]>
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); | |
e0bd6a96 | 66 | static void alert(char *event, char *dev, char *disc, struct alert_info *info); |
a90e1050 | 67 | static int check_array(struct state *st, struct mdstat_ent *mdstat, |
e0bd6a96 | 68 | int test, struct alert_info *info, |
c2ecf5f6 | 69 | int increments, char *prefer); |
83f3bc5f | 70 | static int add_new_arrays(struct mdstat_ent *mdstat, struct state **statelist, |
e0bd6a96 N |
71 | int test, struct alert_info *info); |
72 | static void try_spare_migration(struct state *statelist, struct alert_info *info); | |
c3621c0a | 73 | static void link_containers_with_subarrays(struct state *list); |
2e0172b1 | 74 | |
a655e550 | 75 | int Monitor(struct mddev_dev *devlist, |
52826846 | 76 | char *mailaddr, char *alert_cmd, |
95c50205 N |
77 | struct context *c, |
78 | int daemonise, int oneshot, | |
79 | int dosyslog, char *pidfile, int increments, | |
80 | int share) | |
52826846 NB |
81 | { |
82 | /* | |
83 | * Every few seconds, scan every md device looking for changes | |
84 | * When a change is found, log it, possibly run the alert command, | |
85 | * and possibly send Email | |
86 | * | |
87 | * For each array, we record: | |
88 | * Update time | |
89 | * active/working/failed/spare drives | |
90 | * State of each device. | |
e0d19036 | 91 | * %rebuilt if rebuilding |
52826846 NB |
92 | * |
93 | * If the update time changes, check out all the data again | |
94 | * It is possible that we cannot get the state of each device | |
95 | * due to bugs in the md kernel module. | |
e0d19036 NB |
96 | * We also read /proc/mdstat to get rebuild percent, |
97 | * and to get state on all active devices incase of kernel bug. | |
52826846 | 98 | * |
e0d19036 NB |
99 | * Events are: |
100 | * Fail | |
101 | * An active device had Faulty set or Active/Sync removed | |
102 | * FailSpare | |
103 | * A spare device had Faulty set | |
104 | * SpareActive | |
105 | * An active device had a reverse transition | |
106 | * RebuildStarted | |
107 | * percent went from -1 to +ve | |
9a36a9b7 ZB |
108 | * RebuildNN |
109 | * percent went from below to not-below NN% | |
e0d19036 NB |
110 | * DeviceDisappeared |
111 | * Couldn't access a device which was previously visible | |
52826846 NB |
112 | * |
113 | * if we detect an array with active<raid and spare==0 | |
114 | * we look at other arrays that have same spare-group | |
115 | * If we find one with active==raid and spare>0, | |
116 | * and if we can get_disk_info and find a name | |
117 | * Then we hot-remove and hot-add to the other array | |
118 | * | |
e0d19036 NB |
119 | * If devlist is NULL, then we can monitor everything because --scan |
120 | * was given. We get an initial list from config file and add anything | |
121 | * that appears in /proc/mdstat | |
52826846 NB |
122 | */ |
123 | ||
2e0172b1 | 124 | struct state *statelist = NULL; |
b657208c | 125 | struct state *st2; |
52826846 | 126 | int finished = 0; |
e0d19036 | 127 | struct mdstat_ent *mdstat = NULL; |
72362f18 | 128 | char *mailfrom; |
e0bd6a96 | 129 | struct alert_info info; |
9f3dd454 | 130 | struct mddev_ident *mdlist; |
cab9c67d | 131 | int delay_for_event = c->delay; |
e0d19036 | 132 | |
d013a55e | 133 | if (!mailaddr) { |
8aec876d | 134 | mailaddr = conf_get_mailaddr(); |
95c50205 | 135 | if (mailaddr && ! c->scan) |
e7b84f9d | 136 | pr_err("Monitor using email address \"%s\" from config file\n", |
d013a55e NB |
137 | mailaddr); |
138 | } | |
8aec876d | 139 | mailfrom = conf_get_mailfrom(); |
4948b8f7 | 140 | |
d013a55e | 141 | if (!alert_cmd) { |
8aec876d | 142 | alert_cmd = conf_get_program(); |
f566ef45 | 143 | if (alert_cmd && !c->scan) |
e7b84f9d | 144 | pr_err("Monitor using program \"%s\" from config file\n", |
d013a55e NB |
145 | alert_cmd); |
146 | } | |
95c50205 | 147 | if (c->scan && !mailaddr && !alert_cmd && !dosyslog) { |
e7b84f9d | 148 | pr_err("No mail address or alert command - not monitoring.\n"); |
d013a55e | 149 | return 1; |
56eedc1a | 150 | } |
e0bd6a96 N |
151 | info.alert_cmd = alert_cmd; |
152 | info.mailaddr = mailaddr; | |
153 | info.mailfrom = mailfrom; | |
154 | info.dosyslog = dosyslog; | |
d013a55e | 155 | |
44d337f0 AC |
156 | if (daemonise) { |
157 | int rv = make_daemon(pidfile); | |
158 | if (rv >= 0) | |
159 | return rv; | |
160 | } | |
e0d19036 | 161 | |
1011e834 | 162 | if (share) |
95c50205 | 163 | if (check_one_sharer(c->scan)) |
2e0172b1 | 164 | return 1; |
edde9560 | 165 | |
e0d19036 | 166 | if (devlist == NULL) { |
9f3dd454 | 167 | mdlist = conf_get_ident(NULL); |
f566ef45 | 168 | for (; mdlist; mdlist = mdlist->next) { |
fe056d1f | 169 | struct state *st; |
f566ef45 | 170 | |
fe056d1f N |
171 | if (mdlist->devname == NULL) |
172 | continue; | |
112cace6 N |
173 | if (strcasecmp(mdlist->devname, "<ignore>") == 0) |
174 | continue; | |
503975b9 | 175 | st = xcalloc(1, sizeof *st); |
db2d001c | 176 | if (mdlist->devname[0] == '/') |
503975b9 | 177 | st->devname = xstrdup(mdlist->devname); |
db2d001c | 178 | else { |
503975b9 | 179 | st->devname = xmalloc(8+strlen(mdlist->devname)+1); |
db2d001c N |
180 | strcpy(strcpy(st->devname, "/dev/md/"), |
181 | mdlist->devname); | |
182 | } | |
e0d19036 | 183 | st->next = statelist; |
4dd2df09 | 184 | st->devnm[0] = 0; |
9dad51d4 | 185 | st->percent = RESYNC_UNKNOWN; |
