]> git.ipfire.org Git - thirdparty/mdadm.git/blame - Create.c
Factor common code into new "start_mdmon".
[thirdparty/mdadm.git] / Create.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"
682c7051
NB
31#include "md_u.h"
32#include "md_p.h"
60248f74 33#include <ctype.h>
64c4757e 34
82d9eba6 35int Create(struct supertype *st, char *mddev, int mdfd,
5dd497ee 36 int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
3d3dd91e 37 char *name, char *homehost, int *uuid,
cd29a5c8 38 int subdevs, mddev_dev_t devlist,
47d79ef8 39 int runstop, int verbose, int force, int assume_clean,
dfd4d8ee 40 char *bitmap_file, int bitmap_chunk, int write_behind, int delay)
64c4757e 41{
682c7051
NB
42 /*
43 * Create a new raid array.
44 *
45 * First check that necessary details are available
46 * (i.e. level, raid-disks)
47 *
48 * Then check each disk to see what might be on it
49 * and report anything interesting.
50 *
51 * If anything looks odd, and runstop not set,
52 * abort.
53 *
54 * SET_ARRAY_INFO and ADD_NEW_DISK, and
82d9eba6 55 * if runstop==run, or raiddisks disks were used,
682c7051
NB
56 * RUN_ARRAY
57 */
b5e64645 58 unsigned long long minsize=0, maxsize=0;
cd29a5c8
NB
59 char *mindisc = NULL;
60 char *maxdisc = NULL;
c913b90e 61 int dnum;
cd29a5c8 62 mddev_dev_t dv;
682c7051 63 int fail=0, warn=0;
82b27616 64 struct stat stb;
82d9eba6 65 int first_missing = subdevs * 2;
6fb79233 66 int second_missing = subdevs * 2;
52826846 67 int missing_disks = 0;
82d9eba6 68 int insert_point = subdevs * 2; /* where to insert a missing drive */
d2ca6449 69 int total_slots;
4b1ac34b 70 int pass;
82d9eba6
NB
71 int vers;
72 int rv;
c82f047c 73 int bitmap_fd;
5f8097be 74 int have_container = 0;
a931db9e
NB
75 int container_fd;
76 int need_mdmon = 0;
f9c25f1d 77 unsigned long long bitmapsize;
d03373f1 78 struct mdinfo *sra;
d2ca6449 79 struct mdinfo info, *infos;
b8ac1967 80 int did_default = 0;
682c7051 81
943eafef 82 int major_num = BITMAP_MAJOR_HI;
dcec9ee5 83
06c7f68e 84 memset(&info, 0, sizeof(info));
29e766a5 85
82d9eba6
NB
86 vers = md_get_version(mdfd);
87 if (vers < 9000) {
dd0781e5 88 fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n");
52826846 89 return 1;
32e5a4ee
DL
90 } else {
91 mdu_array_info_t inf;
92 memset(&inf, 0, sizeof(inf));
93 ioctl(mdfd, GET_ARRAY_INFO, &inf);
94 if (inf.working_disks != 0) {
95 fprintf(stderr, Name ": another array by this name"
96 " is already running.\n");
97 return 1;
98 }
682c7051 99 }
a322f70c 100 if (level == UnSet) {
3cfe6247 101 /* "ddf" and "imsm" metadata only supports one level - should possibly
a322f70c
DW
102 * push this into metadata handler??
