void print_colheaders(FILE *output, RbtHandle rbcolumns)
{
int colcount;
RbtIterator colhandle;
colcount = 1; /* Remember the hostname column */
/* Group column headings */
fprintf(output, "<TR>");
fprintf(output, "<TD ROWSPAN=2> </TD>\n"); /* For the prio column - in both row headers+dash rows */
fprintf(output, "<TD ROWSPAN=2> </TD>\n"); /* For the host column - in both row headers+dash rows */
for (colhandle = rbtBegin(rbcolumns); (colhandle != rbtEnd(rbcolumns)); colhandle = rbtNext(rbcolumns, colhandle)) {
void *k1, *k2;
char *colname;
rbtKeyValue(rbcolumns, colhandle, &k1, &k2);
colname = (char *)k1;
colcount++;
fprintf(output, " <TD ALIGN=CENTER VALIGN=BOTTOM WIDTH=45>\n");
fprintf(output, " <A HREF=\"%s\"><FONT %s><B>%s</B></FONT></A> </TD>\n",
columnlink(colname), xgetenv("XYMONPAGECOLFONT"), colname);
}
fprintf(output, "</TR>\n");
fprintf(output, "<TR><TD COLSPAN=%d><HR WIDTH=\"100%%\"></TD></TR>\n\n", colcount);
}
void print_oneprio(FILE *output, RbtHandle rbstate, RbtHandle rbcolumns, int prio)
{
RbtIterator hhandle;
int firsthost = 1;
char *curhost = "";
/* Then output each host and their column status */
for (hhandle = rbtBegin(rbstate); (hhandle != rbtEnd(rbstate)); hhandle = rbtNext(rbstate, hhandle)) {
void *k1, *k2;
hstatus_t *itm;
rbtKeyValue(rbstate, hhandle, &k1, &k2);
itm = (hstatus_t *)k2;
if (itm->config->priority != prio) continue;
if (strcmp(curhost, itm->hostname) == 0) continue;
/* New host */
curhost = itm->hostname;
print_hoststatus(output, itm, rbcolumns, prio, firsthost);
firsthost = 0;
}
/* If we did output any hosts, make some room for the next priority */
if (!firsthost) fprintf(output, "<TR><TD> </TD></TR>\n");
}
void generate_critpage(FILE *output, char *hfprefix)
{
RbtIterator hhandle;
int color = COL_GREEN;
int maxprio = 0;
/* Determine background color and max. priority */
for (hhandle = rbtBegin(rbstate); (hhandle != rbtEnd(rbstate)); hhandle = rbtNext(rbstate, hhandle)) {
void *k1, *k2;
hstatus_t *itm;
rbtKeyValue(rbstate, hhandle, &k1, &k2);
itm = (hstatus_t *)k2;
if (itm->color > color) color = itm->color;
if (itm->config->priority > maxprio) maxprio = itm->config->priority;
}
headfoot(output, hfprefix, "", "header", color);
fprintf(output, "<center>\n");
if (color != COL_GREEN) {
RbtHandle rbcolumns;
int prio;
rbcolumns = columnlist(rbstate);
fprintf(output, "<TABLE BORDER=0 CELLPADDING=4 SUMMARY=\"Critical status display\">\n");
print_colheaders(output, rbcolumns);
for (prio = 1; (prio <= maxprio); prio++) {
print_oneprio(output, rbstate, rbcolumns, prio);
}
fprintf(output, "</TABLE>\n");
rbtDelete(rbcolumns);
}
else {
/* "All Monitored Systems OK */
fprintf(output, "%s", xgetenv("XYMONALLOKTEXT"));
}
fprintf(output, "</center>\n");
headfoot(output, hfprefix, "", "footer", color);
}
RbtHandle columnlist(RbtHandle statetree)
{
RbtHandle rbcolumns;
RbtIterator hhandle;
rbcolumns = rbtNew(name_compare);
for (hhandle = rbtBegin(statetree); (hhandle != rbtEnd(statetree)); hhandle = rbtNext(statetree, hhandle)) {
void *k1, *k2;
hstatus_t *itm;
RbtStatus status;
rbtKeyValue(statetree, hhandle, &k1, &k2);
itm = (hstatus_t *)k2;
status = rbtInsert(rbcolumns, itm->testname, NULL);
}
return rbcolumns;
}
int delete_nkconfig(char *dropkey, int evenifcloned)
{
RbtHandle handle;
void *k1, *k2;
handle = rbtFind(rbconf, dropkey);
if (handle == rbtEnd(rbconf)) return 0;
if (!evenifcloned) {
/* Check if this record has any clones attached to it */
char *hostname, *p;
hostname = strdup(dropkey);
p = strchr(hostname, '|'); if (p) *p = '\0';
handle = rbtBegin(rbconf);
while (handle != rbtEnd(rbconf)) {
void *k1, *k2;
char *key, *ptr;
rbtKeyValue(rbconf, handle, &k1, &k2);
