X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/570e43c8285a4e5e3f31edd8198b8e03ce63f46c..fda32cb6013c65f765fa9b203bed23e07d35ba59:/redis.c diff --git a/redis.c b/redis.c index f2f54316..380924de 100644 --- a/redis.c +++ b/redis.c @@ -27,7 +27,7 @@ * POSSIBILITY OF SUCH DAMAGE. */ -#define REDIS_VERSION "1.3.7" +#define REDIS_VERSION "2.1.0" #include "fmacros.h" #include "config.h" @@ -37,8 +37,6 @@ #include #include #include -#define __USE_POSIX199309 -#define __USE_UNIX98 #include #ifdef HAVE_BACKTRACE @@ -59,6 +57,7 @@ #include #include #include +#include #include #include @@ -75,7 +74,9 @@ #include "zmalloc.h" /* total memory usage aware version of malloc/free */ #include "lzf.h" /* LZF compression library */ #include "pqsort.h" /* Partial qsort for SORT+LIMIT */ -#include "zipmap.h" +#include "zipmap.h" /* Compact dictionary-alike data structure */ +#include "sha1.h" /* SHA1 is used for DEBUG DIGEST */ +#include "release.h" /* Release and/or git repository information */ /* Error codes */ #define REDIS_OK 0 @@ -91,7 +92,7 @@ #define REDIS_CONFIGLINE_MAX 1024 #define REDIS_OBJFREELIST_MAX 1000000 /* Max number of objects to cache */ #define REDIS_MAX_SYNC_TIME 60 /* Slave can't take more to sync */ -#define REDIS_EXPIRELOOKUPS_PER_CRON 10 /* try to expire 10 keys/loop */ +#define REDIS_EXPIRELOOKUPS_PER_CRON 10 /* lookup 10 expires per loop */ #define REDIS_MAX_WRITE_PER_EVENT (1024*64) #define REDIS_REQUEST_MAX_SIZE (1024*1024*256) /* max bytes in inline command */ @@ -111,6 +112,7 @@ config file and the server is using more than maxmemory bytes of memory. In short this commands are denied on low memory conditions. */ #define REDIS_CMD_DENYOOM 4 +#define REDIS_CMD_FORCE_REPLICATION 8 /* Force replication even if dirty is 0 */ /* Object types */ #define REDIS_STRING 0 @@ -236,7 +238,9 @@ static char* strencoding[] = { /* We can print the stacktrace, so our assert is defined this way: */ #define redisAssert(_e) ((_e)?(void)0 : (_redisAssert(#_e,__FILE__,__LINE__),_exit(1))) +#define redisPanic(_e) _redisPanic(#_e,__FILE__,__LINE__),_exit(1) static void _redisAssert(char *estr, char *file, int line); +static void _redisPanic(char *msg, char *file, int line); /*================================= Data types ============================== */ @@ -327,6 +331,8 @@ typedef struct redisClient { * is >= blockingto then the operation timed out. */ list *io_keys; /* Keys this client is waiting to be loaded from the * swap file in order to continue. */ + dict *pubsub_channels; /* channels a client is interested in (SUBSCRIBE) */ + list *pubsub_patterns; /* patterns a client is interested in (SUBSCRIBE) */ } redisClient; struct saveparam { @@ -339,8 +345,6 @@ struct redisServer { int port; int fd; redisDb *db; - dict *sharingpool; /* Poll used for object sharing */ - unsigned int sharingpoolsize; long long dirty; /* changes to DB from the last save */ list *clients; list *slaves, *monitors; @@ -362,6 +366,7 @@ struct redisServer { int daemonize; int appendonly; int appendfsync; + int shutdown_asap; time_t lastfsync; int appendfd; int appendseldb; @@ -369,6 +374,7 @@ struct redisServer { pid_t bgsavechildpid; pid_t bgrewritechildpid; sds bgrewritebuf; /* buffer taken by parent during oppend only rewrite */ + sds aofbuf; /* AOF buffer, written before entering the event loop */ struct saveparam *saveparams; int saveparamslen; char *logfile; @@ -376,8 +382,8 @@ struct redisServer { char *dbfilename; char *appendfilename; char *requirepass; - int shareobjects; int rdbcompression; + int activerehashing; /* Replication related */ int isslave; char *masterauth; @@ -435,10 +441,20 @@ struct redisServer { unsigned long long vm_stats_swapped_objects; unsigned long long vm_stats_swapouts; unsigned long long vm_stats_swapins; + /* Pubsub */ + dict *pubsub_channels; /* Map channels to list of subscribed clients */ + list *pubsub_patterns; /* A list of pubsub_patterns */ + /* Misc */ FILE *devnull; }; +typedef struct pubsubPattern { + redisClient *client; + robj *pattern; +} pubsubPattern; + typedef void redisCommandProc(redisClient *c); +typedef void redisVmPreloadProc(redisClient *c, struct redisCommand *cmd, int argc, robj **argv); struct redisCommand { char *name; redisCommandProc *proc; @@ -447,7 +463,7 @@ struct redisCommand { /* Use a function to determine which keys need to be loaded * in the background prior to executing this command. Takes precedence * over vm_firstkey and others, ignored when NULL */ - redisCommandProc *vm_preload_proc; + redisVmPreloadProc *vm_preload_proc; /* What keys should be loaded in background when calling this command? */ int vm_firstkey; /* The first argument that's a key (0 = no keys) */ int vm_lastkey; /* THe last argument that's a key */ @@ -495,13 +511,17 @@ typedef struct zset { /* Our shared "common" objects */ +#define REDIS_SHARED_INTEGERS 10000 struct sharedObjectsStruct { robj *crlf, *ok, *err, *emptybulk, *czero, *cone, *pong, *space, *colon, *nullbulk, *nullmultibulk, *queued, *emptymultibulk, *wrongtypeerr, *nokeyerr, *syntaxerr, *sameobjecterr, *outofrangeerr, *plus, *select0, *select1, *select2, *select3, *select4, - *select5, *select6, *select7, *select8, *select9; + *select5, *select6, *select7, *select8, *select9, + *messagebulk, *pmessagebulk, *subscribebulk, *unsubscribebulk, *mbulk3, + *mbulk4, *psubscribebulk, *punsubscribebulk, + *integers[REDIS_SHARED_INTEGERS]; } shared; /* Global vars that are actally used as constants. The following double @@ -542,10 +562,11 @@ static int rdbSaveBackground(char *filename); static robj *createStringObject(char *ptr, size_t len); static robj *dupStringObject(robj *o); static void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc); +static void replicationFeedMonitors(list *monitors, int dictid, robj **argv, int argc); +static void flushAppendOnlyFile(void); static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc); static int syncWithMaster(void); -static robj *tryObjectSharing(robj *o); -static int tryObjectEncoding(robj *o); +static robj *tryObjectEncoding(robj *o); static robj *getDecodedObject(robj *o); static int removeExpire(redisDb *db, robj *key); static int expireIfNeeded(redisDb *db, robj *key); @@ -592,8 +613,9 @@ static robj *vmReadObjectFromSwap(off_t page, int type); static void waitEmptyIOJobsQueue(void); static void vmReopenSwapFile(void); static int vmFreePage(off_t page); -static void zunionInterBlockClientOnSwappedKeys(redisClient *c); -static int blockClientOnSwappedKeys(struct redisCommand *cmd, redisClient *c); +static void zunionInterBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv); +static void execBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv); +static int blockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd); static int dontWaitForSwappedKey(redisClient *c, robj *key); static void handleClientsBlockedOnSwappedKey(redisDb *db, robj *key); static void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask); @@ -601,12 +623,23 @@ static struct redisCommand *lookupCommand(char *name); static void call(redisClient *c, struct redisCommand *cmd); static void resetClient(redisClient *c); static void convertToRealHash(robj *o); +static int pubsubUnsubscribeAllChannels(redisClient *c, int notify); +static int pubsubUnsubscribeAllPatterns(redisClient *c, int notify); +static void freePubsubPattern(void *p); +static int listMatchPubsubPattern(void *a, void *b); +static int compareStringObjects(robj *a, robj *b); +static int equalStringObjects(robj *a, robj *b); +static void usage(); +static int rewriteAppendOnlyFileBackground(void); +static int vmSwapObjectBlocking(robj *key, robj *val); +static int prepareForShutdown(); static void authCommand(redisClient *c); static void pingCommand(redisClient *c); static void echoCommand(redisClient *c); static void setCommand(redisClient *c); static void setnxCommand(redisClient *c); +static void setexCommand(redisClient *c); static void getCommand(redisClient *c); static void delCommand(redisClient *c); static void existsCommand(redisClient *c); @@ -686,18 +719,26 @@ static void substrCommand(redisClient *c); static void zrankCommand(redisClient *c); static void zrevrankCommand(redisClient *c); static void hsetCommand(redisClient *c); +static void hsetnxCommand(redisClient *c); static void hgetCommand(redisClient *c); +static void hmsetCommand(redisClient *c); +static void hmgetCommand(redisClient *c); static void hdelCommand(redisClient *c); static void hlenCommand(redisClient *c); static void zremrangebyrankCommand(redisClient *c); -static void zunionCommand(redisClient *c); -static void zinterCommand(redisClient *c); +static void zunionstoreCommand(redisClient *c); +static void zinterstoreCommand(redisClient *c); static void hkeysCommand(redisClient *c); static void hvalsCommand(redisClient *c); static void hgetallCommand(redisClient *c); static void hexistsCommand(redisClient *c); static void configCommand(redisClient *c); static void hincrbyCommand(redisClient *c); +static void subscribeCommand(redisClient *c); +static void unsubscribeCommand(redisClient *c); +static void psubscribeCommand(redisClient *c); +static void punsubscribeCommand(redisClient *c); +static void publishCommand(redisClient *c); /*================================= Globals ================================= */ @@ -707,6 +748,7 @@ static struct redisCommand cmdTable[] = { {"get",getCommand,2,REDIS_CMD_INLINE,NULL,1,1,1}, {"set",setCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,0,0,0}, {"setnx",setnxCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,0,0,0}, + {"setex",setexCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,0,0,0}, {"append",appendCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1}, {"substr",substrCommand,4,REDIS_CMD_INLINE,NULL,1,1,1}, {"del",delCommand,-2,REDIS_CMD_INLINE,NULL,0,0,0}, @@ -746,8 +788,8 @@ static struct redisCommand cmdTable[] = { {"zrem",zremCommand,3,REDIS_CMD_BULK,NULL,1,1,1}, {"zremrangebyscore",zremrangebyscoreCommand,4,REDIS_CMD_INLINE,NULL,1,1,1}, {"zremrangebyrank",zremrangebyrankCommand,4,REDIS_CMD_INLINE,NULL,1,1,1}, - {"zunion",zunionCommand,-4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0}, - {"zinter",zinterCommand,-4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0}, + {"zunionstore",zunionstoreCommand,-4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0}, + {"zinterstore",zinterstoreCommand,-4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,zunionInterBlockClientOnSwappedKeys,0,0,0}, {"zrange",zrangeCommand,-4,REDIS_CMD_INLINE,NULL,1,1,1}, {"zrangebyscore",zrangebyscoreCommand,-4,REDIS_CMD_INLINE,NULL,1,1,1}, {"zcount",zcountCommand,4,REDIS_CMD_INLINE,NULL,1,1,1}, @@ -757,8 +799,11 @@ static struct redisCommand cmdTable[] = { {"zrank",zrankCommand,3,REDIS_CMD_BULK,NULL,1,1,1}, {"zrevrank",zrevrankCommand,3,REDIS_CMD_BULK,NULL,1,1,1}, {"hset",hsetCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1}, - {"hincrby",hincrbyCommand,4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,NULL,1,1,1}, + {"hsetnx",hsetnxCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1}, {"hget",hgetCommand,3,REDIS_CMD_BULK,NULL,1,1,1}, + {"hmset",hmsetCommand,-4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM,NULL,1,1,1}, + {"hmget",hmgetCommand,-3,REDIS_CMD_BULK,NULL,1,1,1}, + {"hincrby",hincrbyCommand,4,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,NULL,1,1,1}, {"hdel",hdelCommand,3,REDIS_CMD_BULK,NULL,1,1,1}, {"hlen",hlenCommand,2,REDIS_CMD_INLINE,NULL,1,1,1}, {"hkeys",hkeysCommand,2,REDIS_CMD_INLINE,NULL,1,1,1}, @@ -789,7 +834,7 @@ static struct redisCommand cmdTable[] = { {"lastsave",lastsaveCommand,1,REDIS_CMD_INLINE,NULL,0,0,0}, {"type",typeCommand,2,REDIS_CMD_INLINE,NULL,1,1,1}, {"multi",multiCommand,1,REDIS_CMD_INLINE,NULL,0,0,0}, - {"exec",execCommand,1,REDIS_CMD_INLINE,NULL,0,0,0}, + {"exec",execCommand,1,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM,execBlockClientOnSwappedKeys,0,0,0}, {"discard",discardCommand,1,REDIS_CMD_INLINE,NULL,0,0,0}, {"sync",syncCommand,1,REDIS_CMD_INLINE,NULL,0,0,0}, {"flushdb",flushdbCommand,1,REDIS_CMD_INLINE,NULL,0,0,0}, @@ -801,11 +846,14 @@ static struct redisCommand cmdTable[] = { {"slaveof",slaveofCommand,3,REDIS_CMD_INLINE,NULL,0,0,0}, {"debug",debugCommand,-2,REDIS_CMD_INLINE,NULL,0,0,0}, {"config",configCommand,-2,REDIS_CMD_BULK,NULL,0,0,0}, + {"subscribe",subscribeCommand,-2,REDIS_CMD_INLINE,NULL,0,0,0}, + {"unsubscribe",unsubscribeCommand,-1,REDIS_CMD_INLINE,NULL,0,0,0}, + {"psubscribe",psubscribeCommand,-2,REDIS_CMD_INLINE,NULL,0,0,0}, + {"punsubscribe",punsubscribeCommand,-1,REDIS_CMD_INLINE,NULL,0,0,0}, + {"publish",publishCommand,3,REDIS_CMD_BULK|REDIS_CMD_FORCE_REPLICATION,NULL,0,0,0}, {NULL,NULL,0,0,NULL,0,0,0} }; -static void usage(); - /*============================ Utility functions ============================ */ /* Glob-style pattern matching. */ @@ -935,6 +983,77 @@ static int stringmatch(const char *pattern, const char *string, int nocase) { return stringmatchlen(pattern,strlen(pattern),string,strlen(string),nocase); } +/* Convert a string representing an amount of memory into the number of + * bytes, so for instance memtoll("1Gi") will return 1073741824 that is + * (1024*1024*1024). + * + * On parsing error, if *err is not NULL, it's set to 1, otherwise it's + * set to 0 */ +static long long memtoll(const char *p, int *err) { + const char *u; + char buf[128]; + long mul; /* unit multiplier */ + long long val; + unsigned int digits; + + if (err) *err = 0; + /* Search the first non digit character. */ + u = p; + if (*u == '-') u++; + while(*u && isdigit(*u)) u++; + if (*u == '\0' || !strcasecmp(u,"b")) { + mul = 1; + } else if (!strcasecmp(u,"k")) { + mul = 1000; + } else if (!strcasecmp(u,"kb")) { + mul = 1024; + } else if (!strcasecmp(u,"m")) { + mul = 1000*1000; + } else if (!strcasecmp(u,"mb")) { + mul = 1024*1024; + } else if (!strcasecmp(u,"g")) { + mul = 1000L*1000*1000; + } else if (!strcasecmp(u,"gb")) { + mul = 1024L*1024*1024; + } else { + if (err) *err = 1; + mul = 1; + } + digits = u-p; + if (digits >= sizeof(buf)) { + if (err) *err = 1; + return LLONG_MAX; + } + memcpy(buf,p,digits); + buf[digits] = '\0'; + val = strtoll(buf,NULL,10); + return val*mul; +} + +/* Convert a long long into a string. Returns the number of + * characters needed to represent the number, that can be shorter if passed + * buffer length is not enough to store the whole number. */ +static int ll2string(char *s, size_t len, long long value) { + char buf[32], *p; + unsigned long long v; + size_t l; + + if (len == 0) return 0; + v = (value < 0) ? -value : value; + p = buf+31; /* point to the last character */ + do { + *p-- = '0'+(v%10); + v /= 10; + } while(v); + if (value < 0) *p-- = '-'; + p++; + l = 32-(p-buf); + if (l+1 > len) l = len-1; /* Make sure it fits, including the nul term */ + memcpy(s,p,l); + s[l] = '\0'; + return l; +} + static void redisLog(int level, const char *fmt, ...) { va_list ap; FILE *fp; @@ -1017,8 +1136,8 @@ static int dictEncObjKeyCompare(void *privdata, const void *key1, int cmp; if (o1->encoding == REDIS_ENCODING_INT && - o2->encoding == REDIS_ENCODING_INT && - o1->ptr == o2->ptr) return 1; + o2->encoding == REDIS_ENCODING_INT) + return o1->ptr == o2->ptr; o1 = getDecodedObject(o1); o2 = getDecodedObject(o2); @@ -1038,7 +1157,7 @@ static unsigned int dictEncObjHash(const void *key) { char buf[32]; int len; - len = snprintf(buf,32,"%ld",(long)o->ptr); + len = ll2string(buf,32,(long)o->ptr); return dictGenHashFunction((unsigned char*)buf, len); } else { unsigned int hash; @@ -1141,7 +1260,9 @@ static void closeTimedoutClients(void) { if (server.maxidletime && !(c->flags & REDIS_SLAVE) && /* no timeout for slaves */ !(c->flags & REDIS_MASTER) && /* no timeout for masters */ - (now - c->lastinteraction > server.maxidletime)) + dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */ + listLength(c->pubsub_patterns) == 0 && + (now - c->lastinteraction > server.maxidletime)) { redisLog(REDIS_VERBOSE,"Closing idle client"); freeClient(c); @@ -1169,16 +1290,28 @@ static void tryResizeHashTables(void) { int j; for (j = 0; j < server.dbnum; j++) { - if (htNeedsResize(server.db[j].dict)) { - redisLog(REDIS_VERBOSE,"The hash table %d is too sparse, resize it...",j); + if (htNeedsResize(server.db[j].dict)) dictResize(server.db[j].dict); - redisLog(REDIS_VERBOSE,"Hash table %d resized.",j); - } if (htNeedsResize(server.db[j].expires)) dictResize(server.db[j].expires); } } +/* Our hash table implementation performs rehashing incrementally while + * we write/read from the hash table. Still if the server is idle, the hash + * table will use two tables for a long time. So we try to use 1 millisecond + * of CPU time at every serverCron() loop in order to rehash some key. */ +static void incrementallyRehash(void) { + int j; + + for (j = 0; j < server.dbnum; j++) { + if (dictIsRehashing(server.db[j].dict)) { + dictRehashMilliseconds(server.db[j].dict,1); + break; /* already used our millisecond for this loop... */ + } + } +} + /* A background saving child (BGSAVE) terminated its work. Handle this. */ void backgroundSaveDoneHandler(int statloc) { int exitcode = WEXITSTATUS(statloc); @@ -1193,7 +1326,7 @@ void backgroundSaveDoneHandler(int statloc) { redisLog(REDIS_WARNING, "Background saving error"); } else { redisLog(REDIS_WARNING, - "Background saving terminated by signal"); + "Background saving terminated by signal %d", WTERMSIG(statloc)); rdbRemoveTempFile(server.bgsavechildpid); } server.bgsavechildpid = -1; @@ -1254,7 +1387,8 @@ void backgroundRewriteDoneHandler(int statloc) { redisLog(REDIS_WARNING, "Background append only file rewriting error"); } else { redisLog(REDIS_WARNING, - "Background append only file rewriting terminated by signal"); + "Background append only file rewriting terminated by signal %d", + WTERMSIG(statloc)); } cleanup: sdsfree(server.bgrewritebuf); @@ -1263,6 +1397,19 @@ cleanup: server.bgrewritechildpid = -1; } +/* This function is called once a background process of some kind terminates, + * as we want to avoid resizing the hash tables when there is a child in order + * to play well with copy-on-write (otherwise when a resize happens lots of + * memory pages are copied). The goal of this function is to update the ability + * for dict.c to resize the hash tables accordingly to the fact we have o not + * running childs. */ +static void updateDictResizePolicy(void) { + if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1) + dictEnableResize(); + else + dictDisableResize(); +} + static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) { int j, loops = server.cronloops++; REDIS_NOTUSED(eventLoop); @@ -1275,6 +1422,13 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD * To access a global var is faster than calling time(NULL) */ server.unixtime = time(NULL); + /* We received a SIGTERM, shutting down here in a safe way, as it is + * not ok doing so inside the signal handler. */ + if (server.shutdown_asap) { + if (prepareForShutdown() == REDIS_OK) exit(0); + redisLog(REDIS_WARNING,"SIGTERM received but errors trying to shut down the server, check the logs for more information"); + } + /* Show some info about non-empty databases */ for (j = 0; j < server.dbnum; j++) { long long size, used, vkeys; @@ -1294,15 +1448,17 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD * if we resize the HT while there is the saving child at work actually * a lot of memory movements in the parent will cause a lot of pages * copied. */ - if (server.bgsavechildpid == -1 && !