]>
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 */ |
4b1ac34b | 69 | int pass; |
82d9eba6 NB |
70 | int vers; |
71 | int rv; | |
c82f047c | 72 | int bitmap_fd; |
5f8097be | 73 | int have_container = 0; |
a931db9e NB |
74 | int container_fd; |
75 | int need_mdmon = 0; | |
f9c25f1d | 76 | unsigned long long bitmapsize; |
d03373f1 | 77 | struct mdinfo *sra; |
06c7f68e | 78 | struct mdinfo info; |
b8ac1967 | 79 | int did_default = 0; |
682c7051 | 80 | |
943eafef | 81 | int major_num = BITMAP_MAJOR_HI; |
dcec9ee5 | 82 | |
06c7f68e | 83 | memset(&info, 0, sizeof(info)); |
29e766a5 | 84 | |
82d9eba6 NB |
85 | vers = md_get_version(mdfd); |
86 | if (vers < 9000) { | |
dd0781e5 | 87 | fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n"); |
52826846 | 88 | return 1; |
32e5a4ee DL |
89 | } else { |
90 | mdu_array_info_t inf; | |
91 | memset(&inf, 0, sizeof(inf)); | |
92 | ioctl(mdfd, GET_ARRAY_INFO, &inf); | |
93 | if (inf.working_disks != 0) { | |
94 | fprintf(stderr, Name ": another array by this name" | |
95 | " is already running.\n"); | |
96 | return 1; | |
97 | } | |
682c7051 | 98 | } |
a322f70c | 99 | if (level == UnSet) { |
3cfe6247 | 100 | /* "ddf" and "imsm" metadata only supports one level - should possibly |
a322f70c DW |
101 | * push this into metadata handler?? |
102 | */ | |
3cfe6247 | 103 | if (st && (st->ss == &super_ddf || st->ss == &super_imsm)) |
a322f70c DW |
104 | level = LEVEL_CONTAINER; |
105 | } | |
106 | ||
98c6faba | 107 | if (level == UnSet) { |
682c7051 NB |
108 | fprintf(stderr, |
109 | Name ": a RAID level is needed to create an array.\n"); | |
110 | return 1; | |
111 | } | |
98c6faba NB |
112 | if (raiddisks < 4 && level == 6) { |
113 | fprintf(stderr, | |
114 | Name ": at least 4 raid-devices needed for level 6\n"); | |
115 | return 1; | |
116 | } | |
117 | if (raiddisks > 256 && level == 6) { | |
118 | fprintf(stderr, | |
119 | Name ": no more than 256 raid-devices supported for level 6\n"); | |
120 | return 1; | |
121 | } | |
52826846 NB |
122 | if (raiddisks < 2 && level >= 4) { |
123 | fprintf(stderr, | |
98c6faba | 124 | Name ": at least 2 raid-devices needed for level 4 or 5\n"); |
52826846 NB |
125 | return 1; |
126 | } | |
570510ba NB |
127 | if (level <= 0 && sparedisks) { |
128 | fprintf(stderr, | |
129 | Name ": This level does not support spare devices\n"); | |
130 | return 1; | |
131 | } | |
5f8097be NB |
132 | |
133 | if (subdevs == 1 && strcmp(devlist->devname, "missing") != 0) { | |
134 | /* If given a single device, it might be a container, and we can | |
135 | * extract a device list from there | |
136 | */ | |
137 | mdu_array_info_t inf; | |
138 | int fd; | |
139 | ||
140 | memset(&inf, 0, sizeof(inf)); | |
141 | fd = open(devlist->devname, O_RDONLY, 0); | |
142 | if (fd >= 0 && | |
143 | ioctl(fd, GET_ARRAY_INFO, &inf) == 0 && | |
144 | inf.raid_disks == 0) { | |
145 | /* yep, looks like a container */ | |
146 | if (st) { | |
147 | rv = st->ss->load_super(st, fd, | |
148 | devlist->devname); | |
149 | if (rv == 0) | |
150 | have_container = 1; | |
151 | } else { | |
152 | st = guess_super(fd); | |
153 | if (st && !(rv = st->ss-> | |
154 | load_super(st, fd, | |
155 | devlist->devname))) | |
156 | have_container = 1; | |
