change back 0644 permission for Grow.c
[thirdparty/mdadm.git] / util.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
6f02172d 4 * Copyright (C) 2001-2013 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>
a7dec3fd 31#include <sys/resource.h>
a9c15847
N
32#include <sys/vfs.h>
33#include <linux/magic.h>
986b8688 34#include <poll.h>
98c6faba 35#include <ctype.h>
a322f70c 36#include <dirent.h>
a931db9e 37#include <signal.h>
7716570e 38#include <dlfcn.h>
7716570e 39
0a816ef9
NB
40
41/*
42 * following taken from linux/blkpg.h because they aren't
43 * anywhere else and it isn't safe to #include linux/ * stuff.
44 */
45
46#define BLKPG _IO(0x12,105)
47
48/* The argument structure */
49struct blkpg_ioctl_arg {
5d500228
N
50 int op;
51 int flags;
52 int datalen;
53 void *data;
0a816ef9
NB
54};
55
56/* The subfunctions (for the op field) */
57#define BLKPG_ADD_PARTITION 1
58#define BLKPG_DEL_PARTITION 2
59
60/* Sizes of name fields. Unused at present. */
61#define BLKPG_DEVNAMELTH 64
62#define BLKPG_VOLNAMELTH 64
63
64/* The data structure for ADD_PARTITION and DEL_PARTITION */
65struct blkpg_partition {
66 long long start; /* starting offset in bytes */
67 long long length; /* length in bytes */
68 int pno; /* partition number */
69 char devname[BLKPG_DEVNAMELTH]; /* partition name, like sda5 or c0d1p2,
70 to be used in kernel messages */
71 char volname[BLKPG_VOLNAMELTH]; /* volume label */
72};
64c4757e 73
0f22b998 74#include "part.h"
056b331e
N
75
76/* Force a compilation error if condition is true */
77#define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
78
79/* Force a compilation error if condition is true, but also produce a
80 result (of value 0 and type size_t), so the expression can be used
81 e.g. in a structure initializer (or where-ever else comma expressions
82 aren't permitted). */
83#define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))
84
d15a1f72 85static int is_dlm_hooks_ready = 0;
7071320a
N
86
87int dlm_funs_ready(void)
88{
89 return is_dlm_hooks_ready ? 1 : 0;
90}
91
7071320a 92static struct dlm_hooks *dlm_hooks = NULL;
d15a1f72
GJ
93struct dlm_lock_resource *dlm_lock_res = NULL;
94static int ast_called = 0;
95
96struct dlm_lock_resource {
97 dlm_lshandle_t *ls;
98 struct dlm_lksb lksb;
99};
100
d15a1f72
GJ
101/* Using poll(2) to wait for and dispatch ASTs */
102static int poll_for_ast(dlm_lshandle_t ls)
103{
104 struct pollfd pfd;
105
106 pfd.fd = dlm_hooks->ls_get_fd(ls);
107 pfd.events = POLLIN;
108
109 while (!ast_called)
110 {
111 if (poll(&pfd, 1, 0) < 0)
112 {
113 perror("poll");
114 return -1;
115 }
116 dlm_hooks->dispatch(dlm_hooks->ls_get_fd(ls));
117 }
118 ast_called = 0;
119
120 return 0;
121}
122
123static void dlm_ast(void *arg)
124{
125 ast_called = 1;
126}
127
81a8a694 128static char *cluster_name = NULL;
d15a1f72 129/* Create the lockspace, take bitmapXXX locks on all the bitmaps. */
81a8a694 130int cluster_get_dlmlock(int *lockid)
d15a1f72
GJ
131{
132 int ret = -1;
133 char str[64];
134 int flags = LKF_NOQUEUE;
135
81a8a694
GJ
136 ret = get_cluster_name(&cluster_name);
137 if (ret) {
138 pr_err("The md can't get cluster name\n");
139 return -1;
140 }
141
d15a1f72 142 dlm_lock_res = xmalloc(sizeof(struct dlm_lock_resource));
81a8a694 143 dlm_lock_res->ls = dlm_hooks->create_lockspace(cluster_name, O_RDWR);
d15a1f72 144 if (!dlm_lock_res->ls) {
81a8a694 145 pr_err("%s failed to create lockspace\n", cluster_name);
32539f74 146 return -ENOMEM;
d15a1f72
GJ
147 }
148
81a8a694 149 snprintf(str, 64, "bitmap%s", cluster_name);
d15a1f72
GJ
150 ret = dlm_hooks->ls_lock(dlm_lock_res->ls, LKM_PWMODE, &dlm_lock_res->lksb,
151 flags, str, strlen(str), 0, dlm_ast,
152 dlm_lock_res, NULL, NULL);
153 if (ret) {
154 pr_err("error %d when get PW mode on lock %s\n", errno, str);
32539f74
GJ
155 dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1);
156 return ret;
d15a1f72
GJ
157 }
158
159 /* Wait for it to complete */
160 poll_for_ast(dlm_lock_res->ls);
161 *lockid = dlm_lock_res->lksb.sb_lkid;
162
32539f74 163 return dlm_lock_res->lksb.sb_status;
d15a1f72
GJ
164}
165
81a8a694 166int cluster_release_dlmlock(int lockid)
d15a1f72
GJ
167{
168 int ret = -1;
169
81a8a694
GJ
170 if (!cluster_name)
171 return -1;
172
d15a1f72
GJ
173 ret = dlm_hooks->ls_unlock(dlm_lock_res->ls, lockid, 0,
174 &dlm_lock_res->lksb, dlm_lock_res);
175 if (ret) {
176 pr_err("error %d happened when unlock\n", errno);
177 /* XXX make sure the lock is unlocked eventually */
178 goto out;
179 }
180
181 /* Wait for it to complete */
182 poll_for_ast(dlm_lock_res->ls);
183
184 errno = dlm_lock_res->lksb.sb_status;
185 if (errno != EUNLOCK) {
186 pr_err("error %d happened in ast when unlock lockspace\n", errno);
187 /* XXX make sure the lockspace is unlocked eventually */
188 goto out;
189 }
190
81a8a694 191 ret = dlm_hooks->release_lockspace(cluster_name, dlm_lock_res->ls, 1);
d15a1f72
GJ
192 if (ret) {
193 pr_err("error %d happened when release lockspace\n", errno);
194 /* XXX make sure the lockspace is released eventually */
195 goto out;
196 }
197 free(dlm_lock_res);
198
199out:
200 return ret;
201}
202
3ab8f4bf
JS
203int md_array_active(int fd)
204{
205 struct mdinfo *sra;
206 struct mdu_array_info_s array;
207 int ret;
208
209 sra = sysfs_read(fd, NULL, GET_ARRAY_STATE);
210 if (sra) {
211 if (sra->array_state != ARRAY_CLEAR &&
212 sra->array_state != ARRAY_INACTIVE &&
213 sra->array_state != ARRAY_UNKNOWN_STATE)
214 ret = 0;
215 else
216 ret = -ENODEV;
217
218 free(sra);
219 } else {
220 /*
221 * GET_ARRAY_INFO doesn't provide access to the proper state
222 * information, so fallback to a basic check for raid_disks != 0
223 */
224 ret = ioctl(fd, GET_ARRAY_INFO, &array);
225 }
226
227 return !ret;
228}
229
9cd39f01
JS
230/*
231 * Get array info from the kernel. Longer term we want to deprecate the
232 * ioctl and get it from sysfs.
233 */
234int md_get_array_info(int fd, struct mdu_array_info_s *array)
235{
236 return ioctl(fd, GET_ARRAY_INFO, array);
237}
238
018a4882
JS
239/*
240 * Set array info
241 */
242int md_set_array_info(int fd, struct mdu_array_info_s *array)
243{
244 return ioctl(fd, SET_ARRAY_INFO, array);
245}
246
d97572f5
JS
247/*
248 * Get disk info from the kernel.
249 */
250int md_get_disk_info(int fd, struct mdu_disk_info_s *disk)
251{
252 return ioctl(fd, GET_DISK_INFO, disk);
253}
254
64c4757e
NB
255/*
256 * Parse a 128 bit uuid in 4 integers
257 * format is 32 hexx nibbles with options :.<space> separator
258 * If not exactly 32 hex digits are found, return 0
259 * else return 1
260 */
261int parse_uuid(char *str, int uuid[4])
262{
aba69144
NB
263 int hit = 0; /* number of Hex digIT */
264 int i;
265 char c;
ca3b6696
N
266 for (i = 0; i < 4; i++)
267 uuid[i] = 0;
aba69144 268
ca3b6696 269 while ((c = *str++) != 0) {
aba69144 270 int n;
ca3b6696 271 if (c >= '0' && c <= '9')
aba69144 272 n = c-'0';
ca3b6696 273 else if (c >= 'a' && c <= 'f')
aba69144 274 n = 10 + c - 'a';
ca3b6696 275 else if (c >= 'A' && c <= 'F')
aba69144
NB
276 n = 10 + c - 'A';
277 else if (strchr(":. -", c))
278 continue;
279 else return 0;
280
281 if (hit<32) {
282 uuid[hit/8] <<= 4;
283 uuid[hit/8] += n;
284 }
285 hit++;
82b27616 286 }
aba69144
NB
287 if (hit == 32)
288 return 1;
289 return 0;
64c4757e
NB
290}
291
64c4757e
NB
292int get_linux_version()
293{
294 struct utsname name;
98c6faba 295 char *cp;
f161d047 296 int a = 0, b = 0,c = 0;
64c4757e
NB
297 if (uname(&name) <0)
298 return -1;
299
98c6faba
NB
300 cp = name.release;
301 a = strtoul(cp, &cp, 10);
f161d047
N
302 if (*cp == '.')
303 b = strtoul(cp+1, &cp, 10);
304 if (*cp == '.')
305 c = strtoul(cp+1, &cp, 10);
98c6faba 306
682c7051 307 return (a*1000000)+(b*1000)+c;
64c4757e
NB
308}
309
bc77ed53
DW
310int mdadm_version(char *version)
311{
312 int a, b, c;
313 char *cp;
314
315 if (!version)
316 version = Version;
317
318 cp = strchr(version, '-');
319 if (!cp || *(cp+1) != ' ' || *(cp+2) != 'v')
320 return -1;
321 cp += 3;
322 a = strtoul(cp, &cp, 10);
323 if (*cp != '.')
324 return -1;
325 b = strtoul(cp+1, &cp, 10);
326 if (*cp == '.')
327 c = strtoul(cp+1, &cp, 10);
328 else
329 c = 0;
330 if (*cp != ' ' && *cp != '-')
331 return -1;
332 return (a*1000000)+(b*1000)+c;
333}
334
51d4261c 335unsigned long long parse_size(char *size)
84e11361
N
336{
337 /* parse 'size' which should be a number optionally
338 * followed by 'K', 'M', or 'G'.
