// Berkeley Open Infrastructure for Network Computing // http://boinc.berkeley.edu // Copyright (C) 2005 University of California // // This is free software; you can redistribute it and/or // modify it under the terms of the GNU Lesser General Public // License as published by the Free Software Foundation; // either version 2.1 of the License, or (at your option) any later version. // // This software is distributed in the hope that it will be useful, // but WITHOUT ANY WARRANTY; without even the implied warranty of // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. // See the GNU Lesser General Public License for more details. // // To view the GNU Lesser General Public License visit // http://www.gnu.org/copyleft/lesser.html // or write to the Free Software Foundation, Inc., // 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA // Handle a scheduling server RPC #include "config.h" #include #include #include #include #include #include using namespace std; #include #include #include #include #include #include #include "boinc_db.h" #include "backend_lib.h" #include "error_numbers.h" #include "parse.h" #include "str_util.h" #include "util.h" #include "filesys.h" #include "main.h" #include "server_types.h" #include "sched_util.h" #include "handle_request.h" #include "sched_msgs.h" #include "sched_resend.h" #include "sched_send.h" #include "sched_config.h" #include "sched_locality.h" #include "sched_result.h" #include "time_stats_log.h" #ifdef _USING_FCGI_ #include "fcgi_stdio.h" #endif // find the user's most recently-created host with given various characteristics // static bool find_host_by_other(DB_USER& user, HOST req_host, DB_HOST& host) { char buf[2048]; char dn[512], ip[512], os[512], pm[512]; // Only check if the fields are populated if (strlen(req_host.domain_name) && strlen(req_host.last_ip_addr) && strlen(req_host.os_name) && strlen(req_host.p_model)) { strcpy(dn, req_host.domain_name); escape_string(dn, 512); strcpy(ip, req_host.last_ip_addr); escape_string(ip, 512); strcpy(os, req_host.os_name); escape_string(os, 512); strcpy(pm, req_host.p_model); escape_string(pm, 512); sprintf(buf, "where userid=%d and id>%d and domain_name='%s' and last_ip_addr = '%s' and os_name = '%s' and p_model = '%s' and m_nbytes = %lf order by id desc", user.id, req_host.id, dn, ip, os, pm, req_host.m_nbytes ); if (!host.enumerate(buf)) { host.end_enumerate(); return true; } } return false; } static void get_weak_auth(USER& user, char* buf) { char buf2[256], out[256]; sprintf(buf2, "%s%s", user.authenticator, user.passwd_hash); md5_block((unsigned char*)buf2, strlen(buf2), out); sprintf(buf, "%d_%s", user.id, out); } static void send_error_message(SCHEDULER_REPLY& reply, char* msg, int delay) { USER_MESSAGE um(msg, "low"); reply.insert_message(um); reply.set_delay(delay); reply.nucleus_only = true; } // Try to lock a file with name based on host ID, // to prevent 2 schedulers from running at same time for same host. // Return: // 0 if successful // In this case store file descriptor in reply struct so we can unlock later // In other cases store -1 in reply struct // PID (>0) if another process has lock // -1 if error (e.g. can't create file) // int lock_sched(SCHEDULER_REPLY& reply) { char filename[256]; char pid_string[16]; int fd, pid, count; reply.lockfile_fd=-1; sprintf(filename, "%s/CGI_%07d", config.sched_lockfile_dir, reply.host.id); fd = open(filename, O_WRONLY|O_CREAT, S_IRUSR|S_IWUSR|S_IRGRP|S_IROTH); if (fd < 0) return -1; // if we can't get an advisory write lock on the file, // return the PID of the process that DOES hold the lock. // (or -1 if failure) // pid = mylockf(fd); if (pid) { close(fd); return pid; } // write PID into the CGI_ file and flush to disk // count = sprintf(pid_string, "%d\n", getpid()); write(fd, pid_string, count); fsync(fd); reply.lockfile_fd = fd; return 0; } // unlock and delete per-host lockfile // void unlock_sched(SCHEDULER_REPLY& reply) { char filename[256]; if (reply.lockfile_fd < 0) return; sprintf(filename, "%s/CGI_%07d", config.sched_lockfile_dir, reply.host.id); unlink(filename); close(reply.lockfile_fd); } // find