]>
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" | |
61018da0 N |
32 | #include <dirent.h> |
33 | #include <ctype.h> | |
8382f19b | 34 | |
2bf740b3 N |
35 | static int count_active(struct supertype *st, struct mdinfo *sra, |
36 | int mdfd, char **availp, | |
8382f19b | 37 | struct mdinfo *info); |
7e0f6979 | 38 | static void find_reject(int mdfd, struct supertype *st, struct mdinfo *sra, |
8382f19b NB |
39 | int number, __u64 events, int verbose, |
40 | char *array_name); | |
f08605b3 | 41 | static int try_spare(char *devname, int *dfdp, struct dev_policy *pol, |
d2db3045 | 42 | struct map_ent *target, |
f08605b3 | 43 | struct supertype *st, int verbose); |
8382f19b | 44 | |
7d91c3f5 | 45 | static int Incremental_container(struct supertype *st, char *devname, |
11b6d91d N |
46 | struct context *c); |
47 | ||
48 | int Incremental(char *devname, struct context *c, | |
49 | struct supertype *st) | |
8382f19b NB |
50 | { |
51 | /* Add this device to an array, creating the array if necessary | |
598f0d58 | 52 | * and starting the array if sensible or - if runstop>0 - if possible. |
8382f19b NB |
53 | * |
54 | * This has several steps: | |
55 | * | |
56 | * 1/ Check if device is permitted by mdadm.conf, reject if not. | |
57 | * 2/ Find metadata, reject if none appropriate (check | |
58 | * version/name from args) | |
59 | * 3/ Check if there is a match in mdadm.conf | |
215bb3f7 N |
60 | * 3a/ if not, check for homehost match. If no match, assemble as |
61 | * a 'foreign' array. | |
8382f19b NB |
62 | * 4/ Determine device number. |
63 | * - If in mdadm.conf with std name, use that | |
64 | * - UUID in /var/run/mdadm.map use that | |
65 | * - If name is suggestive, use that. unless in use with different uuid. | |
66 | * - Choose a free, high number. | |
67 | * - Use a partitioned device unless strong suggestion not to. | |
68 | * e.g. auto=md | |
352452c3 | 69 | * Don't choose partitioned for containers. |
8382f19b NB |
70 | * 5/ Find out if array already exists |
71 | * 5a/ if it does not | |
72 | * - choose a name, from mdadm.conf or 'name' field in array. | |
73 | * - create the array | |
74 | * - add the device | |
75 | * 5b/ if it does | |
76 | * - check one drive in array to make sure metadata is a reasonably | |
77 | * close match. Reject if not (e.g. different type) | |
78 | * - add the device | |
79 | * 6/ Make sure /var/run/mdadm.map contains this array. | |
80 | * 7/ Is there enough devices to possibly start the array? | |
352452c3 | 81 | * For a container, this means running Incremental_container. |
8382f19b NB |
82 | * 7a/ if not, finish with success. |
83 | * 7b/ if yes, | |
84 | * - read all metadata and arrange devices like -A does | |
85 | * - if number of OK devices match expected, or -R and there are enough, | |
86 | * start the array (auto-readonly). | |
87 | */ | |
88 | struct stat stb; | |
7e83544b | 89 | struct mdinfo info, dinfo; |
de6ae750 | 90 | struct mdinfo *sra = NULL, *d; |
fa56eddb | 91 | struct mddev_ident *match; |
8382f19b | 92 | char chosen_name[1024]; |
15d4a7e4 | 93 | int rv = 1; |
8382f19b | 94 | struct map_ent *mp, *map = NULL; |
15d4a7e4 | 95 | int dfd = -1, mdfd = -1; |
de6ae750 | 96 | char *avail = NULL; |
8382f19b | 97 | int active_disks; |
47c74f3f | 98 | int trustworthy; |
d7288ddc | 99 | char *name_to_use; |
ad5bc697 | 100 | mdu_array_info_t ainf; |
7e83544b | 101 | struct dev_policy *policy = NULL; |
d2db3045 N |
102 | struct map_ent target_array; |
103 | int have_target; | |
8382f19b NB |
104 | |
105 | struct createinfo *ci = conf_get_create_info(); | |
106 | ||
47c74f3f | 107 | if (stat(devname, &stb) < 0) { |
11b6d91d | 108 | if (c->verbose >= 0) |
e7b84f9d | 109 | pr_err("stat failed for %s: %s.\n", |
47c74f3f N |
110 | devname, strerror(errno)); |
111 | return rv; | |
112 | } | |
113 | if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
11b6d91d | 114 | if (c->verbose >= 0) |
e7b84f9d | 115 | pr_err("%s is not a block device.\n", |
47c74f3f N |
116 | devname); |
117 | return rv; | |
118 | } | |
0431869c | 119 | dfd = dev_open(devname, O_RDONLY); |
47c74f3f | 120 | if (dfd < 0) { |
11b6d91d | 121 | if (c->verbose >= 0) |
e7b84f9d | 122 | pr_err("cannot open %s: %s.\n", |
47c74f3f N |
123 | devname, strerror(errno)); |
124 | return rv; | |
125 | } | |
126 | /* If the device is a container, we do something very different */ | |
3a371610 | 127 | if (must_be_container(dfd)) { |
47c74f3f N |
128 | if (!st) |
129 | st = super_by_fd(dfd, NULL); | |
cb8f6859 N |
130 | if (st) |
131 | st->ignore_hw_compat = 1; | |
a92b2112 | 132 | if (st && st->ss->load_container) |
47c74f3f N |
133 | rv = st->ss->load_container(st, dfd, NULL); |
134 | ||
135 | close(dfd); | |
5fc8cff3 JS |
136 | if (!rv && st->ss->container_content) { |
137 | if (map_lock(&map)) | |
e7b84f9d N |
138 | pr_err("failed to get " |
139 | "exclusive lock on mapfile\n"); | |
11b6d91d | 140 | rv = Incremental_container(st, devname, c); |
5fc8cff3 JS |
141 | map_unlock(&map); |
142 | return rv; | |
143 | } | |
47c74f3f | 144 | |
e7b84f9d | 145 | pr_err("%s is not part of an md array.\n", |
47c74f3f N |
146 | devname); |
147 | return rv; | |
148 | } | |
8382f19b | 149 | |
352452c3 | 150 | /* 1/ Check if device is permitted by mdadm.conf */ |
8382f19b NB |
151 | |
152 | if (!conf_test_dev(devname)) { | |
11b6d91d | 153 | if (c->verbose >= 0) |
e7b84f9d N |
154 | pr_err("%s not permitted by mdadm.conf.\n", |
155 | devname); | |
15d4a7e4 | 156 | goto out; |
8382f19b NB |
157 | } |
158 | ||
159 | /* 2/ Find metadata, reject if none appropriate (check | |
160 | * version/name from args) */ | |
161 | ||
8382f19b | 162 | if (fstat(dfd, &stb) < 0) { |
11b6d91d | 163 | if (c->verbose >= 0) |
e7b84f9d | 164 | pr_err("fstat failed for %s: %s.\n", |
8382f19b | 165 | devname, strerror(errno)); |
15d4a7e4 | 166 | goto out; |
8382f19b NB |
167 | } |
168 | if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
11b6d91d | 169 | if (c->verbose >= 0) |
e7b84f9d | 170 | pr_err("%s is not a block device.\n", |
8382f19b | 171 | devname); |
15d4a7e4 | 172 | goto out; |
8382f19b NB |
173 | } |
174 | ||
7e83544b N |
175 | dinfo.disk.major = major(stb.st_rdev); |
176 | dinfo.disk.minor = minor(stb.st_rdev); | |
177 | ||
178 | policy = disk_policy(&dinfo); | |
d2db3045 | 179 | have_target = policy_check_path(&dinfo, &target_array); |
7e83544b | 180 | |
8382f19b | 181 | if (st == NULL && (st = guess_super(dfd)) == NULL) { |
11b6d91d | 182 | if (c->verbose >= 0) |
e7b84f9d N |
183 | pr_err("no recognisable superblock on %s.\n", |
184 | devname); | |
d2db3045 N |
185 | rv = try_spare(devname, &dfd, policy, |
186 | have_target ? &target_array : NULL, | |
11b6d91d | 187 | st, c->verbose); |
15d4a7e4 | 188 | goto out; |
8382f19b | 189 | } |
cb8f6859 | 190 | st->ignore_hw_compat = 1; |
0f22b998 N |
191 | if (st->ss->compare_super == NULL || |
192 | st->ss->load_super(st, dfd, NULL)) { | |
11b6d91d | 193 | if (c->verbose >= 0) |
e7b84f9d | 194 | pr_err("no RAID superblock on %s.\n", |
8382f19b | 195 | devname); |
d2db3045 N |
196 | rv = try_spare(devname, &dfd, policy, |
197 | have_target ? &target_array : NULL, | |
11b6d91d | 198 | st, c->verbose); |
0f22b998 | 199 | free(st); |
15d4a7e4 | 200 | goto out; |
8382f19b | 201 | } |
15d4a7e4 | 202 | close (dfd); dfd = -1; |
8382f19b | 203 | |
a5d85af7 | 204 | st->ss->getinfo_super(st, &info, NULL); |
8382f19b | 205 | |
3a97f210 | 206 | /* 3/ Check if there is a match in mdadm.conf */ |
11b6d91d | 207 | match = conf_match(st, &info, devname, c->verbose, &rv); |
3a97f210 N |
208 | if (!match && rv == 2) |
209 | goto out; | |
8382f19b | 210 | |
112cace6 N |
211 | if (match && match->devname |
212 | && strcasecmp(match->devname, "<ignore>") == 0) { | |
11b6d91d | 213 | if (c->verbose >= 0) |
e7b84f9d | 214 | pr_err("array containing %s is explicitly" |
112cace6 N |
215 | " ignored by mdadm.conf\n", |
