]>
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 | |
e27d562b NB |
33 | /* The largest number of disks current arrays can manage is 384 |
34 | * This really should be dynamically, but that will have to wait | |
35 | * At least it isn't MD_SB_DISKS. | |
36 | */ | |
37 | #define MaxDisks 384 | |
2e0172b1 N |
38 | struct state { |
39 | char *devname; | |
40 | int devnum; /* to sync with mdstat info */ | |
41 | long utime; | |
42 | int err; | |
43 | char *spare_group; | |
44 | int active, working, failed, spare, raid; | |
eb486760 | 45 | int from_config; |
2e0172b1 N |
46 | int expected_spares; |
47 | int devstate[MaxDisks]; | |
0f0749ad | 48 | dev_t devid[MaxDisks]; |
2e0172b1 N |
49 | int percent; |
50 | int parent_dev; /* For subarray, devnum of parent. | |
51 | * For others, NoMdDev | |
52 | */ | |
53 | struct supertype *metadata; | |
c3621c0a ML |
54 | struct state *subarray;/* for a container it is a link to first subarray |
55 | * for a subarray it is a link to next subarray | |
56 | * in the same container */ | |
57 | struct state *parent; /* for a subarray it is a link to its container | |
58 | */ | |
2e0172b1 N |
59 | struct state *next; |
60 | }; | |
61 | ||
e0bd6a96 N |
62 | struct alert_info { |
63 | char *mailaddr; | |
64 | char *mailfrom; | |
65 | char *alert_cmd; | |
66 | int dosyslog; | |
67 | }; | |
2e0172b1 N |
68 | static int make_daemon(char *pidfile); |
69 | static int check_one_sharer(int scan); | |
e0bd6a96 | 70 | static void alert(char *event, char *dev, char *disc, struct alert_info *info); |
a90e1050 | 71 | static int check_array(struct state *st, struct mdstat_ent *mdstat, |
e0bd6a96 | 72 | int test, struct alert_info *info, |
c2ecf5f6 | 73 | int increments, char *prefer); |
83f3bc5f | 74 | static int add_new_arrays(struct mdstat_ent *mdstat, struct state **statelist, |
e0bd6a96 N |
75 | int test, struct alert_info *info); |
76 | static void try_spare_migration(struct state *statelist, struct alert_info *info); | |
c3621c0a | 77 | static void link_containers_with_subarrays(struct state *list); |
2e0172b1 | 78 | |
a655e550 | 79 | int Monitor(struct mddev_dev *devlist, |
52826846 | 80 | char *mailaddr, char *alert_cmd, |
aa88f531 | 81 | int period, int daemonise, int scan, int oneshot, |
edde9560 | 82 | int dosyslog, int test, char *pidfile, int increments, |
c2ecf5f6 | 83 | int share, char *prefer) |
52826846 NB |
84 | { |
85 | /* | |
86 | * Every few seconds, scan every md device looking for changes | |
87 | * When a change is found, log it, possibly run the alert command, | |
88 | * and possibly send Email | |
89 | * | |
90 | * For each array, we record: | |
91 | * Update time | |
92 | * active/working/failed/spare drives | |
93 | * State of each device. | |
e0d19036 | 94 | * %rebuilt if rebuilding |
52826846 NB |
95 | * |
96 | * If the update time changes, check out all the data again | |
97 | * It is possible that we cannot get the state of each device | |
98 | * due to bugs in the md kernel module. | |
e0d19036 NB |
99 | * We also read /proc/mdstat to get rebuild percent, |
100 | * and to get state on all active devices incase of kernel bug. | |
52826846 | 101 | * |
e0d19036 NB |
102 | * Events are: |
103 | * Fail | |
104 | * An active device had Faulty set or Active/Sync removed | |
105 | * FailSpare | |
106 | * A spare device had Faulty set | |
107 | * SpareActive | |
108 | * An active device had a reverse transition | |
109 | * RebuildStarted | |
110 | * percent went from -1 to +ve | |
9a36a9b7 ZB |
111 | * RebuildNN |
112 | * percent went from below to not-below NN% | |
e0d19036 NB |
113 | * DeviceDisappeared |
114 | * Couldn't access a device which was previously visible | |
52826846 NB |
115 | * |
116 | * if we detect an array with active<raid and spare==0 | |
117 | * we look at other arrays that have same spare-group | |
118 | * If we find one with active==raid and spare>0, | |
119 | * and if we can get_disk_info and find a name | |
120 | * Then we hot-remove and hot-add to the other array | |
121 | * | |
e0d19036 NB |
122 | * If devlist is NULL, then we can monitor everything because --scan |
123 | * was given. We get an initial list from config file and add anything | |
124 | * that appears in /proc/mdstat | |
52826846 NB |
125 | */ |
126 | ||
2e0172b1 | 127 | struct state *statelist = NULL; |
b657208c | 128 | struct state *st2; |
52826846 | 129 | int finished = 0; |
e0d19036 | 130 | struct mdstat_ent *mdstat = NULL; |
4948b8f7 | 131 | char *mailfrom = NULL; |
e0bd6a96 | 132 | struct alert_info info; |
e0d19036 | 133 | |
d013a55e | 134 | if (!mailaddr) { |
8aec876d | 135 | mailaddr = conf_get_mailaddr(); |
d013a55e | 136 | if (mailaddr && ! scan) |
56eedc1a | 137 | fprintf(stderr, Name ": Monitor using email address \"%s\" from config file\n", |
d013a55e NB |
138 | mailaddr); |
139 | } | |
8aec876d | 140 | mailfrom = conf_get_mailfrom(); |
4948b8f7 | 141 | |
