]>
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> |
4450e59f | 29 | #include <limits.h> |
773135f5 | 30 | #include <syslog.h> |
af3396da | 31 | #ifndef NO_LIBUDEV |
49b69533 | 32 | #include <libudev.h> |
af3396da | 33 | #endif |
52826846 | 34 | |
2e0172b1 N |
35 | struct state { |
36 | char *devname; | |
4dd2df09 | 37 | char devnm[32]; /* to sync with mdstat info */ |
1d13b599 | 38 | unsigned int utime; |
2e0172b1 N |
39 | int err; |
40 | char *spare_group; | |
41 | int active, working, failed, spare, raid; | |
f1661bd7 | 42 | int from_config; |
73ff0732 | 43 | int from_auto; |
2e0172b1 | 44 | int expected_spares; |
b0599bda N |
45 | int devstate[MAX_DISKS]; |
46 | dev_t devid[MAX_DISKS]; | |
2e0172b1 | 47 | int percent; |
4dd2df09 N |
48 | char parent_devnm[32]; /* For subarray, devnm of parent. |
49 | * For others, "" | |
50 | */ | |
2e0172b1 | 51 | struct supertype *metadata; |
c3621c0a ML |
52 | struct state *subarray;/* for a container it is a link to first subarray |
53 | * for a subarray it is a link to next subarray | |
54 | * in the same container */ | |
55 | struct state *parent; /* for a subarray it is a link to its container | |
56 | */ | |
2e0172b1 N |
57 | struct state *next; |
58 | }; | |
59 | ||
e0bd6a96 N |
60 | struct alert_info { |
61 | char *mailaddr; | |
62 | char *mailfrom; | |
63 | char *alert_cmd; | |
64 | int dosyslog; | |
65 | }; | |
2e0172b1 N |
66 | static int make_daemon(char *pidfile); |
67 | static int check_one_sharer(int scan); | |
7f3b2d1d | 68 | static void write_autorebuild_pid(void); |
e0bd6a96 | 69 | static void alert(char *event, char *dev, char *disc, struct alert_info *info); |
a90e1050 | 70 | static int check_array(struct state *st, struct mdstat_ent *mdstat, |
e0bd6a96 | 71 | int test, struct alert_info *info, |
c2ecf5f6 | 72 | int increments, char *prefer); |
83f3bc5f | 73 | static int add_new_arrays(struct mdstat_ent *mdstat, struct state **statelist, |
e0bd6a96 N |
74 | int test, struct alert_info *info); |
75 | static void try_spare_migration(struct state *statelist, struct alert_info *info); | |
c3621c0a | 76 | static void link_containers_with_subarrays(struct state *list); |
af3396da | 77 | #ifndef NO_LIBUDEV |
49b69533 | 78 | static int check_udev_activity(void); |
af3396da | 79 | #endif |
2e0172b1 | 80 | |
a655e550 | 81 | int Monitor(struct mddev_dev *devlist, |
52826846 | 82 | char *mailaddr, char *alert_cmd, |
95c50205 N |
83 | struct context *c, |
84 | int daemonise, int oneshot, | |
85 | int dosyslog, char *pidfile, int increments, | |
86 | int share) | |
52826846 NB |
87 | { |
88 | /* | |
89 | * Every few seconds, scan every md device looking for changes | |
90 | * When a change is found, log it, possibly run the alert command, | |
91 | * and possibly send Email | |
92 | * | |
93 | * For each array, we record: | |
94 | * Update time | |
95 | * active/working/failed/spare drives | |
96 | * State of each device. | |
e0d19036 | 97 | * %rebuilt if rebuilding |
52826846 NB |
98 | * |
99 | * If the update time changes, check out all the data again | |
100 | * It is possible that we cannot get the state of each device | |
101 | * due to bugs in the md kernel module. | |
e0d19036 NB |
102 | * We also read /proc/mdstat to get rebuild percent, |
103 | * and to get state on all active devices incase of kernel bug. | |
52826846 | 104 | * |
e0d19036 NB |
105 | * Events are: |
106 | * Fail | |
107 | * An active device had Faulty set or Active/Sync removed | |
108 | * FailSpare | |
109 | * A spare device had Faulty set | |
110 | * SpareActive | |
111 | * An active device had a reverse transition | |
112 | * RebuildStarted | |
113 | * percent went from -1 to +ve | |
9a36a9b7 ZB |
114 | * RebuildNN |
115 | * percent went from below to not-below NN% | |
e0d19036 NB |
116 | * DeviceDisappeared |
117 | * Couldn't access a device which was previously visible | |
52826846 NB |
118 | * |
119 | * if we detect an array with active<raid and spare==0 | |
120 | * we look at other arrays that have same spare-group | |
121 | * If we find one with active==raid and spare>0, | |
122 | * and if we can get_disk_info and find a name | |
123 | * Then we hot-remove and hot-add to the other array | |
124 | * | |
e0d19036 NB |
125 | * If devlist is NULL, then we can monitor everything because --scan |
126 | * was given. We get an initial list from config file and add anything | |
127 | * that appears in /proc/mdstat | |
52826846 NB |
128 | */ |
129 | ||
2e0172b1 | 130 | struct state *statelist = NULL; |
b657208c | 131 | struct state *st2; |
52826846 | 132 | int finished = 0; |
e0d19036 | 133 | struct mdstat_ent *mdstat = NULL; |
72362f18 | 134 | char *mailfrom; |
e0bd6a96 | 135 | struct alert_info info; |
9f3dd454 | 136 | struct mddev_ident *mdlist; |
af3396da | 137 | int delay_for_event = c->delay; |
e0d19036 | 138 | |
f5ff2988 | 139 | if (!mailaddr) |
8aec876d | 140 | mailaddr = conf_get_mailaddr(); |
4948b8f7 | 141 | |
f5ff2988 | 142 | if (!alert_cmd) |
8aec876d | 143 | alert_cmd = conf_get_program(); |
f5ff2988 KT |
144 | |
145 | mailfrom = conf_get_mailfrom(); | |
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 | } |
f5ff2988 KT |
151 | |
152 | if (c->verbose) { | |
153 | pr_err("Monitor is started with delay %ds\n", c->delay); | |
154 | if (mailaddr) | |
155 | pr_err("Monitor using email address %s\n", mailaddr); | |
156 | if (alert_cmd) | |
157 | pr_err("Monitor using program %s\n", alert_cmd); | |
158 | } | |
159 | ||
e0bd6a96 N |
160 | info.alert_cmd = alert_cmd; |
161 | info.mailaddr = mailaddr; | |
162 | info.mailfrom = mailfrom; | |
163 | info.dosyslog = dosyslog; | |
d013a55e | 164 | |
7f3b2d1d BK |
165 | if (share){ |
166 | if (check_one_sharer(c->scan)) | |
167 | return 1; | |
168 | } | |
169 | ||
44d337f0 AC |
170 | if (daemonise) { |
171 | int rv = make_daemon(pidfile); | |
