]>
Commit | Line | Data |
---|---|---|
64c4757e | 1 | /* |
9a9dab36 | 2 | * mdadm - manage Linux "md" devices aka RAID arrays. |
64c4757e | 3 | * |
e736b623 | 4 | * Copyright (C) 2001-2009 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 | ||
9a9dab36 | 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 | ||
34 | if (st && st->ss->default_layout) | |
35 | layout = st->ss->default_layout(level); | |
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) | |
45 | fprintf(stderr, | |
46 | Name ": layout defaults to n1\n"); | |
47 | break; | |
48 | case 5: | |
49 | case 6: | |
50 | layout = map_name(r5layout, "default"); | |
51 | if (verbose > 0) | |
52 | fprintf(stderr, | |
53 | Name ": layout defaults to %s\n", map_num(r5layout, layout)); | |
54 | break; | |
55 | case LEVEL_FAULTY: | |
56 | layout = map_name(faultylayout, "default"); | |
57 | ||
58 | if (verbose > 0) | |
59 | fprintf(stderr, | |
60 | Name ": layout defaults to %s\n", map_num(faultylayout, layout)); | |
61 | break; | |
62 | } | |
63 | ||
64 | return layout; | |
65 | } | |
66 | ||
67 | ||
7f91af49 | 68 | int Create(struct supertype *st, char *mddev, |
5dd497ee | 69 | int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks, |
3d3dd91e | 70 | char *name, char *homehost, int *uuid, |
cd29a5c8 | 71 | int subdevs, mddev_dev_t devlist, |
47d79ef8 | 72 | int runstop, int verbose, int force, int assume_clean, |
7f91af49 | 73 | char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof) |
64c4757e | 74 | { |
682c7051 NB |
75 | /* |
76 | * Create a new raid array. | |
77 | * | |
78 | * First check that necessary details are available | |
79 | * (i.e. level, raid-disks) | |
80 | * | |
81 | * Then check each disk to see what might be on it | |
82 | * and report anything interesting. | |
83 | * | |
84 | * If anything looks odd, and runstop not set, | |
85 | * abort. | |
86 | * | |
87 | * SET_ARRAY_INFO and ADD_NEW_DISK, and | |
82d9eba6 | 88 | * if runstop==run, or raiddisks disks were used, |
682c7051 NB |
89 | * RUN_ARRAY |
90 | */ | |
7f91af49 | 91 | int mdfd; |
b5e64645 | 92 | unsigned long long minsize=0, maxsize=0; |
cd29a5c8 NB |
93 | char *mindisc = NULL; |
94 | char *maxdisc = NULL; | |
c913b90e | 95 | int dnum; |
cd29a5c8 | 96 | mddev_dev_t dv; |
682c7051 | 97 | int fail=0, warn=0; |
82b27616 | 98 | struct stat stb; |
82d9eba6 | 99 | int first_missing = subdevs * 2; |
6fb79233 | 100 | int second_missing = subdevs * 2; |
52826846 | 101 | int missing_disks = 0; |
82d9eba6 | 102 | int insert_point = subdevs * 2; /* where to insert a missing drive */ |
d2ca6449 | 103 | int total_slots; |
4b1ac34b | 104 | int pass; |
82d9eba6 NB |
105 | int vers; |
106 | int rv; | |
c82f047c | 107 | int bitmap_fd; |
5f8097be | 108 | int have_container = 0; |
0e600426 | 109 | int container_fd = -1; |
a931db9e | 110 | int need_mdmon = 0; |
f9c25f1d | 111 | unsigned long long bitmapsize; |
d2ca6449 | 112 | struct mdinfo info, *infos; |
b8ac1967 | 113 | int did_default = 0; |
a18a888e | 114 | int do_default_layout = 0; |
8ed3e5e1 | 115 | unsigned long safe_mode_delay = 0; |
a04d5763 | 116 | char chosen_name[1024]; |
ad5bc697 | 117 | struct map_ent *map = NULL; |
8592f29d | 118 | unsigned long long newsize; |
682c7051 | 119 | |
943eafef | 120 | int major_num = BITMAP_MAJOR_HI; |
dcec9ee5 | 121 | |
06c7f68e | 122 | memset(&info, 0, sizeof(info)); |
29e766a5 | 123 | |
a322f70c | 124 | if (level == UnSet) { |
3cfe6247 | 125 | /* "ddf" and "imsm" metadata only supports one level - should possibly |
a322f70c DW |
126 | * push this into metadata handler?? |
127 | */ | |
3cfe6247 | 128 | if (st && (st->ss == &super_ddf || st->ss == &super_imsm)) |
a322f70c DW |
129 | level = LEVEL_CONTAINER; |
130 | } | |
131 | ||
98c6faba | 132 | if (level == UnSet) { |
682c7051 NB |
133 | fprintf(stderr, |
134 | Name ": a RAID level is needed to create an array.\n"); | |
135 | return 1; | |
136 | } | |
98c6faba NB |
137 | if (raiddisks < 4 && level == 6) { |
138 | fprintf(stderr, | |
139 | Name ": at least 4 raid-devices needed for level 6\n"); | |
140 | return 1; | |
141 | } | |
142 | if (raiddisks > 256 && level == 6) { | |
143 | fprintf(stderr, | |
