static AvahiProbeJob* find_history_job(AvahiProbeScheduler *s, AvahiRecord *record) {
AvahiProbeJob *pj;
assert(s);
assert(record);
for (pj = s->history; pj; pj = pj->jobs_next) {
assert(pj->done);
if (avahi_record_equal_no_ttl(pj->record, record)) {
/* Check whether this entry is outdated */
if (avahi_age(&pj->delivery) > AVAHI_PROBE_HISTORY_MSEC*1000) {
/* it is outdated, so let's remove it */
job_free(s, pj);
return NULL;
}
return pj;
}
}
return NULL;
}
int job_reap( bool interactive )
{
ASSERT_IS_MAIN_THREAD();
job_t *jnext;
int found=0;
static int locked = 0;
locked++;
/*
job_read may fire an event handler, we do not want to call
ourselves recursively (to avoid infinite recursion).
*/
if( locked>1 )
return 0;
job_iterator_t jobs;
jnext = jobs.next();
while (jnext)
{
job_t *j = jnext;
jnext = jobs.next();
process_t *p;
/*
If we are reaping only jobs who do not need status messages
sent to the console, do not consider reaping jobs that need
status messages
*/
if( (!job_get_flag( j, JOB_SKIP_NOTIFICATION ) ) && (!interactive) && (!job_get_flag( j, JOB_FOREGROUND )))
{
continue;
}
for( p=j->first_process; p; p=p->next )
{
int s;
if( !p->completed )
continue;
if( !p->pid )
continue;
s = p->status;
proc_fire_event( L"PROCESS_EXIT", EVENT_EXIT, p->pid, ( WIFSIGNALED(s)?-1:WEXITSTATUS( s )) );
if( WIFSIGNALED(s) )
{
/*
Ignore signal SIGPIPE.We issue it ourselves to the pipe
writer when the pipe reader dies.
*/
if( WTERMSIG(s) != SIGPIPE )
{
int proc_is_job = ((p==j->first_process) && (p->next == 0));
if( proc_is_job )
job_set_flag( j, JOB_NOTIFIED, 1 );
if( !job_get_flag( j, JOB_SKIP_NOTIFICATION ) )
{
if( proc_is_job )
fwprintf( stdout,
_( L"%ls: Job %d, \'%ls\' terminated by signal %ls (%ls)" ),
program_name,
j->job_id,
j->command_wcstr(),
sig2wcs(WTERMSIG(p->status)),
signal_get_desc( WTERMSIG(p->status) ) );
else
fwprintf( stdout,
_( L"%ls: Process %d, \'%ls\' from job %d, \'%ls\' terminated by signal %ls (%ls)" ),
program_name,
p->pid,
p->argv0(),
j->job_id,
j->command_wcstr(),
sig2wcs(WTERMSIG(p->status)),
signal_get_desc( WTERMSIG(p->status) ) );
tputs(clr_eol,1,&writeb);
fwprintf (stdout, L"\n" );
found=1;
}
/*
Clear status so it is not reported more than once
*/
p->status = 0;
}
}
}
/*
If all processes have completed, tell the user the job has
completed and delete it from the active job list.
*/
if( job_is_completed( j ) )
{
if( !job_get_flag( j, JOB_FOREGROUND) && !job_get_flag( j, JOB_NOTIFIED ) && !job_get_flag( j, JOB_SKIP_NOTIFICATION ) )
{
format_job_info( j, _( L"ended" ) );
found=1;
}
proc_fire_event( L"JOB_EXIT", EVENT_EXIT, -j->pgid, 0 );
proc_fire_event( L"JOB_EXIT", EVENT_JOB_ID, j->job_id, 0 );
job_free(j);
}
else if( job_is_stopped( j ) && !job_get_flag( j, JOB_NOTIFIED ) )
{
/*
Notify the user about newly stopped jobs.
