bool
shepherd_read_exit_status_file(int *return_code)
{
bool ret = true;
FILE *fp = NULL;
const char *const filename = "exit_status";
fp = fopen(filename, "r");
if (fp != NULL) {
int arguments = fscanf(fp, "%d\n", return_code);
/* retrieve first exit status from exit status file */
if (arguments != 1) {
shepherd_trace("could not read exit_status file");
*return_code = ESSTATE_NO_EXITSTATUS;
ret = false;
}
} else {
shepherd_error(1, MSG_FILE_NOOPEN_SS, filename, strerror(errno));
ret = false;
}
FCLOSE(fp);
return ret;
FCLOSE_ERROR:
shepherd_error(1, MSG_FILE_NOCLOSE_SS, filename, strerror(errno));
return false;
}
bool
shepherd_read_processor_set_number_file(int *proc_set)
{
bool ret = true;
FILE *fp = NULL;
const char *const filename = "processor_set_number";
fp = fopen(filename, "r");
if (fp != NULL) {
int arguments = fscanf(fp, "%d", proc_set);
if (arguments != 1) {
shepherd_trace("could not read processor_set_number file");
*proc_set = 0;
ret = false;
}
} else {
shepherd_error(1, MSG_FILE_NOOPEN_SS, filename, strerror(errno));
ret = false;
}
FCLOSE(fp);
return ret;
FCLOSE_ERROR:
shepherd_error(1, MSG_FILE_NOCLOSE_SS, filename, strerror(errno));
return false;
}
void sge_pset_create_processor_set(void)
{
#if defined(__sgi) || defined(ALPHA) || defined(SOLARIS64) || defined(SOLARISAMD64)
char err_str[2*SGE_PATH_MAX+128];
/* SGI IRIX processor set stuff */
if (strcasecmp("UNDEFINED",get_conf_val("processors"))) {
int ret;
errno = 0;
if (sge_switch2start_user()) {
shepherd_trace("can't switch user in sge_pset_create_processor_set");
shepherd_state = SSTATE_PROCSET_NOTSET;
shepherd_error(1, strerror(errno));
return;
}
if ((ret=set_processor_range(get_conf_val("processors"),
(int) strtol(get_conf_val("job_id"), NULL, 10),
err_str)) != PROC_SET_OK) {
sge_switch2admin_user();
if (ret == PROC_SET_WARNING) /* not critical - e.g. not root */
shepherd_trace("warning: processor set not set in set_processor_range");
else { /* critical --> use err_str to indicate error */
shepherd_trace("critical error in set_processor_range - bailing out");
shepherd_state = SSTATE_PROCSET_NOTSET;
shepherd_error(1, err_str);
}
} else {
sge_switch2admin_user();
}
}
#endif
}
void sge_pset_free_processor_set(void)
{
#if defined(__sgi) || defined(ALPHA) || defined(SOLARIS64) || defined(SOLARISAMD64)
/* SGI IRIX processor set stuff */
if (strcasecmp("UNDEFINED",get_conf_val("processors"))) {
char err_str[2*SGE_PATH_MAX+128];
int ret;
errno = 0;
if (sge_switch2start_user()) {
shepherd_trace("failed to switch user in free_processor_set: %s",
strerror(errno));
shepherd_state = SSTATE_PROCSET_NOTFREED;
shepherd_error(1, strerror(errno));
return;
}
if ((ret=free_processor_set(err_str)) != PROC_SET_OK) {
sge_switch2admin_user();
switch (ret) {
case PROC_SET_WARNING: /* not critical - e.g. not root */
shepherd_trace("warning: processor set not freed in free_processor_set - "
"did no exist, probably");
break;
case PROC_SET_ERROR: /* critical - err_str indicates error */
shepherd_trace("critical error in free_processor_set - bailing out");
shepherd_state = SSTATE_PROCSET_NOTFREED;
shepherd_error(1, err_str);
break;
case PROC_SET_BUSY: /* still processes running in processor set */