(loops % 10)) tryResizeHashTables(); + if (server.bgsavechildpid == -1 && server.bgrewritechildpid == -1) { + if (!(loops % 10)) tryResizeHashTables(); + if (server.activerehashing) incrementallyRehash(); + } /* Show information about connected clients */ if (!(loops % 50)) { - redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use, %d shared objects", + redisLog(REDIS_VERBOSE,"%d clients connected (%d slaves), %zu bytes in use", listLength(server.clients)-listLength(server.slaves), listLength(server.slaves), - zmalloc_used_memory(), - dictSize(server.sharingpool)); + zmalloc_used_memory()); } /* Close connections of timedout clients */ @@ -1320,6 +1476,7 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD } else { backgroundRewriteDoneHandler(statloc); } + updateDictResizePolicy(); } } else { /* If there is not a background saving in progress check if @@ -1401,6 +1558,7 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD redisLog(REDIS_NOTICE,"Connecting to MASTER..."); if (syncWithMaster() == REDIS_OK) { redisLog(REDIS_NOTICE,"MASTER <-> SLAVE sync succeeded"); + if (server.appendonly) rewriteAppendOnlyFileBackground(); } } return 100; @@ -1412,6 +1570,7 @@ static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientD static void beforeSleep(struct aeEventLoop *eventLoop) { REDIS_NOTUSED(eventLoop); + /* Awake clients that got all the swapped keys they requested */ if (server.vm_enabled && listLength(server.io_ready_clients)) { listIter li; listNode *ln; @@ -1436,9 +1595,13 @@ static void beforeSleep(struct aeEventLoop *eventLoop) { processInputBuffer(c); } } + /* Write the AOF buffer on disk */ + flushAppendOnlyFile(); } static void createSharedObjects(void) { + int j; + shared.crlf = createObject(REDIS_STRING,sdsnew("\r\n")); shared.ok = createObject(REDIS_STRING,sdsnew("+OK\r\n")); shared.err = createObject(REDIS_STRING,sdsnew("-ERR\r\n")); @@ -1473,6 +1636,18 @@ static void createSharedObjects(void) { shared.select7 = createStringObject("select 7\r\n",10); shared.select8 = createStringObject("select 8\r\n",10); shared.select9 = createStringObject("select 9\r\n",10); + shared.messagebulk = createStringObject("$7\r\nmessage\r\n",13); + shared.pmessagebulk = createStringObject("$8\r\npmessage\r\n",14); + shared.subscribebulk = createStringObject("$9\r\nsubscribe\r\n",15); + shared.unsubscribebulk = createStringObject("$11\r\nunsubscribe\r\n",18); + shared.psubscribebulk = createStringObject("$10\r\npsubscribe\r\n",17); + shared.punsubscribebulk = createStringObject("$12\r\npunsubscribe\r\n",19); + shared.mbulk3 = createStringObject("*3\r\n",4); + shared.mbulk4 = createStringObject("*4\r\n",4); + for (j = 0; j < REDIS_SHARED_INTEGERS; j++) { + shared.integers[j] = createObject(REDIS_STRING,(void*)(long)j); + shared.integers[j]->encoding = REDIS_ENCODING_INT; + } } static void appendServerSaveParams(time_t seconds, int changes) { @@ -1499,7 +1674,7 @@ static void initServerConfig() { server.glueoutputbuf = 1; server.daemonize = 0; server.appendonly = 0; - server.appendfsync = APPENDFSYNC_ALWAYS; + server.appendfsync = APPENDFSYNC_EVERYSEC; server.lastfsync = time(NULL); server.appendfd = -1; server.appendseldb = -1; /* Make sure the first time will not match */ @@ -1507,9 +1682,8 @@ static void initServerConfig() { server.dbfilename = zstrdup("dump.rdb"); server.appendfilename = zstrdup("appendonly.aof"); server.requirepass = NULL; - server.shareobjects = 0; server.rdbcompression = 1; - server.sharingpoolsize = 1024; + server.activerehashing = 1; server.maxclients = 0; server.blpop_blocked_clients = 0; server.maxmemory = 0; @@ -1522,6 +1696,7 @@ static void initServerConfig() { server.vm_blocked_clients = 0; server.hash_max_zipmap_entries = REDIS_HASH_MAX_ZIPMAP_ENTRIES; server.hash_max_zipmap_value = REDIS_HASH_MAX_ZIPMAP_VALUE; + server.shutdown_asap = 0; resetServerSaveParams(); @@ -1562,7 +1737,6 @@ static void initServer() { createSharedObjects(); server.el = aeCreateEventLoop(); server.db = zmalloc(sizeof(redisDb)*server.dbnum); - server.sharingpool = dictCreate(&setDictType,NULL); server.fd = anetTcpServer(server.neterr, server.port, server.bindaddr); if (server.fd == -1) { redisLog(REDIS_WARNING, "Opening TCP port: %s", server.neterr); @@ -1576,10 +1750,15 @@ static void initServer() { server.db[j].io_keys = dictCreate(&keylistDictType,NULL); server.db[j].id = j; } + server.pubsub_channels = dictCreate(&keylistDictType,NULL); + server.pubsub_patterns = listCreate(); + listSetFreeMethod(server.pubsub_patterns,freePubsubPattern); + listSetMatchMethod(server.pubsub_patterns,listMatchPubsubPattern); server.cronloops = 0; server.bgsavechildpid = -1; server.bgrewritechildpid = -1; server.bgrewritebuf = sdsempty(); + server.aofbuf = sdsempty(); server.lastsave = time(NULL); server.dirty = 0; server.stat_numcommands = 0; @@ -1629,15 +1808,12 @@ static void loadServerConfig(char *filename) { char buf[REDIS_CONFIGLINE_MAX+1], *err = NULL; int linenum = 0; sds line = NULL; - char *errormsg = "Fatal error, can't open config file '%s'"; - char *errorbuf = zmalloc(sizeof(char)*(strlen(errormsg)+strlen(filename))); - sprintf(errorbuf, errormsg, filename); if (filename[0] == '-' && filename[1] == '\0') fp = stdin; else { if ((fp = fopen(filename,"r")) == NULL) { - redisLog(REDIS_WARNING, errorbuf); + redisLog(REDIS_WARNING, "Fatal error, can't open config file '%s'", filename); exit(1); } } @@ -1724,7 +1900,7 @@ 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 = strtoll(argv[1], NULL, 10); + server.maxmemory = memtoll(argv[1],NULL); } else if (!strcasecmp(argv[0],"slaveof") && argc == 3) { server.masterhost = sdsnew(argv[1]); server.masterport = atoi(argv[2]); @@ -1735,18 +1911,13 @@ static void loadServerConfig(char *filename) { if ((server.glueoutputbuf = yesnotoi(argv[1])) == -1) { err = "argument must be 'yes' or 'no'"; goto loaderr; } - } else if (!strcasecmp(argv[0],"shareobjects") && argc == 2) { - if ((server.shareobjects = yesnotoi(argv[1])) == -1) { - err = "argument must be 'yes' or 'no'"; goto loaderr; - } } else if (!strcasecmp(argv[0],"rdbcompression") && argc == 2) { if ((server.rdbcompression = yesnotoi(argv[1])) == -1) { err = "argument must be 'yes' or 'no'"; goto loaderr; } - } else if (!strcasecmp(argv[0],"shareobjectspoolsize") && argc == 2) { - server.sharingpoolsize = atoi(argv[1]); - if (server.sharingpoolsize < 1) { - err = "invalid object sharing pool size"; goto loaderr; + } else if (!strcasecmp(argv[0],"activerehashing") && argc == 2) { + if ((server.activerehashing = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; } } else if (!strcasecmp(argv[0],"daemonize") && argc == 2) { if ((server.daemonize = yesnotoi(argv[1])) == -1) { @@ -1756,6 +1927,9 @@ static void loadServerConfig(char *filename) { if ((server.appendonly = yesnotoi(argv[1])) == -1) { err = "argument must be 'yes' or 'no'"; goto loaderr; } + } else if (!strcasecmp(argv[0],"appendfilename") && argc == 2) { + zfree(server.appendfilename); + server.appendfilename = zstrdup(argv[1]); } else if (!strcasecmp(argv[0],"appendfsync") && argc == 2) { if (!strcasecmp(argv[1],"no")) { server.appendfsync = APPENDFSYNC_NO; @@ -1783,19 +1957,17 @@ static void loadServerConfig(char *filename) { zfree(server.vm_swap_file); server.vm_swap_file = zstrdup(argv[1]); } else if (!strcasecmp(argv[0],"vm-max-memory") && argc == 2) { - server.vm_max_memory = strtoll(argv[1], NULL, 10); + server.vm_max_memory = memtoll(argv[1],NULL); } else if (!strcasecmp(argv[0],"vm-page-size") && argc == 2) { - server.vm_page_size = strtoll(argv[1], NULL, 10); + server.vm_page_size = memtoll(argv[1], NULL); } else if (!strcasecmp(argv[0],"vm-pages") && argc == 2) { - server.vm_pages = strtoll(argv[1], NULL, 10); + server.vm_pages = memtoll(argv[1], NULL); } else if (!strcasecmp(argv[0],"vm-max-threads") && argc == 2) { server.vm_max_threads = strtoll(argv[1], NULL, 10); } else if (!strcasecmp(argv[0],"hash-max-zipmap-entries") && argc == 2){ - server.hash_max_zipmap_entries = strtol(argv[1], NULL, 10); + server.hash_max_zipmap_entries = memtoll(argv[1], NULL); } else if (!strcasecmp(argv[0],"hash-max-zipmap-value") && argc == 2){ - server.hash_max_zipmap_value = strtol(argv[1], NULL, 10); - } else if (!strcasecmp(argv[0],"vm-max-threads") && argc == 2) { - server.vm_max_threads = strtoll(argv[1], NULL, 10); + server.hash_max_zipmap_value = memtoll(argv[1], NULL); } else { err = "Bad directive or wrong number of arguments"; goto loaderr; } @@ -1839,6 +2011,12 @@ static void freeClient(redisClient *c) { if (c->flags & REDIS_BLOCKED) unblockClientWaitingData(c); + /* Unsubscribe from all the pubsub channels */ + pubsubUnsubscribeAllChannels(c,0); + pubsubUnsubscribeAllPatterns(c,0); + dictRelease(c->pubsub_channels); + listRelease(c->pubsub_patterns); + /* Obvious cleanup */ aeDeleteFileEvent(server.el,c->fd,AE_READABLE); aeDeleteFileEvent(server.el,c->fd,AE_WRITABLE); listRelease(c->reply); @@ -1861,7 +2039,7 @@ static void freeClient(redisClient *c) { dontWaitForSwappedKey(c,ln->value); } listRelease(c->io_keys); - /* Other cleanup */ + /* Master/slave cleanup */ if (c->flags & REDIS_SLAVE) { if (c->replstate == REDIS_REPL_SEND_BULK && c->repldbfd != -1) close(c->repldbfd); @@ -1874,6 +2052,7 @@ static void freeClient(redisClient *c) { server.master = NULL; server.replstate = REDIS_REPL_CONNECT; } + /* Release memory */ zfree(c->argv); zfree(c->mbargv); freeClientMultiState(c); @@ -1917,7 +2096,7 @@ static void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) /* Use writev() if we have enough buffers to send */ if (!server.glueoutputbuf && - listLength(c->reply) > REDIS_WRITEV_THRESHOLD && + listLength(c->reply) > REDIS_WRITEV_THRESHOLD && !(c->flags & REDIS_MASTER)) { sendReplyToClientWritev(el, fd, privdata, mask); @@ -1995,7 +2174,7 @@ static void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int o = listNodeValue(node); objlen = sdslen(o->ptr); - if (totwritten + objlen - offset > REDIS_MAX_WRITE_PER_EVENT) + if (totwritten + objlen - offset > REDIS_MAX_WRITE_PER_EVENT) break; if(ion == REDIS_WRITEV_IOVEC_COUNT) @@ -2043,7 +2222,7 @@ static void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int } } - if (totwritten > 0) + if (totwritten > 0) c->lastinteraction = time(NULL); if (listLength(c->reply) == 0) { @@ -2074,12 +2253,15 @@ static void call(redisClient *c, struct redisCommand *cmd) { dirty = server.dirty; cmd->proc(c); - if (server.appendonly && server.dirty-dirty) + dirty = server.dirty-dirty; + + if (server.appendonly && dirty) feedAppendOnlyFile(cmd,c->db->id,c->argv,c->argc); - if (server.dirty-dirty && listLength(server.slaves)) + if ((dirty || cmd->flags & REDIS_CMD_FORCE_REPLICATION) && + listLength(server.slaves)) replicationFeedSlaves(server.slaves,c->db->id,c->argv,c->argc); if (listLength(server.monitors)) - replicationFeedSlaves(server.monitors,c->db->id,c->argv,c->argc); + replicationFeedMonitors(server.monitors,c->db->id,c->argv,c->argc); server.stat_numcommands++; } @@ -2217,15 +2399,9 @@ static int processCommand(redisClient *c) { return 1; } } - /* Let's try to share objects on the command arguments vector */ - if (server.shareobjects) { - int j; - 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]); + c->argv[c->argc-1] = tryObjectEncoding(c->argv[c->argc-1]); /* Check if the user is authenticated */ if (server.requirepass && !c->authenticated && cmd->proc != authCommand) { @@ -2243,13 +2419,23 @@ static int processCommand(redisClient *c) { return 1; } + /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */ + if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0) + && + cmd->proc != subscribeCommand && cmd->proc != unsubscribeCommand && + cmd->proc != psubscribeCommand && cmd->proc != punsubscribeCommand) { + addReplySds(c,sdsnew("-ERR only (P)SUBSCRIBE / (P)UNSUBSCRIBE / QUIT allowed in this context\r\n")); + resetClient(c); + return 1; + } + /* Exec the command */ if (c->flags & REDIS_MULTI && cmd->proc != execCommand && cmd->proc != discardCommand) { queueMultiCommand(c,cmd); addReply(c,shared.queued); } else { if (server.vm_enabled && server.vm_max_threads > 0 && - blockClientOnSwappedKeys(cmd,c)) return 1; + blockClientOnSwappedKeys(c,cmd)) return 1; call(c,cmd); } @@ -2330,6 +2516,64 @@ static void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int arg if (outv != static_outv) zfree(outv); } +static sds sdscatrepr(sds s, char *p, size_t len) { + s = sdscatlen(s,"\"",1); + while(len--) { + switch(*p) { + case '\\': + case '"': + s = sdscatprintf(s,"\\%c",*p); + break; + case '\n': s = sdscatlen(s,"\\n",1); break; + case '\r': s = sdscatlen(s,"\\r",1); break; + case '\t': s = sdscatlen(s,"\\t",1); break; + case '\a': s = sdscatlen(s,"\\a",1); break; + case '\b': s = sdscatlen(s,"\\b",1); break; + default: + if (isprint(*p)) + s = sdscatprintf(s,"%c",*p); + else + s = sdscatprintf(s,"\\x%02x",(unsigned char)*p); + break; + } + p++; + } + return sdscatlen(s,"\"",1); +} + +static void replicationFeedMonitors(list *monitors, int dictid, robj **argv, int argc) { + listNode *ln; + listIter li; + int j; + sds cmdrepr = sdsnew("+"); + robj *cmdobj; + struct timeval tv; + + gettimeofday(&tv,NULL); + cmdrepr = sdscatprintf(cmdrepr,"%ld.%ld ",(long)tv.tv_sec,(long)tv.tv_usec); + if (dictid != 0) cmdrepr = sdscatprintf(cmdrepr,"(db %d) ", dictid); + + for (j = 0; j < argc; j++) { + if (argv[j]->encoding == REDIS_ENCODING_INT) { + cmdrepr = sdscatprintf(cmdrepr, "%ld", (long)argv[j]->ptr); + } else { + cmdrepr = sdscatrepr(cmdrepr,(char*)argv[j]->ptr, + sdslen(argv[j]->ptr)); + } + if (j != argc-1) + cmdrepr = sdscatlen(cmdrepr," ",1); + } + cmdrepr = sdscatlen(cmdrepr,"\r\n",2); + cmdobj = createObject(REDIS_STRING,cmdrepr); + + listRewind(monitors,&li); + while((ln = listNext(&li))) { + redisClient *monitor = ln->value; + addReply(monitor,cmdobj); + } + decrRefCount(cmdobj); +} + static void processInputBuffer(redisClient *c) { again: /* Before to process the input buffer, make sure the client is not @@ -2347,7 +2591,7 @@ again: if (p) { sds query, *argv; int argc, j; - + query = c->querybuf; c->querybuf = sdsempty(); querylen = 1+(p-(query)); @@ -2454,6 +2698,10 @@ static void *dupClientReplyValue(void *o) { return o; } +static int listMatchObjects(void *a, void *b) { + return equalStringObjects(a,b); +} + static redisClient *createClient(int fd) { redisClient *c = zmalloc(sizeof(*c)); @@ -2481,6 +2729,10 @@ static redisClient *createClient(int fd) { c->blockingkeysnum = 0; c->io_keys = listCreate(); listSetFreeMethod(c->io_keys,decrRefCount); + c->pubsub_channels = dictCreate(&setDictType,NULL); + c->pubsub_patterns = listCreate(); + listSetFreeMethod(c->pubsub_patterns,decrRefCount); + listSetMatchMethod(c->pubsub_patterns,listMatchObjects); if (aeCreateFileEvent(server.el, c->fd, AE_READABLE, readQueryFromClient, c) == AE_ERR) { freeClient(c); @@ -2519,19 +2771,22 @@ static void addReplyDouble(redisClient *c, double d) { (unsigned long) strlen(buf),buf)); } -static void addReplyLong(redisClient *c, long l) { +static void addReplyLongLong(redisClient *c, long long ll) { char buf[128]; size_t len; - if (l == 0) { + if (ll == 0) { addReply(c,shared.czero); return; - } else if (l == 1) { + } else if (ll == 1) { addReply(c,shared.cone); return; } - len = snprintf(buf,sizeof(buf),":%ld\r\n",l); - addReplySds(c,sdsnewlen(buf,len)); + buf[0] = ':'; + len = ll2string(buf+1,sizeof(buf)-1,ll); + buf[len+1] = '\r'; + buf[len+2] = '\n'; + addReplySds(c,sdsnewlen(buf,len+3)); } static void addReplyUlong(redisClient *c, unsigned long ul) { @@ -2550,7 +2805,8 @@ static void addReplyUlong(redisClient *c, unsigned long ul) { } static void addReplyBulkLen(redisClient *c, robj *obj) { - size_t len; + size_t len, intlen; + char buf[128]; if (obj->encoding == REDIS_ENCODING_RAW) { len = sdslen(obj->ptr); @@ -2567,7 +2823,11 @@ static void addReplyBulkLen(redisClient *c, robj *obj) { len++; } } - addReplySds(c,sdscatprintf(sdsempty(),"$%lu\r\n",(unsigned long)len)); + buf[0] = '$'; + intlen = ll2string(buf+1,sizeof(buf)-1,(long long)len); + buf[intlen+1] = '\r'; + buf[intlen+2] = '\n'; + addReplySds(c,sdsnewlen(buf,intlen+3)); } static void addReplyBulk(redisClient *c, robj *obj) { @@ -2661,6 +2921,23 @@ static robj *createStringObject(char *ptr, size_t len) { return createObject(REDIS_STRING,sdsnewlen(ptr,len)); } +static robj *createStringObjectFromLongLong(long long value) { + robj *o; + if (value >= 0 && value < REDIS_SHARED_INTEGERS) { + incrRefCount(shared.integers[value]); + o = shared.integers[value]; + } else { + if (value >= LONG_MIN && value <= LONG_MAX) { + o = createObject(REDIS_STRING, NULL); + o->encoding = REDIS_ENCODING_INT; + o->ptr = (void*)((long)value); + } else { + o = createObject(REDIS_STRING,sdsfromlonglong(value)); + } + } + return o; +} + static robj *dupStringObject(robj *o) { assert(o->encoding == REDIS_ENCODING_RAW); return createStringObject(o->ptr,sdslen(o->ptr)); @@ -2727,27 +3004,24 @@ static void freeHashObject(robj *o) { zfree(o->ptr); break; default: - redisAssert(0); + redisPanic("Unknown hash encoding type"); break; } } static void incrRefCount(robj *o) { - redisAssert(!server.vm_enabled || o->storage == REDIS_VM_MEMORY); o->refcount++; } static void decrRefCount(void *obj) { robj *o = obj; + if (o->refcount <= 0) redisPanic("decrRefCount against refcount <= 0"); /* Object is a key of a swapped out value, or in the process of being * loaded. */ if (server.vm_enabled && (o->storage == REDIS_VM_SWAPPED || o->storage == REDIS_VM_LOADING)) { - if (o->storage == REDIS_VM_SWAPPED || o->storage == REDIS_VM_LOADING) { - redisAssert(o->refcount == 1); - } if (o->storage == REDIS_VM_LOADING) vmCancelThreadedIOJob(obj); redisAssert(o->type == REDIS_STRING); freeStringObject(o); @@ -2770,7 +3044,7 @@ static void decrRefCount(void *obj) { case REDIS_SET: freeSetObject(o); break; case REDIS_ZSET: freeZsetObject(o); break; case REDIS_HASH: freeHashObject(o); break; - default: redisAssert(0); break; + default: redisPanic("Unknown object type"); break; } if (server.vm_enabled) pthread_mutex_lock(&server.obj_freelist_mutex); if (listLength(server.objfreelist) > REDIS_OBJFREELIST_MAX || @@ -2860,51 +3134,6 @@ static int deleteKey(redisDb *db, robj *key) { return retval == DICT_OK; } -/* Try to share an object against the shared objects pool */ -static robj *tryObjectSharing(robj *o) { - struct dictEntry *de; - unsigned long c; - - if (o == NULL || server.shareobjects == 0) return o; - - redisAssert(o->type == REDIS_STRING); - de = dictFind(server.sharingpool,o); - if (de) { - robj *shared = dictGetEntryKey(de); - - c = ((unsigned long) dictGetEntryVal(de))+1; - dictGetEntryVal(de) = (void*) c; - incrRefCount(shared); - decrRefCount(o); - return shared; - } else { - /* Here we are using a stream algorihtm: Every time an object is - * shared we increment its count, everytime there is a miss we - * recrement the counter of a random object. If this object reaches - * zero we remove the object and put the current object instead. */ - if (dictSize(server.sharingpool) >= - server.sharingpoolsize) { - de = dictGetRandomKey(server.sharingpool); - redisAssert(de != NULL); - c = ((unsigned long) dictGetEntryVal(de))-1; - dictGetEntryVal(de) = (void*) c; - if (c == 0) { - dictDelete(server.sharingpool,de->key); - } - } else { - c = 0; /* If the pool is empty we want to add this object */ - } - if (c == 0) { - int retval; - - retval = dictAdd(server.sharingpool,o,(void*)1); - redisAssert(retval == DICT_OK); - incrRefCount(o); - } - return o; - } -} - /* Check if the nul-terminated string 's' can be represented by a long * (that is, is a number that fits into long without any other space or * character before or after the digits). @@ -2915,10 +3144,10 @@ static int isStringRepresentableAsLong(sds s, long *longval) { char buf[32], *endptr; long value; int slen; - + value = strtol(s, &endptr, 10); if (endptr[0] != '\0') return REDIS_ERR; - slen = snprintf(buf,32,"%ld",value); + slen = ll2string(buf,32,value); /* If the number converted back into a string is not identical * then it's not possible to encode the string as integer */ @@ -2928,36 +3157,42 @@ static int isStringRepresentableAsLong(sds s, long *longval) { } /* Try to encode a string object in order to save space */ -static int tryObjectEncoding(robj *o) { +static robj *tryObjectEncoding(robj *o) { long value; sds s = o->ptr; if (o->encoding != REDIS_ENCODING_RAW) - return REDIS_ERR; /* Already encoded */ + return o; /* Already encoded */ - /* It's not save to encode shared objects: shared objects can be shared + /* It's not safe 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; + if (o->refcount > 1) return o; /* Currently we try to encode only strings */ redisAssert(o->type == REDIS_STRING); /* Check if we can represent this string as a long integer */ - if (isStringRepresentableAsLong(s,&value) == REDIS_ERR) return REDIS_ERR; + if (isStringRepresentableAsLong(s,&value) == REDIS_ERR) return o; /* Ok, this object can be encoded */ - o->encoding = REDIS_ENCODING_INT; - sdsfree(o->ptr); - o->ptr = (void*) value; - return REDIS_OK; + if (value >= 0 && value < REDIS_SHARED_INTEGERS) { + decrRefCount(o); + incrRefCount(shared.integers[value]); + return shared.integers[value]; + } else { + o->encoding = REDIS_ENCODING_INT; + sdsfree(o->ptr); + o->ptr = (void*) value; + return o; + } } /* Get a decoded version of an encoded object (returned as a new object). * If the object is already raw-encoded just increment the ref count. */ static robj *getDecodedObject(robj *o) { robj *dec; - + if (o->encoding == REDIS_ENCODING_RAW) { incrRefCount(o); return o; @@ -2965,17 +3200,17 @@ static robj *getDecodedObject(robj *o) { if (o->type == REDIS_STRING && o->encoding == REDIS_ENCODING_INT) { char buf[32]; - snprintf(buf,32,"%ld",(long)o->ptr); + ll2string(buf,32,(long)o->ptr); dec = createStringObject(buf,strlen(buf)); return dec; } else { - redisAssert(1 != 1); + redisPanic("Unknown encoding type"); } } /* 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 + * use ll2string() to get a string representation of the numbers on the stack * and compare the strings, it's much faster than calling getDecodedObject(). * * Important note: if objects are not integer encoded, but binary-safe strings, @@ -2988,14 +3223,14 @@ static int compareStringObjects(robj *a, robj *b) { if (a == b) return 0; if (a->encoding != REDIS_ENCODING_RAW) { - snprintf(bufa,sizeof(bufa),"%ld",(long) a->ptr); + ll2string(bufa,sizeof(bufa),(long) a->ptr); astr = bufa; bothsds = 0; } else { astr = a->ptr; } if (b->encoding != REDIS_ENCODING_RAW) { - snprintf(bufb,sizeof(bufb),"%ld",(long) b->ptr); + ll2string(bufb,sizeof(bufb),(long) b->ptr); bstr = bufb; bothsds = 0; } else { @@ -3004,6 +3239,18 @@ static int compareStringObjects(robj *a, robj *b) { return bothsds ? sdscmp(astr,bstr) : strcmp(astr,bstr); } +/* Equal string objects return 1 if the two objects are the same from the + * point of view of a string comparison, otherwise 0 is returned. Note that + * this function is faster then checking for (compareStringObject(a,b) == 0) + * because it can perform some more optimization. */ +static int equalStringObjects(robj *a, robj *b) { + if (a->encoding != REDIS_ENCODING_RAW && b->encoding != REDIS_ENCODING_RAW){ + return a->ptr == b->ptr; + } else { + return compareStringObjects(a,b) == 0; + } +} + static size_t stringObjectLen(robj *o) { redisAssert(o->type == REDIS_STRING); if (o->encoding == REDIS_ENCODING_RAW) { @@ -3011,8 +3258,99 @@ static size_t stringObjectLen(robj *o) { } else { char buf[32]; - return snprintf(buf,32,"%ld",(long)o->ptr); + return ll2string(buf,32,(long)o->ptr); + } +} + +static int getDoubleFromObject(robj *o, double *target) { + double value; + char *eptr; + + if (o == NULL) { + value = 0; + } else { + redisAssert(o->type == REDIS_STRING); + if (o->encoding == REDIS_ENCODING_RAW) { + value = strtod(o->ptr, &eptr); + if (eptr[0] != '\0') return REDIS_ERR; + } else if (o->encoding == REDIS_ENCODING_INT) { + value = (long)o->ptr; + } else { + redisPanic("Unknown string encoding"); + } + } + + *target = value; + return REDIS_OK; +} + +static int getDoubleFromObjectOrReply(redisClient *c, robj *o, double *target, const char *msg) { + double value; + if (getDoubleFromObject(o, &value) != REDIS_OK) { + if (msg != NULL) { + addReplySds(c, sdscatprintf(sdsempty(), "-ERR %s\r\n", msg)); + } else { + addReplySds(c, sdsnew("-ERR value is not a double\r\n")); + } + return REDIS_ERR; + } + + *target = value; + return REDIS_OK; +} + +static int getLongLongFromObject(robj *o, long long *target) { + long long value; + char *eptr; + + if (o == NULL) { + value = 0; + } else { + redisAssert(o->type == REDIS_STRING); + if (o->encoding == REDIS_ENCODING_RAW) { + value = strtoll(o->ptr, &eptr, 10); + if (eptr[0] != '\0') return REDIS_ERR; + } else if (o->encoding == REDIS_ENCODING_INT) { + value = (long)o->ptr; + } else { + redisPanic("Unknown string encoding"); + } + } + + *target = value; + return REDIS_OK; +} + +static int getLongLongFromObjectOrReply(redisClient *c, robj *o, long long *target, const char *msg) { + long long value; + if (getLongLongFromObject(o, &value) != REDIS_OK) { + if (msg != NULL) { + addReplySds(c, sdscatprintf(sdsempty(), "-ERR %s\r\n", msg)); + } else { + addReplySds(c, sdsnew("-ERR value is not an integer\r\n")); + } + return REDIS_ERR; + } + + *target = value; + return REDIS_OK; +} + +static int getLongFromObjectOrReply(redisClient *c, robj *o, long *target, const char *msg) { + long long value; + + if (getLongLongFromObjectOrReply(c, o, &value, msg) != REDIS_OK) return REDIS_ERR; + if (value < LONG_MIN || value > LONG_MAX) { + if (msg != NULL) { + addReplySds(c, sdscatprintf(sdsempty(), "-ERR %s\r\n", msg)); + } else { + addReplySds(c, sdsnew("-ERR value is out of range\r\n")); + } + return REDIS_ERR; } + + *target = value; + return REDIS_OK; } /*============================ RDB saving/loading =========================== */ @@ -3051,22 +3389,12 @@ static int rdbSaveLen(FILE *fp, uint32_t len) { return 0; } -/* String objects in the form "2391" "-100" without any space and with a - * range of values that can fit in an 8, 16 or 32 bit signed value can be - * encoded as integers to save space */ -static int rdbTryIntegerEncoding(char *s, size_t len, unsigned char *enc) { - long long value; - char *endptr, buf[32]; - - /* Check if it's possible to encode this value as a number */ - value = strtoll(s, &endptr, 10); - if (endptr[0] != '\0') return 0; - snprintf(buf,32,"%lld",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) != len || memcmp(buf,s,len)) return 0; - +/* Encode 'value' as an integer if possible (if integer will fit the + * supported range). If the function sucessful encoded the integer + * then the (up to 5 bytes) encoded representation is written in the + * string pointed by 'enc' and the length is returned. Otherwise + * 0 is returned. */ +static int rdbEncodeInteger(long long value, unsigned char *enc) { /* Finally check if it fits in our ranges */ if (value >= -(1<<7) && value <= (1<<7)-1) { enc[0] = (REDIS_RDB_ENCVAL<<6)|REDIS_RDB_ENC_INT8; @@ -3089,6 +3417,25 @@ static int rdbTryIntegerEncoding(char *s, size_t len, unsigned char *enc) { } } +/* String objects in the form "2391" "-100" without any space and with a + * range of values that can fit in an 8, 16 or 32 bit signed value can be + * encoded as integers to save space */ +static int rdbTryIntegerEncoding(char *s, size_t len, unsigned char *enc) { + long long value; + char *endptr, buf[32]; + + /* Check if it's possible to encode this value as a number */ + value = strtoll(s, &endptr, 10); + if (endptr[0] != '\0') return 0; + ll2string(buf,32,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) != len || memcmp(buf,s,len)) return 0; + + return rdbEncodeInteger(value,enc); +} + static int rdbSaveLzfStringObject(FILE *fp, unsigned char *s, size_t len) { size_t comprlen, outlen; unsigned char byte; @@ -3152,6 +3499,21 @@ static int rdbSaveRawString(FILE *fp, unsigned char *s, size_t len) { static int rdbSaveStringObject(FILE *fp, robj *obj) { int retval; + /* Avoid to decode the object, then encode it again, if the + * object is alrady integer encoded. */ + if (obj->encoding == REDIS_ENCODING_INT) { + long val = (long) obj->ptr; + unsigned char buf[5]; + int enclen; + + if ((enclen = rdbEncodeInteger(val,buf)) > 0) { + if (fwrite(buf,enclen,1,fp) == 0) return -1; + return 0; + } + /* otherwise... fall throught and continue with the usual + * code path. */ + } + /* Avoid incr/decr ref count business when possible. * This plays well with copy-on-write given that we are probably * in a child process (BGSAVE). Also this makes sure key objects @@ -3186,7 +3548,23 @@ static int rdbSaveDoubleValue(FILE *fp, double val) { len = 1; buf[0] = (val < 0) ? 255 : 254; } else { - snprintf((char*)buf+1,sizeof(buf)-1,"%.17g",val); +#if (DBL_MANT_DIG >= 52) && (LLONG_MAX == 0x7fffffffffffffffLL) + /* Check if the float is in a safe range to be casted into a + * long long. We are assuming that long long is 64 bit here. + * Also we are assuming that there are no implementations around where + * double has precision < 52 bit. + * + * Under this assumptions we test if a double is inside an interval + * where casting to long long is safe. Then using two castings we + * make sure the decimal part is zero. If all this is true we use + * integer printing function that is much faster. */ + double min = -4503599627370495; /* (2^52)-1 */ + double max = 4503599627370496; /* -(2^52) */ + if (val > min && val < max && val == ((double)((long long)val))) + ll2string((char*)buf+1,sizeof(buf),(long long)val); + else +#endif + snprintf((char*)buf+1,sizeof(buf)-1,"%.17g",val); buf[0] = strlen((char*)buf+1); len = buf[0]+1; } @@ -3268,7 +3646,7 @@ static int rdbSaveObject(FILE *fp, robj *o) { dictReleaseIterator(di); } } else { - redisAssert(0); + redisPanic("Unknown object type"); } return 0; } @@ -3287,7 +3665,7 @@ static off_t rdbSavedObjectLen(robj *o, FILE *fp) { /* Return the number of pages required to save this object in the swap file */ static off_t rdbSavedObjectPages(robj *o, FILE *fp) { off_t bytes = rdbSavedObjectLen(o,fp); - + return (bytes+(server.vm_page_size-1))/server.vm_page_size; } @@ -3370,7 +3748,7 @@ static int rdbSave(char *filename) { fflush(fp); fsync(fileno(fp)); fclose(fp); - + /* Use RENAME to make sure the DB file is changed atomically only * if the generate DB file is ok. */ if (rename(tmpfile,filename) == -1) { @@ -3414,6 +3792,7 @@ static int rdbSaveBackground(char *filename) { } redisLog(REDIS_NOTICE,"Background saving started by pid %d",childpid); server.bgsavechildpid = childpid; + updateDictResizePolicy(); return REDIS_OK; } return REDIS_OK; /* unreached */ @@ -3469,7 +3848,11 @@ static uint32_t rdbLoadLen(FILE *fp, int *isencoded) { } } -static robj *rdbLoadIntegerObject(FILE *fp, int enctype) { +/* Load an integer-encoded object from file 'fp', with the specified + * encoding type 'enctype'. If encode is true the function may return + * an integer-encoded object as reply, otherwise the returned object + * will always be encoded as a raw string. */ +static robj *rdbLoadIntegerObject(FILE *fp, int enctype, int encode) { unsigned char enc[4]; long long val; @@ -3488,9 +3871,12 @@ static robj *rdbLoadIntegerObject(FILE *fp, int enctype) { val = (int32_t)v; } else { val = 0; /* anti-warning */ - redisAssert(0); + redisPanic("Unknown RDB integer encoding type"); } - return createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",val)); + if (encode) + return createStringObjectFromLongLong(val); + else + return createObject(REDIS_STRING,sdsfromlonglong(val)); } static robj *rdbLoadLzfStringObject(FILE*fp) { @@ -3512,7 +3898,7 @@ err: return NULL; } -static robj *rdbLoadStringObject(FILE*fp) { +static robj *rdbGenericLoadStringObject(FILE*fp, int encode) { int isencoded; uint32_t len; sds val; @@ -3523,11 +3909,11 @@ static robj *rdbLoadStringObject(FILE*fp) { case REDIS_RDB_ENC_INT8: case REDIS_RDB_ENC_INT16: case REDIS_RDB_ENC_INT32: - return tryObjectSharing(rdbLoadIntegerObject(fp,len)); + return rdbLoadIntegerObject(fp,len,encode); case REDIS_RDB_ENC_LZF: - return tryObjectSharing(rdbLoadLzfStringObject(fp)); + return rdbLoadLzfStringObject(fp); default: - redisAssert(0); + redisPanic("Unknown RDB encoding type"); } } @@ -3537,7 +3923,15 @@ static robj *rdbLoadStringObject(FILE*fp) { sdsfree(val); return NULL; } - return tryObjectSharing(createObject(REDIS_STRING,val)); + return createObject(REDIS_STRING,val); +} + +static robj *rdbLoadStringObject(FILE *fp) { + return rdbGenericLoadStringObject(fp,0); +} + +static robj *rdbLoadEncodedStringObject(FILE *fp) { + return rdbGenericLoadStringObject(fp,1); } /* For information about double serialization check rdbSaveDoubleValue() */ @@ -3566,8 +3960,8 @@ static robj *rdbLoadObject(int type, FILE *fp) { redisLog(REDIS_DEBUG,"LOADING OBJECT %d (at %d)\n",type,ftell(fp)); if (type == REDIS_STRING) { /* Read string value */ - if ((o = rdbLoadStringObject(fp)) == NULL) return NULL; - tryObjectEncoding(o); + if ((o = rdbLoadEncodedStringObject(fp)) == NULL) return NULL; + o = tryObjectEncoding(o); } else if (type == REDIS_LIST || type == REDIS_SET) { /* Read list/set value */ uint32_t listlen; @@ -3582,8 +3976,8 @@ static robj *rdbLoadObject(int type, FILE *fp) { while(listlen--) { robj *ele; - if ((ele = rdbLoadStringObject(fp)) == NULL) return NULL; - tryObjectEncoding(ele); + if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL; + ele = tryObjectEncoding(ele); if (type == REDIS_LIST) { listAddNodeTail((list*)o->ptr,ele); } else { @@ -3603,8 +3997,8 @@ static robj *rdbLoadObject(int type, FILE *fp) { robj *ele; double *score = zmalloc(sizeof(double)); - if ((ele = rdbLoadStringObject(fp)) == NULL) return NULL; - tryObjectEncoding(ele); + if ((ele = rdbLoadEncodedStringObject(fp)) == NULL) return NULL; + ele = tryObjectEncoding(ele); if (rdbLoadDoubleValue(fp,score) == -1) return NULL; dictAdd(zs->dict,ele,score); zslInsert(zs->zsl,*score,ele); @@ -3643,26 +4037,26 @@ static robj *rdbLoadObject(int type, FILE *fp) { decrRefCount(key); decrRefCount(val); } else { - tryObjectEncoding(key); - tryObjectEncoding(val); + key = tryObjectEncoding(key); + val = tryObjectEncoding(val); dictAdd((dict*)o->ptr,key,val); } } } else { - redisAssert(0); + redisPanic("Unknown object type"); } return o; } static int rdbLoad(char *filename) { FILE *fp; - robj *keyobj = NULL; uint32_t dbid; int