103 */
3cfe6247 104 if (st && (st->ss == &super_ddf || st->ss == &super_imsm))
a322f70c
DW
105 level = LEVEL_CONTAINER;
106 }
107
98c6faba 108 if (level == UnSet) {
682c7051
NB
109 fprintf(stderr,
110 Name ": a RAID level is needed to create an array.\n");
111 return 1;
112 }
98c6faba
NB
113 if (raiddisks < 4 && level == 6) {
114 fprintf(stderr,
115 Name ": at least 4 raid-devices needed for level 6\n");
116 return 1;
117 }
118 if (raiddisks > 256 && level == 6) {
119 fprintf(stderr,
120 Name ": no more than 256 raid-devices supported for level 6\n");
121 return 1;
122 }
52826846
NB
123 if (raiddisks < 2 && level >= 4) {
124 fprintf(stderr,
98c6faba 125 Name ": at least 2 raid-devices needed for level 4 or 5\n");
52826846
NB
126 return 1;
127 }
570510ba
NB
128 if (level <= 0 && sparedisks) {
129 fprintf(stderr,
130 Name ": This level does not support spare devices\n");
131 return 1;
132 }
5f8097be
NB
133
134 if (subdevs == 1 && strcmp(devlist->devname, "missing") != 0) {
135 /* If given a single device, it might be a container, and we can
136 * extract a device list from there
137 */
138 mdu_array_info_t inf;
139 int fd;
140
141 memset(&inf, 0, sizeof(inf));
142 fd = open(devlist->devname, O_RDONLY, 0);
143 if (fd >= 0 &&
144 ioctl(fd, GET_ARRAY_INFO, &inf) == 0 &&
145 inf.raid_disks == 0) {
146 /* yep, looks like a container */
147 if (st) {
148 rv = st->ss->load_super(st, fd,
149 devlist->devname);
150 if (rv == 0)
151 have_container = 1;
152 } else {
153 st = guess_super(fd);
154 if (st && !(rv = st->ss->
155 load_super(st, fd,
156 devlist->devname)))
157 have_container = 1;
158 else
159 st = NULL;
160 }
161 }
162 if (fd >= 0)
163 close(fd);
164 if (have_container) {
165 subdevs = 0;
166 devlist = NULL;
167 }
168 }
682c7051 169 if (subdevs > raiddisks+sparedisks) {
b83d95f3 170 fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks);
52826846 171 return 1;
682c7051 172 }
5f8097be 173 if (!have_container && subdevs < raiddisks+sparedisks) {
52826846
NB
174 fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n");
175 return 1;
176 }
5b28bd56
NB
177 if (bitmap_file && level <= 0) {
178 fprintf(stderr, Name ": bitmaps not meaningful with level %s\n",
179 map_num(pers, level)?:"given");
180 return 1;
181 }
52826846 182
682c7051 183 /* now set some defaults */
98c6faba 184 if (layout == UnSet)
682c7051
NB
185 switch(level) {
186 default: /* no layout */
187 layout = 0;
188 break;
e5329c37
NB
189 case 10:
190 layout = 0x102; /* near=2, far=1 */
dab6685f 191 if (verbose > 0)
e5329c37
NB
192 fprintf(stderr,
193 Name ": layout defaults to n1\n");
194 break;
682c7051 195 case 5:
98c6faba 196 case 6:
682c7051 197 layout = map_name(r5layout, "default");
dab6685f 198 if (verbose > 0)
682c7051
NB
199 fprintf(stderr,
200 Name ": layout defaults to %s\n", map_num(r5layout, layout));
201 break;
b5e64645
NB
202 case LEVEL_FAULTY:
203 layout = map_name(faultylayout, "default");
204
dab6685f 205 if (verbose > 0)
b5e64645
NB
206 fprintf(stderr,
207 Name ": layout defaults to %s\n", map_num(faultylayout, layout));
208 break;
682c7051
NB
209 }
210
e5329c37
NB
211 if (level == 10)
212 /* check layout fits in array*/
213 if ((layout&255) * ((layout>>8)&255) > raiddisks) {
214 fprintf(stderr, Name ": that layout requires at least %d devices\n",
215 (layout&255) * ((layout>>8)&255));
216 return 1;
217 }
218
aa88f531
NB
219 switch(level) {
220 case 4:
221 case 5:
e5329c37 222 case 10:
98c6faba 223 case 6:
aa88f531 224 case 0:
570510ba 225 case LEVEL_LINEAR: /* linear */
aa88f531
NB
226 if (chunk == 0) {
227 chunk = 64;
dab6685f 228 if (verbose > 0)
aa88f531
NB
229 fprintf(stderr, Name ": chunk size defaults to 64K\n");
230 }
231 break;
570510ba 232 case 1:
08e43379 233 case LEVEL_FAULTY:
570510ba 234 case LEVEL_MULTIPATH:
17f25ca6 235 case LEVEL_CONTAINER:
aa88f531
NB
236 if (chunk) {
237 chunk = 0;
dab6685f 238 if (verbose > 0)
aa88f531
NB
239 fprintf(stderr, Name ": chunk size ignored for this level\n");
240 }
241 break;
570510ba
NB
242 default:
243 fprintf(stderr, Name ": unknown level %d\n", level);
244 return 1;
682c7051
NB
245 }
246
17f25ca6 247 if (st && ! st->ss->validate_geometry(st, level, layout, raiddisks,
2c514b71 248 chunk, size, NULL, NULL, verbose>=0))
17f25ca6
NB
249 return 1;
250
682c7051 251 /* now look at the subdevs */
06c7f68e
NB
252 info.array.active_disks = 0;
253 info.array.working_disks = 0;
c913b90e
NB
254 dnum = 0;
255 for (dv=devlist; dv; dv=dv->next, dnum++) {
cd29a5c8 256 char *dname = dv->devname;
17f25ca6 257 unsigned long long freesize;
cd29a5c8 258 if (strcasecmp(dname, "missing")==0) {
c913b90e
NB
259 if (first_missing > dnum)
260 first_missing = dnum;
6fb79233
NB
261 if (second_missing > dnum && dnum > first_missing)
262 second_missing = dnum;
52826846
NB
263 missing_disks ++;
264 continue;
265 }
06c7f68e 266 info.array.working_disks++;
c913b90e 267 if (dnum < raiddisks)
06c7f68e 268 info.array.active_disks++;
058574b1 269 if (st == NULL) {
8aec876d 270 struct createinfo *ci = conf_get_create_info();
058574b1
NB
271 if (ci)
272 st = ci->supertype;
273 }
576d6d83
NB
274 if (st == NULL) {
275 /* Need to choose a default metadata, which is different
17f25ca6 276 * depending on geometry of array.
576d6d83
NB
277 */
278 int i;
279 char *name = "default";
17f25ca6 280 for(i=0; !st && superlist[i]; i++) {
576d6d83 281 st = superlist[i]->match_metadata_desc(name);
17f25ca6
NB
282 if (st && !st->ss->validate_geometry
283 (st, level, layout, raiddisks,
2c514b71
NB
284 chunk, size, dname, &freesize,
285 verbose > 0))
17f25ca6
NB
286 st = NULL;
287 }
576d6d83
NB
288
289 if (!st) {
17f25ca6
NB
290 fprintf(stderr, Name ": device %s not suitable "
291 "for any style of array\n",
292 dname);
576d6d83
NB
293 exit(2);
294 }
b8ac1967 295 if (st->ss != &super0 ||
576d6d83 296 st->minor_version != 90)
b8ac1967 297 did_default = 1;
17f25ca6
NB
298 } else {
299 if (!st->ss->validate_geometry(st, level, layout,
300 raiddisks,
301 chunk, size, dname,
2c514b71
NB
302 &freesize,
303 verbose > 0)) {
17f25ca6
NB
304
305 fprintf(stderr,
306 Name ": %s is not suitable for "
307 "this array.\n",
308 dname);
309 fail = 1;
310 continue;
311 }
682c7051 312 }
82d9eba6
NB
313
314 freesize /= 2; /* convert to K */
d2cd3ffc
NB
315 if (chunk) {
316 /* round to chunk size */
317 freesize = freesize & ~(chunk-1);
318 }
682c7051
NB
319
320 if (size && freesize < size) {
52826846 321 fprintf(stderr, Name ": %s is smaller that given size."
17f25ca6
NB
322 " %lluK < %lluK + metadata\n",
323 dname, freesize, size);
52826846 324 fail = 1;
52826846 325 continue;
682c7051 326 }
cd29a5c8
NB
327 if (maxdisc == NULL || (maxdisc && freesize > maxsize)) {
328 maxdisc = dname;
52826846 329 maxsize = freesize;
682c7051 330 }
cd29a5c8
NB
331 if (mindisc ==NULL || (mindisc && freesize < minsize)) {
332 mindisc = dname;
52826846 333 minsize = freesize;
682c7051 334 }
dab6685f 335 if (runstop != 1 || verbose >= 0) {
b6e63da4
NB
336 int fd = open(dname, O_RDONLY, 0);
337 if (fd <0 ) {
338 fprintf(stderr, Name ": Cannot open %s: %s\n",
339 dname, strerror(errno));
340 fail=1;
341 continue;
342 }
dab6685f
NB
343 warn |= check_ext2(fd, dname);
344 warn |= check_reiser(fd, dname);
345 warn |= check_raid(fd, dname);
b6e63da4 346 close(fd);
dab6685f 347 }
682c7051
NB
348 }
349 if (fail) {
52826846
NB
350 fprintf(stderr, Name ": create aborted\n");
351 return 1;
682c7051
NB
352 }
353 if (size == 0) {
5f8097be 354 if (mindisc == NULL && !have_container) {
