]> git.ipfire.org Git - thirdparty/mdadm.git/blame - Create.c
Assemble: fix call to wait_for
[thirdparty/mdadm.git] / Create.c
CommitLineData
64c4757e 1/*
9a9dab36 2 * mdadm - manage Linux "md" devices aka RAID arrays.
64c4757e 3 *
ca3b6696 4 * Copyright (C) 2001-2012 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
ca3b6696 25#include "mdadm.h"
682c7051
NB
26#include "md_u.h"
27#include "md_p.h"
60248f74 28#include <ctype.h>
64c4757e 29
a18a888e
DW
30static int default_layout(struct supertype *st, int level, int verbose)
31{
32 int layout = UnSet;
33
30f58b22
DW
34 if (st && st->ss->default_geometry)
35 st->ss->default_geometry(st, &level, &layout, NULL);
a18a888e
DW
36
37 if (layout == UnSet)
38 switch(level) {
39 default: /* no layout */
40 layout = 0;
41 break;
42 case 10:
43 layout = 0x102; /* near=2, far=1 */
44 if (verbose > 0)
e7b84f9d 45 pr_err("layout defaults to n2\n");
a18a888e
DW
46 break;
47 case 5:
48 case 6:
49 layout = map_name(r5layout, "default");
50 if (verbose > 0)
e7b84f9d 51 pr_err("layout defaults to %s\n", map_num(r5layout, layout));
a18a888e
DW
52 break;
53 case LEVEL_FAULTY:
54 layout = map_name(faultylayout, "default");
55
56 if (verbose > 0)
e7b84f9d 57 pr_err("layout defaults to %s\n", map_num(faultylayout, layout));
a18a888e
DW
58 break;
59 }
60
61 return layout;
62}
63
7f91af49 64int Create(struct supertype *st, char *mddev,
171dccc8 65 char *name, int *uuid,
a655e550 66 int subdevs, struct mddev_dev *devlist,
99cc42f4 67 struct shape *s,
40c9a66a 68 struct context *c, unsigned long long data_offset)
64c4757e 69{
682c7051
NB
70 /*
71 * Create a new raid array.
72 *
73 * First check that necessary details are available
74 * (i.e. level, raid-disks)
75 *
76 * Then check each disk to see what might be on it
77 * and report anything interesting.
78 *
79 * If anything looks odd, and runstop not set,
80 * abort.
81 *
82 * SET_ARRAY_INFO and ADD_NEW_DISK, and
82d9eba6 83 * if runstop==run, or raiddisks disks were used,
682c7051
NB
84 * RUN_ARRAY
85 */
7f91af49 86 int mdfd;
b5e64645 87 unsigned long long minsize=0, maxsize=0;
cd29a5c8
NB
88 char *mindisc = NULL;
89 char *maxdisc = NULL;
c913b90e 90 int dnum;
a655e550 91 struct mddev_dev *dv;
682c7051 92 int fail=0, warn=0;
82b27616 93 struct stat stb;
82d9eba6 94 int first_missing = subdevs * 2;
6fb79233 95 int second_missing = subdevs * 2;
52826846 96 int missing_disks = 0;
82d9eba6 97 int insert_point = subdevs * 2; /* where to insert a missing drive */
d2ca6449 98 int total_slots;
4b1ac34b 99 int pass;
82d9eba6
NB
100 int vers;
101 int rv;
c82f047c 102 int bitmap_fd;
5f8097be 103 int have_container = 0;
0e600426 104 int container_fd = -1;
a931db9e 105 int need_mdmon = 0;
f9c25f1d 106 unsigned long long bitmapsize;
d2ca6449 107 struct mdinfo info, *infos;
b8ac1967 108 int did_default = 0;
a18a888e 109 int do_default_layout = 0;
c21e737b 110 int do_default_chunk = 0;
8ed3e5e1 111 unsigned long safe_mode_delay = 0;
a04d5763 112 char chosen_name[1024];
ad5bc697 113 struct map_ent *map = NULL;
8592f29d 114 unsigned long long newsize;
682c7051 115
943eafef 116 int major_num = BITMAP_MAJOR_HI;
dcec9ee5 117
06c7f68e 118 memset(&info, 0, sizeof(info));
99cc42f4
N
119 if (s->level == UnSet && st && st->ss->default_geometry)
120 st->ss->default_geometry(st, &s->level, NULL, NULL);
121 if (s->level == UnSet) {
e7b84f9d 122 pr_err("a RAID level is needed to create an array.\n");
682c7051
NB
123 return 1;
124 }
99cc42f4 125 if (s->raiddisks < 4 && s->level == 6) {
e7b84f9d 126 pr_err("at least 4 raid-devices needed for level 6\n");
98c6faba
NB
127 return 1;
128 }
99cc42f4 129 if (s->raiddisks > 256 && s->level == 6) {
e7b84f9d 130 pr_err("no more than 256 raid-devices supported for level 6\n");
98c6faba
NB
131 return 1;
132 }
99cc42f4 133 if (s->raiddisks < 2 && s->level >= 4) {
e7b84f9d 134 pr_err("at least 2 raid-devices needed for level 4 or 5\n");
52826846
NB
135 return 1;
136 }
99cc42f4 137 if (s->level <= 0 && s->sparedisks) {
e7b84f9d 138 pr_err("This level does not support spare devices\n");
570510ba
NB
139 return 1;
140 }
5f8097be
NB
141
142 if (subdevs == 1 && strcmp(devlist->devname, "missing") != 0) {
143 /* If given a single device, it might be a container, and we can
144 * extract a device list from there
145 */
146 mdu_array_info_t inf;
