END_TEST
START_TEST(find_job_by_array_with_removed_record_test)
{
int result;
all_jobs alljobs;
struct job *test_job1 = job_alloc();
strcpy(test_job1->ji_qs.ji_jobid, "test_job1");
result = insert_job(&alljobs,test_job1);
fail_unless(result == PBSE_NONE, "job insert fail1");
struct job *test_job2 = job_alloc();
strcpy(test_job2->ji_qs.ji_jobid, "test_job2");
result = insert_job(&alljobs,test_job2);
fail_unless(result == PBSE_NONE, "job insert fail2");
struct job *test_job3 = job_alloc();
strcpy(test_job3->ji_qs.ji_jobid, "test_job3");
result = insert_job(&alljobs,test_job3);
fail_unless(result == PBSE_NONE, "job insert fai3");
struct job *test_job4 = job_alloc();
strcpy(test_job4->ji_qs.ji_jobid, "test_job4");
result = insert_job(&alljobs,test_job4);
fail_unless(result == PBSE_NONE, "job insert fail4");
struct job *test_job5 = job_alloc();
strcpy(test_job5->ji_qs.ji_jobid, "test_job5");
result = insert_job(&alljobs,test_job5);
fail_unless(result == PBSE_NONE, "job insert fail5");
}
END_TEST
START_TEST(swap_jobs_test)
{
all_jobs alljobs;
struct job *test_job;
struct job *second_test_job;
int result;
test_job = job_alloc();
second_test_job = job_alloc();
strcpy(test_job->ji_qs.ji_jobid,"test");
strcpy(second_test_job->ji_qs.ji_jobid,"second_test");
result = swap_jobs(&alljobs,NULL,test_job);
fail_unless(result != PBSE_NONE, "NULL first input job fail");
result = insert_job_after(&alljobs,test_job,NULL);
fail_unless(result != PBSE_NONE, "NULL second input job fail");
insert_job(&alljobs, test_job);
insert_job(&alljobs, second_test_job);
result = swap_jobs(&alljobs, test_job,second_test_job);
fail_unless(result == PBSE_NONE, "swap jobs fail");
}
END_TEST
START_TEST(swap_jobs_test)
{
struct all_jobs alljobs;
struct job *test_job;
struct job *second_test_job;
int result;
initialize_all_jobs_array(&alljobs);
test_job = job_alloc();
second_test_job = job_alloc();
result = swap_jobs(&alljobs,NULL,test_job);
fail_unless(result != PBSE_NONE, "NULL first input job fail");
result = insert_job_after(&alljobs,test_job,NULL);
fail_unless(result != PBSE_NONE, "NULL second input job fail");
insert_job(&alljobs, test_job);
insert_job(&alljobs, second_test_job);
result = swap_jobs(&alljobs, test_job,second_test_job);
fail_unless(result == PBSE_NONE, "swap jobs fail");
}
END_TEST
START_TEST(test_remove_some_recycle_jobs)
{
job *pjobs[1000];
initialize_recycler();
for (int i = 0; i < 1000; i++)
{
pjobs[i] = job_alloc();
fail_unless(insert_into_recycler(pjobs[i]) == PBSE_NONE);
// make the first 700 get removed
if (i < 700)
pjobs[i]->ji_momstat = 0;
}
pthread_t t1;
pthread_t t2;
pthread_t t3;
pthread_create(&t1, NULL, remove_some_recycle_jobs, NULL);
pthread_create(&t2, NULL, remove_some_recycle_jobs, NULL);
pthread_create(&t3, NULL, remove_some_recycle_jobs, NULL);
pthread_join(t1, NULL);
pthread_join(t2, NULL);
pthread_join(t3, NULL);
// 300 should be left
fail_unless(recycler.rc_jobs.count() == 300);
}
END_TEST
START_TEST(insert_job_after_test)
{
all_jobs alljobs;
struct job *test_job;
int result;
test_job = job_alloc();
strcpy(test_job->ji_qs.ji_jobid,"mylittlejob");
result = insert_job_after(NULL,test_job,test_job);
fail_unless(result != PBSE_NONE, "insert into null array fail");
result = insert_job_after(&alljobs,(char *)NULL,test_job);
fail_unless(result != PBSE_NONE, "NULL job after insert fail");
result = insert_job_after(&alljobs,test_job,NULL);
fail_unless(result != PBSE_NONE, "NULL job to insert fail");
insert_job(&alljobs,test_job);
result = insert_job_after(&alljobs,test_job,test_job);
fail_unless(result == PBSE_NONE, "job insert fail");
}
END_TEST
START_TEST(insert_job_after_test)
{
struct all_jobs alljobs;
struct job *test_job;
int result;
initialize_all_jobs_array(&alljobs);
