X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/942a39618b8a7fe59435f632ccfb5e7495564fa1..1fd9bc8adc3ae42e7e33564d6f76d76f89dfb78b:/redis.c diff --git a/redis.c b/redis.c index 010dfb8f..dff33754 100644 --- a/redis.c +++ b/redis.c @@ -27,7 +27,7 @@ * POSSIBILITY OF SUCH DAMAGE. */ -#define REDIS_VERSION "1.000" +#define REDIS_VERSION "1.07" #include "fmacros.h" #include "config.h" @@ -56,7 +56,13 @@ #include #include #include +#include #include +#include + +#if defined(__sun) +#include "solarisfixes.h" +#endif #include "redis.h" #include "ae.h" /* Event driven programming library */ @@ -84,7 +90,12 @@ #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 */ +#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% */ @@ -102,7 +113,8 @@ #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 */ @@ -166,11 +178,8 @@ /* 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 */ @@ -181,6 +190,13 @@ /* 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 ============================== */ @@ -206,9 +222,10 @@ typedef struct redisClient { redisDb *db; int dictid; sds querybuf; - robj **argv; - int argc; + 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 */ @@ -252,24 +269,32 @@ 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; + pid_t bgrewritechildpid; + sds bgrewritebuf; /* buffer taken by parent during oppend only rewrite */ 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; /* client that is master for this slave */ int replstate; unsigned int maxclients; - unsigned int maxmemory; + 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; @@ -303,6 +328,28 @@ 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, @@ -312,6 +359,12 @@ 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); @@ -327,10 +380,11 @@ 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 robj *getDecodedObject(robj *o); static int removeExpire(redisDb *db, robj *key); static int expireIfNeeded(redisDb *db, robj *key); static int deleteIfVolatile(redisDb *db, robj *key); @@ -342,6 +396,13 @@ static void freeMemoryIfNeeded(void); static int processCommand(redisClient *c); static void setupSigSegvAction(void); static void rdbRemoveTempFile(pid_t childpid); +static void aofRemoveTempFile(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); @@ -362,6 +423,7 @@ static void dbsizeCommand(redisClient *c); static void lastsaveCommand(redisClient *c); static void saveCommand(redisClient *c); static void bgsaveCommand(redisClient *c); +static void bgrewriteaofCommand(redisClient *c); static void shutdownCommand(redisClient *c); static void moveCommand(redisClient *c); static void renameCommand(redisClient *c); @@ -382,6 +444,7 @@ 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); @@ -393,14 +456,28 @@ 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 getSetCommand(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 zincrbyCommand(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 ================================= */ /* Global vars */ @@ -424,12 +501,14 @@ static struct redisCommand cmdTable[] = { {"lrange",lrangeCommand,4,REDIS_CMD_INLINE}, {"ltrim",ltrimCommand,4,REDIS_CMD_INLINE}, {"lrem",lremCommand,4,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}, {"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}, @@ -437,15 +516,27 @@ static struct redisCommand cmdTable[] = { {"sdiff",sdiffCommand,-2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, {"sdiffstore",sdiffstoreCommand,-3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM}, {"smembers",sinterCommand,2,REDIS_CMD_INLINE}, + {"zadd",zaddCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, + {"zincrby",zincrbyCommand,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}, + {"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}, @@ -453,6 +544,7 @@ static struct redisCommand cmdTable[] = { {"echo",echoCommand,2,REDIS_CMD_BULK}, {"save",saveCommand,1,REDIS_CMD_INLINE}, {"bgsave",bgsaveCommand,1,REDIS_CMD_INLINE}, + {"bgrewriteaof",bgrewriteaofCommand,1,REDIS_CMD_INLINE}, {"shutdown",shutdownCommand,1,REDIS_CMD_INLINE}, {"lastsave",lastsaveCommand,1,REDIS_CMD_INLINE}, {"type",typeCommand,2,REDIS_CMD_INLINE}, @@ -467,6 +559,7 @@ static struct redisCommand cmdTable[] = { {"debug",debugCommand,-2,REDIS_CMD_INLINE}, {NULL,NULL,0,0} }; + /*============================ Utility functions ============================ */ /* Glob-style pattern matching. */ @@ -606,7 +699,7 @@ static void redisLog(int level, const char *fmt, ...) { time_t now; now = time(NULL); - strftime(buf,64,"%d %b %H:%M:%S",gmtime(&now)); + strftime(buf,64,"%d %b %H:%M:%S",localtime(&now)); fprintf(fp,"%s %c ",buf,c[level]); vfprintf(fp, fmt, ap); fprintf(fp,"\n"); @@ -623,6 +716,12 @@ static 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) { @@ -657,37 +756,24 @@ static unsigned int dictObjHash(const void *key) { 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; + robj *o1 = (robj*) key1, *o2 = (robj*) key2; + 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; - } + o1 = getDecodedObject(o1); + o2 = getDecodedObject(o2); + cmp = sdsDictKeyCompare(privdata,o1->ptr,o2->ptr); + decrRefCount(o1); + decrRefCount(o2); + return cmp; } static unsigned int dictEncObjHash(const void *key) { - const robj *o = key; + robj *o = (robj*) 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; - } + o = getDecodedObject(o); + unsigned int hash = dictGenHashFunction(o->ptr, sdslen((sds)o->ptr)); + decrRefCount(o); + return hash; } static dictType setDictType = { @@ -699,6 +785,15 @@ static dictType setDictType = { 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 = { dictObjHash, /* hash function */ NULL, /* key dup */ @@ -765,6 +860,90 @@ static void tryResizeHashTables(void) { } } +/* A background saving child (BGSAVE) terminated its work. Handle this. */ +void backgroundSaveDoneHandler(int statloc) { + int exitcode = WEXITSTATUS(statloc); + 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 terminated by signal"); + rdbRemoveTempFile(server.bgsavechildpid); + } + server.bgsavechildpid = -1; + /* Possibly there are slaves waiting for a BGSAVE in order to be served + * (the first stage of SYNC is a bulk transfer of dump.rdb) */ + updateSlavesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR); +} + +/* A background append only file rewriting (BGREWRITEAOF) terminated its work. + * Handle this. */ +void backgroundRewriteDoneHandler(int statloc) { + int exitcode = WEXITSTATUS(statloc); + int bysignal = WIFSIGNALED(statloc); + + if (!bysignal && exitcode == 0) { + int fd; + char tmpfile[256]; + + redisLog(REDIS_NOTICE, + "Background append only file rewriting terminated with success"); + /* Now it's time to flush the differences accumulated by the parent */ + snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) server.bgrewritechildpid); + fd = open(tmpfile,O_WRONLY|O_APPEND); + if (fd == -1) { + redisLog(REDIS_WARNING, "Not able to open the temp append only file produced by the child: %s", strerror(errno)); + goto cleanup; + } + /* Flush our data... */ + if (write(fd,server.bgrewritebuf,sdslen(server.bgrewritebuf)) != + (signed) sdslen(server.bgrewritebuf)) { + redisLog(REDIS_WARNING, "Error or short write trying to flush the parent diff of the append log file in the child temp file: %s", strerror(errno)); + close(fd); + goto cleanup; + } + redisLog(REDIS_WARNING,"Parent diff flushed into the new append log file with success"); + /* Now our work is to rename the temp file into the stable file. And + * switch the file descriptor used by the server for append only. */ + if (rename(tmpfile,server.appendfilename) == -1) { + redisLog(REDIS_WARNING,"Can't