]>
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; |
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; |
6d9c7c25 GR |
117 | if (s->bitmap_file && strcmp(s->bitmap_file, "clustered") == 0) |
118 | major_num = BITMAP_MAJOR_CLUSTERED; | |
dcec9ee5 | 119 | |
06c7f68e | 120 | memset(&info, 0, sizeof(info)); |
99cc42f4 N |
121 | if (s->level == UnSet && st && st->ss->default_geometry) |
122 | st->ss->default_geometry(st, &s->level, NULL, NULL); | |
123 | if (s->level == UnSet) { | |
e7b84f9d | 124 | pr_err("a RAID level is needed to create an array.\n"); |
682c7051 NB |
125 | return 1; |
126 | } | |
99cc42f4 | 127 | if (s->raiddisks < 4 && s->level == 6) { |
e7b84f9d | 128 | pr_err("at least 4 raid-devices needed for level 6\n"); |
98c6faba NB |
129 | return 1; |
130 | } | |
99cc42f4 | 131 | if (s->raiddisks > 256 && s->level == 6) { |
e7b84f9d | 132 | pr_err("no more than 256 raid-devices supported for level 6\n"); |
98c6faba NB |
133 | return 1; |
134 | } | |
99cc42f4 | 135 | if (s->raiddisks < 2 && s->level >= 4) { |
e7b84f9d | 136 | pr_err("at least 2 raid-devices needed for level 4 or 5\n"); |
52826846 NB |
137 | return 1; |
138 | } | |
99cc42f4 | 139 | if (s->level <= 0 && s->sparedisks) { |
e7b84f9d | 140 | pr_err("This level does not support spare devices\n"); |
570510ba NB |
141 | return 1; |
142 | } | |
5f8097be NB |
143 | |
144 | if (subdevs == 1 && strcmp(devlist->devname, "missing") != 0) { | |
145 | /* If given a single device, it might be a container, and we can | |
146 | * extract a device list from there | |
147 | */ | |
148 | mdu_array_info_t inf; | |
149 | int fd; | |
150 | ||
151 | memset(&inf, 0, sizeof(inf)); | |
37ea3936 | 152 | fd = open(devlist->devname, O_RDONLY); |
5f8097be NB |
153 | if (fd >= 0 && |
154 | ioctl(fd, GET_ARRAY_INFO, &inf) == 0 && | |
155 | inf.raid_disks == 0) { | |
156 | /* yep, looks like a container */ | |
157 | if (st) { | |
0fb69d1d N |
158 | rv = st->ss->load_container(st, fd, |
159 | devlist->devname); | |
5f8097be NB |
160 | if (rv == 0) |
161 | have_container = 1; | |
162 | } else { | |
0fb69d1d | 163 | st = super_by_fd(fd, NULL); |
5f8097be | 164 | if (st && !(rv = st->ss-> |
0fb69d1d N |
165 | load_container(st, fd, |
166 | devlist->devname))) | |
5f8097be NB |
167 | have_container = 1; |
168 | else | |
169 | st = NULL; | |
170 | } | |
18fde300 | 171 | if (have_container) { |
99cc42f4 | 172 | subdevs = s->raiddisks; |
18fde300 DW |
173 | first_missing = subdevs * 2; |
174 | second_missing = subdevs * 2; | |
175 | insert_point = subdevs * 2; | |
176 | } | |
5f8097be NB |
177 | } |
178 | if (fd >= 0) | |
179 | close(fd); | |
5f8097be | 180 | } |
99cc42f4 | 181 | if (st && st->ss->external && s->sparedisks) { |
7a862a02 | 182 | pr_err("This metadata type does not support spare disks at create time\n"); |
ffcfc735 N |
183 | return 1; |
184 | } | |
99cc42f4 N |
185 | if (subdevs > s->raiddisks+s->sparedisks) { |
186 | pr_err("You have listed more devices (%d) than are in the array(%d)!\n", subdevs, s->raiddisks+s->sparedisks); | |
52826846 | 187 | return 1; |
682c7051 | 188 | } |
99cc42f4 | 189 | if (!have_container && subdevs < s->raiddisks+s->sparedisks) { |
e7b84f9d | 190 | pr_err("You haven't given enough devices (real or missing) to create this array\n"); |
52826846 NB |
191 | return 1; |
192 | } | |
99cc42f4 | 193 | if (s->bitmap_file && s->level <= 0) { |
e7b84f9d | 194 | pr_err("bitmaps not meaningful with level %s\n", |
99cc42f4 | 195 | map_num(pers, s->level)?:"given"); |
5b28bd56 NB |
196 | return 1; |
197 | } | |
52826846 | 198 | |
682c7051 | 199 | /* now set some defaults */ |
b5e64645 | 200 | |
99cc42f4 | 201 | if (s->layout == UnSet) { |
a18a888e | 202 | do_default_layout = 1; |
99cc42f4 | 203 | s->layout = default_layout(st, s->level, c->verbose); |
a18a888e | 204 | } |
682c7051 | 205 | |
99cc42f4 | 206 | if (s->level == 10) |
e5329c37 | 207 | /* check layout fits in array*/ |
99cc42f4 | 208 | if ((s->layout&255) * ((s->layout>>8)&255) > s->raiddisks) { |
e7b84f9d | 209 | pr_err("that layout requires at least %d devices\n", |
99cc42f4 | 210 | (s->layout&255) * ((s->layout>>8)&255)); |
e5329c37 NB |
211 | return 1; |
212 | } | |
213 | ||
99cc42f4 | 214 | switch(s->level) { |
aa88f531 NB |
215 | case 4: |
216 | case 5: | |
e5329c37 | 217 | case 10: |
98c6faba | 218 | case 6: |
aa88f531 | 219 | case 0: |
99cc42f4 N |
220 | if (s->chunk == 0 || s->chunk == UnSet) { |
221 | s->chunk = UnSet; | |
