]> git.ipfire.org Git - thirdparty/mdadm.git/blame - Assemble.c
Assemble: small cleanup of error checking.
[thirdparty/mdadm.git] / Assemble.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
e736b623 4 * Copyright (C) 2001-2009 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"
41a3b72a 26#include <ctype.h>
64c4757e 27
624920bb
NB
28static int name_matches(char *found, char *required, char *homehost)
29{
30 /* See if the name found matches the required name, possibly
31 * prefixed with 'homehost'
32 */
33 char fnd[33];
34
35 strncpy(fnd, found, 32);
36 fnd[32] = 0;
37 if (strcmp(found, required)==0)
38 return 1;
39 if (homehost) {
40 int l = strlen(homehost);
41 if (l < 32 && fnd[l] == ':' &&
42 strcmp(fnd+l+1, required)==0)
43 return 1;
44 }
45 return 0;
46}
47
9008ed1c 48static int is_member_busy(char *metadata_version)
2de8884f
DW
49{
50 /* check if the given member array is active */
51 struct mdstat_ent *mdstat = mdstat_read(1, 0);
52 struct mdstat_ent *ent;
53 int busy = 0;
54
55 for (ent = mdstat; ent; ent = ent->next) {
56 if (ent->metadata_version == NULL)
57 continue;
58 if (strncmp(ent->metadata_version, "external:", 9) != 0)
59 continue;
60 if (!is_subarray(&ent->metadata_version[9]))
61 continue;
62 /* Skip first char - it can be '/' or '-' */
63 if (strcmp(&ent->metadata_version[10], metadata_version+1) == 0) {
64 busy = 1;
65 break;
66 }
67 }
68 free_mdstat(mdstat);
69
70 return busy;
71}
72
7f91af49 73int Assemble(struct supertype *st, char *mddev,
8aec876d 74 mddev_ident_t ident,
06b0d786 75 mddev_dev_t devlist, char *backup_file,
64c4757e 76 int readonly, int runstop,
0ac91628 77 char *update, char *homehost, int require_homehost,
64c4757e
NB
78 int verbose, int force)
79{
80 /*
52826846
NB
81 * The task of Assemble is to find a collection of
82 * devices that should (according to their superblocks)
83 * form an array, and to give this collection to the MD driver.
84 * In Linux-2.4 and later, this involves submitting a
64c4757e
NB
85 * SET_ARRAY_INFO ioctl with no arg - to prepare
86 * the array - and then submit a number of
87 * ADD_NEW_DISK ioctls to add disks into
88 * the array. Finally RUN_ARRAY might
89 * be submitted to start the array.
90 *
91 * Much of the work of Assemble is in finding and/or
92 * checking the disks to make sure they look right.
93 *
4b1ac34b 94 * If mddev is not set, then scan must be set and we
64c4757e
NB
95 * read through the config file for dev+uuid mapping
96 * We recurse, setting mddev, for each device that
97 * - isn't running
4b1ac34b 98 * - has a valid uuid (or any uuid if !uuidset)
64c4757e
NB
99 *
100 * If mddev is set, we try to determine state of md.
101 * check version - must be at least 0.90.0
102 * check kernel version. must be at least 2.4.
103 * If not, we can possibly fall back on START_ARRAY
104 * Try to GET_ARRAY_INFO.
105 * If possible, give up
106 * If not, try to STOP_ARRAY just to make sure
107 *
108 * If !uuidset and scan, look in conf-file for uuid
109 * If not found, give up
b8a8ccf9 110 * If !devlist and scan and uuidset, get list of devs from conf-file
64c4757e
NB
111 *
112 * For each device:
113 * Check superblock - discard if bad
114 * Check uuid (set if we don't have one) - discard if no match
5787fa49 115 * Check superblock similarity if we have a superblock - discard if different
4b1ac34b 116 * Record events, devicenum
64c4757e 117 * This should give us a list of devices for the array
4b1ac34b 118 * We should collect the most recent event number
64c4757e
NB
119 *
120 * Count disks with recent enough event count
121 * While force && !enough disks
122 * Choose newest rejected disks, update event count
123 * mark clean and rewrite superblock
124 * If recent kernel:
125 * SET_ARRAY_INFO
126 * foreach device with recent events : ADD_NEW_DISK
127 * if runstop == 1 || "enough" disks and runstop==0 -> RUN_ARRAY
128 * If old kernel:
129 * Check the device numbers in superblock are right
130 * update superblock if any changes
131 * START_ARRAY
132 *
133 */
c30e5369
N
134 int mdfd;
135 int clean;
f05641cf
N
136 int auto_assem = (mddev == NULL && !ident->uuid_set &&
137 ident->super_minor == UnSet && ident->name[0] == 0
aa7c284c 138 && (ident->container == NULL || ident->member == NULL));
64c4757e 139 int old_linux = 0;
c30e5369 140 int vers = vers; /* Keep gcc quite - it really is initialised */
64c4757e
NB
141 struct {
142 char *devname;
213ee40b
NB
143 int uptodate; /* set once we decide that this device is as
144 * recent as everything else in the array.
145 */
146 struct mdinfo i;
5787fa49 147 } *devices;
02e7c5b7 148 char *devmap;
56eedc1a 149 int *best = NULL; /* indexed by raid_disk */
f21e18ca 150 int bestcnt = 0;
98c6faba 151 int devcnt = 0;
1ff98339 152 unsigned int okcnt, sparecnt, rebuilding_cnt;
98c6faba 153 unsigned int req_cnt;
f21e18ca 154 int i;
64c4757e 155 int most_recent = 0;
cd29a5c8 156 int chosen_drive;
52826846 157 int change = 0;
cd29a5c8 158 int inargv = 0;
52437b4f 159 int report_missmatch;
bf4fb153 160 int bitmap_done;
7f91af49
N
161 int start_partial_ok = (runstop >= 0) &&
162 (force || devlist==NULL || auto_assem);
98c6faba 163 unsigned int num_devs;
5787fa49 164 mddev_dev_t tmpdev;
4b1ac34b 165 struct mdinfo info;
98dbd966 166 struct mdinfo *content = NULL;
265e0f17 167 char *avail;
308e1801 168 int nextspare = 0;
197e3eb6 169 char *name = NULL;
69207ff6 170 int trustworthy;
a04d5763 171 char chosen_name[1024];
8a46fe84 172
682c7051 173 if (get_linux_version() < 2004000)
64c4757e
NB
174 old_linux = 1;
175
64c4757e 176 /*
52826846
NB
177 * If any subdevs are listed, then any that don't
178 * match ident are discarded. Remainder must all match and
179 * become the array.
180 * If no subdevs, then we scan all devices in the config file, but
181 * there must be something in the identity
64c4757e 182 */
64c4757e 183
cd29a5c8 184 if (!devlist &&
52826846 185 ident->uuid_set == 0 &&
f21e18ca 186 (ident->super_minor < 0 || ident->super_minor == UnSet) &&
e0fe762a
N
187 ident->name[0] == 0 &&
188 (ident->container == NULL || ident->member == NULL) &&
52826846
NB
189 ident->devices == NULL) {
190 fprintf(stderr, Name ": No identity information available for %s - cannot assemble.\n",
8a46fe84 191 mddev ? mddev : "further assembly");
52826846 192 return 1;
64c4757e 193 }
71d60c48 194
cd29a5c8 195 if (devlist == NULL)
8aec876d 196 devlist = conf_get_devs();
7f91af49 197 else if (mddev)
da6b5ca9 198 inargv = 1;
64c4757e 199
52437b4f 200 report_missmatch = ((inargv && verbose >= 0) || verbose > 0);
60e1bc1a 201 try_again:
c30e5369
N
202 /* We come back here when doing auto-assembly and attempting some
203 * set of devices failed. Those are now marked as ->used==2 and
204 * we ignore them and try again
205 */
60e1bc1a 206
5787fa49
NB
207 tmpdev = devlist; num_devs = 0;
208 while (tmpdev) {
da6b5ca9
NB
209 if (tmpdev->used)
210 tmpdev->used = 2;
211 else
212 num_devs++;
5787fa49
NB
213 tmpdev = tmpdev->next;
214 }
5787fa49 215
82d9eba6 216 if (!st && ident->st) st = ident->st;
64c4757e 217
dab6685f 218 if (verbose>0)
82b27616 219 fprintf(stderr, Name ": looking for devices for %s\n",
8a46fe84 220 mddev ? mddev : "further assembly");
82b27616 221
811e6cbe
NB
222 /* first walk the list of devices to find a consistent set
223 * that match the criterea, if that is possible.
c30e5369 224 * We flag the ones we like with 'used'.
