int add_walltime_remaining(
int index,
attribute *pattr,
tlist_head *phead)
{
int len;
char buf[MAXPATHLEN];
char *pname;
svrattrl *pal;
resource *pres;
int found = 0;
unsigned long remaining = 0;
unsigned long upperBound = 0;
/* encode walltime remaining, this is custom because walltime
* remaining isn't an attribute */
if ((pattr + JOB_ATR_state)->at_val.at_char != 'R')
{
/* only for running jobs, do nothing */
return(PBSE_NONE);
}
if (((pattr + JOB_ATR_resource)->at_val.at_list.ll_next != NULL) &&
((pattr + JOB_ATR_resource)->at_flags & ATR_VFLAG_SET))
{
pres = (resource *)GET_NEXT((pattr + JOB_ATR_resource)->at_val.at_list);
if ((pattr + JOB_ATR_comp_time)->at_flags & ATR_VFLAG_SET)
upperBound = (pattr + JOB_ATR_comp_time)->at_val.at_long;
else
upperBound = (unsigned long)time_now;
/* find the walltime resource */
for (;pres != NULL;pres = (resource *)GET_NEXT(pres->rs_link))
{
pname = pres->rs_defin->rs_name;
if (strcmp(pname, "walltime") == 0)
{
/* found walltime */
unsigned long value = (unsigned long)pres->rs_value.at_val.at_long;
remaining = value - (time_now - (pattr + index)->at_val.at_long);
found = TRUE;
break;
}
}
}
if (found == TRUE)
{
snprintf(buf,sizeof(buf),"%lu",remaining);
len = strlen(buf) + 1;
pal = attrlist_create("Walltime","Remaining",len);
if (pal != NULL)
{
memcpy(pal->al_value,buf,len);
pal->al_flags = ATR_VFLAG_SET;
append_link(phead,&pal->al_link,pal);
}
}
return(PBSE_NONE);
} /* END add_walltime_remaining() */
int encode_svrstate(
pbs_attribute *pattr, /* ptr to pbs_attribute */
tlist_head *phead, /* head of attrlist list */
const char *atname, /* pbs_attribute name */
const char *rsname, /* null */
int mode, /* encode mode */
int perm) /* only used for resources */
{
svrattrl *pal;
const char *psname;
if (pattr == NULL)
{
/* FAILURE */
return(-1);
}
if ((mode == ATR_ENCODE_SAVE) ||
(pattr->at_val.at_long <= SV_STATE_DOWN) ||
(pattr->at_val.at_long > SV_STATE_SHUTSIG))
{
/* SUCCESS */
return(0); /* don't bother to encode it */
}
psname = svr_state_names[pattr->at_val.at_long];
if (pattr->at_val.at_long == SV_STATE_RUN)
{
pthread_mutex_lock(server.sv_attr_mutex);
if (server.sv_attr[SRV_ATR_scheduling].at_val.at_long == 0)
psname = svr_idle;
else
{
pthread_mutex_lock(scheduler_sock_jobct_mutex);
if (scheduler_sock != -1)
psname = svr_sched;
pthread_mutex_unlock(scheduler_sock_jobct_mutex);
}
pthread_mutex_unlock(server.sv_attr_mutex);
}
pal = attrlist_create(atname, rsname, strlen(psname) + 1);
if (pal == NULL)
{
/* FAILURE */
return(-1);
}
strcpy(pal->al_value, psname);
pal->al_flags = pattr->at_flags;
append_link(phead, &pal->al_link, pal);
/* SUCCESS */
return(1);
} /* END encode_svrstate() */
int
decode_DIS_svrattrl(int sock, tlist_head *phead)
{
unsigned int i;
unsigned int hasresc;
size_t ls;
unsigned int data_len;
unsigned int numattr;
svrattrl *psvrat = NULL;
int rc;
size_t tsize;
numattr = disrui(sock, &rc); /* number of attributes in set */
if (rc) return rc;
for (i = 0; i < numattr; ++i)
{
data_len = disrui(sock, &rc); /* here it is used */
if (rc) return rc;
tsize = sizeof(svrattrl) + data_len;
if ((psvrat = (svrattrl *)malloc(tsize)) == 0)
return DIS_NOMALLOC;
CLEAR_LINK(psvrat->al_link);
psvrat->al_atopl.next = 0;
psvrat->al_tsize = tsize;
psvrat->al_name = (char *)psvrat + sizeof(svrattrl);
psvrat->al_resc = 0;
psvrat->al_value = 0;
psvrat->al_nameln = 0;
psvrat->al_rescln = 0;
psvrat->al_valln = 0;
psvrat->al_flags = 0;
if ((rc = disrfcs(sock, &ls, data_len, psvrat->al_name)))
break;
*(psvrat->al_name + ls++) = '\0';
psvrat->al_nameln = (int)ls;
data_len -= ls;
hasresc = disrui(sock, &rc);
if (rc)
break;
if (hasresc)
{
psvrat->al_resc = psvrat->al_name + ls;
if ((rc = disrfcs(sock, &ls, data_len, psvrat->al_resc)))
break;
*(psvrat->al_resc + ls++) = '\0';
psvrat->al_rescln = (int)ls;
data_len -= ls;
}
psvrat->al_value = psvrat->al_name + psvrat->al_nameln +
psvrat->al_rescln;
if ((rc = disrfcs(sock, &ls, data_len, psvrat->al_value)))
break;
*(psvrat->al_value + ls++) = '\0';
psvrat->al_valln = (int)ls;
psvrat->al_op = (enum batch_op)disrui(sock, &rc);
if (rc) break;
append_link(phead, &psvrat->al_link, psvrat);
}
if (rc)
{
(void)free(psvrat);
}
return (rc);
}
static int status_que(
pbs_queue *pque, /* ptr to que to status */
struct batch_request *preq,
tlist_head *pstathd) /* head of list to append status to */
{
struct brp_status *pstat;
svrattrl *pal;
int bad = 0;
if ((preq->rq_perm & ATR_DFLAG_RDACC) == 0)
{
return(PBSE_PERM);
}
/* ok going to do status, update count and state counts from qu_qs */
pque->qu_attr[QA_ATR_TotalJobs].at_val.at_long = pque->qu_numjobs;
pque->qu_attr[QA_ATR_TotalJobs].at_flags |= ATR_VFLAG_SET;
update_state_ct(
&pque->qu_attr[QA_ATR_JobsByState],
pque->qu_njstate,
pque->qu_jobstbuf);
/* allocate status sub-structure and fill in header portion */
pstat = (struct brp_status *)calloc(1, sizeof(struct brp_status));
if (pstat == NULL)
{
return(PBSE_SYSTEM);
}
memset(pstat, 0, sizeof(struct brp_status));
pstat->brp_objtype = MGR_OBJ_QUEUE;
strcpy(pstat->brp_objname, pque->qu_qs.qu_name);
CLEAR_LINK(pstat->brp_stlink);
CLEAR_HEAD(pstat->brp_attr);
append_link(pstathd, &pstat->brp_stlink, pstat);
/* add attributes to the status reply */
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
if (status_attrib(
pal,
que_attr_def,
pque->qu_attr,
QA_ATR_LAST,
preq->rq_perm,
&pstat->brp_attr,
&bad,
1) != 0) /* IsOwner == TRUE */
{
return(bad);
}
return(PBSE_NONE);
} /* END status_que() */
int req_stat_svr(
struct batch_request *preq) /* ptr to the decoded request */
{
svrattrl *pal;
struct batch_reply *preply;
struct brp_status *pstat;
int bad = 0;
char nc_buf[128];
int numjobs;
int netrates[3];
memset(netrates, 0, sizeof(netrates));
/* update count and state counts from sv_numjobs and sv_jobstates */
lock_sv_qs_mutex(server.sv_qs_mutex, __func__);
numjobs = server.sv_qs.sv_numjobs;
unlock_sv_qs_mutex(server.sv_qs_mutex, __func__);
pthread_mutex_lock(server.sv_attr_mutex);
server.sv_attr[SRV_ATR_TotalJobs].at_val.at_long = numjobs;
server.sv_attr[SRV_ATR_TotalJobs].at_flags |= ATR_VFLAG_SET;
pthread_mutex_lock(server.sv_jobstates_mutex);
update_state_ct(
&server.sv_attr[SRV_ATR_JobsByState],
server.sv_jobstates,
server.sv_jobstbuf);
pthread_mutex_unlock(server.sv_jobstates_mutex);
netcounter_get(netrates);
snprintf(nc_buf, 127, "%d %d %d", netrates[0], netrates[1], netrates[2]);
if (server.sv_attr[SRV_ATR_NetCounter].at_val.at_str != NULL)
free(server.sv_attr[SRV_ATR_NetCounter].at_val.at_str);
server.sv_attr[SRV_ATR_NetCounter].at_val.at_str = strdup(nc_buf);
if (server.sv_attr[SRV_ATR_NetCounter].at_val.at_str != NULL)
server.sv_attr[SRV_ATR_NetCounter].at_flags |= ATR_VFLAG_SET;
pthread_mutex_unlock(server.sv_attr_mutex);
/* allocate a reply structure and a status sub-structure */
preply = &preq->rq_reply;
preply->brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preply->brp_un.brp_status);
pstat = (struct brp_status *)calloc(1, sizeof(struct brp_status));
if (pstat == NULL)
{
reply_free(preply);
req_reject(PBSE_SYSTEM, 0, preq, NULL, NULL);
pthread_mutex_unlock(server.sv_attr_mutex);
return(PBSE_SYSTEM);
}
CLEAR_LINK(pstat->brp_stlink);
strcpy(pstat->brp_objname, server_name);
pstat->brp_objtype = MGR_OBJ_SERVER;
CLEAR_HEAD(pstat->brp_attr);
append_link(&preply->brp_un.brp_status, &pstat->brp_stlink, pstat);
/* add attributes to the status reply */
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
if (status_attrib(
pal,
svr_attr_def,
server.sv_attr,
SRV_ATR_LAST,
preq->rq_perm,
&pstat->brp_attr,
&bad,
1)) /* IsOwner == TRUE */
{
reply_badattr(PBSE_NOATTR, bad, pal, preq);
}
else
{
reply_send_svr(preq);
}
return(PBSE_NONE);
} /* END req_stat_svr() */
void req_commit(
struct batch_request *preq) /* I */
{
job *pj;
pj = locate_new_job(preq->rq_conn, preq->rq_ind.rq_commit);
if (LOGLEVEL >= 6)
{
log_record(
PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
(pj != NULL) ? pj->ji_qs.ji_jobid : "NULL",
"committing job");
}
if (pj == NULL)
{
req_reject(PBSE_UNKJOBID, 0, preq, NULL, NULL);
return;
}
if (pj->ji_qs.ji_substate != JOB_SUBSTATE_TRANSICM)
{
log_err(errno, "req_commit", "cannot commit job in unexpected state");
req_reject(PBSE_IVALREQ, 0, preq, NULL, NULL);
return;
}
/* move job from new job list to "all" job list, set to running state */
delete_link(&pj->ji_alljobs);
append_link(&svr_alljobs, &pj->ji_alljobs, pj);
/*
** Set JOB_SVFLG_HERE to indicate that this is Mother Superior.
*/
pj->ji_qs.ji_svrflags |= JOB_SVFLG_HERE;
pj->ji_qs.ji_state = JOB_STATE_RUNNING;
pj->ji_qs.ji_substate = JOB_SUBSTATE_PRERUN;
pj->ji_qs.ji_un_type = JOB_UNION_TYPE_MOM;
pj->ji_qs.ji_un.ji_momt.ji_svraddr = get_connectaddr(preq->rq_conn);
pj->ji_qs.ji_un.ji_momt.ji_exitstat = 0;
/* For MOM - start up the job (blocks) */
if (LOGLEVEL >= 6)
{
log_record(
PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
(pj != NULL) ? pj->ji_qs.ji_jobid : "NULL",
"starting job execution");
}
start_exec(pj);
if (LOGLEVEL >= 6)
{
log_record(
PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
(pj != NULL) ? pj->ji_qs.ji_jobid : "NULL",
"job execution started");
}
/* if start request fails, reply with failure string */
if (pj->ji_qs.ji_substate == JOB_SUBSTATE_EXITING)
{
char tmpLine[1024];
if ((pj->ji_hosts != NULL) &&
(pj->ji_nodekill >= 0) &&
(pj->ji_hosts[pj->ji_nodekill].hn_host != NULL))
{
sprintf(tmpLine, "start failed on node %s",
pj->ji_hosts[pj->ji_nodekill].hn_host);
}
else
{
sprintf(tmpLine, "start failed on unknown node");
}
if (LOGLEVEL >= 6)
{
log_record(
PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
(pj != NULL) ? pj->ji_qs.ji_jobid : "NULL",
tmpLine);
}
reply_text(preq, 0, tmpLine);
}
else
{
reply_jobid(preq, pj->ji_qs.ji_jobid, BATCH_REPLY_CHOICE_Commit);
}
job_save(pj, SAVEJOB_FULL);
/* NOTE: we used to flag JOB_ATR_errpath, JOB_ATR_outpath,
* JOB_ATR_session_id, and JOB_ATR_altid as modified at this point to make sure
* pbs_server got these attr values. This worked fine before TORQUE modified
* job launched into an async process. At 2.0.0p6, a new attribute "SEND" flag
* was added to handle this process. */
return;
} /* END req_commit() */
int encode_state(
pbs_attribute *pattr, /*struct pbs_attribute being encoded */
tlist_head *ph, /*head of a list of "svrattrl" structs
which are to be returned*/
char *aname, /*pbs_attribute's name */
char *rname, /*resource's name (null if none) */
int mode, /*mode code, unused here */
int perm) /* only used for resources */
{
int i;
svrattrl *pal;
short state;
char state_str[MAX_ENCODE_BFR];
if (!pattr)
{
return -(PBSE_INTERNAL);
}
if (!(pattr->at_flags & ATR_VFLAG_SET))
{
/* SUCCESS - pbs_attribute not set */
return(0);
}
state = pattr->at_val.at_short & (INUSE_SUBNODE_MASK | INUSE_UNKNOWN);
if (!state)
{
strcpy(state_str, ND_free);
}
else
{
state_str[0] = '\0';
for (i = 0;ns[i].name;i++)
{
if (state & ns[i].bit)
{
if (state_str[0] != '\0')
strcat(state_str, ",");
strcat(state_str, ns[i].name);
}
}
}
pal = attrlist_create(aname, rname, (int)strlen(state_str) + 1);
if (pal == NULL)
{
return -(PBSE_SYSTEM);
}
strcpy(pal->al_value, state_str);
pal->al_flags = ATR_VFLAG_SET;
append_link(ph, &pal->al_link, pal);
/* SUCCESS */
return(0);
} /* END encode_state */
int
status_job(job *pjob, struct batch_request *preq, svrattrl *pal, pbs_list_head *pstathd, int *bad)
{
struct brp_status *pstat;
time_t tm;
long oldtime = 0;
int old_elig_flags = 0;
int old_atyp_flags = 0;
/* see if the client is authorized to status this job */
if (! server.sv_attr[(int)SRV_ATR_query_others].at_val.at_long)
if (svr_authorize_jobreq(preq, pjob))
return (PBSE_PERM);
if (pjob->ji_qs.ji_svrflags & JOB_SVFLG_ArrayJob) {
/* for Array Job, if array_indices_remaining is modified */
/* then need to recalculate the string value */
char *pnewstr;
attribute *premain;
premain = &pjob->ji_wattr[(int)JOB_ATR_array_indices_remaining];
if (premain->at_flags & ATR_VFLAG_MODCACHE) {
pnewstr = cvt_range(pjob->ji_ajtrk, JOB_STATE_QUEUED);
if (pnewstr == NULL)
pnewstr = "-";
job_attr_def[JOB_ATR_array_indices_remaining].at_free(premain);
job_attr_def[JOB_ATR_array_indices_remaining].at_decode(premain, 0, 0, pnewstr);
/* also update value of attribute "array_state_count" */
update_subjob_state_ct(pjob);
}
}
/* calc eligible time on the fly and return, don't save. */
if (server.sv_attr[(int)SRV_ATR_EligibleTimeEnable].at_val.at_long != 0) {
if (pjob->ji_wattr[(int)JOB_ATR_accrue_type].at_val.at_long == JOB_ELIGIBLE) {
time(&tm);
oldtime = pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_val.at_long;
pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_val.at_long += ((long)tm - pjob->ji_wattr[(int)JOB_ATR_sample_starttime].at_val.at_long);
pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_flags |= ATR_VFLAG_MODCACHE;
/* Note: ATR_VFLAG_MODCACHE must be set because of svr_cached() does */
/* not correctly check ATR_VFLAG_SET */
}
} else {
/* eligible_time_enable is off so, */
/* clear set flag so that eligible_time and accrue type dont show */
old_elig_flags = pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_flags;
pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_flags &= ~ATR_VFLAG_SET;
pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_flags |= (ATR_VFLAG_MODCACHE);
old_atyp_flags = pjob->ji_wattr[(int)JOB_ATR_accrue_type].at_flags;
pjob->ji_wattr[(int)JOB_ATR_accrue_type].at_flags &= ~ATR_VFLAG_SET;
pjob->ji_wattr[(int)JOB_ATR_accrue_type].at_flags |= (ATR_VFLAG_MODCACHE);
/* Note: ATR_VFLAG_MODCACHE must be set because of svr_cached() does */
/* not correctly check ATR_VFLAG_SET */
}
/* allocate reply structure and fill in header portion */
pstat = (struct brp_status *)malloc(sizeof(struct brp_status));
if (pstat == (struct brp_status *)0)
return (PBSE_SYSTEM);
CLEAR_LINK(pstat->brp_stlink);
pstat->brp_objtype = MGR_OBJ_JOB;
(void)strcpy(pstat->brp_objname, pjob->ji_qs.ji_jobid);
CLEAR_HEAD(pstat->brp_attr);
append_link(pstathd, &pstat->brp_stlink, pstat);
/* add attributes to the status reply */
*bad = 0;
if (status_attrib(pal, job_attr_def, pjob->ji_wattr, JOB_ATR_LAST,
preq->rq_perm, &pstat->brp_attr, bad))
return (PBSE_NOATTR);
/* reset eligible time, it was calctd on the fly, real calctn only when accrue_type changes */
if (server.sv_attr[(int)SRV_ATR_EligibleTimeEnable].at_val.at_long != 0) {
if (pjob->ji_wattr[(int)JOB_ATR_accrue_type].at_val.at_long == JOB_ELIGIBLE) {
pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_val.at_long = oldtime;
pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_flags |= ATR_VFLAG_MODCACHE;
/* Note: ATR_VFLAG_MODCACHE must be set because of svr_cached() does */
/* not correctly check ATR_VFLAG_SET */
}
} else {
/* reset the set flags */
pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_flags = old_elig_flags;
pjob->ji_wattr[(int)JOB_ATR_accrue_type].at_flags = old_atyp_flags;
}
return (0);
}
/**
* @brief
* status_subjob - status a single subjob (of an Array Job)
* Works by statusing the parrent unless subjob is actually running.
