]> git.ipfire.org Git - thirdparty/mdadm.git/blame - Assemble.c
Use load_container in Incremental assembly.
[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);
268 } else if (tst->ss->compare_super == NULL) {
269 if (report_missmatch)
270 fprintf(stderr, Name ": Cannot assemble %s metadata on %s\n",
271 tst->ss->name, devname);
272 tst->ss->free_super(tst);
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);
e5c99c08 281 tst->ss->free_super(tst);
31015d57 282 tmpdev->used = 2;
52826846 283 } else {
98dbd966
DW
284 content = &info;
285 memset(content, 0, sizeof(*content));
a5d85af7 286 tst->ss->getinfo_super(tst, content, NULL);
64c4757e 287 }
c06487ce 288 if (dfd >= 0) close(dfd);
52826846 289
9008ed1c
N
290 if (tst && tst->sb && tst->ss->container_content
291 && tst->loaded_container) {
292 /* tmpdev is a container. We need to be either
293 * looking for a member, or auto-assembling
294 */
295 if (st) {
296 /* already found some components, this cannot
297 * be another one.
298 */
299 if (report_missmatch)
300 fprintf(stderr, Name ": %s is a container, but we are looking for components\n",
301 devname);
302 goto loop;
303 }
304
305 if (ident->container) {
306 if (ident->container[0] == '/' &&
307 !same_dev(ident->container, devname)) {
308 if (report_missmatch)
309 fprintf(stderr, Name ": %s is not the container required (%s)\n",
310 devname, ident->container);
311 goto loop;
312 }
313 if (ident->container[0] != '/') {
314 /* we have a uuid */
315 int uuid[4];
316 if (!parse_uuid(ident->container, uuid) ||
317 !same_uuid(content->uuid, uuid, tst->ss->swapuuid)) {
318 if (report_missmatch)
319 fprintf(stderr, Name ": %s has wrong UUID to be required container\n",
320 devname);
321 goto loop;
322 }
323 }
324 }
325 /* It is worth looking inside this container.
326 */
7636b5a8
N
327 if (verbose > 0)
328 fprintf(stderr, Name ": looking in container %s\n",
329 devname);
9008ed1c
N
330 next_member:
331 if (tmpdev->content)
332 content = tmpdev->content;
333 else
00bbdbda 334 content = tst->ss->container_content(tst, NULL);
3ef383aa
DW
335 if (!content)
336 goto loop; /* empty container */
9008ed1c
N
337
338 tmpdev->content = content->next;
339 if (tmpdev->content == NULL)
f7ad3ccc 340 tmpdev->used = 2;
9008ed1c
N
341
342 } else if (ident->container || ident->member) {
343 /* No chance of this matching if we don't have
344 * a container */
345 if (report_missmatch)
346 fprintf(stderr, Name "%s is not a container, and one is required.\n",
347 devname);
348 goto loop;
349 }
350
838acbc2 351 if (ident->uuid_set && (!update || strcmp(update, "uuid")!= 0) &&
3da92f27 352 (!tst || !tst->sb ||
98dbd966 353 same_uuid(content->uuid, ident->uuid, tst->ss->swapuuid)==0)) {
52437b4f 354 if (report_missmatch)
52826846
NB
355 fprintf(stderr, Name ": %s has wrong uuid.\n",
356 devname);
df37ffc0 357 goto loop;
82b27616 358 }
c4f12c13 359 if (ident->name[0] && (!update || strcmp(update, "name")!= 0) &&
3da92f27 360 (!tst || !tst->sb ||
98dbd966 361 name_matches(content->name, ident->name, homehost)==0)) {
52437b4f 362 if (report_missmatch)
947fd4dd
NB
363 fprintf(stderr, Name ": %s has wrong name.\n",
364 devname);
df37ffc0 365 goto loop;
947fd4dd 366 }
98c6faba 367 if (ident->super_minor != UnSet &&
3da92f27 368 (!tst || !tst->sb ||
98dbd966 369 ident->super_minor != content->array.md_minor)) {
52437b4f 370 if (report_missmatch)
52826846 371 fprintf(stderr, Name ": %s has wrong super-minor.\n",
64c4757e 372 devname);
df37ffc0 373 goto loop;
64c4757e 374 }
98c6faba 375 if (ident->level != UnSet &&
3da92f27 376 (!tst || !tst->sb ||
98dbd966 377 ident->level != content->array.level)) {
52437b4f 378 if (report_missmatch)
cd29a5c8
NB
379 fprintf(stderr, Name ": %s has wrong raid level.\n",
380 devname);
df37ffc0 381 goto loop;
cd29a5c8 382 }
98c6faba 383 if (ident->raid_disks != UnSet &&
3da92f27 384 (!tst || !tst->sb ||
98dbd966 385 ident->raid_disks!= content->array.raid_disks)) {
52437b4f 386 if (report_missmatch)
cd29a5c8
NB
387 fprintf(stderr, Name ": %s requires wrong number of drives.\n",
388 devname);
df37ffc0 389 goto loop;
cd29a5c8 390 }
c30e5369 391 if (auto_assem) {
3da92f27 392 if (tst == NULL || tst->sb == NULL)
da6b5ca9 393 continue;
da6b5ca9 394 }
83b6208e 395 /* If we are this far, then we are nearly commited to this device.
52826846 396 * If the super_block doesn't exist, or doesn't match others,
83b6208e
NB
397 * then we probably cannot continue
398 * However if one of the arrays is for the homehost, and
399 * the other isn't that can disambiguate.
52826846 400 */
52826846 401
3da92f27 402 if (!tst || !tst->sb) {
52826846
NB
403 fprintf(stderr, Name ": %s has no superblock - assembly aborted\n",
404 devname);
ddd1a492
NB
405 if (st)
406 st->ss->free_super(st);
4e8d9f0a 407 dev_policy_free(pol);
52826846
NB
408 return 1;
409 }
838acbc2 410
9008ed1c
N
411 if (tst && tst->sb && tst->ss->container_content
412 && tst->loaded_container) {
413 /* we have the one container we need, don't keep
414 * looking. If the chosen member is active, skip.