key = (char *)k1; ptr = (char *)k2;
if ((*(key + strlen(key) - 1) == '=') && (strcmp(hostname, ptr) == 0)) {
xfree(hostname);
return 1;
}
handle = rbtNext(rbconf, handle);
}
xfree(hostname);
}
handle = rbtFind(rbconf, dropkey);
if (handle != rbtEnd(rbconf)) {
rbtKeyValue(rbconf, handle, &k1, &k2);
rbtErase(rbconf, handle);
flushrec(k1, k2);
}
return 0;
}
void locator_flushcache(enum locator_servicetype_t svc, char *key)
{
RbtIterator handle;
if (!havecache[svc]) return;
if (key) {
handle = rbtFind(locatorcache[svc], key);
if (handle != rbtEnd(locatorcache[svc])) {
cacheitm_t *itm = (cacheitm_t *)gettreeitem(locatorcache[svc], handle);
itm->tstamp = 0;
}
}
else {
for (handle = rbtBegin(locatorcache[svc]); (handle != rbtEnd(locatorcache[svc])); handle = rbtNext(locatorcache[svc], handle)) {
cacheitm_t *itm = (cacheitm_t *)gettreeitem(locatorcache[svc], handle);
itm->tstamp = 0;
}
}
}
int main(int argc, char *argv[])
{
int argi;
struct sigaction sa;
namelist_t *hostwalk;
time_t nexttimeout;
for (argi=1; (argi < argc); argi++) {
if (argnmatch(argv[argi], "--server=")) {
char *p = strchr(argv[argi], '=');
serverip = strdup(p+1);
}
else if (argnmatch(argv[argi], "--interval=")) {
char *p = strchr(argv[argi], '=');
pollinterval = atoi(p+1);
}
else if (argnmatch(argv[argi], "--log-interval=")) {
char *p = strchr(argv[argi], '=');
errorloginterval = atoi(p+1);
}
else if (argnmatch(argv[argi], "--id=")) {
char *p = strchr(argv[argi], '=');
serverid = atoi(p+1);
}
else if (strcmp(argv[argi], "--debug") == 0) {
debug = 1;
}
}
setup_signalhandler("hobbitfetch");
memset(&sa, 0, sizeof(sa));
sa.sa_handler = sigmisc_handler;
sigaction(SIGHUP, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
sigaction(SIGUSR1, &sa, NULL); /* SIGUSR1 triggers logging of active requests */
clients = rbtNew(name_compare);
nexttimeout = time(NULL) + 60;
{
/* Seed the random number generator */
struct timeval tv;
struct timezone tz;
gettimeofday(&tv, &tz);
srandom(tv.tv_usec);
}
do {
RbtIterator handle;
conn_t *connwalk, *cprev;
fd_set fdread, fdwrite;
int n, maxfd;
struct timeval tmo;
time_t now;
now = time(NULL);
if (now > reloadtime) {
/* Time to reload the bb-hosts file */
reloadtime = now + 600;
load_hostnames(xgetenv("BBHOSTS"), NULL, get_fqdn());
for (hostwalk = first_host(); (hostwalk); hostwalk = hostwalk->next) {
char *hname;
clients_t *newclient;
if (!bbh_item(hostwalk, BBH_FLAG_PULLDATA)) continue;
hname = bbh_item(hostwalk, BBH_HOSTNAME);
handle = rbtFind(clients, hname);
if (handle == rbtEnd(clients)) {
newclient = (clients_t *)calloc(1, sizeof(clients_t));
newclient->hostname = strdup(hname);
rbtInsert(clients, newclient->hostname, newclient);
whentoqueue = now;
}
}
}
now = time(NULL);
if (now > nexttimeout) {
/* Check for connections that have timed out */
nexttimeout = now + 60;
for (connwalk = chead; (connwalk); connwalk = connwalk->next) {
if ((connwalk->tstamp + 60) < now) {
if (debug || (connwalk->client->nexterrortxt < now)) {
errprintf("Timeout while talking to %s (req %lu): Aborting session\n",
addrstring(&connwalk->caddr), connwalk->seq);
connwalk->client->nexterrortxt = now + errorloginterval;
}
flag_cleanup(connwalk);
}
}
}
if (needcleanup) {
/* Remove any finished requests */
needcleanup = 0;
connwalk = chead; cprev = NULL;
dbgprintf("Doing cleanup\n");
while (connwalk) {
conn_t *zombie;
if ((connwalk->action == C_READING) || (connwalk->action == C_WRITING)) {
/* Active connection - skip to the next conn_t record */
cprev = connwalk;
connwalk = connwalk->next;
continue;
}
if (connwalk->action == C_CLEANUP) {
if (connwalk->ctype == C_CLIENT) {
/*
* Finished getting data from a client,
* flag idle and set next poll time.