147 int fd;
148
149 memset(&inf, 0, sizeof(inf));
37ea3936 150 fd = open(devlist->devname, O_RDONLY);
5f8097be
NB
151 if (fd >= 0 &&
152 ioctl(fd, GET_ARRAY_INFO, &inf) == 0 &&
153 inf.raid_disks == 0) {
154 /* yep, looks like a container */
155 if (st) {
0fb69d1d
N
156 rv = st->ss->load_container(st, fd,
157 devlist->devname);
5f8097be
NB
158 if (rv == 0)
159 have_container = 1;
160 } else {
0fb69d1d 161 st = super_by_fd(fd, NULL);
5f8097be 162 if (st && !(rv = st->ss->
0fb69d1d
N
163 load_container(st, fd,
164 devlist->devname)))
5f8097be
NB
165 have_container = 1;
166 else
167 st = NULL;
168 }
18fde300 169 if (have_container) {
99cc42f4 170 subdevs = s->raiddisks;
18fde300
DW
171 first_missing = subdevs * 2;
172 second_missing = subdevs * 2;
173 insert_point = subdevs * 2;
174 }
5f8097be
NB
175 }
176 if (fd >= 0)
177 close(fd);
5f8097be 178 }
99cc42f4 179 if (st && st->ss->external && s->sparedisks) {
e7b84f9d
N
180 pr_err("This metadata type does not support "
181 "spare disks at create time\n");
ffcfc735
N
182 return 1;
183 }
99cc42f4
N
184 if (subdevs > s->raiddisks+s->sparedisks) {
185 pr_err("You have listed more devices (%d) than are in the array(%d)!\n", subdevs, s->raiddisks+s->sparedisks);
52826846 186 return 1;
682c7051 187 }
99cc42f4 188 if (!have_container && subdevs < s->raiddisks+s->sparedisks) {
e7b84f9d 189 pr_err("You haven't given enough devices (real or missing) to create this array\n");
52826846
NB
190 return 1;
191 }
99cc42f4 192 if (s->bitmap_file && s->level <= 0) {
e7b84f9d 193 pr_err("bitmaps not meaningful with level %s\n",
99cc42f4 194 map_num(pers, s->level)?:"given");
5b28bd56
NB
195 return 1;
196 }
52826846 197
682c7051 198 /* now set some defaults */
b5e64645 199
99cc42f4 200 if (s->layout == UnSet) {
a18a888e 201 do_default_layout = 1;
99cc42f4 202 s->layout = default_layout(st, s->level, c->verbose);
a18a888e 203 }
682c7051 204
99cc42f4 205 if (s->level == 10)
e5329c37 206 /* check layout fits in array*/
99cc42f4 207 if ((s->layout&255) * ((s->layout>>8)&255) > s->raiddisks) {
e7b84f9d 208 pr_err("that layout requires at least %d devices\n",
99cc42f4 209 (s->layout&255) * ((s->layout>>8)&255));
e5329c37
NB
210 return 1;
211 }
212
99cc42f4 213 switch(s->level) {
aa88f531
NB
214 case 4:
215 case 5:
e5329c37 216 case 10:
98c6faba 217 case 6:
aa88f531 218 case 0:
99cc42f4
N
219 if (s->chunk == 0 || s->chunk == UnSet) {
220 s->chunk = UnSet;
c21e737b
CA
221 do_default_chunk = 1;
222 /* chunk will be set later */
bb7295f1 223 }
5f175898
N
224 break;
225 case LEVEL_LINEAR:
226 /* a chunksize of zero 0s perfectly valid (and preferred) since 2.6.16 */
99cc42f4
N
227 if (get_linux_version() < 2006016 && s->chunk == 0) {
228 s->chunk = 64;
171dccc8 229 if (c->verbose > 0)
e7b84f9d 230 pr_err("chunk size defaults to 64K\n");
aa88f531
NB
231 }
232 break;
570510ba 233 case 1:
08e43379 234 case LEVEL_FAULTY:
570510ba 235 case LEVEL_MULTIPATH:
17f25ca6 236 case LEVEL_CONTAINER:
99cc42f4
N
237 if (s->chunk) {
238 s->chunk = 0;
171dccc8 239 if (c->verbose > 0)
e7b84f9d 240 pr_err("chunk size ignored for this level\n");
aa88f531
NB
241 }
242 break;
570510ba 243 default:
99cc42f4 244 pr_err("unknown level %d\n", s->level);
570510ba 245 return 1;
682c7051 246 }
99cc42f4 247 if (s->size == MAX_SIZE)
d04f65f4 248 /* use '0' to mean 'max' now... */
99cc42f4
N
249 s->size = 0;
250 if (s->size && s->chunk && s->chunk != UnSet)
251 s->size &= ~(unsigned long long)(s->chunk - 1);
252 newsize = s->size * 2;
253 if (st && ! st->ss->validate_geometry(st, s->level, s->layout, s->raiddisks,
af4348dd 254 &s->chunk, s->size*2,
40c9a66a 255 data_offset, NULL,
af4348dd 256 &newsize, c->verbose>=0))
17f25ca6 257 return 1;
bb7295f1 258
99cc42f4
N
259 if (s->chunk && s->chunk != UnSet) {
260 newsize &= ~(unsigned long long)(s->chunk*2 - 1);
64385908
CA
261 if (do_default_chunk) {
262 /* default chunk was just set */
171dccc8 263 if (c->verbose > 0)
e7b84f9d 264 pr_err("chunk size "
99cc42f4
N
265 "defaults to %dK\n", s->chunk);
266 s->size &= ~(unsigned long long)(s->chunk - 1);
64385908
CA
267 do_default_chunk = 0;
268 }
bb7295f1 269 }
64385908 270
99cc42f4
N
271 if (s->size == 0) {
272 s->size = newsize / 2;
273 if (s->level == 1)
ae6c05ad
N
274 /* If this is ever reshaped to RAID5, we will
275 * need a chunksize. So round it off a bit