shepherd_trace("error in releasing processor set");
shepherd_state = SSTATE_PROCSET_NOTFREED;
shepherd_error(1, err_str);
break;
default: /* should not occur */
sprintf(err_str,
"internal error after free_processor_set - ret=%d", ret);
shepherd_state = SSTATE_PROCSET_NOTFREED;
shepherd_error(1, err_str);
break;
}
} else {
sge_switch2admin_user();
}
}
#endif
}
bool
shepherd_read_osjobid_file(
#if (IRIX)
ash_t *return_code,
#elif defined(NECSX4) || defined(NECSX5)
id_t *return_code,
#elif defined(CRAY)
int *return_code,
#endif
bool is_error
)
{
bool ret = true;
FILE *fp = NULL;
const char *const filename = "osjobid";
fp = fopen(filename, "r");
if (fp != NULL) {
int arguments = 0;
#if defined(IRIX)
arguments = fscanf(fp, "%lld\n", return_code);
#else
arguments = fscanf(fp, "%d\n", return_code);
#endif
if (arguments != 1) {
shepherd_trace("could not read osjobid file");
*return_code = 0;
ret = false;
}
FCLOSE(fp);
} else {
if (is_error == true) {
shepherd_error(1, MSG_FILE_NOOPEN_SS, filename, strerror(errno));
} else {
shepherd_trace(MSG_FILE_NOOPEN_SS, filename, strerror(errno));
}
ret = false;
}
return ret;
FCLOSE_ERROR:
shepherd_error(1, MSG_FILE_NOCLOSE_SS, filename, strerror(errno));
return false;
}
bool shepherd_write_osjobid_file(const char *osjobid)
{
bool ret = true;
const char *const filename = "osjobid";
FILE *fp = NULL;
fp = fopen(filename, "w");
if (fp != NULL) {
FPRINTF((fp, "%s\n", osjobid));
FCLOSE(fp);
} else {
shepherd_error(1, MSG_FILE_NOOPEN_SS, filename, strerror(errno));
ret = false;
}
return ret;
FPRINTF_ERROR:
FCLOSE_ERROR:
shepherd_error(1, MSG_FILE_NOCLOSE_SS, filename, strerror(errno));
return false;
}
void
create_checkpointed_file(int ckpt_is_in_arena)
{
const char *const filename = "checkpointed";
FILE *fp = NULL;
fp = fopen(filename, "w");
if (fp != NULL) {
if (ckpt_is_in_arena) {
FPRINTF((fp, "1\n"));
}
FCLOSE(fp);
} else {
shepherd_error(1, MSG_FILE_NOOPEN_SS, filename, strerror(errno));
}
return;
FPRINTF_ERROR:
FCLOSE_ERROR:
shepherd_error(1, MSG_FILE_NOCLOSE_SS, filename, strerror(errno));
return;
}
bool
shepherd_write_processor_set_number_file(int proc_set)
{
bool ret = true;
const char *const filename = "processor_set_number";
FILE *fp = NULL;
fp = fopen(filename, "w");
if (fp != NULL) {
FPRINTF((fp, "%d\n", proc_set));
FCLOSE(fp);
} else {
shepherd_error(1, MSG_FILE_NOOPEN_SS, filename, strerror(errno));
ret = false;
}
return ret;
FPRINTF_ERROR:
FCLOSE_ERROR:
shepherd_error(1, MSG_FILE_NOCLOSE_SS, filename, strerror(errno));
return false;
}
bool
shepherd_write_sig_info_file(const char *filename, const char *task_id,
u_long32 exit_status)
{
bool ret = true;
FILE *fp = NULL;
fp = fopen(filename, "a");
if (fp != NULL) {
FPRINTF((fp, "%s "sge_u32"\n", task_id, exit_status));
FCLOSE(fp);
} else {
shepherd_error(1, MSG_FILE_NOOPEN_SS, filename, strerror(errno));
ret = false;
}
return ret;
FPRINTF_ERROR:
FCLOSE_ERROR:
shepherd_error(1, MSG_FILE_NOCLOSE_SS, filename, strerror(errno));
return false;
}
bool
shepherd_write_shepherd_about_to_exit_file(void)
{
bool ret = true;
const char *const filename = "shepherd_about_to_exit";
FILE *fp;
fp = fopen(filename, "w");
if (fp != NULL)
{
FCLOSE(fp);
}
