]> git.ipfire.org Git - thirdparty/mdadm.git/blob - Monitor.c
Fix for segfault when reading /proc/mdstat
[thirdparty/mdadm.git] / Monitor.c
1 /*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
4 * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
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
22 * Email: <neilb@cse.unsw.edu.au>
23 * Paper: Neil Brown
24 * School of Computer Science and Engineering
25 * The University of New South Wales
26 * Sydney, 2052
27 * Australia
28 */
29
30 #include "mdadm.h"
31 #include "md_p.h"
32 #include "md_u.h"
33 #include <sys/wait.h>
34 #include <signal.h>
35 #include <limits.h>
36 #include <syslog.h>
37
38 static void alert(char *event, char *dev, char *disc, char *mailaddr, char *mailfrom,
39 char *cmd, int dosyslog);
40
41 static char *percentalerts[] = {
42 "RebuildStarted",
43 "Rebuild20",
44 "Rebuild40",
45 "Rebuild60",
46 "Rebuild80",
47 };
48
49 /* The largest number of disks current arrays can manage is 384
50 * This really should be dynamically, but that will have to wait
51 * At least it isn't MD_SB_DISKS.
52 */
53 #define MaxDisks 384
54 int Monitor(mddev_dev_t devlist,
55 char *mailaddr, char *alert_cmd,
56 int period, int daemonise, int scan, int oneshot,
57 int dosyslog, int test, char* pidfile)
58 {
59 /*
60 * Every few seconds, scan every md device looking for changes
61 * When a change is found, log it, possibly run the alert command,
62 * and possibly send Email
63 *
64 * For each array, we record:
65 * Update time
66 * active/working/failed/spare drives
67 * State of each device.
68 * %rebuilt if rebuilding
69 *
70 * If the update time changes, check out all the data again
71 * It is possible that we cannot get the state of each device
72 * due to bugs in the md kernel module.
73 * We also read /proc/mdstat to get rebuild percent,
74 * and to get state on all active devices incase of kernel bug.
75 *
76 * Events are:
77 * Fail
78 * An active device had Faulty set or Active/Sync removed
79 * FailSpare
80 * A spare device had Faulty set
81 * SpareActive
82 * An active device had a reverse transition
83 * RebuildStarted
84 * percent went from -1 to +ve
85 * Rebuild20 Rebuild40 Rebuild60 Rebuild80
86 * percent went from below to not-below that number
87 * DeviceDisappeared
88 * Couldn't access a device which was previously visible
89 *
90 * if we detect an array with active<raid and spare==0
91 * we look at other arrays that have same spare-group
92 * If we find one with active==raid and spare>0,
93 * and if we can get_disk_info and find a name
94 * Then we hot-remove and hot-add to the other array
95 *
96 * If devlist is NULL, then we can monitor everything because --scan
97 * was given. We get an initial list from config file and add anything
98 * that appears in /proc/mdstat
99 */
100
101 struct state {
102 char *devname;
103 int devnum; /* to sync with mdstat info */
104 long utime;
105 int err;
106 char *spare_group;
107 int active, working, failed, spare, raid;
108 int expected_spares;
109 int devstate[MaxDisks];
110 int devid[MaxDisks];
111 int percent;
112 struct state *next;
113 } *statelist = NULL;
114 int finished = 0;
115 struct mdstat_ent *mdstat = NULL;
116 char *mailfrom = NULL;
117
118 if (!mailaddr) {
119 mailaddr = conf_get_mailaddr();
120 if (mailaddr && ! scan)
121 fprintf(stderr, Name ": Monitor using email address \"%s\" from config file\n",
122 mailaddr);
123 }
124 mailfrom = conf_get_mailfrom();
125
126 if (!alert_cmd) {
127 alert_cmd = conf_get_program();
128 if (alert_cmd && ! scan)
129 fprintf(stderr, Name ": Monitor using program \"%s\" from config file\n",
130 alert_cmd);
131 }
132 if (scan && !mailaddr && !alert_cmd) {
133 fprintf(stderr, Name ": No mail address or alert command - not monitoring.\n");
134 return 1;
135 }
136
137 if (daemonise) {
138 int pid = fork();
139 if (pid > 0) {
140 if (!pidfile)
141 printf("%d\n", pid);
142 else {
143 FILE *pid_file;
144 pid_file=fopen(pidfile, "w");
145 if (!pid_file)
146 perror("cannot create pid file");
147 else {
148 fprintf(pid_file,"%d\n", pid);
149 fclose(pid_file);
150 }
151 }
152 return 0;
153 }
154 if (pid < 0) {
155 perror("daemonise");
156 return 1;
157 }
158 close(0);
159 open("/dev/null", 3);
