]> git.ipfire.org Git - thirdparty/mdadm.git/blob - Monitor.c
projects / thirdparty / mdadm.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
config: support "ARRAY <ignore> ..." lines in mdadm.conf
[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", O_RDWR);
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;
169 if (mdlist->devname == NULL)
170 continue;
171 if (strcasecmp(mdlist->devname, "<ignore>") == 0)
172 continue;
173 st = malloc(sizeof *st);
174 if (st == NULL)
175 continue;
176 st->devname = strdup(mdlist->devname);
177 st->utime = 0;
178 st->next = statelist;
179 st->err = 0;
180 st->devnum = INT_MAX;
181 st->percent = -2;
182 st->expected_spares = mdlist->spare_disks;
183 if (mdlist->spare_group)
184 st->spare_group = strdup(mdlist->spare_group);
185 else
186 st->spare_group = NULL;
187 statelist = st;
188 }
189 } else {
190 mddev_dev_t dv;
191 for (dv=devlist ; dv; dv=dv->next) {
192 mddev_ident_t mdlist = conf_get_ident(dv->devname);
193 struct state *st = malloc(sizeof *st);
194 if (st == NULL)
195 continue;
196 st->devname = strdup(dv->devname);
197 st->utime = 0;
198 st->next = statelist;
199 st->err = 0;
200 st->devnum = INT_MAX;
201 st->percent = -2;
202 st->expected_spares = -1;
203 st->spare_group = NULL;
204 if (mdlist) {
205 st->expected_spares = mdlist->spare_disks;
206 if (mdlist->spare_group)
207 st->spare_group = strdup(mdlist->spare_group);
208 }
209 statelist = st;
210 }
211 }
212
213
214 while (! finished) {
215 int new_found = 0;
216 struct state *st;
217
218 if (mdstat)
219 free_mdstat(mdstat);
220 mdstat = mdstat_read(oneshot?0:1, 0);
221
222 for (st=statelist; st; st=st->next) {
223 struct { int state, major, minor; } info[MaxDisks];
224 mdu_array_info_t array;
225 struct mdstat_ent *mse = NULL, *mse2;
226 char *dev = st->devname;
227 int fd;
228 unsigned int i;
229
230 if (test)
231 alert("TestMessage", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
232 fd = open(dev, O_RDONLY);
233 if (fd < 0) {
234 if (!st->err)
235 alert("DeviceDisappeared", dev, NULL,
236 mailaddr, mailfrom, alert_cmd, dosyslog);
237 /* fprintf(stderr, Name ": cannot open %s: %s\n",
238 dev, strerror(errno));
239 */ st->err=1;
240 continue;
241 }
242 fcntl(fd, F_SETFD, FD_CLOEXEC);
243 if (ioctl(fd, GET_ARRAY_INFO, &array)<0) {
244 if (!st->err)
245 alert("DeviceDisappeared", dev, NULL,
246 mailaddr, mailfrom, alert_cmd, dosyslog);
247 /* fprintf(stderr, Name ": cannot get array info for %s: %s\n",
248 dev, strerror(errno));
249 */ st->err=1;
250 close(fd);
251 continue;
252 }
253 /* It's much easier to list what array levels can't
254 * have a device disappear than all of them that can
255 */
256 if (array.level == 0 || array.level == -1) {
257 if (!st->err)
258 alert("DeviceDisappeared", dev, "Wrong-Level",
259 mailaddr, mailfrom, alert_cmd, dosyslog);
260 st->err = 1;
261 close(fd);
262 continue;
263 }
264 if (st->devnum == INT_MAX) {
265 struct stat stb;
266 if (fstat(fd, &stb) == 0 &&
267 (S_IFMT&stb.st_mode)==S_IFBLK) {
