]> git.ipfire.org Git - thirdparty/mdadm.git/blame - Create.c
Make sure NOFILE resource limit is big enough.
[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 }
8baab049 302 if (data_offset == VARIABLE_OFFSET) {
72ca9bcf
N
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 }
066e92f0
LD
411 if (!freesize) {
412 pr_err("no free space left on %s\n", dname);
413 fail = 1;
414 continue;
415 }
682c7051 416
99cc42f4 417 if (s->size && freesize < s->size) {
e7b84f9d 418 pr_err("%s is smaller than given size."
17f25ca6 419 " %lluK < %lluK + metadata\n",
99cc42f4 420 dname, freesize, s->size);
52826846 421 fail = 1;
52826846 422 continue;
682c7051 423 }
cd29a5c8
NB
424 if (maxdisc == NULL || (maxdisc && freesize > maxsize)) {
425 maxdisc = dname;
52826846 426 maxsize = freesize;
682c7051 427 }
cd29a5c8
NB
428 if (mindisc ==NULL || (mindisc && freesize < minsize)) {
429 mindisc = dname;
52826846 430 minsize = freesize;
682c7051 431 }
171dccc8 432 if (c->runstop != 1 || c->verbose >= 0) {
37ea3936 433 int fd = open(dname, O_RDONLY);
b6e63da4 434 if (fd <0 ) {
e7b84f9d 435 pr_err("Cannot open %s: %s\n",
b6e63da4
NB
436 dname, strerror(errno));
437 fail=1;
438 continue;
439 }
dab6685f
NB
440 warn |= check_ext2(fd, dname);
441 warn |= check_reiser(fd, dname);
442 warn |= check_raid(fd, dname);
034b203a
TM
443 if (strcmp(st->ss->name, "1.x") == 0 &&
444 st->minor_version >= 1)
445 /* metadata at front */
53ed6ac3 446 warn |= check_partitions(fd, dname, 0, 0);
99cc42f4
N
447 else if (s->level == 1 || s->level == LEVEL_CONTAINER
448 || (s->level == 0 && s->raiddisks == 1))
034b203a 449 /* partitions could be meaningful */
99cc42f4 450 warn |= check_partitions(fd, dname, freesize*2, s->size*2);
034b203a
TM
451 else
452 /* partitions cannot be meaningful */
53ed6ac3 453 warn |= check_partitions(fd, dname, 0, 0);
034b203a 454 if (strcmp(st->ss->name, "1.x") == 0 &&
a0962fe9
N
455 st->minor_version >= 1 &&
456 did_default &&
99cc42f4 457 s->level == 1 &&
034b203a
TM
458 (warn & 1024) == 0) {
459 warn |= 1024;
e7b84f9d 460 pr_err("Note: this array has metadata at the start and\n"
a0962fe9 461 " may not be suitable as a boot device. If you plan to\n"
cc86f89c 462 " store '/boot' on this device please ensure that\n"
a0962fe9 463 " your boot-loader understands md/v1.x metadata, or use\n"
cc86f89c 464 " --metadata=0.90\n");
a0962fe9 465 }
b6e63da4 466 close(fd);
dab6685f 467 }
682c7051 468 }
99cc42f4 469 if (s->raiddisks + s->sparedisks > st->max_devs) {
e7b84f9d 470 pr_err("Too many devices:"
acab7bb1
N
471 " %s metadata only supports %d\n",
472 st->ss->name, st->max_devs);
473 return 1;
474 }
8592f29d 475 if (have_container)
99cc42f4 476 info.array.working_disks = s->raiddisks;
682c7051 477 if (fail) {
e7b84f9d 478 pr_err("create aborted\n");
52826846 479 return 1;
682c7051 480 }
99cc42f4 481 if (s->size == 0) {
5f8097be 482 if (mindisc == NULL && !have_container) {
e7b84f9d 483 pr_err("no size and no drives given - aborting create.\n");
52826846
NB
484 return 1;
485 }
99cc42f4
N
486 if (s->level > 0 || s->level == LEVEL_MULTIPATH
487 || s->level == LEVEL_FAULTY
b8ac1967 488 || st->ss->external ) {
b5e64645 489 /* size is meaningful */
99cc42f4
N
490 if (!st->ss->validate_geometry(st, s->level, s->layout,
491 s->raiddisks,
492 &s->chunk, minsize*2,
40c9a66a 493 data_offset,
2c514b71 494 NULL, NULL, 0)) {