160 dup2(0,1);
161 dup2(0,2);
162 setsid();
163 }
164
165 if (devlist == NULL) {
166 mddev_ident_t mdlist = conf_get_ident(NULL);
167 for (; mdlist; mdlist=mdlist->next) {
168 struct state *st = malloc(sizeof *st);
169 if (st == NULL)
170 continue;
171 st->devname = strdup(mdlist->devname);
172 st->utime = 0;
173 st->next = statelist;
174 st->err = 0;
175 st->devnum = INT_MAX;
176 st->percent = -2;
177 st->expected_spares = mdlist->spare_disks;
178 if (mdlist->spare_group)
179 st->spare_group = strdup(mdlist->spare_group);
180 else
181 st->spare_group = NULL;
182 statelist = st;
183 }
184 } else {
185 mddev_dev_t dv;
186 for (dv=devlist ; dv; dv=dv->next) {
187 mddev_ident_t mdlist = conf_get_ident(dv->devname);
188 struct state *st = malloc(sizeof *st);
189 if (st == NULL)
190 continue;
191 st->devname = strdup(dv->devname);
192 st->utime = 0;
193 st->next = statelist;
194 st->err = 0;
195 st->devnum = INT_MAX;
196 st->percent = -2;
197 st->expected_spares = -1;
198 st->spare_group = NULL;
199 if (mdlist) {
200 st->expected_spares = mdlist->spare_disks;
201 if (mdlist->spare_group)
202 st->spare_group = strdup(mdlist->spare_group);
203 }
204 statelist = st;
205 }
206 }
207
208
209 while (! finished) {
210 int new_found = 0;
211 struct state *st;
212
213 if (mdstat)
214 free_mdstat(mdstat);
215 mdstat = mdstat_read(oneshot?0:1, 0);
216
217 for (st=statelist; st; st=st->next) {
218 struct { int state, major, minor; } info[MaxDisks];
219 mdu_array_info_t array;
220 struct mdstat_ent *mse = NULL, *mse2;
221 char *dev = st->devname;
222 int fd;
223 unsigned int i;
224
225 if (test)
226 alert("TestMessage", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
227 fd = open(dev, O_RDONLY);
228 if (fd < 0) {
229 if (!st->err)
230 alert("DeviceDisappeared", dev, NULL,
231 mailaddr, mailfrom, alert_cmd, dosyslog);
232 /* fprintf(stderr, Name ": cannot open %s: %s\n",
233 dev, strerror(errno));
234 */ st->err=1;
235 continue;
236 }
237 fcntl(fd, F_SETFD, FD_CLOEXEC);
238 if (ioctl(fd, GET_ARRAY_INFO, &array)<0) {
239 if (!st->err)
240 alert("DeviceDisappeared", dev, NULL,
241 mailaddr, mailfrom, alert_cmd, dosyslog);
242 /* fprintf(stderr, Name ": cannot get array info for %s: %s\n",
243 dev, strerror(errno));
244 */ st->err=1;
245 close(fd);
246 continue;
247 }
248 /* It's much easier to list what array levels can't
249 * have a device disappear than all of them that can
250 */
251 if (array.level == 0 || array.level == -1) {
252 if (!st->err)
253 alert("DeviceDisappeared", dev, "Wrong-Level",
254 mailaddr, mailfrom, alert_cmd, dosyslog);
255 st->err = 1;
256 close(fd);
257 continue;
258 }
259 if (st->devnum == INT_MAX) {
260 struct stat stb;
261 if (fstat(fd, &stb) == 0 &&
262 (S_IFMT&stb.st_mode)==S_IFBLK) {
263 if (major(stb.st_rdev) == MD_MAJOR)
264 st->devnum = minor(stb.st_rdev);
265 else
266 st->devnum = -1- (minor(stb.st_rdev)>>6);
267 }
268 }
269
270 for (mse2 = mdstat ; mse2 ; mse2=mse2->next)
271 if (mse2->devnum == st->devnum) {
272 mse2->devnum = INT_MAX; /* flag it as "used" */
273 mse = mse2;
274 }
275
276 if (st->utime == array.utime &&
277 st->failed == array.failed_disks &&
278 st->working == array.working_disks &&
279 st->spare == array.spare_disks &&
280 (mse == NULL || (
281 mse->percent == st->percent
282 ))) {
283 close(fd);
284 st->err = 0;
285 continue;
286 }
287 if (st->utime == 0 && /* new array */
288 mse && /* is in /proc/mdstat */
289 mse->pattern && strchr(mse->pattern, '_') /* degraded */
290 )
291 alert("DegradedArray", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
292
293 if (st->utime == 0 && /* new array */
294 st->expected_spares > 0 &&
295 array.spare_disks < st->expected_spares)
296 alert("SparesMissing", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
297 if (mse &&
298 st->percent == -1 &&
299 mse->percent >= 0)
300 alert("RebuildStarted", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
301 if (mse &&
302 st->percent >= 0 &&
303 mse->percent >= 0 &&
304 (mse->percent / 20) > (st->percent / 20))
305 alert(percentalerts[mse->percent/20],
306 dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
307
308 if (mse &&
309 mse->percent == -1 &&
310 st->percent >= 0) {
311 /* Rebuild/sync/whatever just finished.
