]>
Commit | Line | Data |
---|---|---|
64c4757e | 1 | /* |
9a9dab36 | 2 | * mdadm - manage Linux "md" devices aka RAID arrays. |
64c4757e | 3 | * |
c61b1c0b | 4 | * Copyright (C) 2001-2016 Neil Brown <neilb@suse.com> |
64c4757e NB |
5 | * |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License as published by | |
9 | * the Free Software Foundation; either version 2 of the License, or | |
10 | * (at your option) any later version. | |
11 | * | |
12 | * This program is distributed in the hope that it will be useful, | |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | * GNU General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with this program; if not, write to the Free Software | |
19 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
20 | * | |
21 | * Author: Neil Brown | |
e736b623 | 22 | * Email: <neilb@suse.de> |
64c4757e NB |
23 | */ |
24 | ||
9a9dab36 | 25 | #include "mdadm.h" |
41a3b72a | 26 | #include <ctype.h> |
64c4757e | 27 | |
86b77ddf | 28 | static int name_matches(char *found, char *required, char *homehost, int require_homehost) |
624920bb NB |
29 | { |
30 | /* See if the name found matches the required name, possibly | |
31 | * prefixed with 'homehost' | |
32 | */ | |
00f23a88 N |
33 | char *sep; |
34 | unsigned int l; | |
624920bb | 35 | |
624920bb NB |
36 | if (strcmp(found, required)==0) |
37 | return 1; | |
00f23a88 N |
38 | sep = strchr(found, ':'); |
39 | if (!sep) | |
40 | return 0; | |
41 | l = sep - found; | |
42 | if (strncmp(found, "any:", 4) == 0 || | |
43 | (homehost && strcmp(homehost, "any") == 0) || | |
86b77ddf | 44 | !require_homehost || |
00f23a88 N |
45 | (homehost && strlen(homehost) == l && |
46 | strncmp(found, homehost, l) == 0)) { | |
47 | /* matching homehost */ | |
48 | if (strcmp(sep+1, required) == 0) | |
624920bb NB |
49 | return 1; |
50 | } | |
51 | return 0; | |
52 | } | |
53 | ||
9008ed1c | 54 | static int is_member_busy(char *metadata_version) |
2de8884f DW |
55 | { |
56 | /* check if the given member array is active */ | |
b4924f46 | 57 | struct mdstat_ent *mdstat = mdstat_read(0, 0); |
2de8884f DW |
58 | struct mdstat_ent *ent; |
59 | int busy = 0; | |
60 | ||
61 | for (ent = mdstat; ent; ent = ent->next) { | |
62 | if (ent->metadata_version == NULL) | |
63 | continue; | |
64 | if (strncmp(ent->metadata_version, "external:", 9) != 0) | |
65 | continue; | |
66 | if (!is_subarray(&ent->metadata_version[9])) | |
67 | continue; | |
68 | /* Skip first char - it can be '/' or '-' */ | |
69 | if (strcmp(&ent->metadata_version[10], metadata_version+1) == 0) { | |
70 | busy = 1; | |
71 | break; | |
72 | } | |
73 | } | |
74 | free_mdstat(mdstat); | |
75 | ||
76 | return busy; | |
77 | } | |
78 | ||
fa56eddb | 79 | static int ident_matches(struct mddev_ident *ident, |
08fb91a3 N |
80 | struct mdinfo *content, |
81 | struct supertype *tst, | |
86b77ddf | 82 | char *homehost, int require_homehost, |
08fb91a3 N |
83 | char *update, char *devname) |
84 | { | |
85 | ||
86 | if (ident->uuid_set && (!update || strcmp(update, "uuid")!= 0) && | |
cbeeb0e5 AC |
87 | same_uuid(content->uuid, ident->uuid, tst->ss->swapuuid)==0 && |
88 | memcmp(content->uuid, uuid_zero, sizeof(int[4])) != 0) { | |
08fb91a3 | 89 | if (devname) |
e7b84f9d | 90 | pr_err("%s has wrong uuid.\n", devname); |
08fb91a3 N |
91 | return 0; |
92 | } | |
93 | if (ident->name[0] && (!update || strcmp(update, "name")!= 0) && | |
86b77ddf | 94 | name_matches(content->name, ident->name, homehost, require_homehost)==0) { |
08fb91a3 | 95 | if (devname) |
e7b84f9d | 96 | pr_err("%s has wrong name.\n", devname); |
08fb91a3 N |
97 | return 0; |
98 | } | |
99 | if (ident->super_minor != UnSet && | |
100 | ident->super_minor != content->array.md_minor) { | |
101 | if (devname) | |
e7b84f9d N |
102 | pr_err("%s has wrong super-minor.\n", |
103 | devname); | |
08fb91a3 N |
104 | return 0; |
105 | } | |
106 | if (ident->level != UnSet && | |
107 | ident->level != content->array.level) { | |
108 | if (devname) | |
e7b84f9d N |
109 | pr_err("%s has wrong raid level.\n", |
110 | devname); | |
08fb91a3 N |
111 | return 0; |
112 | } | |
113 | if (ident->raid_disks != UnSet && | |
babb8dd4 | 114 | content->array.raid_disks != 0 && /* metadata doesn't know how many to expect */ |
08fb91a3 N |
115 | ident->raid_disks!= content->array.raid_disks) { |
116 | if (devname) | |
e7b84f9d N |
117 | pr_err("%s requires wrong number of drives.\n", |
118 | devname); | |
08fb91a3 N |
119 | return 0; |
120 | } | |
805d30b2 N |
121 | if (ident->member && ident->member[0]) { |
122 | /* content->text_version must match */ | |
123 | char *s = strchr(content->text_version+1, '/'); | |
124 | if (s == NULL) { | |
125 | if (devname) | |
e7b84f9d N |
126 | pr_err("%s is not a container and one is required.\n", |
127 | devname); | |
805d30b2 N |
128 | return 0; |
129 | } else if (strcmp(ident->member, s+1) != 0) { | |
130 | if (devname) | |
e7b84f9d N |
131 | pr_err("skipping wrong member %s is %s\n", |
132 | content->text_version, devname); | |
805d30b2 N |
133 | return 0; |
134 | } | |
135 | } | |
08fb91a3 N |
136 | return 1; |
137 | } | |
08fb91a3 | 138 | |
95425a89 N |
139 | static int select_devices(struct mddev_dev *devlist, |
140 | struct mddev_ident *ident, | |
141 | struct supertype **stp, | |
142 | struct mdinfo **contentp, | |
143 | struct context *c, | |
144 | int inargv, int auto_assem) | |
64c4757e | 145 | { |
a655e550 | 146 | struct mddev_dev *tmpdev; |
95425a89 N |
147 | int num_devs; |
148 | struct supertype *st = *stp; | |
98dbd966 | 149 | struct mdinfo *content = NULL; |
8cf2eb96 | 150 | int report_mismatch = ((inargv && c->verbose >= 0) || c->verbose > 0); |
cbeeb0e5 | 151 | struct domainlist *domains = NULL; |
0a6bff09 | 152 | dev_t rdev; |
60e1bc1a | 153 | |
5787fa49 NB |
154 | tmpdev = devlist; num_devs = 0; |
155 | while (tmpdev) { | |
da6b5ca9 NB |
156 | if (tmpdev->used) |
157 | tmpdev->used = 2; | |
158 | else | |
159 | num_devs++; | |
0431869c | 160 | tmpdev->disposition = 0; |
5787fa49 NB |
161 | tmpdev = tmpdev->next; |
162 | } | |
5787fa49 | 163 | |
811e6cbe NB |
164 | /* first walk the list of devices to find a consistent set |
165 | * that match the criterea, if that is possible. | |
c30e5369 | 166 | * We flag the ones we like with 'used'. |
811e6cbe NB |
167 | */ |
168 | for (tmpdev = devlist; | |
169 | tmpdev; | |
5083d66b | 170 | tmpdev = tmpdev ? tmpdev->next : NULL) { |
811e6cbe | 171 | char *devname = tmpdev->devname; |
64c4757e | 172 | int dfd; |
518a60f3 | 173 | struct supertype *tst; |
4e8d9f0a | 174 | struct dev_policy *pol = NULL; |
5083d66b | 175 | int found_container = 0; |
52826846 | 176 | |
95425a89 N |
177 | if (tmpdev->used > 1) |
178 | continue; | |
da6b5ca9 | 179 | |
c39b2e63 N |
180 | if (ident->container) { |
181 | if (ident->container[0] == '/' && | |
182 | !same_dev(ident->container, devname)) { | |
183 | if (report_mismatch) | |
184 | pr_err("%s is not the container required (%s)\n", | |
185 | devname, ident->container); | |
186 | continue; | |
187 | } | |
188 | } else if (ident->devices && | |
189 | !match_oneof(ident->devices, devname)) { | |
190 | /* Note that we ignore the "device=" identifier if a | |
191 | * "container=" is given. Checking both is unnecessarily | |
192 | * complicated. | |
193 | */ | |
8cf2eb96 | 194 | if (report_mismatch) |
e7b84f9d | 195 | pr_err("%s is not one of %s\n", devname, ident->devices); |
52826846 | 196 | continue; |
56eedc1a | 197 | } |
4b1ac34b | 198 | |
518a60f3 JS |
199 | tst = dup_super(st); |
200 | ||
56d18859 | 201 | dfd = dev_open(devname, O_RDONLY); |
64c4757e | 202 | if (dfd < 0) { |
8cf2eb96 | 203 | if (report_mismatch) |
e7b84f9d N |
204 | pr_err("cannot open device %s: %s\n", |
205 | devname, strerror(errno)); | |
da6b5ca9 | 206 | tmpdev->used = 2; |
0a6bff09 | 207 | } else if (!fstat_is_blkdev(dfd, devname, &rdev)) { |
da6b5ca9 | 208 | tmpdev->used = 2; |
5083d66b N |
209 | } else if (must_be_container(dfd)) { |
210 | if (st) { | |
211 | /* already found some components, this cannot | |
212 | * be another one. | |
213 | */ | |
8cf2eb96 | 214 | if (report_mismatch) |
e7b84f9d N |
215 | pr_err("%s is a container, but we are looking for components\n", |
216 | devname); | |
5083d66b N |
217 | tmpdev->used = 2; |
218 | } if (!tst && (tst = super_by_fd(dfd, NULL)) == NULL) { | |
8cf2eb96 | 219 | if (report_mismatch) |
e7b84f9d N |
220 | pr_err("not a recognisable container: %s\n", |
221 | devname); | |
5083d66b | 222 | tmpdev->used = 2; |
d7be7d87 JS |
223 | } else if (!tst->ss->load_container || |
224 | tst->ss->load_container(tst, dfd, NULL)) { | |
8cf2eb96 | 225 | if (report_mismatch) |
e7b84f9d N |
226 | pr_err("no correct container type: %s\n", |
227 | devname); | |
5083d66b N |
228 | tmpdev->used = 2; |
229 | } else if (auto_assem && | |
0a6bff09 ZL |
230 | !conf_test_metadata(tst->ss->name, |
231 | (pol = devid_policy(rdev)), | |
4977146a | 232 | tst->ss->match_home(tst, c->homehost) == 1)) { |
8cf2eb96 | 233 | if (report_mismatch) |
7a862a02 | 234 | pr_err("%s has metadata type %s for which auto-assembly is disabled\n", |
e7b84f9d | 235 | devname, tst->ss->name); |
5083d66b N |
236 | tmpdev->used = 2; |
237 | } else | |
238 | found_container = 1; | |
52826846 | 239 | } else { |
5083d66b | 240 | if (!tst && (tst = guess_super(dfd)) == NULL) { |
8cf2eb96 | 241 | if (report_mismatch) |
e7b84f9d N |
242 | pr_err("no recogniseable superblock on %s\n", |
243 | devname); | |
5083d66b | 244 | tmpdev->used = 2; |
83607604 N |
245 | } else if ((tst->ignore_hw_compat = 0), |
246 | tst->ss->load_super(tst, dfd, | |
247 | report_mismatch ? devname : NULL)) { | |
8cf2eb96 | 248 | if (report_mismatch) |
e7b84f9d N |
249 | pr_err("no RAID superblock on %s\n", |
250 | devname); | |
5083d66b N |
251 | tmpdev->used = 2; |
252 | } else if (tst->ss->compare_super == NULL) { | |
8cf2eb96 | 253 | if (report_mismatch) |
e7b84f9d N |
254 | pr_err("Cannot assemble %s metadata on %s\n", |
255 | tst->ss->name, devname); | |
5083d66b N |
256 | tmpdev->used = 2; |
257 | } else if (auto_assem && st == NULL && | |
0a6bff09 ZL |
258 | !conf_test_metadata(tst->ss->name, |
259 | (pol = devid_policy(rdev)), | |
4977146a | 260 | tst->ss->match_home(tst, c->homehost) == 1)) { |
8cf2eb96 | 261 | if (report_mismatch) |
7a862a02 | 262 | pr_err("%s has metadata type %s for which auto-assembly is disabled\n", |
e7b84f9d | 263 | devname, tst->ss->name); |
5083d66b N |
264 | tmpdev->used = 2; |
265 | } | |
64c4757e | 266 | } |
c06487ce | 267 | if (dfd >= 0) close(dfd); |
d68ea4d7 | 268 | if (tmpdev->used == 2) { |
d4386799 | 269 | if (auto_assem || !inargv) |
d68ea4d7 N |
