X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/ca734d17ad9b7ea97a87bd048ad1c1e253ef67a6..bfe85f7ca97259256e8089349e1a462b6c7dbd00:/src/redis.c?ds=sidebyside diff --git a/src/redis.c b/src/redis.c index 99ff1898..63b41ba8 100644 --- a/src/redis.c +++ b/src/redis.c @@ -68,147 +68,179 @@ double R_Zero, R_PosInf, R_NegInf, R_Nan; /* Global vars */ struct redisServer server; /* server global state */ struct redisCommand *commandTable; -struct redisCommand readonlyCommandTable[] = { - {"get",getCommand,2,0,NULL,1,1,1}, - {"set",setCommand,3,REDIS_CMD_DENYOOM,NULL,0,0,0}, - {"setnx",setnxCommand,3,REDIS_CMD_DENYOOM,NULL,0,0,0}, - {"setex",setexCommand,4,REDIS_CMD_DENYOOM,NULL,0,0,0}, - {"append",appendCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"substr",substrCommand,4,0,NULL,1,1,1}, - {"strlen",strlenCommand,2,0,NULL,1,1,1}, - {"del",delCommand,-2,0,NULL,0,0,0}, - {"exists",existsCommand,2,0,NULL,1,1,1}, - {"incr",incrCommand,2,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"decr",decrCommand,2,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"mget",mgetCommand,-2,0,NULL,1,-1,1}, - {"rpush",rpushCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"lpush",lpushCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"rpushx",rpushxCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"lpushx",lpushxCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"linsert",linsertCommand,5,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"rpop",rpopCommand,2,0,NULL,1,1,1}, - {"lpop",lpopCommand,2,0,NULL,1,1,1}, - {"brpop",brpopCommand,-3,0,NULL,1,1,1}, - {"blpop",blpopCommand,-3,0,NULL,1,1,1}, - {"llen",llenCommand,2,0,NULL,1,1,1}, - {"lindex",lindexCommand,3,0,NULL,1,1,1}, - {"lset",lsetCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"lrange",lrangeCommand,4,0,NULL,1,1,1}, - {"ltrim",ltrimCommand,4,0,NULL,1,1,1}, - {"lrem",lremCommand,4,0,NULL,1,1,1}, - {"rpoplpush",rpoplpushcommand,3,REDIS_CMD_DENYOOM,NULL,1,2,1}, - {"sadd",saddCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"srem",sremCommand,3,0,NULL,1,1,1}, - {"smove",smoveCommand,4,0,NULL,1,2,1}, - {"sismember",sismemberCommand,3,0,NULL,1,1,1}, - {"scard",scardCommand,2,0,NULL,1,1,1}, - {"spop",spopCommand,2,0,NULL,1,1,1}, - {"srandmember",srandmemberCommand,2,0,NULL,1,1,1}, - {"sinter",sinterCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1}, - {"sinterstore",sinterstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1}, - {"sunion",sunionCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1}, - {"sunionstore",sunionstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1}, - {"sdiff",sdiffCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1}, - {"sdiffstore",sdiffstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1}, - {"smembers",sinterCommand,2,0,NULL,1,1,1}, - {"zadd",zaddCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"zincrby",zincrbyCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"zrem",zremCommand,3,0,NULL,1,1,1}, - {"zremrangebyscore",zremrangebyscoreCommand,4,0,NULL,1,1,1}, - {"zremrangebyrank",zremrangebyrankCommand,4,0,NULL,1,1,1}, - {"zunionstore",zunionstoreCommand,-4,REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0}, - {"zinterstore",zinterstoreCommand,-4,REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0}, - {"zrange",zrangeCommand,-4,0,NULL,1,1,1}, - {"zrangebyscore",zrangebyscoreCommand,-4,0,NULL,1,1,1}, - {"zrevrangebyscore",zrevrangebyscoreCommand,-4,0,NULL,1,1,1}, - {"zcount",zcountCommand,4,0,NULL,1,1,1}, - {"zrevrange",zrevrangeCommand,-4,0,NULL,1,1,1}, - {"zcard",zcardCommand,2,0,NULL,1,1,1}, - {"zscore",zscoreCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"zrank",zrankCommand,3,0,NULL,1,1,1}, - {"zrevrank",zrevrankCommand,3,0,NULL,1,1,1}, - {"hset",hsetCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"hsetnx",hsetnxCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"hget",hgetCommand,3,0,NULL,1,1,1}, - {"hmset",hmsetCommand,-4,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"hmget",hmgetCommand,-3,0,NULL,1,1,1}, - {"hincrby",hincrbyCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"hdel",hdelCommand,3,0,NULL,1,1,1}, - {"hlen",hlenCommand,2,0,NULL,1,1,1}, - {"hkeys",hkeysCommand,2,0,NULL,1,1,1}, - {"hvals",hvalsCommand,2,0,NULL,1,1,1}, - {"hgetall",hgetallCommand,2,0,NULL,1,1,1}, - {"hexists",hexistsCommand,3,0,NULL,1,1,1}, - {"incrby",incrbyCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"decrby",decrbyCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"getset",getsetCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"mset",msetCommand,-3,REDIS_CMD_DENYOOM,NULL,1,-1,2}, - {"msetnx",msetnxCommand,-3,REDIS_CMD_DENYOOM,NULL,1,-1,2}, - {"randomkey",randomkeyCommand,1,0,NULL,0,0,0}, - {"select",selectCommand,2,0,NULL,0,0,0}, - {"move",moveCommand,3,0,NULL,1,1,1}, - {"rename",renameCommand,3,0,NULL,1,1,1}, - {"renamenx",renamenxCommand,3,0,NULL,1,1,1}, - {"expire",expireCommand,3,0,NULL,0,0,0}, - {"expireat",expireatCommand,3,0,NULL,0,0,0}, - {"keys",keysCommand,2,0,NULL,0,0,0}, - {"dbsize",dbsizeCommand,1,0,NULL,0,0,0}, - {"auth",authCommand,2,0,NULL,0,0,0}, - {"ping",pingCommand,1,0,NULL,0,0,0}, - {"echo",echoCommand,2,0,NULL,0,0,0}, - {"save",saveCommand,1,0,NULL,0,0,0}, - {"bgsave",bgsaveCommand,1,0,NULL,0,0,0}, - {"bgrewriteaof",bgrewriteaofCommand,1,0,NULL,0,0,0}, - {"shutdown",shutdownCommand,1,0,NULL,0,0,0}, - {"lastsave",lastsaveCommand,1,0,NULL,0,0,0}, - {"type",typeCommand,2,0,NULL,1,1,1}, - {"multi",multiCommand,1,0,NULL,0,0,0}, - {"exec",execCommand,1,REDIS_CMD_DENYOOM,execBlockClientOnSwappedKeys,0,0,0}, - {"discard",discardCommand,1,0,NULL,0,0,0}, - {"sync",syncCommand,1,0,NULL,0,0,0}, - {"flushdb",flushdbCommand,1,0,NULL,0,0,0}, - {"flushall",flushallCommand,1,0,NULL,0,0,0}, - {"sort",sortCommand,-2,REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"info",infoCommand,1,0,NULL,0,0,0}, - {"monitor",monitorCommand,1,0,NULL,0,0,0}, - {"ttl",ttlCommand,2,0,NULL,1,1,1}, - {"persist",persistCommand,2,0,NULL,1,1,1}, - {"slaveof",slaveofCommand,3,0,NULL,0,0,0}, - {"debug",debugCommand,-2,0,NULL,0,0,0}, - {"config",configCommand,-2,0,NULL,0,0,0}, - {"subscribe",subscribeCommand,-2,0,NULL,0,0,0}, - {"unsubscribe",unsubscribeCommand,-1,0,NULL,0,0,0}, - {"psubscribe",psubscribeCommand,-2,0,NULL,0,0,0}, - {"punsubscribe",punsubscribeCommand,-1,0,NULL,0,0,0}, - {"publish",publishCommand,3,REDIS_CMD_FORCE_REPLICATION,NULL,0,0,0}, - {"watch",watchCommand,-2,0,NULL,0,0,0}, - {"unwatch",unwatchCommand,1,0,NULL,0,0,0} +struct redisCommand