f1661bd7 | 186 | st->from_config = 1; |
feb716e9 | 187 | st->expected_spares = mdlist->spare_disks; |
e0d19036 | 188 | if (mdlist->spare_group) |
503975b9 | 189 | st->spare_group = xstrdup(mdlist->spare_group); |
e0d19036 NB |
190 | statelist = st; |
191 | } | |
192 | } else { | |
a655e550 | 193 | struct mddev_dev *dv; |
f566ef45 JS |
194 | |
195 | for (dv = devlist; dv; dv = dv->next) { | |
503975b9 | 196 | struct state *st = xcalloc(1, sizeof *st); |
9f3dd454 | 197 | mdlist = conf_get_ident(dv->devname); |
503975b9 | 198 | st->devname = xstrdup(dv->devname); |
e0d19036 | 199 | st->next = statelist; |
4dd2df09 | 200 | st->devnm[0] = 0; |
9dad51d4 | 201 | st->percent = RESYNC_UNKNOWN; |
feb716e9 | 202 | st->expected_spares = -1; |
e5329c37 NB |
203 | if (mdlist) { |
204 | st->expected_spares = mdlist->spare_disks; | |
205 | if (mdlist->spare_group) | |
503975b9 | 206 | st->spare_group = xstrdup(mdlist->spare_group); |
e5329c37 | 207 | } |
e0d19036 NB |
208 | statelist = st; |
209 | } | |
210 | } | |
211 | ||
f566ef45 | 212 | while (!finished) { |
aa88f531 | 213 | int new_found = 0; |
73ff0732 | 214 | struct state *st, **stp; |
a90e1050 | 215 | int anydegraded = 0; |
007087d0 | 216 | int anyredundant = 0; |
e0d19036 NB |
217 | |
218 | if (mdstat) | |
219 | free_mdstat(mdstat); | |
f566ef45 | 220 | mdstat = mdstat_read(oneshot ? 0 : 1, 0); |
e0d19036 | 221 | |
007087d0 | 222 | for (st = statelist; st; st = st->next) { |
95c50205 N |
223 | if (check_array(st, mdstat, c->test, &info, |
224 | increments, c->prefer)) | |
a90e1050 | 225 | anydegraded = 1; |
007087d0 MT |
226 | /* for external arrays, metadata is filled for |
227 | * containers only | |
228 | */ | |
229 | if (st->metadata && st->metadata->ss->external) | |
230 | continue; | |
231 | if (st->err == 0 && !anyredundant) | |
232 | anyredundant = 1; | |
233 | } | |
1011e834 | 234 | |
e0d19036 | 235 | /* now check if there are any new devices found in mdstat */ |
95c50205 N |
236 | if (c->scan) |
237 | new_found = add_new_arrays(mdstat, &statelist, c->test, | |
e0bd6a96 | 238 | &info); |
2e0172b1 | 239 | |
e0d19036 NB |
240 | /* If an array has active < raid && spare == 0 && spare_group != NULL |
241 | * Look for another array with spare > 0 and active == raid and same spare_group | |
242 | * if found, choose a device and hotremove/hotadd | |
243 | */ | |
a90e1050 | 244 | if (share && anydegraded) |
e0bd6a96 | 245 | try_spare_migration(statelist, &info); |
aa88f531 NB |
246 | if (!new_found) { |
247 | if (oneshot) | |
248 | break; | |
007087d0 MT |
249 | else if (!anyredundant) { |
250 | break; | |
251 | } | |
e2308733 | 252 | else { |
cab9c67d BK |
253 | int wait_result = mdstat_wait(delay_for_event); |
254 | ||
255 | /* | |
256 | * If mdmonitor is awaken by event, set small delay once | |
257 | * to deal with udev and mdadm. | |
258 | */ | |
259 | if (wait_result != 0) { | |
260 | if (c->delay > 5) | |
261 | delay_for_event = 5; | |
262 | } else | |
263 | delay_for_event = c->delay; | |
264 | ||
e2308733 MT |
265 | mdstat_close(); |
266 | } | |
aa88f531 | 267 | } |
95c50205 | 268 | c->test = 0; |
73ff0732 N |
269 | |
270 | for (stp = &statelist; (st = *stp) != NULL; ) { | |
271 | if (st->from_auto && st->err > 5) { | |
272 | *stp = st->next; | |
273 | free(st->devname); | |
274 | free(st->spare_group); | |
275 | free(st); | |
276 | } else | |
277 | stp = &st->next; | |
278 | } | |
52826846 | 279 | } |
b657208c JS |
280 | for (st2 = statelist; st2; st2 = statelist) { |
281 | statelist = st2->next; | |
282 | free(st2); | |
283 | } | |
284 | ||
b5e64645 NB |
285 | if (pidfile) |
286 | unlink(pidfile); | |
52826846 NB |
287 | return 0; |
288 | } | |
289 | ||
2e0172b1 N |
290 | static int make_daemon(char *pidfile) |
291 | { | |
44d337f0 AC |
292 | /* Return: |
293 | * -1 in the forked daemon | |
294 | * 0 in the parent | |
295 | * 1 on error | |
296 | * so a none-negative becomes the exit code. | |
297 | */ | |
2e0172b1 N |
298 | int pid = fork(); |
299 | if (pid > 0) { | |
300 | if (!pidfile) | |
301 | printf("%d\n", pid); | |
302 | else { | |
303 | FILE *pid_file; | |
304 | pid_file=fopen(pidfile, "w"); | |
305 | if (!pid_file) | |
306 | perror("cannot create pid file"); | |
307 | else { | |
308 | fprintf(pid_file,"%d\n", pid); | |
309 | fclose(pid_file); | |
310 | } | |
311 | } | |
312 | return 0; | |
313 | } | |
314 | if (pid < 0) { | |
315 | perror("daemonise"); | |
316 | return 1; | |
317 | } | |
318 | close(0); | |
319 | open("/dev/null", O_RDWR); | |
f566ef45 JS |
320 | dup2(0, 1); |
321 | dup2(0, 2); | |
2e0172b1 | 322 | setsid(); |
44d337f0 | 323 | return -1; |
2e0172b1 N |
324 | } |
325 | ||
326 | static int check_one_sharer(int scan) | |
327 | { | |
185ec439 CL |
328 | int pid; |
329 | FILE *comm_fp; | |
2e0172b1 | 330 | FILE *fp; |
185ec439 | 331 | char comm_path[100]; |
0011874f | 332 | char path[100]; |
185ec439 CL |
333 | char comm[20]; |
334 | ||
0011874f JS |
335 | sprintf(path, "%s/autorebuild.pid", MDMON_DIR); |
336 | fp = fopen(path, "r"); | |
2e0172b1 | 337 | if (fp) { |
71204a50 N |
338 | if (fscanf(fp, "%d", &pid) != 1) |
339 | pid = -1; | |
185ec439 CL |
340 | snprintf(comm_path, sizeof(comm_path), |
341 | "/proc/%d/comm", pid); | |
342 | comm_fp = fopen(comm_path, "r"); | |
343 | if (comm_fp) { | |
344 | if (fscanf(comm_fp, "%s", comm) && | |
