]>
Commit | Line | Data |
---|---|---|
52826846 | 1 | /* |
9a9dab36 | 2 | * mdadm - manage Linux "md" devices aka RAID arrays. |
52826846 | 3 | * |
e736b623 | 4 | * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de> |
52826846 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> |
52826846 NB |
23 | */ |
24 | ||
9a9dab36 | 25 | #include "mdadm.h" |
52826846 NB |
26 | #include "md_p.h" |
27 | #include "md_u.h" | |
e0d19036 | 28 | #include <sys/wait.h> |
98127a6d | 29 | #include <signal.h> |
4450e59f | 30 | #include <limits.h> |
773135f5 | 31 | #include <syslog.h> |
52826846 | 32 | |
2e0172b1 N |
33 | struct state { |
34 | char *devname; | |
4dd2df09 | 35 | char devnm[32]; /* to sync with mdstat info */ |
1d13b599 | 36 | unsigned int utime; |
2e0172b1 N |
37 | int err; |
38 | char *spare_group; | |
39 | int active, working, failed, spare, raid; | |
f1661bd7 | 40 | int from_config; |
73ff0732 | 41 | int from_auto; |
2e0172b1 | 42 | int expected_spares; |
b0599bda N |
43 | int devstate[MAX_DISKS]; |
44 | dev_t devid[MAX_DISKS]; | |
2e0172b1 | 45 | int percent; |
4dd2df09 N |
46 | char parent_devnm[32]; /* For subarray, devnm of parent. |
47 | * For others, "" | |
48 | */ | |
2e0172b1 | 49 | struct supertype *metadata; |
c3621c0a ML |
50 | struct state *subarray;/* for a container it is a link to first subarray |
51 | * for a subarray it is a link to next subarray | |
52 | * in the same container */ | |
53 | struct state *parent; /* for a subarray it is a link to its container | |
54 | */ | |
2e0172b1 N |
55 | struct state *next; |
56 | }; | |
57 | ||
e0bd6a96 N |
58 | struct alert_info { |
59 | char *mailaddr; | |
60 | char *mailfrom; | |
61 | char *alert_cmd; | |
62 | int dosyslog; | |
63 | }; | |
2e0172b1 N |
64 | static int make_daemon(char *pidfile); |
65 | static int check_one_sharer(int scan); | |
7f3b2d1d | 66 | static void write_autorebuild_pid(void); |
e0bd6a96 | 67 | static void alert(char *event, char *dev, char *disc, struct alert_info *info); |
a90e1050 | 68 | static int check_array(struct state *st, struct mdstat_ent *mdstat, |
e0bd6a96 | 69 | int test, struct alert_info *info, |
c2ecf5f6 | 70 | int increments, char *prefer); |
83f3bc5f | 71 | static int add_new_arrays(struct mdstat_ent *mdstat, struct state **statelist, |
e0bd6a96 N |
72 | int test, struct alert_info *info); |
73 | static void try_spare_migration(struct state *statelist, struct alert_info *info); | |
c3621c0a | 74 | static void link_containers_with_subarrays(struct state *list); |
2e0172b1 | 75 | |
a655e550 | 76 | int Monitor(struct mddev_dev *devlist, |
52826846 | 77 | char *mailaddr, char *alert_cmd, |
95c50205 N |
78 | struct context *c, |
79 | int daemonise, int oneshot, | |
80 | int dosyslog, char *pidfile, int increments, | |
81 | int share) | |
52826846 NB |
82 | { |
83 | /* | |
84 | * Every few seconds, scan every md device looking for changes | |
85 | * When a change is found, log it, possibly run the alert command, | |
86 | * and possibly send Email | |
87 | * | |
88 | * For each array, we record: | |
89 | * Update time | |
90 | * active/working/failed/spare drives | |
91 | * State of each device. | |
e0d19036 | 92 | * %rebuilt if rebuilding |
52826846 NB |
93 | * |
94 | * If the update time changes, check out all the data again | |
95 | * It is possible that we cannot get the state of each device | |
96 | * due to bugs in the md kernel module. | |
e0d19036 NB |
97 | * We also read /proc/mdstat to get rebuild percent, |
98 | * and to get state on all active devices incase of kernel bug. | |
52826846 | 99 | * |
e0d19036 NB |
100 | * Events are: |
101 | * Fail | |
102 | * An active device had Faulty set or Active/Sync removed | |
103 | * FailSpare | |
104 | * A spare device had Faulty set | |
105 | * SpareActive | |
106 | * An active device had a reverse transition | |
107 | * RebuildStarted | |
108 | * percent went from -1 to +ve | |
9a36a9b7 ZB |
109 | * RebuildNN |
110 | * percent went from below to not-below NN% | |
e0d19036 NB |
111 | * DeviceDisappeared |
112 | * Couldn't access a device which was previously visible | |
52826846 NB |
113 | * |
114 | * if we detect an array with active<raid and spare==0 | |
115 | * we look at other arrays that have same spare-group | |
116 | * If we find one with active==raid and spare>0, | |
117 | * and if we can get_disk_info and find a name | |
118 | * Then we hot-remove and hot-add to the other array | |
119 | * | |
e0d19036 NB |
120 | * If devlist is NULL, then we can monitor everything because --scan |
121 | * was given. We get an initial list from config file and add anything | |
122 | * that appears in /proc/mdstat | |
52826846 NB |
123 | */ |
124 | ||
2e0172b1 | 125 | struct state *statelist = NULL; |
b657208c | 126 | struct state *st2; |
52826846 | 127 | int finished = 0; |
e0d19036 | 128 | struct mdstat_ent *mdstat = NULL; |
72362f18 | 129 | char *mailfrom; |
e0bd6a96 | 130 | struct alert_info info; |
9f3dd454 | 131 | struct mddev_ident *mdlist; |
cab9c67d | 132 | int delay_for_event = c->delay; |
e0d19036 | 133 | |
d013a55e | 134 | if (!mailaddr) { |
8aec876d | 135 | mailaddr = conf_get_mailaddr(); |
95c50205 | 136 | if (mailaddr && ! c->scan) |
e7b84f9d | 137 | pr_err("Monitor using email address \"%s\" from config file\n", |
d013a55e NB |
138 | mailaddr); |
139 | } | |
8aec876d | 140 | mailfrom = conf_get_mailfrom(); |
4948b8f7 | 141 | |
d013a55e | 142 | if (!alert_cmd) { |
8aec876d | 143 | alert_cmd = conf_get_program(); |
f566ef45 | 144 | if (alert_cmd && !c->scan) |
e7b84f9d | 145 | pr_err("Monitor using program \"%s\" from config file\n", |
