]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
4 | * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de> | |
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 | ||
30 | #include "mdadm.h" | |
31 | #include "md_u.h" | |
32 | #include "md_p.h" | |
33 | #include <ctype.h> | |
34 | ||
35 | int Create(struct supertype *st, char *mddev, int mdfd, | |
36 | int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks, | |
37 | char *name, char *homehost, int *uuid, | |
38 | int subdevs, mddev_dev_t devlist, | |
39 | int runstop, int verbose, int force, int assume_clean, | |
40 | char *bitmap_file, int bitmap_chunk, int write_behind, int delay) | |
41 | { | |
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 | |
55 | * if runstop==run, or raiddisks disks were used, | |
56 | * RUN_ARRAY | |
57 | */ | |
58 | unsigned long long minsize=0, maxsize=0; | |
59 | char *mindisc = NULL; | |
60 | char *maxdisc = NULL; | |
61 | int dnum; | |
62 | mddev_dev_t dv; | |
63 | int fail=0, warn=0; | |
64 | struct stat stb; | |
65 | int first_missing = subdevs * 2; | |
66 | int missing_disks = 0; | |
67 | int insert_point = subdevs * 2; /* where to insert a missing drive */ | |
68 | int pass; | |
69 | int vers; | |
70 | int rv; | |
71 | int bitmap_fd; | |
72 | unsigned long long bitmapsize; | |
73 | struct mdinfo info; | |
74 | ||
75 | int major_num = BITMAP_MAJOR_HI; | |
76 | ||
77 | memset(&info, 0, sizeof(info)); | |
78 | ||
79 | vers = md_get_version(mdfd); | |
80 | if (vers < 9000) { | |
81 | fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n"); | |
82 | return 1; | |
83 | } else { | |
84 | mdu_array_info_t inf; | |
85 | memset(&inf, 0, sizeof(inf)); | |
86 | ioctl(mdfd, GET_ARRAY_INFO, &inf); | |
87 | if (inf.working_disks != 0) { | |
88 | fprintf(stderr, Name ": another array by this name" | |
89 | " is already running.\n"); | |
90 | return 1; | |
91 | } | |
92 | } | |
93 | if (level == UnSet) { | |
94 | fprintf(stderr, | |
95 | Name ": a RAID level is needed to create an array.\n"); | |
96 | return 1; | |
97 | } | |
98 | if (raiddisks < 4 && level == 6) { | |
99 | fprintf(stderr, | |
100 | Name ": at least 4 raid-devices needed for level 6\n"); | |
101 | return 1; | |
102 | } | |
103 | if (raiddisks > 256 && level == 6) { | |
104 | fprintf(stderr, | |
105 | Name ": no more than 256 raid-devices supported for level 6\n"); | |
106 | return 1; | |
107 | } | |
108 | if (raiddisks < 2 && level >= 4) { | |
109 | fprintf(stderr, | |
110 | Name ": at least 2 raid-devices needed for level 4 or 5\n"); | |
111 | return 1; | |
112 | } | |
113 | if (level <= 0 && sparedisks) { | |
114 | fprintf(stderr, | |
115 | Name ": This level does not support spare devices\n"); | |
116 | return 1; | |
117 | } | |
118 | if (subdevs > raiddisks+sparedisks) { | |
119 | fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks); | |
120 | return 1; | |
121 | } | |
122 | if (subdevs < raiddisks+sparedisks) { | |
123 | fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n"); | |
124 | return 1; | |
125 | } | |
126 | if (bitmap_file && level <= 0) { | |
127 | fprintf(stderr, Name ": bitmaps not meaningful with level %s\n", | |
128 | map_num(pers, level)?:"given"); | |
129 | return 1; | |
130 | } | |
131 | ||
132 | /* now set some defaults */ | |
133 | if (layout == UnSet) | |
134 | switch(level) { | |
135 | default: /* no layout */ | |
136 | layout = 0; | |
137 | break; | |
138 | case 10: | |
139 | layout = 0x102; /* near=2, far=1 */ | |
140 | if (verbose > 0) | |
141 | fprintf(stderr, | |
142 | Name ": layout defaults to n1\n"); | |
143 | break; | |
144 | case 5: | |
145 | case 6: | |
146 | layout = map_name(r5layout, "default"); | |