216 | devname); | |
15d4a7e4 | 217 | goto out; |
112cace6 N |
218 | } |
219 | ||
7b403fef N |
220 | /* 3a/ if not, check for homehost match. If no match, continue |
221 | * but don't trust the 'name' in the array. Thus a 'random' minor | |
222 | * number will be assigned, and the device name will be based | |
223 | * on that. */ | |
215bb3f7 N |
224 | if (match) |
225 | trustworthy = LOCAL; | |
11b6d91d | 226 | else if (st->ss->match_home(st, c->homehost) == 1) |
d1d3482b N |
227 | trustworthy = LOCAL; |
228 | else if (st->ss->match_home(st, "any") == 1) | |
229 | trustworthy = LOCAL_ANY; | |
215bb3f7 | 230 | else |
d1d3482b N |
231 | trustworthy = FOREIGN; |
232 | ||
4e8d9f0a | 233 | if (!match && !conf_test_metadata(st->ss->name, policy, |
d1d3482b | 234 | (trustworthy == LOCAL))) { |
11b6d91d | 235 | if (c->verbose >= 1) |
e7b84f9d N |
236 | pr_err("%s has metadata type %s for which " |
237 | "auto-assembly is disabled\n", | |
238 | devname, st->ss->name); | |
15d4a7e4 | 239 | goto out; |
d1d3482b N |
240 | } |
241 | if (trustworthy == LOCAL_ANY) | |
ecb02e31 | 242 | trustworthy = LOCAL; |
e12e1d1d | 243 | |
4ef2f11e N |
244 | /* There are three possible sources for 'autof': command line, |
245 | * ARRAY line in mdadm.conf, or CREATE line in mdadm.conf. | |
350ac35d N |
246 | * ARRAY takes precedence, then command line, then |
247 | * CREATE. | |
4ef2f11e | 248 | */ |
350ac35d | 249 | if (match && match->autof) |
11b6d91d N |
250 | c->autof = match->autof; |
251 | if (c->autof == 0) | |
252 | c->autof = ci->autof; | |
4ef2f11e | 253 | |
7cdc0872 | 254 | name_to_use = info.name; |
0ac91628 | 255 | if (name_to_use[0] == 0 && |
0c4304ca | 256 | info.array.level == LEVEL_CONTAINER) { |
ecb02e31 N |
257 | name_to_use = info.text_version; |
258 | trustworthy = METADATA; | |
259 | } | |
0ac91628 | 260 | if (name_to_use[0] && trustworthy != LOCAL && |
11b6d91d | 261 | ! c->require_homehost && |
0ac91628 N |
262 | conf_name_is_free(name_to_use)) |
263 | trustworthy = LOCAL; | |
ecb02e31 | 264 | |
7cdc0872 N |
265 | /* strip "hostname:" prefix from name if we have decided |
266 | * to treat it as LOCAL | |
267 | */ | |
268 | if (trustworthy == LOCAL && strchr(name_to_use, ':') != NULL) | |
269 | name_to_use = strchr(name_to_use, ':')+1; | |
270 | ||
ad5bc697 | 271 | /* 4/ Check if array exists. |
215bb3f7 | 272 | */ |
93c861ee | 273 | if (map_lock(&map)) |
e7b84f9d | 274 | pr_err("failed to get exclusive lock on " |
93c861ee | 275 | "mapfile\n"); |
0431869c N |
276 | /* Now check we can get O_EXCL. If not, probably "mdadm -A" has |
277 | * taken over | |
278 | */ | |
279 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
280 | if (dfd < 0) { | |
281 | if (c->verbose >= 0) | |
282 | pr_err("cannot reopen %s: %s.\n", | |
283 | devname, strerror(errno)); | |
284 | goto out_unlock; | |
285 | } | |
286 | /* Cannot hold it open while we add the device to the array, | |
287 | * so we must release the O_EXCL and depend on the map_lock() | |
288 | */ | |
289 | close(dfd); | |
290 | dfd = -1; | |
291 | ||
215bb3f7 | 292 | mp = map_by_uuid(&map, info.uuid); |
339c2d6c N |
293 | if (mp) |
294 | mdfd = open_dev(mp->devnum); | |
295 | else | |
215bb3f7 | 296 | mdfd = -1; |
d7288ddc | 297 | |
8382f19b | 298 | if (mdfd < 0) { |
8382f19b | 299 | |
215bb3f7 | 300 | /* Couldn't find an existing array, maybe make a new one */ |
ffb3e494 | 301 | mdfd = create_mddev(match ? match->devname : NULL, |
11b6d91d | 302 | name_to_use, c->autof, trustworthy, chosen_name); |
215bb3f7 N |
303 | |
304 | if (mdfd < 0) | |
25824e2d | 305 | goto out_unlock; |
215bb3f7 N |
306 | |
307 | sysfs_init(&info, mdfd, 0); | |
308 | ||
f35f2525 | 309 | if (set_array_info(mdfd, st, &info) != 0) { |
e7b84f9d | 310 | pr_err("failed to set array info for %s: %s\n", |
8382f19b | 311 | chosen_name, strerror(errno)); |
15d4a7e4 | 312 | rv = 2; |
25824e2d | 313 | goto out_unlock; |
8382f19b | 314 | } |
7801ac20 | 315 | |
c5afc314 N |
316 | dinfo = info; |
317 | dinfo.disk.major = major(stb.st_rdev); | |
318 | dinfo.disk.minor = minor(stb.st_rdev); | |
319 | if (add_disk(mdfd, st, &info, &dinfo) != 0) { | |
75a410f6 | 320 | pr_err("failed to add %s to new array %s: %s.\n", |
8382f19b NB |
321 | devname, chosen_name, strerror(errno)); |
322 | ioctl(mdfd, STOP_ARRAY, 0); | |
15d4a7e4 | 323 | rv = 2; |
25824e2d | 324 | goto out_unlock; |
8382f19b | 325 | } |
2bf740b3 N |
326 | sra = sysfs_read(mdfd, -1, (GET_DEVS | GET_STATE | |
327 | GET_OFFSET | GET_SIZE)); | |
ca3b6696 | 328 | |
06c7f68e | 329 | if (!sra || !sra->devs || sra->devs->disk.raid_disk >= 0) { |
8382f19b NB |
330 | /* It really should be 'none' - must be old buggy |
331 | * kernel, and mdadm -I may not be able to complete. | |
332 | * So reject it. | |
333 | */ | |
334 | ioctl(mdfd, STOP_ARRAY, NULL); | |
e7b84f9d N |
335 | pr_err("You have an old buggy kernel which cannot support\n" |
336 | " --incremental reliably. Aborting.\n"); | |
15d4a7e4 | 337 | rv = 2; |
25824e2d | 338 | goto out_unlock; |
8382f19b | 339 | } |
c5afc314 | 340 | info.array.working_disks = 1; |
ad5bc697 N |
341 | /* 6/ Make sure /var/run/mdadm.map contains this array. */ |
342 | map_update(&map, fd2devnum(mdfd), | |
343 | info.text_version, | |
344 | info.uuid, chosen_name); | |
8382f19b NB |
345 | } else { |
346 | /* 5b/ if it does */ | |
347 | /* - check one drive in array to make sure metadata is a reasonably */ | |
348 | /* close match. Reject if not (e.g. different type) */ | |
349 | /* - add the device */ | |
350 | char dn[20]; | |
351 | int dfd2; | |
8382f19b | 352 | int err; |
3da92f27 | 353 | struct supertype *st2; |
c5afc314 | 354 | struct mdinfo info2, *d; |
215bb3f7 | 355 | |
2bf740b3 N |
356 | sra = sysfs_read(mdfd, -1, (GET_DEVS | GET_STATE | |
357 | GET_OFFSET | GET_SIZE)); | |
ca3b6696 | 358 | |
339c2d6c N |
359 | if (mp->path) |
360 | strcpy(chosen_name, mp->path); | |
361 | else | |
362 | strcpy(chosen_name, devnum2devname(mp->devnum)); | |
215bb3f7 | 363 | |
ef83fe7c N |
364 | /* It is generally not OK to add non-spare drives to a |
365 | * running array as they are probably missing because | |
366 | * they failed. However if runstop is 1, then the | |
7e83544b N |
367 | * array was possibly started early and our best bet is |
368 | * to add this anyway. | |
369 | * Also if action policy is re-add or better we allow | |
e5e5d7ce N |
370 | * re-add. |
371 | * This doesn't apply to containers as the 'non-spare' | |
372 | * flag has a different meaning. The test has to happen | |
373 | * at the device level there | |
3a6ec29a | 374 | */ |
e5e5d7ce N |
375 | if (!st->ss->external |
376 | && (info.disk.state & (1<<MD_DISK_SYNC)) != 0 | |
7e83544b N |
377 | && ! policy_action_allows(policy, st->ss->name, |
378 | act_re_add) | |
11b6d91d | 379 | && c->runstop < 1) { |
e5e5d7ce | 380 | if (ioctl(mdfd, GET_ARRAY_INFO, &ainf) == 0) { |
e7b84f9d N |
381 | pr_err("not adding %s to active array (without --run) %s\n", |
382 | devname, chosen_name); | |
15d4a7e4 | 383 | rv = 2; |
25824e2d | 384 | goto out_unlock; |
3a6ec29a N |
385 | } |
386 | } | |
15d4a7e4 N |
387 | if (!sra) { |
388 | rv = 2; | |
25824e2d | 389 | goto out_unlock; |
15d4a7e4 | 390 | } |
7c548327 N |
391 | if (sra->devs) { |
392 | sprintf(dn, "%d:%d", sra->devs->disk.major, | |
393 | sra->devs->disk.minor); | |
394 | dfd2 = dev_open(dn, O_RDONLY); | |
fb745c4b | 395 | if (dfd2 < 0) { |
e7b84f9d | 396 | pr_err("unable to open %s\n", devname); |
fb745c4b JS |
397 | rv = 2; |
398 | goto out_unlock; | |
399 | } | |
7c548327 N |
400 | st2 = dup_super(st); |
401 | if (st2->ss->load_super(st2, dfd2, NULL) || | |
402 | st->ss->compare_super(st, st2) != 0) { | |
e7b84f9d N |
403 | pr_err("metadata mismatch between %s and " |
404 | "chosen array %s\n", | |
405 | devname, chosen_name); | |
7c548327 | 406 | close(dfd2); |
15d4a7e4 | 407 | rv = 2; |
25824e2d | 408 | goto out_unlock; |
7c548327 | 409 | } |
