super1: make aread/awrite always use an aligned buffer.
[thirdparty/mdadm.git] / util.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
e736b623 4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
64c4757e
NB
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
e736b623 22 * Email: <neilb@suse.de>
64c4757e
NB
23 */
24
9a9dab36 25#include "mdadm.h"
64c4757e 26#include "md_p.h"
edd8d13c 27#include <sys/socket.h>
64c4757e 28#include <sys/utsname.h>
9fe32043 29#include <sys/wait.h>
edd8d13c 30#include <sys/un.h>
98c6faba 31#include <ctype.h>
a322f70c 32#include <dirent.h>
a931db9e 33#include <signal.h>
0a816ef9 34
a0963a86
JS
35int __offroot;
36
0a816ef9
NB
37/*
38 * following taken from linux/blkpg.h because they aren't
39 * anywhere else and it isn't safe to #include linux/ * stuff.
40 */
41
42#define BLKPG _IO(0x12,105)
43
44/* The argument structure */
45struct blkpg_ioctl_arg {
46 int op;
47 int flags;
48 int datalen;
49 void *data;
50};
51
52/* The subfunctions (for the op field) */
53#define BLKPG_ADD_PARTITION 1
54#define BLKPG_DEL_PARTITION 2
55
56/* Sizes of name fields. Unused at present. */
57#define BLKPG_DEVNAMELTH 64
58#define BLKPG_VOLNAMELTH 64
59
60/* The data structure for ADD_PARTITION and DEL_PARTITION */
61struct blkpg_partition {
62 long long start; /* starting offset in bytes */
63 long long length; /* length in bytes */
64 int pno; /* partition number */
65 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
66 to be used in kernel messages */
67 char volname[BLKPG_VOLNAMELTH]; /* volume label */
68};
64c4757e 69
0f22b998 70#include "part.h"
056b331e
N
71
72/* Force a compilation error if condition is true */
73#define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
74
75/* Force a compilation error if condition is true, but also produce a
76 result (of value 0 and type size_t), so the expression can be used
77 e.g. in a structure initializer (or where-ever else comma expressions
78 aren't permitted). */
79#define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
80
64c4757e
NB
81/*
82 * Parse a 128 bit uuid in 4 integers
83 * format is 32 hexx nibbles with options :.<space> separator
84 * If not exactly 32 hex digits are found, return 0
85 * else return 1
86 */
87int parse_uuid(char *str, int uuid[4])
88{
aba69144
NB
89 int hit = 0; /* number of Hex digIT */
90 int i;
91 char c;
92 for (i=0; i<4; i++) uuid[i]=0;
93
94 while ((c= *str++)) {
95 int n;
96 if (c>='0' && c<='9')
97 n = c-'0';
98 else if (c>='a' && c <= 'f')
99 n = 10 + c - 'a';
100 else if (c>='A' && c <= 'F')
101 n = 10 + c - 'A';
102 else if (strchr(":. -", c))
103 continue;
104 else return 0;
105
106 if (hit<32) {
107 uuid[hit/8] <<= 4;
108 uuid[hit/8] += n;
109 }
110 hit++;
82b27616 111 }
aba69144
NB
112 if (hit == 32)
113 return 1;
114 return 0;
64c4757e
NB
115}
116
117
118/*
119 * Get the md version number.
120 * We use the RAID_VERSION ioctl if it is supported
121 * If not, but we have a block device with major '9', we assume
122 * 0.36.0
123 *
124 * Return version number as 24 but number - assume version parts
125 * always < 255
126 */
127
128int md_get_version(int fd)
129{
130 struct stat stb;
131 mdu_version_t vers;
132
133 if (fstat(fd, &stb)<0)
134 return -1;
135 if ((S_IFMT&stb.st_mode) != S_IFBLK)
136 return -1;
137
138 if (ioctl(fd, RAID_VERSION, &vers) == 0)
682c7051 139 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
5787fa49
NB
140 if (errno == EACCES)
141 return -1;
0df46c2a 142 if (major(stb.st_rdev) == MD_MAJOR)
682c7051 143 return (3600);
64c4757e
NB
144 return -1;
145}
146
64c4757e
NB
147int get_linux_version()
148{
149 struct utsname name;
98c6faba 150 char *cp;
f161d047 151 int a = 0, b = 0,c = 0;
64c4757e
NB
152 if (uname(&name) <0)
153 return -1;
154
98c6faba
NB
155 cp = name.release;
156 a = strtoul(cp, &cp, 10);
f161d047
N
157 if (*cp == '.')
158 b = strtoul(cp+1, &cp, 10);
159 if (*cp == '.')
160 c = strtoul(cp+1, &cp, 10);
98c6faba 161
682c7051 162 return (a*1000000)+(b*1000)+c;
64c4757e
NB
163}
164
71204a50 165#ifndef MDASSEMBLE
bc77ed53
DW
166int mdadm_version(char *version)
167{
168 int a, b, c;
169 char *cp;
170
171 if (!version)
172 version = Version;
173
174 cp = strchr(version, '-');
175 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
176 return -1;
177 cp += 3;
178 a = strtoul(cp, &cp, 10);
179 if (*cp != '.')
180 return -1;
181 b = strtoul(cp+1, &cp, 10);
182 if (*cp == '.')
183 c = strtoul(cp+1, &cp, 10);
184 else
185 c = 0;
186 if (*cp != ' ' && *cp != '-')
187 return -1;
188 return (a*1000000)+(b*1000)+c;
189}
190
84e11361
N
191long long parse_size(char *size)
192{
193 /* parse 'size' which should be a number optionally
194 * followed by 'K', 'M', or 'G'.
195 * Without a suffix, K is assumed.
196 * Number returned is in sectors (half-K)
197 */
198 char *c;
199 long long s = strtoll(size, &c, 10);
200 if (s > 0) {
201 switch (*c) {
202 case 'K':
203 c++;
204 default:
205 s *= 2;
206 break;
207 case 'M':
208 c++;
209 s *= 1024 * 2;
210 break;
211 case 'G':
212 c++;
213 s *= 1024 * 1024 * 2;
214 break;
215 }
216 }
217 if (*c)
218 s = 0;
219 return s;
220}
221
4a06e2c2
N
222int parse_layout_10(char *layout)
223{
224 int copies, rv;
225 char *cp;
226 /* Parse the layout string for raid10 */
227 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
228 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
229 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
230 copies > 200 ||
231 *cp)
232 return -1;
233 if (layout[0] == 'n')
234 rv = 256 + copies;
235 else if (layout[0] == 'o')
236 rv = 0x10000 + (copies<<8) + 1;
237 else
238 rv = 1 + (copies<<8);
239 return rv;
240}
241
242int parse_layout_faulty(char *layout)
243{
244 /* Parse the layout string for 'faulty' */
245 int ln = strcspn(layout, "0123456789");
246 char *m = strdup(layout);
247 int mode;
248 m[ln] = 0;
249 mode = map_name(faultylayout, m);
250 if (mode == UnSet)
251 return -1;
252
253 return mode | (atoi(layout+ln)<< ModeShift);
254}
eb3929a4 255#endif
4a06e2c2 256
0430ed48
NB
257void remove_partitions(int fd)
258{
259 /* remove partitions from this block devices.
