]>
Commit | Line | Data |
---|---|---|
8382f19b NB |
1 | /* |
2 | * Incremental.c - support --incremental. Part of: | |
3 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
4 | * | |
ca3b6696 | 5 | * Copyright (C) 2006-2012 Neil Brown <neilb@suse.de> |
8382f19b NB |
6 | * |
7 | * | |
8 | * This program is free software; you can redistribute it and/or modify | |
9 | * it under the terms of the GNU General Public License as published by | |
10 | * the Free Software Foundation; either version 2 of the License, or | |
11 | * (at your option) any later version. | |
12 | * | |
13 | * This program is distributed in the hope that it will be useful, | |
14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | * GNU General Public License for more details. | |
17 | * | |
18 | * You should have received a copy of the GNU General Public License | |
19 | * along with this program; if not, write to the Free Software | |
20 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
21 | * | |
22 | * Author: Neil Brown | |
23 | * Email: <neilb@suse.de> | |
24 | * Paper: Neil Brown | |
25 | * Novell Inc | |
26 | * GPO Box Q1283 | |
27 | * QVB Post Office, NSW 1230 | |
28 | * Australia | |
29 | */ | |
30 | ||
31 | #include "mdadm.h" | |
6b63c1a4 | 32 | #include <sys/wait.h> |
61018da0 N |
33 | #include <dirent.h> |
34 | #include <ctype.h> | |
8382f19b | 35 | |
2bf740b3 N |
36 | static int count_active(struct supertype *st, struct mdinfo *sra, |
37 | int mdfd, char **availp, | |
8382f19b | 38 | struct mdinfo *info); |
7e0f6979 | 39 | static void find_reject(int mdfd, struct supertype *st, struct mdinfo *sra, |
8382f19b NB |
40 | int number, __u64 events, int verbose, |
41 | char *array_name); | |
f08605b3 | 42 | static int try_spare(char *devname, int *dfdp, struct dev_policy *pol, |
d2db3045 | 43 | struct map_ent *target, |
f08605b3 | 44 | struct supertype *st, int verbose); |
8382f19b | 45 | |
7d91c3f5 | 46 | static int Incremental_container(struct supertype *st, char *devname, |
11b6d91d N |
47 | struct context *c); |
48 | ||
49 | int Incremental(char *devname, struct context *c, | |
50 | struct supertype *st) | |
8382f19b NB |
51 | { |
52 | /* Add this device to an array, creating the array if necessary | |
598f0d58 | 53 | * and starting the array if sensible or - if runstop>0 - if possible. |
8382f19b NB |
54 | * |
55 | * This has several steps: | |
56 | * | |
57 | * 1/ Check if device is permitted by mdadm.conf, reject if not. | |
58 | * 2/ Find metadata, reject if none appropriate (check | |
59 | * version/name from args) | |
60 | * 3/ Check if there is a match in mdadm.conf | |
215bb3f7 N |
61 | * 3a/ if not, check for homehost match. If no match, assemble as |
62 | * a 'foreign' array. | |
8382f19b NB |
63 | * 4/ Determine device number. |
64 | * - If in mdadm.conf with std name, use that | |
65 | * - UUID in /var/run/mdadm.map use that | |
66 | * - If name is suggestive, use that. unless in use with different uuid. | |
67 | * - Choose a free, high number. | |
68 | * - Use a partitioned device unless strong suggestion not to. | |
69 | * e.g. auto=md | |
352452c3 | 70 | * Don't choose partitioned for containers. |
8382f19b NB |
71 | * 5/ Find out if array already exists |
72 | * 5a/ if it does not | |
73 | * - choose a name, from mdadm.conf or 'name' field in array. | |
74 | * - create the array | |
75 | * - add the device | |
76 | * 5b/ if it does | |
77 | * - check one drive in array to make sure metadata is a reasonably | |
78 | * close match. Reject if not (e.g. different type) | |
79 | * - add the device | |
80 | * 6/ Make sure /var/run/mdadm.map contains this array. | |
81 | * 7/ Is there enough devices to possibly start the array? | |
352452c3 | 82 | * For a container, this means running Incremental_container. |
8382f19b NB |
83 | * 7a/ if not, finish with success. |
84 | * 7b/ if yes, | |
85 | * - read all metadata and arrange devices like -A does | |
86 | * - if number of OK devices match expected, or -R and there are enough, | |
87 | * start the array (auto-readonly). | |
88 | */ | |
89 | struct stat stb; | |
7e83544b | 90 | struct mdinfo info, dinfo; |
de6ae750 | 91 | struct mdinfo *sra = NULL, *d; |
fa56eddb | 92 | struct mddev_ident *match; |
8382f19b | 93 | char chosen_name[1024]; |
15d4a7e4 | 94 | int rv = 1; |
8382f19b | 95 | struct map_ent *mp, *map = NULL; |
15d4a7e4 | 96 | int dfd = -1, mdfd = -1; |
de6ae750 | 97 | char *avail = NULL; |
8382f19b | 98 | int active_disks; |
47c74f3f | 99 | int trustworthy; |
d7288ddc | 100 | char *name_to_use; |
ad5bc697 | 101 | mdu_array_info_t ainf; |
7e83544b | 102 | struct dev_policy *policy = NULL; |
d2db3045 N |
103 | struct map_ent target_array; |
104 | int have_target; | |
8382f19b NB |
105 | |
106 | struct createinfo *ci = conf_get_create_info(); | |
107 | ||
47c74f3f | 108 | if (stat(devname, &stb) < 0) { |
11b6d91d | 109 | if (c->verbose >= 0) |
e7b84f9d | 110 | pr_err("stat failed for %s: %s.\n", |
47c74f3f N |
111 | devname, strerror(errno)); |
112 | return rv; | |
113 | } | |
114 | if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
11b6d91d | 115 | if (c->verbose >= 0) |
e7b84f9d | 116 | pr_err("%s is not a block device.\n", |
47c74f3f N |
117 | devname); |
118 | return rv; | |
119 | } | |
0431869c | 120 | dfd = dev_open(devname, O_RDONLY); |
47c74f3f | 121 | if (dfd < 0) { |
11b6d91d | 122 | if (c->verbose >= 0) |
e7b84f9d | 123 | pr_err("cannot open %s: %s.\n", |
47c74f3f N |
124 | devname, strerror(errno)); |
125 | return rv; | |
126 | } | |
127 | /* If the device is a container, we do something very different */ | |
3a371610 | 128 | if (must_be_container(dfd)) { |
47c74f3f N |
129 | if (!st) |
130 | st = super_by_fd(dfd, NULL); | |
cb8f6859 N |
131 | if (st) |
132 | st->ignore_hw_compat = 1; | |
a92b2112 | 133 | if (st && st->ss->load_container) |
47c74f3f N |
134 | rv = st->ss->load_container(st, dfd, NULL); |
135 | ||
136 | close(dfd); | |
5fc8cff3 JS |
137 | if (!rv && st->ss->container_content) { |
138 | if (map_lock(&map)) | |
e7b84f9d N |
139 | pr_err("failed to get " |
140 | "exclusive lock on mapfile\n"); | |
11b6d91d | 141 | rv = Incremental_container(st, devname, c); |
5fc8cff3 JS |
142 | map_unlock(&map); |
143 | return rv; | |
144 | } | |
47c74f3f | 145 | |
e7b84f9d | 146 | pr_err("%s is not part of an md array.\n", |
47c74f3f N |
147 | devname); |
148 | return rv; | |
149 | } | |
8382f19b | 150 | |
352452c3 | 151 | /* 1/ Check if device is permitted by mdadm.conf */ |
8382f19b NB |
152 | |
153 | if (!conf_test_dev(devname)) { | |
11b6d91d | 154 | if (c->verbose >= 0) |
e7b84f9d N |
155 | pr_err("%s not permitted by mdadm.conf.\n", |
156 | devname); | |
15d4a7e4 | 157 | goto out; |
8382f19b NB |
158 | } |
159 | ||
160 | /* 2/ Find metadata, reject if none appropriate (check | |
161 | * version/name from args) */ | |
162 | ||
8382f19b | 163 | if (fstat(dfd, &stb) < 0) { |
11b6d91d | 164 | if (c->verbose >= 0) |
e7b84f9d | 165 | pr_err("fstat failed for %s: %s.\n", |
8382f19b | 166 | devname, strerror(errno)); |
15d4a7e4 | 167 | goto out; |
8382f19b NB |
168 | } |
169 | if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
11b6d91d | 170 | if (c->verbose >= 0) |
e7b84f9d | 171 | pr_err("%s is not a block device.\n", |
8382f19b | 172 | devname); |
15d4a7e4 | 173 | goto out; |
8382f19b NB |
174 | } |
175 | ||
7e83544b N |
176 | dinfo.disk.major = major(stb.st_rdev); |
177 | dinfo.disk.minor = minor(stb.st_rdev); | |
178 | ||
179 | policy = disk_policy(&dinfo); | |
d2db3045 | 180 | have_target = policy_check_path(&dinfo, &target_array); |
7e83544b | 181 | |
8382f19b | 182 | if (st == NULL && (st = guess_super(dfd)) == NULL) { |
11b6d91d | 183 | if (c->verbose >= 0) |
e7b84f9d N |
184 | pr_err("no recognisable superblock on %s.\n", |
185 | devname); | |
d2db3045 N |
186 | rv = try_spare(devname, &dfd, policy, |
187 | have_target ? &target_array : NULL, | |
11b6d91d | 188 | st, c->verbose); |
15d4a7e4 | 189 | goto out; |
8382f19b | 190 | } |
cb8f6859 | 191 | st->ignore_hw_compat = 1; |
0f22b998 N |
192 | if (st->ss->compare_super == NULL || |
193 | st->ss->load_super(st, dfd, NULL)) { | |
11b6d91d | 194 | if (c->verbose >= 0) |
e7b84f9d | 195 | pr_err("no RAID superblock on %s.\n", |
8382f19b | 196 | devname); |
d2db3045 N |
197 | rv = try_spare(devname, &dfd, policy, |