345 | strncmp(basename(comm), Name, strlen(Name)) == 0) { | |
346 | if (scan) { | |
347 | pr_err("Only one autorebuild process allowed in scan mode, aborting\n"); | |
348 | fclose(comm_fp); | |
349 | fclose(fp); | |
350 | return 1; | |
351 | } else { | |
352 | pr_err("Warning: One autorebuild process already running.\n"); | |
353 | } | |
2e0172b1 | 354 | } |
185ec439 | 355 | fclose(comm_fp); |
2e0172b1 N |
356 | } |
357 | fclose(fp); | |
358 | } | |
359 | if (scan) { | |
f566ef45 | 360 | if (mkdir(MDMON_DIR, S_IRWXU) < 0 && errno != EEXIST) { |
7a862a02 | 361 | pr_err("Can't create autorebuild.pid file\n"); |
a1bb2065 | 362 | } else { |
0011874f | 363 | fp = fopen(path, "w"); |
a1bb2065 | 364 | if (!fp) |
7a862a02 | 365 | pr_err("Cannot create autorebuild.pidfile\n"); |
a1bb2065 AC |
366 | else { |
367 | pid = getpid(); | |
368 | fprintf(fp, "%d\n", pid); | |
369 | fclose(fp); | |
370 | } | |
2e0172b1 N |
371 | } |
372 | } | |
373 | return 0; | |
374 | } | |
52826846 | 375 | |
e0bd6a96 | 376 | static void alert(char *event, char *dev, char *disc, struct alert_info *info) |
52826846 | 377 | { |
773135f5 NB |
378 | int priority; |
379 | ||
b1bac75b | 380 | if (!info->alert_cmd && !info->mailaddr && !info->dosyslog) { |
cd29a5c8 | 381 | time_t now = time(0); |
aba69144 | 382 | |
f566ef45 JS |
383 | printf("%1.15s: %s on %s %s\n", ctime(&now) + 4, |
384 | event, dev, disc?disc:"unknown device"); | |
cd29a5c8 | 385 | } |
e0bd6a96 | 386 | if (info->alert_cmd) { |
52826846 NB |
387 | int pid = fork(); |
388 | switch(pid) { | |
389 | default: | |
390 | waitpid(pid, NULL, 0); | |
391 | break; | |
392 | case -1: | |
393 | break; | |
394 | case 0: | |
e0bd6a96 N |
395 | execl(info->alert_cmd, info->alert_cmd, |
396 | event, dev, disc, NULL); | |
52826846 NB |
397 | exit(2); |
398 | } | |
399 | } | |
f566ef45 JS |
400 | if (info->mailaddr && (strncmp(event, "Fail", 4) == 0 || |
401 | strncmp(event, "Test", 4) == 0 || | |
402 | strncmp(event, "Spares", 6) == 0 || | |
403 | strncmp(event, "Degrade", 7) == 0)) { | |
52826846 NB |
404 | FILE *mp = popen(Sendmail, "w"); |
405 | if (mp) { | |
008e1100 | 406 | FILE *mdstat; |
52826846 NB |
407 | char hname[256]; |
408 | gethostname(hname, sizeof(hname)); | |
409 | signal(SIGPIPE, SIG_IGN); | |
e0bd6a96 N |
410 | if (info->mailfrom) |
411 | fprintf(mp, "From: %s\n", info->mailfrom); | |
4948b8f7 | 412 | else |
f566ef45 JS |
413 | fprintf(mp, "From: %s monitoring <root>\n", |
414 | Name); | |
e0bd6a96 N |
415 | fprintf(mp, "To: %s\n", info->mailaddr); |
416 | fprintf(mp, "Subject: %s event on %s:%s\n\n", | |
417 | event, dev, hname); | |
52826846 | 418 | |
e0bd6a96 | 419 | fprintf(mp, |
7a862a02 | 420 | "This is an automatically generated mail message from %s\n", Name); |
52826846 NB |
421 | fprintf(mp, "running on %s\n\n", hname); |
422 | ||
e0bd6a96 | 423 | fprintf(mp, |
7a862a02 | 424 | "A %s event had been detected on md device %s.\n\n", event, dev); |
52826846 | 425 | |
37dfc3d6 | 426 | if (disc && disc[0] != ' ') |
e0bd6a96 | 427 | fprintf(mp, |
7a862a02 | 428 | "It could be related to component device %s.\n\n", disc); |
37dfc3d6 NB |
429 | if (disc && disc[0] == ' ') |
430 | fprintf(mp, "Extra information:%s.\n\n", disc); | |
52826846 NB |
431 | |
432 | fprintf(mp, "Faithfully yours, etc.\n"); | |
008e1100 NB |
433 | |
434 | mdstat = fopen("/proc/mdstat", "r"); | |
435 | if (mdstat) { | |
436 | char buf[8192]; | |
437 | int n; | |
e0bd6a96 | 438 | fprintf(mp, |
7a862a02 | 439 | "\nP.S. The /proc/mdstat file currently contains the following:\n\n"); |
f566ef45 JS |
440 | while ((n = fread(buf, 1, sizeof(buf), |
441 | mdstat)) > 0) | |
442 | n = fwrite(buf, 1, n, mp); | |
008e1100 NB |
443 | fclose(mdstat); |
444 | } | |
6278fb3a | 445 | pclose(mp); |
52826846 | 446 | } |
52826846 | 447 | } |
773135f5 NB |
448 | |
449 | /* log the event to syslog maybe */ | |
e0bd6a96 | 450 | if (info->dosyslog) { |
773135f5 NB |
451 | /* Log at a different severity depending on the event. |
452 | * | |
453 | * These are the critical events: */ | |
f566ef45 JS |
454 | if (strncmp(event, "Fail", 4) == 0 || |
455 | strncmp(event, "Degrade", 7) == 0 || | |
456 | strncmp(event, "DeviceDisappeared", 17) == 0) | |
773135f5 NB |
457 | priority = LOG_CRIT; |
458 | /* Good to know about, but are not failures: */ | |
f566ef45 JS |
459 | else if (strncmp(event, "Rebuild", 7) == 0 || |
460 | strncmp(event, "MoveSpare", 9) == 0 || | |
d1732eeb | 461 | strncmp(event, "Spares", 6) != 0) |
773135f5 NB |
462 | priority = LOG_WARNING; |
463 | /* Everything else: */ | |
464 | else | |
465 | priority = LOG_INFO; | |
466 | ||
b033913a | 467 | if (disc && disc[0] != ' ') |
e0bd6a96 | 468 | syslog(priority, |
7a862a02 | 469 | "%s event detected on md device %s, component device %s", event, dev, disc); |
b033913a N |
470 | else if (disc) |
471 | syslog(priority, | |
472 | "%s event detected on md device %s: %s", | |
473 | event, dev, disc); | |
773135f5 | 474 | else |
e0bd6a96 N |
475 | syslog(priority, |
476 | "%s event detected on md device %s", | |
477 | event, dev); | |
773135f5 | 478 | } |
52826846 | 479 | } |
b90c0e9a | 480 | |
a90e1050 | 481 | static int check_array(struct state *st, struct mdstat_ent *mdstat, |
e0bd6a96 | 482 | int test, struct alert_info *ainfo, |
c2ecf5f6 | 483 | int increments, char *prefer) |
