Incremental: allow --quiet to silence from errors from "-If"
[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 523 name);
d33f1518 524 cont_err("size=%dK mtime=%s",
682c7051
NB
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;
d33f1518 548 cont_err("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-";
d33f1518
N
571 cont_err("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
4dd2df09
N
780
781int devnm2devid(char *devnm)
782{
783 /* First look in /sys/block/$DEVNM/dev for %d:%d
784 * If that fails, try parsing out a number
785 */
786 char path[100];
787 char *ep;
788 int fd;
789 int mjr,mnr;
790
791 sprintf(path, "/sys/block/%s/dev", devnm);
792 fd = open(path, O_RDONLY);
793 if (fd >= 0) {
794 char buf[20];
795 int n = read(fd, buf, sizeof(buf));
796 close(fd);
797 if (n > 0)
798 buf[n] = 0;
799 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
800 return makedev(mjr, mnr);
801 }
802 if (strncmp(devnm, "md_d", 4) == 0 &&
803 isdigit(devnm[4]) &&
804 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
805 ep > devnm && *ep == 0)
806 return makedev(get_mdp_major(), mnr << MdpMinorShift);
807
808 if (strncmp(devnm, "md", 2) == 0 &&
809 isdigit(devnm[2]) &&
810 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
811 ep > devnm && *ep == 0)
812 return makedev(MD_MAJOR, mnr);
813
814 return 0;
815}
816
5a23a06e 817#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
4dd2df09 818char *get_md_name(char *devnm)
e0d19036
NB
819{
820 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 821 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 822
e0d19036
NB
823 static char devname[50];
824 struct stat stb;
4dd2df09 825 dev_t rdev = devnm2devid(devnm);
dd0781e5 826 char *dn;
98c6faba 827
4dd2df09
N
828 if (rdev == 0)
829 return 0;
830 if (strncmp(devnm, "md_", 3) == 0) {
831 snprintf(devname, sizeof(devname), "/dev/md/%s",
832 devnm + 3);
98c6faba
NB
833 if (stat(devname, &stb) == 0
834 && (S_IFMT&stb.st_mode) == S_IFBLK
835 && (stb.st_rdev == rdev))
836 return devname;
837 }
4dd2df09
N
838 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
839 if (stat(devname, &stb) == 0
840 && (S_IFMT&stb.st_mode) == S_IFBLK
841 && (stb.st_rdev == rdev))
842 return devname;
843
844 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
845 if (stat(devname, &stb) == 0
846 && (S_IFMT&stb.st_mode) == S_IFBLK
847 && (stb.st_rdev == rdev))
848 return devname;
849
16c6fa80 850 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
851 if (dn)
852 return dn;
4dd2df09 853 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 854 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
855 if (errno != EEXIST)
856 return NULL;
e0d19036
NB
857
858 if (stat(devname, &stb) == 0
859 && (S_IFMT&stb.st_mode) == S_IFBLK
860 && (stb.st_rdev == rdev))
861 return devname;
862 unlink(devname);
863 return NULL;
864}
865
866void put_md_name(char *name)
867{
ca3b6696 868 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
869 unlink(name);
870}
ea24acd0 871
4dd2df09 872char *find_free_devnm(int use_partitions)
ea24acd0 873{
4dd2df09 874 static char devnm[32];
ea24acd0
NB
875 int devnum;
876 for (devnum = 127; devnum != 128;
a56fb7ec 877 devnum = devnum ? devnum-1 : (1<<20)-1) {
ea24acd0 878
4dd2df09
N
879 if (use_partitions)
880 sprintf(devnm, "md_d%d", devnum);
881 else
882 sprintf(devnm, "md%d", devnum);
883 if (mddev_busy(devnm))
ea24acd0 884 continue;
4dd2df09 885 if (!conf_name_is_free(devnm))
fb52f245 886 continue;
06d2ffc3
N
887 if (!use_udev()) {
888 /* make sure it is new to /dev too, at least as a
889 * non-standard */
4dd2df09
N
890 int devid = devnm2devid(devnm);
891 if (devid) {
892 char *dn = map_dev(major(devid),
893 minor(devid), 0);
894 if (dn && ! is_standard(dn, NULL))
895 continue;
896 }
06d2ffc3 897 }
ea24acd0
NB
898 break;
899 }
900 if (devnum == 128)
4dd2df09
N
901 return NULL;
902 return devnm;
ea24acd0 903}
435d4ebb 904#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 905
8b0dabea
NB
906int dev_open(char *dev, int flags)
907{
908 /* like 'open', but if 'dev' matches %d:%d, create a temp
909 * block device and open that
910 */
911 char *e;
912 int fd = -1;
913 char devname[32];
e81cdd9f 914 int major;