198 | have_target ? &target_array : NULL, | |
11b6d91d | 199 | st, c->verbose); |
0f22b998 | 200 | free(st); |
15d4a7e4 | 201 | goto out; |
8382f19b | 202 | } |
15d4a7e4 | 203 | close (dfd); dfd = -1; |
8382f19b | 204 | |
a5d85af7 | 205 | st->ss->getinfo_super(st, &info, NULL); |
8382f19b | 206 | |
3a97f210 | 207 | /* 3/ Check if there is a match in mdadm.conf */ |
11b6d91d | 208 | match = conf_match(st, &info, devname, c->verbose, &rv); |
3a97f210 N |
209 | if (!match && rv == 2) |
210 | goto out; | |
8382f19b | 211 | |
112cace6 N |
212 | if (match && match->devname |
213 | && strcasecmp(match->devname, "<ignore>") == 0) { | |
11b6d91d | 214 | if (c->verbose >= 0) |
e7b84f9d | 215 | pr_err("array containing %s is explicitly" |
112cace6 N |
216 | " ignored by mdadm.conf\n", |
217 | devname); | |
15d4a7e4 | 218 | goto out; |
112cace6 N |
219 | } |
220 | ||
7b403fef N |
221 | /* 3a/ if not, check for homehost match. If no match, continue |
222 | * but don't trust the 'name' in the array. Thus a 'random' minor | |
223 | * number will be assigned, and the device name will be based | |
224 | * on that. */ | |
215bb3f7 N |
225 | if (match) |
226 | trustworthy = LOCAL; | |
11b6d91d | 227 | else if (st->ss->match_home(st, c->homehost) == 1) |
d1d3482b N |
228 | trustworthy = LOCAL; |
229 | else if (st->ss->match_home(st, "any") == 1) | |
230 | trustworthy = LOCAL_ANY; | |
215bb3f7 | 231 | else |
d1d3482b N |
232 | trustworthy = FOREIGN; |
233 | ||
4e8d9f0a | 234 | if (!match && !conf_test_metadata(st->ss->name, policy, |
d1d3482b | 235 | (trustworthy == LOCAL))) { |
11b6d91d | 236 | if (c->verbose >= 1) |
e7b84f9d N |
237 | pr_err("%s has metadata type %s for which " |
238 | "auto-assembly is disabled\n", | |
239 | devname, st->ss->name); | |
15d4a7e4 | 240 | goto out; |
d1d3482b N |
241 | } |
242 | if (trustworthy == LOCAL_ANY) | |
ecb02e31 | 243 | trustworthy = LOCAL; |
e12e1d1d | 244 | |
4ef2f11e N |
245 | /* There are three possible sources for 'autof': command line, |
246 | * ARRAY line in mdadm.conf, or CREATE line in mdadm.conf. | |
350ac35d N |
247 | * ARRAY takes precedence, then command line, then |
248 | * CREATE. | |
4ef2f11e | 249 | */ |
350ac35d | 250 | if (match && match->autof) |
11b6d91d N |
251 | c->autof = match->autof; |
252 | if (c->autof == 0) | |
253 | c->autof = ci->autof; | |
4ef2f11e | 254 | |
7cdc0872 | 255 | name_to_use = info.name; |
0ac91628 | 256 | if (name_to_use[0] == 0 && |
0c4304ca | 257 | info.array.level == LEVEL_CONTAINER) { |
ecb02e31 N |
258 | name_to_use = info.text_version; |
259 | trustworthy = METADATA; | |
260 | } | |
0ac91628 | 261 | if (name_to_use[0] && trustworthy != LOCAL && |
11b6d91d | 262 | ! c->require_homehost && |
0ac91628 N |
263 | conf_name_is_free(name_to_use)) |
264 | trustworthy = LOCAL; | |
ecb02e31 | 265 | |
7cdc0872 N |
266 | /* strip "hostname:" prefix from name if we have decided |
267 | * to treat it as LOCAL | |
268 | */ | |
269 | if (trustworthy == LOCAL && strchr(name_to_use, ':') != NULL) | |
270 | name_to_use = strchr(name_to_use, ':')+1; | |
271 | ||
ad5bc697 | 272 | /* 4/ Check if array exists. |
215bb3f7 | 273 | */ |
93c861ee | 274 | if (map_lock(&map)) |
e7b84f9d | 275 | pr_err("failed to get exclusive lock on " |
93c861ee | 276 | "mapfile\n"); |
0431869c N |
277 | /* Now check we can get O_EXCL. If not, probably "mdadm -A" has |
278 | * taken over | |
279 | */ | |
280 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
281 | if (dfd < 0) { | |
282 | if (c->verbose >= 0) | |
283 | pr_err("cannot reopen %s: %s.\n", | |
284 | devname, strerror(errno)); | |
285 | goto out_unlock; | |
286 | } | |
287 | /* Cannot hold it open while we add the device to the array, | |
288 | * so we must release the O_EXCL and depend on the map_lock() | |
83785d30 | 289 | * So now is the best time to remove any partitions. |
0431869c | 290 | */ |
83785d30 | 291 | remove_partitions(dfd); |
0431869c N |
292 | close(dfd); |
293 | dfd = -1; | |
294 | ||
215bb3f7 | 295 | mp = map_by_uuid(&map, info.uuid); |
339c2d6c | 296 | if (mp) |
4dd2df09 | 297 | mdfd = open_dev(mp->devnm); |
339c2d6c | 298 | else |
215bb3f7 | 299 | mdfd = -1; |
d7288ddc | 300 | |
8382f19b | 301 | if (mdfd < 0) { |
8382f19b | 302 | |
215bb3f7 | 303 | /* Couldn't find an existing array, maybe make a new one */ |
ffb3e494 | 304 | mdfd = create_mddev(match ? match->devname : NULL, |
11b6d91d | 305 | name_to_use, c->autof, trustworthy, chosen_name); |
215bb3f7 N |
306 | |
307 | if (mdfd < 0) | |
25824e2d | 308 | goto out_unlock; |
215bb3f7 | 309 | |
4dd2df09 | 310 | sysfs_init(&info, mdfd, NULL); |
215bb3f7 | 311 | |
f35f2525 | 312 | if (set_array_info(mdfd, st, &info) != 0) { |
e7b84f9d | 313 | pr_err("failed to set array info for %s: %s\n", |
8382f19b | 314 | chosen_name, strerror(errno)); |
15d4a7e4 | 315 | rv = 2; |
25824e2d | 316 | goto out_unlock; |
8382f19b | 317 | } |
7801ac20 | 318 | |
c5afc314 N |
319 | dinfo = info; |
320 | dinfo.disk.major = major(stb.st_rdev); | |
321 | dinfo.disk.minor = minor(stb.st_rdev); | |
322 | if (add_disk(mdfd, st, &info, &dinfo) != 0) { | |
75a410f6 | 323 | pr_err("failed to add %s to new array %s: %s.\n", |
8382f19b NB |
324 | devname, chosen_name, strerror(errno)); |
325 | ioctl(mdfd, STOP_ARRAY, 0); | |
15d4a7e4 | 326 | rv = 2; |
25824e2d | 327 | goto out_unlock; |
8382f19b | 328 | } |
4dd2df09 N |
329 | sra = sysfs_read(mdfd, NULL, (GET_DEVS | GET_STATE | |
330 | GET_OFFSET | GET_SIZE)); | |
ca3b6696 | 331 | |
06c7f68e | 332 | if (!sra || !sra->devs || sra->devs->disk.raid_disk >= 0) { |
8382f19b NB |
333 | /* It really should be 'none' - must be old buggy |
334 | * kernel, and mdadm -I may not be able to complete. | |
335 | * So reject it. | |
336 | */ | |
337 | ioctl(mdfd, STOP_ARRAY, NULL); | |
e7b84f9d N |
338 | pr_err("You have an old buggy kernel which cannot support\n" |
339 | " --incremental reliably. Aborting.\n"); | |
15d4a7e4 | 340 | rv = 2; |
25824e2d | 341 | goto out_unlock; |
8382f19b | 342 | } |
c5afc314 | 343 | info.array.working_disks = 1; |
ad5bc697 | 344 | /* 6/ Make sure /var/run/mdadm.map contains this array. */ |
4dd2df09 | 345 | map_update(&map, fd2devnm(mdfd), |
ad5bc697 N |
346 | info.text_version, |
347 | info.uuid, chosen_name); | |
8382f19b NB |
348 | } else { |
349 | /* 5b/ if it does */ | |
350 | /* - check one drive in array to make sure metadata is a reasonably */ | |
351 | /* close match. Reject if not (e.g. different type) */ | |
352 | /* - add the device */ | |
353 | char dn[20]; | |
354 | int dfd2; | |
8382f19b | 355 | int err; |
3da92f27 | 356 | struct supertype *st2; |
c5afc314 | 357 | struct mdinfo info2, *d; |
215bb3f7 | 358 | |
4dd2df09 | 359 | sra = sysfs_read(mdfd, NULL, (GET_DEVS | GET_STATE | |
2bf740b3 | 360 | GET_OFFSET | GET_SIZE)); |
ca3b6696 | 361 | |
339c2d6c N |
362 | if (mp->path) |
363 | strcpy(chosen_name, mp->path); | |
364 | else | |
4dd2df09 | 365 | strcpy(chosen_name, mp->devnm); |
215bb3f7 | 366 | |
ef83fe7c N |
367 | /* It is generally not OK to add non-spare drives to a |
368 | * running array as they are probably missing because | |
369 | * they failed. However if runstop is 1, then the | |
7e83544b N |
370 | * array was possibly started early and our best bet is |
371 | * to add this anyway. | |
372 | * Also if action policy is re-add or better we allow | |
e5e5d7ce N |
373 | * re-add. |
374 | * This doesn't apply to containers as the 'non-spare' | |
375 | * flag has a different meaning. The test has to happen | |
376 | * at the device level there | |
3a6ec29a | 377 | */ |
e5e5d7ce N |
378 | if (!st->ss->external |
379 | && (info.disk.state & (1<<MD_DISK_SYNC)) != 0 | |
7e83544b N |
380 | && ! policy_action_allows(policy, st->ss->name, |
381 | act_re_add) | |
11b6d91d | 382 | && c->runstop < 1) { |
e5e5d7ce | 383 | if (ioctl(mdfd, GET_ARRAY_INFO, &ainf) == 0) { |
e7b84f9d N |
384 | pr_err("not adding %s to active array (without --run) %s\n", |
385 | devname, chosen_name); | |
15d4a7e4 | 386 | rv = 2; |
25824e2d | 387 | goto out_unlock; |
3a6ec29a N |
388 | } |
389 | } | |
15d4a7e4 N |
390 | if (!sra) { |
391 | rv = 2; | |
25824e2d | 392 | goto out_unlock; |
15d4a7e4 | 393 | } |
7c548327 N |
394 | if (sra->devs) { |
395 | sprintf(dn, "%d:%d", sra->devs->disk.major, | |
396 | sra->devs->disk.minor); | |
397 | dfd2 = dev_open(dn, O_RDONLY); | |