2e0172b1 | 484 | { |
ff044d6b AC |
485 | /* Update the state 'st' to reflect any changes shown in mdstat, |
486 | * or found by directly examining the array, and return | |
487 | * '1' if the array is degraded, or '0' if it is optimal (or dead). | |
488 | */ | |
b0599bda | 489 | struct { int state, major, minor; } info[MAX_DISKS]; |
aed5f5c3 | 490 | struct mdinfo *sra = NULL; |
2e0172b1 N |
491 | mdu_array_info_t array; |
492 | struct mdstat_ent *mse = NULL, *mse2; | |
493 | char *dev = st->devname; | |
b9a0309c | 494 | int fd; |
2e0172b1 | 495 | int i; |
9e6d9291 N |
496 | int remaining_disks; |
497 | int last_disk; | |
721b662b | 498 | int new_array = 0; |
1830e74b | 499 | int retval; |
802961a2 | 500 | int is_container = 0; |
b3ab4e4d | 501 | unsigned long redundancy_only_flags = 0; |
2e0172b1 N |
502 | |
503 | if (test) | |
e0bd6a96 | 504 | alert("TestMessage", dev, NULL, ainfo); |
b9a0309c | 505 | |
1830e74b JS |
506 | retval = 0; |
507 | ||
2e0172b1 | 508 | fd = open(dev, O_RDONLY); |
13e5d845 JS |
509 | if (fd < 0) |
510 | goto disappeared; | |
b9a0309c | 511 | |
802961a2 MT |
512 | if (st->devnm[0] == 0) |
513 | strcpy(st->devnm, fd2devnm(fd)); | |
514 | ||
515 | for (mse2 = mdstat; mse2; mse2 = mse2->next) | |
516 | if (strcmp(mse2->devnm, st->devnm) == 0) { | |
517 | mse2->devnm[0] = 0; /* flag it as "used" */ | |
518 | mse = mse2; | |
519 | } | |
520 | ||
521 | if (!mse) { | |
522 | /* duplicated array in statelist | |
523 | * or re-created after reading mdstat | |
524 | */ | |
525 | st->err++; | |
526 | goto out; | |
527 | } | |
528 | ||
529 | if (mse->level == NULL) | |
530 | is_container = 1; | |
531 | ||
2dab69c9 | 532 | if (!is_container && !md_array_active(fd)) |
13e5d845 | 533 | goto disappeared; |
b9a0309c | 534 | |
2e0172b1 | 535 | fcntl(fd, F_SETFD, FD_CLOEXEC); |
13e5d845 JS |
536 | if (md_get_array_info(fd, &array) < 0) |
537 | goto disappeared; | |
538 | ||
b3ab4e4d MT |
539 | if (!is_container && map_name(pers, mse->level) > 0) |
540 | redundancy_only_flags |= GET_MISMATCH; | |
802961a2 MT |
541 | |
542 | sra = sysfs_read(-1, st->devnm, GET_LEVEL | GET_DISKS | GET_DEVS | | |
b3ab4e4d | 543 | GET_STATE | redundancy_only_flags); |
aed5f5c3 | 544 | |
aed5f5c3 JS |
545 | if (!sra) |
546 | goto disappeared; | |
547 | ||
2e0172b1 N |
548 | /* It's much easier to list what array levels can't |
549 | * have a device disappear than all of them that can | |
550 | */ | |
48bc2ade | 551 | if (sra->array.level == 0 || sra->array.level == -1) { |
f1661bd7 | 552 | if (!st->err && !st->from_config) |
04e27c20 | 553 | alert("DeviceDisappeared", dev, " Wrong-Level", ainfo); |
73ff0732 | 554 | st->err++; |
1830e74b | 555 | goto out; |
2e0172b1 | 556 | } |
2e0172b1 | 557 | |
2e0172b1 N |
558 | /* this array is in /proc/mdstat */ |
559 | if (array.utime == 0) | |
560 | /* external arrays don't update utime, so | |
561 | * just make sure it is always different. */ | |
562 | array.utime = st->utime + 1;; | |
563 | ||
0f760384 | 564 | if (st->err) { |
73ff0732 | 565 | /* New array appeared where previously had an error */ |
0f760384 N |
566 | st->err = 0; |
567 | st->percent = RESYNC_NONE; | |
721b662b | 568 | new_array = 1; |
007087d0 MT |
569 | if (!is_container) |
570 | alert("NewArray", st->devname, NULL, ainfo); | |
0f760384 N |
571 | } |
572 | ||
b8e5713c | 573 | if (st->utime == array.utime && st->failed == sra->array.failed_disks && |
e5eb6857 | 574 | st->working == sra->array.working_disks && |
b98943a4 | 575 | st->spare == sra->array.spare_disks && |
d7be7d87 | 576 | (mse == NULL || (mse->percent == st->percent))) { |
ff044d6b | 577 | if ((st->active < st->raid) && st->spare == 0) |
1830e74b JS |
578 | retval = 1; |
579 | goto out; | |
2e0172b1 N |
580 | } |
581 | if (st->utime == 0 && /* new array */ | |
f27904a5 | 582 | mse->pattern && strchr(mse->pattern, '_') /* degraded */) |
e0bd6a96 | 583 | alert("DegradedArray", dev, NULL, ainfo); |
2e0172b1 | 584 | |
b98943a4 JS |
585 | if (st->utime == 0 && /* new array */ st->expected_spares > 0 && |
586 | sra->array.spare_disks < st->expected_spares) | |
e0bd6a96 | 587 | alert("SparesMissing", dev, NULL, ainfo); |
9dad51d4 | 588 | if (st->percent < 0 && st->percent != RESYNC_UNKNOWN && |
2e0172b1 | 589 | mse->percent >= 0) |
e0bd6a96 | 590 | alert("RebuildStarted", dev, NULL, ainfo); |
f27904a5 | 591 | if (st->percent >= 0 && mse->percent >= 0 && |
2e0172b1 | 592 | (mse->percent / increments) > (st->percent / increments)) { |
97589839 | 593 | char percentalert[18]; |
f27904a5 JS |
594 | /* |
595 | * "RebuildNN" (10 chars) or "RebuildStarted" (15 chars) | |
596 | */ | |
2e0172b1 N |
597 | |
598 | if((mse->percent / increments) == 0) | |
f27904a5 JS |
599 | snprintf(percentalert, sizeof(percentalert), |
600 | "RebuildStarted"); | |
2e0172b1 | 601 | else |
f27904a5 JS |
602 | snprintf(percentalert, sizeof(percentalert), |
603 | "Rebuild%02d", mse->percent); | |
2e0172b1 | 604 | |
e0bd6a96 | 605 | alert(percentalert, dev, NULL, ainfo); |
2e0172b1 N |
606 | } |
607 | ||
f27904a5 | 608 | if (mse->percent == RESYNC_NONE && st->percent >= 0) { |
2e0172b1 N |
609 | /* Rebuild/sync/whatever just finished. |
610 | * If there is a number in /mismatch_cnt, | |
611 | * we should report that. | |
612 | */ | |
2e0172b1 | 613 | if (sra && sra->mismatch_cnt > 0) { |
