bool do_pass() {
int retval;
DB_TRANSITIONER_ITEM_SET transitioner;
std::vector<TRANSITIONER_ITEM> items;
bool did_something = false;
if (!one_pass) check_stop_daemons();
// loop over entries that are due to be checked
//
while (1) {
retval = transitioner.enumerate(
(int)time(0), SELECT_LIMIT, mod_n, mod_i, items
);
if (retval) {
if (retval != ERR_DB_NOT_FOUND) {
log_messages.printf(MSG_CRITICAL,
"WU enum error: %s; exiting\n", boincerror(retval)
);
exit(1);
}
break;
}
did_something = true;
TRANSITIONER_ITEM& wu_item = items[0];
retval = handle_wu(transitioner, items);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"[WU#%u %s] handle_wu: %s; quitting\n",
wu_item.id, wu_item.name, boincerror(retval)
);
exit(1);
}
if (!one_pass) check_stop_daemons();
}
return did_something;
}
// delete files in antique files list, and empty the list.
// Returns number of files deleted, or negative for error.
//
// TODO: the list contains filenames, and we convert these to paths.
// This is wacked. The list should contain paths.
//
int delete_antique_files() {
int nfiles=0;
log_messages.printf(MSG_DEBUG,
"delete_antique_files(): start (%d files)\n",
(int)files_to_delete.size()
);
while (!files_to_delete.empty()) {
char timestamp[128];
char pathname[1024];
int retval;
FILE_RECORD fr = files_to_delete.front();
check_stop_daemons();
retval = get_file_path(
fr.name.c_str(), config.upload_dir,
config.uldl_dir_fanout, pathname
);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"get_file_path(%s) failed: %s\n", fr.name.c_str(), boincerror(retval)
);
return retval;
}
strcpy(timestamp, time_to_string(fr.date_modified));
log_messages.printf(MSG_DEBUG,
"deleting [antique %s] %s\n",
timestamp, pathname
);
if (unlink(pathname)) {
int save_error=errno;
log_messages.printf(MSG_CRITICAL,
"unlink(%s) failed: %s\n",
pathname, strerror(save_error)
);
return retval;
} else {
nfiles++;
files_to_delete.pop_front();
}
}
log_messages.printf(MSG_DEBUG,
"delete_antique_files(): done, deleted %d files\n", nfiles
);
return 0;
}
int delete_host_file(int host_id, const char* file_name) {
DB_MSG_TO_HOST mth;
int retval;
mth.clear();
mth.create_time = time(0);
mth.hostid = host_id;
mth.handled = false;
sprintf(mth.xml, "<delete_file_info>%s</delete_file_info>\n", file_name);
sprintf(mth.variety, "delete_file");
retval = mth.insert();
if (retval) {
fprintf(stderr, "msg_to_host.insert(): %s\n", boincerror(retval));
return retval;
}
return 0;
}
int handle_message(MSG_FROM_HOST& mfh) {
int retval;
printf("got message \n%s\n", mfh.xml);
DB_MSG_TO_HOST mth;
mth.clear();
mth.create_time = time(0);
mth.hostid = mfh.hostid;
mth.handled = false;
safe_strcpy(mth.xml, mfh.xml);
retval = mth.insert();
if (retval) {
printf("insert failed %s\n", boincerror(retval));
}
return 0;
}
int ACTIVE_TASK::copy_output_files() {
for (size_t i = 0; i < result->output_files.size(); ++i) {
FILE_REF& fref = result->output_files[i];
if (!fref.copy_file) {
continue;
}
const FILE_INFO* fip = fref.file_info;
std::string slotfile = slot_dir + std::string("/") + std::string(fref.open_name);
std::string projfile = get_pathname(fip);
int retval = boinc_rename(slotfile.c_str(), projfile.c_str());
if (retval) {
msg_printf(wup->project, MSG_INTERNAL_ERROR, "Can't rename output file %s to %s: %s",
fip->name.c_str(), projfile.c_str(), boincerror(retval));
}
}
return 0;
}
// Here when a result has been validated.
// - update consecutive_valid
// - udpdate turnaround stats
// - insert credited_job record if needed
//
int is_valid(
DB_HOST& host, RESULT& result, WORKUNIT& wu, DB_HOST_APP_VERSION& hav
) {
DB_CREDITED_JOB credited_job;
int retval;
double turnaround = result.received_time - result.sent_time;
compute_avg_turnaround(host, turnaround);
// increment daily quota
//
hav.max_jobs_per_day++;
// increment consecutive_valid, but only if unreplicated
//
if (!is_unreplicated(wu)) {
hav.consecutive_valid++;
log_messages.printf(MSG_DEBUG,
"[HAV#%lu] consecutive valid now %d\n",
hav.app_version_id, hav.consecutive_valid
);
}
if (update_credited_job) {
credited_job.userid = host.userid;
credited_job.workunitid = long(wu.opaque);
if (dry_run) {
log_messages.printf(MSG_NORMAL, "DB not updated (dry run)\n");
} else {
retval = credited_job.insert();
if (retval) {
log_messages.printf(MSG_CRITICAL,
"[RESULT#%lu] Warning: credited_job insert failed (userid: %lu workunit: %f err: %s)\n",
result.id, host.userid, wu.opaque, boincerror(retval)
);
} else {
log_messages.printf(MSG_DEBUG,
"[RESULT#%lu %s] added credited_job record [WU#%lu OPAQUE#%f USER#%lu]\n",
result.id, result.name, wu.id, wu.opaque, host.userid
);
}
}
}
return 0;
}
int main_loop() {
int retval;
bool did_something;
char buf[256];
retval = boinc_db.open(
config.db_name, config.db_host, config.db_user, config.db_passwd
);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"boinc_db.open failed: %s\n", boincerror(retval)
);
exit(1);
}
sprintf(buf, "where name='%s'", app_name);
while (1) {
check_stop_daemons();
// look up app within the loop,
// in case its min_avg_pfc has been changed by the feeder
//
retval = app.lookup(buf);
if (retval) {
log_messages.printf(MSG_CRITICAL, "can't find app %s\n", app_name);
exit(1);
}
did_something = do_validate_scan();
if (!did_something) {
write_modified_app_versions(app_versions);
if (one_pass) break;
#ifdef GCL_SIMULATOR
char nameforsim[64];
sprintf(nameforsim, "validator%i", app.id);
continue_simulation(nameforsim);
signal(SIGUSR2, simulator_signal_handler);
pause();
#else
sleep(sleep_interval);
#endif
}
}
return 0;
}
int main(int argc, char** argv) {
int i, retval;
bool one_pass = false;
check_stop_daemons();
for (i=1; i<argc; i++) {
if (is_arg(argv[i], "one_pass")) {
one_pass = true;
} else if (is_arg(argv[i], "d")) {
if (!argv[++i]) {
log_messages.printf(MSG_CRITICAL, "%s requires an argument\n\n", argv[--i]);
usage(argv[0]);
exit(1);
}
int dl = atoi(argv[i]);
log_messages.set_debug_level(dl);
if (dl == 4) g_print_queries = true;
} else if (is_arg(argv[i], "h") || is_arg(argv[i], "help")) {
usage(argv[0]);
exit(0);
} else if (is_arg(argv[i], "v") || is_arg(argv[i], "version")) {
printf("%s\n", SVN_VERSION);
exit(0);
} else {
log_messages.printf(MSG_CRITICAL, "unknown command line argument: %s\n\n", argv[i]);
usage(argv[0]);
exit(1);
}
}
retval = config.parse_file();
if (retval) {
log_messages.printf(MSG_CRITICAL,
"Can't parse config.xml: %s\n", boincerror(retval)
);
exit(1);
}
log_messages.printf(MSG_NORMAL, "Starting message handler\n");
install_stop_signal_handler();
main_loop(one_pass);
}
int main() {
SCHED_SHMEM* ssp;
int retval;
void* p;
retval = config.parse_file();
if (retval) {
printf("Can't parse config.xml: %s\n", boincerror(retval));
exit(1);
}
retval = attach_shmem(config.shmem_key, &p);
if (retval) {
printf("can't attach shmem: key %x\n", config.shmem_key);
exit(1);
}
ssp = (SCHED_SHMEM*)p;
retval = ssp->verify();
ssp->show(stdout);
}
// Ask the WU generator to make more WUs for this file.
// Returns nonzero if can't make more work.
// Returns zero if it *might* have made more work
// (no way to be sure if it succeeded).
//
int make_more_work_for_file(char* filename) {
int retval = 0;
DB_SCHED_TRIGGER trigger;
if (!retrieve_single_trigger_by_fileset_name(filename, trigger)) {
// trigger retrieval failed (message logged by previous method)
return -1;
}
// Check if there's remaining work for this WU
if (trigger.no_work_available) {
// Give up trying to interact with the WU generator.
if (config.debug_locality) {
log_messages.printf(MSG_NORMAL,
"[locality] work generator says no work remaining for trigger %s\n", filename
);
}
return -1;
}
// // FIXME: should we reset these? The old code didn't do any consistency checks...
// trigger.work_available = false;
// trigger.no_work_available = false;
// trigger.working_set_removal = false;
// set trigger state to need_work as a way of indicating that we need work
// for this fileset. If this operation fails, don't worry or tarry!
retval = trigger.update_single_state(DB_SCHED_TRIGGER::state_need_work, true);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"unable to set need_work state for trigger %s (error: %s)\n",
filename, boincerror(retval)
);
return -1;
}
return 0;
}
static void flag_for_possible_removal(char* fileset_name) {
int retval = 0;
DB_SCHED_TRIGGER trigger;
if (!retrieve_single_trigger_by_fileset_name(fileset_name, trigger)) {
// trigger retrieval failed (message logged by previous method)
return;
}
// // FIXME: should we reset these? The old code didn't do any consistency checks...
