/* This function has a unit test. */
int
string_to_port( const char* addr )
{
const char* acc = addr;
const char* tmp;
int port = 0;
if( ! (addr && is_valid_sinful(addr)) ) {
return 0;
}
if (*acc != '<')
return 0;
acc++;
if (*acc == '[') {
tmp = strchr(acc, ']');
if (!tmp)
return 0;
acc = tmp + 1;
}
tmp = strchr(acc, ':');
if (!tmp)
return 0;
tmp++;
port = atoi(tmp);
return port;
}
JICLocalSchedd::JICLocalSchedd( const char* classad_filename,
const char* schedd_address,
int cluster, int proc, int subproc )
: JICLocalFile( classad_filename, cluster, proc, subproc )
{
// initialize this to something reasonable. we'll change it
// if anything special happens which needs a different value.
exit_code = JOB_EXITED;
if( ! is_valid_sinful(schedd_address) ) {
EXCEPT("schedd_addr not specified with valid address");
}
schedd_addr = strdup( schedd_address );
dprintf( D_ALWAYS,
"Starter running a job under a schedd listening at %s\n",
schedd_addr );
job_updater = NULL;
m_cleanup_retry_tid = -1;
m_num_cleanup_retries = 0;
// NOTE: these config knobs are very similar to
// SHADOW_MAX_JOB_CLEANUP_RETRIES and
// SHADOW_JOB_CLEANUP_RETRY_DELAY in the shadow.
m_max_cleanup_retries = param_integer("LOCAL_UNIVERSE_MAX_JOB_CLEANUP_RETRIES", 5);
m_cleanup_retry_delay = param_integer("LOCAL_UNIVERSE_JOB_CLEANUP_RETRY_DELAY", 30);
this->starter_user_policy = NULL;
m_tried_notify_job_exit = false;
}
static bool test_no_angle_brackets() {
emit_test("Is the string correctly rejected if there are no angle brackets?");
char* input = strdup( "209.172.63.167:8080" );
emit_input_header();
emit_param("SINFUL", input);
int result = is_valid_sinful( input );
free( input );
emit_output_expected_header();
emit_retval("%s", tfstr(FALSE));
emit_output_actual_header();
emit_retval("%s", tfstr(result));
if(result != FALSE) {
FAIL;
}
PASS;
}
static bool test_hostname() {
emit_test("Are hostnames instead of IP addresses rejected like they should be?");
char* input = strdup( "<balthazar.cs.wisc.edu:47>" );
emit_input_header();
emit_param("SINFUL", input);
int result = is_valid_sinful( input );
free( input );
emit_output_expected_header();
emit_retval("%s", tfstr(FALSE));
emit_output_actual_header();
emit_retval("%s", tfstr(result));
if(result != FALSE) {
FAIL;
}
PASS;
}
static bool test_normal_case() {
emit_test("Is normal input identified correctly?");
char* input = strdup( "<208.122.19.56:47>" );
emit_input_header();
emit_param("SINFUL", input);
int result = is_valid_sinful( input );
free( input );
emit_output_expected_header();
emit_retval("%s", tfstr(TRUE));
emit_output_actual_header();
emit_retval("%s", tfstr(result));
if(result != TRUE) {
FAIL;
}
PASS;
}
char*
getAddrFromClaimId( const char* id )
{
char* tmp;
char* addr;
char* copy = strdup( id );
tmp = strchr( copy, '#' );
if( tmp ) {
*tmp = '\0';
if( is_valid_sinful(copy) ) {
addr = strdup( copy );
free( copy );
return addr;
}
}
free( copy );
return NULL;
}
bool
DCStarter::initFromClassAd( ClassAd* ad )
{
char* tmp = NULL;
if( ! ad ) {
dprintf( D_ALWAYS,
"ERROR: DCStarter::initFromClassAd() called with NULL ad\n" );
return false;
}
ad->LookupString( ATTR_STARTER_IP_ADDR, &tmp );
if( ! tmp ) {
// If that's not defined, try ATTR_MY_ADDRESS
ad->LookupString( ATTR_MY_ADDRESS, &tmp );
}
if( ! tmp ) {
dprintf( D_FULLDEBUG, "ERROR: DCStarter::initFromClassAd(): "
"Can't find starter address in ad\n" );
return false;
} else {
if( is_valid_sinful(tmp) ) {
New_addr( strnewp(tmp) );
is_initialized = true;
} else {
dprintf( D_FULLDEBUG,
"ERROR: DCStarter::initFromClassAd(): invalid %s in ad (%s)\n",
ATTR_STARTER_IP_ADDR, tmp );
}
free( tmp );
tmp = NULL;
}
if( ad->LookupString(ATTR_VERSION, &tmp) ) {
New_version( strnewp(tmp) );
free( tmp );
tmp = NULL;
}
return is_initialized;
}
int
main(int argc, char *argv[])
{
char *arg;
int nArgs = 0; // number of args
int i, result;
char* pool = NULL;
char* scheddName = NULL;
char* scheddAddr = NULL;
MyString method;
char *tmp;
myDistro->Init( argc, argv );
MyName = condor_basename(argv[0]);
config();
#if !defined(WIN32)
install_sig_handler(SIGPIPE, SIG_IGN );
#endif
// dig around in the config file looking for what the config file says
// about getting files from Condor. This defaults with the global variable
// initialization.