260 * This is used for components added to an array
261 */
262#ifdef BLKPG_DEL_PARTITION
263 struct blkpg_ioctl_arg a;
264 struct blkpg_partition p;
265
266 a.op = BLKPG_DEL_PARTITION;
267 a.data = (void*)&p;
268 a.datalen = sizeof(p);
269 a.flags = 0;
270 memset(a.data, 0, a.datalen);
271 for (p.pno=0; p.pno < 16; p.pno++)
272 ioctl(fd, BLKPG, &a);
273#endif
274}
275
691c6ee1
N
276int test_partition(int fd)
277{
278 /* Check if fd is a whole-disk or a partition.
279 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
280 * will return ENXIO on an invalid partition number.
281 */
282 struct blkpg_ioctl_arg a;
283 struct blkpg_partition p;
284 a.op = BLKPG_DEL_PARTITION;
285 a.data = (void*)&p;
286 a.datalen = sizeof(p);
287 a.flags = 0;
288 memset(a.data, 0, a.datalen);
289 p.pno = 1<<30;
290 if (ioctl(fd, BLKPG, &a) == 0)
291 /* Very unlikely, but not a partition */
292 return 0;
293 if (errno == ENXIO)
294 /* not a partition */
295 return 0;
296
297 return 1;
298}
299
bfd76b93
CA
300int test_partition_from_id(dev_t id)
301{
302 char buf[20];
303 int fd, rv;
304
305 sprintf(buf, "%d:%d", major(id), minor(id));
306 fd = dev_open(buf, O_RDONLY);
307 if (fd < 0)
308 return -1;
309 rv = test_partition(fd);
310 close(fd);
311 return rv;
312}
691c6ee1 313
583315d9 314int enough(int level, int raid_disks, int layout, int clean,
265e0f17 315 char *avail, int avail_disks)
64c4757e 316{
265e0f17 317 int copies, first;
64c4757e 318 switch (level) {
265e0f17
NB
319 case 10:
320 /* This is the tricky one - we need to check
321 * which actual disks are present.
322 */
702b557b 323 copies = (layout&255)* ((layout>>8) & 255);
265e0f17
NB
324 first=0;
325 do {
326 /* there must be one of the 'copies' form 'first' */
327 int n = copies;
328 int cnt=0;
329 while (n--) {
330 if (avail[first])
331 cnt++;
332 first = (first+1) % raid_disks;
333 }
334 if (cnt == 0)
335 return 0;
336
337 } while (first != 0);
338 return 1;
e5329c37 339
df0d4ea0 340 case LEVEL_MULTIPATH:
e0d19036 341 return avail_disks>= 1;
df0d4ea0 342 case LEVEL_LINEAR:
64c4757e
NB
343 case 0:
344 return avail_disks == raid_disks;
345 case 1:
346 return avail_disks >= 1;
347 case 4:
348 case 5:
583315d9
NB
349 if (clean)
350 return avail_disks >= raid_disks-1;
351 else
352 return avail_disks >= raid_disks;
98c6faba 353 case 6:
583315d9
NB
354 if (clean)
355 return avail_disks >= raid_disks-2;
356 else
357 return avail_disks >= raid_disks;
64c4757e
NB
358 default:
359 return 0;
360 }
361}
362
8453e704
N
363int enough_fd(int fd)
364{
365 struct mdu_array_info_s array;
366 struct mdu_disk_info_s disk;
367 int avail_disks = 0;
aabe020d 368 int i, rv;
8453e704
N
369 char *avail;
370
371 if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 ||
372 array.raid_disks <= 0)
373 return 0;
374 avail = calloc(array.raid_disks, 1);
9e6d9291 375 for (i=0; i < 1024 && array.nr_disks > 0; i++) {
8453e704
N
376 disk.number = i;
377 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
378 continue;
9e6d9291
N
379 if (disk.major == 0 && disk.minor == 0)
380 continue;
381 array.nr_disks--;
382
8453e704
N
383 if (! (disk.state & (1<<MD_DISK_SYNC)))
384 continue;
385 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
386 continue;
387 avail_disks++;
388 avail[disk.raid_disk] = 1;
389 }
390 /* This is used on an active array, so assume it is clean */
aabe020d
JS
391 rv = enough(array.level, array.raid_disks, array.layout,
392 1, avail, avail_disks);
393 free(avail);
394 return rv;
8453e704
N
395}
396
397
22e263f6
AC
398const int uuid_zero[4] = { 0, 0, 0, 0 };
399
f277ce36 400int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 401{
f277ce36
NB
402 if (swapuuid) {
403 /* parse uuids are hostendian.
404 * uuid's from some superblocks are big-ending
aba69144 405 * if there is a difference, we need to swap..
f277ce36
NB
406 */
407 unsigned char *ac = (unsigned char *)a;
408 unsigned char *bc = (unsigned char *)b;
409 int i;
410 for (i=0; i<16; i+= 4) {
411 if (ac[i+0] != bc[i+3] ||
412 ac[i+1] != bc[i+2] ||
413 ac[i+2] != bc[i+1] ||
414 ac[i+3] != bc[i+0])
415 return 0;
416 }
417 return 1;
418 } else {
419 if (a[0]==b[0] &&
420 a[1]==b[1] &&
421 a[2]==b[2] &&
422 a[3]==b[3])
423 return 1;
424 return 0;
425 }
64c4757e 426}
350f29f9
NB
427void copy_uuid(void *a, int b[4], int swapuuid)
428{
429 if (swapuuid) {
430 /* parse uuids are hostendian.
431 * uuid's from some superblocks are big-ending
432 * if there is a difference, we need to swap..
433 */
434 unsigned char *ac = (unsigned char *)a;
435 unsigned char *bc = (unsigned char *)b;
436 int i;
437 for (i=0; i<16; i+= 4) {
438 ac[i+0] = bc[i+3];
439 ac[i+1] = bc[i+2];
440 ac[i+2] = bc[i+1];
441 ac[i+3] = bc[i+0];
442 }
443 } else
444 memcpy(a, b, 16);
445}
64c4757e 446
aae5a112 447char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 448{
9968e376 449 int i, j;
d7288ddc
N
450 char uuid[16];
451 char *c = buf;
452 strcpy(c, "UUID-");
453 c += strlen(c);
aae5a112 454 copy_uuid(uuid, id, swap);
9968e376 455 for (i = 0; i < 4; i++) {
9968e376 456 if (i)
ff54de6e 457 *c++ = sep;
9968e376
DW
458 for (j = 3; j >= 0; j--) {
459 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
460 c+= 2;
461 }
d7288ddc
N
462 }
463 return buf;
aae5a112
DW
464
465}
466
467char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
468{
86983cce
N
469 // dirty hack to work around an issue with super1 superblocks...
470 // super1 superblocks need swapuuid set in order for assembly to
471 // work, but can't have it set if we want this printout to match
472 // all the other uuid printouts in super1.c, so we force swapuuid
473 // to 1 to make our printout match the rest of super1
474 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
475}
476
435d4ebb 477#ifndef MDASSEMBLE
682c7051
NB
478int check_ext2(int fd, char *name)
479{
480 /*
481 * Check for an ext2fs file system.
