mdmon: Fix crash if /proc/mdstat lists 0.9 superblocks
[thirdparty/mdadm.git] / util.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
e736b623 4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
64c4757e
NB
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
e736b623 22 * Email: <neilb@suse.de>
64c4757e
NB
23 */
24
9a9dab36 25#include "mdadm.h"
64c4757e 26#include "md_p.h"
edd8d13c 27#include <sys/socket.h>
64c4757e 28#include <sys/utsname.h>
9fe32043 29#include <sys/wait.h>
edd8d13c 30#include <sys/un.h>
98c6faba 31#include <ctype.h>
a322f70c 32#include <dirent.h>
a931db9e 33#include <signal.h>
0a816ef9
NB
34
35/*
36 * following taken from linux/blkpg.h because they aren't
37 * anywhere else and it isn't safe to #include linux/ * stuff.
38 */
39
40#define BLKPG _IO(0x12,105)
41
42/* The argument structure */
43struct blkpg_ioctl_arg {
44 int op;
45 int flags;
46 int datalen;
47 void *data;
48};
49
50/* The subfunctions (for the op field) */
51#define BLKPG_ADD_PARTITION 1
52#define BLKPG_DEL_PARTITION 2
53
54/* Sizes of name fields. Unused at present. */
55#define BLKPG_DEVNAMELTH 64
56#define BLKPG_VOLNAMELTH 64
57
58/* The data structure for ADD_PARTITION and DEL_PARTITION */
59struct blkpg_partition {
60 long long start; /* starting offset in bytes */
61 long long length; /* length in bytes */
62 int pno; /* partition number */
63 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
64 to be used in kernel messages */
65 char volname[BLKPG_VOLNAMELTH]; /* volume label */
66};
64c4757e 67
0f22b998 68#include "part.h"
056b331e
N
69
70/* Force a compilation error if condition is true */
71#define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
72
73/* Force a compilation error if condition is true, but also produce a
74 result (of value 0 and type size_t), so the expression can be used
75 e.g. in a structure initializer (or where-ever else comma expressions
76 aren't permitted). */
77#define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
78
64c4757e
NB
79/*
80 * Parse a 128 bit uuid in 4 integers
81 * format is 32 hexx nibbles with options :.<space> separator
82 * If not exactly 32 hex digits are found, return 0
83 * else return 1
84 */
85int parse_uuid(char *str, int uuid[4])
86{
aba69144
NB
87 int hit = 0; /* number of Hex digIT */
88 int i;
89 char c;
90 for (i=0; i<4; i++) uuid[i]=0;
91
92 while ((c= *str++)) {
93 int n;
94 if (c>='0' && c<='9')
95 n = c-'0';
96 else if (c>='a' && c <= 'f')
97 n = 10 + c - 'a';
98 else if (c>='A' && c <= 'F')
99 n = 10 + c - 'A';
100 else if (strchr(":. -", c))
101 continue;
102 else return 0;
103
104 if (hit<32) {
105 uuid[hit/8] <<= 4;
106 uuid[hit/8] += n;
107 }
108 hit++;
82b27616 109 }
aba69144
NB
110 if (hit == 32)
111 return 1;
112 return 0;
64c4757e
NB
113}
114
115
116/*
117 * Get the md version number.
118 * We use the RAID_VERSION ioctl if it is supported
119 * If not, but we have a block device with major '9', we assume
120 * 0.36.0
121 *
122 * Return version number as 24 but number - assume version parts
123 * always < 255
124 */
125
126int md_get_version(int fd)
127{
128 struct stat stb;
129 mdu_version_t vers;
130
131 if (fstat(fd, &stb)<0)
132 return -1;
133 if ((S_IFMT&stb.st_mode) != S_IFBLK)
134 return -1;
135
136 if (ioctl(fd, RAID_VERSION, &vers) == 0)
682c7051 137 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
5787fa49
NB
138 if (errno == EACCES)
139 return -1;
0df46c2a 140 if (major(stb.st_rdev) == MD_MAJOR)
682c7051 141 return (3600);
64c4757e
NB
142 return -1;
143}
144
64c4757e
NB
145int get_linux_version()
146{
147 struct utsname name;
98c6faba 148 char *cp;
64c4757e
NB
149 int a,b,c;
150 if (uname(&name) <0)
151 return -1;
152
98c6faba
NB
153 cp = name.release;
154 a = strtoul(cp, &cp, 10);
155 if (*cp != '.') return -1;
156 b = strtoul(cp+1, &cp, 10);
157 if (*cp != '.') return -1;
158 c = strtoul(cp+1, NULL, 10);
159
682c7051 160 return (a*1000000)+(b*1000)+c;
64c4757e
NB
161}
162
71204a50 163#ifndef MDASSEMBLE
bc77ed53
DW
164int mdadm_version(char *version)
165{
166 int a, b, c;
167 char *cp;
168
169 if (!version)
170 version = Version;
171
172 cp = strchr(version, '-');
173 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
174 return -1;
175 cp += 3;
176 a = strtoul(cp, &cp, 10);
177 if (*cp != '.')
178 return -1;
179 b = strtoul(cp+1, &cp, 10);
180 if (*cp == '.')
181 c = strtoul(cp+1, &cp, 10);
182 else
183 c = 0;
184 if (*cp != ' ' && *cp != '-')
185 return -1;
186 return (a*1000000)+(b*1000)+c;
187}
188
84e11361
N
189long long parse_size(char *size)
190{
191 /* parse 'size' which should be a number optionally
192 * followed by 'K', 'M', or 'G'.
193 * Without a suffix, K is assumed.
194 * Number returned is in sectors (half-K)
195 */
196 char *c;
197 long long s = strtoll(size, &c, 10);
198 if (s > 0) {
199 switch (*c) {
200 case 'K':
201 c++;
202 default:
203 s *= 2;
204 break;
205 case 'M':
206 c++;
207 s *= 1024 * 2;
208 break;
209 case 'G':
210 c++;
211 s *= 1024 * 1024 * 2;
212 break;
213 }
214 }
215 if (*c)
216 s = 0;
217 return s;
218}
219
4a06e2c2
N
220int parse_layout_10(char *layout)
221{
222 int copies, rv;
223 char *cp;
224 /* Parse the layout string for raid10 */
225 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
226 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
227 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
228 copies > 200 ||
229 *cp)
230 return -1;
231 if (layout[0] == 'n')
232 rv = 256 + copies;
233 else if (layout[0] == 'o')
234 rv = 0x10000 + (copies<<8) + 1;
235 else
236 rv = 1 + (copies<<8);
237 return rv;
238}
239
240int parse_layout_faulty(char *layout)
241{
242 /* Parse the layout string for 'faulty' */
243 int ln = strcspn(layout, "0123456789");
244 char *m = strdup(layout);
245 int mode;
246 m[ln] = 0;
247 mode = map_name(faultylayout, m);
248 if (mode == UnSet)
249 return -1;
250
251 return mode | (atoi(layout+ln)<< ModeShift);
252}
eb3929a4 253#endif
4a06e2c2 254
0430ed48
NB
255void remove_partitions(int fd)
256{
257 /* remove partitions from this block devices.
