/* * Copyright 1995-2014 Andrew Smith * Copyright 2014 Con Kolivas * * This program is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License as published by the Free * Software Foundation; either version 3 of the License, or (at your option) * any later version. See COPYING for more details. */ #include "ckdb.h" /* TODO: any tree/list accessed in new threads needs * to ensure all code using those trees/lists use locks * This code's lock implementation is equivalent to table level locking * Consider adding row level locking (a per kitem usage count) if needed * TODO: verify all tables with multithread access are locked */ /* Startup * ------- * During startup we load the DB and track where it is up to with * dbstatus, we then reload "ckpool's ckdb logfiles" (CCLs) based * on dbstatus * Once the DB is loaded, we can immediately start receiving ckpool * messages since ckpool already has logged all messages to the CLLs * and ckpool only verifies authorise responses * Thus we can queue all messages: * workinfo, shares, shareerror, ageworkinfo, poolstats, userstats * and block * with an ok.queued reply to ckpool, to be processed after the reload * completes and just process authorise messages immediately while the * reload runs * This can't cause a duplicate process of an authorise message since a * reload will ignore any messages before the last DB auths message, * however, if ckdb and ckpool get out of sync due to ckpool starting * during the reload (as mentioned below) it is possible for ckdb to * find an authorise message in the CCLs that was processed in the * message queue and thus is already in the DB. * This error would be very rare and also not an issue * To avoid this, we start the ckpool message queue after loading * the users, auths, idcontrol and workers DB tables, before loading the * much larger sharesummary, workinfo, userstats and poolstats DB tables * so that ckdb is effectively ready for messages almost immediately * The first ckpool message also allows us to know where ckpool is up to * in the CCLs and thus where to stop processing the CCLs to stay in * sync with ckpool * If ckpool isn't running, then the reload will complete at the end of * the last CCL file, however if the 1st message arrives from ckpool while * processing the CCLs, that will mark the point where to stop processing * but can also produce a fatal error at the end of processing, reporting * the ckpool message, if the message was not found in the CCL processing * after the message was received * This can be caused by two circumstances: * 1) the disk had not yet written it to the CCL when ckdb read EOF and * ckpool was started at about the same time as the reload completed. * This can be seen if the message displayed in the fatal error IS NOT * in ckdb's message logfile. * A ckdb restart will resolve this * 2) ckpool was started at the time of the end of the reload, but the * message was written to disk and found in the CCL before it was * processed in the message queue. * This can be seen if the message displayed in the fatal error IS in * ckdb's message logfile and means the messages after it in ckdb's * message logfile have already been processed. * Again, a ckdb restart will resolve this * In both the above (very rare) cases, if ckdb was to continue running, * it would break the synchronisation and could cause DB problems, so * ckdb aborting and needing a complete restart resolves it * The users table, required for the authorise messages, is always updated * immediately and is not affected by ckpool messages until we * During the reload, when checking the timeframe for summarisation, we * use the current last userstats createdate as 'now' to avoid touching a * timeframe where data could still be waiting to be loaded */ /* Reload data needed * ------------------ * After the DB load completes, load "ckpool's ckdb logfile" (CCL), and * all later CCLs, that contains the oldest date of all of the following: * RAM shares: oldest DB sharesummary firstshare where complete='n' * All shares before this have been summarised to the DB with * complete='a' (or 'y') and were deleted from RAM * If there are none with complete='n' but are others in the DB, * then the newest firstshare is used * RAM shareerrors: as above * DB+RAM sharesummary: created from shares, so as above * Some shares after this may have been summarised to other * sharesummary complete='n', but for any such sharesummary * we reset it back to the first share found and it will * correct itself during the CCL reload * TODO: Verify that all DB sharesummaries with complete='n' * have done this * DB+RAM workinfo: start from newest DB createdate workinfo * DB+RAM auths: start from newest DB createdate auths * DB+RAM poolstats: newest createdate poolstats * TODO: subtract how much we need in RAM of the 'between' * non db records - will depend on TODO: pool stats reporting * requirements * DB+RAM userstats: start of the time band of the latest DB record, * since all data before this has been summarised to the DB * The userstats summarisation always processes the oldest * RAM data to the DB * TODO: multiple pools is not yet handled by ckdb * TODO: handle a pool restart with a different instance name * since this would always make the userstats reload point * just after the instance name was changed - however * this can be handled for now by simply ignoring the * poolinstance * DB+RAM workers: created by auths so auths will resolve it * DB+RAM blocks: resolved by workinfo - any unsaved blocks (if any) * will be after the last DB workinfo * DB+RAM accountbalance (TODO): resolved by shares/workinfo/blocks * RAM workerstatus: all except last_idle are set at the end of the * CCL reload * Code currently doesn't use last_idle * * idcontrol: only userid reuse is critical and the user is added * immeditately to the DB before replying to the add message * * Tables that are/will be written straight to the DB, so are OK: * users, useraccounts, paymentaddresses, payments, * accountadjustment, optioncontrol, miningpayouts, * eventlog * * The code deals with the issue of 'now' when reloading by: * createdate is considered 'now' for all data during a reload and is * a mandatory field always checked for in ckpool data * Implementing this is simple by assuming 'now' is always createdate * i.e. during reload but also during normal running to avoid timing * issues with ckpool data * Other data supplies the current time to the functions that require * 'now' since that data is not part of a reload * * During reload, any data before the calculated reload stamp for * that data is discarded * Any data that matches the reload stamp is processed with an * ignore duplicates flag for all except as below. * Any data after the stamp, is processed normally for all except: * 1) userstats: any record that falls in a DB userstats that would * summarise that record is discarded, * otherwise the userstats is processed normally * 2) shares/shareerrors: any record that matches an incomplete DB * sharesummary that hasn't been reset, will reset the sharesummary * so that the sharesummary will be recalculated * The record is processed normally with or without the reset * If the sharesummary is complete, the record is discarded * 3) ageworkinfo records are also handled by the shares date * while processing, any records already aged are not updated * and a warning is displayed if there were any matching shares */ char *EMPTY = ""; static char *db_name; static char *db_user; static char *db_pass; // Currently hard coded at 4 characters static char *status_chars = "|/-\\"; static char *restorefrom; // disallow: '/' '.' '_' and FLDSEP const char *userpatt = "^[^/\\._"FLDSEPSTR"]*$"; const char *mailpatt = "^[A-Za-z0-9_-][A-Za-z0-9_\\.-]*@[A-Za-z0-9][A-Za-z0-9\\.-]*[A-Za-z0-9]$"; const char *idpatt = "^[_A-Za-z][_A-Za-z0-9]*$"; const char *intpatt = "^[0-9][0-9]*$"; const char *hashpatt = "^[A-Fa-f0-9]*$"; // TODO: bitcoind will check it properly const char *addrpatt = "^[A-Za-z0-9]*$"; // So the records below have the same 'name' as the klist const char Transfer[] = "Transfer"; // older version missing field defaults static TRANSFER auth_1 = { "poolinstance", "", auth_1.svalue }; K_ITEM auth_poolinstance = { Transfer, NULL, NULL, (void *)(&auth_1) }; static TRANSFER auth_2 = { "preauth", FALSE_STR, auth_2.svalue }; K_ITEM auth_preauth = { Transfer, NULL, NULL, (void *)(&auth_2) }; static TRANSFER poolstats_1 = { "elapsed", "0", poolstats_1.svalue }; K_ITEM poolstats_elapsed = { Transfer, NULL, NULL, (void *)(&poolstats_1) }; static TRANSFER userstats_1 = { "elapsed", "0", userstats_1.svalue }; K_ITEM userstats_elapsed = { Transfer, NULL, NULL, (void *)(&userstats_1) }; static TRANSFER userstats_2 = { "workername", "all", userstats_2.svalue }; K_ITEM userstats_workername = { Transfer, NULL, NULL, (void *)(&userstats_2) }; static TRANSFER userstats_3 = { "idle", FALSE_STR, userstats_3.svalue }; K_ITEM userstats_idle = { Transfer, NULL, NULL, (void *)(&userstats_3) }; static TRANSFER userstats_4 = { "eos", TRUE_STR, userstats_4.svalue }; K_ITEM userstats_eos = { Transfer, NULL, NULL, (void *)(&userstats_4) }; static TRANSFER shares_1 = { "secondaryuserid", TRUE_STR, shares_1.svalue }; K_ITEM shares_secondaryuserid = { Transfer, NULL, NULL, (void *)(&shares_1) }; static TRANSFER shareerrors_1 = { "secondaryuserid", TRUE_STR, shareerrors_1.svalue }; K_ITEM shareerrors_secondaryuserid = { Transfer, NULL, NULL, (void *)(&shareerrors_1) }; // Time limit that this problem occurred // 24-Aug-2014 05:20+00 (1st one shortly after this) tv_t missing_secuser_min = { 1408857600L, 0L }; // 24-Aug-2014 06:00+00 tv_t missing_secuser_max = { 1408860000L, 0L }; LOADSTATUS dbstatus; POOLSTATUS pool = { 0, START_POOL_HEIGHT, 0, 0, 0, 0, 0, 0 }; // Ensure long long and int64_t are both 8 bytes (and thus also the same) #define ASSERT1(condition) __maybe_unused static char sizeof_longlong_must_be_8[(condition)?1:-1] #define ASSERT2(condition) __maybe_unused static char sizeof_int64_t_must_be_8[(condition)?1:-1] ASSERT1(sizeof(long long) == 8); ASSERT2(sizeof(int64_t) == 8); const tv_t default_expiry = { DEFAULT_EXPIRY, 0L }; const tv_t date_eot = { DATE_S_EOT, DATE_uS_EOT }; const tv_t date_begin = { DATE_BEGIN, 0L }; // argv -y - don't run in ckdb mode, just confirm sharesummaries bool confirm_sharesummary; /* Optional workinfoid range -Y to supply when confirming sharesummaries * N.B. if you specify -Y it will enable -y, so -y isn't also required * * Default (NULL) is to confirm all aged sharesummaries * Default should normally be used every time * The below options are mainly for debugging or * a quicker confirm