144 | Name ": no more than 256 raid-devices supported for level 6\n"); | |
145 | return 1; | |
146 | } | |
52826846 NB |
147 | if (raiddisks < 2 && level >= 4) { |
148 | fprintf(stderr, | |
98c6faba | 149 | Name ": at least 2 raid-devices needed for level 4 or 5\n"); |
52826846 NB |
150 | return 1; |
151 | } | |
570510ba NB |
152 | if (level <= 0 && sparedisks) { |
153 | fprintf(stderr, | |
154 | Name ": This level does not support spare devices\n"); | |
155 | return 1; | |
156 | } | |
5f8097be NB |
157 | |
158 | if (subdevs == 1 && strcmp(devlist->devname, "missing") != 0) { | |
159 | /* If given a single device, it might be a container, and we can | |
160 | * extract a device list from there | |
161 | */ | |
162 | mdu_array_info_t inf; | |
163 | int fd; | |
164 | ||
165 | memset(&inf, 0, sizeof(inf)); | |
37ea3936 | 166 | fd = open(devlist->devname, O_RDONLY); |
5f8097be NB |
167 | if (fd >= 0 && |
168 | ioctl(fd, GET_ARRAY_INFO, &inf) == 0 && | |
169 | inf.raid_disks == 0) { | |
170 | /* yep, looks like a container */ | |
171 | if (st) { | |
172 | rv = st->ss->load_super(st, fd, | |
173 | devlist->devname); | |
174 | if (rv == 0) | |
175 | have_container = 1; | |
176 | } else { | |
177 | st = guess_super(fd); | |
178 | if (st && !(rv = st->ss-> | |
179 | load_super(st, fd, | |
180 | devlist->devname))) | |
181 | have_container = 1; | |
182 | else | |
183 | st = NULL; | |
184 | } | |
18fde300 | 185 | if (have_container) { |
8592f29d | 186 | subdevs = raiddisks; |
18fde300 DW |
187 | first_missing = subdevs * 2; |
188 | second_missing = subdevs * 2; | |
189 | insert_point = subdevs * 2; | |
190 | } | |
5f8097be NB |
191 | } |
192 | if (fd >= 0) | |
193 | close(fd); | |
5f8097be | 194 | } |
ffcfc735 N |
195 | if (st && st->ss->external && sparedisks) { |
196 | fprintf(stderr, | |
197 | Name ": This metadata type does not support " | |
b42f577a | 198 | "spare disks at create time\n"); |
ffcfc735 N |
199 | return 1; |
200 | } | |
682c7051 | 201 | if (subdevs > raiddisks+sparedisks) { |
b83d95f3 | 202 | fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks); |
52826846 | 203 | return 1; |
682c7051 | 204 | } |
5f8097be | 205 | if (!have_container && subdevs < raiddisks+sparedisks) { |
52826846 NB |
206 | fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n"); |
207 | return 1; | |
208 | } | |
5b28bd56 NB |
209 | if (bitmap_file && level <= 0) { |
210 | fprintf(stderr, Name ": bitmaps not meaningful with level %s\n", | |
211 | map_num(pers, level)?:"given"); | |
212 | return 1; | |
213 | } | |
52826846 | 214 | |
682c7051 | 215 | /* now set some defaults */ |
b5e64645 | 216 | |
a18a888e DW |
217 | |
218 | if (layout == UnSet) { | |
219 | do_default_layout = 1; | |
220 | layout = default_layout(st, level, verbose); | |
221 | } | |
682c7051 | 222 | |
e5329c37 NB |
223 | if (level == 10) |
224 | /* check layout fits in array*/ | |
225 | if ((layout&255) * ((layout>>8)&255) > raiddisks) { | |
226 | fprintf(stderr, Name ": that layout requires at least %d devices\n", | |
227 | (layout&255) * ((layout>>8)&255)); | |
228 | return 1; | |
229 | } | |
230 | ||
aa88f531 NB |
231 | switch(level) { |
232 | case 4: | |
233 | case 5: | |
e5329c37 | 234 | case 10: |
98c6faba | 235 | case 6: |
aa88f531 | 236 | case 0: |
aa88f531 | 237 | if (chunk == 0) { |
5f175898 N |
238 | chunk = 512; |
239 | if (verbose > 0) | |
240 | fprintf(stderr, Name ": chunk size defaults to 512K\n"); | |
241 | } | |
242 | break; | |
243 | case LEVEL_LINEAR: | |
244 | /* a chunksize of zero 0s perfectly valid (and preferred) since 2.6.16 */ | |
245 | if (get_linux_version() < 2006016 && chunk == 0) { | |
aa88f531 | 246 | chunk = 64; |
dab6685f | 247 | if (verbose > 0) |
aa88f531 NB |
248 | fprintf(stderr, Name ": chunk size defaults to 64K\n"); |
249 | } | |
250 | break; | |
570510ba | 251 | case 1: |
08e43379 | 252 | case LEVEL_FAULTY: |
570510ba | 253 | case LEVEL_MULTIPATH: |
17f25ca6 | 254 | case LEVEL_CONTAINER: |
aa88f531 NB |
255 | if (chunk) { |
256 | chunk = 0; | |
dab6685f | 257 | if (verbose > 0) |
aa88f531 NB |
258 | fprintf(stderr, Name ": chunk size ignored for this level\n"); |