147 | if (verbose > 0) | |
148 | fprintf(stderr, | |
149 | Name ": layout defaults to %s\n", map_num(r5layout, layout)); | |
150 | break; | |
151 | case LEVEL_FAULTY: | |
152 | layout = map_name(faultylayout, "default"); | |
153 | ||
154 | if (verbose > 0) | |
155 | fprintf(stderr, | |
156 | Name ": layout defaults to %s\n", map_num(faultylayout, layout)); | |
157 | break; | |
158 | } | |
159 | ||
160 | if (level == 10) | |
161 | /* check layout fits in array*/ | |
162 | if ((layout&255) * ((layout>>8)&255) > raiddisks) { | |
163 | fprintf(stderr, Name ": that layout requires at least %d devices\n", | |
164 | (layout&255) * ((layout>>8)&255)); | |
165 | return 1; | |
166 | } | |
167 | ||
168 | switch(level) { | |
169 | case 4: | |
170 | case 5: | |
171 | case 10: | |
172 | case 6: | |
173 | case 0: | |
174 | case LEVEL_LINEAR: /* linear */ | |
175 | if (chunk == 0) { | |
176 | chunk = 64; | |
177 | if (verbose > 0) | |
178 | fprintf(stderr, Name ": chunk size defaults to 64K\n"); | |
179 | } | |
180 | break; | |
181 | case 1: | |
182 | case LEVEL_FAULTY: | |
183 | case LEVEL_MULTIPATH: | |
184 | if (chunk) { | |
185 | chunk = 0; | |
186 | if (verbose > 0) | |
187 | fprintf(stderr, Name ": chunk size ignored for this level\n"); | |
188 | } | |
189 | break; | |
190 | default: | |
191 | fprintf(stderr, Name ": unknown level %d\n", level); | |
192 | return 1; | |
193 | } | |
194 | ||
195 | /* now look at the subdevs */ | |
196 | info.array.active_disks = 0; | |
197 | info.array.working_disks = 0; | |
198 | dnum = 0; | |
199 | for (dv=devlist; dv; dv=dv->next, dnum++) { | |
200 | char *dname = dv->devname; | |
201 | unsigned long long ldsize, freesize; | |
202 | int fd; | |
203 | if (strcasecmp(dname, "missing")==0) { | |
204 | if (first_missing > dnum) | |
205 | first_missing = dnum; | |
206 | missing_disks ++; | |
207 | continue; | |
208 | } | |
209 | info.array.working_disks++; | |
210 | if (dnum < raiddisks) | |
211 | info.array.active_disks++; | |
212 | fd = open(dname, O_RDONLY|O_EXCL, 0); | |
213 | if (fd <0 ) { | |
214 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
215 | dname, strerror(errno)); | |
216 | fail=1; | |
217 | continue; | |
218 | } | |
219 | if (!get_dev_size(fd, dname, &ldsize)) { | |
220 | fail = 1; | |
221 | close(fd); | |
222 | continue; | |
223 | } | |
224 | if (st == NULL) { | |
225 | struct createinfo *ci = conf_get_create_info(); | |
226 | if (ci) | |
227 | st = ci->supertype; | |
228 | } | |
229 | if (st == NULL) { | |
230 | /* Need to choose a default metadata, which is different | |
231 | * depending on the sizes of devices | |
232 | */ | |
233 | int i; | |
234 | char *name = "default"; | |
235 | if (level >= 1 && ldsize > (0x7fffffffULL<<10)) | |
236 | name = "default/large"; | |
237 | for(i=0; !st && superlist[i]; i++) | |
238 | st = superlist[i]->match_metadata_desc(name); | |
239 | ||
240 | if (!st) { | |
241 | fprintf(stderr, Name ": internal error - no default metadata style\n"); | |
242 | exit(2); | |
243 | } | |
244 | if (st->ss->major != 0 || | |
245 | st->minor_version != 90) | |
246 | fprintf(stderr, Name ": Defaulting to version" | |
247 | " %d.%d metadata\n", | |
248 | st->ss->major, | |
249 | st->minor_version); | |
250 | } | |
251 | freesize = st->ss->avail_size(st, ldsize >> 9); | |
252 | if (freesize == 0) { | |
253 | fprintf(stderr, Name ": %s is too small: %luK\n", | |
254 | dname, (unsigned long)(ldsize>>10)); | |
255 | fail = 1; | |
256 | close(fd); | |
257 | continue; | |
258 | } | |
259 | ||
260 | freesize /= 2; /* convert to K */ | |
261 | if (chunk) { | |
262 | /* round to chunk size */ | |
263 | freesize = freesize & ~(chunk-1); | |
264 | } | |
265 | ||
266 | if (size && freesize < size) { | |