if required due to not running confirms regularly * TODO: ... once the code includes flagging confirmed sharesummaries * Valid options are: * bNNN - confirm all workinfoid's from the previous db block before * NNN (or 0) up to the workinfoid of the 1st db block height * equal or after NNN * wNNN - confirm all workinfoid's from NNN up to the last aged * sharesummary * rNNN-MMM - confirm all workinfoid's from NNN to MMM inclusive * cNNN-MMM - check the CCL record timestamps then exit * i - just show the DB load information then exit */ static char *confirm_range; static int confirm_block; static int64_t confirm_range_start; static int64_t confirm_range_finish; static bool confirm_check_createdate; static int64_t ccl_mismatch_abs; static int64_t ccl_mismatch; static double ccl_mismatch_min; static double ccl_mismatch_max; static int64_t ccl_unordered_abs; static int64_t ccl_unordered; static double ccl_unordered_most; // The workinfoid range we are processing int64_t confirm_first_workinfoid; int64_t confirm_last_workinfoid; /* Stop the reload 11min after the 'last' workinfoid+1 appears * ckpool uses 10min - but add 1min to be sure */ #define WORKINFO_AGE 660 static tv_t confirm_finish; static tv_t reload_timestamp; // DB users,workers,auth load is complete bool db_auths_complete = false; // DB load is complete bool db_load_complete = false; // Different input data handling bool reloading = false; // Data load is complete bool startup_complete = false; // Tell everyone to die bool everyone_die = false; static cklock_t fpm_lock; static char *first_pool_message; static sem_t socketer_sem; char *by_default = "code"; char *inet_default = "127.0.0.1"; // LOGQUEUE K_LIST *logqueue_free; K_STORE *logqueue_store; // WORKQUEUE K_LIST *workqueue_free; K_STORE *workqueue_store; pthread_mutex_t wq_waitlock; pthread_cond_t wq_waitcond; // TRANSFER K_LIST *transfer_free; // USERS K_TREE *users_root; K_TREE *userid_root; K_LIST *users_free; K_STORE *users_store; // USERATTS K_TREE *useratts_root; K_LIST *useratts_free; K_STORE *useratts_store; // WORKERS K_TREE *workers_root; K_LIST *workers_free; K_STORE *workers_store; // PAYMENTADDRESSES K_TREE *paymentaddresses_root; K_LIST *paymentaddresses_free; K_STORE *paymentaddresses_store; // PAYMENTS K_TREE *payments_root; K_LIST *payments_free; K_STORE *payments_store; /* unused yet // ACCOUNTBALANCE K_TREE *accountbalance_root; K_LIST *accountbalance_free; K_STORE *accountbalance_store; // ACCOUNTADJUSTMENT K_TREE *accountadjustment_root; K_LIST *accountadjustment_free; K_STORE *accountadjustment_store; */ // IDCONTROL // These are only used for db access - not stored in memory //K_TREE *idcontrol_root; K_LIST *idcontrol_free; K_STORE *idcontrol_store; // OPTIONCONTROL K_TREE *optioncontrol_root; K_LIST *optioncontrol_free; K_STORE *optioncontrol_store; // WORKINFO workinfo.id.json={...} K_TREE *workinfo_root; // created during data load then destroyed since not needed later K_TREE *workinfo_height_root; K_LIST *workinfo_free; K_STORE *workinfo_store; // one in the current block K_ITEM *workinfo_current; // first workinfo of current block tv_t last_bc; // current network diff double current_ndiff; // SHARES shares.id.json={...} K_TREE *shares_root; K_LIST *shares_free; K_STORE *shares_store; // SHAREERRORS shareerrors.id.json={...} K_TREE *shareerrors_root; K_LIST *shareerrors_free; K_STORE *shareerrors_store; // SHARESUMMARY K_TREE *sharesummary_root; K_TREE *sharesummary_workinfoid_root; K_LIST *sharesummary_free; K_STORE *sharesummary_store; // BLOCKS block.id.json={...} const char *blocks_new = "New"; const char *blocks_confirm = "1-Confirm"; const char *blocks_42 = "42-Confirm"; const char *blocks_orphan = "Orphan"; const char *blocks_unknown = "?Unknown?"; K_TREE *blocks_root; K_LIST *blocks_free; K_STORE *blocks_store; // MININGPAYOUTS K_TREE *miningpayouts_root; K_LIST *miningpayouts_free; K_STORE *miningpayouts_store; /* // EVENTLOG K_TREE *eventlog_root; K_LIST *eventlog_free; K_STORE *eventlog_store; */ // AUTHS authorise.id.json={...} K_TREE *auths_root; K_LIST *auths_free; K_STORE *auths_store; // POOLSTATS poolstats.id.json={...} // TODO: redo like userstats, but every 10min K_TREE *poolstats_root; K_LIST *poolstats_free; K_STORE *poolstats_store; // USERSTATS userstats.id.json={...} K_TREE *userstats_root; K_TREE *userstats_statsdate_root; // ordered by statsdate first K_TREE *userstats_workerstatus_root; // during data load K_LIST *userstats_free; K_STORE *userstats_store; // Awaiting EOS K_STORE *userstats_eos_store; // Temporary while summarising K_STORE *userstats_summ; // WORKERSTATUS from various incoming data K_TREE *workerstatus_root; K_LIST *workerstatus_free; K_STORE *workerstatus_store; static char logname[512]; static char *dbcode; // low spec version of rotating_log() - no locking static bool rotating_log_nolock(char *msg) { char *filename; FILE *fp; bool ok = false; filename = rotating_filename(logname, time(NULL)); fp = fopen(filename, "a+e"); if (unlikely(!fp)) { LOGERR("Failed to fopen %s in rotating_log!", filename); goto stageleft; } fprintf(fp, "%s\n", msg); fclose(fp); ok = true; stageleft: free(filename); return ok; } static void log_queue_message(char *msg) { K_ITEM *lq_item; LOGQUEUE *lq; K_WLOCK(logqueue_free); lq_item = k_unlink_head(logqueue_free); DATA_LOGQUEUE(lq, lq_item); lq->msg = strdup(msg); if (!(lq->msg)) quithere(1, "malloc (%d) OOM", (int)strlen(msg)); k_add_tail(logqueue_store, lq_item); K_WUNLOCK(logqueue_free); } void logmsg(int loglevel, const char *fmt, ...) { int logfd = 0; char *buf = NULL; struct tm tm; time_t now_t; va_list ap; char stamp[128]; char *extra = EMPTY; char tzinfo[16]; char tzch; long minoff, hroff; if (loglevel > global_ckp->loglevel) return; now_t = time(NULL); localtime_r(&now_t, &tm); minoff = timezone / 60; if (minoff < 0) { tzch = '+'; minoff *= -1; } else tzch = '-'; hroff = minoff / 60; if (minoff % 60) { snprintf(tzinfo, sizeof(tzinfo), "%c%02ld:%02ld", tzch, hroff, minoff % 60); } else { snprintf(tzinfo, sizeof(tzinfo), "%c%02ld", tzch, hroff); } snprintf(stamp, sizeof(stamp), "[%d-%02d-%02d %02d:%02d:%02d%s]", tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, tm.tm_hour, tm.tm_min, tm.tm_sec, tzinfo); if (!fmt) { fprintf(stderr, "%s %s() called without fmt\n", stamp, __func__); return; } if (!global_ckp) extra = " !!NULL global_ckp!!"; else logfd = global_ckp->logfd; va_start(ap, fmt); VASPRINTF(&buf, fmt, ap); va_end(ap); if (logfd) { FILE *LOGFP = global_ckp->logfp; flock(logfd, LOCK_EX); fprintf(LOGFP, "%s %s", stamp, buf); if (loglevel <= LOG_ERR && errno != 0) fprintf(LOGFP, " with errno %d: %s", errno, strerror(errno)); errno = 0; fprintf(LOGFP, "\n"); flock(logfd, LOCK_UN); } if (loglevel <= LOG_WARNING) { if (loglevel <= LOG_ERR && errno != 0) { fprintf(stderr, "%s %s with errno %d: %s%s\n", stamp, buf, errno, strerror(errno), extra); errno = 0; } else fprintf(stderr, "%s %s%s\n", stamp, buf, extra); fflush(stderr); } free(buf); } static void setnow(tv_t *now) { ts_t spec; spec.tv_sec = 0; spec.tv_nsec = 0; clock_gettime(CLOCK_REALTIME, &spec); now->tv_sec = spec.tv_sec; now->tv_usec = spec.tv_nsec / 1000; } // Limits are all +/-1s since on the live machine all were well within that static void check_createdate_ccl(char *cmd, tv_t *cd) { static tv_t last_cd; static char last_cmd[CMD_SIZ+1]; char cd_buf1[DATE_BUFSIZ], cd_buf2[DATE_BUFSIZ]; char *filename; double td; if (cd->tv_sec < reload_timestamp.tv_sec || cd->tv_sec >= (reload_timestamp.tv_sec + ROLL_S)) { ccl_mismatch_abs++; td = tvdiff(cd, &reload_timestamp); if (td < -1 || td > ROLL_S + 1) { ccl_mismatch++; filename = rotating_filename("", reload_timestamp.tv_sec); tv_to_buf(cd, cd_buf1, sizeof(cd_buf1)); LOGERR("%s(): CCL contains mismatch data: cmd:%s CCL:%.10s cd:%s", __func__, cmd, filename, cd_buf1); free(filename); } if (ccl_mismatch_min > td) ccl_mismatch_min = td; if (ccl_mismatch_max < td) ccl_mismatch_max = td; } td = tvdiff(cd, &last_cd); if (td < 0.0) { ccl_unordered_abs++; if (ccl_unordered_most > td) ccl_unordered_most = td; } if (td < -1.0) { ccl_unordered++; tv_to_buf(&last_cd, cd_buf1, sizeof(cd_buf1)); tv_to_buf(cd, cd_buf2, sizeof(cd_buf2)); LOGERR("%s(): CCL time unordered: %s<->%s %ld,%ld<->%ld,%ld %s<->%s", __func__, last_cmd, cmd, last_cd.tv_sec,last_cd.tv_usec, cd->tv_sec, cd->tv_usec, cd_buf1, cd_buf2); } copy_tv(&last_cd, cd); STRNCPY(last_cmd, cmd); } static uint64_t ticks; static time_t last_tick; void tick() { time_t now; char ch; now = time(NULL); if (now != last_tick) { last_tick = now; ch = status_chars[ticks++ & 0x3]; putchar(ch); putchar('\r'); fflush(stdout); } } PGconn *dbconnect() { char conninfo[256]; PGconn *conn; int i, retry = 10; snprintf(conninfo, sizeof(conninfo), "host=127.0.0.1 dbname=%s user=%s%s%s", db_name, db_user, db_pass ? " password=" : "", db_pass ? db_pass : ""); conn = PQconnectdb(conninfo); if (PQstatus(conn) != CONNECTION_OK) { LOGERR("%s(): Failed 1st connect to db '%s'", __func__, pqerrmsg(conn)); LOGERR("%s(): Retrying for %d seconds ...", __func__, retry); for (i = 0; i < retry; i++) { sleep(1); conn = PQconnectdb(conninfo); if (PQstatus(conn) == CONNECTION_OK) { LOGWARNING("%s(): Connected on attempt %d", __func__, i+2); return conn; } } quithere(1, "ERR: Failed to connect %d times to db '%s'", retry+1, pqerrmsg(conn)); } return conn; } /* Load tables required to support auths,adduser,chkpass and newid * N.B. idcontrol is DB internal so is always ready * OptionControl is loaded first in case it is needed by other loads * (though not yet) */ static bool getdata1() { PGconn *conn = dbconnect(); bool ok = true; if (!(ok = check_db_version(conn))) goto matane; if (!(ok = optioncontrol_fill(conn))) goto matane; if (!(ok = users_fill(conn))) goto matane; if (!(ok = workers_fill(conn))) goto matane; if (!confirm_sharesummary) ok = auths_fill(conn); matane: PQfinish(conn); return ok; } static bool getdata2() { PGconn *conn = dbconnect(); bool ok = true; if (!(ok = blocks_fill(conn)) || everyone_die) goto sukamudai; if (!confirm_sharesummary) { if (!(ok = paymentaddresses_fill(conn)) || everyone_die) goto sukamudai; if (!(ok = payments_fill(conn)) || everyone_die) goto sukamudai; } if (!(ok = workinfo_fill(conn)) || everyone_die) goto sukamudai; if (!(ok = sharesummary_fill(conn)) || everyone_die) goto sukamudai; if (!confirm_sharesummary) { if (!(ok = useratts_fill(conn)) || everyone_die) goto sukamudai; if (!