test_job = job_alloc();
result = insert_job_after(NULL,test_job,test_job);
fail_unless(result != PBSE_NONE, "insert into null array fail");
result = insert_job_after(&alljobs,NULL,test_job);
fail_unless(result != PBSE_NONE, "NULL job after insert fail");
result = insert_job_after(&alljobs,test_job,NULL);
fail_unless(result != PBSE_NONE, "NULL job to insert fail");
insert_job(&alljobs,test_job);
result = insert_job_after(&alljobs,test_job,test_job);
fail_unless(result == PBSE_NONE, "job insert fail");
}
END_TEST
START_TEST(test_pop_job_from_recycler)
{
job *pjobs[10];
initialize_recycler();
while (recycler.rc_jobs.count() > 0)
pop_job_from_recycler(&recycler.rc_jobs);
for (int i = 0; i < 10; i++)
{
pjobs[i] = job_alloc();
fail_unless(insert_into_recycler(pjobs[i]) == PBSE_NONE);
}
for (unsigned int i = 0; i < 10; i++)
{
job *pjob = pop_job_from_recycler(&recycler.rc_jobs);
fail_unless(pjob == pjobs[i]);
fail_unless(recycler.rc_jobs.count() == 9 - i);
}
for (int i = 0; i < 3; i++)
fail_unless(pop_job_from_recycler(&recycler.rc_jobs) == NULL);
// test for records already freed or that it has not been recycled
for (int i = 0; i < 5; i++)
{
pjobs[i] = job_alloc();
fail_unless(insert_into_recycler(pjobs[i]) == PBSE_NONE);
}
job *pjob = pop_job_from_recycler(&recycler.rc_jobs);
int count = 0;
char buf[80];
while(pjob)
{
count++;
snprintf(buf,sizeof(buf),"%016lx",(long)pjob);
fail_unless(strcmp(pjob->ji_qs.ji_jobid, buf)==0, pjob->ji_qs.ji_jobid);
pjob = pop_job_from_recycler(&recycler.rc_jobs);
}
fail_unless(count == 5, "count of valid recycled jobs were wrong: %d", count);
}
END_TEST
START_TEST(job_alloc_test)
{
struct job *result = NULL;
result = job_alloc();
fail_unless(result != NULL, "job was not allocated");
}
END_TEST /* handle_aborted_job_test */
START_TEST(job_alloc_test)
{
struct job *result = NULL;
result = job_alloc();
fail_unless(result != NULL, "job was not allocated");
}
job *job_recov(
char *filename) /* I */ /* pathname to job save file */
{
job *pj;
char namebuf[MAXPATHLEN];
char log_buf[LOCAL_LOG_BUF_SIZE];
int rc;
pj = job_alloc(); /* allocate & initialize job structure space */
if (pj == NULL)
{
/* FAILURE - cannot alloc memory */
return(NULL);
}
snprintf(namebuf, MAXPATHLEN, "%s%s", path_jobs, filename); /* job directory path, filename */
size_t logBufLen = sizeof(log_buf);
if ((rc = job_recov_xml(namebuf, &pj, log_buf, logBufLen)) && rc == PBSE_INVALID_SYNTAX)
rc = job_recov_binary(namebuf, &pj, log_buf, logBufLen);
if (rc == PBSE_NONE)
rc = set_array_job_ids(&pj, log_buf, logBufLen);
if (rc != PBSE_NONE)
{
if (rc == -1)
{
log_err(errno, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "1", LOGLEVEL);
free(pj->ji_mutex);
#endif
free((char *)pj);
} /* sometime pjob is freed by abt_job() */
return(NULL);
}
pj->ji_commit_done = 1;
/* all done recovering the job */
#ifdef PBS_MOM
job_save(pj, SAVEJOB_FULL, (multi_mom == 0)?0:pbs_rm_port);
#else
job_save(pj, SAVEJOB_FULL, 0);
#endif
return(pj);
} /* END job_recov() */
END_TEST
START_TEST(cleanup_restart_file_test)
{
struct job* test_job = NULL;
cleanup_restart_file(test_job); /*TODO: add a kind of assert*/
test_job = job_alloc();
cleanup_restart_file(test_job);
}
END_TEST
START_TEST(remove_checkpoint_test)
{
struct job* test_job = NULL;
remove_checkpoint(NULL); /*TODO: add a kind of assert*/
remove_checkpoint(&test_job);
test_job = job_alloc();
remove_checkpoint(&test_job);
}
END_TEST
START_TEST(svr_job_purge_test)
{
struct job* test_job = NULL;
int result = svr_job_purge(test_job);
fail_unless(result != 0, "NULL job input fail");
test_job = job_alloc();
svr_job_purge(test_job);
fail_unless(result >= -1, "empty job input fail: %d", result);/*TODO: fix -1 via log_job_record mock*/
}
END_TEST
START_TEST(copy_job_test)
{
