]>
Commit | Line | Data |
---|---|---|
64c4757e | 1 | /* |
9a9dab36 | 2 | * mdadm - manage Linux "md" devices aka RAID arrays. |
64c4757e | 3 | * |
4f589ad0 | 4 | * Copyright (C) 2001-2006 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 | |
22 | * Email: <neilb@cse.unsw.edu.au> | |
23 | * Paper: Neil Brown | |
24 | * School of Computer Science and Engineering | |
25 | * The University of New South Wales | |
26 | * Sydney, 2052 | |
27 | * Australia | |
28 | */ | |
29 | ||
9a9dab36 | 30 | #include "mdadm.h" |
682c7051 NB |
31 | #include "md_u.h" |
32 | #include "md_p.h" | |
60248f74 | 33 | #include <ctype.h> |
64c4757e | 34 | |
82d9eba6 | 35 | int Create(struct supertype *st, char *mddev, int mdfd, |
5dd497ee | 36 | int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks, |
3d3dd91e | 37 | char *name, char *homehost, int *uuid, |
cd29a5c8 | 38 | int subdevs, mddev_dev_t devlist, |
47d79ef8 | 39 | int runstop, int verbose, int force, int assume_clean, |
dfd4d8ee | 40 | char *bitmap_file, int bitmap_chunk, int write_behind, int delay) |
64c4757e | 41 | { |
682c7051 NB |
42 | /* |
43 | * Create a new raid array. | |
44 | * | |
45 | * First check that necessary details are available | |
46 | * (i.e. level, raid-disks) | |
47 | * | |
48 | * Then check each disk to see what might be on it | |
49 | * and report anything interesting. | |
50 | * | |
51 | * If anything looks odd, and runstop not set, | |
52 | * abort. | |
53 | * | |
54 | * SET_ARRAY_INFO and ADD_NEW_DISK, and | |
82d9eba6 | 55 | * if runstop==run, or raiddisks disks were used, |
682c7051 NB |
56 | * RUN_ARRAY |
57 | */ | |
b5e64645 | 58 | unsigned long long minsize=0, maxsize=0; |
cd29a5c8 NB |
59 | char *mindisc = NULL; |
60 | char *maxdisc = NULL; | |
c913b90e | 61 | int dnum; |
cd29a5c8 | 62 | mddev_dev_t dv; |
682c7051 | 63 | int fail=0, warn=0; |
82b27616 | 64 | struct stat stb; |
82d9eba6 | 65 | int first_missing = subdevs * 2; |
6fb79233 | 66 | int second_missing = subdevs * 2; |
52826846 | 67 | int missing_disks = 0; |
82d9eba6 | 68 | int insert_point = subdevs * 2; /* where to insert a missing drive */ |
d2ca6449 | 69 | int total_slots; |
4b1ac34b | 70 | int pass; |
82d9eba6 NB |
71 | int vers; |
72 | int rv; | |
c82f047c | 73 | int bitmap_fd; |
5f8097be | 74 | int have_container = 0; |
0e600426 | 75 | int container_fd = -1; |
a931db9e | 76 | int need_mdmon = 0; |
f9c25f1d | 77 | unsigned long long bitmapsize; |
d2ca6449 | 78 | struct mdinfo info, *infos; |
b8ac1967 | 79 | int did_default = 0; |
8ed3e5e1 | 80 | unsigned long safe_mode_delay = 0; |
682c7051 | 81 | |
943eafef | 82 | int major_num = BITMAP_MAJOR_HI; |
dcec9ee5 | 83 | |
06c7f68e | 84 | memset(&info, 0, sizeof(info)); |
29e766a5 | 85 | |
82d9eba6 NB |
86 | vers = md_get_version(mdfd); |
87 | if (vers < 9000) { | |
dd0781e5 | 88 | fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n"); |
52826846 | 89 | return 1; |
32e5a4ee DL |
90 | } else { |
91 | mdu_array_info_t inf; | |
92 | memset(&inf, 0, sizeof(inf)); | |
93 | ioctl(mdfd, GET_ARRAY_INFO, &inf); | |
94 | if (inf.working_disks != 0) { | |
95 | fprintf(stderr, Name ": another array by this name" | |
96 | " is already running.\n"); | |
97 | return 1; | |
98 | } | |
682c7051 | 99 | } |
a322f70c | 100 | if (level == UnSet) { |
3cfe6247 | 101 | /* "ddf" and "imsm" metadata only supports one level - should possibly |
a322f70c DW |
102 | * push this into metadata handler?? |
103 | */ | |
3cfe6247 | 104 | if (st && (st->ss == &super_ddf || st->ss == &super_imsm)) |
a322f70c DW |