157 | else | |
158 | st = NULL; | |
159 | } | |
160 | } | |
161 | if (fd >= 0) | |
162 | close(fd); | |
163 | if (have_container) { | |
164 | subdevs = 0; | |
165 | devlist = NULL; | |
166 | } | |
167 | } | |
682c7051 | 168 | if (subdevs > raiddisks+sparedisks) { |
b83d95f3 | 169 | fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks); |
52826846 | 170 | return 1; |
682c7051 | 171 | } |
5f8097be | 172 | if (!have_container && subdevs < raiddisks+sparedisks) { |
52826846 NB |
173 | fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n"); |
174 | return 1; | |
175 | } | |
5b28bd56 NB |
176 | if (bitmap_file && level <= 0) { |
177 | fprintf(stderr, Name ": bitmaps not meaningful with level %s\n", | |
178 | map_num(pers, level)?:"given"); | |
179 | return 1; | |
180 | } | |
52826846 | 181 | |
682c7051 | 182 | /* now set some defaults */ |
98c6faba | 183 | if (layout == UnSet) |
682c7051 NB |
184 | switch(level) { |
185 | default: /* no layout */ | |
186 | layout = 0; | |
187 | break; | |
e5329c37 NB |
188 | case 10: |
189 | layout = 0x102; /* near=2, far=1 */ | |
dab6685f | 190 | if (verbose > 0) |
e5329c37 NB |
191 | fprintf(stderr, |
192 | Name ": layout defaults to n1\n"); | |
193 | break; | |
682c7051 | 194 | case 5: |
98c6faba | 195 | case 6: |
682c7051 | 196 | layout = map_name(r5layout, "default"); |
dab6685f | 197 | if (verbose > 0) |
682c7051 NB |
198 | fprintf(stderr, |
199 | Name ": layout defaults to %s\n", map_num(r5layout, layout)); | |
200 | break; | |
b5e64645 NB |
201 | case LEVEL_FAULTY: |
202 | layout = map_name(faultylayout, "default"); | |
203 | ||
dab6685f | 204 | if (verbose > 0) |
b5e64645 NB |
205 | fprintf(stderr, |
206 | Name ": layout defaults to %s\n", map_num(faultylayout, layout)); | |
207 | break; | |
682c7051 NB |
208 | } |
209 | ||
e5329c37 NB |
210 | if (level == 10) |
211 | /* check layout fits in array*/ | |
212 | if ((layout&255) * ((layout>>8)&255) > raiddisks) { | |
213 | fprintf(stderr, Name ": that layout requires at least %d devices\n", | |
214 | (layout&255) * ((layout>>8)&255)); | |
215 | return 1; | |
216 | } | |
217 | ||
aa88f531 NB |
218 | switch(level) { |
219 | case 4: | |
220 | case 5: | |
e5329c37 | 221 | case 10: |
98c6faba | 222 | case 6: |
aa88f531 | 223 | case 0: |
570510ba | 224 | case LEVEL_LINEAR: /* linear */ |
aa88f531 NB |
225 | if (chunk == 0) { |
226 | chunk = 64; | |
dab6685f | 227 | if (verbose > 0) |
aa88f531 NB |
228 | fprintf(stderr, Name ": chunk size defaults to 64K\n"); |
229 | } | |
230 | break; | |
570510ba | 231 | case 1: |
08e43379 | 232 | case LEVEL_FAULTY: |
570510ba | 233 | case LEVEL_MULTIPATH: |
17f25ca6 | 234 | case LEVEL_CONTAINER: |
aa88f531 NB |
235 | if (chunk) { |
236 | chunk = 0; | |
dab6685f | 237 | if (verbose > 0) |
aa88f531 NB |
238 | fprintf(stderr, Name ": chunk size ignored for this level\n"); |
239 | } | |
240 | break; | |
570510ba NB |
241 | default: |
242 | fprintf(stderr, Name ": unknown level %d\n", level); | |
243 | return 1; | |
682c7051 NB |
244 | } |
245 | ||
17f25ca6 NB |
246 | if (st && ! st->ss->validate_geometry(st, level, layout, raiddisks, |
247 | chunk, size, NULL, NULL)) | |
248 | return 1; | |
249 | ||
682c7051 | 250 | /* now look at the subdevs */ |
