X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/325d1eb4a86b59eaa8c867b2bb0a7a827d98aaf0..b892cabed4079de8a1c6ac18041e3ea835c80300:/redis.c diff --git a/redis.c b/redis.c index eb0a1730..df3b4131 100644 --- a/redis.c +++ b/redis.c @@ -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% */ @@ -167,11 +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 */ @@ -185,6 +193,11 @@ #define ZSKIPLIST_MAXLEVEL 32 /* Should be enough for 2^32 elements */ #define ZSKIPLIST_P 0.25 /* Skiplist P = 1/4 */ +/* Append only defines */ +#define APPENDFSYNC_NO 0 +#define APPENDFSYNC_ALWAYS 1 +#define APPENDFSYNC_EVERYSEC 2 + /*================================= Data types ============================== */ /* A redis object, that is a type able to hold a string / list / set */ @@ -256,6 +269,11 @@ 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; @@ -264,10 +282,12 @@ struct redisServer { 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 */ @@ -318,7 +338,7 @@ typedef struct zskiplistNode { typedef struct zskiplist { struct zskiplistNode *header, *tail; - long length; + unsigned long length; int level; } zskiplist; @@ -338,6 +358,12 @@ struct sharedObjectsStruct { *select5, *select6, *select7, *select8, *select9; } shared; +/* Global vars that are actally used as constants. The following double + * values are used for double on-disk serialization, and are initialized + * at runtime to avoid strange compiler optimizations. */ + +static double R_Zero, R_PosInf, R_NegInf, R_Nan; + /*================================ Prototypes =============================== */ static void freeStringObject(robj *o); @@ -353,6 +379,7 @@ 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); @@ -372,6 +399,8 @@ 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); @@ -424,10 +453,12 @@ static void flushdbCommand(redisClient *c); static void flushallCommand(redisClient *c); static void sortCommand(redisClient *c); static void lremCommand(redisClient *c); +static void rpoplpushcommand(redisClient *c); static void infoCommand(redisClient *c); static void mgetCommand(redisClient *c); static void monitorCommand(redisClient *c); static void expireCommand(redisClient *c); +static void expireatCommand(redisClient *c); static void getsetCommand(redisClient *c); static void ttlCommand(redisClient *c); static void slaveofCommand(redisClient *c); @@ -436,9 +467,12 @@ static void msetCommand(redisClient *c); static void msetnxCommand(redisClient *c); static void zaddCommand(redisClient *c); static void zrangeCommand(redisClient *c); +static void zrangebyscoreCommand(redisClient *c); static void zrevrangeCommand(redisClient *c); -static void zlenCommand(redisClient *c); +static void zcardCommand(redisClient *c); static void zremCommand(redisClient *c); +static void zscoreCommand(redisClient *c); +static void zremrangebyscoreCommand(redisClient *c); /*================================= Globals ================================= */ @@ -463,6 +497,7 @@ 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}, @@ -479,9 +514,12 @@ static struct redisCommand cmdTable[] = { {"smembers",sinterCommand,2,REDIS_CMD_INLINE}, {"zadd",zaddCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM}, {"zrem",zremCommand,3,REDIS_CMD_BULK}, + {"zremrangebyscore",zremrangebyscoreCommand,4,REDIS_CMD_INLINE}, {"zrange",zrangeCommand,4,REDIS_CMD_INLINE}, + {"zrangebyscore",zrangebyscoreCommand,4,REDIS_CMD_INLINE}, {"zrevrange",zrevrangeCommand,4,REDIS_CMD_INLINE}, - {"zlen",zlenCommand,2,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}, @@ -493,6 +531,7 @@ static struct redisCommand cmdTable[] = { {"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}, @@ -514,6 +553,7 @@ static struct redisCommand cmdTable[] = { {"debug",debugCommand,-2,REDIS_CMD_INLINE}, {NULL,NULL,0,0} }; + /*============================ Utility functions ============================ */ /* Glob-style pattern matching. */ @@ -653,7 +693,7 @@ 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"); @@ -873,7 +913,7 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD /* Check if a background saving