*
* @param[in,out] pjob - ptr to parent Array
* @param[in] preq - request structure
* @param[in] pal - specific attributes to status
* @param[in] subj - if not = -1 then include subjob [n]
* @param[in,out] pstathd - RETURN: head of list to append status to
* @param[out] bad - RETURN: index of first bad attribute
*
* @return int
* @retval 0 : success
* @retval PBSE_PERM : client is not authorized to status the job
* @retval PBSE_SYSTEM : memory allocation error
* @retval PBSE_IVALREQ : something wrong with the flags
*/
int
status_subjob(job *pjob, struct batch_request *preq, svrattrl *pal, int subj, pbs_list_head *pstathd, int *bad)
{
int limit = (int)JOB_ATR_LAST;
struct brp_status *pstat;
job *psubjob; /* ptr to job to status */
char realstate;
int rc = 0;
int oldeligflags = 0;
int oldatypflags = 0;
int subjob_state = -1;
char *old_subjob_comment = NULL;
/* see if the client is authorized to status this job */
if (! server.sv_attr[(int)SRV_ATR_query_others].at_val.at_long)
if (svr_authorize_jobreq(preq, pjob))
return (PBSE_PERM);
if ((pjob->ji_qs.ji_svrflags & JOB_SVFLG_ArrayJob) == 0)
return PBSE_IVALREQ;
/* if subjob is running, use real job structure */
if (get_subjob_state(pjob, subj) == JOB_STATE_RUNNING) {
psubjob = find_job(mk_subjob_id(pjob, subj));
if (psubjob)
status_job(psubjob, preq, pal, pstathd, bad);
return 0;
}
/* otherwise we fake it with info from the parent */
/* allocate reply structure and fill in header portion */
/* for the general case, we don't want to include the parent's */
/* array related attrbutes as they belong only to the Array */
if (pal == NULL)
limit = JOB_ATR_array;
pstat = (struct brp_status *)malloc(sizeof(struct brp_status));
if (pstat == (struct brp_status *)0)
return (PBSE_SYSTEM);
CLEAR_LINK(pstat->brp_stlink);
pstat->brp_objtype = MGR_OBJ_JOB;
(void)strcpy(pstat->brp_objname, mk_subjob_id(pjob, subj));
CLEAR_HEAD(pstat->brp_attr);
append_link(pstathd, &pstat->brp_stlink, pstat);
/* add attributes to the status reply */
*bad = 0;
/*
* fake the job state and comment by setting the parent job's state
* and comment to that of the subjob
*/
subjob_state = get_subjob_state(pjob, subj);
realstate = pjob->ji_wattr[(int)JOB_ATR_state].at_val.at_char;
pjob->ji_wattr[(int)JOB_ATR_state].at_val.at_char = statechars[subjob_state];
pjob->ji_wattr[(int)JOB_ATR_state].at_flags |= ATR_VFLAG_MODCACHE;
if (subjob_state == JOB_STATE_EXPIRED || subjob_state == JOB_STATE_FINISHED) {
if (pjob->ji_ajtrk->tkm_tbl[subj].trk_substate == JOB_SUBSTATE_FINISHED) {
if (pjob->ji_wattr[(int)JOB_ATR_Comment].at_flags & ATR_VFLAG_SET) {
old_subjob_comment = strdup(pjob->ji_wattr[(int)JOB_ATR_Comment].at_val.at_str);
if (old_subjob_comment == (char *)0)
return (PBSE_SYSTEM);
}
if (job_attr_def[(int)JOB_ATR_Comment].at_decode(&pjob->ji_wattr[(int)JOB_ATR_Comment],
(char *)0, (char *)0, "Subjob finished") == PBSE_SYSTEM) {
free(old_subjob_comment);
return (PBSE_SYSTEM);
}
} else if (pjob->ji_ajtrk->tkm_tbl[subj].trk_substate == JOB_SUBSTATE_FAILED) {
if (pjob->ji_wattr[(int)JOB_ATR_Comment].at_flags & ATR_VFLAG_SET) {
old_subjob_comment = strdup(pjob->ji_wattr[(int)JOB_ATR_Comment].at_val.at_str);
if (old_subjob_comment == (char *)0)
return (PBSE_SYSTEM);
}
if (job_attr_def[(int)JOB_ATR_Comment].at_decode(&pjob->ji_wattr[(int)JOB_ATR_Comment],
(char *)0, (char *)0, "Subjob failed") == PBSE_SYSTEM) {
free(old_subjob_comment);
return (PBSE_SYSTEM);
}
} else if (pjob->ji_ajtrk->tkm_tbl[subj].trk_substate == JOB_SUBSTATE_TERMINATED) {
if (pjob->ji_wattr[(int)JOB_ATR_Comment].at_flags & ATR_VFLAG_SET) {
old_subjob_comment = strdup(pjob->ji_wattr[(int)JOB_ATR_Comment].at_val.at_str);
if (old_subjob_comment == (char *)0)
return (PBSE_SYSTEM);
}
if (job_attr_def[(int)JOB_ATR_Comment].at_decode(&pjob->ji_wattr[(int)JOB_ATR_Comment],
(char *)0, (char *)0, "Subjob terminated") == PBSE_SYSTEM) {
free(old_subjob_comment);
return (PBSE_SYSTEM);
}
}
}
/* when eligible_time_enable is off, */
/* clear the set flag so that eligible_time and accrue_type dont show */
if (server.sv_attr[(int)SRV_ATR_EligibleTimeEnable].at_val.at_long == 0) {
oldeligflags = pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_flags;
pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_flags &= ~ATR_VFLAG_SET;
pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_flags |= ATR_VFLAG_MODCACHE;
oldatypflags = pjob->ji_wattr[(int)JOB_ATR_accrue_type].at_flags;
pjob->ji_wattr[(int)JOB_ATR_accrue_type].at_flags &= ~ATR_VFLAG_SET;
pjob->ji_wattr[(int)JOB_ATR_accrue_type].at_flags |= ATR_VFLAG_MODCACHE;
/* Note: ATR_VFLAG_MODCACHE must be set because of svr_cached() does */
/* not correctly check ATR_VFLAG_SET */
}
if (status_attrib(pal, job_attr_def, pjob->ji_wattr, limit,
preq->rq_perm, &pstat->brp_attr, bad))
rc = PBSE_NOATTR;
/* Set the parent state back to what it really is */
pjob->ji_wattr[(int)JOB_ATR_state].at_val.at_char = realstate;
pjob->ji_wattr[(int)JOB_ATR_state].at_flags |= ATR_VFLAG_MODCACHE;
/* Set the parent comment back to what it really is */
if (old_subjob_comment != NULL) {
if (job_attr_def[(int)JOB_ATR_Comment].at_decode(&pjob->ji_wattr[(int)JOB_ATR_Comment],
(char *)0, (char *)0, old_subjob_comment) == PBSE_SYSTEM) {
free(old_subjob_comment);
return (PBSE_SYSTEM);
}
free(old_subjob_comment);
}
/* reset the flags */
if (server.sv_attr[(int)SRV_ATR_EligibleTimeEnable].at_val.at_long == 0) {
pjob->ji_wattr[(int)JOB_ATR_eligible_time].at_flags = oldeligflags;
pjob->ji_wattr[(int)JOB_ATR_accrue_type].at_flags = oldatypflags;
}
return (rc);
}
job *job_recov(
char *filename) /* I */ /* pathname to job save file */
{
int fds;
job *pj;
char *pn;
char namebuf[MAXPATHLEN];
int qs_upgrade;
#ifndef PBS_MOM
char parent_id[PBS_MAXSVRJOBID + 1];
job_array *pa;
#endif
qs_upgrade = FALSE;
pj = job_alloc(); /* allocate & initialize job structure space */
if (pj == NULL)
{
/* FAILURE - cannot alloc memory */
return(NULL);
}
strcpy(namebuf, path_jobs); /* job directory path */
strcat(namebuf, filename);
fds = open(namebuf, O_RDONLY, 0);
if (fds < 0)
{
sprintf(log_buffer, "unable to open %s",
namebuf);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
/* FAILURE - cannot open job file */
return(NULL);
}
/* read in job quick save sub-structure */
if (read(fds, (char *)&pj->ji_qs, quicksize) != (ssize_t)quicksize &&
pj->ji_qs.qs_version == PBS_QS_VERSION)
{
sprintf(log_buffer, "Unable to read %s",
namebuf);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
close(fds);
return(NULL);
}
/* is ji_qs the version we expect? */
if (pj->ji_qs.qs_version != PBS_QS_VERSION)
{
/* ji_qs is older version */
sprintf(log_buffer,
"%s appears to be from an old version. Attempting to convert.\n",
namebuf);
log_err(-1, "job_recov", log_buffer);
if (job_qs_upgrade(pj, fds, namebuf, pj->ji_qs.qs_version) != 0)
{
sprintf(log_buffer, "unable to upgrade %s\n", namebuf);
log_err(-1, "job_recov", log_buffer);
free((char *)pj);
close(fds);
return(NULL);
}
qs_upgrade = TRUE;
} /* END if (pj->ji_qs.qs_version != PBS_QS_VERSION) */
/* Does file name match the internal name? */
/* This detects ghost files */
pn = strrchr(namebuf, (int)'/') + 1;
if (strncmp(pn, pj->ji_qs.ji_fileprefix, strlen(pj->ji_qs.ji_fileprefix)) != 0)
{
/* mismatch, discard job */
sprintf(log_buffer, "Job Id %s does not match file name for %s",
pj->ji_qs.ji_jobid,
namebuf);
log_err(-1, "job_recov", log_buffer);
free((char *)pj);
close(fds);
return(NULL);
}
/* read in working attributes */
if (recov_attr(
fds,
pj,
job_attr_def,
pj->ji_wattr,
(int)JOB_ATR_LAST,
(int)JOB_ATR_UNKN,
TRUE) != 0)
{
sprintf(log_buffer, "unable to recover %s (file is likely corrupted)",
namebuf);
log_err(-1, "job_recov", log_buffer);
job_free(pj);
close(fds);
return(NULL);
}
#ifdef PBS_MOM
/* read in tm sockets and ips */
if (recov_tmsock(fds, pj) != 0)
{
sprintf(log_buffer, "warning: tmsockets not recovered from %s (written by an older pbs_mom?)",
namebuf);
log_err(-1, "job_recov", log_buffer);
}
if (recov_roottask(fds, pj) != 0)
{
sprintf(log_buffer, "warning: root task not recovered from %s (written by an older pbs_mom?)",
namebuf);
log_err(-1, "job_recov", log_buffer);
}
if (recov_jobflags(fds, pj) != 0)
{
sprintf(log_buffer, "warning: job flags not recovered from %s (written by an older pbs_mom?)", namebuf);
log_err(-1, "job_recov", log_buffer);
}
#else /* PBS_MOM */
if (pj->ji_wattr[(int)JOB_ATR_job_array_request].at_flags & ATR_VFLAG_SET)
{
/* job is part of an array. We need to put a link back to the server
job array struct for this array. We also have to link this job into
the linked list of jobs belonging to the array. */
array_get_parent_id(pj->ji_qs.ji_jobid, parent_id);
pa = get_array(parent_id);
if (strcmp(parent_id, pj->ji_qs.ji_jobid) == 0)
{
pj->ji_isparent = TRUE;
}
else
{
if (pa == NULL)
{
/* couldn't find array struct, it must not have been recovered,
treat job as indepentent job? perhaps we should delete the job
XXX_JOB_ARRAY: should I unset this?*/
pj->ji_wattr[(int)JOB_ATR_job_array_request].at_flags &= ~ATR_VFLAG_SET;
}
else
{
CLEAR_LINK(pj->ji_arrayjobs);
append_link(&pa->array_alljobs, &pj->ji_arrayjobs, (void*)pj);
pj->ji_arraystruct = pa;
pa->jobs_recovered++;
}
}
}
#endif
close(fds);
/* all done recovering the job */
if (qs_upgrade == TRUE)
{
job_save(pj, SAVEJOB_FULL);
}
return(pj);
} /* END job_recov() */
int encode_time(
attribute *attr, /* ptr to attribute (value in attr->at_val.at_long) */
tlist_head *phead, /* head of attrlist list (optional) */
char *atname, /* attribute name */
char *rsname, /* resource name (optional) */
int mode) /* encode mode (not used) */
{
size_t ct;
char cvnbuf[CVNBUFSZ];
int hr;
int min;
long n;
svrattrl *pal;
int sec;
char *pv;
if (attr == NULL)
{
/* FAILURE */
return(-1);
}
if (!(attr->at_flags & ATR_VFLAG_SET))
{
return(0);
}
n = attr->at_val.at_long;
hr = n / 3600;
n = n % 3600;
min = n / 60;
n = n % 60;
sec = n;
pv = cvnbuf;
sprintf(pv, "%02d:%02d:%02d",
hr, min, sec);
pv += strlen(pv);
ct = strlen(cvnbuf) + 1;
if (phead != NULL)
{
pal = attrlist_create(atname, rsname, ct);
if (pal == NULL)
{
/* FAILURE */
return(-1);
}
memcpy(pal->al_value, cvnbuf, ct);
pal->al_flags = attr->at_flags;
append_link(phead, &pal->al_link, pal);
}
else
{
strcpy(atname, cvnbuf);
}
/* SUCCESS */
return(1);
}
/**
* attempt_delete()
* deletes a job differently depending on the job's state
*
* @return TRUE if the job was deleted, FALSE if skipped
* @param pjob - a pointer to the job being handled
*/
int attempt_delete(
void *j) /* I */
{
int skipped = FALSE;
struct work_task *pwtold;
struct work_task *pwtnew;
job *pjob;
/* job considered deleted if null */
if (j == NULL)
return(TRUE);
pjob = (job *)j;
if (pjob->ji_qs.ji_state == JOB_STATE_TRANSIT)
{
/*
* Find pid of router from existing work task entry,
* then establish another work task on same child.