d013a55e | 142 | if (!alert_cmd) { |
8aec876d | 143 | alert_cmd = conf_get_program(); |
d013a55e | 144 | if (alert_cmd && ! scan) |
56eedc1a | 145 | fprintf(stderr, Name ": Monitor using program \"%s\" from config file\n", |
d013a55e NB |
146 | alert_cmd); |
147 | } | |
b1bac75b | 148 | if (scan && !mailaddr && !alert_cmd && !dosyslog) { |
56eedc1a | 149 | fprintf(stderr, Name ": No mail address or alert command - not monitoring.\n"); |
d013a55e | 150 | return 1; |
56eedc1a | 151 | } |
e0bd6a96 N |
152 | info.alert_cmd = alert_cmd; |
153 | info.mailaddr = mailaddr; | |
154 | info.mailfrom = mailfrom; | |
155 | info.dosyslog = dosyslog; | |
d013a55e | 156 | |
44d337f0 AC |
157 | if (daemonise) { |
158 | int rv = make_daemon(pidfile); | |
159 | if (rv >= 0) | |
160 | return rv; | |
161 | } | |
e0d19036 | 162 | |
2e0172b1 N |
163 | if (share) |
164 | if (check_one_sharer(scan)) | |
165 | return 1; | |
edde9560 | 166 | |
e0d19036 | 167 | if (devlist == NULL) { |
fa56eddb | 168 | struct mddev_ident *mdlist = conf_get_ident(NULL); |
e0d19036 | 169 | for (; mdlist; mdlist=mdlist->next) { |
fe056d1f N |
170 | struct state *st; |
171 | if (mdlist->devname == NULL) | |
172 | continue; | |
112cace6 N |
173 | if (strcasecmp(mdlist->devname, "<ignore>") == 0) |
174 | continue; | |
9bfc6a7d | 175 | st = calloc(1, sizeof *st); |
e0d19036 NB |
176 | if (st == NULL) |
177 | continue; | |
db2d001c N |
178 | if (mdlist->devname[0] == '/') |
179 | st->devname = strdup(mdlist->devname); | |
180 | else { | |
181 | st->devname = malloc(8+strlen(mdlist->devname)+1); | |
182 | strcpy(strcpy(st->devname, "/dev/md/"), | |
183 | mdlist->devname); | |
184 | } | |
e0d19036 | 185 | st->next = statelist; |
4450e59f | 186 | st->devnum = INT_MAX; |
fcf21959 | 187 | st->percent = RESYNC_UNKNOWN; |
eb486760 | 188 | st->from_config = 1; |
feb716e9 | 189 | st->expected_spares = mdlist->spare_disks; |
e0d19036 NB |
190 | if (mdlist->spare_group) |
191 | st->spare_group = strdup(mdlist->spare_group); | |
e0d19036 NB |
192 | statelist = st; |
193 | } | |
194 | } else { | |
a655e550 | 195 | struct mddev_dev *dv; |
e0d19036 | 196 | for (dv=devlist ; dv; dv=dv->next) { |
fa56eddb | 197 | struct mddev_ident *mdlist = conf_get_ident(dv->devname); |
9bfc6a7d | 198 | struct state *st = calloc(1, sizeof *st); |
e0d19036 NB |
199 | if (st == NULL) |
200 | continue; | |
201 | st->devname = strdup(dv->devname); | |
e0d19036 | 202 | st->next = statelist; |
4450e59f | 203 | st->devnum = INT_MAX; |
fcf21959 | 204 | st->percent = RESYNC_UNKNOWN; |
feb716e9 | 205 | st->expected_spares = -1; |
e5329c37 NB |
206 | if (mdlist) { |
207 | st->expected_spares = mdlist->spare_disks; | |
208 | if (mdlist->spare_group) | |
209 | st->spare_group = strdup(mdlist->spare_group); | |
210 | } | |
e0d19036 NB |
211 | statelist = st; |
212 | } | |
213 | } | |
214 | ||
215 | ||
216 | while (! finished) { | |
aa88f531 | 217 | int new_found = 0; |
e0d19036 | 218 | struct state *st; |
a90e1050 | 219 | int anydegraded = 0; |
e0d19036 NB |
220 | |
221 | if (mdstat) | |
222 | free_mdstat(mdstat); | |
22a88995 | 223 | mdstat = mdstat_read(oneshot?0:1, 0); |
e0d19036 | 224 | |
2e0172b1 | 225 | for (st=statelist; st; st=st->next) |
c2ecf5f6 N |
226 | if (check_array(st, mdstat, test, &info, |
227 | increments, prefer)) | |
a90e1050 | 228 | anydegraded = 1; |
2e0172b1 | 229 | |
e0d19036 | 230 | /* now check if there are any new devices found in mdstat */ |
2e0172b1 | 231 | if (scan) |
83f3bc5f | 232 | new_found = add_new_arrays(mdstat, &statelist, test, |
e0bd6a96 | 233 | &info); |
2e0172b1 | 234 | |
e0d19036 NB |
235 | /* If an array has active < raid && spare == 0 && spare_group != NULL |
236 | * Look for another array with spare > 0 and active == raid and same spare_group | |
237 | * if found, choose a device and hotremove/hotadd | |
238 | */ | |
a90e1050 | 239 | if (share && anydegraded) |
e0bd6a96 | 240 | try_spare_migration(statelist, &info); |
aa88f531 NB |
241 | if (!new_found) { |
242 | if (oneshot) | |
243 | break; | |
244 | else | |
dd0781e5 | 245 | mdstat_wait(period); |
aa88f531 | 246 | } |
98c6faba | 247 | test = 0; |
52826846 | 248 | } |
b657208c JS |
249 | for (st2 = statelist; st2; st2 = statelist) { |
250 | statelist = st2->next; | |
251 | free(st2); | |
252 | } | |
253 | ||
b5e64645 NB |
254 | if (pidfile) |
255 | unlink(pidfile); | |
52826846 NB |
256 | return 0; |
257 | } | |
258 | ||
2e0172b1 N |
259 | static int make_daemon(char *pidfile) |
260 | { | |
44d337f0 AC |
261 | /* Return: |
262 | * -1 in the forked daemon | |
263 | * 0 in the parent | |
264 | * 1 on error | |
265 | * so a none-negative becomes the exit code. | |
266 | */ | |
2e0172b1 N |
267 | int pid = fork(); |
268 | if (pid > 0) { | |
269 | if (!pidfile) | |
270 | printf("%d\n", pid); | |
271 | else { | |
272 | FILE *pid_file; | |
273 | pid_file=fopen(pidfile, "w"); | |
274 | if (!pid_file) | |