811e6cbe
NB
225 */
226 for (tmpdev = devlist;
227 tmpdev;
228 tmpdev = tmpdev->next) {
229 char *devname = tmpdev->devname;
64c4757e
NB
230 int dfd;
231 struct stat stb;
3da92f27 232 struct supertype *tst = dup_super(st);
4e8d9f0a 233 struct dev_policy *pol = NULL;
52826846 234
da6b5ca9
NB
235 if (tmpdev->used > 1) continue;
236
52826846 237 if (ident->devices &&
56eedc1a 238 !match_oneof(ident->devices, devname)) {
52437b4f 239 if (report_missmatch)
56eedc1a 240 fprintf(stderr, Name ": %s is not one of %s\n", devname, ident->devices);
52826846 241 continue;
56eedc1a 242 }
4b1ac34b 243
8b0dabea 244 dfd = dev_open(devname, O_RDONLY|O_EXCL);
64c4757e 245 if (dfd < 0) {
52437b4f 246 if (report_missmatch)
682c7051 247 fprintf(stderr, Name ": cannot open device %s: %s\n",
64c4757e 248 devname, strerror(errno));
da6b5ca9 249 tmpdev->used = 2;
52826846
NB
250 } else if (fstat(dfd, &stb)< 0) {
251 /* Impossible! */
252 fprintf(stderr, Name ": fstat failed for %s: %s\n",
253 devname, strerror(errno));
da6b5ca9 254 tmpdev->used = 2;
cd29a5c8
NB
255 } else if ((stb.st_mode & S_IFMT) != S_IFBLK) {
256 fprintf(stderr, Name ": %s is not a block device.\n",
52826846 257 devname);
da6b5ca9 258 tmpdev->used = 2;
82d9eba6 259 } else if (!tst && (tst = guess_super(dfd)) == NULL) {
52437b4f 260 if (report_missmatch)
da6b5ca9
NB
261 fprintf(stderr, Name ": no recogniseable superblock on %s\n",
262 devname);
263 tmpdev->used = 2;
1ff98339
N
264 } else if (tst->ss->load_super(tst,dfd, NULL)) {
265 if (report_missmatch)
0f22b998
N
266 fprintf(stderr, Name ": no RAID superblock on %s\n",
267 devname);
d68ea4d7 268 tmpdev->used = 2;
0f22b998
N
269 } else if (tst->ss->compare_super == NULL) {
270 if (report_missmatch)
271 fprintf(stderr, Name ": Cannot assemble %s metadata on %s\n",
272 tst->ss->name, devname);
0f22b998 273 tmpdev->used = 2;
31015d57 274 } else if (auto_assem && st == NULL &&
4e8d9f0a 275 !conf_test_metadata(tst->ss->name, (pol = devnum_policy(stb.st_rdev)),
d1d3482b 276 tst->ss->match_home(tst, homehost) == 1)) {
31015d57
N
277 if (report_missmatch)
278 fprintf(stderr, Name ": %s has metadata type %s for which "
279 "auto-assembly is disabled\n",
280 devname, tst->ss->name);
281 tmpdev->used = 2;
52826846 282 } else {
98dbd966
DW
283 content = &info;
284 memset(content, 0, sizeof(*content));
a5d85af7 285 tst->ss->getinfo_super(tst, content, NULL);
64c4757e 286 }
c06487ce 287 if (dfd >= 0) close(dfd);
d68ea4d7
N
288 if (tmpdev->used == 2) {
289 if (auto_assem)
290 /* Ignore unrecognised devices during auto-assembly */
291 goto loop;
292 if (ident->uuid_set || ident->name[0] ||
293 ident->super_minor != UnSet)
294 /* Ignore unrecognised device if looking for
295 * specific array */
296 goto loop;
297
298
299 fprintf(stderr, Name ": %s has no superblock - assembly aborted\n",
300 devname);
301 if (st)
302 st->ss->free_super(st);
303 dev_policy_free(pol);
304 return 1;
305 }
52826846 306
d68ea4d7 307 if (tst->ss->container_content
9008ed1c
N
308 && tst->loaded_container) {
309 /* tmpdev is a container. We need to be either
310 * looking for a member, or auto-assembling
311 */
312 if (st) {
313 /* already found some components, this cannot
314 * be another one.
315 */
316 if (report_missmatch)
317 fprintf(stderr, Name ": %s is a container, but we are looking for components\n",
318 devname);
319 goto loop;
320 }
321
322 if (ident->container) {
323 if (ident->container[0] == '/' &&
324 !same_dev(ident->container, devname)) {
325 if (report_missmatch)
326 fprintf(stderr, Name ": %s is not the container required (%s)\n",
327 devname, ident->container);
328 goto loop;
329 }
330 if (ident->container[0] != '/') {
331 /* we have a uuid */
332 int uuid[4];
333 if (!parse_uuid(ident->container, uuid) ||
334 !same_uuid(content->uuid, uuid, tst->ss->swapuuid)) {
335 if (report_missmatch)
336 fprintf(stderr, Name ": %s has wrong UUID to be required container\n",
337 devname);
338 goto loop;
339 }
340 }
341 }
342 /* It is worth looking inside this container.
343 */
7636b5a8
N
344 if (verbose > 0)
345 fprintf(stderr, Name ": looking in container %s\n",
346 devname);
9008ed1c
N
347 next_member:
348 if (tmpdev->content)
349 content = tmpdev->content;
350 else
00bbdbda 351 content = tst->ss->container_content(tst, NULL);
3ef383aa
DW
352 if (!content)
353 goto loop; /* empty container */
9008ed1c
N
354
355 tmpdev->content = content->next;
356 if (tmpdev->content == NULL)
f7ad3ccc 357 tmpdev->used = 2;
9008ed1c
N
358
359 } else if (ident->container || ident->member) {
360 /* No chance of this matching if we don't have
361 * a container */
362 if (report_missmatch)
363 fprintf(stderr, Name "%s is not a container, and one is required.\n",
364 devname);
365 goto loop;
366 }
367
838acbc2 368 if (ident->uuid_set && (!update || strcmp(update, "uuid")!= 0) &&
d68ea4d7 369 same_uuid(content->uuid, ident->uuid, tst->ss->swapuuid)==0) {
52437b4f 370 if (report_missmatch)
52826846
NB
371 fprintf(stderr, Name ": %s has wrong uuid.\n",
372 devname);
df37ffc0 373 goto loop;
82b27616 374 }
c4f12c13 375 if (ident->name[0] && (!update || strcmp(update, "name")!= 0) &&
d68ea4d7 376 name_matches(content->name, ident->name, homehost)==0) {
52437b4f 377 if (report_missmatch)
947fd4dd
NB
378 fprintf(stderr, Name ": %s has wrong name.\n",
379 devname);
df37ffc0 380 goto loop;
947fd4dd 381 }
98c6faba 382 if (ident->super_minor != UnSet &&
d68ea4d7 383 ident->super_minor != content->array.md_minor) {
52437b4f 384 if (report_missmatch)
52826846 385 fprintf(stderr, Name ": %s has wrong super-minor.\n",
64c4757e 386 devname);
df37ffc0 387 goto loop;
64c4757e 388 }
98c6faba 389 if (ident->level != UnSet &&
d68ea4d7 390 ident->level != content->array.level) {
52437b4f 391 if (report_missmatch)
cd29a5c8
NB
392 fprintf(stderr, Name ": %s has wrong raid level.\n",
393 devname);
df37ffc0 394 goto loop;
cd29a5c8 395 }
98c6faba 396 if (ident->raid_disks != UnSet &&
d68ea4d7 397 ident->raid_disks!= content->array.raid_disks) {
52437b4f 398 if (report_missmatch)
cd29a5c8
NB
399 fprintf(stderr, Name ": %s requires wrong number of drives.\n",
400 devname);
df37ffc0 401 goto loop;
cd29a5c8 402 }
838acbc2 403
d68ea4d7 404 if (tst->ss->container_content
9008ed1c
N
405 && tst->loaded_container) {
406 /* we have the one container we need, don't keep
407 * looking. If the chosen member is active, skip.
