Add --offroot argument to mdmon
[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;
f161d047 149 int a = 0, b = 0,c = 0;
64c4757e
NB
150 if (uname(&name) <0)
151 return -1;
152
98c6faba
NB
153 cp = name.release;
154 a = strtoul(cp, &cp, 10);
f161d047
N
155 if (*cp == '.')
156 b = strtoul(cp+1, &cp, 10);
157 if (*cp == '.')
158 c = strtoul(cp+1, &cp, 10);
98c6faba 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;
aabe020d 366 int i, rv;
8453e704
N
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 */
aabe020d
JS
389 rv = enough(array.level, array.raid_disks, array.layout,
390 1, avail, avail_disks);
391 free(avail);
392 return rv;
8453e704
N
393}
394
395
22e263f6
AC
396const int uuid_zero[4] = { 0, 0, 0, 0 };
397
f277ce36 398int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 399{
f277ce36
NB
400 if (swapuuid) {
401 /* parse uuids are hostendian.
402 * uuid's from some superblocks are big-ending
aba69144 403 * if there is a difference, we need to swap..
f277ce36
NB
404 */
405 unsigned char *ac = (unsigned char *)a;
406 unsigned char *bc = (unsigned char *)b;
407 int i;
408 for (i=0; i<16; i+= 4) {
409 if (ac[i+0] != bc[i+3] ||
410 ac[i+1] != bc[i+2] ||
411 ac[i+2] != bc[i+1] ||
412 ac[i+3] != bc[i+0])
413 return 0;
414 }
415 return 1;
416 } else {
417 if (a[0]==b[0] &&
418 a[1]==b[1] &&
419 a[2]==b[2] &&
420 a[3]==b[3])
421 return 1;
422 return 0;
423 }
64c4757e 424}
350f29f9
NB
425void copy_uuid(void *a, int b[4], int swapuuid)
426{
427 if (swapuuid) {
428 /* parse uuids are hostendian.
429 * uuid's from some superblocks are big-ending
430 * if there is a difference, we need to swap..
431 */
432 unsigned char *ac = (unsigned char *)a;
433 unsigned char *bc = (unsigned char *)b;
434 int i;
435 for (i=0; i<16; i+= 4) {
436 ac[i+0] = bc[i+3];
437 ac[i+1] = bc[i+2];
438 ac[i+2] = bc[i+1];
439 ac[i+3] = bc[i+0];
440 }
441 } else
442 memcpy(a, b, 16);
443}
64c4757e 444
aae5a112 445char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 446{
9968e376 447 int i, j;
d7288ddc
N
448 char uuid[16];
449 char *c = buf;
450 strcpy(c, "UUID-");
451 c += strlen(c);
aae5a112 452 copy_uuid(uuid, id, swap);
9968e376 453 for (i = 0; i < 4; i++) {
9968e376 454 if (i)
ff54de6e 455 *c++ = sep;
9968e376
DW
456 for (j = 3; j >= 0; j--) {
457 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
458 c+= 2;
459 }
d7288ddc
N
460 }
461 return buf;
aae5a112
DW
462
463}
464
465char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
466{
86983cce
N
467 // dirty hack to work around an issue with super1 superblocks...
468 // super1 superblocks need swapuuid set in order for assembly to
469 // work, but can't have it set if we want this printout to match
470 // all the other uuid printouts in super1.c, so we force swapuuid
471 // to 1 to make our printout match the rest of super1
472 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
473}
474
435d4ebb 475#ifndef MDASSEMBLE
682c7051
NB
476int check_ext2(int fd, char *name)
477{
478 /*
479 * Check for an ext2fs file system.
480 * Superblock is always 1K at 1K offset
481 *
482 * s_magic is le16 at 56 == 0xEF53
483 * report mtime - le32 at 44
484 * blocks - le32 at 4
485 * logblksize - le32 at 24
486 */
487 unsigned char sb[1024];
488 time_t mtime;
489 int size, bsize;
490 if (lseek(fd, 1024,0)!= 1024)
491 return 0;
492 if (read(fd, sb, 1024)!= 1024)
493 return 0;
494 if (sb[56] != 0x53 || sb[57] != 0xef)
495 return 0;
496
497 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
498 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
499 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
500 fprintf(stderr, Name ": %s appears to contain an ext2fs file system\n",
501 name);
502 fprintf(stderr," size=%dK mtime=%s",
503 size*(1<<bsize), ctime(&mtime));
504 return 1;
505}
506
507int check_reiser(int fd, char *name)
508{
509 /*
510 * superblock is at 64K
511 * size is 1024;
512 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
513 *
514 */
515 unsigned char sb[1024];
881990a2 516 unsigned long size;
682c7051
NB
517 if (lseek(fd, 64*1024, 0) != 64*1024)
518 return 0;
519 if (read(fd, sb, 1024) != 1024)
520 return 0;
a46f4061
NB
521 if (strncmp((char*)sb+52, "ReIsErFs",8)!=0 &&
522 strncmp((char*)sb+52, "ReIsEr2Fs",9)!=0)
682c7051
NB
523 return 0;
524 fprintf(stderr, Name ": %s appears to contain a reiserfs file system\n",name);
525 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
881990a2 526 fprintf(stderr, " size = %luK\n", size*4);
aba69144 527
682c7051
NB
528 return 1;
529}
530
531int check_raid(int fd, char *name)
532{
4b1ac34b 533 struct mdinfo info;
682c7051 534 time_t crtime;
d078d77c 535 char *level;
82d9eba6 536 struct supertype *st = guess_super(fd);
f9ce90ba 537
82d9eba6 538 if (!st) return 0;
4a96d9ff 539 st->ignore_hw_compat = 1;
3da92f27 540 st->ss->load_super(st, fd, name);
82d9eba6
NB
541 /* Looks like a raid array .. */
542 fprintf(stderr, Name ": %s appears to be part of a raid array:\n",
543 name);
a5d85af7 544 st->ss->getinfo_super(st, &info, NULL);
3da92f27 545 st->ss->free_super(st);
82d9eba6 546 crtime = info.array.ctime;
d078d77c
NB
547 level = map_num(pers, info.array.level);
548 if (!level) level = "-unknown-";
549 fprintf(stderr, " level=%s devices=%d ctime=%s",
550 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 551 return 1;
682c7051
NB
552}
553
682c7051
NB
554int ask(char *mesg)
555{
556 char *add = "";
557 int i;
558 for (i=0; i<5; i++) {
559 char buf[100];
560 fprintf(stderr, "%s%s", mesg, add);
561 fflush(stderr);
562 if (fgets(buf, 100, stdin)==NULL)
563 return 0;
564 if (buf[0]=='y' || buf[0]=='Y')
565 return 1;
566 if (buf[0]=='n' || buf[0]=='N')
567 return 0;
568 add = "(y/n) ";
569 }
570 fprintf(stderr, Name ": assuming 'no'\n");
571 return 0;
572}
435d4ebb 573#endif /* MDASSEMBLE */
682c7051 574
8d80900b 575int is_standard(char *dev, int *nump)
e5329c37
NB
576{
577 /* tests if dev is a "standard" md dev name.