52826846
NB
355 fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
356 return 1;
357 }
5f8097be
NB
358 if (level > 0 || level == LEVEL_MULTIPATH
359 || level == LEVEL_FAULTY
b8ac1967 360 || st->ss->external ) {
b5e64645 361 /* size is meaningful */
b8ac1967
NB
362 if (!st->ss->validate_geometry(st, level, layout,
363 raiddisks,
364 chunk, minsize,
2c514b71 365 NULL, NULL, 0)) {
aba69144 366 fprintf(stderr, Name ": devices too large for RAID level %d\n", level);
b5e64645
NB
367 return 1;
368 }
369 size = minsize;
dab6685f 370 if (verbose > 0)
5dd497ee 371 fprintf(stderr, Name ": size set to %lluK\n", size);
b5e64645 372 }
682c7051 373 }
b5e64645 374 if (level > 0 && ((maxsize-size)*100 > maxsize)) {
dab6685f 375 if (runstop != 1 || verbose >= 0)
5dd497ee 376 fprintf(stderr, Name ": largest drive (%s) exceed size (%lluK) by more than 1%%\n",
dab6685f 377 maxdisc, size);
52826846 378 warn = 1;
682c7051
NB
379 }
380
381 if (warn) {
52826846
NB
382 if (runstop!= 1) {
383 if (!ask("Continue creating array? ")) {
384 fprintf(stderr, Name ": create aborted.\n");
385 return 1;
386 }
387 } else {
dab6685f 388 if (verbose > 0)
52826846 389 fprintf(stderr, Name ": creation continuing despite oddities due to --run\n");
682c7051 390 }
682c7051
NB
391 }
392
66f8bbbe 393 /* If this is raid4/5, we want to configure the last active slot
52826846 394 * as missing, so that a reconstruct happens (faster than re-parity)
98c6faba 395 * FIX: Can we do this for raid6 as well?
52826846 396 */
47d79ef8 397 if (assume_clean==0 && force == 0 && first_missing >= raiddisks) {
98c6faba 398 switch ( level ) {
66f8bbbe 399 case 4:
98c6faba
NB
400 case 5:
401 insert_point = raiddisks-1;
402 sparedisks++;
06c7f68e 403 info.array.active_disks--;
98c6faba
NB
404 missing_disks++;
405 break;
406 default:
407 break;
408 }
52826846 409 }
6fb79233
NB
410 /* For raid6, if creating with 1 missing drive, make a good drive
411 * into a spare, else the create will fail
412 */
413 if (assume_clean == 0 && force == 0 && first_missing < raiddisks &&
414 second_missing >= raiddisks && level == 6) {
415 insert_point = raiddisks - 1;
416 if (insert_point == first_missing)
417 insert_point--;
418 sparedisks ++;
419 info.array.active_disks--;
420 missing_disks++;
421 }
570510ba 422
5f8097be 423 if (level <= 0 && first_missing < subdevs * 2) {
570510ba
NB
424 fprintf(stderr,
425 Name ": This level does not support missing devices\n");
426 return 1;
427 }
aba69144 428
682c7051
NB
429 /* Ok, lets try some ioctls */
430
06c7f68e
NB
431 info.array.level = level;
432 info.array.size = size;
433 info.array.raid_disks = raiddisks;
82b27616
NB
434 /* The kernel should *know* what md_minor we are dealing
435 * with, but it chooses to trust me instead. Sigh
436 */
06c7f68e 437 info.array.md_minor = 0;
82b27616 438 if (fstat(mdfd, &stb)==0)
06c7f68e
NB
439 info.array.md_minor = minor(stb.st_rdev);
440 info.array.not_persistent = 0;
6fb79233 441
66f8bbbe 442 if ( ( (level == 4 || level == 5) &&
b5e64645
NB
443 (insert_point < raiddisks || first_missing < raiddisks) )
444 ||
6fb79233
NB
445 ( level == 6 && (insert_point < raiddisks
446 || second_missing < raiddisks))
47d79ef8 447 ||
d024b0a7
NB
448 ( level <= 0 )
449 ||
47d79ef8 450 assume_clean
103f2410 451 ) {
06c7f68e 452 info.array.state = 1; /* clean, but one+ drive will be missing*/
103f2410
NB
453 info.resync_start = ~0ULL;
454 } else {
06c7f68e 455 info.array.state = 0; /* not clean, but no errors */
103f2410
NB
456 info.resync_start = 0;
457 }
f9c25f1d
NB
458 if (level == 10) {
459 /* for raid10, the bitmap size is the capacity of the array,
460 * which is array.size * raid_disks / ncopies;
461 * .. but convert to sectors.