*/
connwalk->client->busy = 0;
set_polltime(connwalk->client);
}
else if (connwalk->ctype == C_SERVER) {
/* Nothing needed for server cleanups */
}
}
/* Unlink the request from the list of active connections */
zombie = connwalk;
if (cprev == NULL) {
chead = zombie->next;
connwalk = chead;
cprev = NULL;
}
else {
cprev->next = zombie->next;
connwalk = zombie->next;
}
/* Purge the zombie */
dbgprintf("Request completed: req %lu, peer %s, action was %d, type was %d\n",
zombie->seq, addrstring(&zombie->caddr),
zombie->action, zombie->ctype);
close(zombie->sockfd);
freestrbuffer(zombie->msgbuf);
xfree(zombie);
}
/* Set the tail pointer correctly */
ctail = chead;
if (ctail) { while (ctail->next) ctail = ctail->next; }
}
if (dumpsessions) {
/* Set by SIGUSR1 - dump the list of active requests */
dumpsessions = 0;
for (connwalk = chead; (connwalk); connwalk = connwalk->next) {
char *ctypestr, *actionstr;
char timestr[30];
switch (connwalk->ctype) {
case C_CLIENT: ctypestr = "client"; break;
case C_SERVER: ctypestr = "server"; break;
}
switch (connwalk->action) {
case C_READING: actionstr = "reading"; break;
case C_WRITING: actionstr = "writing"; break;
case C_CLEANUP: actionstr = "cleanup"; break;
}
strftime(timestr, sizeof(timestr), "%Y-%m-%d %H:%M:%S",
localtime(&connwalk->tstamp));
errprintf("Request %lu: state %s/%s, peer %s, started %s (%lu secs ago)\n",
connwalk->seq, ctypestr, actionstr, addrstring(&connwalk->caddr),
timestr, (now - connwalk->tstamp));
}
}
now = time(NULL);
if (now >= whentoqueue) {
/* Scan host-tree for clients we need to contact */
for (handle = rbtBegin(clients); (handle != rbtEnd(clients)); handle = rbtNext(clients, handle)) {
clients_t *clientwalk;
char msgline[100];
strbuffer_t *request;
char *pullstr, *ip;
int port;
clientwalk = (clients_t *)gettreeitem(clients, handle);
if (clientwalk->busy) continue;
if (clientwalk->nextpoll > now) continue;
/* Deleted hosts stay in our tree - but should disappear from the known hosts */
hostwalk = hostinfo(clientwalk->hostname); if (!hostwalk) continue;
pullstr = bbh_item(hostwalk, BBH_FLAG_PULLDATA); if (!pullstr) continue;
ip = strchr(pullstr, '=');
port = atoi(xgetenv("BBPORT"));
if (!ip) {
ip = strdup(bbh_item(hostwalk, BBH_IP));
}
else {
/* There is an explicit IP setting in the pulldata tag */
char *p;
ip++; /* Skip the '=' */
ip = strdup(ip);
p = strchr(ip, ':');
if (p) { *p = '\0'; port = atoi(p+1); }
if (*ip == '\0') {
/* No IP given, just a port number */
xfree(ip);
ip = strdup(bbh_item(hostwalk, BBH_IP));
}
}
if (strcmp(ip, "0.0.0.0") == 0) {
struct hostent *hent;
xfree(ip); ip = NULL;
hent = gethostbyname(clientwalk->hostname);
if (hent) {
struct in_addr addr;
memcpy(&addr, *(hent->h_addr_list), sizeof(addr));
ip = strdup(inet_ntoa(addr));
}
}
if (!ip) continue;
/*
* Build the "pullclient" request, which includes the latest
* clientdata config we got from the server. Keep the clientdata
* here - we send "pullclient" requests more often that we actually
* contact the server, but we should provide the config data always.
*/
request = newstrbuffer(0);
sprintf(msgline, "pullclient %d\n", serverid);
addtobuffer(request, msgline);
if (clientwalk->clientdata) addtobuffer(request, clientwalk->clientdata);
/* Put the request on the connection queue */
addrequest(C_CLIENT, ip, port, request, clientwalk);
clientwalk->busy = 1;
xfree(ip);
}
}
/* Handle request queue */
FD_ZERO(&fdread);
FD_ZERO(&fdwrite);
maxfd = -1;
for (connwalk = chead; (connwalk); connwalk = connwalk->next) {
switch (connwalk->action) {
case C_READING:
FD_SET(connwalk->sockfd, &fdread);
if (connwalk->sockfd > maxfd) maxfd = connwalk->sockfd;
break;
case C_WRITING:
FD_SET(connwalk->sockfd, &fdwrite);
if (connwalk->sockfd > maxfd) maxfd = connwalk->sockfd;
break;
case C_CLEANUP:
break;
}
}
/* Do select with a 1 second timeout */
tmo.tv_sec = 1;
tmo.tv_usec = 0;
n = select(maxfd+1, &fdread, &fdwrite, NULL, &tmo);
if (n == -1) {
if (errno == EINTR) continue; /* Interrupted, e.g. a SIGHUP */
/* This is a "cannot-happen" failure. Bail out */
errprintf("select failure: %s\n", strerror(errno));
return 0;
