mdadm-1.6.0
[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 */
aa88f531 55 unsigned 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;
120 case 5:
98c6faba 121 case 6:
682c7051
NB
122 layout = map_name(r5layout, "default");
123 if (verbose)
124 fprintf(stderr,
125 Name ": layout defaults to %s\n", map_num(r5layout, layout));
126 break;
127 }
128
aa88f531
NB
129 switch(level) {
130 case 4:
131 case 5:
98c6faba 132 case 6:
aa88f531
NB
133 case 0:
134 case -1: /* linear */
135 if (chunk == 0) {
136 chunk = 64;
137 if (verbose)
138 fprintf(stderr, Name ": chunk size defaults to 64K\n");
139 }
140 break;
141 default: /* raid1, multipath */
142 if (chunk) {
143 chunk = 0;
144 if (verbose)
145 fprintf(stderr, Name ": chunk size ignored for this level\n");
146 }
147 break;
682c7051
NB
148 }
149
150 /* now look at the subdevs */
52826846
NB
151 array.active_disks = 0;
152 array.working_disks = 0;
c913b90e
NB
153 dnum = 0;
154 for (dv=devlist; dv; dv=dv->next, dnum++) {
cd29a5c8 155 char *dname = dv->devname;
aa88f531 156 unsigned long dsize, freesize;
52826846 157 int fd;
cd29a5c8 158 if (strcasecmp(dname, "missing")==0) {
c913b90e
NB
159 if (first_missing > dnum)
160 first_missing = dnum;
52826846
NB
161 missing_disks ++;
162 continue;
163 }
164 array.working_disks++;
c913b90e 165 if (dnum < raiddisks)
52826846
NB
166 array.active_disks++;
167 fd = open(dname, O_RDONLY, 0);
682c7051
NB
168 if (fd <0 ) {
169 fprintf(stderr, Name ": Cannot open %s: %s\n",
170 dname, strerror(errno));
171 fail=1;
172 continue;
173 }
174 if (ioctl(fd, BLKGETSIZE, &dsize)) {
175 fprintf(stderr, Name ": Cannot get size of %s: %s\n",
176 dname, strerror(errno));
177 fail = 1;
178 close(fd);
179 continue;
180 }
181 if (dsize < MD_RESERVED_SECTORS*2) {
aa88f531 182 fprintf(stderr, Name ": %s is too small: %luK\n",
52826846
NB
183 dname, dsize/2);
184 fail = 1;
185 close(fd);
186 continue;
682c7051
NB
187 }
188 freesize = MD_NEW_SIZE_SECTORS(dsize);
189 freesize /= 2;
190
191 if (size && freesize < size) {
52826846 192 fprintf(stderr, Name ": %s is smaller that given size."
aa88f531 193 " %luK < %luK + superblock\n", dname, freesize, size);
52826846
NB
194 fail = 1;
195 close(fd);
196 continue;
682c7051 197 }
cd29a5c8
NB
198 if (maxdisc == NULL || (maxdisc && freesize > maxsize)) {
199 maxdisc = dname;
52826846 200 maxsize = freesize;
682c7051 201 }
cd29a5c8
NB
202 if (mindisc ==NULL || (mindisc && freesize < minsize)) {
203 mindisc = dname;
52826846 204 minsize = freesize;
682c7051
NB
205 }
206 warn |= check_ext2(fd, dname);
207 warn |= check_reiser(fd, dname);
208 warn |= check_raid(fd, dname);
209 close(fd);
210 }
211 if (fail) {
52826846
NB
212 fprintf(stderr, Name ": create aborted\n");
213 return 1;
682c7051
NB
214 }
215 if (size == 0) {
cd29a5c8 216 if (mindisc == NULL) {
52826846
NB
217 fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
218 return 1;
219 }
220 size = minsize;
221 if (verbose && level>0)
aa88f531 222 fprintf(stderr, Name ": size set to %luK\n", size);
682c7051 223 }
56eb10c0 224 if (level >= 1 && ((maxsize-size)*100 > maxsize)) {
aa88f531 225 fprintf(stderr, Name ": largest drive (%s) exceed size (%luK) by more than 1%%\n",
cd29a5c8 226 maxdisc, size);
52826846 227 warn = 1;
682c7051
NB
228 }
229
230 if (warn) {
52826846
NB
231 if (runstop!= 1) {
232 if (!ask("Continue creating array? ")) {
233 fprintf(stderr, Name ": create aborted.\n");
234 return 1;
235 }
236 } else {
237 if (verbose)
238 fprintf(stderr, Name ": creation continuing despite oddities due to --run\n");
682c7051 239 }
682c7051
NB
240 }
241
52826846
NB
242 /* If this is raid5, we want to configure the last active slot
243 * as missing, so that a reconstruct happens (faster than re-parity)
98c6faba 244 * FIX: Can we do this for raid6 as well?
