void
pipe_setup( char *cluster, char *proc, char * /*capability*/ )
{
static char host[1024];
UsePipes = TRUE;
dprintf( D_ALWAYS, "Job = %s.%s\n", cluster, proc );
if( Spool ) {
free( Spool );
}
Spool = param( "SPOOL" );
if( Spool == NULL ) {
EXCEPT( "Spool directory not specified in config file" );
}
snprintf( host, 1024, "%s", get_local_hostname().Value() );
ExecutingHost = host;
open_named_pipe( "/tmp/syscall_req", O_RDONLY, REQ_SOCK );
dprintf( D_ALWAYS, "Shadow: REQ_SOCK connected, fd = %d\n", REQ_SOCK );
open_named_pipe( "/tmp/syscall_rpl", O_WRONLY, RPL_SOCK );
dprintf( D_ALWAYS, "Shadow: RPL_SOCK connected, fd = %d\n", RPL_SOCK );
open_named_pipe( "/tmp/log", O_RDONLY, CLIENT_LOG );
dprintf( D_ALWAYS, "Shadow: CLIENT_LOG connected, fd = %d\n", CLIENT_LOG );
sock_RSC1 = RSC_ShadowInit( RSC_SOCK, CLIENT_LOG );
start_job( cluster, proc );
}
/*
* NOTE: suppress_lock must be held.
*/
static void
log_writer_last_msg_flush(LogWriter *self)
{
LogMessage *m;
LogPathOptions path_options = LOG_PATH_OPTIONS_INIT;
gchar buf[1024];
gssize len;
const gchar *p;
msg_debug("Suppress timer elapsed, emitting suppression summary",
NULL);
m = log_msg_new_empty();
m->timestamps[LM_TS_STAMP] = m->timestamps[LM_TS_RECVD];
m->pri = self->last_msg->pri;
m->flags = LF_INTERNAL | LF_LOCAL;
p = log_msg_get_value(self->last_msg, LM_V_HOST, &len);
log_msg_set_value(m, LM_V_HOST, p, len);
p = log_msg_get_value(self->last_msg, LM_V_PROGRAM, &len);
log_msg_set_value(m, LM_V_PROGRAM, p, len);
len = g_snprintf(buf, sizeof(buf), "Last message '%.20s' repeated %d times, suppressed by syslog-ng on %s",
log_msg_get_value(self->last_msg, LM_V_MESSAGE, NULL),
self->last_msg_count,
get_local_hostname(NULL));
log_msg_set_value(m, LM_V_MESSAGE, buf, len);
path_options.ack_needed = FALSE;
log_queue_push_tail(self->queue, m, &path_options);
log_writer_last_msg_release(self);
}
// Return our hostname in a static data buffer.
const char *
my_hostname()
{
// if( ! hostnames_initialized ) {
// init_hostnames();
// }
// return hostname;
static MyString __my_hostname;
__my_hostname = get_local_hostname();
return __my_hostname.Value();
}
void
Starter::exited(int status)
{
ClassAd *jobAd = NULL;
bool jobAdNeedsFree = true;
if (s_claim && s_claim->ad()) {
// real jobs in the startd have claims and job ads, boinc and perhaps others won't
jobAd = s_claim->ad();
jobAdNeedsFree = false;
} else {
// Dummy up an ad
int now = (int) time(0);
jobAd = new ClassAd();
jobAd->SetMyTypeName("Job");
jobAd->SetTargetTypeName("Machine");
jobAd->Assign(ATTR_CLUSTER_ID, now);
jobAd->Assign(ATTR_PROC_ID, 1);
jobAd->Assign(ATTR_OWNER, "boinc");
jobAd->Assign(ATTR_Q_DATE, (int)s_birthdate);
jobAd->Assign(ATTR_JOB_PRIO, 0);
jobAd->Assign(ATTR_IMAGE_SIZE, 0);
jobAd->Assign(ATTR_JOB_CMD, "boinc");
MyString gjid;
gjid.formatstr("%s#%d#%d#%d", get_local_hostname().Value(),
now, 1, now);
jobAd->Assign(ATTR_GLOBAL_JOB_ID, gjid);
}
// First, patch up the ad a little bit
jobAd->Assign(ATTR_COMPLETION_DATE, (int)time(0));
int runtime = time(0) - s_birthdate;
jobAd->Assign(ATTR_JOB_REMOTE_WALL_CLOCK, runtime);
int jobStatus = COMPLETED;
if (WIFSIGNALED(status)) {
jobStatus = REMOVED;
}
jobAd->Assign(ATTR_JOB_STATUS, jobStatus);
AppendHistory(jobAd);
WritePerJobHistoryFile(jobAd, true /* use gjid for filename*/);
if (jobAdNeedsFree) {
delete jobAd;
}
// Make sure our time isn't going to go off.
cancelKillTimer();
// Just for good measure, try to kill what's left of our whole
// pid family.
if (daemonCore->Kill_Family(s_pid) == FALSE) {
dprintf(D_ALWAYS,
"error killing process family of starter with pid %u\n",
s_pid);
}
// Now, delete any files lying around.
ASSERT( executeDir() );
cleanup_execute_dir( s_pid, executeDir() );
#if defined(LINUX)
if( param_boolean( "GLEXEC_STARTER", false ) ) {
cleanupAfterGlexec();
}
#endif
}
void
main_init(int argc, char* argv[])
{
char** ptr;
MyString job_queue_name;
int argc_count = 1;
for(ptr = argv + 1, argc_count = 1; argc_count<argc && *ptr; ptr++,argc_count++)
{
if(ptr[0][0] != '-')
{
usage(argv[0]);
}
switch(ptr[0][1])
{
case 'n':
if (Name) {
free(Name);
}
Name = build_valid_daemon_name( *(++ptr) );
break;
default:
usage(argv[0]);
}
}
// Tell Attrlist to publish the server time
AttrList_setPublishServerTime( true );
// Initialize DaemonCore's use of ProcFamily. We do this so that we
// launch a ProcD if necessary so that any Starters that we launch
// for Local Universe jobs can share a single ProcD, instead of
// each creating their own
daemonCore->Proc_Family_Init();
#if defined(HAVE_DLOPEN)
ClassAdLogPluginManager::Load();
ScheddPluginManager::Load();
ScheddPluginManager::EarlyInitialize();
ClassAdLogPluginManager::EarlyInitialize();
#endif
/* schedd doesn't care about other daemons. only that it has the ability
* to run jobs. so the following code is for now not needed.