172 | if (rv >= 0) | |
173 | return rv; | |
174 | } | |
e0d19036 | 175 | |
1011e834 | 176 | if (share) |
7f3b2d1d | 177 | write_autorebuild_pid(); |
edde9560 | 178 | |
e0d19036 | 179 | if (devlist == NULL) { |
9f3dd454 | 180 | mdlist = conf_get_ident(NULL); |
f566ef45 | 181 | for (; mdlist; mdlist = mdlist->next) { |
fe056d1f | 182 | struct state *st; |
f566ef45 | 183 | |
fe056d1f N |
184 | if (mdlist->devname == NULL) |
185 | continue; | |
112cace6 N |
186 | if (strcasecmp(mdlist->devname, "<ignore>") == 0) |
187 | continue; | |
e702f392 | 188 | |
503975b9 | 189 | st = xcalloc(1, sizeof *st); |
db2d001c | 190 | if (mdlist->devname[0] == '/') |
503975b9 | 191 | st->devname = xstrdup(mdlist->devname); |
db2d001c | 192 | else { |
503975b9 | 193 | st->devname = xmalloc(8+strlen(mdlist->devname)+1); |
db2d001c N |
194 | strcpy(strcpy(st->devname, "/dev/md/"), |
195 | mdlist->devname); | |
196 | } | |
e702f392 KT |
197 | if (!is_mddev(mdlist->devname)) |
198 | return 1; | |
e0d19036 | 199 | st->next = statelist; |
4dd2df09 | 200 | st->devnm[0] = 0; |
9dad51d4 | 201 | st->percent = RESYNC_UNKNOWN; |
f1661bd7 | 202 | st->from_config = 1; |
feb716e9 | 203 | st->expected_spares = mdlist->spare_disks; |
e0d19036 | 204 | if (mdlist->spare_group) |
503975b9 | 205 | st->spare_group = xstrdup(mdlist->spare_group); |
e0d19036 NB |
206 | statelist = st; |
207 | } | |
208 | } else { | |
a655e550 | 209 | struct mddev_dev *dv; |
f566ef45 JS |
210 | |
211 | for (dv = devlist; dv; dv = dv->next) { | |
e702f392 KT |
212 | struct state *st; |
213 | ||
214 | if (!is_mddev(dv->devname)) | |
215 | return 1; | |
216 | ||
217 | st = xcalloc(1, sizeof *st); | |
9f3dd454 | 218 | mdlist = conf_get_ident(dv->devname); |
503975b9 | 219 | st->devname = xstrdup(dv->devname); |
e0d19036 | 220 | st->next = statelist; |
4dd2df09 | 221 | st->devnm[0] = 0; |
9dad51d4 | 222 | st->percent = RESYNC_UNKNOWN; |
feb716e9 | 223 | st->expected_spares = -1; |
e5329c37 NB |
224 | if (mdlist) { |
225 | st->expected_spares = mdlist->spare_disks; | |
226 | if (mdlist->spare_group) | |
503975b9 | 227 | st->spare_group = xstrdup(mdlist->spare_group); |
e5329c37 | 228 | } |
e0d19036 NB |
229 | statelist = st; |
230 | } | |
231 | } | |
232 | ||
f566ef45 | 233 | while (!finished) { |
aa88f531 | 234 | int new_found = 0; |
73ff0732 | 235 | struct state *st, **stp; |
a90e1050 | 236 | int anydegraded = 0; |
007087d0 | 237 | int anyredundant = 0; |
e0d19036 NB |
238 | |
239 | if (mdstat) | |
240 | free_mdstat(mdstat); | |
f566ef45 | 241 | mdstat = mdstat_read(oneshot ? 0 : 1, 0); |
e0d19036 | 242 | |
007087d0 | 243 | for (st = statelist; st; st = st->next) { |
95c50205 N |
244 | if (check_array(st, mdstat, c->test, &info, |
245 | increments, c->prefer)) | |
a90e1050 | 246 | anydegraded = 1; |
007087d0 MT |
247 | /* for external arrays, metadata is filled for |
248 | * containers only | |
249 | */ | |
250 | if (st->metadata && st->metadata->ss->external) | |
251 | continue; | |
252 | if (st->err == 0 && !anyredundant) | |
253 | anyredundant = 1; | |
254 | } | |
1011e834 | 255 | |
e0d19036 | 256 | /* now check if there are any new devices found in mdstat */ |
95c50205 N |
257 | if (c->scan) |
258 | new_found = add_new_arrays(mdstat, &statelist, c->test, | |
e0bd6a96 | 259 | &info); |
2e0172b1 | 260 | |
e0d19036 NB |
261 | /* If an array has active < raid && spare == 0 && spare_group != NULL |
262 | * Look for another array with spare > 0 and active == raid and same spare_group | |
49b69533 | 263 | * if found, choose a device and hotremove/hotadd |
e0d19036 | 264 | */ |
a90e1050 | 265 | if (share && anydegraded) |
e0bd6a96 | 266 | try_spare_migration(statelist, &info); |
aa88f531 NB |
267 | if (!new_found) { |
268 | if (oneshot) | |
269 | break; | |
007087d0 | 270 | else if (!anyredundant) { |
8c80d305 | 271 | pr_err("No array with redundancy detected, stopping\n"); |
007087d0 MT |
272 | break; |
273 | } | |
e2308733 | 274 | else { |
af3396da | 275 | #ifndef NO_LIBUDEV |
cab9c67d | 276 | /* |
af3396da MT |
277 | * Wait for udevd to finish new devices |
278 | * processing. | |
cab9c67d | 279 | */ |
af3396da MT |
280 | if (mdstat_wait(delay_for_event) && |
281 | check_udev_activity()) | |
49b69533 | 282 | pr_err("Error while waiting for UDEV to complete new devices processing\n"); |
af3396da MT |
283 | #else |
284 | int wait_result = mdstat_wait(delay_for_event); | |
285 | /* | |
286 | * Give chance to process new device | |
287 | */ | |
288 | if (wait_result != 0) { | |
289 | if (c->delay > 5) | |
290 | delay_for_event = 5; | |
291 | } else | |
292 | delay_for_event = c->delay; | |
293 | #endif | |
e2308733 MT |
294 | mdstat_close(); |
295 | } | |
aa88f531 | 296 | } |
95c50205 | 297 | c->test = 0; |
73ff0732 N |
298 | |
299 | for (stp = &statelist; (st = *stp) != NULL; ) { | |
300 | if (st->from_auto && st->err > 5) { | |
301 | *stp = st->next; | |
302 | free(st->devname); | |
303 | free(st->spare_group); | |
304 | free(st); | |
305 | } else | |
306 | stp = &st->next; | |
307 | } | |
52826846 | 308 | } |
b657208c JS |
309 | for (st2 = statelist; st2; st2 = statelist) { |
310 | statelist = st2->next; | |
311 | free(st2); | |
312 | } | |
313 | ||
b5e64645 NB |
314 | if (pidfile) |
315 | unlink(pidfile); | |
52826846 NB |
316 | return 0; |
317 | } | |
318 | ||
2e0172b1 N |
319 | static int make_daemon(char *pidfile) |
320 | { | |
44d337f0 AC |
321 | /* Return: |
322 | * -1 in the forked daemon | |
323 | * 0 in the parent | |
324 | * 1 on error | |
325 | * so a none-negative becomes the exit code. | |
326 | */ | |
2e0172b1 N |
327 | int pid = fork(); |
328 | if (pid > 0) { | |
329 | if (!pidfile) | |
330 | printf("%d\n", pid); | |
331 | else { | |
ca4b156b MT |
332 | FILE *pid_file = NULL; |
333 | int fd = open(pidfile, O_WRONLY | O_CREAT | O_TRUNC, | |
334 | 0644); | |
335 | if (fd >= 0) | |