8b0dabea 915 int minor;
e81cdd9f
NB
916
917 if (!dev) return -1;
6df6a774 918 flags |= O_DIRECT;
e81cdd9f
NB
919
920 major = strtoul(dev, &e, 0);
8b0dabea
NB
921 if (e > dev && *e == ':' && e[1] &&
922 (minor = strtoul(e+1, &e, 0)) >= 0 &&
923 *e == 0) {
9dc7d357
N
924 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
925 (int)getpid(), major, minor);
926 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
927 fd = open(devname, flags);
928 unlink(devname);
6df6a774
N
929 }
930 if (fd < 0) {
9dc7d357 931 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
932 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
933 (int)getpid(), major, minor);
ca3b6696 934 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
935 fd = open(devname, flags);
936 unlink(devname);
937 }
8b0dabea
NB
938 }
939 } else
6df6a774 940 fd = open(dev, flags);
8b0dabea
NB
941 return fd;
942}
f9ce90ba 943
4dd2df09 944int open_dev_flags(char *devnm, int flags)
e8a70c89 945{
4dd2df09 946 int devid;
e8a70c89
N
947 char buf[20];
948
4dd2df09
N
949 devid = devnm2devid(devnm);
950 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
951 return dev_open(buf, flags);
952}
953
4dd2df09 954int open_dev(char *devnm)
d998b738 955{
4dd2df09 956 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
957}
958
4dd2df09 959int open_dev_excl(char *devnm)
a931db9e
NB
960{
961 char buf[20];
962 int i;
7187750e 963 int flags = O_RDWR;
4dd2df09 964 int devid = devnm2devid(devnm);
a931db9e 965
4dd2df09 966 sprintf(buf, "%d:%d", major(devid), minor(devid));
ca3b6696 967 for (i = 0 ; i < 25 ; i++) {
7187750e 968 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
969 if (fd >= 0)
970 return fd;
7187750e
N
971 if (errno == EACCES && flags == O_RDWR) {
972 flags = O_RDONLY;
973 continue;
974 }
a931db9e
NB
975 if (errno != EBUSY)
976 return fd;
977 usleep(200000);
978 }
979 return -1;
980}
981
9008ed1c
N
982int same_dev(char *one, char *two)
983{
984 struct stat st1, st2;
985 if (stat(one, &st1) != 0)
986 return 0;
987 if (stat(two, &st2) != 0)
988 return 0;
989 if ((st1.st_mode & S_IFMT) != S_IFBLK)
990 return 0;
991 if ((st2.st_mode & S_IFMT) != S_IFBLK)
992 return 0;
993 return st1.st_rdev == st2.st_rdev;
994}
995
a7c6e3fb 996void wait_for(char *dev, int fd)
a714580e
N
997{
998 int i;
a7c6e3fb
N
999 struct stat stb_want;
1000
1001 if (fstat(fd, &stb_want) != 0 ||
1002 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1003 return;
a714580e 1004
ca3b6696 1005 for (i = 0 ; i < 25 ; i++) {
a714580e 1006 struct stat stb;
a7c6e3fb
N
1007 if (stat(dev, &stb) == 0 &&
1008 (stb.st_mode & S_IFMT) == S_IFBLK &&
1009 (stb.st_rdev == stb_want.st_rdev))
a714580e
N
1010 return;
1011 usleep(200000);
1012 }
436305c6
DW
1013 if (i == 25)
1014 dprintf("%s: timeout waiting for %s\n", __func__, dev);
a714580e
N
1015}
1016
0f22b998
N
1017struct superswitch *superlist[] =
1018{
1019 &super0, &super1,
1020 &super_ddf, &super_imsm,
0592faeb 1021 &mbr, &gpt,
0f22b998 1022 NULL };
f9ce90ba 1023
ea24acd0 1024#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 1025
4725bc31 1026struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1027{
1686dc25
NB
1028 mdu_array_info_t array;
1029 int vers;
1030 int minor;
1031 struct supertype *st = NULL;
7e0f6979 1032 struct mdinfo *sra;
142cb9e1 1033 char *verstr;
1686dc25
NB
1034 char version[20];
1035 int i;
f7e7067b 1036 char *subarray = NULL;
4dd2df09 1037 char container[32] = "";
1686dc25 1038
4dd2df09 1039 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1040
1041 if (sra) {
7e0f6979
NB
1042 vers = sra->array.major_version;
1043 minor = sra->array.minor_version;
142cb9e1 1044 verstr = sra->text_version;
1686dc25
NB
1045 } else {
1046 if (ioctl(fd, GET_ARRAY_INFO, &array))
1047 array.major_version = array.minor_version = 0;
1048 vers = array.major_version;
1049 minor = array.minor_version;
142cb9e1 1050 verstr = "";
6fbba4c9 1051 }
82d9eba6 1052
1686dc25
NB
1053 if (vers != -1) {
1054 sprintf(version, "%d.%d", vers, minor);
1055 verstr = version;
6fbba4c9 1056 }
3c558363 1057 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1058 char *dev = verstr+1;
5f7e44b2 1059
f7e7067b 1060 subarray = strchr(dev, '/');