// ZKM HACK TO MAKE SURE SCHEDD HAS USER CREDENTIALS
//
// if we are using the credd and credmon, we need to init them before
// doing anything!
char* p = param("SEC_CREDENTIAL_DIRECTORY");
if(p) {
free(p);
dprintf(D_ALWAYS, "SCHEDD: INITIALIZING USER CREDS\n");
Daemon *my_credd;
// we will abort if we can't locate the credd, so let's try a
// few times. locate() caches the result so we have to destroy
// the object and make a new one each time.
int retries = 20;
bool success = false;
do {
// allocate a credd
my_credd = new Daemon(DT_CREDD);
if(my_credd) {
// call locate
bool loc_rc = my_credd->locate();
if(loc_rc) {
// get a connected relisock
CondorError errstack;
ReliSock* r = (ReliSock*)my_credd->startCommand(
CREDD_REFRESH_ALL, Stream::reli_sock, 20, &errstack);
if ( r ) {
// ask the credd to get us some fresh user creds
ClassAd ad;
putClassAd(r, ad);
r->end_of_message();
r->decode();
getClassAd(r, ad);
r->end_of_message();
dprintf(D_SECURITY | D_FULLDEBUG, "SCHEDD: received ad from CREDD:\n");
dPrintAd(D_SECURITY | D_FULLDEBUG, ad);
MyString result;
ad.LookupString("Result", result);
if(result == "success") {
success = true;
} else {
dprintf(D_FULLDEBUG, "SCHEDD: warning, creddmon returned failure.\n");
}
// clean up.
delete r;
} else {
dprintf(D_FULLDEBUG, "SCHEDD: warning, startCommand failed, %s\n", errstack.getFullText(true).c_str());
}
} else {
dprintf(D_FULLDEBUG, "SCHEDD: warning, locate failed.\n");
}
// clean up.
delete my_credd;
} else {
dprintf(D_FULLDEBUG, "SCHEDD: warning, new Daemon(DT_CREDD) failed.\n");
}
// if something went wrong, sleep and retry (finit number of times)
if(!success) {
dprintf(D_FULLDEBUG, "SCHEDD: sleeping and trying again %i times.\n", retries);
sleep(1);
retries--;
}
} while ((retries > 0) && (success == false));
// except if fail
if (!success) {
EXCEPT("FAILED TO INITIALIZE USER CREDS (locate failed)");
}
}
// END ZKM HACK
*/
#ifndef WIN32
// if using the SEC_CREDENTIAL_DIRECTORY, confirm we are "up-to-date".
// at the moment, we take an "all-or-nothing" approach. ultimately, this
// should be per-user, and the SchedD should start normally and run jobs
// for users who DO have valid credentials, and simply holding on to jobs
// in idle state for users who do NOT have valid credentials.
//
char* p = param("SEC_CREDENTIAL_DIRECTORY");
if(p) {
free(p);
bool success = false;
int retries = 60;
do {
// look for existence of file that says everything is up-to-date.
success = credmon_poll(NULL, false, false);
if(!success) {
dprintf(D_ALWAYS, "SCHEDD: User credentials not up-to-date. Start-up delayed. Waiting 10 seconds and trying %i more times.\n", retries);
sleep(10);
retries--;
}
} while ((!success) && (retries > 0));
// we tried, we give up.
if(!success) {
EXCEPT("User credentials unavailable after 10 minutes");
}
}
// User creds good to go, let's start this thing up!
#endif // WIN32
// Initialize all the modules
scheduler.Init();
scheduler.Register();
// Initialize the job queue
char *job_queue_param_name = param("JOB_QUEUE_LOG");
if (job_queue_param_name == NULL) {
// the default place for the job_queue.log is in spool
job_queue_name.formatstr( "%s/job_queue.log", Spool);
} else {
job_queue_name = job_queue_param_name; // convert char * to MyString
free(job_queue_param_name);
}
// Make a backup of the job queue?
if ( param_boolean_crufty("SCHEDD_BACKUP_SPOOL", false) ) {
MyString hostname;
hostname = get_local_hostname();
MyString job_queue_backup;
job_queue_backup.formatstr( "%s/job_queue.bak.%s.%ld",
Spool, hostname.Value(), (long)time(NULL) );
if ( copy_file( job_queue_name.Value(), job_queue_backup.Value() ) ) {
dprintf( D_ALWAYS, "Failed to backup spool to '%s'\n",
job_queue_backup.Value() );
} else {
dprintf( D_FULLDEBUG, "Spool backed up to '%s'\n",
job_queue_backup.Value() );
}
}
int max_historical_logs = param_integer( "MAX_JOB_QUEUE_LOG_ROTATIONS", DEFAULT_MAX_JOB_QUEUE_LOG_ROTATIONS );
InitJobQueue(job_queue_name.Value(),max_historical_logs);
PostInitJobQueue();
// Initialize the dedicated scheduler stuff
dedicated_scheduler.initialize();
// Do a timeout now at startup to get the ball rolling...