d013a55e NB |
146 | alert_cmd); |
147 | } | |
95c50205 | 148 | if (c->scan && !mailaddr && !alert_cmd && !dosyslog) { |
e7b84f9d | 149 | pr_err("No mail address or alert command - not monitoring.\n"); |
d013a55e | 150 | return 1; |
56eedc1a | 151 | } |
e0bd6a96 N |
152 | info.alert_cmd = alert_cmd; |
153 | info.mailaddr = mailaddr; | |
154 | info.mailfrom = mailfrom; | |
155 | info.dosyslog = dosyslog; | |
d013a55e | 156 | |
7f3b2d1d BK |
157 | if (share){ |
158 | if (check_one_sharer(c->scan)) | |
159 | return 1; | |
160 | } | |
161 | ||
44d337f0 AC |
162 | if (daemonise) { |
163 | int rv = make_daemon(pidfile); | |
164 | if (rv >= 0) | |
165 | return rv; | |
166 | } | |
e0d19036 | 167 | |
1011e834 | 168 | if (share) |
7f3b2d1d | 169 | write_autorebuild_pid(); |
edde9560 | 170 | |
e0d19036 | 171 | if (devlist == NULL) { |
9f3dd454 | 172 | mdlist = conf_get_ident(NULL); |
f566ef45 | 173 | for (; mdlist; mdlist = mdlist->next) { |
fe056d1f | 174 | struct state *st; |
f566ef45 | 175 | |
fe056d1f N |
176 | if (mdlist->devname == NULL) |
177 | continue; | |
112cace6 N |
178 | if (strcasecmp(mdlist->devname, "<ignore>") == 0) |
179 | continue; | |
503975b9 | 180 | st = xcalloc(1, sizeof *st); |
db2d001c | 181 | if (mdlist->devname[0] == '/') |
503975b9 | 182 | st->devname = xstrdup(mdlist->devname); |
db2d001c | 183 | else { |
503975b9 | 184 | st->devname = xmalloc(8+strlen(mdlist->devname)+1); |
db2d001c N |
185 | strcpy(strcpy(st->devname, "/dev/md/"), |
186 | mdlist->devname); | |
187 | } | |
e0d19036 | 188 | st->next = statelist; |
4dd2df09 | 189 | st->devnm[0] = 0; |
9dad51d4 | 190 | st->percent = RESYNC_UNKNOWN; |
f1661bd7 | 191 | st->from_config = 1; |
feb716e9 | 192 | st->expected_spares = mdlist->spare_disks; |
e0d19036 | 193 | if (mdlist->spare_group) |
503975b9 | 194 | st->spare_group = xstrdup(mdlist->spare_group); |
e0d19036 NB |
195 | statelist = st; |
196 | } | |
197 | } else { | |
a655e550 | 198 | struct mddev_dev *dv; |
f566ef45 JS |
199 | |
200 | for (dv = devlist; dv; dv = dv->next) { | |
503975b9 | 201 | struct state *st = xcalloc(1, sizeof *st); |
9f3dd454 | 202 | mdlist = conf_get_ident(dv->devname); |
503975b9 | 203 | st->devname = xstrdup(dv->devname); |
e0d19036 | 204 | st->next = statelist; |
4dd2df09 | 205 | st->devnm[0] = 0; |
9dad51d4 | 206 | st->percent = RESYNC_UNKNOWN; |
feb716e9 | 207 | st->expected_spares = -1; |
e5329c37 NB |
208 | if (mdlist) { |
209 | st->expected_spares = mdlist->spare_disks; | |
210 | if (mdlist->spare_group) | |
503975b9 | 211 | st->spare_group = xstrdup(mdlist->spare_group); |
e5329c37 | 212 | } |
e0d19036 NB |
213 | statelist = st; |
214 | } | |
215 | } | |
216 | ||
f566ef45 | 217 | while (!finished) { |
aa88f531 | 218 | int new_found = 0; |
73ff0732 | 219 | struct state *st, **stp; |
a90e1050 | 220 | int anydegraded = 0; |
007087d0 | 221 | int anyredundant = 0; |
e0d19036 NB |
222 | |
223 | if (mdstat) | |
224 | free_mdstat(mdstat); | |
f566ef45 | 225 | mdstat = mdstat_read(oneshot ? 0 : 1, 0); |
e0d19036 | 226 | |
007087d0 | 227 | for (st = statelist; st; st = st->next) { |
95c50205 N |
228 | if (check_array(st, mdstat, c->test, &info, |
229 | increments, c->prefer)) | |
a90e1050 | 230 | anydegraded = 1; |
007087d0 MT |
231 | /* for external arrays, metadata is filled for |
232 | * containers only | |
233 | */ | |
234 | if (st->metadata && st->metadata->ss->external) | |
235 | continue; | |
236 | if (st->err == 0 && !anyredundant) | |
237 | anyredundant = 1; | |
238 | } | |
1011e834 | 239 | |
e0d19036 | 240 | /* now check if there are any new devices found in mdstat */ |
95c50205 N |
241 | if (c->scan) |
242 | new_found = add_new_arrays(mdstat, &statelist, c->test, | |
e0bd6a96 | 243 | &info); |
2e0172b1 | 244 | |
e0d19036 NB |
245 | /* If an array has active < raid && spare == 0 && spare_group != NULL |
246 | * Look for another array with spare > 0 and active == raid and same spare_group | |
247 | * if found, choose a device and hotremove/hotadd | |
248 | */ | |
a90e1050 | 249 | if (share && anydegraded) |
e0bd6a96 | 250 | try_spare_migration(statelist, &info); |
aa88f531 NB |
251 | if (!new_found) { |
252 | if (oneshot) | |
253 | break; | |
007087d0 MT |
254 | else if (!anyredundant) { |
255 | break; | |
256 | } | |
e2308733 | 257 | else { |
cab9c67d BK |
258 | int wait_result = mdstat_wait(delay_for_event); |
259 | ||
260 | /* | |
261 | * If mdmonitor is awaken by event, set small delay once | |
262 | * to deal with udev and mdadm. | |
263 | */ | |
264 | if (wait_result != 0) { | |
265 | if (c->delay > 5) | |
266 | delay_for_event = 5; | |
267 | } else | |
268 | delay_for_event = c->delay; | |
269 | ||
e2308733 MT |
270 | mdstat_close(); |
271 | } | |
aa88f531 | 272 | } |
95c50205 | 273 | c->test = 0; |
73ff0732 N |
274 | |
275 | for (stp = &statelist; (st = *stp) != NULL; ) { | |
276 | if (st->from_auto && st->err > 5) { | |
277 | *stp = st->next; | |
278 | free(st->devname); | |
279 | free(st->spare_group); | |
280 | free(st); | |
281 | } else | |
282 | stp = &st->next; | |
283 | } | |
52826846 | 284 | } |
b657208c JS |
285 | for (st2 = statelist; st2; st2 = statelist) { |
286 | statelist = st2->next; | |
287 | free(st2); | |
288 | } | |
289 | ||
b5e64645 NB |
290 | if (pidfile) |
291 | unlink(pidfile); | |
52826846 NB |
292 | return 0; |
293 | } | |
294 | ||
2e0172b1 N |
295 | static int make_daemon(char *pidfile) |
296 | { | |
44d337f0 AC |
297 | /* Return: |
298 | * -1 in the forked daemon | |
299 | * 0 in the parent | |
300 | * 1 on error | |
301 | * so a none-negative becomes the exit code. | |
302 | */ | |
2e0172b1 N |
303 | int pid = fork(); |
