1128int must_be_container(int fd)
1129{
1130 unsigned long long size;
1131 if (md_get_version(fd) < 0)
1132 return 0;
1133 if (get_dev_size(fd, NULL, &size) == 0)
1134 return 1;
1135 if (size == 0)
1136 return 1;
1137 return 0;
1138}
034b203a
TM
1139
1140/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1141 * Returns: 1 if successful
1142 * -1 for unknown partition type
1143 * 0 for other errors
1144 */
1145static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1146{
056b331e 1147 struct GPT gpt;
034b203a 1148 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1149 struct GPT_part_entry *part;
1150 char buf[512];
034b203a
TM
1151 unsigned long long curr_part_end;
1152 unsigned all_partitions, entry_size;
f21e18ca 1153 unsigned part_nr;
034b203a
TM
1154
1155 *endofpart = 0;
1156
056b331e 1157 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1158 /* skip protective MBR */
034b203a 1159 lseek(fd, 512, SEEK_SET);
73e658d8 1160 /* read GPT header */
056b331e 1161 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1162 return 0;
1163
1164 /* get the number of partition entries and the entry size */
056b331e
N
1165 all_partitions = __le32_to_cpu(gpt.part_cnt);
1166 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1167
1168 /* Check GPT signature*/
056b331e 1169 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1170 return -1;
1171
1172 /* sanity checks */
1173 if (all_partitions > 1024 ||
db7fdfe4 1174 entry_size > sizeof(buf))
034b203a
TM
1175 return -1;
1176
db7fdfe4
JS
1177 part = (struct GPT_part_entry *)buf;
1178
034b203a 1179 for (part_nr=0; part_nr < all_partitions; part_nr++) {
73e658d8 1180 /* read partition entry */
db7fdfe4 1181 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1182 return 0;
1183
034b203a 1184 /* is this valid partition? */
db7fdfe4 1185 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1186 /* check the last lba for the current partition */
db7fdfe4 1187 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1188 if (curr_part_end > *endofpart)
1189 *endofpart = curr_part_end;
1190 }
1191
034b203a
TM
1192 }
1193 return 1;
1194}
1195
1196/* Sets endofpart parameter to the last block used by the last partition on the device.
1197 * Returns: 1 if successful
1198 * -1 for unknown partition type
1199 * 0 for other errors
1200 */
1201static int get_last_partition_end(int fd, unsigned long long *endofpart)
1202{
056b331e 1203 struct MBR boot_sect;
034b203a
TM
1204 struct MBR_part_record *part;
1205 unsigned long long curr_part_end;
f21e18ca 1206 unsigned part_nr;
034b203a
TM
1207 int retval = 0;
1208
1209 *endofpart = 0;
1210
056b331e 1211 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1212 /* read MBR */
1213 lseek(fd, 0, 0);
056b331e 1214 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1215 goto abort;
1216
1217 /* check MBP signature */
056b331e 1218 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1219 retval = 1;
1220 /* found the correct signature */
056b331e 1221 part = boot_sect.parts;
034b203a
TM
1222
1223 for (part_nr=0; part_nr < MBR_PARTITIONS; part_nr++) {
1224 /* check for GPT type */
1225 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1226 retval = get_gpt_last_partition_end(fd, endofpart);