415 */
416 if (is_member_busy(content->text_version)) {
f7ad3ccc
N
417 if (report_missmatch)
418 fprintf(stderr, Name ": member %s in %s is already assembled\n",
419 content->text_version,
420 devname);
8a0a0ded
N
421 skip:
422 if (tmpdev->content)
423 goto next_member;
f7ad3ccc
N
424 tst->ss->free_super(tst);
425 tst = NULL;
426 content = NULL;
9008ed1c
N
427 if (auto_assem)
428 goto loop;
4e8d9f0a 429 dev_policy_free(pol);
9008ed1c
N
430 return 1;
431 }
8a0a0ded
N
432 if (ident->member && ident->member[0]) {
433 char *s = strchr(content->text_version+1, '/');
434 if (s == NULL) {
435 fprintf(stderr, Name ": badly formatted version: %s\n",
436 content->text_version);
437 goto skip;
438 }
439 if (strcmp(ident->member, s+1) != 0) {
440 if (report_missmatch)
441 fprintf(stderr,
442 Name ": skipping wrong member %s\n",
443 content->text_version);
444 goto skip;
445 }
446 }
9008ed1c 447 st = tst; tst = NULL;
4c1c3ad8 448 if (!auto_assem && inargv && tmpdev->next != NULL) {
9008ed1c
N
449 fprintf(stderr, Name ": %s is a container, but is not "
450 "only device given: confused and aborting\n",
451 devname);
452 st->ss->free_super(st);
4e8d9f0a 453 dev_policy_free(pol);
9008ed1c
N
454 return 1;
455 }
7636b5a8
N
456 if (verbose > 0)
457 fprintf(stderr, Name ": found match on member %s in %s\n",
458 content->text_version, devname);
9008ed1c
N
459 break;
460 }
83b6208e 461 if (st == NULL)
3da92f27
NB
462 st = dup_super(tst);
463 if (st->minor_version == -1)
464 st->minor_version = tst->minor_version;
83b6208e
NB
465 if (st->ss != tst->ss ||
466 st->minor_version != tst->minor_version ||
64557c33 467 st->ss->compare_super(st, tst) != 0) {
83b6208e 468 /* Some mismatch. If exactly one array matches this host,
da6b5ca9
NB
469 * we can resolve on that one.
470 * Or, if we are auto assembling, we just ignore the second
471 * for now.
83b6208e 472 */
c30e5369 473 if (auto_assem)
df37ffc0 474 goto loop;
83b6208e 475 if (homehost) {
3da92f27
NB
476 int first = st->ss->match_home(st, homehost);
477 int last = tst->ss->match_home(tst, homehost);
9362c1c8
N
478 if (first != last &&
479 (first == 1 || last == 1)) {
83b6208e
NB
480 /* We can do something */
481 if (first) {/* just ignore this one */
52437b4f 482 if (report_missmatch)
83b6208e
NB
483 fprintf(stderr, Name ": %s misses out due to wrong homehost\n",
484 devname);
df37ffc0 485 goto loop;
83b6208e
NB
486 } else { /* reject all those sofar */
487 mddev_dev_t td;
52437b4f 488 if (report_missmatch)
83b6208e
NB
489 fprintf(stderr, Name ": %s overrides previous devices due to good homehost\n",
490 devname);
491 for (td=devlist; td != tmpdev; td=td->next)
492 if (td->used == 1)
493 td->used = 0;
494 tmpdev->used = 1;
df37ffc0 495 goto loop;
83b6208e
NB
496 }
497 }
498 }
52826846
NB
499 fprintf(stderr, Name ": superblock on %s doesn't match others - assembly aborted\n",
500 devname);
3da92f27
NB
501 tst->ss->free_super(tst);
502 st->ss->free_super(st);
4e8d9f0a 503 dev_policy_free(pol);
52826846 504 return 1;
64c4757e
NB
505 }
506
811e6cbe 507 tmpdev->used = 1;
df37ffc0
NB
508
509 loop:
4e8d9f0a
N
510 dev_policy_free(pol);
511 pol = NULL;
9008ed1c
N
512 if (tmpdev->content)
513 goto next_member;
3d2b16e7
NB
514 if (tst)
515 tst->ss->free_super(tst);
811e6cbe
NB
516 }
517
98dbd966 518 if (!st || !st->sb || !content)
c30e5369
N
519 return 2;
520
05833051 521 /* Now need to open the array device. Use create_mddev */
98dbd966 522 if (content == &info)
a5d85af7 523 st->ss->getinfo_super(st, content, NULL);
3d2c4fc7 524
69207ff6 525 trustworthy = FOREIGN;
05833051 526 name = content->name;
745f72f6
N
527 switch (st->ss->match_home(st, homehost)
528 ?: st->ss->match_home(st, "any")) {
69207ff6
N
529 case 1:
530 trustworthy = LOCAL;
98dbd966 531 name = strchr(content->name, ':');
69207ff6
N
532 if (name)
533 name++;
3d2c4fc7 534 else
98dbd966 535 name = content->name;
69207ff6 536 break;
c30e5369 537 }
05833051 538 if (!auto_assem)
aa7c284c 539 /* If the array is listed in mdadm.conf or on
ac2ecf55
N
540 * command line, then we trust the name
541 * even if the array doesn't look local
542 */
543 trustworthy = LOCAL;
544
05833051 545 if (name[0] == 0 &&
98dbd966
DW
546 content->array.level == LEVEL_CONTAINER) {
547 name = content->text_version;
a4bc1720
N
548 trustworthy = METADATA;
549 }
0ac91628
N
550
551 if (name[0] && trustworthy != LOCAL &&
552 ! require_homehost &&
553 conf_name_is_free(name))
554 trustworthy = LOCAL;
555
7cdc0872
N
556 if (trustworthy == LOCAL &&
557 strchr(name, ':'))
558 /* Ignore 'host:' prefix of name */
559 name = strchr(name, ':')+1;
560
a04d5763
N
561 mdfd = create_mddev(mddev, name, ident->autof, trustworthy,
562 chosen_name);
c30e5369
N
563 if (mdfd < 0) {
564 st->ss->free_super(st);
c30e5369 565 if (auto_assem)
60e1bc1a 566 goto try_again;
c30e5369
N
567 return 1;
568 }
a04d5763 569 mddev = chosen_name;
c30e5369
N
570 vers = md_get_version(mdfd);
571 if (vers < 9000) {
572 fprintf(stderr, Name ": Assemble requires driver version 0.90.0 or later.\n"
573 " Upgrade your kernel or try --build\n");
574 close(mdfd);
575 return 1;
576 }
66afdfa9 577 if (mddev_busy(fd2devnum(mdfd))) {
c30e5369
N
578 fprintf(stderr, Name ": %s already active, cannot restart it!\n",
579 mddev);
580 for (tmpdev = devlist ;
581 tmpdev && tmpdev->used != 1;
582 tmpdev = tmpdev->next)
583 ;
584 if (tmpdev && auto_assem)
585 fprintf(stderr, Name ": %s needed for %s...\n",
586 mddev, tmpdev->devname);
587 close(mdfd);
588 mdfd = -3;
589 st->ss->free_super(st);
c30e5369
N
590 if (auto_assem)
591 goto try_again;
592 return 1;
8a46fe84 593 }
c30e5369 594 ioctl(mdfd, STOP_ARRAY, NULL); /* just incase it was started but has no content */
8a46fe84 595
9008ed1c
N
596#ifndef MDASSEMBLE
597 if (content != &info) {
598 /* This is a member of a container. Try starting the array. */