258 * This is used for components added to an array
259 */
260#ifdef BLKPG_DEL_PARTITION
261 struct blkpg_ioctl_arg a;
262 struct blkpg_partition p;
263
264 a.op = BLKPG_DEL_PARTITION;
265 a.data = (void*)&p;
266 a.datalen = sizeof(p);
267 a.flags = 0;
268 memset(a.data, 0, a.datalen);
269 for (p.pno=0; p.pno < 16; p.pno++)
270 ioctl(fd, BLKPG, &a);
271#endif
272}
273
691c6ee1
N
274int test_partition(int fd)
275{
276 /* Check if fd is a whole-disk or a partition.
277 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
278 * will return ENXIO on an invalid partition number.
279 */
280 struct blkpg_ioctl_arg a;
281 struct blkpg_partition p;
282 a.op = BLKPG_DEL_PARTITION;
283 a.data = (void*)&p;
284 a.datalen = sizeof(p);
285 a.flags = 0;
286 memset(a.data, 0, a.datalen);
287 p.pno = 1<<30;
288 if (ioctl(fd, BLKPG, &a) == 0)
289 /* Very unlikely, but not a partition */
290 return 0;
291 if (errno == ENXIO)
292 /* not a partition */
293 return 0;
294
295 return 1;
296}
297
bfd76b93
CA
298int test_partition_from_id(dev_t id)
299{
300 char buf[20];
301 int fd, rv;
302
303 sprintf(buf, "%d:%d", major(id), minor(id));
304 fd = dev_open(buf, O_RDONLY);
305 if (fd < 0)
306 return -1;
307 rv = test_partition(fd);
308 close(fd);
309 return rv;
310}
691c6ee1 311
583315d9 312int enough(int level, int raid_disks, int layout, int clean,
265e0f17 313 char *avail, int avail_disks)
64c4757e 314{
265e0f17 315 int copies, first;
64c4757e 316 switch (level) {
265e0f17
NB
317 case 10:
318 /* This is the tricky one - we need to check
319 * which actual disks are present.
320 */
702b557b 321 copies = (layout&255)* ((layout>>8) & 255);
265e0f17
NB
322 first=0;
323 do {
324 /* there must be one of the 'copies' form 'first' */
325 int n = copies;
326 int cnt=0;
327 while (n--) {
328 if (avail[first])
329 cnt++;
330 first = (first+1) % raid_disks;
331 }
332 if (cnt == 0)
333 return 0;
334
335 } while (first != 0);
336 return 1;
e5329c37 337
df0d4ea0 338 case LEVEL_MULTIPATH:
e0d19036 339 return avail_disks>= 1;
df0d4ea0 340 case LEVEL_LINEAR:
64c4757e
NB
341 case 0:
342 return avail_disks == raid_disks;
343 case 1:
344 return avail_disks >= 1;
345 case 4:
346 case 5:
583315d9
NB
347 if (clean)
348 return avail_disks >= raid_disks-1;
349 else
350 return avail_disks >= raid_disks;
98c6faba 351 case 6:
583315d9
NB
352 if (clean)
353 return avail_disks >= raid_disks-2;
354 else
355 return avail_disks >= raid_disks;
64c4757e
NB
356 default:
357 return 0;
358 }
359}
360
8453e704
N
361int enough_fd(int fd)
362{
363 struct mdu_array_info_s array;
364 struct mdu_disk_info_s disk;
365 int avail_disks = 0;
366 int i;
367 char *avail;
368
369 if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 ||
370 array.raid_disks <= 0)
371 return 0;
372 avail = calloc(array.raid_disks, 1);
373 for (i=0; i<array.raid_disks + array.nr_disks; i++) {
374 disk.number = i;
375 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
376 continue;
377 if (! (disk.state & (1<<MD_DISK_SYNC)))
378 continue;
379 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
380 continue;
381 avail_disks++;
382 avail[disk.raid_disk] = 1;
383 }
384 /* This is used on an active array, so assume it is clean */
385 return enough(array.level, array.raid_disks, array.layout,
386 1,
387 avail, avail_disks);
388}
389
390
22e263f6
AC
391const int uuid_zero[4] = { 0, 0, 0, 0 };
392
f277ce36 393int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 394{
f277ce36
NB
395 if (swapuuid) {
396 /* parse uuids are hostendian.
397 * uuid's from some superblocks are big-ending
aba69144 398 * if there is a difference, we need to swap..
f277ce36
NB
399 */
400 unsigned char *ac = (unsigned char *)a;
401 unsigned char *bc = (unsigned char *)b;
402 int i;
403 for (i=0; i<16; i+= 4) {
404 if (ac[i+0] != bc[i+3] ||
405 ac[i+1] != bc[i+2] ||
406 ac[i+2] != bc[i+1] ||
407 ac[i+3] != bc[i+0])
408 return 0;
409 }
410 return 1;
411 } else {
412 if (a[0]==b[0] &&
413 a[1]==b[1] &&
414 a[2]==b[2] &&
415 a[3]==b[3])
416 return 1;
417 return 0;
418 }
64c4757e 419}
350f29f9
NB
420void copy_uuid(void *a, int b[4], int swapuuid)
421{
422 if (swapuuid) {
423 /* parse uuids are hostendian.
424 * uuid's from some superblocks are big-ending
425 * if there is a difference, we need to swap..
426 */
427 unsigned char *ac = (unsigned char *)a;
428 unsigned char *bc = (unsigned char *)b;
429 int i;
430 for (i=0; i<16; i+= 4) {
431 ac[i+0] = bc[i+3];
432 ac[i+1] = bc[i+2];
433 ac[i+2] = bc[i+1];
434 ac[i+3] = bc[i+0];
435 }
436 } else
437 memcpy(a, b, 16);
438}
64c4757e 439
aae5a112 440char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 441{
9968e376 442 int i, j;
d7288ddc
N
443 char uuid[16];
444 char *c = buf;
445 strcpy(c, "UUID-");
446 c += strlen(c);
aae5a112 447 copy_uuid(uuid, id, swap);
9968e376 448 for (i = 0; i < 4; i++) {
9968e376 449 if (i)
ff54de6e 450 *c++ = sep;
9968e376
DW
451 for (j = 3; j >= 0; j--) {
452 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
453 c+= 2;
454 }
d7288ddc
N
455 }
456 return buf;
aae5a112
DW
457
458}
459
460char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
461{
86983cce
N
462 // dirty hack to work around an issue with super1 superblocks...
463 // super1 superblocks need swapuuid set in order for assembly to
464 // work, but can't have it set if we want this printout to match
465 // all the other uuid printouts in super1.c, so we force swapuuid
466 // to 1 to make our printout match the rest of super1
467 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
468}
469
435d4ebb 470#ifndef MDASSEMBLE
682c7051
NB
471int check_ext2(int fd, char *name)
472{
473 /*
474 * Check for an ext2fs file system.