339 * Without a suffix, K is assumed.
340 * Number returned is in sectors (half-K)
822e393a 341 * INVALID_SECTORS returned on error.
84e11361
N
342 */
343 char *c;
344 long long s = strtoll(size, &c, 10);
345 if (s > 0) {
346 switch (*c) {
347 case 'K':
348 c++;
349 default:
350 s *= 2;
351 break;
352 case 'M':
353 c++;
354 s *= 1024 * 2;
355 break;
356 case 'G':
357 c++;
358 s *= 1024 * 1024 * 2;
359 break;
72ca9bcf
N
360 case 's': /* sectors */
361 c++;
362 break;
84e11361 363 }
51d4261c 364 } else
822e393a 365 s = INVALID_SECTORS;
84e11361 366 if (*c)
822e393a 367 s = INVALID_SECTORS;
84e11361
N
368 return s;
369}
370
4a06e2c2
N
371int parse_layout_10(char *layout)
372{
373 int copies, rv;
374 char *cp;
375 /* Parse the layout string for raid10 */
376 /* 'f', 'o' or 'n' followed by a number <= raid_disks */
377 if ((layout[0] != 'n' && layout[0] != 'f' && layout[0] != 'o') ||
378 (copies = strtoul(layout+1, &cp, 10)) < 1 ||
379 copies > 200 ||
380 *cp)
381 return -1;
382 if (layout[0] == 'n')
383 rv = 256 + copies;
384 else if (layout[0] == 'o')
385 rv = 0x10000 + (copies<<8) + 1;
386 else
387 rv = 1 + (copies<<8);
388 return rv;
389}
390
391int parse_layout_faulty(char *layout)
392{
393 /* Parse the layout string for 'faulty' */
394 int ln = strcspn(layout, "0123456789");
503975b9 395 char *m = xstrdup(layout);
4a06e2c2
N
396 int mode;
397 m[ln] = 0;
398 mode = map_name(faultylayout, m);
399 if (mode == UnSet)
400 return -1;
401
402 return mode | (atoi(layout+ln)<< ModeShift);
403}
79868890
N
404
405long parse_num(char *num)
406{
407 /* Either return a valid number, or -1 */
408 char *c;
409 long rv = strtol(num, &c, 10);
410 if (rv < 0 || *c || !num[0])
411 return -1;
412 else
413 return rv;
414}
4a06e2c2 415
4de90913
GJ
416int parse_cluster_confirm_arg(char *input, char **devname, int *slot)
417{
418 char *dev;
419 *slot = strtoul(input, &dev, 10);
420 if (dev == input || dev[0] != ':')
421 return -1;
422 *devname = dev+1;
423 return 0;
424}
425
0430ed48
NB
426void remove_partitions(int fd)
427{
428 /* remove partitions from this block devices.
429 * This is used for components added to an array
430 */
431#ifdef BLKPG_DEL_PARTITION
432 struct blkpg_ioctl_arg a;
433 struct blkpg_partition p;
434
435 a.op = BLKPG_DEL_PARTITION;
436 a.data = (void*)&p;
437 a.datalen = sizeof(p);
438 a.flags = 0;
439 memset(a.data, 0, a.datalen);
ca3b6696 440 for (p.pno = 0; p.pno < 16; p.pno++)
0430ed48
NB
441 ioctl(fd, BLKPG, &a);
442#endif
443}
444
691c6ee1
N
445int test_partition(int fd)
446{
447 /* Check if fd is a whole-disk or a partition.
448 * BLKPG will return EINVAL on a partition, and BLKPG_DEL_PARTITION
449 * will return ENXIO on an invalid partition number.
450 */
451 struct blkpg_ioctl_arg a;
452 struct blkpg_partition p;
453 a.op = BLKPG_DEL_PARTITION;
454 a.data = (void*)&p;
455 a.datalen = sizeof(p);
456 a.flags = 0;
457 memset(a.data, 0, a.datalen);
458 p.pno = 1<<30;
459 if (ioctl(fd, BLKPG, &a) == 0)
460 /* Very unlikely, but not a partition */
461 return 0;
357ac106 462 if (errno == ENXIO || errno == ENOTTY)
691c6ee1
N
463 /* not a partition */
464 return 0;
465
466 return 1;
467}
468
bfd76b93
CA
469int test_partition_from_id(dev_t id)
470{
471 char buf[20];
472 int fd, rv;
473
474 sprintf(buf, "%d:%d", major(id), minor(id));
475 fd = dev_open(buf, O_RDONLY);
476 if (fd < 0)
477 return -1;
478 rv = test_partition(fd);
479 close(fd);
480 return rv;
481}
691c6ee1 482
de5a472e 483int enough(int level, int raid_disks, int layout, int clean, char *avail)
64c4757e 484{
265e0f17 485 int copies, first;
de5a472e
N
486 int i;
487 int avail_disks = 0;
488
489 for (i = 0; i < raid_disks; i++)
490 avail_disks += !!avail[i];
491
64c4757e 492 switch (level) {
265e0f17
NB
493 case 10:
494 /* This is the tricky one - we need to check
495 * which actual disks are present.
496 */
702b557b 497 copies = (layout&255)* ((layout>>8) & 255);
ca3b6696 498 first = 0;
265e0f17
NB
499 do {
500 /* there must be one of the 'copies' form 'first' */
501 int n = copies;
ca3b6696 502 int cnt = 0;
b3ec716d 503 int this = first;
265e0f17 504 while (n--) {
b3ec716d 505 if (avail[this])
265e0f17 506 cnt++;
b3ec716d 507 this = (this+1) % raid_disks;
265e0f17
NB
508 }
509 if (cnt == 0)
510 return 0;
b3ec716d 511 first = (first+(layout&255)) % raid_disks;
265e0f17
NB
512 } while (first != 0);
513 return 1;
e5329c37 514
df0d4ea0 515 case LEVEL_MULTIPATH:
e0d19036 516 return avail_disks>= 1;
df0d4ea0 517 case LEVEL_LINEAR:
64c4757e
NB
518 case 0:
519 return avail_disks == raid_disks;
520 case 1:
521 return avail_disks >= 1;
522 case 4:
330d6900
N
523 if (avail_disks == raid_disks - 1 &&
524 !avail[raid_disks - 1])
525 /* If just the parity device is missing, then we
526 * have enough, even if not clean
527 */
528 return 1;
529 /* FALL THROUGH */
64c4757e 530 case 5:
583315d9
NB
531 if (clean)
532 return avail_disks >= raid_disks-1;
533 else
534 return avail_disks >= raid_disks;
98c6faba 535 case 6:
583315d9
NB
536 if (clean)
537 return avail_disks >= raid_disks-2;
538 else
539 return avail_disks >= raid_disks;
64c4757e
NB
540 default:
541 return 0;
542 }
543}
544
22e263f6
AC
545const int uuid_zero[4] = { 0, 0, 0, 0 };
546
f277ce36 547int same_uuid(int a[4], int b[4], int swapuuid)
64c4757e 548{
f277ce36
NB
549 if (swapuuid) {
550 /* parse uuids are hostendian.
551 * uuid's from some superblocks are big-ending
aba69144 552 * if there is a difference, we need to swap..
f277ce36
NB
553 */
554 unsigned char *ac = (unsigned char *)a;
555 unsigned char *bc = (unsigned char *)b;
556 int i;
ca3b6696 557 for (i = 0; i < 16; i += 4) {
f277ce36
NB
558 if (ac[i+0] != bc[i+3] ||
559 ac[i+1] != bc[i+2] ||
560 ac[i+2] != bc[i+1] ||
561 ac[i+3] != bc[i+0])
562 return 0;
563 }
564 return 1;
565 } else {
566 if (a[0]==b[0] &&
567 a[1]==b[1] &&
568 a[2]==b[2] &&
569 a[3]==b[3])
570 return 1;
571 return 0;
572 }
64c4757e 573}
ca3b6696 574
350f29f9
NB
575void copy_uuid(void *a, int b[4], int swapuuid)
576{
577 if (swapuuid) {
578 /* parse uuids are hostendian.
579 * uuid's from some superblocks are big-ending
580 * if there is a difference, we need to swap..
581 */
582 unsigned char *ac = (unsigned char *)a;
583 unsigned char *bc = (unsigned char *)b;
584 int i;
ca3b6696 585 for (i = 0; i < 16; i += 4) {
350f29f9
NB
586 ac[i+0] = bc[i+3];
587 ac[i+1] = bc[i+2];
588 ac[i+2] = bc[i+1];
589 ac[i+3] = bc[i+0];
590 }
591 } else
592 memcpy(a, b, 16);
593}
64c4757e 594
aae5a112 595char *__fname_from_uuid(int id[4], int swap, char *buf, char sep)
d7288ddc 596{
9968e376 597 int i, j;
d7288ddc
N
598 char uuid[16];
599 char *c = buf;
600 strcpy(c, "UUID-");
601 c += strlen(c);
aae5a112 602 copy_uuid(uuid, id, swap);
9968e376 603 for (i = 0; i < 4; i++) {
9968e376 604 if (i)
ff54de6e 605 *c++ = sep;
9968e376
DW
606 for (j = 3; j >= 0; j--) {
607 sprintf(c,"%02x", (unsigned char) uuid[j+4*i]);
608 c+= 2;
609 }
d7288ddc
N
610 }
611 return buf;
aae5a112
DW
612
613}
614
615char *fname_from_uuid(struct supertype *st, struct mdinfo *info, char *buf, char sep)
616{
86983cce
N
617 // dirty hack to work around an issue with super1 superblocks...
618 // super1 superblocks need swapuuid set in order for assembly to
619 // work, but can't have it set if we want this printout to match
620 // all the other uuid printouts in super1.c, so we force swapuuid
621 // to 1 to make our printout match the rest of super1
622 return __fname_from_uuid(info->uuid, (st->ss == &super1) ? 1 : st->ss->swapuuid, buf, sep);
d7288ddc
N
623}
624
682c7051
NB
625int check_ext2(int fd, char *name)
626{
627 /*
628 * Check for an ext2fs file system.