599 return assemble_container_content(st, mdfd, content, runstop,
600 chosen_name, verbose);
601 }
602#endif
811e6cbe 603 /* Ok, no bad inconsistancy, we can try updating etc */
bf4fb153 604 bitmap_done = 0;
6e46bf34 605 content->update_private = NULL;
d7f7ebb7 606 devices = malloc(num_devs * sizeof(*devices));
02e7c5b7 607 devmap = calloc(num_devs * content->array.raid_disks, 1);
da6b5ca9 608 for (tmpdev = devlist; tmpdev; tmpdev=tmpdev->next) if (tmpdev->used == 1) {
811e6cbe
NB
609 char *devname = tmpdev->devname;
610 struct stat stb;
5787fa49 611 /* looks like a good enough match to update the super block if needed */
d1f1011b 612#ifndef MDASSEMBLE
5787fa49 613 if (update) {
811e6cbe 614 int dfd;
4b1ac34b
NB
615 /* prepare useful information in info structures */
616 struct stat stb2;
3da92f27 617 struct supertype *tst;
1e2b2765 618 int err;
4b1ac34b 619 fstat(mdfd, &stb2);
7d99579f
NB
620
621 if (strcmp(update, "uuid")==0 &&
622 !ident->uuid_set) {
623 int rfd;
624 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
625 read(rfd, ident->uuid, 16) != 16) {
626 *(__u32*)(ident->uuid) = random();
627 *(__u32*)(ident->uuid+1) = random();
628 *(__u32*)(ident->uuid+2) = random();
629 *(__u32*)(ident->uuid+3) = random();
630 }
631 if (rfd >= 0) close(rfd);
7d99579f 632 }
811e6cbe
NB
633 dfd = dev_open(devname, O_RDWR|O_EXCL);
634
0430ed48
NB
635 remove_partitions(dfd);
636
3da92f27 637 tst = dup_super(st);
9f22b13f
N
638 if (dfd < 0 || tst->ss->load_super(tst, dfd, NULL) != 0) {
639 fprintf(stderr, Name ": cannot re-read metadata from %s - aborting\n",
640 devname);
641 if (dfd >= 0)
642 close(dfd);
643 close(mdfd);
d7f7ebb7 644 free(devices);
02e7c5b7 645 free(devmap);
9f22b13f
N
646 return 1;
647 }
02e7c5b7 648 tst->ss->getinfo_super(tst, content, devmap + devcnt * content->array.raid_disks);
811e6cbe 649
98dbd966
DW
650 memcpy(content->uuid, ident->uuid, 16);
651 strcpy(content->name, ident->name);
652 content->array.md_minor = minor(stb2.st_rdev);
811e6cbe 653
1e2b2765
N
654 if (strcmp(update, "byteorder") == 0)
655 err = 0;
656 else
657 err = tst->ss->update_super(tst, content, update,
658 devname, verbose,
659 ident->uuid_set,
660 homehost);
661 if (err < 0) {
662 fprintf(stderr,
663 Name ": --update=%s not understood"
664 " for %s metadata\n",
665 update, tst->ss->name);
666 tst->ss->free_super(tst);
667 free(tst);
668 close(mdfd);
669 close(dfd);
d7f7ebb7 670 free(devices);
02e7c5b7 671 free(devmap);
1e2b2765
N
672 return 1;
673 }
e5eac01f
NB
674 if (strcmp(update, "uuid")==0 &&
675 !ident->uuid_set) {
676 ident->uuid_set = 1;
98dbd966 677 memcpy(ident->uuid, content->uuid, 16);
e5eac01f 678 }
1e2b2765 679 if (tst->ss->store_super(tst, dfd))
5787fa49
NB
680 fprintf(stderr, Name ": Could not re-write superblock on %s.\n",
681 devname);
1e2b2765 682 close(dfd);
8131b493
NB
683
684 if (strcmp(update, "uuid")==0 &&
bf4fb153 685 ident->bitmap_fd >= 0 && !bitmap_done) {
3da92f27 686 if (bitmap_update_uuid(ident->bitmap_fd,
98dbd966 687 content->uuid,
3da92f27 688 tst->ss->swapuuid) != 0)
bf4fb153
NB
689 fprintf(stderr, Name ": Could not update uuid on external bitmap.\n");
690 else
691 bitmap_done = 1;
692 }
3da92f27 693 tst->ss->free_super(tst);
d1f1011b
NB
694 } else
695#endif
696 {
30e1b9a5 697 struct supertype *tst = dup_super(st);
da6b5ca9
NB
698 int dfd;
699 dfd = dev_open(devname, O_RDWR|O_EXCL);
700
0430ed48
NB
701 remove_partitions(dfd);
702
9f22b13f
N
703 if (dfd < 0 || tst->ss->load_super(tst, dfd, NULL) != 0) {
704 fprintf(stderr, Name ": cannot re-read metadata from %s - aborting\n",
705 devname);
706 if (dfd >= 0)
707 close(dfd);
708 close(mdfd);
d7f7ebb7 709 free(devices);
02e7c5b7 710 free(devmap);
9f22b13f
N
711 return 1;
712 }
02e7c5b7 713 tst->ss->getinfo_super(tst, content, devmap + devcnt * content->array.raid_disks);
3da92f27 714 tst->ss->free_super(tst);
da6b5ca9 715 close(dfd);
5787fa49
NB
716 }
717
811e6cbe
NB
718 stat(devname, &stb);
719
dab6685f 720 if (verbose > 0)
cd29a5c8 721 fprintf(stderr, Name ": %s is identified as a member of %s, slot %d.\n",
98dbd966 722 devname, mddev, content->disk.raid_disk);
64c4757e 723 devices[devcnt].devname = devname;
64c4757e 724 devices[devcnt].uptodate = 0;
98dbd966 725 devices[devcnt].i = *content;
213ee40b
NB
726 devices[devcnt].i.disk.major = major(stb.st_rdev);
727 devices[devcnt].i.disk.minor = minor(stb.st_rdev);
64c4757e 728 if (most_recent < devcnt) {
213ee40b
NB
729 if (devices[devcnt].i.events
730 > devices[most_recent].i.events)
64c4757e
NB
731 most_recent = devcnt;
732 }
df0d4ea0 733 if (content->array.level == LEVEL_MULTIPATH)
5787fa49
NB
734 /* with multipath, the raid_disk from the superblock is meaningless */
735 i = devcnt;
736 else
213ee40b 737 i = devices[devcnt].i.disk.raid_disk;
308e1801 738 if (i+1 == 0) {
98dbd966
DW
739 if (nextspare < content->array.raid_disks)
740 nextspare = content->array.raid_disks;
308e1801 741 i = nextspare++;
08110d41 742 } else {
98dbd966 743 if (i >= content->array.raid_disks &&
08110d41
NB
744 i >= nextspare)
745 nextspare = i+1;
308e1801 746 }
98c6faba 747 if (i < 10000) {
56eedc1a 748 if (i >= bestcnt) {
f21e18ca 749 int newbestcnt = i+10;
56eedc1a 750 int *newbest = malloc(sizeof(int)*newbestcnt);
f21e18ca 751 int c;
56eedc1a
NB
752 for (c=0; c < newbestcnt; c++)
753 if (c < bestcnt)
754 newbest[c] = best[c];
755 else
756 newbest[c] = -1;
757 if (best)free(best);
758 best = newbest;
759 bestcnt = newbestcnt;
760 }
6a255b69 761 if (best[i] >=0 &&
213ee40b
NB
762 devices[best[i]].i.events
763 == devices[devcnt].i.events
764 && (devices[best[i]].i.disk.minor
765 != devices[devcnt].i.disk.minor)
b8ac1967 766 && st->ss == &super0
98dbd966 767 && content->array.level != LEVEL_MULTIPATH) {
6a255b69
NB
768 /* two different devices with identical superblock.
