]>
Commit | Line | Data |
---|---|---|
64c4757e | 1 | /* |
9a9dab36 | 2 | * mdadm - manage Linux "md" devices aka RAID arrays. |
64c4757e | 3 | * |
6f02172d | 4 | * Copyright (C) 2001-2013 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 |
30 | static 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 | 64 | int 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; |
98dbf73c | 87 | unsigned long long minsize = 0, maxsize = 0; |
cd29a5c8 NB |
88 | char *mindisc = NULL; |
89 | char *maxdisc = NULL; | |
cc1799c3 | 90 | int dnum, raid_disk_num; |
a655e550 | 91 | struct mddev_dev *dv; |
0a6bff09 | 92 | dev_t rdev; |
98dbf73c | 93 | int fail = 0, warn = 0; |
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 | 100 | int rv; |
c82f047c | 101 | int bitmap_fd; |
5f8097be | 102 | int have_container = 0; |
0e600426 | 103 | int container_fd = -1; |
a931db9e | 104 | int need_mdmon = 0; |
f9c25f1d | 105 | unsigned long long bitmapsize; |
d2ca6449 | 106 | struct mdinfo info, *infos; |
b8ac1967 | 107 | int did_default = 0; |
a18a888e | 108 | int do_default_layout = 0; |
c21e737b | 109 | int do_default_chunk = 0; |
8ed3e5e1 | 110 | unsigned long safe_mode_delay = 0; |
a04d5763 | 111 | char chosen_name[1024]; |
ad5bc697 | 112 | struct map_ent *map = NULL; |
8592f29d | 113 | unsigned long long newsize; |
5f4cc239 | 114 | mdu_array_info_t inf; |
682c7051 | 115 | |
943eafef | 116 | int major_num = BITMAP_MAJOR_HI; |
82d9485e | 117 | if (s->bitmap_file && strcmp(s->bitmap_file, "clustered") == 0) { |
6d9c7c25 | 118 | major_num = BITMAP_MAJOR_CLUSTERED; |
82d9485e GJ |
119 | if (c->nodes <= 1) { |
120 | pr_err("At least 2 nodes are needed for cluster-md\n"); | |
121 | return 1; | |
122 | } | |
123 | } | |
dcec9ee5 | 124 | |
06c7f68e | 125 | memset(&info, 0, sizeof(info)); |
99cc42f4 N |
126 | if (s->level == UnSet && st && st->ss->default_geometry) |
127 | st->ss->default_geometry(st, &s->level, NULL, NULL); | |
128 | if (s->level == UnSet) { | |
e7b84f9d | 129 | pr_err("a RAID level is needed to create an array.\n"); |
682c7051 NB |
130 | return 1; |
131 | } | |
99cc42f4 | 132 | if (s->raiddisks < 4 && s->level == 6) { |
e7b84f9d | 133 | pr_err("at least 4 raid-devices needed for level 6\n"); |
98c6faba NB |
134 | return 1; |
135 | } | |
99cc42f4 | 136 | if (s->raiddisks > 256 && s->level == 6) { |
e7b84f9d | 137 | pr_err("no more than 256 raid-devices supported for level 6\n"); |
98c6faba NB |
138 | return 1; |
139 | } | |
99cc42f4 | 140 | if (s->raiddisks < 2 && s->level >= 4) { |
e7b84f9d | 141 | pr_err("at least 2 raid-devices needed for level 4 or 5\n"); |
52826846 NB |
142 | return 1; |
143 | } | |
99cc42f4 | 144 | if (s->level <= 0 && s->sparedisks) { |
e7b84f9d | 145 | pr_err("This level does not support spare devices\n"); |
570510ba NB |
146 | return 1; |
147 | } | |
5f8097be NB |
148 | |
149 | if (subdevs == 1 && strcmp(devlist->devname, "missing") != 0) { | |
150 | /* If given a single device, it might be a container, and we can | |
151 | * extract a device list from there | |
152 | */ | |
5f8097be NB |
153 | int fd; |
154 | ||
155 | memset(&inf, 0, sizeof(inf)); | |
37ea3936 | 156 | fd = open(devlist->devname, O_RDONLY); |
5f8097be | 157 | if (fd >= 0 && |
9cd39f01 | 158 | md_get_array_info(fd, &inf) == 0 && inf.raid_disks == 0) { |
5f8097be NB |
159 | /* yep, looks like a container */ |
160 | if (st) { | |
0fb69d1d N |
161 | rv = st->ss->load_container(st, fd, |
162 | devlist->devname); | |
5f8097be NB |
163 | if (rv == 0) |
164 | have_container = 1; | |
165 | } else { | |
0fb69d1d | 166 | st = super_by_fd(fd, NULL); |
5f8097be | 167 | if (st && !(rv = st->ss-> |
0fb69d1d N |
168 | load_container(st, fd, |
169 | devlist->devname))) | |
5f8097be NB |
170 | have_container = 1; |
171 | else | |
172 | st = NULL; | |
173 | } | |
18fde300 | 174 | if (have_container) { |
99cc42f4 | 175 | subdevs = s->raiddisks; |
18fde300 DW |
176 | first_missing = subdevs * 2; |
177 | second_missing = subdevs * 2; | |
178 | insert_point = subdevs * 2; | |
179 | } | |
5f8097be NB |
180 | } |
181 | if (fd >= 0) | |
182 | close(fd); | |
5f8097be | 183 | } |
99cc42f4 | 184 | if (st && st->ss->external && s->sparedisks) { |
7a862a02 | 185 | pr_err("This metadata type does not support spare disks at create time\n"); |
ffcfc735 N |
186 | return 1; |
187 | } | |
cc1799c3 | 188 | if (subdevs > s->raiddisks+s->sparedisks+s->journaldisks) { |
99cc42f4 | 189 | pr_err("You have listed more devices (%d) than are in the array(%d)!\n", subdevs, s->raiddisks+s->sparedisks); |
52826846 | 190 | return 1; |
682c7051 | 191 | } |
cc1799c3 | 192 | if (!have_container && subdevs < s->raiddisks+s->sparedisks+s->journaldisks) { |
e7b84f9d | 193 | pr_err("You haven't given enough devices (real or missing) to create this array\n"); |
52826846 NB |
194 | return 1; |
195 | } | |
99cc42f4 | 196 | if (s->bitmap_file && s->level <= 0) { |
e7b84f9d | 197 | pr_err("bitmaps not meaningful with level %s\n", |
99cc42f4 | 198 | map_num(pers, s->level)?:"given"); |
5b28bd56 NB |
199 | return 1; |
200 | } | |
52826846 | 201 | |
