assemble: combine the two create_mddev calls in to one.
[thirdparty/mdadm.git] / util.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
4f589ad0 4 * Copyright (C) 2001-2006 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
22 * Email: <neilb@cse.unsw.edu.au>
23 * Paper: Neil Brown
24 * School of Computer Science and Engineering
25 * The University of New South Wales
26 * Sydney, 2052
27 * Australia
28 */
29
9a9dab36 30#include "mdadm.h"
64c4757e 31#include "md_p.h"
edd8d13c 32#include <sys/socket.h>
64c4757e 33#include <sys/utsname.h>
9fe32043 34#include <sys/wait.h>
edd8d13c 35#include <sys/un.h>
98c6faba 36#include <ctype.h>
a322f70c 37#include <dirent.h>
a931db9e 38#include <signal.h>
0a816ef9
NB
39
40/*
41 * following taken from linux/blkpg.h because they aren't
42 * anywhere else and it isn't safe to #include linux/ * stuff.
43 */
44
45#define BLKPG _IO(0x12,105)
46
47/* The argument structure */
48struct blkpg_ioctl_arg {
49 int op;
50 int flags;
51 int datalen;
52 void *data;
53};
54
55/* The subfunctions (for the op field) */
56#define BLKPG_ADD_PARTITION 1
57#define BLKPG_DEL_PARTITION 2
58
59/* Sizes of name fields. Unused at present. */
60#define BLKPG_DEVNAMELTH 64
61#define BLKPG_VOLNAMELTH 64
62
63/* The data structure for ADD_PARTITION and DEL_PARTITION */
64struct blkpg_partition {
65 long long start; /* starting offset in bytes */
66 long long length; /* length in bytes */
67 int pno; /* partition number */
68 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
69 to be used in kernel messages */
70 char volname[BLKPG_VOLNAMELTH]; /* volume label */
71};
64c4757e
NB
72
73/*
74 * Parse a 128 bit uuid in 4 integers
75 * format is 32 hexx nibbles with options :.<space> separator
76 * If not exactly 32 hex digits are found, return 0
77 * else return 1
78 */
79int parse_uuid(char *str, int uuid[4])
80{
aba69144
NB
81 int hit = 0; /* number of Hex digIT */
82 int i;
83 char c;
84 for (i=0; i<4; i++) uuid[i]=0;
85
86 while ((c= *str++)) {
87 int n;
88 if (c>='0' && c<='9')
89 n = c-'0';
90 else if (c>='a' && c <= 'f')
91 n = 10 + c - 'a';
92 else if (c>='A' && c <= 'F')
93 n = 10 + c - 'A';
94 else if (strchr(":. -", c))
95 continue;
96 else return 0;
97
98 if (hit<32) {
99 uuid[hit/8] <<= 4;
100 uuid[hit/8] += n;
101 }
102 hit++;
82b27616 103 }
aba69144
NB
104 if (hit == 32)
105 return 1;
106 return 0;
64c4757e
NB
107}
108
109
110/*
111 * Get the md version number.
112 * We use the RAID_VERSION ioctl if it is supported
113 * If not, but we have a block device with major '9', we assume
114 * 0.36.0
115 *
116 * Return version number as 24 but number - assume version parts
117 * always < 255
118 */
119
120int md_get_version(int fd)
121{
122 struct stat stb;
123 mdu_version_t vers;
124
125 if (fstat(fd, &stb)<0)
126 return -1;
127 if ((S_IFMT&stb.st_mode) != S_IFBLK)
128 return -1;
129
130 if (ioctl(fd, RAID_VERSION, &vers) == 0)
682c7051 131 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
5787fa49
NB
132 if (errno == EACCES)
133 return -1;
0df46c2a 134 if (major(stb.st_rdev) == MD_MAJOR)
682c7051 135 return (3600);
64c4757e
NB
136 return -1;
137}
138
64c4757e
NB
139int get_linux_version()
140{
141 struct utsname name;
98c6faba 142 char *cp;
64c4757e
NB
143 int a,b,c;
144 if (uname(&name) <0)
145 return -1;
146
98c6faba
NB
147 cp = name.release;
148 a = strtoul(cp, &cp, 10);
149 if (*cp != '.') return -1;
150 b = strtoul(cp+1, &cp, 10);
151 if (*cp != '.') return -1;
152 c = strtoul(cp+1, NULL, 10);
153
682c7051 154 return (a*1000000)+(b*1000)+c;
64c4757e
NB
155}
156
0430ed48
NB
157void remove_partitions(int fd)
158{
159 /* remove partitions from this block devices.
160 * This is used for components added to an array
161 */
162#ifdef BLKPG_DEL_PARTITION
163 struct blkpg_ioctl_arg a;
164 struct blkpg_partition p;
165
166 a.op = BLKPG_DEL_PARTITION;
167 a.data = (void*)&p;
168 a.datalen = sizeof(p);
169 a.flags = 0;
170 memset(a.data, 0, a.datalen);
171 for (p.pno=0; p.pno < 16; p.pno++)
172 ioctl(fd, BLKPG, &a);
173#endif
174}
175
583315d9 176int enough(int level, int raid_disks, int layout, int clean,
265e0f17 177 char *avail, int avail_disks)
64c4757e 178{
265e0f17 179 int copies, first;
64c4757e 180 switch (level) {
265e0f17
NB
181 case 10:
182 /* This is the tricky one - we need to check
183 * which actual disks are present.
