X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/c9a111acf47cb5bb2138d1f699253f87d68e53e8..638e42aca477816ab4754ffb4e1e2816bdcc163c:/redis.c diff --git a/redis.c b/redis.c index bcc8e5dc..74986528 100644 --- a/redis.c +++ b/redis.c @@ -27,9 +27,10 @@ * POSSIBILITY OF SUCH DAMAGE. */ -#define REDIS_VERSION "0.101" +#define REDIS_VERSION "1.001" #include "fmacros.h" +#include "config.h" #include #include @@ -38,8 +39,12 @@ #include #define __USE_POSIX199309 #include + +#ifdef HAVE_BACKTRACE #include #include +#endif /* HAVE_BACKTRACE */ + #include #include #include @@ -52,7 +57,6 @@ #include #include #include -#include #include "redis.h" #include "ae.h" /* Event driven programming library */ @@ -64,8 +68,6 @@ #include "lzf.h" /* LZF compression library */ #include "pqsort.h" /* Partial qsort for SORT+LIMIT */ -#include "config.h" - /* Error codes */ #define REDIS_OK 0 #define REDIS_ERR -1 @@ -102,6 +104,10 @@ #define REDIS_SET 2 #define REDIS_HASH 3 +/* Objects encoding */ +#define REDIS_ENCODING_RAW 0 /* Raw representation */ +#define REDIS_ENCODING_INT 1 /* Encoded as integer */ + /* Object types only used for dumping to disk */ #define REDIS_EXPIRETIME 253 #define REDIS_SELECTDB 254 @@ -181,7 +187,9 @@ /* A redis object, that is a type able to hold a string / list / set */ typedef struct redisObject { void *ptr; - int type; + unsigned char type; + unsigned char encoding; + unsigned char notused[2]; int refcount; } robj; @@ -198,9 +206,10 @@ 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 */ @@ -261,7 +270,7 @@ struct redisServer { 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; @@ -321,16 +330,21 @@ static robj *createStringObject(char *ptr, size_t len); static void replicationFeedSlaves(list *slaves, struct redisCommand *cmd, int dictid, robj **argv, int argc); static int syncWithMaster(void); static robj *tryObjectSharing(robj *o); +static int tryObjectEncoding(robj *o); +static robj *getDecodedObject(const robj *o); static int removeExpire(redisDb *db, robj *key); static int expireIfNeeded(redisDb *db, robj *key); static int deleteIfVolatile(redisDb *db, robj *key); static int deleteKey(redisDb *db, robj *key); static time_t getExpire(redisDb *db, robj *key); static int setExpire(redisDb *db, robj *key, time_t when); -static void updateSalvesWaitingBgsave(int bgsaveerr); +static void updateSlavesWaitingBgsave(int bgsaveerr); static void freeMemoryIfNeeded(void); static int processCommand(redisClient *c); static void setupSigSegvAction(void); +static void rdbRemoveTempFile(pid_t childpid); +static size_t stringObjectLen(robj *o); +static void processInputBuffer(redisClient *c); static void authCommand(redisClient *c); static void pingCommand(redisClient *c); @@ -386,10 +400,13 @@ 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 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); + /*================================= Globals ================================= */ /* Global vars */ @@ -428,7 +445,9 @@ static struct redisCommand cmdTable[] = { {"smembers",sinterCommand,2,REDIS_CMD_INLINE}, {"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}, @@ -631,32 +650,68 @@ static void dictRedisObjectDestructor(void *privdata, void *val) decrRefCount(val); } -static int dictSdsKeyCompare(void *privdata, const void *key1, +static int dictObjKeyCompare(void *privdata, const void *key1, const void *key2) { const robj *o1 = key1, *o2 = key2; return sdsDictKeyCompare(privdata,o1->ptr,o2->ptr); } -static unsigned int dictSdsHash(const void *key) { +static unsigned int dictObjHash(const void *key) { const robj *o = key; return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr)); } +static int dictEncObjKeyCompare(void *privdata, const void *key1, + const void *key2) +{ + const robj *o1 = key1, *o2 = key2; + + if (o1->encoding == REDIS_ENCODING_RAW && + o2->encoding == REDIS_ENCODING_RAW) + return sdsDictKeyCompare(privdata,o1->ptr,o2->ptr); + else { + robj *dec1, *dec2; + int cmp; + + dec1 = o1->encoding != REDIS_ENCODING_RAW ? + getDecodedObject(o1) : (robj*)o1; + dec2 = o2->encoding != REDIS_ENCODING_RAW ? + getDecodedObject(o2) : (robj*)o2; + cmp = sdsDictKeyCompare(privdata,dec1->ptr,dec2->ptr); + if (dec1 != o1) decrRefCount(dec1); + if (dec2 != o2) decrRefCount(dec2); + return cmp; + } +} + +static unsigned int dictEncObjHash(const void *key) { + const robj *o = key; + + if (o->encoding == REDIS_ENCODING_RAW) + return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr)); + else { + robj *dec = getDecodedObject(o); + unsigned int hash = dictGenHashFunction(dec->ptr, sdslen((sds)dec->ptr)); + decrRefCount(dec); + return hash; + } +} + static dictType setDictType = { - dictSdsHash, /* hash function */ + dictEncObjHash, /* hash function */ NULL, /* key dup */ NULL, /* val dup */ - dictSdsKeyCompare, /* key compare */ + dictEncObjKeyCompare, /* key compare */ dictRedisObjectDestructor, /* key destructor */ NULL /* val destructor */ }; static dictType hashDictType = { - dictSdsHash, /* hash function */ + dictObjHash, /* hash function */ NULL, /* key dup */ NULL, /* val dup */ - dictSdsKeyCompare, /* key compare */ + dictObjKeyCompare, /* key compare */ dictRedisObjectDestructor, /* key destructor */ dictRedisObjectDestructor /* val destructor */ }; @@ -734,8 +789,8 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD size = dictSlots(server.db[j].dict); used = dictSize(server.db[j].dict); vkeys = dictSize(server.db[j].expires); - if (!(loops % 5) && used > 0) { - redisLog(REDIS_DEBUG,"DB %d: %d keys (%d volatile) in %d slots HT.",j,used,vkeys,size); + if (!(loops % 5) && (used || vkeys)) { + redisLog(REDIS_DEBUG,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size); /* dictPrintStats(server.dict); */ } } @@ -750,7 +805,7 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD /* Show information about connected clients */ if (!(loops % 5)) { - redisLog(REDIS_DEBUG,"%d clients connected (%d slaves), %zu bytes in use", + redisLog(REDIS_DEBUG,"%d clients connected (%d slaves), %zu bytes in use, %d shared objects", listLength(server.clients)-listLength(server.slaves), listLength(server.slaves), server.usedmemory, @@ -778,10 +833,11 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD } else { redisLog(REDIS_WARNING, "Background saving terminated by signal"); + rdbRemoveTempFile(server.bgsavechildpid); } server.bgsaveinprogress = 0; server.bgsavechildpid = -1; - updateSalvesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR); + updateSlavesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR); } } else { /* If there is not a background saving in progress check if @@ -898,6 +954,7 @@ static void initServerConfig() { server.dbfilename = "dump.rdb"; server.requirepass = NULL; server.shareobjects = 0; + server.sharingpoolsize = 1024; server.maxclients = 0; server.maxmemory = 0; ResetServerSaveParams(); @@ -928,7 +985,6 @@ static void initServer() { server.el = aeCreateEventLoop(); server.db = zmalloc(sizeof(redisDb)*server.dbnum); server.sharingpool = dictCreate(&setDictType,NULL); - server.sharingpoolsize = 1024; if (!server.db || !server.clients || !server.slaves || !server.monitors || !server.el || !server.objfreelist) oom("server initialization"); /* Fatal OOM */ server.fd = anetTcpServer(server.neterr, server.port, server.bindaddr); @@ -1068,7 +1124,7 @@ 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]); @@ -1120,7 +1176,10 @@ 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) { @@ -1148,6 +1207,7 @@ static void freeClient(redisClient *c) { server.replstate = REDIS_REPL_CONNECT; } zfree(c->argv); + zfree(c->mbargv); zfree(c); } @@ -1250,6 +1310,7 @@ 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 @@ -1267,6 +1328,82 @@ 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 */ + /* + if (sdslen(c->querybuf) > 0) + processInputBuffer(c); + */ + 