475 * Superblock is always 1K at 1K offset
476 *
477 * s_magic is le16 at 56 == 0xEF53
478 * report mtime - le32 at 44
479 * blocks - le32 at 4
480 * logblksize - le32 at 24
481 */
482 unsigned char sb[1024];
483 time_t mtime;
484 int size, bsize;
485 if (lseek(fd, 1024,0)!= 1024)
486 return 0;
487 if (read(fd, sb, 1024)!= 1024)
488 return 0;
489 if (sb[56] != 0x53 || sb[57] != 0xef)
490 return 0;
491
492 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
493 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
494 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
495 fprintf(stderr, Name ": %s appears to contain an ext2fs file system\n",
496 name);
497 fprintf(stderr," size=%dK mtime=%s",
498 size*(1<<bsize), ctime(&mtime));
499 return 1;
500}
501
502int check_reiser(int fd, char *name)
503{
504 /*
505 * superblock is at 64K
506 * size is 1024;
507 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
508 *
509 */
510 unsigned char sb[1024];
881990a2 511 unsigned long size;
682c7051
NB
512 if (lseek(fd, 64*1024, 0) != 64*1024)
513 return 0;
514 if (read(fd, sb, 1024) != 1024)
515 return 0;
a46f4061
NB
516 if (strncmp((char*)sb+52, "ReIsErFs",8)!=0 &&
517 strncmp((char*)sb+52, "ReIsEr2Fs",9)!=0)
682c7051
NB
518 return 0;
519 fprintf(stderr, Name ": %s appears to contain a reiserfs file system\n",name);
520 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
881990a2 521 fprintf(stderr, " size = %luK\n", size*4);
aba69144 522
682c7051
NB
523 return 1;
524}
525
526int check_raid(int fd, char *name)
527{
4b1ac34b 528 struct mdinfo info;
682c7051 529 time_t crtime;
d078d77c 530 char *level;
82d9eba6 531 struct supertype *st = guess_super(fd);
f9ce90ba 532
82d9eba6 533 if (!st) return 0;
3da92f27 534 st->ss->load_super(st, fd, name);
82d9eba6
NB
535 /* Looks like a raid array .. */
536 fprintf(stderr, Name ": %s appears to be part of a raid array:\n",
537 name);
a5d85af7 538 st->ss->getinfo_super(st, &info, NULL);
3da92f27 539 st->ss->free_super(st);
82d9eba6 540 crtime = info.array.ctime;
d078d77c
NB
541 level = map_num(pers, info.array.level);
542 if (!level) level = "-unknown-";
543 fprintf(stderr, " level=%s devices=%d ctime=%s",
544 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 545 return 1;
682c7051
NB
546}
547
682c7051
NB
548int ask(char *mesg)
549{
550 char *add = "";
551 int i;
552 for (i=0; i<5; i++) {
553 char buf[100];
554 fprintf(stderr, "%s%s", mesg, add);
555 fflush(stderr);
556 if (fgets(buf, 100, stdin)==NULL)
557 return 0;
558 if (buf[0]=='y' || buf[0]=='Y')
559 return 1;
560 if (buf[0]=='n' || buf[0]=='N')
561 return 0;
562 add = "(y/n) ";
563 }
564 fprintf(stderr, Name ": assuming 'no'\n");
565 return 0;
566}
435d4ebb 567#endif /* MDASSEMBLE */
682c7051 568
8d80900b 569int is_standard(char *dev, int *nump)
e5329c37
NB
570{
571 /* tests if dev is a "standard" md dev name.
572 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 573 * where NN is a string of digits
598f0d58
NB
574 * Returns 1 if a partitionable standard,
575 * -1 if non-partitonable,
576 * 0 if not a standard name.
e5329c37 577 */
8d80900b
NB
578 char *d = strrchr(dev, '/');
579 int type=0;
580 int num;
581 if (!d)
e5329c37 582 return 0;
8d80900b
NB
583 if (strncmp(d, "/d",2)==0)
584 d += 2, type=1; /* /dev/md/dN{pM} */
585 else if (strncmp(d, "/md_d", 5)==0)
2b4ca8f0 586 d += 5, type=1; /* /dev/md_dN{pM} */
8d80900b
NB
587 else if (strncmp(d, "/md", 3)==0)
588 d += 3, type=-1; /* /dev/mdN */
589 else if (d-dev > 3 && strncmp(d-2, "md/", 3)==0)
5a6d1148 590 d += 1, type=-1; /* /dev/md/N */
e5329c37
NB
591 else
592 return 0;
8d80900b 593 if (!*d)
e5329c37 594 return 0;
8d80900b
NB
595 num = atoi(d);
596 while (isdigit(*d))
597 d++;
598 if (*d)
e5329c37 599 return 0;
8d80900b
NB
600 if (nump) *nump = num;
601
602 return type;
e5329c37
NB
603}
604
4b1ac34b 605unsigned long calc_csum(void *super, int bytes)
82b27616 606{
56eb10c0 607 unsigned long long newcsum = 0;
82b27616 608 int i;
4b1ac34b
NB
609 unsigned int csum;
610 unsigned int *superc = (unsigned int*) super;
82b27616 611
4b1ac34b 612 for(i=0; i<bytes/4; i++)
82b27616
NB
613 newcsum+= superc[i];
614 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 615#ifdef __alpha__
aba69144 616/* The in-kernel checksum calculation is always 16bit on
570c0542
NB
617 * the alpha, though it is 32 bit on i386...
618 * I wonder what it is elsewhere... (it uses and API in
619 * a way that it shouldn't).
620 */
621 csum = (csum & 0xffff) + (csum >> 16);
622 csum = (csum & 0xffff) + (csum >> 16);
623#endif
82b27616
NB
624 return csum;
625}
cd29a5c8 626
435d4ebb 627#ifndef MDASSEMBLE
56eb10c0 628char *human_size(long long bytes)
cd29a5c8
NB
629{
630 static char buf[30];
d5d3721e
NB
631
632 /* We convert bytes to either centi-M{ega,ibi}bytes or
633 * centi-G{igi,ibi}bytes, with appropriate rounding,
634 * and then print 1/100th of those as a decimal.
635 * We allow upto 2048Megabytes before converting to
636 * gigabytes, as that shows more precision and isn't
637 * too large a number.
638 * Terrabytes are not yet handled.