578 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 579 * where NN is a string of digits
598f0d58
NB
580 * Returns 1 if a partitionable standard,
581 * -1 if non-partitonable,
582 * 0 if not a standard name.
e5329c37 583 */
8d80900b
NB
584 char *d = strrchr(dev, '/');
585 int type=0;
586 int num;
587 if (!d)
e5329c37 588 return 0;
8d80900b
NB
589 if (strncmp(d, "/d",2)==0)
590 d += 2, type=1; /* /dev/md/dN{pM} */
591 else if (strncmp(d, "/md_d", 5)==0)
2b4ca8f0 592 d += 5, type=1; /* /dev/md_dN{pM} */
8d80900b
NB
593 else if (strncmp(d, "/md", 3)==0)
594 d += 3, type=-1; /* /dev/mdN */
595 else if (d-dev > 3 && strncmp(d-2, "md/", 3)==0)
5a6d1148 596 d += 1, type=-1; /* /dev/md/N */
e5329c37
NB
597 else
598 return 0;
8d80900b 599 if (!*d)
e5329c37 600 return 0;
8d80900b
NB
601 num = atoi(d);
602 while (isdigit(*d))
603 d++;
604 if (*d)
e5329c37 605 return 0;
8d80900b
NB
606 if (nump) *nump = num;
607
608 return type;
e5329c37
NB
609}
610
4b1ac34b 611unsigned long calc_csum(void *super, int bytes)
82b27616 612{
56eb10c0 613 unsigned long long newcsum = 0;
82b27616 614 int i;
4b1ac34b
NB
615 unsigned int csum;
616 unsigned int *superc = (unsigned int*) super;
82b27616 617
4b1ac34b 618 for(i=0; i<bytes/4; i++)
82b27616
NB
619 newcsum+= superc[i];
620 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 621#ifdef __alpha__
aba69144 622/* The in-kernel checksum calculation is always 16bit on
570c0542
NB
623 * the alpha, though it is 32 bit on i386...
624 * I wonder what it is elsewhere... (it uses and API in
625 * a way that it shouldn't).
626 */
627 csum = (csum & 0xffff) + (csum >> 16);
628 csum = (csum & 0xffff) + (csum >> 16);
629#endif
82b27616
NB
630 return csum;
631}
cd29a5c8 632
435d4ebb 633#ifndef MDASSEMBLE
56eb10c0 634char *human_size(long long bytes)
cd29a5c8
NB
635{
636 static char buf[30];
d5d3721e
NB
637
638 /* We convert bytes to either centi-M{ega,ibi}bytes or
639 * centi-G{igi,ibi}bytes, with appropriate rounding,
640 * and then print 1/100th of those as a decimal.
641 * We allow upto 2048Megabytes before converting to
642 * gigabytes, as that shows more precision and isn't
643 * too large a number.
01619b48 644 * Terabytes are not yet handled.
d5d3721e 645 */
cd29a5c8 646
56eb10c0 647 if (bytes < 5000*1024)
cd29a5c8 648 buf[0]=0;
d5d3721e
NB
649 else if (bytes < 2*1024LL*1024LL*1024LL) {
650 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
651 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 652 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
653 cMiB/100 , cMiB % 100,
654 cMB/100, cMB % 100);
655 } else {
656 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
657 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 658 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
659 cGiB/100 , cGiB % 100,
660 cGB/100, cGB % 100);
661 }
cd29a5c8
NB
662 return buf;
663}
e0d19036
NB
664
665char *human_size_brief(long long bytes)
666{
667 static char buf[30];
e0d19036
NB
668
669 if (bytes < 5000*1024)
8f23b0b3 670 snprintf(buf, sizeof(buf), "%ld.%02ldKiB",
bd526cee 671 (long)(bytes>>10), (long)(((bytes&1023)*100+512)/1024)
e0d19036
NB
672 );
673 else if (bytes < 2*1024LL*1024LL*1024LL)
8f23b0b3 674 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
e0d19036 675 (long)(bytes>>20),
bd526cee 676 (long)((bytes&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
677 );
678 else
8f23b0b3 679 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
e0d19036 680 (long)(bytes>>30),
bd526cee 681 (long)(((bytes>>10)&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
682 );
683 return buf;
684}
e4965ef8
N
685
686void print_r10_layout(int layout)
687{
688 int near = layout & 255;
689 int far = (layout >> 8) & 255;
690 int offset = (layout&0x10000);
691 char *sep = "";
692
693 if (near != 1) {
694 printf("%s near=%d", sep, near);
695 sep = ",";
696 }
697 if (far != 1)
698 printf("%s %s=%d", sep, offset?"offset":"far", far);
699 if (near*far == 1)
700 printf("NO REDUNDANCY");
701}
435d4ebb 702#endif
e0d19036 703
5f8097be
NB
704unsigned long long calc_array_size(int level, int raid_disks, int layout,
705 int chunksize, unsigned long long devsize)
577e8448
AK
706{
707 devsize &= ~(unsigned long long)((chunksize>>9)-1);
708 return get_data_disks(level, layout, raid_disks) * devsize;
709}
710
711int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
712{
713 int data_disks = 0;
714 switch (level) {
715 case 0: data_disks = raid_disks; break;
716 case 1: data_disks = 1; break;
717 case 4:
718 case 5: data_disks = raid_disks - 1; break;
719 case 6: data_disks = raid_disks - 2; break;
720 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
721 break;
722 }
577e8448
AK
723
724 return data_disks;
5f8097be
NB
725}
726
0e600426 727#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
e0d19036
NB
728char *get_md_name(int dev)
729{
730 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 731 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
e0d19036
NB
732 static char devname[50];
733 struct stat stb;
98c6faba 734 dev_t rdev;
dd0781e5 735 char *dn;
98c6faba
NB
736
737 if (dev < 0) {
dd0781e5
NB
738 int mdp = get_mdp_major();
739 if (mdp < 0) return NULL;
