]> git.ipfire.org Git - thirdparty/mdadm.git/blame - Assemble.c
Improve type names for mddev_dev
[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
fa56eddb 73static int ident_matches(struct mddev_ident *ident,
08fb91a3
N
74 struct mdinfo *content,
75 struct supertype *tst,
76 char *homehost,
77 char *update, char *devname)
78{
79
80 if (ident->uuid_set && (!update || strcmp(update, "uuid")!= 0) &&
81 same_uuid(content->uuid, ident->uuid, tst->ss->swapuuid)==0) {
82 if (devname)
83 fprintf(stderr, Name ": %s has wrong uuid.\n",
84 devname);
85 return 0;
86 }
87 if (ident->name[0] && (!update || strcmp(update, "name")!= 0) &&
88 name_matches(content->name, ident->name, homehost)==0) {
89 if (devname)
90 fprintf(stderr, Name ": %s has wrong name.\n",
91 devname);
92 return 0;
93 }
94 if (ident->super_minor != UnSet &&
95 ident->super_minor != content->array.md_minor) {
96 if (devname)
97 fprintf(stderr, Name ": %s has wrong super-minor.\n",
98 devname);
99 return 0;
100 }
101 if (ident->level != UnSet &&
102 ident->level != content->array.level) {
103 if (devname)
104 fprintf(stderr, Name ": %s has wrong raid level.\n",
105 devname);
106 return 0;
107 }
108 if (ident->raid_disks != UnSet &&
109 ident->raid_disks!= content->array.raid_disks) {
110 if (devname)
111 fprintf(stderr, Name ": %s requires wrong number of drives.\n",
112 devname);
113 return 0;
114 }
115 return 1;
116}
117
118
7f91af49 119int Assemble(struct supertype *st, char *mddev,
fa56eddb 120 struct mddev_ident *ident,
a655e550 121 struct mddev_dev *devlist, char *backup_file,
64c4757e 122 int readonly, int runstop,
0ac91628 123 char *update, char *homehost, int require_homehost,
64c4757e
NB
124 int verbose, int force)
125{
126 /*
52826846
NB
127 * The task of Assemble is to find a collection of
128 * devices that should (according to their superblocks)
129 * form an array, and to give this collection to the MD driver.
130 * In Linux-2.4 and later, this involves submitting a
64c4757e
NB
131 * SET_ARRAY_INFO ioctl with no arg - to prepare
132 * the array - and then submit a number of
133 * ADD_NEW_DISK ioctls to add disks into
134 * the array. Finally RUN_ARRAY might
135 * be submitted to start the array.
136 *
137 * Much of the work of Assemble is in finding and/or
138 * checking the disks to make sure they look right.
139 *
4b1ac34b 140 * If mddev is not set, then scan must be set and we
64c4757e
NB
141 * read through the config file for dev+uuid mapping
142 * We recurse, setting mddev, for each device that
143 * - isn't running
4b1ac34b 144 * - has a valid uuid (or any uuid if !uuidset)
64c4757e
NB
145 *
146 * If mddev is set, we try to determine state of md.
147 * check version - must be at least 0.90.0
148 * check kernel version. must be at least 2.4.
149 * If not, we can possibly fall back on START_ARRAY
150 * Try to GET_ARRAY_INFO.
151 * If possible, give up
152 * If not, try to STOP_ARRAY just to make sure
153 *
154 * If !uuidset and scan, look in conf-file for uuid
155 * If not found, give up
b8a8ccf9 156 * If !devlist and scan and uuidset, get list of devs from conf-file
64c4757e
NB
157 *
158 * For each device:
159 * Check superblock - discard if bad
160 * Check uuid (set if we don't have one) - discard if no match
5787fa49 161 * Check superblock similarity if we have a superblock - discard if different
4b1ac34b 162 * Record events, devicenum
64c4757e 163 * This should give us a list of devices for the array
4b1ac34b 164 * We should collect the most recent event number
64c4757e
NB
165 *
166 * Count disks with recent enough event count
167 * While force && !enough disks
168 * Choose newest rejected disks, update event count
169 * mark clean and rewrite superblock
170 * If recent kernel:
171 * SET_ARRAY_INFO
172 * foreach device with recent events : ADD_NEW_DISK
173 * if runstop == 1 || "enough" disks and runstop==0 -> RUN_ARRAY
174 * If old kernel:
175 * Check the device numbers in superblock are right
176 * update superblock if any changes
177 * START_ARRAY
178 *
179 */
c30e5369
N
180 int mdfd;
181 int clean;
f05641cf
N
182 int auto_assem = (mddev == NULL && !ident->uuid_set &&
183 ident->super_minor == UnSet && ident->name[0] == 0
aa7c284c 184 && (ident->container == NULL || ident->member == NULL));
64c4757e 185 int old_linux = 0;
c30e5369 186 int vers = vers; /* Keep gcc quite - it really is initialised */
64c4757e
NB
187 struct {
188 char *devname;
213ee40b
NB
189 int uptodate; /* set once we decide that this device is as
190 * recent as everything else in the array.
191 */
192 struct mdinfo i;
5787fa49 193 } *devices;
02e7c5b7 194 char *devmap;
56eedc1a 195 int *best = NULL; /* indexed by raid_disk */
f21e18ca 196 int bestcnt = 0;
98c6faba 197 int devcnt = 0;
1ff98339 198 unsigned int okcnt, sparecnt, rebuilding_cnt;
98c6faba 199 unsigned int req_cnt;
f21e18ca 200 int i;
64c4757e 201 int most_recent = 0;
cd29a5c8 202 int chosen_drive;
52826846 203 int change = 0;
cd29a5c8 204 int inargv = 0;
52437b4f 205 int report_missmatch;
bf4fb153 206 int bitmap_done;
7f91af49
N
207 int start_partial_ok = (runstop >= 0) &&
208 (force || devlist==NULL || auto_assem);
98c6faba 209 unsigned int num_devs;
a655e550 210 struct mddev_dev *tmpdev;
4b1ac34b 211 struct mdinfo info;
98dbd966 212 struct mdinfo *content = NULL;
265e0f17 213 char *avail;
308e1801 214 int nextspare = 0;
197e3eb6 215 char *name = NULL;
69207ff6 216 int trustworthy;
a04d5763 217 char chosen_name[1024];
8a46fe84 218
682c7051 219 if (get_linux_version() < 2004000)
64c4757e
NB
220 old_linux = 1;
221
64c4757e 222 /*
52826846
NB
223 * If any subdevs are listed, then any that don't
224 * match ident are discarded. Remainder must all match and
225 * become the array.
226 * If no subdevs, then we scan all devices in the config file, but
227 * there must be something in the identity
64c4757e 228 */
64c4757e 229
cd29a5c8 230 if (!devlist &&
52826846 231 ident->uuid_set == 0 &&
f21e18ca 232 (ident->super_minor < 0 || ident->super_minor == UnSet) &&
e0fe762a
N
233 ident->name[0] == 0 &&
234 (ident->container == NULL || ident->member == NULL) &&
52826846
NB
235 ident->devices == NULL) {
236 fprintf(stderr, Name ": No identity information available for %s - cannot assemble.\n",
8a46fe84 237 mddev ? mddev : "further assembly");
52826846 238 return 1;
64c4757e 239 }
71d60c48 240
cd29a5c8 241 if (devlist == NULL)
8aec876d 242 devlist = conf_get_devs();
7f91af49 243 else if (mddev)
da6b5ca9 244 inargv = 1;
64c4757e 245
52437b4f 246 report_missmatch = ((inargv && verbose >= 0) || verbose > 0);
60e1bc1a 247 try_again:
c30e5369
N
248 /* We come back here when doing auto-assembly and attempting some
249 * set of devices failed. Those are now marked as ->used==2 and
250 * we ignore them and try again
251 */
60e1bc1a 252
5787fa49
NB
253 tmpdev = devlist; num_devs = 0;
254 while (tmpdev) {
da6b5ca9
NB
255 if (tmpdev->used)
256 tmpdev->used = 2;
257 else
258 num_devs++;
5787fa49
NB
259 tmpdev = tmpdev->next;
260 }
5787fa49 261
82d9eba6 262 if (!st && ident->st) st = ident->st;
64c4757e 263
dab6685f 264 if (verbose>0)
82b27616 265 fprintf(stderr, Name ": looking for devices for %s\n",
8a46fe84 266 mddev ? mddev : "further assembly");
82b27616 267
811e6cbe
NB
268 /* first walk the list of devices to find a consistent set
269 * that match the criterea, if that is possible.
c30e5369 270 * We flag the ones we like with 'used'.