304 | if (pid > 0) { | |
305 | if (!pidfile) | |
306 | printf("%d\n", pid); | |
307 | else { | |
ca4b156b MT |
308 | FILE *pid_file = NULL; |
309 | int fd = open(pidfile, O_WRONLY | O_CREAT | O_TRUNC, | |
310 | 0644); | |
311 | if (fd >= 0) | |
312 | pid_file = fdopen(fd, "w"); | |
2e0172b1 N |
313 | if (!pid_file) |
314 | perror("cannot create pid file"); | |
315 | else { | |
316 | fprintf(pid_file,"%d\n", pid); | |
317 | fclose(pid_file); | |
318 | } | |
319 | } | |
320 | return 0; | |
321 | } | |
322 | if (pid < 0) { | |
323 | perror("daemonise"); | |
324 | return 1; | |
325 | } | |
ff6bb131 | 326 | manage_fork_fds(0); |
2e0172b1 | 327 | setsid(); |
44d337f0 | 328 | return -1; |
2e0172b1 N |
329 | } |
330 | ||
331 | static int check_one_sharer(int scan) | |
332 | { | |
185ec439 CL |
333 | int pid; |
334 | FILE *comm_fp; | |
2e0172b1 | 335 | FILE *fp; |
7f3b2d1d BK |
336 | char comm_path[PATH_MAX]; |
337 | char path[PATH_MAX]; | |
185ec439 CL |
338 | char comm[20]; |
339 | ||
0011874f JS |
340 | sprintf(path, "%s/autorebuild.pid", MDMON_DIR); |
341 | fp = fopen(path, "r"); | |
2e0172b1 | 342 | if (fp) { |
71204a50 N |
343 | if (fscanf(fp, "%d", &pid) != 1) |
344 | pid = -1; | |
185ec439 CL |
345 | snprintf(comm_path, sizeof(comm_path), |
346 | "/proc/%d/comm", pid); | |
347 | comm_fp = fopen(comm_path, "r"); | |
348 | if (comm_fp) { | |
349 | if (fscanf(comm_fp, "%s", comm) && | |
350 | strncmp(basename(comm), Name, strlen(Name)) == 0) { | |
351 | if (scan) { | |
352 | pr_err("Only one autorebuild process allowed in scan mode, aborting\n"); | |
353 | fclose(comm_fp); | |
354 | fclose(fp); | |
355 | return 1; | |
356 | } else { | |
357 | pr_err("Warning: One autorebuild process already running.\n"); | |
358 | } | |
2e0172b1 | 359 | } |
185ec439 | 360 | fclose(comm_fp); |
2e0172b1 N |
361 | } |
362 | fclose(fp); | |
363 | } | |
7f3b2d1d BK |
364 | return 0; |
365 | } | |
366 | ||
367 | static void write_autorebuild_pid() | |
368 | { | |
369 | char path[PATH_MAX]; | |
370 | int pid; | |
ca4b156b | 371 | FILE *fp = NULL; |
7f3b2d1d BK |
372 | sprintf(path, "%s/autorebuild.pid", MDMON_DIR); |
373 | ||
ca4b156b | 374 | if (mkdir(MDMON_DIR, 0700) < 0 && errno != EEXIST) { |
7f3b2d1d BK |
375 | pr_err("Can't create autorebuild.pid file\n"); |
376 | } else { | |
ca4b156b MT |
377 | int fd = open(path, O_WRONLY | O_CREAT | O_TRUNC, 0700); |
378 | ||
379 | if (fd >= 0) | |
380 | fp = fdopen(fd, "w"); | |
381 | ||
7f3b2d1d | 382 | if (!fp) |
7a862a02 | 383 | pr_err("Can't create autorebuild.pid file\n"); |
7f3b2d1d BK |
384 | else { |
385 | pid = getpid(); | |
386 | fprintf(fp, "%d\n", pid); | |
387 | fclose(fp); | |
2e0172b1 N |
388 | } |
389 | } | |
2e0172b1 | 390 | } |
52826846 | 391 | |
e0bd6a96 | 392 | static void alert(char *event, char *dev, char *disc, struct alert_info *info) |
52826846 | 393 | { |
773135f5 NB |
394 | int priority; |
395 | ||
b1bac75b | 396 | if (!info->alert_cmd && !info->mailaddr && !info->dosyslog) { |
cd29a5c8 | 397 | time_t now = time(0); |
aba69144 | 398 | |
f566ef45 JS |
399 | printf("%1.15s: %s on %s %s\n", ctime(&now) + 4, |
400 | event, dev, disc?disc:"unknown device"); | |
cd29a5c8 | 401 | } |
e0bd6a96 | 402 | if (info->alert_cmd) { |
52826846 NB |
403 | int pid = fork(); |
404 | switch(pid) { | |
405 | default: | |
406 | waitpid(pid, NULL, 0); | |
407 | break; | |
408 | case -1: | |
409 | break; | |
410 | case 0: | |
e0bd6a96 N |
411 | execl(info->alert_cmd, info->alert_cmd, |
412 | event, dev, disc, NULL); | |
52826846 NB |
413 | exit(2); |
414 | } | |
415 | } | |
f566ef45 JS |
416 | if (info->mailaddr && (strncmp(event, "Fail", 4) == 0 || |
417 | strncmp(event, "Test", 4) == 0 || | |
418 | strncmp(event, "Spares", 6) == 0 || | |
419 | strncmp(event, "Degrade", 7) == 0)) { | |
52826846 NB |
420 | FILE *mp = popen(Sendmail, "w"); |
421 | if (mp) { | |
008e1100 | 422 | FILE *mdstat; |
52826846 NB |
423 | char hname[256]; |
424 | gethostname(hname, sizeof(hname)); | |
425 | signal(SIGPIPE, SIG_IGN); | |
e0bd6a96 N |
426 | if (info->mailfrom) |
427 | fprintf(mp, "From: %s\n", info->mailfrom); | |
4948b8f7 | 428 | else |
f566ef45 JS |
429 | fprintf(mp, "From: %s monitoring <root>\n", |
430 | Name); | |
e0bd6a96 N |
431 | fprintf(mp, "To: %s\n", info->mailaddr); |
432 | fprintf(mp, "Subject: %s event on %s:%s\n\n", | |
433 | event, dev, hname); | |
52826846 | 434 | |
e0bd6a96 | 435 | fprintf(mp, |
7a862a02 | 436 | "This is an automatically generated mail message from %s\n", Name); |
52826846 NB |
437 | fprintf(mp, "running on %s\n\n", hname); |
438 | ||
e0bd6a96 | 439 | fprintf(mp, |
7a862a02 | 440 | "A %s event had been detected on md device %s.\n\n", event, dev); |
52826846 | 441 | |
37dfc3d6 | 442 | if (disc && disc[0] != ' ') |
e0bd6a96 | 443 | fprintf(mp, |
7a862a02 | 444 | "It could be related to component device %s.\n\n", disc); |
37dfc3d6 NB |
445 | if (disc && disc[0] == ' ') |
446 | fprintf(mp, "Extra information:%s.\n\n", disc); | |
52826846 NB |
447 | |
448 | fprintf(mp, "Faithfully yours, etc.\n"); | |
008e1100 NB |
449 | |
450 | mdstat = fopen("/proc/mdstat", "r"); | |
451 | if (mdstat) { | |
452 | char buf[8192]; | |
453 | int n; | |
e0bd6a96 | 454 | fprintf(mp, |
7a862a02 | 455 | "\nP.S. The /proc/mdstat file currently contains the following:\n\n"); |
f566ef45 JS |
456 | while ((n = fread(buf, 1, sizeof(buf), |
457 | mdstat)) > 0) | |
458 | n = fwrite(buf, 1, n, mp); | |
008e1100 NB |
459 | fclose(mdstat); |
460 | } | |
6278fb3a | 461 | pclose(mp); |
52826846 | 462 | } |
52826846 | 463 | } |
773135f5 NB |
464 | |
465 | /* log the event to syslog maybe */ | |