482 * Superblock is always 1K at 1K offset
483 *
484 * s_magic is le16 at 56 == 0xEF53
485 * report mtime - le32 at 44
486 * blocks - le32 at 4
487 * logblksize - le32 at 24
488 */
489 unsigned char sb[1024];
490 time_t mtime;
491 int size, bsize;
492 if (lseek(fd, 1024,0)!= 1024)
493 return 0;
494 if (read(fd, sb, 1024)!= 1024)
495 return 0;
496 if (sb[56] != 0x53 || sb[57] != 0xef)
497 return 0;
498
499 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
500 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
501 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
502 fprintf(stderr, Name ": %s appears to contain an ext2fs file system\n",
503 name);
504 fprintf(stderr," size=%dK mtime=%s",
505 size*(1<<bsize), ctime(&mtime));
506 return 1;
507}
508
509int check_reiser(int fd, char *name)
510{
511 /*
512 * superblock is at 64K
513 * size is 1024;
514 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
515 *
516 */
517 unsigned char sb[1024];
881990a2 518 unsigned long size;
682c7051
NB
519 if (lseek(fd, 64*1024, 0) != 64*1024)
520 return 0;
521 if (read(fd, sb, 1024) != 1024)
522 return 0;
a46f4061
NB
523 if (strncmp((char*)sb+52, "ReIsErFs",8)!=0 &&
524 strncmp((char*)sb+52, "ReIsEr2Fs",9)!=0)
682c7051
NB
525 return 0;
526 fprintf(stderr, Name ": %s appears to contain a reiserfs file system\n",name);
527 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
881990a2 528 fprintf(stderr, " size = %luK\n", size*4);
aba69144 529
682c7051
NB
530 return 1;
531}
532
533int check_raid(int fd, char *name)
534{
4b1ac34b 535 struct mdinfo info;
682c7051 536 time_t crtime;
d078d77c 537 char *level;
82d9eba6 538 struct supertype *st = guess_super(fd);
f9ce90ba 539
82d9eba6 540 if (!st) return 0;
4a96d9ff 541 st->ignore_hw_compat = 1;
3da92f27 542 st->ss->load_super(st, fd, name);
82d9eba6
NB
543 /* Looks like a raid array .. */
544 fprintf(stderr, Name ": %s appears to be part of a raid array:\n",
545 name);
a5d85af7 546 st->ss->getinfo_super(st, &info, NULL);
3da92f27 547 st->ss->free_super(st);
82d9eba6 548 crtime = info.array.ctime;
d078d77c
NB
549 level = map_num(pers, info.array.level);
550 if (!level) level = "-unknown-";
551 fprintf(stderr, " level=%s devices=%d ctime=%s",
552 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 553 return 1;
682c7051
NB
554}
555
682c7051
NB
556int ask(char *mesg)
557{
558 char *add = "";
559 int i;
560 for (i=0; i<5; i++) {
561 char buf[100];
562 fprintf(stderr, "%s%s", mesg, add);
563 fflush(stderr);
564 if (fgets(buf, 100, stdin)==NULL)
565 return 0;
566 if (buf[0]=='y' || buf[0]=='Y')
567 return 1;
568 if (buf[0]=='n' || buf[0]=='N')
569 return 0;
570 add = "(y/n) ";
571 }
572 fprintf(stderr, Name ": assuming 'no'\n");
573 return 0;
574}
435d4ebb 575#endif /* MDASSEMBLE */
682c7051 576
8d80900b 577int is_standard(char *dev, int *nump)
e5329c37
NB
578{
579 /* tests if dev is a "standard" md dev name.
580 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 581 * where NN is a string of digits
598f0d58
NB
582 * Returns 1 if a partitionable standard,
583 * -1 if non-partitonable,
584 * 0 if not a standard name.
e5329c37 585 */
8d80900b
NB
586 char *d = strrchr(dev, '/');
587 int type=0;
588 int num;
589 if (!d)
e5329c37 590 return 0;
8d80900b
NB
591 if (strncmp(d, "/d",2)==0)
592 d += 2, type=1; /* /dev/md/dN{pM} */
593 else if (strncmp(d, "/md_d", 5)==0)
2b4ca8f0 594 d += 5, type=1; /* /dev/md_dN{pM} */
8d80900b
NB
595 else if (strncmp(d, "/md", 3)==0)
596 d += 3, type=-1; /* /dev/mdN */
597 else if (d-dev > 3 && strncmp(d-2, "md/", 3)==0)
5a6d1148 598 d += 1, type=-1; /* /dev/md/N */
e5329c37
NB
599 else
600 return 0;
8d80900b 601 if (!*d)
e5329c37 602 return 0;
8d80900b
NB
603 num = atoi(d);
604 while (isdigit(*d))
605 d++;
606 if (*d)
e5329c37 607 return 0;
8d80900b
NB
608 if (nump) *nump = num;
609
610 return type;
e5329c37
NB
611}
612
4b1ac34b 613unsigned long calc_csum(void *super, int bytes)
82b27616 614{
56eb10c0 615 unsigned long long newcsum = 0;
82b27616 616 int i;
4b1ac34b
NB
617 unsigned int csum;
618 unsigned int *superc = (unsigned int*) super;
82b27616 619
4b1ac34b 620 for(i=0; i<bytes/4; i++)
82b27616
NB
621 newcsum+= superc[i];
622 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 623#ifdef __alpha__
aba69144 624/* The in-kernel checksum calculation is always 16bit on
570c0542
NB
625 * the alpha, though it is 32 bit on i386...
626 * I wonder what it is elsewhere... (it uses and API in
627 * a way that it shouldn't).
628 */
629 csum = (csum & 0xffff) + (csum >> 16);
630 csum = (csum & 0xffff) + (csum >> 16);
631#endif
82b27616
NB
632 return csum;
633}
cd29a5c8 634
435d4ebb 635#ifndef MDASSEMBLE
56eb10c0 636char *human_size(long long bytes)
cd29a5c8
NB
637{
638 static char buf[30];
d5d3721e
NB
639
640 /* We convert bytes to either centi-M{ega,ibi}bytes or
641 * centi-G{igi,ibi}bytes, with appropriate rounding,
642 * and then print 1/100th of those as a decimal.
643 * We allow upto 2048Megabytes before converting to
644 * gigabytes, as that shows more precision and isn't
645 * too large a number.