184 */
702b557b 185 copies = (layout&255)* ((layout>>8) & 255);
265e0f17
NB
186 first=0;
187 do {
188 /* there must be one of the 'copies' form 'first' */
189 int n = copies;
190 int cnt=0;
191 while (n--) {
192 if (avail[first])
193 cnt++;
194 first = (first+1) % raid_disks;
195 }
196 if (cnt == 0)
197 return 0;
198
199 } while (first != 0);
200 return 1;
e5329c37 201
e0d19036
NB
202 case -4:
203 return avail_disks>= 1;
64c4757e
NB
204 case -1:
205 case 0:
206 return avail_disks == raid_disks;
207 case 1:
208 return avail_disks >= 1;
209 case 4:
210 case 5:
583315d9
NB
211 if (clean)
212 return avail_disks >= raid_disks-1;
213 else
214 return avail_disks >= raid_disks;
98c6faba 215 case 6:
583315d9
NB
216 if (clean)
217 return avail_disks >= raid_disks-2;
218 else
219 return avail_disks >= raid_disks;
64c4757e
NB
220 default:
221 return 0;
222 }
223}
224
36ba7d48 225const int uuid_match_any[4] = { ~0, ~0, ~0, ~0 };
f277ce36 226int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 227{
36ba7d48
DW
228 if (memcmp(a, uuid_match_any, sizeof(int[4])) == 0 ||
229 memcmp(b, uuid_match_any, sizeof(int[4])) == 0)
230 return 1;
231
f277ce36
NB
232 if (swapuuid) {
233 /* parse uuids are hostendian.
234 * uuid's from some superblocks are big-ending
aba69144 235 * if there is a difference, we need to swap..
f277ce36
NB
236 */
237 unsigned char *ac = (unsigned char *)a;
238 unsigned char *bc = (unsigned char *)b;
239 int i;
240 for (i=0; i<16; i+= 4) {
241 if (ac[i+0] != bc[i+3] ||
242 ac[i+1] != bc[i+2] ||
243 ac[i+2] != bc[i+1] ||
244 ac[i+3] != bc[i+0])
245 return 0;
246 }
247 return 1;
248 } else {
249 if (a[0]==b[0] &&
250 a[1]==b[1] &&
251 a[2]==b[2] &&
252 a[3]==b[3])
253 return 1;
254 return 0;
255 }
64c4757e 256}
350f29f9
NB
257void copy_uuid(void *a, int b[4], int swapuuid)
258{
259 if (swapuuid) {
260 /* parse uuids are hostendian.
261 * uuid's from some superblocks are big-ending
262 * if there is a difference, we need to swap..
263 */
264 unsigned char *ac = (unsigned char *)a;
265 unsigned char *bc = (unsigned char *)b;
266 int i;
267 for (i=0; i<16; i+= 4) {
268 ac[i+0] = bc[i+3];
269 ac[i+1] = bc[i+2];
270 ac[i+2] = bc[i+1];
271 ac[i+3] = bc[i+0];
272 }
273 } else
274 memcpy(a, b, 16);
275}
64c4757e 276
ff54de6e 277char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
d7288ddc 278{
9968e376
DW
279 int i, j;
280 int id;
d7288ddc
N
281 char uuid[16];
282 char *c = buf;
283 strcpy(c, "UUID-");
284 c += strlen(c);
285 copy_uuid(uuid, info->uuid, st->ss->swapuuid);
9968e376
DW
286 for (i = 0; i < 4; i++) {
287 id = uuid[i];
288 if (i)
ff54de6e 289 *c++ = sep;
9968e376
DW
290 for (j = 3; j >= 0; j--) {
291 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
292 c+= 2;
293 }
d7288ddc
N
294 }
295 return buf;
296}
297
435d4ebb 298#ifndef MDASSEMBLE
682c7051
NB
299int check_ext2(int fd, char *name)
300{
301 /*
302 * Check for an ext2fs file system.
303 * Superblock is always 1K at 1K offset
304 *
305 * s_magic is le16 at 56 == 0xEF53
306 * report mtime - le32 at 44
307 * blocks - le32 at 4
308 * logblksize - le32 at 24
309 */
310 unsigned char sb[1024];
311 time_t mtime;
312 int size, bsize;
313 if (lseek(fd, 1024,0)!= 1024)
314 return 0;
315 if (read(fd, sb, 1024)!= 1024)
316 return 0;
317 if (sb[56] != 0x53 || sb[57] != 0xef)
318 return 0;
319
320 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
321 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
322 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
323 fprintf(stderr, Name ": %s appears to contain an ext2fs file system\n",
324 name);
325 fprintf(stderr," size=%dK mtime=%s",
326 size*(1<<bsize), ctime(&mtime));
327 return 1;
328}
329
330int check_reiser(int fd, char *name)
331{
332 /*
333 * superblock is at 64K
334 * size is 1024;
335 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
336 *
337 */
338 unsigned char sb[1024];
881990a2 339 unsigned long size;
682c7051
NB
340 if (lseek(fd, 64*1024, 0) != 64*1024)
341 return 0;
342 if (read(fd, sb, 1024) != 1024)
343 return 0;
a46f4061
NB
344 if (strncmp((char*)sb+52, "ReIsErFs",8)!=0 &&
345 strncmp((char*)sb+52, "ReIsEr2Fs",9)!=0)
682c7051
NB
346 return 0;
347 fprintf(stderr, Name ": %s appears to contain a reiserfs file system\n",name);
348 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
881990a2 349 fprintf(stderr, " size = %luK\n", size*4);
aba69144 350
682c7051
NB
351 return 1;
352}
353
354int check_raid(int fd, char *name)