276 * now just to be safe
277 */
99cc42f4 278 s->size &= ~(64ULL-1);
ae6c05ad 279
99cc42f4
N
280 if (s->size && c->verbose > 0)
281 pr_err("setting size to %lluK\n", s->size);
8592f29d 282 }
17f25ca6 283
682c7051 284 /* now look at the subdevs */
06c7f68e
NB
285 info.array.active_disks = 0;
286 info.array.working_disks = 0;
c913b90e 287 dnum = 0;
8592f29d 288 for (dv=devlist; dv && !have_container; dv=dv->next, dnum++) {
cd29a5c8 289 char *dname = dv->devname;
17f25ca6 290 unsigned long long freesize;
e9b11fee 291 int dfd;
72ca9bcf 292 char *doff;
e9b11fee 293
cd29a5c8 294 if (strcasecmp(dname, "missing")==0) {
c913b90e
NB
295 if (first_missing > dnum)
296 first_missing = dnum;
6fb79233
NB
297 if (second_missing > dnum && dnum > first_missing)
298 second_missing = dnum;
52826846
NB
299 missing_disks ++;
300 continue;
301 }
72ca9bcf
N
302 if (data_offset != VARIABLE_OFFSET) {
303 doff = strchr(dname, ':');
304 if (doff) {
305 *doff++ = 0;
306 dv->data_offset = parse_size(doff);
307 } else
308 dv->data_offset = INVALID_SECTORS;
309 } else
310 dv->data_offset = data_offset;
311
58b3c697 312 dfd = open(dname, O_RDONLY);
e9b11fee 313 if (dfd < 0) {
e7b84f9d 314 pr_err("cannot open %s: %s\n",
e9b11fee
N
315 dname, strerror(errno));
316 exit(2);
317 }
318 if (fstat(dfd, &stb) != 0 ||
319 (stb.st_mode & S_IFMT) != S_IFBLK) {
320 close(dfd);
e7b84f9d 321 pr_err("%s is not a block device\n",
e9b11fee
N
322 dname);
323 exit(2);
324 }
325 close(dfd);
06c7f68e 326 info.array.working_disks++;
99cc42f4 327 if (dnum < s->raiddisks)
06c7f68e 328 info.array.active_disks++;
058574b1 329 if (st == NULL) {
8aec876d 330 struct createinfo *ci = conf_get_create_info();
058574b1
NB
331 if (ci)
332 st = ci->supertype;
333 }
576d6d83
NB
334 if (st == NULL) {
335 /* Need to choose a default metadata, which is different
17f25ca6 336 * depending on geometry of array.
576d6d83
NB
337 */
338 int i;
339 char *name = "default";
17f25ca6 340 for(i=0; !st && superlist[i]; i++) {
576d6d83 341 st = superlist[i]->match_metadata_desc(name);
ecbd9e81
N
342 if (!st)
343 continue;
a18a888e 344 if (do_default_layout)
99cc42f4 345 s->layout = default_layout(st, s->level, c->verbose);
ecbd9e81 346 switch (st->ss->validate_geometry(
99cc42f4 347 st, s->level, s->layout, s->raiddisks,
af4348dd 348 &s->chunk, s->size*2,
72ca9bcf 349 dv->data_offset, dname,
af4348dd 350 &freesize, c->verbose > 0)) {
ecbd9e81
N
351 case -1: /* Not valid, message printed, and not
352 * worth checking any further */
353 exit(2);
354 break;
355 case 0: /* Geometry not valid */
55425f27 356 free(st);
17f25ca6 357 st = NULL;
99cc42f4 358 s->chunk = do_default_chunk ? UnSet : s->chunk;
ecbd9e81
N
359 break;
360 case 1: /* All happy */
361 break;
55425f27 362 }
17f25ca6 363 }
576d6d83
NB
364
365 if (!st) {
58b3c697
N
366 int dfd = open(dname, O_RDONLY|O_EXCL);
367 if (dfd < 0) {
e7b84f9d 368 pr_err("cannot open %s: %s\n",
58b3c697
N
369 dname, strerror(errno));
370 exit(2);
371 }
e7b84f9d 372 pr_err("device %s not suitable "
17f25ca6
NB
373 "for any style of array\n",
374 dname);
576d6d83
NB
375 exit(2);
376 }
b8ac1967 377 if (st->ss != &super0 ||
576d6d83 378 st->minor_version != 90)
b8ac1967 379 did_default = 1;
17f25ca6 380 } else {
a18a888e 381 if (do_default_layout)
99cc42f4
N
382 s->layout = default_layout(st, s->level, 0);
383 if (!st->ss->validate_geometry(st, s->level, s->layout,
384 s->raiddisks,
af4348dd 385 &s->chunk, s->size*2,
72ca9bcf 386 dv->data_offset,
af4348dd 387 dname, &freesize,
171dccc8 388 c->verbose >= 0)) {
17f25ca6 389
e7b84f9d
N
390 pr_err("%s is not suitable for "
391 "this array.\n",
392 dname);
17f25ca6
NB
393 fail = 1;
394 continue;
395 }
682c7051 396 }
82d9eba6
NB
397
398 freesize /= 2; /* convert to K */
99cc42f4 399 if (s->chunk && s->chunk != UnSet) {
d2cd3ffc 400 /* round to chunk size */
99cc42f4 401 freesize = freesize & ~(s->chunk-1);
64385908
CA
402 if (do_default_chunk) {
403 /* default chunk was just set */
171dccc8 404 if (c->verbose > 0)
e7b84f9d 405 pr_err("chunk size "
99cc42f4
N
406 "defaults to %dK\n", s->chunk);
407 s->size &= ~(unsigned long long)(s->chunk - 1);
64385908
CA
408 do_default_chunk = 0;
409 }
d2cd3ffc 410 }
682c7051 411
99cc42f4 412 if (s->size && freesize < s->size) {
e7b84f9d 413 pr_err("%s is smaller than given size."