struct job* result = copy_job(NULL);
struct job *parent = job_alloc();
struct job *child = NULL;
fail_unless(result == NULL, "NULL input check fail");
child = copy_job(parent);
fail_unless(child != NULL, "job was not copied");
/* TODO: add check for correctness of the copy */
}
END_TEST
START_TEST(job_free_test)
{
struct job *result = NULL;
/* NULL value test */
job_free(result,0);
result = job_alloc();
job_free(result,0);
/* TODO: add some check for the free operation
or refactor function interface, suggestions:
1) pointer to pointer pass, then null the actual pointer
2) return value, could be a code or the actual ptr from 1*/
}
END_TEST
START_TEST(get_jobs_array_test)
{
struct job *test_job = NULL;
struct job_array *result = get_jobs_array(NULL);
fail_unless(result == NULL, "NULL input pointer to pointer to job fail");
result = get_jobs_array(&test_job);
fail_unless(result == NULL, "NULL input pointer to job fail");
test_job = job_alloc();
result = get_jobs_array(&test_job);
fail_unless(result == NULL, "get job array fail");
}
END_TEST
START_TEST(get_jobs_queue_test)
{
struct job *test_job = NULL;
struct pbs_queue *result = get_jobs_queue(NULL);
fail_unless(result == NULL, "NULL input pointer to pointer to job fail");
result = get_jobs_queue(&test_job);
fail_unless(result == NULL, "NULL input pointer to job fail");
test_job = job_alloc();
result = get_jobs_queue(&test_job);
fail_unless(result == NULL, "get job queue fail");
}
END_TEST
START_TEST(delete_dependency_job_test)
{
job *pjob = job_alloc();
batch_request preq;
memset(&preq, 0, sizeof(preq));
strcpy(preq.rq_ind.rq_register.rq_parent, job1);
strcpy(preq.rq_ind.rq_register.rq_child, job1);
fail_unless(delete_dependency_job(&preq, &pjob) == PBSE_IVALREQ);
strcpy(preq.rq_ind.rq_register.rq_child, job2);
fail_unless(delete_dependency_job(&preq, &pjob) == PBSE_NONE);
fail_unless(pjob == NULL);
}
END_TEST
START_TEST(remove_job_test)
{
all_jobs alljobs;
int result;
struct job *test_job = job_alloc();
result = remove_job(NULL,test_job);
fail_unless(result != PBSE_NONE, "remove from null array fail");
result = remove_job(&alljobs,NULL);
fail_unless(result != PBSE_NONE, "NULL job remove fail");
insert_job(&alljobs,test_job);
result = remove_job(&alljobs,test_job);
fail_unless(result == PBSE_NONE, "job remove fail");
}
END_TEST
START_TEST(insert_job_first_test)
{
all_jobs alljobs;
struct job *test_job = job_alloc();
int result;
result = insert_job_first(NULL,test_job);
fail_unless(result != PBSE_NONE, "insert into null array fail");
result = insert_job_first(&alljobs,NULL);
fail_unless(result != PBSE_NONE, "NULL job insert fail");
result = insert_job_first(&alljobs,test_job);
fail_unless(result == PBSE_NONE, "job insert fail");
}
END_TEST
START_TEST(has_job_test)
{
all_jobs alljobs;
struct job *test_job = job_alloc();
int result;
result = has_job(NULL,test_job);
fail_unless(result != PBSE_NONE, "null input array fail");
result = has_job(&alljobs,NULL);
fail_unless(result != PBSE_NONE, "NULL input job fail");
insert_job(&alljobs, test_job);
result = has_job(&alljobs, test_job);
fail_unless(result == TRUE, "has_job fail");
}
END_TEST
START_TEST(get_jobs_index_test)
{
struct all_jobs alljobs;
struct job *test_job = job_alloc();
int result;
initialize_all_jobs_array(&alljobs);
result = get_jobs_index(NULL,test_job);
fail_unless(result == -1 * PBSE_BAD_PARAMETER, "null input array fail");
result = get_jobs_index(&alljobs,NULL);
fail_unless(result == -1 * PBSE_BAD_PARAMETER, "NULL input job fail");
insert_job(&alljobs, test_job);
result = get_jobs_index(&alljobs, test_job);
fail_unless(result == PBSE_NONE, "get_jobs_index fail");
}
END_TEST
START_TEST(cpy_checkpoint_test)
{
struct job *test_job = job_alloc();
struct batch_request *result = cpy_checkpoint(NULL,