tmp = param( "SANDBOX_TRANSFER_METHOD" );
if ( tmp != NULL ) {
method = tmp;
free( tmp );
string_to_stm( method, st_method );
}
char **args = (char **)malloc(sizeof(char *) * argc); // args
if ( ! args) exit(2);
// parse the arguments.
for( argv++; (arg = *argv); argv++ ) {
if( arg[0] == '-' ) {
if( ! arg[1] ) {
usage();
}
switch( arg[1] ) {
case 'd':
// dprintf to console
dprintf_set_tool_debug("TOOL", 0);
break;
case 'c':
args[nArgs] = arg;
nArgs++;
argv++;
if( ! *argv ) {
fprintf( stderr,
"%s: -constraint requires another argument\n",
MyName);
exit(1);
}
args[nArgs] = *argv;
nArgs++;
break;
case 'a':
if( arg[2] && arg[2] == 'd' ) {
argv++;
if( ! *argv ) {
fprintf( stderr,
"%s: -addr requires another argument\n",
MyName);
exit(1);
}
if( is_valid_sinful(*argv) ) {
scheddAddr = strdup(*argv);
if( ! scheddAddr ) {
fprintf( stderr, "Out of Memory!\n" );
exit(1);
}
} else {
fprintf( stderr,
"%s: \"%s\" is not a valid address\n",
MyName, *argv );
fprintf( stderr, "Should be of the form "
"<ip.address.here:port>\n" );
fprintf( stderr,
"For example: <123.456.789.123:6789>\n" );
exit( 1 );
}
break;
}
All = true;
break;
case 'n':
// use the given name as the schedd name to connect to
argv++;
if( ! *argv ) {
fprintf( stderr, "%s: -name requires another argument\n",
MyName);
exit(1);
}
if ( scheddName ) free(scheddName);
scheddName = strdup(*argv);
break;
case 'p':
// use the given name as the central manager to query
argv++;
if( ! *argv ) {
fprintf( stderr, "%s: -pool requires another argument\n",
MyName);
exit(1);
}
if( pool ) {
free( pool );
}
pool = strdup( *argv );
break;
case 's':
argv++;
if( ! *argv ) {
fprintf( stderr, "%s: -stm requires another argument\n",
MyName);
exit(1);
}
method = *argv;
string_to_stm(method, st_method);
break;
case 'v':
version();
break;
case 'h':
usage(0);
break;
default:
fprintf( stderr, "Unrecognized option: %s\n", arg );
usage();
break;
}
} else {
if( All ) {
// If -all is set, there should be no other
// constraint arguments.
usage();
}
args[nArgs] = arg;
nArgs++;
}
}
// Check to make sure we have a valid sandbox transfer mechanism.
if (st_method == STM_UNKNOWN) {
fprintf( stderr,
"%s: Unknown sandbox transfer method: %s\n", MyName,
method.Value());
usage();
exit(1);
}
if( ! (All || nArgs) ) {
// We got no indication of what to act on
fprintf( stderr, "You did not specify any jobs\n" );
usage();
}
// We're done parsing args, now make sure we know how to
// contact the schedd.
if( ! scheddAddr ) {
// This will always do the right thing, even if either or
// both of scheddName or pool are NULL.
schedd = new DCSchedd( scheddName, pool );
} else {
schedd = new DCSchedd( scheddAddr );
}
if( ! schedd->locate() ) {
fprintf( stderr, "%s: %s\n", MyName, schedd->error() );
exit( 1 );
}
// Process the args.
if( All ) {
handleAll();
} else {
for(i = 0; i < nArgs; i++) {
if( match_prefix( args[i], "-constraint" ) ) {
i++;
addConstraint( args[i] );
} else {
procArg(args[i]);
}
}
}
// Sanity check: make certain we now have a constraint
if ( global_constraint.Length() <= 0 ) {
fprintf( stderr, "Unable to create a job constraint!\n");
exit(1);
}
fprintf(stdout,"Fetching data files...\n");
switch(st_method) {
case STM_USE_SCHEDD_ONLY:
{ // start block
// Get the sandbox directly from the schedd.