fb745c4b | 398 | if (dfd2 < 0) { |
e7b84f9d | 399 | pr_err("unable to open %s\n", devname); |
fb745c4b JS |
400 | rv = 2; |
401 | goto out_unlock; | |
402 | } | |
7c548327 N |
403 | st2 = dup_super(st); |
404 | if (st2->ss->load_super(st2, dfd2, NULL) || | |
405 | st->ss->compare_super(st, st2) != 0) { | |
e7b84f9d N |
406 | pr_err("metadata mismatch between %s and " |
407 | "chosen array %s\n", | |
408 | devname, chosen_name); | |
7c548327 | 409 | close(dfd2); |
15d4a7e4 | 410 | rv = 2; |
25824e2d | 411 | goto out_unlock; |
7c548327 | 412 | } |
8382f19b | 413 | close(dfd2); |
a5d85af7 | 414 | st2->ss->getinfo_super(st2, &info2, NULL); |
7c548327 N |
415 | st2->ss->free_super(st2); |
416 | if (info.array.level != info2.array.level || | |
417 | memcmp(info.uuid, info2.uuid, 16) != 0 || | |
418 | info.array.raid_disks != info2.array.raid_disks) { | |
e7b84f9d N |
419 | pr_err("unexpected difference between %s and %s.\n", |
420 | chosen_name, devname); | |
15d4a7e4 | 421 | rv = 2; |
25824e2d | 422 | goto out_unlock; |
7c548327 | 423 | } |
8382f19b | 424 | } |
69fe207e N |
425 | info.disk.major = major(stb.st_rdev); |
426 | info.disk.minor = minor(stb.st_rdev); | |
c5afc314 N |
427 | /* add disk needs to know about containers */ |
428 | if (st->ss->external) | |
429 | sra->array.level = LEVEL_CONTAINER; | |
69fe207e | 430 | err = add_disk(mdfd, st, sra, &info); |
8382f19b NB |
431 | if (err < 0 && errno == EBUSY) { |
432 | /* could be another device present with the same | |
433 | * disk.number. Find and reject any such | |
434 | */ | |
435 | find_reject(mdfd, st, sra, info.disk.number, | |
11b6d91d | 436 | info.events, c->verbose, chosen_name); |
69fe207e | 437 | err = add_disk(mdfd, st, sra, &info); |
8382f19b | 438 | } |
75a410f6 N |
439 | if (err < 0 && errno == EINVAL && |
440 | info.disk.state & (1<<MD_DISK_SYNC)) { | |
441 | /* Maybe it needs to be added as a spare */ | |
442 | if (policy_action_allows(policy, st->ss->name, | |
443 | act_force_spare)) { | |
444 | info.disk.state &= ~(1<<MD_DISK_SYNC); | |
445 | err = add_disk(mdfd, st, sra, &info); | |
446 | } else | |
447 | if (c->verbose >= 0) | |
448 | pr_err("can only add %s to %s as a spare, and force-spare is not set.\n", | |
449 | devname, chosen_name); | |
450 | } | |
8382f19b | 451 | if (err < 0) { |
75a410f6 | 452 | pr_err("failed to add %s to existing array %s: %s.\n", |
8382f19b | 453 | devname, chosen_name, strerror(errno)); |
15d4a7e4 | 454 | rv = 2; |
25824e2d | 455 | goto out_unlock; |
8382f19b | 456 | } |
c5afc314 N |
457 | info.array.working_disks = 0; |
458 | for (d = sra->devs; d; d=d->next) | |
459 | info.array.working_disks ++; | |
ca3b6696 | 460 | |
8382f19b | 461 | } |
8382f19b NB |
462 | |
463 | /* 7/ Is there enough devices to possibly start the array? */ | |
464 | /* 7a/ if not, finish with success. */ | |
352452c3 | 465 | if (info.array.level == LEVEL_CONTAINER) { |
4dd2df09 | 466 | char devnm[32]; |
352452c3 | 467 | /* Try to assemble within the container */ |
a408d66c | 468 | sysfs_uevent(sra, "change"); |
11b6d91d | 469 | if (c->verbose >= 0) |
e7b84f9d N |
470 | pr_err("container %s now has %d device%s\n", |
471 | chosen_name, info.array.working_disks, | |
ca3b6696 | 472 | info.array.working_disks == 1?"":"s"); |
a7c6e3fb | 473 | wait_for(chosen_name, mdfd); |
ed690d36 | 474 | if (st->ss->external) |
4dd2df09 | 475 | strcpy(devnm, fd2devnm(mdfd)); |
5fc8cff3 JS |
476 | if (st->ss->load_container) |
477 | rv = st->ss->load_container(st, mdfd, NULL); | |
a7c6e3fb | 478 | close(mdfd); |
2bf740b3 | 479 | sysfs_free(sra); |
5fc8cff3 | 480 | if (!rv) |
11b6d91d | 481 | rv = Incremental_container(st, chosen_name, c); |
5fc8cff3 | 482 | map_unlock(&map); |
03b7f6c6 N |
483 | if (rv == 1) |
484 | /* Don't fail the whole -I if a subarray didn't | |
485 | * have enough devices to start yet | |
486 | */ | |
487 | rv = 0; | |
ed690d36 PC |
488 | /* after spare is added, ping monitor for external metadata |
489 | * so that it can eg. try to rebuild degraded array */ | |
983fff45 | 490 | if (st->ss->external) |
4dd2df09 | 491 | ping_monitor(devnm); |
03b7f6c6 | 492 | return rv; |
352452c3 | 493 | } |
de6ae750 | 494 | |
2bf740b3 N |
495 | /* We have added something to the array, so need to re-read the |
496 | * state. Eventually this state should be kept up-to-date as | |
497 | * things change. | |
498 | */ | |
499 | sysfs_free(sra); | |
4dd2df09 | 500 | sra = sysfs_read(mdfd, NULL, (GET_DEVS | GET_STATE | |
2bf740b3 N |
501 | GET_OFFSET | GET_SIZE)); |
502 | active_disks = count_active(st, sra, mdfd, &avail, &info); | |
8382f19b NB |
503 | if (enough(info.array.level, info.array.raid_disks, |
504 | info.array.layout, info.array.state & 1, | |
de5a472e | 505 | avail) == 0) { |
11b6d91d | 506 | if (c->verbose >= 0) |
e7b84f9d N |
507 | pr_err("%s attached to %s, not enough to start (%d).\n", |
508 | devname, chosen_name, active_disks); | |
15d4a7e4 | 509 | rv = 0; |
25824e2d | 510 | goto out_unlock; |
8382f19b | 511 | } |
8382f19b NB |
512 | |
513 | /* 7b/ if yes, */ | |
514 | /* - if number of OK devices match expected, or -R and there */ | |
515 | /* are enough, */ | |
516 | /* + add any bitmap file */ | |
517 | /* + start the array (auto-readonly). */ | |
8382f19b NB |
518 | |
519 | if (ioctl(mdfd, GET_ARRAY_INFO, &ainf) == 0) { | |
11b6d91d | 520 | if (c->verbose >= 0) |
e7b84f9d N |
521 | pr_err("%s attached to %s which is already active.\n", |
522 | devname, chosen_name); | |
15d4a7e4 | 523 | rv = 0; |
25824e2d | 524 | goto out_unlock; |
8382f19b | 525 | } |
ad5bc697 N |
526 | |
527 | map_unlock(&map); | |
11b6d91d | 528 | if (c->runstop > 0 || active_disks >= info.array.working_disks) { |
2bf740b3 | 529 | struct mdinfo *dsk; |
8382f19b | 530 | /* Let's try to start it */ |
5e88ab2e N |
531 | |
532 | if (info.reshape_active && !(info.reshape_active & RESHAPE_NO_BACKUP)) { | |
ed503f89 N |
533 | pr_err("%s: This array is being reshaped and cannot be started\n", |
534 | chosen_name); | |
535 | cont_err("by --incremental. Please use --assemble\n"); | |
5e88ab2e N |
536 | goto out; |
537 | } | |
8382f19b NB |
538 | if (match && match->bitmap_file) { |
539 | int bmfd = open(match->bitmap_file, O_RDWR); | |
540 | if (bmfd < 0) { | |
e7b84f9d N |
541 | pr_err("Could not open bitmap file %s.\n", |
542 | match->bitmap_file); | |
15d4a7e4 | 543 | goto out; |
8382f19b NB |
544 | } |
545 | if (ioctl(mdfd, SET_BITMAP_FILE, bmfd) != 0) { | |
546 | close(bmfd); | |
e7b84f9d N |
547 | pr_err("Failed to set bitmapfile for %s.\n", |
548 | chosen_name); | |
15d4a7e4 | 549 | goto out; |
8382f19b NB |
550 | } |
551 | close(bmfd); | |
552 | } | |
de6ae750 N |
553 | /* Need to remove from the array any devices which |
554 | * 'count_active' discerned were too old or inappropriate | |
555 | */ | |
556 | for (d = sra ? sra->devs : NULL ; d ; d = d->next) | |
557 | if (d->disk.state & (1<<MD_DISK_REMOVED)) | |
558 | remove_disk(mdfd, st, sra, d); | |
2bf740b3 | 559 | |
7b403fef | 560 | if ((sra == NULL || active_disks >= info.array.working_disks) |
215bb3f7 | 561 | && trustworthy != FOREIGN) |
8382f19b NB |
562 | rv = ioctl(mdfd, RUN_ARRAY, NULL); |
563 | else | |
564 | rv = sysfs_set_str(sra, NULL, | |
565 | "array_state", "read-auto"); | |
566 | if (rv == 0) { | |
11b6d91d | 567 | if (c->verbose >= 0) |
e7b84f9d N |
568 | pr_err("%s attached to %s, which has been started.\n", |
569 | devname, chosen_name); | |
8382f19b | 570 | rv = 0; |
a7c6e3fb | 571 | wait_for(chosen_name, mdfd); |
7e83544b N |
572 | /* We just started the array, so some devices |
573 | * might have been evicted from the array | |
574 | * because their event counts were too old. | |
575 | * If the action=re-add policy is in-force for | |
576 | * those devices we should re-add them now. | |
577 | */ | |
578 | for (dsk = sra->devs; dsk ; dsk = dsk->next) { | |
579 | if (disk_action_allows(dsk, st->ss->name, act_re_add) && | |
580 | add_disk(mdfd, st, sra, dsk) == 0) | |
e7b84f9d N |
581 | pr_err("%s re-added to %s\n", |
582 | dsk->sys_name, chosen_name); | |
7e83544b | 583 | } |
8382f19b | 584 | } else { |
e7b84f9d N |
585 | pr_err("%s attached to %s, but failed to start: %s.\n", |
586 | devname, chosen_name, strerror(errno)); | |
8382f19b NB |
587 | rv = 1; |
588 | } | |
589 | } else { | |
11b6d91d | 590 | if (c->verbose >= 0) |
