* POSSIBILITY OF SUCH DAMAGE.
*/
-#define REDIS_VERSION "0.091"
+#define REDIS_VERSION "1.050"
#include "fmacros.h"
+#include "config.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <unistd.h>
+#define __USE_POSIX199309
#include <signal.h>
+
+#ifdef HAVE_BACKTRACE
+#include <execinfo.h>
+#include <ucontext.h>
+#endif /* HAVE_BACKTRACE */
+
#include <sys/wait.h>
#include <errno.h>
#include <assert.h>
#include <sys/time.h>
#include <sys/resource.h>
#include <limits.h>
+#include <math.h>
+
+#if defined(__sun)
+#include "solarisfixes.h"
+#endif
+#include "redis.h"
#include "ae.h" /* Event driven programming library */
#include "sds.h" /* Dynamic safe strings */
#include "anet.h" /* Networking the easy way */
#include "dict.h" /* Hash tables */
#include "adlist.h" /* Linked lists */
#include "zmalloc.h" /* total memory usage aware version of malloc/free */
-#include "lzf.h"
+#include "lzf.h" /* LZF compression library */
+#include "pqsort.h" /* Partial qsort for SORT+LIMIT */
/* Error codes */
#define REDIS_OK 0
#define REDIS_MAXIDLETIME (60*5) /* default client timeout */
#define REDIS_IOBUF_LEN 1024
#define REDIS_LOADBUF_LEN 1024
-#define REDIS_MAX_ARGS 16
+#define REDIS_STATIC_ARGS 4
#define REDIS_DEFAULT_DBNUM 16
#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 100 /* try to expire 100 keys/second */
+#define REDIS_MAX_WRITE_PER_EVENT (1024*64)
+#define REDIS_REQUEST_MAX_SIZE (1024*1024*256) /* max bytes in inline command */
/* Hash table parameters */
#define REDIS_HT_MINFILL 10 /* Minimal hash table fill 10% */
-#define REDIS_HT_MINSLOTS 16384 /* Never resize the HT under this */
/* Command flags */
-#define REDIS_CMD_BULK 1
-#define REDIS_CMD_INLINE 2
+#define REDIS_CMD_BULK 1 /* Bulk write command */
+#define REDIS_CMD_INLINE 2 /* Inline command */
+/* REDIS_CMD_DENYOOM reserves a longer comment: all the commands marked with
+ this flags will return an error when the 'maxmemory' option is set in the
+ 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
/* Object types */
#define REDIS_STRING 0
#define REDIS_LIST 1
#define REDIS_SET 2
-#define REDIS_HASH 3
+#define REDIS_ZSET 3
+#define REDIS_HASH 4
+
+/* Objects encoding */
+#define REDIS_ENCODING_RAW 0 /* Raw representation */
+#define REDIS_ENCODING_INT 1 /* Encoded as integer */
/* Object types only used for dumping to disk */
#define REDIS_EXPIRETIME 253
/* Sort operations */
#define REDIS_SORT_GET 0
-#define REDIS_SORT_DEL 1
-#define REDIS_SORT_INCR 2
-#define REDIS_SORT_DECR 3
-#define REDIS_SORT_ASC 4
-#define REDIS_SORT_DESC 5
+#define REDIS_SORT_ASC 1
+#define REDIS_SORT_DESC 2
#define REDIS_SORTKEY_MAX 1024
/* Log levels */
/* Anti-warning macro... */
#define REDIS_NOTUSED(V) ((void) V)
+#define ZSKIPLIST_MAXLEVEL 32 /* Should be enough for 2^32 elements */
+#define ZSKIPLIST_P 0.25 /* Skiplist P = 1/4 */
+
+/* Append only defines */
+#define APPENDFSYNC_NO 0
+#define APPENDFSYNC_ALWAYS 1
+#define APPENDFSYNC_EVERYSEC 2
+
/*================================= Data types ============================== */
/* A redis object, that is a type able to hold a string / list / set */
typedef struct redisObject {
void *ptr;
- int type;
+ unsigned char type;
+ unsigned char encoding;
+ unsigned char notused[2];
int refcount;
} robj;
redisDb *db;
int dictid;
sds querybuf;
- robj *argv[REDIS_MAX_ARGS];
- int argc;
+ robj **argv, **mbargv;
+ int argc, mbargc;
int bulklen; /* bulk read len. -1 if not in bulk read mode */
+ int multibulk; /* multi bulk command format active */
list *reply;
int sentlen;
time_t lastinteraction; /* time of the last interaction, used for timeout */
int maxidletime;
int dbnum;
int daemonize;
+ int appendonly;
+ int appendfsync;
+ time_t lastfsync;
+ int appendfd;
+ int appendseldb;
char *pidfile;
int bgsaveinprogress;
+ pid_t bgsavechildpid;
struct saveparam *saveparams;
int saveparamslen;
char *logfile;
char *bindaddr;
char *dbfilename;
+ char *appendfilename;
char *requirepass;
int shareobjects;
/* Replication related */
int isslave;
+ char *masterauth;
char *masterhost;
int masterport;
redisClient *master; /* client that is master for this slave */
int replstate;
+ unsigned int maxclients;
+ unsigned long maxmemory;
/* Sort parameters - qsort_r() is only available under BSD so we
* have to take this state global, in order to pass it to sortCompare() */
int sort_desc;
int flags;
};
+struct redisFunctionSym {
+ char *name;
+ unsigned long pointer;
+};
+
typedef struct _redisSortObject {
robj *obj;
union {
robj *pattern;
} redisSortOperation;
+/* ZSETs use a specialized version of Skiplists */
+
+typedef struct zskiplistNode {
+ struct zskiplistNode **forward;
+ struct zskiplistNode *backward;
+ double score;
+ robj *obj;
+} zskiplistNode;
+
+typedef struct zskiplist {
+ struct zskiplistNode *header, *tail;
+ unsigned long length;
+ int level;
+} zskiplist;
+
+typedef struct zset {
+ dict *dict;
+ zskiplist *zsl;
+} zset;
+
+/* Our shared "common" objects */
+
struct sharedObjectsStruct {
robj *crlf, *ok, *err, *emptybulk, *czero, *cone, *pong, *space,
*colon, *nullbulk, *nullmultibulk,
*select5, *select6, *select7, *select8, *select9;
} shared;
+/* Global vars that are actally used as constants. The following double
+ * values are used for double on-disk serialization, and are initialized
+ * at runtime to avoid strange compiler optimizations. */
+
+static double R_Zero, R_PosInf, R_NegInf, R_Nan;
+
/*================================ Prototypes =============================== */
static void freeStringObject(robj *o);
static int rdbSaveBackground(char *filename);
static robj *createStringObject(char *ptr, size_t len);
static void replicationFeedSlaves(list *slaves, struct redisCommand *cmd, int dictid, robj **argv, int argc);
+static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc);
static int syncWithMaster(void);
static robj *tryObjectSharing(robj *o);
+static int tryObjectEncoding(robj *o);
+static robj *getDecodedObject(const robj *o);
static int removeExpire(redisDb *db, robj *key);
static int expireIfNeeded(redisDb *db, robj *key);
static int deleteIfVolatile(redisDb *db, robj *key);
static int deleteKey(redisDb *db, robj *key);
static time_t getExpire(redisDb *db, robj *key);
static int setExpire(redisDb *db, robj *key, time_t when);
-static void updateSalvesWaitingBgsave(int bgsaveerr);
+static void updateSlavesWaitingBgsave(int bgsaveerr);
+static void freeMemoryIfNeeded(void);
+static int processCommand(redisClient *c);
+static void setupSigSegvAction(void);
+static void rdbRemoveTempFile(pid_t childpid);
+static size_t stringObjectLen(robj *o);
+static void processInputBuffer(redisClient *c);
+static zskiplist *zslCreate(void);
+static void zslFree(zskiplist *zsl);
+static void zslInsert(zskiplist *zsl, double score, robj *obj);
static void authCommand(redisClient *c);
static void pingCommand(redisClient *c);
static void smoveCommand(redisClient *c);
static void sismemberCommand(redisClient *c);
static void scardCommand(redisClient *c);
+static void spopCommand(redisClient *c);
+static void srandmemberCommand(redisClient *c);
static void sinterCommand(redisClient *c);
static void sinterstoreCommand(redisClient *c);
static void sunionCommand(redisClient *c);
static void sunionstoreCommand(redisClient *c);
+static void sdiffCommand(redisClient *c);
+static void sdiffstoreCommand(redisClient *c);
static void syncCommand(redisClient *c);
static void flushdbCommand(redisClient *c);
static void flushallCommand(redisClient *c);
static void sortCommand(redisClient *c);
static void lremCommand(redisClient *c);
+static void rpoplpushcommand(redisClient *c);
static void infoCommand(redisClient *c);
static void mgetCommand(redisClient *c);
static void monitorCommand(redisClient *c);
static void expireCommand(redisClient *c);
-static void getSetCommand(redisClient *c);
+static void expireatCommand(redisClient *c);
+static void getsetCommand(redisClient *c);
+static void ttlCommand(redisClient *c);
+static void slaveofCommand(redisClient *c);
+static void debugCommand(redisClient *c);
+static void msetCommand(redisClient *c);
+static void msetnxCommand(redisClient *c);
+static void zaddCommand(redisClient *c);
+static void zrangeCommand(redisClient *c);
+static void zrangebyscoreCommand(redisClient *c);
+static void zrevrangeCommand(redisClient *c);
+static void zcardCommand(redisClient *c);
+static void zremCommand(redisClient *c);
+static void zscoreCommand(redisClient *c);
+static void zremrangebyscoreCommand(redisClient *c);
/*================================= Globals ================================= */
static struct redisServer server; /* server global state */
static struct redisCommand cmdTable[] = {
{"get",getCommand,2,REDIS_CMD_INLINE},
- {"set",setCommand,3,REDIS_CMD_BULK},
- {"setnx",setnxCommand,3,REDIS_CMD_BULK},
- {"del",delCommand,2,REDIS_CMD_INLINE},
+ {"set",setCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
+ {"setnx",setnxCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
+ {"del",delCommand,-2,REDIS_CMD_INLINE},
{"exists",existsCommand,2,REDIS_CMD_INLINE},
- {"incr",incrCommand,2,REDIS_CMD_INLINE},
- {"decr",decrCommand,2,REDIS_CMD_INLINE},
+ {"incr",incrCommand,2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
+ {"decr",decrCommand,2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
{"mget",mgetCommand,-2,REDIS_CMD_INLINE},
- {"rpush",rpushCommand,3,REDIS_CMD_BULK},
- {"lpush",lpushCommand,3,REDIS_CMD_BULK},
+ {"rpush",rpushCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
+ {"lpush",lpushCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
{"rpop",rpopCommand,2,REDIS_CMD_INLINE},
{"lpop",lpopCommand,2,REDIS_CMD_INLINE},
{"llen",llenCommand,2,REDIS_CMD_INLINE},
{"lindex",lindexCommand,3,REDIS_CMD_INLINE},
- {"lset",lsetCommand,4,REDIS_CMD_BULK},
+ {"lset",lsetCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
{"lrange",lrangeCommand,4,REDIS_CMD_INLINE},
{"ltrim",ltrimCommand,4,REDIS_CMD_INLINE},
{"lrem",lremCommand,4,REDIS_CMD_BULK},
- {"sadd",saddCommand,3,REDIS_CMD_BULK},
+ {"rpoplpush",rpoplpushcommand,3,REDIS_CMD_BULK},
+ {"sadd",saddCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
{"srem",sremCommand,3,REDIS_CMD_BULK},
{"smove",smoveCommand,4,REDIS_CMD_BULK},
{"sismember",sismemberCommand,3,REDIS_CMD_BULK},
{"scard",scardCommand,2,REDIS_CMD_INLINE},
- {"sinter",sinterCommand,-2,REDIS_CMD_INLINE},
- {"sinterstore",sinterstoreCommand,-3,REDIS_CMD_INLINE},
- {"sunion",sunionCommand,-2,REDIS_CMD_INLINE},
- {"sunionstore",sunionstoreCommand,-3,REDIS_CMD_INLINE},
+ {"spop",spopCommand,2,REDIS_CMD_INLINE},
+ {"srandmember",srandmemberCommand,2,REDIS_CMD_INLINE},
+ {"sinter",sinterCommand,-2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
+ {"sinterstore",sinterstoreCommand,-3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
+ {"sunion",sunionCommand,-2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
+ {"sunionstore",sunionstoreCommand,-3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
+ {"sdiff",sdiffCommand,-2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
+ {"sdiffstore",sdiffstoreCommand,-3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
{"smembers",sinterCommand,2,REDIS_CMD_INLINE},
- {"incrby",incrbyCommand,3,REDIS_CMD_INLINE},
- {"decrby",decrbyCommand,3,REDIS_CMD_INLINE},
- {"getset",getSetCommand,3,REDIS_CMD_BULK},
+ {"zadd",zaddCommand,4,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
+ {"zrem",zremCommand,3,REDIS_CMD_BULK},
+ {"zremrangebyscore",zremrangebyscoreCommand,4,REDIS_CMD_INLINE},
+ {"zrange",zrangeCommand,4,REDIS_CMD_INLINE},
+ {"zrangebyscore",zrangebyscoreCommand,4,REDIS_CMD_INLINE},
+ {"zrevrange",zrevrangeCommand,4,REDIS_CMD_INLINE},
+ {"zcard",zcardCommand,2,REDIS_CMD_INLINE},
+ {"zscore",zscoreCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
+ {"incrby",incrbyCommand,3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
+ {"decrby",decrbyCommand,3,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
+ {"getset",getsetCommand,3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
+ {"mset",msetCommand,-3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
+ {"msetnx",msetnxCommand,-3,REDIS_CMD_BULK|REDIS_CMD_DENYOOM},
{"randomkey",randomkeyCommand,1,REDIS_CMD_INLINE},
{"select",selectCommand,2,REDIS_CMD_INLINE},
{"move",moveCommand,3,REDIS_CMD_INLINE},
{"rename",renameCommand,3,REDIS_CMD_INLINE},
{"renamenx",renamenxCommand,3,REDIS_CMD_INLINE},
+ {"expire",expireCommand,3,REDIS_CMD_INLINE},
+ {"expireat",expireatCommand,3,REDIS_CMD_INLINE},
{"keys",keysCommand,2,REDIS_CMD_INLINE},
{"dbsize",dbsizeCommand,1,REDIS_CMD_INLINE},
{"auth",authCommand,2,REDIS_CMD_INLINE},
{"sync",syncCommand,1,REDIS_CMD_INLINE},
{"flushdb",flushdbCommand,1,REDIS_CMD_INLINE},
{"flushall",flushallCommand,1,REDIS_CMD_INLINE},
- {"sort",sortCommand,-2,REDIS_CMD_INLINE},
+ {"sort",sortCommand,-2,REDIS_CMD_INLINE|REDIS_CMD_DENYOOM},
{"info",infoCommand,1,REDIS_CMD_INLINE},
{"monitor",monitorCommand,1,REDIS_CMD_INLINE},
- {"expire",expireCommand,3,REDIS_CMD_INLINE},
+ {"ttl",ttlCommand,2,REDIS_CMD_INLINE},
+ {"slaveof",slaveofCommand,3,REDIS_CMD_INLINE},
+ {"debug",debugCommand,-2,REDIS_CMD_INLINE},
{NULL,NULL,0,0}
};
return 0;
}
-void redisLog(int level, const char *fmt, ...)
