]> git.ipfire.org Git - thirdparty/mdadm.git/blame - Assemble.c
Monitor: reduce default poll interval if mdstat is pollable.
[thirdparty/mdadm.git] / Assemble.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
4f589ad0 4 * Copyright (C) 2001-2006 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
22 * Email: <neilb@cse.unsw.edu.au>
23 * Paper: Neil Brown
24 * School of Computer Science and Engineering
25 * The University of New South Wales
26 * Sydney, 2052
27 * Australia
28 */
29
9a9dab36 30#include "mdadm.h"
41a3b72a 31#include <ctype.h>
64c4757e 32
624920bb
NB
33static int name_matches(char *found, char *required, char *homehost)
34{
35 /* See if the name found matches the required name, possibly
36 * prefixed with 'homehost'
37 */
38 char fnd[33];
39
40 strncpy(fnd, found, 32);
41 fnd[32] = 0;
42 if (strcmp(found, required)==0)
43 return 1;
44 if (homehost) {
45 int l = strlen(homehost);
46 if (l < 32 && fnd[l] == ':' &&
47 strcmp(fnd+l+1, required)==0)
48 return 1;
49 }
50 return 0;
51}
52
9008ed1c 53static int is_member_busy(char *metadata_version)
2de8884f
DW
54{
55 /* check if the given member array is active */
56 struct mdstat_ent *mdstat = mdstat_read(1, 0);
57 struct mdstat_ent *ent;
58 int busy = 0;
59
60 for (ent = mdstat; ent; ent = ent->next) {
61 if (ent->metadata_version == NULL)
62 continue;
63 if (strncmp(ent->metadata_version, "external:", 9) != 0)
64 continue;
65 if (!is_subarray(&ent->metadata_version[9]))
66 continue;
67 /* Skip first char - it can be '/' or '-' */
68 if (strcmp(&ent->metadata_version[10], metadata_version+1) == 0) {
69 busy = 1;
70 break;
71 }
72 }
73 free_mdstat(mdstat);
74
75 return busy;
76}
77
7f91af49 78int Assemble(struct supertype *st, char *mddev,
8aec876d 79 mddev_ident_t ident,
06b0d786 80 mddev_dev_t devlist, char *backup_file,
64c4757e 81 int readonly, int runstop,
0ac91628 82 char *update, char *homehost, int require_homehost,
64c4757e
NB
83 int verbose, int force)
84{
85 /*
52826846
NB
86 * The task of Assemble is to find a collection of
87 * devices that should (according to their superblocks)
88 * form an array, and to give this collection to the MD driver.
89 * In Linux-2.4 and later, this involves submitting a
64c4757e
NB
90 * SET_ARRAY_INFO ioctl with no arg - to prepare
91 * the array - and then submit a number of
92 * ADD_NEW_DISK ioctls to add disks into
93 * the array. Finally RUN_ARRAY might
94 * be submitted to start the array.
95 *
96 * Much of the work of Assemble is in finding and/or
97 * checking the disks to make sure they look right.
98 *
4b1ac34b 99 * If mddev is not set, then scan must be set and we
64c4757e
NB
100 * read through the config file for dev+uuid mapping
101 * We recurse, setting mddev, for each device that
102 * - isn't running
4b1ac34b 103 * - has a valid uuid (or any uuid if !uuidset)
64c4757e
NB
104 *
105 * If mddev is set, we try to determine state of md.
106 * check version - must be at least 0.90.0
107 * check kernel version. must be at least 2.4.
108 * If not, we can possibly fall back on START_ARRAY
109 * Try to GET_ARRAY_INFO.
110 * If possible, give up
111 * If not, try to STOP_ARRAY just to make sure
112 *
113 * If !uuidset and scan, look in conf-file for uuid
114 * If not found, give up
b8a8ccf9 115 * If !devlist and scan and uuidset, get list of devs from conf-file
64c4757e
NB
116 *
117 * For each device:
118 * Check superblock - discard if bad
119 * Check uuid (set if we don't have one) - discard if no match
5787fa49 120 * Check superblock similarity if we have a superblock - discard if different
4b1ac34b 121 * Record events, devicenum
64c4757e 122 * This should give us a list of devices for the array
4b1ac34b 123 * We should collect the most recent event number
64c4757e
NB
124 *
125 * Count disks with recent enough event count
126 * While force && !enough disks
127 * Choose newest rejected disks, update event count
128 * mark clean and rewrite superblock
129 * If recent kernel:
130 * SET_ARRAY_INFO
131 * foreach device with recent events : ADD_NEW_DISK
132 * if runstop == 1 || "enough" disks and runstop==0 -> RUN_ARRAY
133 * If old kernel:
134 * Check the device numbers in superblock are right
135 * update superblock if any changes
136 * START_ARRAY
137 *
138 */
c30e5369
N
139 int mdfd;
140 int clean;
f05641cf
N
141 int auto_assem = (mddev == NULL && !ident->uuid_set &&
142 ident->super_minor == UnSet && ident->name[0] == 0
aa7c284c 143 && (ident->container == NULL || ident->member == NULL));
64c4757e 144 int old_linux = 0;
c30e5369 145 int vers = vers; /* Keep gcc quite - it really is initialised */
64c4757e
NB
146 struct {
147 char *devname;
213ee40b
NB
148 int uptodate; /* set once we decide that this device is as
149 * recent as everything else in the array.
150 */
151 struct mdinfo i;
5787fa49 152 } *devices;
56eedc1a 153 int *best = NULL; /* indexed by raid_disk */
98c6faba
NB
154 unsigned int bestcnt = 0;
155 int devcnt = 0;
156 unsigned int okcnt, sparecnt;
157 unsigned int req_cnt;
158 unsigned int i;
64c4757e 159 int most_recent = 0;
cd29a5c8 160 int chosen_drive;
52826846 161 int change = 0;
cd29a5c8 162 int inargv = 0;
52437b4f 163 int report_missmatch;
bf4fb153 164 int bitmap_done;
7f91af49
N
165 int start_partial_ok = (runstop >= 0) &&
166 (force || devlist==NULL || auto_assem);
98c6faba 167 unsigned int num_devs;
5787fa49 168 mddev_dev_t tmpdev;
4b1ac34b 169 struct mdinfo info;
98dbd966 170 struct mdinfo *content = NULL;
265e0f17 171 char *avail;
308e1801 172 int nextspare = 0;
197e3eb6 173 char *name = NULL;
69207ff6 174 int trustworthy;
a04d5763 175 char chosen_name[1024];
8a46fe84 176
682c7051 177 if (get_linux_version() < 2004000)
64c4757e
NB
178 old_linux = 1;
179
64c4757e 180 /*
52826846
NB
181 * If any subdevs are listed, then any that don't
182 * match ident are discarded. Remainder must all match and
183 * become the array.