629 * Superblock is always 1K at 1K offset
630 *
631 * s_magic is le16 at 56 == 0xEF53
632 * report mtime - le32 at 44
633 * blocks - le32 at 4
634 * logblksize - le32 at 24
635 */
636 unsigned char sb[1024];
637 time_t mtime;
2538ba2a
N
638 unsigned long long size;
639 int bsize;
682c7051
NB
640 if (lseek(fd, 1024,0)!= 1024)
641 return 0;
642 if (read(fd, sb, 1024)!= 1024)
643 return 0;
644 if (sb[56] != 0x53 || sb[57] != 0xef)
645 return 0;
646
647 mtime = sb[44]|(sb[45]|(sb[46]|sb[47]<<8)<<8)<<8;
648 bsize = sb[24]|(sb[25]|(sb[26]|sb[27]<<8)<<8)<<8;
649 size = sb[4]|(sb[5]|(sb[6]|sb[7]<<8)<<8)<<8;
2538ba2a 650 size <<= bsize;
e7b84f9d 651 pr_err("%s appears to contain an ext2fs file system\n",
682c7051 652 name);
2538ba2a 653 cont_err("size=%lluK mtime=%s", size, ctime(&mtime));
682c7051
NB
654 return 1;
655}
656
657int check_reiser(int fd, char *name)
658{
659 /*
660 * superblock is at 64K
661 * size is 1024;
662 * Magic string "ReIsErFs" or "ReIsEr2Fs" at 52
663 *
664 */
665 unsigned char sb[1024];
2538ba2a 666 unsigned long long size;
682c7051
NB
667 if (lseek(fd, 64*1024, 0) != 64*1024)
668 return 0;
669 if (read(fd, sb, 1024) != 1024)
670 return 0;
ca3b6696
N
671 if (strncmp((char*)sb+52, "ReIsErFs",8) != 0 &&
672 strncmp((char*)sb+52, "ReIsEr2Fs",9) != 0)
682c7051 673 return 0;
e7b84f9d 674 pr_err("%s appears to contain a reiserfs file system\n",name);
682c7051 675 size = sb[0]|(sb[1]|(sb[2]|sb[3]<<8)<<8)<<8;
2538ba2a 676 cont_err("size = %lluK\n", size*4);
aba69144 677
682c7051
NB
678 return 1;
679}
680
681int check_raid(int fd, char *name)
682{
4b1ac34b 683 struct mdinfo info;
682c7051 684 time_t crtime;
d078d77c 685 char *level;
82d9eba6 686 struct supertype *st = guess_super(fd);
f9ce90ba 687
cb8f6859
N
688 if (!st)
689 return 0;
fa219dd2
MD
690 if (st->ss->add_to_super != NULL) {
691 st->ss->load_super(st, fd, name);
692 /* Looks like a raid array .. */
693 pr_err("%s appears to be part of a raid array:\n", name);
694 st->ss->getinfo_super(st, &info, NULL);
695 st->ss->free_super(st);
696 crtime = info.array.ctime;
697 level = map_num(pers, info.array.level);
698 if (!level)
699 level = "-unknown-";
700 cont_err("level=%s devices=%d ctime=%s",
701 level, info.array.raid_disks, ctime(&crtime));
702 } else {
703 /* Looks like GPT or MBR */
704 pr_err("partition table exists on %s\n", name);
705 }
82d9eba6 706 return 1;
682c7051
NB
707}
708
682c7051
NB
709int ask(char *mesg)
710{
711 char *add = "";
712 int i;
ca3b6696 713 for (i = 0; i < 5; i++) {
682c7051
NB
714 char buf[100];
715 fprintf(stderr, "%s%s", mesg, add);
716 fflush(stderr);
717 if (fgets(buf, 100, stdin)==NULL)
718 return 0;
719 if (buf[0]=='y' || buf[0]=='Y')
720 return 1;
721 if (buf[0]=='n' || buf[0]=='N')
722 return 0;
723 add = "(y/n) ";
724 }
e7b84f9d 725 pr_err("assuming 'no'\n");
682c7051
NB
726 return 0;
727}
728
8d80900b 729int is_standard(char *dev, int *nump)
e5329c37
NB
730{
731 /* tests if dev is a "standard" md dev name.
732 * i.e if the last component is "/dNN" or "/mdNN",
aba69144 733 * where NN is a string of digits
598f0d58
NB
734 * Returns 1 if a partitionable standard,
735 * -1 if non-partitonable,
736 * 0 if not a standard name.
e5329c37 737 */
8d80900b 738 char *d = strrchr(dev, '/');
ca3b6696 739 int type = 0;
8d80900b
NB
740 int num;
741 if (!d)
e5329c37 742 return 0;
ca3b6696
N
743 if (strncmp(d, "/d",2) == 0)
744 d += 2, type = 1; /* /dev/md/dN{pM} */
745 else if (strncmp(d, "/md_d", 5) == 0)
746 d += 5, type = 1; /* /dev/md_dN{pM} */
747 else if (strncmp(d, "/md", 3) == 0)
748 d += 3, type = -1; /* /dev/mdN */
749 else if (d-dev > 3 && strncmp(d-2, "md/", 3) == 0)
750 d += 1, type = -1; /* /dev/md/N */
e5329c37
NB
751 else
752 return 0;
8d80900b 753 if (!*d)
e5329c37 754 return 0;
8d80900b
NB
755 num = atoi(d);
756 while (isdigit(*d))
757 d++;
758 if (*d)
e5329c37 759 return 0;
8d80900b
NB
760 if (nump) *nump = num;
761
762 return type;
e5329c37
NB
763}
764
4b1ac34b 765unsigned long calc_csum(void *super, int bytes)
82b27616 766{
56eb10c0 767 unsigned long long newcsum = 0;
82b27616 768 int i;
4b1ac34b
NB
769 unsigned int csum;
770 unsigned int *superc = (unsigned int*) super;
82b27616 771
ca3b6696
N
772 for(i = 0; i < bytes/4; i++)
773 newcsum += superc[i];
82b27616 774 csum = (newcsum& 0xffffffff) + (newcsum>>32);
570c0542 775#ifdef __alpha__
aba69144 776/* The in-kernel checksum calculation is always 16bit on
570c0542 777 * the alpha, though it is 32 bit on i386...
ca3b6696 778 * I wonder what it is elsewhere... (it uses an API in
570c0542
NB
779 * a way that it shouldn't).
780 */
781 csum = (csum & 0xffff) + (csum >> 16);
782 csum = (csum & 0xffff) + (csum >> 16);
783#endif
82b27616
NB
784 return csum;
785}
cd29a5c8 786
56eb10c0 787char *human_size(long long bytes)
cd29a5c8 788{
ff9239ee 789 static char buf[47];
d5d3721e
NB
790
791 /* We convert bytes to either centi-M{ega,ibi}bytes or
792 * centi-G{igi,ibi}bytes, with appropriate rounding,
793 * and then print 1/100th of those as a decimal.
794 * We allow upto 2048Megabytes before converting to
795 * gigabytes, as that shows more precision and isn't
796 * too large a number.
01619b48 797 * Terabytes are not yet handled.
d5d3721e 798 */
cd29a5c8 799
56eb10c0 800 if (bytes < 5000*1024)
ca3b6696 801 buf[0] = 0;
d5d3721e 802 else if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 803 long cMiB = (bytes * 200LL / (1LL<<20) + 1) / 2;
d5d3721e 804 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
8f23b0b3 805 snprintf(buf, sizeof(buf), " (%ld.%02ld MiB %ld.%02ld MB)",
efa29530 806 cMiB/100, cMiB % 100, cMB/100, cMB % 100);
d5d3721e 807 } else {
93d3bd3b 808 long cGiB = (bytes * 200LL / (1LL<<30) +1) / 2;
d5d3721e 809 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
8f23b0b3 810 snprintf(buf, sizeof(buf), " (%ld.%02ld GiB %ld.%02ld GB)",
efa29530 811 cGiB/100, cGiB % 100, cGB/100, cGB % 100);
d5d3721e 812 }
cd29a5c8
NB
813 return buf;
814}
e0d19036 815
f0ec6710 816char *human_size_brief(long long bytes, int prefix)
e0d19036
NB
817{
818 static char buf[30];
e0d19036 819
570abc6f
MN
820 /* We convert bytes to either centi-M{ega,ibi}bytes or
821 * centi-G{igi,ibi}bytes, with appropriate rounding,
822 * and then print 1/100th of those as a decimal.
823 * We allow upto 2048Megabytes before converting to
824 * gigabytes, as that shows more precision and isn't
825 * too large a number.
826 * Terabytes are not yet handled.
f0ec6710
MN
827 *
828 * If prefix == IEC, we mean prefixes like kibi,mebi,gibi etc.
829 * If prefix == JEDEC, we mean prefixes like kilo,mega,giga etc.