99cc42f4 495 pr_err("devices too large for RAID level %d\n", s->level);
b5e64645
NB
496 return 1;
497 }
99cc42f4
N
498 s->size = minsize;
499 if (s->level == 1)
ae6c05ad
N
500 /* If this is ever reshaped to RAID5, we will
501 * need a chunksize. So round it off a bit
502 * now just to be safe
503 */
99cc42f4 504 s->size &= ~(64ULL-1);
171dccc8 505 if (c->verbose > 0)
99cc42f4 506 pr_err("size set to %lluK\n", s->size);
b5e64645 507 }
682c7051 508 }
748952f7
N
509
510 if (!s->bitmap_file &&
511 s->level >= 1 &&
512 (s->write_behind || s->size > 100*1024*1024ULL)) {
513 if (c->verbose > 0)
514 pr_err("automatically enabling write-intent bitmap on large array\n");
515 s->bitmap_file = "internal";
516 }
517 if (s->bitmap_file && strcmp(s->bitmap_file, "none") == 0)
518 s->bitmap_file = NULL;
519
99cc42f4 520 if (!have_container && s->level > 0 && ((maxsize-s->size)*100 > maxsize)) {
171dccc8 521 if (c->runstop != 1 || c->verbose >= 0)
e7b84f9d 522 pr_err("largest drive (%s) exceeds size (%lluK) by more than 1%%\n",
99cc42f4 523 maxdisc, s->size);
52826846 524 warn = 1;
682c7051
NB
525 }
526
9eafa1de 527 if (st->ss->detail_platform && st->ss->detail_platform(0, 1, NULL) != 0) {
171dccc8 528 if (c->runstop != 1 || c->verbose >= 0)
e7b84f9d 529 pr_err("%s unable to enumerate platform support\n"
5615172f
DW
530 " array may not be compatible with hardware/firmware\n",
531 st->ss->name);
532 warn = 1;
533 }
534
682c7051 535 if (warn) {
171dccc8 536 if (c->runstop!= 1) {
52826846 537 if (!ask("Continue creating array? ")) {
e7b84f9d 538 pr_err("create aborted.\n");
52826846
NB
539 return 1;
540 }
541 } else {
171dccc8 542 if (c->verbose > 0)
e7b84f9d 543 pr_err("creation continuing despite oddities due to --run\n");
682c7051 544 }
682c7051
NB
545 }
546
66f8bbbe 547 /* If this is raid4/5, we want to configure the last active slot
52826846 548 * as missing, so that a reconstruct happens (faster than re-parity)
98c6faba 549 * FIX: Can we do this for raid6 as well?
52826846 550 */
87b47257 551 if (st->ss->external == 0 &&
99cc42f4
N
552 s->assume_clean==0 && c->force == 0 && first_missing >= s->raiddisks) {
553 switch ( s->level ) {
66f8bbbe 554 case 4:
98c6faba 555 case 5:
99cc42f4
N
556 insert_point = s->raiddisks-1;
557 s->sparedisks++;
06c7f68e 558 info.array.active_disks--;
98c6faba
NB
559 missing_disks++;
560 break;
561 default:
562 break;
563 }
52826846 564 }
6fb79233
NB
565 /* For raid6, if creating with 1 missing drive, make a good drive
566 * into a spare, else the create will fail
567 */
99cc42f4 568 if (s->assume_clean == 0 && c->force == 0 && first_missing < s->raiddisks &&
ffcfc735 569 st->ss->external == 0 &&
99cc42f4
N
570 second_missing >= s->raiddisks && s->level == 6) {
571 insert_point = s->raiddisks - 1;
6fb79233
NB
572 if (insert_point == first_missing)
573 insert_point--;
99cc42f4 574 s->sparedisks ++;
6fb79233
NB
575 info.array.active_disks--;
576 missing_disks++;
577 }
570510ba 578
99cc42f4 579 if (s->level <= 0 && first_missing < subdevs * 2) {
e7b84f9d 580 pr_err("This level does not support missing devices\n");
570510ba
NB
581 return 1;
582 }
aba69144 583
7f91af49 584 /* We need to create the device */
ad5bc697 585 map_lock(&map);
171dccc8 586 mdfd = create_mddev(mddev, name, c->autof, LOCAL, chosen_name);
e06af9dd
JS
587 if (mdfd < 0) {
588 map_unlock(&map);