105 | level = LEVEL_CONTAINER; |
106 | } | |
107 | ||
98c6faba | 108 | if (level == UnSet) { |
682c7051 NB |
109 | fprintf(stderr, |
110 | Name ": a RAID level is needed to create an array.\n"); | |
111 | return 1; | |
112 | } | |
98c6faba NB |
113 | if (raiddisks < 4 && level == 6) { |
114 | fprintf(stderr, | |
115 | Name ": at least 4 raid-devices needed for level 6\n"); | |
116 | return 1; | |
117 | } | |
118 | if (raiddisks > 256 && level == 6) { | |
119 | fprintf(stderr, | |
120 | Name ": no more than 256 raid-devices supported for level 6\n"); | |
121 | return 1; | |
122 | } | |
52826846 NB |
123 | if (raiddisks < 2 && level >= 4) { |
124 | fprintf(stderr, | |
98c6faba | 125 | Name ": at least 2 raid-devices needed for level 4 or 5\n"); |
52826846 NB |
126 | return 1; |
127 | } | |
570510ba NB |
128 | if (level <= 0 && sparedisks) { |
129 | fprintf(stderr, | |
130 | Name ": This level does not support spare devices\n"); | |
131 | return 1; | |
132 | } | |
5f8097be NB |
133 | |
134 | if (subdevs == 1 && strcmp(devlist->devname, "missing") != 0) { | |
135 | /* If given a single device, it might be a container, and we can | |
136 | * extract a device list from there | |
137 | */ | |
138 | mdu_array_info_t inf; | |
139 | int fd; | |
140 | ||
141 | memset(&inf, 0, sizeof(inf)); | |
37ea3936 | 142 | fd = open(devlist->devname, O_RDONLY); |
5f8097be NB |
143 | if (fd >= 0 && |
144 | ioctl(fd, GET_ARRAY_INFO, &inf) == 0 && | |
145 | inf.raid_disks == 0) { | |
146 | /* yep, looks like a container */ | |
147 | if (st) { | |
148 | rv = st->ss->load_super(st, fd, | |
149 | devlist->devname); | |
150 | if (rv == 0) | |
151 | have_container = 1; | |
152 | } else { | |
153 | st = guess_super(fd); | |
154 | if (st && !(rv = st->ss-> | |
155 | load_super(st, fd, | |
156 | devlist->devname))) | |
157 | have_container = 1; | |
158 | else | |
159 | st = NULL; | |
160 | } | |
161 | } | |
162 | if (fd >= 0) | |
163 | close(fd); | |
164 | if (have_container) { | |
165 | subdevs = 0; | |
166 | devlist = NULL; | |
167 | } | |
168 | } | |
ffcfc735 N |
169 | if (st && st->ss->external && sparedisks) { |
170 | fprintf(stderr, | |
171 | Name ": This metadata type does not support " | |
172 | "spare disks are create time\n"); | |
173 | return 1; | |
174 | } | |
682c7051 | 175 | if (subdevs > raiddisks+sparedisks) { |
b83d95f3 | 176 | fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks); |
52826846 | 177 | return 1; |
682c7051 | 178 | } |
5f8097be | 179 | if (!have_container && subdevs < raiddisks+sparedisks) { |
52826846 NB |
180 | fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n"); |
181 | return 1; | |
182 | } | |
5b28bd56 NB |
183 | if (bitmap_file && level <= 0) { |
184 | fprintf(stderr, Name ": bitmaps not meaningful with level %s\n", | |
185 | map_num(pers, level)?:"given"); | |
186 | return 1; | |
187 | } | |
52826846 | 188 | |
682c7051 | 189 | /* now set some defaults */ |
98c6faba | 190 | if (layout == UnSet) |
682c7051 NB |
191 | switch(level) { |
192 | default: /* no layout */ | |
193 | layout = 0; | |
194 | break; | |
e5329c37 NB |
195 | case 10: |
196 | layout = 0x102; /* near=2, far=1 */ | |
dab6685f | 197 | if (verbose > 0) |
e5329c37 NB |
198 | fprintf(stderr, |
199 | Name ": layout defaults to n1\n"); | |
200 | break; | |
682c7051 | 201 | case 5: |
98c6faba | 202 | case 6: |
682c7051 | 203 | layout = map_name(r5layout, "default"); |
dab6685f | 204 | if (verbose > 0) |
682c7051 NB |
205 | fprintf(stderr, |
206 | Name ": layout defaults to %s\n", map_num(r5layout, layout)); | |