17f25ca6 414 " %lluK < %lluK + metadata\n",
99cc42f4 415 dname, freesize, s->size);
52826846 416 fail = 1;
52826846 417 continue;
682c7051 418 }
cd29a5c8
NB
419 if (maxdisc == NULL || (maxdisc && freesize > maxsize)) {
420 maxdisc = dname;
52826846 421 maxsize = freesize;
682c7051 422 }
cd29a5c8
NB
423 if (mindisc ==NULL || (mindisc && freesize < minsize)) {
424 mindisc = dname;
52826846 425 minsize = freesize;
682c7051 426 }
171dccc8 427 if (c->runstop != 1 || c->verbose >= 0) {
37ea3936 428 int fd = open(dname, O_RDONLY);
b6e63da4 429 if (fd <0 ) {
e7b84f9d 430 pr_err("Cannot open %s: %s\n",
b6e63da4
NB
431 dname, strerror(errno));
432 fail=1;
433 continue;
434 }
dab6685f
NB
435 warn |= check_ext2(fd, dname);
436 warn |= check_reiser(fd, dname);
437 warn |= check_raid(fd, dname);
034b203a
TM
438 if (strcmp(st->ss->name, "1.x") == 0 &&
439 st->minor_version >= 1)
440 /* metadata at front */
53ed6ac3 441 warn |= check_partitions(fd, dname, 0, 0);
99cc42f4
N
442 else if (s->level == 1 || s->level == LEVEL_CONTAINER
443 || (s->level == 0 && s->raiddisks == 1))
034b203a 444 /* partitions could be meaningful */
99cc42f4 445 warn |= check_partitions(fd, dname, freesize*2, s->size*2);
034b203a
TM
446 else
447 /* partitions cannot be meaningful */
53ed6ac3 448 warn |= check_partitions(fd, dname, 0, 0);
034b203a 449 if (strcmp(st->ss->name, "1.x") == 0 &&
a0962fe9
N
450 st->minor_version >= 1 &&
451 did_default &&
99cc42f4 452 s->level == 1 &&
034b203a
TM
453 (warn & 1024) == 0) {
454 warn |= 1024;
e7b84f9d 455 pr_err("Note: this array has metadata at the start and\n"
a0962fe9 456 " may not be suitable as a boot device. If you plan to\n"
cc86f89c 457 " store '/boot' on this device please ensure that\n"
a0962fe9 458 " your boot-loader understands md/v1.x metadata, or use\n"
cc86f89c 459 " --metadata=0.90\n");
a0962fe9 460 }
b6e63da4 461 close(fd);
dab6685f 462 }
682c7051 463 }
99cc42f4 464 if (s->raiddisks + s->sparedisks > st->max_devs) {
e7b84f9d 465 pr_err("Too many devices:"
acab7bb1
N
466 " %s metadata only supports %d\n",
467 st->ss->name, st->max_devs);
468 return 1;
469 }
8592f29d 470 if (have_container)
99cc42f4 471 info.array.working_disks = s->raiddisks;
682c7051 472 if (fail) {
e7b84f9d 473 pr_err("create aborted\n");
52826846 474 return 1;
682c7051 475 }
99cc42f4 476 if (s->size == 0) {
5f8097be 477 if (mindisc == NULL && !have_container) {
e7b84f9d 478 pr_err("no size and no drives given - aborting create.\n");
52826846
NB
479 return 1;
480 }
99cc42f4
N
481 if (s->level > 0 || s->level == LEVEL_MULTIPATH
482 || s->level == LEVEL_FAULTY
b8ac1967 483 || st->ss->external ) {
b5e64645 484 /* size is meaningful */
99cc42f4
N
485 if (!st->ss->validate_geometry(st, s->level, s->layout,
486 s->raiddisks,
487 &s->chunk, minsize*2,
40c9a66a 488 data_offset,
2c514b71 489 NULL, NULL, 0)) {
99cc42f4 490 pr_err("devices too large for RAID level %d\n", s->level);
b5e64645
NB
491 return 1;
492 }
99cc42f4
N
493 s->size = minsize;
494 if (s->level == 1)
ae6c05ad
N
495 /* If this is ever reshaped to RAID5, we will
496 * need a chunksize. So round it off a bit
497 * now just to be safe
498 */
99cc42f4 499 s->size &= ~(64ULL-1);
171dccc8 500 if (c->verbose > 0)
99cc42f4 501 pr_err("size set to %lluK\n", s->size);
b5e64645 502 }
682c7051 503 }
99cc42f4 504 if (!have_container && s->level > 0 && ((maxsize-s->size)*100 > maxsize)) {
171dccc8 505 if (c->runstop != 1 || c->verbose >= 0)
e7b84f9d 506 pr_err("largest drive (%s) exceeds size (%lluK) by more than 1%%\n",
99cc42f4 507 maxdisc, s->size);
52826846 508 warn = 1;
682c7051
NB
509 }
510
9eafa1de 511 if (st->ss->detail_platform && st->ss->detail_platform(0, 1, NULL) != 0) {
171dccc8 512 if (c->runstop != 1 || c->verbose >= 0)
e7b84f9d 513 pr_err("%s unable to enumerate platform support\n"
5615172f
DW
514 " array may not be compatible with hardware/firmware\n",
515 st->ss->name);
516 warn = 1;
517 }
518
682c7051 519 if (warn) {
171dccc8 520 if (c->runstop!= 1) {
52826846 521 if (!ask("Continue creating array? ")) {
e7b84f9d 522 pr_err("create aborted.\n");
52826846
NB
523 return 1;
524 }
525 } else {
171dccc8 526 if (c->verbose > 0)
e7b84f9d 527 pr_err("creation continuing despite oddities due to --run\n");
682c7051 528 }
682c7051
NB
529 }
530
66f8bbbe 531 /* If this is raid4/5, we want to configure the last active slot
52826846 532 * as missing, so that a reconstruct happens (faster than re-parity)
98c6faba 533 * FIX: Can we do this for raid6 as well?
