]> git.ipfire.org Git - thirdparty/mdadm.git/blob - util.c
Don't #include blkpg.h
[thirdparty/mdadm.git] / util.c
1 /*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
4 * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
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
30 #include "mdadm.h"
31 #include "md_p.h"
32 #include <sys/utsname.h>
33 #include <ctype.h>
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 */
43 struct blkpg_ioctl_arg {
44 int op;
45 int flags;
46 int datalen;
47 void *data;
48 };
49
50 /* The subfunctions (for the op field) */
51 #define BLKPG_ADD_PARTITION 1
52 #define BLKPG_DEL_PARTITION 2
53
54 /* Sizes of name fields. Unused at present. */
55 #define BLKPG_DEVNAMELTH 64
56 #define BLKPG_VOLNAMELTH 64
57
58 /* The data structure for ADD_PARTITION and DEL_PARTITION */
59 struct 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 };
67
68 /*
69 * Parse a 128 bit uuid in 4 integers
70 * format is 32 hexx nibbles with options :.<space> separator
71 * If not exactly 32 hex digits are found, return 0
72 * else return 1
73 */
74 int parse_uuid(char *str, int uuid[4])
75 {
76 int hit = 0; /* number of Hex digIT */
77 int i;
78 char c;
79 for (i=0; i<4; i++) uuid[i]=0;
80
81 while ((c= *str++)) {
82 int n;
83 if (c>='0' && c<='9')
84 n = c-'0';
85 else if (c>='a' && c <= 'f')
86 n = 10 + c - 'a';
87 else if (c>='A' && c <= 'F')
88 n = 10 + c - 'A';
89 else if (strchr(":. -", c))
90 continue;
91 else return 0;
92
93 if (hit<32) {
94 uuid[hit/8] <<= 4;
95 uuid[hit/8] += n;
96 }
97 hit++;
98 }
99 if (hit == 32)
100 return 1;
101 return 0;
102
103 }
104
105
106 /*
107 * Get the md version number.
108 * We use the RAID_VERSION ioctl if it is supported
109 * If not, but we have a block device with major '9', we assume
110 * 0.36.0
111 *
112 * Return version number as 24 but number - assume version parts
113 * always < 255
114 */
115
116 int md_get_version(int fd)
117 {
118 struct stat stb;
119 mdu_version_t vers;
120
121 if (fstat(fd, &stb)<0)
122 return -1;
123 if ((S_IFMT&stb.st_mode) != S_IFBLK)
124 return -1;
125
126 if (ioctl(fd, RAID_VERSION, &vers) == 0)
127 return (vers.major*10000) + (vers.minor*100) + vers.patchlevel;
128 if (errno == EACCES)
129 return -1;
130 if (major(stb.st_rdev) == MD_MAJOR)
131 return (3600);
132 return -1;
133 }
134
135
136 int get_linux_version()
137 {
138 struct utsname name;
139 char *cp;
140 int a,b,c;
141 if (uname(&name) <0)
142 return -1;
143
144 cp = name.release;
145 a = strtoul(cp, &cp, 10);
146 if (*cp != '.') return -1;
147 b = strtoul(cp+1, &cp, 10);
148 if (*cp != '.') return -1;
149 c = strtoul(cp+1, NULL, 10);
150
151 return (a*1000000)+(b*1000)+c;
152 }
153
154 void remove_partitions(int fd)
155 {
156 /* remove partitions from this block devices.
157 * This is used for components added to an array
158 */
159 #ifdef BLKPG_DEL_PARTITION
160 struct blkpg_ioctl_arg a;
161 struct blkpg_partition p;
162
163 a.op = BLKPG_DEL_PARTITION;
164 a.data = (void*)&p;
165 a.datalen = sizeof(p);
166 a.flags = 0;
167 memset(a.data, 0, a.datalen);
168 for (p.pno=0; p.pno < 16; p.pno++)
169 ioctl(fd, BLKPG, &a);
170 #endif
171 }
172
173 int enough(int level, int raid_disks, int layout,
174 char *avail, int avail_disks)
175 {
176 int copies, first;
177 switch (level) {
178 case 10:
179 /* This is the tricky one - we need to check
180 * which actual disks are present.