769 * Could be a mis-detection caused by overlapping
770 * partitions. fail-safe.
771 */
772 fprintf(stderr, Name ": WARNING %s and %s appear"
773 " to have very similar superblocks.\n"
774 " If they are really different, "
775 "please --zero the superblock on one\n"
d2df86e0
NB
776 " If they are the same or overlap,"
777 " please remove one from %s.\n",
778 devices[best[i]].devname, devname,
779 inargv ? "the list" :
780 "the\n DEVICE list in mdadm.conf"
781 );
7f91af49 782 close(mdfd);
d7f7ebb7 783 free(devices);
02e7c5b7 784 free(devmap);
6a255b69
NB
785 return 1;
786 }
52826846 787 if (best[i] == -1
213ee40b
NB
788 || (devices[best[i]].i.events
789 < devices[devcnt].i.events))
52826846 790 best[i] = devcnt;
56eedc1a 791 }
64c4757e 792 devcnt++;
df37ffc0 793 }
6e46bf34
DW
794 free(content->update_private);
795 content->update_private = NULL;
4b1ac34b 796
64c4757e 797 if (devcnt == 0) {
682c7051 798 fprintf(stderr, Name ": no devices found for %s\n",
64c4757e 799 mddev);
3d2b16e7
NB
800 if (st)
801 st->ss->free_super(st);
7f91af49 802 close(mdfd);
d7f7ebb7 803 free(devices);
02e7c5b7 804 free(devmap);
64c4757e
NB
805 return 1;
806 }
4b1ac34b 807
3d2b16e7
NB
808 if (update && strcmp(update, "byteorder")==0)
809 st->minor_version = 90;
810
a5d85af7 811 st->ss->getinfo_super(st, content, NULL);
98dbd966 812 clean = content->array.state & 1;
4b1ac34b 813
64c4757e
NB
814 /* now we have some devices that might be suitable.
815 * I wonder how many
816 */
98dbd966
DW
817 avail = malloc(content->array.raid_disks);
818 memset(avail, 0, content->array.raid_disks);
64c4757e 819 okcnt = 0;
52826846 820 sparecnt=0;
1ff98339 821 rebuilding_cnt=0;
f21e18ca 822 for (i=0; i< bestcnt; i++) {
64c4757e 823 int j = best[i];
ee04451c
NB
824 int event_margin = 1; /* always allow a difference of '1'
825 * like the kernel does
826 */
64c4757e 827 if (j < 0) continue;
d013a55e
NB
828 /* note: we ignore error flags in multipath arrays
829 * as they don't make sense
830 */
df0d4ea0 831 if (content->array.level != LEVEL_MULTIPATH)
f22385f9 832 if (!(devices[j].i.disk.state & (1<<MD_DISK_ACTIVE))) {
213ee40b
NB
833 if (!(devices[j].i.disk.state
834 & (1<<MD_DISK_FAULTY)))
aa88f531 835 sparecnt++;
d013a55e 836 continue;
aa88f531 837 }
02e7c5b7
N
838 /* If this devices thinks that 'most_recent' has failed, then
839 * we must reject this device.
840 */
841 if (j != most_recent &&
842 content->array.raid_disks > 0 &&
843 devices[most_recent].i.disk.raid_disk >= 0 &&
844 devmap[j * content->array.raid_disks + devices[most_recent].i.disk.raid_disk] == 0) {
845 if (verbose > -1)
846 fprintf(stderr, Name ": ignoring %s as it reports %s as failed\n",
847 devices[j].devname, devices[most_recent].devname);
848 best[i] = -1;
849 continue;
850 }
213ee40b
NB
851 if (devices[j].i.events+event_margin >=
852 devices[most_recent].i.events) {
64c4757e 853 devices[j].uptodate = 1;
1ff98339
N
854 if (i < content->array.raid_disks) {
855 if (devices[j].i.recovery_start == MaxSector) {
856 okcnt++;
857 avail[i]=1;
858 } else
859 rebuilding_cnt++;
265e0f17 860 } else
52826846 861 sparecnt++;
64c4757e
NB
862 }
863 }
02e7c5b7 864 free(devmap);
98dbd966
DW
865 while (force && !enough(content->array.level, content->array.raid_disks,
866 content->array.layout, 1,
265e0f17 867 avail, okcnt)) {
64c4757e
NB
868 /* Choose the newest best drive which is
869 * not up-to-date, update the superblock
870 * and add it.