1227 break;
1228 }
1229 /* check the last used lba for the current partition */
1230 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1231 __le32_to_cpu(part->blocks_num);
1232 if (curr_part_end > *endofpart)
1233 *endofpart = curr_part_end;
1234
1235 part++;
1236 }
1237 } else {
1238 /* Unknown partition table */
1239 retval = -1;
1240 }
1241 abort:
1242 return retval;
1243}
1244
53ed6ac3
KW
1245int check_partitions(int fd, char *dname, unsigned long long freesize,
1246 unsigned long long size)
034b203a
TM
1247{
1248 /*
1249 * Check where the last partition ends
1250 */
1251 unsigned long long endofpart;
1252 int ret;
1253
1254 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1255 /* There appears to be a partition table here */
1256 if (freesize == 0) {
1257 /* partitions will not be visible in new device */
e7b84f9d
N
1258 pr_err("partition table exists on %s but will be lost or\n"
1259 " meaningless after creating array\n",
1260 dname);
034b203a
TM
1261 return 1;
1262 } else if (endofpart > freesize) {
1263 /* last partition overlaps metadata */
e7b84f9d
N
1264 pr_err("metadata will over-write last partition on %s.\n",
1265 dname);
034b203a 1266 return 1;
53ed6ac3
KW
1267 } else if (size && endofpart > size) {
1268 /* partitions will be truncated in new device */
e7b84f9d
N
1269 pr_err("array size is too small to cover all partitions on %s.\n",
1270 dname);
53ed6ac3 1271 return 1;
034b203a
TM
1272 }
1273 }
1274 return 0;
1275}
1276
8382f19b
NB
1277void get_one_disk(int mdfd, mdu_array_info_t *ainf, mdu_disk_info_t *disk)
1278{
1279 int d;
9e6d9291 1280
8382f19b 1281 ioctl(mdfd, GET_ARRAY_INFO, ainf);
480f3566 1282 for (d = 0 ; d < MAX_DISKS ; d++) {
9e6d9291
N
1283 if (ioctl(mdfd, GET_DISK_INFO, disk) == 0 &&
1284 (disk->major || disk->minor))
8382f19b 1285 return;
9e6d9291 1286 }
8382f19b 1287}
63152c1b 1288
a322f70c
DW
1289int open_container(int fd)
1290{
1291 /* 'fd' is a block device. Find out if it is in use
1292 * by a container, and return an open fd on that container.
1293 */
1294 char path[256];
1295 char *e;
1296 DIR *dir;
1297 struct dirent *de;
1298 int dfd, n;
1299 char buf[200];
1300 int major, minor;
1301 struct stat st;
1302
1303 if (fstat(fd, &st) != 0)
1304 return -1;
1305 sprintf(path, "/sys/dev/block/%d:%d/holders",
1306 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1307 e = path + strlen(path);
1308
1309 dir = opendir(path);
1310 if (!dir)
1311 return -1;
1312 while ((de = readdir(dir))) {
1313 if (de->d_ino == 0)
1314 continue;
1315 if (de->d_name[0] == '.')
1316 continue;
1317 sprintf(e, "/%s/dev", de->d_name);
1318 dfd = open(path, O_RDONLY);
1319 if (dfd < 0)
1320 continue;
1321 n = read(dfd, buf, sizeof(buf));
1322 close(dfd);
f21e18ca 1323 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1324 continue;
1325 buf[n] = 0;
1326 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1327 continue;
1328 sprintf(buf, "%d:%d", major, minor);
1329 dfd = dev_open(buf, O_RDONLY);
1330 if (dfd >= 0) {
1331 closedir(dir);
1332 return dfd;
1333 }
1334 }
355726fa 1335 closedir(dir);
a322f70c
DW
1336 return -1;
1337}
1338
33414a01
DW
1339struct superswitch *version_to_superswitch(char *vers)
1340{
1341 int i;
1342