redisCommandTable[] = { + {"get",getCommand,2,0,NULL,1,1,1,0,0}, + {"set",setCommand,3,REDIS_CMD_DENYOOM,noPreloadGetKeys,1,1,1,0,0}, + {"setnx",setnxCommand,3,REDIS_CMD_DENYOOM,noPreloadGetKeys,1,1,1,0,0}, + {"setex",setexCommand,4,REDIS_CMD_DENYOOM,noPreloadGetKeys,2,2,1,0,0}, + {"append",appendCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"strlen",strlenCommand,2,0,NULL,1,1,1,0,0}, + {"del",delCommand,-2,0,noPreloadGetKeys,1,-1,1,0,0}, + {"exists",existsCommand,2,0,NULL,1,1,1,0,0}, + {"setbit",setbitCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"getbit",getbitCommand,3,0,NULL,1,1,1,0,0}, + {"setrange",setrangeCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"getrange",getrangeCommand,4,0,NULL,1,1,1,0,0}, + {"substr",getrangeCommand,4,0,NULL,1,1,1,0,0}, + {"incr",incrCommand,2,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"decr",decrCommand,2,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"mget",mgetCommand,-2,0,NULL,1,-1,1,0,0}, + {"rpush",rpushCommand,-3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"lpush",lpushCommand,-3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"rpushx",rpushxCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"lpushx",lpushxCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"linsert",linsertCommand,5,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"rpop",rpopCommand,2,0,NULL,1,1,1,0,0}, + {"lpop",lpopCommand,2,0,NULL,1,1,1,0,0}, + {"brpop",brpopCommand,-3,0,NULL,1,1,1,0,0}, + {"brpoplpush",brpoplpushCommand,4,REDIS_CMD_DENYOOM,NULL,1,2,1,0,0}, + {"blpop",blpopCommand,-3,0,NULL,1,-2,1,0,0}, + {"llen",llenCommand,2,0,NULL,1,1,1,0,0}, + {"lindex",lindexCommand,3,0,NULL,1,1,1,0,0}, + {"lset",lsetCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"lrange",lrangeCommand,4,0,NULL,1,1,1,0,0}, + {"ltrim",ltrimCommand,4,0,NULL,1,1,1,0,0}, + {"lrem",lremCommand,4,0,NULL,1,1,1,0,0}, + {"rpoplpush",rpoplpushCommand,3,REDIS_CMD_DENYOOM,NULL,1,2,1,0,0}, + {"sadd",saddCommand,-3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"srem",sremCommand,-3,0,NULL,1,1,1,0,0}, + {"smove",smoveCommand,4,0,NULL,1,2,1,0,0}, + {"sismember",sismemberCommand,3,0,NULL,1,1,1,0,0}, + {"scard",scardCommand,2,0,NULL,1,1,1,0,0}, + {"spop",spopCommand,2,0,NULL,1,1,1,0,0}, + {"srandmember",srandmemberCommand,2,0,NULL,1,1,1,0,0}, + {"sinter",sinterCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1,0,0}, + {"sinterstore",sinterstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1,0,0}, + {"sunion",sunionCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1,0,0}, + {"sunionstore",sunionstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1,0,0}, + {"sdiff",sdiffCommand,-2,REDIS_CMD_DENYOOM,NULL,1,-1,1,0,0}, + {"sdiffstore",sdiffstoreCommand,-3,REDIS_CMD_DENYOOM,NULL,2,-1,1,0,0}, + {"smembers",sinterCommand,2,0,NULL,1,1,1,0,0}, + {"zadd",zaddCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"zincrby",zincrbyCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"zrem",zremCommand,3,0,NULL,1,1,1,0,0}, + {"zremrangebyscore",zremrangebyscoreCommand,4,0,NULL,1,1,1,0,0}, + {"zremrangebyrank",zremrangebyrankCommand,4,0,NULL,1,1,1,0,0}, + {"zunionstore",zunionstoreCommand,-4,REDIS_CMD_DENYOOM,zunionInterGetKeys,0,0,0,0,0}, + {"zinterstore",zinterstoreCommand,-4,REDIS_CMD_DENYOOM,zunionInterGetKeys,0,0,0,0,0}, + {"zrange",zrangeCommand,-4,0,NULL,1,1,1,0,0}, + {"zrangebyscore",zrangebyscoreCommand,-4,0,NULL,1,1,1,0,0}, + {"zrevrangebyscore",zrevrangebyscoreCommand,-4,0,NULL,1,1,1,0,0}, + {"zcount",zcountCommand,4,0,NULL,1,1,1,0,0}, + {"zrevrange",zrevrangeCommand,-4,0,NULL,1,1,1,0,0}, + {"zcard",zcardCommand,2,0,NULL,1,1,1,0,0}, + {"zscore",zscoreCommand,3,0,NULL,1,1,1,0,0}, + {"zrank",zrankCommand,3,0,NULL,1,1,1,0,0}, + {"zrevrank",zrevrankCommand,3,0,NULL,1,1,1,0,0}, + {"hset",hsetCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"hsetnx",hsetnxCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"hget",hgetCommand,3,0,NULL,1,1,1,0,0}, + {"hmset",hmsetCommand,-4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"hmget",hmgetCommand,-3,0,NULL,1,1,1,0,0}, + {"hincrby",hincrbyCommand,4,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"hdel",hdelCommand,-3,0,NULL,1,1,1,0,0}, + {"hlen",hlenCommand,2,0,NULL,1,1,1,0,0}, + {"hkeys",hkeysCommand,2,0,NULL,1,1,1,0,0}, + {"hvals",hvalsCommand,2,0,NULL,1,1,1,0,0}, + {"hgetall",hgetallCommand,2,0,NULL,1,1,1,0,0}, + {"hexists",hexistsCommand,3,0,NULL,1,1,1,0,0}, + {"incrby",incrbyCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"decrby",decrbyCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"getset",getsetCommand,3,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"mset",msetCommand,-3,REDIS_CMD_DENYOOM,NULL,1,-1,2,0,0}, + {"msetnx",msetnxCommand,-3,REDIS_CMD_DENYOOM,NULL,1,-1,2,0,0}, + {"randomkey",randomkeyCommand,1,0,NULL,0,0,0,0,0}, + {"select",selectCommand,2,0,NULL,0,0,0,0,0}, + {"move",moveCommand,3,0,NULL,1,1,1,0,0}, + {"rename",renameCommand,3,0,renameGetKeys,1,2,1,0,0}, + {"renamenx",renamenxCommand,3,0,renameGetKeys,1,2,1,0,0}, + {"expire",expireCommand,3,0,NULL,1,1,1,0,0}, + {"expireat",expireatCommand,3,0,NULL,1,1,1,0,0}, + {"keys",keysCommand,2,0,NULL,0,0,0,0,0}, + {"dbsize",dbsizeCommand,1,0,NULL,0,0,0,0,0}, + {"auth",authCommand,2,0,NULL,0,0,0,0,0}, + {"ping",pingCommand,1,0,NULL,0,0,0,0,0}, + {"echo",echoCommand,2,0,NULL,0,0,0,0,0}, + {"save",saveCommand,1,0,NULL,0,0,0,0,0}, + {"bgsave",bgsaveCommand,1,0,NULL,0,0,0,0,0}, + {"bgrewriteaof",bgrewriteaofCommand,1,0,NULL,0,0,0,0,0}, + {"shutdown",shutdownCommand,1,0,NULL,0,0,0,0,0}, + {"lastsave",lastsaveCommand,1,0,NULL,0,0,0,0,0}, + {"type",typeCommand,2,0,NULL,1,1,1,0,0}, + {"multi",multiCommand,1,0,NULL,0,0,0,0,0}, + {"exec",execCommand,1,REDIS_CMD_DENYOOM,NULL,0,0,0,0,0}, + {"discard",discardCommand,1,0,NULL,0,0,0,0,0}, + {"sync",syncCommand,1,0,NULL,0,0,0,0,0}, + {"flushdb",flushdbCommand,1,0,NULL,0,0,0,0,0}, + {"flushall",flushallCommand,1,0,NULL,0,0,0,0,0}, + {"sort",sortCommand,-2,REDIS_CMD_DENYOOM,NULL,1,1,1,0,0}, + {"info",infoCommand,-1,0,NULL,0,0,0,0,0}, + {"monitor",monitorCommand,1,0,NULL,0,0,0,0,0}, + {"ttl",ttlCommand,2,0,NULL,1,1,1,0,0}, + {"persist",persistCommand,2,0,NULL,1,1,1,0,0}, + {"slaveof",slaveofCommand,3,0,NULL,0,0,0,0,0}, + {"debug",debugCommand,-2,0,NULL,0,0,0,0,0}, + {"config",configCommand,-2,0,NULL,0,0,0,0,0}, + {"subscribe",subscribeCommand,-2,0,NULL,0,0,0,0,0}, + {"unsubscribe",unsubscribeCommand,-1,0,NULL,0,0,0,0,0}, + {"psubscribe",psubscribeCommand,-2,0,NULL,0,0,0,0,0}, + {"punsubscribe",punsubscribeCommand,-1,0,NULL,0,0,0,0,0}, + {"publish",publishCommand,3,REDIS_CMD_FORCE_REPLICATION,NULL,0,0,0,0,0}, + {"watch",watchCommand,-2,0,noPreloadGetKeys,1,-1,1,0,0}, + {"unwatch",unwatchCommand,1,0,NULL,0,0,0,0,0}, + {"cluster",clusterCommand,-2,0,NULL,0,0,0,0,0}, + {"restore",restoreCommand,4,0,NULL,0,0,0,0,0}, + {"migrate",migrateCommand,6,0,NULL,0,0,0,0,0}, + {"dump",dumpCommand,2,0,NULL,0,0,0,0,0}, + {"object",objectCommand,-2,0,NULL,0,0,0,0,0}, + {"client",clientCommand,-2,0,NULL,0,0,0,0,0} }; /*============================ Utility functions ============================ */ -void redisLog(int level, const char *fmt, ...) { - va_list ap; +/* Low level logging. To use only for very big messages, otherwise + * redisLog() is to prefer. */ +void redisLogRaw(int level, const char *msg) { + const int syslogLevelMap[] = { LOG_DEBUG, LOG_INFO, LOG_NOTICE, LOG_WARNING }; + const char *c = ".-*#"; + time_t now = time(NULL); FILE *fp; - char *c = ".