-{
+static void redisLog(int level, const char *fmt, ...) {
va_list ap;
FILE *fp;
va_start(ap, fmt);
if (level >= server.verbosity) {
char *c = ".-*";
- fprintf(fp,"%c ",c[level]);
+ char buf[64];
+ time_t now;
+
+ now = time(NULL);
+ strftime(buf,64,"%d %b %H:%M:%S",localtime(&now));
+ fprintf(fp,"%s %c ",buf,c[level]);
vfprintf(fp, fmt, ap);
fprintf(fp,"\n");
fflush(fp);
* keys and radis objects as values (objects can hold SDS strings,
* lists, sets). */
+static void dictVanillaFree(void *privdata, void *val)
+{
+ DICT_NOTUSED(privdata);
+ zfree(val);
+}
+
static int sdsDictKeyCompare(void *privdata, const void *key1,
const void *key2)
{
decrRefCount(val);
}
-static int dictSdsKeyCompare(void *privdata, const void *key1,
+static int dictObjKeyCompare(void *privdata, const void *key1,
const void *key2)
{
const robj *o1 = key1, *o2 = key2;
return sdsDictKeyCompare(privdata,o1->ptr,o2->ptr);
}
-static unsigned int dictSdsHash(const void *key) {
+static unsigned int dictObjHash(const void *key) {
const robj *o = key;
return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
}
+static int dictEncObjKeyCompare(void *privdata, const void *key1,
+ const void *key2)
+{
+ const robj *o1 = key1, *o2 = key2;
+
+ if (o1->encoding == REDIS_ENCODING_RAW &&
+ o2->encoding == REDIS_ENCODING_RAW)
+ return sdsDictKeyCompare(privdata,o1->ptr,o2->ptr);
+ else {
+ robj *dec1, *dec2;
+ int cmp;
+
+ dec1 = o1->encoding != REDIS_ENCODING_RAW ?
+ getDecodedObject(o1) : (robj*)o1;
+ dec2 = o2->encoding != REDIS_ENCODING_RAW ?
+ getDecodedObject(o2) : (robj*)o2;
+ cmp = sdsDictKeyCompare(privdata,dec1->ptr,dec2->ptr);
+ if (dec1 != o1) decrRefCount(dec1);
+ if (dec2 != o2) decrRefCount(dec2);
+ return cmp;
+ }
+}
+
+static unsigned int dictEncObjHash(const void *key) {
+ const robj *o = key;
+
+ if (o->encoding == REDIS_ENCODING_RAW)
+ return dictGenHashFunction(o->ptr, sdslen((sds)o->ptr));
+ else {
+ robj *dec = getDecodedObject(o);
+ unsigned int hash = dictGenHashFunction(dec->ptr, sdslen((sds)dec->ptr));
+ decrRefCount(dec);
+ return hash;
+ }
+}
+
static dictType setDictType = {
- dictSdsHash, /* hash function */
+ dictEncObjHash, /* hash function */
NULL, /* key dup */
NULL, /* val dup */
- dictSdsKeyCompare, /* key compare */
+ dictEncObjKeyCompare, /* key compare */
dictRedisObjectDestructor, /* key destructor */
NULL /* val destructor */
};
+static dictType zsetDictType = {
+ dictEncObjHash, /* hash function */
+ NULL, /* key dup */
+ NULL, /* val dup */
+ dictEncObjKeyCompare, /* key compare */
+ dictRedisObjectDestructor, /* key destructor */
+ dictVanillaFree /* val destructor */
+};
+
static dictType hashDictType = {
- dictSdsHash, /* hash function */
+ dictObjHash, /* hash function */
NULL, /* key dup */
NULL, /* val dup */
- dictSdsKeyCompare, /* key compare */
+ dictObjKeyCompare, /* key compare */
dictRedisObjectDestructor, /* key destructor */
dictRedisObjectDestructor /* val destructor */
};
}
/* ====================== Redis server networking stuff ===================== */
-void closeTimedoutClients(void) {
+static void closeTimedoutClients(void) {
redisClient *c;
listNode *ln;
time_t now = time(NULL);
while ((ln = listYield(server.clients)) != NULL) {
c = listNodeValue(ln);
if (!(c->flags & REDIS_SLAVE) && /* no timeout for slaves */
+ !(c->flags & REDIS_MASTER) && /* no timeout for masters */
(now - c->lastinteraction > server.maxidletime)) {
redisLog(REDIS_DEBUG,"Closing idle client");
freeClient(c);
}
}
-int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
+static int htNeedsResize(dict *dict) {
+ long long size, used;
+
+ size = dictSlots(dict);
+ used = dictSize(dict);
+ return (size && used && size > DICT_HT_INITIAL_SIZE &&
+ (used*100/size < REDIS_HT_MINFILL));
+}
+
+/* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
+ * we resize the hash table to save memory */
+static void tryResizeHashTables(void) {
+ int j;
+
+ for (j = 0; j < server.dbnum; j++) {
+ if (htNeedsResize(server.db[j].dict)) {
+ redisLog(REDIS_DEBUG,"The hash table %d is too sparse, resize it...",j);
+ dictResize(server.db[j].dict);
+ redisLog(REDIS_DEBUG,"Hash table %d resized.",j);
+ }
+ if (htNeedsResize(server.db[j].expires))
+ dictResize(server.db[j].expires);
+ }
+}
+
+static int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
int j, loops = server.cronloops++;
REDIS_NOTUSED(eventLoop);
REDIS_NOTUSED(id);
/* Update the global state with the amount of used memory */
server.usedmemory = zmalloc_used_memory();
- /* If the percentage of used slots in the HT reaches REDIS_HT_MINFILL
- * we resize the hash table to save memory */
+ /* Show some info about non-empty databases */
for (j = 0; j < server.dbnum; j++) {
- int size, used, vkeys;
+ long long size, used, vkeys;
size = dictSlots(server.db[j].dict);
used = dictSize(server.db[j].dict);
vkeys = dictSize(server.db[j].expires);
- if (!(loops % 5) && used > 0) {
- redisLog(REDIS_DEBUG,"DB %d: %d keys (%d volatile) in %d slots HT.",j,used,vkeys,size);
+ if (!(loops % 5) && (used || vkeys)) {
+ redisLog(REDIS_DEBUG,"DB %d: %lld keys (%lld volatile) in %lld slots HT.",j,used,vkeys,size);
/* dictPrintStats(server.dict); */
}
- if (size && used && size > REDIS_HT_MINSLOTS &&
- (used*100/size < REDIS_HT_MINFILL)) {
- redisLog(REDIS_NOTICE,"The hash table %d is too sparse, resize it...",j);
- dictResize(server.db[j].dict);
- redisLog(REDIS_NOTICE,"Hash table %d resized.",j);
- }
}
+ /* We don't want to resize the hash tables while a bacground saving
+ * is in progress: the saving child is created using fork() that is
+ * implemented with a copy-on-write semantic in most modern systems, so
+ * if we resize the HT while there is the saving child at work actually
+ * a lot of memory movements in the parent will cause a lot of pages
+ * copied. */
+ if (!server.bgsaveinprogress) tryResizeHashTables();
+
/* Show information about connected clients */
if (!(loops % 5)) {
- redisLog(REDIS_DEBUG,"%d clients connected (%d slaves), %zu bytes in use",
+ redisLog(REDIS_DEBUG,"%d clients connected (%d slaves), %zu bytes in use, %d shared objects",
listLength(server.clients)-listLength(server.slaves),
listLength(server.slaves),
server.usedmemory,
}
/* Close connections of timedout clients */
- if (!(loops % 10))
+ if (server.maxidletime && !(loops % 10))
closeTimedoutClients();
/* Check if a background saving in progress terminated */
if (server.bgsaveinprogress) {
int statloc;
- /* XXX: TODO handle the case of the saving child killed */
if (wait4(-1,&statloc,WNOHANG,NULL)) {
int exitcode = WEXITSTATUS(statloc);
- if (exitcode == 0) {
+ int bysignal = WIFSIGNALED(statloc);
+
+ if (!bysignal && exitcode == 0) {
redisLog(REDIS_NOTICE,
"Background saving terminated with success");
server.dirty = 0;
server.lastsave = time(NULL);
+ } else if (!bysignal && exitcode != 0) {
+ redisLog(REDIS_WARNING, "Background saving error");
} else {
redisLog(REDIS_WARNING,
- "Background saving error");
+ "Background saving terminated by signal");
+ rdbRemoveTempFile(server.bgsavechildpid);
}
server.bgsaveinprogress = 0;
- updateSalvesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR);
+ server.bgsavechildpid = -1;
+ updateSlavesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR);
}
} else {
/* If there is not a background saving in progress check if
}
}
- /* Try to expire a few timed out keys */
+ /* Try to expire a few timed out keys. The algorithm used is adaptive and
+ * will use few CPU cycles if there are few expiring keys, otherwise
+ * it will get more aggressive to avoid that too much memory is used by
+ * keys that can be removed from the keyspace. */
for (j = 0; j < server.dbnum; j++) {
+ int expired;
redisDb *db = server.db+j;
- int num = dictSize(db->expires);
- if (num) {
+ /* Continue to expire if at the end of the cycle more than 25%
+ * of the keys were expired. */
+ do {
+ int num = dictSize(db->expires);
time_t now = time(NULL);
+ expired = 0;
if (num > REDIS_EXPIRELOOKUPS_PER_CRON)
num = REDIS_EXPIRELOOKUPS_PER_CRON;
while (num--) {
t = (time_t) dictGetEntryVal(de);
if (now > t) {
deleteKey(db,dictGetEntryKey(de));
+ expired++;
}
}
- }
+ } while (expired > REDIS_EXPIRELOOKUPS_PER_CRON/4);
}
/* Check if we should connect to a MASTER */
static void appendServerSaveParams(time_t seconds, int changes) {
server.saveparams = zrealloc(server.saveparams,sizeof(struct saveparam)*(server.saveparamslen+1));
- if (server.saveparams == NULL) oom("appendServerSaveParams");
server.saveparams[server.saveparamslen].seconds = seconds;
server.saveparams[server.saveparamslen].changes = changes;
server.saveparamslen++;
}
-static void ResetServerSaveParams() {
+static void resetServerSaveParams() {
zfree(server.saveparams);
server.saveparams = NULL;
server.saveparamslen = 0;
server.bindaddr = NULL;
server.glueoutputbuf = 1;
server.daemonize = 0;
+ server.appendonly = 0;
+ server.appendfsync = APPENDFSYNC_ALWAYS;
+ server.lastfsync = time(NULL);
+ server.appendfd = -1;
+ server.appendseldb = -1; /* Make sure the first time will not match */
server.pidfile = "/var/run/redis.pid";
server.dbfilename = "dump.rdb";
+ server.appendfilename = "appendonly.log";
server.requirepass = NULL;
server.shareobjects = 0;
- ResetServerSaveParams();
+ server.sharingpoolsize = 1024;
+ server.maxclients = 0;
+ server.maxmemory = 0;
+ resetServerSaveParams();
appendServerSaveParams(60*60,1); /* save after 1 hour and 1 change */
appendServerSaveParams(300,100); /* save after 5 minutes and 100 changes */
appendServerSaveParams(60,10000); /* save after 1 minute and 10000 changes */
/* Replication related */
server.isslave = 0;
+ server.masterauth = NULL;
server.masterhost = NULL;
server.masterport = 6379;
server.master = NULL;
server.replstate = REDIS_REPL_NONE;
+
+ /* Double constants initialization */
+ R_Zero = 0.0;
+ R_PosInf = 1.0/R_Zero;
+ R_NegInf = -1.0/R_Zero;
+ R_Nan = R_Zero/R_Zero;
}
static void initServer() {
signal(SIGHUP, SIG_IGN);
signal(SIGPIPE, SIG_IGN);
+ setupSigSegvAction();
server.clients = listCreate();
server.slaves = listCreate();
server.el = aeCreateEventLoop();
server.db = zmalloc(sizeof(redisDb)*server.dbnum);
server.sharingpool = dictCreate(&setDictType,NULL);
- server.sharingpoolsize = 1024;
- if (!server.db || !server.clients || !server.slaves || !server.monitors || !server.el || !server.objfreelist)
- oom("server initialization"); /* Fatal OOM */
server.fd = anetTcpServer(server.neterr, server.port, server.bindaddr);
if (server.fd == -1) {
redisLog(REDIS_WARNING, "Opening TCP port: %s", server.neterr);
}
server.cronloops = 0;
server.bgsaveinprogress = 0;
+ server.bgsavechildpid = -1;
server.lastsave = time(NULL);
server.dirty = 0;
server.usedmemory = 0;
server.stat_numcommands = 0;
server.stat_numconnections = 0;
server.stat_starttime = time(NULL);
- aeCreateTimeEvent(server.el, 1000, serverCron, NULL, NULL);
+ aeCreateTimeEvent(server.el, 1, serverCron, NULL, NULL);
+
+ if (server.appendonly) {
+ server.appendfd = open(server.appendfilename,O_WRONLY|O_APPEND|O_CREAT,0644);
+ if (server.appendfd == -1) {
+ redisLog(REDIS_WARNING, "Can't open the append-only file: %s",
+ strerror(errno));
+ exit(1);
+ }
+ }
}
/* Empty the whole database */
/* I agree, this is a very rudimental way to load a configuration...