207 | break; | |
b5e64645 NB |
208 | case LEVEL_FAULTY: |
209 | layout = map_name(faultylayout, "default"); | |
210 | ||
dab6685f | 211 | if (verbose > 0) |
b5e64645 NB |
212 | fprintf(stderr, |
213 | Name ": layout defaults to %s\n", map_num(faultylayout, layout)); | |
214 | break; | |
682c7051 NB |
215 | } |
216 | ||
e5329c37 NB |
217 | if (level == 10) |
218 | /* check layout fits in array*/ | |
219 | if ((layout&255) * ((layout>>8)&255) > raiddisks) { | |
220 | fprintf(stderr, Name ": that layout requires at least %d devices\n", | |
221 | (layout&255) * ((layout>>8)&255)); | |
222 | return 1; | |
223 | } | |
224 | ||
aa88f531 NB |
225 | switch(level) { |
226 | case 4: | |
227 | case 5: | |
e5329c37 | 228 | case 10: |
98c6faba | 229 | case 6: |
aa88f531 | 230 | case 0: |
570510ba | 231 | case LEVEL_LINEAR: /* linear */ |
aa88f531 NB |
232 | if (chunk == 0) { |
233 | chunk = 64; | |
dab6685f | 234 | if (verbose > 0) |
aa88f531 NB |
235 | fprintf(stderr, Name ": chunk size defaults to 64K\n"); |
236 | } | |
237 | break; | |
570510ba | 238 | case 1: |
08e43379 | 239 | case LEVEL_FAULTY: |
570510ba | 240 | case LEVEL_MULTIPATH: |
17f25ca6 | 241 | case LEVEL_CONTAINER: |
aa88f531 NB |
242 | if (chunk) { |
243 | chunk = 0; | |
dab6685f | 244 | if (verbose > 0) |
aa88f531 NB |
245 | fprintf(stderr, Name ": chunk size ignored for this level\n"); |
246 | } | |
247 | break; | |
570510ba NB |
248 | default: |
249 | fprintf(stderr, Name ": unknown level %d\n", level); | |
250 | return 1; | |
682c7051 NB |
251 | } |
252 | ||
17f25ca6 | 253 | if (st && ! st->ss->validate_geometry(st, level, layout, raiddisks, |
2c514b71 | 254 | chunk, size, NULL, NULL, verbose>=0)) |
17f25ca6 NB |
255 | return 1; |
256 | ||
682c7051 | 257 | /* now look at the subdevs */ |
06c7f68e NB |
258 | info.array.active_disks = 0; |
259 | info.array.working_disks = 0; | |
c913b90e NB |
260 | dnum = 0; |
261 | for (dv=devlist; dv; dv=dv->next, dnum++) { | |
cd29a5c8 | 262 | char *dname = dv->devname; |
17f25ca6 | 263 | unsigned long long freesize; |
cd29a5c8 | 264 | if (strcasecmp(dname, "missing")==0) { |
c913b90e NB |
265 | if (first_missing > dnum) |
266 | first_missing = dnum; | |
6fb79233 NB |
267 | if (second_missing > dnum && dnum > first_missing) |
268 | second_missing = dnum; | |
52826846 NB |
269 | missing_disks ++; |
270 | continue; | |
271 | } | |
06c7f68e | 272 | info.array.working_disks++; |
c913b90e | 273 | if (dnum < raiddisks) |
06c7f68e | 274 | info.array.active_disks++; |
058574b1 | 275 | if (st == NULL) { |
8aec876d | 276 | struct createinfo *ci = conf_get_create_info(); |
058574b1 NB |
277 | if (ci) |
278 | st = ci->supertype; | |
279 | } | |
576d6d83 NB |
280 | if (st == NULL) { |
281 | /* Need to choose a default metadata, which is different | |
17f25ca6 | 282 | * depending on geometry of array. |
576d6d83 NB |
283 | */ |
284 | int i; | |
285 | char *name = "default"; | |
17f25ca6 | 286 | for(i=0; !st && superlist[i]; i++) { |
576d6d83 | 287 | st = superlist[i]->match_metadata_desc(name); |
17f25ca6 NB |
288 | if (st && !st->ss->validate_geometry |
289 | (st, level, layout, raiddisks, | |
2c514b71 NB |
290 | chunk, size, dname, &freesize, |
291 | verbose > 0)) | |
17f25ca6 NB |
292 | st = NULL; |
293 | } | |
576d6d83 NB |
294 | |
295 | if (!st) { | |
17f25ca6 NB |
296 | fprintf(stderr, Name ": device %s not suitable " |
297 | "for any style of array\n", | |
298 | dname); | |
576d6d83 NB |
299 | exit(2); |
300 | } | |
b8ac1967 | 301 | if (st->ss != &super0 || |
