X-Git-Url: https://git.saurik.com/redis.git/blobdiff_plain/8bca8773b4e2542f9537b8403764867aa76273a5..abce5627d7ce3bc1f5d2a15f65af11874bec169a:/redis.c?ds=inline diff --git a/redis.c b/redis.c index 04d8b0be..90fe6f95 100644 --- a/redis.c +++ b/redis.c @@ -27,7 +27,7 @@ * POSSIBILITY OF SUCH DAMAGE. */ -#define REDIS_VERSION "1.3.8" +#define REDIS_VERSION "1.3.10" #include "fmacros.h" #include "config.h" @@ -37,8 +37,6 @@ #include #include #include -#define __USE_POSIX199309 -#define __USE_UNIX98 #include #ifdef HAVE_BACKTRACE @@ -91,7 +89,7 @@ #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 +109,7 @@ 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 +235,9 @@ 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 ============================== */ @@ -341,8 +342,6 @@ 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; @@ -371,6 +370,7 @@ 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; @@ -378,8 +378,8 @@ struct redisServer { char *dbfilename; char *appendfilename; char *requirepass; - int shareobjects; int rdbcompression; + int activerehashing; /* Replication related */ int isslave; char *masterauth; @@ -506,6 +506,7 @@ 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, @@ -513,8 +514,9 @@ struct sharedObjectsStruct { *outofrangeerr, *plus, *select0, *select1, *select2, *select3, *select4, *select5, *select6, *select7, *select8, *select9, - *messagebulk, *subscribebulk, *unsubscribebulk, *mbulk3, - *psubscribebulk, *punsubscribebulk; + *messagebulk, *pmessagebulk, *subscribebulk, *unsubscribebulk, *mbulk3, + *mbulk4, *psubscribebulk, *punsubscribebulk, + *integers[REDIS_SHARED_INTEGERS]; } shared; /* Global vars that are actally used as constants. The following double @@ -555,10 +557,11 @@ 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); @@ -619,13 +622,17 @@ 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 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); @@ -705,7 +712,10 @@ 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); @@ -731,6 +741,7 @@ 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}, @@ -781,8 +792,11 @@ 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}, @@ -829,7 +843,7 @@ static struct redisCommand cmdTable[] = { {"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,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} }; @@ -962,6 +976,77 @@ 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; @@ -1044,8 +1129,8 @@ 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); @@ -1065,7 +1150,7 @@ 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; @@ -1198,16 +1283,28 @@ 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); @@ -1222,7 +1319,7 @@ 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; @@ -1283,7 +1380,8 @@ 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); @@ -1292,6 +1390,19 @@ 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); @@ -1323,15 +1434,17 @@ 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 */ @@ -1349,6 +1462,7 @@ 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 @@ -1430,6 +1544,7 @@ 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; @@ -1441,6 +1556,7 @@ 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; @@ -1465,9 +1581,13 @@ 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")); @@ -1503,11 +1623,17 @@ static void createSharedObjects(void) { 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) { @@ -1534,7 +1660,7 @@ 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 */ @@ -1542,9 +1668,8 @@ 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; @@ -1597,7 +1722,6 @@ 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); @@ -1619,6 +1743,7 @@ static void initServer() { server.bgsavechildpid = -1; server.bgrewritechildpid = -1; server.bgrewritebuf = sdsempty(); + server.aofbuf = sdsempty(); server.lastsave = time(NULL); server.dirty = 0; server.stat_numcommands = 0; @@ -1668,15 +1793,12 @@ 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); } } @@ -1763,7 +1885,7 @@ 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]); @@ -1774,18 +1896,13 @@ 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) { @@ -1822,19 +1939,17 @@ 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; } @@ -1963,7 +2078,7 @@ 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); @@ -2041,7 +2156,7 @@ 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) @@ -2089,7 +2204,7 @@ 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) { @@ -2120,12 +2235,15 @@ 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++; } @@ -2263,15 +2381,9 @@ 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) { @@ -2290,7 +2402,8 @@ static int processCommand(redisClient *c) { } /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */ - if (dictSize(c->pubsub_channels) > 0 && + 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")); @@ -2385,6 +2498,64 @@ 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 @@ -2402,7 +2573,7 @@ again: if (p) { sds query, *argv; int argc, j; - + query = c->querybuf; c->querybuf = sdsempty(); querylen = 1+(p-(query)); @@ -2510,7 +2681,7 @@ static void *dupClientReplyValue(void *o) { } static int listMatchObjects(void *a, void *b) { - return compareStringObjects(a,b) == 0; + return equalStringObjects(a,b); } static redisClient *createClient(int fd) { @@ -2597,6 +2768,21 @@ static void addReplyLong(redisClient *c, long l) { addReplySds(c,sdsnewlen(buf,len)); } +static void addReplyLongLong(redisClient *c, long long ll) { + char buf[128]; + size_t len; + + if (ll == 0) { + addReply(c,shared.czero); + return; + } else if (ll == 1) { + addReply(c,shared.cone); + return; + } + len = snprintf(buf,sizeof(buf),":%lld\r\n",ll); + addReplySds(c,sdsnewlen(buf,len)); +} + static void addReplyUlong(redisClient *c, unsigned long ul) { char buf[128]; size_t len; @@ -2724,6 +2910,23 @@ 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 { + o = createObject(REDIS_STRING, NULL); + if (value >= LONG_MIN && value <= LONG_MAX) { + 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)); @@ -2790,27 +2993,24 @@ 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); @@ -2833,7 +3033,7 @@ 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 || @@ -2923,51 +3123,6 @@ 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). @@ -2978,10 +3133,10 @@ 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 */ @@ -2991,36 +3146,42 @@ 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; @@ -3028,17 +3189,17 @@ 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, @@ -3051,14 +3212,14 @@ 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 { @@ -3067,6 +3228,18 @@ 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) { @@ -3074,8 +3247,99 @@ 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 =========================== */ @@ -3124,7 +3388,7 @@ static int rdbTryIntegerEncoding(char *s, size_t len, unsigned char *enc) { /* 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); + 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 */ @@ -3331,7 +3595,7 @@ static int rdbSaveObject(FILE *fp, robj *o) { dictReleaseIterator(di); } } else { - redisAssert(0); + redisPanic("Unknown object type"); } return 0; } @@ -3350,7 +3614,7 @@ 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; } @@ -3433,7 +3697,7 @@ 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) { @@ -3477,6 +3741,7 @@ 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 */ @@ -3532,7 +3797,11 @@ 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; @@ -3551,9 +3820,12 @@ 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) { @@ -3575,7 +3847,7 @@ err: return NULL; } -static robj *rdbLoadStringObject(FILE*fp) { +static robj *rdbGenericLoadStringObject(FILE*fp, int encode) { int isencoded; uint32_t len; sds val; @@ -3586,11 +3858,11 @@ 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"); } } @@ -3600,7 +3872,15 @@ 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() */ @@ -3629,8 +3909,8 @@ 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; @@ -3645,8 +3925,8 @@ 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 { @@ -3666,8 +3946,8 @@ 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); @@ -3706,26 +3986,26 @@ 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"); @@ -3744,8 +4024,9 @@ 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) { @@ -3767,36 +4048,61 @@ 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 */ @@ -3824,45 +4130,60 @@ static void echoCommand(redisClient *c) { /*=================================== Strings =============================== */ -static void setGenericCommand(redisClient *c, int nx) { +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; + } + } - 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; @@ -3893,7 +4214,7 @@ 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]); @@ -3933,7 +4254,7 @@ 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]); @@ -3960,28 +4281,14 @@ 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 = tryObjectEncoding(o); retval = dictAdd(c->db->dict,c->argv[1],o); if (retval == DICT_ERR) { dictReplace(c->db->dict,c->argv[1],o); @@ -4004,12 +4311,16 @@ 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); } @@ -4027,7 +4338,7 @@ static void appendCommand(redisClient *c) { totlen = stringObjectLen(c->argv[2]); } else { dictEntry *de; - + de = dictFind(c->db->dict,c->argv[1]); assert(de != NULL); @@ -4116,7 +4427,7 @@ static void existsCommand(redisClient *c) { 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 { @@ -4126,18 +4437,25 @@ 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); } } @@ -4246,7 +4564,6 @@ static void shutdownCommand(redisClient *c) { 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 @@ -4254,7 +4571,7 @@ static void shutdownCommand(redisClient *c) { * 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"); + 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")); } @@ -4398,7 +4715,7 @@ 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)); } @@ -4489,8 +4806,8 @@ 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); @@ -4585,7 +4902,7 @@ 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++; @@ -4754,7 +5071,7 @@ 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)); } @@ -4823,7 +5140,7 @@ 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; } @@ -5045,7 +5362,7 @@ 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; @@ -5087,7 +5404,7 @@ 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; @@ -5294,7 +5611,7 @@ 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; } } @@ -5378,7 +5695,7 @@ 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); @@ -5407,14 +5724,14 @@ 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); } @@ -5448,12 +5765,15 @@ 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; @@ -5466,13 +5786,16 @@ static void zremrangebyscoreCommand(redisClient *c) { } 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; @@ -5526,7 +5849,7 @@ 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"); } } @@ -5579,7 +5902,8 @@ 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--; @@ -5701,8 +6025,8 @@ static void zinterCommand(redisClient *c) { 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; @@ -5711,6 +6035,9 @@ 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) { @@ -5718,8 +6045,8 @@ 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; @@ -5825,7 +6152,7 @@ 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); @@ -5962,295 +6289,389 @@ 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. */ - if (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) { - long long value = 0, incr = 0; - robj *o = lookupKeyWrite(c->db,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 (dictFind(o->ptr,key) != NULL) { + return 1; + } + } + return 0; +} + +/* 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) { + 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 { + if (dictReplace(o->ptr,key,value)) { + /* Insert */ + incrRefCount(key); + } else { + /* Update */ + update = 1; + } + incrRefCount(value); + } + return update; +} + +/* 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; +} + +/* 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); +} + +/* 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; + + 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; +} + +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 { + o = createStringObject((char*)hi->zv,hi->zvlen); + } + } else { + if (what & REDIS_HASH_KEY) { + o = dictGetEntryKey(hi->de); + } else { + o = dictGetEntryVal(hi->de); + } + incrRefCount(o); + } + return o; +} + +static robj *hashLookupWriteOrCreate(redisClient *c, robj *key) { + robj *o = lookupKeyWrite(c->db,key); if (o == NULL) { o = createHashObject(); - dictAdd(c->db->dict,c->argv[1],o); - incrRefCount(c->argv[1]); + dictAdd(c->db->dict,key,o); + incrRefCount(key); } else { if (o->type != REDIS_HASH) { addReply(c,shared.wrongtypeerr); - return; + return NULL; } } + return o; +} - robj *o_incr = getDecodedObject(c->argv[3]); - incr = strtoll(o_incr->ptr, NULL, 10); - decrRefCount(o_incr); +/* ============================= Hash commands ============================== */ +static void hsetCommand(redisClient *c) { + int update; + robj *o; - 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),NULL); - sdsfree(svalue); - o->ptr = zm; - - /* Check if the zipmap needs to be converted. */ - if (zipmapLen(zm) > server.hash_max_zipmap_entries) - convertToRealHash(o); + 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++; +} + +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 { - robj *hval; - dictEntry *de; + hashTryObjectEncoding(o,&c->argv[2], &c->argv[3]); + hashSet(o,c->argv[2],c->argv[3]); + addReply(c, shared.cone); + server.dirty++; + } +} - /* 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); - } +static void hmsetCommand(redisClient *c) { + int i; + robj *o; + + if ((c->argc % 2) == 1) { + addReplySds(c,sdsnew("-ERR wrong number of arguments for HMSET\r\n")); + return; + } + + 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++; +} - value += incr; - hval = createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",value)); - tryObjectEncoding(hval); - if (dictReplace(o->ptr,c->argv[2],hval)) { - incrRefCount(c->argv[2]); +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; } + value += incr; + new = createStringObjectFromLongLong(value); + hashTryObjectEncoding(o,&c->argv[2],NULL); + hashSet(o,c->argv[2],new); + decrRefCount(new); + addReplyLongLong(c,value); server.dirty++; - addReplyLong(c, value); } static void hgetCommand(redisClient *c) { - robj *o; - + robj *o, *value; if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullbulk)) == NULL || checkType(c,o,REDIS_HASH)) return; - 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; - } else { - addReply(c,shared.nullbulk); - decrRefCount(field); - return; - } + if ((value = hashGet(o,c->argv[2])) != NULL) { + addReplyBulk(c,value); + decrRefCount(value); } else { - struct dictEntry *de; + addReply(c,shared.nullbulk); + } +} - de = dictFind(o->ptr,c->argv[2]); - if (de == NULL) { - addReply(c,shared.nullbulk); - } else { - robj *e = dictGetEntryVal(de); +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); + } - addReplyBulk(c,e); + /* 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; - int deleted = 0; - if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,o,REDIS_HASH)) return; - 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]); + if (hashDelete(o,c->argv[2])) { + if (hashLength(o) == 0) deleteKey(c->db,c->argv[1]); + addReply(c,shared.cone); + server.dirty++; } 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]); + addReply(c,shared.czero); } - if (deleted) server.dirty++; - addReply(c,deleted ? shared.cone : shared.czero); } static void hlenCommand(redisClient *c) { robj *o; - unsigned long len; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || checkType(c,o,REDIS_HASH)) return; - len = (o->encoding == REDIS_ENCODING_ZIPMAP) ? - zipmapLen((unsigned char*)o->ptr) : dictSize((dict*)o->ptr); - addReplyUlong(c,len); + addReplyUlong(c,hashLength(o)); } -#define REDIS_GETALL_KEYS 1 -#define REDIS_GETALL_VALS 2 static void genericHgetallCommand(redisClient *c, int flags) { - robj *o, *lenobj; + robj *o, *lenobj, *obj; unsigned long count = 0; + hashIterator *hi; - if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.nullmultibulk)) == NULL + 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); - 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++; - } + 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++; } - } else { - dictIterator *di = dictGetIterator(o->ptr); - dictEntry *de; - - while((de = dictNext(di)) != NULL) { - robj *fieldobj = dictGetEntryKey(de); - robj *valobj = dictGetEntryVal(de); - - if (flags & REDIS_GETALL_KEYS) { - addReplyBulk(c,fieldobj); - count++; - } - if (flags & REDIS_GETALL_VALS) { - addReplyBulk(c,valobj); - 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) { @@ -6265,8 +6686,8 @@ 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; @@ -6302,23 +6723,26 @@ 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; } @@ -6335,20 +6759,47 @@ 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 @@ -6383,14 +6834,8 @@ 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; @@ -6411,7 +6856,7 @@ 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 && @@ -6473,7 +6918,7 @@ 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; @@ -6517,38 +6962,36 @@ 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); + } } } @@ -6590,10 +7033,11 @@ 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 */ @@ -6620,11 +7064,14 @@ 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 */ @@ -6645,7 +7092,7 @@ 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); @@ -6868,15 +7315,20 @@ 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; @@ -6893,11 +7345,11 @@ 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) { @@ -6969,6 +7421,20 @@ 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; @@ -6979,6 +7445,13 @@ 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)); @@ -6992,6 +7465,10 @@ 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 ========================= */ @@ -7141,7 +7618,7 @@ 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; @@ -7378,7 +7855,7 @@ 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"); @@ -7614,11 +8091,55 @@ 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 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 @@ -7664,23 +8185,11 @@ static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv decrRefCount(argv[j]); } - /* We want to perform a single write. This should be guaranteed atomic - * at least if the filesystem we are writing is a real physical