Grow: fix problem with reshaping RAID4 to RAID0.
[thirdparty/mdadm.git] / util.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
ca3b6696 4 * Copyright (C) 2001-2012 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 {
5d500228
N
44 int op;
45 int flags;
46 int datalen;
47 void *data;
0a816ef9
NB
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;
ca3b6696
N
90 for (i = 0; i < 4; i++)
91 uuid[i] = 0;
aba69144 92
ca3b6696 93 while ((c = *str++) != 0) {
aba69144 94 int n;
ca3b6696 95 if (c >= '0' && c <= '9')
aba69144 96 n = c-'0';
ca3b6696 97 else if (c >= 'a' && c <= 'f')
aba69144 98 n = 10 + c - 'a';
ca3b6696 99 else if (c >= 'A' && c <= 'F')
aba69144
NB
100 n = 10 + c - 'A';
101 else if (strchr(":. -", c))
102 continue;
103 else return 0;
104
105 if (hit<32) {
106 uuid[hit/8] <<= 4;
107 uuid[hit/8] += n;
108 }
109 hit++;
82b27616 110 }
aba69144
NB
111 if (hit == 32)
112 return 1;
113 return 0;
64c4757e
NB
114}
115
64c4757e
NB
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{
5d500228
N
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)
137 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
138 if (errno == EACCES)
139 return -1;
140 if (major(stb.st_rdev) == MD_MAJOR)
141 return (3600);
64c4757e 142 return -1;
64c4757e
NB
143}
144
64c4757e
NB
145int get_linux_version()
146{
147 struct utsname name;
98c6faba 148 char *cp;
f161d047 149 int a = 0, b = 0,c = 0;
64c4757e
NB
150 if (uname(&name) <0)
151 return -1;
152
98c6faba
NB
153 cp = name.release;
154 a = strtoul(cp, &cp, 10);
f161d047
N
155 if (*cp == '.')
156 b = strtoul(cp+1, &cp, 10);
157 if (*cp == '.')
158 c = strtoul(cp+1, &cp, 10);
98c6faba 159
682c7051 160 return (a*1000000)+(b*1000)+c;
64c4757e
NB
161}
162
71204a50 163#ifndef MDASSEMBLE
bc77ed53
DW
164int mdadm_version(char *version)
165{
166 int a, b, c;
167 char *cp;
168
169 if (!version)
170 version = Version;
171
172 cp = strchr(version, '-');
173 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
174 return -1;
175 cp += 3;
176 a = strtoul(cp, &cp, 10);
177 if (*cp != '.')
178 return -1;
179 b = strtoul(cp+1, &cp, 10);
180 if (*cp == '.')
181 c = strtoul(cp+1, &cp, 10);
182 else
183 c = 0;
184 if (*cp != ' ' && *cp != '-')
185 return -1;
186 return (a*1000000)+(b*1000)+c;
187}
188
51d4261c 189unsigned long long parse_size(char *size)
84e11361
N
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)
822e393a 195 * INVALID_SECTORS returned on error.
84e11361
N
196 */
197 char *c;
198 long long s = strtoll(size, &c, 10);
199 if (s > 0) {
200 switch (*c) {
201 case 'K':
202 c++;
203 default:
204 s *= 2;
205 break;
206 case 'M':
207 c++;
208 s *= 1024 * 2;
209 break;
210 case 'G':
211 c++;
212 s *= 1024 * 1024 * 2;
213 break;
72ca9bcf
N
214 case 's': /* sectors */
215 c++;
216 break;
84e11361 217 }
51d4261c 218 } else
822e393a 219 s = INVALID_SECTORS;
84e11361 220 if (*c)
822e393a 221 s = INVALID_SECTORS;
84e11361
N
222 return s;
223}
224
4a06e2c2
N
225int parse_layout_10(char *layout)
226{
227 int copies, rv;
228 char *cp;
229 /* Parse the layout string for raid10 */
230 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
231 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
232 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
233 copies > 200 ||
234 *cp)
235 return -1;
236 if (layout[0] == 'n')
237 rv = 256 + copies;
238 else if (layout[0] == 'o')
239 rv = 0x10000 + (copies<<8) + 1;
240 else
241 rv = 1 + (copies<<8);
242 return rv;
243}
244
245int parse_layout_faulty(char *layout)
246{
247 /* Parse the layout string for 'faulty' */
248 int ln = strcspn(layout, "0123456789");
503975b9 249 char *m = xstrdup(layout);
4a06e2c2
N
250 int mode;
251 m[ln] = 0;
252 mode = map_name(faultylayout, m);
253 if (mode == UnSet)
254 return -1;
255
256 return mode | (atoi(layout+ln)<< ModeShift);
257}
79868890
N
258
259long parse_num(char *num)
260{
261 /* Either return a valid number, or -1 */
262 char *c;
263 long rv = strtol(num, &c, 10);
264 if (rv < 0 || *c || !num[0])
265 return -1;
266 else
267 return rv;
268}
eb3929a4 269#endif
4a06e2c2 270
0430ed48
NB
271void remove_partitions(int fd)
272{
273 /* remove partitions from this block devices.
274 * This is used for components added to an array
275 */
276#ifdef BLKPG_DEL_PARTITION
277 struct blkpg_ioctl_arg a;
278 struct blkpg_partition p;
279
280 a.op = BLKPG_DEL_PARTITION;
281 a.data = (void*)&p;
282 a.datalen = sizeof(p);
283 a.flags = 0;
284 memset(a.data, 0, a.datalen);
ca3b6696 285 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
286 ioctl(fd, BLKPG, &a);
287#endif
288}
289
691c6ee1
N
290int test_partition(int fd)
291{
292 /* Check if fd is a whole-disk or a partition.
293 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
294 * will return ENXIO on an invalid partition number.
295 */
296 struct blkpg_ioctl_arg a;
297 struct blkpg_partition p;
298 a.op = BLKPG_DEL_PARTITION;
299 a.data = (void*)&p;
300 a.datalen = sizeof(p);
301 a.flags = 0;
302 memset(a.data, 0, a.datalen);
303 p.pno = 1<<30;
304 if (ioctl(fd, BLKPG, &a) == 0)
305 /* Very unlikely, but not a partition */
306 return 0;
307 if (errno == ENXIO)
308 /* not a partition */
309 return 0;
310
311 return 1;
312}
313
bfd76b93
CA
314int test_partition_from_id(dev_t id)
315{
316 char buf[20];
317 int fd, rv;
318
319 sprintf(buf, "%d:%d", major(id), minor(id));
320 fd = dev_open(buf, O_RDONLY);
321 if (fd < 0)
322 return -1;
323 rv = test_partition(fd);
324 close(fd);
325 return rv;
326}
691c6ee1 327
de5a472e 328int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 329{
265e0f17 330 int copies, first;
de5a472e
N
331 int i;
332 int avail_disks = 0;
333
334 for (i = 0; i < raid_disks; i++)
335 avail_disks += !!avail[i];
336
64c4757e 337 switch (level) {
265e0f17
NB
338 case 10:
339 /* This is the tricky one - we need to check
340 * which actual disks are present.