7f91af49 589 return 1;
e06af9dd 590 }
3e9df86a
AK
591 /* verify if chosen_name is not in use,
592 * it could be in conflict with already existing device
593 * e.g. container, array
594 */
595 if (strncmp(chosen_name, "/dev/md/", 8) == 0
596 && map_by_name(&map, chosen_name+8) != NULL) {
e7b84f9d 597 pr_err("Array name %s is in use already.\n",
3e9df86a
AK
598 chosen_name);
599 close(mdfd);
600 map_unlock(&map);
601 return 1;
602 }
f2be09f1 603 mddev = chosen_name;
7f91af49
N
604
605 vers = md_get_version(mdfd);
606 if (vers < 9000) {
e7b84f9d 607 pr_err("Create requires md driver version 0.90.0 or later\n");
e06af9dd 608 goto abort_locked;
7f91af49
N
609 } else {
610 mdu_array_info_t inf;
611 memset(&inf, 0, sizeof(inf));
612 ioctl(mdfd, GET_ARRAY_INFO, &inf);
613 if (inf.working_disks != 0) {
e7b84f9d 614 pr_err("another array by this name"
7f91af49 615 " is already running.\n");
e06af9dd 616 goto abort_locked;
7f91af49
N
617 }
618 }
a04d5763 619
682c7051
NB
620 /* Ok, lets try some ioctls */
621
99cc42f4
N
622 info.array.level = s->level;
623 info.array.size = s->size;
624 info.array.raid_disks = s->raiddisks;
82b27616
NB
625 /* The kernel should *know* what md_minor we are dealing
626 * with, but it chooses to trust me instead. Sigh
627 */
06c7f68e 628 info.array.md_minor = 0;
82b27616 629 if (fstat(mdfd, &stb)==0)
06c7f68e
NB
630 info.array.md_minor = minor(stb.st_rdev);
631 info.array.not_persistent = 0;
6fb79233 632
99cc42f4
N
633 if ( ( (s->level == 4 || s->level == 5) &&
634 (insert_point < s->raiddisks || first_missing < s->raiddisks) )
b5e64645 635 ||
99cc42f4
N
636 ( s->level == 6 && (insert_point < s->raiddisks
637 || second_missing < s->raiddisks))
47d79ef8 638 ||
99cc42f4 639 ( s->level <= 0 )
d024b0a7 640 ||
99cc42f4 641 s->assume_clean
103f2410 642 ) {
06c7f68e 643 info.array.state = 1; /* clean, but one+ drive will be missing*/
b7528a20 644 info.resync_start = MaxSector;
103f2410 645 } else {
06c7f68e 646 info.array.state = 0; /* not clean, but no errors */
103f2410
NB
647 info.resync_start = 0;
648 }
99cc42f4 649 if (s->level == 10) {
f9c25f1d
NB
650 /* for raid10, the bitmap size is the capacity of the array,
651 * which is array.size * raid_disks / ncopies;
652 * .. but convert to sectors.
653 */
99cc42f4
N
654 int ncopies = ((s->layout>>8) & 255) * (s->layout & 255);
655 bitmapsize = s->size * s->raiddisks / ncopies * 2;
656/* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, s->size, s->raiddisks, ncopies);*/
f9c25f1d 657 } else
99cc42f4 658 bitmapsize = s->size * 2;
f9c25f1d 659
82b27616
NB
660 /* There is lots of redundancy in these disk counts,
661 * raid_disks is the most meaningful value
662 * it describes the geometry of the array
663 * it is constant
664 * nr_disks is total number of used slots.
665 * it should be raid_disks+spare_disks
666 * spare_disks is the number of extra disks present
667 * see above
668 * active_disks is the number of working disks in
669 * active slots. (With raid_disks)
670 * working_disks is the total number of working disks,
671 * including spares
672 * failed_disks is the number of disks marked failed
673 *
5d500228 674 * Ideally, the kernel would keep these (except raid_disks)
82b27616
NB
675 * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
676 * So for now, we assume that all raid and spare
677 * devices will be given.