268 if (major(stb.st_rdev) == MD_MAJOR)
269 st->devnum = minor(stb.st_rdev);
270 else
271 st->devnum = -1- (minor(stb.st_rdev)>>6);
272 }
273 }
274
275 for (mse2 = mdstat ; mse2 ; mse2=mse2->next)
276 if (mse2->devnum == st->devnum) {
277 mse2->devnum = INT_MAX; /* flag it as "used" */
278 mse = mse2;
279 }
280
281 if (st->utime == array.utime &&
282 st->failed == array.failed_disks &&
283 st->working == array.working_disks &&
284 st->spare == array.spare_disks &&
285 (mse == NULL || (
286 mse->percent == st->percent
287 ))) {
288 close(fd);
289 st->err = 0;
290 continue;
291 }
292 if (st->utime == 0 && /* new array */
293 mse && /* is in /proc/mdstat */
294 mse->pattern && strchr(mse->pattern, '_') /* degraded */
295 )
296 alert("DegradedArray", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
297
298 if (st->utime == 0 && /* new array */
299 st->expected_spares > 0 &&
300 array.spare_disks < st->expected_spares)
301 alert("SparesMissing", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
302 if (mse &&
303 st->percent == -1 &&
304 mse->percent >= 0)
305 alert("RebuildStarted", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
306 if (mse &&
307 st->percent >= 0 &&
308 mse->percent >= 0 &&
309 (mse->percent / 20) > (st->percent / 20))
310 alert(percentalerts[mse->percent/20],
311 dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
312
313 if (mse &&
314 mse->percent == -1 &&
315 st->percent >= 0) {
316 /* Rebuild/sync/whatever just finished.
317 * If there is a number in /mismatch_cnt,
318 * we should report that.
319 */
320 struct mdinfo *sra =
321 sysfs_read(-1, st->devnum, GET_MISMATCH);
322 if (sra && sra->mismatch_cnt > 0) {
323 char cnt[40];
324 sprintf(cnt, " mismatches found: %d", sra->mismatch_cnt);
325 alert("RebuildFinished", dev, cnt, mailaddr, mailfrom, alert_cmd, dosyslog);
326 } else
327 alert("RebuildFinished", dev, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
328 if (sra)
329 free(sra);
330 }
331
332 if (mse)
333 st->percent = mse->percent;
334
335
336 for (i=0; i<MaxDisks && i <= array.raid_disks + array.nr_disks;
337 i++) {
338 mdu_disk_info_t disc;
339 disc.number = i;
340 if (ioctl(fd, GET_DISK_INFO, &disc) >= 0) {
341 info[i].state = disc.state;
342 info[i].major = disc.major;
343 info[i].minor = disc.minor;
344 } else
345 info[i].major = info[i].minor = 0;
346 }
347 close(fd);
348
349 for (i=0; i<MaxDisks; i++) {
350 mdu_disk_info_t disc = {0};
351 int newstate=0;
352 int change;
353 char *dv = NULL;
354 disc.number = i;
355 if (i > array.raid_disks + array.nr_disks) {
356 newstate = 0;
357 disc.major = disc.minor = 0;
358 } else if (info[i].major || info[i].minor) {
359 newstate = info[i].state;
360 dv = map_dev(info[i].major, info[i].minor, 1);
361 disc.state = newstate;
362 disc.major = info[i].major;
363 disc.minor = info[i].minor;
364 } else if (mse && mse->pattern && i < strlen(mse->pattern)) {
365 switch(mse->pattern[i]) {
366 case 'U': newstate = 6 /* ACTIVE/SYNC */; break;
367 case '_': newstate = 0; break;
368 }
369 disc.major = disc.minor = 0;