811e6cbe
NB
271 */
272 for (tmpdev = devlist;
273 tmpdev;
274 tmpdev = tmpdev->next) {
275 char *devname = tmpdev->devname;
64c4757e
NB
276 int dfd;
277 struct stat stb;
3da92f27 278 struct supertype *tst = dup_super(st);
4e8d9f0a 279 struct dev_policy *pol = NULL;
52826846 280
da6b5ca9
NB
281 if (tmpdev->used > 1) continue;
282
52826846 283 if (ident->devices &&
56eedc1a 284 !match_oneof(ident->devices, devname)) {
52437b4f 285 if (report_missmatch)
56eedc1a 286 fprintf(stderr, Name ": %s is not one of %s\n", devname, ident->devices);
52826846 287 continue;
56eedc1a 288 }
4b1ac34b 289
8b0dabea 290 dfd = dev_open(devname, O_RDONLY|O_EXCL);
64c4757e 291 if (dfd < 0) {
52437b4f 292 if (report_missmatch)
682c7051 293 fprintf(stderr, Name ": cannot open device %s: %s\n",
64c4757e 294 devname, strerror(errno));
da6b5ca9 295 tmpdev->used = 2;
52826846
NB
296 } else if (fstat(dfd, &stb)< 0) {
297 /* Impossible! */
298 fprintf(stderr, Name ": fstat failed for %s: %s\n",
299 devname, strerror(errno));
da6b5ca9 300 tmpdev->used = 2;
cd29a5c8
NB
301 } else if ((stb.st_mode & S_IFMT) != S_IFBLK) {
302 fprintf(stderr, Name ": %s is not a block device.\n",
52826846 303 devname);
da6b5ca9 304 tmpdev->used = 2;
82d9eba6 305 } else if (!tst && (tst = guess_super(dfd)) == NULL) {
52437b4f 306 if (report_missmatch)
da6b5ca9
NB
307 fprintf(stderr, Name ": no recogniseable superblock on %s\n",
308 devname);
309 tmpdev->used = 2;
1ff98339
N
310 } else if (tst->ss->load_super(tst,dfd, NULL)) {
311 if (report_missmatch)
0f22b998
N
312 fprintf(stderr, Name ": no RAID superblock on %s\n",
313 devname);
d68ea4d7 314 tmpdev->used = 2;
0f22b998
N
315 } else if (tst->ss->compare_super == NULL) {
316 if (report_missmatch)
317 fprintf(stderr, Name ": Cannot assemble %s metadata on %s\n",
318 tst->ss->name, devname);
0f22b998 319 tmpdev->used = 2;
31015d57 320 } else if (auto_assem && st == NULL &&
4e8d9f0a 321 !conf_test_metadata(tst->ss->name, (pol = devnum_policy(stb.st_rdev)),
d1d3482b 322 tst->ss->match_home(tst, homehost) == 1)) {
31015d57
N
323 if (report_missmatch)
324 fprintf(stderr, Name ": %s has metadata type %s for which "
325 "auto-assembly is disabled\n",
326 devname, tst->ss->name);
327 tmpdev->used = 2;
52826846 328 } else {
98dbd966
DW
329 content = &info;
330 memset(content, 0, sizeof(*content));
a5d85af7 331 tst->ss->getinfo_super(tst, content, NULL);
64c4757e 332 }
c06487ce 333 if (dfd >= 0) close(dfd);
d68ea4d7
N
334 if (tmpdev->used == 2) {
335 if (auto_assem)
336 /* Ignore unrecognised devices during auto-assembly */
337 goto loop;
338 if (ident->uuid_set || ident->name[0] ||
339 ident->super_minor != UnSet)
340 /* Ignore unrecognised device if looking for
341 * specific array */
342 goto loop;
343
344
345 fprintf(stderr, Name ": %s has no superblock - assembly aborted\n",
346 devname);
347 if (st)
348 st->ss->free_super(st);
349 dev_policy_free(pol);
350 return 1;
351 }
52826846 352
d68ea4d7 353 if (tst->ss->container_content
9008ed1c
N
354 && tst->loaded_container) {
355 /* tmpdev is a container. We need to be either
356 * looking for a member, or auto-assembling
357 */
358 if (st) {
359 /* already found some components, this cannot
360 * be another one.
361 */
362 if (report_missmatch)
363 fprintf(stderr, Name ": %s is a container, but we are looking for components\n",
364 devname);
365 goto loop;
366 }
367
368 if (ident->container) {
369 if (ident->container[0] == '/' &&
370 !same_dev(ident->container, devname)) {
371 if (report_missmatch)
372 fprintf(stderr, Name ": %s is not the container required (%s)\n",
373 devname, ident->container);
374 goto loop;
375 }
376 if (ident->container[0] != '/') {
377 /* we have a uuid */
378 int uuid[4];
379 if (!parse_uuid(ident->container, uuid) ||
380 !same_uuid(content->uuid, uuid, tst->ss->swapuuid)) {
381 if (report_missmatch)
382 fprintf(stderr, Name ": %s has wrong UUID to be required container\n",
383 devname);
384 goto loop;
385 }
386 }
387 }
388 /* It is worth looking inside this container.
389 */
7636b5a8
N
390 if (verbose > 0)
391 fprintf(stderr, Name ": looking in container %s\n",
392 devname);
9008ed1c
N
393 next_member:
394 if (tmpdev->content)
395 content = tmpdev->content;
396 else
00bbdbda 397 content = tst->ss->container_content(tst, NULL);
3ef383aa
DW
398 if (!content)
399 goto loop; /* empty container */
9008ed1c
N
400
401 tmpdev->content = content->next;
402 if (tmpdev->content == NULL)
f7ad3ccc 403 tmpdev->used = 2;
9008ed1c
N
404
405 } else if (ident->container || ident->member) {
406 /* No chance of this matching if we don't have
407 * a container */
408 if (report_missmatch)
409 fprintf(stderr, Name "%s is not a container, and one is required.\n",
410 devname);
411 goto loop;
412 }
413
08fb91a3
N
414 if (!ident_matches(ident, content, tst,
415 homehost, update,
416 report_missmatch ? devname : NULL))
df37ffc0 417 goto loop;
838acbc2 418
d68ea4d7 419 if (tst->ss->container_content
9008ed1c
N
420 && tst->loaded_container) {
421 /* we have the one container we need, don't keep
422 * looking. If the chosen member is active, skip.