will improve later if the config gets more complex */
static void loadServerConfig(char *filename) {
- FILE *fp = fopen(filename,"r");
+ FILE *fp;
char buf[REDIS_CONFIGLINE_MAX+1], *err = NULL;
int linenum = 0;
sds line = NULL;
-
- if (!fp) {
- redisLog(REDIS_WARNING,"Fatal error, can't open config file");
- exit(1);
+
+ if (filename[0] == '-' && filename[1] == '\0')
+ fp = stdin;
+ else {
+ if ((fp = fopen(filename,"r")) == NULL) {
+ redisLog(REDIS_WARNING,"Fatal error, can't open config file");
+ exit(1);
+ }
}
+
while(fgets(buf,REDIS_CONFIGLINE_MAX+1,fp) != NULL) {
sds *argv;
int argc, j;
/* Execute config directives */
if (!strcasecmp(argv[0],"timeout") && argc == 2) {
server.maxidletime = atoi(argv[1]);
- if (server.maxidletime < 1) {
+ if (server.maxidletime < 0) {
err = "Invalid timeout value"; goto loaderr;
}
} else if (!strcasecmp(argv[0],"port") && argc == 2) {
goto loaderr;
}
} else if (!strcasecmp(argv[0],"logfile") && argc == 2) {
- FILE *fp;
+ FILE *logfp;
server.logfile = zstrdup(argv[1]);
if (!strcasecmp(server.logfile,"stdout")) {
if (server.logfile) {
/* Test if we are able to open the file. The server will not
* be able to abort just for this problem later... */
- fp = fopen(server.logfile,"a");
- if (fp == NULL) {
+ logfp = fopen(server.logfile,"a");
+ if (logfp == NULL) {
err = sdscatprintf(sdsempty(),
"Can't open the log file: %s", strerror(errno));
goto loaderr;
}
- fclose(fp);
+ fclose(logfp);
}
} else if (!strcasecmp(argv[0],"databases") && argc == 2) {
server.dbnum = atoi(argv[1]);
if (server.dbnum < 1) {
err = "Invalid number of databases"; goto loaderr;
}
+ } 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);
} else if (!strcasecmp(argv[0],"slaveof") && argc == 3) {
server.masterhost = sdsnew(argv[1]);
server.masterport = atoi(argv[2]);
server.replstate = REDIS_REPL_CONNECT;
+ } else if (!strcasecmp(argv[0],"masterauth") && argc == 2) {
+ server.masterauth = zstrdup(argv[1]);
} else if (!strcasecmp(argv[0],"glueoutputbuf") && argc == 2) {
if ((server.glueoutputbuf = yesnotoi(argv[1])) == -1) {
err = "argument must be 'yes' or 'no'"; goto loaderr;
if ((server.shareobjects = 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],"daemonize") && argc == 2) {
if ((server.daemonize = yesnotoi(argv[1])) == -1) {
err = "argument must be 'yes' or 'no'"; goto loaderr;
}
+ } else if (!strcasecmp(argv[0],"appendonly") && argc == 2) {
+ if ((server.appendonly = yesnotoi(argv[1])) == -1) {
+ err = "argument must be 'yes' or 'no'"; goto loaderr;
+ }
+ } else if (!strcasecmp(argv[0],"appendfsync") && argc == 2) {
+ if (!strcasecmp(argv[1],"no")) {
+ server.appendfsync = APPENDFSYNC_NO;
+ } else if (!strcasecmp(argv[1],"always")) {
+ server.appendfsync = APPENDFSYNC_ALWAYS;
+ } else if (!strcasecmp(argv[1],"everysec")) {
+ server.appendfsync = APPENDFSYNC_EVERYSEC;
+ } else {
+ err = "argument must be 'no', 'always' or 'everysec'";
+ goto loaderr;
+ }
} else if (!strcasecmp(argv[0],"requirepass") && argc == 2) {
server.requirepass = zstrdup(argv[1]);
} else if (!strcasecmp(argv[0],"pidfile") && argc == 2) {
zfree(argv);
sdsfree(line);
}
- fclose(fp);
+ if (fp != stdin) fclose(fp);
return;
loaderr:
for (j = 0; j < c->argc; j++)
decrRefCount(c->argv[j]);
+ for (j = 0; j < c->mbargc; j++)
+ decrRefCount(c->mbargv[j]);
c->argc = 0;
+ c->mbargc = 0;
}
static void freeClient(redisClient *c) {
server.master = NULL;
server.replstate = REDIS_REPL_CONNECT;
}
+ zfree(c->argv);
+ zfree(c->mbargv);
zfree(c);
}
listDelNode(c->reply,ln);
}
/* Now the output buffer is empty, add the new single element */
- addReplySds(c,sdsnewlen(buf,totlen));
+ o = createObject(REDIS_STRING,sdsnewlen(buf,totlen));
+ listAddNodeTail(c->reply,o);
}
}
}
if (c->flags & REDIS_MASTER) {
+ /* Don't reply to a master */
nwritten = objlen - c->sentlen;
} else {
nwritten = write(fd, ((char*)o->ptr)+c->sentlen, objlen - c->sentlen);
listDelNode(c->reply,listFirst(c->reply));
c->sentlen = 0;
}
+ /* Note that we avoid to send more thank REDIS_MAX_WRITE_PER_EVENT
+ * bytes, in a single threaded server it's a good idea to serve
+ * other clients as well, even if a very large request comes from
+ * super fast link that is always able to accept data (in real world
+ * scenario think about 'KEYS *' against the loopback interfae) */
+ if (totwritten > REDIS_MAX_WRITE_PER_EVENT) break;
}
if (nwritten == -1) {
if (errno == EAGAIN) {
static void resetClient(redisClient *c) {
freeClientArgv(c);
c->bulklen = -1;
+ c->multibulk = 0;
}
/* If this function gets called we already read a whole
struct redisCommand *cmd;
long long dirty;
+ /* Free some memory if needed (maxmemory setting) */
+ if (server.maxmemory) freeMemoryIfNeeded();
+
+ /* Handle the multi bulk command type. This is an alternative protocol
+ * supported by Redis in order to receive commands that are composed of
+ * multiple binary-safe "bulk" arguments. The latency of processing is
+ * a bit higher but this allows things like multi-sets, so if this
+ * protocol is used only for MSET and similar commands this is a big win. */
+ if (c->multibulk == 0 && c->argc == 1 && ((char*)(c->argv[0]->ptr))[0] == '*') {
+ c->multibulk = atoi(((char*)c->argv[0]->ptr)+1);
+ if (c->multibulk <= 0) {
+ resetClient(c);
+ return 1;
+ } else {
+ decrRefCount(c->argv[c->argc-1]);
+ c->argc--;
+ return 1;
+ }
+ } else if (c->multibulk) {
+ if (c->bulklen == -1) {
+ if (((char*)c->argv[0]->ptr)[0] != '$') {
+ addReplySds(c,sdsnew("-ERR multi bulk protocol error\r\n"));
+ resetClient(c);
+ return 1;
+ } else {
+ int bulklen = atoi(((char*)c->argv[0]->ptr)+1);
+ decrRefCount(c->argv[0]);
+ if (bulklen < 0 || bulklen > 1024*1024*1024) {
+ c->argc--;
+ addReplySds(c,sdsnew("-ERR invalid bulk write count\r\n"));
+ resetClient(c);
+ return 1;
+ }
+ c->argc--;
+ c->bulklen = bulklen+2; /* add two bytes for CR+LF */
+ return 1;
+ }
+ } else {
+ c->mbargv = zrealloc(c->mbargv,(sizeof(robj*))*(c->mbargc+1));
+ c->mbargv[c->mbargc] = c->argv[0];
+ c->mbargc++;
+ c->argc--;
+ c->multibulk--;
+ if (c->multibulk == 0) {
+ robj **auxargv;
+ int auxargc;
+
+ /* Here we need to swap the multi-bulk argc/argv with the
+ * normal argc/argv of the client structure. */
+ auxargv = c->argv;
+ c->argv = c->mbargv;
+ c->mbargv = auxargv;
+
+ auxargc = c->argc;
+ c->argc = c->mbargc;
+ c->mbargc = auxargc;
+
+ /* We need to set bulklen to something different than -1
+ * in order for the code below to process the command without
+ * to try to read the last argument of a bulk command as
+ * a special argument. */
+ c->bulklen = 0;
+ /* continue below and process the command */
+ } else {
+ c->bulklen = -1;
+ return 1;
+ }
+ }
+ }
+ /* -- end of multi bulk commands processing -- */
+
/* The QUIT command is handled as a special case. Normal command
* procs are unable to close the client connection safely */
if (!strcasecmp(c->argv[0]->ptr,"quit")) {
addReplySds(c,sdsnew("-ERR wrong number of arguments\r\n"));
resetClient(c);
return 1;
+ } else if (server.maxmemory && cmd->flags & REDIS_CMD_DENYOOM && zmalloc_used_memory() > server.maxmemory) {
+ addReplySds(c,sdsnew("-ERR command not allowed when used memory > 'maxmemory'\r\n"));
+ resetClient(c);
+ return 1;
} else if (cmd->flags & REDIS_CMD_BULK && c->bulklen == -1) {
int bulklen = atoi(c->argv[c->argc-1]->ptr);
c->argc--;
c->bulklen = bulklen+2; /* add two bytes for CR+LF */
/* It is possible that the bulk read is already in the
- * buffer. Check this condition and handle it accordingly */
+ * buffer. Check this condition and handle it accordingly.
+ * This is just a fast path, alternative to call processInputBuffer().