8382f19b | 410 | close(dfd2); |
a5d85af7 | 411 | st2->ss->getinfo_super(st2, &info2, NULL); |
7c548327 N |
412 | st2->ss->free_super(st2); |
413 | if (info.array.level != info2.array.level || | |
414 | memcmp(info.uuid, info2.uuid, 16) != 0 || | |
415 | info.array.raid_disks != info2.array.raid_disks) { | |
e7b84f9d N |
416 | pr_err("unexpected difference between %s and %s.\n", |
417 | chosen_name, devname); | |
15d4a7e4 | 418 | rv = 2; |
25824e2d | 419 | goto out_unlock; |
7c548327 | 420 | } |
8382f19b | 421 | } |
69fe207e N |
422 | info.disk.major = major(stb.st_rdev); |
423 | info.disk.minor = minor(stb.st_rdev); | |
c5afc314 N |
424 | /* add disk needs to know about containers */ |
425 | if (st->ss->external) | |
426 | sra->array.level = LEVEL_CONTAINER; | |
69fe207e | 427 | err = add_disk(mdfd, st, sra, &info); |
8382f19b NB |
428 | if (err < 0 && errno == EBUSY) { |
429 | /* could be another device present with the same | |
430 | * disk.number. Find and reject any such | |
431 | */ | |
432 | find_reject(mdfd, st, sra, info.disk.number, | |
11b6d91d | 433 | info.events, c->verbose, chosen_name); |
69fe207e | 434 | err = add_disk(mdfd, st, sra, &info); |
8382f19b | 435 | } |
75a410f6 N |
436 | if (err < 0 && errno == EINVAL && |
437 | info.disk.state & (1<<MD_DISK_SYNC)) { | |
438 | /* Maybe it needs to be added as a spare */ | |
439 | if (policy_action_allows(policy, st->ss->name, | |
440 | act_force_spare)) { | |
441 | info.disk.state &= ~(1<<MD_DISK_SYNC); | |
442 | err = add_disk(mdfd, st, sra, &info); | |
443 | } else | |
444 | if (c->verbose >= 0) | |
445 | pr_err("can only add %s to %s as a spare, and force-spare is not set.\n", | |
446 | devname, chosen_name); | |
447 | } | |
8382f19b | 448 | if (err < 0) { |
75a410f6 | 449 | pr_err("failed to add %s to existing array %s: %s.\n", |
8382f19b | 450 | devname, chosen_name, strerror(errno)); |
15d4a7e4 | 451 | rv = 2; |
25824e2d | 452 | goto out_unlock; |
8382f19b | 453 | } |
c5afc314 N |
454 | info.array.working_disks = 0; |
455 | for (d = sra->devs; d; d=d->next) | |
456 | info.array.working_disks ++; | |
ca3b6696 | 457 | |
8382f19b | 458 | } |
8382f19b NB |
459 | |
460 | /* 7/ Is there enough devices to possibly start the array? */ | |
461 | /* 7a/ if not, finish with success. */ | |
352452c3 | 462 | if (info.array.level == LEVEL_CONTAINER) { |
7b0bbd0f | 463 | int devnum = devnum; /* defined and used iff ->external */ |
352452c3 | 464 | /* Try to assemble within the container */ |
a408d66c | 465 | sysfs_uevent(sra, "change"); |
11b6d91d | 466 | if (c->verbose >= 0) |
e7b84f9d N |
467 | pr_err("container %s now has %d device%s\n", |
468 | chosen_name, info.array.working_disks, | |
ca3b6696 | 469 | info.array.working_disks == 1?"":"s"); |
a7c6e3fb | 470 | wait_for(chosen_name, mdfd); |
ed690d36 | 471 | if (st->ss->external) |
983fff45 | 472 | devnum = fd2devnum(mdfd); |
5fc8cff3 JS |
473 | if (st->ss->load_container) |
474 | rv = st->ss->load_container(st, mdfd, NULL); | |
a7c6e3fb | 475 | close(mdfd); |
2bf740b3 | 476 | sysfs_free(sra); |
5fc8cff3 | 477 | if (!rv) |
11b6d91d | 478 | rv = Incremental_container(st, chosen_name, c); |
5fc8cff3 | 479 | map_unlock(&map); |
03b7f6c6 N |
480 | if (rv == 1) |
481 | /* Don't fail the whole -I if a subarray didn't | |
482 | * have enough devices to start yet | |
483 | */ | |
484 | rv = 0; | |
ed690d36 PC |
485 | /* after spare is added, ping monitor for external metadata |
486 | * so that it can eg. try to rebuild degraded array */ | |
983fff45 AK |
487 | if (st->ss->external) |
488 | ping_monitor_by_id(devnum); | |
03b7f6c6 | 489 | return rv; |
352452c3 | 490 | } |
de6ae750 | 491 | |
2bf740b3 N |
492 | /* We have added something to the array, so need to re-read the |
493 | * state. Eventually this state should be kept up-to-date as | |
494 | * things change. | |
495 | */ | |
496 | sysfs_free(sra); | |
497 | sra = sysfs_read(mdfd, -1, (GET_DEVS | GET_STATE | | |
498 | GET_OFFSET | GET_SIZE)); | |
499 | active_disks = count_active(st, sra, mdfd, &avail, &info); | |
8382f19b NB |
500 | if (enough(info.array.level, info.array.raid_disks, |
501 | info.array.layout, info.array.state & 1, | |
de5a472e | 502 | avail) == 0) { |
11b6d91d | 503 | if (c->verbose >= 0) |
e7b84f9d N |
504 | pr_err("%s attached to %s, not enough to start (%d).\n", |
505 | devname, chosen_name, active_disks); | |
15d4a7e4 | 506 | rv = 0; |
25824e2d | 507 | goto out_unlock; |
8382f19b | 508 | } |
8382f19b NB |
509 | |
510 | /* 7b/ if yes, */ | |
511 | /* - if number of OK devices match expected, or -R and there */ | |
512 | /* are enough, */ | |
513 | /* + add any bitmap file */ | |
514 | /* + start the array (auto-readonly). */ | |
8382f19b NB |
515 | |
516 | if (ioctl(mdfd, GET_ARRAY_INFO, &ainf) == 0) { | |
11b6d91d | 517 | if (c->verbose >= 0) |
e7b84f9d N |
518 | pr_err("%s attached to %s which is already active.\n", |
519 | devname, chosen_name); | |
15d4a7e4 | 520 | rv = 0; |
25824e2d | 521 | goto out_unlock; |
8382f19b | 522 | } |
ad5bc697 N |
523 | |
524 | map_unlock(&map); | |
11b6d91d | 525 | if (c->runstop > 0 || active_disks >= info.array.working_disks) { |
2bf740b3 | 526 | struct mdinfo *dsk; |
8382f19b | 527 | /* Let's try to start it */ |
5e88ab2e N |
528 | |
529 | if (info.reshape_active && !(info.reshape_active & RESHAPE_NO_BACKUP)) { | |
530 | fprintf(stderr, Name | |
531 | ": %s: This array is being reshaped and cannot be started\n" | |
532 | " by --incremental. Please use --assemble\n", | |
533 | chosen_name); | |
534 | goto out; | |
535 | } | |
8382f19b NB |
536 | if (match && match->bitmap_file) { |
537 | int bmfd = open(match->bitmap_file, O_RDWR); | |
538 | if (bmfd < 0) { | |
e7b84f9d N |
539 | pr_err("Could not open bitmap file %s.\n", |
540 | match->bitmap_file); | |
15d4a7e4 | 541 | goto out; |
8382f19b NB |
542 | } |
543 | if (ioctl(mdfd, SET_BITMAP_FILE, bmfd) != 0) { | |
544 | close(bmfd); | |
e7b84f9d N |
545 | pr_err("Failed to set bitmapfile for %s.\n", |
546 | chosen_name); | |
15d4a7e4 | 547 | goto out; |
8382f19b NB |
548 | } |
549 | close(bmfd); | |
550 | } | |
de6ae750 N |
551 | /* Need to remove from the array any devices which |
552 | * 'count_active' discerned were too old or inappropriate | |
553 | */ | |
554 | for (d = sra ? sra->devs : NULL ; d ; d = d->next) | |
555 | if (d->disk.state & (1<<MD_DISK_REMOVED)) | |
556 | remove_disk(mdfd, st, sra, d); | |
2bf740b3 | 557 | |
7b403fef | 558 | if ((sra == NULL || active_disks >= info.array.working_disks) |
215bb3f7 | 559 | && trustworthy != FOREIGN) |
8382f19b NB |
560 | rv = ioctl(mdfd, RUN_ARRAY, NULL); |
561 | else | |
562 | rv = sysfs_set_str(sra, NULL, | |
563 | "array_state", "read-auto"); | |
564 | if (rv == 0) { | |
11b6d91d | 565 | if (c->verbose >= 0) |
e7b84f9d N |
566 | pr_err("%s attached to %s, which has been started.\n", |
567 | devname, chosen_name); | |
8382f19b | 568 | rv = 0; |
a7c6e3fb | 569 | wait_for(chosen_name, mdfd); |
7e83544b N |
570 | /* We just started the array, so some devices |
571 | * might have been evicted from the array | |
572 | * because their event counts were too old. | |
573 | * If the action=re-add policy is in-force for | |
574 | * those devices we should re-add them now. | |
575 | */ | |
576 | for (dsk = sra->devs; dsk ; dsk = dsk->next) { | |
577 | if (disk_action_allows(dsk, st->ss->name, act_re_add) && | |
578 | add_disk(mdfd, st, sra, dsk) == 0) | |
e7b84f9d N |
579 | pr_err("%s re-added to %s\n", |
580 | dsk->sys_name, chosen_name); | |
7e83544b | 581 | } |
8382f19b | 582 | } else { |
e7b84f9d N |
583 | pr_err("%s attached to %s, but failed to start: %s.\n", |
584 | devname, chosen_name, strerror(errno)); | |
8382f19b NB |
585 | rv = 1; |
586 | } | |
587 | } else { | |
11b6d91d | 588 | if (c->verbose >= 0) |
e7b84f9d N |
589 | pr_err("%s attached to %s, not enough to start safely.\n", |
590 | devname, chosen_name); | |