52826846 534 */
87b47257 535 if (st->ss->external == 0 &&
99cc42f4
N
536 s->assume_clean==0 && c->force == 0 && first_missing >= s->raiddisks) {
537 switch ( s->level ) {
66f8bbbe 538 case 4:
98c6faba 539 case 5:
99cc42f4
N
540 insert_point = s->raiddisks-1;
541 s->sparedisks++;
06c7f68e 542 info.array.active_disks--;
98c6faba
NB
543 missing_disks++;
544 break;
545 default:
546 break;
547 }
52826846 548 }
6fb79233
NB
549 /* For raid6, if creating with 1 missing drive, make a good drive
550 * into a spare, else the create will fail
551 */
99cc42f4 552 if (s->assume_clean == 0 && c->force == 0 && first_missing < s->raiddisks &&
ffcfc735 553 st->ss->external == 0 &&
99cc42f4
N
554 second_missing >= s->raiddisks && s->level == 6) {
555 insert_point = s->raiddisks - 1;
6fb79233
NB
556 if (insert_point == first_missing)
557 insert_point--;
99cc42f4 558 s->sparedisks ++;
6fb79233
NB
559 info.array.active_disks--;
560 missing_disks++;
561 }
570510ba 562
99cc42f4 563 if (s->level <= 0 && first_missing < subdevs * 2) {
e7b84f9d 564 pr_err("This level does not support missing devices\n");
570510ba
NB
565 return 1;
566 }
aba69144 567
7f91af49 568 /* We need to create the device */
ad5bc697 569 map_lock(&map);
171dccc8 570 mdfd = create_mddev(mddev, name, c->autof, LOCAL, chosen_name);
e06af9dd
JS
571 if (mdfd < 0) {
572 map_unlock(&map);
7f91af49 573 return 1;
e06af9dd 574 }
3e9df86a
AK
575 /* verify if chosen_name is not in use,
576 * it could be in conflict with already existing device
577 * e.g. container, array
578 */
579 if (strncmp(chosen_name, "/dev/md/", 8) == 0
580 && map_by_name(&map, chosen_name+8) != NULL) {
e7b84f9d 581 pr_err("Array name %s is in use already.\n",
3e9df86a
AK
582 chosen_name);
583 close(mdfd);
584 map_unlock(&map);
585 return 1;
586 }
f2be09f1 587 mddev = chosen_name;
7f91af49
N
588
589 vers = md_get_version(mdfd);
590 if (vers < 9000) {
e7b84f9d 591 pr_err("Create requires md driver version 0.90.0 or later\n");
e06af9dd 592 goto abort_locked;
7f91af49
N
593 } else {
594 mdu_array_info_t inf;
595 memset(&inf, 0, sizeof(inf));
596 ioctl(mdfd, GET_ARRAY_INFO, &inf);
597 if (inf.working_disks != 0) {
e7b84f9d 598 pr_err("another array by this name"
7f91af49 599 " is already running.\n");
e06af9dd 600 goto abort_locked;
7f91af49
N
601 }
602 }
a04d5763 603
682c7051
NB
604 /* Ok, lets try some ioctls */
605
99cc42f4
N
606 info.array.level = s->level;
607 info.array.size = s->size;
608 info.array.raid_disks = s->raiddisks;
82b27616
NB
609 /* The kernel should *know* what md_minor we are dealing
610 * with, but it chooses to trust me instead. Sigh
611 */
06c7f68e 612 info.array.md_minor = 0;
82b27616 613 if (fstat(mdfd, &stb)==0)
06c7f68e
NB
614 info.array.md_minor = minor(stb.st_rdev);
615 info.array.not_persistent = 0;
6fb79233 616
99cc42f4
N
617 if ( ( (s->level == 4 || s->level == 5) &&
618 (insert_point < s->raiddisks || first_missing < s->raiddisks) )
b5e64645 619 ||
99cc42f4
N
620 ( s->level == 6 && (insert_point < s->raiddisks
621 || second_missing < s->raiddisks))
47d79ef8 622 ||
99cc42f4 623 ( s->level <= 0 )
d024b0a7 624 ||
99cc42f4 625 s->assume_clean
103f2410 626 ) {
06c7f68e 627 info.array.state = 1; /* clean, but one+ drive will be missing*/
b7528a20 628 info.resync_start = MaxSector;
103f2410 629 } else {
06c7f68e 630 info.array.state = 0; /* not clean, but no errors */
103f2410
NB
631 info.resync_start = 0;
632 }
99cc42f4 633 if (s->level == 10) {
f9c25f1d
NB
634 /* for raid10, the bitmap size is the capacity of the array,
635 * which is array.size * raid_disks / ncopies;
636 * .. but convert to sectors.
637 */
99cc42f4
N
638 int ncopies = ((s->layout>>8) & 255) * (s->layout & 255);
639 bitmapsize = s->size * s->raiddisks / ncopies * 2;
640/* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, s->size, s->raiddisks, ncopies);*/
f9c25f1d 641 } else
99cc42f4 642 bitmapsize = s->size * 2;
f9c25f1d 643
82b27616
NB
644 /* There is lots of redundancy in these disk counts,
645 * raid_disks is the most meaningful value
646 * it describes the geometry of the array
647 * it is constant
648 * nr_disks is total number of used slots.