0df46c2a 740 rdev = makedev(mdp, (-1-dev)<<6);
8f23b0b3 741 snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev);
98c6faba
NB
742 if (stat(devname, &stb) == 0
743 && (S_IFMT&stb.st_mode) == S_IFBLK
744 && (stb.st_rdev == rdev))
745 return devname;
746 } else {
0df46c2a 747 rdev = makedev(MD_MAJOR, dev);
8f23b0b3 748 snprintf(devname, sizeof(devname), "/dev/md%d", dev);
98c6faba
NB
749 if (stat(devname, &stb) == 0
750 && (S_IFMT&stb.st_mode) == S_IFBLK
751 && (stb.st_rdev == rdev))
752 return devname;
753
8f23b0b3 754 snprintf(devname, sizeof(devname), "/dev/md/%d", dev);
98c6faba
NB
755 if (stat(devname, &stb) == 0
756 && (S_IFMT&stb.st_mode) == S_IFBLK
757 && (stb.st_rdev == rdev))
758 return devname;
759 }
16c6fa80 760 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
761 if (dn)
762 return dn;
8f23b0b3 763 snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev);
e0d19036 764 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
765 if (errno != EEXIST)
766 return NULL;
e0d19036
NB
767
768 if (stat(devname, &stb) == 0
769 && (S_IFMT&stb.st_mode) == S_IFBLK
770 && (stb.st_rdev == rdev))
771 return devname;
772 unlink(devname);
773 return NULL;
774}
775
776void put_md_name(char *name)
777{
778 if (strncmp(name, "/dev/.tmp.md", 12)==0)
779 unlink(name);
780}
ea24acd0 781
ea24acd0
NB
782int find_free_devnum(int use_partitions)
783{
784 int devnum;
785 for (devnum = 127; devnum != 128;
a56fb7ec 786 devnum = devnum ? devnum-1 : (1<<20)-1) {
ea24acd0
NB
787 char *dn;
788 int _devnum;
789
790 _devnum = use_partitions ? (-1-devnum) : devnum;
791 if (mddev_busy(_devnum))
792 continue;
793 /* make sure it is new to /dev too, at least as a
794 * non-standard */
795 dn = map_dev(dev2major(_devnum), dev2minor(_devnum), 0);
796 if (dn && ! is_standard(dn, NULL))
797 continue;
798 break;
799 }
800 if (devnum == 128)
801 return NoMdDev;
802 return use_partitions ? (-1-devnum) : devnum;
803}
435d4ebb 804#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 805
8b0dabea
NB
806int dev_open(char *dev, int flags)
807{
808 /* like 'open', but if 'dev' matches %d:%d, create a temp
809 * block device and open that
810 */
811 char *e;
812 int fd = -1;
813 char devname[32];
e81cdd9f 814 int major;
8b0dabea 815 int minor;
e81cdd9f
NB
816
817 if (!dev) return -1;
6df6a774 818 flags |= O_DIRECT;
e81cdd9f
NB
819
820 major = strtoul(dev, &e, 0);
8b0dabea
NB
821 if (e > dev && *e == ':' && e[1] &&
822 (minor = strtoul(e+1, &e, 0)) >= 0 &&
823 *e == 0) {
6df6a774
N
824 char *path = map_dev(major, minor, 0);
825 if (path)
826 fd = open(path, flags);
827 if (fd < 0) {
828 snprintf(devname, sizeof(devname), "/dev/.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 }
834 }
835 if (fd < 0) {
836 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
837 (int)getpid(), major, minor);
838 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
839 fd = open(devname, flags);
840 unlink(devname);
841 }
8b0dabea
NB
842 }
843 } else
6df6a774 844 fd = open(dev, flags);
8b0dabea
NB
845 return fd;
846}
f9ce90ba 847
d998b738 848int open_dev_flags(int devnum, int flags)
e8a70c89
N
849{
850 char buf[20];
851
852 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
d998b738
N
853 return dev_open(buf, flags);
854}
855
856int open_dev(int devnum)
857{
858 return open_dev_flags(devnum, O_RDONLY);
e8a70c89
N
859}
860
a931db9e
NB
861int open_dev_excl(int devnum)
862{
863 char buf[20];
864 int i;
7187750e 865 int flags = O_RDWR;
a931db9e
NB
866
867 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
868 for (i=0 ; i<25 ; i++) {
7187750e 869 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
870 if (fd >= 0)
871 return fd;
7187750e
N
872 if (errno == EACCES && flags == O_RDWR) {
873 flags = O_RDONLY;
874 continue;
875 }
a931db9e
NB
876 if (errno != EBUSY)
877 return fd;
878 usleep(200000);
879 }
880 return -1;
881}
882
9008ed1c
N
883int same_dev(char *one, char *two)
884{
885 struct stat st1, st2;
886 if (stat(one, &st1) != 0)
887 return 0;
888 if (stat(two, &st2) != 0)
889 return 0;
890 if ((st1.st_mode & S_IFMT) != S_IFBLK)
891 return 0;
892 if ((st2.st_mode & S_IFMT) != S_IFBLK)
893 return 0;
894 return st1.st_rdev == st2.st_rdev;
895}
896
a7c6e3fb 897void wait_for(char *dev, int fd)
a714580e
N
898{
899 int i;
a7c6e3fb
N
900 struct stat stb_want;
901
902 if (fstat(fd, &stb_want) != 0 ||
903 (stb_want.st_mode & S_IFMT) != S_IFBLK)
904 return;
a714580e
N
905
906 for (i=0 ; i<25 ; i++) {
907 struct stat stb;
a7c6e3fb
N
908 if (stat(dev, &stb) == 0 &&
909 (stb.st_mode & S_IFMT) == S_IFBLK &&
910 (stb.st_rdev == stb_want.st_rdev))
a714580e
N
911 return;
912 usleep(200000);
913 }
436305c6
DW
914 if (i == 25)
915 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
916}
917
0f22b998
N
918struct superswitch *superlist[] =
919{
920 &super0, &super1,
921 &super_ddf, &super_imsm,
0592faeb 922 &mbr, &gpt,
0f22b998 923 NULL };
f9ce90ba 924
ea24acd0 925#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 926
4725bc31 927struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 928{
1686dc25
NB
929 mdu_array_info_t array;
930 int vers;
931 int minor;
932 struct supertype *st = NULL;