c21e737b CA |
222 | do_default_chunk = 1; |
223 | /* chunk will be set later */ | |
bb7295f1 | 224 | } |
5f175898 N |
225 | break; |
226 | case LEVEL_LINEAR: | |
227 | /* a chunksize of zero 0s perfectly valid (and preferred) since 2.6.16 */ | |
99cc42f4 N |
228 | if (get_linux_version() < 2006016 && s->chunk == 0) { |
229 | s->chunk = 64; | |
171dccc8 | 230 | if (c->verbose > 0) |
e7b84f9d | 231 | pr_err("chunk size defaults to 64K\n"); |
aa88f531 NB |
232 | } |
233 | break; | |
570510ba | 234 | case 1: |
08e43379 | 235 | case LEVEL_FAULTY: |
570510ba | 236 | case LEVEL_MULTIPATH: |
17f25ca6 | 237 | case LEVEL_CONTAINER: |
99cc42f4 N |
238 | if (s->chunk) { |
239 | s->chunk = 0; | |
171dccc8 | 240 | if (c->verbose > 0) |
e7b84f9d | 241 | pr_err("chunk size ignored for this level\n"); |
aa88f531 NB |
242 | } |
243 | break; | |
570510ba | 244 | default: |
99cc42f4 | 245 | pr_err("unknown level %d\n", s->level); |
570510ba | 246 | return 1; |
682c7051 | 247 | } |
99cc42f4 | 248 | if (s->size == MAX_SIZE) |
d04f65f4 | 249 | /* use '0' to mean 'max' now... */ |
99cc42f4 N |
250 | s->size = 0; |
251 | if (s->size && s->chunk && s->chunk != UnSet) | |
252 | s->size &= ~(unsigned long long)(s->chunk - 1); | |
253 | newsize = s->size * 2; | |
254 | if (st && ! st->ss->validate_geometry(st, s->level, s->layout, s->raiddisks, | |
af4348dd | 255 | &s->chunk, s->size*2, |
40c9a66a | 256 | data_offset, NULL, |
af4348dd | 257 | &newsize, c->verbose>=0)) |
17f25ca6 | 258 | return 1; |
bb7295f1 | 259 | |
99cc42f4 N |
260 | if (s->chunk && s->chunk != UnSet) { |
261 | newsize &= ~(unsigned long long)(s->chunk*2 - 1); | |
64385908 CA |
262 | if (do_default_chunk) { |
263 | /* default chunk was just set */ | |
171dccc8 | 264 | if (c->verbose > 0) |
7a862a02 | 265 | pr_err("chunk size defaults to %dK\n", s->chunk); |
99cc42f4 | 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; |
476066a3 N |
288 | for (dv = devlist; dv ; dv = dv->next) |
289 | if (data_offset == VARIABLE_OFFSET) | |
290 | dv->data_offset = INVALID_SECTORS; | |
291 | else | |
292 | dv->data_offset = data_offset; | |
293 | ||
8592f29d | 294 | for (dv=devlist; dv && !have_container; dv=dv->next, dnum++) { |
cd29a5c8 | 295 | char *dname = dv->devname; |
17f25ca6 | 296 | unsigned long long freesize; |
e9b11fee | 297 | int dfd; |
72ca9bcf | 298 | char *doff; |
e9b11fee | 299 | |
cd29a5c8 | 300 | if (strcasecmp(dname, "missing")==0) { |
c913b90e NB |
301 | if (first_missing > dnum) |
302 | first_missing = dnum; | |
6fb79233 NB |
303 | if (second_missing > dnum && dnum > first_missing) |
304 | second_missing = dnum; | |
52826846 NB |
305 | missing_disks ++; |
306 | continue; | |
307 | } | |
8baab049 | 308 | if (data_offset == VARIABLE_OFFSET) { |
72ca9bcf N |
309 | doff = strchr(dname, ':'); |
310 | if (doff) { | |
311 | *doff++ = 0; | |
312 | dv->data_offset = parse_size(doff); | |
313 | } else | |
314 | dv->data_offset = INVALID_SECTORS; | |
315 | } else | |
316 | dv->data_offset = data_offset; | |
317 | ||
58b3c697 | 318 | dfd = open(dname, O_RDONLY); |
e9b11fee | 319 | if (dfd < 0) { |
e7b84f9d | 320 | pr_err("cannot open %s: %s\n", |
e9b11fee N |
321 | dname, strerror(errno)); |
322 | exit(2); | |
323 | } | |
324 | if (fstat(dfd, &stb) != 0 || | |
325 | (stb.st_mode & S_IFMT) != S_IFBLK) { | |
326 | close(dfd); | |
e7b84f9d | 327 | pr_err("%s is not a block device\n", |
e9b11fee N |
328 | dname); |
329 | exit(2); | |
330 | } | |
331 | close(dfd); | |
06c7f68e | 332 | info.array.working_disks++; |
99cc42f4 | 333 | if (dnum < s->raiddisks) |
06c7f68e | 334 | info.array.active_disks++; |
058574b1 | 335 | if (st == NULL) { |
8aec876d | 336 | struct createinfo *ci = conf_get_create_info(); |
058574b1 NB |
337 | if (ci) |
338 | st = ci->supertype; | |
339 | } | |
576d6d83 NB |
340 | if (st == NULL) { |
341 | /* Need to choose a default metadata, which is different | |
17f25ca6 | 342 | * depending on geometry of array. |
576d6d83 NB |
343 | */ |
344 | int i; | |
345 | char *name = "default"; | |
17f25ca6 | 346 | for(i=0; !st && superlist[i]; i++) { |
576d6d83 | 347 | st = superlist[i]->match_metadata_desc(name); |
ecbd9e81 N |
348 | if (!st) |
349 | continue; | |
a18a888e | 350 | if (do_default_layout) |
99cc42f4 | 351 | s->layout = default_layout(st, s->level, c->verbose); |
ecbd9e81 | 352 | switch (st->ss->validate_geometry( |
99cc42f4 | 353 | st, s->level, s->layout, s->raiddisks, |
af4348dd | 354 | &s->chunk, s->size*2, |