408 */
409 if (is_member_busy(content->text_version)) {
f7ad3ccc
N
410 if (report_missmatch)
411 fprintf(stderr, Name ": member %s in %s is already assembled\n",
412 content->text_version,
413 devname);
8a0a0ded
N
414 skip:
415 if (tmpdev->content)
416 goto next_member;
f7ad3ccc
N
417 tst->ss->free_super(tst);
418 tst = NULL;
419 content = NULL;
9008ed1c
N
420 if (auto_assem)
421 goto loop;
4e8d9f0a 422 dev_policy_free(pol);
9008ed1c
N
423 return 1;
424 }
8a0a0ded
N
425 if (ident->member && ident->member[0]) {
426 char *s = strchr(content->text_version+1, '/');
427 if (s == NULL) {
428 fprintf(stderr, Name ": badly formatted version: %s\n",
429 content->text_version);
430 goto skip;
431 }
432 if (strcmp(ident->member, s+1) != 0) {
433 if (report_missmatch)
434 fprintf(stderr,
435 Name ": skipping wrong member %s\n",
436 content->text_version);
437 goto skip;
438 }
439 }
9008ed1c 440 st = tst; tst = NULL;
4c1c3ad8 441 if (!auto_assem && inargv && tmpdev->next != NULL) {
9008ed1c
N
442 fprintf(stderr, Name ": %s is a container, but is not "
443 "only device given: confused and aborting\n",
444 devname);
445 st->ss->free_super(st);
4e8d9f0a 446 dev_policy_free(pol);
9008ed1c
N
447 return 1;
448 }
7636b5a8
N
449 if (verbose > 0)
450 fprintf(stderr, Name ": found match on member %s in %s\n",
451 content->text_version, devname);
9008ed1c
N
452 break;
453 }
83b6208e 454 if (st == NULL)
3da92f27
NB
455 st = dup_super(tst);
456 if (st->minor_version == -1)
457 st->minor_version = tst->minor_version;
83b6208e
NB
458 if (st->ss != tst->ss ||
459 st->minor_version != tst->minor_version ||
64557c33 460 st->ss->compare_super(st, tst) != 0) {
83b6208e 461 /* Some mismatch. If exactly one array matches this host,
da6b5ca9
NB
462 * we can resolve on that one.
463 * Or, if we are auto assembling, we just ignore the second
464 * for now.
83b6208e 465 */
c30e5369 466 if (auto_assem)
df37ffc0 467 goto loop;
83b6208e 468 if (homehost) {
3da92f27
NB
469 int first = st->ss->match_home(st, homehost);
470 int last = tst->ss->match_home(tst, homehost);
9362c1c8
N
471 if (first != last &&
472 (first == 1 || last == 1)) {
83b6208e
NB
473 /* We can do something */
474 if (first) {/* just ignore this one */
52437b4f 475 if (report_missmatch)
83b6208e
NB
476 fprintf(stderr, Name ": %s misses out due to wrong homehost\n",
477 devname);
df37ffc0 478 goto loop;
83b6208e
NB
479 } else { /* reject all those sofar */
480 mddev_dev_t td;
52437b4f 481 if (report_missmatch)
83b6208e
NB
482 fprintf(stderr, Name ": %s overrides previous devices due to good homehost\n",
483 devname);
484 for (td=devlist; td != tmpdev; td=td->next)
485 if (td->used == 1)
486 td->used = 0;
487 tmpdev->used = 1;
df37ffc0 488 goto loop;
83b6208e
NB
489 }
490 }
491 }
52826846
NB
492 fprintf(stderr, Name ": superblock on %s doesn't match others - assembly aborted\n",
493 devname);
3da92f27
NB
494 tst->ss->free_super(tst);
495 st->ss->free_super(st);
4e8d9f0a 496 dev_policy_free(pol);
52826846 497 return 1;
64c4757e
NB
498 }
499
811e6cbe 500 tmpdev->used = 1;
df37ffc0
NB
501
502 loop:
4e8d9f0a
N
503 dev_policy_free(pol);
504 pol = NULL;
9008ed1c
N
505 if (tmpdev->content)
506 goto next_member;
3d2b16e7
NB
507 if (tst)
508 tst->ss->free_super(tst);
811e6cbe
NB
509 }
510
98dbd966 511 if (!st || !st->sb || !content)
c30e5369
N
512 return 2;
513
05833051 514 /* Now need to open the array device. Use create_mddev */
98dbd966 515 if (content == &info)
a5d85af7 516 st->ss->getinfo_super(st, content, NULL);
3d2c4fc7 517
69207ff6 518 trustworthy = FOREIGN;
05833051 519 name = content->name;
745f72f6
N
520 switch (st->ss->match_home(st, homehost)
521 ?: st->ss->match_home(st, "any")) {
69207ff6
N
522 case 1:
523 trustworthy = LOCAL;
98dbd966 524 name = strchr(content->name, ':');
69207ff6
N
525 if (name)
526 name++;
3d2c4fc7 527 else
98dbd966 528 name = content->name;
69207ff6 529 break;
c30e5369 530 }
05833051 531 if (!auto_assem)
aa7c284c 532 /* If the array is listed in mdadm.conf or on
ac2ecf55
N
533 * command line, then we trust the name
534 * even if the array doesn't look local
535 */
536 trustworthy = LOCAL;
537
05833051 538 if (name[0] == 0 &&
98dbd966
DW
539 content->array.level == LEVEL_CONTAINER) {
540 name = content->text_version;
a4bc1720
N
541 trustworthy = METADATA;
542 }
0ac91628
N
543
544 if (name[0] && trustworthy != LOCAL &&
545 ! require_homehost &&
546 conf_name_is_free(name))
547 trustworthy = LOCAL;
548
7cdc0872
N
549 if (trustworthy == LOCAL &&
550 strchr(name, ':'))
551 /* Ignore 'host:' prefix of name */
552 name = strchr(name, ':')+1;
553
a04d5763
N
554 mdfd = create_mddev(mddev, name, ident->autof, trustworthy,
555 chosen_name);
c30e5369
N
556 if (mdfd < 0) {
557 st->ss->free_super(st);
c30e5369 558 if (auto_assem)
60e1bc1a 559 goto try_again;
c30e5369
N
560 return 1;
561 }
a04d5763 562 mddev = chosen_name;
c30e5369
N
563 vers = md_get_version(mdfd);
564 if (vers < 9000) {
565 fprintf(stderr, Name ": Assemble requires driver version 0.90.0 or later.\n"
566 " Upgrade your kernel or try --build\n");
567 close(mdfd);
568 return 1;
569 }
66afdfa9 570 if (mddev_busy(fd2devnum(mdfd))) {
c30e5369
N
571 fprintf(stderr, Name ": %s already active, cannot restart it!\n",
572 mddev);
573 for (tmpdev = devlist ;
574 tmpdev && tmpdev->used != 1;
575 tmpdev = tmpdev->next)
576 ;
577 if (tmpdev && auto_assem)
578 fprintf(stderr, Name ": %s needed for %s...\n",
579 mddev, tmpdev->devname);
580 close(mdfd);
581 mdfd = -3;
582 st->ss->free_super(st);
c30e5369
N
583 if (auto_assem)
584 goto try_again;
585 return 1;
8a46fe84 586 }
c30e5369 587 ioctl(mdfd, STOP_ARRAY, NULL); /* just incase it was started but has no content */
8a46fe84 588
9008ed1c
N
589#ifndef MDASSEMBLE
590 if (content != &info) {
591 /* This is a member of a container. Try starting the array. */