// And now, do the work.
CondorError errstack;
result = schedd->receiveJobSandbox(global_constraint.Value(),
&errstack);
if ( !result ) {
fprintf( stderr, "\n%s\n", errstack.getFullText(true).c_str() );
fprintf( stderr, "ERROR: Failed to spool job files.\n" );
exit(1);
}
// All done
return 0;
} //end block
break;
case STM_USE_TRANSFERD:
{ // start block
// NEW METHOD where we ask the schedd for a transferd, then get the
// files from the transferd
CondorError errstack;
ClassAd respad;
int invalid;
MyString reason;
MyString td_sinful;
MyString td_cap;
result = schedd->requestSandboxLocation(FTPD_DOWNLOAD,
global_constraint, FTP_CFTP, &respad, &errstack);
if ( !result ) {
fprintf( stderr, "\n%s\n", errstack.getFullText(true).c_str() );
fprintf( stderr, "ERROR: Failed to spool job files.\n" );
exit(1);
}
respad.LookupInteger(ATTR_TREQ_INVALID_REQUEST, invalid);
if (invalid == TRUE) {
fprintf( stderr, "ERROR: Failed to spool job files.\n" );
respad.LookupString(ATTR_TREQ_INVALID_REASON, reason);
fprintf( stderr, "%s\n", reason.Value());
exit(EXIT_FAILURE);
}
respad.LookupString(ATTR_TREQ_TD_SINFUL, td_sinful);
respad.LookupString(ATTR_TREQ_CAPABILITY, td_cap);
dprintf(D_ALWAYS,
"td: %s, cap: %s\n", td_sinful.Value(), td_cap.Value());
DCTransferD dctd(td_sinful.Value());
result = dctd.download_job_files(&respad, &errstack);
if ( !result ) {
fprintf( stderr, "\n%s\n", errstack.getFullText(true).c_str() );
fprintf( stderr, "ERROR: Failed to spool job files.\n" );
exit(1);
}
} // end block
break;
default:
EXCEPT("PROGRAMMER ERROR: st_method must be known.");
break;
}
// All done
return 0;
}
void
BaseShadow::baseInit( ClassAd *job_ad, const char* schedd_addr, const char *xfer_queue_contact_info )
{
int pending = FALSE;
if( ! job_ad ) {
EXCEPT("baseInit() called with NULL job_ad!");
}
jobAd = job_ad;
if (sendUpdatesToSchedd && ! is_valid_sinful(schedd_addr)) {
EXCEPT("schedd_addr not specified with valid address");
}
scheddAddr = sendUpdatesToSchedd ? strdup( schedd_addr ) : strdup("noschedd");
m_xfer_queue_contact_info = xfer_queue_contact_info;
if ( !jobAd->LookupString(ATTR_OWNER, owner)) {
EXCEPT("Job ad doesn't contain an %s attribute.", ATTR_OWNER);
}
if( !jobAd->LookupInteger(ATTR_CLUSTER_ID, cluster)) {
EXCEPT("Job ad doesn't contain a %s attribute.", ATTR_CLUSTER_ID);
}
if( !jobAd->LookupInteger(ATTR_PROC_ID, proc)) {
EXCEPT("Job ad doesn't contain a %s attribute.", ATTR_PROC_ID);
}
// Grab the GlobalJobId if we've got it.
if( ! jobAd->LookupString(ATTR_GLOBAL_JOB_ID, &gjid) ) {
gjid = NULL;
}
// grab the NT domain if we've got it
jobAd->LookupString(ATTR_NT_DOMAIN, domain);
if ( !jobAd->LookupString(ATTR_JOB_IWD, iwd)) {
EXCEPT("Job ad doesn't contain an %s attribute.", ATTR_JOB_IWD);
}
if( !jobAd->LookupFloat(ATTR_BYTES_SENT, prev_run_bytes_sent) ) {
prev_run_bytes_sent = 0;
}
if( !jobAd->LookupFloat(ATTR_BYTES_RECVD, prev_run_bytes_recvd) ) {
prev_run_bytes_recvd = 0;
}
// construct the core file name we'd get if we had one.