184 * If no subdevs, then we scan all devices in the config file, but
185 * there must be something in the identity
64c4757e 186 */
64c4757e 187
cd29a5c8 188 if (!devlist &&
52826846
NB
189 ident->uuid_set == 0 &&
190 ident->super_minor < 0 &&
e0fe762a
N
191 ident->name[0] == 0 &&
192 (ident->container == NULL || ident->member == NULL) &&
52826846
NB
193 ident->devices == NULL) {
194 fprintf(stderr, Name ": No identity information available for %s - cannot assemble.\n",
8a46fe84 195 mddev ? mddev : "further assembly");
52826846 196 return 1;
64c4757e 197 }
71d60c48 198
cd29a5c8 199 if (devlist == NULL)
8aec876d 200 devlist = conf_get_devs();
7f91af49 201 else if (mddev)
da6b5ca9 202 inargv = 1;
64c4757e 203
52437b4f 204 report_missmatch = ((inargv && verbose >= 0) || verbose > 0);
60e1bc1a 205 try_again:
c30e5369
N
206 /* We come back here when doing auto-assembly and attempting some
207 * set of devices failed. Those are now marked as ->used==2 and
208 * we ignore them and try again
209 */
60e1bc1a 210
5787fa49
NB
211 tmpdev = devlist; num_devs = 0;
212 while (tmpdev) {
da6b5ca9
NB
213 if (tmpdev->used)
214 tmpdev->used = 2;
215 else
216 num_devs++;
5787fa49
NB
217 tmpdev = tmpdev->next;
218 }
5787fa49
NB
219 devices = malloc(num_devs * sizeof(*devices));
220
82d9eba6 221 if (!st && ident->st) st = ident->st;
64c4757e 222
dab6685f 223 if (verbose>0)
82b27616 224 fprintf(stderr, Name ": looking for devices for %s\n",
8a46fe84 225 mddev ? mddev : "further assembly");
82b27616 226
811e6cbe
NB
227 /* first walk the list of devices to find a consistent set
228 * that match the criterea, if that is possible.
c30e5369 229 * We flag the ones we like with 'used'.
811e6cbe
NB
230 */
231 for (tmpdev = devlist;
232 tmpdev;
233 tmpdev = tmpdev->next) {
234 char *devname = tmpdev->devname;
64c4757e
NB
235 int dfd;
236 struct stat stb;
3da92f27 237 struct supertype *tst = dup_super(st);
52826846 238
da6b5ca9
NB
239 if (tmpdev->used > 1) continue;
240
52826846 241 if (ident->devices &&
56eedc1a 242 !match_oneof(ident->devices, devname)) {
52437b4f 243 if (report_missmatch)
56eedc1a 244 fprintf(stderr, Name ": %s is not one of %s\n", devname, ident->devices);
52826846 245 continue;
56eedc1a 246 }
4b1ac34b 247
8b0dabea 248 dfd = dev_open(devname, O_RDONLY|O_EXCL);
64c4757e 249 if (dfd < 0) {
52437b4f 250 if (report_missmatch)
682c7051 251 fprintf(stderr, Name ": cannot open device %s: %s\n",
64c4757e 252 devname, strerror(errno));
da6b5ca9 253 tmpdev->used = 2;
52826846
NB
254 } else if (fstat(dfd, &stb)< 0) {
255 /* Impossible! */
256 fprintf(stderr, Name ": fstat failed for %s: %s\n",
257 devname, strerror(errno));
da6b5ca9 258 tmpdev->used = 2;
cd29a5c8
NB
259 } else if ((stb.st_mode & S_IFMT) != S_IFBLK) {
260 fprintf(stderr, Name ": %s is not a block device.\n",
52826846 261 devname);
da6b5ca9 262 tmpdev->used = 2;
82d9eba6 263 } else if (!tst && (tst = guess_super(dfd)) == NULL) {
52437b4f 264 if (report_missmatch)
da6b5ca9
NB
265 fprintf(stderr, Name ": no recogniseable superblock on %s\n",
266 devname);
267 tmpdev->used = 2;
31015d57
N
268 } else if (auto_assem && st == NULL &&
269 !conf_test_metadata(tst->ss->name)) {
270 if (report_missmatch)
271 fprintf(stderr, Name ": %s has metadata type %s for which "
272 "auto-assembly is disabled\n",
273 devname, tst->ss->name);
274 tmpdev->used = 2;
3da92f27 275 } else if (tst->ss->load_super(tst,dfd, NULL)) {
52437b4f 276 if (report_missmatch)
682c7051 277 fprintf( stderr, Name ": no RAID superblock on %s\n",
64c4757e 278 devname);
52826846 279 } else {
98dbd966
DW
280 content = &info;
281 memset(content, 0, sizeof(*content));
282 tst->ss->getinfo_super(tst, content);
64c4757e 283 }
c06487ce 284 if (dfd >= 0) close(dfd);
52826846 285
9008ed1c
N
286 if (tst && tst->sb && tst->ss->container_content
287 && tst->loaded_container) {
288 /* tmpdev is a container. We need to be either
289 * looking for a member, or auto-assembling
290 */
291 if (st) {
292 /* already found some components, this cannot
293 * be another one.
294 */
295 if (report_missmatch)
296 fprintf(stderr, Name ": %s is a container, but we are looking for components\n",
297 devname);
298 goto loop;
299 }
300
301 if (ident->container) {
302 if (ident->container[0] == '/' &&
303 !same_dev(ident->container, devname)) {
304 if (report_missmatch)
305 fprintf(stderr, Name ": %s is not the container required (%s)\n",
306 devname, ident->container);
307 goto loop;
308 }
309 if (ident->container[0] != '/') {
310 /* we have a uuid */
311 int uuid[4];
312 if (!parse_uuid(ident->container, uuid) ||
313 !same_uuid(content->uuid, uuid, tst->ss->swapuuid)) {
314 if (report_missmatch)
315 fprintf(stderr, Name ": %s has wrong UUID to be required container\n",
316 devname);
317 goto loop;
318 }
319 }
320 }
321 /* It is worth looking inside this container.