462 */
702b557b 463 int ncopies = ((layout>>8) & 255) * (layout & 255);
5dd497ee
NB
464 bitmapsize = (unsigned long long)size * raiddisks / ncopies * 2;
465/* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, size, raiddisks, ncopies);*/
f9c25f1d 466 } else
5dd497ee 467 bitmapsize = (unsigned long long)size * 2;
f9c25f1d 468
82b27616
NB
469 /* There is lots of redundancy in these disk counts,
470 * raid_disks is the most meaningful value
471 * it describes the geometry of the array
472 * it is constant
473 * nr_disks is total number of used slots.
474 * it should be raid_disks+spare_disks
475 * spare_disks is the number of extra disks present
476 * see above
477 * active_disks is the number of working disks in
478 * active slots. (With raid_disks)
479 * working_disks is the total number of working disks,
480 * including spares
481 * failed_disks is the number of disks marked failed
482 *
483 * Ideally, the kernel would keep these (except raid_disks)
484 * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
485 * So for now, we assume that all raid and spare
486 * devices will be given.
487 */
06c7f68e
NB
488 info.array.spare_disks=sparedisks;
489 info.array.failed_disks=missing_disks;
490 info.array.nr_disks = info.array.working_disks
491 + info.array.failed_disks;
492 info.array.layout = layout;
493 info.array.chunk_size = chunk*1024;
82d9eba6 494
b3b33eb5
NB
495 if (name == NULL || *name == 0) {
496 /* base name on mddev */
d1e80164
NB
497 /* /dev/md0 -> 0
498 * /dev/md_d0 -> d0
499 * /dev/md/1 -> 1
500 * /dev/md/d1 -> d1
501 * /dev/md/home -> home
502 * /dev/mdhome -> home
503 */
b3b33eb5
NB
504 name = strrchr(mddev, '/');
505 if (name) {
506 name++;
60248f74
NB
507 if (strncmp(name, "md_d", 4)==0 &&
508 strlen(name) > 4 &&
509 isdigit(name[4]) &&
b3b33eb5 510 (name-mddev) == 5 /* /dev/ */)
d1e80164 511 name += 3;
60248f74
NB
512 else if (strncmp(name, "md", 2)==0 &&
513 strlen(name) > 2 &&
514 isdigit(name[2]) &&
515 (name-mddev) == 5 /* /dev/ */)
b3b33eb5
NB
516 name += 2;
517 }
518 }
06c7f68e 519 if (!st->ss->init_super(st, &info.array, size, name, homehost, uuid))
82d9eba6 520 return 1;
682c7051 521
d2ca6449 522 total_slots = info.array.nr_disks;
159c3a1a
NB
523 st->ss->getinfo_super(st, &info);
524
f7f1b6a1
NB
525 if (did_default && verbose >= 0) {
526 if (info.text_version[0] == '/') {
527 int dnum = devname2devnum(info.text_version+1);
528 char *path;
529 int mdp = get_mdp_major();
530 struct mdinfo *mdi;
531 if (dnum > 0)
532 path = map_dev(MD_MAJOR, dnum, 1);
533 else
534 path = map_dev(mdp, (-1-dnum)<< 6, 1);
535
536 mdi = sysfs_read(-1, dnum, GET_VERSION);
537
538 fprintf(stderr, Name ": Creating array inside "
539 "%s container %s\n",
540 mdi?mdi->text_version:"managed", path);
541 sysfs_free(mdi);
542 } else
543 fprintf(stderr, Name ": Defaulting to version"
544 " %s metadata\n", info.text_version);
545 }
b8ac1967 546
dcec9ee5 547 if (bitmap_file && vers < 9003) {
943eafef 548 major_num = BITMAP_MAJOR_HOSTENDIAN;
dcec9ee5
NB
549#ifdef __BIG_ENDIAN
550 fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n"
551 " between different architectured. Consider upgrading the Linux kernel.\n");
552#endif
553 }
554
55935d51
NB
555 if (bitmap_file && strcmp(bitmap_file, "internal")==0) {
556 if ((vers%100) < 2) {
557 fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n");
558 return 1;
559 }
3da92f27 560 if (!st->ss->add_internal_bitmap(st, &bitmap_chunk,
199171a2 561 delay, write_behind,
943eafef 562 bitmapsize, 1, major_num)) {
55935d51
NB
563 fprintf(stderr, Name ": Given bitmap chunk size not supported.\n");
564 return 1;
565 }
566 bitmap_file = NULL;
567 }
568
569
d03373f1 570 sra = sysfs_read(mdfd, 0, 0);
55935d51 571
d03373f1
NB
572 if (st->ss->external) {
573 char ver[100];
dd15dc4a
NB
574 strcat(strcpy(ver, "external:"),
575 info.text_version);
f7e7067b 576 if (st->ss->external && st->subarray[0]) {
2f6079dc 577 /* member */
dd15dc4a 578
a931db9e
NB
579 /* When creating a member, we need to be careful
580 * to negotiate with mdmon properly.