576d6d83 | 302 | st->minor_version != 90) |
b8ac1967 | 303 | did_default = 1; |
17f25ca6 NB |
304 | } else { |
305 | if (!st->ss->validate_geometry(st, level, layout, | |
306 | raiddisks, | |
307 | chunk, size, dname, | |
2c514b71 NB |
308 | &freesize, |
309 | verbose > 0)) { | |
17f25ca6 NB |
310 | |
311 | fprintf(stderr, | |
312 | Name ": %s is not suitable for " | |
313 | "this array.\n", | |
314 | dname); | |
315 | fail = 1; | |
316 | continue; | |
317 | } | |
682c7051 | 318 | } |
82d9eba6 NB |
319 | |
320 | freesize /= 2; /* convert to K */ | |
d2cd3ffc NB |
321 | if (chunk) { |
322 | /* round to chunk size */ | |
323 | freesize = freesize & ~(chunk-1); | |
324 | } | |
682c7051 NB |
325 | |
326 | if (size && freesize < size) { | |
52826846 | 327 | fprintf(stderr, Name ": %s is smaller that given size." |
17f25ca6 NB |
328 | " %lluK < %lluK + metadata\n", |
329 | dname, freesize, size); | |
52826846 | 330 | fail = 1; |
52826846 | 331 | continue; |
682c7051 | 332 | } |
cd29a5c8 NB |
333 | if (maxdisc == NULL || (maxdisc && freesize > maxsize)) { |
334 | maxdisc = dname; | |
52826846 | 335 | maxsize = freesize; |
682c7051 | 336 | } |
cd29a5c8 NB |
337 | if (mindisc ==NULL || (mindisc && freesize < minsize)) { |
338 | mindisc = dname; | |
52826846 | 339 | minsize = freesize; |
682c7051 | 340 | } |
dab6685f | 341 | if (runstop != 1 || verbose >= 0) { |
37ea3936 | 342 | int fd = open(dname, O_RDONLY); |
b6e63da4 NB |
343 | if (fd <0 ) { |
344 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
345 | dname, strerror(errno)); | |
346 | fail=1; | |
347 | continue; | |
348 | } | |
dab6685f NB |
349 | warn |= check_ext2(fd, dname); |
350 | warn |= check_reiser(fd, dname); | |
351 | warn |= check_raid(fd, dname); | |
b6e63da4 | 352 | close(fd); |
dab6685f | 353 | } |
682c7051 NB |
354 | } |
355 | if (fail) { | |
52826846 NB |
356 | fprintf(stderr, Name ": create aborted\n"); |
357 | return 1; | |
682c7051 NB |
358 | } |
359 | if (size == 0) { | |
5f8097be | 360 | if (mindisc == NULL && !have_container) { |
52826846 NB |
361 | fprintf(stderr, Name ": no size and no drives given - aborting create.\n"); |
362 | return 1; | |
363 | } | |
5f8097be NB |
364 | if (level > 0 || level == LEVEL_MULTIPATH |
365 | || level == LEVEL_FAULTY | |
b8ac1967 | 366 | || st->ss->external ) { |
b5e64645 | 367 | /* size is meaningful */ |
b8ac1967 NB |
368 | if (!st->ss->validate_geometry(st, level, layout, |
369 | raiddisks, | |
370 | chunk, minsize, | |
2c514b71 | 371 | NULL, NULL, 0)) { |
aba69144 | 372 | fprintf(stderr, Name ": devices too large for RAID level %d\n", level); |
b5e64645 NB |
373 | return 1; |
374 | } | |
375 | size = minsize; | |
dab6685f | 376 | if (verbose > 0) |
5dd497ee | 377 | fprintf(stderr, Name ": size set to %lluK\n", size); |
b5e64645 | 378 | } |
682c7051 | 379 | } |
b5e64645 | 380 | if (level > 0 && ((maxsize-size)*100 > maxsize)) { |
dab6685f | 381 | if (runstop != 1 || verbose >= 0) |
5dd497ee | 382 | fprintf(stderr, Name ": largest drive (%s) exceed size (%lluK) by more than 1%%\n", |
dab6685f | 383 | maxdisc, size); |
52826846 | 384 | warn = 1; |
682c7051 NB |
385 | } |
386 | ||
387 | if (warn) { | |
52826846 NB |
388 | if (runstop!= 1) { |
389 | if (!ask("Continue creating array? ")) { | |
390 | fprintf(stderr, Name ": create aborted.\n"); | |
391 | return 1; | |
392 | } | |
393 | } else { | |
dab6685f | 394 | if (verbose > 0) |
52826846 | 395 | fprintf(stderr, Name ": creation continuing despite oddities due to --run\n"); |
682c7051 | 396 | } |
682c7051 NB |
397 | } |
398 | ||