341 */
702b557b 342 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 343 first = 0;
265e0f17
NB
344 do {
345 /* there must be one of the 'copies' form 'first' */
346 int n = copies;
ca3b6696 347 int cnt = 0;
b3ec716d 348 int this = first;
265e0f17 349 while (n--) {
b3ec716d 350 if (avail[this])
265e0f17 351 cnt++;
b3ec716d 352 this = (this+1) % raid_disks;
265e0f17
NB
353 }
354 if (cnt == 0)
355 return 0;
b3ec716d 356 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
357 } while (first != 0);
358 return 1;
e5329c37 359
df0d4ea0 360 case LEVEL_MULTIPATH:
e0d19036 361 return avail_disks>= 1;
df0d4ea0 362 case LEVEL_LINEAR:
64c4757e
NB
363 case 0:
364 return avail_disks == raid_disks;
365 case 1:
366 return avail_disks >= 1;
367 case 4:
368 case 5:
583315d9
NB
369 if (clean)
370 return avail_disks >= raid_disks-1;
371 else
372 return avail_disks >= raid_disks;
98c6faba 373 case 6:
583315d9
NB
374 if (clean)
375 return avail_disks >= raid_disks-2;
376 else
377 return avail_disks >= raid_disks;
64c4757e
NB
378 default:
379 return 0;
380 }
381}
382
8453e704
N
383int enough_fd(int fd)
384{
385 struct mdu_array_info_s array;
386 struct mdu_disk_info_s disk;
387 int avail_disks = 0;
aabe020d 388 int i, rv;
8453e704
N
389 char *avail;
390
391 if (ioctl(fd, GET_ARRAY_INFO, &array) != 0 ||
392 array.raid_disks <= 0)
393 return 0;
503975b9 394 avail = xcalloc(array.raid_disks, 1);
ca3b6696 395 for (i = 0; i < MAX_DISKS && array.nr_disks > 0; i++) {
8453e704
N
396 disk.number = i;
397 if (ioctl(fd, GET_DISK_INFO, &disk) != 0)
398 continue;
9e6d9291
N
399 if (disk.major == 0 && disk.minor == 0)
400 continue;
401 array.nr_disks--;
402
8453e704
N
403 if (! (disk.state & (1<<MD_DISK_SYNC)))
404 continue;
405 if (disk.raid_disk < 0 || disk.raid_disk >= array.raid_disks)
406 continue;
407 avail_disks++;
408 avail[disk.raid_disk] = 1;
409 }
410 /* This is used on an active array, so assume it is clean */
aabe020d 411 rv = enough(array.level, array.raid_disks, array.layout,
de5a472e 412 1, avail);
aabe020d
JS
413 free(avail);
414 return rv;
8453e704
N
415}
416
22e263f6
AC
417const int uuid_zero[4] = { 0, 0, 0, 0 };
418
f277ce36 419int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 420{
f277ce36
NB
421 if (swapuuid) {
422 /* parse uuids are hostendian.
423 * uuid's from some superblocks are big-ending
aba69144 424 * if there is a difference, we need to swap..
f277ce36
NB
425 */
426 unsigned char *ac = (unsigned char *)a;
427 unsigned char *bc = (unsigned char *)b;
428 int i;
ca3b6696 429 for (i = 0; i < 16; i += 4) {
f277ce36
NB
430 if (ac[i+0] != bc[i+3] ||
431 ac[i+1] != bc[i+2] ||
432 ac[i+2] != bc[i+1] ||
433 ac[i+3] != bc[i+0])
434 return 0;
435 }
436 return 1;
437 } else {
438 if (a[0]==b[0] &&
439 a[1]==b[1] &&
440 a[2]==b[2] &&
441 a[3]==b[3])
442 return 1;
443 return 0;
444 }
64c4757e 445}
ca3b6696 446
350f29f9
NB
447void copy_uuid(void *a, int b[4], int swapuuid)
448{
449 if (swapuuid) {
450 /* parse uuids are hostendian.
451 * uuid's from some superblocks are big-ending
452 * if there is a difference, we need to swap..
453 */
454 unsigned char *ac = (unsigned char *)a;
455 unsigned char *bc = (unsigned char *)b;
456 int i;
ca3b6696 457 for (i = 0; i < 16; i += 4) {
350f29f9
NB
458 ac[i+0] = bc[i+3];
459 ac[i+1] = bc[i+2];
460 ac[i+2] = bc[i+1];
461 ac[i+3] = bc[i+0];
462 }
463 } else
464 memcpy(a, b, 16);
465}
64c4757e 466
aae5a112 467char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 468{
9968e376 469 int i, j;
d7288ddc
N
470 char uuid[16];
471 char *c = buf;
472 strcpy(c, "UUID-");
473 c += strlen(c);
aae5a112 474 copy_uuid(uuid, id, swap);
9968e376 475 for (i = 0; i < 4; i++) {
9968e376 476 if (i)
ff54de6e 477 *c++ = sep;
9968e376
DW
478 for (j = 3; j >= 0; j--) {
479 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
480 c+= 2;
481 }
d7288ddc
N
482 }
483 return buf;
aae5a112
DW
484
485}
486
487char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
488{
86983cce
N
489 // dirty hack to work around an issue with super1 superblocks...
490 // super1 superblocks need swapuuid set in order for assembly to
491 // work, but can't have it set if we want this printout to match
492 // all the other uuid printouts in super1.c, so we force swapuuid
493 // to 1 to make our printout match the rest of super1
494 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
495}
496
435d4ebb 497#ifndef MDASSEMBLE
682c7051
NB
498int check_ext2(int fd, char *name)
499{
500 /*
501 * Check for an ext2fs file system.
502 * Superblock is always 1K at 1K offset
503 *
504 * s_magic is le16 at 56 == 0xEF53
505 * report mtime - le32 at 44
506 * blocks - le32 at 4
507 * logblksize - le32 at 24
508 */
509 unsigned char sb[1024];
510 time_t mtime;
511 int size, bsize;
512 if (lseek(fd, 1024,0)!= 1024)
513 return 0;
514 if (read(fd, sb, 1024)!= 1024)
515 return 0;
516 if (sb[56] != 0x53 || sb[57] != 0xef)
517 return 0;
518
519 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
520 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
521 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
e7b84f9d 522 pr_err("%s appears to contain an ext2fs file system\n",
682c7051
NB
523 name);
524 fprintf(stderr," size=%dK mtime=%s",
525 size*(1<<bsize), ctime(&mtime));
526 return 1;
527}
528
529int check_reiser(int fd, char *name)
530{
531 /*
532 * superblock is at 64K
533 * size is 1024;
534 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
535 *
536 */
537 unsigned char sb[1024];
881990a2 538 unsigned long size;
682c7051
NB
539 if (lseek(fd, 64*1024, 0) != 64*1024)
540 return 0;
541 if (read(fd, sb, 1024) != 1024)
542 return 0;
ca3b6696
N
543 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
544 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 545 return 0;
e7b84f9d 546 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 547 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
881990a2 548 fprintf(stderr, " size = %luK\n", size*4);
aba69144 549
682c7051
NB
550 return 1;
551}
552
553int check_raid(int fd, char *name)
554{
4b1ac34b 555 struct mdinfo info;
682c7051 556 time_t crtime;
d078d77c 557 char *level;
82d9eba6 558 struct supertype *st = guess_super(fd);
f9ce90ba 559
cb8f6859
N
560 if (!st)
561 return 0;
3da92f27 562 st->ss->load_super(st, fd, name);
82d9eba6 563 /* Looks like a raid array .. */
e7b84f9d 564 pr_err("%s appears to be part of a raid array:\n",
82d9eba6 565 name);
a5d85af7 566 st->ss->getinfo_super(st, &info, NULL);
3da92f27 567 st->ss->free_super(st);
82d9eba6 568 crtime = info.array.ctime;
d078d77c
NB
569 level = map_num(pers, info.array.level);
570 if (!level) level = "-unknown-";
571 fprintf(stderr, " level=%s devices=%d ctime=%s",
572 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 573 return 1;
682c7051
NB
574}
575
682c7051
NB
576int ask(char *mesg)
577{
578 char *add = "";
579 int i;
ca3b6696 580 for (i = 0; i < 5; i++) {
682c7051
NB
581 char buf[100];
582 fprintf(stderr, "%s%s", mesg, add);
583 fflush(stderr);
584 if (fgets(buf, 100, stdin)==NULL)
585 return 0;
586 if (buf[0]=='y' || buf[0]=='Y')
587 return 1;
588 if (buf[0]=='n' || buf[0]=='N')
589 return 0;
590 add = "(y/n) ";
591 }
e7b84f9d 592 pr_err("assuming 'no'\n");
682c7051
NB
593 return 0;
594}
435d4ebb 595#endif /* MDASSEMBLE */
682c7051 596
8d80900b 597int is_standard(char *dev, int *nump)
e5329c37
NB
598{
599 /* tests if dev is a "standard" md dev name.