682c7051 | 202 | /* now set some defaults */ |
b5e64645 | 203 | |
99cc42f4 | 204 | if (s->layout == UnSet) { |
a18a888e | 205 | do_default_layout = 1; |
99cc42f4 | 206 | s->layout = default_layout(st, s->level, c->verbose); |
a18a888e | 207 | } |
682c7051 | 208 | |
99cc42f4 | 209 | if (s->level == 10) |
e5329c37 | 210 | /* check layout fits in array*/ |
99cc42f4 | 211 | if ((s->layout&255) * ((s->layout>>8)&255) > s->raiddisks) { |
e7b84f9d | 212 | pr_err("that layout requires at least %d devices\n", |
99cc42f4 | 213 | (s->layout&255) * ((s->layout>>8)&255)); |
e5329c37 NB |
214 | return 1; |
215 | } | |
216 | ||
99cc42f4 | 217 | switch(s->level) { |
aa88f531 NB |
218 | case 4: |
219 | case 5: | |
e5329c37 | 220 | case 10: |
98c6faba | 221 | case 6: |
aa88f531 | 222 | case 0: |
99cc42f4 N |
223 | if (s->chunk == 0 || s->chunk == UnSet) { |
224 | s->chunk = UnSet; | |
c21e737b CA |
225 | do_default_chunk = 1; |
226 | /* chunk will be set later */ | |
bb7295f1 | 227 | } |
5f175898 N |
228 | break; |
229 | case LEVEL_LINEAR: | |
230 | /* a chunksize of zero 0s perfectly valid (and preferred) since 2.6.16 */ | |
99cc42f4 N |
231 | if (get_linux_version() < 2006016 && s->chunk == 0) { |
232 | s->chunk = 64; | |
171dccc8 | 233 | if (c->verbose > 0) |
e7b84f9d | 234 | pr_err("chunk size defaults to 64K\n"); |
aa88f531 NB |
235 | } |
236 | break; | |
570510ba | 237 | case 1: |
08e43379 | 238 | case LEVEL_FAULTY: |
570510ba | 239 | case LEVEL_MULTIPATH: |
17f25ca6 | 240 | case LEVEL_CONTAINER: |
99cc42f4 N |
241 | if (s->chunk) { |
242 | s->chunk = 0; | |
171dccc8 | 243 | if (c->verbose > 0) |
e7b84f9d | 244 | pr_err("chunk size ignored for this level\n"); |
aa88f531 NB |
245 | } |
246 | break; | |
570510ba | 247 | default: |
99cc42f4 | 248 | pr_err("unknown level %d\n", s->level); |
570510ba | 249 | return 1; |
682c7051 | 250 | } |
99cc42f4 | 251 | if (s->size == MAX_SIZE) |
d04f65f4 | 252 | /* use '0' to mean 'max' now... */ |
99cc42f4 N |
253 | s->size = 0; |
254 | if (s->size && s->chunk && s->chunk != UnSet) | |
255 | s->size &= ~(unsigned long long)(s->chunk - 1); | |
256 | newsize = s->size * 2; | |
257 | if (st && ! st->ss->validate_geometry(st, s->level, s->layout, s->raiddisks, | |
af4348dd | 258 | &s->chunk, s->size*2, |
40c9a66a | 259 | data_offset, NULL, |
5308f117 | 260 | &newsize, s->consistency_policy, |
98dbf73c | 261 | c->verbose >= 0)) |
17f25ca6 | 262 | return 1; |
bb7295f1 | 263 | |
99cc42f4 N |
264 | if (s->chunk && s->chunk != UnSet) { |
265 | newsize &= ~(unsigned long long)(s->chunk*2 - 1); | |
64385908 CA |
266 | if (do_default_chunk) { |
267 | /* default chunk was just set */ | |
171dccc8 | 268 | if (c->verbose > 0) |
7a862a02 | 269 | pr_err("chunk size defaults to %dK\n", s->chunk); |
99cc42f4 | 270 | s->size &= ~(unsigned long long)(s->chunk - 1); |
64385908 CA |
271 | do_default_chunk = 0; |
272 | } | |
bb7295f1 | 273 | } |
64385908 | 274 | |
99cc42f4 N |
275 | if (s->size == 0) { |
276 | s->size = newsize / 2; | |
277 | if (s->level == 1) | |
ae6c05ad N |
278 | /* If this is ever reshaped to RAID5, we will |
279 | * need a chunksize. So round it off a bit | |
280 | * now just to be safe | |
281 | */ | |
99cc42f4 | 282 | s->size &= ~(64ULL-1); |
ae6c05ad | 283 | |
99cc42f4 N |
284 | if (s->size && c->verbose > 0) |
285 | pr_err("setting size to %lluK\n", s->size); | |
8592f29d | 286 | } |
17f25ca6 | 287 | |
682c7051 | 288 | /* now look at the subdevs */ |
06c7f68e NB |
289 | info.array.active_disks = 0; |
290 | info.array.working_disks = 0; | |
c913b90e | 291 | dnum = 0; |
98dbf73c | 292 | for (dv = devlist; dv; dv = dv->next) |
476066a3 N |
293 | if (data_offset == VARIABLE_OFFSET) |
294 | dv->data_offset = INVALID_SECTORS; | |
295 | else | |
296 | dv->data_offset = data_offset; | |
297 | ||
8592f29d | 298 | for (dv=devlist; dv && !have_container; dv=dv->next, dnum++) { |
cd29a5c8 | 299 | char *dname = dv->devname; |
17f25ca6 | 300 | unsigned long long freesize; |
e9b11fee | 301 | int dfd; |
72ca9bcf | 302 | char *doff; |
e9b11fee | 303 | |
98dbf73c | 304 | if (strcasecmp(dname, "missing") == 0) { |
c913b90e NB |
305 | if (first_missing > dnum) |
306 | first_missing = dnum; | |
6fb79233 NB |
307 | if (second_missing > dnum && dnum > first_missing) |
308 | second_missing = dnum; | |
52826846 NB |
309 | missing_disks ++; |
310 | continue; | |
311 | } | |
8baab049 | 312 | if (data_offset == VARIABLE_OFFSET) { |
72ca9bcf N |
313 | doff = strchr(dname, ':'); |
314 | if (doff) { | |
315 | *doff++ = 0; | |
316 | dv->data_offset = parse_size(doff); | |
317 | } else | |
318 | dv->data_offset = INVALID_SECTORS; | |
319 | } else | |
320 | dv->data_offset = data_offset; | |
321 | ||
58b3c697 | 322 | dfd = open(dname, O_RDONLY); |
e9b11fee | 323 | if (dfd < 0) { |
e7b84f9d | 324 | pr_err("cannot open %s: %s\n", |
e9b11fee N |
325 | dname, strerror(errno)); |
326 | exit(2); | |
327 | } | |
0a6bff09 | 328 | if (!fstat_is_blkdev(dfd, dname, NULL)) { |
e9b11fee | 329 | close(dfd); |
e9b11fee N |
330 | exit(2); |
331 | } | |
332 | close(dfd); | |