MyString tmp_name = iwd;
tmp_name += DIR_DELIM_CHAR;
tmp_name += "core.";
tmp_name += cluster;
tmp_name += '.';
tmp_name += proc;
core_file_name = strdup( tmp_name.Value() );
// put the shadow's sinful string into the jobAd. Helpful for
// the mpi shadow, at least...and a good idea in general.
MyString tmp_addr = ATTR_MY_ADDRESS;
tmp_addr += "=\"";
tmp_addr += daemonCore->InfoCommandSinfulString();
tmp_addr += '"';
if ( !jobAd->Insert( tmp_addr.Value() )) {
EXCEPT( "Failed to insert %s!", ATTR_MY_ADDRESS );
}
DebugId = display_dprintf_header;
config();
// Make sure we've got enough swap space to run
checkSwap();
// handle system calls with Owner's privilege
// XXX this belong here? We'll see...
// Calling init_user_ids() while in user priv causes badness.
// Make sure we're in another priv state.
set_condor_priv();
if ( !init_user_ids(owner.Value(), domain.Value())) {
dprintf(D_ALWAYS, "init_user_ids() failed as user %s\n",owner.Value() );
// uids.C will EXCEPT when we set_user_priv() now
// so there's not much we can do at this point
#if ! defined(WIN32)
if ( param_boolean( "SHADOW_RUN_UNKNOWN_USER_JOBS", false ) )
{
dprintf(D_ALWAYS, "trying init_user_ids() as user nobody\n" );
owner="nobody";
domain=NULL;
if (!init_user_ids(owner.Value(), domain.Value()))
{
dprintf(D_ALWAYS, "init_user_ids() failed!\n");
}
else
{
jobAd->Assign( ATTR_JOB_RUNAS_OWNER, "FALSE" );
m_RunAsNobody=true;
dprintf(D_ALWAYS, "init_user_ids() now running as user nobody\n");
}
}
#endif
}
set_user_priv();
daemonCore->Register_Priv_State( PRIV_USER );
dumpClassad( "BaseShadow::baseInit()", this->jobAd, D_JOB );
// initialize the UserPolicy object
shadow_user_policy.init( jobAd, this );
// setup an object to keep our job ad updated to the schedd's
// permanent job queue. this clears all the dirty bits on our
// copy of the classad, so anything we touch after this will
// be updated to the schedd when appropriate.
// Unless we got a command line arg asking us not to
if (sendUpdatesToSchedd) {
// the usual case
job_updater = new QmgrJobUpdater( jobAd, scheddAddr, CondorVersion() );
} else {
job_updater = new NullQmgrJobUpdater( jobAd, scheddAddr, CondorVersion() );
}
// init user log; hold on failure
// NOTE: job_updater must be initialized _before_ initUserLog(),
// in order to handle the case of the job going on hold as a
// result of failure in initUserLog().
initUserLog();
// change directory; hold on failure
if ( cdToIwd() == -1 ) {
EXCEPT("Could not cd to initial working directory");
}
// check to see if this invocation of the shadow is just to write
// a terminate event and exit since this job had been recorded as
// pending termination, but somehow the job didn't leave the queue
// and the schedd is trying to restart it again..
if( jobAd->LookupInteger(ATTR_TERMINATION_PENDING, pending)) {
if (pending == TRUE) {
// If the classad of this job "thinks" that this job should be
// finished already, let's enact that belief.
// This function does not return.