92d49ecf 1061 if (subarray) {
f7e7067b 1062 *subarray++ = '\0';
503975b9 1063 subarray = xstrdup(subarray);
92d49ecf 1064 }
4dd2df09 1065 strcpy(container, dev);
f7e7067b
NB
1066 if (sra)
1067 sysfs_free(sra);
5f7e44b2 1068 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1069 if (sra && sra->text_version[0])
1070 verstr = sra->text_version;
1071 else
1072 verstr = "-no-metadata-";
f7e7067b
NB
1073 }
1074
1075 for (i = 0; st == NULL && superlist[i] ; i++)
1076 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
1077
1078 if (sra)
1079 sysfs_free(sra);
f7e7067b 1080 if (st) {
3b0896f8 1081 st->sb = NULL;
1f49fb3a
N
1082 if (subarrayp)
1083 *subarrayp = subarray;
4dd2df09
N
1084 strcpy(st->container_devnm, container);
1085 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1086 } else
1087 free(subarray);
5f7e44b2 1088
82d9eba6 1089 return st;
f9ce90ba 1090}
ea24acd0
NB
1091#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
1092
0f0749ad 1093int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1094{
1095 char buf[20];
1096 int fd;
1097
1098 sprintf(buf, "%d:%d", major(id), minor(id));
1099 fd = dev_open(buf, O_RDONLY);
1100 if (fd < 0)
1101 return 0;
1102 if (get_dev_size(fd, NULL, size)) {
1103 close(fd);
1104 return 1;
1105 }
1106 close(fd);
1107 return 0;
1108}
f9ce90ba 1109
159c3a1a 1110struct supertype *dup_super(struct supertype *orig)
3da92f27 1111{
159c3a1a 1112 struct supertype *st;
1686dc25 1113
d2ca6449
NB
1114 if (!orig)
1115 return orig;
503975b9 1116 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1117 st->ss = orig->ss;
1118 st->max_devs = orig->max_devs;
1119 st->minor_version = orig->minor_version;
cb8f6859 1120 st->ignore_hw_compat = orig->ignore_hw_compat;
159c3a1a
NB
1121 st->sb = NULL;
1122 st->info = NULL;
1123 return st;
3da92f27
NB
1124}
1125
54887ad8 1126struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1127{
1128 /* try each load_super to find the best match,
1129 * and return the best superswitch
1130 */
82d9eba6
NB
1131 struct superswitch *ss;
1132 struct supertype *st;
f21e18ca 1133 time_t besttime = 0;
570c0542 1134 int bestsuper = -1;
f9ce90ba
NB
1135 int i;
1136
503975b9 1137 st = xcalloc(1, sizeof(*st));
4dd2df09 1138 st->container_devnm[0] = 0;
d1d599ea 1139
ca3b6696 1140 for (i = 0 ; superlist[i]; i++) {
f9ce90ba
NB
1141 int rv;
1142 ss = superlist[i];
54887ad8
N
1143 if (guess_type == guess_array && ss->add_to_super == NULL)
1144 continue;
1145 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1146 continue;
ef609477 1147 memset(st, 0, sizeof(*st));
df3346e6 1148 st->ignore_hw_compat = 1;
3da92f27 1149 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1150 if (rv == 0) {
1151 struct mdinfo info;
a5d85af7 1152 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1153 if (bestsuper == -1 ||
1154 besttime < info.array.ctime) {
1155 bestsuper = i;
1156 besttime = info.array.ctime;
570c0542 1157 }
3da92f27 1158 ss->free_super(st);
570c0542
NB
1159 }
1160 }
1161 if (bestsuper != -1) {
1162 int rv;
ef609477 1163 memset(st, 0, sizeof(*st));
df3346e6 1164 st->ignore_hw_compat = 1;
3da92f27 1165 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1166 if (rv == 0) {
5e747af2 1167 superlist[bestsuper]->free_super(st);
82d9eba6 1168 return st;
f9ce90ba
NB
1169 }
1170 }
570c0542 1171 free(st);
f9ce90ba
NB
1172 return NULL;
1173}
fe6729fa 1174
beae1dfe
NB
1175/* Return size of device in bytes */
1176int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1177{
1178 unsigned long long ldsize;
c2c9bb6f
NB
1179 struct stat st;
1180
1181 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1182 ldsize = (unsigned long long)st.st_size;
1183 else
beae1dfe
NB
1184#ifdef BLKGETSIZE64
1185 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1186#endif
1187 {
1188 unsigned long dsize;
1189 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1190 ldsize = dsize;
1191 ldsize <<= 9;
1192 } else {
1193 if (dname)
e7b84f9d 1194 pr_err("Cannot get size of %s: %s\b",
beae1dfe
NB
1195 dname, strerror(errno));
1196 return 0;
1197 }
1198 }
1199 *sizep = ldsize;
1200 return 1;
1201}
8fac0577 1202
3a371610
N
1203/* Return true if this can only be a container, not a member device.