else
{
shepherd_error(1, MSG_FILE_NOOPEN_SS, filename, strerror(errno));
ret = false;
}
return ret;
FCLOSE_ERROR:
shepherd_error(1, MSG_FILE_NOCLOSE_SS, filename, strerror(errno));
return false;
}
bool
shepherd_write_usage_file(u_long32 wait_status, int exit_status,
int child_signal, u_long32 start_time,
u_long32 end_time, struct rusage *rusage)
{
bool ret = true;
const char *const filename = "usage";
FILE *fp = NULL;
shepherd_trace("writing usage file to \"usage\"");
fp = fopen(filename, "w");
if (fp != NULL) {
/*
* the wait status is returned by japi_wait()
* see sge_reportL.h for bitmask and makro definition
*/
FPRINTF((fp, "wait_status="sge_u32"\n", wait_status));
FPRINTF((fp, "exit_status=%d\n", exit_status));
FPRINTF((fp, "signal=%d\n", child_signal));
FPRINTF((fp, "start_time=%d\n", (int) start_time));
FPRINTF((fp, "end_time=%d\n", (int) end_time));
FPRINTF((fp, "ru_wallclock="sge_u32"\n", (u_long32) end_time-start_time));
#if defined(NEC_ACCOUNTING_ENTRIES)
/* Additional accounting information for NEC SX-4 SX-5 */
#if defined(NECSX4) || defined(NECSX5)
#if defined(NECSX4)
FPRINTF((fp, "necsx_necsx4="sge_u32"\n", 1));
#elif defined(NECSX5)
FPRINTF((fp, "necsx_necsx5="sge_u32"\n", 1));
#endif
FPRINTF((fp, "necsx_base_prty="sge_u32"\n", 0));
FPRINTF((fp, "necsx_time_slice="sge_u32"\n", 0));
FPRINTF((fp, "necsx_num_procs="sge_u32"\n", 0));
FPRINTF((fp, "necsx_kcore_min="sge_u32"\n", 0));
FPRINTF((fp, "necsx_mean_size="sge_u32"\n", 0));
FPRINTF((fp, "necsx_maxmem_size="sge_u32"\n", 0));
FPRINTF((fp, "necsx_chars_trnsfd="sge_u32"\n", 0));
FPRINTF((fp, "necsx_blocks_rw="sge_u32"\n", 0));
FPRINTF((fp, "necsx_inst="sge_u32"\n", 0));
FPRINTF((fp, "necsx_vector_inst="sge_u32"\n", 0));
FPRINTF((fp, "necsx_vector_elmt="sge_u32"\n", 0));
FPRINTF((fp, "necsx_vec_exe="sge_u32"\n", 0));
FPRINTF((fp, "necsx_flops="sge_u32"\n", 0));
FPRINTF((fp, "necsx_conc_flops="sge_u32"\n", 0));
FPRINTF((fp, "necsx_fpec="sge_u32"\n", 0));
FPRINTF((fp, "necsx_cmcc="sge_u32"\n", 0));
FPRINTF((fp, "necsx_bccc="sge_u32"\n", 0));
FPRINTF((fp, "necsx_mt_open="sge_u32"\n", 0));
FPRINTF((fp, "necsx_io_blocks="sge_u32"\n", 0));
FPRINTF((fp, "necsx_multi_single="sge_u32"\n", 0));
FPRINTF((fp, "necsx_max_nproc="sge_u32"\n", 0));
#endif
#endif
FPRINTF((fp, "ru_utime=%f\n", (double)rusage->ru_utime.tv_sec + (double)rusage->ru_utime.tv_usec / 1000000.0));
FPRINTF((fp, "ru_stime=%f\n", (double)rusage->ru_stime.tv_sec + (double)rusage->ru_stime.tv_usec / 1000000.0));
FPRINTF((fp, "ru_maxrss=%ld\n", rusage->ru_maxrss));
FPRINTF((fp, "ru_ixrss=%ld\n", rusage->ru_ixrss));
#if defined(ultrix)
FPRINTF((fp, "ru_ismrss=%ld\n", rusage->ru_ismrss));
#endif
FPRINTF((fp, "ru_idrss=%ld\n", rusage->ru_idrss));
FPRINTF((fp, "ru_isrss=%ld\n", rusage->ru_isrss));
FPRINTF((fp, "ru_minflt=%ld\n", rusage->ru_minflt));
FPRINTF((fp, "ru_majflt=%ld\n", rusage->ru_majflt));
FPRINTF((fp, "ru_nswap=%ld\n", rusage->ru_nswap));
FPRINTF((fp, "ru_inblock=%ld\n", rusage->ru_inblock));
FPRINTF((fp, "ru_oublock=%ld\n", rusage->ru_oublock));
FPRINTF((fp, "ru_msgsnd=%ld\n", rusage->ru_msgsnd));
FPRINTF((fp, "ru_msgrcv=%ld\n", rusage->ru_msgrcv));
FPRINTF((fp, "ru_nsignals=%ld\n", rusage->ru_nsignals));