581 * If it is already running, we cannot write to
582 * the devices and must ask it to do that part.
583 * If it isn't running, we write to the devices,
584 * and then start it.
585 * We hold an exclusive open on the container
586 * device to make sure mdmon doesn't exit after
587 * we checked that it is running.
588 *
589 * For now, fail if it is already running.
590 */
591 container_fd = open_dev_excl(st->container_dev);
592 if (container_fd < 0) {
593 fprintf(stderr, Name ": Cannot get exclusive "
594 "open on container - weird.\n");
595 return 1;
596 }
597 if (mdmon_running(st->container_dev)) {
edd8d13c
NB
598 if (verbose)
599 fprintf(stderr, Name ": reusing mdmon "
600 "for %s.\n",
601 devnum2devname(st->container_dev));
602 st->update_tail = &st->updates;
603 } else
604 need_mdmon = 1;
2f6079dc 605 }
d03373f1
NB
606 if ((vers % 100) < 2 ||
607 sra == NULL ||
608 sysfs_set_str(sra, NULL, "metadata_version",
609 ver) < 0) {
610 fprintf(stderr, Name ": This kernel does not "
611 "support external metadata.\n");
612 return 1;
613 }
2503d23b 614 rv = sysfs_set_array(sra, &info);
d03373f1 615 } else if ((vers % 100) >= 1) { /* can use different versions */
82d9eba6
NB
616 mdu_array_info_t inf;
617 memset(&inf, 0, sizeof(inf));
b8ac1967
NB
618 inf.major_version = info.array.major_version;
619 inf.minor_version = info.array.minor_version;
82d9eba6 620 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
aba69144 621 } else
82d9eba6
NB
622 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
623 if (rv) {
52826846
NB
624 fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
625 mddev, strerror(errno));
626 return 1;
682c7051 627 }
52826846 628
c82f047c
NB
629 if (bitmap_file) {
630 int uuid[4];
55935d51 631
3da92f27 632 st->ss->uuid_from_super(st, uuid);
dfd4d8ee 633 if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk,
dcec9ee5 634 delay, write_behind,
f9c25f1d 635 bitmapsize,
943eafef 636 major_num)) {
c82f047c
NB
637 return 1;
638 }
639 bitmap_fd = open(bitmap_file, O_RDWR);
640 if (bitmap_fd < 0) {
641 fprintf(stderr, Name ": weird: %s cannot be openned\n",
dcec9ee5 642 bitmap_file);
c82f047c
NB
643 return 1;
644 }
645 if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) {
646 fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n",
647 mddev, strerror(errno));
648 return 1;
649 }
650 }
651
d2ca6449 652 infos = malloc(sizeof(*infos) * total_slots);
4b1ac34b 653
4b1ac34b
NB
654 for (pass=1; pass <=2 ; pass++) {
655 mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */
656
657 for (dnum=0, dv = devlist ; dv ;
658 dv=(dv->next)?(dv->next):moved_disk, dnum++) {
659 int fd;
660 struct stat stb;
d2ca6449 661 struct mdinfo *inf = &infos[dnum];
4b1ac34b 662
d2ca6449
NB
663 if (dnum >= total_slots)
664 abort();
4b1ac34b
NB
665 if (dnum == insert_point) {
666 moved_disk = dv;
52826846 667 }
4b1ac34b 668 if (dnum == insert_point ||
111d01fc
NB
669 strcasecmp(dv->devname, "missing")==0)
670 continue;
671
d2ca6449
NB
672 switch(pass) {
673 case 1:
674 *inf = info;