275 | perror("cannot create pid file"); | |
276 | else { | |
277 | fprintf(pid_file,"%d\n", pid); | |
278 | fclose(pid_file); | |
279 | } | |
280 | } | |
281 | return 0; | |
282 | } | |
283 | if (pid < 0) { | |
284 | perror("daemonise"); | |
285 | return 1; | |
286 | } | |
287 | close(0); | |
288 | open("/dev/null", O_RDWR); | |
289 | dup2(0,1); | |
290 | dup2(0,2); | |
291 | setsid(); | |
44d337f0 | 292 | return -1; |
2e0172b1 N |
293 | } |
294 | ||
295 | static int check_one_sharer(int scan) | |
296 | { | |
297 | int pid, rv; | |
298 | FILE *fp; | |
299 | char dir[20]; | |
0011874f | 300 | char path[100]; |
2e0172b1 | 301 | struct stat buf; |
0011874f JS |
302 | sprintf(path, "%s/autorebuild.pid", MDMON_DIR); |
303 | fp = fopen(path, "r"); | |
2e0172b1 | 304 | if (fp) { |
71204a50 N |
305 | if (fscanf(fp, "%d", &pid) != 1) |
306 | pid = -1; | |
2e0172b1 N |
307 | sprintf(dir, "/proc/%d", pid); |
308 | rv = stat(dir, &buf); | |
309 | if (rv != -1) { | |
310 | if (scan) { | |
311 | fprintf(stderr, Name ": Only one " | |
312 | "autorebuild process allowed" | |
313 | " in scan mode, aborting\n"); | |
314 | fclose(fp); | |
315 | return 1; | |
316 | } else { | |
317 | fprintf(stderr, Name ": Warning: One" | |
318 | " autorebuild process already" | |
a1bb2065 | 319 | " running.\n"); |
2e0172b1 N |
320 | } |
321 | } | |
322 | fclose(fp); | |
323 | } | |
324 | if (scan) { | |
0011874f | 325 | if (mkdir(MDMON_DIR, S_IRWXU) < 0 && |
a1bb2065 AC |
326 | errno != EEXIST) { |
327 | fprintf(stderr, Name ": Can't create " | |
328 | "autorebuild.pid file\n"); | |
329 | } else { | |
0011874f | 330 | fp = fopen(path, "w"); |
a1bb2065 AC |
331 | if (!fp) |
332 | fprintf(stderr, Name ": Cannot create" | |
333 | " autorebuild.pid" | |
334 | "file\n"); | |
335 | else { | |
336 | pid = getpid(); | |
337 | fprintf(fp, "%d\n", pid); | |
338 | fclose(fp); | |
339 | } | |
2e0172b1 N |
340 | } |
341 | } | |
342 | return 0; | |
343 | } | |
52826846 | 344 | |
e0bd6a96 | 345 | static void alert(char *event, char *dev, char *disc, struct alert_info *info) |
52826846 | 346 | { |
773135f5 NB |
347 | int priority; |
348 | ||
b1bac75b | 349 | if (!info->alert_cmd && !info->mailaddr && !info->dosyslog) { |
cd29a5c8 | 350 | time_t now = time(0); |
aba69144 | 351 | |
e0d19036 | 352 | printf("%1.15s: %s on %s %s\n", ctime(&now)+4, event, dev, disc?disc:"unknown device"); |
cd29a5c8 | 353 | } |
e0bd6a96 | 354 | if (info->alert_cmd) { |
52826846 NB |
355 | int pid = fork(); |
356 | switch(pid) { | |
357 | default: | |
358 | waitpid(pid, NULL, 0); | |
359 | break; | |
360 | case -1: | |
361 | break; | |
362 | case 0: | |
e0bd6a96 N |
363 | execl(info->alert_cmd, info->alert_cmd, |
364 | event, dev, disc, NULL); | |
52826846 NB |
365 | exit(2); |
366 | } | |
367 | } | |
e0bd6a96 | 368 | if (info->mailaddr && |
aba69144 | 369 | (strncmp(event, "Fail", 4)==0 || |
98c6faba | 370 | strncmp(event, "Test", 4)==0 || |
d1732eeb | 371 | strncmp(event, "Spares", 6)==0 || |
aa88f531 | 372 | strncmp(event, "Degrade", 7)==0)) { |
52826846 NB |
373 | FILE *mp = popen(Sendmail, "w"); |
374 | if (mp) { | |
008e1100 | 375 | FILE *mdstat; |
52826846 NB |
376 | char hname[256]; |
377 | gethostname(hname, sizeof(hname)); | |
378 | signal(SIGPIPE, SIG_IGN); | |
e0bd6a96 N |
379 | if (info->mailfrom) |
380 | fprintf(mp, "From: %s\n", info->mailfrom); | |
4948b8f7 NB |
381 | else |
382 | fprintf(mp, "From: " Name " monitoring <root>\n"); | |
e0bd6a96 N |
383 | fprintf(mp, "To: %s\n", info->mailaddr); |
384 | fprintf(mp, "Subject: %s event on %s:%s\n\n", | |
385 | event, dev, hname); | |
52826846 | 386 | |
e0bd6a96 N |
387 | fprintf(mp, |
388 | "This is an automatically generated" | |
389 | " mail message from " Name "\n"); | |
52826846 NB |
390 | fprintf(mp, "running on %s\n\n", hname); |
391 | ||
e0bd6a96 N |
392 | fprintf(mp, |
393 | "A %s event had been detected on" | |
394 | " md device %s.\n\n", event, dev); | |
52826846 | 395 | |
37dfc3d6 | 396 | if (disc && disc[0] != ' ') |
e0bd6a96 N |
397 | fprintf(mp, |
398 | "It could be related to" | |
399 | " component device %s.\n\n", disc); | |
37dfc3d6 NB |
400 | if (disc && disc[0] == ' ') |
401 | fprintf(mp, "Extra information:%s.\n\n", disc); | |
52826846 NB |
402 | |
403 | fprintf(mp, "Faithfully yours, etc.\n"); | |
008e1100 NB |
404 | |
405 | mdstat = fopen("/proc/mdstat", "r"); | |
406 | if (mdstat) { | |
407 | char buf[8192]; | |
408 | int n; | |
e0bd6a96 N |
409 | fprintf(mp, |
410 | "\nP.S. The /proc/mdstat file" | |
411 | " currently contains the following:\n\n"); | |
008e1100 | 412 | while ( (n=fread(buf, 1, sizeof(buf), mdstat)) > 0) |
e0bd6a96 | 413 | n=fwrite(buf, 1, n, mp); |
008e1100 NB |
414 | fclose(mdstat); |
415 | } | |
6278fb3a | 416 | pclose(mp); |
52826846 | 417 | } |
52826846 | 418 | } |
773135f5 NB |
419 | |
420 | /* log the event to syslog maybe */ | |
e0bd6a96 | 421 | if (info->dosyslog) { |