type, retval, rdbver; + int swap_all_values = 0; dict *d = server.db[0].dict; redisDb *db = server.db+0; char buf[1024]; - time_t expiretime = -1, now = time(NULL); + time_t expiretime, now = time(NULL); long long loadedkeys = 0; fp = fopen(filename,"r"); @@ -3681,8 +4075,9 @@ static int rdbLoad(char *filename) { return REDIS_ERR; } while(1) { - robj *o; + robj *key, *val; + expiretime = -1; /* Read type. */ if ((type = rdbLoadType(fp)) == -1) goto eoferr; if (type == REDIS_EXPIRETIME) { @@ -3704,102 +4099,177 @@ static int rdbLoad(char *filename) { continue; } /* Read key */ - if ((keyobj = rdbLoadStringObject(fp)) == NULL) goto eoferr; + if ((key = rdbLoadStringObject(fp)) == NULL) goto eoferr; /* Read value */ - if ((o = rdbLoadObject(type,fp)) == NULL) goto eoferr; + if ((val = rdbLoadObject(type,fp)) == NULL) goto eoferr; + /* Check if the key already expired */ + if (expiretime != -1 && expiretime < now) { + decrRefCount(key); + decrRefCount(val); + continue; + } /* Add the new object in the hash table */ - retval = dictAdd(d,keyobj,o); + retval = dictAdd(d,key,val); if (retval == DICT_ERR) { - redisLog(REDIS_WARNING,"Loading DB, duplicated key (%s) found! Unrecoverable error, exiting now.", keyobj->ptr); + redisLog(REDIS_WARNING,"Loading DB, duplicated key (%s) found! Unrecoverable error, exiting now.", key->ptr); exit(1); } + loadedkeys++; /* Set the expire time if needed */ - if (expiretime != -1) { - setExpire(db,keyobj,expiretime); - /* Delete this key if already expired */ - if (expiretime < now) deleteKey(db,keyobj); - expiretime = -1; - } - keyobj = o = NULL; + if (expiretime != -1) setExpire(db,key,expiretime); + /* Handle swapping while loading big datasets when VM is on */ - loadedkeys++; - if (server.vm_enabled && (loadedkeys % 5000) == 0) { + + /* If we detecter we are hopeless about fitting something in memory + * we just swap every new key on disk. Directly... + * Note that's important to check for this condition before resorting + * to random sampling, otherwise we may try to swap already + * swapped keys. */ + if (swap_all_values) { + dictEntry *de = dictFind(d,key); + + /* de may be NULL since the key already expired */ + if (de) { + key = dictGetEntryKey(de); + val = dictGetEntryVal(de); + + if (vmSwapObjectBlocking(key,val) == REDIS_OK) { + dictGetEntryVal(de) = NULL; + } + } + continue; + } + + /* If we have still some hope of having some value fitting memory + * then we try random sampling. */ + if (!swap_all_values && server.vm_enabled && (loadedkeys % 5000) == 0) { while (zmalloc_used_memory() > server.vm_max_memory) { if (vmSwapOneObjectBlocking() == REDIS_ERR) break; } + if (zmalloc_used_memory() > server.vm_max_memory) + swap_all_values = 1; /* We are already using too much mem */ } } fclose(fp); return REDIS_OK; 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, aborting now."); exit(1); return REDIS_ERR; /* Just to avoid warning */ } -/*================================== Commands =============================== */ - -static void authCommand(redisClient *c) { - if (!server.requirepass || !strcmp(c->argv[1]->ptr, server.requirepass)) { - c->authenticated = 1; - addReply(c,shared.ok); - } else { - c->authenticated = 0; - addReplySds(c,sdscatprintf(sdsempty(),"-ERR invalid password\r\n")); +/*================================== Shutdown =============================== */ +static int prepareForShutdown() { + 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.bgsavechildpid != -1) { + redisLog(REDIS_WARNING,"There is a live saving child. Killing it!"); + kill(server.bgsavechildpid,SIGKILL); + rdbRemoveTempFile(server.bgsavechildpid); + } + if (server.appendonly) { + /* Append only file: fsync() the AOF and exit */ + fsync(server.appendfd); + if (server.vm_enabled) unlink(server.vm_swap_file); + } else { + /* Snapshotting. Perform a SYNC SAVE and exit */ + if (rdbSave(server.dbfilename) == REDIS_OK) { + if (server.daemonize) + unlink(server.pidfile); + redisLog(REDIS_WARNING,"%zu bytes used at exit",zmalloc_used_memory()); + } else { + /* Ooops.. error saving! The best we can do is to continue + * operating. Note that if there was a background saving process, + * in the next cron() Redis will be notified that the background + * saving aborted, handling special stuff like slaves pending for + * synchronization... */ + redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit"); + return REDIS_ERR; + } + } + redisLog(REDIS_WARNING,"Server exit now, bye bye..."); + return REDIS_OK; +} + +/*================================== Commands =============================== */ + +static void authCommand(redisClient *c) { + if (!server.requirepass || !strcmp(c->argv[1]->ptr, server.requirepass)) { + c->authenticated = 1; + addReply(c,shared.ok); + } else { + c->authenticated = 0; + addReplySds(c,sdscatprintf(sdsempty(),"-ERR invalid password\r\n")); + } +} + +static void pingCommand(redisClient *c) { + addReply(c,shared.pong); +} + +static void echoCommand(redisClient *c) { + addReplyBulk(c,c->argv[1]); +} + +/*=================================== Strings =============================== */ + +static void setGenericCommand(redisClient *c, int nx, robj *key, robj *val, robj *expire) { + int retval; + long seconds = 0; /* initialized to avoid an harmness warning */ + + if (expire) { + if (getLongFromObjectOrReply(c, expire, &seconds, NULL) != REDIS_OK) + return; + if (seconds <= 0) { + addReplySds(c,sdsnew("-ERR invalid expire time in SETEX\r\n")); + return; + } } -} - -static void pingCommand(redisClient *c) { - addReply(c,shared.pong); -} - -static void echoCommand(redisClient *c) { - addReplyBulk(c,c->argv[1]); -} - -/*=================================== Strings =============================== */ - -static void setGenericCommand(redisClient *c, int nx) { - int retval; - if (nx) deleteIfVolatile(c->db,c->argv[1]); - retval = dictAdd(c->db->dict,c->argv[1],c->argv[2]); + if (nx) deleteIfVolatile(c->db,key); + retval = dictAdd(c->db->dict,key,val); if (retval == DICT_ERR) { if (!nx) { /* If the key is about a swapped value, we want a new key object * to overwrite the old. So we delete the old key in the database. * This will also make sure that swap pages about the old object * will be marked as free. */ - if (server.vm_enabled && deleteIfSwapped(c->db,c->argv[1])) - incrRefCount(c->argv[1]); - dictReplace(c->db->dict,c->argv[1],c->argv[2]); - incrRefCount(c->argv[2]); + if (server.vm_enabled && deleteIfSwapped(c->db,key)) + incrRefCount(key); + dictReplace(c->db->dict,key,val); + incrRefCount(val); } else { addReply(c,shared.czero); return; } } else { - incrRefCount(c->argv[1]); - incrRefCount(c->argv[2]); + incrRefCount(key); + incrRefCount(val); } server.dirty++; - removeExpire(c->db,c->argv[1]); + removeExpire(c->db,key); + if (expire) setExpire(c->db,key,time(NULL)+seconds); addReply(c, nx ? shared.cone : shared.ok); } static void setCommand(redisClient *c) { - setGenericCommand(c,0); + setGenericCommand(c,0,c->argv[1],c->argv[2],NULL); } static void setnxCommand(redisClient *c) { - setGenericCommand(c,1); + setGenericCommand(c,1,c->argv[1],c->argv[2],NULL); +} + +static void setexCommand(redisClient *c) { + setGenericCommand(c,0,c->argv[1],c->argv[3],c->argv[2]); } static int getGenericCommand(redisClient *c) { robj *o; - + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL) return REDIS_OK; @@ -3830,7 +4300,7 @@ static void getsetCommand(redisClient *c) { static void mgetCommand(redisClient *c) { int j; - + addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",c->argc-1)); for (j = 1; j < c->argc; j++) { robj *o = lookupKeyRead(c->db,c->argv[j]); @@ -3870,7 +4340,7 @@ static void msetGenericCommand(redisClient *c, int nx) { for (j = 1; j < c->argc; j += 2) { int retval; - tryObjectEncoding(c->argv[j+1]); + c->argv[j+1] = tryObjectEncoding(c->argv[j+1]); retval = dictAdd(c->db->dict,c->argv[j],c->argv[j+1]); if (retval == DICT_ERR) { dictReplace(c->db->dict,c->argv[j],c->argv[j+1]); @@ -3897,28 +4367,13 @@ static void incrDecrCommand(redisClient *c, long long incr) { long long value; int retval; robj *o; - - o = lookupKeyWrite(c->db,c->argv[1]); - if (o == NULL) { - value = 0; - } else { - if (o->type != REDIS_STRING) { - value = 0; - } else { - char *eptr; - if (o->encoding == REDIS_ENCODING_RAW) - value = strtoll(o->ptr, &eptr, 10); - else if (o->encoding == REDIS_ENCODING_INT) - value = (long)o->ptr; - else - redisAssert(1 != 1); - } - } + o = lookupKeyWrite(c->db,c->argv[1]); + if (o != NULL && checkType(c,o,REDIS_STRING)) return; + if (getLongLongFromObjectOrReply(c,o,&value,NULL) != REDIS_OK) return; value += incr; - o = createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",value)); - tryObjectEncoding(o); + o = createStringObjectFromLongLong(value); retval = dictAdd(c->db->dict,c->argv[1],o); if (retval == DICT_ERR) { dictReplace(c->db->dict,c->argv[1],o); @@ -3941,12 +4396,16 @@ static void decrCommand(redisClient *c) { } static void incrbyCommand(redisClient *c) { - long long incr = strtoll(c->argv[2]->ptr, NULL, 10); + long long incr; + + if (getLongLongFromObjectOrReply(c, c->argv[2], &incr, NULL) != REDIS_OK) return; incrDecrCommand(c,incr); } static void decrbyCommand(redisClient *c) { - long long incr = strtoll(c->argv[2]->ptr, NULL, 10); + long long incr; + + if (getLongLongFromObjectOrReply(c, c->argv[2], &incr, NULL) != REDIS_OK) return; incrDecrCommand(c,-incr); } @@ -3964,7 +4423,7 @@ static void appendCommand(redisClient *c) { totlen = stringObjectLen(c->argv[2]); } else { dictEntry *de; - + de = dictFind(c->db->dict,c->argv[1]); assert(de != NULL); @@ -4044,16 +4503,21 @@ static void delCommand(redisClient *c) { deleted++; } } - addReplyLong(c,deleted); + addReplyLongLong(c,deleted); } static void existsCommand(redisClient *c) { - addReply(c,lookupKeyRead(c->db,c->argv[1]) ? shared.cone : shared.czero); + expireIfNeeded(c->db,c->argv[1]); + if (dictFind(c->db->dict,c->argv[1])) { + addReply(c, shared.cone); + } else { + addReply(c, shared.czero); + } } static void selectCommand(redisClient *c) { int id = atoi(c->argv[1]->ptr); - + if (selectDb(c,id) == REDIS_ERR) { addReplySds(c,sdsnew("-ERR invalid DB index\r\n")); } else { @@ -4063,18 +4527,25 @@ static void selectCommand(redisClient *c) { static void randomkeyCommand(redisClient *c) { dictEntry *de; - + robj *key; + while(1) { de = dictGetRandomKey(c->db->dict); if (!de || expireIfNeeded(c->db,dictGetEntryKey(de)) == 0) break; } + if (de == NULL) { - addReply(c,shared.plus); - addReply(c,shared.crlf); + addReply(c,shared.nullbulk); + return; + } + + key = dictGetEntryKey(de); + if (server.vm_enabled) { + key = dupStringObject(key); + addReplyBulk(c,key); + decrRefCount(key); } else { - addReply(c,shared.plus); - addReply(c,dictGetEntryKey(de)); - addReply(c,shared.crlf); + addReplyBulk(c,key); } } @@ -4162,40 +4633,9 @@ 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.bgsavechildpid != -1) { - redisLog(REDIS_WARNING,"There is a live saving child. Killing it!"); - kill(server.bgsavechildpid,SIGKILL); - rdbRemoveTempFile(server.bgsavechildpid); - } - if (server.appendonly) { - /* Append only file: fsync() the AOF and exit */ - fsync(server.appendfd); - if (server.vm_enabled) unlink(server.vm_swap_file); + if (prepareForShutdown() == REDIS_OK) exit(0); - } else { - /* Snapshotting. Perform a SYNC SAVE and exit */ - if (rdbSave(server.dbfilename) == REDIS_OK) { - if (server.daemonize) - unlink(server.pidfile); - redisLog(REDIS_WARNING,"%zu bytes used at exit",zmalloc_used_memory()); - redisLog(REDIS_WARNING,"Server exit now, bye bye..."); - if (server.vm_enabled) unlink(server.vm_swap_file); - exit(0); - } else { - /* Ooops.. error saving! The best we can do is to continue - * operating. Note that if there was a background saving process, - * in the next cron() Redis will be notified that the background - * saving aborted, handling special stuff like slaves pending for - * synchronization... */ - redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit"); - addReplySds(c, - sdsnew("-ERR can't quit, problems saving the DB\r\n")); - } - } + addReplySds(c, sdsnew("-ERR Errors trying to SHUTDOWN. Check logs.\r\n")); } static void renameGenericCommand(redisClient *c, int nx) { @@ -4318,7 +4758,7 @@ static void pushGenericCommand(redisClient *c, int where) { incrRefCount(c->argv[2]); } server.dirty++; - addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",listLength(list))); + addReplyLongLong(c,listLength(list)); } static void lpushCommand(redisClient *c) { @@ -4335,7 +4775,7 @@ static void llenCommand(redisClient *c) { if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,o,REDIS_LIST)) return; - + l = o->ptr; addReplyUlong(c,listLength(l)); } @@ -4426,8 +4866,8 @@ static void lrangeCommand(redisClient *c) { listNode *ln; robj *ele; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullmultibulk)) == NULL || - checkType(c,o,REDIS_LIST)) return; + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptymultibulk)) == NULL + || checkType(c,o,REDIS_LIST)) return; list = o->ptr; llen = listLength(list); @@ -4522,7 +4962,7 @@ static void lremCommand(redisClient *c) { robj *ele = listNodeValue(ln); next = fromtail ? ln->prev : ln->next; - if (compareStringObjects(ele,c->argv[3]) == 0) { + if (equalStringObjects(ele,c->argv[3])) { listDelNode(list,ln); server.dirty++; removed++; @@ -4691,7 +5131,7 @@ static void scardCommand(redisClient *c) { if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,o,REDIS_SET)) return; - + s = o->ptr; addReplyUlong(c,dictSize(s)); } @@ -4760,7 +5200,7 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, unsigned long server.dirty++; addReply(c,shared.czero); } else { - addReply(c,shared.nullmultibulk); + addReply(c,shared.emptymultibulk); } return; } @@ -4820,7 +5260,7 @@ static void sinterGenericCommand(redisClient *c, robj **setskeys, unsigned long if (dictSize((dict*)dstset->ptr) > 0) { dictAdd(c->db->dict,dstkey,dstset); incrRefCount(dstkey); - addReplyLong(c,dictSize((dict*)dstset->ptr)); + addReplyLongLong(c,dictSize((dict*)dstset->ptr)); } else { decrRefCount(dstset); addReply(c,shared.czero); @@ -4923,7 +5363,7 @@ static void sunionDiffGenericCommand(redisClient *c, robj **setskeys, int setsnu if (dictSize((dict*)dstset->ptr) > 0) { dictAdd(c->db->dict,dstkey,dstset); incrRefCount(dstkey); - addReplyLong(c,dictSize((dict*)dstset->ptr)); + addReplyLongLong(c,dictSize((dict*)dstset->ptr)); } else { decrRefCount(dstset); addReply(c,shared.czero); @@ -4972,8 +5412,10 @@ static zskiplistNode *zslCreateNode(int level, double score, robj *obj) { zskiplistNode *zn = zmalloc(sizeof(*zn)); zn->forward = zmalloc(sizeof(zskiplistNode*) * level); - if (level > 0) + if (level > 1) zn->span = zmalloc(sizeof(unsigned int) * (level - 1)); + else + zn->span = NULL; zn->score = score; zn->obj = obj; return zn; @@ -4982,7 +5424,7 @@ static zskiplistNode *zslCreateNode(int level, double score, robj *obj) { static zskiplist *zslCreate(void) { int j; zskiplist *zsl; - + zsl = zmalloc(sizeof(*zsl)); zsl->level = 1; zsl->length = 0; @@ -5024,7 +5466,7 @@ static int zslRandomLevel(void) { int level = 1; while ((random()&0xFFFF) < (ZSKIPLIST_P * 0xFFFF)) level += 1; - return level; + return (levelforward[0]; - if (x && score == x->score && compareStringObjects(x->obj,obj) == 0) { + if (x && score == x->score && equalStringObjects(x->obj,obj)) { zslDeleteNode(zsl, x, update); zslFreeNode(x); return 1; @@ -5231,7 +5673,7 @@ static unsigned long zslGetRank(zskiplist *zsl, double score, robj *o) { } /* x might be equal to zsl->header, so test if obj is non-NULL */ - if (x->obj && compareStringObjects(x->obj,o) == 0) { + if (x->obj && equalStringObjects(x->obj,o)) { return rank; } } @@ -5315,7 +5757,7 @@ static void zaddGenericCommand(redisClient *c, robj *key, robj *ele, double scor } else { dictEntry *de; double *oldscore; - + /* case 2: Score update operation */ de = dictFind(zs->dict,ele); redisAssert(de != NULL); @@ -5344,14 +5786,14 @@ static void zaddGenericCommand(redisClient *c, robj *key, robj *ele, double scor static void zaddCommand(redisClient *c) { double scoreval; - scoreval = strtod(c->argv[2]->ptr,NULL); + if (getDoubleFromObjectOrReply(c, c->argv[2], &scoreval, NULL) != REDIS_OK) return; zaddGenericCommand(c,c->argv[1],c->argv[3],scoreval,0); } static void zincrbyCommand(redisClient *c) { double scoreval; - scoreval = strtod(c->argv[2]->ptr,NULL); + if (getDoubleFromObjectOrReply(c, c->argv[2], &scoreval, NULL) != REDIS_OK) return; zaddGenericCommand(c,c->argv[1],c->argv[3],scoreval,1); } @@ -5385,12 +5827,15 @@ 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); + double min; + double max; long deleted; robj *zsetobj; zset *zs; + if ((getDoubleFromObjectOrReply(c, c->argv[2], &min, NULL) != REDIS_OK) || + (getDoubleFromObjectOrReply(c, c->argv[3], &max, NULL) != REDIS_OK)) return; + if ((zsetobj = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,zsetobj,REDIS_ZSET)) return; @@ -5399,17 +5844,20 @@ static void zremrangebyscoreCommand(redisClient *c) { if (htNeedsResize(zs->dict)) dictResize(zs->dict); if (dictSize(zs->dict) == 0) deleteKey(c->db,c->argv[1]); server.dirty += deleted; - addReplyLong(c,deleted); + addReplyLongLong(c,deleted); } static void zremrangebyrankCommand(redisClient *c) { - int start = atoi(c->argv[2]->ptr); - int end = atoi(c->argv[3]->ptr); + long start; + long end; int llen; long deleted; robj *zsetobj; zset *zs; + if ((getLongFromObjectOrReply(c, c->argv[2], &start, NULL) != REDIS_OK) || + (getLongFromObjectOrReply(c, c->argv[3], &end, NULL) != REDIS_OK)) return; + if ((zsetobj = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,zsetobj,REDIS_ZSET)) return; zs = zsetobj->ptr; @@ -5434,7 +5882,7 @@ static void zremrangebyrankCommand(redisClient *c) { if (htNeedsResize(zs->dict)) dictResize(zs->dict); if (dictSize(zs->dict) == 0) deleteKey(c->db,c->argv[1]); server.dirty += deleted; - addReplyLong(c, deleted); + addReplyLongLong(c, deleted); } typedef struct { @@ -5463,7 +5911,7 @@ inline static void