66f8bbbe | 399 | /* If this is raid4/5, we want to configure the last active slot |
52826846 | 400 | * as missing, so that a reconstruct happens (faster than re-parity) |
98c6faba | 401 | * FIX: Can we do this for raid6 as well? |
52826846 | 402 | */ |
47d79ef8 | 403 | if (assume_clean==0 && force == 0 && first_missing >= raiddisks) { |
98c6faba | 404 | switch ( level ) { |
66f8bbbe | 405 | case 4: |
98c6faba NB |
406 | case 5: |
407 | insert_point = raiddisks-1; | |
408 | sparedisks++; | |
06c7f68e | 409 | info.array.active_disks--; |
98c6faba NB |
410 | missing_disks++; |
411 | break; | |
412 | default: | |
413 | break; | |
414 | } | |
52826846 | 415 | } |
6fb79233 NB |
416 | /* For raid6, if creating with 1 missing drive, make a good drive |
417 | * into a spare, else the create will fail | |
418 | */ | |
419 | if (assume_clean == 0 && force == 0 && first_missing < raiddisks && | |
ffcfc735 | 420 | st->ss->external == 0 && |
6fb79233 NB |
421 | second_missing >= raiddisks && level == 6) { |
422 | insert_point = raiddisks - 1; | |
423 | if (insert_point == first_missing) | |
424 | insert_point--; | |
425 | sparedisks ++; | |
426 | info.array.active_disks--; | |
427 | missing_disks++; | |
428 | } | |
570510ba | 429 | |
5f8097be | 430 | if (level <= 0 && first_missing < subdevs * 2) { |
570510ba NB |
431 | fprintf(stderr, |
432 | Name ": This level does not support missing devices\n"); | |
433 | return 1; | |
434 | } | |
aba69144 | 435 | |
682c7051 NB |
436 | /* Ok, lets try some ioctls */ |
437 | ||
06c7f68e NB |
438 | info.array.level = level; |
439 | info.array.size = size; | |
440 | info.array.raid_disks = raiddisks; | |
82b27616 NB |
441 | /* The kernel should *know* what md_minor we are dealing |
442 | * with, but it chooses to trust me instead. Sigh | |
443 | */ | |
06c7f68e | 444 | info.array.md_minor = 0; |
82b27616 | 445 | if (fstat(mdfd, &stb)==0) |
06c7f68e NB |
446 | info.array.md_minor = minor(stb.st_rdev); |
447 | info.array.not_persistent = 0; | |
6fb79233 | 448 | |
66f8bbbe | 449 | if ( ( (level == 4 || level == 5) && |
b5e64645 NB |
450 | (insert_point < raiddisks || first_missing < raiddisks) ) |
451 | || | |
6fb79233 NB |
452 | ( level == 6 && (insert_point < raiddisks |
453 | || second_missing < raiddisks)) | |
47d79ef8 | 454 | || |
d024b0a7 NB |
455 | ( level <= 0 ) |
456 | || | |
47d79ef8 | 457 | assume_clean |
103f2410 | 458 | ) { |
06c7f68e | 459 | info.array.state = 1; /* clean, but one+ drive will be missing*/ |
103f2410 NB |
460 | info.resync_start = ~0ULL; |
461 | } else { | |
06c7f68e | 462 | info.array.state = 0; /* not clean, but no errors */ |
103f2410 NB |
463 | info.resync_start = 0; |
464 | } | |
f9c25f1d NB |
465 | if (level == 10) { |
466 | /* for raid10, the bitmap size is the capacity of the array, | |
467 | * which is array.size * raid_disks / ncopies; | |
468 | * .. but convert to sectors. | |
469 | */ | |
702b557b | 470 | int ncopies = ((layout>>8) & 255) * (layout & 255); |
5dd497ee NB |
471 | bitmapsize = (unsigned long long)size * raiddisks / ncopies * 2; |
472 | /* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, size, raiddisks, ncopies);*/ | |
f9c25f1d | 473 | } else |
5dd497ee | 474 | bitmapsize = (unsigned long long)size * 2; |
f9c25f1d | 475 | |
82b27616 NB |
476 | /* There is lots of redundancy in these disk counts, |
477 | * raid_disks is the most meaningful value | |
478 | * it describes the geometry of the array | |
479 | * it is constant | |
480 | * nr_disks is total number of used slots. | |