// trigger.need_work = false;
// trigger.work_available = false;
// trigger.no_work_available = false;
// set trigger state to working_set_removal
retval = trigger.update_single_state(DB_SCHED_TRIGGER::state_working_set_removal, true);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"unable to set working_set_removal state for trigger %s (error: %s)\n",
fileset_name, boincerror(retval)
);
}
}
int main(int argc, char** argv) {
char path[256];
int retval;
if ( (argc == 1) || !strcmp(argv[1], "-h") || !strcmp(argv[1],"--help") || (argc != 2) ) {
printf (usage);
exit(1);
}
retval = config.parse_file();
if (retval) {
fprintf(stderr, "Can't parse config.xml: %s\n", boincerror(retval));
exit(1);
}
retval = dir_hier_path(
argv[1], config.download_dir, config.uldl_dir_fanout, path, true
);
if (retval) {
fprintf(stderr, "dir_hier_path(): %d\n", retval);
exit(1);
}
printf("%s\n", path);
}
int result_delete_files(RESULT& result) {
char* p;
char filename[256], pathname[256], buf[BLOB_SIZE];
bool no_delete=false;
int count_deleted = 0, retval, mthd_retval = 0;
safe_strcpy(buf, result.xml_doc_in);
p = strtok(buf,"\n");
while (p) {
if (parse_str(p, "<name>", filename, sizeof(filename))) {
} else if (match_tag(p, "<file_info>")) {
no_delete = false;
strcpy(filename, "");
} else if (match_tag(p, "<no_delete/>")) {
no_delete = true;
} else if (match_tag(p, "</file_info>")) {
if (!no_delete) {
retval = get_file_path(
filename, config.upload_dir, config.uldl_dir_fanout,
pathname
);
if (retval == ERR_OPENDIR) {
mthd_retval = ERR_OPENDIR;
log_messages.printf(MSG_CRITICAL,
"[RESULT#%d] missing dir for %s\n",
result.id, pathname
);
} else if (retval) {
// the fact that no result files were found is a critical
// error if this was a successful result, but is to be
// expected if the result outcome was failure, since in
// that case there may well be no output file produced.
//
int debug_or_crit;
if (RESULT_OUTCOME_SUCCESS == result.outcome) {
debug_or_crit=MSG_CRITICAL;
} else {
debug_or_crit=MSG_DEBUG;
}
log_messages.printf(debug_or_crit,
"[RESULT#%d] outcome=%d client_state=%d No file %s to delete\n",
result.id, result.outcome, result.client_state, filename
);
} else {
retval = unlink(pathname);
if (retval) {
mthd_retval = ERR_UNLINK;
log_messages.printf(MSG_CRITICAL,
"[RESULT#%d] unlink %s error: %s %s\n",
result.id, pathname, boincerror(retval),
(retval && errno)?strerror(errno):""
);
} else {
count_deleted++;
log_messages.printf(MSG_NORMAL,
"[RESULT#%d] unlinked %s\n", result.id, pathname
);
}
}
}
}
p = strtok(0, "\n");
}
log_messages.printf(MSG_DEBUG,
"[RESULT#%d] deleted %d file(s)\n", result.id, count_deleted
);
return mthd_retval;
}
// Called by validator when canonical result has been selected.
// Compute credit for valid instances.
// This is called exactly once for each valid result.
//
int assign_credit_set(
WORKUNIT& wu, vector<RESULT>& results,
DB_APP& app,
vector<DB_APP_VERSION>& app_versions,
vector<DB_HOST_APP_VERSION>& host_app_versions,
double max_granted_credit, double& credit
) {
unsigned int i;
int mode, retval;
double pfc;
vector<double> normal;
vector<double> approx;
for (i=0; i<results.size(); i++) {
RESULT& r = results[i];
if (r.validate_state != VALIDATE_STATE_VALID) continue;
DB_HOST_APP_VERSION& hav = host_app_versions[i];
retval = get_pfc(r, wu, app, app_versions, hav, pfc, mode);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"get_pfc() error: %s\n", boincerror(retval)
);
continue;
} else {
if (config.debug_credit) {
log_messages.printf(MSG_NORMAL,
"[credit] [RESULT#%d] get_pfc() returns credit %g mode %s\n",
r.id, pfc*COBBLESTONE_SCALE, (mode==PFC_MODE_NORMAL)?"normal":"approx"
);
}
}
if (pfc > wu.rsc_fpops_bound) {
log_messages.printf(MSG_NORMAL,
"[credit] PFC too high: %f\n", pfc*COBBLESTONE_SCALE
);
pfc = wu_estimated_pfc(wu, app);
}
// max_granted_credit trumps rsc_fpops_bound;
// the latter may be set absurdly high
//
if (max_granted_credit && pfc*COBBLESTONE_SCALE > max_granted_credit) {
log_messages.printf(MSG_NORMAL,
"[credit] Credit too high: %f\n", pfc*COBBLESTONE_SCALE
);
pfc = max_granted_credit/COBBLESTONE_SCALE;
}
if (mode == PFC_MODE_NORMAL) {
normal.push_back(pfc);
} else {
approx.push_back(pfc);
}
}
// averaging policy: if there is least one normal result,
// use the "low average" of normal results.
// Otherwise use the min of all results
//
double x;
if (normal.size()) {
x = low_average(normal);
} else if (approx.size()) {
x = vec_min(approx);
} else {
x = 0;
}
x *= COBBLESTONE_SCALE;
if (config.debug_credit) {
log_messages.printf(MSG_NORMAL,
"[credit] [WU#%d] assign_credit_set: credit %g\n",
wu.id, x
);
}
credit = x;
return 0;
}
int wu_delete_files(WORKUNIT& wu) {
char* p;
char filename[256], pathname[256], buf[BLOB_SIZE];
bool no_delete=false;
int count_deleted = 0, retval, mthd_retval = 0;
if (strstr(wu.name, "nodelete")) return 0;
safe_strcpy(buf, wu.xml_doc);
p = strtok(buf, "\n");
strcpy(filename, "");
while (p) {
if (parse_str(p, "<name>", filename, sizeof(filename))) {
} else if (match_tag(p, "<file_info>")) {
no_delete = false;
strcpy(filename, "");
} else if (match_tag(p, "<no_delete/>")) {
no_delete = true;
} else if (match_tag(p, "</file_info>")) {
if (!no_delete) {
retval = get_file_path(
filename, config.download_dir, config.uldl_dir_fanout,
pathname
);
if (retval == ERR_OPENDIR) {
log_messages.printf(MSG_CRITICAL,
"[WU#%d] missing dir for %s\n",
wu.id, filename
);
mthd_retval = ERR_UNLINK;
} else if (retval) {
log_messages.printf(MSG_CRITICAL,
"[WU#%d] get_file_path: %s: %s\n",
wu.id, filename, boincerror(retval)
);
} else {
log_messages.printf(MSG_NORMAL,
"[WU#%d] deleting %s\n", wu.id, filename
);
retval = unlink(pathname);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"[WU#%d] unlink %s failed: %s\n",
wu.id, filename, boincerror(retval)
);
mthd_retval = ERR_UNLINK;
} else {
count_deleted++;
}
// delete the cached MD5 file if needed
//
if (config.cache_md5_info) {
strcat(pathname,".md5");
log_messages.printf(MSG_NORMAL,
"[WU#%d] deleting %s\n", wu.id, filename
);
retval = unlink(pathname);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"[WU#%d] unlink %s failed: %s\n",
wu.id, filename, boincerror(retval)
);
}
}
}
}
}
p = strtok(0, "\n");
}
log_messages.printf(MSG_DEBUG,
"[WU#%d] deleted %d file(s)\n", wu.id, count_deleted
);
return mthd_retval;
}
int main(int argc, char** argv) {
int retval;
int i;
check_stop_daemons();
for (i=1; i<argc; i++) {
if (is_arg(argv[i], "d") || is_arg(argv[i], "debug_level")) {
if (!argv[++i]) {
log_messages.printf(MSG_CRITICAL, "%s requires an argument\n\n", argv[--i]);
usage(argv[0]);
exit(1);
}
int dl = atoi(argv[i]);
log_messages.set_debug_level(dl);
if (dl == 4) g_print_queries = true;
} else if (is_arg(argv[i], "dry_run")) {
antiques_deletion_dry_run = true;
} else if (is_arg(argv[i], "usleep")) {
antique_usleep = atoi(argv[++i]);
} else if (is_arg(argv[i], "h") || is_arg(argv[i], "help")) {
usage(argv[0]);
exit(0);
} else if (is_arg(argv[i], "v") || is_arg(argv[i], "version")) {
printf("%s\n", SVN_VERSION);
exit(0);
} else {
log_messages.printf(MSG_CRITICAL, "unknown command line argument: %s\n\n", argv[i]);
usage(argv[0]);
exit(1);
}
}
retval = config.parse_file();
if (retval) {
log_messages.printf(MSG_CRITICAL,
"Can't parse config.xml: %s\n", boincerror(retval)
);
exit(1);
}
log_messages.printf(MSG_NORMAL, "Starting\n");
retval = boinc_db.open(
config.replica_db_name,
config.replica_db_host,
config.replica_db_user,
config.replica_db_passwd
);
if (retval) {
log_messages.printf(MSG_CRITICAL, "can't open DB\n");
exit(1);
}
retval = boinc_db.set_isolation_level(READ_UNCOMMITTED);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"boinc_db.set_isolation_level: %s; %s\n",
boincerror(retval), boinc_db.error_string()
);
}
install_stop_signal_handler();
retval = delete_antiques();
if (retval) {
log_messages.printf(MSG_CRITICAL,
"delete_antiques() returned with error %d\n",
retval
);
}
log_messages.printf(MSG_NORMAL, "Done\n");
return retval;
}
// Grant the host (and associated user and team)
// the given amount of credit for work that started at the given time.
// Update the user and team records,
// but not the host record (caller must update)
//
int grant_credit(DB_HOST& host, double start_time, double credit) {
DB_USER user;
DB_TEAM team;
int retval;
char buf[256];
double now = dtime();
// first, process the host
update_average(
now,
start_time, credit, CREDIT_HALF_LIFE,
host.expavg_credit, host.expavg_time
);
host.total_credit += credit;
// then the user
retval = user.lookup_id(host.userid);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"lookup of user %d failed: %s\n",
host.userid, boincerror(retval)
);
return retval;
}
update_average(
now,
start_time, credit, CREDIT_HALF_LIFE,
user.expavg_credit, user.expavg_time
);
sprintf(
buf, "total_credit=total_credit+%.15e, expavg_credit=%.15e, expavg_time=%.15e",
credit, user.expavg_credit, user.expavg_time
);
retval = user.update_field(buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"update of user %d failed: %s\n",
host.userid, boincerror(retval)
);
}
// and finally the team
if (user.teamid) {
retval = team.lookup_id(user.teamid);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"lookup of team %d failed: %s\n",
user.teamid, boincerror(retval)
);
return retval;
}
update_average(
now,
start_time, credit, CREDIT_HALF_LIFE,
team.expavg_credit, team.expavg_time
);
sprintf(buf,
"total_credit=total_credit+%.15e, expavg_credit=%.15e, expavg_time=%.15e",
credit, team.expavg_credit, team.expavg_time
);
retval = team.update_field(buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"update of team %d failed: %s\n",
team.id, boincerror(retval)
);
}
}
return 0;
}
int main(int argc, char** argv) {
#ifndef _USING_FCGI_
FILE* fin, *fout;
#else
FCGI_FILE *fin, *fout;
#endif
int i, retval;
char req_path[MAXPATHLEN], reply_path[MAXPATHLEN];
char log_path[MAXPATHLEN], path[MAXPATHLEN];
unsigned int counter=0;
char* code_sign_key;
int length = -1;
log_messages.pid = getpid();
bool debug_log = false;
for (i=1; i<argc; i++) {
if (!strcmp(argv[i], "--batch")) {
batch = true;
continue;
} else if (!strcmp(argv[i], "--mark_jobs_done")) {
mark_jobs_done = true;
} else if (!strcmp(argv[i], "--debug_log")) {
debug_log = true;
#ifdef GCL_SIMULATOR
} else if (!strcmp(argv[i], "--simulator")) {
if(!argv[++i]) {
log_messages.printf(MSG_CRITICAL, "%s requires an argument\n\n", argv[--i]);
usage(argv[0]);
exit(1);
}
simtime = atof(argv[i]);
#endif
} else if(!strcmp(argv[i], "-h") || !strcmp(argv[i], "--help")) {
usage(argv[0]);
exit(0);
} else if(!strcmp(argv[i], "-v") || !strcmp(argv[i], "--version")) {
printf("%s\n", SVN_VERSION);
exit(0);
} else if (strlen(argv[i])){
log_messages.printf(MSG_CRITICAL, "unknown command line argument: %s\n\n", argv[i]);
usage(argv[0]);
exit(1);
}
}
// install a signal handler that catches SIGTERMS sent by Apache if the CGI
// times out.