72ca9bcf | 355 | dv->data_offset, dname, |
af4348dd | 356 | &freesize, c->verbose > 0)) { |
ecbd9e81 N |
357 | case -1: /* Not valid, message printed, and not |
358 | * worth checking any further */ | |
359 | exit(2); | |
360 | break; | |
361 | case 0: /* Geometry not valid */ | |
55425f27 | 362 | free(st); |
17f25ca6 | 363 | st = NULL; |
99cc42f4 | 364 | s->chunk = do_default_chunk ? UnSet : s->chunk; |
ecbd9e81 N |
365 | break; |
366 | case 1: /* All happy */ | |
367 | break; | |
55425f27 | 368 | } |
17f25ca6 | 369 | } |
576d6d83 NB |
370 | |
371 | if (!st) { | |
58b3c697 N |
372 | int dfd = open(dname, O_RDONLY|O_EXCL); |
373 | if (dfd < 0) { | |
e7b84f9d | 374 | pr_err("cannot open %s: %s\n", |
58b3c697 N |
375 | dname, strerror(errno)); |
376 | exit(2); | |
377 | } | |
7a862a02 | 378 | pr_err("device %s not suitable for any style of array\n", |
17f25ca6 | 379 | dname); |
576d6d83 NB |
380 | exit(2); |
381 | } | |
b8ac1967 | 382 | if (st->ss != &super0 || |
576d6d83 | 383 | st->minor_version != 90) |
b8ac1967 | 384 | did_default = 1; |
17f25ca6 | 385 | } else { |
a18a888e | 386 | if (do_default_layout) |
99cc42f4 N |
387 | s->layout = default_layout(st, s->level, 0); |
388 | if (!st->ss->validate_geometry(st, s->level, s->layout, | |
389 | s->raiddisks, | |
af4348dd | 390 | &s->chunk, s->size*2, |
72ca9bcf | 391 | dv->data_offset, |
af4348dd | 392 | dname, &freesize, |
171dccc8 | 393 | c->verbose >= 0)) { |
17f25ca6 | 394 | |
7a862a02 | 395 | pr_err("%s is not suitable for this array.\n", |
e7b84f9d | 396 | dname); |
17f25ca6 NB |
397 | fail = 1; |
398 | continue; | |
399 | } | |
682c7051 | 400 | } |
82d9eba6 NB |
401 | |
402 | freesize /= 2; /* convert to K */ | |
99cc42f4 | 403 | if (s->chunk && s->chunk != UnSet) { |
d2cd3ffc | 404 | /* round to chunk size */ |
99cc42f4 | 405 | freesize = freesize & ~(s->chunk-1); |
64385908 CA |
406 | if (do_default_chunk) { |
407 | /* default chunk was just set */ | |
171dccc8 | 408 | if (c->verbose > 0) |
7a862a02 | 409 | pr_err("chunk size defaults to %dK\n", s->chunk); |
99cc42f4 | 410 | s->size &= ~(unsigned long long)(s->chunk - 1); |
64385908 CA |
411 | do_default_chunk = 0; |
412 | } | |
d2cd3ffc | 413 | } |
066e92f0 LD |
414 | if (!freesize) { |
415 | pr_err("no free space left on %s\n", dname); | |
416 | fail = 1; | |
417 | continue; | |
418 | } | |
682c7051 | 419 | |
99cc42f4 | 420 | if (s->size && freesize < s->size) { |
7a862a02 | 421 | pr_err("%s is smaller than given size. %lluK < %lluK + metadata\n", |
99cc42f4 | 422 | dname, freesize, s->size); |
52826846 | 423 | fail = 1; |
52826846 | 424 | continue; |
682c7051 | 425 | } |
cd29a5c8 NB |
426 | if (maxdisc == NULL || (maxdisc && freesize > maxsize)) { |
427 | maxdisc = dname; | |
52826846 | 428 | maxsize = freesize; |
682c7051 | 429 | } |
cd29a5c8 NB |
430 | if (mindisc ==NULL || (mindisc && freesize < minsize)) { |
431 | mindisc = dname; | |
52826846 | 432 | minsize = freesize; |
682c7051 | 433 | } |
171dccc8 | 434 | if (c->runstop != 1 || c->verbose >= 0) { |
37ea3936 | 435 | int fd = open(dname, O_RDONLY); |
b6e63da4 | 436 | if (fd <0 ) { |
e7b84f9d | 437 | pr_err("Cannot open %s: %s\n", |
b6e63da4 NB |
438 | dname, strerror(errno)); |
439 | fail=1; | |
440 | continue; | |
441 | } | |
dab6685f NB |
442 | warn |= check_ext2(fd, dname); |
443 | warn |= check_reiser(fd, dname); | |
444 | warn |= check_raid(fd, dname); | |
034b203a TM |
445 | if (strcmp(st->ss->name, "1.x") == 0 && |
446 | st->minor_version >= 1) | |
447 | /* metadata at front */ | |
53ed6ac3 | 448 | warn |= check_partitions(fd, dname, 0, 0); |
99cc42f4 N |
449 | else if (s->level == 1 || s->level == LEVEL_CONTAINER |
450 | || (s->level == 0 && s->raiddisks == 1)) | |
034b203a | 451 | /* partitions could be meaningful */ |
99cc42f4 | 452 | warn |= check_partitions(fd, dname, freesize*2, s->size*2); |
034b203a TM |
453 | else |
454 | /* partitions cannot be meaningful */ | |
53ed6ac3 | 455 | warn |= check_partitions(fd, dname, 0, 0); |
034b203a | 456 | if (strcmp(st->ss->name, "1.x") == 0 && |
a0962fe9 N |
457 | st->minor_version >= 1 && |
458 | did_default && | |
99cc42f4 | 459 | s->level == 1 && |
034b203a TM |
460 | (warn & 1024) == 0) { |
461 | warn |= 1024; | |
e7b84f9d | 462 | pr_err("Note: this array has metadata at the start and\n" |
a0962fe9 | 463 | " may not be suitable as a boot device. If you plan to\n" |
cc86f89c | 464 | " store '/boot' on this device please ensure that\n" |