06c7f68e | 333 | info.array.working_disks++; |
dfd7822c | 334 | if (dnum < s->raiddisks && dv->disposition != 'j') |
06c7f68e | 335 | info.array.active_disks++; |
058574b1 | 336 | if (st == NULL) { |
8aec876d | 337 | struct createinfo *ci = conf_get_create_info(); |
058574b1 NB |
338 | if (ci) |
339 | st = ci->supertype; | |
340 | } | |
576d6d83 NB |
341 | if (st == NULL) { |
342 | /* Need to choose a default metadata, which is different | |
17f25ca6 | 343 | * depending on geometry of array. |
576d6d83 NB |
344 | */ |
345 | int i; | |
346 | char *name = "default"; | |
98dbf73c | 347 | for(i = 0; !st && superlist[i]; i++) { |
576d6d83 | 348 | st = superlist[i]->match_metadata_desc(name); |
ecbd9e81 N |
349 | if (!st) |
350 | continue; | |
a18a888e | 351 | if (do_default_layout) |
99cc42f4 | 352 | s->layout = default_layout(st, s->level, c->verbose); |
ecbd9e81 | 353 | switch (st->ss->validate_geometry( |
99cc42f4 | 354 | st, s->level, s->layout, s->raiddisks, |
af4348dd | 355 | &s->chunk, s->size*2, |
72ca9bcf | 356 | dv->data_offset, dname, |
5308f117 AP |
357 | &freesize, s->consistency_policy, |
358 | c->verbose > 0)) { | |
ecbd9e81 N |
359 | case -1: /* Not valid, message printed, and not |
360 | * worth checking any further */ | |
361 | exit(2); | |
362 | break; | |
363 | case 0: /* Geometry not valid */ | |
55425f27 | 364 | free(st); |
17f25ca6 | 365 | st = NULL; |
99cc42f4 | 366 | s->chunk = do_default_chunk ? UnSet : s->chunk; |
ecbd9e81 N |
367 | break; |
368 | case 1: /* All happy */ | |
369 | break; | |
55425f27 | 370 | } |
17f25ca6 | 371 | } |
576d6d83 NB |
372 | |
373 | if (!st) { | |
58b3c697 N |
374 | int dfd = open(dname, O_RDONLY|O_EXCL); |
375 | if (dfd < 0) { | |
e7b84f9d | 376 | pr_err("cannot open %s: %s\n", |
58b3c697 N |
377 | dname, strerror(errno)); |
378 | exit(2); | |
379 | } | |
7a862a02 | 380 | pr_err("device %s not suitable for any style of array\n", |
17f25ca6 | 381 | dname); |
576d6d83 NB |
382 | exit(2); |
383 | } | |
b8ac1967 | 384 | if (st->ss != &super0 || |
576d6d83 | 385 | st->minor_version != 90) |
b8ac1967 | 386 | did_default = 1; |
17f25ca6 | 387 | } else { |
a18a888e | 388 | if (do_default_layout) |
99cc42f4 N |
389 | s->layout = default_layout(st, s->level, 0); |
390 | if (!st->ss->validate_geometry(st, s->level, s->layout, | |
391 | s->raiddisks, | |
af4348dd | 392 | &s->chunk, s->size*2, |
72ca9bcf | 393 | dv->data_offset, |
af4348dd | 394 | dname, &freesize, |
5308f117 | 395 | s->consistency_policy, |
171dccc8 | 396 | c->verbose >= 0)) { |
17f25ca6 | 397 | |
7a862a02 | 398 | pr_err("%s is not suitable for this array.\n", |
e7b84f9d | 399 | dname); |
17f25ca6 NB |
400 | fail = 1; |
401 | continue; | |
402 | } | |
682c7051 | 403 | } |
82d9eba6 | 404 | |
cc1799c3 | 405 | if (dv->disposition == 'j') |
dfd7822c | 406 | goto skip_size_check; /* skip write journal for size check */ |
cc1799c3 | 407 | |
82d9eba6 | 408 | freesize /= 2; /* convert to K */ |
99cc42f4 | 409 | if (s->chunk && s->chunk != UnSet) { |
d2cd3ffc | 410 | /* round to chunk size */ |
99cc42f4 | 411 | freesize = freesize & ~(s->chunk-1); |
64385908 CA |
412 | if (do_default_chunk) { |
413 | /* default chunk was just set */ | |
171dccc8 | 414 | if (c->verbose > 0) |
7a862a02 | 415 | pr_err("chunk size defaults to %dK\n", s->chunk); |
99cc42f4 | 416 | s->size &= ~(unsigned long long)(s->chunk - 1); |
64385908 CA |
417 | do_default_chunk = 0; |
418 | } | |
d2cd3ffc | 419 | } |
066e92f0 LD |
420 | if (!freesize) { |
421 | pr_err("no free space left on %s\n", dname); | |
422 | fail = 1; | |
423 | continue; | |
424 | } | |
682c7051 | 425 | |
99cc42f4 | 426 | if (s->size && freesize < s->size) { |
7a862a02 | 427 | pr_err("%s is smaller than given size. %lluK < %lluK + metadata\n", |
99cc42f4 | 428 | dname, freesize, s->size); |
52826846 | 429 | fail = 1; |
52826846 | 430 | continue; |
682c7051 | 431 | } |
cd29a5c8 NB |
432 | if (maxdisc == NULL || (maxdisc && freesize > maxsize)) { |
433 | maxdisc = dname; | |
52826846 | 434 | maxsize = freesize; |
682c7051 | 435 | } |
cd29a5c8 NB |
436 | if (mindisc ==NULL || (mindisc && freesize < minsize)) { |
437 | mindisc = dname; | |
52826846 | 438 | minsize = freesize; |
682c7051 | 439 | } |
dfd7822c | 440 | skip_size_check: |
171dccc8 | 441 | if (c->runstop != 1 || c->verbose >= 0) { |
37ea3936 | 442 | int fd = open(dname, O_RDONLY); |
98dbf73c | 443 | if (fd < 0) { |
e7b84f9d | 444 | pr_err("Cannot open %s: %s\n", |
b6e63da4 | 445 | dname, strerror(errno)); |
98dbf73c | 446 | fail = 1; |
b6e63da4 NB |
447 | continue; |
448 | } | |
dab6685f NB |
449 | warn |= check_ext2(fd, dname); |
450 | warn |= check_reiser(fd, dname); | |
451 | warn |= check_raid(fd, dname); | |
034b203a TM |
452 | if (strcmp(st->ss->name, "1.x") == 0 && |
453 | st->minor_version >= 1) | |
454 | /* metadata at front */ | |
53ed6ac3 | 455 | warn |= check_partitions(fd, dname, 0, 0); |
cf622ec1 JS |
456 | else if (s->level == 1 || s->level == LEVEL_CONTAINER || |
457 | (s->level == 0 && s->raiddisks == 1)) | |
034b203a | 458 | /* partitions could be meaningful */ |
