man mdadm: Add note about auto-assembly during array reshape
[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
NB
34
35/*
36 * following taken from linux/blkpg.h because they aren't
37 * anywhere else and it isn't safe to #include linux/ * stuff.
38 */
39
40#define BLKPG _IO(0x12,105)
41
42/* The argument structure */
43struct blkpg_ioctl_arg {
44 int op;
45 int flags;
46 int datalen;
47 void *data;
48};
49
50/* The subfunctions (for the op field) */
51#define BLKPG_ADD_PARTITION 1
52#define BLKPG_DEL_PARTITION 2
53
54/* Sizes of name fields. Unused at present. */
55#define BLKPG_DEVNAMELTH 64
56#define BLKPG_VOLNAMELTH 64
57
58/* The data structure for ADD_PARTITION and DEL_PARTITION */
59struct blkpg_partition {
60 long long start; /* starting offset in bytes */
61 long long length; /* length in bytes */
62 int pno; /* partition number */
63 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
64 to be used in kernel messages */
65 char volname[BLKPG_VOLNAMELTH]; /* volume label */
66};
64c4757e 67
0f22b998 68#include "part.h"
056b331e
N
69
70/* Force a compilation error if condition is true */
71#define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
72
73/* Force a compilation error if condition is true, but also produce a
74 result (of value 0 and type size_t), so the expression can be used
75 e.g. in a structure initializer (or where-ever else comma expressions
76 aren't permitted). */
77#define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
78
64c4757e
NB
79/*
80 * Parse a 128 bit uuid in 4 integers
81 * format is 32 hexx nibbles with options :.<space> separator
82 * If not exactly 32 hex digits are found, return 0
83 * else return 1
84 */
85int parse_uuid(char *str, int uuid[4])
86{
aba69144
NB
87 int hit = 0; /* number of Hex digIT */
88 int i;
89 char c;
90 for (i=0; i<4; i++) uuid[i]=0;
91
92 while ((c= *str++)) {
93 int n;
94 if (c>='0' && c<='9')
95 n = c-'0';
96 else if (c>='a' && c <= 'f')
97 n = 10 + c - 'a';
98 else if (c>='A' && c <= 'F')
99 n = 10 + c - 'A';
100 else if (strchr(":. -", c))
101 continue;
102 else return 0;
103
104 if (hit<32) {
105 uuid[hit/8] <<= 4;
106 uuid[hit/8] += n;
107 }
108 hit++;
82b27616 109 }
aba69144
NB
110 if (hit == 32)
111 return 1;
112 return 0;
64c4757e
NB
113}
114
115
116/*
117 * Get the md version number.
118 * We use the RAID_VERSION ioctl if it is supported
119 * If not, but we have a block device with major '9', we assume
120 * 0.36.0
121 *
122 * Return version number as 24 but number - assume version parts
123 * always < 255
124 */
125
126int md_get_version(int fd)
127{
128 struct stat stb;
129 mdu_version_t vers;
130
131 if (fstat(fd, &stb)<0)
132 return -1;
133 if ((S_IFMT&stb.st_mode) != S_IFBLK)
134 return -1;
135
136 if (ioctl(fd, RAID_VERSION, &vers) == 0)
682c7051 137 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
5787fa49
NB
138 if (errno == EACCES)
139 return -1;
0df46c2a 140 if (major(stb.st_rdev) == MD_MAJOR)
682c7051 141 return (3600);
64c4757e
NB
142 return -1;
143}
144
64c4757e
NB
145int get_linux_version()
146{
147 struct utsname name;
98c6faba 148 char *cp;
64c4757e
NB
149 int a,b,c;
150 if (uname(&name) <0)
151 return -1;
152
98c6faba
NB
153 cp = name.release;
154 a = strtoul(cp, &cp, 10);
155 if (*cp != '.') return -1;
156 b = strtoul(cp+1, &cp, 10);
157 if (*cp != '.') return -1;
158 c = strtoul(cp+1, NULL, 10);
159
682c7051 160 return (a*1000000)+(b*1000)+c;
64c4757e
NB
161}
162
71204a50 163#ifndef MDASSEMBLE
bc77ed53
DW
164int mdadm_version(char *version)
165{
166 int a, b, c;
167 char *cp;
168
169 if (!version)
170 version = Version;
171
172 cp = strchr(version, '-');
173 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
174 return -1;
175 cp += 3;
176 a = strtoul(cp, &cp, 10);
177 if (*cp != '.')
178 return -1;
179 b = strtoul(cp+1, &cp, 10);
180 if (*cp == '.')
181 c = strtoul(cp+1, &cp, 10);
182 else
183 c = 0;
184 if (*cp != ' ' && *cp != '-')
185 return -1;
186 return (a*1000000)+(b*1000)+c;
187}
188
84e11361
N
189long long parse_size(char *size)
190{
191 /* parse 'size' which should be a number optionally
192 * followed by 'K', 'M', or 'G'.
193 * Without a suffix, K is assumed.
194 * Number returned is in sectors (half-K)
195 */
196 char *c;
197 long long s = strtoll(size, &c, 10);
198 if (s > 0) {
199 switch (*c) {
200 case 'K':
201 c++;
202 default:
203 s *= 2;
204 break;
205 case 'M':
206 c++;
207 s *= 1024 * 2;
208 break;
209 case 'G':
210 c++;
211 s *= 1024 * 1024 * 2;
212 break;
213 }
214 }
215 if (*c)
216 s = 0;
217 return s;
218}
219
4a06e2c2
N
220int parse_layout_10(char *layout)
221{
222 int copies, rv;
223 char *cp;
224 /* Parse the layout string for raid10 */
225 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
226 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
227 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
228 copies > 200 ||
229 *cp)
230 return -1;
231 if (layout[0] == 'n')
232 rv = 256 + copies;
233 else if (layout[0] == 'o')
234 rv = 0x10000 + (copies<<8) + 1;
235 else
236 rv = 1 + (copies<<8);
237 return rv;
238}
239
240int parse_layout_faulty(char *layout)
241{
242 /* Parse the layout string for 'faulty' */
243 int ln = strcspn(layout, "0123456789");
244 char *m = strdup(layout);
245 int mode;
246 m[ln] = 0;
247 mode = map_name(faultylayout, m);
248 if (mode == UnSet)
249 return -1;
250
251 return mode | (atoi(layout+ln)<< ModeShift);
252}
eb3929a4 253#endif
4a06e2c2 254
0430ed48
NB
255void remove_partitions(int fd)
256{
257 /* remove partitions from this block devices.
258 * This is used for components added to an array
259 */
260#ifdef BLKPG_DEL_PARTITION
261 struct blkpg_ioctl_arg a;
262 struct blkpg_partition p;
263
264 a.op = BLKPG_DEL_PARTITION;
265 a.data = (void*)&p;
266 a.datalen = sizeof(p);
267 a.flags = 0;
268 memset(a.data, 0, a.datalen);
269 for (p.pno=0; p.pno < 16; p.pno++)
270 ioctl(fd, BLKPG, &a);
271#endif
272}
273
691c6ee1
N
274int test_partition(int fd)
275{
276 /* Check if fd is a whole-disk or a partition.
277 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
278 * will return ENXIO on an invalid partition number.
279 */
280 struct blkpg_ioctl_arg a;
281 struct blkpg_partition p;
282 a.op = BLKPG_DEL_PARTITION;
283 a.data = (void*)&p;
284 a.datalen = sizeof(p);
285 a.flags = 0;
286 memset(a.data, 0, a.datalen);
287 p.pno = 1<<30;
288 if (ioctl(fd, BLKPG, &a) == 0)
289 /* Very unlikely, but not a partition */
290 return 0;
291 if (errno == ENXIO)
292 /* not a partition */
293 return 0;
294
295 return 1;
296}
297
bfd76b93
CA
298int test_partition_from_id(dev_t id)
299{
300 char buf[20];
301 int fd, rv;
302
303 sprintf(buf, "%d:%d", major(id), minor(id));
304 fd = dev_open(buf, O_RDONLY);
305 if (fd < 0)
306 return -1;
307 rv = test_partition(fd);
308 close(fd);
309 return rv;
310}
691c6ee1 311
583315d9 312int enough(int level, int raid_disks, int layout, int clean,
265e0f17 313 char *avail, int avail_disks)
64c4757e 314{
265e0f17 315 int copies, first;
64c4757e 316 switch (level) {
265e0f17
NB
317 case 10:
318 /* This is the tricky one - we need to check
319 * which actual disks are present.