06c7f68e NB |
251 | info.array.active_disks = 0; |
252 | info.array.working_disks = 0; | |
c913b90e NB |
253 | dnum = 0; |
254 | for (dv=devlist; dv; dv=dv->next, dnum++) { | |
cd29a5c8 | 255 | char *dname = dv->devname; |
17f25ca6 | 256 | unsigned long long freesize; |
cd29a5c8 | 257 | if (strcasecmp(dname, "missing")==0) { |
c913b90e NB |
258 | if (first_missing > dnum) |
259 | first_missing = dnum; | |
6fb79233 NB |
260 | if (second_missing > dnum && dnum > first_missing) |
261 | second_missing = dnum; | |
52826846 NB |
262 | missing_disks ++; |
263 | continue; | |
264 | } | |
06c7f68e | 265 | info.array.working_disks++; |
c913b90e | 266 | if (dnum < raiddisks) |
06c7f68e | 267 | info.array.active_disks++; |
058574b1 | 268 | if (st == NULL) { |
8aec876d | 269 | struct createinfo *ci = conf_get_create_info(); |
058574b1 NB |
270 | if (ci) |
271 | st = ci->supertype; | |
272 | } | |
576d6d83 NB |
273 | if (st == NULL) { |
274 | /* Need to choose a default metadata, which is different | |
17f25ca6 | 275 | * depending on geometry of array. |
576d6d83 NB |
276 | */ |
277 | int i; | |
278 | char *name = "default"; | |
17f25ca6 | 279 | for(i=0; !st && superlist[i]; i++) { |
576d6d83 | 280 | st = superlist[i]->match_metadata_desc(name); |
17f25ca6 NB |
281 | if (st && !st->ss->validate_geometry |
282 | (st, level, layout, raiddisks, | |
283 | chunk, size, dname, &freesize)) | |
284 | st = NULL; | |
285 | } | |
576d6d83 NB |
286 | |
287 | if (!st) { | |
17f25ca6 NB |
288 | fprintf(stderr, Name ": device %s not suitable " |
289 | "for any style of array\n", | |
290 | dname); | |
576d6d83 NB |
291 | exit(2); |
292 | } | |
b8ac1967 | 293 | if (st->ss != &super0 || |
576d6d83 | 294 | st->minor_version != 90) |
b8ac1967 | 295 | did_default = 1; |
17f25ca6 NB |
296 | } else { |
297 | if (!st->ss->validate_geometry(st, level, layout, | |
298 | raiddisks, | |
299 | chunk, size, dname, | |
300 | &freesize)) { | |
301 | ||
302 | fprintf(stderr, | |
303 | Name ": %s is not suitable for " | |
304 | "this array.\n", | |
305 | dname); | |
306 | fail = 1; | |
307 | continue; | |
308 | } | |
682c7051 | 309 | } |
82d9eba6 NB |
310 | |
311 | freesize /= 2; /* convert to K */ | |
d2cd3ffc NB |
312 | if (chunk) { |
313 | /* round to chunk size */ | |
314 | freesize = freesize & ~(chunk-1); | |
315 | } | |
682c7051 NB |
316 | |
317 | if (size && freesize < size) { | |
52826846 | 318 | fprintf(stderr, Name ": %s is smaller that given size." |
17f25ca6 NB |
319 | " %lluK < %lluK + metadata\n", |
320 | dname, freesize, size); | |
52826846 | 321 | fail = 1; |
52826846 | 322 | continue; |
682c7051 | 323 | } |
cd29a5c8 NB |
324 | if (maxdisc == NULL || (maxdisc && freesize > maxsize)) { |
325 | maxdisc = dname; | |
52826846 | 326 | maxsize = freesize; |
682c7051 | 327 | } |
cd29a5c8 NB |
328 | if (mindisc ==NULL || (mindisc && freesize < minsize)) { |
329 | mindisc = dname; | |
52826846 | 330 | minsize = freesize; |
682c7051 | 331 | } |
dab6685f | 332 | if (runstop != 1 || verbose >= 0) { |
b6e63da4 NB |
333 | int fd = open(dname, O_RDONLY, 0); |
334 | if (fd <0 ) { | |
335 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
336 | dname, strerror(errno)); | |
337 | fail=1; | |
338 | continue; | |
339 | } | |