e7b84f9d N |
591 | pr_err("%s attached to %s, not enough to start safely.\n", |
592 | devname, chosen_name); | |
8382f19b NB |
593 | rv = 0; |
594 | } | |
15d4a7e4 | 595 | out: |
de6ae750 | 596 | free(avail); |
15d4a7e4 N |
597 | if (dfd >= 0) |
598 | close(dfd); | |
599 | if (mdfd >= 0) | |
600 | close(mdfd); | |
7e83544b N |
601 | if (policy) |
602 | dev_policy_free(policy); | |
2bf740b3 N |
603 | if (sra) |
604 | sysfs_free(sra); | |
8382f19b | 605 | return rv; |
25824e2d JS |
606 | out_unlock: |
607 | map_unlock(&map); | |
608 | goto out; | |
8382f19b NB |
609 | } |
610 | ||
7e0f6979 | 611 | static void find_reject(int mdfd, struct supertype *st, struct mdinfo *sra, |
8382f19b NB |
612 | int number, __u64 events, int verbose, |
613 | char *array_name) | |
614 | { | |
3da92f27 | 615 | /* Find a device attached to this array with a disk.number of number |
8382f19b NB |
616 | * and events less than the passed events, and remove the device. |
617 | */ | |
06c7f68e | 618 | struct mdinfo *d; |
8382f19b NB |
619 | mdu_array_info_t ra; |
620 | ||
621 | if (ioctl(mdfd, GET_ARRAY_INFO, &ra) == 0) | |
622 | return; /* not safe to remove from active arrays | |
623 | * without thinking more */ | |
624 | ||
625 | for (d = sra->devs; d ; d = d->next) { | |
626 | char dn[10]; | |
627 | int dfd; | |
8382f19b | 628 | struct mdinfo info; |
06c7f68e | 629 | sprintf(dn, "%d:%d", d->disk.major, d->disk.minor); |
8382f19b NB |
630 | dfd = dev_open(dn, O_RDONLY); |
631 | if (dfd < 0) | |
632 | continue; | |
3da92f27 | 633 | if (st->ss->load_super(st, dfd, NULL)) { |
8382f19b NB |
634 | close(dfd); |
635 | continue; | |
636 | } | |
a5d85af7 | 637 | st->ss->getinfo_super(st, &info, NULL); |
3da92f27 | 638 | st->ss->free_super(st); |
8382f19b NB |
639 | close(dfd); |
640 | ||
641 | if (info.disk.number != number || | |
642 | info.events >= events) | |
643 | continue; | |
644 | ||
06c7f68e | 645 | if (d->disk.raid_disk > -1) |
8382f19b NB |
646 | sysfs_set_str(sra, d, "slot", "none"); |
647 | if (sysfs_set_str(sra, d, "state", "remove") == 0) | |
648 | if (verbose >= 0) | |
e7b84f9d N |
649 | pr_err("removing old device %s from %s\n", |
650 | d->sys_name+4, array_name); | |
8382f19b NB |
651 | } |
652 | } | |
653 | ||
2bf740b3 N |
654 | static int count_active(struct supertype *st, struct mdinfo *sra, |
655 | int mdfd, char **availp, | |
8382f19b NB |
656 | struct mdinfo *bestinfo) |
657 | { | |
658 | /* count how many devices in sra think they are active */ | |
06c7f68e | 659 | struct mdinfo *d; |
de6ae750 | 660 | int cnt = 0; |
72e7fb13 | 661 | int replcnt = 0; |
8382f19b | 662 | __u64 max_events = 0; |
8382f19b | 663 | char *avail = NULL; |
e4c72d1d | 664 | int *best = NULL; |
de6ae750 N |
665 | char *devmap = NULL; |
666 | int numdevs = 0; | |
667 | int devnum; | |
668 | int b, i; | |
669 | int raid_disks = 0; | |
8382f19b | 670 | |
b526e52d DW |
671 | if (!sra) |
672 | return 0; | |
673 | ||
de6ae750 N |
674 | for (d = sra->devs ; d ; d = d->next) |
675 | numdevs++; | |
ca3b6696 | 676 | for (d = sra->devs, devnum = 0 ; d ; d = d->next, devnum++) { |
8382f19b NB |
677 | char dn[30]; |
678 | int dfd; | |
8382f19b NB |
679 | int ok; |
680 | struct mdinfo info; | |
681 | ||
06c7f68e | 682 | sprintf(dn, "%d:%d", d->disk.major, d->disk.minor); |
8382f19b NB |
683 | dfd = dev_open(dn, O_RDONLY); |
684 | if (dfd < 0) | |
685 | continue; | |
3da92f27 | 686 | ok = st->ss->load_super(st, dfd, NULL); |
8382f19b NB |
687 | close(dfd); |
688 | if (ok != 0) | |
689 | continue; | |
de6ae750 N |
690 | info.array.raid_disks = raid_disks; |
691 | st->ss->getinfo_super(st, &info, devmap + raid_disks * devnum); | |
43aaf431 | 692 | if (!avail) { |
de6ae750 | 693 | raid_disks = info.array.raid_disks; |
503975b9 | 694 | avail = xcalloc(raid_disks, 1); |
43aaf431 | 695 | *availp = avail; |
de6ae750 | 696 | |
503975b9 N |
697 | best = xcalloc(raid_disks, sizeof(int)); |
698 | devmap = xcalloc(raid_disks, numdevs); | |
de6ae750 N |
699 | |
700 | st->ss->getinfo_super(st, &info, devmap); | |
43aaf431 DL |
701 | } |
702 | ||
8382f19b NB |
703 | if (info.disk.state & (1<<MD_DISK_SYNC)) |
704 | { | |
8382f19b NB |
705 | if (cnt == 0) { |
706 | cnt++; | |
707 | max_events = info.events; | |
708 | avail[info.disk.raid_disk] = 2; | |
de6ae750 | 709 | best[info.disk.raid_disk] = devnum; |
a5d85af7 | 710 | st->ss->getinfo_super(st, bestinfo, NULL); |
8382f19b | 711 | } else if (info.events == max_events) { |
8382f19b | 712 | avail[info.disk.raid_disk] = 2; |
de6ae750 | 713 | best[info.disk.raid_disk] = devnum; |
8382f19b | 714 | } else if (info.events == max_events-1) { |
de6ae750 N |
715 | if (avail[info.disk.raid_disk] == 0) { |
716 | avail[info.disk.raid_disk] = 1; | |
717 | best[info.disk.raid_disk] = devnum; | |
718 | } | |
8382f19b NB |
719 | } else if (info.events < max_events - 1) |
720 | ; | |
721 | else if (info.events == max_events+1) { | |
722 | int i; | |
8382f19b | 723 | max_events = info.events; |
ca3b6696 | 724 | for (i = 0; i < raid_disks; i++) |
8382f19b NB |
725 | if (avail[i]) |
726 | avail[i]--; | |
727 | avail[info.disk.raid_disk] = 2; | |
de6ae750 | 728 | best[info.disk.raid_disk] = devnum; |
a5d85af7 | 729 | st->ss->getinfo_super(st, bestinfo, NULL); |
8382f19b | 730 | } else { /* info.events much bigger */ |
47c7a4be | 731 | memset(avail, 0, raid_disks); |
8382f19b | 732 | max_events = info.events; |
1d1a9f87 | 733 | avail[info.disk.raid_disk] = 2; |
47c7a4be | 734 | best[info.disk.raid_disk] = devnum; |
a5d85af7 | 735 | st->ss->getinfo_super(st, bestinfo, NULL); |
8382f19b | 736 | } |
72e7fb13 N |
737 | } else if (info.disk.state & (1<<MD_DISK_REPLACEMENT)) |
738 | replcnt++; | |
3da92f27 | 739 | st->ss->free_super(st); |
8382f19b | 740 | } |
de6ae750 N |
741 | if (!avail) |
742 | return 0; | |
743 | /* We need to reject any device that thinks the best device is | |
744 | * failed or missing */ | |
745 | for (b = 0; b < raid_disks; b++) | |
746 | if (avail[b] == 2) | |
747 | break; | |
748 | cnt = 0; | |
749 | for (i = 0 ; i < raid_disks ; i++) { | |
750 | if (i != b && avail[i]) | |
751 | if (devmap[raid_disks * best[i] + b] == 0) { | |
752 | /* This device thinks 'b' is failed - | |
753 | * don't use it */ | |
754 | devnum = best[i]; | |
755 | for (d=sra->devs ; devnum; d = d->next) | |
756 | devnum--; | |
757 | d->disk.state |= (1 << MD_DISK_REMOVED); | |
758 | avail[i] = 0; | |
759 | } | |
760 | if (avail[i]) | |
761 | cnt++; | |
762 | } | |
f80057ae N |
763 | /* Also need to reject any spare device with an event count that |
764 | * is too high | |
765 | */ | |
766 | for (d = sra->devs; d; d = d->next) { | |
767 | if (!(d->disk.state & (1<<MD_DISK_SYNC)) && | |
768 | d->events > max_events) | |
769 | d->disk.state |= (1 << MD_DISK_REMOVED); | |
770 | } | |
de6ae750 N |
771 | free(best); |
772 | free(devmap); | |
72e7fb13 | 773 | return cnt + replcnt; |
8382f19b NB |
774 | } |
775 | ||
5be68a07 PC |
776 | /* test if container has degraded member(s) */ |
777 | static int container_members_max_degradation(struct map_ent *map, struct map_ent *me) | |
778 | { | |
779 | mdu_array_info_t array; | |
780 | int afd; | |
781 | int max_degraded = 0; | |
782 | ||
783 | for(; map; map = map->next) { | |
4dd2df09 | 784 | if (!metadata_container_matches(map->metadata, me->devnm)) |
5be68a07 | 785 | continue; |
4dd2df09 | 786 | afd = open_dev(map->devnm); |
5be68a07 PC |
787 | if (afd < 0) |
788 | continue; | |
789 | /* most accurate information regarding array degradation */ | |
790 | if (ioctl(afd, GET_ARRAY_INFO, &array) >= 0) { | |
791 | int degraded = array.raid_disks - array.active_disks - | |
792 | array.spare_disks; | |
793 | if (degraded > max_degraded) | |
794 | max_degraded = degraded; | |
795 | } | |
796 | close(afd); | |
797 | } | |
798 | return (max_degraded); | |
799 | } | |
800 | ||
56e8be85 | 801 | static int array_try_spare(char *devname, int *dfdp, struct dev_policy *pol, |
d2db3045 | 802 | struct map_ent *target, int bare, |
56e8be85 | 803 | struct supertype *st, int verbose) |
f08605b3 N |
804 | { |
805 | /* This device doesn't have any md metadata | |
d2db3045 N |
806 | * The device policy allows 'spare' and if !bare, it allows spare-same-slot. |
807 | * If 'st' is not set, then we only know that some metadata allows this, | |
808 | * others possibly don't. | |