the user's most recently-created host with given host CPID // static bool find_host_by_cpid(DB_USER& user, char* host_cpid, DB_HOST& host) { char buf[256], buf2[256]; sprintf(buf, "%s%s", host_cpid, user.email_addr); md5_block((const unsigned char*)buf, strlen(buf), buf2); sprintf(buf, "where userid=%d and host_cpid='%s' order by id desc", user.id, buf2 ); if (!host.enumerate(buf)) { host.end_enumerate(); return true; } return false; } // Called when there's evidence that the host has detached. // Mark in-progress results for the given host // as server state OVER, outcome CLIENT_DETACHED. // This serves two purposes: // 1) make sure we don't resend these results to the host // (they may be the reason the user detached) // 2) trigger the generation of new results for these WUs // static void mark_results_over(DB_HOST& host) { char buf[256], buf2[256]; DB_RESULT result; sprintf(buf, "where hostid=%d and server_state=%d", host.id, RESULT_SERVER_STATE_IN_PROGRESS ); while (!result.enumerate(buf)) { sprintf(buf2, "server_state=%d, outcome=%d", RESULT_SERVER_STATE_OVER, RESULT_OUTCOME_CLIENT_DETACHED ); result.update_field(buf2); // and trigger WU transition // DB_WORKUNIT wu; wu.id = result.workunitid; sprintf(buf2, "transition_time=%d", (int)time(0)); wu.update_field(buf2); log_messages.printf( SCHED_MSG_LOG::MSG_CRITICAL, "[HOST#%d] [RESULT#%d] [WU#%d] changed CPID: marking in-progress result %s as client error!\n", host.id, result.id, result.workunitid, result.name ); } } // 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 sreq.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(SCHEDULER_REQUEST& sreq, SCHEDULER_REPLY& reply) { int retval; char buf[256]; DB_HOST host; DB_USER user; DB_TEAM team; if (sreq.hostid) { retval = host.lookup_id(sreq.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) { reply.hostid = host.id; log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "[HOST#%d] forwarding to new host ID %d\n", sreq.hostid, host.id ); } } if (retval) { USER_MESSAGE um("Can't find host record", "low"); reply.insert_message(um); log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "[HOST#%d?] can't find host\n", sreq.hostid ); sreq.hostid = 0; goto lookup_user_and_make_new_host; } reply.host = host; log_messages.printf( SCHED_MSG_LOG::MSG_DEBUG, "Request [HOST#%d] Database [HOST#%d] Request [RPC#%d] Database [RPC#%d]\n", sreq.hostid, host.id, sreq.rpc_seqno, host.rpc_seqno ); // look up user based on the ID in host record, // and see if the authenticator matches (regular or weak) // sprintf(buf, "where id=%d", host.userid); retval = user.lookup(buf); if (!retval && !strcmp(user.authenticator, sreq.authenticator)) { // req auth matches user auth - go on } else { bool weak_auth = false; if (!retval) { // user for host.userid exists - check weak auth // get_weak_auth(user, buf); if (!strcmp(buf, sreq.authenticator)) { weak_auth = true; log_messages.printf( SCHED_MSG_LOG::MSG_DEBUG, "[HOST#%d] accepting weak authenticator\n", host.id ); } } if (!weak_auth) { // weak auth failed - look up user based on authenticator // strlcpy( user.authenticator, sreq.authenticator, sizeof(user.authenticator) ); sprintf(buf, "where authenticator='%s'", user.authenticator); retval = user.lookup(buf); if (retval) { USER_MESSAGE um("Invalid or missing account key. " "Visit this project's web site to get an account key.", "high" ); reply.insert_message(um); reply.set_delay(DELAY_MISSING_KEY); reply.nucleus_only = true; log_messages.printf( SCHED_MSG_LOG::MSG_CRITICAL, "[HOST#%d] [USER#%d] Bad authenticator '%s'\n", host.id, user.id, sreq.authenticator ); return ERR_AUTHENTICATOR; } } } 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( SCHED_MSG_LOG::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 && sreq.rpc_seqno < reply.host.rpc_seqno) { sreq.hostid = 0; log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "[HOST#%d] [USER#%d] RPC seqno %d less than expected %d; creating new host\n", reply.host.id, user.id, sreq.rpc_seqno, 