rename the temp append only file into the stable one: %s", strerror(errno)); + close(fd); + goto cleanup; + } + /* Mission completed... almost */ + redisLog(REDIS_NOTICE,"Append only file successfully rewritten."); + if (server.appendfd != -1) { + /* If append only is actually enabled... */ + close(server.appendfd); + server.appendfd = fd; + fsync(fd); + server.appendseldb = -1; /* Make sure it will issue SELECT */ + redisLog(REDIS_NOTICE,"The new append only file was selected for future appends."); + } else { + /* If append only is disabled we just generate a dump in this + * format. Why not? */ + close(fd); + } + } else if (!bysignal && exitcode != 0) { + redisLog(REDIS_WARNING, "Background append only file rewriting error"); + } else { + redisLog(REDIS_WARNING, + "Background append only file rewriting terminated by signal"); + } +cleanup: + sdsfree(server.bgrewritebuf); + server.bgrewritebuf = sdsempty(); + aofRemoveTempFile(server.bgrewritechildpid); + server.bgrewritechildpid = -1; +} + static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { int j, loops = server.cronloops++; REDIS_NOTUSED(eventLoop); @@ -793,7 +972,7 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD * 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(); + if (server.bgsavechildpid == -1) tryResizeHashTables(); /* Show information about connected clients */ if (!(loops % 5)) { @@ -808,28 +987,17 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD if (server.maxidletime && !(loops % 10)) closeTimedoutClients(); - /* Check if a background saving in progress terminated */ - if (server.bgsaveinprogress) { + /* Check if a background saving or AOF rewrite in progress terminated */ + if (server.bgsavechildpid != -1 || server.bgrewritechildpid != -1) { int statloc; - if (wait4(-1,&statloc,WNOHANG,NULL)) { - int exitcode = WEXITSTATUS(statloc); - 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"); + pid_t pid; + + if ((pid = wait3(&statloc,WNOHANG,NULL)) != 0) { + if (pid == server.bgsavechildpid) { + backgroundSaveDoneHandler(statloc); } else { - redisLog(REDIS_WARNING, - "Background saving terminated by signal"); - rdbRemoveTempFile(server.bgsavechildpid); + backgroundRewriteDoneHandler(statloc); } - 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 @@ -848,14 +1016,21 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD } } - /* Try to expire a few timed out keys */ + /* 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; - int num = dictSize(db->expires); - if (num) { + /* 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--) { @@ -866,9 +1041,10 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD 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 */ @@ -920,13 +1096,12 @@ 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; @@ -942,24 +1117,37 @@ 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.aof"; server.requirepass = NULL; server.shareobjects = 0; server.sharingpoolsize = 1024; server.maxclients = 0; server.maxmemory = 0; - ResetServerSaveParams(); + 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() { @@ -977,8 +1165,6 @@ static void initServer() { server.el = aeCreateEventLoop(); server.db = zmalloc(sizeof(redisDb)*server.dbnum); server.sharingpool = dictCreate(&setDictType,NULL); - 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); @@ -990,15 +1176,25 @@ static void initServer() { server.db[j].id = j; } server.cronloops = 0; - server.bgsaveinprogress = 0; server.bgsavechildpid = -1; + server.bgrewritechildpid = -1; + server.bgrewritebuf = sdsempty(); 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 */ @@ -1116,11 +1312,13 @@ static void loadServerConfig(char *filename) { } else if (!strcasecmp(argv[0],"maxclients") && argc == 2) { server.maxclients = atoi(argv[1]); } else if (!strcasecmp(argv[0],"maxmemory") && argc == 2) { - server.maxmemory = atoi(argv[1]); + 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 (!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; @@ -1138,6 +1336,21 @@ static void loadServerConfig(char *filename) { 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 (!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 (!strcasecmp(argv[0],"pidfile") && argc == 2) { @@ -1168,7 +1381,10 @@ 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) { @@ -1196,37 +1412,35 @@ static void freeClient(redisClient *c) { 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; + int copylen = 0; + char buf[GLUEREPLY_UP_TO]; listNode *ln; robj *o; listRewind(c->reply); while((ln = listYield(c->reply))) { + int objlen; + o = ln->value; - totlen += sdslen(o->ptr); - /* 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; - - listRewind(c->reply); - while((ln = listYield(c->reply))) { - 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); + } else { + if (copylen == 0) return; + break; } - /* Now the output buffer is empty, add the new single element */ - o = createObject(REDIS_STRING,sdsnewlen(buf,totlen)); - if (!listAddNodeTail(c->reply,o)) oom("listAddNodeTail"); } + /* 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) { @@ -1236,9 +1450,19 @@ 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); @@ -1262,10 +1486,10 @@ static void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) 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 server + * 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 - * terms think to 'KEYS *' against the loopback interfae) */ + * scenario think about 'KEYS *' against the loopback interfae) */ if (totwritten > REDIS_MAX_WRITE_PER_EVENT) break; } if (nwritten == -1) { @@ -1285,6 +1509,84 @@ 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) { @@ -1298,6 +1600,7 @@ 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 @@ -1315,6 +1618,74 @@ static int processCommand(redisClient *c) { /* 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 (!strcasecmp(c->argv[0]->ptr,"quit")) { @@ -1348,7 +1719,10 @@ 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++; @@ -1377,7 +1751,9 @@ 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); @@ -1403,7 +1779,6 @@ static void replicationFeedSlaves(list *slaves, struct redisCommand *cmd, int di outv = static_outv; } else { outv = zmalloc(sizeof(robj*)*(argc*2+1)); - if (!outv) oom("replicationFeedSlaves"); } for (j = 0; j < argc; j++) { @@ -1412,7 +1787,8 @@ 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; } @@ -1461,34 +1837,7 @@ static void replicationFeedSlaves(list *slaves, struct redisCommand *cmd, int di if (outv != static_outv) zfree(outv); } -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; - } - +static void processInputBuffer(redisClient *c) { again: if (c->bulklen == -1) { /* Read the first line of the query */ @@ -1517,12 +1866,10 @@ again: return; } argv = sdssplitlen(query,sdslen(query)," ",1,&argc); - if (argv == NULL) oom("sdssplitlen"); sdsfree(query); if (c->argv) zfree(c->argv); c->argv = zmalloc(sizeof(robj*)*argc); - if (c->argv == NULL) oom("allocating arguments list for client"); for (j = 0; j < argc; j++) { if (sdslen(argv[j])) { @@ -1555,10 +1902,43 @@ 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) { @@ -1585,20 +1965,23 @@ static redisClient *createClient(int fd) { 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; c->replstate = REDIS_REPL_NONE; - if ((c->reply = listCreate()) == NULL) oom("listCreate"); + c->reply = listCreate(); listSetFreeMethod(c->reply,decrRefCount); listSetDupMethod(c->reply,dupClientReplyValue); if (aeCreateFileEvent(server.el, c->fd, AE_READABLE, - readQueryFromClient, c, NULL) == AE_ERR) { + readQueryFromClient, c) == AE_ERR) { freeClient(c); return NULL; } - if (!listAddNodeTail(server.clients,c)) oom("listAddNodeTail"); + listAddNodeTail(server.clients,c); return c; } @@ -1607,13 +1990,8 @@ static void addReply(redisClient *c, robj *obj) { (c->replstate == REDIS_REPL_NONE || c->replstate == REDIS_REPL_ONLINE) && aeCreateFileEvent(server.el, c->fd, AE_WRITABLE, - sendReplyToClient, c, NULL) == AE_ERR) return; - if (obj->encoding != REDIS_ENCODING_RAW) { - obj = getDecodedObject(obj); - } else { - incrRefCount(obj); - } - if (!listAddNodeTail(c->reply,obj)) oom("listAddNodeTail"); + sendReplyToClient, c) == AE_ERR) return; + listAddNodeTail(c->reply,getDecodedObject(obj)); } static void addReplySds(redisClient *c, sds s) { @@ -1622,6 +2000,14 @@ static void addReplySds(redisClient *c, sds s) { decrRefCount(o); } +static void addReplyDouble(redisClient *c, double d) { + char buf[128]; + + snprintf(buf,sizeof(buf),"%.17g",d); + addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n%s\r\n", + strlen(buf),buf)); +} + static void addReplyBulkLen(redisClient *c, robj *obj) { size_t len; @@ -1669,7 +2055,9 @@ static void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask) { char *err = "-ERR max number of clients reached\r\n"; /* That's a best effort error message, don't check write errors */ - (void) write(c->fd,err,strlen(err)); + if (write(c->fd,err,strlen(err)) == -1) { + /* Nothing to do, Just to avoid the warning... */ + } freeClient(c); return; } @@ -1688,7 +2076,6 @@ 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; @@ -1703,17 +2090,23 @@ 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) { if (o->encoding == REDIS_ENCODING_RAW) { sdsfree(o->ptr); @@ -1728,6 +2121,14 @@ 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); } @@ -1752,6 +2153,7 @@ static void decrRefCount(void *obj) { 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; } @@ -1836,10 +2238,31 @@ static robj *tryObjectSharing(robj *o) { } } +/* 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; +} + /* Try to encode a string object in order to save space */ static int tryObjectEncoding(robj *o) { long value; - char *endptr, buf[32]; sds s = o->ptr; if (o->encoding != REDIS_ENCODING_RAW) @@ -1853,15 +2276,8 @@ static int tryObjectEncoding(robj *o) { /* Currently we try to encode only strings */ assert(o->type == REDIS_STRING); - /* Check if it's possible to encode this value as a long. We are assuming - * that sizeof(long) = sizeof(void) in all the supported archs. */ - value = strtol(s, &endptr, 10); - if (endptr[0] != '\0') return REDIS_ERR; - 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 (strlen(buf) != sdslen(s) || memcmp(buf,s,sdslen(s))) return REDIS_ERR; + /* 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; @@ -1870,11 +2286,15 @@ static int tryObjectEncoding(robj *o) { return REDIS_OK; } -/* Get a decoded version of an encoded object (returned as a new object) */ -static robj *getDecodedObject(const robj *o) { +/* Get a decoded version of an encoded object (returned as a new object). + * If the object is already raw-encoded just increment the ref count. */ +static robj *getDecodedObject(robj *o) { robj *dec; - assert(o->encoding != REDIS_ENCODING_RAW); + if (o->encoding == REDIS_ENCODING_RAW) { + incrRefCount(o); + return o; + } if (o->type == REDIS_STRING && o->encoding == REDIS_ENCODING_INT) { char buf[32]; @@ -1886,6 +2306,48 @@ static robj *getDecodedObject(const robj *o) { } } +/* 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(). + * + * Important note: if objects are not integer encoded, but binary-safe strings, + * sdscmp() from sds.c will apply memcmp() so this function ca be considered + * binary safe. */ +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 ============================ */ static int rdbSaveType(FILE *fp, unsigned char type) { @@ -2025,16 +2487,38 @@ static int rdbSaveStringObjectRaw(FILE *fp, robj *obj) { /* 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; + obj = getDecodedObject(obj); + retval = rdbSaveStringObjectRaw(fp,obj); + decrRefCount(obj); + return retval; +} + +/* 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 { - return rdbSaveStringObjectRaw(fp,obj); + snprintf((char*)buf+1,sizeof(buf)-1,"%.17g",val); + buf[0] = strlen((char*)buf+1); + 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 */ @@ -2104,7 +2588,6 @@ 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); @@ -2112,6 +2595,21 @@ 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); } @@ -2129,7 +2627,7 @@ 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; } @@ -2149,7 +2647,7 @@ werr: static int rdbSaveBackground(char *filename) { pid_t childpid; - if (server.bgsaveinprogress) return REDIS_ERR; + if (server.bgsavechildpid != -1) return REDIS_ERR; if ((childpid = fork()) == 0) { /* Child */ close(server.fd); @@ -2166,7 +2664,6 @@ static int rdbSaveBackground(char *filename) { return REDIS_ERR; } redisLog(REDIS_NOTICE,"Background saving started by pid %d",childpid); - server.bgsaveinprogress = 1; server.bgsavechildpid = childpid; return REDIS_OK; } @@ -2300,6 +2797,23 @@ 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; @@ -2369,16 +2883,35 @@ static int rdbLoad(char *filename) { 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 { - assert(0 != 0); - } + } 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); + } /* Add the new object in the hash table */ retval = dictAdd(d,keyobj,o); if (retval == DICT_ERR) { @@ -2399,7 +2932,7 @@ 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 */ } @@ -2412,7 +2945,7 @@ static void authCommand(redisClient *c) { addReply(c,shared.ok); } else { c->authenticated = 0; - addReply(c,shared.err); + addReplySds(c,sdscatprintf(sdsempty(),"-ERR invalid password\r\n")); } } @@ -2473,7 +3006,7 @@ static void getCommand(redisClient *c) { } } -static void getSetCommand(redisClient *c) { +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]); @@ -2505,6 +3038,49 @@ static void mgetCommand(redisClient *c) { } } +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; + } + } + } + + 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); +} + static void incrDecrCommand(redisClient *c, long long incr) { long long value; int retval; @@ -2626,7 +3202,6 @@ 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) { @@ -2671,6 +3246,7 @@ 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; } } @@ -2679,7 +3255,7 @@ static void typeCommand(redisClient *c) { } static void saveCommand(redisClient *c) { - if (server.bgsaveinprogress) { + if (server.bgsavechildpid != -1) { addReplySds(c,sdsnew("-ERR background save in progress\r\n")); return; } @@ -2691,7 +3267,7 @@ static void saveCommand(redisClient *c) { } static void bgsaveCommand(redisClient *c) { - if (server.bgsaveinprogress) { + if (server.bgsavechildpid != -1) { addReplySds(c,sdsnew("-ERR background save already in progress\r\n")); return; } @@ -2707,7 +3283,7 @@ static void shutdownCommand(redisClient *c) { /* 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) { + if (server.bgsavechildpid != -1) { redisLog(REDIS_WARNING,"There is a live saving child. Killing it!"); kill(server.bgsavechildpid,SIGKILL); rdbRemoveTempFile(server.bgsavechildpid); @@ -2822,9 +3398,9 @@ 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]); @@ -2836,9 +3412,9 @@ 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]); } @@ -3090,7 +3666,7 @@ 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++; @@ -3103,6 +3679,70 @@ 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) { @@ -3249,6 +3889,31 @@ static void spopCommand(redisClient *c) { } } +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; @@ -3262,7 +3927,6 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, int