322 */
323 next_member:
324 if (tmpdev->content)
325 content = tmpdev->content;
326 else
327 content = tst->ss->container_content(tst);
328
329 tmpdev->content = content->next;
330 if (tmpdev->content == NULL)
f7ad3ccc 331 tmpdev->used = 2;
9008ed1c
N
332
333 } else if (ident->container || ident->member) {
334 /* No chance of this matching if we don't have
335 * a container */
336 if (report_missmatch)
337 fprintf(stderr, Name "%s is not a container, and one is required.\n",
338 devname);
339 goto loop;
340 }
341
838acbc2 342 if (ident->uuid_set && (!update || strcmp(update, "uuid")!= 0) &&
3da92f27 343 (!tst || !tst->sb ||
98dbd966 344 same_uuid(content->uuid, ident->uuid, tst->ss->swapuuid)==0)) {
52437b4f 345 if (report_missmatch)
52826846
NB
346 fprintf(stderr, Name ": %s has wrong uuid.\n",
347 devname);
df37ffc0 348 goto loop;
82b27616 349 }
c4f12c13 350 if (ident->name[0] && (!update || strcmp(update, "name")!= 0) &&
3da92f27 351 (!tst || !tst->sb ||
98dbd966 352 name_matches(content->name, ident->name, homehost)==0)) {
52437b4f 353 if (report_missmatch)
947fd4dd
NB
354 fprintf(stderr, Name ": %s has wrong name.\n",
355 devname);
df37ffc0 356 goto loop;
947fd4dd 357 }
98c6faba 358 if (ident->super_minor != UnSet &&
3da92f27 359 (!tst || !tst->sb ||
98dbd966 360 ident->super_minor != content->array.md_minor)) {
52437b4f 361 if (report_missmatch)
52826846 362 fprintf(stderr, Name ": %s has wrong super-minor.\n",
64c4757e 363 devname);
df37ffc0 364 goto loop;
64c4757e 365 }
98c6faba 366 if (ident->level != UnSet &&
3da92f27 367 (!tst || !tst->sb ||
98dbd966 368 ident->level != content->array.level)) {
52437b4f 369 if (report_missmatch)
cd29a5c8
NB
370 fprintf(stderr, Name ": %s has wrong raid level.\n",
371 devname);
df37ffc0 372 goto loop;
cd29a5c8 373 }
98c6faba 374 if (ident->raid_disks != UnSet &&
3da92f27 375 (!tst || !tst->sb ||
98dbd966 376 ident->raid_disks!= content->array.raid_disks)) {
52437b4f 377 if (report_missmatch)
cd29a5c8
NB
378 fprintf(stderr, Name ": %s requires wrong number of drives.\n",
379 devname);
df37ffc0 380 goto loop;
cd29a5c8 381 }
c30e5369 382 if (auto_assem) {
3da92f27 383 if (tst == NULL || tst->sb == NULL)
da6b5ca9 384 continue;
da6b5ca9 385 }
83b6208e 386 /* If we are this far, then we are nearly commited to this device.
52826846 387 * If the super_block doesn't exist, or doesn't match others,
83b6208e
NB
388 * then we probably cannot continue
389 * However if one of the arrays is for the homehost, and
390 * the other isn't that can disambiguate.
52826846 391 */
52826846 392
3da92f27 393 if (!tst || !tst->sb) {
52826846
NB
394 fprintf(stderr, Name ": %s has no superblock - assembly aborted\n",
395 devname);
ddd1a492
NB
396 if (st)
397 st->ss->free_super(st);
52826846
NB
398 return 1;
399 }
838acbc2 400
9008ed1c
N
401 if (tst && tst->sb && tst->ss->container_content
402 && tst->loaded_container) {
403 /* we have the one container we need, don't keep
404 * looking. If the chosen member is active, skip.
405 */
406 if (is_member_busy(content->text_version)) {
f7ad3ccc
N
407 if (report_missmatch)
408 fprintf(stderr, Name ": member %s in %s is already assembled\n",
409 content->text_version,
410 devname);
411 tst->ss->free_super(tst);
412 tst = NULL;
413 content = NULL;
9008ed1c
N
414 if (auto_assem)
415 goto loop;
9008ed1c
N
416 return 1;
417 }
418 st = tst; tst = NULL;
419 if (!auto_assem && tmpdev->next != NULL) {
420 fprintf(stderr, Name ": %s is a container, but is not "
421 "only device given: confused and aborting\n",
422 devname);
423 st->ss->free_super(st);
424 return 1;
425 }
426 break;
427 }
83b6208e 428 if (st == NULL)
3da92f27
NB
429 st = dup_super(tst);
430 if (st->minor_version == -1)
431 st->minor_version = tst->minor_version;
83b6208e
NB
432 if (st->ss != tst->ss ||
433 st->minor_version != tst->minor_version ||
64557c33 434 st->ss->compare_super(st, tst) != 0) {
83b6208e 435 /* Some mismatch. If exactly one array matches this host,
da6b5ca9
NB
436 * we can resolve on that one.
437 * Or, if we are auto assembling, we just ignore the second
438 * for now.
83b6208e 439 */
c30e5369 440 if (auto_assem)
df37ffc0 441 goto loop;
83b6208e 442 if (homehost) {
3da92f27
NB
443 int first = st->ss->match_home(st, homehost);
444 int last = tst->ss->match_home(tst, homehost);
9362c1c8
N
445 if (first != last &&
446 (first == 1 || last == 1)) {
83b6208e
NB
447 /* We can do something */
448 if (first) {/* just ignore this one */
52437b4f 449 if (report_missmatch)
83b6208e
NB
450 fprintf(stderr, Name ": %s misses out due to wrong homehost\n",
451 devname);
df37ffc0 452 goto loop;
83b6208e
NB
453 } else { /* reject all those sofar */
454 mddev_dev_t td;
52437b4f 455 if (report_missmatch)
83b6208e
NB
456 fprintf(stderr, Name ": %s overrides previous devices due to good homehost\n",
457 devname);
458 for (td=devlist; td != tmpdev; td=td->next)
459 if (td->used == 1)
460 td->used = 0;
461 tmpdev->used = 1;
df37ffc0 462 goto loop;
83b6208e
NB
463 }
464 }
465 }
52826846
NB
466 fprintf(stderr, Name ": superblock on %s doesn't match others - assembly aborted\n",
467 devname);
3da92f27
NB
468 tst->ss->free_super(tst);
469 st->ss->free_super(st);
52826846 470 return 1;
64c4757e
NB
471 }
472
811e6cbe 473 tmpdev->used = 1;
df37ffc0
NB
474
475 loop:
9008ed1c
N
476 if (tmpdev->content)
477 goto next_member;
3d2b16e7
NB
478 if (tst)
479 tst->ss->free_super(tst);
811e6cbe
NB
480 }
481
98dbd966 482 if (!st || !st->sb || !content)
c30e5369
N
483 return 2;
484
05833051 485 /* Now need to open the array device. Use create_mddev */
98dbd966
DW
486 if (content == &info)
487 st->ss->getinfo_super(st, content);
3d2c4fc7 488
69207ff6 489 trustworthy = FOREIGN;
05833051 490 name = content->name;
745f72f6
N
491 switch (st->ss->match_home(st, homehost)
492 ?: st->ss->match_home(st, "any")) {
69207ff6
N
493 case 1:
494 trustworthy = LOCAL;
98dbd966 495 name = strchr(content->name, ':');
69207ff6
N
496 if (name)
497 name++;
3d2c4fc7 498 else
98dbd966 499 name = content->name;
69207ff6 500 break;
c30e5369 501 }
05833051 502 if (!auto_assem)
aa7c284c 503 /* If the array is listed in mdadm.conf or on
ac2ecf55
N
504 * command line, then we trust the name
505 * even if the array doesn't look local
506 */
507 trustworthy = LOCAL;
508
05833051 509 if (name[0] == 0 &&
98dbd966
DW
510 content->array.level == LEVEL_CONTAINER) {