scheduler.timeout();
#if defined(HAVE_DLOPEN)
ScheddPluginManager::Initialize();
ClassAdLogPluginManager::Initialize();
#endif
daemonCore->InstallAuditingCallback( AuditLogNewConnection );
}
gboolean
log_macro_expand(GString *result, gint id, gboolean escape, LogTemplateOptions *opts, gint tz, gint32 seq_num, const gchar *context_id, LogMessage *msg)
{
static LogTemplateOptions default_opts = { TRUE, TS_FMT_BSD, 0, { NULL, NULL }, { NULL, NULL } };
if (!opts)
opts = &default_opts;
switch (id)
{
case M_FACILITY:
{
/* facility */
const char *n;
n = syslog_name_lookup_name_by_value(msg->pri & LOG_FACMASK, sl_facilities);
if (n)
{
g_string_append(result, n);
}
else
{
format_uint32_padded(result, 0, 0, 16, (msg->pri & LOG_FACMASK) >> 3);
}
break;
}
case M_FACILITY_NUM:
{
format_uint32_padded(result, 0, 0, 10, (msg->pri & LOG_FACMASK) >> 3);
break;
}
case M_LEVEL:
{
/* level */
const char *n;
n = syslog_name_lookup_name_by_value(msg->pri & LOG_PRIMASK, sl_levels);
if (n)
{
g_string_append(result, n);
}
else
{
format_uint32_padded(result, 0, 0, 10, msg->pri & LOG_PRIMASK);
}
break;
}
case M_LEVEL_NUM:
{
format_uint32_padded(result, 0, 0, 10, msg->pri & LOG_PRIMASK);
break;
}
case M_TAG:
{
format_uint32_padded(result, 2, '0', 16, msg->pri);
break;
}
case M_TAGS:
{
log_msg_print_tags(msg, result);
break;
}
case M_BSDTAG:
{
format_uint32_padded(result, 0, 0, 10, (msg->pri & LOG_PRIMASK));
g_string_append_c(result, (((msg->pri & LOG_FACMASK) >> 3) + 'A'));
break;
}
case M_PRI:
{
format_uint32_padded(result, 0, 0, 10, msg->pri);
break;
}
case M_HOST:
{
if (msg->flags & LF_CHAINED_HOSTNAME)
{
/* host */
const gchar *p1, *p2;
int remaining, length;
gssize host_len;
const gchar *host = log_msg_get_value(msg, LM_V_HOST, &host_len);
p1 = memchr(host, '@', host_len);
if (p1)
p1++;
else
p1 = host;
remaining = host_len - (p1 - host);
p2 = memchr(p1, '/', remaining);
length = p2 ? p2 - p1
: host_len - (p1 - host);
result_append(result, p1, length, escape);
}
else
{
result_append_value(result, msg, LM_V_HOST, escape);
}
break;
}
case M_SDATA:
if (escape)
{
GString *sdstr = g_string_sized_new(0);
log_msg_append_format_sdata(msg, sdstr, seq_num);
result_append(result, sdstr->str, sdstr->len, TRUE);
g_string_free(sdstr, TRUE);
}
else
{
log_msg_append_format_sdata(msg, result, seq_num);
}
break;
case M_MSGHDR:
if ((msg->flags & LF_LEGACY_MSGHDR))
{
/* fast path for now, as most messages come from legacy devices */
result_append_value(result, msg, LM_V_LEGACY_MSGHDR, escape);
}
else
{
/* message, complete with program name and pid */
gssize len;
len = result->len;
result_append_value(result, msg, LM_V_PROGRAM, escape);
if (len != result->len)
{
const gchar *pid = log_msg_get_value(msg, LM_V_PID, &len);
if (len > 0)
{
result_append(result, "[", 1, FALSE);
result_append(result, pid, len, escape);
result_append(result, "]", 1, FALSE);
}
result_append(result, ": ", 2, FALSE);
}
}
break;
case M_MESSAGE:
if (cfg_is_config_version_older(configuration, 0x0300))
log_macro_expand(result, M_MSGHDR, escape, opts, tz, seq_num, context_id, msg);
result_append_value(result, msg, LM_V_MESSAGE, escape);
break;
case M_SOURCE_IP:
{
gchar *ip;
if (msg->saddr && (g_sockaddr_inet_check(msg->saddr) ||
#if ENABLE_IPV6
g_sockaddr_inet6_check(msg->saddr))
#else
0)
#endif
)
{
gchar buf[MAX_SOCKADDR_STRING];
g_sockaddr_format(msg->saddr, buf, sizeof(buf), GSA_ADDRESS_ONLY);
ip = buf;
}
else
{
ip = "127.0.0.1";
}
result_append(result, ip, strlen(ip), escape);
break;
}
case M_SEQNUM:
{
if (seq_num)
{
format_uint32_padded(result, 0, 0, 10, seq_num);
}
break;
}
case M_CONTEXT_ID:
{
if (context_id)
{
result_append(result, context_id, strlen(context_id), escape);
}
break;
}
case M_LOGHOST:
{
gsize hname_len;
const gchar *hname = get_local_hostname(&hname_len);
result_append(result, hname, hname_len, escape);
break;
}
case M_SYSUPTIME:
{
GTimeVal ct;
g_get_current_time(&ct);
format_uint64_padded(result, 0, 0, 10, g_time_val_diff(&ct, &app_uptime) / 1000 / 10);
break;
}
default:
{
/* year, month, day */
struct tm *tm, tm_storage;
gchar buf[64];
gint length;
time_t t;
LogStamp *stamp, sstamp;
glong zone_ofs;
guint tmp_hour;
if (id >= M_TIME_FIRST && id <= M_TIME_LAST)
{
stamp = &msg->timestamps[LM_TS_STAMP];
}
else if (id >= M_TIME_FIRST + M_RECVD_OFS && id <= M_TIME_LAST + M_RECVD_OFS)
{
id -= M_RECVD_OFS;
stamp = &msg->timestamps[LM_TS_RECVD];
}
else if (id >= M_TIME_FIRST + M_STAMP_OFS && id <= M_TIME_LAST + M_STAMP_OFS)
{
id -= M_STAMP_OFS;
stamp = &msg->timestamps[LM_TS_STAMP];
}
else if (id >= M_TIME_FIRST + M_CSTAMP_OFS && id <= M_TIME_LAST + M_CSTAMP_OFS)
{
GTimeVal tv;
id -= M_CSTAMP_OFS;
cached_g_current_time(&tv);
sstamp.tv_sec = tv.tv_sec;
sstamp.tv_usec = tv.tv_usec;
sstamp.zone_offset = -1;
stamp = &sstamp;
}
else
{
g_assert_not_reached();
break;
}
/* try to use the following zone values in order:
* destination specific timezone, if one is specified
* message specific timezone, if one is specified
* local timezone
*/
zone_ofs = (opts->time_zone_info[tz] != NULL ? time_zone_info_get_offset(opts->time_zone_info[tz], stamp->tv_sec) : stamp->zone_offset);
if (zone_ofs == -1)
zone_ofs = stamp->zone_offset;
t = stamp->tv_sec + zone_ofs;