592 return assemble_container_content(st, mdfd, content, runstop,
593 chosen_name, verbose);
594 }
595#endif
811e6cbe 596 /* Ok, no bad inconsistancy, we can try updating etc */
bf4fb153 597 bitmap_done = 0;
6e46bf34 598 content->update_private = NULL;
d7f7ebb7 599 devices = malloc(num_devs * sizeof(*devices));
02e7c5b7 600 devmap = calloc(num_devs * content->array.raid_disks, 1);
da6b5ca9 601 for (tmpdev = devlist; tmpdev; tmpdev=tmpdev->next) if (tmpdev->used == 1) {
811e6cbe
NB
602 char *devname = tmpdev->devname;
603 struct stat stb;
5787fa49 604 /* looks like a good enough match to update the super block if needed */
d1f1011b 605#ifndef MDASSEMBLE
5787fa49 606 if (update) {
811e6cbe 607 int dfd;
4b1ac34b
NB
608 /* prepare useful information in info structures */
609 struct stat stb2;
3da92f27 610 struct supertype *tst;
1e2b2765 611 int err;
4b1ac34b 612 fstat(mdfd, &stb2);
7d99579f
NB
613
614 if (strcmp(update, "uuid")==0 &&
615 !ident->uuid_set) {
616 int rfd;
617 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
618 read(rfd, ident->uuid, 16) != 16) {
619 *(__u32*)(ident->uuid) = random();
620 *(__u32*)(ident->uuid+1) = random();
621 *(__u32*)(ident->uuid+2) = random();
622 *(__u32*)(ident->uuid+3) = random();
623 }
624 if (rfd >= 0) close(rfd);
7d99579f 625 }
811e6cbe
NB
626 dfd = dev_open(devname, O_RDWR|O_EXCL);
627
0430ed48
NB
628 remove_partitions(dfd);
629
3da92f27 630 tst = dup_super(st);
9f22b13f
N
631 if (dfd < 0 || tst->ss->load_super(tst, dfd, NULL) != 0) {
632 fprintf(stderr, Name ": cannot re-read metadata from %s - aborting\n",
633 devname);
634 if (dfd >= 0)
635 close(dfd);
636 close(mdfd);
d7f7ebb7 637 free(devices);
02e7c5b7 638 free(devmap);
9f22b13f
N
639 return 1;
640 }
02e7c5b7 641 tst->ss->getinfo_super(tst, content, devmap + devcnt * content->array.raid_disks);
811e6cbe 642
98dbd966
DW
643 memcpy(content->uuid, ident->uuid, 16);
644 strcpy(content->name, ident->name);
645 content->array.md_minor = minor(stb2.st_rdev);
811e6cbe 646
1e2b2765
N
647 if (strcmp(update, "byteorder") == 0)
648 err = 0;
649 else
650 err = tst->ss->update_super(tst, content, update,
651 devname, verbose,
652 ident->uuid_set,
653 homehost);
654 if (err < 0) {
655 fprintf(stderr,
656 Name ": --update=%s not understood"
657 " for %s metadata\n",
658 update, tst->ss->name);
659 tst->ss->free_super(tst);
660 free(tst);
661 close(mdfd);
662 close(dfd);
d7f7ebb7 663 free(devices);
02e7c5b7 664 free(devmap);
1e2b2765
N
665 return 1;
666 }
e5eac01f
NB
667 if (strcmp(update, "uuid")==0 &&
668 !ident->uuid_set) {
669 ident->uuid_set = 1;
98dbd966 670 memcpy(ident->uuid, content->uuid, 16);
e5eac01f 671 }
1e2b2765 672 if (tst->ss->store_super(tst, dfd))
5787fa49
NB
673 fprintf(stderr, Name ": Could not re-write superblock on %s.\n",
674 devname);
1e2b2765 675 close(dfd);
8131b493
NB
676
677 if (strcmp(update, "uuid")==0 &&
bf4fb153 678 ident->bitmap_fd >= 0 && !bitmap_done) {
3da92f27 679 if (bitmap_update_uuid(ident->bitmap_fd,
98dbd966 680 content->uuid,
3da92f27 681 tst->ss->swapuuid) != 0)
bf4fb153
NB
682 fprintf(stderr, Name ": Could not update uuid on external bitmap.\n");
683 else
684 bitmap_done = 1;
685 }
3da92f27 686 tst->ss->free_super(tst);
d1f1011b
NB
687 } else
688#endif
689 {
30e1b9a5 690 struct supertype *tst = dup_super(st);
da6b5ca9
NB
691 int dfd;
692 dfd = dev_open(devname, O_RDWR|O_EXCL);
693
0430ed48
NB
694 remove_partitions(dfd);
695
9f22b13f
N
696 if (dfd < 0 || tst->ss->load_super(tst, dfd, NULL) != 0) {
697 fprintf(stderr, Name ": cannot re-read metadata from %s - aborting\n",
698 devname);
699 if (dfd >= 0)
700 close(dfd);
701 close(mdfd);
d7f7ebb7 702 free(devices);
02e7c5b7 703 free(devmap);
9f22b13f
N
704 return 1;
705 }
02e7c5b7 706 tst->ss->getinfo_super(tst, content, devmap + devcnt * content->array.raid_disks);
3da92f27 707 tst->ss->free_super(tst);
da6b5ca9 708 close(dfd);
5787fa49
NB
709 }
710
811e6cbe
NB
711 stat(devname, &stb);
712
dab6685f 713 if (verbose > 0)
cd29a5c8 714 fprintf(stderr, Name ": %s is identified as a member of %s, slot %d.\n",
98dbd966 715 devname, mddev, content->disk.raid_disk);
64c4757e 716 devices[devcnt].devname = devname;
64c4757e 717 devices[devcnt].uptodate = 0;
98dbd966 718 devices[devcnt].i = *content;
213ee40b
NB
719 devices[devcnt].i.disk.major = major(stb.st_rdev);
720 devices[devcnt].i.disk.minor = minor(stb.st_rdev);
64c4757e 721 if (most_recent < devcnt) {
213ee40b
NB
722 if (devices[devcnt].i.events
723 > devices[most_recent].i.events)
64c4757e
NB
724 most_recent = devcnt;
725 }
df0d4ea0 726 if (content->array.level == LEVEL_MULTIPATH)
5787fa49
NB
727 /* with multipath, the raid_disk from the superblock is meaningless */
728 i = devcnt;
729 else
213ee40b 730 i = devices[devcnt].i.disk.raid_disk;
308e1801 731 if (i+1 == 0) {
98dbd966
DW
732 if (nextspare < content->array.raid_disks)
733 nextspare = content->array.raid_disks;
308e1801 734 i = nextspare++;
08110d41 735 } else {
98dbd966 736 if (i >= content->array.raid_disks &&
08110d41
NB
737 i >= nextspare)
738 nextspare = i+1;
308e1801 739 }
98c6faba 740 if (i < 10000) {
56eedc1a 741 if (i >= bestcnt) {
f21e18ca 742 int newbestcnt = i+10;
56eedc1a 743 int *newbest = malloc(sizeof(int)*newbestcnt);
f21e18ca 744 int c;
56eedc1a
NB
745 for (c=0; c < newbestcnt; c++)
746 if (c < bestcnt)
747 newbest[c] = best[c];
748 else
749 newbest[c] = -1;
750 if (best)free(best);
751 best = newbest;
752 bestcnt = newbestcnt;
753 }
6a255b69 754 if (best[i] >=0 &&
213ee40b
NB
755 devices[best[i]].i.events
756 == devices[devcnt].i.events
757 && (devices[best[i]].i.disk.minor
758 != devices[devcnt].i.disk.minor)
b8ac1967 759 && st->ss == &super0
98dbd966 760 && content->array.level != LEVEL_MULTIPATH) {
6a255b69
NB
761 /* two different devices with identical superblock.