259 | } | |
260 | break; | |
570510ba NB |
261 | default: |
262 | fprintf(stderr, Name ": unknown level %d\n", level); | |
263 | return 1; | |
682c7051 | 264 | } |
8592f29d | 265 | |
9f3bd60c N |
266 | if (size && chunk) |
267 | size &= ~(unsigned long long)(chunk - 1); | |
8592f29d | 268 | newsize = size * 2; |
17f25ca6 | 269 | if (st && ! st->ss->validate_geometry(st, level, layout, raiddisks, |
8592f29d | 270 | chunk, size*2, NULL, &newsize, verbose>=0)) |
17f25ca6 | 271 | return 1; |
8592f29d N |
272 | if (size == 0) { |
273 | size = newsize / 2; | |
274 | if (size && verbose > 0) | |
275 | fprintf(stderr, Name ": setting size to %lluK\n", | |
276 | (unsigned long long)size); | |
277 | } | |
17f25ca6 | 278 | |
682c7051 | 279 | /* now look at the subdevs */ |
06c7f68e NB |
280 | info.array.active_disks = 0; |
281 | info.array.working_disks = 0; | |
c913b90e | 282 | dnum = 0; |
8592f29d | 283 | for (dv=devlist; dv && !have_container; dv=dv->next, dnum++) { |
cd29a5c8 | 284 | char *dname = dv->devname; |
17f25ca6 | 285 | unsigned long long freesize; |
cd29a5c8 | 286 | if (strcasecmp(dname, "missing")==0) { |
c913b90e NB |
287 | if (first_missing > dnum) |
288 | first_missing = dnum; | |
6fb79233 NB |
289 | if (second_missing > dnum && dnum > first_missing) |
290 | second_missing = dnum; | |
52826846 NB |
291 | missing_disks ++; |
292 | continue; | |
293 | } | |
06c7f68e | 294 | info.array.working_disks++; |
c913b90e | 295 | if (dnum < raiddisks) |
06c7f68e | 296 | info.array.active_disks++; |
058574b1 | 297 | if (st == NULL) { |
8aec876d | 298 | struct createinfo *ci = conf_get_create_info(); |
058574b1 NB |
299 | if (ci) |
300 | st = ci->supertype; | |
301 | } | |
576d6d83 NB |
302 | if (st == NULL) { |
303 | /* Need to choose a default metadata, which is different | |
17f25ca6 | 304 | * depending on geometry of array. |
576d6d83 NB |
305 | */ |
306 | int i; | |
307 | char *name = "default"; | |
17f25ca6 | 308 | for(i=0; !st && superlist[i]; i++) { |
576d6d83 | 309 | st = superlist[i]->match_metadata_desc(name); |
a18a888e DW |
310 | if (do_default_layout) |
311 | layout = default_layout(st, level, verbose); | |
17f25ca6 NB |
312 | if (st && !st->ss->validate_geometry |
313 | (st, level, layout, raiddisks, | |
e46273eb | 314 | chunk, size*2, dname, &freesize, |
2c514b71 | 315 | verbose > 0)) |
17f25ca6 NB |
316 | st = NULL; |
317 | } | |
576d6d83 NB |
318 | |
319 | if (!st) { | |
17f25ca6 NB |
320 | fprintf(stderr, Name ": device %s not suitable " |
321 | "for any style of array\n", | |
322 | dname); | |
576d6d83 NB |
323 | exit(2); |
324 | } | |
b8ac1967 | 325 | if (st->ss != &super0 || |
576d6d83 | 326 | st->minor_version != 90) |
b8ac1967 | 327 | did_default = 1; |
17f25ca6 | 328 | } else { |
a18a888e DW |
329 | if (do_default_layout) |
330 | layout = default_layout(st, level, verbose); | |
17f25ca6 NB |
331 | if (!st->ss->validate_geometry(st, level, layout, |
332 | raiddisks, | |
e46273eb | 333 | chunk, size*2, dname, |
2c514b71 | 334 | &freesize, |
b42f577a | 335 | verbose >= 0)) { |
17f25ca6 NB |
336 | |
337 | fprintf(stderr, | |
338 | Name ": %s is not suitable for " | |
339 | "this array.\n", | |
340 | dname); | |
341 | fail = 1; | |
342 | continue; | |
343 | } | |
682c7051 | 344 | } |
82d9eba6 NB |
345 | |
346 | freesize /= 2; /* convert to K */ | |
d2cd3ffc NB |
347 | if (chunk) { |
348 | /* round to chunk size */ | |
349 | freesize = freesize & ~(chunk-1); | |
350 | } | |
682c7051 NB |
351 | |
352 | if (size && freesize < size) { | |
42dc2744 | 353 | fprintf(stderr, Name ": %s is smaller than given size." |
17f25ca6 NB |
354 | " %lluK < %lluK + metadata\n", |
355 | dname, freesize, size); | |
52826846 | 356 | fail = 1; |
52826846 | 357 | continue; |
682c7051 | 358 | } |
cd29a5c8 NB |
359 | if (maxdisc == NULL || (maxdisc && freesize > maxsize)) { |
360 | maxdisc = dname; | |
52826846 | 361 | maxsize = freesize; |
682c7051 | 362 | } |
cd29a5c8 NB |
363 | if (mindisc ==NULL || (mindisc && freesize < minsize)) { |
364 | mindisc = dname; | |
52826846 | 365 | minsize = freesize; |
682c7051 | 366 | } |
dab6685f | 367 | if (runstop != 1 || verbose >= 0) { |