1204 * i.e. is and md device and size is zero
1205 */
1206int must_be_container(int fd)
1207{
1208 unsigned long long size;
1209 if (md_get_version(fd) < 0)
1210 return 0;
1211 if (get_dev_size(fd, NULL, &size) == 0)
1212 return 1;
1213 if (size == 0)
1214 return 1;
1215 return 0;
1216}
034b203a
TM
1217
1218/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1219 * Returns: 1 if successful
1220 * -1 for unknown partition type
1221 * 0 for other errors
1222 */
1223static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1224{
056b331e 1225 struct GPT gpt;
034b203a 1226 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1227 struct GPT_part_entry *part;
1228 char buf[512];
034b203a
TM
1229 unsigned long long curr_part_end;
1230 unsigned all_partitions, entry_size;
f21e18ca 1231 unsigned part_nr;
034b203a
TM
1232
1233 *endofpart = 0;
1234
056b331e 1235 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1236 /* skip protective MBR */
034b203a 1237 lseek(fd, 512, SEEK_SET);
73e658d8 1238 /* read GPT header */
056b331e 1239 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1240 return 0;
1241
1242 /* get the number of partition entries and the entry size */
056b331e
N
1243 all_partitions = __le32_to_cpu(gpt.part_cnt);
1244 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1245
1246 /* Check GPT signature*/
056b331e 1247 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1248 return -1;
1249
1250 /* sanity checks */
1251 if (all_partitions > 1024 ||
db7fdfe4 1252 entry_size > sizeof(buf))
034b203a
TM
1253 return -1;
1254
db7fdfe4
JS
1255 part = (struct GPT_part_entry *)buf;
1256
ca3b6696 1257 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1258 /* read partition entry */
db7fdfe4 1259 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1260 return 0;
1261
034b203a 1262 /* is this valid partition? */
db7fdfe4 1263 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1264 /* check the last lba for the current partition */
db7fdfe4 1265 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1266 if (curr_part_end > *endofpart)
1267 *endofpart = curr_part_end;
1268 }
1269
034b203a
TM
1270 }
1271 return 1;
1272}
1273
1274/* Sets endofpart parameter to the last block used by the last partition on the device.
1275 * Returns: 1 if successful
1276 * -1 for unknown partition type
1277 * 0 for other errors
1278 */
1279static int get_last_partition_end(int fd, unsigned long long *endofpart)
1280{
056b331e 1281 struct MBR boot_sect;
034b203a
TM
1282 struct MBR_part_record *part;
1283 unsigned long long curr_part_end;
f21e18ca 1284 unsigned part_nr;
034b203a
TM
1285 int retval = 0;
1286
1287 *endofpart = 0;
1288
056b331e 1289 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1290 /* read MBR */
1291 lseek(fd, 0, 0);
056b331e 1292 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1293 goto abort;
1294
1295 /* check MBP signature */
056b331e 1296 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1297 retval = 1;
1298 /* found the correct signature */
056b331e 1299 part = boot_sect.parts;
034b203a 1300
ca3b6696 1301 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
034b203a
TM
1302 /* check for GPT type */
1303 if (part->part_type == MBR_GPT_PARTITION_TYPE) {
1304 retval = get_gpt_last_partition_end(fd, endofpart);
1305 break;
1306 }
1307 /* check the last used lba for the current partition */
1308 curr_part_end = __le32_to_cpu(part->first_sect_lba) +
1309 __le32_to_cpu(part->blocks_num);
1310 if (curr_part_end > *endofpart)
1311 *endofpart = curr_part_end;
1312
1313 part++;
1314 }
1315 } else {
1316 /* Unknown partition table */
1317 retval = -1;
1318 }
1319 abort:
1320 return retval;
1321}
1322
53ed6ac3
KW
1323int check_partitions(int fd, char *dname, unsigned long long freesize,
1324 unsigned long long size)
034b203a
TM
1325{
1326 /*
1327 * Check where the last partition ends
1328 */
1329 unsigned long long endofpart;
1330 int ret;
1331
1332 if ((ret = get_last_partition_end(fd, &endofpart)) > 0) {
1333 /* There appears to be a partition table here */
1334 if (freesize == 0) {
1335 /* partitions will not be visible in new device */
e7b84f9d
N
1336 pr_err("partition table exists on %s but will be lost or\n"
1337 " meaningless after creating array\n",
1338 dname);
034b203a
TM
1339 return 1;
1340 } else if (endofpart > freesize) {
1341 /* last partition overlaps metadata */
e7b84f9d
N
1342 pr_err("metadata will over-write last partition on %s.\n",
1343 dname);
034b203a 1344 return 1;
53ed6ac3
KW
1345 } else if (size && endofpart > size) {
1346 /* partitions will be truncated in new device */
e7b84f9d
N
1347 pr_err("array size is too small to cover all partitions on %s.\n",
1348 dname);
53ed6ac3 1349 return 1;
034b203a
TM
1350 }
1351 }
1352 return 0;
1353}
1354
a322f70c
DW
1355int open_container(int fd)
1356{
1357 /* 'fd' is a block device. Find out if it is in use
1358 * by a container, and return an open fd on that container.