99cc42f4 | 459 | warn |= check_partitions(fd, dname, freesize*2, s->size*2); |
034b203a TM |
460 | else |
461 | /* partitions cannot be meaningful */ | |
53ed6ac3 | 462 | warn |= check_partitions(fd, dname, 0, 0); |
034b203a | 463 | if (strcmp(st->ss->name, "1.x") == 0 && |
a0962fe9 N |
464 | st->minor_version >= 1 && |
465 | did_default && | |
99cc42f4 | 466 | s->level == 1 && |
034b203a TM |
467 | (warn & 1024) == 0) { |
468 | warn |= 1024; | |
e7b84f9d | 469 | pr_err("Note: this array has metadata at the start and\n" |
a0962fe9 | 470 | " may not be suitable as a boot device. If you plan to\n" |
cc86f89c | 471 | " store '/boot' on this device please ensure that\n" |
a0962fe9 | 472 | " your boot-loader understands md/v1.x metadata, or use\n" |
cc86f89c | 473 | " --metadata=0.90\n"); |
a0962fe9 | 474 | } |
b6e63da4 | 475 | close(fd); |
dab6685f | 476 | } |
682c7051 | 477 | } |
99cc42f4 | 478 | if (s->raiddisks + s->sparedisks > st->max_devs) { |
7a862a02 | 479 | pr_err("Too many devices: %s metadata only supports %d\n", |
acab7bb1 N |
480 | st->ss->name, st->max_devs); |
481 | return 1; | |
482 | } | |
8592f29d | 483 | if (have_container) |
99cc42f4 | 484 | info.array.working_disks = s->raiddisks; |
682c7051 | 485 | if (fail) { |
e7b84f9d | 486 | pr_err("create aborted\n"); |
52826846 | 487 | return 1; |
682c7051 | 488 | } |
99cc42f4 | 489 | if (s->size == 0) { |
5f8097be | 490 | if (mindisc == NULL && !have_container) { |
e7b84f9d | 491 | pr_err("no size and no drives given - aborting create.\n"); |
52826846 NB |
492 | return 1; |
493 | } | |
cf622ec1 | 494 | if (s->level > 0 || s->level == LEVEL_MULTIPATH || |
98dbf73c | 495 | s->level == LEVEL_FAULTY || st->ss->external) { |
b5e64645 | 496 | /* size is meaningful */ |
99cc42f4 N |
497 | if (!st->ss->validate_geometry(st, s->level, s->layout, |
498 | s->raiddisks, | |
499 | &s->chunk, minsize*2, | |
40c9a66a | 500 | data_offset, |
5308f117 AP |
501 | NULL, NULL, |
502 | s->consistency_policy, 0)) { | |
99cc42f4 | 503 | pr_err("devices too large for RAID level %d\n", s->level); |
b5e64645 NB |
504 | return 1; |
505 | } | |
99cc42f4 N |
506 | s->size = minsize; |
507 | if (s->level == 1) | |
ae6c05ad N |
508 | /* If this is ever reshaped to RAID5, we will |
509 | * need a chunksize. So round it off a bit | |
510 | * now just to be safe | |
511 | */ | |
99cc42f4 | 512 | s->size &= ~(64ULL-1); |
171dccc8 | 513 | if (c->verbose > 0) |
99cc42f4 | 514 | pr_err("size set to %lluK\n", s->size); |
b5e64645 | 515 | } |
682c7051 | 516 | } |
748952f7 N |
517 | |
518 | if (!s->bitmap_file && | |
519 | s->level >= 1 && | |
39917e56 | 520 | st->ss->add_internal_bitmap && |
e97a7cd0 AP |
521 | (s->consistency_policy != CONSISTENCY_POLICY_RESYNC && |
522 | s->consistency_policy != CONSISTENCY_POLICY_PPL) && | |
748952f7 N |
523 | (s->write_behind || s->size > 100*1024*1024ULL)) { |
524 | if (c->verbose > 0) | |
525 | pr_err("automatically enabling write-intent bitmap on large array\n"); | |
526 | s->bitmap_file = "internal"; | |
527 | } | |
528 | if (s->bitmap_file && strcmp(s->bitmap_file, "none") == 0) | |
529 | s->bitmap_file = NULL; | |
530 | ||
5308f117 AP |
531 | if (s->consistency_policy == CONSISTENCY_POLICY_PPL && |
532 | !st->ss->write_init_ppl) { | |
533 | pr_err("%s metadata does not support PPL\n", st->ss->name); | |
534 | return 1; | |
535 | } | |
536 | ||
99cc42f4 | 537 | if (!have_container && s->level > 0 && ((maxsize-s->size)*100 > maxsize)) { |
171dccc8 | 538 | if (c->runstop != 1 || c->verbose >= 0) |
e7b84f9d | 539 | pr_err("largest drive (%s) exceeds size (%lluK) by more than 1%%\n", |
99cc42f4 | 540 | maxdisc, s->size); |
52826846 | 541 | warn = 1; |
682c7051 NB |
542 | } |
543 | ||
9eafa1de | 544 | if (st->ss->detail_platform && st->ss->detail_platform(0, 1, NULL) != 0) { |
171dccc8 | 545 | if (c->runstop != 1 || c->verbose >= 0) |
e7b84f9d | 546 | pr_err("%s unable to enumerate platform support\n" |
5615172f DW |
547 | " array may not be compatible with hardware/firmware\n", |
548 | st->ss->name); | |
549 | warn = 1; | |
550 | } | |
529e2aa5 | 551 | st->nodes = c->nodes; |
7716570e | 552 | st->cluster_name = c->homecluster; |
5615172f | 553 | |
682c7051 | 554 | if (warn) { |
171dccc8 | 555 | if (c->runstop!= 1) { |
52826846 | 556 | if (!ask("Continue creating array? ")) { |
e7b84f9d | 557 | pr_err("create aborted.\n"); |
52826846 NB |
558 | return 1; |
559 | } | |
560 | } else { | |
171dccc8 | 561 | if (c->verbose > 0) |
e7b84f9d | 562 | pr_err("creation continuing despite oddities due to --run\n"); |
682c7051 | 563 | } |
682c7051 NB |
564 | } |
565 | ||
66f8bbbe | 566 | /* If this is raid4/5, we want to configure the last active slot |
52826846 | 567 | * as missing, so that a reconstruct happens (faster than re-parity) |
98c6faba | 568 | * FIX: Can we do this for raid6 as well? |
52826846 | 569 | */ |
98dbf73c JS |
570 | if (st->ss->external == 0 && s->assume_clean == 0 && |
571 | c->force == 0 && first_missing >= s->raiddisks) { | |
572 | switch (s->level) { | |
66f8bbbe | 573 | case 4: |
