Release mdadm-3.2.2
[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);
9e6d9291 373 for (i=0; i < 1024 && array.nr_disks > 0; i++) {
8453e704
N
374 disk.number = i;
375 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
376 continue;
9e6d9291
N
377 if (disk.major == 0 && disk.minor == 0)
378 continue;
379 array.nr_disks--;
380
8453e704
N
381 if (! (disk.state & (1<<MD_DISK_SYNC)))
382 continue;
383 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
384 continue;
385 avail_disks++;
386 avail[disk.raid_disk] = 1;
387 }
388 /* This is used on an active array, so assume it is clean */
389 return enough(array.level, array.raid_disks, array.layout,
390 1,
391 avail, avail_disks);
392}
393
394
22e263f6
AC
395const int uuid_zero[4] = { 0, 0, 0, 0 };
396
f277ce36 397int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 398{
f277ce36
NB
399 if (swapuuid) {
400 /* parse uuids are hostendian.
401 * uuid's from some superblocks are big-ending
aba69144 402 * if there is a difference, we need to swap..
f277ce36
NB
403 */
404 unsigned char *ac = (unsigned char *)a;
405 unsigned char *bc = (unsigned char *)b;
406 int i;
407 for (i=0; i<16; i+= 4) {
408 if (ac[i+0] != bc[i+3] ||
409 ac[i+1] != bc[i+2] ||
410 ac[i+2] != bc[i+1] ||
411 ac[i+3] != bc[i+0])
412 return 0;
413 }
414 return 1;
415 } else {
416 if (a[0]==b[0] &&
417 a[1]==b[1] &&
418 a[2]==b[2] &&
419 a[3]==b[3])
420 return 1;
421 return 0;
422 }
64c4757e 423}
350f29f9
NB
424void copy_uuid(void *a, int b[4], int swapuuid)
425{
426 if (swapuuid) {
427 /* parse uuids are hostendian.
428 * uuid's from some superblocks are big-ending
429 * if there is a difference, we need to swap..
430 */
431 unsigned char *ac = (unsigned char *)a;
432 unsigned char *bc = (unsigned char *)b;
433 int i;
434 for (i=0; i<16; i+= 4) {
435 ac[i+0] = bc[i+3];
436 ac[i+1] = bc[i+2];
437 ac[i+2] = bc[i+1];
438 ac[i+3] = bc[i+0];
439 }
440 } else
441 memcpy(a, b, 16);
442}
64c4757e 443
aae5a112 444char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 445{
9968e376 446 int i, j;
d7288ddc
N
447 char uuid[16];
448 char *c = buf;
449 strcpy(c, "UUID-");
450 c += strlen(c);
aae5a112 451 copy_uuid(uuid, id, swap);
9968e376 452 for (i = 0; i < 4; i++) {
9968e376 453 if (i)
ff54de6e 454 *c++ = sep;
9968e376
DW
455 for (j = 3; j >= 0; j--) {
456 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
457 c+= 2;
458 }
d7288ddc
N
459 }
460 return buf;
aae5a112
DW
461
462}
463
464char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
465{
86983cce
N
466 // dirty hack to work around an issue with super1 superblocks...
467 // super1 superblocks need swapuuid set in order for assembly to
468 // work, but can't have it set if we want this printout to match
469 // all the other uuid printouts in super1.c, so we force swapuuid
470 // to 1 to make our printout match the rest of super1
471 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
472}
473
435d4ebb 474#ifndef MDASSEMBLE
682c7051
NB
475int check_ext2(int fd, char *name)
476{
477 /*
478 * Check for an ext2fs file system.
479 * Superblock is always 1K at 1K offset
480 *
481 * s_magic is le16 at 56 == 0xEF53
482 * report mtime - le32 at 44
483 * blocks - le32 at 4
484 * logblksize - le32 at 24
485 */
486 unsigned char sb[1024];
487 time_t mtime;
488 int size, bsize;
489 if (lseek(fd, 1024,0)!= 1024)
490 return 0;
491 if (read(fd, sb, 1024)!= 1024)
492 return 0;
493 if (sb[56] != 0x53 || sb[57] != 0xef)
494 return 0;
495
496 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
497 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
498 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
499 fprintf(stderr, Name ": %s appears to contain an ext2fs file system\n",
500 name);
501 fprintf(stderr," size=%dK mtime=%s",
502 size*(1<<bsize), ctime(&mtime));
503 return 1;
504}
505
506int check_reiser(int fd, char *name)
507{
508 /*
509 * superblock is at 64K
510 * size is 1024;
511 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
512 *
513 */
514 unsigned char sb[1024];
881990a2 515 unsigned long size;
682c7051
NB
516 if (lseek(fd, 64*1024, 0) != 64*1024)
517 return 0;
518 if (read(fd, sb, 1024) != 1024)
519 return 0;
a46f4061
NB
520 if (strncmp((char*)sb+52, "ReIsErFs",8)!=0 &&
521 strncmp((char*)sb+52, "ReIsEr2Fs",9)!=0)
682c7051
NB
522 return 0;
523 fprintf(stderr, Name ": %s appears to contain a reiserfs file system\n",name);
524 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
881990a2 525 fprintf(stderr, " size = %luK\n", size*4);
aba69144 526
682c7051
NB
527 return 1;
528}
529
530int check_raid(int fd, char *name)
531{
4b1ac34b 532 struct mdinfo info;
682c7051 533 time_t crtime;
d078d77c 534 char *level;
82d9eba6 535 struct supertype *st = guess_super(fd);
f9ce90ba 536
82d9eba6 537 if (!st) return 0;
3da92f27 538 st->ss->load_super(st, fd, name);
82d9eba6
NB
539 /* Looks like a raid array .. */
540 fprintf(stderr, Name ": %s appears to be part of a raid array:\n",
541 name);
a5d85af7 542 st->ss->getinfo_super(st, &info, NULL);
3da92f27 543 st->ss->free_super(st);
82d9eba6 544 crtime = info.array.ctime;
d078d77c
NB
545 level = map_num(pers, info.array.level);
546 if (!level) level = "-unknown-";
547 fprintf(stderr, " level=%s devices=%d ctime=%s",
548 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 549 return 1;
682c7051
NB
550}
551
682c7051
NB
552int ask(char *mesg)
553{
554 char *add = "";
555 int i;
556 for (i=0; i<5; i++) {
557 char buf[100];
558 fprintf(stderr, "%s%s", mesg, add);
559 fflush(stderr);
560 if (fgets(buf, 100, stdin)==NULL)
561 return 0;
562 if (buf[0]=='y' || buf[0]=='Y')
563 return 1;
564 if (buf[0]=='n' || buf[0]=='N')
565 return 0;
566 add = "(y/n) ";
567 }
568 fprintf(stderr, Name ": assuming 'no'\n");
569 return 0;
570}
435d4ebb 571#endif /* MDASSEMBLE */
682c7051 572
8d80900b 573int is_standard(char *dev, int *nump)
e5329c37
NB
574{
575 /* tests if dev is a "standard" md dev name.