423 */
424 if (is_member_busy(content->text_version)) {
f7ad3ccc
N
425 if (report_missmatch)
426 fprintf(stderr, Name ": member %s in %s is already assembled\n",
427 content->text_version,
428 devname);
8a0a0ded
N
429 skip:
430 if (tmpdev->content)
431 goto next_member;
f7ad3ccc
N
432 tst->ss->free_super(tst);
433 tst = NULL;
434 content = NULL;
9008ed1c
N
435 if (auto_assem)
436 goto loop;
4e8d9f0a 437 dev_policy_free(pol);
9008ed1c
N
438 return 1;
439 }
8a0a0ded
N
440 if (ident->member && ident->member[0]) {
441 char *s = strchr(content->text_version+1, '/');
442 if (s == NULL) {
443 fprintf(stderr, Name ": badly formatted version: %s\n",
444 content->text_version);
445 goto skip;
446 }
447 if (strcmp(ident->member, s+1) != 0) {
448 if (report_missmatch)
449 fprintf(stderr,
450 Name ": skipping wrong member %s\n",
451 content->text_version);
452 goto skip;
453 }
454 }
9008ed1c 455 st = tst; tst = NULL;
4c1c3ad8 456 if (!auto_assem && inargv && tmpdev->next != NULL) {
9008ed1c
N
457 fprintf(stderr, Name ": %s is a container, but is not "
458 "only device given: confused and aborting\n",
459 devname);
460 st->ss->free_super(st);
4e8d9f0a 461 dev_policy_free(pol);
9008ed1c
N
462 return 1;
463 }
7636b5a8
N
464 if (verbose > 0)
465 fprintf(stderr, Name ": found match on member %s in %s\n",
466 content->text_version, devname);
9008ed1c
N
467 break;
468 }
83b6208e 469 if (st == NULL)
3da92f27
NB
470 st = dup_super(tst);
471 if (st->minor_version == -1)
472 st->minor_version = tst->minor_version;
83b6208e
NB
473 if (st->ss != tst->ss ||
474 st->minor_version != tst->minor_version ||
64557c33 475 st->ss->compare_super(st, tst) != 0) {
83b6208e 476 /* Some mismatch. If exactly one array matches this host,
da6b5ca9
NB
477 * we can resolve on that one.
478 * Or, if we are auto assembling, we just ignore the second
479 * for now.
83b6208e 480 */
c30e5369 481 if (auto_assem)
df37ffc0 482 goto loop;
83b6208e 483 if (homehost) {
3da92f27
NB
484 int first = st->ss->match_home(st, homehost);
485 int last = tst->ss->match_home(tst, homehost);
9362c1c8
N
486 if (first != last &&
487 (first == 1 || last == 1)) {
83b6208e
NB
488 /* We can do something */
489 if (first) {/* just ignore this one */
52437b4f 490 if (report_missmatch)
83b6208e
NB
491 fprintf(stderr, Name ": %s misses out due to wrong homehost\n",
492 devname);
df37ffc0 493 goto loop;
83b6208e 494 } else { /* reject all those sofar */
a655e550 495 struct mddev_dev *td;
52437b4f 496 if (report_missmatch)
83b6208e
NB
497 fprintf(stderr, Name ": %s overrides previous devices due to good homehost\n",
498 devname);
499 for (td=devlist; td != tmpdev; td=td->next)
500 if (td->used == 1)
501 td->used = 0;
502 tmpdev->used = 1;
df37ffc0 503 goto loop;
83b6208e
NB
504 }
505 }
506 }
52826846
NB
507 fprintf(stderr, Name ": superblock on %s doesn't match others - assembly aborted\n",
508 devname);
3da92f27
NB
509 tst->ss->free_super(tst);
510 st->ss->free_super(st);
4e8d9f0a 511 dev_policy_free(pol);
52826846 512 return 1;
64c4757e
NB
513 }
514
811e6cbe 515 tmpdev->used = 1;
df37ffc0
NB
516
517 loop:
4e8d9f0a
N
518 dev_policy_free(pol);
519 pol = NULL;
9008ed1c
N
520 if (tmpdev->content)
521 goto next_member;
3d2b16e7
NB
522 if (tst)
523 tst->ss->free_super(tst);
811e6cbe
NB
524 }
525
98dbd966 526 if (!st || !st->sb || !content)
c30e5369
N
527 return 2;
528
05833051 529 /* Now need to open the array device. Use create_mddev */
98dbd966 530 if (content == &info)
a5d85af7 531 st->ss->getinfo_super(st, content, NULL);
3d2c4fc7 532
69207ff6 533 trustworthy = FOREIGN;
05833051 534 name = content->name;
745f72f6
N
535 switch (st->ss->match_home(st, homehost)
536 ?: st->ss->match_home(st, "any")) {
69207ff6
N
537 case 1:
538 trustworthy = LOCAL;
98dbd966 539 name = strchr(content->name, ':');
69207ff6
N
540 if (name)
541 name++;
3d2c4fc7 542 else
98dbd966 543 name = content->name;
69207ff6 544 break;
c30e5369 545 }
05833051 546 if (!auto_assem)
aa7c284c 547 /* If the array is listed in mdadm.conf or on
ac2ecf55
N
548 * command line, then we trust the name
549 * even if the array doesn't look local
550 */
551 trustworthy = LOCAL;
552
05833051 553 if (name[0] == 0 &&
98dbd966
DW
554 content->array.level == LEVEL_CONTAINER) {
555 name = content->text_version;
a4bc1720
N
556 trustworthy = METADATA;
557 }
0ac91628
N
558
559 if (name[0] && trustworthy != LOCAL &&
560 ! require_homehost &&
561 conf_name_is_free(name))
562 trustworthy = LOCAL;
563
7cdc0872
N
564 if (trustworthy == LOCAL &&
565 strchr(name, ':'))
566 /* Ignore 'host:' prefix of name */
567 name = strchr(name, ':')+1;
568
a04d5763
N
569 mdfd = create_mddev(mddev, name, ident->autof, trustworthy,
570 chosen_name);
c30e5369
N
571 if (mdfd < 0) {
572 st->ss->free_super(st);
c30e5369 573 if (auto_assem)
60e1bc1a 574 goto try_again;
c30e5369
N
575 return 1;
576 }
a04d5763 577 mddev = chosen_name;
c30e5369
N
578 vers = md_get_version(mdfd);
579 if (vers < 9000) {
580 fprintf(stderr, Name ": Assemble requires driver version 0.90.0 or later.\n"
581 " Upgrade your kernel or try --build\n");
582 close(mdfd);
583 return 1;
584 }
66afdfa9 585 if (mddev_busy(fd2devnum(mdfd))) {
c30e5369
N
586 fprintf(stderr, Name ": %s already active, cannot restart it!\n",
587 mddev);
588 for (tmpdev = devlist ;
589 tmpdev && tmpdev->used != 1;
590 tmpdev = tmpdev->next)
591 ;
592 if (tmpdev && auto_assem)
593 fprintf(stderr, Name ": %s needed for %s...\n",
594 mddev, tmpdev->devname);
595 close(mdfd);
596 mdfd = -3;
597 st->ss->free_super(st);
c30e5369
N
598 if (auto_assem)
599 goto try_again;
600 return 1;
8a46fe84 601 }
c30e5369 602 ioctl(mdfd, STOP_ARRAY, NULL); /* just incase it was started but has no content */
8a46fe84 603
9008ed1c
N
604#ifndef MDASSEMBLE
605 if (content != &info) {
606 /* This is a member of a container. Try starting the array. */