320 */
702b557b 321 copies = (layout&255)* ((layout>>8) & 255);
265e0f17
NB
322 first=0;
323 do {
324 /* there must be one of the 'copies' form 'first' */
325 int n = copies;
326 int cnt=0;
327 while (n--) {
328 if (avail[first])
329 cnt++;
330 first = (first+1) % raid_disks;
331 }
332 if (cnt == 0)
333 return 0;
334
335 } while (first != 0);
336 return 1;
e5329c37 337
df0d4ea0 338 case LEVEL_MULTIPATH:
e0d19036 339 return avail_disks>= 1;
df0d4ea0 340 case LEVEL_LINEAR:
64c4757e
NB
341 case 0:
342 return avail_disks == raid_disks;
343 case 1:
344 return avail_disks >= 1;
345 case 4:
346 case 5:
583315d9
NB
347 if (clean)
348 return avail_disks >= raid_disks-1;
349 else
350 return avail_disks >= raid_disks;
98c6faba 351 case 6:
583315d9
NB
352 if (clean)
353 return avail_disks >= raid_disks-2;
354 else
355 return avail_disks >= raid_disks;
64c4757e
NB
356 default:
357 return 0;
358 }
359}
360
8453e704
N
361int enough_fd(int fd)
362{
363 struct mdu_array_info_s array;
364 struct mdu_disk_info_s disk;
365 int avail_disks = 0;
366 int i;
367 char *avail;
368
369 if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 ||
370 array.raid_disks <= 0)
371 return 0;
372 avail = calloc(array.raid_disks, 1);
373 for (i=0; i<array.raid_disks + array.nr_disks; i++) {
374 disk.number = i;
375 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
376 continue;
377 if (! (disk.state & (1<<MD_DISK_SYNC)))
378 continue;
379 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
380 continue;
381 avail_disks++;
382 avail[disk.raid_disk] = 1;
383 }
384 /* This is used on an active array, so assume it is clean */
385 return enough(array.level, array.raid_disks, array.layout,
386 1,
387 avail, avail_disks);
388}
389
390
22e263f6
AC
391const int uuid_zero[4] = { 0, 0, 0, 0 };
392
f277ce36 393int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 394{
f277ce36
NB
395 if (swapuuid) {
396 /* parse uuids are hostendian.
397 * uuid's from some superblocks are big-ending
aba69144 398 * if there is a difference, we need to swap..
f277ce36
NB
399 */
400 unsigned char *ac = (unsigned char *)a;
401 unsigned char *bc = (unsigned char *)b;
402 int i;
403 for (i=0; i<16; i+= 4) {
404 if (ac[i+0] != bc[i+3] ||
405 ac[i+1] != bc[i+2] ||
406 ac[i+2] != bc[i+1] ||
407 ac[i+3] != bc[i+0])
408 return 0;
409 }
410 return 1;
411 } else {
412 if (a[0]==b[0] &&
413 a[1]==b[1] &&
414 a[2]==b[2] &&
415 a[3]==b[3])
416 return 1;
417 return 0;
418 }
64c4757e 419}
350f29f9
NB
420void copy_uuid(void *a, int b[4], int swapuuid)
421{
422 if (swapuuid) {
423 /* parse uuids are hostendian.
424 * uuid's from some superblocks are big-ending
425 * if there is a difference, we need to swap..
426 */
427 unsigned char *ac = (unsigned char *)a;
428 unsigned char *bc = (unsigned char *)b;
429 int i;
430 for (i=0; i<16; i+= 4) {
431 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 }
436 } else
437 memcpy(a, b, 16);
438}
64c4757e 439
aae5a112 440char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 441{
9968e376 442 int i, j;
d7288ddc
N
443 char uuid[16];
444 char *c = buf;
445 strcpy(c, "UUID-");
446 c += strlen(c);
aae5a112 447 copy_uuid(uuid, id, swap);
9968e376 448 for (i = 0; i < 4; i++) {
9968e376 449 if (i)
ff54de6e 450 *c++ = sep;
9968e376
DW
451 for (j = 3; j >= 0; j--) {
452 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
453 c+= 2;
454 }
d7288ddc
N
455 }
456 return buf;
aae5a112
DW
457
458}
459
460char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
461{
86983cce
N
462 // dirty hack to work around an issue with super1 superblocks...
463 // super1 superblocks need swapuuid set in order for assembly to
464 // work, but can't have it set if we want this printout to match
465 // all the other uuid printouts in super1.c, so we force swapuuid
466 // to 1 to make our printout match the rest of super1
467 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
468}
469
435d4ebb 470#ifndef MDASSEMBLE
682c7051
NB
471int check_ext2(int fd, char *name)
472{
473 /*
474 * Check for an ext2fs file system.
475 * Superblock is always 1K at 1K offset
476 *
477 * s_magic is le16 at 56 == 0xEF53
478 * report mtime - le32 at 44
479 * blocks - le32 at 4
480 * logblksize - le32 at 24
481 */
482 unsigned char sb[1024];
483 time_t mtime;
484 int size, bsize;
485 if (lseek(fd, 1024,0)!= 1024)
486 return 0;
487 if (read(fd, sb, 1024)!= 1024)
488 return 0;
489 if (sb[56] != 0x53 || sb[57] != 0xef)
490 return 0;
491
492 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
493 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
494 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
495 fprintf(stderr, Name ": %s appears to contain an ext2fs file system\n",
496 name);
497 fprintf(stderr," size=%dK mtime=%s",
498 size*(1<<bsize), ctime(&mtime));
499 return 1;
500}
501
502int check_reiser(int fd, char *name)
503{
504 /*
505 * superblock is at 64K
506 * size is 1024;
507 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
508 *
509 */
510 unsigned char sb[1024];
881990a2 511 unsigned long size;
682c7051
NB
512 if (lseek(fd, 64*1024, 0) != 64*1024)
513 return 0;
514 if (read(fd, sb, 1024) != 1024)
515 return 0;
a46f4061
NB
516 if (strncmp((char*)sb+52, "ReIsErFs",8)!=0 &&
517 strncmp((char*)sb+52, "ReIsEr2Fs",9)!=0)
682c7051
NB
518 return 0;
519 fprintf(stderr, Name ": %s appears to contain a reiserfs file system\n",name);
520 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
881990a2 521 fprintf(stderr, " size = %luK\n", size*4);
aba69144 522
682c7051
NB
523 return 1;
524}
525
526int check_raid(int fd, char *name)
527{
4b1ac34b 528 struct mdinfo info;
682c7051 529 time_t crtime;
d078d77c 530 char *level;
82d9eba6 531 struct supertype *st = guess_super(fd);
f9ce90ba 532
82d9eba6 533 if (!st) return 0;
3da92f27 534 st->ss->load_super(st, fd, name);
82d9eba6
NB
535 /* Looks like a raid array .. */
536 fprintf(stderr, Name ": %s appears to be part of a raid array:\n",
537 name);
a5d85af7 538 st->ss->getinfo_super(st, &info, NULL);
3da92f27 539 st->ss->free_super(st);
82d9eba6 540 crtime = info.array.ctime;
d078d77c
NB
541 level = map_num(pers, info.array.level);
542 if (!level) level = "-unknown-";
543 fprintf(stderr, " level=%s devices=%d ctime=%s",
544 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 545 return 1;
682c7051
NB
546}
547
682c7051
NB
548int ask(char *mesg)
549{
550 char *add = "";
551 int i;
552 for (i=0; i<5; i++) {
553 char buf[100];
554 fprintf(stderr, "%s%s", mesg, add);
555 fflush(stderr);
556 if (fgets(buf, 100, stdin)==NULL)
557 return 0;
558 if (buf[0]=='y' || buf[0]=='Y')
559 return 1;
560 if (buf[0]=='n' || buf[0]=='N')
561 return 0;
562 add = "(y/n) ";
563 }
564 fprintf(stderr, Name ": assuming 'no'\n");
565 return 0;
566}
435d4ebb 567#endif /* MDASSEMBLE */
682c7051
NB
568
569char *map_num(mapping_t *map, int num)
570{
571 while (map->name) {
572 if (map->num == num)
573 return map->name;
574 map++;
575 }
576 return NULL;
577}
578
579int map_name(mapping_t *map, char *name)
580{
581 while (map->name) {
582 if (strcmp(map->name, name)==0)
583 return map->num;
584 map++;
585 }
98c6faba 586 return UnSet;
682c7051 587}
82b27616 588
e5329c37 589
8d80900b 590int is_standard(char *dev, int *nump)
e5329c37
NB
591{
592 /* tests if dev is a "standard" md dev name.