576 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 577 * where NN is a string of digits
598f0d58
NB
578 * Returns 1 if a partitionable standard,
579 * -1 if non-partitonable,
580 * 0 if not a standard name.
e5329c37 581 */
8d80900b
NB
582 char *d = strrchr(dev, '/');
583 int type=0;
584 int num;
585 if (!d)
e5329c37 586 return 0;
8d80900b
NB
587 if (strncmp(d, "/d",2)==0)
588 d += 2, type=1; /* /dev/md/dN{pM} */
589 else if (strncmp(d, "/md_d", 5)==0)
2b4ca8f0 590 d += 5, type=1; /* /dev/md_dN{pM} */
8d80900b
NB
591 else if (strncmp(d, "/md", 3)==0)
592 d += 3, type=-1; /* /dev/mdN */
593 else if (d-dev > 3 && strncmp(d-2, "md/", 3)==0)
5a6d1148 594 d += 1, type=-1; /* /dev/md/N */
e5329c37
NB
595 else
596 return 0;
8d80900b 597 if (!*d)
e5329c37 598 return 0;
8d80900b
NB
599 num = atoi(d);
600 while (isdigit(*d))
601 d++;
602 if (*d)
e5329c37 603 return 0;
8d80900b
NB
604 if (nump) *nump = num;
605
606 return type;
e5329c37
NB
607}
608
4b1ac34b 609unsigned long calc_csum(void *super, int bytes)
82b27616 610{
56eb10c0 611 unsigned long long newcsum = 0;
82b27616 612 int i;
4b1ac34b
NB
613 unsigned int csum;
614 unsigned int *superc = (unsigned int*) super;
82b27616 615
4b1ac34b 616 for(i=0; i<bytes/4; i++)
82b27616
NB
617 newcsum+= superc[i];
618 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 619#ifdef __alpha__
aba69144 620/* The in-kernel checksum calculation is always 16bit on
570c0542
NB
621 * the alpha, though it is 32 bit on i386...
622 * I wonder what it is elsewhere... (it uses and API in
623 * a way that it shouldn't).
624 */
625 csum = (csum & 0xffff) + (csum >> 16);
626 csum = (csum & 0xffff) + (csum >> 16);
627#endif
82b27616
NB
628 return csum;
629}
cd29a5c8 630
435d4ebb 631#ifndef MDASSEMBLE
56eb10c0 632char *human_size(long long bytes)
cd29a5c8
NB
633{
634 static char buf[30];
d5d3721e
NB
635
636 /* We convert bytes to either centi-M{ega,ibi}bytes or
637 * centi-G{igi,ibi}bytes, with appropriate rounding,
638 * and then print 1/100th of those as a decimal.
639 * We allow upto 2048Megabytes before converting to
640 * gigabytes, as that shows more precision and isn't
641 * too large a number.
642 * Terrabytes are not yet handled.
643 */
cd29a5c8 644
56eb10c0 645 if (bytes < 5000*1024)
cd29a5c8 646 buf[0]=0;
d5d3721e
NB
647 else if (bytes < 2*1024LL*1024LL*1024LL) {
648 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
649 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 650 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
651 cMiB/100 , cMiB % 100,
652 cMB/100, cMB % 100);
653 } else {
654 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
655 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 656 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
657 cGiB/100 , cGiB % 100,
658 cGB/100, cGB % 100);
659 }
cd29a5c8
NB
660 return buf;
661}
e0d19036
NB
662
663char *human_size_brief(long long bytes)
664{
665 static char buf[30];
e0d19036
NB
666
667 if (bytes < 5000*1024)
8f23b0b3 668 snprintf(buf, sizeof(buf), "%ld.%02ldKiB",
bd526cee 669 (long)(bytes>>10), (long)(((bytes&1023)*100+512)/1024)
e0d19036
NB
670 );
671 else if (bytes < 2*1024LL*1024LL*1024LL)
8f23b0b3 672 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
e0d19036 673 (long)(bytes>>20),
bd526cee 674 (long)((bytes&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
675 );
676 else
8f23b0b3 677 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
e0d19036 678 (long)(bytes>>30),
bd526cee 679 (long)(((bytes>>10)&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
680 );
681 return buf;
682}
e4965ef8
N
683
684void print_r10_layout(int layout)
685{
686 int near = layout & 255;
687 int far = (layout >> 8) & 255;
688 int offset = (layout&0x10000);
689 char *sep = "";
690
691 if (near != 1) {
692 printf("%s near=%d", sep, near);
693 sep = ",";
694 }
695 if (far != 1)
696 printf("%s %s=%d", sep, offset?"offset":"far", far);
697 if (near*far == 1)
698 printf("NO REDUNDANCY");
699}
435d4ebb 700#endif
e0d19036 701
5f8097be
NB
702unsigned long long calc_array_size(int level, int raid_disks, int layout,
703 int chunksize, unsigned long long devsize)
704{
705 int data_disks = 0;
706 switch (level) {
707 case 0: data_disks = raid_disks; break;
708 case 1: data_disks = 1; break;
709 case 4:
710 case 5: data_disks = raid_disks - 1; break;
711 case 6: data_disks = raid_disks - 2; break;