+ * It's a good idea since the code is small and this condition
+ * happens most of the times. */
if ((signed)sdslen(c->querybuf) >= c->bulklen) {
c->argv[c->argc] = createStringObject(c->querybuf,c->bulklen-2);
c->argc++;
for(j = 1; j < c->argc; j++)
c->argv[j] = tryObjectSharing(c->argv[j]);
}
+ /* Let's try to encode the bulk object to save space. */
+ if (cmd->flags & REDIS_CMD_BULK)
+ tryObjectEncoding(c->argv[c->argc-1]);
+
/* Check if the user is authenticated */
if (server.requirepass && !c->authenticated && cmd->proc != authCommand) {
addReplySds(c,sdsnew("-ERR operation not permitted\r\n"));
/* Exec the command */
dirty = server.dirty;
cmd->proc(c);
- if (server.dirty-dirty != 0 && listLength(server.slaves))
+ if (server.appendonly && server.dirty-dirty)
+ feedAppendOnlyFile(cmd,c->db->id,c->argv,c->argc);
+ if (server.dirty-dirty && listLength(server.slaves))
replicationFeedSlaves(server.slaves,cmd,c->db->id,c->argv,c->argc);
if (listLength(server.monitors))
replicationFeedSlaves(server.monitors,cmd,c->db->id,c->argv,c->argc);
static void replicationFeedSlaves(list *slaves, struct redisCommand *cmd, int dictid, robj **argv, int argc) {
listNode *ln;
- robj *outv[REDIS_MAX_ARGS*4]; /* enough room for args, spaces, newlines */
int outc = 0, j;
+ robj **outv;
+ /* (args*2)+1 is enough room for args, spaces, newlines */
+ robj *static_outv[REDIS_STATIC_ARGS*2+1];
+
+ if (argc <= REDIS_STATIC_ARGS) {
+ outv = static_outv;
+ } else {
+ outv = zmalloc(sizeof(robj*)*(argc*2+1));
+ }
for (j = 0; j < argc; j++) {
if (j != 0) outv[outc++] = shared.space;
robj *lenobj;
lenobj = createObject(REDIS_STRING,
- sdscatprintf(sdsempty(),"%d\r\n",sdslen(argv[j]->ptr)));
+ sdscatprintf(sdsempty(),"%d\r\n",
+ stringObjectLen(argv[j])));
lenobj->refcount = 0;
outv[outc++] = lenobj;
}
for (j = 0; j < outc; j++) addReply(slave,outv[j]);
}
for (j = 0; j < outc; j++) decrRefCount(outv[j]);
+ if (outv != static_outv) zfree(outv);
}
-static void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask) {
- redisClient *c = (redisClient*) privdata;
- char buf[REDIS_IOBUF_LEN];
- int nread;
- REDIS_NOTUSED(el);
- REDIS_NOTUSED(mask);
-
- nread = read(fd, buf, REDIS_IOBUF_LEN);
- if (nread == -1) {
- if (errno == EAGAIN) {
- nread = 0;
- } else {
- redisLog(REDIS_DEBUG, "Reading from client: %s",strerror(errno));
- freeClient(c);
- return;
- }
- } else if (nread == 0) {
- redisLog(REDIS_DEBUG, "Client closed connection");
- freeClient(c);
- return;
- }
- if (nread) {
- c->querybuf = sdscatlen(c->querybuf, buf, nread);
- c->lastinteraction = time(NULL);
- } else {
- return;
- }
-
+static void processInputBuffer(redisClient *c) {
again:
if (c->bulklen == -1) {
/* Read the first line of the query */
char *p = strchr(c->querybuf,'\n');
size_t querylen;
+
if (p) {
sds query, *argv;
int argc, j;
}
argv = sdssplitlen(query,sdslen(query)," ",1,&argc);
sdsfree(query);
- if (argv == NULL) oom("sdssplitlen");
- for (j = 0; j < argc && j < REDIS_MAX_ARGS; j++) {
+
+ if (c->argv) zfree(c->argv);
+ c->argv = zmalloc(sizeof(robj*)*argc);
+
+ for (j = 0; j < argc; j++) {
if (sdslen(argv[j])) {
c->argv[c->argc] = createObject(REDIS_STRING,argv[j]);
c->argc++;
/* Execute the command. If the client is still valid
* after processCommand() return and there is something
* on the query buffer try to process the next command. */
- if (processCommand(c) && sdslen(c->querybuf)) goto again;
+ if (c->argc && processCommand(c) && sdslen(c->querybuf)) goto again;
return;
- } else if (sdslen(c->querybuf) >= 1024) {
+ } else if (sdslen(c->querybuf) >= REDIS_REQUEST_MAX_SIZE) {
redisLog(REDIS_DEBUG, "Client protocol error");
freeClient(c);
return;
c->argv[c->argc] = createStringObject(c->querybuf,c->bulklen-2);
c->argc++;
c->querybuf = sdsrange(c->querybuf,c->bulklen,-1);
- processCommand(c);
+ /* Process the command. If the client is still valid after
+ * the processing and there is more data in the buffer
+ * try to parse it. */
+ if (processCommand(c) && sdslen(c->querybuf)) goto again;
+ return;
+ }
+ }
+}
+
+static void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask) {
+ redisClient *c = (redisClient*) privdata;
+ char buf[REDIS_IOBUF_LEN];
+ int nread;
+ REDIS_NOTUSED(el);
+ REDIS_NOTUSED(mask);
+
+ nread = read(fd, buf, REDIS_IOBUF_LEN);
+ if (nread == -1) {
+ if (errno == EAGAIN) {
+ nread = 0;
+ } else {
+ redisLog(REDIS_DEBUG, "Reading from client: %s",strerror(errno));
+ freeClient(c);
return;
}
+ } else if (nread == 0) {
+ redisLog(REDIS_DEBUG, "Client closed connection");
+ freeClient(c);
+ return;
+ }
+ if (nread) {
+ c->querybuf = sdscatlen(c->querybuf, buf, nread);
+ c->lastinteraction = time(NULL);
+ } else {
+ return;
}
+ processInputBuffer(c);
}
static int selectDb(redisClient *c, int id) {
c->fd = fd;
c->querybuf = sdsempty();
c->argc = 0;
+ c->argv = NULL;
c->bulklen = -1;
+ c->multibulk = 0;
+ c->mbargc = 0;
+ c->mbargv = NULL;
c->sentlen = 0;
c->flags = 0;
c->lastinteraction = time(NULL);
c->authenticated = 0;
c->replstate = REDIS_REPL_NONE;
- if ((c->reply = listCreate()) == NULL) oom("listCreate");
+ c->reply = listCreate();
listSetFreeMethod(c->reply,decrRefCount);
listSetDupMethod(c->reply,dupClientReplyValue);
if (aeCreateFileEvent(server.el, c->fd, AE_READABLE,
freeClient(c);
return NULL;
}
- if (!listAddNodeTail(server.clients,c)) oom("listAddNodeTail");
+ listAddNodeTail(server.clients,c);
return c;
}
c->replstate == REDIS_REPL_ONLINE) &&
aeCreateFileEvent(server.el, c->fd, AE_WRITABLE,
sendReplyToClient, c, NULL) == AE_ERR) return;
- if (!listAddNodeTail(c->reply,obj)) oom("listAddNodeTail");
- incrRefCount(obj);
+ if (obj->encoding != REDIS_ENCODING_RAW) {
+ obj = getDecodedObject(obj);
+ } else {
+ incrRefCount(obj);
+ }
+ listAddNodeTail(c->reply,obj);
}
static void addReplySds(redisClient *c, sds s) {
decrRefCount(o);
}
+static void addReplyBulkLen(redisClient *c, robj *obj) {
+ size_t len;
+
+ if (obj->encoding == REDIS_ENCODING_RAW) {
+ len = sdslen(obj->ptr);
+ } else {
+ long n = (long)obj->ptr;
+
+ len = 1;
+ if (n < 0) {
+ len++;
+ n = -n;
+ }
+ while((n = n/10) != 0) {
+ len++;
+ }
+ }
+ addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",len));
+}
+
static void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
int cport, cfd;
char cip[128];
+ redisClient *c;
REDIS_NOTUSED(el);
REDIS_NOTUSED(mask);
REDIS_NOTUSED(privdata);
return;
}
redisLog(REDIS_DEBUG,"Accepted %s:%d", cip, cport);
- if (createClient(cfd) == NULL) {
+ if ((c = createClient(cfd)) == NULL) {
redisLog(REDIS_WARNING,"Error allocating resoures for the client");
close(cfd); /* May be already closed, just ingore errors */
return;
}
+ /* If maxclient directive is set and this is one client more... close the
+ * connection. Note that we create the client instead to check before
+ * for this condition, since now the socket is already set in nonblocking
+ * mode and we can send an error for free using the Kernel I/O */
+ if (server.maxclients && listLength(server.clients) > server.maxclients) {
+ char *err = "-ERR max number of clients reached\r\n";
+
+ /* That's a best effort error message, don't check write errors */
+ if (write(c->fd,err,strlen(err)) == -1) {
+ /* Nothing to do, Just to avoid the warning... */
+ }
+ freeClient(c);
+ return;
+ }
server.stat_numconnections++;
}
} else {
o = zmalloc(sizeof(*o));
}
- if (!o) oom("createObject");
o->type = type;
+ o->encoding = REDIS_ENCODING_RAW;
o->ptr = ptr;
o->refcount = 1;
return o;
static robj *createListObject(void) {
list *l = listCreate();
- if (!l) oom("listCreate");
listSetFreeMethod(l,decrRefCount);
return createObject(REDIS_LIST,l);
}
static robj *createSetObject(void) {
dict *d = dictCreate(&setDictType,NULL);
- if (!d) oom("dictCreate");
return createObject(REDIS_SET,d);
}
+static robj *createZsetObject(void) {
+ zset *zs = zmalloc(sizeof(*zs));
+
+ zs->dict = dictCreate(&zsetDictType,NULL);
+ zs->zsl = zslCreate();
+ return createObject(REDIS_ZSET,zs);
+}
+
static void freeStringObject(robj *o) {
- sdsfree(o->ptr);
+ if (o->encoding == REDIS_ENCODING_RAW) {
+ sdsfree(o->ptr);
+ }
}
static void freeListObject(robj *o) {
dictRelease((dict*) o->ptr);
}
+static void freeZsetObject(robj *o) {
+ zset *zs = o->ptr;
+
+ dictRelease(zs->dict);
+ zslFree(zs->zsl);
+ zfree(zs);
+}
+
static void freeHashObject(robj *o) {
dictRelease((dict*) o->ptr);
}
case REDIS_STRING: freeStringObject(o); break;
case REDIS_LIST: freeListObject(o); break;
case REDIS_SET: freeSetObject(o); break;
+ case REDIS_ZSET: freeZsetObject(o); break;
case REDIS_HASH: freeHashObject(o); break;
default: assert(0 != 0); break;
}
}
}
+static robj *lookupKey(redisDb *db, robj *key) {
+ dictEntry *de = dictFind(db->dict,key);
+ return de ? dictGetEntryVal(de) : NULL;
+}
+
+static robj *lookupKeyRead(redisDb *db, robj *key) {
+ expireIfNeeded(db,key);
+ return lookupKey(db,key);
+}
+
+static robj *lookupKeyWrite(redisDb *db, robj *key) {
+ deleteIfVolatile(db,key);
+ return lookupKey(db,key);
+}
+
+static int deleteKey(redisDb *db, robj *key) {
+ int retval;
+
+ /* We need to protect key from destruction: after the first dictDelete()
+ * it may happen that 'key' is no longer valid if we don't increment
+ * it's count. This may happen when we get the object reference directly
+ * from the hash table with dictRandomKey() or dict iterators */
+ incrRefCount(key);
+ if (dictSize(db->expires)) dictDelete(db->expires,key);
+ retval = dictDelete(db->dict,key);
+ decrRefCount(key);
+
+ return retval == DICT_OK;
+}
+
/* Try to share an object against the shared objects pool */
static robj *tryObjectSharing(robj *o) {
struct dictEntry *de;
}
}
-static robj *lookupKey(redisDb *db, robj *key) {
- dictEntry *de = dictFind(db->dict,key);
- return de ? dictGetEntryVal(de) : NULL;
-}
+/* Check if the nul-terminated string 's' can be represented by a long
+ * (that is, is a number that fits into long without any other space or
+ * character before or after the digits).
+ *
+ * If so, the function returns REDIS_OK and *longval is set to the value
+ * of the number. Otherwise REDIS_ERR is returned */
+static int isStringRepresentableAsLong(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);
-static robj *lookupKeyRead(redisDb *db, robj *key) {
- expireIfNeeded(db,key);
- return lookupKey(db,key);
+ /* If the number converted back into a string is not identical
+ * then it's not possible to encode the string as integer */
+ if (sdslen(s) != (unsigned)slen || memcmp(buf,s,slen)) return REDIS_ERR;
+ if (longval) *longval = value;
+ return REDIS_OK;
}
-static robj *lookupKeyWrite(redisDb *db, robj *key) {
- deleteIfVolatile(db,key);
- return lookupKey(db,key);
-}
+/* Try to encode a string object in order to save space */
+static int tryObjectEncoding(robj *o) {
+ long value;
+ sds s = o->ptr;
-static int deleteKey(redisDb *db, robj *key) {
- int retval;
+ if (o->encoding != REDIS_ENCODING_RAW)
+ return REDIS_ERR; /* Already encoded */
- /* We need to protect key from destruction: after the first dictDelete()
- * it may happen that 'key' is no longer valid if we don't increment
- * it's count. This may happen when we get the object reference directly
- * from the hash table with dictRandomKey() or dict iterators */
- incrRefCount(key);
- if (dictSize(db->expires)) dictDelete(db->expires,key);
- retval = dictDelete(db->dict,key);
- decrRefCount(key);
+ /* It's not save to encode shared objects: shared objects can be shared
+ * everywhere in the "object space" of Redis. Encoded objects can only
+ * appear as "values" (and not, for instance, as keys) */
+ if (o->refcount > 1) return REDIS_ERR;
- return retval == DICT_OK;
-}
+ /* Currently we try to encode only strings */
+ assert(o->type == REDIS_STRING);
-/*============================ DB saving/loading ============================ */
+ /* Check if we can represent this string as a long integer */
+ if (isStringRepresentableAsLong(s,&value) == REDIS_ERR) return REDIS_ERR;
-static int rdbSaveType(FILE *fp, unsigned char type) {
+ /* Ok, this object can be encoded */
+ o->encoding = REDIS_ENCODING_INT;
+ sdsfree(o->ptr);
+ o->ptr = (void*) value;
+ return REDIS_OK;
+}
+
+/* Get a decoded version of an encoded object (returned as a new object) */
+static robj *getDecodedObject(const robj *o) {
+ robj *dec;
+
+ assert(o->encoding != REDIS_ENCODING_RAW);
+ if (o->type == REDIS_STRING && o->encoding == REDIS_ENCODING_INT) {
+ char buf[32];
+
+ snprintf(buf,32,"%ld",(long)o->ptr);
+ dec = createStringObject(buf,strlen(buf));
+ return dec;
+ } else {
+ assert(1 != 1);
+ }
+}
+
+/* Compare two string objects via strcmp() or alike.
+ * Note that the objects may be integer-encoded. In such a case we
+ * use snprintf() to get a string representation of the numbers on the stack
+ * and compare the strings, it's much faster than calling getDecodedObject(). */
+static int compareStringObjects(robj *a, robj *b) {
+ assert(a->type == REDIS_STRING && b->type == REDIS_STRING);
+ char bufa[128], bufb[128], *astr, *bstr;
+ int bothsds = 1;
+
+ if (a == b) return 0;
+ if (a->encoding != REDIS_ENCODING_RAW) {
+ snprintf(bufa,sizeof(bufa),"%ld",(long) a->ptr);
+ astr = bufa;
+ bothsds = 0;
+ } else {
+ astr = a->ptr;
+ }
+ if (b->encoding != REDIS_ENCODING_RAW) {
+ snprintf(bufb,sizeof(bufb),"%ld",(long) b->ptr);
+ bstr = bufb;
+ bothsds = 0;
+ } else {
+ bstr = b->ptr;
+ }
+ return bothsds ? sdscmp(astr,bstr) : strcmp(astr,bstr);
+}
+
+static size_t stringObjectLen(robj *o) {
+ assert(o->type == REDIS_STRING);
+ if (o->encoding == REDIS_ENCODING_RAW) {
+ return sdslen(o->ptr);
+ } else {
+ char buf[32];
+
+ return snprintf(buf,32,"%ld",(long)o->ptr);
+ }
+}
+
+/*============================ DB saving/loading ============================ */
+
+static int rdbSaveType(FILE *fp, unsigned char type) {
if (fwrite(&type,1,1,fp) == 0) return -1;
return 0;
}
/* String objects in the form "2391" "-100" without any space and with a
* range of values that can fit in an 8, 16 or 32 bit signed value can be
* encoded as integers to save space */
-int rdbTryIntegerEncoding(sds s, unsigned char *enc) {
+static int rdbTryIntegerEncoding(sds s, unsigned char *enc) {
long long value;
char *endptr, buf[32];
/* Save a string objet as [len][data] on disk. If the object is a string
* representation of an integer value we try to safe it in a special form */
-static int rdbSaveStringObject(FILE *fp, robj *obj) {
- size_t len = sdslen(obj->ptr);
+static int rdbSaveStringObjectRaw(FILE *fp, robj *obj) {
+ size_t len;
int enclen;
+ len = sdslen(obj->ptr);
+
/* Try integer encoding */
if (len <= 11) {
unsigned char buf[5];
/* Try LZF compression - under 20 bytes it's unable to compress even
* aaaaaaaaaaaaaaaaaa so skip it */
- if (1 && len > 20) {
+ if (len > 20) {
int retval;
retval = rdbSaveLzfStringObject(fp,obj);
return 0;
}
+/* Like rdbSaveStringObjectRaw() but handle encoded objects */
+static int rdbSaveStringObject(FILE *fp, robj *obj) {
+ int retval;
+ robj *dec;
+
+ if (obj->encoding != REDIS_ENCODING_RAW) {
+ dec = getDecodedObject(obj);
+ retval = rdbSaveStringObjectRaw(fp,dec);
+ decrRefCount(dec);
+ return retval;
+ } else {
+ return rdbSaveStringObjectRaw(fp,obj);
+ }
+}
+
+/* Save a double value. Doubles are saved as strings prefixed by an unsigned
+ * 8 bit integer specifing the length of the representation.