511 name = content->text_version;
a4bc1720
N
512 trustworthy = METADATA;
513 }
0ac91628
N
514
515 if (name[0] && trustworthy != LOCAL &&
516 ! require_homehost &&
517 conf_name_is_free(name))
518 trustworthy = LOCAL;
519
7cdc0872
N
520 if (trustworthy == LOCAL &&
521 strchr(name, ':'))
522 /* Ignore 'host:' prefix of name */
523 name = strchr(name, ':')+1;
524
a04d5763
N
525 mdfd = create_mddev(mddev, name, ident->autof, trustworthy,
526 chosen_name);
c30e5369
N
527 if (mdfd < 0) {
528 st->ss->free_super(st);
529 free(devices);
530 if (auto_assem)
60e1bc1a 531 goto try_again;
c30e5369
N
532 return 1;
533 }
a04d5763 534 mddev = chosen_name;
c30e5369
N
535 vers = md_get_version(mdfd);
536 if (vers < 9000) {
537 fprintf(stderr, Name ": Assemble requires driver version 0.90.0 or later.\n"
538 " Upgrade your kernel or try --build\n");
539 close(mdfd);
540 return 1;
541 }
66afdfa9 542 if (mddev_busy(fd2devnum(mdfd))) {
c30e5369
N
543 fprintf(stderr, Name ": %s already active, cannot restart it!\n",
544 mddev);
545 for (tmpdev = devlist ;
546 tmpdev && tmpdev->used != 1;
547 tmpdev = tmpdev->next)
548 ;
549 if (tmpdev && auto_assem)
550 fprintf(stderr, Name ": %s needed for %s...\n",
551 mddev, tmpdev->devname);
552 close(mdfd);
553 mdfd = -3;
554 st->ss->free_super(st);
555 free(devices);
556 if (auto_assem)
557 goto try_again;
558 return 1;
8a46fe84 559 }
c30e5369 560 ioctl(mdfd, STOP_ARRAY, NULL); /* just incase it was started but has no content */
8a46fe84 561
9008ed1c
N
562#ifndef MDASSEMBLE
563 if (content != &info) {
564 /* This is a member of a container. Try starting the array. */
565 return assemble_container_content(st, mdfd, content, runstop,
566 chosen_name, verbose);
567 }
568#endif
811e6cbe 569 /* Ok, no bad inconsistancy, we can try updating etc */
bf4fb153 570 bitmap_done = 0;
da6b5ca9 571 for (tmpdev = devlist; tmpdev; tmpdev=tmpdev->next) if (tmpdev->used == 1) {
811e6cbe
NB
572 char *devname = tmpdev->devname;
573 struct stat stb;
5787fa49 574 /* looks like a good enough match to update the super block if needed */
d1f1011b 575#ifndef MDASSEMBLE
5787fa49 576 if (update) {
811e6cbe 577 int dfd;
4b1ac34b
NB
578 /* prepare useful information in info structures */
579 struct stat stb2;
3da92f27 580 struct supertype *tst;
4b1ac34b 581 fstat(mdfd, &stb2);
7d99579f
NB
582
583 if (strcmp(update, "uuid")==0 &&
584 !ident->uuid_set) {
585 int rfd;
586 if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
587 read(rfd, ident->uuid, 16) != 16) {
588 *(__u32*)(ident->uuid) = random();
589 *(__u32*)(ident->uuid+1) = random();
590 *(__u32*)(ident->uuid+2) = random();
591 *(__u32*)(ident->uuid+3) = random();
592 }
593 if (rfd >= 0) close(rfd);
7d99579f 594 }
811e6cbe
NB
595 dfd = dev_open(devname, O_RDWR|O_EXCL);
596
0430ed48
NB
597 remove_partitions(dfd);
598
3da92f27
NB
599 tst = dup_super(st);
600 tst->ss->load_super(tst, dfd, NULL);
98dbd966 601 tst->ss->getinfo_super(tst, content);
811e6cbe 602
98dbd966
DW
603 memcpy(content->uuid, ident->uuid, 16);
604 strcpy(content->name, ident->name);
605 content->array.md_minor = minor(stb2.st_rdev);
811e6cbe 606
98dbd966 607 tst->ss->update_super(tst, content, update,
3da92f27
NB
608 devname, verbose,
609 ident->uuid_set, homehost);
e5eac01f
NB
610 if (strcmp(update, "uuid")==0 &&
611 !ident->uuid_set) {
612 ident->uuid_set = 1;
98dbd966 613 memcpy(ident->uuid, content->uuid, 16);
e5eac01f 614 }
b8a8ccf9 615 if (dfd < 0)
5787fa49
NB
616 fprintf(stderr, Name ": Cannot open %s for superblock update\n",
617 devname);
3da92f27 618 else if (tst->ss->store_super(tst, dfd))
5787fa49
NB
619 fprintf(stderr, Name ": Could not re-write superblock on %s.\n",
620 devname);
621 if (dfd >= 0)
622 close(dfd);
8131b493
NB
623
624 if (strcmp(update, "uuid")==0 &&
bf4fb153 625 ident->bitmap_fd >= 0 && !bitmap_done) {
3da92f27 626 if (bitmap_update_uuid(ident->bitmap_fd,
98dbd966 627 content->uuid,
3da92f27 628 tst->ss->swapuuid) != 0)
bf4fb153
NB
629 fprintf(stderr, Name ": Could not update uuid on external bitmap.\n");
630 else
631 bitmap_done = 1;
632 }
3da92f27 633 tst->ss->free_super(tst);
d1f1011b
NB
634 } else
635#endif
636 {
30e1b9a5 637 struct supertype *tst = dup_super(st);
da6b5ca9
NB
638 int dfd;
639 dfd = dev_open(devname, O_RDWR|O_EXCL);
640
0430ed48
NB
641 remove_partitions(dfd);
642
3da92f27 643 tst->ss->load_super(tst, dfd, NULL);
98dbd966 644 tst->ss->getinfo_super(tst, content);
3da92f27 645 tst->ss->free_super(tst);
da6b5ca9 646 close(dfd);
5787fa49
NB
647 }
648
811e6cbe
NB
649 stat(devname, &stb);
650
dab6685f 651 if (verbose > 0)
cd29a5c8 652 fprintf(stderr, Name ": %s is identified as a member of %s, slot %d.\n",
98dbd966 653 devname, mddev, content->disk.raid_disk);
64c4757e 654 devices[devcnt].devname = devname;
64c4757e 655 devices[devcnt].uptodate = 0;
98dbd966 656 devices[devcnt].i = *content;
213ee40b
NB
657 devices[devcnt].i.disk.major = major(stb.st_rdev);
658 devices[devcnt].i.disk.minor = minor(stb.st_rdev);
64c4757e 659 if (most_recent < devcnt) {
213ee40b
NB
660 if (devices[devcnt].i.events
661 > devices[most_recent].i.events)
64c4757e
NB
662 most_recent = devcnt;
663 }
98dbd966 664 if (content->array.level == -4)
5787fa49
NB
665 /* with multipath, the raid_disk from the superblock is meaningless */
666 i = devcnt;
667 else
213ee40b 668 i = devices[devcnt].i.disk.raid_disk;
308e1801 669 if (i+1 == 0) {
98dbd966
DW
670 if (nextspare < content->array.raid_disks)
671 nextspare = content->array.raid_disks;
308e1801 672 i = nextspare++;
08110d41 673 } else {
98dbd966 674 if (i >= content->array.raid_disks &&
08110d41
NB
675 i >= nextspare)
676 nextspare = i+1;
308e1801 677 }
98c6faba 678 if (i < 10000) {
56eedc1a 679 if (i >= bestcnt) {
98c6faba 680 unsigned int newbestcnt = i+10;
56eedc1a 681 int *newbest = malloc(sizeof(int)*newbestcnt);
98c6faba 682 unsigned int c;
56eedc1a
NB
683 for (c=0; c < newbestcnt; c++)
684 if (c < bestcnt)
685 newbest[c] = best[c];
686 else
687 newbest[c] = -1;
688 if (best)free(best);
689 best = newbest;
690 bestcnt = newbestcnt;
691 }
6a255b69 692 if (best[i] >=0 &&
213ee40b
NB
693 devices[best[i]].i.events
694 == devices[devcnt].i.events
695 && (devices[best[i]].i.disk.minor
696 != devices[devcnt].i.disk.minor)
b8ac1967 697 && st->ss == &super0
98dbd966 698 && content->array.level != LEVEL_MULTIPATH) {
6a255b69
NB
699 /* two different devices with identical superblock.