481 | * it should be raid_disks+spare_disks | |
482 | * spare_disks is the number of extra disks present | |
483 | * see above | |
484 | * active_disks is the number of working disks in | |
485 | * active slots. (With raid_disks) | |
486 | * working_disks is the total number of working disks, | |
487 | * including spares | |
488 | * failed_disks is the number of disks marked failed | |
489 | * | |
490 | * Ideally, the kernel would keep these (except raid_disks) | |
491 | * up-to-date as we ADD_NEW_DISK, but it doesn't (yet). | |
492 | * So for now, we assume that all raid and spare | |
493 | * devices will be given. | |
494 | */ | |
06c7f68e NB |
495 | info.array.spare_disks=sparedisks; |
496 | info.array.failed_disks=missing_disks; | |
497 | info.array.nr_disks = info.array.working_disks | |
498 | + info.array.failed_disks; | |
499 | info.array.layout = layout; | |
500 | info.array.chunk_size = chunk*1024; | |
82d9eba6 | 501 | |
b3b33eb5 NB |
502 | if (name == NULL || *name == 0) { |
503 | /* base name on mddev */ | |
d1e80164 NB |
504 | /* /dev/md0 -> 0 |
505 | * /dev/md_d0 -> d0 | |
506 | * /dev/md/1 -> 1 | |
507 | * /dev/md/d1 -> d1 | |
508 | * /dev/md/home -> home | |
509 | * /dev/mdhome -> home | |
510 | */ | |
b3b33eb5 NB |
511 | name = strrchr(mddev, '/'); |
512 | if (name) { | |
513 | name++; | |
60248f74 NB |
514 | if (strncmp(name, "md_d", 4)==0 && |
515 | strlen(name) > 4 && | |
516 | isdigit(name[4]) && | |
b3b33eb5 | 517 | (name-mddev) == 5 /* /dev/ */) |
d1e80164 | 518 | name += 3; |
60248f74 NB |
519 | else if (strncmp(name, "md", 2)==0 && |
520 | strlen(name) > 2 && | |
521 | isdigit(name[2]) && | |
522 | (name-mddev) == 5 /* /dev/ */) | |
b3b33eb5 NB |
523 | name += 2; |
524 | } | |
525 | } | |
06c7f68e | 526 | if (!st->ss->init_super(st, &info.array, size, name, homehost, uuid)) |
82d9eba6 | 527 | return 1; |
682c7051 | 528 | |
d2ca6449 | 529 | total_slots = info.array.nr_disks; |
f35f2525 | 530 | sysfs_init(&info, mdfd, 0); |
159c3a1a NB |
531 | st->ss->getinfo_super(st, &info); |
532 | ||
f7f1b6a1 | 533 | if (did_default && verbose >= 0) { |
3c558363 | 534 | if (is_subarray(info.text_version)) { |
f7f1b6a1 NB |
535 | int dnum = devname2devnum(info.text_version+1); |
536 | char *path; | |
537 | int mdp = get_mdp_major(); | |
538 | struct mdinfo *mdi; | |
539 | if (dnum > 0) | |
540 | path = map_dev(MD_MAJOR, dnum, 1); | |
541 | else | |
542 | path = map_dev(mdp, (-1-dnum)<< 6, 1); | |
543 | ||
544 | mdi = sysfs_read(-1, dnum, GET_VERSION); | |
545 | ||
546 | fprintf(stderr, Name ": Creating array inside " | |
547 | "%s container %s\n", | |
548 | mdi?mdi->text_version:"managed", path); | |
549 | sysfs_free(mdi); | |
550 | } else | |
551 | fprintf(stderr, Name ": Defaulting to version" | |
552 | " %s metadata\n", info.text_version); | |
553 | } | |
b8ac1967 | 554 | |
dcec9ee5 | 555 | if (bitmap_file && vers < 9003) { |
943eafef | 556 | major_num = BITMAP_MAJOR_HOSTENDIAN; |
dcec9ee5 NB |
557 | #ifdef __BIG_ENDIAN |
558 | fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n" | |
559 | " between different architectured. Consider upgrading the Linux kernel.\n"); | |
560 | #endif | |
561 | } | |
562 | ||
55935d51 NB |
563 | if (bitmap_file && strcmp(bitmap_file, "internal")==0) { |
564 | if ((vers%100) < 2) { | |
565 | fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n"); | |
566 | return 1; | |
567 | } | |
3da92f27 | 568 | if (!st->ss->add_internal_bitmap(st, &bitmap_chunk, |
199171a2 | 569 | delay, write_behind, |
943eafef | 570 | bitmapsize, 1, major_num)) { |