600 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 601 * where NN is a string of digits
598f0d58
NB
602 * Returns 1 if a partitionable standard,
603 * -1 if non-partitonable,
604 * 0 if not a standard name.
e5329c37 605 */
8d80900b 606 char *d = strrchr(dev, '/');
ca3b6696 607 int type = 0;
8d80900b
NB
608 int num;
609 if (!d)
e5329c37 610 return 0;
ca3b6696
N
611 if (strncmp(d, "/d",2) == 0)
612 d += 2, type = 1; /* /dev/md/dN{pM} */
613 else if (strncmp(d, "/md_d", 5) == 0)
614 d += 5, type = 1; /* /dev/md_dN{pM} */
615 else if (strncmp(d, "/md", 3) == 0)
616 d += 3, type = -1; /* /dev/mdN */
617 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
618 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
619 else
620 return 0;
8d80900b 621 if (!*d)
e5329c37 622 return 0;
8d80900b
NB
623 num = atoi(d);
624 while (isdigit(*d))
625 d++;
626 if (*d)
e5329c37 627 return 0;
8d80900b
NB
628 if (nump) *nump = num;
629
630 return type;
e5329c37
NB
631}
632
4b1ac34b 633unsigned long calc_csum(void *super, int bytes)
82b27616 634{
56eb10c0 635 unsigned long long newcsum = 0;
82b27616 636 int i;
4b1ac34b
NB
637 unsigned int csum;
638 unsigned int *superc = (unsigned int*) super;
82b27616 639
ca3b6696
N
640 for(i = 0; i < bytes/4; i++)
641 newcsum += superc[i];
82b27616 642 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 643#ifdef __alpha__
aba69144 644/* The in-kernel checksum calculation is always 16bit on
570c0542 645 * the alpha, though it is 32 bit on i386...
ca3b6696 646 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
647 * a way that it shouldn't).
648 */
649 csum = (csum & 0xffff) + (csum >> 16);
650 csum = (csum & 0xffff) + (csum >> 16);
651#endif
82b27616
NB
652 return csum;
653}
cd29a5c8 654
435d4ebb 655#ifndef MDASSEMBLE
56eb10c0 656char *human_size(long long bytes)
cd29a5c8
NB
657{
658 static char buf[30];
d5d3721e
NB
659
660 /* We convert bytes to either centi-M{ega,ibi}bytes or
661 * centi-G{igi,ibi}bytes, with appropriate rounding,
662 * and then print 1/100th of those as a decimal.
663 * We allow upto 2048Megabytes before converting to
664 * gigabytes, as that shows more precision and isn't
665 * too large a number.
01619b48 666 * Terabytes are not yet handled.
d5d3721e 667 */
cd29a5c8 668
56eb10c0 669 if (bytes < 5000*1024)
ca3b6696 670 buf[0] = 0;
d5d3721e
NB
671 else if (bytes < 2*1024LL*1024LL*1024LL) {
672 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
673 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 674 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
675 cMiB/100 , cMiB % 100,
676 cMB/100, cMB % 100);
677 } else {
678 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
679 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 680 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
681 cGiB/100 , cGiB % 100,
682 cGB/100, cGB % 100);
683 }
cd29a5c8
NB
684 return buf;
685}
e0d19036 686
f0ec6710 687char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
688{
689 static char buf[30];
e0d19036 690
570abc6f
MN
691 /* We convert bytes to either centi-M{ega,ibi}bytes or
692 * centi-G{igi,ibi}bytes, with appropriate rounding,
693 * and then print 1/100th of those as a decimal.
694 * We allow upto 2048Megabytes before converting to
695 * gigabytes, as that shows more precision and isn't
696 * too large a number.
697 * Terabytes are not yet handled.