270 | /* Ignore unrecognised devices during auto-assembly */ |
271 | goto loop; | |
272 | if (ident->uuid_set || ident->name[0] || | |
273 | ident->super_minor != UnSet) | |
274 | /* Ignore unrecognised device if looking for | |
275 | * specific array */ | |
276 | goto loop; | |
d68ea4d7 | 277 | |
e7b84f9d | 278 | pr_err("%s has no superblock - assembly aborted\n", |
79f9f56d | 279 | devname); |
d68ea4d7 N |
280 | if (st) |
281 | st->ss->free_super(st); | |
282 | dev_policy_free(pol); | |
cbeeb0e5 | 283 | domain_free(domains); |
4a670aab AY |
284 | if (tst) |
285 | tst->ss->free_super(tst); | |
95425a89 | 286 | return -1; |
d68ea4d7 | 287 | } |
52826846 | 288 | |
5083d66b | 289 | if (found_container) { |
9008ed1c N |
290 | /* tmpdev is a container. We need to be either |
291 | * looking for a member, or auto-assembling | |
292 | */ | |
56d18859 N |
293 | /* should be safe to try an exclusive open now, we |
294 | * have rejected anything that some other mdadm might | |
295 | * be looking at | |
296 | */ | |
297 | dfd = dev_open(devname, O_RDONLY | O_EXCL); | |
298 | if (dfd < 0) { | |
8cf2eb96 | 299 | if (report_mismatch) |
e7b84f9d | 300 | pr_err("%s is busy - skipping\n", devname); |
56d18859 N |
301 | goto loop; |
302 | } | |
303 | close(dfd); | |
9008ed1c | 304 | |
c39b2e63 N |
305 | if (ident->container && ident->container[0] != '/') { |
306 | /* we have a uuid */ | |
307 | int uuid[4]; | |
87477e6d | 308 | |
c39b2e63 N |
309 | content = *contentp; |
310 | tst->ss->getinfo_super(tst, content, NULL); | |
87477e6d | 311 | |
c39b2e63 N |
312 | if (!parse_uuid(ident->container, uuid) || |
313 | !same_uuid(content->uuid, uuid, tst->ss->swapuuid)) { | |
314 | if (report_mismatch) | |
315 | pr_err("%s has wrong UUID to be required container\n", | |
316 | devname); | |
317 | goto loop; | |
9008ed1c N |
318 | } |
319 | } | |
320 | /* It is worth looking inside this container. | |
321 | */ | |
4977146a | 322 | if (c->verbose > 0) |
e7b84f9d | 323 | pr_err("looking in container %s\n", |
79f9f56d | 324 | devname); |
1415fe4b | 325 | |
88cef9b3 N |
326 | for (content = tst->ss->container_content(tst, NULL); |
327 | content; | |
328 | content = content->next) { | |
1415fe4b | 329 | |
88cef9b3 | 330 | if (!ident_matches(ident, content, tst, |
86b77ddf N |
331 | c->homehost, c->require_homehost, |
332 | c->update, | |
8cf2eb96 | 333 | report_mismatch ? devname : NULL)) |
88cef9b3 N |
334 | /* message already printed */; |
335 | else if (is_member_busy(content->text_version)) { | |
8cf2eb96 | 336 | if (report_mismatch) |
e7b84f9d | 337 | pr_err("member %s in %s is already assembled\n", |
79f9f56d N |
338 | content->text_version, |
339 | devname); | |
81219e70 LM |
340 | } else if (content->array.state & (1<<MD_SB_BLOCK_VOLUME)) { |
341 | /* do not assemble arrays with unsupported configurations */ | |
e7b84f9d | 342 | pr_err("Cannot activate member %s in %s.\n", |
79f9f56d N |
343 | content->text_version, |
344 | devname); | |
88cef9b3 N |
345 | } else |
346 | break; | |
347 | } | |
348 | if (!content) { | |
f7ad3ccc | 349 | tmpdev->used = 2; |
88cef9b3 | 350 | goto loop; /* empty container */ |
fa031239 | 351 | } |
88cef9b3 | 352 | |
9008ed1c | 353 | st = tst; tst = NULL; |
4c1c3ad8 | 354 | if (!auto_assem && inargv && tmpdev->next != NULL) { |
7a862a02 | 355 | pr_err("%s is a container, but is not only device given: confused and aborting\n", |
79f9f56d | 356 | devname); |
9008ed1c | 357 | st->ss->free_super(st); |
4e8d9f0a | 358 | dev_policy_free(pol); |
cbeeb0e5 | 359 | domain_free(domains); |
95425a89 | 360 | return -1; |
9008ed1c | 361 | } |
4977146a | 362 | if (c->verbose > 0) |
e7b84f9d | 363 | pr_err("found match on member %s in %s\n", |
79f9f56d | 364 | content->text_version, devname); |
bac0d92e | 365 | |
5083d66b N |
366 | /* make sure we finished the loop */ |
367 | tmpdev = NULL; | |
bac0d92e | 368 | goto loop; |
5083d66b | 369 | } else { |
95425a89 | 370 | content = *contentp; |
5083d66b N |
371 | tst->ss->getinfo_super(tst, content, NULL); |
372 | ||
373 | if (!ident_matches(ident, content, tst, | |
86b77ddf N |
374 | c->homehost, c->require_homehost, |
375 | c->update, | |
8cf2eb96 | 376 | report_mismatch ? devname : NULL)) |
df37ffc0 | 377 | goto loop; |
66eb2c93 | 378 | |
5141638c N |
379 | if (auto_assem) { |
380 | /* Never auto-assemble things that conflict | |
381 | * with mdadm.conf in some way | |
1d04e275 | 382 | */ |
5141638c N |
383 | struct mddev_ident *match; |
384 | int rv = 0; | |
385 | ||
386 | match = conf_match(tst, content, devname, | |
387 | report_mismatch ? c->verbose : -1, | |
388 | &rv); | |
389 | if (!match && rv == 2) | |
390 | goto loop; | |
391 | if (match && match->devname && | |
392 | strcasecmp(match->devname, "<ignore>") == 0) { | |
393 | if (report_mismatch) | |
394 | pr_err("%s is a member of an explicitly ignored array\n", | |
395 | devname); | |
396 | goto loop; | |
397 | } | |
398 | if (match && !ident_matches(match, content, tst, | |
86b77ddf N |
399 | c->homehost, c->require_homehost, |
400 | c->update, | |
5141638c N |
401 | report_mismatch ? devname : NULL)) |
402 | /* Array exists in mdadm.conf but some | |
403 | * details don't match, so reject it | |
404 | */ | |
405 | goto loop; | |
406 | } | |
66eb2c93 | 407 | |
56d18859 N |
408 | /* should be safe to try an exclusive open now, we |
409 | * have rejected anything that some other mdadm might | |
410 | * be looking at | |
411 | */ | |
412 | dfd = dev_open(devname, O_RDONLY | O_EXCL); | |
413 | if (dfd < 0) { | |
8cf2eb96 | 414 | if (report_mismatch) |
e7b84f9d | 415 | pr_err("%s is busy - skipping\n", devname); |
56d18859 N |
416 | goto loop; |
417 | } | |
418 | close(dfd); | |
419 | ||
5083d66b N |
420 | if (st == NULL) |
421 | st = dup_super(tst); | |
422 | if (st->minor_version == -1) | |
423 | st->minor_version = tst->minor_version; | |
ed7fc6b4 AC |
424 | |
425 | if (memcmp(content->uuid, uuid_zero, | |
426 | sizeof(int[4])) == 0) { | |
427 | /* this is a floating spare. It cannot define | |
428 | * an array unless there are no more arrays of | |
429 | * this type to be found. It can be included | |
430 | * in an array of this type though. | |
431 | */ | |
432 | tmpdev->used = 3; | |
433 | goto loop; | |
434 | } | |
435 | ||
5083d66b N |
436 | if (st->ss != tst->ss || |
437 | st->minor_version != tst->minor_version || | |
438 | st->ss->compare_super(st, tst) != 0) { | |
439 | /* Some mismatch. If exactly one array matches this host, | |
440 | * we can resolve on that one. | |
441 | * Or, if we are auto assembling, we just ignore the second | |
442 | * for now. | |
443 | */ | |
444 | if (auto_assem) | |
445 | goto loop; | |
4977146a N |
446 | if (c->homehost) { |
447 | int first = st->ss->match_home(st, c->homehost); | |
448 | int last = tst->ss->match_home(tst, c->homehost); | |
5083d66b N |
449 | if (first != last && |
450 | (first == 1 || last == 1)) { | |
451 | /* We can do something */ | |
452 | if (first) {/* just ignore this one */ | |
8cf2eb96 | 453 | if (report_mismatch) |
e7b84f9d | 454 | pr_err("%s misses out due to wrong homehost\n", |
79f9f56d | 455 | devname); |
5083d66b N |
456 | goto loop; |
457 | } else { /* reject all those sofar */ | |
458 | struct mddev_dev *td; | |
8cf2eb96 | 459 | if (report_mismatch) |
e7b84f9d | 460 | pr_err("%s overrides previous devices due to good homehost\n", |
79f9f56d | 461 | devname); |
5083d66b N |
462 | for (td=devlist; td != tmpdev; td=td->next) |
463 | if (td->used == 1) | |
464 | td->used = 0; | |
465 | tmpdev->used = 1; | |
466 | goto loop; | |
467 | } | |
83b6208e NB |
468 | } |
469 | } | |
e7b84f9d | 470 | pr_err("superblock on %s doesn't match others - assembly aborted\n", |
79f9f56d | 471 | devname); |
5083d66b N |
472 | tst->ss->free_super(tst); |
473 | st->ss->free_super(st); | |
474 | dev_policy_free(pol); | |
cbeeb0e5 | 475 | domain_free(domains); |
95425a89 | 476 | return -1; |
83b6208e | 477 | } |
5083d66b | 478 | tmpdev->used = 1; |
64c4757e | 479 | } |
df37ffc0 | 480 | loop: |
cbeeb0e5 | 481 | /* Collect domain information from members only */ |
26b05aea AC |
482 | if (tmpdev && tmpdev->used == 1) { |
483 | if (!pol) | |
0a6bff09 | 484 | pol = devid_policy(rdev); |
cbeeb0e5 | 485 | domain_merge(&domains, pol, tst?tst->ss->name:NULL); |
26b05aea | 486 | } |
4e8d9f0a N |
487 | dev_policy_free(pol); |
488 | pol = NULL; | |
3d2b16e7 NB |
489 | if (tst) |
490 | tst->ss->free_super(tst); | |
811e6cbe NB |
491 | } |
492 | ||
ed7fc6b4 | 493 | /* Check if we found some imsm spares but no members */ |
3c7b4a25 CA |
494 | if ((auto_assem || |
495 | (ident->uuid_set && | |
496 | memcmp(uuid_zero, ident->uuid,sizeof(uuid_zero)) == 0)) && | |
497 | (!st || !st->sb)) | |
ed7fc6b4 AC |
498 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
499 | if (tmpdev->used != 3) | |
500 | continue; | |
501 | tmpdev->used = 1; | |
95425a89 | 502 | content = *contentp; |
ed7fc6b4 AC |
503 | |
504 | if (!st->sb) { | |
505 | /* we need sb from one of the spares */ | |
506 | int dfd = dev_open(tmpdev->devname, O_RDONLY); | |
507 | if (dfd < 0 || | |
508 | st->ss->load_super(st, dfd, NULL)) | |
509 | tmpdev->used = 2; | |
510 | if (dfd > 0) | |
511 | close(dfd); | |
512 | } | |
513 | } | |
514 | ||
cbeeb0e5 AC |
515 | /* Now reject spares that don't match domains of identified members */ |
516 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
cbeeb0e5 AC |
517 | if (tmpdev->used != 3) |
518 | continue; | |
9e04ac1c | 519 | if (!stat_is_blkdev(tmpdev->devname, &rdev)) { |
cbeeb0e5 AC |
520 | tmpdev->used = 2; |
521 | } else { | |
9e04ac1c | 522 | struct dev_policy *pol = devid_policy(rdev); |
a5d10dce N |
523 | int dt = domain_test(domains, pol, NULL); |
524 | if (inargv && dt != 0) | |
525 | /* take this spare as domains match | |
526 | * if there are any */ | |
527 | tmpdev->used = 1; | |
528 | else if (!inargv && dt == 1) | |
529 | /* device wasn't explicitly listed, so need | |
530 | * explicit domain match - which we have */ | |
cbeeb0e5 AC |
531 | tmpdev->used = 1; |
532 | else | |
533 | /* if domains don't match mark as unused */ | |
534 | tmpdev->used = 0; | |
535 | dev_policy_free(pol); | |
536 | } | |
537 | } | |
538 | domain_free(domains); | |
95425a89 N |
539 | *stp = st; |
540 | if (st && st->sb && content == *contentp) | |
541 | st->ss->getinfo_super(st, content, NULL); | |
542 | *contentp = content; | |
543 | ||
544 | return num_devs; | |
545 | } | |
546 | ||
2c355c22 N |
547 | struct devs { |
548 | char *devname; | |
549 | int uptodate; /* set once we decide that this device is as | |
550 | * recent as everything else in the array. | |
551 | */ | |