01619b48 646 * Terabytes are not yet handled.
d5d3721e 647 */
cd29a5c8 648
56eb10c0 649 if (bytes < 5000*1024)
cd29a5c8 650 buf[0]=0;
d5d3721e
NB
651 else if (bytes < 2*1024LL*1024LL*1024LL) {
652 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
653 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 654 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
655 cMiB/100 , cMiB % 100,
656 cMB/100, cMB % 100);
657 } else {
658 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
659 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 660 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
661 cGiB/100 , cGiB % 100,
662 cGB/100, cGB % 100);
663 }
cd29a5c8
NB
664 return buf;
665}
e0d19036
NB
666
667char *human_size_brief(long long bytes)
668{
669 static char buf[30];
e0d19036
NB
670
671 if (bytes < 5000*1024)
8f23b0b3 672 snprintf(buf, sizeof(buf), "%ld.%02ldKiB",
bd526cee 673 (long)(bytes>>10), (long)(((bytes&1023)*100+512)/1024)
e0d19036
NB
674 );
675 else if (bytes < 2*1024LL*1024LL*1024LL)
8f23b0b3 676 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
e0d19036 677 (long)(bytes>>20),
bd526cee 678 (long)((bytes&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
679 );
680 else
8f23b0b3 681 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
e0d19036 682 (long)(bytes>>30),
bd526cee 683 (long)(((bytes>>10)&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
684 );
685 return buf;
686}
e4965ef8
N
687
688void print_r10_layout(int layout)
689{
690 int near = layout & 255;
691 int far = (layout >> 8) & 255;
692 int offset = (layout&0x10000);
693 char *sep = "";
694
695 if (near != 1) {
696 printf("%s near=%d", sep, near);
697 sep = ",";
698 }
699 if (far != 1)
700 printf("%s %s=%d", sep, offset?"offset":"far", far);
701 if (near*far == 1)
702 printf("NO REDUNDANCY");
703}
435d4ebb 704#endif
e0d19036 705
5f8097be
NB
706unsigned long long calc_array_size(int level, int raid_disks, int layout,
707 int chunksize, unsigned long long devsize)
577e8448
AK
708{
709 devsize &= ~(unsigned long long)((chunksize>>9)-1);
710 return get_data_disks(level, layout, raid_disks) * devsize;
711}
712
713int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
714{
715 int data_disks = 0;
716 switch (level) {
717 case 0: data_disks = raid_disks; break;
718 case 1: data_disks = 1; break;
719 case 4:
720 case 5: data_disks = raid_disks - 1; break;
721 case 6: data_disks = raid_disks - 2; break;
722 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
723 break;
724 }
577e8448
AK
725
726 return data_disks;
5f8097be
NB
727}
728
0e600426 729#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
e0d19036
NB
730char *get_md_name(int dev)
731{
732 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 733 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
e0d19036
NB
734 static char devname[50];
735 struct stat stb;
98c6faba 736 dev_t rdev;
dd0781e5 737 char *dn;
98c6faba
NB
738
739 if (dev < 0) {
dd0781e5
NB
740 int mdp = get_mdp_major();
741 if (mdp < 0) return NULL;
0df46c2a 742 rdev = makedev(mdp, (-1-dev)<<6);
8f23b0b3 743 snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev);
98c6faba
NB
744 if (stat(devname, &stb) == 0
745 && (S_IFMT&stb.st_mode) == S_IFBLK
746 && (stb.st_rdev == rdev))
747 return devname;
748 } else {
0df46c2a 749 rdev = makedev(MD_MAJOR, dev);
8f23b0b3 750 snprintf(devname, sizeof(devname), "/dev/md%d", dev);
98c6faba
NB
751 if (stat(devname, &stb) == 0
752 && (S_IFMT&stb.st_mode) == S_IFBLK
753 && (stb.st_rdev == rdev))
754 return devname;
755
8f23b0b3 756 snprintf(devname, sizeof(devname), "/dev/md/%d", dev);
98c6faba
NB
757 if (stat(devname, &stb) == 0
758 && (S_IFMT&stb.st_mode) == S_IFBLK
759 && (stb.st_rdev == rdev))
760 return devname;
761 }
16c6fa80 762 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
763 if (dn)
764 return dn;
8f23b0b3 765 snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev);
e0d19036 766 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
767 if (errno != EEXIST)
768 return NULL;
e0d19036
NB
769
770 if (stat(devname, &stb) == 0
771 && (S_IFMT&stb.st_mode) == S_IFBLK
772 && (stb.st_rdev == rdev))
773 return devname;
774 unlink(devname);
775 return NULL;
776}
777
778void put_md_name(char *name)
779{
780 if (strncmp(name, "/dev/.tmp.md", 12)==0)
781 unlink(name);
782}
ea24acd0 783
ea24acd0
NB
784int find_free_devnum(int use_partitions)
785{
786 int devnum;
787 for (devnum = 127; devnum != 128;
a56fb7ec 788 devnum = devnum ? devnum-1 : (1<<20)-1) {
ea24acd0
NB
789 char *dn;
790 int _devnum;
791
792 _devnum = use_partitions ? (-1-devnum) : devnum;
793 if (mddev_busy(_devnum))
794 continue;
795 /* make sure it is new to /dev too, at least as a
796 * non-standard */
797 dn = map_dev(dev2major(_devnum), dev2minor(_devnum), 0);
798 if (dn && ! is_standard(dn, NULL))
799 continue;
800 break;
801 }
802 if (devnum == 128)
803 return NoMdDev;
804 return use_partitions ? (-1-devnum) : devnum;
805}
435d4ebb 806#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 807
8b0dabea
NB
808int dev_open(char *dev, int flags)
809{
810 /* like 'open', but if 'dev' matches %d:%d, create a temp
811 * block device and open that
812 */
813 char *e;
814 int fd = -1;
815 char devname[32];
e81cdd9f 816 int major;
8b0dabea 817 int minor;
e81cdd9f
NB
818
819 if (!dev) return -1;
6df6a774 820 flags |= O_DIRECT;
e81cdd9f
NB
821
822 major = strtoul(dev, &e, 0);
8b0dabea
NB
823 if (e > dev && *e == ':' && e[1] &&
824 (minor = strtoul(e+1, &e, 0)) >= 0 &&
825 *e == 0) {
6df6a774
N
826 char *path = map_dev(major, minor, 0);
827 if (path)
828 fd = open(path, flags);
829 if (fd < 0) {
830 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
831 (int)getpid(), major, minor);
832 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
833 fd = open(devname, flags);
834 unlink(devname);
835 }
836 }
837 if (fd < 0) {
838 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
839 (int)getpid(), major, minor);
840 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
841 fd = open(devname, flags);
842 unlink(devname);
843 }
8b0dabea
NB
844 }
845 } else
6df6a774 846 fd = open(dev, flags);
8b0dabea
NB
847 return fd;
848}
f9ce90ba 849
d998b738 850int open_dev_flags(int devnum, int flags)
e8a70c89
N
851{
852 char buf[20];
853
854 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
d998b738
N
855 return dev_open(buf, flags);
856}
857
858int open_dev(int devnum)
859{
860 return open_dev_flags(devnum, O_RDONLY);
e8a70c89
N
861}
862
a931db9e
NB
863int open_dev_excl(int devnum)
864{
865 char buf[20];
866 int i;
7187750e 867 int flags = O_RDWR;
a931db9e
NB
868
869 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