639 */
cd29a5c8 640
56eb10c0 641 if (bytes < 5000*1024)
cd29a5c8 642 buf[0]=0;
d5d3721e
NB
643 else if (bytes < 2*1024LL*1024LL*1024LL) {
644 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
645 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 646 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
647 cMiB/100 , cMiB % 100,
648 cMB/100, cMB % 100);
649 } else {
650 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
651 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 652 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
653 cGiB/100 , cGiB % 100,
654 cGB/100, cGB % 100);
655 }
cd29a5c8
NB
656 return buf;
657}
e0d19036
NB
658
659char *human_size_brief(long long bytes)
660{
661 static char buf[30];
e0d19036
NB
662
663 if (bytes < 5000*1024)
8f23b0b3 664 snprintf(buf, sizeof(buf), "%ld.%02ldKiB",
bd526cee 665 (long)(bytes>>10), (long)(((bytes&1023)*100+512)/1024)
e0d19036
NB
666 );
667 else if (bytes < 2*1024LL*1024LL*1024LL)
8f23b0b3 668 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
e0d19036 669 (long)(bytes>>20),
bd526cee 670 (long)((bytes&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
671 );
672 else
8f23b0b3 673 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
e0d19036 674 (long)(bytes>>30),
bd526cee 675 (long)(((bytes>>10)&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
676 );
677 return buf;
678}
e4965ef8
N
679
680void print_r10_layout(int layout)
681{
682 int near = layout & 255;
683 int far = (layout >> 8) & 255;
684 int offset = (layout&0x10000);
685 char *sep = "";
686
687 if (near != 1) {
688 printf("%s near=%d", sep, near);
689 sep = ",";
690 }
691 if (far != 1)
692 printf("%s %s=%d", sep, offset?"offset":"far", far);
693 if (near*far == 1)
694 printf("NO REDUNDANCY");
695}
435d4ebb 696#endif
e0d19036 697
5f8097be
NB
698unsigned long long calc_array_size(int level, int raid_disks, int layout,
699 int chunksize, unsigned long long devsize)
700{
701 int data_disks = 0;
702 switch (level) {
703 case 0: data_disks = raid_disks; break;
704 case 1: data_disks = 1; break;
705 case 4:
706 case 5: data_disks = raid_disks - 1; break;
707 case 6: data_disks = raid_disks - 2; break;
708 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
709 break;
710 }
711 devsize &= ~(unsigned long long)((chunksize>>9)-1);
712 return data_disks * devsize;
713}
714
0e600426 715#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
e0d19036
NB
716char *get_md_name(int dev)
717{
718 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 719 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
e0d19036
NB
720 static char devname[50];
721 struct stat stb;
98c6faba 722 dev_t rdev;
dd0781e5 723 char *dn;
98c6faba
NB
724
725 if (dev < 0) {
dd0781e5
NB
726 int mdp = get_mdp_major();
727 if (mdp < 0) return NULL;
0df46c2a 728 rdev = makedev(mdp, (-1-dev)<<6);
8f23b0b3 729 snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev);
98c6faba
NB
730 if (stat(devname, &stb) == 0
731 && (S_IFMT&stb.st_mode) == S_IFBLK
732 && (stb.st_rdev == rdev))
733 return devname;
734 } else {
0df46c2a 735 rdev = makedev(MD_MAJOR, dev);
8f23b0b3 736 snprintf(devname, sizeof(devname), "/dev/md%d", dev);
98c6faba
NB
737 if (stat(devname, &stb) == 0
738 && (S_IFMT&stb.st_mode) == S_IFBLK
739 && (stb.st_rdev == rdev))
740 return devname;
741
8f23b0b3 742 snprintf(devname, sizeof(devname), "/dev/md/%d", dev);
98c6faba
NB
743 if (stat(devname, &stb) == 0
744 && (S_IFMT&stb.st_mode) == S_IFBLK
745 && (stb.st_rdev == rdev))
746 return devname;
747 }
16c6fa80 748 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
749 if (dn)
750 return dn;
8f23b0b3 751 snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev);
e0d19036 752 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
753 if (errno != EEXIST)
754 return NULL;
e0d19036
NB
755
756 if (stat(devname, &stb) == 0
757 && (S_IFMT&stb.st_mode) == S_IFBLK
758 && (stb.st_rdev == rdev))
759 return devname;
760 unlink(devname);
761 return NULL;
762}
763
764void put_md_name(char *name)
765{
766 if (strncmp(name, "/dev/.tmp.md", 12)==0)
767 unlink(name);
768}
ea24acd0 769
ea24acd0
NB
770int find_free_devnum(int use_partitions)
771{
772 int devnum;
773 for (devnum = 127; devnum != 128;
a56fb7ec 774 devnum = devnum ? devnum-1 : (1<<20)-1) {
ea24acd0
NB
775 char *dn;
776 int _devnum;
777
778 _devnum = use_partitions ? (-1-devnum) : devnum;
779 if (mddev_busy(_devnum))
780 continue;
781 /* make sure it is new to /dev too, at least as a
782 * non-standard */
783 dn = map_dev(dev2major(_devnum), dev2minor(_devnum), 0);
784 if (dn && ! is_standard(dn, NULL))
785 continue;
786 break;
787 }
788 if (devnum == 128)
789 return NoMdDev;
790 return use_partitions ? (-1-devnum) : devnum;
791}
435d4ebb 792#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 793
8b0dabea
NB
794int dev_open(char *dev, int flags)
795{
796 /* like 'open', but if 'dev' matches %d:%d, create a temp
797 * block device and open that
798 */
799 char *e;
800 int fd = -1;
801 char devname[32];
e81cdd9f 802 int major;
8b0dabea 803 int minor;
e81cdd9f
NB
804
805 if (!dev) return -1;
6df6a774 806 flags |= O_DIRECT;
e81cdd9f
NB
807
808 major = strtoul(dev, &e, 0);
8b0dabea
NB
809 if (e > dev && *e == ':' && e[1] &&
810 (minor = strtoul(e+1, &e, 0)) >= 0 &&
811 *e == 0) {
6df6a774
N
812 char *path = map_dev(major, minor, 0);
813 if (path)
814 fd = open(path, flags);
815 if (fd < 0) {
816 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
817 (int)getpid(), major, minor);
818 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
819 fd = open(devname, flags);
820 unlink(devname);
821 }
822 }
823 if (fd < 0) {
824 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
825 (int)getpid(), major, minor);
826 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
827 fd = open(devname, flags);
828 unlink(devname);
829 }
8b0dabea
NB
830 }
831 } else
6df6a774 832 fd = open(dev, flags);
8b0dabea
NB
833 return fd;
834}
f9ce90ba 835
d998b738 836int open_dev_flags(int devnum, int flags)
e8a70c89
N
837{
838 char buf[20];
839
840 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
d998b738
N
841 return dev_open(buf, flags);
842}
843
844int open_dev(int devnum)
845{
846 return open_dev_flags(devnum, O_RDONLY);
e8a70c89
N
847}
848
a931db9e
NB
849int open_dev_excl(int devnum)
850{
851 char buf[20];
852 int i;
7187750e 853 int flags = O_RDWR;
a931db9e
NB
854
855 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
856 for (i=0 ; i<25 ; i++) {
7187750e 857 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
858 if (fd >= 0)
