int
status_job (mbdReqType reqType,
struct jobCard *jp, int newStatus, sbdReplyType err)
{
static char fname[] = "status_job()";
static int seq = 1;
static char lastHost[MAXHOSTNAMELEN];
int reply;
char *request_buf;
char *reply_buf = NULL;
XDR xdrs;
struct LSFHeader hdr;
int cc;
struct statusReq statusReq;
int flags;
int i;
int len;
struct lsfAuth *auth = NULL;
if ((logclass & LC_TRACE) && (logclass & LC_SIGNAL))
ls_syslog (LOG_DEBUG, "%s: Entering ... regType %d jobId %s",
fname, reqType, lsb_jobid2str (jp->jobSpecs.jobId));
if (newStatus == JOB_STAT_EXIT)
{
jp->userJobSucc = FALSE;
}
if (MASK_STATUS (newStatus) == JOB_STAT_DONE)
{
jp->userJobSucc = TRUE;
}
if (IS_POST_FINISH (newStatus))
{
if (jp->userJobSucc != TRUE)
{
return 0;
}
}
if (masterHost == NULL)
return -1;
if (jp->notReported < 0)
{
jp->notReported = -INFINIT_INT;
return (0);
}
statusReq.jobId = jp->jobSpecs.jobId;
statusReq.actPid = jp->jobSpecs.actPid;
statusReq.jobPid = jp->jobSpecs.jobPid;
statusReq.jobPGid = jp->jobSpecs.jobPGid;
statusReq.newStatus = newStatus;
statusReq.reason = jp->jobSpecs.reasons;
statusReq.subreasons = jp->jobSpecs.subreasons;
statusReq.sbdReply = err;
statusReq.lsfRusage = jp->lsfRusage;
statusReq.execUid = jp->jobSpecs.execUid;
statusReq.numExecHosts = 0;
statusReq.execHosts = NULL;
statusReq.exitStatus = jp->w_status;
statusReq.execCwd = jp->jobSpecs.execCwd;
statusReq.execHome = jp->jobSpecs.execHome;
statusReq.execUsername = jp->execUsername;
statusReq.queuePostCmd = "";
statusReq.queuePreCmd = "";
statusReq.msgId = jp->delieveredMsgId;
if (IS_FINISH (newStatus))
{
if (jp->maxRusage.mem > jp->runRusage.mem)
jp->runRusage.mem = jp->maxRusage.mem;
if (jp->maxRusage.swap > jp->runRusage.swap)
jp->runRusage.swap = jp->maxRusage.swap;
if (jp->maxRusage.stime > jp->runRusage.stime)
jp->runRusage.stime = jp->maxRusage.stime;
if (jp->maxRusage.utime > jp->runRusage.utime)
jp->runRusage.utime = jp->maxRusage.utime;
}
statusReq.runRusage.mem = jp->runRusage.mem;
statusReq.runRusage.swap = jp->runRusage.swap;
statusReq.runRusage.utime = jp->runRusage.utime;
statusReq.runRusage.stime = jp->runRusage.stime;
statusReq.runRusage.npids = jp->runRusage.npids;
statusReq.runRusage.pidInfo = jp->runRusage.pidInfo;
statusReq.runRusage.npgids = jp->runRusage.npgids;
statusReq.runRusage.pgid = jp->runRusage.pgid;
statusReq.actStatus = jp->actStatus;
statusReq.sigValue = jp->jobSpecs.actValue;
statusReq.seq = seq;
seq++;
if (seq >= MAX_SEQ_NUM)
seq = 1;
len = 1024 + ALIGNWORD_ (sizeof (struct statusReq));
len += ALIGNWORD_ (strlen (statusReq.execHome)) + 4 +
ALIGNWORD_ (strlen (statusReq.execCwd)) + 4 +
ALIGNWORD_ (strlen (statusReq.execUsername)) + 4;
for (i = 0; i < statusReq.runRusage.npids; i++)
len += ALIGNWORD_ (sizeof (struct pidInfo)) + 4;
for (i = 0; i < statusReq.runRusage.npgids; i++)
len += ALIGNWORD_ (sizeof (int)) + 4;
if (logclass & (LC_TRACE | LC_COMM))