870 for (i=0 ; i<25 ; i++) {
7187750e 871 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
872 if (fd >= 0)
873 return fd;
7187750e
N
874 if (errno == EACCES && flags == O_RDWR) {
875 flags = O_RDONLY;
876 continue;
877 }
a931db9e
NB
878 if (errno != EBUSY)
879 return fd;
880 usleep(200000);
881 }
882 return -1;
883}
884
9008ed1c
N
885int same_dev(char *one, char *two)
886{
887 struct stat st1, st2;
888 if (stat(one, &st1) != 0)
889 return 0;
890 if (stat(two, &st2) != 0)
891 return 0;
892 if ((st1.st_mode & S_IFMT) != S_IFBLK)
893 return 0;
894 if ((st2.st_mode & S_IFMT) != S_IFBLK)
895 return 0;
896 return st1.st_rdev == st2.st_rdev;
897}
898
a7c6e3fb 899void wait_for(char *dev, int fd)
a714580e
N
900{
901 int i;
a7c6e3fb
N
902 struct stat stb_want;
903
904 if (fstat(fd, &stb_want) != 0 ||
905 (stb_want.st_mode & S_IFMT) != S_IFBLK)
906 return;
a714580e
N
907
908 for (i=0 ; i<25 ; i++) {
909 struct stat stb;
a7c6e3fb
N
910 if (stat(dev, &stb) == 0 &&
911 (stb.st_mode & S_IFMT) == S_IFBLK &&
912 (stb.st_rdev == stb_want.st_rdev))
a714580e
N
913 return;
914 usleep(200000);
915 }
436305c6
DW
916 if (i == 25)
917 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
918}
919
0f22b998
N
920struct superswitch *superlist[] =
921{
922 &super0, &super1,
923 &super_ddf, &super_imsm,
0592faeb 924 &mbr, &gpt,
0f22b998 925 NULL };
f9ce90ba 926
ea24acd0 927#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 928
4725bc31 929struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 930{
1686dc25
NB
931 mdu_array_info_t array;
932 int vers;
933 int minor;
934 struct supertype *st = NULL;
7e0f6979 935 struct mdinfo *sra;
142cb9e1 936 char *verstr;
1686dc25
NB
937 char version[20];
938 int i;
f7e7067b 939 char *subarray = NULL;
5f7e44b2 940 int container = NoMdDev;
1686dc25
NB
941
942 sra = sysfs_read(fd, 0, GET_VERSION);
943
944 if (sra) {
7e0f6979
NB
945 vers = sra->array.major_version;
946 minor = sra->array.minor_version;
142cb9e1 947 verstr = sra->text_version;
1686dc25
NB
948 } else {
949 if (ioctl(fd, GET_ARRAY_INFO, &array))
950 array.major_version = array.minor_version = 0;
951 vers = array.major_version;
952 minor = array.minor_version;
142cb9e1 953 verstr = "";
6fbba4c9 954 }
82d9eba6 955
1686dc25
NB
956 if (vers != -1) {
957 sprintf(version, "%d.%d", vers, minor);
958 verstr = version;
6fbba4c9 959 }
3c558363 960 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 961 char *dev = verstr+1;
5f7e44b2 962
f7e7067b 963 subarray = strchr(dev, '/');
f7e7067b
NB
964 if (subarray)
965 *subarray++ = '\0';
f7e7067b 966 subarray = strdup(subarray);
5f7e44b2 967 container = devname2devnum(dev);
f7e7067b
NB
968 if (sra)
969 sysfs_free(sra);
5f7e44b2 970 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
971 if (sra && sra->text_version[0])
972 verstr = sra->text_version;
973 else
974 verstr = "-no-metadata-";
f7e7067b
NB
975 }
976
977 for (i = 0; st == NULL && superlist[i] ; i++)
978 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
979
980 if (sra)
981 sysfs_free(sra);
f7e7067b 982 if (st) {
3b0896f8 983 st->sb = NULL;
1f49fb3a
N
984 if (subarrayp)
985 *subarrayp = subarray;
5f7e44b2
DW
986 st->container_dev = container;
987 st->devnum = fd2devnum(fd);
4725bc31
N
988 } else
989 free(subarray);
5f7e44b2 990
82d9eba6 991 return st;
f9ce90ba 992}
ea24acd0
NB
993#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
994
0f0749ad 995int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
996{
997 char buf[20];
998 int fd;
999
1000 sprintf(buf, "%d:%d", major(id), minor(id));
1001 fd = dev_open(buf, O_RDONLY);
1002 if (fd < 0)
1003 return 0;
1004 if (get_dev_size(fd, NULL, size)) {
1005 close(fd);
1006 return 1;
1007 }
1008 close(fd);
1009 return 0;
1010}
f9ce90ba 1011
159c3a1a 1012struct supertype *dup_super(struct supertype *orig)
3da92f27 1013{
159c3a1a 1014 struct supertype *st;
1686dc25 1015
d2ca6449
NB
1016 if (!orig)
1017 return orig;
159c3a1a 1018 st = malloc(sizeof(*st));
3da92f27
NB
1019 if (!st)
1020 return st;
ef609477 1021 memset(st, 0, sizeof(*st));
159c3a1a
NB
1022 st->ss = orig->ss;
1023 st->max_devs = orig->max_devs;
1024 st->minor_version = orig->minor_version;
1025 st->sb = NULL;
1026 st->info = NULL;
1027 return st;
3da92f27
NB
1028}
1029
54887ad8 1030struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1031{
1032 /* try each load_super to find the best match,
1033 * and return the best superswitch
1034 */
82d9eba6
NB
1035 struct superswitch *ss;
1036 struct supertype *st;
f21e18ca 1037 time_t besttime = 0;
570c0542 1038 int bestsuper = -1;
f9ce90ba
NB
1039 int i;
1040
82d9eba6 1041 st = malloc(sizeof(*st));
d1d599ea
N
1042 memset(st, 0, sizeof(*st));
1043 st->container_dev = NoMdDev;
1044
f9ce90ba
NB
1045 for (i=0 ; superlist[i]; i++) {
1046 int rv;
1047 ss = superlist[i];
54887ad8
N
1048 if (guess_type == guess_array && ss->add_to_super == NULL)
1049 continue;
1050 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1051 continue;
ef609477 1052 memset(st, 0, sizeof(*st));
df3346e6 1053 st->ignore_hw_compat = 1;
3da92f27 1054 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1055 if (rv == 0) {
1056 struct mdinfo info;
a5d85af7 1057 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1058 if (bestsuper == -1 ||
1059 besttime < info.array.ctime) {
1060 bestsuper = i;
1061 besttime = info.array.ctime;
570c0542 1062 }
3da92f27 1063 ss->free_super(st);
570c0542
NB
1064 }
1065 }
1066 if (bestsuper != -1) {
1067 int rv;
ef609477 1068 memset(st, 0, sizeof(*st));
df3346e6 1069 st->ignore_hw_compat = 1;
3da92f27 1070 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1071 if (rv == 0) {
5e747af2 1072 superlist[bestsuper]->free_super(st);
df3346e6 1073 st->ignore_hw_compat = 0;
82d9eba6 1074 return st;
f9ce90ba
NB
1075 }
1076 }
570c0542 1077 free(st);
f9ce90ba
NB
1078 return NULL;
1079}
fe6729fa 1080
beae1dfe
NB
1081/* Return size of device in bytes */
1082int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1083{
1084 unsigned long long ldsize;
c2c9bb6f
NB
1085 struct stat st;
1086
1087 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1088 ldsize = (unsigned long long)st.st_size;
1089 else
beae1dfe
NB
1090#ifdef BLKGETSIZE64
1091 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1092#endif
1093 {
1094 unsigned long dsize;
1095 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1096 ldsize = dsize;
1097 ldsize <<= 9;
1098 } else {
1099 if (dname)
1100 fprintf(stderr, Name ": Cannot get size of %s: %s\b",
1101 dname, strerror(errno));
1102 return 0;
1103 }
1104 }
1105 *sizep = ldsize;
1106 return 1;
1107}
8fac0577 1108
3a371610
N
1109/* Return true if this can only be a container, not a member device.