setsnum, r robj *lenobj = NULL, *dstset = NULL; int j, cardinality = 0; - if (!dv) oom("sinterGenericCommand"); for (j = 0; j < setsnum; j++) { robj *setobj; @@ -3309,7 +3973,6 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, int setsnum, r * 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; @@ -3366,7 +4029,6 @@ static void sunionDiffGenericCommand(redisClient *c, robj **setskeys, int setsnu robj *dstset = NULL; int j, cardinality = 0; - if (!dv) oom("sunionDiffGenericCommand"); for (j = 0; j < setsnum; j++) { robj *setobj; @@ -3397,7 +4059,6 @@ static void sunionDiffGenericCommand(redisClient *c, robj **setskeys, int setsnu if (!dv[j]) continue; /* non existing keys are like empty sets */ di = dictGetIterator(dv[j]); - if (!di) oom("dictGetIterator"); while((de = dictNext(di)) != NULL) { robj *ele; @@ -3424,7 +4085,6 @@ static void sunionDiffGenericCommand(redisClient *c, robj **setskeys, int setsnu if (!dstkey) { addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",cardinality)); di = dictGetIterator(dstset->ptr); - if (!di) oom("dictGetIterator"); while((de = dictNext(di)) != NULL) { robj *ele; @@ -3469,161 +4129,679 @@ static void sdiffstoreCommand(redisClient *c) { sunionDiffGenericCommand(c,c->argv+2,c->argc-2,c->argv[1],REDIS_OP_DIFF); } -static void flushdbCommand(redisClient *c) { - server.dirty += dictSize(c->db->dict); - dictEmpty(c->db->dict); - dictEmpty(c->db->expires); - addReply(c,shared.ok); -} +/* ==================================== ZSets =============================== */ -static void flushallCommand(redisClient *c) { - server.dirty += emptyDb(); - addReply(c,shared.ok); - rdbSave(server.dbfilename); - server.dirty++; -} +/* 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"). */ -static redisSortOperation *createSortOperation(int type, robj *pattern) { - redisSortOperation *so = zmalloc(sizeof(*so)); - if (!so) oom("createSortOperation"); - so->type = type; - so->pattern = pattern; - return so; +/* 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; } -/* Return the value associated to the key with a name obtained - * substituting the first occurence of '*' in 'pattern' with 'subst' */ -static robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { - char *p; - sds spat, ssub; - robj keyobj; - int prefixlen, sublen, postfixlen; - /* Expoit the internal sds representation to create a sds string allocated on the stack in order to make this function faster */ - struct { - long len; - long free; - char buf[REDIS_SORTKEY_MAX+1]; - } keyname; +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 */ + +/* This generic command implements both ZADD and ZINCRBY. + * scoreval is the score if the operation is a ZADD (doincrement == 0) or + * the increment if the operation is a ZINCRBY (doincrement == 1). */ +static void zaddGenericCommand(redisClient *c, robj *key, robj *ele, double scoreval, int doincrement) { + robj *zsetobj; + zset *zs; + double *score; + + zsetobj = lookupKeyWrite(c->db,key); + if (zsetobj == NULL) { + zsetobj = createZsetObject(); + dictAdd(c->db->dict,key,zsetobj); + incrRefCount(key); + } else { + if (zsetobj->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + return; + } + } + zs = zsetobj->ptr; - if (subst->encoding == REDIS_ENCODING_RAW) - incrRefCount(subst); - else { - subst = getDecodedObject(subst); + /* Ok now since we implement both ZADD and ZINCRBY here the code + * needs to handle the two different conditions. It's all about setting + * '*score', that is, the new score to set, to the right value. */ + score = zmalloc(sizeof(double)); + if (doincrement) { + dictEntry *de; + + /* Read the old score. If the element was not present starts from 0 */ + de = dictFind(zs->dict,ele); + if (de) { + double *oldscore = dictGetEntryVal(de); + *score = *oldscore + scoreval; + } else { + *score = scoreval; + } + } else { + *score = scoreval; } - spat = pattern->ptr; - ssub = subst->ptr; - if (sdslen(spat)+sdslen(ssub)-1 > REDIS_SORTKEY_MAX) return NULL; - p = strchr(spat,'*'); - if (!p) return NULL; + /* What follows is a simple remove and re-insert operation that is common + * to both ZADD and ZINCRBY... */ + if (dictAdd(zs->dict,ele,score) == DICT_OK) { + /* case 1: New element */ + incrRefCount(ele); /* added to hash */ + zslInsert(zs->zsl,*score,ele); + incrRefCount(ele); /* added to skiplist */ + server.dirty++; + if (doincrement) + addReplyDouble(c,*score); + else + addReply(c,shared.cone); + } else { + dictEntry *de; + double *oldscore; + + /* case 2: Score update operation */ + de = dictFind(zs->dict,ele); + assert(de != NULL); + oldscore = dictGetEntryVal(de); + if (*score != *oldscore) { + int deleted; + + /* Remove and insert the element in the skip list with new score */ + deleted = zslDelete(zs->zsl,*oldscore,ele); + assert(deleted != 0); + zslInsert(zs->zsl,*score,ele); + incrRefCount(ele); + /* Update the score in the hash table */ + dictReplace(zs->dict,ele,score); + server.dirty++; + } else { + zfree(score); + } + if (doincrement) + addReplyDouble(c,*score); + else + addReply(c,shared.czero); + } +} - prefixlen = p-spat; - sublen = sdslen(ssub); - postfixlen = sdslen(spat)-(prefixlen+1); - memcpy(keyname.buf,spat,prefixlen); - memcpy(keyname.buf+prefixlen,ssub,sublen); - memcpy(keyname.buf+prefixlen+sublen,p+1,postfixlen); - keyname.buf[prefixlen+sublen+postfixlen] = '\0'; - keyname.len = prefixlen+sublen+postfixlen; +static void zaddCommand(redisClient *c) { + double scoreval; - keyobj.refcount = 1; - keyobj.type = REDIS_STRING; - keyobj.ptr = ((char*)&keyname)+(sizeof(long)*2); + scoreval = strtod(c->argv[2]->ptr,NULL); + zaddGenericCommand(c,c->argv[1],c->argv[3],scoreval,0); +} - decrRefCount(subst); +static void zincrbyCommand(redisClient *c) { + double scoreval; - /* printf("lookup '%s' => %p\n", keyname.buf,de); */ - return lookupKeyRead(db,&keyobj); + scoreval = strtod(c->argv[2]->ptr,NULL); + zaddGenericCommand(c,c->argv[1],c->argv[3],scoreval,1); } -/* sortCompare() is used by qsort in sortCommand(). Given that qsort_r with - * the additional parameter is not standard but a BSD-specific we have to - * pass sorting parameters via the global 'server' structure */ -static int sortCompare(const void *s1, const void *s2) { - const redisSortObject *so1 = s1, *so2 = s2; - int cmp; +static void zremCommand(redisClient *c) { + robj *zsetobj; + zset *zs; - if (!server.sort_alpha) { - /* Numeric sorting. Here it's trivial as we precomputed scores */ - if (so1->u.score > so2->u.score) { - cmp = 1; - } else if (so1->u.score < so2->u.score) { - cmp = -1; - } else { - cmp = 0; - } + zsetobj = lookupKeyWrite(c->db,c->argv[1]); + if (zsetobj == NULL) { + addReply(c,shared.czero); } else { - /* Alphanumeric sorting */ - if (server.sort_bypattern) { - if (!so1->u.cmpobj || !so2->u.cmpobj) { - /* At least one compare object is NULL */ - if (so1->u.cmpobj == so2->u.cmpobj) - cmp = 0; - else if (so1->u.cmpobj == NULL) - cmp = -1; - else - cmp = 1; - } else { - /* We have both the objects, use strcoll */ - cmp = strcoll(so1->u.cmpobj->ptr,so2->u.cmpobj->ptr); - } - } else { - /* Compare elements directly */ - 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); - } + 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); } - return server.sort_desc ? -cmp : cmp; } -/* The SORT command is the most complex command in Redis. Warning: this code - * is optimized for speed and a bit less for readability */ -static void sortCommand(redisClient *c) { - list *operations; - int outputlen = 0; - int desc = 0, alpha = 0; - int limit_start = 0, limit_count = -1, start, end; - int j, dontsort = 0, vectorlen; - int getop = 