267 | fprintf(stderr, Name ": %s is smaller that given size." | |
268 | " %lluK < %lluK + superblock\n", dname, freesize, size); | |
269 | fail = 1; | |
270 | close(fd); | |
271 | continue; | |
272 | } | |
273 | if (maxdisc == NULL || (maxdisc && freesize > maxsize)) { | |
274 | maxdisc = dname; | |
275 | maxsize = freesize; | |
276 | } | |
277 | if (mindisc ==NULL || (mindisc && freesize < minsize)) { | |
278 | mindisc = dname; | |
279 | minsize = freesize; | |
280 | } | |
281 | if (runstop != 1 || verbose >= 0) { | |
282 | warn |= check_ext2(fd, dname); | |
283 | warn |= check_reiser(fd, dname); | |
284 | warn |= check_raid(fd, dname); | |
285 | } | |
286 | close(fd); | |
287 | } | |
288 | if (fail) { | |
289 | fprintf(stderr, Name ": create aborted\n"); | |
290 | return 1; | |
291 | } | |
292 | if (size == 0) { | |
293 | if (mindisc == NULL) { | |
294 | fprintf(stderr, Name ": no size and no drives given - aborting create.\n"); | |
295 | return 1; | |
296 | } | |
297 | if (level > 0 || level == LEVEL_MULTIPATH || level == LEVEL_FAULTY) { | |
298 | /* size is meaningful */ | |
299 | if (minsize > 0x100000000ULL && st->ss->major == 0) { | |
300 | fprintf(stderr, Name ": devices too large for RAID level %d\n", level); | |
301 | return 1; | |
302 | } | |
303 | size = minsize; | |
304 | if (verbose > 0) | |
305 | fprintf(stderr, Name ": size set to %lluK\n", size); | |
306 | } | |
307 | } | |
308 | if (level > 0 && ((maxsize-size)*100 > maxsize)) { | |
309 | if (runstop != 1 || verbose >= 0) | |
310 | fprintf(stderr, Name ": largest drive (%s) exceed size (%lluK) by more than 1%%\n", | |
311 | maxdisc, size); | |
312 | warn = 1; | |
313 | } | |
314 | ||
315 | if (warn) { | |
316 | if (runstop!= 1) { | |
317 | if (!ask("Continue creating array? ")) { | |
318 | fprintf(stderr, Name ": create aborted.\n"); | |
319 | return 1; | |
320 | } | |
321 | } else { | |
322 | if (verbose > 0) | |
323 | fprintf(stderr, Name ": creation continuing despite oddities due to --run\n"); | |
324 | } | |
325 | } | |
326 | ||
327 | /* If this is raid4/5, we want to configure the last active slot | |
328 | * as missing, so that a reconstruct happens (faster than re-parity) | |
329 | * FIX: Can we do this for raid6 as well? | |
330 | */ | |
331 | if (assume_clean==0 && force == 0 && first_missing >= raiddisks) { | |
332 | switch ( level ) { | |
333 | case 4: | |
334 | case 5: | |
335 | insert_point = raiddisks-1; | |
336 | sparedisks++; | |
337 | info.array.active_disks--; | |
338 | missing_disks++; | |
339 | break; | |
340 | default: | |
341 | break; | |
342 | } | |
343 | } | |
344 | ||
345 | if (level <= 0 && first_missing != subdevs * 2) { | |
346 | fprintf(stderr, | |
347 | Name ": This level does not support missing devices\n"); | |
348 | return 1; | |
349 | } | |
350 | ||
351 | /* Ok, lets try some ioctls */ | |
352 | ||
353 | info.array.level = level; | |
354 | info.array.size = size; | |
355 | info.array.raid_disks = raiddisks; | |
356 | /* The kernel should *know* what md_minor we are dealing | |
357 | * with, but it chooses to trust me instead. Sigh | |
358 | */ | |
359 | info.array.md_minor = 0; | |
360 | if (fstat(mdfd, &stb)==0) | |
361 | info.array.md_minor = minor(stb.st_rdev); | |
362 | info.array.not_persistent = 0; | |
363 | /*** FIX: Need to do something about RAID-6 here ***/ | |
364 | if ( ( (level == 4 || level == 5) && | |
365 | (insert_point < raiddisks || first_missing < raiddisks) ) | |
366 | || | |
367 | ( level == 6 && missing_disks == 2) | |
368 | || | |
369 | assume_clean | |
370 | ) | |
371 | info.array.state = 1; /* clean, but one+ drive will be missing*/ | |
372 | else | |