355{
4b1ac34b 356 struct mdinfo info;
682c7051 357 time_t crtime;
d078d77c 358 char *level;
82d9eba6 359 struct supertype *st = guess_super(fd);
f9ce90ba 360
82d9eba6 361 if (!st) return 0;
3da92f27 362 st->ss->load_super(st, fd, name);
82d9eba6
NB
363 /* Looks like a raid array .. */
364 fprintf(stderr, Name ": %s appears to be part of a raid array:\n",
365 name);
3da92f27
NB
366 st->ss->getinfo_super(st, &info);
367 st->ss->free_super(st);
82d9eba6 368 crtime = info.array.ctime;
d078d77c
NB
369 level = map_num(pers, info.array.level);
370 if (!level) level = "-unknown-";
371 fprintf(stderr, " level=%s devices=%d ctime=%s",
372 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 373 return 1;
682c7051
NB
374}
375
682c7051
NB
376int ask(char *mesg)
377{
378 char *add = "";
379 int i;
380 for (i=0; i<5; i++) {
381 char buf[100];
382 fprintf(stderr, "%s%s", mesg, add);
383 fflush(stderr);
384 if (fgets(buf, 100, stdin)==NULL)
385 return 0;
386 if (buf[0]=='y' || buf[0]=='Y')
387 return 1;
388 if (buf[0]=='n' || buf[0]=='N')
389 return 0;
390 add = "(y/n) ";
391 }
392 fprintf(stderr, Name ": assuming 'no'\n");
393 return 0;
394}
435d4ebb 395#endif /* MDASSEMBLE */
682c7051
NB
396
397char *map_num(mapping_t *map, int num)
398{
399 while (map->name) {
400 if (map->num == num)
401 return map->name;
402 map++;
403 }
404 return NULL;
405}
406
407int map_name(mapping_t *map, char *name)
408{
409 while (map->name) {
410 if (strcmp(map->name, name)==0)
411 return map->num;
412 map++;
413 }
98c6faba 414 return UnSet;
682c7051 415}
82b27616 416
e5329c37 417
8d80900b 418int is_standard(char *dev, int *nump)
e5329c37
NB
419{
420 /* tests if dev is a "standard" md dev name.
421 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 422 * where NN is a string of digits
598f0d58
NB
423 * Returns 1 if a partitionable standard,
424 * -1 if non-partitonable,
425 * 0 if not a standard name.
e5329c37 426 */
8d80900b
NB
427 char *d = strrchr(dev, '/');
428 int type=0;
429 int num;
430 if (!d)
e5329c37 431 return 0;
8d80900b
NB
432 if (strncmp(d, "/d",2)==0)
433 d += 2, type=1; /* /dev/md/dN{pM} */
434 else if (strncmp(d, "/md_d", 5)==0)
2b4ca8f0 435 d += 5, type=1; /* /dev/md_dN{pM} */
8d80900b
NB
436 else if (strncmp(d, "/md", 3)==0)
437 d += 3, type=-1; /* /dev/mdN */
438 else if (d-dev > 3 && strncmp(d-2, "md/", 3)==0)
5a6d1148 439 d += 1, type=-1; /* /dev/md/N */
e5329c37
NB
440 else
441 return 0;
8d80900b 442 if (!*d)
e5329c37 443 return 0;
8d80900b
NB
444 num = atoi(d);
445 while (isdigit(*d))
446 d++;
447 if (*d)
e5329c37 448 return 0;
8d80900b
NB
449 if (nump) *nump = num;
450
451 return type;
e5329c37
NB
452}
453
454
82b27616
NB
455/*
456 * convert a major/minor pair for a block device into a name in /dev, if possible.
457 * On the first call, walk /dev collecting name.
458 * Put them in a simple linked listfor now.
459 */
460struct devmap {
461 int major, minor;
462 char *name;
463 struct devmap *next;
464} *devlist = NULL;
465int devlist_ready = 0;
466
82b27616
NB
467int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s)
468{
bed256c2
NB
469 struct stat st;
470 if (S_ISLNK(stb->st_mode)) {
471 stat(name, &st);
472 stb = &st;
82b27616 473 }
bed256c2
NB
474
475 if ((stb->st_mode&S_IFMT)== S_IFBLK) {
476 char *n = strdup(name);
477 struct devmap *dm = malloc(sizeof(*dm));
478 if (strncmp(n, "/dev/./", 7)==0)
479 strcpy(n+4, name+6);
480 if (dm) {
481 dm->major = major(stb->st_rdev);
482 dm->minor = minor(stb->st_rdev);
483 dm->name = n;
484 dm->next = devlist;
485 devlist = dm;
486 }
487 }
488 return 0;
82b27616
NB
489}
490
45e878bb
NB
491#ifndef HAVE_NFTW
492#ifdef HAVE_FTW
493int add_dev_1(const char *name, const struct stat *stb, int flag)
494{
495 return add_dev(name, stb, flag, NULL);
496}
497int nftw(const char *path, int (*han)(const char *name, const struct stat *stb, int flag, struct FTW *s), int nopenfd, int flags)
498{
499 return ftw(path, add_dev_1, nopenfd);
500}
501#else
45e878bb
NB
502int nftw(const char *path, int (*han)(const char *name, const struct stat *stb, int flag, struct FTW *s), int nopenfd, int flags)
503{
504 return 0;
505}
506#endif /* HAVE_FTW */
507#endif /* HAVE_NFTW */
508
dd0781e5
NB
509/*
510 * Find a block device with the right major/minor number.
b79713f8
NB
511 * If we find multiple names, choose the shortest.
512 * If we find a non-standard name, it is probably there
513 * deliberately so prefer it over a standard name.