*/
if( !job_get_flag( j, JOB_SKIP_NOTIFICATION ) )
{
format_job_info( j, _( L"stopped" ) );
found=1;
}
job_set_flag( j, JOB_NOTIFIED, 1 );
}
}
if( found )
fflush( stdout );
locked = 0;
return found;
}
int main(int argc, char** args) {
char* default_configfn = "astrometry.cfg";
char* default_config_path = "../etc";
int c;
char* configfn = NULL;
int i;
engine_t* engine;
char* mydir = NULL;
char* basedir = NULL;
char* me;
anbool help = FALSE;
sl* strings = sl_new(4);
char* cancelfn = NULL;
char* solvedfn = NULL;
int loglvl = LOG_MSG;
anbool tostderr = FALSE;
char* infn = NULL;
FILE* fin = NULL;
anbool fromstdin = FALSE;
bl* opts = opts_from_array(myopts, sizeof(myopts)/sizeof(an_option_t), NULL);
sl* inds = sl_new(4);
char* datalog = NULL;
engine = engine_new();
while (1) {
c = opts_getopt(opts, argc, args);
if (c == -1)
break;
switch (c) {
case 'D':
datalog = optarg;
break;
case 'p':
engine->inparallel = TRUE;
break;
case 'i':
sl_append(inds, optarg);
break;
case 'd':
basedir = optarg;
break;
case 'f':
infn = optarg;
fromstdin = streq(infn, "-");
break;
case 'E':
tostderr = TRUE;
break;
case 'h':
help = TRUE;
break;
case 'v':
loglvl++;
break;
case 's':
solvedfn = optarg;
case 'C':
cancelfn = optarg;
break;
case 'c':
configfn = strdup(optarg);
break;
case '?':
break;
default:
printf("Unknown flag %c\n", c);
exit( -1);
}
}
if (optind == argc && !infn) {
// Need extra args: filename
printf("You must specify at least one input file!\n\n");
help = TRUE;
}
if (help) {
print_help(args[0], opts);
exit(0);
}
bl_free(opts);
gslutils_use_error_system();
log_init(loglvl);
if (tostderr)
log_to(stderr);
if (datalog) {
datalogfid = fopen(datalog, "wb");
if (!datalogfid) {
SYSERROR("Failed to open data log file \"%s\" for writing", datalog);
return -1;
}
atexit(close_datalogfid);
data_log_init(100);
data_log_enable_all();
data_log_to(datalogfid);
data_log_start();
}
if (infn) {
logverb("Reading input filenames from %s\n", (fromstdin ? "stdin" : infn));
if (!fromstdin) {
fin = fopen(infn, "rb");
if (!fin) {
ERROR("Failed to open file %s for reading input filenames", infn);
exit(-1);
}
} else
fin = stdin;
}
// directory containing the 'engine' executable:
me = find_executable(args[0], NULL);
if (!me)
me = strdup(args[0]);
mydir = sl_append(strings, dirname(me));
free(me);
// Read config file
if (!configfn) {
int i;
sl* trycf = sl_new(4);
sl_appendf(trycf, "%s/%s/%s", mydir, default_config_path, default_configfn);
// if I'm in /usr/bin, look for config file in /etc
if (streq(mydir, "/usr/bin")) {
sl_appendf(trycf, "/etc/%s", default_configfn);
}
sl_appendf(trycf, "%s/%s", mydir, default_configfn);
sl_appendf(trycf, "./%s", default_configfn);
sl_appendf(trycf, "./%s/%s", default_config_path, default_configfn);
for (i=0; i<sl_size(trycf); i++) {
char* cf = sl_get(trycf, i);
if (file_exists(cf)) {
configfn = strdup(cf);
logverb("Using config file \"%s\"\n", cf);
break;
} else {
logverb("Config file \"%s\" doesn't exist.\n", cf);
}
}
if (!configfn) {
char* cflist = sl_join(trycf, "\n ");
logerr("Couldn't find config file: tried:\n %s\n", cflist);
free(cflist);
}
sl_free2(trycf);
}
if (!streq(configfn, "none")) {
if (engine_parse_config_file(engine, configfn)) {
logerr("Failed to parse (or encountered an error while interpreting) config file \"%s\"\n", configfn);
exit( -1);
}
}
if (sl_size(inds)) {
// Expand globs.
for (i=0; i<sl_size(inds); i++) {
char* s = sl_get(inds, i);
glob_t myglob;
int flags = GLOB_TILDE | GLOB_BRACE;
if (glob(s, flags, NULL, &myglob)) {
SYSERROR("Failed to expand wildcards in index-file path \"%s\"", s);
exit(-1);
}
for (c=0; c<myglob.gl_pathc; c++) {
if (engine_add_index(engine, myglob.gl_pathv[c])) {
ERROR("Failed to add index \"%s\"", myglob.gl_pathv[c]);
exit(-1);
}
}
globfree(&myglob);
}
}
if (!pl_size(engine->indexes)) {
logerr("\n\n"
"---------------------------------------------------------------------\n"
"You must list at least one index in the config file (%s)\n\n"
"See http://astrometry.net/use.html about how to get some index files.\n"
"---------------------------------------------------------------------\n"
"\n", configfn);
exit(-1);
}
if (engine->minwidth <= 0.0 || engine->maxwidth <= 0.0) {
logerr("\"minwidth\" and \"maxwidth\" in the config file %s must be positive!\n", configfn);
exit(-1);
}
free(configfn);
if (!il_size(engine->default_depths)) {
parse_depth_string(engine->default_depths,
"10 20 30 40 50 60 70 80 90 100 "
"110 120 130 140 150 160 170 180 190 200");
}
engine->cancelfn = cancelfn;
engine->solvedfn = solvedfn;
i = optind;
while (1) {
char* jobfn;
job_t* job;
struct timeval tv1, tv2;
if (infn) {
// Read name of next input file to be read.