cached_gmtime(&t, &tm_storage);
tm = &tm_storage;
switch (id)
{
case M_WEEK_DAY_ABBREV:
g_string_append_len(result, weekday_names_abbrev[tm->tm_wday], 3);
break;
case M_WEEK_DAY_NAME:
g_string_append(result, weekday_names[tm->tm_wday]);
break;
case M_WEEK_DAY:
format_uint32_padded(result, 0, 0, 10, tm->tm_wday + 1);
break;
case M_WEEK:
format_uint32_padded(result, 2, '0', 10, (tm->tm_yday - (tm->tm_wday - 1 + 7) % 7 + 7) / 7);
break;
case M_YEAR:
format_uint32_padded(result, 4, '0', 10, tm->tm_year + 1900);
break;
case M_YEAR_DAY:
format_uint32_padded(result, 3, '0', 10, tm->tm_yday + 1);
break;
case M_MONTH:
format_uint32_padded(result, 2, '0', 10, tm->tm_mon + 1);
break;
case M_MONTH_WEEK:
format_uint32_padded(result, 0, 0, 10, ((tm->tm_mday / 7) + ((tm->tm_wday > 0) && ((tm->tm_mday % 7) >= tm->tm_wday))));
break;
case M_MONTH_ABBREV:
g_string_append_len(result, month_names_abbrev[tm->tm_mon], 3);
break;
case M_MONTH_NAME:
g_string_append(result, month_names[tm->tm_mon]);
break;
case M_DAY:
format_uint32_padded(result, 2, '0', 10, tm->tm_mday);
break;
case M_HOUR:
format_uint32_padded(result, 2, '0', 10, tm->tm_hour);
break;
case M_HOUR12:
if (tm->tm_hour < 12)
tmp_hour = tm->tm_hour;
else
tmp_hour = tm->tm_hour - 12;
if (tmp_hour == 0)
tmp_hour = 12;
format_uint32_padded(result, 2, '0', 10, tmp_hour);
break;
case M_MIN:
format_uint32_padded(result, 2, '0', 10, tm->tm_min);
break;
case M_SEC:
format_uint32_padded(result, 2, '0', 10, tm->tm_sec);
break;
case M_MSEC:
format_uint32_padded(result, 3, '0', 10, stamp->tv_usec/1000);
break;
case M_USEC:
format_uint32_padded(result, 6, '0', 10, stamp->tv_usec);
break;
case M_AMPM:
g_string_append(result, tm->tm_hour < 12 ? "AM" : "PM");
break;
case M_DATE:
case M_STAMP:
case M_ISODATE:
case M_FULLDATE:
case M_UNIXTIME:
{
gint format = id == M_DATE ? TS_FMT_BSD :
id == M_ISODATE ? TS_FMT_ISO :
id == M_FULLDATE ? TS_FMT_FULL :
id == M_UNIXTIME ? TS_FMT_UNIX :
opts->ts_format;
log_stamp_append_format(stamp, result, format, zone_ofs, opts->frac_digits);
break;
}
case M_TZ:
case M_TZOFFSET:
length = format_zone_info(buf, sizeof(buf), zone_ofs);
g_string_append_len(result, buf, length);
break;
}
break;
}
}
void
main_init(int argc, char* argv[])
{
char** ptr;
MyString job_queue_name;
int argc_count = 1;
for(ptr = argv + 1, argc_count = 1; argc_count<argc && *ptr; ptr++,argc_count++)
{
if(ptr[0][0] != '-')
{
usage(argv[0]);
}
switch(ptr[0][1])
{
case 'n':
Name = build_valid_daemon_name( *(++ptr) );
break;
default:
usage(argv[0]);
}
}
// Tell Attrlist to publish the server time
AttrList_setPublishServerTime( true );
// Initialize DaemonCore's use of ProcFamily. We do this so that we
// launch a ProcD if necessary so that any Starters that we launch
// for Local Universe jobs can share a single ProcD, instead of
// each creating their own
daemonCore->Proc_Family_Init();
#if defined(WANT_CONTRIB) && defined(WITH_MANAGEMENT)
#if defined(HAVE_DLOPEN)
// Intialization of the plugin manager, i.e. loading all
// plugins, should be performed before the job queue log is
// read so plugins have a chance to learn about all jobs
// already in the queue
ClassAdLogPluginManager::Load();
// Load all ScheddPlugins. In reality this doesn't do much
// since initializing any plugin manager loads plugins for all
// plugin manager.
ScheddPluginManager::Load();
// Tell all ScheddPlugins to initialze themselves
ScheddPluginManager::EarlyInitialize();
// Tell all plugins to initialize themselves
ClassAdLogPluginManager::EarlyInitialize();
#endif
#endif
// Initialize all the modules
scheduler.Init();
scheduler.Register();
// Initialize the job queue
char *job_queue_param_name = param("JOB_QUEUE_LOG");
if (job_queue_param_name == NULL) {
// the default place for the job_queue.log is in spool
job_queue_name.sprintf( "%s/job_queue.log", Spool);
} else {
job_queue_name = job_queue_param_name; // convert char * to MyString
free(job_queue_param_name);
}
// Make a backup of the job queue?
if ( param_boolean_crufty("SCHEDD_BACKUP_SPOOL", false) ) {
MyString hostname;
UtcTime now(true);
hostname = get_local_hostname();
MyString job_queue_backup;
job_queue_backup.sprintf( "%s/job_queue.bak.%s.%ld",
Spool, hostname.Value(), now.seconds() );
if ( copy_file( job_queue_name.Value(), job_queue_backup.Value() ) ) {
dprintf( D_ALWAYS, "Failed to backup spool to '%s'\n",
job_queue_backup.Value() );
} else {
dprintf( D_FULLDEBUG, "Spool backed up to '%s'\n",
job_queue_backup.Value() );
}
}
int max_historical_logs = param_integer( "MAX_JOB_QUEUE_LOG_ROTATIONS", DEFAULT_MAX_JOB_QUEUE_LOG_ROTATIONS );
InitJobQueue(job_queue_name.Value(),max_historical_logs);
mark_jobs_idle();
// The below must happen _after_ InitJobQueue is called.
if ( scheduler.autocluster.config() ) {
// clear out auto cluster id attributes
WalkJobQueue( (int(*)(ClassAd *))clear_autocluster_id );
}
//
// Update the SchedDInterval attributes in jobs if they
// have it defined. This will be for JobDeferral and
// CronTab jobs
//
WalkJobQueue( (int(*)(ClassAd *))::updateSchedDInterval );
// Initialize the dedicated scheduler stuff
dedicated_scheduler.initialize();
// Do a timeout now at startup to get the ball rolling...