f0ec6710
MN
698 *
699 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
700 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
701 */
702
e0d19036 703 if (bytes < 5000*1024)
570abc6f 704 buf[0] = 0;
f0ec6710
MN
705 else if (prefix == IEC) {
706 if (bytes < 2*1024LL*1024LL*1024LL) {
707 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
708 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
709 cMiB/100 , cMiB % 100);
710 } else {
711 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
712 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
713 cGiB/100 , cGiB % 100);
714 }
715 }
716 else if (prefix == JEDEC) {
717 if (bytes < 2*1024LL*1024LL*1024LL) {
718 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
719 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
720 cMB/100, cMB % 100);
721 } else {
722 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
723 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
724 cGB/100 , cGB % 100);
725 }
570abc6f 726 }
f0ec6710
MN
727 else
728 buf[0] = 0;
570abc6f 729
e0d19036
NB
730 return buf;
731}
e4965ef8
N
732
733void print_r10_layout(int layout)
734{
735 int near = layout & 255;
736 int far = (layout >> 8) & 255;
737 int offset = (layout&0x10000);
738 char *sep = "";
739
740 if (near != 1) {
741 printf("%s near=%d", sep, near);
742 sep = ",";
743 }
744 if (far != 1)
745 printf("%s %s=%d", sep, offset?"offset":"far", far);
746 if (near*far == 1)
747 printf("NO REDUNDANCY");
748}
435d4ebb 749#endif
e0d19036 750
5f8097be
NB
751unsigned long long calc_array_size(int level, int raid_disks, int layout,
752 int chunksize, unsigned long long devsize)
577e8448 753{
e03640bd
CA
754 if (level == 1)
755 return devsize;
577e8448
AK
756 devsize &= ~(unsigned long long)((chunksize>>9)-1);
757 return get_data_disks(level, layout, raid_disks) * devsize;
758}
759
760int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
761{
762 int data_disks = 0;
763 switch (level) {
ca3b6696
N
764 case 0: data_disks = raid_disks;
765 break;
766 case 1: data_disks = 1;
767 break;
5f8097be 768 case 4:
ca3b6696
N
769 case 5: data_disks = raid_disks - 1;
770 break;
771 case 6: data_disks = raid_disks - 2;
772 break;
5f8097be
NB
773 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
774 break;
775 }
577e8448
AK
776
777 return data_disks;
5f8097be
NB
778}
779
0e600426 780#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
e0d19036
NB
781char *get_md_name(int dev)
782{
783 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 784 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
e0d19036
NB
785 static char devname[50];
786 struct stat stb;
98c6faba 787 dev_t rdev;
dd0781e5 788 char *dn;
98c6faba
NB
789
790 if (dev < 0) {
dd0781e5
NB
791 int mdp = get_mdp_major();
792 if (mdp < 0) return NULL;
0df46c2a 793 rdev = makedev(mdp, (-1-dev)<<6);
8f23b0b3 794 snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev);
98c6faba
NB
795 if (stat(devname, &stb) == 0
796 && (S_IFMT&stb.st_mode) == S_IFBLK
797 && (stb.st_rdev == rdev))
798 return devname;
799 } else {
0df46c2a 800 rdev = makedev(MD_MAJOR, dev);
8f23b0b3 801 snprintf(devname, sizeof(devname), "/dev/md%d", dev);
98c6faba
NB
802 if (stat(devname, &stb) == 0
803 && (S_IFMT&stb.st_mode) == S_IFBLK
804 && (stb.st_rdev == rdev))
805 return devname;
806
8f23b0b3 807 snprintf(devname, sizeof(devname), "/dev/md/%d", dev);
98c6faba
NB
808 if (stat(devname, &stb) == 0
809 && (S_IFMT&stb.st_mode) == S_IFBLK
810 && (stb.st_rdev == rdev))
811 return devname;
812 }
16c6fa80 813 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
814 if (dn)
815 return dn;
8f23b0b3 816 snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev);
e0d19036 817 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
818 if (errno != EEXIST)
819 return NULL;
e0d19036
NB
820
821 if (stat(devname, &stb) == 0
822 && (S_IFMT&stb.st_mode) == S_IFBLK
823 && (stb.st_rdev == rdev))
824 return devname;
825 unlink(devname);
826 return NULL;
827}
828
829void put_md_name(char *name)
830{
ca3b6696 831 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
832 unlink(name);
833}
ea24acd0 834
ea24acd0
NB
835int find_free_devnum(int use_partitions)
836{
837 int devnum;
838 for (devnum = 127; devnum != 128;
a56fb7ec 839 devnum = devnum ? devnum-1 : (1<<20)-1) {
ea24acd0 840 int _devnum;
fb52f245 841 char nbuf[50];
ea24acd0
NB
842
843 _devnum = use_partitions ? (-1-devnum) : devnum;
844 if (mddev_busy(_devnum))
845 continue;
fb52f245
N
846 sprintf(nbuf, "%s%d", use_partitions?"mdp":"md", devnum);
847 if (!conf_name_is_free(nbuf))
848 continue;
06d2ffc3
N
849 if (!use_udev()) {
850 /* make sure it is new to /dev too, at least as a
851 * non-standard */
852 char *dn = map_dev(dev2major(_devnum),
853 dev2minor(_devnum), 0);
854 if (dn && ! is_standard(dn, NULL))
855 continue;
856 }
ea24acd0
NB
857 break;
858 }
859 if (devnum == 128)
860 return NoMdDev;
861 return use_partitions ? (-1-devnum) : devnum;
862}
435d4ebb 863#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 864
8b0dabea
NB
865int dev_open(char *dev, int flags)
866{
867 /* like 'open', but if 'dev' matches %d:%d, create a temp
868 * block device and open that
869 */
870 char *e;
871 int fd = -1;
872 char devname[32];
e81cdd9f 873 int major;
8b0dabea 874 int minor;
e81cdd9f
NB
875
876 if (!dev) return -1;
6df6a774 877 flags |= O_DIRECT;
e81cdd9f
NB
878
879 major = strtoul(dev, &e, 0);
8b0dabea
NB
880 if (e > dev && *e == ':' && e[1] &&
881 (minor = strtoul(e+1, &e, 0)) >= 0 &&
882 *e == 0) {
9dc7d357
N
883 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
884 (int)getpid(), major, minor);
885 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
886 fd = open(devname, flags);
887 unlink(devname);
6df6a774
N
888 }
889 if (fd < 0) {
9dc7d357 890 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
891 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
892 (int)getpid(), major, minor);
ca3b6696 893 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
894 fd = open(devname, flags);
895 unlink(devname);
896 }
8b0dabea
NB
897 }
898 } else
6df6a774 899 fd = open(dev, flags);
8b0dabea
NB
900 return fd;
901}
f9ce90ba 902
d998b738 903int open_dev_flags(int devnum, int flags)
e8a70c89
N
904{
905 char buf[20];
906
907 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
d998b738
N
908 return dev_open(buf, flags);
909}
910
911int open_dev(int devnum)
912{
913 return open_dev_flags(devnum, O_RDONLY);
e8a70c89
N
914}
915
a931db9e
NB
916int open_dev_excl(int devnum)
917{
918 char buf[20];
919 int i;
7187750e 920 int flags = O_RDWR;
a931db9e
NB
921
922 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
ca3b6696 923 for (i = 0 ; i < 25 ; i++) {
7187750e 924 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
925 if (fd >= 0)
926 return fd;
7187750e
N
927 if (errno == EACCES && flags == O_RDWR) {
928 flags = O_RDONLY;
929 continue;
930 }
a931db9e
NB
931 if (errno != EBUSY)
932 return fd;
933 usleep(200000);
934 }
935 return -1;
936}
937
9008ed1c
N
938int same_dev(char *one, char *two)
939{
940 struct stat st1, st2;
941 if (stat(one, &st1) != 0)
942 return 0;
943 if (stat(two, &st2) != 0)
944 return 0;
945 if ((st1.st_mode & S_IFMT) != S_IFBLK)
946 return 0;
947 if ((st2.st_mode & S_IFMT) != S_IFBLK)
948 return 0;
949 return st1.st_rdev == st2.st_rdev;
950}
951
a7c6e3fb 952void wait_for(char *dev, int fd)
a714580e
N
953{
954 int i;
a7c6e3fb
N
955 struct stat stb_want;
956
957 if (fstat(fd, &stb_want) != 0 ||
958 (stb_want.st_mode & S_IFMT) != S_IFBLK)
959 return;
a714580e 960
ca3b6696 961 for (i = 0 ; i < 25 ; i++) {
a714580e 962 struct stat stb;
a7c6e3fb
N
963 if (stat(dev, &stb) == 0 &&
964 (stb.st_mode & S_IFMT) == S_IFBLK &&
965 (stb.st_rdev == stb_want.st_rdev))
a714580e
N
966 return;
967 usleep(200000);
968 }
436305c6
DW
969 if (i == 25)