762 * Could be a mis-detection caused by overlapping
763 * partitions. fail-safe.
764 */
765 fprintf(stderr, Name ": WARNING %s and %s appear"
766 " to have very similar superblocks.\n"
767 " If they are really different, "
768 "please --zero the superblock on one\n"
d2df86e0
NB
769 " If they are the same or overlap,"
770 " please remove one from %s.\n",
771 devices[best[i]].devname, devname,
772 inargv ? "the list" :
773 "the\n DEVICE list in mdadm.conf"
774 );
7f91af49 775 close(mdfd);
d7f7ebb7 776 free(devices);
02e7c5b7 777 free(devmap);
6a255b69
NB
778 return 1;
779 }
52826846 780 if (best[i] == -1
213ee40b
NB
781 || (devices[best[i]].i.events
782 < devices[devcnt].i.events))
52826846 783 best[i] = devcnt;
56eedc1a 784 }
64c4757e 785 devcnt++;
df37ffc0 786 }
6e46bf34
DW
787 free(content->update_private);
788 content->update_private = NULL;
4b1ac34b 789
64c4757e 790 if (devcnt == 0) {
682c7051 791 fprintf(stderr, Name ": no devices found for %s\n",
64c4757e 792 mddev);
3d2b16e7
NB
793 if (st)
794 st->ss->free_super(st);
7f91af49 795 close(mdfd);
d7f7ebb7 796 free(devices);
02e7c5b7 797 free(devmap);
64c4757e
NB
798 return 1;
799 }
4b1ac34b 800
3d2b16e7
NB
801 if (update && strcmp(update, "byteorder")==0)
802 st->minor_version = 90;
803
a5d85af7 804 st->ss->getinfo_super(st, content, NULL);
98dbd966 805 clean = content->array.state & 1;
4b1ac34b 806
64c4757e
NB
807 /* now we have some devices that might be suitable.
808 * I wonder how many
809 */
98dbd966
DW
810 avail = malloc(content->array.raid_disks);
811 memset(avail, 0, content->array.raid_disks);
64c4757e 812 okcnt = 0;
52826846 813 sparecnt=0;
1ff98339 814 rebuilding_cnt=0;
f21e18ca 815 for (i=0; i< bestcnt; i++) {
64c4757e 816 int j = best[i];
ee04451c
NB
817 int event_margin = 1; /* always allow a difference of '1'
818 * like the kernel does
819 */
64c4757e 820 if (j < 0) continue;
d013a55e
NB
821 /* note: we ignore error flags in multipath arrays
822 * as they don't make sense
823 */
df0d4ea0 824 if (content->array.level != LEVEL_MULTIPATH)
f22385f9 825 if (!(devices[j].i.disk.state & (1<<MD_DISK_ACTIVE))) {
213ee40b
NB
826 if (!(devices[j].i.disk.state
827 & (1<<MD_DISK_FAULTY)))
aa88f531 828 sparecnt++;
d013a55e 829 continue;
aa88f531 830 }
02e7c5b7
N
831 /* If this devices thinks that 'most_recent' has failed, then
832 * we must reject this device.
833 */
834 if (j != most_recent &&
835 content->array.raid_disks > 0 &&
836 devices[most_recent].i.disk.raid_disk >= 0 &&
837 devmap[j * content->array.raid_disks + devices[most_recent].i.disk.raid_disk] == 0) {
838 if (verbose > -1)
839 fprintf(stderr, Name ": ignoring %s as it reports %s as failed\n",
840 devices[j].devname, devices[most_recent].devname);
841 best[i] = -1;
842 continue;
843 }
213ee40b
NB
844 if (devices[j].i.events+event_margin >=
845 devices[most_recent].i.events) {
64c4757e 846 devices[j].uptodate = 1;
1ff98339
N
847 if (i < content->array.raid_disks) {
848 if (devices[j].i.recovery_start == MaxSector) {
849 okcnt++;
850 avail[i]=1;
851 } else
852 rebuilding_cnt++;
265e0f17 853 } else
52826846 854 sparecnt++;
64c4757e
NB
855 }
856 }
02e7c5b7 857 free(devmap);
98dbd966
DW
858 while (force && !enough(content->array.level, content->array.raid_disks,
859 content->array.layout, 1,
265e0f17 860 avail, okcnt)) {
64c4757e
NB
861 /* Choose the newest best drive which is
862 * not up-to-date, update the superblock
863 * and add it.
864 */
52826846 865 int fd;
3da92f27 866 struct supertype *tst;
f21e18ca 867 unsigned long long current_events;
cd29a5c8 868 chosen_drive = -1;
f21e18ca 869 for (i = 0; i < content->array.raid_disks && i < bestcnt; i++) {
52826846
NB
870 int j = best[i];
871 if (j>=0 &&
872 !devices[j].uptodate &&
921d9e16 873 devices[j].i.recovery_start == MaxSector &&
52826846 874 (chosen_drive < 0 ||
213ee40b
NB
875 devices[j].i.events
876 > devices[chosen_drive].i.events))
52826846
NB
877 chosen_drive = j;
878 }
879 if (chosen_drive < 0)
880 break;
213ee40b 881 current_events = devices[chosen_drive].i.events;
c5a6f9a6 882 add_another:
dab6685f
NB
883 if (verbose >= 0)
884 fprintf(stderr, Name ": forcing event count in %s(%d) from %d upto %d\n",
213ee40b
NB
885 devices[chosen_drive].devname,
886 devices[chosen_drive].i.disk.raid_disk,
887 (int)(devices[chosen_drive].i.events),
888 (int)(devices[most_recent].i.events));
8b0dabea 889 fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
52826846
NB
890 if (fd < 0) {
891 fprintf(stderr, Name ": Couldn't open %s for write - not updating\n",
892 devices[chosen_drive].devname);
213ee40b 893 devices[chosen_drive].i.events = 0;
52826846
NB
894 continue;
895 }
3da92f27 896 tst = dup_super(st);
60b435db 897 if (tst->ss->load_super(tst,fd, NULL)) {
52826846
NB
898 close(fd);
899 fprintf(stderr, Name ": RAID superblock disappeared from %s - not updating.\n",
900 devices[chosen_drive].devname);
213ee40b 901 devices[chosen_drive].i.events = 0;
52826846
NB
902 continue;
903 }
98dbd966
DW
904 content->events = devices[most_recent].i.events;
905 tst->ss->update_super(tst, content, "force-one",
67a8c82d
NB
906 devices[chosen_drive].devname, verbose,
907 0, NULL);
4b1ac34b 908
3da92f27 909 if (tst->ss->store_super(tst, fd)) {
52826846
NB
910 close(fd);
911 fprintf(stderr, Name ": Could not re-write superblock on %s\n",
912 devices[chosen_drive].devname);
213ee40b 913 devices[chosen_drive].i.events = 0;
3da92f27 914 tst->ss->free_super(tst);
52826846
NB
915 continue;
916 }
917 close(fd);
213ee40b 918 devices[chosen_drive].i.events = devices[most_recent].i.events;
52826846 919 devices[chosen_drive].uptodate = 1;
265e0f17 920 avail[chosen_drive] = 1;
52826846 921 okcnt++;
3da92f27 922 tst->ss->free_super(tst);
c5a6f9a6
NB
923
924 /* If there are any other drives of the same vintage,
925 * add them in as well. We can't lose and we might gain
926 */
f21e18ca 927 for (i = 0; i < content->array.raid_disks && i < bestcnt ; i++) {
c5a6f9a6
NB
928 int j = best[i];
929 if (j >= 0 &&
930 !devices[j].uptodate &&
213ee40b 931 devices[j].i.events == current_events) {
c5a6f9a6
NB
932 chosen_drive = j;
933 goto add_another;
934 }
935 }
64c4757e 936 }
52826846
NB
937
938 /* Now we want to look at the superblock which the kernel will base things on
939 * and compare the devices that we think are working with the devices that the
940 * superblock thinks are working.