7e0f6979 933 struct mdinfo *sra;
142cb9e1 934 char *verstr;
1686dc25
NB
935 char version[20];
936 int i;
f7e7067b 937 char *subarray = NULL;
5f7e44b2 938 int container = NoMdDev;
1686dc25
NB
939
940 sra = sysfs_read(fd, 0, GET_VERSION);
941
942 if (sra) {
7e0f6979
NB
943 vers = sra->array.major_version;
944 minor = sra->array.minor_version;
142cb9e1 945 verstr = sra->text_version;
1686dc25
NB
946 } else {
947 if (ioctl(fd, GET_ARRAY_INFO, &array))
948 array.major_version = array.minor_version = 0;
949 vers = array.major_version;
950 minor = array.minor_version;
142cb9e1 951 verstr = "";
6fbba4c9 952 }
82d9eba6 953
1686dc25
NB
954 if (vers != -1) {
955 sprintf(version, "%d.%d", vers, minor);
956 verstr = version;
6fbba4c9 957 }
3c558363 958 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 959 char *dev = verstr+1;
5f7e44b2 960
f7e7067b 961 subarray = strchr(dev, '/');
f7e7067b
NB
962 if (subarray)
963 *subarray++ = '\0';
f7e7067b 964 subarray = strdup(subarray);
5f7e44b2 965 container = devname2devnum(dev);
f7e7067b
NB
966 if (sra)
967 sysfs_free(sra);
5f7e44b2 968 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
969 if (sra && sra->text_version[0])
970 verstr = sra->text_version;
971 else
972 verstr = "-no-metadata-";
f7e7067b
NB
973 }
974
975 for (i = 0; st == NULL && superlist[i] ; i++)
976 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
977
978 if (sra)
979 sysfs_free(sra);
f7e7067b 980 if (st) {
3b0896f8 981 st->sb = NULL;
1f49fb3a
N
982 if (subarrayp)
983 *subarrayp = subarray;
5f7e44b2
DW
984 st->container_dev = container;
985 st->devnum = fd2devnum(fd);
4725bc31
N
986 } else
987 free(subarray);
5f7e44b2 988
82d9eba6 989 return st;
f9ce90ba 990}
ea24acd0
NB
991#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
992
0f0749ad 993int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
994{
995 char buf[20];
996 int fd;
997
998 sprintf(buf, "%d:%d", major(id), minor(id));
999 fd = dev_open(buf, O_RDONLY);
1000 if (fd < 0)
1001 return 0;
1002 if (get_dev_size(fd, NULL, size)) {
1003 close(fd);
1004 return 1;
1005 }
1006 close(fd);
1007 return 0;
1008}
f9ce90ba 1009
159c3a1a 1010struct supertype *dup_super(struct supertype *orig)
3da92f27 1011{
159c3a1a 1012 struct supertype *st;
1686dc25 1013
d2ca6449
NB
1014 if (!orig)
1015 return orig;
159c3a1a 1016 st = malloc(sizeof(*st));
3da92f27
NB
1017 if (!st)
1018 return st;
ef609477 1019 memset(st, 0, sizeof(*st));
159c3a1a
NB
1020 st->ss = orig->ss;
1021 st->max_devs = orig->max_devs;
1022 st->minor_version = orig->minor_version;
1023 st->sb = NULL;
1024 st->info = NULL;
1025 return st;
3da92f27
NB
1026}
1027
54887ad8 1028struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1029{
1030 /* try each load_super to find the best match,
1031 * and return the best superswitch
1032 */
82d9eba6
NB
1033 struct superswitch *ss;
1034 struct supertype *st;
f21e18ca 1035 time_t besttime = 0;
570c0542 1036 int bestsuper = -1;
f9ce90ba
NB
1037 int i;
1038
82d9eba6 1039 st = malloc(sizeof(*st));
d1d599ea
N
1040 memset(st, 0, sizeof(*st));
1041 st->container_dev = NoMdDev;
1042
f9ce90ba
NB
1043 for (i=0 ; superlist[i]; i++) {
1044 int rv;
1045 ss = superlist[i];
54887ad8
N
1046 if (guess_type == guess_array && ss->add_to_super == NULL)
1047 continue;
1048 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1049 continue;
ef609477 1050 memset(st, 0, sizeof(*st));
df3346e6 1051 st->ignore_hw_compat = 1;
3da92f27 1052 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1053 if (rv == 0) {
1054 struct mdinfo info;
a5d85af7 1055 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1056 if (bestsuper == -1 ||
1057 besttime < info.array.ctime) {
1058 bestsuper = i;
1059 besttime = info.array.ctime;
570c0542 1060 }
3da92f27 1061 ss->free_super(st);
570c0542
NB
1062 }
1063 }
1064 if (bestsuper != -1) {
1065 int rv;
ef609477 1066 memset(st, 0, sizeof(*st));
df3346e6 1067 st->ignore_hw_compat = 1;
3da92f27 1068 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1069 if (rv == 0) {
5e747af2 1070 superlist[bestsuper]->free_super(st);
df3346e6 1071 st->ignore_hw_compat = 0;
82d9eba6 1072 return st;
f9ce90ba
NB
1073 }
1074 }
570c0542 1075 free(st);
f9ce90ba
NB
1076 return NULL;
1077}
fe6729fa 1078
beae1dfe
NB
1079/* Return size of device in bytes */
1080int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1081{
1082 unsigned long long ldsize;
c2c9bb6f
NB
1083 struct stat st;
1084
1085 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1086 ldsize = (unsigned long long)st.st_size;
1087 else
beae1dfe
NB
1088#ifdef BLKGETSIZE64
1089 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1090#endif
1091 {
1092 unsigned long dsize;
1093 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1094 ldsize = dsize;
1095 ldsize <<= 9;
1096 } else {
1097 if (dname)
1098 fprintf(stderr, Name ": Cannot get size of %s: %s\b",
1099 dname, strerror(errno));
1100 return 0;
1101 }
1102 }
1103 *sizep = ldsize;
1104 return 1;
1105}
8fac0577 1106
3a371610
N
1107/* Return true if this can only be a container, not a member device.
1108 * i.e. is and md device and size is zero
1109 */
1110int must_be_container(int fd)
1111{
1112 unsigned long long size;