712 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
713 break;
714 }
715 devsize &= ~(unsigned long long)((chunksize>>9)-1);
716 return data_disks * devsize;
717}
718
0e600426 719#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
e0d19036
NB
720char *get_md_name(int dev)
721{
722 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 723 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
e0d19036
NB
724 static char devname[50];
725 struct stat stb;
98c6faba 726 dev_t rdev;
dd0781e5 727 char *dn;
98c6faba
NB
728
729 if (dev < 0) {
dd0781e5
NB
730 int mdp = get_mdp_major();
731 if (mdp < 0) return NULL;
0df46c2a 732 rdev = makedev(mdp, (-1-dev)<<6);
8f23b0b3 733 snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev);
98c6faba
NB
734 if (stat(devname, &stb) == 0
735 && (S_IFMT&stb.st_mode) == S_IFBLK
736 && (stb.st_rdev == rdev))
737 return devname;
738 } else {
0df46c2a 739 rdev = makedev(MD_MAJOR, dev);
8f23b0b3 740 snprintf(devname, sizeof(devname), "/dev/md%d", dev);
98c6faba
NB
741 if (stat(devname, &stb) == 0
742 && (S_IFMT&stb.st_mode) == S_IFBLK
743 && (stb.st_rdev == rdev))
744 return devname;
745
8f23b0b3 746 snprintf(devname, sizeof(devname), "/dev/md/%d", dev);
98c6faba
NB
747 if (stat(devname, &stb) == 0
748 && (S_IFMT&stb.st_mode) == S_IFBLK
749 && (stb.st_rdev == rdev))
750 return devname;
751 }
16c6fa80 752 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
753 if (dn)
754 return dn;
8f23b0b3 755 snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev);
e0d19036 756 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
757 if (errno != EEXIST)
758 return NULL;
e0d19036
NB
759
760 if (stat(devname, &stb) == 0
761 && (S_IFMT&stb.st_mode) == S_IFBLK
762 && (stb.st_rdev == rdev))
763 return devname;
764 unlink(devname);
765 return NULL;
766}
767
768void put_md_name(char *name)
769{
770 if (strncmp(name, "/dev/.tmp.md", 12)==0)
771 unlink(name);
772}
ea24acd0 773
ea24acd0
NB
774int find_free_devnum(int use_partitions)
775{
776 int devnum;
777 for (devnum = 127; devnum != 128;
a56fb7ec 778 devnum = devnum ? devnum-1 : (1<<20)-1) {
ea24acd0
NB
779 char *dn;
780 int _devnum;
781
782 _devnum = use_partitions ? (-1-devnum) : devnum;
783 if (mddev_busy(_devnum))
784 continue;
785 /* make sure it is new to /dev too, at least as a
786 * non-standard */
787 dn = map_dev(dev2major(_devnum), dev2minor(_devnum), 0);
788 if (dn && ! is_standard(dn, NULL))
789 continue;
790 break;
791 }
792 if (devnum == 128)
793 return NoMdDev;
794 return use_partitions ? (-1-devnum) : devnum;
795}
435d4ebb 796#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 797
8b0dabea
NB
798int dev_open(char *dev, int flags)
799{
800 /* like 'open', but if 'dev' matches %d:%d, create a temp
801 * block device and open that
802 */
803 char *e;
804 int fd = -1;
805 char devname[32];
e81cdd9f 806 int major;
8b0dabea 807 int minor;
e81cdd9f
NB
808
809 if (!dev) return -1;
6df6a774 810 flags |= O_DIRECT;
e81cdd9f
NB
811
812 major = strtoul(dev, &e, 0);
8b0dabea
NB
813 if (e > dev && *e == ':' && e[1] &&
814 (minor = strtoul(e+1, &e, 0)) >= 0 &&
815 *e == 0) {
6df6a774
N
816 char *path = map_dev(major, minor, 0);
817 if (path)
818 fd = open(path, flags);
819 if (fd < 0) {
820 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
821 (int)getpid(), major, minor);
822 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
823 fd = open(devname, flags);
824 unlink(devname);
825 }
826 }
827 if (fd < 0) {
828 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
829 (int)getpid(), major, minor);
830 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
831 fd = open(devname, flags);
832 unlink(devname);
833 }
8b0dabea
NB
834 }
835 } else
6df6a774 836 fd = open(dev, flags);
8b0dabea
NB
837 return fd;
838}
f9ce90ba 839
d998b738 840int open_dev_flags(int devnum, int flags)
e8a70c89
N
841{
842 char buf[20];
843
844 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
d998b738
N
845 return dev_open(buf, flags);
846}
847
848int open_dev(int devnum)
849{
850 return open_dev_flags(devnum, O_RDONLY);
e8a70c89
N
851}
852
a931db9e
NB
853int open_dev_excl(int devnum)
854{
855 char buf[20];
856 int i;
7187750e 857 int flags = O_RDWR;
a931db9e
NB
858
859 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
860 for (i=0 ; i<25 ; i++) {
7187750e 861 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