552 | int included; /* set if the device is already in the array | |
553 | * due to a previous '-I' | |
554 | */ | |
555 | struct mdinfo i; | |
556 | }; | |
557 | ||
558 | static int load_devices(struct devs *devices, char *devmap, | |
df842e69 | 559 | struct mddev_ident *ident, struct supertype **stp, |
2c355c22 N |
560 | struct mddev_dev *devlist, struct context *c, |
561 | struct mdinfo *content, | |
562 | int mdfd, char *mddev, | |
563 | int *most_recentp, int *bestcntp, int **bestp, | |
564 | int inargv) | |
565 | { | |
566 | struct mddev_dev *tmpdev; | |
567 | int devcnt = 0; | |
568 | int nextspare = 0; | |
2c355c22 | 569 | int bitmap_done = 0; |
a792ece6 | 570 | int most_recent = -1; |
2c355c22 N |
571 | int bestcnt = 0; |
572 | int *best = *bestp; | |
df842e69 | 573 | struct supertype *st = *stp; |
2c355c22 N |
574 | |
575 | for (tmpdev = devlist; tmpdev; tmpdev=tmpdev->next) { | |
576 | char *devname = tmpdev->devname; | |
577 | struct stat stb; | |
9ee314da | 578 | struct supertype *tst; |
2c355c22 | 579 | int i; |
e9e6894d | 580 | int dfd; |
563ac108 | 581 | int disk_state; |
2c355c22 N |
582 | |
583 | if (tmpdev->used != 1) | |
584 | continue; | |
585 | /* looks like a good enough match to update the super block if needed */ | |
2c355c22 | 586 | if (c->update) { |
2c355c22 N |
587 | /* prepare useful information in info structures */ |
588 | struct stat stb2; | |
2c355c22 N |
589 | int err; |
590 | fstat(mdfd, &stb2); | |
591 | ||
c5f71c24 JS |
592 | if (strcmp(c->update, "uuid") == 0 && !ident->uuid_set) |
593 | random_uuid((__u8 *)ident->uuid); | |
594 | ||
e6e9dd3f AP |
595 | if (strcmp(c->update, "ppl") == 0 && |
596 | ident->bitmap_fd >= 0) { | |
597 | pr_err("PPL is not compatible with bitmap\n"); | |
d8b01738 GJ |
598 | close(mdfd); |
599 | free(devices); | |
600 | free(devmap); | |
e6e9dd3f AP |
601 | return -1; |
602 | } | |
603 | ||
2c355c22 N |
604 | dfd = dev_open(devname, |
605 | tmpdev->disposition == 'I' | |
606 | ? O_RDWR : (O_RDWR|O_EXCL)); | |
607 | ||
608 | tst = dup_super(st); | |
609 | if (dfd < 0 || tst->ss->load_super(tst, dfd, NULL) != 0) { | |
610 | pr_err("cannot re-read metadata from %s - aborting\n", | |
611 | devname); | |
612 | if (dfd >= 0) | |
613 | close(dfd); | |
614 | close(mdfd); | |
615 | free(devices); | |
616 | free(devmap); | |
9ee314da N |
617 | tst->ss->free_super(tst); |
618 | free(tst); | |
df842e69 | 619 | *stp = st; |
2c355c22 N |
620 | return -1; |
621 | } | |
622 | tst->ss->getinfo_super(tst, content, devmap + devcnt * content->array.raid_disks); | |
623 | ||
624 | memcpy(content->uuid, ident->uuid, 16); | |
625 | strcpy(content->name, ident->name); | |
626 | content->array.md_minor = minor(stb2.st_rdev); | |
627 | ||
628 | if (strcmp(c->update, "byteorder") == 0) | |
629 | err = 0; | |
0aa2f15b GJ |
630 | else if (strcmp(c->update, "home-cluster") == 0) { |
631 | tst->cluster_name = c->homecluster; | |
d80f7aa9 | 632 | err = tst->ss->write_bitmap(tst, dfd, NameUpdate); |
7e6e839a GJ |
633 | } else if (strcmp(c->update, "nodes") == 0) { |
634 | tst->nodes = c->nodes; | |
635 | err = tst->ss->write_bitmap(tst, dfd, NodeNumUpdate); | |
d5ff855d N |
636 | } else if (strcmp(c->update, "revert-reshape") == 0 && |
637 | c->invalid_backup) | |
638 | err = tst->ss->update_super(tst, content, | |
639 | "revert-reshape-nobackup", | |
640 | devname, c->verbose, | |
641 | ident->uuid_set, | |
642 | c->homehost); | |
643 | else | |
2c355c22 N |
644 | err = tst->ss->update_super(tst, content, c->update, |
645 | devname, c->verbose, | |
646 | ident->uuid_set, | |
647 | c->homehost); | |
648 | if (err < 0) { | |
afa368f4 | 649 | if (err == -1) |
7a862a02 | 650 | pr_err("--update=%s not understood for %s metadata\n", |
afa368f4 | 651 | c->update, tst->ss->name); |
2c355c22 N |
652 | tst->ss->free_super(tst); |
653 | free(tst); | |
654 | close(mdfd); | |
655 | close(dfd); | |
656 | free(devices); | |
657 | free(devmap); | |
df842e69 | 658 | *stp = st; |
2c355c22 N |
659 | return -1; |
660 | } | |
661 | if (strcmp(c->update, "uuid")==0 && | |
662 | !ident->uuid_set) { | |
663 | ident->uuid_set = 1; | |
664 | memcpy(ident->uuid, content->uuid, 16); | |
665 | } | |
666 | if (tst->ss->store_super(tst, dfd)) | |
667 | pr_err("Could not re-write superblock on %s.\n", | |
668 | devname); | |
2c355c22 N |
669 | |
670 | if (strcmp(c->update, "uuid")==0 && | |
671 | ident->bitmap_fd >= 0 && !bitmap_done) { | |
672 | if (bitmap_update_uuid(ident->bitmap_fd, | |
673 | content->uuid, | |
674 | tst->ss->swapuuid) != 0) | |
675 | pr_err("Could not update uuid on external bitmap.\n"); | |
676 | else | |
677 | bitmap_done = 1; | |
678 | } | |
32141c17 | 679 | } else { |
e9e6894d N |
680 | dfd = dev_open(devname, |
681 | tmpdev->disposition == 'I' | |
682 | ? O_RDWR : (O_RDWR|O_EXCL)); | |
9ee314da | 683 | tst = dup_super(st); |
2c355c22 N |
684 | |
685 | if (dfd < 0 || tst->ss->load_super(tst, dfd, NULL) != 0) { | |
686 | pr_err("cannot re-read metadata from %s - aborting\n", | |
687 | devname); | |
688 | if (dfd >= 0) | |
689 | close(dfd); | |
690 | close(mdfd); | |
691 | free(devices); | |
692 | free(devmap); | |
9ee314da N |
693 | tst->ss->free_super(tst); |
694 | free(tst); | |
df842e69 | 695 | *stp = st; |
2c355c22 N |
696 | return -1; |
697 | } | |
698 | tst->ss->getinfo_super(tst, content, devmap + devcnt * content->array.raid_disks); | |
2c355c22 N |
699 | } |
700 | ||
e9e6894d N |
701 | fstat(dfd, &stb); |
702 | close(dfd); | |
2c355c22 N |
703 | |
704 | if (c->verbose > 0) | |
aacb2f81 N |
705 | pr_err("%s is identified as a member of %s, slot %d%s.\n", |
706 | devname, mddev, content->disk.raid_disk, | |
707 | (content->disk.state & (1<<MD_DISK_REPLACEMENT)) ? " replacement":""); | |
2c355c22 N |
708 | devices[devcnt].devname = devname; |
709 | devices[devcnt].uptodate = 0; | |
710 | devices[devcnt].included = (tmpdev->disposition == 'I'); | |
711 | devices[devcnt].i = *content; | |
712 | devices[devcnt].i.disk.major = major(stb.st_rdev); | |
713 | devices[devcnt].i.disk.minor = minor(stb.st_rdev); | |
f80057ae | 714 | |
563ac108 GK |
715 | disk_state = devices[devcnt].i.disk.state & ~((1<<MD_DISK_FAILFAST) | |
716 | (1<<MD_DISK_WRITEMOSTLY)); | |
717 | if (disk_state == ((1<<MD_DISK_ACTIVE) | (1<<MD_DISK_SYNC))) { | |
a792ece6 N |
718 | if (most_recent < 0 || |
719 | devices[devcnt].i.events | |
56bbc588 N |
720 | > devices[most_recent].i.events) { |
721 | struct supertype *tmp = tst; | |
722 | tst = st; | |
723 | st = tmp; | |
2c355c22 | 724 | most_recent = devcnt; |
56bbc588 | 725 | } |
a792ece6 | 726 | } |
9ee314da N |
727 | tst->ss->free_super(tst); |
728 | free(tst); | |
f80057ae | 729 | |
2c355c22 N |
730 | if (content->array.level == LEVEL_MULTIPATH) |
731 | /* with multipath, the raid_disk from the superblock is meaningless */ | |
732 | i = devcnt; | |
733 | else | |
734 | i = devices[devcnt].i.disk.raid_disk; | |
69a48116 | 735 | if (i+1 == 0 || i == MD_DISK_ROLE_JOURNAL) { |
aacb2f81 N |
736 | if (nextspare < content->array.raid_disks*2) |
737 | nextspare = content->array.raid_disks*2; | |
2c355c22 N |
738 | i = nextspare++; |
739 | } else { | |
aacb2f81 N |
740 | /* i is raid_disk - double it so there is room for |
741 | * replacements */ | |
742 | i *= 2; | |
743 | if (devices[devcnt].i.disk.state & (1<<MD_DISK_REPLACEMENT)) | |
744 | i++; | |
745 | if (i >= content->array.raid_disks*2 && | |
2c355c22 N |
746 | i >= nextspare) |
747 | nextspare = i+1; | |
748 | } | |
749 | if (i < 10000) { | |
750 | if (i >= bestcnt) { | |
751 | int newbestcnt = i+10; | |
752 | int *newbest = xmalloc(sizeof(int)*newbestcnt); | |
753 | int c; | |
754 | for (c=0; c < newbestcnt; c++) | |
755 | if (c < bestcnt) | |
756 | newbest[c] = best[c]; | |
757 | else | |
758 | newbest[c] = -1; | |
759 | if (best)free(best); | |
760 | best = newbest; | |
761 | bestcnt = newbestcnt; | |
762 | } | |
763 | if (best[i] >=0 && | |
fc54fe7a JS |
764 | devices[best[i]].i.events == |
765 | devices[devcnt].i.events && | |
d16a7494 JS |
766 | (devices[best[i]].i.disk.minor != |
767 | devices[devcnt].i.disk.minor) && | |
fc54fe7a JS |
768 | st->ss == &super0 && |
769 | content->array.level != LEVEL_MULTIPATH) { | |
2c355c22 N |
770 | /* two different devices with identical superblock. |
771 | * Could be a mis-detection caused by overlapping | |
772 | * partitions. fail-safe. | |
773 | */ | |
7a862a02 N |
774 | pr_err("WARNING %s and %s appear to have very similar superblocks.\n" |
775 | " If they are really different, please --zero the superblock on one\n" | |
776 | " If they are the same or overlap, please remove one from %s.\n", | |
2c355c22 N |
777 | devices[best[i]].devname, devname, |
778 | inargv ? "the list" : | |
779 | "the\n DEVICE list in mdadm.conf" | |
780 | ); | |
781 | close(mdfd); | |
782 | free(devices); | |
783 | free(devmap); | |
df842e69 | 784 | *stp = st; |
2c355c22 N |
785 | return -1; |
786 | } | |
d7be7d87 JS |
787 | if (best[i] == -1 || (devices[best[i]].i.events |
788 | < devices[devcnt].i.events)) | |
2c355c22 | 789 | best[i] = devcnt; |
7298c9a6 MT |
790 | else if (st->ss == &super_imsm) |
791 | best[i+1] = devcnt; | |
2c355c22 N |
792 | } |
793 | devcnt++; | |
794 | } | |
a792ece6 N |
795 | if (most_recent >= 0) |
796 | *most_recentp = most_recent; | |
2c355c22 N |
797 | *bestcntp = bestcnt; |
798 | *bestp = best; | |
df842e69 | 799 | *stp = st; |
2c355c22 N |
800 | return devcnt; |
801 | } | |
802 | ||
9f5470ce N |
803 | static int force_array(struct mdinfo *content, |
804 | struct devs *devices, | |
805 | int *best, int bestcnt, char *avail, | |
806 | int most_recent, | |
807 | struct supertype *st, | |
808 | struct context *c) | |
809 | { | |
810 | int okcnt = 0; | |
811 | while (!enough(content->array.level, content->array.raid_disks, | |
812 | content->array.layout, 1, | |
fc54fe7a | 813 | avail) || |
9f5470ce N |
814 | (content->reshape_active && content->delta_disks > 0 && |
815 | !enough(content->array.level, (content->array.raid_disks | |
816 | - content->delta_disks), | |
fc54fe7a | 817 | content->new_layout, 1, avail))) { |
9f5470ce N |
818 | /* Choose the newest best drive which is |
819 | * not up-to-date, update the superblock | |
820 | * and add it. | |
821 | */ | |
822 | int fd; | |
823 | struct supertype *tst; | |
824 | unsigned long long current_events; | |
825 | int chosen_drive = -1; | |
826 | int i; | |
827 | ||
f81a2b56 N |
828 | for (i = 0; |
829 | i < content->array.raid_disks * 2 && i < bestcnt; | |
830 | i += 2) { | |
9f5470ce | 831 | int j = best[i]; |
87af7267 N |
832 | if (j < 0) |
833 | continue; | |
834 | if (devices[j].uptodate) | |
835 | continue; | |