test_job,
JOB_ATR_checkpoint_name,
CKPT_DIR_IN);
struct batch_request *initial = alloc_br(/*PBS_BATCH_CheckpointJob*/0);
fail_unless(result == NULL, "NULL batch_request input fail");
result = cpy_checkpoint(initial,
NULL,
JOB_ATR_checkpoint_name,
CKPT_DIR_IN);
fail_unless(result == NULL, "NULL job input fail");
/*TODO: add test for valid input, invalid dir value*/
}
END_TEST
START_TEST(svr_job_purge_test)
{
struct job* test_job = NULL;
int result = svr_job_purge(test_job);
fail_unless(result != 0, "NULL job input fail");
called_remove_job = 0;
dequejob_rc = PBSE_JOB_NOT_IN_QUEUE;
test_job = job_alloc();
test_job->ji_qs.ji_substate = JOB_SUBSTATE_QUEUED;
test_job->ji_qs.ji_state = JOB_STATE_QUEUED;
result = svr_job_purge(test_job);
fail_unless(result == 0, "non-queued job fail", result);
// called_remove_job once means we didn't call job_free
fail_unless(called_remove_job == 1);
dequejob_rc = 0;
result = svr_job_purge(test_job);
fail_unless(result == 0, "queued job fail: %d", result);
// Calling remove_job twice means we did call job_free
fail_unless(called_remove_job == 3);
}
void mom_req_quejob(
batch_request *preq) /* ptr to the decoded request */
{
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 */
int resc_access_perm = ATR_DFLAG_USWR | ATR_DFLAG_Creat;
memset(basename, 0, sizeof(basename));
if (PBSNodeCheckProlog)
{
check_state(1);
if (internal_state & INUSE_DOWN)
{
req_reject(PBSE_BADMOMSTATE, 0, preq, NULL, NULL);
return;
}
}
if (reject_job_submit == TRUE)
{
req_reject(-1, 0, preq, NULL, "This mom is configured not to run jobs");
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, __func__, (char *)"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 = mom_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 pbs_attribute.
*/
psatl = (svrattrl *)GET_NEXT(preq->rq_ind.rq_queuejob.rq_attr);
while (psatl != NULL)
{
if (!strcmp(psatl->al_name,ATTR_hashname))
{
snprintf(basename, sizeof(basename), "%s", psatl->al_value);
break;
}
psatl = (svrattrl *)GET_NEXT(psatl->al_link);
}
if (basename[0] == '\0')
snprintf(basename, sizeof(basename), "%s", jid);
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, __func__, (char *)"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
{
/* reject the job. It is already working here. */
sprintf(log_buffer, "Job already exists. State: %d substate: %d", pj->ji_qs.ji_state, pj->ji_qs.ji_substate);
log_err(-1, __func__, log_buffer);
sprintf(log_buffer, "Job %s already on mom", pj->ji_qs.ji_jobid);
req_reject(PBSE_JOBEXIST, 0, preq, NULL, log_buffer);
return;
}
} /* END if (pj != NULL) */
else
{
/* if not already here, allocate job struct */
if ((pj = job_alloc()) == NULL)
{
/* FAILURE */
req_reject(PBSE_MEM_MALLOC, 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;
/* changing the union type overwrites the euid for the job, and if
* ji_grpcache is set this potentially allows jobs to run as root. Unsetting
* ji_grpcache fixes this problem --dbeer */
if (pj->ji_grpcache != NULL)
{
free(pj->ji_grpcache);
pj->ji_grpcache = NULL;
}
}
/* 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 pbs_attribute by name */
index = find_attr(job_attr_def,psatl->al_name,JOB_ATR_LAST);
if (index < 0)
{
/* FAILURE */
/* didn`t recognize the name */
mom_job_purge(pj); /* CRI - 12/20/2004 */
reply_badattr(PBSE_NOATTR, 1, psatl, preq);
return;
}
pdef = &job_attr_def[index];
/* Is pbs_attribute not writeable by manager or by a server? */
if ((pdef->at_flags & resc_access_perm) == 0)
{
/* FAILURE */
mom_job_purge(pj);
reply_badattr(PBSE_ATTRRO, 1, psatl, preq);
return;
}
/* decode pbs_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,
resc_access_perm);
}
if (rc != 0)
{
/* FAILURE */
/* all errors are fatal for MOM */