37ea3936 | 368 | int fd = open(dname, O_RDONLY); |
b6e63da4 NB |
369 | if (fd <0 ) { |
370 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
371 | dname, strerror(errno)); | |
372 | fail=1; | |
373 | continue; | |
374 | } | |
dab6685f NB |
375 | warn |= check_ext2(fd, dname); |
376 | warn |= check_reiser(fd, dname); | |
377 | warn |= check_raid(fd, dname); | |
a0962fe9 N |
378 | if (st && strcmp(st->ss->name, "1.x") == 0 && |
379 | st->minor_version >= 1 && | |
380 | did_default && | |
381 | level == 1) { | |
382 | warn = 1; | |
383 | fprintf(stderr, Name ": Note: this array has metadata at the start and\n" | |
384 | " may not be suitable as a boot device. If you plan to\n" | |
385 | " store '/' or '/boot' on this device please ensure that\n" | |
386 | " your boot-loader understands md/v1.x metadata, or use\n" | |
387 | " --metadata=1.0\n"); | |
388 | } | |
b6e63da4 | 389 | close(fd); |
dab6685f | 390 | } |
682c7051 | 391 | } |
8592f29d N |
392 | if (have_container) |
393 | info.array.working_disks = raiddisks; | |
682c7051 | 394 | if (fail) { |
52826846 NB |
395 | fprintf(stderr, Name ": create aborted\n"); |
396 | return 1; | |
682c7051 NB |
397 | } |
398 | if (size == 0) { | |
5f8097be | 399 | if (mindisc == NULL && !have_container) { |
52826846 NB |
400 | fprintf(stderr, Name ": no size and no drives given - aborting create.\n"); |
401 | return 1; | |
402 | } | |
5f8097be NB |
403 | if (level > 0 || level == LEVEL_MULTIPATH |
404 | || level == LEVEL_FAULTY | |
b8ac1967 | 405 | || st->ss->external ) { |
b5e64645 | 406 | /* size is meaningful */ |
b8ac1967 NB |
407 | if (!st->ss->validate_geometry(st, level, layout, |
408 | raiddisks, | |
e46273eb | 409 | chunk, minsize*2, |
2c514b71 | 410 | NULL, NULL, 0)) { |
aba69144 | 411 | fprintf(stderr, Name ": devices too large for RAID level %d\n", level); |
b5e64645 NB |
412 | return 1; |
413 | } | |
414 | size = minsize; | |
dab6685f | 415 | if (verbose > 0) |
5dd497ee | 416 | fprintf(stderr, Name ": size set to %lluK\n", size); |
b5e64645 | 417 | } |
682c7051 | 418 | } |
8592f29d | 419 | if (!have_container && level > 0 && ((maxsize-size)*100 > maxsize)) { |
dab6685f | 420 | if (runstop != 1 || verbose >= 0) |
8592f29d | 421 | fprintf(stderr, Name ": largest drive (%s) exceeds size (%lluK) by more than 1%%\n", |
dab6685f | 422 | maxdisc, size); |
52826846 | 423 | warn = 1; |
682c7051 NB |
424 | } |
425 | ||
5615172f DW |
426 | if (st->ss->detail_platform && st->ss->detail_platform(0, 1) != 0) { |
427 | if (runstop != 1 || verbose >= 0) | |
428 | fprintf(stderr, Name ": %s unable to enumerate platform support\n" | |
429 | " array may not be compatible with hardware/firmware\n", | |
430 | st->ss->name); | |
431 | warn = 1; | |
432 | } | |
433 | ||
682c7051 | 434 | if (warn) { |
52826846 NB |
435 | if (runstop!= 1) { |
436 | if (!ask("Continue creating array? ")) { | |
437 | fprintf(stderr, Name ": create aborted.\n"); | |
438 | return 1; | |
439 | } | |
440 | } else { | |
dab6685f | 441 | if (verbose > 0) |
52826846 | 442 | fprintf(stderr, Name ": creation continuing despite oddities due to --run\n"); |
682c7051 | 443 | } |
682c7051 NB |
444 | } |
445 | ||
66f8bbbe | 446 | /* If this is raid4/5, we want to configure the last active slot |
52826846 | 447 | * as missing, so that a reconstruct happens (faster than re-parity) |
98c6faba | 448 | * FIX: Can we do this for raid6 as well? |
52826846 | 449 | */ |
87b47257 N |
450 | if (st->ss->external == 0 && |
451 | assume_clean==0 && force == 0 && first_missing >= raiddisks) { | |
98c6faba | 452 | switch ( level ) { |
66f8bbbe | 453 | case 4: |
98c6faba NB |
454 | case 5: |
455 | insert_point = raiddisks-1; | |
456 | sparedisks++; | |
06c7f68e | 457 | info.array.active_disks--; |
98c6faba NB |
458 | missing_disks++; |
459 | break; | |
460 | default: | |
461 | break; | |
462 | } | |
52826846 | 463 | } |
6fb79233 NB |
464 | /* For raid6, if creating with 1 missing drive, make a good drive |
465 | * into a spare, else the create will fail | |
466 | */ | |
467 | if (assume_clean == 0 && force == 0 && first_missing < raiddisks && | |
ffcfc735 | 468 | st->ss->external == 0 && |