+ * This 8 bit integer has special values in order to specify the following
+ * conditions:
+ * 253: not a number
+ * 254: + inf
+ * 255: - inf
+ */
+static int rdbSaveDoubleValue(FILE *fp, double val) {
+ unsigned char buf[128];
+ int len;
+
+ if (isnan(val)) {
+ buf[0] = 253;
+ len = 1;
+ } else if (!isfinite(val)) {
+ len = 1;
+ buf[0] = (val < 0) ? 255 : 254;
+ } else {
+ snprintf((char*)buf+1,sizeof(buf)-1,"%.17g",val);
+ buf[0] = strlen((char*)buf);
+ len = buf[0]+1;
+ }
+ if (fwrite(buf,len,1,fp) == 0) return -1;
+ return 0;
+}
+
/* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success */
static int rdbSave(char *filename) {
dictIterator *di = NULL;
int j;
time_t now = time(NULL);
- snprintf(tmpfile,256,"temp-%d.%ld.rdb",(int)time(NULL),(long int)random());
+ snprintf(tmpfile,256,"temp-%d.rdb", (int) getpid());
fp = fopen(tmpfile,"w");
if (!fp) {
redisLog(REDIS_WARNING, "Failed saving the DB: %s", strerror(errno));
dictIterator *di = dictGetIterator(set);
dictEntry *de;
- if (!set) oom("dictGetIteraotr");
if (rdbSaveLen(fp,dictSize(set)) == -1) goto werr;
while((de = dictNext(di)) != NULL) {
robj *eleobj = dictGetEntryKey(de);
if (rdbSaveStringObject(fp,eleobj) == -1) goto werr;
}
dictReleaseIterator(di);
+ } else if (o->type == REDIS_ZSET) {
+ /* Save a set value */
+ zset *zs = o->ptr;
+ dictIterator *di = dictGetIterator(zs->dict);
+ dictEntry *de;
+
+ if (rdbSaveLen(fp,dictSize(zs->dict)) == -1) goto werr;
+ while((de = dictNext(di)) != NULL) {
+ robj *eleobj = dictGetEntryKey(de);
+ double *score = dictGetEntryVal(de);
+
+ if (rdbSaveStringObject(fp,eleobj) == -1) goto werr;
+ if (rdbSaveDoubleValue(fp,*score) == -1) goto werr;
+ }
+ dictReleaseIterator(di);
} else {
assert(0 != 0);
}
/* Use RENAME to make sure the DB file is changed atomically only
* if the generate DB file is ok. */
if (rename(tmpfile,filename) == -1) {
- redisLog(REDIS_WARNING,"Error moving temp DB file on the final destionation: %s", strerror(errno));
+ redisLog(REDIS_WARNING,"Error moving temp DB file on the final destination: %s", strerror(errno));
unlink(tmpfile);
return REDIS_ERR;
}
}
} else {
/* Parent */
+ if (childpid == -1) {
+ redisLog(REDIS_WARNING,"Can't save in background: fork: %s",
+ strerror(errno));
+ return REDIS_ERR;
+ }
redisLog(REDIS_NOTICE,"Background saving started by pid %d",childpid);
server.bgsaveinprogress = 1;
+ server.bgsavechildpid = childpid;
return REDIS_OK;
}
return REDIS_OK; /* unreached */
}
+static void rdbRemoveTempFile(pid_t childpid) {
+ char tmpfile[256];
+
+ snprintf(tmpfile,256,"temp-%d.rdb", (int) childpid);
+ unlink(tmpfile);
+}
+
static int rdbLoadType(FILE *fp) {
unsigned char type;
if (fread(&type,1,1,fp) == 0) return -1;
if ((val = sdsnewlen(NULL,len)) == NULL) goto err;
if (fread(c,clen,1,fp) == 0) goto err;
if (lzf_decompress(c,clen,val,len) == 0) goto err;
+ zfree(c);
return createObject(REDIS_STRING,val);
err:
zfree(c);
return tryObjectSharing(createObject(REDIS_STRING,val));
}
+/* For information about double serialization check rdbSaveDoubleValue() */
+static int rdbLoadDoubleValue(FILE *fp, double *val) {
+ char buf[128];
+ unsigned char len;
+
+ if (fread(&len,1,1,fp) == 0) return -1;
+ switch(len) {
+ case 255: *val = R_NegInf; return 0;
+ case 254: *val = R_PosInf; return 0;
+ case 253: *val = R_Nan; return 0;
+ default:
+ if (fread(buf,len,1,fp) == 0) return -1;
+ sscanf(buf, "%lg", val);
+ return 0;
+ }
+}
+
static int rdbLoad(char *filename) {
FILE *fp;
robj *keyobj = NULL;
if (type == REDIS_STRING) {
/* Read string value */
if ((o = rdbLoadStringObject(fp,rdbver)) == NULL) goto eoferr;
+ tryObjectEncoding(o);
} else if (type == REDIS_LIST || type == REDIS_SET) {
/* Read list/set value */
uint32_t listlen;
robj *ele;
if ((ele = rdbLoadStringObject(fp,rdbver)) == NULL) goto eoferr;
+ tryObjectEncoding(ele);
if (type == REDIS_LIST) {
- if (!listAddNodeTail((list*)o->ptr,ele))
- oom("listAddNodeTail");
+ listAddNodeTail((list*)o->ptr,ele);
} else {
- if (dictAdd((dict*)o->ptr,ele,NULL) == DICT_ERR)
- oom("dictAdd");
+ dictAdd((dict*)o->ptr,ele,NULL);
}
}
+ } else if (type == REDIS_ZSET) {
+ /* Read list/set value */
+ uint32_t zsetlen;
+ zset *zs;
+
+ if ((zsetlen = rdbLoadLen(fp,rdbver,NULL)) == REDIS_RDB_LENERR)
+ goto eoferr;
+ o = createZsetObject();
+ zs = o->ptr;
+ /* Load every single element of the list/set */
+ while(zsetlen--) {
+ robj *ele;
+ double *score = zmalloc(sizeof(double));
+
+ if ((ele = rdbLoadStringObject(fp,rdbver)) == NULL) goto eoferr;
+ tryObjectEncoding(ele);
+ if (rdbLoadDoubleValue(fp,score) == -1) goto eoferr;
+ dictAdd(zs->dict,ele,score);
+ zslInsert(zs->zsl,*score,ele);
+ incrRefCount(ele); /* added to skiplist */
+ }
} else {
assert(0 != 0);
}
eoferr: /* unexpected end of file is handled here with a fatal exit */
if (keyobj) decrRefCount(keyobj);
- redisLog(REDIS_WARNING,"Short read or OOM loading DB. Unrecoverable error, exiting now.");
+ redisLog(REDIS_WARNING,"Short read or OOM loading DB. Unrecoverable error, aborting now.");
exit(1);
return REDIS_ERR; /* Just to avoid warning */
}
addReply(c,shared.ok);
} else {
c->authenticated = 0;
- addReply(c,shared.err);
+ addReplySds(c,sdscatprintf(sdsempty(),"-ERR invalid password\r\n"));
}
}
}
static void echoCommand(redisClient *c) {
- addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",
- (int)sdslen(c->argv[1]->ptr)));
+ addReplyBulkLen(c,c->argv[1]);
addReply(c,c->argv[1]);
addReply(c,shared.crlf);
}
if (o->type != REDIS_STRING) {
addReply(c,shared.wrongtypeerr);
} else {
- addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(o->ptr)));
+ addReplyBulkLen(c,o);
addReply(c,o);
addReply(c,shared.crlf);
}
}
}
-static void getSetCommand(redisClient *c) {
+static void getsetCommand(redisClient *c) {
getCommand(c);
if (dictAdd(c->db->dict,c->argv[1],c->argv[2]) == DICT_ERR) {
dictReplace(c->db->dict,c->argv[1],c->argv[2]);
if (o->type != REDIS_STRING) {
addReply(c,shared.nullbulk);
} else {
- addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(o->ptr)));
+ addReplyBulkLen(c,o);
addReply(c,o);
addReply(c,shared.crlf);
}
} else {
char *eptr;
- value = strtoll(o->ptr, &eptr, 10);
+ if (o->encoding == REDIS_ENCODING_RAW)
+ value = strtoll(o->ptr, &eptr, 10);
+ else if (o->encoding == REDIS_ENCODING_INT)
+ value = (long)o->ptr;
+ else
+ assert(1 != 1);
}
}
value += incr;
o = createObject(REDIS_STRING,sdscatprintf(sdsempty(),"%lld",value));
+ tryObjectEncoding(o);
retval = dictAdd(c->db->dict,c->argv[1],o);
if (retval == DICT_ERR) {
dictReplace(c->db->dict,c->argv[1],o);
/* ========================= Type agnostic commands ========================= */
static void delCommand(redisClient *c) {
- if (deleteKey(c->db,c->argv[1])) {
- server.dirty++;
- addReply(c,shared.cone);
- } else {
+ int deleted = 0, j;
+
+ for (j = 1; j < c->argc; j++) {
+ if (deleteKey(c->db,c->argv[j])) {
+ server.dirty++;
+ deleted++;
+ }
+ }
+ switch(deleted) {
+ case 0:
addReply(c,shared.czero);
+ break;
+ case 1:
+ addReply(c,shared.cone);
+ break;
+ default:
+ addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",deleted));
+ break;
}
}
robj *lenobj = createObject(REDIS_STRING,NULL);
di = dictGetIterator(c->db->dict);
- if (!di) oom("dictGetIterator");
addReply(c,lenobj);
decrRefCount(lenobj);
while((de = dictNext(di)) != NULL) {
case REDIS_STRING: type = "+string"; break;
case REDIS_LIST: type = "+list"; break;
case REDIS_SET: type = "+set"; break;
+ case REDIS_ZSET: type = "+zset"; break;
default: type = "unknown"; break;
}
}
static void shutdownCommand(redisClient *c) {
redisLog(REDIS_WARNING,"User requested shutdown, saving DB...");
- /* XXX: TODO kill the child if there is a bgsave in progress */
+ /* Kill the saving child if there is a background saving in progress.
+ We want to avoid race conditions, for instance our saving child may
+ overwrite the synchronous saving did by SHUTDOWN. */
+ if (server.bgsaveinprogress) {
+ redisLog(REDIS_WARNING,"There is a live saving child. Killing it!");
+ kill(server.bgsavechildpid,SIGKILL);
+ rdbRemoveTempFile(server.bgsavechildpid);
+ }
+ /* SYNC SAVE */
if (rdbSave(server.dbfilename) == REDIS_OK) {
- if (server.daemonize) {
+ if (server.daemonize)
unlink(server.pidfile);
- }
redisLog(REDIS_WARNING,"%zu bytes used at exit",zmalloc_used_memory());
redisLog(REDIS_WARNING,"Server exit now, bye bye...");
exit(1);
} else {
+ /* Ooops.. error saving! The best we can do is to continue operating.
+ * Note that if there was a background saving process, in the next
+ * cron() Redis will be notified that the background saving aborted,
+ * handling special stuff like slaves pending for synchronization... */
redisLog(REDIS_WARNING,"Error trying to save the DB, can't exit");
addReplySds(c,sdsnew("-ERR can't quit, problems saving the DB\r\n"));
}
lobj = createListObject();
list = lobj->ptr;
if (where == REDIS_HEAD) {
- if (!listAddNodeHead(list,c->argv[2])) oom("listAddNodeHead");
+ listAddNodeHead(list,c->argv[2]);
} else {
- if (!listAddNodeTail(list,c->argv[2])) oom("listAddNodeTail");
+ listAddNodeTail(list,c->argv[2]);
}
dictAdd(c->db->dict,c->argv[1],lobj);
incrRefCount(c->argv[1]);
}
list = lobj->ptr;
if (where == REDIS_HEAD) {
- if (!listAddNodeHead(list,c->argv[2])) oom("listAddNodeHead");
+ listAddNodeHead(list,c->argv[2]);
} else {
- if (!listAddNodeTail(list,c->argv[2])) oom("listAddNodeTail");
+ listAddNodeTail(list,c->argv[2]);
}
incrRefCount(c->argv[2]);
}
addReply(c,shared.nullbulk);
} else {
robj *ele = listNodeValue(ln);
- addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(ele->ptr)));
+ addReplyBulkLen(c,ele);
addReply(c,ele);
addReply(c,shared.crlf);
}
addReply(c,shared.nullbulk);
} else {
robj *ele = listNodeValue(ln);
- addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(ele->ptr)));
+ addReplyBulkLen(c,ele);
addReply(c,ele);
addReply(c,shared.crlf);
listDelNode(list,ln);
addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",rangelen));
for (j = 0; j < rangelen; j++) {
ele = listNodeValue(ln);
- addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",(int)sdslen(ele->ptr)));
+ addReplyBulkLen(c,ele);
addReply(c,ele);
addReply(c,shared.crlf);
ln = ln->next;
ln = listLast(list);
listDelNode(list,ln);
}
- addReply(c,shared.ok);
server.dirty++;
+ addReply(c,shared.ok);
}
}
}
o = lookupKeyWrite(c->db,c->argv[1]);
if (o == NULL) {
- addReply(c,shared.nokeyerr);
+ addReply(c,shared.czero);
} else {
if (o->type != REDIS_LIST) {
addReply(c,shared.wrongtypeerr);
robj *ele = listNodeValue(ln);
next = fromtail ? ln->prev : ln->next;
- if (sdscmp(ele->ptr,c->argv[3]->ptr) == 0) {
+ if (compareStringObjects(ele,c->argv[3]) == 0) {
listDelNode(list,ln);
server.dirty++;
removed++;
}
}
+/* This is the semantic of this command:
+ * RPOPLPUSH srclist dstlist:
+ * IF LLEN(srclist) > 0
+ * element = RPOP srclist
+ * LPUSH dstlist element
+ * RETURN element
+ * ELSE
+ * RETURN nil
+ * END
+ * END
+ *
+ * The idea is to be able to get an element from a list in a reliable way
+ * since the element is not just returned but pushed against another list
+ * as well. This command was originally proposed by Ezra Zygmuntowicz.
