monitor: don't mark dirty on resync complete
[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
f277ce36 225int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 226{
f277ce36
NB
227 if (swapuuid) {
228 /* parse uuids are hostendian.
229 * uuid's from some superblocks are big-ending
aba69144 230 * if there is a difference, we need to swap..
f277ce36
NB
231 */
232 unsigned char *ac = (unsigned char *)a;
233 unsigned char *bc = (unsigned char *)b;
234 int i;
235 for (i=0; i<16; i+= 4) {
236 if (ac[i+0] != bc[i+3] ||
237 ac[i+1] != bc[i+2] ||
238 ac[i+2] != bc[i+1] ||
239 ac[i+3] != bc[i+0])
240 return 0;
241 }
242 return 1;
243 } else {
244 if (a[0]==b[0] &&
245 a[1]==b[1] &&
246 a[2]==b[2] &&
247 a[3]==b[3])
248 return 1;
249 return 0;
250 }
64c4757e 251}
350f29f9
NB
252void copy_uuid(void *a, int b[4], int swapuuid)
253{
254 if (swapuuid) {
255 /* parse uuids are hostendian.
256 * uuid's from some superblocks are big-ending
257 * if there is a difference, we need to swap..
258 */
259 unsigned char *ac = (unsigned char *)a;
260 unsigned char *bc = (unsigned char *)b;
261 int i;
262 for (i=0; i<16; i+= 4) {
263 ac[i+0] = bc[i+3];
264 ac[i+1] = bc[i+2];
265 ac[i+2] = bc[i+1];
266 ac[i+3] = bc[i+0];
267 }
268 } else
269 memcpy(a, b, 16);
270}
64c4757e 271
435d4ebb 272#ifndef MDASSEMBLE
682c7051
NB
273int check_ext2(int fd, char *name)
274{
275 /*
276 * Check for an ext2fs file system.
277 * Superblock is always 1K at 1K offset
278 *
279 * s_magic is le16 at 56 == 0xEF53
280 * report mtime - le32 at 44
281 * blocks - le32 at 4
282 * logblksize - le32 at 24
283 */
284 unsigned char sb[1024];
285 time_t mtime;
286 int size, bsize;
287 if (lseek(fd, 1024,0)!= 1024)
288 return 0;
289 if (read(fd, sb, 1024)!= 1024)
290 return 0;
291 if (sb[56] != 0x53 || sb[57] != 0xef)
292 return 0;
293
294 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
295 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
296 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
297 fprintf(stderr, Name ": %s appears to contain an ext2fs file system\n",
298 name);
299 fprintf(stderr," size=%dK mtime=%s",
300 size*(1<<bsize), ctime(&mtime));
301 return 1;
302}
303
304int check_reiser(int fd, char *name)
305{
306 /*
307 * superblock is at 64K
308 * size is 1024;
309 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
310 *
311 */
312 unsigned char sb[1024];
881990a2 313 unsigned long size;
682c7051
NB
314 if (lseek(fd, 64*1024, 0) != 64*1024)
315 return 0;
316 if (read(fd, sb, 1024) != 1024)
317 return 0;
a46f4061
NB
318 if (strncmp((char*)sb+52, "ReIsErFs",8)!=0 &&
319 strncmp((char*)sb+52, "ReIsEr2Fs",9)!=0)
682c7051
NB
320 return 0;
321 fprintf(stderr, Name ": %s appears to contain a reiserfs file system\n",name);
322 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
881990a2 323 fprintf(stderr, " size = %luK\n", size*4);
aba69144 324
682c7051
NB
325 return 1;
326}
327
328int check_raid(int fd, char *name)
329{
4b1ac34b 330 struct mdinfo info;
682c7051 331 time_t crtime;
d078d77c 332 char *level;
82d9eba6 333 struct supertype *st = guess_super(fd);
f9ce90ba 334
82d9eba6 335 if (!st) return 0;
3da92f27 336 st->ss->load_super(st, fd, name);
82d9eba6
NB
337 /* Looks like a raid array .. */
338 fprintf(stderr, Name ": %s appears to be part of a raid array:\n",
339 name);
3da92f27
NB
340 st->ss->getinfo_super(st, &info);
341 st->ss->free_super(st);
82d9eba6 342 crtime = info.array.ctime;
d078d77c
NB
343 level = map_num(pers, info.array.level);
344 if (!level) level = "-unknown-";
345 fprintf(stderr, " level=%s devices=%d ctime=%s",