970 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
971}
972
0f22b998
N
973struct superswitch *superlist[] =
974{
975 &super0, &super1,
976 &super_ddf, &super_imsm,
0592faeb 977 &mbr, &gpt,
0f22b998 978 NULL };
f9ce90ba 979
ea24acd0 980#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 981
4725bc31 982struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 983{
1686dc25
NB
984 mdu_array_info_t array;
985 int vers;
986 int minor;
987 struct supertype *st = NULL;
7e0f6979 988 struct mdinfo *sra;
142cb9e1 989 char *verstr;
1686dc25
NB
990 char version[20];
991 int i;
f7e7067b 992 char *subarray = NULL;
5f7e44b2 993 int container = NoMdDev;
1686dc25
NB
994
995 sra = sysfs_read(fd, 0, GET_VERSION);
996
997 if (sra) {
7e0f6979
NB
998 vers = sra->array.major_version;
999 minor = sra->array.minor_version;
142cb9e1 1000 verstr = sra->text_version;
1686dc25
NB
1001 } else {
1002 if (ioctl(fd, GET_ARRAY_INFO, &array))
1003 array.major_version = array.minor_version = 0;
1004 vers = array.major_version;
1005 minor = array.minor_version;
142cb9e1 1006 verstr = "";
6fbba4c9 1007 }
82d9eba6 1008
1686dc25
NB
1009 if (vers != -1) {
1010 sprintf(version, "%d.%d", vers, minor);
1011 verstr = version;
6fbba4c9 1012 }
3c558363 1013 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1014 char *dev = verstr+1;
5f7e44b2 1015
f7e7067b 1016 subarray = strchr(dev, '/');
92d49ecf 1017 if (subarray) {
f7e7067b 1018 *subarray++ = '\0';
503975b9 1019 subarray = xstrdup(subarray);
92d49ecf 1020 }
5f7e44b2 1021 container = devname2devnum(dev);
f7e7067b
NB
1022 if (sra)
1023 sysfs_free(sra);
5f7e44b2 1024 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1025 if (sra && sra->text_version[0])
1026 verstr = sra->text_version;
1027 else
1028 verstr = "-no-metadata-";
f7e7067b
NB
1029 }
1030
1031 for (i = 0; st == NULL && superlist[i] ; i++)
1032 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
1033
1034 if (sra)
1035 sysfs_free(sra);
f7e7067b 1036 if (st) {
3b0896f8 1037 st->sb = NULL;
1f49fb3a
N
1038 if (subarrayp)
1039 *subarrayp = subarray;
5f7e44b2
DW
1040 st->container_dev = container;
1041 st->devnum = fd2devnum(fd);
4725bc31
N
1042 } else
1043 free(subarray);
5f7e44b2 1044
82d9eba6 1045 return st;
f9ce90ba 1046}
ea24acd0
NB
1047#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1048
0f0749ad 1049int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1050{
1051 char buf[20];
1052 int fd;
1053
1054 sprintf(buf, "%d:%d", major(id), minor(id));
1055 fd = dev_open(buf, O_RDONLY);
1056 if (fd < 0)
1057 return 0;
1058 if (get_dev_size(fd, NULL, size)) {
1059 close(fd);
1060 return 1;
1061 }
1062 close(fd);
1063 return 0;
1064}
f9ce90ba 1065
159c3a1a 1066struct supertype *dup_super(struct supertype *orig)
3da92f27 1067{
159c3a1a 1068 struct supertype *st;
1686dc25 1069
d2ca6449
NB
1070 if (!orig)
1071 return orig;
503975b9 1072 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1073 st->ss = orig->ss;
1074 st->max_devs = orig->max_devs;
1075 st->minor_version = orig->minor_version;
cb8f6859 1076 st->ignore_hw_compat = orig->ignore_hw_compat;
159c3a1a
NB
1077 st->sb = NULL;
1078 st->info = NULL;
1079 return st;
3da92f27
NB
1080}
1081
54887ad8 1082struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1083{
1084 /* try each load_super to find the best match,
1085 * and return the best superswitch
1086 */
82d9eba6
NB
1087 struct superswitch *ss;
1088 struct supertype *st;
f21e18ca 1089 time_t besttime = 0;
570c0542 1090 int bestsuper = -1;
f9ce90ba
NB
1091 int i;
1092
503975b9 1093 st = xcalloc(1, sizeof(*st));
d1d599ea
N
1094 st->container_dev = NoMdDev;
1095
ca3b6696 1096 for (i = 0 ; superlist[i]; i++) {
f9ce90ba
NB
1097 int rv;
1098 ss = superlist[i];
54887ad8
N
1099 if (guess_type == guess_array && ss->add_to_super == NULL)
1100 continue;
1101 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1102 continue;
ef609477 1103 memset(st, 0, sizeof(*st));
df3346e6 1104 st->ignore_hw_compat = 1;
3da92f27 1105 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1106 if (rv == 0) {
1107 struct mdinfo info;
a5d85af7 1108 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1109 if (bestsuper == -1 ||
1110 besttime < info.array.ctime) {
1111 bestsuper = i;
1112 besttime = info.array.ctime;
570c0542 1113 }
3da92f27 1114 ss->free_super(st);
570c0542
NB
1115 }
1116 }
1117 if (bestsuper != -1) {
1118 int rv;
ef609477 1119 memset(st, 0, sizeof(*st));
df3346e6 1120 st->ignore_hw_compat = 1;
3da92f27 1121 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1122 if (rv == 0) {
5e747af2 1123 superlist[bestsuper]->free_super(st);
82d9eba6 1124 return st;
f9ce90ba
NB
1125 }
1126 }
570c0542 1127 free(st);
f9ce90ba
NB
1128 return NULL;
1129}
fe6729fa 1130
beae1dfe
NB
1131/* Return size of device in bytes */
1132int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1133{
1134 unsigned long long ldsize;
c2c9bb6f
NB
1135 struct stat st;
1136
1137 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1138 ldsize = (unsigned long long)st.st_size;
1139 else
beae1dfe
NB
1140#ifdef BLKGETSIZE64
1141 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1142#endif
1143 {
1144 unsigned long dsize;
1145 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1146 ldsize = dsize;
1147 ldsize <<= 9;
1148 } else {
1149 if (dname)
e7b84f9d 1150 pr_err("Cannot get size of %s: %s\b",
beae1dfe
NB
1151 dname, strerror(errno));
1152 return 0;
1153 }
1154 }
1155 *sizep = ldsize;
1156 return 1;
1157}
8fac0577 1158
3a371610
N
1159/* Return true if this can only be a container, not a member device.
1160 * i.e. is and md device and size is zero
1161 */
1162int must_be_container(int fd)
1163{
1164 unsigned long long size;
1165 if (md_get_version(fd) < 0)
1166 return 0;
1167 if (get_dev_size(fd, NULL, &size) == 0)
1168 return 1;
1169 if (size == 0)
1170 return 1;
1171 return 0;
1172}
034b203a
TM
1173
1174/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1175 * Returns: 1 if successful
1176 * -1 for unknown partition type
1177 * 0 for other errors
1178 */
1179static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1180{
056b331e 1181 struct GPT gpt;
034b203a 1182 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1183 struct GPT_part_entry *part;
1184 char buf[512];
034b203a
TM
1185 unsigned long long curr_part_end;
1186 unsigned all_partitions, entry_size;
f21e18ca 1187 unsigned part_nr;
034b203a
TM
1188
1189 *endofpart = 0;
1190
056b331e 1191 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1192 /* skip protective MBR */
034b203a 1193 lseek(fd, 512, SEEK_SET);
73e658d8 1194 /* read GPT header */
056b331e 1195 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1196 return 0;
1197
1198 /* get the number of partition entries and the entry size */
056b331e
N
1199 all_partitions = __le32_to_cpu(gpt.part_cnt);
1200 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1201
1202 /* Check GPT signature*/
056b331e 1203 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1204 return -1;
1205
1206 /* sanity checks */
1207 if (all_partitions > 1024 ||
db7fdfe4 1208 entry_size > sizeof(buf))
034b203a
TM
1209 return -1;
1210
db7fdfe4
JS
1211 part = (struct GPT_part_entry *)buf;
1212
ca3b6696 1213 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1214 /* read partition entry */
db7fdfe4 1215 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1216 return 0;
1217
034b203a 1218 /* is this valid partition? */
db7fdfe4 1219 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1220 /* check the last lba for the current partition */
db7fdfe4 1221 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1222 if (curr_part_end > *endofpart)
1223 *endofpart = curr_part_end;
1224 }
1225
034b203a
TM
1226 }
1227 return 1;
1228}
1229
1230/* Sets endofpart parameter to the last block used by the last partition on the device.