logverb("\nWaiting for next input filename...\n");
jobfn = read_string_terminated(fin, "\n\r\0", 3, FALSE);
if (strlen(jobfn) == 0)
break;
} else {
if (i == argc)
break;
jobfn = args[i];
i++;
}
gettimeofday(&tv1, NULL);
logmsg("Reading file \"%s\"...\n", jobfn);
job = engine_read_job_file(engine, jobfn);
if (!job) {
ERROR("Failed to read job file \"%s\"", jobfn);
exit(-1);
}
if (basedir) {
logverb("Setting job's output base directory to %s\n", basedir);
job_set_output_base_dir(job, basedir);
}
if (engine_run_job(engine, job))
logerr("Failed to run_job()\n");
job_free(job);
gettimeofday(&tv2, NULL);
logverb("Spent %g seconds on this field.\n", millis_between(&tv1, &tv2)/1000.0);
}
engine_free(engine);
sl_free2(strings);
sl_free2(inds);
if (fin && !fromstdin)
fclose(fin);
return 0;
}
void job_purge(
job *pjob) /* M */
{
static char id[] = "job_purge";
job_file_delete_info *jfdi;
jfdi = malloc(sizeof(job_file_delete_info));
if (jfdi == NULL)
{
log_err(ENOMEM,id,"No space to allocate info for job file deletion");
return;
}
/* initialize struct information */
if (pjob->ji_flags & MOM_HAS_TMPDIR)
{
jfdi->has_temp_dir = TRUE;
pjob->ji_flags &= ~MOM_HAS_TMPDIR;
}
else
jfdi->has_temp_dir = FALSE;
if (pjob->ji_flags & MOM_HAS_NODEFILE)
{
jfdi->has_node_file = TRUE;
pjob->ji_flags &= ~MOM_HAS_NODEFILE;
}
else
jfdi->has_node_file = FALSE;
strcpy(jfdi->jobid,pjob->ji_qs.ji_jobid);
strcpy(jfdi->prefix,pjob->ji_qs.ji_fileprefix);
if ((pjob->ji_wattr[JOB_ATR_checkpoint_dir].at_flags & ATR_VFLAG_SET) &&
(pjob->ji_wattr[JOB_ATR_checkpoint_name].at_flags & ATR_VFLAG_SET))
jfdi->checkpoint_dir = strdup(pjob->ji_wattr[JOB_ATR_checkpoint_dir].at_val.at_str);
else
jfdi->checkpoint_dir = NULL;
jfdi->gid = pjob->ji_qs.ji_un.ji_momt.ji_exgid;
jfdi->uid = pjob->ji_qs.ji_un.ji_momt.ji_exuid;
if (thread_unlink_calls == TRUE)
enqueue_threadpool_request(delete_job_files,jfdi);
else
delete_job_files(jfdi);
/* remove this job from the global queue */
delete_link(&pjob->ji_jobque);
delete_link(&pjob->ji_alljobs);
if (LOGLEVEL >= 6)
{
sprintf(log_buffer,"removing job");
log_record(PBSEVENT_DEBUG,PBS_EVENTCLASS_JOB,pjob->ji_qs.ji_jobid,log_buffer);
}
#if IBM_SP2==2 /* IBM SP PSSP 3.1 */
unload_sp_switch(pjob);
#endif /* IBM SP */
job_free(pjob);
/* if no jobs are left, check if MOM should be restarted */
if (((job *)GET_NEXT(svr_alljobs)) == NULL)
MOMCheckRestart();
} /* END job_purge() */
static void elapse_callback(AVAHI_GCC_UNUSED AvahiTimeEvent *e, void* data) {
AvahiProbeJob *pj = data, *next;
AvahiProbeScheduler *s;
AvahiDnsPacket *p;
unsigned n;
assert(pj);
s = pj->scheduler;
if (pj->done) {
/* Lets remove it from the history */
job_free(s, pj);
return;
}
if (!(p = avahi_dns_packet_new_query(s->interface->hardware->mtu)))
return; /* OOM */
n = 1;
/* Add the import probe */
if (!packet_add_probe_query(s, p, pj)) {
size_t size;
AvahiKey *k;
int b;
avahi_dns_packet_free(p);
/* The probe didn't fit in the package, so let's allocate a larger one */
size =
avahi_key_get_estimate_size(pj->record->key) +
avahi_record_get_estimate_size(pj->record) +
AVAHI_DNS_PACKET_HEADER_SIZE;
if (size > AVAHI_DNS_PACKET_SIZE_MAX)
size = AVAHI_DNS_PACKET_SIZE_MAX;
if (!(p = avahi_dns_packet_new_query(size)))
return; /* OOM */
if (!(k = avahi_key_new(pj->record->key->name, pj->record->key->clazz, AVAHI_DNS_TYPE_ANY))) {
avahi_dns_packet_free(p);
return; /* OOM */
}
b = avahi_dns_packet_append_key(p, k, 0) && avahi_dns_packet_append_record(p, pj->record, 0, 0);
avahi_key_unref(k);
if (b) {
avahi_dns_packet_set_field(p, AVAHI_DNS_FIELD_NSCOUNT, 1);
avahi_dns_packet_set_field(p, AVAHI_DNS_FIELD_QDCOUNT, 1);
avahi_interface_send_packet(s->interface, p);
} else
avahi_log_warn("Probe record too large, cannot send");
avahi_dns_packet_free(p);
job_mark_done(s, pj);
return;
}
/* Try to fill up packet with more probes, if available */
for (pj = s->jobs; pj; pj = pj->jobs_next) {
if (pj->chosen)
continue;
if (!packet_add_probe_query(s, p, pj))
break;
n++;
}
avahi_dns_packet_set_field(p, AVAHI_DNS_FIELD_QDCOUNT, n);
n = 0;
/* Now add the chosen records to the authorative section */
for (pj = s->jobs; pj; pj = next) {
next = pj->jobs_next;
if (!pj->chosen)
continue;
if (!avahi_dns_packet_append_record(p, pj->record, 0, 0)) {
/* avahi_log_warn("Bad probe size estimate!"); */
/* Unmark all following jobs */
for (; pj; pj = pj->jobs_next)
pj->chosen = 0;
break;
}