678 */
99cc42f4 679 info.array.spare_disks=s->sparedisks;
06c7f68e
NB
680 info.array.failed_disks=missing_disks;
681 info.array.nr_disks = info.array.working_disks
682 + info.array.failed_disks;
99cc42f4
N
683 info.array.layout = s->layout;
684 info.array.chunk_size = s->chunk*1024;
82d9eba6 685
b3b33eb5
NB
686 if (name == NULL || *name == 0) {
687 /* base name on mddev */
d1e80164
NB
688 /* /dev/md0 -> 0
689 * /dev/md_d0 -> d0
eca944fa 690 * /dev/md_foo -> foo
d1e80164
NB
691 * /dev/md/1 -> 1
692 * /dev/md/d1 -> d1
693 * /dev/md/home -> home
694 * /dev/mdhome -> home
695 */
69207ff6 696 /* FIXME compare this with rules in create_mddev */
b3b33eb5
NB
697 name = strrchr(mddev, '/');
698 if (name) {
699 name++;
eca944fa
N
700 if (strncmp(name, "md_", 3)==0 &&
701 strlen(name) > 3 &&
b3b33eb5 702 (name-mddev) == 5 /* /dev/ */)
d1e80164 703 name += 3;
60248f74
NB
704 else if (strncmp(name, "md", 2)==0 &&
705 strlen(name) > 2 &&
706 isdigit(name[2]) &&
707 (name-mddev) == 5 /* /dev/ */)
b3b33eb5
NB
708 name += 2;
709 }
710 }
83cd1e97 711 if (!st->ss->init_super(st, &info.array, s->size, name, c->homehost, uuid,
40c9a66a 712 data_offset))
e06af9dd 713 goto abort_locked;
682c7051 714
d2ca6449 715 total_slots = info.array.nr_disks;
a5d85af7 716 st->ss->getinfo_super(st, &info, NULL);
4dd2df09 717 sysfs_init(&info, mdfd, NULL);
159c3a1a 718
171dccc8 719 if (did_default && c->verbose >= 0) {
3c558363 720 if (is_subarray(info.text_version)) {
4dd2df09
N
721 char devnm[32];
722 char *ep;
f7f1b6a1 723 struct mdinfo *mdi;
f7f1b6a1 724
4dd2df09
N
725 strncpy(devnm, info.text_version+1, 32);
726 devnm[31] = 0;
727 ep = strchr(devnm, '/');
728 if (ep)
729 *ep = 0;
730
731 mdi = sysfs_read(-1, devnm, GET_VERSION);
f7f1b6a1 732
4dd2df09
N
733 pr_err("Creating array inside %s container %s\n",
734 mdi?mdi->text_version:"managed", devnm);
f7f1b6a1
NB
735 sysfs_free(mdi);
736 } else
e7b84f9d 737 pr_err("Defaulting to version"
f7f1b6a1
NB
738 " %s metadata\n", info.text_version);
739 }
b8ac1967 740
4dd2df09 741 map_update(&map, fd2devnm(mdfd), info.text_version,
a04d5763 742 info.uuid, chosen_name);
ad5bc697 743 map_unlock(&map);
a04d5763 744
99cc42f4 745 if (s->bitmap_file && vers < 9003) {
943eafef 746 major_num = BITMAP_MAJOR_HOSTENDIAN;
dcec9ee5 747#ifdef __BIG_ENDIAN
e7b84f9d 748 pr_err("Warning - bitmaps created on this kernel are not portable\n"
dcec9ee5
NB
749 " between different architectured. Consider upgrading the Linux kernel.\n");
750#endif
751 }
752
99cc42f4 753 if (s->bitmap_file && strcmp(s->bitmap_file, "internal")==0) {
55935d51 754 if ((vers%100) < 2) {
e7b84f9d 755 pr_err("internal bitmaps not supported by this kernel.\n");
7f91af49 756 goto abort;
55935d51 757 }
ebeb3663 758 if (!st->ss->add_internal_bitmap) {
e7b84f9d 759 pr_err("internal bitmaps not supported with %s metadata\n",
ebeb3663
N
760 st->ss->name);
761 goto abort;
762 }
99cc42f4
N
763 if (!st->ss->add_internal_bitmap(st, &s->bitmap_chunk,
764 c->delay, s->write_behind,
943eafef 765 bitmapsize, 1, major_num)) {
e7b84f9d 766 pr_err("Given bitmap chunk size not supported.\n");
7f91af49 767 goto abort;
55935d51 768 }
99cc42f4 769 s->bitmap_file = NULL;
55935d51
NB
770 }
771
4dd2df09 772 sysfs_init(&info, mdfd, NULL);
55935d51 773
4dd2df09 774 if (st->ss->external && st->container_devnm[0]) {
f35f2525
N
775 /* member */
776
777 /* When creating a member, we need to be careful
778 * to negotiate with mdmon properly.
