]>
Commit | Line | Data |
---|---|---|
64c4757e | 1 | /* |
9a9dab36 | 2 | * mdadm - manage Linux "md" devices aka RAID arrays. |
64c4757e | 3 | * |
cd29a5c8 | 4 | * Copyright (C) 2001-2002 Neil Brown <neilb@cse.unsw.edu.au> |
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" | |
64c4757e | 33 | |
82d9eba6 | 34 | int Create(struct supertype *st, char *mddev, int mdfd, |
aa88f531 | 35 | int chunk, int level, int layout, unsigned long size, int raiddisks, int sparedisks, |
947fd4dd | 36 | char *name, |
cd29a5c8 | 37 | int subdevs, mddev_dev_t devlist, |
c82f047c | 38 | int runstop, int verbose, int force, |
dfd4d8ee | 39 | char *bitmap_file, int bitmap_chunk, int write_behind, int delay) |
64c4757e | 40 | { |
682c7051 NB |
41 | /* |
42 | * Create a new raid array. | |
43 | * | |
44 | * First check that necessary details are available | |
45 | * (i.e. level, raid-disks) | |
46 | * | |
47 | * Then check each disk to see what might be on it | |
48 | * and report anything interesting. | |
49 | * | |
50 | * If anything looks odd, and runstop not set, | |
51 | * abort. | |
52 | * | |
53 | * SET_ARRAY_INFO and ADD_NEW_DISK, and | |
82d9eba6 | 54 | * if runstop==run, or raiddisks disks were used, |
682c7051 NB |
55 | * RUN_ARRAY |
56 | */ | |
b5e64645 | 57 | unsigned long long minsize=0, maxsize=0; |
cd29a5c8 NB |
58 | char *mindisc = NULL; |
59 | char *maxdisc = NULL; | |
c913b90e | 60 | int dnum; |
cd29a5c8 | 61 | mddev_dev_t dv; |
682c7051 | 62 | int fail=0, warn=0; |
82b27616 | 63 | struct stat stb; |
82d9eba6 | 64 | int first_missing = subdevs * 2; |
52826846 | 65 | int missing_disks = 0; |
82d9eba6 | 66 | int insert_point = subdevs * 2; /* where to insert a missing drive */ |
4b1ac34b NB |
67 | void *super; |
68 | int pass; | |
82d9eba6 NB |
69 | int vers; |
70 | int rv; | |
c82f047c | 71 | int bitmap_fd; |
682c7051 NB |
72 | |
73 | mdu_array_info_t array; | |
682c7051 | 74 | |
82d9eba6 NB |
75 | vers = md_get_version(mdfd); |
76 | if (vers < 9000) { | |
dd0781e5 | 77 | fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n"); |
52826846 | 78 | return 1; |
682c7051 | 79 | } |
98c6faba | 80 | if (level == UnSet) { |
682c7051 NB |
81 | fprintf(stderr, |
82 | Name ": a RAID level is needed to create an array.\n"); | |
83 | return 1; | |
84 | } | |
85 | if (raiddisks < 1) { | |
86 | fprintf(stderr, | |
b83d95f3 | 87 | Name ": a number of --raid-devices must be given to create an array\n"); |
682c7051 NB |
88 | return 1; |
89 | } | |
98c6faba NB |
90 | if (raiddisks < 4 && level == 6) { |
91 | fprintf(stderr, | |
92 | Name ": at least 4 raid-devices needed for level 6\n"); | |
93 | return 1; | |
94 | } | |
95 | if (raiddisks > 256 && level == 6) { | |
96 | fprintf(stderr, | |
97 | Name ": no more than 256 raid-devices supported for level 6\n"); | |
98 | return 1; | |
99 | } | |
52826846 NB |
100 | if (raiddisks < 2 && level >= 4) { |
101 | fprintf(stderr, | |
98c6faba | 102 | Name ": at least 2 raid-devices needed for level 4 or 5\n"); |
52826846 NB |
103 | return 1; |
104 | } | |
682c7051 | 105 | if (subdevs > raiddisks+sparedisks) { |
b83d95f3 | 106 | fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks); |
52826846 | 107 | return 1; |
682c7051 | 108 | } |
cd29a5c8 | 109 | if (subdevs < raiddisks+sparedisks) { |
52826846 NB |
110 | fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n"); |
111 | return 1; | |
112 | } | |
113 | ||
682c7051 | 114 | /* now set some defaults */ |
