Use O_EXCL when opening component devices to be assembled into an array
[thirdparty/mdadm.git] / Create.c
CommitLineData
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
NB
33
34int Create(char *mddev, int mdfd,
aa88f531 35 int chunk, int level, int layout, unsigned long size, int raiddisks, int sparedisks,
cd29a5c8 36 int subdevs, mddev_dev_t devlist,
52826846 37 int runstop, int verbose, int force)
64c4757e 38{
682c7051
NB
39 /*
40 * Create a new raid array.
41 *
42 * First check that necessary details are available
43 * (i.e. level, raid-disks)
44 *
45 * Then check each disk to see what might be on it
46 * and report anything interesting.
47 *
48 * If anything looks odd, and runstop not set,
49 * abort.
50 *
51 * SET_ARRAY_INFO and ADD_NEW_DISK, and
52 * if runstop==run, or raiddisks diskswere used,
53 * RUN_ARRAY
54 */
b5e64645 55 unsigned long long minsize=0, maxsize=0;
cd29a5c8
NB
56 char *mindisc = NULL;
57 char *maxdisc = NULL;
c913b90e 58 int dnum;
cd29a5c8 59 mddev_dev_t dv;
682c7051 60 int fail=0, warn=0;
82b27616 61 struct stat stb;
52826846
NB
62 int first_missing = MD_SB_DISKS*2;
63 int missing_disks = 0;
64 int insert_point = MD_SB_DISKS*2; /* where to insert a missing drive */
2d465520 65 mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */
682c7051
NB
66
67 mdu_array_info_t array;
682c7051
NB
68
69
70 if (md_get_version(mdfd) < 9000) {
dd0781e5 71 fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n");
52826846 72 return 1;
682c7051 73 }
98c6faba 74 if (level == UnSet) {
682c7051
NB
75 fprintf(stderr,
76 Name ": a RAID level is needed to create an array.\n");
77 return 1;
78 }
79 if (raiddisks < 1) {
80 fprintf(stderr,
b83d95f3 81 Name ": a number of --raid-devices must be given to create an array\n");
682c7051
NB
82 return 1;
83 }
98c6faba
NB
84 if (raiddisks < 4 && level == 6) {
85 fprintf(stderr,
86 Name ": at least 4 raid-devices needed for level 6\n");
87 return 1;
88 }
89 if (raiddisks > 256 && level == 6) {
90 fprintf(stderr,
91 Name ": no more than 256 raid-devices supported for level 6\n");
92 return 1;
93 }
52826846
NB
94 if (raiddisks < 2 && level >= 4) {
95 fprintf(stderr,
98c6faba 96 Name ": at least 2 raid-devices needed for level 4 or 5\n");
52826846
NB
97 return 1;
98 }
682c7051
NB
99 if (raiddisks+sparedisks > MD_SB_DISKS) {
100 fprintf(stderr,
b83d95f3 101 Name ": too many devices requested: %d+%d > %d\n",
682c7051
NB
102 raiddisks, sparedisks, MD_SB_DISKS);
103 return 1;
104 }
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 */
122 if (verbose)
123 fprintf(stderr,
124 Name ": layout defaults to n1\n");
125 break;
682c7051 126 case 5:
98c6faba 127 case 6:
682c7051
NB
128 layout = map_name(r5layout, "default");
129 if (verbose)
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
136 if (verbose)
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;
159 if (verbose)
160 fprintf(stderr, Name ": chunk size defaults to 64K\n");
161 }
162 break;
163 default: /* raid1, multipath */
164 if (chunk) {
165 chunk = 0;
166 if (verbose)
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;
211 dsize <<= 9;
212 }
213 if (ldsize < MD_RESERVED_SECTORS*2LL*512LL) {
aa88f531 214 fprintf(stderr, Name ": %s is too small: %luK\n",
b5e64645 215 dname, (unsigned long)(ldsize>>10));
52826846
NB
216 fail = 1;
217 close(fd);
218 continue;
682c7051 219 }
b5e64645 220 freesize = MD_NEW_SIZE_SECTORS((ldsize>>9));
682c7051
NB
221 freesize /= 2;
222
223 if (size && freesize < size) {
52826846 224 fprintf(stderr, Name ": %s is smaller that given size."
b5e64645 225 " %lluK < %luK + superblock\n", dname, freesize, size);
52826846
NB
226 fail = 1;
227 close(fd);
228 continue;
682c7051 229 }
cd29a5c8
NB
230 if (maxdisc == NULL || (maxdisc && freesize > maxsize)) {
231 maxdisc = dname;
52826846 232 maxsize = freesize;
682c7051 233 }
cd29a5c8
NB
234 if (mindisc ==NULL || (mindisc && freesize < minsize)) {
235 mindisc = dname;
52826846 236 minsize = freesize;
682c7051
NB
237 }
238 warn |= check_ext2(fd, dname);
239 warn |= check_reiser(fd, dname);
240 warn |= check_raid(fd, dname);
241 close(fd);
242 }
243 if (fail) {
52826846
NB
244 fprintf(stderr, Name ": create aborted\n");
245 return 1;
682c7051
NB
246 }
247 if (size == 0) {
cd29a5c8 248 if (mindisc == NULL) {
52826846
NB
249 fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
250 return 1;
251 }
b5e64645
NB
252 if (level > 0) {
253 /* size is meaningful */
254 if (minsize > 0x100000000ULL) {
255 fprintf(stderr, Name ": devices too large for RAID level %d\n", level);
256 return 1;
257 }
258 size = minsize;
259 if (verbose)
260 fprintf(stderr, Name ": size set to %luK\n", size);
261 }
682c7051 262 }
b5e64645 263 if (level > 0 && ((maxsize-size)*100 > maxsize)) {
aa88f531 264 fprintf(stderr, Name ": largest drive (%s) exceed size (%luK) by more than 1%%\n",
cd29a5c8 265 maxdisc, size);
52826846 266 warn = 1;
682c7051
NB
267 }
268
269 if (warn) {
52826846
NB
270 if (runstop!= 1) {
271 if (!ask("Continue creating array? ")) {
272 fprintf(stderr, Name ": create aborted.\n");
273 return 1;
274 }
275 } else {
276 if (verbose)
277 fprintf(stderr, Name ": creation continuing despite oddities due to --run\n");
682c7051 278 }
682c7051
NB
279 }
280
52826846
NB
281 /* If this is raid5, we want to configure the last active slot
282 * as missing, so that a reconstruct happens (faster than re-parity)
98c6faba 283 * FIX: Can we do this for raid6 as well?