336 | pid_file = fdopen(fd, "w"); | |
2e0172b1 N |
337 | if (!pid_file) |
338 | perror("cannot create pid file"); | |
339 | else { | |
340 | fprintf(pid_file,"%d\n", pid); | |
341 | fclose(pid_file); | |
342 | } | |
343 | } | |
344 | return 0; | |
345 | } | |
346 | if (pid < 0) { | |
347 | perror("daemonise"); | |
348 | return 1; | |
349 | } | |
ff6bb131 | 350 | manage_fork_fds(0); |
2e0172b1 | 351 | setsid(); |
44d337f0 | 352 | return -1; |
2e0172b1 N |
353 | } |
354 | ||
355 | static int check_one_sharer(int scan) | |
356 | { | |
185ec439 CL |
357 | int pid; |
358 | FILE *comm_fp; | |
2e0172b1 | 359 | FILE *fp; |
7f3b2d1d BK |
360 | char comm_path[PATH_MAX]; |
361 | char path[PATH_MAX]; | |
185ec439 CL |
362 | char comm[20]; |
363 | ||
0011874f JS |
364 | sprintf(path, "%s/autorebuild.pid", MDMON_DIR); |
365 | fp = fopen(path, "r"); | |
2e0172b1 | 366 | if (fp) { |
71204a50 N |
367 | if (fscanf(fp, "%d", &pid) != 1) |
368 | pid = -1; | |
185ec439 CL |
369 | snprintf(comm_path, sizeof(comm_path), |
370 | "/proc/%d/comm", pid); | |
371 | comm_fp = fopen(comm_path, "r"); | |
372 | if (comm_fp) { | |
1c66260d | 373 | if (fscanf(comm_fp, "%19s", comm) && |
185ec439 CL |
374 | strncmp(basename(comm), Name, strlen(Name)) == 0) { |
375 | if (scan) { | |
376 | pr_err("Only one autorebuild process allowed in scan mode, aborting\n"); | |
377 | fclose(comm_fp); | |
378 | fclose(fp); | |
379 | return 1; | |
380 | } else { | |
381 | pr_err("Warning: One autorebuild process already running.\n"); | |
382 | } | |
2e0172b1 | 383 | } |
185ec439 | 384 | fclose(comm_fp); |
2e0172b1 N |
385 | } |
386 | fclose(fp); | |
387 | } | |
7f3b2d1d BK |
388 | return 0; |
389 | } | |
390 | ||
391 | static void write_autorebuild_pid() | |
392 | { | |
393 | char path[PATH_MAX]; | |
394 | int pid; | |
ca4b156b | 395 | FILE *fp = NULL; |
7f3b2d1d BK |
396 | sprintf(path, "%s/autorebuild.pid", MDMON_DIR); |
397 | ||
ca4b156b | 398 | if (mkdir(MDMON_DIR, 0700) < 0 && errno != EEXIST) { |
7f3b2d1d BK |
399 | pr_err("Can't create autorebuild.pid file\n"); |
400 | } else { | |
ca4b156b MT |
401 | int fd = open(path, O_WRONLY | O_CREAT | O_TRUNC, 0700); |
402 | ||
403 | if (fd >= 0) | |
404 | fp = fdopen(fd, "w"); | |
405 | ||
7f3b2d1d | 406 | if (!fp) |
7a862a02 | 407 | pr_err("Can't create autorebuild.pid file\n"); |
7f3b2d1d BK |
408 | else { |
409 | pid = getpid(); | |
410 | fprintf(fp, "%d\n", pid); | |
411 | fclose(fp); | |
2e0172b1 N |
412 | } |
413 | } | |
2e0172b1 | 414 | } |
52826846 | 415 | |
e0bd6a96 | 416 | static void alert(char *event, char *dev, char *disc, struct alert_info *info) |
52826846 | 417 | { |
773135f5 NB |
418 | int priority; |
419 | ||
b1bac75b | 420 | if (!info->alert_cmd && !info->mailaddr && !info->dosyslog) { |
cd29a5c8 | 421 | time_t now = time(0); |
aba69144 | 422 | |
f566ef45 JS |
423 | printf("%1.15s: %s on %s %s\n", ctime(&now) + 4, |
424 | event, dev, disc?disc:"unknown device"); | |
cd29a5c8 | 425 | } |
e0bd6a96 | 426 | if (info->alert_cmd) { |
52826846 NB |
427 | int pid = fork(); |
428 | switch(pid) { | |
429 | default: | |
430 | waitpid(pid, NULL, 0); | |
431 | break; | |
432 | case -1: | |
433 | break; | |
434 | case 0: | |
e0bd6a96 N |
435 | execl(info->alert_cmd, info->alert_cmd, |
436 | event, dev, disc, NULL); | |
52826846 NB |
437 | exit(2); |
438 | } | |
439 | } | |
f566ef45 JS |
440 | if (info->mailaddr && (strncmp(event, "Fail", 4) == 0 || |
441 | strncmp(event, "Test", 4) == 0 || | |
442 | strncmp(event, "Spares", 6) == 0 || | |
443 | strncmp(event, "Degrade", 7) == 0)) { | |
52826846 NB |
444 | FILE *mp = popen(Sendmail, "w"); |
445 | if (mp) { | |
008e1100 | 446 | FILE *mdstat; |
52826846 | 447 | char hname[256]; |
83a379cf | 448 | |
52826846 | 449 | gethostname(hname, sizeof(hname)); |
83a379cf LF |
450 | signal_s(SIGPIPE, SIG_IGN); |
451 | ||
e0bd6a96 N |
452 | if (info->mailfrom) |
453 | fprintf(mp, "From: %s\n", info->mailfrom); | |
4948b8f7 | 454 | else |
f566ef45 JS |
455 | fprintf(mp, "From: %s monitoring <root>\n", |
456 | Name); | |
e0bd6a96 N |
457 | fprintf(mp, "To: %s\n", info->mailaddr); |
458 | fprintf(mp, "Subject: %s event on %s:%s\n\n", | |
459 | event, dev, hname); | |
52826846 | 460 | |
e0bd6a96 | 461 | fprintf(mp, |
7a862a02 | 462 | "This is an automatically generated mail message from %s\n", Name); |
52826846 NB |
463 | fprintf(mp, "running on %s\n\n", hname); |
464 | ||
e0bd6a96 | 465 | fprintf(mp, |
7a862a02 | 466 | "A %s event had been detected on md device %s.\n\n", event, dev); |
52826846 | 467 | |
37dfc3d6 | 468 | if (disc && disc[0] != ' ') |
e0bd6a96 | 469 | fprintf(mp, |
7a862a02 | 470 | "It could be related to component device %s.\n\n", disc); |
37dfc3d6 NB |
471 | if (disc && disc[0] == ' ') |
472 | fprintf(mp, "Extra information:%s.\n\n", disc); | |
52826846 NB |
473 | |
474 | fprintf(mp, "Faithfully yours, etc.\n"); | |
008e1100 NB |
475 | |
476 | mdstat = fopen("/proc/mdstat", "r"); | |
477 | if (mdstat) { | |
478 | char buf[8192]; | |
479 | int n; | |
e0bd6a96 | 480 | fprintf(mp, |
7a862a02 | 481 | "\nP.S. The /proc/mdstat file currently contains the following:\n\n"); |
f566ef45 JS |
482 | while ((n = fread(buf, 1, sizeof(buf), |
483 | mdstat)) > 0) | |
484 | n = fwrite(buf, 1, n, mp); | |
008e1100 NB |
485 | fclose(mdstat); |
486 | } | |
6278fb3a | 487 | pclose(mp); |
52826846 | 488 | } |
52826846 | 489 | } |
773135f5 NB |
490 | |
491 | /* log the event to syslog maybe */ | |
e0bd6a96 | 492 | if (info->dosyslog) { |
773135f5 NB |
493 | /* Log at a different severity depending on the event. |
494 | * | |
495 | * These are the critical events: */ | |
f566ef45 JS |
496 | if (strncmp(event, "Fail", 4) == 0 || |
497 | strncmp(event, "Degrade", 7) == 0 || | |
498 | strncmp(event, "DeviceDisappeared", 17) == 0) | |
773135f5 NB |
499 | priority = LOG_CRIT; |