570abc6f
MN
830 */
831
e0d19036 832 if (bytes < 5000*1024)
570abc6f 833 buf[0] = 0;
f0ec6710
MN
834 else if (prefix == IEC) {
835 if (bytes < 2*1024LL*1024LL*1024LL) {
93d3bd3b 836 long cMiB = (bytes * 200LL / (1LL<<20) +1) /2;
f0ec6710 837 snprintf(buf, sizeof(buf), "%ld.%02ldMiB",
efa29530 838 cMiB/100, cMiB % 100);
f0ec6710 839 } else {
93d3bd3b 840 long cGiB = (bytes * 200LL / (1LL<<30) +1) /2;
f0ec6710 841 snprintf(buf, sizeof(buf), "%ld.%02ldGiB",
efa29530 842 cGiB/100, cGiB % 100);
f0ec6710
MN
843 }
844 }
845 else if (prefix == JEDEC) {
846 if (bytes < 2*1024LL*1024LL*1024LL) {
847 long cMB = (bytes / ( 1000000LL / 200LL ) +1) /2;
848 snprintf(buf, sizeof(buf), "%ld.%02ldMB",
efa29530 849 cMB/100, cMB % 100);
f0ec6710
MN
850 } else {
851 long cGB = (bytes / (1000000000LL/200LL ) +1) /2;
852 snprintf(buf, sizeof(buf), "%ld.%02ldGB",
efa29530 853 cGB/100, cGB % 100);
f0ec6710 854 }
570abc6f 855 }
f0ec6710
MN
856 else
857 buf[0] = 0;
570abc6f 858
e0d19036
NB
859 return buf;
860}
e4965ef8
N
861
862void print_r10_layout(int layout)
863{
864 int near = layout & 255;
865 int far = (layout >> 8) & 255;
866 int offset = (layout&0x10000);
867 char *sep = "";
868
869 if (near != 1) {
870 printf("%s near=%d", sep, near);
871 sep = ",";
872 }
873 if (far != 1)
874 printf("%s %s=%d", sep, offset?"offset":"far", far);
875 if (near*far == 1)
876 printf("NO REDUNDANCY");
877}
e0d19036 878
5f8097be
NB
879unsigned long long calc_array_size(int level, int raid_disks, int layout,
880 int chunksize, unsigned long long devsize)
577e8448 881{
e03640bd
CA
882 if (level == 1)
883 return devsize;
577e8448
AK
884 devsize &= ~(unsigned long long)((chunksize>>9)-1);
885 return get_data_disks(level, layout, raid_disks) * devsize;
886}
887
888int get_data_disks(int level, int layout, int raid_disks)
5f8097be
NB
889{
890 int data_disks = 0;
891 switch (level) {
ca3b6696
N
892 case 0: data_disks = raid_disks;
893 break;
894 case 1: data_disks = 1;
895 break;
5f8097be 896 case 4:
ca3b6696
N
897 case 5: data_disks = raid_disks - 1;
898 break;
899 case 6: data_disks = raid_disks - 2;
900 break;
5f8097be
NB
901 case 10: data_disks = raid_disks / (layout & 255) / ((layout>>8)&255);
902 break;
903 }
577e8448
AK
904
905 return data_disks;
5f8097be
NB
906}
907
13db17bd 908dev_t devnm2devid(char *devnm)
4dd2df09
N
909{
910 /* First look in /sys/block/$DEVNM/dev for %d:%d
911 * If that fails, try parsing out a number
912 */
913 char path[100];
914 char *ep;
915 int fd;
916 int mjr,mnr;
917
918 sprintf(path, "/sys/block/%s/dev", devnm);
919 fd = open(path, O_RDONLY);
920 if (fd >= 0) {
921 char buf[20];
922 int n = read(fd, buf, sizeof(buf));
923 close(fd);
924 if (n > 0)
925 buf[n] = 0;
926 if (n > 0 && sscanf(buf, "%d:%d\n", &mjr, &mnr) == 2)
927 return makedev(mjr, mnr);
928 }
929 if (strncmp(devnm, "md_d", 4) == 0 &&
930 isdigit(devnm[4]) &&
931 (mnr = strtoul(devnm+4, &ep, 10)) >= 0 &&
932 ep > devnm && *ep == 0)
933 return makedev(get_mdp_major(), mnr << MdpMinorShift);
934
935 if (strncmp(devnm, "md", 2) == 0 &&
936 isdigit(devnm[2]) &&
937 (mnr = strtoul(devnm+2, &ep, 10)) >= 0 &&
938 ep > devnm && *ep == 0)
939 return makedev(MD_MAJOR, mnr);
940
941 return 0;
942}
943
944char *get_md_name(char *devnm)
e0d19036
NB
945{
946 /* find /dev/md%d or /dev/md/%d or make a device /dev/.tmp.md%d */
98c6faba 947 /* if dev < 0, want /dev/md/d%d or find mdp in /proc/devices ... */
4dd2df09 948
e0d19036
NB
949 static char devname[50];
950 struct stat stb;
4dd2df09 951 dev_t rdev = devnm2devid(devnm);
dd0781e5 952 char *dn;
98c6faba 953
4dd2df09
N
954 if (rdev == 0)
955 return 0;
956 if (strncmp(devnm, "md_", 3) == 0) {
957 snprintf(devname, sizeof(devname), "/dev/md/%s",
958 devnm + 3);
98c6faba
NB
959 if (stat(devname, &stb) == 0
960 && (S_IFMT&stb.st_mode) == S_IFBLK
961 && (stb.st_rdev == rdev))
962 return devname;
963 }
4dd2df09
N
964 snprintf(devname, sizeof(devname), "/dev/%s", devnm);
965 if (stat(devname, &stb) == 0
966 && (S_IFMT&stb.st_mode) == S_IFBLK
967 && (stb.st_rdev == rdev))
968 return devname;
969
970 snprintf(devname, sizeof(devname), "/dev/md/%s", devnm+2);
971 if (stat(devname, &stb) == 0
972 && (S_IFMT&stb.st_mode) == S_IFBLK
973 && (stb.st_rdev == rdev))
974 return devname;
975
16c6fa80 976 dn = map_dev(major(rdev), minor(rdev), 0);
dd0781e5
NB
977 if (dn)
978 return dn;
4dd2df09 979 snprintf(devname, sizeof(devname), "/dev/.tmp.%s", devnm);
e0d19036 980 if (mknod(devname, S_IFBLK | 0600, rdev) == -1)
dd0781e5
NB
981 if (errno != EEXIST)
982 return NULL;
e0d19036
NB
983
984 if (stat(devname, &stb) == 0
985 && (S_IFMT&stb.st_mode) == S_IFBLK
986 && (stb.st_rdev == rdev))
987 return devname;
988 unlink(devname);
989 return NULL;
990}
991
992void put_md_name(char *name)
993{
ca3b6696 994 if (strncmp(name, "/dev/.tmp.md", 12) == 0)
e0d19036
NB
995 unlink(name);
996}
f9ce90ba 997
cc742d38
N
998int get_maj_min(char *dev, int *major, int *minor)
999{
1000 char *e;
1001 *major = strtoul(dev, &e, 0);
1002 return (e > dev && *e == ':' && e[1] &&
1003 (*minor = strtoul(e+1, &e, 0)) >= 0 &&
1004 *e == 0);
1005}
1006
8b0dabea
NB
1007int dev_open(char *dev, int flags)
1008{
1009 /* like 'open', but if 'dev' matches %d:%d, create a temp
1010 * block device and open that
1011 */
8b0dabea
NB
1012 int fd = -1;
1013 char devname[32];
e81cdd9f 1014 int major;
8b0dabea 1015 int minor;
e81cdd9f 1016
9f0ad56b
JS
1017 if (!dev)
1018 return -1;
6df6a774 1019 flags |= O_DIRECT;
e81cdd9f 1020
cc742d38 1021 if (get_maj_min(dev, &major, &minor)) {
9dc7d357
N
1022 snprintf(devname, sizeof(devname), "/dev/.tmp.md.%d:%d:%d",
1023 (int)getpid(), major, minor);
1024 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
1025 fd = open(devname, flags);
1026 unlink(devname);
6df6a774
N
1027 }
1028 if (fd < 0) {
9dc7d357 1029 /* Try /tmp as /dev appear to be read-only */
6df6a774
N
1030 snprintf(devname, sizeof(devname), "/tmp/.tmp.md.%d:%d:%d",
1031 (int)getpid(), major, minor);
ca3b6696 1032 if (mknod(devname, S_IFBLK|0600, makedev(major, minor)) == 0) {
6df6a774
N
1033 fd = open(devname, flags);
1034 unlink(devname);
1035 }
8b0dabea
NB
1036 }
1037 } else
6df6a774 1038 fd = open(dev, flags);
8b0dabea
NB
1039 return fd;
1040}
f9ce90ba 1041
4dd2df09 1042int open_dev_flags(char *devnm, int flags)
e8a70c89 1043{
13db17bd 1044 dev_t devid;
e8a70c89
N
1045 char buf[20];
1046
4dd2df09
N
1047 devid = devnm2devid(devnm);
1048 sprintf(buf, "%d:%d", major(devid), minor(devid));
d998b738
N
1049 return dev_open(buf, flags);
1050}
1051
4dd2df09 1052int open_dev(char *devnm)
d998b738 1053{
4dd2df09 1054 return open_dev_flags(devnm, O_RDONLY);
e8a70c89
N
1055}
1056
4dd2df09 1057int open_dev_excl(char *devnm)
a931db9e
NB
1058{
1059 char buf[20];
1060 int i;
7187750e 1061 int flags = O_RDWR;
13db17bd 1062 dev_t devid = devnm2devid(devnm);
399e0b97 1063 long delay = 1000;
a931db9e 1064
4dd2df09 1065 sprintf(buf, "%d:%d", major(devid), minor(devid));
efa29530 1066 for (i = 0; i < 25; i++) {
7187750e 1067 int fd = dev_open(buf, flags|O_EXCL);
a931db9e
NB
1068 if (fd >= 0)
1069 return fd;
7187750e
N
1070 if (errno == EACCES && flags == O_RDWR) {
1071 flags = O_RDONLY;
1072 continue;
1073 }
a931db9e
NB
1074 if (errno != EBUSY)
1075 return fd;
399e0b97
N
1076 usleep(delay);
1077 if (delay < 200000)
1078 delay *= 2;
a931db9e
NB
1079 }
1080 return -1;
1081}
1082
9008ed1c
N
1083int same_dev(char *one, char *two)
1084{
1085 struct stat st1, st2;
1086 if (stat(one, &st1) != 0)
1087 return 0;
1088 if (stat(two, &st2) != 0)
1089 return 0;
1090 if ((st1.st_mode & S_IFMT) != S_IFBLK)
1091 return 0;
1092 if ((st2.st_mode & S_IFMT) != S_IFBLK)
1093 return 0;
1094 return st1.st_rdev == st2.st_rdev;
1095}
1096
a7c6e3fb 1097void wait_for(char *dev, int fd)
a714580e
N
1098{
1099 int i;
a7c6e3fb 1100 struct stat stb_want;
399e0b97 1101 long delay = 1000;
a7c6e3fb
N
1102
1103 if (fstat(fd, &stb_want) != 0 ||
1104 (stb_want.st_mode & S_IFMT) != S_IFBLK)
1105 return;
a714580e 1106
efa29530 1107 for (i = 0; i < 25; i++) {
a714580e 1108 struct stat stb;
a7c6e3fb
N
1109 if (stat(dev, &stb) == 0 &&
1110 (stb.st_mode & S_IFMT) == S_IFBLK &&
1111 (stb.st_rdev == stb_want.st_rdev))
a714580e 1112 return;
399e0b97
N
1113 usleep(delay);
1114 if (delay < 200000)
1115 delay *= 2;
a714580e 1116 }
436305c6 1117 if (i == 25)
1ade5cc1 1118 dprintf("timeout waiting for %s\n", dev);
a714580e
N
1119}
1120
0f22b998
N
1121struct superswitch *superlist[] =
1122{
1123 &super0, &super1,
1124 &super_ddf, &super_imsm,
0592faeb 1125 &mbr, &gpt,