(ok = poolstats_fill(conn)) || everyone_die) goto sukamudai; ok = userstats_fill(conn); } sukamudai: PQfinish(conn); return ok; } static bool reload_from(tv_t *start); static bool reload() { char buf[DATE_BUFSIZ+1]; char *filename; tv_t start; char *reason; FILE *fp; tv_to_buf(&(dbstatus.oldest_sharesummary_firstshare_n), buf, sizeof(buf)); LOGWARNING("%s(): %s oldest DB incomplete sharesummary", __func__, buf); tv_to_buf(&(dbstatus.newest_sharesummary_firstshare_ay), buf, sizeof(buf)); LOGWARNING("%s(): %s newest DB complete sharesummary", __func__, buf); tv_to_buf(&(dbstatus.newest_createdate_workinfo), buf, sizeof(buf)); LOGWARNING("%s(): %s newest DB workinfo", __func__, buf); tv_to_buf(&(dbstatus.newest_createdate_auths), buf, sizeof(buf)); LOGWARNING("%s(): %s newest DB auths", __func__, buf); tv_to_buf(&(dbstatus.newest_createdate_poolstats), buf, sizeof(buf)); LOGWARNING("%s(): %s newest DB poolstats", __func__, buf); tv_to_buf(&(dbstatus.newest_starttimeband_userstats), buf, sizeof(buf)); LOGWARNING("%s(): %s newest DB userstats start timeband", __func__, buf); tv_to_buf(&(dbstatus.newest_createdate_blocks), buf, sizeof(buf)); LOGWARNING("%s(): %s newest DB blocks (ignored)", __func__, buf); if (dbstatus.oldest_sharesummary_firstshare_n.tv_sec) copy_tv(&(dbstatus.sharesummary_firstshare), &(dbstatus.oldest_sharesummary_firstshare_n)); else copy_tv(&(dbstatus.sharesummary_firstshare), &(dbstatus.newest_sharesummary_firstshare_ay)); copy_tv(&start, &(dbstatus.sharesummary_firstshare)); reason = "sharesummary"; if (!tv_newer(&start, &(dbstatus.newest_createdate_workinfo))) { copy_tv(&start, &(dbstatus.newest_createdate_workinfo)); reason = "workinfo"; } if (!tv_newer(&start, &(dbstatus.newest_createdate_auths))) { copy_tv(&start, &(dbstatus.newest_createdate_auths)); reason = "auths"; } if (!tv_newer(&start, &(dbstatus.newest_createdate_poolstats))) { copy_tv(&start, &(dbstatus.newest_createdate_poolstats)); reason = "poolstats"; } if (!tv_newer(&start, &(dbstatus.newest_starttimeband_userstats))) { copy_tv(&start, &(dbstatus.newest_starttimeband_userstats)); reason = "userstats"; } tv_to_buf(&start, buf, sizeof(buf)); LOGWARNING("%s() restart timestamp %s for %s", __func__, buf, reason); if (start.tv_sec < DATE_BEGIN) { start.tv_sec = DATE_BEGIN; start.tv_usec = 0L; filename = rotating_filename(restorefrom, start.tv_sec); fp = fopen(filename, "re"); if (fp) fclose(fp); else { mode_t mode = S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH; int fd = open(filename, O_CREAT|O_RDONLY, mode); if (fd == -1) { int ern = errno; quithere(1, "Couldn't create '%s' (%d) %s", filename, ern, strerror(ern)); } close(fd); } free(filename); } return reload_from(&start); } /* Open the file in path, check if there is a pid in there that still exists * and if not, write the pid into that file. */ static bool write_pid(ckpool_t *ckp, const char *path, pid_t pid) { struct stat statbuf; FILE *fp; int ret; if (!stat(path, &statbuf)) { int oldpid; LOGWARNING("File %s exists", path); fp = fopen(path, "re"); if (!fp) { LOGEMERG("Failed to open file %s", path); return false; } ret = fscanf(fp, "%d", &oldpid); fclose(fp); if (ret == 1 && !(kill(oldpid, 0))) { if (!ckp->killold) { LOGEMERG("Process %s pid %d still exists, start ckpool with -k if you wish to kill it", path, oldpid); return false; } if (kill(oldpid, 9)) { LOGEMERG("Unable to kill old process %s pid %d", path, oldpid); return false; } LOGWARNING("Killing off old process %s pid %d", path, oldpid); } } fp = fopen(path, "we"); if (!fp) { LOGERR("Failed to open file %s", path); return false; } fprintf(fp, "%d", pid); fclose(fp); return true; } static void create_process_unixsock(proc_instance_t *pi) { unixsock_t *us = &pi->us; us->path = strdup(pi->ckp->socket_dir); realloc_strcat(&us->path, pi->sockname); LOGDEBUG("Opening %s", us->path); us->sockd = open_unix_server(us->path); if (unlikely(us->sockd < 0)) quit(1, "Failed to open %s socket", pi->sockname); } static void write_namepid(proc_instance_t *pi) { char s[256]; pi->pid = getpid(); sprintf(s, "%s%s.pid", pi->ckp->socket_dir, pi->processname); if (!write_pid(pi->ckp, s, pi->pid)) quit(1, "Failed to write %s pid %d", pi->processname, pi->pid); } static void rm_namepid(proc_instance_t *pi) { char s[256]; sprintf(s, "%s%s.pid", pi->ckp->socket_dir, pi->processname); unlink(s); } static void clean_up(ckpool_t *ckp) { rm_namepid(&ckp->main); dealloc(ckp->socket_dir); fclose(ckp->logfp); } static void alloc_storage() { workqueue_free = k_new_list("WorkQueue", sizeof(WORKQUEUE), ALLOC_WORKQUEUE, LIMIT_WORKQUEUE, true); workqueue_store = k_new_store(workqueue_free); transfer_free = k_new_list(Transfer, sizeof(TRANSFER), ALLOC_TRANSFER, LIMIT_TRANSFER, true); transfer_free->dsp_func = dsp_transfer; users_free = k_new_list("Users", sizeof(USERS), ALLOC_USERS, LIMIT_USERS, true); users_store = k_new_store(users_free); users_root = new_ktree(); userid_root = new_ktree(); useratts_free = k_new_list("Useratts", sizeof(USERATTS), ALLOC_USERATTS, LIMIT_USERATTS, true); useratts_store = k_new_store(useratts_free); useratts_root = new_ktree(); optioncontrol_free = k_new_list("OptionControl", sizeof(OPTIONCONTROL), ALLOC_OPTIONCONTROL, LIMIT_OPTIONCONTROL, true); optioncontrol_store = k_new_store(optioncontrol_free); optioncontrol_root = new_ktree(); workers_free = k_new_list("Workers", sizeof(WORKERS), ALLOC_WORKERS, LIMIT_WORKERS, true); workers_store = k_new_store(workers_free); workers_root = new_ktree(); paymentaddresses_free = k_new_list("PaymentAddresses", sizeof(PAYMENTADDRESSES), ALLOC_PAYMENTADDRESSES, LIMIT_PAYMENTADDRESSES, true); paymentaddresses_store = k_new_store(paymentaddresses_free); paymentaddresses_root = new_ktree(); payments_free = k_new_list("Payments", sizeof(PAYMENTS), ALLOC_PAYMENTS, LIMIT_PAYMENTS, true); payments_store = k_new_store(payments_free); payments_root = new_ktree(); idcontrol_free = k_new_list("IDControl", sizeof(IDCONTROL), ALLOC_IDCONTROL, LIMIT_IDCONTROL, true); idcontrol_store = k_new_store(idcontrol_free); workinfo_free = k_new_list("WorkInfo", sizeof(WORKINFO), ALLOC_WORKINFO, LIMIT_WORKINFO, true); workinfo_store = k_new_store(workinfo_free); workinfo_root = new_ktree(); if (!confirm_sharesummary) workinfo_height_root = new_ktree(); shares_free = k_new_list("Shares", sizeof(SHARES), ALLOC_SHARES, LIMIT_SHARES, true); shares_store = k_new_store(shares_free); shares_root = new_ktree(); shareerrors_free = k_new_list("ShareErrors", sizeof(SHAREERRORS), ALLOC_SHAREERRORS, LIMIT_SHAREERRORS, true); shareerrors_store = k_new_store(shareerrors_free); shareerrors_root = new_ktree(); sharesummary_free = k_new_list("ShareSummary", sizeof(SHARESUMMARY), ALLOC_SHARESUMMARY, LIMIT_SHARESUMMARY, true); sharesummary_store = k_new_store(sharesummary_free); sharesummary_root = new_ktree(); sharesummary_workinfoid_root = new_ktree(); sharesummary_free->dsp_func = dsp_sharesummary; blocks_free = k_new_list("Blocks", sizeof(BLOCKS), ALLOC_BLOCKS, LIMIT_BLOCKS, true); blocks_store = k_new_store(blocks_free); blocks_root = new_ktree(); blocks_free->dsp_func = dsp_blocks; miningpayouts_free = k_new_list("MiningPayouts", sizeof(MININGPAYOUTS), ALLOC_MININGPAYOUTS, LIMIT_MININGPAYOUTS, true); miningpayouts_store = k_new_store(miningpayouts_free); miningpayouts_root = new_ktree(); auths_free = k_new_list("Auths", sizeof(AUTHS), ALLOC_AUTHS, LIMIT_AUTHS, true); auths_store = k_new_store(auths_free); auths_root = new_ktree(); poolstats_free = k_new_list("PoolStats", sizeof(POOLSTATS), ALLOC_POOLSTATS, LIMIT_POOLSTATS, true); poolstats_store = k_new_store(poolstats_free); poolstats_root = new_ktree(); userstats_free = k_new_list("UserStats", sizeof(USERSTATS), ALLOC_USERSTATS, LIMIT_USERSTATS, true); userstats_store = k_new_store(userstats_free); userstats_eos_store = k_new_store(userstats_free); userstats_summ = k_new_store(userstats_free); userstats_root = new_ktree(); userstats_statsdate_root = new_ktree(); userstats_workerstatus_root = new_ktree(); userstats_free->dsp_func = dsp_userstats; workerstatus_free = k_new_list("WorkerStatus", sizeof(WORKERSTATUS), ALLOC_WORKERSTATUS, LIMIT_WORKERSTATUS, true); workerstatus_store = k_new_store(workerstatus_free); workerstatus_root = new_ktree(); } static bool setup_data() { K_TREE_CTX ctx[1]; K_ITEM look, *found; WORKINFO wi, *wic, *wif; cklock_init(&fpm_lock); cksem_init(&socketer_sem); mutex_init(&wq_waitlock); cond_init(&wq_waitcond); alloc_storage(); if (!getdata1() || everyone_die) return false; db_auths_complete = true; cksem_post(&socketer_sem); if (!getdata2() || everyone_die) return false; db_load_complete = true; if (!reload() || everyone_die) return false; set_block_share_counters(); if (everyone_die) return false; workerstatus_ready(); userstats_workerstatus_root = free_ktree(userstats_workerstatus_root, NULL); workinfo_current = last_in_ktree(workinfo_height_root, ctx); if (workinfo_current) { DATA_WORKINFO(wic, workinfo_current); STRNCPY(wi.coinbase1, wic->coinbase1); wi.createdate.tv_sec = 0L; wi.createdate.tv_usec = 0L; INIT_WORKINFO(&look); look.data = (void *)(&wi); // Find the first workinfo for this height found = find_after_in_ktree(workinfo_height_root, &look, cmp_workinfo_height, ctx); if (found) { DATA_WORKINFO(wif, found); copy_tv(&last_bc, &(wif->createdate)); } // No longer needed workinfo_height_root = free_ktree(workinfo_height_root, NULL); } return true; } static enum cmd_values breakdown(K_TREE **trf_root, K_STORE **trf_store, char *buf, int *which_cmds, char *cmd, char *id, tv_t *cd) { char reply[1024] = ""; TRANSFER *transfer; K_TREE_CTX ctx[1]; K_ITEM *item; char *cmdptr, *idptr, *next, *eq; char *data = NULL; bool noid = false; *trf_root = NULL; *trf_store = NULL; *which_cmds = CMD_UNSET; *cmd = *id = '\0'; cd->tv_sec = 0; cd->tv_usec = 0; cmdptr = strdup(buf); idptr = strchr(cmdptr, '.'); if (!idptr || !