X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/3305306f0933bdfbd958cb12bd559636d50e4dc7..37f68e61024151b1a7461910fe1fcef887b0900b:/redis.c diff --git a/redis.c b/redis.c index d1d0a289..a45f2831 100644 --- a/redis.c +++ b/redis.c @@ -27,16 +27,24 @@ * POSSIBILITY OF SUCH DAMAGE. */ -#define REDIS_VERSION "0.09" +#define REDIS_VERSION "1.050" #include "fmacros.h" +#include "config.h" #include #include #include #include #include +#define __USE_POSIX199309 #include + +#ifdef HAVE_BACKTRACE +#include +#include +#endif /* HAVE_BACKTRACE */ + #include #include #include @@ -48,14 +56,23 @@ #include #include #include +#include #include +#include +#if defined(__sun) +#include "solarisfixes.h" +#endif + +#include "redis.h" #include "ae.h" /* Event driven programming library */ #include "sds.h" /* Dynamic safe strings */ #include "anet.h" /* Networking the easy way */ #include "dict.h" /* Hash tables */ #include "adlist.h" /* Linked lists */ #include "zmalloc.h" /* total memory usage aware version of malloc/free */ +#include "lzf.h" /* LZF compression library */ +#include "pqsort.h" /* Partial qsort for SORT+LIMIT */ /* Error codes */ #define REDIS_OK 0 @@ -64,29 +81,47 @@ /* Static server configuration */ #define REDIS_SERVERPORT 6379 /* TCP port */ #define REDIS_MAXIDLETIME (60*5) /* default client timeout */ -#define REDIS_QUERYBUF_LEN 1024 +#define REDIS_IOBUF_LEN 1024 #define REDIS_LOADBUF_LEN 1024 -#define REDIS_MAX_ARGS 16 +#define REDIS_STATIC_ARGS 4 #define REDIS_DEFAULT_DBNUM 16 #define REDIS_CONFIGLINE_MAX 1024 #define REDIS_OBJFREELIST_MAX 1000000 /* Max number of objects to cache */ #define REDIS_MAX_SYNC_TIME 60 /* Slave can't take more to sync */ +#define REDIS_EXPIRELOOKUPS_PER_CRON 100 /* try to expire 100 keys/second */ +#define REDIS_MAX_WRITE_PER_EVENT (1024*64) +#define REDIS_REQUEST_MAX_SIZE (1024*1024*256) /* max bytes in inline command */ + +/* If more then REDIS_WRITEV_THRESHOLD write packets are pending use writev */ +#define REDIS_WRITEV_THRESHOLD 3 +/* Max number of iovecs used for each writev call */ +#define REDIS_WRITEV_IOVEC_COUNT 256 /* Hash table parameters */ #define REDIS_HT_MINFILL 10 /* Minimal hash table fill 10% */ -#define REDIS_HT_MINSLOTS 16384 /* Never resize the HT under this */ /* Command flags */ -#define REDIS_CMD_BULK 1 -#define REDIS_CMD_INLINE 2 +#define REDIS_CMD_BULK 1 /* Bulk write command */ +#define REDIS_CMD_INLINE 2 /* Inline command */ +/* REDIS_CMD_DENYOOM reserves a longer comment: all the commands marked with + this flags will return an error when the 'maxmemory' option is set in the + config file and the server is using more than maxmemory bytes of memory. + In short this commands are denied on low memory conditions. */ +#define REDIS_CMD_DENYOOM 4 /* Object types */ #define REDIS_STRING 0 #define REDIS_LIST 1 #define REDIS_SET 2 -#define REDIS_HASH 3 +#define REDIS_ZSET 3 +#define REDIS_HASH 4 + +/* Objects encoding */ +#define REDIS_ENCODING_RAW 0 /* Raw representation */ +#define REDIS_ENCODING_INT 1 /* Encoded as integer */ /* Object types only used for dumping to disk */ +#define REDIS_EXPIRETIME 253 #define REDIS_SELECTDB 254 #define REDIS_EOF 255 @@ -115,7 +150,7 @@ #define REDIS_RDB_ENC_INT8 0 /* 8 bit signed integer */ #define REDIS_RDB_ENC_INT16 1 /* 16 bit signed integer */ #define REDIS_RDB_ENC_INT32 2 /* 32 bit signed integer */ -#define REDIS_RDB_ENC_FLZ 3 /* string compressed with FASTLZ */ +#define REDIS_RDB_ENC_LZF 3 /* string compressed with FASTLZ */ /* Client flags */ #define REDIS_CLOSE 1 /* This client connection should be closed ASAP */ @@ -123,22 +158,28 @@ #define REDIS_MASTER 4 /* This client is a master server */ #define REDIS_MONITOR 8 /* This client is a slave monitor, see MONITOR */ -/* Server replication state */ +/* Slave replication state - slave side */ #define REDIS_REPL_NONE 0 /* No active replication */ #define REDIS_REPL_CONNECT 1 /* Must connect to master */ #define REDIS_REPL_CONNECTED 2 /* Connected to master */ +/* Slave replication state - from the point of view of master + * Note that in SEND_BULK and ONLINE state the slave receives new updates + * in its output queue. In the WAIT_BGSAVE state instead the server is waiting + * to start the next background saving in order to send updates to it. */ +#define REDIS_REPL_WAIT_BGSAVE_START 3 /* master waits bgsave to start feeding it */ +#define REDIS_REPL_WAIT_BGSAVE_END 4 /* master waits bgsave to start bulk DB transmission */ +#define REDIS_REPL_SEND_BULK 5 /* master is sending the bulk DB */ +#define REDIS_REPL_ONLINE 6 /* bulk DB already transmitted, receive updates */ + /* List related stuff */ #define REDIS_HEAD 0 #define REDIS_TAIL 1 /* Sort operations */ #define REDIS_SORT_GET 0 -#define REDIS_SORT_DEL 1 -#define REDIS_SORT_INCR 2 -#define REDIS_SORT_DECR 3 -#define REDIS_SORT_ASC 4 -#define REDIS_SORT_DESC 5 +#define REDIS_SORT_ASC 1 +#define REDIS_SORT_DESC 2 #define REDIS_SORTKEY_MAX 1024 /* Log levels */ @@ -149,12 +190,22 @@ /* Anti-warning macro... */ #define REDIS_NOTUSED(V) ((void) V) +#define ZSKIPLIST_MAXLEVEL 32 /* Should be enough for 2^32 elements */ +#define ZSKIPLIST_P 0.25 /* Skiplist P = 1/4 */ + +/* Append only defines */ +#define APPENDFSYNC_NO 0 +#define APPENDFSYNC_ALWAYS 1 +#define APPENDFSYNC_EVERYSEC 2 + /*================================= Data types ============================== */ /* A redis object, that is a type able to hold a string / list / set */ typedef struct redisObject { - int type; void *ptr; + unsigned char type; + unsigned char encoding; + unsigned char notused[2]; int refcount; } robj; @@ -171,15 +222,20 @@ typedef struct redisClient { redisDb *db; int dictid; sds querybuf; - robj *argv[REDIS_MAX_ARGS]; - int argc; - int bulklen; /* bulk read len. -1 if not in bulk read mode */ + robj **argv, **mbargv; + int argc, mbargc; + int bulklen; /* bulk read len. -1 if not in bulk read mode */ + int multibulk; /* multi bulk command format active */ list *reply; int sentlen; time_t lastinteraction; /* time of the last interaction, used for timeout */ - int flags; /* REDIS_CLOSE | REDIS_SLAVE | REDIS_MONITOR */ - int slaveseldb; /* slave selected db, if this client is a slave */ - int authenticated; /* when requirepass is non-NULL */ + int flags; /* REDIS_CLOSE | REDIS_SLAVE | REDIS_MONITOR */ + int slaveseldb; /* slave selected db, if this client is a slave */ + int authenticated; /* when requirepass is non-NULL */ + int replstate; /* replication state if this is a slave */ + int repldbfd; /* replication DB file descriptor */ + long repldboff; /* replication DB file offset */ + off_t repldbsize; /* replication DB file size */ } redisClient; struct saveparam { @@ -202,7 +258,7 @@ struct redisServer { int cronloops; /* number of times the cron function run */ list *objfreelist; /* A list of freed objects to avoid malloc() */ time_t lastsave; /* Unix time of last save succeeede */ - int usedmemory; /* Used memory in megabytes */ + size_t usedmemory; /* Used memory in megabytes */ /* Fields used only for stats */ time_t stat_starttime; /* server start time */ long long stat_numcommands; /* number of processed commands */ @@ -213,21 +269,31 @@ struct redisServer { int maxidletime; int dbnum; int daemonize; + int appendonly; + int appendfsync; + time_t lastfsync; + int appendfd; + int appendseldb; char *pidfile; int bgsaveinprogress; + pid_t bgsavechildpid; struct saveparam *saveparams; int saveparamslen; char *logfile; char *bindaddr; char *dbfilename; + char *appendfilename; char *requirepass; int shareobjects; /* Replication related */ int isslave; + char *masterauth; char *masterhost; int masterport; - redisClient *master; + redisClient *master; /* client that is master for this slave */ int replstate; + unsigned int maxclients; + unsigned long maxmemory; /* Sort parameters - qsort_r() is only available under BSD so we * have to take this state global, in order to pass it to sortCompare() */ int sort_desc; @@ -243,6 +309,11 @@ struct redisCommand { int flags; }; +struct redisFunctionSym { + char *name; + unsigned long pointer; +}; + typedef struct _redisSortObject { robj *obj; union { @@ -256,6 +327,28 @@ typedef struct _redisSortOperation { robj *pattern; } redisSortOperation; +/* ZSETs use a specialized version of Skiplists */ + +typedef struct zskiplistNode { + struct zskiplistNode **forward; + struct zskiplistNode *backward; + double score; + robj *obj; +} zskiplistNode; + +typedef struct zskiplist { + struct zskiplistNode *header, *tail; + unsigned long length; + int level; +} zskiplist; + +typedef struct zset { + dict *dict; + zskiplist *zsl; +} zset; + +/* Our shared "common" objects */ + struct sharedObjectsStruct { robj *crlf, *ok, *err, *emptybulk, *czero, *cone, *pong, *space, *colon, *nullbulk, *nullmultibulk, @@ -265,6 +358,12 @@ struct sharedObjectsStruct { *select5, *select6, *select7, *select8, *select9; } shared; +/* Global vars that are actally used as constants. The following double + * values are used for double on-disk serialization, and are initialized + * at runtime to avoid strange compiler optimizations. */ + +static double R_Zero, R_PosInf, R_NegInf, R_Nan; + /*================================ Prototypes =============================== */ static void freeStringObject(robj *o); @@ -280,11 +379,28 @@ static void incrRefCount(robj *o); static int rdbSaveBackground(char *filename); static robj *createStringObject(char *ptr, size_t len); static void replicationFeedSlaves(list *slaves, struct redisCommand *cmd, int dictid, robj **argv, int argc); +static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc); static int syncWithMaster(void); static robj *tryObjectSharing(robj *o); +static int tryObjectEncoding(robj *o); +static robj *getDecodedObject(const robj *o); static int removeExpire(redisDb *db, robj *key); static int expireIfNeeded(redisDb *db, robj *key); static int deleteIfVolatile(redisDb *db, robj *key); +static int deleteKey(redisDb *db, robj *key); +static time_t getExpire(redisDb *db, robj *key); +static int setExpire(redisDb *db, robj *key, time_t when); +static void updateSlavesWaitingBgsave(int bgsaveerr); +static void freeMemoryIfNeeded(void); +static int processCommand(redisClient *c); +static void setupSigSegvAction(void); +static void rdbRemoveTempFile(pid_t childpid); +static size_t stringObjectLen(robj *o); +static void processInputBuffer(redisClient *c); +static zskiplist *zslCreate(void); +static void zslFree(zskiplist *zsl); +static void zslInsert(zskiplist *zsl, double score, robj *obj); +static void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int mask); static void authCommand(redisClient *c); static void pingCommand(redisClient *c); @@ -321,19 +437,42 @@ static void typeCommand(redisClient *c); static void lsetCommand(redisClient *c); static void saddCommand(redisClient *c); static void sremCommand(redisClient *c); +static void smoveCommand(redisClient *c); static void sismemberCommand(redisClient *c); static void scardCommand(redisClient *c); +static void spopCommand(redisClient *c); +static void srandmemberCommand(redisClient *c); static void sinterCommand(redisClient *c); static void sinterstoreCommand(redisClient *c); +static void sunionCommand(redisClient *c); +static void sunionstoreCommand(redisClient *c); +static void sdiffCommand(redisClient *c); +static void sdiffstoreCommand(redisClient *c); static void syncCommand(redisClient *c); static void flushdbCommand(redisClient *c); static void flushallCommand(redisClient *c); static void sortCommand(redisClient *c); static void lremCommand(redisClient *c); +static void rpoplpushcommand(redisClient *c); static void infoCommand(redisClient *c); static void mgetCommand(redisClient *c); static void monitorCommand(redisClient *c); static void expireCommand(redisClient *c); +static void expireatCommand(redisClient *c); +static void getsetCommand(redisClient *c); +static void ttlCommand(redisClient *c); +static void slaveofCommand(redisClient *c); +static void debugCommand(redisClient *c); +static void msetCommand(redisClient *c); +static void msetnxCommand(redisClient *c); +static void zaddCommand(redisClient *c); +static void zrangeCommand(redisClient *c); +static void zrangebyscoreCommand(redisClient *c); +static void zrevrangeCommand(redisClient *c); +static void zcardCommand(redisClient *c); +static void zremCommand(redisClient *c); +static void zscoreCommand(redisClient *c); +static void zremrangebyscoreCommand(redisClient *c); /*================================= Globals ================================= */ @@ -341,37 +480,58 @@ static void expireCommand(redisClient *c); static struct redisServer server; /* server global state */ static struct redisCommand cmdTable[] = { {"get",getCommand,2,REDIS_CMD_INLINE}, - {"set",setCommand,3,REDIS_CMD_BULK}, - {"setnx",setnxCommand,3,REDIS_CMD_BULK}, - {"del",delCommand,2,REDIS_CMD_INLINE}, + {"set",setCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, + {"setnx",setnxCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, + {"del",delCommand,-2,REDIS_CMD_INLINE}, {"exists",existsCommand,2,REDIS_CMD_INLINE}, - {"incr",incrCommand,2,REDIS_CMD_INLINE}, - {"decr",decrCommand,2,REDIS_CMD_INLINE}, + {"incr",incrCommand,2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, + {"decr",decrCommand,2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, {"mget",mgetCommand,-2,REDIS_CMD_INLINE}, - {"rpush",rpushCommand,3,REDIS_CMD_BULK}, - {"lpush",lpushCommand,3,REDIS_CMD_BULK}, + {"rpush",rpushCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, + {"lpush",lpushCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, {"rpop",rpopCommand,2,REDIS_CMD_INLINE}, {"lpop",lpopCommand,2,REDIS_CMD_INLINE}, {"llen",llenCommand,2,REDIS_CMD_INLINE}, {"lindex",lindexCommand,3,REDIS_CMD_INLINE}, - {"lset",lsetCommand,4,REDIS_CMD_BULK}, + {"lset",lsetCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, {"lrange",lrangeCommand,4,REDIS_CMD_INLINE}, {"ltrim",ltrimCommand,4,REDIS_CMD_INLINE}, {"lrem",lremCommand,4,REDIS_CMD_BULK}, - {"sadd",saddCommand,3,REDIS_CMD_BULK}, + {"rpoplpush",rpoplpushcommand,3,REDIS_CMD_BULK}, + {"sadd",saddCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, {"srem",sremCommand,3,REDIS_CMD_BULK}, + {"smove",smoveCommand,4,REDIS_CMD_BULK}, {"sismember",sismemberCommand,3,REDIS_CMD_BULK}, {"scard",scardCommand,2,REDIS_CMD_INLINE}, - {"sinter",sinterCommand,-2,REDIS_CMD_INLINE}, - {"sinterstore",sinterstoreCommand,-3,REDIS_CMD_INLINE}, + {"spop",spopCommand,2,REDIS_CMD_INLINE}, + {"srandmember",srandmemberCommand,2,REDIS_CMD_INLINE}, + {"sinter",sinterCommand,-2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, + {"sinterstore",sinterstoreCommand,-3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, + {"sunion",sunionCommand,-2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, + {"sunionstore",sunionstoreCommand,-3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, + {"sdiff",sdiffCommand,-2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, + {"sdiffstore",sdiffstoreCommand,-3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, {"smembers",sinterCommand,2,REDIS_CMD_INLINE}, - {"incrby",incrbyCommand,3,REDIS_CMD_INLINE}, - {"decrby",decrbyCommand,3,REDIS_CMD_INLINE}, + {"zadd",zaddCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, + {"zrem",zremCommand,3,REDIS_CMD_BULK}, + {"zremrangebyscore",zremrangebyscoreCommand,4,REDIS_CMD_INLINE}, + {"zrange",zrangeCommand,4,REDIS_CMD_INLINE}, + {"zrangebyscore",zrangebyscoreCommand,4,REDIS_CMD_INLINE}, + {"zrevrange",zrevrangeCommand,4,REDIS_CMD_INLINE}, + {"zcard",zcardCommand,2,REDIS_CMD_INLINE}, + {"zscore",zscoreCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, + {"incrby",incrbyCommand,3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, + {"decrby",decrbyCommand,3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, + {"getset",getsetCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, + {"mset",msetCommand,-3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, + {"msetnx",msetnxCommand,-3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, {"randomkey",randomkeyCommand,1,REDIS_CMD_INLINE}, {"select",selectCommand,2,REDIS_CMD_INLINE}, {"move",moveCommand,3,REDIS_CMD_INLINE}, {"rename",renameCommand,3,REDIS_CMD_INLINE}, {"renamenx",renamenxCommand,3,REDIS_CMD_INLINE}, + {"expire",expireCommand,3,REDIS_CMD_INLINE}, + {"expireat",expireatCommand,3,REDIS_CMD_INLINE}, {"keys",keysCommand,2,REDIS_CMD_INLINE}, {"dbsize",dbsizeCommand,1,REDIS_CMD_INLINE}, {"auth",authCommand,2,REDIS_CMD_INLINE}, @@ -385,10 +545,12 @@ static struct redisCommand cmdTable[] = { {"sync",syncCommand,1,REDIS_CMD_INLINE}, {"flushdb",flushdbCommand,1,REDIS_CMD_INLINE}, {"flushall",flushallCommand,1,REDIS_CMD_INLINE}, - {"sort",sortCommand,-2,REDIS_CMD_INLINE}, + {"sort",sortCommand,-2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, {"info",infoCommand,1,REDIS_CMD_INLINE}, {"monitor",monitorCommand,1,REDIS_CMD_INLINE}, - {"expire",expireCommand,3,REDIS_CMD_INLINE}, + {"ttl",ttlCommand,2,REDIS_CMD_INLINE}, + {"slaveof",slaveofCommand,3,REDIS_CMD_INLINE}, + {"debug",debugCommand,-2,REDIS_CMD_INLINE}, {NULL,NULL,0,0} }; @@ -517,8 +679,7 @@ int stringmatchlen(const char *pattern, int patternLen, return 0; } -void redisLog(int level, const char *fmt, ...) -{ +static void redisLog(int level, const char *fmt, ...) { va_list ap; FILE *fp; @@ -528,7 +689,12 @@ void redisLog(int level, const char *fmt, ...) va_start(ap, fmt); if (level >= server.verbosity) { char *c = ".