5f8097be 675
d2ca6449
NB
676 inf->disk.number = dnum;
677 inf->disk.raid_disk = dnum;
678 if (inf->disk.raid_disk < raiddisks)
679 inf->disk.state = (1<<MD_DISK_ACTIVE) |
680 (1<<MD_DISK_SYNC);
681 else
682 inf->disk.state = 0;
683
684 if (dv->writemostly)
685 inf->disk.state |= (1<<MD_DISK_WRITEMOSTLY);
686
687 if (st->ss->external && st->subarray[0])
688 fd = open(dv->devname, O_RDWR, 0);
689 else
690 fd = open(dv->devname, O_RDWR|O_EXCL,0);
691
692 if (fd < 0) {
693 fprintf(stderr, Name ": failed to open %s "
694 "after earlier success - aborting\n",
695 dv->devname);
696 return 1;
697 }
698 fstat(fd, &stb);
699 inf->disk.major = major(stb.st_rdev);
700 inf->disk.minor = minor(stb.st_rdev);
111d01fc 701
111d01fc 702 remove_partitions(fd);
d2ca6449 703 st->ss->add_to_super(st, &inf->disk,
111d01fc 704 fd, dv->devname);
d2ca6449
NB
705 st->ss->getinfo_super(st, inf);
706
707 /* getinfo_super might have lost these ... */
708 inf->disk.major = major(stb.st_rdev);
709 inf->disk.minor = minor(stb.st_rdev);
4b1ac34b
NB
710 break;
711 case 2:
d2ca6449 712 inf->errors = 0;
2503d23b 713 rv = 0;
4b1ac34b 714
d2ca6449
NB
715 if (st->ss->external)
716 rv = sysfs_add_disk(sra, inf);
717 else
2503d23b 718 rv = ioctl(mdfd, ADD_NEW_DISK,
d2ca6449
NB
719 &inf->disk);
720
2503d23b 721 if (rv) {
d03373f1
NB
722 fprintf(stderr,
723 Name ": ADD_NEW_DISK for %s "
724 "failed: %s\n",
4b1ac34b 725 dv->devname, strerror(errno));
3da92f27 726 st->ss->free_super(st);
4b1ac34b
NB
727 return 1;
728 }
4b1ac34b
NB
729 break;
730 }
731 if (dv == moved_disk && dnum != insert_point) break;
52826846 732 }
edd8d13c 733 if (pass == 1) {
111d01fc 734 st->ss->write_init_super(st);
edd8d13c
NB
735 flush_metadata_updates(st);
736 }
5e52ae9e 737 }
d2ca6449 738 free(infos);
3da92f27 739 st->ss->free_super(st);
5e52ae9e 740
682c7051 741 /* param is not actually used */
17f25ca6
NB
742 if (level == LEVEL_CONTAINER)
743 /* No need to start */
744 ;
745 else if (runstop == 1 || subdevs >= raiddisks) {
598f0d58
NB
746 if (st->ss->external) {
747 switch(level) {
748 case LEVEL_LINEAR:
749 case LEVEL_MULTIPATH:
750 case 0:
751 sysfs_set_str(sra, NULL, "array_state",
752 "active");
a931db9e 753 need_mdmon = 0;
598f0d58
NB
754 break;
755 default:
756 sysfs_set_str(sra, NULL, "array_state",
757 "readonly");
758 break;
759 }
760 } else {
761 mdu_param_t param;
762 if (ioctl(mdfd, RUN_ARRAY, &param)) {
763 fprintf(stderr, Name ": RUN_ARRAY failed: %s\n",
764 strerror(errno));
765 Manage_runstop(mddev, mdfd, -1, 0);
766 return 1;
767 }
682c7051 768 }
dab6685f
NB
769 if (verbose >= 0)
770 fprintf(stderr, Name ": array %s started.\n", mddev);
f7e7067b 771 if (st->ss->external && st->subarray[0]) {
8850ee3e
N
772 if (need_mdmon)
773 start_mdmon(st->container_dev);
774
775 ping_monitor(devnum2devname(st->container_dev));
a931db9e
NB
776 close(container_fd);
777 }
682c7051 778 } else {
b83d95f3 779 fprintf(stderr, Name ": not starting array - not enough devices.\n");
682c7051
NB
780 }
781 return 0;
64c4757e 782}