1113 if (md_get_version(fd) < 0)
1114 return 0;
1115 if (get_dev_size(fd, NULL, &size) == 0)
1116 return 1;
1117 if (size == 0)
1118 return 1;
1119 return 0;
1120}
034b203a
TM
1121
1122/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1123 * Returns: 1 if successful
1124 * -1 for unknown partition type
1125 * 0 for other errors
1126 */
1127static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1128{
056b331e 1129 struct GPT gpt;
034b203a 1130 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1131 struct GPT_part_entry *part;
1132 char buf[512];
034b203a
TM
1133 unsigned long long curr_part_end;
1134 unsigned all_partitions, entry_size;
f21e18ca 1135 unsigned part_nr;
034b203a
TM
1136
1137 *endofpart = 0;
1138
056b331e 1139 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1140 /* skip protective MBR */
034b203a 1141 lseek(fd, 512, SEEK_SET);
73e658d8 1142 /* read GPT header */
056b331e 1143 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1144 return 0;
1145
1146 /* get the number of partition entries and the entry size */
056b331e
N
1147 all_partitions = __le32_to_cpu(gpt.part_cnt);
1148 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1149
1150 /* Check GPT signature*/
056b331e 1151 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1152 return -1;
1153
1154 /* sanity checks */
1155 if (all_partitions > 1024 ||
db7fdfe4 1156 entry_size > sizeof(buf))
034b203a
TM
1157 return -1;
1158
db7fdfe4
JS
1159 part = (struct GPT_part_entry *)buf;
1160
034b203a 1161 for (part_nr=0; part_nr < all_partitions; part_nr++) {
73e658d8 1162 /* read partition entry */
db7fdfe4 1163 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1164 return 0;
1165
034b203a 1166 /* is this valid partition? */
db7fdfe4 1167 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1168 /* check the last lba for the current partition */
db7fdfe4 1169 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1170 if (curr_part_end > *endofpart)
1171 *endofpart = curr_part_end;
1172 }
1173
034b203a
TM
1174 }
1175 return 1;
1176}
1177
1178/* Sets endofpart parameter to the last block used by the last partition on the device.
1179 * Returns: 1 if successful
1180 * -1 for unknown partition type
1181 * 0 for other errors
1182 */
1183static int get_last_partition_end(int fd, unsigned long long *endofpart)
1184{
056b331e 1185 struct MBR boot_sect;
034b203a
TM
1186 struct MBR_part_record *part;
1187 unsigned long long curr_part_end;
f21e18ca 1188 unsigned part_nr;
034b203a
TM
1189 int retval = 0;
1190
1191 *endofpart = 0;
1192
056b331e 1193 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1194 /* read MBR */
1195 lseek(fd, 0, 0);
056b331e 1196 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1197 goto abort;
1198
1199 /* check MBP signature */
056b331e 1200 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1201 retval = 1;
1202 /* found the correct signature */
056b331e 1203 part = boot_sect.parts;
034b203a
TM
1204
1205 for (part_nr=0; part_nr < MBR_PARTITIONS; part_nr++) {
1206 /* check for GPT type */
1207 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1208 retval = get_gpt_last_partition_end(fd, endofpart);
1209 break;
1210 }
1211 /* check the last used lba for the current partition */
1212 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1213 __le32_to_cpu(part->blocks_num);
1214 if (curr_part_end > *endofpart)
1215 *endofpart = curr_part_end;
1216
1217 part++;
1218 }
1219 } else {
1220 /* Unknown partition table */
1221 retval = -1;
1222 }
1223 abort:
1224 return retval;
1225}
1226
53ed6ac3
KW
1227int check_partitions(int fd, char *dname, unsigned long long freesize,
1228 unsigned long long size)
034b203a
TM
1229{
1230 /*
1231 * Check where the last partition ends
1232 */
1233 unsigned long long endofpart;
1234 int ret;
1235
1236 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1237 /* There appears to be a partition table here */
1238 if (freesize == 0) {
1239 /* partitions will not be visible in new device */
1240 fprintf(stderr,
1241 Name ": partition table exists on %s but will be lost or\n"
1242 " meaningless after creating array\n",
1243 dname);
1244 return 1;
1245 } else if (endofpart > freesize) {
1246 /* last partition overlaps metadata */
1247 fprintf(stderr,
1248 Name ": metadata will over-write last partition on %s.\n",
1249 dname);
1250 return 1;
53ed6ac3
KW
1251 } else if (size && endofpart > size) {
1252 /* partitions will be truncated in new device */
1253 fprintf(stderr,
1254 Name ": array size is too small to cover all partitions on %s.\n",
1255 dname);
1256 return 1;
034b203a
TM
1257 }
1258 }
1259 return 0;
1260}
1261
8382f19b
NB
1262void get_one_disk(int mdfd, mdu_array_info_t *ainf, mdu_disk_info_t *disk)
1263{
1264 int d;
9e6d9291 1265
8382f19b 1266 ioctl(mdfd, GET_ARRAY_INFO, ainf);
9e6d9291
N
1267 for (d = 0 ; d < 1024 ; d++) {
1268 if (ioctl(mdfd, GET_DISK_INFO, disk) == 0 &&
1269 (disk->major || disk->minor))
8382f19b 1270 return;
9e6d9291 1271 }
8382f19b 1272}
63152c1b 1273
a322f70c
DW
1274int open_container(int fd)
1275{
1276 /* 'fd' is a block device. Find out if it is in use
1277 * by a container, and return an open fd on that container.