* Next, signal the router and wait for its completion;
*/
pwtold = (struct work_task *)GET_NEXT(pjob->ji_svrtask);
while (pwtold != NULL)
{
if ((pwtold->wt_type == WORK_Deferred_Child) ||
(pwtold->wt_type == WORK_Deferred_Cmp))
{
kill((pid_t)pwtold->wt_event, SIGTERM);
pjob->ji_qs.ji_substate = JOB_SUBSTATE_ABORT;
}
pwtold = (struct work_task *)GET_NEXT(pwtold->wt_linkobj);
}
skipped = TRUE;
return(!skipped);
} /* END if (pjob->ji_qs.ji_state == JOB_SUBSTATE_TRANSIT) */
else if (pjob->ji_qs.ji_substate == JOB_SUBSTATE_PRERUN)
{
/* we'll wait for the mom to get this job, then delete it */
skipped = TRUE;
} /* END if (pjob->ji_qs.ji_substate == JOB_SUBSTATE_PRERUN) */
else if (pjob->ji_qs.ji_state == JOB_STATE_RUNNING)
{
/* set up nanny */
if (!has_job_delete_nanny(pjob))
{
apply_job_delete_nanny(pjob, time_now + 60);
/* need to issue a signal to the mom, but we don't want to sent an ack to the
* client when the mom replies */
issue_signal(pjob, "SIGTERM", post_delete, NULL);
}
if ((pjob->ji_qs.ji_svrflags & JOB_SVFLG_CHECKPOINT_FILE) != 0)
{
/* job has restart file at mom, change restart comment if failed */
change_restart_comment_if_needed(pjob);
}
return(!skipped);
} /* END if (pjob->ji_qs.ji_state == JOB_STATE_RUNNING) */
if ((pjob->ji_qs.ji_svrflags & JOB_SVFLG_CHECKPOINT_FILE) != 0)
{
/* job has restart file at mom, change restart comment if failed */
change_restart_comment_if_needed(pjob);
/* job has restart file at mom, do end job processing */
svr_setjobstate(pjob, JOB_STATE_EXITING, JOB_SUBSTATE_EXITING);
pjob->ji_momhandle = -1;
/* force new connection */
pwtnew = set_task(WORK_Immed, 0, on_job_exit, (void *)pjob);
if (pwtnew)
{
append_link(&pjob->ji_svrtask, &pwtnew->wt_linkobj, pwtnew);
}
}
else if ((pjob->ji_qs.ji_svrflags & JOB_SVFLG_StagedIn) != 0)
{
/* job has staged-in file, should remove them */
remove_stagein(pjob);
job_abt(&pjob, NULL);
}
else
{
/*
* the job is not transitting (though it may have been) and
* is not running, so put in into a complete state.
*/
struct work_task *ptask;
struct pbs_queue *pque;
int KeepSeconds = 0;
svr_setjobstate(pjob, JOB_STATE_COMPLETE, JOB_SUBSTATE_COMPLETE);
if ((pque = pjob->ji_qhdr) && (pque != NULL))
{
pque->qu_numcompleted++;
}
KeepSeconds = attr_ifelse_long(
&pque->qu_attr[QE_ATR_KeepCompleted],
&server.sv_attr[SRV_ATR_KeepCompleted],
0);
ptask = set_task(WORK_Timed, time_now + KeepSeconds, on_job_exit, pjob);
if (ptask != NULL)
{
append_link(&pjob->ji_svrtask, &ptask->wt_linkobj, ptask);
}
}
return(!skipped);
} /* END attempt_delete() */
void array_delete_wt(struct work_task *ptask)
{
struct batch_request *preq;
job_array *pa;
/*struct work_task *pnew_task;*/
struct work_task *pwtnew;
int i;
static int last_check = 0;
static char *last_id = NULL;
preq = ptask->wt_parm1;
pa = get_array(preq->rq_ind.rq_delete.rq_objname);
if (pa == NULL)
{
/* jobs must have exited already */
reply_ack(preq);
last_check = 0;
free(last_id);
last_id = NULL;
return;
}
if (last_id == NULL)
{
last_id = strdup(preq->rq_ind.rq_delete.rq_objname);
last_check = time_now;
}
else if (strcmp(last_id, preq->rq_ind.rq_delete.rq_objname) != 0)
{
last_check = time_now;
free(last_id);
last_id = strdup(preq->rq_ind.rq_delete.rq_objname);
}
else if (time_now - last_check > 10)
{
int num_jobs;
int num_prerun;
job *pjob;
num_jobs = 0;
num_prerun = 0;
for (i = 0; i < pa->ai_qs.array_size; i++)
{
if (pa->jobs[i] == NULL)
continue;
pjob = (job *)pa->jobs[i];
num_jobs++;
if (pjob->ji_qs.ji_substate == JOB_SUBSTATE_PRERUN)
{
num_prerun++;
/* mom still hasn't gotten job?? delete anyway */
if ((pjob->ji_qs.ji_svrflags & JOB_SVFLG_CHECKPOINT_FILE) != 0)
{
/* job has restart file at mom, do end job processing */
change_restart_comment_if_needed(pjob);
svr_setjobstate(pjob, JOB_STATE_EXITING, JOB_SUBSTATE_EXITING);
pjob->ji_momhandle = -1;
/* force new connection */
pwtnew = set_task(WORK_Immed, 0, on_job_exit, (void *)pjob);
if (pwtnew)
{
append_link(&pjob->ji_svrtask, &pwtnew->wt_linkobj, pwtnew);
}
}
else if ((pjob->ji_qs.ji_svrflags & JOB_SVFLG_StagedIn) != 0)
{
/* job has staged-in file, should remove them */
remove_stagein(pjob);
job_abt(&pjob, NULL);
}
else
{
job_abt(&pjob, NULL);
}
}
}
if (num_jobs == num_prerun)
{
reply_ack(preq);
free(last_id);
last_id = NULL;
return;
}
}
req_deletearray(preq);
}
int encode_arst(
pbs_attribute *attr, /* I ptr to pbs_attribute to encode */
tlist_head *phead, /* O ptr to head of attrlist list */
char *atname, /* I pbs_attribute name */
char *rsname, /* I resource name or NULL (optional) */
int mode, /* I encode mode */
int perm) /* only used for resources */
{
char *end;
int i;
int j;
svrattrl *pal;
char *pc;
char *pfrom;
char separator;
if (attr == NULL)
{
/* FAILURE - invalid parameters */
return(-2);
}
if (((attr->at_flags & ATR_VFLAG_SET) == 0) ||
!attr->at_val.at_arst ||
!attr->at_val.at_arst->as_usedptr)
{
/* SUCCESS - empty list */
return(0); /* no values */
}
i = (int)(attr->at_val.at_arst->as_next - attr->at_val.at_arst->as_buf);
if (mode == ATR_ENCODE_SAVE)
{
separator = '\n'; /* new-line for encode_acl */
/* allow one extra byte for final new-line */
j = i + 1;
}
else
{
separator = ','; /* normally a comma is the separator */
j = i;
}
/* how many back-slashes are required */
for (pc = attr->at_val.at_arst->as_buf;pc < attr->at_val.at_arst->as_next;++pc)
{
if ((*pc == '"') || (*pc == '\'') || (*pc == ',') || (*pc == '\\') || (*pc == '\n'))
++j;
}
pal = attrlist_create(atname, rsname, j + 1);
if (pal == NULL)
{
return(-1);
}
pal->al_flags = attr->at_flags;
pc = pal->al_value;
pfrom = attr->at_val.at_arst->as_buf;
/* replace nulls between sub-strings with separater characters */
/* escape any embedded special character */
end = attr->at_val.at_arst->as_next;
while (pfrom < end)
{
switch (*pfrom)
{
case '\0':
*pc = separator;
break;
case '"':
case '\'':
case ',':
case '\\':
case '\n':
*pc++ = '\\';
default:
*pc = *pfrom;
break;
}
pc++;
pfrom++;
}
/* convert the last null to separator only if going to new-lines */
if (mode == ATR_ENCODE_SAVE)
*pc = '\0'; /* insure string terminator */
else
*(pc - 1) = '\0';
append_link(phead, &pal->al_link, pal);
return(1);
} /* END encode_arst() */
/**
* @brief
* main - the initialization and main loop of pbs_daemon
*/
int
main(int argc, char *argv[])
{
char jobfile[MAXPATHLEN+1];
char jobfile_full[MAXPATHLEN+1];
pbs_net_t hostaddr = 0;
int port = -1;
int move_type = -1;
pbs_list_head attrl;
enum conn_type cntype = ToServerDIS;
int con = -1;
char *destin;
int encode_type;
int i;
job *jobp;
char job_id[PBS_MAXSVRJOBID+1];
attribute *pattr;
struct attropl *pqjatr; /* list (single) of attropl for quejob */
char script_name[MAXPATHLEN+1];
int in_server = -1;
char *param_name, *param_val;
char buf[4096];
struct hostent *hp;
struct in_addr addr;
char *credbuf = NULL;
size_t credlen = 0;
int prot = PROT_TCP;
/*the real deal or output version and exit?*/
execution_mode(argc, argv);
/* If we are not run with real and effective uid of 0, forget it */
pbs_loadconf(0);
if (!isAdminPrivilege(getlogin())) {
fprintf(stderr, "%s: Must be run by root\n", argv[0]);
exit(SEND_JOB_FATAL);
}
/* initialize the pointers in the resource_def array */
for (i = 0; i < (svr_resc_size - 1); ++i)
svr_resc_def[i].rs_next = &svr_resc_def[i+1];
/* last entry is left with null pointer */
/* set single threaded mode */
pbs_client_thread_set_single_threaded_mode();
/* disable attribute verification */
set_no_attribute_verification();
/* initialize the thread context */
if (pbs_client_thread_init_thread_context() != 0) {
fprintf(stderr, "%s: Unable to initialize thread context\n",
argv[0]);
exit(SEND_JOB_FATAL);
}
if(set_msgdaemonname("PBS_send_job")) {
fprintf(stderr, "Out of memory\n");
return 1;
}
winsock_init();
connection_init();
while (fgets(buf, sizeof(buf), stdin) != NULL) {
buf[strlen(buf)-1] = '\0'; /* gets rid of newline */
param_name = buf;
param_val = strchr(buf, '=');
if (param_val) {
*param_val = '\0';
param_val++;
} else { /* bad param_val -- skipping */
break;
}
if (strcmp(param_name, "jobfile") == 0) {
jobfile[0] = '\0';
strncpy(jobfile, param_val, MAXPATHLEN);
} else if (strcmp(param_name, "destaddr") == 0) {
hostaddr = atol(param_val);
} else if (strcmp(param_name, "destport") == 0) {
port = atoi(param_val);
} else if (strcmp(param_name, "move_type") == 0) {
move_type = atoi(param_val);
} else if (strcmp(param_name, "in_server") == 0) {
in_server = atoi(param_val);
} else if (strcmp(param_name, "server_name") == 0) {
server_name[0] = '\0';
strncpy(server_name, param_val, PBS_MAXSERVERNAME);
} else if (strcmp(param_name, "server_host") == 0) {
server_host[0] = '\0';
strncpy(server_host, param_val, (sizeof(server_host) - 1));
} else if (strcmp(param_name, "server_addr") == 0) {
pbs_server_addr = atol(param_val);
} else if (strcmp(param_name, "server_port") == 0) {
pbs_server_port_dis = atoi(param_val);
} else if (strcmp(param_name, "log_file") == 0) {
log_file = strdup(param_val);
} else if (strcmp(param_name, "path_log") == 0) {
path_log[0] = '\0';
strncpy(path_log, param_val, MAXPATHLEN);
} else if (strcmp(param_name, "path_jobs") == 0) {
path_jobs = strdup(param_val);
} else if (strcmp(param_name, "path_spool") == 0) {
path_spool = strdup(param_val);
} else if (strcmp(param_name, "path_rescdef") == 0) {
path_rescdef = strdup(param_val);
} else if (strcmp(param_name, "path_users") == 0) {
path_users = strdup(param_val);
} else if (strcmp(param_name, "path_hooks_workdir") == 0) {
path_hooks_workdir = strdup(param_val);
if (path_hooks_workdir == NULL)
exit(SEND_JOB_FATAL);
} else if (strcmp(param_name, "svr_history_enable") == 0) {
svr_history_enable = atol(param_val);
} else if (strcmp(param_name, "svr_history_duration") == 0) {
svr_history_duration = atol(param_val);
} else if (strcmp(param_name, "single_signon_password_enable") == 0) {
if (decode_b(&server.sv_attr[(int)SRV_ATR_ssignon_enable],
NULL, NULL, param_val) != 0) {
fprintf(stderr, "%s: failed to set ssignon_password_enable\n", argv[0]);
exit(SEND_JOB_FATAL);
}
} else if (strcmp(param_name, "script_name") == 0) {
strncpy(script_name, param_val, MAXPATHLEN + 1);
} else
break;
}
time(&time_now);
(void)log_open_main(log_file, path_log, 1); /* silent open */
if (setup_resc(1) == -1) {
/* log_buffer set in setup_resc */
log_err(-1, "pbsd_send_job(setup_resc)", log_buffer);
return (-1);
}
if( strlen(jobfile) == 0 || hostaddr == 0 || port == 0 || move_type == -1 || \