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; + /* + if (sdslen(c->querybuf) > 0) + processInputBuffer(c); + */ + 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")) { @@ -1315,6 +1452,10 @@ static int processCommand(redisClient *c) { for(j = 1; j < c->argc; j++) c->argv[j] = tryObjectSharing(c->argv[j]); } + /* Let's try to encode the bulk object to save space. */ + if (cmd->flags & REDIS_CMD_BULK) + tryObjectEncoding(c->argv[c->argc-1]); + /* Check if the user is authenticated */ if (server.requirepass && !c->authenticated && cmd->proc != authCommand) { addReplySds(c,sdsnew("-ERR operation not permitted\r\n")); @@ -1360,7 +1501,8 @@ 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; } @@ -1409,34 +1551,7 @@ 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 */ @@ -1484,7 +1599,7 @@ again: /* Execute the command. If the client is still valid * after processCommand() return and there is something * on the query buffer try to process the next command. */ - if (processCommand(c) && sdslen(c->querybuf)) goto again; + if (c->argc && processCommand(c) && sdslen(c->querybuf)) goto again; return; } else if (sdslen(c->querybuf) >= REDIS_REQUEST_MAX_SIZE) { redisLog(REDIS_DEBUG, "Client protocol error"); @@ -1503,12 +1618,45 @@ 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) { if (id < 0 || id >= server.dbnum) return REDIS_ERR; @@ -1533,6 +1681,9 @@ 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); @@ -1556,8 +1707,12 @@ static void addReply(redisClient *c, robj *obj) { 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"); - incrRefCount(obj); } static void addReplySds(redisClient *c, sds s) { @@ -1566,6 +1721,26 @@ static void addReplySds(redisClient *c, sds s) { decrRefCount(o); } +static void addReplyBulkLen(redisClient *c, robj *obj) { + size_t len; + + if (obj->encoding == REDIS_ENCODING_RAW) { + len = sdslen(obj->ptr); + } else { + long n = (long)obj->ptr; + + len = 1; + if (n < 0) { + len++; + n = -n; + } + while((n = n/10) != 0) { + len++; + } + } + addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",len)); +} + static void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask) { int cport, cfd; char cip[128]; @@ -1614,6 +1789,7 @@ static robj *createObject(int type, void *ptr) { } if (!o) oom("createObject"); o->type = type; + o->encoding = REDIS_ENCODING_RAW; o->ptr = ptr; o->refcount = 1; return o; @@ -1638,7 +1814,9 @@ static robj *createSetObject(void) { } static void freeStringObject(robj *o) { - sdsfree(o->ptr); + if (o->encoding == REDIS_ENCODING_RAW) { + sdsfree(o->ptr); + } } static void freeListObject(robj *o) { @@ -1682,6 +1860,36 @@ static void decrRefCount(void *obj) { } } +static robj *lookupKey(redisDb *db, robj *key) { + dictEntry *de = dictFind(db->dict,key); + return de ? dictGetEntryVal(de) : NULL; +} + +static robj *lookupKeyRead(redisDb *db, robj *key) { + expireIfNeeded(db,key); + return lookupKey(db,key); +} + +static robj *lookupKeyWrite(redisDb *db, robj *key) { + deleteIfVolatile(db,key); + return lookupKey(db,key); +} + +static int deleteKey(redisDb *db, robj *key) { + int retval; + + /* We need to protect key from destruction: after the first dictDelete() + * it may happen that 'key' is no longer valid if we don't increment + * it's count. This may happen when we get the object reference directly + * from the hash table with dictRandomKey() or dict iterators */ + incrRefCount(key); + if (dictSize(db->expires)) dictDelete(db->expires,key); + retval = dictDelete(db->dict,key); + decrRefCount(key); + + return retval == DICT_OK; +} + /* Try to share an object against the shared objects pool */ static robj *tryObjectSharing(robj *o) { struct dictEntry *de; @@ -1727,34 +1935,98 @@ static robj *tryObjectSharing(robj *o) { } } -static robj *lookupKey(redisDb *db, robj *key) { - dictEntry *de = dictFind(db->dict,key); - return de ? dictGetEntryVal(de) : NULL; +/* Check if the nul-terminated string 