836 | if (devices[j].i.recovery_start != MaxSector) { | |
837 | int delta; | |
838 | if (!devices[j].i.reshape_active || | |
839 | devices[j].i.delta_disks <= 0) | |
840 | continue; | |
841 | /* When increasing number of devices, an | |
842 | * added device also appears to be | |
843 | * recovering. It is safe to include it | |
844 | * as long as it won't be a source of | |
845 | * data. | |
846 | * For now, just allow for last data | |
847 | * devices in RAID4 or last devices in RAID4/5/6. | |
848 | */ | |
849 | delta = devices[j].i.delta_disks; | |
850 | if (devices[j].i.array.level >= 4 && | |
851 | devices[j].i.array.level <= 6 && | |
852 | i/2 >= content->array.raid_disks - delta) | |
853 | /* OK */; | |
854 | else if (devices[j].i.array.level == 4 && | |
855 | i/2 >= content->array.raid_disks - delta - 1) | |
856 | /* OK */; | |
857 | else | |
858 | continue; | |
62f1aee7 BC |
859 | } else if (devices[j].i.reshape_active != |
860 | content->reshape_active || | |
861 | (devices[j].i.reshape_active && | |
862 | devices[j].i.reshape_progress != | |
863 | content->reshape_progress)) | |
864 | /* Here, it may be a source of data. If two | |
865 | * devices claim different progresses, it | |
866 | * means that reshape boundaries differ for | |
867 | * their own devices. Kernel will only treat | |
868 | * the first one as reshape progress and | |
869 | * go on. It may cause disaster, so avoid it. | |
870 | */ | |
871 | continue; | |
87af7267 | 872 | if (chosen_drive < 0 || |
9f5470ce | 873 | devices[j].i.events |
87af7267 | 874 | > devices[chosen_drive].i.events) |
9f5470ce N |
875 | chosen_drive = j; |
876 | } | |
877 | if (chosen_drive < 0) | |
878 | break; | |
879 | current_events = devices[chosen_drive].i.events; | |
880 | add_another: | |
881 | if (c->verbose >= 0) | |
ebf3be99 | 882 | pr_err("forcing event count in %s(%d) from %d up to %d\n", |
79f9f56d N |
883 | devices[chosen_drive].devname, |
884 | devices[chosen_drive].i.disk.raid_disk, | |
885 | (int)(devices[chosen_drive].i.events), | |
886 | (int)(devices[most_recent].i.events)); | |
9f5470ce N |
887 | fd = dev_open(devices[chosen_drive].devname, |
888 | devices[chosen_drive].included ? O_RDWR | |
889 | : (O_RDWR|O_EXCL)); | |
890 | if (fd < 0) { | |
891 | pr_err("Couldn't open %s for write - not updating\n", | |
79f9f56d | 892 | devices[chosen_drive].devname); |
9f5470ce N |
893 | devices[chosen_drive].i.events = 0; |
894 | continue; | |
895 | } | |
896 | tst = dup_super(st); | |
897 | if (tst->ss->load_super(tst,fd, NULL)) { | |
898 | close(fd); | |
899 | pr_err("RAID superblock disappeared from %s - not updating.\n", | |
79f9f56d | 900 | devices[chosen_drive].devname); |
9f5470ce N |
901 | devices[chosen_drive].i.events = 0; |
902 | continue; | |
903 | } | |
904 | content->events = devices[most_recent].i.events; | |
905 | tst->ss->update_super(tst, content, "force-one", | |
79f9f56d N |
906 | devices[chosen_drive].devname, c->verbose, |
907 | 0, NULL); | |
9f5470ce N |
908 | |
909 | if (tst->ss->store_super(tst, fd)) { | |
910 | close(fd); | |
911 | pr_err("Could not re-write superblock on %s\n", | |
79f9f56d | 912 | devices[chosen_drive].devname); |
9f5470ce N |
913 | devices[chosen_drive].i.events = 0; |
914 | tst->ss->free_super(tst); | |
915 | continue; | |
916 | } | |
917 | close(fd); | |
918 | devices[chosen_drive].i.events = devices[most_recent].i.events; | |
919 | devices[chosen_drive].uptodate = 1; | |
920 | avail[chosen_drive] = 1; | |
921 | okcnt++; | |
922 | tst->ss->free_super(tst); | |
9f5470ce N |
923 | /* If there are any other drives of the same vintage, |
924 | * add them in as well. We can't lose and we might gain | |
925 | */ | |
f81a2b56 N |
926 | for (i = 0; |
927 | i < content->array.raid_disks * 2 && i < bestcnt ; | |
928 | i += 2) { | |
9f5470ce N |
929 | int j = best[i]; |
930 | if (j >= 0 && | |
931 | !devices[j].uptodate && | |
932 | devices[j].i.recovery_start == MaxSector && | |
62f1aee7 BC |
933 | devices[j].i.events == current_events && |
934 | ((!devices[j].i.reshape_active && | |
935 | !content->reshape_active) || | |
936 | (devices[j].i.reshape_active == | |
937 | content->reshape_active && | |
938 | devices[j].i.reshape_progress == | |
939 | content->reshape_progress))) { | |
9f5470ce N |
940 | chosen_drive = j; |
941 | goto add_another; | |
942 | } | |
943 | } | |
944 | } | |
945 | return okcnt; | |
946 | } | |
947 | ||
ddc1b11f N |
948 | static int start_array(int mdfd, |
949 | char *mddev, | |
950 | struct mdinfo *content, | |
951 | struct supertype *st, | |
952 | struct mddev_ident *ident, | |
953 | int *best, int bestcnt, | |
954 | int chosen_drive, | |
955 | struct devs *devices, | |
956 | unsigned int okcnt, | |
957 | unsigned int sparecnt, | |
958 | unsigned int rebuilding_cnt, | |
69a48116 | 959 | unsigned int journalcnt, |
ddc1b11f N |
960 | struct context *c, |
961 | int clean, char *avail, | |
8cde842b | 962 | int start_partial_ok, |
71417de6 | 963 | int err_ok, |
051f3265 | 964 | int was_forced |
ddc1b11f N |
965 | ) |
966 | { | |
967 | int rv; | |
968 | int i; | |
969 | unsigned int req_cnt; | |
970 | ||
051f3265 | 971 | if (content->journal_device_required && (content->journal_clean == 0)) { |
69a48116 SL |
972 | if (!c->force) { |
973 | pr_err("Not safe to assemble with missing or stale journal device, consider --force.\n"); | |
974 | return 1; | |
975 | } | |
976 | pr_err("Journal is missing or stale, starting array read only.\n"); | |
977 | c->readonly = 1; | |
978 | } | |
979 | ||
e97a7cd0 AP |
980 | if (content->consistency_policy == CONSISTENCY_POLICY_PPL) |
981 | clean = 1; | |
982 | ||
ddc1b11f | 983 | rv = set_array_info(mdfd, st, content); |
71417de6 | 984 | if (rv && !err_ok) { |
ddc1b11f N |
985 | pr_err("failed to set array info for %s: %s\n", |
986 | mddev, strerror(errno)); | |
987 | return 1; | |
988 | } | |
989 | if (ident->bitmap_fd >= 0) { | |
990 | if (ioctl(mdfd, SET_BITMAP_FILE, ident->bitmap_fd) != 0) { | |
991 | pr_err("SET_BITMAP_FILE failed.\n"); | |
992 | return 1; | |
993 | } | |
994 | } else if (ident->bitmap_file) { | |
995 | /* From config file */ | |
996 | int bmfd = open(ident->bitmap_file, O_RDWR); | |
997 | if (bmfd < 0) { | |
998 | pr_err("Could not open bitmap file %s\n", | |
999 | ident->bitmap_file); | |
1000 | return 1; | |
1001 | } | |
1002 | if (ioctl(mdfd, SET_BITMAP_FILE, bmfd) != 0) { | |
1003 | pr_err("Failed to set bitmapfile for %s\n", mddev); | |
1004 | close(bmfd); | |
1005 | return 1; | |
1006 | } | |
1007 | close(bmfd); | |
1008 | } | |
1009 | ||
1010 | /* First, add the raid disks, but add the chosen one last */ | |
94b53b77 | 1011 | for (i = 0; i <= bestcnt; i++) { |
ddc1b11f N |
1012 | int j; |
1013 | if (i < bestcnt) { | |
1014 | j = best[i]; | |
1015 | if (j == chosen_drive) | |
1016 | continue; | |
1017 | } else | |
1018 | j = chosen_drive; | |
1019 | ||
1020 | if (j >= 0 && !devices[j].included) { | |
94b53b77 JS |
1021 | int dfd; |
1022 | ||
1023 | dfd = dev_open(devices[j].devname, O_RDWR|O_EXCL); | |
ddc1b11f N |
1024 | if (dfd >= 0) { |
1025 | remove_partitions(dfd); | |
1026 | close(dfd); | |
1027 | } | |
1028 | rv = add_disk(mdfd, st, content, &devices[j].i); | |
1029 | ||
1030 | if (rv) { | |
7a862a02 | 1031 | pr_err("failed to add %s to %s: %s\n", |
94b53b77 | 1032 | devices[j].devname, mddev, |
ddc1b11f | 1033 | strerror(errno)); |
94b53b77 JS |
1034 | if (i < content->array.raid_disks * 2 || |
1035 | i == bestcnt) | |
ddc1b11f N |
1036 | okcnt--; |
1037 | else | |
1038 | sparecnt--; | |
94b53b77 | 1039 | } else if (c->verbose > 0) { |
aacb2f81 | 1040 | pr_err("added %s to %s as %d%s%s\n", |
ddc1b11f N |
1041 | devices[j].devname, mddev, |
1042 | devices[j].i.disk.raid_disk, | |
1043 | devices[j].uptodate?"": | |
aacb2f81 | 1044 | " (possibly out of date)", |
94b53b77 JS |
1045 | (devices[j].i.disk.state & |
1046 | (1<<MD_DISK_REPLACEMENT)) ? | |
1047 | " replacement":""); | |
1048 | } | |
ddc1b11f N |
1049 | } else if (j >= 0) { |
1050 | if (c->verbose > 0) | |
1051 | pr_err("%s is already in %s as %d\n", | |
1052 | devices[j].devname, mddev, | |
1053 | devices[j].i.disk.raid_disk); | |
94b53b77 JS |
1054 | } else if (c->verbose > 0 && |
1055 | i < content->array.raid_disks * 2 && (i & 1) == 0) | |
ddc1b11f | 1056 | pr_err("no uptodate device for slot %d of %s\n", |
ee466574 | 1057 | i/2, mddev); |
ddc1b11f N |
1058 | } |
1059 | ||
1060 | if (content->array.level == LEVEL_CONTAINER) { | |
1061 | if (c->verbose >= 0) { | |
7a862a02 | 1062 | pr_err("Container %s has been assembled with %d drive%s", |
94b53b77 JS |
1063 | mddev, okcnt + sparecnt + journalcnt, |
1064 | okcnt + sparecnt + journalcnt == 1 ? "" : "s"); | |
ddc1b11f | 1065 | if (okcnt < (unsigned)content->array.raid_disks) |
b0681598 | 1066 | fprintf(stderr, " (out of %d)\n", |
ddc1b11f | 1067 | content->array.raid_disks); |
b0681598 MD |
1068 | else { |
1069 | fprintf(stderr, "\n"); | |
1070 | sysfs_rules_apply(mddev, content); | |
1071 | } | |
ddc1b11f | 1072 | } |
0c21b485 PB |
1073 | |
1074 | if (st->ss->validate_container) { | |
1075 | struct mdinfo *devices_list; | |
94b53b77 | 1076 | struct mdinfo *info_devices; |
0c21b485 | 1077 | unsigned int count; |
94b53b77 | 1078 | |
0c21b485 | 1079 | devices_list = NULL; |
94b53b77 JS |
1080 | info_devices = xmalloc(sizeof(struct mdinfo) * |
1081 | (okcnt + sparecnt)); | |
1082 | for (count = 0; count < okcnt + sparecnt; count++) { | |
0c21b485 PB |
1083 | info_devices[count] = devices[count].i; |
1084 | info_devices[count].next = devices_list; | |
1085 | devices_list = &info_devices[count]; | |
1086 | } | |
1087 | if (st->ss->validate_container(devices_list)) | |
1088 | pr_err("Mismatch detected!\n"); | |
1089 | free(info_devices); | |
1090 | } | |
1091 | ||
ddc1b11f N |
1092 | st->ss->free_super(st); |
1093 | sysfs_uevent(content, "change"); | |
284546ef N |
1094 | if (err_ok && okcnt < (unsigned)content->array.raid_disks) |
1095 | /* Was partial, is still partial, so signal an error | |
1096 | * to ensure we don't retry */ | |
1097 | return 1; | |
ddc1b11f N |
1098 | return 0; |
1099 | } | |
1100 | ||
1101 | /* Get number of in-sync devices according to the superblock. | |
1102 | * We must have this number to start the array without -s or -R | |
1103 | */ | |
1104 | req_cnt = content->array.working_disks; | |
1105 | ||
1106 | if (c->runstop == 1 || | |
1107 | (c->runstop <= 0 && | |
94b53b77 JS |
1108 | (enough(content->array.level, content->array.raid_disks, |
1109 | content->array.layout, clean, avail) && | |
1110 | (okcnt + rebuilding_cnt >= req_cnt || start_partial_ok)))) { | |
ddc1b11f N |