862 if (fd >= 0)
863 return fd;
7187750e
N
864 if (errno == EACCES && flags == O_RDWR) {
865 flags = O_RDONLY;
866 continue;
867 }
a931db9e
NB
868 if (errno != EBUSY)
869 return fd;
870 usleep(200000);
871 }
872 return -1;
873}
874
9008ed1c
N
875int same_dev(char *one, char *two)
876{
877 struct stat st1, st2;
878 if (stat(one, &st1) != 0)
879 return 0;
880 if (stat(two, &st2) != 0)
881 return 0;
882 if ((st1.st_mode & S_IFMT) != S_IFBLK)
883 return 0;
884 if ((st2.st_mode & S_IFMT) != S_IFBLK)
885 return 0;
886 return st1.st_rdev == st2.st_rdev;
887}
888
a7c6e3fb 889void wait_for(char *dev, int fd)
a714580e
N
890{
891 int i;
a7c6e3fb
N
892 struct stat stb_want;
893
894 if (fstat(fd, &stb_want) != 0 ||
895 (stb_want.st_mode & S_IFMT) != S_IFBLK)
896 return;
a714580e
N
897
898 for (i=0 ; i<25 ; i++) {
899 struct stat stb;
a7c6e3fb
N
900 if (stat(dev, &stb) == 0 &&
901 (stb.st_mode & S_IFMT) == S_IFBLK &&
902 (stb.st_rdev == stb_want.st_rdev))
a714580e
N
903 return;
904 usleep(200000);
905 }
436305c6
DW
906 if (i == 25)
907 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
908}
909
0f22b998
N
910struct superswitch *superlist[] =
911{
912 &super0, &super1,
913 &super_ddf, &super_imsm,
0592faeb 914 &mbr, &gpt,
0f22b998 915 NULL };
f9ce90ba 916
ea24acd0 917#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 918
4725bc31 919struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 920{
1686dc25
NB
921 mdu_array_info_t array;
922 int vers;
923 int minor;
924 struct supertype *st = NULL;
7e0f6979 925 struct mdinfo *sra;
142cb9e1 926 char *verstr;
1686dc25
NB
927 char version[20];
928 int i;
f7e7067b 929 char *subarray = NULL;
5f7e44b2 930 int container = NoMdDev;
1686dc25
NB
931
932 sra = sysfs_read(fd, 0, GET_VERSION);
933
934 if (sra) {
7e0f6979
NB
935 vers = sra->array.major_version;
936 minor = sra->array.minor_version;
142cb9e1 937 verstr = sra->text_version;
1686dc25
NB
938 } else {
939 if (ioctl(fd, GET_ARRAY_INFO, &array))
940 array.major_version = array.minor_version = 0;
941 vers = array.major_version;
942 minor = array.minor_version;
142cb9e1 943 verstr = "";
6fbba4c9 944 }
82d9eba6 945
1686dc25
NB
946 if (vers != -1) {
947 sprintf(version, "%d.%d", vers, minor);
948 verstr = version;
6fbba4c9 949 }
3c558363 950 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 951 char *dev = verstr+1;
5f7e44b2 952
f7e7067b 953 subarray = strchr(dev, '/');
f7e7067b
NB
954 if (subarray)
955 *subarray++ = '\0';
f7e7067b 956 subarray = strdup(subarray);
5f7e44b2 957 container = devname2devnum(dev);
f7e7067b
NB
958 if (sra)
959 sysfs_free(sra);
5f7e44b2 960 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
961 if (sra && sra->text_version[0])
962 verstr = sra->text_version;
963 else
964 verstr = "-no-metadata-";
f7e7067b
NB
965 }
966
967 for (i = 0; st == NULL && superlist[i] ; i++)
968 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
969
970 if (sra)
971 sysfs_free(sra);
f7e7067b 972 if (st) {
3b0896f8 973 st->sb = NULL;
1f49fb3a
N
974 if (subarrayp)
975 *subarrayp = subarray;
5f7e44b2
DW
976 st->container_dev = container;
977 st->devnum = fd2devnum(fd);
4725bc31
N
978 } else
979 free(subarray);
5f7e44b2 980
82d9eba6 981 return st;
f9ce90ba 982}
ea24acd0
NB
983#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
984
0f0749ad 985int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
986{
987 char buf[20];
988 int fd;
989
990 sprintf(buf, "%d:%d", major(id), minor(id));
991 fd = dev_open(buf, O_RDONLY);
992 if (fd < 0)
993 return 0;
994 if (get_dev_size(fd, NULL, size)) {
995 close(fd);
996 return 1;
997 }
998 close(fd);
999 return 0;
1000}
f9ce90ba 1001
159c3a1a 1002struct supertype *dup_super(struct supertype *orig)
3da92f27 1003{
159c3a1a 1004 struct supertype *st;
1686dc25 1005
d2ca6449
NB
1006 if (!orig)
1007 return orig;
159c3a1a 1008 st = malloc(sizeof(*st));
3da92f27
NB
1009 if (!st)
1010 return st;
ef609477 1011 memset(st, 0, sizeof(*st));
159c3a1a
NB
1012 st->ss = orig->ss;
1013 st->max_devs = orig->max_devs;
1014 st->minor_version = orig->minor_version;
1015 st->sb = NULL;
1016 st->info = NULL;
1017 return st;
3da92f27
NB
1018}
1019
54887ad8 1020struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1021{
1022 /* try each load_super to find the best match,
1023 * and return the best superswitch
1024 */
82d9eba6
NB
1025 struct superswitch *ss;
1026 struct supertype *st;