0; /* GET operation counter */ - robj *sortval, *sortby = NULL; - redisSortObject *vector; /* Resulting vector to sort */ +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; - /* Lookup the key to sort. It must be of the right types */ - sortval = lookupKeyRead(c->db,c->argv[1]); - if (sortval == NULL) { - addReply(c,shared.nokeyerr); - return; - } - if (sortval->type != REDIS_SET && sortval->type != REDIS_LIST) { - addReply(c,shared.wrongtypeerr); - return; - } + zsetobj = lookupKeyWrite(c->db,c->argv[1]); + if (zsetobj == NULL) { + addReply(c,shared.czero); + } else { + long deleted; - /* Create a list of operations to perform for every sorted element. - * Operations can be GET/DEL/INCR/DECR */ - operations = listCreate(); - listSetFreeMethod(operations,zfree); - j = 2; + 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)); + } +} - /* Now we need to protect sortval incrementing its count, in the future - * SORT may have options able to overwrite/delete keys during the sorting - * and the sorted key itself may get destroied */ - incrRefCount(sortval); +static void zrangeGenericCommand(redisClient *c, int reverse) { + robj *o; + int start = atoi(c->argv[2]->ptr); + int end = atoi(c->argv[3]->ptr); - /* The SORT command has an SQL-alike syntax, parse it */ - while(j < c->argc) { + 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 { + double *score = dictGetEntryVal(de); + + addReplyDouble(c,*score); + } + } + } +} + +/* ========================= 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); +} + +static void flushallCommand(redisClient *c) { + server.dirty += emptyDb(); + addReply(c,shared.ok); + rdbSave(server.dbfilename); + server.dirty++; +} + +static redisSortOperation *createSortOperation(int type, robj *pattern) { + redisSortOperation *so = zmalloc(sizeof(*so)); + so->type = type; + so->pattern = pattern; + return so; +} + +/* Return the value associated to the key with a name obtained + * substituting the first occurence of '*' in 'pattern' with 'subst' */ +static robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { + char *p; + sds spat, ssub; + robj keyobj; + int prefixlen, sublen, postfixlen; + /* Expoit the internal sds representation to create a sds string allocated on the stack in order to make this function faster */ + struct { + long len; + long free; + char buf[REDIS_SORTKEY_MAX+1]; + } keyname; + + /* If the pattern is "#" return the substitution object itself in order + * to implement the "SORT ... GET #" feature. */ + spat = pattern->ptr; + if (spat[0] == '#' && spat[1] == '\0') { + return subst; + } + + /* The substitution object may be specially encoded. If so we create + * a decoded object on the fly. Otherwise getDecodedObject will just + * increment the ref count, that we'll decrement later. */ + subst = getDecodedObject(subst); + + ssub = subst->ptr; + if (sdslen(spat)+sdslen(ssub)-1 > REDIS_SORTKEY_MAX) return NULL; + p = strchr(spat,'*'); + if (!p) { + decrRefCount(subst); + return NULL; + } + + prefixlen = p-spat; + sublen = sdslen(ssub); + postfixlen = sdslen(spat)-(prefixlen+1); + memcpy(keyname.buf,spat,prefixlen); + memcpy(keyname.buf+prefixlen,ssub,sublen); + memcpy(keyname.buf+prefixlen+sublen,p+1,postfixlen); + keyname.buf[prefixlen+sublen+postfixlen] = '\0'; + keyname.len = prefixlen+sublen+postfixlen; + + keyobj.refcount = 1; + 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); +} + +/* sortCompare() is used by qsort in sortCommand(). Given that qsort_r with + * the additional parameter is not standard but a BSD-specific we have to + * pass sorting parameters via the global 'server' structure */ +static int sortCompare(const void *s1, const void *s2) { + const redisSortObject *so1 = s1, *so2 = s2; + int cmp; + + if (!server.sort_alpha) { + /* Numeric sorting. Here it's trivial as we precomputed scores */ + if (so1->u.score > so2->u.score) { + cmp = 1; + } else if (so1->u.score < so2->u.score) { + cmp = -1; + } else { + cmp = 0; + } + } else { + /* Alphanumeric sorting */ + if (server.sort_bypattern) { + if (!so1->u.cmpobj || !so2->u.cmpobj) { + /* At least one compare object is NULL */ + if (so1->u.cmpobj == so2->u.cmpobj) + cmp = 0; + else if (so1->u.cmpobj == NULL) + cmp = -1; + else + cmp = 1; + } else { + /* We have both the objects, use strcoll */ + cmp = strcoll(so1->u.cmpobj->ptr,so2->u.cmpobj->ptr); + } + } else { + /* Compare elements directly */ + robj *dec1, *dec2; + + dec1 = getDecodedObject(so1->obj); + dec2 = getDecodedObject(so2->obj); + cmp = strcoll(dec1->ptr,dec2->ptr); + decrRefCount(dec1); + decrRefCount(dec2); + } + } + return server.sort_desc ? -cmp : cmp; +} + +/* The SORT command is the most complex command in Redis. Warning: this code + * is optimized for speed and a bit less for readability */ +static void sortCommand(redisClient *c) { + list *operations; + int outputlen = 0; + int desc = 0, alpha = 0; + int limit_start = 0, limit_count = -1, start, end; + int j, dontsort = 0, vectorlen; + int getop = 0; /* GET operation counter */ + robj *sortval, *sortby = NULL, *storekey = NULL; + redisSortObject *vector; /* Resulting vector to sort */ + + /* Lookup the key to sort. It must be of the right types */ + sortval = lookupKeyRead(c->db,c->argv[1]); + if (sortval == NULL) { + addReply(c,shared.nokeyerr); + return; + } + if (sortval->type != REDIS_SET && sortval->type != REDIS_LIST) { + addReply(c,shared.wrongtypeerr); + return; + } + + /* Create a list of operations to perform for every sorted element. + * Operations can be GET/DEL/INCR/DECR */ + operations = listCreate(); + listSetFreeMethod(operations,zfree); + j = 2; + + /* Now we need to protect sortval incrementing its count, in the future + * SORT may have options able to overwrite/delete keys during the sorting + * and the sorted key itself may get destroied */ + incrRefCount(sortval); + + /* The SORT command has an SQL-alike syntax, parse it */ + while(j < c->argc) { int leftargs = c->argc-j-1; if (!strcasecmp(c->argv[j]->ptr,"asc")) { desc = 0; @@ -3635,6 +4813,9 @@ 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, @@ -3646,18 +4827,6 @@ 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); @@ -3672,7 +4841,6 @@ 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; @@ -3692,7 +4860,6 @@ 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; @@ -3712,19 +4879,17 @@ static void sortCommand(redisClient *c) { byval = lookupKeyByPattern(c->db,sortby,vector[j].obj); if (!byval || byval->type != REDIS_STRING) continue; if (alpha) { - if (byval->encoding == REDIS_ENCODING_RAW) { - vector[j].u.cmpobj = byval; - incrRefCount(byval); - } else { - vector[j].u.cmpobj = getDecodedObject(byval); - } + vector[j].u.cmpobj = getDecodedObject(byval); } else { if (byval->encoding == REDIS_ENCODING_RAW) { vector[j].u.score = strtod(byval->ptr,NULL); } else { - if (byval->encoding == REDIS_ENCODING_INT) + /* Don't need to decode the object if it's + * integer-encoded (the only encoding supported) so + * far. We can just cast it */ + if (byval->encoding == REDIS_ENCODING_INT) { vector[j].u.score = (long)byval->ptr; - else + } else assert(1 != 1); } } @@ -3766,32 +4931,72 @@ static void sortCommand(redisClient *c) { /* 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; - if (!getop) { - addReplyBulkLen(c,vector[j].obj); - 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 */ + } + } } - listRewind(operations); - while((ln = listYield(operations))) { - 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 { - addReplyBulkLen(c,val); - addReply(c,val); - addReply(c,shared.crlf); + assert(sop->type == REDIS_SORT_GET); /* always fails */ } - } else if (sop->type == REDIS_SORT_DEL) { - /* TODO */ } } + 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 */ @@ -3811,6 +5016,7 @@ static void infoCommand(redisClient *c) { 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" @@ -3823,13 +5029,14 @@ static void infoCommand(redisClient *c) { "total_commands_processed:%lld\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.bgsavechildpid != -1, server.lastsave, server.stat_numconnections, server.stat_numcommands, @@ -3845,7 +5052,7 @@ static void infoCommand(redisClient *c) { server.masterport, (server.replstate == REDIS_REPL_CONNECTED) ? "up" : "down", - (int)(time(NULL)-server.master->lastinteraction) + server.master ? ((int)(time(NULL)-server.master->lastinteraction)) : -1 ); } for (j = 0; j < server.dbnum; j++) { @@ -3854,7 +5061,7 @@ static void infoCommand(redisClient *c) { 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", + info = sdscatprintf(info, "db%d:keys=%lld,expires=%lld\r\n", j, keys, vkeys); } } @@ -3869,7 +5076,7 @@ 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); } @@ -3933,21 +5140,21 @@ static int deleteIfVolatile(redisDb *db, robj *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); server.dirty++; } else { @@ -3957,6 +5164,14 @@ static void expireCommand(redisClient *c) { } } +static void expireCommand(redisClient *c) { + expireGenericCommand(c,c->argv[1],strtol(c->argv[2]->ptr,NULL,10)); +} + +static void expireatCommand(redisClient *c) { + expireGenericCommand(c,c->argv[1],strtol(c->argv[2]->ptr,NULL,10)-time(NULL)); +} + static void ttlCommand(redisClient *c) { time_t expire; int ttl = -1; @@ -4049,7 +5264,7 @@ static void syncCommand(redisClient *c) { 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) { + if (server.bgsavechildpid != -1) { /* 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 */ @@ -4066,7 +5281,6 @@ static void syncCommand(redisClient *c) { * another slave. Set the right state, and copy the buffer. */ listRelease(c->reply); c->reply = listDup(slave->reply); - if (!c->reply) oom("listDup copying slave reply list"); c->replstate = REDIS_REPL_WAIT_BGSAVE_END; redisLog(REDIS_NOTICE,"Waiting for end of BGSAVE for SYNC"); } else { @@ -4088,7 +5302,7 @@ static void syncCommand(redisClient *c) { c->repldbfd = -1; c->flags |= REDIS_SLAVE; c->slaveseldb = 0; - if (!listAddNodeTail(server.slaves,c)) oom("listAddNodeTail"); + listAddNodeTail(server.slaves,c); return; } @@ -4137,7 +5351,7 @@ static void sendBulkToSlave(aeEventLoop *el, int fd, void *privdata, int mask) { 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) { + sendReplyToClient, slave) == AE_ERR) { freeClient(slave); return; } @@ -4181,7 +5395,7 @@ static void updateSlavesWaitingBgsave(int bgsaveerr) { 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) { + if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, sendBulkToSlave, slave) == AE_ERR) { freeClient(slave); continue; } @@ -4202,7 +5416,7 @@ static void updateSlavesWaitingBgsave(int bgsaveerr) { } 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; @@ -4212,6 +5426,30 @@ 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); @@ -4226,6 +5464,11 @@ static int syncWithMaster(void) { 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 */ @@ -4297,56 +5540,490 @@ static void slaveofCommand(redisClient *c) { 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 +/* ============================ 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]; + + 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); + 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); + } + /* If a background append only file rewriting is in progress we want to + * accumulate the differences between the child DB and the current one + * in a buffer, so that when the child process will do its work we + * can append the differences to the new append only file. */ + if (server.bgrewritechildpid != -1) + server.bgrewritebuf = sdscatlen(server.bgrewritebuf,buf,sdslen(buf)); + + sdsfree(buf); + 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 (len && 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); +} + +/* Write an object into a file in the bulk format $\r\n\r\n */ +static int fwriteBulk(FILE *fp, robj *obj) { + char buf[128]; + obj = getDecodedObject(obj); + snprintf(buf,sizeof(buf),"$%ld\r\n",(long)sdslen(obj->ptr)); + if (fwrite(buf,strlen(buf),1,fp) == 0) goto err; + if (fwrite(obj->ptr,sdslen(obj->ptr),1,fp) == 0) goto err; + if (fwrite("\r\n",2,1,fp) == 0) goto err; + decrRefCount(obj); + return 1; +err: + decrRefCount(obj); + return 0; +} + +/* Write a double value in bulk format $\r\n\r\n */ +static int fwriteBulkDouble(FILE *fp, double d) { + char buf[128], dbuf[128]; + + snprintf(dbuf,sizeof(dbuf),"%.17g\r\n",d); + snprintf(buf,sizeof(buf),"$%lu\r\n",(unsigned long)strlen(dbuf)-2); + if (fwrite(buf,strlen(buf),1,fp) == 0) return 0; + if (fwrite(dbuf,strlen(dbuf),1,fp) == 0) return 0; + return 1; +} + +/* Write a long value in bulk format $\r\n\r\n */ +static int fwriteBulkLong(FILE *fp, long l) { + char buf[128], lbuf[128]; + + snprintf(lbuf,sizeof(lbuf),"%ld\r\n",l); + snprintf(buf,sizeof(buf),"$%lu\r\n",(unsigned long)strlen(lbuf)-2); + if (fwrite(buf,strlen(buf),1,fp) == 0) return 0; + if (fwrite(lbuf,strlen(lbuf),1,fp) == 0) return 0; + return 1; +} + +/* Write a sequence of commands able to fully rebuild the dataset into + * "filename". Used both by REWRITEAOF and BGREWRITEAOF. */ +static int rewriteAppendOnlyFile(char *filename) { + dictIterator *di = NULL; + dictEntry *de; + FILE *fp; + char tmpfile[256]; + int j; + time_t now = time(NULL); + + /* Note that we have to use a different temp name here compared to the + * one used by rewriteAppendOnlyFileBackground() function. */ + snprintf(tmpfile,256,"temp-rewriteaof-%d.aof", (int) getpid()); + fp = fopen(tmpfile,"w"); + if (!fp) { + redisLog(REDIS_WARNING, "Failed rewriting the append only file: %s", strerror(errno)); + return REDIS_ERR; + } + for (j = 0; j < server.dbnum; j++) { + char selectcmd[] = "*2\r\n$6\r\nSELECT\r\n"; + redisDb *db = server.db+j; + dict *d = db->dict; + if (dictSize(d) == 0) continue; + di = dictGetIterator(d); + if (!di) { + fclose(fp); + return REDIS_ERR; + } + + /* SELECT the new DB */ + if (fwrite(selectcmd,sizeof(selectcmd)-1,1,fp) == 0) goto werr; + if (fwriteBulkLong(fp,j) == 0) goto werr; + + /* Iterate this DB writing every entry */ + while((de = dictNext(di)) != NULL) { + robj *key = dictGetEntryKey(de); + robj *o = dictGetEntryVal(de); + time_t expiretime = getExpire(db,key); + + /* Save the key and associated value */ + if (o->type == REDIS_STRING) { + /* Emit a SET command */ + char cmd[]="*3\r\n$3\r\nSET\r\n"; + if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr; + /* Key and value */ + if (fwriteBulk(fp,key) == 0) goto werr; + if (fwriteBulk(fp,o) == 0) goto werr; + } else if (o->type == REDIS_LIST) { + /* Emit the RPUSHes needed to rebuild the list */ + list *list = o->ptr; + listNode *ln; + + listRewind(list); + while((ln = listYield(list))) { + char cmd[]="*3\r\n$5\r\nRPUSH\r\n"; + robj *eleobj = listNodeValue(ln); + + if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr; + if (fwriteBulk(fp,key) == 0) goto werr; + if (fwriteBulk(fp,eleobj) == 0) goto werr; + } + } else if (o->type == REDIS_SET) { + /* Emit the SADDs needed to rebuild the set */ + dict *set = o->ptr; + dictIterator *di = dictGetIterator(set); + dictEntry *de; + + while((de = dictNext(di)) != NULL) { + char cmd[]="*3\r\n$4\r\nSADD\r\n"; + robj *eleobj = dictGetEntryKey(de); + + if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr; + if (fwriteBulk(fp,key) == 0) goto werr; + if (fwriteBulk(fp,eleobj) == 0) goto werr; + } + dictReleaseIterator(di); + } else if (o->type == REDIS_ZSET) { + /* Emit the ZADDs needed to rebuild the sorted set */ + zset *zs = o->ptr; + dictIterator *di = dictGetIterator(zs->dict); + dictEntry *de; + + while((de = dictNext(di)) != NULL) { + char cmd[]="*4\r\n$4\r\nZADD\r\n"; + robj *eleobj = dictGetEntryKey(de); + double *score = dictGetEntryVal(de); + + if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr; + if (fwriteBulk(fp,key) == 0) goto werr; + if (fwriteBulkDouble(fp,*score) == 0) goto werr; + if (fwriteBulk(fp,eleobj) == 0) goto werr; + } + dictReleaseIterator(di); + } else { + assert(0 != 0); + } + /* Save the expire time */ + if (expiretime != -1) { + char cmd[]="*3\r\n$6\r\nEXPIRE\r\n"; + /* If this key is already expired skip it */ + if (expiretime < now) continue; + if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr; + if (fwriteBulk(fp,key) == 0) goto werr; + if (fwriteBulkLong(fp,expiretime) == 0) goto werr; + } + } + dictReleaseIterator(di); + } + + /* Make sure data will not remain on the OS's output buffers */ + fflush(fp); + fsync(fileno(fp)); + fclose(fp); + + /* 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 append only file on the final destination: %s", strerror(errno)); + unlink(tmpfile); + return REDIS_ERR; + } + redisLog(REDIS_NOTICE,"SYNC append only file rewrite performed"); + return REDIS_OK; + +werr: + fclose(fp); + unlink(tmpfile); + redisLog(REDIS_WARNING,"Write error writing append only fileon disk: %s", strerror(errno)); + if (di) dictReleaseIterator(di); + return REDIS_ERR; +} + +/* This is how rewriting of the append only file in background works: * - * 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. + * 1) The user calls BGREWRITEAOF + * 2) Redis calls this function, that forks(): + * 2a) the child rewrite the append only file in a temp file. + * 2b) the parent accumulates differences in server.bgrewritebuf. + * 3) When the child finished '2a' exists. + * 4) The parent will trap the exit code, if it's OK, will append the + * data accumulated into server.bgrewritebuf into the temp file, and + * finally will rename(2) the temp file in the actual file name. + * The the new file is reopened as the new append only file. Profit! */ -static void freeMemoryIfNeeded(void) { - while (server.maxmemory && zmalloc_used_memory() > server.maxmemory) { - if (listLength(server.objfreelist)) { - robj *o; +static int rewriteAppendOnlyFileBackground(void) { + pid_t childpid; - listNode *head = listFirst(server.objfreelist); - o = listNodeValue(head); - listDelNode(server.objfreelist,head); - zfree(o); + if (server.bgrewritechildpid != -1) return REDIS_ERR; + if ((childpid = fork()) == 0) { + /* Child */ + char tmpfile[256]; + close(server.fd); + + snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) getpid()); + if (rewriteAppendOnlyFile(tmpfile) == REDIS_OK) { + exit(0); } else { - int j, k, freed = 0; + exit(1); + } + } else { + /* Parent */ + if (childpid == -1) { + redisLog(REDIS_WARNING, + "Can't rewrite append only file in background: fork: %s", + strerror(errno)); + return REDIS_ERR; + } + redisLog(REDIS_NOTICE, + "Background append only file rewriting started by pid %d",childpid); + server.bgrewritechildpid = childpid; + /* We set appendseldb to -1 in order to force the next call to the + * feedAppendOnlyFile() to issue a SELECT command, so the differences + * accumulated by the parent into server.bgrewritebuf will start + * with a SELECT statement and it will be safe to merge. */ + server.appendseldb = -1; + return REDIS_OK; + } + return REDIS_OK; /* unreached */ +} - for (j = 0; j < server.dbnum; j++) { - int minttl = -1; - robj *minkey = NULL; - struct dictEntry *de; +static void bgrewriteaofCommand(redisClient *c) { + if (server.bgrewritechildpid != -1) { + addReplySds(c,sdsnew("-ERR background append only file rewriting already in progress\r\n")); + return; + } + if (rewriteAppendOnlyFileBackground() == REDIS_OK) { + addReply(c,shared.ok); + } else { + addReply(c,shared.err); + } +} - 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; +static void aofRemoveTempFile(pid_t childpid) { + char tmpfile[256]; - 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... */ - } - } + snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) childpid); + unlink(tmpfile); } /* ================================= Debugging ============================== */ @@ -4354,6 +6031,18 @@ static void freeMemoryIfNeeded(void) { static void debugCommand(redisClient *c) { if (!strcasecmp(c->argv[1]->ptr,"segfault")) { *((char*)-1) = 'x'; + } else if (!strcasecmp(c->argv[1]->ptr,"reload")) { + if (rdbSave(server.dbfilename) != REDIS_OK) { + addReply(c,shared.err); + return; + } + emptyDb(); + if (rdbLoad(server.dbfilename) != REDIS_OK) { + addReply(c,shared.err); + return; + } + redisLog(REDIS_WARNING,"DB reloaded by DEBUG RELOAD"); + addReply(c,shared.ok); } else if (!strcasecmp(c->argv[1]->ptr,"object") && c->argc == 3) { dictEntry *de = dictFind(c->db->dict,c->argv[2]); robj *key, *val; @@ -4369,131 +6058,95 @@ static void debugCommand(redisClient *c) { key, key->refcount, val, val->refcount, val->encoding)); } else { addReplySds(c,sdsnew( - "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT ]\r\n")); + "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT |RELOAD]\r\n")); } } -#ifdef HAVE_BACKTRACE -static struct redisFunctionSym symsTable[] = { -{"dictEncObjKeyCompare", (unsigned long)dictEncObjKeyCompare}, -{"dictEncObjHash", (unsigned long)dictEncObjHash}, -{"incrDecrCommand", (unsigned long)incrDecrCommand}, -{"freeStringObject", (unsigned long)freeStringObject}, -{"freeListObject", (unsigned long)freeListObject}, -{"freeSetObject", (unsigned long)freeSetObject}, -{"decrRefCount", (unsigned long)decrRefCount}, -{"createObject", (unsigned long)createObject}, -{"freeClient", (unsigned long)freeClient}, -{"rdbLoad", (unsigned long)rdbLoad}, -{"rdbSaveStringObject", (unsigned long)rdbSaveStringObject}, -{"rdbSaveStringObjectRaw", (unsigned long)rdbSaveStringObjectRaw}, -{"addReply", (unsigned long)addReply}, -{"addReplySds", (unsigned long)addReplySds}, -{"incrRefCount", (unsigned long)incrRefCount}, -{"rdbSaveBackground", (unsigned long)rdbSaveBackground}, -{"createStringObject", (unsigned long)createStringObject}, -{"replicationFeedSlaves", (unsigned long)replicationFeedSlaves}, -{"syncWithMaster", (unsigned long)syncWithMaster}, -{"tryObjectSharing", (unsigned long)tryObjectSharing}, -{"tryObjectEncoding", (unsigned long)tryObjectEncoding}, -{"getDecodedObject", (unsigned long)getDecodedObject}, -{"removeExpire", (unsigned long)removeExpire}, -{"expireIfNeeded", (unsigned long)expireIfNeeded}, -{"deleteIfVolatile", (unsigned long)deleteIfVolatile}, -{"deleteKey", (unsigned long)deleteKey}, -{"getExpire", (unsigned long)getExpire}, -{"setExpire", (unsigned long)setExpire}, -{"updateSlavesWaitingBgsave", (unsigned long)updateSlavesWaitingBgsave}, -{"freeMemoryIfNeeded", (unsigned long)freeMemoryIfNeeded}, -{"authCommand", (unsigned long)authCommand}, -{"pingCommand", (unsigned long)pingCommand}, -{"echoCommand", (unsigned long)echoCommand}, -{"setCommand", (unsigned long)setCommand}, -{"setnxCommand", (unsigned long)setnxCommand}, -{"getCommand", (unsigned long)getCommand}, -{"delCommand", (unsigned long)delCommand}, -{"existsCommand", (unsigned long)existsCommand}, -{"incrCommand", (unsigned long)incrCommand}, -{"decrCommand", (unsigned long)decrCommand}, -{"incrbyCommand", (unsigned long)incrbyCommand}, -{"decrbyCommand", (unsigned long)decrbyCommand}, -{"selectCommand", (unsigned