593 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 594 * where NN is a string of digits
598f0d58
NB
595 * Returns 1 if a partitionable standard,
596 * -1 if non-partitonable,
597 * 0 if not a standard name.
e5329c37 598 */
8d80900b
NB
599 char *d = strrchr(dev, '/');
600 int type=0;
601 int num;
602 if (!d)
e5329c37 603 return 0;
8d80900b
NB
604 if (strncmp(d, "/d",2)==0)
605 d += 2, type=1; /* /dev/md/dN{pM} */
606 else if (strncmp(d, "/md_d", 5)==0)
2b4ca8f0 607 d += 5, type=1; /* /dev/md_dN{pM} */
8d80900b
NB
608 else if (strncmp(d, "/md", 3)==0)
609 d += 3, type=-1; /* /dev/mdN */
610 else if (d-dev > 3 && strncmp(d-2, "md/", 3)==0)
5a6d1148 611 d += 1, type=-1; /* /dev/md/N */
e5329c37
NB
612 else
613 return 0;
8d80900b 614 if (!*d)
e5329c37 615 return 0;
8d80900b
NB
616 num = atoi(d);
617 while (isdigit(*d))
618 d++;
619 if (*d)
e5329c37 620 return 0;
8d80900b
NB
621 if (nump) *nump = num;
622
623 return type;
e5329c37
NB
624}
625
626
82b27616
NB
627/*
628 * convert a major/minor pair for a block device into a name in /dev, if possible.
629 * On the first call, walk /dev collecting name.
630 * Put them in a simple linked listfor now.
631 */
632struct devmap {
633 int major, minor;
634 char *name;
635 struct devmap *next;
636} *devlist = NULL;
637int devlist_ready = 0;
638
82b27616
NB
639int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s)
640{
bed256c2 641 struct stat st;
bf68e9d9 642
bed256c2 643 if (S_ISLNK(stb->st_mode)) {
bf68e9d9
DW
644 if (stat(name, &st) != 0)
645 return 0;
bed256c2 646 stb = &st;
82b27616 647 }
bed256c2
NB
648
649 if ((stb->st_mode&S_IFMT)== S_IFBLK) {
650 char *n = strdup(name);
651 struct devmap *dm = malloc(sizeof(*dm));
652 if (strncmp(n, "/dev/./", 7)==0)
653 strcpy(n+4, name+6);
654 if (dm) {
655 dm->major = major(stb->st_rdev);
656 dm->minor = minor(stb->st_rdev);
657 dm->name = n;
658 dm->next = devlist;
659 devlist = dm;
660 }
661 }
662 return 0;
82b27616
NB
663}
664
45e878bb
NB
665#ifndef HAVE_NFTW
666#ifdef HAVE_FTW
667int add_dev_1(const char *name, const struct stat *stb, int flag)
668{
669 return add_dev(name, stb, flag, NULL);
670}
671int nftw(const char *path, int (*han)(const char *name, const struct stat *stb, int flag, struct FTW *s), int nopenfd, int flags)
672{
673 return ftw(path, add_dev_1, nopenfd);
674}
675#else
45e878bb
NB
676int nftw(const char *path, int (*han)(const char *name, const struct stat *stb, int flag, struct FTW *s), int nopenfd, int flags)
677{
678 return 0;
679}
680#endif /* HAVE_FTW */
681#endif /* HAVE_NFTW */
682
dd0781e5
NB
683/*
684 * Find a block device with the right major/minor number.
b79713f8 685 * If we find multiple names, choose the shortest.
70ef16db 686 * If we find a name in /dev/md/, we prefer that.
b79713f8 687 * This applies only to names for MD devices.
dd0781e5 688 */
16c6fa80 689char *map_dev(int major, int minor, int create)
82b27616 690{
dd0781e5 691 struct devmap *p;
70ef16db 692 char *regular = NULL, *preferred=NULL;
e7bb5d23 693 int did_check = 0;
eed35d66 694
e81cdd9f 695 if (major == 0 && minor == 0)
eed35d66 696 return NULL;
e81cdd9f 697
e7bb5d23 698 retry:
dd0781e5 699 if (!devlist_ready) {
0a416ec3
NB
700 char *dev = "/dev";
701 struct stat stb;
eed35d66
NB
702 while(devlist) {
703 struct devmap *d = devlist;
704 devlist = d->next;
705 free(d->name);
706 free(d);
707 }
0a416ec3
NB
708 if (lstat(dev, &stb)==0 &&
709 S_ISLNK(stb.st_mode))
710 dev = "/dev/.";
711 nftw(dev, add_dev, 10, FTW_PHYS);
dd0781e5 712 devlist_ready=1;
e7bb5d23 713 did_check = 1;
dd0781e5 714 }
82b27616 715
dd0781e5
NB
716 for (p=devlist; p; p=p->next)
717 if (p->major == major &&
718 p->minor == minor) {
70ef16db
N
719 if (strncmp(p->name, "/dev/md/",8) == 0) {
720 if (preferred == NULL ||
721 strlen(p->name) < strlen(preferred))
722 preferred = p->name;
b79713f8 723 } else {
70ef16db
N
724 if (regular == NULL ||
725 strlen(p->name) < strlen(regular))
726 regular = p->name;
b79713f8 727 }
dd0781e5 728 }
70ef16db 729 if (!regular && !preferred && !did_check) {
e7bb5d23
NB
730 devlist_ready = 0;
731 goto retry;
732 }
70ef16db 733 if (create && !regular && !preferred) {
16c6fa80 734 static char buf[30];
382245c3 735 snprintf(buf, sizeof(buf), "%d:%d", major, minor);
70ef16db 736 regular = buf;
16c6fa80
NB
737 }
738
70ef16db 739 return preferred ? preferred : regular;
82b27616
NB
740}
741
4b1ac34b 742unsigned long calc_csum(void *super, int bytes)
82b27616 743{
56eb10c0 744 unsigned long long newcsum = 0;
82b27616 745 int i;
4b1ac34b
NB
746 unsigned int csum;
747 unsigned int *superc = (unsigned int*) super;
82b27616 748
4b1ac34b 749 for(i=0; i<bytes/4; i++)
82b27616
NB
750 newcsum+= superc[i];
751 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 752#ifdef __alpha__
aba69144 753/* The in-kernel checksum calculation is always 16bit on
570c0542
NB
754 * the alpha, though it is 32 bit on i386...
755 * I wonder what it is elsewhere... (it uses and API in
756 * a way that it shouldn't).
757 */
758 csum = (csum & 0xffff) + (csum >> 16);
759 csum = (csum & 0xffff) + (csum >> 16);
760#endif
82b27616
NB
761 return csum;
762}
cd29a5c8 763
435d4ebb 764#ifndef MDASSEMBLE
56eb10c0 765char *human_size(long long bytes)
cd29a5c8
NB
766{
767 static char buf[30];
d5d3721e
NB
768
769 /* We convert bytes to either centi-M{ega,ibi}bytes or
770 * centi-G{igi,ibi}bytes, with appropriate rounding,
771 * and then print 1/100th of those as a decimal.