dab6685f NB |
340 | warn |= check_ext2(fd, dname); |
341 | warn |= check_reiser(fd, dname); | |
342 | warn |= check_raid(fd, dname); | |
b6e63da4 | 343 | close(fd); |
dab6685f | 344 | } |
682c7051 NB |
345 | } |
346 | if (fail) { | |
52826846 NB |
347 | fprintf(stderr, Name ": create aborted\n"); |
348 | return 1; | |
682c7051 NB |
349 | } |
350 | if (size == 0) { | |
5f8097be | 351 | if (mindisc == NULL && !have_container) { |
52826846 NB |
352 | fprintf(stderr, Name ": no size and no drives given - aborting create.\n"); |
353 | return 1; | |
354 | } | |
5f8097be NB |
355 | if (level > 0 || level == LEVEL_MULTIPATH |
356 | || level == LEVEL_FAULTY | |
b8ac1967 | 357 | || st->ss->external ) { |
b5e64645 | 358 | /* size is meaningful */ |
b8ac1967 NB |
359 | if (!st->ss->validate_geometry(st, level, layout, |
360 | raiddisks, | |
361 | chunk, minsize, | |
362 | NULL, NULL)) { | |
aba69144 | 363 | fprintf(stderr, Name ": devices too large for RAID level %d\n", level); |
b5e64645 NB |
364 | return 1; |
365 | } | |
366 | size = minsize; | |
dab6685f | 367 | if (verbose > 0) |
5dd497ee | 368 | fprintf(stderr, Name ": size set to %lluK\n", size); |
b5e64645 | 369 | } |
682c7051 | 370 | } |
b5e64645 | 371 | if (level > 0 && ((maxsize-size)*100 > maxsize)) { |
dab6685f | 372 | if (runstop != 1 || verbose >= 0) |
5dd497ee | 373 | fprintf(stderr, Name ": largest drive (%s) exceed size (%lluK) by more than 1%%\n", |
dab6685f | 374 | maxdisc, size); |
52826846 | 375 | warn = 1; |
682c7051 NB |
376 | } |
377 | ||
378 | if (warn) { | |
52826846 NB |
379 | if (runstop!= 1) { |
380 | if (!ask("Continue creating array? ")) { | |
381 | fprintf(stderr, Name ": create aborted.\n"); | |
382 | return 1; | |
383 | } | |
384 | } else { | |
dab6685f | 385 | if (verbose > 0) |
52826846 | 386 | fprintf(stderr, Name ": creation continuing despite oddities due to --run\n"); |
682c7051 | 387 | } |
682c7051 NB |
388 | } |
389 | ||
66f8bbbe | 390 | /* If this is raid4/5, we want to configure the last active slot |
52826846 | 391 | * as missing, so that a reconstruct happens (faster than re-parity) |
98c6faba | 392 | * FIX: Can we do this for raid6 as well? |
52826846 | 393 | */ |
47d79ef8 | 394 | if (assume_clean==0 && force == 0 && first_missing >= raiddisks) { |
98c6faba | 395 | switch ( level ) { |
66f8bbbe | 396 | case 4: |
98c6faba NB |
397 | case 5: |
398 | insert_point = raiddisks-1; | |
399 | sparedisks++; | |
06c7f68e | 400 | info.array.active_disks--; |
98c6faba NB |
401 | missing_disks++; |
402 | break; | |
403 | default: | |
404 | break; | |
405 | } | |
52826846 | 406 | } |
6fb79233 NB |
407 | /* For raid6, if creating with 1 missing drive, make a good drive |
408 | * into a spare, else the create will fail | |
409 | */ | |
410 | if (assume_clean == 0 && force == 0 && first_missing < raiddisks && | |
411 | second_missing >= raiddisks && level == 6) { | |
412 | insert_point = raiddisks - 1; | |
413 | if (insert_point == first_missing) | |
414 | insert_point--; | |
415 | sparedisks ++; | |
416 | info.array.active_disks--; | |
417 | missing_disks++; | |
418 | } | |
570510ba | 419 | |
5f8097be | 420 | if (level <= 0 && first_missing < subdevs * 2) { |
570510ba NB |
421 | fprintf(stderr, |