700 * Could be a mis-detection caused by overlapping
701 * partitions. fail-safe.
702 */
703 fprintf(stderr, Name ": WARNING %s and %s appear"
704 " to have very similar superblocks.\n"
705 " If they are really different, "
706 "please --zero the superblock on one\n"
d2df86e0
NB
707 " If they are the same or overlap,"
708 " please remove one from %s.\n",
709 devices[best[i]].devname, devname,
710 inargv ? "the list" :
711 "the\n DEVICE list in mdadm.conf"
712 );
7f91af49 713 close(mdfd);
6a255b69
NB
714 return 1;
715 }
52826846 716 if (best[i] == -1
213ee40b
NB
717 || (devices[best[i]].i.events
718 < devices[devcnt].i.events))
52826846 719 best[i] = devcnt;
56eedc1a 720 }
64c4757e 721 devcnt++;
df37ffc0 722 }
4b1ac34b 723
64c4757e 724 if (devcnt == 0) {
682c7051 725 fprintf(stderr, Name ": no devices found for %s\n",
64c4757e 726 mddev);
3d2b16e7
NB
727 if (st)
728 st->ss->free_super(st);
7f91af49 729 close(mdfd);
64c4757e
NB
730 return 1;
731 }
4b1ac34b 732
3d2b16e7
NB
733 if (update && strcmp(update, "byteorder")==0)
734 st->minor_version = 90;
735
98dbd966
DW
736 st->ss->getinfo_super(st, content);
737 clean = content->array.state & 1;
4b1ac34b 738
64c4757e
NB
739 /* now we have some devices that might be suitable.
740 * I wonder how many
741 */
98dbd966
DW
742 avail = malloc(content->array.raid_disks);
743 memset(avail, 0, content->array.raid_disks);
64c4757e 744 okcnt = 0;
52826846 745 sparecnt=0;
56eedc1a 746 for (i=0; i< bestcnt ;i++) {
64c4757e 747 int j = best[i];
ee04451c
NB
748 int event_margin = 1; /* always allow a difference of '1'
749 * like the kernel does
750 */
64c4757e 751 if (j < 0) continue;
d013a55e
NB
752 /* note: we ignore error flags in multipath arrays
753 * as they don't make sense
754 */
98dbd966 755 if (content->array.level != -4)
213ee40b
NB
756 if (!(devices[j].i.disk.state & (1<<MD_DISK_SYNC))) {
757 if (!(devices[j].i.disk.state
758 & (1<<MD_DISK_FAULTY)))
aa88f531 759 sparecnt++;
d013a55e 760 continue;
aa88f531 761 }
213ee40b
NB
762 if (devices[j].i.events+event_margin >=
763 devices[most_recent].i.events) {
64c4757e 764 devices[j].uptodate = 1;
98dbd966 765 if (i < content->array.raid_disks) {
52826846 766 okcnt++;
265e0f17
NB
767 avail[i]=1;
768 } else
52826846 769 sparecnt++;
64c4757e
NB
770 }
771 }
98dbd966
DW
772 while (force && !enough(content->array.level, content->array.raid_disks,
773 content->array.layout, 1,
265e0f17 774 avail, okcnt)) {
64c4757e
NB
775 /* Choose the newest best drive which is
776 * not up-to-date, update the superblock
777 * and add it.
778 */
52826846 779 int fd;
3da92f27 780 struct supertype *tst;
c5a6f9a6 781 long long current_events;
cd29a5c8 782 chosen_drive = -1;
98dbd966 783 for (i=0; i<content->array.raid_disks && i < bestcnt; i++) {
52826846
NB
784 int j = best[i];
785 if (j>=0 &&
786 !devices[j].uptodate &&
52826846 787 (chosen_drive < 0 ||
213ee40b
NB
788 devices[j].i.events
789 > devices[chosen_drive].i.events))
52826846
NB
790 chosen_drive = j;
791 }
792 if (chosen_drive < 0)
793 break;
213ee40b 794 current_events = devices[chosen_drive].i.events;
c5a6f9a6 795 add_another:
dab6685f
NB
796 if (verbose >= 0)
797 fprintf(stderr, Name ": forcing event count in %s(%d) from %d upto %d\n",
213ee40b
NB
798 devices[chosen_drive].devname,
799 devices[chosen_drive].i.disk.raid_disk,
800 (int)(devices[chosen_drive].i.events),
801 (int)(devices[most_recent].i.events));
8b0dabea 802 fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
52826846
NB
803 if (fd < 0) {
804 fprintf(stderr, Name ": Couldn't open %s for write - not updating\n",
805 devices[chosen_drive].devname);
213ee40b 806 devices[chosen_drive].i.events = 0;
52826846
NB
807 continue;
808 }
3da92f27 809 tst = dup_super(st);
60b435db 810 if (tst->ss->load_super(tst,fd, NULL)) {
52826846
NB
811 close(fd);
812 fprintf(stderr, Name ": RAID superblock disappeared from %s - not updating.\n",
813 devices[chosen_drive].devname);
213ee40b 814 devices[chosen_drive].i.events = 0;
52826846
NB
815 continue;
816 }
98dbd966
DW
817 content->events = devices[most_recent].i.events;
818 tst->ss->update_super(tst, content, "force-one",
67a8c82d
NB
819 devices[chosen_drive].devname, verbose,
820 0, NULL);
4b1ac34b 821
3da92f27 822 if (tst->ss->store_super(tst, fd)) {
52826846
NB
823 close(fd);
824 fprintf(stderr, Name ": Could not re-write superblock on %s\n",
825 devices[chosen_drive].devname);
213ee40b 826 devices[chosen_drive].i.events = 0;
3da92f27 827 tst->ss->free_super(tst);
52826846
NB
828 continue;
829 }
830 close(fd);
213ee40b 831 devices[chosen_drive].i.events = devices[most_recent].i.events;
52826846 832 devices[chosen_drive].uptodate = 1;
265e0f17 833 avail[chosen_drive] = 1;
52826846 834 okcnt++;
3da92f27 835 tst->ss->free_super(tst);
c5a6f9a6
NB
836
837 /* If there are any other drives of the same vintage,
838 * add them in as well. We can't lose and we might gain
839 */
98dbd966 840 for (i=0; i<content->array.raid_disks && i < bestcnt ; i++) {
c5a6f9a6
NB
841 int j = best[i];
842 if (j >= 0 &&
843 !devices[j].uptodate &&
213ee40b 844 devices[j].i.events == current_events) {
c5a6f9a6
NB
845 chosen_drive = j;
846 goto add_another;
847 }
848 }
64c4757e 849 }
52826846
NB
850
851 /* Now we want to look at the superblock which the kernel will base things on
852 * and compare the devices that we think are working with the devices that the
853 * superblock thinks are working.