}
if (n == 0) continue; /* Timeout */
for (connwalk = chead; (connwalk); connwalk = connwalk->next) {
switch (connwalk->action) {
case C_READING:
if (FD_ISSET(connwalk->sockfd, &fdread)) grabdata(connwalk);
break;
case C_WRITING:
if (FD_ISSET(connwalk->sockfd, &fdwrite)) senddata(connwalk);
break;
case C_CLEANUP:
break;
}
}
} while (running);
return 0;
}
void print_hoststatus(FILE *output, hstatus_t *itm, RbtHandle columns, int prio, int firsthost)
{
void *hinfo;
char *dispname, *ip, *key;
time_t now;
RbtIterator colhandle;
now = getcurrenttime(NULL);
hinfo = hostinfo(itm->hostname);
dispname = xmh_item(hinfo, XMH_DISPLAYNAME);
ip = xmh_item(hinfo, XMH_IP);
fprintf(output, "<TR>\n");
/* Print the priority */
fprintf(output, "<TD ALIGN=LEFT VALIGN=TOP WIDTH=25%% NOWRAP>");
if (firsthost)
fprintf(output, "<FONT %s>PRIO %d</FONT>", xgetenv("XYMONPAGEROWFONT"), prio);
else
fprintf(output, " ");
fprintf(output, "</TD>\n");
/* Print the hostname with a link to the critical systems info page */
fprintf(output, "<TD ALIGN=LEFT>%s</TD>\n", hostnamehtml(itm->hostname, NULL, usetooltips));
key = (char *)malloc(strlen(itm->hostname) + 1024);
for (colhandle = rbtBegin(columns); (colhandle != rbtEnd(columns)); colhandle = rbtNext(columns, colhandle)) {
void *k1, *k2;
char *colname;
RbtIterator sthandle;
fprintf(output, "<TD ALIGN=CENTER>");
rbtKeyValue(columns, colhandle, &k1, &k2);
colname = (char *)k1;
sprintf(key, "%s|%s", itm->hostname, colname);
sthandle = rbtFind(rbstate, key);
if (sthandle == rbtEnd(rbstate)) {
fprintf(output, "-");
}
else {
hstatus_t *column;
char *htmlalttag;
char *htmlackstr;
rbtKeyValue(rbstate, sthandle, &k1, &k2);
column = (hstatus_t *)k2;
if (column->config->priority != prio)
fprintf(output, "-");
else {
time_t age = now - column->lastchange;
char *htmlgroupstr;
char *htmlextrastr;
htmlalttag = alttag(colname, column->color, 0, 1, agestring(age));
htmlackstr = (column->ackmsg ? column->ackmsg : "");
htmlgroupstr = strdup(urlencode(column->config->ttgroup ? column->config->ttgroup : ""));
htmlextrastr = strdup(urlencode(column->config->ttextra ? column->config->ttextra : ""));
fprintf(output, "<A HREF=\"%s&NKPRIO=%d&NKTTGROUP=%s&NKTTEXTRA=%s\">",
hostsvcurl(itm->hostname, colname, 1),
prio,
htmlgroupstr, htmlextrastr);
fprintf(output, "<IMG SRC=\"%s/%s\" ALT=\"%s\" TITLE=\"%s %s\" HEIGHT=\"%s\" WIDTH=\"%s\" BORDER=0></A>",
xgetenv("XYMONSKIN"),
dotgiffilename(column->color, (column->acktime > 0), (age > oldlimit)),
colorname(column->color), htmlalttag, htmlackstr,
xgetenv("DOTHEIGHT"), xgetenv("DOTWIDTH"));
xfree(htmlgroupstr);
xfree(htmlextrastr);
}
}
fprintf(output, "</TD>\n");
}
xfree(key);
fprintf(output, "</TR>\n");
}
int load_nkconfig(char *fn)
{
static void *configfiles = NULL;
static int firsttime = 1;
FILE *fd;
strbuffer_t *inbuf;
/* Setup the default configuration filename */
if (!fn) {
if (!defaultfn) {
char *bbhome = xgetenv("BBHOME");
defaultfn = (char *)malloc(strlen(bbhome) + strlen(DEFAULTCONFIG) + 2);
sprintf(defaultfn, "%s/%s", bbhome, DEFAULTCONFIG);
}
fn = defaultfn;
}
if (configfn) xfree(configfn);
configfn = strdup(fn);
/* First check if there were no modifications at all */
if (configfiles) {
if (!stackfmodified(configfiles)){
dbgprintf("No files modified, skipping reload of %s\n", fn);
return 0;
}
else {
stackfclist(&configfiles);
configfiles = NULL;
}
}
if (!firsttime) {
/* Clean up existing datatree */
RbtHandle handle;
void *k1, *k2;
for (handle = rbtBegin(rbconf); (handle != rbtEnd(rbconf)); handle = rbtNext(rbconf, handle)) {
rbtKeyValue(rbconf, handle, &k1, &k2);
flushrec(k1, k2);
}
rbtDelete(rbconf);
}
firsttime = 0;
rbconf = rbtNew(name_compare);
fd = stackfopen(fn, "r", &configfiles);
if (fd == NULL) return 1;
inbuf = newstrbuffer(0);
while (stackfgets(inbuf, NULL)) {
/* Full record : Host service START END TIMESPEC TTPrio TTGroup TTExtra */
/* Clone record: Host =HOST */
char *ehost, *eservice, *estart, *eend, *etime, *ttgroup, *ttextra, *updinfo;
int ttprio = 0;
nkconf_t *newitem;
RbtStatus status;
int idx = 0;
ehost = gettok(STRBUF(inbuf), "|\n"); if (!ehost) continue;