f21e18ca 1027 time_t besttime = 0;
570c0542 1028 int bestsuper = -1;
f9ce90ba
NB
1029 int i;
1030
82d9eba6 1031 st = malloc(sizeof(*st));
d1d599ea
N
1032 memset(st, 0, sizeof(*st));
1033 st->container_dev = NoMdDev;
1034
f9ce90ba
NB
1035 for (i=0 ; superlist[i]; i++) {
1036 int rv;
1037 ss = superlist[i];
54887ad8
N
1038 if (guess_type == guess_array && ss->add_to_super == NULL)
1039 continue;
1040 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1041 continue;
ef609477 1042 memset(st, 0, sizeof(*st));
df3346e6 1043 st->ignore_hw_compat = 1;
3da92f27 1044 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1045 if (rv == 0) {
1046 struct mdinfo info;
a5d85af7 1047 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1048 if (bestsuper == -1 ||
1049 besttime < info.array.ctime) {
1050 bestsuper = i;
1051 besttime = info.array.ctime;
570c0542 1052 }
3da92f27 1053 ss->free_super(st);
570c0542
NB
1054 }
1055 }
1056 if (bestsuper != -1) {
1057 int rv;
ef609477 1058 memset(st, 0, sizeof(*st));
df3346e6 1059 st->ignore_hw_compat = 1;
3da92f27 1060 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1061 if (rv == 0) {
5e747af2 1062 superlist[bestsuper]->free_super(st);
df3346e6 1063 st->ignore_hw_compat = 0;
82d9eba6 1064 return st;
f9ce90ba
NB
1065 }
1066 }
570c0542 1067 free(st);
f9ce90ba
NB
1068 return NULL;
1069}
fe6729fa 1070
beae1dfe
NB
1071/* Return size of device in bytes */
1072int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1073{
1074 unsigned long long ldsize;
c2c9bb6f
NB
1075 struct stat st;
1076
1077 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1078 ldsize = (unsigned long long)st.st_size;
1079 else
beae1dfe
NB
1080#ifdef BLKGETSIZE64
1081 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1082#endif
1083 {
1084 unsigned long dsize;
1085 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1086 ldsize = dsize;
1087 ldsize <<= 9;
1088 } else {
1089 if (dname)
1090 fprintf(stderr, Name ": Cannot get size of %s: %s\b",
1091 dname, strerror(errno));
1092 return 0;
1093 }
1094 }
1095 *sizep = ldsize;
1096 return 1;
1097}
8fac0577 1098
3a371610
N
1099/* Return true if this can only be a container, not a member device.
1100 * i.e. is and md device and size is zero
1101 */
1102int must_be_container(int fd)
1103{
1104 unsigned long long size;
1105 if (md_get_version(fd) < 0)
1106 return 0;
1107 if (get_dev_size(fd, NULL, &size) == 0)
1108 return 1;
1109 if (size == 0)
1110 return 1;
1111 return 0;
1112}
034b203a
TM
1113
1114/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1115 * Returns: 1 if successful
1116 * -1 for unknown partition type
1117 * 0 for other errors
1118 */
1119static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1120{
056b331e 1121 struct GPT gpt;
034b203a 1122 unsigned char empty_gpt_entry[16]= {0};
73e658d8 1123 struct GPT_part_entry part;
034b203a
TM
1124 unsigned long long curr_part_end;
1125 unsigned all_partitions, entry_size;
f21e18ca 1126 unsigned part_nr;
034b203a
TM
1127
1128 *endofpart = 0;
1129
056b331e 1130 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1131 /* skip protective MBR */
034b203a 1132 lseek(fd, 512, SEEK_SET);
73e658d8 1133 /* read GPT header */
056b331e 1134 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1135 return 0;
1136
1137 /* get the number of partition entries and the entry size */
056b331e
N
1138 all_partitions = __le32_to_cpu(gpt.part_cnt);
1139 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1140
1141 /* Check GPT signature*/
056b331e 1142 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1143 return -1;
1144
1145 /* sanity checks */
1146 if (all_partitions > 1024 ||
1147 entry_size > 512)
1148 return -1;
1149
034b203a 1150 for (part_nr=0; part_nr < all_partitions; part_nr++) {
73e658d8
LB
1151 /* read partition entry */
1152 if (read(fd, &part, entry_size) != (ssize_t)entry_size)
1153 return 0;
1154
034b203a 1155 /* is this valid partition? */
73e658d8 1156 if (memcmp(part.type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1157 /* check the last lba for the current partition */
73e658d8 1158 curr_part_end = __le64_to_cpu(part.ending_lba);
034b203a
TM
1159 if (curr_part_end > *endofpart)
1160 *endofpart = curr_part_end;
1161 }
1162
034b203a
TM
1163 }
1164 return 1;
1165}
1166
1167/* Sets endofpart parameter to the last block used by the last partition on the device.