854 * If there are differences and --force is given, then update this chosen
855 * superblock.
856 */
cd29a5c8 857 chosen_drive = -1;
3da92f27 858 st->ss->free_super(st);
56eedc1a 859 for (i=0; chosen_drive < 0 && i<bestcnt; i++) {
52826846
NB
860 int j = best[i];
861 int fd;
4b1ac34b 862
52826846
NB
863 if (j<0)
864 continue;
865 if (!devices[j].uptodate)
866 continue;
867 chosen_drive = j;
8b0dabea 868 if ((fd=dev_open(devices[j].devname, O_RDONLY|O_EXCL))< 0) {
52826846
NB
869 fprintf(stderr, Name ": Cannot open %s: %s\n",
870 devices[j].devname, strerror(errno));
7f91af49 871 close(mdfd);
52826846
NB
872 return 1;
873 }
3da92f27 874 if (st->ss->load_super(st,fd, NULL)) {
52826846
NB
875 close(fd);
876 fprintf(stderr, Name ": RAID superblock has disappeared from %s\n",
877 devices[j].devname);
7f91af49 878 close(mdfd);
52826846
NB
879 return 1;
880 }
881 close(fd);
882 }
3da92f27 883 if (st->sb == NULL) {
4b1ac34b 884 fprintf(stderr, Name ": No suitable drives found for %s\n", mddev);
7f91af49 885 close(mdfd);
4b1ac34b
NB
886 return 1;
887 }
98dbd966 888 st->ss->getinfo_super(st, content);
f35f2525 889#ifndef MDASSEMBLE
98dbd966 890 sysfs_init(content, mdfd, 0);
f35f2525 891#endif
56eedc1a 892 for (i=0; i<bestcnt; i++) {
52826846 893 int j = best[i];
98c6faba 894 unsigned int desired_state;
11a3e71d 895
98dbd966 896 if (i < content->array.raid_disks)
11a3e71d
NB
897 desired_state = (1<<MD_DISK_ACTIVE) | (1<<MD_DISK_SYNC);
898 else
899 desired_state = 0;
900
52826846
NB
901 if (j<0)
902 continue;
903 if (!devices[j].uptodate)
904 continue;
4b1ac34b 905
213ee40b 906 devices[j].i.disk.state = desired_state;
4e9a6ff7
N
907 if (!(devices[j].i.array.state & 1))
908 clean = 0;
213ee40b
NB
909
910 if (st->ss->update_super(st, &devices[j].i, "assemble", NULL,
68c7d6d7 911 verbose, 0, NULL)) {
52826846 912 if (force) {
dab6685f
NB
913 if (verbose >= 0)
914 fprintf(stderr, Name ": "
915 "clearing FAULTY flag for device %d in %s for %s\n",
916 j, mddev, devices[j].devname);
4b1ac34b 917 change = 1;
52826846 918 } else {
dab6685f
NB
919 if (verbose >= -1)
920 fprintf(stderr, Name ": "
921 "device %d in %s has wrong state in superblock, but %s seems ok\n",
922 i, mddev, devices[j].devname);
52826846
NB
923 }
924 }
4b1ac34b 925#if 0
213ee40b 926 if (!(super.disks[i].i.disk.state & (1 << MD_DISK_FAULTY))) {
52826846
NB
927 fprintf(stderr, Name ": devices %d of %s is not marked FAULTY in superblock, but cannot be found\n",
928 i, mddev);
929 }
4b1ac34b 930#endif
52826846 931 }
c5a6f9a6 932 if (force && !clean &&
98dbd966
DW
933 !enough(content->array.level, content->array.raid_disks,
934 content->array.layout, clean,
c5a6f9a6 935 avail, okcnt)) {
98dbd966 936 change += st->ss->update_super(st, content, "force-array",
67a8c82d
NB
937 devices[chosen_drive].devname, verbose,
938 0, NULL);
583315d9 939 clean = 1;
d013a55e 940 }
52826846 941
4b1ac34b 942 if (change) {
52826846 943 int fd;
8b0dabea 944 fd = dev_open(devices[chosen_drive].devname, O_RDWR|O_EXCL);
52826846 945 if (fd < 0) {
353632d9 946 fprintf(stderr, Name ": Could not open %s for write - cannot Assemble array.\n",
52826846 947 devices[chosen_drive].devname);
7f91af49 948 close(mdfd);
52826846
NB
949 return 1;
950 }
3da92f27 951 if (st->ss->store_super(st, fd)) {
52826846
NB
952 close(fd);
953 fprintf(stderr, Name ": Could not re-write superblock on %s\n",
954 devices[chosen_drive].devname);
7f91af49 955 close(mdfd);
52826846
NB
956 return 1;
957 }
958 close(fd);
52826846
NB
959 }
960
353632d9
NB
961 /* If we are in the middle of a reshape we may need to restore saved data
962 * that was moved aside due to the reshape overwriting live data
963 * The code of doing this lives in Grow.c
964 */
39c5a909 965#ifndef MDASSEMBLE
98dbd966 966 if (content->reshape_active) {
353632d9
NB
967 int err = 0;
968 int *fdlist = malloc(sizeof(int)* bestcnt);
969 for (i=0; i<bestcnt; i++) {
970 int j = best[i];
971 if (j >= 0) {
972 fdlist[i] = dev_open(devices[j].devname, O_RDWR|O_EXCL);
973 if (fdlist[i] < 0) {
974 fprintf(stderr, Name ": Could not open %s for write - cannot Assemble array.\n",
975 devices[j].devname);
976 err = 1;
977 break;
978 }
979 } else
980 fdlist[i] = -1;
981 }
982 if (!err)
98dbd966 983 err = Grow_restart(st, content, fdlist, bestcnt, backup_file);
353632d9
NB
984 while (i>0) {
985 i--;
986 if (fdlist[i]>=0) close(fdlist[i]);
987 }
988 if (err) {
989 fprintf(stderr, Name ": Failed to restore critical section for reshape, sorry.\n");
7f91af49 990 close(mdfd);
353632d9
NB
991 return err;
992 }
993 }
39c5a909 994#endif
aa88f531
NB
995 /* count number of in-sync devices according to the superblock.
996 * We must have this number to start the array without -s or -R
997 */
98dbd966 998 req_cnt = content->array.working_disks;
aa88f531 999
64c4757e
NB
1000 /* Almost ready to actually *do* something */
1001 if (!old_linux) {
fbf8a0b7 1002 int rv;
a19c88b8 1003
a04d5763
N
1004 /* First, fill in the map, so that udev can find our name
1005 * as soon as we become active.