779 * If it is already running, we cannot write to
780 * the devices and must ask it to do that part.
781 * If it isn't running, we write to the devices,
782 * and then start it.
783 * We hold an exclusive open on the container
784 * device to make sure mdmon doesn't exit after
785 * we checked that it is running.
786 *
787 * For now, fail if it is already running.
788 */
4dd2df09 789 container_fd = open_dev_excl(st->container_devnm);
f35f2525 790 if (container_fd < 0) {
e7b84f9d 791 pr_err("Cannot get exclusive "
f35f2525 792 "open on container - weird.\n");
7f91af49 793 goto abort;
d03373f1 794 }
4dd2df09 795 if (mdmon_running(st->container_devnm)) {
171dccc8 796 if (c->verbose)
e7b84f9d 797 pr_err("reusing mdmon "
f35f2525 798 "for %s.\n",
4dd2df09 799 st->container_devnm);
f35f2525
N
800 st->update_tail = &st->updates;
801 } else
802 need_mdmon = 1;
803 }
804 rv = set_array_info(mdfd, st, &info);
82d9eba6 805 if (rv) {
e7b84f9d 806 pr_err("failed to set array info for %s: %s\n",
52826846 807 mddev, strerror(errno));
7f91af49 808 goto abort;
682c7051 809 }
52826846 810
99cc42f4 811 if (s->bitmap_file) {
c82f047c 812 int uuid[4];
55935d51 813
3da92f27 814 st->ss->uuid_from_super(st, uuid);
99cc42f4
N
815 if (CreateBitmap(s->bitmap_file, c->force, (char*)uuid, s->bitmap_chunk,
816 c->delay, s->write_behind,
f9c25f1d 817 bitmapsize,
943eafef 818 major_num)) {
7f91af49 819 goto abort;
c82f047c 820 }
99cc42f4 821 bitmap_fd = open(s->bitmap_file, O_RDWR);
c82f047c 822 if (bitmap_fd < 0) {
e7b84f9d 823 pr_err("weird: %s cannot be openned\n",
99cc42f4 824 s->bitmap_file);
7f91af49 825 goto abort;
c82f047c
NB
826 }
827 if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) {
e7b84f9d 828 pr_err("Cannot set bitmap file for %s: %s\n",
c82f047c 829 mddev, strerror(errno));
7f91af49 830 goto abort;
c82f047c
NB
831 }
832 }
833
503975b9 834 infos = xmalloc(sizeof(*infos) * total_slots);
a7dec3fd 835 enable_fds(total_slots);
4b1ac34b 836 for (pass=1; pass <=2 ; pass++) {
a655e550 837 struct mddev_dev *moved_disk = NULL; /* the disk that was moved out of the insert point */
4b1ac34b
NB
838
839 for (dnum=0, dv = devlist ; dv ;
840 dv=(dv->next)?(dv->next):moved_disk, dnum++) {
841 int fd;
842 struct stat stb;
d2ca6449 843 struct mdinfo *inf = &infos[dnum];
4b1ac34b 844
d2ca6449
NB
845 if (dnum >= total_slots)
846 abort();
4b1ac34b
NB
847 if (dnum == insert_point) {
848 moved_disk = dv;
8592f29d 849 continue;
52826846 850 }
8592f29d 851 if (strcasecmp(dv->devname, "missing")==0)
111d01fc 852 continue;
8592f29d
N
853 if (have_container)
854 moved_disk = NULL;