871 */
52826846 872 int fd;
3da92f27 873 struct supertype *tst;
f21e18ca 874 unsigned long long current_events;
cd29a5c8 875 chosen_drive = -1;
f21e18ca 876 for (i = 0; i < content->array.raid_disks && i < bestcnt; i++) {
52826846
NB
877 int j = best[i];
878 if (j>=0 &&
879 !devices[j].uptodate &&
921d9e16 880 devices[j].i.recovery_start == MaxSector &&
52826846 881 (chosen_drive < 0 ||
213ee40b
NB
882 devices[j].i.events
883 > devices[chosen_drive].i.events))
52826846
NB
884 chosen_drive = j;
885 }
886 if (chosen_drive < 0)
887 break;
213ee40b 888 current_events = devices[chosen_drive].i.events;
c5a6f9a6 889 add_another:
dab6685f
NB
890 if (verbose >= 0)
891 fprintf(stderr, Name ": forcing event count in %s(%d) from %d upto %d\n",
213ee40b
NB
892 devices[chosen_drive].devname,
893 devices[chosen_drive].i.disk.raid_disk,
894 (int)(devices[chosen_drive].i.events),
895 (int)(devices[most_recent].i.events));
8b0dabea 896 fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
52826846
NB
897 if (fd < 0) {
898 fprintf(stderr, Name ": Couldn't open %s for write - not updating\n",
899 devices[chosen_drive].devname);
213ee40b 900 devices[chosen_drive].i.events = 0;
52826846
NB
901 continue;
902 }
3da92f27 903 tst = dup_super(st);
60b435db 904 if (tst->ss->load_super(tst,fd, NULL)) {
52826846
NB
905 close(fd);
906 fprintf(stderr, Name ": RAID superblock disappeared from %s - not updating.\n",
907 devices[chosen_drive].devname);
213ee40b 908 devices[chosen_drive].i.events = 0;
52826846
NB
909 continue;
910 }
98dbd966
DW
911 content->events = devices[most_recent].i.events;
912 tst->ss->update_super(tst, content, "force-one",
67a8c82d
NB
913 devices[chosen_drive].devname, verbose,
914 0, NULL);
4b1ac34b 915
3da92f27 916 if (tst->ss->store_super(tst, fd)) {
52826846
NB
917 close(fd);
918 fprintf(stderr, Name ": Could not re-write superblock on %s\n",
919 devices[chosen_drive].devname);
213ee40b 920 devices[chosen_drive].i.events = 0;
3da92f27 921 tst->ss->free_super(tst);
52826846
NB
922 continue;
923 }
924 close(fd);
213ee40b 925 devices[chosen_drive].i.events = devices[most_recent].i.events;
52826846 926 devices[chosen_drive].uptodate = 1;
265e0f17 927 avail[chosen_drive] = 1;
52826846 928 okcnt++;
3da92f27 929 tst->ss->free_super(tst);
c5a6f9a6
NB
930
931 /* If there are any other drives of the same vintage,
932 * add them in as well. We can't lose and we might gain
933 */
f21e18ca 934 for (i = 0; i < content->array.raid_disks && i < bestcnt ; i++) {
c5a6f9a6
NB
935 int j = best[i];
936 if (j >= 0 &&
937 !devices[j].uptodate &&
213ee40b 938 devices[j].i.events == current_events) {
c5a6f9a6
NB
939 chosen_drive = j;
940 goto add_another;
941 }
942 }
64c4757e 943 }
52826846
NB
944
945 /* Now we want to look at the superblock which the kernel will base things on
946 * and compare the devices that we think are working with the devices that the
947 * superblock thinks are working.
948 * If there are differences and --force is given, then update this chosen
949 * superblock.
950 */
cd29a5c8 951 chosen_drive = -1;
3da92f27 952 st->ss->free_super(st);
56eedc1a 953 for (i=0; chosen_drive < 0 && i<bestcnt; i++) {
52826846
NB
954 int j = best[i];
955 int fd;
4b1ac34b 956
52826846
NB
957 if (j<0)
958 continue;
959 if (!devices[j].uptodate)
960 continue;
961 chosen_drive = j;
8b0dabea 962 if ((fd=dev_open(devices[j].devname, O_RDONLY|O_EXCL))< 0) {
52826846
NB
963 fprintf(stderr, Name ": Cannot open %s: %s\n",
964 devices[j].devname, strerror(errno));
7f91af49 965 close(mdfd);
d7f7ebb7 966 free(devices);
52826846
NB
967 return 1;
968 }
3da92f27 969 if (st->ss->load_super(st,fd, NULL)) {
52826846
NB
970 close(fd);
971 fprintf(stderr, Name ": RAID superblock has disappeared from %s\n",
972 devices[j].devname);
7f91af49 973 close(mdfd);
d7f7ebb7 974 free(devices);
52826846
NB
975 return 1;
976 }
977 close(fd);
978 }
3da92f27 979 if (st->sb == NULL) {
4b1ac34b 980 fprintf(stderr, Name ": No suitable drives found for %s\n", mddev);
7f91af49 981 close(mdfd);
d7f7ebb7 982 free(devices);
4b1ac34b
NB
983 return 1;
984 }
a5d85af7 985 st->ss->getinfo_super(st, content, NULL);
f35f2525 986#ifndef MDASSEMBLE
98dbd966 987 sysfs_init(content, mdfd, 0);
f35f2525 988#endif
56eedc1a 989 for (i=0; i<bestcnt; i++) {
52826846 990 int j = best[i];
98c6faba 991 unsigned int desired_state;
11a3e71d 992
98dbd966 993 if (i < content->array.raid_disks)
11a3e71d
NB
994 desired_state = (1<<MD_DISK_ACTIVE) | (1<<MD_DISK_SYNC);
995 else
996 desired_state = 0;
997
52826846
NB
998 if (j<0)
999 continue;
1000 if (!devices[j].uptodate)
1001 continue;
4b1ac34b 1002
213ee40b 1003 devices[j].i.disk.state = desired_state;
4e9a6ff7
N
1004 if (!(devices[j].i.array.state & 1))
1005 clean = 0;
213ee40b
NB
1006
1007 if (st->ss->update_super(st, &devices[j].i, "assemble", NULL,
68c7d6d7 1008 verbose, 0, NULL)) {
52826846 1009 if (force) {
dab6685f
NB
1010 if (verbose >= 0)
1011 fprintf(stderr, Name ": "
1012 "clearing FAULTY flag for device %d in %s for %s\n",
1013 j, mddev, devices[j].devname);
4b1ac34b 1014 change = 1;
52826846 1015 } else {
dab6685f
NB
1016 if (verbose >= -1)
1017 fprintf(stderr, Name ": "
1018 "device %d in %s has wrong state in superblock, but %s seems ok\n",
1019 i, mddev, devices[j].devname);
52826846
NB
1020 }
1021 }
4b1ac34b 1022#if 0
213ee40b 1023 if (!(super.disks[i].i.disk.state & (1 << MD_DISK_FAULTY))) {
52826846
NB
1024 fprintf(stderr, Name ": devices %d of %s is not marked FAULTY in superblock, but cannot be found\n",
1025 i, mddev);
1026 }
4b1ac34b 1027#endif
52826846 1028 }
c5a6f9a6 1029 if (force && !clean &&
98dbd966
DW
1030 !enough(content->array.level, content->array.raid_disks,
1031 content->array.layout, clean,
c5a6f9a6 1032 avail, okcnt)) {
98dbd966 1033 change += st->ss->update_super(st, content, "force-array",
67a8c82d
NB
1034 devices[chosen_drive].devname, verbose,
1035 0, NULL);
583315d9 1036 clean = 1;
d013a55e 1037 }
52826846 1038
4b1ac34b 1039 if (change) {
52826846 1040 int fd;
8b0dabea 1041 fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
52826846 1042 if (fd < 0) {