zunionInterAggregate(double *target, double val, int aggregat *target = val > *target ? val : *target; } else { /* safety net */ - redisAssert(0 != 0); + redisPanic("Unknown ZUNION/INTER aggregate type"); } } @@ -5479,7 +5927,7 @@ static void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) { /* expect zsetnum input keys to be given */ zsetnum = atoi(c->argv[2]->ptr); if (zsetnum < 1) { - addReplySds(c,sdsnew("-ERR at least 1 input key is needed for ZUNION/ZINTER\r\n")); + addReplySds(c,sdsnew("-ERR at least 1 input key is needed for ZUNIONSTORE/ZINTERSTORE\r\n")); return; } @@ -5516,7 +5964,8 @@ static void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) { if (remaining >= (zsetnum + 1) && !strcasecmp(c->argv[j]->ptr,"weights")) { j++; remaining--; for (i = 0; i < zsetnum; i++, j++, remaining--) { - src[i].weight = strtod(c->argv[j]->ptr, NULL); + if (getDoubleFromObjectOrReply(c, c->argv[j], &src[i].weight, NULL) != REDIS_OK) + return; } } else if (remaining >= 2 && !strcasecmp(c->argv[j]->ptr,"aggregate")) { j++; remaining--; @@ -5619,27 +6068,27 @@ static void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) { if (dstzset->zsl->length) { dictAdd(c->db->dict,dstkey,dstobj); incrRefCount(dstkey); - addReplyLong(c, dstzset->zsl->length); + addReplyLongLong(c, dstzset->zsl->length); server.dirty++; } else { - decrRefCount(dstzset); + decrRefCount(dstobj); addReply(c, shared.czero); } zfree(src); } -static void zunionCommand(redisClient *c) { +static void zunionstoreCommand(redisClient *c) { zunionInterGenericCommand(c,c->argv[1], REDIS_OP_UNION); } -static void zinterCommand(redisClient *c) { +static void zinterstoreCommand(redisClient *c) { zunionInterGenericCommand(c,c->argv[1], REDIS_OP_INTER); } static void zrangeGenericCommand(redisClient *c, int reverse) { robj *o; - int start = atoi(c->argv[2]->ptr); - int end = atoi(c->argv[3]->ptr); + long start; + long end; int withscores = 0; int llen; int rangelen, j; @@ -5648,6 +6097,9 @@ static void zrangeGenericCommand(redisClient *c, int reverse) { zskiplistNode *ln; robj *ele; + if ((getLongFromObjectOrReply(c, c->argv[2], &start, NULL) != REDIS_OK) || + (getLongFromObjectOrReply(c, c->argv[3], &end, NULL) != REDIS_OK)) return; + if (c->argc == 5 && !strcasecmp(c->argv[4]->ptr,"withscores")) { withscores = 1; } else if (c->argc >= 5) { @@ -5655,8 +6107,8 @@ static void zrangeGenericCommand(redisClient *c, int reverse) { return; } - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullmultibulk)) == NULL || - checkType(c,o,REDIS_ZSET)) return; + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptymultibulk)) == NULL + || checkType(c,o,REDIS_ZSET)) return; zsetobj = o->ptr; zsl = zsetobj->zsl; llen = zsl->length; @@ -5762,7 +6214,7 @@ static void genericZrangebyscoreCommand(redisClient *c, int justcount) { /* Ok, lookup the key and get the range */ o = lookupKeyRead(c->db,c->argv[1]); if (o == NULL) { - addReply(c,justcount ? shared.czero : shared.nullmultibulk); + addReply(c,justcount ? shared.czero : shared.emptymultibulk); } else { if (o->type != REDIS_ZSET) { addReply(c,shared.wrongtypeerr); @@ -5812,7 +6264,7 @@ static void genericZrangebyscoreCommand(redisClient *c, int justcount) { if (limit > 0) limit--; } if (justcount) { - addReplyLong(c,(long)rangelen); + addReplyLongLong(c,(long)rangelen); } else { lenobj->ptr = sdscatprintf(sdsempty(),"*%lu\r\n", withscores ? (rangelen*2) : rangelen); @@ -5882,9 +6334,9 @@ static void zrankGenericCommand(redisClient *c, int reverse) { rank = zslGetRank(zsl, *score, c->argv[2]); if (rank) { if (reverse) { - addReplyLong(c, zsl->length - rank); + addReplyLongLong(c, zsl->length - rank); } else { - addReplyLong(c, rank-1); + addReplyLongLong(c, rank-1); } } else { addReply(c,shared.nullbulk); @@ -5899,299 +6351,389 @@ static void zrevrankCommand(redisClient *c) { zrankGenericCommand(c, 1); } -/* =================================== Hashes =============================== */ -static void hsetCommand(redisClient *c) { - int update = 0; - robj *o = lookupKeyWrite(c->db,c->argv[1]); +/* ========================= Hashes utility functions ======================= */ +#define REDIS_HASH_KEY 1 +#define REDIS_HASH_VALUE 2 - if (o == NULL) { - o = createHashObject(); - dictAdd(c->db->dict,c->argv[1],o); - incrRefCount(c->argv[1]); - } else { - if (o->type != REDIS_HASH) { - addReply(c,shared.wrongtypeerr); +/* Check the length of a number of objects to see if we need to convert a + * zipmap to a real hash. Note that we only check string encoded objects + * as their string length can be queried in constant time. */ +static void hashTryConversion(robj *subject, robj **argv, int start, int end) { + int i; + if (subject->encoding != REDIS_ENCODING_ZIPMAP) return; + + for (i = start; i <= end; i++) { + if (argv[i]->encoding == REDIS_ENCODING_RAW && + sdslen(argv[i]->ptr) > server.hash_max_zipmap_value) + { + convertToRealHash(subject); return; } } - /* We want to convert the zipmap into an hash table right now if the - * entry to be added is too big. Note that we check if the object - * is integer encoded before to try fetching the length in the test below. - * This is because integers are small, but currently stringObjectLen() - * performs a slow conversion: not worth it. */ - if (o->encoding == REDIS_ENCODING_ZIPMAP && - ((c->argv[2]->encoding == REDIS_ENCODING_RAW && - sdslen(c->argv[2]->ptr) > server.hash_max_zipmap_value) || - (c->argv[3]->encoding == REDIS_ENCODING_RAW && - sdslen(c->argv[3]->ptr) > server.hash_max_zipmap_value))) - { - convertToRealHash(o); +} + +/* Encode given objects in-place when the hash uses a dict. */ +static void hashTryObjectEncoding(robj *subject, robj **o1, robj **o2) { + if (subject->encoding == REDIS_ENCODING_HT) { + if (o1) *o1 = tryObjectEncoding(*o1); + if (o2) *o2 = tryObjectEncoding(*o2); } +} +/* Get the value from a hash identified by key. Returns either a string + * object or NULL if the value cannot be found. The refcount of the object + * is always increased by 1 when the value was found. */ +static robj *hashGet(robj *o, robj *key) { + robj *value = NULL; if (o->encoding == REDIS_ENCODING_ZIPMAP) { - unsigned char *zm = o->ptr; - robj *valobj = getDecodedObject(c->argv[3]); - - zm = zipmapSet(zm,c->argv[2]->ptr,sdslen(c->argv[2]->ptr), - valobj->ptr,sdslen(valobj->ptr),&update); - decrRefCount(valobj); - o->ptr = zm; - - /* And here there is the second check for hash conversion... - * we want to do it only if the operation was not just an update as - * zipmapLen() is O(N). */ - if (!update && zipmapLen(zm) > server.hash_max_zipmap_entries) - convertToRealHash(o); + unsigned char *v; + unsigned int vlen; + key = getDecodedObject(key); + if (zipmapGet(o->ptr,key->ptr,sdslen(key->ptr),&v,&vlen)) { + value = createStringObject((char*)v,vlen); + } + decrRefCount(key); } else { - tryObjectEncoding(c->argv[2]); - /* note that c->argv[3] is already encoded, as the latest arg - * of a bulk command is always integer encoded if possible. */ - if (dictReplace(o->ptr,c->argv[2],c->argv[3])) { - incrRefCount(c->argv[2]); - } else { - update = 1; + dictEntry *de = dictFind(o->ptr,key); + if (de != NULL) { + value = dictGetEntryVal(de); + incrRefCount(value); } - incrRefCount(c->argv[3]); } - server.dirty++; - addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",update == 0)); + return value; } -static void hincrbyCommand(redisClient *c) { - int update = 0; - long long value = 0, incr = 0; - robj *o = lookupKeyWrite(c->db,c->argv[1]); - - if (o == NULL) { - o = createHashObject(); - dictAdd(c->db->dict,c->argv[1],o); - incrRefCount(c->argv[1]); +/* Test if the key exists in the given hash. Returns 1 if the key + * exists and 0 when it doesn't. */ +static int hashExists(robj *o, robj *key) { + if (o->encoding == REDIS_ENCODING_ZIPMAP) { + key = getDecodedObject(key); + if (zipmapExists(o->ptr,key->ptr,sdslen(key->ptr))) { + decrRefCount(key); + return 1; + } + decrRefCount(key); } else { - if (o->type != REDIS_HASH) { - addReply(c,shared.wrongtypeerr); - return; + if (dictFind(o->ptr,key) != NULL) { + return 1; } } + return 0; +} - robj *o_incr = getDecodedObject(c->argv[3]); - incr = strtoll(o_incr->ptr, NULL, 10); - decrRefCount(o_incr); - +/* Add an element, discard the old if the key already exists. + * Return 0 on insert and 1 on update. */ +static int hashSet(robj *o, robj *key, robj *value) { + int update = 0; if (o->encoding == REDIS_ENCODING_ZIPMAP) { - unsigned char *zm = o->ptr; - unsigned char *zval; - unsigned int zvlen; - - /* Find value if already present in hash */ - if (zipmapGet(zm,c->argv[2]->ptr,sdslen(c->argv[2]->ptr), - &zval,&zvlen)) { - /* strtoll needs the char* to have a trailing \0, but - * the zipmap doesn't include them. */ - sds szval = sdsnewlen(zval, zvlen); - value = strtoll(szval,NULL,10); - sdsfree(szval); - } - - value += incr; - sds svalue = sdscatprintf(sdsempty(),"%lld",value); - zm = zipmapSet(zm,c->argv[2]->ptr,sdslen(c->argv[2]->ptr), - (unsigned char*)svalue,sdslen(svalue),&update); - sdsfree(svalue); - o->ptr = zm; - - /* Check if the zipmap needs to be converted - * if this was not an update. */ - if (!update && zipmapLen(zm) > server.hash_max_zipmap_entries) + key = getDecodedObject(key); + value = getDecodedObject(value); + o->ptr = zipmapSet(o->ptr, + key->ptr,sdslen(key->ptr), + value->ptr,sdslen(value->ptr), &update); + decrRefCount(key); + decrRefCount(value); + + /* Check if the zipmap needs to be upgraded to a real hash table */ + if (zipmapLen(o->ptr) > server.hash_max_zipmap_entries) convertToRealHash(o); } else { - robj *hval; - dictEntry *de; - - /* Find value if already present in hash */ - de = dictFind(o->ptr,c->argv[2]); - if (de != NULL) { - hval = dictGetEntryVal(de); - if (hval->encoding == REDIS_ENCODING_RAW) - value = strtoll(hval->ptr,NULL,10); - else if (hval->encoding == REDIS_ENCODING_INT) - value = (long)hval->ptr; - else - redisAssert(1 != 1); + if (dictReplace(o->ptr,key,value)) { + /* Insert */ + incrRefCount(key); + } else { + /* Update */ + update = 1; } + incrRefCount(value); + } + return update; +} - value += incr; - hval = createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",value)); - tryObjectEncoding(hval); - if (dictReplace(o->ptr,c->argv[2],hval)) { - incrRefCount(c->argv[2]); - } +/* Delete an element from a hash. + * Return 1 on deleted and 0 on not found. */ +static int hashDelete(robj *o, robj *key) { + int deleted = 0; + if (o->encoding == REDIS_ENCODING_ZIPMAP) { + key = getDecodedObject(key); + o->ptr = zipmapDel(o->ptr,key->ptr,sdslen(key->ptr), &deleted); + decrRefCount(key); + } else { + deleted = dictDelete((dict*)o->ptr,key) == DICT_OK; + /* Always check if the dictionary needs a resize after a delete. */ + if (deleted && htNeedsResize(o->ptr)) dictResize(o->ptr); } + return deleted; +} - server.dirty++; - addReplyLong(c, value); +/* Return the number of elements in a hash. */ +static unsigned long hashLength(robj *o) { + return (o->encoding == REDIS_ENCODING_ZIPMAP) ? + zipmapLen((unsigned char*)o->ptr) : dictSize((dict*)o->ptr); } -static void hgetCommand(redisClient *c) { - robj *o; +/* Structure to hold hash iteration abstration. Note that iteration over + * hashes involves both fields and values. Because it is possible that + * not both are required, store pointers in the iterator to avoid + * unnecessary memory allocation for fields/values. */ +typedef struct { + int encoding; + unsigned char *zi; + unsigned char *zk, *zv; + unsigned int zklen, zvlen; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL || - checkType(c,o,REDIS_HASH)) return; + dictIterator *di; + dictEntry *de; +} hashIterator; + +static hashIterator *hashInitIterator(robj *subject) { + hashIterator *hi = zmalloc(sizeof(hashIterator)); + hi->encoding = subject->encoding; + if (hi->encoding == REDIS_ENCODING_ZIPMAP) { + hi->zi = zipmapRewind(subject->ptr); + } else if (hi->encoding == REDIS_ENCODING_HT) { + hi->di = dictGetIterator(subject->ptr); + } else { + redisAssert(NULL); + } + return hi; +} - if (o->encoding == REDIS_ENCODING_ZIPMAP) { - unsigned char *zm = o->ptr; - unsigned char *val; - unsigned int vlen; - robj *field; - - field = getDecodedObject(c->argv[2]); - if (zipmapGet(zm,field->ptr,sdslen(field->ptr), &val,&vlen)) { - addReplySds(c,sdscatprintf(sdsempty(),"$%u\r\n", vlen)); - addReplySds(c,sdsnewlen(val,vlen)); - addReply(c,shared.crlf); - decrRefCount(field); - return; +static void hashReleaseIterator(hashIterator *hi) { + if (hi->encoding == REDIS_ENCODING_HT) { + dictReleaseIterator(hi->di); + } + zfree(hi); +} + +/* Move to the next entry in the hash. Return REDIS_OK when the next entry + * could be found and REDIS_ERR when the iterator reaches the end. */ +static int hashNext(hashIterator *hi) { + if (hi->encoding == REDIS_ENCODING_ZIPMAP) { + if ((hi->zi = zipmapNext(hi->zi, &hi->zk, &hi->zklen, + &hi->zv, &hi->zvlen)) == NULL) return REDIS_ERR; + } else { + if ((hi->de = dictNext(hi->di)) == NULL) return REDIS_ERR; + } + return REDIS_OK; +} + +/* Get key or value object at current iteration position. + * This increases the refcount of the field object by 1. */ +static robj *hashCurrent(hashIterator *hi, int what) { + robj *o; + if (hi->encoding == REDIS_ENCODING_ZIPMAP) { + if (what & REDIS_HASH_KEY) { + o = createStringObject((char*)hi->zk,hi->zklen); } else { - addReply(c,shared.nullbulk); - decrRefCount(field); - return; + o = createStringObject((char*)hi->zv,hi->zvlen); } } else { - struct dictEntry *de; - - de = dictFind(o->ptr,c->argv[2]); - if (de == NULL) { - addReply(c,shared.nullbulk); + if (what & REDIS_HASH_KEY) { + o = dictGetEntryKey(hi->de); } else { - robj *e = dictGetEntryVal(de); + o = dictGetEntryVal(hi->de); + } + incrRefCount(o); + } + return o; +} - addReplyBulk(c,e); +static robj *hashLookupWriteOrCreate(redisClient *c, robj *key) { + robj *o = lookupKeyWrite(c->db,key); + if (o == NULL) { + o = createHashObject(); + dictAdd(c->db->dict,key,o); + incrRefCount(key); + } else { + if (o->type != REDIS_HASH) { + addReply(c,shared.wrongtypeerr); + return NULL; } } + return o; } -static void hdelCommand(redisClient *c) { +/* ============================= Hash commands ============================== */ +static void hsetCommand(redisClient *c) { + int update; robj *o; - int deleted = 0; - if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || - checkType(c,o,REDIS_HASH)) return; + if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return; + hashTryConversion(o,c->argv,2,3); + hashTryObjectEncoding(o,&c->argv[2], &c->argv[3]); + update = hashSet(o,c->argv[2],c->argv[3]); + addReply(c, update ? shared.czero : shared.cone); + server.dirty++; +} - if (o->encoding == REDIS_ENCODING_ZIPMAP) { - robj *field = getDecodedObject(c->argv[2]); - - o->ptr = zipmapDel((unsigned char*) o->ptr, - (unsigned char*) field->ptr, - sdslen(field->ptr), &deleted); - decrRefCount(field); - if (zipmapLen((unsigned char*) o->ptr) == 0) - deleteKey(c->db,c->argv[1]); +static void hsetnxCommand(redisClient *c) { + robj *o; + if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return; + hashTryConversion(o,c->argv,2,3); + + if (hashExists(o, c->argv[2])) { + addReply(c, shared.czero); } else { - deleted = dictDelete((dict*)o->ptr,c->argv[2]) == DICT_OK; - if (htNeedsResize(o->ptr)) dictResize(o->ptr); - if (dictSize((dict*)o->ptr) == 0) deleteKey(c->db,c->argv[1]); + hashTryObjectEncoding(o,&c->argv[2], &c->argv[3]); + hashSet(o,c->argv[2],c->argv[3]); + addReply(c, shared.cone); + server.dirty++; } - if (deleted) server.dirty++; - addReply(c,deleted ? shared.cone : shared.czero); } -static void hlenCommand(redisClient *c) { +static void hmsetCommand(redisClient *c) { + int i; robj *o; - unsigned long len; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || - checkType(c,o,REDIS_HASH)) return; + if ((c->argc % 2) == 1) { + addReplySds(c,sdsnew("-ERR wrong number of arguments for HMSET\r\n")); + return; + } - len = (o->encoding == REDIS_ENCODING_ZIPMAP) ? - zipmapLen((unsigned char*)o->ptr) : dictSize((dict*)o->ptr); - addReplyUlong(c,len); + if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return; + hashTryConversion(o,c->argv,2,c->argc-1); + for (i = 2; i < c->argc; i += 2) { + hashTryObjectEncoding(o,&c->argv[i], &c->argv[i+1]); + hashSet(o,c->argv[i],c->argv[i+1]); + } + addReply(c, shared.ok); + server.dirty++; } -#define REDIS_GETALL_KEYS 1 -#define REDIS_GETALL_VALS 2 -static void genericHgetallCommand(redisClient *c, int flags) { - robj *o, *lenobj; - unsigned long count = 0; - - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullmultibulk)) == NULL - || checkType(c,o,REDIS_HASH)) return; - - lenobj = createObject(REDIS_STRING,NULL); - addReply(c,lenobj); - decrRefCount(lenobj); +static void hincrbyCommand(redisClient *c) { + long long value, incr; + robj *o, *current, *new; + + if (getLongLongFromObjectOrReply(c,c->argv[3],&incr,NULL) != REDIS_OK) return; + if ((o = hashLookupWriteOrCreate(c,c->argv[1])) == NULL) return; + if ((current = hashGet(o,c->argv[2])) != NULL) { + if (getLongLongFromObjectOrReply(c,current,&value, + "hash value is not an integer") != REDIS_OK) { + decrRefCount(current); + return; + } + decrRefCount(current); + } else { + value = 0; + } - if (o->encoding == REDIS_ENCODING_ZIPMAP) { - unsigned char *p = zipmapRewind(o->ptr); - unsigned char *field, *val; - unsigned int flen, vlen; - - while((p = zipmapNext(p,&field,&flen,&val,&vlen)) != NULL) { - robj *aux; - - if (flags & REDIS_GETALL_KEYS) { - aux = createStringObject((char*)field,flen); - addReplyBulk(c,aux); - decrRefCount(aux); - count++; - } - if (flags & REDIS_GETALL_VALS) { - aux = createStringObject((char*)val,vlen); - addReplyBulk(c,aux); - decrRefCount(aux); - count++; - } + value += incr; + new = createStringObjectFromLongLong(value); + hashTryObjectEncoding(o,&c->argv[2],NULL); + hashSet(o,c->argv[2],new); + decrRefCount(new); + addReplyLongLong(c,value); + server.dirty++; +} + +static void hgetCommand(redisClient *c) { + robj *o, *value; + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL || + checkType(c,o,REDIS_HASH)) return; + + if ((value = hashGet(o,c->argv[2])) != NULL) { + addReplyBulk(c,value); + decrRefCount(value); + } else { + addReply(c,shared.nullbulk); + } +} + +static void hmgetCommand(redisClient *c) { + int i; + robj *o, *value; + o = lookupKeyRead(c->db,c->argv[1]); + if (o != NULL && o->type != REDIS_HASH) { + addReply(c,shared.wrongtypeerr); + } + + /* Note the check for o != NULL happens inside the loop. This is + * done because objects that cannot be found are considered to be + * an empty hash. The reply should then be a series of NULLs. */ + addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",c->argc-2)); + for (i = 2; i < c->argc; i++) { + if (o != NULL && (value = hashGet(o,c->argv[i])) != NULL) { + addReplyBulk(c,value); + decrRefCount(value); + } else { + addReply(c,shared.nullbulk); } + } +} + +static void hdelCommand(redisClient *c) { + robj *o; + if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || + checkType(c,o,REDIS_HASH)) return; + + if (hashDelete(o,c->argv[2])) { + if (hashLength(o) == 0) deleteKey(c->db,c->argv[1]); + addReply(c,shared.cone); + server.dirty++; } else { - dictIterator *di = dictGetIterator(o->ptr); - dictEntry *de; + addReply(c,shared.czero); + } +} - while((de = dictNext(di)) != NULL) { - robj *fieldobj = dictGetEntryKey(de); - robj *valobj = dictGetEntryVal(de); +static void hlenCommand(redisClient *c) { + robj *o; + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || + checkType(c,o,REDIS_HASH)) return; - if (flags & REDIS_GETALL_KEYS) { - addReplyBulk(c,fieldobj); - count++; - } - if (flags & REDIS_GETALL_VALS) { - addReplyBulk(c,valobj); - count++; - } + addReplyUlong(c,hashLength(o)); +} + +static void genericHgetallCommand(redisClient *c, int flags) { + robj *o, *lenobj, *obj; + unsigned long count = 0; + hashIterator *hi; + + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptymultibulk)) == NULL + || checkType(c,o,REDIS_HASH)) return; + + lenobj = createObject(REDIS_STRING,NULL); + addReply(c,lenobj); + decrRefCount(lenobj); + + hi = hashInitIterator(o); + while (hashNext(hi) != REDIS_ERR) { + if (flags & REDIS_HASH_KEY) { + obj = hashCurrent(hi,REDIS_HASH_KEY); + addReplyBulk(c,obj); + decrRefCount(obj); + count++; + } + if (flags & REDIS_HASH_VALUE) { + obj = hashCurrent(hi,REDIS_HASH_VALUE); + addReplyBulk(c,obj); + decrRefCount(obj); + count++; } - dictReleaseIterator(di); } + hashReleaseIterator(hi); + lenobj->ptr = sdscatprintf(sdsempty(),"*%lu\r\n",count); } static void hkeysCommand(redisClient *c) { - genericHgetallCommand(c,REDIS_GETALL_KEYS); + genericHgetallCommand(c,REDIS_HASH_KEY); } static void hvalsCommand(redisClient *c) { - genericHgetallCommand(c,REDIS_GETALL_VALS); + genericHgetallCommand(c,REDIS_HASH_VALUE); } static void hgetallCommand(redisClient *c) { - genericHgetallCommand(c,REDIS_GETALL_KEYS|REDIS_GETALL_VALS); + genericHgetallCommand(c,REDIS_HASH_KEY|REDIS_HASH_VALUE); } static void hexistsCommand(redisClient *c) { robj *o; - int exists = 0; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,o,REDIS_HASH)) return; - if (o->encoding == REDIS_ENCODING_ZIPMAP) { - robj *field; - unsigned char *zm = o->ptr; - - field = getDecodedObject(c->argv[2]); - exists = zipmapExists(zm,field->ptr,sdslen(field->ptr)); - decrRefCount(field); - } else { - exists = dictFind(o->ptr,c->argv[2]) != NULL; - } - addReply(c,exists ? shared.cone : shared.czero); + addReply(c, hashExists(o,c->argv[2]) ? shared.cone : shared.czero); } static void convertToRealHash(robj *o) { @@ -6206,8 +6748,8 @@ static void convertToRealHash(robj *o) { keyobj = createStringObject((char*)key,klen); valobj = createStringObject((char*)val,vlen); - tryObjectEncoding(keyobj); - tryObjectEncoding(valobj); + keyobj = tryObjectEncoding(keyobj); + valobj = tryObjectEncoding(valobj); dictAdd(dict,keyobj,valobj); } o->encoding = REDIS_ENCODING_HT; @@ -6243,23 +6785,26 @@ static redisSortOperation *createSortOperation(int type, robj *pattern) { } /* Return the value associated to the key with a name obtained - * substituting the first occurence of '*' in 'pattern' with 'subst' */ + * substituting the first occurence of '*' in 'pattern' with 'subst'. + * The returned object will always have its refcount increased by 1 + * when it is non-NULL. */ static robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { - char *p; + char *p, *f; sds spat, ssub; - robj keyobj; - int prefixlen, sublen, postfixlen; + robj keyobj, fieldobj, *o; + int prefixlen, sublen, postfixlen, fieldlen; /* Expoit the internal sds representation to create a sds string allocated on the stack in order to make this function faster */ struct { long len; long free; char buf[REDIS_SORTKEY_MAX+1]; - } keyname; + } keyname, fieldname; /* If the pattern is "#" return the substitution object itself in order * to implement the "SORT ... GET #" feature. */ spat = pattern->ptr; if (spat[0] == '#' && spat[1] == '\0') { + incrRefCount(subst); return subst; } @@ -6276,20 +6821,47 @@ static robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) { return NULL; } + /* Find out if we're dealing with a hash dereference. */ + if ((f = strstr(p+1, "->")) != NULL) { + fieldlen = sdslen(spat)-(f-spat); + /* this also copies \0 character */ + memcpy(fieldname.buf,f+2,fieldlen-1); + fieldname.len = fieldlen-2; + } else { + fieldlen = 0; + } + prefixlen = p-spat; sublen = sdslen(ssub); - postfixlen = sdslen(spat)-(prefixlen+1); + postfixlen = sdslen(spat)-(prefixlen+1)-fieldlen; memcpy(keyname.buf,spat,prefixlen); memcpy(keyname.buf+prefixlen,ssub,sublen); memcpy(keyname.buf+prefixlen+sublen,p+1,postfixlen); keyname.buf[prefixlen+sublen+postfixlen] = '\0'; keyname.len = prefixlen+sublen+postfixlen; - - initStaticStringObject(keyobj,((char*)&keyname)+(sizeof(long)*2)) decrRefCount(subst); - /* printf("lookup '%s' => %p\n", keyname.buf,de); */ - return lookupKeyRead(db,&keyobj); + /* Lookup substituted key */ + initStaticStringObject(keyobj,((char*)&keyname)+(sizeof(long)*2)); + o = lookupKeyRead(db,&keyobj); + if (o == NULL) return NULL; + + if (fieldlen > 0) { + if (o->type != REDIS_HASH || fieldname.len < 1) return NULL; + + /* Retrieve value from hash by the field name. This operation + * already increases the refcount of the returned object. */ + initStaticStringObject(fieldobj,((char*)&fieldname)+(sizeof(long)*2)); + o = hashGet(o, &fieldobj); + } else { + if (o->type != REDIS_STRING) return NULL; + + /* Every object that this function returns needs to have its refcount + * increased. sortCommand decreases it again. */ + incrRefCount(o); + } + + return o; } /* sortCompare() is used by qsort in sortCommand(). Given that qsort_r with @@ -6324,14 +6896,8 @@ static int sortCompare(const void *s1, const void *s2) { cmp = strcoll(so1->u.cmpobj->ptr,so2->u.cmpobj->ptr); } } else { - /* Compare elements directly */ - robj *dec1, *dec2; - - dec1 = getDecodedObject(so1->obj); - dec2 = getDecodedObject(so2->obj); - cmp = strcoll(dec1->ptr,dec2->ptr); - decrRefCount(dec1); - decrRefCount(dec2); + /* Compare elements directly. */ + cmp = compareStringObjects(so1->obj,so2->obj); } } return server.sort_desc ? -cmp : cmp; @@ -6352,7 +6918,7 @@ static void sortCommand(redisClient *c) { /* Lookup the key to sort. It must be of the right types */ sortval = lookupKeyRead(c->db,c->argv[1]); if (sortval == NULL) { - addReply(c,shared.nullmultibulk); + addReply(c,shared.emptymultibulk); return; } if (sortval->type != REDIS_SET && sortval->type != REDIS_LIST && @@ -6414,7 +6980,7 @@ static void sortCommand(redisClient *c) { case REDIS_LIST: vectorlen = listLength((list*)sortval->ptr); break; case REDIS_SET: vectorlen = dictSize((dict*)sortval->ptr); break; case REDIS_ZSET: vectorlen = dictSize(((zset*)sortval->ptr)->dict); break; - default: vectorlen = 0; redisAssert(0); /* Avoid GCC warning */ + default: vectorlen = 0; redisPanic("Bad SORT type"); /* Avoid GCC warning */ } vector = zmalloc(sizeof(redisSortObject)*vectorlen); j = 0; @@ -6458,38 +7024,36 @@ static void sortCommand(redisClient *c) { /* Now it's time to load the right scores in the sorting vector */ if (dontsort == 0) { for (j = 0; j < vectorlen; j++) { + robj *byval; if (sortby) { - robj *byval; - + /* lookup value to sort by */ byval = lookupKeyByPattern(c->db,sortby,vector[j].obj); - if (!byval || byval->type != REDIS_STRING) continue; - if (alpha) { - vector[j].u.cmpobj = getDecodedObject(byval); - } else { - if (byval->encoding == REDIS_ENCODING_RAW) { - vector[j].u.score = strtod(byval->ptr,NULL); - } else { - /* Don't need to decode the object if it's - * integer-encoded (the only encoding supported) so - * far. We can just cast it */ - if (byval->encoding == REDIS_ENCODING_INT) { - vector[j].u.score = (long)byval->ptr; - } else - redisAssert(1 != 1); - } - } + if (!byval) continue; } else { - 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 - redisAssert(1 != 1); - } + /* use object itself to sort by */ + byval = vector[j].obj; + } + + if (alpha) { + if (sortby) vector[j].u.cmpobj = getDecodedObject(byval); + } else { + if (byval->encoding == REDIS_ENCODING_RAW) { + vector[j].u.score = strtod(byval->ptr,NULL); + } else if (byval->encoding == REDIS_ENCODING_INT) { + /* Don't need to decode the object if it's + * integer-encoded (the only encoding supported) so + * far. We can just cast it */ + vector[j].u.score = (long)byval->ptr; + } else { + redisAssert(1 != 1); } } + + /* when the object was retrieved using lookupKeyByPattern, + * its refcount needs to be decreased. */ + if (sortby) { + decrRefCount(byval); + } } } @@ -6531,10 +7095,11 @@ static void sortCommand(redisClient *c) { vector[j].obj); if (sop->type == REDIS_SORT_GET) { - if (!val || val->type != REDIS_STRING) { + if (!val) { addReply(c,shared.nullbulk); } else { addReplyBulk(c,val); + decrRefCount(val); } } else { redisAssert(sop->type == REDIS_SORT_GET); /* always fails */ @@ -6561,11 +7126,14 @@ static void sortCommand(redisClient *c) { vector[j].obj); if (sop->type == REDIS_SORT_GET) { - if (!val || val->type != REDIS_STRING) { + if (!val) { listAddNodeTail(listPtr,createStringObject("",0)); } else { + /* We should do a incrRefCount on val because it is + * added to the list, but also a decrRefCount because + * it is returned by lookupKeyByPattern. This results + * in doing nothing at all. */ listAddNodeTail(listPtr,val); - incrRefCount(val); } } else { redisAssert(sop->type == REDIS_SORT_GET); /* always fails */ @@ -6586,7 +7154,7 @@ static void sortCommand(redisClient *c) { decrRefCount(sortval); listRelease(operations); for (j = 0; j < vectorlen; j++) { - if (sortby && alpha && vector[j].u.cmpobj) + if (alpha && vector[j].u.cmpobj) decrRefCount(vector[j].u.cmpobj); } zfree(vector); @@ -6625,6 +7193,8 @@ static sds genRedisInfoString(void) { bytesToHuman(hmem,zmalloc_used_memory()); info = sdscatprintf(sdsempty(), "redis_version:%s\r\n" + "redis_git_sha1:%s\r\n" + "redis_git_dirty:%d\r\n" "arch_bits:%s\r\n" "multiplexing_api:%s\r\n" "process_id:%ld\r\n" @@ -6642,11 +7212,15 @@ static sds genRedisInfoString(void) { "total_connections_received:%lld\r\n" "total_commands_processed:%lld\r\n" "expired_keys:%lld\r\n" - "hash_max_zipmap_entries:%ld\r\n" - "hash_max_zipmap_value:%ld\r\n" + "hash_max_zipmap_entries:%zu\r\n" + "hash_max_zipmap_value:%zu\r\n" + "pubsub_channels:%ld\r\n" + "pubsub_patterns:%u\r\n" "vm_enabled:%d\r\n" "role:%s\r\n" ,REDIS_VERSION, + REDIS_GIT_SHA1, + strtol(REDIS_GIT_DIRTY,NULL,10) > 0, (sizeof(long) == 8) ? "64" : "32", aeGetApiName(), (long) getpid(), @@ -6666,6 +7240,8 @@ static sds genRedisInfoString(void) { server.stat_expiredkeys, server.hash_max_zipmap_entries, server.hash_max_zipmap_value, + dictSize(server.pubsub_channels), + listLength(server.pubsub_patterns), server.vm_enabled != 0, server.masterhost == NULL ? "master" : "slave" ); @@ -6805,15 +7381,20 @@ static int deleteIfVolatile(redisDb *db, robj *key) { return dictDelete(db->dict,key) == DICT_OK; } -static void expireGenericCommand(redisClient *c, robj *key, time_t seconds) { +static void expireGenericCommand(redisClient *c, robj *key, robj *param, long offset) { dictEntry *de; + time_t seconds; + + if (getLongFromObjectOrReply(c, param, &seconds, NULL) != REDIS_OK) return; + + seconds -= offset; de = dictFind(c->db->dict,key); if (de == NULL) { addReply(c,shared.czero); return; } - if (seconds < 0) { + if (seconds <= 0) { if (deleteKey(c->db,key)) server.dirty++; addReply(c, shared.cone); return; @@ -6830,11 +7411,11 @@ static void expireGenericCommand(redisClient *c, robj *key, time_t seconds) { } static void expireCommand(redisClient *c) { - expireGenericCommand(c,c->argv[1],strtol(c->argv[2]->ptr,NULL,10)); + expireGenericCommand(c,c->argv[1],c->argv[2],0); } static void expireatCommand(redisClient *c) { - expireGenericCommand(c,c->argv[1],strtol(c->argv[2]->ptr,NULL,10)-time(NULL)); + expireGenericCommand(c,c->argv[1],c->argv[2],time(NULL)); } static void ttlCommand(redisClient *c) { @@ -6906,6 +7487,20 @@ static void discardCommand(redisClient *c) { addReply(c,shared.ok); } +/* Send a MULTI command to all the slaves and AOF file. Check the execCommand + * implememntation for more information. */ +static void execCommandReplicateMulti(redisClient *c) { + struct redisCommand *cmd; + robj *multistring = createStringObject("MULTI",5); + + cmd = lookupCommand("multi"); + if (server.appendonly) + feedAppendOnlyFile(cmd,c->db->id,&multistring,1); + if (listLength(server.slaves)) + replicationFeedSlaves(server.slaves,c->db->id,&multistring,1); + decrRefCount(multistring); +} + static void execCommand(redisClient *c) { int j; robj **orig_argv; @@ -6916,6 +7511,13 @@ static void execCommand(redisClient *c) { return; } + /* Replicate a MULTI request now that we are sure the block is executed. + * This way we'll deliver the MULTI/..../EXEC block as a whole and + * both the AOF and the replication link will have the same consistency + * and atomicity guarantees. */ + execCommandReplicateMulti(c); + + /* Exec all the queued commands */ orig_argv = c->argv; orig_argc = c->argc; addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",c->mstate.count)); @@ -6929,6 +7531,10 @@ static void execCommand(redisClient *c) { freeClientMultiState(c); initClientMultiState(c); c->flags &= (~REDIS_MULTI); + /* Make sure the EXEC command is always replicated / AOF, since we + * always send the MULTI command (we can't know beforehand if the + * next operations will contain at least a modification to the DB). */ + server.dirty++; } /* =========================== Blocking Operations ========================= */ @@ -7078,7 +7684,7 @@ static void blockingPopGenericCommand(redisClient *c, int where) { * non-blocking POP operation */ robj *argv[2], **orig_argv; int orig_argc; - + /* We need to alter the command arguments before to call * popGenericCommand() as the command takes a single key. */ orig_argv = c->argv; @@ -7315,7 +7921,7 @@ static void updateSlavesWaitingBgsave(int bgsaveerr) { slave->replstate = REDIS_REPL_WAIT_BGSAVE_END; } else if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) { struct redis_stat buf; - + if (bgsaveerr != REDIS_OK) { freeClient(slave); redisLog(REDIS_WARNING,"SYNC failed. BGSAVE child returned an error"); @@ -7551,11 +8157,87 @@ static void freeMemoryIfNeeded(void) { /* ============================== Append Only file ========================== */ +/* Write the append only file buffer on disk. + * + * Since we are required to write the AOF before replying to the client, + * and the only way the client socket can get a write is entering when the + * the event loop, we accumulate all the AOF writes in a memory + * buffer and write it on disk using this function just before entering + * the event loop again. */ +static void flushAppendOnlyFile(void) { + time_t now; + ssize_t nwritten; + + if (sdslen(server.aofbuf) == 0) return; + + /* 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,server.aofbuf,sdslen(server.aofbuf)); + if (nwritten != (signed)sdslen(server.aofbuf)) { + /* Ooops, we are in troubles. The best thing to do for now is + * aborting instead of giving 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); + } + sdsfree(server.aofbuf); + server.aofbuf = sdsempty(); + + /* Fsync if needed */ + now = time(NULL); + if (server.appendfsync == APPENDFSYNC_ALWAYS || + (server.appendfsync == APPENDFSYNC_EVERYSEC && + now-server.lastfsync > 1)) + { + /* aof_fsync is defined as fdatasync() for Linux in order to avoid + * flushing metadata. */ + aof_fsync(server.appendfd); /* Let's try to get this data on the disk */ + server.lastfsync = now; + } +} + +static sds catAppendOnlyGenericCommand(sds buf, int argc, robj **argv) { + int j; + buf = sdscatprintf(buf,"*%d\r\n",argc); + for (j = 0; j < argc; j++) { + robj *o = getDecodedObject(argv[j]); + buf = sdscatprintf(buf,"$%lu\r\n",(unsigned long)sdslen(o->ptr)); + buf = sdscatlen(buf,o->ptr,sdslen(o->ptr)); + buf = sdscatlen(buf,"\r\n",2); + decrRefCount(o); + } + return buf; +} + +static sds catAppendOnlyExpireAtCommand(sds buf, robj *key, robj *seconds) { + int argc = 3; + long when; + robj *argv[3]; + + /* Make sure we can use strtol */ + seconds = getDecodedObject(seconds); + when = time(NULL)+strtol(seconds->ptr,NULL,10); + decrRefCount(seconds); + + argv[0] = createStringObject("EXPIREAT",8); + argv[1] = key; + argv[2] = createObject(REDIS_STRING, + sdscatprintf(sdsempty(),"%ld",when)); + buf = catAppendOnlyGenericCommand(buf, argc, argv); + decrRefCount(argv[0]); + decrRefCount(argv[2]); + return buf; +} + static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc) { sds buf = sdsempty(); - int j; - ssize_t nwritten; - time_t now; robj *tmpargv[3]; /* The DB this command was targetting is not the same as the last command @@ -7569,55 +8251,26 @@ static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv server.appendseldb = dictid; } - /* "Fix" the argv vector if the command is EXPIRE. We want to translate - * EXPIREs into EXPIREATs calls */ if (cmd->proc == expireCommand) { - long when; - - tmpargv[0] = createStringObject("EXPIREAT",8); + /* Translate EXPIRE into EXPIREAT */ + buf = catAppendOnlyExpireAtCommand(buf,argv[1],argv[2]); + } else if (cmd->proc == setexCommand) { + /* Translate SETEX to SET and EXPIREAT */ + tmpargv[0] = createStringObject("SET",3); tmpargv[1] = argv[1]; - incrRefCount(argv[1]); - when = time(NULL)+strtol(argv[2]->ptr,NULL,10); - tmpargv[2] = createObject(REDIS_STRING, - sdscatprintf(sdsempty(),"%ld",when)); - argv = tmpargv; - } - - /* Append the actual command */ - buf = sdscatprintf(buf,"*%d\r\n",argc); - for (j = 0; j < argc; j++) { - robj *o = argv[j]; - - o = getDecodedObject(o); - buf = sdscatprintf(buf,"$%lu\r\n",(unsigned long)sdslen(o->ptr)); - buf = sdscatlen(buf,o->ptr,sdslen(o->ptr)); - buf = sdscatlen(buf,"\r\n",2); - decrRefCount(o); + tmpargv[2] = argv[3]; + buf = catAppendOnlyGenericCommand(buf,3,tmpargv); + decrRefCount(tmpargv[0]); + buf = catAppendOnlyExpireAtCommand(buf,argv[1],argv[2]); + } else { + buf = catAppendOnlyGenericCommand(buf,argc,argv); } - /* Free the objects from the modified argv for EXPIREAT */ - if (cmd->proc == expireCommand) { - for (j = 0; j < 3; j++) - decrRefCount(argv[j]); - } + /* Append to the AOF buffer. This will be flushed on disk just before + * of re-entering the event loop, so before the client will get a + * positive reply about the operation performed. */ + server.aofbuf = sdscatlen(server.aofbuf,buf,sdslen(buf)); - /* We want to perform a single write. This should be guaranteed atomic - * at least if the filesystem we are writing is a real physical one. - * While this will save us against the server being killed I don't think - * there is much to do about the whole server stopping for power problems - * or alike */ - nwritten = write(server.appendfd,buf,sdslen(buf)); - if (nwritten != (signed)sdslen(buf)) { - /* Ooops, we are in troubles. The best thing to do for now is - * to simply exit instead to give the illusion that everything is - * working as expected. */ - if (nwritten == -1) { - redisLog(REDIS_WARNING,"Exiting on error writing to the append-only file: %s",strerror(errno)); - } else { - redisLog(REDIS_WARNING,"Exiting on short write while writing to the append-only file: %s",strerror(errno)); - } - exit(1); - } /* If a background append only file rewriting is in progress we want to * accumulate the differences between the child DB and the current one * in a buffer, so that when the child process will do its work we @@ -7626,14 +8279,6 @@ static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv server.bgrewritebuf = sdscatlen(server.bgrewritebuf,buf,sdslen(buf)); sdsfree(buf); - now = time(NULL); - if (server.appendfsync == APPENDFSYNC_ALWAYS || - (server.appendfsync == APPENDFSYNC_EVERYSEC && - now-server.lastfsync > 1)) - { - fsync(server.appendfd); /* Let's try to get this data on the disk */ - server.lastfsync = now; - } } /* In Redis commands are always executed in the context of a client, so in @@ -7653,12 +8298,14 @@ static struct redisClient *createFakeClient(void) { c->reply = listCreate(); listSetFreeMethod(c->reply,decrRefCount); listSetDupMethod(c->reply,dupClientReplyValue); + initClientMultiState(c); return c; } static void freeFakeClient(struct redisClient *c) { sdsfree(c->querybuf); listRelease(c->reply); + freeClientMultiState(c); zfree(c); } @@ -7670,6 +8317,7 @@ int loadAppendOnlyFile(char *filename) { FILE *fp = fopen(filename,"r"); struct redis_stat sb; unsigned long long loadedkeys = 0; + int appendonly = server.appendonly; if (redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0) return REDIS_ERR; @@ -7679,6 +8327,10 @@ int loadAppendOnlyFile(char *filename) { exit(1); } + /* Temporarily disable AOF, to prevent EXEC from feeding a MULTI + * to the same file we're about to read. */ + server.appendonly = 0; + fakeClient = createFakeClient(); while(1) { int argc, j; @@ -7713,14 +8365,9 @@ int loadAppendOnlyFile(char *filename) { 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]); - } + /* Try object encoding */ if (cmd->flags & REDIS_CMD_BULK) - tryObjectEncoding(argv[argc-1]); + argv[argc-1] = tryObjectEncoding(argv[argc-1]); /* Run the command in the context of a fake client */ fakeClient->argc = argc; fakeClient->argv = argv; @@ -7739,8 +8386,14 @@ int loadAppendOnlyFile(char *filename) { } } } + + /* This point can only be reached when EOF is reached without errors. + * If the client is in the middle of a MULTI/EXEC, log error and quit. */ + if (fakeClient->flags & REDIS_MULTI) goto readerr; + fclose(fp); freeFakeClient(fakeClient); + server.appendonly = appendonly; return REDIS_OK; readerr: @@ -7957,7 +8610,7 @@ static int rewriteAppendOnlyFile(char *filename) { dictReleaseIterator(di); } } else { - redisAssert(0); + redisPanic("Unknown object type"); } /* Save the expire time */ if (expiretime != -1) { @@ -7977,7 +8630,7 @@ static int rewriteAppendOnlyFile(char *filename) { fflush(fp); fsync(fileno(fp)); fclose(fp); - + /* Use RENAME to make sure the DB file is changed atomically only * if the generate DB file is ok. */ if (rename(tmpfile,filename) == -1) { @@ -8036,6 +8689,7 @@ static int rewriteAppendOnlyFileBackground(void) { redisLog(REDIS_NOTICE, "Background append only file rewriting started by pid %d",childpid); server.bgrewritechildpid = childpid; + updateDictResizePolicy(); /* We set appendseldb to -1 in order to force the next call to the * feedAppendOnlyFile() to issue a SELECT command, so the differences * accumulated by the parent into server.bgrewritebuf will start @@ -8087,44 +8741,82 @@ static void aofRemoveTempFile(pid_t childpid) { * as a fully non-blocking VM. */ -/* =================== Virtual Memory - Blocking Side ====================== */ +/* Called when the user switches from "appendonly yes" to "appendonly no" + * at runtime using the CONFIG command. */ +static void stopAppendOnly(void) { + flushAppendOnlyFile(); + fsync(server.appendfd); + close(server.appendfd); -/* substitute the first occurrence of '%p' with the process pid in the - * swap file name. */ -static void expandVmSwapFilename(void) { - char *p = strstr(server.vm_swap_file,"%p"); - sds new; - - if (!p) return; - new = sdsempty(); - *p = '\0'; - new = sdscat(new,server.vm_swap_file); - new = sdscatprintf(new,"%ld",(long) getpid()); - new = sdscat(new,p+2); - zfree(server.vm_swap_file); - server.vm_swap_file = new; + server.appendfd = -1; + server.appendseldb = -1; + server.appendonly = 0; + /* rewrite operation in progress? kill it, wait child exit */ + if (server.bgsavechildpid != -1) { + int statloc; + + if (kill(server.bgsavechildpid,SIGKILL) != -1) + wait3(&statloc,0,NULL); + /* reset the buffer accumulating changes while the child saves */ + sdsfree(server.bgrewritebuf); + server.bgrewritebuf = sdsempty(); + server.bgsavechildpid = -1; + } } +/* Called when the user switches from "appendonly no" to "appendonly yes" + * at runtime using the CONFIG command. */ +static int startAppendOnly(void) { + server.appendonly = 1; + server.lastfsync = time(NULL); + server.appendfd = open(server.appendfilename,O_WRONLY|O_APPEND|O_CREAT,0644); + if (server.appendfd == -1) { + redisLog(REDIS_WARNING,"Used tried to switch on AOF via CONFIG, but I can't open the AOF file: %s",strerror(errno)); + return REDIS_ERR; + } + if (rewriteAppendOnlyFileBackground() == REDIS_ERR) { + server.appendonly = 0; + close(server.appendfd); + redisLog(REDIS_WARNING,"Used tried to switch on AOF via CONFIG, I can't trigger a background AOF rewrite operation. Check the above logs for more info about the error.",strerror(errno)); + return REDIS_ERR; + } + return REDIS_OK; +} + +/* =================== Virtual Memory - Blocking Side ====================== */ + static void vmInit(void) { off_t totsize; int pipefds[2]; size_t stacksize; + struct flock fl; if (server.vm_max_threads != 0) zmalloc_enable_thread_safeness(); /* we need thread safe zmalloc() */ - expandVmSwapFilename(); redisLog(REDIS_NOTICE,"Using '%s' as swap file",server.vm_swap_file); + /* Try to open the old swap file, otherwise create it */ if ((server.vm_fp = fopen(server.vm_swap_file,"r+b")) == NULL) { server.vm_fp = fopen(server.vm_swap_file,"w+b"); } if (server.vm_fp == NULL) { redisLog(REDIS_WARNING, - "Impossible to open the swap file: %s. Exiting.", + "Can't open the swap file: %s. Exiting.", strerror(errno)); exit(1); } server.vm_fd = fileno(server.vm_fp); + /* Lock the swap file for writing, this is useful in order to avoid + * another instance to use the same swap file for a config error. */ + fl.l_type = F_WRLCK; + fl.l_whence = SEEK_SET; + fl.l_start = fl.l_len = 0; + if (fcntl(server.vm_fd,F_SETLK,&fl) == -1) { + redisLog(REDIS_WARNING, + "Can't lock the swap file at '%s': %s. Make sure it is not used by another Redis instance.", server.vm_swap_file, strerror(errno)); + exit(1); + } + /* Initialize */ server.vm_next_page = 0; server.vm_near_pages = 0; server.vm_stats_used_pages = 0; @@ -8219,7 +8911,7 @@ static int vmFreePage(off_t page) { } /* Find N contiguous free pages storing the first page of the cluster in *first. - * Returns REDIS_OK if it was able to find N contiguous pages, otherwise + * Returns REDIS_OK if it was able to find N contiguous pages, otherwise * REDIS_ERR is returned. * * This function uses a simple algorithm: we try to allocate @@ -8230,7 +8922,7 @@ static int vmFreePage(off_t page) { * we try to find less populated places doing a forward jump of * REDIS_VM_MAX_RANDOM_JUMP, then we start scanning again a few pages * without hurry, and then we jump again and so forth... - * + * * This function can be improved using a free list to avoid to guess * too much, since we could collect data about freed pages. * @@ -8603,7 +9295,10 @@ static void freeIOJob(iojob *j) { j->type == REDIS_IOJOB_DO_SWAP || j->type == REDIS_IOJOB_LOAD) && j->val != NULL) decrRefCount(j->val); - decrRefCount(j->key); + /* We don't decrRefCount the j->key field as we did't incremented + * the count creating IO Jobs. This is because the key field here is + * just used as an indentifier and if a key is removed the Job should + * never be touched again. */ zfree(j); } @@ -8776,7 +9471,7 @@ again: iojob *job = ln->value; if (job->canceled) continue; /* Skip this, already canceled. */ - if (compareStringObjects(job->key,o) == 0) { + if (job->key == o) { redisLog(REDIS_DEBUG,"*** CANCELED %p (%s) (type %d) (LIST ID %d)\n", (void*)job, (char*)o->ptr, job->type, i); /* Mark the pages as free since the swap didn't happened @@ -8881,7 +9576,7 @@ static void *IOThreadEntryPoint(void *arg) { listDelNode(server.io_processing,ln); listAddNodeTail(server.io_processed,j); unlockThreadedIO(); - + /* Signal the main thread there is new stuff to process */ assert(write(server.io_ready_pipe_write,"x",1) == 1); } @@ -8959,14 +9654,14 @@ static void queueIOJob(iojob *j) { static int vmSwapObjectThreaded(robj *key, robj *val, redisDb *db) { iojob *j; - + assert(key->storage == REDIS_VM_MEMORY); assert(key->refcount == 1); j = zmalloc(sizeof(*j)); j->type = REDIS_IOJOB_PREPARE_SWAP; j->db = db; - j->key = dupStringObject(key); + j->key = key; j->val = val; incrRefCount(val); j->canceled = 0; @@ -9003,7 +9698,7 @@ static int waitForSwappedKey(redisClient *c, robj *key) { vmCancelThreadedIOJob(o); return 0; } - + /* OK: the key is either swapped, or being loaded just now. */ /* Add the key to the list of keys this client is waiting for. @@ -9034,7 +9729,7 @@ static int waitForSwappedKey(redisClient *c, robj *key) { j = zmalloc(sizeof(*j)); j->type = REDIS_IOJOB_LOAD; j->db = c->db; - j->key = dupStringObject(key); + j->key = o; j->key->vtype = o->vtype; j->page = o->vm.page; j->val = NULL; @@ -9047,12 +9742,56 @@ static int waitForSwappedKey(redisClient *c, robj *key) { return 1; } -/* Preload keys needed for the ZUNION and ZINTER commands. */ -static void zunionInterBlockClientOnSwappedKeys(redisClient *c) { +/* Preload keys for any command with first, last and step values for + * the command keys prototype, as defined in the command table. */ +static void waitForMultipleSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv) { + int j, last; + if (cmd->vm_firstkey == 0) return; + last = cmd->vm_lastkey; + if (last < 0) last = argc+last; + for (j = cmd->vm_firstkey; j <= last; j += cmd->vm_keystep) { + redisAssert(j < argc); + waitForSwappedKey(c,argv[j]); + } +} + +/* Preload keys needed for the ZUNIONSTORE and ZINTERSTORE commands. + * Note that the number of keys to preload is user-defined, so we need to + * apply a sanity check against argc. */ +static void zunionInterBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv) { int i, num; - num = atoi(c->argv[2]->ptr); + REDIS_NOTUSED(cmd); + + num = atoi(argv[2]->ptr); + if (num > (argc-3)) return; for (i = 0; i < num; i++) { - waitForSwappedKey(c,c->argv[3+i]); + waitForSwappedKey(c,argv[3+i]); + } +} + +/* Preload keys needed to execute the entire MULTI/EXEC block. + * + * This function is called by blockClientOnSwappedKeys when EXEC is issued, + * and will block the client when any command requires a swapped out value. */ +static void execBlockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd, int argc, robj **argv) { + int i, margc; + struct redisCommand *mcmd; + robj **margv; + REDIS_NOTUSED(cmd); + REDIS_NOTUSED(argc); + REDIS_NOTUSED(argv); + + if (!(c->flags & REDIS_MULTI)) return; + for (i = 0; i < c->mstate.count; i++) { + mcmd = c->mstate.commands[i].cmd; + margc = c->mstate.commands[i].argc; + margv = c->mstate.commands[i].argv; + + if (mcmd->vm_preload_proc != NULL) { + mcmd->vm_preload_proc(c,mcmd,margc,margv); + } else { + waitForMultipleSwappedKeys(c,mcmd,margc,margv); + } } } @@ -9066,17 +9805,11 @@ static void zunionInterBlockClientOnSwappedKeys(redisClient *c) { * * Return 1 if the client is marked as blocked, 0 if the client can * continue as the keys it is going to access appear to be in memory. */ -static int blockClientOnSwappedKeys(struct redisCommand *cmd, redisClient *c) { - int j, last; - +static int blockClientOnSwappedKeys(redisClient *c, struct redisCommand *cmd) { if (cmd->vm_preload_proc != NULL) { - cmd->vm_preload_proc(c); + cmd->vm_preload_proc(c,cmd,c->argc,c->argv); } else { - if (cmd->vm_firstkey == 0) return 0; - last = cmd->vm_lastkey; - if (last < 0) last = c->argc+last; - for (j = cmd->vm_firstkey; j <= last; j += cmd->vm_keystep) - waitForSwappedKey(c,c->argv[j]); + waitForMultipleSwappedKeys(c,cmd,c->argc,c->argv); } /* If the client was blocked for at least one key, mark it as blocked. */ @@ -9103,7 +9836,7 @@ static int dontWaitForSwappedKey(redisClient *c, robj *key) { /* Remove the key from the list of keys this client is waiting for. */ listRewind(c->io_keys,&li); while ((ln = listNext(&li)) != NULL) { - if (compareStringObjects(ln->value,key) == 0) { + if (equalStringObjects(ln->value,key)) { listDelNode(c->io_keys,ln); break; } @@ -9152,6 +9885,8 @@ static void handleClientsBlockedOnSwappedKey(redisDb *db, robj *key) { static void configSetCommand(redisClient *c) { robj *o = getDecodedObject(c->argv[3]); + long long ll; + if (!strcasecmp(c->argv[2]->ptr,"dbfilename")) { zfree(server.dbfilename); server.dbfilename = zstrdup(o->ptr); @@ -9162,7 +9897,74 @@ static void configSetCommand(redisClient *c) { zfree(server.masterauth); server.masterauth = zstrdup(o->ptr); } else if (!strcasecmp(c->argv[2]->ptr,"maxmemory")) { - server.maxmemory = strtoll(o->ptr, NULL, 10); + if (getLongLongFromObject(o,&ll) == REDIS_ERR || + ll < 0) goto badfmt; + server.maxmemory = ll; + } else if (!strcasecmp(c->argv[2]->ptr,"timeout")) { + if (getLongLongFromObject(o,&ll) == REDIS_ERR || + ll < 0 || ll > LONG_MAX) goto badfmt; + server.maxidletime = ll; + } else if (!strcasecmp(c->argv[2]->ptr,"appendfsync")) { + if (!strcasecmp(o->ptr,"no")) { + server.appendfsync = APPENDFSYNC_NO; + } else if (!strcasecmp(o->ptr,"everysec")) { + server.appendfsync = APPENDFSYNC_EVERYSEC; + } else if (!strcasecmp(o->ptr,"always")) { + server.appendfsync = APPENDFSYNC_ALWAYS; + } else { + goto badfmt; + } + } else if (!strcasecmp(c->argv[2]->ptr,"appendonly")) { + int old = server.appendonly; + int new = yesnotoi(o->ptr); + + if (new == -1) goto badfmt; + if (old != new) { + if (new == 0) { + stopAppendOnly(); + } else { + if (startAppendOnly() == REDIS_ERR) { + addReplySds(c,sdscatprintf(sdsempty(), + "-ERR Unable to turn on AOF. Check server logs.