859 return fd;
7187750e
N
860 if (errno == EACCES && flags == O_RDWR) {
861 flags = O_RDONLY;
862 continue;
863 }
a931db9e
NB
864 if (errno != EBUSY)
865 return fd;
866 usleep(200000);
867 }
868 return -1;
869}
870
9008ed1c
N
871int same_dev(char *one, char *two)
872{
873 struct stat st1, st2;
874 if (stat(one, &st1) != 0)
875 return 0;
876 if (stat(two, &st2) != 0)
877 return 0;
878 if ((st1.st_mode & S_IFMT) != S_IFBLK)
879 return 0;
880 if ((st2.st_mode & S_IFMT) != S_IFBLK)
881 return 0;
882 return st1.st_rdev == st2.st_rdev;
883}
884
a7c6e3fb 885void wait_for(char *dev, int fd)
a714580e
N
886{
887 int i;
a7c6e3fb
N
888 struct stat stb_want;
889
890 if (fstat(fd, &stb_want) != 0 ||
891 (stb_want.st_mode & S_IFMT) != S_IFBLK)
892 return;
a714580e
N
893
894 for (i=0 ; i<25 ; i++) {
895 struct stat stb;
a7c6e3fb
N
896 if (stat(dev, &stb) == 0 &&
897 (stb.st_mode & S_IFMT) == S_IFBLK &&
898 (stb.st_rdev == stb_want.st_rdev))
a714580e
N
899 return;
900 usleep(200000);
901 }
436305c6
DW
902 if (i == 25)
903 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
904}
905
0f22b998
N
906struct superswitch *superlist[] =
907{
908 &super0, &super1,
909 &super_ddf, &super_imsm,
0592faeb 910 &mbr, &gpt,
0f22b998 911 NULL };
f9ce90ba 912
ea24acd0 913#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 914
4725bc31 915struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 916{
1686dc25
NB
917 mdu_array_info_t array;
918 int vers;
919 int minor;
920 struct supertype *st = NULL;
7e0f6979 921 struct mdinfo *sra;
142cb9e1 922 char *verstr;
1686dc25
NB
923 char version[20];
924 int i;
f7e7067b 925 char *subarray = NULL;
5f7e44b2 926 int container = NoMdDev;
1686dc25
NB
927
928 sra = sysfs_read(fd, 0, GET_VERSION);
929
930 if (sra) {
7e0f6979
NB
931 vers = sra->array.major_version;
932 minor = sra->array.minor_version;
142cb9e1 933 verstr = sra->text_version;
1686dc25
NB
934 } else {
935 if (ioctl(fd, GET_ARRAY_INFO, &array))
936 array.major_version = array.minor_version = 0;
937 vers = array.major_version;
938 minor = array.minor_version;
142cb9e1 939 verstr = "";
6fbba4c9 940 }
82d9eba6 941
1686dc25
NB
942 if (vers != -1) {
943 sprintf(version, "%d.%d", vers, minor);
944 verstr = version;
6fbba4c9 945 }
3c558363 946 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 947 char *dev = verstr+1;
5f7e44b2 948
f7e7067b 949 subarray = strchr(dev, '/');
f7e7067b
NB
950 if (subarray)
951 *subarray++ = '\0';
f7e7067b 952 subarray = strdup(subarray);
5f7e44b2 953 container = devname2devnum(dev);
f7e7067b
NB
954 if (sra)
955 sysfs_free(sra);
5f7e44b2 956 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
957 if (sra && sra->text_version[0])
958 verstr = sra->text_version;
959 else
960 verstr = "-no-metadata-";
f7e7067b
NB
961 }
962
963 for (i = 0; st == NULL && superlist[i] ; i++)
964 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
965
966 if (sra)
967 sysfs_free(sra);
f7e7067b 968 if (st) {
3b0896f8 969 st->sb = NULL;
1f49fb3a
N
970 if (subarrayp)
971 *subarrayp = subarray;
5f7e44b2
DW
972 st->container_dev = container;
973 st->devnum = fd2devnum(fd);
4725bc31
N
974 } else
975 free(subarray);
5f7e44b2 976
82d9eba6 977 return st;
f9ce90ba 978}
ea24acd0
NB
979#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
980
0f0749ad 981int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
982{
983 char buf[20];
984 int fd;
985
986 sprintf(buf, "%d:%d", major(id), minor(id));
987 fd = dev_open(buf, O_RDONLY);
988 if (fd < 0)
989 return 0;
990 if (get_dev_size(fd, NULL, size)) {
991 close(fd);
992 return 1;
993 }
994 close(fd);
995 return 0;
996}
f9ce90ba 997
159c3a1a 998struct supertype *dup_super(struct supertype *orig)
3da92f27 999{
159c3a1a 1000 struct supertype *st;
1686dc25 1001
d2ca6449
NB
1002 if (!orig)
1003 return orig;
159c3a1a 1004 st = malloc(sizeof(*st));
3da92f27
NB
1005 if (!st)
1006 return st;
ef609477 1007 memset(st, 0, sizeof(*st));
159c3a1a
NB
1008 st->ss = orig->ss;
1009 st->max_devs = orig->max_devs;
1010 st->minor_version = orig->minor_version;
1011 st->sb = NULL;
1012 st->info = NULL;
1013 return st;
3da92f27
NB
1014}
1015
54887ad8 1016struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1017{
1018 /* try each load_super to find the best match,
1019 * and return the best superswitch
1020 */
82d9eba6
NB
1021 struct superswitch *ss;
1022 struct supertype *st;
f21e18ca 1023 time_t besttime = 0;
570c0542 1024 int bestsuper = -1;
f9ce90ba
NB
1025 int i;
1026
82d9eba6 1027 st = malloc(sizeof(*st));
d1d599ea
N
1028 memset(st, 0, sizeof(*st));
1029 st->container_dev = NoMdDev;
1030
f9ce90ba
NB
1031 for (i=0 ; superlist[i]; i++) {
1032 int rv;
1033 ss = superlist[i];
54887ad8
N
1034 if (guess_type == guess_array && ss->add_to_super == NULL)
1035 continue;
1036 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1037 continue;
ef609477 1038 memset(st, 0, sizeof(*st));
df3346e6 1039 st->ignore_hw_compat = 1;
3da92f27 1040 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1041 if (rv == 0) {
1042 struct mdinfo info;
a5d85af7 1043 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1044 if (bestsuper == -1 ||
1045 besttime < info.array.ctime) {
1046 bestsuper = i;
1047 besttime = info.array.ctime;
570c0542 1048 }
3da92f27 1049 ss->free_super(st);
570c0542
NB
1050 }
1051 }
1052 if (bestsuper != -1) {
1053 int rv;
ef609477 1054 memset(st, 0, sizeof(*st));
df3346e6 1055 st->ignore_hw_compat = 1;
3da92f27 1056 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1057 if (rv == 0) {
5e747af2 1058 superlist[bestsuper]->free_super(st);
df3346e6 1059 st->ignore_hw_compat = 0;
82d9eba6 1060 return st;
f9ce90ba
NB
1061 }
1062 }
570c0542 1063 free(st);
f9ce90ba
NB
1064 return NULL;
1065}
fe6729fa 1066
beae1dfe
NB
1067/* Return size of device in bytes */
1068int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1069{
1070 unsigned long long ldsize;
c2c9bb6f
NB
1071 struct stat st;
1072
1073 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1074 ldsize = (unsigned long long)st.st_size;
1075 else
beae1dfe
NB
1076#ifdef BLKGETSIZE64
1077 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1078#endif
1079 {
1080 unsigned long dsize;
1081 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1082 ldsize = dsize;
1083 ldsize <<= 9;
1084 } else {
1085 if (dname)
1086 fprintf(stderr, Name ": Cannot get size of %s: %s\b",
1087 dname, strerror(errno));
1088 return 0;
1089 }
1090 }
1091 *sizep = ldsize;
1092 return 1;
1093}
8fac0577 1094
3a371610
N
1095/* Return true if this can only be a container, not a member device.