1359 */
1360 char path[256];
1361 char *e;
1362 DIR *dir;
1363 struct dirent *de;
1364 int dfd, n;
1365 char buf[200];
1366 int major, minor;
1367 struct stat st;
1368
1369 if (fstat(fd, &st) != 0)
1370 return -1;
1371 sprintf(path, "/sys/dev/block/%d:%d/holders",
1372 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1373 e = path + strlen(path);
1374
1375 dir = opendir(path);
1376 if (!dir)
1377 return -1;
1378 while ((de = readdir(dir))) {
1379 if (de->d_ino == 0)
1380 continue;
1381 if (de->d_name[0] == '.')
1382 continue;
a994592d
N
1383 /* Need to make sure it is a container and not a volume */
1384 sprintf(e, "/%s/md/metadata_version", de->d_name);
1385 dfd = open(path, O_RDONLY);
1386 if (dfd < 0)
1387 continue;
1388 n = read(dfd, buf, sizeof(buf));
1389 close(dfd);
1390 if (n <= 0 || (unsigned)n >= sizeof(buf))
1391 continue;
1392 buf[n] = 0;
1393 if (strncmp(buf, "external", 8) != 0 ||
1394 n < 10 ||
1395 buf[9] == '/')
1396 continue;
a322f70c
DW
1397 sprintf(e, "/%s/dev", de->d_name);
1398 dfd = open(path, O_RDONLY);
1399 if (dfd < 0)
1400 continue;
1401 n = read(dfd, buf, sizeof(buf));
1402 close(dfd);
f21e18ca 1403 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1404 continue;
1405 buf[n] = 0;
1406 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1407 continue;
1408 sprintf(buf, "%d:%d", major, minor);
1409 dfd = dev_open(buf, O_RDONLY);
1410 if (dfd >= 0) {
1411 closedir(dir);
1412 return dfd;
1413 }
1414 }
355726fa 1415 closedir(dir);
a322f70c
DW
1416 return -1;
1417}
1418
33414a01
DW
1419struct superswitch *version_to_superswitch(char *vers)
1420{
1421 int i;
1422
1423 for (i = 0; superlist[i]; i++) {
1424 struct superswitch *ss = superlist[i];
1425
1426 if (strcmp(vers, ss->name) == 0)
1427 return ss;
1428 }
1429
1430 return NULL;
1431}
1432
4dd2df09
N
1433int metadata_container_matches(char *metadata, char *devnm)
1434{
1435 /* Check if 'devnm' is the container named in 'metadata'
1436 * which is
1437 * /containername/componentname or
1438 * -containername/componentname
1439 */
1440 int l;
1441 if (*metadata != '/' && *metadata != '-')
1442 return 0;
1443 l = strlen(devnm);
1444 if (strncmp(metadata+1, devnm, l) != 0)
1445 return 0;
1446 if (metadata[l+1] != '/')
1447 return 0;
1448 return 1;
1449}
1450
1451int metadata_subdev_matches(char *metadata, char *devnm)
1452{
1453 /* Check if 'devnm' is the subdev named in 'metadata'
1454 * which is
1455 * /containername/subdev or
1456 * -containername/subdev
1457 */
1458 char *sl;
1459 if (*metadata != '/' && *metadata != '-')
1460 return 0;
1461 sl = strchr(metadata+1, '/');
1462 if (!sl)
1463 return 0;
1464 if (strcmp(sl+1, devnm) == 0)
1465 return 1;
1466 return 0;
1467}
1468
33414a01
DW
1469int is_container_member(struct mdstat_ent *mdstat, char *container)
1470{
1471 if (mdstat->metadata_version == NULL ||
1472 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1473 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1474 return 0;
1475
1476 return 1;
1477}
1478
1479int is_subarray_active(char *subarray, char *container)
1480{
1481 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1482 struct mdstat_ent *ent;
1483
e5408a32
DW
1484 for (ent = mdstat; ent; ent = ent->next)
1485 if (is_container_member(ent, container))
e5e5d7ce 1486 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1487 break;
33414a01
DW
1488
1489 free_mdstat(mdstat);
1490
1491 return ent != NULL;
1492}
1493
1494/* open_subarray - opens a subarray in a container
1495 * @dev: container device name
feab51f8 1496 * @st: empty supertype
33414a01
DW
1497 * @quiet: block reporting errors flag
1498 *
1499 * On success returns an fd to a container and fills in *st
1500 */