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)
{
mom_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)
{
remove_from_job_list(pj);
append_link(&svr_newjobs,&pj->ji_alljobs,pj);
if (pj->ji_grpcache != NULL)
{
free(pj->ji_grpcache);
pj->ji_grpcache = NULL;
}
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,FALSE);
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, FALSE);
}
/* 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[JOB_ATR_mtime].at_val.at_long = (long)time_now;
pj->ji_wattr[JOB_ATR_mtime].at_flags |= ATR_VFLAG_SET;
if (pj->ji_grpcache != NULL)
{
free(pj->ji_grpcache);
pj->ji_grpcache = NULL;
}
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,FALSE);
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 */
// call mom_job_purge first so that double-frees don't happen
// when the on_close function is called
mom_job_purge(pj);
close_conn(sock, FALSE);
return;
}
/* link job into server's new jobs list request */
append_link(&svr_newjobs, &pj->ji_alljobs, pj);
return;
} /* END mom_req_quejob() */
void*
job_or_resv_recov_fs(char *filename, int objtype)
{
int fds;
job *pj;
void *pobj = NULL;
#ifndef PBS_MOM
resc_resv *presv;
#endif
void *p_fixed = NULL;
int fixed_size;
char *prefix = NULL;
char *path = NULL;
char *err_msg;
char *ptcs; /*text control string for err msg*/
char *pobjID = NULL;
char *pn; /*name of the file "root" (prefix)*/
attribute *wattr = NULL;
attribute_def *p_attr_def = NULL;
int final_attr;
int attr_unkn;
char namebuf[MAXPATHLEN];
char err_buf[80];
if (objtype == RESC_RESV_OBJECT) {
#ifndef PBS_MOM /*MOM doesn't know about resource reservations*/
presv = resc_resv_alloc(); /* allocate & init resc_rescv struct */
if (presv == (resc_resv *)0) {
return ((void *)0);
}
pobj = (void *)presv;
path = path_resvs;
err_msg = "error opening reservation file";
ptcs = "reservation Id %s does not match file name for %s";
pobjID = presv->ri_qs.ri_resvID;
p_fixed = (void *)&presv->ri_qs;
fixed_size = sizeof(struct resvfix);
prefix = presv->ri_qs.ri_fileprefix;
p_attr_def = resv_attr_def;
wattr = presv->ri_wattr;
attr_unkn = RESV_ATR_UNKN;
final_attr = RESV_ATR_LAST;
#else /* PBS_MOM only: This will never come here for MOM!!! */
return ((void *)0);
#endif
} else {
pj = job_alloc(); /* allocate & initialize job struct */
if (pj == (job *)0) {
return ((void *)0);
}
pobj = (void *)pj;
path = path_jobs;
err_msg = "error opening job file";
ptcs = "Job Id %s does not match file name for %s";
pobjID = pj->ji_qs.ji_jobid;
p_fixed = (void *)&pj->ji_qs;
fixed_size = sizeof(struct jobfix);
if (*pj->ji_qs.ji_fileprefix != '\0')
prefix = pj->ji_qs.ji_fileprefix;
else
prefix = pj->ji_qs.ji_jobid;
p_attr_def = job_attr_def;
wattr = pj->ji_wattr;
attr_unkn = JOB_ATR_UNKN;
final_attr = JOB_ATR_LAST;
}
(void)strcpy(namebuf, path); /* job (reservation) directory path */
(void)strcat(namebuf, filename);
#ifdef WIN32
fix_perms(namebuf);
#endif
fds = open(namebuf, O_RDONLY, 0);
if (fds < 0) {
sprintf(log_buffer, "%s on %s", err_msg, namebuf);
log_err(errno, "job_or_resv_recov", log_buffer);
free((char *)pobj);
return ((void *)0);
}
#ifdef WIN32
setmode(fds, O_BINARY);
#endif
/* read in job or resc_resv quick save sub-structure */
if (read(fds, (char *)p_fixed, fixed_size) != fixed_size) {
(void)sprintf(err_buf, "problem reading %s", namebuf);
log_err(errno, "job_or_resv_recov", err_buf);
free((char *)pobj);
(void)close(fds);
return ((void *)0);
}
/* Does file name match the internal name? */
/* This detects ghost files */
#ifdef WIN32
pn = strrchr(namebuf, (int)'/');
if (pn == NULL)
pn = strrchr(namebuf, (int)'\\');
if (pn == NULL) {
sprintf(log_buffer, "bad path %s", namebuf);
log_err(errno, "job_or_resv_recov", log_buffer);
free((char *)pj);
(void)close(fds);
return ((job *)0);