1111 | /* This array is good-to-go. |
1112 | * If a reshape is in progress then we might need to | |
1113 | * continue monitoring it. In that case we start | |
1114 | * it read-only and let the grow code make it writable. | |
1115 | */ | |
1116 | int rv; | |
94b53b77 | 1117 | |
ddc1b11f N |
1118 | if (content->reshape_active && |
1119 | !(content->reshape_active & RESHAPE_NO_BACKUP) && | |
1120 | content->delta_disks <= 0) { | |
be7c26b4 N |
1121 | if (!c->backup_file) { |
1122 | pr_err("%s: Need a backup file to complete reshape of this array.\n", | |
1123 | mddev); | |
1124 | pr_err("Please provided one with \"--backup-file=...\"\n"); | |
1125 | if (c->update && | |
1126 | strcmp(c->update, "revert-reshape") == 0) | |
1127 | pr_err("(Don't specify --update=revert-reshape again, that part succeeded.)\n"); | |
1128 | return 1; | |
1129 | } | |
ddc1b11f N |
1130 | rv = sysfs_set_str(content, NULL, |
1131 | "array_state", "readonly"); | |
1132 | if (rv == 0) | |
1133 | rv = Grow_continue(mdfd, st, content, | |
06e293d0 | 1134 | c->backup_file, 0, |
ddc1b11f N |
1135 | c->freeze_reshape); |
1136 | } else if (c->readonly && | |
94b53b77 JS |
1137 | sysfs_attribute_available(content, NULL, |
1138 | "array_state")) { | |
ddc1b11f N |
1139 | rv = sysfs_set_str(content, NULL, |
1140 | "array_state", "readonly"); | |
1141 | } else | |
ddc1b11f | 1142 | rv = ioctl(mdfd, RUN_ARRAY, NULL); |
8832342d | 1143 | reopen_mddev(mdfd); /* drop O_EXCL */ |
ddc1b11f | 1144 | if (rv == 0) { |
b0681598 | 1145 | sysfs_rules_apply(mddev, content); |
ddc1b11f N |
1146 | if (c->verbose >= 0) { |
1147 | pr_err("%s has been started with %d drive%s", | |
1148 | mddev, okcnt, okcnt==1?"":"s"); | |
1149 | if (okcnt < (unsigned)content->array.raid_disks) | |
94b53b77 JS |
1150 | fprintf(stderr, " (out of %d)", |
1151 | content->array.raid_disks); | |
ddc1b11f | 1152 | if (rebuilding_cnt) |
94b53b77 JS |
1153 | fprintf(stderr, "%s %d rebuilding", |
1154 | sparecnt?",":" and", | |
1155 | rebuilding_cnt); | |
ddc1b11f | 1156 | if (sparecnt) |
94b53b77 JS |
1157 | fprintf(stderr, " and %d spare%s", |
1158 | sparecnt, | |
1159 | sparecnt == 1 ? "" : "s"); | |
051f3265 | 1160 | if (content->journal_clean) |
94b53b77 JS |
1161 | fprintf(stderr, " and %d journal", |
1162 | journalcnt); | |
ddc1b11f N |
1163 | fprintf(stderr, ".\n"); |
1164 | } | |
1165 | if (content->reshape_active && | |
1166 | content->array.level >= 4 && | |
1167 | content->array.level <= 6) { | |
1168 | /* might need to increase the size | |
1169 | * of the stripe cache - default is 256 | |
1170 | */ | |
56fcbcbb | 1171 | int chunk_size = content->array.chunk_size; |
94b53b77 | 1172 | |
56fcbcbb N |
1173 | if (content->reshape_active && |
1174 | content->new_chunk > chunk_size) | |
1175 | chunk_size = content->new_chunk; | |
29a312f2 | 1176 | if (256 < 4 * ((chunk_size+4065)/4096)) { |
94b53b77 JS |
1177 | struct mdinfo *sra; |
1178 | ||
1179 | sra = sysfs_read(mdfd, NULL, 0); | |
ddc1b11f N |
1180 | if (sra) |
1181 | sysfs_set_num(sra, NULL, | |
1182 | "stripe_cache_size", | |
29a312f2 | 1183 | (4 * chunk_size / 4096) + 1); |
ddc1b11f N |
1184 | sysfs_free(sra); |
1185 | } | |
1186 | } | |
1187 | if (okcnt < (unsigned)content->array.raid_disks) { | |
1188 | /* If any devices did not get added | |
1189 | * because the kernel rejected them based | |
1190 | * on event count, try adding them | |
1191 | * again providing the action policy is | |
1192 | * 're-add' or greater. The bitmap | |
1193 | * might allow them to be included, or | |
1194 | * they will become spares. | |
1195 | */ | |
1196 | for (i = 0; i < bestcnt; i++) { | |
1197 | int j = best[i]; | |
1198 | if (j >= 0 && !devices[j].uptodate) { | |
1199 | if (!disk_action_allows(&devices[j].i, st->ss->name, act_re_add)) | |
1200 | continue; | |
1201 | rv = add_disk(mdfd, st, content, | |
1202 | &devices[j].i); | |
1203 | if (rv == 0 && c->verbose >= 0) | |
1204 | pr_err("%s has been re-added.\n", | |
1205 | devices[j].devname); | |
1206 | } | |
1207 | } | |
1208 | } | |
8cde842b N |
1209 | if (content->array.level == 6 && |
1210 | okcnt + 1 == (unsigned)content->array.raid_disks && | |
1211 | was_forced) { | |
94b53b77 JS |
1212 | struct mdinfo *sra; |
1213 | ||
1214 | sra = sysfs_read(mdfd, NULL, 0); | |
8cde842b N |
1215 | if (sra) |
1216 | sysfs_set_str(sra, NULL, | |
1217 | "sync_action", "repair"); | |
1218 | sysfs_free(sra); | |
1219 | } | |
ddc1b11f N |
1220 | return 0; |
1221 | } | |
94b53b77 | 1222 | pr_err("failed to RUN_ARRAY %s: %s\n", mddev, strerror(errno)); |
ddc1b11f N |
1223 | |
1224 | if (!enough(content->array.level, content->array.raid_disks, | |
1225 | content->array.layout, 1, avail)) | |
7a862a02 | 1226 | pr_err("Not enough devices to start the array.\n"); |
ddc1b11f N |
1227 | else if (!enough(content->array.level, |
1228 | content->array.raid_disks, | |
94b53b77 | 1229 | content->array.layout, clean, avail)) |
7a862a02 | 1230 | pr_err("Not enough devices to start the array while not clean - consider --force.\n"); |
ddc1b11f N |
1231 | |
1232 | return 1; | |
1233 | } | |
1234 | if (c->runstop == -1) { | |
1235 | pr_err("%s assembled from %d drive%s", | |
94b53b77 | 1236 | mddev, okcnt, okcnt == 1 ? "" : "s"); |
ddc1b11f | 1237 | if (okcnt != (unsigned)content->array.raid_disks) |
94b53b77 JS |
1238 | fprintf(stderr, " (out of %d)", |
1239 | content->array.raid_disks); | |
ddc1b11f N |
1240 | fprintf(stderr, ", but not started.\n"); |
1241 | return 2; | |
1242 | } | |
1243 | if (c->verbose >= -1) { | |
94b53b77 JS |
1244 | pr_err("%s assembled from %d drive%s", |
1245 | mddev, okcnt, okcnt == 1 ? "" : "s"); | |
ddc1b11f | 1246 | if (rebuilding_cnt) |
94b53b77 JS |
1247 | fprintf(stderr, "%s %d rebuilding", |
1248 | sparecnt ? "," : " and", rebuilding_cnt); | |
ddc1b11f | 1249 | if (sparecnt) |
94b53b77 JS |
1250 | fprintf(stderr, " and %d spare%s", sparecnt, |
1251 | sparecnt == 1 ? "" : "s"); | |
ddc1b11f N |
1252 | if (!enough(content->array.level, content->array.raid_disks, |
1253 | content->array.layout, 1, avail)) | |
1254 | fprintf(stderr, " - not enough to start the array.\n"); | |
1255 | else if (!enough(content->array.level, | |
1256 | content->array.raid_disks, | |
94b53b77 | 1257 | content->array.layout, clean, avail)) |
7a862a02 | 1258 | fprintf(stderr, " - not enough to start the array while not clean - consider --force.\n"); |
ddc1b11f N |
1259 | else { |
1260 | if (req_cnt == (unsigned)content->array.raid_disks) | |
94b53b77 JS |
1261 | fprintf(stderr, " - need all %d to start it", |
1262 | req_cnt); | |
ddc1b11f | 1263 | else |
aacb2f81 | 1264 | fprintf(stderr, " - need %d to start", req_cnt); |
ddc1b11f N |
1265 | fprintf(stderr, " (use --run to insist).\n"); |
1266 | } | |
1267 | } | |
1268 | return 1; | |
1269 | } | |
1270 | ||
95425a89 N |
1271 | int Assemble(struct supertype *st, char *mddev, |
1272 | struct mddev_ident *ident, | |
1273 | struct mddev_dev *devlist, | |
1274 | struct context *c) | |
1275 | { | |
1276 | /* | |
1277 | * The task of Assemble is to find a collection of | |
1278 | * devices that should (according to their superblocks) | |
1279 | * form an array, and to give this collection to the MD driver. | |
1280 | * In Linux-2.4 and later, this involves submitting a | |
1281 | * SET_ARRAY_INFO ioctl with no arg - to prepare | |
1282 | * the array - and then submit a number of | |
1283 | * ADD_NEW_DISK ioctls to add disks into | |
1284 | * the array. Finally RUN_ARRAY might | |
1285 | * be submitted to start the array. | |
1286 | * | |
1287 | * Much of the work of Assemble is in finding and/or | |
1288 | * checking the disks to make sure they look right. | |
1289 | * | |
1290 | * If mddev is not set, then scan must be set and we | |
1291 | * read through the config file for dev+uuid mapping | |
1292 | * We recurse, setting mddev, for each device that | |
1293 | * - isn't running | |
1294 | * - has a valid uuid (or any uuid if !uuidset) | |
1295 | * | |
1296 | * If mddev is set, we try to determine state of md. | |
1297 | * check version - must be at least 0.90.0 | |
1298 | * check kernel version. must be at least 2.4. | |
1299 | * If not, we can possibly fall back on START_ARRAY | |
1300 | * Try to GET_ARRAY_INFO. | |
1301 | * If possible, give up | |
1302 | * If not, try to STOP_ARRAY just to make sure | |
1303 | * | |
1304 | * If !uuidset and scan, look in conf-file for uuid | |
1305 | * If not found, give up | |
1306 | * If !devlist and scan and uuidset, get list of devs from conf-file | |
1307 | * | |
1308 | * For each device: | |
1309 | * Check superblock - discard if bad | |
1310 | * Check uuid (set if we don't have one) - discard if no match | |
1311 | * Check superblock similarity if we have a superblock - discard if different | |
1312 | * Record events, devicenum | |
1313 | * This should give us a list of devices for the array | |
1314 | * We should collect the most recent event number | |
1315 | * | |
1316 | * Count disks with recent enough event count | |
1317 | * While force && !enough disks | |
1318 | * Choose newest rejected disks, update event count | |
1319 | * mark clean and rewrite superblock | |
1320 | * If recent kernel: | |
1321 | * SET_ARRAY_INFO | |
1322 | * foreach device with recent events : ADD_NEW_DISK | |
1323 | * if runstop == 1 || "enough" disks and runstop==0 -> RUN_ARRAY | |
1324 | * If old kernel: | |
1325 | * Check the device numbers in superblock are right | |
1326 | * update superblock if any changes | |
1327 | * START_ARRAY | |
1328 | * | |
1329 | */ | |
57908e9e GJ |
1330 | int rv = -1; |
1331 | int mdfd = -1; | |
95425a89 N |
1332 | int clean; |
1333 | int auto_assem = (mddev == NULL && !ident->uuid_set && | |
fc54fe7a JS |
1334 | ident->super_minor == UnSet && ident->name[0] == 0 && |
1335 | (ident->container == NULL || ident->member == NULL)); | |
57908e9e | 1336 | struct devs *devices = NULL; |
95425a89 N |
1337 | char *devmap; |
1338 | int *best = NULL; /* indexed by raid_disk */ | |
1339 | int bestcnt = 0; | |
2c355c22 | 1340 | int devcnt; |
69a48116 | 1341 | unsigned int okcnt, sparecnt, rebuilding_cnt, replcnt, journalcnt; |
dbfbca43 | 1342 | int journal_clean = 0; |
95425a89 | 1343 | int i; |
8cde842b | 1344 | int was_forced = 0; |
95425a89 N |
1345 | int most_recent = 0; |
1346 | int chosen_drive; | |
1347 | int change = 0; | |
1348 | int inargv = 0; | |
95425a89 N |
1349 | int start_partial_ok = (c->runstop >= 0) && |
1350 | (c->force || devlist==NULL || auto_assem); | |
1351 | int num_devs; | |
1352 | struct mddev_dev *tmpdev; | |
1353 | struct mdinfo info; | |
1354 | struct mdinfo *content = NULL; | |
1355 | struct mdinfo *pre_exist = NULL; | |