772 * We allow upto 2048Megabytes before converting to
773 * gigabytes, as that shows more precision and isn't
774 * too large a number.
775 * Terrabytes are not yet handled.
776 */
cd29a5c8 777
56eb10c0 778 if (bytes < 5000*1024)
cd29a5c8 779 buf[0]=0;
d5d3721e
NB
780 else if (bytes < 2*1024LL*1024LL*1024LL) {
781 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
782 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 783 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
784 cMiB/100 , cMiB % 100,
785 cMB/100, cMB % 100);
786 } else {
787 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
788 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 789 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
790 cGiB/100 , cGiB % 100,
791 cGB/100, cGB % 100);
792 }
cd29a5c8
NB
793 return buf;
794}
e0d19036
NB
795
796char *human_size_brief(long long bytes)
797{
798 static char buf[30];
e0d19036
NB
799
800 if (bytes < 5000*1024)
8f23b0b3 801 snprintf(buf, sizeof(buf), "%ld.%02ldKiB",
bd526cee 802 (long)(bytes>>10), (long)(((bytes&1023)*100+512)/1024)
e0d19036
NB
803 );
804 else if (bytes < 2*1024LL*1024LL*1024LL)
8f23b0b3 805 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
e0d19036 806 (long)(bytes>>20),
bd526cee 807 (long)((bytes&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
808 );
809 else
8f23b0b3 810 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
e0d19036 811 (long)(bytes>>30),
bd526cee 812 (long)(((bytes>>10)&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
813 );
814 return buf;
815}
e4965ef8
N
816
817void print_r10_layout(int layout)
818{
819 int near = layout & 255;
820 int far = (layout >> 8) & 255;
821 int offset = (layout&0x10000);
822 char *sep = "";
823
824 if (near != 1) {
825 printf("%s near=%d", sep, near);
826 sep = ",";
827 }
828 if (far != 1)
829 printf("%s %s=%d", sep, offset?"offset":"far", far);
830 if (near*far == 1)
831 printf("NO REDUNDANCY");
832}
435d4ebb 833#endif
e0d19036 834
5f8097be
NB
835unsigned long long calc_array_size(int level, int raid_disks, int layout,
836 int chunksize, unsigned long long devsize)
837{
838 int data_disks = 0;
839 switch (level) {
840 case 0: data_disks = raid_disks; break;
841 case 1: data_disks = 1; break;
842 case 4:
843 case 5: data_disks = raid_disks - 1; break;
844 case 6: data_disks = raid_disks - 2; break;
845 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
846 break;
847 }
848 devsize &= ~(unsigned long long)((chunksize>>9)-1);
849 return data_disks * devsize;
850}
851
dd0781e5 852int get_mdp_major(void)
98c6faba 853{
dd0781e5
NB
854static int mdp_major = -1;
855 FILE *fl;
98c6faba
NB
856 char *w;
857 int have_block = 0;
858 int have_devices = 0;
859 int last_num = -1;
dd0781e5
NB
860
861 if (mdp_major != -1)
862 return mdp_major;
863 fl = fopen("/proc/devices", "r");
98c6faba 864 if (!fl)
dd0781e5 865 return -1;
98c6faba
NB
866 while ((w = conf_word(fl, 1))) {
867 if (have_block && strcmp(w, "devices:")==0)
868 have_devices = 1;
869 have_block = (strcmp(w, "Block")==0);
870 if (isdigit(w[0]))
871 last_num = atoi(w);
872 if (have_devices && strcmp(w, "mdp")==0)
873 mdp_major = last_num;
874 free(w);
875 }
876 fclose(fl);
dd0781e5 877 return mdp_major;
98c6faba
NB
878}
879
0e600426 880#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
e0d19036
NB
881char *get_md_name(int dev)
882{
883 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 884 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
e0d19036
NB
885 static char devname[50];
886 struct stat stb;
98c6faba 887 dev_t rdev;
dd0781e5 888 char *dn;
98c6faba
NB
889
890 if (dev < 0) {
dd0781e5
NB
891 int mdp = get_mdp_major();
892 if (mdp < 0) return NULL;
0df46c2a 893 rdev = makedev(mdp, (-1-dev)<<6);
8f23b0b3 894 snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev);
98c6faba
NB
895 if (stat(devname, &stb) == 0
896 && (S_IFMT&stb.st_mode) == S_IFBLK
897 && (stb.st_rdev == rdev))
898 return devname;
899 } else {
0df46c2a 900 rdev = makedev(MD_MAJOR, dev);
8f23b0b3 901 snprintf(devname, sizeof(devname), "/dev/md%d", dev);
98c6faba
NB
902 if (stat(devname, &stb) == 0
903 && (S_IFMT&stb.st_mode) == S_IFBLK
904 && (stb.st_rdev == rdev))
905 return devname;
906
8f23b0b3 907 snprintf(devname, sizeof(devname), "/dev/md/%d", dev);
98c6faba
NB
908 if (stat(devname, &stb) == 0
909 && (S_IFMT&stb.st_mode) == S_IFBLK
910 && (stb.st_rdev == rdev))
911 return devname;
912 }
16c6fa80 913 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
914 if (dn)
915 return dn;
8f23b0b3 916 snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev);
e0d19036 917 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
918 if (errno != EEXIST)
919 return NULL;
e0d19036
NB
920
921 if (stat(devname, &stb) == 0
922 && (S_IFMT&stb.st_mode) == S_IFBLK
923 && (stb.st_rdev == rdev))
924 return devname;
925 unlink(devname);
926 return NULL;
927}
928
929void put_md_name(char *name)
930{
931 if (strncmp(name, "/dev/.tmp.md", 12)==0)
932 unlink(name);
933}
ea24acd0 934
ea24acd0
NB
935int find_free_devnum(int use_partitions)
936{
937 int devnum;
938 for (devnum = 127; devnum != 128;
a56fb7ec 939 devnum = devnum ? devnum-1 : (1<<20)-1) {
ea24acd0
NB
940 char *dn;
941 int _devnum;
942
943 _devnum = use_partitions ? (-1-devnum) : devnum;
944 if (mddev_busy(_devnum))
945 continue;
946 /* make sure it is new to /dev too, at least as a
947 * non-standard */
948 dn = map_dev(dev2major(_devnum), dev2minor(_devnum), 0);
949 if (dn && ! is_standard(dn, NULL))
950 continue;
951 break;
952 }
953 if (devnum == 128)
954 return NoMdDev;
955 return use_partitions ? (-1-devnum) : devnum;
956}
435d4ebb 957#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 958
8b0dabea
NB
959int dev_open(char *dev, int flags)
960{
961 /* like 'open', but if 'dev' matches %d:%d, create a temp
962 * block device and open that
963 */
964 char *e;
965 int fd = -1;
966 char devname[32];
e81cdd9f 967 int major;
8b0dabea 968 int minor;
e81cdd9f
NB
969
970 if (!dev) return -1;
6df6a774 971 flags |= O_DIRECT;
e81cdd9f
NB
972
973 major = strtoul(dev, &e, 0);
8b0dabea
NB
974 if (e > dev && *e == ':' && e[1] &&
975 (minor = strtoul(e+1, &e, 0)) >= 0 &&
976 *e == 0) {
6df6a774
N
977 char *path = map_dev(major, minor, 0);
978 if (path)
979 fd = open(path, flags);
980 if (fd < 0) {
981 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
982 (int)getpid(), major, minor);
983 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
984 fd = open(devname, flags);
985 unlink(devname);
986 }
987 }
988 if (fd < 0) {
989 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
990 (int)getpid(), major, minor);
991 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
992 fd = open(devname, flags);
993 unlink(devname);
994 }
8b0dabea
NB
995 }
996 } else
6df6a774 997 fd = open(dev, flags);
8b0dabea
NB
998 return fd;
999}
f9ce90ba 1000
d998b738 1001int open_dev_flags(int devnum, int flags)
e8a70c89
N
1002{
1003 char buf[20];
1004