312 * If there is a number in /mismatch_cnt,
313 * we should report that.
314 */
315 struct mdinfo *sra =
316 sysfs_read(-1, st->devnum, GET_MISMATCH);
317 if (sra && sra->mismatch_cnt > 0) {
318 char cnt[40];
319 sprintf(cnt, " mismatches found: %d", sra->mismatch_cnt);
320 alert("RebuildFinished", dev, cnt, mailaddr, mailfrom, alert_cmd, dosyslog);
321 } else
322 alert("RebuildFinished", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
323 if (sra)
324 free(sra);
325 }
326
327 if (mse)
328 st->percent = mse->percent;
329
330
331 for (i=0; i<MaxDisks && i <= array.raid_disks + array.nr_disks;
332 i++) {
333 mdu_disk_info_t disc;
334 disc.number = i;
335 if (ioctl(fd, GET_DISK_INFO, &disc) >= 0) {
336 info[i].state = disc.state;
337 info[i].major = disc.major;
338 info[i].minor = disc.minor;
339 } else
340 info[i].major = info[i].minor = 0;
341 }
342 close(fd);
343
344 for (i=0; i<MaxDisks; i++) {
345 mdu_disk_info_t disc = {0};
346 int newstate=0;
347 int change;
348 char *dv = NULL;
349 disc.number = i;
350 if (i > array.raid_disks + array.nr_disks) {
351 newstate = 0;
352 disc.major = disc.minor = 0;
353 } else if (info[i].major || info[i].minor) {
354 newstate = info[i].state;
355 dv = map_dev(info[i].major, info[i].minor, 1);
356 disc.state = newstate;
357 disc.major = info[i].major;
358 disc.minor = info[i].minor;
359 } else if (mse && mse->pattern && i < strlen(mse->pattern)) {
360 switch(mse->pattern[i]) {
361 case 'U': newstate = 6 /* ACTIVE/SYNC */; break;
362 case '_': newstate = 0; break;
363 }
364 disc.major = disc.minor = 0;
365 }
366 if (dv == NULL && st->devid[i])
367 dv = map_dev(major(st->devid[i]),
368 minor(st->devid[i]), 1);
369 change = newstate ^ st->devstate[i];
370 if (st->utime && change && !st->err) {
371 if (i < (unsigned)array.raid_disks &&
372 (((newstate&change)&(1<<MD_DISK_FAULTY)) ||
373 ((st->devstate[i]&change)&(1<<MD_DISK_ACTIVE)) ||
374 ((st->devstate[i]&change)&(1<<MD_DISK_SYNC)))
375 )
376 alert("Fail", dev, dv, mailaddr, mailfrom, alert_cmd, dosyslog);
377 else if (i >= (unsigned)array.raid_disks &&
378 (disc.major || disc.minor) &&
379 st->devid[i] == makedev(disc.major, disc.minor) &&
380 ((newstate&change)&(1<<MD_DISK_FAULTY))
381 )
382 alert("FailSpare", dev, dv, mailaddr, mailfrom, alert_cmd, dosyslog);
383 else if (i < (unsigned)array.raid_disks &&
384 (((st->devstate[i]&change)&(1<<MD_DISK_FAULTY)) ||
385 ((newstate&change)&(1<<MD_DISK_ACTIVE)) ||
386 ((newstate&change)&(1<<MD_DISK_SYNC)))
387 )
388 alert("SpareActive", dev, dv, mailaddr, mailfrom, alert_cmd, dosyslog);
389 }
390 st->devstate[i] = newstate;
391 st->devid[i] = makedev(disc.major, disc.minor);
392 }
393 st->active = array.active_disks;
394 st->working = array.working_disks;
395 st->spare = array.spare_disks;
396 st->failed = array.failed_disks;
397 st->utime = array.utime;
398 st->raid = array.raid_disks;
399 st->err = 0;
400 }
401 /* now check if there are any new devices found in mdstat */