514 * This applies only to names for MD devices.
dd0781e5 515 */
16c6fa80 516char *map_dev(int major, int minor, int create)
82b27616 517{
dd0781e5 518 struct devmap *p;
b79713f8 519 char *std = NULL, *nonstd=NULL;
e7bb5d23 520 int did_check = 0;
eed35d66 521
e81cdd9f 522 if (major == 0 && minor == 0)
eed35d66 523 return NULL;
e81cdd9f 524
e7bb5d23 525 retry:
dd0781e5 526 if (!devlist_ready) {
0a416ec3
NB
527 char *dev = "/dev";
528 struct stat stb;
eed35d66
NB
529 while(devlist) {
530 struct devmap *d = devlist;
531 devlist = d->next;
532 free(d->name);
533 free(d);
534 }
0a416ec3
NB
535 if (lstat(dev, &stb)==0 &&
536 S_ISLNK(stb.st_mode))
537 dev = "/dev/.";
538 nftw(dev, add_dev, 10, FTW_PHYS);
dd0781e5 539 devlist_ready=1;
e7bb5d23 540 did_check = 1;
dd0781e5 541 }
82b27616 542
dd0781e5
NB
543 for (p=devlist; p; p=p->next)
544 if (p->major == major &&
545 p->minor == minor) {
b79713f8
NB
546 if (is_standard(p->name, NULL)) {
547 if (std == NULL ||
548 strlen(p->name) < strlen(std))
549 std = p->name;
550 } else {
551 if (nonstd == NULL ||
552 strlen(p->name) < strlen(nonstd))
553 nonstd = p->name;
554 }
dd0781e5 555 }
e7bb5d23
NB
556 if (!std && !nonstd && !did_check) {
557 devlist_ready = 0;
558 goto retry;
559 }
16c6fa80
NB
560 if (create && !std && !nonstd) {
561 static char buf[30];
382245c3 562 snprintf(buf, sizeof(buf), "%d:%d", major, minor);
16c6fa80
NB
563 nonstd = buf;
564 }
565
b79713f8 566 return nonstd ? nonstd : std;
82b27616
NB
567}
568
4b1ac34b 569unsigned long calc_csum(void *super, int bytes)
82b27616 570{
56eb10c0 571 unsigned long long newcsum = 0;
82b27616 572 int i;
4b1ac34b
NB
573 unsigned int csum;
574 unsigned int *superc = (unsigned int*) super;
82b27616 575
4b1ac34b 576 for(i=0; i<bytes/4; i++)
82b27616
NB
577 newcsum+= superc[i];
578 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 579#ifdef __alpha__
aba69144 580/* The in-kernel checksum calculation is always 16bit on
570c0542
NB
581 * the alpha, though it is 32 bit on i386...
582 * I wonder what it is elsewhere... (it uses and API in
583 * a way that it shouldn't).
584 */
585 csum = (csum & 0xffff) + (csum >> 16);
586 csum = (csum & 0xffff) + (csum >> 16);
587#endif
82b27616
NB
588 return csum;
589}
cd29a5c8 590
435d4ebb 591#ifndef MDASSEMBLE
56eb10c0 592char *human_size(long long bytes)
cd29a5c8
NB
593{
594 static char buf[30];
d5d3721e
NB
595
596 /* We convert bytes to either centi-M{ega,ibi}bytes or
597 * centi-G{igi,ibi}bytes, with appropriate rounding,
598 * and then print 1/100th of those as a decimal.
599 * We allow upto 2048Megabytes before converting to
600 * gigabytes, as that shows more precision and isn't
601 * too large a number.
602 * Terrabytes are not yet handled.
603 */
cd29a5c8 604
56eb10c0 605 if (bytes < 5000*1024)
cd29a5c8 606 buf[0]=0;
d5d3721e
NB
607 else if (bytes < 2*1024LL*1024LL*1024LL) {
608 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
609 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 610 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
611 cMiB/100 , cMiB % 100,
612 cMB/100, cMB % 100);
613 } else {
614 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
615 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 616 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
617 cGiB/100 , cGiB % 100,
618 cGB/100, cGB % 100);
619 }
cd29a5c8
NB
620 return buf;
621}
e0d19036
NB
622
623char *human_size_brief(long long bytes)
624{
625 static char buf[30];
e0d19036
NB
626
627 if (bytes < 5000*1024)
8f23b0b3 628 snprintf(buf, sizeof(buf), "%ld.%02ldKiB",
bd526cee 629 (long)(bytes>>10), (long)(((bytes&1023)*100+512)/1024)
e0d19036
NB
630 );
631 else if (bytes < 2*1024LL*1024LL*1024LL)
8f23b0b3 632 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
e0d19036 633 (long)(bytes>>20),
bd526cee 634 (long)((bytes&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
635 );
636 else
8f23b0b3 637 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
e0d19036 638 (long)(bytes>>30),
bd526cee 639 (long)(((bytes>>10)&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
640 );
641 return buf;
642}
e4965ef8
N
643
644void print_r10_layout(int layout)
645{
646 int near = layout & 255;
647 int far = (layout >> 8) & 255;
648 int offset = (layout&0x10000);
649 char *sep = "";
650
651 if (near != 1) {
652 printf("%s near=%d", sep, near);
653 sep = ",";
654 }
655 if (far != 1)
656 printf("%s %s=%d", sep, offset?"offset":"far", far);
657 if (near*far == 1)
658 printf("NO REDUNDANCY");
659}
435d4ebb 660#endif
e0d19036 661
5f8097be
NB
662unsigned long long calc_array_size(int level, int raid_disks, int layout,
663 int chunksize, unsigned long long devsize)
664{
665 int data_disks = 0;
666 switch (level) {
667 case 0: data_disks = raid_disks; break;