-*"; - fprintf(fp,"%c ",c[level]); + char buf[64]; + time_t now; + + now = time(NULL); + strftime(buf,64,"%d %b %H:%M:%S",localtime(&now)); + fprintf(fp,"%s %c ",buf,c[level]); vfprintf(fp, fmt, ap); fprintf(fp,"\n"); fflush(fp); @@ -544,6 +710,12 @@ void redisLog(int level, const char *fmt, ...) * keys and radis objects as values (objects can hold SDS strings, * lists, sets). */ +static void dictVanillaFree(void *privdata, void *val) +{ + DICT_NOTUSED(privdata); + zfree(val); +} + static int sdsDictKeyCompare(void *privdata, const void *key1, const void *key2) { @@ -563,32 +735,77 @@ static void dictRedisObjectDestructor(void *privdata, void *val) decrRefCount(val); } -static int dictSdsKeyCompare(void *privdata, const void *key1, +static int dictObjKeyCompare(void *privdata, const void *key1, const void *key2) { const robj *o1 = key1, *o2 = key2; return sdsDictKeyCompare(privdata,o1->ptr,o2->ptr); } -static unsigned int dictSdsHash(const void *key) { +static unsigned int dictObjHash(const void *key) { const robj *o = key; return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr)); } +static int dictEncObjKeyCompare(void *privdata, const void *key1, + const void *key2) +{ + const robj *o1 = key1, *o2 = key2; + + if (o1->encoding == REDIS_ENCODING_RAW && + o2->encoding == REDIS_ENCODING_RAW) + return sdsDictKeyCompare(privdata,o1->ptr,o2->ptr); + else { + robj *dec1, *dec2; + int cmp; + + dec1 = o1->encoding != REDIS_ENCODING_RAW ? + getDecodedObject(o1) : (robj*)o1; + dec2 = o2->encoding != REDIS_ENCODING_RAW ? + getDecodedObject(o2) : (robj*)o2; + cmp = sdsDictKeyCompare(privdata,dec1->ptr,dec2->ptr); + if (dec1 != o1) decrRefCount(dec1); + if (dec2 != o2) decrRefCount(dec2); + return cmp; + } +} + +static unsigned int dictEncObjHash(const void *key) { + const robj *o = key; + + if (o->encoding == REDIS_ENCODING_RAW) + return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr)); + else { + robj *dec = getDecodedObject(o); + unsigned int hash = dictGenHashFunction(dec->ptr, sdslen((sds)dec->ptr)); + decrRefCount(dec); + return hash; + } +} + static dictType setDictType = { - dictSdsHash, /* hash function */ + dictEncObjHash, /* hash function */ NULL, /* key dup */ NULL, /* val dup */ - dictSdsKeyCompare, /* key compare */ + dictEncObjKeyCompare, /* key compare */ dictRedisObjectDestructor, /* key destructor */ NULL /* val destructor */ }; +static dictType zsetDictType = { + dictEncObjHash, /* hash function */ + NULL, /* key dup */ + NULL, /* val dup */ + dictEncObjKeyCompare, /* key compare */ + dictRedisObjectDestructor, /* key destructor */ + dictVanillaFree /* val destructor */ +}; + static dictType hashDictType = { - dictSdsHash, /* hash function */ + dictObjHash, /* hash function */ NULL, /* key dup */ NULL, /* val dup */ - dictSdsKeyCompare, /* key compare */ + dictObjKeyCompare, /* key compare */ dictRedisObjectDestructor, /* key destructor */ dictRedisObjectDestructor /* val destructor */ }; @@ -608,27 +825,50 @@ static void oom(const char *msg) { } /* ====================== Redis server networking stuff ===================== */ -void closeTimedoutClients(void) { +static void closeTimedoutClients(void) { redisClient *c; - listIter *li; listNode *ln; time_t now = time(NULL); - li = listGetIterator(server.clients,AL_START_HEAD); - if (!li) return; - while ((ln = listNextElement(li)) != NULL) { + listRewind(server.clients); + while ((ln = listYield(server.clients)) != NULL) { c = listNodeValue(ln); if (!(c->flags & REDIS_SLAVE) && /* no timeout for slaves */ + !(c->flags & REDIS_MASTER) && /* no timeout for masters */ (now - c->lastinteraction > server.maxidletime)) { redisLog(REDIS_DEBUG,"Closing idle client"); freeClient(c); } } - listReleaseIterator(li); } -int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { - int j, size, used, loops = server.cronloops++; +static int htNeedsResize(dict *dict) { + long long size, used; + + size = dictSlots(dict); + used = dictSize(dict); + return (size && used && size > DICT_HT_INITIAL_SIZE && + (used*100/size < REDIS_HT_MINFILL)); +} + +/* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL + * we resize the hash table to save memory */ +static void tryResizeHashTables(void) { + int j; + + for (j = 0; j < server.dbnum; j++) { + if (htNeedsResize(server.db[j].dict)) { + redisLog(REDIS_DEBUG,"The hash table %d is too sparse, resize it...",j); + dictResize(server.db[j].dict); + redisLog(REDIS_DEBUG,"Hash table %d resized.",j); + } + if (htNeedsResize(server.db[j].expires)) + dictResize(server.db[j].expires); + } +} + +static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { + int j, loops = server.cronloops++; REDIS_NOTUSED(eventLoop); REDIS_NOTUSED(id); REDIS_NOTUSED(clientData); @@ -636,26 +876,30 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { /* Update the global state with the amount of used memory */ server.usedmemory = zmalloc_used_memory(); - /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL - * we resize the hash table to save memory */ + /* Show some info about non-empty databases */ for (j = 0; j < server.dbnum; j++) { + long long size, used, vkeys; + size = dictSlots(server.db[j].dict); used = dictSize(server.db[j].dict); - if (!(loops % 5) && used > 0) { - redisLog(REDIS_DEBUG,"DB %d: %d keys in %d slots HT.",j,used,size); + vkeys = dictSize(server.db[j].expires); + if (!(loops % 5) && (used || vkeys)) { + redisLog(REDIS_DEBUG,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size); /* dictPrintStats(server.dict); */ } - if (size && used && size > REDIS_HT_MINSLOTS && - (used*100/size < REDIS_HT_MINFILL)) { - redisLog(REDIS_NOTICE,"The hash table %d is too sparse, resize it...",j); - dictResize(server.db[j].dict); - redisLog(REDIS_NOTICE,"Hash table %d resized.",j); - } } + /* We don't want to resize the hash tables while a bacground saving + * is in progress: the saving child is created using fork() that is + * implemented with a copy-on-write semantic in most modern systems, so + * if we resize the HT while there is the saving child at work actually + * a lot of memory movements in the parent will cause a lot of pages + * copied. */ + if (!server.bgsaveinprogress) tryResizeHashTables(); + /* Show information about connected clients */ if (!(loops % 5)) { - redisLog(REDIS_DEBUG,"%d clients connected (%d slaves), %d bytes in use", + redisLog(REDIS_DEBUG,"%d clients connected (%d slaves), %zu bytes in use, %d shared objects", listLength(server.clients)-listLength(server.slaves), listLength(server.slaves), server.usedmemory, @@ -663,24 +907,31 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { } /* Close connections of timedout clients */ - if (!(loops % 10)) + if (server.maxidletime && !(loops % 10)) closeTimedoutClients(); /* Check if a background saving in progress terminated */ if (server.bgsaveinprogress) { int statloc; - if (wait4(-1,&statloc,WNOHANG,NULL)) { + if (wait3(&statloc,WNOHANG,NULL)) { int exitcode = WEXITSTATUS(statloc); - if (exitcode == 0) { + int bysignal = WIFSIGNALED(statloc); + + if (!bysignal && exitcode == 0) { redisLog(REDIS_NOTICE, "Background saving terminated with success"); server.dirty = 0; server.lastsave = time(NULL); + } else if (!bysignal && exitcode != 0) { + redisLog(REDIS_WARNING, "Background saving error"); } else { redisLog(REDIS_WARNING, - "Background saving error"); + "Background saving terminated by signal"); + rdbRemoveTempFile(server.bgsavechildpid); } server.bgsaveinprogress = 0; + server.bgsavechildpid = -1; + updateSlavesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR); } } else { /* If there is not a background saving in progress check if @@ -698,6 +949,38 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { } } } + + /* Try to expire a few timed out keys. The algorithm used is adaptive and + * will use few CPU cycles if there are few expiring keys, otherwise + * it will get more aggressive to avoid that too much memory is used by + * keys that can be removed from the keyspace. */ + for (j = 0; j < server.dbnum; j++) { + int expired; + redisDb *db = server.db+j; + + /* Continue to expire if at the end of the cycle more than 25% + * of the keys were expired. */ + do { + int num = dictSize(db->expires); + time_t now = time(NULL); + + expired = 0; + if (num > REDIS_EXPIRELOOKUPS_PER_CRON) + num = REDIS_EXPIRELOOKUPS_PER_CRON; + while (num--) { + dictEntry *de; + time_t t; + + if ((de = dictGetRandomKey(db->expires)) == NULL) break; + t = (time_t) dictGetEntryVal(de); + if (now > t) { + deleteKey(db,dictGetEntryKey(de)); + expired++; + } + } + } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4); + } + /* Check if we should connect to a MASTER */ if (server.replstate == REDIS_REPL_CONNECT) { redisLog(REDIS_NOTICE,"Connecting to MASTER..."); @@ -747,13 +1030,12 @@ static void createSharedObjects(void) { static void appendServerSaveParams(time_t seconds, int changes) { server.saveparams = zrealloc(server.saveparams,sizeof(struct saveparam)*(server.saveparamslen+1)); - if (server.saveparams == NULL) oom("appendServerSaveParams"); server.saveparams[server.saveparamslen].seconds = seconds; server.saveparams[server.saveparamslen].changes = changes; server.saveparamslen++; } -static void ResetServerSaveParams() { +static void resetServerSaveParams() { zfree(server.saveparams); server.saveparams = NULL; server.saveparamslen = 0; @@ -769,21 +1051,37 @@ static void initServerConfig() { server.bindaddr = NULL; server.glueoutputbuf = 1; server.daemonize = 0; + server.appendonly = 0; + server.appendfsync = APPENDFSYNC_ALWAYS; + server.lastfsync = time(NULL); + server.appendfd = -1; + server.appendseldb = -1; /* Make sure the first time will not match */ server.pidfile = "/var/run/redis.pid"; server.dbfilename = "dump.rdb"; + server.appendfilename = "appendonly.log"; server.requirepass = NULL; server.shareobjects = 0; - ResetServerSaveParams(); + server.sharingpoolsize = 1024; + server.maxclients = 0; + server.maxmemory = 0; + resetServerSaveParams(); appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */ appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */ appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */ /* Replication related */ server.isslave = 0; + server.masterauth = NULL; server.masterhost = NULL; server.masterport = 6379; server.master = NULL; server.replstate = REDIS_REPL_NONE; + + /* 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; } static void initServer() { @@ -791,6 +1089,7 @@ static void initServer() { signal(SIGHUP, SIG_IGN); signal(SIGPIPE, SIG_IGN); + setupSigSegvAction(); server.clients = listCreate(); server.slaves = listCreate(); @@ -800,9 +1099,6 @@ static void initServer() { server.el = aeCreateEventLoop(); server.db = zmalloc(sizeof(redisDb)*server.dbnum); server.sharingpool = dictCreate(&setDictType,NULL); - server.sharingpoolsize = 1024; - if (!server.db || !server.clients || !server.slaves || !server.monitors || !server.el || !server.objfreelist) - oom("server initialization"); /* Fatal OOM */ server.fd = anetTcpServer(server.neterr, server.port, server.bindaddr); if (server.fd == -1) { redisLog(REDIS_WARNING, "Opening TCP port: %s", server.neterr); @@ -815,37 +1111,61 @@ static void initServer() { } server.cronloops = 0; server.bgsaveinprogress = 0; + server.bgsavechildpid = -1; server.lastsave = time(NULL); server.dirty = 0; server.usedmemory = 0; server.stat_numcommands = 0; server.stat_numconnections = 0; server.stat_starttime = time(NULL); - aeCreateTimeEvent(server.el, 1000, serverCron, NULL, NULL); + aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL); + + if (server.appendonly) { + server.appendfd = open(server.appendfilename,O_WRONLY|O_APPEND|O_CREAT,0644); + if (server.appendfd == -1) { + redisLog(REDIS_WARNING, "Can't open the append-only file: %s", + strerror(errno)); + exit(1); + } + } } /* Empty the whole database */ -static void emptyDb() { +static long long emptyDb() { int j; + long long removed = 0; for (j = 0; j < server.dbnum; j++) { + removed += dictSize(server.db[j].dict); dictEmpty(server.db[j].dict); dictEmpty(server.db[j].expires); } + return removed; +} + +static int yesnotoi(char *s) { + if (!strcasecmp(s,"yes")) return 1; + else if (!strcasecmp(s,"no")) return 0; + else return -1; } /* I agree, this is a very rudimental way to load a configuration... will improve later if the config gets more complex */ static void loadServerConfig(char *filename) { - FILE *fp = fopen(filename,"r"); + FILE *fp; char buf[REDIS_CONFIGLINE_MAX+1], *err = NULL; int linenum = 0; sds line = NULL; - - if (!fp) { - redisLog(REDIS_WARNING,"Fatal error, can't open config file"); - exit(1); + + if (filename[0] == '-' && filename[1] == '\0') + fp = stdin; + else { + if ((fp = fopen(filename,"r")) == NULL) { + redisLog(REDIS_WARNING,"Fatal error, can't open config file"); + exit(1); + } } + while(fgets(buf,REDIS_CONFIGLINE_MAX+1,fp) != NULL) { sds *argv; int argc, j; @@ -865,92 +1185,111 @@ static void loadServerConfig(char *filename) { sdstolower(argv[0]); /* Execute config directives */ - if (!strcmp(argv[0],"timeout") && argc == 2) { + if (!strcasecmp(argv[0],"timeout") && argc == 2) { server.maxidletime = atoi(argv[1]); - if (server.maxidletime < 1) { + if (server.maxidletime < 0) { err = "Invalid timeout value"; goto loaderr; } - } else if (!strcmp(argv[0],"port") && argc == 2) { + } else if (!strcasecmp(argv[0],"port") && argc == 2) { server.port = atoi(argv[1]); if (server.port < 1 || server.port > 65535) { err = "Invalid port"; goto loaderr; } - } else if (!strcmp(argv[0],"bind") && argc == 2) { + } else if (!strcasecmp(argv[0],"bind") && argc == 2) { server.bindaddr = zstrdup(argv[1]); - } else if (!strcmp(argv[0],"save") && argc == 3) { + } else if (!strcasecmp(argv[0],"save") && argc == 3) { int seconds = atoi(argv[1]); int changes = atoi(argv[2]); if (seconds < 1 || changes < 0) { err = "Invalid save parameters"; goto loaderr; } appendServerSaveParams(seconds,changes); - } else if (!strcmp(argv[0],"dir") && argc == 2) { + } else if (!strcasecmp(argv[0],"dir") && argc == 2) { if (chdir(argv[1]) == -1) { redisLog(REDIS_WARNING,"Can't chdir to '%s': %s", argv[1], strerror(errno)); exit(1); } - } else if (!strcmp(argv[0],"loglevel") && argc == 2) { - if (!strcmp(argv[1],"debug")) server.verbosity = REDIS_DEBUG; - else if (!strcmp(argv[1],"notice")) server.verbosity = REDIS_NOTICE; - else if (!strcmp(argv[1],"warning")) server.verbosity = REDIS_WARNING; + } else if (!strcasecmp(argv[0],"loglevel") && argc == 2) { + if (!strcasecmp(argv[1],"debug")) server.verbosity = REDIS_DEBUG; + else if (!strcasecmp(argv[1],"notice")) server.verbosity = REDIS_NOTICE; + else if (!strcasecmp(argv[1],"warning")) server.verbosity = REDIS_WARNING; else { err = "Invalid log level. Must be one of debug, notice, warning"; goto loaderr; } - } else if (!strcmp(argv[0],"logfile") && argc == 2) { - FILE *fp; + } else if (!strcasecmp(argv[0],"logfile") && argc == 2) { + FILE *logfp; server.logfile = zstrdup(argv[1]); - if (!strcmp(server.logfile,"stdout")) { + if (!strcasecmp(server.logfile,"stdout")) { zfree(server.logfile); server.logfile = NULL; } if (server.logfile) { /* Test if we are able to open the file. The server will not * be able to abort just for this problem later... */ - fp = fopen(server.logfile,"a"); - if (fp == NULL) { + logfp = fopen(server.logfile,"a"); + if (logfp == NULL) { err = sdscatprintf(sdsempty(), "Can't open the log file: %s", strerror(errno)); goto loaderr; } - fclose(fp); + fclose(logfp); } - } else if (!strcmp(argv[0],"databases") && argc == 2) { + } else if (!strcasecmp(argv[0],"databases") && argc == 2) { server.dbnum = atoi(argv[1]); if (server.dbnum < 1) { err = "Invalid number of databases"; goto loaderr; } - } else if (!strcmp(argv[0],"slaveof") && argc == 3) { + } else if (!strcasecmp(argv[0],"maxclients") && argc == 2) { + server.maxclients = atoi(argv[1]); + } else if (!strcasecmp(argv[0],"maxmemory") && argc == 2) { + server.maxmemory = strtoll(argv[1], NULL, 10); + } else if (!strcasecmp(argv[0],"slaveof") && argc == 3) { server.masterhost = sdsnew(argv[1]); server.masterport = atoi(argv[2]); server.replstate = REDIS_REPL_CONNECT; - } else if (!strcmp(argv[0],"glueoutputbuf") && argc == 2) { - sdstolower(argv[1]); - if (!strcmp(argv[1],"yes")) server.glueoutputbuf = 1; - else if (!strcmp(argv[1],"no")) server.glueoutputbuf = 0; - else { + } else if (!strcasecmp(argv[0],"masterauth") && argc == 2) { + server.masterauth = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"glueoutputbuf") && argc == 2) { + if ((server.glueoutputbuf = yesnotoi(argv[1])) == -1) { err = "argument must be 'yes' or 'no'"; goto loaderr; } - } else if (!strcmp(argv[0],"shareobjects") && argc == 2) { - sdstolower(argv[1]); - if (!strcmp(argv[1],"yes")) server.shareobjects = 1; - else if (!strcmp(argv[1],"no")) server.shareobjects = 0; - else { + } else if (!strcasecmp(argv[0],"shareobjects") && argc == 2) { + if ((server.shareobjects = yesnotoi(argv[1])) == -1) { err = "argument must be 'yes' or 'no'"; goto loaderr; } - } else if (!strcmp(argv[0],"daemonize") && argc == 2) { - sdstolower(argv[1]); - if (!strcmp(argv[1],"yes")) server.daemonize = 1; - else if (!strcmp(argv[1],"no")) server.daemonize = 0; - else { + } else if (!strcasecmp(argv[0],"shareobjectspoolsize") && argc == 2) { + server.sharingpoolsize = atoi(argv[1]); + if (server.sharingpoolsize < 1) { + err = "invalid object sharing pool size"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"daemonize") && argc == 2) { + if ((server.daemonize = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"appendonly") && argc == 2) { + if ((server.appendonly = yesnotoi(argv[1])) == -1) { err = "argument must be 'yes' or 'no'"; goto loaderr; } - } else if (!strcmp(argv[0],"requirepass") && argc == 