607 return assemble_container_content(st, mdfd, content, runstop,
608 chosen_name, verbose);
609 }
610#endif
811e6cbe 611 /* Ok, no bad inconsistancy, we can try updating etc */
bf4fb153 612 bitmap_done = 0;
6e46bf34 613 content->update_private = NULL;
d7f7ebb7 614 devices = malloc(num_devs * sizeof(*devices));
02e7c5b7 615 devmap = calloc(num_devs * content->array.raid_disks, 1);
da6b5ca9 616 for (tmpdev = devlist; tmpdev; tmpdev=tmpdev->next) if (tmpdev->used == 1) {
811e6cbe
NB
617 char *devname = tmpdev->devname;
618 struct stat stb;
5787fa49 619 /* looks like a good enough match to update the super block if needed */
d1f1011b 620#ifndef MDASSEMBLE
5787fa49 621 if (update) {
811e6cbe 622 int dfd;
4b1ac34b
NB
623 /* prepare useful information in info structures */
624 struct stat stb2;
3da92f27 625 struct supertype *tst;
1e2b2765 626 int err;
4b1ac34b 627 fstat(mdfd, &stb2);
7d99579f
NB
628
629 if (strcmp(update, "uuid")==0 &&
630 !ident->uuid_set) {
631 int rfd;
632 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
633 read(rfd, ident->uuid, 16) != 16) {
634 *(__u32*)(ident->uuid) = random();
635 *(__u32*)(ident->uuid+1) = random();
636 *(__u32*)(ident->uuid+2) = random();
637 *(__u32*)(ident->uuid+3) = random();
638 }
639 if (rfd >= 0) close(rfd);
7d99579f 640 }
811e6cbe
NB
641 dfd = dev_open(devname, O_RDWR|O_EXCL);
642
0430ed48
NB
643 remove_partitions(dfd);
644
3da92f27 645 tst = dup_super(st);
9f22b13f
N
646 if (dfd < 0 || tst->ss->load_super(tst, dfd, NULL) != 0) {
647 fprintf(stderr, Name ": cannot re-read metadata from %s - aborting\n",
648 devname);
649 if (dfd >= 0)
650 close(dfd);
651 close(mdfd);
d7f7ebb7 652 free(devices);
02e7c5b7 653 free(devmap);
9f22b13f
N
654 return 1;
655 }
02e7c5b7 656 tst->ss->getinfo_super(tst, content, devmap + devcnt * content->array.raid_disks);
811e6cbe 657
98dbd966
DW
658 memcpy(content->uuid, ident->uuid, 16);
659 strcpy(content->name, ident->name);
660 content->array.md_minor = minor(stb2.st_rdev);
811e6cbe 661
1e2b2765
N
662 if (strcmp(update, "byteorder") == 0)
663 err = 0;
664 else
665 err = tst->ss->update_super(tst, content, update,
666 devname, verbose,
667 ident->uuid_set,
668 homehost);
669 if (err < 0) {
670 fprintf(stderr,
671 Name ": --update=%s not understood"
672 " for %s metadata\n",
673 update, tst->ss->name);
674 tst->ss->free_super(tst);
675 free(tst);
676 close(mdfd);
677 close(dfd);
d7f7ebb7 678 free(devices);
02e7c5b7 679 free(devmap);
1e2b2765
N
680 return 1;
681 }
e5eac01f
NB
682 if (strcmp(update, "uuid")==0 &&
683 !ident->uuid_set) {
684 ident->uuid_set = 1;
98dbd966 685 memcpy(ident->uuid, content->uuid, 16);
e5eac01f 686 }
1e2b2765 687 if (tst->ss->store_super(tst, dfd))
5787fa49
NB
688 fprintf(stderr, Name ": Could not re-write superblock on %s.\n",
689 devname);
1e2b2765 690 close(dfd);
8131b493
NB
691
692 if (strcmp(update, "uuid")==0 &&
bf4fb153 693 ident->bitmap_fd >= 0 && !bitmap_done) {
3da92f27 694 if (bitmap_update_uuid(ident->bitmap_fd,
98dbd966 695 content->uuid,
3da92f27 696 tst->ss->swapuuid) != 0)
bf4fb153
NB
697 fprintf(stderr, Name ": Could not update uuid on external bitmap.\n");
698 else
699 bitmap_done = 1;
700 }
3da92f27 701 tst->ss->free_super(tst);
d1f1011b
NB
702 } else
703#endif
704 {
30e1b9a5 705 struct supertype *tst = dup_super(st);
da6b5ca9
NB
706 int dfd;
707 dfd = dev_open(devname, O_RDWR|O_EXCL);
708
0430ed48
NB
709 remove_partitions(dfd);
710
9f22b13f
N
711 if (dfd < 0 || tst->ss->load_super(tst, dfd, NULL) != 0) {
712 fprintf(stderr, Name ": cannot re-read metadata from %s - aborting\n",
713 devname);
714 if (dfd >= 0)
715 close(dfd);
716 close(mdfd);
d7f7ebb7 717 free(devices);
02e7c5b7 718 free(devmap);
9f22b13f
N
719 return 1;
720 }
02e7c5b7 721 tst->ss->getinfo_super(tst, content, devmap + devcnt * content->array.raid_disks);
3da92f27 722 tst->ss->free_super(tst);
da6b5ca9 723 close(dfd);
5787fa49
NB
724 }
725
811e6cbe
NB
726 stat(devname, &stb);
727
dab6685f 728 if (verbose > 0)
cd29a5c8 729 fprintf(stderr, Name ": %s is identified as a member of %s, slot %d.\n",
98dbd966 730 devname, mddev, content->disk.raid_disk);
64c4757e 731 devices[devcnt].devname = devname;
64c4757e 732 devices[devcnt].uptodate = 0;
98dbd966 733 devices[devcnt].i = *content;
213ee40b
NB
734 devices[devcnt].i.disk.major = major(stb.st_rdev);
735 devices[devcnt].i.disk.minor = minor(stb.st_rdev);
64c4757e 736 if (most_recent < devcnt) {
213ee40b
NB
737 if (devices[devcnt].i.events
738 > devices[most_recent].i.events)
64c4757e
NB
739 most_recent = devcnt;
740 }
df0d4ea0 741 if (content->array.level == LEVEL_MULTIPATH)
5787fa49
NB
742 /* with multipath, the raid_disk from the superblock is meaningless */
743 i = devcnt;
744 else
213ee40b 745 i = devices[devcnt].i.disk.raid_disk;
308e1801 746 if (i+1 == 0) {
98dbd966
DW
747 if (nextspare < content->array.raid_disks)
748 nextspare = content->array.raid_disks;
308e1801 749 i = nextspare++;
08110d41 750 } else {
98dbd966 751 if (i >= content->array.raid_disks &&
08110d41
NB
752 i >= nextspare)
753 nextspare = i+1;
308e1801 754 }
98c6faba 755 if (i < 10000) {
56eedc1a 756 if (i >= bestcnt) {
f21e18ca 757 int newbestcnt = i+10;
56eedc1a 758 int *newbest = malloc(sizeof(int)*newbestcnt);
f21e18ca 759 int c;
56eedc1a
NB
760 for (c=0; c < newbestcnt; c++)
761 if (c < bestcnt)
762 newbest[c] = best[c];
763 else
764 newbest[c] = -1;
765 if (best)free(best);
766 best = newbest;
767 bestcnt = newbestcnt;
768 }
6a255b69 769 if (best[i] >=0 &&
213ee40b
NB
770 devices[best[i]].i.events
771 == devices[devcnt].i.events
772 && (devices[best[i]].i.disk.minor
773 != devices[devcnt].i.disk.minor)
b8ac1967 774 && st->ss == &super0
98dbd966 775 && content->array.level != LEVEL_MULTIPATH) {
6a255b69
NB
776 /* two different devices with identical superblock.
777 * Could be a mis-detection caused by overlapping
778 * partitions. fail-safe.