in_server == -1 || strlen(server_name) == 0 || strlen(server_host) == 0 || \
pbs_server_addr == 0 || pbs_server_port_dis == 0 || \
strlen(path_log) == 0 || path_jobs == NULL || \
path_spool == NULL || path_users == NULL ) {
log_err(-1, "pbs_send_job", "error on one of the parameters");
log_close(0); /* silent close */
exit(SEND_JOB_FATAL);
}
CLEAR_HEAD(task_list_immed);
CLEAR_HEAD(task_list_timed);
CLEAR_HEAD(task_list_event);
CLEAR_HEAD(svr_queues);
CLEAR_HEAD(svr_alljobs);
CLEAR_HEAD(svr_newjobs);
CLEAR_HEAD(svr_allresvs);
CLEAR_HEAD(svr_newresvs);
CLEAR_HEAD(svr_deferred_req);
CLEAR_HEAD(svr_unlicensedjobs);
strcpy(jobfile_full, path_jobs);
strcat(jobfile_full, jobfile);
if (chk_save_file(jobfile_full) != 0) {
sprintf(log_buffer, "Error opening jobfile=%s", jobfile);
log_err(-1, __func__, log_buffer);
goto fatal_exit;
}
if ((jobp=job_recov_fs(jobfile, RECOV_SUBJOB)) == NULL) {
sprintf(log_buffer, "Failed to recreate job in jobfile=%s", jobfile);
log_err(-1, __func__, log_buffer);
goto fatal_exit;
}
/* now delete the temp job file that was created by job_save_fs in server code
* jobs are in database now, no need to keep in filesystem
*/
unlink(jobfile_full);
if (in_server)
append_link(&svr_alljobs, &jobp->ji_alljobs, jobp);
/* select attributes/resources to send based on move type */
if (move_type == MOVE_TYPE_Exec) {
resc_access_perm = ATR_DFLAG_MOM;
encode_type = ATR_ENCODE_MOM;
cntype = ToServerDIS;
} else {
resc_access_perm = ATR_DFLAG_USWR | ATR_DFLAG_OPWR |
ATR_DFLAG_MGWR | ATR_DFLAG_SvRD;
encode_type = ATR_ENCODE_SVR;
svr_dequejob(jobp);
}
CLEAR_HEAD(attrl);
pattr = jobp->ji_wattr;
for (i=0; i < (int)JOB_ATR_LAST; i++) {
if ((job_attr_def+i)->at_flags & resc_access_perm) {
(void)(job_attr_def+i)->at_encode(pattr+i, &attrl,
(job_attr_def+i)->at_name, NULL,
encode_type, NULL);
}
}
attrl_fixlink(&attrl);
/* script name is passed from parent */
/* get host name */
pbs_loadconf(0);
addr.s_addr = htonl(hostaddr);
hp = gethostbyaddr((void *)&addr, sizeof(struct in_addr), AF_INET);
if (hp == NULL) {
sprintf(log_buffer, "%s: h_errno=%d",
inet_ntoa(addr), h_errno);
log_err(-1, __func__, log_buffer);
}
else {
/* read any credential file */
(void)get_credential(hp->h_name, jobp, PBS_GC_BATREQ,
&credbuf, &credlen);
}
/* save the job id for when after we purge the job */
(void)strcpy(job_id, jobp->ji_qs.ji_jobid);
con = -1;
DIS_tcparray_init();
for (i=0; i<RETRY; i++) {
pbs_errno = 0;
/* connect to receiving server with retries */
if (i > 0) { /* recycle after an error */
if (con >= 0)
svr_disconnect(con);
if (should_retry_route(pbs_errno) == -1) {
goto fatal_exit; /* fatal error, don't retry */
}
sleep(1<<i);
}
if ((con = svr_connect(hostaddr, port, 0, cntype, prot)) ==
PBS_NET_RC_FATAL) {
(void)sprintf(log_buffer, "send_job failed to %lx port %d",
hostaddr, port);
log_err(pbs_errno, __func__, log_buffer);
goto fatal_exit;
} else if (con == PBS_NET_RC_RETRY) {
pbs_errno = WSAECONNREFUSED; /* should retry */
continue;
}
/*
* if the job is substate JOB_SUBSTATE_TRNOUTCM which means
* we are recovering after being down or a late failure, we
* just want to send the "read-to-commit/commit"
*/
if (jobp->ji_qs.ji_substate != JOB_SUBSTATE_TRNOUTCM) {
if (jobp->ji_qs.ji_substate != JOB_SUBSTATE_TRNOUT) {
jobp->ji_qs.ji_substate = JOB_SUBSTATE_TRNOUT;
}
pqjatr = &((svrattrl *)GET_NEXT(attrl))->al_atopl;
destin = jobp->ji_qs.ji_destin;
if (PBSD_queuejob(con, jobp->ji_qs.ji_jobid, destin, pqjatr, NULL, prot, NULL)== 0) {
if (pbs_errno == PBSE_JOBEXIST && move_type == MOVE_TYPE_Exec) {
/* already running, mark it so */
log_event(PBSEVENT_ERROR,
PBS_EVENTCLASS_JOB, LOG_INFO,
jobp->ji_qs.ji_jobid, "Mom reports job already running");
goto ok_exit;
} else if ((pbs_errno == PBSE_HOOKERROR) ||
(pbs_errno == PBSE_HOOK_REJECT) ||
(pbs_errno == PBSE_HOOK_REJECT_RERUNJOB) ||
(pbs_errno == PBSE_HOOK_REJECT_DELETEJOB)) {
char name_buf[MAXPATHLEN+1];
int rfd;
int len;
char *reject_msg;
int err;
err = pbs_errno;
reject_msg = pbs_geterrmsg(con);
(void)snprintf(log_buffer, sizeof(log_buffer),
"send of job to %s failed error = %d reject_msg=%s",
destin, err,
reject_msg?reject_msg:"");
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB,
LOG_INFO, jobp->ji_qs.ji_jobid,
log_buffer);
(void)strcpy(name_buf, path_hooks_workdir);
(void)strcat(name_buf, jobp->ji_qs.ji_jobid);
(void)strcat(name_buf, HOOK_REJECT_SUFFIX);
if ((reject_msg != NULL) &&
(reject_msg[0] != '\0')) {
if ((rfd = open(name_buf,
O_RDWR|O_CREAT|O_TRUNC, 0600)) == -1) {
snprintf(log_buffer,
sizeof(log_buffer),
"open of reject file %s failed: errno %d",
name_buf, errno);
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB, LOG_INFO, jobp->ji_qs.ji_jobid, log_buffer);
} else {
secure_file(name_buf, "Administrators",
READS_MASK|WRITES_MASK|STANDARD_RIGHTS_REQUIRED);
setmode(rfd, O_BINARY);
len = strlen(reject_msg)+1;
/* write also trailing null char */
if (write(rfd, reject_msg, len) != len) {
snprintf(log_buffer,
sizeof(log_buffer),
"write to file %s incomplete: errno %d", name_buf, errno);
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB, LOG_INFO, jobp->ji_qs.ji_jobid, log_buffer);
}
close(rfd);
}
}
if (err == PBSE_HOOKERROR)
exit(SEND_JOB_HOOKERR);
if (err == PBSE_HOOK_REJECT)
exit(SEND_JOB_HOOK_REJECT);
if (err == PBSE_HOOK_REJECT_RERUNJOB)
exit(SEND_JOB_HOOK_REJECT_RERUNJOB);
if (err == PBSE_HOOK_REJECT_DELETEJOB)
exit(SEND_JOB_HOOK_REJECT_DELETEJOB);
} else {
(void)sprintf(log_buffer, "send of job to %s failed error = %d", destin, pbs_errno);
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB, LOG_INFO, jobp->ji_qs.ji_jobid, log_buffer);
continue;
}
}
if (jobp->ji_qs.ji_svrflags & JOB_SVFLG_SCRIPT) {
if (PBSD_jscript(con, script_name, prot, NULL) != 0)
continue;
}
if (credlen > 0) {
int ret;
ret = PBSD_jcred(con,
jobp->ji_extended.ji_ext.ji_credtype,
credbuf, credlen, prot, NULL);
if ((ret == 0) || (i == (RETRY - 1)))
free(credbuf); /* free credbuf if credbuf is sent successfully OR */
/* at the end of all retry attempts */
if (ret != 0)
continue;
}
if ((move_type == MOVE_TYPE_Exec) &&
(jobp->ji_qs.ji_svrflags & JOB_SVFLG_HASRUN) &&
(hostaddr != pbs_server_addr)) {
/* send files created on prior run */
if ((move_job_file(con, jobp, StdOut, prot) != 0) ||
(move_job_file(con, jobp, StdErr, prot) != 0) ||
(move_job_file(con, jobp, Chkpt, prot) != 0))
continue;
}
jobp->ji_qs.ji_substate = JOB_SUBSTATE_TRNOUTCM;
}
if (PBSD_rdytocmt(con, job_id, prot, NULL) != 0)
continue;
if (PBSD_commit(con, job_id, prot, NULL) != 0)
goto fatal_exit;
goto ok_exit; /* This child process is all done */
}
if (con >= 0)
svr_disconnect(con);
/*
* If connection is actively refused by the execution node(or mother superior) OR
* the execution node(or mother superior) is rejecting request with error
* PBSE_BADHOST(failing to authorize server host), the node should be marked down.
*/
if ((move_type == MOVE_TYPE_Exec) && (pbs_errno == WSAECONNREFUSED || pbs_errno == PBSE_BADHOST)) {
i = SEND_JOB_NODEDW;
} else if (should_retry_route(pbs_errno) == -1) {
i = SEND_JOB_FATAL;
} else {
i = SEND_JOB_RETRY;
}
(void)sprintf(log_buffer, "send_job failed with error %d", pbs_errno);
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_JOB, LOG_NOTICE,
jobp->ji_qs.ji_jobid, log_buffer);
log_close(0);
net_close(-1);
unlink(script_name);
exit(i);
fatal_exit:
if (con >= 0)
svr_disconnect(con);
log_close(0);
net_close(-1);
unlink(script_name);
exit(SEND_JOB_FATAL);
ok_exit:
if (con >= 0)
svr_disconnect(con);
log_close(0);
net_close(-1);
unlink(script_name);
exit(SEND_JOB_OK);
}
static void
svrcached(attribute *pat, pbs_list_head *phead, attribute_def *pdef)
{
svrattrl *working = NULL;
svrattrl *wcopy;
svrattrl *encoded;
if (resc_access_perm & PRIV_READ)
encoded = pat->at_priv_encoded;
else
encoded = pat->at_user_encoded;
if (pat->at_flags & ATR_VFLAG_MODCACHE)
/* free old cache value if the value has changed */
free_svrcache(pat);
if ((encoded == NULL) || (pat->at_flags & ATR_VFLAG_MODCACHE)) {
if (pat->at_flags & ATR_VFLAG_SET) {
/* encode and cache new svrattrl structure */
(void)pdef->at_encode(pat, phead, pdef->at_name,
(char *)0, ATR_ENCODE_CLIENT, &working);
if (resc_access_perm & PRIV_READ)
pat->at_priv_encoded = working;
else
pat->at_user_encoded = working;
pat->at_flags &= ~ATR_VFLAG_MODCACHE;
while (working) {
working->al_refct++; /* incr ref count */
working = working->al_sister;
}
}
} else {
/* can use the existing cached svrattrl struture */
working = encoded;
if (working->al_refct < 2) {
while (working) {
CLEAR_LINK(working->al_link);
append_link(phead, &working->al_link, working);
working->al_refct++; /* incr ref count */
working = working->al_sister;
}
} else {
/*
* already linked in, must make a copy to link
* NOTE: the copy points to the original's data
* so it should be freed by itself, hence the
* ref count is set to 1 and the sisters are not
* linked in
*/
while (working) {
wcopy = malloc(sizeof(struct svrattrl));
if (wcopy) {
*wcopy = *working;
working = working->al_sister;
CLEAR_LINK(wcopy->al_link);
append_link(phead, &wcopy->al_link, wcopy);
wcopy->al_refct = 1;
wcopy->al_sister = NULL;
}
}
}
}
}
void req_quejob(
struct batch_request *preq) /* ptr to the decoded request */
{
char *id = "req_quejob";
char basename[PBS_JOBBASE + 1];
int created_here = 0;
int index;
char *jid;
attribute_def *pdef;
job *pj;
svrattrl *psatl;
int rc;
int sock = preq->rq_conn;
int IsCheckpoint = 0;
/* set basic (user) level access permission */
resc_access_perm = ATR_DFLAG_USWR | ATR_DFLAG_Creat;
if (PBSNodeCheckProlog)
{
check_state(1);
mom_server_all_update_stat();
if (internal_state & INUSE_DOWN)
{
req_reject(PBSE_MOMREJECT,0,preq,NULL,NULL);
return;
}
}
if (preq->rq_fromsvr)
{
/* from another server - accept the extra attributes */
resc_access_perm |= ATR_DFLAG_MGWR | ATR_DFLAG_SvWR | ATR_DFLAG_MOM;
jid = preq->rq_ind.rq_queuejob.rq_jid;
}
else
{
/* request must be from server */
log_err(errno, id, "request not from server");
req_reject(PBSE_IVALREQ, 0, preq, NULL, "request not received from server");
return;
}
/* does job already exist, check both old and new jobs */
if ((pj = find_job(jid)) == NULL)
{
pj = (job *)GET_NEXT(svr_newjobs);
while (pj != NULL)
{
if (!strcmp(pj->ji_qs.ji_jobid, jid))
break;
pj = (job *)GET_NEXT(pj->ji_alljobs);
}
}
/*
* New job ...
*
* for MOM - rather than make up a hashname, we use the name sent
* to us by the server as an attribute.