855 if (have_container && dnum < info.array.raid_disks - 1)
856 /* repeatedly use the container */
857 moved_disk = dv;
111d01fc 858
d2ca6449
NB
859 switch(pass) {
860 case 1:
861 *inf = info;
5f8097be 862
d2ca6449
NB
863 inf->disk.number = dnum;
864 inf->disk.raid_disk = dnum;
99cc42f4 865 if (inf->disk.raid_disk < s->raiddisks)
d2ca6449
NB
866 inf->disk.state = (1<<MD_DISK_ACTIVE) |
867 (1<<MD_DISK_SYNC);
868 else
869 inf->disk.state = 0;
870
b01b06bd 871 if (dv->writemostly == 1)
d2ca6449
NB
872 inf->disk.state |= (1<<MD_DISK_WRITEMOSTLY);
873
8592f29d
N
874 if (have_container)
875 fd = -1;
876 else {
d1d599ea 877 if (st->ss->external &&
4dd2df09 878 st->container_devnm[0])
8592f29d
N
879 fd = open(dv->devname, O_RDWR);
880 else
881 fd = open(dv->devname, O_RDWR|O_EXCL);
882
883 if (fd < 0) {
e7b84f9d 884 pr_err("failed to open %s "
8592f29d
N
885 "after earlier success - aborting\n",
886 dv->devname);
887 goto abort;
888 }
889 fstat(fd, &stb);
890 inf->disk.major = major(stb.st_rdev);
891 inf->disk.minor = minor(stb.st_rdev);
d2ca6449 892 }
8592f29d
N
893 if (fd >= 0)
894 remove_partitions(fd);
f20c3968 895 if (st->ss->add_to_super(st, &inf->disk,
72ca9bcf
N
896 fd, dv->devname,
897 dv->data_offset)) {
4eb26970 898 ioctl(mdfd, STOP_ARRAY, NULL);
f20c3968 899 goto abort;
4eb26970 900 }
a5d85af7 901 st->ss->getinfo_super(st, inf, NULL);
8ed3e5e1 902 safe_mode_delay = inf->safe_mode_delay;
d2ca6449 903
171dccc8 904 if (have_container && c->verbose > 0)
e7b84f9d 905 pr_err("Using %s for device %d\n",
8592f29d
N
906 map_dev(inf->disk.major,
907 inf->disk.minor,
908 0), dnum);
909
910 if (!have_container) {
911 /* getinfo_super might have lost these ... */
912 inf->disk.major = major(stb.st_rdev);
913 inf->disk.minor = minor(stb.st_rdev);
914 }
4b1ac34b
NB
915 break;
916 case 2:
d2ca6449 917 inf->errors = 0;
4b1ac34b 918
f35f2525 919 rv = add_disk(mdfd, st, &info, inf);
d2ca6449 920
2503d23b 921 if (rv) {
e7b84f9d
N
922 pr_err("ADD_NEW_DISK for %s "
923 "failed: %s\n",
924 dv->devname, strerror(errno));
7f91af49 925 goto abort;
4b1ac34b 926 }
4b1ac34b
NB
927 break;
928 }
8592f29d
N
929 if (!have_container &&
930 dv == moved_disk && dnum != insert_point) break;
52826846 931 }
edd8d13c 932 if (pass == 1) {
9b1fb677
DW
933 struct mdinfo info_new;
934 struct map_ent *me = NULL;
935
936 /* check to see if the uuid has changed due to these
937 * metadata changes, and if so update the member array
938 * and container uuid. Note ->write_init_super clears
939 * the subarray cursor such that ->getinfo_super once
940 * again returns container info.