feab51f8 1501int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1502{
1503 struct mdinfo *mdi;
a951a4f7 1504 struct mdinfo *info;
33414a01 1505 int fd, err = 1;
4dd2df09 1506 char *_devnm;
33414a01
DW
1507
1508 fd = open(dev, O_RDWR|O_EXCL);
1509 if (fd < 0) {
1510 if (!quiet)
e7b84f9d 1511 pr_err("Couldn't open %s, aborting\n",
33414a01 1512 dev);
b990032d 1513 return -1;
33414a01
DW
1514 }
1515
4dd2df09
N
1516 _devnm = fd2devnm(fd);
1517 if (_devnm == NULL) {
33414a01 1518 if (!quiet)
e7b84f9d
N
1519 pr_err("Failed to determine device number for %s\n",
1520 dev);
33414a01
DW
1521 goto close_fd;
1522 }
4dd2df09 1523 strcpy(st->devnm, _devnm);
33414a01 1524
4dd2df09 1525 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1526 if (!mdi) {
1527 if (!quiet)
e7b84f9d 1528 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1529 dev);
1530 goto close_fd;
1531 }
1532
1533 if (mdi->array.level != UnSet) {
1534 if (!quiet)
e7b84f9d 1535 pr_err("%s is not a container\n", dev);
33414a01
DW
1536 goto free_sysfs;
1537 }
1538
1539 st->ss = version_to_superswitch(mdi->text_version);
1540 if (!st->ss) {
1541 if (!quiet)
e7b84f9d
N
1542 pr_err("Operation not supported for %s metadata\n",
1543 mdi->text_version);
33414a01
DW
1544 goto free_sysfs;
1545 }
1546
4dd2df09 1547 if (st->devnm[0] == 0) {
33414a01 1548 if (!quiet)
e7b84f9d 1549 pr_err("Failed to allocate device name\n");
33414a01
DW
1550 goto free_sysfs;
1551 }
1552
db20d413 1553 if (!st->ss->load_container) {
33414a01 1554 if (!quiet)
e7b84f9d 1555 pr_err("%s is not a container\n", dev);
4dd2df09 1556 goto free_sysfs;
33414a01
DW
1557 }
1558
db20d413 1559 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1560 if (!quiet)
e7b84f9d 1561 pr_err("Failed to load metadata for %s\n",
db20d413 1562 dev);
4dd2df09 1563 goto free_sysfs;
33414a01
DW
1564 }
1565
a951a4f7
N
1566 info = st->ss->container_content(st, subarray);
1567 if (!info) {
1568 if (!quiet)
e7b84f9d 1569 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1570 subarray, dev);
1571 goto free_super;
1572 }
1573 free(info);
1574
33414a01
DW
1575 err = 0;
1576
1577 free_super:
1578 if (err)
1579 st->ss->free_super(st);
33414a01
DW
1580 free_sysfs:
1581 sysfs_free(mdi);
1582 close_fd:
1583 if (err)
1584 close(fd);
1585
1586 if (err)
1587 return -1;
1588 else
1589 return fd;
1590}
1591
7801ac20
N
1592int add_disk(int mdfd, struct supertype *st,
1593 struct mdinfo *sra, struct mdinfo *info)
1594{
1595 /* Add a device to an array, in one of 2 ways. */
1596 int rv;
1597#ifndef MDASSEMBLE
1598 if (st->ss->external) {
d23534e4
DW
1599 if (info->disk.state & (1<<MD_DISK_SYNC))
1600 info->recovery_start = MaxSector;
1601 else
1602 info->recovery_start = 0;
2904b26f 1603 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1604 if (! rv) {
1605 struct mdinfo *sd2;
f35f2525
N
1606 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1607 if (sd2 == info)
1608 break;
1609 if (sd2 == NULL) {
503975b9 1610 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1611 *sd2 = *info;
1612 sd2->next = sra->devs;
1613 sra->devs = sd2;
1614 }
7801ac20
N
1615 }
1616 } else
1617#endif
1618 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1619 return rv;
1620}
1621
de6ae750
N
1622int remove_disk(int mdfd, struct supertype *st,
1623 struct mdinfo *sra, struct mdinfo *info)
1624{
1625 int rv;
1626 /* Remove the disk given by 'info' from the array */
1627#ifndef MDASSEMBLE
1628 if (st->ss->external)
1629 rv = sysfs_set_str(sra, info, "slot", "none");
1630 else
1631#endif
1632 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1633 info->disk.minor));
1634 return rv;
1635}
1636
f35f2525
N
1637int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1638{
1639 /* Initialise kernel's knowledge of array.