668 case 1: data_disks = 1; break;
669 case 4:
670 case 5: data_disks = raid_disks - 1; break;
671 case 6: data_disks = raid_disks - 2; break;
672 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
673 break;
674 }
675 devsize &= ~(unsigned long long)((chunksize>>9)-1);
676 return data_disks * devsize;
677}
678
dd0781e5 679int get_mdp_major(void)
98c6faba 680{
dd0781e5
NB
681static int mdp_major = -1;
682 FILE *fl;
98c6faba
NB
683 char *w;
684 int have_block = 0;
685 int have_devices = 0;
686 int last_num = -1;
dd0781e5
NB
687
688 if (mdp_major != -1)
689 return mdp_major;
690 fl = fopen("/proc/devices", "r");
98c6faba 691 if (!fl)
dd0781e5 692 return -1;
98c6faba
NB
693 while ((w = conf_word(fl, 1))) {
694 if (have_block && strcmp(w, "devices:")==0)
695 have_devices = 1;
696 have_block = (strcmp(w, "Block")==0);
697 if (isdigit(w[0]))
698 last_num = atoi(w);
699 if (have_devices && strcmp(w, "mdp")==0)
700 mdp_major = last_num;
701 free(w);
702 }
703 fclose(fl);
dd0781e5 704 return mdp_major;
98c6faba
NB
705}
706
0e600426 707#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
e0d19036
NB
708char *get_md_name(int dev)
709{
710 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 711 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
e0d19036
NB
712 static char devname[50];
713 struct stat stb;
98c6faba 714 dev_t rdev;
dd0781e5 715 char *dn;
98c6faba
NB
716
717 if (dev < 0) {
dd0781e5
NB
718 int mdp = get_mdp_major();
719 if (mdp < 0) return NULL;
0df46c2a 720 rdev = makedev(mdp, (-1-dev)<<6);
8f23b0b3 721 snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev);
98c6faba
NB
722 if (stat(devname, &stb) == 0
723 && (S_IFMT&stb.st_mode) == S_IFBLK
724 && (stb.st_rdev == rdev))
725 return devname;
726 } else {
0df46c2a 727 rdev = makedev(MD_MAJOR, dev);
8f23b0b3 728 snprintf(devname, sizeof(devname), "/dev/md%d", dev);
98c6faba
NB
729 if (stat(devname, &stb) == 0
730 && (S_IFMT&stb.st_mode) == S_IFBLK
731 && (stb.st_rdev == rdev))
732 return devname;
733
8f23b0b3 734 snprintf(devname, sizeof(devname), "/dev/md/%d", dev);
98c6faba
NB
735 if (stat(devname, &stb) == 0
736 && (S_IFMT&stb.st_mode) == S_IFBLK
737 && (stb.st_rdev == rdev))
738 return devname;
739 }
16c6fa80 740 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
741 if (dn)
742 return dn;
8f23b0b3 743 snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev);
e0d19036 744 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
745 if (errno != EEXIST)
746 return NULL;
e0d19036
NB
747
748 if (stat(devname, &stb) == 0
749 && (S_IFMT&stb.st_mode) == S_IFBLK
750 && (stb.st_rdev == rdev))
751 return devname;
752 unlink(devname);
753 return NULL;
754}
755
756void put_md_name(char *name)
757{
758 if (strncmp(name, "/dev/.tmp.md", 12)==0)
759 unlink(name);
760}
ea24acd0 761
ea24acd0
NB
762int find_free_devnum(int use_partitions)
763{
764 int devnum;
765 for (devnum = 127; devnum != 128;
766 devnum = devnum ? devnum-1 : (1<<22)-1) {
767 char *dn;
768 int _devnum;
769
770 _devnum = use_partitions ? (-1-devnum) : devnum;
771 if (mddev_busy(_devnum))
772 continue;
773 /* make sure it is new to /dev too, at least as a
774 * non-standard */
775 dn = map_dev(dev2major(_devnum), dev2minor(_devnum), 0);
776 if (dn && ! is_standard(dn, NULL))
777 continue;
778 break;
779 }
780 if (devnum == 128)
781 return NoMdDev;
782 return use_partitions ? (-1-devnum) : devnum;
783}
435d4ebb 784#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 785
8b0dabea
NB
786int dev_open(char *dev, int flags)
787{
788 /* like 'open', but if 'dev' matches %d:%d, create a temp
789 * block device and open that
790 */
791 char *e;
792 int fd = -1;
793 char devname[32];
e81cdd9f 794 int major;
8b0dabea 795 int minor;
e81cdd9f
NB
796
797 if (!dev) return -1;
798
799 major = strtoul(dev, &e, 0);
8b0dabea
NB
800 if (e > dev && *e == ':' && e[1] &&
801 (minor = strtoul(e+1, &e, 0)) >= 0 &&
802 *e == 0) {
8c210183
NB
803 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
804 (int)getpid(), major, minor);
8b0dabea 805 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
6416d527 806 fd = open(devname, flags|O_DIRECT);
8b0dabea
NB
807 unlink(devname);
808 }
809 } else
6416d527 810 fd = open(dev, flags|O_DIRECT);
8b0dabea
NB
811 return fd;
812}
f9ce90ba 813
a931db9e
NB
814int open_dev_excl(int devnum)
815{
816 char buf[20];
817 int i;
818
819 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
820 for (i=0 ; i<25 ; i++) {
821 int fd = dev_open(buf, O_RDWR|O_EXCL);
822 if (fd >= 0)
823 return fd;
824 if (errno != EBUSY)
825 return fd;
826 usleep(200000);
827 }
828 return -1;
829}
830
cdddbdbc 831struct superswitch *superlist[] = { &super0, &super1, &super_ddf, &super_imsm, NULL };
f9ce90ba 832