scheduler.timeout();
#if defined(WANT_CONTRIB) && defined(WITH_MANAGEMENT)
#if defined(HAVE_DLOPEN)
// Tell all ScheddPlugins to initialze themselves
ScheddPluginManager::Initialize();
// Tell all plugins to initialize themselves
ClassAdLogPluginManager::Initialize();
#endif
#endif
}
int main(int argc, char *argv[])
{
configuration_t configuration;
int i, j;
struct timeval tm;
struct timezone tz;
measurement_t *measurement;
struct timeval next, wait;
int subject_id, flow_id;
unsigned long long packets, bytes;
double mbps;
char command[MAX_COMMAND+1];
char hostname_interface[MAX_HOSTNAME_INTERFACE+1]; /* DiMAPI connect string
as "hostname:interface,..." */
struct timeval tv_start, tv_stop; /* to measure how fast mapi_read_result()
responds */
int tv_diff_pkt, tv_diff_byte; /* time used by mapi_read_results() */
int tv_diff_threshold; /* 1 if threshold was reached */
mapi_results_t *pkt_counter_res;
mapi_results_t *byte_counter_res;
unsigned long long pkt_counter;
unsigned long long byte_counter;
int scope_size;
double pkt_sec; /* seconds from previous packet result */
double byte_sec; /* seconds from previous byte result */
mapi_flow_info_t info;
mapi_device_info_t dinfo;
openlog("abw", LOG_PID, LOG_LOCAL0);
syslog(LOG_DEBUG, "starting abw");
memset((void *)&configuration, 0, (size_t)(sizeof(configuration)));
/* Create global configuration */
if ((configuration.global=malloc(sizeof(global_t)))==NULL) {
fprintf(stderr, "%s: malloc() failed\n", __func__);
return -1;
}
memset(configuration.global, 0, sizeof(global_t));
/* Create first subject, scope, parameters and measurement so that they
can be filled-in by command-line options */
/* if ((configuration.subject=new_subject())==NULL) {
fprintf(stderr, "%s: new_subject() failed\n", __func__);
return -1;
}
if ((configuration.scope=new_scope())==NULL) {
fprintf(stderr, "%s: new_subject() failed\n", __func__);
return -1;
}
if ((configuration.parameters=new_parameters())==NULL) {
fprintf(stderr, "%s: new_parameters() failed\n", __func__);
return -1;
}
if ((configuration.measurement=new_measurement())==NULL) {
fprintf(stderr, "%s: new_measurement() failed\n", __func__);
return -1;
} */
/* Read command line */
if (read_command_line(argc, argv, &configuration)<0) {
fprintf(stderr, "%s: read_command_line() failed\n", __func__);
return -1;
}
/* Read configuration file */
if (configuration.global->conf_filename) {
if (read_conf_file(&configuration)<0) {
fprintf(stderr, "%s: read_conf_file() failed\n", __func__);
return -1;
}
}
/* Fill-in local hostname */
if (get_local_hostname(&(configuration.global->hostname))<0) {
fprintf(stderr, "%s: get_local_hostname() failed\n", __func__);
return -1;
}
/* Check if specified values are within acceptable limits */
if (check_conf(&configuration)<0) {
fprintf(stderr, "%s: check_conf() failed\n", __func__);
exit(-1);
}
/* Print configuration */
if (debug)
print_conf(&configuration);
if (daemonize) {
printf("Switching to daemon\n");
if (continue_as_daemon()<0) {
fprintf(stderr, "%s: continue_as_daemon() failed\n", __func__);
return -1;
}
printf("Continuing as daemon\n");
}
/*
* Create RRD files
*/
/* Go over all measurements */
measurement=configuration.measurement;
while (measurement) {
int parameters_id;
char *filename;
parameters_id = measurement->parameters_id;
/* Go over all protocols */
j=0;
while (protocols[j].protocol) {
if ((filename=
abw_rrd_create_filename(measurement->scope,
parameters_id, protocols[j].protocol))==NULL) {
fprintf(stderr, "%s: rrd_create_filename() failed\n",
__func__);
return -1;
}
if (abw_rrd_create_file(filename)<0) {
fprintf(stderr, "%s: abw_rrd_create_file() failed\n", __func__);
return -1;
}
j++;
} /* Go over all protocols */
/* Go over all tracked protocols */
j=0;
while (tracked_protocols[j].protocol) {
if ((filename=
abw_rrd_create_filename(measurement->scope,
parameters_id, tracked_protocols[j].protocol))==NULL) {
fprintf(stderr, "%s: rrd_create_filename() failed\n",
__func__);
return -1;
}
if (abw_rrd_create_file(filename)<0) {
fprintf(stderr, "%s: abw_rrd_create_file() failed\n", __func__);
return -1;
}
j++;
} /* Go over all tracked protocols */
/* Create RRD file for "all" protocol (all traffic together) */
if ((filename=
abw_rrd_create_filename(measurement->scope,
parameters_id, "all"))==NULL) {
fprintf(stderr, "%s: rrd_create_filename() failed\n",
__func__);
return -1;
}
if (abw_rrd_create_file(filename)<0) {
fprintf(stderr, "%s: abw_rrd_create_file() failed\n", __func__);
return -1;
}
measurement=measurement->next;
} /* while (measurement) */
/*
* Create MAPI flows
*/
flow_id=0;
/* Go over all measurements */
measurement=configuration.measurement;
while (measurement) {
/* Go over all monitored protocols */
i=0;
while (measurement->protocols_array[i] && i<MAX_PROTOCOLS) {
int parameters_id;
char *protocol;
/* Create data structure to maintain MAPI information */
if (flow_id>=MAX_FLOWS) {
fprintf(stderr, "%s: more than %d flows requested\n", __func__,
MAX_FLOWS);
return -1;
}
if ((flow[flow_id]=new_flow())==NULL) {
fprintf(stderr, "%s: new_flow() failed\n", __func__);
return -1;
}
flow[flow_id]->measurement=measurement;
flow[flow_id]->protocol=measurement->protocols_array[i];
parameters_id = measurement->parameters_id;
protocol = measurement->protocols_array[i];
if ((flow[flow_id]->rrd_filename=
abw_rrd_create_filename(measurement->scope,
parameters_id, protocol))==NULL) {
fprintf(stderr, "%s: rrd_create_filename() failed\n",
__func__);
return -1;
}
/*
* If scope has only one subject and if hostname is "localhost" or
* equal to local hostname, then use MAPI connect string (not DiMAPI)
*/
if (!(measurement->scope->subject[1]) &&
(!strcmp(measurement->scope->subject[0]->hostname, "localhost") ||
!strcmp(measurement->scope->subject[0]->hostname,
configuration.global->hostname)))
strcpy(hostname_interface,
measurement->scope->subject[0]->interface);
/*
* Prepare DiMAPI connect string as hostname:interface, ...