55935d51 NB |
571 | fprintf(stderr, Name ": Given bitmap chunk size not supported.\n"); |
572 | return 1; | |
573 | } | |
574 | bitmap_file = NULL; | |
575 | } | |
576 | ||
577 | ||
f35f2525 | 578 | sysfs_init(&info, mdfd, 0); |
55935d51 | 579 | |
f35f2525 N |
580 | if (st->ss->external && st->subarray[0]) { |
581 | /* member */ | |
582 | ||
583 | /* When creating a member, we need to be careful | |
584 | * to negotiate with mdmon properly. | |
585 | * If it is already running, we cannot write to | |
586 | * the devices and must ask it to do that part. | |
587 | * If it isn't running, we write to the devices, | |
588 | * and then start it. | |
589 | * We hold an exclusive open on the container | |
590 | * device to make sure mdmon doesn't exit after | |
591 | * we checked that it is running. | |
592 | * | |
593 | * For now, fail if it is already running. | |
594 | */ | |
595 | container_fd = open_dev_excl(st->container_dev); | |
596 | if (container_fd < 0) { | |
597 | fprintf(stderr, Name ": Cannot get exclusive " | |
598 | "open on container - weird.\n"); | |
d03373f1 NB |
599 | return 1; |
600 | } | |
f35f2525 N |
601 | if (mdmon_running(st->container_dev)) { |
602 | if (verbose) | |
603 | fprintf(stderr, Name ": reusing mdmon " | |
604 | "for %s.\n", | |
605 | devnum2devname(st->container_dev)); | |
606 | st->update_tail = &st->updates; | |
607 | } else | |
608 | need_mdmon = 1; | |
609 | } | |
610 | rv = set_array_info(mdfd, st, &info); | |
82d9eba6 | 611 | if (rv) { |
f35f2525 | 612 | fprintf(stderr, Name ": failed to set array info for %s: %s\n", |
52826846 NB |
613 | mddev, strerror(errno)); |
614 | return 1; | |
682c7051 | 615 | } |
52826846 | 616 | |
c82f047c NB |
617 | if (bitmap_file) { |
618 | int uuid[4]; | |
55935d51 | 619 | |
3da92f27 | 620 | st->ss->uuid_from_super(st, uuid); |
dfd4d8ee | 621 | if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk, |
dcec9ee5 | 622 | delay, write_behind, |
f9c25f1d | 623 | bitmapsize, |
943eafef | 624 | major_num)) { |
c82f047c NB |
625 | return 1; |
626 | } | |
627 | bitmap_fd = open(bitmap_file, O_RDWR); | |
628 | if (bitmap_fd < 0) { | |
629 | fprintf(stderr, Name ": weird: %s cannot be openned\n", | |
dcec9ee5 | 630 | bitmap_file); |
c82f047c NB |
631 | return 1; |
632 | } | |
633 | if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) { | |
634 | fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n", | |
635 | mddev, strerror(errno)); | |
636 | return 1; | |
637 | } | |
638 | } | |
639 | ||
d2ca6449 | 640 | infos = malloc(sizeof(*infos) * total_slots); |
4b1ac34b | 641 | |
4b1ac34b NB |
642 | for (pass=1; pass <=2 ; pass++) { |
643 | mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */ | |
644 | ||
645 | for (dnum=0, dv = devlist ; dv ; | |
646 | dv=(dv->next)?(dv->next):moved_disk, dnum++) { | |
647 | int fd; | |
648 | struct stat stb; | |
d2ca6449 | 649 | struct mdinfo *inf = &infos[dnum]; |
4b1ac34b | 650 | |
d2ca6449 NB |
651 | if (dnum >= total_slots) |
652 | abort(); | |
4b1ac34b NB |
653 | if (dnum == insert_point) { |
654 | moved_disk = dv; | |
52826846 | 655 | } |
4b1ac34b | 656 | if (dnum == insert_point || |
111d01fc NB |
657 | strcasecmp(dv->devname, "missing")==0) |
658 | continue; | |
659 | ||
d2ca6449 NB |
660 | switch(pass) { |
661 | case 1: | |
662 | *inf = info; | |
5f8097be | 663 | |
d2ca6449 NB |
664 | inf->disk.number = dnum; |
665 | inf->disk.raid_disk = dnum; | |
666 | if (inf->disk.raid_disk < raiddisks) | |
667 | inf->disk.state = (1<<MD_DISK_ACTIVE) | | |
668 | (1<<MD_DISK_SYNC); | |