this->terminateJob(US_TERMINATE_PENDING);
}
}
// If we need to claim the startd before activating the claim
int wantClaiming = 0;
jobAd->LookupBool(ATTR_CLAIM_STARTD, wantClaiming);
if (wantClaiming) {
MyString startdSinful;
MyString claimid;
// Pull startd addr and claimid out of the jobad
jobAd->LookupString(ATTR_STARTD_IP_ADDR, startdSinful);
jobAd->LookupString(ATTR_CLAIM_ID, claimid);
dprintf(D_ALWAYS, "%s is true, trying to claim startd %s\n", ATTR_CLAIM_STARTD, startdSinful.Value());
classy_counted_ptr<DCStartd> startd = new DCStartd("description", NULL, startdSinful.Value(), claimid.Value());
classy_counted_ptr<DCMsgCallback> cb =
new DCMsgCallback((DCMsgCallback::CppFunction)&BaseShadow::startdClaimedCB,
this, jobAd);
// this can't fail, will always call the callback
startd->asyncRequestOpportunisticClaim(jobAd,
"description",
daemonCore->InfoCommandSinfulString(),
1200 /*alive interval*/,
20 /* net timeout*/,
100 /*total timeout*/,
cb);
}
}
void
parseArgv( int /*argc*/, char* argv[] )
{
char** tmp = argv;
for( tmp++; *tmp; tmp++ ) {
if( (*tmp)[0] != '-' ) {
// If it doesn't start with '-', skip it
continue;
}
switch( (*tmp)[1] ) {
// // // // // // // // // // // // // // // //
// Shared options that make sense to all cmds
// // // // // // // // // // // // // // // //
case 'v':
if( strncmp("-version", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
version();
break;
case 'h':
if( strncmp("-help", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
usage( my_name, 0);
break;
case 'd':
if( strncmp("-debug", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
dprintf_set_tool_debug("TOOL", 0);
break;
case 'a':
if( cmd != CA_REQUEST_CLAIM ) {
invalid( *tmp );
}
if( strncmp("-address", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
tmp++;
if( ! (tmp && *tmp) ) {
another( "-address" );
}
if( ! is_valid_sinful(*tmp) ) {
fprintf( stderr, "%s: '%s' is not a valid address\n",
my_name, *tmp );
exit( 1 );
}
if (addr) {
free(addr);
}
addr = strdup( *tmp );
break;
case 'n':
if( cmd != CA_REQUEST_CLAIM ) {
invalid( *tmp );
}
if( strncmp("-name", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
tmp++;
if( ! (tmp && *tmp) ) {
another( "-name" );
}
if (name) {
free(name);
}
name = get_daemon_name( *tmp );
if( ! name ) {
fprintf( stderr, "%s: unknown host %s\n", my_name,
get_host_part(*tmp) );
exit( 1 );
}
break;
// // // // // // // // // // // // // // // //
// Switches that only make sense to some cmds
// // // // // // // // // // // // // // // //
case 'f':
if( !((cmd == CA_RELEASE_CLAIM) ||
(cmd == CA_DEACTIVATE_CLAIM)) )
{
invalid( *tmp );
}
if( strncmp("-fast", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
vacate_type = VACATE_FAST;
break;
case 'r':
if( !((cmd == CA_REQUEST_CLAIM) ||
(cmd == CA_ACTIVATE_CLAIM)) )
{
invalid( *tmp );
}
if( strncmp("-requirements", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
tmp++;
if( ! (tmp && *tmp) ) {
another( "-requirements" );
}
if (requirements) {
free(requirements);
}
requirements = strdup( *tmp );
break;
case 'i':
if( cmd == CA_REQUEST_CLAIM ) {
invalid( *tmp );
}
if( strncmp("-id", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
tmp++;
if( ! (tmp && *tmp) ) {
another( "-id" );
}
if (claim_id) {
free(claim_id);
}
claim_id = strdup( *tmp );
break;
case 'j':
if( cmd != CA_ACTIVATE_CLAIM ) {
invalid( *tmp );
}
if( strncmp("-jobad", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
tmp++;
if( ! (tmp && *tmp) ) {
another( "-jobad" );
}
if (jobad_path) {
free(jobad_path);
}
jobad_path = strdup( *tmp );
break;
case 'k':
if( cmd != CA_ACTIVATE_CLAIM ) {
invalid( *tmp );
}
if( strncmp("-keyword", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
tmp++;
if( ! (tmp && *tmp) ) {
another( "-keyword" );
}
if (job_keyword) {
free(job_keyword);
}
job_keyword = strdup( *tmp );
break;
// // // // // // // // // // // // // // // // // //
// P and C are complicated, since they are ambiguous
// in the case of activate, but not others. so, they
// have their own methods to make it easier to
// understand what the hell's going on. :)
// // // // // // // // // // // // // // // // // //
case 'l':
if( strncmp("-lease", *tmp, strlen(*tmp)) == 0 ) {
if( cmd != CA_REQUEST_CLAIM ) {
invalid( *tmp );
}
tmp++;
if( ! (tmp && *tmp) ) {
another( "-lease" );
}
lease_time = atoi( *tmp );
}
else {
invalid( *tmp );
}
break;
case 't':
if( strncmp("-timeout", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
tmp++;
if( ! (tmp && *tmp) ) {
another( "-timeout" );
}
timeout = atoi( *tmp );
break;
case 'x':
if( strncmp("-x509proxy", *tmp, strlen(*tmp)) ) {
invalid( *tmp );
}
tmp++;
if( ! (tmp && *tmp) ) {
another( "-x509proxy" );
}
proxy_file = *tmp;
break;
case 'p':
parsePOpt( tmp[0], tmp[1] );
tmp++;
break;
case 'c':
parseCOpt( tmp[0], tmp[1] );
tmp++;
break;
default:
invalid( *tmp );
}
}
// Now that we're done parsing, make sure it all makes sense
if( needs_id && ! claim_id ) {
fprintf( stderr, "ERROR: You must specify a ClaimID with "
"-id for %s\n", my_name );
usage( my_name );
}
if( addr && name ) {
fprintf( stderr,
"ERROR: You cannot specify both -name and -address\n" );
usage( my_name );
}
if( addr ) {
target = addr;
} else if( name ) {
target = name;
} else if( claim_id ) {
// This is the last resort, because claim ids are
// no longer considered to be the correct place to
// get the startd's address.