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(sreq.authenticator, '_')) { int userid = atoi(sreq.authenticator); retval = user.lookup_id(userid); if (!retval) { get_weak_auth(user, buf); if (strcmp(buf, sreq.authenticator)) { retval = ERR_AUTHENTICATOR; } } } else { strlcpy( user.authenticator, sreq.authenticator, sizeof(user.authenticator) ); sprintf(buf, "where authenticator='%s'", user.authenticator); retval = user.lookup(buf); } if (retval) { USER_MESSAGE um( "Invalid or missing account key. " "Visit this project's web site to get an account key.", "low" ); reply.insert_message(um); reply.set_delay(DELAY_MISSING_KEY); log_messages.printf( SCHED_MSG_LOG::MSG_CRITICAL, "[HOST#] Bad authenticator '%s': %d\n", sreq.authenticator, retval ); return ERR_AUTHENTICATOR; } 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(sreq.host.host_cpid)) { if (find_host_by_cpid(user, sreq.host.host_cpid, host)) { log_messages.printf( SCHED_MSG_LOG::MSG_CRITICAL, "[HOST#%d] [USER#%d] User has another host with same CPID.\n", host.id, host.userid ); 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. // if (find_host_by_other(user, sreq.host, host)) { log_messages.printf( SCHED_MSG_LOG::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. // reply.user is filled in and valid at this point // host = sreq.host; host.id = 0; host.create_time = time(0); host.userid = reply.user.id; host.rpc_seqno = 0; host.expavg_time = time(0); host.error_rate = 0.1; strcpy(host.venue, reply.user.venue); host.fix_nans(); retval = host.insert(); if (retval) { USER_MESSAGE um("Couldn't create host record in database", "low"); reply.insert_message(um); boinc_db.print_error("host.insert()"); log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL, "host.insert() failed\n"); return retval; } host.id = boinc_db.insert_id(); got_host: reply.host = host; reply.hostid = reply.host.id; // this tells client to updates its host ID sreq.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 reply.user at this point // if (reply.user.teamid) { retval = team.lookup_id(reply.user.teamid); if (!retval) reply.team = team; } // compute email hash // md5_block( (unsigned char*)reply.user.email_addr, strlen(reply.user.email_addr), reply.email_hash ); // if new user CPID, update user record // if (strlen(sreq.cross_project_id)) { if (strcmp(sreq.cross_project_id, reply.user.cross_project_id)) { user.id = reply.user.id; escape_string(sreq.cross_project_id, sizeof(sreq.cross_project_id)); sprintf(buf, "cross_project_id='%s'", sreq.cross_project_id); unescape_string(sreq.cross_project_id, sizeof(sreq.cross_project_id)); user.update_field(buf); } } return 0; } // somewhat arbitrary formula for credit as a function of CPU time. // Could also include terms for RAM size, network speed etc. // static void compute_credit_rating(HOST& host) { double fpw, intw, scale, x; if (config.use_benchmark_weights) { fpw = config.fp_benchmark_weight; intw = 1. - fpw; // FP benchmark is 2x int benchmark, on average. // Compute a scaling factor the gives the same credit per day // no matter how benchmarks are weighted // scale = 1.5 / (2*intw + fpw); } else { fpw = .5; intw = .5; scale = 1; } x = fpw*fabs(host.p_fpops) + intw*fabs(host.p_iops); x /= 1e9; x *= COBBLESTONE_FACTOR; x /= SECONDS_PER_DAY; x *= scale; host.claimed_credit_per_cpu_sec = x; } // modify host struct based on request. // Copy all fields that are determined by the client. // static int modify_host_struct(SCHEDULER_REQUEST& sreq, HOST& host) { host.timezone = sreq.host.timezone; strncpy(host.domain_name, sreq.host.domain_name, sizeof(host.domain_name)); strncpy(host.serialnum, sreq.host.serialnum, sizeof(host.serialnum)); if (strcmp(host.last_ip_addr, sreq.host.last_ip_addr)) { strncpy(host.last_ip_addr, sreq.host.last_ip_addr, sizeof(host.last_ip_addr)); } else { host.nsame_ip_addr++; } host.on_frac = sreq.host.on_frac; host.connected_frac = sreq.host.connected_frac; host.active_frac = sreq.host.active_frac; host.cpu_efficiency = sreq.host.cpu_efficiency; host.duration_correction_factor = sreq.host.duration_correction_factor; host.p_ncpus = sreq.host.p_ncpus; strncpy(host.p_vendor, sreq.host.p_vendor, sizeof(host.p_vendor)); // unlikely this will change strncpy(host.p_model, sreq.host.p_model, sizeof(host.p_model)); host.p_fpops = sreq.host.p_fpops; host.p_iops = sreq.host.p_iops; host.p_membw = sreq.host.p_membw; strncpy(host.os_name, sreq.host.os_name, sizeof(host.os_name)); strncpy(host.os_version, sreq.host.os_version, sizeof(host.os_version)); host.m_nbytes = sreq.host.m_nbytes; host.m_cache = sreq.host.m_cache; host.m_swap = sreq.host.m_swap; host.d_total = sreq.host.d_total; host.d_free = sreq.host.d_free; host.d_boinc_used_total = sreq.host.d_boinc_used_total; host.d_boinc_used_project = sreq.host.d_boinc_used_project; host.n_bwup = sreq.host.n_bwup; host.n_bwdown = sreq.host.n_bwdown; if (strlen(sreq.host.host_cpid)) { strcpy(host.host_cpid, sreq.host.host_cpid); } host.fix_nans(); compute_credit_rating(host); sreq.host.claimed_credit_per_cpu_sec = host.claimed_credit_per_cpu_sec; return 0; } // update the DB record to the values in "xhost" // "initial_host" stores the current DB values; // update only those fields that have changed // static int update_host_record(HOST& initial_host, HOST& xhost, USER& user) { DB_HOST host; int retval; char buf[1024]; host = xhost; // hash the CPID reported by the host with the user's email address. // This prevents one user from spoofing another one's host. // if (strlen(host.host_cpid)) { sprintf(buf, "%s%s", host.host_cpid, user.email_addr); md5_block((const unsigned char*)buf, strlen(buf), host.host_cpid); } char* p = getenv("REMOTE_ADDR"); if (p) { strlcpy(host.external_ip_addr, p, sizeof(host.external_ip_addr)); } retval = host.update_diff(initial_host); if (retval) { log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL, "host.update() failed: %d\n", retval); } return 0; } // Figure out which of the results the host currently has // should be aborted outright, or aborted if not started yet // int send_result_abort( SCHEDULER_REQUEST& sreq, SCHEDULER_REPLY& reply, SCHED_SHMEM& ) { int aborts_sent = 0; int retval = 0; DB_IN_PROGRESS_RESULT result; std::string result_names; unsigned int i; if (sreq.other_results.size() == 0) { return 0; } // initially mark all results for abort and build list of results to query // for (i=0; i 0) result_names.append(", "); result_names.append("'"); result_names.append(orp.name); result_names.append("'"); } // query the db for the results and set the appropriate flag // while (!(retval = result.enumerate(reply.host.id, result_names.c_str()))) { for (i=0; i from request msg // - from request message // - prefs from user DB record // and parse them into sreq.global_prefs. // 2) update prefs in user record if needed // 2) send global prefs in reply msg if needed // int handle_global_prefs(SCHEDULER_REQUEST& sreq, SCHEDULER_REPLY& reply) { char buf[LARGE_BLOB_SIZE]; reply.send_global_prefs = false; bool have_working_prefs = (strlen(sreq.working_global_prefs_xml)>0); bool have_master_prefs = (strlen(sreq.global_prefs_xml)>0); bool have_db_prefs = (strlen(reply.user.global_prefs)>0); bool same_account = !strcmp( sreq.global_prefs_source_email_hash, reply.email_hash ); double master_mod_time=0, db_mod_time=0; if (have_master_prefs) { parse_double(sreq.global_prefs_xml, "", master_mod_time); if (master_mod_time > dtime()) master_mod_time = dtime(); } if (have_db_prefs) { parse_double(reply.user.global_prefs, "", db_mod_time); if (db_mod_time > dtime()) db_mod_time = dtime(); } //log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "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 sreq.global_prefs // if (have_working_prefs) { sreq.global_prefs.parse(sreq.working_global_prefs_xml, ""); log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "using working prefs\n"); } else { if (have_master_prefs) { if (have_db_prefs && db_mod_time > master_mod_time) { sreq.global_prefs.parse(reply.user.global_prefs, reply.host.venue); log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "using db prefs - more recent\n"); } else { sreq.global_prefs.parse(sreq.global_prefs_xml, reply.host.venue); log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "using master prefs\n"); } } else { if (have_db_prefs) { sreq.global_prefs.parse(reply.user.global_prefs, reply.host.venue); log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "using db prefs\n"); } else { sreq.global_prefs.defaults(); log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "using default prefs\n"); } } } // decide whether to update DB // if (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) { log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "updating db prefs\n"); strcpy(reply.user.global_prefs, sreq.global_prefs_xml); DB_USER user; user.id = reply.user.id; escape_string(sreq.global_prefs_xml, sizeof(sreq.global_prefs_xml)); sprintf(buf, "global_prefs='%s'", sreq.global_prefs_xml); unescape_string(sreq.global_prefs_xml, sizeof(sreq.global_prefs_xml)); int retval = user.update_field(buf); if (retval) { log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL, "user.update_field() failed: %d\n", retval ); } } } // decide whether to send DB prefs in reply msg // //log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "have db %d; dbmod %f; global mod %f\n", have_db_prefs, db_mod_time, sreq.global_prefs.mod_time); if (have_db_prefs && db_mod_time > master_mod_time) { log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "sending db prefs in reply\n"); reply.send_global_prefs = true; } return 0; } // if the client has an old code sign public key, // send it the new one, with a signature based on the old one. // If they don't have a code sign key, send them one // void send_code_sign_key( SCHEDULER_REQUEST& sreq, SCHEDULER_REPLY& reply, char* code_sign_key ) { char* oldkey, *signature; int i, retval; char path[256]; if (strlen(sreq.code_sign_key)) { if (strcmp(sreq.code_sign_key, code_sign_key)) { log_messages.printf(SCHED_MSG_LOG::MSG_NORMAL, "received old code sign key\n"); // look for a signature file // for (i=0; ; i++) { sprintf(path, "%s/old_key_%d", config.key_dir, i); retval = read_file_malloc(path, oldkey); if (retval) { USER_MESSAGE um( "You may have an outdated code verification key. " "This may prevent you from accepting new executables. " "If the problem persists, detach/attach the project. ", "high" ); reply.insert_message(um); return; } if (!strcmp(oldkey, sreq.code_sign_key)) { sprintf(path, "%s/signature_%d", config.key_dir, i); retval = read_file_malloc(path, signature); if (retval) { USER_MESSAGE um( "You may have an outdated code verification key. " "This may prevent you from accepting new executables. " "If the problem persists, detach/attach the project. ", "high" ); reply.insert_message(um); } else { safe_strcpy(reply.code_sign_key, code_sign_key); safe_strcpy(reply.code_sign_key_signature, signature); free(signature); } } free(oldkey); return; } } } else { safe_strcpy(reply.code_sign_key, code_sign_key); } } // This routine examines the value // from config.xml. If set, and the core client version is less than // this version, send a warning to users to upgrade before deadline // given in in Unix time(2) format // expires. // void warn_user_if_core_client_upgrade_scheduled( SCHEDULER_REQUEST& sreq, SCHEDULER_REPLY& reply ) { int core_ver; core_ver = sreq.core_client_major_version*100; core_ver += sreq.core_client_minor_version; if (core_ver < config.min_core_client_version_announced) { // time remaining in hours, before upgrade required int remaining = config.min_core_client_upgrade_deadline-time(0); remaining /= 3600; if (0 < remaining) { char msg[512]; int days = remaining / 24; int hours = remaining % 24; sprintf(msg, "Starting in %d days and %d hours, project will require a minimum " "BOINC core client version of %d.%d.0. You are currently using " "version %d.%d.