ea24acd0 833#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 834
1686dc25 835struct supertype *super_by_fd(int fd)
f9ce90ba 836{
1686dc25
NB
837 mdu_array_info_t array;
838 int vers;
839 int minor;
840 struct supertype *st = NULL;
7e0f6979 841 struct mdinfo *sra;
142cb9e1 842 char *verstr;
1686dc25
NB
843 char version[20];
844 int i;
f7e7067b 845 char *subarray = NULL;
1686dc25
NB
846
847 sra = sysfs_read(fd, 0, GET_VERSION);
848
849 if (sra) {
7e0f6979
NB
850 vers = sra->array.major_version;
851 minor = sra->array.minor_version;
142cb9e1 852 verstr = sra->text_version;
1686dc25
NB
853 } else {
854 if (ioctl(fd, GET_ARRAY_INFO, &array))
855 array.major_version = array.minor_version = 0;
856 vers = array.major_version;
857 minor = array.minor_version;
142cb9e1 858 verstr = "";
6fbba4c9 859 }
82d9eba6 860
1686dc25
NB
861 if (vers != -1) {
862 sprintf(version, "%d.%d", vers, minor);
863 verstr = version;
6fbba4c9 864 }
3c558363 865 if (minor == -2 && is_subarray(verstr)) {
f7e7067b
NB
866 char *dev = verstr+1;
867 subarray = strchr(dev, '/');
868 int devnum;
869 if (subarray)
870 *subarray++ = '\0';
77472ff8 871 devnum = devname2devnum(dev);
f7e7067b
NB
872 subarray = strdup(subarray);
873 if (sra)
874 sysfs_free(sra);
875 sra = sysfs_read(-1, devnum, GET_VERSION);
876 verstr = sra->text_version ? : "-no-metadata-";
877 }
878
879 for (i = 0; st == NULL && superlist[i] ; i++)
880 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
881
882 if (sra)
883 sysfs_free(sra);
f7e7067b 884 if (st) {
3b0896f8 885 st->sb = NULL;
f7e7067b
NB
886 if (subarray) {
887 strncpy(st->subarray, subarray, 32);
888 st->subarray[31] = 0;
889 free(subarray);
890 } else
891 st->subarray[0] = 0;
892 }
82d9eba6 893 return st;
f9ce90ba 894}
ea24acd0
NB
895#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
896
f9ce90ba 897
159c3a1a 898struct supertype *dup_super(struct supertype *orig)
3da92f27 899{
159c3a1a 900 struct supertype *st;
1686dc25 901
d2ca6449
NB
902 if (!orig)
903 return orig;
159c3a1a 904 st = malloc(sizeof(*st));
3da92f27
NB
905 if (!st)
906 return st;
ef609477 907 memset(st, 0, sizeof(*st));
159c3a1a
NB
908 st->ss = orig->ss;
909 st->max_devs = orig->max_devs;
910 st->minor_version = orig->minor_version;
f7e7067b 911 strcpy(st->subarray, orig->subarray);
159c3a1a
NB
912 st->sb = NULL;
913 st->info = NULL;
914 return st;
3da92f27
NB
915}
916
82d9eba6 917struct supertype *guess_super(int fd)
f9ce90ba
NB
918{
919 /* try each load_super to find the best match,
920 * and return the best superswitch
921 */
82d9eba6
NB
922 struct superswitch *ss;
923 struct supertype *st;
570c0542
NB
924 unsigned long besttime = 0;
925 int bestsuper = -1;
f9ce90ba
NB
926 int i;
927
82d9eba6 928 st = malloc(sizeof(*st));
f9ce90ba
NB
929 for (i=0 ; superlist[i]; i++) {
930 int rv;
931 ss = superlist[i];
ef609477 932 memset(st, 0, sizeof(*st));
3da92f27 933 rv = ss->load_super(st, fd, NULL);
570c0542
NB
934 if (rv == 0) {
935 struct mdinfo info;
3da92f27 936 st->ss->getinfo_super(st, &info);
570c0542
NB
937 if (bestsuper == -1 ||
938 besttime < info.array.ctime) {
939 bestsuper = i;
940 besttime = info.array.ctime;
570c0542 941 }
3da92f27 942 ss->free_super(st);
570c0542
NB
943 }
944 }
945 if (bestsuper != -1) {
946 int rv;
ef609477 947 memset(st, 0, sizeof(*st));
3da92f27 948 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 949 if (rv == 0) {
5e747af2 950 superlist[bestsuper]->free_super(st);
82d9eba6 951 return st;
f9ce90ba
NB
952 }
953 }
570c0542 954 free(st);
f9ce90ba
NB
955 return NULL;
956}
fe6729fa 957
beae1dfe
NB
958/* Return size of device in bytes */
959int get_dev_size(int fd, char *dname, unsigned long long *sizep)
960{
961 unsigned long long ldsize;
c2c9bb6f
NB
962 struct stat st;
963
964 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
965 ldsize = (unsigned long long)st.st_size;
966 else
beae1dfe
NB
967#ifdef BLKGETSIZE64
968 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
969#endif
970 {
971 unsigned long dsize;
972 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
973 ldsize = dsize;
974 ldsize <<= 9;
975 } else {
976 if (dname)
977 fprintf(stderr, Name ": Cannot get size of %s: %s\b",
978 dname, strerror(errno));
979 return 0;
980 }
981 }
982 *sizep = ldsize;
983 return 1;
984}
8fac0577 985
8382f19b
NB
986void get_one_disk(int mdfd, mdu_array_info_t *ainf, mdu_disk_info_t *disk)
987{
988 int d;
989 ioctl(mdfd, GET_ARRAY_INFO, ainf);
990 for (d = 0 ; d < ainf->raid_disks + ainf->nr_disks ; d++)
991 if (ioctl(mdfd, GET_DISK_INFO, disk) == 0)
992 return;
993}
63152c1b 994
a322f70c
DW
995int open_container(int fd)
996{
997 /* 'fd' is a block device. Find out if it is in use
998 * by a container, and return an open fd on that container.