779 */
780 fprintf(stderr, Name ": WARNING %s and %s appear"
781 " to have very similar superblocks.\n"
782 " If they are really different, "
783 "please --zero the superblock on one\n"
d2df86e0
NB
784 " If they are the same or overlap,"
785 " please remove one from %s.\n",
786 devices[best[i]].devname, devname,
787 inargv ? "the list" :
788 "the\n DEVICE list in mdadm.conf"
789 );
7f91af49 790 close(mdfd);
d7f7ebb7 791 free(devices);
02e7c5b7 792 free(devmap);
6a255b69
NB
793 return 1;
794 }
52826846 795 if (best[i] == -1
213ee40b
NB
796 || (devices[best[i]].i.events
797 < devices[devcnt].i.events))
52826846 798 best[i] = devcnt;
56eedc1a 799 }
64c4757e 800 devcnt++;
df37ffc0 801 }
6e46bf34
DW
802 free(content->update_private);
803 content->update_private = NULL;
4b1ac34b 804
64c4757e 805 if (devcnt == 0) {
682c7051 806 fprintf(stderr, Name ": no devices found for %s\n",
64c4757e 807 mddev);
3d2b16e7
NB
808 if (st)
809 st->ss->free_super(st);
7f91af49 810 close(mdfd);
d7f7ebb7 811 free(devices);
02e7c5b7 812 free(devmap);
64c4757e
NB
813 return 1;
814 }
4b1ac34b 815
3d2b16e7
NB
816 if (update && strcmp(update, "byteorder")==0)
817 st->minor_version = 90;
818
a5d85af7 819 st->ss->getinfo_super(st, content, NULL);
98dbd966 820 clean = content->array.state & 1;
4b1ac34b 821
64c4757e
NB
822 /* now we have some devices that might be suitable.
823 * I wonder how many
824 */
98dbd966
DW
825 avail = malloc(content->array.raid_disks);
826 memset(avail, 0, content->array.raid_disks);
64c4757e 827 okcnt = 0;
52826846 828 sparecnt=0;
1ff98339 829 rebuilding_cnt=0;
f21e18ca 830 for (i=0; i< bestcnt; i++) {
64c4757e 831 int j = best[i];
ee04451c
NB
832 int event_margin = 1; /* always allow a difference of '1'
833 * like the kernel does
834 */
64c4757e 835 if (j < 0) continue;
d013a55e
NB
836 /* note: we ignore error flags in multipath arrays
837 * as they don't make sense
838 */
df0d4ea0 839 if (content->array.level != LEVEL_MULTIPATH)
f22385f9 840 if (!(devices[j].i.disk.state & (1<<MD_DISK_ACTIVE))) {
213ee40b
NB
841 if (!(devices[j].i.disk.state
842 & (1<<MD_DISK_FAULTY)))
aa88f531 843 sparecnt++;
d013a55e 844 continue;
aa88f531 845 }
02e7c5b7
N
846 /* If this devices thinks that 'most_recent' has failed, then
847 * we must reject this device.
848 */
849 if (j != most_recent &&
850 content->array.raid_disks > 0 &&
851 devices[most_recent].i.disk.raid_disk >= 0 &&
852 devmap[j * content->array.raid_disks + devices[most_recent].i.disk.raid_disk] == 0) {
853 if (verbose > -1)
854 fprintf(stderr, Name ": ignoring %s as it reports %s as failed\n",
855 devices[j].devname, devices[most_recent].devname);
856 best[i] = -1;
857 continue;
858 }
213ee40b
NB
859 if (devices[j].i.events+event_margin >=
860 devices[most_recent].i.events) {
64c4757e 861 devices[j].uptodate = 1;
1ff98339
N
862 if (i < content->array.raid_disks) {
863 if (devices[j].i.recovery_start == MaxSector) {
864 okcnt++;
865 avail[i]=1;
866 } else
867 rebuilding_cnt++;
265e0f17 868 } else
52826846 869 sparecnt++;
64c4757e
NB
870 }
871 }
02e7c5b7 872 free(devmap);
98dbd966
DW
873 while (force && !enough(content->array.level, content->array.raid_disks,
874 content->array.layout, 1,
265e0f17 875 avail, okcnt)) {
64c4757e
NB
876 /* Choose the newest best drive which is
877 * not up-to-date, update the superblock
878 * and add it.
879 */
52826846 880 int fd;
3da92f27 881 struct supertype *tst;
f21e18ca 882 unsigned long long current_events;
cd29a5c8 883 chosen_drive = -1;
f21e18ca 884 for (i = 0; i < content->array.raid_disks && i < bestcnt; i++) {
52826846
NB
885 int j = best[i];
886 if (j>=0 &&
887 !devices[j].uptodate &&
921d9e16 888 devices[j].i.recovery_start == MaxSector &&
52826846 889 (chosen_drive < 0 ||
213ee40b
NB
890 devices[j].i.events
891 > devices[chosen_drive].i.events))
52826846
NB
892 chosen_drive = j;
893 }
894 if (chosen_drive < 0)
895 break;
213ee40b 896 current_events = devices[chosen_drive].i.events;
c5a6f9a6 897 add_another:
dab6685f
NB
898 if (verbose >= 0)
899 fprintf(stderr, Name ": forcing event count in %s(%d) from %d upto %d\n",
213ee40b
NB
900 devices[chosen_drive].devname,
901 devices[chosen_drive].i.disk.raid_disk,
902 (int)(devices[chosen_drive].i.events),
903 (int)(devices[most_recent].i.events));
8b0dabea 904 fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
52826846
NB
905 if (fd < 0) {
906 fprintf(stderr, Name ": Couldn't open %s for write - not updating\n",
907 devices[chosen_drive].devname);
213ee40b 908 devices[chosen_drive].i.events = 0;
52826846
NB
909 continue;
910 }
3da92f27 911 tst = dup_super(st);
60b435db 912 if (tst->ss->load_super(tst,fd, NULL)) {
52826846
NB
913 close(fd);
914 fprintf(stderr, Name ": RAID superblock disappeared from %s - not updating.\n",
915 devices[chosen_drive].devname);
213ee40b 916 devices[chosen_drive].i.events = 0;
52826846
NB
917 continue;
918 }
98dbd966
DW
919 content->events = devices[most_recent].i.events;
920 tst->ss->update_super(tst, content, "force-one",
67a8c82d
NB
921 devices[chosen_drive].devname, verbose,
922 0, NULL);
4b1ac34b 923
3da92f27 924 if (tst->ss->store_super(tst, fd)) {
52826846
NB
925 close(fd);
926 fprintf(stderr, Name ": Could not re-write superblock on %s\n",
927 devices[chosen_drive].devname);
213ee40b 928 devices[chosen_drive].i.events = 0;
3da92f27 929 tst->ss->free_super(tst);
52826846
NB
930 continue;
931 }
932 close(fd);
213ee40b 933 devices[chosen_drive].i.events = devices[most_recent].i.events;
52826846 934 devices[chosen_drive].uptodate = 1;
265e0f17 935 avail[chosen_drive] = 1;
52826846 936 okcnt++;
3da92f27 937 tst->ss->free_super(tst);
c5a6f9a6
NB
938
939 /* If there are any other drives of the same vintage,
940 * add them in as well. We can't lose and we might gain
941 */
f21e18ca 942 for (i = 0; i < content->array.raid_disks && i < bestcnt ; i++) {
c5a6f9a6
NB
943 int j = best[i];
944 if (j >= 0 &&
945 !devices[j].uptodate &&
213ee40b 946 devices[j].i.events == current_events) {
c5a6f9a6
NB
947 chosen_drive = j;
948 goto add_another;
949 }
950 }
64c4757e 951 }
52826846
NB
952
953 /* Now we want to look at the superblock which the kernel will base things on
954 * and compare the devices that we think are working with the devices that the
955 * superblock thinks are working.
956 * If there are differences and --force is given, then update this chosen
957 * superblock.