500 | /* Good to know about, but are not failures: */ | |
f566ef45 JS |
501 | else if (strncmp(event, "Rebuild", 7) == 0 || |
502 | strncmp(event, "MoveSpare", 9) == 0 || | |
d1732eeb | 503 | strncmp(event, "Spares", 6) != 0) |
773135f5 NB |
504 | priority = LOG_WARNING; |
505 | /* Everything else: */ | |
506 | else | |
507 | priority = LOG_INFO; | |
508 | ||
b033913a | 509 | if (disc && disc[0] != ' ') |
e0bd6a96 | 510 | syslog(priority, |
7a862a02 | 511 | "%s event detected on md device %s, component device %s", event, dev, disc); |
b033913a N |
512 | else if (disc) |
513 | syslog(priority, | |
514 | "%s event detected on md device %s: %s", | |
515 | event, dev, disc); | |
773135f5 | 516 | else |
e0bd6a96 N |
517 | syslog(priority, |
518 | "%s event detected on md device %s", | |
519 | event, dev); | |
773135f5 | 520 | } |
52826846 | 521 | } |
b90c0e9a | 522 | |
a90e1050 | 523 | static int check_array(struct state *st, struct mdstat_ent *mdstat, |
e0bd6a96 | 524 | int test, struct alert_info *ainfo, |
c2ecf5f6 | 525 | int increments, char *prefer) |
2e0172b1 | 526 | { |
ff044d6b AC |
527 | /* Update the state 'st' to reflect any changes shown in mdstat, |
528 | * or found by directly examining the array, and return | |
529 | * '1' if the array is degraded, or '0' if it is optimal (or dead). | |
530 | */ | |
b0599bda | 531 | struct { int state, major, minor; } info[MAX_DISKS]; |
aed5f5c3 | 532 | struct mdinfo *sra = NULL; |
2e0172b1 N |
533 | mdu_array_info_t array; |
534 | struct mdstat_ent *mse = NULL, *mse2; | |
535 | char *dev = st->devname; | |
b9a0309c | 536 | int fd; |
2e0172b1 | 537 | int i; |
9e6d9291 N |
538 | int remaining_disks; |
539 | int last_disk; | |
721b662b | 540 | int new_array = 0; |
1830e74b | 541 | int retval; |
802961a2 | 542 | int is_container = 0; |
b3ab4e4d | 543 | unsigned long redundancy_only_flags = 0; |
2e0172b1 N |
544 | |
545 | if (test) | |
e0bd6a96 | 546 | alert("TestMessage", dev, NULL, ainfo); |
b9a0309c | 547 | |
1830e74b JS |
548 | retval = 0; |
549 | ||
2e0172b1 | 550 | fd = open(dev, O_RDONLY); |
13e5d845 JS |
551 | if (fd < 0) |
552 | goto disappeared; | |
b9a0309c | 553 | |
802961a2 MT |
554 | if (st->devnm[0] == 0) |
555 | strcpy(st->devnm, fd2devnm(fd)); | |
556 | ||
557 | for (mse2 = mdstat; mse2; mse2 = mse2->next) | |
558 | if (strcmp(mse2->devnm, st->devnm) == 0) { | |
559 | mse2->devnm[0] = 0; /* flag it as "used" */ | |
560 | mse = mse2; | |
561 | } | |
562 | ||
563 | if (!mse) { | |
564 | /* duplicated array in statelist | |
565 | * or re-created after reading mdstat | |
566 | */ | |
567 | st->err++; | |
568 | goto out; | |
569 | } | |
570 | ||
571 | if (mse->level == NULL) | |
572 | is_container = 1; | |
573 | ||
2dab69c9 | 574 | if (!is_container && !md_array_active(fd)) |
13e5d845 | 575 | goto disappeared; |
b9a0309c | 576 | |
2e0172b1 | 577 | fcntl(fd, F_SETFD, FD_CLOEXEC); |
13e5d845 JS |
578 | if (md_get_array_info(fd, &array) < 0) |
579 | goto disappeared; | |
580 | ||
b3ab4e4d MT |
581 | if (!is_container && map_name(pers, mse->level) > 0) |
582 | redundancy_only_flags |= GET_MISMATCH; | |
802961a2 MT |
583 | |
584 | sra = sysfs_read(-1, st->devnm, GET_LEVEL | GET_DISKS | GET_DEVS | | |
b3ab4e4d | 585 | GET_STATE | redundancy_only_flags); |
aed5f5c3 | 586 | |
aed5f5c3 JS |
587 | if (!sra) |
588 | goto disappeared; | |
589 | ||
2e0172b1 N |
590 | /* It's much easier to list what array levels can't |
591 | * have a device disappear than all of them that can | |
592 | */ | |
48bc2ade | 593 | if (sra->array.level == 0 || sra->array.level == -1) { |
f1661bd7 | 594 | if (!st->err && !st->from_config) |
04e27c20 | 595 | alert("DeviceDisappeared", dev, " Wrong-Level", ainfo); |
73ff0732 | 596 | st->err++; |
1830e74b | 597 | goto out; |
2e0172b1 | 598 | } |
2e0172b1 | 599 | |
2e0172b1 N |
600 | /* this array is in /proc/mdstat */ |
601 | if (array.utime == 0) | |
602 | /* external arrays don't update utime, so | |
603 | * just make sure it is always different. */ | |
604 | array.utime = st->utime + 1;; | |
605 | ||
0f760384 | 606 | if (st->err) { |
73ff0732 | 607 | /* New array appeared where previously had an error */ |
0f760384 N |
608 | st->err = 0; |
609 | st->percent = RESYNC_NONE; | |
721b662b | 610 | new_array = 1; |
007087d0 MT |
611 | if (!is_container) |
612 | alert("NewArray", st->devname, NULL, ainfo); | |
0f760384 N |
613 | } |
614 | ||
b8e5713c | 615 | if (st->utime == array.utime && st->failed == sra->array.failed_disks && |
e5eb6857 | 616 | st->working == sra->array.working_disks && |
b98943a4 | 617 | st->spare == sra->array.spare_disks && |
d7be7d87 | 618 | (mse == NULL || (mse->percent == st->percent))) { |
ff044d6b | 619 | if ((st->active < st->raid) && st->spare == 0) |
1830e74b JS |
620 | retval = 1; |
621 | goto out; | |
2e0172b1 N |
622 | } |
623 | if (st->utime == 0 && /* new array */ | |
f27904a5 | 624 | mse->pattern && strchr(mse->pattern, '_') /* degraded */) |
e0bd6a96 | 625 | alert("DegradedArray", dev, NULL, ainfo); |
2e0172b1 | 626 | |
b98943a4 JS |
627 | if (st->utime == 0 && /* new array */ st->expected_spares > 0 && |
628 | sra->array.spare_disks < st->expected_spares) | |
e0bd6a96 | 629 | alert("SparesMissing", dev, NULL, ainfo); |
9dad51d4 | 630 | if (st->percent < 0 && st->percent != RESYNC_UNKNOWN && |
2e0172b1 | 631 | mse->percent >= 0) |
e0bd6a96 | 632 | alert("RebuildStarted", dev, NULL, ainfo); |
f27904a5 | 633 | if (st->percent >= 0 && mse->percent >= 0 && |
2e0172b1 | 634 | (mse->percent / increments) > (st->percent / increments)) { |
97589839 | 635 | char percentalert[18]; |
f27904a5 JS |
636 | /* |
637 | * "RebuildNN" (10 chars) or "RebuildStarted" (15 chars) | |
638 | */ | |
2e0172b1 N |
639 | |
640 | if((mse->percent / increments) == 0) | |
f27904a5 JS |
641 | snprintf(percentalert, sizeof(percentalert), |
642 | "RebuildStarted"); | |
2e0172b1 | 643 | else |