6fb79233 NB |
469 | second_missing >= raiddisks && level == 6) { |
470 | insert_point = raiddisks - 1; | |
471 | if (insert_point == first_missing) | |
472 | insert_point--; | |
473 | sparedisks ++; | |
474 | info.array.active_disks--; | |
475 | missing_disks++; | |
476 | } | |
570510ba | 477 | |
5f8097be | 478 | if (level <= 0 && first_missing < subdevs * 2) { |
570510ba NB |
479 | fprintf(stderr, |
480 | Name ": This level does not support missing devices\n"); | |
481 | return 1; | |
482 | } | |
aba69144 | 483 | |
7f91af49 | 484 | /* We need to create the device */ |
ad5bc697 | 485 | map_lock(&map); |
a04d5763 | 486 | mdfd = create_mddev(mddev, name, autof, LOCAL, chosen_name); |
7f91af49 N |
487 | if (mdfd < 0) |
488 | return 1; | |
f2be09f1 | 489 | mddev = chosen_name; |
7f91af49 N |
490 | |
491 | vers = md_get_version(mdfd); | |
492 | if (vers < 9000) { | |
493 | fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n"); | |
494 | goto abort; | |
495 | } else { | |
496 | mdu_array_info_t inf; | |
497 | memset(&inf, 0, sizeof(inf)); | |
498 | ioctl(mdfd, GET_ARRAY_INFO, &inf); | |
499 | if (inf.working_disks != 0) { | |
500 | fprintf(stderr, Name ": another array by this name" | |
501 | " is already running.\n"); | |
502 | goto abort; | |
503 | } | |
504 | } | |
a04d5763 | 505 | |
682c7051 NB |
506 | /* Ok, lets try some ioctls */ |
507 | ||
06c7f68e NB |
508 | info.array.level = level; |
509 | info.array.size = size; | |
510 | info.array.raid_disks = raiddisks; | |
82b27616 NB |
511 | /* The kernel should *know* what md_minor we are dealing |
512 | * with, but it chooses to trust me instead. Sigh | |
513 | */ | |
06c7f68e | 514 | info.array.md_minor = 0; |
82b27616 | 515 | if (fstat(mdfd, &stb)==0) |
06c7f68e NB |
516 | info.array.md_minor = minor(stb.st_rdev); |
517 | info.array.not_persistent = 0; | |
6fb79233 | 518 | |
66f8bbbe | 519 | if ( ( (level == 4 || level == 5) && |
b5e64645 NB |
520 | (insert_point < raiddisks || first_missing < raiddisks) ) |
521 | || | |
6fb79233 NB |
522 | ( level == 6 && (insert_point < raiddisks |
523 | || second_missing < raiddisks)) | |
47d79ef8 | 524 | || |
d024b0a7 NB |
525 | ( level <= 0 ) |
526 | || | |
47d79ef8 | 527 | assume_clean |
103f2410 | 528 | ) { |
06c7f68e | 529 | info.array.state = 1; /* clean, but one+ drive will be missing*/ |
103f2410 NB |
530 | info.resync_start = ~0ULL; |
531 | } else { | |
06c7f68e | 532 | info.array.state = 0; /* not clean, but no errors */ |
103f2410 NB |
533 | info.resync_start = 0; |
534 | } | |
f9c25f1d NB |
535 | if (level == 10) { |
536 | /* for raid10, the bitmap size is the capacity of the array, | |
537 | * which is array.size * raid_disks / ncopies; | |
538 | * .. but convert to sectors. | |
539 | */ | |
702b557b | 540 | int ncopies = ((layout>>8) & 255) * (layout & 255); |
5dd497ee NB |
541 | bitmapsize = (unsigned long long)size * raiddisks / ncopies * 2; |
542 | /* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, size, raiddisks, ncopies);*/ | |
f9c25f1d | 543 | } else |
5dd497ee | 544 | bitmapsize = (unsigned long long)size * 2; |
f9c25f1d | 545 | |
82b27616 NB |
546 | /* There is lots of redundancy in these disk counts, |
547 | * raid_disks is the most meaningful value | |
548 | * it describes the geometry of the array | |
549 | * it is constant | |
550 | * nr_disks is total number of used slots. | |
551 | * it should be raid_disks+spare_disks | |
552 | * spare_disks is the number of extra disks present | |
553 | * see above | |
554 | * active_disks is the number of working disks in | |
555 | * active slots. (With raid_disks) | |
556 | * working_disks is the total number of working disks, | |
557 | * including spares | |
558 | * failed_disks is the number of disks marked failed | |
559 | * | |
560 | * Ideally, the kernel would keep these (except raid_disks) | |
561 | * up-to-date as we ADD_NEW_DISK, but it doesn't (yet). | |
562 | * So for now, we assume that all raid and spare | |
563 | * devices will be given. | |
564 | */ | |
06c7f68e NB |
565 | info.array.spare_disks=sparedisks; |
566 | info.array.failed_disks=missing_disks; | |
567 | info.array.nr_disks = info.array.working_disks | |