353632d9 1043 fprintf(stderr, Name ": Could not open %s for write - cannot Assemble array.\n",
52826846 1044 devices[chosen_drive].devname);
7f91af49 1045 close(mdfd);
d7f7ebb7 1046 free(devices);
52826846
NB
1047 return 1;
1048 }
3da92f27 1049 if (st->ss->store_super(st, fd)) {
52826846
NB
1050 close(fd);
1051 fprintf(stderr, Name ": Could not re-write superblock on %s\n",
1052 devices[chosen_drive].devname);
7f91af49 1053 close(mdfd);
d7f7ebb7 1054 free(devices);
52826846
NB
1055 return 1;
1056 }
1057 close(fd);
52826846
NB
1058 }
1059
353632d9
NB
1060 /* If we are in the middle of a reshape we may need to restore saved data
1061 * that was moved aside due to the reshape overwriting live data
1062 * The code of doing this lives in Grow.c
1063 */
39c5a909 1064#ifndef MDASSEMBLE
98dbd966 1065 if (content->reshape_active) {
353632d9
NB
1066 int err = 0;
1067 int *fdlist = malloc(sizeof(int)* bestcnt);
cd77ac4e 1068 if (verbose > 0)
ea0ebe96
N
1069 fprintf(stderr, Name ":%s has an active reshape - checking "
1070 "if critical section needs to be restored\n",
1071 chosen_name);
353632d9
NB
1072 for (i=0; i<bestcnt; i++) {
1073 int j = best[i];
1074 if (j >= 0) {
1075 fdlist[i] = dev_open(devices[j].devname, O_RDWR|O_EXCL);
1076 if (fdlist[i] < 0) {
1077 fprintf(stderr, Name ": Could not open %s for write - cannot Assemble array.\n",
1078 devices[j].devname);
1079 err = 1;
1080 break;
1081 }
1082 } else
1083 fdlist[i] = -1;
1084 }
1085 if (!err)
cd77ac4e 1086 err = Grow_restart(st, content, fdlist, bestcnt, backup_file, verbose > 0);
353632d9
NB
1087 while (i>0) {
1088 i--;
1089 if (fdlist[i]>=0) close(fdlist[i]);
1090 }
1091 if (err) {
1092 fprintf(stderr, Name ": Failed to restore critical section for reshape, sorry.\n");
e9e43ec3
N
1093 if (backup_file == NULL)
1094 fprintf(stderr," Possibly you needed to specify the --backup-file\n");
7f91af49 1095 close(mdfd);
d7f7ebb7 1096 free(devices);
353632d9
NB
1097 return err;
1098 }
1099 }
39c5a909 1100#endif
aa88f531
NB
1101 /* count number of in-sync devices according to the superblock.
1102 * We must have this number to start the array without -s or -R
1103 */
98dbd966 1104 req_cnt = content->array.working_disks;
aa88f531 1105
64c4757e
NB
1106 /* Almost ready to actually *do* something */
1107 if (!old_linux) {
fbf8a0b7 1108 int rv;
a19c88b8 1109
a04d5763
N
1110 /* First, fill in the map, so that udev can find our name
1111 * as soon as we become active.
1112 */
98dbd966
DW
1113 map_update(NULL, fd2devnum(mdfd), content->text_version,
1114 content->uuid, chosen_name);
a04d5763 1115
98dbd966 1116 rv = set_array_info(mdfd, st, content);
fbf8a0b7 1117 if (rv) {
f35f2525 1118 fprintf(stderr, Name ": failed to set array info for %s: %s\n",
64c4757e 1119 mddev, strerror(errno));
24af7a87 1120 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1121 close(mdfd);
d7f7ebb7 1122 free(devices);
64c4757e
NB
1123 return 1;
1124 }
11484a63 1125 if (ident->bitmap_fd >= 0) {
c82f047c
NB
1126 if (ioctl(mdfd, SET_BITMAP_FILE, ident->bitmap_fd) != 0) {
1127 fprintf(stderr, Name ": SET_BITMAP_FILE failed.\n");
24af7a87 1128 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1129 close(mdfd);
d7f7ebb7 1130 free(devices);
c82f047c
NB
1131 return 1;
1132 }
7ef02d01
NB
1133 } else if (ident->bitmap_file) {
1134 /* From config file */
1135 int bmfd = open(ident->bitmap_file, O_RDWR);
1136 if (bmfd < 0) {
1137 fprintf(stderr, Name ": Could not open bitmap file %s\n",
1138 ident->bitmap_file);
24af7a87 1139 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1140 close(mdfd);
d7f7ebb7 1141 free(devices);
7ef02d01
NB
1142 return 1;
1143 }
1144 if (ioctl(mdfd, SET_BITMAP_FILE, bmfd) != 0) {
1145 fprintf(stderr, Name ": Failed to set bitmapfile for %s\n", mddev);
1146 close(bmfd);
24af7a87 1147 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1148 close(mdfd);
d7f7ebb7 1149 free(devices);
7ef02d01
NB
1150 return 1;
1151 }
1152 close(bmfd);
c82f047c 1153 }
7ef02d01 1154
52826846 1155 /* First, add the raid disks, but add the chosen one last */
56eedc1a 1156 for (i=0; i<= bestcnt; i++) {
52826846 1157 int j;
56eedc1a 1158 if (i < bestcnt) {
52826846
NB
1159 j = best[i];
1160 if (j == chosen_drive)
1161 continue;
1162 } else
1163 j = chosen_drive;
1164
56eedc1a 1165 if (j >= 0 /* && devices[j].uptodate */) {
98dbd966 1166 rv = add_disk(mdfd, st, content, &devices[j].i);
7801ac20 1167
a19c88b8 1168 if (rv) {
213ee40b
NB
1169 fprintf(stderr, Name ": failed to add "
1170 "%s to %s: %s\n",
64c4757e
NB
1171 devices[j].devname,
1172 mddev,
1173 strerror(errno));
98dbd966 1174 if (i < content->array.raid_disks
213ee40b 1175 || i == bestcnt)
52826846
NB
1176 okcnt--;
1177 else
1178 sparecnt--;
dab6685f 1179 } else if (verbose > 0)
213ee40b
NB
1180 fprintf(stderr, Name ": added %s "
1181 "to %s as %d\n",
1182 devices[j].devname, mddev,
1183 devices[j].i.disk.raid_disk);
98dbd966 1184 } else if (verbose > 0 && i < content->array.raid_disks)
213ee40b
NB
1185 fprintf(stderr, Name ": no uptodate device for "
1186 "slot %d of %s\n",
64c4757e
NB
1187 i, mddev);
1188 }
aba69144 1189
98dbd966 1190 if (content->array.level == LEVEL_CONTAINER) {
598f0d58
NB
1191 if (verbose >= 0) {
1192 fprintf(stderr, Name ": Container %s has been "
1193 "assembled with %d drive%s",
57ed8c91 1194 mddev, okcnt+sparecnt, okcnt+sparecnt==1?"":"s");
f21e18ca 1195 if (okcnt < (unsigned)content->array.raid_disks)
598f0d58 1196 fprintf(stderr, " (out of %d)",
98dbd966 1197 content->array.raid_disks);
598f0d58
NB
1198 fprintf(stderr, "\n");
1199 }
98dbd966 1200 sysfs_uevent(content, "change");
a7c6e3fb 1201 wait_for(chosen_name, mdfd);
7f91af49 1202 close(mdfd);
d7f7ebb7 1203 free(devices);
598f0d58
NB
1204 return 0;
1205 }
1206
64c4757e 1207 if (runstop == 1 ||
b8a8ccf9 1208 (runstop <= 0 &&
98dbd966
DW
1209 ( enough(content->array.level, content->array.raid_disks,
1210 content->array.layout, clean, avail, okcnt) &&
1ff98339 1211 (okcnt + rebuilding_cnt >= req_cnt || start_partial_ok)
aa88f531 1212 ))) {
e9e43ec3
N
1213 /* This array is good-to-go.