a0962fe9 | 465 | " your boot-loader understands md/v1.x metadata, or use\n" |
cc86f89c | 466 | " --metadata=0.90\n"); |
a0962fe9 | 467 | } |
b6e63da4 | 468 | close(fd); |
dab6685f | 469 | } |
682c7051 | 470 | } |
99cc42f4 | 471 | if (s->raiddisks + s->sparedisks > st->max_devs) { |
7a862a02 | 472 | pr_err("Too many devices: %s metadata only supports %d\n", |
acab7bb1 N |
473 | st->ss->name, st->max_devs); |
474 | return 1; | |
475 | } | |
8592f29d | 476 | if (have_container) |
99cc42f4 | 477 | info.array.working_disks = s->raiddisks; |
682c7051 | 478 | if (fail) { |
e7b84f9d | 479 | pr_err("create aborted\n"); |
52826846 | 480 | return 1; |
682c7051 | 481 | } |
99cc42f4 | 482 | if (s->size == 0) { |
5f8097be | 483 | if (mindisc == NULL && !have_container) { |
e7b84f9d | 484 | pr_err("no size and no drives given - aborting create.\n"); |
52826846 NB |
485 | return 1; |
486 | } | |
99cc42f4 N |
487 | if (s->level > 0 || s->level == LEVEL_MULTIPATH |
488 | || s->level == LEVEL_FAULTY | |
b8ac1967 | 489 | || st->ss->external ) { |
b5e64645 | 490 | /* size is meaningful */ |
99cc42f4 N |
491 | if (!st->ss->validate_geometry(st, s->level, s->layout, |
492 | s->raiddisks, | |
493 | &s->chunk, minsize*2, | |
40c9a66a | 494 | data_offset, |
2c514b71 | 495 | NULL, NULL, 0)) { |
99cc42f4 | 496 | pr_err("devices too large for RAID level %d\n", s->level); |
b5e64645 NB |
497 | return 1; |
498 | } | |
99cc42f4 N |
499 | s->size = minsize; |
500 | if (s->level == 1) | |
ae6c05ad N |
501 | /* If this is ever reshaped to RAID5, we will |
502 | * need a chunksize. So round it off a bit | |
503 | * now just to be safe | |
504 | */ | |
99cc42f4 | 505 | s->size &= ~(64ULL-1); |
171dccc8 | 506 | if (c->verbose > 0) |
99cc42f4 | 507 | pr_err("size set to %lluK\n", s->size); |
b5e64645 | 508 | } |
682c7051 | 509 | } |
748952f7 N |
510 | |
511 | if (!s->bitmap_file && | |
512 | s->level >= 1 && | |
39917e56 | 513 | st->ss->add_internal_bitmap && |
748952f7 N |
514 | (s->write_behind || s->size > 100*1024*1024ULL)) { |
515 | if (c->verbose > 0) | |
516 | pr_err("automatically enabling write-intent bitmap on large array\n"); | |
517 | s->bitmap_file = "internal"; | |
518 | } | |
519 | if (s->bitmap_file && strcmp(s->bitmap_file, "none") == 0) | |
520 | s->bitmap_file = NULL; | |
521 | ||
99cc42f4 | 522 | if (!have_container && s->level > 0 && ((maxsize-s->size)*100 > maxsize)) { |
171dccc8 | 523 | if (c->runstop != 1 || c->verbose >= 0) |
e7b84f9d | 524 | pr_err("largest drive (%s) exceeds size (%lluK) by more than 1%%\n", |
99cc42f4 | 525 | maxdisc, s->size); |
52826846 | 526 | warn = 1; |
682c7051 NB |
527 | } |
528 | ||
9eafa1de | 529 | if (st->ss->detail_platform && st->ss->detail_platform(0, 1, NULL) != 0) { |
171dccc8 | 530 | if (c->runstop != 1 || c->verbose >= 0) |
e7b84f9d | 531 | pr_err("%s unable to enumerate platform support\n" |
5615172f DW |
532 | " array may not be compatible with hardware/firmware\n", |
533 | st->ss->name); | |
534 | warn = 1; | |
535 | } | |
529e2aa5 | 536 | st->nodes = c->nodes; |
7716570e | 537 | st->cluster_name = c->homecluster; |
5615172f | 538 | |
682c7051 | 539 | if (warn) { |
171dccc8 | 540 | if (c->runstop!= 1) { |
52826846 | 541 | if (!ask("Continue creating array? ")) { |
e7b84f9d | 542 | pr_err("create aborted.\n"); |
52826846 NB |
543 | return 1; |
544 | } | |
545 | } else { | |
171dccc8 | 546 | if (c->verbose > 0) |
e7b84f9d | 547 | pr_err("creation continuing despite oddities due to --run\n"); |
682c7051 | 548 | } |
682c7051 NB |
549 | } |
550 | ||
66f8bbbe | 551 | /* If this is raid4/5, we want to configure the last active slot |
52826846 | 552 | * as missing, so that a reconstruct happens (faster than re-parity) |
98c6faba | 553 | * FIX: Can we do this for raid6 as well? |
52826846 | 554 | */ |
87b47257 | 555 | if (st->ss->external == 0 && |
99cc42f4 N |
556 | s->assume_clean==0 && c->force == 0 && first_missing >= s->raiddisks) { |
557 | switch ( s->level ) { | |
66f8bbbe | 558 | case 4: |
98c6faba | 559 | case 5: |
99cc42f4 N |
560 | insert_point = s->raiddisks-1; |
561 | s->sparedisks++; | |
06c7f68e | 562 | info.array.active_disks--; |
98c6faba NB |
563 | missing_disks++; |
564 | break; | |
565 | default: | |
566 | break; | |
567 | } | |
52826846 | 568 | } |
6fb79233 NB |
569 | /* For raid6, if creating with 1 missing drive, make a good drive |
570 | * into a spare, else the create will fail | |
571 | */ | |