1231 * Returns: 1 if successful
1232 * -1 for unknown partition type
1233 * 0 for other errors
1234 */
1235static int get_last_partition_end(int fd, unsigned long long *endofpart)
1236{
056b331e 1237 struct MBR boot_sect;
034b203a
TM
1238 struct MBR_part_record *part;
1239 unsigned long long curr_part_end;
f21e18ca 1240 unsigned part_nr;
034b203a
TM
1241 int retval = 0;
1242
1243 *endofpart = 0;
1244
056b331e 1245 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1246 /* read MBR */
1247 lseek(fd, 0, 0);
056b331e 1248 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1249 goto abort;
1250
1251 /* check MBP signature */
056b331e 1252 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1253 retval = 1;
1254 /* found the correct signature */
056b331e 1255 part = boot_sect.parts;
034b203a 1256
ca3b6696 1257 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
034b203a
TM
1258 /* check for GPT type */
1259 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1260 retval = get_gpt_last_partition_end(fd, endofpart);
1261 break;
1262 }
1263 /* check the last used lba for the current partition */
1264 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1265 __le32_to_cpu(part->blocks_num);
1266 if (curr_part_end > *endofpart)
1267 *endofpart = curr_part_end;
1268
1269 part++;
1270 }
1271 } else {
1272 /* Unknown partition table */
1273 retval = -1;
1274 }
1275 abort:
1276 return retval;
1277}
1278
53ed6ac3
KW
1279int check_partitions(int fd, char *dname, unsigned long long freesize,
1280 unsigned long long size)
034b203a
TM
1281{
1282 /*
1283 * Check where the last partition ends
1284 */
1285 unsigned long long endofpart;
1286 int ret;
1287
1288 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1289 /* There appears to be a partition table here */
1290 if (freesize == 0) {
1291 /* partitions will not be visible in new device */
e7b84f9d
N
1292 pr_err("partition table exists on %s but will be lost or\n"
1293 " meaningless after creating array\n",
1294 dname);
034b203a
TM
1295 return 1;
1296 } else if (endofpart > freesize) {
1297 /* last partition overlaps metadata */
e7b84f9d
N
1298 pr_err("metadata will over-write last partition on %s.\n",
1299 dname);
034b203a 1300 return 1;
53ed6ac3
KW
1301 } else if (size && endofpart > size) {
1302 /* partitions will be truncated in new device */
e7b84f9d
N
1303 pr_err("array size is too small to cover all partitions on %s.\n",
1304 dname);
53ed6ac3 1305 return 1;
034b203a
TM
1306 }
1307 }
1308 return 0;
1309}
1310
a322f70c
DW
1311int open_container(int fd)
1312{
1313 /* 'fd' is a block device. Find out if it is in use
1314 * by a container, and return an open fd on that container.
1315 */
1316 char path[256];
1317 char *e;
1318 DIR *dir;
1319 struct dirent *de;
1320 int dfd, n;
1321 char buf[200];
1322 int major, minor;
1323 struct stat st;
1324
1325 if (fstat(fd, &st) != 0)
1326 return -1;
1327 sprintf(path, "/sys/dev/block/%d:%d/holders",
1328 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1329 e = path + strlen(path);
1330
1331 dir = opendir(path);
1332 if (!dir)
1333 return -1;
1334 while ((de = readdir(dir))) {
1335 if (de->d_ino == 0)
1336 continue;
1337 if (de->d_name[0] == '.')
1338 continue;
a994592d
N
1339 /* Need to make sure it is a container and not a volume */
1340 sprintf(e, "/%s/md/metadata_version", de->d_name);
1341 dfd = open(path, O_RDONLY);
1342 if (dfd < 0)
1343 continue;
1344 n = read(dfd, buf, sizeof(buf));
1345 close(dfd);
1346 if (n <= 0 || (unsigned)n >= sizeof(buf))
1347 continue;
1348 buf[n] = 0;
1349 if (strncmp(buf, "external", 8) != 0 ||
1350 n < 10 ||
1351 buf[9] == '/')
1352 continue;
a322f70c
DW
1353 sprintf(e, "/%s/dev", de->d_name);
1354 dfd = open(path, O_RDONLY);
1355 if (dfd < 0)
1356 continue;
1357 n = read(dfd, buf, sizeof(buf));
1358 close(dfd);
f21e18ca 1359 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1360 continue;
1361 buf[n] = 0;
1362 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1363 continue;
1364 sprintf(buf, "%d:%d", major, minor);
1365 dfd = dev_open(buf, O_RDONLY);
1366 if (dfd >= 0) {
1367 closedir(dir);
1368 return dfd;
1369 }
1370 }
355726fa 1371 closedir(dir);
a322f70c
DW
1372 return -1;
1373}
1374
33414a01
DW
1375struct superswitch *version_to_superswitch(char *vers)
1376{
1377 int i;
1378
1379 for (i = 0; superlist[i]; i++) {
1380 struct superswitch *ss = superlist[i];
1381
1382 if (strcmp(vers, ss->name) == 0)
1383 return ss;
1384 }
1385
1386 return NULL;
1387}
1388
1389int is_container_member(struct mdstat_ent *mdstat, char *container)
1390{
1391 if (mdstat->metadata_version == NULL ||
1392 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
1393 !is_subarray(mdstat->metadata_version+9) ||
1394 strncmp(mdstat->metadata_version+10, container, strlen(container)) != 0 ||
1395 mdstat->metadata_version[10+strlen(container)] != '/')
1396 return 0;
1397
1398 return 1;
1399}
1400
1401int is_subarray_active(char *subarray, char *container)
1402{
1403 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1404 struct mdstat_ent *ent;
1405
e5408a32
DW
1406 for (ent = mdstat; ent; ent = ent->next)
1407 if (is_container_member(ent, container))
e5e5d7ce 1408 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1409 break;
33414a01
DW
1410
1411 free_mdstat(mdstat);
1412
1413 return ent != NULL;
1414}
1415
1416/* open_subarray - opens a subarray in a container
1417 * @dev: container device name
feab51f8 1418 * @st: empty supertype
33414a01
DW
1419 * @quiet: block reporting errors flag
1420 *
1421 * On success returns an fd to a container and fills in *st
1422 */
feab51f8 1423int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1424{
1425 struct mdinfo *mdi;
a951a4f7 1426 struct mdinfo *info;
33414a01
DW
1427 int fd, err = 1;
1428
1429 fd = open(dev, O_RDWR|O_EXCL);
1430 if (fd < 0) {
1431 if (!quiet)
e7b84f9d 1432 pr_err("Couldn't open %s, aborting\n",
33414a01 1433 dev);