1168 * Returns: 1 if successful
1169 * -1 for unknown partition type
1170 * 0 for other errors
1171 */
1172static int get_last_partition_end(int fd, unsigned long long *endofpart)
1173{
056b331e 1174 struct MBR boot_sect;
034b203a
TM
1175 struct MBR_part_record *part;
1176 unsigned long long curr_part_end;
f21e18ca 1177 unsigned part_nr;
034b203a
TM
1178 int retval = 0;
1179
1180 *endofpart = 0;
1181
056b331e 1182 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1183 /* read MBR */
1184 lseek(fd, 0, 0);
056b331e 1185 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1186 goto abort;
1187
1188 /* check MBP signature */
056b331e 1189 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1190 retval = 1;
1191 /* found the correct signature */
056b331e 1192 part = boot_sect.parts;
034b203a
TM
1193
1194 for (part_nr=0; part_nr < MBR_PARTITIONS; part_nr++) {
1195 /* check for GPT type */
1196 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1197 retval = get_gpt_last_partition_end(fd, endofpart);
1198 break;
1199 }
1200 /* check the last used lba for the current partition */
1201 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1202 __le32_to_cpu(part->blocks_num);
1203 if (curr_part_end > *endofpart)
1204 *endofpart = curr_part_end;
1205
1206 part++;
1207 }
1208 } else {
1209 /* Unknown partition table */
1210 retval = -1;
1211 }
1212 abort:
1213 return retval;
1214}
1215
53ed6ac3
KW
1216int check_partitions(int fd, char *dname, unsigned long long freesize,
1217 unsigned long long size)
034b203a
TM
1218{
1219 /*
1220 * Check where the last partition ends
1221 */
1222 unsigned long long endofpart;
1223 int ret;
1224
1225 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1226 /* There appears to be a partition table here */
1227 if (freesize == 0) {
1228 /* partitions will not be visible in new device */
1229 fprintf(stderr,
1230 Name ": partition table exists on %s but will be lost or\n"
1231 " meaningless after creating array\n",
1232 dname);
1233 return 1;
1234 } else if (endofpart > freesize) {
1235 /* last partition overlaps metadata */
1236 fprintf(stderr,
1237 Name ": metadata will over-write last partition on %s.\n",
1238 dname);
1239 return 1;
53ed6ac3
KW
1240 } else if (size && endofpart > size) {
1241 /* partitions will be truncated in new device */
1242 fprintf(stderr,
1243 Name ": array size is too small to cover all partitions on %s.\n",
1244 dname);
1245 return 1;
034b203a
TM
1246 }
1247 }
1248 return 0;
1249}
1250
8382f19b
NB
1251void get_one_disk(int mdfd, mdu_array_info_t *ainf, mdu_disk_info_t *disk)
1252{
1253 int d;
9e6d9291 1254
8382f19b 1255 ioctl(mdfd, GET_ARRAY_INFO, ainf);
9e6d9291
N
1256 for (d = 0 ; d < 1024 ; d++) {
1257 if (ioctl(mdfd, GET_DISK_INFO, disk) == 0 &&
1258 (disk->major || disk->minor))
8382f19b 1259 return;
9e6d9291 1260 }
8382f19b 1261}
63152c1b 1262
a322f70c
DW
1263int open_container(int fd)
1264{
1265 /* 'fd' is a block device. Find out if it is in use
1266 * by a container, and return an open fd on that container.
1267 */
1268 char path[256];
1269 char *e;
1270 DIR *dir;
1271 struct dirent *de;
1272 int dfd, n;
1273 char buf[200];
1274 int major, minor;
1275 struct stat st;
1276
1277 if (fstat(fd, &st) != 0)
1278 return -1;
1279 sprintf(path, "/sys/dev/block/%d:%d/holders",
1280 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1281 e = path + strlen(path);
1282
1283 dir = opendir(path);
1284 if (!dir)
1285 return -1;
1286 while ((de = readdir(dir))) {
1287 if (de->d_ino == 0)
1288 continue;
1289 if (de->d_name[0] == '.')
1290 continue;
1291 sprintf(e, "/%s/dev", de->d_name);
1292 dfd = open(path, O_RDONLY);
1293 if (dfd < 0)
1294 continue;
1295 n = read(dfd, buf, sizeof(buf));
1296 close(dfd);
f21e18ca 1297 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1298 continue;
1299 buf[n] = 0;
1300 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1301 continue;
1302 sprintf(buf, "%d:%d", major, minor);
1303 dfd = dev_open(buf, O_RDONLY);
1304 if (dfd >= 0) {
1305 closedir(dir);
1306 return dfd;
1307 }
1308 }
355726fa 1309 closedir(dir);
a322f70c
DW
1310 return -1;
1311}
1312
33414a01
DW
1313struct superswitch *version_to_superswitch(char *vers)
1314{
1315 int i;
1316
1317 for (i = 0; superlist[i]; i++) {
1318 struct superswitch *ss = superlist[i];
1319
1320 if (strcmp(vers, ss->name) == 0)
1321 return ss;
1322 }
1323
1324 return NULL;
1325}
1326
1327int is_container_member(struct mdstat_ent *mdstat, char *container)
1328{
1329 if (mdstat->metadata_version == NULL ||
1330 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
1331 !is_subarray(mdstat->metadata_version+9) ||
1332 strncmp(mdstat->metadata_version+10, container, strlen(container)) != 0 ||
1333 mdstat->metadata_version[10+strlen(container)] != '/')
1334 return 0;
1335
1336 return 1;
1337}
1338
1339int is_subarray_active(char *subarray, char *container)
1340{
1341 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1342 struct mdstat_ent *ent;
1343
e5408a32
DW
1344 for (ent = mdstat; ent; ent = ent->next)
1345 if (is_container_member(ent, container))
e5e5d7ce 1346 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1347 break;
33414a01
DW
1348
1349 free_mdstat(mdstat);
1350
1351 return ent != NULL;
1352}
1353
1354/* open_subarray - opens a subarray in a container