773135f5 NB |
422 | /* Log at a different severity depending on the event. |
423 | * | |
424 | * These are the critical events: */ | |
425 | if (strncmp(event, "Fail", 4)==0 || | |
426 | strncmp(event, "Degrade", 7)==0 || | |
427 | strncmp(event, "DeviceDisappeared", 17)==0) | |
428 | priority = LOG_CRIT; | |
429 | /* Good to know about, but are not failures: */ | |
430 | else if (strncmp(event, "Rebuild", 7)==0 || | |
d1732eeb NB |
431 | strncmp(event, "MoveSpare", 9)==0 || |
432 | strncmp(event, "Spares", 6) != 0) | |
773135f5 NB |
433 | priority = LOG_WARNING; |
434 | /* Everything else: */ | |
435 | else | |
436 | priority = LOG_INFO; | |
437 | ||
438 | if (disc) | |
e0bd6a96 N |
439 | syslog(priority, |
440 | "%s event detected on md device %s," | |
441 | " component device %s", event, dev, disc); | |
773135f5 | 442 | else |
e0bd6a96 N |
443 | syslog(priority, |
444 | "%s event detected on md device %s", | |
445 | event, dev); | |
773135f5 | 446 | } |
52826846 | 447 | } |
b90c0e9a | 448 | |
a90e1050 | 449 | static int check_array(struct state *st, struct mdstat_ent *mdstat, |
e0bd6a96 | 450 | int test, struct alert_info *ainfo, |
c2ecf5f6 | 451 | int increments, char *prefer) |
2e0172b1 | 452 | { |
ff044d6b AC |
453 | /* Update the state 'st' to reflect any changes shown in mdstat, |
454 | * or found by directly examining the array, and return | |
455 | * '1' if the array is degraded, or '0' if it is optimal (or dead). | |
456 | */ | |
2e0172b1 N |
457 | struct { int state, major, minor; } info[MaxDisks]; |
458 | mdu_array_info_t array; | |
459 | struct mdstat_ent *mse = NULL, *mse2; | |
460 | char *dev = st->devname; | |
461 | int fd; | |
462 | int i; | |
9e6d9291 N |
463 | int remaining_disks; |
464 | int last_disk; | |
6f51b1cc | 465 | int new_array = 0; |
2e0172b1 N |
466 | |
467 | if (test) | |
e0bd6a96 | 468 | alert("TestMessage", dev, NULL, ainfo); |
2e0172b1 N |
469 | fd = open(dev, O_RDONLY); |
470 | if (fd < 0) { | |
471 | if (!st->err) | |
e0bd6a96 N |
472 | alert("DeviceDisappeared", dev, NULL, ainfo); |
473 | st->err=1; | |
a90e1050 | 474 | return 0; |
2e0172b1 N |
475 | } |
476 | fcntl(fd, F_SETFD, FD_CLOEXEC); | |
477 | if (ioctl(fd, GET_ARRAY_INFO, &array)<0) { | |
478 | if (!st->err) | |
e0bd6a96 N |
479 | alert("DeviceDisappeared", dev, NULL, ainfo); |
480 | st->err=1; | |
2e0172b1 | 481 | close(fd); |
a90e1050 | 482 | return 0; |
2e0172b1 N |
483 | } |
484 | /* It's much easier to list what array levels can't | |
485 | * have a device disappear than all of them that can | |
486 | */ | |
487 | if (array.level == 0 || array.level == -1) { | |
eb486760 | 488 | if (!st->err && !st->from_config) |
e0bd6a96 | 489 | alert("DeviceDisappeared", dev, "Wrong-Level", ainfo); |
2e0172b1 N |
490 | st->err = 1; |
491 | close(fd); | |
a90e1050 | 492 | return 0; |
2e0172b1 N |
493 | } |
494 | if (st->devnum == INT_MAX) { | |
495 | struct stat stb; | |
496 | if (fstat(fd, &stb) == 0 && | |
497 | (S_IFMT&stb.st_mode)==S_IFBLK) { | |
498 | if (major(stb.st_rdev) == MD_MAJOR) | |
499 | st->devnum = minor(stb.st_rdev); | |
500 | else | |
501 | st->devnum = -1- (minor(stb.st_rdev)>>6); | |
502 | } | |
503 | } | |
504 | ||
505 | for (mse2 = mdstat ; mse2 ; mse2=mse2->next) | |
506 | if (mse2->devnum == st->devnum) { | |
507 | mse2->devnum = INT_MAX; /* flag it as "used" */ | |
508 | mse = mse2; | |
509 | } | |
510 | ||
511 | if (!mse) { | |
512 | /* duplicated array in statelist | |
513 | * or re-created after reading mdstat*/ | |
514 | st->err = 1; | |
515 | close(fd); | |
a90e1050 | 516 | return 0; |
2e0172b1 N |
517 | } |
518 | /* this array is in /proc/mdstat */ | |
519 | if (array.utime == 0) | |
520 | /* external arrays don't update utime, so | |
521 | * just make sure it is always different. */ | |
522 | array.utime = st->utime + 1;; | |
523 | ||
5f862fb9 N |
524 | if (st->err) { |
525 | /* New array appeared where previously had and error */ | |
526 | st->err = 0; | |
527 | st->percent = RESYNC_NONE; | |
6f51b1cc | 528 | new_array = 1; |
5f862fb9 N |
529 | alert("NewArray", st->devname, NULL, ainfo); |
530 | } | |
531 | ||
2e0172b1 N |
532 | if (st->utime == array.utime && |
533 | st->failed == array.failed_disks && | |
534 | st->working == array.working_disks && | |
535 | st->spare == array.spare_disks && | |
536 | (mse == NULL || ( | |
537 | mse->percent == st->percent | |
538 | ))) { | |
539 | close(fd); | |
ff044d6b AC |
540 | if ((st->active < st->raid) && st->spare == 0) |
541 | return 1; | |
542 | else | |
543 | return 0; | |
2e0172b1 N |
544 | } |
545 | if (st->utime == 0 && /* new array */ | |
546 | mse->pattern && strchr(mse->pattern, '_') /* degraded */ | |
547 | ) | |
e0bd6a96 | 548 | alert("DegradedArray", dev, NULL, ainfo); |
2e0172b1 N |
549 | |
550 | if (st->utime == 0 && /* new array */ | |
551 | st->expected_spares > 0 && | |
552 | array.spare_disks < st->expected_spares) | |