2) { + } else if (!strcasecmp(argv[0],"appendfsync") && argc == 2) { + if (!strcasecmp(argv[1],"no")) { + server.appendfsync = APPENDFSYNC_NO; + } else if (!strcasecmp(argv[1],"always")) { + server.appendfsync = APPENDFSYNC_ALWAYS; + } else if (!strcasecmp(argv[1],"everysec")) { + server.appendfsync = APPENDFSYNC_EVERYSEC; + } else { + err = "argument must be 'no', 'always' or 'everysec'"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"requirepass") && argc == 2) { server.requirepass = zstrdup(argv[1]); - } else if (!strcmp(argv[0],"pidfile") && argc == 2) { + } else if (!strcasecmp(argv[0],"pidfile") && argc == 2) { server.pidfile = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"dbfilename") && argc == 2) { + server.dbfilename = zstrdup(argv[1]); } else { err = "Bad directive or wrong number of arguments"; goto loaderr; } @@ -959,7 +1298,7 @@ static void loadServerConfig(char *filename) { zfree(argv); sdsfree(line); } - fclose(fp); + if (fp != stdin) fclose(fp); return; loaderr: @@ -975,7 +1314,10 @@ static void freeClientArgv(redisClient *c) { for (j = 0; j < c->argc; j++) decrRefCount(c->argv[j]); + for (j = 0; j < c->mbargc; j++) + decrRefCount(c->mbargv[j]); c->argc = 0; + c->mbargc = 0; } static void freeClient(redisClient *c) { @@ -991,6 +1333,8 @@ static void freeClient(redisClient *c) { assert(ln != NULL); listDelNode(server.clients,ln); if (c->flags & REDIS_SLAVE) { + if (c->replstate == REDIS_REPL_SEND_BULK && c->repldbfd != -1) + close(c->repldbfd); list *l = (c->flags & REDIS_MONITOR) ? server.monitors : server.slaves; ln = listSearchKey(l,c); assert(ln != NULL); @@ -1000,38 +1344,36 @@ static void freeClient(redisClient *c) { server.master = NULL; server.replstate = REDIS_REPL_CONNECT; } + zfree(c->argv); + zfree(c->mbargv); zfree(c); } +#define GLUEREPLY_UP_TO (1024) static void glueReplyBuffersIfNeeded(redisClient *c) { - int totlen = 0; - listNode *ln = c->reply->head, *next; + int copylen = 0; + char buf[GLUEREPLY_UP_TO]; + listNode *ln; robj *o; - while(ln) { + listRewind(c->reply); + while((ln = listYield(c->reply))) { + int objlen; + o = ln->value; - totlen += sdslen(o->ptr); - ln = ln->next; - /* This optimization makes more sense if we don't have to copy - * too much data */ - if (totlen > 1024) return; - } - if (totlen > 0) { - char buf[1024]; - int copylen = 0; - - ln = c->reply->head; - while(ln) { - next = ln->next; - o = ln->value; - memcpy(buf+copylen,o->ptr,sdslen(o->ptr)); - copylen += sdslen(o->ptr); + objlen = sdslen(o->ptr); + if (copylen + objlen <= GLUEREPLY_UP_TO) { + memcpy(buf+copylen,o->ptr,objlen); + copylen += objlen; listDelNode(c->reply,ln); - ln = next; + } else { + if (copylen == 0) return; + break; } - /* Now the output buffer is empty, add the new single element */ - addReplySds(c,sdsnewlen(buf,totlen)); } + /* Now the output buffer is empty, add the new single element */ + o = createObject(REDIS_STRING,sdsnewlen(buf,copylen)); + listAddNodeHead(c->reply,o); } static void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) { @@ -1041,9 +1383,20 @@ static void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) REDIS_NOTUSED(el); REDIS_NOTUSED(mask); - if (server.glueoutputbuf && listLength(c->reply) > 1) - glueReplyBuffersIfNeeded(c); + + /* Use writev() if we have enough buffers to send */ + if (!server.glueoutputbuf && + listLength(c->reply) > REDIS_WRITEV_THRESHOLD && + !(c->flags & REDIS_MASTER)) + { + sendReplyToClientWritev(el, fd, privdata, mask); + return; + } + while(listLength(c->reply)) { + if (server.glueoutputbuf && listLength(c->reply) > 1) + glueReplyBuffersIfNeeded(c); + o = listNodeValue(listFirst(c->reply)); objlen = sdslen(o->ptr); @@ -1053,6 +1406,7 @@ static void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) } if (c->flags & REDIS_MASTER) { + /* Don't reply to a master */ nwritten = objlen - c->sentlen; } else { nwritten = write(fd, ((char*)o->ptr)+c->sentlen, objlen - c->sentlen); @@ -1065,6 +1419,12 @@ static void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) listDelNode(c->reply,listFirst(c->reply)); c->sentlen = 0; } + /* Note that we avoid to send more thank REDIS_MAX_WRITE_PER_EVENT + * bytes, in a single threaded server it's a good idea to serve + * other clients as well, even if a very large request comes from + * super fast link that is always able to accept data (in real world + * scenario think about 'KEYS *' against the loopback interfae) */ + if (totwritten > REDIS_MAX_WRITE_PER_EVENT) break; } if (nwritten == -1) { if (errno == EAGAIN) { @@ -1083,10 +1443,88 @@ static void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) } } +static void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int mask) +{ + redisClient *c = privdata; + int nwritten = 0, totwritten = 0, objlen, willwrite; + robj *o; + struct iovec iov[REDIS_WRITEV_IOVEC_COUNT]; + int offset, ion = 0; + REDIS_NOTUSED(el); + REDIS_NOTUSED(mask); + + listNode *node; + while (listLength(c->reply)) { + offset = c->sentlen; + ion = 0; + willwrite = 0; + + /* fill-in the iov[] array */ + for(node = listFirst(c->reply); node; node = listNextNode(node)) { + o = listNodeValue(node); + objlen = sdslen(o->ptr); + + if (totwritten + objlen - offset > REDIS_MAX_WRITE_PER_EVENT) + break; + + if(ion == REDIS_WRITEV_IOVEC_COUNT) + break; /* no more iovecs */ + + iov[ion].iov_base = ((char*)o->ptr) + offset; + iov[ion].iov_len = objlen - offset; + willwrite += objlen - offset; + offset = 0; /* just for the first item */ + ion++; + } + + if(willwrite == 0) + break; + + /* write all collected blocks at once */ + if((nwritten = writev(fd, iov, ion)) < 0) { + if (errno != EAGAIN) { + redisLog(REDIS_DEBUG, + "Error writing to client: %s", strerror(errno)); + freeClient(c); + return; + } + break; + } + + totwritten += nwritten; + offset = c->sentlen; + + /* remove written robjs from c->reply */ + while (nwritten && listLength(c->reply)) { + o = listNodeValue(listFirst(c->reply)); + objlen = sdslen(o->ptr); + + if(nwritten >= objlen - offset) { + listDelNode(c->reply, listFirst(c->reply)); + nwritten -= objlen - offset; + c->sentlen = 0; + } else { + /* partial write */ + c->sentlen += nwritten; + break; + } + offset = 0; + } + } + + if (totwritten > 0) + c->lastinteraction = time(NULL); + + if (listLength(c->reply) == 0) { + c->sentlen = 0; + aeDeleteFileEvent(server.el,c->fd,AE_WRITABLE); + } +} + static struct redisCommand *lookupCommand(char *name) { int j = 0; while(cmdTable[j].name != NULL) { - if (!strcmp(name,cmdTable[j].name)) return &cmdTable[j]; + if (!strcasecmp(name,cmdTable[j].name)) return &cmdTable[j]; j++; } return NULL; @@ -1096,6 +1534,7 @@ static struct redisCommand *lookupCommand(char *name) { static void resetClient(redisClient *c) { freeClientArgv(c); c->bulklen = -1; + c->multibulk = 0; } /* If this function gets called we already read a whole @@ -1110,10 +1549,80 @@ static int processCommand(redisClient *c) { struct redisCommand *cmd; long long dirty; - sdstolower(c->argv[0]->ptr); + /* Free some memory if needed (maxmemory setting) */ + if (server.maxmemory) freeMemoryIfNeeded(); + + /* Handle the multi bulk command type. This is an alternative protocol + * supported by Redis in order to receive commands that are composed of + * multiple binary-safe "bulk" arguments. The latency of processing is + * a bit higher but this allows things like multi-sets, so if this + * protocol is used only for MSET and similar commands this is a big win. */ + if (c->multibulk == 0 && c->argc == 1 && ((char*)(c->argv[0]->ptr))[0] == '*') { + c->multibulk = atoi(((char*)c->argv[0]->ptr)+1); + if (c->multibulk <= 0) { + resetClient(c); + return 1; + } else { + decrRefCount(c->argv[c->argc-1]); + c->argc--; + return 1; + } + } else if (c->multibulk) { + if (c->bulklen == -1) { + if (((char*)c->argv[0]->ptr)[0] != '$') { + addReplySds(c,sdsnew("-ERR multi bulk protocol error\r\n")); + resetClient(c); + return 1; + } else { + int bulklen = atoi(((char*)c->argv[0]->ptr)+1); + decrRefCount(c->argv[0]); + if (bulklen < 0 || bulklen > 1024*1024*1024) { + c->argc--; + addReplySds(c,sdsnew("-ERR invalid bulk write count\r\n")); + resetClient(c); + return 1; + } + c->argc--; + c->bulklen = bulklen+2; /* add two bytes for CR+LF */ + return 1; + } + } else { + c->mbargv = zrealloc(c->mbargv,(sizeof(robj*))*(c->mbargc+1)); + c->mbargv[c->mbargc] = c->argv[0]; + c->mbargc++; + c->argc--; + c->multibulk--; + if (c->multibulk == 0) { + robj **auxargv; + int auxargc; + + /* Here we need to swap the multi-bulk argc/argv with the + * normal argc/argv of the client structure. */ + auxargv = c->argv; + c->argv = c->mbargv; + c->mbargv = auxargv; + + auxargc = c->argc; + c->argc = c->mbargc; + c->mbargc = auxargc; + + /* We need to set bulklen to something different than -1 + * in order for the code below to process the command without + * to try to read the last argument of a bulk command as + * a special argument. */ + c->bulklen = 0; + /* continue below and process the command */ + } else { + c->bulklen = -1; + return 1; + } + } + } + /* -- end of multi bulk commands processing -- */ + /* The QUIT command is handled as a special case. Normal command * procs are unable to close the client connection safely */ - if (!strcmp(c->argv[0]->ptr,"quit")) { + if (!strcasecmp(c->argv[0]->ptr,"quit")) { freeClient(c); return 0; } @@ -1127,6 +1636,10 @@ static int processCommand(redisClient *c) { addReplySds(c,sdsnew("-ERR wrong number of arguments\r\n")); resetClient(c); return 1; + } else if (server.maxmemory && cmd->flags & REDIS_CMD_DENYOOM && zmalloc_used_memory() > server.maxmemory) { + addReplySds(c,sdsnew("-ERR command not allowed when used memory > 'maxmemory'\r\n")); + resetClient(c); + return 1; } else if (cmd->flags & REDIS_CMD_BULK && c->bulklen == -1) { int bulklen = atoi(c->argv[c->argc-1]->ptr); @@ -1140,7 +1653,10 @@ static int processCommand(redisClient *c) { c->argc--; c->bulklen = bulklen+2; /* add two bytes for CR+LF */ /* It is possible that the bulk read is already in the - * buffer. Check this condition and handle it accordingly */ + * buffer. Check this condition and handle it accordingly. + * This is just a fast path, alternative to call processInputBuffer(). + * It's a good idea since the code is small and this condition + * happens most of the times. */ if ((signed)sdslen(c->querybuf) >= c->bulklen) { c->argv[c->argc] = createStringObject(c->querybuf,c->bulklen-2); c->argc++; @@ -1155,6 +1671,10 @@ static int processCommand(redisClient *c) { for(j = 1; j < c->argc; j++) c->argv[j] = tryObjectSharing(c->argv[j]); } + /* Let's try to encode the bulk object to save space. */ + if (cmd->flags & REDIS_CMD_BULK) + tryObjectEncoding(c->argv[c->argc-1]); + /* Check if the user is authenticated */ if (server.requirepass && !c->authenticated && cmd->proc != authCommand) { addReplySds(c,sdsnew("-ERR operation not permitted\r\n")); @@ -1165,7 +1685,9 @@ static int processCommand(redisClient *c) { /* Exec the command */ dirty = server.dirty; cmd->proc(c); - if (server.dirty-dirty != 0 && listLength(server.slaves)) + if (server.appendonly && server.dirty-dirty) + feedAppendOnlyFile(cmd,c->db->id,c->argv,c->argc); + if (server.dirty-dirty && listLength(server.slaves)) replicationFeedSlaves(server.slaves,cmd,c->db->id,c->argv,c->argc); if (listLength(server.monitors)) replicationFeedSlaves(server.monitors,cmd,c->db->id,c->argv,c->argc); @@ -1181,9 +1703,17 @@ static int processCommand(redisClient *c) { } static void replicationFeedSlaves(list *slaves, struct redisCommand *cmd, int dictid, robj **argv, int argc) { - listNode *ln = slaves->head; - robj *outv[REDIS_MAX_ARGS*4]; /* enough room for args, spaces, newlines */ + listNode *ln; int outc = 0, j; + robj **outv; + /* (args*2)+1 is enough room for args, spaces, newlines */ + robj *static_outv[REDIS_STATIC_ARGS*2+1]; + + if (argc <= REDIS_STATIC_ARGS) { + outv = static_outv; + } else { + outv = zmalloc(sizeof(robj*)*(argc*2+1)); + } for (j = 0; j < argc; j++) { if (j != 0) outv[outc++] = shared.space; @@ -1191,7 +1721,8 @@ static void replicationFeedSlaves(list *slaves, struct redisCommand *cmd, int di robj *lenobj; lenobj = createObject(REDIS_STRING, - sdscatprintf(sdsempty(),"%d\r\n",sdslen(argv[j]->ptr))); + sdscatprintf(sdsempty(),"%d\r\n", + stringObjectLen(argv[j]))); lenobj->refcount = 0; outv[outc++] = lenobj; } @@ -1199,8 +1730,18 @@ static void replicationFeedSlaves(list *slaves, struct redisCommand *cmd, int di } outv[outc++] = shared.crlf; - while(ln) { + /* Increment all the refcounts at start and decrement at end in order to + * be sure to free objects if there is no slave in a replication state + * able to be feed with commands */ + for (j = 0; j < outc; j++) incrRefCount(outv[j]); + listRewind(slaves); + while((ln = listYield(slaves))) { redisClient *slave = ln->value; + + /* Don't feed slaves that are still waiting for BGSAVE to start */ + if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) continue; + + /* Feed all the other slaves, MONITORs and so on */ if (slave->slaveseldb != dictid) { robj *selectcmd; @@ -1225,43 +1766,18 @@ static void replicationFeedSlaves(list *slaves, struct redisCommand *cmd, int di slave->slaveseldb = dictid; } for (j = 0; j < outc; j++) addReply(slave,outv[j]); - ln = ln->next; } + for (j = 0; j < outc; j++) decrRefCount(outv[j]); + if (outv != static_outv) zfree(outv); } -static void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask) { - redisClient *c = (redisClient*) privdata; - char buf[REDIS_QUERYBUF_LEN]; - int nread; - REDIS_NOTUSED(el); - REDIS_NOTUSED(mask); - - nread = read(fd, buf, REDIS_QUERYBUF_LEN); - if (nread == -1) { - if (errno == EAGAIN) { - nread = 0; - } else { - redisLog(REDIS_DEBUG, "Reading from client: %s",strerror(errno)); - freeClient(c); - return; - } - } else if (nread == 0) { - redisLog(REDIS_DEBUG, "Client closed connection"); - freeClient(c); - return; - } - if (nread) { - c->querybuf = sdscatlen(c->querybuf, buf, nread); - c->lastinteraction = time(NULL); - } else { - return; - } - +static void processInputBuffer(redisClient *c) { again: if (c->bulklen == -1) { /* Read the first line of the query */ char *p = strchr(c->querybuf,'\n'); size_t querylen; + if (p) { sds query, *argv; int argc, j; @@ -1285,8 +1801,11 @@ again: } argv = sdssplitlen(query,sdslen(query)," ",1,&argc); sdsfree(query); - if (argv == NULL) oom("sdssplitlen"); - for (j = 0; j < argc && j < REDIS_MAX_ARGS; j++) { + + if (c->argv) zfree(c->argv); + c->argv = zmalloc(sizeof(robj*)*argc); + + for (j = 0; j < argc; j++) { if (sdslen(argv[j])) { c->argv[c->argc] = createObject(REDIS_STRING,argv[j]); c->argc++; @@ -1298,9 +1817,9 @@ again: /* Execute the command. If the client is still valid * after processCommand() return and there is something * on the query buffer try to process the next command. */ - if (processCommand(c) && sdslen(c->querybuf)) goto again; + if (c->argc && processCommand(c) && sdslen(c->querybuf)) goto again; return; - } else if (sdslen(c->querybuf) >= 1024) { + } else if (sdslen(c->querybuf) >= REDIS_REQUEST_MAX_SIZE) { redisLog(REDIS_DEBUG, "Client protocol error"); freeClient(c); return; @@ -1317,10 +1836,43 @@ again: c->argv[c->argc] = createStringObject(c->querybuf,c->bulklen-2); c->argc++; c->querybuf = sdsrange(c->querybuf,c->bulklen,-1); - processCommand(c); + /* Process the command. If the client is still valid after + * the processing and there is more data in the buffer + * try to parse it. */ + if (processCommand(c) && sdslen(c->querybuf)) goto again; + return; + } + } +} + +static void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask) { + redisClient *c = (redisClient*) privdata; + char buf[REDIS_IOBUF_LEN]; + int nread; + REDIS_NOTUSED(el); + REDIS_NOTUSED(mask); + + nread = read(fd, buf, REDIS_IOBUF_LEN); + if (nread == -1) { + if (errno == EAGAIN) { + nread = 0; + } else { + redisLog(REDIS_DEBUG, "Reading from client: %s",strerror(errno)); + freeClient(c); return; } + } else if (nread == 0) { + redisLog(REDIS_DEBUG, "Client closed connection"); + freeClient(c); + return; } + if (nread) { + c->querybuf = sdscatlen(c->querybuf, buf, nread); + c->lastinteraction = time(NULL); + } else { + return; + } + processInputBuffer(c); } static int selectDb(redisClient *c, int id) { @@ -1330,6 +1882,11 @@ static int selectDb(redisClient *c, int id) { return REDIS_OK; } +static void *dupClientReplyValue(void *o) { + incrRefCount((robj*)o); + return 0; +} + static redisClient *createClient(int fd) { redisClient *c = zmalloc(sizeof(*c)); @@ -1340,28 +1897,40 @@ static redisClient *createClient(int fd) { c->fd = fd; c->querybuf = sdsempty(); c->argc = 0; + c->argv = NULL; c->bulklen = -1; + c->multibulk = 0; + c->mbargc = 0; + c->mbargv = NULL; c->sentlen = 0; c->flags = 0; c->lastinteraction = time(NULL); c->authenticated = 0; - if ((c->reply = listCreate()) == NULL) oom("listCreate"); + c->replstate = REDIS_REPL_NONE; + c->reply = listCreate(); listSetFreeMethod(c->reply,decrRefCount); + listSetDupMethod(c->reply,dupClientReplyValue); if (aeCreateFileEvent(server.el, c->fd, AE_READABLE, readQueryFromClient, c, NULL) == AE_ERR) { freeClient(c); return NULL; } - if (!listAddNodeTail(server.clients,c)) oom("listAddNodeTail"); + listAddNodeTail(server.clients,c); return c; } static void addReply(redisClient *c, robj *obj) { if (listLength(c->reply) == 0 && + (c->replstate == REDIS_REPL_NONE || + c->replstate == REDIS_REPL_ONLINE) && aeCreateFileEvent(server.el, c->fd, AE_WRITABLE, sendReplyToClient, c, NULL) == AE_ERR) return; - if (!listAddNodeTail(c->reply,obj)) oom("listAddNodeTail"); - incrRefCount(obj); + if (obj->encoding != REDIS_ENCODING_RAW) { + obj = getDecodedObject(obj); + } else { + incrRefCount(obj); + } + listAddNodeTail(c->reply,obj); } static void addReplySds(redisClient *c, sds s) { @@ -1370,9 +1939,30 @@ static void addReplySds(redisClient *c, sds s) { decrRefCount(o); } +static void addReplyBulkLen(redisClient *c, robj *obj) { + size_t len; + + if (obj->encoding == REDIS_ENCODING_RAW) { + len = sdslen(obj->ptr); + } else { + long n = (long)obj->ptr; + + len = 1; + if (n < 0) { + len++; + n = -n; + } + while((n = n/10) != 0) { + len++; + } + } + addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",len)); +} + static void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask) { int cport, cfd; char cip[128]; + redisClient *c; REDIS_NOTUSED(el); REDIS_NOTUSED(mask); REDIS_NOTUSED(privdata); @@ -1383,11 +1973,25 @@ static void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask) { return; } redisLog(REDIS_DEBUG,"Accepted %s:%d", cip, cport); - if (createClient(cfd) == NULL) { + if ((c = createClient(cfd)) == NULL) { redisLog(REDIS_WARNING,"Error allocating resoures for the client"); close(cfd); /* May be already closed, just ingore errors */ return; } + /* If maxclient directive is set and this is one client more... close the + * connection. Note that we create the client instead to check before + * for this condition, since now the socket is already set in nonblocking + * mode and we can send an error for free using the Kernel I/O */ + if (server.maxclients && listLength(server.clients) > server.maxclients) { + char *err = "-ERR max number of clients reached\r\n"; + + /* That's a best effort error message, don't check write errors */ + if (write(c->fd,err,strlen(err)) == -1) { + /* Nothing to do, Just to avoid the warning... */ + } + freeClient(c); + return; + } server.stat_numconnections++; } @@ -1403,8 +2007,8 @@ static robj *createObject(int type, void *ptr) { } else { o = zmalloc(sizeof(*o)); } - if (!o) oom("createObject"); o->type = type; + o->encoding = REDIS_ENCODING_RAW; o->ptr = ptr; o->refcount = 1; return o; @@ -1417,19 +2021,27 @@ static robj *createStringObject(char *ptr, size_t len) { static robj *createListObject(void) { list *l = listCreate(); - if (!l) oom("listCreate"); listSetFreeMethod(l,decrRefCount); return createObject(REDIS_LIST,l); } static robj *createSetObject(void) { dict *d = dictCreate(&setDictType,NULL); - if (!d) oom("dictCreate"); return createObject(REDIS_SET,d); } +static robj *createZsetObject(void) { + zset *zs = zmalloc(sizeof(*zs)); + + zs->dict = dictCreate(&zsetDictType,NULL); + zs->zsl = zslCreate(); + return createObject(REDIS_ZSET,zs); +} + static void freeStringObject(robj *o) { - sdsfree(o->ptr); + if (o->encoding == REDIS_ENCODING_RAW) { + sdsfree(o->ptr); + } } static void freeListObject(robj *o) { @@ -1440,21 +2052,39 @@ static void freeSetObject(robj *o) { dictRelease((dict*) o->ptr); } +static void freeZsetObject(robj *o) { + zset *zs = o->ptr; + + dictRelease(zs->dict); + zslFree(zs->zsl); + zfree(zs); +} + static void freeHashObject(robj *o) { dictRelease((dict*) o->ptr); } static void incrRefCount(robj *o) { o->refcount++; +#ifdef DEBUG_REFCOUNT + if (o->type == REDIS_STRING) + printf("Increment '%s'(%p), now is: %d\n",o->ptr,o,o->refcount); +#endif } static void decrRefCount(void *obj) { robj *o = obj; + +#ifdef DEBUG_REFCOUNT + if (o->type == REDIS_STRING) + printf("Decrement '%s'(%p), now is: %d\n",o->ptr,o,o->refcount-1); +#endif if (--(o->refcount) == 0) { switch(o->type) { case REDIS_STRING: freeStringObject(o); break; case REDIS_LIST: freeListObject(o); break; case REDIS_SET: freeSetObject(o); break; + case REDIS_ZSET: freeZsetObject(o); break; case REDIS_HASH: freeHashObject(o); break; default: assert(0 != 0); break; } @@ -1464,6 +2094,36 @@ static void decrRefCount(void *obj) { } } +static robj *lookupKey(redisDb *db, robj *key) { + dictEntry *de = dictFind(db->dict,key); + return de ? dictGetEntryVal(de) : NULL; +} + +static robj *lookupKeyRead(redisDb *db, robj *key) { + expireIfNeeded(db,key); + return lookupKey(db,key); +} + +static robj *lookupKeyWrite(redisDb *db, robj *key) { + deleteIfVolatile(db,key); + return lookupKey(db,key); +} + +static int deleteKey(redisDb *db, robj *key) { + int retval; + + /* We need to protect key from destruction: after the first dictDelete() + * it may happen that 'key' is no longer valid if we don't increment + * it's count. This may happen when we get the object reference directly + * from the hash table with dictRandomKey() or dict iterators */ + incrRefCount(key); + if (dictSize(db->expires)) dictDelete(db->expires,key); + retval = dictDelete(db->dict,key); + decrRefCount(key); + + return retval == DICT_OK; +} + /* Try to share an object against the shared objects pool */ static robj *tryObjectSharing(robj *o) { struct dictEntry *de; @@ -1509,24 +2169,106 @@ static robj *tryObjectSharing(robj *o) { } } -static robj *lookupKey(redisDb *db, robj *key) { - dictEntry *de = dictFind(db->dict,key); - return de ? dictGetEntryVal(de) : NULL; +/* Check if the nul-terminated string 's' can be represented by a long + * (that is, is a number that fits into long without any other space or + * character before or after the digits). + * + * If so, the function returns REDIS_OK and *longval is set to the value + * of the number. Otherwise REDIS_ERR is returned */ +static int isStringRepresentableAsLong(sds s, long *longval) { + char buf[32], *endptr; + long value; + int slen; + + value = strtol(s, &endptr, 10); + if (endptr[0] != '\0') return REDIS_ERR; + slen = snprintf(buf,32,"%ld",value); + + /* If the number converted back into a string is not identical + * then it's not possible to encode the string as integer */ + if (sdslen(s) != (unsigned)slen || memcmp(buf,s,slen)) return REDIS_ERR; + if (longval) *longval = value; + return REDIS_OK; } -static robj *lookupKeyRead(redisDb *db, robj *key) { - expireIfNeeded(db,key); - return lookupKey(db,key); +/* Try to encode a string object in order to save space */ +static int tryObjectEncoding(robj *o) { + long value; + sds s = o->ptr; + + if (o->encoding != REDIS_ENCODING_RAW) + return REDIS_ERR; /* Already encoded */ + + /* It's not save to encode shared objects: shared objects can be shared + * everywhere in the "object space" of Redis. Encoded objects can only + * appear as "values" (and not, for instance, as keys) */ + if (o->refcount > 1) return REDIS_ERR; + + /* Currently we try to encode only strings */ + assert(o->type == REDIS_STRING); + + /* Check if we can represent this string as a long integer */ + if (isStringRepresentableAsLong(s,&value) == REDIS_ERR) return REDIS_ERR; + + /* Ok, this object can be encoded */ + o->encoding = REDIS_ENCODING_INT; + sdsfree(o->ptr); + o->ptr = (void*) value; + return REDIS_OK; } -static robj *lookupKeyWrite(redisDb *db, robj *key) { - deleteIfVolatile(db,key); - return lookupKey(db,key); +/* Get a decoded version of an encoded object (returned as a new object) */ +static robj *getDecodedObject(const robj *o) { + robj *dec; + + assert(o->encoding != REDIS_ENCODING_RAW); + if (o->type == REDIS_STRING && o->encoding == REDIS_ENCODING_INT) { + char buf[32]; + + snprintf(buf,32,"%ld",(long)o->ptr); + dec = createStringObject(buf,strlen(buf)); + return dec; + } else { + assert(1 != 1); + } } -static int deleteKey(redisDb *db, robj *key) { - if (dictSize(db->expires)) dictDelete(db->expires,key); - return dictDelete(db->dict,key) == DICT_OK; +/* Compare two string objects via strcmp() or alike. + * Note that the objects may be integer-encoded. In such a case we + * use snprintf() to get a string representation of the numbers on the stack + * and compare the strings, it's much faster than calling getDecodedObject(). */ +static int compareStringObjects(robj *a, robj *b) { + assert(a->type == REDIS_STRING && b->type == REDIS_STRING); + char bufa[128], bufb[128], *astr, *bstr; + int bothsds = 1; + + if (a == b) return 0; + if (a->encoding != REDIS_ENCODING_RAW) { + snprintf(bufa,sizeof(bufa),"%ld",(long) a->ptr); + astr = bufa; + bothsds = 0; + } else { + astr = a->ptr; + } + if (b->encoding != REDIS_ENCODING_RAW) { + snprintf(bufb,sizeof(bufb),"%ld",(long) b->ptr); + bstr = bufb; + bothsds = 0; + } else { + bstr = b->ptr; + } + return bothsds ? sdscmp(astr,bstr) : strcmp(astr,bstr); +} + +static size_t stringObjectLen(robj *o) { + assert(o->type == REDIS_STRING); + if (o->encoding == REDIS_ENCODING_RAW) { + return sdslen(o->ptr); + } else { + char buf[32]; + + return snprintf(buf,32,"%ld",(long)o->ptr); + } } /*============================ DB saving/loading ============================ */ @@ -1536,6 +2278,12 @@ static int rdbSaveType(FILE *fp, unsigned char type) { return 0; } +static int rdbSaveTime(FILE *fp, time_t t) { + int32_t t32 = (int32_t) t; + if (fwrite(&t32,4,1,fp) == 0) return -1; + return 0; +} + /* check rdbLoadLen() comments for more info */ static int rdbSaveLen(FILE *fp, uint32_t len) { unsigned char buf[2]; @@ -1562,7 +2310,7 @@ static int rdbSaveLen(FILE *fp, uint32_t len) { /* String objects in the form "2391" "-100" without any space and with a * range of values that can fit in an 8, 16 or 32 bit signed value can be * encoded as integers to save space */ -int rdbTryIntegerEncoding(sds s, unsigned char *enc) { +static int rdbTryIntegerEncoding(sds s, unsigned char *enc) { long long value; char *endptr, buf[32]; @@ -1597,12 +2345,43 @@ int rdbTryIntegerEncoding(sds s, unsigned char *enc) { } } +static int rdbSaveLzfStringObject(FILE *fp, robj *obj) { + unsigned int comprlen, outlen; + unsigned char byte; + void *out; + + /* We require at least four bytes compression for this to be worth it */ + outlen = sdslen(obj->ptr)-4; + if (outlen <= 0) return 0; + if ((out = zmalloc(outlen+1)) == NULL) return 0; + comprlen = lzf_compress(obj->ptr, sdslen(obj->ptr), out, outlen); + if (comprlen == 0) { + zfree(out); + return 0; + } + /* Data compressed! Let's save it on disk */ + byte = (REDIS_RDB_ENCVAL<<6)|REDIS_RDB_ENC_LZF; + if (fwrite(&byte,1,1,fp) == 0) goto writeerr; + if (rdbSaveLen(fp,comprlen) == -1) goto writeerr; + if (rdbSaveLen(fp,sdslen(obj->ptr)) == -1) goto writeerr; + if (fwrite(out,comprlen,1,fp) == 0) goto writeerr; + zfree(out); + return comprlen; + +writeerr: + zfree(out); + return -1; +} + /* Save a string objet as [len][data] on disk. If the object is a string * representation of an integer value we try to safe it in a special form */ -static int rdbSaveStringObject(FILE *fp, robj *obj) { - size_t len = sdslen(obj->ptr); +static int rdbSaveStringObjectRaw(FILE *fp, robj *obj) { + size_t len; int enclen; + len = sdslen(obj->ptr); + + /* Try integer encoding */ if (len <= 11) { unsigned char buf[5]; if ((enclen = rdbTryIntegerEncoding(obj->ptr,buf)) > 0) { @@ -1610,11 +2389,66 @@ static int rdbSaveStringObject(FILE *fp, robj *obj) { return 0; } } + + /* Try LZF compression - under 20 bytes it's unable to compress even + * aaaaaaaaaaaaaaaaaa so skip it */ + if (len > 20) { + int retval; + + retval = rdbSaveLzfStringObject(fp,obj); + if (retval == -1) return -1; + if (retval > 0) return 0; + /* retval == 0 means data can't be compressed, save the old way */ + } + + /* Store verbatim */ if (rdbSaveLen(fp,len) == -1) return -1; if (len && fwrite(obj->ptr,len,1,fp) == 0) return -1; return 0; } +/* Like rdbSaveStringObjectRaw() but handle encoded objects */ +static int rdbSaveStringObject(FILE *fp, robj *obj) { + int retval; + robj *dec; + + if (obj->encoding != REDIS_ENCODING_RAW) { + dec = getDecodedObject(obj); + retval = rdbSaveStringObjectRaw(fp,dec); + decrRefCount(dec); + return retval; + } else { + return rdbSaveStringObjectRaw(fp,obj); + } +} + +/* Save a double value. Doubles are saved as strings prefixed by an unsigned + * 8 bit integer specifing the length of the representation. + * This 8 bit integer has special values in order to specify the following + * conditions: + * 253: not a number + * 254: + inf + * 255: - inf + */ +static int rdbSaveDoubleValue(FILE *fp, double val) { + unsigned char buf[128]; + int len; + + if (isnan(val)) { + buf[0] = 253; + len = 1; + } else if (!isfinite(val)) { + len = 1; + buf[0] = (val < 0) ? 255 : 254; + } else { + snprintf((char*)buf+1,sizeof(buf)-1,"%.17g",val); + buf[0] = strlen((char*)buf); + len = buf[0]+1; + } + if (fwrite(buf,len,1,fp) == 0) return -1; + return 0; +} + /* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success */ static int rdbSave(char *filename) { dictIterator *di = NULL; @@ -1622,8 +2456,9 @@ static int rdbSave(char *filename) { FILE *fp; char tmpfile[256]; int j; + time_t now = time(NULL); - snprintf(tmpfile,256,"temp-%d.%ld.rdb",(int)time(NULL),(long int)random()); + snprintf(tmpfile,256,"temp-%d.rdb", (int) getpid()); fp = fopen(tmpfile,"w"); if (!fp) { redisLog(REDIS_WARNING, "Failed saving the DB: %s", strerror(errno)); @@ -1631,7 +2466,8 @@ static int rdbSave(char *filename) { } if (fwrite("REDIS0001",9,1,fp) == 0) goto werr; for (j = 0; j < server.dbnum; j++) { - dict *d = server.db[j].dict; + redisDb *db = server.db+j; + dict *d = db->dict; if (dictSize(d) == 0) continue; di = dictGetIterator(d); if (!di) { @@ -1647,7 +2483,16 @@ static int rdbSave(char *filename) { while((de = dictNext(di)) != NULL) { robj *key = dictGetEntryKey(de); robj *o = dictGetEntryVal(de); - + time_t expiretime = getExpire(db,key); + + /* Save the expire time */ + if (expiretime != -1) { + /* If this key is already expired skip it */ + if (expiretime < now) continue; + if (rdbSaveType(fp,REDIS_EXPIRETIME) == -1) goto werr; + if (rdbSaveTime(fp,expiretime) == -1) goto werr; + } + /* Save the key and associated value */ if (rdbSaveType(fp,o->type) == -1) goto werr; if (rdbSaveStringObject(fp,key) == -1) goto werr; if (o->type == REDIS_STRING) { @@ -1656,14 +2501,14 @@ static int rdbSave(char *filename) { } else if (o->type == REDIS_LIST) { /* Save a list value */ list *list = o->ptr; - listNode *ln = list->head; + listNode *ln; + listRewind(list); if (rdbSaveLen(fp,listLength(list)) == -1) goto werr; - while(ln) { + while((ln = listYield(list))) { robj *eleobj = listNodeValue(ln); if (rdbSaveStringObject(fp,eleobj) == -1) goto werr; - ln = ln->next; } } else if (o->type == REDIS_SET) { /* Save a set value */ @@ -1671,7 +2516,6 @@ static int rdbSave(char *filename) { dictIterator *di = dictGetIterator(set); dictEntry *de; - if (!set) oom("dictGetIteraotr"); if (rdbSaveLen(fp,dictSize(set)) == -1) goto werr; while((de = dictNext(di)) != NULL) { robj *eleobj = dictGetEntryKey(de); @@ -1679,6 +2523,21 @@ static int rdbSave(char *filename) { if (rdbSaveStringObject(fp,eleobj) == -1) goto werr; } dictReleaseIterator(di); + } else if (o->type == REDIS_ZSET) { + /* Save a set value */ + zset *zs = o->ptr; + dictIterator *di = dictGetIterator(zs->dict); + dictEntry *de; + + if (rdbSaveLen(fp,dictSize(zs->dict)) == -1) goto werr; + while((de = dictNext(di)) != NULL) { + robj *eleobj = dictGetEntryKey(de); + double *score = dictGetEntryVal(de); + + if (rdbSaveStringObject(fp,eleobj) == -1) goto werr; + if (rdbSaveDoubleValue(fp,*score) == -1) goto werr; + } + dictReleaseIterator(di); } else { assert(0 != 0); } @@ -1696,7 +2555,7 @@ static int rdbSave(char *filename) { /* Use RENAME to make sure the DB file is changed atomically only * if the generate DB file is ok. */ if (rename(tmpfile,filename) == -1) { - redisLog(REDIS_WARNING,"Error moving temp DB file on the final destionation: %s", strerror(errno)); + redisLog(REDIS_WARNING,"Error moving temp DB file on the final destination: %s", strerror(errno)); unlink(tmpfile); return REDIS_ERR; } @@ -1727,19 +2586,38 @@ static int rdbSaveBackground(char *filename) { } } else { /* Parent */ + if (childpid == -1) { + redisLog(REDIS_WARNING,"Can't save in background: fork: %s", + strerror(errno)); + return REDIS_ERR; + } redisLog(REDIS_NOTICE,"Background saving started by pid %d",childpid); server.bgsaveinprogress = 1; + server.bgsavechildpid = childpid; return REDIS_OK; } return REDIS_OK; /* unreached */ } +static void rdbRemoveTempFile(pid_t childpid) { + char tmpfile[256]; + + snprintf(tmpfile,256,"temp-%d.rdb", (int) childpid); + unlink(tmpfile); +} + static int rdbLoadType(FILE *fp) { unsigned char type; if (fread(&type,1,1,fp) == 0) return -1; return type; } +static time_t rdbLoadTime(FILE *fp) { + int32_t t32; + if (fread(&t32,4,1,fp) == 0) return -1; + return (time_t) t32; +} + /* Load an encoded length from the DB, see the REDIS_RDB_* defines on the top * of this file for a description of how this are stored on disk. * @@ -1801,6 +2679,25 @@ static robj *rdbLoadIntegerObject(FILE *fp, int enctype) { return createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",val)); } +static robj *rdbLoadLzfStringObject(FILE*fp, int rdbver) { + unsigned int len, clen; + unsigned char *c = NULL; + sds val = NULL; + + if ((clen = rdbLoadLen(fp,rdbver,NULL)) == REDIS_RDB_LENERR) return NULL; + if ((len = rdbLoadLen(fp,rdbver,NULL)) == REDIS_RDB_LENERR) return NULL; + if ((c = zmalloc(clen)) == NULL) goto err; + if ((val = sdsnewlen(NULL,len)) == NULL) goto err; + if (fread(c,clen,1,fp) == 0) goto err; + if (lzf_decompress(c,clen,val,len) == 0) goto err; + zfree(c); + return createObject(REDIS_STRING,val); +err: + zfree(c); + sdsfree(val); + return NULL; +} + static robj *rdbLoadStringObject(FILE*fp, int rdbver) { int isencoded; uint32_t len; @@ -1813,6 +2710,8 @@ static robj *rdbLoadStringObject(FILE*fp, int rdbver) { case REDIS_RDB_ENC_INT16: case REDIS_RDB_ENC_INT32: return tryObjectSharing(rdbLoadIntegerObject(fp,len)); + case REDIS_RDB_ENC_LZF: + return tryObjectSharing(rdbLoadLzfStringObject(fp,rdbver)); default: assert(0!