b990032d 1434 return -1;
33414a01
DW
1435 }
1436
1437 st->devnum = fd2devnum(fd);
1438 if (st->devnum == NoMdDev) {
1439 if (!quiet)
e7b84f9d
N
1440 pr_err("Failed to determine device number for %s\n",
1441 dev);
33414a01
DW
1442 goto close_fd;
1443 }
1444
1445 mdi = sysfs_read(fd, st->devnum, GET_VERSION|GET_LEVEL);
1446 if (!mdi) {
1447 if (!quiet)
e7b84f9d 1448 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1449 dev);
1450 goto close_fd;
1451 }
1452
1453 if (mdi->array.level != UnSet) {
1454 if (!quiet)
e7b84f9d 1455 pr_err("%s is not a container\n", dev);
33414a01
DW
1456 goto free_sysfs;
1457 }
1458
1459 st->ss = version_to_superswitch(mdi->text_version);
1460 if (!st->ss) {
1461 if (!quiet)
e7b84f9d
N
1462 pr_err("Operation not supported for %s metadata\n",
1463 mdi->text_version);
33414a01
DW
1464 goto free_sysfs;
1465 }
1466
1467 st->devname = devnum2devname(st->devnum);
1468 if (!st->devname) {
1469 if (!quiet)
e7b84f9d 1470 pr_err("Failed to allocate device name\n");
33414a01
DW
1471 goto free_sysfs;
1472 }
1473
db20d413 1474 if (!st->ss->load_container) {
33414a01 1475 if (!quiet)
e7b84f9d 1476 pr_err("%s is not a container\n", dev);
33414a01
DW
1477 goto free_name;
1478 }
1479
db20d413 1480 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1481 if (!quiet)
e7b84f9d 1482 pr_err("Failed to load metadata for %s\n",
db20d413
N
1483 dev);
1484 goto free_name;
33414a01
DW
1485 }
1486
a951a4f7
N
1487 info = st->ss->container_content(st, subarray);
1488 if (!info) {
1489 if (!quiet)
e7b84f9d 1490 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1491 subarray, dev);
1492 goto free_super;
1493 }
1494 free(info);
1495
33414a01
DW
1496 err = 0;
1497
1498 free_super:
1499 if (err)
1500 st->ss->free_super(st);
1501 free_name:
1502 if (err)
1503 free(st->devname);
1504 free_sysfs:
1505 sysfs_free(mdi);
1506 close_fd:
1507 if (err)
1508 close(fd);
1509
1510 if (err)
1511 return -1;
1512 else
1513 return fd;
1514}
1515
7801ac20
N
1516int add_disk(int mdfd, struct supertype *st,
1517 struct mdinfo *sra, struct mdinfo *info)
1518{
1519 /* Add a device to an array, in one of 2 ways. */
1520 int rv;
1521#ifndef MDASSEMBLE
1522 if (st->ss->external) {
d23534e4
DW
1523 if (info->disk.state & (1<<MD_DISK_SYNC))
1524 info->recovery_start = MaxSector;
1525 else
1526 info->recovery_start = 0;
2904b26f 1527 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1528 if (! rv) {
1529 struct mdinfo *sd2;
f35f2525
N
1530 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1531 if (sd2 == info)
1532 break;
1533 if (sd2 == NULL) {
503975b9 1534 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1535 *sd2 = *info;
1536 sd2->next = sra->devs;
1537 sra->devs = sd2;
1538 }
7801ac20
N
1539 }
1540 } else
1541#endif
1542 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1543 return rv;
1544}
1545
de6ae750
N
1546int remove_disk(int mdfd, struct supertype *st,
1547 struct mdinfo *sra, struct mdinfo *info)
1548{
1549 int rv;
1550 /* Remove the disk given by 'info' from the array */
1551#ifndef MDASSEMBLE
1552 if (st->ss->external)
1553 rv = sysfs_set_str(sra, info, "slot", "none");
1554 else
1555#endif
1556 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1557 info->disk.minor));
1558 return rv;
1559}
1560
f35f2525
N
1561int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1562{
1563 /* Initialise kernel's knowledge of array.
1564 * This varies between externally managed arrays
1565 * and older kernels
1566 */
1567 int vers = md_get_version(mdfd);
1568 int rv;
1569
1570#ifndef MDASSEMBLE
1571 if (st->ss->external)
1572 rv = sysfs_set_array(info, vers);
1573 else
1574#endif
1575 if ((vers % 100) >= 1) { /* can use different versions */
1576 mdu_array_info_t inf;
1577 memset(&inf, 0, sizeof(inf));
1578 inf.major_version = info->array.major_version;
1579 inf.minor_version = info->array.minor_version;
1580 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1581 } else
1582 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1583 return rv;
1584}
1585
1e5c6983
DW
1586unsigned long long min_recovery_start(struct mdinfo *array)
1587{
1588 /* find the minimum recovery_start in an array for metadata
1589 * formats that only record per-array recovery progress instead
1590 * of per-device
1591 */
1592 unsigned long long recovery_start = MaxSector;
1593 struct mdinfo *d;
1594
1595 for (d = array->devs; d; d = d->next)
1596 recovery_start = min(recovery_start, d->recovery_start);
1597
1598 return recovery_start;
1599}
1600
24f6f99b 1601int mdmon_pid(int devnum)
a931db9e
NB
1602{
1603 char path[100];
1604 char pid[10];
1605 int fd;
1606 int n;
10013317
PHC
1607 char *devname = devnum2devname(devnum);
1608
753cf905 1609 sprintf(path, "%s/%s.pid", MDMON_DIR, devname);
10013317
PHC
1610 free(devname);
1611
24f6f99b 1612 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1613
1614 if (fd < 0)
cf556303 1615 return -1;
a931db9e
NB
1616 n = read(fd, pid, 9);
1617 close(fd);
1618 if (n <= 0)
cf556303 1619 return -1;
24f6f99b 1620 return atoi(pid);
a931db9e
NB
1621}
1622
24f6f99b 1623int mdmon_running(int devnum)
a931db9e 1624{
24f6f99b
N
1625 int pid = mdmon_pid(devnum);
1626 if (pid <= 0)
a931db9e 1627 return 0;
24f6f99b 1628 if (kill(pid, 0) == 0)
a931db9e
NB
1629 return 1;
1630 return 0;
1631}
1632
8850ee3e
N
1633int start_mdmon(int devnum)
1634{
1913c325 1635 int i, skipped;
44d2e365 1636 int len;
ca3b6696 1637 pid_t pid;
9fe32043 1638 int status;
44d2e365
N
1639 char pathbuf[1024];
1640 char *paths[4] = {
1641 pathbuf,
1642 "/sbin/mdmon",
1643 "mdmon",
1644 NULL
1645 };
8850ee3e 1646
40ebbb9c 1647 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1648 return 0;
1649
9cf014ec 1650 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1651 if (len > 0) {
1652 char *sl;
1653 pathbuf[len] = 0;
1654 sl = strrchr(pathbuf, '/');
1655 if (sl)
1656 sl++;
1657 else
1658 sl = pathbuf;