ls_syslog (LOG_DEBUG, "%s: The length of the job message is: <%d>", fname,
len);
if ((request_buf = malloc (len)) == NULL)
{
ls_syslog (LOG_ERR, I18N_FUNC_FAIL_M, fname, "malloc");
return (-1);
}
xdrmem_create (&xdrs, request_buf, len, XDR_ENCODE);
initLSFHeader_ (&hdr);
hdr.opCode = reqType;
if (!xdr_encodeMsg (&xdrs, (char *) &statusReq, &hdr, xdr_statusReq, 0,
auth))
{
ls_syslog (LOG_ERR, I18N_JOB_FAIL_S_M,
fname, lsb_jobid2str (jp->jobSpecs.jobId), "xdr_statusReq");
lsb_merr2 (I18N_FUNC_FAIL, fname, "xdr_statusReq");
xdr_destroy (&xdrs);
FREEUP (request_buf);
relife ();
}
flags = CALL_SERVER_NO_HANDSHAKE;
if (statusChan >= 0)
flags |= CALL_SERVER_USE_SOCKET;
if (reqType == BATCH_RUSAGE_JOB)
flags |= CALL_SERVER_NO_WAIT_REPLY;
if (logclass & LC_COMM)
ls_syslog (LOG_DEBUG1, "%s: before call_server statusChan=%d flags=%d",
fname, statusChan, flags);
cc = call_server (masterHost,
mbd_port,
request_buf,
XDR_GETPOS (&xdrs),
&reply_buf,
&hdr,
connTimeout, readTimeout, &statusChan, NULL, NULL, flags);
if (cc < 0)
{
statusChan = -1;
if (!equalHost_ (masterHost, lastHost))
{
if (errno != EINTR)
ls_syslog (LOG_DEBUG,
"%s: Failed to reach mbatchd on host <%s> for job <%s>: %s",
fname, masterHost, lsb_jobid2str (jp->jobSpecs.jobId),
lsb_sysmsg ());
strcpy (lastHost, masterHost);
}
xdr_destroy (&xdrs);
FREEUP (request_buf);
failcnt++;
return (-1);
}
else if (cc == 0)
{
}
failcnt = 0;
lastHost[0] = '\0';
xdr_destroy (&xdrs);
FREEUP (request_buf);
if (cc)
free (reply_buf);
if (flags & CALL_SERVER_NO_WAIT_REPLY)
{
struct timeval timeval;
timeval.tv_sec = 0;
timeval.tv_usec = 0;
if (rd_select_ (chanSock_ (statusChan), &timeval) == 0)
{
jp->needReportRU = FALSE;
jp->lastStatusMbdTime = now;
return 0;
}
CLOSECD (statusChan);
if (logclass & LC_COMM)
ls_syslog (LOG_DEBUG1,
"%s: Job <%s> rd_select() failed, assume connection broken",
fname, lsb_jobid2str (jp->jobSpecs.jobId));
return (-1);
}
reply = hdr.opCode;
switch (reply)
{
case LSBE_NO_ERROR:
case LSBE_LOCK_JOB:
jp->needReportRU = FALSE;
jp->lastStatusMbdTime = now;
if (reply == LSBE_LOCK_JOB)
{
if (IS_SUSP (jp->jobSpecs.jStatus))
jp->jobSpecs.reasons |= SUSP_MBD_LOCK;
else
ls_syslog (LOG_ERR, _i18n_msg_get (ls_catd, NL_SETN, 5204, "%s: Job <%s> is in status <%x> and mbatchd wants to lock it, ignored."), /* catgets 5204 */
fname,
lsb_jobid2str (jp->jobSpecs.jobId),
jp->jobSpecs.jStatus);
}
return (0);
case LSBE_NO_JOB:
if (!IS_POST_FINISH (jp->jobSpecs.jStatus))
{
ls_syslog (LOG_ERR, _i18n_msg_get (ls_catd, NL_SETN, 5205, "%s: Job <%s> is forgotten by mbatchd on host <%s>, ignored."), fname, lsb_jobid2str (jp->jobSpecs.jobId), masterHost); /* catgets 5205 */
}
jp->notReported = -INFINIT_INT;
return (0);
case LSBE_STOP_JOB:
if (jobsig (jp, SIGSTOP, TRUE) < 0)
SET_STATE (jp->jobSpecs.jStatus, JOB_STAT_EXIT);
else
{
SET_STATE (jp->jobSpecs.jStatus, JOB_STAT_USUSP);
jp->jobSpecs.reasons |= SUSP_USER_STOP;
}
return (-1);
case LSBE_SBATCHD:
ls_syslog (LOG_ERR, _i18n_msg_get (ls_catd, NL_SETN, 5206, "%s: mbatchd on host <%s> doesn't think I'm configured as a batch server when I report the status for job <%s>"), /* catgets 5206 */
fname, masterHost, lsb_jobid2str (jp->jobSpecs.jobId));
return (-1);
default:
ls_syslog (LOG_ERR, _i18n_msg_get (ls_catd, NL_SETN, 5207, "%s: Illegal reply code <%d> from mbatchd on host <%s> for job <%s>"), /* catgets 5207 */
fname,
reply, masterHost, lsb_jobid2str (jp->jobSpecs.jobId));
return (-1);
}
}
void
die(int sig)
{
static char fname[] = "die";
char myhost[MAXHOSTNAMELEN];
if (debug > 1)
fprintf(stderr, "%s: signal %d\n",
fname,
sig);
if (masterme) {
releaseElogLock();
}
if (gethostname(myhost, MAXHOSTNAMELEN) <0) {
ls_syslog(LOG_ERR, I18N_FUNC_S_FAIL_M, fname, "gethostname", myhost);
strcpy(myhost, "localhost");
}
if (sig > 0 && sig < 100) {
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 8216,
"Daemon on host <%s> received signal <%d>; exiting"), /* catgets 8216 */
myhost, sig);
} else {
switch (sig) {
case MASTER_RESIGN:
ls_syslog(LOG_INFO, (_i18n_msg_get(ls_catd , NL_SETN, 8272, "Master daemon on host <%s> resigned; exiting")), myhost); /* catgets 8272 */
break;
case MASTER_RECONFIG:
ls_syslog(LOG_INFO, (_i18n_msg_get(ls_catd , NL_SETN, 8273, "Master daemon on host <%s> exiting for reconfiguration")), myhost); /* catgets 8273 */
break;
case SLAVE_MEM:
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 8217,
"Slave daemon on host <%s> failed in memory allocation; fatal error - exiting"), myhost); /* catgets 8217 */
lsb_merr1(_i18n_msg_get(ls_catd , NL_SETN, 8217,
"Slave daemon on host <%s> failed in memory allocation; fatal error - exiting"), myhost);
break;
case MASTER_MEM:
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 8218,
"Master daemon on host <%s> failed in memory allocation; fatal error - exiting"), myhost); /* catgets 8218 */
break;
case SLAVE_FATAL:
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 8219,
"Slave daemon on host <%s> dying; fatal error - see above messages for reason"), myhost); /* catgets 8219 */
break;
case MASTER_FATAL:
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 8220,
"Master daemon on host <%s> dying; fatal error - see above messages for reason"), myhost); /* catgets 8220 */
break;
case MASTER_CONF:
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 8221,
"Master daemon on host <%s> died of bad configuration file"), myhost); /* catgets 8221 */
break;
case SLAVE_RESTART:
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 8222,
"Slave daemon on host <%s> restarting"), myhost); /* catgets 8222 */
break;
case SLAVE_SHUTDOWN:
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 8223,
"Slave daemon on host <%s> shutdown"), myhost); /* catgets 8223 */
break;
default:
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 8224,
"Daemon on host <%s> exiting; cause code <%d> unknown"), myhost, sig); /* catgets 8224 */
break;
}
}