in progress terminated */ if (server.bgsaveinprogress) { int statloc; - if (wait4(-1,&statloc,WNOHANG,NULL)) { + if (wait3(&statloc,WNOHANG,NULL)) { int exitcode = WEXITSTATUS(statloc); int bysignal = WIFSIGNALED(statloc); @@ -910,14 +950,21 @@ 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--) { @@ -928,9 +975,10 @@ 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 */ @@ -987,7 +1035,7 @@ static void appendServerSaveParams(time_t seconds, int changes) { server.saveparamslen++; } -static void ResetServerSaveParams() { +static void resetServerSaveParams() { zfree(server.saveparams); server.saveparams = NULL; server.saveparamslen = 0; @@ -1003,24 +1051,37 @@ static void initServerConfig() { server.bindaddr = NULL; server.glueoutputbuf = 1; server.daemonize = 0; + server.appendonly = 0; + server.appendfsync = APPENDFSYNC_ALWAYS; + server.lastfsync = time(NULL); + server.appendfd = -1; + server.appendseldb = -1; /* Make sure the first time will not match */ server.pidfile = "/var/run/redis.pid"; server.dbfilename = "dump.rdb"; + server.appendfilename = "appendonly.log"; server.requirepass = NULL; server.shareobjects = 0; 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() { @@ -1057,7 +1118,16 @@ static void initServer() { 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 */ @@ -1180,6 +1250,8 @@ static void loadServerConfig(char *filename) { 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; @@ -1197,6 +1269,21 @@ 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) { @@ -1262,34 +1349,31 @@ static void freeClient(redisClient *c) { 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)); - listAddNodeTail(c->reply,o); } + /* 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) { @@ -1299,9 +1383,22 @@ 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 0 + if (!server.glueoutputbuf && + listLength(c->reply) > REDIS_WRITEV_THRESHOLD && + !(c->flags & REDIS_MASTER)) + { + sendReplyToClientWritev(el, fd, privdata, mask); + return; + } + #endif + while(listLength(c->reply)) { + if (server.glueoutputbuf && listLength(c->reply) > 1) + glueReplyBuffersIfNeeded(c); + o = listNodeValue(listFirst(c->reply)); objlen = sdslen(o->ptr); @@ -1325,10 +1422,10 @@ 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) { @@ -1348,6 +1445,84 @@ 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) { @@ -1512,7 +1687,9 @@ 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); @@ -1811,7 +1988,9 @@ 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; } @@ -2056,24 +2235,31 @@ 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(). */ 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_INT && b->encoding == REDIS_ENCODING_INT){ - return (long)a->ptr - (long)b->ptr; + if (a->encoding != REDIS_ENCODING_RAW) { + snprintf(bufa,sizeof(bufa),"%ld",(long) a->ptr); + astr = bufa; + bothsds = 0; } else { - int retval; - - incrRefCount(a); - incrRefCount(b); - if (a->encoding != REDIS_ENCODING_RAW) a = getDecodedObject(a); - if (b->encoding != REDIS_ENCODING_RAW) b = getDecodedObject(a); - retval = sdscmp(a->ptr,b->ptr); - decrRefCount(a); - decrRefCount(b); - return retval; + 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) { @@ -2238,6 +2424,33 @@ static int rdbSaveStringObject(FILE *fp, robj *obj) { } } +/* Save a double value. Doubles are saved as strings prefixed by an unsigned + * 8 bit integer specifing the length of the representation. + * This 8 bit integer has special values in order to specify the following + * conditions: + * 253: not a number + * 254: + inf + * 255: - inf + */ +static int rdbSaveDoubleValue(FILE *fp, double val) { + unsigned char buf[128]; + int len; + + if (isnan(val)) { + buf[0] = 253; + len = 1; + } else if (!isfinite(val)) { + len = 1; + buf[0] = (val < 0) ? 