1096 * i.e. is and md device and size is zero
1097 */
1098int must_be_container(int fd)
1099{
1100 unsigned long long size;
1101 if (md_get_version(fd) < 0)
1102 return 0;
1103 if (get_dev_size(fd, NULL, &size) == 0)
1104 return 1;
1105 if (size == 0)
1106 return 1;
1107 return 0;
1108}
034b203a
TM
1109
1110/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1111 * Returns: 1 if successful
1112 * -1 for unknown partition type
1113 * 0 for other errors
1114 */
1115static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1116{
056b331e 1117 struct GPT gpt;
034b203a
TM
1118 unsigned char buf[512];
1119 unsigned char empty_gpt_entry[16]= {0};
1120 struct GPT_part_entry *part;
1121 unsigned long long curr_part_end;
1122 unsigned all_partitions, entry_size;
f21e18ca 1123 unsigned part_nr;
034b203a
TM
1124
1125 *endofpart = 0;
1126
056b331e 1127 BUILD_BUG_ON(sizeof(gpt) != 512);
034b203a
TM
1128 /* read GPT header */
1129 lseek(fd, 512, SEEK_SET);
056b331e 1130 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1131 return 0;
1132
1133 /* get the number of partition entries and the entry size */
056b331e
N
1134 all_partitions = __le32_to_cpu(gpt.part_cnt);
1135 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1136
1137 /* Check GPT signature*/
056b331e 1138 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1139 return -1;
1140
1141 /* sanity checks */
1142 if (all_partitions > 1024 ||
1143 entry_size > 512)
1144 return -1;
1145
1146 /* read first GPT partition entries */
1147 if (read(fd, buf, 512) != 512)
1148 return 0;
1149
1150 part = (struct GPT_part_entry*)buf;
1151
1152 for (part_nr=0; part_nr < all_partitions; part_nr++) {
1153 /* is this valid partition? */
1154 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
1155 /* check the last lba for the current partition */
056b331e 1156 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1157 if (curr_part_end > *endofpart)
1158 *endofpart = curr_part_end;
1159 }
1160
1161 part = (struct GPT_part_entry*)((unsigned char*)part + entry_size);
1162
1163 if ((unsigned char *)part >= buf + 512) {
1164 if (read(fd, buf, 512) != 512)
1165 return 0;
1166 part = (struct GPT_part_entry*)buf;
1167 }
1168 }
1169 return 1;
1170}
1171
1172/* Sets endofpart parameter to the last block used by the last partition on the device.
1173 * Returns: 1 if successful
1174 * -1 for unknown partition type
1175 * 0 for other errors
1176 */
1177static int get_last_partition_end(int fd, unsigned long long *endofpart)
1178{
056b331e 1179 struct MBR boot_sect;
034b203a
TM
1180 struct MBR_part_record *part;
1181 unsigned long long curr_part_end;
f21e18ca 1182 unsigned part_nr;
034b203a
TM
1183 int retval = 0;
1184
1185 *endofpart = 0;
1186
056b331e 1187 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1188 /* read MBR */
1189 lseek(fd, 0, 0);
056b331e 1190 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1191 goto abort;
1192
1193 /* check MBP signature */
056b331e 1194 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1195 retval = 1;
1196 /* found the correct signature */
056b331e 1197 part = boot_sect.parts;
034b203a
TM
1198
1199 for (part_nr=0; part_nr < MBR_PARTITIONS; part_nr++) {
1200 /* check for GPT type */
1201 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1202 retval = get_gpt_last_partition_end(fd, endofpart);
1203 break;
1204 }
1205 /* check the last used lba for the current partition */
1206 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1207 __le32_to_cpu(part->blocks_num);
1208 if (curr_part_end > *endofpart)
1209 *endofpart = curr_part_end;
1210
1211 part++;
1212 }
1213 } else {
1214 /* Unknown partition table */
1215 retval = -1;
1216 }
1217 abort:
1218 return retval;
1219}
1220
53ed6ac3
KW
1221int check_partitions(int fd, char *dname, unsigned long long freesize,
1222 unsigned long long size)
034b203a
TM
1223{
1224 /*
1225 * Check where the last partition ends
1226 */
1227 unsigned long long endofpart;
1228 int ret;
1229
1230 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1231 /* There appears to be a partition table here */
1232 if (freesize == 0) {
1233 /* partitions will not be visible in new device */
1234 fprintf(stderr,
1235 Name ": partition table exists on %s but will be lost or\n"
1236 " meaningless after creating array\n",
1237 dname);
1238 return 1;
1239 } else if (endofpart > freesize) {
1240 /* last partition overlaps metadata */
1241 fprintf(stderr,
1242 Name ": metadata will over-write last partition on %s.\n",
1243 dname);
1244 return 1;
53ed6ac3
KW
1245 } else if (size && endofpart > size) {
1246 /* partitions will be truncated in new device */
1247 fprintf(stderr,
1248 Name ": array size is too small to cover all partitions on %s.\n",
1249 dname);
1250 return 1;
034b203a
TM
1251 }
1252 }
1253 return 0;
1254}
1255
8382f19b
NB
1256void get_one_disk(int mdfd, mdu_array_info_t *ainf, mdu_disk_info_t *disk)
1257{
1258 int d;
1259 ioctl(mdfd, GET_ARRAY_INFO, ainf);
1260 for (d = 0 ; d < ainf->raid_disks + ainf->nr_disks ; d++)
1261 if (ioctl(mdfd, GET_DISK_INFO, disk) == 0)
1262 return;
1263}
63152c1b 1264
a322f70c
DW
1265int open_container(int fd)
1266{
1267 /* 'fd' is a block device. Find out if it is in use
1268 * by a container, and return an open fd on that container.
1269 */
1270 char path[256];
1271 char *e;
1272 DIR *dir;
1273 struct dirent *de;
1274 int dfd, n;
1275 char buf[200];
1276 int major, minor;
1277 struct stat st;
1278
1279 if (fstat(fd, &st) != 0)
1280 return -1;
1281 sprintf(path, "/sys/dev/block/%d:%d/holders",
1282 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1283 e = path + strlen(path);
1284
1285 dir = opendir(path);
1286 if (!dir)
1287 return -1;
1288 while ((de = readdir(dir))) {
1289 if (de->d_ino == 0)
1290 continue;
1291 if (de->d_name[0] == '.')