370 }
371 if (dv == NULL && st->devid[i])
372 dv = map_dev(major(st->devid[i]),
373 minor(st->devid[i]), 1);
374 change = newstate ^ st->devstate[i];
375 if (st->utime && change && !st->err) {
376 if (i < (unsigned)array.raid_disks &&
377 (((newstate&change)&(1<<MD_DISK_FAULTY)) ||
378 ((st->devstate[i]&change)&(1<<MD_DISK_ACTIVE)) ||
379 ((st->devstate[i]&change)&(1<<MD_DISK_SYNC)))
380 )
381 alert("Fail", dev, dv, mailaddr, mailfrom, alert_cmd, dosyslog);
382 else if (i >= (unsigned)array.raid_disks &&
383 (disc.major || disc.minor) &&
384 st->devid[i] == makedev(disc.major, disc.minor) &&
385 ((newstate&change)&(1<<MD_DISK_FAULTY))
386 )
387 alert("FailSpare", dev, dv, mailaddr, mailfrom, alert_cmd, dosyslog);
388 else if (i < (unsigned)array.raid_disks &&
389 (((st->devstate[i]&change)&(1<<MD_DISK_FAULTY)) ||
390 ((newstate&change)&(1<<MD_DISK_ACTIVE)) ||
391 ((newstate&change)&(1<<MD_DISK_SYNC)))
392 )
393 alert("SpareActive", dev, dv, mailaddr, mailfrom, alert_cmd, dosyslog);
394 }
395 st->devstate[i] = newstate;
396 st->devid[i] = makedev(disc.major, disc.minor);
397 }
398 st->active = array.active_disks;
399 st->working = array.working_disks;
400 st->spare = array.spare_disks;
401 st->failed = array.failed_disks;
402 st->utime = array.utime;
403 st->raid = array.raid_disks;
404 st->err = 0;
405 }
406 /* now check if there are any new devices found in mdstat */
407 if (scan) {
408 struct mdstat_ent *mse;
409 for (mse=mdstat; mse; mse=mse->next)
410 if (mse->devnum != INT_MAX &&
411 mse->level &&
412 (strcmp(mse->level, "raid0")!=0 &&
413 strcmp(mse->level, "linear")!=0)
414 ) {
415 struct state *st = malloc(sizeof *st);
416 mdu_array_info_t array;
417 int fd;
418 if (st == NULL)
419 continue;
420 st->devname = strdup(get_md_name(mse->devnum));
421 if ((fd = open(st->devname, O_RDONLY)) < 0 ||
422 ioctl(fd, GET_ARRAY_INFO, &array)< 0) {
423 /* no such array */
424 if (fd >=0) close(fd);
425 put_md_name(st->devname);
426 free(st->devname);
427 free(st);
428 continue;
429 }
430 close(fd);
431 st->utime = 0;
432 st->next = statelist;
433 st->err = 1;
434 st->devnum = mse->devnum;
435 st->percent = -2;
436 st->spare_group = NULL;
437 st->expected_spares = -1;
438 statelist = st;
439 if (test)
440 alert("TestMessage", st->devname, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
441 alert("NewArray", st->devname, NULL, mailaddr, mailfrom, alert_cmd, dosyslog);
442 new_found = 1;
443 }
444 }
445 /* If an array has active < raid && spare == 0 && spare_group != NULL
446 * Look for another array with spare > 0 and active == raid and same spare_group
447 * if found, choose a device and hotremove/hotadd
448 */
449 for (st = statelist; st; st=st->next)
450 if (st->active < st->raid &&
451 st->spare == 0 &&
452 st->spare_group != NULL) {
453 struct state *st2;
454 for (st2=statelist ; st2 ; st2=st2->next)
455 if (st2 != st &&
456 st2->spare > 0 &&
457 st2->active == st2->raid &&
458 st2->spare_group != NULL &&
459 strcmp(st->spare_group, st2->spare_group) == 0) {