f27904a5 JS |
644 | snprintf(percentalert, sizeof(percentalert), |
645 | "Rebuild%02d", mse->percent); | |
2e0172b1 | 646 | |
e0bd6a96 | 647 | alert(percentalert, dev, NULL, ainfo); |
2e0172b1 N |
648 | } |
649 | ||
f27904a5 | 650 | if (mse->percent == RESYNC_NONE && st->percent >= 0) { |
2e0172b1 N |
651 | /* Rebuild/sync/whatever just finished. |
652 | * If there is a number in /mismatch_cnt, | |
653 | * we should report that. | |
654 | */ | |
2e0172b1 | 655 | if (sra && sra->mismatch_cnt > 0) { |
8453f8d0 LD |
656 | char cnt[80]; |
657 | snprintf(cnt, sizeof(cnt), | |
658 | " mismatches found: %d (on raid level %d)", | |
f566ef45 | 659 | sra->mismatch_cnt, sra->array.level); |
e0bd6a96 | 660 | alert("RebuildFinished", dev, cnt, ainfo); |
2e0172b1 | 661 | } else |
e0bd6a96 | 662 | alert("RebuildFinished", dev, NULL, ainfo); |
2e0172b1 N |
663 | } |
664 | st->percent = mse->percent; | |
665 | ||
b98943a4 | 666 | remaining_disks = sra->array.nr_disks; |
f27904a5 | 667 | for (i = 0; i < MAX_DISKS && remaining_disks > 0; i++) { |
2e0172b1 N |
668 | mdu_disk_info_t disc; |
669 | disc.number = i; | |
d97572f5 | 670 | if (md_get_disk_info(fd, &disc) >= 0) { |
2e0172b1 N |
671 | info[i].state = disc.state; |
672 | info[i].major = disc.major; | |
673 | info[i].minor = disc.minor; | |
9e6d9291 N |
674 | if (disc.major || disc.minor) |
675 | remaining_disks --; | |
2e0172b1 N |
676 | } else |
677 | info[i].major = info[i].minor = 0; | |
678 | } | |
9e6d9291 | 679 | last_disk = i; |
2e0172b1 | 680 | |
4019ad07 JL |
681 | if (mse->metadata_version && |
682 | strncmp(mse->metadata_version, "external:", 9) == 0 && | |
4dd2df09 N |
683 | is_subarray(mse->metadata_version+9)) { |
684 | char *sl; | |
f566ef45 | 685 | strcpy(st->parent_devnm, mse->metadata_version + 10); |
4dd2df09 N |
686 | sl = strchr(st->parent_devnm, '/'); |
687 | if (sl) | |
688 | *sl = 0; | |
689 | } else | |
690 | st->parent_devnm[0] = 0; | |
f27904a5 | 691 | if (st->metadata == NULL && st->parent_devnm[0] == 0) |
2e0172b1 N |
692 | st->metadata = super_by_fd(fd, NULL); |
693 | ||
f566ef45 JS |
694 | for (i = 0; i < MAX_DISKS; i++) { |
695 | mdu_disk_info_t disc = {0, 0, 0, 0, 0}; | |
696 | int newstate = 0; | |
2e0172b1 N |
697 | int change; |
698 | char *dv = NULL; | |
699 | disc.number = i; | |
f27904a5 | 700 | if (i < last_disk && (info[i].major || info[i].minor)) { |
2e0172b1 | 701 | newstate = info[i].state; |
f27904a5 JS |
702 | dv = map_dev_preferred(info[i].major, info[i].minor, 1, |
703 | prefer); | |
2e0172b1 N |
704 | disc.state = newstate; |
705 | disc.major = info[i].major; | |
706 | disc.minor = info[i].minor; | |
721b662b N |
707 | } else |
708 | newstate = (1 << MD_DISK_REMOVED); | |
709 | ||
2e0172b1 | 710 | if (dv == NULL && st->devid[i]) |
f27904a5 JS |
711 | dv = map_dev_preferred(major(st->devid[i]), |
712 | minor(st->devid[i]), 1, prefer); | |
2e0172b1 | 713 | change = newstate ^ st->devstate[i]; |
721b662b | 714 | if (st->utime && change && !st->err && !new_array) { |
f27904a5 | 715 | if ((st->devstate[i]&change) & (1 << MD_DISK_SYNC)) |
e0bd6a96 | 716 | alert("Fail", dev, dv, ainfo); |
f27904a5 | 717 | else if ((newstate & (1 << MD_DISK_FAULTY)) && |
2e0172b1 | 718 | (disc.major || disc.minor) && |
f27904a5 JS |
719 | st->devid[i] == makedev(disc.major, |
720 | disc.minor)) | |
e0bd6a96 | 721 | alert("FailSpare", dev, dv, ainfo); |
f27904a5 | 722 | else if ((newstate&change) & (1 << MD_DISK_SYNC)) |
e0bd6a96 | 723 | alert("SpareActive", dev, dv, ainfo); |
2e0172b1 N |
724 | } |
725 | st->devstate[i] = newstate; | |
726 | st->devid[i] = makedev(disc.major, disc.minor); | |
727 | } | |
b98943a4 | 728 | st->active = sra->array.active_disks; |
e5eb6857 | 729 | st->working = sra->array.working_disks; |
b98943a4 | 730 | st->spare = sra->array.spare_disks; |
b8e5713c | 731 | st->failed = sra->array.failed_disks; |
2e0172b1 | 732 | st->utime = array.utime; |
12a9d21f | 733 | st->raid = sra->array.raid_disks; |
2e0172b1 | 734 | st->err = 0; |
a90e1050 | 735 | if ((st->active < st->raid) && st->spare == 0) |
1830e74b JS |
736 | retval = 1; |
737 | ||
738 | out: | |
aed5f5c3 JS |
739 | if (sra) |
740 | sysfs_free(sra); | |
802961a2 | 741 | if (fd >= 0) |
13e5d845 | 742 | close(fd); |
1830e74b | 743 | return retval; |
13e5d845 JS |
744 | |
745 | disappeared: | |
007087d0 | 746 | if (!st->err && !is_container) |
13e5d845 JS |
747 | alert("DeviceDisappeared", dev, NULL, ainfo); |
748 | st->err++; | |
749 | goto out; | |
2e0172b1 N |
750 | } |
751 | ||
83f3bc5f | 752 | static int add_new_arrays(struct mdstat_ent *mdstat, struct state **statelist, |
e0bd6a96 | 753 | int test, struct alert_info *info) |
2e0172b1 N |
754 | { |
755 | struct mdstat_ent *mse; | |
756 | int new_found = 0; | |
1e08717f | 757 | char *name; |
2e0172b1 | 758 | |
f566ef45 | 759 | for (mse = mdstat; mse; mse = mse->next) |
d7be7d87 | 760 | if (mse->devnm[0] && (!mse->level || /* retrieve containers */ |
f566ef45 JS |
761 | (strcmp(mse->level, "raid0") != 0 && |
762 | strcmp(mse->level, "linear") != 0))) { | |
503975b9 | 763 | struct state *st = xcalloc(1, sizeof *st); |
2e0172b1 N |
764 | mdu_array_info_t array; |
765 | int fd; | |
1e08717f SV |
766 | |
767 | name = get_md_name(mse->devnm); | |
768 | if (!name) { | |
769 | free(st); | |
770 | continue; | |
771 | } | |
772 | ||
773 | st->devname = xstrdup(name); | |
2e0172b1 | 774 | if ((fd = open(st->devname, O_RDONLY)) < 0 || |
9cd39f01 | 775 | md_get_array_info(fd, &array) < 0) { |
2e0172b1 | 776 | /* no such array */ |
9cd39f01 JS |
777 | if (fd >= 0) |
778 | close(fd); | |
2e0172b1 N |
779 | put_md_name(st->devname); |
780 | free(st->devname); | |
781 | if (st->metadata) { | |
782 | st->metadata->ss->free_super(st->metadata); | |
783 | free(st->metadata); | |
784 | } | |
785 | free(st); | |
786 | continue; | |
787 | } | |