-*#"; char buf[64]; - time_t now; + int rawmode = (level & REDIS_LOG_RAW); + level &= 0xff; /* clear flags */ if (level < server.verbosity) return; fp = (server.logfile == NULL) ? stdout : fopen(server.logfile,"a"); if (!fp) return; - va_start(ap, fmt); - now = time(NULL); - strftime(buf,64,"%d %b %H:%M:%S",localtime(&now)); - fprintf(fp,"[%d] %s %c ",(int)getpid(),buf,c[level]); - vfprintf(fp, fmt, ap); - fprintf(fp,"\n"); + if (rawmode) { + fprintf(fp,"%s",msg); + } else { + strftime(buf,sizeof(buf),"%d %b %H:%M:%S",localtime(&now)); + fprintf(fp,"[%d] %s %c %s\n",(int)getpid(),buf,c[level],msg); + } fflush(fp); - va_end(ap); if (server.logfile) fclose(fp); + + if (server.syslog_enabled) syslog(syslogLevelMap[level], "%s", msg); +} + +/* Like redisLogRaw() but with printf-alike support. This is the funciton that + * is used across the code. The raw version is only used in order to dump + * the INFO output on crash. */ +void redisLog(int level, const char *fmt, ...) { + va_list ap; + char msg[REDIS_MAX_LOGMSG_LEN]; + + if ((level&0xff) < server.verbosity) return; + + va_start(ap, fmt); + vsnprintf(msg, sizeof(msg), fmt, ap); + va_end(ap); + + redisLogRaw(level,msg); } /* Redis generally does not try to recover from out of memory conditions @@ -222,6 +254,17 @@ void oom(const char *msg) { abort(); } +/* Return the UNIX time in microseconds */ +long long ustime(void) { + struct timeval tv; + long long ust; + + gettimeofday(&tv, NULL); + ust = ((long long)tv.tv_sec)*1000000; + ust += tv.tv_usec; + return ust; +} + /*====================== Hash table type implementation ==================== */ /* This is an hash table type that uses the SDS dynamic strings libary as @@ -252,6 +295,15 @@ int dictSdsKeyCompare(void *privdata, const void *key1, return memcmp(key1, key2, l1) == 0; } +/* A case insensitive version used for the command lookup table. */ +int dictSdsKeyCaseCompare(void *privdata, const void *key1, + const void *key2) +{ + DICT_NOTUSED(privdata); + + return strcasecmp(key1, key2) == 0; +} + void dictRedisObjectDestructor(void *privdata, void *val) { DICT_NOTUSED(privdata); @@ -283,6 +335,10 @@ unsigned int dictSdsHash(const void *key) { return dictGenHashFunction((unsigned char*)key, sdslen((char*)key)); } +unsigned int dictSdsCaseHash(const void *key) { + return dictGenCaseHashFunction((unsigned char*)key, sdslen((char*)key)); +} + int dictEncObjKeyCompare(void *privdata, const void *key1, const void *key2) { @@ -324,7 +380,7 @@ unsigned int dictEncObjHash(const void *key) { } } -/* Sets type */ +/* Sets type and diskstore negative caching hash table */ dictType setDictType = { dictEncObjHash, /* hash function */ NULL, /* key dup */ @@ -364,6 +420,16 @@ dictType keyptrDictType = { NULL /* val destructor */ }; +/* Command table. sds string -> command struct pointer. */ +dictType commandTableDictType = { + dictSdsCaseHash, /* hash function */ + NULL, /* key dup */ + NULL, /* val dup */ + dictSdsKeyCaseCompare, /* key compare */ + dictSdsDestructor, /* key destructor */ + NULL /* val destructor */ +}; + /* Hash type hash table (note that small hashes are represented with zimpaps) */ dictType hashDictType = { dictEncObjHash, /* hash function */ @@ -386,6 +452,17 @@ dictType keylistDictType = { dictListDestructor /* val destructor */ }; +/* Cluster nodes hash table, mapping nodes addresses 1.2.3.4:6379 to + * clusterNode structures. */ +dictType clusterNodesDictType = { + dictSdsHash, /* hash function */ + NULL, /* key dup */ + NULL, /* val dup */ + dictSdsKeyCompare, /* key compare */ + dictSdsDestructor, /* key destructor */ + NULL /* val destructor */ +}; + int htNeedsResize(dict *dict) { long long size, used; @@ -485,7 +562,7 @@ void updateLRUClock(void) { } int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { - int j, loops = server.cronloops++; + int j, loops = server.cronloops; REDIS_NOTUSED(eventLoop); REDIS_NOTUSED(id); REDIS_NOTUSED(clientData); @@ -509,6 +586,10 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { */ updateLRUClock(); + /* Record the max memory used since the server was started. */ + if (zmalloc_used_memory() > server.stat_peak_memory) + server.stat_peak_memory = zmalloc_used_memory(); + /* We received a SIGTERM, shutting down here in a safe way, as it is * not ok doing so inside the signal handler. */ if (server.shutdown_asap) { @@ -549,23 +630,43 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { } /* Close connections of timedout clients */ - if ((server.maxidletime && !(loops % 100)) || server.blpop_blocked_clients) + if ((server.maxidletime && !(loops % 100)) || server.bpop_blocked_clients) closeTimedoutClients(); - /* Check if a background saving or AOF rewrite in progress terminated */ + /* Check if a background saving or AOF rewrite in progress terminated. */ if (server.bgsavechildpid != -1 || server.bgrewritechildpid != -1) { int statloc; pid_t pid; if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) { + int exitcode = WEXITSTATUS(statloc); + int bysignal = 0; + + if (WIFSIGNALED(statloc)) bysignal = WTERMSIG(statloc); + if (pid == server.bgsavechildpid) { - backgroundSaveDoneHandler(statloc); + backgroundSaveDoneHandler(exitcode,bysignal); } else { - backgroundRewriteDoneHandler(statloc); + backgroundRewriteDoneHandler(exitcode,bysignal); } updateDictResizePolicy(); } - } else { + } else if (server.bgsavethread != (pthread_t) -1) { + if (server.bgsavethread != (pthread_t) -1) { + int state; + + pthread_mutex_lock(&server.bgsavethread_mutex); + state = server.bgsavethread_state; + pthread_mutex_unlock(&server.bgsavethread_mutex); + + if (state == REDIS_BGSAVE_THREAD_DONE_OK || + state == REDIS_BGSAVE_THREAD_DONE_ERR) + { + backgroundSaveDoneHandler( + (state == REDIS_BGSAVE_THREAD_DONE_OK) ? 0 : 1, 0); + } + } + } else if (!server.ds_enabled) { /* If there is not a background saving in progress check if * we have to save now */ time_t now = time(NULL); @@ -587,43 +688,18 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { * in order to guarantee a strict consistency. */ if (server.masterhost == NULL) activeExpireCycle(); - /* Swap a few keys on disk if we are over the memory limit and VM - * is enbled. Try to free objects from the free list first. */ - if (vmCanSwapOut()) { - while (server.vm_enabled && zmalloc_used_memory() > - server.vm_max_memory) - { - int retval; - - if (tryFreeOneObjectFromFreelist() == REDIS_OK) continue; - retval = (server.vm_max_threads == 0) ? - vmSwapOneObjectBlocking() : - vmSwapOneObjectThreaded(); - if (retval == REDIS_ERR && !(loops % 300) && - zmalloc_used_memory() > - (server.vm_max_memory+server.vm_max_memory/10)) - { - redisLog(REDIS_WARNING,"WARNING: vm-max-memory limit exceeded by more than 10%% but unable to swap more objects out!"); - } - /* Note that when using threade I/O we free just one object, - * because anyway when the I/O thread in charge to swap this - * object out will finish, the handler of completed jobs - * will try to swap more objects if we are still out of memory. */ - if (retval == REDIS_ERR || server.vm_max_threads > 0) break; - } - } + /* Remove a few cached objects from memory if we are over the + * configured memory limit */ + if (server.ds_enabled) cacheCron(); - /* Check if we should connect to a MASTER */ - if (server.replstate == REDIS_REPL_CONNECT && !(loops % 10)) { - redisLog(REDIS_NOTICE,"Connecting to MASTER..."); - if (syncWithMaster() == REDIS_OK) { - redisLog(REDIS_NOTICE,"MASTER <-> SLAVE sync succeeded"); - if (server.appendonly) rewriteAppendOnlyFileBackground(); - } - } + /* Replication cron function -- used to reconnect to master and + * to detect transfer failures. */ + if (!