346 level, info.array.raid_disks, ctime(&crtime));
82d9eba6 347 return 1;
682c7051
NB
348}
349
682c7051
NB
350int ask(char *mesg)
351{
352 char *add = "";
353 int i;
354 for (i=0; i<5; i++) {
355 char buf[100];
356 fprintf(stderr, "%s%s", mesg, add);
357 fflush(stderr);
358 if (fgets(buf, 100, stdin)==NULL)
359 return 0;
360 if (buf[0]=='y' || buf[0]=='Y')
361 return 1;
362 if (buf[0]=='n' || buf[0]=='N')
363 return 0;
364 add = "(y/n) ";
365 }
366 fprintf(stderr, Name ": assuming 'no'\n");
367 return 0;
368}
435d4ebb 369#endif /* MDASSEMBLE */
682c7051
NB
370
371char *map_num(mapping_t *map, int num)
372{
373 while (map->name) {
374 if (map->num == num)
375 return map->name;
376 map++;
377 }
378 return NULL;
379}
380
381int map_name(mapping_t *map, char *name)
382{
383 while (map->name) {
384 if (strcmp(map->name, name)==0)
385 return map->num;
386 map++;
387 }
98c6faba 388 return UnSet;
682c7051 389}
82b27616 390
e5329c37 391
8d80900b 392int is_standard(char *dev, int *nump)
e5329c37
NB
393{
394 /* tests if dev is a "standard" md dev name.
395 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 396 * where NN is a string of digits
598f0d58
NB
397 * Returns 1 if a partitionable standard,
398 * -1 if non-partitonable,
399 * 0 if not a standard name.
e5329c37 400 */
8d80900b
NB
401 char *d = strrchr(dev, '/');
402 int type=0;
403 int num;
404 if (!d)
e5329c37 405 return 0;
8d80900b
NB
406 if (strncmp(d, "/d",2)==0)
407 d += 2, type=1; /* /dev/md/dN{pM} */
408 else if (strncmp(d, "/md_d", 5)==0)
409 d += 5, type=1; /* /dev/md_dNpM */
410 else if (strncmp(d, "/md", 3)==0)
411 d += 3, type=-1; /* /dev/mdN */
412 else if (d-dev > 3 && strncmp(d-2, "md/", 3)==0)
5a6d1148 413 d += 1, type=-1; /* /dev/md/N */
e5329c37
NB
414 else
415 return 0;
8d80900b 416 if (!*d)
e5329c37 417 return 0;
8d80900b
NB
418 num = atoi(d);
419 while (isdigit(*d))
420 d++;
421 if (*d)
e5329c37 422 return 0;
8d80900b
NB
423 if (nump) *nump = num;
424
425 return type;
e5329c37
NB
426}
427
428
82b27616
NB
429/*
430 * convert a major/minor pair for a block device into a name in /dev, if possible.
431 * On the first call, walk /dev collecting name.
432 * Put them in a simple linked listfor now.
433 */
434struct devmap {
435 int major, minor;
436 char *name;
437 struct devmap *next;
438} *devlist = NULL;
439int devlist_ready = 0;
440
82b27616
NB
441int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s)
442{
bed256c2
NB
443 struct stat st;
444 if (S_ISLNK(stb->st_mode)) {
445 stat(name, &st);
446 stb = &st;
82b27616 447 }
bed256c2
NB
448
449 if ((stb->st_mode&S_IFMT)== S_IFBLK) {
450 char *n = strdup(name);
451 struct devmap *dm = malloc(sizeof(*dm));
452 if (strncmp(n, "/dev/./", 7)==0)
453 strcpy(n+4, name+6);
454 if (dm) {
455 dm->major = major(stb->st_rdev);
456 dm->minor = minor(stb->st_rdev);
457 dm->name = n;
458 dm->next = devlist;
459 devlist = dm;
460 }
461 }
462 return 0;
82b27616
NB
463}
464
45e878bb
NB
465#ifndef HAVE_NFTW
466#ifdef HAVE_FTW
467int add_dev_1(const char *name, const struct stat *stb, int flag)
468{
469 return add_dev(name, stb, flag, NULL);
470}
471int nftw(const char *path, int (*han)(const char *name, const struct stat *stb, int flag, struct FTW *s), int nopenfd, int flags)
472{
473 return ftw(path, add_dev_1, nopenfd);
474}
475#else
45e878bb
NB
476int nftw(const char *path, int (*han)(const char *name, const struct stat *stb, int flag, struct FTW *s), int nopenfd, int flags)
477{
478 return 0;
479}
480#endif /* HAVE_FTW */
481#endif /* HAVE_NFTW */
482
dd0781e5
NB
483/*
484 * Find a block device with the right major/minor number.