1214 * If a reshape is in progress then we might need to
1215 * continue monitoring it. In that case we start
1216 * it read-only and let the grow code make it writable.
1217 */
1218 int rv;
eb3929a4 1219#ifndef MDASSEMBLE
e9e43ec3
N
1220 if (content->reshape_active &&
1221 content->delta_disks <= 0)
1222 rv = Grow_continue(mdfd, st, content, backup_file);
1223 else
eb3929a4 1224#endif
e9e43ec3
N
1225 rv = ioctl(mdfd, RUN_ARRAY, NULL);
1226 if (rv == 0) {
dab6685f
NB
1227 if (verbose >= 0) {
1228 fprintf(stderr, Name ": %s has been started with %d drive%s",
1229 mddev, okcnt, okcnt==1?"":"s");
f21e18ca 1230 if (okcnt < (unsigned)content->array.raid_disks)
98dbd966 1231 fprintf(stderr, " (out of %d)", content->array.raid_disks);
1ff98339
N
1232 if (rebuilding_cnt)
1233 fprintf(stderr, "%s %d rebuilding", sparecnt?",":" and", rebuilding_cnt);
dab6685f
NB
1234 if (sparecnt)
1235 fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
1236 fprintf(stderr, ".\n");
1237 }
8a659c33
N
1238 if (content->reshape_active &&
1239 content->array.level >= 4 &&
1240 content->array.level <= 6) {
acee8e89
N
1241 /* might need to increase the size
1242 * of the stripe cache - default is 256
1243 */
8a659c33 1244 if (256 < 4 * (content->array.chunk_size/4096)) {
acee8e89
N
1245 struct mdinfo *sra = sysfs_read(mdfd, 0, 0);
1246 if (sra)
1247 sysfs_set_num(sra, NULL,
1248 "stripe_cache_size",
8a659c33 1249 (4 * content->array.chunk_size / 4096) + 1);
acee8e89
N
1250 }
1251 }
7e83544b
N
1252 if (okcnt < (unsigned)content->array.raid_disks) {
1253 /* If any devices did not get added
1254 * because the kernel rejected them based
1255 * on event count, try adding them
1256 * again providing the action policy is
1257 * 're-add' or greater. The bitmap
1258 * might allow them to be included, or
1259 * they will become spares.
1260 */
1261 for (i = 0; i <= bestcnt; i++) {
1262 int j = best[i];
1263 if (j >= 0 && !devices[j].uptodate) {
1264 if (!disk_action_allows(&devices[j].i, st->ss->name, act_re_add))
1265 continue;
1266 rv = add_disk(mdfd, st, content,
1267 &devices[j].i);
1268 if (rv == 0 && verbose >= 0)
1269 fprintf(stderr,
1270 Name ": %s has been re-added.\n",
1271 devices[j].devname);
1272 }
1273 }
1274 }
a7c6e3fb 1275 wait_for(mddev, mdfd);
7f91af49
N
1276 close(mdfd);
1277 if (auto_assem) {
589395d6 1278 int usecs = 1;
589395d6
NB
1279 /* There is a nasty race with 'mdadm --monitor'.
1280 * If it opens this device before we close it,
1281 * it gets an incomplete open on which IO
598f0d58
NB
1282 * doesn't work and the capacity is
1283 * wrong.
589395d6
NB
1284 * If we reopen (to check for layered devices)
1285 * before --monitor closes, we loose.
1286 *
1287 * So: wait upto 1 second for there to be
1288 * a non-zero capacity.