job_mark_done(s, pj);
n ++;
}
avahi_dns_packet_set_field(p, AVAHI_DNS_FIELD_NSCOUNT, n);
/* Send it now */
avahi_interface_send_packet(s->interface, p);
avahi_dns_packet_free(p);
}
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);
}
int avahi_response_scheduler_post(AvahiResponseScheduler *s, AvahiRecord *record, int flush_cache, const AvahiAddress *querier, int immediately) {
AvahiResponseJob *rj;
struct timeval tv;
/* char *t; */
assert(s);
assert(record);
assert(!avahi_key_is_pattern(record->key));
/* t = avahi_record_to_string(record); */
/* avahi_log_debug("post %i %s", immediately, t); */
/* avahi_free(t); */
/* Check whether this response is suppressed */
if (querier &&
(rj = find_suppressed_job(s, record, querier)) &&
avahi_record_is_goodbye(record) == avahi_record_is_goodbye(rj->record) &&
rj->record->ttl >= record->ttl/2) {
/* avahi_log_debug("Response suppressed by known answer suppression."); */
return 0;
}
/* Check if we already sent this response recently */
if ((rj = find_history_job(s, record))) {
if (avahi_record_is_goodbye(record) == avahi_record_is_goodbye(rj->record) &&
rj->record->ttl >= record->ttl/2 &&
(rj->flush_cache || !flush_cache)) {
/* avahi_log_debug("Response suppressed by local duplicate suppression (history)"); */
return 0;
}
/* Outdated ... */
job_free(s, rj);
}
avahi_elapse_time(&tv, immediately ? 0 : AVAHI_RESPONSE_DEFER_MSEC, immediately ? 0 : AVAHI_RESPONSE_JITTER_MSEC);
if ((rj = find_scheduled_job(s, record))) {
/* avahi_log_debug("Response suppressed by local duplicate suppression (scheduled)"); */
/* Update a little ... */
/* Update the time if the new is prior to the old */
if (avahi_timeval_compare(&tv, &rj->delivery) < 0) {
rj->delivery = tv;
avahi_time_event_update(rj->time_event, &rj->delivery);
}
/* Update the flush cache bit */
if (flush_cache)
rj->flush_cache = 1;
/* Update the querier field */
if (!querier || (rj->querier_valid && avahi_address_cmp(querier, &rj->querier) != 0))
rj->querier_valid = 0;
/* Update record data (just for the TTL) */
avahi_record_unref(rj->record);
rj->record = avahi_record_ref(record);
return 1;
} else {
/* avahi_log_debug("Accepted new response job."); */
/* Create a new job and schedule it */
if (!(rj = job_new(s, record, AVAHI_SCHEDULED)))
return 0; /* OOM */
rj->delivery = tv;
rj->time_event = avahi_time_event_new(s->time_event_queue, &rj->delivery, elapse_callback, rj);
rj->flush_cache = flush_cache;
if ((rj->querier_valid = !!querier))
rj->querier = *querier;
return 1;
}
}
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() */
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);
}
int job_reap(bool interactive)
{
ASSERT_IS_MAIN_THREAD();
job_t *jnext;
int found=0;
/* job_reap may fire an event handler, we do not want to call ourselves recursively (to avoid infinite recursion). */
static bool locked = false;
if (locked)
{
return 0;
}
locked = true;
process_mark_finished_children(false);
/* Preserve the exit status */
const int saved_status = proc_get_last_status();
job_iterator_t jobs;
const size_t job_count = jobs.count();
jnext = jobs.next();
while (jnext)
{
job_t *j = jnext;
jnext = jobs.next();
/*
If we are reaping only jobs who do not need status messages
sent to the console, do not consider reaping jobs that need
status messages
*/
if ((!job_get_flag(j, JOB_SKIP_NOTIFICATION)) && (!interactive) && (!job_get_flag(j, JOB_FOREGROUND)))
{
continue;
}
for (process_t *p = j->first_process; p; p=p->next)
{
int s;
if (!p->completed)
continue;
if (!p->pid)
continue;
s = p->status;
proc_fire_event(L"PROCESS_EXIT", EVENT_EXIT, p->pid, (WIFSIGNALED(s)?-1:WEXITSTATUS(s)));
if (WIFSIGNALED(s))
{
/*
Ignore signal SIGPIPE.We issue it ourselves to the pipe
writer when the pipe reader dies.
*/
if (WTERMSIG(s) != SIGPIPE)
{
int proc_is_job = ((p==j->first_process) && (p->next == 0));
if (proc_is_job)
job_set_flag(j, JOB_NOTIFIED, 1);
if (!job_get_flag(j, JOB_SKIP_NOTIFICATION))
{
/* Print nothing if we get SIGINT in the foreground process group, to avoid spamming obvious stuff on the console (#1119). If we get SIGINT for the foreground process, assume the user typed ^C and can see it working. It's possible they didn't, and the signal was delivered via pkill, etc., but the SIGINT/SIGTERM distinction is precisely to allow INT to be from a UI and TERM to be programmatic, so this assumption is keeping with the design of signals.