99cc42f4 | 572 | if (s->assume_clean == 0 && c->force == 0 && first_missing < s->raiddisks && |
ffcfc735 | 573 | st->ss->external == 0 && |
99cc42f4 N |
574 | second_missing >= s->raiddisks && s->level == 6) { |
575 | insert_point = s->raiddisks - 1; | |
6fb79233 NB |
576 | if (insert_point == first_missing) |
577 | insert_point--; | |
99cc42f4 | 578 | s->sparedisks ++; |
6fb79233 NB |
579 | info.array.active_disks--; |
580 | missing_disks++; | |
581 | } | |
570510ba | 582 | |
99cc42f4 | 583 | if (s->level <= 0 && first_missing < subdevs * 2) { |
e7b84f9d | 584 | pr_err("This level does not support missing devices\n"); |
570510ba NB |
585 | return 1; |
586 | } | |
aba69144 | 587 | |
7f91af49 | 588 | /* We need to create the device */ |
ad5bc697 | 589 | map_lock(&map); |
171dccc8 | 590 | mdfd = create_mddev(mddev, name, c->autof, LOCAL, chosen_name); |
e06af9dd JS |
591 | if (mdfd < 0) { |
592 | map_unlock(&map); | |
7f91af49 | 593 | return 1; |
e06af9dd | 594 | } |
3e9df86a AK |
595 | /* verify if chosen_name is not in use, |
596 | * it could be in conflict with already existing device | |
597 | * e.g. container, array | |
598 | */ | |
599 | if (strncmp(chosen_name, "/dev/md/", 8) == 0 | |
600 | && map_by_name(&map, chosen_name+8) != NULL) { | |
e7b84f9d | 601 | pr_err("Array name %s is in use already.\n", |
3e9df86a AK |
602 | chosen_name); |
603 | close(mdfd); | |
604 | map_unlock(&map); | |
605 | return 1; | |
606 | } | |
f2be09f1 | 607 | mddev = chosen_name; |
7f91af49 N |
608 | |
609 | vers = md_get_version(mdfd); | |
610 | if (vers < 9000) { | |
e7b84f9d | 611 | pr_err("Create requires md driver version 0.90.0 or later\n"); |
e06af9dd | 612 | goto abort_locked; |
7f91af49 N |
613 | } else { |
614 | mdu_array_info_t inf; | |
615 | memset(&inf, 0, sizeof(inf)); | |
616 | ioctl(mdfd, GET_ARRAY_INFO, &inf); | |
617 | if (inf.working_disks != 0) { | |
7a862a02 | 618 | pr_err("another array by this name is already running.\n"); |
e06af9dd | 619 | goto abort_locked; |
7f91af49 N |
620 | } |
621 | } | |
a04d5763 | 622 | |
682c7051 NB |
623 | /* Ok, lets try some ioctls */ |
624 | ||
99cc42f4 N |
625 | info.array.level = s->level; |
626 | info.array.size = s->size; | |
627 | info.array.raid_disks = s->raiddisks; | |
82b27616 NB |
628 | /* The kernel should *know* what md_minor we are dealing |
629 | * with, but it chooses to trust me instead. Sigh | |
630 | */ | |
06c7f68e | 631 | info.array.md_minor = 0; |
82b27616 | 632 | if (fstat(mdfd, &stb)==0) |
06c7f68e NB |
633 | info.array.md_minor = minor(stb.st_rdev); |
634 | info.array.not_persistent = 0; | |
6fb79233 | 635 | |
99cc42f4 N |
636 | if ( ( (s->level == 4 || s->level == 5) && |
637 | (insert_point < s->raiddisks || first_missing < s->raiddisks) ) | |
b5e64645 | 638 | || |
99cc42f4 N |
639 | ( s->level == 6 && (insert_point < s->raiddisks |
640 | || second_missing < s->raiddisks)) | |
47d79ef8 | 641 | || |
99cc42f4 | 642 | ( s->level <= 0 ) |
d024b0a7 | 643 | || |
99cc42f4 | 644 | s->assume_clean |
103f2410 | 645 | ) { |
06c7f68e | 646 | info.array.state = 1; /* clean, but one+ drive will be missing*/ |
b7528a20 | 647 | info.resync_start = MaxSector; |
103f2410 | 648 | } else { |
06c7f68e | 649 | info.array.state = 0; /* not clean, but no errors */ |
103f2410 NB |
650 | info.resync_start = 0; |
651 | } | |
99cc42f4 | 652 | if (s->level == 10) { |
f9c25f1d NB |
653 | /* for raid10, the bitmap size is the capacity of the array, |
654 | * which is array.size * raid_disks / ncopies; | |
655 | * .. but convert to sectors. | |
656 | */ | |
99cc42f4 N |
657 | int ncopies = ((s->layout>>8) & 255) * (s->layout & 255); |
658 | bitmapsize = s->size * s->raiddisks / ncopies * 2; | |
659 | /* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, s->size, s->raiddisks, ncopies);*/ | |
f9c25f1d | 660 | } else |
99cc42f4 | 661 | bitmapsize = s->size * 2; |
f9c25f1d | 662 | |
82b27616 NB |
663 | /* There is lots of redundancy in these disk counts, |
664 | * raid_disks is the most meaningful value | |
665 | * it describes the geometry of the array | |
666 | * it is constant | |
667 | * nr_disks is total number of used slots. | |
668 | * it should be raid_disks+spare_disks | |
669 | * spare_disks is the number of extra disks present | |
670 | * see above | |
671 | * active_disks is the number of working disks in | |
672 | * active slots. (With raid_disks) | |
673 | * working_disks is the total number of working disks, | |
674 | * including spares | |