1289 */
1290 while (usecs < 1000) {
1291 mdfd = open(mddev, O_RDONLY);
1292 if (mdfd >= 0) {
beae1dfe
NB
1293 unsigned long long size;
1294 if (get_dev_size(mdfd, NULL, &size) &&
589395d6
NB
1295 size > 0)
1296 break;
1297 close(mdfd);
1298 }
1299 usleep(usecs);
1300 usecs <<= 1;
1301 }
1302 }
d7f7ebb7 1303 free(devices);
64c4757e 1304 return 0;
82b27616 1305 }
682c7051 1306 fprintf(stderr, Name ": failed to RUN_ARRAY %s: %s\n",
64c4757e 1307 mddev, strerror(errno));
583315d9 1308
98dbd966
DW
1309 if (!enough(content->array.level, content->array.raid_disks,
1310 content->array.layout, 1, avail, okcnt))
583315d9
NB
1311 fprintf(stderr, Name ": Not enough devices to "
1312 "start the array.\n");
98dbd966
DW
1313 else if (!enough(content->array.level,
1314 content->array.raid_disks,
1315 content->array.layout, clean,
583315d9
NB
1316 avail, okcnt))
1317 fprintf(stderr, Name ": Not enough devices to "
1318 "start the array while not clean "
1319 "- consider --force.\n");
1320
7f91af49 1321 if (auto_assem)
1c203a4b 1322 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1323 close(mdfd);
d7f7ebb7 1324 free(devices);
64c4757e
NB
1325 return 1;
1326 }
82b27616 1327 if (runstop == -1) {
b8a8ccf9 1328 fprintf(stderr, Name ": %s assembled from %d drive%s",
52826846 1329 mddev, okcnt, okcnt==1?"":"s");
f21e18ca 1330 if (okcnt != (unsigned)content->array.raid_disks)
98dbd966 1331 fprintf(stderr, " (out of %d)", content->array.raid_disks);
b8a8ccf9 1332 fprintf(stderr, ", but not started.\n");
7f91af49 1333 close(mdfd);
d7f7ebb7 1334 free(devices);
64c4757e 1335 return 0;
82b27616 1336 }
4855f95c 1337 if (verbose >= -1) {
dab6685f 1338 fprintf(stderr, Name ": %s assembled from %d drive%s", mddev, okcnt, okcnt==1?"":"s");
1ff98339
N
1339 if (rebuilding_cnt)
1340 fprintf(stderr, "%s %d rebuilding", sparecnt?", ":" and ", rebuilding_cnt);
dab6685f
NB
1341 if (sparecnt)
1342 fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
98dbd966
DW
1343 if (!enough(content->array.level, content->array.raid_disks,
1344 content->array.layout, 1, avail, okcnt))
dab6685f 1345 fprintf(stderr, " - not enough to start the array.\n");
98dbd966
DW
1346 else if (!enough(content->array.level,
1347 content->array.raid_disks,
1348 content->array.layout, clean,
583315d9
NB
1349 avail, okcnt))
1350 fprintf(stderr, " - not enough to start the "
1351 "array while not clean - consider "
1352 "--force.\n");
dab6685f 1353 else {
f21e18ca 1354 if (req_cnt == (unsigned)content->array.raid_disks)
dab6685f
NB
1355 fprintf(stderr, " - need all %d to start it", req_cnt);
1356 else
98dbd966 1357 fprintf(stderr, " - need %d of %d to start", req_cnt, content->array.raid_disks);
dab6685f
NB
1358 fprintf(stderr, " (use --run to insist).\n");
1359 }
aa88f531 1360 }
7f91af49 1361 if (auto_assem)
1c203a4b 1362 ioctl(mdfd, STOP_ARRAY, NULL);
56a8da69 1363 close(mdfd);
d7f7ebb7 1364 free(devices);
64c4757e 1365 return 1;
82b27616 1366 } else {
52826846
NB
1367 /* The "chosen_drive" is a good choice, and if necessary, the superblock has
1368 * been updated to point to the current locations of devices.
1369 * so we can just start the array
82b27616 1370 */
cd29a5c8 1371 unsigned long dev;
213ee40b
NB
1372 dev = makedev(devices[chosen_drive].i.disk.major,
1373 devices[chosen_drive].i.disk.minor);
82b27616
NB
1374 if (ioctl(mdfd, START_ARRAY, dev)) {
1375 fprintf(stderr, Name ": Cannot start array: %s\n",
1376 strerror(errno));
1377 }
aba69144 1378
64c4757e 1379 }
7f91af49 1380 close(mdfd);
d7f7ebb7 1381 free(devices);
e0d19036 1382 return 0;
64c4757e 1383}
6234c63c
DW
1384
1385#ifndef MDASSEMBLE
1386int assemble_container_content(struct supertype *st, int mdfd,
1387 struct mdinfo *content, int runstop,
1388 char *chosen_name, int verbose)
1389{
1390 struct mdinfo *dev, *sra;
1391 int working = 0, preexist = 0;
1392 struct map_ent *map = NULL;
1393
1394 sysfs_init(content, mdfd, 0);
1395
1396 sra = sysfs_read(mdfd, 0, GET_VERSION);
1397 if (sra == NULL || strcmp(sra->text_version, content->text_version) != 0)
4408ee76
N
1398 if (sysfs_set_array(content, md_get_version(mdfd)) != 0) {
1399 close(mdfd);
6234c63c 1400 return 1;
4408ee76 1401 }
6234c63c
DW
1402 if (sra)
1403 sysfs_free(sra);
1404
1405 for (dev = content->devs; dev; dev = dev->next)
462906cd 1406 if (sysfs_add_disk(content, dev, 1) == 0)
6234c63c
DW
1407 working++;
1408 else if (errno == EEXIST)
1409 preexist++;
4408ee76
N
1410 if (working == 0) {
1411 close(mdfd);
7cb2aa33 1412 return 1;/* Nothing new, don't try to start */
8b4e5ea9
N
1413 }
1414
1415 map_update(&map, fd2devnum(mdfd),
1416 content->text_version,
1417 content->uuid, chosen_name);
1418
1419 if (runstop > 0 ||
6234c63c 1420 (working + preexist) >= content->array.working_disks) {
bb50e5d3 1421 int err;
a714580e 1422
6234c63c
DW
1423 switch(content->array.level) {
1424 case LEVEL_LINEAR:
1425 case LEVEL_MULTIPATH:
1426 case 0:
bb50e5d3
N
1427 err = sysfs_set_str(content, NULL, "array_state",
1428 "active");
6234c63c
DW
1429 break;
1430 default:
bb50e5d3 1431 err = sysfs_set_str(content, NULL, "array_state",
6234c63c
DW
1432 "readonly");
1433 /* start mdmon if needed. */
bb50e5d3
N
1434 if (!err) {
1435 if (!mdmon_running(st->container_dev))
1436 start_mdmon(st->container_dev);
1437 ping_monitor(devnum2devname(st->container_dev));
1438 }
6234c63c
DW
1439 break;
1440 }
bb50e5d3
N
1441 if (!err)
1442 sysfs_set_safemode(content, content->safe_mode_delay);
6234c63c 1443 if (verbose >= 0) {
bb50e5d3
N
1444 if (err)
1445 fprintf(stderr, Name
1446 ": array %s now has %d devices",
1447 chosen_name, working + preexist);
1448 else
1449 fprintf(stderr, Name
1450 ": Started %s with %d devices",
1451 chosen_name, working + preexist);
6234c63c
DW
1452 if (preexist)
1453 fprintf(stderr, " (%d new)", working);
1454 fprintf(stderr, "\n");
1455 }
bb50e5d3 1456 if (!err)
a7c6e3fb 1457 wait_for(chosen_name, mdfd);
4408ee76 1458 close(mdfd);
7cb2aa33 1459 return 0;
6234c63c 1460 /* FIXME should have an O_EXCL and wait for read-auto */
7cb2aa33 1461 } else {
6234c63c
DW
1462 if (verbose >= 0)
1463 fprintf(stderr, Name
1464 ": %s assembled with %d devices but "
1465 "not started\n",
1466 chosen_name, working);
4408ee76 1467 close(mdfd);
7cb2aa33
N
1468 return 1;
1469 }
6234c63c
DW
1470}
1471#endif
1472