}
pn++;
#else
pn = strrchr(namebuf, (int)'/') + 1;
#endif
if (strncmp(pn, prefix, strlen(prefix)) != 0) {
/* mismatch, discard job (reservation) */
(void)sprintf(log_buffer, ptcs, pobjID, namebuf);
log_err(-1, "job_or_resv_recov", log_buffer);
free((char *)pobj);
(void)close(fds);
return ((void *)0);
}
/* read in working attributes */
if (recov_attr_fs(fds, pobj, p_attr_def, wattr,
final_attr, attr_unkn) != 0) {
log_err(errno, "job_or_resv_recov", "error from recov_attr");
if (objtype == RESC_RESV_OBJECT) {
#ifndef PBS_MOM /*MOM doesn't know about resource reservations*/
resv_free((resc_resv *)pobj);
#endif
} else {
job_free((job *)pobj);
}
(void)close(fds);
return ((void *)0);
}
(void)close(fds);
#if defined(PBS_MOM) && defined(WIN32)
/* get a handle to the job (may not exist) */
pj->ji_hJob = OpenJobObject(JOB_OBJECT_ALL_ACCESS, FALSE,
pj->ji_qs.ji_jobid);
#endif
/* all done recovering the job (reservation) */
return (pobj);
}
job *
job_recov_fs(char *filename, int recov_subjob)
{
int fds;
char basen[MAXPATHLEN+1];
job *pj;
char *pn;
char *psuffix;
pj = job_alloc(); /* allocate & initialize job structure space */
if (pj == (job *)0) {
return ((job *)0);
}
(void)strcpy(pbs_recov_filename, path_jobs); /* job directory path */
(void)strcat(pbs_recov_filename, filename);
#ifdef WIN32
fix_perms(pbs_recov_filename);
#endif
/* change file name in case recovery fails so we don't try same file */
(void)strcpy(basen, pbs_recov_filename);
psuffix = basen + strlen(basen) - strlen(JOB_BAD_SUFFIX);
(void)strcpy(psuffix, JOB_BAD_SUFFIX);
#ifdef WIN32
if (MoveFileEx(pbs_recov_filename, basen,
MOVEFILE_REPLACE_EXISTING | MOVEFILE_WRITE_THROUGH) == 0) {
errno = GetLastError();
sprintf(log_buffer, "MoveFileEx(%s, %s) failed!",
pbs_recov_filename, basen);
log_err(errno, "nodes", log_buffer);
}
secure_file(basen, "Administrators",
READS_MASK|WRITES_MASK|STANDARD_RIGHTS_REQUIRED);
#else
if (rename(pbs_recov_filename, basen) == -1) {
sprintf(log_buffer, "error renaming job file %s",
pbs_recov_filename);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
return ((job *)0);
}
#endif
fds = open(basen, O_RDONLY, 0);
if (fds < 0) {
sprintf(log_buffer, "error opening of job file %s",
pbs_recov_filename);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
return ((job *)0);
}
#ifdef WIN32
setmode(fds, O_BINARY);
#endif
/* read in job fixed sub-structure */
errno = -1;
if (read(fds, (char *)&pj->ji_qs, fixedsize) != (int)fixedsize) {
sprintf(log_buffer, "error reading fixed portion of %s",
pbs_recov_filename);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
(void)close(fds);
return ((job *)0);
}
/* Does file name match the internal name? */
/* This detects ghost files */
#ifdef WIN32
pn = strrchr(pbs_recov_filename, (int)'/');
if (pn == NULL)
pn = strrchr(pbs_recov_filename, (int)'\\');
if (pn == NULL) {
sprintf(log_buffer, "bad path %s", pbs_recov_filename);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
(void)close(fds);
return ((job *)0);
}
pn++;
#else
pn = strrchr(pbs_recov_filename, (int)'/') + 1;
#endif
if (strncmp(pn, pj->ji_qs.ji_jobid, strlen(pn)-3) != 0) {
/* mismatch, discard job */
(void)sprintf(log_buffer,
"Job Id %s does not match file name for %s",
pj->ji_qs.ji_jobid,
pbs_recov_filename);
log_err(-1, "job_recov", log_buffer);
free((char *)pj);
(void)close(fds);
return ((job *)0);
}
/* unless directed, don't recover Array Sub jobs */
if ((pj->ji_qs.ji_svrflags & JOB_SVFLG_SubJob) &&
(recov_subjob == NO_RECOV_SUBJOB)) {
free((char *)pj);
(void)close(fds);
return ((job *)0);
}
/* read in extended save area depending on VERSION */
errno = -1;
DBPRT(("Job save version %d\n", pj->ji_qs.ji_jsversion))
if (pj->ji_qs.ji_jsversion < JSVERSION_514) {
/* If really old version, it wasn't there, abort out */