98c6faba | 115 | if (layout == UnSet) |
682c7051 NB |
116 | switch(level) { |
117 | default: /* no layout */ | |
118 | layout = 0; | |
119 | break; | |
e5329c37 NB |
120 | case 10: |
121 | layout = 0x102; /* near=2, far=1 */ | |
dab6685f | 122 | if (verbose > 0) |
e5329c37 NB |
123 | fprintf(stderr, |
124 | Name ": layout defaults to n1\n"); | |
125 | break; | |
682c7051 | 126 | case 5: |
98c6faba | 127 | case 6: |
682c7051 | 128 | layout = map_name(r5layout, "default"); |
dab6685f | 129 | if (verbose > 0) |
682c7051 NB |
130 | fprintf(stderr, |
131 | Name ": layout defaults to %s\n", map_num(r5layout, layout)); | |
132 | break; | |
b5e64645 NB |
133 | case LEVEL_FAULTY: |
134 | layout = map_name(faultylayout, "default"); | |
135 | ||
dab6685f | 136 | if (verbose > 0) |
b5e64645 NB |
137 | fprintf(stderr, |
138 | Name ": layout defaults to %s\n", map_num(faultylayout, layout)); | |
139 | break; | |
682c7051 NB |
140 | } |
141 | ||
e5329c37 NB |
142 | if (level == 10) |
143 | /* check layout fits in array*/ | |
144 | if ((layout&255) * ((layout>>8)&255) > raiddisks) { | |
145 | fprintf(stderr, Name ": that layout requires at least %d devices\n", | |
146 | (layout&255) * ((layout>>8)&255)); | |
147 | return 1; | |
148 | } | |
149 | ||
aa88f531 NB |
150 | switch(level) { |
151 | case 4: | |
152 | case 5: | |
e5329c37 | 153 | case 10: |
98c6faba | 154 | case 6: |
aa88f531 NB |
155 | case 0: |
156 | case -1: /* linear */ | |
157 | if (chunk == 0) { | |
158 | chunk = 64; | |
dab6685f | 159 | if (verbose > 0) |
aa88f531 NB |
160 | fprintf(stderr, Name ": chunk size defaults to 64K\n"); |
161 | } | |
162 | break; | |
163 | default: /* raid1, multipath */ | |
164 | if (chunk) { | |
165 | chunk = 0; | |
dab6685f | 166 | if (verbose > 0) |
aa88f531 NB |
167 | fprintf(stderr, Name ": chunk size ignored for this level\n"); |
168 | } | |
169 | break; | |
682c7051 NB |
170 | } |
171 | ||
172 | /* now look at the subdevs */ | |
52826846 NB |
173 | array.active_disks = 0; |
174 | array.working_disks = 0; | |
c913b90e NB |
175 | dnum = 0; |
176 | for (dv=devlist; dv; dv=dv->next, dnum++) { | |
cd29a5c8 | 177 | char *dname = dv->devname; |
b5e64645 NB |
178 | unsigned long dsize; |
179 | unsigned long long ldsize, freesize; | |
52826846 | 180 | int fd; |
cd29a5c8 | 181 | if (strcasecmp(dname, "missing")==0) { |
c913b90e NB |
182 | if (first_missing > dnum) |
183 | first_missing = dnum; | |
52826846 NB |
184 | missing_disks ++; |
185 | continue; | |
186 | } | |
187 | array.working_disks++; | |
c913b90e | 188 | if (dnum < raiddisks) |
52826846 | 189 | array.active_disks++; |
d7eaf49f | 190 | fd = open(dname, O_RDONLY|O_EXCL, 0); |
682c7051 NB |
191 | if (fd <0 ) { |
192 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
193 | dname, strerror(errno)); | |
194 | fail=1; | |
195 | continue; | |
196 | } | |
b5e64645 NB |
197 | #ifdef BLKGETSIZE64 |
198 | if (ioctl(fd, BLKGETSIZE64, &ldsize)==0) | |
199 | ; | |
200 | else | |
201 | #endif | |
682c7051 NB |
202 | if (ioctl(fd, BLKGETSIZE, &dsize)) { |
203 | fprintf(stderr, Name ": Cannot get size of %s: %s\n", | |
204 | dname, strerror(errno)); | |
205 | fail = 1; | |
206 | close(fd); | |
207 | continue; | |
208 | } | |
b5e64645 NB |
209 | else { |
210 | ldsize = dsize; | |
570c0542 | 211 | ldsize <<= 9; |
b5e64645 | 212 | } |
34163fc7 | 213 | freesize = st->ss->avail_size(st, ldsize >> 9, 64*2); |