941 */
942 map_lock(&map);
a5d85af7 943 st->ss->getinfo_super(st, &info_new, NULL);
99cc42f4 944 if (st->ss->external && s->level != LEVEL_CONTAINER &&
9b1fb677 945 !same_uuid(info_new.uuid, info.uuid, 0)) {
4dd2df09 946 map_update(&map, fd2devnm(mdfd),
9b1fb677
DW
947 info_new.text_version,
948 info_new.uuid, chosen_name);
4dd2df09 949 me = map_by_devnm(&map, st->container_devnm);
9b1fb677
DW
950 }
951
ab80e597 952 if (st->ss->write_init_super(st)) {
ab80e597 953 st->ss->free_super(st);
e06af9dd 954 goto abort_locked;
ab80e597 955 }
9b1fb677
DW
956
957 /* update parent container uuid */
958 if (me) {
503975b9 959 char *path = xstrdup(me->path);
9b1fb677 960
a5d85af7 961 st->ss->getinfo_super(st, &info_new, NULL);
4dd2df09 962 map_update(&map, st->container_devnm,
9b1fb677
DW
963 info_new.text_version,
964 info_new.uuid, path);
965 free(path);
966 }
967 map_unlock(&map);
968
edd8d13c 969 flush_metadata_updates(st);
6946681d 970 st->ss->free_super(st);
edd8d13c 971 }
5e52ae9e 972 }
d2ca6449 973 free(infos);
5e52ae9e 974
99cc42f4 975 if (s->level == LEVEL_CONTAINER) {
97590376
N
976 /* No need to start. But we should signal udev to
977 * create links */
978 sysfs_uevent(&info, "change");
171dccc8 979 if (c->verbose >= 0)
e7b84f9d 980 pr_err("container %s prepared.\n", mddev);
a7c6e3fb 981 wait_for(chosen_name, mdfd);
99cc42f4 982 } else if (c->runstop == 1 || subdevs >= s->raiddisks) {
598f0d58 983 if (st->ss->external) {
4c821454 984 int err;
99cc42f4 985 switch(s->level) {
598f0d58
NB
986 case LEVEL_LINEAR:
987 case LEVEL_MULTIPATH:
988 case 0:
4c821454 989 err = sysfs_set_str(&info, NULL, "array_state",
171dccc8 990 c->readonly
72d566f6
N
991 ? "readonly"
992 : "active");
a931db9e 993 need_mdmon = 0;
598f0d58
NB
994 break;
995 default:
4c821454
N
996 err = sysfs_set_str(&info, NULL, "array_state",
997 "readonly");
598f0d58
NB
998 break;
999 }
f35f2525 1000 sysfs_set_safemode(&info, safe_mode_delay);
4c821454 1001 if (err) {
e7b84f9d
N
1002 pr_err("failed to"
1003 " activate array.\n");
4c821454
N
1004 ioctl(mdfd, STOP_ARRAY, NULL);
1005 goto abort;
1006 }
171dccc8 1007 } else if (c->readonly &&
72d566f6
N
1008 sysfs_attribute_available(
1009 &info, NULL, "array_state")) {
1010 if (sysfs_set_str(&info, NULL,
1011 "array_state", "readonly") < 0) {
1012 pr_err("Failed to start array: %s\n",
1013 strerror(errno));
1014 ioctl(mdfd, STOP_ARRAY, NULL);
1015 goto abort;
1016 }
598f0d58 1017 } else {
97590376 1018 /* param is not actually used */
598f0d58
NB
1019 mdu_param_t param;
1020 if (ioctl(mdfd, RUN_ARRAY, &param)) {
e7b84f9d 1021 pr_err("RUN_ARRAY failed: %s\n",
72d566f6 1022 strerror(errno));
a252c078 1023 if (info.array.chunk_size & (info.array.chunk_size-1)) {
e7b84f9d
N
1024 cont_err("Problem may be that "
1025 "chunk size is not a power of 2\n");
a252c078 1026 }
4eb26970 1027 ioctl(mdfd, STOP_ARRAY, NULL);
7f91af49 1028 goto abort;
598f0d58 1029 }
a21e848a
N
1030 /* if start_ro module parameter is set, array is
1031 * auto-read-only, which is bad as the resync won't
1032 * start. So lets make it read-write now.
1033 */
1034 ioctl(mdfd, RESTART_ARRAY_RW, NULL);
682c7051 1035 }
171dccc8 1036 if (c->verbose >= 0)
e7b84f9d 1037 pr_err("array %s started.\n", mddev);
4dd2df09 1038 if (st->ss->external && st->container_devnm[0]) {
8850ee3e 1039 if (need_mdmon)
4dd2df09 1040 start_mdmon(st->container_devnm);
8850ee3e 1041
4dd2df09 1042 ping_monitor(st->container_devnm);
a931db9e
NB
1043 close(container_fd);
1044 }
a7c6e3fb 1045 wait_for(chosen_name, mdfd);
682c7051 1046 } else {
e7b84f9d 1047 pr_err("not starting array - not enough devices.\n");
682c7051 1048 }
7f91af49 1049 close(mdfd);
682c7051 1050 return 0;
7f91af49
N
1051
1052 abort:
4eb26970 1053 map_lock(&map);
e06af9dd 1054 abort_locked:
4dd2df09 1055 map_remove(&map, fd2devnm(mdfd));
4eb26970
DW
1056 map_unlock(&map);
1057
7f91af49
N
1058 if (mdfd >= 0)
1059 close(mdfd);
1060 return 1;
64c4757e 1061}