1005 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
d998b738
N
1006 return dev_open(buf, flags);
1007}
1008
1009int open_dev(int devnum)
1010{
1011 return open_dev_flags(devnum, O_RDONLY);
e8a70c89
N
1012}
1013
a931db9e
NB
1014int open_dev_excl(int devnum)
1015{
1016 char buf[20];
1017 int i;
1018
1019 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
1020 for (i=0 ; i<25 ; i++) {
1021 int fd = dev_open(buf, O_RDWR|O_EXCL);
1022 if (fd >= 0)
1023 return fd;
1024 if (errno != EBUSY)
1025 return fd;
1026 usleep(200000);
1027 }
1028 return -1;
1029}
1030
9008ed1c
N
1031int same_dev(char *one, char *two)
1032{
1033 struct stat st1, st2;
1034 if (stat(one, &st1) != 0)
1035 return 0;
1036 if (stat(two, &st2) != 0)
1037 return 0;
1038 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1039 return 0;
1040 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1041 return 0;
1042 return st1.st_rdev == st2.st_rdev;
1043}
1044
a7c6e3fb 1045void wait_for(char *dev, int fd)
a714580e
N
1046{
1047 int i;
a7c6e3fb
N
1048 struct stat stb_want;
1049
1050 if (fstat(fd, &stb_want) != 0 ||
1051 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1052 return;
a714580e
N
1053
1054 for (i=0 ; i<25 ; i++) {
1055 struct stat stb;
a7c6e3fb
N
1056 if (stat(dev, &stb) == 0 &&
1057 (stb.st_mode & S_IFMT) == S_IFBLK &&
1058 (stb.st_rdev == stb_want.st_rdev))
a714580e
N
1059 return;
1060 usleep(200000);
1061 }
436305c6
DW
1062 if (i == 25)
1063 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
1064}
1065
0f22b998
N
1066struct superswitch *superlist[] =
1067{
1068 &super0, &super1,
1069 &super_ddf, &super_imsm,
0592faeb 1070 &mbr, &gpt,
0f22b998 1071 NULL };
f9ce90ba 1072
ea24acd0 1073#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1074
4725bc31 1075struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1076{
1686dc25
NB
1077 mdu_array_info_t array;
1078 int vers;
1079 int minor;
1080 struct supertype *st = NULL;
7e0f6979 1081 struct mdinfo *sra;
142cb9e1 1082 char *verstr;
1686dc25
NB
1083 char version[20];
1084 int i;
f7e7067b 1085 char *subarray = NULL;
5f7e44b2 1086 int container = NoMdDev;
1686dc25
NB
1087
1088 sra = sysfs_read(fd, 0, GET_VERSION);
1089
1090 if (sra) {
7e0f6979
NB
1091 vers = sra->array.major_version;
1092 minor = sra->array.minor_version;
142cb9e1 1093 verstr = sra->text_version;
1686dc25
NB
1094 } else {
1095 if (ioctl(fd, GET_ARRAY_INFO, &array))
1096 array.major_version = array.minor_version = 0;
1097 vers = array.major_version;
1098 minor = array.minor_version;
142cb9e1 1099 verstr = "";
6fbba4c9 1100 }
82d9eba6 1101
1686dc25
NB
1102 if (vers != -1) {
1103 sprintf(version, "%d.%d", vers, minor);
1104 verstr = version;
6fbba4c9 1105 }
3c558363 1106 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1107 char *dev = verstr+1;
5f7e44b2 1108
f7e7067b 1109 subarray = strchr(dev, '/');
f7e7067b
NB
1110 if (subarray)
1111 *subarray++ = '\0';
f7e7067b 1112 subarray = strdup(subarray);
5f7e44b2 1113 container = devname2devnum(dev);
f7e7067b
NB
1114 if (sra)
1115 sysfs_free(sra);
5f7e44b2 1116 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1117 if (sra && sra->text_version[0])
1118 verstr = sra->text_version;
1119 else
1120 verstr = "-no-metadata-";
f7e7067b
NB
1121 }
1122
1123 for (i = 0; st == NULL && superlist[i] ; i++)
1124 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
1125
1126 if (sra)
1127 sysfs_free(sra);
f7e7067b 1128 if (st) {
3b0896f8 1129 st->sb = NULL;
1f49fb3a
N
1130 if (subarrayp)
1131 *subarrayp = subarray;
5f7e44b2
DW
1132 st->container_dev = container;
1133 st->devnum = fd2devnum(fd);
4725bc31
N
1134 } else
1135 free(subarray);
5f7e44b2 1136
82d9eba6 1137 return st;
f9ce90ba 1138}
ea24acd0
NB
1139#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1140
0f0749ad 1141int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1142{
1143 char buf[20];
1144 int fd;
1145
1146 sprintf(buf, "%d:%d", major(id), minor(id));
1147 fd = dev_open(buf, O_RDONLY);
1148 if (fd < 0)
1149 return 0;
1150 if (get_dev_size(fd, NULL, size)) {
1151 close(fd);
1152 return 1;
1153 }
1154 close(fd);
1155 return 0;
1156}
f9ce90ba 1157
159c3a1a 1158struct supertype *dup_super(struct supertype *orig)
3da92f27 1159{
159c3a1a 1160 struct supertype *st;
1686dc25 1161
d2ca6449
NB
1162 if (!orig)
1163 return orig;
159c3a1a 1164 st = malloc(sizeof(*st));
3da92f27
NB
1165 if (!st)
1166 return st;
ef609477 1167 memset(st, 0, sizeof(*st));
159c3a1a
NB
1168 st->ss = orig->ss;
1169 st->max_devs = orig->max_devs;
1170 st->minor_version = orig->minor_version;
1171 st->sb = NULL;
1172 st->info = NULL;
1173 return st;
3da92f27
NB
1174}
1175
54887ad8 1176struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1177{
1178 /* try each load_super to find the best match,
1179 * and return the best superswitch
1180 */
82d9eba6
NB
1181 struct superswitch *ss;
1182 struct supertype *st;
f21e18ca 1183 time_t besttime = 0;
570c0542 1184 int bestsuper = -1;
f9ce90ba
NB
1185 int i;
1186
82d9eba6 1187 st = malloc(sizeof(*st));
d1d599ea
N
1188 memset(st, 0, sizeof(*st));
1189 st->container_dev = NoMdDev;
1190
f9ce90ba
NB
1191 for (i=0 ; superlist[i]; i++) {
1192 int rv;
1193 ss = superlist[i];
54887ad8
N
1194 if (guess_type == guess_array && ss->add_to_super == NULL)
1195 continue;
1196 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1197 continue;
ef609477 1198 memset(st, 0, sizeof(*st));
3da92f27 1199 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1200 if (rv == 0) {
1201 struct mdinfo info;
a5d85af7 1202 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1203 if (bestsuper == -1 ||
1204 besttime < info.array.ctime) {
1205 bestsuper = i;
1206 besttime = info.array.ctime;
570c0542 1207 }
3da92f27 1208 ss->free_super(st);
570c0542
NB
1209 }
1210 }
1211 if (bestsuper != -1) {
1212 int rv;
ef609477 1213 memset(st, 0, sizeof(*st));
3da92f27 1214 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1215 if (rv == 0) {
5e747af2 1216 superlist[bestsuper]->free_super(st);
82d9eba6 1217 return st;
f9ce90ba
NB
1218 }
1219 }
570c0542 1220 free(st);
f9ce90ba
NB
1221 return NULL;
1222}
fe6729fa 1223
beae1dfe
NB
1224/* Return size of device in bytes */
1225int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1226{
1227 unsigned long long ldsize;
c2c9bb6f
NB
1228 struct stat st;
1229
1230 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1231 ldsize = (unsigned long long)st.st_size;
1232 else
beae1dfe
NB
1233#ifdef BLKGETSIZE64
1234 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1235#endif
1236 {
1237 unsigned long dsize;
1238 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1239 ldsize = dsize;
1240 ldsize <<= 9;
1241 } else {
1242 if (dname)
1243 fprintf(stderr, Name ": Cannot get size of %s: %s\b",
1244 dname, strerror(errno));
1245 return 0;
1246 }
1247 }
1248 *sizep = ldsize;
1249 return 1;
1250}
8fac0577 1251
3a371610
N
1252/* Return true if this can only be a container, not a member device.