e0bd6a96 | 466 | if (info->dosyslog) { |
773135f5 NB |
467 | /* Log at a different severity depending on the event. |
468 | * | |
469 | * These are the critical events: */ | |
f566ef45 JS |
470 | if (strncmp(event, "Fail", 4) == 0 || |
471 | strncmp(event, "Degrade", 7) == 0 || | |
472 | strncmp(event, "DeviceDisappeared", 17) == 0) | |
773135f5 NB |
473 | priority = LOG_CRIT; |
474 | /* Good to know about, but are not failures: */ | |
f566ef45 JS |
475 | else if (strncmp(event, "Rebuild", 7) == 0 || |
476 | strncmp(event, "MoveSpare", 9) == 0 || | |
d1732eeb | 477 | strncmp(event, "Spares", 6) != 0) |
773135f5 NB |
478 | priority = LOG_WARNING; |
479 | /* Everything else: */ | |
480 | else | |
481 | priority = LOG_INFO; | |
482 | ||
b033913a | 483 | if (disc && disc[0] != ' ') |
e0bd6a96 | 484 | syslog(priority, |
7a862a02 | 485 | "%s event detected on md device %s, component device %s", event, dev, disc); |
b033913a N |
486 | else if (disc) |
487 | syslog(priority, | |
488 | "%s event detected on md device %s: %s", | |
489 | event, dev, disc); | |
773135f5 | 490 | else |
e0bd6a96 N |
491 | syslog(priority, |
492 | "%s event detected on md device %s", | |
493 | event, dev); | |
773135f5 | 494 | } |
52826846 | 495 | } |
b90c0e9a | 496 | |
a90e1050 | 497 | static int check_array(struct state *st, struct mdstat_ent *mdstat, |
e0bd6a96 | 498 | int test, struct alert_info *ainfo, |
c2ecf5f6 | 499 | int increments, char *prefer) |
2e0172b1 | 500 | { |
ff044d6b AC |
501 | /* Update the state 'st' to reflect any changes shown in mdstat, |
502 | * or found by directly examining the array, and return | |
503 | * '1' if the array is degraded, or '0' if it is optimal (or dead). | |
504 | */ | |
b0599bda | 505 | struct { int state, major, minor; } info[MAX_DISKS]; |
aed5f5c3 | 506 | struct mdinfo *sra = NULL; |
2e0172b1 N |
507 | mdu_array_info_t array; |
508 | struct mdstat_ent *mse = NULL, *mse2; | |
509 | char *dev = st->devname; | |
b9a0309c | 510 | int fd; |
2e0172b1 | 511 | int i; |
9e6d9291 N |
512 | int remaining_disks; |
513 | int last_disk; | |
721b662b | 514 | int new_array = 0; |
1830e74b | 515 | int retval; |
802961a2 | 516 | int is_container = 0; |
b3ab4e4d | 517 | unsigned long redundancy_only_flags = 0; |
2e0172b1 N |
518 | |
519 | if (test) | |
e0bd6a96 | 520 | alert("TestMessage", dev, NULL, ainfo); |
b9a0309c | 521 | |
1830e74b JS |
522 | retval = 0; |
523 | ||
2e0172b1 | 524 | fd = open(dev, O_RDONLY); |
13e5d845 JS |
525 | if (fd < 0) |
526 | goto disappeared; | |
b9a0309c | 527 | |
802961a2 MT |
528 | if (st->devnm[0] == 0) |
529 | strcpy(st->devnm, fd2devnm(fd)); | |
530 | ||
531 | for (mse2 = mdstat; mse2; mse2 = mse2->next) | |
532 | if (strcmp(mse2->devnm, st->devnm) == 0) { | |
533 | mse2->devnm[0] = 0; /* flag it as "used" */ | |
534 | mse = mse2; | |
535 | } | |
536 | ||
537 | if (!mse) { | |
538 | /* duplicated array in statelist | |
539 | * or re-created after reading mdstat | |
540 | */ | |
541 | st->err++; | |
542 | goto out; | |
543 | } | |
544 | ||
545 | if (mse->level == NULL) | |
546 | is_container = 1; | |
547 | ||
2dab69c9 | 548 | if (!is_container && !md_array_active(fd)) |
13e5d845 | 549 | goto disappeared; |
b9a0309c | 550 | |
2e0172b1 | 551 | fcntl(fd, F_SETFD, FD_CLOEXEC); |
13e5d845 JS |
552 | if (md_get_array_info(fd, &array) < 0) |
553 | goto disappeared; | |
554 | ||
b3ab4e4d MT |
555 | if (!is_container && map_name(pers, mse->level) > 0) |
556 | redundancy_only_flags |= GET_MISMATCH; | |
802961a2 MT |
557 | |
558 | sra = sysfs_read(-1, st->devnm, GET_LEVEL | GET_DISKS | GET_DEVS | | |
b3ab4e4d | 559 | GET_STATE | redundancy_only_flags); |
aed5f5c3 | 560 | |
aed5f5c3 JS |
561 | if (!sra) |
562 | goto disappeared; | |
563 | ||
2e0172b1 N |
564 | /* It's much easier to list what array levels can't |
565 | * have a device disappear than all of them that can | |
566 | */ | |
48bc2ade | 567 | if (sra->array.level == 0 || sra->array.level == -1) { |
f1661bd7 | 568 | if (!st->err && !st->from_config) |
04e27c20 | 569 | alert("DeviceDisappeared", dev, " Wrong-Level", ainfo); |
73ff0732 | 570 | st->err++; |
1830e74b | 571 | goto out; |
2e0172b1 | 572 | } |
2e0172b1 | 573 | |
2e0172b1 N |
574 | /* this array is in /proc/mdstat */ |
575 | if (array.utime == 0) | |
576 | /* external arrays don't update utime, so | |
577 | * just make sure it is always different. */ | |
578 | array.utime = st->utime + 1;; | |
579 | ||
0f760384 | 580 | if (st->err) { |
73ff0732 | 581 | /* New array appeared where previously had an error */ |
0f760384 N |
582 | st->err = 0; |
583 | st->percent = RESYNC_NONE; | |
721b662b | 584 | new_array = 1; |
007087d0 MT |
585 | if (!is_container) |
586 | alert("NewArray", st->devname, NULL, ainfo); | |
0f760384 N |
587 | } |
588 | ||
b8e5713c | 589 | if (st->utime == array.utime && st->failed == sra->array.failed_disks && |
e5eb6857 | 590 | st->working == sra->array.working_disks && |
b98943a4 | 591 | st->spare == sra->array.spare_disks && |
d7be7d87 | 592 | (mse == NULL || (mse->percent == st->percent))) { |
ff044d6b | 593 | if ((st->active < st->raid) && st->spare == 0) |
1830e74b JS |
594 | retval = 1; |
595 | goto out; | |
2e0172b1 N |
596 | } |
597 | if (st->utime == 0 && /* new array */ | |
f27904a5 | 598 | mse->pattern && strchr(mse->pattern, '_') /* degraded */) |
e0bd6a96 | 599 | alert("DegradedArray", dev, NULL, ainfo); |
2e0172b1 | 600 | |
b98943a4 JS |
601 | if (st->utime == 0 && /* new array */ st->expected_spares > 0 && |
602 | sra->array.spare_disks < st->expected_spares) | |
e0bd6a96 | 603 | alert("SparesMissing", dev, NULL, ainfo); |