*/
psatl = (svrattrl *)GET_NEXT(preq->rq_ind.rq_queuejob.rq_attr);
while (psatl != NULL)
{
if (!strcmp(psatl->al_name,ATTR_hashname))
{
strcpy(basename,psatl->al_value);
break;
}
psatl = (svrattrl *)GET_NEXT(psatl->al_link);
}
if (pj != NULL)
{
/* newly queued job already exists */
if (pj->ji_qs.ji_substate == JOB_SUBSTATE_RUNNING)
{
/* FAILURE - job exists and is running */
log_err(errno,id,"cannot queue new job, job exists and is running");
req_reject(PBSE_JOBEXIST,0,preq,NULL,"job is running");
return;
}
/* if checkpointed, then keep old and skip rest of process */
if (pj->ji_qs.ji_svrflags & JOB_SVFLG_CHECKPOINT_FILE)
{
IsCheckpoint = 1;
} /* END if (pj->ji_qs.ji_svrflags & JOB_SVFLG_CHECKPOINT_FILE) */
else
{
/* unlink job from svr_alljobs since it will be placed on newjobs */
delete_link(&pj->ji_alljobs);
}
} /* END if (pj != NULL) */
else
{
/* if not already here, allocate job struct */
if ((pj = job_alloc()) == NULL)
{
/* FAILURE */
req_reject(PBSE_SYSTEM, 0, preq, NULL, "cannot allocate new job structure");
return;
}
} /* END else (pj != NULL) */
if (IsCheckpoint == 0)
{
strcpy(pj->ji_qs.ji_jobid,jid);
strcpy(pj->ji_qs.ji_fileprefix,basename);
pj->ji_modified = 1;
pj->ji_qs.ji_svrflags = created_here;
pj->ji_qs.ji_un_type = JOB_UNION_TYPE_NEW;
}
/* decode attributes from request into job structure */
psatl = (svrattrl *)GET_NEXT(preq->rq_ind.rq_queuejob.rq_attr);
while (psatl != NULL)
{
if (IsCheckpoint == 1)
{
if (strcmp(psatl->al_name,ATTR_checkpoint_name) &&
strcmp(psatl->al_name,ATTR_v))
{
psatl = (svrattrl *)GET_NEXT(psatl->al_link);
continue;
}
}
/* identify the attribute by name */
index = find_attr(job_attr_def,psatl->al_name,JOB_ATR_LAST);
if (index < 0)
{
/* FAILURE */
/* didn`t recognize the name */
job_purge(pj); /* CRI - 12/20/2004 */
reply_badattr(PBSE_NOATTR,1,psatl,preq);
return;
}
pdef = &job_attr_def[index];
/* Is attribute not writeable by manager or by a server? */
if ((pdef->at_flags & resc_access_perm) == 0)
{
/* FAILURE */
job_purge(pj);
reply_badattr(PBSE_ATTRRO,1,psatl,preq);
return;
}
/* decode attribute */
if (!strcmp(psatl->al_name,ATTR_v))
{
rc = decode_arst_merge(
&pj->ji_wattr[index],
psatl->al_name,
psatl->al_resc,
psatl->al_value);
}
else
{
rc = pdef->at_decode(
&pj->ji_wattr[index],
psatl->al_name,
psatl->al_resc,
psatl->al_value);
}
if (rc != 0)
{
/* FAILURE */
/* all errors are fatal for MOM */
job_purge(pj);
reply_badattr(rc,1,psatl,preq);
return;
}
if (psatl->al_op == DFLT)
{
if (psatl->al_resc)
{
resource *presc;
resource_def *prdef;
prdef = find_resc_def(svr_resc_def,psatl->al_resc,svr_resc_size);
if (prdef == NULL)
{
job_purge(pj);
reply_badattr(rc,1,psatl, preq);
return;
}
presc = find_resc_entry(&pj->ji_wattr[index],prdef);
if (presc != NULL)
presc->rs_value.at_flags |= ATR_VFLAG_DEFLT;
}
else
{
pj->ji_wattr[index].at_flags |= ATR_VFLAG_DEFLT;
}
} /* END if (psatl->al_op == DFLT) */
psatl = (svrattrl *)GET_NEXT(psatl->al_link);
} /* END while (psatl != NULL) */
if (IsCheckpoint == 1)
{
pj->ji_qs.ji_substate = JOB_SUBSTATE_TRANSIN;
if (reply_jobid(preq,pj->ji_qs.ji_jobid,BATCH_REPLY_CHOICE_Queue) == 0)
{
delete_link(&pj->ji_alljobs);
append_link(&svr_newjobs,&pj->ji_alljobs,pj);
pj->ji_qs.ji_un_type = JOB_UNION_TYPE_NEW;
pj->ji_qs.ji_un.ji_newt.ji_fromsock = sock;
pj->ji_qs.ji_un.ji_newt.ji_fromaddr = get_connectaddr(sock);
pj->ji_qs.ji_un.ji_newt.ji_scriptsz = 0;
/* Per Eric R., req_mvjobfile was giving error in open_std_file,
showed up as fishy error message */
if (pj->ji_grpcache != NULL)
{
free(pj->ji_grpcache);
pj->ji_grpcache = NULL;
}
}
else
{
close_conn(sock);
}
/* SUCCESS */
return;
}
/* set remaining job structure elements */
pj->ji_qs.ji_state = JOB_STATE_TRANSIT;
pj->ji_qs.ji_substate = JOB_SUBSTATE_TRANSIN;
pj->ji_wattr[(int)JOB_ATR_mtime].at_val.at_long = (long)time_now;
pj->ji_wattr[(int)JOB_ATR_mtime].at_flags |= ATR_VFLAG_SET;
pj->ji_qs.ji_un_type = JOB_UNION_TYPE_NEW;
pj->ji_qs.ji_un.ji_newt.ji_fromsock = sock;
pj->ji_qs.ji_un.ji_newt.ji_fromaddr = get_connectaddr(sock);
pj->ji_qs.ji_un.ji_newt.ji_scriptsz = 0;
/* acknowledge the request with the job id */
if (reply_jobid(preq, pj->ji_qs.ji_jobid, BATCH_REPLY_CHOICE_Queue) != 0)
{
/* reply failed, purge the job and close the connection */
close_conn(sock);
job_purge(pj);
return;
}
/* link job into server's new jobs list request */
append_link(&svr_newjobs, &pj->ji_alljobs, pj);
return;
} /* END req_quejob() */
int send_job(
job *jobp,
pbs_net_t hostaddr, /* host address, host byte order */
int port, /* service port, host byte order */
int move_type, /* move, route, or execute */
void (*post_func)(struct work_task *), /* after move */
void *data) /* ptr to optional batch_request to be put */
/* in the work task structure */
{
tlist_head attrl;
enum conn_type cntype = ToServerDIS;
int con;
char *destin = jobp->ji_qs.ji_destin;
int encode_type;
int i;
int NumRetries;
char *id = "send_job";
attribute *pattr;
pid_t pid;
struct attropl *pqjatr; /* list (single) of attropl for quejob */
char *safail = "sigaction failed\n";
char *spfail = "sigprocmask failed\n";
char script_name[MAXPATHLEN + 1];
sigset_t child_set, all_set;
struct sigaction child_action;
struct work_task *ptask;
mbool_t Timeout = FALSE;
char *pc;
sigemptyset(&child_set);
sigaddset(&child_set, SIGCHLD);
sigfillset(&all_set);
/* block SIGCHLD until work task is established */
if (sigprocmask(SIG_BLOCK, &child_set, NULL) == -1)
{
log_err(errno,id,spfail);
pbs_errno = PBSE_SYSTEM;
log_event(
PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
jobp->ji_qs.ji_jobid,
"cannot set signal mask");
return(-1);
}
if (LOGLEVEL >= 6)
{
sprintf(log_buffer,"about to send job - type=%d",
move_type);
log_event(
PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
jobp->ji_qs.ji_jobid,
"forking in send_job");
}
pid = fork();
if (pid == -1)
{
/* error on fork */
log_err(errno, id, "fork failed\n");
if (sigprocmask(SIG_UNBLOCK, &child_set, NULL) == -1)
log_err(errno, id, spfail);
pbs_errno = PBSE_SYSTEM;
return(-1);
}
if (pid != 0)
{
/* The parent (main server) */
/* create task to monitor job startup */
/* CRI: need way to report to scheduler job is starting, not started */
ptask = set_task(WORK_Deferred_Child, pid, post_func, jobp);
if (ptask == NULL)
{
log_err(errno, id, msg_err_malloc);
return(-1);
}
ptask->wt_parm2 = data;
append_link(
&((job *)jobp)->ji_svrtask,
&ptask->wt_linkobj,
ptask);
/* now can unblock SIGCHLD */
if (sigprocmask(SIG_UNBLOCK, &child_set, NULL) == -1)
log_err(errno, id, spfail);
if (LOGLEVEL >= 1)
{
extern long DispatchTime[];
extern job *DispatchJob[];
extern char *DispatchNode[];
extern time_t time_now;
struct pbsnode *NP;
/* record job dispatch time */
int jindex;
for (jindex = 0;jindex < 20;jindex++)
{
if (DispatchJob[jindex] == NULL)
{
DispatchTime[jindex] = time_now;
DispatchJob[jindex] = jobp;
if ((NP = PGetNodeFromAddr(hostaddr)) != NULL)
DispatchNode[jindex] = NP->nd_name;
else
DispatchNode[jindex] = NULL;
break;
}
}
}
/* SUCCESS */
return(2);
} /* END if (pid != 0) */
/*
* the child process
*
* set up signal catcher for error return
*/
rpp_terminate();
child_action.sa_handler = net_move_die;
sigfillset(&child_action.sa_mask);
child_action.sa_flags = 0;
if (sigaction(SIGHUP, &child_action, NULL))
log_err(errno, id, safail);
if (sigaction(SIGINT, &child_action, NULL))
log_err(errno, id, safail);
if (sigaction(SIGQUIT, &child_action, NULL))
log_err(errno, id, safail);
/* signal handling is set, now unblock */
if (sigprocmask(SIG_UNBLOCK, &child_set, NULL) == -1)
log_err(errno, id, spfail);
/* encode job attributes to be moved */
CLEAR_HEAD(attrl);
/* select attributes/resources to send based on move type */
if (move_type == MOVE_TYPE_Exec)
{
/* moving job to MOM - ie job start */
resc_access_perm = ATR_DFLAG_MOM;
encode_type = ATR_ENCODE_MOM;
cntype = ToServerDIS;
}
else
{
/* moving job to alternate server? */
resc_access_perm =
ATR_DFLAG_USWR |
ATR_DFLAG_OPWR |
ATR_DFLAG_MGWR |
ATR_DFLAG_SvRD;
encode_type = ATR_ENCODE_SVR;
/* clear default resource settings */
svr_dequejob(jobp);
}
pattr = jobp->ji_wattr;
for (i = 0;i < JOB_ATR_LAST;i++)
{
if (((job_attr_def + i)->at_flags & resc_access_perm) ||
((strncmp((job_attr_def + i)->at_name,"session_id",10) == 0) &&
(jobp->ji_wattr[JOB_ATR_checkpoint_name].at_flags & ATR_VFLAG_SET)))
{
(job_attr_def + i)->at_encode(
pattr + i,
&attrl,
(job_attr_def + i)->at_name,
NULL,
encode_type);
}
} /* END for (i) */
attrl_fixlink(&attrl);
/* put together the job script file name */
strcpy(script_name, path_jobs);
if (jobp->ji_wattr[JOB_ATR_job_array_request].at_flags & ATR_VFLAG_SET)
{
strcat(script_name, jobp->ji_arraystruct->ai_qs.fileprefix);
}
else
{
strcat(script_name, jobp->ji_qs.ji_fileprefix);
}
strcat(script_name, JOB_SCRIPT_SUFFIX);
pbs_errno = 0;
con = -1;
for (NumRetries = 0;NumRetries < RETRY;NumRetries++)
{
int rc;
/* connect to receiving server with retries */
if (NumRetries > 0)
{
/* recycle after an error */
if (con >= 0)
svr_disconnect(con);
/* check pbs_errno from previous attempt */
if (should_retry_route(pbs_errno) == -1)
{
sprintf(log_buffer, "child failed in previous commit request for job %s",
jobp->ji_qs.ji_jobid);
log_err(pbs_errno, id, log_buffer);
exit(1); /* fatal error, don't retry */
}
sleep(1 << NumRetries);
}
/* NOTE: on node hangs, svr_connect is successful */
if ((con = svr_connect(hostaddr, port, 0, cntype)) == PBS_NET_RC_FATAL)
{
sprintf(log_buffer, "send_job failed to %lx port %d",
hostaddr,
port);
log_err(pbs_errno, id, log_buffer);
exit(1);
}
if (con == PBS_NET_RC_RETRY)
{
pbs_errno = 0; /* should retry */
continue;
}
/*
* if the job is substate JOB_SUBSTATE_TRNOUTCM which means
* we are recovering after being down or a late failure, we
* just want to send the "ready-to-commit/commit"
*/
if (jobp->ji_qs.ji_substate != JOB_SUBSTATE_TRNOUTCM)
{
if (jobp->ji_qs.ji_substate != JOB_SUBSTATE_TRNOUT)
{
jobp->ji_qs.ji_substate = JOB_SUBSTATE_TRNOUT;
job_save(jobp, SAVEJOB_QUICK, 0);
}
pqjatr = &((svrattrl *)GET_NEXT(attrl))->al_atopl;
if ((pc = PBSD_queuejob(
con,
jobp->ji_qs.ji_jobid,
destin,
pqjatr,
NULL)) == NULL)
{
if ((pbs_errno == PBSE_EXPIRED) || (pbs_errno == PBSE_READ_REPLY_TIMEOUT))
{
/* queue job timeout based on pbs_tcp_timeout */
Timeout = TRUE;
}
if ((pbs_errno == PBSE_JOBEXIST) && (move_type == MOVE_TYPE_Exec))
{
/* already running, mark it so */
log_event(
PBSEVENT_ERROR,
PBS_EVENTCLASS_JOB,
jobp->ji_qs.ji_jobid,
"MOM reports job already running");
exit(0);
}
sprintf(log_buffer, "send of job to %s failed error = %d",
destin,
pbs_errno);
log_event(
PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
jobp->ji_qs.ji_jobid,
log_buffer);
continue;
} /* END if ((pc = PBSD_queuejob() == NULL) */
free(pc);
if (jobp->ji_qs.ji_svrflags & JOB_SVFLG_SCRIPT)
{
if (PBSD_jscript(con, script_name, jobp->ji_qs.ji_jobid) != 0)
continue;
}
/* XXX may need to change the logic below, if we are sending the job to
a mom on the same host and the mom and server are not sharing the same
spool directory, then we still need to move the file */
if ((move_type == MOVE_TYPE_Exec) &&
(jobp->ji_qs.ji_svrflags & JOB_SVFLG_HASRUN) &&
(hostaddr != pbs_server_addr))
{
/* send files created on prior run */
if ((move_job_file(con,jobp,StdOut) != 0) ||
(move_job_file(con,jobp,StdErr) != 0) ||
(move_job_file(con,jobp,Checkpoint) != 0))
{
continue;
}
}
/* ignore signals */
if (sigprocmask(SIG_BLOCK, &all_set, NULL) == -1)
log_err(errno, id, "sigprocmask\n");
jobp->ji_qs.ji_substate = JOB_SUBSTATE_TRNOUTCM;
job_save(jobp, SAVEJOB_QUICK, 0);
}
else
{
/* ignore signals */
if (sigprocmask(SIG_BLOCK, &all_set, NULL) == -1)
log_err(errno, id, "sigprocmask\n");
}
if (PBSD_rdytocmt(con, jobp->ji_qs.ji_jobid) != 0)
{
if (sigprocmask(SIG_UNBLOCK, &all_set, NULL) == -1)
log_err(errno, id, "sigprocmask\n");
continue;
}
if ((rc = PBSD_commit(con, jobp->ji_qs.ji_jobid)) != 0)
{
int errno2;
/* NOTE: errno is modified by log_err */
errno2 = errno;
sprintf(log_buffer, "send_job commit failed, rc=%d (%s)",
rc,
(connection[con].ch_errtxt != NULL) ? connection[con].ch_errtxt : "N/A");
log_ext(errno2, id, log_buffer, LOG_WARNING);
/* if failure occurs, pbs_mom should purge job and pbs_server should set *
job state to idle w/error msg */
if (errno2 == EINPROGRESS)
{
/* request is still being processed */
/* increase tcp_timeout in qmgr? */
Timeout = TRUE;
/* do we need a continue here? */
sprintf(log_buffer, "child commit request timed-out for job %s, increase tcp_timeout?",
jobp->ji_qs.ji_jobid);
log_ext(errno2, id, log_buffer, LOG_WARNING);
/* don't retry on timeout--break out and report error! */
break;
}
else
{
sprintf(log_buffer, "child failed in commit request for job %s",
jobp->ji_qs.ji_jobid);
log_ext(errno2, id, log_buffer, LOG_CRIT);
/* FAILURE */
exit(1);
}
} /* END if ((rc = PBSD_commit(con,jobp->ji_qs.ji_jobid)) != 0) */
svr_disconnect(con);
/* child process is done */
/* SUCCESS */
exit(0);
} /* END for (NumRetries) */
if (con >= 0)
svr_disconnect(con);
if (Timeout == TRUE)
{
/* 10 indicates that job migrate timed out, server will mark node down *
and abort the job - see post_sendmom() */
sprintf(log_buffer, "child timed-out attempting to start job %s",
jobp->ji_qs.ji_jobid);
log_ext(pbs_errno, id, log_buffer, LOG_WARNING);
exit(10);
}
if (should_retry_route(pbs_errno) == -1)
{
sprintf(log_buffer, "child failed and will not retry job %s",
jobp->ji_qs.ji_jobid);
log_err(pbs_errno, id, log_buffer);
exit(1);
}
exit(2);
/*NOTREACHED*/
return(0);
} /* END send_job() */
void
req_register(struct batch_request *preq)
{
int made;
attribute *pattr;
struct depend *pdep;
struct depend_job *pdj;
job *pjob;
char *ps;
struct work_task *ptask;
int rc = 0;
int revtype;
int type;
int savetype = SAVEJOB_FULL;
/* make sure request is from a server */
if (!preq->rq_fromsvr) {
#ifdef NAS /* localmod 109 */
sprintf(log_buffer, "Dependency request not from server");
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_JOB, LOG_INFO,
preq->rq_ind.rq_register.rq_parent, log_buffer);
#endif /* localmod 109 */
req_reject(PBSE_IVALREQ, 0, preq);
return;
}
/* find the "parent" job specified in the request */
if ((pjob = find_job(preq->rq_ind.rq_register.rq_parent)) == NULL) {
/*
* job not found... if server is initializing, it may not
* yet recovered, that is not an error.