568 | + info.array.failed_disks; | |
569 | info.array.layout = layout; | |
570 | info.array.chunk_size = chunk*1024; | |
82d9eba6 | 571 | |
b3b33eb5 NB |
572 | if (name == NULL || *name == 0) { |
573 | /* base name on mddev */ | |
d1e80164 NB |
574 | /* /dev/md0 -> 0 |
575 | * /dev/md_d0 -> d0 | |
576 | * /dev/md/1 -> 1 | |
577 | * /dev/md/d1 -> d1 | |
578 | * /dev/md/home -> home | |
579 | * /dev/mdhome -> home | |
580 | */ | |
69207ff6 | 581 | /* FIXME compare this with rules in create_mddev */ |
b3b33eb5 NB |
582 | name = strrchr(mddev, '/'); |
583 | if (name) { | |
584 | name++; | |
60248f74 NB |
585 | if (strncmp(name, "md_d", 4)==0 && |
586 | strlen(name) > 4 && | |
587 | isdigit(name[4]) && | |
b3b33eb5 | 588 | (name-mddev) == 5 /* /dev/ */) |
d1e80164 | 589 | name += 3; |
60248f74 NB |
590 | else if (strncmp(name, "md", 2)==0 && |
591 | strlen(name) > 2 && | |
592 | isdigit(name[2]) && | |
593 | (name-mddev) == 5 /* /dev/ */) | |
b3b33eb5 NB |
594 | name += 2; |
595 | } | |
596 | } | |
06c7f68e | 597 | if (!st->ss->init_super(st, &info.array, size, name, homehost, uuid)) |
7f91af49 | 598 | goto abort; |
682c7051 | 599 | |
d2ca6449 | 600 | total_slots = info.array.nr_disks; |
f35f2525 | 601 | sysfs_init(&info, mdfd, 0); |
159c3a1a NB |
602 | st->ss->getinfo_super(st, &info); |
603 | ||
f7f1b6a1 | 604 | if (did_default && verbose >= 0) { |
3c558363 | 605 | if (is_subarray(info.text_version)) { |
f7f1b6a1 NB |
606 | int dnum = devname2devnum(info.text_version+1); |
607 | char *path; | |
608 | int mdp = get_mdp_major(); | |
609 | struct mdinfo *mdi; | |
610 | if (dnum > 0) | |
611 | path = map_dev(MD_MAJOR, dnum, 1); | |
612 | else | |
613 | path = map_dev(mdp, (-1-dnum)<< 6, 1); | |
614 | ||
615 | mdi = sysfs_read(-1, dnum, GET_VERSION); | |
616 | ||
617 | fprintf(stderr, Name ": Creating array inside " | |
618 | "%s container %s\n", | |
619 | mdi?mdi->text_version:"managed", path); | |
620 | sysfs_free(mdi); | |
621 | } else | |
622 | fprintf(stderr, Name ": Defaulting to version" | |
623 | " %s metadata\n", info.text_version); | |
624 | } | |
b8ac1967 | 625 | |
ad5bc697 | 626 | map_update(&map, fd2devnum(mdfd), info.text_version, |
a04d5763 | 627 | info.uuid, chosen_name); |
ad5bc697 | 628 | map_unlock(&map); |
a04d5763 | 629 | |
dcec9ee5 | 630 | if (bitmap_file && vers < 9003) { |
943eafef | 631 | major_num = BITMAP_MAJOR_HOSTENDIAN; |
dcec9ee5 NB |
632 | #ifdef __BIG_ENDIAN |
633 | fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n" | |
634 | " between different architectured. Consider upgrading the Linux kernel.\n"); | |
635 | #endif | |
636 | } | |
637 | ||
55935d51 NB |
638 | if (bitmap_file && strcmp(bitmap_file, "internal")==0) { |
639 | if ((vers%100) < 2) { | |
640 | fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n"); | |
7f91af49 | 641 | goto abort; |
55935d51 | 642 | } |
3da92f27 | 643 | if (!st->ss->add_internal_bitmap(st, &bitmap_chunk, |
199171a2 | 644 | delay, write_behind, |
943eafef | 645 | bitmapsize, 1, major_num)) { |
55935d51 | 646 | fprintf(stderr, Name ": Given bitmap chunk size not supported.\n"); |
7f91af49 | 647 | goto abort; |
55935d51 NB |
648 | } |
649 | bitmap_file = NULL; | |
650 | } | |
651 | ||
652 | ||
f35f2525 | 653 | sysfs_init(&info, mdfd, 0); |
55935d51 | 654 | |
f35f2525 N |
655 | if (st->ss->external && st->subarray[0]) { |
656 | /* member */ | |
657 | ||
658 | /* When creating a member, we need to be careful | |
659 | * to negotiate with mdmon properly. | |
660 | * If it is already running, we cannot write to | |
661 | * the devices and must ask it to do that part. | |
662 | * If it isn't running, we write to the devices, | |
663 | * and then start it. | |
664 | * We hold an exclusive open on the container | |
665 | * device to make sure mdmon doesn't exit after | |
666 | * we checked that it is running. | |
667 | * | |
668 | * For now, fail if it is already running. | |
669 | */ | |
670 | container_fd = open_dev_excl(st->container_dev); | |
671 | if (container_fd < 0) { | |
672 | fprintf(stderr, Name ": Cannot get exclusive " | |
673 | "open on container - weird.\n"); | |
7f91af49 | 674 | goto abort; |