181 */
182 copies = (layout&255)* (layout>>8);
183 first=0;
184 do {
185 /* there must be one of the 'copies' form 'first' */
186 int n = copies;
187 int cnt=0;
188 while (n--) {
189 if (avail[first])
190 cnt++;
191 first = (first+1) % raid_disks;
192 }
193 if (cnt == 0)
194 return 0;
195
196 } while (first != 0);
197 return 1;
198
199 case -4:
200 return avail_disks>= 1;
201 case -1:
202 case 0:
203 return avail_disks == raid_disks;
204 case 1:
205 return avail_disks >= 1;
206 case 4:
207 case 5:
208 return avail_disks >= raid_disks-1;
209 case 6:
210 return avail_disks >= raid_disks-2;
211 default:
212 return 0;
213 }
214 }
215
216 int same_uuid(int a[4], int b[4], int swapuuid)
217 {
218 if (swapuuid) {
219 /* parse uuids are hostendian.
220 * uuid's from some superblocks are big-ending
221 * if there is a difference, we need to swap..
222 */
223 unsigned char *ac = (unsigned char *)a;
224 unsigned char *bc = (unsigned char *)b;
225 int i;
226 for (i=0; i<16; i+= 4) {
227 if (ac[i+0] != bc[i+3] ||
228 ac[i+1] != bc[i+2] ||
229 ac[i+2] != bc[i+1] ||
230 ac[i+3] != bc[i+0])
231 return 0;
232 }
233 return 1;
234 } else {
235 if (a[0]==b[0] &&
236 a[1]==b[1] &&
237 a[2]==b[2] &&
238 a[3]==b[3])
239 return 1;
240 return 0;
241 }
242 }
243
244 int check_ext2(int fd, char *name)
245 {
246 /*
247 * Check for an ext2fs file system.
248 * Superblock is always 1K at 1K offset
249 *
250 * s_magic is le16 at 56 == 0xEF53
251 * report mtime - le32 at 44
252 * blocks - le32 at 4
253 * logblksize - le32 at 24
254 */
255 unsigned char sb[1024];
256 time_t mtime;
257 int size, bsize;
258 if (lseek(fd, 1024,0)!= 1024)
259 return 0;
260 if (read(fd, sb, 1024)!= 1024)
261 return 0;
262 if (sb[56] != 0x53 || sb[57] != 0xef)
263 return 0;
264
265 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
266 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
267 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
268 fprintf(stderr, Name ": %s appears to contain an ext2fs file system\n",
269 name);
270 fprintf(stderr," size=%dK mtime=%s",
271 size*(1<<bsize), ctime(&mtime));
272 return 1;
273 }
274
275 int check_reiser(int fd, char *name)
276 {
277 /*
278 * superblock is at 64K
279 * size is 1024;
280 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
281 *
282 */
283 unsigned char sb[1024];
284 unsigned long size;
285 if (lseek(fd, 64*1024, 0) != 64*1024)
286 return 0;
287 if (read(fd, sb, 1024) != 1024)
288 return 0;
289 if (strncmp((char*)sb+52, "ReIsErFs",8)!=0 &&
290 strncmp((char*)sb+52, "ReIsEr2Fs",9)!=0)
291 return 0;
292 fprintf(stderr, Name ": %s appears to contain a reiserfs file system\n",name);
293 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
294 fprintf(stderr, " size = %luK\n", size*4);
295
296 return 1;
297 }
298
299 int check_raid(int fd, char *name)
300 {
301 void *super;
302 struct mdinfo info;
303 time_t crtime;
304 char *level;
305 struct supertype *st = guess_super(fd);
306
307 if (!st) return 0;
308 st->ss->load_super(st, fd, &super, name);
309 /* Looks like a raid array .. */
310 fprintf(stderr, Name ": %s appears to be part of a raid array:\n",
311 name);
312 st->ss->getinfo_super(&info, super);
313 free(super);
314 crtime = info.array.ctime;
315 level = map_num(pers, info.array.level);
316 if (!level) level = "-unknown-";
317 fprintf(stderr, " level=%s devices=%d ctime=%s",
318 level, info.array.raid_disks, ctime(&crtime));
319 return 1;
320 }
321
322 int ask(char *mesg)
323 {
324 char *add = "";
325 int i;
326 for (i=0; i<5; i++) {
327 char buf[100];
328 fprintf(stderr, "%s%s", mesg, add);
329 fflush(stderr);
330 if (fgets(buf, 100, stdin)==NULL)
331 return 0;
332 if (buf[0]=='y' || buf[0]=='Y')
333 return 1;
334 if (buf[0]=='n' || buf[0]=='N')
335 return 0;
336 add = "(y/n) ";
337 }
338 fprintf(stderr, Name ": assuming 'no'\n");
339 return 0;
340 }
341
342 char *map_num(mapping_t *map, int num)
343 {
344 while (map->name) {
345 if (map->num == num)
346 return map->name;
347 map++;
348 }
349 return NULL;
350 }
351
352 int map_name(mapping_t *map, char *name)
353 {
354 while (map->name) {
355 if (strcmp(map->name, name)==0)
356 return map->num;
357 map++;
358 }
359 return UnSet;
360 }
361
362
363 int is_standard(char *dev, int *nump)
364 {
365 /* tests if dev is a "standard" md dev name.