*/
else {
j=0; hostname_interface[0]='\0';
while (measurement->scope->subject[j] && j<MAX_SUBJECTS) {
/* Append comma "," */
if (hostname_interface[0]) {
if (strlen(hostname_interface)+1>=MAX_HOSTNAME_INTERFACE) {
fprintf(stderr, "%s: DiMAPI connect string is longer than %d characters\n", __func__, MAX_HOSTNAME_INTERFACE);
return -1;
}
strcat(hostname_interface, ",");
}
/* Append next hostname:interface */
if (strlen(hostname_interface) +
strlen(measurement->scope->subject[j]->hostname) +
strlen(measurement->scope->subject[j]->interface)
>= MAX_HOSTNAME_INTERFACE) {
fprintf(stderr, "%s: DiMAPI connect string is longer than %d characters\n", __func__, MAX_HOSTNAME_INTERFACE);
return -1;
}
sprintf(hostname_interface + strlen(hostname_interface), "%s:%s",
measurement->scope->subject[j]->hostname,
measurement->scope->subject[j]->interface);
j++;
} /* while (measurement->scope->subject[j] && j<MAX_SUBJECTS) */
} /* Creating DiMAPI connect string */
/* Create a new MAPI flow */
if (debug)
printf("%s: mapi_create_flow(%s)\n", __func__, hostname_interface);
if ((flow[flow_id]->fd=mapi_create_flow(hostname_interface))<0) {
fprintf(stderr, "%s: mapi_create_flow(%s) failed\n", __func__,
hostname_interface);
fprintf(stderr, "%s: Do you run mapid daemon on the machine where you connect to?\n", __func__);
fprintf(stderr, "%s: Do you run mapicommd daemon on the machine where you connect to? (if you are connecting to a non-local machine or to multiple machines)\n", __func__);
return -1;
}
/* If this is a MAPI flow (not DiMAPI flow), then set MPLS and VLAN
flags according to mapi.conf. Otherwise the flags were set in
abw.conf */
if (!strchr(hostname_interface, ':')) {
if (debug)
printf("%s: MAPI flow on \"%s\", setting MPLS and VLAN flags from mapi.conf\n", __func__, hostname_interface);
if ((mapi_get_flow_info(flow[flow_id]->fd, &info)) < 0){
fprintf(stderr, "%s: mapi_get_flow_info() failed\n", __func__);
return -1;
}
if ((mapi_get_device_info(info.devid, &dinfo)) < 0) {
fprintf(stderr, "%s: mapi_get_device_info() failed\n", __func__);
return -1;
}
measurement->scope->mpls = dinfo.mpls;
measurement->scope->vlan = dinfo.vlan;
}
else
if (debug)
printf("%s: DiMAPI flow on \"%s\", setting MPLS and VLAN flags from abw.conf\n", __func__, hostname_interface);
/* Prepare header filter for this protocol */
if ((flow[flow_id]->tracked_protocol=
protocol_filter(measurement->parameters->header_filter,
flow[flow_id]->protocol, measurement->scope->mpls,
measurement->scope->vlan,
&(flow[flow_id]->header_filter)))<0) {
fprintf(stderr, "%s: protocol_filter() failed\n", __func__);
return -1;
}
if (debug)
printf("measurement->parameters->header_filter: %s, flow[flow_id]->protocol: %s, flow[flow_id]->header_filter: %s, track_function: %s\n", (measurement->parameters->header_filter)?measurement->parameters->header_filter:"NULL", flow[flow_id]->protocol, (flow[flow_id]->header_filter)?flow[flow_id]->header_filter:"NULL", (flow[flow_id]->tracked_protocol)?tracked_protocols[flow[flow_id]->tracked_protocol-1].track_function:"none");
/* Filter based on input port, we can use port number in the first
subject of the scope, because all subjects in a scope must have
the same port number */
if (measurement->scope->subject[0]->port >= 0) {
if ((flow[flow_id]->interface_fid=mapi_apply_function(flow[flow_id]->fd, "INTERFACE", measurement->scope->subject[0]->port))<0) {
fprintf(stderr, "%s: INTERFACE failed\n", __func__);
return -1;
}
}
/* Note that BPF_FILTER uses compiled header filter that
selects packets of the given protocol */
/* BPF_FILTER is applied if a) header_filter was specified in
[parameters] section or b) protocol other than "all" and other than
some that requires tracking was specified in [parameters] section or
c) MPLS is used on links in this [scope] */
if (flow[flow_id]->header_filter) {
if (debug)
printf("%s: mapi_apply_function(%d, BPF_FILTER, \"%s\")\n",
__func__, flow[flow_id]->fd, flow[flow_id]->header_filter);
if ((flow[flow_id]->bpf_filter_fid=
mapi_apply_function(flow[flow_id]->fd, "BPF_FILTER",
flow[flow_id]->header_filter))<0) {
fprintf(stderr, "%s: BPF_FILTER (\"%s\") failed\n",
__func__, flow[flow_id]->header_filter);
return -1;
}
}
/* Track application protocol, BPF_FILTER could have been applied
before */
if (flow[flow_id]->tracked_protocol) {
if (debug)
printf("%s: mapi_apply_function(%d, %s)\n", __func__,
flow[flow_id]->fd,
tracked_protocols[flow[flow_id]->tracked_protocol-1].
track_function);
if ((flow[flow_id]->track_function_fid=
mapi_apply_function(flow[flow_id]->fd,
tracked_protocols[flow[flow_id]->tracked_protocol-1].
track_function))<0) {
fprintf(stderr, "%s: tracking (%s) failed\n", __func__,
tracked_protocols[flow[flow_id]->tracked_protocol-1].