8453f8d0 LD |
614 | char cnt[80]; |
615 | snprintf(cnt, sizeof(cnt), | |
616 | " mismatches found: %d (on raid level %d)", | |
f566ef45 | 617 | sra->mismatch_cnt, sra->array.level); |
e0bd6a96 | 618 | alert("RebuildFinished", dev, cnt, ainfo); |
2e0172b1 | 619 | } else |
e0bd6a96 | 620 | alert("RebuildFinished", dev, NULL, ainfo); |
2e0172b1 N |
621 | } |
622 | st->percent = mse->percent; | |
623 | ||
b98943a4 | 624 | remaining_disks = sra->array.nr_disks; |
f27904a5 | 625 | for (i = 0; i < MAX_DISKS && remaining_disks > 0; i++) { |
2e0172b1 N |
626 | mdu_disk_info_t disc; |
627 | disc.number = i; | |
d97572f5 | 628 | if (md_get_disk_info(fd, &disc) >= 0) { |
2e0172b1 N |
629 | info[i].state = disc.state; |
630 | info[i].major = disc.major; | |
631 | info[i].minor = disc.minor; | |
9e6d9291 N |
632 | if (disc.major || disc.minor) |
633 | remaining_disks --; | |
2e0172b1 N |
634 | } else |
635 | info[i].major = info[i].minor = 0; | |
636 | } | |
9e6d9291 | 637 | last_disk = i; |
2e0172b1 | 638 | |
4019ad07 JL |
639 | if (mse->metadata_version && |
640 | strncmp(mse->metadata_version, "external:", 9) == 0 && | |
4dd2df09 N |
641 | is_subarray(mse->metadata_version+9)) { |
642 | char *sl; | |
f566ef45 | 643 | strcpy(st->parent_devnm, mse->metadata_version + 10); |
4dd2df09 N |
644 | sl = strchr(st->parent_devnm, '/'); |
645 | if (sl) | |
646 | *sl = 0; | |
647 | } else | |
648 | st->parent_devnm[0] = 0; | |
f27904a5 | 649 | if (st->metadata == NULL && st->parent_devnm[0] == 0) |
2e0172b1 N |
650 | st->metadata = super_by_fd(fd, NULL); |
651 | ||
f566ef45 JS |
652 | for (i = 0; i < MAX_DISKS; i++) { |
653 | mdu_disk_info_t disc = {0, 0, 0, 0, 0}; | |
654 | int newstate = 0; | |
2e0172b1 N |
655 | int change; |
656 | char *dv = NULL; | |
657 | disc.number = i; | |
f27904a5 | 658 | if (i < last_disk && (info[i].major || info[i].minor)) { |
2e0172b1 | 659 | newstate = info[i].state; |
f27904a5 JS |
660 | dv = map_dev_preferred(info[i].major, info[i].minor, 1, |
661 | prefer); | |
2e0172b1 N |
662 | disc.state = newstate; |
663 | disc.major = info[i].major; | |
664 | disc.minor = info[i].minor; | |
721b662b N |
665 | } else |
666 | newstate = (1 << MD_DISK_REMOVED); | |
667 | ||
2e0172b1 | 668 | if (dv == NULL && st->devid[i]) |
f27904a5 JS |
669 | dv = map_dev_preferred(major(st->devid[i]), |
670 | minor(st->devid[i]), 1, prefer); | |
2e0172b1 | 671 | change = newstate ^ st->devstate[i]; |
721b662b | 672 | if (st->utime && change && !st->err && !new_array) { |
f27904a5 | 673 | if ((st->devstate[i]&change) & (1 << MD_DISK_SYNC)) |
e0bd6a96 | 674 | alert("Fail", dev, dv, ainfo); |
f27904a5 | 675 | else if ((newstate & (1 << MD_DISK_FAULTY)) && |
2e0172b1 | 676 | (disc.major || disc.minor) && |
f27904a5 JS |
677 | st->devid[i] == makedev(disc.major, |
678 | disc.minor)) | |
e0bd6a96 | 679 | alert("FailSpare", dev, dv, ainfo); |
f27904a5 | 680 | else if ((newstate&change) & (1 << MD_DISK_SYNC)) |
e0bd6a96 | 681 | alert("SpareActive", dev, dv, ainfo); |
2e0172b1 N |
682 | } |
683 | st->devstate[i] = newstate; | |
684 | st->devid[i] = makedev(disc.major, disc.minor); | |
685 | } | |
b98943a4 | 686 | st->active = sra->array.active_disks; |
e5eb6857 | 687 | st->working = sra->array.working_disks; |
b98943a4 | 688 | st->spare = sra->array.spare_disks; |
b8e5713c | 689 | st->failed = sra->array.failed_disks; |
2e0172b1 | 690 | st->utime = array.utime; |
12a9d21f | 691 | st->raid = sra->array.raid_disks; |
2e0172b1 | 692 | st->err = 0; |
a90e1050 | 693 | if ((st->active < st->raid) && st->spare == 0) |
1830e74b JS |
694 | retval = 1; |
695 | ||
696 | out: | |
aed5f5c3 JS |
697 | if (sra) |
698 | sysfs_free(sra); | |
802961a2 | 699 | if (fd >= 0) |
13e5d845 | 700 | close(fd); |
1830e74b | 701 | return retval; |
13e5d845 JS |
702 | |
703 | disappeared: | |
007087d0 | 704 | if (!st->err && !is_container) |
13e5d845 JS |
705 | alert("DeviceDisappeared", dev, NULL, ainfo); |
706 | st->err++; | |
707 | goto out; | |
2e0172b1 N |
708 | } |
709 | ||
83f3bc5f | 710 | static int add_new_arrays(struct mdstat_ent *mdstat, struct state **statelist, |
e0bd6a96 | 711 | int test, struct alert_info *info) |
2e0172b1 N |
712 | { |
713 | struct mdstat_ent *mse; | |
714 | int new_found = 0; | |
1e08717f | 715 | char *name; |
2e0172b1 | 716 | |
f566ef45 | 717 | for (mse = mdstat; mse; mse = mse->next) |
d7be7d87 | 718 | if (mse->devnm[0] && (!mse->level || /* retrieve containers */ |
f566ef45 JS |
719 | (strcmp(mse->level, "raid0") != 0 && |
720 | strcmp(mse->level, "linear") != 0))) { | |
503975b9 | 721 | struct state *st = xcalloc(1, sizeof *st); |
2e0172b1 N |
722 | mdu_array_info_t array; |
723 | int fd; | |
1e08717f SV |
724 | |
725 | name = get_md_name(mse->devnm); | |
726 | if (!name) { | |
727 | free(st); | |
728 | continue; | |
729 | } | |
730 | ||
731 | st->devname = xstrdup(name); | |
2e0172b1 | 732 | if ((fd = open(st->devname, O_RDONLY)) < 0 || |
9cd39f01 | 733 | md_get_array_info(fd, &array) < 0) { |
2e0172b1 | 734 | /* no such array */ |
9cd39f01 JS |
735 | if (fd >= 0) |
736 | close(fd); | |
2e0172b1 N |
737 | put_md_name(st->devname); |
738 | free(st->devname); | |
739 | if (st->metadata) { | |
740 | st->metadata->ss->free_super(st->metadata); | |
741 | free(st->metadata); | |
742 | } | |
743 | free(st); | |