809 | * So look for a container or array to attach the device to. | |
810 | * Prefer 'target' if that is set and the array is found. | |
811 | * | |
f08605b3 N |
812 | * If st is set, then only arrays of that type are considered |
813 | * Return 0 on success, or some exit code on failure, probably 1. | |
814 | */ | |
8659d089 | 815 | int rv = 1; |
f08605b3 N |
816 | struct stat stb; |
817 | struct map_ent *mp, *map = NULL; | |
818 | struct mdinfo *chosen = NULL; | |
819 | int dfd = *dfdp; | |
820 | ||
f08605b3 N |
821 | if (fstat(dfd, &stb) != 0) |
822 | return 1; | |
f08605b3 | 823 | |
56e8be85 | 824 | /* |
f08605b3 N |
825 | * Now we need to find a suitable array to add this to. |
826 | * We only accept arrays that: | |
827 | * - match 'st' | |
828 | * - are in the same domains as the device | |
829 | * - are of an size for which the device will be useful | |
830 | * and we choose the one that is the most degraded | |
831 | */ | |
832 | ||
833 | if (map_lock(&map)) { | |
e7b84f9d | 834 | pr_err("failed to get exclusive lock on " |
f08605b3 N |
835 | "mapfile\n"); |
836 | return 1; | |
837 | } | |
838 | for (mp = map ; mp ; mp = mp->next) { | |
839 | struct supertype *st2; | |
840 | struct domainlist *dl = NULL; | |
841 | struct mdinfo *sra; | |
842 | unsigned long long devsize; | |
71204a50 | 843 | unsigned long long component_size = 0; |
f08605b3 N |
844 | |
845 | if (is_subarray(mp->metadata)) | |
846 | continue; | |
847 | if (st) { | |
848 | st2 = st->ss->match_metadata_desc(mp->metadata); | |
849 | if (!st2 || | |
850 | (st->minor_version >= 0 && | |
851 | st->minor_version != st2->minor_version)) { | |
852 | if (verbose > 1) | |
e7b84f9d | 853 | pr_err("not adding %s to %s as metadata type doesn't match\n", |
f08605b3 N |
854 | devname, mp->path); |
855 | free(st2); | |
856 | continue; | |
857 | } | |
858 | free(st2); | |
859 | } | |
4dd2df09 | 860 | sra = sysfs_read(-1, mp->devnm, |
f08605b3 N |
861 | GET_DEVS|GET_OFFSET|GET_SIZE|GET_STATE| |
862 | GET_DEGRADED|GET_COMPONENT|GET_VERSION); | |
863 | if (!sra) { | |
864 | /* Probably a container - no degraded info */ | |
4dd2df09 | 865 | sra = sysfs_read(-1, mp->devnm, |
f08605b3 N |
866 | GET_DEVS|GET_OFFSET|GET_SIZE|GET_STATE| |
867 | GET_COMPONENT|GET_VERSION); | |
868 | if (sra) | |
5be68a07 | 869 | sra->array.failed_disks = -1; |
f08605b3 N |
870 | } |
871 | if (!sra) | |
872 | continue; | |
873 | if (st == NULL) { | |
874 | int i; | |
875 | st2 = NULL; | |
ca3b6696 | 876 | for(i = 0; !st2 && superlist[i]; i++) |
f08605b3 N |
877 | st2 = superlist[i]->match_metadata_desc( |
878 | sra->text_version); | |
6e57f80a N |
879 | if (!st2) { |
880 | if (verbose > 1) | |
e7b84f9d | 881 | pr_err("not adding %s to %s" |
6e57f80a N |
882 | " as metadata not recognised.\n", |
883 | devname, mp->path); | |
884 | goto next; | |
885 | } | |
625b2507 N |
886 | /* Need to double check the 'act_spare' permissions applies |
887 | * to this metadata. | |
888 | */ | |
889 | if (!policy_action_allows(pol, st2->ss->name, act_spare)) | |
890 | goto next; | |
d2db3045 N |
891 | if (!bare && !policy_action_allows(pol, st2->ss->name, |
892 | act_spare_same_slot)) | |
893 | goto next; | |
f08605b3 N |
894 | } else |
895 | st2 = st; | |
5be68a07 PC |
896 | /* update number of failed disks for mostly degraded |
897 | * container member */ | |
898 | if (sra->array.failed_disks == -1) | |
899 | sra->array.failed_disks = container_members_max_degradation(map, mp); | |
900 | ||
f08605b3 | 901 | get_dev_size(dfd, NULL, &devsize); |
48865704 PC |
902 | if (sra->component_size == 0) { |
903 | /* true for containers, here we must read superblock | |
904 | * to obtain minimum spare size */ | |
905 | struct supertype *st3 = dup_super(st2); | |
4dd2df09 | 906 | int mdfd = open_dev(mp->devnm); |
f56128b9 | 907 | if (mdfd < 0) { |
1fdeb8a0 | 908 | free(st3); |
48865704 | 909 | goto next; |
1fdeb8a0 | 910 | } |
48865704 PC |
911 | if (st3->ss->load_container && |
912 | !st3->ss->load_container(st3, mdfd, mp->path)) { | |
913 | component_size = st3->ss->min_acceptable_spare_size(st3); | |
914 | st3->ss->free_super(st3); | |
915 | } | |
916 | free(st3); | |
917 | close(mdfd); | |
918 | } | |
919 | if ((sra->component_size > 0 && | |
387fcd59 N |
920 | st2->ss->avail_size(st2, devsize, |
921 | sra->devs | |
922 | ? sra->devs->data_offset | |
923 | : INVALID_SECTORS) | |
924 | < sra->component_size) | |
48865704 PC |
925 | || |
926 | (sra->component_size == 0 && devsize < component_size)) { | |
f08605b3 | 927 | if (verbose > 1) |
e7b84f9d | 928 | pr_err("not adding %s to %s as it is too small\n", |
f08605b3 N |
929 | devname, mp->path); |
930 | goto next; | |
931 | } | |
d2db3045 N |
932 | /* test against target. |
933 | * If 'target' is set and 'bare' is false, we only accept | |
934 | * arrays/containers that match 'target'. | |
935 | * If 'target' is set and 'bare' is true, we prefer the | |
936 | * array which matches 'target'. | |
b4924220 | 937 | * target is considered only if we deal with degraded array |
d2db3045 | 938 | */ |
fee6a49e PC |
939 | if (target && policy_action_allows(pol, st2->ss->name, |
940 | act_spare_same_slot)) { | |
d2db3045 N |
941 | if (strcmp(target->metadata, mp->metadata) == 0 && |
942 | memcmp(target->uuid, mp->uuid, | |
b4924220 PC |
943 | sizeof(target->uuid)) == 0 && |
944 | sra->array.failed_disks > 0) { | |
d2db3045 N |
945 | /* This is our target!! */ |
946 | if (chosen) | |
947 | sysfs_free(chosen); | |
948 | chosen = sra; | |
949 | sra = NULL; | |
950 | /* skip to end so we don't check any more */ | |
951 | while (mp->next) | |
952 | mp = mp->next; | |
953 | goto next; | |
954 | } | |
955 | /* not our target */ | |
956 | if (!bare) | |
957 | goto next; | |
958 | } | |
959 | ||
b6b3f0f7 | 960 | dl = domain_from_array(sra, st2->ss->name); |
e5508b36 | 961 | if (domain_test(dl, pol, st2->ss->name) != 1) { |
b6b3f0f7 PC |
962 | /* domain test fails */ |
963 | if (verbose > 1) | |
e7b84f9d | 964 | pr_err("not adding %s to %s as" |
b6b3f0f7 PC |
965 | " it is not in a compatible domain\n", |
966 | devname, mp->path); | |
967 | ||
968 | goto next; | |
969 | } | |
f08605b3 N |
970 | /* all tests passed, OK to add to this array */ |
971 | if (!chosen) { | |
972 | chosen = sra; | |
973 | sra = NULL; | |
974 | } else if (chosen->array.failed_disks < sra->array.failed_disks) { | |
975 | sysfs_free(chosen); | |
976 | chosen = sra; | |
977 | sra = NULL; | |
978 | } | |
979 | next: | |
980 | if (sra) | |
981 | sysfs_free(sra); | |
982 | if (st != st2) | |
983 | free(st2); | |
984 | if (dl) | |
985 | domain_free(dl); | |
986 | } | |
987 | if (chosen) { | |
988 | /* add current device to chosen array as a spare */ | |
4dd2df09 | 989 | int mdfd = open_dev(chosen->sys_name); |
f08605b3 | 990 | if (mdfd >= 0) { |
a655e550 | 991 | struct mddev_dev devlist; |
f08605b3 N |
992 | char devname[20]; |
993 | devlist.next = NULL; | |
994 | devlist.used = 0; | |
f08605b3 N |
995 | devlist.writemostly = 0; |
996 | devlist.devname = devname; | |
997 | sprintf(devname, "%d:%d", major(stb.st_rdev), | |
998 | minor(stb.st_rdev)); | |
999 | devlist.disposition = 'a'; | |
1000 | close(dfd); | |
1001 | *dfdp = -1; | |
1002 | rv = Manage_subdevs(chosen->sys_name, mdfd, &devlist, | |
11b391ec | 1003 | -1, 0, NULL, 0); |
f08605b3 N |
1004 | close(mdfd); |
1005 | } | |
1006 | if (verbose > 0) { | |
1007 | if (rv == 0) | |
e7b84f9d | 1008 | pr_err("added %s as spare for %s\n", |
f08605b3 N |
1009 | devname, chosen->sys_name); |
1010 | else | |
e7b84f9d | 1011 | pr_err("failed to add %s as spare for %s\n", |
f08605b3 N |
1012 | devname, chosen->sys_name); |
1013 | } | |
1014 | sysfs_free(chosen); | |
1015 | } | |
015da8f5 | 1016 | map_unlock(&map); |
8659d089 | 1017 | return rv; |
f08605b3 N |
1018 | } |
1019 | ||
56e8be85 N |
1020 | static int partition_try_spare(char *devname, int *dfdp, struct dev_policy *pol, |
1021 | struct supertype *st, int verbose) | |
1022 | { | |
61018da0 N |
1023 | /* we know that at least one partition virtual-metadata is |
1024 | * allowed to incorporate spares like this device. We need to | |
1025 | * find a suitable device to copy partition information from. | |
1026 | * | |
1027 | * Getting a list of all disk (not partition) devices is | |
1028 | * slightly non-trivial. We could look at /sys/block, but | |