1355 * @dev: container device name
feab51f8 1356 * @st: empty supertype
33414a01
DW
1357 * @quiet: block reporting errors flag
1358 *
1359 * On success returns an fd to a container and fills in *st
1360 */
feab51f8 1361int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1362{
1363 struct mdinfo *mdi;
a951a4f7 1364 struct mdinfo *info;
33414a01
DW
1365 int fd, err = 1;
1366
1367 fd = open(dev, O_RDWR|O_EXCL);
1368 if (fd < 0) {
1369 if (!quiet)
1370 fprintf(stderr, Name ": Couldn't open %s, aborting\n",
1371 dev);
1372 return 2;
1373 }
1374
1375 st->devnum = fd2devnum(fd);
1376 if (st->devnum == NoMdDev) {
1377 if (!quiet)
1378 fprintf(stderr,
1379 Name ": Failed to determine device number for %s\n",
1380 dev);
1381 goto close_fd;
1382 }
1383
1384 mdi = sysfs_read(fd, st->devnum, GET_VERSION|GET_LEVEL);
1385 if (!mdi) {
1386 if (!quiet)
1387 fprintf(stderr, Name ": Failed to read sysfs for %s\n",
1388 dev);
1389 goto close_fd;
1390 }
1391
1392 if (mdi->array.level != UnSet) {
1393 if (!quiet)
1394 fprintf(stderr, Name ": %s is not a container\n", dev);
1395 goto free_sysfs;
1396 }
1397
1398 st->ss = version_to_superswitch(mdi->text_version);
1399 if (!st->ss) {
1400 if (!quiet)
1401 fprintf(stderr,
1402 Name ": Operation not supported for %s metadata\n",
1403 mdi->text_version);
1404 goto free_sysfs;
1405 }
1406
1407 st->devname = devnum2devname(st->devnum);
1408 if (!st->devname) {
1409 if (!quiet)
1410 fprintf(stderr, Name ": Failed to allocate device name\n");
1411 goto free_sysfs;
1412 }
1413
db20d413 1414 if (!st->ss->load_container) {
33414a01 1415 if (!quiet)
db20d413 1416 fprintf(stderr, Name ": %s is not a container\n", dev);
33414a01
DW
1417 goto free_name;
1418 }
1419
db20d413 1420 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1421 if (!quiet)
db20d413
N
1422 fprintf(stderr, Name ": Failed to load metadata for %s\n",
1423 dev);
1424 goto free_name;
33414a01
DW
1425 }
1426
a951a4f7
N
1427 info = st->ss->container_content(st, subarray);
1428 if (!info) {
1429 if (!quiet)
1430 fprintf(stderr, Name ": Failed to find subarray-%s in %s\n",
1431 subarray, dev);
1432 goto free_super;
1433 }
1434 free(info);
1435
33414a01
DW
1436 err = 0;
1437
1438 free_super:
1439 if (err)
1440 st->ss->free_super(st);
1441 free_name:
1442 if (err)
1443 free(st->devname);
1444 free_sysfs:
1445 sysfs_free(mdi);
1446 close_fd:
1447 if (err)
1448 close(fd);
1449
1450 if (err)
1451 return -1;
1452 else
1453 return fd;
1454}
1455
7801ac20
N
1456int add_disk(int mdfd, struct supertype *st,
1457 struct mdinfo *sra, struct mdinfo *info)
1458{
1459 /* Add a device to an array, in one of 2 ways. */
1460 int rv;
1461#ifndef MDASSEMBLE
1462 if (st->ss->external) {
d23534e4
DW
1463 if (info->disk.state & (1<<MD_DISK_SYNC))
1464 info->recovery_start = MaxSector;
1465 else
1466 info->recovery_start = 0;
2904b26f 1467 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1468 if (! rv) {
1469 struct mdinfo *sd2;
f35f2525
N
1470 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1471 if (sd2 == info)
1472 break;
1473 if (sd2 == NULL) {
1474 sd2 = malloc(sizeof(*sd2));
1475 *sd2 = *info;
1476 sd2->next = sra->devs;
1477 sra->devs = sd2;
1478 }
7801ac20
N
1479 }
1480 } else
1481#endif
1482 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1483 return rv;
1484}
1485
de6ae750
N
1486int remove_disk(int mdfd, struct supertype *st,
1487 struct mdinfo *sra, struct mdinfo *info)
1488{
1489 int rv;
1490 /* Remove the disk given by 'info' from the array */
1491#ifndef MDASSEMBLE
1492 if (st->ss->external)
1493 rv = sysfs_set_str(sra, info, "slot", "none");
1494 else
1495#endif
1496 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1497 info->disk.minor));
1498 return rv;
1499}
1500
f35f2525
N
1501int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1502{
1503 /* Initialise kernel's knowledge of array.
1504 * This varies between externally managed arrays
1505 * and older kernels
1506 */
1507 int vers = md_get_version(mdfd);
1508 int rv;
1509
1510#ifndef MDASSEMBLE
1511 if (st->ss->external)
1512 rv = sysfs_set_array(info, vers);
1513 else
1514#endif
1515 if ((vers % 100) >= 1) { /* can use different versions */
1516 mdu_array_info_t inf;
1517 memset(&inf, 0, sizeof(inf));
1518 inf.major_version = info->array.major_version;
1519 inf.minor_version = info->array.minor_version;
1520 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1521 } else
1522 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1523 return rv;
1524}
1525
1e5c6983
DW
1526unsigned long long min_recovery_start(struct mdinfo *array)
1527{
1528 /* find the minimum recovery_start in an array for metadata
1529 * formats that only record per-array recovery progress instead
1530 * of per-device
1531 */
1532 unsigned long long recovery_start = MaxSector;
1533 struct mdinfo *d;
1534
1535 for (d = array->devs; d; d = d->next)
1536 recovery_start = min(recovery_start, d->recovery_start);
1537
1538 return recovery_start;
1539}
1540
24f6f99b 1541int mdmon_pid(int devnum)
a931db9e
NB
1542{
1543 char path[100];
1544 char pid[10];
1545 int fd;
1546 int n;
10013317
PHC
1547 char *devname = devnum2devname(devnum);
1548
753cf905 1549 sprintf(path, "%s/%s.pid", MDMON_DIR, devname);
10013317
PHC
1550 free(devname);
1551