track_function);
return -1;
}
}
/* Sampling */
if (measurement->parameters->sau_mode == 'd' &&
(unsigned int)(measurement->parameters->sau_threshold) != 1) {
if ((flow[flow_id]->sample_fid=
mapi_apply_function(flow[flow_id]->fd, "SAMPLE",
measurement->parameters->sau_threshold, PERIODIC))<0) {
fprintf(stderr, "%s: SAMPLE (PERIODIC, %.02f) failed\n",
__func__, measurement->parameters->sau_threshold);
return -1;
}
}
else if (measurement->parameters->sau_mode == 'p' &&
(unsigned int)(measurement->parameters->sau_threshold) != 1) {
if ((flow[flow_id]->sample_fid=
mapi_apply_function(flow[flow_id]->fd, "SAMPLE",
(measurement->parameters->sau_threshold)*100,
PROBABILISTIC))<0) {
fprintf(stderr, "%s: SAMPLE (PROBABILISTIC, %.02f) failed\n",
__func__, (measurement->parameters->sau_threshold)*100);
return -1;
}
}
/* Payload searching */
if (measurement->parameters->payload_strings[0]) {
if ((flow[flow_id]->str_search_fid=
mapi_apply_function(flow[flow_id]->fd, "STR_SEARCH",
measurement->parameters->payload_strings[0], 0, 0))<0) {
fprintf(stderr, "%s: STR_SEARCH (%s) failed\n",
__func__, measurement->parameters->payload_strings[0]);
return -1;
}
}
/* Counting packets and bytes */
if ((flow[flow_id]->pkt_counter_fid=
mapi_apply_function(flow[flow_id]->fd, "PKT_COUNTER"))<0) {
fprintf(stderr, "%s: PKT_COUNTER failed\n", __func__);
return -1;
}
/* Simultaneous use of PKT_COUNTER and BYTE_COUNTER does not
work with DAG4.3GE. Temporary hack: always use stflib version */
if ((flow[flow_id]->byte_counter_fid=
mapi_apply_function(flow[flow_id]->fd, "stdflib:BYTE_COUNTER"))<0) {
fprintf(stderr, "%s: BYTE_COUNTER failed\n",
__func__);
return -1;
}
/* Connect to flow */
if (!configuration.global->no_measure) {
if (mapi_connect(flow[flow_id]->fd)<0) {
fprintf(stderr, "%s: mapi_connect() (%s) failed\n",
__func__, hostname_interface);
return -1;
}
if ((scope_size=mapi_get_scope_size(flow[flow_id]->fd)) !=
flow[flow_id]->measurement->scope->subject_no) {
fprintf(stderr, "%s: mapi_get_scope_size() returned %d for %d subjects\n", __func__, scope_size, flow[flow_id]->measurement->scope->subject_no);
return -1;
}
}
i++;
flow_id++;
} /* while (measurement->protocols_array[i] && i<MAX_PROTOCOLS) */
measurement=measurement->next;
} /* while (measurement) */
if (configuration.global->no_measure || !configuration.measurement)
return 0;
/* Periodically get results from MAPI flows */
while (1) {
if (gettimeofday(&tm, &tz)<0) {
fprintf(stderr, "%s: gettimeofday() failed\n", __func__);
return -1;
}
flow_id=0;
while (flow[flow_id] && flow_id<MAX_FLOWS) {
int scope_packets, scope_bytes;
if (!configuration.global->no_stdout) {
printf("%d %u.%u", flow[flow_id]->measurement->scope->id,
(unsigned int)(tm.tv_sec),
(unsigned int)(tm.tv_usec));
if (!configuration.global->stdout_simple)
printf(" %s\n", flow[flow_id]->protocol);
}
gettimeofday(&tv_start, NULL);
if ((pkt_counter_res=mapi_read_results(flow[flow_id]->fd,
flow[flow_id]->pkt_counter_fid))==NULL) {
fprintf(stderr, "%s: mapi_read_results() for flow %d failed\n",
__func__, flow_id);
return -1;
}
gettimeofday(&tv_stop, NULL);
tv_diff_pkt=timestamp_diff(&tv_start, &tv_stop);
gettimeofday(&tv_start, NULL);
if ((byte_counter_res=mapi_read_results(flow[flow_id]->fd,
flow[flow_id]->byte_counter_fid))==NULL) {
fprintf(stderr, "%s: mapi_read_results() for flow %d failed\n",
__func__, flow_id);
return -1;
}
gettimeofday(&tv_stop, NULL);
tv_diff_byte=timestamp_diff(&tv_start, &tv_stop);
if (tv_diff_pkt>=TV_DIFF_THRESHOLD ||
tv_diff_byte>=TV_DIFF_THRESHOLD)
tv_diff_threshold=1;
else
tv_diff_threshold=0;
if (tv_diff_pkt>=TV_DIFF_THRESHOLD)
syslog(LOG_DEBUG, "mapi_read_result() for PKT_COUNTER takes %d us for measurement ID %d and protocol %s (threshold %d us reached)", tv_diff_pkt, flow[flow_id]->measurement->id, flow[flow_id]->protocol, TV_DIFF_THRESHOLD);
if (tv_diff_byte>=TV_DIFF_THRESHOLD)
syslog(LOG_DEBUG, "mapi_read_result() for BYTE_COUNTER takes %d us for measurement ID %d and protocol %s (threshold %d us reached)", tv_diff_byte, flow[flow_id]->measurement->id, flow[flow_id]->protocol, TV_DIFF_THRESHOLD);
scope_size = flow[flow_id]->measurement->scope->subject_no;
scope_packets=0;
scope_bytes=0;
for (subject_id=0; subject_id<scope_size; subject_id++) {
pkt_counter=
*((unsigned long long*)(pkt_counter_res[subject_id].res));
byte_counter=
*((unsigned long long*)(byte_counter_res[subject_id].res));
packets=pkt_counter - flow[flow_id]->pkt_counter[subject_id];
bytes=byte_counter - flow[flow_id]->byte_counter[subject_id];
mbps=(double)bytes*8/1000000;
flow[flow_id]->pkt_counter[subject_id]=pkt_counter;
flow[flow_id]->byte_counter[subject_id]=byte_counter;
/* Determine seconds from previous result */
if (flow[flow_id]->pkt_ts[subject_id])
pkt_sec=(double)(pkt_counter_res[subject_id].ts -
flow[flow_id]->pkt_ts[subject_id])/1000000;