1029 | * that is theoretically due to be removed. Maybe best to use | |
1030 | * /dev/disk/by-path/?* and ignore names ending '-partNN' as | |
1031 | * we depend on this directory of 'path' info. But that fails | |
1032 | * to find loop devices and probably others. Maybe don't | |
1033 | * worry about that, they aren't the real target. | |
1034 | * | |
1035 | * So: check things in /dev/disk/by-path to see if they are in | |
1036 | * a compatible domain, then load the partition table and see | |
1037 | * if it is OK for the new device, and choose the largest | |
1038 | * partition table that fits. | |
1039 | */ | |
1040 | DIR *dir; | |
1041 | struct dirent *de; | |
1042 | char *chosen = NULL; | |
71204a50 | 1043 | unsigned long long chosen_size = 0; |
61018da0 N |
1044 | struct supertype *chosen_st = NULL; |
1045 | int fd; | |
1046 | ||
1047 | dir = opendir("/dev/disk/by-path"); | |
1048 | if (!dir) | |
1049 | return 1; | |
1050 | while ((de = readdir(dir)) != NULL) { | |
1051 | char *ep; | |
1052 | struct dev_policy *pol2 = NULL; | |
1053 | struct domainlist *domlist = NULL; | |
1054 | int fd = -1; | |
1055 | struct mdinfo info; | |
1056 | struct supertype *st2 = NULL; | |
1057 | char *devname = NULL; | |
1058 | unsigned long long devsectors; | |
1059 | ||
1060 | if (de->d_ino == 0 || | |
1061 | de->d_name[0] == '.' || | |
1062 | (de->d_type != DT_LNK && de->d_type != DT_UNKNOWN)) | |
1063 | goto next; | |
1064 | ||
1065 | ep = de->d_name + strlen(de->d_name); | |
1066 | while (ep > de->d_name && | |
1067 | isdigit(ep[-1])) | |
1068 | ep--; | |
1069 | if (ep > de->d_name + 5 && | |
1070 | strncmp(ep-5, "-part", 5) == 0) | |
1071 | /* This is a partition - skip it */ | |
1072 | goto next; | |
1073 | ||
1074 | pol2 = path_policy(de->d_name, type_disk); | |
1075 | ||
1076 | domain_merge(&domlist, pol2, st ? st->ss->name : NULL); | |
e5508b36 | 1077 | if (domain_test(domlist, pol, st ? st->ss->name : NULL) != 1) |
61018da0 N |
1078 | /* new device is incompatible with this device. */ |
1079 | goto next; | |
1080 | ||
1081 | domain_free(domlist); | |
1082 | domlist = NULL; | |
1083 | ||
71204a50 N |
1084 | if (asprintf(&devname, "/dev/disk/by-path/%s", de->d_name) != 1) { |
1085 | devname = NULL; | |
1086 | goto next; | |
1087 | } | |
61018da0 N |
1088 | fd = open(devname, O_RDONLY); |
1089 | if (fd < 0) | |
1090 | goto next; | |
1091 | if (get_dev_size(fd, devname, &devsectors) == 0) | |
1092 | goto next; | |
1093 | devsectors >>= 9; | |
1094 | ||
1095 | if (st) | |
1096 | st2 = dup_super(st); | |
1097 | else | |
1098 | st2 = guess_super_type(fd, guess_partitions); | |
1099 | if (st2 == NULL || | |
1100 | st2->ss->load_super(st2, fd, NULL) < 0) | |
1101 | goto next; | |
1102 | ||
1103 | if (!st) { | |
1104 | /* Check domain policy again, this time referring to metadata */ | |
1105 | domain_merge(&domlist, pol2, st2->ss->name); | |
e5508b36 | 1106 | if (domain_test(domlist, pol, st2->ss->name) != 1) |
61018da0 N |
1107 | /* Incompatible devices for this metadata type */ |
1108 | goto next; | |
625b2507 N |
1109 | if (!policy_action_allows(pol, st2->ss->name, act_spare)) |
1110 | /* Some partition types allow sparing, but not | |
1111 | * this one. | |
1112 | */ | |
1113 | goto next; | |
61018da0 N |
1114 | } |
1115 | ||
a5d85af7 | 1116 | st2->ss->getinfo_super(st2, &info, NULL); |
61018da0 N |
1117 | if (info.component_size > devsectors) |
1118 | /* This partitioning doesn't fit in the device */ | |
1119 | goto next; | |
1120 | ||
1121 | /* This is an acceptable device to copy partition | |
1122 | * metadata from. We could just stop here, but I | |
1123 | * think I want to keep looking incase a larger | |
1124 | * metadata which makes better use of the device can | |
1125 | * be found. | |
1126 | */ | |
1127 | if (chosen == NULL || | |
1128 | chosen_size < info.component_size) { | |
1129 | chosen_size = info.component_size; | |
1130 | free(chosen); | |
1131 | chosen = devname; | |
1132 | devname = NULL; | |
1133 | if (chosen_st) { | |
1134 | chosen_st->ss->free_super(chosen_st); | |
1135 | free(chosen_st); | |
1136 | } | |
1137 | chosen_st = st2; | |
1138 | st2 = NULL; | |
1139 | } | |
1140 | ||
1141 | next: | |
1142 | free(devname); | |
1143 | domain_free(domlist); | |
1144 | dev_policy_free(pol2); | |
1145 | if (st2) | |
1146 | st2->ss->free_super(st2); | |
1147 | free(st2); | |
1148 | ||
1149 | if (fd >= 0) | |
1150 | close(fd); | |
1151 | } | |
1152 | ||
be5c60e3 JS |
1153 | closedir(dir); |
1154 | ||
61018da0 N |
1155 | if (!chosen) |
1156 | return 1; | |
1157 | ||
1158 | /* 'chosen' is the best device we can find. Let's write its | |
1159 | * metadata to devname dfd is read-only so don't use that | |
1160 | */ | |
1161 | fd = open(devname, O_RDWR); | |
1162 | if (fd >= 0) { | |
1163 | chosen_st->ss->store_super(chosen_st, fd); | |
1164 | close(fd); | |
1165 | } | |
1166 | free(chosen); | |
1167 | chosen_st->ss->free_super(chosen_st); | |
1168 | free(chosen_st); | |
1169 | return 0; | |
56e8be85 N |
1170 | } |
1171 | ||
52e965c2 N |
1172 | static int is_bare(int dfd) |
1173 | { | |
1174 | unsigned long long size = 0; | |
1175 | char bufpad[4096 + 4096]; | |
1176 | char *buf = (char*)(((long)bufpad + 4096) & ~4095); | |
1177 | ||
1178 | if (lseek(dfd, 0, SEEK_SET) != 0 || | |
1179 | read(dfd, buf, 4096) != 4096) | |
1180 | return 0; | |
1181 | ||
1182 | if (buf[0] != '\0' && buf[0] != '\x5a' && buf[0] != '\xff') | |
1183 | return 0; | |
1184 | if (memcmp(buf, buf+1, 4095) != 0) | |
1185 | return 0; | |
1186 | ||
1187 | /* OK, first 4K appear blank, try the end. */ | |
1188 | get_dev_size(dfd, NULL, &size); | |
1189 | if (lseek(dfd, size-4096, SEEK_SET) < 0 || | |
1190 | read(dfd, buf, 4096) != 4096) | |
1191 | return 0; | |
1192 | ||
1193 | if (buf[0] != '\0' && buf[0] != '\x5a' && buf[0] != '\xff') | |
1194 | return 0; | |
1195 | if (memcmp(buf, buf+1, 4095) != 0) | |
1196 | return 0; | |
1197 | ||
1198 | return 1; | |
1199 | } | |
56e8be85 N |
1200 | |
1201 | /* adding a spare to a regular array is quite different from adding one to | |
1202 | * a set-of-partitions virtual array. | |
1203 | * This function determines which is worth trying and tries as appropriate. | |
1204 | * Arrays are given priority over partitions. | |
1205 | */ | |
1206 | static int try_spare(char *devname, int *dfdp, struct dev_policy *pol, | |
d2db3045 | 1207 | struct map_ent *target, |
56e8be85 N |
1208 | struct supertype *st, int verbose) |
1209 | { | |
1210 | int i; | |
1211 | int rv; | |
1212 | int arrays_ok = 0; | |
1213 | int partitions_ok = 0; | |
56e8be85 | 1214 | int dfd = *dfdp; |
d2db3045 | 1215 | int bare; |
56e8be85 | 1216 | |
d2db3045 | 1217 | /* Can only add a spare if device has at least one domain */ |
56e8be85 N |
1218 | if (pol_find(pol, pol_domain) == NULL) |
1219 | return 1; | |
1220 | /* And only if some action allows spares */ | |
1221 | if (!policy_action_allows(pol, st?st->ss->name:NULL, act_spare)) | |
1222 | return 1; | |
1223 | ||
d2db3045 N |
1224 | /* Now check if the device is bare. |
1225 | * bare devices can always be added as a spare | |
1226 | * non-bare devices can only be added if spare-same-slot is permitted, | |
1227 | * and this device is replacing a previous device - in which case 'target' | |
1228 | * will be set. | |
56e8be85 | 1229 | */ |
52e965c2 | 1230 | if (!is_bare(dfd)) { |
d2db3045 N |
1231 | /* Must have a target and allow same_slot */ |
1232 | /* Later - may allow force_spare without target */ | |
1233 | if (!target || | |
1234 | !policy_action_allows(pol, st?st->ss->name:NULL, | |
1235 | act_spare_same_slot)) { | |
1236 | if (verbose > 1) | |
e7b84f9d | 1237 | pr_err("%s is not bare, so not " |
d2db3045 N |
1238 | "considering as a spare\n", |
1239 | devname); | |
1240 | return 1; | |
1241 | } | |
1242 | bare = 0; | |
1243 | } else | |
1244 | bare = 1; | |
56e8be85 | 1245 | |
d2db3045 N |
1246 | /* It might be OK to add this device to an array - need to see |
1247 | * what arrays might be candidates. | |
56e8be85 N |
1248 | */ |
1249 | if (st) { | |
1250 | /* just try try 'array' or 'partition' based on this metadata */ | |
1251 | if (st->ss->add_to_super) | |
d2db3045 | 1252 | return array_try_spare(devname, dfdp, pol, target, bare, |
56e8be85 N |
1253 | st, verbose); |
1254 | else | |