1253 * i.e. is and md device and size is zero
1254 */
1255int must_be_container(int fd)
1256{
1257 unsigned long long size;
1258 if (md_get_version(fd) < 0)
1259 return 0;
1260 if (get_dev_size(fd, NULL, &size) == 0)
1261 return 1;
1262 if (size == 0)
1263 return 1;
1264 return 0;
1265}
034b203a
TM
1266
1267/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1268 * Returns: 1 if successful
1269 * -1 for unknown partition type
1270 * 0 for other errors
1271 */
1272static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1273{
056b331e 1274 struct GPT gpt;
034b203a
TM
1275 unsigned char buf[512];
1276 unsigned char empty_gpt_entry[16]= {0};
1277 struct GPT_part_entry *part;
1278 unsigned long long curr_part_end;
1279 unsigned all_partitions, entry_size;
f21e18ca 1280 unsigned part_nr;
034b203a
TM
1281
1282 *endofpart = 0;
1283
056b331e 1284 BUILD_BUG_ON(sizeof(gpt) != 512);
034b203a
TM
1285 /* read GPT header */
1286 lseek(fd, 512, SEEK_SET);
056b331e 1287 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1288 return 0;
1289
1290 /* get the number of partition entries and the entry size */
056b331e
N
1291 all_partitions = __le32_to_cpu(gpt.part_cnt);
1292 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1293
1294 /* Check GPT signature*/
056b331e 1295 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1296 return -1;
1297
1298 /* sanity checks */
1299 if (all_partitions > 1024 ||
1300 entry_size > 512)
1301 return -1;
1302
1303 /* read first GPT partition entries */
1304 if (read(fd, buf, 512) != 512)
1305 return 0;
1306
1307 part = (struct GPT_part_entry*)buf;
1308
1309 for (part_nr=0; part_nr < all_partitions; part_nr++) {
1310 /* is this valid partition? */
1311 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
1312 /* check the last lba for the current partition */
056b331e 1313 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1314 if (curr_part_end > *endofpart)
1315 *endofpart = curr_part_end;
1316 }
1317
1318 part = (struct GPT_part_entry*)((unsigned char*)part + entry_size);
1319
1320 if ((unsigned char *)part >= buf + 512) {
1321 if (read(fd, buf, 512) != 512)
1322 return 0;
1323 part = (struct GPT_part_entry*)buf;
1324 }
1325 }
1326 return 1;
1327}
1328
1329/* Sets endofpart parameter to the last block used by the last partition on the device.
1330 * Returns: 1 if successful
1331 * -1 for unknown partition type
1332 * 0 for other errors
1333 */
1334static int get_last_partition_end(int fd, unsigned long long *endofpart)
1335{
056b331e 1336 struct MBR boot_sect;
034b203a
TM
1337 struct MBR_part_record *part;
1338 unsigned long long curr_part_end;
f21e18ca 1339 unsigned part_nr;
034b203a
TM
1340 int retval = 0;
1341
1342 *endofpart = 0;
1343
056b331e 1344 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1345 /* read MBR */
1346 lseek(fd, 0, 0);
056b331e 1347 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1348 goto abort;
1349
1350 /* check MBP signature */
056b331e 1351 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1352 retval = 1;
1353 /* found the correct signature */
056b331e 1354 part = boot_sect.parts;
034b203a
TM
1355
1356 for (part_nr=0; part_nr < MBR_PARTITIONS; part_nr++) {
1357 /* check for GPT type */
1358 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1359 retval = get_gpt_last_partition_end(fd, endofpart);
1360 break;
1361 }
1362 /* check the last used lba for the current partition */
1363 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1364 __le32_to_cpu(part->blocks_num);
1365 if (curr_part_end > *endofpart)
1366 *endofpart = curr_part_end;
1367
1368 part++;
1369 }
1370 } else {
1371 /* Unknown partition table */
1372 retval = -1;
1373 }
1374 abort:
1375 return retval;
1376}
1377
53ed6ac3
KW
1378int check_partitions(int fd, char *dname, unsigned long long freesize,
1379 unsigned long long size)
034b203a
TM
1380{
1381 /*
1382 * Check where the last partition ends
1383 */
1384 unsigned long long endofpart;
1385 int ret;
1386
1387 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1388 /* There appears to be a partition table here */
1389 if (freesize == 0) {
1390 /* partitions will not be visible in new device */
1391 fprintf(stderr,
1392 Name ": partition table exists on %s but will be lost or\n"
1393 " meaningless after creating array\n",
1394 dname);
1395 return 1;
1396 } else if (endofpart > freesize) {
1397 /* last partition overlaps metadata */
1398 fprintf(stderr,
1399 Name ": metadata will over-write last partition on %s.\n",
1400 dname);
1401 return 1;
53ed6ac3
KW
1402 } else if (size && endofpart > size) {
1403 /* partitions will be truncated in new device */
1404 fprintf(stderr,
1405 Name ": array size is too small to cover all partitions on %s.\n",
1406 dname);
1407 return 1;
034b203a
TM
1408 }
1409 }
1410 return 0;
1411}
1412
8382f19b
NB
1413void get_one_disk(int mdfd, mdu_array_info_t *ainf, mdu_disk_info_t *disk)
1414{
1415 int d;
1416 ioctl(mdfd, GET_ARRAY_INFO, ainf);
1417 for (d = 0 ; d < ainf->raid_disks + ainf->nr_disks ; d++)
1418 if (ioctl(mdfd, GET_DISK_INFO, disk) == 0)
1419 return;
1420}
63152c1b 1421
a322f70c
DW
1422int open_container(int fd)
1423{
1424 /* 'fd' is a block device. Find out if it is in use
1425 * by a container, and return an open fd on that container.
1426 */
1427 char path[256];
1428 char *e;
1429 DIR *dir;
1430 struct dirent *de;
1431 int dfd, n;
1432 char buf[200];
1433 int major, minor;
1434 struct stat st;
1435
1436 if (fstat(fd, &st) != 0)
1437 return -1;
1438 sprintf(path, "/sys/dev/block/%d:%d/holders",
1439 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1440 e = path + strlen(path);
1441
1442 dir = opendir(path);
1443 if (!dir)
1444 return -1;
1445 while ((de = readdir(dir))) {
1446 if (de->d_ino == 0)
1447 continue;
1448 if (de->d_name[0] == '.')