*idptr) noid = true; else { *(idptr++) = '\0'; data = strchr(idptr, '.'); if (data) *(data++) = '\0'; STRNCPYSIZ(id, idptr, ID_SIZ); } STRNCPYSIZ(cmd, cmdptr, CMD_SIZ); for (*which_cmds = 0; ckdb_cmds[*which_cmds].cmd_val != CMD_END; (*which_cmds)++) { if (strcasecmp(cmd, ckdb_cmds[*which_cmds].cmd_str) == 0) break; } if (ckdb_cmds[*which_cmds].cmd_val == CMD_END) { LOGERR("Listener received unknown command: '%s'", buf); free(cmdptr); return CMD_REPLY; } if (noid) { if (ckdb_cmds[*which_cmds].noid) { *id = '\0'; free(cmdptr); return ckdb_cmds[*which_cmds].cmd_val; } STRNCPYSIZ(id, cmdptr, ID_SIZ); LOGERR("Listener received invalid (noid) message: '%s'", buf); free(cmdptr); return CMD_REPLY; } *trf_root = new_ktree(); *trf_store = k_new_store(transfer_free); next = data; if (next && strncmp(next, JSON_TRANSFER, JSON_TRANSFER_LEN) == 0) { json_t *json_data; json_error_t err_val; void *json_iter; const char *json_key, *json_str; json_t *json_value; int json_typ; size_t siz; bool ok; next += JSON_TRANSFER_LEN; json_data = json_loads(next, JSON_DISABLE_EOF_CHECK, &err_val); if (!json_data) { /* This REALLY shouldn't ever get an error since the input * is a json generated string * If that happens then dump lots of information */ char *text = safe_text(next); LOGERR("Json decode error from command: '%s' " "json_err=(%d:%d:%d)%s:%s input='%s'", cmd, err_val.line, err_val.column, err_val.position, err_val.source, err_val.text, text); free(text); free(cmdptr); return CMD_REPLY; } json_iter = json_object_iter(json_data); K_WLOCK(transfer_free); while (json_iter) { json_key = json_object_iter_key(json_iter); json_value = json_object_iter_value(json_iter); item = k_unlink_head(transfer_free); DATA_TRANSFER(transfer, item); ok = true; json_typ = json_typeof(json_value); switch (json_typ) { case JSON_STRING: json_str = json_string_value(json_value); siz = strlen(json_str); if (siz >= sizeof(transfer->svalue)) transfer->mvalue = strdup(json_str); else { STRNCPY(transfer->svalue, json_str); transfer->mvalue = transfer->svalue; } break; case JSON_REAL: snprintf(transfer->svalue, sizeof(transfer->svalue), "%f", json_real_value(json_value)); transfer->mvalue = transfer->svalue; break; case JSON_INTEGER: snprintf(transfer->svalue, sizeof(transfer->svalue), "%"PRId64, (int64_t)json_integer_value(json_value)); transfer->mvalue = transfer->svalue; break; case JSON_TRUE: case JSON_FALSE: snprintf(transfer->svalue, sizeof(transfer->svalue), "%s", (json_typ == JSON_TRUE) ? TRUE_STR : FALSE_STR); transfer->mvalue = transfer->svalue; break; case JSON_ARRAY: { /* only one level array of strings for now (merkletree) * ignore other data */ size_t i, len, off, count = json_array_size(json_value); json_t *json_element; bool first = true; APPEND_REALLOC_INIT(transfer->mvalue, off, len); for (i = 0; i < count; i++) { json_element = json_array_get(json_value, i); if (json_is_string(json_element)) { json_str = json_string_value(json_element); siz = strlen(json_str); if (first) first = false; else { APPEND_REALLOC(transfer->mvalue, off, len, " "); } APPEND_REALLOC(transfer->mvalue, off, len, json_str); } else LOGERR("%s() unhandled json type %d in array %s" " in cmd %s", __func__, json_typ, json_key, cmd); } } break; default: LOGERR("%s() unhandled json type %d in cmd %s", __func__, json_typ, cmd); ok = false; break; } if (ok) STRNCPY(transfer->name, json_key); if (!ok || find_in_ktree(*trf_root, item, cmp_transfer, ctx)) { if (transfer->mvalue != transfer->svalue) free(transfer->mvalue); k_add_head(transfer_free, item); } else { *trf_root = add_to_ktree(*trf_root, item, cmp_transfer); k_add_head(*trf_store, item); } json_iter = json_object_iter_next(json_data, json_iter); } K_WUNLOCK(transfer_free); json_decref(json_data); } else { K_WLOCK(transfer_free); while (next && *next) { data = next; next = strchr(data, FLDSEP); if (next) *(next++) = '\0'; eq = strchr(data, '='); if (!eq) eq = EMPTY; else *(eq++) = '\0'; item = k_unlink_head(transfer_free); DATA_TRANSFER(transfer, item); STRNCPY(transfer->name, data); STRNCPY(transfer->svalue, eq); transfer->mvalue = transfer->svalue; if (find_in_ktree(*trf_root, item, cmp_transfer, ctx)) { if (transfer->mvalue != transfer->svalue) free(transfer->mvalue); k_add_head(transfer_free, item); } else { *trf_root = add_to_ktree(*trf_root, item, cmp_transfer); k_add_head(*trf_store, item); } } K_WUNLOCK(transfer_free); } if (ckdb_cmds[*which_cmds].createdate) { item = require_name(*trf_root, "createdate", 10, NULL, reply, sizeof(reply)); if (!item) return CMD_REPLY; DATA_TRANSFER(transfer, item); txt_to_ctv("createdate", transfer->mvalue, cd, sizeof(*cd)); if (cd->tv_sec == 0) { LOGERR("%s(): failed, %s has invalid createdate '%s'", __func__, cmdptr, transfer->mvalue); free(cmdptr); return CMD_REPLY; } if (confirm_check_createdate) check_createdate_ccl(cmd, cd); } free(cmdptr); return ckdb_cmds[*which_cmds].cmd_val; } static void summarise_blocks() { K_ITEM *b_item, *b_prev, *wi_item, ss_look, *ss_item; K_TREE_CTX ctx[1], ss_ctx[1]; double diffacc, diffinv, shareacc, shareinv; tv_t now, elapsed_start, elapsed_finish; int64_t elapsed, wi_start, wi_finish; BLOCKS *blocks, *prev_blocks; WORKINFO *prev_workinfo; SHARESUMMARY looksharesummary, *sharesummary; int32_t hi, prev_hi; bool ok; setnow(&now); K_RLOCK(blocks_free); // Find the oldest, stats pending, not new, block b_item = first_in_ktree(blocks_root, ctx); while (b_item) { DATA_BLOCKS(blocks, b_item); if (CURRENT(&(blocks->expirydate)) && blocks->statsconfirmed[0] == BLOCKS_STATSPENDING && blocks->confirmed[0] != BLOCKS_NEW) break; b_item = next_in_ktree(ctx); } K_RUNLOCK(blocks_free); // None if (!b_item) return; wi_finish = blocks->workinfoid; hi = 0; K_RLOCK(workinfo_free); if (workinfo_current) { WORKINFO *wic; DATA_WORKINFO(wic, workinfo_current); hi = coinbase1height(wic->coinbase1); } K_RUNLOCK(workinfo_free); // Wait at least for the (badly named) '2nd' confirm if (hi == 0 || blocks->height >= (hi - 1)) return; diffacc = diffinv = shareacc = shareinv = 0; elapsed = 0; K_RLOCK(blocks_free); b_prev = find_prev_blocks(blocks->height); K_RUNLOCK(blocks_free); if (!b_prev) { wi_start = 0; elapsed_start.tv_sec = elapsed_start.tv_usec = 0L; prev_hi = 0; } else { DATA_BLOCKS(prev_blocks, b_prev); wi_start = prev_blocks->workinfoid; wi_item = find_workinfo(wi_start); if (!wi_item) { // This will repeat until fixed ... LOGERR("%s() block %d, but prev %d wid " "%"PRId64" is missing", __func__, blocks->height, prev_blocks->height, prev_blocks->workinfoid); return; } DATA_WORKINFO(prev_workinfo, wi_item); copy_tv(&elapsed_start, &(prev_workinfo->createdate)); prev_hi = prev_blocks->height; } elapsed_finish.tv_sec = elapsed_finish.tv_usec = 0L; // Add up the sharesummaries, abort if any SUMMARY_NEW looksharesummary.workinfoid = wi_finish; looksharesummary.userid = MAXID; looksharesummary.workername[0] = '\0'; INIT_SHARESUMMARY(&ss_look); ss_look.data = (void *)(&looksharesummary); K_RLOCK(sharesummary_free); ss_item = find_before_in_ktree(sharesummary_workinfoid_root, &ss_look, cmp_sharesummary_workinfoid, ss_ctx); if (!ss_item) { K_RUNLOCK(sharesummary_free); // This will repeat each call here until fixed ... LOGERR("%s() block %d, prev %d no sharesummaries " "on or before %"PRId64, __func__, blocks->height, prev_hi, wi_finish); return; } DATA_SHARESUMMARY(sharesummary, ss_item); while (ss_item && sharesummary->workinfoid > wi_start) { if (sharesummary->complete[0] == SUMMARY_NEW) { // Not aged yet K_RUNLOCK(sharesummary_free); return; } if (elapsed_start.tv_sec == 0 || !tv_newer(&elapsed_start, &(sharesummary->firstshare))) { copy_tv(&elapsed_start, &(sharesummary->firstshare)); } if (tv_newer(&elapsed_finish, &(sharesummary->lastshare))) copy_tv(&elapsed_finish, &(sharesummary->lastshare)); diffacc += sharesummary->diffacc; diffinv += sharesummary->diffsta + sharesummary->diffdup + sharesummary->diffhi + sharesummary-> diffrej; shareacc += sharesummary->shareacc; shareinv += sharesummary->sharesta + sharesummary->sharedup + sharesummary->sharehi + sharesummary-> sharerej; ss_item = prev_in_ktree(ss_ctx); DATA_SHARESUMMARY_NULL(sharesummary, ss_item); } K_RUNLOCK(sharesummary_free); elapsed = (int64_t)(tvdiff(&elapsed_finish, &elapsed_start) + 0.5); ok = blocks_stats(NULL, blocks->height, blocks->blockhash, diffacc, diffinv, shareacc, shareinv, elapsed, by_default, (char *)__func__, inet_default, &now); if (ok) { LOGWARNING("%s() block %d, stats confirmed " "%0.f/%.0f/%.0f/%.0f/%"PRId64, __func__, blocks->height, diffacc, diffinv, shareacc, shareinv, elapsed); } else { LOGERR("%s() block %d, failed to confirm stats", __func__, blocks->height); } } static void summarise_poolstats() { // TODO } // TODO: daily // TODO: consider limiting how much/how long this processes each time static void summarise_userstats() { K_TREE_CTX ctx[1], ctx2[1]; K_ITEM *first, *last, *new, *next, *tmp; USERSTATS *userstats, *us_first, *us_last, *us_next; double statrange, factor; bool locked, upgrade; tv_t now, process, when; PGconn *conn = NULL; int count; char error[1024]; char tvbuf1[DATE_BUFSIZ], tvbuf2[DATE_BUFSIZ]; upgrade = false; locked = false; while (1764) { error[0] = '\0'; setnow(&now); upgrade = false; locked = true; K_ILOCK(userstats_free); // confirm_summaries() doesn't call this if (!reloading) copy_tv(&process, &now); else { // During reload, base the check date on the newest statsdate last = last_in_ktree(userstats_statsdate_root, ctx); if (!last) break; DATA_USERSTATS(us_last, last); copy_tv(&process, &us_last->statsdate); } first = first_in_ktree(userstats_statsdate_root, ctx); DATA_USERSTATS_NULL(us_first, first); // Oldest non DB stat // TODO: make the index start with summarylevel? so can find faster while (first && us_first->summarylevel[0] != SUMMARY_NONE) { first = next_in_ktree(ctx); DATA_USERSTATS_NULL(us_first, first); } if (!first) break; statrange = tvdiff(&process, &(us_first->statsdate)); // Is there data ready for summarising? if (statrange <= USERSTATS_AGE) break; copy_tv(&when, &(us_first->statsdate)); /* Convert when to the start of the timeframe after the one it is in * assume timeval ignores leapseconds ... */ when.tv_sec = when.tv_sec - (when.tv_sec % USERSTATS_DB_S) + USERSTATS_DB_S; when.tv_usec = 0; // Is the whole timerange up to before 'when' ready for summarising? statrange = tvdiff(&process, &when); if (statrange < USERSTATS_AGE) break; next = next_in_ktree(ctx); upgrade = true; K_ULOCK(userstats_free); new = k_unlink_head(userstats_free); DATA_USERSTATS(userstats, new); memcpy(userstats, us_first, sizeof(USERSTATS)); userstats_root = remove_from_ktree(userstats_root, first, cmp_userstats, ctx2); userstats_statsdate_root = remove_from_ktree(userstats_statsdate_root, first, cmp_userstats_statsdate, ctx2); k_unlink_item(userstats_store, first); k_add_head(userstats_summ, first); count = 1; while (next) { DATA_USERSTATS(us_next, next); statrange = tvdiff(&when, &(us_next->statsdate)); if (statrange <= 0) break; tmp = next_in_ktree(ctx); if (us_next->summarylevel[0] == SUMMARY_NONE && us_next->userid == userstats->userid && strcmp(us_next->workername, userstats->workername) == 0) { count++; userstats->hashrate += us_next->hashrate; userstats->hashrate5m += us_next->hashrate5m; userstats->hashrate1hr += us_next->hashrate1hr; userstats->hashrate24hr += us_next->hashrate24hr; if (userstats->elapsed > us_next->elapsed) userstats->elapsed = us_next->elapsed; userstats->summarycount += us_next->summarycount; userstats_root = remove_from_ktree(userstats_root, next, cmp_userstats, ctx2); userstats_statsdate_root = remove_from_ktree(userstats_statsdate_root, next, cmp_userstats_statsdate, ctx2); k_unlink_item(userstats_store, next); k_add_head(userstats_summ, next); } next = tmp; } // Can temporarily release the lock since all our data is now not part of the lock if (upgrade) K_WUNLOCK(userstats_free); else K_IUNLOCK(userstats_free); upgrade = false; locked = false; if (userstats->hashrate5m > 0.0 || userstats->hashrate1hr > 0.0) userstats->idle = false; else userstats->idle = true; userstats->summarylevel[0] = SUMMARY_DB; userstats->summarylevel[1] = '\0'; // Expect 6 per poolinstance factor = (double)count / 6.0; userstats->hashrate *= factor; userstats->hashrate5m *= factor; userstats->hashrate1hr *= factor; userstats->hashrate24hr *= factor; copy_tv(&(userstats->statsdate), &when); // Stats to the end of this timeframe userstats->statsdate.tv_sec -= 1; userstats->statsdate.tv_usec = 999999; // This is simply when it was written, so 'now' is fine SIMPLEDATEDEFAULT(userstats, &now); if (!conn) conn = dbconnect(); if (!userstats_add_db(conn, userstats)) { /* This should only happen if a restart finds data that wasn't found during the reload but is in the same timeframe as DB data i.e. it shouldn't happen, but keep the summary anyway */ when.tv_sec -= USERSTATS_DB_S; tv_to_buf(&when, tvbuf1, sizeof(tvbuf1)); tv_to_buf(&(userstats->statsdate), tvbuf2, sizeof(tvbuf2)); snprintf(error, sizeof(error), "Userid %"PRId64" Worker %s, %d userstats record%s " "discarded from %s to %s", userstats->userid, userstats->workername, count, (count == 1 ? "" : "s"), tvbuf1, tvbuf2); } // The flags are not needed //upgrade = true; //locked = true; K_WLOCK(userstats_free); k_list_transfer_to_tail(userstats_summ, userstats_free); k_add_head(userstats_store, new); userstats_root = add_to_ktree(userstats_root, new, cmp_userstats); userstats_statsdate_root = add_to_ktree(userstats_statsdate_root, new, cmp_userstats_statsdate); K_WUNLOCK(userstats_free); //locked = false; //upgrade = false; if (error[0]) LOGERR(error); } if (locked) { if (upgrade) K_WUNLOCK(userstats_free); else K_IUNLOCK(userstats_free); } if (conn) PQfinish(conn); } static void *summariser(__maybe_unused void *arg) { pthread_detach(pthread_self()); rename_proc("db_summariser"); while (!everyone_die && !db_load_complete) cksleep_ms(42); while (!everyone_die) { sleep(5); if (!everyone_die) summarise_blocks(); sleep(4); if (!everyone_die) summarise_poolstats(); sleep(4); if (!everyone_die) summarise_userstats(); } return NULL; } static void *logger(__maybe_unused void *arg) { K_ITEM *lq_item; LOGQUEUE *lq; char buf[128]; tv_t now; pthread_detach(pthread_self()); snprintf(buf, sizeof(buf), "db%s_logger", dbcode); rename_proc(buf); setnow(&now); snprintf(buf, sizeof(buf), "logstart.%ld,%ld", now.tv_sec, now.tv_usec); LOGFILE(buf); while (!everyone_die) { K_WLOCK(logqueue_free); lq_item = k_unlink_head(logqueue_store); K_WUNLOCK(logqueue_free); while (lq_item) { DATA_LOGQUEUE(lq, lq_item); LOGFILE(lq->msg); free(lq->msg); K_WLOCK(logqueue_free); k_add_head(logqueue_free, lq_item); if (!everyone_die) lq_item = k_unlink_head(logqueue_store); else lq_item = NULL; K_WUNLOCK(logqueue_free); } cksleep_ms(42); } K_WLOCK(logqueue_free); setnow(&now); snprintf(buf, sizeof(buf), "logstopping.%d.%ld,%ld", logqueue_store->count, now.tv_sec, now.tv_usec); LOGFILE(buf); while((lq_item = k_unlink_head(logqueue_store))) { DATA_LOGQUEUE(lq, lq_item); LOGFILE(lq->msg); } K_WUNLOCK(logqueue_free); setnow(&now); snprintf(buf, sizeof(buf), "logstop.%ld,%ld", now.tv_sec, now.tv_usec); LOGFILE(buf); return NULL; } #define STORELASTREPLY(_cmd) do { \ if (last_ ## _cmd) \ free(last_ ## _cmd); \ last_ ## _cmd = buf; \ buf = NULL; \ if (reply_ ## _cmd) \ free(reply_ ## _cmd); \ reply_ ## _cmd = rep; \ } while (0) static void *socketer(__maybe_unused void *arg) { proc_instance_t *pi = (proc_instance_t *)arg; unixsock_t *us = &pi->us; char *end, *ans = NULL, *rep = NULL, *buf = NULL, *dot; char cmd[CMD_SIZ+1], id[ID_SIZ+1], reply[1024+1]; char *last_auth = NULL, *reply_auth = NULL; char *last_addrauth = NULL, *reply_addrauth = NULL; char *last_chkpass = NULL, *reply_chkpass = NULL; char *last_adduser = NULL, *reply_adduser = NULL; char *last_newpass = NULL, *reply_newpass = NULL; char *last_userset = NULL, *reply_userset = NULL; char *last_workerset = NULL, *reply_workerset = NULL; char *last_newid = NULL, *reply_newid = NULL; char *last_setatts = NULL, *reply_setatts = NULL; char *last_setopts = NULL, *reply_setopts = NULL; char *last_web = NULL, *reply_web = NULL; char *reply_last, duptype[CMD_SIZ+1]; enum cmd_values cmdnum; int sockd, which_cmds; WORKQUEUE *workqueue; TRANSFER *transfer; K_STORE *trf_store; K_TREE *trf_root; K_ITEM *item; size_t siz; tv_t now, cd; bool dup, want_first, show_dup; int loglevel, oldloglevel; pthread_detach(pthread_self()); rename_proc("db_socketer"); while (!everyone_die && !db_auths_complete) cksem_mswait(&socketer_sem, 420); want_first = true; while (!everyone_die) { if (buf) dealloc(buf); sockd = accept(us->sockd, NULL, NULL); if (sockd < 0) { LOGERR("Failed to accept on socket in listener"); break; } cmdnum = CMD_UNSET; trf_root = NULL; trf_store = NULL; buf = recv_unix_msg(sockd); // Once we've read the message setnow(&now); if (buf) { end = buf + strlen(buf) - 1; // strip trailing \n and \r while (end >= buf && (*end == '\n' || *end == '\r')) *(end--) = '\0'; } if (!buf || !*buf) { // An empty message wont get a reply if (!buf) LOGWARNING("Failed to get message in listener"); else LOGWARNING("Empty message in listener"); } else { /* For duplicates: * Queued pool messages are handled by the queue code * but since they reply ok.queued that message can * be returned every time here * System: repeat process them * Web: current php web sends a timestamp of seconds * so duplicate code will only trigger if the same * message is sent within the same second and thus * will effectively reduce the processing load for * sequential duplicates * As per the 'if' list below, * remember individual last messages and replies and * repeat the reply without reprocessing the message * The rest are remembered in the same buffer 'web' * so a duplicate will not be seen if another 'web' * command arrived between two duplicate commands */ dup = false; show_dup = true; // These are ordered approximately most likely first if (last_auth && strcmp(last_auth, buf) == 0) { reply_last = reply_auth; dup = true; } else if (last_chkpass && strcmp(last_chkpass, buf) == 0) { reply_last = reply_chkpass; dup = true; } else if (last_adduser && strcmp(last_adduser, buf) == 0) { reply_last = reply_adduser; dup = true; } else if (last_newpass && strcmp(last_newpass, buf) == 0) { reply_last = reply_newpass; dup = true; } else if (last_newid && strcmp(last_newid, buf) == 0) { reply_last = reply_newid; dup = true; } else if (last_addrauth && strcmp(last_addrauth, buf) == 0) { reply_last = reply_addrauth; dup = true; } else if (last_userset && strcmp(last_userset, buf) == 0) { reply_last = reply_userset; dup = true; } else if (last_workerset && strcmp(last_workerset, buf) == 0) { reply_last = reply_workerset; dup = true; } else if (last_setatts && strcmp(last_setatts, buf) == 0) { reply_last = reply_setatts; dup = true; } else if (last_setopts && strcmp(last_setopts, buf) == 0) { reply_last = reply_setopts; dup = true; } else if (last_web && strcmp(last_web, buf) == 0) { reply_last = reply_web; dup = true; show_dup = false; } if (dup) { send_unix_msg(sockd, reply_last); STRNCPY(duptype, buf); dot = strchr(duptype, '.'); if (dot) *dot = '\0'; snprintf(reply, sizeof(reply), "%s%ld,%ld.%s", LOGDUP, now.tv_sec, now.tv_usec, duptype); LOGQUE(reply); if (show_dup) LOGWARNING("Duplicate '%s' message received", duptype); else LOGDEBUG("Duplicate '%s' message received", duptype); } else { LOGQUE(buf); cmdnum = breakdown(&trf_root, &trf_store, buf, &which_cmds, cmd, id, &cd); switch (cmdnum) { case CMD_REPLY: snprintf(reply, sizeof(reply), "%s.%ld.?.", id, now.tv_sec); send_unix_msg(sockd, reply); break; case CMD_SHUTDOWN: LOGWARNING("Listener received shutdown message, terminating ckdb"); snprintf(reply, sizeof(reply), "%s.%ld.ok.exiting", id, now.tv_sec); send_unix_msg(sockd, reply); everyone_die = true; break; case CMD_PING: LOGDEBUG("Listener received ping request"); snprintf(reply, sizeof(reply), "%s.%ld.ok.pong", id, now.tv_sec); send_unix_msg(sockd, reply); break; case CMD_VERSION: LOGDEBUG("Listener received version request"); snprintf(reply, sizeof(reply), "%s.%ld.ok.CKDB V%s", id, now.tv_sec, CKDB_VERSION); send_unix_msg(sockd, reply); break; case CMD_LOGLEVEL: if (!*id) { LOGDEBUG("Listener received loglevel, currently %d", pi->ckp->loglevel); snprintf(reply, sizeof(reply), "%s.%ld.ok.loglevel currently %d", id, now.tv_sec, pi->ckp->loglevel); } else { oldloglevel = pi->ckp->loglevel; loglevel = atoi(id); LOGDEBUG("Listener received loglevel %d currently %d A", loglevel, oldloglevel); if (loglevel < LOG_EMERG || loglevel > LOG_DEBUG) { snprintf(reply, sizeof(reply), "%s.%ld.ERR.invalid loglevel %d" " - currently %d", id, now.tv_sec, loglevel, oldloglevel); } else { pi->ckp->loglevel = loglevel; snprintf(reply, sizeof(reply), "%s.%ld.ok.loglevel now %d - was %d", id, now.tv_sec, pi->ckp->loglevel, oldloglevel); } // Do this twice since the loglevel may have changed LOGDEBUG("Listener received loglevel %d currently %d B", loglevel, oldloglevel); } send_unix_msg(sockd, reply); break; // Always process immediately: case CMD_AUTH: case CMD_ADDRAUTH: case CMD_HEARTBEAT: // First message from the pool if (want_first) { ck_wlock(&fpm_lock); first_pool_message = strdup(buf); ck_wunlock(&fpm_lock); want_first = false; } case CMD_CHKPASS: case CMD_ADDUSER: case CMD_NEWPASS: case CMD_USERSET: case CMD_WORKERSET: case CMD_GETATTS: case CMD_SETATTS: case CMD_EXPATTS: case CMD_GETOPTS: case CMD_SETOPTS: case CMD_BLOCKLIST: case CMD_NEWID: case CMD_STATS: ans = ckdb_cmds[which_cmds].func(NULL, cmd, id, &now, by_default, (char *)__func__, inet_default, &cd, trf_root); siz = strlen(ans) + strlen(id) + 32; rep = malloc(siz); snprintf(rep, siz, "%s.%ld.%s", id, now.tv_sec, ans); send_unix_msg(sockd, rep); free(ans); ans = NULL; switch (cmdnum) { case CMD_AUTH: STORELASTREPLY(auth); break; case CMD_ADDRAUTH: STORELASTREPLY(addrauth); break; case CMD_CHKPASS: STORELASTREPLY(chkpass); break; case CMD_ADDUSER: STORELASTREPLY(adduser); break; case CMD_NEWPASS: STORELASTREPLY(newpass); break; case CMD_USERSET: STORELASTREPLY(userset); break; case CMD_WORKERSET: STORELASTREPLY(workerset); break; case CMD_NEWID: STORELASTREPLY(newid); break; case CMD_SETATTS: STORELASTREPLY(setatts); break; case CMD_SETOPTS: STORELASTREPLY(setopts); break; // The rest default: free(rep); } rep = NULL; break; // Process, but reject (loading) until startup_complete case CMD_HOMEPAGE: case CMD_ALLUSERS: case CMD_WORKERS: case CMD_PAYMENTS: case CMD_PPLNS: case CMD_DSP: case CMD_BLOCKSTATUS: if (!startup_complete) { snprintf(reply, sizeof(reply), "%s.