82d9eba6 | 214 | if (freesize == 0) { |
aa88f531 | 215 | fprintf(stderr, Name ": %s is too small: %luK\n", |
b5e64645 | 216 | dname, (unsigned long)(ldsize>>10)); |
52826846 NB |
217 | fail = 1; |
218 | close(fd); | |
219 | continue; | |
682c7051 | 220 | } |
82d9eba6 NB |
221 | |
222 | freesize /= 2; /* convert to K */ | |
682c7051 NB |
223 | |
224 | if (size && freesize < size) { | |
52826846 | 225 | fprintf(stderr, Name ": %s is smaller that given size." |
b5e64645 | 226 | " %lluK < %luK + superblock\n", dname, freesize, size); |
52826846 NB |
227 | fail = 1; |
228 | close(fd); | |
229 | continue; | |
682c7051 | 230 | } |
cd29a5c8 NB |
231 | if (maxdisc == NULL || (maxdisc && freesize > maxsize)) { |
232 | maxdisc = dname; | |
52826846 | 233 | maxsize = freesize; |
682c7051 | 234 | } |
cd29a5c8 NB |
235 | if (mindisc ==NULL || (mindisc && freesize < minsize)) { |
236 | mindisc = dname; | |
52826846 | 237 | minsize = freesize; |
682c7051 | 238 | } |
dab6685f NB |
239 | if (runstop != 1 || verbose >= 0) { |
240 | warn |= check_ext2(fd, dname); | |
241 | warn |= check_reiser(fd, dname); | |
242 | warn |= check_raid(fd, dname); | |
243 | } | |
682c7051 NB |
244 | close(fd); |
245 | } | |
246 | if (fail) { | |
52826846 NB |
247 | fprintf(stderr, Name ": create aborted\n"); |
248 | return 1; | |
682c7051 NB |
249 | } |
250 | if (size == 0) { | |
cd29a5c8 | 251 | if (mindisc == NULL) { |
52826846 NB |
252 | fprintf(stderr, Name ": no size and no drives given - aborting create.\n"); |
253 | return 1; | |
254 | } | |
c13c45e9 | 255 | if (level > 0 || level == LEVEL_MULTIPATH || level == LEVEL_FAULTY) { |
b5e64645 NB |
256 | /* size is meaningful */ |
257 | if (minsize > 0x100000000ULL) { | |
258 | fprintf(stderr, Name ": devices too large for RAID level %d\n", level); | |
259 | return 1; | |
260 | } | |
261 | size = minsize; | |
dab6685f | 262 | if (verbose > 0) |
b5e64645 NB |
263 | fprintf(stderr, Name ": size set to %luK\n", size); |
264 | } | |
682c7051 | 265 | } |
b5e64645 | 266 | if (level > 0 && ((maxsize-size)*100 > maxsize)) { |
dab6685f NB |
267 | if (runstop != 1 || verbose >= 0) |
268 | fprintf(stderr, Name ": largest drive (%s) exceed size (%luK) by more than 1%%\n", | |
269 | maxdisc, size); | |
52826846 | 270 | warn = 1; |
682c7051 NB |
271 | } |
272 | ||
273 | if (warn) { | |
52826846 NB |
274 | if (runstop!= 1) { |
275 | if (!ask("Continue creating array? ")) { | |
276 | fprintf(stderr, Name ": create aborted.\n"); | |
277 | return 1; | |
278 | } | |
279 | } else { | |
dab6685f | 280 | if (verbose > 0) |
52826846 | 281 | fprintf(stderr, Name ": creation continuing despite oddities due to --run\n"); |
682c7051 | 282 | } |
682c7051 NB |
283 | } |
284 | ||
52826846 NB |
285 | /* If this is raid5, we want to configure the last active slot |
286 | * as missing, so that a reconstruct happens (faster than re-parity) | |
98c6faba | 287 | * FIX: Can we do this for raid6 as well? |
52826846 | 288 | */ |
98c6faba NB |
289 | if (force == 0 && first_missing >= raiddisks) { |
290 | switch ( level ) { | |
291 | case 5: | |
292 | insert_point = raiddisks-1; | |
293 | sparedisks++; | |
294 | array.active_disks--; | |
295 | missing_disks++; | |
296 | break; | |
297 | default: | |
298 | break; | |
299 | } | |
52826846 NB |
300 | } |
301 | ||
682c7051 NB |
302 | /* Ok, lets try some ioctls */ |
303 | ||
304 | array.level = level; | |
305 | array.size = size; | |
682c7051 | 306 | array.raid_disks = raiddisks; |