24f6f99b 1552 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1553
1554 if (fd < 0)
cf556303 1555 return -1;
a931db9e
NB
1556 n = read(fd, pid, 9);
1557 close(fd);
1558 if (n <= 0)
cf556303 1559 return -1;
24f6f99b 1560 return atoi(pid);
a931db9e
NB
1561}
1562
24f6f99b 1563int mdmon_running(int devnum)
a931db9e 1564{
24f6f99b
N
1565 int pid = mdmon_pid(devnum);
1566 if (pid <= 0)
a931db9e 1567 return 0;
24f6f99b 1568 if (kill(pid, 0) == 0)
a931db9e
NB
1569 return 1;
1570 return 0;
1571}
1572
8850ee3e
N
1573int start_mdmon(int devnum)
1574{
1575 int i;
44d2e365 1576 int len;
9fe32043
N
1577 pid_t pid;
1578 int status;
44d2e365
N
1579 char pathbuf[1024];
1580 char *paths[4] = {
1581 pathbuf,
1582 "/sbin/mdmon",
1583 "mdmon",
1584 NULL
1585 };
8850ee3e 1586
40ebbb9c 1587 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1588 return 0;
1589
44d2e365
N
1590 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf));
1591 if (len > 0) {
1592 char *sl;
1593 pathbuf[len] = 0;
1594 sl = strrchr(pathbuf, '/');
1595 if (sl)
1596 sl++;
1597 else
1598 sl = pathbuf;
1599 strcpy(sl, "mdmon");
1600 } else
1601 pathbuf[0] = '\0';
1602
8850ee3e
N
1603 switch(fork()) {
1604 case 0:
1605 /* FIXME yuk. CLOSE_EXEC?? */
1606 for (i=3; i < 100; i++)
1607 close(i);
44d2e365
N
1608 for (i=0; paths[i]; i++)
1609 if (paths[i][0])
1610 execl(paths[i], "mdmon",
e8a70c89
N
1611 devnum2devname(devnum),
1612 NULL);
8850ee3e
N
1613 exit(1);
1614 case -1: fprintf(stderr, Name ": cannot run mdmon. "
1615 "Array remains readonly\n");
1616 return -1;
9fe32043
N
1617 default: /* parent - good */
1618 pid = wait(&status);
1619 if (pid < 0 || status != 0)
1620 return -1;
8850ee3e
N
1621 }
1622 return 0;
1623}
1624
40ebbb9c 1625int check_env(char *name)
5dcfcb71 1626{
40ebbb9c 1627 char *val = getenv(name);
5dcfcb71
DW
1628
1629 if (val && atoi(val) == 1)
1630 return 1;
1631
1632 return 0;
1633}
1634
148acb7b
DW
1635__u32 random32(void)
1636{
1637 __u32 rv;
1638 int rfd = open("/dev/urandom", O_RDONLY);
1639 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1640 rv = random();
1641 if (rfd >= 0)
1642 close(rfd);
1643 return rv;
1644}
1645
0e600426 1646#ifndef MDASSEMBLE
edd8d13c
NB
1647int flush_metadata_updates(struct supertype *st)
1648{
1649 int sfd;
1650 if (!st->updates) {
1651 st->update_tail = NULL;
1652 return -1;
1653 }
1654
1655 sfd = connect_monitor(devnum2devname(st->container_dev));
1656 if (sfd < 0)
1657 return -1;
1658
1659 while (st->updates) {
1660 struct metadata_update *mu = st->updates;
1661 st->updates = mu->next;
1662
1663 send_message(sfd, mu, 0);
1664 wait_reply(sfd, 0);
1665 free(mu->buf);
1666 free(mu);
1667 }
1668 ack(sfd, 0);
1669 wait_reply(sfd, 0);
1670 close(sfd);
1671 st->update_tail = NULL;
1672 return 0;
1673}
1674
1675void append_metadata_update(struct supertype *st, void *buf, int len)
1676{
1677
1678 struct metadata_update *mu = malloc(sizeof(*mu));
1679
1680 mu->buf = buf;
1681 mu->len = len;
1682 mu->space = NULL;
cb23f1f4 1683 mu->space_list = NULL;
edd8d13c
NB
1684 mu->next = NULL;
1685 *st->update_tail = mu;
1686 st->update_tail = &mu->next;
1687}
0e600426 1688#endif /* MDASSEMBLE */
a931db9e 1689
fe6729fa
NB
1690#ifdef __TINYC__
1691/* tinyc doesn't optimize this check in ioctl.h out ... */
1692unsigned int __invalid_size_argument_for_IOC = 0;
1693#endif
1694
6d11ec6f
AK
1695int experimental(void)
1696{
1697 if (check_env("MDADM_EXPERIMENTAL"))
1698 return 1;
1699 else {
1700 fprintf(stderr, Name ": To use this feature MDADM_EXPERIMENTAL enviroment variable has to defined.\n");
1701 return 0;
1702 }
1703}
1704
326727d9
AC
1705/* Pick all spares matching given criteria from a container
1706 * if min_size == 0 do not check size
1707 * if domlist == NULL do not check domains
1708 * if spare_group given add it to domains of each spare
1709 * metadata allows to test domains using metadata of destination array */
1710struct mdinfo *container_choose_spares(struct supertype *st,
1711 unsigned long long min_size,
1712 struct domainlist *domlist,
1713 char *spare_group,
1714 const char *metadata, int get_one)
1715{
1716 struct mdinfo *d, **dp, *disks = NULL;
1717
1718 /* get list of all disks in container */
1719 if (st->ss->getinfo_super_disks)
1720 disks = st->ss->getinfo_super_disks(st);
1721
1722 if (!disks)
1723 return disks;
1724 /* find spare devices on the list */
1725 dp = &disks->devs;
1726 disks->array.spare_disks = 0;
1727 while (*dp) {
1728 int found = 0;
1729 d = *dp;
1730 if (d->disk.state == 0) {
1731 /* check if size is acceptable */
1732 unsigned long long dev_size;
1733 dev_t dev = makedev(d->disk.major,d->disk.minor);
1734
1735 if (!min_size ||
1736 (dev_size_from_id(dev, &dev_size) &&
1737 dev_size >= min_size))
1738 found = 1;
1739 /* check if domain matches */
1740 if (found && domlist) {
1741 struct dev_policy *pol = devnum_policy(dev);
1742 if (spare_group)
1743 pol_add(&pol, pol_domain,
1744 spare_group, NULL);
e5508b36 1745 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1746 found = 0;
1747 dev_policy_free(pol);
1748 }
1749 }
1750 if (found) {
1751 dp = &d->next;
1752 disks->array.spare_disks++;
1753 if (get_one) {
1754 sysfs_free(*dp);
1755 d->next = NULL;
1756 }
1757 } else {
1758 *dp = d->next;
1759 d->next = NULL;
1760 sysfs_free(d);
1761 }
1762 }
1763 return disks;
1764}