1343 for (i = 0; superlist[i]; i++) {
1344 struct superswitch *ss = superlist[i];
1345
1346 if (strcmp(vers, ss->name) == 0)
1347 return ss;
1348 }
1349
1350 return NULL;
1351}
1352
1353int is_container_member(struct mdstat_ent *mdstat, char *container)
1354{
1355 if (mdstat->metadata_version == NULL ||
1356 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
1357 !is_subarray(mdstat->metadata_version+9) ||
1358 strncmp(mdstat->metadata_version+10, container, strlen(container)) != 0 ||
1359 mdstat->metadata_version[10+strlen(container)] != '/')
1360 return 0;
1361
1362 return 1;
1363}
1364
1365int is_subarray_active(char *subarray, char *container)
1366{
1367 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1368 struct mdstat_ent *ent;
1369
e5408a32
DW
1370 for (ent = mdstat; ent; ent = ent->next)
1371 if (is_container_member(ent, container))
e5e5d7ce 1372 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1373 break;
33414a01
DW
1374
1375 free_mdstat(mdstat);
1376
1377 return ent != NULL;
1378}
1379
1380/* open_subarray - opens a subarray in a container
1381 * @dev: container device name
feab51f8 1382 * @st: empty supertype
33414a01
DW
1383 * @quiet: block reporting errors flag
1384 *
1385 * On success returns an fd to a container and fills in *st
1386 */
feab51f8 1387int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1388{
1389 struct mdinfo *mdi;
a951a4f7 1390 struct mdinfo *info;
33414a01
DW
1391 int fd, err = 1;
1392
1393 fd = open(dev, O_RDWR|O_EXCL);
1394 if (fd < 0) {
1395 if (!quiet)
e7b84f9d 1396 pr_err("Couldn't open %s, aborting\n",
33414a01 1397 dev);
b990032d 1398 return -1;
33414a01
DW
1399 }
1400
1401 st->devnum = fd2devnum(fd);
1402 if (st->devnum == NoMdDev) {
1403 if (!quiet)
e7b84f9d
N
1404 pr_err("Failed to determine device number for %s\n",
1405 dev);
33414a01
DW
1406 goto close_fd;
1407 }
1408
1409 mdi = sysfs_read(fd, st->devnum, GET_VERSION|GET_LEVEL);
1410 if (!mdi) {
1411 if (!quiet)
e7b84f9d 1412 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1413 dev);
1414 goto close_fd;
1415 }
1416
1417 if (mdi->array.level != UnSet) {
1418 if (!quiet)
e7b84f9d 1419 pr_err("%s is not a container\n", dev);
33414a01
DW
1420 goto free_sysfs;
1421 }
1422
1423 st->ss = version_to_superswitch(mdi->text_version);
1424 if (!st->ss) {
1425 if (!quiet)
e7b84f9d
N
1426 pr_err("Operation not supported for %s metadata\n",
1427 mdi->text_version);
33414a01
DW
1428 goto free_sysfs;
1429 }
1430
1431 st->devname = devnum2devname(st->devnum);
1432 if (!st->devname) {
1433 if (!quiet)
e7b84f9d 1434 pr_err("Failed to allocate device name\n");
33414a01
DW
1435 goto free_sysfs;
1436 }
1437
db20d413 1438 if (!st->ss->load_container) {
33414a01 1439 if (!quiet)
e7b84f9d 1440 pr_err("%s is not a container\n", dev);
33414a01
DW
1441 goto free_name;
1442 }
1443
db20d413 1444 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1445 if (!quiet)
e7b84f9d 1446 pr_err("Failed to load metadata for %s\n",
db20d413
N
1447 dev);
1448 goto free_name;
33414a01
DW
1449 }
1450
a951a4f7
N
1451 info = st->ss->container_content(st, subarray);
1452 if (!info) {
1453 if (!quiet)
e7b84f9d 1454 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1455 subarray, dev);
1456 goto free_super;
1457 }