eservice = gettok(NULL, "|\n"); if (!eservice) continue;
if (*eservice == '=') {
char *key = (char *)malloc(strlen(ehost) + 2);
char *pointsto = strdup(eservice+1);
sprintf(key, "%s=", ehost);
status = rbtInsert(rbconf, key, pointsto);
}
else {
estart = gettok(NULL, "|\n"); if (!estart) continue;
eend = gettok(NULL, "|\n"); if (!eend) continue;
etime = gettok(NULL, "|\n"); if (!etime) continue;
ttprio = atoi(gettok(NULL, "|\n")); if (ttprio == 0) continue;
ttgroup = gettok(NULL, "|\n");
ttextra = gettok(NULL, "|\n");
updinfo = gettok(NULL, "|\n");
newitem = (nkconf_t *)malloc(sizeof(nkconf_t));
newitem->key = (char *)malloc(strlen(ehost) + strlen(eservice) + 15);
sprintf(newitem->key, "%s|%s", ehost, eservice);
newitem->starttime= ((estart && *estart) ? atoi(estart) : 0);
newitem->endtime = ((eend && *eend) ? atoi(eend) : 0);
newitem->nktime = ((etime && *etime) ? strdup(etime) : NULL);
newitem->priority = ttprio;
newitem->ttgroup = strdup(ttgroup);
newitem->ttextra = strdup(ttextra);
newitem->updinfo = strdup(updinfo);
status = rbtInsert(rbconf, newitem->key, newitem);
while (status == RBT_STATUS_DUPLICATE_KEY) {
idx++;
sprintf(newitem->key, "%s|%s|%d", ehost, eservice, idx);
status = rbtInsert(rbconf, newitem->key, newitem);
}
}
}
stackfclose(fd);
freestrbuffer(inbuf);
if (debug) {
RbtHandle handle;
handle = rbtBegin(rbconf);
while (handle != rbtEnd(rbconf)) {
void *k1, *k2;
rbtKeyValue(rbconf, handle, &k1, &k2);
printf("%s\n", (char *)k1);
handle = rbtNext(rbconf, handle);
}
}
return 0;
}
int update_nkconfig(nkconf_t *rec)
{
char *bakfn;
FILE *bakfd;
unsigned char buf[8192];
int n;
struct stat st;
struct utimbuf ut;
RbtHandle handle;
FILE *fd;
int result = 0;
/* First, copy the old file */
bakfn = (char *)malloc(strlen(configfn) + 5);
sprintf(bakfn, "%s.bak", configfn);
if (stat(configfn, &st) == 0) {
ut.actime = st.st_atime;
ut.modtime = st.st_mtime;
}
else ut.actime = ut.modtime = getcurrenttime(NULL);
fd = fopen(configfn, "r");
if (fd) {
bakfd = fopen(bakfn, "w");
if (bakfd) {
while ((n = fread(buf, 1, sizeof(buf), fd)) > 0) fwrite(buf, 1, n, bakfd);
fclose(bakfd);
utime(bakfn, &ut);
}
fclose(fd);
}
xfree(bakfn);
fd = fopen(configfn, "w");
if (fd == NULL) {
errprintf("Cannot open output file %s\n", configfn);
return 1;
}
if (rec) {
handle = rbtFind(rbconf, rec->key);
if (handle == rbtEnd(rbconf)) rbtInsert(rbconf, rec->key, rec);
}
handle = rbtBegin(rbconf);
while (handle != rbtEnd(rbconf)) {
void *k1, *k2;
char *onekey;
rbtKeyValue(rbconf, handle, &k1, &k2);
onekey = (char *)k1;
if (*(onekey + strlen(onekey) - 1) == '=') {
char *pointsto = (char *)k2;
char *hostname;
hostname = strdup(onekey);
*(hostname + strlen(hostname) - 1) = '\0';
fprintf(fd, "%s|=%s\n", hostname, pointsto);
}
else {
nkconf_t *onerec = (nkconf_t *)k2;
char startstr[20], endstr[20];
*startstr = *endstr = '\0';
if (onerec->starttime > 0) sprintf(startstr, "%d", (int)onerec->starttime);
if (onerec->endtime > 0) sprintf(endstr, "%d", (int)onerec->endtime);
fprintf(fd, "%s|%s|%s|%s|%d|%s|%s|%s\n",
onekey,
startstr, endstr,
(onerec->nktime ? onerec->nktime : ""),
onerec->priority,
(onerec->ttgroup ? onerec->ttgroup : ""),
(onerec->ttextra ? onerec->ttextra : ""),
(onerec->updinfo ? onerec->updinfo : ""));
}
handle = rbtNext(rbconf, handle);
}
fclose(fd);
return result;
}
nkconf_t *get_nkconfig(char *key, int flags, char **resultkey)
{
static RbtHandle handle;
static char *realkey = NULL;
void *k1, *k2;
nkconf_t *result = NULL;
int isclone;
if (resultkey) *resultkey = NULL;
switch (flags) {
case NKCONF_TIMEFILTER:
handle = findrec(key);
/* We may have hit a cloned record, so use the real key for further searches */
if (handle != rbtEnd(rbconf)) {
rbtKeyValue(rbconf, handle, &k1, &k2);
realkey = k1;
}
while (handle != rbtEnd(rbconf)) {
rbtKeyValue(rbconf, handle, &k1, &k2);
result = (nkconf_t *)k2;
if (timecheck(result->starttime, result->endtime, result->nktime)) return result;
/* Go to the next */
handle = rbtNext(rbconf, handle);
if (handle != rbtEnd(rbconf)) {
rbtKeyValue(rbconf, handle, &k1, &k2);
if (strncmp(realkey, ((nkconf_t *)k2)->key, strlen(realkey)) != 0) handle=rbtEnd(rbconf);