+ */
+static void rpoplpushcommand(redisClient *c) {
+ robj *sobj;
+
+ sobj = lookupKeyWrite(c->db,c->argv[1]);
+ if (sobj == NULL) {
+ addReply(c,shared.nullbulk);
+ } else {
+ if (sobj->type != REDIS_LIST) {
+ addReply(c,shared.wrongtypeerr);
+ } else {
+ list *srclist = sobj->ptr;
+ listNode *ln = listLast(srclist);
+
+ if (ln == NULL) {
+ addReply(c,shared.nullbulk);
+ } else {
+ robj *dobj = lookupKeyWrite(c->db,c->argv[2]);
+ robj *ele = listNodeValue(ln);
+ list *dstlist;
+
+ if (dobj == NULL) {
+
+ /* Create the list if the key does not exist */
+ dobj = createListObject();
+ dictAdd(c->db->dict,c->argv[2],dobj);
+ incrRefCount(c->argv[2]);
+ } else if (dobj->type != REDIS_LIST) {
+ addReply(c,shared.wrongtypeerr);
+ return;
+ }
+ /* Add the element to the target list */
+ dstlist = dobj->ptr;
+ listAddNodeHead(dstlist,ele);
+ incrRefCount(ele);
+
+ /* Send the element to the client as reply as well */
+ addReplyBulkLen(c,ele);
+ addReply(c,ele);
+ addReply(c,shared.crlf);
+
+ /* Finally remove the element from the source list */
+ listDelNode(srclist,ln);
+ server.dirty++;
+ }
+ }
+ }
+}
+
+
/* ==================================== Sets ================================ */
static void saddCommand(redisClient *c) {
}
if (dictDelete(set->ptr,c->argv[2]) == DICT_OK) {
server.dirty++;
+ if (htNeedsResize(set->ptr)) dictResize(set->ptr);
addReply(c,shared.cone);
} else {
addReply(c,shared.czero);
}
}
+static void spopCommand(redisClient *c) {
+ robj *set;
+ dictEntry *de;
+
+ set = lookupKeyWrite(c->db,c->argv[1]);
+ if (set == NULL) {
+ addReply(c,shared.nullbulk);
+ } else {
+ if (set->type != REDIS_SET) {
+ addReply(c,shared.wrongtypeerr);
+ return;
+ }
+ de = dictGetRandomKey(set->ptr);
+ if (de == NULL) {
+ addReply(c,shared.nullbulk);
+ } else {
+ robj *ele = dictGetEntryKey(de);
+
+ addReplyBulkLen(c,ele);
+ addReply(c,ele);
+ addReply(c,shared.crlf);
+ dictDelete(set->ptr,ele);
+ if (htNeedsResize(set->ptr)) dictResize(set->ptr);
+ server.dirty++;
+ }
+ }
+}
+
+static void srandmemberCommand(redisClient *c) {
+ robj *set;
+ dictEntry *de;
+
+ set = lookupKeyRead(c->db,c->argv[1]);
+ if (set == NULL) {
+ addReply(c,shared.nullbulk);
+ } else {
+ if (set->type != REDIS_SET) {
+ addReply(c,shared.wrongtypeerr);
+ return;
+ }
+ de = dictGetRandomKey(set->ptr);
+ if (de == NULL) {
+ addReply(c,shared.nullbulk);
+ } else {
+ robj *ele = dictGetEntryKey(de);
+
+ addReplyBulkLen(c,ele);
+ addReply(c,ele);
+ addReply(c,shared.crlf);
+ }
+ }
+}
+
static int qsortCompareSetsByCardinality(const void *s1, const void *s2) {
dict **d1 = (void*) s1, **d2 = (void*) s2;
robj *lenobj = NULL, *dstset = NULL;
int j, cardinality = 0;
- if (!dv) oom("sinterCommand");
for (j = 0; j < setsnum; j++) {
robj *setobj;
/* If we have a target key where to store the resulting set
* create this key with an empty set inside */
dstset = createSetObject();
- deleteKey(c->db,dstkey);
- dictAdd(c->db->dict,dstkey,dstset);
- incrRefCount(dstkey);
}
/* Iterate all the elements of the first (smallest) set, and test
* the element against all the other sets, if at least one set does
* not include the element it is discarded */
di = dictGetIterator(dv[0]);
- if (!di) oom("dictGetIterator");
while((de = dictNext(di)) != NULL) {
robj *ele;
continue; /* at least one set does not contain the member */
ele = dictGetEntryKey(de);
if (!dstkey) {
- addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",sdslen(ele->ptr)));
+ addReplyBulkLen(c,ele);
addReply(c,ele);
addReply(c,shared.crlf);
cardinality++;
}
dictReleaseIterator(di);
+ if (dstkey) {
+ /* Store the resulting set into the target */
+ deleteKey(c->db,dstkey);
+ dictAdd(c->db->dict,dstkey,dstset);
+ incrRefCount(dstkey);
+ }
+
if (!dstkey) {
lenobj->ptr = sdscatprintf(sdsempty(),"*%d\r\n",cardinality);
} else {
- addReply(c,shared.ok);
+ addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",
+ dictSize((dict*)dstset->ptr)));
server.dirty++;
}
zfree(dv);
sinterGenericCommand(c,c->argv+2,c->argc-2,c->argv[1]);
}
-static void sunionGenericCommand(redisClient *c, robj **setskeys, int setsnum, robj *dstkey) {
+#define REDIS_OP_UNION 0
+#define REDIS_OP_DIFF 1
+
+static void sunionDiffGenericCommand(redisClient *c, robj **setskeys, int setsnum, robj *dstkey, int op) {
dict **dv = zmalloc(sizeof(dict*)*setsnum);
dictIterator *di;
dictEntry *de;
- robj *lenobj = NULL, *dstset = NULL;
+ robj *dstset = NULL;
int j, cardinality = 0;
- if (!dv) oom("sunionCommand");
for (j = 0; j < setsnum; j++) {
robj *setobj;
* this set object will be the resulting object to set into the target key*/
dstset = createSetObject();
- /* The first thing we should output is the total number of elements...
- * since this is a multi-bulk write, but at this stage we don't know
- * the intersection set size, so we use a trick, append an empty object
- * to the output list and save the pointer to later modify it with the
- * right length */
- if (!dstkey) {
- lenobj = createObject(REDIS_STRING,NULL);
- addReply(c,lenobj);
- decrRefCount(lenobj);
- } else {
- /* If we have a target key where to store the resulting set
- * create this key with an empty set inside */
- deleteKey(c->db,dstkey);
- dictAdd(c->db->dict,dstkey,dstset);
- incrRefCount(dstkey);
- server.dirty++;
- }
-
/* Iterate all the elements of all the sets, add every element a single
* time to the result set */
for (j = 0; j < setsnum; j++) {
+ if (op == REDIS_OP_DIFF && j == 0 && !dv[j]) break; /* result set is empty */
if (!dv[j]) continue; /* non existing keys are like empty sets */
di = dictGetIterator(dv[j]);
- if (!di) oom("dictGetIterator");
while((de = dictNext(di)) != NULL) {
robj *ele;
/* dictAdd will not add the same element multiple times */
ele = dictGetEntryKey(de);
- if (dictAdd(dstset->ptr,ele,NULL) == DICT_OK) {
- incrRefCount(ele);
- if (!dstkey) {
- addReplySds(c,sdscatprintf(sdsempty(),
- "$%d\r\n",sdslen(ele->ptr)));
- addReply(c,ele);
- addReply(c,shared.crlf);
+ if (op == REDIS_OP_UNION || j == 0) {
+ if (dictAdd(dstset->ptr,ele,NULL) == DICT_OK) {
+ incrRefCount(ele);
cardinality++;
}
+ } else if (op == REDIS_OP_DIFF) {
+ if (dictDelete(dstset->ptr,ele) == DICT_OK) {
+ cardinality--;
+ }
}
}
dictReleaseIterator(di);
+
+ if (op == REDIS_OP_DIFF && cardinality == 0) break; /* result set is empty */
}
+ /* Output the content of the resulting set, if not in STORE mode */
+ if (!dstkey) {
+ addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",cardinality));
+ di = dictGetIterator(dstset->ptr);
+ while((de = dictNext(di)) != NULL) {
+ robj *ele;
+
+ ele = dictGetEntryKey(de);
+ addReplyBulkLen(c,ele);
+ addReply(c,ele);
+ addReply(c,shared.crlf);
+ }
+ dictReleaseIterator(di);
+ } else {
+ /* If we have a target key where to store the resulting set
+ * create this key with the result set inside */
+ deleteKey(c->db,dstkey);
+ dictAdd(c->db->dict,dstkey,dstset);
+ incrRefCount(dstkey);
+ }
+
+ /* Cleanup */
if (!dstkey) {
- lenobj->ptr = sdscatprintf(sdsempty(),"*%d\r\n",cardinality);
decrRefCount(dstset);
} else {
- addReply(c,shared.ok);
+ addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",
+ dictSize((dict*)dstset->ptr)));
server.dirty++;
}
zfree(dv);
}
static void sunionCommand(redisClient *c) {
- sunionGenericCommand(c,c->argv+1,c->argc-1,NULL);
+ sunionDiffGenericCommand(c,c->argv+1,c->argc-1,NULL,REDIS_OP_UNION);
}
static void sunionstoreCommand(redisClient *c) {
- sunionGenericCommand(c,c->argv+2,c->argc-2,c->argv[1]);
-}
-
-static void flushdbCommand(redisClient *c) {
- server.dirty += dictSize(c->db->dict);
- dictEmpty(c->db->dict);
- dictEmpty(c->db->expires);
- addReply(c,shared.ok);
+ sunionDiffGenericCommand(c,c->argv+2,c->argc-2,c->argv[1],REDIS_OP_UNION);
}
-static void flushallCommand(redisClient *c) {
- server.dirty += emptyDb();
- addReply(c,shared.ok);
- rdbSave(server.dbfilename);
- server.dirty++;
+static void sdiffCommand(redisClient *c) {
+ sunionDiffGenericCommand(c,c->argv+1,c->argc-1,NULL,REDIS_OP_DIFF);
}
-redisSortOperation *createSortOperation(int type, robj *pattern) {
- redisSortOperation *so = zmalloc(sizeof(*so));
- if (!so) oom("createSortOperation");
- so->type = type;
- so->pattern = pattern;
- return so;
+static void sdiffstoreCommand(redisClient *c) {
+ sunionDiffGenericCommand(c,c->argv+2,c->argc-2,c->argv[1],REDIS_OP_DIFF);
}
-/* Return the value associated to the key with a name obtained
- * substituting the first occurence of '*' in 'pattern' with 'subst' */
-robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) {
- char *p;
- sds spat, ssub;
- robj keyobj;
- int prefixlen, sublen, postfixlen;
- /* Expoit the internal sds representation to create a sds string allocated on the stack in order to make this function faster */
- struct {
- long len;
- long free;
- char buf[REDIS_SORTKEY_MAX+1];
- } keyname;
+/* ==================================== ZSets =============================== */
- spat = pattern->ptr;
- ssub = subst->ptr;
- if (sdslen(spat)+sdslen(ssub)-1 > REDIS_SORTKEY_MAX) return NULL;
- p = strchr(spat,'*');
- if (!p) return NULL;
+/* ZSETs are ordered sets using two data structures to hold the same elements
+ * in order to get O(log(N)) INSERT and REMOVE operations into a sorted
+ * data structure.
+ *
+ * The elements are added to an hash table mapping Redis objects to scores.
+ * At the same time the elements are added to a skip list mapping scores
+ * to Redis objects (so objects are sorted by scores in this "view"). */
- prefixlen = p-spat;
- sublen = sdslen(ssub);
- postfixlen = sdslen(spat)-(prefixlen+1);
- memcpy(keyname.buf,spat,prefixlen);
- memcpy(keyname.buf+prefixlen,ssub,sublen);
- memcpy(keyname.buf+prefixlen+sublen,p+1,postfixlen);
- keyname.buf[prefixlen+sublen+postfixlen] = '\0';
- keyname.len = prefixlen+sublen+postfixlen;
+/* This skiplist implementation is almost a C translation of the original
+ * algorithm described by William Pugh in "Skip Lists: A Probabilistic
+ * Alternative to Balanced Trees", modified in three ways:
+ * a) this implementation allows for repeated values.
+ * b) the comparison is not just by key (our 'score') but by satellite data.