649 * it should be raid_disks+spare_disks
650 * spare_disks is the number of extra disks present
651 * see above
652 * active_disks is the number of working disks in
653 * active slots. (With raid_disks)
654 * working_disks is the total number of working disks,
655 * including spares
656 * failed_disks is the number of disks marked failed
657 *
5d500228 658 * Ideally, the kernel would keep these (except raid_disks)
82b27616
NB
659 * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
660 * So for now, we assume that all raid and spare
661 * devices will be given.
662 */
99cc42f4 663 info.array.spare_disks=s->sparedisks;
06c7f68e
NB
664 info.array.failed_disks=missing_disks;
665 info.array.nr_disks = info.array.working_disks
666 + info.array.failed_disks;
99cc42f4
N
667 info.array.layout = s->layout;
668 info.array.chunk_size = s->chunk*1024;
82d9eba6 669
b3b33eb5
NB
670 if (name == NULL || *name == 0) {
671 /* base name on mddev */
d1e80164
NB
672 /* /dev/md0 -> 0
673 * /dev/md_d0 -> d0
674 * /dev/md/1 -> 1
675 * /dev/md/d1 -> d1
676 * /dev/md/home -> home
677 * /dev/mdhome -> home
678 */
69207ff6 679 /* FIXME compare this with rules in create_mddev */
b3b33eb5
NB
680 name = strrchr(mddev, '/');
681 if (name) {
682 name++;
60248f74
NB
683 if (strncmp(name, "md_d", 4)==0 &&
684 strlen(name) > 4 &&
685 isdigit(name[4]) &&
b3b33eb5 686 (name-mddev) == 5 /* /dev/ */)
d1e80164 687 name += 3;
60248f74
NB
688 else if (strncmp(name, "md", 2)==0 &&
689 strlen(name) > 2 &&
690 isdigit(name[2]) &&
691 (name-mddev) == 5 /* /dev/ */)
b3b33eb5
NB
692 name += 2;
693 }
694 }
83cd1e97 695 if (!st->ss->init_super(st, &info.array, s->size, name, c->homehost, uuid,
40c9a66a 696 data_offset))
e06af9dd 697 goto abort_locked;
682c7051 698
d2ca6449 699 total_slots = info.array.nr_disks;
a5d85af7 700 st->ss->getinfo_super(st, &info, NULL);
95eeceeb 701 sysfs_init(&info, mdfd, 0);
159c3a1a 702
171dccc8 703 if (did_default && c->verbose >= 0) {
3c558363 704 if (is_subarray(info.text_version)) {
f7f1b6a1
NB
705 int dnum = devname2devnum(info.text_version+1);
706 char *path;
707 int mdp = get_mdp_major();
708 struct mdinfo *mdi;
709 if (dnum > 0)
710 path = map_dev(MD_MAJOR, dnum, 1);
711 else
712 path = map_dev(mdp, (-1-dnum)<< 6, 1);
713
714 mdi = sysfs_read(-1, dnum, GET_VERSION);
715
e7b84f9d 716 pr_err("Creating array inside "
ca3b6696 717 "%s container %s\n",
f7f1b6a1
NB
718 mdi?mdi->text_version:"managed", path);
719 sysfs_free(mdi);
720 } else
e7b84f9d 721 pr_err("Defaulting to version"
f7f1b6a1
NB
722 " %s metadata\n", info.text_version);
723 }
b8ac1967 724
ad5bc697 725 map_update(&map, fd2devnum(mdfd), info.text_version,
a04d5763 726 info.uuid, chosen_name);
ad5bc697 727 map_unlock(&map);
a04d5763 728
99cc42f4 729 if (s->bitmap_file && vers < 9003) {
943eafef 730 major_num = BITMAP_MAJOR_HOSTENDIAN;
dcec9ee5 731#ifdef __BIG_ENDIAN
e7b84f9d 732 pr_err("Warning - bitmaps created on this kernel are not portable\n"
dcec9ee5
NB
733 " between different architectured. Consider upgrading the Linux kernel.\n");
734#endif
735 }
736
99cc42f4 737 if (s->bitmap_file && strcmp(s->bitmap_file, "internal")==0) {
55935d51 738 if ((vers%100) < 2) {
e7b84f9d 739 pr_err("internal bitmaps not supported by this kernel.\n");
7f91af49 740 goto abort;
55935d51 741 }
ebeb3663 742 if (!st->ss->add_internal_bitmap) {
e7b84f9d 743 pr_err("internal bitmaps not supported with %s metadata\n",
ebeb3663
N
744 st->ss->name);
745 goto abort;
746 }
99cc42f4
N
747 if (!st->ss->add_internal_bitmap(st, &s->bitmap_chunk,
748 c->delay, s->write_behind,
943eafef 749 bitmapsize, 1, major_num)) {
e7b84f9d 750 pr_err("Given bitmap chunk size not supported.\n");
7f91af49 751 goto abort;
55935d51 752 }
99cc42f4 753 s->bitmap_file = NULL;
55935d51
NB
754 }
755
f35f2525 756 sysfs_init(&info, mdfd, 0);
55935d51 757
d1d599ea 758 if (st->ss->external && st->container_dev != NoMdDev) {
f35f2525
N
759 /* member */
760
761 /* When creating a member, we need to be careful
762 * to negotiate with mdmon properly.
763 * If it is already running, we cannot write to
764 * the devices and must ask it to do that part.
765 * If it isn't running, we write to the devices,
766 * and then start it.
767 * We hold an exclusive open on the container
768 * device to make sure mdmon doesn't exit after
769 * we checked that it is running.
770 *
771 * For now, fail if it is already running.