32141c17
JS
1126 NULL
1127};
f7dd881f 1128
4725bc31 1129struct supertype *super_by_fd(int fd, char **subarrayp)
f9ce90ba 1130{
1686dc25
NB
1131 mdu_array_info_t array;
1132 int vers;
1133 int minor;
1134 struct supertype *st = NULL;
7e0f6979 1135 struct mdinfo *sra;
142cb9e1 1136 char *verstr;
1686dc25
NB
1137 char version[20];
1138 int i;
f7e7067b 1139 char *subarray = NULL;
4dd2df09 1140 char container[32] = "";
1686dc25 1141
4dd2df09 1142 sra = sysfs_read(fd, NULL, GET_VERSION);
1686dc25
NB
1143
1144 if (sra) {
7e0f6979
NB
1145 vers = sra->array.major_version;
1146 minor = sra->array.minor_version;
142cb9e1 1147 verstr = sra->text_version;
1686dc25 1148 } else {
9cd39f01 1149 if (md_get_array_info(fd, &array))
1686dc25
NB
1150 array.major_version = array.minor_version = 0;
1151 vers = array.major_version;
1152 minor = array.minor_version;
142cb9e1 1153 verstr = "";
6fbba4c9 1154 }
82d9eba6 1155
1686dc25
NB
1156 if (vers != -1) {
1157 sprintf(version, "%d.%d", vers, minor);
1158 verstr = version;
6fbba4c9 1159 }
3c558363 1160 if (minor == -2 && is_subarray(verstr)) {
f7e7067b 1161 char *dev = verstr+1;
5f7e44b2 1162
f7e7067b 1163 subarray = strchr(dev, '/');
92d49ecf 1164 if (subarray) {
f7e7067b 1165 *subarray++ = '\0';
503975b9 1166 subarray = xstrdup(subarray);
92d49ecf 1167 }
4dd2df09 1168 strcpy(container, dev);
15d230f7 1169 sysfs_free(sra);
5f7e44b2 1170 sra = sysfs_read(-1, container, GET_VERSION);
603f24a0
N
1171 if (sra && sra->text_version[0])
1172 verstr = sra->text_version;
1173 else
1174 verstr = "-no-metadata-";
f7e7067b
NB
1175 }
1176
efa29530 1177 for (i = 0; st == NULL && superlist[i]; i++)
f7e7067b 1178 st = superlist[i]->match_metadata_desc(verstr);
1686dc25 1179
15d230f7 1180 sysfs_free(sra);
f7e7067b 1181 if (st) {
3b0896f8 1182 st->sb = NULL;
1f49fb3a
N
1183 if (subarrayp)
1184 *subarrayp = subarray;
4dd2df09
N
1185 strcpy(st->container_devnm, container);
1186 strcpy(st->devnm, fd2devnm(fd));
4725bc31
N
1187 } else
1188 free(subarray);
5f7e44b2 1189
82d9eba6 1190 return st;
f9ce90ba 1191}
ea24acd0 1192
0f0749ad 1193int dev_size_from_id(dev_t id, unsigned long long *size)
52d5d101
AC
1194{
1195 char buf[20];
1196 int fd;
1197
1198 sprintf(buf, "%d:%d", major(id), minor(id));
1199 fd = dev_open(buf, O_RDONLY);
1200 if (fd < 0)
1201 return 0;
1202 if (get_dev_size(fd, NULL, size)) {
1203 close(fd);
1204 return 1;
1205 }
1206 close(fd);
1207 return 0;
1208}
f9ce90ba 1209
159c3a1a 1210struct supertype *dup_super(struct supertype *orig)
3da92f27 1211{
159c3a1a 1212 struct supertype *st;
1686dc25 1213
d2ca6449
NB
1214 if (!orig)
1215 return orig;
503975b9 1216 st = xcalloc(1, sizeof(*st));
159c3a1a
NB
1217 st->ss = orig->ss;
1218 st->max_devs = orig->max_devs;
1219 st->minor_version = orig->minor_version;
cb8f6859 1220 st->ignore_hw_compat = orig->ignore_hw_compat;
23bf42cc 1221 st->data_offset = orig->data_offset;
159c3a1a
NB
1222 st->sb = NULL;
1223 st->info = NULL;
1224 return st;
3da92f27
NB
1225}
1226
54887ad8 1227struct supertype *guess_super_type(int fd, enum guess_types guess_type)
f9ce90ba
NB
1228{
1229 /* try each load_super to find the best match,
1230 * and return the best superswitch
1231 */
82d9eba6
NB
1232 struct superswitch *ss;
1233 struct supertype *st;
1d13b599 1234 unsigned int besttime = 0;
570c0542 1235 int bestsuper = -1;
f9ce90ba
NB
1236 int i;
1237
503975b9 1238 st = xcalloc(1, sizeof(*st));
4dd2df09 1239 st->container_devnm[0] = 0;
d1d599ea 1240
efa29530 1241 for (i = 0; superlist[i]; i++) {
f9ce90ba
NB
1242 int rv;
1243 ss = superlist[i];
54887ad8
N
1244 if (guess_type == guess_array && ss->add_to_super == NULL)
1245 continue;
1246 if (guess_type == guess_partitions && ss->add_to_super != NULL)
1247 continue;
ef609477 1248 memset(st, 0, sizeof(*st));
df3346e6 1249 st->ignore_hw_compat = 1;
3da92f27 1250 rv = ss->load_super(st, fd, NULL);
570c0542
NB
1251 if (rv == 0) {
1252 struct mdinfo info;
a5d85af7 1253 st->ss->getinfo_super(st, &info, NULL);
570c0542
NB
1254 if (bestsuper == -1 ||
1255 besttime < info.array.ctime) {
1256 bestsuper = i;
1257 besttime = info.array.ctime;
570c0542 1258 }
3da92f27 1259 ss->free_super(st);
570c0542
NB
1260 }
1261 }
1262 if (bestsuper != -1) {
1263 int rv;
ef609477 1264 memset(st, 0, sizeof(*st));
df3346e6 1265 st->ignore_hw_compat = 1;
3da92f27 1266 rv = superlist[bestsuper]->load_super(st, fd, NULL);
f9ce90ba 1267 if (rv == 0) {
5e747af2 1268 superlist[bestsuper]->free_super(st);
82d9eba6 1269 return st;
f9ce90ba
NB
1270 }
1271 }
570c0542 1272 free(st);
f9ce90ba
NB
1273 return NULL;
1274}
fe6729fa 1275
beae1dfe
NB
1276/* Return size of device in bytes */
1277int get_dev_size(int fd, char *dname, unsigned long long *sizep)
1278{
1279 unsigned long long ldsize;
c2c9bb6f
NB
1280 struct stat st;
1281
1282 if (fstat(fd, &st) != -1 && S_ISREG(st.st_mode))
1283 ldsize = (unsigned long long)st.st_size;
1284 else
beae1dfe
NB
1285#ifdef BLKGETSIZE64
1286 if (ioctl(fd, BLKGETSIZE64, &ldsize) != 0)
1287#endif
1288 {
1289 unsigned long dsize;
1290 if (ioctl(fd, BLKGETSIZE, &dsize) == 0) {
1291 ldsize = dsize;
1292 ldsize <<= 9;
1293 } else {
1294 if (dname)
32971509 1295 pr_err("Cannot get size of %s: %s\n",
beae1dfe
NB
1296 dname, strerror(errno));
1297 return 0;
1298 }
1299 }
1300 *sizep = ldsize;
1301 return 1;
1302}
8fac0577 1303
32971509
PB
1304/* Return sector size of device in bytes */
1305int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep)
1306{
1307 unsigned int sectsize;
1308
1309 if (ioctl(fd, BLKSSZGET, &sectsize) != 0) {
1310 if (dname)
1311 pr_err("Cannot get sector size of %s: %s\n",
1312 dname, strerror(errno));
1313 return 0;
1314 }
1315
1316 *sectsizep = sectsize;
1317 return 1;
1318}
1319
3a371610
N
1320/* Return true if this can only be a container, not a member device.
1321 * i.e. is and md device and size is zero
1322 */
1323int must_be_container(int fd)
1324{
f5c924f4 1325 struct mdinfo *mdi;
3a371610 1326 unsigned long long size;
f5c924f4
JS
1327
1328 mdi = sysfs_read(fd, NULL, GET_VERSION);
1329 if (!mdi)
3a371610 1330 return 0;
f5c924f4
JS
1331 sysfs_free(mdi);
1332
3a371610
N
1333 if (get_dev_size(fd, NULL, &size) == 0)
1334 return 1;
1335 if (size == 0)
1336 return 1;
1337 return 0;
1338}
034b203a
TM
1339
1340/* Sets endofpart parameter to the last block used by the last GPT partition on the device.
1341 * Returns: 1 if successful
1342 * -1 for unknown partition type
1343 * 0 for other errors
1344 */
1345static int get_gpt_last_partition_end(int fd, unsigned long long *endofpart)
1346{
056b331e 1347 struct GPT gpt;
034b203a 1348 unsigned char empty_gpt_entry[16]= {0};
db7fdfe4
JS
1349 struct GPT_part_entry *part;
1350 char buf[512];
034b203a
TM
1351 unsigned long long curr_part_end;
1352 unsigned all_partitions, entry_size;
f21e18ca 1353 unsigned part_nr;
41b06495 1354 unsigned int sector_size = 0;
034b203a
TM
1355
1356 *endofpart = 0;
1357
056b331e 1358 BUILD_BUG_ON(sizeof(gpt) != 512);
73e658d8 1359 /* skip protective MBR */
41b06495
MD
1360 if (!get_dev_sector_size(fd, NULL, &sector_size))
1361 return 0;
1362 lseek(fd, sector_size, SEEK_SET);
73e658d8 1363 /* read GPT header */
056b331e 1364 if (read(fd, &gpt, 512) != 512)
034b203a
TM
1365 return 0;
1366
1367 /* get the number of partition entries and the entry size */
056b331e
N
1368 all_partitions = __le32_to_cpu(gpt.part_cnt);
1369 entry_size = __le32_to_cpu(gpt.part_size);
034b203a
TM
1370
1371 /* Check GPT signature*/
056b331e 1372 if (gpt.magic != GPT_SIGNATURE_MAGIC)
034b203a
TM
1373 return -1;
1374
1375 /* sanity checks */
1376 if (all_partitions > 1024 ||
db7fdfe4 1377 entry_size > sizeof(buf))
034b203a
TM
1378 return -1;
1379
db7fdfe4
JS
1380 part = (struct GPT_part_entry *)buf;
1381
41b06495
MD
1382 /* set offset to third block (GPT entries) */
1383 lseek(fd, sector_size*2, SEEK_SET);
ca3b6696 1384 for (part_nr = 0; part_nr < all_partitions; part_nr++) {
73e658d8 1385 /* read partition entry */
db7fdfe4 1386 if (read(fd, buf, entry_size) != (ssize_t)entry_size)
73e658d8
LB
1387 return 0;
1388
034b203a 1389 /* is this valid partition? */
db7fdfe4 1390 if (memcmp(part->type_guid, empty_gpt_entry, 16) != 0) {
034b203a 1391 /* check the last lba for the current partition */
db7fdfe4 1392 curr_part_end = __le64_to_cpu(part->ending_lba);
034b203a
TM
1393 if (curr_part_end > *endofpart)
1394 *endofpart = curr_part_end;
1395 }
1396
034b203a
TM
1397 }
1398 return 1;
1399}
1400
1401/* Sets endofpart parameter to the last block used by the last partition on the device.