+ * c) there is a back pointer, so it's a doubly linked list with the back
+ * pointers being only at "level 1". This allows to traverse the list
+ * from tail to head, useful for ZREVRANGE. */
- keyobj.refcount = 1;
- keyobj.type = REDIS_STRING;
- keyobj.ptr = ((char*)&keyname)+(sizeof(long)*2);
+static zskiplistNode *zslCreateNode(int level, double score, robj *obj) {
+ zskiplistNode *zn = zmalloc(sizeof(*zn));
- /* printf("lookup '%s' => %p\n", keyname.buf,de); */
- return lookupKeyRead(db,&keyobj);
+ zn->forward = zmalloc(sizeof(zskiplistNode*) * level);
+ zn->score = score;
+ zn->obj = obj;
+ return zn;
+}
+
+static zskiplist *zslCreate(void) {
+ int j;
+ zskiplist *zsl;
+
+ zsl = zmalloc(sizeof(*zsl));
+ zsl->level = 1;
+ zsl->length = 0;
+ zsl->header = zslCreateNode(ZSKIPLIST_MAXLEVEL,0,NULL);
+ for (j = 0; j < ZSKIPLIST_MAXLEVEL; j++)
+ zsl->header->forward[j] = NULL;
+ zsl->header->backward = NULL;
+ zsl->tail = NULL;
+ return zsl;
+}
+
+static void zslFreeNode(zskiplistNode *node) {
+ decrRefCount(node->obj);
+ zfree(node->forward);
+ zfree(node);
+}
+
+static void zslFree(zskiplist *zsl) {
+ zskiplistNode *node = zsl->header->forward[0], *next;
+
+ zfree(zsl->header->forward);
+ zfree(zsl->header);
+ while(node) {
+ next = node->forward[0];
+ zslFreeNode(node);
+ node = next;
+ }
+ zfree(zsl);
+}
+
+static int zslRandomLevel(void) {
+ int level = 1;
+ while ((random()&0xFFFF) < (ZSKIPLIST_P * 0xFFFF))
+ level += 1;
+ return level;
+}
+
+static void zslInsert(zskiplist *zsl, double score, robj *obj) {
+ zskiplistNode *update[ZSKIPLIST_MAXLEVEL], *x;
+ int i, level;
+
+ x = zsl->header;
+ for (i = zsl->level-1; i >= 0; i--) {
+ while (x->forward[i] &&
+ (x->forward[i]->score < score ||
+ (x->forward[i]->score == score &&
+ compareStringObjects(x->forward[i]->obj,obj) < 0)))
+ x = x->forward[i];
+ update[i] = x;
+ }
+ /* we assume the key is not already inside, since we allow duplicated
+ * scores, and the re-insertion of score and redis object should never
+ * happpen since the caller of zslInsert() should test in the hash table
+ * if the element is already inside or not. */
+ level = zslRandomLevel();
+ if (level > zsl->level) {
+ for (i = zsl->level; i < level; i++)
+ update[i] = zsl->header;
+ zsl->level = level;
+ }
+ x = zslCreateNode(level,score,obj);
+ for (i = 0; i < level; i++) {
+ x->forward[i] = update[i]->forward[i];
+ update[i]->forward[i] = x;
+ }
+ x->backward = (update[0] == zsl->header) ? NULL : update[0];
+ if (x->forward[0])
+ x->forward[0]->backward = x;
+ else
+ zsl->tail = x;
+ zsl->length++;
+}
+
+/* Delete an element with matching score/object from the skiplist. */
+static int zslDelete(zskiplist *zsl, double score, robj *obj) {
+ zskiplistNode *update[ZSKIPLIST_MAXLEVEL], *x;
+ int i;
+
+ x = zsl->header;
+ for (i = zsl->level-1; i >= 0; i--) {
+ while (x->forward[i] &&
+ (x->forward[i]->score < score ||
+ (x->forward[i]->score == score &&
+ compareStringObjects(x->forward[i]->obj,obj) < 0)))
+ x = x->forward[i];
+ update[i] = x;
+ }
+ /* We may have multiple elements with the same score, what we need
+ * is to find the element with both the right score and object. */
+ x = x->forward[0];
+ if (x && score == x->score && compareStringObjects(x->obj,obj) == 0) {
+ for (i = 0; i < zsl->level; i++) {
+ if (update[i]->forward[i] != x) break;
+ update[i]->forward[i] = x->forward[i];
+ }
+ if (x->forward[0]) {
+ x->forward[0]->backward = (x->backward == zsl->header) ?
+ NULL : x->backward;
+ } else {
+ zsl->tail = x->backward;
+ }
+ zslFreeNode(x);
+ while(zsl->level > 1 && zsl->header->forward[zsl->level-1] == NULL)
+ zsl->level--;
+ zsl->length--;
+ return 1;
+ } else {
+ return 0; /* not found */
+ }
+ return 0; /* not found */
+}
+
+/* Delete all the elements with score between min and max from the skiplist.
+ * Min and mx are inclusive, so a score >= min || score <= max is deleted.
+ * Note that this function takes the reference to the hash table view of the
+ * sorted set, in order to remove the elements from the hash table too. */
+static unsigned long zslDeleteRange(zskiplist *zsl, double min, double max, dict *dict) {
+ zskiplistNode *update[ZSKIPLIST_MAXLEVEL], *x;
+ unsigned long removed = 0;
+ int i;
+
+ x = zsl->header;
+ for (i = zsl->level-1; i >= 0; i--) {
+ while (x->forward[i] && x->forward[i]->score < min)
+ x = x->forward[i];
+ update[i] = x;
+ }
+ /* We may have multiple elements with the same score, what we need
+ * is to find the element with both the right score and object. */
+ x = x->forward[0];
+ while (x && x->score <= max) {
+ zskiplistNode *next;
+
+ for (i = 0; i < zsl->level; i++) {
+ if (update[i]->forward[i] != x) break;
+ update[i]->forward[i] = x->forward[i];
+ }
+ if (x->forward[0]) {
+ x->forward[0]->backward = (x->backward == zsl->header) ?
+ NULL : x->backward;
+ } else {
+ zsl->tail = x->backward;
+ }
+ next = x->forward[0];
+ dictDelete(dict,x->obj);
+ zslFreeNode(x);
+ while(zsl->level > 1 && zsl->header->forward[zsl->level-1] == NULL)
+ zsl->level--;
+ zsl->length--;
+ removed++;
+ x = next;
+ }
+ return removed; /* not found */
+}
+
+/* Find the first node having a score equal or greater than the specified one.
+ * Returns NULL if there is no match. */
+static zskiplistNode *zslFirstWithScore(zskiplist *zsl, double score) {
+ zskiplistNode *x;
+ int i;
+
+ x = zsl->header;
+ for (i = zsl->level-1; i >= 0; i--) {
+ while (x->forward[i] && x->forward[i]->score < score)
+ x = x->forward[i];
+ }
+ /* We may have multiple elements with the same score, what we need
+ * is to find the element with both the right score and object. */
+ return x->forward[0];
+}
+
+/* The actual Z-commands implementations */
+
+static void zaddCommand(redisClient *c) {
+ robj *zsetobj;
+ zset *zs;
+ double *score;
+
+ zsetobj = lookupKeyWrite(c->db,c->argv[1]);
+ if (zsetobj == NULL) {
+ zsetobj = createZsetObject();
+ dictAdd(c->db->dict,c->argv[1],zsetobj);
+ incrRefCount(c->argv[1]);
+ } else {
+ if (zsetobj->type != REDIS_ZSET) {
+ addReply(c,shared.wrongtypeerr);
+ return;
+ }
+ }
+ score = zmalloc(sizeof(double));
+ *score = strtod(c->argv[2]->ptr,NULL);
+ zs = zsetobj->ptr;
+ if (dictAdd(zs->dict,c->argv[3],score) == DICT_OK) {
+ /* case 1: New element */
+ incrRefCount(c->argv[3]); /* added to hash */
+ zslInsert(zs->zsl,*score,c->argv[3]);
+ incrRefCount(c->argv[3]); /* added to skiplist */
+ server.dirty++;
+ addReply(c,shared.cone);
+ } else {
+ dictEntry *de;
+ double *oldscore;
+
+ /* case 2: Score update operation */
+ de = dictFind(zs->dict,c->argv[3]);
+ assert(de != NULL);
+ oldscore = dictGetEntryVal(de);
+ if (*score != *oldscore) {
+ int deleted;
+
+ deleted = zslDelete(zs->zsl,*oldscore,c->argv[3]);
+ assert(deleted != 0);
+ zslInsert(zs->zsl,*score,c->argv[3]);
+ incrRefCount(c->argv[3]);
+ dictReplace(zs->dict,c->argv[3],score);
+ server.dirty++;
+ } else {
+ zfree(score);
+ }
+ addReply(c,shared.czero);
+ }
+}
+
+static void zremCommand(redisClient *c) {
+ robj *zsetobj;
+ zset *zs;
+
+ zsetobj = lookupKeyWrite(c->db,c->argv[1]);
+ if (zsetobj == NULL) {
+ addReply(c,shared.czero);
+ } else {
+ dictEntry *de;
+ double *oldscore;
+ int deleted;
+
+ if (zsetobj->type != REDIS_ZSET) {
+ addReply(c,shared.wrongtypeerr);
+ return;
+ }
+ zs = zsetobj->ptr;
+ de = dictFind(zs->dict,c->argv[2]);
+ if (de == NULL) {
+ addReply(c,shared.czero);
+ return;
+ }
+ /* Delete from the skiplist */
+ oldscore = dictGetEntryVal(de);
+ deleted = zslDelete(zs->zsl,*oldscore,c->argv[2]);
+ assert(deleted != 0);
+
+ /* Delete from the hash table */
+ dictDelete(zs->dict,c->argv[2]);
+ if (htNeedsResize(zs->dict)) dictResize(zs->dict);
+ server.dirty++;
+ addReply(c,shared.cone);
+ }
+}
+
+static void zremrangebyscoreCommand(redisClient *c) {
+ double min = strtod(c->argv[2]->ptr,NULL);
+ double max = strtod(c->argv[3]->ptr,NULL);
+ robj *zsetobj;
+ zset *zs;
+
+ zsetobj = lookupKeyWrite(c->db,c->argv[1]);
+ if (zsetobj == NULL) {
+ addReply(c,shared.czero);
+ } else {
+ long deleted;
+
+ if (zsetobj->type != REDIS_ZSET) {
+ addReply(c,shared.wrongtypeerr);
+ return;
+ }
+ zs = zsetobj->ptr;
+ deleted = zslDeleteRange(zs->zsl,min,max,zs->dict);
+ if (htNeedsResize(zs->dict)) dictResize(zs->dict);
+ server.dirty += deleted;
+ addReplySds(c,sdscatprintf(sdsempty(),":%lu\r\n",deleted));
+ }
+}
+
+static void zrangeGenericCommand(redisClient *c, int reverse) {
+ robj *o;
+ int start = atoi(c->argv[2]->ptr);
+ int end = atoi(c->argv[3]->ptr);
+
+ o = lookupKeyRead(c->db,c->argv[1]);
+ if (o == NULL) {
+ addReply(c,shared.nullmultibulk);
+ } else {
+ if (o->type != REDIS_ZSET) {
+ addReply(c,shared.wrongtypeerr);
+ } else {
+ zset *zsetobj = o->ptr;
+ zskiplist *zsl = zsetobj->zsl;
+ zskiplistNode *ln;
+
+ int llen = zsl->length;
+ int rangelen, j;
+ robj *ele;
+
+ /* convert negative indexes */
+ if (start < 0) start = llen+start;
+ if (end < 0) end = llen+end;
+ if (start < 0) start = 0;
+ if (end < 0) end = 0;
+
+ /* indexes sanity checks */
+ if (start > end || start >= llen) {
+ /* Out of range start or start > end result in empty list */
+ addReply(c,shared.emptymultibulk);
+ return;
+ }
+ if (end >= llen) end = llen-1;
+ rangelen = (end-start)+1;
+
+ /* Return the result in form of a multi-bulk reply */
+ if (reverse) {
+ ln = zsl->tail;
+ while (start--)
+ ln = ln->backward;
+ } else {
+ ln = zsl->header->forward[0];
+ while (start--)
+ ln = ln->forward[0];
+ }
+
+ addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",rangelen));
+ for (j = 0; j < rangelen; j++) {
+ ele = ln->obj;
+ addReplyBulkLen(c,ele);
+ addReply(c,ele);
+ addReply(c,shared.crlf);
+ ln = reverse ? ln->backward : ln->forward[0];
+ }
+ }
+ }
+}
+
+static void zrangeCommand(redisClient *c) {
+ zrangeGenericCommand(c,0);
+}
+
+static void zrevrangeCommand(redisClient *c) {
+ zrangeGenericCommand(c,1);
+}
+
+static void zrangebyscoreCommand(redisClient *c) {
+ robj *o;
+ double min = strtod(c->argv[2]->ptr,NULL);
+ double max = strtod(c->argv[3]->ptr,NULL);
+
+ o = lookupKeyRead(c->db,c->argv[1]);
+ if (o == NULL) {
+ addReply(c,shared.nullmultibulk);
+ } else {
+ if (o->type != REDIS_ZSET) {
+ addReply(c,shared.wrongtypeerr);
+ } else {
+ zset *zsetobj = o->ptr;
+ zskiplist *zsl = zsetobj->zsl;
+ zskiplistNode *ln;
+ robj *ele, *lenobj;
+ unsigned int rangelen = 0;
+
+ /* Get the first node with the score >= min */
+ ln = zslFirstWithScore(zsl,min);
+ if (ln == NULL) {
+ /* No element matching the speciifed interval */
+ addReply(c,shared.emptymultibulk);
+ return;
+ }
+
+ /* We don't know in advance how many matching elements there
+ * are in the list, so we push this object that will represent