If echoctl is on, then the terminal will have written ^C to the console. If off, it won't have. We don't echo ^C either way, so as to respect the user's preference. */
if (WTERMSIG(p->status) != SIGINT || ! job_get_flag(j, JOB_FOREGROUND))
{
if (proc_is_job)
{
// We want to report the job number, unless it's the only job, in which case we don't need to
const wcstring job_number_desc = (job_count == 1) ? wcstring() : format_string(L"Job %d, ", j->job_id);
fwprintf(stdout,
_(L"%ls: %ls\'%ls\' terminated by signal %ls (%ls)"),
program_name,
job_number_desc.c_str(),
truncate_command(j->command()).c_str(),
sig2wcs(WTERMSIG(p->status)),
signal_get_desc(WTERMSIG(p->status)));
}
else
{
const wcstring job_number_desc = (job_count == 1) ? wcstring() : format_string(L"from job %d, ", j->job_id);
fwprintf(stdout,
_(L"%ls: Process %d, \'%ls\' %ls\'%ls\' terminated by signal %ls (%ls)"),
program_name,
p->pid,
p->argv0(),
job_number_desc.c_str(),
truncate_command(j->command()).c_str(),
sig2wcs(WTERMSIG(p->status)),
signal_get_desc(WTERMSIG(p->status)));
}
tputs(clr_eol,1,&writeb);
fwprintf(stdout, L"\n");
}
found=1;
}
/*
Clear status so it is not reported more than once
*/
p->status = 0;
}
}
}
/*
If all processes have completed, tell the user the job has
completed and delete it from the active job list.
*/
if (job_is_completed(j))
{
if (!job_get_flag(j, JOB_FOREGROUND) && !job_get_flag(j, JOB_NOTIFIED) && !job_get_flag(j, JOB_SKIP_NOTIFICATION))
{
format_job_info(j, _(L"ended"), job_count);
found=1;
}
proc_fire_event(L"JOB_EXIT", EVENT_EXIT, -j->pgid, 0);
proc_fire_event(L"JOB_EXIT", EVENT_JOB_ID, j->job_id, 0);
job_free(j);
}
else if (job_is_stopped(j) && !job_get_flag(j, JOB_NOTIFIED))
{
/*
Notify the user about newly stopped jobs.
*/
if (!job_get_flag(j, JOB_SKIP_NOTIFICATION))
{
format_job_info(j, _(L"stopped"), job_count);
found=1;
}
job_set_flag(j, JOB_NOTIFIED, 1);
}
}
if (found)
fflush(stdout);
/* Restore the exit status. */
proc_set_last_status(saved_status);
locked = false;
return found;
}
job *job_recov(
char *filename) /* I */ /* pathname to job save file */
{
int fds;
job *pj;
char *pn;
char namebuf[MAXPATHLEN];
char log_buf[LOCAL_LOG_BUF_SIZE];
#ifndef PBS_MOM
char parent_id[PBS_MAXSVRJOBID + 1];
job_array *pa;
#endif
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 */
fds = open(namebuf, O_RDONLY, 0);
if (fds < 0)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "unable to open %s", namebuf);
log_err(errno, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "1", LOGLEVEL);
free(pj->ji_mutex);
#endif
free((char *)pj);
/* FAILURE - cannot open job file */
return(NULL);
}
/* read in job quick save sub-structure */
if (read_ac_socket(fds, (char *)&pj->ji_qs, sizeof(pj->ji_qs)) != sizeof(pj->ji_qs) &&
pj->ji_qs.qs_version == PBS_QS_VERSION)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "Unable to read %s", namebuf);
log_err(errno, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "2", LOGLEVEL);
free(pj->ji_mutex);
#endif
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 */
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"%s appears to be from an old version. Attempting to convert.\n",
namebuf);
log_err(-1, __func__, log_buf);
if (job_qs_upgrade(pj, fds, namebuf, pj->ji_qs.qs_version) != 0)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "unable to upgrade %s\n", namebuf);
log_err(-1, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "3", LOGLEVEL);
free(pj->ji_mutex);
#endif
free((char *)pj);
close(fds);
return(NULL);
}
} /* 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 */
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "Job Id %s does not match file name for %s",
pj->ji_qs.ji_jobid,
namebuf);
log_err(-1, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "4", LOGLEVEL);
free(pj->ji_mutex);
#endif
free((char *)pj);
close(fds);
return(NULL);
}
/* read in working attributes */
if (recov_attr(
fds,
pj,
job_attr_def,
pj->ji_wattr,
JOB_ATR_LAST,
JOB_ATR_UNKN,
TRUE) != 0)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "unable to recover %s (file is likely corrupted)", namebuf);
log_err(-1, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "5", LOGLEVEL);
job_free(pj, FALSE);
#else
mom_job_free(pj);
#endif
close(fds);
return(NULL);
}
#ifndef PBS_MOM
/* Comment out the mother superior tracking. Will be debugged later
if (pj->ji_wattr[JOB_ATR_exec_host].at_val.at_str != NULL)
{*/
/* add job to the mother superior list for it's node */
/* char *ms = strdup(pj->ji_wattr[JOB_ATR_exec_host].at_val.at_str);
char *end = strchr(ms, '/');
if (end != NULL)
*end = '\0';
if ((end = strchr(ms, '+')) != NULL)
*end = '\0';
add_to_ms_list(ms, pj);
free(ms);
}*/
#endif
#ifdef PBS_MOM
/* read in tm sockets and ips */
if (recov_tmsock(fds, pj) != 0)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"warning: tmsockets not recovered from %s (written by an older pbs_mom?)",