52826846 245 */
98c6faba
NB
246 if (force == 0 && first_missing >= raiddisks) {
247 switch ( level ) {
248 case 5:
249 insert_point = raiddisks-1;
250 sparedisks++;
251 array.active_disks--;
252 missing_disks++;
253 break;
254 default:
255 break;
256 }
52826846
NB
257 }
258
682c7051
NB
259 /* Ok, lets try some ioctls */
260
261 array.level = level;
262 array.size = size;
682c7051 263 array.raid_disks = raiddisks;
82b27616
NB
264 /* The kernel should *know* what md_minor we are dealing
265 * with, but it chooses to trust me instead. Sigh
266 */
682c7051 267 array.md_minor = 0;
82b27616
NB
268 if (fstat(mdfd, &stb)==0)
269 array.md_minor = MINOR(stb.st_rdev);
682c7051 270 array.not_persistent = 0;
98c6faba
NB
271 /*** FIX: Need to do something about RAID-6 here ***/
272 if ( (level == 5 || level == 6) &&
273 (insert_point < raiddisks || first_missing < raiddisks) )
274 array.state = 1; /* clean, but one+ drive will be missing */
82b27616 275 else
52826846 276 array.state = 0; /* not clean, but no errors */
82b27616
NB
277
278 /* There is lots of redundancy in these disk counts,
279 * raid_disks is the most meaningful value
280 * it describes the geometry of the array
281 * it is constant
282 * nr_disks is total number of used slots.
283 * it should be raid_disks+spare_disks
284 * spare_disks is the number of extra disks present
285 * see above
286 * active_disks is the number of working disks in
287 * active slots. (With raid_disks)
288 * working_disks is the total number of working disks,
289 * including spares
290 * failed_disks is the number of disks marked failed
291 *
292 * Ideally, the kernel would keep these (except raid_disks)
293 * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
294 * So for now, we assume that all raid and spare
295 * devices will be given.
296 */
52826846
NB
297 array.spare_disks=sparedisks;
298 array.failed_disks=missing_disks;
299 array.nr_disks = array.working_disks + array.failed_disks;
682c7051
NB
300 array.layout = layout;
301 array.chunk_size = chunk*1024;
302
303 if (ioctl(mdfd, SET_ARRAY_INFO, &array)) {
52826846
NB
304 fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
305 mddev, strerror(errno));
306 return 1;
682c7051
NB
307 }
308
2d465520 309 for (dnum=0, dv = devlist ; dv ; dv=(dv->next)?(dv->next):moved_disk, dnum++) {
52826846
NB
310 int fd;
311 struct stat stb;
312 mdu_disk_info_t disk;
313
c913b90e 314 disk.number = dnum;
2d465520
NB
315 if (dnum == insert_point) {
316 moved_disk = dv;
317 }
52826846
NB
318 disk.raid_disk = disk.number;
319 if (disk.raid_disk < raiddisks)
320 disk.state = 6; /* active and in sync */
321 else
322 disk.state = 0;
2d465520
NB
323 if (dnum == insert_point ||
324 strcasecmp(dv->devname, "missing")==0) {
52826846
NB
325 disk.major = 0;
326 disk.minor = 0;
327 disk.state = 1; /* faulty */
328 } else {
cd29a5c8 329 fd = open(dv->devname, O_RDONLY, 0);
52826846
NB
330 if (fd < 0) {
331 fprintf(stderr, Name ": failed to open %s after earlier success - aborting\n",
cd29a5c8 332 dv->devname);
52826846
NB
333 return 1;
334 }
335 fstat(fd, &stb);
336 disk.major = MAJOR(stb.st_rdev);
337 disk.minor = MINOR(stb.st_rdev);
338 close(fd);
339 }
340 if (ioctl(mdfd, ADD_NEW_DISK, &disk)) {
341 fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\n",
cd29a5c8 342 dv->devname, strerror(errno));
52826846
NB
343 return 1;
344 }
2d465520 345 if (dv == moved_disk && dnum != insert_point) break;
5e52ae9e
NB
346 }
347
682c7051
NB
348 /* param is not actually used */
349 if (runstop == 1 || subdevs >= raiddisks) {
82b27616 350 mdu_param_t param;
682c7051
NB
351 if (ioctl(mdfd, RUN_ARRAY, &param)) {
352 fprintf(stderr, Name ": RUN_ARRAY failed: %s\n",
353 strerror(errno));
dd0781e5 354 Manage_runstop(mddev, mdfd, -1);
682c7051
NB
355 return 1;
356 }
357 fprintf(stderr, Name ": array %s started.\n", mddev);
358 } else {
b83d95f3 359 fprintf(stderr, Name ": not starting array - not enough devices.\n");
682c7051
NB
360 }
361 return 0;
64c4757e 362}