b79713f8
NB
485 * If we find multiple names, choose the shortest.
486 * If we find a non-standard name, it is probably there
487 * deliberately so prefer it over a standard name.
488 * This applies only to names for MD devices.
dd0781e5 489 */
16c6fa80 490char *map_dev(int major, int minor, int create)
82b27616 491{
dd0781e5 492 struct devmap *p;
b79713f8 493 char *std = NULL, *nonstd=NULL;
e7bb5d23 494 int did_check = 0;
eed35d66 495
e81cdd9f 496 if (major == 0 && minor == 0)
eed35d66 497 return NULL;
e81cdd9f 498
e7bb5d23 499 retry:
dd0781e5 500 if (!devlist_ready) {
0a416ec3
NB
501 char *dev = "/dev";
502 struct stat stb;
eed35d66
NB
503 while(devlist) {
504 struct devmap *d = devlist;
505 devlist = d->next;
506 free(d->name);
507 free(d);
508 }
0a416ec3
NB
509 if (lstat(dev, &stb)==0 &&
510 S_ISLNK(stb.st_mode))
511 dev = "/dev/.";
512 nftw(dev, add_dev, 10, FTW_PHYS);
dd0781e5 513 devlist_ready=1;
e7bb5d23 514 did_check = 1;
dd0781e5 515 }
82b27616 516
dd0781e5
NB
517 for (p=devlist; p; p=p->next)
518 if (p->major == major &&
519 p->minor == minor) {
b79713f8
NB
520 if (is_standard(p->name, NULL)) {
521 if (std == NULL ||
522 strlen(p->name) < strlen(std))
523 std = p->name;
524 } else {
525 if (nonstd == NULL ||
526 strlen(p->name) < strlen(nonstd))
527 nonstd = p->name;
528 }
dd0781e5 529 }
e7bb5d23
NB
530 if (!std && !nonstd && !did_check) {
531 devlist_ready = 0;
532 goto retry;
533 }
16c6fa80
NB
534 if (create && !std && !nonstd) {
535 static char buf[30];
382245c3 536 snprintf(buf, sizeof(buf), "%d:%d", major, minor);
16c6fa80
NB
537 nonstd = buf;
538 }
539
b79713f8 540 return nonstd ? nonstd : std;
82b27616
NB
541}
542
4b1ac34b 543unsigned long calc_csum(void *super, int bytes)
82b27616 544{
56eb10c0 545 unsigned long long newcsum = 0;
82b27616 546 int i;
4b1ac34b
NB
547 unsigned int csum;
548 unsigned int *superc = (unsigned int*) super;
82b27616 549
4b1ac34b 550 for(i=0; i<bytes/4; i++)
82b27616
NB
551 newcsum+= superc[i];
552 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 553#ifdef __alpha__
aba69144 554/* The in-kernel checksum calculation is always 16bit on
570c0542
NB
555 * the alpha, though it is 32 bit on i386...
556 * I wonder what it is elsewhere... (it uses and API in
557 * a way that it shouldn't).
558 */
559 csum = (csum & 0xffff) + (csum >> 16);
560 csum = (csum & 0xffff) + (csum >> 16);
561#endif
82b27616
NB
562 return csum;
563}
cd29a5c8 564
435d4ebb 565#ifndef MDASSEMBLE
56eb10c0 566char *human_size(long long bytes)
cd29a5c8
NB
567{
568 static char buf[30];
d5d3721e
NB
569
570 /* We convert bytes to either centi-M{ega,ibi}bytes or
571 * centi-G{igi,ibi}bytes, with appropriate rounding,
572 * and then print 1/100th of those as a decimal.