e0bd6a96 | 553 | alert("SparesMissing", dev, NULL, ainfo); |
fcf21959 | 554 | if (st->percent < 0 && st->percent != RESYNC_UNKNOWN && |
2e0172b1 | 555 | mse->percent >= 0) |
e0bd6a96 | 556 | alert("RebuildStarted", dev, NULL, ainfo); |
2e0172b1 N |
557 | if (st->percent >= 0 && |
558 | mse->percent >= 0 && | |
559 | (mse->percent / increments) > (st->percent / increments)) { | |
560 | char percentalert[15]; // "RebuildNN" (10 chars) or "RebuildStarted" (15 chars) | |
561 | ||
562 | if((mse->percent / increments) == 0) | |
563 | snprintf(percentalert, sizeof(percentalert), "RebuildStarted"); | |
564 | else | |
565 | snprintf(percentalert, sizeof(percentalert), "Rebuild%02d", mse->percent); | |
566 | ||
e0bd6a96 | 567 | alert(percentalert, dev, NULL, ainfo); |
2e0172b1 N |
568 | } |
569 | ||
fcf21959 | 570 | if (mse->percent == RESYNC_NONE && |
2e0172b1 N |
571 | st->percent >= 0) { |
572 | /* Rebuild/sync/whatever just finished. | |
573 | * If there is a number in /mismatch_cnt, | |
574 | * we should report that. | |
575 | */ | |
576 | struct mdinfo *sra = | |
577 | sysfs_read(-1, st->devnum, GET_MISMATCH); | |
578 | if (sra && sra->mismatch_cnt > 0) { | |
8453f8d0 LD |
579 | char cnt[80]; |
580 | snprintf(cnt, sizeof(cnt), | |
581 | " mismatches found: %d (on raid level %d)", | |
d97a5e60 | 582 | sra->mismatch_cnt, array.level); |
e0bd6a96 | 583 | alert("RebuildFinished", dev, cnt, ainfo); |
2e0172b1 | 584 | } else |
e0bd6a96 | 585 | alert("RebuildFinished", dev, NULL, ainfo); |
2e0172b1 N |
586 | if (sra) |
587 | free(sra); | |
588 | } | |
589 | st->percent = mse->percent; | |
590 | ||
9e6d9291 N |
591 | remaining_disks = array.nr_disks; |
592 | for (i=0; i<MaxDisks && remaining_disks > 0; | |
2e0172b1 N |
593 | i++) { |
594 | mdu_disk_info_t disc; | |
595 | disc.number = i; | |
596 | if (ioctl(fd, GET_DISK_INFO, &disc) >= 0) { | |
597 | info[i].state = disc.state; | |
598 | info[i].major = disc.major; | |
599 | info[i].minor = disc.minor; | |
9e6d9291 N |
600 | if (disc.major || disc.minor) |
601 | remaining_disks --; | |
2e0172b1 N |
602 | } else |
603 | info[i].major = info[i].minor = 0; | |
604 | } | |
9e6d9291 | 605 | last_disk = i; |
2e0172b1 | 606 | |
4019ad07 JL |
607 | if (mse->metadata_version && |
608 | strncmp(mse->metadata_version, "external:", 9) == 0 && | |
2e0172b1 N |
609 | is_subarray(mse->metadata_version+9)) |
610 | st->parent_dev = | |
611 | devname2devnum(mse->metadata_version+10); | |
612 | else | |
613 | st->parent_dev = NoMdDev; | |
614 | if (st->metadata == NULL && | |
615 | st->parent_dev == NoMdDev) | |
616 | st->metadata = super_by_fd(fd, NULL); | |
617 | ||
618 | close(fd); | |
619 | ||
620 | for (i=0; i<MaxDisks; i++) { | |
621 | mdu_disk_info_t disc = {0,0,0,0,0}; | |
622 | int newstate=0; | |
623 | int change; | |
624 | char *dv = NULL; | |
625 | disc.number = i; | |
6f51b1cc N |
626 | if (i < last_disk && |
627 | (info[i].major || info[i].minor)) { | |
2e0172b1 | 628 | newstate = info[i].state; |
c2ecf5f6 N |
629 | dv = map_dev_preferred( |
630 | info[i].major, info[i].minor, 1, | |
631 | prefer); | |
2e0172b1 N |
632 | disc.state = newstate; |
633 | disc.major = info[i].major; | |
634 | disc.minor = info[i].minor; | |
6f51b1cc N |
635 | } else |
636 | newstate = (1 << MD_DISK_REMOVED); | |
637 | ||
2e0172b1 | 638 | if (dv == NULL && st->devid[i]) |
c2ecf5f6 N |
639 | dv = map_dev_preferred( |
640 | major(st->devid[i]), | |
641 | minor(st->devid[i]), 1, prefer); | |
2e0172b1 | 642 | change = newstate ^ st->devstate[i]; |
6f51b1cc N |
643 | if (st->utime && change && !st->err && !new_array) { |
644 | if ((st->devstate[i]&change)&(1<<MD_DISK_SYNC)) | |
e0bd6a96 | 645 | alert("Fail", dev, dv, ainfo); |
6f51b1cc | 646 | else if ((newstate & (1<<MD_DISK_FAULTY)) && |
2e0172b1 | 647 | (disc.major || disc.minor) && |
6f51b1cc | 648 | st->devid[i] == makedev(disc.major, disc.minor)) |
e0bd6a96 | 649 | alert("FailSpare", dev, dv, ainfo); |
6f51b1cc | 650 | else if ((newstate&change)&(1<<MD_DISK_SYNC)) |
e0bd6a96 | 651 | alert("SpareActive", dev, dv, ainfo); |
2e0172b1 N |
652 | } |
653 | st->devstate[i] = newstate; | |
654 | st->devid[i] = makedev(disc.major, disc.minor); | |
655 | } | |
656 | st->active = array.active_disks; | |
657 | st->working = array.working_disks; | |
658 | st->spare = array.spare_disks; | |
659 | st->failed = array.failed_disks; | |
660 | st->utime = array.utime; | |
661 | st->raid = array.raid_disks; | |
662 | st->err = 0; | |
a90e1050 N |
663 | if ((st->active < st->raid) && st->spare == 0) |
664 | return 1; | |
665 | return 0; | |
2e0172b1 N |
666 | } |
667 | ||
83f3bc5f | 668 | static int add_new_arrays(struct mdstat_ent *mdstat, struct state **statelist, |
e0bd6a96 | 669 | int test, struct alert_info *info) |
2e0172b1 N |
670 | { |
671 | struct mdstat_ent *mse; | |
672 | int new_found = 0; | |
673 | ||
674 | for (mse=mdstat; mse; mse=mse->next) | |