%ld.loading.%s", id, now.tv_sec, cmd); send_unix_msg(sockd, reply); } else { ans = ckdb_cmds[which_cmds].func(NULL, cmd, id, &now, by_default, (char *)__func__, inet_default, &cd, trf_root); siz = strlen(ans) + strlen(id) + 32; rep = malloc(siz); snprintf(rep, siz, "%s.%ld.%s", id, now.tv_sec, ans); send_unix_msg(sockd, rep); free(ans); ans = NULL; if (cmdnum == CMD_DSP) free(rep); else { if (last_web) free(last_web); last_web = buf; buf = NULL; if (reply_web) free(reply_web); reply_web = rep; } rep = NULL; } break; // Always queue (ok.queued) case CMD_SHARELOG: case CMD_POOLSTAT: case CMD_USERSTAT: case CMD_BLOCK: // First message from the pool if (want_first) { ck_wlock(&fpm_lock); first_pool_message = strdup(buf); ck_wunlock(&fpm_lock); want_first = false; } snprintf(reply, sizeof(reply), "%s.%ld.ok.queued", id, now.tv_sec); send_unix_msg(sockd, reply); K_WLOCK(workqueue_free); item = k_unlink_head(workqueue_free); K_WUNLOCK(workqueue_free); DATA_WORKQUEUE(workqueue, item); workqueue->buf = buf; buf = NULL; workqueue->which_cmds = which_cmds; workqueue->cmdnum = cmdnum; STRNCPY(workqueue->cmd, cmd); STRNCPY(workqueue->id, id); copy_tv(&(workqueue->now), &now); STRNCPY(workqueue->by, by_default); STRNCPY(workqueue->code, __func__); STRNCPY(workqueue->inet, inet_default); copy_tv(&(workqueue->cd), &cd); workqueue->trf_root = trf_root; trf_root = NULL; workqueue->trf_store = trf_store; trf_store = NULL; K_WLOCK(workqueue_free); k_add_tail(workqueue_store, item); K_WUNLOCK(workqueue_free); mutex_lock(&wq_waitlock); pthread_cond_signal(&wq_waitcond); mutex_unlock(&wq_waitlock); break; // Code error default: LOGEMERG("%s() CODE ERROR unhandled message %d %.32s...", __func__, cmdnum, buf); snprintf(reply, sizeof(reply), "%s.%ld.failed.code", id, now.tv_sec); send_unix_msg(sockd, reply); break; } } } close(sockd); tick(); if (trf_root) trf_root = free_ktree(trf_root, NULL); if (trf_store) { item = trf_store->head; while (item) { DATA_TRANSFER(transfer, item); if (transfer->mvalue != transfer->svalue) free(transfer->mvalue); item = item->next; } K_WLOCK(transfer_free); k_list_transfer_to_head(trf_store, transfer_free); trf_store = k_free_store(trf_store); if (transfer_free->count == transfer_free->total && transfer_free->total > ALLOC_TRANSFER * CULL_TRANSFER) k_cull_list(transfer_free); K_WUNLOCK(transfer_free); } } if (buf) dealloc(buf); // TODO: if anyone cares, free all the dup buffers :P close_unix_socket(us->sockd, us->path); return NULL; } static bool reload_line(PGconn *conn, char *filename, uint64_t count, char *buf) { char cmd[CMD_SIZ+1], id[ID_SIZ+1]; enum cmd_values cmdnum; char *end, *ans; int which_cmds; K_STORE *trf_store = NULL; K_TREE *trf_root = NULL; TRANSFER *transfer; K_ITEM *item; tv_t now, cd; bool finished; // Once we've read the message setnow(&now); if (buf) { end = buf + strlen(buf) - 1; // strip trailing \n and \r while (end >= buf && (*end == '\n' || *end == '\r')) *(end--) = '\0'; } if (!buf || !*buf) { if (!buf) LOGERR("%s() NULL message line %"PRIu64, __func__, count); else LOGERR("%s() Empty message line %"PRIu64, __func__, count); } else { finished = false; ck_wlock(&fpm_lock); if (first_pool_message && strcmp(first_pool_message, buf) == 0) finished = true; ck_wunlock(&fpm_lock); if (finished) { LOGERR("%s() reload completed, ckpool queue match at line %"PRIu64, __func__, count); return true; } LOGQUE(buf); cmdnum = breakdown(&trf_root, &trf_store, buf, &which_cmds, cmd, id, &cd); switch (cmdnum) { // Ignore case CMD_REPLY: break; // Shouldn't be there case CMD_SHUTDOWN: case CMD_PING: case CMD_VERSION: case CMD_LOGLEVEL: // Non pool commands, shouldn't be there case CMD_ADDUSER: case CMD_NEWPASS: case CMD_CHKPASS: case CMD_USERSET: case CMD_WORKERSET: case CMD_BLOCKLIST: case CMD_BLOCKSTATUS: case CMD_NEWID: case CMD_PAYMENTS: case CMD_WORKERS: case CMD_ALLUSERS: case CMD_HOMEPAGE: case CMD_GETATTS: case CMD_SETATTS: case CMD_EXPATTS: case CMD_GETOPTS: case CMD_SETOPTS: case CMD_DSP: case CMD_STATS: case CMD_PPLNS: LOGERR("%s() Message line %"PRIu64" '%s' - invalid - ignored", __func__, count, cmd); break; case CMD_AUTH: case CMD_ADDRAUTH: case CMD_HEARTBEAT: case CMD_POOLSTAT: case CMD_USERSTAT: case CMD_BLOCK: if (confirm_sharesummary) break; case CMD_SHARELOG: ans = ckdb_cmds[which_cmds].func(conn, cmd, id, &now, by_default, (char *)__func__, inet_default, &cd, trf_root); if (ans) free(ans); break; default: // Force this switch to be updated if new cmds are added quithere(1, "%s line %"PRIu64" '%s' - not handled by reload", filename, count, cmd); break; } if (trf_root) trf_root = free_ktree(trf_root, NULL); if (trf_store) { item = trf_store->head; while (item) { DATA_TRANSFER(transfer, item); if (transfer->mvalue != transfer->svalue) free(transfer->mvalue); item = item->next; } K_WLOCK(transfer_free); k_list_transfer_to_head(trf_store, transfer_free); K_WUNLOCK(transfer_free); trf_store = k_free_store(trf_store); } } tick(); return false; } // 10Mb for now - transactiontree can be large #define MAX_READ (10 * 1024 * 1024) static char *reload_buf; /* If the reload start file is missing and -r was specified correctly: * touch the filename reported in "Failed to open 'filename'", * if ckdb aborts at the beginning of the reload, then start again */ static bool reload_from(tv_t *start) { PGconn *conn = NULL; char buf[DATE_BUFSIZ+1], run[DATE_BUFSIZ+1]; size_t rflen = strlen(restorefrom); char *missingfirst = NULL, *missinglast = NULL; int missing_count; int processing; bool finished = false, matched = false, ret = true; char *filename = NULL; uint64_t count, total; tv_t now; FILE *fp = NULL; reload_buf = malloc(MAX_READ); if (!reload_buf) quithere(1, "(%d) OOM", MAX_READ); reloading = true; copy_tv(&reload_timestamp, start); reload_timestamp.tv_sec -= reload_timestamp.tv_sec % ROLL_S; tv_to_buf(start, buf, sizeof(buf)); tv_to_buf(&reload_timestamp, run, sizeof(run)); LOGWARNING("%s(): from %s (stamp %s)", __func__, buf, run); filename = rotating_filename(restorefrom, reload_timestamp.tv_sec); fp = fopen(filename, "re"); if (!fp) quithere(1, "Failed to open '%s'", filename); setnow(&now); tvs_to_buf(&now, run, sizeof(run)); snprintf(reload_buf, MAX_READ, "reload.%s.s0", run); LOGQUE(reload_buf); conn = dbconnect(); total = 0; processing = 0; while (!everyone_die && !finished) { LOGWARNING("%s(): processing %s", __func__, filename); processing++; count = 0; while (!everyone_die && !matched && fgets_unlocked(reload_buf, MAX_READ, fp)) matched = reload_line(conn, filename, ++count, reload_buf); if (ferror(fp)) { int err = errno; quithere(1, "Read failed on %s (%d) '%s'", filename, err, strerror(err)); } LOGWARNING("%s(): %sread %"PRIu64" line%s from %s", __func__, everyone_die ? "Shutdown, aborting - " : "", count, count == 1 ? "" : "s", filename); total += count; fclose(fp); free(filename); if (everyone_die || matched) break; reload_timestamp.tv_sec += ROLL_S; if (confirm_sharesummary && tv_newer(&confirm_finish, &reload_timestamp)) { LOGWARNING("%s(): confirm range complete", __func__); break; } filename = rotating_filename(restorefrom, reload_timestamp.tv_sec); fp = fopen(filename, "re"); if (!fp) { missingfirst = strdup(filename); free(filename); filename = NULL; errno = 0; missing_count = 1; setnow(&now); now.tv_sec += ROLL_S; while (42) { reload_timestamp.tv_sec += ROLL_S; /* WARNING: if the system clock is wrong, any CCLs * missing or not created due to a ckpool outage of * an hour or more can stop the reload early and * cause DB problems! Though, the clock being wrong * can screw up ckpool and ckdb anyway ... */ if (!tv_newer(&reload_timestamp, &now)) { finished = true; break; } filename = rotating_filename(restorefrom, reload_timestamp.tv_sec); fp = fopen(filename, "re"); if (fp) break; errno = 0; if (missing_count++ > 1) free(missinglast); missinglast = strdup(filename); free(filename); filename = NULL; } if (missing_count == 1) LOGWARNING("%s(): skipped %s", __func__, missingfirst+rflen); else { LOGWARNING("%s(): skipped %d files from %s to %s", __func__, missing_count, missingfirst+rflen, missinglast+rflen); free(missinglast); missinglast = NULL; } free(missingfirst); missingfirst = NULL; } } PQfinish(conn); snprintf(reload_buf, MAX_READ, "reload.%s.%"PRIu64, run, total); LOGQUE(reload_buf); LOGWARNING("%s(): read %d file%s, total %"PRIu64" line%s", __func__, processing, processing == 1 ? "" : "s", total, total == 1 ? "" : "s"); if (everyone_die) return true; if (!matched) { ck_wlock(&fpm_lock); if (first_pool_message) { LOGERR("%s() reload completed without finding ckpool queue match '%.32s'...", __func__, first_pool_message); LOGERR("%s() restart ckdb to resolve this", __func__); ret = false; } ck_wunlock(&fpm_lock); } reloading = false; free(reload_buf); reload_buf = NULL; return ret; } static void process_queued(PGconn *conn, K_ITEM *wq_item) { static char *last_buf = NULL; WORKQUEUE *workqueue; TRANSFER *transfer; K_ITEM *item; char *ans; DATA_WORKQUEUE(workqueue, wq_item); // Simply ignore the (very rare) duplicates if (!last_buf || strcmp(workqueue->buf, last_buf)) { ans = ckdb_cmds[workqueue->which_cmds].func(conn, workqueue->cmd, workqueue->id, &(workqueue->now), workqueue->by, workqueue->code, workqueue->inet, &(workqueue->cd), workqueue->trf_root); free(ans); } if (last_buf) free(last_buf); last_buf = workqueue->buf; workqueue->trf_root = free_ktree(workqueue->trf_root, NULL); item = workqueue->trf_store->head; while (item) { DATA_TRANSFER(transfer, item); if (transfer->mvalue != transfer->svalue) free(transfer->mvalue); item = item->next; } K_WLOCK(transfer_free); k_list_transfer_to_head(workqueue->trf_store, transfer_free); K_WUNLOCK(transfer_free); workqueue->trf_store = k_free_store(workqueue->trf_store); K_WLOCK(workqueue_free); k_add_head(workqueue_free, wq_item); if (workqueue_free->count == workqueue_free->total && workqueue_free->total > ALLOC_WORKQUEUE * CULL_WORKQUEUE) k_cull_list(workqueue_free); K_WUNLOCK(workqueue_free); } // TODO: equivalent of api_allow static void *listener(void *arg) { PGconn *conn = NULL; pthread_t log_pt; pthread_t sock_pt; pthread_t summ_pt; K_ITEM *wq_item; time_t now; int wqcount, wqgot; logqueue_free = k_new_list("LogQueue", sizeof(LOGQUEUE), ALLOC_LOGQUEUE, LIMIT_LOGQUEUE, true); logqueue_store = k_new_store(logqueue_free); create_pthread(&log_pt, logger, NULL); create_pthread(&sock_pt, socketer, arg); create_pthread(&summ_pt, summariser, NULL); rename_proc("db_listener"); if (!setup_data()) { if (!everyone_die) { LOGEMERG("ABORTING"); everyone_die = true; } return NULL; } if (!everyone_die) { K_RLOCK(workqueue_store); wqcount = workqueue_store->count; K_RUNLOCK(workqueue_store); LOGWARNING("%s(): ckdb ready, queue %d", __func__, wqcount); startup_complete = true; } // Process queued work conn = dbconnect(); now = time(NULL); wqgot = 0; while (!everyone_die) { K_WLOCK(workqueue_store); wq_item = k_unlink_head(workqueue_store); K_WUNLOCK(workqueue_store); /* Don't keep a connection for more than ~10s or ~1000 items * but always have a connection open */ if ((time(NULL) - now) > 10 || wqgot > 1000) { PQfinish(conn); conn = dbconnect(); now = time(NULL); wqgot = 0; } if (wq_item) { wqgot++; process_queued(conn, wq_item); tick(); } else { const ts_t tsdiff = {0, 420000000}; tv_t now; ts_t abs; tv_time(&now); tv_to_ts(&abs, &now); timeraddspec(&abs, &tsdiff); mutex_lock(&wq_waitlock); pthread_cond_timedwait(&wq_waitcond, &wq_waitlock, &abs); mutex_unlock(&wq_waitlock); } } if (conn) PQfinish(conn); return NULL; } /* TODO: This will be way faster traversing both trees simultaneously * rather than traversing one and searching the other, then repeating * in reverse. Will change it later */ static void compare_summaries(K_TREE *leftsum, char *leftname, K_TREE *rightsum, char *rightname, bool show_missing, bool show_diff) { K_TREE_CTX ctxl[1], ctxr[1]; K_ITEM look, *lss, *rss; char cd_buf[DATE_BUFSIZ]; SHARESUMMARY looksharesummary, *l_ss, *r_ss; uint64_t total, ok, missing, diff; uint64_t first_used = 0, last_used = 0; int64_t miss_first = 0, miss_last = 0; tv_t miss_first_cd = {0,0}, miss_last_cd = {0,0}; int64_t diff_first = 0, diff_last = 0; tv_t diff_first_cd = {0,0}, diff_last_cd = {0,0}; char cd_buf1[DATE_BUFSIZ], cd_buf2[DATE_BUFSIZ]; looksharesummary.workinfoid = confirm_first_workinfoid; looksharesummary.userid = -1; looksharesummary.workername[0] = '\0'; INIT_SHARESUMMARY(&look); look.data = (void *)(&looksharesummary); total = ok = missing = diff = 0; lss = find_after_in_ktree(leftsum, &look, cmp_sharesummary_workinfoid, ctxl); while (lss) { DATA_SHARESUMMARY(l_ss, lss); if (l_ss->workinfoid > confirm_last_workinfoid) break; total++; if (first_used == 0) first_used = l_ss->workinfoid; last_used = l_ss->workinfoid; rss = find_in_ktree(rightsum, lss, cmp_sharesummary_workinfoid, ctxr); DATA_SHARESUMMARY_NULL(r_ss, rss); if (!rss) { missing++; if (miss_first == 0) { miss_first = l_ss->workinfoid; copy_tv(&miss_first_cd, &(l_ss->createdate)); } miss_last = l_ss->workinfoid; copy_tv(&miss_last_cd, &(l_ss->createdate)); if (show_missing) { LOGERR("ERROR: %s %"PRId64"/%s/%ld,%ld %.19s missing from %s", leftname, l_ss->workinfoid, l_ss->workername, l_ss->createdate.tv_sec, l_ss->createdate.tv_usec, tv_to_buf(&(l_ss->createdate), cd_buf, sizeof(cd_buf)), rightname); } } else if (r_ss->diffacc != l_ss->diffacc) { diff++; if (show_diff) { if (diff_first == 0) { diff_first = l_ss->workinfoid; copy_tv(&diff_first_cd, &(l_ss->createdate)); } diff_last = l_ss->workinfoid; copy_tv(&diff_last_cd, &(l_ss->createdate)); LOGERR("ERROR: %"PRId64"/%s/%ld,%ld %.19s - diffacc: %s: %.0f %s: %.0f", l_ss->workinfoid, l_ss->workername, l_ss->createdate.tv_sec, l_ss->createdate.tv_usec, tv_to_buf(&(l_ss->createdate), cd_buf, sizeof(cd_buf)), leftname, l_ss->diffacc, rightname, r_ss->diffacc); } } else ok++; lss = next_in_ktree(ctxl); } LOGERR("RESULT: %s->%s Total %"PRIu64" workinfoid %"PRId64"-%"PRId64 " %s missing: %"PRIu64" different: %"PRIu64, leftname, rightname, total, first_used, last_used, rightname, missing, diff); if (miss_first) { tv_to_buf(&miss_first_cd, cd_buf1, sizeof(cd_buf1)); tv_to_buf(&miss_last_cd, cd_buf2, sizeof(cd_buf2)); LOGERR(" workinfoid range for missing: %"PRId64"-%"PRId64 " (%s .. %s)", miss_first, miss_last, cd_buf1, cd_buf2); } if (show_diff && diff_first) { tv_to_buf(&diff_first_cd, cd_buf1, sizeof(cd_buf1)); tv_to_buf(&diff_last_cd, cd_buf2, sizeof(cd_buf2)); LOGERR(" workinfoid range for differences: %"PRId64"-%"PRId64 " (%s .. %s)", diff_first, diff_last, cd_buf1, cd_buf2); } } /* TODO: have a seperate option to find/store missing workinfo/shares/etc * from the reload files, in a supplied UTC time range * since there is no automatic way to get them in the DB after later ones * have been stored e.g. a database failure/recovery or short outage but * later workinfos/shares/etc have been stored so earlier ones will not be * picked up by the reload * However, will need to deal with, via reporting an error and/or abort, * if a stored workinfoid is in a range that has already been paid * and the payment is now wrong */ static void confirm_reload() { K_TREE *sharesummary_workinfoid_save; __maybe_unused K_TREE *sharesummary_save; __maybe_unused K_TREE *workinfo_save; K_ITEM b_look, wi_look, *wi_item, *wif_item, *wil_item; K_ITEM *b_begin_item, *b_end_item; K_ITEM *ss_begin_item, *ss_end_item; WORKINFO lookworkinfo, *workinfo; BLOCKS lookblocks, *b_blocks, *e_blocks; SHARESUMMARY *b_ss, *e_ss; K_TREE_CTX ctx[1]; char buf[DATE_BUFSIZ+1]; char *first_reason; char *last_reason; char cd_buf[DATE_BUFSIZ]; char first_buf[64], last_buf[64]; char *filename; tv_t start; FILE *fp; // TODO: // abort reload when we get an age after the end of a workinfo after the Xs after the last workinfo before the end wif_item = first_in_ktree(workinfo_root, ctx); wil_item = last_in_ktree(workinfo_root, ctx); if (!wif_item || !wil_item) { LOGWARNING("%s(): DB contains no workinfo records", __func__); return; } DATA_WORKINFO(workinfo, wif_item); tv_to_buf(&(workinfo->createdate), cd_buf, sizeof(cd_buf)); LOGWARNING("%s(): DB first workinfoid %"PRId64" %s", __func__, workinfo->workinfoid, cd_buf); DATA_WORKINFO(workinfo, wil_item); tv_to_buf(&(workinfo->createdate), cd_buf, sizeof(cd_buf)); LOGWARNING("%s(): DB last workinfoid %"PRId64" %s", __func__, workinfo->workinfoid, cd_buf); b_begin_item = first_in_ktree(blocks_root, ctx); b_end_item = last_in_ktree(blocks_root, ctx); if (!b_begin_item || !b_end_item) LOGWARNING("%s(): DB contains no blocks :(", __func__); else { DATA_BLOCKS(b_blocks, b_begin_item); DATA_BLOCKS(e_blocks, b_end_item); tv_to_buf(&(b_blocks->createdate), cd_buf, sizeof(cd_buf)); LOGWARNING("%s(): DB first block %d/%"PRId64" %s", __func__, b_blocks->height, b_blocks->workinfoid, cd_buf); tv_to_buf(&(e_blocks->createdate), cd_buf, sizeof(cd_buf)); LOGWARNING("%s(): DB last block %d/%"PRId64" %s", __func__, e_blocks->height, e_blocks->workinfoid, cd_buf); } ss_begin_item = first_in_ktree(sharesummary_workinfoid_root, ctx); ss_end_item = last_in_ktree(sharesummary_workinfoid_root, ctx); if (!ss_begin_item || !ss_end_item) LOGWARNING("%s(): DB contains no sharesummary records", __func__); else { DATA_SHARESUMMARY(b_ss, ss_begin_item); DATA_SHARESUMMARY(e_ss, ss_end_item); tv_to_buf(&(b_ss->createdate), cd_buf, sizeof(cd_buf)); LOGWARNING("%s(): DB first sharesummary %"PRId64"/%s %s", __func__, b_ss->workinfoid, b_ss->workername, cd_buf); tv_to_buf(&(e_ss->createdate), cd_buf, sizeof(cd_buf)); LOGWARNING("%s(): DB last sharesummary %"PRId64"/%s %s", __func__, e_ss->workinfoid, e_ss->workername, cd_buf); } /* The first workinfo we should process * With no y records we should start from the beginning (0) * With any y records, we should start from the oldest of: y+1 and a * which can produce y records as reload a's, if a is used */ if (dbstatus.newest_workinfoid_y > 0) { confirm_first_workinfoid = dbstatus.newest_workinfoid_y + 1; if (confirm_first_workinfoid > dbstatus.oldest_workinfoid_a) { confirm_first_workinfoid = dbstatus.oldest_workinfoid_a; first_reason = "oldest aged"; } else first_reason = "newest confirmed+1"; } else first_reason = "0 - none confirmed"; /* The last workinfo we should process * The reason for going past the last 'a' up to before * the first 'n' is in case there were shares missed between them - * but that should only be the case with a code bug - * so it checks that */ if (dbstatus.newest_workinfoid_a > 0) { confirm_last_workinfoid = dbstatus.newest_workinfoid_a; last_reason = "newest aged"; } if (confirm_last_workinfoid < dbstatus.oldest_workinfoid_n) { confirm_last_workinfoid = dbstatus.oldest_workinfoid_n - 1; last_reason = "oldest new-1"; } if (confirm_last_workinfoid == 0) { LOGWARNING("%s(): there are no unconfirmed sharesummary records in the DB", __func__, buf); return; } INIT_BLOCKS(&b_look); INIT_WORKINFO(&wi_look); // Do this after above code for checking and so we can use the results if (confirm_range && *confirm_range) { switch(tolower(confirm_range[0])) { case 'b': // First DB record of the block = or after confirm_block lookblocks.height = confirm_block; lookblocks.blockhash[0] = '\0'; b_look.data = (void *)(&lookblocks); b_end_item = find_after_in_ktree(blocks_root, &b_look, cmp_blocks, ctx); if (!b_end_item) { LOGWARNING("%s(): no DB block height found matching or after %d", __func__, confirm_block); return; } DATA_BLOCKS(e_blocks, b_end_item); confirm_last_workinfoid = e_blocks->workinfoid; // Now find the last DB record of the previous block lookblocks.height = e_blocks->height; lookblocks.blockhash[0] = '\0'; b_look.data = (void *)(&lookblocks); b_begin_item = find_before_in_ktree(blocks_root, &b_look, cmp_blocks, ctx); if (!b_begin_item) confirm_first_workinfoid = 0; else { DATA_BLOCKS(b_blocks, b_begin_item); // First DB record of the block 'begin' lookblocks.height = b_blocks->height; lookblocks.blockhash[0] = '\0'; b_look.data = (void *)(&lookblocks); b_begin_item = find_after_in_ktree(blocks_root, &b_look, cmp_blocks, ctx); // Not possible if (!b_begin_item) confirm_first_workinfoid = 0; else { DATA_BLOCKS(b_blocks, b_begin_item); confirm_first_workinfoid = b_blocks->workinfoid; } } snprintf(first_buf, sizeof(first_buf), "block %d", b_begin_item ? b_blocks->height : 0); first_reason = first_buf; snprintf(last_buf, sizeof(last_buf), "block %d", e_blocks->height); last_reason = last_buf; break; case 'i': LOGWARNING("%s(): info displayed - exiting", __func__); exit(0); case 'c': case 'r': confirm_first_workinfoid = confirm_range_start; confirm_last_workinfoid = confirm_range_finish; first_reason = "start range"; last_reason = "end range"; break; case 'w': confirm_first_workinfoid = confirm_range_start; // last from default if (confirm_last_workinfoid < confirm_first_workinfoid) { LOGWARNING("%s(): no unconfirmed sharesummary records before start", __func__, buf); return; } first_reason = "start range"; break; default: quithere(1, "Code fail"); } } /* These two below find the closest valid workinfo to the ones chosen * however we still use the original ones chosen to select/ignore data */ /* Find the workinfo before confirm_first_workinfoid+1 * i.e. the one we want or the previous before it */ lookworkinfo.workinfoid = confirm_first_workinfoid + 1; lookworkinfo.expirydate.tv_sec = date_begin.tv_sec; lookworkinfo.expirydate.tv_usec = date_begin.tv_usec; wi_look.data = (void *)(&lookworkinfo); wi_item = find_before_in_ktree(workinfo_root, &wi_look, cmp_workinfo, ctx); if (wi_item) { DATA_WORKINFO(workinfo, wi_item); copy_tv(&start, &(workinfo->createdate)); if (workinfo->workinfoid != confirm_first_workinfoid) { LOGWARNING("%s() start workinfo not found ... using time of %"PRId64, __func__, workinfo->workinfoid); } } else { start.tv_sec = start.tv_usec = 0; LOGWARNING("%s() no start workinfo found ... using time 0", __func__); } /* Find the workinfo after confirm_last_workinfoid-1 * i.e. the one we want or the next after it */ lookworkinfo.workinfoid = confirm_last_workinfoid - 1; lookworkinfo.expirydate.tv_sec = date_eot.tv_sec; lookworkinfo.expirydate.tv_usec = date_eot.tv_usec; wi_look.data = (void *)(&lookworkinfo); wi_item = find_after_in_ktree(workinfo_root, &wi_look, cmp_workinfo, ctx); if (wi_item) { DATA_WORKINFO(workinfo, wi_item); /* Now find the one after the one we found to determine the * confirm_finish timestamp */ lookworkinfo.workinfoid = workinfo->workinfoid; lookworkinfo.expirydate.tv_sec = date_eot.tv_sec; lookworkinfo.expirydate.tv_usec = date_eot.tv_usec; wi_look.data = (void *)(&lookworkinfo); wi_item = find_after_in_ktree(workinfo_root, &wi_look, cmp_workinfo, ctx); if (wi_item) { DATA_WORKINFO(workinfo, wi_item); copy_tv(&confirm_finish, &(workinfo->createdate)); confirm_finish.tv_sec += WORKINFO_AGE; } else { confirm_finish.tv_sec = date_eot.tv_sec; confirm_finish.tv_usec = date_eot.tv_usec; } } else { confirm_finish.tv_sec = date_eot.tv_sec; confirm_finish.tv_usec = date_eot.tv_usec; LOGWARNING("%s() no finish workinfo found ... using EOT", __func__); } LOGWARNING("%s() workinfo range: %"PRId64" to %"PRId64" ('%s' to '%s')", __func__, confirm_first_workinfoid, confirm_last_workinfoid, first_reason, last_reason); tv_to_buf(&start, buf, sizeof(buf)); LOGWARNING("%s() load start timestamp %s", __func__, buf); tv_to_buf(&confirm_finish, buf, sizeof(buf)); LOGWARNING("%s() load finish timestamp %s", __func__, buf); /* Save the DB info for comparing to the reload * i.e. the reload will generate from scratch all the * sharesummaries and workinfo from the CCLs */ sharesummary_workinfoid_save = sharesummary_workinfoid_root; sharesummary_save = sharesummary_root; workinfo_save = workinfo_root; sharesummary_workinfoid_root = new_ktree(); sharesummary_root = new_ktree(); workinfo_root = new_ktree(); if (start.tv_sec < DATE_BEGIN) { start.tv_sec = DATE_BEGIN; start.tv_usec = 0L; filename = rotating_filename(restorefrom, start.tv_sec); fp = fopen(filename, "re"); if (fp) fclose(fp); else { mode_t mode = S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH; int fd = open(filename, O_CREAT|O_RDONLY, mode); if (fd == -1) { int ern = errno; quithere(1, "Couldn't create '%s' (%d) %s", filename, ern, strerror(ern)); } close(fd); } free(filename); } if (!reload_from(&start)) { LOGEMERG("%s() ABORTING from reload_from()", __func__); return; } if (confirm_check_createdate) { LOGERR("%s(): CCL mismatches %"PRId64"/%"PRId64" %.6f/%.6f unordered " "%"PRId64"/%"PRId64" %.6f", __func__, ccl_mismatch, ccl_mismatch_abs, ccl_mismatch_min, ccl_mismatch_max, ccl_unordered, ccl_unordered_abs, ccl_unordered_most); return; } compare_summaries(sharesummary_workinfoid_save, "DB", sharesummary_workinfoid_root, "ReLoad", true, true); compare_summaries(sharesummary_workinfoid_root, "ReLoad", sharesummary_workinfoid_save, "DB", true, false); } static void confirm_summaries() { pthread_t log_pt; char *range, *minus; // Simple value check to abort early if (confirm_range && *confirm_range) { switch(tolower(confirm_range[0])) { case 'b': case 'c': case 'r': case 'w': if (strlen(confirm_range) < 2) { LOGEMERG("%s() invalid confirm range length '%s'", __func__, confirm_range); return; } break; case 'i': break; default: LOGEMERG("%s() invalid confirm range '%s'", __func__, confirm_range); return; } switch(tolower(confirm_range[0])) { case 'b': confirm_block = atoi(confirm_range+1); if (confirm_block <= 0) { LOGEMERG("%s() invalid confirm block '%s' - must be >0", __func__, confirm_range); return; } break; case 'i': break; case 'c': confirm_check_createdate = true; case 'r': range = strdup(confirm_range); minus = strchr(range+1, '-'); if (!minus || minus == range+1) { LOGEMERG("%s() invalid confirm range '%s' - must be %cNNN-MMM", __func__, confirm_range, tolower(confirm_range[0])); return; } *(minus++) = '\0'; confirm_range_start = atoll(range+1); if (confirm_range_start <= 0) { LOGEMERG("%s() invalid confirm start in '%s' - must be >0", __func__, confirm_range); return; } confirm_range_finish = atoll(minus); if (confirm_range_finish <= 0) { LOGEMERG("%s() invalid confirm finish in '%s' - must be >0", __func__, confirm_range); return; } if (confirm_range_finish < confirm_range_start) { LOGEMERG("%s() invalid confirm range in '%s' - finish < start", __func__, confirm_range); return; } free(range); break; case 'w': confirm_range_start = atoll(confirm_range+1); if (confirm_range_start <= 0) { LOGEMERG("%s() invalid confirm start '%s' - must be >0", __func__, confirm_range); return; } } } logqueue_free = k_new_list("LogQueue", sizeof(LOGQUEUE), ALLOC_LOGQUEUE, LIMIT_LOGQUEUE, true); logqueue_store = k_new_store(logqueue_free); create_pthread(&log_pt, logger, NULL); rename_proc("dby_confirmer"); alloc_storage(); if (!getdata1()) { LOGEMERG("%s() ABORTING from getdata1()", __func__); return; } if (!getdata2()) { LOGEMERG("%s() ABORTING from getdata2()", __func__); return; } confirm_reload(); } static void check_restore_dir(char *name) { struct stat statbuf; if (!restorefrom) { restorefrom = strdup("logs"); if (!restorefrom) quithere(1, "OOM"); } if (!(*restorefrom)) quit(1, "ERR: '-r dir' can't be empty"); trail_slash(&restorefrom); if (stat(restorefrom, &statbuf)) quit(1, "ERR: -r '%s' directory doesn't exist", restorefrom); restorefrom = realloc(restorefrom, strlen(restorefrom)+strlen(name)+1); if (!restorefrom) quithere(1, "OOM"); strcat(restorefrom, name); } static struct option long_options[] = { { "dbprefix", required_argument, 0, 'b' }, { "config", required_argument, 0, 'c' }, { "dbname", required_argument, 0, 'd' }, { "help", no_argument, 0, 'h' }, { "killold", no_argument, 0, 'k' }, { "loglevel", required_argument, 0, 'l' }, { "name", required_argument, 0, 'n' }, { "dbpass", required_argument, 0, 'p' }, { "ckpool-logdir", required_argument, 0, 'r' }, { "logdir", required_argument, 0, 'R' }, { "sockdir", required_argument, 0, 's' }, { "dbuser", required_argument, 0, 'u' }, { "version", no_argument, 0, 'v' }, { "confirm", no_argument, 0, 'y' }, { "confirmrange", required_argument, 0, 'Y' }, { 0, 0, 0, 0 } }; static void sighandler(int sig) { LOGWARNING("Received signal %d, shutting down", sig); everyone_die = true; cksleep_ms(420); exit(0); } int main(int argc, char **argv) { struct sigaction handler; char buf[512]; ckpool_t ckp; int c, ret, i = 0, j; char *kill; tv_t now; printf("CKDB Master V%s (C) Kano (see source code)\n", CKDB_VERSION); feenableexcept(FE_DIVBYZERO | FE_INVALID | FE_OVERFLOW); global_ckp = &ckp; memset(&ckp, 0, sizeof(ckp)); ckp.loglevel = LOG_NOTICE; while ((c = getopt_long(argc, argv, "c:d:hkl:n:p:r:R:s:u:vyY:", long_options, &i)) != -1) { switch(c) { case 'c': ckp.config = strdup(optarg); break; case 'd': db_name = strdup(optarg); kill = optarg; while (*kill) *(kill++) = ' '; break; case 'h': for (j = 0; long_options[j].val; j++) { struct option *jopt = &long_options[j]; if (jopt->has_arg) { char *upper = alloca(strlen(jopt->name) + 1); int offset = 0; do { upper[offset] = toupper(jopt->name[offset]); } while (upper[offset++] != '\0'); printf("-%c %s | --%s %s\n", jopt->val, upper, jopt->name, upper); } else printf("-%c | --%s\n", jopt->val, jopt->name); } exit(0); case 'k': ckp.killold = true; break; case 'l': ckp.loglevel = atoi(optarg); if (ckp.loglevel < LOG_EMERG || ckp.loglevel > LOG_DEBUG) { quit(1, "Invalid loglevel (range %d - %d): %d", LOG_EMERG, LOG_DEBUG, ckp.loglevel); } break; case 'n': ckp.name = strdup(optarg); break; case 'p': db_pass = strdup(optarg); kill = optarg; if (*kill) *(kill++) = ' '; while (*kill) *(kill++) = '\0'; break; case 'r': restorefrom = strdup(optarg); break; case 'R': ckp.logdir = strdup(optarg); break; case 's': ckp.socket_dir = strdup(optarg); break; case 'u': db_user = strdup(optarg); kill = optarg; while (*kill) *(kill++) = ' '; break; case 'v': exit(0); case 'y': confirm_sharesummary = true; break; case 'Y': confirm_range = strdup(optarg); // Auto enable it also confirm_sharesummary = true; break; } } if (confirm_sharesummary) dbcode = "y"; else dbcode = ""; if (!db_name) db_name = "ckdb"; if (!db_user) db_user = "postgres"; if (!ckp.name) ckp.name = "ckdb"; snprintf(buf, 15, "%s%s", ckp.name, dbcode); prctl(PR_SET_NAME, buf, 0, 0, 0); memset(buf, 0, 15); check_restore_dir(ckp.name); if (!ckp.config) { ckp.config = strdup(ckp.name); realloc_strcat(&ckp.config, ".conf"); } if (!ckp.socket_dir) { ckp.socket_dir = strdup("/opt/"); realloc_strcat(&ckp.socket_dir, ckp.name); } trail_slash(&ckp.socket_dir); /* Ignore sigpipe */ signal(SIGPIPE, SIG_IGN); ret = mkdir(ckp.socket_dir, 0770); if (ret && errno != EEXIST) quit(1, "Failed to make directory %s", ckp.socket_dir); // parse_config(&ckp); if (!ckp.logdir) ckp.logdir = strdup("dblogs"); /* Create the log directory */ trail_slash(&ckp.logdir); ret = mkdir(ckp.logdir, 0700); if (ret && errno != EEXIST) quit(1, "Failed to make log directory %s", ckp.logdir); /* Create the logfile */ sprintf(buf, "%s%s%s.log", ckp.logdir, ckp.name, dbcode); ckp.logfp = fopen(buf, "ae"); if (!ckp.logfp) quit(1, "Failed to open log file %s", buf); ckp.logfd = fileno(ckp.logfp); snprintf(logname, sizeof(logname), "%s%s-db%s-", ckp.logdir, ckp.name, dbcode); setnow(&now); srandom((unsigned int)(now.tv_usec * 4096 + now.tv_sec % 4096)); ckp.main.ckp = &ckp; ckp.main.processname = strdup("main"); if (confirm_sharesummary) { // TODO: add a system lock to stop running 2 at once? confirm_summaries(); } else { ckp.main.sockname = strdup("listener"); write_namepid(&ckp.main); create_process_unixsock(&ckp.main); create_pthread(&ckp.pth_listener, listener, &ckp.main); handler.sa_handler = sighandler; handler.sa_flags = 0; sigemptyset(&handler.sa_mask); sigaction(SIGTERM, &handler, NULL); sigaction(SIGINT, &handler, NULL); /* Shutdown from here if the listener is sent a shutdown message */ join_pthread(ckp.pth_listener); } clean_up(&ckp); return 0; }