1449 continue;
1450 sprintf(e, "/%s/dev", de->d_name);
1451 dfd = open(path, O_RDONLY);
1452 if (dfd < 0)
1453 continue;
1454 n = read(dfd, buf, sizeof(buf));
1455 close(dfd);
f21e18ca 1456 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1457 continue;
1458 buf[n] = 0;
1459 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1460 continue;
1461 sprintf(buf, "%d:%d", major, minor);
1462 dfd = dev_open(buf, O_RDONLY);
1463 if (dfd >= 0) {
1464 closedir(dir);
1465 return dfd;
1466 }
1467 }
355726fa 1468 closedir(dir);
a322f70c
DW
1469 return -1;
1470}
1471
33414a01
DW
1472struct superswitch *version_to_superswitch(char *vers)
1473{
1474 int i;
1475
1476 for (i = 0; superlist[i]; i++) {
1477 struct superswitch *ss = superlist[i];
1478
1479 if (strcmp(vers, ss->name) == 0)
1480 return ss;
1481 }
1482
1483 return NULL;
1484}
1485
1486int is_container_member(struct mdstat_ent *mdstat, char *container)
1487{
1488 if (mdstat->metadata_version == NULL ||
1489 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
1490 !is_subarray(mdstat->metadata_version+9) ||
1491 strncmp(mdstat->metadata_version+10, container, strlen(container)) != 0 ||
1492 mdstat->metadata_version[10+strlen(container)] != '/')
1493 return 0;
1494
1495 return 1;
1496}
1497
1498int is_subarray_active(char *subarray, char *container)
1499{
1500 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1501 struct mdstat_ent *ent;
1502
e5408a32
DW
1503 for (ent = mdstat; ent; ent = ent->next)
1504 if (is_container_member(ent, container))
e5e5d7ce 1505 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1506 break;
33414a01
DW
1507
1508 free_mdstat(mdstat);
1509
1510 return ent != NULL;
1511}
1512
1513/* open_subarray - opens a subarray in a container
1514 * @dev: container device name
feab51f8 1515 * @st: empty supertype
33414a01
DW
1516 * @quiet: block reporting errors flag
1517 *
1518 * On success returns an fd to a container and fills in *st
1519 */
feab51f8 1520int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1521{
1522 struct mdinfo *mdi;
a951a4f7 1523 struct mdinfo *info;
33414a01
DW
1524 int fd, err = 1;
1525
1526 fd = open(dev, O_RDWR|O_EXCL);
1527 if (fd < 0) {
1528 if (!quiet)
1529 fprintf(stderr, Name ": Couldn't open %s, aborting\n",
1530 dev);
1531 return 2;
1532 }
1533
1534 st->devnum = fd2devnum(fd);
1535 if (st->devnum == NoMdDev) {
1536 if (!quiet)
1537 fprintf(stderr,
1538 Name ": Failed to determine device number for %s\n",
1539 dev);
1540 goto close_fd;
1541 }
1542
1543 mdi = sysfs_read(fd, st->devnum, GET_VERSION|GET_LEVEL);
1544 if (!mdi) {
1545 if (!quiet)
1546 fprintf(stderr, Name ": Failed to read sysfs for %s\n",
1547 dev);
1548 goto close_fd;
1549 }
1550
1551 if (mdi->array.level != UnSet) {
1552 if (!quiet)
1553 fprintf(stderr, Name ": %s is not a container\n", dev);
1554 goto free_sysfs;
1555 }
1556
1557 st->ss = version_to_superswitch(mdi->text_version);
1558 if (!st->ss) {
1559 if (!quiet)
1560 fprintf(stderr,
1561 Name ": Operation not supported for %s metadata\n",
1562 mdi->text_version);
1563 goto free_sysfs;
1564 }
1565
1566 st->devname = devnum2devname(st->devnum);
1567 if (!st->devname) {
1568 if (!quiet)
1569 fprintf(stderr, Name ": Failed to allocate device name\n");
1570 goto free_sysfs;
1571 }
1572
db20d413 1573 if (!st->ss->load_container) {
33414a01 1574 if (!quiet)
db20d413 1575 fprintf(stderr, Name ": %s is not a container\n", dev);
33414a01
DW
1576 goto free_name;
1577 }
1578
db20d413 1579 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1580 if (!quiet)
db20d413
N
1581 fprintf(stderr, Name ": Failed to load metadata for %s\n",
1582 dev);
1583 goto free_name;
33414a01
DW
1584 }
1585
a951a4f7
N
1586 info = st->ss->container_content(st, subarray);
1587 if (!info) {
1588 if (!quiet)
1589 fprintf(stderr, Name ": Failed to find subarray-%s in %s\n",
1590 subarray, dev);
1591 goto free_super;
1592 }
1593 free(info);
1594
33414a01
DW
1595 err = 0;
1596
1597 free_super:
1598 if (err)
1599 st->ss->free_super(st);
1600 free_name:
1601 if (err)
1602 free(st->devname);
1603 free_sysfs:
1604 sysfs_free(mdi);
1605 close_fd:
1606 if (err)
1607 close(fd);
1608
1609 if (err)
1610 return -1;
1611 else
1612 return fd;
1613}
1614
7801ac20
N
1615int add_disk(int mdfd, struct supertype *st,
1616 struct mdinfo *sra, struct mdinfo *info)
1617{
1618 /* Add a device to an array, in one of 2 ways. */
1619 int rv;
1620#ifndef MDASSEMBLE
1621 if (st->ss->external) {
d23534e4
DW
1622 if (info->disk.state & (1<<MD_DISK_SYNC))
1623 info->recovery_start = MaxSector;
1624 else
1625 info->recovery_start = 0;
2904b26f 1626 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1627 if (! rv) {
1628 struct mdinfo *sd2;
f35f2525
N
1629 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1630 if (sd2 == info)
1631 break;
1632 if (sd2 == NULL) {
1633 sd2 = malloc(sizeof(*sd2));
1634 *sd2 = *info;
1635 sd2->next = sra->devs;
1636 sra->devs = sd2;
1637 }
7801ac20
N
1638 }
1639 } else
1640#endif
1641 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1642 return rv;
1643}
1644
de6ae750
N
1645int remove_disk(int mdfd, struct supertype *st,
1646 struct mdinfo *sra, struct mdinfo *info)
1647{
1648 int rv;
1649 /* Remove the disk given by 'info' from the array */
1650#ifndef MDASSEMBLE
1651 if (st->ss->external)
1652 rv = sysfs_set_str(sra, info, "slot", "none");
1653 else
1654#endif
1655 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1656 info->disk.minor));
1657 return rv;
1658}
1659
f35f2525
N
1660int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1661{
1662 /* Initialise kernel's knowledge of array.
1663 * This varies between externally managed arrays
1664 * and older kernels
1665 */
1666 int vers = md_get_version(mdfd);
1667 int rv;
1668
1669#ifndef MDASSEMBLE
1670 if (st->ss->external)
1671 rv = sysfs_set_array(info, vers);
1672 else
1673#endif
1674 if ((vers % 100) >= 1) { /* can use different versions */
1675 mdu_array_info_t inf;
1676 memset(&inf, 0, sizeof(inf));
1677 inf.major_version = info->array.major_version;
1678 inf.minor_version = info->array.minor_version;
1679 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1680 } else
1681 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1682 return rv;
1683}
1684
1e5c6983
DW
1685unsigned long long min_recovery_start(struct mdinfo *array)
1686{
1687 /* find the minimum recovery_start in an array for metadata
1688 * formats that only record per-array recovery progress instead
1689 * of per-device
1690 */
1691 unsigned long long recovery_start = MaxSector;
1692 struct mdinfo *d;
1693
1694 for (d = array->devs; d; d = d->next)
1695 recovery_start = min(recovery_start, d->recovery_start);
1696
1697 return recovery_start;
1698}
1699
11877f4d 1700void fmt_devname(char *name, int num)
2f6079dc 1701{
1f0769d7 1702 if (num >= 0)
2f6079dc
NB
1703 sprintf(name, "md%d", num);
1704 else
1705 sprintf(name, "md_d%d", -1-num);
11877f4d
N
1706}
1707
1708char *devnum2devname(int num)
1709{
1710 char name[100];
1711 fmt_devname(name,num);
2f6079dc
NB
1712 return strdup(name);