788 | close(fd); | |
83f3bc5f | 789 | st->next = *statelist; |
2e0172b1 | 790 | st->err = 1; |
73ff0732 | 791 | st->from_auto = 1; |
4dd2df09 | 792 | strcpy(st->devnm, mse->devnm); |
9dad51d4 | 793 | st->percent = RESYNC_UNKNOWN; |
2e0172b1 | 794 | st->expected_spares = -1; |
eb28e119 | 795 | if (mse->metadata_version && |
f566ef45 JS |
796 | strncmp(mse->metadata_version, |
797 | "external:", 9) == 0 && | |
4dd2df09 N |
798 | is_subarray(mse->metadata_version+9)) { |
799 | char *sl; | |
800 | strcpy(st->parent_devnm, | |
801 | mse->metadata_version+10); | |
802 | sl = strchr(st->parent_devnm, '/'); | |
803 | *sl = 0; | |
804 | } else | |
805 | st->parent_devnm[0] = 0; | |
83f3bc5f | 806 | *statelist = st; |
2e0172b1 | 807 | if (test) |
e0bd6a96 | 808 | alert("TestMessage", st->devname, NULL, info); |
2e0172b1 N |
809 | new_found = 1; |
810 | } | |
811 | return new_found; | |
812 | } | |
813 | ||
fbfdcb06 AO |
814 | static int get_required_spare_criteria(struct state *st, |
815 | struct spare_criteria *sc) | |
80e7f8c3 AC |
816 | { |
817 | int fd; | |
80e7f8c3 | 818 | |
f566ef45 | 819 | if (!st->metadata || !st->metadata->ss->get_spare_criteria) { |
fbfdcb06 | 820 | sc->min_size = 0; |
4b57ecf6 | 821 | sc->sector_size = 0; |
de697acc | 822 | return 0; |
300f5033 | 823 | } |
80e7f8c3 AC |
824 | |
825 | fd = open(st->devname, O_RDONLY); | |
826 | if (fd < 0) | |
de697acc AC |
827 | return 1; |
828 | if (st->metadata->ss->external) | |
829 | st->metadata->ss->load_container(st->metadata, fd, st->devname); | |
830 | else | |
831 | st->metadata->ss->load_super(st->metadata, fd, st->devname); | |
80e7f8c3 | 832 | close(fd); |
de697acc AC |
833 | if (!st->metadata->sb) |
834 | return 1; | |
fbfdcb06 AO |
835 | |
836 | st->metadata->ss->get_spare_criteria(st->metadata, sc); | |
80e7f8c3 AC |
837 | st->metadata->ss->free_super(st->metadata); |
838 | ||
de697acc | 839 | return 0; |
80e7f8c3 AC |
840 | } |
841 | ||
5ec0f373 | 842 | static int check_donor(struct state *from, struct state *to) |
2feb22ef | 843 | { |
66f5c4b6 N |
844 | struct state *sub; |
845 | ||
2feb22ef N |
846 | if (from == to) |
847 | return 0; | |
66f5c4b6 N |
848 | if (from->parent) |
849 | /* Cannot move from a member */ | |
2feb22ef | 850 | return 0; |
ff044d6b AC |
851 | if (from->err) |
852 | return 0; | |
66f5c4b6 N |
853 | for (sub = from->subarray; sub; sub = sub->subarray) |
854 | /* If source array has degraded subarrays, don't | |
855 | * remove anything | |
856 | */ | |
857 | if (sub->active < sub->raid) | |
858 | return 0; | |
859 | if (from->metadata->ss->external == 0) | |
860 | if (from->active < from->raid) | |
861 | return 0; | |
2feb22ef N |
862 | if (from->spare <= 0) |
863 | return 0; | |
e78dda3b | 864 | return 1; |
2feb22ef N |
865 | } |
866 | ||
0f0749ad | 867 | static dev_t choose_spare(struct state *from, struct state *to, |
f566ef45 | 868 | struct domainlist *domlist, struct spare_criteria *sc) |
0fa21e85 N |
869 | { |
870 | int d; | |
0f0749ad | 871 | dev_t dev = 0; |
0fa21e85 | 872 | |
b0599bda | 873 | for (d = from->raid; !dev && d < MAX_DISKS; d++) { |
f566ef45 | 874 | if (from->devid[d] > 0 && from->devstate[d] == 0) { |
0fa21e85 N |
875 | struct dev_policy *pol; |
876 | unsigned long long dev_size; | |
4b57ecf6 | 877 | unsigned int dev_sector_size; |
0fa21e85 | 878 | |
bfd76b93 CA |
879 | if (to->metadata->ss->external && |
880 | test_partition_from_id(from->devid[d])) | |
881 | continue; | |
882 | ||
fbfdcb06 | 883 | if (sc->min_size && |
0fa21e85 | 884 | dev_size_from_id(from->devid[d], &dev_size) && |
fbfdcb06 | 885 | dev_size < sc->min_size) |
0fa21e85 N |
886 | continue; |
887 | ||
4b57ecf6 AO |
888 | if (sc->sector_size && |
889 | dev_sector_size_from_id(from->devid[d], | |
890 | &dev_sector_size) && | |
891 | sc->sector_size != dev_sector_size) | |
892 | continue; | |
893 | ||
4dd2df09 | 894 | pol = devid_policy(from->devid[d]); |
0fa21e85 N |
895 | if (from->spare_group) |
896 | pol_add(&pol, pol_domain, | |
897 | from->spare_group, NULL); | |
f566ef45 JS |
898 | if (domain_test(domlist, pol, |
899 | to->metadata->ss->name) == 1) | |
0fa21e85 N |
900 | dev = from->devid[d]; |
901 | dev_policy_free(pol); | |
902 | } | |
903 | } | |
904 | return dev; | |
905 | } | |
906 | ||
0f0749ad | 907 | static dev_t container_choose_spare(struct state *from, struct state *to, |
f0b85306 | 908 | struct domainlist *domlist, |
fbfdcb06 | 909 | struct spare_criteria *sc, int active) |
5739e0d0 N |
910 | { |
911 | /* This is similar to choose_spare, but we cannot trust devstate, | |
912 | * so we need to read the metadata instead | |
913 | */ | |
326727d9 | 914 | struct mdinfo *list; |
5739e0d0 | 915 | struct supertype *st = from->metadata; |
ff044d6b | 916 | int fd = open(from->devname, O_RDONLY); |
5739e0d0 | 917 | int err; |
0f0749ad | 918 | dev_t dev = 0; |
5739e0d0 N |
919 | |
920 | if (fd < 0) | |
921 | return 0; | |
326727d9 AC |
922 | if (!st->ss->getinfo_super_disks) { |
923 | close(fd); | |
5739e0d0 | 924 | return 0; |
326727d9 | 925 | } |
1011e834 | 926 | |
5739e0d0 N |
927 | err = st->ss->load_container(st, fd, NULL); |
928 | close(fd); | |
929 | if (err) | |
930 | return 0; | |
1011e834 | 931 | |
a1e49d69 CA |
932 | if (from == to) { |
933 | /* We must check if number of active disks has not increased | |
934 | * since ioctl in main loop. mdmon may have added spare | |
935 | * to subarray. If so we do not need to look for more spares | |
936 | * so return non zero value */ | |
937 | int active_cnt = 0; | |
938 | struct mdinfo *dp; | |
939 | list = st->ss->getinfo_super_disks(st); | |
940 | if (!list) { | |
941 | st->ss->free_super(st); | |
942 | return 1; | |
943 | } | |
944 | dp = list->devs; | |
945 | while (dp) { | |
f566ef45 JS |
946 | if (dp->disk.state & (1 << MD_DISK_SYNC) && |
947 | !(dp->disk.state & (1 << MD_DISK_FAULTY))) | |