373 | info.array.state = 0; /* not clean, but no errors */ | |
374 | ||
375 | if (level == 10) { | |
376 | /* for raid10, the bitmap size is the capacity of the array, | |
377 | * which is array.size * raid_disks / ncopies; | |
378 | * .. but convert to sectors. | |
379 | */ | |
380 | int ncopies = ((layout>>8) & 255) * (layout & 255); | |
381 | bitmapsize = (unsigned long long)size * raiddisks / ncopies * 2; | |
382 | /* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, size, raiddisks, ncopies);*/ | |
383 | } else | |
384 | bitmapsize = (unsigned long long)size * 2; | |
385 | ||
386 | /* There is lots of redundancy in these disk counts, | |
387 | * raid_disks is the most meaningful value | |
388 | * it describes the geometry of the array | |
389 | * it is constant | |
390 | * nr_disks is total number of used slots. | |
391 | * it should be raid_disks+spare_disks | |
392 | * spare_disks is the number of extra disks present | |
393 | * see above | |
394 | * active_disks is the number of working disks in | |
395 | * active slots. (With raid_disks) | |
396 | * working_disks is the total number of working disks, | |
397 | * including spares | |
398 | * failed_disks is the number of disks marked failed | |
399 | * | |
400 | * Ideally, the kernel would keep these (except raid_disks) | |
401 | * up-to-date as we ADD_NEW_DISK, but it doesn't (yet). | |
402 | * So for now, we assume that all raid and spare | |
403 | * devices will be given. | |
404 | */ | |
405 | info.array.spare_disks=sparedisks; | |
406 | info.array.failed_disks=missing_disks; | |
407 | info.array.nr_disks = info.array.working_disks | |
408 | + info.array.failed_disks; | |
409 | info.array.layout = layout; | |
410 | info.array.chunk_size = chunk*1024; | |
411 | info.array.major_version = st->ss->major; | |
412 | ||
413 | if (name == NULL || *name == 0) { | |
414 | /* base name on mddev */ | |
415 | /* /dev/md0 -> 0 | |
416 | * /dev/md_d0 -> d0 | |
417 | * /dev/md/1 -> 1 | |
418 | * /dev/md/d1 -> d1 | |
419 | * /dev/md/home -> home | |
420 | * /dev/mdhome -> home | |
421 | */ | |
422 | name = strrchr(mddev, '/'); | |
423 | if (name) { | |
424 | name++; | |
425 | if (strncmp(name, "md_d", 4)==0 && | |
426 | strlen(name) > 4 && | |
427 | isdigit(name[4]) && | |
428 | (name-mddev) == 5 /* /dev/ */) | |
429 | name += 3; | |
430 | else if (strncmp(name, "md", 2)==0 && | |
431 | strlen(name) > 2 && | |
432 | isdigit(name[2]) && | |
433 | (name-mddev) == 5 /* /dev/ */) | |
434 | name += 2; | |
435 | } | |
436 | } | |
437 | if (!st->ss->init_super(st, &info.array, size, name, homehost, uuid)) | |
438 | return 1; | |
439 | ||
440 | if (bitmap_file && vers < 9003) { | |
441 | major_num = BITMAP_MAJOR_HOSTENDIAN; | |
442 | #ifdef __BIG_ENDIAN | |
443 | fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n" | |
444 | " between different architectured. Consider upgrading the Linux kernel.\n"); | |
445 | #endif | |
446 | } | |
447 | ||
448 | if (bitmap_file && strcmp(bitmap_file, "internal")==0) { | |
449 | if ((vers%100) < 2) { | |
450 | fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n"); | |
451 | return 1; | |
452 | } | |
453 | if (!st->ss->add_internal_bitmap(st, &bitmap_chunk, | |
454 | delay, write_behind, | |
455 | bitmapsize, 1, major_num)) { | |
456 | fprintf(stderr, Name ": Given bitmap chunk size not supported.\n"); | |
457 | return 1; | |
458 | } | |
459 | bitmap_file = NULL; | |
460 | } | |
461 | ||
462 | ||
463 | ||
464 | if ((vers % 100) >= 1) { /* can use different versions */ | |
465 | mdu_array_info_t inf; | |
466 | memset(&inf, 0, sizeof(inf)); | |
467 | inf.major_version = st->ss->major; | |
468 | inf.minor_version = st->minor_version; | |