1255 | return partition_try_spare(devname, dfdp, pol, | |
1256 | st, verbose); | |
1257 | } | |
d2db3045 N |
1258 | /* No metadata was specified or found so options are open. |
1259 | * Check for whether any array metadata, or any partition metadata | |
1260 | * might allow adding the spare. This check is just help to avoid | |
1261 | * a more costly scan of all arrays when we can be sure that will | |
1262 | * fail. | |
1263 | */ | |
56e8be85 N |
1264 | for (i = 0; (!arrays_ok || !partitions_ok) && superlist[i] ; i++) { |
1265 | if (superlist[i]->add_to_super && !arrays_ok && | |
1266 | policy_action_allows(pol, superlist[i]->name, act_spare)) | |
1267 | arrays_ok = 1; | |
1268 | if (superlist[i]->add_to_super == NULL && !partitions_ok && | |
1269 | policy_action_allows(pol, superlist[i]->name, act_spare)) | |
1270 | partitions_ok = 1; | |
1271 | } | |
aaccda44 | 1272 | rv = 1; |
56e8be85 | 1273 | if (arrays_ok) |
d2db3045 N |
1274 | rv = array_try_spare(devname, dfdp, pol, target, bare, |
1275 | st, verbose); | |
aaccda44 | 1276 | if (rv != 0 && partitions_ok) |
56e8be85 N |
1277 | rv = partition_try_spare(devname, dfdp, pol, st, verbose); |
1278 | return rv; | |
1279 | } | |
1280 | ||
8382f19b NB |
1281 | int IncrementalScan(int verbose) |
1282 | { | |
1283 | /* look at every device listed in the 'map' file. | |
1284 | * If one is found that is not running then: | |
1285 | * look in mdadm.conf for bitmap file. | |
1286 | * if one exists, but array has none, add it. | |
1287 | * try to start array in auto-readonly mode | |
1288 | */ | |
1289 | struct map_ent *mapl = NULL; | |
1290 | struct map_ent *me; | |
fa56eddb | 1291 | struct mddev_ident *devs, *mddev; |
8382f19b NB |
1292 | int rv = 0; |
1293 | ||
1294 | map_read(&mapl); | |
1295 | devs = conf_get_ident(NULL); | |
1296 | ||
1297 | for (me = mapl ; me ; me = me->next) { | |
8382f19b NB |
1298 | mdu_array_info_t array; |
1299 | mdu_bitmap_file_t bmf; | |
7e0f6979 | 1300 | struct mdinfo *sra; |
4dd2df09 | 1301 | int mdfd = open_dev(me->devnm); |
215bb3f7 | 1302 | |
8382f19b NB |
1303 | if (mdfd < 0) |
1304 | continue; | |
1305 | if (ioctl(mdfd, GET_ARRAY_INFO, &array) == 0 || | |
1306 | errno != ENODEV) { | |
1307 | close(mdfd); | |
1308 | continue; | |
1309 | } | |
1310 | /* Ok, we can try this one. Maybe it needs a bitmap */ | |
1311 | for (mddev = devs ; mddev ; mddev = mddev->next) | |
339c2d6c | 1312 | if (mddev->devname && me->path |
2400e6eb | 1313 | && devname_matches(mddev->devname, me->path)) |
8382f19b NB |
1314 | break; |
1315 | if (mddev && mddev->bitmap_file) { | |
1316 | /* | |
1317 | * Note: early kernels will wrongly fail this, so it | |
1318 | * is a hint only | |
1319 | */ | |
1320 | int added = -1; | |
1321 | if (ioctl(mdfd, GET_ARRAY_INFO, &bmf) < 0) { | |
1322 | int bmfd = open(mddev->bitmap_file, O_RDWR); | |
1323 | if (bmfd >= 0) { | |
1324 | added = ioctl(mdfd, SET_BITMAP_FILE, | |
1325 | bmfd); | |
1326 | close(bmfd); | |
1327 | } | |
1328 | } | |
1329 | if (verbose >= 0) { | |
1330 | if (added == 0) | |
e7b84f9d N |
1331 | pr_err("Added bitmap %s to %s\n", |
1332 | mddev->bitmap_file, me->path); | |
8382f19b | 1333 | else if (errno != EEXIST) |
e7b84f9d N |
1334 | pr_err("Failed to add bitmap to %s: %s\n", |
1335 | me->path, strerror(errno)); | |
8382f19b NB |
1336 | } |
1337 | } | |
5e88ab2e N |
1338 | /* FIXME check for reshape_active and consider not |
1339 | * starting array. | |
1340 | */ | |
4dd2df09 | 1341 | sra = sysfs_read(mdfd, NULL, 0); |
8382f19b NB |
1342 | if (sra) { |
1343 | if (sysfs_set_str(sra, NULL, | |
1344 | "array_state", "read-auto") == 0) { | |
1345 | if (verbose >= 0) | |
e7b84f9d | 1346 | pr_err("started array %s\n", |
4dd2df09 | 1347 | me->path ?: me->devnm); |
8382f19b | 1348 | } else { |
e7b84f9d | 1349 | pr_err("failed to start array %s: %s\n", |
4dd2df09 | 1350 | me->path ?: me->devnm, |
e7b84f9d | 1351 | strerror(errno)); |
8382f19b NB |
1352 | rv = 1; |
1353 | } | |
b1efa6c2 | 1354 | sysfs_free(sra); |
8382f19b NB |
1355 | } |
1356 | } | |
1357 | return rv; | |
1358 | } | |
598f0d58 | 1359 | |
1771a6e2 N |
1360 | static char *container2devname(char *devname) |
1361 | { | |
1362 | char *mdname = NULL; | |
1363 | ||
1364 | if (devname[0] == '/') { | |
1365 | int fd = open(devname, O_RDONLY); | |
1366 | if (fd >= 0) { | |
4dd2df09 | 1367 | mdname = xstrdup(fd2devnm(fd)); |
1771a6e2 N |
1368 | close(fd); |
1369 | } | |
1370 | } else { | |
1371 | int uuid[4]; | |
1372 | struct map_ent *mp, *map = NULL; | |
ca3b6696 | 1373 | |
1771a6e2 N |
1374 | if (!parse_uuid(devname, uuid)) |
1375 | return mdname; | |
1376 | mp = map_by_uuid(&map, uuid); | |
1377 | if (mp) | |
4dd2df09 | 1378 | mdname = xstrdup(mp->devnm); |
1771a6e2 N |
1379 | map_free(map); |
1380 | } | |
1381 | ||
1382 | return mdname; | |
1383 | } | |
1384 | ||
47c74f3f | 1385 | static int Incremental_container(struct supertype *st, char *devname, |
11b6d91d | 1386 | struct context *c) |
598f0d58 NB |
1387 | { |
1388 | /* Collect the contents of this container and for each | |
1389 | * array, choose a device name and assemble the array. | |
1390 | */ | |
1391 | ||
47c74f3f | 1392 | struct mdinfo *list; |
598f0d58 | 1393 | struct mdinfo *ra; |
ad5bc697 | 1394 | struct map_ent *map = NULL; |
47c74f3f N |
1395 | struct mdinfo info; |
1396 | int trustworthy; | |
fa56eddb | 1397 | struct mddev_ident *match; |
47c74f3f | 1398 | int rv = 0; |
037e12d6 AC |
1399 | struct domainlist *domains; |
1400 | struct map_ent *smp; | |
1401 | int suuid[4]; | |
1402 | int sfd; | |
81219e70 LM |
1403 | int ra_blocked = 0; |
1404 | int ra_all = 0; | |
47c74f3f | 1405 | |
47c74f3f N |
1406 | st->ss->getinfo_super(st, &info, NULL); |
1407 | ||
11b6d91d | 1408 | if ((c->runstop > 0 && info.container_enough >= 0) || |
47c74f3f N |
1409 | info.container_enough > 0) |
1410 | /* pass */; | |
1411 | else { | |
11b6d91d | 1412 | if (c->verbose) |
e7b84f9d | 1413 | pr_err("not enough devices to start the container\n"); |
47c74f3f N |
1414 | return 0; |
1415 | } | |
1416 | ||
11b6d91d | 1417 | match = conf_match(st, &info, devname, c->verbose, &rv); |
47c74f3f N |
1418 | if (match == NULL && rv == 2) |
1419 | return rv; | |
1420 | ||
1421 | /* Need to compute 'trustworthy' */ | |
1422 | if (match) | |
1423 | trustworthy = LOCAL; | |
11b6d91d | 1424 | else if (st->ss->match_home(st, c->homehost) == 1) |
47c74f3f N |
1425 | trustworthy = LOCAL; |
1426 | else if (st->ss->match_home(st, "any") == 1) | |
1427 | trustworthy = LOCAL; | |
1428 | else | |
1429 | trustworthy = FOREIGN; | |
ad5bc697 | 1430 | |
47c74f3f | 1431 | list = st->ss->container_content(st, NULL); |
81219e70 LM |
1432 | /* when nothing to activate - quit */ |
1433 | if (list == NULL) | |
1434 | return 0; | |
598f0d58 | 1435 | for (ra = list ; ra ; ra = ra->next) { |
598f0d58 NB |
1436 | int mdfd; |
1437 | char chosen_name[1024]; | |
ad5bc697 | 1438 | struct map_ent *mp; |
fa56eddb | 1439 | struct mddev_ident *match = NULL; |
598f0d58 | 1440 | |
81219e70 LM |
1441 | ra_all++; |
1442 | /* do not activate arrays blocked by metadata handler */ | |
1443 | if (ra->array.state & (1 << MD_SB_BLOCK_VOLUME)) { | |
e7b84f9d | 1444 | pr_err("Cannot activate array %s in %s.\n", |
81219e70 LM |
1445 | ra->text_version, devname); |
1446 | ra_blocked++; | |
1447 | continue; | |
1448 | } | |
c5afc314 | 1449 | mp = map_by_uuid(&map, ra->uuid); |
d7288ddc | 1450 | |
30926600 | 1451 | if (mp) { |
4dd2df09 | 1452 | mdfd = open_dev(mp->devnm); |
339c2d6c N |
1453 | if (mp->path) |
1454 | strcpy(chosen_name, mp->path); | |
1455 | else | |
4dd2df09 | 1456 | strcpy(chosen_name, mp->devnm); |
30926600 | 1457 | } else { |
dbb44303 | 1458 | |
112cace6 | 1459 | /* Check in mdadm.conf for container == devname and |
dbb44303 N |
1460 | * member == ra->text_version after second slash. |
1461 | */ | |
1462 | char *sub = strchr(ra->text_version+1, '/'); | |
fa56eddb | 1463 | struct mddev_ident *array_list; |
dbb44303 N |
1464 | if (sub) { |
1465 | sub++; | |
1466 | array_list = conf_get_ident(NULL); | |
1467 | } else | |
1468 | array_list = NULL; | |
1469 | for(; array_list ; array_list = array_list->next) { | |
dbb44303 N |
1470 | char *dn; |
1471 | if (array_list->member == NULL || | |
1472 | array_list->container == NULL) | |
1473 | continue; | |
1474 | if (strcmp(array_list->member, sub) != 0) | |