}
}
realkey = NULL;
break;
case NKCONF_FIRSTMATCH:
handle = findrec(key);
realkey = NULL;
if (handle != rbtEnd(rbconf)) {
rbtKeyValue(rbconf, handle, &k1, &k2);
realkey = (char *)k1;
}
break;
case NKCONF_FIRST:
realkey = NULL;
handle = rbtBegin(rbconf);
if (handle == rbtEnd(rbconf)) return NULL;
do {
rbtKeyValue(rbconf, handle, &k1, &k2);
realkey = (char *)k1;
isclone = (*(realkey + strlen(realkey) - 1) == '=');
if (isclone) handle = rbtNext(rbconf, handle);
} while (isclone && (handle != rbtEnd(rbconf)));
break;
case NKCONF_NEXT:
if (!realkey || (handle == rbtEnd(rbconf))) return NULL;
isclone = 1;
while (isclone && (handle != rbtEnd(rbconf))) {
handle = rbtNext(rbconf, handle);
if (handle) {
rbtKeyValue(rbconf, handle, &k1, &k2);
realkey = (char *)k1;
isclone = (*(realkey + strlen(realkey) - 1) == '=');
}
}
break;
case NKCONF_RAW_FIRST:
handle = rbtBegin(rbconf);
realkey = NULL;
break;
case NKCONF_RAW_NEXT:
handle = rbtNext(rbconf, handle);
realkey = NULL;
break;
case NKCONF_FIRSTHOSTMATCH:
do {
int found = 0;
char *delim;
realkey = NULL;
handle = rbtBegin(rbconf);
while (!found && (handle != rbtEnd(rbconf))) {
rbtKeyValue(rbconf, handle, &k1, &k2);
realkey = (char *)k1;
delim = realkey + strlen(key); /* OK even if past end of realkey */
found = ((strncmp(realkey, key, strlen(key)) == 0) &&
((*delim == '|') || (*delim == '=')));
if (!found) { handle = rbtNext(rbconf, handle); realkey = NULL; }
}
if ((handle != rbtEnd(rbconf)) && (*(realkey + strlen(realkey) - 1) == '=')) {
key = (char *)k2;
isclone = 1;
}
else isclone = 0;
} while (isclone && (handle != rbtEnd(rbconf)));
break;
}
if (handle == rbtEnd(rbconf)) { realkey = NULL; return NULL; }
rbtKeyValue(rbconf, handle, &k1, &k2);
if (resultkey) *resultkey = (char *)k1;
result = (nkconf_t *)k2;
return result;
}
int main(int argc, char *argv[])
{
int daemonize = 0;
char *pidfile = NULL;
char *envarea = NULL;
int cnid = -1;
pcre *msgfilter = NULL;
pcre *stdfilter = NULL;
int argi;
struct sigaction sa;
RbtIterator handle;
/* Dont save the error buffer */
save_errbuf = 0;
/* Create the peer container */
peers = rbtNew(name_compare);
for (argi=1; (argi < argc); argi++) {
if (argnmatch(argv[argi], "--debug")) {
debug = 1;
}
else if (argnmatch(argv[argi], "--channel=")) {
char *cn = strchr(argv[argi], '=') + 1;
for (cnid = C_STATUS; (channelnames[cnid] && strcmp(channelnames[cnid], cn)); cnid++) ;
if (channelnames[cnid] == NULL) cnid = -1;
}
else if (argnmatch(argv[argi], "--daemon")) {
daemonize = 1;
}
else if (argnmatch(argv[argi], "--no-daemon")) {
daemonize = 0;
}
else if (argnmatch(argv[argi], "--pidfile=")) {
char *p = strchr(argv[argi], '=');
pidfile = strdup(p+1);
}
else if (argnmatch(argv[argi], "--log=")) {
char *p = strchr(argv[argi], '=');
logfn = strdup(p+1);
}
else if (argnmatch(argv[argi], "--env=")) {
char *p = strchr(argv[argi], '=');
loadenv(p+1, envarea);
}
else if (argnmatch(argv[argi], "--area=")) {
char *p = strchr(argv[argi], '=');
envarea = strdup(p+1);
}
else if (argnmatch(argv[argi], "--locator=")) {
char *p = strchr(argv[argi], '=');
locator_init(p+1);
locatorbased = 1;
}
else if (argnmatch(argv[argi], "--service=")) {
char *p = strchr(argv[argi], '=');
locatorservice = get_servicetype(p+1);
}
else if (argnmatch(argv[argi], "--filter=")) {
char *p = strchr(argv[argi], '=');
msgfilter = compileregex(p+1);
if (!msgfilter) {
errprintf("Invalid filter (bad expression): %s\n", p+1);
}
else {
stdfilter = compileregex("^@@(logrotate|shutdown|drophost|droptest|renamehost|renametest)");
}
}
else {
char *childcmd;
char **childargs;
int i = 0;
childcmd = argv[argi];
childargs = (char **) calloc((1 + argc - argi), sizeof(char *));
while (argi < argc) { childargs[i++] = argv[argi++]; }
addlocalpeer(childcmd, childargs);
}
}
/* Sanity checks */
if (cnid == -1) {
errprintf("No channel/unknown channel specified\n");
return 1;
}
if (locatorbased && (locatorservice == ST_MAX)) {
errprintf("Must specify --service when using locator\n");
return 1;
}
if (!locatorbased && (rbtBegin(peers) == rbtEnd(peers))) {
errprintf("Must specify command for local worker\n");