573 * We allow upto 2048Megabytes before converting to
574 * gigabytes, as that shows more precision and isn't
575 * too large a number.
576 * Terrabytes are not yet handled.
577 */
cd29a5c8 578
56eb10c0 579 if (bytes < 5000*1024)
cd29a5c8 580 buf[0]=0;
d5d3721e
NB
581 else if (bytes < 2*1024LL*1024LL*1024LL) {
582 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
583 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 584 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
d5d3721e
NB
585 cMiB/100 , cMiB % 100,
586 cMB/100, cMB % 100);
587 } else {
588 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
589 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 590 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
d5d3721e
NB
591 cGiB/100 , cGiB % 100,
592 cGB/100, cGB % 100);
593 }
cd29a5c8
NB
594 return buf;
595}
e0d19036
NB
596
597char *human_size_brief(long long bytes)
598{
599 static char buf[30];
e0d19036
NB
600
601 if (bytes < 5000*1024)
8f23b0b3 602 snprintf(buf, sizeof(buf), "%ld.%02ldKiB",
bd526cee 603 (long)(bytes>>10), (long)(((bytes&1023)*100+512)/1024)
e0d19036
NB
604 );
605 else if (bytes < 2*1024LL*1024LL*1024LL)
8f23b0b3 606 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
e0d19036 607 (long)(bytes>>20),
bd526cee 608 (long)((bytes&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
609 );
610 else
8f23b0b3 611 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
e0d19036 612 (long)(bytes>>30),
bd526cee 613 (long)(((bytes>>10)&0xfffff)+0x100000/200)/(0x100000/100)
e0d19036
NB
614 );
615 return buf;
616}
435d4ebb 617#endif
e0d19036 618
5f8097be
NB
619unsigned long long calc_array_size(int level, int raid_disks, int layout,
620 int chunksize, unsigned long long devsize)
621{
622 int data_disks = 0;
623 switch (level) {
624 case 0: data_disks = raid_disks; break;
625 case 1: data_disks = 1; break;
626 case 4:
627 case 5: data_disks = raid_disks - 1; break;
628 case 6: data_disks = raid_disks - 2; break;
629 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
630 break;
631 }
632 devsize &= ~(unsigned long long)((chunksize>>9)-1);
633 return data_disks * devsize;
634}
635
435d4ebb 636#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
dd0781e5 637int get_mdp_major(void)
98c6faba 638{
dd0781e5
NB
639static int mdp_major = -1;
640 FILE *fl;
98c6faba
NB
641 char *w;
642 int have_block = 0;
643 int have_devices = 0;
644 int last_num = -1;
dd0781e5
NB
645
646 if (mdp_major != -1)
647 return mdp_major;
648 fl = fopen("/proc/devices", "r");
98c6faba 649 if (!fl)
dd0781e5 650 return -1;
98c6faba
NB
651 while ((w = conf_word(fl, 1))) {
652 if (have_block && strcmp(w, "devices:")==0)
653 have_devices = 1;
654 have_block = (strcmp(w, "Block")==0);
655 if (isdigit(w[0]))
656 last_num = atoi(w);
657 if (have_devices && strcmp(w, "mdp")==0)
658 mdp_major = last_num;
659 free(w);
660 }
661 fclose(fl);
dd0781e5 662 return mdp_major;
98c6faba
NB
663}
664
665
e0d19036 666
e0d19036
NB
667char *get_md_name(int dev)
668{
669 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 670 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
e0d19036
NB
671 static char devname[50];
672 struct stat stb;
98c6faba 673 dev_t rdev;
dd0781e5 674 char *dn;
98c6faba
NB
675
676 if (dev < 0) {
dd0781e5
NB