1458 free(info);
1459
33414a01
DW
1460 err = 0;
1461
1462 free_super:
1463 if (err)
1464 st->ss->free_super(st);
1465 free_name:
1466 if (err)
1467 free(st->devname);
1468 free_sysfs:
1469 sysfs_free(mdi);
1470 close_fd:
1471 if (err)
1472 close(fd);
1473
1474 if (err)
1475 return -1;
1476 else
1477 return fd;
1478}
1479
7801ac20
N
1480int add_disk(int mdfd, struct supertype *st,
1481 struct mdinfo *sra, struct mdinfo *info)
1482{
1483 /* Add a device to an array, in one of 2 ways. */
1484 int rv;
1485#ifndef MDASSEMBLE
1486 if (st->ss->external) {
d23534e4
DW
1487 if (info->disk.state & (1<<MD_DISK_SYNC))
1488 info->recovery_start = MaxSector;
1489 else
1490 info->recovery_start = 0;
2904b26f 1491 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1492 if (! rv) {
1493 struct mdinfo *sd2;
f35f2525
N
1494 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1495 if (sd2 == info)
1496 break;
1497 if (sd2 == NULL) {
503975b9 1498 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1499 *sd2 = *info;
1500 sd2->next = sra->devs;
1501 sra->devs = sd2;
1502 }
7801ac20
N
1503 }
1504 } else
1505#endif
1506 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1507 return rv;
1508}
1509
de6ae750
N
1510int remove_disk(int mdfd, struct supertype *st,
1511 struct mdinfo *sra, struct mdinfo *info)
1512{
1513 int rv;
1514 /* Remove the disk given by 'info' from the array */
1515#ifndef MDASSEMBLE
1516 if (st->ss->external)
1517 rv = sysfs_set_str(sra, info, "slot", "none");
1518 else
1519#endif
1520 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1521 info->disk.minor));
1522 return rv;
1523}
1524
f35f2525
N
1525int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1526{
1527 /* Initialise kernel's knowledge of array.
1528 * This varies between externally managed arrays
1529 * and older kernels
1530 */
1531 int vers = md_get_version(mdfd);
1532 int rv;
1533
1534#ifndef MDASSEMBLE
1535 if (st->ss->external)
1536 rv = sysfs_set_array(info, vers);
1537 else
1538#endif
1539 if ((vers % 100) >= 1) { /* can use different versions */
1540 mdu_array_info_t inf;
1541 memset(&inf, 0, sizeof(inf));
1542 inf.major_version = info->array.major_version;
1543 inf.minor_version = info->array.minor_version;
1544 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1545 } else
1546 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1547 return rv;
1548}
1549
1e5c6983
DW
1550unsigned long long min_recovery_start(struct mdinfo *array)
1551{
1552 /* find the minimum recovery_start in an array for metadata
1553 * formats that only record per-array recovery progress instead
1554 * of per-device
1555 */
1556 unsigned long long recovery_start = MaxSector;
1557 struct mdinfo *d;
1558
1559 for (d = array->devs; d; d = d->next)
1560 recovery_start = min(recovery_start, d->recovery_start);
1561
1562 return recovery_start;
1563}
1564
24f6f99b 1565int mdmon_pid(int devnum)
a931db9e
NB
1566{
1567 char path[100];
1568 char pid[10];
1569 int fd;
1570 int n;
10013317
PHC
1571 char *devname = devnum2devname(devnum);
1572
753cf905 1573 sprintf(path, "%s/%s.pid", MDMON_DIR, devname);
10013317
PHC
1574 free(devname);
1575
24f6f99b 1576 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1577
1578 if (fd < 0)