return 1;
}
/* Do cache responses to avoid doing too many lookups */
if (locatorbased) locator_prepcache(locatorservice, 0);
/* Go daemon */
if (daemonize) {
/* Become a daemon */
pid_t daemonpid = fork();
if (daemonpid < 0) {
/* Fork failed */
errprintf("Could not fork child\n");
exit(1);
}
else if (daemonpid > 0) {
/* Parent creates PID file and exits */
FILE *fd = NULL;
if (pidfile) fd = fopen(pidfile, "w");
if (fd) {
fprintf(fd, "%d\n", (int)daemonpid);
fclose(fd);
}
exit(0);
}
/* Child (daemon) continues here */
setsid();
}
/* Catch signals */
setup_signalhandler("xymond_channel");
memset(&sa, 0, sizeof(sa));
sa.sa_handler = sig_handler;
sigaction(SIGINT, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
sigaction(SIGCHLD, &sa, NULL);
signal(SIGALRM, SIG_IGN);
/* Switch stdout/stderr to the logfile, if one was specified */
freopen("/dev/null", "r", stdin); /* xymond_channel's stdin is not used */
if (logfn) {
freopen(logfn, "a", stdout);
freopen(logfn, "a", stderr);
}
/* Attach to the channel */
channel = setup_channel(cnid, CHAN_CLIENT);
if (channel == NULL) {
errprintf("Channel not available\n");
running = 0;
}
while (running) {
/*
* Wait for GOCLIENT to go up.
*
* Note that we use IPC_NOWAIT if there are messages in the
* queue, because then we just want to pick up a message if
* there is one, and if not we want to continue pushing the
* queued data to the worker.
*/
struct sembuf s;
int n;
s.sem_num = GOCLIENT; s.sem_op = -1; s.sem_flg = ((pendingcount > 0) ? IPC_NOWAIT : 0);
n = semop(channel->semid, &s, 1);
if (n == 0) {
/*
* GOCLIENT went high, and so we got alerted about a new
* message arriving. Copy the message to our own buffer queue.
*/
char *inbuf = NULL;
if (!msgfilter || matchregex(channel->channelbuf, msgfilter) || matchregex(channel->channelbuf, stdfilter)) {
inbuf = strdup(channel->channelbuf);
}
/*
* Now we have safely stored the new message in our buffer.
* Wait until any other clients on the same channel have picked up
* this message (GOCLIENT reaches 0).
*
* We wrap this into an alarm handler, because it can occasionally
* fail, causing the whole system to lock up. We dont want that....
* We'll set the alarm to trigger after 1 second. Experience shows
* that we'll either succeed in a few milliseconds, or fail completely
* and wait the full alarm-timer duration.
*/
gotalarm = 0; signal(SIGALRM, sig_handler); alarm(2);
do {
s.sem_num = GOCLIENT; s.sem_op = 0; s.sem_flg = 0;
n = semop(channel->semid, &s, 1);
} while ((n == -1) && (errno == EAGAIN) && running && (!gotalarm));
signal(SIGALRM, SIG_IGN);
if (gotalarm) {
errprintf("Gave up waiting for GOCLIENT to go low.\n");
}
/*
* Let master know we got it by downing BOARDBUSY.
* This should not block, since BOARDBUSY is upped
* by the master just before he ups GOCLIENT.
*/
do {
s.sem_num = BOARDBUSY; s.sem_op = -1; s.sem_flg = IPC_NOWAIT;
n = semop(channel->semid, &s, 1);
} while ((n == -1) && (errno == EINTR));
if (n == -1) {
errprintf("Tried to down BOARDBUSY: %s\n", strerror(errno));
}
if (inbuf) {
/*
* See if they want us to rotate logs. We pass this on to
* the worker module as well, but must handle our own logfile.
*/
if (strncmp(inbuf, "@@logrotate", 11) == 0) {
freopen(logfn, "a", stdout);
freopen(logfn, "a", stderr);
}
/*
* Put the new message on our outbound queue.
*/
if (addmessage(inbuf) != 0) {
/* Failed to queue message, free the buffer */
xfree(inbuf);
}
}
}
else {
if (errno != EAGAIN) {
dbgprintf("Semaphore wait aborted: %s\n", strerror(errno));
continue;
}
}
/*
* We've picked up messages from the master. Now we
* must push them to the worker process. Since there
* is no way to hang off both a semaphore and select(),
* we'll just push as much data as possible into the
* pipe. If we get to a point where we would block,
* then wait a teeny bit of time and restart the
* whole loop with checking for new messages from the
* master etc.
*
* In theory, this could become an almost busy-wait loop.
* In practice, however, the queue will be empty most
* of the time because we'll just shove the data to the
* worker child.