9dad51d4 | 604 | if (st->percent < 0 && st->percent != RESYNC_UNKNOWN && |
2e0172b1 | 605 | mse->percent >= 0) |
e0bd6a96 | 606 | alert("RebuildStarted", dev, NULL, ainfo); |
f27904a5 | 607 | if (st->percent >= 0 && mse->percent >= 0 && |
2e0172b1 | 608 | (mse->percent / increments) > (st->percent / increments)) { |
97589839 | 609 | char percentalert[18]; |
f27904a5 JS |
610 | /* |
611 | * "RebuildNN" (10 chars) or "RebuildStarted" (15 chars) | |
612 | */ | |
2e0172b1 N |
613 | |
614 | if((mse->percent / increments) == 0) | |
f27904a5 JS |
615 | snprintf(percentalert, sizeof(percentalert), |
616 | "RebuildStarted"); | |
2e0172b1 | 617 | else |
f27904a5 JS |
618 | snprintf(percentalert, sizeof(percentalert), |
619 | "Rebuild%02d", mse->percent); | |
2e0172b1 | 620 | |
e0bd6a96 | 621 | alert(percentalert, dev, NULL, ainfo); |
2e0172b1 N |
622 | } |
623 | ||
f27904a5 | 624 | if (mse->percent == RESYNC_NONE && st->percent >= 0) { |
2e0172b1 N |
625 | /* Rebuild/sync/whatever just finished. |
626 | * If there is a number in /mismatch_cnt, | |
627 | * we should report that. | |
628 | */ | |
2e0172b1 | 629 | if (sra && sra->mismatch_cnt > 0) { |
8453f8d0 LD |
630 | char cnt[80]; |
631 | snprintf(cnt, sizeof(cnt), | |
632 | " mismatches found: %d (on raid level %d)", | |
f566ef45 | 633 | sra->mismatch_cnt, sra->array.level); |
e0bd6a96 | 634 | alert("RebuildFinished", dev, cnt, ainfo); |
2e0172b1 | 635 | } else |
e0bd6a96 | 636 | alert("RebuildFinished", dev, NULL, ainfo); |
2e0172b1 N |
637 | } |
638 | st->percent = mse->percent; | |
639 | ||
b98943a4 | 640 | remaining_disks = sra->array.nr_disks; |
f27904a5 | 641 | for (i = 0; i < MAX_DISKS && remaining_disks > 0; i++) { |
2e0172b1 N |
642 | mdu_disk_info_t disc; |
643 | disc.number = i; | |
d97572f5 | 644 | if (md_get_disk_info(fd, &disc) >= 0) { |
2e0172b1 N |
645 | info[i].state = disc.state; |
646 | info[i].major = disc.major; | |
647 | info[i].minor = disc.minor; | |
9e6d9291 N |
648 | if (disc.major || disc.minor) |
649 | remaining_disks --; | |
2e0172b1 N |
650 | } else |
651 | info[i].major = info[i].minor = 0; | |
652 | } | |
9e6d9291 | 653 | last_disk = i; |
2e0172b1 | 654 | |
4019ad07 JL |
655 | if (mse->metadata_version && |
656 | strncmp(mse->metadata_version, "external:", 9) == 0 && | |
4dd2df09 N |
657 | is_subarray(mse->metadata_version+9)) { |
658 | char *sl; | |
f566ef45 | 659 | strcpy(st->parent_devnm, mse->metadata_version + 10); |
4dd2df09 N |
660 | sl = strchr(st->parent_devnm, '/'); |
661 | if (sl) | |
662 | *sl = 0; | |
663 | } else | |
664 | st->parent_devnm[0] = 0; | |
f27904a5 | 665 | if (st->metadata == NULL && st->parent_devnm[0] == 0) |
2e0172b1 N |
666 | st->metadata = super_by_fd(fd, NULL); |
667 | ||
f566ef45 JS |
668 | for (i = 0; i < MAX_DISKS; i++) { |
669 | mdu_disk_info_t disc = {0, 0, 0, 0, 0}; | |
670 | int newstate = 0; | |
2e0172b1 N |
671 | int change; |
672 | char *dv = NULL; | |
673 | disc.number = i; | |
f27904a5 | 674 | if (i < last_disk && (info[i].major || info[i].minor)) { |
2e0172b1 | 675 | newstate = info[i].state; |
f27904a5 JS |
676 | dv = map_dev_preferred(info[i].major, info[i].minor, 1, |
677 | prefer); | |
2e0172b1 N |
678 | disc.state = newstate; |
679 | disc.major = info[i].major; | |
680 | disc.minor = info[i].minor; | |
721b662b N |
681 | } else |
682 | newstate = (1 << MD_DISK_REMOVED); | |
683 | ||
2e0172b1 | 684 | if (dv == NULL && st->devid[i]) |
f27904a5 JS |
685 | dv = map_dev_preferred(major(st->devid[i]), |
686 | minor(st->devid[i]), 1, prefer); | |
2e0172b1 | 687 | change = newstate ^ st->devstate[i]; |
721b662b | 688 | if (st->utime && change && !st->err && !new_array) { |
f27904a5 | 689 | if ((st->devstate[i]&change) & (1 << MD_DISK_SYNC)) |
e0bd6a96 | 690 | alert("Fail", dev, dv, ainfo); |
f27904a5 | 691 | else if ((newstate & (1 << MD_DISK_FAULTY)) && |
2e0172b1 | 692 | (disc.major || disc.minor) && |
f27904a5 JS |
693 | st->devid[i] == makedev(disc.major, |
694 | disc.minor)) | |
e0bd6a96 | 695 | alert("FailSpare", dev, dv, ainfo); |
f27904a5 | 696 | else if ((newstate&change) & (1 << MD_DISK_SYNC)) |
e0bd6a96 | 697 | alert("SpareActive", dev, dv, ainfo); |
2e0172b1 N |
698 | } |
699 | st->devstate[i] = newstate; | |
700 | st->devid[i] = makedev(disc.major, disc.minor); | |
701 | } | |
b98943a4 | 702 | st->active = sra->array.active_disks; |
e5eb6857 | 703 | st->working = sra->array.working_disks; |
b98943a4 | 704 | st->spare = sra->array.spare_disks; |
b8e5713c | 705 | st->failed = sra->array.failed_disks; |
2e0172b1 | 706 | st->utime = array.utime; |
12a9d21f | 707 | st->raid = sra->array.raid_disks; |
2e0172b1 | 708 | st->err = 0; |
a90e1050 | 709 | if ((st->active < st->raid) && st->spare == 0) |
1830e74b JS |
710 | retval = 1; |
711 | ||
712 | out: | |
aed5f5c3 JS |
713 | if (sra) |
714 | sysfs_free(sra); | |
802961a2 | 715 | if (fd >= 0) |
13e5d845 | 716 | close(fd); |
1830e74b | 717 | return retval; |
13e5d845 JS |
718 | |
719 | disappeared: | |
007087d0 | 720 | if (!st->err && !is_container) |
13e5d845 JS |
721 | alert("DeviceDisappeared", dev, NULL, ainfo); |
722 | st->err++; | |
723 | goto out; | |
2e0172b1 N |
724 | } |
725 | ||
83f3bc5f | 726 | static int add_new_arrays(struct mdstat_ent *mdstat, struct state **statelist, |
e0bd6a96 | 727 | int test, struct alert_info *info) |
2e0172b1 N |
728 | { |
729 | struct mdstat_ent *mse; | |
730 | int new_found = 0; | |
1e08717f | 731 | char *name; |
2e0172b1 | 732 | |
f566ef45 | 733 | for (mse = mdstat; mse; mse = mse->next) |
d7be7d87 | 734 | if (mse->devnm[0] && (!mse->level || /* retrieve containers */ |
f566ef45 JS |