675 | * failed_disks is the number of disks marked failed | |
676 | * | |
5d500228 | 677 | * Ideally, the kernel would keep these (except raid_disks) |
82b27616 NB |
678 | * up-to-date as we ADD_NEW_DISK, but it doesn't (yet). |
679 | * So for now, we assume that all raid and spare | |
680 | * devices will be given. | |
681 | */ | |
99cc42f4 | 682 | info.array.spare_disks=s->sparedisks; |
06c7f68e NB |
683 | info.array.failed_disks=missing_disks; |
684 | info.array.nr_disks = info.array.working_disks | |
685 | + info.array.failed_disks; | |
99cc42f4 N |
686 | info.array.layout = s->layout; |
687 | info.array.chunk_size = s->chunk*1024; | |
82d9eba6 | 688 | |
b3b33eb5 NB |
689 | if (name == NULL || *name == 0) { |
690 | /* base name on mddev */ | |
d1e80164 NB |
691 | /* /dev/md0 -> 0 |
692 | * /dev/md_d0 -> d0 | |
eca944fa | 693 | * /dev/md_foo -> foo |
d1e80164 NB |
694 | * /dev/md/1 -> 1 |
695 | * /dev/md/d1 -> d1 | |
696 | * /dev/md/home -> home | |
697 | * /dev/mdhome -> home | |
698 | */ | |
69207ff6 | 699 | /* FIXME compare this with rules in create_mddev */ |
b3b33eb5 NB |
700 | name = strrchr(mddev, '/'); |
701 | if (name) { | |
702 | name++; | |
eca944fa N |
703 | if (strncmp(name, "md_", 3)==0 && |
704 | strlen(name) > 3 && | |
b3b33eb5 | 705 | (name-mddev) == 5 /* /dev/ */) |
d1e80164 | 706 | name += 3; |
60248f74 NB |
707 | else if (strncmp(name, "md", 2)==0 && |
708 | strlen(name) > 2 && | |
709 | isdigit(name[2]) && | |
710 | (name-mddev) == 5 /* /dev/ */) | |
b3b33eb5 NB |
711 | name += 2; |
712 | } | |
713 | } | |
83cd1e97 | 714 | if (!st->ss->init_super(st, &info.array, s->size, name, c->homehost, uuid, |
40c9a66a | 715 | data_offset)) |
e06af9dd | 716 | goto abort_locked; |
682c7051 | 717 | |
d2ca6449 | 718 | total_slots = info.array.nr_disks; |
a5d85af7 | 719 | st->ss->getinfo_super(st, &info, NULL); |
4dd2df09 | 720 | sysfs_init(&info, mdfd, NULL); |
159c3a1a | 721 | |
171dccc8 | 722 | if (did_default && c->verbose >= 0) { |
3c558363 | 723 | if (is_subarray(info.text_version)) { |
4dd2df09 N |
724 | char devnm[32]; |
725 | char *ep; | |
f7f1b6a1 | 726 | struct mdinfo *mdi; |
f7f1b6a1 | 727 | |
4dd2df09 N |
728 | strncpy(devnm, info.text_version+1, 32); |
729 | devnm[31] = 0; | |
730 | ep = strchr(devnm, '/'); | |
731 | if (ep) | |
732 | *ep = 0; | |
733 | ||
734 | mdi = sysfs_read(-1, devnm, GET_VERSION); | |
f7f1b6a1 | 735 | |
4dd2df09 N |
736 | pr_err("Creating array inside %s container %s\n", |
737 | mdi?mdi->text_version:"managed", devnm); | |
f7f1b6a1 NB |
738 | sysfs_free(mdi); |
739 | } else | |
7a862a02 | 740 | pr_err("Defaulting to version %s metadata\n", info.text_version); |
f7f1b6a1 | 741 | } |
b8ac1967 | 742 | |
4dd2df09 | 743 | map_update(&map, fd2devnm(mdfd), info.text_version, |
a04d5763 | 744 | info.uuid, chosen_name); |
19ad4b2c AP |
745 | /* Keep map locked until devices have been added to array |
746 | * to stop another mdadm from finding and using those devices. | |
747 | */ | |
a04d5763 | 748 | |
99cc42f4 | 749 | if (s->bitmap_file && vers < 9003) { |
943eafef | 750 | major_num = BITMAP_MAJOR_HOSTENDIAN; |
dcec9ee5 | 751 | #ifdef __BIG_ENDIAN |
e7b84f9d | 752 | pr_err("Warning - bitmaps created on this kernel are not portable\n" |
dcec9ee5 NB |
753 | " between different architectured. Consider upgrading the Linux kernel.\n"); |
754 | #endif | |
755 | } | |
756 | ||
95a05b37 GJ |
757 | if (s->bitmap_file && (strcmp(s->bitmap_file, "internal")==0 || |
758 | strcmp(s->bitmap_file, "clustered")==0)) { | |
55935d51 | 759 | if ((vers%100) < 2) { |
e7b84f9d | 760 | pr_err("internal bitmaps not supported by this kernel.\n"); |
19ad4b2c | 761 | goto abort_locked; |
55935d51 | 762 | } |
ebeb3663 | 763 | if (!st->ss->add_internal_bitmap) { |
e7b84f9d | 764 | pr_err("internal bitmaps not supported with %s metadata\n", |
ebeb3663 | 765 | st->ss->name); |
19ad4b2c | 766 | goto abort_locked; |
ebeb3663 | 767 | } |
99cc42f4 N |
768 | if (!st->ss->add_internal_bitmap(st, &s->bitmap_chunk, |
769 | c->delay, s->write_behind, | |
943eafef | 770 | bitmapsize, 1, major_num)) { |
e7b84f9d | 771 | pr_err("Given bitmap chunk size not supported.\n"); |
19ad4b2c | 772 | goto abort_locked; |
55935d51 | 773 | } |
99cc42f4 | 774 | s->bitmap_file = NULL; |
55935d51 NB |
775 | } |
776 | ||
4dd2df09 | 777 | sysfs_init(&info, mdfd, NULL); |
55935d51 | 778 | |
4dd2df09 | 779 | if (st->ss->external && st->container_devnm[0]) { |
f35f2525 N |
780 | /* member */ |