677 int mdp = get_mdp_major();
678 if (mdp < 0) return NULL;
0df46c2a 679 rdev = makedev(mdp, (-1-dev)<<6);
8f23b0b3 680 snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev);
98c6faba
NB
681 if (stat(devname, &stb) == 0
682 && (S_IFMT&stb.st_mode) == S_IFBLK
683 && (stb.st_rdev == rdev))
684 return devname;
685 } else {
0df46c2a 686 rdev = makedev(MD_MAJOR, dev);
8f23b0b3 687 snprintf(devname, sizeof(devname), "/dev/md%d", dev);
98c6faba
NB
688 if (stat(devname, &stb) == 0
689 && (S_IFMT&stb.st_mode) == S_IFBLK
690 && (stb.st_rdev == rdev))
691 return devname;
692
8f23b0b3 693 snprintf(devname, sizeof(devname), "/dev/md/%d", dev);
98c6faba
NB
694 if (stat(devname, &stb) == 0
695 && (S_IFMT&stb.st_mode) == S_IFBLK
696 && (stb.st_rdev == rdev))
697 return devname;
698 }
16c6fa80 699 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
700 if (dn)
701 return dn;
8f23b0b3 702 snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev);
e0d19036 703 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
704 if (errno != EEXIST)
705 return NULL;
e0d19036
NB
706
707 if (stat(devname, &stb) == 0
708 && (S_IFMT&stb.st_mode) == S_IFBLK
709 && (stb.st_rdev == rdev))
710 return devname;
711 unlink(devname);
712 return NULL;
713}
714
715void put_md_name(char *name)
716{
717 if (strncmp(name, "/dev/.tmp.md", 12)==0)
718 unlink(name);
719}
ea24acd0 720
ea24acd0
NB
721int find_free_devnum(int use_partitions)
722{
723 int devnum;
724 for (devnum = 127; devnum != 128;
725 devnum = devnum ? devnum-1 : (1<<22)-1) {
726 char *dn;
727 int _devnum;
728
729 _devnum = use_partitions ? (-1-devnum) : devnum;
730 if (mddev_busy(_devnum))
731 continue;
732 /* make sure it is new to /dev too, at least as a
733 * non-standard */
734 dn = map_dev(dev2major(_devnum), dev2minor(_devnum), 0);
735 if (dn && ! is_standard(dn, NULL))
736 continue;
737 break;
738 }
739 if (devnum == 128)
740 return NoMdDev;
741 return use_partitions ? (-1-devnum) : devnum;
742}
435d4ebb 743#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
f9ce90ba 744
8b0dabea
NB
745int dev_open(char *dev, int flags)
746{
747 /* like 'open', but if 'dev' matches %d:%d, create a temp
748 * block device and open that
749 */
750 char *e;
751 int fd = -1;
752 char devname[32];
e81cdd9f 753 int major;
8b0dabea 754 int minor;
e81cdd9f
NB
755
756 if (!dev) return -1;
757
758 major = strtoul(dev, &e, 0);
8b0dabea
NB
759 if (e > dev && *e == ':' && e[1] &&
760 (minor = strtoul(e+1, &e, 0)) >= 0 &&
761 *e == 0) {
8c210183
NB
762 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
763 (int)getpid(), major, minor);
8b0dabea 764 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
6416d527 765 fd = open(devname, flags|O_DIRECT);
8b0dabea
NB
766 unlink(devname);
767 }
768 } else
6416d527 769 fd = open(dev, flags|O_DIRECT);
8b0dabea
NB
770 return fd;
771}
f9ce90ba 772
a931db9e
NB
773int open_dev_excl(int devnum)
774{
775 char buf[20];
776 int i;
777
778 sprintf(buf, "%d:%d", dev2major(devnum), dev2minor(devnum));
779 for (i=0 ; i<25 ; i++) {
780 int fd = dev_open(buf, O_RDWR|O_EXCL);
781 if (fd >= 0)
782 return fd;
783 if (errno != EBUSY)
784 return fd;
785 usleep(200000);
786 }
787 return -1;
788}
789
cdddbdbc 790struct superswitch *superlist[] = { &super0, &super1, &super_ddf, &super_imsm, NULL };