one. - * While this will save us against the server being killed I don't think - * there is much to do about the whole server stopping for power problems - * or alike */ - nwritten = write(server.appendfd,buf,sdslen(buf)); - if (nwritten != (signed)sdslen(buf)) { - /* Ooops, we are in troubles. The best thing to do for now is - * to simply exit instead to give the illusion that everything is - * working as expected. */ - if (nwritten == -1) { - redisLog(REDIS_WARNING,"Exiting on error writing to the append-only file: %s",strerror(errno)); - } else { - redisLog(REDIS_WARNING,"Exiting on short write while writing to the append-only file: %s",strerror(errno)); - } - exit(1); - } + /* 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)); + /* 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 @@ -7689,14 +8198,6 @@ 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 @@ -7716,12 +8217,14 @@ 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); } @@ -7733,6 +8236,7 @@ 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; @@ -7742,6 +8246,10 @@ 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; @@ -7776,14 +8284,9 @@ 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; @@ -7802,8 +8305,14 @@ 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: @@ -8020,7 +8529,7 @@ static int rewriteAppendOnlyFile(char *filename) { dictReleaseIterator(di); } } else { - redisAssert(0); + redisPanic("Unknown object type"); } /* Save the expire time */ if (expiretime != -1) { @@ -8040,7 +8549,7 @@ 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) { @@ -8099,6 +8608,7 @@ 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 @@ -8152,42 +8662,38 @@ static void aofRemoveTempFile(pid_t childpid) { /* =================== Virtual Memory - Blocking Side ====================== */ -/* 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; -} - 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; @@ -8282,7 +8788,7 @@ 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 @@ -8293,7 +8799,7 @@ 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. * @@ -8666,7 +9172,10 @@ 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); } @@ -8839,7 +9348,7 @@ 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 @@ -8944,7 +9453,7 @@ 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); } @@ -9022,14 +9531,14 @@ 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; @@ -9066,7 +9575,7 @@ 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. @@ -9097,7 +9606,7 @@ 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; @@ -9166,7 +9675,7 @@ 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; } @@ -9226,6 +9735,50 @@ static void configSetCommand(redisClient *c) { server.masterauth = zstrdup(o->ptr); } else if (!strcasecmp(c->argv[2]->ptr,"maxmemory")) { server.maxmemory = strtoll(o->ptr, NULL, 10); + } 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,"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", @@ -9235,6 +9788,14 @@ 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) { @@ -9269,6 +9830,35 @@ static void configGetCommand(redisClient *c) { addReplyBulkCString(c,buf); 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); } @@ -9312,7 +9902,7 @@ static int listMatchPubsubPattern(void *a, void *b) { pubsubPattern *pa = a, *pb = b; return (pa->client == pb->client) && - (compareStringObjects(pa->pattern,pb->pattern) == 0); + (equalStringObjects(pa->pattern,pb->pattern)); } /* Subscribe a client to a channel. Returns 1 if the operation succeeded, or @@ -9503,8 +10093,9 @@ static int pubsubPublishMessage(robj *channel, robj *message) { sdslen(pat->pattern->ptr), (char*)channel->ptr, sdslen(channel->ptr),0)) { - addReply(pat->client,shared.mbulk3); - addReply(pat->client,shared.messagebulk); + addReply(pat->client,shared.mbulk4); + addReply(pat->client,shared.pmessagebulk); + addReplyBulk(pat->client,pat->pattern); addReplyBulk(pat->client,channel); addReplyBulk(pat->client,message); receivers++; @@ -9616,6 +10207,9 @@ 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; @@ -9645,9 +10239,28 @@ 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 { 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")); } } @@ -9660,6 +10273,15 @@ static void _redisAssert(char *estr, char *file, int line) { #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'; +#endif +} + /* =================================== Main! ================================ */ #ifdef __linux__ @@ -9679,7 +10301,7 @@ 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__ */ @@ -9776,7 +10398,7 @@ 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 */ @@ -9801,7 +10423,7 @@ 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) {