shutdown(chanSock_(batchSock), 2);
exit(sig);
}
int
ls_initrex(int num, int options)
{
struct servent *sv;
if (geteuid() == 0)
rootuid_ = TRUE;
if (initenv_(NULL, NULL)<0) {
if (rootuid_ && !(options & KEEPUID))
lsfSetUid(getuid());
return(-1);
}
inithostsock_();
lsQueueInit_(&requestQ, lsReqCmp_, NULL);
if (requestQ == NULL) {
lserrno = LSE_MALLOC;
return(-1);
}
res_addr_.sin_family = AF_INET;
if (genParams_[LSF_RES_PORT].paramValue) {
if ((res_addr_.sin_port = atoi(genParams_[LSF_RES_PORT].paramValue))
!= 0)
res_addr_.sin_port = htons(res_addr_.sin_port);
else
goto res_init_fail;
} else if (genParams_[LSF_RES_DEBUG].paramValue) {
res_addr_.sin_port = htons(RES_PORT);
} else {
# if defined(_COMPANY_X_)
if ((res_addr_.sin_port =
get_port_number(RES_SERVICE,(char *)NULL)) == -1) {
# else
if ((sv = getservbyname("res", "tcp")) != NULL)
res_addr_.sin_port = sv->s_port;
else {
# endif
res_init_fail:
lserrno = LSE_RES_NREG;
if (rootuid_ && !(options & KEEPUID))
lsfSetUid(getuid());
return (-1);
}
}
initconntbl_();
FD_ZERO(&connection_ok_);
if ((rootuid_) && (genParams_[LSF_AUTH].paramValue == NULL)) {
int i;
i = opensocks_(num);
if (!(options & KEEPUID))
lsfSetUid(getuid());
return (i);
} else {
return (num);
}
}
int
opensocks_(int num)
{
static char fname[] = "opensocks_";
int s;
int nextdescr;
int i;
totsockets_ = (num <= 0 || num > MAXCONNECT) ? LSF_DEFAULT_SOCKS : num;
if (logclass & LC_COMM)
ls_syslog(LOG_DEBUG,"%s: try to allocate num <%d> of socks",fname,num);
nextdescr = FIRST_RES_SOCK;
for (i = 0; i < totsockets_; i++) {
if ((s = CreateSock_(SOCK_STREAM)) < 0) {
if (logclass & LC_COMM)
ls_syslog(LOG_DEBUG,
"%s: CreateSock_ failed, iter:<%d> %s",
fname,i,strerror(errno));
totsockets_ = i;
if (i > 0) {
break;
} else {
return(-1);
}
}
if (s != nextdescr) {
if (dup2(s,nextdescr) < 0) {
if (logclass & LC_COMM)
ls_syslog(LOG_DEBUG,
"%s: dup2() failed, old:<%d>, new<%d>, iter:<%d> %s",
fname,s,nextdescr,i,strerror(errno));
close(s);
lserrno = LSE_SOCK_SYS;
totsockets_ = i;
if (i > 0)
break;
else
return (-1);
}
#if defined(FD_CLOEXEC)
fcntl(nextdescr, F_SETFD, (fcntl(nextdescr, F_GETFD)
| FD_CLOEXEC)) ;
#else
#if defined(FIOCLEX)
(void)ioctl(nextdescr, FIOCLEX, (char *)NULL);
#endif
#endif
close(s);
}
nextdescr++;
}
currentsocket_ = FIRST_RES_SOCK;
if (logclass & LC_COMM)
ls_syslog(LOG_DEBUG,"%s: returning num=<%d>",fname,totsockets_);
return (totsockets_);
}
/* ls_fdbusy()
*/
int
ls_fdbusy(int fd)
{
sTab hashSearchPtr;
hEnt *hEntPtr;
if (fd == chanSock_(limchans_[PRIMARY])
|| fd == chanSock_(limchans_[MASTER])
|| fd == chanSock_(limchans_[UNBOUND]))
return TRUE;
if (fd == cli_nios_fd[0])
return TRUE;
hEntPtr = h_firstEnt_(&conn_table, &hashSearchPtr);
while (hEntPtr) {
int *pfd;
pfd = hEntPtr->hData;
if (fd == pfd[0]
|| fd == pfd[1])
return (TRUE);
hEntPtr = h_nextEnt_(&hashSearchPtr);
}
if (rootuid_
&& fd >= currentsocket_
&& fd < FIRST_RES_SOCK + totsockets_)
return TRUE;
return FALSE;
}