669 | else | |
670 | inf->disk.state = 0; | |
671 | ||
672 | if (dv->writemostly) | |
673 | inf->disk.state |= (1<<MD_DISK_WRITEMOSTLY); | |
674 | ||
675 | if (st->ss->external && st->subarray[0]) | |
37ea3936 | 676 | fd = open(dv->devname, O_RDWR); |
d2ca6449 | 677 | else |
37ea3936 | 678 | fd = open(dv->devname, O_RDWR|O_EXCL); |
d2ca6449 NB |
679 | |
680 | if (fd < 0) { | |
681 | fprintf(stderr, Name ": failed to open %s " | |
682 | "after earlier success - aborting\n", | |
683 | dv->devname); | |
684 | return 1; | |
685 | } | |
686 | fstat(fd, &stb); | |
687 | inf->disk.major = major(stb.st_rdev); | |
688 | inf->disk.minor = minor(stb.st_rdev); | |
111d01fc | 689 | |
111d01fc | 690 | remove_partitions(fd); |
d2ca6449 | 691 | st->ss->add_to_super(st, &inf->disk, |
111d01fc | 692 | fd, dv->devname); |
d2ca6449 | 693 | st->ss->getinfo_super(st, inf); |
8ed3e5e1 | 694 | safe_mode_delay = inf->safe_mode_delay; |
d2ca6449 NB |
695 | |
696 | /* getinfo_super might have lost these ... */ | |
697 | inf->disk.major = major(stb.st_rdev); | |
698 | inf->disk.minor = minor(stb.st_rdev); | |
4b1ac34b NB |
699 | break; |
700 | case 2: | |
d2ca6449 | 701 | inf->errors = 0; |
2503d23b | 702 | rv = 0; |
4b1ac34b | 703 | |
f35f2525 | 704 | rv = add_disk(mdfd, st, &info, inf); |
d2ca6449 | 705 | |
2503d23b | 706 | if (rv) { |
d03373f1 NB |
707 | fprintf(stderr, |
708 | Name ": ADD_NEW_DISK for %s " | |
709 | "failed: %s\n", | |
4b1ac34b | 710 | dv->devname, strerror(errno)); |
3da92f27 | 711 | st->ss->free_super(st); |
4b1ac34b NB |
712 | return 1; |
713 | } | |
4b1ac34b NB |
714 | break; |
715 | } | |
716 | if (dv == moved_disk && dnum != insert_point) break; | |
52826846 | 717 | } |
edd8d13c | 718 | if (pass == 1) { |
111d01fc | 719 | st->ss->write_init_super(st); |
edd8d13c NB |
720 | flush_metadata_updates(st); |
721 | } | |
5e52ae9e | 722 | } |
d2ca6449 | 723 | free(infos); |
3da92f27 | 724 | st->ss->free_super(st); |
5e52ae9e | 725 | |
682c7051 | 726 | /* param is not actually used */ |
17f25ca6 NB |
727 | if (level == LEVEL_CONTAINER) |
728 | /* No need to start */ | |
729 | ; | |
730 | else if (runstop == 1 || subdevs >= raiddisks) { | |
598f0d58 NB |
731 | if (st->ss->external) { |
732 | switch(level) { | |
733 | case LEVEL_LINEAR: | |
734 | case LEVEL_MULTIPATH: | |
735 | case 0: | |
f35f2525 | 736 | sysfs_set_str(&info, NULL, "array_state", |
598f0d58 | 737 | "active"); |
a931db9e | 738 | need_mdmon = 0; |
598f0d58 NB |
739 | break; |
740 | default: | |
f35f2525 | 741 | sysfs_set_str(&info, NULL, "array_state", |
598f0d58 NB |
742 | "readonly"); |
743 | break; | |
744 | } | |
f35f2525 | 745 | sysfs_set_safemode(&info, safe_mode_delay); |
598f0d58 NB |
746 | } else { |
747 | mdu_param_t param; | |
748 | if (ioctl(mdfd, RUN_ARRAY, ¶m)) { | |
749 | fprintf(stderr, Name ": RUN_ARRAY failed: %s\n", | |
750 | strerror(errno)); | |
751 | Manage_runstop(mddev, mdfd, -1, 0); | |
752 | return 1; | |
753 | } | |
682c7051 | 754 | } |
dab6685f NB |
755 | if (verbose >= 0) |
756 | fprintf(stderr, Name ": array %s started.\n", mddev); | |
f7e7067b | 757 | if (st->ss->external && st->subarray[0]) { |
8850ee3e N |
758 | if (need_mdmon) |
759 | start_mdmon(st->container_dev); | |
760 | ||
761 | ping_monitor(devnum2devname(st->container_dev)); | |
a931db9e NB |
762 | close(container_fd); |
763 | } | |
682c7051 | 764 | } else { |
b83d95f3 | 765 | fprintf(stderr, Name ": not starting array - not enough devices.\n"); |
682c7051 NB |
766 | } |
767 | return 0; | |
64c4757e | 768 | } |