366 * i.e if the last component is "/dNN" or "/mdNN",
367 * where NN is a string of digits
368 */
369 char *d = strrchr(dev, '/');
370 int type=0;
371 int num;
372 if (!d)
373 return 0;
374 if (strncmp(d, "/d",2)==0)
375 d += 2, type=1; /* /dev/md/dN{pM} */
376 else if (strncmp(d, "/md_d", 5)==0)
377 d += 5, type=1; /* /dev/md_dNpM */
378 else if (strncmp(d, "/md", 3)==0)
379 d += 3, type=-1; /* /dev/mdN */
380 else if (d-dev > 3 && strncmp(d-2, "md/", 3)==0)
381 d += 1, type=-1; /* /dev/md/N */
382 else
383 return 0;
384 if (!*d)
385 return 0;
386 num = atoi(d);
387 while (isdigit(*d))
388 d++;
389 if (*d)
390 return 0;
391 if (nump) *nump = num;
392
393 return type;
394 }
395
396
397 /*
398 * convert a major/minor pair for a block device into a name in /dev, if possible.
399 * On the first call, walk /dev collecting name.
400 * Put them in a simple linked listfor now.
401 */
402 struct devmap {
403 int major, minor;
404 char *name;
405 struct devmap *next;
406 } *devlist = NULL;
407 int devlist_ready = 0;
408
409 int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s)
410 {
411 struct stat st;
412 if (S_ISLNK(stb->st_mode)) {
413 stat(name, &st);
414 stb = &st;
415 }
416
417 if ((stb->st_mode&S_IFMT)== S_IFBLK) {
418 char *n = strdup(name);
419 struct devmap *dm = malloc(sizeof(*dm));
420 if (strncmp(n, "/dev/./", 7)==0)
421 strcpy(n+4, name+6);
422 if (dm) {
423 dm->major = major(stb->st_rdev);
424 dm->minor = minor(stb->st_rdev);
425 dm->name = n;
426 dm->next = devlist;
427 devlist = dm;
428 }
429 }
430 return 0;
431 }
432
433 #ifndef HAVE_NFTW
434 #ifdef HAVE_FTW
435 int add_dev_1(const char *name, const struct stat *stb, int flag)
436 {
437 return add_dev(name, stb, flag, NULL);
438 }
439 int nftw(const char *path, int (*han)(const char *name, const struct stat *stb, int flag, struct FTW *s), int nopenfd, int flags)
440 {
441 return ftw(path, add_dev_1, nopenfd);
442 }
443 #else
444 int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s)
445 {
446 return 0;
447 }
448 int nftw(const char *path, int (*han)(const char *name, const struct stat *stb, int flag, struct FTW *s), int nopenfd, int flags)
449 {
450 return 0;
451 }
452 #endif /* HAVE_FTW */
453 #endif /* HAVE_NFTW */
454
455 /*
456 * Find a block device with the right major/minor number.
457 * If we find multiple names, choose the shortest.
458 * If we find a non-standard name, it is probably there
459 * deliberately so prefer it over a standard name.
460 * This applies only to names for MD devices.