*/
if (server.sv_attr[(int)SRV_ATR_State].at_val.at_long != SV_STATE_INIT) {
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_JOB, LOG_INFO,
preq->rq_ind.rq_register.rq_parent,
msg_unkjobid);
req_reject(PBSE_UNKJOBID, 0, preq);
} else {
reply_ack(preq);
}
return;
}
pattr = &pjob->ji_wattr[(int)JOB_ATR_depend];
type = preq->rq_ind.rq_register.rq_dependtype;
pjob->ji_modified = 1;
/* more of the server:port fix kludge */
ps = strchr(preq->rq_ind.rq_register.rq_child, (int)'@');
if (ps != NULL) {
(void)strcpy(preq->rq_ind.rq_register.rq_svr, ps+1);
*ps = '\0';
} else {
(void)strcpy(preq->rq_ind.rq_register.rq_svr, preq->rq_host);
}
if (pjob->ji_qs.ji_state == JOB_STATE_MOVED) {
snprintf(log_buffer, sizeof(log_buffer), "Parent %s%s", msg_movejob, pjob->ji_qs.ji_destin);
log_event(PBSEVENT_DEBUG|PBSEVENT_SYSTEM|PBSEVENT_ERROR,
PBS_EVENTCLASS_REQUEST, LOG_INFO,
preq->rq_ind.rq_register.rq_child, log_buffer);
req_reject(PBSE_JOB_MOVED, 0, preq);
return;
}
switch (preq->rq_ind.rq_register.rq_op) {
/*
* Register a dependency
*/
case JOB_DEPEND_OP_REGISTER:
switch (type) {
case JOB_DEPEND_TYPE_AFTERSTART:
if (pjob->ji_qs.ji_substate >= JOB_SUBSTATE_RUNNING) {
/* job already running, setup task to send */
/* release back to child and continue with */
/* registration process */
ptask = set_task(WORK_Immed, 0, post_run_depend,
(void *)pjob);
if (ptask)
append_link(&pjob->ji_svrtask,
&ptask->wt_linkobj, ptask);
}
/* fall through to complete registration */
case JOB_DEPEND_TYPE_AFTERANY:
case JOB_DEPEND_TYPE_AFTEROK:
case JOB_DEPEND_TYPE_AFTERNOTOK:
rc = register_dep(pattr, preq, type, &made);
break;
case JOB_DEPEND_TYPE_BEFORESTART:
case JOB_DEPEND_TYPE_BEFOREANY:
case JOB_DEPEND_TYPE_BEFOREOK:
case JOB_DEPEND_TYPE_BEFORENOTOK:
/*
* Check job owner for permission, use the real
* job owner, not the sending server's name.
*/
(void)strcpy(preq->rq_user,
preq->rq_ind.rq_register.rq_owner);
if (svr_chk_owner(preq, pjob)) {
rc = PBSE_PERM; /* not same user */
} else {
/* ok owner, see if job has "on" */
pdep = find_depend(JOB_DEPEND_TYPE_ON, pattr);
if (pdep == 0) {
/* on "on", see if child already registered */
revtype = type ^ (JOB_DEPEND_TYPE_BEFORESTART -
JOB_DEPEND_TYPE_AFTERSTART);
pdep = find_depend(revtype, pattr);
if (pdep == 0) {
/* no "on" and no prior - return error */
rc = PBSE_BADDEPEND;
} else {
pdj = find_dependjob(pdep,
preq->rq_ind.rq_register.rq_child);
if (pdj) {
/* has prior register, update it */
(void)strcpy(pdj->dc_svr,
preq->rq_ind.rq_register.rq_svr);
}
}
} else if ((rc=register_dep(pattr, preq, type, &made)) == 0) {
if (made) { /* first time registered */
if (--pdep->dp_numexp <= 0)
del_depend(pdep);
}
}
}
break;
default:
#ifdef NAS /* localmod 109 */
sprintf(log_buffer, "Unknown dep. op: %d", preq->rq_ind.rq_register.rq_op);
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_JOB, LOG_INFO,
preq->rq_ind.rq_register.rq_parent, log_buffer);
#endif /* localmod 109 */
rc = PBSE_IVALREQ;
break;
}
break;
/*
* Release a dependency so job might run
*/
case JOB_DEPEND_OP_RELEASE:
switch (type) {
case JOB_DEPEND_TYPE_BEFORESTART:
case JOB_DEPEND_TYPE_BEFOREANY:
case JOB_DEPEND_TYPE_BEFOREOK:
case JOB_DEPEND_TYPE_BEFORENOTOK:
/* predecessor sent release-reduce "on", */
/* see if this job can now run */
type ^= (JOB_DEPEND_TYPE_BEFORESTART -
JOB_DEPEND_TYPE_AFTERSTART);
if ((pdep = find_depend(type, pattr)) != NULL) {
pdj = find_dependjob(pdep,
preq->rq_ind.rq_register.rq_child);
if (pdj) {
del_depend_job(pdj);
pattr->at_flags |= ATR_VFLAG_MODIFY | ATR_VFLAG_MODCACHE;
savetype = SAVEJOB_FULLFORCE;
(void)sprintf(log_buffer, msg_registerrel,
preq->rq_ind.rq_register.rq_child);
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB,
LOG_INFO,
pjob->ji_qs.ji_jobid, log_buffer);
if (GET_NEXT(pdep->dp_jobs) == 0) {
/* no more dependencies of this type */
del_depend(pdep);
set_depend_hold(pjob, pattr);
}
break;
}
#ifdef NAS /* localmod 109 */
sprintf(log_buffer, "Dep.rls. job not found: %d/%s", type, preq->rq_ind.rq_register.rq_child);
} else {
sprintf(log_buffer, "Dep.rls. type not found: %d", type);
#endif /* localmod 109 */
}
#ifdef NAS /* localmod 109 */
log_event(PBSEVENT_DEBUG, PBS_EVENTCLASS_JOB, LOG_INFO,
preq->rq_ind.rq_register.rq_parent, log_buffer);
#endif /* localmod 109 */
rc = PBSE_IVALREQ;
break;
}
break;
case JOB_DEPEND_OP_READY:
rc = PBSE_NOSYNCMSTR;
break;
case JOB_DEPEND_OP_DELETE:
(void)sprintf(log_buffer, msg_registerdel,
preq->rq_ind.rq_register.rq_child);
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB, LOG_INFO,
pjob->ji_qs.ji_jobid, log_buffer);
job_abt(pjob, log_buffer);
break;
case JOB_DEPEND_OP_UNREG:
unregister_dep(pattr, preq);
set_depend_hold(pjob, pattr);
break;
default:
sprintf(log_buffer, msg_illregister,
preq->rq_ind.rq_register.rq_parent);
log_event(PBSEVENT_DEBUG|PBSEVENT_SYSTEM|PBSEVENT_ERROR,
PBS_EVENTCLASS_REQUEST, LOG_INFO,
preq->rq_host, log_buffer);
rc = PBSE_IVALREQ;
break;;
}
if (rc) {
pjob->ji_modified = 0;
req_reject(rc, 0, preq);
} else {
/* If this is an array job, forcibly save it to ensure
* dependencies are recorded.
*/
if (pjob->ji_qs.ji_svrflags & JOB_SVFLG_ArrayJob)
savetype = SAVEJOB_FULLFORCE;
if (pjob->ji_modified)
(void)job_save(pjob, savetype);
reply_ack(preq);
}
return;
}
int reply_send(
struct batch_request *request) /* I (freed) */
{
int rc = 0;
int sfds = request->rq_conn; /* socket */
#ifndef PBS_MOM
static char *id = "reply_send";
struct work_task *ptask;
#endif /* PBS_MOM */
/* determine where the reply should go, remote or local */
if (sfds == PBS_LOCAL_CONNECTION)
{
#ifndef PBS_MOM
/*
* reply stays local, find work task and move it to
* the immediate list for dispatching.
*/
ptask = (struct work_task *)GET_NEXT(task_list_event);
while (ptask != NULL)
{
if ((ptask->wt_type == WORK_Deferred_Local) &&
(ptask->wt_parm1 == (void *)request))
{
delete_link(&ptask->wt_linkall);
append_link(&task_list_immed, &ptask->wt_linkall, ptask);
return(0);
}
ptask = (struct work_task *)GET_NEXT(ptask->wt_linkall);
}
/* should have found a task and didn't */
log_err(-1, id, "did not find work task for local request");
#endif /* PBS_MOM */
rc = PBSE_SYSTEM;
}
else if (sfds >= 0)
{
/* Otherwise, the reply is to be sent to a remote client */
#ifndef PBS_MOM
if (request->rq_noreply != TRUE)
{
#endif
rc = dis_reply_write(sfds, &request->rq_reply);
#ifndef PBS_MOM
if (LOGLEVEL >= 7)
{
sprintf(log_buffer, "Reply sent for request type %s on socket %d",
reqtype_to_txt(request->rq_type),
sfds);
log_record(
PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
id,
log_buffer);
}
#ifndef PBS_MOM
}
#endif
#endif
}
#ifndef PBS_MOM
if ((request->rq_type != PBS_BATCH_AsyModifyJob) || (request->rq_noreply == TRUE))
{
#endif
free_br(request);
#ifndef PBS_MOM
}
#endif
return(rc);
} /* END reply_send() */
int encode_jobs(
pbs_attribute *pattr, /*struct pbs_attribute being encoded */
tlist_head *ph, /*head of a list of "svrattrl" structs
which are to be returned*/
char *aname, /*pbs_attribute's name */
char *rname, /*resource's name (null if none) */
int mode, /*mode code, unused here */
int perm) /* only used for resources */
{
svrattrl *pal;
struct jobinfo *jip;
struct pbsnode *pnode;
struct pbssubn *psubn;
int i;
int jobcnt; /*number of jobs using the node */
int strsize; /*computed string size */
char *job_str; /*holds comma separated list of jobs*/
int job_str_len = 0;
if (pattr == NULL)
{
return (-1);
}
if (!(pattr->at_flags & ATR_VFLAG_SET) || !pattr->at_val.at_jinfo)
{
return(0); /* nothing to report back */
}
/* cnt number of jobs and estimate size of string buffer required */
jobcnt = 0;
strsize = 1; /*allow for terminating null char*/
pnode = pattr->at_val.at_jinfo;
for (psubn = pnode->nd_psn;psubn != NULL;psubn = psubn->next)
{
for (jip = psubn->jobs;jip != NULL;jip = jip->next)
{
jobcnt++;
strsize += strlen(jip->jobid) + PCONST_ENCOVERHEAD;
}
} /* END for (psubn) */
if (jobcnt == 0)
{
/* no jobs currently on this node */
return(0);
}
if ((job_str = (char *)calloc(1, strsize+1)) == NULL)
{
return -(PBSE_SYSTEM);
}
i = 0;
for (psubn = pnode->nd_psn;psubn != NULL;psubn = psubn->next)
{
for (jip = psubn->jobs;jip != NULL;jip = jip->next)
{
if (i != 0)
strcat(job_str, ", ");
else
i++;
sprintf(job_str + strlen(job_str), "%d/%s",
psubn->index, jip->jobid);
}
} /* END for (psubn) */
job_str_len = strlen(job_str);
pal = attrlist_create(aname, rname, job_str_len+1);
if (pal == NULL)
{
free(job_str);
return -(PBSE_SYSTEM);
}
strncpy(pal->al_value, job_str, job_str_len);
pal->al_flags = ATR_VFLAG_SET;
free(job_str);
append_link(ph, &pal->al_link, pal);
return(0); /* success */
} /* END encode_jobs() */
int encode_time(
pbs_attribute *attr, /* ptr to pbs_attribute (value in attr->at_val.at_long) */
tlist_head *phead, /* head of attrlist list (optional) */
const char *atname, /* pbs_attribute name */
const char *rsname, /* resource name (optional) */
int mode, /* encode mode (not used) */
int perm) /* only used for resources */
{
size_t ct;
char cvnbuf[ENCODE_TIME_SIZE];
int hr;
int min;
long n;
svrattrl *pal;
int sec;
char *pv;
if ((attr == NULL)||(phead == NULL))
{
/* FAILURE */
return(-1);
}
if (!(attr->at_flags & ATR_VFLAG_SET))
{
return(0);
}
n = attr->at_val.at_long;
hr = n / 3600;
n = n % 3600;
min = n / 60;
n = n % 60;
sec = n;
sprintf(cvnbuf, "%02d:%02d:%02d",
hr, min, sec);
pv = cvnbuf;
pv += strlen(pv);
ct = strlen(cvnbuf);
pal = attrlist_create(atname, rsname, ct+1);
if (pal == NULL)
{
/* FAILURE */
return(-1);
}
memcpy(pal->al_value, cvnbuf, ct);
pal->al_flags = attr->at_flags;
append_link(phead, &pal->al_link, pal);
/* SUCCESS */
return(1);
}
int status_node(
struct pbsnode *pnode, /* ptr to node receiving status query */
struct batch_request *preq,
int *bad, /* O */
tlist_head *pstathd) /* head of list to append status to */
{
int rc = 0;
struct brp_status *pstat;
svrattrl *pal;
if ((preq->rq_perm & ATR_DFLAG_RDACC) == 0)
{
return(PBSE_PERM);
}
/* allocate status sub-structure and fill in header portion */
pstat = (struct brp_status *)calloc(1, sizeof(struct brp_status));
if (pstat == NULL)
{
return(PBSE_SYSTEM);
}
pstat->brp_objtype = MGR_OBJ_NODE;
strncpy(pstat->brp_objname, pnode->nd_name, sizeof(pstat->brp_objname)-1);
CLEAR_LINK(pstat->brp_stlink);
CLEAR_HEAD(pstat->brp_attr);
/*add this new brp_status structure to the list hanging off*/
/*the request's reply substructure */
append_link(pstathd, &pstat->brp_stlink, pstat);
/*point to the list of node-attributes about which we want status*/
/*hang that status information from the brp_attr field for this */
/*brp_status structure */
*bad = 0; /*global variable*/
if (preq->rq_ind.rq_status.rq_attr.ll_struct != NULL)
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
else
pal = NULL;
rc = status_nodeattrib(
pal,
node_attr_def,
pnode,
ND_ATR_LAST,
preq->rq_perm,
&pstat->brp_attr,
bad);
return(rc);
} /* END status_node() */
/* array_recov reads in an array struct saved to disk and inserts it into
the servers list of arrays */
int array_recov(
char *path,
job_array **new_pa)
{
extern tlist_head svr_jobarrays;
job_array *pa;
array_request_node *rn;
int fd;
int old_version;
int num_tokens;
int i;
int len;
int rc;
old_version = ARRAY_QS_STRUCT_VERSION;
/* allocate the storage for the struct */
pa = (job_array*)calloc(1,sizeof(job_array));
if (pa == NULL)
{
return PBSE_SYSTEM;
}
/* initialize the linked list nodes */
CLEAR_LINK(pa->all_arrays);
CLEAR_HEAD(pa->request_tokens);
fd = open(path, O_RDONLY, 0);
if ( array_259_upgrade )
{
rc = read_and_convert_259_array(fd, pa, path);
if(rc != PBSE_NONE)
{
free(pa);
close(fd);
return rc;
}
}
else
{
/* read the file into the struct previously allocated.