1356 | char *avail; | |
95425a89 N |
1357 | char *name = NULL; |
1358 | char chosen_name[1024]; | |
1359 | struct map_ent *map = NULL; | |
1360 | struct map_ent *mp; | |
1361 | ||
95425a89 N |
1362 | /* |
1363 | * If any subdevs are listed, then any that don't | |
1364 | * match ident are discarded. Remainder must all match and | |
1365 | * become the array. | |
1366 | * If no subdevs, then we scan all devices in the config file, but | |
1367 | * there must be something in the identity | |
1368 | */ | |
1369 | ||
1370 | if (!devlist && | |
1371 | ident->uuid_set == 0 && | |
1372 | (ident->super_minor < 0 || ident->super_minor == UnSet) && | |
1373 | ident->name[0] == 0 && | |
1374 | (ident->container == NULL || ident->member == NULL) && | |
1375 | ident->devices == NULL) { | |
1376 | pr_err("No identity information available for %s - cannot assemble.\n", | |
1377 | mddev ? mddev : "further assembly"); | |
1378 | return 1; | |
1379 | } | |
1380 | ||
1381 | if (devlist == NULL) | |
1382 | devlist = conf_get_devs(); | |
1383 | else if (mddev) | |
1384 | inargv = 1; | |
1385 | ||
79f9f56d | 1386 | try_again: |
95425a89 N |
1387 | /* We come back here when doing auto-assembly and attempting some |
1388 | * set of devices failed. Those are now marked as ->used==2 and | |
1389 | * we ignore them and try again | |
1390 | */ | |
1391 | if (!st && ident->st) | |
1392 | st = ident->st; | |
1393 | if (c->verbose>0) | |
1394 | pr_err("looking for devices for %s\n", | |
1395 | mddev ? mddev : "further assembly"); | |
1396 | ||
1397 | content = &info; | |
9f2e55a4 | 1398 | if (st && c->force) |
cb8f6859 | 1399 | st->ignore_hw_compat = 1; |
95425a89 N |
1400 | num_devs = select_devices(devlist, ident, &st, &content, c, |
1401 | inargv, auto_assem); | |
1402 | if (num_devs < 0) | |
1403 | return 1; | |
1011e834 | 1404 | |
98dbd966 | 1405 | if (!st || !st->sb || !content) |
c30e5369 N |
1406 | return 2; |
1407 | ||
0431869c N |
1408 | /* We have a full set of devices - we now need to find the |
1409 | * array device. | |
1410 | * However there is a risk that we are racing with "mdadm -I" | |
1411 | * and the array is already partially assembled - we will have | |
1412 | * rejected any devices already in this address. | |
1413 | * So we take a lock on the map file - to prevent further races - | |
1414 | * and look for the uuid in there. If found and the array is | |
1415 | * active, we abort. If found and the array is not active | |
1416 | * we commit to that md device and add all the contained devices | |
1417 | * to our list. We flag them so that we don't try to re-add, | |
1418 | * but can remove if they turn out to not be wanted. | |
1419 | */ | |
1420 | if (map_lock(&map)) | |
1421 | pr_err("failed to get exclusive lock on mapfile - continue anyway...\n"); | |
ec6db5ba N |
1422 | if (c->update && strcmp(c->update,"uuid") == 0) |
1423 | mp = NULL; | |
1424 | else | |
1425 | mp = map_by_uuid(&map, content->uuid); | |
0431869c N |
1426 | if (mp) { |
1427 | struct mdinfo *dv; | |
1428 | /* array already exists. */ | |
4dd2df09 | 1429 | pre_exist = sysfs_read(-1, mp->devnm, GET_LEVEL|GET_DEVS); |
0431869c N |
1430 | if (pre_exist->array.level != UnSet) { |
1431 | pr_err("Found some drive for an array that is already active: %s\n", | |
1432 | mp->path); | |
1433 | pr_err("giving up.\n"); | |
57908e9e | 1434 | goto out; |
0431869c N |
1435 | } |
1436 | for (dv = pre_exist->devs; dv; dv = dv->next) { | |
1437 | /* We want to add this device to our list, | |
1438 | * but it could already be there if "mdadm -I" | |
1439 | * started *after* we checked for O_EXCL. | |
1440 | * If we add it to the top of the list | |
1441 | * it will be preferred over later copies. | |
1442 | */ | |
1443 | struct mddev_dev *newdev; | |
1444 | char *devname = map_dev(dv->disk.major, | |
1445 | dv->disk.minor, | |
1446 | 0); | |
1447 | if (!devname) | |
1448 | continue; | |
1449 | newdev = xmalloc(sizeof(*newdev)); | |
1450 | newdev->devname = devname; | |
1451 | newdev->disposition = 'I'; | |
1452 | newdev->used = 1; | |
1453 | newdev->next = devlist; | |
1454 | devlist = newdev; | |
1455 | num_devs++; | |
1456 | } | |
1457 | strcpy(chosen_name, mp->path); | |
1458 | if (c->verbose > 0 || mddev == NULL || | |
1459 | strcmp(mddev, chosen_name) != 0) | |
1460 | pr_err("Merging with already-assembled %s\n", | |
1461 | chosen_name); | |
4dd2df09 | 1462 | mdfd = open_dev_excl(mp->devnm); |
0431869c N |
1463 | } else { |
1464 | int trustworthy = FOREIGN; | |
1465 | name = content->name; | |
1466 | switch (st->ss->match_home(st, c->homehost) | |
1467 | ?: st->ss->match_home(st, "any")) { | |
1468 | case 1: | |
1469 | trustworthy = LOCAL; | |
1470 | name = strchr(content->name, ':'); | |
1471 | if (name) | |
1472 | name++; | |
1473 | else | |
1474 | name = content->name; | |
1475 | break; | |
1476 | } | |
1477 | if (!auto_assem) | |
1478 | /* If the array is listed in mdadm.conf or on | |
1479 | * command line, then we trust the name | |
1480 | * even if the array doesn't look local | |
1481 | */ | |
1482 | trustworthy = LOCAL; | |
ac2ecf55 | 1483 | |
0431869c N |
1484 | if (name[0] == 0 && |
1485 | content->array.level == LEVEL_CONTAINER) { | |
1486 | name = content->text_version; | |
1487 | trustworthy = METADATA; | |
1488 | } | |
0ac91628 | 1489 | |
0431869c N |
1490 | if (name[0] && trustworthy != LOCAL && |
1491 | ! c->require_homehost && | |
1492 | conf_name_is_free(name)) | |
1493 | trustworthy = LOCAL; | |
0ac91628 | 1494 | |
0431869c N |
1495 | if (trustworthy == LOCAL && |
1496 | strchr(name, ':')) | |
1497 | /* Ignore 'host:' prefix of name */ | |
1498 | name = strchr(name, ':')+1; | |
7cdc0872 | 1499 | |
0431869c | 1500 | mdfd = create_mddev(mddev, name, ident->autof, trustworthy, |
cd6cbb08 | 1501 | chosen_name, 0); |
0431869c | 1502 | } |
c30e5369 N |
1503 | if (mdfd < 0) { |
1504 | st->ss->free_super(st); | |
c30e5369 | 1505 | if (auto_assem) |
60e1bc1a | 1506 | goto try_again; |
57908e9e | 1507 | goto out; |
c30e5369 | 1508 | } |
a04d5763 | 1509 | mddev = chosen_name; |
0431869c | 1510 | if (pre_exist == NULL) { |
4dd2df09 | 1511 | if (mddev_busy(fd2devnm(mdfd))) { |
0431869c N |
1512 | pr_err("%s already active, cannot restart it!\n", |
1513 | mddev); | |
1514 | for (tmpdev = devlist ; | |
1515 | tmpdev && tmpdev->used != 1; | |
1516 | tmpdev = tmpdev->next) | |
1517 | ; | |
1518 | if (tmpdev && auto_assem) | |
1519 | pr_err("%s needed for %s...\n", | |
1520 | mddev, tmpdev->devname); | |
1521 | close(mdfd); | |
1522 | mdfd = -3; | |
1523 | st->ss->free_super(st); | |
1524 | if (auto_assem) | |
1525 | goto try_again; | |
57908e9e | 1526 | goto out; |
0431869c N |
1527 | } |
1528 | /* just incase it was started but has no content */ | |
1529 | ioctl(mdfd, STOP_ARRAY, NULL); | |
8a46fe84 NB |
1530 | } |
1531 | ||
9008ed1c N |
1532 | if (content != &info) { |
1533 | /* This is a member of a container. Try starting the array. */ | |
588bebfc | 1534 | int err; |
11b6d91d | 1535 | err = assemble_container_content(st, mdfd, content, c, |
9ca39acb | 1536 | chosen_name, NULL); |
588bebfc | 1537 | close(mdfd); |
588bebfc | 1538 | return err; |
9008ed1c | 1539 | } |
32141c17 | 1540 | |
811e6cbe | 1541 | /* Ok, no bad inconsistancy, we can try updating etc */ |
503975b9 N |
1542 | devices = xcalloc(num_devs, sizeof(*devices)); |
1543 | devmap = xcalloc(num_devs, content->array.raid_disks); | |
df842e69 | 1544 | devcnt = load_devices(devices, devmap, ident, &st, devlist, |
2c355c22 N |
1545 | c, content, mdfd, mddev, |
1546 | &most_recent, &bestcnt, &best, inargv); | |
76781701 | 1547 | if (devcnt < 0) { |
57908e9e | 1548 | mdfd = -3; |
783a4a93 GJ |
1549 | /* |
1550 | * devices is already freed in load_devices, so set devices | |
1551 | * to NULL to avoid double free devices. | |
1552 | */ | |
1553 | devices = NULL; | |
57908e9e | 1554 | goto out; |
76781701 | 1555 | } |
4b1ac34b | 1556 | |
64c4757e | 1557 | if (devcnt == 0) { |
e7b84f9d | 1558 | pr_err("no devices found for %s\n", |
79f9f56d | 1559 | mddev); |
3d2b16e7 NB |
1560 | if (st) |
1561 | st->ss->free_super(st); | |
02e7c5b7 | 1562 | free(devmap); |
57908e9e | 1563 | goto out; |
64c4757e | 1564 | } |
4b1ac34b | 1565 | |
4977146a | 1566 | if (c->update && strcmp(c->update, "byteorder")==0) |
3d2b16e7 NB |
1567 | st->minor_version = 90; |
1568 | ||
a5d85af7 | 1569 | st->ss->getinfo_super(st, content, NULL); |
98dbd966 | 1570 | clean = content->array.state & 1; |
4b1ac34b | 1571 | |
64c4757e NB |
1572 | /* now we have some devices that might be suitable. |
1573 | * I wonder how many | |
1574 | */ | |
503975b9 | 1575 | avail = xcalloc(content->array.raid_disks, 1); |
64c4757e | 1576 | okcnt = 0; |
aacb2f81 | 1577 | replcnt = 0; |
52826846 | 1578 | sparecnt=0; |
69a48116 | 1579 | journalcnt=0; |
1ff98339 | 1580 | rebuilding_cnt=0; |
f21e18ca | 1581 | for (i=0; i< bestcnt; i++) { |
64c4757e | 1582 | int j = best[i]; |
ee04451c NB |
1583 | int event_margin = 1; /* always allow a difference of '1' |
1584 | * like the kernel does | |
1585 | */ | |
64c4757e | 1586 | if (j < 0) continue; |
d013a55e NB |
1587 | /* note: we ignore error flags in multipath arrays |
1588 | * as they don't make sense | |
1589 | */ | |
69a48116 SL |
1590 | if (content->array.level != LEVEL_MULTIPATH) { |
1591 | if (devices[j].i.disk.state & (1<<MD_DISK_JOURNAL)) { | |
051f3265 | 1592 | if (content->journal_device_required) |
69a48116 SL |
1593 | journalcnt++; |
1594 | else /* unexpected journal, mark as faulty */ | |
1595 | devices[j].i.disk.state |= (1<<MD_DISK_FAULTY); | |
1596 | } else if (!(devices[j].i.disk.state & (1<<MD_DISK_ACTIVE))) { | |
213ee40b | 1597 | if (!(devices[j].i.disk.state |
ed7fc6b4 AC |
1598 | & (1<<MD_DISK_FAULTY))) { |
1599 | devices[j].uptodate = 1; | |
aa88f531 | 1600 | sparecnt++; |
ed7fc6b4 | 1601 | } |
d013a55e | 1602 | continue; |
aa88f531 | 1603 | } |
69a48116 | 1604 | } |
de5a472e | 1605 | /* If this device thinks that 'most_recent' has failed, then |
02e7c5b7 N |
1606 | * we must reject this device. |
1607 | */ | |
eb20ecf1 | 1608 | if (j != most_recent && !c->force && |
02e7c5b7 N |
1609 | content->array.raid_disks > 0 && |
1610 | devices[most_recent].i.disk.raid_disk >= 0 && | |
1611 | devmap[j * content->array.raid_disks + devices[most_recent].i.disk.raid_disk] == 0) { | |
4977146a | 1612 | if (c->verbose > -1) |
e7b84f9d | 1613 | pr_err("ignoring %s as it reports %s as failed\n", |
79f9f56d | 1614 | devices[j].devname, devices[most_recent].devname); |
02e7c5b7 N |
1615 | best[i] = -1; |
1616 | continue; | |
1617 | } | |
f80057ae N |
1618 | /* Require event counter to be same as, or just less than, |