1006 */
98dbd966
DW
1007 map_update(NULL, fd2devnum(mdfd), content->text_version,
1008 content->uuid, chosen_name);
a04d5763 1009
98dbd966 1010 rv = set_array_info(mdfd, st, content);
fbf8a0b7 1011 if (rv) {
f35f2525 1012 fprintf(stderr, Name ": failed to set array info for %s: %s\n",
64c4757e 1013 mddev, strerror(errno));
7f91af49 1014 close(mdfd);
64c4757e
NB
1015 return 1;
1016 }
11484a63 1017 if (ident->bitmap_fd >= 0) {
c82f047c
NB
1018 if (ioctl(mdfd, SET_BITMAP_FILE, ident->bitmap_fd) != 0) {
1019 fprintf(stderr, Name ": SET_BITMAP_FILE failed.\n");
7f91af49 1020 close(mdfd);
c82f047c
NB
1021 return 1;
1022 }
7ef02d01
NB
1023 } else if (ident->bitmap_file) {
1024 /* From config file */
1025 int bmfd = open(ident->bitmap_file, O_RDWR);
1026 if (bmfd < 0) {
1027 fprintf(stderr, Name ": Could not open bitmap file %s\n",
1028 ident->bitmap_file);
7f91af49 1029 close(mdfd);
7ef02d01
NB
1030 return 1;
1031 }
1032 if (ioctl(mdfd, SET_BITMAP_FILE, bmfd) != 0) {
1033 fprintf(stderr, Name ": Failed to set bitmapfile for %s\n", mddev);
1034 close(bmfd);
7f91af49 1035 close(mdfd);
7ef02d01
NB
1036 return 1;
1037 }
1038 close(bmfd);
c82f047c 1039 }
7ef02d01 1040
52826846 1041 /* First, add the raid disks, but add the chosen one last */
56eedc1a 1042 for (i=0; i<= bestcnt; i++) {
52826846 1043 int j;
56eedc1a 1044 if (i < bestcnt) {
52826846
NB
1045 j = best[i];
1046 if (j == chosen_drive)
1047 continue;
1048 } else
1049 j = chosen_drive;
1050
56eedc1a 1051 if (j >= 0 /* && devices[j].uptodate */) {
98dbd966 1052 rv = add_disk(mdfd, st, content, &devices[j].i);
7801ac20 1053
a19c88b8 1054 if (rv) {
213ee40b
NB
1055 fprintf(stderr, Name ": failed to add "
1056 "%s to %s: %s\n",
64c4757e
NB
1057 devices[j].devname,
1058 mddev,
1059 strerror(errno));
98dbd966 1060 if (i < content->array.raid_disks
213ee40b 1061 || i == bestcnt)
52826846
NB
1062 okcnt--;
1063 else
1064 sparecnt--;
dab6685f 1065 } else if (verbose > 0)
213ee40b
NB
1066 fprintf(stderr, Name ": added %s "
1067 "to %s as %d\n",
1068 devices[j].devname, mddev,
1069 devices[j].i.disk.raid_disk);
98dbd966 1070 } else if (verbose > 0 && i < content->array.raid_disks)
213ee40b
NB
1071 fprintf(stderr, Name ": no uptodate device for "
1072 "slot %d of %s\n",
64c4757e
NB
1073 i, mddev);
1074 }
aba69144 1075
98dbd966 1076 if (content->array.level == LEVEL_CONTAINER) {
598f0d58
NB
1077 if (verbose >= 0) {
1078 fprintf(stderr, Name ": Container %s has been "
1079 "assembled with %d drive%s",
57ed8c91 1080 mddev, okcnt+sparecnt, okcnt+sparecnt==1?"":"s");
98dbd966 1081 if (okcnt < content->array.raid_disks)
598f0d58 1082 fprintf(stderr, " (out of %d)",
98dbd966 1083 content->array.raid_disks);
598f0d58
NB
1084 fprintf(stderr, "\n");
1085 }
98dbd966 1086 sysfs_uevent(content, "change");
a7c6e3fb 1087 wait_for(chosen_name, mdfd);
7f91af49 1088 close(mdfd);
598f0d58
NB
1089 return 0;
1090 }
1091
64c4757e 1092 if (runstop == 1 ||
b8a8ccf9 1093 (runstop <= 0 &&
98dbd966
DW
1094 ( enough(content->array.level, content->array.raid_disks,
1095 content->array.layout, clean, avail, okcnt) &&
aa88f531
NB
1096 (okcnt >= req_cnt || start_partial_ok)
1097 ))) {
82b27616 1098 if (ioctl(mdfd, RUN_ARRAY, NULL)==0) {
dab6685f
NB
1099 if (verbose >= 0) {
1100 fprintf(stderr, Name ": %s has been started with %d drive%s",
1101 mddev, okcnt, okcnt==1?"":"s");
98dbd966
DW
1102 if (okcnt < content->array.raid_disks)
1103 fprintf(stderr, " (out of %d)", content->array.raid_disks);
dab6685f
NB
1104 if (sparecnt)
1105 fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
1106 fprintf(stderr, ".\n");
1107 }
8a659c33
N
1108 if (content->reshape_active &&
1109 content->array.level >= 4 &&
1110 content->array.level <= 6) {
acee8e89
N
1111 /* might need to increase the size
1112 * of the stripe cache - default is 256
1113 */
8a659c33 1114 if (256 < 4 * (content->array.chunk_size/4096)) {
acee8e89
N
1115 struct mdinfo *sra = sysfs_read(mdfd, 0, 0);
1116 if (sra)
1117 sysfs_set_num(sra, NULL,
1118 "stripe_cache_size",
8a659c33 1119 (4 * content->array.chunk_size / 4096) + 1);
acee8e89
N
1120 }
1121 }
a7c6e3fb 1122 wait_for(mddev, mdfd);
7f91af49
N
1123 close(mdfd);
1124 if (auto_assem) {
589395d6 1125 int usecs = 1;
589395d6
NB
1126 /* There is a nasty race with 'mdadm --monitor'.
1127 * If it opens this device before we close it,
1128 * it gets an incomplete open on which IO
598f0d58
NB
1129 * doesn't work and the capacity is
1130 * wrong.
589395d6
NB
1131 * If we reopen (to check for layered devices)
1132 * before --monitor closes, we loose.
1133 *
1134 * So: wait upto 1 second for there to be
1135 * a non-zero capacity.