422 | Name ": This level does not support missing devices\n"); | |
423 | return 1; | |
424 | } | |
aba69144 | 425 | |
682c7051 NB |
426 | /* Ok, lets try some ioctls */ |
427 | ||
06c7f68e NB |
428 | info.array.level = level; |
429 | info.array.size = size; | |
430 | info.array.raid_disks = raiddisks; | |
82b27616 NB |
431 | /* The kernel should *know* what md_minor we are dealing |
432 | * with, but it chooses to trust me instead. Sigh | |
433 | */ | |
06c7f68e | 434 | info.array.md_minor = 0; |
82b27616 | 435 | if (fstat(mdfd, &stb)==0) |
06c7f68e NB |
436 | info.array.md_minor = minor(stb.st_rdev); |
437 | info.array.not_persistent = 0; | |
6fb79233 | 438 | |
66f8bbbe | 439 | if ( ( (level == 4 || level == 5) && |
b5e64645 NB |
440 | (insert_point < raiddisks || first_missing < raiddisks) ) |
441 | || | |
6fb79233 NB |
442 | ( level == 6 && (insert_point < raiddisks |
443 | || second_missing < raiddisks)) | |
47d79ef8 | 444 | || |
d024b0a7 NB |
445 | ( level <= 0 ) |
446 | || | |
47d79ef8 | 447 | assume_clean |
b5e64645 | 448 | ) |
06c7f68e | 449 | info.array.state = 1; /* clean, but one+ drive will be missing*/ |
82b27616 | 450 | else |
06c7f68e | 451 | info.array.state = 0; /* not clean, but no errors */ |
82b27616 | 452 | |
f9c25f1d NB |
453 | if (level == 10) { |
454 | /* for raid10, the bitmap size is the capacity of the array, | |
455 | * which is array.size * raid_disks / ncopies; | |
456 | * .. but convert to sectors. | |
457 | */ | |
702b557b | 458 | int ncopies = ((layout>>8) & 255) * (layout & 255); |
5dd497ee NB |
459 | bitmapsize = (unsigned long long)size * raiddisks / ncopies * 2; |
460 | /* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, size, raiddisks, ncopies);*/ | |
f9c25f1d | 461 | } else |
5dd497ee | 462 | bitmapsize = (unsigned long long)size * 2; |
f9c25f1d | 463 | |
82b27616 NB |
464 | /* There is lots of redundancy in these disk counts, |
465 | * raid_disks is the most meaningful value | |
466 | * it describes the geometry of the array | |
467 | * it is constant | |
468 | * nr_disks is total number of used slots. | |
469 | * it should be raid_disks+spare_disks | |
470 | * spare_disks is the number of extra disks present | |
471 | * see above | |
472 | * active_disks is the number of working disks in | |
473 | * active slots. (With raid_disks) | |
474 | * working_disks is the total number of working disks, | |
475 | * including spares | |
476 | * failed_disks is the number of disks marked failed | |
477 | * | |
478 | * Ideally, the kernel would keep these (except raid_disks) | |
479 | * up-to-date as we ADD_NEW_DISK, but it doesn't (yet). | |
480 | * So for now, we assume that all raid and spare | |
481 | * devices will be given. | |
482 | */ | |
06c7f68e NB |
483 | info.array.spare_disks=sparedisks; |
484 | info.array.failed_disks=missing_disks; | |
485 | info.array.nr_disks = info.array.working_disks | |
486 | + info.array.failed_disks; | |
487 | info.array.layout = layout; | |
488 | info.array.chunk_size = chunk*1024; | |
b8ac1967 | 489 | info.array.major_version = 0; /* Flag to say "not zeroing superblock" */ |
82d9eba6 | 490 | |
b3b33eb5 NB |
491 | if (name == NULL || *name == 0) { |
492 | /* base name on mddev */ | |
d1e80164 NB |