//
signal(SIGTERM, sigterm_handler);
if (debug_log) {
if (!freopen("debug_log", "w", stderr)) {
fprintf(stderr, "Can't redirect stderr\n");
exit(1);
}
} else {
char *stderr_buffer;
get_log_path(path, "scheduler.log");
#ifndef _USING_FCGI_
char buf[256];
if (!freopen(path, "a", stderr)) {
fprintf(stderr, "Can't redirect stderr\n");
sprintf(buf, "Server can't open log file (%s)", path);
send_message(buf, 3600);
exit(1);
}
#else
FCGI_FILE* f = FCGI::fopen(path, "a");
if (f) {
log_messages.redirect(f);
} else {
char buf[256];
fprintf(stderr, "Can't redirect FCGI log messages\n");
sprintf(buf, "Server can't open log file for FCGI (%s)", path);
send_message(buf, 3600);
exit(1);
}
#endif
// install a larger buffer for stderr. This ensures that
// log information from different scheduler requests running
// in parallel aren't intermingled in the log file.
//
if (config.scheduler_log_buffer) {
stderr_buffer = (char*)malloc(config.scheduler_log_buffer);
if (!stderr_buffer) {
log_messages.printf(MSG_CRITICAL,
"Unable to allocate stderr buffer\n"
);
} else {
#ifdef _USING_FCGI_
retval = setvbuf(
f->stdio_stream, stderr_buffer, _IOFBF,
config.scheduler_log_buffer
);
#else
retval = setvbuf(
stderr, stderr_buffer, _IOFBF, config.scheduler_log_buffer
);
#endif
if (retval) {
log_messages.printf(MSG_CRITICAL,
"Unable to change stderr buffering\n"
);
}
}
}
}
srand(time(0)+getpid());
log_messages.set_debug_level(DEBUG_LEVEL);
#if DUMP_CORE_ON_SEGV
set_core_dump_size_limit();
#endif
retval = config.parse_file();
if (retval) {
log_messages.printf(MSG_CRITICAL,
"Can't parse config.xml: %s\n", boincerror(retval)
);
send_message("Server can't parse configuration file", 3600);
exit(0);
}
log_messages.set_debug_level(config.sched_debug_level);
if (config.sched_debug_level == 4) g_print_queries = true;
gui_urls.init();
project_files.init();
init_file_delete_regex();
sprintf(path, "%s/code_sign_public", config.key_dir);
retval = read_file_malloc(path, code_sign_key);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"Can't read code sign key file (%s)\n", path
);
send_message("Server can't find key file", 3600);
exit(0);
}
strip_whitespace(code_sign_key);
g_pid = getpid();
#ifdef _USING_FCGI_
//while(FCGI_Accept() >= 0 && counter < MAX_FCGI_COUNT) {
while(FCGI_Accept() >= 0) {
counter++;
log_messages.set_indent_level(0);
#endif
if (config.debug_request_headers) {
log_request_headers(length);
}
if (!debug_log && check_stop_sched()) {
send_message("Project is temporarily shut down for maintenance", 3600);
goto done;
}
if (!ssp) {
attach_to_feeder_shmem();
}
if (!ssp) {
send_message("Server error: can't attach shared memory", 3600);
goto done;
}
if (strlen(config.debug_req_reply_dir)) {
struct stat statbuf;
// the code below is convoluted because,
// instead of going from stdin to stdout directly,
// we go via a pair of disk files
// (this makes it easy to save the input,
// and to know the length of the output).
// NOTE: to use this, you must create group-writeable dirs
// boinc_req and boinc_reply in the project dir
//
sprintf(req_path, "%s/%d_%u_sched_request.xml", config.debug_req_reply_dir, g_pid, counter);
sprintf(reply_path, "%s/%d_%u_sched_reply.xml", config.debug_req_reply_dir, g_pid, counter);
// keep an own 'log' per PID in case general logging fails
// this allows to associate at leas the scheduler request with the client
// IP address (as shown in httpd error log) in case of a crash
sprintf(log_path, "%s/%d_%u_sched.log", config.debug_req_reply_dir, g_pid, counter);
#ifndef _USING_FCGI_
fout = fopen(log_path, "a");
#else
fout = FCGI::fopen(log_path,"a");
#endif
fprintf(fout, "PID: %d Client IP: %s\n", g_pid, get_remote_addr());
fclose(fout);
log_messages.printf(MSG_DEBUG,
"keeping sched_request in %s, sched_reply in %s, custom log in %s\n",
req_path, reply_path, log_path
);
#ifndef _USING_FCGI_
fout = fopen(req_path, "w");
#else
fout = FCGI::fopen(req_path,"w");
#endif
if (!fout) {
log_messages.printf(MSG_CRITICAL,
"can't write request file\n"
);
exit(1);
}
copy_stream(stdin, fout);
fclose(fout);
stat(req_path, &statbuf);
if (length>=0 && (statbuf.st_size != length)) {
log_messages.printf(MSG_CRITICAL,
"Request length %d != CONTENT_LENGTH %d\n",
(int)statbuf.st_size, length
);
}
#ifndef _USING_FCGI_
fin = fopen(req_path, "r");
#else
fin = FCGI::fopen(req_path,"r");
#endif
if (!fin) {
log_messages.printf(MSG_CRITICAL,
"can't read request file\n"
);
exit(1);
}
#ifndef _USING_FCGI_
fout = fopen(reply_path, "w");
#else
fout = FCGI::fopen(reply_path, "w");
#endif
if (!fout) {
log_messages.printf(MSG_CRITICAL,
"can't write reply file\n"
);
exit(1);
}
handle_request(fin, fout, code_sign_key);
fclose(fin);
fclose(fout);
#ifndef _USING_FCGI_
fin = fopen(reply_path, "r");
#else
fin = FCGI::fopen(reply_path, "r");
#endif
if (!fin) {
log_messages.printf(MSG_CRITICAL,
"can't read reply file\n"
);
exit(1);
}
copy_stream(fin, stdout);
fclose(fin);
// if not contacted from a client, don't keep the log files
/* not sure what lead to the assumption of a client setting
CONTENT_LENGTH, but it's wrong at least on our current
project / Apache / Client configuration. Commented out.
if (getenv("CONTENT_LENGTH")) {
unlink(req_path);
unlink(reply_path);
}
*/
#ifndef _USING_FCGI_
} else if (batch) {
while (!feof(stdin)) {
handle_request(stdin, stdout, code_sign_key);
fflush(stdout);
}
#endif
} else {
handle_request(stdin, stdout, code_sign_key);
fflush(stderr);
}
done:
#ifdef _USING_FCGI_
if (config.debug_fcgi) {
log_messages.printf(MSG_NORMAL,
"FCGI: counter: %d\n", counter
);
log_messages.flush();
}
} // do()
if (counter == MAX_FCGI_COUNT) {
fprintf(stderr, "FCGI: counter passed MAX_FCGI_COUNT - exiting..\n");
} else {
fprintf(stderr, "FCGI: FCGI_Accept failed - exiting..\n");
}
// when exiting, write headers back to apache so it won't complain
// about "incomplete headers"
fprintf(stdout,"Content-type: text/plain\n\n");
#endif
if (db_opened) {
boinc_db.close();
}
}
int wu_delete_files(WORKUNIT& wu) {
char* p;
char filename[256], path[MAXPATHLEN], buf[BLOB_SIZE];
char path_gz[MAXPATHLEN], path_md5[MAXPATHLEN];
bool no_delete=false;
int count_deleted = 0, retval, mthd_retval = 0;
if (strstr(wu.name, "nodelete")) return 0;
safe_strcpy(buf, wu.xml_doc);
p = strtok(buf, "\n");
strcpy(filename, "");
// TODO: use the XML parser. Yuck!