8382f19b NB |
591 | rv = 0; |
592 | } | |
15d4a7e4 | 593 | out: |
de6ae750 | 594 | free(avail); |
15d4a7e4 N |
595 | if (dfd >= 0) |
596 | close(dfd); | |
597 | if (mdfd >= 0) | |
598 | close(mdfd); | |
7e83544b N |
599 | if (policy) |
600 | dev_policy_free(policy); | |
2bf740b3 N |
601 | if (sra) |
602 | sysfs_free(sra); | |
8382f19b | 603 | return rv; |
25824e2d JS |
604 | out_unlock: |
605 | map_unlock(&map); | |
606 | goto out; | |
8382f19b NB |
607 | } |
608 | ||
7e0f6979 | 609 | static void find_reject(int mdfd, struct supertype *st, struct mdinfo *sra, |
8382f19b NB |
610 | int number, __u64 events, int verbose, |
611 | char *array_name) | |
612 | { | |
3da92f27 | 613 | /* Find a device attached to this array with a disk.number of number |
8382f19b NB |
614 | * and events less than the passed events, and remove the device. |
615 | */ | |
06c7f68e | 616 | struct mdinfo *d; |
8382f19b NB |
617 | mdu_array_info_t ra; |
618 | ||
619 | if (ioctl(mdfd, GET_ARRAY_INFO, &ra) == 0) | |
620 | return; /* not safe to remove from active arrays | |
621 | * without thinking more */ | |
622 | ||
623 | for (d = sra->devs; d ; d = d->next) { | |
624 | char dn[10]; | |
625 | int dfd; | |
8382f19b | 626 | struct mdinfo info; |
06c7f68e | 627 | sprintf(dn, "%d:%d", d->disk.major, d->disk.minor); |
8382f19b NB |
628 | dfd = dev_open(dn, O_RDONLY); |
629 | if (dfd < 0) | |
630 | continue; | |
3da92f27 | 631 | if (st->ss->load_super(st, dfd, NULL)) { |
8382f19b NB |
632 | close(dfd); |
633 | continue; | |
634 | } | |
a5d85af7 | 635 | st->ss->getinfo_super(st, &info, NULL); |
3da92f27 | 636 | st->ss->free_super(st); |
8382f19b NB |
637 | close(dfd); |
638 | ||
639 | if (info.disk.number != number || | |
640 | info.events >= events) | |
641 | continue; | |
642 | ||
06c7f68e | 643 | if (d->disk.raid_disk > -1) |
8382f19b NB |
644 | sysfs_set_str(sra, d, "slot", "none"); |
645 | if (sysfs_set_str(sra, d, "state", "remove") == 0) | |
646 | if (verbose >= 0) | |
e7b84f9d N |
647 | pr_err("removing old device %s from %s\n", |
648 | d->sys_name+4, array_name); | |
8382f19b NB |
649 | } |
650 | } | |
651 | ||
2bf740b3 N |
652 | static int count_active(struct supertype *st, struct mdinfo *sra, |
653 | int mdfd, char **availp, | |
8382f19b NB |
654 | struct mdinfo *bestinfo) |
655 | { | |
656 | /* count how many devices in sra think they are active */ | |
06c7f68e | 657 | struct mdinfo *d; |
de6ae750 | 658 | int cnt = 0; |
72e7fb13 | 659 | int replcnt = 0; |
8382f19b | 660 | __u64 max_events = 0; |
8382f19b | 661 | char *avail = NULL; |
e4c72d1d | 662 | int *best = NULL; |
de6ae750 N |
663 | char *devmap = NULL; |
664 | int numdevs = 0; | |
665 | int devnum; | |
666 | int b, i; | |
667 | int raid_disks = 0; | |
8382f19b | 668 | |
b526e52d DW |
669 | if (!sra) |
670 | return 0; | |
671 | ||
de6ae750 N |
672 | for (d = sra->devs ; d ; d = d->next) |
673 | numdevs++; | |
ca3b6696 | 674 | for (d = sra->devs, devnum = 0 ; d ; d = d->next, devnum++) { |
8382f19b NB |
675 | char dn[30]; |
676 | int dfd; | |
8382f19b NB |
677 | int ok; |
678 | struct mdinfo info; | |
679 | ||
06c7f68e | 680 | sprintf(dn, "%d:%d", d->disk.major, d->disk.minor); |
8382f19b NB |
681 | dfd = dev_open(dn, O_RDONLY); |
682 | if (dfd < 0) | |
683 | continue; | |
3da92f27 | 684 | ok = st->ss->load_super(st, dfd, NULL); |
8382f19b NB |
685 | close(dfd); |
686 | if (ok != 0) | |
687 | continue; | |
de6ae750 N |
688 | info.array.raid_disks = raid_disks; |
689 | st->ss->getinfo_super(st, &info, devmap + raid_disks * devnum); | |
43aaf431 | 690 | if (!avail) { |
de6ae750 | 691 | raid_disks = info.array.raid_disks; |
503975b9 | 692 | avail = xcalloc(raid_disks, 1); |
43aaf431 | 693 | *availp = avail; |
de6ae750 | 694 | |
503975b9 N |
695 | best = xcalloc(raid_disks, sizeof(int)); |
696 | devmap = xcalloc(raid_disks, numdevs); | |
de6ae750 N |
697 | |
698 | st->ss->getinfo_super(st, &info, devmap); | |
43aaf431 DL |
699 | } |
700 | ||
8382f19b NB |
701 | if (info.disk.state & (1<<MD_DISK_SYNC)) |
702 | { | |
8382f19b NB |
703 | if (cnt == 0) { |
704 | cnt++; | |
705 | max_events = info.events; | |
706 | avail[info.disk.raid_disk] = 2; | |
de6ae750 | 707 | best[info.disk.raid_disk] = devnum; |
a5d85af7 | 708 | st->ss->getinfo_super(st, bestinfo, NULL); |
8382f19b | 709 | } else if (info.events == max_events) { |
8382f19b | 710 | avail[info.disk.raid_disk] = 2; |
de6ae750 | 711 | best[info.disk.raid_disk] = devnum; |
8382f19b | 712 | } else if (info.events == max_events-1) { |
de6ae750 N |
713 | if (avail[info.disk.raid_disk] == 0) { |
714 | avail[info.disk.raid_disk] = 1; | |
715 | best[info.disk.raid_disk] = devnum; | |
716 | } | |
8382f19b NB |
717 | } else if (info.events < max_events - 1) |
718 | ; | |
719 | else if (info.events == max_events+1) { | |
720 | int i; | |
8382f19b | 721 | max_events = info.events; |
ca3b6696 | 722 | for (i = 0; i < raid_disks; i++) |
8382f19b NB |
723 | if (avail[i]) |
724 | avail[i]--; | |
725 | avail[info.disk.raid_disk] = 2; | |
de6ae750 | 726 | best[info.disk.raid_disk] = devnum; |
a5d85af7 | 727 | st->ss->getinfo_super(st, bestinfo, NULL); |
8382f19b | 728 | } else { /* info.events much bigger */ |
47c7a4be | 729 | memset(avail, 0, raid_disks); |
8382f19b | 730 | max_events = info.events; |
1d1a9f87 | 731 | avail[info.disk.raid_disk] = 2; |
47c7a4be | 732 | best[info.disk.raid_disk] = devnum; |
a5d85af7 | 733 | st->ss->getinfo_super(st, bestinfo, NULL); |
8382f19b | 734 | } |
72e7fb13 N |
735 | } else if (info.disk.state & (1<<MD_DISK_REPLACEMENT)) |
736 | replcnt++; | |
3da92f27 | 737 | st->ss->free_super(st); |
8382f19b | 738 | } |
de6ae750 N |
739 | if (!avail) |
740 | return 0; | |
741 | /* We need to reject any device that thinks the best device is | |
742 | * failed or missing */ | |
743 | for (b = 0; b < raid_disks; b++) | |
744 | if (avail[b] == 2) | |
745 | break; | |
746 | cnt = 0; | |
747 | for (i = 0 ; i < raid_disks ; i++) { | |
748 | if (i != b && avail[i]) | |
749 | if (devmap[raid_disks * best[i] + b] == 0) { | |
750 | /* This device thinks 'b' is failed - | |
751 | * don't use it */ | |
752 | devnum = best[i]; | |
753 | for (d=sra->devs ; devnum; d = d->next) | |
754 | devnum--; | |
755 | d->disk.state |= (1 << MD_DISK_REMOVED); | |
756 | avail[i] = 0; | |
757 | } | |
758 | if (avail[i]) | |
759 | cnt++; | |
760 | } | |
761 | free(best); | |
762 | free(devmap); | |
72e7fb13 | 763 | return cnt + replcnt; |
8382f19b NB |
764 | } |
765 | ||
5be68a07 PC |
766 | /* test if container has degraded member(s) */ |
767 | static int container_members_max_degradation(struct map_ent *map, struct map_ent *me) | |
768 | { | |
769 | mdu_array_info_t array; | |
770 | int afd; | |
771 | int max_degraded = 0; | |
772 | ||
773 | for(; map; map = map->next) { | |
774 | if (!is_subarray(map->metadata) || | |
775 | devname2devnum(map->metadata+1) != me->devnum) | |
776 | continue; | |
777 | afd = open_dev(map->devnum); | |
778 | if (afd < 0) | |
779 | continue; | |
780 | /* most accurate information regarding array degradation */ | |
781 | if (ioctl(afd, GET_ARRAY_INFO, &array) >= 0) { | |
782 | int degraded = array.raid_disks - array.active_disks - | |
783 | array.spare_disks; | |
784 | if (degraded > max_degraded) | |
785 | max_degraded = degraded; | |
786 | } | |
787 | close(afd); | |
788 | } | |
789 | return (max_degraded); | |
790 | } | |
791 | ||
56e8be85 | 792 | static int array_try_spare(char *devname, int *dfdp, struct dev_policy *pol, |
d2db3045 | 793 | struct map_ent *target, int bare, |
56e8be85 | 794 | struct supertype *st, int verbose) |
f08605b3 N |
795 | { |
796 | /* This device doesn't have any md metadata | |
d2db3045 N |
797 | * The device policy allows 'spare' and if !bare, it allows spare-same-slot. |
798 | * If 'st' is not set, then we only know that some metadata allows this, | |
799 | * others possibly don't. | |