98c6faba | 574 | case 5: |
99cc42f4 N |
575 | insert_point = s->raiddisks-1; |
576 | s->sparedisks++; | |
06c7f68e | 577 | info.array.active_disks--; |
98c6faba NB |
578 | missing_disks++; |
579 | break; | |
580 | default: | |
581 | break; | |
582 | } | |
52826846 | 583 | } |
6fb79233 NB |
584 | /* For raid6, if creating with 1 missing drive, make a good drive |
585 | * into a spare, else the create will fail | |
586 | */ | |
99cc42f4 | 587 | if (s->assume_clean == 0 && c->force == 0 && first_missing < s->raiddisks && |
ffcfc735 | 588 | st->ss->external == 0 && |
99cc42f4 N |
589 | second_missing >= s->raiddisks && s->level == 6) { |
590 | insert_point = s->raiddisks - 1; | |
6fb79233 NB |
591 | if (insert_point == first_missing) |
592 | insert_point--; | |
99cc42f4 | 593 | s->sparedisks ++; |
6fb79233 NB |
594 | info.array.active_disks--; |
595 | missing_disks++; | |
596 | } | |
570510ba | 597 | |
99cc42f4 | 598 | if (s->level <= 0 && first_missing < subdevs * 2) { |
e7b84f9d | 599 | pr_err("This level does not support missing devices\n"); |
570510ba NB |
600 | return 1; |
601 | } | |
aba69144 | 602 | |
7f91af49 | 603 | /* We need to create the device */ |
ad5bc697 | 604 | map_lock(&map); |
cd6cbb08 | 605 | mdfd = create_mddev(mddev, name, c->autof, LOCAL, chosen_name, 1); |
e06af9dd JS |
606 | if (mdfd < 0) { |
607 | map_unlock(&map); | |
7f91af49 | 608 | return 1; |
e06af9dd | 609 | } |
3e9df86a AK |
610 | /* verify if chosen_name is not in use, |
611 | * it could be in conflict with already existing device | |
612 | * e.g. container, array | |
613 | */ | |
cf622ec1 JS |
614 | if (strncmp(chosen_name, "/dev/md/", 8) == 0 && |
615 | map_by_name(&map, chosen_name+8) != NULL) { | |
e7b84f9d | 616 | pr_err("Array name %s is in use already.\n", |
3e9df86a AK |
617 | chosen_name); |
618 | close(mdfd); | |
619 | map_unlock(&map); | |
cd6cbb08 | 620 | udev_unblock(); |
3e9df86a AK |
621 | return 1; |
622 | } | |
f2be09f1 | 623 | mddev = chosen_name; |
7f91af49 | 624 | |
5f4cc239 JS |
625 | memset(&inf, 0, sizeof(inf)); |
626 | md_get_array_info(mdfd, &inf); | |
627 | if (inf.working_disks != 0) { | |
628 | pr_err("another array by this name is already running.\n"); | |
e06af9dd | 629 | goto abort_locked; |
7f91af49 | 630 | } |
a04d5763 | 631 | |
682c7051 NB |
632 | /* Ok, lets try some ioctls */ |
633 | ||
99cc42f4 N |
634 | info.array.level = s->level; |
635 | info.array.size = s->size; | |
636 | info.array.raid_disks = s->raiddisks; | |
82b27616 NB |
637 | /* The kernel should *know* what md_minor we are dealing |
638 | * with, but it chooses to trust me instead. Sigh | |
639 | */ | |
06c7f68e | 640 | info.array.md_minor = 0; |
0a6bff09 ZL |
641 | if (fstat_is_blkdev(mdfd, mddev, &rdev)) |
642 | info.array.md_minor = minor(rdev); | |
06c7f68e | 643 | info.array.not_persistent = 0; |
6fb79233 | 644 | |
cf622ec1 JS |
645 | if (((s->level == 4 || s->level == 5) && |
646 | (insert_point < s->raiddisks || first_missing < s->raiddisks)) || | |
647 | (s->level == 6 && (insert_point < s->raiddisks || | |
648 | second_missing < s->raiddisks)) || | |
649 | (s->level <= 0) || s->assume_clean) { | |
06c7f68e | 650 | info.array.state = 1; /* clean, but one+ drive will be missing*/ |
b7528a20 | 651 | info.resync_start = MaxSector; |
103f2410 | 652 | } else { |
06c7f68e | 653 | info.array.state = 0; /* not clean, but no errors */ |
103f2410 NB |
654 | info.resync_start = 0; |
655 | } | |
99cc42f4 | 656 | if (s->level == 10) { |
f9c25f1d NB |
657 | /* for raid10, the bitmap size is the capacity of the array, |
658 | * which is array.size * raid_disks / ncopies; | |
659 | * .. but convert to sectors. | |
660 | */ | |
99cc42f4 N |
661 | int ncopies = ((s->layout>>8) & 255) * (s->layout & 255); |
662 | bitmapsize = s->size * s->raiddisks / ncopies * 2; | |
663 | /* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, s->size, s->raiddisks, ncopies);*/ | |
f9c25f1d | 664 | } else |
99cc42f4 | 665 | bitmapsize = s->size * 2; |
f9c25f1d | 666 | |
82b27616 NB |
667 | /* There is lots of redundancy in these disk counts, |
668 | * raid_disks is the most meaningful value | |
669 | * it describes the geometry of the array | |
670 | * it is constant | |
671 | * nr_disks is total number of used slots. | |
672 | * it should be raid_disks+spare_disks | |
673 | * spare_disks is the number of extra disks present | |
674 | * see above | |
675 | * active_disks is the number of working disks in | |
676 | * active slots. (With raid_disks) | |
677 | * working_disks is the total number of working disks, | |
678 | * including spares | |
679 | * failed_disks is the number of disks marked failed | |
680 | * | |
5d500228 | 681 | * Ideally, the kernel would keep these (except raid_disks) |
82b27616 NB |
682 | * up-to-date as we ADD_NEW_DISK, but it doesn't (yet). |
683 | * So for now, we assume that all raid and spare | |
684 | * devices will be given. | |
685 | */ | |
99cc42f4 | 686 | info.array.spare_disks=s->sparedisks; |
06c7f68e NB |
687 | info.array.failed_disks=missing_disks; |
688 | info.array.nr_disks = info.array.working_disks | |