999 */
1000 char path[256];
1001 char *e;
1002 DIR *dir;
1003 struct dirent *de;
1004 int dfd, n;
1005 char buf[200];
1006 int major, minor;
1007 struct stat st;
1008
1009 if (fstat(fd, &st) != 0)
1010 return -1;
1011 sprintf(path, "/sys/dev/block/%d:%d/holders",
1012 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1013 e = path + strlen(path);
1014
1015 dir = opendir(path);
1016 if (!dir)
1017 return -1;
1018 while ((de = readdir(dir))) {
1019 if (de->d_ino == 0)
1020 continue;
1021 if (de->d_name[0] == '.')
1022 continue;
1023 sprintf(e, "/%s/dev", de->d_name);
1024 dfd = open(path, O_RDONLY);
1025 if (dfd < 0)
1026 continue;
1027 n = read(dfd, buf, sizeof(buf));
1028 close(dfd);
1029 if (n <= 0 || n >= sizeof(buf))
1030 continue;
1031 buf[n] = 0;
1032 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1033 continue;
1034 sprintf(buf, "%d:%d", major, minor);
1035 dfd = dev_open(buf, O_RDONLY);
1036 if (dfd >= 0) {
1037 closedir(dir);
1038 return dfd;
1039 }
1040 }
355726fa 1041 closedir(dir);
a322f70c
DW
1042 return -1;
1043}
1044
7801ac20
N
1045int add_disk(int mdfd, struct supertype *st,
1046 struct mdinfo *sra, struct mdinfo *info)
1047{
1048 /* Add a device to an array, in one of 2 ways. */
1049 int rv;
1050#ifndef MDASSEMBLE
1051 if (st->ss->external) {
1052 rv = sysfs_add_disk(sra, info);
1053 if (! rv) {
1054 struct mdinfo *sd2;
f35f2525
N
1055 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1056 if (sd2 == info)
1057 break;
1058 if (sd2 == NULL) {
1059 sd2 = malloc(sizeof(*sd2));
1060 *sd2 = *info;
1061 sd2->next = sra->devs;
1062 sra->devs = sd2;
1063 }
7801ac20
N
1064 }
1065 } else
1066#endif
1067 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1068 return rv;
1069}
1070
f35f2525
N
1071int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1072{
1073 /* Initialise kernel's knowledge of array.
1074 * This varies between externally managed arrays
1075 * and older kernels
1076 */
1077 int vers = md_get_version(mdfd);
1078 int rv;
1079
1080#ifndef MDASSEMBLE
1081 if (st->ss->external)
1082 rv = sysfs_set_array(info, vers);
1083 else
1084#endif
1085 if ((vers % 100) >= 1) { /* can use different versions */
1086 mdu_array_info_t inf;
1087 memset(&inf, 0, sizeof(inf));
1088 inf.major_version = info->array.major_version;
1089 inf.minor_version = info->array.minor_version;
1090 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
1091 } else
1092 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
1093 return rv;
1094}
1095
2f6079dc
NB
1096char *devnum2devname(int num)
1097{
1098 char name[100];
1099 if (num > 0)
1100 sprintf(name, "md%d", num);
1101 else
1102 sprintf(name, "md_d%d", -1-num);
1103 return strdup(name);
1104}
1105
77472ff8
NB
1106int devname2devnum(char *name)
1107{
1108 char *ep;
1109 int num;
1110 if (strncmp(name, "md_d", 4)==0)
1111 num = -1-strtoul(name+4, &ep, 10);
1112 else
1113 num = strtoul(name+2, &ep, 10);
1114 return num;
1115}
1116
c94709e8 1117int stat2devnum(struct stat *st)
2f6079dc 1118{
d7ab966b
N
1119 char path[30];
1120 char link[200];
1121 char *cp;
1122 int n;
1123
c94709e8
DW
1124 if ((S_IFMT & st->st_mode) == S_IFBLK) {
1125 if (major(st->st_rdev) == MD_MAJOR)
1126 return minor(st->st_rdev);
d7ab966b
N
1127 else if (major(st->st_rdev) == get_mdp_major())
1128 return -1- (minor(st->st_rdev)>>MdpMinorShift);
1129
1130 /* must be an extended-minor partition. Look at the
1131 * /sys/dev/block/%d:%d link which must look like
1132 * ../../block/mdXXX/mdXXXpYY
1133 */
1134 sprintf(path, "/sys/dev/block/%d:%d", major(st->st_rdev),
1135 minor(st->st_rdev));
1136 n = readlink(path, link, sizeof(link)-1);
1137 if (n <= 0)
1138 return NoMdDev;
1139 link[n] = 0;
1140 cp = strrchr(link, '/');
1141 if (cp) *cp = 0;
1142 cp = strchr(link, '/');
1143 if (cp && strncmp(cp, "/md", 3) == 0)