493 | /* /dev/md0 -> 0 |
494 | * /dev/md_d0 -> d0 | |
495 | * /dev/md/1 -> 1 | |
496 | * /dev/md/d1 -> d1 | |
497 | * /dev/md/home -> home | |
498 | * /dev/mdhome -> home | |
499 | */ | |
b3b33eb5 NB |
500 | name = strrchr(mddev, '/'); |
501 | if (name) { | |
502 | name++; | |
60248f74 NB |
503 | if (strncmp(name, "md_d", 4)==0 && |
504 | strlen(name) > 4 && | |
505 | isdigit(name[4]) && | |
b3b33eb5 | 506 | (name-mddev) == 5 /* /dev/ */) |
d1e80164 | 507 | name += 3; |
60248f74 NB |
508 | else if (strncmp(name, "md", 2)==0 && |
509 | strlen(name) > 2 && | |
510 | isdigit(name[2]) && | |
511 | (name-mddev) == 5 /* /dev/ */) | |
b3b33eb5 NB |
512 | name += 2; |
513 | } | |
514 | } | |
06c7f68e | 515 | if (!st->ss->init_super(st, &info.array, size, name, homehost, uuid)) |
82d9eba6 | 516 | return 1; |
682c7051 | 517 | |
159c3a1a NB |
518 | st->ss->getinfo_super(st, &info); |
519 | ||
b8ac1967 NB |
520 | if (did_default) |
521 | fprintf(stderr, Name ": Defaulting to version" | |
522 | " %s metadata\n", info.text_version); | |
523 | ||
dcec9ee5 | 524 | if (bitmap_file && vers < 9003) { |
943eafef | 525 | major_num = BITMAP_MAJOR_HOSTENDIAN; |
dcec9ee5 NB |
526 | #ifdef __BIG_ENDIAN |
527 | fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n" | |
528 | " between different architectured. Consider upgrading the Linux kernel.\n"); | |
529 | #endif | |
530 | } | |
531 | ||
55935d51 NB |
532 | if (bitmap_file && strcmp(bitmap_file, "internal")==0) { |
533 | if ((vers%100) < 2) { | |
534 | fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n"); | |
535 | return 1; | |
536 | } | |
3da92f27 | 537 | if (!st->ss->add_internal_bitmap(st, &bitmap_chunk, |
199171a2 | 538 | delay, write_behind, |
943eafef | 539 | bitmapsize, 1, major_num)) { |
55935d51 NB |
540 | fprintf(stderr, Name ": Given bitmap chunk size not supported.\n"); |
541 | return 1; | |
542 | } | |
543 | bitmap_file = NULL; | |
544 | } | |
545 | ||
546 | ||
d03373f1 | 547 | sra = sysfs_read(mdfd, 0, 0); |
55935d51 | 548 | |
d03373f1 NB |
549 | if (st->ss->external) { |
550 | char ver[100]; | |
dd15dc4a NB |
551 | strcat(strcpy(ver, "external:"), |
552 | info.text_version); | |
553 | if (st->ss->external == 2) { | |
2f6079dc | 554 | /* member */ |
dd15dc4a | 555 | |
a931db9e NB |
556 | /* When creating a member, we need to be careful |
557 | * to negotiate with mdmon properly. | |
558 | * If it is already running, we cannot write to | |
559 | * the devices and must ask it to do that part. | |
560 | * If it isn't running, we write to the devices, | |
561 | * and then start it. | |
562 | * We hold an exclusive open on the container | |
563 | * device to make sure mdmon doesn't exit after | |
564 | * we checked that it is running. | |
565 | * | |
566 | * For now, fail if it is already running. | |
567 | */ | |
568 | container_fd = open_dev_excl(st->container_dev); | |
569 | if (container_fd < 0) { | |
570 | fprintf(stderr, Name ": Cannot get exclusive " | |
571 | "open on container - weird.\n"); | |
572 | return 1; | |
573 | } | |
574 | if (mdmon_running(st->container_dev)) { | |
575 | fprintf(stderr, Name ": mdmon already running " | |
576 | "for %s - sorry\n", | |
577 | devnum2devname(st->container_dev)); | |
578 | return 1; | |