800 | * So look for a container or array to attach the device to. | |
801 | * Prefer 'target' if that is set and the array is found. | |
802 | * | |
f08605b3 N |
803 | * If st is set, then only arrays of that type are considered |
804 | * Return 0 on success, or some exit code on failure, probably 1. | |
805 | */ | |
8659d089 | 806 | int rv = 1; |
f08605b3 N |
807 | struct stat stb; |
808 | struct map_ent *mp, *map = NULL; | |
809 | struct mdinfo *chosen = NULL; | |
810 | int dfd = *dfdp; | |
811 | ||
f08605b3 N |
812 | if (fstat(dfd, &stb) != 0) |
813 | return 1; | |
f08605b3 | 814 | |
56e8be85 | 815 | /* |
f08605b3 N |
816 | * Now we need to find a suitable array to add this to. |
817 | * We only accept arrays that: | |
818 | * - match 'st' | |
819 | * - are in the same domains as the device | |
820 | * - are of an size for which the device will be useful | |
821 | * and we choose the one that is the most degraded | |
822 | */ | |
823 | ||
824 | if (map_lock(&map)) { | |
e7b84f9d | 825 | pr_err("failed to get exclusive lock on " |
f08605b3 N |
826 | "mapfile\n"); |
827 | return 1; | |
828 | } | |
829 | for (mp = map ; mp ; mp = mp->next) { | |
830 | struct supertype *st2; | |
831 | struct domainlist *dl = NULL; | |
832 | struct mdinfo *sra; | |
833 | unsigned long long devsize; | |
71204a50 | 834 | unsigned long long component_size = 0; |
f08605b3 N |
835 | |
836 | if (is_subarray(mp->metadata)) | |
837 | continue; | |
838 | if (st) { | |
839 | st2 = st->ss->match_metadata_desc(mp->metadata); | |
840 | if (!st2 || | |
841 | (st->minor_version >= 0 && | |
842 | st->minor_version != st2->minor_version)) { | |
843 | if (verbose > 1) | |
e7b84f9d | 844 | pr_err("not adding %s to %s as metadata type doesn't match\n", |
f08605b3 N |
845 | devname, mp->path); |
846 | free(st2); | |
847 | continue; | |
848 | } | |
849 | free(st2); | |
850 | } | |
851 | sra = sysfs_read(-1, mp->devnum, | |
852 | GET_DEVS|GET_OFFSET|GET_SIZE|GET_STATE| | |
853 | GET_DEGRADED|GET_COMPONENT|GET_VERSION); | |
854 | if (!sra) { | |
855 | /* Probably a container - no degraded info */ | |
856 | sra = sysfs_read(-1, mp->devnum, | |
857 | GET_DEVS|GET_OFFSET|GET_SIZE|GET_STATE| | |
858 | GET_COMPONENT|GET_VERSION); | |
859 | if (sra) | |
5be68a07 | 860 | sra->array.failed_disks = -1; |
f08605b3 N |
861 | } |
862 | if (!sra) | |
863 | continue; | |
864 | if (st == NULL) { | |
865 | int i; | |
866 | st2 = NULL; | |
ca3b6696 | 867 | for(i = 0; !st2 && superlist[i]; i++) |
f08605b3 N |
868 | st2 = superlist[i]->match_metadata_desc( |
869 | sra->text_version); | |
6e57f80a N |
870 | if (!st2) { |
871 | if (verbose > 1) | |
e7b84f9d | 872 | pr_err("not adding %s to %s" |
6e57f80a N |
873 | " as metadata not recognised.\n", |
874 | devname, mp->path); | |
875 | goto next; | |
876 | } | |
625b2507 N |
877 | /* Need to double check the 'act_spare' permissions applies |
878 | * to this metadata. | |
879 | */ | |
880 | if (!policy_action_allows(pol, st2->ss->name, act_spare)) | |
881 | goto next; | |
d2db3045 N |
882 | if (!bare && !policy_action_allows(pol, st2->ss->name, |
883 | act_spare_same_slot)) | |
884 | goto next; | |
f08605b3 N |
885 | } else |
886 | st2 = st; | |
5be68a07 PC |
887 | /* update number of failed disks for mostly degraded |
888 | * container member */ | |
889 | if (sra->array.failed_disks == -1) | |
890 | sra->array.failed_disks = container_members_max_degradation(map, mp); | |
891 | ||
f08605b3 | 892 | get_dev_size(dfd, NULL, &devsize); |
48865704 PC |
893 | if (sra->component_size == 0) { |
894 | /* true for containers, here we must read superblock | |
895 | * to obtain minimum spare size */ | |
896 | struct supertype *st3 = dup_super(st2); | |
897 | int mdfd = open_dev(mp->devnum); | |
f56128b9 | 898 | if (mdfd < 0) { |
1fdeb8a0 | 899 | free(st3); |
48865704 | 900 | goto next; |
1fdeb8a0 | 901 | } |
48865704 PC |
902 | if (st3->ss->load_container && |
903 | !st3->ss->load_container(st3, mdfd, mp->path)) { | |
904 | component_size = st3->ss->min_acceptable_spare_size(st3); | |
905 | st3->ss->free_super(st3); | |
906 | } | |
907 | free(st3); | |
908 | close(mdfd); | |
909 | } | |
910 | if ((sra->component_size > 0 && | |
387fcd59 N |
911 | st2->ss->avail_size(st2, devsize, |
912 | sra->devs | |
913 | ? sra->devs->data_offset | |
914 | : INVALID_SECTORS) | |
915 | < sra->component_size) | |
48865704 PC |
916 | || |
917 | (sra->component_size == 0 && devsize < component_size)) { | |
f08605b3 | 918 | if (verbose > 1) |
e7b84f9d | 919 | pr_err("not adding %s to %s as it is too small\n", |
f08605b3 N |
920 | devname, mp->path); |
921 | goto next; | |
922 | } | |
d2db3045 N |
923 | /* test against target. |
924 | * If 'target' is set and 'bare' is false, we only accept | |
925 | * arrays/containers that match 'target'. | |
926 | * If 'target' is set and 'bare' is true, we prefer the | |
927 | * array which matches 'target'. | |
b4924220 | 928 | * target is considered only if we deal with degraded array |
d2db3045 | 929 | */ |
fee6a49e PC |
930 | if (target && policy_action_allows(pol, st2->ss->name, |
931 | act_spare_same_slot)) { | |
d2db3045 N |
932 | if (strcmp(target->metadata, mp->metadata) == 0 && |
933 | memcmp(target->uuid, mp->uuid, | |
b4924220 PC |
934 | sizeof(target->uuid)) == 0 && |
935 | sra->array.failed_disks > 0) { | |
d2db3045 N |
936 | /* This is our target!! */ |
937 | if (chosen) | |
938 | sysfs_free(chosen); | |
939 | chosen = sra; | |
940 | sra = NULL; | |
941 | /* skip to end so we don't check any more */ | |
942 | while (mp->next) | |
943 | mp = mp->next; | |
944 | goto next; | |
945 | } | |
946 | /* not our target */ | |
947 | if (!bare) | |
948 | goto next; | |
949 | } | |
950 | ||
b6b3f0f7 | 951 | dl = domain_from_array(sra, st2->ss->name); |
e5508b36 | 952 | if (domain_test(dl, pol, st2->ss->name) != 1) { |
b6b3f0f7 PC |
953 | /* domain test fails */ |
954 | if (verbose > 1) | |
e7b84f9d | 955 | pr_err("not adding %s to %s as" |
b6b3f0f7 PC |
956 | " it is not in a compatible domain\n", |
957 | devname, mp->path); | |
958 | ||
959 | goto next; | |
960 | } | |
f08605b3 N |
961 | /* all tests passed, OK to add to this array */ |
962 | if (!chosen) { | |
963 | chosen = sra; | |
964 | sra = NULL; | |
965 | } else if (chosen->array.failed_disks < sra->array.failed_disks) { | |
966 | sysfs_free(chosen); | |
967 | chosen = sra; | |
968 | sra = NULL; | |
969 | } | |
970 | next: | |
971 | if (sra) | |
972 | sysfs_free(sra); | |
973 | if (st != st2) | |
974 | free(st2); | |
975 | if (dl) | |
976 | domain_free(dl); | |
977 | } | |
978 | if (chosen) { | |
979 | /* add current device to chosen array as a spare */ | |
980 | int mdfd = open_dev(devname2devnum(chosen->sys_name)); | |
981 | if (mdfd >= 0) { | |
a655e550 | 982 | struct mddev_dev devlist; |
f08605b3 N |
983 | char devname[20]; |
984 | devlist.next = NULL; | |
985 | devlist.used = 0; | |
f08605b3 N |
986 | devlist.writemostly = 0; |
987 | devlist.devname = devname; | |
988 | sprintf(devname, "%d:%d", major(stb.st_rdev), | |
989 | minor(stb.st_rdev)); | |
990 | devlist.disposition = 'a'; | |
991 | close(dfd); | |
992 | *dfdp = -1; | |
993 | rv = Manage_subdevs(chosen->sys_name, mdfd, &devlist, | |
11b391ec | 994 | -1, 0, NULL, 0); |
f08605b3 N |
995 | close(mdfd); |
996 | } | |
997 | if (verbose > 0) { | |
998 | if (rv == 0) | |
e7b84f9d | 999 | pr_err("added %s as spare for %s\n", |
f08605b3 N |
1000 | devname, chosen->sys_name); |
1001 | else | |
e7b84f9d | 1002 | pr_err("failed to add %s as spare for %s\n", |
f08605b3 N |
1003 | devname, chosen->sys_name); |
1004 | } | |
1005 | sysfs_free(chosen); | |
1006 | } | |
015da8f5 | 1007 | map_unlock(&map); |
8659d089 | 1008 | return rv; |
f08605b3 N |
1009 | } |
1010 | ||
56e8be85 N |
1011 | static int partition_try_spare(char *devname, int *dfdp, struct dev_policy *pol, |
1012 | struct supertype *st, int verbose) | |
1013 | { | |