460 /* try to remove and add */
461 int fd1 = open(st->devname, O_RDONLY);
462 int fd2 = open(st2->devname, O_RDONLY);
463 int dev = -1;
464 int d;
465 if (fd1 < 0 || fd2 < 0) {
466 if (fd1>=0) close(fd1);
467 if (fd2>=0) close(fd2);
468 continue;
469 }
470 for (d=st2->raid; d < MaxDisks; d++) {
471 if (st2->devid[d] > 0 &&
472 st2->devstate[d] == 0) {
473 dev = st2->devid[d];
474 break;
475 }
476 }
477 if (dev > 0) {
478 if (ioctl(fd2, HOT_REMOVE_DISK,
479 (unsigned long)dev) == 0) {
480 if (ioctl(fd1, HOT_ADD_DISK,
481 (unsigned long)dev) == 0) {
482 alert("MoveSpare", st->devname, st2->devname, mailaddr, mailfrom, alert_cmd, dosyslog);
483 close(fd1);
484 close(fd2);
485 break;
486 }
487 else ioctl(fd2, HOT_ADD_DISK, (unsigned long) dev);
488 }
489 }
490 close(fd1);
491 close(fd2);
492 }
493 }
494 if (!new_found) {
495 if (oneshot)
496 break;
497 else
498 mdstat_wait(period);
499 }
500 test = 0;
501 }
502 if (pidfile)
503 unlink(pidfile);
504 return 0;
505 }
506
507
508 static void alert(char *event, char *dev, char *disc, char *mailaddr, char *mailfrom, char *cmd,
509 int dosyslog)
510 {
511 int priority;
512
513 if (!cmd && !mailaddr) {
514 time_t now = time(0);
515
516 printf("%1.15s: %s on %s %s\n", ctime(&now)+4, event, dev, disc?disc:"unknown device");
517 }
518 if (cmd) {
519 int pid = fork();
520 switch(pid) {
521 default:
522 waitpid(pid, NULL, 0);
523 break;
524 case -1:
525 break;
526 case 0:
527 execl(cmd, cmd, event, dev, disc, NULL);
528 exit(2);
529 }
530 }
531 if (mailaddr &&
532 (strncmp(event, "Fail", 4)==0 ||
533 strncmp(event, "Test", 4)==0 ||
534 strncmp(event, "Spares", 6)==0 ||
535 strncmp(event, "Degrade", 7)==0)) {
536 FILE *mp = popen(Sendmail, "w");
537 if (mp) {
538 FILE *mdstat;
539 char hname[256];
540 gethostname(hname, sizeof(hname));
541 signal(SIGPIPE, SIG_IGN);
542 if (mailfrom)
543 fprintf(mp, "From: %s\n", mailfrom);
544 else
545 fprintf(mp, "From: " Name " monitoring <root>\n");
546 fprintf(mp, "To: %s\n", mailaddr);
547 fprintf(mp, "Subject: %s event on %s:%s\n\n", event, dev, hname);
548
549 fprintf(mp, "This is an automatically generated mail message from " Name "\n");
550 fprintf(mp, "running on %s\n\n", hname);
551
552 fprintf(mp, "A %s event had been detected on md device %s.\n\n", event, dev);
553
554 if (disc && disc[0] != ' ')
555 fprintf(mp, "It could be related to component device %s.\n\n", disc);
556 if (disc && disc[0] == ' ')
557 fprintf(mp, "Extra information:%s.\n\n", disc);
558
559 fprintf(mp, "Faithfully yours, etc.\n");
560
561 mdstat = fopen("/proc/mdstat", "r");
562 if (mdstat) {
563 char buf[8192];
564 int n;
565 fprintf(mp, "\nP.S. The /proc/mdstat file currently contains the following:\n\n");
566 while ( (n=fread(buf, 1, sizeof(buf), mdstat)) > 0)
567 n=fwrite(buf, 1, n, mp); /* yes, i don't care about the result */
568 fclose(mdstat);
569 }
570 fclose(mp);
571 }
572
573 }
574
575 /* log the event to syslog maybe */
576 if (dosyslog) {
577 /* Log at a different severity depending on the event.