1278 */
1279 char path[256];
1280 char *e;
1281 DIR *dir;
1282 struct dirent *de;
1283 int dfd, n;
1284 char buf[200];
1285 int major, minor;
1286 struct stat st;
1287
1288 if (fstat(fd, &st) != 0)
1289 return -1;
1290 sprintf(path, "/sys/dev/block/%d:%d/holders",
1291 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1292 e = path + strlen(path);
1293
1294 dir = opendir(path);
1295 if (!dir)
1296 return -1;
1297 while ((de = readdir(dir))) {
1298 if (de->d_ino == 0)
1299 continue;
1300 if (de->d_name[0] == '.')
1301 continue;
1302 sprintf(e, "/%s/dev", de->d_name);
1303 dfd = open(path, O_RDONLY);
1304 if (dfd < 0)
1305 continue;
1306 n = read(dfd, buf, sizeof(buf));
1307 close(dfd);
f21e18ca 1308 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1309 continue;
1310 buf[n] = 0;
1311 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1312 continue;
1313 sprintf(buf, "%d:%d", major, minor);
1314 dfd = dev_open(buf, O_RDONLY);
1315 if (dfd >= 0) {
1316 closedir(dir);
1317 return dfd;
1318 }
1319 }
355726fa 1320 closedir(dir);
a322f70c
DW
1321 return -1;
1322}
1323
33414a01
DW
1324struct superswitch *version_to_superswitch(char *vers)
1325{
1326 int i;
1327
1328 for (i = 0; superlist[i]; i++) {
1329 struct superswitch *ss = superlist[i];
1330
1331 if (strcmp(vers, ss->name) == 0)
1332 return ss;
1333 }
1334
1335 return NULL;
1336}
1337
1338int is_container_member(struct mdstat_ent *mdstat, char *container)
1339{
1340 if (mdstat->metadata_version == NULL ||
1341 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
1342 !is_subarray(mdstat->metadata_version+9) ||
1343 strncmp(mdstat->metadata_version+10, container, strlen(container)) != 0 ||
1344 mdstat->metadata_version[10+strlen(container)] != '/')
1345 return 0;
1346
1347 return 1;
1348}
1349
1350int is_subarray_active(char *subarray, char *container)
1351{
1352 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1353 struct mdstat_ent *ent;
1354
e5408a32
DW
1355 for (ent = mdstat; ent; ent = ent->next)
1356 if (is_container_member(ent, container))
e5e5d7ce 1357 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1358 break;
33414a01
DW
1359
1360 free_mdstat(mdstat);
1361
1362 return ent != NULL;
1363}
1364
1365/* open_subarray - opens a subarray in a container
1366 * @dev: container device name
feab51f8 1367 * @st: empty supertype
33414a01
DW
1368 * @quiet: block reporting errors flag
1369 *
1370 * On success returns an fd to a container and fills in *st
1371 */
feab51f8 1372int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1373{
1374 struct mdinfo *mdi;
a951a4f7 1375 struct mdinfo *info;
33414a01
DW
1376 int fd, err = 1;
1377
1378 fd = open(dev, O_RDWR|O_EXCL);
1379 if (fd < 0) {
1380 if (!quiet)
1381 fprintf(stderr, Name ": Couldn't open %s, aborting\n",
1382 dev);
b990032d 1383 return -1;
33414a01
DW
1384 }
1385
1386 st->devnum = fd2devnum(fd);
1387 if (st->devnum == NoMdDev) {
1388 if (!quiet)
1389 fprintf(stderr,
1390 Name ": Failed to determine device number for %s\n",
1391 dev);
1392 goto close_fd;
1393 }
1394
1395 mdi = sysfs_read(fd, st->devnum, GET_VERSION|GET_LEVEL);
1396 if (!mdi) {
1397 if (!quiet)
1398 fprintf(stderr, Name ": Failed to read sysfs for %s\n",
1399 dev);
1400 goto close_fd;
1401 }
1402
1403 if (mdi->array.level != UnSet) {
1404 if (!quiet)
1405 fprintf(stderr, Name ": %s is not a container\n", dev);
1406 goto free_sysfs;
1407 }
1408
1409 st->ss = version_to_superswitch(mdi->text_version);
1410 if (!st->ss) {
1411 if (!quiet)
1412 fprintf(stderr,
1413 Name ": Operation not supported for %s metadata\n",
1414 mdi->text_version);
1415 goto free_sysfs;
1416 }
1417
1418 st->devname = devnum2devname(st->devnum);
1419 if (!st->devname) {
1420 if (!quiet)
1421 fprintf(stderr, Name ": Failed to allocate device name\n");
1422 goto free_sysfs;
1423 }
1424
db20d413 1425 if (!st->ss->load_container) {
33414a01 1426 if (!quiet)
db20d413 1427 fprintf(stderr, Name ": %s is not a container\n", dev);
33414a01
DW
1428 goto free_name;
1429 }
1430
db20d413 1431 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1432 if (!quiet)
db20d413
N
1433 fprintf(stderr, Name ": Failed to load metadata for %s\n",
1434 dev);
1435 goto free_name;
33414a01
DW
1436 }
1437
a951a4f7
N
1438 info = st->ss->container_content(st, subarray);
1439 if (!info) {
1440 if (!quiet)