1640 * This varies between externally managed arrays
1641 * and older kernels
1642 */
1643 int vers = md_get_version(mdfd);
1644 int rv;
1645
1646#ifndef MDASSEMBLE
1647 if (st->ss->external)
1648 rv = sysfs_set_array(info, vers);
1649 else
1650#endif
1651 if ((vers % 100) >= 1) { /* can use different versions */
1652 mdu_array_info_t inf;
1653 memset(&inf, 0, sizeof(inf));
1654 inf.major_version = info->array.major_version;
1655 inf.minor_version = info->array.minor_version;
1656 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1657 } else
1658 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1659 return rv;
1660}
1661
1e5c6983
DW
1662unsigned long long min_recovery_start(struct mdinfo *array)
1663{
1664 /* find the minimum recovery_start in an array for metadata
1665 * formats that only record per-array recovery progress instead
1666 * of per-device
1667 */
1668 unsigned long long recovery_start = MaxSector;
1669 struct mdinfo *d;
1670
1671 for (d = array->devs; d; d = d->next)
1672 recovery_start = min(recovery_start, d->recovery_start);
1673
1674 return recovery_start;
1675}
1676
4dd2df09 1677int mdmon_pid(char *devnm)
a931db9e
NB
1678{
1679 char path[100];
1680 char pid[10];
1681 int fd;
1682 int n;
10013317 1683
4dd2df09 1684 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1685
24f6f99b 1686 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1687
1688 if (fd < 0)
cf556303 1689 return -1;
a931db9e
NB
1690 n = read(fd, pid, 9);
1691 close(fd);
1692 if (n <= 0)
cf556303 1693 return -1;
24f6f99b 1694 return atoi(pid);
a931db9e
NB
1695}
1696
4dd2df09 1697int mdmon_running(char *devnm)
a931db9e 1698{
4dd2df09 1699 int pid = mdmon_pid(devnm);
24f6f99b 1700 if (pid <= 0)
a931db9e 1701 return 0;
24f6f99b 1702 if (kill(pid, 0) == 0)
a931db9e
NB
1703 return 1;
1704 return 0;
1705}
1706
4dd2df09 1707int start_mdmon(char *devnm)
8850ee3e 1708{
1913c325 1709 int i, skipped;
44d2e365 1710 int len;
ca3b6696 1711 pid_t pid;
9fe32043 1712 int status;
44d2e365
N
1713 char pathbuf[1024];
1714 char *paths[4] = {
1715 pathbuf,
1716 "/sbin/mdmon",
1717 "mdmon",
1718 NULL
1719 };
8850ee3e 1720
40ebbb9c 1721 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1722 return 0;
1723
9cf014ec 1724 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1725 if (len > 0) {
1726 char *sl;
1727 pathbuf[len] = 0;
1728 sl = strrchr(pathbuf, '/');
1729 if (sl)
1730 sl++;
1731 else
1732 sl = pathbuf;
1733 strcpy(sl, "mdmon");
1734 } else
1735 pathbuf[0] = '\0';
1736
0f7bdf89
JS
1737 /* First try to run systemctl */
1738 switch(fork()) {
1739 case 0:
1740 /* FIXME yuk. CLOSE_EXEC?? */
1741 skipped = 0;
1742 for (i = 3; skipped < 20; i++)
1743 if (close(i) < 0)
1744 skipped++;
1745 else
1746 skipped = 0;
1747
701d5b4a
N
1748 /* Don't want to see error messages from systemctl.
1749 * If the service doesn't exist, we start mdmon ourselves.
1750 */
1751 close(2);
1752 open("/dev/null", O_WRONLY);
0f7bdf89 1753 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
4dd2df09 1754 devnm);
0f7bdf89
JS
1755 status = execl("/usr/bin/systemctl", "systemctl", "start",
1756 pathbuf, NULL);
1757 status = execl("/bin/systemctl", "systemctl", "start",
1758 pathbuf, NULL);
1759 exit(1);
1760 case -1: pr_err("cannot run mdmon. "
1761 "Array remains readonly\n");
1762 return -1;
1763 default: /* parent - good */
1764 pid = wait(&status);
1765 if (pid >= 0 && status == 0)
1766 return 0;
1767 }
1768
1769 /* That failed, try running mdmon directly */
8850ee3e
N
1770 switch(fork()) {
1771 case 0:
1772 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1773 skipped = 0;
ca3b6696 1774 for (i = 3; skipped < 20; i++)
1913c325
N
1775 if (close(i) < 0)
1776 skipped++;
1777 else
1778 skipped = 0;
1779
ca3b6696 1780 for (i = 0; paths[i]; i++)
a0963a86 1781 if (paths[i][0]) {
3e23ba9d 1782 execl(paths[i], "mdmon",
4dd2df09 1783 devnm, NULL);
a0963a86 1784 }
8850ee3e 1785 exit(1);
e7b84f9d 1786 case -1: pr_err("cannot run mdmon. "
8850ee3e
N
1787 "Array remains readonly\n");
1788 return -1;
9fe32043
N
1789 default: /* parent - good */
1790 pid = wait(&status);
15c10423
JS
1791 if (pid < 0 || status != 0) {
1792 pr_err("failed to launch mdmon. "
1793 "Array remains readonly\n");
9fe32043 1794 return -1;
15c10423 1795 }
8850ee3e
N
1796 }