255 : 254; + } else { + snprintf((char*)buf+1,sizeof(buf)-1,"%.17g",val); + buf[0] = strlen((char*)buf); + len = buf[0]+1; + } + if (fwrite(buf,len,1,fp) == 0) return -1; + return 0; +} + /* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success */ static int rdbSave(char *filename) { dictIterator *di = NULL; @@ -2312,6 +2525,21 @@ 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); } @@ -2500,6 +2728,23 @@ 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; @@ -2574,6 +2819,27 @@ static int rdbLoad(char *filename) { dictAdd((dict*)o->ptr,ele,NULL); } } + } else if (type == REDIS_ZSET) { + /* Read list/set value */ + uint32_t zsetlen; + zset *zs; + + if ((zsetlen = rdbLoadLen(fp,rdbver,NULL)) == REDIS_RDB_LENERR) + goto eoferr; + o = createZsetObject(); + zs = o->ptr; + /* Load every single element of the list/set */ + while(zsetlen--) { + robj *ele; + double *score = zmalloc(sizeof(double)); + + if ((ele = rdbLoadStringObject(fp,rdbver)) == NULL) goto eoferr; + tryObjectEncoding(ele); + if (rdbLoadDoubleValue(fp,score) == -1) goto eoferr; + dictAdd(zs->dict,ele,score); + zslInsert(zs->zsl,*score,ele); + incrRefCount(ele); /* added to skiplist */ + } } else { assert(0 != 0); } @@ -2597,7 +2863,7 @@ 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 */ } @@ -2610,7 +2876,7 @@ 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")); } } @@ -2868,6 +3134,7 @@ 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; } } @@ -3300,6 +3567,70 @@ 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) { @@ -3731,17 +4062,21 @@ static zskiplist *zslCreate(void) { static void zslFreeNode(zskiplistNode *node) { decrRefCount(node->obj); + zfree(node->forward); zfree(node); } static void zslFree(zskiplist *zsl) { - zskiplistNode *node = zsl->header->forward[1], *next; + zskiplistNode *node = zsl->header->forward[0], *next; + zfree(zsl->header->forward); + zfree(zsl->header); while(node) { - next = node->forward[1]; + next = node->forward[0]; zslFreeNode(node); node = next; } + zfree(zsl); } static int zslRandomLevel(void) { @@ -3757,7 +4092,10 @@ static void zslInsert(zskiplist *zsl, double score, robj *obj) { x = zsl->header; for (i = zsl->level-1; i >= 0; i--) { - while (x->forward[i] && x->forward[i]->score < score) + 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; } @@ -3776,7 +4114,7 @@ static void zslInsert(zskiplist *zsl, double score, robj *obj) { x->forward[i] = update[i]->forward[i]; update[i]->forward[i] = x; } - x->backward = (update[0] == zsl->header) ? NULL : update[i]; + x->backward = (update[0] == zsl->header) ? NULL : update[0]; if (x->forward[0]) x->forward[0]->backward = x; else @@ -3784,44 +4122,104 @@ static void zslInsert(zskiplist *zsl, double score, robj *obj) { 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) + 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]; - while(x->score == score) { - if (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; + 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 { - x = x->forward[0]; - if (!x) return 0; /* end of the list reached, not found */ + zsl->tail = x->backward; } + zslFreeNode(x); + while(zsl->level > 1 && zsl->header->forward[zsl->level-1] == NULL) + zsl->level--; + zsl->length--; + return 1; + } else { + return 0; /* not found */ } return 0; /* not found */ } +/* Delete all the elements with score between min and max from the skiplist. + * Min and mx are inclusive, so a score >= min || score <= max is deleted. + * Note that this function takes the reference to the hash table view of the + * sorted set, in order to remove the elements from the hash table too. */ +static unsigned long zslDeleteRange(zskiplist *zsl, double min, double max, dict *dict) { + zskiplistNode *update[ZSKIPLIST_MAXLEVEL], *x; + unsigned long removed = 0; + int i; + + x = zsl->header; + for (i = zsl->level-1; i >= 0; i--) { + while (x->forward[i] && x->forward[i]->score < min) + x = x->forward[i]; + update[i] = x; + } + /* We may have multiple elements with the same score, what we need + * is to find the element with both the right score and object. */ + x = x->forward[0]; + while (x && x->score <= max) { + zskiplistNode *next; + + for (i = 0; i < zsl->level; i++) { + if (update[i]->forward[i] != x) break; + update[i]->forward[i] = x->forward[i]; + } + if (x->forward[0]) { + x->forward[0]->backward = (x->backward == zsl->header) ? + NULL : x->backward; + } else { + zsl->tail = x->backward; + } + next = x->forward[0]; + dictDelete(dict,x->obj); + zslFreeNode(x); + while(zsl->level > 1 && zsl->header->forward[zsl->level-1] == NULL) + zsl->level--; + zsl->length--; + removed++; + x = next; + } + return removed; /* not found */ +} + +/* Find the first node having a score equal or greater than the specified one. + * Returns NULL if there is no match. */ +static zskiplistNode *zslFirstWithScore(zskiplist *zsl, double score) { + zskiplistNode *x; + int i; + + x = zsl->header; + for (i = zsl->level-1; i >= 0; i--) { + while (x->forward[i] && x->forward[i]->score < score) + x = x->forward[i]; + } + /* We may have multiple elements with the same score, what we need + * is to find the element with both the right score and object. */ + return x->forward[0]; +} + /* The actual Z-commands implementations */ static void zaddCommand(redisClient *c) { @@ -3865,7 +4263,10 @@ static void zaddCommand(redisClient *c) { assert(deleted != 0); zslInsert(zs->zsl,*score,c->argv[3]); incrRefCount(c->argv[3]); + dictReplace(zs->dict,c->argv[3],score); server.dirty++; + } else { + zfree(score); } addReply(c,shared.czero); } @@ -3906,6 +4307,30 @@ static void zremCommand(redisClient *c) { } } +static void zremrangebyscoreCommand(redisClient *c) { + double min = strtod(c->argv[2]->ptr,NULL); + double max = strtod(c->argv[3]->ptr,NULL); + robj *zsetobj; + zset *zs; + + zsetobj = lookupKeyWrite(c->db,c->argv[1]); + if (zsetobj == NULL) { + addReply(c,shared.czero); + } else { + long deleted; + + if (zsetobj->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + return; + } + zs = zsetobj->ptr; + deleted = zslDeleteRange(zs->zsl,min,max,zs->dict); + if (htNeedsResize(zs->dict)) dictResize(zs->dict); + server.dirty += deleted; + addReplySds(c,sdscatprintf(sdsempty(),":%lu\r\n",deleted)); + } +} + static void zrangeGenericCommand(redisClient *c, int reverse) { robj *o; int start = atoi(c->argv[2]->ptr); @@ -3972,7 +4397,53 @@ static void zrevrangeCommand(redisClient *c) { zrangeGenericCommand(c,1); } -static void zlenCommand(redisClient *c) { +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; @@ -3990,6 +4461,36 @@ static void zlenCommand(redisClient *c) { } } +static void zscoreCommand(redisClient *c) { + robj *o; + zset *zs; + + o = lookupKeyRead(c->db,c->argv[1]); + if (o == NULL) { + addReply(c,shared.nullbulk); + return; + } else { + if (o->type != REDIS_ZSET) { + addReply(c,shared.wrongtypeerr); + } else { + dictEntry *de; + + zs = o->ptr; + de = dictFind(zs->dict,c->argv[2]); + if (!de) { + addReply(c,shared.nullbulk); + } else { + char buf[128]; + double *score = dictGetEntryVal(de); + + snprintf(buf,sizeof(buf),"%.17g",*score); + addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n%s\r\n", + strlen(buf),buf)); + } + } + } +} + /* ========================= Non type-specific commands ==================== */ static void flushdbCommand(redisClient *c) { @@ -4119,7 +4620,7 @@ static void sortCommand(redisClient *c) { int limit_start = 0, limit_count = -1, start, end; int j, dontsort = 0, vectorlen; int getop = 0; /* GET operation counter */ - robj *sortval, *sortby = NULL; + robj *sortval, *sortby = NULL, *storekey = NULL; redisSortObject *vector; /* Resulting vector to sort */ /* Lookup the key to sort. It must be of the right types */ @@ -4157,6 +4658,9 @@ 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, @@ -4168,18 +4672,6 @@ 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); @@ -4286,37 +4778,77 @@ 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); - } - 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); + 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 { - 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 */ } } - } + } else { + robj *listObject = createListObject(); + list *listPtr = (list*) listObject->ptr; - /* Cleanup */ - decrRefCount(sortval); - listRelease(operations); + /* 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 { + assert(sop->type == REDIS_SORT_GET); /* always fails */ + } + } + } + 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 */ + decrRefCount(sortval); + listRelease(operations); for (j = 0; j < vectorlen; j++) { if (sortby && alpha && vector[j].u.cmpobj) decrRefCount(vector[j].u.cmpobj); @@ -4367,7 +4899,7 @@ 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++) { @@ -4455,21 +4987,21 @@ 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 { @@ -4479,6 +5011,14 @@ 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; @@ -4766,7 +5306,7 @@ 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; @@ -4776,6 +5316,30 @@ 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); @@ -4790,6 +5354,11 @@ 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 */ @@ -4913,6 +5482,198 @@ static void freeMemoryIfNeeded(void) { } } +/* ============================== Append Only file ========================== */ + +static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc) { + sds buf = sdsempty(); + int j; + ssize_t nwritten; + time_t now; + robj *tmpargv[3]; + + /* The DB this command was targetting is not the same as the last command + * we appendend. To issue a SELECT command is needed. */ + if (dictid != server.appendseldb) { + char seldb[64]; + + snprintf(seldb,sizeof(seldb),"%d",dictid); + buf = sdscatprintf(buf,"*2\r\n$6\r\nSELECT\r\n$%d\r\n%s\r\n", + strlen(seldb),seldb); + server.appendseldb = dictid; + } + + /* "Fix" the argv vector if the command is EXPIRE. We want to translate + * EXPIREs into EXPIREATs calls */ + if (cmd->proc == expireCommand) { + long when; + + tmpargv[0] = createStringObject("EXPIREAT",8); + tmpargv[1] = argv[1]; + incrRefCount(argv[1]); + when = time(NULL)+strtol(argv[2]->ptr,NULL,10); + tmpargv[2] = createObject(REDIS_STRING, + sdscatprintf(sdsempty(),"%ld",when)); + argv = tmpargv; + } + + /* Append the actual command */ + buf = sdscatprintf(buf,"*%d\r\n",argc); + for (j = 0; j < argc; j++) { + robj *o = argv[j]; + + if (o->encoding != REDIS_ENCODING_RAW) + o = getDecodedObject(o); + buf = sdscatprintf(buf,"$%d\r\n",sdslen(o->ptr)); + buf = sdscatlen(buf,o->ptr,sdslen(o->ptr)); + buf = sdscatlen(buf,"\r\n",2); + if (o != argv[j]) + decrRefCount(o); + } + + /* Free the objects from the modified argv for EXPIREAT */ + if (cmd->proc == expireCommand) { + for (j = 0; j < 3; j++) + decrRefCount(argv[j]); + } + + /* We want to perform a single write. This should be guaranteed atomic + * at least if the filesystem we are writing is a real physical one. + * While this will save us against the server being killed I don't think + * there is much to do about the whole server stopping for power problems + * or alike */ + nwritten = write(server.appendfd,buf,sdslen(buf)); + if (nwritten != (signed)sdslen(buf)) { + /* Ooops, we are in troubles. The best thing to do for now is + * to simply exit instead to give the illusion that everything is + * working as expected. */ + if (nwritten == -1) { + redisLog(REDIS_WARNING,"Exiting on error writing to the append-only file: %s",strerror(errno)); + } else { + redisLog(REDIS_WARNING,"Exiting on short write while writing to the append-only file: %s",strerror(errno)); + } + exit(1); + } + now = time(NULL); + if (server.appendfsync == APPENDFSYNC_ALWAYS || + (server.appendfsync == APPENDFSYNC_EVERYSEC && + now-server.lastfsync > 1)) + { + fsync(server.appendfd); /* Let's try to get this data on the disk */ + server.lastfsync = now; + } +} + +/* In Redis commands are always executed in the context of a client, so in + * order to load the append only file we need to create a fake client. */ +static struct redisClient *createFakeClient(void) { + struct redisClient *c = zmalloc(sizeof(*c)); + + selectDb(c,0); + c->fd = -1; + c->querybuf = sdsempty(); + c->argc = 0; + c->argv = NULL; + c->flags = 0; + /* We set the fake client as a slave waiting for the synchronization + * so that Redis will not try to send replies to this client. */ + c->replstate = REDIS_REPL_WAIT_BGSAVE_START; + c->reply = listCreate(); + listSetFreeMethod(c->reply,decrRefCount); + listSetDupMethod(c->reply,dupClientReplyValue); + return c; +} + +static void freeFakeClient(struct redisClient *c) { + sdsfree(c->querybuf); + listRelease(c->reply); + zfree(c); +} + +/* Replay the append log file. On error REDIS_OK is returned. On non fatal + * error (the append only file is zero-length) REDIS_ERR is returned. On + * fatal error an error message is logged and the program exists. */ +int loadAppendOnlyFile(char *filename) { + struct redisClient *fakeClient; + FILE *fp = fopen(filename,"r"); + struct redis_stat sb; + + if (redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0) + return REDIS_ERR; + + if (fp == NULL) { + redisLog(REDIS_WARNING,"Fatal error: can't open the append log file for reading: %s",strerror(errno)); + exit(1); + } + + fakeClient = createFakeClient(); + while(1) { + int argc, j; + unsigned long len; + robj **argv; + char buf[128]; + sds argsds; + struct redisCommand *cmd; + + if (fgets(buf,sizeof(buf),fp) == NULL) { + if (feof(fp)) + break; + else + goto readerr; + } + if (buf[0] != '*') goto fmterr; + argc = atoi(buf+1); + argv = zmalloc(sizeof(robj*)*argc); + for (j = 0; j < argc; j++) { + if (fgets(buf,sizeof(buf),fp) == NULL) goto readerr; + if (buf[0] != '$') goto fmterr; + len = strtol(buf+1,NULL,10); + argsds = sdsnewlen(NULL,len); + if (fread(argsds,len,1,fp) == 0) goto fmterr; + argv[j] = createObject(REDIS_STRING,argsds); + if (fread(buf,2,1,fp) == 0) goto fmterr; /* discard CRLF */ + } + + /* Command lookup */ + cmd = lookupCommand(argv[0]->ptr); + if (!cmd) { + redisLog(REDIS_WARNING,"Unknown command '%s' reading the append only file", argv[0]->ptr); + exit(1); + } + /* Try object sharing and encoding */ + if (server.shareobjects) { + int j; + for(j = 1; j < argc; j++) + argv[j] = tryObjectSharing(argv[j]); + } + if (cmd->flags & REDIS_CMD_BULK) + tryObjectEncoding(argv[argc-1]); + /* Run the command in the context of a fake client */ + fakeClient->argc = argc; + fakeClient->argv = argv; + cmd->proc(fakeClient); + /* Discard the reply objects list from the fake client */ + while(listLength(fakeClient->reply)) + listDelNode(fakeClient->reply,listFirst(fakeClient->reply)); + /* Clean up, ready for the next command */ + for (j = 0; j < argc; j++) decrRefCount(argv[j]); + zfree(argv); + } + fclose(fp); + freeFakeClient(fakeClient); + return REDIS_OK; + +readerr: + if (feof(fp)) { + redisLog(REDIS_WARNING,"Unexpected end of file reading the append only file"); + } else { + redisLog(REDIS_WARNING,"Unrecoverable error reading the append only file: %s", strerror(errno)); + } + exit(1); +fmterr: + redisLog(REDIS_WARNING,"Bad file format reading the append only file"); + exit(1); +} + /* ================================= Debugging ============================== */ static void debugCommand(redisClient *c) { @@ -4937,146 +5698,91 @@ static void debugCommand(redisClient *c) { } } -#ifdef HAVE_BACKTRACE -static struct redisFunctionSym symsTable[] = { -{"compareStringObjects", (unsigned long)compareStringObjects}, -{"isStringRepresentableAsLong", (unsigned long)isStringRepresentableAsLong}, -{"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}, -{"srandmemberCommand", (unsigned long)srandmemberCommand}, -{"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}, -{"msetGenericCommand", (unsigned long)msetGenericCommand}, -{"msetCommand", (unsigned long)msetCommand}, -{"msetnxCommand", (unsigned long)msetnxCommand}, -{"zslCreateNode", (unsigned long)zslCreateNode}, -{"zslCreate", (unsigned long)zslCreate}, -{"zslFreeNode",(unsigned long)zslFreeNode}, -{"zslFree",(unsigned long)zslFree}, -{"zslRandomLevel",(unsigned long)zslRandomLevel}, -{"zslInsert",(unsigned long)zslInsert}, -{"zslDelete",(unsigned long)zslDelete}, -{"createZsetObject",(unsigned long)createZsetObject}, -{"zaddCommand",(unsigned long)zaddCommand}, -{"zrangeGenericCommand",(unsigned long)zrangeGenericCommand}, -{"zrangeCommand",(unsigned long)zrangeCommand}, -{"zrevrangeCommand",(unsigned long)zrevrangeCommand}, -{"zremCommand",(unsigned long)zremCommand}, -{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, 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; @@ -5171,82 +5877,39 @@ 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; -}