1402 * Returns: 1 if successful
1403 * -1 for unknown partition type
1404 * 0 for other errors
1405 */
1406static int get_last_partition_end(int fd, unsigned long long *endofpart)
1407{
056b331e 1408 struct MBR boot_sect;
034b203a 1409 unsigned long long curr_part_end;
f21e18ca 1410 unsigned part_nr;
31208db9 1411 unsigned int sector_size;
034b203a
TM
1412 int retval = 0;
1413
1414 *endofpart = 0;
1415
056b331e 1416 BUILD_BUG_ON(sizeof(boot_sect) != 512);
034b203a
TM
1417 /* read MBR */
1418 lseek(fd, 0, 0);
056b331e 1419 if (read(fd, &boot_sect, 512) != 512)
034b203a
TM
1420 goto abort;
1421
1422 /* check MBP signature */
056b331e 1423 if (boot_sect.magic == MBR_SIGNATURE_MAGIC) {
034b203a
TM
1424 retval = 1;
1425 /* found the correct signature */
034b203a 1426
ca3b6696 1427 for (part_nr = 0; part_nr < MBR_PARTITIONS; part_nr++) {
8e2bca51
JC
1428 /*
1429 * Have to make every access through boot_sect rather
1430 * than using a pointer to the partition table (or an
1431 * entry), since the entries are not properly aligned.
1432 */
1433
034b203a 1434 /* check for GPT type */
8e2bca51
JC
1435 if (boot_sect.parts[part_nr].part_type ==
1436 MBR_GPT_PARTITION_TYPE) {
034b203a
TM
1437 retval = get_gpt_last_partition_end(fd, endofpart);
1438 break;
1439 }
1440 /* check the last used lba for the current partition */
8e2bca51
JC
1441 curr_part_end =
1442 __le32_to_cpu(boot_sect.parts[part_nr].first_sect_lba) +
1443 __le32_to_cpu(boot_sect.parts[part_nr].blocks_num);
034b203a
TM
1444 if (curr_part_end > *endofpart)
1445 *endofpart = curr_part_end;
034b203a
TM
1446 }
1447 } else {
1448 /* Unknown partition table */
1449 retval = -1;
1450 }
31208db9
MD
1451 /* calculate number of 512-byte blocks */
1452 if (get_dev_sector_size(fd, NULL, &sector_size))
1453 *endofpart *= (sector_size / 512);
034b203a
TM
1454 abort:
1455 return retval;
1456}
1457
53ed6ac3
KW
1458int check_partitions(int fd, char *dname, unsigned long long freesize,
1459 unsigned long long size)
034b203a
TM
1460{
1461 /*
1462 * Check where the last partition ends
1463 */
1464 unsigned long long endofpart;
034b203a 1465
31208db9 1466 if (get_last_partition_end(fd, &endofpart) > 0) {
034b203a
TM
1467 /* There appears to be a partition table here */
1468 if (freesize == 0) {
1469 /* partitions will not be visible in new device */
e7b84f9d
N
1470 pr_err("partition table exists on %s but will be lost or\n"
1471 " meaningless after creating array\n",
1472 dname);
034b203a
TM
1473 return 1;
1474 } else if (endofpart > freesize) {
1475 /* last partition overlaps metadata */
e7b84f9d
N
1476 pr_err("metadata will over-write last partition on %s.\n",
1477 dname);
034b203a 1478 return 1;
53ed6ac3
KW
1479 } else if (size && endofpart > size) {
1480 /* partitions will be truncated in new device */
e7b84f9d
N
1481 pr_err("array size is too small to cover all partitions on %s.\n",
1482 dname);
53ed6ac3 1483 return 1;
034b203a
TM
1484 }
1485 }
1486 return 0;
1487}
1488
a322f70c
DW
1489int open_container(int fd)
1490{
1491 /* 'fd' is a block device. Find out if it is in use
1492 * by a container, and return an open fd on that container.
1493 */
1494 char path[256];
1495 char *e;
1496 DIR *dir;
1497 struct dirent *de;
1498 int dfd, n;
1499 char buf[200];
1500 int major, minor;
1501 struct stat st;
1502
1503 if (fstat(fd, &st) != 0)
1504 return -1;
1505 sprintf(path, "/sys/dev/block/%d:%d/holders",
1506 (int)major(st.st_rdev), (int)minor(st.st_rdev));
1507 e = path + strlen(path);
1508
1509 dir = opendir(path);
1510 if (!dir)
1511 return -1;
1512 while ((de = readdir(dir))) {
1513 if (de->d_ino == 0)
1514 continue;
1515 if (de->d_name[0] == '.')
1516 continue;
a994592d
N
1517 /* Need to make sure it is a container and not a volume */
1518 sprintf(e, "/%s/md/metadata_version", de->d_name);
1519 dfd = open(path, O_RDONLY);
1520 if (dfd < 0)
1521 continue;
1522 n = read(dfd, buf, sizeof(buf));
1523 close(dfd);
1524 if (n <= 0 || (unsigned)n >= sizeof(buf))
1525 continue;
1526 buf[n] = 0;
1527 if (strncmp(buf, "external", 8) != 0 ||
1528 n < 10 ||
1529 buf[9] == '/')
1530 continue;
a322f70c
DW
1531 sprintf(e, "/%s/dev", de->d_name);
1532 dfd = open(path, O_RDONLY);
1533 if (dfd < 0)
1534 continue;
1535 n = read(dfd, buf, sizeof(buf));
1536 close(dfd);
f21e18ca 1537 if (n <= 0 || (unsigned)n >= sizeof(buf))
a322f70c
DW
1538 continue;
1539 buf[n] = 0;
1540 if (sscanf(buf, "%d:%d", &major, &minor) != 2)
1541 continue;
1542 sprintf(buf, "%d:%d", major, minor);
1543 dfd = dev_open(buf, O_RDONLY);
1544 if (dfd >= 0) {
1545 closedir(dir);
1546 return dfd;
1547 }
1548 }
355726fa 1549 closedir(dir);
a322f70c
DW
1550 return -1;
1551}
1552
33414a01
DW
1553struct superswitch *version_to_superswitch(char *vers)
1554{
1555 int i;
1556
1557 for (i = 0; superlist[i]; i++) {
1558 struct superswitch *ss = superlist[i];
1559
1560 if (strcmp(vers, ss->name) == 0)
1561 return ss;
1562 }
1563
1564 return NULL;
1565}
1566
4dd2df09
N
1567int metadata_container_matches(char *metadata, char *devnm)
1568{
1569 /* Check if 'devnm' is the container named in 'metadata'
1570 * which is
1571 * /containername/componentname or
1572 * -containername/componentname
1573 */
1574 int l;
1575 if (*metadata != '/' && *metadata != '-')
1576 return 0;
1577 l = strlen(devnm);
1578 if (strncmp(metadata+1, devnm, l) != 0)
1579 return 0;
1580 if (metadata[l+1] != '/')
1581 return 0;
1582 return 1;
1583}
1584
1585int metadata_subdev_matches(char *metadata, char *devnm)
1586{
1587 /* Check if 'devnm' is the subdev named in 'metadata'
1588 * which is
1589 * /containername/subdev or
1590 * -containername/subdev
1591 */
1592 char *sl;
1593 if (*metadata != '/' && *metadata != '-')
1594 return 0;
1595 sl = strchr(metadata+1, '/');
1596 if (!sl)
1597 return 0;
1598 if (strcmp(sl+1, devnm) == 0)
1599 return 1;
1600 return 0;
1601}
1602
33414a01
DW
1603int is_container_member(struct mdstat_ent *mdstat, char *container)
1604{
1605 if (mdstat->metadata_version == NULL ||
1606 strncmp(mdstat->metadata_version, "external:", 9) != 0 ||
4dd2df09 1607 !metadata_container_matches(mdstat->metadata_version+9, container))
33414a01
DW
1608 return 0;
1609
1610 return 1;
1611}
1612
1613int is_subarray_active(char *subarray, char *container)
1614{
1615 struct mdstat_ent *mdstat = mdstat_read(0, 0);
1616 struct mdstat_ent *ent;
1617
e5408a32
DW
1618 for (ent = mdstat; ent; ent = ent->next)
1619 if (is_container_member(ent, container))
e5e5d7ce 1620 if (strcmp(to_subarray(ent, container), subarray) == 0)
33414a01 1621 break;
33414a01
DW
1622
1623 free_mdstat(mdstat);
1624
1625 return ent != NULL;
1626}
1627
1628/* open_subarray - opens a subarray in a container
1629 * @dev: container device name
feab51f8 1630 * @st: empty supertype
33414a01
DW
1631 * @quiet: block reporting errors flag
1632 *
1633 * On success returns an fd to a container and fills in *st
1634 */
feab51f8 1635int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet)
33414a01
DW
1636{
1637 struct mdinfo *mdi;
a951a4f7 1638 struct mdinfo *info;
33414a01 1639 int fd, err = 1;
4dd2df09 1640 char *_devnm;
33414a01
DW
1641
1642 fd = open(dev, O_RDWR|O_EXCL);
1643 if (fd < 0) {
1644 if (!quiet)