1110 * i.e. is and md device and size is zero
1111 */
1112int must_be_container(int fd)
1113{
1114 unsigned long long size;
1115 if (md_get_version(fd) < 0)
1116 return 0;
1117 if (get_dev_size(fd, NULL, &size) == 0)
1118 return 1;
1119 if (size == 0)
1120 return 1;
1121 return 0;
1122}
034b203a
TM
1123
1124/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1125 * Returns: 1 if successful
1126 * -1 for unknown partition type
1127 * 0 for other errors
1128 */
1129static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1130{
056b331e 1131 struct GPT gpt;
034b203a 1132 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1133 struct GPT_part_entry *part;
1134 char buf[512];
034b203a
TM
1135 unsigned long long curr_part_end;
1136 unsigned all_partitions, entry_size;
f21e18ca 1137 unsigned part_nr;
034b203a
TM
1138
1139 *endofpart = 0;
1140
056b331e 1141 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1142 /* skip protective MBR */
034b203a 1143 lseek(fd, 512, SEEK_SET);
73e658d8 1144 /* read GPT header */
056b331e 1145 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1146 return 0;
1147
1148 /* get the number of partition entries and the entry size */
056b331e
N
1149 all_partitions = __le32_to_cpu(gpt.part_cnt);
1150 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1151
1152 /* Check GPT signature*/
056b331e 1153 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1154 return -1;
1155
1156 /* sanity checks */
1157 if (all_partitions > 1024 ||
db7fdfe4 1158 entry_size > sizeof(buf))
034b203a
TM
1159 return -1;
1160
db7fdfe4
JS
1161 part = (struct GPT_part_entry *)buf;
1162
034b203a 1163 for (part_nr=0; part_nr < all_partitions; part_nr++) {
73e658d8 1164 /* read partition entry */
db7fdfe4 1165 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1166 return 0;
1167
034b203a 1168 /* is this valid partition? */
db7fdfe4 1169 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1170 /* check the last lba for the current partition */
db7fdfe4 1171 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1172 if (curr_part_end > *endofpart)
1173 *endofpart = curr_part_end;
1174 }
1175
034b203a
TM
1176 }
1177 return 1;
1178}
1179
1180/* Sets endofpart parameter to the last block used by the last partition on the device.
1181 * Returns: 1 if successful
1182 * -1 for unknown partition type
1183 * 0 for other errors
1184 */
1185static int get_last_partition_end(int fd, unsigned long long *endofpart)
1186{
056b331e 1187 struct MBR boot_sect;
034b203a
TM
1188 struct MBR_part_record *part;
1189 unsigned long long curr_part_end;
f21e18ca 1190 unsigned part_nr;
034b203a
TM
1191 int retval = 0;
1192
1193 *endofpart = 0;
1194
056b331e 1195 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1196 /* read MBR */
1197 lseek(fd, 0, 0);
056b331e 1198 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1199 goto abort;
1200
1201 /* check MBP signature */
056b331e 1202 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1203 retval = 1;
1204 /* found the correct signature */
056b331e 1205 part = boot_sect.parts;
034b203a
TM
1206
1207 for (part_nr=0; part_nr < MBR_PARTITIONS; part_nr++) {
1208 /* check for GPT type */
1209 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1210 retval = get_gpt_last_partition_end(fd, endofpart);
1211 break;
1212 }
1213 /* check the last used lba for the current partition */
1214 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1215 __le32_to_cpu(part->blocks_num);
1216 if (curr_part_end > *endofpart)
1217 *endofpart = curr_part_end;
1218
1219 part++;
1220 }
1221 } else {
1222 /* Unknown partition table */
1223 retval = -1;
1224 }
1225 abort:
1226 return retval;
1227}
1228
53ed6ac3
KW
1229int check_partitions(int fd, char *dname, unsigned long long freesize,
1230 unsigned long long size)
034b203a
TM
1231{
1232 /*
1233 * Check where the last partition ends
1234 */
1235 unsigned long long endofpart;
1236 int ret;
1237
1238 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1239 /* There appears to be a partition table here */
1240 if (freesize == 0) {
1241 /* partitions will not be visible in new device */
1242 fprintf(stderr,
1243 Name ": partition table exists on %s but will be lost or\n"
1244 " meaningless after creating array\n",
1245 dname);
1246 return 1;
1247 } else if (endofpart > freesize) {
1248 /* last partition overlaps metadata */
1249 fprintf(stderr,
1250 Name ": metadata will over-write last partition on %s.\n",
1251 dname);
1252 return 1;
53ed6ac3
KW
1253 } else if (size && endofpart > size) {
1254 /* partitions will be truncated in new device */
1255 fprintf(stderr,
1256 Name ": array size is too small to cover all partitions on %s.\n",
1257 dname);
1258 return 1;
034b203a
TM
1259 }
1260 }
1261 return 0;
1262}
1263
8382f19b
NB
1264void get_one_disk(int mdfd, mdu_array_info_t *ainf, mdu_disk_info_t *disk)
1265{
1266 int d;
9e6d9291 1267
8382f19b 1268 ioctl(mdfd, GET_ARRAY_INFO, ainf);
9e6d9291
N
1269 for (d = 0 ; d < 1024 ; d++) {
1270 if (ioctl(mdfd, GET_DISK_INFO, disk) == 0 &&
1271 (disk->major || disk->minor))
8382f19b 1272 return;
9e6d9291 1273 }
8382f19b 1274}
63152c1b 1275
a322f70c
DW
1276int open_container(int fd)
1277{
1278 /* 'fd' is a block device. Find out if it is in use
1279 * by a container, and return an open fd on that container.
1280 */
1281 char path[256];
1282 char *e;
1283 DIR *dir;
1284 struct dirent *de;
1285 int dfd, n;
1286 char buf[200];
1287 int major, minor;
1288 struct stat st;
1289
1290 if (fstat(fd, &st) != 0)
1291 return -1;
1292 sprintf(path, "/sys/dev/block/%d:%d/holders",
1293 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1294 e = path + strlen(path);
1295
1296 dir = opendir(path);
1297 if (!dir)
1298 return -1;
1299 while ((de = readdir(dir))) {
1300 if (de->d_ino == 0)
1301 continue;
1302 if (de->d_name[0] == '.')