1136 */
1137 while (usecs < 1000) {
1138 mdfd = open(mddev, O_RDONLY);
1139 if (mdfd >= 0) {
beae1dfe
NB
1140 unsigned long long size;
1141 if (get_dev_size(mdfd, NULL, &size) &&
589395d6
NB
1142 size > 0)
1143 break;
1144 close(mdfd);
1145 }
1146 usleep(usecs);
1147 usecs <<= 1;
1148 }
1149 }
64c4757e 1150 return 0;
82b27616 1151 }
682c7051 1152 fprintf(stderr, Name ": failed to RUN_ARRAY %s: %s\n",
64c4757e 1153 mddev, strerror(errno));
583315d9 1154
98dbd966
DW
1155 if (!enough(content->array.level, content->array.raid_disks,
1156 content->array.layout, 1, avail, okcnt))
583315d9
NB
1157 fprintf(stderr, Name ": Not enough devices to "
1158 "start the array.\n");
98dbd966
DW
1159 else if (!enough(content->array.level,
1160 content->array.raid_disks,
1161 content->array.layout, clean,
583315d9
NB
1162 avail, okcnt))
1163 fprintf(stderr, Name ": Not enough devices to "
1164 "start the array while not clean "
1165 "- consider --force.\n");
1166
7f91af49 1167 if (auto_assem)
1c203a4b 1168 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1169 close(mdfd);
64c4757e
NB
1170 return 1;
1171 }
82b27616 1172 if (runstop == -1) {
b8a8ccf9 1173 fprintf(stderr, Name ": %s assembled from %d drive%s",
52826846 1174 mddev, okcnt, okcnt==1?"":"s");
98dbd966
DW
1175 if (okcnt != content->array.raid_disks)
1176 fprintf(stderr, " (out of %d)", content->array.raid_disks);
b8a8ccf9 1177 fprintf(stderr, ", but not started.\n");
7f91af49 1178 close(mdfd);
64c4757e 1179 return 0;
82b27616 1180 }
4855f95c 1181 if (verbose >= -1) {
dab6685f
NB
1182 fprintf(stderr, Name ": %s assembled from %d drive%s", mddev, okcnt, okcnt==1?"":"s");
1183 if (sparecnt)
1184 fprintf(stderr, " and %d spare%s", sparecnt, sparecnt==1?"":"s");
98dbd966
DW
1185 if (!enough(content->array.level, content->array.raid_disks,
1186 content->array.layout, 1, avail, okcnt))
dab6685f 1187 fprintf(stderr, " - not enough to start the array.\n");
98dbd966
DW
1188 else if (!enough(content->array.level,
1189 content->array.raid_disks,
1190 content->array.layout, clean,
583315d9
NB
1191 avail, okcnt))
1192 fprintf(stderr, " - not enough to start the "
1193 "array while not clean - consider "
1194 "--force.\n");
dab6685f 1195 else {
98dbd966 1196 if (req_cnt == content->array.raid_disks)
dab6685f
NB
1197 fprintf(stderr, " - need all %d to start it", req_cnt);
1198 else
98dbd966 1199 fprintf(stderr, " - need %d of %d to start", req_cnt, content->array.raid_disks);
dab6685f
NB
1200 fprintf(stderr, " (use --run to insist).\n");
1201 }
aa88f531 1202 }
7f91af49 1203 if (auto_assem)
1c203a4b 1204 ioctl(mdfd, STOP_ARRAY, NULL);
56a8da69 1205 close(mdfd);
64c4757e 1206 return 1;
82b27616 1207 } else {
52826846
NB
1208 /* The "chosen_drive" is a good choice, and if necessary, the superblock has
1209 * been updated to point to the current locations of devices.
1210 * so we can just start the array
82b27616 1211 */
cd29a5c8 1212 unsigned long dev;
213ee40b
NB
1213 dev = makedev(devices[chosen_drive].i.disk.major,
1214 devices[chosen_drive].i.disk.minor);
82b27616
NB
1215 if (ioctl(mdfd, START_ARRAY, dev)) {
1216 fprintf(stderr, Name ": Cannot start array: %s\n",
1217 strerror(errno));
1218 }
aba69144 1219
64c4757e 1220 }
7f91af49 1221 close(mdfd);
e0d19036 1222 return 0;
64c4757e 1223}
6234c63c
DW
1224
1225#ifndef MDASSEMBLE
1226int assemble_container_content(struct supertype *st, int mdfd,
1227 struct mdinfo *content, int runstop,
1228 char *chosen_name, int verbose)
1229{
1230 struct mdinfo *dev, *sra;
1231 int working = 0, preexist = 0;
1232 struct map_ent *map = NULL;
1233
1234 sysfs_init(content, mdfd, 0);
1235
1236 sra = sysfs_read(mdfd, 0, GET_VERSION);
1237 if (sra == NULL || strcmp(sra->text_version, content->text_version) != 0)
4408ee76
N
1238 if (sysfs_set_array(content, md_get_version(mdfd)) != 0) {
1239 close(mdfd);
6234c63c 1240 return 1;
4408ee76 1241 }
6234c63c
DW
1242 if (sra)
1243 sysfs_free(sra);
1244
1245 for (dev = content->devs; dev; dev = dev->next)
462906cd 1246 if (sysfs_add_disk(content, dev, 1) == 0)
6234c63c
DW
1247 working++;
1248 else if (errno == EEXIST)
1249 preexist++;
4408ee76
N
1250 if (working == 0) {
1251 close(mdfd);
7cb2aa33 1252 return 1;/* Nothing new, don't try to start */
8b4e5ea9
N
1253 }
1254
1255 map_update(&map, fd2devnum(mdfd),
1256 content->text_version,
1257 content->uuid, chosen_name);
1258
1259 if (runstop > 0 ||
6234c63c 1260 (working + preexist) >= content->array.working_disks) {
bb50e5d3 1261 int err;
a714580e 1262
6234c63c
DW
1263 switch(content->array.level) {
1264 case LEVEL_LINEAR:
1265 case LEVEL_MULTIPATH:
1266 case 0:
bb50e5d3
N
1267 err = sysfs_set_str(content, NULL, "array_state",
1268 "active");
6234c63c
DW
1269 break;
1270 default:
bb50e5d3 1271 err = sysfs_set_str(content, NULL, "array_state",
6234c63c
DW
1272 "readonly");
1273 /* start mdmon if needed. */
bb50e5d3
N
1274 if (!err) {
1275 if (!mdmon_running(st->container_dev))
1276 start_mdmon(st->container_dev);
1277 ping_monitor(devnum2devname(st->container_dev));
1278 }
6234c63c
DW
1279 break;
1280 }
bb50e5d3
N
1281 if (!err)
1282 sysfs_set_safemode(content, content->safe_mode_delay);
6234c63c 1283 if (verbose >= 0) {
bb50e5d3
N
1284 if (err)
1285 fprintf(stderr, Name
1286 ": array %s now has %d devices",
1287 chosen_name, working + preexist);
1288 else
1289 fprintf(stderr, Name
1290 ": Started %s with %d devices",
1291 chosen_name, working + preexist);
6234c63c
DW
1292 if (preexist)
1293 fprintf(stderr, " (%d new)", working);
1294 fprintf(stderr, "\n");
1295 }
bb50e5d3 1296 if (!err)
a7c6e3fb 1297 wait_for(chosen_name, mdfd);
4408ee76 1298 close(mdfd);
7cb2aa33 1299 return 0;
6234c63c 1300 /* FIXME should have an O_EXCL and wait for read-auto */
7cb2aa33 1301 } else {
6234c63c
DW
1302 if (verbose >= 0)
1303 fprintf(stderr, Name
1304 ": %s assembled with %d devices but "
1305 "not started\n",
1306 chosen_name, working);
4408ee76 1307 close(mdfd);
7cb2aa33
N
1308 return 1;
1309 }
6234c63c
DW
1310}
1311#endif
1312