target = getAddrFromClaimId( claim_id );
} else {
// local startd
target = NULL;
}
if( cmd == CA_ACTIVATE_CLAIM && ! (job_keyword || jobad_path) ) {
fprintf( stderr,
"ERROR: You must specify -keyword or -jobad for %s\n",
my_name );
usage( my_name );
}
if (cmd == DELEGATE_GSI_CRED_STARTD && !proxy_file) {
proxy_file = get_x509_proxy_filename();
if (!proxy_file) {
fprintf( stderr,
"\nERROR: can't determine proxy filename to delegate\n" );
exit(1);
}
}
if( jobad_path ) {
if( ! strcmp(jobad_path, "-") ) {
JOBAD_PATH = stdin;
} else {
JOBAD_PATH = safe_fopen_wrapper_follow( jobad_path, "r" );
if( !JOBAD_PATH ) {
fprintf( stderr,
"ERROR: failed to open '%s': errno %d (%s)\n",
jobad_path, errno, strerror(errno) );
exit( 1 );
}
}
}
if( classad_path ) {
CA_PATH = safe_fopen_wrapper_follow( classad_path, "w" );
if( !CA_PATH ) {
fprintf( stderr,
"ERROR: failed to open '%s': errno %d (%s)\n",
classad_path, errno, strerror(errno) );
exit( 1 );
}
}
}
void
parseArgs( int argc, char *argv[] )
{
char *opt;
char** tmp = argv;
for( tmp++; *tmp; tmp++ ) {
opt = tmp[0];
if( sscanf(opt, "%d.%d", &cluster, &proc) == 2 ) {
if( cluster < 0 || proc < 0 ) {
dprintf(D_ALWAYS,
"ERROR: invalid cluster.proc specified: %s\n", opt);
usage(argc, argv);
}
continue;
}
if( opt[0] == '<' ) {
// might be the schedd's address
if( is_valid_sinful(opt)) {
schedd_addr = opt;
continue;
} else {
dprintf(D_ALWAYS,
"ERROR: invalid shadow-private schedd_addr specified: %s\n", opt);
usage(argc, argv);
}
}
if( !strcmp(opt, "--reconnect") || !strcmp(opt, "-reconnect") ) {
is_reconnect = true;
continue;
}
if (strncmp(opt, "--schedd", 8) == 0) {
char *ptr = strchr(opt, '<');
if (ptr && is_valid_sinful(ptr)) {
public_schedd_addr = ptr;
continue;
}
else {
dprintf(D_ALWAYS,
"ERROR: invalid public schedd_addr specified: %s\n",
opt);
usage(argc, argv);
}
}
if (strncmp(opt, "--xfer-queue=", 13) == 0) {
xfer_queue_contact_info = opt+13;
continue;
}
if (strcmp(opt, "--no-schedd-updates") == 0) {
sendUpdatesToSchedd = false;
continue;
}
// the only other argument we understand is the
// filename we should read our ClassAd from, "-" for
// STDIN. There's no further checking we need to do
if( job_ad_file ) {
// already were here, bail out
dprintf( D_ALWAYS, "ERROR: unrecognized option (%s)\n", opt );
usage(argc, argv);
}
job_ad_file = opt;
}
// A proper model of arguments should be presented here and
// used to validate the provided arguments. It would be
// something like:
// if no cluster/proc, who cares
// if no job_ad_file, fail
// And that might be it.
// The validation used to count arguments processed, which was
// easily fooled.