461 */
462 char *map_dev(int major, int minor, int create)
463 {
464 struct devmap *p;
465 char *std = NULL, *nonstd=NULL;
466 int did_check = 0;
467
468 if (major == 0 && minor == 0)
469 return NULL;
470
471 retry:
472 if (!devlist_ready) {
473 char *dev = "/dev";
474 struct stat stb;
475 while(devlist) {
476 struct devmap *d = devlist;
477 devlist = d->next;
478 free(d->name);
479 free(d);
480 }
481 if (lstat(dev, &stb)==0 &&
482 S_ISLNK(stb.st_mode))
483 dev = "/dev/.";
484 nftw(dev, add_dev, 10, FTW_PHYS);
485 devlist_ready=1;
486 did_check = 1;
487 }
488
489 for (p=devlist; p; p=p->next)
490 if (p->major == major &&
491 p->minor == minor) {
492 if (is_standard(p->name, NULL)) {
493 if (std == NULL ||
494 strlen(p->name) < strlen(std))
495 std = p->name;
496 } else {
497 if (nonstd == NULL ||
498 strlen(p->name) < strlen(nonstd))
499 nonstd = p->name;
500 }
501 }
502 if (!std && !nonstd && !did_check) {
503 devlist_ready = 0;
504 goto retry;
505 }
506 if (create && !std && !nonstd) {
507 static char buf[30];
508 snprintf(buf, sizeof(buf), "%d:%d", major, minor);
509 nonstd = buf;
510 }
511
512 return nonstd ? nonstd : std;
513 }
514
515 unsigned long calc_csum(void *super, int bytes)
516 {
517 unsigned long long newcsum = 0;
518 int i;
519 unsigned int csum;
520 unsigned int *superc = (unsigned int*) super;
521
522 for(i=0; i<bytes/4; i++)
523 newcsum+= superc[i];
524 csum = (newcsum& 0xffffffff) + (newcsum>>32);
525 #ifdef __alpha__
526 /* The in-kernel checksum calculation is always 16bit on
527 * the alpha, though it is 32 bit on i386...
528 * I wonder what it is elsewhere... (it uses and API in
529 * a way that it shouldn't).
530 */
531 csum = (csum & 0xffff) + (csum >> 16);
532 csum = (csum & 0xffff) + (csum >> 16);
533 #endif
534 return csum;
535 }
536
537 char *human_size(long long bytes)
538 {
539 static char buf[30];
540
541 /* We convert bytes to either centi-M{ega,ibi}bytes or
542 * centi-G{igi,ibi}bytes, with appropriate rounding,
543 * and then print 1/100th of those as a decimal.
544 * We allow upto 2048Megabytes before converting to
545 * gigabytes, as that shows more precision and isn't
546 * too large a number.
547 * Terrabytes are not yet handled.
548 */
549
550 if (bytes < 5000*1024)
551 buf[0]=0;
552 else if (bytes < 2*1024LL*1024LL*1024LL) {
553 long cMiB = (bytes / ( (1LL<<20) / 200LL ) +1) /2;
554 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
555 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
556 cMiB/100 , cMiB % 100,
557 cMB/100, cMB % 100);
558 } else {
559 long cGiB = (bytes / ( (1LL<<30) / 200LL ) +1) /2;
560 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
561 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
562 cGiB/100 , cGiB % 100,
563 cGB/100, cGB % 100);
564 }
565 return buf;
566 }
567
568 char *human_size_brief(long long bytes)
569 {
570 static char buf[30];
571
572
573 if (bytes < 5000*1024)
574 snprintf(buf, sizeof(buf), "%ld.%02ldKiB",
575 (long)(bytes>>10), (long)(((bytes&1023)*100+512)/1024)
576 );
577 else if (bytes < 2*1024LL*1024LL*1024LL)
578 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
579 (long)(bytes>>20),
580 (long)((bytes&0xfffff)+0x100000/200)/(0x100000/100)
581 );
582 else
583 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
584 (long)(bytes>>30),
585 (long)(((bytes>>10)&0xfffff)+0x100000/200)/(0x100000/100)
586 );
587 return buf;
588 }
589
590 int get_mdp_major(void)
591 {
592 static int mdp_major = -1;
593 FILE *fl;
594 char *w;
595 int have_block = 0;
596 int have_devices = 0;
597 int last_num = -1;
598
599 if (mdp_major != -1)
600 return mdp_major;
601 fl = fopen("/proc/devices", "r");
602 if (!fl)
603 return -1;
604 while ((w = conf_word(fl, 1))) {
605 if (have_block && strcmp(w, "devices:")==0)
606 have_devices = 1;
607 have_block = (strcmp(w, "Block")==0);