61018da0 N |
1014 | /* we know that at least one partition virtual-metadata is |
1015 | * allowed to incorporate spares like this device. We need to | |
1016 | * find a suitable device to copy partition information from. | |
1017 | * | |
1018 | * Getting a list of all disk (not partition) devices is | |
1019 | * slightly non-trivial. We could look at /sys/block, but | |
1020 | * that is theoretically due to be removed. Maybe best to use | |
1021 | * /dev/disk/by-path/?* and ignore names ending '-partNN' as | |
1022 | * we depend on this directory of 'path' info. But that fails | |
1023 | * to find loop devices and probably others. Maybe don't | |
1024 | * worry about that, they aren't the real target. | |
1025 | * | |
1026 | * So: check things in /dev/disk/by-path to see if they are in | |
1027 | * a compatible domain, then load the partition table and see | |
1028 | * if it is OK for the new device, and choose the largest | |
1029 | * partition table that fits. | |
1030 | */ | |
1031 | DIR *dir; | |
1032 | struct dirent *de; | |
1033 | char *chosen = NULL; | |
71204a50 | 1034 | unsigned long long chosen_size = 0; |
61018da0 N |
1035 | struct supertype *chosen_st = NULL; |
1036 | int fd; | |
1037 | ||
1038 | dir = opendir("/dev/disk/by-path"); | |
1039 | if (!dir) | |
1040 | return 1; | |
1041 | while ((de = readdir(dir)) != NULL) { | |
1042 | char *ep; | |
1043 | struct dev_policy *pol2 = NULL; | |
1044 | struct domainlist *domlist = NULL; | |
1045 | int fd = -1; | |
1046 | struct mdinfo info; | |
1047 | struct supertype *st2 = NULL; | |
1048 | char *devname = NULL; | |
1049 | unsigned long long devsectors; | |
1050 | ||
1051 | if (de->d_ino == 0 || | |
1052 | de->d_name[0] == '.' || | |
1053 | (de->d_type != DT_LNK && de->d_type != DT_UNKNOWN)) | |
1054 | goto next; | |
1055 | ||
1056 | ep = de->d_name + strlen(de->d_name); | |
1057 | while (ep > de->d_name && | |
1058 | isdigit(ep[-1])) | |
1059 | ep--; | |
1060 | if (ep > de->d_name + 5 && | |
1061 | strncmp(ep-5, "-part", 5) == 0) | |
1062 | /* This is a partition - skip it */ | |
1063 | goto next; | |
1064 | ||
1065 | pol2 = path_policy(de->d_name, type_disk); | |
1066 | ||
1067 | domain_merge(&domlist, pol2, st ? st->ss->name : NULL); | |
e5508b36 | 1068 | if (domain_test(domlist, pol, st ? st->ss->name : NULL) != 1) |
61018da0 N |
1069 | /* new device is incompatible with this device. */ |
1070 | goto next; | |
1071 | ||
1072 | domain_free(domlist); | |
1073 | domlist = NULL; | |
1074 | ||
71204a50 N |
1075 | if (asprintf(&devname, "/dev/disk/by-path/%s", de->d_name) != 1) { |
1076 | devname = NULL; | |
1077 | goto next; | |
1078 | } | |
61018da0 N |
1079 | fd = open(devname, O_RDONLY); |
1080 | if (fd < 0) | |
1081 | goto next; | |
1082 | if (get_dev_size(fd, devname, &devsectors) == 0) | |
1083 | goto next; | |
1084 | devsectors >>= 9; | |
1085 | ||
1086 | if (st) | |
1087 | st2 = dup_super(st); | |
1088 | else | |
1089 | st2 = guess_super_type(fd, guess_partitions); | |
1090 | if (st2 == NULL || | |
1091 | st2->ss->load_super(st2, fd, NULL) < 0) | |
1092 | goto next; | |
1093 | ||
1094 | if (!st) { | |
1095 | /* Check domain policy again, this time referring to metadata */ | |
1096 | domain_merge(&domlist, pol2, st2->ss->name); | |
e5508b36 | 1097 | if (domain_test(domlist, pol, st2->ss->name) != 1) |
61018da0 N |
1098 | /* Incompatible devices for this metadata type */ |
1099 | goto next; | |
625b2507 N |
1100 | if (!policy_action_allows(pol, st2->ss->name, act_spare)) |
1101 | /* Some partition types allow sparing, but not | |
1102 | * this one. | |
1103 | */ | |
1104 | goto next; | |
61018da0 N |
1105 | } |
1106 | ||
a5d85af7 | 1107 | st2->ss->getinfo_super(st2, &info, NULL); |
61018da0 N |
1108 | if (info.component_size > devsectors) |
1109 | /* This partitioning doesn't fit in the device */ | |
1110 | goto next; | |
1111 | ||
1112 | /* This is an acceptable device to copy partition | |
1113 | * metadata from. We could just stop here, but I | |
1114 | * think I want to keep looking incase a larger | |
1115 | * metadata which makes better use of the device can | |
1116 | * be found. | |
1117 | */ | |
1118 | if (chosen == NULL || | |
1119 | chosen_size < info.component_size) { | |
1120 | chosen_size = info.component_size; | |
1121 | free(chosen); | |
1122 | chosen = devname; | |
1123 | devname = NULL; | |
1124 | if (chosen_st) { | |
1125 | chosen_st->ss->free_super(chosen_st); | |
1126 | free(chosen_st); | |
1127 | } | |
1128 | chosen_st = st2; | |
1129 | st2 = NULL; | |
1130 | } | |
1131 | ||
1132 | next: | |
1133 | free(devname); | |
1134 | domain_free(domlist); | |
1135 | dev_policy_free(pol2); | |
1136 | if (st2) | |
1137 | st2->ss->free_super(st2); | |
1138 | free(st2); | |
1139 | ||
1140 | if (fd >= 0) | |
1141 | close(fd); | |
1142 | } | |
1143 | ||
be5c60e3 JS |
1144 | closedir(dir); |
1145 | ||
61018da0 N |
1146 | if (!chosen) |
1147 | return 1; | |
1148 | ||
1149 | /* 'chosen' is the best device we can find. Let's write its | |
1150 | * metadata to devname dfd is read-only so don't use that | |
1151 | */ | |
1152 | fd = open(devname, O_RDWR); | |
1153 | if (fd >= 0) { | |
1154 | chosen_st->ss->store_super(chosen_st, fd); | |
1155 | close(fd); | |
1156 | } | |
1157 | free(chosen); | |
1158 | chosen_st->ss->free_super(chosen_st); | |
1159 | free(chosen_st); | |
1160 | return 0; | |
56e8be85 N |
1161 | } |
1162 | ||
52e965c2 N |
1163 | static int is_bare(int dfd) |
1164 | { | |
1165 | unsigned long long size = 0; | |
1166 | char bufpad[4096 + 4096]; | |
1167 | char *buf = (char*)(((long)bufpad + 4096) & ~4095); | |
1168 | ||
1169 | if (lseek(dfd, 0, SEEK_SET) != 0 || | |
1170 | read(dfd, buf, 4096) != 4096) | |
1171 | return 0; | |
1172 | ||
1173 | if (buf[0] != '\0' && buf[0] != '\x5a' && buf[0] != '\xff') | |
1174 | return 0; | |
1175 | if (memcmp(buf, buf+1, 4095) != 0) | |
1176 | return 0; | |
1177 | ||
1178 | /* OK, first 4K appear blank, try the end. */ | |
1179 | get_dev_size(dfd, NULL, &size); | |
1180 | if (lseek(dfd, size-4096, SEEK_SET) < 0 || | |
1181 | read(dfd, buf, 4096) != 4096) | |
1182 | return 0; | |
1183 | ||
1184 | if (buf[0] != '\0' && buf[0] != '\x5a' && buf[0] != '\xff') | |
1185 | return 0; | |
1186 | if (memcmp(buf, buf+1, 4095) != 0) | |
1187 | return 0; | |
1188 | ||
1189 | return 1; | |
1190 | } | |
56e8be85 N |
1191 | |
1192 | /* adding a spare to a regular array is quite different from adding one to | |
1193 | * a set-of-partitions virtual array. | |
1194 | * This function determines which is worth trying and tries as appropriate. | |
1195 | * Arrays are given priority over partitions. | |
1196 | */ | |
1197 | static int try_spare(char *devname, int *dfdp, struct dev_policy *pol, | |
d2db3045 | 1198 | struct map_ent *target, |
56e8be85 N |
1199 | struct supertype *st, int verbose) |
1200 | { | |
1201 | int i; | |
1202 | int rv; | |
1203 | int arrays_ok = 0; | |
1204 | int partitions_ok = 0; | |
56e8be85 | 1205 | int dfd = *dfdp; |
d2db3045 | 1206 | int bare; |
56e8be85 | 1207 | |
d2db3045 | 1208 | /* Can only add a spare if device has at least one domain */ |
56e8be85 N |
1209 | if (pol_find(pol, pol_domain) == NULL) |
1210 | return 1; | |
1211 | /* And only if some action allows spares */ | |
1212 | if (!policy_action_allows(pol, st?st->ss->name:NULL, act_spare)) | |
1213 | return 1; | |
1214 | ||
d2db3045 N |
1215 | /* Now check if the device is bare. |
1216 | * bare devices can always be added as a spare | |
1217 | * non-bare devices can only be added if spare-same-slot is permitted, | |
1218 | * and this device is replacing a previous device - in which case 'target' | |
1219 | * will be set. | |
56e8be85 | 1220 | */ |
52e965c2 | 1221 | if (!is_bare(dfd)) { |
d2db3045 N |
1222 | /* Must have a target and allow same_slot */ |
1223 | /* Later - may allow force_spare without target */ | |
1224 | if (!target || | |