//
while (p) {
if (parse_str(p, "<name>", filename, sizeof(filename))) {
} else if (match_tag(p, "<file_info>")) {
no_delete = false;
strcpy(filename, "");
} else if (match_tag(p, "<no_delete/>")) {
no_delete = true;
} else if (match_tag(p, "</file_info>")) {
if (!no_delete) {
retval = get_file_path(
filename, download_dir, config.uldl_dir_fanout,
path
);
if (retval == ERR_OPENDIR) {
log_messages.printf(MSG_CRITICAL,
"[WU#%lu] missing dir for %s\n",
wu.id, filename
);
mthd_retval = ERR_UNLINK;
} else if (retval) {
log_messages.printf(MSG_CRITICAL,
"[WU#%lu] get_file_path: %s: %s\n",
wu.id, filename, boincerror(retval)
);
} else {
log_messages.printf(MSG_NORMAL,
"[WU#%lu] deleting %s\n", wu.id, filename
);
retval = unlink(path);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"[WU#%lu] unlink %s failed: %s\n",
wu.id, filename, boincerror(retval)
);
mthd_retval = ERR_UNLINK;
} else {
count_deleted++;
}
// delete the gzipped version of the file
//
sprintf(path_gz, "%s.gz", path);
retval = unlink(path_gz);
if (!retval) {
log_messages.printf(MSG_NORMAL,
"[WU#%lu] deleted %s.gz\n", wu.id, filename
);
}
// delete the cached MD5 file if needed
//
if (config.cache_md5_info) {
sprintf(path_md5, "%s.md5", path);
log_messages.printf(MSG_NORMAL,
"[WU#%lu] deleting %s.md5\n", wu.id, filename
);
retval = unlink(path_md5);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"[WU#%lu] unlink %s.md5 failed: %s\n",
wu.id, filename, boincerror(retval)
);
}
}
}
}
}
p = strtok(0, "\n");
}
log_messages.printf(MSG_DEBUG,
"[WU#%lu] deleted %d file(s)\n", wu.id, count_deleted
);
return mthd_retval;
}
// resend any jobs that:
// 1) we already sent to this host;
// 2) are still in progress (i.e. haven't timed out) and
// 3) aren't present on the host
// Return true if there were any such jobs
//
bool resend_lost_work() {
SCHED_DB_RESULT result;
std::vector<DB_RESULT>results;
unsigned int i;
char buf[256];
char warning_msg[256];
bool did_any = false;
int num_eligible_to_resend=0;
int num_resent=0;
BEST_APP_VERSION* bavp = NULL;
APP* app = NULL;
int retval;
sprintf(buf, " where hostid=%d and server_state=%d ",
g_reply->host.id, RESULT_SERVER_STATE_IN_PROGRESS
);
while (!result.enumerate(buf)) {
if (!work_needed(false)) {
result.end_enumerate();
break;
}
bool found = false;
for (i=0; i<g_request->other_results.size(); i++) {
OTHER_RESULT& orp = g_request->other_results[i];
if (!strcmp(orp.name, result.name)) {
found = true;
break;
}
}
if (found) continue;
num_eligible_to_resend++;
if (config.debug_resend) {
log_messages.printf(MSG_NORMAL,
"[resend] [HOST#%d] found lost [RESULT#%u]: %s\n",
g_reply->host.id, result.id, result.name
);
}
DB_WORKUNIT wu;
bool can_resend = true;
retval = wu.lookup_id(result.workunitid);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"[HOST#%d] can't resend - WU not found for [RESULT#%u]\n",
g_reply->host.id, result.id
);
can_resend = false;
}
if (can_resend) {
app = ssp->lookup_app(wu.appid);
bavp = get_app_version(wu, true, false);
if (!bavp) {
if (config.debug_resend) {
log_messages.printf(MSG_NORMAL,
"[HOST#%d] can't resend [RESULT#%u]: no app version for %s\n",
g_reply->host.id, result.id, app->name
);
}
can_resend = false;
}
}
if (can_resend && wu.error_mask) {
if (config.debug_resend) {
log_messages.printf(MSG_NORMAL,
"[resend] skipping [RESULT#%u]: WU error mask %d\n",
result.id, wu.error_mask
);
}
can_resend = false;
}
if (can_resend && wu.canonical_resultid) {
if (config.debug_resend) {
log_messages.printf(MSG_NORMAL,
"[resend] skipping [RESULT#%u]: already have canonical result\n",
result.id
);
}
can_resend = false;
}
if (can_resend && wu_is_infeasible_fast(
wu, result.server_state, result.priority, result.report_deadline,
*app, *bavp
)) {
if (config.debug_resend) {
log_messages.printf(MSG_NORMAL,
"[resend] skipping [RESULT#%u]: feasibility check failed\n",
result.id
);
}
can_resend = false;
}
if (can_resend && possibly_give_result_new_deadline(result, wu, *bavp)) {
if (config.debug_resend) {
log_messages.printf(MSG_NORMAL,
"[resend] skipping [RESULT#%u]: deadline assignment failed\n",
result.id
);
}
can_resend = false;
}
// If we can't resend this job for any of the above reasons,
// make it time out so that the transitioner does the right thing.
//
if (!can_resend) {
result.report_deadline = time(0)-1;
retval = result.mark_as_sent(result.server_state, config.report_grace_period);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"resend_lost_work: can't update result deadline: %s\n",
boincerror(retval)
);
continue;
}
retval = update_wu_on_send(
wu, result.report_deadline + config.report_grace_period,
*app, *bavp
);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"resend_lost_result: can't update WU transition time: %s\n",
boincerror(retval)
);
continue;
}
sprintf(warning_msg,
"Didn't resend lost task %s (expired)", result.name
);
g_reply->insert_message(warning_msg, "low");
} else {
retval = add_result_to_reply(result, wu, bavp, false);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"[HOST#%d] failed to send [RESULT#%u]\n",
g_reply->host.id, result.id
);
continue;
}
sprintf(warning_msg, "Resent lost task %s", result.name);
g_reply->insert_message(warning_msg, "low");
num_resent++;
did_any = true;
if (g_wreq->njobs_sent >= config.max_wus_to_send) {
result.end_enumerate();
break;
}
}
}
if (num_eligible_to_resend && config.debug_resend) {
log_messages.printf(MSG_NORMAL,
"[resend] [HOST#%d] %d lost results, resent %d\n",
g_reply->host.id, num_eligible_to_resend, num_resent
);
}
return did_any;
}
int main(int argc, char** argv) {
int i, retval;
char file_name[256];
int host_id;
host_id = 0;
strcpy(file_name, "");
check_stop_daemons();
for (i=1; i<argc; i++) {
if (is_arg(argv[i], "host_id")) {
if (!argv[++i]) {
fprintf(stderr, "%s requires an argument\n\n", argv[--i]);
usage(argv[0]);
exit(1);
}
host_id = atoi(argv[i]);
} else if (is_arg(argv[i], "file_name")) {
if (!argv[++i]) {
fprintf(stderr, "%s requires an argument\n\n", argv[--i]);
usage(argv[0]);
exit(1);
}
strcpy(file_name, argv[i]);
} else if (is_arg(argv[i], "help") || is_arg(argv[i], "h")) {
usage(argv[0]);
exit(0);
} else if (is_arg(argv[i], "version") || is_arg(argv[i], "v")) {
printf("%s\n", SVN_VERSION);
exit(0);
} else {
fprintf(stderr, "unknown command line argument: %s\n\n", argv[i]);
usage(argv[0]);
exit(1);
}
}
if (!strlen(file_name) || host_id == 0) {
usage(argv[0]);
exit(1);
}
retval = config.parse_file();
if (retval) {
fprintf(stderr, "Can't parse config.xml: %s\n", boincerror(retval));
exit(1);
}
retval = boinc_db.open(
config.db_name, config.db_host, config.db_user, config.db_passwd
);
if (retval) {
fprintf(stderr, "boinc_db.open failed: %s\n", boincerror(retval));
exit(1);
}
retval = create_delete_file_msg(host_id, file_name);
boinc_db.close();
return retval;
}
// return true if we changed the file_delete_state of a WU or a result
//
bool do_pass(bool retry_error) {
DB_WORKUNIT wu;
DB_RESULT result;
bool did_something = false;
char buf[256];
char clause[256];
int retval, new_state;
check_stop_daemons();
strcpy(clause, "");
if (id_modulus) {
sprintf(clause, " and id %% %d = %d ", id_modulus, id_remainder);
}
if (dont_delete_batches) {
strcat(clause, " and batch <= 0 ");
}
if (appid) {
sprintf(buf, " and appid = %d ", appid);
strcat(clause, buf);
}
sprintf(buf,
"where file_delete_state=%d %s limit %d",
retry_error?FILE_DELETE_ERROR:FILE_DELETE_READY,
clause, WUS_PER_ENUM
);
while (do_input_files) {
retval = wu.enumerate(buf);
if (retval) {
if (retval != ERR_DB_NOT_FOUND) {
log_messages.printf(MSG_DEBUG, "DB connection lost, exiting\n");
exit(0);
}
break;
}
if (preserve_wu_files) {
retval = 0;
} else {
retval = wu_delete_files(wu);
}
if (retval) {
new_state = FILE_DELETE_ERROR;
log_messages.printf(MSG_CRITICAL,
"[WU#%d] file deletion failed: %s\n", wu.id, boincerror(retval)
);
} else {
new_state = FILE_DELETE_DONE;
}
if (new_state != wu.file_delete_state) {
sprintf(buf, "file_delete_state=%d", new_state);
retval = wu.update_field(buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"[WU#%d] update failed: %s\n", wu.id, boincerror(retval)
);
} else {
log_messages.printf(MSG_DEBUG,
"[WU#%d] file_delete_state updated\n", wu.id
);
did_something = true;
}
}
}
sprintf(buf,
"where file_delete_state=%d %s limit %d",
retry_error?FILE_DELETE_ERROR:FILE_DELETE_READY,
clause, RESULTS_PER_ENUM
);
while (do_output_files) {
retval = result.enumerate(buf);
if (retval) {
if (retval != ERR_DB_NOT_FOUND) {
log_messages.printf(MSG_DEBUG, "DB connection lost, exiting\n");
exit(0);
}
break;
}
if (preserve_result_files) {
retval = 0;
} else {
retval = result_delete_files(result);
}
if (retval) {
new_state = FILE_DELETE_ERROR;
log_messages.printf(MSG_CRITICAL,
"[RESULT#%d] file deletion failed: %s\n", result.id, boincerror(retval)
);
} else {
new_state = FILE_DELETE_DONE;
}
if (new_state != result.file_delete_state) {
sprintf(buf, "file_delete_state=%d", new_state);
retval = result.update_field(buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"[RESULT#%d] update failed: %s\n", result.id, boincerror(retval)
);
} else {
log_messages.printf(MSG_DEBUG,
"[RESULT#%d] file_delete_state updated\n", result.id
);
did_something = true;
}
}
}
return did_something;
}
// Do checks that require DB access for whether we can send this job,
// and return:
// 0 if OK to send
// 1 if can't send to this host
// 2 if can't send to ANY host
//
int slow_check(
WU_RESULT& wu_result, // the job cache entry.
// We may refresh its hr_class and app_version_id fields.
APP* app,
BEST_APP_VERSION* bavp // the app version to be used
) {
int n, retval;
DB_RESULT result;
char buf[256];
WORKUNIT& wu = wu_result.workunit;
// Don't send if we've already sent a result of this WU to this user.
//
if (config.one_result_per_user_per_wu) {
sprintf(buf,
"where workunitid=%d and userid=%d", wu.id, g_reply->user.id
);
retval = result.count(n, buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"send_work: can't get result count (%s)\n", boincerror(retval)
);
return 1;
} else {
if (n>0) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [USER#%d] already has %d result(s) for [WU#%u]\n",
g_reply->user.id, n, wu.id
);
}
return 1;
}
}
} else if (config.one_result_per_host_per_wu) {
// Don't send if we've already sent a result of this WU to this host.