long)selectCommand}, -{"randomkeyCommand", (unsigned long)randomkeyCommand}, -{"keysCommand", (unsigned long)keysCommand}, -{"dbsizeCommand", (unsigned long)dbsizeCommand}, -{"lastsaveCommand", (unsigned long)lastsaveCommand}, -{"saveCommand", (unsigned long)saveCommand}, -{"bgsaveCommand", (unsigned long)bgsaveCommand}, -{"shutdownCommand", (unsigned long)shutdownCommand}, -{"moveCommand", (unsigned long)moveCommand}, -{"renameCommand", (unsigned long)renameCommand}, -{"renamenxCommand", (unsigned long)renamenxCommand}, -{"lpushCommand", (unsigned long)lpushCommand}, -{"rpushCommand", (unsigned long)rpushCommand}, -{"lpopCommand", (unsigned long)lpopCommand}, -{"rpopCommand", (unsigned long)rpopCommand}, -{"llenCommand", (unsigned long)llenCommand}, -{"lindexCommand", (unsigned long)lindexCommand}, -{"lrangeCommand", (unsigned long)lrangeCommand}, -{"ltrimCommand", (unsigned long)ltrimCommand}, -{"typeCommand", (unsigned long)typeCommand}, -{"lsetCommand", (unsigned long)lsetCommand}, -{"saddCommand", (unsigned long)saddCommand}, -{"sremCommand", (unsigned long)sremCommand}, -{"smoveCommand", (unsigned long)smoveCommand}, -{"sismemberCommand", (unsigned long)sismemberCommand}, -{"scardCommand", (unsigned long)scardCommand}, -{"spopCommand", (unsigned long)spopCommand}, -{"sinterCommand", (unsigned long)sinterCommand}, -{"sinterstoreCommand", (unsigned long)sinterstoreCommand}, -{"sunionCommand", (unsigned long)sunionCommand}, -{"sunionstoreCommand", (unsigned long)sunionstoreCommand}, -{"sdiffCommand", (unsigned long)sdiffCommand}, -{"sdiffstoreCommand", (unsigned long)sdiffstoreCommand}, -{"syncCommand", (unsigned long)syncCommand}, -{"flushdbCommand", (unsigned long)flushdbCommand}, -{"flushallCommand", (unsigned long)flushallCommand}, -{"sortCommand", (unsigned long)sortCommand}, -{"lremCommand", (unsigned long)lremCommand}, -{"infoCommand", (unsigned long)infoCommand}, -{"mgetCommand", (unsigned long)mgetCommand}, -{"monitorCommand", (unsigned long)monitorCommand}, -{"expireCommand", (unsigned long)expireCommand}, -{"getSetCommand", (unsigned long)getSetCommand}, -{"ttlCommand", (unsigned long)ttlCommand}, -{"slaveofCommand", (unsigned long)slaveofCommand}, -{"debugCommand", (unsigned long)debugCommand}, -{"processCommand", (unsigned long)processCommand}, -{"setupSigSegvAction", (unsigned long)setupSigSegvAction}, -{"readQueryFromClient", (unsigned long)readQueryFromClient}, -{"rdbRemoveTempFile", (unsigned long)rdbRemoveTempFile}, -{NULL,0} -}; +/* =================================== Main! ================================ */ -/* This function try to convert a pointer into a function name. It's used in - * oreder to provide a backtrace under segmentation fault that's able to - * display functions declared as static (otherwise the backtrace is useless). */ -static char *findFuncName(void *pointer, unsigned long *offset){ - int i, ret = -1; - unsigned long off, minoff = 0; +#ifdef __linux__ +int linuxOvercommitMemoryValue(void) { + FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r"); + char buf[64]; - /* Try to match against the Symbol with the smallest offset */ - for (i=0; symsTable[i].pointer; i++) { - unsigned long lp = (unsigned long) pointer; + if (!fp) return -1; + if (fgets(buf,64,fp) == NULL) { + fclose(fp); + return -1; + } + fclose(fp); - if (lp != (unsigned long)-1 && lp >= symsTable[i].pointer) { - off=lp-symsTable[i].pointer; - if (ret < 0 || off < minoff) { - minoff=off; - ret=i; - } - } + 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."); } - if (ret == -1) return NULL; - *offset = minoff; - return symsTable[ret].name; +} +#endif /* __linux__ */ + +static void daemonize(void) { + int fd; + FILE *fp; + + if (fork() != 0) exit(0); /* parent exits */ + setsid(); /* create a new session */ + + /* Every output goes to /dev/null. If Redis is daemonized but + * the 'logfile' is set to 'stdout' in the configuration file + * it will not log at all. */ + if ((fd = open("/dev/null", O_RDWR, 0)) != -1) { + dup2(fd, STDIN_FILENO); + dup2(fd, STDOUT_FILENO); + dup2(fd, STDERR_FILENO); + if (fd > STDERR_FILENO) close(fd); + } + /* Try to write the pid file */ + fp = fopen(server.pidfile,"w"); + if (fp) { + fprintf(fp,"%d\n",getpid()); + fclose(fp); + } +} + +int main(int argc, char **argv) { + initServerConfig(); + if (argc == 2) { + 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); +#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) == AE_ERR) oom("creating file event"); + redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port); + aeMain(server.el); + 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; @@ -4502,7 +6155,7 @@ static void *getMcontextEip(ucontext_t *uc) { #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) - #ifdef _STRUCT_X86_THREAD_STATE64 + #if defined(_STRUCT_X86_THREAD_STATE64) && !defined(__i386__) return (void*) uc->uc_mcontext->__ss.__rip; #else return (void*) uc->uc_mcontext->__ss.__eip; @@ -4545,7 +6198,7 @@ static void segvHandler(int sig, siginfo_t *info, void *secret) { listLength(server.slaves), server.usedmemory, server.dirty, - server.bgsaveinprogress, + server.bgsavechildpid != -1, server.lastsave, server.stat_numconnections, server.stat_numcommands, @@ -4588,82 +6241,39 @@ static void setupSigSegvAction(void) { sigaction (SIGBUS, &act, NULL); return; } -#else /* HAVE_BACKTRACE */ -static void setupSigSegvAction(void) { -} -#endif /* HAVE_BACKTRACE */ -/* =================================== Main! ================================ */ +#include "staticsymbols.h" +/* This function try to convert a pointer into a function name. It's used in + * oreder to provide a backtrace under segmentation fault that's able to + * display functions declared as static (otherwise the backtrace is useless). */ +static char *findFuncName(void *pointer, unsigned long *offset){ + int i, ret = -1; + unsigned long off, minoff = 0; -#ifdef __linux__ -int linuxOvercommitMemoryValue(void) { - FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r"); - char buf[64]; + /* Try to match against the Symbol with the smallest offset */ + for (i=0; symsTable[i].pointer; i++) { + unsigned long lp = (unsigned long) pointer; - if (!fp) return -1; - if (fgets(buf,64,fp) == NULL) { - fclose(fp); - return -1; + if (lp != (unsigned long)-1 && lp >= symsTable[i].pointer) { + off=lp-symsTable[i].pointer; + if (ret < 0 || off < minoff) { + minoff=off; + ret=i; + } + } } - fclose(fp); - - return atoi(buf); + if (ret == -1) return NULL; + *offset = minoff; + return symsTable[ret].name; } - -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."); - } +#else /* HAVE_BACKTRACE */ +static void setupSigSegvAction(void) { } -#endif /* __linux__ */ +#endif /* HAVE_BACKTRACE */ -static void daemonize(void) { - int fd; - FILE *fp; - if (fork() != 0) exit(0); /* parent exits */ - setsid(); /* create a new session */ - /* Every output goes to /dev/null. If Redis is daemonized but - * the 'logfile' is set to 'stdout' in the configuration file - * it will not log at all. */ - if ((fd = open("/dev/null", O_RDWR, 0)) != -1) { - dup2(fd, STDIN_FILENO); - dup2(fd, STDOUT_FILENO); - dup2(fd, STDERR_FILENO); - if (fd > STDERR_FILENO) close(fd); - } - /* Try to write the pid file */ - fp = fopen(server.pidfile,"w"); - if (fp) { - fprintf(fp,"%d\n",getpid()); - fclose(fp); - } -} +/* The End */ + + -int main(int argc, char **argv) { - initServerConfig(); - if (argc == 2) { - 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); -#ifdef __linux__ - linuxOvercommitMemoryWarning(); -#endif - 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); - aeMain(server.el); - aeDeleteEventLoop(server.el); - return 0; -}