a1e49d69 CA |
948 | active_cnt++; |
949 | dp = dp->next; | |
950 | } | |
951 | sysfs_free(list); | |
952 | if (active < active_cnt) { | |
953 | /* Spare just activated.*/ | |
954 | st->ss->free_super(st); | |
955 | return 1; | |
956 | } | |
957 | } | |
958 | ||
326727d9 | 959 | /* We only need one spare so full list not needed */ |
fbfdcb06 | 960 | list = container_choose_spares(st, sc, domlist, from->spare_group, |
326727d9 AC |
961 | to->metadata->ss->name, 1); |
962 | if (list) { | |
963 | struct mdinfo *disks = list->devs; | |
964 | if (disks) | |
965 | dev = makedev(disks->disk.major, disks->disk.minor); | |
966 | sysfs_free(list); | |
5739e0d0 | 967 | } |
326727d9 | 968 | st->ss->free_super(st); |
5739e0d0 N |
969 | return dev; |
970 | } | |
971 | ||
e0bd6a96 | 972 | static void try_spare_migration(struct state *statelist, struct alert_info *info) |
2e0172b1 | 973 | { |
66f5c4b6 N |
974 | struct state *from; |
975 | struct state *st; | |
fbfdcb06 | 976 | struct spare_criteria sc; |
c3621c0a ML |
977 | |
978 | link_containers_with_subarrays(statelist); | |
66f5c4b6 | 979 | for (st = statelist; st; st = st->next) |
f566ef45 | 980 | if (st->active < st->raid && st->spare == 0 && !st->err) { |
e78dda3b N |
981 | struct domainlist *domlist = NULL; |
982 | int d; | |
66f5c4b6 N |
983 | struct state *to = st; |
984 | ||
4dd2df09 | 985 | if (to->parent_devnm[0] && !to->parent) |
c0dc0ad5 CA |
986 | /* subarray monitored without parent container |
987 | * we can't move spares here */ | |
988 | continue; | |
1011e834 | 989 | |
66f5c4b6 N |
990 | if (to->parent) |
991 | /* member of a container */ | |
992 | to = to->parent; | |
e78dda3b | 993 | |
fbfdcb06 | 994 | if (get_required_spare_criteria(to, &sc)) |
de697acc | 995 | continue; |
e9a2ac02 AC |
996 | if (to->metadata->ss->external) { |
997 | /* We must make sure there is | |
998 | * no suitable spare in container already. | |
999 | * If there is we don't add more */ | |
1000 | dev_t devid = container_choose_spare( | |
fbfdcb06 | 1001 | to, to, NULL, &sc, st->active); |
e9a2ac02 AC |
1002 | if (devid > 0) |
1003 | continue; | |
1004 | } | |
b0599bda | 1005 | for (d = 0; d < MAX_DISKS; d++) |
e78dda3b N |
1006 | if (to->devid[d]) |
1007 | domainlist_add_dev(&domlist, | |
1008 | to->devid[d], | |
1009 | to->metadata->ss->name); | |
1010 | if (to->spare_group) | |
1011 | domain_add(&domlist, to->spare_group); | |
5ec0f373 ML |
1012 | /* |
1013 | * No spare migration if the destination | |
1014 | * has no domain. Skip this array. | |
1015 | */ | |
1016 | if (!domlist) | |
1017 | continue; | |
0fa21e85 | 1018 | for (from=statelist ; from ; from=from->next) { |
0f0749ad | 1019 | dev_t devid; |
5ec0f373 | 1020 | if (!check_donor(from, to)) |
0fa21e85 | 1021 | continue; |
5739e0d0 N |
1022 | if (from->metadata->ss->external) |
1023 | devid = container_choose_spare( | |
fbfdcb06 | 1024 | from, to, domlist, &sc, 0); |
5739e0d0 | 1025 | else |
f0b85306 | 1026 | devid = choose_spare(from, to, domlist, |
fbfdcb06 | 1027 | &sc); |
f566ef45 JS |
1028 | if (devid > 0 && |
1029 | move_spare(from->devname, to->devname, | |
1030 | devid)) { | |
1031 | alert("MoveSpare", to->devname, | |
1032 | from->devname, info); | |
d52bb542 AC |
1033 | break; |
1034 | } | |
0fa21e85 | 1035 | } |
e78dda3b | 1036 | domain_free(domlist); |
2e0172b1 N |
1037 | } |
1038 | } | |
c3621c0a ML |
1039 | |
1040 | /* search the statelist to connect external | |
1041 | * metadata subarrays with their containers | |
1042 | * We always completely rebuild the tree from scratch as | |
1043 | * that is safest considering the possibility of entries | |
1044 | * disappearing or changing. | |
1045 | */ | |
1046 | static void link_containers_with_subarrays(struct state *list) | |
1047 | { | |
1048 | struct state *st; | |
1049 | struct state *cont; | |
1050 | for (st = list; st; st = st->next) { | |
1051 | st->parent = NULL; | |
1052 | st->subarray = NULL; | |
1053 | } | |
1054 | for (st = list; st; st = st->next) | |
4dd2df09 | 1055 | if (st->parent_devnm[0]) |
c3621c0a | 1056 | for (cont = list; cont; cont = cont->next) |
f566ef45 | 1057 | if (!cont->err && cont->parent_devnm[0] == 0 && |
4dd2df09 | 1058 | strcmp(cont->devnm, st->parent_devnm) == 0) { |
c3621c0a ML |
1059 | st->parent = cont; |
1060 | st->subarray = cont->subarray; | |
1061 | cont->subarray = st; | |
1062 | break; | |
1063 | } | |
1064 | } | |
1065 | ||
af3396da | 1066 | #ifndef NO_LIBUDEV |
49b69533 OS |
1067 | /* function: check_udev_activity |
1068 | * Description: Function waits for udev to finish | |
1069 | * events processing. | |
1070 | * Returns: | |
1071 | * 1 - detected error while opening udev | |
1072 | * 2 - timeout | |
1073 | * 0 - successfull completion | |
1074 | */ | |
1075 | static int check_udev_activity(void) | |
1076 | { | |
1077 | struct udev *udev = NULL; | |
1078 | struct udev_queue *udev_queue = NULL; | |
1079 | int timeout_cnt = 30; | |
1080 | int rc = 0; | |
1081 | ||
1082 | /* | |
1083 | * In rare cases systemd may not have udevm, | |
1084 | * in such cases just exit with rc 0 | |
1085 | */ | |
1086 | if (!use_udev()) | |
1087 | goto out; | |
1088 | ||
1089 | udev = udev_new(); | |
1090 | if (!udev) { | |
1091 | rc = 1; | |
1092 | goto out; | |
1093 | } | |
1094 | ||
1095 | udev_queue = udev_queue_new(udev); | |
1096 | if (!udev_queue) { | |
1097 | rc = 1; | |
1098 | goto out; | |
1099 | } | |
1100 | ||
1101 | if (udev_queue_get_queue_is_empty(udev_queue)) | |
1102 | goto out; | |
1103 | ||
1104 | while (!udev_queue_get_queue_is_empty(udev_queue)) { | |
1105 | sleep(1); | |
1106 | ||
1107 | if (timeout_cnt) | |
1108 | timeout_cnt--; | |
1109 | else { | |
1110 | rc = 2; | |
1111 | goto out; | |
1112 | } | |
1113 | } | |
1114 | ||
1115 | out: | |
1116 | if (udev_queue) | |
1117 | udev_queue_unref(udev_queue); | |
1118 | if (udev) | |