1225 | !policy_action_allows(pol, st?st->ss->name:NULL, | |
1226 | act_spare_same_slot)) { | |
1227 | if (verbose > 1) | |
e7b84f9d | 1228 | pr_err("%s is not bare, so not " |
d2db3045 N |
1229 | "considering as a spare\n", |
1230 | devname); | |
1231 | return 1; | |
1232 | } | |
1233 | bare = 0; | |
1234 | } else | |
1235 | bare = 1; | |
56e8be85 | 1236 | |
d2db3045 N |
1237 | /* It might be OK to add this device to an array - need to see |
1238 | * what arrays might be candidates. | |
56e8be85 N |
1239 | */ |
1240 | if (st) { | |
1241 | /* just try try 'array' or 'partition' based on this metadata */ | |
1242 | if (st->ss->add_to_super) | |
d2db3045 | 1243 | return array_try_spare(devname, dfdp, pol, target, bare, |
56e8be85 N |
1244 | st, verbose); |
1245 | else | |
1246 | return partition_try_spare(devname, dfdp, pol, | |
1247 | st, verbose); | |
1248 | } | |
d2db3045 N |
1249 | /* No metadata was specified or found so options are open. |
1250 | * Check for whether any array metadata, or any partition metadata | |
1251 | * might allow adding the spare. This check is just help to avoid | |
1252 | * a more costly scan of all arrays when we can be sure that will | |
1253 | * fail. | |
1254 | */ | |
56e8be85 N |
1255 | for (i = 0; (!arrays_ok || !partitions_ok) && superlist[i] ; i++) { |
1256 | if (superlist[i]->add_to_super && !arrays_ok && | |
1257 | policy_action_allows(pol, superlist[i]->name, act_spare)) | |
1258 | arrays_ok = 1; | |
1259 | if (superlist[i]->add_to_super == NULL && !partitions_ok && | |
1260 | policy_action_allows(pol, superlist[i]->name, act_spare)) | |
1261 | partitions_ok = 1; | |
1262 | } | |
aaccda44 | 1263 | rv = 1; |
56e8be85 | 1264 | if (arrays_ok) |
d2db3045 N |
1265 | rv = array_try_spare(devname, dfdp, pol, target, bare, |
1266 | st, verbose); | |
aaccda44 | 1267 | if (rv != 0 && partitions_ok) |
56e8be85 N |
1268 | rv = partition_try_spare(devname, dfdp, pol, st, verbose); |
1269 | return rv; | |
1270 | } | |
1271 | ||
8382f19b NB |
1272 | int IncrementalScan(int verbose) |
1273 | { | |
1274 | /* look at every device listed in the 'map' file. | |
1275 | * If one is found that is not running then: | |
1276 | * look in mdadm.conf for bitmap file. | |
1277 | * if one exists, but array has none, add it. | |
1278 | * try to start array in auto-readonly mode | |
1279 | */ | |
1280 | struct map_ent *mapl = NULL; | |
1281 | struct map_ent *me; | |
fa56eddb | 1282 | struct mddev_ident *devs, *mddev; |
8382f19b NB |
1283 | int rv = 0; |
1284 | ||
1285 | map_read(&mapl); | |
1286 | devs = conf_get_ident(NULL); | |
1287 | ||
1288 | for (me = mapl ; me ; me = me->next) { | |
8382f19b NB |
1289 | mdu_array_info_t array; |
1290 | mdu_bitmap_file_t bmf; | |
7e0f6979 | 1291 | struct mdinfo *sra; |
339c2d6c | 1292 | int mdfd = open_dev(me->devnum); |
215bb3f7 | 1293 | |
8382f19b NB |
1294 | if (mdfd < 0) |
1295 | continue; | |
1296 | if (ioctl(mdfd, GET_ARRAY_INFO, &array) == 0 || | |
1297 | errno != ENODEV) { | |
1298 | close(mdfd); | |
1299 | continue; | |
1300 | } | |
1301 | /* Ok, we can try this one. Maybe it needs a bitmap */ | |
1302 | for (mddev = devs ; mddev ; mddev = mddev->next) | |
339c2d6c | 1303 | if (mddev->devname && me->path |
2400e6eb | 1304 | && devname_matches(mddev->devname, me->path)) |
8382f19b NB |
1305 | break; |
1306 | if (mddev && mddev->bitmap_file) { | |
1307 | /* | |
1308 | * Note: early kernels will wrongly fail this, so it | |
1309 | * is a hint only | |
1310 | */ | |
1311 | int added = -1; | |
1312 | if (ioctl(mdfd, GET_ARRAY_INFO, &bmf) < 0) { | |
1313 | int bmfd = open(mddev->bitmap_file, O_RDWR); | |
1314 | if (bmfd >= 0) { | |
1315 | added = ioctl(mdfd, SET_BITMAP_FILE, | |
1316 | bmfd); | |
1317 | close(bmfd); | |
1318 | } | |
1319 | } | |
1320 | if (verbose >= 0) { | |
1321 | if (added == 0) | |
e7b84f9d N |
1322 | pr_err("Added bitmap %s to %s\n", |
1323 | mddev->bitmap_file, me->path); | |
8382f19b | 1324 | else if (errno != EEXIST) |
e7b84f9d N |
1325 | pr_err("Failed to add bitmap to %s: %s\n", |
1326 | me->path, strerror(errno)); | |
8382f19b NB |
1327 | } |
1328 | } | |
5e88ab2e N |
1329 | /* FIXME check for reshape_active and consider not |
1330 | * starting array. | |
1331 | */ | |
8382f19b NB |
1332 | sra = sysfs_read(mdfd, 0, 0); |
1333 | if (sra) { | |
1334 | if (sysfs_set_str(sra, NULL, | |
1335 | "array_state", "read-auto") == 0) { | |
1336 | if (verbose >= 0) | |
e7b84f9d N |
1337 | pr_err("started array %s\n", |
1338 | me->path ?: devnum2devname(me->devnum)); | |
8382f19b | 1339 | } else { |
e7b84f9d N |
1340 | pr_err("failed to start array %s: %s\n", |
1341 | me->path ?: devnum2devname(me->devnum), | |
1342 | strerror(errno)); | |
8382f19b NB |
1343 | rv = 1; |
1344 | } | |
b1efa6c2 | 1345 | sysfs_free(sra); |
8382f19b NB |
1346 | } |
1347 | } | |
1348 | return rv; | |
1349 | } | |
598f0d58 | 1350 | |
1771a6e2 N |
1351 | static char *container2devname(char *devname) |
1352 | { | |
1353 | char *mdname = NULL; | |
1354 | ||
1355 | if (devname[0] == '/') { | |
1356 | int fd = open(devname, O_RDONLY); | |
1357 | if (fd >= 0) { | |
1358 | mdname = devnum2devname(fd2devnum(fd)); | |
1359 | close(fd); | |
1360 | } | |
1361 | } else { | |
1362 | int uuid[4]; | |
1363 | struct map_ent *mp, *map = NULL; | |
ca3b6696 | 1364 | |
1771a6e2 N |
1365 | if (!parse_uuid(devname, uuid)) |
1366 | return mdname; | |
1367 | mp = map_by_uuid(&map, uuid); | |
1368 | if (mp) | |
1369 | mdname = devnum2devname(mp->devnum); | |
1370 | map_free(map); | |
1371 | } | |
1372 | ||
1373 | return mdname; | |
1374 | } | |
1375 | ||
47c74f3f | 1376 | static int Incremental_container(struct supertype *st, char *devname, |
11b6d91d | 1377 | struct context *c) |
598f0d58 NB |
1378 | { |
1379 | /* Collect the contents of this container and for each | |
1380 | * array, choose a device name and assemble the array. | |
1381 | */ | |
1382 | ||
47c74f3f | 1383 | struct mdinfo *list; |
598f0d58 | 1384 | struct mdinfo *ra; |
ad5bc697 | 1385 | struct map_ent *map = NULL; |
47c74f3f N |
1386 | struct mdinfo info; |
1387 | int trustworthy; | |
fa56eddb | 1388 | struct mddev_ident *match; |
47c74f3f | 1389 | int rv = 0; |
037e12d6 AC |
1390 | struct domainlist *domains; |
1391 | struct map_ent *smp; | |
1392 | int suuid[4]; | |
1393 | int sfd; | |
81219e70 LM |
1394 | int ra_blocked = 0; |
1395 | int ra_all = 0; | |
47c74f3f | 1396 | |
47c74f3f N |
1397 | st->ss->getinfo_super(st, &info, NULL); |
1398 | ||
11b6d91d | 1399 | if ((c->runstop > 0 && info.container_enough >= 0) || |
47c74f3f N |
1400 | info.container_enough > 0) |
1401 | /* pass */; | |
1402 | else { | |
11b6d91d | 1403 | if (c->verbose) |
e7b84f9d | 1404 | pr_err("not enough devices to start the container\n"); |
47c74f3f N |
1405 | return 0; |
1406 | } | |
1407 | ||
11b6d91d | 1408 | match = conf_match(st, &info, devname, c->verbose, &rv); |
47c74f3f N |
1409 | if (match == NULL && rv == 2) |
1410 | return rv; | |
1411 | ||
1412 | /* Need to compute 'trustworthy' */ | |
1413 | if (match) | |
1414 | trustworthy = LOCAL; | |
11b6d91d | 1415 | else if (st->ss->match_home(st, c->homehost) == 1) |
47c74f3f N |
1416 | trustworthy = LOCAL; |
1417 | else if (st->ss->match_home(st, "any") == 1) | |
1418 | trustworthy = LOCAL; | |
1419 | else | |
1420 | trustworthy = FOREIGN; | |
ad5bc697 | 1421 | |
47c74f3f | 1422 | list = st->ss->container_content(st, NULL); |
81219e70 LM |
1423 | /* when nothing to activate - quit */ |
1424 | if (list == NULL) | |
1425 | return 0; | |
598f0d58 | 1426 | for (ra = list ; ra ; ra = ra->next) { |
598f0d58 NB |
1427 | int mdfd; |
1428 | char chosen_name[1024]; | |
ad5bc697 | 1429 | struct map_ent *mp; |
fa56eddb | 1430 | struct mddev_ident *match = NULL; |
598f0d58 | 1431 | |
81219e70 LM |
1432 | ra_all++; |
1433 | /* do not activate arrays blocked by metadata handler */ | |
1434 | if (ra->array.state & (1 << MD_SB_BLOCK_VOLUME)) { | |