1303 continue;
1304 sprintf(e, "/%s/dev", de->d_name);
1305 dfd = open(path, O_RDONLY);
1306 if (dfd < 0)
1307 continue;
1308 n = read(dfd, buf, sizeof(buf));
1309 close(dfd);
f21e18ca 1310 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1311 continue;
1312 buf[n] = 0;
1313 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1314 continue;
1315 sprintf(buf, "%d:%d", major, minor);
1316 dfd = dev_open(buf, O_RDONLY);
1317 if (dfd >= 0) {
1318 closedir(dir);
1319 return dfd;
1320 }
1321 }
355726fa 1322 closedir(dir);
a322f70c
DW
1323 return -1;
1324}
1325
33414a01
DW
1326struct superswitch *version_to_superswitch(char *vers)
1327{
1328 int i;
1329
1330 for (i = 0; superlist[i]; i++) {
1331 struct superswitch *ss = superlist[i];
1332
1333 if (strcmp(vers, ss->name) == 0)
1334 return ss;
1335 }
1336
1337 return NULL;
1338}
1339
1340int is_container_member(struct mdstat_ent *mdstat, char *container)
1341{
1342 if (mdstat->metadata_version == NULL ||
1343 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
1344 !is_subarray(mdstat->metadata_version+9) ||
1345 strncmp(mdstat->metadata_version+10, container, strlen(container)) != 0 ||
1346 mdstat->metadata_version[10+strlen(container)] != '/')
1347 return 0;
1348
1349 return 1;
1350}
1351
1352int is_subarray_active(char *subarray, char *container)
1353{
1354 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1355 struct mdstat_ent *ent;
1356
e5408a32
DW
1357 for (ent = mdstat; ent; ent = ent->next)
1358 if (is_container_member(ent, container))
e5e5d7ce 1359 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1360 break;
33414a01
DW
1361
1362 free_mdstat(mdstat);
1363
1364 return ent != NULL;
1365}
1366
1367/* open_subarray - opens a subarray in a container
1368 * @dev: container device name
feab51f8 1369 * @st: empty supertype
33414a01
DW
1370 * @quiet: block reporting errors flag
1371 *
1372 * On success returns an fd to a container and fills in *st
1373 */
feab51f8 1374int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1375{
1376 struct mdinfo *mdi;
a951a4f7 1377 struct mdinfo *info;
33414a01
DW
1378 int fd, err = 1;
1379
1380 fd = open(dev, O_RDWR|O_EXCL);
1381 if (fd < 0) {
1382 if (!quiet)
1383 fprintf(stderr, Name ": Couldn't open %s, aborting\n",
1384 dev);
b990032d 1385 return -1;
33414a01
DW
1386 }
1387
1388 st->devnum = fd2devnum(fd);
1389 if (st->devnum == NoMdDev) {
1390 if (!quiet)
1391 fprintf(stderr,
1392 Name ": Failed to determine device number for %s\n",
1393 dev);
1394 goto close_fd;
1395 }
1396
1397 mdi = sysfs_read(fd, st->devnum, GET_VERSION|GET_LEVEL);
1398 if (!mdi) {
1399 if (!quiet)
1400 fprintf(stderr, Name ": Failed to read sysfs for %s\n",
1401 dev);
1402 goto close_fd;
1403 }
1404
1405 if (mdi->array.level != UnSet) {
1406 if (!quiet)
1407 fprintf(stderr, Name ": %s is not a container\n", dev);
1408 goto free_sysfs;
1409 }
1410
1411 st->ss = version_to_superswitch(mdi->text_version);
1412 if (!st->ss) {
1413 if (!quiet)
1414 fprintf(stderr,
1415 Name ": Operation not supported for %s metadata\n",
1416 mdi->text_version);
1417 goto free_sysfs;
1418 }
1419
1420 st->devname = devnum2devname(st->devnum);
1421 if (!st->devname) {
1422 if (!quiet)
1423 fprintf(stderr, Name ": Failed to allocate device name\n");
1424 goto free_sysfs;
1425 }
1426
db20d413 1427 if (!st->ss->load_container) {
33414a01 1428 if (!quiet)
db20d413 1429 fprintf(stderr, Name ": %s is not a container\n", dev);
33414a01
DW
1430 goto free_name;
1431 }
1432
db20d413 1433 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1434 if (!quiet)
db20d413
N
1435 fprintf(stderr, Name ": Failed to load metadata for %s\n",
1436 dev);
1437 goto free_name;
33414a01
DW
1438 }
1439
a951a4f7
N
1440 info = st->ss->container_content(st, subarray);
1441 if (!info) {
1442 if (!quiet)
1443 fprintf(stderr, Name ": Failed to find subarray-%s in %s\n",
1444 subarray, dev);
1445 goto free_super;
1446 }
1447 free(info);
1448
33414a01
DW
1449 err = 0;
1450
1451 free_super:
1452 if (err)
1453 st->ss->free_super(st);
1454 free_name:
1455 if (err)
1456 free(st->devname);
1457 free_sysfs:
1458 sysfs_free(mdi);
1459 close_fd:
1460 if (err)
1461 close(fd);
1462
1463 if (err)
1464 return -1;
1465 else
1466 return fd;
1467}
1468
7801ac20
N
1469int add_disk(int mdfd, struct supertype *st,
1470 struct mdinfo *sra, struct mdinfo *info)
1471{
1472 /* Add a device to an array, in one of 2 ways. */
1473 int rv;
1474#ifndef MDASSEMBLE
1475 if (st->ss->external) {
d23534e4
DW
1476 if (info->disk.state & (1<<MD_DISK_SYNC))
1477 info->recovery_start = MaxSector;
1478 else
1479 info->recovery_start = 0;
2904b26f 1480 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1481 if (! rv) {
1482 struct mdinfo *sd2;
f35f2525
N
1483 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1484 if (sd2 == info)
1485 break;
1486 if (sd2 == NULL) {
1487 sd2 = malloc(sizeof(*sd2));
1488 *sd2 = *info;
1489 sd2->next = sra->devs;
1490 sra->devs = sd2;
1491 }
7801ac20
N
1492 }
1493 } else
1494#endif
1495 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1496 return rv;
1497}
1498
de6ae750
N
1499int remove_disk(int mdfd, struct supertype *st,
1500 struct mdinfo *sra, struct mdinfo *info)
1501{
1502 int rv;
1503 /* Remove the disk given by 'info' from the array */
1504#ifndef MDASSEMBLE
1505 if (st->ss->external)
1506 rv = sysfs_set_str(sra, info, "slot", "none");
1507 else
1508#endif
1509 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1510 info->disk.minor));
1511 return rv;
1512}
1513
f35f2525
N
1514int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1515{
1516 /* Initialise kernel's knowledge of array.