// We only have to check this if we don't send one result per user.
//
sprintf(buf,
"where workunitid=%d and hostid=%d", wu.id, g_reply->host.id
);
retval = result.count(n, buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"send_work: can't get result count (%s)\n", boincerror(retval)
);
return 1;
} else {
if (n>0) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [HOST#%d] already has %d result(s) for [WU#%u]\n",
g_reply->host.id, n, wu.id
);
}
return 1;
}
}
}
// Checks that require looking up the WU.
// Lump these together so we only do 1 lookup
//
if (app_hr_type(*app) || app->homogeneous_app_version) {
DB_WORKUNIT db_wu;
db_wu.id = wu.id;
int vals[3];
retval = db_wu.get_field_ints(
"hr_class, app_version_id, error_mask", 3, vals
);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"can't get fields for [WU#%u]: %s\n", db_wu.id, boincerror(retval)
);
return 1;
}
// check wu.error_mask
//
if (vals[2] != 0) {
return 2;
}
if (app_hr_type(*app)) {
wu.hr_class = vals[0];
if (already_sent_to_different_hr_class(wu, *app)) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [HOST#%d] [WU#%u %s] is assigned to different HR class\n",
g_reply->host.id, wu.id, wu.name
);
}
// Mark the workunit as infeasible.
// This ensures that jobs already assigned to an HR class
// are processed first.
//
wu_result.infeasible_count++;
return 1;
}
}
if (app->homogeneous_app_version) {
int wu_avid = vals[1];
wu.app_version_id = wu_avid;
if (wu_avid && wu_avid != bavp->avp->id) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [HOST#%d] [WU#%u %s] is assigned to different app version\n",
g_reply->host.id, wu.id, wu.name
);
}
wu_result.infeasible_count++;
return 1;
}
}
}
return 0;
}
int main(int argc, char** argv) {
int retval;
bool one_pass = false;
int i;
DB_APP app;
check_stop_daemons();
*app.name='\0';
for (i=1; i<argc; i++) {
if (is_arg(argv[i], "one_pass")) {
one_pass = true;
} else if (is_arg(argv[i], "dont_retry_errors")) {
dont_retry_errors = true;
} else if (is_arg(argv[i], "preserve_wu_files")) {
preserve_wu_files = true;
} else if (is_arg(argv[i], "preserve_result_files")) {
preserve_result_files = true;
} else if (is_arg(argv[i], "app")) {
strcpy(app.name, argv[++i]);
} else if (is_arg(argv[i], "appid")) {
if (!argv[++i]) {
log_messages.printf(MSG_CRITICAL, "%s requires an argument\n\n", argv[--i]);
usage(argv[0]);
exit(1);
}
appid = atoi(argv[i]);
} else if (is_arg(argv[i], "d") || is_arg(argv[i], "debug_level")) {
if (!argv[++i]) {
log_messages.printf(MSG_CRITICAL, "%s requires an argument\n\n", argv[--i]);
usage(argv[0]);
exit(1);
}
int dl = atoi(argv[i]);
log_messages.set_debug_level(dl);
if (dl == 4) g_print_queries = true;
} else if (is_arg(argv[i], "mod")) {
if (!argv[i+1] || !argv[i+2]) {
log_messages.printf(MSG_CRITICAL, "%s requires two arguments\n\n", argv[i]);
usage(argv[0]);
exit(1);
}
id_modulus = atoi(argv[++i]);
id_remainder = atoi(argv[++i]);
} else if (is_arg(argv[i], "dont_delete_antiques")) {
dont_delete_antiques = true;
} else if (is_arg(argv[i], "delete_antiques_interval")) {
antique_interval = atoi(argv[++i]);
} else if (is_arg(argv[i], "delete_antiques_limit")) {
antique_limit = atoi(argv[++i]);
} else if (is_arg(argv[i], "dont_delete_batches")) {
dont_delete_batches = true;
} else if (is_arg(argv[i], "delete_antiques_now")) {
antique_delay = 0;
} else if (is_arg(argv[i], "input_files_only")) {
do_output_files = false;
dont_delete_antiques = true;
} else if (is_arg(argv[i], "output_files_only")) {
do_input_files = false;
} else if (is_arg(argv[i], "sleep_interval")) {
if (!argv[++i]) {
log_messages.printf(MSG_CRITICAL, "%s requires an argument\n\n", argv[--i]);
usage(argv[0]);
exit(1);
}
sleep_interval = atoi(argv[i]);
} else if (is_arg(argv[i], "h") || is_arg(argv[i], "help")) {
usage(argv[0]);
exit(0);
} else if (is_arg(argv[i], "v") || is_arg(argv[i], "version")) {
printf("%s\n", SVN_VERSION);
exit(0);
} else {
log_messages.printf(MSG_CRITICAL, "unknown command line argument: %s\n\n", argv[i]);
usage(argv[0]);
exit(1);
}
}
if (id_modulus) {
log_messages.printf(MSG_DEBUG,
"Using mod'ed WU/result enumeration. mod = %d rem = %d\n",
id_modulus, id_remainder
);
}
retval = config.parse_file();
if (retval) {
log_messages.printf(MSG_CRITICAL,
"Can't parse config.xml: %s\n", boincerror(retval)
);
exit(1);
}
log_messages.printf(MSG_NORMAL, "Starting\n");
retval = boinc_db.open(config.db_name, config.db_host, config.db_user, config.db_passwd);
if (retval) {
log_messages.printf(MSG_CRITICAL, "can't open DB\n");
exit(1);
}
retval = boinc_db.set_isolation_level(READ_UNCOMMITTED);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"boinc_db.set_isolation_level: %s; %s\n",
boincerror(retval), boinc_db.error_string()
);
}
if (*app.name && !appid) {
char buf[256];
sprintf(buf, "where name='%s'", app.name);
retval = app.lookup(buf);
if (retval) {
log_messages.printf(MSG_CRITICAL, "Can't find app\n");
exit(1);
}
appid=app.id;
log_messages.printf(MSG_DEBUG, "Deleting files of appid %d\n",appid);
}
install_stop_signal_handler();
bool retry_errors_now = !dont_retry_errors;
double next_error_time=0;
double next_antique_time = dtime() + antique_delay;
while (1) {
bool got_any = do_pass(false);
if (retry_errors_now) {
bool got_any_errors = do_pass(true);
if (got_any_errors) {
got_any = true;
} else {
retry_errors_now = false;
next_error_time = dtime() + ERROR_INTERVAL;
log_messages.printf(MSG_DEBUG,
"ending retry of previous errors\n"
);
}
}
if (!got_any) {
if (one_pass) break;
sleep(sleep_interval);
}
if (!dont_delete_antiques && (dtime() > next_antique_time)) {
log_messages.printf(MSG_DEBUG,
"Doing antique deletion pass\n"
);
do_antique_pass();
next_antique_time = dtime() + antique_interval;
}
if (!dont_retry_errors && !retry_errors_now && (dtime() > next_error_time)) {
retry_errors_now = true;
log_messages.printf(MSG_DEBUG,
"starting retry of previous errors\n"
);
}
}
}
int main(int argc, char** argv) {
int retval;
char buf[256];
for (int i=1; i<argc; i++) {
if (!strcmp(argv[i], "--app_name")) {
app_name = argv[++i];
} else if (!strcmp(argv[i], "--lo")) {
lo = atoi(argv[++i]);
} else if (!strcmp(argv[i], "--hi")) {
hi = atoi(argv[++i]);
} else if (!strcmp(argv[i], "-d")) {
log_messages.set_debug_level(atoi(argv[++i]));
} else if (!strcmp(argv[i], "--debug_leveld")) {
log_messages.set_debug_level(atoi(argv[++i]));
} else if (!strcmp(argv[i], "--sleep_time")) {
sleep_time = atoi(argv[++i]);
} else if (!strcmp(argv[i], "--random_order")) {
order_clause = " order by random ";
} else if (!strcmp(argv[i], "--priority_asc")) {
order_clause = " order by priority asc ";
} else if (!strcmp(argv[i], "--priority_order")) {
order_clause = " order by priority desc ";
} else if (!strcmp(argv[i], "--priority_order_create_time")) {
order_clause = " order by priority desc, workunitid ";
} else {
usage();
}
}
if (!app_name || !lo || !hi || !sleep_time) {
usage();
}
log_messages.printf(MSG_NORMAL, "Starting\n");
retval = config.parse_file();
if (retval) {
log_messages.printf(MSG_CRITICAL,
"Can't parse config.xml: %s\n", boincerror(retval)
);
exit(1);
}
retval = boinc_db.open(
config.db_name, config.db_host, config.db_user, config.db_passwd
);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"boinc_db.open: %d; %s\n", retval, boinc_db.error_string()
);
exit(1);
}
sprintf(buf, "where name='%s'", app_name);
if (app.lookup(buf)) {
log_messages.printf(MSG_CRITICAL, "no such app: %s\n", app_name);
exit(1);
}
if (app.n_size_classes < 2) {
log_messages.printf(MSG_CRITICAL, "app '%s' is not multi-size\n", app_name);
exit(1);
}
while (1) {
bool action;
retval = do_pass(action);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"do_pass(): %s", boincerror(retval)
);
exit(1);
}
if (!action) {
log_messages.printf(MSG_NORMAL, "sleeping\n");
daemon_sleep(sleep_time);
}
}
}
// Write the client_state.xml file
//
int CLIENT_STATE::write_state_file() {
MFILE mf;
int retval, ret1, ret2, attempt;
#ifdef _WIN32
char win_error_msg[4096];
#endif
for (attempt=1; attempt<=MAX_STATE_FILE_WRITE_ATTEMPTS; attempt++) {
if (attempt > 1) boinc_sleep(1.0);
if (log_flags.statefile_debug) {
msg_printf(0, MSG_INFO,
"[statefile] Writing state file"
);
}
#ifdef _WIN32
retval = mf.open(STATE_FILE_NEXT, "wc");
#else
retval = mf.open(STATE_FILE_NEXT, "w");
#endif
if (retval) {
if ((attempt == MAX_STATE_FILE_WRITE_ATTEMPTS) || log_flags.statefile_debug) {
msg_printf(0, MSG_INTERNAL_ERROR,
"Can't open %s: %s",
STATE_FILE_NEXT, boincerror(retval)
);
}
if (attempt < MAX_STATE_FILE_WRITE_ATTEMPTS) continue;
return ERR_FOPEN;
}
MIOFILE miof;
miof.init_mfile(&mf);
ret1 = write_state(miof);
ret2 = mf.close();
if (ret1) {
if ((attempt == MAX_STATE_FILE_WRITE_ATTEMPTS) || log_flags.statefile_debug) {
msg_printf(NULL, MSG_INTERNAL_ERROR,
"Couldn't write state file: %s", boincerror(retval)
);
}
if (attempt < MAX_STATE_FILE_WRITE_ATTEMPTS) continue;
return ret1;
}
if (ret2) {
if (attempt < MAX_STATE_FILE_WRITE_ATTEMPTS) continue;
return ret2;
}
// only attempt to rename the current state file if it exists.