e7b84f9d 1645 pr_err("Couldn't open %s, aborting\n",
33414a01 1646 dev);
b990032d 1647 return -1;
33414a01
DW
1648 }
1649
4dd2df09
N
1650 _devnm = fd2devnm(fd);
1651 if (_devnm == NULL) {
33414a01 1652 if (!quiet)
e7b84f9d
N
1653 pr_err("Failed to determine device number for %s\n",
1654 dev);
33414a01
DW
1655 goto close_fd;
1656 }
4dd2df09 1657 strcpy(st->devnm, _devnm);
33414a01 1658
4dd2df09 1659 mdi = sysfs_read(fd, st->devnm, GET_VERSION|GET_LEVEL);
33414a01
DW
1660 if (!mdi) {
1661 if (!quiet)
e7b84f9d 1662 pr_err("Failed to read sysfs for %s\n",
33414a01
DW
1663 dev);
1664 goto close_fd;
1665 }
1666
1667 if (mdi->array.level != UnSet) {
1668 if (!quiet)
e7b84f9d 1669 pr_err("%s is not a container\n", dev);
33414a01
DW
1670 goto free_sysfs;
1671 }
1672
1673 st->ss = version_to_superswitch(mdi->text_version);
1674 if (!st->ss) {
1675 if (!quiet)
e7b84f9d
N
1676 pr_err("Operation not supported for %s metadata\n",
1677 mdi->text_version);
33414a01
DW
1678 goto free_sysfs;
1679 }
1680
4dd2df09 1681 if (st->devnm[0] == 0) {
33414a01 1682 if (!quiet)
e7b84f9d 1683 pr_err("Failed to allocate device name\n");
33414a01
DW
1684 goto free_sysfs;
1685 }
1686
db20d413 1687 if (!st->ss->load_container) {
33414a01 1688 if (!quiet)
e7b84f9d 1689 pr_err("%s is not a container\n", dev);
4dd2df09 1690 goto free_sysfs;
33414a01
DW
1691 }
1692
db20d413 1693 if (st->ss->load_container(st, fd, NULL)) {
33414a01 1694 if (!quiet)
e7b84f9d 1695 pr_err("Failed to load metadata for %s\n",
db20d413 1696 dev);
4dd2df09 1697 goto free_sysfs;
33414a01
DW
1698 }
1699
a951a4f7
N
1700 info = st->ss->container_content(st, subarray);
1701 if (!info) {
1702 if (!quiet)
e7b84f9d 1703 pr_err("Failed to find subarray-%s in %s\n",
a951a4f7
N
1704 subarray, dev);
1705 goto free_super;
1706 }
1707 free(info);
1708
33414a01
DW
1709 err = 0;
1710
1711 free_super:
1712 if (err)
1713 st->ss->free_super(st);
33414a01
DW
1714 free_sysfs:
1715 sysfs_free(mdi);
1716 close_fd:
1717 if (err)
1718 close(fd);
1719
1720 if (err)
1721 return -1;
1722 else
1723 return fd;
1724}
1725
7801ac20
N
1726int add_disk(int mdfd, struct supertype *st,
1727 struct mdinfo *sra, struct mdinfo *info)
1728{
1729 /* Add a device to an array, in one of 2 ways. */
1730 int rv;
32141c17 1731
7801ac20 1732 if (st->ss->external) {
d23534e4
DW
1733 if (info->disk.state & (1<<MD_DISK_SYNC))
1734 info->recovery_start = MaxSector;
1735 else
1736 info->recovery_start = 0;
2904b26f 1737 rv = sysfs_add_disk(sra, info, 0);
7801ac20
N
1738 if (! rv) {
1739 struct mdinfo *sd2;
f35f2525
N
1740 for (sd2 = sra->devs; sd2; sd2=sd2->next)
1741 if (sd2 == info)
1742 break;
1743 if (sd2 == NULL) {
503975b9 1744 sd2 = xmalloc(sizeof(*sd2));
f35f2525
N
1745 *sd2 = *info;
1746 sd2->next = sra->devs;
1747 sra->devs = sd2;
1748 }
7801ac20
N
1749 }
1750 } else
7801ac20
N
1751 rv = ioctl(mdfd, ADD_NEW_DISK, &info->disk);
1752 return rv;
1753}
1754
de6ae750
N
1755int remove_disk(int mdfd, struct supertype *st,
1756 struct mdinfo *sra, struct mdinfo *info)
1757{
1758 int rv;
32141c17 1759
de6ae750 1760 /* Remove the disk given by 'info' from the array */
de6ae750
N
1761 if (st->ss->external)
1762 rv = sysfs_set_str(sra, info, "slot", "none");
1763 else
de6ae750
N
1764 rv = ioctl(mdfd, HOT_REMOVE_DISK, makedev(info->disk.major,
1765 info->disk.minor));
1766 return rv;
1767}
1768
1ab9ed2a 1769int hot_remove_disk(int mdfd, unsigned long dev, int force)
2dd271fe 1770{
1ab9ed2a 1771 int cnt = force ? 500 : 5;
2dd271fe
N
1772 int ret;
1773
1774 /* HOT_REMOVE_DISK can fail with EBUSY if there are
1775 * outstanding IO requests to the device.
1776 * In this case, it can be helpful to wait a little while,
1ab9ed2a 1777 * up to 5 seconds if 'force' is set, or 50 msec if not.
2dd271fe
N
1778 */
1779 while ((ret = ioctl(mdfd, HOT_REMOVE_DISK, dev)) == -1 &&
1780 errno == EBUSY &&
1781 cnt-- > 0)
1782 usleep(10000);
1783
1784 return ret;
1785}
1786
1ab9ed2a 1787int sys_hot_remove_disk(int statefd, int force)
fdd01569 1788{
1ab9ed2a 1789 int cnt = force ? 500 : 5;
fdd01569
N
1790 int ret;
1791
1792 while ((ret = write(statefd, "remove", 6)) == -1 &&
1793 errno == EBUSY &&
1794 cnt-- > 0)
1795 usleep(10000);
1796 return ret == 6 ? 0 : -1;
1797}
1798
f35f2525
N
1799int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info)
1800{
1801 /* Initialise kernel's knowledge of array.
1802 * This varies between externally managed arrays
1803 * and older kernels
1804 */
700483a2 1805 mdu_array_info_t inf;
f35f2525
N
1806 int rv;
1807
f35f2525 1808 if (st->ss->external)
700483a2 1809 return sysfs_set_array(info, 9003);
700483a2
JS
1810
1811 memset(&inf, 0, sizeof(inf));
1812 inf.major_version = info->array.major_version;
1813 inf.minor_version = info->array.minor_version;
1814 rv = md_set_array_info(mdfd, &inf);
1815
f35f2525
N
1816 return rv;
1817}
1818
1e5c6983
DW
1819unsigned long long min_recovery_start(struct mdinfo *array)
1820{
1821 /* find the minimum recovery_start in an array for metadata
1822 * formats that only record per-array recovery progress instead
1823 * of per-device
1824 */
1825 unsigned long long recovery_start = MaxSector;
1826 struct mdinfo *d;
1827
1828 for (d = array->devs; d; d = d->next)
1829 recovery_start = min(recovery_start, d->recovery_start);
1830
1831 return recovery_start;
1832}
1833
4dd2df09 1834int mdmon_pid(char *devnm)
a931db9e
NB
1835{
1836 char path[100];
1837 char pid[10];
1838 int fd;
1839 int n;
10013317 1840
4dd2df09 1841 sprintf(path, "%s/%s.pid", MDMON_DIR, devnm);
10013317 1842
24f6f99b 1843 fd = open(path, O_RDONLY | O_NOATIME, 0);
a931db9e
NB
1844
1845 if (fd < 0)
cf556303 1846 return -1;
a931db9e
NB
1847 n = read(fd, pid, 9);
1848 close(fd);
1849 if (n <= 0)
cf556303 1850 return -1;
24f6f99b 1851 return atoi(pid);
a931db9e
NB
1852}
1853
4dd2df09 1854int mdmon_running(char *devnm)
a931db9e 1855{
4dd2df09 1856 int pid = mdmon_pid(devnm);
24f6f99b 1857 if (pid <= 0)
a931db9e 1858 return 0;
24f6f99b 1859 if (kill(pid, 0) == 0)
a931db9e
NB
1860 return 1;
1861 return 0;
1862}
1863
4dd2df09 1864int start_mdmon(char *devnm)
8850ee3e 1865{
1913c325 1866 int i, skipped;
44d2e365 1867 int len;
ca3b6696 1868 pid_t pid;
9fe32043 1869 int status;
44d2e365
N
1870 char pathbuf[1024];
1871 char *paths[4] = {
1872 pathbuf,
85945e19 1873 BINDIR "/mdmon",
2f1bcf43 1874 "./mdmon",
44d2e365
N
1875 NULL
1876 };
8850ee3e 1877
40ebbb9c 1878 if (check_env("MDADM_NO_MDMON"))
8850ee3e
N
1879 return 0;
1880
9cf014ec 1881 len = readlink("/proc/self/exe", pathbuf, sizeof(pathbuf)-1);
44d2e365
N
1882 if (len > 0) {
1883 char *sl;
1884 pathbuf[len] = 0;
1885 sl = strrchr(pathbuf, '/');
1886 if (sl)
1887 sl++;
1888 else
1889 sl = pathbuf;
1890 strcpy(sl, "mdmon");
1891 } else
1892 pathbuf[0] = '\0';
1893
0f7bdf89 1894 /* First try to run systemctl */
9540cc24
N
1895 if (!check_env("MDADM_NO_SYSTEMCTL"))
1896 switch(fork()) {
1897 case 0:
1898 /* FIXME yuk. CLOSE_EXEC?? */
1899 skipped = 0;
1900 for (i = 3; skipped < 20; i++)
1901 if (close(i) < 0)
1902 skipped++;
1903 else
1904 skipped = 0;
1905
1906 /* Don't want to see error messages from
1907 * systemctl. If the service doesn't exist,
1908 * we start mdmon ourselves.