675 | if (mse->devnum != INT_MAX && | |
676 | (!mse->level || /* retrieve containers */ | |
677 | (strcmp(mse->level, "raid0") != 0 && | |
678 | strcmp(mse->level, "linear") != 0)) | |
679 | ) { | |
9bfc6a7d | 680 | struct state *st = calloc(1, sizeof *st); |
2e0172b1 N |
681 | mdu_array_info_t array; |
682 | int fd; | |
683 | if (st == NULL) | |
684 | continue; | |
685 | st->devname = strdup(get_md_name(mse->devnum)); | |
686 | if ((fd = open(st->devname, O_RDONLY)) < 0 || | |
687 | ioctl(fd, GET_ARRAY_INFO, &array)< 0) { | |
688 | /* no such array */ | |
689 | if (fd >=0) close(fd); | |
690 | put_md_name(st->devname); | |
691 | free(st->devname); | |
692 | if (st->metadata) { | |
693 | st->metadata->ss->free_super(st->metadata); | |
694 | free(st->metadata); | |
695 | } | |
696 | free(st); | |
697 | continue; | |
698 | } | |
699 | close(fd); | |
83f3bc5f | 700 | st->next = *statelist; |
2e0172b1 N |
701 | st->err = 1; |
702 | st->devnum = mse->devnum; | |
fcf21959 | 703 | st->percent = RESYNC_UNKNOWN; |
2e0172b1 | 704 | st->expected_spares = -1; |
eb28e119 N |
705 | if (mse->metadata_version && |
706 | strncmp(mse->metadata_version, "external:", 9) == 0 && | |
2e0172b1 N |
707 | is_subarray(mse->metadata_version+9)) |
708 | st->parent_dev = | |
709 | devname2devnum(mse->metadata_version+10); | |
710 | else | |
711 | st->parent_dev = NoMdDev; | |
83f3bc5f | 712 | *statelist = st; |
2e0172b1 | 713 | if (test) |
e0bd6a96 | 714 | alert("TestMessage", st->devname, NULL, info); |
2e0172b1 N |
715 | new_found = 1; |
716 | } | |
717 | return new_found; | |
718 | } | |
719 | ||
de697acc | 720 | static int get_min_spare_size_required(struct state *st, unsigned long long *sizep) |
80e7f8c3 AC |
721 | { |
722 | int fd; | |
80e7f8c3 AC |
723 | |
724 | if (!st->metadata || | |
300f5033 CA |
725 | !st->metadata->ss->min_acceptable_spare_size) { |
726 | *sizep = 0; | |
de697acc | 727 | return 0; |
300f5033 | 728 | } |
80e7f8c3 AC |
729 | |
730 | fd = open(st->devname, O_RDONLY); | |
731 | if (fd < 0) | |
de697acc AC |
732 | return 1; |
733 | if (st->metadata->ss->external) | |
734 | st->metadata->ss->load_container(st->metadata, fd, st->devname); | |
735 | else | |
736 | st->metadata->ss->load_super(st->metadata, fd, st->devname); | |
80e7f8c3 | 737 | close(fd); |
de697acc AC |
738 | if (!st->metadata->sb) |
739 | return 1; | |
740 | *sizep = st->metadata->ss->min_acceptable_spare_size(st->metadata); | |
80e7f8c3 AC |
741 | st->metadata->ss->free_super(st->metadata); |
742 | ||
de697acc | 743 | return 0; |
80e7f8c3 AC |
744 | } |
745 | ||
5ec0f373 | 746 | static int check_donor(struct state *from, struct state *to) |
2feb22ef | 747 | { |
66f5c4b6 N |
748 | struct state *sub; |
749 | ||
2feb22ef N |
750 | if (from == to) |
751 | return 0; | |
66f5c4b6 N |
752 | if (from->parent) |
753 | /* Cannot move from a member */ | |
2feb22ef | 754 | return 0; |
ff044d6b AC |
755 | if (from->err) |
756 | return 0; | |
66f5c4b6 N |
757 | for (sub = from->subarray; sub; sub = sub->subarray) |
758 | /* If source array has degraded subarrays, don't | |
759 | * remove anything | |
760 | */ | |
761 | if (sub->active < sub->raid) | |
762 | return 0; | |
763 | if (from->metadata->ss->external == 0) | |
764 | if (from->active < from->raid) | |
765 | return 0; | |
2feb22ef N |
766 | if (from->spare <= 0) |
767 | return 0; | |
e78dda3b | 768 | return 1; |
2feb22ef N |
769 | } |
770 | ||
0f0749ad | 771 | static dev_t choose_spare(struct state *from, struct state *to, |
f0b85306 | 772 | struct domainlist *domlist, unsigned long long min_size) |
0fa21e85 N |
773 | { |
774 | int d; | |
0f0749ad | 775 | dev_t dev = 0; |
0fa21e85 N |
776 | |
777 | for (d = from->raid; !dev && d < MaxDisks; d++) { | |
778 | if (from->devid[d] > 0 && | |
779 | from->devstate[d] == 0) { | |
780 | struct dev_policy *pol; | |
781 | unsigned long long dev_size; | |
782 | ||
bfd76b93 CA |
783 | if (to->metadata->ss->external && |
784 | test_partition_from_id(from->devid[d])) | |
785 | continue; | |
786 | ||
0fa21e85 N |
787 | if (min_size && |
788 | dev_size_from_id(from->devid[d], &dev_size) && | |
789 | dev_size < min_size) | |
790 | continue; | |
791 | ||
792 | pol = devnum_policy(from->devid[d]); | |
793 | if (from->spare_group) | |
794 | pol_add(&pol, pol_domain, | |
795 | from->spare_group, NULL); | |
e5508b36 | 796 | if (domain_test(domlist, pol, to->metadata->ss->name) == 1) |
0fa21e85 N |
797 | dev = from->devid[d]; |
798 | dev_policy_free(pol); | |
799 | } | |
800 | } | |
801 | return dev; | |
802 | } | |
803 | ||
0f0749ad | 804 | static dev_t container_choose_spare(struct state *from, struct state *to, |
f0b85306 | 805 | struct domainlist *domlist, |
a1e49d69 | 806 | unsigned long long min_size, int active) |
5739e0d0 N |
807 | { |
808 | /* This is similar to choose_spare, but we cannot trust devstate, | |
809 | * so we need to read the metadata instead | |