//
if (boinc_file_exists(STATE_FILE_NAME)) {
if (boinc_file_exists(STATE_FILE_PREV)) {
retval = boinc_delete_file(STATE_FILE_PREV);
if (retval) {
if ((attempt == MAX_STATE_FILE_WRITE_ATTEMPTS) || log_flags.statefile_debug) {
#ifdef _WIN32
msg_printf(0, MSG_INFO,
"Can't delete previous state file; %s",
windows_format_error_string(GetLastError(), win_error_msg, sizeof(win_error_msg))
);
#else
msg_printf(0, MSG_INFO,
"Can't delete previous state file: %s",
strerror(errno)
);
#endif
}
if (attempt < MAX_STATE_FILE_WRITE_ATTEMPTS) continue;
}
}
retval = boinc_rename(STATE_FILE_NAME, STATE_FILE_PREV);
if (retval) {
if ((attempt == MAX_STATE_FILE_WRITE_ATTEMPTS) || log_flags.statefile_debug) {
#ifdef _WIN32
msg_printf(0, MSG_INFO,
"Can't rename current state file to previous state file; %s",
windows_format_error_string(GetLastError(), win_error_msg, sizeof(win_error_msg))
);
#else
msg_printf(0, MSG_INFO,
"Can't rename current state file to previous state file: %s",
strerror(errno)
);
#endif
}
if (attempt < MAX_STATE_FILE_WRITE_ATTEMPTS) continue;
}
}
retval = boinc_rename(STATE_FILE_NEXT, STATE_FILE_NAME);
if (log_flags.statefile_debug) {
msg_printf(0, MSG_INFO,
"[statefile] Done writing state file"
);
}
if (!retval) break; // Success!
if ((attempt == MAX_STATE_FILE_WRITE_ATTEMPTS) || log_flags.statefile_debug) {
#ifdef _WIN32
msg_printf(0, MSG_INFO,
"rename error: %s",
windows_format_error_string(GetLastError(), win_error_msg, sizeof(win_error_msg))
);
#elif defined (__APPLE__)
if (log_flags.statefile_debug) {
system("ls -al /Library/Application\\ Support/BOINC\\ Data/client*.*");
}
#endif
}
if (attempt < MAX_STATE_FILE_WRITE_ATTEMPTS) continue;
return ERR_RENAME;
}
return 0;
}
// Based on the info in the request message,
// look up the host and its user, and make sure the authenticator matches.
// Some special cases:
// 1) If no host ID is supplied, or if RPC seqno mismatch,
// create a new host record
// 2) If the host record specified by g_request->hostid is a "zombie"
// (i.e. it was merged with another host via the web site)
// then follow links to find the proper host
//
// POSTCONDITION:
// If this function returns zero, then:
// - reply.host contains a valid host record (possibly new)
// - reply.user contains a valid user record
// - if user belongs to a team, reply.team contains team record
//
int authenticate_user() {
int retval;
char buf[1024];
DB_HOST host;
DB_USER user;
DB_TEAM team;
if (g_request->hostid) {
retval = host.lookup_id(g_request->hostid);
while (!retval && host.userid==0) {
// if host record is zombie, follow link to new host
//
retval = host.lookup_id(host.rpc_seqno);
if (!retval) {
g_reply->hostid = host.id;
log_messages.printf(MSG_NORMAL,
"[HOST#%d] forwarding to new host ID %d\n",
g_request->hostid, host.id
);
}
}
if (retval) {
g_reply->insert_message("Can't find host record", "low");
log_messages.printf(MSG_NORMAL,
"[HOST#%d?] can't find host\n",
g_request->hostid
);
g_request->hostid = 0;
goto lookup_user_and_make_new_host;
}
g_reply->host = host;
// look up user based on the ID in host record,
// and see if the authenticator matches (regular or weak)
//
g_request->using_weak_auth = false;
sprintf(buf, "where id=%d", host.userid);
retval = user.lookup(buf);
if (!retval && !strcmp(user.authenticator, g_request->authenticator)) {
// req auth matches user auth - go on
} else {
if (!retval) {
// user for host.userid exists - check weak auth
//
get_weak_auth(user, buf);
if (!strcmp(buf, g_request->authenticator)) {
g_request->using_weak_auth = true;
log_messages.printf(MSG_DEBUG,
"[HOST#%d] accepting weak authenticator\n",
host.id
);
}
}
if (!g_request->using_weak_auth) {
// weak auth failed - look up user based on authenticator
//
strlcpy(
user.authenticator, g_request->authenticator, sizeof(user.authenticator)
);
escape_string(user.authenticator, sizeof(user.authenticator));
sprintf(buf, "where authenticator='%s'", user.authenticator);
retval = user.lookup(buf);
if (retval) {
g_reply->insert_message(
_("Invalid or missing account key. To fix, remove and add this project."),
"notice"
);
g_reply->set_delay(DELAY_MISSING_KEY);
g_reply->nucleus_only = true;
log_messages.printf(MSG_CRITICAL,
"[HOST#%d] [USER#%d] Bad authenticator '%s'\n",
host.id, user.id, g_request->authenticator
);
return ERR_AUTHENTICATOR;
}
}
}
g_reply->user = user;
if (host.userid != user.id) {
// If the request's host ID isn't consistent with the authenticator,
// create a new host record.
//
log_messages.printf(MSG_NORMAL,
"[HOST#%d] [USER#%d] inconsistent host ID; creating new host\n",
host.id, user.id
);
goto make_new_host;
}
// If the seqno from the host is less than what we expect,
// the user must have copied the state file to a different host.
// Make a new host record.
//
if (!batch && g_request->rpc_seqno < g_reply->host.rpc_seqno) {
g_request->hostid = 0;
log_messages.printf(MSG_NORMAL,
"[HOST#%d] [USER#%d] RPC seqno %d less than expected %d; creating new host\n",
g_reply->host.id, user.id, g_request->rpc_seqno, g_reply->host.rpc_seqno
);
goto make_new_host;
}
} else {
// Here no hostid was given, or the ID was bad.
// Look up the user, then create a new host record
//
lookup_user_and_make_new_host:
// if authenticator contains _, it's a weak auth
//
if (strchr(g_request->authenticator, '_')) {
int userid = atoi(g_request->authenticator);
retval = user.lookup_id(userid);
if (!retval) {
get_weak_auth(user, buf);
if (strcmp(buf, g_request->authenticator)) {
retval = ERR_AUTHENTICATOR;
}
}
} else {
strlcpy(
user.authenticator, g_request->authenticator,
sizeof(user.authenticator)
);
escape_string(user.authenticator, sizeof(user.authenticator));
sprintf(buf, "where authenticator='%s'", user.authenticator);
retval = user.lookup(buf);
}
if (retval) {
g_reply->insert_message(
"Invalid or missing account key. To fix, remove and add this project .",
"low"
);
g_reply->set_delay(DELAY_MISSING_KEY);
log_messages.printf(MSG_CRITICAL,
"[HOST#<none>] Bad authenticator '%s': %s\n",
g_request->authenticator, boincerror(retval)
);
return ERR_AUTHENTICATOR;
}
g_reply->user = user;
// If host CPID is present,
// scan backwards through this user's hosts,
// looking for one with the same host CPID.
// If we find one, it means the user detached and reattached.
// Use the existing host record,
// and mark in-progress results as over.
//
if (strlen(g_request->host.host_cpid)) {
if (find_host_by_cpid(user, g_request->host.host_cpid, host)) {
log_messages.printf(MSG_NORMAL,
"[HOST#%d] [USER#%d] No host ID in request, but host with matching CPID found.\n",
host.id, host.userid
);
if ((g_request->allow_multiple_clients != 1)
&& (g_request->other_results.size() == 0)
) {
mark_results_over(host);
}
goto got_host;
}
}
make_new_host:
// One final attempt to locate an existing host record:
// scan backwards through this user's hosts,
// looking for one with the same host name,
// IP address, processor and amount of RAM.
// If found, use the existing host record,
// and mark in-progress results as over.
//
// NOTE: If the client was run with --allow_multiple_clients, skip this.
//
if ((g_request->allow_multiple_clients != 1)
&& find_host_by_other(user, g_request->host, host)
) {
log_messages.printf(MSG_NORMAL,
"[HOST#%d] [USER#%d] Found similar existing host for this user - assigned.\n",
host.id, host.userid
);
mark_results_over(host);
goto got_host;
}
// either of the above cases,
// or host ID didn't match user ID,
// or RPC seqno was too low.
//
// Create a new host.
// g_reply->user is filled in and valid at this point
//
host = g_request->host;
host.id = 0;
host.create_time = time(0);
host.userid = g_reply->user.id;
host.rpc_seqno = 0;
host.expavg_time = time(0);
safe_strcpy(host.venue, g_reply->user.venue);
host.fix_nans();
retval = host.insert();
if (retval) {
g_reply->insert_message(
"Couldn't create host record in database", "low"
);
boinc_db.print_error("host.insert()");
log_messages.printf(MSG_CRITICAL, "host.insert() failed\n");
return retval;
}
host.id = boinc_db.insert_id();
got_host:
g_reply->host = host;
g_reply->hostid = g_reply->host.id;
// this tells client to updates its host ID
g_request->rpc_seqno = 0;
// this value eventually gets written to host DB record;
// for new hosts it must be zero.
// This kludge forces this.
}
// have user record in g_reply->user at this point
//
if (g_reply->user.teamid) {
retval = team.lookup_id(g_reply->user.teamid);
if (!retval) g_reply->team = team;
}
// compute email hash
//
md5_block(
(unsigned char*)g_reply->user.email_addr,
strlen(g_reply->user.email_addr),
g_reply->email_hash
);
// if new user CPID, update user record
//
if (!g_request->using_weak_auth && strlen(g_request->cross_project_id)) {
if (strcmp(g_request->cross_project_id, g_reply->user.cross_project_id)) {
user.id = g_reply->user.id;
escape_string(g_request->cross_project_id, sizeof(g_request->cross_project_id));
sprintf(buf, "cross_project_id='%s'", g_request->cross_project_id);
unescape_string(g_request->cross_project_id, sizeof(g_request->cross_project_id));
user.update_field(buf);
}
}
return 0;
}
/// Start a task in a slot directory.