958 */
cd29a5c8 959 chosen_drive = -1;
3da92f27 960 st->ss->free_super(st);
56eedc1a 961 for (i=0; chosen_drive < 0 && i<bestcnt; i++) {
52826846
NB
962 int j = best[i];
963 int fd;
4b1ac34b 964
52826846
NB
965 if (j<0)
966 continue;
967 if (!devices[j].uptodate)
968 continue;
969 chosen_drive = j;
8b0dabea 970 if ((fd=dev_open(devices[j].devname, O_RDONLY|O_EXCL))< 0) {
52826846
NB
971 fprintf(stderr, Name ": Cannot open %s: %s\n",
972 devices[j].devname, strerror(errno));
7f91af49 973 close(mdfd);
d7f7ebb7 974 free(devices);
52826846
NB
975 return 1;
976 }
3da92f27 977 if (st->ss->load_super(st,fd, NULL)) {
52826846
NB
978 close(fd);
979 fprintf(stderr, Name ": RAID superblock has disappeared from %s\n",
980 devices[j].devname);
7f91af49 981 close(mdfd);
d7f7ebb7 982 free(devices);
52826846
NB
983 return 1;
984 }
985 close(fd);
986 }
3da92f27 987 if (st->sb == NULL) {
4b1ac34b 988 fprintf(stderr, Name ": No suitable drives found for %s\n", mddev);
7f91af49 989 close(mdfd);
d7f7ebb7 990 free(devices);
4b1ac34b
NB
991 return 1;
992 }
a5d85af7 993 st->ss->getinfo_super(st, content, NULL);
f35f2525 994#ifndef MDASSEMBLE
98dbd966 995 sysfs_init(content, mdfd, 0);
f35f2525 996#endif
56eedc1a 997 for (i=0; i<bestcnt; i++) {
52826846 998 int j = best[i];
98c6faba 999 unsigned int desired_state;
11a3e71d 1000
98dbd966 1001 if (i < content->array.raid_disks)
11a3e71d
NB
1002 desired_state = (1<<MD_DISK_ACTIVE) | (1<<MD_DISK_SYNC);
1003 else
1004 desired_state = 0;
1005
52826846
NB
1006 if (j<0)
1007 continue;
1008 if (!devices[j].uptodate)
1009 continue;
4b1ac34b 1010
213ee40b 1011 devices[j].i.disk.state = desired_state;
4e9a6ff7
N
1012 if (!(devices[j].i.array.state & 1))
1013 clean = 0;
213ee40b
NB
1014
1015 if (st->ss->update_super(st, &devices[j].i, "assemble", NULL,
68c7d6d7 1016 verbose, 0, NULL)) {
52826846 1017 if (force) {
dab6685f
NB
1018 if (verbose >= 0)
1019 fprintf(stderr, Name ": "
1020 "clearing FAULTY flag for device %d in %s for %s\n",
1021 j, mddev, devices[j].devname);
4b1ac34b 1022 change = 1;
52826846 1023 } else {
dab6685f
NB
1024 if (verbose >= -1)
1025 fprintf(stderr, Name ": "
1026 "device %d in %s has wrong state in superblock, but %s seems ok\n",
1027 i, mddev, devices[j].devname);
52826846
NB
1028 }
1029 }
4b1ac34b 1030#if 0
213ee40b 1031 if (!(super.disks[i].i.disk.state & (1 << MD_DISK_FAULTY))) {
52826846
NB
1032 fprintf(stderr, Name ": devices %d of %s is not marked FAULTY in superblock, but cannot be found\n",
1033 i, mddev);
1034 }
4b1ac34b 1035#endif
52826846 1036 }
c5a6f9a6 1037 if (force && !clean &&
98dbd966
DW
1038 !enough(content->array.level, content->array.raid_disks,
1039 content->array.layout, clean,
c5a6f9a6 1040 avail, okcnt)) {
98dbd966 1041 change += st->ss->update_super(st, content, "force-array",
67a8c82d
NB
1042 devices[chosen_drive].devname, verbose,
1043 0, NULL);
583315d9 1044 clean = 1;
d013a55e 1045 }
52826846 1046
4b1ac34b 1047 if (change) {
52826846 1048 int fd;
8b0dabea 1049 fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
52826846 1050 if (fd < 0) {
353632d9 1051 fprintf(stderr, Name ": Could not open %s for write - cannot Assemble array.\n",
52826846 1052 devices[chosen_drive].devname);
7f91af49 1053 close(mdfd);
d7f7ebb7 1054 free(devices);
52826846
NB
1055 return 1;
1056 }
3da92f27 1057 if (st->ss->store_super(st, fd)) {
52826846
NB
1058 close(fd);
1059 fprintf(stderr, Name ": Could not re-write superblock on %s\n",
1060 devices[chosen_drive].devname);
7f91af49 1061 close(mdfd);
d7f7ebb7 1062 free(devices);
52826846
NB
1063 return 1;
1064 }
1065 close(fd);
52826846
NB
1066 }
1067
353632d9
NB
1068 /* If we are in the middle of a reshape we may need to restore saved data
1069 * that was moved aside due to the reshape overwriting live data
1070 * The code of doing this lives in Grow.c
1071 */
39c5a909 1072#ifndef MDASSEMBLE
98dbd966 1073 if (content->reshape_active) {
353632d9
NB
1074 int err = 0;
1075 int *fdlist = malloc(sizeof(int)* bestcnt);
cd77ac4e 1076 if (verbose > 0)
ea0ebe96
N
1077 fprintf(stderr, Name ":%s has an active reshape - checking "
1078 "if critical section needs to be restored\n",
1079 chosen_name);
353632d9
NB
1080 for (i=0; i<bestcnt; i++) {
1081 int j = best[i];
1082 if (j >= 0) {
1083 fdlist[i] = dev_open(devices[j].devname, O_RDWR|O_EXCL);
1084 if (fdlist[i] < 0) {
1085 fprintf(stderr, Name ": Could not open %s for write - cannot Assemble array.\n",
1086 devices[j].devname);
1087 err = 1;
1088 break;
1089 }
1090 } else
1091 fdlist[i] = -1;
1092 }
1093 if (!err)
cd77ac4e 1094 err = Grow_restart(st, content, fdlist, bestcnt, backup_file, verbose > 0);
353632d9
NB
1095 while (i>0) {
1096 i--;
1097 if (fdlist[i]>=0) close(fdlist[i]);
1098 }
1099 if (err) {
1100 fprintf(stderr, Name ": Failed to restore critical section for reshape, sorry.\n");
e9e43ec3
N
1101 if (backup_file == NULL)
1102 fprintf(stderr," Possibly you needed to specify the --backup-file\n");
7f91af49 1103 close(mdfd);
d7f7ebb7 1104 free(devices);
353632d9
NB
1105 return err;
1106 }
1107 }
39c5a909 1108#endif
aa88f531
NB
1109 /* count number of in-sync devices according to the superblock.
1110 * We must have this number to start the array without -s or -R
1111 */
98dbd966 1112 req_cnt = content->array.working_disks;
aa88f531 1113
64c4757e
NB
1114 /* Almost ready to actually *do* something */
1115 if (!old_linux) {
fbf8a0b7 1116 int rv;
a19c88b8 1117
a04d5763
N
1118 /* First, fill in the map, so that udev can find our name
1119 * as soon as we become active.