cf556303 1579 return -1;
a931db9e
NB
1580 n = read(fd, pid, 9);
1581 close(fd);
1582 if (n <= 0)
cf556303 1583 return -1;
24f6f99b 1584 return atoi(pid);
a931db9e
NB
1585}
1586
24f6f99b 1587int mdmon_running(int devnum)
a931db9e 1588{
24f6f99b
N
1589 int pid = mdmon_pid(devnum);
1590 if (pid <= 0)
a931db9e 1591 return 0;
24f6f99b 1592 if (kill(pid, 0) == 0)
a931db9e
NB
1593 return 1;
1594 return 0;
1595}
1596
8850ee3e
N
1597int start_mdmon(int devnum)
1598{
1913c325 1599 int i, skipped;
44d2e365 1600 int len;
9fe32043
N
1601 pid_t pid;
1602 int status;
44d2e365
N
1603 char pathbuf[1024];
1604 char *paths[4] = {
1605 pathbuf,
1606 "/sbin/mdmon",
1607 "mdmon",
1608 NULL
1609 };
8850ee3e 1610
40ebbb9c 1611 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1612 return 0;
1613
9cf014ec 1614 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1615 if (len > 0) {
1616 char *sl;
1617 pathbuf[len] = 0;
1618 sl = strrchr(pathbuf, '/');
1619 if (sl)
1620 sl++;
1621 else
1622 sl = pathbuf;
1623 strcpy(sl, "mdmon");
1624 } else
1625 pathbuf[0] = '\0';
1626
8850ee3e
N
1627 switch(fork()) {
1628 case 0:
1629 /* FIXME yuk. CLOSE_EXEC?? */
1913c325
N
1630 skipped = 0;
1631 for (i=3; skipped < 20; i++)
1632 if (close(i) < 0)
1633 skipped++;
1634 else
1635 skipped = 0;
1636
44d2e365 1637 for (i=0; paths[i]; i++)
a0963a86
JS
1638 if (paths[i][0]) {
1639 if (__offroot) {
1640 execl(paths[i], "mdmon", "--offroot",
1641 devnum2devname(devnum),
1642 NULL);
1643 } else {
1644 execl(paths[i], "mdmon",
1645 devnum2devname(devnum),
1646 NULL);
1647 }
1648 }
8850ee3e 1649 exit(1);
e7b84f9d 1650 case -1: pr_err("cannot run mdmon. "
8850ee3e
N
1651 "Array remains readonly\n");
1652 return -1;
9fe32043
N
1653 default: /* parent - good */
1654 pid = wait(&status);
1655 if (pid < 0 || status != 0)
1656 return -1;
8850ee3e
N
1657 }
1658 return 0;
1659}
1660
40ebbb9c 1661int check_env(char *name)
5dcfcb71 1662{
40ebbb9c 1663 char *val = getenv(name);
5dcfcb71
DW
1664
1665 if (val && atoi(val) == 1)
1666 return 1;
1667
1668 return 0;
1669}
1670
148acb7b
DW
1671__u32 random32(void)
1672{
1673 __u32 rv;
1674 int rfd = open("/dev/urandom", O_RDONLY);
1675 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1676 rv = random();
1677 if (rfd >= 0)
1678 close(rfd);
1679 return rv;
1680}
1681
0e600426 1682#ifndef MDASSEMBLE
edd8d13c
NB
1683int flush_metadata_updates(struct supertype *st)
1684{
1685 int sfd;
1686 if (!st->updates) {
1687 st->update_tail = NULL;
1688 return -1;
1689 }
1690
1691 sfd = connect_monitor(devnum2devname(st->container_dev));
1692 if (sfd < 0)
1693 return -1;
1694
1695 while (st->updates) {
1696 struct metadata_update *mu = st->updates;
1697 st->updates = mu->next;
1698
1699 send_message(sfd, mu, 0);
1700 wait_reply(sfd, 0);
1701 free(mu->buf);
1702 free(mu);
1703 }
1704 ack(sfd, 0);
1705 wait_reply(sfd, 0);
1706 close(sfd);
1707 st->update_tail = NULL;
1708 return 0;
1709}
1710
1711void append_metadata_update(struct supertype *st, void *buf, int len)
1712{
1713
503975b9 1714 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
1715
1716 mu->buf = buf;
1717 mu->len = len;
1718 mu->space = NULL;
cb23f1f4 1719 mu->space_list = NULL;
edd8d13c
NB
1720 mu->next = NULL;