1713}
1714
77472ff8
NB
1715int devname2devnum(char *name)
1716{
1717 char *ep;
1718 int num;
1719 if (strncmp(name, "md_d", 4)==0)
1720 num = -1-strtoul(name+4, &ep, 10);
1721 else
1722 num = strtoul(name+2, &ep, 10);
1723 return num;
1724}
1725
c94709e8 1726int stat2devnum(struct stat *st)
2f6079dc 1727{
d7ab966b
N
1728 char path[30];
1729 char link[200];
1730 char *cp;
1731 int n;
1732
c94709e8
DW
1733 if ((S_IFMT & st->st_mode) == S_IFBLK) {
1734 if (major(st->st_rdev) == MD_MAJOR)
1735 return minor(st->st_rdev);
f21e18ca 1736 else if (major(st->st_rdev) == (unsigned)get_mdp_major())
d7ab966b
N
1737 return -1- (minor(st->st_rdev)>>MdpMinorShift);
1738
1739 /* must be an extended-minor partition. Look at the
1740 * /sys/dev/block/%d:%d link which must look like
1741 * ../../block/mdXXX/mdXXXpYY
1742 */
1743 sprintf(path, "/sys/dev/block/%d:%d", major(st->st_rdev),
1744 minor(st->st_rdev));
1745 n = readlink(path, link, sizeof(link)-1);
1746 if (n <= 0)
1747 return NoMdDev;
1748 link[n] = 0;
1749 cp = strrchr(link, '/');
1750 if (cp) *cp = 0;
aa4cab51 1751 cp = strrchr(link, '/');
d7ab966b
N
1752 if (cp && strncmp(cp, "/md", 3) == 0)
1753 return devname2devnum(cp+1);
2f6079dc 1754 }
d7ab966b 1755 return NoMdDev;
c94709e8
DW
1756
1757}
1758
1759int fd2devnum(int fd)
1760{
1761 struct stat stb;
1762 if (fstat(fd, &stb) == 0)
1763 return stat2devnum(&stb);
d7ab966b 1764 return NoMdDev;
2f6079dc
NB
1765}
1766
24f6f99b 1767int mdmon_pid(int devnum)
a931db9e
NB
1768{
1769 char path[100];
1770 char pid[10];
1771 int fd;
1772 int n;
10013317
PHC
1773 char *devname = devnum2devname(devnum);
1774
753cf905 1775 sprintf(path, "%s/%s.pid", MDMON_DIR, devname);
10013317
PHC
1776 free(devname);
1777
24f6f99b 1778 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1779
1780 if (fd < 0)
cf556303 1781 return -1;
a931db9e
NB
1782 n = read(fd, pid, 9);
1783 close(fd);
1784 if (n <= 0)
cf556303 1785 return -1;
24f6f99b 1786 return atoi(pid);
a931db9e
NB
1787}
1788
24f6f99b 1789int mdmon_running(int devnum)
a931db9e 1790{
24f6f99b
N
1791 int pid = mdmon_pid(devnum);
1792 if (pid <= 0)
a931db9e 1793 return 0;
24f6f99b 1794 if (kill(pid, 0) == 0)
a931db9e
NB
1795 return 1;
1796 return 0;
1797}
1798
8850ee3e
N
1799int start_mdmon(int devnum)
1800{
1801 int i;
44d2e365 1802 int len;
9fe32043
N
1803 pid_t pid;
1804 int status;
44d2e365
N
1805 char pathbuf[1024];
1806 char *paths[4] = {
1807 pathbuf,
1808 "/sbin/mdmon",
1809 "mdmon",
1810 NULL
1811 };
8850ee3e 1812
40ebbb9c 1813 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1814 return 0;
1815
44d2e365
N
1816 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf));
1817 if (len > 0) {
1818 char *sl;
1819 pathbuf[len] = 0;
1820 sl = strrchr(pathbuf, '/');
1821 if (sl)
1822 sl++;
1823 else
1824 sl = pathbuf;
1825 strcpy(sl, "mdmon");
1826 } else
1827 pathbuf[0] = '\0';
1828
8850ee3e
N
1829 switch(fork()) {
1830 case 0:
1831 /* FIXME yuk. CLOSE_EXEC?? */
1832 for (i=3; i < 100; i++)
1833 close(i);
44d2e365
N
1834 for (i=0; paths[i]; i++)
1835 if (paths[i][0])
1836 execl(paths[i], "mdmon",
e8a70c89
N
1837 devnum2devname(devnum),
1838 NULL);
8850ee3e
N
1839 exit(1);
1840 case -1: fprintf(stderr, Name ": cannot run mdmon. "
1841 "Array remains readonly\n");
1842 return -1;
9fe32043
N
1843 default: /* parent - good */
1844 pid = wait(&status);
1845 if (pid < 0 || status != 0)
1846 return -1;
8850ee3e
N
1847 }
1848 return 0;
1849}
1850
40ebbb9c 1851int check_env(char *name)
5dcfcb71 1852{
40ebbb9c 1853 char *val = getenv(name);
5dcfcb71
DW
1854
1855 if (val && atoi(val) == 1)
1856 return 1;
1857
1858 return 0;
1859}
1860
148acb7b
DW
1861__u32 random32(void)
1862{
1863 __u32 rv;
1864 int rfd = open("/dev/urandom", O_RDONLY);
1865 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1866 rv = random();
1867 if (rfd >= 0)
1868 close(rfd);
1869 return rv;
1870}
1871
0e600426 1872#ifndef MDASSEMBLE
edd8d13c
NB
1873int flush_metadata_updates(struct supertype *st)
1874{
1875 int sfd;
1876 if (!st->updates) {
1877 st->update_tail = NULL;
1878 return -1;
1879 }
1880
1881 sfd = connect_monitor(devnum2devname(st->container_dev));
1882 if (sfd < 0)
1883 return -1;
1884
1885 while (st->updates) {
1886 struct metadata_update *mu = st->updates;
1887 st->updates = mu->next;
1888
1889 send_message(sfd, mu, 0);
1890 wait_reply(sfd, 0);
1891 free(mu->buf);
1892 free(mu);
1893 }
1894 ack(sfd, 0);
1895 wait_reply(sfd, 0);
1896 close(sfd);
1897 st->update_tail = NULL;
1898 return 0;
1899}
1900
1901void append_metadata_update(struct supertype *st, void *buf, int len)
1902{
1903
1904 struct metadata_update *mu = malloc(sizeof(*mu));
1905
1906 mu->buf = buf;
1907 mu->len = len;
1908 mu->space = NULL;
cb23f1f4 1909 mu->space_list = NULL;
edd8d13c
NB
1910 mu->next = NULL;
1911 *st->update_tail = mu;
1912 st->update_tail = &mu->next;
1913}
0e600426 1914#endif /* MDASSEMBLE */
a931db9e 1915
fe6729fa
NB
1916#ifdef __TINYC__
1917/* tinyc doesn't optimize this check in ioctl.h out ... */
1918unsigned int __invalid_size_argument_for_IOC = 0;
1919#endif
1920
6d11ec6f
AK
1921int experimental(void)
1922{
1923 if (check_env("MDADM_EXPERIMENTAL"))
1924 return 1;
1925 else {
1926 fprintf(stderr, Name ": To use this feature MDADM_EXPERIMENTAL enviroment variable has to defined.\n");
1927 return 0;
1928 }
1929}
1930
326727d9
AC
1931/* Pick all spares matching given criteria from a container
1932 * if min_size == 0 do not check size
1933 * if domlist == NULL do not check domains
1934 * if spare_group given add it to domains of each spare
1935 * metadata allows to test domains using metadata of destination array */
1936struct mdinfo *container_choose_spares(struct supertype *st,
1937 unsigned long long min_size,
1938 struct domainlist *domlist,
1939 char *spare_group,
1940 const char *metadata, int get_one)
1941{
1942 struct mdinfo *d, **dp, *disks = NULL;
1943
1944 /* get list of all disks in container */
1945 if (st->ss->getinfo_super_disks)
1946 disks = st->ss->getinfo_super_disks(st);
1947
1948 if (!disks)
1949 return disks;
1950 /* find spare devices on the list */
1951 dp = &disks->devs;
1952 disks->array.spare_disks = 0;
1953 while (*dp) {
1954 int found = 0;
1955 d = *dp;
1956 if (d->disk.state == 0) {
1957 /* check if size is acceptable */
1958 unsigned long long dev_size;
1959 dev_t dev = makedev(d->disk.major,d->disk.minor);
1960
1961 if (!min_size ||
1962 (dev_size_from_id(dev, &dev_size) &&
1963 dev_size >= min_size))
1964 found = 1;
1965 /* check if domain matches */
1966 if (found && domlist) {
1967 struct dev_policy *pol = devnum_policy(dev);
1968 if (spare_group)
1969 pol_add(&pol, pol_domain,
1970 spare_group, NULL);
e5508b36 1971 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1972 found = 0;
1973 dev_policy_free(pol);
1974 }
1975 }
1976 if (found) {
1977 dp = &d->next;
1978 disks->array.spare_disks++;
1979 if (get_one) {
1980 sysfs_free(*dp);
1981 d->next = NULL;
1982 }
1983 } else {
1984 *dp = d->next;
1985 d->next = NULL;
1986 sysfs_free(d);
1987 }
1988 }
1989 return disks;
1990}