689 | + info.array.failed_disks; | |
99cc42f4 N |
690 | info.array.layout = s->layout; |
691 | info.array.chunk_size = s->chunk*1024; | |
82d9eba6 | 692 | |
b3b33eb5 NB |
693 | if (name == NULL || *name == 0) { |
694 | /* base name on mddev */ | |
d1e80164 NB |
695 | /* /dev/md0 -> 0 |
696 | * /dev/md_d0 -> d0 | |
eca944fa | 697 | * /dev/md_foo -> foo |
d1e80164 NB |
698 | * /dev/md/1 -> 1 |
699 | * /dev/md/d1 -> d1 | |
700 | * /dev/md/home -> home | |
701 | * /dev/mdhome -> home | |
702 | */ | |
69207ff6 | 703 | /* FIXME compare this with rules in create_mddev */ |
b3b33eb5 NB |
704 | name = strrchr(mddev, '/'); |
705 | if (name) { | |
706 | name++; | |
98dbf73c JS |
707 | if (strncmp(name, "md_", 3) == 0 && |
708 | strlen(name) > 3 && (name-mddev) == 5 /* /dev/ */) | |
d1e80164 | 709 | name += 3; |
98dbf73c JS |
710 | else if (strncmp(name, "md", 2) == 0 && |
711 | strlen(name) > 2 && isdigit(name[2]) && | |
60248f74 | 712 | (name-mddev) == 5 /* /dev/ */) |
b3b33eb5 NB |
713 | name += 2; |
714 | } | |
715 | } | |
5308f117 | 716 | if (!st->ss->init_super(st, &info.array, s, name, c->homehost, uuid, |
40c9a66a | 717 | data_offset)) |
e06af9dd | 718 | goto abort_locked; |
682c7051 | 719 | |
d2ca6449 | 720 | total_slots = info.array.nr_disks; |
a5d85af7 | 721 | st->ss->getinfo_super(st, &info, NULL); |
dae13137 JS |
722 | if (sysfs_init(&info, mdfd, NULL)) { |
723 | pr_err("unable to initialize sysfs\n"); | |
724 | goto abort_locked; | |
725 | } | |
159c3a1a | 726 | |
171dccc8 | 727 | if (did_default && c->verbose >= 0) { |
3c558363 | 728 | if (is_subarray(info.text_version)) { |
4dd2df09 N |
729 | char devnm[32]; |
730 | char *ep; | |
f7f1b6a1 | 731 | struct mdinfo *mdi; |
f7f1b6a1 | 732 | |
4dd2df09 N |
733 | strncpy(devnm, info.text_version+1, 32); |
734 | devnm[31] = 0; | |
735 | ep = strchr(devnm, '/'); | |
736 | if (ep) | |
737 | *ep = 0; | |
738 | ||
739 | mdi = sysfs_read(-1, devnm, GET_VERSION); | |
f7f1b6a1 | 740 | |
4dd2df09 N |
741 | pr_err("Creating array inside %s container %s\n", |
742 | mdi?mdi->text_version:"managed", devnm); | |
f7f1b6a1 NB |
743 | sysfs_free(mdi); |
744 | } else | |
7a862a02 | 745 | pr_err("Defaulting to version %s metadata\n", info.text_version); |
f7f1b6a1 | 746 | } |
b8ac1967 | 747 | |
4dd2df09 | 748 | map_update(&map, fd2devnm(mdfd), info.text_version, |
a04d5763 | 749 | info.uuid, chosen_name); |
19ad4b2c AP |
750 | /* Keep map locked until devices have been added to array |
751 | * to stop another mdadm from finding and using those devices. | |
752 | */ | |
a04d5763 | 753 | |
98dbf73c JS |
754 | if (s->bitmap_file && (strcmp(s->bitmap_file, "internal") == 0 || |
755 | strcmp(s->bitmap_file, "clustered") == 0)) { | |
ebeb3663 | 756 | if (!st->ss->add_internal_bitmap) { |
e7b84f9d | 757 | pr_err("internal bitmaps not supported with %s metadata\n", |
ebeb3663 | 758 | st->ss->name); |
19ad4b2c | 759 | goto abort_locked; |
ebeb3663 | 760 | } |
2ec2b7e9 JS |
761 | if (st->ss->add_internal_bitmap(st, &s->bitmap_chunk, |
762 | c->delay, s->write_behind, | |
763 | bitmapsize, 1, major_num)) { | |
e7b84f9d | 764 | pr_err("Given bitmap chunk size not supported.\n"); |
19ad4b2c | 765 | goto abort_locked; |
55935d51 | 766 | } |
99cc42f4 | 767 | s->bitmap_file = NULL; |
55935d51 NB |
768 | } |
769 | ||
dae13137 JS |
770 | if (sysfs_init(&info, mdfd, NULL)) { |
771 | pr_err("unable to initialize sysfs\n"); | |
772 | goto abort_locked; | |
773 | } | |
55935d51 | 774 | |
4dd2df09 | 775 | if (st->ss->external && st->container_devnm[0]) { |
f35f2525 N |
776 | /* member */ |
777 | ||
778 | /* When creating a member, we need to be careful | |
779 | * to negotiate with mdmon properly. | |
780 | * If it is already running, we cannot write to | |
781 | * the devices and must ask it to do that part. | |
782 | * If it isn't running, we write to the devices, | |
783 | * and then start it. | |
784 | * We hold an exclusive open on the container | |
785 | * device to make sure mdmon doesn't exit after | |
786 | * we checked that it is running. | |
787 | * | |
788 | * For now, fail if it is already running. | |
789 | */ | |
4dd2df09 | 790 | container_fd = open_dev_excl(st->container_devnm); |
f35f2525 | 791 | if (container_fd < 0) { |
7a862a02 | 792 | pr_err("Cannot get exclusive open on container - weird.\n"); |
19ad4b2c | 793 | goto abort_locked; |
d03373f1 | 794 | } |
4dd2df09 | 795 | if (mdmon_running(st->container_devnm)) { |
171dccc8 | 796 | if (c->verbose) |
7a862a02 | 797 | pr_err("reusing mdmon for %s.\n", |
4dd2df09 | 798 | st->container_devnm); |
f35f2525 N |
799 | st->update_tail = &st->updates; |
800 | } else | |
801 | need_mdmon = 1; | |
802 | } | |
803 | rv = set_array_info(mdfd, st, &info); | |
82d9eba6 | 804 | if (rv) { |
e7b84f9d | 805 | pr_err("failed to set array info for %s: %s\n", |
52826846 | 806 | mddev, strerror(errno)); |
19ad4b2c | 807 | goto abort_locked; |
682c7051 | 808 | } |
52826846 | 809 | |
99cc42f4 | 810 | if (s->bitmap_file) { |
c82f047c | 811 | int uuid[4]; |
55935d51 | 812 | |