772 */
773 container_fd = open_dev_excl(st->container_dev);
774 if (container_fd < 0) {
e7b84f9d 775 pr_err("Cannot get exclusive "
f35f2525 776 "open on container - weird.\n");
7f91af49 777 goto abort;
d03373f1 778 }
f35f2525 779 if (mdmon_running(st->container_dev)) {
171dccc8 780 if (c->verbose)
e7b84f9d 781 pr_err("reusing mdmon "
f35f2525
N
782 "for %s.\n",
783 devnum2devname(st->container_dev));
784 st->update_tail = &st->updates;
785 } else
786 need_mdmon = 1;
787 }
788 rv = set_array_info(mdfd, st, &info);
82d9eba6 789 if (rv) {
e7b84f9d 790 pr_err("failed to set array info for %s: %s\n",
52826846 791 mddev, strerror(errno));
7f91af49 792 goto abort;
682c7051 793 }
52826846 794
99cc42f4 795 if (s->bitmap_file) {
c82f047c 796 int uuid[4];
55935d51 797
3da92f27 798 st->ss->uuid_from_super(st, uuid);
99cc42f4
N
799 if (CreateBitmap(s->bitmap_file, c->force, (char*)uuid, s->bitmap_chunk,
800 c->delay, s->write_behind,
f9c25f1d 801 bitmapsize,
943eafef 802 major_num)) {
7f91af49 803 goto abort;
c82f047c 804 }
99cc42f4 805 bitmap_fd = open(s->bitmap_file, O_RDWR);
c82f047c 806 if (bitmap_fd < 0) {
e7b84f9d 807 pr_err("weird: %s cannot be openned\n",
99cc42f4 808 s->bitmap_file);
7f91af49 809 goto abort;
c82f047c
NB
810 }
811 if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) {
e7b84f9d 812 pr_err("Cannot set bitmap file for %s: %s\n",
c82f047c 813 mddev, strerror(errno));
7f91af49 814 goto abort;
c82f047c
NB
815 }
816 }
817
503975b9 818 infos = xmalloc(sizeof(*infos) * total_slots);
4b1ac34b 819
4b1ac34b 820 for (pass=1; pass <=2 ; pass++) {
a655e550 821 struct mddev_dev *moved_disk = NULL; /* the disk that was moved out of the insert point */
4b1ac34b
NB
822
823 for (dnum=0, dv = devlist ; dv ;
824 dv=(dv->next)?(dv->next):moved_disk, dnum++) {
825 int fd;
826 struct stat stb;
d2ca6449 827 struct mdinfo *inf = &infos[dnum];
4b1ac34b 828
d2ca6449
NB
829 if (dnum >= total_slots)
830 abort();
4b1ac34b
NB
831 if (dnum == insert_point) {
832 moved_disk = dv;
8592f29d 833 continue;
52826846 834 }
8592f29d 835 if (strcasecmp(dv->devname, "missing")==0)
111d01fc 836 continue;
8592f29d
N
837 if (have_container)
838 moved_disk = NULL;
839 if (have_container && dnum < info.array.raid_disks - 1)
840 /* repeatedly use the container */
841 moved_disk = dv;
111d01fc 842
d2ca6449
NB
843 switch(pass) {
844 case 1:
845 *inf = info;
5f8097be 846
d2ca6449
NB
847 inf->disk.number = dnum;
848 inf->disk.raid_disk = dnum;
99cc42f4 849 if (inf->disk.raid_disk < s->raiddisks)
d2ca6449
NB
850 inf->disk.state = (1<<MD_DISK_ACTIVE) |
851 (1<<MD_DISK_SYNC);
852 else
853 inf->disk.state = 0;
854
b01b06bd 855 if (dv->writemostly == 1)
d2ca6449
NB
856 inf->disk.state |= (1<<MD_DISK_WRITEMOSTLY);
857
8592f29d
N
858 if (have_container)
859 fd = -1;
860 else {
d1d599ea
N
861 if (st->ss->external &&
862 st->container_dev != NoMdDev)
8592f29d
N
863 fd = open(dv->devname, O_RDWR);
864 else
865 fd = open(dv->devname, O_RDWR|O_EXCL);
866
867 if (fd < 0) {
e7b84f9d 868 pr_err("failed to open %s "
8592f29d
N
869 "after earlier success - aborting\n",
870 dv->devname);
871 goto abort;
872 }
873 fstat(fd, &stb);
874 inf->disk.major = major(stb.st_rdev);
875 inf->disk.minor = minor(stb.st_rdev);
d2ca6449 876 }
8592f29d
N
877 if (fd >= 0)
878 remove_partitions(fd);
f20c3968 879 if (st->ss->add_to_super(st, &inf->disk,
72ca9bcf
N
880 fd, dv->devname,
881 dv->data_offset)) {
4eb26970 882 ioctl(mdfd, STOP_ARRAY, NULL);
f20c3968 883 goto abort;
4eb26970 884 }
a5d85af7 885 st->ss->getinfo_super(st, inf, NULL);
8ed3e5e1 886 safe_mode_delay = inf->safe_mode_delay;
d2ca6449 887
171dccc8 888 if (have_container && c->verbose > 0)
e7b84f9d 889 pr_err("Using %s for device %d\n",
8592f29d
N
890 map_dev(inf->disk.major,
891 inf->disk.minor,
892 0), dnum);
893
894 if (!have_container) {
895 /* getinfo_super might have lost these ... */
896 inf->disk.major = major(stb.st_rdev);
897 inf->disk.minor = minor(stb.st_rdev);
898 }
4b1ac34b
NB
899 break;
900 case 2:
d2ca6449 901 inf->errors = 0;
4b1ac34b 902
f35f2525 903 rv = add_disk(mdfd, st, &info, inf);
d2ca6449 904
2503d23b 905 if (rv) {
e7b84f9d
N
906 pr_err("ADD_NEW_DISK for %s "
907 "failed: %s\n",
908 dv->devname, strerror(errno));
7f91af49 909 goto abort;
4b1ac34b 910 }
4b1ac34b
NB
911 break;
912 }
8592f29d
N
913 if (!have_container &&
914 dv == moved_disk && dnum != insert_point) break;
52826846 915 }
edd8d13c 916 if (pass == 1) {
9b1fb677
DW
917 struct mdinfo info_new;
918 struct map_ent *me = NULL;
919
920 /* check to see if the uuid has changed due to these
921 * metadata changes, and if so update the member array
922 * and container uuid. Note ->write_init_super clears
923 * the subarray cursor such that ->getinfo_super once
924 * again returns container info.