namebuf);
log_err(-1, __func__, log_buf);
}
#else /* not PBS_MOM */
if (strchr(pj->ji_qs.ji_jobid, '[') != NULL)
{
/* 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 (pa == NULL)
{
job_abt(&pj, (char *)"Array job missing array struct, aborting job");
close(fds);
return NULL;
}
strcpy(pj->ji_arraystructid, parent_id);
if (strcmp(parent_id, pj->ji_qs.ji_jobid) == 0)
{
pj->ji_is_array_template = TRUE;
}
else
{
pa->job_ids[(int)pj->ji_wattr[JOB_ATR_job_array_id].at_val.at_long] = strdup(pj->ji_qs.ji_jobid);
pa->jobs_recovered++;
/* This is a bit of a kluge, but for some reason if an array job was
on hold when the server went down the ji_wattr[JOB_ATR_hold].at_val.at_long
value is 0 on recovery even though pj->ji_qs.ji_state is JOB_STATE_HELD and
the substate is JOB_SUBSTATE_HELD
*/
if ((pj->ji_qs.ji_state == JOB_STATE_HELD) &&
(pj->ji_qs.ji_substate == JOB_SUBSTATE_HELD))
{
pj->ji_wattr[JOB_ATR_hold].at_val.at_long = HOLD_l;
pj->ji_wattr[JOB_ATR_hold].at_flags = ATR_VFLAG_SET;
}
}
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
}
#endif
close(fds);
pj->ji_commit_done = 1;
/* all done recovering the job */
job_save(pj, SAVEJOB_FULL, 0);
return(pj);
} /* END job_recov() */
/* command_exec(cmd, pass_pipefd)
*
* Execute the single command specified in the 'cmd' command structure.
*
* The 'pass_pipefd' argument is used for pipes.
* On input, '*pass_pipefd' is the file descriptor used to read the
* previous command's output. That is, it's the read end of the previous
* pipe. It equals STDIN_FILENO if there was no previous pipe.
* On output, command_exec should set '*pass_pipefd' to the file descriptor
* used for reading from THIS command's pipe.
* If this command didn't have a pipe -- that is, if cmd->commandop != PIPE
* -- then it should set '*pass_pipefd = STDIN_FILENO'.
*
* Returns the process ID of the forked child, or < 0 if some system call
* fails.
*
* You must also handle the internal commands "cd" and "exit".
* These are special because they must execute in the shell process, rather
* than a child. (Why?)
*
* However, these special commands still have a status!
* For example, "cd DIR" should return status 0 if we successfully change
* to the DIR directory, and status 1 otherwise.
* Thus, "cd /tmp && echo /tmp exists" should print "/tmp exists" to stdout
* iff the /tmp directory exists.
* Not only this, but redirections should work too!
* For example, "cd /tmp > foo" should create an empty file named 'foo';
* and "cd /tmp 2> foo" should print any error messages to 'foo'.
*
* How can you return a status, and do redirections, for a command executed
* in the parent shell?
* Hint: It is easiest if you fork a child ANYWAY!
* You should divide functionality between the parent and the child.
* Some functions will be executed in each process.
*/
static pid_t
command_exec(command_t *cmd, int *pass_pipefd)
{
pid_t pid = -1; // process ID for child
int pipefd[2]; // file descriptors for this process's pipe
int status;
/* EXERCISE: Complete this function!
* We've written some of the skeleton for you, but feel free to
* change it.
*/
// Create a pipe, if this command is the left-hand side of a pipe.
// Return -1 if the pipe fails.
/* Your code here. */
if(cmd->subshell && cmd->argv[0] && !(strcmp(cmd->argv[0],"") == 0)){
goto error;
}
if (cmd->controlop == CMD_PIPE) {
if(pipe(pipefd)){
printf("Error creating pipe\n");
return -1;
}
}
// Fork the child and execute the command in that child.
// You will handle all redirections by manipulating file descriptors.
//
// This section is fairly long. It is probably best to implement this
// part in stages, checking it after each step. For instance, first
// implement just the fork and the execute in the child. This should
// allow you to execute simple commands like 'ls'. Then add support
// for redirections: commands like 'ls > foo' and 'cat < foo'. Then
// add parentheses, then pipes, and finally the internal commands
// 'cd' and 'exit'.
//
// In the child, you should:
// 1. Set up stdout to point to this command's pipe, if necessary.
// 2. Set up stdin to point to the PREVIOUS command's pipe (that
// is, *pass_pipefd), if appropriate.
// 3. Close some file descriptors. Hint: Consider the read end
// of this process's pipe.
// 4. Set up redirections.
// Hint: For output redirections (stdout and stderr), the 'mode'
// argument of open() should be set to 0666.
// 5. Execute the command.
// There are some special cases:
// a. Parentheses. Execute cmd->subshell. (How?)
// b. A null command (no subshell, no arguments).
// Exit with status 0.
// c. "exit".
// d. "cd".
//
// In the parent, you should:
// 1. Close some file descriptors. Hint: Consider the write end
// of this command's pipe, and one other fd as well.
// 2. Handle the special "exit" and "cd" commands.
// 3. Set *pass_pipefd as appropriate.