1441 fprintf(stderr, Name ": Failed to find subarray-%s in %s\n",
1442 subarray, dev);
1443 goto free_super;
1444 }
1445 free(info);
1446
33414a01
DW
1447 err = 0;
1448
1449 free_super:
1450 if (err)
1451 st->ss->free_super(st);
1452 free_name:
1453 if (err)
1454 free(st->devname);
1455 free_sysfs:
1456 sysfs_free(mdi);
1457 close_fd:
1458 if (err)
1459 close(fd);
1460
1461 if (err)
1462 return -1;
1463 else
1464 return fd;
1465}
1466
7801ac20
N
1467int add_disk(int mdfd, struct supertype *st,
1468 struct mdinfo *sra, struct mdinfo *info)
1469{
1470 /* Add a device to an array, in one of 2 ways. */
1471 int rv;
1472#ifndef MDASSEMBLE
1473 if (st->ss->external) {
d23534e4
DW
1474 if (info->disk.state & (1<<MD_DISK_SYNC))
1475 info->recovery_start = MaxSector;
1476 else
1477 info->recovery_start = 0;
2904b26f 1478 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1479 if (! rv) {
1480 struct mdinfo *sd2;
f35f2525
N
1481 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1482 if (sd2 == info)
1483 break;
1484 if (sd2 == NULL) {
1485 sd2 = malloc(sizeof(*sd2));
1486 *sd2 = *info;
1487 sd2->next = sra->devs;
1488 sra->devs = sd2;
1489 }
7801ac20
N
1490 }
1491 } else
1492#endif
1493 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1494 return rv;
1495}
1496
de6ae750
N
1497int remove_disk(int mdfd, struct supertype *st,
1498 struct mdinfo *sra, struct mdinfo *info)
1499{
1500 int rv;
1501 /* Remove the disk given by 'info' from the array */
1502#ifndef MDASSEMBLE
1503 if (st->ss->external)
1504 rv = sysfs_set_str(sra, info, "slot", "none");
1505 else
1506#endif
1507 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1508 info->disk.minor));
1509 return rv;
1510}
1511
f35f2525
N
1512int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1513{
1514 /* Initialise kernel's knowledge of array.
1515 * This varies between externally managed arrays
1516 * and older kernels
1517 */
1518 int vers = md_get_version(mdfd);
1519 int rv;
1520
1521#ifndef MDASSEMBLE
1522 if (st->ss->external)
1523 rv = sysfs_set_array(info, vers);
1524 else
1525#endif
1526 if ((vers % 100) >= 1) { /* can use different versions */
1527 mdu_array_info_t inf;
1528 memset(&inf, 0, sizeof(inf));
1529 inf.major_version = info->array.major_version;
1530 inf.minor_version = info->array.minor_version;
1531 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1532 } else
1533 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1534 return rv;
1535}
1536
1e5c6983
DW
1537unsigned long long min_recovery_start(struct mdinfo *array)
1538{
1539 /* find the minimum recovery_start in an array for metadata
1540 * formats that only record per-array recovery progress instead
1541 * of per-device
1542 */
1543 unsigned long long recovery_start = MaxSector;
1544 struct mdinfo *d;
1545
1546 for (d = array->devs; d; d = d->next)
1547 recovery_start = min(recovery_start, d->recovery_start);
1548
1549 return recovery_start;
1550}
1551
24f6f99b 1552int mdmon_pid(int devnum)
a931db9e
NB
1553{
1554 char path[100];
1555 char pid[10];
1556 int fd;
1557 int n;
10013317
PHC
1558 char *devname = devnum2devname(devnum);
1559
753cf905 1560 sprintf(path, "%s/%s.pid", MDMON_DIR, devname);
10013317
PHC
1561 free(devname);
1562
24f6f99b 1563 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1564
1565 if (fd < 0)
cf556303 1566 return -1;
a931db9e
NB
1567 n = read(fd, pid, 9);
1568 close(fd);
1569 if (n <= 0)
cf556303 1570 return -1;
24f6f99b 1571 return atoi(pid);
a931db9e
NB
1572}
1573
24f6f99b 1574int mdmon_running(int devnum)
a931db9e 1575{
24f6f99b
N
1576 int pid = mdmon_pid(devnum);
1577 if (pid <= 0)
a931db9e 1578 return 0;
24f6f99b 1579 if (kill(pid, 0) == 0)
a931db9e
NB
1580 return 1;
1581 return 0;
1582}
1583
8850ee3e
N
1584int start_mdmon(int devnum)
1585{
1913c325 1586 int i, skipped;
44d2e365 1587 int len;
9fe32043
N
1588 pid_t pid;
1589 int status;
44d2e365
N
1590 char pathbuf[1024];
1591 char *paths[4] = {
1592 pathbuf,
1593 "/sbin/mdmon",
1594 "mdmon",
1595 NULL
1596 };
8850ee3e 1597
40ebbb9c 1598 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1599 return 0;
1600
9cf014ec 1601 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1602 if (len > 0) {
1603 char *sl;
1604 pathbuf[len] = 0;
1605 sl = strrchr(pathbuf, '/');
1606 if (sl)
1607 sl++;
1608 else
1609 sl = pathbuf;
1610 strcpy(sl, "mdmon");
1611 } else