%d; please upgrade before this time.", days, hours, config.min_core_client_version_announced / 100, config.min_core_client_version_announced % 100, sreq.core_client_major_version, sreq.core_client_minor_version, sreq.core_client_release ); // make this low priority until three days are left. Then // bump to high. // if (days<3) { USER_MESSAGE um(msg, "high"); reply.insert_message(um); } else { USER_MESSAGE um(msg, "low"); reply.insert_message(um); } log_messages.printf( SCHED_MSG_LOG::MSG_DEBUG, "Sending warning: upgrade client %d.%d.%d within %d days %d hours\n", sreq.core_client_major_version, sreq.core_client_minor_version, sreq.core_client_release, days, hours ); } } return; } bool unacceptable_os( SCHEDULER_REQUEST& sreq, SCHEDULER_REPLY& reply ) { if (config.no_darwin_6) { log_messages.printf( SCHED_MSG_LOG::MSG_DEBUG, "OS version %s %s\n", sreq.host.os_name, sreq.host.os_version ); if (!strcmp(sreq.host.os_name, "Darwin") && (!strncmp(sreq.host.os_version, "5.", 2) || !strncmp(sreq.host.os_version, "6.", 2) ) ) { log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "Unacceptable OS %s %s\n", sreq.host.os_name, sreq.host.os_version ); USER_MESSAGE um( "Project only supports MacOS Darwin versions 7.X and above", "low" ); reply.insert_message(um); reply.set_delay(DELAY_UNACCEPTABLE_OS); return true; } } return false; } bool unacceptable_cpu( SCHEDULER_REQUEST& sreq, SCHEDULER_REPLY& reply ) { if (config.no_amd_k6) { if (strstr(sreq.host.p_vendor, "AMD") && strstr(sreq.host.p_model, "Family 5 Model 8 Stepping 0")) { log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "Unacceptable CPU %s %s\n", sreq.host.p_vendor, sreq.host.p_model ); USER_MESSAGE um("Project doesn't support AMD K6 CPUs", "low"); reply.insert_message(um); reply.set_delay(DELAY_UNACCEPTABLE_OS); return true; } } return false; } bool wrong_core_client_version( SCHEDULER_REQUEST& sreq, SCHEDULER_REPLY& reply ) { char msg[256]; bool wrong_version = false; if (config.min_core_client_version) { int major = config.min_core_client_version/100; int minor = config.min_core_client_version % 100; if (sreq.core_client_major_version < major || ((sreq.core_client_major_version == major) && (sreq.core_client_minor_version < minor))) { wrong_version = true; sprintf(msg, "Need version %d.%d or higher of the BOINC core client. You have %d.%d.", major, minor, sreq.core_client_major_version, sreq.core_client_minor_version ); log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "[HOST#%d] [auth %s] Wrong minor version from user: wanted %d, got %d\n", sreq.hostid, sreq.authenticator, minor, sreq.core_client_minor_version ); } } if (wrong_version) { USER_MESSAGE um(msg, "low"); reply.insert_message(um); reply.set_delay(DELAY_BAD_CLIENT_VERSION); return true; } return false; } inline static const char* get_remote_addr() { const char * r = getenv("REMOTE_ADDR"); return r ? r : "?.?.?.?"; } void handle_msgs_from_host(SCHEDULER_REQUEST& sreq, SCHEDULER_REPLY& reply) { unsigned int i; DB_MSG_FROM_HOST mfh; int retval; for (i=0; i 0) && have_no_work && (sreq.results.size() == 0) && (sreq.hostid != 0) ) { USER_MESSAGE um("No work available", "low"); reply.insert_message(um); reply.set_delay(DELAY_NO_WORK_SKIP); if (!config.msg_to_host) { log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "No work - skipping DB access\n" ); return; } } // FROM HERE ON DON'T RETURN; goto leave instead // because we've tagged an entry in the work array with our process ID // now open the database // retval = open_database(); if (retval) { send_error_message(reply, "Server can't open database", 3600); goto leave; } retval = authenticate_user(sreq, reply); if (retval) goto leave; if (reply.user.id == 0) { log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL, "No user ID!\n"); } initial_host = reply.host; reply.host.rpc_seqno = sreq.rpc_seqno; log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "Processing request from [USER#%d] [HOST#%d] [IP %s] [RPC#%d] core client version %d.%d.