735 | (strcmp(mse->level, "raid0") != 0 && |
736 | strcmp(mse->level, "linear") != 0))) { | |
503975b9 | 737 | struct state *st = xcalloc(1, sizeof *st); |
2e0172b1 N |
738 | mdu_array_info_t array; |
739 | int fd; | |
1e08717f SV |
740 | |
741 | name = get_md_name(mse->devnm); | |
742 | if (!name) { | |
743 | free(st); | |
744 | continue; | |
745 | } | |
746 | ||
747 | st->devname = xstrdup(name); | |
2e0172b1 | 748 | if ((fd = open(st->devname, O_RDONLY)) < 0 || |
9cd39f01 | 749 | md_get_array_info(fd, &array) < 0) { |
2e0172b1 | 750 | /* no such array */ |
9cd39f01 JS |
751 | if (fd >= 0) |
752 | close(fd); | |
2e0172b1 N |
753 | put_md_name(st->devname); |
754 | free(st->devname); | |
755 | if (st->metadata) { | |
756 | st->metadata->ss->free_super(st->metadata); | |
757 | free(st->metadata); | |
758 | } | |
759 | free(st); | |
760 | continue; | |
761 | } | |
762 | close(fd); | |
83f3bc5f | 763 | st->next = *statelist; |
2e0172b1 | 764 | st->err = 1; |
73ff0732 | 765 | st->from_auto = 1; |
4dd2df09 | 766 | strcpy(st->devnm, mse->devnm); |
9dad51d4 | 767 | st->percent = RESYNC_UNKNOWN; |
2e0172b1 | 768 | st->expected_spares = -1; |
eb28e119 | 769 | if (mse->metadata_version && |
f566ef45 JS |
770 | strncmp(mse->metadata_version, |
771 | "external:", 9) == 0 && | |
4dd2df09 N |
772 | is_subarray(mse->metadata_version+9)) { |
773 | char *sl; | |
774 | strcpy(st->parent_devnm, | |
775 | mse->metadata_version+10); | |
776 | sl = strchr(st->parent_devnm, '/'); | |
777 | *sl = 0; | |
778 | } else | |
779 | st->parent_devnm[0] = 0; | |
83f3bc5f | 780 | *statelist = st; |
2e0172b1 | 781 | if (test) |
e0bd6a96 | 782 | alert("TestMessage", st->devname, NULL, info); |
2e0172b1 N |
783 | new_found = 1; |
784 | } | |
785 | return new_found; | |
786 | } | |
787 | ||
fbfdcb06 AO |
788 | static int get_required_spare_criteria(struct state *st, |
789 | struct spare_criteria *sc) | |
80e7f8c3 AC |
790 | { |
791 | int fd; | |
80e7f8c3 | 792 | |
f566ef45 | 793 | if (!st->metadata || !st->metadata->ss->get_spare_criteria) { |
fbfdcb06 | 794 | sc->min_size = 0; |
4b57ecf6 | 795 | sc->sector_size = 0; |
de697acc | 796 | return 0; |
300f5033 | 797 | } |
80e7f8c3 AC |
798 | |
799 | fd = open(st->devname, O_RDONLY); | |
800 | if (fd < 0) | |
de697acc AC |
801 | return 1; |
802 | if (st->metadata->ss->external) | |
803 | st->metadata->ss->load_container(st->metadata, fd, st->devname); | |
804 | else | |
805 | st->metadata->ss->load_super(st->metadata, fd, st->devname); | |
80e7f8c3 | 806 | close(fd); |
de697acc AC |
807 | if (!st->metadata->sb) |
808 | return 1; | |
fbfdcb06 AO |
809 | |
810 | st->metadata->ss->get_spare_criteria(st->metadata, sc); | |
80e7f8c3 AC |
811 | st->metadata->ss->free_super(st->metadata); |
812 | ||
de697acc | 813 | return 0; |
80e7f8c3 AC |
814 | } |
815 | ||
5ec0f373 | 816 | static int check_donor(struct state *from, struct state *to) |
2feb22ef | 817 | { |
66f5c4b6 N |
818 | struct state *sub; |
819 | ||
2feb22ef N |
820 | if (from == to) |
821 | return 0; | |
66f5c4b6 N |
822 | if (from->parent) |
823 | /* Cannot move from a member */ | |
2feb22ef | 824 | return 0; |
ff044d6b AC |
825 | if (from->err) |
826 | return 0; | |
66f5c4b6 N |
827 | for (sub = from->subarray; sub; sub = sub->subarray) |
828 | /* If source array has degraded subarrays, don't | |
829 | * remove anything | |
830 | */ | |
831 | if (sub->active < sub->raid) | |
832 | return 0; | |
833 | if (from->metadata->ss->external == 0) | |
834 | if (from->active < from->raid) | |
835 | return 0; | |
2feb22ef N |
836 | if (from->spare <= 0) |
837 | return 0; | |
e78dda3b | 838 | return 1; |
2feb22ef N |
839 | } |
840 | ||
0f0749ad | 841 | static dev_t choose_spare(struct state *from, struct state *to, |
f566ef45 | 842 | struct domainlist *domlist, struct spare_criteria *sc) |
0fa21e85 N |
843 | { |
844 | int d; | |
0f0749ad | 845 | dev_t dev = 0; |
0fa21e85 | 846 | |
b0599bda | 847 | for (d = from->raid; !dev && d < MAX_DISKS; d++) { |
f566ef45 | 848 | if (from->devid[d] > 0 && from->devstate[d] == 0) { |
0fa21e85 N |
849 | struct dev_policy *pol; |
850 | unsigned long long dev_size; | |
4b57ecf6 | 851 | unsigned int dev_sector_size; |
0fa21e85 | 852 | |
bfd76b93 CA |
853 | if (to->metadata->ss->external && |
854 | test_partition_from_id(from->devid[d])) | |
855 | continue; | |
856 | ||
fbfdcb06 | 857 | if (sc->min_size && |
0fa21e85 | 858 | dev_size_from_id(from->devid[d], &dev_size) && |
fbfdcb06 | 859 | dev_size < sc->min_size) |
0fa21e85 N |
860 | continue; |
861 | ||
4b57ecf6 AO |
862 | if (sc->sector_size && |
863 | dev_sector_size_from_id(from->devid[d], | |
864 | &dev_sector_size) && | |
865 | sc->sector_size != dev_sector_size) | |
866 | continue; | |
867 | ||
4dd2df09 | 868 | pol = devid_policy(from->devid[d]); |
0fa21e85 N |
869 | if (from->spare_group) |
870 | pol_add(&pol, pol_domain, | |
871 | from->spare_group, NULL); | |
f566ef45 JS |
872 | if (domain_test(domlist, pol, |
873 | to->metadata->ss->name) == 1) | |
0fa21e85 N |
874 | dev = from->devid[d]; |
875 | dev_policy_free(pol); | |
876 | } | |
877 | } | |
878 | return dev; | |
879 | } | |
880 | ||
0f0749ad | 881 | static dev_t container_choose_spare(struct state *from, struct state *to, |
f0b85306 | 882 | struct domainlist *domlist, |
fbfdcb06 | 883 | struct spare_criteria *sc, int active) |
5739e0d0 N |
884 | { |
885 | /* This is similar to choose_spare, but we cannot trust devstate, | |
886 | * so we need to read the metadata instead | |
887 | */ | |
326727d9 | 888 | struct mdinfo *list; |
5739e0d0 | 889 | struct supertype *st = from->metadata; |