941 * If there are differences and --force is given, then update this chosen
942 * superblock.
943 */
cd29a5c8 944 chosen_drive = -1;
3da92f27 945 st->ss->free_super(st);
56eedc1a 946 for (i=0; chosen_drive < 0 && i<bestcnt; i++) {
52826846
NB
947 int j = best[i];
948 int fd;
4b1ac34b 949
52826846
NB
950 if (j<0)
951 continue;
952 if (!devices[j].uptodate)
953 continue;
954 chosen_drive = j;
8b0dabea 955 if ((fd=dev_open(devices[j].devname, O_RDONLY|O_EXCL))< 0) {
52826846
NB
956 fprintf(stderr, Name ": Cannot open %s: %s\n",
957 devices[j].devname, strerror(errno));
7f91af49 958 close(mdfd);
d7f7ebb7 959 free(devices);
52826846
NB
960 return 1;
961 }
3da92f27 962 if (st->ss->load_super(st,fd, NULL)) {
52826846
NB
963 close(fd);
964 fprintf(stderr, Name ": RAID superblock has disappeared from %s\n",
965 devices[j].devname);
7f91af49 966 close(mdfd);
d7f7ebb7 967 free(devices);
52826846
NB
968 return 1;
969 }
970 close(fd);
971 }
3da92f27 972 if (st->sb == NULL) {
4b1ac34b 973 fprintf(stderr, Name ": No suitable drives found for %s\n", mddev);
7f91af49 974 close(mdfd);
d7f7ebb7 975 free(devices);
4b1ac34b
NB
976 return 1;
977 }
a5d85af7 978 st->ss->getinfo_super(st, content, NULL);
f35f2525 979#ifndef MDASSEMBLE
98dbd966 980 sysfs_init(content, mdfd, 0);
f35f2525 981#endif
56eedc1a 982 for (i=0; i<bestcnt; i++) {
52826846 983 int j = best[i];
98c6faba 984 unsigned int desired_state;
11a3e71d 985
98dbd966 986 if (i < content->array.raid_disks)
11a3e71d
NB
987 desired_state = (1<<MD_DISK_ACTIVE) | (1<<MD_DISK_SYNC);
988 else
989 desired_state = 0;
990
52826846
NB
991 if (j<0)
992 continue;
993 if (!devices[j].uptodate)
994 continue;
4b1ac34b 995
213ee40b 996 devices[j].i.disk.state = desired_state;
4e9a6ff7
N
997 if (!(devices[j].i.array.state & 1))
998 clean = 0;
213ee40b
NB
999
1000 if (st->ss->update_super(st, &devices[j].i, "assemble", NULL,
68c7d6d7 1001 verbose, 0, NULL)) {
52826846 1002 if (force) {
dab6685f
NB
1003 if (verbose >= 0)
1004 fprintf(stderr, Name ": "
1005 "clearing FAULTY flag for device %d in %s for %s\n",
1006 j, mddev, devices[j].devname);
4b1ac34b 1007 change = 1;
52826846 1008 } else {
dab6685f
NB
1009 if (verbose >= -1)
1010 fprintf(stderr, Name ": "
1011 "device %d in %s has wrong state in superblock, but %s seems ok\n",
1012 i, mddev, devices[j].devname);
52826846
NB
1013 }
1014 }
4b1ac34b 1015#if 0
213ee40b 1016 if (!(super.disks[i].i.disk.state & (1 << MD_DISK_FAULTY))) {
52826846
NB
1017 fprintf(stderr, Name ": devices %d of %s is not marked FAULTY in superblock, but cannot be found\n",
1018 i, mddev);
1019 }
4b1ac34b 1020#endif
52826846 1021 }
c5a6f9a6 1022 if (force && !clean &&
98dbd966
DW
1023 !enough(content->array.level, content->array.raid_disks,
1024 content->array.layout, clean,
c5a6f9a6 1025 avail, okcnt)) {
98dbd966 1026 change += st->ss->update_super(st, content, "force-array",
67a8c82d
NB
1027 devices[chosen_drive].devname, verbose,
1028 0, NULL);
583315d9 1029 clean = 1;
d013a55e 1030 }
52826846 1031
4b1ac34b 1032 if (change) {
52826846 1033 int fd;
8b0dabea 1034 fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
52826846 1035 if (fd < 0) {
353632d9 1036 fprintf(stderr, Name ": Could not open %s for write - cannot Assemble array.\n",
52826846 1037 devices[chosen_drive].devname);
7f91af49 1038 close(mdfd);
d7f7ebb7 1039 free(devices);
52826846
NB
1040 return 1;
1041 }
3da92f27 1042 if (st->ss->store_super(st, fd)) {
52826846
NB
1043 close(fd);
1044 fprintf(stderr, Name ": Could not re-write superblock on %s\n",
1045 devices[chosen_drive].devname);
7f91af49 1046 close(mdfd);
d7f7ebb7 1047 free(devices);
52826846
NB
1048 return 1;
1049 }
1050 close(fd);
52826846
NB
1051 }
1052
353632d9
NB
1053 /* If we are in the middle of a reshape we may need to restore saved data
1054 * that was moved aside due to the reshape overwriting live data
1055 * The code of doing this lives in Grow.c
1056 */
39c5a909 1057#ifndef MDASSEMBLE
98dbd966 1058 if (content->reshape_active) {
353632d9
NB
1059 int err = 0;
1060 int *fdlist = malloc(sizeof(int)* bestcnt);
cd77ac4e 1061 if (verbose > 0)
ea0ebe96
N
1062 fprintf(stderr, Name ":%s has an active reshape - checking "
1063 "if critical section needs to be restored\n",
1064 chosen_name);
353632d9
NB
1065 for (i=0; i<bestcnt; i++) {
1066 int j = best[i];
1067 if (j >= 0) {
1068 fdlist[i] = dev_open(devices[j].devname, O_RDWR|O_EXCL);
1069 if (fdlist[i] < 0) {
1070 fprintf(stderr, Name ": Could not open %s for write - cannot Assemble array.\n",
1071 devices[j].devname);
1072 err = 1;
1073 break;
1074 }
1075 } else
1076 fdlist[i] = -1;
1077 }
1078 if (!err)
cd77ac4e 1079 err = Grow_restart(st, content, fdlist, bestcnt, backup_file, verbose > 0);
353632d9
NB
1080 while (i>0) {
1081 i--;
1082 if (fdlist[i]>=0) close(fdlist[i]);
1083 }
1084 if (err) {
1085 fprintf(stderr, Name ": Failed to restore critical section for reshape, sorry.\n");
e9e43ec3
N
1086 if (backup_file == NULL)
1087 fprintf(stderr," Possibly you needed to specify the --backup-file\n");
7f91af49 1088 close(mdfd);
d7f7ebb7 1089 free(devices);
353632d9
NB
1090 return err;
1091 }
1092 }
39c5a909 1093#endif
aa88f531
NB
1094 /* count number of in-sync devices according to the superblock.
1095 * We must have this number to start the array without -s or -R
1096 */
98dbd966 1097 req_cnt = content->array.working_disks;
aa88f531 1098
64c4757e
NB
1099 /* Almost ready to actually *do* something */
1100 if (!old_linux) {
fbf8a0b7 1101 int rv;
a19c88b8 1102
a04d5763
N
1103 /* First, fill in the map, so that udev can find our name
1104 * as soon as we become active.