810 | */ | |
326727d9 | 811 | struct mdinfo *list; |
5739e0d0 | 812 | struct supertype *st = from->metadata; |
ff044d6b | 813 | int fd = open(from->devname, O_RDONLY); |
5739e0d0 | 814 | int err; |
0f0749ad | 815 | dev_t dev = 0; |
5739e0d0 N |
816 | |
817 | if (fd < 0) | |
818 | return 0; | |
326727d9 AC |
819 | if (!st->ss->getinfo_super_disks) { |
820 | close(fd); | |
5739e0d0 | 821 | return 0; |
326727d9 | 822 | } |
5739e0d0 N |
823 | |
824 | err = st->ss->load_container(st, fd, NULL); | |
825 | close(fd); | |
826 | if (err) | |
827 | return 0; | |
5739e0d0 | 828 | |
a1e49d69 CA |
829 | if (from == to) { |
830 | /* We must check if number of active disks has not increased | |
831 | * since ioctl in main loop. mdmon may have added spare | |
832 | * to subarray. If so we do not need to look for more spares | |
833 | * so return non zero value */ | |
834 | int active_cnt = 0; | |
835 | struct mdinfo *dp; | |
836 | list = st->ss->getinfo_super_disks(st); | |
837 | if (!list) { | |
838 | st->ss->free_super(st); | |
839 | return 1; | |
840 | } | |
841 | dp = list->devs; | |
842 | while (dp) { | |
843 | if (dp->disk.state & (1<<MD_DISK_SYNC) && | |
844 | !(dp->disk.state & (1<<MD_DISK_FAULTY))) | |
845 | active_cnt++; | |
846 | dp = dp->next; | |
847 | } | |
848 | sysfs_free(list); | |
849 | if (active < active_cnt) { | |
850 | /* Spare just activated.*/ | |
851 | st->ss->free_super(st); | |
852 | return 1; | |
853 | } | |
854 | } | |
855 | ||
326727d9 AC |
856 | /* We only need one spare so full list not needed */ |
857 | list = container_choose_spares(st, min_size, domlist, from->spare_group, | |
858 | to->metadata->ss->name, 1); | |
859 | if (list) { | |
860 | struct mdinfo *disks = list->devs; | |
861 | if (disks) | |
862 | dev = makedev(disks->disk.major, disks->disk.minor); | |
863 | sysfs_free(list); | |
5739e0d0 | 864 | } |
326727d9 | 865 | st->ss->free_super(st); |
5739e0d0 N |
866 | return dev; |
867 | } | |
868 | ||
869 | ||
e0bd6a96 | 870 | static void try_spare_migration(struct state *statelist, struct alert_info *info) |
2e0172b1 | 871 | { |
66f5c4b6 N |
872 | struct state *from; |
873 | struct state *st; | |
c3621c0a ML |
874 | |
875 | link_containers_with_subarrays(statelist); | |
66f5c4b6 N |
876 | for (st = statelist; st; st = st->next) |
877 | if (st->active < st->raid && | |
ef15641f | 878 | st->spare == 0 && !st->err) { |
e78dda3b N |
879 | struct domainlist *domlist = NULL; |
880 | int d; | |
66f5c4b6 | 881 | struct state *to = st; |
f0b85306 | 882 | unsigned long long min_size; |
66f5c4b6 | 883 | |
c0dc0ad5 CA |
884 | if (to->parent_dev != NoMdDev && !to->parent) |
885 | /* subarray monitored without parent container | |
886 | * we can't move spares here */ | |
887 | continue; | |
888 | ||
66f5c4b6 N |
889 | if (to->parent) |
890 | /* member of a container */ | |
891 | to = to->parent; | |
e78dda3b | 892 | |
de697acc AC |
893 | if (get_min_spare_size_required(to, &min_size)) |
894 | continue; | |
e9a2ac02 AC |
895 | if (to->metadata->ss->external) { |
896 | /* We must make sure there is | |
897 | * no suitable spare in container already. | |
898 | * If there is we don't add more */ | |
899 | dev_t devid = container_choose_spare( | |
a1e49d69 | 900 | to, to, NULL, min_size, st->active); |
e9a2ac02 AC |
901 | if (devid > 0) |
902 | continue; | |
903 | } | |
e78dda3b N |
904 | for (d = 0; d < MaxDisks; d++) |
905 | if (to->devid[d]) | |
906 | domainlist_add_dev(&domlist, | |
907 | to->devid[d], | |
908 | to->metadata->ss->name); | |
909 | if (to->spare_group) | |
910 | domain_add(&domlist, to->spare_group); | |
5ec0f373 ML |
911 | /* |
912 | * No spare migration if the destination | |
913 | * has no domain. Skip this array. | |
914 | */ | |
915 | if (!domlist) | |
916 | continue; | |
0fa21e85 | 917 | for (from=statelist ; from ; from=from->next) { |
0f0749ad | 918 | dev_t devid; |
5ec0f373 | 919 | if (!check_donor(from, to)) |
0fa21e85 | 920 | continue; |
5739e0d0 N |
921 | if (from->metadata->ss->external) |
922 | devid = container_choose_spare( | |
a1e49d69 | 923 | from, to, domlist, min_size, 0); |
5739e0d0 | 924 | else |
f0b85306 AC |
925 | devid = choose_spare(from, to, domlist, |
926 | min_size); | |
0fa21e85 | 927 | if (devid > 0 |
d52bb542 AC |
928 | && move_spare(from->devname, to->devname, devid)) { |
929 | alert("MoveSpare", to->devname, from->devname, info); | |
930 | break; | |
931 | } | |
0fa21e85 | 932 | } |
e78dda3b | 933 | domain_free(domlist); |
2e0172b1 N |
934 | } |
935 | } | |
c3621c0a ML |
936 | |
937 | /* search the statelist to connect external | |
938 | * metadata subarrays with their containers | |
939 | * We always completely rebuild the tree from scratch as | |
940 | * that is safest considering the possibility of entries | |
941 | * disappearing or changing. | |
942 | */ | |
943 | static void link_containers_with_subarrays(struct state *list) | |