1659 strcpy(sl, "mdmon");
1660 } else
1661 pathbuf[0] = '\0';
1662
0f7bdf89
JS
1663 /* First try to run systemctl */
1664 switch(fork()) {
1665 case 0:
1666 /* FIXME yuk. CLOSE_EXEC?? */
1667 skipped = 0;
1668 for (i = 3; skipped < 20; i++)
1669 if (close(i) < 0)
1670 skipped++;
1671 else
1672 skipped = 0;
1673
1674 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1675 devnum2devname(devnum));
1676 status = execl("/usr/bin/systemctl", "systemctl", "start",
1677 pathbuf, NULL);
1678 status = execl("/bin/systemctl", "systemctl", "start",
1679 pathbuf, NULL);
1680 exit(1);
1681 case -1: pr_err("cannot run mdmon. "
1682 "Array remains readonly\n");
1683 return -1;
1684 default: /* parent - good */
1685 pid = wait(&status);
1686 if (pid >= 0 && status == 0)
1687 return 0;
1688 }
1689
1690 /* That failed, try running mdmon directly */
8850ee3e
N
1691 switch(fork()) {
1692 case 0:
1693 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1694 skipped = 0;
ca3b6696 1695 for (i = 3; skipped < 20; i++)
1913c325
N
1696 if (close(i) < 0)
1697 skipped++;
1698 else
1699 skipped = 0;
1700
ca3b6696 1701 for (i = 0; paths[i]; i++)
a0963a86 1702 if (paths[i][0]) {
3e23ba9d
JS
1703 execl(paths[i], "mdmon",
1704 devnum2devname(devnum), NULL);
a0963a86 1705 }
8850ee3e 1706 exit(1);
e7b84f9d 1707 case -1: pr_err("cannot run mdmon. "
8850ee3e
N
1708 "Array remains readonly\n");
1709 return -1;
9fe32043
N
1710 default: /* parent - good */
1711 pid = wait(&status);
15c10423
JS
1712 if (pid < 0 || status != 0) {
1713 pr_err("failed to launch mdmon. "
1714 "Array remains readonly\n");
9fe32043 1715 return -1;
15c10423 1716 }
8850ee3e
N
1717 }
1718 return 0;
1719}
1720
40ebbb9c 1721int check_env(char *name)
5dcfcb71 1722{
40ebbb9c 1723 char *val = getenv(name);
5dcfcb71
DW
1724
1725 if (val && atoi(val) == 1)
1726 return 1;
1727
1728 return 0;
1729}
1730
148acb7b
DW
1731__u32 random32(void)
1732{
1733 __u32 rv;
1734 int rfd = open("/dev/urandom", O_RDONLY);
1735 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1736 rv = random();
1737 if (rfd >= 0)
1738 close(rfd);
1739 return rv;
1740}
1741
0e600426 1742#ifndef MDASSEMBLE
edd8d13c
NB
1743int flush_metadata_updates(struct supertype *st)
1744{
1745 int sfd;
1746 if (!st->updates) {
1747 st->update_tail = NULL;
1748 return -1;
1749 }
1750
1751 sfd = connect_monitor(devnum2devname(st->container_dev));
1752 if (sfd < 0)
1753 return -1;
1754
1755 while (st->updates) {
1756 struct metadata_update *mu = st->updates;
1757 st->updates = mu->next;
1758
1759 send_message(sfd, mu, 0);
1760 wait_reply(sfd, 0);
1761 free(mu->buf);
1762 free(mu);
1763 }
1764 ack(sfd, 0);
1765 wait_reply(sfd, 0);
1766 close(sfd);
1767 st->update_tail = NULL;
1768 return 0;
1769}
1770
1771void append_metadata_update(struct supertype *st, void *buf, int len)
1772{
1773
503975b9 1774 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
1775
1776 mu->buf = buf;
1777 mu->len = len;
1778 mu->space = NULL;
cb23f1f4 1779 mu->space_list = NULL;
edd8d13c
NB
1780 mu->next = NULL;
1781 *st->update_tail = mu;
1782 st->update_tail = &mu->next;
1783}
0e600426 1784#endif /* MDASSEMBLE */
a931db9e 1785
fe6729fa
NB
1786#ifdef __TINYC__
1787/* tinyc doesn't optimize this check in ioctl.h out ... */
1788unsigned int __invalid_size_argument_for_IOC = 0;
1789#endif
1790
6d11ec6f
AK
1791int experimental(void)
1792{
1793 if (check_env("MDADM_EXPERIMENTAL"))
1794 return 1;
1795 else {
e7b84f9d 1796 pr_err("To use this feature MDADM_EXPERIMENTAL"
65c83a80 1797 " environment variable has to be defined.\n");
6d11ec6f
AK
1798 return 0;
1799 }
1800}
1801
326727d9
AC
1802/* Pick all spares matching given criteria from a container
1803 * if min_size == 0 do not check size
1804 * if domlist == NULL do not check domains
1805 * if spare_group given add it to domains of each spare
1806 * metadata allows to test domains using metadata of destination array */
1807struct mdinfo *container_choose_spares(struct supertype *st,
1808 unsigned long long min_size,
1809 struct domainlist *domlist,
1810 char *spare_group,
1811 const char *metadata, int get_one)
1812{
1813 struct mdinfo *d, **dp, *disks = NULL;
1814
1815 /* get list of all disks in container */
1816 if (st->ss->getinfo_super_disks)
1817 disks = st->ss->getinfo_super_disks(st);
1818
1819 if (!disks)
1820 return disks;
1821 /* find spare devices on the list */
1822 dp = &disks->devs;
1823 disks->array.spare_disks = 0;
1824 while (*dp) {
1825 int found = 0;
1826 d = *dp;
1827 if (d->disk.state == 0) {
1828 /* check if size is acceptable */
1829 unsigned long long dev_size;
1830 dev_t dev = makedev(d->disk.major,d->disk.minor);
1831
1832 if (!min_size ||
1833 (dev_size_from_id(dev, &dev_size) &&
1834 dev_size >= min_size))
1835 found = 1;
1836 /* check if domain matches */
1837 if (found && domlist) {
1838 struct dev_policy *pol = devnum_policy(dev);
1839 if (spare_group)
1840 pol_add(&pol, pol_domain,
1841 spare_group, NULL);
e5508b36 1842 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1843 found = 0;
1844 dev_policy_free(pol);
1845 }
1846 }
1847 if (found) {
1848 dp = &d->next;
1849 disks->array.spare_disks++;
1850 if (get_one) {
1851 sysfs_free(*dp);
1852 d->next = NULL;
1853 }
1854 } else {
1855 *dp = d->next;
1856 d->next = NULL;
1857 sysfs_free(d);
1858 }
1859 }
1860 return disks;
1861}
9eafa1de
MN
1862
1863/* Checks if paths point to the same device
1864 * Returns 0 if they do.
1865 * Returns 1 if they don't.
1866 * Returns -1 if something went wrong,
1867 * e.g. paths are empty or the files
1868 * they point to don't exist */
1869int compare_paths (char* path1, char* path2)
1870{
1871 struct stat st1,st2;
1872
1873 if (path1 == NULL || path2 == NULL)
1874 return -1;
1875 if (stat(path1,&st1) != 0)
1876 return -1;
1877 if (stat(path2,&st2) != 0)
1878 return -1;
1879 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
1880 return 0;
1881 return 1;
1882}