(loops % 10)) replicationCron(); - /* Update fragmentation info, used for fast RSS estimation */ - if (!(loops % 10)) server.fragmentation = zmalloc_get_fragmentation_ratio(); + /* Run other sub-systems specific cron jobs */ + if (server.cluster_enabled && !(loops % 10)) clusterCron(); + + server.cronloops++; return 100; } @@ -632,21 +708,22 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { * for ready file descriptors. */ void beforeSleep(struct aeEventLoop *eventLoop) { REDIS_NOTUSED(eventLoop); + listNode *ln; + redisClient *c; - /* Awake clients that got all the swapped keys they requested */ - if (server.vm_enabled && listLength(server.io_ready_clients)) { + /* Awake clients that got all the on disk keys they requested */ + if (server.ds_enabled && listLength(server.io_ready_clients)) { listIter li; - listNode *ln; listRewind(server.io_ready_clients,&li); while((ln = listNext(&li))) { - redisClient *c = ln->value; + c = ln->value; struct redisCommand *cmd; /* Resume the client. */ listDelNode(server.io_ready_clients,ln); c->flags &= (~REDIS_IO_WAIT); - server.vm_blocked_clients--; + server.cache_blocked_clients--; aeCreateFileEvent(server.el, c->fd, AE_READABLE, readQueryFromClient, c); cmd = lookupCommand(c->argv[0]->ptr); @@ -658,6 +735,20 @@ void beforeSleep(struct aeEventLoop *eventLoop) { processInputBuffer(c); } } + + /* Try to process pending commands for clients that were just unblocked. */ + while (listLength(server.unblocked_clients)) { + ln = listFirst(server.unblocked_clients); + redisAssert(ln != NULL); + c = ln->value; + listDelNode(server.unblocked_clients,ln); + c->flags &= ~REDIS_UNBLOCKED; + + /* Process remaining data in the input buffer. */ + if (c->querybuf && sdslen(c->querybuf) > 0) + processInputBuffer(c); + } + /* Write the AOF buffer on disk */ flushAppendOnlyFile(); } @@ -689,6 +780,8 @@ void createSharedObjects(void) { "-ERR source and destination objects are the same\r\n")); shared.outofrangeerr = createObject(REDIS_STRING,sdsnew( "-ERR index out of range\r\n")); + shared.loadingerr = createObject(REDIS_STRING,sdsnew( + "-LOADING Redis is loading the dataset in memory\r\n")); shared.space = createObject(REDIS_STRING,sdsnew(" ")); shared.colon = createObject(REDIS_STRING,sdsnew(":")); shared.plus = createObject(REDIS_STRING,sdsnew("+")); @@ -717,14 +810,20 @@ void createSharedObjects(void) { } void initServerConfig() { - server.dbnum = REDIS_DEFAULT_DBNUM; server.port = REDIS_SERVERPORT; + server.bindaddr = NULL; + server.unixsocket = NULL; + server.ipfd = -1; + server.sofd = -1; + server.dbnum = REDIS_DEFAULT_DBNUM; server.verbosity = REDIS_VERBOSE; server.maxidletime = REDIS_MAXIDLETIME; server.saveparams = NULL; + server.loading = 0; server.logfile = NULL; /* NULL = log on standard output */ - server.bindaddr = NULL; - server.glueoutputbuf = 1; + server.syslog_enabled = 0; + server.syslog_ident = zstrdup("redis"); + server.syslog_facility = LOG_LOCAL0; server.daemonize = 0; server.appendonly = 0; server.appendfsync = APPENDFSYNC_EVERYSEC; @@ -738,25 +837,26 @@ void initServerConfig() { server.requirepass = NULL; server.rdbcompression = 1; server.activerehashing = 1; - server.fragmentation = 1; server.maxclients = 0; - server.blpop_blocked_clients = 0; + server.bpop_blocked_clients = 0; server.maxmemory = 0; server.maxmemory_policy = REDIS_MAXMEMORY_VOLATILE_LRU; server.maxmemory_samples = 3; - server.vm_enabled = 0; - server.vm_swap_file = zstrdup("/tmp/redis-%p.vm"); - server.vm_page_size = 256; /* 256 bytes per page */ - server.vm_pages = 1024*1024*100; /* 104 millions of pages */ - server.vm_max_memory = 1024LL*1024*1024*1; /* 1 GB of RAM */ - server.vm_max_threads = 4; - server.vm_blocked_clients = 0; + server.ds_enabled = 0; + server.ds_path = sdsnew("/tmp/redis.ds"); + server.cache_max_memory = 64LL*1024*1024; /* 64 MB of RAM */ + server.cache_blocked_clients = 0; server.hash_max_zipmap_entries = REDIS_HASH_MAX_ZIPMAP_ENTRIES; server.hash_max_zipmap_value = REDIS_HASH_MAX_ZIPMAP_VALUE; server.list_max_ziplist_entries = REDIS_LIST_MAX_ZIPLIST_ENTRIES; server.list_max_ziplist_value = REDIS_LIST_MAX_ZIPLIST_VALUE; server.set_max_intset_entries = REDIS_SET_MAX_INTSET_ENTRIES; + server.zset_max_ziplist_entries = REDIS_ZSET_MAX_ZIPLIST_ENTRIES; + server.zset_max_ziplist_value = REDIS_ZSET_MAX_ZIPLIST_VALUE; server.shutdown_asap = 0; + server.cache_flush_delay = 0; + server.cluster_enabled = 0; + server.cluster.configfile = zstrdup("nodes.conf"); updateLRUClock(); resetServerSaveParams(); @@ -771,12 +871,21 @@ void initServerConfig() { server.masterport = 6379; server.master = NULL; server.replstate = REDIS_REPL_NONE; + server.repl_serve_stale_data = 1; /* Double constants initialization */ R_Zero = 0.0; R_PosInf = 1.0/R_Zero; R_NegInf = -1.0/R_Zero; R_Nan = R_Zero/R_Zero; + + /* Command table -- we intiialize it here as it is part of the + * initial configuration, since command names may be changed via + * redis.conf using the rename-command directive. */ + server.commands = dictCreate(&commandTableDictType,NULL); + populateCommandTable(); + server.delCommand = lookupCommandByCString("del"); + server.multiCommand = lookupCommandByCString("multi"); } void initServer() { @@ -784,24 +893,41 @@ void initServer() { signal(SIGHUP, SIG_IGN); signal(SIGPIPE, SIG_IGN); - setupSigSegvAction(); + setupSignalHandlers(); - server.mainthread = pthread_self(); - server.devnull = fopen("/dev/null","w"); - if (server.devnull == NULL) { - redisLog(REDIS_WARNING, "Can't open /dev/null: %s", server.neterr); - exit(1); + if (server.syslog_enabled) { + openlog(server.syslog_ident, LOG_PID | LOG_NDELAY | LOG_NOWAIT, + server.syslog_facility); } + + server.mainthread = pthread_self(); server.clients = listCreate(); server.slaves = listCreate(); server.monitors = listCreate(); - server.objfreelist = listCreate(); + server.unblocked_clients = listCreate(); + server.cache_io_queue = listCreate(); + createSharedObjects(); server.el = aeCreateEventLoop(); server.db = zmalloc(sizeof(redisDb)*server.dbnum); - server.fd = anetTcpServer(server.neterr, server.port, server.bindaddr); - if (server.fd == -1) { - redisLog(REDIS_WARNING, "Opening TCP port: %s", server.neterr); + + if (server.port != 0) { + server.ipfd = anetTcpServer(server.neterr,server.port,server.bindaddr); + if (server.ipfd == ANET_ERR) { + redisLog(REDIS_WARNING, "Opening port: %s", server.neterr); + exit(1); + } + } + if (server.unixsocket != NULL) { + unlink(server.unixsocket); /* don't care if this fails */ + server.sofd = anetUnixServer(server.neterr,server.unixsocket); + if (server.sofd == ANET_ERR) { + redisLog(REDIS_WARNING, "Opening socket: %s", server.neterr); + exit(1); + } + } + if (server.ipfd < 0 && server.sofd < 0) { + redisLog(REDIS_WARNING, "Configured to not listen anywhere, exiting."); exit(1); } for (j = 0; j < server.dbnum; j++) { @@ -809,8 +935,11 @@ void initServer() { server.db[j].expires = dictCreate(&keyptrDictType,NULL); server.db[j].blocking_keys = dictCreate(&keylistDictType,NULL); server.db[j].watched_keys = dictCreate(&keylistDictType,NULL); - if (server.vm_enabled) + if (server.ds_enabled) { server.db[j].io_keys = dictCreate(&keylistDictType,NULL); + server.db[j].io_negcache = dictCreate(&setDictType,NULL); + server.db[j].io_queued = dictCreate(&setDictType,NULL); + } server.db[j].id = j; } server.pubsub_channels = dictCreate(&keylistDictType,NULL); @@ -820,6 +949,8 @@ void initServer() { server.cronloops = 0; server.bgsavechildpid = -1; server.bgrewritechildpid = -1; + server.bgsavethread_state = REDIS_BGSAVE_THREAD_UNACTIVE; + server.bgsavethread = (pthread_t) -1; server.bgrewritebuf = sdsempty(); server.aofbuf = sdsempty(); server.lastsave = time(NULL); @@ -827,13 +958,17 @@ void initServer() { server.stat_numcommands = 0; server.stat_numconnections = 0; server.stat_expiredkeys = 0; + server.stat_evictedkeys = 0; server.stat_starttime = time(NULL); server.stat_keyspace_misses = 0; server.stat_keyspace_hits = 0; + server.stat_peak_memory = 0; server.unixtime = time(NULL); aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL); - if (aeCreateFileEvent(server.el, server.fd, AE_READABLE, - acceptHandler, NULL) == AE_ERR) oom("creating file event"); + if (server.ipfd > 0 && aeCreateFileEvent(server.el,server.ipfd,AE_READABLE, + acceptTcpHandler,NULL) == AE_ERR) oom("creating file event"); + if (server.sofd > 0 && aeCreateFileEvent(server.el,server.sofd,AE_READABLE, + acceptUnixHandler,NULL) == AE_ERR) oom("creating file event"); if (server.appendonly) { server.appendfd = open(server.appendfilename,O_WRONLY|O_APPEND|O_CREAT,0644); @@ -844,43 +979,62 @@ void initServer() { } } - if (server.vm_enabled) vmInit(); + if (server.ds_enabled) dsInit(); + if (server.cluster_enabled) clusterInit(); + srand(time(NULL)^getpid()); } -int qsortRedisCommands(const void *r1, const void *r2) { - return strcasecmp( - ((struct redisCommand*)r1)->name, - ((struct redisCommand*)r2)->name); +/* Populates the Redis Command Table starting from the hard coded list + * we have on top of redis.c file. */ +void populateCommandTable(void) { + int j; + int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand); + + for (j = 0; j < numcommands; j++) { + struct redisCommand *c = redisCommandTable+j; + int retval; + + retval = dictAdd(server.commands, sdsnew(c->name), c); + assert(retval == DICT_OK); + } } -void sortCommandTable() { - /* Copy and sort the read-only version of the command table */ - commandTable = (struct redisCommand*)zmalloc(sizeof(readonlyCommandTable)); - memcpy(commandTable,readonlyCommandTable,sizeof(readonlyCommandTable)); - qsort(commandTable, - sizeof(readonlyCommandTable)/sizeof(struct redisCommand), - sizeof(struct redisCommand),qsortRedisCommands); +void resetCommandTableStats(void) { + int numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand); + int j; + + for (j = 0; j < numcommands; j++) { + struct redisCommand *c = redisCommandTable+j; + + c->microseconds = 0; + c->calls = 0; + } } /* ====================== Commands lookup and execution ===================== */ -struct redisCommand *lookupCommand(char *name) { - struct redisCommand tmp = {name,NULL,0,0,NULL,0,0,0}; - return bsearch( - &tmp, - commandTable, - sizeof(readonlyCommandTable)/sizeof(struct redisCommand), - sizeof(struct redisCommand), - qsortRedisCommands); +struct redisCommand *lookupCommand(sds name) { + return dictFetchValue(server.commands, name); +} + +struct redisCommand *lookupCommandByCString(char *s) { + struct redisCommand *cmd; + sds name = sdsnew(s); + + cmd = dictFetchValue(server.commands, name); + sdsfree(name); + return cmd; } /* Call() is the core of Redis execution of a command */ void call(redisClient *c, struct redisCommand *cmd) { - long long dirty; + long long dirty, start = ustime(); dirty = server.dirty; cmd->proc(c); dirty = server.dirty-dirty; + cmd->microseconds += ustime()-start; + cmd->calls++; if (server.appendonly && dirty) feedAppendOnlyFile(cmd,c->db->id,c->argv,c->argc); @@ -933,6 +1087,29 @@ int processCommand(redisClient *c) { return REDIS_OK; } + /* If cluster is enabled, redirect here */ + if (server.cluster_enabled && + !(cmd->getkeys_proc == NULL && cmd->firstkey == 0)) { + int hashslot; + + if (server.cluster.state != REDIS_CLUSTER_OK) { + addReplyError(c,"The cluster is down. Check with CLUSTER INFO for more information"); + return REDIS_OK; + } else { + int ask; + clusterNode *n = getNodeByQuery(c,cmd,c->argv,c->argc,&hashslot,&ask); + if (n == NULL) { + addReplyError(c,"Multi keys request invalid in cluster"); + return REDIS_OK; + } else if (n != server.cluster.myself) { + addReplySds(c,sdscatprintf(sdsempty(), + "-%s %d %s:%d\r\n", ask ? "ASK" : "MOVED", + hashslot,n->ip,n->port)); + return REDIS_OK; + } + } + } + /* Handle the maxmemory directive. * * First we try to free some memory if possible (if there are volatile @@ -955,6 +1132,23 @@ int processCommand(redisClient *c) { return REDIS_OK; } + /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and + * we are a slave with a broken link with master. */ + if (server.masterhost && server.replstate != REDIS_REPL_CONNECTED && + server.repl_serve_stale_data == 0 && + cmd->proc != infoCommand && cmd->proc != slaveofCommand) + { + addReplyError(c, + "link with MASTER is down and slave-serve-stale-data is set to no"); + return REDIS_OK; + } + + /* Loading DB? Return an error if the command is not INFO */ + if (server.loading && cmd->proc != infoCommand) { + addReply(c, shared.loadingerr); + return REDIS_OK; + } + /* Exec the command */ if (c->flags & REDIS_MULTI && cmd->proc != execCommand && cmd->proc != discardCommand && @@ -963,8 +1157,8 @@ int processCommand(redisClient *c) { queueMultiCommand(c,cmd); addReply(c,shared.queued); } else { - if (server.vm_enabled && server.vm_max_threads > 0 && - blockClientOnSwappedKeys(c,cmd)) return REDIS_ERR; + if (server.ds_enabled && blockClientOnSwappedKeys(c,cmd)) + return REDIS_ERR; call(c,cmd); } return REDIS_OK; @@ -982,10 +1176,11 @@ int prepareForShutdown() { kill(server.bgsavechildpid,SIGKILL); rdbRemoveTempFile(server.bgsavechildpid); } - if (server.appendonly) { + if (server.ds_enabled) { + /* FIXME: flush all objects on disk */ + } else if (server.appendonly) { /* Append only file: fsync() the AOF and exit */ aof_fsync(server.appendfd); - if (server.vm_enabled) unlink(server.vm_swap_file); } else if (server.saveparamslen > 0) { /* Snapshotting. Perform a SYNC SAVE and exit */ if (rdbSave(server.dbfilename) != REDIS_OK) { @@ -1049,155 +1244,314 @@ void bytesToHuman(char *s, unsigned long long n) { /* Create the string returned by the INFO command. This is decoupled * by the INFO command itself as we need to report the same information * on memory corruption problems. */ -sds genRedisInfoString(void) { - sds info; +sds genRedisInfoString(char *section) { + sds info = sdsempty(); time_t uptime = time(NULL)-server.stat_starttime; - int j; - char hmem[64]; + int j, numcommands; struct rusage self_ru, c_ru; + unsigned long lol, bib; + int allsections = 0, defsections = 0; + int sections = 0; + + if (section) { + allsections = strcasecmp(section,"all") == 0; + defsections = strcasecmp(section,"default") == 0; + } getrusage(RUSAGE_SELF, &self_ru); getrusage(RUSAGE_CHILDREN, &c_ru); + getClientsMaxBuffers(&lol,&bib); - bytesToHuman(hmem,zmalloc_used_memory()); - info = sdscatprintf(sdsempty(), - "redis_version:%s\r\n" - "redis_git_sha1:%s\r\n" - "redis_git_dirty:%d\r\n" - "arch_bits:%s\r\n" - "multiplexing_api:%s\r\n" - "process_id:%ld\r\n" - "uptime_in_seconds:%ld\r\n" - "uptime_in_days:%ld\r\n" - "lru_clock:%ld\r\n" + /* Server */ + if (allsections || defsections || !strcasecmp(section,"server")) { + if (sections++) info = sdscat(info,"\r\n"); + info = sdscatprintf(info, + "# Server\r\n" + "redis_version:%s\r\n" + "redis_git_sha1:%s\r\n" + "redis_git_dirty:%d\r\n" + "arch_bits:%s\r\n" + "multiplexing_api:%s\r\n" + "process_id:%ld\r\n" + "tcp_port:%d\r\n" + "uptime_in_seconds:%ld\r\n" + "uptime_in_days:%ld\r\n" + "lru_clock:%ld\r\n", + REDIS_VERSION, + redisGitSHA1(), + strtol(redisGitDirty(),NULL,10) > 0, + (sizeof(long) == 8) ? "64" : "32", + aeGetApiName(), + (long) getpid(), + server.port, + uptime, + uptime/(3600*24), + (unsigned long) server.lruclock); + } + + /* Clients */ + if (allsections || defsections || !strcasecmp(section,"clients")) { + if (sections++) info = sdscat(info,"\r\n"); + info = sdscatprintf(info, + "# Clients\r\n" + "connected_clients:%d\r\n" + "client_longest_output_list:%lu\r\n" + "client_biggest_input_buf:%lu\r\n" + "blocked_clients:%d\r\n", + listLength(server.clients)-listLength(server.slaves), + lol, bib, + server.bpop_blocked_clients); + } + + /* Memory */ + if (allsections || defsections || !strcasecmp(section,"memory")) { + char hmem[64]; + char peak_hmem[64]; + + bytesToHuman(hmem,zmalloc_used_memory()); + bytesToHuman(peak_hmem,server.stat_peak_memory); + if (sections++) info = sdscat(info,"\r\n"); + info = sdscatprintf(info, + "# Memory\r\n" + "used_memory:%zu\r\n" + "used_memory_human:%s\r\n" + "used_memory_rss:%zu\r\n" + "used_memory_peak:%zu\r\n" + "used_memory_peak_human:%s\r\n" + "mem_fragmentation_ratio:%.2f\r\n" + "mem_allocator:%s\r\n", + zmalloc_used_memory(), + hmem, + zmalloc_get_rss(), + server.stat_peak_memory, + peak_hmem, + zmalloc_get_fragmentation_ratio(), + REDIS_MALLOC + ); + } + + /* Allocation statistics */ + if (allsections || !strcasecmp(section,"allocstats")) { + if (sections++) info = sdscat(info,"\r\n"); + info = sdscat(info, "# Allocstats\r\nallocation_stats:"); + for (j = 0; j <= ZMALLOC_MAX_ALLOC_STAT; j++) { + size_t count = zmalloc_allocations_for_size(j); + if (count) { + if (info[sdslen(info)-1] != ':') info = sdscatlen(info,",",1); + info = sdscatprintf(info,"%s%d=%zu", + (j == ZMALLOC_MAX_ALLOC_STAT) ? ">=" : "", + j,count); + } + } + info = sdscat(info,"\r\n"); + } + + /* Persistence */ + if (allsections || defsections || !strcasecmp(section,"persistence")) { + if (sections++) info = sdscat(info,"\r\n"); + info = sdscatprintf(info, + "# Persistence\r\n" + "loading:%d\r\n" + "aof_enabled:%d\r\n" + "changes_since_last_save:%lld\r\n" + "bgsave_in_progress:%d\r\n" + "last_save_time:%ld\r\n" + "bgrewriteaof_in_progress:%d\r\n", + server.loading, + server.appendonly, + server.dirty, + server.bgsavechildpid != -1 || + server.bgsavethread != (pthread_t) -1, + server.lastsave, + server.bgrewritechildpid != -1); + + if (server.loading) { + double perc; + time_t eta, elapsed; + off_t remaining_bytes = server.loading_total_bytes- + server.loading_loaded_bytes; + + perc = ((double)server.loading_loaded_bytes / + server.loading_total_bytes) * 100; + + elapsed = time(NULL)-server.loading_start_time; + if (elapsed == 0) { + eta = 1; /* A fake 1 second figure if we don't have + enough info */ + } else { + eta = (elapsed*remaining_bytes)/server.loading_loaded_bytes; + } + + info = sdscatprintf(info, + "loading_start_time:%ld\r\n" + "loading_total_bytes:%llu\r\n" + "loading_loaded_bytes:%llu\r\n" + "loading_loaded_perc:%.2f\r\n" + "loading_eta_seconds:%ld\r\n" + ,(unsigned long) server.loading_start_time, + (unsigned long long) server.loading_total_bytes, + (unsigned long long) server.loading_loaded_bytes, + perc, + eta + ); + } + } + + /* Diskstore */ + if (allsections || defsections || !strcasecmp(section,"diskstore")) { + if (sections++) info = sdscat(info,"\r\n"); + info = sdscatprintf(info, + "# Diskstore\r\n" + "ds_enabled:%d\r\n", + server.ds_enabled != 0); + if (server.ds_enabled) { + lockThreadedIO(); + info = sdscatprintf(info, + "cache_max_memory:%llu\r\n" + "cache_blocked_clients:%lu\r\n" + "cache_io_queue_len:%lu\r\n" + "cache_io_jobs_new:%lu\r\n" + "cache_io_jobs_processing:%lu\r\n" + "cache_io_jobs_processed:%lu\r\n" + "cache_io_ready_clients:%lu\r\n" + ,(unsigned long long) server.cache_max_memory, + (unsigned long) server.cache_blocked_clients, + (unsigned long) listLength(server.cache_io_queue), + (unsigned long) listLength(server.io_newjobs), + (unsigned long) listLength(server.io_processing), + (unsigned long) listLength(server.io_processed), + (unsigned long) listLength(server.io_ready_clients) + ); + unlockThreadedIO(); + } + } + + /* Stats */ + if (allsections || defsections || !strcasecmp(section,"stats")) { + if (sections++) info = sdscat(info,"\r\n"); + info = sdscatprintf(info, + "# Stats\r\n" + "total_connections_received:%lld\r\n" + "total_commands_processed:%lld\r\n" + "expired_keys:%lld\r\n" + "evicted_keys:%lld\r\n" + "keyspace_hits:%lld\r\n" + "keyspace_misses:%lld\r\n" + "pubsub_channels:%ld\r\n" + "pubsub_patterns:%u\r\n", + server.stat_numconnections, + server.stat_numcommands, + server.stat_expiredkeys, + server.stat_evictedkeys, + server.stat_keyspace_hits, + server.stat_keyspace_misses, + dictSize(server.pubsub_channels), + listLength(server.pubsub_patterns)); + } + + /* Replication */ + if (allsections || defsections || !strcasecmp(section,"replication")) { + if (sections++) info = sdscat(info,"\r\n"); + info = sdscatprintf(info, + "# Replication\r\n" + "role:%s\r\n", + server.masterhost == NULL ? "master" : "slave"); + if (server.masterhost) { + info = sdscatprintf(info, + "master_host:%s\r\n" + "master_port:%d\r\n" + "master_link_status:%s\r\n" + "master_last_io_seconds_ago:%d\r\n" + "master_sync_in_progress:%d\r\n" + ,server.masterhost, + server.masterport, + (server.replstate == REDIS_REPL_CONNECTED) ? + "up" : "down", + server.master ? + ((int)(time(NULL)-server.master->lastinteraction)) : -1, + server.replstate == REDIS_REPL_TRANSFER + ); + + if (server.replstate == REDIS_REPL_TRANSFER) { + info = sdscatprintf(info, + "master_sync_left_bytes:%ld\r\n" + "master_sync_last_io_seconds_ago:%d\r\n" + ,(long)server.repl_transfer_left, + (int)(time(NULL)-server.repl_transfer_lastio) + ); + } + } + info = sdscatprintf(info, + "connected_slaves:%d\r\n", + listLength(server.slaves)); + } + + /* CPU */ + if (allsections || defsections || !strcasecmp(section,"cpu")) { + if (sections++) info = sdscat(info,"\r\n"); + info = sdscatprintf(info, + "# CPU\r\n" "used_cpu_sys:%.2f\r\n" "used_cpu_user:%.2f\r\n" "used_cpu_sys_childrens:%.2f\r\n" - "used_cpu_user_childrens:%.2f\r\n" - "connected_clients:%d\r\n" - "connected_slaves:%d\r\n" - "blocked_clients:%d\r\n" - "used_memory:%zu\r\n" - "used_memory_human:%s\r\n" - "used_memory_rss:%zu\r\n" - "used_memory_estimated_rss:%zu\r\n" - "mem_fragmentation_ratio:%.2f\r\n" - "use_tcmalloc:%d\r\n" - "changes_since_last_save:%lld\r\n" - "bgsave_in_progress:%d\r\n" - "last_save_time:%ld\r\n" - "bgrewriteaof_in_progress:%d\r\n" - "total_connections_received:%lld\r\n" - "total_commands_processed:%lld\r\n" - "expired_keys:%lld\r\n" - "keyspace_hits:%lld\r\n" - "keyspace_misses:%lld\r\n" - "hash_max_zipmap_entries:%zu\r\n" - "hash_max_zipmap_value:%zu\r\n" - "pubsub_channels:%ld\r\n" - "pubsub_patterns:%u\r\n" - "vm_enabled:%d\r\n" - "role:%s\r\n" - ,REDIS_VERSION, - redisGitSHA1(), - strtol(redisGitDirty(),NULL,10) > 