781 | ||
782 | /* When creating a member, we need to be careful | |
783 | * to negotiate with mdmon properly. | |
784 | * If it is already running, we cannot write to | |
785 | * the devices and must ask it to do that part. | |
786 | * If it isn't running, we write to the devices, | |
787 | * and then start it. | |
788 | * We hold an exclusive open on the container | |
789 | * device to make sure mdmon doesn't exit after | |
790 | * we checked that it is running. | |
791 | * | |
792 | * For now, fail if it is already running. | |
793 | */ | |
4dd2df09 | 794 | container_fd = open_dev_excl(st->container_devnm); |
f35f2525 | 795 | if (container_fd < 0) { |
7a862a02 | 796 | pr_err("Cannot get exclusive open on container - weird.\n"); |
19ad4b2c | 797 | goto abort_locked; |
d03373f1 | 798 | } |
4dd2df09 | 799 | if (mdmon_running(st->container_devnm)) { |
171dccc8 | 800 | if (c->verbose) |
7a862a02 | 801 | pr_err("reusing mdmon for %s.\n", |
4dd2df09 | 802 | st->container_devnm); |
f35f2525 N |
803 | st->update_tail = &st->updates; |
804 | } else | |
805 | need_mdmon = 1; | |
806 | } | |
807 | rv = set_array_info(mdfd, st, &info); | |
82d9eba6 | 808 | if (rv) { |
e7b84f9d | 809 | pr_err("failed to set array info for %s: %s\n", |
52826846 | 810 | mddev, strerror(errno)); |
19ad4b2c | 811 | goto abort_locked; |
682c7051 | 812 | } |
52826846 | 813 | |
99cc42f4 | 814 | if (s->bitmap_file) { |
c82f047c | 815 | int uuid[4]; |
55935d51 | 816 | |
3da92f27 | 817 | st->ss->uuid_from_super(st, uuid); |
99cc42f4 N |
818 | if (CreateBitmap(s->bitmap_file, c->force, (char*)uuid, s->bitmap_chunk, |
819 | c->delay, s->write_behind, | |
f9c25f1d | 820 | bitmapsize, |
943eafef | 821 | major_num)) { |
19ad4b2c | 822 | goto abort_locked; |
c82f047c | 823 | } |
99cc42f4 | 824 | bitmap_fd = open(s->bitmap_file, O_RDWR); |
c82f047c | 825 | if (bitmap_fd < 0) { |
e7b84f9d | 826 | pr_err("weird: %s cannot be openned\n", |
99cc42f4 | 827 | s->bitmap_file); |
19ad4b2c | 828 | goto abort_locked; |
c82f047c NB |
829 | } |
830 | if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) { | |
e7b84f9d | 831 | pr_err("Cannot set bitmap file for %s: %s\n", |
c82f047c | 832 | mddev, strerror(errno)); |
19ad4b2c | 833 | goto abort_locked; |
c82f047c NB |
834 | } |
835 | } | |
836 | ||
503975b9 | 837 | infos = xmalloc(sizeof(*infos) * total_slots); |
a7dec3fd | 838 | enable_fds(total_slots); |
4b1ac34b | 839 | for (pass=1; pass <=2 ; pass++) { |
a655e550 | 840 | struct mddev_dev *moved_disk = NULL; /* the disk that was moved out of the insert point */ |
4b1ac34b NB |
841 | |
842 | for (dnum=0, dv = devlist ; dv ; | |
843 | dv=(dv->next)?(dv->next):moved_disk, dnum++) { | |
844 | int fd; | |
845 | struct stat stb; | |
d2ca6449 | 846 | struct mdinfo *inf = &infos[dnum]; |
4b1ac34b | 847 | |
d2ca6449 NB |
848 | if (dnum >= total_slots) |
849 | abort(); | |
4b1ac34b NB |
850 | if (dnum == insert_point) { |
851 | moved_disk = dv; | |
8592f29d | 852 | continue; |
52826846 | 853 | } |
8592f29d | 854 | if (strcasecmp(dv->devname, "missing")==0) |
111d01fc | 855 | continue; |
8592f29d N |
856 | if (have_container) |
857 | moved_disk = NULL; | |
858 | if (have_container && dnum < info.array.raid_disks - 1) | |
859 | /* repeatedly use the container */ | |
860 | moved_disk = dv; | |
111d01fc | 861 | |
d2ca6449 NB |
862 | switch(pass) { |
863 | case 1: | |
864 | *inf = info; | |
5f8097be | 865 | |
d2ca6449 NB |
866 | inf->disk.number = dnum; |
867 | inf->disk.raid_disk = dnum; | |
99cc42f4 | 868 | if (inf->disk.raid_disk < s->raiddisks) |
d2ca6449 NB |
869 | inf->disk.state = (1<<MD_DISK_ACTIVE) | |
870 | (1<<MD_DISK_SYNC); | |
871 | else | |
872 | inf->disk.state = 0; | |
873 | ||
b01b06bd | 874 | if (dv->writemostly == 1) |
d2ca6449 NB |
875 | inf->disk.state |= (1<<MD_DISK_WRITEMOSTLY); |
876 | ||
8592f29d N |
877 | if (have_container) |
878 | fd = -1; | |
879 | else { | |
d1d599ea | 880 | if (st->ss->external && |
4dd2df09 | 881 | st->container_devnm[0]) |
8592f29d N |
882 | fd = open(dv->devname, O_RDWR); |
883 | else | |
884 | fd = open(dv->devname, O_RDWR|O_EXCL); | |
885 | ||
886 | if (fd < 0) { | |
7a862a02 | 887 | pr_err("failed to open %s after earlier success - aborting\n", |
8592f29d | 888 | dv->devname); |
19ad4b2c | 889 | goto abort_locked; |
8592f29d N |
890 | } |
891 | fstat(fd, &stb); | |
892 | inf->disk.major = major(stb.st_rdev); | |
893 | inf->disk.minor = minor(stb.st_rdev); | |