}
int
main( int argc, char *argv[] )
{
char *arg;
char **args = (char **)malloc(sizeof(char *)*(argc - 1)); // args
int nArgs = 0; // number of args
int i;
char* cmd_str;
DCCollector* pool = NULL;
char* scheddName = NULL;
char* scheddAddr = NULL;
// Initialize our global variables
has_constraint = false;
myDistro->Init( argc, argv );
MyName = strrchr( argv[0], DIR_DELIM_CHAR );
if( !MyName ) {
MyName = argv[0];
} else {
MyName++;
}
cmd_str = strchr( MyName, '_');
// we match modes based on characters after the '_'. This means
// 'condor_hold.exe' or 'condor_hold_wrapped' are all legal argv[0]'s
// for condor_hold.
if (cmd_str && strncasecmp( cmd_str, "_hold", strlen("_hold") ) == MATCH) {
mode = JA_HOLD_JOBS;
} else if ( cmd_str &&
strncasecmp( cmd_str, "_release", strlen("_release") ) == MATCH ) {
mode = JA_RELEASE_JOBS;
} else if ( cmd_str &&
strncasecmp( cmd_str, "_suspend", strlen("_suspend") ) == MATCH ) {
mode = JA_SUSPEND_JOBS;
} else if ( cmd_str &&
strncasecmp( cmd_str, "_continue", strlen("_continue") ) == MATCH ) {
mode = JA_CONTINUE_JOBS;
}else if ( cmd_str &&
strncasecmp( cmd_str, "_rm", strlen("_rm") ) == MATCH ) {
mode = JA_REMOVE_JOBS;
} else if( cmd_str && ! strncasecmp(cmd_str, "_vacate_job",
strlen("_vacate_job")) ) {
mode = JA_VACATE_JOBS;
} else {
// don't know what mode we're using, so bail.
fprintf( stderr, "Unrecognized command name, \"%s\"\n", MyName );
usage();
}
config();
if( argc < 2 ) {
// We got no indication of what to act on
fprintf( stderr, "You did not specify any jobs\n" );
usage();
}
#if !defined(WIN32)
install_sig_handler(SIGPIPE, SIG_IGN );
#endif
for( argv++; (arg = *argv); argv++ ) {
if( arg[0] == '-' ) {
if (match_prefix(arg, "-debug")) {
// dprintf to console
dprintf_set_tool_debug("TOOL", 0);
} else if (match_prefix(arg, "-constraint")) {
args[nArgs] = arg;
nArgs++;
argv++;
if( ! *argv ) {
fprintf( stderr,
"%s: -constraint requires another argument\n",
MyName);
exit(1);
}
args[nArgs] = *argv;
nArgs++;
ConstraintArg = true;
} else if (match_prefix(arg, "-all")) {
All = true;
} else if (match_prefix(arg, "-addr")) {
argv++;
if( ! *argv ) {
fprintf( stderr,
"%s: -addr requires another argument\n",
MyName);
exit(1);
}
if( is_valid_sinful(*argv) ) {
scheddAddr = strdup(*argv);
if( ! scheddAddr ) {
fprintf( stderr, "Out of memory!\n" );
exit(1);
}
} else {
fprintf( stderr,
"%s: \"%s\" is not a valid address\n",
MyName, *argv );
fprintf( stderr, "Should be of the form "
"<ip.address.here:port>\n" );
fprintf( stderr,
"For example: <123.456.789.123:6789>\n" );
exit( 1 );
}
} else if (match_prefix(arg, "-reason")) {
argv++;
if( ! *argv ) {
fprintf( stderr,
"%s: -reason requires another argument\n",
MyName);
exit(1);
}
actionReason = strdup(*argv);
if( ! actionReason ) {
fprintf( stderr, "Out of memory!\n" );
exit(1);
}
} else if (match_prefix(arg, "-subcode")) {
argv++;
if( ! *argv ) {
fprintf( stderr,
"%s: -subcode requires another argument\n",
MyName);
exit(1);
}
char *end = NULL;
long code = strtol(*argv,&end,10);
if( code == LONG_MIN || !end || *end || end==*argv ) {
fprintf( stderr, "Invalid -subcode %s!\n", *argv );
exit(1);
}
holdReasonSubCode = strdup(*argv);
ASSERT( holdReasonSubCode );
} else if (match_prefix(arg, "-forcex")) {
if( mode == JA_REMOVE_JOBS ) {
mode = JA_REMOVE_X_JOBS;
} else {
fprintf( stderr,
"-forcex is only valid with condor_rm\n" );
usage();
}
} else if (match_prefix(arg, "-fast")) {
if( mode == JA_VACATE_JOBS ) {
mode = JA_VACATE_FAST_JOBS;
} else {
fprintf( stderr,
"-fast is only valid with condor_vacate_job\n" );
usage();
}
} else if (match_prefix(arg, "-name")) {