//
// "cd" error note:
// - Upon syntax errors: Display the message
// "cd: Syntax error on bad number of arguments"
// - Upon system call errors: Call perror("cd")
//
// "cd" Hints:
// For the "cd" command, you should change directories AFTER
// the fork(), not before it. Why?
// Design some tests with 'bash' that will tell you the answer.
// For example, try "cd /tmp ; cd $HOME > foo". In which directory
// does foo appear, /tmp or $HOME? If you chdir() BEFORE the fork,
// in which directory would foo appear, /tmp or $HOME?
//
if(!cmd->subshell && strcmp(cmd->argv[0],"exit") == 0){
// execute exit
exit(0);
}
if(!cmd->subshell && (strcmp(cmd->argv[0],"") == 0))
return -1;
//printf("Spawning new process\n");
// Handle reaping of zombie processes automatically
// struct sigaction sigact;
// sigact.sa_handler = SIG_IGN;
// sigemptyset(&sigact.sa_mask);
// sigact.sa_flags = 0;
// if (sigaction(SIGCHLD, &sigact, 0) == -1) {
// printf("Triggered\n");
// perror(0);
// exit(1);
// }
// if (signal(SIGCHLD, SIG_IGN) == SIG_ERR) {
// perror(0);
// exit(1);
// }
/* Your code here. */
pid = fork();
// parent process
if (pid > 0){
if(cmd->controlop == CMD_PIPE) {
*pass_pipefd = pipefd[0];
close(pipefd[1]);
}
if(cmd->argv[0]){
if(strcmp(cmd->argv[0],"cd") == 0){
if(!cmd->argv[1]){
if(chdir(getenv("HOME")) != 0){
int lsts;
waitpid(pid, &lsts, WNOHANG);
perror("cd");
return -1;
}
return 0;
}
else if ((cmd->argv[1])[0] == '~'){
cmd->argv[1] = strcat(getenv("HOME"),(cmd->argv[1])+1);
}
if(chdir(cmd->argv[1]) != 0){
int lsts;
waitpid(pid, &lsts, WNOHANG);
perror("cd");
return -1;
}
}
else if(strcmp(cmd->argv[0],"jobs") == 0){
if(jobindex) {
printf("PID PPID STATUS\n");
int removelist[100];
int j, k = 0;
for(j = 0 ; j<jobindex;j++){
char *sts;
int stspid;
// printf("Current Status: %d\n",waitpid(joblist[j]->pid,&stspid,WNOHANG));
joblist[j]->status = waitpid(joblist[j]->pid,&stspid,WNOHANG) ? DONE : RUNNING ;
switch(joblist[j]->status){
default: printf("Error Setting value\n");
break;
case RUNNING: sts = "RUNNING";
break;
case DONE: sts = "DONE";
joblist[j]->printstatus = 1;
removelist[k++] = j;
break;
case KILLED: sts = "KILLED";
break;
}
// print the status of background jobs
printf("%d %d %s\n",joblist[j]->pid,joblist[j]->ppid,sts);
}
int l, m = 0;
for(l = 0 ;l<k;l++){
job_free(removelist[l]-m++);
}
}
}
}
return pid;
}
// child process
else if (pid == 0){
FILE *fp;
dup2(*pass_pipefd, STDIN_FILENO);
if(cmd->redirect_filename[0]){
fp = fopen(cmd->redirect_filename[0], "r");
dup2(fileno(fp),0);
}
if(cmd->redirect_filename[1]){
fp =fopen(cmd->redirect_filename[1], "w");
dup2(fileno(fp), 1);
}
if(cmd->redirect_filename[2]){
//close(2);
fp = fopen(cmd->redirect_filename[2], "w");
dup2(fileno(fp), 2);
}
if(cmd->controlop == CMD_PIPE) {
dup2(pipefd[1],STDOUT_FILENO);
close(pipefd[1]);
}
if(cmd->subshell){
//printf("Entered inside: %s\n",(cmd->subshell)->argv[0]);
status = command_line_exec(cmd->subshell);
//printf("Status: %d\n",status);
// Return teh status of the last executed command
if(status)
kill(getpid(),SIGKILL);
//printf("Reached exit error\n");
exit(status);
}
// printf("In child process [0]:%s, %s\n",cmd->argv[0], cmd->redirect_filename[1]);
if(strcmp(cmd->argv[0],"cd") == 0){
// Exit
exit(0);
}
if(strcmp(cmd->argv[0],"jobs") == 0){
// Exit
exit(0);
}
//handle wildcard entries
int z,m=0;
char *local_cmd[512];
local_cmd[m++] = cmd->argv[0];
for(z = 1; z < 512 ;z++){
//printf("Reached %d\n",z );
local_cmd[m++] = cmd->argv[z]?cmd->argv[z]:0;
//printf("Reached %d\n",z );
if(!cmd->argv[z])
break;
// if((strcmp(local_cmd[0],"ls") == 0) && (strstr(cmd->argv[z],"*") != NULL)) {
if(strstr(cmd->argv[z],"*") != NULL) {
m--;
const char *pattern = cmd->argv[z];
glob_t results;
// glob function : glob(const char *pattern, int flags, int (*errfunc) (const char *epath, int eerrno), glob_t *pglob);
if(glob(pattern,0,glob_error, &results) != 0)
exit(1);
unsigned int k;
for (k = 0; k < results.gl_pathc; k++){
//printf("%s\n", results.gl_pathv[k]);
local_cmd[m++] = results.gl_pathv[k];
}
break;
globfree(& results);
}
}
local_cmd[m] = 0;
execvp(local_cmd[0],local_cmd);
// execvp(cmd->argv[0],cmd->argv);
return -1;
}
error:
printf("Error\n");
return -1;
}
// Readrec5 is like readrec, but it reads a record in "version 5"
// of the log format.