52826846 284 */
98c6faba
NB
285 if (force == 0 && first_missing >= raiddisks) {
286 switch ( level ) {
287 case 5:
288 insert_point = raiddisks-1;
289 sparedisks++;
290 array.active_disks--;
291 missing_disks++;
292 break;
293 default:
294 break;
295 }
52826846
NB
296 }
297
682c7051
NB
298 /* Ok, lets try some ioctls */
299
300 array.level = level;
301 array.size = size;
682c7051 302 array.raid_disks = raiddisks;
82b27616
NB
303 /* The kernel should *know* what md_minor we are dealing
304 * with, but it chooses to trust me instead. Sigh
305 */
682c7051 306 array.md_minor = 0;
82b27616 307 if (fstat(mdfd, &stb)==0)
0df46c2a 308 array.md_minor = minor(stb.st_rdev);
682c7051 309 array.not_persistent = 0;
98c6faba 310 /*** FIX: Need to do something about RAID-6 here ***/
b5e64645
NB
311 if ( ( (level == 5) &&
312 (insert_point < raiddisks || first_missing < raiddisks) )
313 ||
314 ( level == 6 && missing_disks == 2)
315 )
98c6faba 316 array.state = 1; /* clean, but one+ drive will be missing */
82b27616 317 else
52826846 318 array.state = 0; /* not clean, but no errors */
82b27616
NB
319
320 /* There is lots of redundancy in these disk counts,
321 * raid_disks is the most meaningful value
322 * it describes the geometry of the array
323 * it is constant
324 * nr_disks is total number of used slots.
325 * it should be raid_disks+spare_disks
326 * spare_disks is the number of extra disks present
327 * see above
328 * active_disks is the number of working disks in
329 * active slots. (With raid_disks)
330 * working_disks is the total number of working disks,
331 * including spares
332 * failed_disks is the number of disks marked failed
333 *
334 * Ideally, the kernel would keep these (except raid_disks)
335 * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
336 * So for now, we assume that all raid and spare
337 * devices will be given.
338 */
52826846
NB
339 array.spare_disks=sparedisks;
340 array.failed_disks=missing_disks;
341 array.nr_disks = array.working_disks + array.failed_disks;
682c7051
NB
342 array.layout = layout;
343 array.chunk_size = chunk*1024;
344
345 if (ioctl(mdfd, SET_ARRAY_INFO, &array)) {
52826846
NB
346 fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
347 mddev, strerror(errno));
348 return 1;
682c7051
NB
349 }
350
2d465520 351 for (dnum=0, dv = devlist ; dv ; dv=(dv->next)?(dv->next):moved_disk, dnum++) {
52826846
NB
352 int fd;
353 struct stat stb;
354 mdu_disk_info_t disk;
355
c913b90e 356 disk.number = dnum;
2d465520
NB
357 if (dnum == insert_point) {
358 moved_disk = dv;
359 }
52826846
NB
360 disk.raid_disk = disk.number;
361 if (disk.raid_disk < raiddisks)
362 disk.state = 6; /* active and in sync */
363 else
364 disk.state = 0;
2d465520
NB
365 if (dnum == insert_point ||
366 strcasecmp(dv->devname, "missing")==0) {
52826846
NB
367 disk.major = 0;
368 disk.minor = 0;
369 disk.state = 1; /* faulty */
370 } else {
d7eaf49f 371 fd = open(dv->devname, O_RDONLY|O_EXCL, 0);
52826846
NB
372 if (fd < 0) {
373 fprintf(stderr, Name ": failed to open %s after earlier success - aborting\n",
cd29a5c8 374 dv->devname);
52826846
NB
375 return 1;
376 }
377 fstat(fd, &stb);
0df46c2a
NB
378 disk.major = major(stb.st_rdev);
379 disk.minor = minor(stb.st_rdev);
52826846
NB
380 close(fd);
381 }
382 if (ioctl(mdfd, ADD_NEW_DISK, &disk)) {
383 fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\n",
cd29a5c8 384 dv->devname, strerror(errno));
52826846
NB
385 return 1;
386 }
2d465520 387 if (dv == moved_disk && dnum != insert_point) break;
5e52ae9e
NB
388 }
389
682c7051
NB
390 /* param is not actually used */
391 if (runstop == 1 || subdevs >= raiddisks) {
82b27616 392 mdu_param_t param;
682c7051
NB
393 if (ioctl(mdfd, RUN_ARRAY, &param)) {
394 fprintf(stderr, Name ": RUN_ARRAY failed: %s\n",
395 strerror(errno));
dd0781e5 396 Manage_runstop(mddev, mdfd, -1);
682c7051
NB
397 return 1;
398 }
399 fprintf(stderr, Name ": array %s started.\n", mddev);
400 } else {
b83d95f3 401 fprintf(stderr, Name ": not starting array - not enough devices.\n");
682c7051
NB
402 }
403 return 0;
64c4757e 404}