744 | continue; | |
745 | } | |
746 | close(fd); | |
83f3bc5f | 747 | st->next = *statelist; |
2e0172b1 | 748 | st->err = 1; |
73ff0732 | 749 | st->from_auto = 1; |
4dd2df09 | 750 | strcpy(st->devnm, mse->devnm); |
9dad51d4 | 751 | st->percent = RESYNC_UNKNOWN; |
2e0172b1 | 752 | st->expected_spares = -1; |
eb28e119 | 753 | if (mse->metadata_version && |
f566ef45 JS |
754 | strncmp(mse->metadata_version, |
755 | "external:", 9) == 0 && | |
4dd2df09 N |
756 | is_subarray(mse->metadata_version+9)) { |
757 | char *sl; | |
758 | strcpy(st->parent_devnm, | |
759 | mse->metadata_version+10); | |
760 | sl = strchr(st->parent_devnm, '/'); | |
761 | *sl = 0; | |
762 | } else | |
763 | st->parent_devnm[0] = 0; | |
83f3bc5f | 764 | *statelist = st; |
2e0172b1 | 765 | if (test) |
e0bd6a96 | 766 | alert("TestMessage", st->devname, NULL, info); |
2e0172b1 N |
767 | new_found = 1; |
768 | } | |
769 | return new_found; | |
770 | } | |
771 | ||
fbfdcb06 AO |
772 | static int get_required_spare_criteria(struct state *st, |
773 | struct spare_criteria *sc) | |
80e7f8c3 AC |
774 | { |
775 | int fd; | |
80e7f8c3 | 776 | |
f566ef45 | 777 | if (!st->metadata || !st->metadata->ss->get_spare_criteria) { |
fbfdcb06 | 778 | sc->min_size = 0; |
4b57ecf6 | 779 | sc->sector_size = 0; |
de697acc | 780 | return 0; |
300f5033 | 781 | } |
80e7f8c3 AC |
782 | |
783 | fd = open(st->devname, O_RDONLY); | |
784 | if (fd < 0) | |
de697acc AC |
785 | return 1; |
786 | if (st->metadata->ss->external) | |
787 | st->metadata->ss->load_container(st->metadata, fd, st->devname); | |
788 | else | |
789 | st->metadata->ss->load_super(st->metadata, fd, st->devname); | |
80e7f8c3 | 790 | close(fd); |
de697acc AC |
791 | if (!st->metadata->sb) |
792 | return 1; | |
fbfdcb06 AO |
793 | |
794 | st->metadata->ss->get_spare_criteria(st->metadata, sc); | |
80e7f8c3 AC |
795 | st->metadata->ss->free_super(st->metadata); |
796 | ||
de697acc | 797 | return 0; |
80e7f8c3 AC |
798 | } |
799 | ||
5ec0f373 | 800 | static int check_donor(struct state *from, struct state *to) |
2feb22ef | 801 | { |
66f5c4b6 N |
802 | struct state *sub; |
803 | ||
2feb22ef N |
804 | if (from == to) |
805 | return 0; | |
66f5c4b6 N |
806 | if (from->parent) |
807 | /* Cannot move from a member */ | |
2feb22ef | 808 | return 0; |
ff044d6b AC |
809 | if (from->err) |
810 | return 0; | |
66f5c4b6 N |
811 | for (sub = from->subarray; sub; sub = sub->subarray) |
812 | /* If source array has degraded subarrays, don't | |
813 | * remove anything | |
814 | */ | |
815 | if (sub->active < sub->raid) | |
816 | return 0; | |
817 | if (from->metadata->ss->external == 0) | |
818 | if (from->active < from->raid) | |
819 | return 0; | |
2feb22ef N |
820 | if (from->spare <= 0) |
821 | return 0; | |
e78dda3b | 822 | return 1; |
2feb22ef N |
823 | } |
824 | ||
0f0749ad | 825 | static dev_t choose_spare(struct state *from, struct state *to, |
f566ef45 | 826 | struct domainlist *domlist, struct spare_criteria *sc) |
0fa21e85 N |
827 | { |
828 | int d; | |
0f0749ad | 829 | dev_t dev = 0; |
0fa21e85 | 830 | |
b0599bda | 831 | for (d = from->raid; !dev && d < MAX_DISKS; d++) { |
f566ef45 | 832 | if (from->devid[d] > 0 && from->devstate[d] == 0) { |
0fa21e85 N |
833 | struct dev_policy *pol; |
834 | unsigned long long dev_size; | |
4b57ecf6 | 835 | unsigned int dev_sector_size; |
0fa21e85 | 836 | |
bfd76b93 CA |
837 | if (to->metadata->ss->external && |
838 | test_partition_from_id(from->devid[d])) | |
839 | continue; | |
840 | ||
fbfdcb06 | 841 | if (sc->min_size && |
0fa21e85 | 842 | dev_size_from_id(from->devid[d], &dev_size) && |
fbfdcb06 | 843 | dev_size < sc->min_size) |
0fa21e85 N |
844 | continue; |
845 | ||
4b57ecf6 AO |
846 | if (sc->sector_size && |
847 | dev_sector_size_from_id(from->devid[d], | |
848 | &dev_sector_size) && | |
849 | sc->sector_size != dev_sector_size) | |
850 | continue; | |
851 | ||
4dd2df09 | 852 | pol = devid_policy(from->devid[d]); |
0fa21e85 N |
853 | if (from->spare_group) |
854 | pol_add(&pol, pol_domain, | |
855 | from->spare_group, NULL); | |
f566ef45 JS |
856 | if (domain_test(domlist, pol, |
857 | to->metadata->ss->name) == 1) | |
0fa21e85 N |
858 | dev = from->devid[d]; |
859 | dev_policy_free(pol); | |
860 | } | |
861 | } | |
862 | return dev; | |
863 | } | |
864 | ||
0f0749ad | 865 | static dev_t container_choose_spare(struct state *from, struct state *to, |
f0b85306 | 866 | struct domainlist *domlist, |
fbfdcb06 | 867 | struct spare_criteria *sc, int active) |
5739e0d0 N |
868 | { |
869 | /* This is similar to choose_spare, but we cannot trust devstate, | |
870 | * so we need to read the metadata instead | |
871 | */ | |
326727d9 | 872 | struct mdinfo *list; |
5739e0d0 | 873 | struct supertype *st = from->metadata; |
ff044d6b | 874 | int fd = open(from->devname, O_RDONLY); |
5739e0d0 | 875 | int err; |
0f0749ad | 876 | dev_t dev = 0; |
5739e0d0 N |
877 | |
878 | if (fd < 0) | |
879 | return 0; | |
326727d9 AC |
880 | if (!st->ss->getinfo_super_disks) { |
881 | close(fd); | |
5739e0d0 | 882 | return 0; |
326727d9 | 883 | } |
1011e834 | 884 | |
5739e0d0 N |
885 | err = st->ss->load_container(st, fd, NULL); |
886 | close(fd); | |
887 | if (err) | |
888 | return 0; | |
1011e834 | 889 | |
a1e49d69 CA |
890 | if (from == to) { |
891 | /* We must check if number of active disks has not increased | |