1292 continue;
1293 sprintf(e, "/%s/dev", de->d_name);
1294 dfd = open(path, O_RDONLY);
1295 if (dfd < 0)
1296 continue;
1297 n = read(dfd, buf, sizeof(buf));
1298 close(dfd);
f21e18ca 1299 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1300 continue;
1301 buf[n] = 0;
1302 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1303 continue;
1304 sprintf(buf, "%d:%d", major, minor);
1305 dfd = dev_open(buf, O_RDONLY);
1306 if (dfd >= 0) {
1307 closedir(dir);
1308 return dfd;
1309 }
1310 }
355726fa 1311 closedir(dir);
a322f70c
DW
1312 return -1;
1313}
1314
33414a01
DW
1315struct superswitch *version_to_superswitch(char *vers)
1316{
1317 int i;
1318
1319 for (i = 0; superlist[i]; i++) {
1320 struct superswitch *ss = superlist[i];
1321
1322 if (strcmp(vers, ss->name) == 0)
1323 return ss;
1324 }
1325
1326 return NULL;
1327}
1328
1329int is_container_member(struct mdstat_ent *mdstat, char *container)
1330{
1331 if (mdstat->metadata_version == NULL ||
1332 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
1333 !is_subarray(mdstat->metadata_version+9) ||
1334 strncmp(mdstat->metadata_version+10, container, strlen(container)) != 0 ||
1335 mdstat->metadata_version[10+strlen(container)] != '/')
1336 return 0;
1337
1338 return 1;
1339}
1340
1341int is_subarray_active(char *subarray, char *container)
1342{
1343 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1344 struct mdstat_ent *ent;
1345
e5408a32
DW
1346 for (ent = mdstat; ent; ent = ent->next)
1347 if (is_container_member(ent, container))
e5e5d7ce 1348 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1349 break;
33414a01
DW
1350
1351 free_mdstat(mdstat);
1352
1353 return ent != NULL;
1354}
1355
1356/* open_subarray - opens a subarray in a container
1357 * @dev: container device name
feab51f8 1358 * @st: empty supertype
33414a01
DW
1359 * @quiet: block reporting errors flag
1360 *
1361 * On success returns an fd to a container and fills in *st
1362 */
feab51f8 1363int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1364{
1365 struct mdinfo *mdi;
a951a4f7 1366 struct mdinfo *info;
33414a01
DW
1367 int fd, err = 1;
1368
1369 fd = open(dev, O_RDWR|O_EXCL);
1370 if (fd < 0) {
1371 if (!quiet)
1372 fprintf(stderr, Name ": Couldn't open %s, aborting\n",
1373 dev);
1374 return 2;
1375 }
1376
1377 st->devnum = fd2devnum(fd);
1378 if (st->devnum == NoMdDev) {
1379 if (!quiet)
1380 fprintf(stderr,
1381 Name ": Failed to determine device number for %s\n",
1382 dev);
1383 goto close_fd;
1384 }
1385
1386 mdi = sysfs_read(fd, st->devnum, GET_VERSION|GET_LEVEL);
1387 if (!mdi) {
1388 if (!quiet)
1389 fprintf(stderr, Name ": Failed to read sysfs for %s\n",
1390 dev);
1391 goto close_fd;
1392 }
1393
1394 if (mdi->array.level != UnSet) {
1395 if (!quiet)
1396 fprintf(stderr, Name ": %s is not a container\n", dev);
1397 goto free_sysfs;
1398 }
1399
1400 st->ss = version_to_superswitch(mdi->text_version);
1401 if (!st->ss) {
1402 if (!quiet)
1403 fprintf(stderr,
1404 Name ": Operation not supported for %s metadata\n",
1405 mdi->text_version);
1406 goto free_sysfs;
1407 }
1408
1409 st->devname = devnum2devname(st->devnum);
1410 if (!st->devname) {
1411 if (!quiet)
1412 fprintf(stderr, Name ": Failed to allocate device name\n");
1413 goto free_sysfs;
1414 }
1415
db20d413 1416 if (!st->ss->load_container) {
33414a01 1417 if (!quiet)
db20d413 1418 fprintf(stderr, Name ": %s is not a container\n", dev);
33414a01
DW
1419 goto free_name;
1420 }
1421
db20d413 1422 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1423 if (!quiet)
db20d413
N
1424 fprintf(stderr, Name ": Failed to load metadata for %s\n",
1425 dev);
1426 goto free_name;
33414a01
DW
1427 }
1428
a951a4f7
N
1429 info = st->ss->container_content(st, subarray);
1430 if (!info) {
1431 if (!quiet)
1432 fprintf(stderr, Name ": Failed to find subarray-%s in %s\n",
1433 subarray, dev);
1434 goto free_super;
1435 }
1436 free(info);
1437
33414a01
DW
1438 err = 0;
1439
1440 free_super:
1441 if (err)
1442 st->ss->free_super(st);
1443 free_name:
1444 if (err)
1445 free(st->devname);
1446 free_sysfs:
1447 sysfs_free(mdi);
1448 close_fd:
1449 if (err)
1450 close(fd);
1451
1452 if (err)
1453 return -1;
1454 else
1455 return fd;
1456}
1457
7801ac20
N
1458int add_disk(int mdfd, struct supertype *st,
1459 struct mdinfo *sra, struct mdinfo *info)
1460{
1461 /* Add a device to an array, in one of 2 ways. */
1462 int rv;
1463#ifndef MDASSEMBLE
1464 if (st->ss->external) {
d23534e4
DW
1465 if (info->disk.state & (1<<MD_DISK_SYNC))
1466 info->recovery_start = MaxSector;
1467 else
1468 info->recovery_start = 0;
2904b26f 1469 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1470 if (! rv) {
1471 struct mdinfo *sd2;
f35f2525
N
1472 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1473 if (sd2 == info)
1474 break;
1475 if (sd2 == NULL) {
1476 sd2 = malloc(sizeof(*sd2));
1477 *sd2 = *info;
1478 sd2->next = sra->devs;
1479 sra->devs = sd2;
1480 }
7801ac20
N
1481 }
1482 } else
1483#endif
1484 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1485 return rv;
1486}
1487
de6ae750
N
1488int remove_disk(int mdfd, struct supertype *st,
1489 struct mdinfo *sra, struct mdinfo *info)
1490{
1491 int rv;
1492 /* Remove the disk given by 'info' from the array */
1493#ifndef MDASSEMBLE
1494 if (st->ss->external)
1495 rv = sysfs_set_str(sra, info, "slot", "none");
1496 else
1497#endif
1498 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1499 info->disk.minor));
1500 return rv;
1501}
1502
f35f2525
N
1503int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1504{
1505 /* Initialise kernel's knowledge of array.