944 | { | |
945 | struct state *st; | |
946 | struct state *cont; | |
947 | for (st = list; st; st = st->next) { | |
948 | st->parent = NULL; | |
949 | st->subarray = NULL; | |
950 | } | |
951 | for (st = list; st; st = st->next) | |
952 | if (st->parent_dev != NoMdDev) | |
953 | for (cont = list; cont; cont = cont->next) | |
954 | if (!cont->err && | |
955 | cont->parent_dev == NoMdDev && | |
956 | cont->devnum == st->parent_dev) { | |
957 | st->parent = cont; | |
958 | st->subarray = cont->subarray; | |
959 | cont->subarray = st; | |
960 | break; | |
961 | } | |
962 | } | |
963 | ||
b90c0e9a NB |
964 | /* Not really Monitor but ... */ |
965 | int Wait(char *dev) | |
966 | { | |
967 | struct stat stb; | |
968 | int devnum; | |
969 | int rv = 1; | |
970 | ||
971 | if (stat(dev, &stb) != 0) { | |
972 | fprintf(stderr, Name ": Cannot find %s: %s\n", dev, | |
973 | strerror(errno)); | |
974 | return 2; | |
975 | } | |
c94709e8 | 976 | devnum = stat2devnum(&stb); |
b90c0e9a NB |
977 | |
978 | while(1) { | |
979 | struct mdstat_ent *ms = mdstat_read(1, 0); | |
980 | struct mdstat_ent *e; | |
981 | ||
982 | for (e=ms ; e; e=e->next) | |
983 | if (e->devnum == devnum) | |
984 | break; | |
985 | ||
986 | if (!e || e->percent < 0) { | |
e7783ee6 | 987 | if (e && e->metadata_version && |
c94709e8 DW |
988 | strncmp(e->metadata_version, "external:", 9) == 0) { |
989 | if (is_subarray(&e->metadata_version[9])) | |
990 | ping_monitor(&e->metadata_version[9]); | |
991 | else | |
983fff45 | 992 | ping_monitor_by_id(devnum); |
c94709e8 | 993 | } |
b90c0e9a NB |
994 | free_mdstat(ms); |
995 | return rv; | |
996 | } | |
89a10d84 | 997 | free_mdstat(ms); |
b90c0e9a NB |
998 | rv = 0; |
999 | mdstat_wait(5); | |
1000 | } | |
1001 | } | |
679eb882 N |
1002 | |
1003 | #ifndef MDASSEMBLE | |
1004 | ||
1005 | static char *clean_states[] = { | |
1006 | "clear", "inactive", "readonly", "read-auto", "clean", NULL }; | |
1007 | ||
1008 | int WaitClean(char *dev, int sock, int verbose) | |
1009 | { | |
1010 | int fd; | |
1011 | struct mdinfo *mdi; | |
1012 | int rv = 1; | |
1013 | int devnum; | |
1014 | ||
1015 | fd = open(dev, O_RDONLY); | |
1016 | if (fd < 0) { | |
1017 | if (verbose) | |
1018 | fprintf(stderr, Name ": Couldn't open %s: %s\n", dev, strerror(errno)); | |
1019 | return 1; | |
1020 | } | |
1021 | ||
1022 | devnum = fd2devnum(fd); | |
1023 | mdi = sysfs_read(fd, devnum, GET_VERSION|GET_LEVEL|GET_SAFEMODE); | |
1024 | if (!mdi) { | |
1025 | if (verbose) | |
1026 | fprintf(stderr, Name ": Failed to read sysfs attributes for " | |
1027 | "%s\n", dev); | |
1028 | close(fd); | |
1029 | return 0; | |
1030 | } | |
1031 | ||
1032 | switch(mdi->array.level) { | |
1033 | case LEVEL_LINEAR: | |
1034 | case LEVEL_MULTIPATH: | |
1035 | case 0: | |
1036 | /* safemode delay is irrelevant for these levels */ | |
1037 | rv = 0; | |
1038 | } | |
1039 | ||
1040 | /* for internal metadata the kernel handles the final clean | |
1041 | * transition, containers can never be dirty | |
1042 | */ | |
1043 | if (!is_subarray(mdi->text_version)) | |
1044 | rv = 0; | |
1045 | ||
1046 | /* safemode disabled ? */ | |
1047 | if (mdi->safe_mode_delay == 0) | |
1048 | rv = 0; | |
1049 | ||
1050 | if (rv) { | |
1051 | int state_fd = sysfs_open(fd2devnum(fd), NULL, "array_state"); | |
1052 | char buf[20]; | |
1053 | fd_set fds; | |
1054 | struct timeval tm; | |
1055 | ||
1056 | /* minimize the safe_mode_delay and prepare to wait up to 5s | |
1057 | * for writes to quiesce | |
1058 | */ | |
1059 | sysfs_set_safemode(mdi, 1); | |
1060 | tm.tv_sec = 5; | |
1061 | tm.tv_usec = 0; | |
1062 | ||
1063 | FD_ZERO(&fds); | |
1064 | ||
1065 | /* wait for array_state to be clean */ | |
1066 | while (1) { | |
1067 | rv = read(state_fd, buf, sizeof(buf)); | |
1068 | if (rv < 0) | |
1069 | break; | |
1070 | if (sysfs_match_word(buf, clean_states) <= 4) | |
1071 | break; | |
1072 | FD_SET(state_fd, &fds); | |
1073 | rv = select(state_fd + 1, NULL, NULL, &fds, &tm); | |
1074 | if (rv < 0 && errno != EINTR) | |
1075 | break; | |
1076 | lseek(state_fd, 0, SEEK_SET); | |
1077 | } | |
1078 | if (rv < 0) | |
1079 | rv = 1; | |
1080 | else if (fping_monitor(sock) == 0 || | |
1081 | ping_monitor(mdi->text_version) == 0) { | |
1082 | /* we need to ping to close the window between array | |
1083 | * state transitioning to clean and the metadata being | |
1084 | * marked clean | |
1085 | */ | |
1086 | rv = 0; | |
1087 | } else | |
1088 | rv = 1; | |
1089 | if (rv && verbose) | |
1090 | fprintf(stderr, Name ": Error waiting for %s to be clean\n", | |
1091 | dev); | |
1092 | ||
1093 | /* restore the original safe_mode_delay */ | |
1094 | sysfs_set_safemode(mdi, mdi->safe_mode_delay); | |
1095 | close(state_fd); | |
1096 | } | |
1097 | ||
1098 | sysfs_free(mdi); | |
1099 | close(fd); | |
1100 | ||
1101 | return rv; | |
1102 | } | |
1103 | #endif /* MDASSEMBLE */ |