892 | * since ioctl in main loop. mdmon may have added spare | |
893 | * to subarray. If so we do not need to look for more spares | |
894 | * so return non zero value */ | |
895 | int active_cnt = 0; | |
896 | struct mdinfo *dp; | |
897 | list = st->ss->getinfo_super_disks(st); | |
898 | if (!list) { | |
899 | st->ss->free_super(st); | |
900 | return 1; | |
901 | } | |
902 | dp = list->devs; | |
903 | while (dp) { | |
f566ef45 JS |
904 | if (dp->disk.state & (1 << MD_DISK_SYNC) && |
905 | !(dp->disk.state & (1 << MD_DISK_FAULTY))) | |
a1e49d69 CA |
906 | active_cnt++; |
907 | dp = dp->next; | |
908 | } | |
909 | sysfs_free(list); | |
910 | if (active < active_cnt) { | |
911 | /* Spare just activated.*/ | |
912 | st->ss->free_super(st); | |
913 | return 1; | |
914 | } | |
915 | } | |
916 | ||
326727d9 | 917 | /* We only need one spare so full list not needed */ |
fbfdcb06 | 918 | list = container_choose_spares(st, sc, domlist, from->spare_group, |
326727d9 AC |
919 | to->metadata->ss->name, 1); |
920 | if (list) { | |
921 | struct mdinfo *disks = list->devs; | |
922 | if (disks) | |
923 | dev = makedev(disks->disk.major, disks->disk.minor); | |
924 | sysfs_free(list); | |
5739e0d0 | 925 | } |
326727d9 | 926 | st->ss->free_super(st); |
5739e0d0 N |
927 | return dev; |
928 | } | |
929 | ||
e0bd6a96 | 930 | static void try_spare_migration(struct state *statelist, struct alert_info *info) |
2e0172b1 | 931 | { |
66f5c4b6 N |
932 | struct state *from; |
933 | struct state *st; | |
fbfdcb06 | 934 | struct spare_criteria sc; |
c3621c0a ML |
935 | |
936 | link_containers_with_subarrays(statelist); | |
66f5c4b6 | 937 | for (st = statelist; st; st = st->next) |
f566ef45 | 938 | if (st->active < st->raid && st->spare == 0 && !st->err) { |
e78dda3b N |
939 | struct domainlist *domlist = NULL; |
940 | int d; | |
66f5c4b6 N |
941 | struct state *to = st; |
942 | ||
4dd2df09 | 943 | if (to->parent_devnm[0] && !to->parent) |
c0dc0ad5 CA |
944 | /* subarray monitored without parent container |
945 | * we can't move spares here */ | |
946 | continue; | |
1011e834 | 947 | |
66f5c4b6 N |
948 | if (to->parent) |
949 | /* member of a container */ | |
950 | to = to->parent; | |
e78dda3b | 951 | |
fbfdcb06 | 952 | if (get_required_spare_criteria(to, &sc)) |
de697acc | 953 | continue; |
e9a2ac02 AC |
954 | if (to->metadata->ss->external) { |
955 | /* We must make sure there is | |
956 | * no suitable spare in container already. | |
957 | * If there is we don't add more */ | |
958 | dev_t devid = container_choose_spare( | |
fbfdcb06 | 959 | to, to, NULL, &sc, st->active); |
e9a2ac02 AC |
960 | if (devid > 0) |
961 | continue; | |
962 | } | |
b0599bda | 963 | for (d = 0; d < MAX_DISKS; d++) |
e78dda3b N |
964 | if (to->devid[d]) |
965 | domainlist_add_dev(&domlist, | |
966 | to->devid[d], | |
967 | to->metadata->ss->name); | |
968 | if (to->spare_group) | |
969 | domain_add(&domlist, to->spare_group); | |
5ec0f373 ML |
970 | /* |
971 | * No spare migration if the destination | |
972 | * has no domain. Skip this array. | |
973 | */ | |
974 | if (!domlist) | |
975 | continue; | |
0fa21e85 | 976 | for (from=statelist ; from ; from=from->next) { |
0f0749ad | 977 | dev_t devid; |
5ec0f373 | 978 | if (!check_donor(from, to)) |
0fa21e85 | 979 | continue; |
5739e0d0 N |
980 | if (from->metadata->ss->external) |
981 | devid = container_choose_spare( | |
fbfdcb06 | 982 | from, to, domlist, &sc, 0); |
5739e0d0 | 983 | else |
f0b85306 | 984 | devid = choose_spare(from, to, domlist, |
fbfdcb06 | 985 | &sc); |
f566ef45 JS |
986 | if (devid > 0 && |
987 | move_spare(from->devname, to->devname, | |
988 | devid)) { | |
989 | alert("MoveSpare", to->devname, | |
990 | from->devname, info); | |
d52bb542 AC |
991 | break; |
992 | } | |
0fa21e85 | 993 | } |
e78dda3b | 994 | domain_free(domlist); |
2e0172b1 N |
995 | } |
996 | } | |
c3621c0a ML |
997 | |
998 | /* search the statelist to connect external | |
999 | * metadata subarrays with their containers | |
1000 | * We always completely rebuild the tree from scratch as | |
1001 | * that is safest considering the possibility of entries | |
1002 | * disappearing or changing. | |
1003 | */ | |
1004 | static void link_containers_with_subarrays(struct state *list) | |
1005 | { | |
1006 | struct state *st; | |
1007 | struct state *cont; | |
1008 | for (st = list; st; st = st->next) { | |
1009 | st->parent = NULL; | |
1010 | st->subarray = NULL; | |
1011 | } | |
1012 | for (st = list; st; st = st->next) | |
4dd2df09 | 1013 | if (st->parent_devnm[0]) |
c3621c0a | 1014 | for (cont = list; cont; cont = cont->next) |
f566ef45 | 1015 | if (!cont->err && cont->parent_devnm[0] == 0 && |
4dd2df09 | 1016 | strcmp(cont->devnm, st->parent_devnm) == 0) { |
c3621c0a ML |
1017 | st->parent = cont; |
1018 | st->subarray = cont->subarray; | |
1019 | cont->subarray = st; | |
1020 | break; | |
1021 | } | |
1022 | } | |
1023 | ||
b90c0e9a NB |
1024 | /* Not really Monitor but ... */ |
1025 | int Wait(char *dev) | |
1026 | { | |
4dd2df09 | 1027 | char devnm[32]; |
cb91230c TM |
1028 | dev_t rdev; |
1029 | char *tmp; | |
b90c0e9a | 1030 | int rv = 1; |
d3f6cf4f | 1031 | int frozen_remaining = 3; |
b90c0e9a | 1032 | |
cb91230c | 1033 | if (!stat_is_blkdev(dev, &rdev)) |
b90c0e9a | 1034 | return 2; |
cb91230c TM |
1035 | |