1612 pathbuf[0] = '\0';
1613
8850ee3e
N
1614 switch(fork()) {
1615 case 0:
1616 /* FIXME yuk. CLOSE_EXEC?? */
1913c325
N
1617 skipped = 0;
1618 for (i=3; skipped < 20; i++)
1619 if (close(i) < 0)
1620 skipped++;
1621 else
1622 skipped = 0;
1623
44d2e365
N
1624 for (i=0; paths[i]; i++)
1625 if (paths[i][0])
1626 execl(paths[i], "mdmon",
e8a70c89
N
1627 devnum2devname(devnum),
1628 NULL);
8850ee3e
N
1629 exit(1);
1630 case -1: fprintf(stderr, Name ": cannot run mdmon. "
1631 "Array remains readonly\n");
1632 return -1;
9fe32043
N
1633 default: /* parent - good */
1634 pid = wait(&status);
1635 if (pid < 0 || status != 0)
1636 return -1;
8850ee3e
N
1637 }
1638 return 0;
1639}
1640
40ebbb9c 1641int check_env(char *name)
5dcfcb71 1642{
40ebbb9c 1643 char *val = getenv(name);
5dcfcb71
DW
1644
1645 if (val && atoi(val) == 1)
1646 return 1;
1647
1648 return 0;
1649}
1650
148acb7b
DW
1651__u32 random32(void)
1652{
1653 __u32 rv;
1654 int rfd = open("/dev/urandom", O_RDONLY);
1655 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1656 rv = random();
1657 if (rfd >= 0)
1658 close(rfd);
1659 return rv;
1660}
1661
0e600426 1662#ifndef MDASSEMBLE
edd8d13c
NB
1663int flush_metadata_updates(struct supertype *st)
1664{
1665 int sfd;
1666 if (!st->updates) {
1667 st->update_tail = NULL;
1668 return -1;
1669 }
1670
1671 sfd = connect_monitor(devnum2devname(st->container_dev));
1672 if (sfd < 0)
1673 return -1;
1674
1675 while (st->updates) {
1676 struct metadata_update *mu = st->updates;
1677 st->updates = mu->next;
1678
1679 send_message(sfd, mu, 0);
1680 wait_reply(sfd, 0);
1681 free(mu->buf);
1682 free(mu);
1683 }
1684 ack(sfd, 0);
1685 wait_reply(sfd, 0);
1686 close(sfd);
1687 st->update_tail = NULL;
1688 return 0;
1689}
1690
1691void append_metadata_update(struct supertype *st, void *buf, int len)
1692{
1693
1694 struct metadata_update *mu = malloc(sizeof(*mu));
1695
1696 mu->buf = buf;
1697 mu->len = len;
1698 mu->space = NULL;
cb23f1f4 1699 mu->space_list = NULL;
edd8d13c
NB
1700 mu->next = NULL;
1701 *st->update_tail = mu;
1702 st->update_tail = &mu->next;
1703}
0e600426 1704#endif /* MDASSEMBLE */
a931db9e 1705
fe6729fa
NB
1706#ifdef __TINYC__
1707/* tinyc doesn't optimize this check in ioctl.h out ... */
1708unsigned int __invalid_size_argument_for_IOC = 0;
1709#endif
1710
6d11ec6f
AK
1711int experimental(void)
1712{
1713 if (check_env("MDADM_EXPERIMENTAL"))
1714 return 1;
1715 else {
65c83a80
LD
1716 fprintf(stderr, Name ": To use this feature MDADM_EXPERIMENTAL"
1717 " environment variable has to be defined.\n");
6d11ec6f
AK
1718 return 0;
1719 }
1720}
1721
326727d9
AC
1722/* Pick all spares matching given criteria from a container
1723 * if min_size == 0 do not check size
1724 * if domlist == NULL do not check domains
1725 * if spare_group given add it to domains of each spare
1726 * metadata allows to test domains using metadata of destination array */
1727struct mdinfo *container_choose_spares(struct supertype *st,
1728 unsigned long long min_size,
1729 struct domainlist *domlist,
1730 char *spare_group,
1731 const char *metadata, int get_one)
1732{
1733 struct mdinfo *d, **dp, *disks = NULL;
1734
1735 /* get list of all disks in container */
1736 if (st->ss->getinfo_super_disks)
1737 disks = st->ss->getinfo_super_disks(st);
1738
1739 if (!disks)
1740 return disks;
1741 /* find spare devices on the list */
1742 dp = &disks->devs;
1743 disks->array.spare_disks = 0;
1744 while (*dp) {
1745 int found = 0;
1746 d = *dp;
1747 if (d->disk.state == 0) {
1748 /* check if size is acceptable */
1749 unsigned long long dev_size;
1750 dev_t dev = makedev(d->disk.major,d->disk.minor);
1751
1752 if (!min_size ||
1753 (dev_size_from_id(dev, &dev_size) &&
1754 dev_size >= min_size))
1755 found = 1;
1756 /* check if domain matches */
1757 if (found && domlist) {
1758 struct dev_policy *pol = devnum_policy(dev);
1759 if (spare_group)
1760 pol_add(&pol, pol_domain,
1761 spare_group, NULL);
e5508b36 1762 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1763 found = 0;
1764 dev_policy_free(pol);
1765 }
1766 }
1767 if (found) {
1768 dp = &d->next;
1769 disks->array.spare_disks++;
1770 if (get_one) {
1771 sysfs_free(*dp);
1772 d->next = NULL;
1773 }
1774 } else {
1775 *dp = d->next;
1776 d->next = NULL;
1777 sysfs_free(d);
1778 }
1779 }
1780 return disks;
1781}