0, - (sizeof(long) == 8) ? "64" : "32", - aeGetApiName(), - (long) getpid(), - uptime, - uptime/(3600*24), - (unsigned long) server.lruclock, + "used_cpu_user_childrens:%.2f\r\n", (float)self_ru.ru_utime.tv_sec+(float)self_ru.ru_utime.tv_usec/1000000, (float)self_ru.ru_stime.tv_sec+(float)self_ru.ru_stime.tv_usec/1000000, (float)c_ru.ru_utime.tv_sec+(float)c_ru.ru_utime.tv_usec/1000000, - (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000, - listLength(server.clients)-listLength(server.slaves), - listLength(server.slaves), - server.blpop_blocked_clients, - zmalloc_used_memory(), - hmem, - zmalloc_get_rss(), - redisEstimateRSS(), - zmalloc_get_fragmentation_ratio(), -#ifdef USE_TCMALLOC - 1, -#else - 0, -#endif - server.dirty, - server.bgsavechildpid != -1, - server.lastsave, - server.bgrewritechildpid != -1, - server.stat_numconnections, - server.stat_numcommands, - server.stat_expiredkeys, - server.stat_keyspace_hits, - server.stat_keyspace_misses, - server.hash_max_zipmap_entries, - server.hash_max_zipmap_value, - dictSize(server.pubsub_channels), - listLength(server.pubsub_patterns), - server.vm_enabled != 0, - server.masterhost == NULL ? "master" : "slave" - ); - if (server.masterhost) { - info = sdscatprintf(info, - "master_host:%s\r\n" - "master_port:%d\r\n" - "master_link_status:%s\r\n" - "master_last_io_seconds_ago:%d\r\n" - ,server.masterhost, - server.masterport, - (server.replstate == REDIS_REPL_CONNECTED) ? - "up" : "down", - server.master ? ((int)(time(NULL)-server.master->lastinteraction)) : -1 - ); + (float)c_ru.ru_stime.tv_sec+(float)c_ru.ru_stime.tv_usec/1000000); } - if (server.vm_enabled) { - lockThreadedIO(); + + /* cmdtime */ + if (allsections || !strcasecmp(section,"commandstats")) { + if (sections++) info = sdscat(info,"\r\n"); + info = sdscatprintf(info, "# Commandstats\r\n"); + numcommands = sizeof(redisCommandTable)/sizeof(struct redisCommand); + for (j = 0; j < numcommands; j++) { + struct redisCommand *c = redisCommandTable+j; + + if (!c->calls) continue; + info = sdscatprintf(info, + "cmdstat_%s:calls=%lld,usec=%lld,usec_per_call=%.2f\r\n", + c->name, c->calls, c->microseconds, + (c->calls == 0) ? 0 : ((float)c->microseconds/c->calls)); + } + } + + /* Clusetr */ + if (allsections || defsections || !strcasecmp(section,"cluster")) { + if (sections++) info = sdscat(info,"\r\n"); info = sdscatprintf(info, - "vm_conf_max_memory:%llu\r\n" - "vm_conf_page_size:%llu\r\n" - "vm_conf_pages:%llu\r\n" - "vm_stats_used_pages:%llu\r\n" - "vm_stats_swapped_objects:%llu\r\n" - "vm_stats_swappin_count:%llu\r\n" - "vm_stats_swappout_count:%llu\r\n" - "vm_stats_io_newjobs_len:%lu\r\n" - "vm_stats_io_processing_len:%lu\r\n" - "vm_stats_io_processed_len:%lu\r\n" - "vm_stats_io_active_threads:%lu\r\n" - "vm_stats_blocked_clients:%lu\r\n" - ,(unsigned long long) server.vm_max_memory, - (unsigned long long) server.vm_page_size, - (unsigned long long) server.vm_pages, - (unsigned long long) server.vm_stats_used_pages, - (unsigned long long) server.vm_stats_swapped_objects, - (unsigned long long) server.vm_stats_swapins, - (unsigned long long) server.vm_stats_swapouts, - (unsigned long) listLength(server.io_newjobs), - (unsigned long) listLength(server.io_processing), - (unsigned long) listLength(server.io_processed), - (unsigned long) server.io_active_threads, - (unsigned long) server.vm_blocked_clients - ); - unlockThreadedIO(); + "# Cluster\r\n" + "cluster_enabled:%d\r\n", + server.cluster_enabled); } - for (j = 0; j < server.dbnum; j++) { - long long keys, vkeys; - keys = dictSize(server.db[j].dict); - vkeys = dictSize(server.db[j].expires); - if (keys || vkeys) { - info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n", - j, keys, vkeys); + /* Key space */ + if (allsections || defsections || !strcasecmp(section,"keyspace")) { + if (sections++) info = sdscat(info,"\r\n"); + info = sdscatprintf(info, "# Keyspace\r\n"); + for (j = 0; j < server.dbnum; j++) { + long long keys, vkeys; + + keys = dictSize(server.db[j].dict); + vkeys = dictSize(server.db[j].expires); + if (keys || vkeys) { + info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n", + j, keys, vkeys); + } } } return info; } void infoCommand(redisClient *c) { - sds info = genRedisInfoString(); + char *section = c->argc == 2 ? c->argv[1]->ptr : "default"; + + if (c->argc > 2) { + addReply(c,shared.syntaxerr); + return; + } + sds info = genRedisInfoString(section); addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n", (unsigned long)sdslen(info))); addReplySds(c,info); @@ -1216,61 +1570,6 @@ void monitorCommand(redisClient *c) { /* ============================ Maxmemory directive ======================== */ -/* Try to free one object form the pre-allocated objects free list. - * This is useful under low mem conditions as by default we take 1 million - * free objects allocated. On success REDIS_OK is returned, otherwise - * REDIS_ERR. */ -int tryFreeOneObjectFromFreelist(void) { - robj *o; - - if (server.vm_enabled) pthread_mutex_lock(&server.obj_freelist_mutex); - if (listLength(server.objfreelist)) { - listNode *head = listFirst(server.objfreelist); - o = listNodeValue(head); - listDelNode(server.objfreelist,head); - if (server.vm_enabled) pthread_mutex_unlock(&server.obj_freelist_mutex); - zfree(o); - return REDIS_OK; - } else { - if (server.vm_enabled) pthread_mutex_unlock(&server.obj_freelist_mutex); - return REDIS_ERR; - } -} - -/* A fast RSS sampling function. - * - * The function is reasonably accurate while fast, since it uses the trick of - * using the server.fragmentation ratio that is computed every second and - * is the ratio between the RSS and our zmalloc() count of allocated bytes. - * - * So in order to compute the current RSS used we just need to multiply - * the zmalloc() memory reporting, that is as fast as reading a counter, - * for the latest estimation of fragmentation. - * - * The behavior of this function is also very desirable because it is - * very responsive to memory changes: while the real RSS is actually measured - * in pages, the RSS estimation will actually change even if just a few bytes - * are freed, and this is a good property when the function is used in order - * to evict keys for Virtual Memory of for 'maxmemory' directive. - * - * Note that when the memory reported by zmalloc is smaller than the RSS - * (that is, fragmentation < 1) this means that something is odd (many pages - * swapped since the Redis instance is idle) and we consider the fragmentation - * ratio 1. */ -size_t redisEstimateRSS(void) { - size_t used = zmalloc_used_memory(); - float maxfrag; - - if (server.fragmentation < 1) return used; - maxfrag = (float)SIZE_MAX / used; - - /* If multiplying memory usage reported by zmalloc per fragmentation - * ratio will lead to an overflow we just return SIZE_MAX. */ - if (maxfrag < server.fragmentation) return SIZE_MAX; - - return (size_t)((used * server.fragmentation)); -} - /* This