579 | } | |
580 | need_mdmon = 1; | |
2f6079dc | 581 | } |
d03373f1 NB |
582 | if ((vers % 100) < 2 || |
583 | sra == NULL || | |
584 | sysfs_set_str(sra, NULL, "metadata_version", | |
585 | ver) < 0) { | |
586 | fprintf(stderr, Name ": This kernel does not " | |
587 | "support external metadata.\n"); | |
588 | return 1; | |
589 | } | |
2503d23b | 590 | rv = sysfs_set_array(sra, &info); |
d03373f1 | 591 | } else if ((vers % 100) >= 1) { /* can use different versions */ |
82d9eba6 NB |
592 | mdu_array_info_t inf; |
593 | memset(&inf, 0, sizeof(inf)); | |
b8ac1967 NB |
594 | inf.major_version = info.array.major_version; |
595 | inf.minor_version = info.array.minor_version; | |
82d9eba6 | 596 | rv = ioctl(mdfd, SET_ARRAY_INFO, &inf); |
aba69144 | 597 | } else |
82d9eba6 NB |
598 | rv = ioctl(mdfd, SET_ARRAY_INFO, NULL); |
599 | if (rv) { | |
52826846 NB |
600 | fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n", |
601 | mddev, strerror(errno)); | |
602 | return 1; | |
682c7051 | 603 | } |
52826846 | 604 | |
c82f047c NB |
605 | if (bitmap_file) { |
606 | int uuid[4]; | |
55935d51 | 607 | |
3da92f27 | 608 | st->ss->uuid_from_super(st, uuid); |
dfd4d8ee | 609 | if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk, |
dcec9ee5 | 610 | delay, write_behind, |
f9c25f1d | 611 | bitmapsize, |
943eafef | 612 | major_num)) { |
c82f047c NB |
613 | return 1; |
614 | } | |
615 | bitmap_fd = open(bitmap_file, O_RDWR); | |
616 | if (bitmap_fd < 0) { | |
617 | fprintf(stderr, Name ": weird: %s cannot be openned\n", | |
dcec9ee5 | 618 | bitmap_file); |
c82f047c NB |
619 | return 1; |
620 | } | |
621 | if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) { | |
622 | fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n", | |
623 | mddev, strerror(errno)); | |
624 | return 1; | |
625 | } | |
626 | } | |
627 | ||
4b1ac34b | 628 | |
4b1ac34b NB |
629 | for (pass=1; pass <=2 ; pass++) { |
630 | mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */ | |
631 | ||
632 | for (dnum=0, dv = devlist ; dv ; | |
633 | dv=(dv->next)?(dv->next):moved_disk, dnum++) { | |
634 | int fd; | |
635 | struct stat stb; | |
4b1ac34b | 636 | |
06c7f68e | 637 | info.disk.number = dnum; |
4b1ac34b NB |
638 | if (dnum == insert_point) { |
639 | moved_disk = dv; | |
52826846 | 640 | } |
06c7f68e NB |
641 | info.disk.raid_disk = info.disk.number; |
642 | if (info.disk.raid_disk < raiddisks) | |
643 | info.disk.state = (1<<MD_DISK_ACTIVE) | | |
dfd4d8ee | 644 | (1<<MD_DISK_SYNC); |
4b1ac34b | 645 | else |
06c7f68e | 646 | info.disk.state = 0; |
dfd4d8ee | 647 | if (dv->writemostly) |
06c7f68e | 648 | info.disk.state |= (1<<MD_DISK_WRITEMOSTLY); |
dfd4d8ee | 649 | |
4b1ac34b | 650 | if (dnum == insert_point || |
111d01fc NB |
651 | strcasecmp(dv->devname, "missing")==0) |
652 | continue; | |
653 | ||
5f8097be NB |
654 | if (st->ss->external == 2) |
655 | fd = open(dv->devname, O_RDWR, 0); | |
656 | else | |
657 | fd = open(dv->devname, O_RDWR|O_EXCL,0); | |
658 | ||
111d01fc NB |
659 | if (fd < 0) { |
660 | fprintf(stderr, Name ": failed to open %s " | |
661 | "after earlier success - aborting\n", | |
662 | dv->devname); | |
663 | return 1; | |
4b1ac34b | 664 | } |
111d01fc NB |
665 | fstat(fd, &stb); |