1119 | udev_unref(udev); | |
1120 | return rc; | |
1121 | } | |
af3396da | 1122 | #endif |
49b69533 | 1123 | |
b90c0e9a NB |
1124 | /* Not really Monitor but ... */ |
1125 | int Wait(char *dev) | |
1126 | { | |
4dd2df09 | 1127 | char devnm[32]; |
cb91230c TM |
1128 | dev_t rdev; |
1129 | char *tmp; | |
b90c0e9a | 1130 | int rv = 1; |
d3f6cf4f | 1131 | int frozen_remaining = 3; |
b90c0e9a | 1132 | |
cb91230c | 1133 | if (!stat_is_blkdev(dev, &rdev)) |
b90c0e9a | 1134 | return 2; |
cb91230c TM |
1135 | |
1136 | tmp = devid2devnm(rdev); | |
1137 | if (!tmp) { | |
1138 | pr_err("Cannot get md device name.\n"); | |
1139 | return 2; | |
1140 | } | |
1141 | ||
1142 | strcpy(devnm, tmp); | |
b90c0e9a NB |
1143 | |
1144 | while(1) { | |
1145 | struct mdstat_ent *ms = mdstat_read(1, 0); | |
1146 | struct mdstat_ent *e; | |
1147 | ||
f566ef45 | 1148 | for (e = ms; e; e = e->next) |
4dd2df09 | 1149 | if (strcmp(e->devnm, devnm) == 0) |
b90c0e9a NB |
1150 | break; |
1151 | ||
d3f6cf4f | 1152 | if (e && e->percent == RESYNC_NONE) { |
276be514 N |
1153 | /* We could be in the brief pause before something |
1154 | * starts. /proc/mdstat doesn't show that, but | |
1155 | * sync_action does. | |
1156 | */ | |
1157 | struct mdinfo mdi; | |
1158 | char buf[21]; | |
dae13137 JS |
1159 | |
1160 | if (sysfs_init(&mdi, -1, devnm)) | |
1161 | return 2; | |
276be514 N |
1162 | if (sysfs_get_str(&mdi, NULL, "sync_action", |
1163 | buf, 20) > 0 && | |
d3f6cf4f | 1164 | strcmp(buf,"idle\n") != 0) { |
276be514 | 1165 | e->percent = RESYNC_UNKNOWN; |
d3f6cf4f N |
1166 | if (strcmp(buf, "frozen\n") == 0) { |
1167 | if (frozen_remaining == 0) | |
1168 | e->percent = RESYNC_NONE; | |
1169 | else | |
1170 | frozen_remaining -= 1; | |
1171 | } | |
1172 | } | |
276be514 | 1173 | } |
639c3c10 | 1174 | if (!e || e->percent == RESYNC_NONE) { |
e7783ee6 | 1175 | if (e && e->metadata_version && |
c94709e8 DW |
1176 | strncmp(e->metadata_version, "external:", 9) == 0) { |
1177 | if (is_subarray(&e->metadata_version[9])) | |
1178 | ping_monitor(&e->metadata_version[9]); | |
1179 | else | |
4dd2df09 | 1180 | ping_monitor(devnm); |
c94709e8 | 1181 | } |
b90c0e9a NB |
1182 | free_mdstat(ms); |
1183 | return rv; | |
1184 | } | |
89a10d84 | 1185 | free_mdstat(ms); |
b90c0e9a NB |
1186 | rv = 0; |
1187 | mdstat_wait(5); | |
1188 | } | |
1189 | } | |
679eb882 | 1190 | |
43ebc910 GP |
1191 | /* The state "broken" is used only for RAID0/LINEAR - it's the same as |
1192 | * "clean", but used in case the array has one or more members missing. | |
1193 | */ | |
679eb882 | 1194 | static char *clean_states[] = { |
43ebc910 | 1195 | "clear", "inactive", "readonly", "read-auto", "clean", "broken", NULL }; |
679eb882 | 1196 | |
1ea04629 | 1197 | int WaitClean(char *dev, int verbose) |
679eb882 N |
1198 | { |
1199 | int fd; | |
1200 | struct mdinfo *mdi; | |
1201 | int rv = 1; | |
4dd2df09 | 1202 | char devnm[32]; |
679eb882 | 1203 | |
9e04ac1c ZL |
1204 | if (!stat_is_blkdev(dev, NULL)) |
1205 | return 2; | |
679eb882 N |
1206 | fd = open(dev, O_RDONLY); |
1207 | if (fd < 0) { | |
1208 | if (verbose) | |
e7b84f9d | 1209 | pr_err("Couldn't open %s: %s\n", dev, strerror(errno)); |
679eb882 N |
1210 | return 1; |
1211 | } | |
1212 | ||
4dd2df09 N |
1213 | strcpy(devnm, fd2devnm(fd)); |
1214 | mdi = sysfs_read(fd, devnm, GET_VERSION|GET_LEVEL|GET_SAFEMODE); | |
679eb882 N |
1215 | if (!mdi) { |
1216 | if (verbose) | |
7a862a02 | 1217 | pr_err("Failed to read sysfs attributes for %s\n", dev); |
679eb882 N |
1218 | close(fd); |
1219 | return 0; | |
1220 | } | |
1221 | ||
1222 | switch(mdi->array.level) { | |
1223 | case LEVEL_LINEAR: | |
1224 | case LEVEL_MULTIPATH: | |
1225 | case 0: | |
1226 | /* safemode delay is irrelevant for these levels */ | |
1227 | rv = 0; | |
1228 | } | |
1229 | ||
1230 | /* for internal metadata the kernel handles the final clean | |
1231 | * transition, containers can never be dirty | |
1232 | */ | |
1233 | if (!is_subarray(mdi->text_version)) | |
1234 | rv = 0; | |
1235 | ||
1236 | /* safemode disabled ? */ | |
1237 | if (mdi->safe_mode_delay == 0) | |
1238 | rv = 0; | |
1239 | ||
1240 | if (rv) { | |
4dd2df09 | 1241 | int state_fd = sysfs_open(fd2devnm(fd), NULL, "array_state"); |
679eb882 | 1242 | char buf[20]; |
efc67e8e | 1243 | int delay = 5000; |
679eb882 N |
1244 | |
1245 | /* minimize the safe_mode_delay and prepare to wait up to 5s | |
1246 | * for writes to quiesce | |
1247 | */ | |
1248 | sysfs_set_safemode(mdi, 1); | |
679eb882 N |
1249 | |
1250 | /* wait for array_state to be clean */ | |
1251 | while (1) { | |
1252 | rv = read(state_fd, buf, sizeof(buf)); | |
1253 | if (rv < 0) | |
1254 | break; | |
43ebc910 GP |
1255 | if (sysfs_match_word(buf, clean_states) < |
1256 | (int)ARRAY_SIZE(clean_states) - 1) | |
679eb882 | 1257 | break; |
efc67e8e | 1258 | rv = sysfs_wait(state_fd, &delay); |
679eb882 N |
1259 | if (rv < 0 && errno != EINTR) |
1260 | break; | |
1261 | lseek(state_fd, 0, SEEK_SET); | |
1262 | } | |
1263 | if (rv < 0) | |
1264 | rv = 1; | |
1ea04629 | 1265 | else if (ping_monitor(mdi->text_version) == 0) { |
679eb882 N |
1266 | /* we need to ping to close the window between array |
1267 | * state transitioning to clean and the metadata being | |
1268 | * marked clean | |
1269 | */ | |
1270 | rv = 0; | |
1ea04629 | 1271 | } else { |
679eb882 | 1272 | rv = 1; |
1ea04629 MT |
1273 | pr_err("Error connecting monitor with %s\n", dev); |
1274 | } | |
679eb882 | 1275 | if (rv && verbose) |
f566ef45 | 1276 | pr_err("Error waiting for %s to be clean\n", dev); |
679eb882 N |
1277 | |
1278 | /* restore the original safe_mode_delay */ | |
1279 | sysfs_set_safemode(mdi, mdi->safe_mode_delay); | |
1280 | close(state_fd); | |
1281 | } | |
1282 | ||
1283 | sysfs_free(mdi); | |
1284 | close(fd); | |
1285 | ||
1286 | return rv; | |
1287 | } |