3da92f27 | 813 | st->ss->uuid_from_super(st, uuid); |
99cc42f4 N |
814 | if (CreateBitmap(s->bitmap_file, c->force, (char*)uuid, s->bitmap_chunk, |
815 | c->delay, s->write_behind, | |
f9c25f1d | 816 | bitmapsize, |
943eafef | 817 | major_num)) { |
19ad4b2c | 818 | goto abort_locked; |
c82f047c | 819 | } |
99cc42f4 | 820 | bitmap_fd = open(s->bitmap_file, O_RDWR); |
c82f047c | 821 | if (bitmap_fd < 0) { |
e7b84f9d | 822 | pr_err("weird: %s cannot be openned\n", |
99cc42f4 | 823 | s->bitmap_file); |
19ad4b2c | 824 | goto abort_locked; |
c82f047c NB |
825 | } |
826 | if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) { | |
e7b84f9d | 827 | pr_err("Cannot set bitmap file for %s: %s\n", |
c82f047c | 828 | mddev, strerror(errno)); |
19ad4b2c | 829 | goto abort_locked; |
c82f047c NB |
830 | } |
831 | } | |
832 | ||
503975b9 | 833 | infos = xmalloc(sizeof(*infos) * total_slots); |
a7dec3fd | 834 | enable_fds(total_slots); |
98dbf73c | 835 | for (pass = 1; pass <= 2; pass++) { |
a655e550 | 836 | struct mddev_dev *moved_disk = NULL; /* the disk that was moved out of the insert point */ |
4b1ac34b | 837 | |
98dbf73c JS |
838 | for (dnum = 0, raid_disk_num = 0, dv = devlist; dv; |
839 | dv = (dv->next) ? (dv->next) : moved_disk, dnum++) { | |
4b1ac34b | 840 | int fd; |
d2ca6449 | 841 | struct mdinfo *inf = &infos[dnum]; |
4b1ac34b | 842 | |
d2ca6449 NB |
843 | if (dnum >= total_slots) |
844 | abort(); | |
4b1ac34b | 845 | if (dnum == insert_point) { |
f170a5a9 | 846 | raid_disk_num += 1; |
4b1ac34b | 847 | moved_disk = dv; |
8592f29d | 848 | continue; |
52826846 | 849 | } |
98dbf73c | 850 | if (strcasecmp(dv->devname, "missing") == 0) { |
f170a5a9 | 851 | raid_disk_num += 1; |
111d01fc | 852 | continue; |
f170a5a9 | 853 | } |
8592f29d N |
854 | if (have_container) |
855 | moved_disk = NULL; | |
856 | if (have_container && dnum < info.array.raid_disks - 1) | |
857 | /* repeatedly use the container */ | |
858 | moved_disk = dv; | |
111d01fc | 859 | |
d2ca6449 NB |
860 | switch(pass) { |
861 | case 1: | |
862 | *inf = info; | |
5f8097be | 863 | |
d2ca6449 | 864 | inf->disk.number = dnum; |
cc1799c3 SL |
865 | inf->disk.raid_disk = raid_disk_num++; |
866 | ||
867 | if (dv->disposition == 'j') { | |
868 | inf->disk.raid_disk = MD_DISK_ROLE_JOURNAL; | |
869 | inf->disk.state = (1<<MD_DISK_JOURNAL); | |
870 | raid_disk_num--; | |
871 | } else if (inf->disk.raid_disk < s->raiddisks) | |
d2ca6449 NB |
872 | inf->disk.state = (1<<MD_DISK_ACTIVE) | |
873 | (1<<MD_DISK_SYNC); | |
874 | else | |
875 | inf->disk.state = 0; | |
876 | ||
e22fe3ae | 877 | if (dv->writemostly == FlagSet) |
d2ca6449 | 878 | inf->disk.state |= (1<<MD_DISK_WRITEMOSTLY); |
e22fe3ae | 879 | if (dv->failfast == FlagSet) |
71574efb | 880 | inf->disk.state |= (1<<MD_DISK_FAILFAST); |
d2ca6449 | 881 | |
8592f29d N |
882 | if (have_container) |
883 | fd = -1; | |
884 | else { | |
d1d599ea | 885 | if (st->ss->external && |
4dd2df09 | 886 | st->container_devnm[0]) |
8592f29d N |
887 | fd = open(dv->devname, O_RDWR); |
888 | else | |
889 | fd = open(dv->devname, O_RDWR|O_EXCL); | |
890 | ||
891 | if (fd < 0) { | |
7a862a02 | 892 | pr_err("failed to open %s after earlier success - aborting\n", |
8592f29d | 893 | dv->devname); |
19ad4b2c | 894 | goto abort_locked; |
8592f29d | 895 | } |
0a6bff09 ZL |
896 | if (!fstat_is_blkdev(fd, dv->devname, &rdev)) |
897 | return 1; | |
898 | inf->disk.major = major(rdev); | |
899 | inf->disk.minor = minor(rdev); | |
d2ca6449 | 900 | } |
8592f29d N |
901 | if (fd >= 0) |
902 | remove_partitions(fd); | |
f20c3968 | 903 | if (st->ss->add_to_super(st, &inf->disk, |
72ca9bcf N |
904 | fd, dv->devname, |
905 | dv->data_offset)) { | |
4eb26970 | 906 | ioctl(mdfd, STOP_ARRAY, NULL); |
19ad4b2c | 907 | goto abort_locked; |
4eb26970 | 908 | } |
a5d85af7 | 909 | st->ss->getinfo_super(st, inf, NULL); |
8ed3e5e1 | 910 | safe_mode_delay = inf->safe_mode_delay; |
d2ca6449 | 911 | |
171dccc8 | 912 | if (have_container && c->verbose > 0) |
e7b84f9d | 913 | pr_err("Using %s for device %d\n", |
8592f29d N |
914 | map_dev(inf->disk.major, |
915 | inf->disk.minor, | |
916 | 0), dnum); | |
917 | ||
918 | if (!have_container) { | |
919 | /* getinfo_super might have lost these ... */ | |
0a6bff09 ZL |
920 | inf->disk.major = major(rdev); |
921 | inf->disk.minor = minor(rdev); | |
8592f29d | 922 | } |
4b1ac34b NB |
923 | break; |
924 | case 2: | |
d2ca6449 | 925 | inf->errors = 0; |
4b1ac34b | 926 | |
f35f2525 | 927 | rv = add_disk(mdfd, st, &info, inf); |
d2ca6449 | 928 | |
2503d23b | 929 | if (rv) { |
7a862a02 | 930 | pr_err("ADD_NEW_DISK for %s failed: %s\n", |
e7b84f9d | 931 | dv->devname, strerror(errno)); |
19ad4b2c | 932 | goto abort_locked; |
4b1ac34b | 933 | } |
4b1ac34b NB |
934 | break; |
935 | } | |
8592f29d N |
936 | if (!have_container && |
937 | dv == moved_disk && dnum != insert_point) break; | |
52826846 | 938 | } |
edd8d13c | 939 | if (pass == 1) { |
9b1fb677 DW |
940 | struct mdinfo info_new; |
941 | struct map_ent *me = NULL; | |
942 | ||
943 | /* check to see if the uuid has changed due to these | |