*/
len = read(fd, &(pa->ai_qs), sizeof(pa->ai_qs));
if ((len < 0) || ((len < (int)sizeof(pa->ai_qs)) && (pa->ai_qs.struct_version == ARRAY_QS_STRUCT_VERSION)))
{
sprintf(log_buffer, "error reading %s", path);
log_err(errno, "array_recov", log_buffer);
free(pa);
close(fd);
return PBSE_SYSTEM;
}
if (pa->ai_qs.struct_version != ARRAY_QS_STRUCT_VERSION)
{
rc = array_upgrade(pa, fd, pa->ai_qs.struct_version, &old_version);
if(rc)
{
sprintf(log_buffer, "Cannot upgrade array version %d to %d", pa->ai_qs.struct_version, ARRAY_QS_STRUCT_VERSION);
log_err(errno, "array_recov", log_buffer);
free(pa);
close(fd);
return rc;
}
}
}
pa->jobs = malloc(sizeof(job *) * pa->ai_qs.array_size);
memset(pa->jobs,0,sizeof(job *) * pa->ai_qs.array_size);
/* check to see if there is any additional info saved in the array file */
/* check if there are any array request tokens that haven't been fully
processed */
if (old_version > 1)
{
if (read(fd, &num_tokens, sizeof(int)) != sizeof(int))
{
sprintf(log_buffer, "error reading token count from %s", path);
log_err(errno, "array_recov", log_buffer);
free(pa);
close(fd);
return PBSE_SYSTEM;
}
for (i = 0; i < num_tokens; i++)
{
rn = (array_request_node*)malloc(sizeof(array_request_node));
if (read(fd, rn, sizeof(array_request_node)) != sizeof(array_request_node))
{
sprintf(log_buffer, "error reading array_request_node from %s", path);
log_err(errno, "array_recov", log_buffer);
free(rn);
for (rn = (array_request_node*)GET_NEXT(pa->request_tokens);
rn != NULL;
rn = (array_request_node*)GET_NEXT(pa->request_tokens))
{
delete_link(&rn->request_tokens_link);
free(rn);
}
free(pa);
close(fd);
return PBSE_SYSTEM;
}
CLEAR_LINK(rn->request_tokens_link);
append_link(&pa->request_tokens, &rn->request_tokens_link, (void*)rn);
}
}
close(fd);
CLEAR_HEAD(pa->ai_qs.deps);
if (old_version != ARRAY_QS_STRUCT_VERSION)
{
/* resave the array struct if the version on disk is older than the current */
array_save(pa);
}
/* link the struct into the servers list of job arrays */
append_link(&svr_jobarrays, &pa->all_arrays, (void*)pa);
*new_pa = pa;
return PBSE_NONE;
}
static void req_stat_job_step2(
struct stat_cntl *cntl) /* I/O (free'd on return) */
{
svrattrl *pal;
job *pjob = NULL;
struct batch_request *preq;
struct batch_reply *preply;
int rc = 0;
enum TJobStatTypeEnum type;
pbs_queue *pque = NULL;
int exec_only = 0;
int bad = 0;
long DTime; /* delta time - only report full pbs_attribute list if J->MTime > DTime */
static svrattrl *dpal = NULL;
int job_array_index = 0;
job_array *pa = NULL;
char log_buf[LOCAL_LOG_BUF_SIZE];
int iter;
time_t time_now = time(NULL);
long poll_jobs = 0;
char job_id[PBS_MAXSVRJOBID+1];
int job_substate = -1;
time_t job_momstattime = -1;
preq = cntl->sc_origrq;
type = (enum TJobStatTypeEnum)cntl->sc_type;
preply = &preq->rq_reply;
/* See pbs_server_attributes(1B) for details on "poll_jobs" behaviour */
if (dpal == NULL)
{
/* build 'delta' pbs_attribute list */
svrattrl *tpal;
tlist_head dalist;
int aindex;
int atrlist[] =
{
JOB_ATR_jobname,
JOB_ATR_resc_used,
JOB_ATR_LAST
};
CLEAR_LINK(dalist);
for (aindex = 0;atrlist[aindex] != JOB_ATR_LAST;aindex++)
{
if ((tpal = attrlist_create("", "", 23)) == NULL)
{
return;
}
tpal->al_valln = atrlist[aindex];
if (dpal == NULL)
dpal = tpal;
append_link(&dalist, &tpal->al_link, tpal);
}
} /* END if (dpal == NULL) */
if (type == tjstArray)
{
pa = get_array(preq->rq_ind.rq_status.rq_id);
if (pa == NULL)
{
req_reject(PBSE_UNKARRAYID, 0, preq, NULL, "unable to find array");
return;
}
}
iter = -1;
get_svr_attr_l(SRV_ATR_PollJobs, &poll_jobs);
if (!poll_jobs)
{
/* polljobs not set - indicates we may need to obtain fresh data from
MOM */
if (cntl->sc_jobid[0] == '\0')
pjob = NULL;
else
pjob = svr_find_job(cntl->sc_jobid, FALSE);
while (1)
{
if (pjob == NULL)
{
/* start from the first job */
if (type == tjstJob)
{
pjob = svr_find_job(preq->rq_ind.rq_status.rq_id, FALSE);
}
else if (type == tjstQueue)
{
pjob = next_job(cntl->sc_pque->qu_jobs,&iter);
}
else if (type == tjstArray)
{
job_array_index = 0;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (job_array_index < pa->ai_qs.array_size)
{
if (pa->job_ids[job_array_index] != NULL)
{
if ((pjob = svr_find_job(pa->job_ids[job_array_index], FALSE)) != NULL)
{
unlock_ji_mutex(pjob, __func__, "2", LOGLEVEL);
break;
}
}
job_array_index++;
}
}
else
{
pjob = next_job(&alljobs,&iter);
}
} /* END if (pjob == NULL) */
else
{
strcpy(job_id, pjob->ji_qs.ji_jobid);
unlock_ji_mutex(pjob, __func__, "3", LOGLEVEL);
if (type == tjstJob)
break;
if (type == tjstQueue)
pjob = next_job(cntl->sc_pque->qu_jobs,&iter);
else if (type == tjstArray)
{
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (++job_array_index < pa->ai_qs.array_size)
{
if (pa->job_ids[job_array_index] != NULL)
{
if ((pjob = svr_find_job(pa->job_ids[job_array_index], FALSE)) != NULL)
{
unlock_ji_mutex(pjob, __func__, "3", LOGLEVEL);
break;
}
}
}
}
else
pjob = next_job(&alljobs,&iter);
}
if (pjob == NULL)
break;
strcpy(job_id, pjob->ji_qs.ji_jobid);
job_substate = pjob->ji_qs.ji_substate;
job_momstattime = pjob->ji_momstat;
strcpy(cntl->sc_jobid, job_id);
unlock_ji_mutex(pjob, __func__, "4", LOGLEVEL);
pjob = NULL;
/* PBS_RESTAT_JOB defaults to 30 seconds */
if ((job_substate == JOB_SUBSTATE_RUNNING) &&
((time_now - job_momstattime) > JobStatRate))
{
/* go to MOM for status */
if ((rc = stat_to_mom(job_id, cntl)) == PBSE_MEM_MALLOC)
break;
if (rc != 0)
{
pjob = svr_find_job(job_id, FALSE);
rc = 0;
continue;
}
if (pa != NULL)
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
return; /* will pick up after mom replies */
}
} /* END while(1) */
if (rc != 0)
{
if (pa != NULL)
unlock_ai_mutex(pa, __func__, "2", LOGLEVEL);
reply_free(preply);
req_reject(rc, 0, preq, NULL, "cannot get update from mom");
return;
}
} /* END if (!server.sv_attr[SRV_ATR_PollJobs].at_val.at_long) */
/*
* now ready for part 3, building the status reply,
* loop through again
*/
if ((type == tjstSummarizeArraysQueue) ||
(type == tjstSummarizeArraysServer))
{
/* No array can be owned for these options */
update_array_statuses();
}
if (type == tjstJob)
pjob = svr_find_job(preq->rq_ind.rq_status.rq_id, FALSE);
else if (type == tjstQueue)
pjob = next_job(cntl->sc_pque->qu_jobs,&iter);
else if (type == tjstSummarizeArraysQueue)
pjob = next_job(cntl->sc_pque->qu_jobs_array_sum,&iter);
else if (type == tjstSummarizeArraysServer)
pjob = next_job(&array_summary,&iter);
else if (type == tjstArray)
{
job_array_index = -1;
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (++job_array_index < pa->ai_qs.array_size)
{
if (pa->job_ids[job_array_index] != NULL)
{
if ((pjob = svr_find_job(pa->job_ids[job_array_index], FALSE)) != NULL)
{
break;
}
}
}
}
else
pjob = next_job(&alljobs,&iter);
DTime = 0;
if (preq->rq_extend != NULL)
{
char *ptr;
/* FORMAT: { EXECQONLY | DELTA:<EPOCHTIME> } */
if (strstr(preq->rq_extend, EXECQUEONLY))
exec_only = 1;
ptr = strstr(preq->rq_extend, "DELTA:");
if (ptr != NULL)
{
ptr += strlen("delta:");
DTime = strtol(ptr, NULL, 10);
}
}
if ((type == tjstTruncatedServer) ||
(type == tjstTruncatedQueue))
{
long sentJobCounter;
long qjcounter;
long qmaxreport;
int iter = -1;
/* loop through all queues */
while ((pque = next_queue(&svr_queues,&iter)) != NULL)
{
qjcounter = 0;
if ((exec_only == 1) &&
(pque->qu_qs.qu_type != QTYPE_Execution))
{
/* ignore routing queues */
unlock_queue(pque, __func__, "ignore queue", LOGLEVEL);
continue;
}
if (((pque->qu_attr[QA_ATR_MaxReport].at_flags & ATR_VFLAG_SET) != 0) &&
(pque->qu_attr[QA_ATR_MaxReport].at_val.at_long >= 0))
{
qmaxreport = pque->qu_attr[QA_ATR_MaxReport].at_val.at_long;
}
else
{
qmaxreport = TMAX_JOB;
}
if (LOGLEVEL >= 5)
{
sprintf(log_buf,"giving scheduler up to %ld idle jobs in queue %s\n",
qmaxreport,
pque->qu_qs.qu_name);
log_event(PBSEVENT_SYSTEM,PBS_EVENTCLASS_QUEUE,pque->qu_qs.qu_name,log_buf);
}
sentJobCounter = 0;
/* loop through jobs in queue */
if (pjob != NULL)
unlock_ji_mutex(pjob, __func__, "5", LOGLEVEL);
iter = -1;
while ((pjob = next_job(pque->qu_jobs,&iter)) != NULL)
{
if ((qjcounter >= qmaxreport) &&
(pjob->ji_qs.ji_state == JOB_STATE_QUEUED))
{
/* max_report of queued jobs reached for queue */
unlock_ji_mutex(pjob, __func__, "6", LOGLEVEL);
continue;
}
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
rc = status_job(
pjob,
preq,
(pjob->ji_wattr[JOB_ATR_mtime].at_val.at_long >= DTime) ? pal : dpal,
&preply->brp_un.brp_status,
&bad);
if ((rc != 0) && (rc != PBSE_PERM))
{
req_reject(rc, bad, preq, NULL, NULL);
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
unlock_ji_mutex(pjob, __func__, "7", LOGLEVEL);
unlock_queue(pque, __func__, "perm", LOGLEVEL);
return;
}
sentJobCounter++;
if (pjob->ji_qs.ji_state == JOB_STATE_QUEUED)
qjcounter++;
unlock_ji_mutex(pjob, __func__, "8", LOGLEVEL);
} /* END foreach (pjob from pque) */
if (LOGLEVEL >= 5)
{
sprintf(log_buf,"sent scheduler %ld total jobs for queue %s\n",
sentJobCounter,
pque->qu_qs.qu_name);
log_event(PBSEVENT_SYSTEM,PBS_EVENTCLASS_QUEUE,pque->qu_qs.qu_name,log_buf);
}
unlock_queue(pque, __func__, "end while", LOGLEVEL);
} /* END for (pque) */
if (pa != NULL)
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
reply_send_svr(preq);
return;
} /* END if ((type == tjstTruncatedServer) || ...) */
while (pjob != NULL)
{
/* go ahead and build the status reply for this job */
if (exec_only)
{
if (cntl->sc_pque != NULL)
{
if (cntl->sc_pque->qu_qs.qu_type != QTYPE_Execution)
goto nextjob;
}
else
{
if (pa != NULL)
pthread_mutex_unlock(pa->ai_mutex);
pque = get_jobs_queue(&pjob);
if (pa != NULL)
pthread_mutex_lock(pa->ai_mutex);
if ((pjob == NULL) ||
(pque == NULL))
goto nextjob;
if (pque->qu_qs.qu_type != QTYPE_Execution)
{
unlock_queue(pque, __func__, "not exec", LOGLEVEL);
goto nextjob;
}
unlock_queue(pque, __func__, "exec", LOGLEVEL);
}
}
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
rc = status_job(
pjob,
preq,
pal,
&preply->brp_un.brp_status,
&bad);
if ((rc != 0) &&
(rc != PBSE_PERM))
{
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
unlock_ji_mutex(pjob, __func__, "9", LOGLEVEL);
req_reject(rc, bad, preq, NULL, NULL);
return;
}
/* get next job */
nextjob:
if (pjob != NULL)
unlock_ji_mutex(pjob, __func__, "10", LOGLEVEL);
if (type == tjstJob)
break;
if (type == tjstQueue)
pjob = next_job(cntl->sc_pque->qu_jobs,&iter);
else if (type == tjstSummarizeArraysQueue)
pjob = next_job(cntl->sc_pque->qu_jobs_array_sum,&iter);
else if (type == tjstSummarizeArraysServer)
pjob = next_job(&array_summary,&iter);
else if (type == tjstArray)
{
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (++job_array_index < pa->ai_qs.array_size)
{
if (pa->job_ids[job_array_index] != NULL)
{
if ((pjob = svr_find_job(pa->job_ids[job_array_index], FALSE)) != NULL)
{
break;
}
}
}
}
else
pjob = next_job(&alljobs,&iter);
rc = 0;
} /* END while (pjob != NULL) */
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
reply_send_svr(preq);
if (LOGLEVEL >= 7)
{
log_event(PBSEVENT_SYSTEM,
PBS_EVENTCLASS_JOB,
"req_statjob",
"Successfully returned the status of queued jobs\n");
}
return;