1619 | * most recent. If it is bigger, it must be a stray spare and | |
1620 | * should be ignored. | |
1621 | */ | |
213ee40b | 1622 | if (devices[j].i.events+event_margin >= |
f80057ae N |
1623 | devices[most_recent].i.events && |
1624 | devices[j].i.events <= | |
1625 | devices[most_recent].i.events | |
1626 | ) { | |
64c4757e | 1627 | devices[j].uptodate = 1; |
69a48116 | 1628 | if (devices[j].i.disk.state & (1<<MD_DISK_JOURNAL)) |
dbfbca43 | 1629 | journal_clean = 1; |
aacb2f81 | 1630 | if (i < content->array.raid_disks * 2) { |
dcc4210f DW |
1631 | if (devices[j].i.recovery_start == MaxSector || |
1632 | (content->reshape_active && | |
d316dba7 | 1633 | i >= content->array.raid_disks - content->delta_disks)) { |
aacb2f81 N |
1634 | if (!avail[i/2]) { |
1635 | okcnt++; | |
1636 | avail[i/2]=1; | |
1637 | } else | |
1638 | replcnt++; | |
1ff98339 N |
1639 | } else |
1640 | rebuilding_cnt++; | |
69a48116 | 1641 | } else if (devices[j].i.disk.raid_disk != MD_DISK_ROLE_JOURNAL) |
52826846 | 1642 | sparecnt++; |
64c4757e NB |
1643 | } |
1644 | } | |
02e7c5b7 | 1645 | free(devmap); |
8cde842b N |
1646 | if (c->force) { |
1647 | int force_ok = force_array(content, devices, best, bestcnt, | |
1648 | avail, most_recent, st, c); | |
1649 | okcnt += force_ok; | |
1650 | if (force_ok) | |
1651 | was_forced = 1; | |
1652 | } | |
52826846 NB |
1653 | /* Now we want to look at the superblock which the kernel will base things on |
1654 | * and compare the devices that we think are working with the devices that the | |
1655 | * superblock thinks are working. | |
1656 | * If there are differences and --force is given, then update this chosen | |
1657 | * superblock. | |
1658 | */ | |
cd29a5c8 | 1659 | chosen_drive = -1; |
3da92f27 | 1660 | st->ss->free_super(st); |
f81a2b56 | 1661 | for (i=0; chosen_drive < 0 && i<bestcnt; i+=2) { |
52826846 NB |
1662 | int j = best[i]; |
1663 | int fd; | |
4b1ac34b | 1664 | |
52826846 NB |
1665 | if (j<0) |
1666 | continue; | |
1667 | if (!devices[j].uptodate) | |
1668 | continue; | |
a28232b8 N |
1669 | if (devices[j].i.events < devices[most_recent].i.events) |
1670 | continue; | |
52826846 | 1671 | chosen_drive = j; |
0431869c N |
1672 | if ((fd=dev_open(devices[j].devname, |
1673 | devices[j].included ? O_RDONLY | |
1674 | : (O_RDONLY|O_EXCL)))< 0) { | |
e7b84f9d | 1675 | pr_err("Cannot open %s: %s\n", |
79f9f56d | 1676 | devices[j].devname, strerror(errno)); |
57908e9e | 1677 | goto out; |
52826846 | 1678 | } |
3da92f27 | 1679 | if (st->ss->load_super(st,fd, NULL)) { |
52826846 | 1680 | close(fd); |
e7b84f9d | 1681 | pr_err("RAID superblock has disappeared from %s\n", |
79f9f56d | 1682 | devices[j].devname); |
57908e9e | 1683 | goto out; |
52826846 NB |
1684 | } |
1685 | close(fd); | |
1686 | } | |
3da92f27 | 1687 | if (st->sb == NULL) { |
e7b84f9d | 1688 | pr_err("No suitable drives found for %s\n", mddev); |
57908e9e | 1689 | goto out; |
4b1ac34b | 1690 | } |
a5d85af7 | 1691 | st->ss->getinfo_super(st, content, NULL); |
dae13137 JS |
1692 | if (sysfs_init(content, mdfd, NULL)) { |
1693 | pr_err("Unable to initialize sysfs\n"); | |
57908e9e | 1694 | goto out; |
dae13137 | 1695 | } |
32141c17 | 1696 | |
dbfbca43 SL |
1697 | /* after reload context, store journal_clean in context */ |
1698 | content->journal_clean = journal_clean; | |
56eedc1a | 1699 | for (i=0; i<bestcnt; i++) { |
52826846 | 1700 | int j = best[i]; |
98c6faba | 1701 | unsigned int desired_state; |
11a3e71d | 1702 | |
31b6f0cd AR |
1703 | if (j < 0) |
1704 | continue; | |
69a48116 SL |
1705 | if (devices[j].i.disk.raid_disk == MD_DISK_ROLE_JOURNAL) |
1706 | desired_state = (1<<MD_DISK_JOURNAL); | |
1707 | else if (i >= content->array.raid_disks * 2) | |
11a3e71d | 1708 | desired_state = 0; |
aacb2f81 N |
1709 | else if (i & 1) |
1710 | desired_state = (1<<MD_DISK_ACTIVE) | (1<<MD_DISK_REPLACEMENT); | |
1711 | else | |
1712 | desired_state = (1<<MD_DISK_ACTIVE) | (1<<MD_DISK_SYNC); | |
11a3e71d | 1713 | |
0833f9c3 GK |
1714 | desired_state |= devices[j].i.disk.state & ((1<<MD_DISK_FAILFAST) | |
1715 | (1<<MD_DISK_WRITEMOSTLY)); | |
1716 | ||
52826846 NB |
1717 | if (!devices[j].uptodate) |
1718 | continue; | |
4b1ac34b | 1719 | |
213ee40b | 1720 | devices[j].i.disk.state = desired_state; |
4e9a6ff7 N |
1721 | if (!(devices[j].i.array.state & 1)) |
1722 | clean = 0; | |
213ee40b NB |
1723 | |
1724 | if (st->ss->update_super(st, &devices[j].i, "assemble", NULL, | |
4977146a N |
1725 | c->verbose, 0, NULL)) { |
1726 | if (c->force) { | |
1727 | if (c->verbose >= 0) | |
e7b84f9d | 1728 | pr_err("clearing FAULTY flag for device %d in %s for %s\n", |
79f9f56d | 1729 | j, mddev, devices[j].devname); |
4b1ac34b | 1730 | change = 1; |
52826846 | 1731 | } else { |
4977146a | 1732 | if (c->verbose >= -1) |
e7b84f9d | 1733 | pr_err("device %d in %s has wrong state in superblock, but %s seems ok\n", |
79f9f56d | 1734 | i, mddev, devices[j].devname); |
52826846 NB |
1735 | } |
1736 | } | |
4b1ac34b | 1737 | #if 0 |
213ee40b | 1738 | if (!(super.disks[i].i.disk.state & (1 << MD_DISK_FAULTY))) { |
e7b84f9d | 1739 | pr_err("devices %d of %s is not marked FAULTY in superblock, but cannot be found\n", |
79f9f56d | 1740 | i, mddev); |
52826846 | 1741 | } |
4b1ac34b | 1742 | #endif |
52826846 | 1743 | } |
4977146a | 1744 | if (c->force && !clean && |
98dbd966 DW |
1745 | !enough(content->array.level, content->array.raid_disks, |
1746 | content->array.layout, clean, | |
de5a472e | 1747 | avail)) { |
98dbd966 | 1748 | change += st->ss->update_super(st, content, "force-array", |
6f4dbdc4 | 1749 | devices[chosen_drive].devname, c->verbose, |
67a8c82d | 1750 | 0, NULL); |
8cde842b | 1751 | was_forced = 1; |
583315d9 | 1752 | clean = 1; |
d013a55e | 1753 | } |
52826846 | 1754 | |
4b1ac34b | 1755 | if (change) { |
52826846 | 1756 | int fd; |
0431869c N |
1757 | fd = dev_open(devices[chosen_drive].devname, |
1758 | devices[chosen_drive].included ? | |
1759 | O_RDWR : (O_RDWR|O_EXCL)); | |
52826846 | 1760 | if (fd < 0) { |
e7b84f9d | 1761 | pr_err("Could not open %s for write - cannot Assemble array.\n", |
79f9f56d | 1762 | devices[chosen_drive].devname); |
57908e9e | 1763 | goto out; |
52826846 | 1764 | } |
3da92f27 | 1765 | if (st->ss->store_super(st, fd)) { |
52826846 | 1766 | close(fd); |
e7b84f9d | 1767 | pr_err("Could not re-write superblock on %s\n", |
79f9f56d | 1768 | devices[chosen_drive].devname); |
57908e9e | 1769 | goto out; |
52826846 | 1770 | } |
4977146a | 1771 | if (c->verbose >= 0) |
e7b84f9d | 1772 | pr_err("Marking array %s as 'clean'\n", |
79f9f56d | 1773 | mddev); |
52826846 | 1774 | close(fd); |
52826846 NB |
1775 | } |
1776 | ||
353632d9 NB |
1777 | /* If we are in the middle of a reshape we may need to restore saved data |
1778 | * that was moved aside due to the reshape overwriting live data | |
1779 | * The code of doing this lives in Grow.c | |
1780 | */ | |
5e88ab2e N |
1781 | if (content->reshape_active && |
1782 | !(content->reshape_active & RESHAPE_NO_BACKUP)) { | |
353632d9 | 1783 | int err = 0; |
503975b9 | 1784 | int *fdlist = xmalloc(sizeof(int)* bestcnt); |
4977146a | 1785 | if (c->verbose > 0) |
c34fef77 | 1786 | pr_err("%s has an active reshape - checking if critical section needs to be restored\n", |
79f9f56d | 1787 | chosen_name); |
54ded86f N |
1788 | if (!c->backup_file) |
1789 | c->backup_file = locate_backup(content->sys_name); | |
a7dec3fd | 1790 | enable_fds(bestcnt/2); |
5e9fd96f N |
1791 | for (i = 0; i < bestcnt/2; i++) { |
1792 | int j = best[i*2]; | |
353632d9 | 1793 | if (j >= 0) { |
0431869c N |
1794 | fdlist[i] = dev_open(devices[j].devname, |
1795 | devices[j].included | |
1796 | ? O_RDWR : (O_RDWR|O_EXCL)); | |
353632d9 | 1797 | if (fdlist[i] < 0) { |
e7b84f9d | 1798 | pr_err("Could not open %s for write - cannot Assemble array.\n", |
79f9f56d | 1799 | devices[j].devname); |
353632d9 NB |
1800 | err = 1; |
1801 | break; | |
1802 | } | |
1803 | } else | |
1804 | fdlist[i] = -1; | |
1805 | } | |
87f26d14 | 1806 | if (!err) { |
ba53ea59 AK |
1807 | if (st->ss->external && st->ss->recover_backup) |
1808 | err = st->ss->recover_backup(st, content); | |
1809 | else | |
5e9fd96f | 1810 | err = Grow_restart(st, content, fdlist, bestcnt/2, |
4977146a N |
1811 | c->backup_file, c->verbose > 0); |
1812 | if (err && c->invalid_backup) { | |
1813 | if (c->verbose > 0) | |
7a862a02 | 1814 | pr_err("continuing without restoring backup\n"); |
87f26d14 N |
1815 | err = 0; |
1816 | } | |
1817 | } | |
353632d9 NB |
1818 | while (i>0) { |
1819 | i--; | |
1820 | if (fdlist[i]>=0) close(fdlist[i]); | |
1821 | } | |
56dcaa6b | 1822 | free(fdlist); |
353632d9 | 1823 | if (err) { |
e7b84f9d | 1824 | pr_err("Failed to restore critical section for reshape, sorry.\n"); |
4977146a | 1825 | if (c->backup_file == NULL) |
e7b84f9d | 1826 | cont_err("Possibly you needed to specify the --backup-file\n"); |
57908e9e | 1827 | goto out; |
353632d9 NB |
1828 | } |
1829 | } | |
aa88f531 | 1830 | |
64c4757e | 1831 | /* Almost ready to actually *do* something */ |
6f4dbdc4 N |
1832 | /* First, fill in the map, so that udev can find our name |
1833 | * as soon as we become active. | |
1834 | */ | |
afa368f4 N |
1835 | if (c->update && strcmp(c->update, "metadata")==0) { |
1836 | content->array.major_version = 1; | |
1837 | content->array.minor_version = 0; | |
1838 | strcpy(content->text_version, "1.0"); | |
1839 | } | |
1840 | ||
4dd2df09 | 1841 | map_update(&map, fd2devnm(mdfd), content->text_version, |
6f4dbdc4 N |
1842 | content->uuid, chosen_name); |
1843 | ||
1844 | rv = start_array(mdfd, mddev, content, | |
1845 | st, ident, best, bestcnt, | |
1846 | chosen_drive, devices, okcnt, sparecnt, | |
69a48116 | 1847 | rebuilding_cnt, journalcnt, |
6f4dbdc4 | 1848 | c, |
71417de6 N |
1849 | clean, avail, start_partial_ok, |
1850 | pre_exist != NULL, | |
051f3265 | 1851 | was_forced); |
6f4dbdc4 N |
1852 | if (rv == 1 && !pre_exist) |
1853 | ioctl(mdfd, STOP_ARRAY, NULL); | |
6f4dbdc4 | 1854 | free(devices); |
57908e9e | 1855 | out: |
898bd1ec | 1856 | map_unlock(&map); |
6f4dbdc4 N |
1857 | if (rv == 0) { |
1858 | wait_for(chosen_name, mdfd); | |
b20c8a50 | 1859 | close(mdfd); |
6f4dbdc4 N |
1860 | if (auto_assem) { |
1861 | int usecs = 1; | |
1862 | /* There is a nasty race with 'mdadm --monitor'. | |
1863 | * If it opens this device before we close it, | |
1864 | * it gets an incomplete open on which IO | |
1865 | * doesn't work and the capacity is | |
1866 | * wrong. | |
1867 | * If we reopen (to check for layered devices) | |
1868 | * before --monitor closes, we loose. | |
1869 | * | |