1909 */
1910 close(2);
1911 open("/dev/null", O_WRONLY);
1912 snprintf(pathbuf, sizeof(pathbuf), "mdmon@%s.service",
1913 devnm);
1914 status = execl("/usr/bin/systemctl", "systemctl",
1915 "start",
1916 pathbuf, NULL);
1917 status = execl("/bin/systemctl", "systemctl", "start",
1918 pathbuf, NULL);
1919 exit(1);
7a862a02 1920 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
9540cc24
N
1921 return -1;
1922 default: /* parent - good */
1923 pid = wait(&status);
1924 if (pid >= 0 && status == 0)
1925 return 0;
1926 }
0f7bdf89
JS
1927
1928 /* That failed, try running mdmon directly */
8850ee3e
N
1929 switch(fork()) {
1930 case 0:
1931 /* FIXME yuk. CLOSE_EXEC?? */
1913c325 1932 skipped = 0;
ca3b6696 1933 for (i = 3; skipped < 20; i++)
1913c325
N
1934 if (close(i) < 0)
1935 skipped++;
1936 else
1937 skipped = 0;
1938
ca3b6696 1939 for (i = 0; paths[i]; i++)
a0963a86 1940 if (paths[i][0]) {
2f1bcf43 1941 execl(paths[i], paths[i],
4dd2df09 1942 devnm, NULL);
a0963a86 1943 }
8850ee3e 1944 exit(1);
7a862a02 1945 case -1: pr_err("cannot run mdmon. Array remains readonly\n");
8850ee3e 1946 return -1;
9fe32043
N
1947 default: /* parent - good */
1948 pid = wait(&status);
15c10423 1949 if (pid < 0 || status != 0) {
7a862a02 1950 pr_err("failed to launch mdmon. Array remains readonly\n");
9fe32043 1951 return -1;
15c10423 1952 }
8850ee3e
N
1953 }
1954 return 0;
1955}
1956
148acb7b
DW
1957__u32 random32(void)
1958{
1959 __u32 rv;
1960 int rfd = open("/dev/urandom", O_RDONLY);
1961 if (rfd < 0 || read(rfd, &rv, 4) != 4)
1962 rv = random();
1963 if (rfd >= 0)
1964 close(rfd);
1965 return rv;
1966}
1967
c5f71c24
JS
1968void random_uuid(__u8 *buf)
1969{
1970 int fd, i, len;
1971 __u32 r[4];
1972
1973 fd = open("/dev/urandom", O_RDONLY);
1974 if (fd < 0)
1975 goto use_random;
1976 len = read(fd, buf, 16);
1977 close(fd);
1978 if (len != 16)
1979 goto use_random;
1980
1981 return;
1982
1983use_random:
1984 for (i = 0; i < 4; i++)
1985 r[i] = random();
1986 memcpy(buf, r, 16);
1987}
1988
edd8d13c
NB
1989int flush_metadata_updates(struct supertype *st)
1990{
1991 int sfd;
1992 if (!st->updates) {
1993 st->update_tail = NULL;
1994 return -1;
1995 }
1996
4dd2df09 1997 sfd = connect_monitor(st->container_devnm);
edd8d13c
NB
1998 if (sfd < 0)
1999 return -1;
2000
2001 while (st->updates) {
2002 struct metadata_update *mu = st->updates;
2003 st->updates = mu->next;
2004
2005 send_message(sfd, mu, 0);
2006 wait_reply(sfd, 0);
2007 free(mu->buf);
2008 free(mu);
2009 }
2010 ack(sfd, 0);
2011 wait_reply(sfd, 0);
2012 close(sfd);
2013 st->update_tail = NULL;
2014 return 0;
2015}
2016
2017void append_metadata_update(struct supertype *st, void *buf, int len)
2018{
2019
503975b9 2020 struct metadata_update *mu = xmalloc(sizeof(*mu));
edd8d13c
NB
2021
2022 mu->buf = buf;
2023 mu->len = len;
2024 mu->space = NULL;
cb23f1f4 2025 mu->space_list = NULL;
edd8d13c
NB
2026 mu->next = NULL;
2027 *st->update_tail = mu;
2028 st->update_tail = &mu->next;
2029}
a931db9e 2030
fe6729fa
NB
2031#ifdef __TINYC__
2032/* tinyc doesn't optimize this check in ioctl.h out ... */
2033unsigned int __invalid_size_argument_for_IOC = 0;
2034#endif
2035
6d11ec6f
AK
2036int experimental(void)
2037{
2038 if (check_env("MDADM_EXPERIMENTAL"))
2039 return 1;
2040 else {
7a862a02 2041 pr_err("To use this feature MDADM_EXPERIMENTAL environment variable has to be defined.\n");
6d11ec6f
AK
2042 return 0;
2043 }
2044}
2045
326727d9
AC
2046/* Pick all spares matching given criteria from a container
2047 * if min_size == 0 do not check size
2048 * if domlist == NULL do not check domains
2049 * if spare_group given add it to domains of each spare
2050 * metadata allows to test domains using metadata of destination array */
2051struct mdinfo *container_choose_spares(struct supertype *st,
2052 unsigned long long min_size,
2053 struct domainlist *domlist,
2054 char *spare_group,
2055 const char *metadata, int get_one)
2056{
2057 struct mdinfo *d, **dp, *disks = NULL;
2058
2059 /* get list of all disks in container */
2060 if (st->ss->getinfo_super_disks)
2061 disks = st->ss->getinfo_super_disks(st);
2062
2063 if (!disks)
2064 return disks;
2065 /* find spare devices on the list */
2066 dp = &disks->devs;
2067 disks->array.spare_disks = 0;
2068 while (*dp) {
2069 int found = 0;
2070 d = *dp;
2071 if (d->disk.state == 0) {
2072 /* check if size is acceptable */
2073 unsigned long long dev_size;
2074 dev_t dev = makedev(d->disk.major,d->disk.minor);
2075
2076 if (!min_size ||
2077 (dev_size_from_id(dev, &dev_size) &&
2078 dev_size >= min_size))
2079 found = 1;
2080 /* check if domain matches */
2081 if (found && domlist) {
4dd2df09 2082 struct dev_policy *pol = devid_policy(dev);
326727d9
AC
2083 if (spare_group)
2084 pol_add(&pol, pol_domain,
2085 spare_group, NULL);
e5508b36 2086 if (domain_test(domlist, pol, metadata) != 1)
326727d9
AC
2087 found = 0;
2088 dev_policy_free(pol);
2089 }
2090 }
2091 if (found) {
2092 dp = &d->next;
2093 disks->array.spare_disks++;
2094 if (get_one) {
2095 sysfs_free(*dp);
2096 d->next = NULL;
2097 }
2098 } else {
2099 *dp = d->next;
2100 d->next = NULL;
2101 sysfs_free(d);
2102 }
2103 }
2104 return disks;
2105}
9eafa1de
MN
2106
2107/* Checks if paths point to the same device
2108 * Returns 0 if they do.
2109 * Returns 1 if they don't.
2110 * Returns -1 if something went wrong,
2111 * e.g. paths are empty or the files
2112 * they point to don't exist */
2113int compare_paths (char* path1, char* path2)
2114{
2115 struct stat st1,st2;
2116
2117 if (path1 == NULL || path2 == NULL)
2118 return -1;
2119 if (stat(path1,&st1) != 0)
2120 return -1;
2121 if (stat(path2,&st2) != 0)
2122 return -1;
2123 if ((st1.st_ino == st2.st_ino) && (st1.st_dev == st2.st_dev))
2124 return 0;
2125 return 1;
2126}
a7dec3fd
N
2127
2128/* Make sure we can open as many devices as needed */
2129void enable_fds(int devices)
2130{
2131 unsigned int fds = 20 + devices;
2132 struct rlimit lim;
2133 if (getrlimit(RLIMIT_NOFILE, &lim) != 0
2134 || lim.rlim_cur >= fds)
2135 return;
2136 if (lim.rlim_max < fds)
2137 lim.rlim_max = fds;
2138 lim.rlim_cur = fds;
2139 setrlimit(RLIMIT_NOFILE, &lim);
2140}
a9c15847
N
2141
2142int in_initrd(void)
2143{
2144 /* This is based on similar function in systemd. */
2145 struct statfs s;
76d0f188
JS
2146 /* statfs.f_type is signed long on s390x and MIPS, causing all
2147 sorts of sign extension problems with RAMFS_MAGIC being
2148 defined as 0x858458f6 */
a9c15847 2149 return statfs("/", &s) >= 0 &&
7ac5d47e 2150 ((unsigned long)s.f_type == TMPFS_MAGIC ||
76d0f188
JS
2151 ((unsigned long)s.f_type & 0xFFFFFFFFUL) ==
2152 ((unsigned long)RAMFS_MAGIC & 0xFFFFFFFFUL));
a9c15847 2153}
8832342d
N
2154
2155void reopen_mddev(int mdfd)
2156{
2157 /* Re-open without any O_EXCL, but keep
2158 * the same fd
2159 */
2160 char *devnm;
2161 int fd;
2162 devnm = fd2devnm(mdfd);
2163 close(mdfd);
2164 fd = open_dev(devnm);
2165 if (fd >= 0 && fd != mdfd)
2166 dup2(fd, mdfd);
2167}
7716570e 2168
e80357f8
GJ
2169static struct cmap_hooks *cmap_hooks = NULL;
2170static int is_cmap_hooks_ready = 0;
7716570e 2171
e80357f8
GJ
2172void set_cmap_hooks(void)
2173{
2174 cmap_hooks = xmalloc(sizeof(struct cmap_hooks));
2175 cmap_hooks->cmap_handle = dlopen("libcmap.so.4", RTLD_NOW | RTLD_LOCAL);
2176 if (!cmap_hooks->cmap_handle)
2177 return;
7716570e 2178
e80357f8
GJ
2179 cmap_hooks->initialize = dlsym(cmap_hooks->cmap_handle, "cmap_initialize");
2180 cmap_hooks->get_string = dlsym(cmap_hooks->cmap_handle, "cmap_get_string");
2181 cmap_hooks->finalize = dlsym(cmap_hooks->cmap_handle, "cmap_finalize");
7716570e 2182
e80357f8
GJ
2183 if (!cmap_hooks->initialize || !cmap_hooks->get_string ||
2184 !cmap_hooks->finalize)
2185 dlclose(cmap_hooks->cmap_handle);
2186 else
2187 is_cmap_hooks_ready = 1;
2188}
7716570e 2189
e80357f8
GJ
2190int get_cluster_name(char **cluster_name)
2191{
2192 int rv = -1;
2193 cmap_handle_t handle;
7716570e 2194
e80357f8
GJ
2195 if (!is_cmap_hooks_ready)
2196 return rv;
7716570e 2197
e80357f8 2198 rv = cmap_hooks->initialize(&handle);
7716570e
GJ
2199 if (rv != CS_OK)
2200 goto out;
2201
e80357f8 2202 rv = cmap_hooks->get_string(handle, "totem.cluster_name", cluster_name);
7716570e
GJ
2203 if (rv != CS_OK) {
2204 free(*cluster_name);
2205 rv = -1;
2206 goto name_err;
2207 }
2208
2209 rv = 0;
2210name_err:
e80357f8 2211 cmap_hooks->finalize(handle);
7716570e 2212out:
7716570e
GJ
2213 return rv;
2214}
d15a1f72
GJ
2215
2216void set_dlm_hooks(void)
2217{
2218 dlm_hooks = xmalloc(sizeof(struct dlm_hooks));
2219 dlm_hooks->dlm_handle = dlopen("libdlm_lt.so.3", RTLD_NOW | RTLD_LOCAL);
2220 if (!dlm_hooks->dlm_handle)
2221 return;
2222
2223 dlm_hooks->create_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_create_lockspace");
2224 dlm_hooks->release_lockspace = dlsym(dlm_hooks->dlm_handle, "dlm_release_lockspace");
2225 dlm_hooks->ls_lock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_lock");
2226 dlm_hooks->ls_unlock = dlsym(dlm_hooks->dlm_handle, "dlm_ls_unlock");
2227 dlm_hooks->ls_get_fd = dlsym(dlm_hooks->dlm_handle, "dlm_ls_get_fd");
2228 dlm_hooks->dispatch = dlsym(dlm_hooks->dlm_handle, "dlm_dispatch");
2229
2230 if (!dlm_hooks->create_lockspace || !dlm_hooks->ls_lock ||
2231 !dlm_hooks->ls_unlock || !dlm_hooks->release_lockspace ||
2232 !dlm_hooks->ls_get_fd || !dlm_hooks->dispatch)
2233 dlclose(dlm_hooks->dlm_handle);
2234 else
2235 is_dlm_hooks_ready = 1;
2236}
e80357f8
GJ
2237
2238void set_hooks(void)
2239{
2240 set_dlm_hooks();
2241 set_cmap_hooks();
2242}