f9ce90ba 791
ea24acd0 792#if !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO)
f7dd881f 793
1686dc25 794struct supertype *super_by_fd(int fd)
f9ce90ba 795{
1686dc25
NB
796 mdu_array_info_t array;
797 int vers;
798 int minor;
799 struct supertype *st = NULL;
7e0f6979 800 struct mdinfo *sra;
142cb9e1 801 char *verstr;
1686dc25
NB
802 char version[20];
803 int i;
f7e7067b 804 char *subarray = NULL;
1686dc25
NB
805
806 sra = sysfs_read(fd, 0, GET_VERSION);
807
808 if (sra) {
7e0f6979
NB
809 vers = sra->array.major_version;
810 minor = sra->array.minor_version;
142cb9e1 811 verstr = sra->text_version;
1686dc25
NB
812 } else {
813 if (ioctl(fd, GET_ARRAY_INFO, &array))
814 array.major_version = array.minor_version = 0;
815 vers = array.major_version;
816 minor = array.minor_version;
142cb9e1 817 verstr = "";
6fbba4c9 818 }
82d9eba6 819
1686dc25
NB
820 if (vers != -1) {
821 sprintf(version, "%d.%d", vers, minor);
822 verstr = version;
6fbba4c9 823 }
3c558363 824 if (minor == -2 && is_subarray(verstr)) {
f7e7067b
NB
825 char *dev = verstr+1;
826 subarray = strchr(dev, '/');
827 int devnum;
828 if (subarray)
829 *subarray++ = '\0';
77472ff8 830 devnum = devname2devnum(dev);
f7e7067b
NB
831 subarray = strdup(subarray);
832 if (sra)
833 sysfs_free(sra);
834 sra = sysfs_read(-1, devnum, GET_VERSION);
835 verstr = sra->text_version ? : "-no-metadata-";
836 }
837
838 for (i = 0; st == NULL && superlist[i] ; i++)
839 st = superlist[i]->match_metadata_desc(verstr);
1686dc25
NB
840
841 if (sra)
842 sysfs_free(sra);
f7e7067b 843 if (st) {
3b0896f8 844 st->sb = NULL;
f7e7067b
NB
845 if (subarray) {
846 strncpy(st->subarray, subarray, 32);
847 st->subarray[31] = 0;
848 free(subarray);
849 } else
850 st->subarray[0] = 0;
851 }
82d9eba6 852 return st;
f9ce90ba 853}
ea24acd0
NB
854#endif /* !defined(MDASSEMBLE) || defined(MDASSEMBLE) && defined(MDASSEMBLE_AUTO) */
855
f9ce90ba 856
159c3a1a 857struct supertype *dup_super(struct supertype *orig)
3da92f27 858{
159c3a1a 859 struct supertype *st;
1686dc25 860
d2ca6449
NB
861 if (!orig)
862 return orig;
159c3a1a 863 st = malloc(sizeof(*st));
3da92f27
NB
864 if (!st)
865 return st;
ef609477 866 memset(st, 0, sizeof(*st));
159c3a1a
NB
867 st->ss = orig->ss;
868 st->max_devs = orig->max_devs;
869 st->minor_version = orig->minor_version;
f7e7067b 870 strcpy(st->subarray, orig->subarray);
159c3a1a
NB
871 st->sb = NULL;
872 st->info = NULL;
873 return st;
3da92f27
NB
874}
875
82d9eba6 876struct supertype *guess_super(int fd)
f9ce90ba
NB
877{
878 /* try each load_super to find the best match,
879 * and return the best superswitch
880 */
82d9eba6
NB
881 struct superswitch *ss;
882 struct supertype *st;
570c0542
NB
883 unsigned long besttime = 0;
884 int bestsuper = -1;
f9ce90ba
NB
885 int i;
886
82d9eba6 887 st = malloc(sizeof(*st));
f9ce90ba
NB
888 for (i=0 ; superlist[i]; i++) {
889 int rv;
890 ss = superlist[i];
ef609477 891 memset(st, 0, sizeof(*st));
3da92f27 892 rv = ss->load_super(st, fd, NULL);
570c0542
NB
893 if (rv == 0) {
894 struct mdinfo info;
3da92f27 895 st->ss->getinfo_super(st, &info);
570c0542
NB
896 if (bestsuper == -1 ||
897 besttime < info.array.ctime) {
898 bestsuper = i;
899 besttime = info.array.ctime;
570c0542 900 }
3da92f27 901 ss->free_super(st);