925 */
926 map_lock(&map);
a5d85af7 927 st->ss->getinfo_super(st, &info_new, NULL);
99cc42f4 928 if (st->ss->external && s->level != LEVEL_CONTAINER &&
9b1fb677
DW
929 !same_uuid(info_new.uuid, info.uuid, 0)) {
930 map_update(&map, fd2devnum(mdfd),
931 info_new.text_version,
932 info_new.uuid, chosen_name);
933 me = map_by_devnum(&map, st->container_dev);
934 }
935
ab80e597 936 if (st->ss->write_init_super(st)) {
ab80e597 937 st->ss->free_super(st);
e06af9dd 938 goto abort_locked;
ab80e597 939 }
9b1fb677
DW
940
941 /* update parent container uuid */
942 if (me) {
503975b9 943 char *path = xstrdup(me->path);
9b1fb677 944
a5d85af7 945 st->ss->getinfo_super(st, &info_new, NULL);
9b1fb677
DW
946 map_update(&map, st->container_dev,
947 info_new.text_version,
948 info_new.uuid, path);
949 free(path);
950 }
951 map_unlock(&map);
952
edd8d13c 953 flush_metadata_updates(st);
6946681d 954 st->ss->free_super(st);
edd8d13c 955 }
5e52ae9e 956 }
d2ca6449 957 free(infos);
5e52ae9e 958
99cc42f4 959 if (s->level == LEVEL_CONTAINER) {
97590376
N
960 /* No need to start. But we should signal udev to
961 * create links */
962 sysfs_uevent(&info, "change");
171dccc8 963 if (c->verbose >= 0)
e7b84f9d 964 pr_err("container %s prepared.\n", mddev);
a7c6e3fb 965 wait_for(chosen_name, mdfd);
99cc42f4 966 } else if (c->runstop == 1 || subdevs >= s->raiddisks) {
598f0d58 967 if (st->ss->external) {
4c821454 968 int err;
99cc42f4 969 switch(s->level) {
598f0d58
NB
970 case LEVEL_LINEAR:
971 case LEVEL_MULTIPATH:
972 case 0:
4c821454 973 err = sysfs_set_str(&info, NULL, "array_state",
171dccc8 974 c->readonly
72d566f6
N
975 ? "readonly"
976 : "active");
a931db9e 977 need_mdmon = 0;
598f0d58
NB
978 break;
979 default:
4c821454
N
980 err = sysfs_set_str(&info, NULL, "array_state",
981 "readonly");
598f0d58
NB
982 break;
983 }
f35f2525 984 sysfs_set_safemode(&info, safe_mode_delay);
4c821454 985 if (err) {
e7b84f9d
N
986 pr_err("failed to"
987 " activate array.\n");
4c821454
N
988 ioctl(mdfd, STOP_ARRAY, NULL);
989 goto abort;
990 }
171dccc8 991 } else if (c->readonly &&
72d566f6
N
992 sysfs_attribute_available(
993 &info, NULL, "array_state")) {
994 if (sysfs_set_str(&info, NULL,
995 "array_state", "readonly") < 0) {
996 pr_err("Failed to start array: %s\n",
997 strerror(errno));
998 ioctl(mdfd, STOP_ARRAY, NULL);
999 goto abort;
1000 }
598f0d58 1001 } else {
97590376 1002 /* param is not actually used */
598f0d58
NB
1003 mdu_param_t param;
1004 if (ioctl(mdfd, RUN_ARRAY, &param)) {
e7b84f9d 1005 pr_err("RUN_ARRAY failed: %s\n",
72d566f6 1006 strerror(errno));
a252c078 1007 if (info.array.chunk_size & (info.array.chunk_size-1)) {
e7b84f9d
N
1008 cont_err("Problem may be that "
1009 "chunk size is not a power of 2\n");
a252c078 1010 }
4eb26970 1011 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1012 goto abort;
598f0d58 1013 }
682c7051 1014 }
171dccc8 1015 if (c->verbose >= 0)
e7b84f9d 1016 pr_err("array %s started.\n", mddev);
d1d599ea 1017 if (st->ss->external && st->container_dev != NoMdDev) {
8850ee3e
N
1018 if (need_mdmon)
1019 start_mdmon(st->container_dev);
1020
983fff45 1021 ping_monitor_by_id(st->container_dev);
a931db9e
NB
1022 close(container_fd);
1023 }
a7c6e3fb 1024 wait_for(chosen_name, mdfd);
682c7051 1025 } else {
e7b84f9d 1026 pr_err("not starting array - not enough devices.\n");
682c7051 1027 }
7f91af49 1028 close(mdfd);
682c7051 1029 return 0;
7f91af49
N
1030
1031 abort:
4eb26970 1032 map_lock(&map);
e06af9dd 1033 abort_locked:
4eb26970
DW
1034 map_remove(&map, fd2devnum(mdfd));
1035 map_unlock(&map);
1036
7f91af49
N
1037 if (mdfd >= 0)
1038 close(mdfd);
1039 return 1;
64c4757e 1040}