1144 return devname2devnum(cp+1);
2f6079dc 1145 }
d7ab966b 1146 return NoMdDev;
c94709e8
DW
1147
1148}
1149
1150int fd2devnum(int fd)
1151{
1152 struct stat stb;
1153 if (fstat(fd, &stb) == 0)
1154 return stat2devnum(&stb);
d7ab966b 1155 return NoMdDev;
2f6079dc
NB
1156}
1157
a931db9e
NB
1158int mdmon_running(int devnum)
1159{
1160 char path[100];
1161 char pid[10];
1162 int fd;
1163 int n;
1164 sprintf(path, "/var/run/mdadm/%s.pid", devnum2devname(devnum));
1165 fd = open(path, O_RDONLY, 0);
1166
1167 if (fd < 0)
1168 return 0;
1169 n = read(fd, pid, 9);
1170 close(fd);
1171 if (n <= 0)
1172 return 0;
1173 if (kill(atoi(pid), 0) == 0)
1174 return 1;
1175 return 0;
1176}
1177
1178int signal_mdmon(int devnum)
1179{
1180 char path[100];
1181 char pid[10];
1182 int fd;
1183 int n;
1184 sprintf(path, "/var/run/mdadm/%s.pid", devnum2devname(devnum));
1185 fd = open(path, O_RDONLY, 0);
1186
1187 if (fd < 0)
1188 return 0;
1189 n = read(fd, pid, 9);
1190 close(fd);
1191 if (n <= 0)
1192 return 0;
1193 if (kill(atoi(pid), SIGUSR1) == 0)
1194 return 1;
1195 return 0;
1196}
1197
8850ee3e
N
1198int start_mdmon(int devnum)
1199{
1200 int i;
44d2e365 1201 int len;
9fe32043
N
1202 pid_t pid;
1203 int status;
44d2e365
N
1204 char pathbuf[1024];
1205 char *paths[4] = {
1206 pathbuf,
1207 "/sbin/mdmon",
1208 "mdmon",
1209 NULL
1210 };
8850ee3e
N
1211
1212 if (env_no_mdmon())
1213 return 0;
1214
44d2e365
N
1215 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf));
1216 if (len > 0) {
1217 char *sl;
1218 pathbuf[len] = 0;
1219 sl = strrchr(pathbuf, '/');
1220 if (sl)
1221 sl++;
1222 else
1223 sl = pathbuf;
1224 strcpy(sl, "mdmon");
1225 } else
1226 pathbuf[0] = '\0';
1227
8850ee3e
N
1228 switch(fork()) {
1229 case 0:
1230 /* FIXME yuk. CLOSE_EXEC?? */
1231 for (i=3; i < 100; i++)
1232 close(i);
44d2e365
N
1233 for (i=0; paths[i]; i++)
1234 if (paths[i][0])
1235 execl(paths[i], "mdmon",
1236 map_dev(dev2major(devnum),
1237 dev2minor(devnum),
1238 1), NULL);
8850ee3e
N
1239 exit(1);
1240 case -1: fprintf(stderr, Name ": cannot run mdmon. "
1241 "Array remains readonly\n");
1242 return -1;
9fe32043
N
1243 default: /* parent - good */
1244 pid = wait(&status);
1245 if (pid < 0 || status != 0)
1246 return -1;
8850ee3e
N
1247 }
1248 return 0;
1249}
1250
5dcfcb71
DW
1251int env_no_mdmon(void)
1252{
1253 char *val = getenv("MDADM_NO_MDMON");
1254
1255 if (val && atoi(val) == 1)
1256 return 1;
1257
1258 return 0;
1259}
1260
0e600426 1261#ifndef MDASSEMBLE
edd8d13c
NB
1262int flush_metadata_updates(struct supertype *st)
1263{
1264 int sfd;
1265 if (!st->updates) {
1266 st->update_tail = NULL;
1267 return -1;
1268 }
1269
1270 sfd = connect_monitor(devnum2devname(st->container_dev));
1271 if (sfd < 0)
1272 return -1;
1273
1274 while (st->updates) {
1275 struct metadata_update *mu = st->updates;
1276 st->updates = mu->next;
1277
1278 send_message(sfd, mu, 0);
1279 wait_reply(sfd, 0);
1280 free(mu->buf);
1281 free(mu);
1282 }
1283 ack(sfd, 0);
1284 wait_reply(sfd, 0);
1285 close(sfd);
1286 st->update_tail = NULL;
1287 return 0;
1288}
1289
1290void append_metadata_update(struct supertype *st, void *buf, int len)
1291{
1292
1293 struct metadata_update *mu = malloc(sizeof(*mu));
1294
1295 mu->buf = buf;
1296 mu->len = len;
1297 mu->space = NULL;
1298 mu->next = NULL;
1299 *st->update_tail = mu;
1300 st->update_tail = &mu->next;
1301}
71d60c48
DW
1302
1303struct superswitch *find_metadata_methods(char *vers)
1304{
1305 if (strcmp(vers, "ddf") == 0)
1306 return &super_ddf;
1307 if (strcmp(vers, "imsm") == 0)
1308 return &super_imsm;
1309 return NULL;
1310}
0e600426 1311#endif /* MDASSEMBLE */
a931db9e 1312
fe6729fa
NB
1313#ifdef __TINYC__
1314/* tinyc doesn't optimize this check in ioctl.h out ... */
1315unsigned int __invalid_size_argument_for_IOC = 0;
1316#endif
1317