578 *
579 * These are the critical events: */
580 if (strncmp(event, "Fail", 4)==0 ||
581 strncmp(event, "Degrade", 7)==0 ||
582 strncmp(event, "DeviceDisappeared", 17)==0)
583 priority = LOG_CRIT;
584 /* Good to know about, but are not failures: */
585 else if (strncmp(event, "Rebuild", 7)==0 ||
586 strncmp(event, "MoveSpare", 9)==0 ||
587 strncmp(event, "Spares", 6) != 0)
588 priority = LOG_WARNING;
589 /* Everything else: */
590 else
591 priority = LOG_INFO;
592
593 if (disc)
594 syslog(priority, "%s event detected on md device %s, component device %s", event, dev, disc);
595 else
596 syslog(priority, "%s event detected on md device %s", event, dev);
597 }
598 }
599
600 /* Not really Monitor but ... */
601 int Wait(char *dev)
602 {
603 struct stat stb;
604 int devnum;
605 int rv = 1;
606
607 if (stat(dev, &stb) != 0) {
608 fprintf(stderr, Name ": Cannot find %s: %s\n", dev,
609 strerror(errno));
610 return 2;
611 }
612 devnum = stat2devnum(&stb);
613
614 while(1) {
615 struct mdstat_ent *ms = mdstat_read(1, 0);
616 struct mdstat_ent *e;
617
618 for (e=ms ; e; e=e->next)
619 if (e->devnum == devnum)
620 break;
621
622 if (!e || e->percent < 0) {
623 if (e && e->metadata_version &&
624 strncmp(e->metadata_version, "external:", 9) == 0) {
625 if (is_subarray(&e->metadata_version[9]))
626 ping_monitor(&e->metadata_version[9]);
627 else
628 ping_monitor(devnum2devname(devnum));
629 }
630 free_mdstat(ms);
631 return rv;
632 }
633 free_mdstat(ms);
634 rv = 0;
635 mdstat_wait(5);
636 }
637 }
638
639 static char *clean_states[] = {
640 "clear", "inactive", "readonly", "read-auto", "clean", NULL };
641
642 int WaitClean(char *dev, int verbose)
643 {
644 int fd;
645 struct mdinfo *mdi;
646 int rv = 1;
647 int devnum;
648
649 fd = open(dev, O_RDONLY);
650 if (fd < 0) {
651 if (verbose)
652 fprintf(stderr, Name ": Couldn't open %s: %s\n", dev, strerror(errno));
653 return 1;
654 }
655
656 devnum = fd2devnum(fd);
657 mdi = sysfs_read(fd, devnum, GET_VERSION|GET_LEVEL|GET_SAFEMODE);
658 if (!mdi) {
659 if (verbose)
660 fprintf(stderr, Name ": Failed to read sysfs attributes for "
661 "%s\n", dev);
662 close(fd);
663 return 0;
664 }
665
666 switch(mdi->array.level) {
667 case LEVEL_LINEAR:
668 case LEVEL_MULTIPATH:
669 case 0:
670 /* safemode delay is irrelevant for these levels */
671 rv = 0;
672
673 }
674
675 /* for internal metadata the kernel handles the final clean
676 * transition, containers can never be dirty
677 */
678 if (!is_subarray(mdi->text_version))
679 rv = 0;
680
681 /* safemode disabled ? */
682 if (mdi->safe_mode_delay == 0)
683 rv = 0;
684
685 if (rv) {
686 int state_fd = sysfs_open(fd2devnum(fd), NULL, "array_state");
687 char buf[20];
688 fd_set fds;
689 struct timeval tm;
690
691 /* minimize the safe_mode_delay and prepare to wait up to 5s
692 * for writes to quiesce
693 */
694 sysfs_set_safemode(mdi, 1);
695 tm.tv_sec = 5;
696 tm.tv_usec = 0;
697
698 /* give mdmon a chance to checkpoint resync */
699 sysfs_set_str(mdi, NULL, "sync_action", "idle");
700
701 FD_ZERO(&fds);
702
703 /* wait for array_state to be clean */
704 while (1) {
705 rv = read(state_fd, buf, sizeof(buf));
706 if (rv < 0)
707 break;
708 if (sysfs_match_word(buf, clean_states) <= 4)
709 break;
710 FD_SET(state_fd, &fds);
711 rv = select(state_fd + 1, NULL, NULL, &fds, &tm);
712 if (rv < 0 && errno != EINTR)
713 break;
714 lseek(state_fd, 0, SEEK_SET);
715 }
716 if (rv < 0)
717 rv = 1;
718 else if (ping_monitor(mdi->text_version) == 0) {
719 /* we need to ping to close the window between array
720 * state transitioning to clean and the metadata being
721 * marked clean
722 */
723 rv = 0;
724 } else
725 rv = 1;
726 if (rv && verbose)
727 fprintf(stderr, Name ": Error waiting for %s to be clean\n",
728 dev);
729
730 /* restore the original safe_mode_delay */
731 sysfs_set_safemode(mdi, mdi->safe_mode_delay);
732 close(state_fd);
733 }
734
735 sysfs_free(mdi);
736 close(fd);
737
738 return rv;
739 }
740
741
Unnamed repository; edit this file 'description' to name the repository.