1506 * This varies between externally managed arrays
1507 * and older kernels
1508 */
1509 int vers = md_get_version(mdfd);
1510 int rv;
1511
1512#ifndef MDASSEMBLE
1513 if (st->ss->external)
1514 rv = sysfs_set_array(info, vers);
1515 else
1516#endif
1517 if ((vers % 100) >= 1) { /* can use different versions */
1518 mdu_array_info_t inf;
1519 memset(&inf, 0, sizeof(inf));
1520 inf.major_version = info->array.major_version;
1521 inf.minor_version = info->array.minor_version;
1522 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1523 } else
1524 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1525 return rv;
1526}
1527
1e5c6983
DW
1528unsigned long long min_recovery_start(struct mdinfo *array)
1529{
1530 /* find the minimum recovery_start in an array for metadata
1531 * formats that only record per-array recovery progress instead
1532 * of per-device
1533 */
1534 unsigned long long recovery_start = MaxSector;
1535 struct mdinfo *d;
1536
1537 for (d = array->devs; d; d = d->next)
1538 recovery_start = min(recovery_start, d->recovery_start);
1539
1540 return recovery_start;
1541}
1542
24f6f99b 1543int mdmon_pid(int devnum)
a931db9e
NB
1544{
1545 char path[100];
1546 char pid[10];
1547 int fd;
1548 int n;
10013317
PHC
1549 char *devname = devnum2devname(devnum);
1550
753cf905 1551 sprintf(path, "%s/%s.pid", MDMON_DIR, devname);
10013317
PHC
1552 free(devname);
1553
24f6f99b 1554 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1555
1556 if (fd < 0)
cf556303 1557 return -1;
a931db9e
NB
1558 n = read(fd, pid, 9);
1559 close(fd);
1560 if (n <= 0)
cf556303 1561 return -1;
24f6f99b 1562 return atoi(pid);
a931db9e
NB
1563}
1564
24f6f99b 1565int mdmon_running(int devnum)
a931db9e 1566{
24f6f99b
N
1567 int pid = mdmon_pid(devnum);
1568 if (pid <= 0)
a931db9e 1569 return 0;
24f6f99b 1570 if (kill(pid, 0) == 0)
a931db9e
NB
1571 return 1;
1572 return 0;
1573}
1574
8850ee3e
N
1575int start_mdmon(int devnum)
1576{
1577 int i;
44d2e365 1578 int len;
9fe32043
N
1579 pid_t pid;
1580 int status;
44d2e365
N
1581 char pathbuf[1024];
1582 char *paths[4] = {
1583 pathbuf,
1584 "/sbin/mdmon",
1585 "mdmon",
1586 NULL
1587 };
8850ee3e 1588
40ebbb9c 1589 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1590 return 0;
1591
44d2e365
N
1592 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf));
1593 if (len > 0) {
1594 char *sl;
1595 pathbuf[len] = 0;
1596 sl = strrchr(pathbuf, '/');
1597 if (sl)
1598 sl++;
1599 else
1600 sl = pathbuf;
1601 strcpy(sl, "mdmon");
1602 } else
1603 pathbuf[0] = '\0';
1604
8850ee3e
N
1605 switch(fork()) {
1606 case 0:
1607 /* FIXME yuk. CLOSE_EXEC?? */
1608 for (i=3; i < 100; i++)
1609 close(i);
44d2e365
N
1610 for (i=0; paths[i]; i++)
1611 if (paths[i][0])
1612 execl(paths[i], "mdmon",
e8a70c89
N
1613 devnum2devname(devnum),
1614 NULL);
8850ee3e
N
1615 exit(1);
1616 case -1: fprintf(stderr, Name ": cannot run mdmon. "
1617 "Array remains readonly\n");
1618 return -1;
9fe32043
N
1619 default: /* parent - good */
1620 pid = wait(&status);
1621 if (pid < 0 || status != 0)
1622 return -1;
8850ee3e
N
1623 }
1624 return 0;
1625}
1626
40ebbb9c 1627int check_env(char *name)
5dcfcb71 1628{
40ebbb9c 1629 char *val = getenv(name);
5dcfcb71
DW
1630
1631 if (val && atoi(val) == 1)
1632 return 1;
1633
1634 return 0;
1635}
1636
148acb7b
DW
1637__u32 random32(void)
1638{
1639 __u32 rv;
1640 int rfd = open("/dev/urandom", O_RDONLY);
1641 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1642 rv = random();
1643 if (rfd >= 0)
1644 close(rfd);
1645 return rv;
1646}
1647
0e600426 1648#ifndef MDASSEMBLE
edd8d13c
NB
1649int flush_metadata_updates(struct supertype *st)
1650{
1651 int sfd;
1652 if (!st->updates) {
1653 st->update_tail = NULL;
1654 return -1;
1655 }
1656
1657 sfd = connect_monitor(devnum2devname(st->container_dev));
1658 if (sfd < 0)
1659 return -1;
1660
1661 while (st->updates) {
1662 struct metadata_update *mu = st->updates;
1663 st->updates = mu->next;
1664
1665 send_message(sfd, mu, 0);
1666 wait_reply(sfd, 0);
1667 free(mu->buf);
1668 free(mu);
1669 }
1670 ack(sfd, 0);
1671 wait_reply(sfd, 0);
1672 close(sfd);
1673 st->update_tail = NULL;
1674 return 0;
1675}
1676
1677void append_metadata_update(struct supertype *st, void *buf, int len)
1678{
1679
1680 struct metadata_update *mu = malloc(sizeof(*mu));
1681
1682 mu->buf = buf;
1683 mu->len = len;
1684 mu->space = NULL;
cb23f1f4 1685 mu->space_list = NULL;
edd8d13c
NB
1686 mu->next = NULL;
1687 *st->update_tail = mu;
1688 st->update_tail = &mu->next;
1689}
0e600426 1690#endif /* MDASSEMBLE */
a931db9e 1691
fe6729fa
NB
1692#ifdef __TINYC__
1693/* tinyc doesn't optimize this check in ioctl.h out ... */
1694unsigned int __invalid_size_argument_for_IOC = 0;
1695#endif
1696
6d11ec6f
AK
1697int experimental(void)
1698{
1699 if (check_env("MDADM_EXPERIMENTAL"))
1700 return 1;
1701 else {
1702 fprintf(stderr, Name ": To use this feature MDADM_EXPERIMENTAL enviroment variable has to defined.\n");
1703 return 0;
1704 }
1705}
1706
326727d9
AC
1707/* Pick all spares matching given criteria from a container
1708 * if min_size == 0 do not check size
1709 * if domlist == NULL do not check domains
1710 * if spare_group given add it to domains of each spare
1711 * metadata allows to test domains using metadata of destination array */
1712struct mdinfo *container_choose_spares(struct supertype *st,
1713 unsigned long long min_size,
1714 struct domainlist *domlist,
1715 char *spare_group,
1716 const char *metadata, int get_one)
1717{
1718 struct mdinfo *d, **dp, *disks = NULL;
1719
1720 /* get list of all disks in container */
1721 if (st->ss->getinfo_super_disks)
1722 disks = st->ss->getinfo_super_disks(st);
1723
1724 if (!disks)
1725 return disks;
1726 /* find spare devices on the list */
1727 dp = &disks->devs;
1728 disks->array.spare_disks = 0;
1729 while (*dp) {
1730 int found = 0;
1731 d = *dp;
1732 if (d->disk.state == 0) {
1733 /* check if size is acceptable */
1734 unsigned long long dev_size;
1735 dev_t dev = makedev(d->disk.major,d->disk.minor);
1736
1737 if (!min_size ||
1738 (dev_size_from_id(dev, &dev_size) &&
1739 dev_size >= min_size))
1740 found = 1;
1741 /* check if domain matches */
1742 if (found && domlist) {
1743 struct dev_policy *pol = devnum_policy(dev);
1744 if (spare_group)
1745 pol_add(&pol, pol_domain,
1746 spare_group, NULL);
e5508b36 1747 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1748 found = 0;
1749 dev_policy_free(pol);
1750 }
1751 }
1752 if (found) {
1753 dp = &d->next;
1754 disks->array.spare_disks++;
1755 if (get_one) {
1756 sysfs_free(*dp);
1757 d->next = NULL;
1758 }
1759 } else {
1760 *dp = d->next;
1761 d->next = NULL;
1762 sysfs_free(d);
1763 }
1764 }
1765 return disks;
1766}