ff044d6b | 890 | int fd = open(from->devname, O_RDONLY); |
5739e0d0 | 891 | int err; |
0f0749ad | 892 | dev_t dev = 0; |
5739e0d0 N |
893 | |
894 | if (fd < 0) | |
895 | return 0; | |
326727d9 AC |
896 | if (!st->ss->getinfo_super_disks) { |
897 | close(fd); | |
5739e0d0 | 898 | return 0; |
326727d9 | 899 | } |
1011e834 | 900 | |
5739e0d0 N |
901 | err = st->ss->load_container(st, fd, NULL); |
902 | close(fd); | |
903 | if (err) | |
904 | return 0; | |
1011e834 | 905 | |
a1e49d69 CA |
906 | if (from == to) { |
907 | /* We must check if number of active disks has not increased | |
908 | * since ioctl in main loop. mdmon may have added spare | |
909 | * to subarray. If so we do not need to look for more spares | |
910 | * so return non zero value */ | |
911 | int active_cnt = 0; | |
912 | struct mdinfo *dp; | |
913 | list = st->ss->getinfo_super_disks(st); | |
914 | if (!list) { | |
915 | st->ss->free_super(st); | |
916 | return 1; | |
917 | } | |
918 | dp = list->devs; | |
919 | while (dp) { | |
f566ef45 JS |
920 | if (dp->disk.state & (1 << MD_DISK_SYNC) && |
921 | !(dp->disk.state & (1 << MD_DISK_FAULTY))) | |
a1e49d69 CA |
922 | active_cnt++; |
923 | dp = dp->next; | |
924 | } | |
925 | sysfs_free(list); | |
926 | if (active < active_cnt) { | |
927 | /* Spare just activated.*/ | |
928 | st->ss->free_super(st); | |
929 | return 1; | |
930 | } | |
931 | } | |
932 | ||
326727d9 | 933 | /* We only need one spare so full list not needed */ |
fbfdcb06 | 934 | list = container_choose_spares(st, sc, domlist, from->spare_group, |
326727d9 AC |
935 | to->metadata->ss->name, 1); |
936 | if (list) { | |
937 | struct mdinfo *disks = list->devs; | |
938 | if (disks) | |
939 | dev = makedev(disks->disk.major, disks->disk.minor); | |
940 | sysfs_free(list); | |
5739e0d0 | 941 | } |
326727d9 | 942 | st->ss->free_super(st); |
5739e0d0 N |
943 | return dev; |
944 | } | |
945 | ||
e0bd6a96 | 946 | static void try_spare_migration(struct state *statelist, struct alert_info *info) |
2e0172b1 | 947 | { |
66f5c4b6 N |
948 | struct state *from; |
949 | struct state *st; | |
fbfdcb06 | 950 | struct spare_criteria sc; |
c3621c0a ML |
951 | |
952 | link_containers_with_subarrays(statelist); | |
66f5c4b6 | 953 | for (st = statelist; st; st = st->next) |
f566ef45 | 954 | if (st->active < st->raid && st->spare == 0 && !st->err) { |
e78dda3b N |
955 | struct domainlist *domlist = NULL; |
956 | int d; | |
66f5c4b6 N |
957 | struct state *to = st; |
958 | ||
4dd2df09 | 959 | if (to->parent_devnm[0] && !to->parent) |
c0dc0ad5 CA |
960 | /* subarray monitored without parent container |
961 | * we can't move spares here */ | |
962 | continue; | |
1011e834 | 963 | |
66f5c4b6 N |
964 | if (to->parent) |
965 | /* member of a container */ | |
966 | to = to->parent; | |
e78dda3b | 967 | |
fbfdcb06 | 968 | if (get_required_spare_criteria(to, &sc)) |
de697acc | 969 | continue; |
e9a2ac02 AC |
970 | if (to->metadata->ss->external) { |
971 | /* We must make sure there is | |
972 | * no suitable spare in container already. | |
973 | * If there is we don't add more */ | |
974 | dev_t devid = container_choose_spare( | |
fbfdcb06 | 975 | to, to, NULL, &sc, st->active); |
e9a2ac02 AC |
976 | if (devid > 0) |
977 | continue; | |
978 | } | |
b0599bda | 979 | for (d = 0; d < MAX_DISKS; d++) |
e78dda3b N |
980 | if (to->devid[d]) |
981 | domainlist_add_dev(&domlist, | |
982 | to->devid[d], | |
983 | to->metadata->ss->name); | |
984 | if (to->spare_group) | |
985 | domain_add(&domlist, to->spare_group); | |
5ec0f373 ML |
986 | /* |
987 | * No spare migration if the destination | |
988 | * has no domain. Skip this array. | |
989 | */ | |
990 | if (!domlist) | |
991 | continue; | |
0fa21e85 | 992 | for (from=statelist ; from ; from=from->next) { |
0f0749ad | 993 | dev_t devid; |
5ec0f373 | 994 | if (!check_donor(from, to)) |
0fa21e85 | 995 | continue; |
5739e0d0 N |
996 | if (from->metadata->ss->external) |
997 | devid = container_choose_spare( | |
fbfdcb06 | 998 | from, to, domlist, &sc, 0); |
5739e0d0 | 999 | else |
f0b85306 | 1000 | devid = choose_spare(from, to, domlist, |
fbfdcb06 | 1001 | &sc); |
f566ef45 JS |
1002 | if (devid > 0 && |
1003 | move_spare(from->devname, to->devname, | |
1004 | devid)) { | |
1005 | alert("MoveSpare", to->devname, | |
1006 | from->devname, info); | |
d52bb542 AC |
1007 | break; |
1008 | } | |
0fa21e85 | 1009 | } |
e78dda3b | 1010 | domain_free(domlist); |
2e0172b1 N |
1011 | } |
1012 | } | |
c3621c0a ML |
1013 | |
1014 | /* search the statelist to connect external | |
1015 | * metadata subarrays with their containers | |
1016 | * We always completely rebuild the tree from scratch as | |
1017 | * that is safest considering the possibility of entries | |
1018 | * disappearing or changing. | |
1019 | */ | |
1020 | static void link_containers_with_subarrays(struct state *list) | |
1021 | { | |
1022 | struct state *st; | |
1023 | struct state *cont; | |
1024 | for (st = list; st; st = st->next) { | |
1025 | st->parent = NULL; | |
1026 | st->subarray = NULL; | |
1027 | } | |
1028 | for (st = list; st; st = st->next) | |
4dd2df09 | 1029 | if (st->parent_devnm[0]) |
c3621c0a | 1030 | for (cont = list; cont; cont = cont->next) |
f566ef45 | 1031 | if (!cont->err && cont->parent_devnm[0] == 0 && |
4dd2df09 | 1032 | strcmp(cont->devnm, st->parent_devnm) == 0) { |
c3621c0a ML |
1033 | st->parent = cont; |
1034 | st->subarray = cont->subarray; | |
1035 | cont->subarray = st; | |
1036 | break; | |
1037 | } | |
1038 | } | |
1039 | ||
b90c0e9a NB |
1040 | /* Not really Monitor but ... */ |
1041 | int Wait(char *dev) | |
1042 | { | |