/// This includes setting up soft links,
/// passing preferences, and starting the process.
///
/// Current dir is top-level Synecdoche dir.
///
/// \post
/// - If any error occurs
/// - #task_state is #PROCESS_COULDNT_START
/// - CLIENT_STATE::report_result_error() is called
/// - else
/// - #task_state is #PROCESS_EXECUTING
///
/// \return 0 on success, nonzero otherwise.
int ACTIVE_TASK::start() {
char exec_name[256], exec_path[256];
unsigned int i;
FILE_REF fref;
int retval;
// F*** goto, need to define some variables here instead of where they are used!
std::ostringstream err_stream;
#ifdef _WIN32
std::string cmd_line;
std::string slotdirpath;
#else
// Needs to be defined here because those gotos would skip the
// initialization of 'cmdline' and 'argv' if it would be defined later.
std::ostringstream cmdline;
std::list<std::string> argv;
#endif
if ((!full_init_done) && (log_flags.task)) {
msg_printf(wup->project, MSG_INFO,
"Starting %s", result->name
);
}
if (log_flags.cpu_sched) {
msg_printf(wup->project, MSG_INFO,
"[cpu_sched] Starting %s%s", result->name, (full_init_done) ? " (resume)" : " (initial)"
);
}
// Always check if all required files are present. If not, trigger
// re-downloads and don't start the science application.
FILE_INFO_PSET missing_file_infos;
retval = gstate.input_files_available(result, true, &missing_file_infos);
if (retval) {
for (FILE_INFO_PSET::iterator it = missing_file_infos.begin(); it != missing_file_infos.end(); ++it) {
FILE_INFO* fip = *it;
if (fip) {
err_stream << "Input file " << fip->name
<< " missing or invalid: " << retval;
} else {
err_stream << "Input file missing or invalid";
// We can't trigger a new download if we don't have
// any file information. Just fail here as before.
goto error;
}
fip->status = FILE_NOT_PRESENT;
}
}
if (!missing_file_infos.empty()) {
// Some files are missing and are set for re-transfer.
// Update status and return without error.
result->set_state(RESULT_FILES_DOWNLOADING, "start");
set_task_state(PROCESS_UNINITIALIZED, "start");
next_scheduler_state = PROCESS_UNINITIALIZED;
return 0;
}
if (!full_init_done) {
checkpoint_cpu_time = 0;
checkpoint_wall_time = gstate.now;
}
current_cpu_time = checkpoint_cpu_time;
episode_start_cpu_time = checkpoint_cpu_time;
debt_interval_start_cpu_time = checkpoint_cpu_time;
graphics_request_queue.init(result->name); // reset message queues
process_control_queue.init(result->name);
if (!app_client_shm.shm) {
retval = get_shmem_seg_name();
if (retval) {
err_stream << "Can't get shared memory segment name: " << boincerror(retval);
goto error;
}
}
// this must go AFTER creating shmem name,
// since the shmem name is part of the file
//
retval = write_app_init_file();
if (retval) {
err_stream << "Can't write init file: " << retval;
goto error;
}
// set up applications files
//
strcpy(exec_name, "");
for (i=0; i<app_version->app_files.size(); i++) {
fref = app_version->app_files[i];
FILE_INFO* fip = fref.file_info;
std::string file_path = get_pathname(fip);
if (fref.main_program) {
if (is_image_file(fip->name)) {
err_stream << "Main program " << fip->name << " is an image file";
retval = ERR_NO_SIGNATURE;
goto error;
}
if (!fip->executable && !wup->project->anonymous_platform) {
err_stream << "Main program " << fip->name << " is not executable";
retval = ERR_NO_SIGNATURE;
goto error;
}
safe_strcpy(exec_name, fip->name.c_str());
safe_strcpy(exec_path, file_path.c_str());
}
// anonymous platform may use different files than
// when the result was started, so link files even if not first time
if ((!full_init_done) || (wup->project->anonymous_platform)) {
retval = setup_file(result->project, fip, fref, file_path, slot_dir, true);
if (retval) {
err_stream << "Can't link input file";
goto error;
}
}
}
if (!strlen(exec_name)) {
err_stream << "No main program specified";
retval = ERR_NOT_FOUND;
goto error;
}
// set up input, output files
if (!full_init_done) {
for (i=0; i<wup->input_files.size(); i++) {
fref = wup->input_files[i];
const FILE_INFO* fip = fref.file_info;
std::string file_path = get_pathname(fref.file_info);
retval = setup_file(result->project, fip, fref, file_path, slot_dir, true);
if (retval) {
err_stream << "Can't link input file";
goto error;
}
}
for (i=0; i<result->output_files.size(); i++) {
fref = result->output_files[i];
if (fref.copy_file) continue;
const FILE_INFO* fip = fref.file_info;
std::string file_path = get_pathname(fref.file_info);
retval = setup_file(result->project, fip, fref, file_path, slot_dir, false);
if (retval) {
err_stream << "Can't link output file";
goto error;
}
}
full_init_done = true;
}
link_user_files();
if (gstate.exit_before_start) {
exit(0);
}
#ifdef _WIN32
PROCESS_INFORMATION process_info;
STARTUPINFO startup_info;
LPVOID environment_block = NULL;
char error_msg[1024];
char error_msg2[1024];
memset(&process_info, 0, sizeof(process_info));
memset(&startup_info, 0, sizeof(startup_info));
startup_info.cb = sizeof(startup_info);
// suppress 2-sec rotating hourglass cursor on startup
//
startup_info.dwFlags = STARTF_FORCEOFFFEEDBACK;
app_client_shm.reset_msgs();
if (config.run_apps_manually) {
// fill in core client's PID so we won't think app has exited
pid = GetCurrentProcessId();
pid_handle = GetCurrentProcess();
set_task_state(PROCESS_EXECUTING, "start");
return 0;
}
// NOTE: in Windows, stderr is redirected in boinc_init_diagnostics();
cmd_line = exec_path + std::string(" ") + wup->command_line;
if (strlen(app_version->cmdline)) {
cmd_line += std::string(" ") + app_version->cmdline;
}
slotdirpath = relative_to_absolute(slot_dir);
bool success = false;
for (i=0; i<5; i++) {
if (sandbox_account_service_token != NULL) {
// Find CreateEnvironmentBlock/DestroyEnvironmentBlock pointers
tCEB pCEB = NULL;
tDEB pDEB = NULL;
HMODULE hUserEnvLib = NULL;
hUserEnvLib = LoadLibrary("userenv.dll");
if (hUserEnvLib) {
pCEB = (tCEB) GetProcAddress(hUserEnvLib, "CreateEnvironmentBlock");
pDEB = (tDEB) GetProcAddress(hUserEnvLib, "DestroyEnvironmentBlock");
}
if (!pCEB(&environment_block, sandbox_account_service_token, FALSE)) {
if (log_flags.task) {
windows_error_string(error_msg, sizeof(error_msg));
msg_printf(wup->project, MSG_INFO,
"Process environment block creation failed: %s", error_msg
);
}
}
if (CreateProcessAsUser(
sandbox_account_service_token,
exec_path,
(LPSTR)cmd_line.c_str(),
NULL,
NULL,
FALSE,
CREATE_NEW_PROCESS_GROUP|CREATE_NO_WINDOW|IDLE_PRIORITY_CLASS|CREATE_UNICODE_ENVIRONMENT,
environment_block,
slotdirpath.c_str(),
&startup_info,
&process_info
)) {
success = true;
break;
} else {
windows_error_string(error_msg, sizeof(error_msg));
msg_printf(wup->project, MSG_INTERNAL_ERROR,
"Process creation failed: %s", error_msg
);
}
if (!pDEB(environment_block)) {
if (log_flags.task) {
windows_error_string(error_msg, sizeof(error_msg2));
msg_printf(wup->project, MSG_INFO,
"Process environment block cleanup failed: %s",
error_msg2
);
}
}
if (hUserEnvLib) {
pCEB = NULL;
pDEB = NULL;
FreeLibrary(hUserEnvLib);
}
} else {
if (CreateProcess(
exec_path,
(LPSTR)cmd_line.c_str(),
NULL,
NULL,
FALSE,
CREATE_NEW_PROCESS_GROUP|CREATE_NO_WINDOW|IDLE_PRIORITY_CLASS,
NULL,
slotdirpath.c_str(),
&startup_info,
&process_info
)) {
success = true;
break;
} else {
windows_error_string(error_msg, sizeof(error_msg));
msg_printf(wup->project, MSG_INTERNAL_ERROR,
"Process creation failed: %s", error_msg
);
}
}
boinc_sleep(drand());
}
if (!success) {
err_stream << "CreateProcess() failed - " << error_msg;
retval = ERR_EXEC;
goto error;
}
pid = process_info.dwProcessId;
pid_handle = process_info.hProcess;
CloseHandle(process_info.hThread); // thread handle is not used
#else
// Unix/Linux/Mac case
// Set up core/app shared memory seg if needed
//
if (!app_client_shm.shm) {
if (app_version->api_major_version() >= 6) {
// Use mmap() shared memory
std::string buf = slot_dir + std::string("/") + std::string(MMAPPED_FILE_NAME);
if (g_use_sandbox) {
if (!boinc_file_exists(buf.c_str())) {
int fd = open(buf.c_str(), O_RDWR | O_CREAT, 0660);
if (fd >= 0) {
close (fd);
#ifdef SANDBOX
set_to_project_group(buf.c_str());
#endif
}
}
}
retval = create_shmem_mmap(
buf.c_str(), sizeof(SHARED_MEM), (void**)&app_client_shm.shm
);
} else {
// Use shmget() shared memory
retval = create_shmem(
shmem_seg_name, sizeof(SHARED_MEM), gstate.boinc_project_gid,
(void**)&app_client_shm.shm
);
if (retval) {
needs_shmem = true;
destroy_shmem(shmem_seg_name);
return retval;
}
}
needs_shmem = false;
}
app_client_shm.reset_msgs();
#if (defined (__APPLE__) && (defined(__i386__) || defined(__x86_64__)))
// PowerPC apps emulated on i386 Macs crash if running graphics
powerpc_emulated_on_i386 = ! is_native_i386_app(exec_path);
#endif
if (config.run_apps_manually) {
pid = getpid(); // use the client's PID
set_task_state(PROCESS_EXECUTING, "start");
return 0;
}
// Prepare command line for the science app:
cmdline << wup->command_line;
if (strlen(app_version->cmdline)) {
cmdline << ' ' << app_version->cmdline;
}
argv = parse_command_line(cmdline.str().c_str());
if (log_flags.task_debug) {
debug_print_argv(argv);
}
pid = fork();
if (pid == -1) {
err_stream << "fork() failed: " << strerror(errno);
retval = ERR_FORK;
goto error;
}
if (pid == 0) {
// from here on we're running in a new process.