} /* END req_stat_job_step2() */
int setup_array_struct(job *pjob)
{
job_array *pa;
/* struct work_task *wt; */
array_request_node *rn;
int bad_token_count;
int array_size;
int rc;
/* setup a link to this job array in the servers all_arrays list */
pa = (job_array *)calloc(1,sizeof(job_array));
pa->ai_qs.struct_version = ARRAY_QS_STRUCT_VERSION;
pa->template_job = pjob;
/*pa->ai_qs.array_size = pjob->ji_wattr[(int)JOB_ATR_job_array_size].at_val.at_long;*/
strcpy(pa->ai_qs.parent_id, pjob->ji_qs.ji_jobid);
strcpy(pa->ai_qs.fileprefix, pjob->ji_qs.ji_fileprefix);
strncpy(pa->ai_qs.owner, pjob->ji_wattr[JOB_ATR_job_owner].at_val.at_str, PBS_MAXUSER + PBS_MAXSERVERNAME + 2);
strncpy(pa->ai_qs.submit_host, get_variable(pjob, pbs_o_host), PBS_MAXSERVERNAME);
pa->ai_qs.num_cloned = 0;
CLEAR_LINK(pa->all_arrays);
CLEAR_HEAD(pa->request_tokens);
append_link(&svr_jobarrays, &pa->all_arrays, (void*)pa);
if (job_save(pjob, SAVEJOB_FULL, 0) != 0)
{
job_purge(pjob);
if (LOGLEVEL >= 6)
{
log_record(
PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
(pjob != NULL) ? pjob->ji_qs.ji_jobid : "NULL",
"cannot save job");
}
return 1;
}
if ((rc = set_slot_limit(pjob->ji_wattr[JOB_ATR_job_array_request].at_val.at_str, pa)))
{
array_delete(pa);
snprintf(log_buffer,sizeof(log_buffer),
"Array %s requested a slot limit above the max limit %ld, rejecting\n",
pa->ai_qs.parent_id,
server.sv_attr[SRV_ATR_MaxSlotLimit].at_val.at_long);
log_event(PBSEVENT_SYSTEM,
PBS_EVENTCLASS_JOB,
pa->ai_qs.parent_id,
log_buffer);
return(INVALID_SLOT_LIMIT);
}
pa->ai_qs.jobs_running = 0;
pa->ai_qs.num_started = 0;
pa->ai_qs.num_failed = 0;
pa->ai_qs.num_successful = 0;
bad_token_count =
parse_array_request(pjob->ji_wattr[JOB_ATR_job_array_request].at_val.at_str,
&(pa->request_tokens));
/* get the number of elements that should be allocated in the array */
rn = (array_request_node *)GET_NEXT(pa->request_tokens);
array_size = 0;
pa->ai_qs.num_jobs = 0;
while (rn != NULL)
{
if (rn->end > array_size)
array_size = rn->end;
/* calculate the actual number of jobs (different from array size) */
pa->ai_qs.num_jobs += rn->end - rn->start + 1;
rn = (array_request_node *)GET_NEXT(rn->request_tokens_link);
}
/* size of array is the biggest index + 1 */
array_size++;
if (server.sv_attr[SRV_ATR_MaxArraySize].at_flags & ATR_VFLAG_SET)
{
int max_array_size = server.sv_attr[SRV_ATR_MaxArraySize].at_val.at_long;
if (max_array_size < pa->ai_qs.num_jobs)
{
array_delete(pa);
return(ARRAY_TOO_LARGE);
}
}
/* initialize the array */
pa->jobs = malloc(array_size * sizeof(job *));
memset(pa->jobs,0,array_size * sizeof(job *));
/* remember array_size */
pa->ai_qs.array_size = array_size;
CLEAR_HEAD(pa->ai_qs.deps);
array_save(pa);
if (bad_token_count > 0)
{
array_delete(pa);
return 2;
}
return 0;
}
int copy_batchrequest(
struct batch_request **newreq,
struct batch_request *preq,
int type,
int jobid)
{
struct batch_request *request;
svrattrl *pal = NULL;
svrattrl *newpal = NULL;
tlist_head *phead = NULL;
char *ptr1;
char *ptr2;
char newjobname[PBS_MAXSVRJOBID+1];
request = alloc_br(type);
if (request)
{
request->rq_type = preq->rq_type;
request->rq_perm = preq->rq_perm;
request->rq_fromsvr = preq->rq_fromsvr;
request->rq_conn = preq->rq_conn;
request->rq_orgconn = preq->rq_orgconn;
request->rq_extsz = preq->rq_extsz;
request->rq_time = preq->rq_time;
strcpy(request->rq_user, preq->rq_user);
strcpy(request->rq_host, preq->rq_host);
request->rq_reply.brp_choice = preq->rq_reply.brp_choice;
request->rq_noreply = preq->rq_noreply;
/* we need to copy rq_extend if there is any data */
if (preq->rq_extend)
{
request->rq_extend = (char *)calloc(1, strlen(preq->rq_extend) + 1);
if (request->rq_extend == NULL)
{
free_br(request);
return(PBSE_SYSTEM);
}
strcpy(request->rq_extend, preq->rq_extend);
}
/* remember the batch_request we copied */
request->rq_extra = (void *)preq;
switch(preq->rq_type)
{
/* This function was created for a modify arracy request (PBS_BATCH_ModifyJob)
the preq->rq_ind structure was allocated in dis_request_read. If other
BATCH types are needed refer to that function to see how the rq_ind structure
was allocated and then copy it here. */
case PBS_BATCH_DeleteJob:
case PBS_BATCH_HoldJob:
case PBS_BATCH_CheckpointJob:
case PBS_BATCH_ModifyJob:
case PBS_BATCH_AsyModifyJob:
/* based on how decode_DIS_Manage allocates data */
CLEAR_HEAD(request->rq_ind.rq_manager.rq_attr);
phead = &request->rq_ind.rq_manager.rq_attr;
request->rq_ind.rq_manager.rq_cmd = preq->rq_ind.rq_manager.rq_cmd;
request->rq_ind.rq_manager.rq_objtype = preq->rq_ind.rq_manager.rq_objtype;
/* If this is a job array it is possible we only have the array name
and not the individual job. We need to find out what we have and
modify the name if needed */
ptr1 = strstr(preq->rq_ind.rq_manager.rq_objname, "[]");
if ((ptr1) && (jobid != -1))
{
ptr1++;
strcpy(newjobname, preq->rq_ind.rq_manager.rq_objname);
ptr2 = strstr(newjobname, "[]");
ptr2++;
*ptr2 = 0;
sprintf(request->rq_ind.rq_manager.rq_objname,"%s%d%s",
newjobname,
jobid,
ptr1);
}
else
strcpy(request->rq_ind.rq_manager.rq_objname, preq->rq_ind.rq_manager.rq_objname);
/* copy the attribute list */
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_manager.rq_attr);
while(pal != NULL)
{
newpal = (svrattrl *)calloc(1, pal->al_tsize + 1);
if (!newpal)
{
free_br(request);
return(PBSE_SYSTEM);
}
CLEAR_LINK(newpal->al_link);
newpal->al_atopl.next = 0;
newpal->al_tsize = pal->al_tsize + 1;
newpal->al_nameln = pal->al_nameln;
newpal->al_flags = pal->al_flags;
newpal->al_atopl.name = (char *)newpal + sizeof(svrattrl);
strcpy(newpal->al_atopl.name, pal->al_atopl.name);
newpal->al_nameln = pal->al_nameln;
newpal->al_atopl.resource = newpal->al_atopl.name + newpal->al_nameln;
if (pal->al_atopl.resource != NULL)
strcpy(newpal->al_atopl.resource, pal->al_atopl.resource);
newpal->al_rescln = pal->al_rescln;
newpal->al_atopl.value = newpal->al_atopl.name + newpal->al_nameln + newpal->al_rescln;
strcpy(newpal->al_atopl.value, pal->al_atopl.value);
newpal->al_valln = pal->al_valln;
newpal->al_atopl.op = pal->al_atopl.op;
pal = (struct svrattrl *)GET_NEXT(pal->al_link);
}
break;
case PBS_BATCH_SignalJob:
strcpy(request->rq_ind.rq_signal.rq_jid, preq->rq_ind.rq_signal.rq_jid);
strcpy(request->rq_ind.rq_signal.rq_signame, preq->rq_ind.rq_signal.rq_signame);
request->rq_extra = strdup(preq->rq_extra);
break;
case PBS_BATCH_MessJob:
strcpy(request->rq_ind.rq_message.rq_jid, preq->rq_ind.rq_message.rq_jid);
request->rq_ind.rq_message.rq_file = preq->rq_ind.rq_message.rq_file;
strcpy(request->rq_ind.rq_message.rq_text, preq->rq_ind.rq_message.rq_text);
break;
default:
break;
}
if ((phead != NULL) &&
(newpal != NULL))
append_link(phead, &newpal->al_link, newpal);
*newreq = request;
return(0);
}
else
return(PBSE_SYSTEM);
}
static int parse_array_request(char *request, tlist_head *tl)
{
char *temp_str;
int num_tokens;
char **tokens;
int i;
int j;
int num_elements;
int start;
int end;
int num_bad_tokens;
int searching;
array_request_node *rn;
array_request_node *rn2;
temp_str = strdup(request);
num_tokens = array_request_token_count(request);
num_bad_tokens = 0;
tokens = (char**)malloc(sizeof(char*) * num_tokens);
j = num_tokens - 1;
/* start from back and scan backwards setting pointers to tokens and changing ',' to '\0' */
for (i = strlen(temp_str) - 1; i >= 0; i--)
{
if (temp_str[i] == ',')
{
tokens[j--] = &temp_str[i+1];
temp_str[i] = '\0';
}
else if (i == 0)
{
tokens[0] = temp_str;
}
}
for (i = 0; i < num_tokens; i++)
{
num_elements = array_request_parse_token(tokens[i], &start, &end);
if (num_elements == 0)
{
num_bad_tokens++;
}
else
{
rn = (array_request_node*)malloc(sizeof(array_request_node));
rn->start = start;
rn->end = end;
CLEAR_LINK(rn->request_tokens_link);
rn2 = GET_NEXT(*tl);
searching = TRUE;
while (searching)
{
if (rn2 == NULL)
{
append_link(tl, &rn->request_tokens_link, (void*)rn);
searching = FALSE;
}
else if (rn->start < rn2->start)
{
insert_link(&rn2->request_tokens_link, &rn->request_tokens_link, (void*)rn,
LINK_INSET_BEFORE);
searching = FALSE;
}
else
{
rn2 = GET_NEXT(rn2->request_tokens_link);
}
}
rn2 = GET_PRIOR(rn->request_tokens_link);
if (rn2 != NULL && rn2->end >= rn->start)
{
num_bad_tokens++;
}
rn2 = GET_NEXT(rn->request_tokens_link);
if (rn2 != NULL && rn2->start <= rn->end)
{
num_bad_tokens++;
}
}
}
free(tokens);
free(temp_str);
return num_bad_tokens;
}
int status_job(
job *pjob, /* ptr to job to status */
batch_request *preq,
svrattrl *pal, /* specific attributes to status */
tlist_head *pstathd, /* RETURN: head of list to append status to */
bool condensed,
int *bad) /* RETURN: index of first bad pbs_attribute */
{
struct brp_status *pstat;
int IsOwner = 0;
long query_others = 0;
long condensed_timeout = JOB_CONDENSED_TIMEOUT;
/* Make sure procct is removed from the job
resource attributes */
remove_procct(pjob);
/* see if the client is authorized to status this job */
if (svr_authorize_jobreq(preq, pjob) == 0)
IsOwner = 1;
get_svr_attr_l(SRV_ATR_query_others, &query_others);
if (!query_others)
{
if (IsOwner == 0)
{
return(PBSE_PERM);
}
}
get_svr_attr_l(SRV_ATR_job_full_report_time, &condensed_timeout);
// if the job has been modified within the timeout, send the full output
if ((condensed == true) &&
(time(NULL) < pjob->ji_mod_time + condensed_timeout))
condensed = false;
/* allocate reply structure and fill in header portion */
if ((pstat = (struct brp_status *)calloc(1, sizeof(struct brp_status))) == NULL)
{
return(PBSE_SYSTEM);
}
CLEAR_LINK(pstat->brp_stlink);
pstat->brp_objtype = MGR_OBJ_JOB;
strcpy(pstat->brp_objname, pjob->ji_qs.ji_jobid);
CLEAR_HEAD(pstat->brp_attr);
append_link(pstathd, &pstat->brp_stlink, pstat);
/* add attributes to the status reply */
*bad = 0;
if (status_attrib(
pal,
job_attr_def,
pjob->ji_wattr,
JOB_ATR_LAST,
preq->rq_perm,
&pstat->brp_attr,
condensed,
bad,
IsOwner))
{
return(PBSE_NOATTR);
}
return (0);
} /* END status_job() */
int status_job(
job *pjob, /* ptr to job to status */
struct batch_request *preq,
svrattrl *pal, /* specific attributes to status */
tlist_head *pstathd, /* RETURN: head of list to append status to */
int *bad) /* RETURN: index of first bad attribute */
{
struct brp_status *pstat;
int IsOwner = 0;
/* see if the client is authorized to status this job */
if (svr_authorize_jobreq(preq, pjob) == 0)
IsOwner = 1;
if (!server.sv_attr[SRV_ATR_query_others].at_val.at_long)
{
if (IsOwner == 0)
{
return(PBSE_PERM);
}
}
/* allocate reply structure and fill in header portion */
pstat = (struct brp_status *)malloc(sizeof(struct brp_status));
if (pstat == NULL)
{
return(PBSE_SYSTEM);
}
memset(pstat, 0, sizeof(struct brp_status));
CLEAR_LINK(pstat->brp_stlink);
pstat->brp_objtype = MGR_OBJ_JOB;
strcpy(pstat->brp_objname, pjob->ji_qs.ji_jobid);
CLEAR_HEAD(pstat->brp_attr);
append_link(pstathd, &pstat->brp_stlink, pstat);
/* add attributes to the status reply */
*bad = 0;
if (status_attrib(
pal,
job_attr_def,
pjob->ji_wattr,
JOB_ATR_LAST,
preq->rq_perm,
&pstat->brp_attr,
bad,
IsOwner))
{
return(PBSE_NOATTR);
}
return (0);
} /* END status_job() */