else
pkt_sec=
flow[flow_id]->measurement->parameters->interval.tv_sec +
(double)(flow[flow_id]->measurement->parameters->interval.tv_usec)/1000000;
if (flow[flow_id]->byte_ts[subject_id])
byte_sec=(double)(byte_counter_res[subject_id].ts -
flow[flow_id]->byte_ts[subject_id])/1000000;
else
byte_sec=
flow[flow_id]->measurement->parameters->interval.tv_sec +
(double)(flow[flow_id]->measurement->parameters->interval.tv_usec)/1000000;
scope_packets+=(packets/pkt_sec);
scope_bytes+=(bytes/byte_sec);
flow[flow_id]->pkt_ts[subject_id]=
pkt_counter_res[subject_id].ts;
flow[flow_id]->byte_ts[subject_id]=
byte_counter_res[subject_id].ts;
if (tv_diff_threshold) {
syslog(LOG_DEBUG, "%s:%s: %.02f seconds from previous result",
flow[flow_id]->measurement->scope->subject[subject_id]->hostname,
flow[flow_id]->measurement->scope->subject[subject_id]->interface,
byte_sec);
}
/* Print result */
if (!configuration.global->no_stdout) {
if (configuration.global->stdout_simple)
printf(" %0.2f %0.2f %0.2f", packets/pkt_sec,
bytes/byte_sec, mbps/byte_sec);
else
printf(" %0.2f packets/s, %0.2f bytes/s, %0.2f Mb/s, time %uus/%uus, interval %0.2fs/%0.2fs\n",
packets/pkt_sec, bytes/byte_sec, mbps/byte_sec,
tv_diff_pkt, tv_diff_byte, pkt_sec, byte_sec);
}
} /* for (subject_id=0; subject_id++; subject_id<scope_size) */
if (!configuration.global->no_stdout)
printf("\n");
/* If interval is at least 1 second, then store results
to RRD file */
if (flow[flow_id]->measurement->parameters->interval.tv_sec) {
sprintf(command, "rrdtool update %s %u:%lu:%lu:%.6f",
flow[flow_id]->rrd_filename,
(unsigned int)(tm.tv_sec),
(unsigned long)(scope_packets),
(unsigned long)(scope_bytes),
(double)scope_bytes*8/1000000);
if (configuration.global->debug > 1)
syslog(LOG_DEBUG, "system(%s)", command);
if (tm.tv_sec == flow[flow_id]->rrd_ts)
syslog(LOG_ERR, "duplicate RRD timestamp %u for scope %d\n", (unsigned int)(tm.tv_sec), flow[flow_id]->measurement->scope->id);
else
flow[flow_id]->rrd_ts=tm.tv_sec;
if (debug)
printf("%s: system(%s)\n", __func__, command);
if (system(command)<0) {
fprintf(stderr, "%s: command(%s) failed\n", __func__,
command);
return -1;
}
}
flow_id++;
} /* while (flow[flow_id] && flow_id<MAX_FLOWS) */
abw_next_timestamp(&(configuration.measurement->parameters->interval),
&next, &wait);
if (!configuration.global->no_stdout &&
!configuration.global->stdout_simple) {
printf("next.tv_sec: %d, next.tv_usec: %d, wait.tv_sec: %d, wait.tv_usec: %d\n", (int)(next.tv_sec), (int)(next.tv_usec), (int)(wait.tv_sec), (int)(wait.tv_usec));
printf("===============================================================================\n");
}
usleep(wait.tv_sec * 1000000 + wait.tv_usec);
} /* while (1) */
return 0;
} /* main() */
void store_pool_cred_handler(void *, int /*i*/, Stream *s)
{
int result;
char *pw = NULL;
char *domain = NULL;
MyString username = POOL_PASSWORD_USERNAME "@";
if (s->type() != Stream::reli_sock) {
dprintf(D_ALWAYS, "ERROR: pool password set attempt via UDP\n");
return;
}
// if we're the CREDD_HOST, make sure any password setting is done locally
// (since knowing what the pool password is on the CREDD_HOST means being
// able to fetch users' passwords)
char *credd_host = param("CREDD_HOST");
if (credd_host) {
MyString my_fqdn_str = get_local_fqdn();
MyString my_hostname_str = get_local_hostname();
MyString my_ip_str = get_local_ipaddr().to_ip_string();
// figure out if we're on the CREDD_HOST
bool on_credd_host = (strcasecmp(my_fqdn_str.Value(), credd_host) == MATCH);
on_credd_host = on_credd_host || (strcasecmp(my_hostname_str.Value(), credd_host) == MATCH);
on_credd_host = on_credd_host || (strcmp(my_ip_str.Value(), credd_host) == MATCH);
if (on_credd_host) {
// we're the CREDD_HOST; make sure the source address matches ours
const char *addr = ((ReliSock*)s)->peer_ip_str();
if (!addr || strcmp(my_ip_str.Value(), addr)) {
dprintf(D_ALWAYS, "ERROR: attempt to set pool password remotely\n");
free(credd_host);
return;
}
}
free(credd_host);
}
s->decode();
if (!s->code(domain) || !s->code(pw) || !s->end_of_message()) {
dprintf(D_ALWAYS, "store_pool_cred: failed to receive all parameters\n");
goto spch_cleanup;
}
if (domain == NULL) {
dprintf(D_ALWAYS, "store_pool_cred_handler: domain is NULL\n");
goto spch_cleanup;
}
// construct the full pool username
username += domain;
// do the real work
if (pw) {
result = store_cred_service(username.Value(), pw, ADD_MODE);
SecureZeroMemory(pw, strlen(pw));
}
else {
result = store_cred_service(username.Value(), NULL, DELETE_MODE);
}
s->encode();
if (!s->code(result)) {
dprintf(D_ALWAYS, "store_pool_cred: Failed to send result.\n");
goto spch_cleanup;
}
if (!s->end_of_message()) {
dprintf(D_ALWAYS, "store_pool_cred: Failed to send end of message.\n");
}
spch_cleanup:
if (pw) free(pw);
if (domain) free(domain);
}