944 | * metadata changes, and if so update the member array | |
945 | * and container uuid. Note ->write_init_super clears | |
946 | * the subarray cursor such that ->getinfo_super once | |
947 | * again returns container info. | |
948 | */ | |
a5d85af7 | 949 | st->ss->getinfo_super(st, &info_new, NULL); |
99cc42f4 | 950 | if (st->ss->external && s->level != LEVEL_CONTAINER && |
9b1fb677 | 951 | !same_uuid(info_new.uuid, info.uuid, 0)) { |
4dd2df09 | 952 | map_update(&map, fd2devnm(mdfd), |
9b1fb677 DW |
953 | info_new.text_version, |
954 | info_new.uuid, chosen_name); | |
4dd2df09 | 955 | me = map_by_devnm(&map, st->container_devnm); |
9b1fb677 DW |
956 | } |
957 | ||
ab80e597 | 958 | if (st->ss->write_init_super(st)) { |
ab80e597 | 959 | st->ss->free_super(st); |
e06af9dd | 960 | goto abort_locked; |
ab80e597 | 961 | } |
9b1fb677 DW |
962 | |
963 | /* update parent container uuid */ | |
964 | if (me) { | |
503975b9 | 965 | char *path = xstrdup(me->path); |
9b1fb677 | 966 | |
a5d85af7 | 967 | st->ss->getinfo_super(st, &info_new, NULL); |
4dd2df09 | 968 | map_update(&map, st->container_devnm, |
9b1fb677 DW |
969 | info_new.text_version, |
970 | info_new.uuid, path); | |
971 | free(path); | |
972 | } | |
9b1fb677 | 973 | |
edd8d13c | 974 | flush_metadata_updates(st); |
6946681d | 975 | st->ss->free_super(st); |
edd8d13c | 976 | } |
5e52ae9e | 977 | } |
19ad4b2c | 978 | map_unlock(&map); |
d2ca6449 | 979 | free(infos); |
5e52ae9e | 980 | |
99cc42f4 | 981 | if (s->level == LEVEL_CONTAINER) { |
97590376 N |
982 | /* No need to start. But we should signal udev to |
983 | * create links */ | |
984 | sysfs_uevent(&info, "change"); | |
171dccc8 | 985 | if (c->verbose >= 0) |
e7b84f9d | 986 | pr_err("container %s prepared.\n", mddev); |
a7c6e3fb | 987 | wait_for(chosen_name, mdfd); |
99cc42f4 | 988 | } else if (c->runstop == 1 || subdevs >= s->raiddisks) { |
598f0d58 | 989 | if (st->ss->external) { |
4c821454 | 990 | int err; |
99cc42f4 | 991 | switch(s->level) { |
598f0d58 NB |
992 | case LEVEL_LINEAR: |
993 | case LEVEL_MULTIPATH: | |
994 | case 0: | |
4c821454 | 995 | err = sysfs_set_str(&info, NULL, "array_state", |
171dccc8 | 996 | c->readonly |
72d566f6 N |
997 | ? "readonly" |
998 | : "active"); | |
a931db9e | 999 | need_mdmon = 0; |
598f0d58 NB |
1000 | break; |
1001 | default: | |
4c821454 N |
1002 | err = sysfs_set_str(&info, NULL, "array_state", |
1003 | "readonly"); | |
598f0d58 NB |
1004 | break; |
1005 | } | |
f35f2525 | 1006 | sysfs_set_safemode(&info, safe_mode_delay); |
4c821454 | 1007 | if (err) { |
7a862a02 | 1008 | pr_err("failed to activate array.\n"); |
4c821454 N |
1009 | ioctl(mdfd, STOP_ARRAY, NULL); |
1010 | goto abort; | |
1011 | } | |
171dccc8 | 1012 | } else if (c->readonly && |
72d566f6 N |
1013 | sysfs_attribute_available( |
1014 | &info, NULL, "array_state")) { | |
1015 | if (sysfs_set_str(&info, NULL, | |
1016 | "array_state", "readonly") < 0) { | |
1017 | pr_err("Failed to start array: %s\n", | |
1018 | strerror(errno)); | |
1019 | ioctl(mdfd, STOP_ARRAY, NULL); | |
1020 | goto abort; | |
1021 | } | |
598f0d58 | 1022 | } else { |
97590376 | 1023 | /* param is not actually used */ |
598f0d58 NB |
1024 | mdu_param_t param; |
1025 | if (ioctl(mdfd, RUN_ARRAY, ¶m)) { | |
e7b84f9d | 1026 | pr_err("RUN_ARRAY failed: %s\n", |
72d566f6 | 1027 | strerror(errno)); |
a252c078 | 1028 | if (info.array.chunk_size & (info.array.chunk_size-1)) { |
7a862a02 | 1029 | cont_err("Problem may be that chunk size is not a power of 2\n"); |
a252c078 | 1030 | } |
4eb26970 | 1031 | ioctl(mdfd, STOP_ARRAY, NULL); |
7f91af49 | 1032 | goto abort; |
598f0d58 | 1033 | } |
a21e848a N |
1034 | /* if start_ro module parameter is set, array is |
1035 | * auto-read-only, which is bad as the resync won't | |
1036 | * start. So lets make it read-write now. | |
1037 | */ | |
1038 | ioctl(mdfd, RESTART_ARRAY_RW, NULL); | |
682c7051 | 1039 | } |
171dccc8 | 1040 | if (c->verbose >= 0) |
e7b84f9d | 1041 | pr_err("array %s started.\n", mddev); |
4dd2df09 | 1042 | if (st->ss->external && st->container_devnm[0]) { |
8850ee3e | 1043 | if (need_mdmon) |
4dd2df09 | 1044 | start_mdmon(st->container_devnm); |
8850ee3e | 1045 | |
4dd2df09 | 1046 | ping_monitor(st->container_devnm); |
a931db9e NB |
1047 | close(container_fd); |
1048 | } | |
a7c6e3fb | 1049 | wait_for(chosen_name, mdfd); |
682c7051 | 1050 | } else { |
e7b84f9d | 1051 | pr_err("not starting array - not enough devices.\n"); |
682c7051 | 1052 | } |
7f91af49 | 1053 | close(mdfd); |
cd6cbb08 N |
1054 | /* Give udev a moment to process the Change event caused |
1055 | * by the close. | |
1056 | */ | |
1057 | usleep(100*1000); | |
1058 | udev_unblock(); | |
682c7051 | 1059 | return 0; |
7f91af49 N |
1060 | |
1061 | abort: | |
cd6cbb08 | 1062 | udev_unblock(); |
4eb26970 | 1063 | map_lock(&map); |
e06af9dd | 1064 | abort_locked: |
4dd2df09 | 1065 | map_remove(&map, fd2devnm(mdfd)); |
4eb26970 DW |
1066 | map_unlock(&map); |
1067 | ||
7f91af49 N |
1068 | if (mdfd >= 0) |
1069 | close(mdfd); | |
1070 | return 1; | |
64c4757e | 1071 | } |