's' can be represented by a long + * (that is, is a number that fits into long without any other space or + * character before or after the digits). + * + * If so, the function returns REDIS_OK and *longval is set to the value + * of the number. Otherwise REDIS_ERR is returned */ +static int isStringRepresentableAsLong(char *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 (strlen(buf) != (unsigned)slen || memcmp(buf,s,slen)) return REDIS_ERR; + if (longval) *longval = value; + return REDIS_OK; } -static robj *lookupKeyRead(redisDb *db, robj *key) { - expireIfNeeded(db,key); - return lookupKey(db,key); +/* Try to encode a string object in order to save space */ +static int tryObjectEncoding(robj *o) { + long value; + sds s = o->ptr; + + if (o->encoding != REDIS_ENCODING_RAW) + return REDIS_ERR; /* Already encoded */ + + /* It's not save to encode shared objects: shared objects can be shared + * everywhere in the "object space" of Redis. Encoded objects can only + * appear as "values" (and not, for instance, as keys) */ + if (o->refcount > 1) return REDIS_ERR; + + /* Currently we try to encode only strings */ + assert(o->type == REDIS_STRING); + + /* Check if we can represent this string as a long integer */ + if (isStringRepresentableAsLong(s,&value) == REDIS_ERR) return REDIS_ERR; + + /* Ok, this object can be encoded */ + o->encoding = REDIS_ENCODING_INT; + sdsfree(o->ptr); + o->ptr = (void*) value; + return REDIS_OK; } -static robj *lookupKeyWrite(redisDb *db, robj *key) { - deleteIfVolatile(db,key); - return lookupKey(db,key); +/* Get a decoded version of an encoded object (returned as a new object) */ +static robj *getDecodedObject(const robj *o) { + robj *dec; + + assert(o->encoding != REDIS_ENCODING_RAW); + if (o->type == REDIS_STRING && o->encoding == REDIS_ENCODING_INT) { + char buf[32]; + + snprintf(buf,32,"%ld",(long)o->ptr); + dec = createStringObject(buf,strlen(buf)); + return dec; + } else { + assert(1 != 1); + } } -static int deleteKey(redisDb *db, robj *key) { - int retval; +static int compareStringObjects(robj *a, robj *b) { + assert(a->type == REDIS_STRING && b->type == REDIS_STRING); - /* We need to protect key from destruction: after the first dictDelete() - * it may happen that 'key' is no longer valid if we don't increment - * it's count. This may happen when we get the object reference directly - * from the hash table with dictRandomKey() or dict iterators */ - incrRefCount(key); - if (dictSize(db->expires)) dictDelete(db->expires,key); - retval = dictDelete(db->dict,key); - decrRefCount(key); + if (a->encoding == REDIS_ENCODING_INT && b->encoding == REDIS_ENCODING_INT){ + return (long)a->ptr - (long)b->ptr; + } else { + int retval; - return retval == DICT_OK; + 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; + } +} + +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 ============================ */ @@ -1861,10 +2133,12 @@ writeerr: /* Save a string objet as [len][data] on disk. If the object is a string * representation of an integer value we try to safe it in a special form */ -static int rdbSaveStringObject(FILE *fp, robj *obj) { - size_t len = sdslen(obj->ptr); +static int rdbSaveStringObjectRaw(FILE *fp, robj *obj) { + size_t len; int enclen; + len = sdslen(obj->ptr); + /* Try integer encoding */ if (len <= 11) { unsigned char buf[5]; @@ -1876,7 +2150,7 @@ static int rdbSaveStringObject(FILE *fp, robj *obj) { /* Try LZF compression - under 20 bytes it's unable to compress even * aaaaaaaaaaaaaaaaaa so skip it */ - if (1 && len > 20) { + if (len > 20) { int retval; retval = rdbSaveLzfStringObject(fp,obj); @@ -1891,6 +2165,21 @@ static int rdbSaveStringObject(FILE *fp, robj *obj) { return 0; } +/* Like rdbSaveStringObjectRaw() but handle encoded objects */ +static int rdbSaveStringObject(FILE *fp, robj *obj) { + int retval; + robj *dec; + + if (obj->encoding != REDIS_ENCODING_RAW) { + dec = getDecodedObject(obj); + retval = rdbSaveStringObjectRaw(fp,dec); + decrRefCount(dec); + return retval; + } else { + return rdbSaveStringObjectRaw(fp,obj); + } +} + /* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success */ static int rdbSave(char *filename) { dictIterator *di = NULL; @@ -1900,7 +2189,7 @@ static int rdbSave(char *filename) { int j; time_t now = time(NULL); - snprintf(tmpfile,256,"temp-%d.