1105 */
98dbd966
DW
1106 map_update(NULL, fd2devnum(mdfd), content->text_version,
1107 content->uuid, chosen_name);
a04d5763 1108
98dbd966 1109 rv = set_array_info(mdfd, st, content);
fbf8a0b7 1110 if (rv) {
f35f2525 1111 fprintf(stderr, Name ": failed to set array info for %s: %s\n",
64c4757e 1112 mddev, strerror(errno));
24af7a87 1113 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1114 close(mdfd);
d7f7ebb7 1115 free(devices);
64c4757e
NB
1116 return 1;
1117 }
11484a63 1118 if (ident->bitmap_fd >= 0) {
c82f047c
NB
1119 if (ioctl(mdfd, SET_BITMAP_FILE, ident->bitmap_fd) != 0) {
1120 fprintf(stderr, Name ": SET_BITMAP_FILE failed.\n");
24af7a87 1121 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1122 close(mdfd);
d7f7ebb7 1123 free(devices);
c82f047c
NB
1124 return 1;
1125 }
7ef02d01
NB
1126 } else if (ident->bitmap_file) {
1127 /* From config file */
1128 int bmfd = open(ident->bitmap_file, O_RDWR);
1129 if (bmfd < 0) {
1130 fprintf(stderr, Name ": Could not open bitmap file %s\n",
1131 ident->bitmap_file);
24af7a87 1132 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1133 close(mdfd);
d7f7ebb7 1134 free(devices);
7ef02d01
NB
1135 return 1;
1136 }
1137 if (ioctl(mdfd, SET_BITMAP_FILE, bmfd) != 0) {
1138 fprintf(stderr, Name ": Failed to set bitmapfile for %s\n", mddev);
1139 close(bmfd);
24af7a87 1140 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1141 close(mdfd);
d7f7ebb7 1142 free(devices);
7ef02d01
NB
1143 return 1;
1144 }
1145 close(bmfd);
c82f047c 1146 }
7ef02d01 1147
52826846 1148 /* First, add the raid disks, but add the chosen one last */
56eedc1a 1149 for (i=0; i<= bestcnt; i++) {
52826846 1150 int j;
56eedc1a 1151 if (i < bestcnt) {
52826846
NB
1152 j = best[i];
1153 if (j == chosen_drive)
1154 continue;
1155 } else
1156 j = chosen_drive;
1157
56eedc1a 1158 if (j >= 0 /* && devices[j].uptodate */) {
98dbd966 1159 rv = add_disk(mdfd, st, content, &devices[j].i);
7801ac20 1160
a19c88b8 1161 if (rv) {
213ee40b
NB
1162 fprintf(stderr, Name ": failed to add "
1163 "%s to %s: %s\n",
64c4757e
NB
1164 devices[j].devname,
1165 mddev,
1166 strerror(errno));
98dbd966 1167 if (i < content->array.raid_disks
213ee40b 1168 || i == bestcnt)
52826846
NB
1169 okcnt--;
1170 else
1171 sparecnt--;
dab6685f 1172 } else if (verbose > 0)
213ee40b
NB
1173 fprintf(stderr, Name ": added %s "
1174 "to %s as %d\n",
1175 devices[j].devname, mddev,
1176 devices[j].i.disk.raid_disk);
98dbd966 1177 } else if (verbose > 0 && i < content->array.raid_disks)
213ee40b
NB
1178 fprintf(stderr, Name ": no uptodate device for "
1179 "slot %d of %s\n",
64c4757e
NB
1180 i, mddev);
1181 }
aba69144 1182
98dbd966 1183 if (content->array.level == LEVEL_CONTAINER) {
598f0d58
NB
1184 if (verbose >= 0) {
1185 fprintf(stderr, Name ": Container %s has been "
1186 "assembled with %d drive%s",
57ed8c91 1187 mddev, okcnt+sparecnt, okcnt+sparecnt==1?"":"s");
f21e18ca 1188 if (okcnt < (unsigned)content->array.raid_disks)
598f0d58 1189 fprintf(stderr, " (out of %d)",
98dbd966 1190 content->array.raid_disks);
598f0d58
NB
1191 fprintf(stderr, "\n");
1192 }
98dbd966 1193 sysfs_uevent(content, "change");
a7c6e3fb 1194 wait_for(chosen_name, mdfd);
7f91af49 1195 close(mdfd);
d7f7ebb7 1196 free(devices);
598f0d58
NB
1197 return 0;
1198 }
1199
64c4757e 1200 if (runstop == 1 ||
b8a8ccf9 1201 (runstop <= 0 &&
98dbd966
DW
1202 ( enough(content->array.level, content->array.raid_disks,
1203 content->array.layout, clean, avail, okcnt) &&
1ff98339 1204 (okcnt + rebuilding_cnt >= req_cnt || start_partial_ok)
aa88f531 1205 ))) {
e9e43ec3
N
1206 /* This array is good-to-go.
1207 * If a reshape is in progress then we might need to
1208 * continue monitoring it. In that case we start
1209 * it read-only and let the grow code make it writable.
1210 */
1211 int rv;
eb3929a4 1212#ifndef MDASSEMBLE
e9e43ec3
N
1213 if (content->reshape_active &&
1214 content->delta_disks <= 0)
1215 rv = Grow_continue(mdfd, st, content, backup_file);
1216 else
eb3929a4 1217#endif
e9e43ec3
N
1218 rv = ioctl(mdfd, RUN_ARRAY, NULL);
1219 if (rv == 0) {
dab6685f
NB
1220 if (verbose >= 0) {
1221 fprintf(stderr, Name ": %s has been started with %d drive%s",
1222 mddev, okcnt, okcnt==1?"":"s");
f21e18ca 1223 if (okcnt < (unsigned)content->array.raid_disks)
98dbd966 1224 fprintf(stderr, " (out of %d)", content->array.raid_disks);
1ff98339
N
1225 if (rebuilding_cnt)
1226 fprintf(stderr, "%s %d rebuilding", sparecnt?",":" and", rebuilding_cnt);
dab6685f
NB
1227 if (sparecnt)
1228 fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
1229 fprintf(stderr, ".\n");
1230 }
8a659c33
N
1231 if (content->reshape_active &&
1232 content->array.level >= 4 &&
1233 content->array.level <= 6) {
acee8e89
N
1234 /* might need to increase the size
1235 * of the stripe cache - default is 256
1236 */
8a659c33 1237 if (256 < 4 * (content->array.chunk_size/4096)) {
acee8e89
N
1238 struct mdinfo *sra = sysfs_read(mdfd, 0, 0);
1239 if (sra)
1240 sysfs_set_num(sra, NULL,
1241 "stripe_cache_size",
8a659c33 1242 (4 * content->array.chunk_size / 4096) + 1);
acee8e89
N
1243 }
1244 }
7e83544b
N
1245 if (okcnt < (unsigned)content->array.raid_disks) {
1246 /* If any devices did not get added
1247 * because the kernel rejected them based
1248 * on event count, try adding them
1249 * again providing the action policy is
1250 * 're-add' or greater. The bitmap
1251 * might allow them to be included, or
1252 * they will become spares.
1253 */
1254 for (i = 0; i <= bestcnt; i++) {
1255 int j = best[i];
1256 if (j >= 0 && !devices[j].uptodate) {
1257 if (!disk_action_allows(&devices[j].i, st->ss->name, act_re_add))
1258 continue;
1259 rv = add_disk(mdfd, st, content,
1260 &devices[j].i);
1261 if (rv == 0 && verbose >= 0)
1262 fprintf(stderr,
1263 Name ": %s has been re-added.\n",
1264 devices[j].devname);
1265 }
1266 }
1267 }
a7c6e3fb 1268 wait_for(mddev, mdfd);
7f91af49
N
1269 close(mdfd);
1270 if (auto_assem) {
589395d6 1271 int usecs = 1;
589395d6
NB
1272 /* There is a nasty race with 'mdadm --monitor'.
1273 * If it opens this device before we close it,
1274 * it gets an incomplete open on which IO
598f0d58
NB
1275 * doesn't work and the capacity is
1276 * wrong.
589395d6
NB
1277 * If we reopen (to check for layered devices)
1278 * before --monitor closes, we loose.
1279 *
1280 * So: wait upto 1 second for there to be
1281 * a non-zero capacity.