402 if (scan) {
403 struct mdstat_ent *mse;
404 for (mse=mdstat; mse; mse=mse->next)
405 if (mse->devnum != INT_MAX &&
406 mse->level &&
407 (strcmp(mse->level, "raid0")!=0 &&
408 strcmp(mse->level, "linear")!=0)
409 ) {
410 struct state *st = malloc(sizeof *st);
411 mdu_array_info_t array;
412 int fd;
413 if (st == NULL)
414 continue;
415 st->devname = strdup(get_md_name(mse->devnum));
416 if ((fd = open(st->devname, O_RDONLY)) < 0 ||
417 ioctl(fd, GET_ARRAY_INFO, &array)< 0) {
418 /* no such array */
419 if (fd >=0) close(fd);
420 put_md_name(st->devname);
421 free(st->devname);
422 free(st);
423 continue;
424 }
425 close(fd);
426 st->utime = 0;
427 st->next = statelist;
428 st->err = 1;
429 st->devnum = mse->devnum;
430 st->percent = -2;
431 st->spare_group = NULL;
432 st->expected_spares = -1;
433 statelist = st;
434 alert("NewArray", st->devname, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
435 new_found = 1;
436 }
437 }
438 /* If an array has active < raid && spare == 0 && spare_group != NULL
439 * Look for another array with spare > 0 and active == raid and same spare_group
440 * if found, choose a device and hotremove/hotadd
441 */
442 for (st = statelist; st; st=st->next)
443 if (st->active < st->raid &&
444 st->spare == 0 &&
445 st->spare_group != NULL) {
446 struct state *st2;
447 for (st2=statelist ; st2 ; st2=st2->next)
448 if (st2 != st &&
449 st2->spare > 0 &&
450 st2->active == st2->raid &&
451 st2->spare_group != NULL &&
452 strcmp(st->spare_group, st2->spare_group) == 0) {
453 /* try to remove and add */
454 int fd1 = open(st->devname, O_RDONLY);
455 int fd2 = open(st2->devname, O_RDONLY);
456 int dev = -1;
457 int d;
458 if (fd1 < 0 || fd2 < 0) {
459 if (fd1>=0) close(fd1);
460 if (fd2>=0) close(fd2);
461 continue;
462 }
463 for (d=st2->raid; d < MaxDisks; d++) {
464 if (st2->devid[d] > 0 &&
465 st2->devstate[d] == 0) {
466 dev = st2->devid[d];
467 break;
468 }
469 }
470 if (dev > 0) {
471 if (ioctl(fd2, HOT_REMOVE_DISK,
472 (unsigned long)dev) == 0) {
473 if (ioctl(fd1, HOT_ADD_DISK,
474 (unsigned long)dev) == 0) {
475 alert("MoveSpare", st->devname, st2->devname, mailaddr, mailfrom, alert_cmd, dosyslog);
476 close(fd1);
477 close(fd2);
478 break;
479 }
480 else ioctl(fd2, HOT_ADD_DISK, (unsigned long) dev);
481 }
482 }
483 close(fd1);
484 close(fd2);
485 }
486 }
487 if (!new_found) {
488 if (oneshot)
489 break;
490 else
491 mdstat_wait(period);
492 }
493 test = 0;
494 }
495 if (pidfile)
496 unlink(pidfile);
497 return 0;
498 }
499
500
501 static void alert(char *event, char *dev, char *disc, char *mailaddr, char *mailfrom, char *cmd,
502 int dosyslog)
503 {
504 int priority;
505
506 if (!cmd && !mailaddr) {
507 time_t now = time(0);
508
509 printf("%1.15s: %s on %s %s\n", ctime(&now)+4, event, dev, disc?disc:"unknown device");
510 }
511 if (cmd) {
512 int pid = fork();
513 switch(pid) {
514 default:
515 waitpid(pid, NULL, 0);
516 break;
517 case -1:
518 break;