1036 | tmp = devid2devnm(rdev); | |
1037 | if (!tmp) { | |
1038 | pr_err("Cannot get md device name.\n"); | |
1039 | return 2; | |
1040 | } | |
1041 | ||
1042 | strcpy(devnm, tmp); | |
b90c0e9a NB |
1043 | |
1044 | while(1) { | |
1045 | struct mdstat_ent *ms = mdstat_read(1, 0); | |
1046 | struct mdstat_ent *e; | |
1047 | ||
f566ef45 | 1048 | for (e = ms; e; e = e->next) |
4dd2df09 | 1049 | if (strcmp(e->devnm, devnm) == 0) |
b90c0e9a NB |
1050 | break; |
1051 | ||
d3f6cf4f | 1052 | if (e && e->percent == RESYNC_NONE) { |
276be514 N |
1053 | /* We could be in the brief pause before something |
1054 | * starts. /proc/mdstat doesn't show that, but | |
1055 | * sync_action does. | |
1056 | */ | |
1057 | struct mdinfo mdi; | |
1058 | char buf[21]; | |
dae13137 JS |
1059 | |
1060 | if (sysfs_init(&mdi, -1, devnm)) | |
1061 | return 2; | |
276be514 N |
1062 | if (sysfs_get_str(&mdi, NULL, "sync_action", |
1063 | buf, 20) > 0 && | |
d3f6cf4f | 1064 | strcmp(buf,"idle\n") != 0) { |
276be514 | 1065 | e->percent = RESYNC_UNKNOWN; |
d3f6cf4f N |
1066 | if (strcmp(buf, "frozen\n") == 0) { |
1067 | if (frozen_remaining == 0) | |
1068 | e->percent = RESYNC_NONE; | |
1069 | else | |
1070 | frozen_remaining -= 1; | |
1071 | } | |
1072 | } | |
276be514 | 1073 | } |
639c3c10 | 1074 | if (!e || e->percent == RESYNC_NONE) { |
e7783ee6 | 1075 | if (e && e->metadata_version && |
c94709e8 DW |
1076 | strncmp(e->metadata_version, "external:", 9) == 0) { |
1077 | if (is_subarray(&e->metadata_version[9])) | |
1078 | ping_monitor(&e->metadata_version[9]); | |
1079 | else | |
4dd2df09 | 1080 | ping_monitor(devnm); |
c94709e8 | 1081 | } |
b90c0e9a NB |
1082 | free_mdstat(ms); |
1083 | return rv; | |
1084 | } | |
89a10d84 | 1085 | free_mdstat(ms); |
b90c0e9a NB |
1086 | rv = 0; |
1087 | mdstat_wait(5); | |
1088 | } | |
1089 | } | |
679eb882 | 1090 | |
43ebc910 GP |
1091 | /* The state "broken" is used only for RAID0/LINEAR - it's the same as |
1092 | * "clean", but used in case the array has one or more members missing. | |
1093 | */ | |
679eb882 | 1094 | static char *clean_states[] = { |
43ebc910 | 1095 | "clear", "inactive", "readonly", "read-auto", "clean", "broken", NULL }; |
679eb882 | 1096 | |
1ea04629 | 1097 | int WaitClean(char *dev, int verbose) |
679eb882 N |
1098 | { |
1099 | int fd; | |
1100 | struct mdinfo *mdi; | |
1101 | int rv = 1; | |
4dd2df09 | 1102 | char devnm[32]; |
679eb882 | 1103 | |
9e04ac1c ZL |
1104 | if (!stat_is_blkdev(dev, NULL)) |
1105 | return 2; | |
679eb882 N |
1106 | fd = open(dev, O_RDONLY); |
1107 | if (fd < 0) { | |
1108 | if (verbose) | |
e7b84f9d | 1109 | pr_err("Couldn't open %s: %s\n", dev, strerror(errno)); |
679eb882 N |
1110 | return 1; |
1111 | } | |
1112 | ||
4dd2df09 N |
1113 | strcpy(devnm, fd2devnm(fd)); |
1114 | mdi = sysfs_read(fd, devnm, GET_VERSION|GET_LEVEL|GET_SAFEMODE); | |
679eb882 N |
1115 | if (!mdi) { |
1116 | if (verbose) | |
7a862a02 | 1117 | pr_err("Failed to read sysfs attributes for %s\n", dev); |
679eb882 N |
1118 | close(fd); |
1119 | return 0; | |
1120 | } | |
1121 | ||
1122 | switch(mdi->array.level) { | |
1123 | case LEVEL_LINEAR: | |
1124 | case LEVEL_MULTIPATH: | |
1125 | case 0: | |
1126 | /* safemode delay is irrelevant for these levels */ | |
1127 | rv = 0; | |
1128 | } | |
1129 | ||
1130 | /* for internal metadata the kernel handles the final clean | |
1131 | * transition, containers can never be dirty | |
1132 | */ | |
1133 | if (!is_subarray(mdi->text_version)) | |
1134 | rv = 0; | |
1135 | ||
1136 | /* safemode disabled ? */ | |
1137 | if (mdi->safe_mode_delay == 0) | |
1138 | rv = 0; | |
1139 | ||
1140 | if (rv) { | |
4dd2df09 | 1141 | int state_fd = sysfs_open(fd2devnm(fd), NULL, "array_state"); |
679eb882 | 1142 | char buf[20]; |
efc67e8e | 1143 | int delay = 5000; |
679eb882 N |
1144 | |
1145 | /* minimize the safe_mode_delay and prepare to wait up to 5s | |
1146 | * for writes to quiesce | |
1147 | */ | |
1148 | sysfs_set_safemode(mdi, 1); | |
679eb882 N |
1149 | |
1150 | /* wait for array_state to be clean */ | |
1151 | while (1) { | |
1152 | rv = read(state_fd, buf, sizeof(buf)); | |
1153 | if (rv < 0) | |
1154 | break; | |
43ebc910 GP |
1155 | if (sysfs_match_word(buf, clean_states) < |
1156 | (int)ARRAY_SIZE(clean_states) - 1) | |
679eb882 | 1157 | break; |
efc67e8e | 1158 | rv = sysfs_wait(state_fd, &delay); |
679eb882 N |
1159 | if (rv < 0 && errno != EINTR) |
1160 | break; | |
1161 | lseek(state_fd, 0, SEEK_SET); | |
1162 | } | |
1163 | if (rv < 0) | |
1164 | rv = 1; | |
1ea04629 | 1165 | else if (ping_monitor(mdi->text_version) == 0) { |
679eb882 N |
1166 | /* we need to ping to close the window between array |
1167 | * state transitioning to clean and the metadata being | |
1168 | * marked clean | |
1169 | */ | |
1170 | rv = 0; | |
1ea04629 | 1171 | } else { |
679eb882 | 1172 | rv = 1; |
1ea04629 MT |
1173 | pr_err("Error connecting monitor with %s\n", dev); |
1174 | } | |
679eb882 | 1175 | if (rv && verbose) |
f566ef45 | 1176 | pr_err("Error waiting for %s to be clean\n", dev); |
679eb882 N |
1177 | |
1178 | /* restore the original safe_mode_delay */ | |
1179 | sysfs_set_safemode(mdi, mdi->safe_mode_delay); | |
1180 | close(state_fd); | |
1181 | } | |
1182 | ||
1183 | sysfs_free(mdi); | |
1184 | close(fd); | |
1185 | ||
1186 | return rv; | |
1187 | } |