82b27616 NB |
307 | /* The kernel should *know* what md_minor we are dealing |
308 | * with, but it chooses to trust me instead. Sigh | |
309 | */ | |
682c7051 | 310 | array.md_minor = 0; |
82b27616 | 311 | if (fstat(mdfd, &stb)==0) |
0df46c2a | 312 | array.md_minor = minor(stb.st_rdev); |
682c7051 | 313 | array.not_persistent = 0; |
98c6faba | 314 | /*** FIX: Need to do something about RAID-6 here ***/ |
b5e64645 NB |
315 | if ( ( (level == 5) && |
316 | (insert_point < raiddisks || first_missing < raiddisks) ) | |
317 | || | |
318 | ( level == 6 && missing_disks == 2) | |
319 | ) | |
98c6faba | 320 | array.state = 1; /* clean, but one+ drive will be missing */ |
82b27616 | 321 | else |
52826846 | 322 | array.state = 0; /* not clean, but no errors */ |
82b27616 NB |
323 | |
324 | /* There is lots of redundancy in these disk counts, | |
325 | * raid_disks is the most meaningful value | |
326 | * it describes the geometry of the array | |
327 | * it is constant | |
328 | * nr_disks is total number of used slots. | |
329 | * it should be raid_disks+spare_disks | |
330 | * spare_disks is the number of extra disks present | |
331 | * see above | |
332 | * active_disks is the number of working disks in | |
333 | * active slots. (With raid_disks) | |
334 | * working_disks is the total number of working disks, | |
335 | * including spares | |
336 | * failed_disks is the number of disks marked failed | |
337 | * | |
338 | * Ideally, the kernel would keep these (except raid_disks) | |
339 | * up-to-date as we ADD_NEW_DISK, but it doesn't (yet). | |
340 | * So for now, we assume that all raid and spare | |
341 | * devices will be given. | |
342 | */ | |
52826846 NB |
343 | array.spare_disks=sparedisks; |
344 | array.failed_disks=missing_disks; | |
345 | array.nr_disks = array.working_disks + array.failed_disks; | |
682c7051 NB |
346 | array.layout = layout; |
347 | array.chunk_size = chunk*1024; | |
55935d51 | 348 | |
82d9eba6 | 349 | |
947fd4dd | 350 | if (!st->ss->init_super(st, &super, &array, name)) |
82d9eba6 | 351 | return 1; |
682c7051 | 352 | |
55935d51 NB |
353 | if (bitmap_file && strcmp(bitmap_file, "internal")==0) { |
354 | if ((vers%100) < 2) { | |
355 | fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n"); | |
356 | return 1; | |
357 | } | |
34163fc7 | 358 | if (!st->ss->add_internal_bitmap(st, super, bitmap_chunk, delay, write_behind, |
55935d51 NB |
359 | size ? size : maxsize)) { |
360 | fprintf(stderr, Name ": Given bitmap chunk size not supported.\n"); | |
361 | return 1; | |
362 | } | |
363 | bitmap_file = NULL; | |
364 | } | |
365 | ||
366 | ||
367 | ||
82d9eba6 NB |
368 | if ((vers % 100) >= 1) { /* can use different versions */ |
369 | mdu_array_info_t inf; | |
370 | memset(&inf, 0, sizeof(inf)); | |
371 | inf.major_version = st->ss->major; | |
372 | inf.minor_version = st->minor_version; | |
373 | rv = ioctl(mdfd, SET_ARRAY_INFO, &inf); | |
374 | } else | |
375 | rv = ioctl(mdfd, SET_ARRAY_INFO, NULL); | |
376 | if (rv) { | |
52826846 NB |
377 | fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n", |
378 | mddev, strerror(errno)); | |
379 | return 1; | |
682c7051 | 380 | } |
52826846 | 381 | |
c82f047c NB |
382 | if (bitmap_file) { |
383 | int uuid[4]; | |
55935d51 NB |
384 | |
385 | if (bitmap_chunk == UnSet) | |
386 | bitmap_chunk = DEFAULT_BITMAP_CHUNK; | |
387 | ||
c82f047c | 388 | st->ss->uuid_from_super(uuid, super); |