1870 | * So: wait upto 1 second for there to be | |
1871 | * a non-zero capacity. | |
1872 | */ | |
1873 | while (usecs < 1000) { | |
1874 | mdfd = open(mddev, O_RDONLY); | |
1875 | if (mdfd >= 0) { | |
1876 | unsigned long long size; | |
1877 | if (get_dev_size(mdfd, NULL, &size) && | |
1878 | size > 0) | |
1879 | break; | |
1880 | close(mdfd); | |
589395d6 | 1881 | } |
6f4dbdc4 N |
1882 | usleep(usecs); |
1883 | usecs <<= 1; | |
dab6685f | 1884 | } |
aa88f531 | 1885 | } |
57908e9e | 1886 | } else if (mdfd >= 0) |
b20c8a50 N |
1887 | close(mdfd); |
1888 | ||
6f4dbdc4 | 1889 | /* '2' means 'OK, but not started yet' */ |
57908e9e GJ |
1890 | if (rv == -1) { |
1891 | free(devices); | |
1892 | return 1; | |
1893 | } | |
6f4dbdc4 | 1894 | return rv == 2 ? 0 : rv; |
64c4757e | 1895 | } |
6234c63c | 1896 | |
6234c63c | 1897 | int assemble_container_content(struct supertype *st, int mdfd, |
11b6d91d | 1898 | struct mdinfo *content, struct context *c, |
9ca39acb | 1899 | char *chosen_name, int *result) |
6234c63c | 1900 | { |
02b70e83 | 1901 | struct mdinfo *dev, *sra, *dev2; |
6234c63c | 1902 | int working = 0, preexist = 0; |
882029c8 | 1903 | int expansion = 0; |
7af03341 | 1904 | int old_raid_disks; |
4aecb54a | 1905 | int start_reshape; |
0a8e239c | 1906 | char *avail; |
c1736844 | 1907 | int err; |
6234c63c | 1908 | |
dae13137 JS |
1909 | if (sysfs_init(content, mdfd, NULL)) { |
1910 | pr_err("Unable to initialize sysfs\n"); | |
1911 | return 1; | |
1912 | } | |
6234c63c | 1913 | |
02b70e83 | 1914 | sra = sysfs_read(mdfd, NULL, GET_VERSION|GET_DEVS); |
0ea8f5b1 N |
1915 | if (sra == NULL || strcmp(sra->text_version, content->text_version) != 0) { |
1916 | if (content->array.major_version == -1 && | |
1917 | content->array.minor_version == -2 && | |
11b6d91d | 1918 | c->readonly && |
0ea8f5b1 N |
1919 | content->text_version[0] == '/') |
1920 | content->text_version[0] = '-'; | |
b6e60be6 | 1921 | if (sysfs_set_array(content, 9003) != 0) { |
30e19bf8 | 1922 | sysfs_free(sra); |
6234c63c | 1923 | return 1; |
22472ee1 | 1924 | } |
0ea8f5b1 | 1925 | } |
588bebfc | 1926 | |
4aecb54a AK |
1927 | /* There are two types of reshape: container wide or sub-array specific |
1928 | * Check if metadata requests blocking container wide reshapes | |
1929 | */ | |
1930 | start_reshape = (content->reshape_active && | |
79f9f56d N |
1931 | !((content->reshape_active == CONTAINER_RESHAPE) && |
1932 | (content->array.state & (1<<MD_SB_BLOCK_CONTAINER_RESHAPE)))); | |
4aecb54a AK |
1933 | |
1934 | /* Block subarray here if it is under reshape now | |
1935 | * Do not allow for any changes in this array | |
1936 | */ | |
1937 | if (st->ss->external && content->recovery_blocked && start_reshape) | |
b8063f07 AK |
1938 | block_subarray(content); |
1939 | ||
02b70e83 N |
1940 | for (dev2 = sra->devs; dev2; dev2 = dev2->next) { |
1941 | for (dev = content->devs; dev; dev = dev->next) | |
1942 | if (dev2->disk.major == dev->disk.major && | |
1943 | dev2->disk.minor == dev->disk.minor) | |
1944 | break; | |
1945 | if (dev) | |
1946 | continue; | |
1947 | /* Don't want this one any more */ | |
1948 | if (sysfs_set_str(sra, dev2, "slot", "none") < 0 && | |
1949 | errno == EBUSY) { | |
1950 | pr_err("Cannot remove old device %s: not updating %s\n", dev2->sys_name, sra->sys_name); | |
1951 | sysfs_free(sra); | |
1952 | return 1; | |
1953 | } | |
1954 | sysfs_set_str(sra, dev2, "state", "remove"); | |
1955 | } | |
7af03341 | 1956 | old_raid_disks = content->array.raid_disks - content->delta_disks; |
c1736844 N |
1957 | avail = xcalloc(content->array.raid_disks, 1); |
1958 | for (dev = content->devs; dev; dev = dev->next) { | |
1959 | if (dev->disk.raid_disk >= 0) | |
1960 | avail[dev->disk.raid_disk] = 1; | |
14032016 | 1961 | if (sysfs_add_disk(content, dev, 1) == 0) { |
7af03341 | 1962 | if (dev->disk.raid_disk >= old_raid_disks && |
14032016 AK |
1963 | content->reshape_active) |
1964 | expansion++; | |
1965 | else | |
1966 | working++; | |
1967 | } else if (errno == EEXIST) | |
6234c63c | 1968 | preexist++; |
c1736844 | 1969 | } |
02b70e83 | 1970 | sysfs_free(sra); |
c1736844 N |
1971 | if (working + expansion == 0 && c->runstop <= 0) { |
1972 | free(avail); | |
7cb2aa33 | 1973 | return 1;/* Nothing new, don't try to start */ |
c1736844 | 1974 | } |
0c79d8ca | 1975 | map_update(NULL, fd2devnm(mdfd), content->text_version, |
8b4e5ea9 N |
1976 | content->uuid, chosen_name); |
1977 | ||
2432ce9b AP |
1978 | if (content->consistency_policy == CONSISTENCY_POLICY_PPL && |
1979 | st->ss->validate_ppl) { | |
1980 | content->array.state |= 1; | |
1981 | err = 0; | |
1982 | ||
1983 | for (dev = content->devs; dev; dev = dev->next) { | |
1984 | int dfd; | |
1985 | char *devpath; | |
1986 | int ret; | |
1987 | ||
1988 | ret = st->ss->validate_ppl(st, content, dev); | |
1989 | if (ret == 0) | |
1990 | continue; | |
1991 | ||
1992 | if (ret < 0) { | |
1993 | err = 1; | |
1994 | break; | |
1995 | } | |
1996 | ||
1997 | if (!c->force) { | |
1998 | pr_err("%s contains invalid PPL - consider --force or --update-subarray with --update=no-ppl\n", | |
1999 | chosen_name); | |
2000 | content->array.state &= ~1; | |
2001 | avail[dev->disk.raid_disk] = 0; | |
2002 | break; | |
2003 | } | |
2004 | ||
2005 | /* have --force - overwrite the invalid ppl */ | |
2006 | devpath = map_dev(dev->disk.major, dev->disk.minor, 0); | |
2007 | dfd = dev_open(devpath, O_RDWR); | |
2008 | if (dfd < 0) { | |
2009 | pr_err("Failed to open %s\n", devpath); | |
2010 | err = 1; | |
2011 | break; | |
2012 | } | |
2013 | ||
2014 | err = st->ss->write_init_ppl(st, content, dfd); | |
2015 | close(dfd); | |
2016 | ||
2017 | if (err) | |
2018 | break; | |
2019 | } | |
2020 | ||
2021 | if (err) { | |
2022 | free(avail); | |
2023 | return err; | |
2024 | } | |
2025 | } | |
2026 | ||
c1736844 N |
2027 | if (enough(content->array.level, content->array.raid_disks, |
2028 | content->array.layout, content->array.state & 1, avail) == 0) { | |
9ca39acb N |
2029 | if (c->export && result) |
2030 | *result |= INCR_NO; | |
2031 | else if (c->verbose >= 0) { | |
c1736844 N |
2032 | pr_err("%s assembled with %d device%s", |
2033 | chosen_name, preexist + working, | |
2034 | preexist + working == 1 ? "":"s"); | |
6234c63c DW |
2035 | if (preexist) |
2036 | fprintf(stderr, " (%d new)", working); | |
c1736844 | 2037 | fprintf(stderr, " but not started\n"); |
6234c63c | 2038 | } |
c1736844 N |
2039 | free(avail); |
2040 | return 1; | |
2041 | } | |
2042 | free(avail); | |
2043 | ||
2044 | if (c->runstop <= 0 && | |
2045 | (working + preexist + expansion) < | |
2046 | content->array.working_disks) { | |
9ca39acb N |
2047 | if (c->export && result) |
2048 | *result |= INCR_UNSAFE; | |
2049 | else if (c->verbose >= 0) { | |
e7b84f9d N |
2050 | pr_err("%s assembled with %d device%s", |
2051 | chosen_name, preexist + working, | |
2052 | preexist + working == 1 ? "":"s"); | |
88716263 N |
2053 | if (preexist) |
2054 | fprintf(stderr, " (%d new)", working); | |
c1736844 | 2055 | fprintf(stderr, " but not safe to start\n"); |
88716263 | 2056 | } |
7cb2aa33 N |
2057 | return 1; |
2058 | } | |
c1736844 N |
2059 | |
2060 | ||
2061 | if (start_reshape) { | |
2062 | int spare = content->array.raid_disks + expansion; | |
2063 | if (restore_backup(st, content, | |
2064 | working, | |
54ded86f | 2065 | spare, &c->backup_file, c->verbose) == 1) |
c1736844 N |
2066 | return 1; |
2067 | ||
2b57e4fe PB |
2068 | if (content->reshape_progress == 0) { |
2069 | /* If reshape progress is 0 - we are assembling the | |
2070 | * array that was stopped, before reshape has started. | |
2071 | * Array needs to be started as active, Grow_continue() | |
2072 | * will start the reshape. | |
2073 | */ | |
2074 | sysfs_set_num(content, NULL, "reshape_position", | |
2075 | MaxSector); | |
2076 | err = sysfs_set_str(content, NULL, | |
2077 | "array_state", "active"); | |
2078 | sysfs_set_num(content, NULL, "reshape_position", 0); | |
2079 | } else { | |
2080 | err = sysfs_set_str(content, NULL, | |
2081 | "array_state", "readonly"); | |
2082 | } | |
2083 | ||
c1736844 N |
2084 | if (err) |
2085 | return 1; | |
2086 | ||
2087 | if (st->ss->external) { | |
2088 | if (!mdmon_running(st->container_devnm)) | |
2089 | start_mdmon(st->container_devnm); | |
2090 | ping_monitor(st->container_devnm); | |
2091 | if (mdmon_running(st->container_devnm) && | |
2092 | st->update_tail == NULL) | |
2093 | st->update_tail = &st->updates; | |
2094 | } | |
2095 | ||
2096 | err = Grow_continue(mdfd, st, content, c->backup_file, | |
06e293d0 | 2097 | 0, c->freeze_reshape); |
c1736844 N |
2098 | } else switch(content->array.level) { |
2099 | case LEVEL_LINEAR: | |
2100 | case LEVEL_MULTIPATH: | |
2101 | case 0: | |
2102 | err = sysfs_set_str(content, NULL, "array_state", | |
2103 | c->readonly ? "readonly" : "active"); | |
2104 | break; | |
2105 | default: | |
2106 | err = sysfs_set_str(content, NULL, "array_state", | |
2107 | "readonly"); | |
2108 | /* start mdmon if needed. */ | |
2109 | if (!err) { | |
2110 | if (!mdmon_running(st->container_devnm)) | |
2111 | start_mdmon(st->container_devnm); | |
2112 | ping_monitor(st->container_devnm); | |
2113 | } | |
2114 | break; | |
2115 | } | |
2116 | if (!err) | |
2117 | sysfs_set_safemode(content, content->safe_mode_delay); | |
2118 | ||
2119 | /* Block subarray here if it is not reshaped now | |
2120 | * It has be blocked a little later to allow mdmon to switch in | |
2121 | * in to R/W state | |
2122 | */ | |
2123 | if (st->ss->external && content->recovery_blocked && | |
2124 | !start_reshape) | |
2125 | block_subarray(content); | |
2126 | ||
9ca39acb N |
2127 | if (c->export && result) { |
2128 | if (err) | |
2129 | *result |= INCR_NO; | |
2130 | else | |
2131 | *result |= INCR_YES; | |
2132 | } else if (c->verbose >= 0) { | |
c1736844 N |
2133 | if (err) |
2134 | pr_err("array %s now has %d device%s", | |
2135 | chosen_name, working + preexist, | |
2136 | working + preexist == 1 ? "":"s"); | |
b0681598 MD |
2137 | else { |
2138 | sysfs_rules_apply(chosen_name, content); | |
c1736844 N |
2139 | pr_err("Started %s with %d device%s", |
2140 | chosen_name, working + preexist, | |
2141 | working + preexist == 1 ? "":"s"); | |
b0681598 | 2142 | } |
c1736844 N |
2143 | if (preexist) |
2144 | fprintf(stderr, " (%d new)", working); | |
2145 | if (expansion) | |
2146 | fprintf(stderr, " ( + %d for expansion)", | |
2147 | expansion); | |
2148 | fprintf(stderr, "\n"); | |
2149 | } | |
2150 | if (!err) | |
2151 | wait_for(chosen_name, mdfd); | |
2152 | return err; | |
2153 | /* FIXME should have an O_EXCL and wait for read-auto */ | |
6234c63c | 2154 | } |