e7b84f9d | 1435 | pr_err("Cannot activate array %s in %s.\n", |
81219e70 LM |
1436 | ra->text_version, devname); |
1437 | ra_blocked++; | |
1438 | continue; | |
1439 | } | |
c5afc314 | 1440 | mp = map_by_uuid(&map, ra->uuid); |
d7288ddc | 1441 | |
30926600 N |
1442 | if (mp) { |
1443 | mdfd = open_dev(mp->devnum); | |
339c2d6c N |
1444 | if (mp->path) |
1445 | strcpy(chosen_name, mp->path); | |
1446 | else | |
1447 | strcpy(chosen_name, devnum2devname(mp->devnum)); | |
30926600 | 1448 | } else { |
dbb44303 | 1449 | |
112cace6 | 1450 | /* Check in mdadm.conf for container == devname and |
dbb44303 N |
1451 | * member == ra->text_version after second slash. |
1452 | */ | |
1453 | char *sub = strchr(ra->text_version+1, '/'); | |
fa56eddb | 1454 | struct mddev_ident *array_list; |
dbb44303 N |
1455 | if (sub) { |
1456 | sub++; | |
1457 | array_list = conf_get_ident(NULL); | |
1458 | } else | |
1459 | array_list = NULL; | |
1460 | for(; array_list ; array_list = array_list->next) { | |
dbb44303 N |
1461 | char *dn; |
1462 | if (array_list->member == NULL || | |
1463 | array_list->container == NULL) | |
1464 | continue; | |
1465 | if (strcmp(array_list->member, sub) != 0) | |
1466 | continue; | |
71d60c48 DW |
1467 | if (array_list->uuid_set && |
1468 | !same_uuid(ra->uuid, array_list->uuid, st->ss->swapuuid)) | |
1469 | continue; | |
1771a6e2 N |
1470 | dn = container2devname(array_list->container); |
1471 | if (dn == NULL) | |
dbb44303 | 1472 | continue; |
dbb44303 N |
1473 | if (strncmp(dn, ra->text_version+1, |
1474 | strlen(dn)) != 0 || | |
1475 | ra->text_version[strlen(dn)+1] != '/') { | |
1476 | free(dn); | |
1477 | continue; | |
1478 | } | |
1479 | free(dn); | |
1480 | /* we have a match */ | |
1481 | match = array_list; | |
11b6d91d | 1482 | if (c->verbose>0) |
e7b84f9d | 1483 | pr_err("match found for member %s\n", |
71d60c48 | 1484 | array_list->member); |
dbb44303 N |
1485 | break; |
1486 | } | |
dbb44303 | 1487 | |
112cace6 N |
1488 | if (match && match->devname && |
1489 | strcasecmp(match->devname, "<ignore>") == 0) { | |
11b6d91d | 1490 | if (c->verbose > 0) |
e7b84f9d N |
1491 | pr_err("array %s/%s is " |
1492 | "explicitly ignored by mdadm.conf\n", | |
1493 | match->container, match->member); | |
112cace6 N |
1494 | return 2; |
1495 | } | |
05833051 N |
1496 | if (match) |
1497 | trustworthy = LOCAL; | |
112cace6 | 1498 | |
30926600 N |
1499 | mdfd = create_mddev(match ? match->devname : NULL, |
1500 | ra->name, | |
11b6d91d | 1501 | c->autof, |
30926600 N |
1502 | trustworthy, |
1503 | chosen_name); | |
1504 | } | |
598f0d58 NB |
1505 | |
1506 | if (mdfd < 0) { | |
e7b84f9d | 1507 | pr_err("failed to open %s: %s.\n", |
598f0d58 NB |
1508 | chosen_name, strerror(errno)); |
1509 | return 2; | |
1510 | } | |
1511 | ||
11b6d91d N |
1512 | assemble_container_content(st, mdfd, ra, c, |
1513 | chosen_name); | |
588bebfc | 1514 | close(mdfd); |
598f0d58 | 1515 | } |
037e12d6 | 1516 | |
81219e70 LM |
1517 | /* don't move spares to container with volume being activated |
1518 | when all volumes are blocked */ | |
ad098cdd | 1519 | if (ra_all == ra_blocked) |
81219e70 | 1520 | return 0; |
81219e70 | 1521 | |
037e12d6 AC |
1522 | /* Now move all suitable spares from spare container */ |
1523 | domains = domain_from_array(list, st->ss->name); | |
1524 | memcpy(suuid, uuid_zero, sizeof(int[4])); | |
1525 | if (domains && | |
1526 | (smp = map_by_uuid(&map, suuid)) != NULL && | |
1527 | (sfd = open(smp->path, O_RDONLY)) >= 0) { | |
1528 | /* spare container found */ | |
1529 | struct supertype *sst = | |
1530 | super_imsm.match_metadata_desc("imsm"); | |
1531 | struct mdinfo *sinfo; | |
1532 | unsigned long long min_size = 0; | |
1533 | if (st->ss->min_acceptable_spare_size) | |
1534 | min_size = st->ss->min_acceptable_spare_size(st); | |
1535 | if (!sst->ss->load_container(sst, sfd, NULL)) { | |
1536 | close(sfd); | |
1537 | sinfo = container_choose_spares(sst, min_size, | |
1538 | domains, NULL, | |
1539 | st->ss->name, 0); | |
1540 | sst->ss->free_super(sst); | |
1541 | if (sinfo){ | |
1542 | int count = 0; | |
1543 | struct mdinfo *disks = sinfo->devs; | |
1544 | while (disks) { | |
1545 | /* move spare from spare | |
1546 | * container to currently | |
1547 | * assembled one | |
1548 | */ | |
1549 | if (move_spare( | |
1550 | smp->path, | |
1551 | devname, | |
1552 | makedev(disks->disk.major, | |
1553 | disks->disk.minor))) | |
1554 | count++; | |
1555 | disks = disks->next; | |
1556 | } | |
1557 | if (count) | |
e7b84f9d N |
1558 | pr_err("Added %d spare%s to %s\n", |
1559 | count, count>1?"s":"", devname); | |
037e12d6 AC |
1560 | } |
1561 | sysfs_free(sinfo); | |
1562 | } else | |
1563 | close(sfd); | |
1564 | } | |
1565 | domain_free(domains); | |
598f0d58 NB |
1566 | return 0; |
1567 | } | |
29ba4804 N |
1568 | |
1569 | /* | |
1570 | * IncrementalRemove - Attempt to see if the passed in device belongs to any | |
1571 | * raid arrays, and if so first fail (if needed) and then remove the device. | |
1572 | * | |
1573 | * @devname - The device we want to remove | |
950bc344 | 1574 | * @id_path - name as found in /dev/disk/by-path for this device |
29ba4804 N |
1575 | * |
1576 | * Note: the device name must be a kernel name like "sda", so | |
1577 | * that we can find it in /proc/mdstat | |
1578 | */ | |
950bc344 | 1579 | int IncrementalRemove(char *devname, char *id_path, int verbose) |
29ba4804 N |
1580 | { |
1581 | int mdfd; | |
aae3cdc3 | 1582 | int rv; |
29ba4804 | 1583 | struct mdstat_ent *ent; |
a655e550 | 1584 | struct mddev_dev devlist; |
29ba4804 | 1585 | |
950bc344 PC |
1586 | if (!id_path) |
1587 | dprintf(Name ": incremental removal without --path <id_path> " | |
1588 | "lacks the possibility to re-add new device in this " | |
1589 | "port\n"); | |
1590 | ||
29ba4804 | 1591 | if (strchr(devname, '/')) { |
e7b84f9d | 1592 | pr_err("incremental removal requires a " |
29ba4804 N |
1593 | "kernel device name, not a file: %s\n", devname); |
1594 | return 1; | |
1595 | } | |
1596 | ent = mdstat_by_component(devname); | |
1597 | if (!ent) { | |
e7b84f9d | 1598 | pr_err("%s does not appear to be a component " |
29ba4804 N |
1599 | "of any array\n", devname); |
1600 | return 1; | |
1601 | } | |
1602 | mdfd = open_dev(ent->devnum); | |
1603 | if (mdfd < 0) { | |
e7b84f9d | 1604 | pr_err("Cannot open array %s!!\n", ent->dev); |
08387a04 | 1605 | free_mdstat(ent); |
29ba4804 N |
1606 | return 1; |
1607 | } | |
403410eb PC |
1608 | |
1609 | if (id_path) { | |
1610 | struct map_ent *map = NULL, *me; | |
1611 | me = map_by_devnum(&map, ent->devnum); | |
1612 | if (me) | |
1613 | policy_save_path(id_path, me); | |
1614 | map_free(map); | |
1615 | } | |
1616 | ||
29ba4804 N |
1617 | memset(&devlist, 0, sizeof(devlist)); |
1618 | devlist.devname = devname; | |
1619 | devlist.disposition = 'f'; | |
08387a04 N |
1620 | /* for a container, we must fail each member array */ |
1621 | if (ent->metadata_version && | |
1622 | strncmp(ent->metadata_version, "external:", 9) == 0) { | |
1623 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
1624 | struct mdstat_ent *memb; | |
1625 | for (memb = mdstat ; memb ; memb = memb->next) | |
1626 | if (is_container_member(memb, ent->dev)) { | |
1627 | int subfd = open_dev(memb->devnum); | |
1628 | if (subfd >= 0) { | |
1629 | Manage_subdevs(memb->dev, subfd, | |
833bb0f8 | 1630 | &devlist, verbose, 0, |
11b391ec | 1631 | NULL, 0); |
08387a04 N |
1632 | close(subfd); |
1633 | } | |
1634 | } | |
1635 | free_mdstat(mdstat); | |
1636 | } else | |
11b391ec | 1637 | Manage_subdevs(ent->dev, mdfd, &devlist, verbose, 0, NULL, 0); |
29ba4804 | 1638 | devlist.disposition = 'r'; |
11b391ec | 1639 | rv = Manage_subdevs(ent->dev, mdfd, &devlist, verbose, 0, NULL, 0); |
aae3cdc3 | 1640 | close(mdfd); |
08387a04 | 1641 | free_mdstat(ent); |
aae3cdc3 | 1642 | return rv; |
29ba4804 | 1643 | } |