1475 | continue; | |
71d60c48 DW |
1476 | if (array_list->uuid_set && |
1477 | !same_uuid(ra->uuid, array_list->uuid, st->ss->swapuuid)) | |
1478 | continue; | |
1771a6e2 N |
1479 | dn = container2devname(array_list->container); |
1480 | if (dn == NULL) | |
dbb44303 | 1481 | continue; |
dbb44303 N |
1482 | if (strncmp(dn, ra->text_version+1, |
1483 | strlen(dn)) != 0 || | |
1484 | ra->text_version[strlen(dn)+1] != '/') { | |
1485 | free(dn); | |
1486 | continue; | |
1487 | } | |
1488 | free(dn); | |
1489 | /* we have a match */ | |
1490 | match = array_list; | |
11b6d91d | 1491 | if (c->verbose>0) |
e7b84f9d | 1492 | pr_err("match found for member %s\n", |
71d60c48 | 1493 | array_list->member); |
dbb44303 N |
1494 | break; |
1495 | } | |
dbb44303 | 1496 | |
112cace6 N |
1497 | if (match && match->devname && |
1498 | strcasecmp(match->devname, "<ignore>") == 0) { | |
11b6d91d | 1499 | if (c->verbose > 0) |
e7b84f9d N |
1500 | pr_err("array %s/%s is " |
1501 | "explicitly ignored by mdadm.conf\n", | |
1502 | match->container, match->member); | |
112cace6 N |
1503 | return 2; |
1504 | } | |
05833051 N |
1505 | if (match) |
1506 | trustworthy = LOCAL; | |
112cace6 | 1507 | |
30926600 N |
1508 | mdfd = create_mddev(match ? match->devname : NULL, |
1509 | ra->name, | |
11b6d91d | 1510 | c->autof, |
30926600 N |
1511 | trustworthy, |
1512 | chosen_name); | |
1513 | } | |
598f0d58 NB |
1514 | |
1515 | if (mdfd < 0) { | |
e7b84f9d | 1516 | pr_err("failed to open %s: %s.\n", |
598f0d58 NB |
1517 | chosen_name, strerror(errno)); |
1518 | return 2; | |
1519 | } | |
1520 | ||
11b6d91d N |
1521 | assemble_container_content(st, mdfd, ra, c, |
1522 | chosen_name); | |
588bebfc | 1523 | close(mdfd); |
598f0d58 | 1524 | } |
037e12d6 | 1525 | |
81219e70 LM |
1526 | /* don't move spares to container with volume being activated |
1527 | when all volumes are blocked */ | |
ad098cdd | 1528 | if (ra_all == ra_blocked) |
81219e70 | 1529 | return 0; |
81219e70 | 1530 | |
037e12d6 AC |
1531 | /* Now move all suitable spares from spare container */ |
1532 | domains = domain_from_array(list, st->ss->name); | |
1533 | memcpy(suuid, uuid_zero, sizeof(int[4])); | |
1534 | if (domains && | |
1535 | (smp = map_by_uuid(&map, suuid)) != NULL && | |
1536 | (sfd = open(smp->path, O_RDONLY)) >= 0) { | |
1537 | /* spare container found */ | |
1538 | struct supertype *sst = | |
1539 | super_imsm.match_metadata_desc("imsm"); | |
1540 | struct mdinfo *sinfo; | |
1541 | unsigned long long min_size = 0; | |
1542 | if (st->ss->min_acceptable_spare_size) | |
1543 | min_size = st->ss->min_acceptable_spare_size(st); | |
1544 | if (!sst->ss->load_container(sst, sfd, NULL)) { | |
1545 | close(sfd); | |
1546 | sinfo = container_choose_spares(sst, min_size, | |
1547 | domains, NULL, | |
1548 | st->ss->name, 0); | |
1549 | sst->ss->free_super(sst); | |
1550 | if (sinfo){ | |
1551 | int count = 0; | |
1552 | struct mdinfo *disks = sinfo->devs; | |
1553 | while (disks) { | |
1554 | /* move spare from spare | |
1555 | * container to currently | |
1556 | * assembled one | |
1557 | */ | |
1558 | if (move_spare( | |
1559 | smp->path, | |
1560 | devname, | |
1561 | makedev(disks->disk.major, | |
1562 | disks->disk.minor))) | |
1563 | count++; | |
1564 | disks = disks->next; | |
1565 | } | |
1566 | if (count) | |
e7b84f9d N |
1567 | pr_err("Added %d spare%s to %s\n", |
1568 | count, count>1?"s":"", devname); | |
037e12d6 AC |
1569 | } |
1570 | sysfs_free(sinfo); | |
1571 | } else | |
1572 | close(sfd); | |
1573 | } | |
1574 | domain_free(domains); | |
598f0d58 NB |
1575 | return 0; |
1576 | } | |
29ba4804 | 1577 | |
6b63c1a4 N |
1578 | static void run_udisks(char *arg1, char *arg2) |
1579 | { | |
1580 | int pid = fork(); | |
1581 | int status; | |
1582 | if (pid == 0) { | |
1583 | execl("/usr/bin/udisks", "udisks", arg1, arg2, NULL); | |
1584 | execl("/bin/udisks", "udisks", arg1, arg2, NULL); | |
1585 | exit(1); | |
1586 | } | |
1587 | while (pid > 0 && wait(&status) != pid) | |
1588 | ; | |
1589 | } | |
1590 | ||
29ba4804 N |
1591 | /* |
1592 | * IncrementalRemove - Attempt to see if the passed in device belongs to any | |
1593 | * raid arrays, and if so first fail (if needed) and then remove the device. | |
1594 | * | |
1595 | * @devname - The device we want to remove | |
950bc344 | 1596 | * @id_path - name as found in /dev/disk/by-path for this device |
29ba4804 N |
1597 | * |
1598 | * Note: the device name must be a kernel name like "sda", so | |
1599 | * that we can find it in /proc/mdstat | |
1600 | */ | |
950bc344 | 1601 | int IncrementalRemove(char *devname, char *id_path, int verbose) |
29ba4804 N |
1602 | { |
1603 | int mdfd; | |
8af530b0 | 1604 | int rv = 0; |
29ba4804 | 1605 | struct mdstat_ent *ent; |
a655e550 | 1606 | struct mddev_dev devlist; |
8af530b0 N |
1607 | struct mdinfo mdi; |
1608 | char buf[32]; | |
29ba4804 | 1609 | |
950bc344 PC |
1610 | if (!id_path) |
1611 | dprintf(Name ": incremental removal without --path <id_path> " | |
1612 | "lacks the possibility to re-add new device in this " | |
1613 | "port\n"); | |
1614 | ||
29ba4804 | 1615 | if (strchr(devname, '/')) { |
e7b84f9d | 1616 | pr_err("incremental removal requires a " |
29ba4804 N |
1617 | "kernel device name, not a file: %s\n", devname); |
1618 | return 1; | |
1619 | } | |
1620 | ent = mdstat_by_component(devname); | |
1621 | if (!ent) { | |
041b815f N |
1622 | if (verbose >= 0) |
1623 | pr_err("%s does not appear to be a component " | |
1624 | "of any array\n", devname); | |
29ba4804 N |
1625 | return 1; |
1626 | } | |
8af530b0 N |
1627 | sysfs_init(&mdi, -1, ent->devnm); |
1628 | if (sysfs_get_str(&mdi, NULL, "array_state", | |
1629 | buf, sizeof(buf)) > 0) { | |
1630 | if (strncmp(buf, "active", 6) == 0 || | |
1631 | strncmp(buf, "clean", 5) == 0) | |
1632 | sysfs_set_str(&mdi, NULL, | |
1633 | "array_state", "read-auto"); | |
1634 | } | |
4dd2df09 | 1635 | mdfd = open_dev(ent->devnm); |
29ba4804 | 1636 | if (mdfd < 0) { |
041b815f N |
1637 | if (verbose >= 0) |
1638 | pr_err("Cannot open array %s!!\n", ent->dev); | |
08387a04 | 1639 | free_mdstat(ent); |
29ba4804 N |
1640 | return 1; |
1641 | } | |
403410eb PC |
1642 | |
1643 | if (id_path) { | |
1644 | struct map_ent *map = NULL, *me; | |
4dd2df09 | 1645 | me = map_by_devnm(&map, ent->devnm); |
403410eb PC |
1646 | if (me) |
1647 | policy_save_path(id_path, me); | |
1648 | map_free(map); | |
1649 | } | |
1650 | ||
29ba4804 N |
1651 | memset(&devlist, 0, sizeof(devlist)); |
1652 | devlist.devname = devname; | |
1653 | devlist.disposition = 'f'; | |
08387a04 N |
1654 | /* for a container, we must fail each member array */ |
1655 | if (ent->metadata_version && | |
1656 | strncmp(ent->metadata_version, "external:", 9) == 0) { | |
1657 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
1658 | struct mdstat_ent *memb; | |
1659 | for (memb = mdstat ; memb ; memb = memb->next) | |
1660 | if (is_container_member(memb, ent->dev)) { | |
4dd2df09 | 1661 | int subfd = open_dev(memb->devnm); |
08387a04 | 1662 | if (subfd >= 0) { |
8af530b0 N |
1663 | rv |= Manage_subdevs( |
1664 | memb->dev, subfd, | |
1665 | &devlist, verbose, 0, | |
1666 | NULL, 0); | |
08387a04 N |
1667 | close(subfd); |
1668 | } | |
1669 | } | |
1670 | free_mdstat(mdstat); | |
1671 | } else | |
8af530b0 N |
1672 | rv |= Manage_subdevs(ent->dev, mdfd, &devlist, |
1673 | verbose, 0, NULL, 0); | |
1674 | if (rv & 2) { | |
6b63c1a4 N |
1675 | /* Failed due to EBUSY, try to stop the array. |
1676 | * Give udisks a chance to unmount it first. | |
8af530b0 | 1677 | */ |
6b63c1a4 N |
1678 | int devid = devnm2devid(ent->devnm); |
1679 | run_udisks("--unmount", map_dev(major(devid),minor(devid), 0)); | |
fe7e0e64 | 1680 | rv = Manage_stop(ent->dev, mdfd, verbose, 1); |
8af530b0 N |
1681 | if (rv) |
1682 | /* At least we can try to trigger a 'remove' */ | |
1683 | sysfs_uevent(&mdi, "remove"); | |
6b63c1a4 N |
1684 | if (verbose) { |
1685 | if (rv) | |
1686 | pr_err("Fail to stop %s too.\n", ent->devnm); | |
1687 | } | |
8af530b0 N |
1688 | } else { |
1689 | devlist.disposition = 'r'; | |
1690 | rv = Manage_subdevs(ent->dev, mdfd, &devlist, | |
1691 | verbose, 0, NULL, 0); | |
1692 | } | |
aae3cdc3 | 1693 | close(mdfd); |
08387a04 | 1694 | free_mdstat(ent); |
aae3cdc3 | 1695 | return rv; |
29ba4804 | 1696 | } |