sprintf(log_buffer,
"Job structure version cannot be recovered for job %s",
pbs_recov_filename);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
(void)close(fds);
return ((job *)0);
} else if (pj->ji_qs.ji_jsversion < JSVERSION_80) {
/* If older version, read and copy extended area */
if (recov_514_extend(fds, pj) != 0) {
sprintf(log_buffer,
"error reading extended portion"
" of %s for prior version",
pbs_recov_filename);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
(void)close(fds);
return ((job *)0);
}
} else {
/* If current version, JSVERSION_80, read into place */
if (read(fds, (char *)&pj->ji_extended,
sizeof(union jobextend)) !=
sizeof(union jobextend)) {
sprintf(log_buffer,
"error reading extended portion of %s",
pbs_recov_filename);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
(void)close(fds);
return ((job *)0);
}
}
#ifndef PBS_MOM
if (pj->ji_qs.ji_svrflags & JOB_SVFLG_ArrayJob) {
size_t xs;
if (read(fds, (char *)&xs, sizeof(xs)) != sizeof(xs)) {
sprintf(log_buffer,
"error reading array section of %s",
pbs_recov_filename);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
(void)close(fds);
return ((job *)0);
}
if ((pj->ji_ajtrk = (struct ajtrkhd *)malloc(xs)) == NULL) {
free((char *)pj);
(void)close(fds);
return ((job *)0);
}
read(fds, (char *)pj->ji_ajtrk + sizeof(xs), xs - sizeof(xs));
pj->ji_ajtrk->tkm_size = xs;
}
#endif /* not PBS_MOM */
/* read in working attributes */
if (recov_attr_fs(fds, pj, job_attr_def, pj->ji_wattr, (int)JOB_ATR_LAST,
(int)JOB_ATR_UNKN) != 0) {
sprintf(log_buffer, "error reading attributes portion of %s",
pbs_recov_filename);
log_err(errno, "job_recov", log_buffer);
job_free(pj);
(void)close(fds);
return ((job *)0);
}
(void)close(fds);
#if defined(PBS_MOM) && defined(WIN32)
/* get a handle to the job (may not exist) */
pj->ji_hJob = OpenJobObject(JOB_OBJECT_ALL_ACCESS, FALSE,
pj->ji_qs.ji_jobid);
#endif
/* all done recovering the job, change file name back to .JB */
#ifdef WIN32
if (MoveFileEx(basen, pbs_recov_filename,
MOVEFILE_REPLACE_EXISTING | MOVEFILE_WRITE_THROUGH) == 0) {
errno = GetLastError();
sprintf(log_buffer, "MoveFileEx(%s, %s) failed!",
basen, pbs_recov_filename);
log_err(errno, "nodes", log_buffer);
}
secure_file(pbs_recov_filename, "Administrators",
READS_MASK|WRITES_MASK|STANDARD_RIGHTS_REQUIRED);
#else
(void)rename(basen, pbs_recov_filename);
#endif
return (pj);
}
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() */
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() */
/**
* @brief
* create_subjob - create a Subjob from the parent Array Job
* Certain attributes are changed or left out
* @param[in] parent - pointer to parent Job
* @param[in] newjid - new job id
* @param[in] rc - return code
* @return pointer to new job
* @retval NULL - error
*/
job *
create_subjob(job *parent, char *newjid, int *rc)
{
pbs_list_head attrl;
int i;
int j;
int indx;
char *index;
attribute_def *pdef;
attribute *ppar;
attribute *psub;
svrattrl *psatl;
job *subj;
long eligibletime;
long time_msec;
#ifdef WIN32
struct _timeb tval;
#else
struct timeval tval;
#endif
if ((parent->ji_qs.ji_svrflags & JOB_SVFLG_ArrayJob) == 0) {
*rc = PBSE_IVALREQ;
return NULL; /* parent not an array job */
}
/* find and copy the index */
if ((index = get_index_from_jid(newjid)) == NULL) {
*rc = PBSE_IVALREQ;
return NULL;
}
if ((indx = subjob_index_to_offset(parent, index)) == -1) {
*rc = PBSE_UNKJOBID;
return NULL;
}
if (parent->ji_ajtrk->tkm_tbl[indx].trk_status != JOB_STATE_QUEUED) {
*rc = PBSE_BADSTATE;
return NULL;
}
/*
* allocate and clear basic structure
* cannot copy job attributes because cannot share strings and other
* malloc-ed data, so copy ji_qs as a whole and then copy the
* non-saved items before ji_qs.