%ld.rdb",(int)time(NULL),(long int)random()); + snprintf(tmpfile,256,"temp-%d.rdb", (int) getpid()); fp = fopen(tmpfile,"w"); if (!fp) { redisLog(REDIS_WARNING, "Failed saving the DB: %s", strerror(errno)); @@ -2027,6 +2316,13 @@ static int rdbSaveBackground(char *filename) { return REDIS_OK; /* unreached */ } +static void rdbRemoveTempFile(pid_t childpid) { + char tmpfile[256]; + + snprintf(tmpfile,256,"temp-%d.rdb", (int) childpid); + unlink(tmpfile); +} + static int rdbLoadType(FILE *fp) { unsigned char type; if (fread(&type,1,1,fp) == 0) return -1; @@ -2201,6 +2497,7 @@ static int rdbLoad(char *filename) { if (type == REDIS_STRING) { /* Read string value */ if ((o = rdbLoadStringObject(fp,rdbver)) == NULL) goto eoferr; + tryObjectEncoding(o); } else if (type == REDIS_LIST || type == REDIS_SET) { /* Read list/set value */ uint32_t listlen; @@ -2213,6 +2510,7 @@ static int rdbLoad(char *filename) { robj *ele; if ((ele = rdbLoadStringObject(fp,rdbver)) == NULL) goto eoferr; + tryObjectEncoding(ele); if (type == REDIS_LIST) { if (!listAddNodeTail((list*)o->ptr,ele)) oom("listAddNodeTail"); @@ -2266,8 +2564,7 @@ static void pingCommand(redisClient *c) { } static void echoCommand(redisClient *c) { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n", - (int)sdslen(c->argv[1]->ptr))); + addReplyBulkLen(c,c->argv[1]); addReply(c,c->argv[1]); addReply(c,shared.crlf); } @@ -2312,14 +2609,14 @@ static void getCommand(redisClient *c) { if (o->type != REDIS_STRING) { addReply(c,shared.wrongtypeerr); } else { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(o->ptr))); + addReplyBulkLen(c,o); addReply(c,o); addReply(c,shared.crlf); } } } -static void getSetCommand(redisClient *c) { +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]); @@ -2343,7 +2640,7 @@ static void mgetCommand(redisClient *c) { if (o->type != REDIS_STRING) { addReply(c,shared.nullbulk); } else { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(o->ptr))); + addReplyBulkLen(c,o); addReply(c,o); addReply(c,shared.crlf); } @@ -2365,12 +2662,18 @@ static void incrDecrCommand(redisClient *c, long long incr) { } else { char *eptr; - value = strtoll(o->ptr, &eptr, 10); + if (o->encoding == REDIS_ENCODING_RAW) + value = strtoll(o->ptr, &eptr, 10); + else if (o->encoding == REDIS_ENCODING_INT) + value = (long)o->ptr; + else + assert(1 != 1); } } value += incr; o = createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",value)); + tryObjectEncoding(o); retval = dictAdd(c->db->dict,c->argv[1],o); if (retval == DICT_ERR) { dictReplace(c->db->dict,c->argv[1],o); @@ -2544,11 +2847,15 @@ static void bgsaveCommand(redisClient *c) { static void shutdownCommand(redisClient *c) { redisLog(REDIS_WARNING,"User requested shutdown, saving DB..."); + /* Kill the saving child if there is a background saving in progress. + We want to avoid race conditions, for instance our saving child may + overwrite the synchronous saving did by SHUTDOWN. */ if (server.bgsaveinprogress) { redisLog(REDIS_WARNING,"There is a live saving child. Killing it!"); - signal(SIGCHLD, SIG_IGN); kill(server.bgsavechildpid,SIGKILL); + rdbRemoveTempFile(server.bgsavechildpid); } + /* SYNC SAVE */ if (rdbSave(server.dbfilename) == REDIS_OK) { if (server.daemonize) unlink(server.pidfile); @@ -2556,7 +2863,10 @@ static void shutdownCommand(redisClient *c) { redisLog(REDIS_WARNING,"Server exit now, bye bye..."); exit(1); } else { - signal(SIGCHLD, SIG_DFL); + /* Ooops.. error saving! The best we can do is to continue operating. + * Note that