1120 */
98dbd966
DW
1121 map_update(NULL, fd2devnum(mdfd), content->text_version,
1122 content->uuid, chosen_name);
a04d5763 1123
98dbd966 1124 rv = set_array_info(mdfd, st, content);
fbf8a0b7 1125 if (rv) {
f35f2525 1126 fprintf(stderr, Name ": failed to set array info for %s: %s\n",
64c4757e 1127 mddev, strerror(errno));
24af7a87 1128 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1129 close(mdfd);
d7f7ebb7 1130 free(devices);
64c4757e
NB
1131 return 1;
1132 }
11484a63 1133 if (ident->bitmap_fd >= 0) {
c82f047c
NB
1134 if (ioctl(mdfd, SET_BITMAP_FILE, ident->bitmap_fd) != 0) {
1135 fprintf(stderr, Name ": SET_BITMAP_FILE failed.\n");
24af7a87 1136 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1137 close(mdfd);
d7f7ebb7 1138 free(devices);
c82f047c
NB
1139 return 1;
1140 }
7ef02d01
NB
1141 } else if (ident->bitmap_file) {
1142 /* From config file */
1143 int bmfd = open(ident->bitmap_file, O_RDWR);
1144 if (bmfd < 0) {
1145 fprintf(stderr, Name ": Could not open bitmap file %s\n",
1146 ident->bitmap_file);
24af7a87 1147 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1148 close(mdfd);
d7f7ebb7 1149 free(devices);
7ef02d01
NB
1150 return 1;
1151 }
1152 if (ioctl(mdfd, SET_BITMAP_FILE, bmfd) != 0) {
1153 fprintf(stderr, Name ": Failed to set bitmapfile for %s\n", mddev);
1154 close(bmfd);
24af7a87 1155 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1156 close(mdfd);
d7f7ebb7 1157 free(devices);
7ef02d01
NB
1158 return 1;
1159 }
1160 close(bmfd);
c82f047c 1161 }
7ef02d01 1162
52826846 1163 /* First, add the raid disks, but add the chosen one last */
56eedc1a 1164 for (i=0; i<= bestcnt; i++) {
52826846 1165 int j;
56eedc1a 1166 if (i < bestcnt) {
52826846
NB
1167 j = best[i];
1168 if (j == chosen_drive)
1169 continue;
1170 } else
1171 j = chosen_drive;
1172
56eedc1a 1173 if (j >= 0 /* && devices[j].uptodate */) {
98dbd966 1174 rv = add_disk(mdfd, st, content, &devices[j].i);
7801ac20 1175
a19c88b8 1176 if (rv) {
213ee40b
NB
1177 fprintf(stderr, Name ": failed to add "
1178 "%s to %s: %s\n",
64c4757e
NB
1179 devices[j].devname,
1180 mddev,
1181 strerror(errno));
98dbd966 1182 if (i < content->array.raid_disks
213ee40b 1183 || i == bestcnt)
52826846
NB
1184 okcnt--;
1185 else
1186 sparecnt--;
dab6685f 1187 } else if (verbose > 0)
213ee40b
NB
1188 fprintf(stderr, Name ": added %s "
1189 "to %s as %d\n",
1190 devices[j].devname, mddev,
1191 devices[j].i.disk.raid_disk);
98dbd966 1192 } else if (verbose > 0 && i < content->array.raid_disks)
213ee40b
NB
1193 fprintf(stderr, Name ": no uptodate device for "
1194 "slot %d of %s\n",
64c4757e
NB
1195 i, mddev);
1196 }
aba69144 1197
98dbd966 1198 if (content->array.level == LEVEL_CONTAINER) {
598f0d58
NB
1199 if (verbose >= 0) {
1200 fprintf(stderr, Name ": Container %s has been "
1201 "assembled with %d drive%s",
57ed8c91 1202 mddev, okcnt+sparecnt, okcnt+sparecnt==1?"":"s");
f21e18ca 1203 if (okcnt < (unsigned)content->array.raid_disks)
598f0d58 1204 fprintf(stderr, " (out of %d)",
98dbd966 1205 content->array.raid_disks);
598f0d58
NB
1206 fprintf(stderr, "\n");
1207 }
98dbd966 1208 sysfs_uevent(content, "change");
a7c6e3fb 1209 wait_for(chosen_name, mdfd);
7f91af49 1210 close(mdfd);
d7f7ebb7 1211 free(devices);
598f0d58
NB
1212 return 0;
1213 }
1214
64c4757e 1215 if (runstop == 1 ||
b8a8ccf9 1216 (runstop <= 0 &&
98dbd966
DW
1217 ( enough(content->array.level, content->array.raid_disks,
1218 content->array.layout, clean, avail, okcnt) &&
1ff98339 1219 (okcnt + rebuilding_cnt >= req_cnt || start_partial_ok)
aa88f531 1220 ))) {
e9e43ec3
N
1221 /* This array is good-to-go.
1222 * If a reshape is in progress then we might need to
1223 * continue monitoring it. In that case we start
1224 * it read-only and let the grow code make it writable.
1225 */
1226 int rv;
eb3929a4 1227#ifndef MDASSEMBLE
e9e43ec3
N
1228 if (content->reshape_active &&
1229 content->delta_disks <= 0)
1230 rv = Grow_continue(mdfd, st, content, backup_file);
1231 else
eb3929a4 1232#endif
e9e43ec3
N
1233 rv = ioctl(mdfd, RUN_ARRAY, NULL);
1234 if (rv == 0) {
dab6685f
NB
1235 if (verbose >= 0) {
1236 fprintf(stderr, Name ": %s has been started with %d drive%s",
1237 mddev, okcnt, okcnt==1?"":"s");
f21e18ca 1238 if (okcnt < (unsigned)content->array.raid_disks)
98dbd966 1239 fprintf(stderr, " (out of %d)", content->array.raid_disks);
1ff98339
N
1240 if (rebuilding_cnt)
1241 fprintf(stderr, "%s %d rebuilding", sparecnt?",":" and", rebuilding_cnt);
dab6685f
NB
1242 if (sparecnt)
1243 fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
1244 fprintf(stderr, ".\n");
1245 }
8a659c33
N
1246 if (content->reshape_active &&
1247 content->array.level >= 4 &&
1248 content->array.level <= 6) {
acee8e89
N
1249 /* might need to increase the size
1250 * of the stripe cache - default is 256
1251 */
8a659c33 1252 if (256 < 4 * (content->array.chunk_size/4096)) {
acee8e89
N
1253 struct mdinfo *sra = sysfs_read(mdfd, 0, 0);
1254 if (sra)
1255 sysfs_set_num(sra, NULL,
1256 "stripe_cache_size",
8a659c33 1257 (4 * content->array.chunk_size / 4096) + 1);
acee8e89
N
1258 }
1259 }
7e83544b
N
1260 if (okcnt < (unsigned)content->array.raid_disks) {
1261 /* If any devices did not get added
1262 * because the kernel rejected them based
1263 * on event count, try adding them
1264 * again providing the action policy is
1265 * 're-add' or greater. The bitmap
1266 * might allow them to be included, or
1267 * they will become spares.
1268 */
1269 for (i = 0; i <= bestcnt; i++) {
1270 int j = best[i];
1271 if (j >= 0 && !devices[j].uptodate) {
1272 if (!disk_action_allows(&devices[j].i, st->ss->name, act_re_add))
1273 continue;
1274 rv = add_disk(mdfd, st, content,
1275 &devices[j].i);
1276 if (rv == 0 && verbose >= 0)
1277 fprintf(stderr,
1278 Name ": %s has been re-added.\n",
1279 devices[j].devname);
1280 }
1281 }
1282 }
a7c6e3fb 1283 wait_for(mddev, mdfd);
7f91af49
N
1284 close(mdfd);
1285 if (auto_assem) {
589395d6 1286 int usecs = 1;
589395d6
NB
1287 /* There is a nasty race with 'mdadm --monitor'.
1288 * If it opens this device before we close it,
1289 * it gets an incomplete open on which IO
598f0d58
NB
1290 * doesn't work and the capacity is
1291 * wrong.
589395d6
NB
1292 * If we reopen (to check for layered devices)
1293 * before --monitor closes, we loose.
1294 *
1295 * So: wait upto 1 second for there to be
1296 * a non-zero capacity.