*/
for (handle = rbtBegin(peers); (handle != rbtEnd(peers)); handle = rbtNext(peers, handle)) {
int canwrite = 1, hasfailed = 0;
xymon_peer_t *pwalk;
time_t msgtimeout = gettimer() - MSGTIMEOUT;
int flushcount = 0;
pwalk = (xymon_peer_t *) gettreeitem(peers, handle);
if (pwalk->msghead == NULL) continue; /* Ignore peers with nothing queued */
switch (pwalk->peerstatus) {
case P_UP:
canwrite = 1;
break;
case P_DOWN:
openconnection(pwalk);
canwrite = (pwalk->peerstatus == P_UP);
break;
case P_FAILED:
canwrite = 0;
break;
}
/* See if we have stale messages queued */
while (pwalk->msghead && (pwalk->msghead->tstamp < msgtimeout)) {
flushmessage(pwalk);
flushcount++;
}
if (flushcount) {
errprintf("Flushed %d stale messages for %s:%d\n",
flushcount,
inet_ntoa(pwalk->peeraddr.sin_addr),
ntohs(pwalk->peeraddr.sin_port));
}
while (pwalk->msghead && canwrite) {
fd_set fdwrite;
struct timeval tmo;
/* Check that this peer is ready for writing. */
FD_ZERO(&fdwrite); FD_SET(pwalk->peersocket, &fdwrite);
tmo.tv_sec = 0; tmo.tv_usec = 2000;
n = select(pwalk->peersocket+1, NULL, &fdwrite, NULL, &tmo);
if (n == -1) {
errprintf("select() failed: %s\n", strerror(errno));
canwrite = 0;
hasfailed = 1;
continue;
}
else if ((n == 0) || (!FD_ISSET(pwalk->peersocket, &fdwrite))) {
canwrite = 0;
continue;
}
n = write(pwalk->peersocket, pwalk->msghead->bufp, pwalk->msghead->buflen);
if (n >= 0) {
pwalk->msghead->bufp += n;
pwalk->msghead->buflen -= n;
if (pwalk->msghead->buflen == 0) flushmessage(pwalk);
}
else if (errno == EAGAIN) {
/*
* Write would block ... stop for now.
*/
canwrite = 0;
}
else {
hasfailed = 1;
}
if (hasfailed) {
/* Write failed, or message grew stale */
errprintf("Peer at %s:%d failed: %s\n",
inet_ntoa(pwalk->peeraddr.sin_addr), ntohs(pwalk->peeraddr.sin_port),
strerror(errno));
canwrite = 0;
shutdownconnection(pwalk);
if (pwalk->peertype == P_NET) locator_serverdown(pwalk->peername, locatorservice);
pwalk->peerstatus = P_FAILED;
}
}
}
}
/* Detach from channels */
close_channel(channel, CHAN_CLIENT);
/* Close peer connections */
for (handle = rbtBegin(peers); (handle != rbtEnd(peers)); handle = rbtNext(peers, handle)) {
xymon_peer_t *pwalk = (xymon_peer_t *) gettreeitem(peers, handle);
shutdownconnection(pwalk);
}
/* Remove the PID file */
if (pidfile) unlink(pidfile);
return 0;
}
int addmessage(char *inbuf)
{
RbtIterator phandle;
xymon_peer_t *peer;
int bcastmsg = 0;
int inlen = strlen(inbuf);
if (locatorbased) {
char *hostname, *hostend, *peerlocation;
/* xymond sends us messages with the KEY in the first field, between a '/' and a '|' */
hostname = inbuf + strcspn(inbuf, "/|\r\n");
if (*hostname != '/') {
errprintf("No key field in message, dropping it\n");
return -1; /* Malformed input */
}
hostname++;
bcastmsg = (*hostname == '*');
if (!bcastmsg) {
/* Lookup which server handles this host */
hostend = hostname + strcspn(hostname, "|\r\n");
if (*hostend != '|') {
errprintf("No delimiter found in input, dropping it\n");
return -1; /* Malformed input */
}
*hostend = '\0';
peerlocation = locator_query(hostname, locatorservice, NULL);
/*
* If we get no response, or an empty response,
* then there is no server capable of handling this
* request.
*/
if (!peerlocation || (*peerlocation == '\0')) {
errprintf("No response from locator for %s/%s, dropping it\n",
servicetype_names[locatorservice], hostname);
return -1;
}
*hostend = '|';
phandle = rbtFind(peers, peerlocation);
if (phandle == rbtEnd(peers)) {
/* New peer - register it */
addnetpeer(peerlocation);
phandle = rbtFind(peers, peerlocation);
}
}
}
else {
phandle = rbtFind(peers, "");
}
if (bcastmsg) {
for (phandle = rbtBegin(peers); (phandle != rbtEnd(peers)); phandle = rbtNext(peers, phandle)) {
peer = (xymon_peer_t *)gettreeitem(peers, phandle);
addmessage_onepeer(peer, inbuf, inlen);
}
}
else {
if (phandle == rbtEnd(peers)) {
errprintf("No peer found to handle message, dropping it\n");
return -1;
}
peer = (xymon_peer_t *)gettreeitem(peers, phandle);
addmessage_onepeer(peer, inbuf, inlen);
}
return 0;
}