=0); } @@ -1827,15 +2726,33 @@ static robj *rdbLoadStringObject(FILE*fp, int rdbver) { return tryObjectSharing(createObject(REDIS_STRING,val)); } +/* For information about double serialization check rdbSaveDoubleValue() */ +static int rdbLoadDoubleValue(FILE *fp, double *val) { + char buf[128]; + unsigned char len; + + if (fread(&len,1,1,fp) == 0) return -1; + switch(len) { + case 255: *val = R_NegInf; return 0; + case 254: *val = R_PosInf; return 0; + case 253: *val = R_Nan; return 0; + default: + if (fread(buf,len,1,fp) == 0) return -1; + sscanf(buf, "%lg", val); + return 0; + } +} + static int rdbLoad(char *filename) { FILE *fp; robj *keyobj = NULL; uint32_t dbid; - int type; - int retval; + int type, retval, rdbver; dict *d = server.db[0].dict; + redisDb *db = server.db+0; char buf[1024]; - int rdbver; + time_t expiretime = -1, now = time(NULL); + fp = fopen(filename,"r"); if (!fp) return REDIS_ERR; if (fread(buf,9,1,fp) == 0) goto eoferr; @@ -1856,6 +2773,11 @@ static int rdbLoad(char *filename) { /* Read type. */ if ((type = rdbLoadType(fp)) == -1) goto eoferr; + if (type == REDIS_EXPIRETIME) { + if ((expiretime = rdbLoadTime(fp)) == -1) goto eoferr; + /* We read the time so we need to read the object type again */ + if ((type = rdbLoadType(fp)) == -1) goto eoferr; + } if (type == REDIS_EOF) break; /* Handle SELECT DB opcode as a special case */ if (type == REDIS_SELECTDB) { @@ -1865,7 +2787,8 @@ static int rdbLoad(char *filename) { redisLog(REDIS_WARNING,"FATAL: Data file was created with a Redis server configured to handle more than %d databases. Exiting\n", server.dbnum); exit(1); } - d = server.db[dbid].dict; + db = server.db+dbid; + d = db->dict; continue; } /* Read key */ @@ -1874,6 +2797,7 @@ static int rdbLoad(char *filename) { if (type == REDIS_STRING) { /* Read string value */ if ((o = rdbLoadStringObject(fp,rdbver)) == NULL) goto eoferr; + tryObjectEncoding(o); } else if (type == REDIS_LIST || type == REDIS_SET) { /* Read list/set value */ uint32_t listlen; @@ -1886,14 +2810,34 @@ static int rdbLoad(char *filename) { robj *ele; if ((ele = rdbLoadStringObject(fp,rdbver)) == NULL) goto eoferr; + tryObjectEncoding(ele); if (type == REDIS_LIST) { - if (!listAddNodeTail((list*)o->ptr,ele)) - oom("listAddNodeTail"); + listAddNodeTail((list*)o->ptr,ele); } else { - if (dictAdd((dict*)o->ptr,ele,NULL) == DICT_ERR) - oom("dictAdd"); + dictAdd((dict*)o->ptr,ele,NULL); } } + } else if (type == REDIS_ZSET) { + /* Read list/set value */ + uint32_t zsetlen; + zset *zs; + + if ((zsetlen = rdbLoadLen(fp,rdbver,NULL)) == REDIS_RDB_LENERR) + goto eoferr; + o = createZsetObject(); + zs = o->ptr; + /* Load every single element of the list/set */ + while(zsetlen--) { + robj *ele; + double *score = zmalloc(sizeof(double)); + + if ((ele = rdbLoadStringObject(fp,rdbver)) == NULL) goto eoferr; + tryObjectEncoding(ele); + if (rdbLoadDoubleValue(fp,score) == -1) goto eoferr; + dictAdd(zs->dict,ele,score); + zslInsert(zs->zsl,*score,ele); + incrRefCount(ele); /* added to skiplist */ + } } else { assert(0 != 0); } @@ -1903,6 +2847,13 @@ static int rdbLoad(char *filename) { redisLog(REDIS_WARNING,"Loading DB, duplicated key (%s) found! Unrecoverable error, exiting now.", keyobj->ptr); exit(1); } + /* Set the expire time if needed */ + if (expiretime != -1) { + setExpire(db,keyobj,expiretime); + /* Delete this key if already expired */ + if (expiretime < now) deleteKey(db,keyobj); + expiretime = -1; + } keyobj = o = NULL; } fclose(fp); @@ -1910,7 +2861,7 @@ static int rdbLoad(char *filename) { eoferr: /* unexpected end of file is handled here with a fatal exit */ if (keyobj) decrRefCount(keyobj); - redisLog(REDIS_WARNING,"Short read or OOM loading DB. Unrecoverable error, exiting now."); + redisLog(REDIS_WARNING,"Short read or OOM loading DB. Unrecoverable error, aborting now."); exit(1); return REDIS_ERR; /* Just to avoid warning */ } @@ -1918,12 +2869,12 @@ eoferr: /* unexpected end of file is handled here with a fatal exit */ /*================================== Commands =============================== */ static void authCommand(redisClient *c) { - if (!strcmp(c->argv[1]->ptr, server.requirepass)) { + if (!server.requirepass || !strcmp(c->argv[1]->ptr, server.requirepass)) { c->authenticated = 1; addReply(c,shared.ok); } else { c->authenticated = 0; - addReply(c,shared.err); + addReplySds(c,sdscatprintf(sdsempty(),"-ERR invalid password\r\n")); } } @@ -1932,8 +2883,7 @@ static void pingCommand(redisClient *c) { } static void echoCommand(redisClient *c) { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n", - (int)sdslen(c->argv[1]->ptr))); + addReplyBulkLen(c,c->argv[1]); addReply(c,c->argv[1]); addReply(c,shared.crlf); } @@ -1978,13 +2928,25 @@ static void getCommand(redisClient *c) { if (o->type != REDIS_STRING) { addReply(c,shared.wrongtypeerr); } else { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(o->ptr))); + addReplyBulkLen(c,o); addReply(c,o); addReply(c,shared.crlf); } } } +static void getsetCommand(redisClient *c) { + getCommand(c); + if (dictAdd(c->db->dict,c->argv[1],c->argv[2]) == DICT_ERR) { + dictReplace(c->db->dict,c->argv[1],c->argv[2]); + } else { + incrRefCount(c->argv[1]); + } + incrRefCount(c->argv[2]); + server.dirty++; + removeExpire(c->db,c->argv[1]); +} + static void mgetCommand(redisClient *c) { int j; @@ -1997,7 +2959,7 @@ static void mgetCommand(redisClient *c) { if (o->type != REDIS_STRING) { addReply(c,shared.nullbulk); } else { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(o->ptr))); + addReplyBulkLen(c,o); addReply(c,o); addReply(c,shared.crlf); } @@ -2005,7 +2967,7 @@ static void mgetCommand(redisClient *c) { } } -static void incrDecrCommand(redisClient *c, int incr) { +static void incrDecrCommand(redisClient *c, long long incr) { long long value; int retval; robj *o; @@ -2019,12 +2981,18 @@ static void incrDecrCommand(redisClient *c, int incr) { } else { char *eptr; - value = strtoll(o->ptr, &eptr, 10); + if (o->encoding == REDIS_ENCODING_RAW) + value = strtoll(o->ptr, &eptr, 10); + else if (o->encoding == REDIS_ENCODING_INT) + value = (long)o->ptr; + else + assert(1 != 1); } } value += incr; o = createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",value)); + tryObjectEncoding(o); retval = dictAdd(c->db->dict,c->argv[1],o); if (retval == DICT_ERR) { dictReplace(c->db->dict,c->argv[1],o); @@ -2047,23 +3015,36 @@ static void decrCommand(redisClient *c) { } static void incrbyCommand(redisClient *c) { - int incr = atoi(c->argv[2]->ptr); + long long incr = strtoll(c->argv[2]->ptr, NULL, 10); incrDecrCommand(c,incr); } static void decrbyCommand(redisClient *c) { - int incr = atoi(c->argv[2]->ptr); + long long incr = strtoll(c->argv[2]->ptr, NULL, 10); incrDecrCommand(c,-incr); } /* ========================= Type agnostic commands ========================= */ static void delCommand(redisClient *c) { - if (deleteKey(c->db,c->argv[1])) { - server.dirty++; - addReply(c,shared.cone); - } else { + int deleted = 0, j; + + for (j = 1; j < c->argc; j++) { + if (deleteKey(c->db,c->argv[j])) { + server.dirty++; + deleted++; + } + } + switch(deleted) { + case 0: addReply(c,shared.czero); + break; + case 1: + addReply(c,shared.cone); + break; + default: + addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",deleted)); + break; } } @@ -2075,7 +3056,7 @@ static void selectCommand(redisClient *c) { int id = atoi(c->argv[1]->ptr); if (selectDb(c,id) == REDIS_ERR) { - addReplySds(c,"-ERR invalid DB index\r\n"); + addReplySds(c,sdsnew("-ERR invalid DB index\r\n")); } else { addReply(c,shared.ok); } @@ -2086,9 +3067,10 @@ static void randomkeyCommand(redisClient *c) { while(1) { de = dictGetRandomKey(c->db->dict); - if (expireIfNeeded(c->db,dictGetEntryKey(de)) == 0) break; + if (!de || expireIfNeeded(c->db,dictGetEntryKey(de)) == 0) break; } if (de == NULL) { + addReply(c,shared.plus); addReply(c,shared.crlf); } else { addReply(c,shared.plus); @@ -2106,7 +3088,6 @@ static void keysCommand(redisClient *c) { robj *lenobj = createObject(REDIS_STRING,NULL); di = dictGetIterator(c->db->dict); - if (!di) oom("dictGetIterator"); addReply(c,lenobj); decrRefCount(lenobj); while((de = dictNext(di)) != NULL) { @@ -2151,6 +3132,7 @@ static void typeCommand(redisClient *c) { case REDIS_STRING: type = "+string"; break; case REDIS_LIST: type = "+list"; break; case REDIS_SET: type = "+set"; break; + case REDIS_ZSET: type = "+zset"; break; default: type = "unknown"; break; } } @@ -2159,6 +3141,10 @@ static void typeCommand(redisClient *c) { } static void saveCommand(redisClient *c) { + if (server.bgsaveinprogress) { + addReplySds(c,sdsnew("-ERR background save in progress\r\n")); + return; + } if (rdbSave(server.dbfilename) == REDIS_OK) { addReply(c,shared.ok); } else { @@ -2180,13 +3166,26 @@ static void bgsaveCommand(redisClient *c) { static void shutdownCommand(redisClient *c) { redisLog(REDIS_WARNING,"User requested shutdown, saving DB..."); + /* Kill the saving child if there is a background saving in progress. + We want to avoid race conditions, for instance our saving child may + overwrite the synchronous saving did by SHUTDOWN. */ + if (server.bgsaveinprogress) { + redisLog(REDIS_WARNING,"There is a live saving child. Killing it!"); + kill(server.bgsavechildpid,SIGKILL); + rdbRemoveTempFile(server.bgsavechildpid); + } + /* SYNC SAVE */ if (rdbSave(server.dbfilename) == REDIS_OK) { - if (server.daemonize) { - unlink(server.pidfile); - } + if (server.daemonize) + unlink(server.pidfile); + redisLog(REDIS_WARNING,"%zu bytes used at exit",zmalloc_used_memory()); redisLog(REDIS_WARNING,"Server exit now, bye bye..."); exit(1); } else { + /* Ooops.. error saving! The best we can do is to continue operating. + * Note that if there was a background saving process, in the next + * cron() Redis will be notified that the background saving aborted, + * handling special stuff like slaves pending for synchronization... */ redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit"); addReplySds(c,sdsnew("-ERR can't quit, problems saving the DB\r\n")); } @@ -2285,9 +3284,9 @@ static void pushGenericCommand(redisClient *c, int where) { lobj = createListObject(); list = lobj->ptr; if (where == REDIS_HEAD) { - if (!listAddNodeHead(list,c->argv[2])) oom("listAddNodeHead"); + listAddNodeHead(list,c->argv[2]); } else { - if (!listAddNodeTail(list,c->argv[2])) oom("listAddNodeTail"); + listAddNodeTail(list,c->argv[2]); } dictAdd(c->db->dict,c->argv[1],lobj); incrRefCount(c->argv[1]); @@ -2299,9 +3298,9 @@ static void pushGenericCommand(redisClient *c, int where) { } list = lobj->ptr; if (where == REDIS_HEAD) { - if (!listAddNodeHead(list,c->argv[2])) oom("listAddNodeHead"); + listAddNodeHead(list,c->argv[2]); } else { - if (!listAddNodeTail(list,c->argv[2])) oom("listAddNodeTail"); + listAddNodeTail(list,c->argv[2]); } incrRefCount(c->argv[2]); } @@ -2354,7 +3353,7 @@ static void lindexCommand(redisClient *c) { addReply(c,shared.nullbulk); } else { robj *ele = listNodeValue(ln); - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(ele->ptr))); + addReplyBulkLen(c,ele); addReply(c,ele); addReply(c,shared.crlf); } @@ -2414,7 +3413,7 @@ static void popGenericCommand(redisClient *c, int where) { addReply(c,shared.nullbulk); } else { robj *ele = listNodeValue(ln); - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(ele->ptr))); + addReplyBulkLen(c,ele); addReply(c,ele); addReply(c,shared.crlf); listDelNode(list,ln); @@ -2470,7 +3469,7 @@ static void lrangeCommand(redisClient *c) { addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",rangelen)); for (j = 0; j < rangelen; j++) { ele = listNodeValue(ln); - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(ele->ptr))); + addReplyBulkLen(c,ele); addReply(c,ele); addReply(c,shared.crlf); ln = ln->next; @@ -2522,8 +3521,8 @@ static void ltrimCommand(redisClient *c) { ln = listLast(list); listDelNode(list,ln); } - addReply(c,shared.ok); server.dirty++; + addReply(c,shared.ok); } } } @@ -2533,7 +3532,7 @@ static void lremCommand(redisClient *c) { o = lookupKeyWrite(c->db,c->argv[1]); if (o == NULL) { - addReply(c,shared.nokeyerr); + addReply(c,shared.czero); } else { if (o->type != REDIS_LIST) { addReply(c,shared.wrongtypeerr); @@ -2553,7 +3552,7 @@ static void lremCommand(redisClient *c) { robj *ele = listNodeValue(ln); next = fromtail ? ln->prev : ln->next; - if (sdscmp(ele->ptr,c->argv[3]->ptr) == 0) { + if (compareStringObjects(ele,c->argv[3]) == 0) { listDelNode(list,ln); server.dirty++; removed++; @@ -2566,6 +3565,70 @@ static void lremCommand(redisClient *c) { } } +/* This is the semantic of this command: + * RPOPLPUSH srclist dstlist: + * IF LLEN(srclist) > 0 + * element = RPOP srclist + * LPUSH dstlist element + * RETURN element + * ELSE + * RETURN nil + * END + * END + * + * The idea is to be able to get an element from a list in a reliable way + * since the element is not just returned but pushed against another list + * as well. This command was originally proposed by Ezra Zygmuntowicz. + */ +static void rpoplpushcommand(redisClient *c) { + robj *sobj; + + sobj = lookupKeyWrite(c->db,c->argv[1]); + if (sobj == NULL) { + addReply(c,shared.nullbulk); + } else { + if (sobj->type != REDIS_LIST) { + addReply(c,shared.wrongtypeerr); + } else { + list *srclist = sobj->ptr; + listNode *ln = listLast(srclist); + + if (ln == NULL) { + addReply(c,shared.nullbulk); + } else { + robj *dobj = lookupKeyWrite(c->db,c->argv[2]); + robj *ele = listNodeValue(ln); + list *dstlist; + + if (dobj == NULL) { + + /* Create the list if the key does not exist */ + dobj = createListObject(); + dictAdd(c->db->dict,c->argv[2],dobj); + incrRefCount(c->argv[2]); + } else if (dobj->type != REDIS_LIST) { + addReply(c,shared.wrongtypeerr); + return; + } + /* Add the element to the target list */ + dstlist = dobj->ptr; + listAddNodeHead(dstlist,ele); + incrRefCount(ele); + + /* Send the element to the client as reply as well */ + addReplyBulkLen(c,ele); + addReply(c,ele); + addReply(c,shared.crlf); + + /* Finally remove the element from the source list */ + listDelNode(srclist,ln); + server.dirty++; + } + } + } +} + + /* ==================================== Sets ================================ */ static void saddCommand(redisClient *c) { @@ -2604,6 +3667,7 @@ static void sremCommand(redisClient *c) { } if (dictDelete(set->ptr,c->argv[2]) == DICT_OK) { server.dirty++; + if (htNeedsResize(set->ptr)) dictResize(set->ptr); addReply(c,shared.cone); } else { addReply(c,shared.czero); @@ -2611,6 +3675,41 @@ static void sremCommand(redisClient *c) { } } +static void smoveCommand(redisClient *c) { + robj *srcset, *dstset; + + srcset = lookupKeyWrite(c->db,c->argv[1]); + dstset = lookupKeyWrite(c->db,c->argv[2]); + + /* If the source key does not exist return 0, if it's of the wrong type + * raise an error */ + if (srcset == NULL || srcset->type != REDIS_SET) { + addReply(c, srcset ? shared.wrongtypeerr : shared.czero); + return; + } + /* Error if the destination key is not a set as well */ + if (dstset && dstset->type != REDIS_SET) { + addReply(c,shared.wrongtypeerr); + return; + } + /* Remove the element from the source set */ + if (dictDelete(srcset->ptr,c->argv[3]) == DICT_ERR) { + /* Key not found in the src set! return zero */ + addReply(c,shared.czero); + return; + } + server.dirty++; + /* Add the element to the destination set */ + if (!dstset) { + dstset = createSetObject(); + dictAdd(c->db->dict,c->argv[2],dstset); + incrRefCount(c->argv[2]); + } + if (dictAdd(dstset->ptr,c->argv[3],NULL) == DICT_OK) + incrRefCount(c->argv[3]); + addReply(c,shared.cone); +} + static void sismemberCommand(redisClient *c) { robj *set; @@ -2648,6 +3747,59 @@ static void scardCommand(redisClient *c) { } } +static void spopCommand(redisClient *c) { + robj *set; + dictEntry *de; + + set = lookupKeyWrite(c->db,c->argv[1]); + if (set == NULL) { + addReply(c,shared.nullbulk); + } else { + if (set->type != REDIS_SET) { + addReply(c,shared.wrongtypeerr); + return; + } + de = dictGetRandomKey(set->ptr); + if (de == NULL) { + addReply(c,shared.nullbulk); + } else { + robj *ele = dictGetEntryKey(de); + + addReplyBulkLen(c,ele); + addReply(c,ele); + addReply(c,shared.crlf); + dictDelete(set->ptr,ele); + if (htNeedsResize(set->ptr)) dictResize(set->ptr); + server.dirty++; + } + } +} + +static void srandmemberCommand(redisClient *c) { + robj *set; + dictEntry *de; + + set = lookupKeyRead(c->db,c->argv[1]); + if (set == NULL) { + addReply(c,shared.nullbulk); + } else { + if (set->type != REDIS_SET) { + addReply(c,shared.wrongtypeerr); + return; + } + de = dictGetRandomKey(set->ptr); + if (de == NULL) { + addReply(c,shared.nullbulk); + } else { + robj *ele = dictGetEntryKey(de); + + addReplyBulkLen(c,ele); + addReply(c,ele); + addReply(c,shared.crlf); + } + } +} + static int qsortCompareSetsByCardinality(const void *s1, const void *s2) { dict **d1 = (void*) s1, **d2 = (void*) s2; @@ -2661,7 +3813,6 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, int setsnum, r robj *lenobj = NULL, *dstset = NULL; int j, cardinality = 0; - if (!dv) oom("sinterCommand"); for (j = 0; j < setsnum; j++) { robj *setobj; @@ -2670,7 +3821,12 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, int setsnum, r lookupKeyRead(c->db,setskeys[j]); if (!setobj) { zfree(dv); - addReply(c,shared.nokeyerr); + if (dstkey) { + deleteKey(c->db,dstkey); + addReply(c,shared.ok); + } else { + addReply(c,shared.nullmultibulk); + } return; } if (setobj->type != REDIS_SET) { @@ -2697,16 +3853,12 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, int setsnum, r /* If we have a target key where to store the resulting set * create this key with an empty