d2ca6449 | 894 | } |
8592f29d N |
895 | if (fd >= 0) |
896 | remove_partitions(fd); | |
f20c3968 | 897 | if (st->ss->add_to_super(st, &inf->disk, |
72ca9bcf N |
898 | fd, dv->devname, |
899 | dv->data_offset)) { | |
4eb26970 | 900 | ioctl(mdfd, STOP_ARRAY, NULL); |
19ad4b2c | 901 | goto abort_locked; |
4eb26970 | 902 | } |
a5d85af7 | 903 | st->ss->getinfo_super(st, inf, NULL); |
8ed3e5e1 | 904 | safe_mode_delay = inf->safe_mode_delay; |
d2ca6449 | 905 | |
171dccc8 | 906 | if (have_container && c->verbose > 0) |
e7b84f9d | 907 | pr_err("Using %s for device %d\n", |
8592f29d N |
908 | map_dev(inf->disk.major, |
909 | inf->disk.minor, | |
910 | 0), dnum); | |
911 | ||
912 | if (!have_container) { | |
913 | /* getinfo_super might have lost these ... */ | |
914 | inf->disk.major = major(stb.st_rdev); | |
915 | inf->disk.minor = minor(stb.st_rdev); | |
916 | } | |
4b1ac34b NB |
917 | break; |
918 | case 2: | |
d2ca6449 | 919 | inf->errors = 0; |
4b1ac34b | 920 | |
f35f2525 | 921 | rv = add_disk(mdfd, st, &info, inf); |
d2ca6449 | 922 | |
2503d23b | 923 | if (rv) { |
7a862a02 | 924 | pr_err("ADD_NEW_DISK for %s failed: %s\n", |
e7b84f9d | 925 | dv->devname, strerror(errno)); |
19ad4b2c | 926 | goto abort_locked; |
4b1ac34b | 927 | } |
4b1ac34b NB |
928 | break; |
929 | } | |
8592f29d N |
930 | if (!have_container && |
931 | dv == moved_disk && dnum != insert_point) break; | |
52826846 | 932 | } |
edd8d13c | 933 | if (pass == 1) { |
9b1fb677 DW |
934 | struct mdinfo info_new; |
935 | struct map_ent *me = NULL; | |
936 | ||
937 | /* check to see if the uuid has changed due to these | |
938 | * metadata changes, and if so update the member array | |
939 | * and container uuid. Note ->write_init_super clears | |
940 | * the subarray cursor such that ->getinfo_super once | |
941 | * again returns container info. | |
942 | */ | |
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 | } | |
9b1fb677 | 967 | |
edd8d13c | 968 | flush_metadata_updates(st); |
6946681d | 969 | st->ss->free_super(st); |
edd8d13c | 970 | } |
5e52ae9e | 971 | } |
19ad4b2c | 972 | map_unlock(&map); |
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) { |
7a862a02 | 1002 | pr_err("failed to activate array.\n"); |
4c821454 N |
1003 | ioctl(mdfd, STOP_ARRAY, NULL); |
1004 | goto abort; | |
1005 | } | |
171dccc8 | 1006 | } else if (c->readonly && |
72d566f6 N |
1007 | sysfs_attribute_available( |
1008 | &info, NULL, "array_state")) { | |
1009 | if (sysfs_set_str(&info, NULL, | |
1010 | "array_state", "readonly") < 0) { | |
1011 | pr_err("Failed to start array: %s\n", | |
1012 | strerror(errno)); | |
1013 | ioctl(mdfd, STOP_ARRAY, NULL); | |
1014 | goto abort; | |
1015 | } | |
598f0d58 | 1016 | } else { |
97590376 | 1017 | /* param is not actually used */ |
598f0d58 NB |
1018 | mdu_param_t param; |
1019 | if (ioctl(mdfd, RUN_ARRAY, ¶m)) { | |
e7b84f9d | 1020 | pr_err("RUN_ARRAY failed: %s\n", |
72d566f6 | 1021 | strerror(errno)); |
a252c078 | 1022 | if (info.array.chunk_size & (info.array.chunk_size-1)) { |
7a862a02 | 1023 | cont_err("Problem may be that chunk size is not a power of 2\n"); |
a252c078 | 1024 | } |
4eb26970 | 1025 | ioctl(mdfd, STOP_ARRAY, NULL); |
7f91af49 | 1026 | goto abort; |
598f0d58 | 1027 | } |
a21e848a N |
1028 | /* if start_ro module parameter is set, array is |
1029 | * auto-read-only, which is bad as the resync won't | |
1030 | * start. So lets make it read-write now. | |
1031 | */ | |
1032 | ioctl(mdfd, RESTART_ARRAY_RW, NULL); | |
682c7051 | 1033 | } |
171dccc8 | 1034 | if (c->verbose >= 0) |
e7b84f9d | 1035 | pr_err("array %s started.\n", mddev); |
4dd2df09 | 1036 | if (st->ss->external && st->container_devnm[0]) { |
8850ee3e | 1037 | if (need_mdmon) |
4dd2df09 | 1038 | start_mdmon(st->container_devnm); |
8850ee3e | 1039 | |
4dd2df09 | 1040 | ping_monitor(st->container_devnm); |
a931db9e NB |
1041 | close(container_fd); |
1042 | } | |
a7c6e3fb | 1043 | wait_for(chosen_name, mdfd); |
682c7051 | 1044 | } else { |
e7b84f9d | 1045 | pr_err("not starting array - not enough devices.\n"); |
682c7051 | 1046 | } |
7f91af49 | 1047 | close(mdfd); |
682c7051 | 1048 | return 0; |
7f91af49 N |
1049 | |
1050 | abort: | |
4eb26970 | 1051 | map_lock(&map); |
e06af9dd | 1052 | abort_locked: |
4dd2df09 | 1053 | map_remove(&map, fd2devnm(mdfd)); |
4eb26970 DW |
1054 | map_unlock(&map); |
1055 | ||
7f91af49 N |
1056 | if (mdfd >= 0) |
1057 | close(mdfd); | |
1058 | return 1; | |
64c4757e | 1059 | } |