dfd4d8ee NB |
389 | if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk, |
390 | delay, write_behind, | |
c82f047c NB |
391 | array.size*2ULL /* FIXME wrong for raid10 */)) { |
392 | return 1; | |
393 | } | |
394 | bitmap_fd = open(bitmap_file, O_RDWR); | |
395 | if (bitmap_fd < 0) { | |
396 | fprintf(stderr, Name ": weird: %s cannot be openned\n", | |
397 | bitmap_file); | |
398 | return 1; | |
399 | } | |
400 | if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) { | |
401 | fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n", | |
402 | mddev, strerror(errno)); | |
403 | return 1; | |
404 | } | |
405 | } | |
406 | ||
4b1ac34b NB |
407 | |
408 | ||
409 | for (pass=1; pass <=2 ; pass++) { | |
410 | mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */ | |
411 | ||
412 | for (dnum=0, dv = devlist ; dv ; | |
413 | dv=(dv->next)?(dv->next):moved_disk, dnum++) { | |
414 | int fd; | |
415 | struct stat stb; | |
416 | mdu_disk_info_t disk; | |
417 | ||
418 | disk.number = dnum; | |
419 | if (dnum == insert_point) { | |
420 | moved_disk = dv; | |
52826846 | 421 | } |
4b1ac34b NB |
422 | disk.raid_disk = disk.number; |
423 | if (disk.raid_disk < raiddisks) | |
dfd4d8ee NB |
424 | disk.state = (1<<MD_DISK_ACTIVE) | |
425 | (1<<MD_DISK_SYNC); | |
4b1ac34b NB |
426 | else |
427 | disk.state = 0; | |
dfd4d8ee NB |
428 | if (dv->writemostly) |
429 | disk.state |= (1<<MD_DISK_WRITEMOSTLY); | |
430 | ||
4b1ac34b NB |
431 | if (dnum == insert_point || |
432 | strcasecmp(dv->devname, "missing")==0) { | |
433 | disk.major = 0; | |
434 | disk.minor = 0; | |
dfd4d8ee | 435 | disk.state = (1<<MD_DISK_FAULTY); |
4b1ac34b NB |
436 | } else { |
437 | fd = open(dv->devname, O_RDONLY|O_EXCL, 0); | |
438 | if (fd < 0) { | |
439 | fprintf(stderr, Name ": failed to open %s after earlier success - aborting\n", | |
440 | dv->devname); | |
441 | return 1; | |
442 | } | |
443 | fstat(fd, &stb); | |
444 | disk.major = major(stb.st_rdev); | |
445 | disk.minor = minor(stb.st_rdev); | |
446 | close(fd); | |
447 | } | |
448 | switch(pass){ | |
449 | case 1: | |
82d9eba6 | 450 | st->ss->add_to_super(super, &disk); |
4b1ac34b NB |
451 | break; |
452 | case 2: | |
eaac7dde | 453 | if (disk.state == 1) break; |
34163fc7 | 454 | st->ss->write_init_super(st, super, &disk, dv->devname, 64*2); |
4b1ac34b NB |
455 | |
456 | if (ioctl(mdfd, ADD_NEW_DISK, &disk)) { | |
457 | fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\n", | |
458 | dv->devname, strerror(errno)); | |
459 | free(super); | |
460 | return 1; | |
461 | } | |
462 | ||
463 | break; | |
464 | } | |
465 | if (dv == moved_disk && dnum != insert_point) break; | |
52826846 | 466 | } |
5e52ae9e | 467 | } |
4b1ac34b | 468 | free(super); |
5e52ae9e | 469 | |
682c7051 NB |
470 | /* param is not actually used */ |
471 | if (runstop == 1 || subdevs >= raiddisks) { | |
82b27616 | 472 | mdu_param_t param; |
682c7051 NB |
473 | if (ioctl(mdfd, RUN_ARRAY, ¶m)) { |
474 | fprintf(stderr, Name ": RUN_ARRAY failed: %s\n", | |
475 | strerror(errno)); | |
91f068bf | 476 | Manage_runstop(mddev, mdfd, -1, 0); |
682c7051 NB |
477 | return 1; |
478 | } | |
dab6685f NB |
479 | if (verbose >= 0) |
480 | fprintf(stderr, Name ": array %s started.\n", mddev); | |
682c7051 | 481 | } else { |
b83d95f3 | 482 | fprintf(stderr, Name ": not starting array - not enough devices.\n"); |
682c7051 NB |
483 | } |
484 | return 0; | |
64c4757e | 485 | } |