1517 * This varies between externally managed arrays
1518 * and older kernels
1519 */
1520 int vers = md_get_version(mdfd);
1521 int rv;
1522
1523#ifndef MDASSEMBLE
1524 if (st->ss->external)
1525 rv = sysfs_set_array(info, vers);
1526 else
1527#endif
1528 if ((vers % 100) >= 1) { /* can use different versions */
1529 mdu_array_info_t inf;
1530 memset(&inf, 0, sizeof(inf));
1531 inf.major_version = info->array.major_version;
1532 inf.minor_version = info->array.minor_version;
1533 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1534 } else
1535 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1536 return rv;
1537}
1538
1e5c6983
DW
1539unsigned long long min_recovery_start(struct mdinfo *array)
1540{
1541 /* find the minimum recovery_start in an array for metadata
1542 * formats that only record per-array recovery progress instead
1543 * of per-device
1544 */
1545 unsigned long long recovery_start = MaxSector;
1546 struct mdinfo *d;
1547
1548 for (d = array->devs; d; d = d->next)
1549 recovery_start = min(recovery_start, d->recovery_start);
1550
1551 return recovery_start;
1552}
1553
24f6f99b 1554int mdmon_pid(int devnum)
a931db9e
NB
1555{
1556 char path[100];
1557 char pid[10];
1558 int fd;
1559 int n;
10013317
PHC
1560 char *devname = devnum2devname(devnum);
1561
753cf905 1562 sprintf(path, "%s/%s.pid", MDMON_DIR, devname);
10013317
PHC
1563 free(devname);
1564
24f6f99b 1565 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1566
1567 if (fd < 0)
cf556303 1568 return -1;
a931db9e
NB
1569 n = read(fd, pid, 9);
1570 close(fd);
1571 if (n <= 0)
cf556303 1572 return -1;
24f6f99b 1573 return atoi(pid);
a931db9e
NB
1574}
1575
24f6f99b 1576int mdmon_running(int devnum)
a931db9e 1577{
24f6f99b
N
1578 int pid = mdmon_pid(devnum);
1579 if (pid <= 0)
a931db9e 1580 return 0;
24f6f99b 1581 if (kill(pid, 0) == 0)
a931db9e
NB
1582 return 1;
1583 return 0;
1584}
1585
8850ee3e
N
1586int start_mdmon(int devnum)
1587{
1913c325 1588 int i, skipped;
44d2e365 1589 int len;
9fe32043
N
1590 pid_t pid;
1591 int status;
44d2e365
N
1592 char pathbuf[1024];
1593 char *paths[4] = {
1594 pathbuf,
1595 "/sbin/mdmon",
1596 "mdmon",
1597 NULL
1598 };
8850ee3e 1599
40ebbb9c 1600 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1601 return 0;
1602
9cf014ec 1603 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1604 if (len > 0) {
1605 char *sl;
1606 pathbuf[len] = 0;
1607 sl = strrchr(pathbuf, '/');
1608 if (sl)
1609 sl++;
1610 else
1611 sl = pathbuf;
1612 strcpy(sl, "mdmon");
1613 } else
1614 pathbuf[0] = '\0';
1615
8850ee3e
N
1616 switch(fork()) {
1617 case 0:
1618 /* FIXME yuk. CLOSE_EXEC?? */
1913c325
N
1619 skipped = 0;
1620 for (i=3; skipped < 20; i++)
1621 if (close(i) < 0)
1622 skipped++;
1623 else
1624 skipped = 0;
1625
44d2e365 1626 for (i=0; paths[i]; i++)
a0963a86
JS
1627 if (paths[i][0]) {
1628 if (__offroot) {
1629 execl(paths[i], "mdmon", "--offroot",
1630 devnum2devname(devnum),
1631 NULL);
1632 } else {
1633 execl(paths[i], "mdmon",
1634 devnum2devname(devnum),
1635 NULL);
1636 }
1637 }
8850ee3e
N
1638 exit(1);
1639 case -1: fprintf(stderr, Name ": cannot run mdmon. "
1640 "Array remains readonly\n");
1641 return -1;
9fe32043
N
1642 default: /* parent - good */
1643 pid = wait(&status);
1644 if (pid < 0 || status != 0)
1645 return -1;
8850ee3e
N
1646 }
1647 return 0;
1648}
1649
40ebbb9c 1650int check_env(char *name)
5dcfcb71 1651{
40ebbb9c 1652 char *val = getenv(name);
5dcfcb71
DW
1653
1654 if (val && atoi(val) == 1)
1655 return 1;
1656
1657 return 0;
1658}
1659
148acb7b
DW
1660__u32 random32(void)
1661{
1662 __u32 rv;
1663 int rfd = open("/dev/urandom", O_RDONLY);
1664 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1665 rv = random();
1666 if (rfd >= 0)
1667 close(rfd);
1668 return rv;
1669}
1670
0e600426 1671#ifndef MDASSEMBLE
edd8d13c
NB
1672int flush_metadata_updates(struct supertype *st)
1673{
1674 int sfd;
1675 if (!st->updates) {
1676 st->update_tail = NULL;
1677 return -1;
1678 }
1679
1680 sfd = connect_monitor(devnum2devname(st->container_dev));
1681 if (sfd < 0)
1682 return -1;
1683
1684 while (st->updates) {
1685 struct metadata_update *mu = st->updates;
1686 st->updates = mu->next;
1687
1688 send_message(sfd, mu, 0);
1689 wait_reply(sfd, 0);
1690 free(mu->buf);
1691 free(mu);
1692 }
1693 ack(sfd, 0);
1694 wait_reply(sfd, 0);
1695 close(sfd);
1696 st->update_tail = NULL;
1697 return 0;
1698}
1699
1700void append_metadata_update(struct supertype *st, void *buf, int len)
1701{
1702
1703 struct metadata_update *mu = malloc(sizeof(*mu));
1704
1705 mu->buf = buf;
1706 mu->len = len;
1707 mu->space = NULL;
cb23f1f4 1708 mu->space_list = NULL;
edd8d13c
NB
1709 mu->next = NULL;
1710 *st->update_tail = mu;
1711 st->update_tail = &mu->next;
1712}
0e600426 1713#endif /* MDASSEMBLE */
a931db9e 1714
fe6729fa
NB
1715#ifdef __TINYC__
1716/* tinyc doesn't optimize this check in ioctl.h out ... */
1717unsigned int __invalid_size_argument_for_IOC = 0;
1718#endif
1719
6d11ec6f
AK
1720int experimental(void)
1721{
1722 if (check_env("MDADM_EXPERIMENTAL"))
1723 return 1;
1724 else {
65c83a80
LD
1725 fprintf(stderr, Name ": To use this feature MDADM_EXPERIMENTAL"
1726 " environment variable has to be defined.\n");
6d11ec6f
AK
1727 return 0;
1728 }
1729}
1730
326727d9
AC
1731/* Pick all spares matching given criteria from a container
1732 * if min_size == 0 do not check size
1733 * if domlist == NULL do not check domains
1734 * if spare_group given add it to domains of each spare
1735 * metadata allows to test domains using metadata of destination array */
1736struct mdinfo *container_choose_spares(struct supertype *st,
1737 unsigned long long min_size,
1738 struct domainlist *domlist,
1739 char *spare_group,
1740 const char *metadata, int get_one)
1741{
1742 struct mdinfo *d, **dp, *disks = NULL;
1743
1744 /* get list of all disks in container */
1745 if (st->ss->getinfo_super_disks)
1746 disks = st->ss->getinfo_super_disks(st);
1747
1748 if (!disks)
1749 return disks;
1750 /* find spare devices on the list */
1751 dp = &disks->devs;
1752 disks->array.spare_disks = 0;
1753 while (*dp) {
1754 int found = 0;
1755 d = *dp;
1756 if (d->disk.state == 0) {
1757 /* check if size is acceptable */
1758 unsigned long long dev_size;
1759 dev_t dev = makedev(d->disk.major,d->disk.minor);
1760
1761 if (!min_size ||
1762 (dev_size_from_id(dev, &dev_size) &&
1763 dev_size >= min_size))
1764 found = 1;
1765 /* check if domain matches */
1766 if (found && domlist) {
1767 struct dev_policy *pol = devnum_policy(dev);
1768 if (spare_group)
1769 pol_add(&pol, pol_domain,
1770 spare_group, NULL);
e5508b36 1771 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1772 found = 0;
1773 dev_policy_free(pol);
1774 }
1775 }
1776 if (found) {
1777 dp = &d->next;
1778 disks->array.spare_disks++;
1779 if (get_one) {
1780 sysfs_free(*dp);
1781 d->next = NULL;
1782 }
1783 } else {
1784 *dp = d->next;
1785 d->next = NULL;
1786 sysfs_free(d);
1787 }
1788 }
1789 return disks;
1790}