// If an error happens,
// exit nonzero so that the core client knows there was a problem.
// don't pass stdout to the app
//
int fd = open("/dev/null", O_RDWR);
dup2(fd, STDOUT_FILENO);
close(fd);
// add to library path:
// - the project dir (../../projects/X)
// - the slot dir (.)
// - the Synecdoche dir (../..)
// We use relative paths in case higher-level dirs
// are not readable to the account under which app runs
//
std::string pdir = get_project_dir(wup->project);
std::ostringstream libpath;
const char* env_lib_path = getenv("LD_LIBRARY_PATH");
if (env_lib_path) {
libpath << env_lib_path << ':';
}
libpath << "../../" << pdir << ":.:../..";
setenv("LD_LIBRARY_PATH", libpath.str().c_str(), 1);
retval = chdir(slot_dir.c_str());
if (retval) {
perror("chdir");
fflush(NULL);
_exit(errno);
}
#if 0
// set stack size limit to the max.
// Some BOINC apps have reported problems with exceeding
// small stack limits (e.g. 8 MB)
// and it seems like the best thing to raise it as high as possible
//
struct rlimit rlim;
#define MIN_STACK_LIMIT 64000000
getrlimit(RLIMIT_STACK, &rlim);
if (rlim.rlim_cur != RLIM_INFINITY && rlim.rlim_cur <= MIN_STACK_LIMIT) {
if (rlim.rlim_max == RLIM_INFINITY || rlim.rlim_max > MIN_STACK_LIMIT) {
rlim.rlim_cur = MIN_STACK_LIMIT;
} else {
rlim.rlim_cur = rlim.rlim_max;
}
setrlimit(RLIMIT_STACK, &rlim);
}
#endif
// hook up stderr to a specially-named file
//
freopen(STDERR_FILE, "a", stderr);
// set idle process priority
#ifdef HAVE_SETPRIORITY
if (setpriority(PRIO_PROCESS, 0, PROCESS_IDLE_PRIORITY)) {
perror("setpriority");
}
#endif
std::string path = std::string("../../") + std::string(exec_path);
if (g_use_sandbox) {
std::ostringstream switcher_path;
switcher_path << "../../" << SWITCHER_DIR << '/' << SWITCHER_FILE_NAME;
argv.push_front(exec_name);
argv.push_front(path);
argv.push_front(SWITCHER_FILE_NAME);
// Files written by projects have user boinc_project and group boinc_project,
// so they must be world-readable so Synecdoche can read them.
umask(2);
retval = do_execv(switcher_path.str(), argv);
} else {
argv.push_front(exec_name);
retval = do_execv(path, argv);
}
msg_printf(wup->project, MSG_INTERNAL_ERROR,
"Process creation (%s) failed: %s, errno=%d\n", path.c_str(), boincerror(retval), errno
);
perror("execv");
fflush(NULL);
_exit(errno);
}
if (log_flags.task_debug) {
msg_printf(wup->project, MSG_INFO,
"[task_debug] ACTIVE_TASK::start(): forked process: pid %d\n", pid
);
}
#endif
set_task_state(PROCESS_EXECUTING, "start");
return 0;
// go here on error; "error_msg" contains error message, "retval" is nonzero
//
error:
// if something failed, it's possible that the executable was munged.
// Verify it to trigger another download.
//
gstate.input_files_available(result, true);
gstate.report_result_error(*result, "%s", err_stream.str().c_str());
set_task_state(PROCESS_COULDNT_START, "start");
return retval;
}
int main(int argc, char *argv[]) {
int retval;
R_RSA_PUBLIC_KEY key;
char log_path[MAXPATHLEN];
#ifdef _USING_FCGI_
unsigned int counter=0;
#endif
for(int c = 1; c < argc; c++) {
std::string option(argv[c]);
if(option == "-v" || option == "--version") {
printf("%s\n", SVN_VERSION);
exit(0);
} else if(option == "-h" || option == "--help") {
usage(argv[0]);
exit(0);
} else if (option.length()){
fprintf(stderr, "unknown command line argument: %s\n\n", argv[c]);
usage(argv[0]);
exit(1);
}
}
installer();
get_log_path(log_path, "file_upload_handler.log");
#ifndef _USING_FCGI_
if (!freopen(log_path, "a", stderr)) {
fprintf(stderr, "Can't open log file '%s' (errno: %d)\n",
log_path, errno
);
return_error(ERR_TRANSIENT, "can't open log file '%s' (errno: %d)",
log_path, errno
);
exit(1);
}
#else
FCGI_FILE *f = FCGI::fopen(log_path, "a");
if (f) {
log_messages.redirect(f);
} else {
fprintf(stderr, "Can't redirect FCGI log messages\n");
return_error(ERR_TRANSIENT, "can't open log file (FCGI)");
exit(1);
}
#endif
retval = config.parse_file();
if (retval) {
fprintf(stderr, "Can't parse config.xml: %s\n", boincerror(retval));
return_error(ERR_TRANSIENT,
"can't parse config file", log_path, errno
);
exit(1);
}
log_messages.pid = getpid();
log_messages.set_debug_level(config.fuh_debug_level);
if (boinc_file_exists(config.project_path("stop_upload"))) {
return_error(ERR_TRANSIENT, "Maintenance underway: file uploads are temporarily disabled.");
exit(1);
}
if (!config.ignore_upload_certificates) {
retval = get_key(key);
if (retval) {
return_error(ERR_TRANSIENT, "can't read key file");
exit(1);
}
}
#ifdef _USING_FCGI_
while(FCGI_Accept() >= 0) {
counter++;
//fprintf(stderr, "file_upload_handler (FCGI): counter: %d\n", counter);
log_messages.set_indent_level(0);
#endif
handle_request(stdin, key);
#ifdef _USING_FCGI_
// flush log for FCGI, otherwise it just buffers a lot
log_messages.flush();
}
// when exiting, write headers back to apache so it won't complain
// about "incomplete headers"
fprintf(stdout,"Content-type: text/plain\n\n");
#endif
return 0;
}
// 1) Decide which global prefs to use for sched decisions: either
// - <working_global_prefs> from request msg
// - <global_prefs> from request message
// - prefs from user DB record
// and parse them into g_request->global_prefs.
// 2) update prefs in user record if needed
// 2) send global prefs in reply msg if needed
//
int handle_global_prefs() {
char buf[BLOB_SIZE+256];
g_reply->send_global_prefs = false;
bool have_working_prefs = (strlen(g_request->working_global_prefs_xml)>0);
bool have_master_prefs = (strlen(g_request->global_prefs_xml)>0);
// absent if the host has host-specific prefs
bool have_db_prefs = (strlen(g_reply->user.global_prefs)>0);
bool same_account = !strcmp(
g_request->global_prefs_source_email_hash, g_reply->email_hash
);
double master_mod_time=0, db_mod_time=0, working_mod_time=0;
if (have_master_prefs) {
parse_double(g_request->global_prefs_xml, "<mod_time>", master_mod_time);
if (master_mod_time > dtime()) master_mod_time = dtime();
}
if (have_working_prefs) {
parse_double(g_request->working_global_prefs_xml, "<mod_time>", working_mod_time);
if (working_mod_time > dtime()) working_mod_time = dtime();
}
if (have_db_prefs) {
parse_double(g_reply->user.global_prefs, "<mod_time>", db_mod_time);
if (db_mod_time > dtime()) db_mod_time = dtime();
}
if (config.debug_prefs) {
log_messages.printf(MSG_NORMAL,
"[prefs] have_master:%d have_working: %d have_db: %d\n",
have_master_prefs, have_working_prefs, have_db_prefs
);
}
// decide which prefs to use for sched decisions,
// and parse them into g_request->global_prefs
//
if (have_working_prefs) {
g_request->global_prefs.parse(g_request->working_global_prefs_xml, "");
if (config.debug_prefs) {
log_messages.printf(MSG_NORMAL, "[prefs] using working prefs\n");
}
} else {
if (have_master_prefs) {
if (have_db_prefs && db_mod_time > master_mod_time) {
g_request->global_prefs.parse(g_reply->user.global_prefs, g_reply->host.venue);
if (config.debug_prefs) {
log_messages.printf(MSG_NORMAL,
"[prefs] using db prefs - more recent\n"
);
}
} else {
g_request->global_prefs.parse(g_request->global_prefs_xml, g_reply->host.venue);
if (config.debug_prefs) {
log_messages.printf(MSG_NORMAL,
"[prefs] using master prefs\n"
);
}
}
} else {
if (have_db_prefs) {
g_request->global_prefs.parse(g_reply->user.global_prefs, g_reply->host.venue);
if (config.debug_prefs) {
log_messages.printf(MSG_NORMAL, "[prefs] using db prefs\n");
}
} else {
g_request->global_prefs.defaults();
if (config.debug_prefs) {
log_messages.printf(MSG_NORMAL, "[prefs] using default prefs\n");
}
}
}
}
// decide whether to update DB
//
if (!g_request->using_weak_auth && have_master_prefs) {
bool update_user_record = false;
if (have_db_prefs) {
if (master_mod_time > db_mod_time && same_account) {
update_user_record = true;
}
} else {
if (same_account) update_user_record = true;
}
if (update_user_record) {
if (config.debug_prefs) {
log_messages.printf(MSG_NORMAL, "[prefs] updating db prefs\n");
}
safe_strcpy(g_reply->user.global_prefs, g_request->global_prefs_xml);
DB_USER user;
user.id = g_reply->user.id;
escape_string(g_request->global_prefs_xml, sizeof(g_request->global_prefs_xml));
sprintf(buf, "global_prefs='%s'", g_request->global_prefs_xml);
unescape_string(g_request->global_prefs_xml, sizeof(g_request->global_prefs_xml));
int retval = user.update_field(buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"user.update_field() failed: %s\n", boincerror(retval)
);
}
}
}
// decide whether to send DB prefs in reply msg
//
if (config.debug_prefs) {
log_messages.printf(MSG_NORMAL,
"[prefs] have DB prefs: %d; dbmod %f; global mod %f; working mod %f\n",
have_db_prefs, db_mod_time, g_request->global_prefs.mod_time, working_mod_time
);
}
if (have_db_prefs && db_mod_time > master_mod_time && db_mod_time > working_mod_time) {
if (config.debug_prefs) {
log_messages.printf(MSG_DEBUG,
"[prefs] sending DB prefs in reply\n"
);
}
g_reply->send_global_prefs = true;
}
return 0;
}