*/
subj = job_alloc();
subj->ji_qs = parent->ji_qs; /* copy the fixed save area */
#ifdef PBS_CRED_GRIDPROXY
subj->ji_gsscontext = parent->ji_gsscontext;
#endif
subj->ji_qhdr = parent->ji_qhdr;
subj->ji_resvp = parent->ji_resvp;
subj->ji_myResv = parent->ji_myResv;
subj->ji_parentaj = parent;
strcpy(subj->ji_qs.ji_jobid, newjid); /* replace job id */
*subj->ji_qs.ji_fileprefix = '\0';
subj->ji_subjindx = indx;
/*
* now that is all done, copy the required attributes by
* encoding and then decoding into the new array. Then add the
* subjob specific attributes.
*/
resc_access_perm = ATR_DFLAG_ACCESS;
CLEAR_HEAD(attrl);
for (i = 0; attrs_to_copy[i] != JOB_ATR_LAST; i++) {
j = (int)attrs_to_copy[i];
ppar = &parent->ji_wattr[j];
psub = &subj->ji_wattr[j];
pdef = &job_attr_def[j];
if (pdef->at_encode(ppar, &attrl, pdef->at_name, NULL,
ATR_ENCODE_MOM, &psatl) > 0) {
for (psatl = (svrattrl *)GET_NEXT(attrl); psatl;
psatl = ((svrattrl *)GET_NEXT(psatl->al_link))) {
pdef->at_decode(psub, psatl->al_name, psatl->al_resc,
psatl->al_value);
}
/* carry forward the default bit if set */
psub->at_flags |= (ppar->at_flags & ATR_VFLAG_DEFLT);
free_attrlist(&attrl);
}
}
psub = &subj->ji_wattr[(int)JOB_ATR_array_id];
job_attr_def[(int)JOB_ATR_array_id].at_decode(psub, NULL, NULL,
parent->ji_qs.ji_jobid);
psub = &subj->ji_wattr[(int)JOB_ATR_array_index];
job_attr_def[(int)JOB_ATR_array_index].at_decode(psub, NULL, NULL, index);
/* Lastly, set or clear a few flags and link in the structure */
subj->ji_qs.ji_svrflags &= ~JOB_SVFLG_ArrayJob;
subj->ji_qs.ji_svrflags |= JOB_SVFLG_SubJob;
subj->ji_modified = 1; /* ** will likely take this out ** */
subj->ji_qs.ji_substate = JOB_SUBSTATE_TRANSICM;
(void)svr_setjobstate(subj, JOB_STATE_QUEUED, JOB_SUBSTATE_QUEUED);
subj->ji_wattr[(int)JOB_ATR_state].at_flags |= ATR_VFLAG_SET;
subj->ji_wattr[(int)JOB_ATR_substate].at_flags |= ATR_VFLAG_SET;
/* subjob needs to borrow eligible time from parent job array.
* expecting only to accrue eligible_time and nothing else.
*/
if (server.sv_attr[(int)SRV_ATR_EligibleTimeEnable].at_val.at_long == 1) {
eligibletime = parent->ji_wattr[(int)JOB_ATR_eligible_time].at_val.at_long;
if (parent->ji_wattr[(int)JOB_ATR_accrue_type].at_val.at_long == JOB_ELIGIBLE)
eligibletime += subj->ji_wattr[(int)JOB_ATR_sample_starttime].at_val.at_long - parent->ji_wattr[(int)JOB_ATR_sample_starttime].at_val.at_long;
subj->ji_wattr[(int)JOB_ATR_eligible_time].at_val.at_long = eligibletime;
subj->ji_wattr[(int)JOB_ATR_eligible_time].at_flags |= ATR_VFLAG_MODIFY | ATR_VFLAG_MODCACHE;
}
#ifdef WIN32
_ftime_s(&tval);
time_msec = (tval.time * 1000L) + tval.millitm;
#else
gettimeofday(&tval, NULL);
time_msec = (tval.tv_sec * 1000L) + (tval.tv_usec/1000L);
#endif
/* set the queue rank attribute */
subj->ji_wattr[(int)JOB_ATR_qrank].at_val.at_long = time_msec;
subj->ji_wattr[(int)JOB_ATR_qrank].at_flags |= ATR_VFLAG_SET|ATR_VFLAG_MODCACHE;
if (svr_enquejob(subj) != 0) {
job_purge(subj);
*rc = PBSE_IVALREQ;
return NULL;
}
*rc = PBSE_NONE;
return subj;
}