function gets called when 'maxmemory' is set on the config file to limit * the max memory used by the server, and we are out of memory. * This function will try to, in order: @@ -1285,12 +1584,11 @@ size_t redisEstimateRSS(void) { void freeMemoryIfNeeded(void) { /* Remove keys accordingly to the active policy as long as we are * over the memory limit. */ + if (server.maxmemory_policy == REDIS_MAXMEMORY_NO_EVICTION) return; + while (server.maxmemory && zmalloc_used_memory() > server.maxmemory) { int j, k, freed = 0; - /* Basic strategy -- remove objects from the free list. */ - if (tryFreeOneObjectFromFreelist() == REDIS_OK) continue; - for (j = 0; j < server.dbnum; j++) { long bestval = 0; /* just to prevent warning */ sds bestkey = NULL; @@ -1326,6 +1624,10 @@ void freeMemoryIfNeeded(void) { de = dictGetRandomKey(dict); thiskey = dictGetEntryKey(de); + /* When policy is volatile-lru we need an additonal lookup + * to locate the real key, as dict is set to db->expires. */ + if (server.maxmemory_policy == REDIS_MAXMEMORY_VOLATILE_LRU) + de = dictFind(db->dict, thiskey); o = dictGetEntryVal(de); thisval = estimateObjectIdleTime(o); @@ -1359,45 +1661,15 @@ void freeMemoryIfNeeded(void) { /* Finally remove the selected key. */ if (bestkey) { robj *keyobj = createStringObject(bestkey,sdslen(bestkey)); + propagateExpire(db,keyobj); dbDelete(db,keyobj); - server.stat_expiredkeys++; + server.stat_evictedkeys++; decrRefCount(keyobj); freed++; } } if (!freed) return; /* nothing to free... */ } - - while(0) { - int j, k, freed = 0; - for (j = 0; j < server.dbnum; j++) { - int minttl = -1; - sds minkey = NULL; - robj *keyobj = NULL; - struct dictEntry *de; - - if (dictSize(server.db[j].expires)) { - freed = 1; - /* From a sample of three keys drop the one nearest to - * the natural expire */ - for (k = 0; k < 3; k++) { - time_t t; - - de = dictGetRandomKey(server.db[j].expires); - t = (time_t) dictGetEntryVal(de); - if (minttl == -1 || t < minttl) { - minkey = dictGetEntryKey(de); - minttl = t; - } - } - keyobj = createStringObject(minkey,sdslen(minkey)); - dbDelete(server.db+j,keyobj); - server.stat_expiredkeys++; - decrRefCount(keyobj); - } - } - if (!freed) return; /* nothing to free... */ - } } /* =================================== Main! ================================ */ @@ -1428,7 +1700,7 @@ void createPidFile(void) { /* Try to write the pid file in a best-effort way. */ FILE *fp = fopen(server.pidfile,"w"); if (fp) { - fprintf(fp,"%d\n",getpid()); + fprintf(fp,"%d\n",(int)getpid()); fclose(fp); } } @@ -1462,11 +1734,27 @@ void usage() { exit(1); } +void redisAsciiArt(void) { +#include "asciilogo.h" + char *buf = zmalloc(1024*16); + + snprintf(buf,1024*16,ascii_logo, + REDIS_VERSION, + redisGitSHA1(), + strtol(redisGitDirty(),NULL,10) > 0, + (sizeof(long) == 8) ? "64" : "32", + server.cluster_enabled ? "cluster" : "stand alone", + server.port, + (long) getpid() + ); + redisLogRaw(REDIS_NOTICE|REDIS_LOG_RAW,buf); + zfree(buf); +} + int main(int argc, char **argv) { - time_t start; + long long start; initServerConfig(); - sortCommandTable(); if (argc == 2) { if (strcmp(argv[1], "-v") == 0 || strcmp(argv[1], "--version") == 0) version(); @@ -1481,29 +1769,33 @@ int main(int argc, char **argv) { if (server.daemonize) daemonize(); initServer(); if (server.daemonize) createPidFile(); + redisAsciiArt(); redisLog(REDIS_NOTICE,"Server started, Redis version " REDIS_VERSION); #ifdef __linux__ linuxOvercommitMemoryWarning(); #endif - start = time(NULL); - if (server.appendonly) { + start = ustime(); + if (server.ds_enabled) { + redisLog(REDIS_NOTICE,"DB not loaded (running with disk back end)"); + } else if (server.appendonly) { if (loadAppendOnlyFile(server.appendfilename) == REDIS_OK) - redisLog(REDIS_NOTICE,"DB loaded from append only file: %ld seconds",time(NULL)-start); + redisLog(REDIS_NOTICE,"DB loaded from append only file: %.3f seconds",(float)(ustime()-start)/1000000); } else { if (rdbLoad(server.dbfilename) == REDIS_OK) - redisLog(REDIS_NOTICE,"DB loaded from disk: %ld seconds",time(NULL)-start); + redisLog(REDIS_NOTICE,"DB loaded from disk: %.3f seconds",(float)(ustime()-start)/1000000); } - redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port); + if (server.ipfd > 0) + redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port); + if (server.sofd > 0) + redisLog(REDIS_NOTICE,"The server is now ready to accept connections at %s", server.unixsocket); aeSetBeforeSleepProc(server.el,beforeSleep); aeMain(server.el); aeDeleteEventLoop(server.el); return 0; } -/* ============================= Backtrace support ========================= */ - #ifdef HAVE_BACKTRACE -void *getMcontextEip(ucontext_t *uc) { +static void *getMcontextEip(ucontext_t *uc) { #if defined(__FreeBSD__) return (void*) uc->uc_mcontext.mc_eip; #elif defined(__dietlibc__) @@ -1531,7 +1823,7 @@ void *getMcontextEip(ucontext_t *uc) { #endif } -void segvHandler(int sig, siginfo_t *info, void *secret) { +static void sigsegvHandler(int sig, siginfo_t *info, void *secret) { void *trace[100]; char **messages = NULL; int i, trace_size = 0; @@ -1542,8 +1834,8 @@ void segvHandler(int sig, siginfo_t *info, void *secret) { redisLog(REDIS_WARNING, "======= Ooops! Redis %s got signal: -%d- =======", REDIS_VERSION, sig); - infostring = genRedisInfoString(); - redisLog(REDIS_WARNING, "%s",infostring); + infostring = genRedisInfoString("all"); + redisLogRaw(REDIS_WARNING, infostring); /* It's not safe to sdsfree() the returned string under memory * corruption conditions. Let it leak as we are going to abort */ @@ -1570,37 +1862,35 @@ void segvHandler(int sig, siginfo_t *info, void *secret) { sigaction (sig, &act, NULL); kill(getpid(),sig); } +#endif /* HAVE_BACKTRACE */ -void sigtermHandler(int sig) { +static void sigtermHandler(int sig) { REDIS_NOTUSED(sig); - redisLog(REDIS_WARNING,"SIGTERM received, scheduling shutting down..."); + redisLog(REDIS_WARNING,"Received SIGTERM, scheduling shutdown..."); server.shutdown_asap = 1; } -void setupSigSegvAction(void) { +void setupSignalHandlers(void) { struct sigaction act; - sigemptyset (&act.sa_mask); - /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction - * is used. Otherwise, sa_handler is used */ - act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO; - act.sa_sigaction = segvHandler; - sigaction (SIGSEGV, &act, NULL); - sigaction (SIGBUS, &act, NULL); - sigaction (SIGFPE, &act, NULL); - sigaction (SIGILL, &act, NULL); - sigaction (SIGBUS, &act, NULL); - + /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction is used. + * Otherwise, sa_handler is used. */ + sigemptyset(&act.sa_mask); act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND; act.sa_handler = sigtermHandler; - sigaction (SIGTERM, &act, NULL); - return; -} + sigaction(SIGTERM, &act, NULL); -#else /* HAVE_BACKTRACE */ -void setupSigSegvAction(void) { +#ifdef HAVE_BACKTRACE + sigemptyset(&act.sa_mask); + act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO; + act.sa_sigaction = sigsegvHandler; + sigaction(SIGSEGV, &act, NULL); + sigaction(SIGBUS, &act, NULL); + sigaction(SIGFPE, &act, NULL); + sigaction(SIGILL, &act, NULL); +#endif + return; } -#endif /* HAVE_BACKTRACE */ /* The End */