1721 *st->update_tail = mu;
1722 st->update_tail = &mu->next;
1723}
0e600426 1724#endif /* MDASSEMBLE */
a931db9e 1725
fe6729fa
NB
1726#ifdef __TINYC__
1727/* tinyc doesn't optimize this check in ioctl.h out ... */
1728unsigned int __invalid_size_argument_for_IOC = 0;
1729#endif
1730
6d11ec6f
AK
1731int experimental(void)
1732{
1733 if (check_env("MDADM_EXPERIMENTAL"))
1734 return 1;
1735 else {
e7b84f9d 1736 pr_err("To use this feature MDADM_EXPERIMENTAL"
65c83a80 1737 " environment variable has to be defined.\n");
6d11ec6f
AK
1738 return 0;
1739 }
1740}
1741
326727d9
AC
1742/* Pick all spares matching given criteria from a container
1743 * if min_size == 0 do not check size
1744 * if domlist == NULL do not check domains
1745 * if spare_group given add it to domains of each spare
1746 * metadata allows to test domains using metadata of destination array */
1747struct mdinfo *container_choose_spares(struct supertype *st,
1748 unsigned long long min_size,
1749 struct domainlist *domlist,
1750 char *spare_group,
1751 const char *metadata, int get_one)
1752{
1753 struct mdinfo *d, **dp, *disks = NULL;
1754
1755 /* get list of all disks in container */
1756 if (st->ss->getinfo_super_disks)
1757 disks = st->ss->getinfo_super_disks(st);
1758
1759 if (!disks)
1760 return disks;
1761 /* find spare devices on the list */
1762 dp = &disks->devs;
1763 disks->array.spare_disks = 0;
1764 while (*dp) {
1765 int found = 0;
1766 d = *dp;
1767 if (d->disk.state == 0) {
1768 /* check if size is acceptable */
1769 unsigned long long dev_size;
1770 dev_t dev = makedev(d->disk.major,d->disk.minor);
1771
1772 if (!min_size ||
1773 (dev_size_from_id(dev, &dev_size) &&
1774 dev_size >= min_size))
1775 found = 1;
1776 /* check if domain matches */
1777 if (found && domlist) {
1778 struct dev_policy *pol = devnum_policy(dev);
1779 if (spare_group)
1780 pol_add(&pol, pol_domain,
1781 spare_group, NULL);
e5508b36 1782 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1783 found = 0;
1784 dev_policy_free(pol);
1785 }
1786 }
1787 if (found) {
1788 dp = &d->next;
1789 disks->array.spare_disks++;
1790 if (get_one) {
1791 sysfs_free(*dp);
1792 d->next = NULL;
1793 }
1794 } else {
1795 *dp = d->next;
1796 d->next = NULL;
1797 sysfs_free(d);
1798 }
1799 }
1800 return disks;
1801}
503975b9
N
1802
1803void *xmalloc(size_t len)
1804{
1805 void *rv = malloc(len);
1806 char *msg;
1807 if (rv)
1808 return rv;
1809 msg = Name ": memory allocation failure - aborting\n";
1810 exit(4+!!write(2, msg, strlen(msg)));
1811}
1812
1813void *xrealloc(void *ptr, size_t len)
1814{
1815 void *rv = realloc(ptr, len);
1816 char *msg;
1817 if (rv)
1818 return rv;
1819 msg = Name ": memory allocation failure - aborting\n";
1820 exit(4+!!write(2, msg, strlen(msg)));
1821}
1822
1823void *xcalloc(size_t num, size_t size)
1824{
1825 void *rv = calloc(num, size);
1826 char *msg;
1827 if (rv)
1828 return rv;
1829 msg = Name ": memory allocation failure - aborting\n";
1830 exit(4+!!write(2, msg, strlen(msg)));
1831}
1832
1833char *xstrdup(const char *str)
1834{
1835 char *rv = strdup(str);
1836 char *msg;
1837 if (rv)
1838 return rv;
1839 msg = Name ": memory allocation failure - aborting\n";
1840 exit(4+!!write(2, msg, strlen(msg)));
1841}