1797 return 0;
1798}
1799
40ebbb9c 1800int check_env(char *name)
5dcfcb71 1801{
40ebbb9c 1802 char *val = getenv(name);
5dcfcb71
DW
1803
1804 if (val && atoi(val) == 1)
1805 return 1;
1806
1807 return 0;
1808}
1809
148acb7b
DW
1810__u32 random32(void)
1811{
1812 __u32 rv;
1813 int rfd = open("/dev/urandom", O_RDONLY);
1814 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1815 rv = random();
1816 if (rfd >= 0)
1817 close(rfd);
1818 return rv;
1819}
1820
0e600426 1821#ifndef MDASSEMBLE
edd8d13c
NB
1822int flush_metadata_updates(struct supertype *st)
1823{
1824 int sfd;
1825 if (!st->updates) {
1826 st->update_tail = NULL;
1827 return -1;
1828 }
1829
4dd2df09 1830 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
1831 if (sfd < 0)
1832 return -1;
1833
1834 while (st->updates) {
1835 struct metadata_update *mu = st->updates;
1836 st->updates = mu->next;
1837
1838 send_message(sfd, mu, 0);
1839 wait_reply(sfd, 0);
1840 free(mu->buf);
1841 free(mu);
1842 }
1843 ack(sfd, 0);
1844 wait_reply(sfd, 0);
1845 close(sfd);
1846 st->update_tail = NULL;
1847 return 0;
1848}
1849
1850void append_metadata_update(struct supertype *st, void *buf, int len)
1851{
1852
503975b9 1853 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
1854
1855 mu->buf = buf;
1856 mu->len = len;
1857 mu->space = NULL;
cb23f1f4 1858 mu->space_list = NULL;
edd8d13c
NB
1859 mu->next = NULL;
1860 *st->update_tail = mu;
1861 st->update_tail = &mu->next;
1862}
0e600426 1863#endif /* MDASSEMBLE */
a931db9e 1864
fe6729fa
NB
1865#ifdef __TINYC__
1866/* tinyc doesn't optimize this check in ioctl.h out ... */
1867unsigned int __invalid_size_argument_for_IOC = 0;
1868#endif
1869
6d11ec6f
AK
1870int experimental(void)
1871{
1872 if (check_env("MDADM_EXPERIMENTAL"))
1873 return 1;
1874 else {
e7b84f9d 1875 pr_err("To use this feature MDADM_EXPERIMENTAL"
65c83a80 1876 " environment variable has to be defined.\n");
6d11ec6f
AK
1877 return 0;
1878 }
1879}
1880
326727d9
AC
1881/* Pick all spares matching given criteria from a container
1882 * if min_size == 0 do not check size
1883 * if domlist == NULL do not check domains
1884 * if spare_group given add it to domains of each spare
1885 * metadata allows to test domains using metadata of destination array */
1886struct mdinfo *container_choose_spares(struct supertype *st,
1887 unsigned long long min_size,
1888 struct domainlist *domlist,
1889 char *spare_group,
1890 const char *metadata, int get_one)
1891{
1892 struct mdinfo *d, **dp, *disks = NULL;
1893
1894 /* get list of all disks in container */
1895 if (st->ss->getinfo_super_disks)
1896 disks = st->ss->getinfo_super_disks(st);
1897
1898 if (!disks)
1899 return disks;
1900 /* find spare devices on the list */
1901 dp = &disks->devs;
1902 disks->array.spare_disks = 0;
1903 while (*dp) {
1904 int found = 0;
1905 d = *dp;
1906 if (d->disk.state == 0) {
1907 /* check if size is acceptable */
1908 unsigned long long dev_size;
1909 dev_t dev = makedev(d->disk.major,d->disk.minor);
1910
1911 if (!min_size ||
1912 (dev_size_from_id(dev, &dev_size) &&
1913 dev_size >= min_size))
1914 found = 1;
1915 /* check if domain matches */
1916 if (found && domlist) {
4dd2df09 1917 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
1918 if (spare_group)
1919 pol_add(&pol, pol_domain,
1920 spare_group, NULL);
e5508b36 1921 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
1922 found = 0;
1923 dev_policy_free(pol);
1924 }
1925 }
1926 if (found) {
1927 dp = &d->next;
1928 disks->array.spare_disks++;
1929 if (get_one) {
1930 sysfs_free(*dp);
1931 d->next = NULL;
1932 }
1933 } else {
1934 *dp = d->next;
1935 d->next = NULL;
1936 sysfs_free(d);
1937 }
1938 }
1939 return disks;
1940}
9eafa1de
MN
1941
1942/* Checks if paths point to the same device
1943 * Returns 0 if they do.
1944 * Returns 1 if they don't.
1945 * Returns -1 if something went wrong,
1946 * e.g. paths are empty or the files
1947 * they point to don't exist */
1948int compare_paths (char* path1, char* path2)
1949{
1950 struct stat st1,st2;
1951
1952 if (path1 == NULL || path2 == NULL)
1953 return -1;
1954 if (stat(path1,&st1) != 0)
1955 return -1;
1956 if (stat(path2,&st2) != 0)
1957 return -1;
1958 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
1959 return 0;
1960 return 1;
1961}