static int
readrec5(File *f, job l, int *err)
{
int r, sz = 0;
size_t namelen;
Jobrec5 jr;
job j;
tube t;
char tubename[MAX_TUBE_NAME_LEN];
r = read(f->fd, &namelen, sizeof(namelen));
if (r == -1) {
twarn("read");
warnpos(f, 0, "error");
*err = 1;
return 0;
}
if (r != sizeof(namelen)) {
return 0;
}
sz += r;
if (namelen >= MAX_TUBE_NAME_LEN) {
warnpos(f, -r, "namelen %zu exceeds maximum of %d", namelen, MAX_TUBE_NAME_LEN - 1);
*err = 1;
return 0;
}
if (namelen) {
r = readfull(f, tubename, namelen, err, "v5 tube name");
if (!r) {
return 0;
}
sz += r;
}
tubename[namelen] = '\0';
r = readfull(f, &jr, Jobrec5size, err, "v5 job struct");
if (!r) {
return 0;
}
sz += r;
// are we reading trailing zeroes?
if (!jr.id) return 0;
j = job_find(jr.id);
if (!(j || namelen)) {
// We read a short record without having seen a
// full record for this job, so the full record
// was in an eariler file that has been deleted.
// Therefore the job itself has either been
// deleted or migrated; either way, this record
// should be ignored.
return 1;
}
switch (jr.state) {
case Reserved:
jr.state = Ready;
case Ready:
case Buried:
case Delayed:
if (!j) {
if (jr.body_size > job_data_size_limit) {
warnpos(f, -r, "job %"PRIu64" is too big (%"PRId32" > %zu)",
jr.id,
jr.body_size,
job_data_size_limit);
goto Error;
}
t = tube_find_or_make(tubename);
j = make_job_with_id(jr.pri, jr.delay, jr.ttr, jr.body_size,
t, jr.id);
j->next = j->prev = j;
j->r.created_at = jr.created_at;
}
j->r.id = jr.id;
j->r.pri = jr.pri;
j->r.delay = jr.delay * 1000; // us => ns
j->r.ttr = jr.ttr * 1000; // us => ns
j->r.body_size = jr.body_size;
j->r.created_at = jr.created_at * 1000; // us => ns
j->r.deadline_at = jr.deadline_at * 1000; // us => ns
j->r.reserve_ct = jr.reserve_ct;
j->r.timeout_ct = jr.timeout_ct;
j->r.release_ct = jr.release_ct;
j->r.bury_ct = jr.bury_ct;
j->r.kick_ct = jr.kick_ct;
j->r.state = jr.state;
job_insert(l, j);
// full record; read the job body
if (namelen) {
if (jr.body_size != j->r.body_size) {
warnpos(f, -r, "job %"PRIu64" size changed", j->r.id);
warnpos(f, -r, "was %"PRId32", now %"PRId32, j->r.body_size, jr.body_size);
goto Error;
}
r = readfull(f, j->body, j->r.body_size, err, "v5 job body");
if (!r) {
goto Error;
}
sz += r;
// since this is a full record, we can move
// the file pointer and decref the old
// file, if any
filermjob(j->file, j);
fileaddjob(f, j);
}
j->walused += sz;
f->w->alive += sz;
return 1;
case Invalid:
if (j) {
job_remove(j);
filermjob(j->file, j);
job_free(j);
}
return 1;
}
Error:
*err = 1;
if (j) {
job_remove(j);
filermjob(j->file, j);
job_free(j);
}
return 0;
}
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);
}
#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 (pa == NULL)
{
job_abt(&pj, "Array job missing array struct, aborting job");
close(fds);
return NULL;
}
if (strcmp(parent_id, pj->ji_qs.ji_jobid) == 0)
{
pj->ji_is_array_template = TRUE;
pj->ji_arraystruct = pa;
}
else
{
pa->jobs[(int)pj->ji_wattr[JOB_ATR_job_array_id].at_val.at_long] = (void *)pj;
pj->ji_arraystruct = pa;
pa->jobs_recovered++;
/* This is a bit of a kluge, but for some reason if an array job was
on hold when the server went down the ji_wattr[JOB_ATR_hold].at_val.at_long
value is 0 on recovery even though pj->ji_qs.ji_state is JOB_STATE_HELD and
the substate is JOB_SUBSTATE_HELD
*/
if ((pj->ji_qs.ji_state == JOB_STATE_HELD) && (pj->ji_qs.ji_substate == JOB_SUBSTATE_HELD))
{
pj->ji_wattr[JOB_ATR_hold].at_val.at_long = HOLD_l;
pj->ji_wattr[JOB_ATR_hold].at_flags = ATR_VFLAG_SET;
}
}
}
#endif
close(fds);
/* all done recovering the job */
job_save(pj, SAVEJOB_FULL);
return(pj);
} /* END job_recov() */