// use the given name as the schedd name to connect to
argv++;
if( ! *argv ) {
fprintf( stderr, "%s: -name requires another argument\n",
MyName);
exit(1);
}
if( !(scheddName = get_daemon_name(*argv)) ) {
fprintf( stderr, "%s: unknown host %s\n",
MyName, get_host_part(*argv) );
exit(1);
}
} else if (match_prefix(arg, "-pool")) {
// use the given name as the central manager to query
argv++;
if( ! *argv ) {
fprintf( stderr, "%s: -pool requires another argument\n",
MyName);
exit(1);
}
if( pool ) {
delete pool;
}
pool = new DCCollector( *argv );
if( ! pool->addr() ) {
fprintf( stderr, "%s: %s\n", MyName, pool->error() );
exit(1);
}
} else if (match_prefix(arg, "-version")) {
version();
} else if (match_prefix(arg, "-help")) {
usage(0);
} else {
fprintf( stderr, "Unrecognized option: %s\n", arg );
usage();
}
} else {
if( All ) {
// If -all is set, there should be no other
// constraint arguments.
usage();
}
args[nArgs] = arg;
nArgs++;
UserJobIdArg = true;
}
}
if( ! (All || nArgs) ) {
// We got no indication of what to act on
fprintf( stderr, "You did not specify any jobs\n" );
usage();
}
if ( ConstraintArg && UserJobIdArg ) {
fprintf( stderr, "You can't use both -constraint and usernames or job ids\n" );
usage();
}
// Pick the default reason if the user didn't specify one
if( actionReason == NULL ) {
switch( mode ) {
case JA_RELEASE_JOBS:
actionReason = strdup("via condor_release");
break;
case JA_REMOVE_X_JOBS:
actionReason = strdup("via condor_rm -forcex");
break;
case JA_REMOVE_JOBS:
actionReason = strdup("via condor_rm");
break;
case JA_HOLD_JOBS:
actionReason = strdup("via condor_hold");
break;
case JA_SUSPEND_JOBS:
actionReason = strdup("via condor_suspend");
break;
case JA_CONTINUE_JOBS:
actionReason = strdup("via condor_continue");
break;
default:
actionReason = NULL;
}
}
// We're done parsing args, now make sure we know how to
// contact the schedd.
if( ! scheddAddr ) {
// This will always do the right thing, even if either or
// both of scheddName or pool are NULL.
schedd = new DCSchedd( scheddName, pool ? pool->addr() : NULL );
} else {
schedd = new DCSchedd( scheddAddr );
}
if( ! schedd->locate() ) {
fprintf( stderr, "%s: %s\n", MyName, schedd->error() );
exit( 1 );
}
// Special case for condor_rm -forcex: a configuration
// setting can disable this functionality. The real
// validation is done in the schedd, but we can catch
// the most common cases here and give a useful error
// message.
if(mode == JA_REMOVE_X_JOBS) {
if( mayUserForceRm() == false) {
fprintf( stderr, "Remove aborted. condor_rm -forcex has been disabled by the administrator.\n" );
exit( 1 );
}
}
// Process the args so we do the work.
if( All ) {
handleAll();
} else {
for(i = 0; i < nArgs; i++) {
if( match_prefix( args[i], "-constraint" ) ) {
i++;
addConstraint( args[i] );
} else {
procArg(args[i]);
}
}
}
// Deal with all the -constraint constraints
handleConstraints();
// Finally, do the actual work for all our args which weren't
// constraints...
if( job_ids ) {
CondorError errstack;
ClassAd* result_ad = doWorkByList( job_ids, &errstack );
if (had_error) {
fprintf( stderr, "%s\n", errstack.getFullText(true).c_str() );
}
printNewMessages( result_ad, job_ids );
delete( result_ad );
}
// If releasing jobs, and no errors happened, do a
// reschedule command now.
if ( mode == JA_RELEASE_JOBS && had_error == false ) {
Daemon my_schedd(DT_SCHEDD, NULL, NULL);
CondorError errstack;
if (!my_schedd.sendCommand(RESCHEDULE, Stream::safe_sock, 0, &errstack)) {
fprintf( stderr, "%s\n", errstack.getFullText(true).c_str() );
}
}
return had_error;
}