set inside */ dstset = createSetObject(); - deleteKey(c->db,dstkey); - dictAdd(c->db->dict,dstkey,dstset); - incrRefCount(dstkey); } /* Iterate all the elements of the first (smallest) set, and test * the element against all the other sets, if at least one set does * not include the element it is discarded */ di = dictGetIterator(dv[0]); - if (!di) oom("dictGetIterator"); while((de = dictNext(di)) != NULL) { robj *ele; @@ -2717,7 +3869,7 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, int setsnum, r continue; /* at least one set does not contain the member */ ele = dictGetEntryKey(de); if (!dstkey) { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",sdslen(ele->ptr))); + addReplyBulkLen(c,ele); addReply(c,ele); addReply(c,shared.crlf); cardinality++; @@ -2728,10 +3880,20 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, int setsnum, r } dictReleaseIterator(di); - if (!dstkey) - lenobj->ptr = sdscatprintf(sdsempty(),"*%d\r\n",cardinality); - else - addReply(c,shared.ok); + if (dstkey) { + /* Store the resulting set into the target */ + deleteKey(c->db,dstkey); + dictAdd(c->db->dict,dstkey,dstset); + incrRefCount(dstkey); + } + + if (!dstkey) { + lenobj->ptr = sdscatprintf(sdsempty(),"*%d\r\n",cardinality); + } else { + addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n", + dictSize((dict*)dstset->ptr))); + server.dirty++; + } zfree(dv); } @@ -2743,22 +3905,608 @@ static void sinterstoreCommand(redisClient *c) { sinterGenericCommand(c,c->argv+2,c->argc-2,c->argv[1]); } +#define REDIS_OP_UNION 0 +#define REDIS_OP_DIFF 1 + +static void sunionDiffGenericCommand(redisClient *c, robj **setskeys, int setsnum, robj *dstkey, int op) { + dict **dv = zmalloc(sizeof(dict*)*setsnum); + dictIterator *di; + dictEntry *de; + robj *dstset = NULL; + int j, cardinality = 0; + + for (j = 0; j < setsnum; j++) { + robj *setobj; + + setobj = dstkey ? + lookupKeyWrite(c->db,setskeys[j]) : + lookupKeyRead(c->db,setskeys[j]); + if (!setobj) { + dv[j] = NULL; + continue; + } + if (setobj->type != REDIS_SET) { + zfree(dv); + addReply(c,shared.wrongtypeerr); + return; + } + dv[j] = setobj->ptr; + } + + /* We need a temp set object to store our union. If the dstkey + * is not NULL (that is, we are inside an SUNIONSTORE operation) then + * this set object will be the resulting object to set into the target key*/ + dstset = createSetObject(); + + /* Iterate all the elements of all the sets, add every element a single + * time to the result set */ + for (j = 0; j < setsnum; j++) { + if (op == REDIS_OP_DIFF && j == 0 && !dv[j]) break; /* result set is empty */ + if (!dv[j]) continue; /* non existing keys are like empty sets */ + + di = dictGetIterator(dv[j]); + + while((de = dictNext(di)) != NULL) { + robj *ele; + + /* dictAdd will not add the same element multiple times */ + ele = dictGetEntryKey(de); + if (op == REDIS_OP_UNION || j == 0) { + if (dictAdd(dstset->ptr,ele,NULL) == DICT_OK) { + incrRefCount(ele); + cardinality++; + } + } else if (op == REDIS_OP_DIFF) { + if (dictDelete(dstset->ptr,ele) == DICT_OK) { + cardinality--; + } + } + } + dictReleaseIterator(di); + + if (op == REDIS_OP_DIFF && cardinality == 0) break; /* result set is empty */ + } + + /* Output the content of the resulting set, if not in STORE mode */ + if (!dstkey) { + addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",cardinality)); + di = dictGetIterator(dstset->ptr); + while((de = dictNext(di)) != NULL) { + robj *ele; + + ele = dictGetEntryKey(de); + addReplyBulkLen(c,ele); + addReply(c,ele); + addReply(c,shared.crlf); + } + dictReleaseIterator(di); + } else { + /* If we have a target key where to store the resulting set + * create this key with the result set inside */ + deleteKey(c->db,dstkey); + dictAdd(c->db->dict,dstkey,dstset); + incrRefCount(dstkey); + } + + /* Cleanup */ + if (!dstkey) { + decrRefCount(dstset); + } else { + addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n", + dictSize((dict*)dstset->ptr))); + server.dirty++; + } + zfree(dv); +} + +static void sunionCommand(redisClient *c) { + sunionDiffGenericCommand(c,c->argv+1,c->argc-1,NULL,REDIS_OP_UNION); +} + +static void sunionstoreCommand(redisClient *c) { + sunionDiffGenericCommand(c,c->argv+2,c->argc-2,c->argv[1],REDIS_OP_UNION); +} + +static void sdiffCommand(redisClient *c) { + sunionDiffGenericCommand(c,c->argv+1,c->argc-1,NULL,REDIS_OP_DIFF); +} + +static void sdiffstoreCommand(redisClient *c) { + sunionDiffGenericCommand(c,c->argv+2,c->argc-2,c->argv[1],REDIS_OP_DIFF); +} + +/* ==================================== ZSets =============================== */ + +/* ZSETs are ordered sets using two data structures to hold the same elements + * in order to get O(log(N)) INSERT and REMOVE operations into a sorted + * data structure. + * + * The elements are added to an hash table mapping Redis objects to scores. + * At the same time the elements are added to a skip list mapping scores + * to Redis objects (so objects are sorted by scores in this "view"). */ + +/* This skiplist implementation is almost a C translation of the original + * algorithm described by William Pugh in "Skip Lists: A Probabilistic + * Alternative to Balanced Trees", modified in three ways: + * a) this implementation allows for repeated values. + * b) the comparison is not just by key (our 'score') but by satellite data. + * c) there is a back pointer, so it's a doubly linked list with the back + * pointers being only at "level 1". This allows to traverse the list + * from tail to head, useful for ZREVRANGE. */ + +static zskiplistNode *zslCreateNode(int level, double score, robj *obj) { + zskiplistNode *zn = zmalloc(sizeof(*zn)); + + zn->forward = zmalloc(sizeof(zskiplistNode*) * level); + zn->score = score; + zn->obj = obj; + return zn; +} + +static zskiplist *zslCreate(void) { + int j; + zskiplist *zsl; + + zsl = zmalloc(sizeof(*zsl)); + zsl->level = 1; + zsl->length = 0; + zsl->header = zslCreateNode(ZSKIPLIST_MAXLEVEL,0,NULL); + for (j = 0; j < ZSKIPLIST_MAXLEVEL; j++) + zsl->header->forward[j] = NULL; + zsl->header->backward = NULL; + zsl->tail = NULL; + return zsl; +} + +static void zslFreeNode(zskiplistNode *node) { + decrRefCount(node->obj); + zfree(node->forward); + zfree(node); +} + +static void zslFree(zskiplist *zsl) { + zskiplistNode *node = zsl->header->forward[0], *next; + + zfree(zsl->header->forward); + zfree(zsl->header); + while(node) { + next = node->forward[0]; + zslFreeNode(node); + node = next; + } + zfree(zsl); +} + +static int zslRandomLevel(void) { + int level = 1; + while ((random()&0xFFFF) < (ZSKIPLIST_P * 0xFFFF)) + level += 1; + return level; +} + +static void zslInsert(zskiplist *zsl, double score, robj *obj) { + zskiplistNode *update[ZSKIPLIST_MAXLEVEL], *x; + int i, level; + + x = zsl->header; + for (i = zsl->level-1; i >= 0; i--) { + while (x->forward[i] && + (x->forward[i]->score < score || + (x->forward[i]->score == score && + compareStringObjects(x->forward[i]->obj,obj) < 0))) + x = x->forward[i]; + update[i] = x; + } + /* we assume the key is not already inside, since we allow duplicated + * scores, and the re-insertion of score and redis object should never + * happpen since the caller of zslInsert() should test in the hash table + * if the element is already inside or not. */ + level = zslRandomLevel(); + if (level > zsl->level) { + for (i = zsl->level; i < level; i++) + update[i] = zsl->header; + zsl->level = level; + } + x = zslCreateNode(level,score,obj); + for (i = 0; i < level; i++) { + x->forward[i] = update[i]->forward[i]; + update[i]->forward[i] = x; + } + x->backward = (update[0] == zsl->header) ? NULL : update[0]; + if (x->forward[0]) + x->forward[0]->backward = x; + else + zsl->tail = x; + zsl->length++; +} + +/* Delete an element with matching score/object from the skiplist. */ +static int zslDelete(zskiplist *zsl, double score, robj *obj) { + zskiplistNode *update[ZSKIPLIST_MAXLEVEL], *x; + int i; + + x = zsl->header; + for (i = zsl->level-1; i >= 0; i--) { + while (x->forward[i] && + (x->forward[i]->score < score || + (x->forward[i]->score == score && + compareStringObjects(x->forward[i]->obj,obj) < 0))) + x = x->forward[i]; + update[i] = x; + } + /* We may have multiple elements with the same score, what we need + * is to find the element with both the right score and object. */ + x = x->forward[0]; + if (x && score == x->score && compareStringObjects(x->obj,obj) == 0) { + for (i = 0; i < zsl->level; i++) { + if (update[i]->forward[i] != x) break; + update[i]->forward[i] = x->forward[i]; + } + if (x->forward[0]) { + x->forward[0]->backward = (x->backward == zsl->header) ? + NULL : x->backward; + } else { + zsl->tail = x->backward; + } + zslFreeNode(x); + while(zsl->level > 1 && zsl->header->forward[zsl->level-1] == NULL) + zsl->level--; + zsl->length--; + return 1; + } else { + return 0; /* not found */ + } + return 0; /* not found */ +} + +/* Delete all the elements with score between min and max from the skiplist. + * Min and mx are inclusive, so a score >= min || score <= max is deleted. + * Note that this function takes the reference to the hash table view of the + * sorted set, in order to remove the elements from the hash table too. */ +static unsigned long zslDeleteRange(zskiplist *zsl, double min, double max, dict *dict) { + zskiplistNode *update[ZSKIPLIST_MAXLEVEL], *x; + unsigned long removed = 0; + int i; + + x = zsl->header; + for (i = zsl->level-1; i >= 0; i--) { + while (x->forward[i] && x->forward[i]->score < min) + x = x->forward[i]; + update[i] = x; + } + /* We may have multiple elements with the same score, what we need + * is to find the element with both the right score and object. */ + x = x->forward[0]; + while (x && x->score <= max) { + zskiplistNode *next; + + for (i = 0; i < zsl->level; i++) { + if (update[i]->forward[i] != x) break; + update[i]->forward[i] = x->forward[i]; + } + if (x->forward[0]) { + x->forward[0]->backward = (x->backward == zsl->header) ? + NULL : x->backward; + } else { + zsl->tail = x->backward; + } + next = x->forward[0]; + dictDelete(dict,x->obj); + zslFreeNode(x); + while(zsl->level > 1 && zsl->header->forward[zsl->level-1] == NULL) + zsl->level--; + zsl->length--; + removed++; + x = next; + } + return removed; /* not found */ +} + +/* Find the first node having a score equal or greater than the specified one. + * Returns NULL if there is no match. */ +static zskiplistNode *zslFirstWithScore(zskiplist *zsl, double score) { + zskiplistNode *x; + int i; + + x = zsl->header; + for (i = zsl->level-1; i >= 0; i--) { + while (x->forward[i] && x->forward[i]->score < score) + x = x->forward[i]; + } + /* We may have multiple elements with the same score, what we need + * is to find the element with both the right score and object. */ + return x->forward[0]; +} + +/* The actual Z-commands implementations */ + +static void zaddCommand(redisClient *c) { + robj *zsetobj; + zset *zs; + double *score; + + zsetobj = lookupKeyWrite(c->db,c->argv[1]); + if (zsetobj == NULL) { + zsetobj = createZsetObject(); + dictAdd(c->db->dict,c->argv[1],zsetobj); + incrRefCount(c->argv[1]); + } else { + if (zsetobj->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + return; + } + } + score = zmalloc(sizeof(double)); + *score = strtod(c->argv[2]->ptr,NULL); + zs = zsetobj->ptr; + if (dictAdd(zs->dict,c->argv[3],score) == DICT_OK) { + /* case 1: New element */ + incrRefCount(c->argv[3]); /* added to hash */ + zslInsert(zs->zsl,*score,c->argv[3]); + incrRefCount(c->argv[3]); /* added to skiplist */ + server.dirty++; + addReply(c,shared.cone); + } else { + dictEntry *de; + double *oldscore; + + /* case 2: Score update operation */ + de = dictFind(zs->dict,c->argv[3]); + assert(de != NULL); + oldscore = dictGetEntryVal(de); + if (*score != *oldscore) { + int deleted; + + deleted = zslDelete(zs->zsl,*oldscore,c->argv[3]); + assert(deleted != 0); + zslInsert(zs->zsl,*score,c->argv[3]); + incrRefCount(c->argv[3]); + dictReplace(zs->dict,c->argv[3],score); + server.dirty++; + } else { + zfree(score); + } + addReply(c,shared.czero); + } +} + +static void zremCommand(redisClient *c) { + robj *zsetobj; + zset *zs; + + zsetobj = lookupKeyWrite(c->db,c->argv[1]); + if (zsetobj == NULL) { + addReply(c,shared.czero); + } else { + dictEntry *de; + double *oldscore; + int deleted; + + if (zsetobj->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + return; + } + zs = zsetobj->ptr; + de = dictFind(zs->dict,c->argv[2]); + if (de == NULL) { + addReply(c,shared.czero); + return; + } + /* Delete from the skiplist */ + oldscore = dictGetEntryVal(de); + deleted = zslDelete(zs->zsl,*oldscore,c->argv[2]); + assert(deleted != 0); + + /* Delete from the hash table */ + dictDelete(zs->dict,c->argv[2]); + if (htNeedsResize(zs->dict)) dictResize(zs->dict); + server.dirty++; + addReply(c,shared.cone); + } +} + +static void zremrangebyscoreCommand(redisClient *c) { + double min = strtod(c->argv[2]->ptr,NULL); + double max = strtod(c->argv[3]->ptr,NULL); + robj *zsetobj; + zset *zs; + + zsetobj = lookupKeyWrite(c->db,c->argv[1]); + if (zsetobj == NULL) { + addReply(c,shared.czero); + } else { + long deleted; + + if (zsetobj->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + return; + } + zs = zsetobj->ptr; + deleted = zslDeleteRange(zs->zsl,min,max,zs->dict); + if (htNeedsResize(zs->dict)) dictResize(zs->dict); + server.dirty += deleted; + addReplySds(c,sdscatprintf(sdsempty(),":%lu\r\n",deleted)); + } +} + +static void zrangeGenericCommand(redisClient *c, int reverse) { + robj *o; + int start = atoi(c->argv[2]->ptr); + int end = atoi(c->argv[3]->ptr); + + o = lookupKeyRead(c->db,c->argv[1]); + if (o == NULL) { + addReply(c,shared.nullmultibulk); + } else { + if (o->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + } else { + zset *zsetobj = o->ptr; + zskiplist *zsl = zsetobj->zsl; + zskiplistNode *ln; + + int llen = zsl->length; + int rangelen, j; + robj *ele; + + /* convert negative indexes */ + if (start < 0) start = llen+start; + if (end < 0) end = llen+end; + if (start < 0) start = 0; + if (end < 0) end = 0; + + /* indexes sanity checks */ + if (start > end || start >= llen) { + /* Out of range start or start > end result in empty list */ + addReply(c,shared.emptymultibulk); + return; + } + if (end >= llen) end = llen-1; + rangelen = (end-start)+1; + + /* Return the result in form of a multi-bulk reply */ + if (reverse) { + ln = zsl->tail; + while (start--) + ln = ln->backward; + } else { + ln = zsl->header->forward[0]; + while (start--) + ln = ln->forward[0]; + } + + addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",rangelen)); + for (j = 0; j < rangelen; j++) { + ele = ln->obj; + addReplyBulkLen(c,ele); + addReply(c,ele); + addReply(c,shared.crlf); + ln = reverse ? ln->backward : ln->forward[0]; + } + } + } +} + +static void zrangeCommand(redisClient *c) { + zrangeGenericCommand(c,0); +} + +static void zrevrangeCommand(redisClient *c) { + zrangeGenericCommand(c,1); +} + +static void zrangebyscoreCommand(redisClient *c) { + robj *o; + double min = strtod(c->argv[2]->ptr,NULL); + double max = strtod(c->argv[3]->ptr,NULL); + + o = lookupKeyRead(c->db,c->argv[1]); + if (o == NULL) { + addReply(c,shared.nullmultibulk); + } else { + if (o->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + } else { + zset *zsetobj = o->ptr; + zskiplist *zsl = zsetobj->zsl; + zskiplistNode *ln; + robj *ele, *lenobj; + unsigned int rangelen = 0; + + /* Get the first node with the score >= min */ + ln = zslFirstWithScore(zsl,min); + if (ln == NULL) { + /* No element matching the speciifed interval */ + addReply(c,shared.emptymultibulk); + return; + } + + /* We don't know in advance how many matching elements there + * are in the list, so we push this object that will represent + * the multi-bulk length in the output buffer, and will "fix" + * it later */ + lenobj = createObject(REDIS_STRING,NULL); + addReply(c,lenobj); + + while(ln && ln->score <= max) { + ele = ln->obj; + addReplyBulkLen(c,ele); + addReply(c,ele); + addReply(c,shared.crlf); + ln = ln->forward[0]; + rangelen++; + } + lenobj->ptr = sdscatprintf(sdsempty(),"*%d\r\n",rangelen); + } + } +} + +static void zcardCommand(redisClient *c) { + robj *o; + zset *zs; + + o = lookupKeyRead(c->db,c->argv[1]); + if (o == NULL) { + addReply(c,shared.czero); + return; + } else { + if (o->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + } else { + zs = o->ptr; + addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",zs->zsl->length)); + } + } +} + +static void zscoreCommand(redisClient *c) { + robj *o; + zset *zs; + + o = lookupKeyRead(c->db,c->argv[1]); + if (o == NULL) { + addReply(c,shared.nullbulk); + return; + } else { + if (o->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + } else { + dictEntry *de; + + zs = o->ptr; + de = dictFind(zs->dict,c->argv[2]); + if (!de) { + addReply(c,shared.nullbulk); + } else { + char buf[128]; + double *score = dictGetEntryVal(de); + + snprintf(buf,sizeof(buf),"%.17g",*score); + addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n%s\r\n", + strlen(buf),buf)); + } + } + } +} + +/* ========================= Non type-specific commands ==================== */ + static void flushdbCommand(redisClient *c) { + server.dirty += dictSize(c->db->dict); dictEmpty(c->db->dict); dictEmpty(c->db->expires); addReply(c,shared.ok); - rdbSave(server.dbfilename); } static void flushallCommand(redisClient *c) { - emptyDb(); + server.dirty += emptyDb(); addReply(c,shared.ok); rdbSave(server.dbfilename); + server.dirty++; } -redisSortOperation *createSortOperation(int type, robj *pattern) { +static redisSortOperation *createSortOperation(int type, robj *pattern) { redisSortOperation *so = zmalloc(sizeof(*so)); - if (!so) oom("createSortOperation"); so->type = type; so->pattern = pattern; return so; @@ -2766,7 +4514,7 @@ redisSortOperation *createSortOperation(int type, robj *pattern) { /* Return the value associated to the key with a name obtained * substituting the first occurence of '*' in 'pattern' with 'subst' */ -robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { +static robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { char *p; sds spat, ssub; robj keyobj; @@ -2778,6 +4526,12 @@ robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { char buf[REDIS_SORTKEY_MAX+1]; } keyname; + if (subst->encoding == REDIS_ENCODING_RAW) + incrRefCount(subst); + else { + subst = getDecodedObject(subst); + } + spat = pattern->ptr; ssub = subst->ptr; if (sdslen(spat)+sdslen(ssub)-1 > REDIS_SORTKEY_MAX) return NULL; @@ -2797,6 +4551,8 @@ robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { keyobj.type = REDIS_STRING; keyobj.ptr = ((char*)&keyname)+(sizeof(long)*2); + decrRefCount(subst); + /* printf("lookup '%s' => %p\n", keyname.buf,de); */ return lookupKeyRead(db,&keyobj); } @@ -2834,7 +4590,20 @@ static int sortCompare(const void *s1, const void *s2) { } } else { /* Compare elements directly */ - cmp = strcoll(so1->obj->ptr,so2->obj->ptr); + if (so1->obj->encoding == REDIS_ENCODING_RAW && + so2->obj->encoding == REDIS_ENCODING_RAW) { + cmp = strcoll(so1->obj->ptr,so2->obj->ptr); + } else { + robj *dec1, *dec2; + + dec1 = so1->obj->encoding == REDIS_ENCODING_RAW ? + so1->obj : getDecodedObject(so1->obj); + dec2 = so2->obj->encoding == REDIS_ENCODING_RAW ? + so2->obj : getDecodedObject(so2->obj); + cmp = strcoll(dec1->ptr,dec2->ptr); + if (dec1 != so1->obj) decrRefCount(dec1); + if (dec2 != so2->obj) decrRefCount(dec2); + } } } return server.sort_desc ? -cmp : cmp; @@ -2849,7 +4618,7 @@ static void sortCommand(redisClient *c) { int limit_start = 0, limit_count = -1, start, end; int j, dontsort = 0, vectorlen; int getop = 0; /* GET operation counter */ - robj *sortval, *sortby = NULL; + robj *sortval, *sortby = NULL, *storekey = NULL; redisSortObject *vector; /* Resulting vector to sort */ /* Lookup the key to sort. It must be of the right types */ @@ -2887,6 +4656,9 @@ static void sortCommand(redisClient *c) { limit_start = atoi(c->argv[j+1]->ptr); limit_count = atoi(c->argv[j+2]->ptr); j+=2; + } else if (!strcasecmp(c->argv[j]->ptr,"store") && leftargs >= 1) { + storekey = c->argv[j+1]; + j++; } else if (!strcasecmp(c->argv[j]->ptr,"by") && leftargs >= 1) { sortby = c->argv[j+1]; /* If the BY pattern does not contain '*', i.e. it is constant, @@ -2898,18 +4670,6 @@ static void sortCommand(redisClient *c) { REDIS_SORT_GET,c->argv[j+1])); getop++; j++; - } else if (!strcasecmp(c->argv[j]->ptr,"del") && leftargs >= 1) { - listAddNodeTail(operations,createSortOperation( - REDIS_SORT_DEL,c->argv[j+1])); - j++; - } else if (!strcasecmp(c->argv[j]->ptr,"incr") && leftargs >= 1) { - listAddNodeTail(operations,createSortOperation( - REDIS_SORT_INCR,c->argv[j+1])); - j++; - } else if (!strcasecmp(c->argv[j]->ptr,"get") && leftargs >= 1) { - listAddNodeTail(operations,createSortOperation( - REDIS_SORT_DECR,c->argv[j+1])); - j++; } else { decrRefCount(sortval); listRelease(operations); @@ -2924,17 +4684,17 @@ static void sortCommand(redisClient *c) { listLength((list*)sortval->ptr) : dictSize((dict*)sortval->ptr); vector = zmalloc(sizeof(redisSortObject)*vectorlen); - if (!vector) oom("allocating objects vector for SORT"); j = 0; if (sortval->type == REDIS_LIST) { list *list = sortval->ptr; - listNode *ln = list->head; - while(ln) { + listNode *ln; + + listRewind(list); + while((ln = listYield(list))) { robj *ele = ln->value; vector[j].obj = ele; vector[j].u.score = 0; vector[j].u.cmpobj = NULL; - ln = ln->next; j++; } } else { @@ -2943,7 +4703,6 @@ static void sortCommand(redisClient *c) { dictEntry *setele; di = dictGetIterator(set); - if (!di) oom("dictGetIterator"); while((setele = dictNext(di)) != NULL) { vector[j].obj = dictGetEntryKey(setele); vector[j].u.score = 0; @@ -2963,13 +4722,33 @@ static void sortCommand(redisClient *c) { byval = lookupKeyByPattern(c->db,sortby,vector[j].obj); if (!byval || byval->type != REDIS_STRING) continue; if (alpha) { - vector[j].u.cmpobj = byval; - incrRefCount(byval); + if (byval->encoding == REDIS_ENCODING_RAW) { + vector[j].u.cmpobj = byval; + incrRefCount(byval); + } else { + vector[j].u.cmpobj = getDecodedObject(byval); + } } else { - vector[j].u.score = strtod(byval->ptr,NULL); + if (byval->encoding == REDIS_ENCODING_RAW) { + vector[j].u.score = strtod(byval->ptr,NULL); + } else { + if (byval->encoding == REDIS_ENCODING_INT) { + vector[j].u.score = (long)byval->ptr; + } else + assert(1 != 1); + } } } else { - if (!alpha) vector[j].u.score = strtod(vector[j].obj->ptr,NULL); + if (!alpha) { + if (vector[j].obj->encoding == REDIS_ENCODING_RAW) + vector[j].u.score = strtod(vector[j].obj->ptr,NULL); + else { + if (vector[j].obj->encoding == REDIS_ENCODING_INT) + vector[j].u.score = (long) vector[j].obj->ptr; + else + assert(1 != 1); + } + } } } } @@ -2988,40 +4767,81 @@ static void sortCommand(redisClient *c) { server.sort_desc = desc; server.sort_alpha = alpha; server.sort_bypattern = sortby ? 1 : 0; - qsort(vector,vectorlen,sizeof(redisSortObject),sortCompare); + if (sortby && (start != 0 || end != vectorlen-1)) + pqsort(vector,vectorlen,sizeof(redisSortObject),sortCompare, start,end); + else + qsort(vector,vectorlen,sizeof(redisSortObject),sortCompare); } /* Send command output to the output buffer, performing the specified * GET/DEL/INCR/DECR operations if any. */ outputlen = getop ? getop*(end-start+1) : end-start+1; - addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",outputlen)); - for (j = start; j <= end; j++) { - listNode *ln = operations->head; - if (!getop) { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n", - sdslen(vector[j].obj->ptr))); - addReply(c,vector[j].obj); - addReply(c,shared.crlf); + if (storekey == NULL) { + /* STORE option not specified, sent the sorting result to client */ + addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",outputlen)); + for (j = start; j <= end; j++) { + listNode *ln; + if (!getop) { + addReplyBulkLen(c,vector[j].obj); + addReply(c,vector[j].obj); + addReply(c,shared.crlf); + } + listRewind(operations); + while((ln = listYield(operations))) { + redisSortOperation *sop = ln->value; + robj *val = lookupKeyByPattern(c->db,sop->pattern, + vector[j].obj); + + if (sop->type == REDIS_SORT_GET) { + if (!val || val->type != REDIS_STRING) { + addReply(c,shared.nullbulk); + } else { + addReplyBulkLen(c,val); + addReply(c,val); + addReply(c,shared.crlf); + } + } else { + assert(sop->type == REDIS_SORT_GET); /* always fails */ + } + } } - while(ln) { - redisSortOperation *sop = ln->value; - robj *val = lookupKeyByPattern(c->db,sop->pattern, - vector[j].obj); + } else { + robj *listObject = createListObject(); + list *listPtr = (list*) listObject->ptr; - if (sop->type == REDIS_SORT_GET) { - if (!val || val->type != REDIS_STRING) { - addReply(c,shared.nullbulk); + /* STORE option specified, set the sorting result as a List object */ + for (j = start; j <= end; j++) { + listNode *ln; + if (!getop) { + listAddNodeTail(listPtr,vector[j].obj); + incrRefCount(vector[j].obj); + } + listRewind(operations); + while((ln = listYield(operations))) { + redisSortOperation *sop = ln->value; + robj *val = lookupKeyByPattern(c->db,sop->pattern, + vector[j].obj); + + if (sop->type == REDIS_SORT_GET) { + if (!val || val->type != REDIS_STRING) { + listAddNodeTail(listPtr,createStringObject("",0)); + } else { + listAddNodeTail(listPtr,val); + incrRefCount(val); + } } else { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n", - sdslen(val->ptr))); - addReply(c,val); - addReply(c,shared.crlf); + assert(sop->type == REDIS_SORT_GET); /* always fails */ } - } else if (sop->type == REDIS_SORT_DEL) { - /* TODO */ } - ln = ln->next; } + if (dictReplace(c->db->dict,storekey,listObject)) { + incrRefCount(storekey); + } + /* Note: we add 1 because the DB is dirty anyway since even if the + * SORT result is empty a new key is set and maybe the old content + * replaced. */ + server.dirty += 1+outputlen; + addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",outputlen)); } /* Cleanup */ @@ -3037,29 +4857,59 @@ static void sortCommand(redisClient *c) { static void infoCommand(redisClient *c) { sds info; time_t uptime = time(NULL)-server.stat_starttime; + int j; info = sdscatprintf(sdsempty(), "redis_version:%s\r\n" + "arch_bits:%s\r\n" + "uptime_in_seconds:%d\r\n" + "uptime_in_days:%d\r\n" "connected_clients:%d\r\n" "connected_slaves:%d\r\n" - "used_memory:%d\r\n" + "used_memory:%zu\r\n" "changes_since_last_save:%lld\r\n" + "bgsave_in_progress:%d\r\n" "last_save_time:%d\r\n" "total_connections_received:%lld\r\n" "total_commands_processed:%lld\r\n" - "uptime_in_seconds:%d\r\n" - "uptime_in_days:%d\r\n" + "role:%s\r\n" ,REDIS_VERSION, + (sizeof(long) == 8) ? "64" : "32", + uptime, + uptime/(3600*24), listLength(server.clients)-listLength(server.slaves), listLength(server.slaves), server.usedmemory, server.dirty, + server.bgsaveinprogress, server.lastsave, server.stat_numconnections, server.stat_numcommands, - uptime, - uptime/(3600*24) + 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 + ); + } + 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); + } + } addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",sdslen(info))); addReplySds(c,info); addReply(c,shared.crlf); @@ -3071,7 +4921,7 @@ static void monitorCommand(redisClient *c) { c->flags |= (REDIS_SLAVE|REDIS_MONITOR); c->slaveseldb = 0; - if (!listAddNodeTail(server.monitors,c)) oom("listAddNodeTail"); + listAddNodeTail(server.monitors,c); addReply(c,shared.ok); } @@ -3093,6 +4943,18 @@ static int setExpire(redisDb *db, robj *key, time_t when) { } } +/* Return the expire time of the specified key, or -1 if no expire + * is associated with this key (i.e. the key is non volatile) */ +static time_t getExpire(redisDb *db, robj *key) { + dictEntry *de; + + /* No expire? return ASAP */ + if (dictSize(db->expires) == 0 || + (de = dictFind(db->expires,key)) == NULL) return -1; + + return (time_t) dictGetEntryVal(de); +} + static int expireIfNeeded(redisDb *db, robj *key) { time_t when; dictEntry *de; @@ -3118,51 +4980,100 @@ static int deleteIfVolatile(redisDb *db, robj *key) { (de = dictFind(db->expires,key)) == NULL) return 0; /* Delete the key */ + server.dirty++; dictDelete(db->expires,key); return dictDelete(db->dict,key) == DICT_OK; } -static void expireCommand(redisClient *c) { +static void expireGenericCommand(redisClient *c, robj *key, time_t seconds) { dictEntry *de; - int seconds = atoi(c->argv[2]->ptr); - de = dictFind(c->db->dict,c->argv[1]); + de = dictFind(c->db->dict,key); if (de == NULL) { addReply(c,shared.czero); return; } - if (seconds <= 0) { - addReply(c, shared.czero); + if (seconds < 0) { + if (deleteKey(c->db,key)) server.dirty++; + addReply(c, shared.cone); return; } else { time_t when = time(NULL)+seconds; - if (setExpire(c->db,c->argv[1],when)) + if (setExpire(c->db,key,when)) { addReply(c,shared.cone); - else + server.dirty++; + } else { addReply(c,shared.czero); + } return; } } -/* =============================== Replication ============================= */ +static void expireCommand(redisClient *c) { + expireGenericCommand(c,c->argv[1],strtol(c->argv[2]->ptr,NULL,10)); +} -/* Send the whole output buffer syncronously to the slave. This a general operation in theory, but it is actually useful only for replication. */ -static int flushClientOutput(redisClient *c) { - int retval; - time_t start = time(NULL); +static void expireatCommand(redisClient *c) { + expireGenericCommand(c,c->argv[1],strtol(c->argv[2]->ptr,NULL,10)-time(NULL)); +} - while(listLength(c->reply)) { - if (time(NULL)-start > 5) return REDIS_ERR; /* 5 seconds timeout */ - retval = aeWait(c->fd,AE_WRITABLE,1000); - if (retval == -1) { - return REDIS_ERR; - } else if (retval & AE_WRITABLE) { - sendReplyToClient(NULL, c->fd, c, AE_WRITABLE); +static void ttlCommand(redisClient *c) { + time_t expire; + int ttl = -1; + + expire = getExpire(c->db,c->argv[1]); + if (expire != -1) { + ttl = (int) (expire-time(NULL)); + if (ttl < 0) ttl = -1; + } + addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",ttl)); +} + +static void msetGenericCommand(redisClient *c, int nx) { + int j; + + if ((c->argc % 2) == 0) { + addReplySds(c,sdsnew("-ERR wrong number of arguments\r\n")); + return; + } + /* Handle the NX flag. The MSETNX semantic is to return zero and don't + * set nothing at all if at least one already key exists. */ + if (nx) { + for (j = 1; j < c->argc; j += 2) { + if (dictFind(c->db->dict,c->argv[j]) != NULL) { + addReply(c, shared.czero); + return; + } } } - return REDIS_OK; + + for (j = 1; j < c->argc; j += 2) { + int retval; + + retval = dictAdd(c->db->dict,c->argv[j],c->argv[j+1]); + if (retval == DICT_ERR) { + dictReplace(c->db->dict,c->argv[j],c->argv[j+1]); + incrRefCount(c->argv[j+1]); + } else { + incrRefCount(c->argv[j]); + incrRefCount(c->argv[j+1]); + } + removeExpire(c->db,c->argv[j]); + } + server.dirty += (c->argc-1)/2; + addReply(c, nx ? shared.cone : shared.ok); +} + +static void msetCommand(redisClient *c) { + msetGenericCommand(c,0); +} + +static void msetnxCommand(redisClient *c) { + msetGenericCommand(c,1); } +/* =============================== Replication ============================= */ + static int syncWrite(int fd, char *ptr, ssize_t size, int timeout) { ssize_t nwritten, ret = size; time_t start = time(NULL); @@ -3226,52 +5137,174 @@ static int syncReadLine(int fd, char *ptr, ssize_t size, int timeout) { } static void syncCommand(redisClient *c) { - struct stat sb; - int fd = -1, len; - time_t start = time(NULL); - char sizebuf[32]; - /* ignore SYNC if aleady slave or in monitor mode */ if (c->flags & REDIS_SLAVE) return; - redisLog(REDIS_NOTICE,"Slave ask for syncronization"); - if (flushClientOutput(c) == REDIS_ERR || - rdbSave(server.dbfilename) != REDIS_OK) - goto closeconn; - - fd = open(server.dbfilename, O_RDONLY); - if (fd == -1 || fstat(fd,&sb) == -1) goto closeconn; - len = sb.st_size; - - snprintf(sizebuf,32,"$%d\r\n",len); - if (syncWrite(c->fd,sizebuf,strlen(sizebuf),5) == -1) goto closeconn; - while(len) { - char buf[1024]; - int nread; - - if (time(NULL)-start > REDIS_MAX_SYNC_TIME) goto closeconn; - nread = read(fd,buf,1024); - if (nread == -1) goto closeconn; - len -= nread; - if (syncWrite(c->fd,buf,nread,5) == -1) goto closeconn; - } - if (syncWrite(c->fd,"\r\n",2,5) == -1) goto closeconn; - close(fd); + /* SYNC can't be issued when the server has pending data to send to + * the client about already issued commands. We need a fresh reply + * buffer registering the differences between the BGSAVE and the current + * dataset, so that we can copy to other slaves if needed. */ + if (listLength(c->reply) != 0) { + addReplySds(c,sdsnew("-ERR SYNC is invalid with pending input\r\n")); + return; + } + + redisLog(REDIS_NOTICE,"Slave ask for synchronization"); + /* Here we need to check if there is a background saving operation + * in progress, or if it is required to start one */ + if (server.bgsaveinprogress) { + /* Ok a background save is in progress. Let's check if it is a good + * one for replication, i.e. if there is another slave that is + * registering differences since the server forked to save */ + redisClient *slave; + listNode *ln; + + listRewind(server.slaves); + while((ln = listYield(server.slaves))) { + slave = ln->value; + if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) break; + } + if (ln) { + /* Perfect, the server is already registering differences for + * another slave. Set the right state, and copy the buffer. */ + listRelease(c->reply); + c->reply = listDup(slave->reply); + c->replstate = REDIS_REPL_WAIT_BGSAVE_END; + redisLog(REDIS_NOTICE,"Waiting for end of BGSAVE for SYNC"); + } else { + /* No way, we need to wait for the next BGSAVE in order to + * register differences */ + c->replstate = REDIS_REPL_WAIT_BGSAVE_START; + redisLog(REDIS_NOTICE,"Waiting for next BGSAVE for SYNC"); + } + } else { + /* Ok we don't have a BGSAVE in progress, let's start one */ + redisLog(REDIS_NOTICE,"Starting BGSAVE for SYNC"); + if (rdbSaveBackground(server.dbfilename) != REDIS_OK) { + redisLog(REDIS_NOTICE,"Replication failed, can't BGSAVE"); + addReplySds(c,sdsnew("-ERR Unalbe to perform background save\r\n")); + return; + } + c->replstate = REDIS_REPL_WAIT_BGSAVE_END; + } + c->repldbfd = -1; c->flags |= REDIS_SLAVE; c->slaveseldb = 0; - if (!listAddNodeTail(server.slaves,c)) oom("listAddNodeTail"); - redisLog(REDIS_NOTICE,"Syncronization with slave succeeded"); + listAddNodeTail(server.slaves,c); return; +} -closeconn: - if (fd != -1) close(fd); - c->flags |= REDIS_CLOSE; - redisLog(REDIS_WARNING,"Syncronization with slave failed"); - return; +static void sendBulkToSlave(aeEventLoop *el, int fd, void *privdata, int mask) { + redisClient *slave = privdata; + REDIS_NOTUSED(el); + REDIS_NOTUSED(mask); + char buf[REDIS_IOBUF_LEN]; + ssize_t nwritten, buflen; + + if (slave->repldboff == 0) { + /* Write the bulk write count before to transfer the DB. In theory here + * we don't know how much room there is in the output buffer of the + * socket, but in pratice SO_SNDLOWAT (the minimum count for output + * operations) will never be smaller than the few bytes we need. */ + sds bulkcount; + + bulkcount = sdscatprintf(sdsempty(),"$%lld\r\n",(unsigned long long) + slave->repldbsize); + if (write(fd,bulkcount,sdslen(bulkcount)) != (signed)sdslen(bulkcount)) + { + sdsfree(bulkcount); + freeClient(slave); + return; + } + sdsfree(bulkcount); + } + lseek(slave->repldbfd,slave->repldboff,SEEK_SET); + buflen = read(slave->repldbfd,buf,REDIS_IOBUF_LEN); + if (buflen <= 0) { + redisLog(REDIS_WARNING,"Read error sending DB to slave: %s", + (buflen == 0) ? "premature EOF" : strerror(errno)); + freeClient(slave); + return; + } + if ((nwritten = write(fd,buf,buflen)) == -1) { + redisLog(REDIS_DEBUG,"Write error sending DB to slave: %s", + strerror(errno)); + freeClient(slave); + return; + } + slave->repldboff += nwritten; + if (slave->repldboff == slave->repldbsize) { + close(slave->repldbfd); + slave->repldbfd = -1; + aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE); + slave->replstate = REDIS_REPL_ONLINE; + if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, + sendReplyToClient, slave, NULL) == AE_ERR) { + freeClient(slave); + return; + } + addReplySds(slave,sdsempty()); + redisLog(REDIS_NOTICE,"Synchronization with slave succeeded"); + } +} + +/* This function is called at the end of every backgrond saving. + * The argument bgsaveerr is REDIS_OK if the background saving succeeded + * otherwise REDIS_ERR is passed to the function. + * + * The goal of this function is to handle slaves waiting for a successful + * background saving in order to perform non-blocking synchronization. */ +static void updateSlavesWaitingBgsave(int bgsaveerr) { + listNode *ln; + int startbgsave = 0; + + listRewind(server.slaves); + while((ln = listYield(server.slaves))) { + redisClient *slave = ln->value; + + if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) { + startbgsave = 1; + slave->replstate = REDIS_REPL_WAIT_BGSAVE_END; + } else if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) { + struct redis_stat buf; + + if (bgsaveerr != REDIS_OK) { + freeClient(slave); + redisLog(REDIS_WARNING,"SYNC failed. BGSAVE child returned an error"); + continue; + } + if ((slave->repldbfd = open(server.dbfilename,O_RDONLY)) == -1 || + redis_fstat(slave->repldbfd,&buf) == -1) { + freeClient(slave); + redisLog(REDIS_WARNING,"SYNC failed. Can't open/stat DB after BGSAVE: %s", strerror(errno)); + continue; + } + slave->repldboff = 0; + slave->repldbsize = buf.st_size; + slave->replstate = REDIS_REPL_SEND_BULK; + aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE); + if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, sendBulkToSlave, slave, NULL) == AE_ERR) { + freeClient(slave); + continue; + } + } + } + if (startbgsave) { + if (rdbSaveBackground(server.dbfilename) != REDIS_OK) { + listRewind(server.slaves); + redisLog(REDIS_WARNING,"SYNC failed. BGSAVE failed"); + while((ln = listYield(server.slaves))) { + redisClient *slave = ln->value; + + if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) + freeClient(slave); + } + } + } } static int syncWithMaster(void) { - char buf[1024], tmpfile[256]; + char buf[1024], tmpfile[256], authcmd[1024]; int dumpsize; int fd = anetTcpConnect(NULL,server.masterhost,server.masterport); int dfd; @@ -3281,6 +5314,30 @@ static int syncWithMaster(void) { strerror(errno)); return REDIS_ERR; } + + /* AUTH with the master if required. */ + if(server.masterauth) { + snprintf(authcmd, 1024, "AUTH %s\r\n", server.masterauth); + if (syncWrite(fd, authcmd, strlen(server.masterauth)+7, 5) == -1) { + close(fd); + redisLog(REDIS_WARNING,"Unable to AUTH to MASTER: %s", + strerror(errno)); + return REDIS_ERR; + } + /* Read the AUTH result. */ + if (syncReadLine(fd,buf,1024,3600) == -1) { + close(fd); + redisLog(REDIS_WARNING,"I/O error reading auth result from MASTER: %s", + strerror(errno)); + return REDIS_ERR; + } + if (buf[0] != '+') { + close(fd); + redisLog(REDIS_WARNING,"Cannot AUTH to MASTER, is the masterauth password correct?"); + return REDIS_ERR; + } + } + /* Issue the SYNC command */ if (syncWrite(fd,"SYNC \r\n",7,5) == -1) { close(fd); @@ -3289,12 +5346,17 @@ static int syncWithMaster(void) { return REDIS_ERR; } /* Read the bulk write count */ - if (syncReadLine(fd,buf,1024,5) == -1) { + if (syncReadLine(fd,buf,1024,3600) == -1) { close(fd); redisLog(REDIS_WARNING,"I/O error reading bulk count from MASTER: %s", strerror(errno)); return REDIS_ERR; } + if (buf[0] != '$') { + close(fd); + redisLog(REDIS_WARNING,"Bad protocol from MASTER, the first byte is not '$', are you sure the host and port are right?"); + return REDIS_ERR; + } dumpsize = atoi(buf+1); redisLog(REDIS_NOTICE,"Receiving %d bytes data dump from MASTER",dumpsize); /* Read the bulk write data on a temp file */ @@ -3344,8 +5406,320 @@ static int syncWithMaster(void) { return REDIS_OK; } +static void slaveofCommand(redisClient *c) { + if (!strcasecmp(c->argv[1]->ptr,"no") && + !strcasecmp(c->argv[2]->ptr,"one")) { + if (server.masterhost) { + sdsfree(server.masterhost); + server.masterhost = NULL; + if (server.master) freeClient(server.master); + server.replstate = REDIS_REPL_NONE; + redisLog(REDIS_NOTICE,"MASTER MODE enabled (user request)"); + } + } else { + sdsfree(server.masterhost); + server.masterhost = sdsdup(c->argv[1]->ptr); + server.masterport = atoi(c->argv[2]->ptr); + if (server.master) freeClient(server.master); + server.replstate = REDIS_REPL_CONNECT; + redisLog(REDIS_NOTICE,"SLAVE OF %s:%d enabled (user request)", + server.masterhost, server.masterport); + } + addReply(c,shared.ok); +} + +/* ============================ Maxmemory directive ======================== */ + +/* 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: + * + * - Free objects from the free list + * - Try to remove keys with an EXPIRE set + * + * It is not possible to free enough memory to reach used-memory < maxmemory + * the server will start refusing commands that will enlarge even more the + * memory usage. + */ +static void freeMemoryIfNeeded(void) { + while (server.maxmemory && zmalloc_used_memory() > server.maxmemory) { + if (listLength(server.objfreelist)) { + robj *o; + + listNode *head = listFirst(server.objfreelist); + o = listNodeValue(head); + listDelNode(server.objfreelist,head); + zfree(o); + } else { + int j, k, freed = 0; + + for (j = 0; j < server.dbnum; j++) { + int minttl = -1; + robj *minkey = 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; + } + } + deleteKey(server.db+j,minkey); + } + } + if (!freed) return; /* nothing to free... */ + } + } +} + +/* ============================== Append Only file ========================== */ + +static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc) { + sds buf = sdsempty(); + int j; + ssize_t nwritten; + time_t now; + robj *tmpargv[3]; + + /* The DB this command was targetting is not the same as the last command + * we appendend. To issue a SELECT command is needed. */ + if (dictid != server.appendseldb) { + char seldb[64]; + + snprintf(seldb,sizeof(seldb),"%d",dictid); + buf = sdscatprintf(buf,"*2\r\n$6\r\nSELECT\r\n$%d\r\n%s\r\n", + strlen(seldb),seldb); + server.appendseldb = dictid; + } + + /* "Fix" the argv vector if the command is EXPIRE. We want to translate + * EXPIREs into EXPIREATs calls */ + if (cmd->proc == expireCommand) { + long when; + + tmpargv[0] = createStringObject("EXPIREAT",8); + tmpargv[1] = argv[1]; + incrRefCount(argv[1]); + when = time(NULL)+strtol(argv[2]->ptr,NULL,10); + tmpargv[2] = createObject(REDIS_STRING, + sdscatprintf(sdsempty(),"%ld",when)); + argv = tmpargv; + } + + /* Append the actual command */ + buf = sdscatprintf(buf,"*%d\r\n",argc); + for (j = 0; j < argc; j++) { + robj *o = argv[j]; + + if (o->encoding != REDIS_ENCODING_RAW) + o = getDecodedObject(o); + buf = sdscatprintf(buf,"$%d\r\n",sdslen(o->ptr)); + buf = sdscatlen(buf,o->ptr,sdslen(o->ptr)); + buf = sdscatlen(buf,"\r\n",2); + if (o != argv[j]) + decrRefCount(o); + } + + /* Free the objects from the modified argv for EXPIREAT */ + if (cmd->proc == expireCommand) { + for (j = 0; j < 3; j++) + decrRefCount(argv[j]); + } + + /* We want to perform a single write. This should be guaranteed atomic + * at least if the filesystem we are writing is a real physical one. + * While this will save us against the server being killed I don't think + * there is much to do about the whole server stopping for power problems + * or alike */ + nwritten = write(server.appendfd,buf,sdslen(buf)); + if (nwritten != (signed)sdslen(buf)) { + /* Ooops, we are in troubles. The best thing to do for now is + * to simply exit instead to give the illusion that everything is + * working as expected. */ + if (nwritten == -1) { + redisLog(REDIS_WARNING,"Exiting on error writing to the append-only file: %s",strerror(errno)); + } else { + redisLog(REDIS_WARNING,"Exiting on short write while writing to the append-only file: %s",strerror(errno)); + } + exit(1); + } + now = time(NULL); + if (server.appendfsync == APPENDFSYNC_ALWAYS || + (server.appendfsync == APPENDFSYNC_EVERYSEC && + now-server.lastfsync > 1)) + { + fsync(server.appendfd); /* Let's try to get this data on the disk */ + server.lastfsync = now; + } +} + +/* In Redis commands are always executed in the context of a client, so in + * order to load the append only file we need to create a fake client. */ +static struct redisClient *createFakeClient(void) { + struct redisClient *c = zmalloc(sizeof(*c)); + + selectDb(c,0); + c->fd = -1; + c->querybuf = sdsempty(); + c->argc = 0; + c->argv = NULL; + c->flags = 0; + /* We set the fake client as a slave waiting for the synchronization + * so that Redis will not try to send replies to this client. */ + c->replstate = REDIS_REPL_WAIT_BGSAVE_START; + c->reply = listCreate(); + listSetFreeMethod(c->reply,decrRefCount); + listSetDupMethod(c->reply,dupClientReplyValue); + return c; +} + +static void freeFakeClient(struct redisClient *c) { + sdsfree(c->querybuf); + listRelease(c->reply); + zfree(c); +} + +/* Replay the append log file. On error REDIS_OK is returned. On non fatal + * error (the append only file is zero-length) REDIS_ERR is returned. On + * fatal error an error message is logged and the program exists. */ +int loadAppendOnlyFile(char *filename) { + struct redisClient *fakeClient; + FILE *fp = fopen(filename,"r"); + struct redis_stat sb; + + if (redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0) + return REDIS_ERR; + + if (fp == NULL) { + redisLog(REDIS_WARNING,"Fatal error: can't open the append log file for reading: %s",strerror(errno)); + exit(1); + } + + fakeClient = createFakeClient(); + while(1) { + int argc, j; + unsigned long len; + robj **argv; + char buf[128]; + sds argsds; + struct redisCommand *cmd; + + if (fgets(buf,sizeof(buf),fp) == NULL) { + if (feof(fp)) + break; + else + goto readerr; + } + if (buf[0] != '*') goto fmterr; + argc = atoi(buf+1); + argv = zmalloc(sizeof(robj*)*argc); + for (j = 0; j < argc; j++) { + if (fgets(buf,sizeof(buf),fp) == NULL) goto readerr; + if (buf[0] != '$') goto fmterr; + len = strtol(buf+1,NULL,10); + argsds = sdsnewlen(NULL,len); + if (fread(argsds,len,1,fp) == 0) goto fmterr; + argv[j] = createObject(REDIS_STRING,argsds); + if (fread(buf,2,1,fp) == 0) goto fmterr; /* discard CRLF */ + } + + /* Command lookup */ + cmd = lookupCommand(argv[0]->ptr); + if (!cmd) { + redisLog(REDIS_WARNING,"Unknown command '%s' reading the append only file", argv[0]->ptr); + exit(1); + } + /* Try object sharing and encoding */ + if (server.shareobjects) { + int j; + for(j = 1; j < argc; j++) + argv[j] = tryObjectSharing(argv[j]); + } + if (cmd->flags & REDIS_CMD_BULK) + tryObjectEncoding(argv[argc-1]); + /* Run the command in the context of a fake client */ + fakeClient->argc = argc; + fakeClient->argv = argv; + cmd->proc(fakeClient); + /* Discard the reply objects list from the fake client */ + while(listLength(fakeClient->reply)) + listDelNode(fakeClient->reply,listFirst(fakeClient->reply)); + /* Clean up, ready for the next command */ + for (j = 0; j < argc; j++) decrRefCount(argv[j]); + zfree(argv); + } + fclose(fp); + freeFakeClient(fakeClient); + return REDIS_OK; + +readerr: + if (feof(fp)) { + redisLog(REDIS_WARNING,"Unexpected end of file reading the append only file"); + } else { + redisLog(REDIS_WARNING,"Unrecoverable error reading the append only file: %s", strerror(errno)); + } + exit(1); +fmterr: + redisLog(REDIS_WARNING,"Bad file format reading the append only file"); + exit(1); +} + +/* ================================= Debugging ============================== */ + +static void debugCommand(redisClient *c) { + if (!strcasecmp(c->argv[1]->ptr,"segfault")) { + *((char*)-1) = 'x'; + } else if (!strcasecmp(c->argv[1]->ptr,"object") && c->argc == 3) { + dictEntry *de = dictFind(c->db->dict,c->argv[2]); + robj *key, *val; + + if (!de) { + addReply(c,shared.nokeyerr); + return; + } + key = dictGetEntryKey(de); + val = dictGetEntryVal(de); + addReplySds(c,sdscatprintf(sdsempty(), + "+Key at:%p refcount:%d, value at:%p refcount:%d encoding:%d\r\n", + key, key->refcount, val, val->refcount, val->encoding)); + } else { + addReplySds(c,sdsnew( + "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT ]\r\n")); + } +} + /* =================================== Main! ================================ */ +#ifdef __linux__ +int linuxOvercommitMemoryValue(void) { + FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r"); + char buf[64]; + + if (!fp) return -1; + if (fgets(buf,64,fp) == NULL) { + fclose(fp); + return -1; + } + fclose(fp); + + return atoi(buf); +} + +void linuxOvercommitMemoryWarning(void) { + if (linuxOvercommitMemoryValue() == 0) { + redisLog(REDIS_WARNING,"WARNING overcommit_memory is set to 0! Background save may fail under low condition memory. To fix this issue add 'vm.overcommit_memory = 1' to /etc/sysctl.conf and then reboot or run the command 'sysctl vm.overcommit_memory=1' for this to take effect."); + } +} +#endif /* __linux__ */ + static void daemonize(void) { int fd; FILE *fp; @@ -3373,17 +5747,27 @@ static void daemonize(void) { int main(int argc, char **argv) { initServerConfig(); if (argc == 2) { - ResetServerSaveParams(); + resetServerSaveParams(); loadServerConfig(argv[1]); } else if (argc > 2) { fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf]\n"); exit(1); + } else { + redisLog(REDIS_WARNING,"Warning: no config file specified, using the default config. In order to specify a config file use 'redis-server /path/to/redis.conf'"); } initServer(); if (server.daemonize) daemonize(); redisLog(REDIS_NOTICE,"Server started, Redis version " REDIS_VERSION); - if (rdbLoad(server.dbfilename) == REDIS_OK) - redisLog(REDIS_NOTICE,"DB loaded from disk"); +#ifdef __linux__ + linuxOvercommitMemoryWarning(); +#endif + if (server.appendonly) { + if (loadAppendOnlyFile(server.appendfilename) == REDIS_OK) + redisLog(REDIS_NOTICE,"DB loaded from append only file"); + } else { + if (rdbLoad(server.dbfilename) == REDIS_OK) + redisLog(REDIS_NOTICE,"DB loaded from disk"); + } if (aeCreateFileEvent(server.el, server.fd, AE_READABLE, acceptHandler, NULL, NULL) == AE_ERR) oom("creating file event"); redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port); @@ -3391,3 +5775,139 @@ int main(int argc, char **argv) { aeDeleteEventLoop(server.el); return 0; } + +/* ============================= Backtrace support ========================= */ + +#ifdef HAVE_BACKTRACE +static char *findFuncName(void *pointer, unsigned long *offset); + +static void *getMcontextEip(ucontext_t *uc) { +#if defined(__FreeBSD__) + return (void*) uc->uc_mcontext.mc_eip; +#elif defined(__dietlibc__) + return (void*) uc->uc_mcontext.eip; +#elif defined(__APPLE__) && !defined(MAC_OS_X_VERSION_10_6) + return (void*) uc->uc_mcontext->__ss.__eip; +#elif defined(__APPLE__) && defined(MAC_OS_X_VERSION_10_6) + #if defined(_STRUCT_X86_THREAD_STATE64) && !defined(__i386__) + return (void*) uc->uc_mcontext->__ss.__rip; + #else + return (void*) uc->uc_mcontext->__ss.__eip; + #endif +#elif defined(__i386__) || defined(__X86_64__) /* Linux x86 */ + return (void*) uc->uc_mcontext.gregs[REG_EIP]; +#elif defined(__ia64__) /* Linux IA64 */ + return (void*) uc->uc_mcontext.sc_ip; +#else + return NULL; +#endif +} + +static void segvHandler(int sig, siginfo_t *info, void *secret) { + void *trace[100]; + char **messages = NULL; + int i, trace_size = 0; + unsigned long offset=0; + time_t uptime = time(NULL)-server.stat_starttime; + ucontext_t *uc = (ucontext_t*) secret; + REDIS_NOTUSED(info); + + redisLog(REDIS_WARNING, + "======= Ooops! Redis %s got signal: -%d- =======", REDIS_VERSION, sig); + redisLog(REDIS_WARNING, "%s", sdscatprintf(sdsempty(), + "redis_version:%s; " + "uptime_in_seconds:%d; " + "connected_clients:%d; " + "connected_slaves:%d; " + "used_memory:%zu; " + "changes_since_last_save:%lld; " + "bgsave_in_progress:%d; " + "last_save_time:%d; " + "total_connections_received:%lld; " + "total_commands_processed:%lld; " + "role:%s;" + ,REDIS_VERSION, + uptime, + listLength(server.clients)-listLength(server.slaves), + listLength(server.slaves), + server.usedmemory, + server.dirty, + server.bgsaveinprogress, + server.lastsave, + server.stat_numconnections, + server.stat_numcommands, + server.masterhost == NULL ? "master" : "slave" + )); + + trace_size = backtrace(trace, 100); + /* overwrite sigaction with caller's address */ + if (getMcontextEip(uc) != NULL) { + trace[1] = getMcontextEip(uc); + } + messages = backtrace_symbols(trace, trace_size); + + for (i=1; i= symsTable[i].pointer) { + off=lp-symsTable[i].pointer; + if (ret < 0 || off < minoff) { + minoff=off; + ret=i; + } + } + } + if (ret == -1) return NULL; + *offset = minoff; + return symsTable[ret].name; +} +#else /* HAVE_BACKTRACE */ +static void setupSigSegvAction(void) { +} +#endif /* HAVE_BACKTRACE */ + + + +/* The End */ + + +