4dd2df09 | 1043 | char devnm[32]; |
cb91230c TM |
1044 | dev_t rdev; |
1045 | char *tmp; | |
b90c0e9a | 1046 | int rv = 1; |
d3f6cf4f | 1047 | int frozen_remaining = 3; |
b90c0e9a | 1048 | |
cb91230c | 1049 | if (!stat_is_blkdev(dev, &rdev)) |
b90c0e9a | 1050 | return 2; |
cb91230c TM |
1051 | |
1052 | tmp = devid2devnm(rdev); | |
1053 | if (!tmp) { | |
1054 | pr_err("Cannot get md device name.\n"); | |
1055 | return 2; | |
1056 | } | |
1057 | ||
1058 | strcpy(devnm, tmp); | |
b90c0e9a NB |
1059 | |
1060 | while(1) { | |
1061 | struct mdstat_ent *ms = mdstat_read(1, 0); | |
1062 | struct mdstat_ent *e; | |
1063 | ||
f566ef45 | 1064 | for (e = ms; e; e = e->next) |
4dd2df09 | 1065 | if (strcmp(e->devnm, devnm) == 0) |
b90c0e9a NB |
1066 | break; |
1067 | ||
d3f6cf4f | 1068 | if (e && e->percent == RESYNC_NONE) { |
276be514 N |
1069 | /* We could be in the brief pause before something |
1070 | * starts. /proc/mdstat doesn't show that, but | |
1071 | * sync_action does. | |
1072 | */ | |
1073 | struct mdinfo mdi; | |
1074 | char buf[21]; | |
dae13137 JS |
1075 | |
1076 | if (sysfs_init(&mdi, -1, devnm)) | |
1077 | return 2; | |
276be514 N |
1078 | if (sysfs_get_str(&mdi, NULL, "sync_action", |
1079 | buf, 20) > 0 && | |
d3f6cf4f | 1080 | strcmp(buf,"idle\n") != 0) { |
276be514 | 1081 | e->percent = RESYNC_UNKNOWN; |
d3f6cf4f N |
1082 | if (strcmp(buf, "frozen\n") == 0) { |
1083 | if (frozen_remaining == 0) | |
1084 | e->percent = RESYNC_NONE; | |
1085 | else | |
1086 | frozen_remaining -= 1; | |
1087 | } | |
1088 | } | |
276be514 | 1089 | } |
639c3c10 | 1090 | if (!e || e->percent == RESYNC_NONE) { |
e7783ee6 | 1091 | if (e && e->metadata_version && |
c94709e8 DW |
1092 | strncmp(e->metadata_version, "external:", 9) == 0) { |
1093 | if (is_subarray(&e->metadata_version[9])) | |
1094 | ping_monitor(&e->metadata_version[9]); | |
1095 | else | |
4dd2df09 | 1096 | ping_monitor(devnm); |
c94709e8 | 1097 | } |
b90c0e9a NB |
1098 | free_mdstat(ms); |
1099 | return rv; | |
1100 | } | |
89a10d84 | 1101 | free_mdstat(ms); |
b90c0e9a NB |
1102 | rv = 0; |
1103 | mdstat_wait(5); | |
1104 | } | |
1105 | } | |
679eb882 | 1106 | |
43ebc910 GP |
1107 | /* The state "broken" is used only for RAID0/LINEAR - it's the same as |
1108 | * "clean", but used in case the array has one or more members missing. | |
1109 | */ | |
679eb882 | 1110 | static char *clean_states[] = { |
43ebc910 | 1111 | "clear", "inactive", "readonly", "read-auto", "clean", "broken", NULL }; |
679eb882 | 1112 | |
1ea04629 | 1113 | int WaitClean(char *dev, int verbose) |
679eb882 N |
1114 | { |
1115 | int fd; | |
1116 | struct mdinfo *mdi; | |
1117 | int rv = 1; | |
4dd2df09 | 1118 | char devnm[32]; |
679eb882 | 1119 | |
9e04ac1c ZL |
1120 | if (!stat_is_blkdev(dev, NULL)) |
1121 | return 2; | |
679eb882 N |
1122 | fd = open(dev, O_RDONLY); |
1123 | if (fd < 0) { | |
1124 | if (verbose) | |
e7b84f9d | 1125 | pr_err("Couldn't open %s: %s\n", dev, strerror(errno)); |
679eb882 N |
1126 | return 1; |
1127 | } | |
1128 | ||
4dd2df09 N |
1129 | strcpy(devnm, fd2devnm(fd)); |
1130 | mdi = sysfs_read(fd, devnm, GET_VERSION|GET_LEVEL|GET_SAFEMODE); | |
679eb882 N |
1131 | if (!mdi) { |
1132 | if (verbose) | |
7a862a02 | 1133 | pr_err("Failed to read sysfs attributes for %s\n", dev); |
679eb882 N |
1134 | close(fd); |
1135 | return 0; | |
1136 | } | |
1137 | ||
1138 | switch(mdi->array.level) { | |
1139 | case LEVEL_LINEAR: | |
1140 | case LEVEL_MULTIPATH: | |
1141 | case 0: | |
1142 | /* safemode delay is irrelevant for these levels */ | |
1143 | rv = 0; | |
1144 | } | |
1145 | ||
1146 | /* for internal metadata the kernel handles the final clean | |
1147 | * transition, containers can never be dirty | |
1148 | */ | |
1149 | if (!is_subarray(mdi->text_version)) | |
1150 | rv = 0; | |
1151 | ||
1152 | /* safemode disabled ? */ | |
1153 | if (mdi->safe_mode_delay == 0) | |
1154 | rv = 0; | |
1155 | ||
1156 | if (rv) { | |
4dd2df09 | 1157 | int state_fd = sysfs_open(fd2devnm(fd), NULL, "array_state"); |
679eb882 | 1158 | char buf[20]; |
efc67e8e | 1159 | int delay = 5000; |
679eb882 N |
1160 | |
1161 | /* minimize the safe_mode_delay and prepare to wait up to 5s | |
1162 | * for writes to quiesce | |
1163 | */ | |
1164 | sysfs_set_safemode(mdi, 1); | |
679eb882 N |
1165 | |
1166 | /* wait for array_state to be clean */ | |
1167 | while (1) { | |
1168 | rv = read(state_fd, buf, sizeof(buf)); | |
1169 | if (rv < 0) | |
1170 | break; | |
43ebc910 GP |
1171 | if (sysfs_match_word(buf, clean_states) < |
1172 | (int)ARRAY_SIZE(clean_states) - 1) | |
679eb882 | 1173 | break; |
efc67e8e | 1174 | rv = sysfs_wait(state_fd, &delay); |
679eb882 N |
1175 | if (rv < 0 && errno != EINTR) |
1176 | break; | |
1177 | lseek(state_fd, 0, SEEK_SET); | |
1178 | } | |
1179 | if (rv < 0) | |
1180 | rv = 1; | |
1ea04629 | 1181 | else if (ping_monitor(mdi->text_version) == 0) { |
679eb882 N |
1182 | /* we need to ping to close the window between array |
1183 | * state transitioning to clean and the metadata being | |
1184 | * marked clean | |
1185 | */ | |
1186 | rv = 0; | |
1ea04629 | 1187 | } else { |
679eb882 | 1188 | rv = 1; |
1ea04629 MT |
1189 | pr_err("Error connecting monitor with %s\n", dev); |
1190 | } | |
679eb882 | 1191 | if (rv && verbose) |
f566ef45 | 1192 | pr_err("Error waiting for %s to be clean\n", dev); |
679eb882 N |
1193 | |
1194 | /* restore the original safe_mode_delay */ | |
1195 | sysfs_set_safemode(mdi, mdi->safe_mode_delay); | |
1196 | close(state_fd); | |
1197 | } | |
1198 | ||
1199 | sysfs_free(mdi); | |
1200 | close(fd); | |
1201 | ||
1202 | return rv; | |
1203 | } |