%d\n", reply.user.id, reply.host.id, get_remote_addr(), sreq.rpc_seqno, sreq.core_client_major_version, sreq.core_client_minor_version, sreq.core_client_release ); ++log_messages; // is host blacklisted? // if (reply.host.max_results_day == -1) { send_error_message( reply, "Not accepting requests from this host", 86400 ); goto leave; } if (strlen(config.sched_lockfile_dir)) { int pid_with_lock = lock_sched(reply); if (pid_with_lock > 0) { log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL, "Another scheduler instance [PID=%d] is running for this host\n", pid_with_lock ); } else if (pid_with_lock) { log_messages.printf(SCHED_MSG_LOG::MSG_CRITICAL, "Error acquiring lock for [HOST#%d]\n", reply.host.id ); } if (pid_with_lock) { send_error_message( reply, "Another scheduler instance is running for this host", 60 ); goto leave; } } last_rpc_time = reply.host.rpc_time; rpc_time_tm = localtime((const time_t*)&reply.host.rpc_time); last_rpc_dayofyear = rpc_time_tm->tm_yday; reply.host.rpc_time = time(0); rpc_time_tm = localtime((const time_t*)&reply.host.rpc_time); current_rpc_dayofyear = rpc_time_tm->tm_yday; if (config.daily_result_quota) { if (reply.host.max_results_day == 0 || reply.host.max_results_day > config.daily_result_quota) { reply.host.max_results_day = config.daily_result_quota; log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "[HOST#%d] Initializing max_results_day to %d\n", reply.host.id, config.daily_result_quota ); } } if (last_rpc_dayofyear != current_rpc_dayofyear) { log_messages.printf(SCHED_MSG_LOG::MSG_DEBUG, "[HOST#%d] Resetting nresults_today\n", reply.host.id ); reply.host.nresults_today = 0; } retval = modify_host_struct(sreq, reply.host); // write time stats to disk if present // if (sreq.have_time_stats_log) { write_time_stats_log(reply); } // look up the client's platform(s) in the DB // platform = ss.lookup_platform(sreq.platform.name); if (platform) sreq.platforms.list.push_back(platform); for (i=0; i 0) { if (config.min_sendwork_interval) { double diff = dtime() - last_rpc_time; if (diff < config.min_sendwork_interval) { ok_to_send_work = false; log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "Not sending work - last request too recent: %f\n", diff ); sprintf(buf, "Not sending work - last request too recent: %d sec", (int)diff ); USER_MESSAGE um(buf, "low"); reply.insert_message(um); // the 1.01 is in case client's clock // is slightly faster than ours // reply.set_delay(1.01*config.min_sendwork_interval); } } if (ok_to_send_work) { send_work(sreq, reply, ss); } } send_code_sign_key(sreq, reply, code_sign_key); handle_msgs_from_host(sreq, reply); if (config.msg_to_host) { handle_msgs_to_host(reply); } update_host_record(initial_host, reply.host, reply.user); leave: if (!have_no_work) { ss.restore_work(g_pid); } } void handle_request( FILE* fin, FILE* fout, SCHED_SHMEM& ss, char* code_sign_key ) { SCHEDULER_REQUEST sreq; SCHEDULER_REPLY sreply; memset(&sreq, 0, sizeof(sreq)); #ifdef _USING_FCGI_ log_messages.set_indent_level(1); #endif if (sreq.parse(fin) == 0){ log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "Handling request: IP %s, auth %s, host %d, platform %s, version %d.%d.%d, work req %d sec\n", get_remote_addr(), sreq.authenticator, sreq.hostid, sreq.platform.name, sreq.core_client_major_version, sreq.core_client_minor_version, sreq.core_client_release, (int)sreq.work_req_seconds ); process_request(sreq, sreply, ss, code_sign_key); #ifdef _USING_FCGI_ log_messages.set_indent_level(2); #endif } else { // BOINC scheduler requests use method POST. // So method GET means that someone is trying a browser. // char *rm=getenv("REQUEST_METHOD"); if (rm && !strcmp(rm, "GET")) { sreply.probable_user_browser=true; } #ifdef _USING_FCGI_ log_messages.set_indent_level(2); #endif log_messages.printf( SCHED_MSG_LOG::MSG_NORMAL, "Incomplete request received %sfrom IP %s, auth %s, platform %s, version %d.%d.%d\n", sreply.probable_user_browser?"(probably a browser) ":"", get_remote_addr(), sreq.authenticator, sreq.platform.name, sreq.core_client_major_version, sreq.core_client_minor_version, sreq.core_client_release ); USER_MESSAGE um("Incomplete request received.", "low"); sreply.insert_message(um); sreply.nucleus_only = true; } if (config.locality_scheduling && !sreply.nucleus_only) { send_file_deletes(sreq, sreply); } // write all messages to log file // for (unsigned int i=0; i