FPRINTF((fp, "ru_nvcsw=%ld\n", rusage->ru_nvcsw));
FPRINTF((fp, "ru_nivcsw=%ld\n", rusage->ru_nivcsw));
FCLOSE(fp);
} else {
shepherd_error(1, MSG_FILE_NOOPEN_SS, filename, strerror(errno));
ret = false;
}
return ret;
FPRINTF_ERROR:
FCLOSE_ERROR:
shepherd_error(1, MSG_FILE_NOCLOSE_SS, filename, strerror(errno));
return false;
}
void shepherd_error_ptr(const char *text)
{
shepherd_error(1, text);
}
void setosjobid(pid_t sid, gid_t *add_grp_id_ptr, struct passwd *pw)
{
FILE *fp=NULL;
shepherd_trace("setosjobid: uid = "pid_t_fmt", euid = "pid_t_fmt, getuid(), geteuid());
# if defined(SOLARIS) || defined(ALPHA) || defined(LINUX) || defined(FREEBSD) || defined(DARWIN)
/* Read SgeId from config-File and create Addgrpid-File */
{
char *cp;
if ((cp = search_conf_val("add_grp_id")))
*add_grp_id_ptr = atol(cp);
else
*add_grp_id_ptr = 0;
}
if ((fp = fopen("addgrpid", "w")) == NULL) {
shepherd_error(1, "can't open \"addgrpid\" file");
}
fprintf(fp, gid_t_fmt"\n", *add_grp_id_ptr);
FCLOSE(fp);
# elif defined(HP1164) || defined(AIX)
{
if ((fp = fopen("addgrpid", "w")) == NULL) {
shepherd_error(1, "can't open \"addgrpid\" file");
}
fprintf(fp, pid_t_fmt"\n", getpgrp());
FCLOSE(fp);
}
# else
{
char osjobid[100];
if ((fp = fopen("osjobid", "w")) == NULL) {
shepherd_error(1, "can't open \"osjobid\" file");
}
if(sge_switch2start_user() == 0) {
# if defined(IRIX)
{
/* The following block contains the operations necessary for
* IRIX6.2 (and later) to set array session handles (ASHs) and
* service provider info (SPI) records
*/
struct acct_spi spi;
int ret;
char *cp;
shepherd_trace("in irix code");
/* get _local_ array session id */
if ((ret=newarraysess())) {
shepherd_error(1, "error: can't create ASH; errno=%d", ret);
}
/* retrieve array session id we just assigned to the process and
* write it to the os-jobid file
*/
sprintf(osjobid, "%lld", getash());
shepherd_trace(osjobid);
/* set service provider information (spi) record */
strncpy(spi.spi_company, "SGE", 8);
strncpy(spi.spi_initiator, get_conf_val("spi_initiator"), 8);
strncpy(spi.spi_origin, get_conf_val("queue"),16);
strcpy(spi.spi_spi, "Job ");
strncat(spi.spi_spi, get_conf_val("job_id"),11);
if ((ret=setspinfo(&spi))) {
shepherd_error(1, "error: can't set SPI; errno=%d", ret);
}
if ((cp = search_conf_val("acct_project"))) {
prid_t proj;
if (strcasecmp(cp, "none") && ((proj = projid(cp)) >= 0)) {
shepherd_trace("setting project \"%s\" to id %lld", cp, proj);
if (setprid(proj) == -1)
shepherd_trace("failed setting project id");
}
else {
shepherd_trace("can't get id for project \"%s\"", cp);
}
} else {
shepherd_trace("can't get configuration entry for projects");
}
}
# elif defined(CRAY)
{
char *cp;
{
int jobid;
if ((jobid=setjob(pw->pw_uid, 0)) < 0) {
shepherd_error(1, "error: can't set job ID; errno = %d", errno);
}
if (sesscntl(jobid, S_ADDFL, S_BATCH) == -1) {
shepherd_error(1, "error: sesscntl(%d, S_ADDFL, S_BATCH) failed,"
" errno = %d", sid, errno);
}
sprintf(osjobid, "%d", jobid);
}
if ((cp = search_conf_val("acct_project"))) {
int proj;
if (strcasecmp(cp, "none") && ((proj = nam2acid(cp)) >= 0)) {
shephed_trace("setting project \"%s\" to acid %d", cp, proj);
if (acctid(0, proj) == -1) {