1282 */
1283 while (usecs < 1000) {
1284 mdfd = open(mddev, O_RDONLY);
1285 if (mdfd >= 0) {
beae1dfe
NB
1286 unsigned long long size;
1287 if (get_dev_size(mdfd, NULL, &size) &&
589395d6
NB
1288 size > 0)
1289 break;
1290 close(mdfd);
1291 }
1292 usleep(usecs);
1293 usecs <<= 1;
1294 }
1295 }
d7f7ebb7 1296 free(devices);
64c4757e 1297 return 0;
82b27616 1298 }
682c7051 1299 fprintf(stderr, Name ": failed to RUN_ARRAY %s: %s\n",
64c4757e 1300 mddev, strerror(errno));
583315d9 1301
98dbd966
DW
1302 if (!enough(content->array.level, content->array.raid_disks,
1303 content->array.layout, 1, avail, okcnt))
583315d9
NB
1304 fprintf(stderr, Name ": Not enough devices to "
1305 "start the array.\n");
98dbd966
DW
1306 else if (!enough(content->array.level,
1307 content->array.raid_disks,
1308 content->array.layout, clean,
583315d9
NB
1309 avail, okcnt))
1310 fprintf(stderr, Name ": Not enough devices to "
1311 "start the array while not clean "
1312 "- consider --force.\n");
1313
7f91af49 1314 if (auto_assem)
1c203a4b 1315 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1316 close(mdfd);
d7f7ebb7 1317 free(devices);
64c4757e
NB
1318 return 1;
1319 }
82b27616 1320 if (runstop == -1) {
b8a8ccf9 1321 fprintf(stderr, Name ": %s assembled from %d drive%s",
52826846 1322 mddev, okcnt, okcnt==1?"":"s");
f21e18ca 1323 if (okcnt != (unsigned)content->array.raid_disks)
98dbd966 1324 fprintf(stderr, " (out of %d)", content->array.raid_disks);
b8a8ccf9 1325 fprintf(stderr, ", but not started.\n");
7f91af49 1326 close(mdfd);
d7f7ebb7 1327 free(devices);
64c4757e 1328 return 0;
82b27616 1329 }
4855f95c 1330 if (verbose >= -1) {
dab6685f 1331 fprintf(stderr, Name ": %s assembled from %d drive%s", mddev, okcnt, okcnt==1?"":"s");
1ff98339
N
1332 if (rebuilding_cnt)
1333 fprintf(stderr, "%s %d rebuilding", sparecnt?", ":" and ", rebuilding_cnt);
dab6685f
NB
1334 if (sparecnt)
1335 fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
98dbd966
DW
1336 if (!enough(content->array.level, content->array.raid_disks,
1337 content->array.layout, 1, avail, okcnt))
dab6685f 1338 fprintf(stderr, " - not enough to start the array.\n");
98dbd966
DW
1339 else if (!enough(content->array.level,
1340 content->array.raid_disks,
1341 content->array.layout, clean,
583315d9
NB
1342 avail, okcnt))
1343 fprintf(stderr, " - not enough to start the "
1344 "array while not clean - consider "
1345 "--force.\n");
dab6685f 1346 else {
f21e18ca 1347 if (req_cnt == (unsigned)content->array.raid_disks)
dab6685f
NB
1348 fprintf(stderr, " - need all %d to start it", req_cnt);
1349 else
98dbd966 1350 fprintf(stderr, " - need %d of %d to start", req_cnt, content->array.raid_disks);
dab6685f
NB
1351 fprintf(stderr, " (use --run to insist).\n");
1352 }
aa88f531 1353 }
7f91af49 1354 if (auto_assem)
1c203a4b 1355 ioctl(mdfd, STOP_ARRAY, NULL);
56a8da69 1356 close(mdfd);
d7f7ebb7 1357 free(devices);
64c4757e 1358 return 1;
82b27616 1359 } else {
52826846
NB
1360 /* The "chosen_drive" is a good choice, and if necessary, the superblock has
1361 * been updated to point to the current locations of devices.
1362 * so we can just start the array
82b27616 1363 */
cd29a5c8 1364 unsigned long dev;
213ee40b
NB
1365 dev = makedev(devices[chosen_drive].i.disk.major,
1366 devices[chosen_drive].i.disk.minor);
82b27616
NB
1367 if (ioctl(mdfd, START_ARRAY, dev)) {
1368 fprintf(stderr, Name ": Cannot start array: %s\n",
1369 strerror(errno));
1370 }
aba69144 1371
64c4757e 1372 }
7f91af49 1373 close(mdfd);
d7f7ebb7 1374 free(devices);
e0d19036 1375 return 0;
64c4757e 1376}
6234c63c
DW
1377
1378#ifndef MDASSEMBLE
1379int assemble_container_content(struct supertype *st, int mdfd,
1380 struct mdinfo *content, int runstop,
1381 char *chosen_name, int verbose)
1382{
1383 struct mdinfo *dev, *sra;
1384 int working = 0, preexist = 0;
1385 struct map_ent *map = NULL;
1386
1387 sysfs_init(content, mdfd, 0);
1388
1389 sra = sysfs_read(mdfd, 0, GET_VERSION);
1390 if (sra == NULL || strcmp(sra->text_version, content->text_version) != 0)
4408ee76
N
1391 if (sysfs_set_array(content, md_get_version(mdfd)) != 0) {
1392 close(mdfd);
6234c63c 1393 return 1;
4408ee76 1394 }
6234c63c
DW
1395 if (sra)
1396 sysfs_free(sra);
1397
1398 for (dev = content->devs; dev; dev = dev->next)
462906cd 1399 if (sysfs_add_disk(content, dev, 1) == 0)
6234c63c
DW
1400 working++;
1401 else if (errno == EEXIST)
1402 preexist++;
4408ee76
N
1403 if (working == 0) {
1404 close(mdfd);
7cb2aa33 1405 return 1;/* Nothing new, don't try to start */
8b4e5ea9
N
1406 }
1407
1408 map_update(&map, fd2devnum(mdfd),
1409 content->text_version,
1410 content->uuid, chosen_name);
1411
1412 if (runstop > 0 ||
6234c63c 1413 (working + preexist) >= content->array.working_disks) {
bb50e5d3 1414 int err;
a714580e 1415
6234c63c
DW
1416 switch(content->array.level) {
1417 case LEVEL_LINEAR:
1418 case LEVEL_MULTIPATH:
1419 case 0:
bb50e5d3
N
1420 err = sysfs_set_str(content, NULL, "array_state",
1421 "active");
6234c63c
DW
1422 break;
1423 default:
bb50e5d3 1424 err = sysfs_set_str(content, NULL, "array_state",
6234c63c
DW
1425 "readonly");
1426 /* start mdmon if needed. */
bb50e5d3
N
1427 if (!err) {
1428 if (!mdmon_running(st->container_dev))
1429 start_mdmon(st->container_dev);
1430 ping_monitor(devnum2devname(st->container_dev));
1431 }
6234c63c
DW
1432 break;
1433 }
bb50e5d3
N
1434 if (!err)
1435 sysfs_set_safemode(content, content->safe_mode_delay);
6234c63c 1436 if (verbose >= 0) {
bb50e5d3
N
1437 if (err)
1438 fprintf(stderr, Name
1439 ": array %s now has %d devices",
1440 chosen_name, working + preexist);
1441 else
1442 fprintf(stderr, Name
1443 ": Started %s with %d devices",
1444 chosen_name, working + preexist);
6234c63c
DW
1445 if (preexist)
1446 fprintf(stderr, " (%d new)", working);
1447 fprintf(stderr, "\n");
1448 }
bb50e5d3 1449 if (!err)
a7c6e3fb 1450 wait_for(chosen_name, mdfd);
4408ee76 1451 close(mdfd);
7cb2aa33 1452 return 0;
6234c63c 1453 /* FIXME should have an O_EXCL and wait for read-auto */
7cb2aa33 1454 } else {
6234c63c
DW
1455 if (verbose >= 0)
1456 fprintf(stderr, Name
1457 ": %s assembled with %d devices but "
1458 "not started\n",
1459 chosen_name, working);
4408ee76 1460 close(mdfd);
7cb2aa33
N
1461 return 1;
1462 }
6234c63c
DW
1463}
1464#endif
1465