\r\n")); + decrRefCount(o); + return; + } + } + } + } else if (!strcasecmp(c->argv[2]->ptr,"save")) { + int vlen, j; + sds *v = sdssplitlen(o->ptr,sdslen(o->ptr)," ",1,&vlen); + + /* Perform sanity check before setting the new config: + * - Even number of args + * - Seconds >= 1, changes >= 0 */ + if (vlen & 1) { + sdsfreesplitres(v,vlen); + goto badfmt; + } + for (j = 0; j < vlen; j++) { + char *eptr; + long val; + + val = strtoll(v[j], &eptr, 10); + if (eptr[0] != '\0' || + ((j & 1) == 0 && val < 1) || + ((j & 1) == 1 && val < 0)) { + sdsfreesplitres(v,vlen); + goto badfmt; + } + } + /* Finally set the new config */ + resetServerSaveParams(); + for (j = 0; j < vlen; j += 2) { + time_t seconds; + int changes; + + seconds = strtoll(v[j],NULL,10); + changes = strtoll(v[j+1],NULL,10); + appendServerSaveParams(seconds, changes); + } + sdsfreesplitres(v,vlen); } else { addReplySds(c,sdscatprintf(sdsempty(), "-ERR not supported CONFIG parameter %s\r\n", @@ -9172,6 +9974,14 @@ static void configSetCommand(redisClient *c) { } decrRefCount(o); addReply(c,shared.ok); + return; + +badfmt: /* Bad format errors */ + addReplySds(c,sdscatprintf(sdsempty(), + "-ERR invalid argument '%s' for CONFIG SET '%s'\r\n", + (char*)o->ptr, + (char*)c->argv[2]->ptr)); + decrRefCount(o); } static void configGetCommand(redisClient *c) { @@ -9201,11 +10011,53 @@ static void configGetCommand(redisClient *c) { if (stringmatch(pattern,"maxmemory",0)) { char buf[128]; - snprintf(buf,128,"%llu\n",server.maxmemory); + ll2string(buf,128,server.maxmemory); addReplyBulkCString(c,"maxmemory"); addReplyBulkCString(c,buf); matches++; } + if (stringmatch(pattern,"timeout",0)) { + char buf[128]; + + ll2string(buf,128,server.maxidletime); + addReplyBulkCString(c,"timeout"); + addReplyBulkCString(c,buf); + matches++; + } + if (stringmatch(pattern,"appendonly",0)) { + addReplyBulkCString(c,"appendonly"); + addReplyBulkCString(c,server.appendonly ? "yes" : "no"); + matches++; + } + if (stringmatch(pattern,"appendfsync",0)) { + char *policy; + + switch(server.appendfsync) { + case APPENDFSYNC_NO: policy = "no"; break; + case APPENDFSYNC_EVERYSEC: policy = "everysec"; break; + case APPENDFSYNC_ALWAYS: policy = "always"; break; + default: policy = "unknown"; break; /* too harmless to panic */ + } + addReplyBulkCString(c,"appendfsync"); + addReplyBulkCString(c,policy); + matches++; + } + if (stringmatch(pattern,"save",0)) { + sds buf = sdsempty(); + int j; + + for (j = 0; j < server.saveparamslen; j++) { + buf = sdscatprintf(buf,"%ld %d", + server.saveparams[j].seconds, + server.saveparams[j].changes); + if (j != server.saveparamslen-1) + buf = sdscatlen(buf," ",1); + } + addReplyBulkCString(c,"save"); + addReplyBulkCString(c,buf); + sdsfree(buf); + matches++; + } decrRefCount(o); lenobj->ptr = sdscatprintf(sdsempty(),"*%d\r\n",matches*2); } @@ -9236,8 +10088,446 @@ badarity: (char*) c->argv[1]->ptr)); } +/* =========================== Pubsub implementation ======================== */ + +static void freePubsubPattern(void *p) { + pubsubPattern *pat = p; + + decrRefCount(pat->pattern); + zfree(pat); +} + +static int listMatchPubsubPattern(void *a, void *b) { + pubsubPattern *pa = a, *pb = b; + + return (pa->client == pb->client) && + (equalStringObjects(pa->pattern,pb->pattern)); +} + +/* Subscribe a client to a channel. Returns 1 if the operation succeeded, or + * 0 if the client was already subscribed to that channel. */ +static int pubsubSubscribeChannel(redisClient *c, robj *channel) { + struct dictEntry *de; + list *clients = NULL; + int retval = 0; + + /* Add the channel to the client -> channels hash table */ + if (dictAdd(c->pubsub_channels,channel,NULL) == DICT_OK) { + retval = 1; + incrRefCount(channel); + /* Add the client to the channel -> list of clients hash table */ + de = dictFind(server.pubsub_channels,channel); + if (de == NULL) { + clients = listCreate(); + dictAdd(server.pubsub_channels,channel,clients); + incrRefCount(channel); + } else { + clients = dictGetEntryVal(de); + } + listAddNodeTail(clients,c); + } + /* Notify the client */ + addReply(c,shared.mbulk3); + addReply(c,shared.subscribebulk); + addReplyBulk(c,channel); + addReplyLongLong(c,dictSize(c->pubsub_channels)+listLength(c->pubsub_patterns)); + return retval; +} + +/* Unsubscribe a client from a channel. Returns 1 if the operation succeeded, or + * 0 if the client was not subscribed to the specified channel. */ +static int pubsubUnsubscribeChannel(redisClient *c, robj *channel, int notify) { + struct dictEntry *de; + list *clients; + listNode *ln; + int retval = 0; + + /* Remove the channel from the client -> channels hash table */ + incrRefCount(channel); /* channel may be just a pointer to the same object + we have in the hash tables. Protect it... */ + if (dictDelete(c->pubsub_channels,channel) == DICT_OK) { + retval = 1; + /* Remove the client from the channel -> clients list hash table */ + de = dictFind(server.pubsub_channels,channel); + assert(de != NULL); + clients = dictGetEntryVal(de); + ln = listSearchKey(clients,c); + assert(ln != NULL); + listDelNode(clients,ln); + if (listLength(clients) == 0) { + /* Free the list and associated hash entry at all if this was + * the latest client, so that it will be possible to abuse + * Redis PUBSUB creating millions of channels. */ + dictDelete(server.pubsub_channels,channel); + } + } + /* Notify the client */ + if (notify) { + addReply(c,shared.mbulk3); + addReply(c,shared.unsubscribebulk); + addReplyBulk(c,channel); + addReplyLongLong(c,dictSize(c->pubsub_channels)+ + listLength(c->pubsub_patterns)); + + } + decrRefCount(channel); /* it is finally safe to release it */ + return retval; +} + +/* Subscribe a client to a pattern. Returns 1 if the operation succeeded, or 0 if the clinet was already subscribed to that pattern. */ +static int pubsubSubscribePattern(redisClient *c, robj *pattern) { + int retval = 0; + + if (listSearchKey(c->pubsub_patterns,pattern) == NULL) { + retval = 1; + pubsubPattern *pat; + listAddNodeTail(c->pubsub_patterns,pattern); + incrRefCount(pattern); + pat = zmalloc(sizeof(*pat)); + pat->pattern = getDecodedObject(pattern); + pat->client = c; + listAddNodeTail(server.pubsub_patterns,pat); + } + /* Notify the client */ + addReply(c,shared.mbulk3); + addReply(c,shared.psubscribebulk); + addReplyBulk(c,pattern); + addReplyLongLong(c,dictSize(c->pubsub_channels)+listLength(c->pubsub_patterns)); + return retval; +} + +/* Unsubscribe a client from a channel. Returns 1 if the operation succeeded, or + * 0 if the client was not subscribed to the specified channel. */ +static int pubsubUnsubscribePattern(redisClient *c, robj *pattern, int notify) { + listNode *ln; + pubsubPattern pat; + int retval = 0; + + incrRefCount(pattern); /* Protect the object. May be the same we remove */ + if ((ln = listSearchKey(c->pubsub_patterns,pattern)) != NULL) { + retval = 1; + listDelNode(c->pubsub_patterns,ln); + pat.client = c; + pat.pattern = pattern; + ln = listSearchKey(server.pubsub_patterns,&pat); + listDelNode(server.pubsub_patterns,ln); + } + /* Notify the client */ + if (notify) { + addReply(c,shared.mbulk3); + addReply(c,shared.punsubscribebulk); + addReplyBulk(c,pattern); + addReplyLongLong(c,dictSize(c->pubsub_channels)+ + listLength(c->pubsub_patterns)); + } + decrRefCount(pattern); + return retval; +} + +/* Unsubscribe from all the channels. Return the number of channels the + * client was subscribed from. */ +static int pubsubUnsubscribeAllChannels(redisClient *c, int notify) { + dictIterator *di = dictGetIterator(c->pubsub_channels); + dictEntry *de; + int count = 0; + + while((de = dictNext(di)) != NULL) { + robj *channel = dictGetEntryKey(de); + + count += pubsubUnsubscribeChannel(c,channel,notify); + } + dictReleaseIterator(di); + return count; +} + +/* Unsubscribe from all the patterns. Return the number of patterns the + * client was subscribed from. */ +static int pubsubUnsubscribeAllPatterns(redisClient *c, int notify) { + listNode *ln; + listIter li; + int count = 0; + + listRewind(c->pubsub_patterns,&li); + while ((ln = listNext(&li)) != NULL) { + robj *pattern = ln->value; + + count += pubsubUnsubscribePattern(c,pattern,notify); + } + return count; +} + +/* Publish a message */ +static int pubsubPublishMessage(robj *channel, robj *message) { + int receivers = 0; + struct dictEntry *de; + listNode *ln; + listIter li; + + /* Send to clients listening for that channel */ + de = dictFind(server.pubsub_channels,channel); + if (de) { + list *list = dictGetEntryVal(de); + listNode *ln; + listIter li; + + listRewind(list,&li); + while ((ln = listNext(&li)) != NULL) { + redisClient *c = ln->value; + + addReply(c,shared.mbulk3); + addReply(c,shared.messagebulk); + addReplyBulk(c,channel); + addReplyBulk(c,message); + receivers++; + } + } + /* Send to clients listening to matching channels */ + if (listLength(server.pubsub_patterns)) { + listRewind(server.pubsub_patterns,&li); + channel = getDecodedObject(channel); + while ((ln = listNext(&li)) != NULL) { + pubsubPattern *pat = ln->value; + + if (stringmatchlen((char*)pat->pattern->ptr, + sdslen(pat->pattern->ptr), + (char*)channel->ptr, + sdslen(channel->ptr),0)) { + addReply(pat->client,shared.mbulk4); + addReply(pat->client,shared.pmessagebulk); + addReplyBulk(pat->client,pat->pattern); + addReplyBulk(pat->client,channel); + addReplyBulk(pat->client,message); + receivers++; + } + } + decrRefCount(channel); + } + return receivers; +} + +static void subscribeCommand(redisClient *c) { + int j; + + for (j = 1; j < c->argc; j++) + pubsubSubscribeChannel(c,c->argv[j]); +} + +static void unsubscribeCommand(redisClient *c) { + if (c->argc == 1) { + pubsubUnsubscribeAllChannels(c,1); + return; + } else { + int j; + + for (j = 1; j < c->argc; j++) + pubsubUnsubscribeChannel(c,c->argv[j],1); + } +} + +static void psubscribeCommand(redisClient *c) { + int j; + + for (j = 1; j < c->argc; j++) + pubsubSubscribePattern(c,c->argv[j]); +} + +static void punsubscribeCommand(redisClient *c) { + if (c->argc == 1) { + pubsubUnsubscribeAllPatterns(c,1); + return; + } else { + int j; + + for (j = 1; j < c->argc; j++) + pubsubUnsubscribePattern(c,c->argv[j],1); + } +} + +static void publishCommand(redisClient *c) { + int receivers = pubsubPublishMessage(c->argv[1],c->argv[2]); + addReplyLongLong(c,receivers); +} + /* ================================= Debugging ============================== */ +/* Compute the sha1 of string at 's' with 'len' bytes long. + * The SHA1 is then xored againt the string pointed by digest. + * Since xor is commutative, this operation is used in order to + * "add" digests relative to unordered elements. + * + * So digest(a,b,c,d) will be the same of digest(b,a,c,d) */ +static void xorDigest(unsigned char *digest, void *ptr, size_t len) { + SHA1_CTX ctx; + unsigned char hash[20], *s = ptr; + int j; + + SHA1Init(&ctx); + SHA1Update(&ctx,s,len); + SHA1Final(hash,&ctx); + + for (j = 0; j < 20; j++) + digest[j] ^= hash[j]; +} + +static void xorObjectDigest(unsigned char *digest, robj *o) { + o = getDecodedObject(o); + xorDigest(digest,o->ptr,sdslen(o->ptr)); + decrRefCount(o); +} + +/* This function instead of just computing the SHA1 and xoring it + * against diget, also perform the digest of "digest" itself and + * replace the old value with the new one. + * + * So the final digest will be: + * + * digest = SHA1(digest xor SHA1(data)) + * + * This function is used every time we want to preserve the order so + * that digest(a,b,c,d) will be different than digest(b,c,d,a) + * + * Also note that mixdigest("foo") followed by mixdigest("bar") + * will lead to a different digest compared to "fo", "obar". + */ +static void mixDigest(unsigned char *digest, void *ptr, size_t len) { + SHA1_CTX ctx; + char *s = ptr; + + xorDigest(digest,s,len); + SHA1Init(&ctx); + SHA1Update(&ctx,digest,20); + SHA1Final(digest,&ctx); +} + +static void mixObjectDigest(unsigned char *digest, robj *o) { + o = getDecodedObject(o); + mixDigest(digest,o->ptr,sdslen(o->ptr)); + decrRefCount(o); +} + +/* Compute the dataset digest. Since keys, sets elements, hashes elements + * are not ordered, we use a trick: every aggregate digest is the xor + * of the digests of their elements. This way the order will not change + * the result. For list instead we use a feedback entering the output digest + * as input in order to ensure that a different ordered list will result in + * a different digest. */ +static void computeDatasetDigest(unsigned char *final) { + unsigned char digest[20]; + char buf[128]; + dictIterator *di = NULL; + dictEntry *de; + int j; + uint32_t aux; + + memset(final,0,20); /* Start with a clean result */ + + for (j = 0; j < server.dbnum; j++) { + redisDb *db = server.db+j; + + if (dictSize(db->dict) == 0) continue; + di = dictGetIterator(db->dict); + + /* hash the DB id, so the same dataset moved in a different + * DB will lead to a different digest */ + aux = htonl(j); + mixDigest(final,&aux,sizeof(aux)); + + /* Iterate this DB writing every entry */ + while((de = dictNext(di)) != NULL) { + robj *key, *o, *kcopy; + time_t expiretime; + + memset(digest,0,20); /* This key-val digest */ + key = dictGetEntryKey(de); + + if (!server.vm_enabled) { + mixObjectDigest(digest,key); + o = dictGetEntryVal(de); + } else { + /* Don't work with the key directly as when VM is active + * this is unsafe: TODO: fix decrRefCount to check if the + * count really reached 0 to avoid this mess */ + kcopy = dupStringObject(key); + mixObjectDigest(digest,kcopy); + o = lookupKeyRead(db,kcopy); + decrRefCount(kcopy); + } + aux = htonl(o->type); + mixDigest(digest,&aux,sizeof(aux)); + expiretime = getExpire(db,key); + + /* Save the key and associated value */ + if (o->type == REDIS_STRING) { + mixObjectDigest(digest,o); + } else if (o->type == REDIS_LIST) { + list *list = o->ptr; + listNode *ln; + listIter li; + + listRewind(list,&li); + while((ln = listNext(&li))) { + robj *eleobj = listNodeValue(ln); + + mixObjectDigest(digest,eleobj); + } + } else if (o->type == REDIS_SET) { + dict *set = o->ptr; + dictIterator *di = dictGetIterator(set); + dictEntry *de; + + while((de = dictNext(di)) != NULL) { + robj *eleobj = dictGetEntryKey(de); + + xorObjectDigest(digest,eleobj); + } + dictReleaseIterator(di); + } else if (o->type == REDIS_ZSET) { + zset *zs = o->ptr; + dictIterator *di = dictGetIterator(zs->dict); + dictEntry *de; + + while((de = dictNext(di)) != NULL) { + robj *eleobj = dictGetEntryKey(de); + double *score = dictGetEntryVal(de); + unsigned char eledigest[20]; + + snprintf(buf,sizeof(buf),"%.17g",*score); + memset(eledigest,0,20); + mixObjectDigest(eledigest,eleobj); + mixDigest(eledigest,buf,strlen(buf)); + xorDigest(digest,eledigest,20); + } + dictReleaseIterator(di); + } else if (o->type == REDIS_HASH) { + hashIterator *hi; + robj *obj; + + hi = hashInitIterator(o); + while (hashNext(hi) != REDIS_ERR) { + unsigned char eledigest[20]; + + memset(eledigest,0,20); + obj = hashCurrent(hi,REDIS_HASH_KEY); + mixObjectDigest(eledigest,obj); + decrRefCount(obj); + obj = hashCurrent(hi,REDIS_HASH_VALUE); + mixObjectDigest(eledigest,obj); + decrRefCount(obj); + xorDigest(digest,eledigest,20); + } + hashReleaseIterator(hi); + } else { + redisPanic("Unknown object type"); + } + /* If the key has an expire, add it to the mix */ + if (expiretime != -1) xorDigest(digest,"!!expire!!",10); + /* We can finally xor the key-val digest to the final digest */ + xorDigest(final,digest,20); + } + dictReleaseIterator(di); + } +} + static void debugCommand(redisClient *c) { if (!strcasecmp(c->argv[1]->ptr,"segfault")) { *((char*)-1) = 'x'; @@ -9294,6 +10584,9 @@ static void debugCommand(redisClient *c) { (void*)key, key->refcount, (unsigned long long) key->vm.page, (unsigned long long) key->vm.usedpages)); } + } else if (!strcasecmp(c->argv[1]->ptr,"swapin") && c->argc == 3) { + lookupKeyRead(c->db,c->argv[2]); + addReply(c,shared.ok); } else if (!strcasecmp(c->argv[1]->ptr,"swapout") && c->argc == 3) { dictEntry *de = dictFind(c->db->dict,c->argv[2]); robj *key, *val; @@ -9323,15 +10616,54 @@ static void debugCommand(redisClient *c) { } else { addReply(c,shared.err); } + } else if (!strcasecmp(c->argv[1]->ptr,"populate") && c->argc == 3) { + long keys, j; + robj *key, *val; + char buf[128]; + + if (getLongFromObjectOrReply(c, c->argv[2], &keys, NULL) != REDIS_OK) + return; + for (j = 0; j < keys; j++) { + snprintf(buf,sizeof(buf),"key:%lu",j); + key = createStringObject(buf,strlen(buf)); + if (lookupKeyRead(c->db,key) != NULL) { + decrRefCount(key); + continue; + } + snprintf(buf,sizeof(buf),"value:%lu",j); + val = createStringObject(buf,strlen(buf)); + dictAdd(c->db->dict,key,val); + } + addReply(c,shared.ok); + } else if (!strcasecmp(c->argv[1]->ptr,"digest") && c->argc == 2) { + unsigned char digest[20]; + sds d = sdsnew("+"); + int j; + + computeDatasetDigest(digest); + for (j = 0; j < 20; j++) + d = sdscatprintf(d, "%02x",digest[j]); + + d = sdscatlen(d,"\r\n",2); + addReplySds(c,d); } else { addReplySds(c,sdsnew( - "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT |SWAPOUT |RELOAD]\r\n")); + "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT |SWAPIN |SWAPOUT |RELOAD]\r\n")); } } static void _redisAssert(char *estr, char *file, int line) { redisLog(REDIS_WARNING,"=== ASSERTION FAILED ==="); - redisLog(REDIS_WARNING,"==> %s:%d '%s' is not true\n",file,line,estr); + redisLog(REDIS_WARNING,"==> %s:%d '%s' is not true",file,line,estr); +#ifdef HAVE_BACKTRACE + redisLog(REDIS_WARNING,"(forcing SIGSEGV in order to print the stack trace)"); + *((char*)-1) = 'x'; +#endif +} + +static void _redisPanic(char *msg, char *file, int line) { + redisLog(REDIS_WARNING,"!!! Software Failure. Press left mouse button to continue"); + redisLog(REDIS_WARNING,"Guru Meditation: %s #%s:%d",msg,file,line); #ifdef HAVE_BACKTRACE redisLog(REDIS_WARNING,"(forcing SIGSEGV in order to print the stack trace)"); *((char*)-1) = 'x'; @@ -9357,7 +10689,7 @@ int linuxOvercommitMemoryValue(void) { 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."); + redisLog(REDIS_WARNING,"WARNING overcommit_memory is set to 0! Background save may fail under low memory condition. 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__ */ @@ -9454,7 +10786,7 @@ static void *getMcontextEip(ucontext_t *uc) { return (void*) uc->uc_mcontext->__ss.__rip; #else return (void*) uc->uc_mcontext->__ss.__eip; - #endif + #endif #elif defined(__i386__) || defined(__X86_64__) || defined(__x86_64__) return (void*) uc->uc_mcontext.gregs[REG_EIP]; /* Linux 32/64 bit */ #elif defined(__ia64__) /* Linux IA64 */ @@ -9479,7 +10811,7 @@ static void segvHandler(int sig, siginfo_t *info, void *secret) { redisLog(REDIS_WARNING, "%s",infostring); /* It's not safe to sdsfree() the returned string under memory * corruption conditions. Let it leak as we are going to abort */ - + trace_size = backtrace(trace, 100); /* overwrite sigaction with caller's address */ if (getMcontextEip(uc) != NULL) { @@ -9501,6 +10833,13 @@ static void segvHandler(int sig, siginfo_t *info, void *secret) { _exit(0); } +static void sigtermHandler(int sig) { + REDIS_NOTUSED(sig); + + redisLog(REDIS_WARNING,"SIGTERM received, scheduling shutting down..."); + server.shutdown_asap = 1; +} + static void setupSigSegvAction(void) { struct sigaction act; @@ -9514,6 +10853,10 @@ static void setupSigSegvAction(void) { sigaction (SIGFPE, &act, NULL); sigaction (SIGILL, &act, NULL); sigaction (SIGBUS, &act, NULL); + + act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND; + act.sa_handler = sigtermHandler; + sigaction (SIGTERM, &act, NULL); return; }