519 case 0:
520 execl(cmd, cmd, event, dev, disc, NULL);
521 exit(2);
522 }
523 }
524 if (mailaddr &&
525 (strncmp(event, "Fail", 4)==0 ||
526 strncmp(event, "Test", 4)==0 ||
527 strncmp(event, "Spares", 6)==0 ||
528 strncmp(event, "Degrade", 7)==0)) {
529 FILE *mp = popen(Sendmail, "w");
530 if (mp) {
531 FILE *mdstat;
532 char hname[256];
533 gethostname(hname, sizeof(hname));
534 signal(SIGPIPE, SIG_IGN);
535 if (mailfrom)
536 fprintf(mp, "From: %s\n", mailfrom);
537 else
538 fprintf(mp, "From: " Name " monitoring <root>\n");
539 fprintf(mp, "To: %s\n", mailaddr);
540 fprintf(mp, "Subject: %s event on %s:%s\n\n", event, dev, hname);
541
542 fprintf(mp, "This is an automatically generated mail message from " Name "\n");
543 fprintf(mp, "running on %s\n\n", hname);
544
545 fprintf(mp, "A %s event had been detected on md device %s.\n\n", event, dev);
546
547 if (disc && disc[0] != ' ')
548 fprintf(mp, "It could be related to component device %s.\n\n", disc);
549 if (disc && disc[0] == ' ')
550 fprintf(mp, "Extra information:%s.\n\n", disc);
551
552 fprintf(mp, "Faithfully yours, etc.\n");
553
554 mdstat = fopen("/proc/mdstat", "r");
555 if (mdstat) {
556 char buf[8192];
557 int n;
558 fprintf(mp, "\nP.S. The /proc/mdstat file currently contains the following:\n\n");
559 while ( (n=fread(buf, 1, sizeof(buf), mdstat)) > 0)
560 n=fwrite(buf, 1, n, mp); /* yes, i don't care about the result */
561 fclose(mdstat);
562 }
563 fclose(mp);
564 }
565
566 }
567
568 /* log the event to syslog maybe */
569 if (dosyslog) {
570 /* Log at a different severity depending on the event.
571 *
572 * These are the critical events: */
573 if (strncmp(event, "Fail", 4)==0 ||
574 strncmp(event, "Degrade", 7)==0 ||
575 strncmp(event, "DeviceDisappeared", 17)==0)
576 priority = LOG_CRIT;
577 /* Good to know about, but are not failures: */
578 else if (strncmp(event, "Rebuild", 7)==0 ||
579 strncmp(event, "MoveSpare", 9)==0 ||
580 strncmp(event, "Spares", 6) != 0)
581 priority = LOG_WARNING;
582 /* Everything else: */
583 else
584 priority = LOG_INFO;
585
586 if (disc)
587 syslog(priority, "%s event detected on md device %s, component device %s", event, dev, disc);
588 else
589 syslog(priority, "%s event detected on md device %s", event, dev);
590 }
591 }
592
593 /* Not really Monitor but ... */
594 int Wait(char *dev)
595 {
596 struct stat stb;
597 int devnum;
598 int rv = 1;
599
600 if (stat(dev, &stb) != 0) {
601 fprintf(stderr, Name ": Cannot find %s: %s\n", dev,
602 strerror(errno));
603 return 2;
604 }
605 if (major(stb.st_rdev) == MD_MAJOR)
606 devnum = minor(stb.st_rdev);
607 else
608 devnum = -1-(minor(stb.st_rdev)/64);
609
610 while(1) {
611 struct mdstat_ent *ms = mdstat_read(1, 0);
612 struct mdstat_ent *e;
613
614 for (e=ms ; e; e=e->next)
615 if (e->devnum == devnum)
616 break;
617
618 if (!e || e->percent < 0) {
619 free_mdstat(ms);
620 return rv;
621 }
622 free(ms);
623 rv = 0;
624 mdstat_wait(5);
625 }
626 }