608 if (isdigit(w[0]))
609 last_num = atoi(w);
610 if (have_devices && strcmp(w, "mdp")==0)
611 mdp_major = last_num;
612 free(w);
613 }
614 fclose(fl);
615 return mdp_major;
616 }
617
618
619
620 char *get_md_name(int dev)
621 {
622 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
623 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
624 static char devname[50];
625 struct stat stb;
626 dev_t rdev;
627 char *dn;
628
629 if (dev < 0) {
630 int mdp = get_mdp_major();
631 if (mdp < 0) return NULL;
632 rdev = makedev(mdp, (-1-dev)<<6);
633 snprintf(devname, sizeof(devname), "/dev/md/d%d", -1-dev);
634 if (stat(devname, &stb) == 0
635 && (S_IFMT&stb.st_mode) == S_IFBLK
636 && (stb.st_rdev == rdev))
637 return devname;
638 } else {
639 rdev = makedev(MD_MAJOR, dev);
640 snprintf(devname, sizeof(devname), "/dev/md%d", dev);
641 if (stat(devname, &stb) == 0
642 && (S_IFMT&stb.st_mode) == S_IFBLK
643 && (stb.st_rdev == rdev))
644 return devname;
645
646 snprintf(devname, sizeof(devname), "/dev/md/%d", dev);
647 if (stat(devname, &stb) == 0
648 && (S_IFMT&stb.st_mode) == S_IFBLK
649 && (stb.st_rdev == rdev))
650 return devname;
651 }
652 dn = map_dev(major(rdev), minor(rdev), 0);
653 if (dn)
654 return dn;
655 snprintf(devname, sizeof(devname), "/dev/.tmp.md%d", dev);
656 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
657 if (errno != EEXIST)
658 return NULL;
659
660 if (stat(devname, &stb) == 0
661 && (S_IFMT&stb.st_mode) == S_IFBLK
662 && (stb.st_rdev == rdev))
663 return devname;
664 unlink(devname);
665 return NULL;
666 }
667
668 void put_md_name(char *name)
669 {
670 if (strncmp(name, "/dev/.tmp.md", 12)==0)
671 unlink(name);
672 }
673
674 int dev_open(char *dev, int flags)
675 {
676 /* like 'open', but if 'dev' matches %d:%d, create a temp
677 * block device and open that
678 */
679 char *e;
680 int fd = -1;
681 char devname[32];
682 int major;
683 int minor;
684
685 if (!dev) return -1;
686
687 major = strtoul(dev, &e, 0);
688 if (e > dev && *e == ':' && e[1] &&
689 (minor = strtoul(e+1, &e, 0)) >= 0 &&
690 *e == 0) {
691 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d", major, minor);
692 if (mknod(devname, S_IFBLK|0600, makedev(major, minor))==0) {
693 fd = open(devname, flags);
694 unlink(devname);
695 }
696 } else
697 fd = open(dev, flags);
698 return fd;
699 }
700
701 struct superswitch *superlist[] = { &super0, &super1, NULL };
702
703 struct supertype *super_by_version(int vers, int minor)
704 {
705 struct supertype *st = malloc(sizeof(*st));
706 if (!st) return st;
707 if (vers == 0) {
708 st->ss = &super0;
709 st->max_devs = MD_SB_DISKS;
710 }
711
712 if (vers == 1) {
713 st->ss = &super1;
714 st->max_devs = 384;
715 }
716 st->minor_version = minor;
717 return st;
718 }
719
720 struct supertype *guess_super(int fd)
721 {
722 /* try each load_super to find the best match,
723 * and return the best superswitch
724 */
725 struct superswitch *ss;
726 struct supertype *st;
727 unsigned long besttime = 0;
728 int bestsuper = -1;
729
730 void *sbp = NULL;
731 int i;
732
733 st = malloc(sizeof(*st));
734 memset(st, 0, sizeof(*st));
735 for (i=0 ; superlist[i]; i++) {
736 int rv;
737 ss = superlist[i];
738 st->ss = NULL;
739 rv = ss->load_super(st, fd, &sbp, NULL);
740 if (rv == 0) {
741 struct mdinfo info;
742 ss->getinfo_super(&info, sbp);
743 if (bestsuper == -1 ||
744 besttime < info.array.ctime) {
745 bestsuper = i;
746 besttime = info.array.ctime;
747 }
748 free(sbp);
749 }
750 }
751 if (bestsuper != -1) {
752 int rv;
753 st->ss = NULL;
754 rv = superlist[bestsuper]->load_super(st, fd, &sbp, NULL);
755 if (rv == 0) {
756 free(sbp);
757 return st;
758 }
759 }
760 free(st);
761 return NULL;
762 }
763
764
765 #ifdef __TINYC__
766 /* tinyc doesn't optimize this check in ioctl.h out ... */
767 unsigned int __invalid_size_argument_for_IOC = 0;
768 #endif
769