469 | rv = ioctl(mdfd, SET_ARRAY_INFO, &inf); | |
470 | } else | |
471 | rv = ioctl(mdfd, SET_ARRAY_INFO, NULL); | |
472 | if (rv) { | |
473 | fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n", | |
474 | mddev, strerror(errno)); | |
475 | return 1; | |
476 | } | |
477 | ||
478 | if (bitmap_file) { | |
479 | int uuid[4]; | |
480 | ||
481 | st->ss->uuid_from_super(st, uuid); | |
482 | if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk, | |
483 | delay, write_behind, | |
484 | bitmapsize, | |
485 | major_num)) { | |
486 | return 1; | |
487 | } | |
488 | bitmap_fd = open(bitmap_file, O_RDWR); | |
489 | if (bitmap_fd < 0) { | |
490 | fprintf(stderr, Name ": weird: %s cannot be openned\n", | |
491 | bitmap_file); | |
492 | return 1; | |
493 | } | |
494 | if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) { | |
495 | fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n", | |
496 | mddev, strerror(errno)); | |
497 | return 1; | |
498 | } | |
499 | } | |
500 | ||
501 | ||
502 | ||
503 | for (pass=1; pass <=2 ; pass++) { | |
504 | mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */ | |
505 | ||
506 | for (dnum=0, dv = devlist ; dv ; | |
507 | dv=(dv->next)?(dv->next):moved_disk, dnum++) { | |
508 | int fd; | |
509 | struct stat stb; | |
510 | ||
511 | info.disk.number = dnum; | |
512 | if (dnum == insert_point) { | |
513 | moved_disk = dv; | |
514 | } | |
515 | info.disk.raid_disk = info.disk.number; | |
516 | if (info.disk.raid_disk < raiddisks) | |
517 | info.disk.state = (1<<MD_DISK_ACTIVE) | | |
518 | (1<<MD_DISK_SYNC); | |
519 | else | |
520 | info.disk.state = 0; | |
521 | if (dv->writemostly) | |
522 | info.disk.state |= (1<<MD_DISK_WRITEMOSTLY); | |
523 | ||
524 | if (dnum == insert_point || | |
525 | strcasecmp(dv->devname, "missing")==0) { | |
526 | info.disk.major = 0; | |
527 | info.disk.minor = 0; | |
528 | info.disk.state = (1<<MD_DISK_FAULTY); | |
529 | } else { | |
530 | fd = open(dv->devname, O_RDONLY|O_EXCL, 0); | |
531 | if (fd < 0) { | |
532 | fprintf(stderr, Name ": failed to open %s after earlier success - aborting\n", | |
533 | dv->devname); | |
534 | return 1; | |
535 | } | |
536 | fstat(fd, &stb); | |
537 | info.disk.major = major(stb.st_rdev); | |
538 | info.disk.minor = minor(stb.st_rdev); | |
539 | remove_partitions(fd); | |
540 | close(fd); | |
541 | } | |
542 | switch(pass){ | |
543 | case 1: | |
544 | st->ss->add_to_super(st, &info.disk); | |
545 | break; | |
546 | case 2: | |
547 | if (info.disk.state == 1) break; | |
548 | Kill(dv->devname, 0, 1); /* Just be sure it is clean */ | |
549 | Kill(dv->devname, 0, 1); /* and again, there could be two superblocks */ | |
550 | st->ss->write_init_super(st, &info.disk, | |
551 | dv->devname); | |
552 | ||
553 | if (ioctl(mdfd, ADD_NEW_DISK, &info.disk)) { | |
554 | fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\n", | |
555 | dv->devname, strerror(errno)); | |
556 | st->ss->free_super(st); | |
557 | return 1; | |
558 | } | |
559 | ||
560 | break; | |
561 | } | |
562 | if (dv == moved_disk && dnum != insert_point) break; | |
563 | } | |
564 | } | |
565 | st->ss->free_super(st); | |
566 | ||
567 | /* param is not actually used */ | |
568 | if (runstop == 1 || subdevs >= raiddisks) { | |
569 | mdu_param_t param; | |
570 | if (ioctl(mdfd, RUN_ARRAY, ¶m)) { | |
571 | fprintf(stderr, Name ": RUN_ARRAY failed: %s\n", | |
572 | strerror(errno)); | |
573 | Manage_runstop(mddev, mdfd, -1, 0); | |
574 | return 1; | |
575 | } | |
576 | if (verbose >= 0) | |
577 | fprintf(stderr, Name ": array %s started.\n", mddev); | |
578 | } else { | |
579 | fprintf(stderr, Name ": not starting array - not enough devices.\n"); | |
580 | } | |
581 | return 0; | |
582 | } |