570c0542
NB
902 }
903 }
904 if (bestsuper != -1) {
905 int rv;
ef609477 906 memset(st, 0, sizeof(*st));
3da92f27 907 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 908 if (rv == 0) {
5e747af2 909 superlist[bestsuper]->free_super(st);
82d9eba6 910 return st;
f9ce90ba
NB
911 }
912 }
570c0542 913 free(st);
f9ce90ba
NB
914 return NULL;
915}
fe6729fa 916
beae1dfe
NB
917/* Return size of device in bytes */
918int get_dev_size(int fd, char *dname, unsigned long long *sizep)
919{
920 unsigned long long ldsize;
c2c9bb6f
NB
921 struct stat st;
922
923 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
924 ldsize = (unsigned long long)st.st_size;
925 else
beae1dfe
NB
926#ifdef BLKGETSIZE64
927 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
928#endif
929 {
930 unsigned long dsize;
931 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
932 ldsize = dsize;
933 ldsize <<= 9;
934 } else {
935 if (dname)
936 fprintf(stderr, Name ": Cannot get size of %s: %s\b",
937 dname, strerror(errno));
938 return 0;
939 }
940 }
941 *sizep = ldsize;
942 return 1;
943}
8fac0577 944
8382f19b
NB
945void get_one_disk(int mdfd, mdu_array_info_t *ainf, mdu_disk_info_t *disk)
946{
947 int d;
948 ioctl(mdfd, GET_ARRAY_INFO, ainf);
949 for (d = 0 ; d < ainf->raid_disks + ainf->nr_disks ; d++)
950 if (ioctl(mdfd, GET_DISK_INFO, disk) == 0)
951 return;
952}
63152c1b 953
a322f70c
DW
954int open_container(int fd)
955{
956 /* 'fd' is a block device. Find out if it is in use
957 * by a container, and return an open fd on that container.
958 */
959 char path[256];
960 char *e;
961 DIR *dir;
962 struct dirent *de;
963 int dfd, n;
964 char buf[200];
965 int major, minor;
966 struct stat st;
967
968 if (fstat(fd, &st) != 0)
969 return -1;
970 sprintf(path, "/sys/dev/block/%d:%d/holders",
971 (int)major(st.st_rdev), (int)minor(st.st_rdev));
972 e = path + strlen(path);
973
974 dir = opendir(path);
975 if (!dir)
976 return -1;
977 while ((de = readdir(dir))) {
978 if (de->d_ino == 0)
979 continue;
980 if (de->d_name[0] == '.')
981 continue;
982 sprintf(e, "/%s/dev", de->d_name);
983 dfd = open(path, O_RDONLY);
984 if (dfd < 0)
985 continue;
986 n = read(dfd, buf, sizeof(buf));
987 close(dfd);
988 if (n <= 0 || n >= sizeof(buf))
989 continue;
990 buf[n] = 0;
991 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
992 continue;
993 sprintf(buf, "%d:%d", major, minor);
994 dfd = dev_open(buf, O_RDONLY);
995 if (dfd >= 0) {
996 closedir(dir);
997 return dfd;
998 }
999 }
355726fa 1000 closedir(dir);
a322f70c
DW
1001 return -1;
1002}
1003
2f6079dc
NB
1004char *devnum2devname(int num)
1005{
1006 char name[100];
1007 if (num > 0)
1008 sprintf(name, "md%d", num);
1009 else
1010 sprintf(name, "md_d%d", -1-num);
1011 return strdup(name);
1012}
1013
77472ff8
NB
1014int devname2devnum(char *name)
1015{
1016 char *ep;
1017 int num;
1018 if (strncmp(name, "md_d", 4)==0)
1019 num = -1-strtoul(name+4, &ep, 10);
1020 else
1021 num = strtoul(name+2, &ep, 10);
1022 return num;
1023}
1024
2f6079dc
NB
1025int fd2devnum(int fd)
1026{
1027 struct stat stb;
1028 if (fstat(fd, &stb) == 0 &&
1029 (S_IFMT&stb.st_mode)==S_IFBLK) {
1030 if (major(stb.st_rdev) == MD_MAJOR)
1031 return minor(stb.st_rdev);
1032 else
1033 return -1- (minor(stb.st_rdev)>>6);
1034 }
1035 return -1;