shepherd_trace("failed setting project id (acctid)");
}
} else {
shepherd_trace("can't get id for project \"%s\"", cp);
}
} else {
shepherd_trace("can't get configuration entry for projects");
}
}
# elif defined(NECSX4) || defined(NECSX5)
{
id_t jobid = 0;
dispset2_t attr;
int value;
/*
* Create new Super-UX job
*/
if (setjid() == -1) {
shepherd_trace("ERROR: can't set jobid: %s[%d]", strerror(errno), errno);
} else {
jobid = getjid(0);
shepherd_trace("Created job with id: "sge_u32, (u_long32) jobid);
}
sprintf(osjobid, sge_u32, (u_long32) jobid);
/*
* We will use limits for the whole job
*/
set_rlimits_os_job_id(jobid);
/*
* The job will use the resources of the configured
* Resource Sharing Group (rsg)
*/
{
char *rsg_id_string;
int rsg_id;
char fsg_dev_string[256];
rsg_id_string = get_conf_val("processors");
rsg_id = atoi(rsg_id_string);
if (rsg_id) {
int fd;
sprintf(fsg_dev_string, "/dev/rsg/%d", rsg_id);
fd = open(fsg_dev_string, O_RDONLY);
if (fd <= 0) {
shepherd_trace("ERROR: can't switch to rsg%d because can't open"
"device: %s[%d]", rsg_id, strerror(errno), errno);
} else {
if (ioctl(fd, RSG_JUMP, NULL) == -1) {
close(fd);
shepherd_trace("ERROR: can't switch to rsg%d: %s[%d]",
rsg_id, strerror(errno), errno);
return;
} else {
close(fd);
shepherd_trace("switched to rsg%d", rsg_id);
}
}
} else {
shepherd_trace("using default rsg", rsg_id);
}
}
/*
* Set scheduling parameter for job
*/
if (((attr.basepri = atoi(get_conf_val("nec_basepriority"))) != NEC_UNDEF_VALUE)
&& ((attr.modcpu = atoi(get_conf_val("nec_modcpu"))) != NEC_UNDEF_VALUE)
&& ((attr.tickcnt = atoi(get_conf_val("nec_tickcnt"))) != NEC_UNDEF_VALUE)
&& ((attr.dcyfctr = atoi(get_conf_val("nec_dcyfctr"))) != NEC_UNDEF_VALUE)
&& ((attr.dcyintvl = atoi(get_conf_val("nec_dcyintvl"))) != NEC_UNDEF_VALUE)
&& ((attr.tmslice = atoi(get_conf_val("nec_timeslice"))) != NEC_UNDEF_VALUE)
&& ((attr.mempri = atoi(get_conf_val("nec_memorypriority"))) != NEC_UNDEF_VALUE)
&& ((attr.szefctmrt = atoi(get_conf_val("nec_mrt_size_effct"))) != NEC_UNDEF_VALUE)
&& ((attr.priefctmrt = atoi(get_conf_val("nec_mrt_pri_effct"))) != NEC_UNDEF_VALUE)
&& ((attr.minmrt = atoi(get_conf_val("nec_mrt_minimum"))) != NEC_UNDEF_VALUE)
&& ((attr.agrange = atoi(get_conf_val("nec_aging_range"))) != NEC_UNDEF_VALUE)
&& ((attr.spinherit = atoi(get_conf_val("nec_slavepriority"))) != NEC_UNDEF_VALUE)
&& ((attr.concpu = atoi(get_conf_val("nec_cpu_count"))) != NEC_UNDEF_VALUE)) {
if (dispcntl(SG_JID, getjid(0), DCNTL_SET2, &attr) == -1) {
shepherd_trace("ERROR: can't set scheduling parameter: %s[%d]",
strerror(errno), errno);
} else {
shepherd_trace("control parameters for active process scheduling modified");
print_scheduling_parameters(attr);
}
} else {
shepherd_trace("we do not control active process scheduling");
}
}
# else
/* write a default os-jobid to file */
sprintf(osjobid, pid_t_fmt, sid);
# endif
sge_switch2admin_user();
}
else /* not running as super user --> we want a default os-jobid */
sprintf(osjobid, "0");
if(fprintf(fp, "%s\n", osjobid) < 0)
shepherd_trace("error writing osjobid file");
FCLOSE(fp); /* Close os-jobid file */
}
# endif
return;
FCLOSE_ERROR:
shepherd_error(1, "can't close file");
}