666 | info.disk.major = major(stb.st_rdev); | |
667 | info.disk.minor = minor(stb.st_rdev); | |
668 | ||
4b1ac34b NB |
669 | switch(pass){ |
670 | case 1: | |
111d01fc NB |
671 | remove_partitions(fd); |
672 | st->ss->add_to_super(st, &info.disk, | |
673 | fd, dv->devname); | |
4b1ac34b NB |
674 | break; |
675 | case 2: | |
2318b9f0 | 676 | close(fd); |
2503d23b NB |
677 | info.component_size = info.array.size * 2; |
678 | info.errors = 0; | |
679 | rv = 0; | |
4b1ac34b | 680 | |
d03373f1 | 681 | if (st->ss->external) { |
598f0d58 | 682 | st->ss->getinfo_super_n(st, &info); |
2318b9f0 | 683 | rv = sysfs_add_disk(sra, &info); |
2503d23b | 684 | } else { |
2503d23b NB |
685 | rv = ioctl(mdfd, ADD_NEW_DISK, |
686 | &info.disk); | |
687 | } | |
688 | if (rv) { | |
d03373f1 NB |
689 | fprintf(stderr, |
690 | Name ": ADD_NEW_DISK for %s " | |
691 | "failed: %s\n", | |
4b1ac34b | 692 | dv->devname, strerror(errno)); |
3da92f27 | 693 | st->ss->free_super(st); |
4b1ac34b NB |
694 | return 1; |
695 | } | |
4b1ac34b NB |
696 | break; |
697 | } | |
698 | if (dv == moved_disk && dnum != insert_point) break; | |
52826846 | 699 | } |
111d01fc NB |
700 | if (pass == 1) |
701 | st->ss->write_init_super(st); | |
5e52ae9e | 702 | } |
3da92f27 | 703 | st->ss->free_super(st); |
5e52ae9e | 704 | |
682c7051 | 705 | /* param is not actually used */ |
17f25ca6 NB |
706 | if (level == LEVEL_CONTAINER) |
707 | /* No need to start */ | |
708 | ; | |
709 | else if (runstop == 1 || subdevs >= raiddisks) { | |
598f0d58 NB |
710 | if (st->ss->external) { |
711 | switch(level) { | |
712 | case LEVEL_LINEAR: | |
713 | case LEVEL_MULTIPATH: | |
714 | case 0: | |
715 | sysfs_set_str(sra, NULL, "array_state", | |
716 | "active"); | |
a931db9e | 717 | need_mdmon = 0; |
598f0d58 NB |
718 | break; |
719 | default: | |
720 | sysfs_set_str(sra, NULL, "array_state", | |
721 | "readonly"); | |
722 | break; | |
723 | } | |
724 | } else { | |
725 | mdu_param_t param; | |
726 | if (ioctl(mdfd, RUN_ARRAY, ¶m)) { | |
727 | fprintf(stderr, Name ": RUN_ARRAY failed: %s\n", | |
728 | strerror(errno)); | |
729 | Manage_runstop(mddev, mdfd, -1, 0); | |
730 | return 1; | |
731 | } | |
682c7051 | 732 | } |
dab6685f NB |
733 | if (verbose >= 0) |
734 | fprintf(stderr, Name ": array %s started.\n", mddev); | |
a931db9e NB |
735 | if (st->ss->external == 2) { |
736 | if (need_mdmon) { | |
737 | int dn = st->container_dev; | |
738 | int i; | |
739 | switch(fork()) { | |
740 | case 0: | |
741 | /* FIXME yuk. CLOSE_EXEC?? */ | |
742 | for (i=3; i < 100; i++) | |
743 | close(i); | |
744 | execl("./mdmon", "mdmon", | |
745 | map_dev(dev2major(dn), | |
746 | dev2minor(dn), | |
747 | 1), NULL); | |
748 | exit(1); | |
749 | case -1: fprintf(stderr, Name ": cannot fork. " | |
750 | "Array remains readonly\n"); | |
751 | return 1; | |
752 | default: ; /* parent - good */ | |
753 | } | |
754 | } else | |
755 | signal_mdmon(st->container_dev); | |
756 | /* FIXME wait for mdmon to set array to read-auto */ | |
757 | sleep(1); | |
758 | close(container_fd); | |
759 | } | |
682c7051 | 760 | } else { |
b83d95f3 | 761 | fprintf(stderr, Name ": not starting array - not enough devices.\n"); |
682c7051 NB |
762 | } |
763 | return 0; | |
64c4757e | 764 | } |