d03373f1 | 675 | } |
f35f2525 N |
676 | if (mdmon_running(st->container_dev)) { |
677 | if (verbose) | |
678 | fprintf(stderr, Name ": reusing mdmon " | |
679 | "for %s.\n", | |
680 | devnum2devname(st->container_dev)); | |
681 | st->update_tail = &st->updates; | |
682 | } else | |
683 | need_mdmon = 1; | |
684 | } | |
685 | rv = set_array_info(mdfd, st, &info); | |
82d9eba6 | 686 | if (rv) { |
f35f2525 | 687 | fprintf(stderr, Name ": failed to set array info for %s: %s\n", |
52826846 | 688 | mddev, strerror(errno)); |
7f91af49 | 689 | goto abort; |
682c7051 | 690 | } |
52826846 | 691 | |
c82f047c NB |
692 | if (bitmap_file) { |
693 | int uuid[4]; | |
55935d51 | 694 | |
3da92f27 | 695 | st->ss->uuid_from_super(st, uuid); |
dfd4d8ee | 696 | if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk, |
dcec9ee5 | 697 | delay, write_behind, |
f9c25f1d | 698 | bitmapsize, |
943eafef | 699 | major_num)) { |
7f91af49 | 700 | goto abort; |
c82f047c NB |
701 | } |
702 | bitmap_fd = open(bitmap_file, O_RDWR); | |
703 | if (bitmap_fd < 0) { | |
704 | fprintf(stderr, Name ": weird: %s cannot be openned\n", | |
dcec9ee5 | 705 | bitmap_file); |
7f91af49 | 706 | goto abort; |
c82f047c NB |
707 | } |
708 | if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) { | |
709 | fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n", | |
710 | mddev, strerror(errno)); | |
7f91af49 | 711 | goto abort; |
c82f047c NB |
712 | } |
713 | } | |
714 | ||
d2ca6449 | 715 | infos = malloc(sizeof(*infos) * total_slots); |
4b1ac34b | 716 | |
4b1ac34b NB |
717 | for (pass=1; pass <=2 ; pass++) { |
718 | mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */ | |
719 | ||
720 | for (dnum=0, dv = devlist ; dv ; | |
721 | dv=(dv->next)?(dv->next):moved_disk, dnum++) { | |
722 | int fd; | |
723 | struct stat stb; | |
d2ca6449 | 724 | struct mdinfo *inf = &infos[dnum]; |
4b1ac34b | 725 | |
d2ca6449 NB |
726 | if (dnum >= total_slots) |
727 | abort(); | |
4b1ac34b NB |
728 | if (dnum == insert_point) { |
729 | moved_disk = dv; | |
8592f29d | 730 | continue; |
52826846 | 731 | } |
8592f29d | 732 | if (strcasecmp(dv->devname, "missing")==0) |
111d01fc | 733 | continue; |
8592f29d N |
734 | if (have_container) |
735 | moved_disk = NULL; | |
736 | if (have_container && dnum < info.array.raid_disks - 1) | |
737 | /* repeatedly use the container */ | |
738 | moved_disk = dv; | |
111d01fc | 739 | |
d2ca6449 NB |
740 | switch(pass) { |
741 | case 1: | |
742 | *inf = info; | |
5f8097be | 743 | |
d2ca6449 NB |
744 | inf->disk.number = dnum; |
745 | inf->disk.raid_disk = dnum; | |
746 | if (inf->disk.raid_disk < raiddisks) | |
747 | inf->disk.state = (1<<MD_DISK_ACTIVE) | | |
748 | (1<<MD_DISK_SYNC); | |
749 | else | |
750 | inf->disk.state = 0; | |
751 | ||
b01b06bd | 752 | if (dv->writemostly == 1) |
d2ca6449 NB |
753 | inf->disk.state |= (1<<MD_DISK_WRITEMOSTLY); |
754 | ||
8592f29d N |
755 | if (have_container) |
756 | fd = -1; | |
757 | else { | |
758 | if (st->ss->external && st->subarray[0]) | |
759 | fd = open(dv->devname, O_RDWR); | |
760 | else | |
761 | fd = open(dv->devname, O_RDWR|O_EXCL); | |
762 | ||
763 | if (fd < 0) { | |
764 | fprintf(stderr, Name ": failed to open %s " | |
765 | "after earlier success - aborting\n", | |
766 | dv->devname); | |
767 | goto abort; | |
768 | } | |
769 | fstat(fd, &stb); | |
770 | inf->disk.major = major(stb.st_rdev); | |
771 | inf->disk.minor = minor(stb.st_rdev); | |
d2ca6449 | 772 | } |
8592f29d N |
773 | if (fd >= 0) |
774 | remove_partitions(fd); | |
f20c3968 | 775 | if (st->ss->add_to_super(st, &inf->disk, |
208933a7 | 776 | fd, dv->devname)) |
f20c3968 | 777 | goto abort; |
d2ca6449 | 778 | st->ss->getinfo_super(st, inf); |
8ed3e5e1 | 779 | safe_mode_delay = inf->safe_mode_delay; |
d2ca6449 | 780 | |
8592f29d N |
781 | if (have_container && verbose > 0) |
782 | fprintf(stderr, Name ": Using %s for device %d\n", | |
783 | map_dev(inf->disk.major, | |
784 | inf->disk.minor, | |
785 | 0), dnum); | |
786 | ||
787 | if (!have_container) { | |
788 | /* getinfo_super might have lost these ... */ | |
789 | inf->disk.major = major(stb.st_rdev); | |