1297 */
1298 while (usecs < 1000) {
1299 mdfd = open(mddev, O_RDONLY);
1300 if (mdfd >= 0) {
beae1dfe
NB
1301 unsigned long long size;
1302 if (get_dev_size(mdfd, NULL, &size) &&
589395d6
NB
1303 size > 0)
1304 break;
1305 close(mdfd);
1306 }
1307 usleep(usecs);
1308 usecs <<= 1;
1309 }
1310 }
d7f7ebb7 1311 free(devices);
64c4757e 1312 return 0;
82b27616 1313 }
682c7051 1314 fprintf(stderr, Name ": failed to RUN_ARRAY %s: %s\n",
64c4757e 1315 mddev, strerror(errno));
583315d9 1316
98dbd966
DW
1317 if (!enough(content->array.level, content->array.raid_disks,
1318 content->array.layout, 1, avail, okcnt))
583315d9
NB
1319 fprintf(stderr, Name ": Not enough devices to "
1320 "start the array.\n");
98dbd966
DW
1321 else if (!enough(content->array.level,
1322 content->array.raid_disks,
1323 content->array.layout, clean,
583315d9
NB
1324 avail, okcnt))
1325 fprintf(stderr, Name ": Not enough devices to "
1326 "start the array while not clean "
1327 "- consider --force.\n");
1328
7f91af49 1329 if (auto_assem)
1c203a4b 1330 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1331 close(mdfd);
d7f7ebb7 1332 free(devices);
64c4757e
NB
1333 return 1;
1334 }
82b27616 1335 if (runstop == -1) {
b8a8ccf9 1336 fprintf(stderr, Name ": %s assembled from %d drive%s",
52826846 1337 mddev, okcnt, okcnt==1?"":"s");
f21e18ca 1338 if (okcnt != (unsigned)content->array.raid_disks)
98dbd966 1339 fprintf(stderr, " (out of %d)", content->array.raid_disks);
b8a8ccf9 1340 fprintf(stderr, ", but not started.\n");
7f91af49 1341 close(mdfd);
d7f7ebb7 1342 free(devices);
64c4757e 1343 return 0;
82b27616 1344 }
4855f95c 1345 if (verbose >= -1) {
dab6685f 1346 fprintf(stderr, Name ": %s assembled from %d drive%s", mddev, okcnt, okcnt==1?"":"s");
1ff98339
N
1347 if (rebuilding_cnt)
1348 fprintf(stderr, "%s %d rebuilding", sparecnt?", ":" and ", rebuilding_cnt);
dab6685f
NB
1349 if (sparecnt)
1350 fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
98dbd966
DW
1351 if (!enough(content->array.level, content->array.raid_disks,
1352 content->array.layout, 1, avail, okcnt))
dab6685f 1353 fprintf(stderr, " - not enough to start the array.\n");
98dbd966
DW
1354 else if (!enough(content->array.level,
1355 content->array.raid_disks,
1356 content->array.layout, clean,
583315d9
NB
1357 avail, okcnt))
1358 fprintf(stderr, " - not enough to start the "
1359 "array while not clean - consider "
1360 "--force.\n");
dab6685f 1361 else {
f21e18ca 1362 if (req_cnt == (unsigned)content->array.raid_disks)
dab6685f
NB
1363 fprintf(stderr, " - need all %d to start it", req_cnt);
1364 else
98dbd966 1365 fprintf(stderr, " - need %d of %d to start", req_cnt, content->array.raid_disks);
dab6685f
NB
1366 fprintf(stderr, " (use --run to insist).\n");
1367 }
aa88f531 1368 }
7f91af49 1369 if (auto_assem)
1c203a4b 1370 ioctl(mdfd, STOP_ARRAY, NULL);
56a8da69 1371 close(mdfd);
d7f7ebb7 1372 free(devices);
64c4757e 1373 return 1;
82b27616 1374 } else {
52826846
NB
1375 /* The "chosen_drive" is a good choice, and if necessary, the superblock has
1376 * been updated to point to the current locations of devices.
1377 * so we can just start the array
82b27616 1378 */
cd29a5c8 1379 unsigned long dev;
213ee40b
NB
1380 dev = makedev(devices[chosen_drive].i.disk.major,
1381 devices[chosen_drive].i.disk.minor);
82b27616
NB
1382 if (ioctl(mdfd, START_ARRAY, dev)) {
1383 fprintf(stderr, Name ": Cannot start array: %s\n",
1384 strerror(errno));
1385 }
aba69144 1386
64c4757e 1387 }
7f91af49 1388 close(mdfd);
d7f7ebb7 1389 free(devices);
e0d19036 1390 return 0;
64c4757e 1391}
6234c63c
DW
1392
1393#ifndef MDASSEMBLE
1394int assemble_container_content(struct supertype *st, int mdfd,
1395 struct mdinfo *content, int runstop,
1396 char *chosen_name, int verbose)
1397{
1398 struct mdinfo *dev, *sra;
1399 int working = 0, preexist = 0;
1400 struct map_ent *map = NULL;
1401
1402 sysfs_init(content, mdfd, 0);
1403
1404 sra = sysfs_read(mdfd, 0, GET_VERSION);
1405 if (sra == NULL || strcmp(sra->text_version, content->text_version) != 0)
4408ee76
N
1406 if (sysfs_set_array(content, md_get_version(mdfd)) != 0) {
1407 close(mdfd);
6234c63c 1408 return 1;
4408ee76 1409 }
6234c63c
DW
1410 if (sra)
1411 sysfs_free(sra);
1412
1413 for (dev = content->devs; dev; dev = dev->next)
462906cd 1414 if (sysfs_add_disk(content, dev, 1) == 0)
6234c63c
DW
1415 working++;
1416 else if (errno == EEXIST)
1417 preexist++;
4408ee76
N
1418 if (working == 0) {
1419 close(mdfd);
7cb2aa33 1420 return 1;/* Nothing new, don't try to start */
8b4e5ea9
N
1421 }
1422
1423 map_update(&map, fd2devnum(mdfd),
1424 content->text_version,
1425 content->uuid, chosen_name);
1426
1427 if (runstop > 0 ||
6234c63c 1428 (working + preexist) >= content->array.working_disks) {
bb50e5d3 1429 int err;
a714580e 1430
6234c63c
DW
1431 switch(content->array.level) {
1432 case LEVEL_LINEAR:
1433 case LEVEL_MULTIPATH:
1434 case 0:
bb50e5d3
N
1435 err = sysfs_set_str(content, NULL, "array_state",
1436 "active");
6234c63c
DW
1437 break;
1438 default:
bb50e5d3 1439 err = sysfs_set_str(content, NULL, "array_state",
6234c63c
DW
1440 "readonly");
1441 /* start mdmon if needed. */
bb50e5d3
N
1442 if (!err) {
1443 if (!mdmon_running(st->container_dev))
1444 start_mdmon(st->container_dev);
1445 ping_monitor(devnum2devname(st->container_dev));
1446 }
6234c63c
DW
1447 break;
1448 }
bb50e5d3
N
1449 if (!err)
1450 sysfs_set_safemode(content, content->safe_mode_delay);
6234c63c 1451 if (verbose >= 0) {
bb50e5d3
N
1452 if (err)
1453 fprintf(stderr, Name
1454 ": array %s now has %d devices",
1455 chosen_name, working + preexist);
1456 else
1457 fprintf(stderr, Name
1458 ": Started %s with %d devices",
1459 chosen_name, working + preexist);
6234c63c
DW
1460 if (preexist)
1461 fprintf(stderr, " (%d new)", working);
1462 fprintf(stderr, "\n");
1463 }
bb50e5d3 1464 if (!err)
a7c6e3fb 1465 wait_for(chosen_name, mdfd);
4408ee76 1466 close(mdfd);
7cb2aa33 1467 return 0;
6234c63c 1468 /* FIXME should have an O_EXCL and wait for read-auto */
7cb2aa33 1469 } else {
6234c63c
DW
1470 if (verbose >= 0)
1471 fprintf(stderr, Name
1472 ": %s assembled with %d devices but "
1473 "not started\n",
1474 chosen_name, working);
4408ee76 1475 close(mdfd);
7cb2aa33
N
1476 return 1;
1477 }
6234c63c
DW
1478}
1479#endif
1480