if there was a background saving process, in the next + * cron() Redis will be notified that the background saving aborted, + * handling special stuff like slaves pending for synchronization... */ redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit"); addReplySds(c,sdsnew("-ERR can't quit, problems saving the DB\r\n")); } @@ -2724,7 +3034,7 @@ static void lindexCommand(redisClient *c) { addReply(c,shared.nullbulk); } else { robj *ele = listNodeValue(ln); - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(ele->ptr))); + addReplyBulkLen(c,ele); addReply(c,ele); addReply(c,shared.crlf); } @@ -2784,7 +3094,7 @@ static void popGenericCommand(redisClient *c, int where) { addReply(c,shared.nullbulk); } else { robj *ele = listNodeValue(ln); - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(ele->ptr))); + addReplyBulkLen(c,ele); addReply(c,ele); addReply(c,shared.crlf); listDelNode(list,ln); @@ -2840,7 +3150,7 @@ static void lrangeCommand(redisClient *c) { addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",rangelen)); for (j = 0; j < rangelen; j++) { ele = listNodeValue(ln); - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(ele->ptr))); + addReplyBulkLen(c,ele); addReply(c,ele); addReply(c,shared.crlf); ln = ln->next; @@ -2892,8 +3202,8 @@ static void ltrimCommand(redisClient *c) { ln = listLast(list); listDelNode(list,ln); } - addReply(c,shared.ok); server.dirty++; + addReply(c,shared.ok); } } } @@ -2923,7 +3233,7 @@ 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++; @@ -3072,7 +3382,7 @@ static void spopCommand(redisClient *c) { } else { robj *ele = dictGetEntryKey(de); - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",sdslen(ele->ptr))); + addReplyBulkLen(c,ele); addReply(c,ele); addReply(c,shared.crlf); dictDelete(set->ptr,ele); @@ -3153,7 +3463,7 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, int setsnum, r continue; /* at least one set does not contain the member */ ele = dictGetEntryKey(de); if (!dstkey) { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",sdslen(ele->ptr))); + addReplyBulkLen(c,ele); addReply(c,ele); addReply(c,shared.crlf); cardinality++; @@ -3262,8 +3572,7 @@ static void sunionDiffGenericCommand(redisClient *c, robj **setskeys, int setsnu robj *ele; ele = dictGetEntryKey(de); - addReplySds(c,sdscatprintf(sdsempty(), - "$%d\r\n",sdslen(ele->ptr))); + addReplyBulkLen(c,ele); addReply(c,ele); addReply(c,shared.crlf); } @@ -3339,6 +3648,12 @@ static robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { char buf[REDIS_SORTKEY_MAX+1]; } keyname; + if (subst->encoding == REDIS_ENCODING_RAW) + incrRefCount(subst); + else { + subst = getDecodedObject(subst); + } + spat = pattern->ptr; ssub = subst->ptr; if (sdslen(spat)+sdslen(ssub)-1 > REDIS_SORTKEY_MAX) return NULL; @@ -3358,6 +3673,8 @@ static robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { keyobj.type = REDIS_STRING; keyobj.ptr = ((char*)&keyname)+(sizeof(long)*2); + decrRefCount(subst); + /* printf("lookup '%s' => %p\n", keyname.buf,de); */ return lookupKeyRead(db,&keyobj); } @@ -3395,7 +3712,20 @@ static int sortCompare(const void *s1, const void *s2) { } } else { /* Compare elements directly */ - cmp = strcoll(so1->obj->ptr,so2->obj->ptr); + if (so1->obj->encoding == REDIS_ENCODING_RAW && + so2->obj->encoding == REDIS_ENCODING_RAW) { + cmp = strcoll(so1->obj->ptr,so2->obj->ptr); + } else { + robj *dec1, *dec2; + + dec1 = so1->obj->encoding == REDIS_ENCODING_RAW ? + so1->obj : getDecodedObject(so1->obj); + dec2 = so2->obj->encoding == REDIS_ENCODING_RAW ? + so2->obj : getDecodedObject(so2->obj); + cmp = strcoll(dec1->ptr,dec2->ptr); + if (dec1 != so1->obj) decrRefCount(dec1); + if (dec2 != so2->obj) decrRefCount(dec2); + } } } return server.sort_desc ? -cmp : cmp; @@ -3525,13 +3855,33 @@ static void sortCommand(redisClient *c) { byval = lookupKeyByPattern(c->db,sortby,vector[j].obj); if (!byval || byval->type != REDIS_STRING) continue; if (alpha) { - vector[j].u.cmpobj = byval; - incrRefCount(byval); + if (byval->encoding == REDIS_ENCODING_RAW) { + vector[j].u.cmpobj = byval; + incrRefCount(byval); + } else { + vector[j].u.cmpobj = getDecodedObject(byval); + } } else { - vector[j].u.score = strtod(byval->ptr,NULL); + if (byval->encoding == REDIS_ENCODING_RAW) { + vector[j].u.score = strtod(byval->ptr,NULL); + } else { + if (byval->encoding == REDIS_ENCODING_INT) { + vector[j].u.score = (long)byval->ptr; + } else + assert(1 != 1); + } } } else { - if (!alpha) vector[j].u.score = strtod(vector[j].obj->ptr,NULL); + if (!alpha) { + if (vector[j].obj->encoding == REDIS_ENCODING_RAW) + vector[j].u.score = strtod(vector[j].obj->ptr,NULL); + else { + if (vector[j].obj->encoding == REDIS_ENCODING_INT) + vector[j].u.score = (long) vector[j].obj->ptr; + else + assert(1 != 1); + } + } } } } @@ -3563,8 +3913,7 @@ static void sortCommand(redisClient *c) { for (j = start; j <= end; j++) { listNode *ln; if (!getop) { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n", - sdslen(vector[j].obj->ptr))); + addReplyBulkLen(c,vector[j].obj); addReply(c,vector[j].obj); addReply(c,shared.crlf); } @@ -3578,8 +3927,7 @@ static void sortCommand(redisClient *c) { if (!val || val->type != REDIS_STRING) { addReply(c,shared.nullbulk); } else { - addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n", - sdslen(val->ptr))); + addReplyBulkLen(c,val); addReply(c,val); addReply(c,shared.crlf); } @@ -3602,9 +3950,11 @@ static void sortCommand(redisClient *c) { static void infoCommand(redisClient *c) { sds info; time_t uptime = time(NULL)-server.stat_starttime; + int j; info = sdscatprintf(sdsempty(), "redis_version:%s\r\n" + "arch_bits:%s\r\n" "uptime_in_seconds:%d\r\n" "uptime_in_days:%d\r\n" "connected_clients:%d\r\n" @@ -3617,6 +3967,7 @@ static void infoCommand(redisClient *c) { "total_commands_processed:%lld\r\n" "role:%s\r\n" ,REDIS_VERSION, + (sizeof(long) == 8) ? "64" : "32", uptime, uptime/(3600*24), listLength(server.clients)-listLength(server.slaves), @@ -3642,6 +3993,16 @@ static void infoCommand(redisClient *c) { (int)(time(NULL)-server.master->lastinteraction) ); } + for (j = 0; j < server.dbnum; j++) { + long long keys, vkeys; + + keys = dictSize(server.db[j].dict); + vkeys = dictSize(server.db[j].expires); + if (keys || vkeys) { + info = sdscatprintf(info, "db%d: keys=%lld,expires=%lld\r\n", + j, keys, vkeys); + } + } addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",sdslen(info))); addReplySds(c,info); addReply(c,shared.crlf); @@ -3731,10 +4092,12 @@ static void expireCommand(redisClient *c) { return; } else { time_t when = time(NULL)+seconds; - if (setExpire(c->db,c->argv[1],when)) + if (setExpire(c->db,c->argv[1],when)) { addReply(c,shared.cone); - else + server.dirty++; + } else { addReply(c,shared.czero); + } return; } } @@ -3751,6 +4114,42 @@ static void ttlCommand(redisClient *c) { addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",ttl)); } +static void msetGenericCommand(redisClient *c, int nx) { + int j; + + if ((c->argc % 2) == 0) { + addReplySds(c,sdsnew("-ERR wrong number of arguments\r\n")); + return; + } + /* Handle the NX flag. The MSETNX semantic is to return zero and don't + * set nothing at all if at least one already key exists. */ + if (nx) { + for (j = 1; j < c->argc; j += 2) { + if (dictFind(c->db->dict,c->argv[j]) != NULL) { + addReply(c, shared.czero); + return; + } + } + } + + for (j = 1; j < c->argc; j += 2) { + dictAdd(c->db->dict,c->argv[j],c->argv[j+1]); + incrRefCount(c->argv[j]); + incrRefCount(c->argv[j+1]); + removeExpire(c->db,c->argv[j]); + } + server.dirty += (c->argc-1)/2; + addReply(c, nx ? shared.cone : shared.ok); +} + +static void msetCommand(redisClient *c) { + msetGenericCommand(c,0); +} + +static void msetnxCommand(redisClient *c) { + msetGenericCommand(c,1); +} + /* =============================== Replication ============================= */ static int syncWrite(int fd, char *ptr, ssize_t size, int timeout) { @@ -3928,7 +4327,13 @@ static void sendBulkToSlave(aeEventLoop *el, int fd, void *privdata, int mask) { } } -static void updateSalvesWaitingBgsave(int bgsaveerr) { +/* This function is called at the end of every backgrond saving. + * The argument bgsaveerr is REDIS_OK if the background saving succeeded + * otherwise REDIS_ERR is passed to the function. + * + * The goal of this function is to handle slaves waiting for a successful + * background saving in order to perform non-blocking synchronization. */ +static void updateSlavesWaitingBgsave(int bgsaveerr) { listNode *ln; int startbgsave = 0; @@ -4141,8 +4546,8 @@ static void debugCommand(redisClient *c) { key = dictGetEntryKey(de); val = dictGetEntryVal(de); addReplySds(c,sdscatprintf(sdsempty(), - "+Key at:%p refcount:%d, value at:%p refcount:%d\r\n", - key, key->refcount, val, val->refcount)); + "+Key at:%p refcount:%d, value at:%p refcount:%d encoding:%d\r\n", + key, key->refcount, val, val->refcount, val->encoding)); } else { addReplySds(c,sdsnew( "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT ]\r\n")); @@ -4151,6 +4556,11 @@ 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}, @@ -4158,6 +4568,8 @@ static struct redisFunctionSym symsTable[] = { {"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}, @@ -4166,13 +4578,15 @@ static struct redisFunctionSym symsTable[] = { {"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}, -{"updateSalvesWaitingBgsave", (unsigned long)updateSalvesWaitingBgsave}, +{"updateSlavesWaitingBgsave", (unsigned long)updateSlavesWaitingBgsave}, {"freeMemoryIfNeeded", (unsigned long)freeMemoryIfNeeded}, {"authCommand", (unsigned long)authCommand}, {"pingCommand", (unsigned long)pingCommand}, @@ -4228,13 +4642,17 @@ static struct redisFunctionSym symsTable[] = { {"mgetCommand", (unsigned long)mgetCommand}, {"monitorCommand", (unsigned long)monitorCommand}, {"expireCommand", (unsigned long)expireCommand}, -{"getSetCommand", (unsigned long)getSetCommand}, +{"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}, {NULL,0} }; @@ -4267,10 +4685,20 @@ static void *getMcontextEip(ucontext_t *uc) { return (void*) uc->uc_mcontext.mc_eip; #elif defined(__dietlibc__) return (void*) uc->uc_mcontext.eip; -#elif defined(__APPLE__) +#elif defined(__APPLE__) && !defined(MAC_OS_X_VERSION_10_6) return (void*) uc->uc_mcontext->__ss.__eip; -#else /* Linux */ +#elif defined(__APPLE__) && defined(MAC_OS_X_VERSION_10_6) + #ifdef _STRUCT_X86_THREAD_STATE64 + return (void*) uc->uc_mcontext->__ss.__rip; + #else + return (void*) uc->uc_mcontext->__ss.__eip; + #endif +#elif defined(__i386__) || defined(__X86_64__) /* Linux x86 */ return (void*) uc->uc_mcontext.gregs[REG_EIP]; +#elif defined(__ia64__) /* Linux IA64 */ + return (void*) uc->uc_mcontext.sc_ip; +#else + return NULL; #endif } @@ -4312,7 +4740,9 @@ static void segvHandler(int sig, siginfo_t *info, void *secret) { trace_size = backtrace(trace, 100); /* overwrite sigaction with caller's address */ - trace[1] = getMcontextEip(uc); + if (getMcontextEip(uc) != NULL) { + trace[1] = getMcontextEip(uc); + } messages = backtrace_symbols(trace, trace_size); for (i=1; i /proc/sys/vm/overcommit_memory' in your init scripts."); + redisLog(REDIS_WARNING,"WARNING overcommit_memory is set to 0! Background save may fail under low condition memory. To fix this issue add 'vm.overcommit_memory = 1' to /etc/sysctl.conf and then reboot or run the command 'sysctl vm.overcommit_memory=1' for this to take effect."); } } #endif /* __linux__ */ @@ -4398,10 +4828,6 @@ static void daemonize(void) { } int main(int argc, char **argv) { -#ifdef __linux__ - linuxOvercommitMemoryWarning(); -#endif - initServerConfig(); if (argc == 2) { ResetServerSaveParams(); @@ -4415,6 +4841,9 @@ int main(int argc, char **argv) { 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,