790 | inf->disk.minor = minor(stb.st_rdev); | |
791 | } | |
4b1ac34b NB |
792 | break; |
793 | case 2: | |
d2ca6449 | 794 | inf->errors = 0; |
2503d23b | 795 | rv = 0; |
4b1ac34b | 796 | |
f35f2525 | 797 | rv = add_disk(mdfd, st, &info, inf); |
d2ca6449 | 798 | |
2503d23b | 799 | if (rv) { |
d03373f1 NB |
800 | fprintf(stderr, |
801 | Name ": ADD_NEW_DISK for %s " | |
802 | "failed: %s\n", | |
4b1ac34b | 803 | dv->devname, strerror(errno)); |
3da92f27 | 804 | st->ss->free_super(st); |
7f91af49 | 805 | goto abort; |
4b1ac34b | 806 | } |
4b1ac34b NB |
807 | break; |
808 | } | |
8592f29d N |
809 | if (!have_container && |
810 | dv == moved_disk && dnum != insert_point) break; | |
52826846 | 811 | } |
edd8d13c | 812 | if (pass == 1) { |
9b1fb677 DW |
813 | struct mdinfo info_new; |
814 | struct map_ent *me = NULL; | |
815 | ||
816 | /* check to see if the uuid has changed due to these | |
817 | * metadata changes, and if so update the member array | |
818 | * and container uuid. Note ->write_init_super clears | |
819 | * the subarray cursor such that ->getinfo_super once | |
820 | * again returns container info. | |
821 | */ | |
822 | map_lock(&map); | |
823 | st->ss->getinfo_super(st, &info_new); | |
824 | if (st->ss->external && level != LEVEL_CONTAINER && | |
825 | !same_uuid(info_new.uuid, info.uuid, 0)) { | |
826 | map_update(&map, fd2devnum(mdfd), | |
827 | info_new.text_version, | |
828 | info_new.uuid, chosen_name); | |
829 | me = map_by_devnum(&map, st->container_dev); | |
830 | } | |
831 | ||
111d01fc | 832 | st->ss->write_init_super(st); |
9b1fb677 DW |
833 | |
834 | /* update parent container uuid */ | |
835 | if (me) { | |
836 | char *path = strdup(me->path); | |
837 | ||
838 | st->ss->getinfo_super(st, &info_new); | |
839 | map_update(&map, st->container_dev, | |
840 | info_new.text_version, | |
841 | info_new.uuid, path); | |
842 | free(path); | |
843 | } | |
844 | map_unlock(&map); | |
845 | ||
edd8d13c NB |
846 | flush_metadata_updates(st); |
847 | } | |
5e52ae9e | 848 | } |
d2ca6449 | 849 | free(infos); |
3da92f27 | 850 | st->ss->free_super(st); |
5e52ae9e | 851 | |
97590376 N |
852 | if (level == LEVEL_CONTAINER) { |
853 | /* No need to start. But we should signal udev to | |
854 | * create links */ | |
855 | sysfs_uevent(&info, "change"); | |
ed034b11 N |
856 | if (verbose >= 0) |
857 | fprintf(stderr, Name ": container %s prepared.\n", mddev); | |
a7c6e3fb | 858 | wait_for(chosen_name, mdfd); |
97590376 | 859 | } else if (runstop == 1 || subdevs >= raiddisks) { |
598f0d58 NB |
860 | if (st->ss->external) { |
861 | switch(level) { | |
862 | case LEVEL_LINEAR: | |
863 | case LEVEL_MULTIPATH: | |
864 | case 0: | |
f35f2525 | 865 | sysfs_set_str(&info, NULL, "array_state", |
598f0d58 | 866 | "active"); |
a931db9e | 867 | need_mdmon = 0; |
598f0d58 NB |
868 | break; |
869 | default: | |
f35f2525 | 870 | sysfs_set_str(&info, NULL, "array_state", |
598f0d58 NB |
871 | "readonly"); |
872 | break; | |
873 | } | |
f35f2525 | 874 | sysfs_set_safemode(&info, safe_mode_delay); |
598f0d58 | 875 | } else { |
97590376 | 876 | /* param is not actually used */ |
598f0d58 NB |
877 | mdu_param_t param; |
878 | if (ioctl(mdfd, RUN_ARRAY, ¶m)) { | |
879 | fprintf(stderr, Name ": RUN_ARRAY failed: %s\n", | |
880 | strerror(errno)); | |
881 | Manage_runstop(mddev, mdfd, -1, 0); | |
7f91af49 | 882 | goto abort; |
598f0d58 | 883 | } |
682c7051 | 884 | } |
dab6685f NB |
885 | if (verbose >= 0) |
886 | fprintf(stderr, Name ": array %s started.\n", mddev); | |
f7e7067b | 887 | if (st->ss->external && st->subarray[0]) { |
8850ee3e N |
888 | if (need_mdmon) |
889 | start_mdmon(st->container_dev); | |
890 | ||
891 | ping_monitor(devnum2devname(st->container_dev)); | |
a931db9e NB |
892 | close(container_fd); |
893 | } | |
a7c6e3fb | 894 | wait_for(chosen_name, mdfd); |
682c7051 | 895 | } else { |
b83d95f3 | 896 | fprintf(stderr, Name ": not starting array - not enough devices.\n"); |
682c7051 | 897 | } |
7f91af49 | 898 | close(mdfd); |
682c7051 | 899 | return 0; |
7f91af49 N |
900 | |
901 | abort: | |
902 | if (mdfd >= 0) | |
903 | close(mdfd); | |
904 | return 1; | |
64c4757e | 905 | } |