1036}
1037
a931db9e
NB
1038int mdmon_running(int devnum)
1039{
1040 char path[100];
1041 char pid[10];
1042 int fd;
1043 int n;
1044 sprintf(path, "/var/run/mdadm/%s.pid", devnum2devname(devnum));
1045 fd = open(path, O_RDONLY, 0);
1046
1047 if (fd < 0)
1048 return 0;
1049 n = read(fd, pid, 9);
1050 close(fd);
1051 if (n <= 0)
1052 return 0;
1053 if (kill(atoi(pid), 0) == 0)
1054 return 1;
1055 return 0;
1056}
1057
1058int signal_mdmon(int devnum)
1059{
1060 char path[100];
1061 char pid[10];
1062 int fd;
1063 int n;
1064 sprintf(path, "/var/run/mdadm/%s.pid", devnum2devname(devnum));
1065 fd = open(path, O_RDONLY, 0);
1066
1067 if (fd < 0)
1068 return 0;
1069 n = read(fd, pid, 9);
1070 close(fd);
1071 if (n <= 0)
1072 return 0;
1073 if (kill(atoi(pid), SIGUSR1) == 0)
1074 return 1;
1075 return 0;
1076}
1077
8850ee3e
N
1078int start_mdmon(int devnum)
1079{
1080 int i;
44d2e365 1081 int len;
9fe32043
N
1082 pid_t pid;
1083 int status;
44d2e365
N
1084 char pathbuf[1024];
1085 char *paths[4] = {
1086 pathbuf,
1087 "/sbin/mdmon",
1088 "mdmon",
1089 NULL
1090 };
8850ee3e
N
1091
1092 if (env_no_mdmon())
1093 return 0;
1094
44d2e365
N
1095 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf));
1096 if (len > 0) {
1097 char *sl;
1098 pathbuf[len] = 0;
1099 sl = strrchr(pathbuf, '/');
1100 if (sl)
1101 sl++;
1102 else
1103 sl = pathbuf;
1104 strcpy(sl, "mdmon");
1105 } else
1106 pathbuf[0] = '\0';
1107
8850ee3e
N
1108 switch(fork()) {
1109 case 0:
1110 /* FIXME yuk. CLOSE_EXEC?? */
1111 for (i=3; i < 100; i++)
1112 close(i);
44d2e365
N
1113 for (i=0; paths[i]; i++)
1114 if (paths[i][0])
1115 execl(paths[i], "mdmon",
1116 map_dev(dev2major(devnum),
1117 dev2minor(devnum),
1118 1), NULL);
8850ee3e
N
1119 exit(1);
1120 case -1: fprintf(stderr, Name ": cannot run mdmon. "
1121 "Array remains readonly\n");
1122 return -1;
9fe32043
N
1123 default: /* parent - good */
1124 pid = wait(&status);
1125 if (pid < 0 || status != 0)
1126 return -1;
8850ee3e
N
1127 }
1128 return 0;
1129}
1130
5dcfcb71
DW
1131int env_no_mdmon(void)
1132{
1133 char *val = getenv("MDADM_NO_MDMON");
1134
1135 if (val && atoi(val) == 1)
1136 return 1;
1137
1138 return 0;
1139}
1140
a931db9e 1141
edd8d13c
NB
1142int flush_metadata_updates(struct supertype *st)
1143{
1144 int sfd;
1145 if (!st->updates) {
1146 st->update_tail = NULL;
1147 return -1;
1148 }
1149
1150 sfd = connect_monitor(devnum2devname(st->container_dev));
1151 if (sfd < 0)
1152 return -1;
1153
1154 while (st->updates) {
1155 struct metadata_update *mu = st->updates;
1156 st->updates = mu->next;
1157
1158 send_message(sfd, mu, 0);
1159 wait_reply(sfd, 0);
1160 free(mu->buf);
1161 free(mu);
1162 }
1163 ack(sfd, 0);
1164 wait_reply(sfd, 0);
1165 close(sfd);
1166 st->update_tail = NULL;
1167 return 0;
1168}
1169
1170void append_metadata_update(struct supertype *st, void *buf, int len)
1171{
1172
1173 struct metadata_update *mu = malloc(sizeof(*mu));
1174
1175 mu->buf = buf;
1176 mu->len = len;
1177 mu->space = NULL;
1178 mu->next = NULL;
1179 *st->update_tail = mu;
1180 st->update_tail = &mu->next;
1181}
1182
a931db9e 1183
fe6729fa
NB
1184#ifdef __TINYC__
1185/* tinyc doesn't optimize this check in ioctl.h out ... */
1186unsigned int __invalid_size_argument_for_IOC = 0;
1187#endif
1188