+ * the multi-bulk length in the output buffer, and will "fix"
+ * it later */
+ lenobj = createObject(REDIS_STRING,NULL);
+ addReply(c,lenobj);
+
+ while(ln && ln->score <= max) {
+ ele = ln->obj;
+ addReplyBulkLen(c,ele);
+ addReply(c,ele);
+ addReply(c,shared.crlf);
+ ln = ln->forward[0];
+ rangelen++;
+ }
+ lenobj->ptr = sdscatprintf(sdsempty(),"*%d\r\n",rangelen);
+ }
+ }
+}
+
+static void zcardCommand(redisClient *c) {
+ robj *o;
+ zset *zs;
+
+ o = lookupKeyRead(c->db,c->argv[1]);
+ if (o == NULL) {
+ addReply(c,shared.czero);
+ return;
+ } else {
+ if (o->type != REDIS_ZSET) {
+ addReply(c,shared.wrongtypeerr);
+ } else {
+ zs = o->ptr;
+ addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",zs->zsl->length));
+ }
+ }
+}
+
+static void zscoreCommand(redisClient *c) {
+ robj *o;
+ zset *zs;
+
+ o = lookupKeyRead(c->db,c->argv[1]);
+ if (o == NULL) {
+ addReply(c,shared.nullbulk);
+ return;
+ } else {
+ if (o->type != REDIS_ZSET) {
+ addReply(c,shared.wrongtypeerr);
+ } else {
+ dictEntry *de;
+
+ zs = o->ptr;
+ de = dictFind(zs->dict,c->argv[2]);
+ if (!de) {
+ addReply(c,shared.nullbulk);
+ } else {
+ char buf[128];
+ double *score = dictGetEntryVal(de);
+
+ snprintf(buf,sizeof(buf),"%.17g",*score);
+ addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n%s\r\n",
+ strlen(buf),buf));
+ }
+ }
+ }
+}
+
+/* ========================= Non type-specific commands ==================== */
+
+static void flushdbCommand(redisClient *c) {
+ server.dirty += dictSize(c->db->dict);
+ dictEmpty(c->db->dict);
+ dictEmpty(c->db->expires);
+ addReply(c,shared.ok);
+}
+
+static void flushallCommand(redisClient *c) {
+ server.dirty += emptyDb();
+ addReply(c,shared.ok);
+ rdbSave(server.dbfilename);
+ server.dirty++;
+}
+
+static redisSortOperation *createSortOperation(int type, robj *pattern) {
+ redisSortOperation *so = zmalloc(sizeof(*so));
+ so->type = type;
+ so->pattern = pattern;
+ return so;
+}
+
+/* Return the value associated to the key with a name obtained
+ * substituting the first occurence of '*' in 'pattern' with 'subst' */
+static robj *lookupKeyByPattern(redisDb *db, robj *pattern, robj *subst) {
+ char *p;
+ sds spat, ssub;
+ robj keyobj;
+ int prefixlen, sublen, postfixlen;
+ /* Expoit the internal sds representation to create a sds string allocated on the stack in order to make this function faster */
+ struct {
+ long len;
+ long free;
+ char buf[REDIS_SORTKEY_MAX+1];
+ } keyname;
+
+ if (subst->encoding == REDIS_ENCODING_RAW)
+ incrRefCount(subst);
+ else {
+ subst = getDecodedObject(subst);
+ }
+
+ spat = pattern->ptr;
+ ssub = subst->ptr;
+ if (sdslen(spat)+sdslen(ssub)-1 > REDIS_SORTKEY_MAX) return NULL;
+ p = strchr(spat,'*');
+ if (!p) return NULL;
+
+ prefixlen = p-spat;
+ sublen = sdslen(ssub);
+ postfixlen = sdslen(spat)-(prefixlen+1);
+ memcpy(keyname.buf,spat,prefixlen);
+ memcpy(keyname.buf+prefixlen,ssub,sublen);
+ memcpy(keyname.buf+prefixlen+sublen,p+1,postfixlen);
+ keyname.buf[prefixlen+sublen+postfixlen] = '\0';
+ keyname.len = prefixlen+sublen+postfixlen;
+
+ keyobj.refcount = 1;
+ keyobj.type = REDIS_STRING;
+ keyobj.ptr = ((char*)&keyname)+(sizeof(long)*2);
+
+ decrRefCount(subst);
+
+ /* printf("lookup '%s' => %p\n", keyname.buf,de); */
+ return lookupKeyRead(db,&keyobj);
}
/* sortCompare() is used by qsort in sortCommand(). Given that qsort_r with
}
} else {
/* Compare elements directly */
- cmp = strcoll(so1->obj->ptr,so2->obj->ptr);
+ if (so1->obj->encoding == REDIS_ENCODING_RAW &&
+ so2->obj->encoding == REDIS_ENCODING_RAW) {
+ cmp = strcoll(so1->obj->ptr,so2->obj->ptr);
+ } else {
+ robj *dec1, *dec2;
+
+ dec1 = so1->obj->encoding == REDIS_ENCODING_RAW ?
+ so1->obj : getDecodedObject(so1->obj);
+ dec2 = so2->obj->encoding == REDIS_ENCODING_RAW ?
+ so2->obj : getDecodedObject(so2->obj);
+ cmp = strcoll(dec1->ptr,dec2->ptr);
+ if (dec1 != so1->obj) decrRefCount(dec1);
+ if (dec2 != so2->obj) decrRefCount(dec2);
+ }
}
}
return server.sort_desc ? -cmp : cmp;
int limit_start = 0, limit_count = -1, start, end;
int j, dontsort = 0, vectorlen;
int getop = 0; /* GET operation counter */
- robj *sortval, *sortby = NULL;
+ robj *sortval, *sortby = NULL, *storekey = NULL;
redisSortObject *vector; /* Resulting vector to sort */
/* Lookup the key to sort. It must be of the right types */
limit_start = atoi(c->argv[j+1]->ptr);
limit_count = atoi(c->argv[j+2]->ptr);
j+=2;
+ } else if (!strcasecmp(c->argv[j]->ptr,"store") && leftargs >= 1) {
+ storekey = c->argv[j+1];
+ j++;
} else if (!strcasecmp(c->argv[j]->ptr,"by") && leftargs >= 1) {
sortby = c->argv[j+1];
/* If the BY pattern does not contain '*', i.e. it is constant,
REDIS_SORT_GET,c->argv[j+1]));
getop++;
j++;
- } else if (!strcasecmp(c->argv[j]->ptr,"del") && leftargs >= 1) {
- listAddNodeTail(operations,createSortOperation(
- REDIS_SORT_DEL,c->argv[j+1]));
- j++;
- } else if (!strcasecmp(c->argv[j]->ptr,"incr") && leftargs >= 1) {
- listAddNodeTail(operations,createSortOperation(
- REDIS_SORT_INCR,c->argv[j+1]));
- j++;
- } else if (!strcasecmp(c->argv[j]->ptr,"get") && leftargs >= 1) {
- listAddNodeTail(operations,createSortOperation(
- REDIS_SORT_DECR,c->argv[j+1]));
- j++;
} else {
decrRefCount(sortval);
listRelease(operations);
listLength((list*)sortval->ptr) :
dictSize((dict*)sortval->ptr);
vector = zmalloc(sizeof(redisSortObject)*vectorlen);
- if (!vector) oom("allocating objects vector for SORT");
j = 0;
if (sortval->type == REDIS_LIST) {
list *list = sortval->ptr;
dictEntry *setele;
di = dictGetIterator(set);
- if (!di) oom("dictGetIterator");
while((setele = dictNext(di)) != NULL) {
vector[j].obj = dictGetEntryKey(setele);
vector[j].u.score = 0;
byval = lookupKeyByPattern(c->db,sortby,vector[j].obj);
if (!byval || byval->type != REDIS_STRING) continue;
if (alpha) {
- vector[j].u.cmpobj = byval;
- incrRefCount(byval);
+ if (byval->encoding == REDIS_ENCODING_RAW) {
+ vector[j].u.cmpobj = byval;
+ incrRefCount(byval);
+ } else {
+ vector[j].u.cmpobj = getDecodedObject(byval);
+ }
} else {
- vector[j].u.score = strtod(byval->ptr,NULL);
+ if (byval->encoding == REDIS_ENCODING_RAW) {
+ vector[j].u.score = strtod(byval->ptr,NULL);
+ } else {
+ if (byval->encoding == REDIS_ENCODING_INT) {
+ vector[j].u.score = (long)byval->ptr;
+ } else
+ assert(1 != 1);
+ }
}
} else {
- if (!alpha) vector[j].u.score = strtod(vector[j].obj->ptr,NULL);
+ if (!alpha) {
+ if (vector[j].obj->encoding == REDIS_ENCODING_RAW)
+ vector[j].u.score = strtod(vector[j].obj->ptr,NULL);
+ else {
+ if (vector[j].obj->encoding == REDIS_ENCODING_INT)
+ vector[j].u.score = (long) vector[j].obj->ptr;
+ else
+ assert(1 != 1);
+ }
+ }
}
}
}
server.sort_desc = desc;
server.sort_alpha = alpha;
server.sort_bypattern = sortby ? 1 : 0;
- qsort(vector,vectorlen,sizeof(redisSortObject),sortCompare);
+ if (sortby && (start != 0 || end != vectorlen-1))
+ pqsort(vector,vectorlen,sizeof(redisSortObject),sortCompare, start,end);
+ else
+ qsort(vector,vectorlen,sizeof(redisSortObject),sortCompare);
}
/* Send command output to the output buffer, performing the specified
* GET/DEL/INCR/DECR operations if any. */
outputlen = getop ? getop*(end-start+1) : end-start+1;
- addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",outputlen));
- for (j = start; j <= end; j++) {
- listNode *ln;
- if (!getop) {
- addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",
- sdslen(vector[j].obj->ptr)));
- addReply(c,vector[j].obj);
- addReply(c,shared.crlf);
+ if (storekey == NULL) {
+ /* STORE option not specified, sent the sorting result to client */
+ addReplySds(c,sdscatprintf(sdsempty(),"*%d\r\n",outputlen));
+ for (j = start; j <= end; j++) {
+ listNode *ln;
+ if (!getop) {
+ addReplyBulkLen(c,vector[j].obj);
+ addReply(c,vector[j].obj);
+ addReply(c,shared.crlf);
+ }
+ listRewind(operations);
+ while((ln = listYield(operations))) {
+ redisSortOperation *sop = ln->value;
+ robj *val = lookupKeyByPattern(c->db,sop->pattern,
+ vector[j].obj);
+
+ if (sop->type == REDIS_SORT_GET) {
+ if (!val || val->type != REDIS_STRING) {
+ addReply(c,shared.nullbulk);
+ } else {
+ addReplyBulkLen(c,val);
+ addReply(c,val);
+ addReply(c,shared.crlf);
+ }
+ } else {
+ assert(sop->type == REDIS_SORT_GET); /* always fails */
+ }
+ }
}
- listRewind(operations);
- while((ln = listYield(operations))) {
- redisSortOperation *sop = ln->value;
- robj *val = lookupKeyByPattern(c->db,sop->pattern,
- vector[j].obj);
+ } else {
+ robj *listObject = createListObject();
+ list *listPtr = (list*) listObject->ptr;
- if (sop->type == REDIS_SORT_GET) {
- if (!val || val->type != REDIS_STRING) {
- addReply(c,shared.nullbulk);
+ /* STORE option specified, set the sorting result as a List object */
+ for (j = start; j <= end; j++) {
+ listNode *ln;
+ if (!getop) {
+ listAddNodeTail(listPtr,vector[j].obj);
+ incrRefCount(vector[j].obj);
+ }
+ listRewind(operations);
+ while((ln = listYield(operations))) {
+ redisSortOperation *sop = ln->value;
+ robj *val = lookupKeyByPattern(c->db,sop->pattern,
+ vector[j].obj);
+
+ if (sop->type == REDIS_SORT_GET) {
+ if (!val || val->type != REDIS_STRING) {
+ listAddNodeTail(listPtr,createStringObject("",0));
+ } else {
+ listAddNodeTail(listPtr,val);
+ incrRefCount(val);
+ }
} else {
- addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",
- sdslen(val->ptr)));
- addReply(c,val);
- addReply(c,shared.crlf);
+ assert(sop->type == REDIS_SORT_GET); /* always fails */
}
- } else if (sop->type == REDIS_SORT_DEL) {
- /* TODO */
}
}
+ if (dictReplace(c->db->dict,storekey,listObject)) {
+ incrRefCount(storekey);
+ }
+ /* Note: we add 1 because the DB is dirty anyway since even if the
+ * SORT result is empty a new key is set and maybe the old content
+ * replaced. */
+ server.dirty += 1+outputlen;
+ addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",outputlen));
}
/* Cleanup */
static void infoCommand(redisClient *c) {
sds info;
time_t uptime = time(NULL)-server.stat_starttime;
+ int j;
info = sdscatprintf(sdsempty(),
"redis_version:%s\r\n"
+ "arch_bits:%s\r\n"
+ "uptime_in_seconds:%d\r\n"
+ "uptime_in_days:%d\r\n"
"connected_clients:%d\r\n"
"connected_slaves:%d\r\n"
"used_memory:%zu\r\n"
"last_save_time:%d\r\n"
"total_connections_received:%lld\r\n"
"total_commands_processed:%lld\r\n"
- "uptime_in_seconds:%d\r\n"
- "uptime_in_days:%d\r\n"
+ "role:%s\r\n"
,REDIS_VERSION,
+ (sizeof(long) == 8) ? "64" : "32",
+ uptime,
+ uptime/(3600*24),
listLength(server.clients)-listLength(server.slaves),
listLength(server.slaves),
server.usedmemory,
server.lastsave,
server.stat_numconnections,
server.stat_numcommands,
- uptime,
- uptime/(3600*24)
+ server.masterhost == NULL ? "master" : "slave"
);
+ if (server.masterhost) {
+ info = sdscatprintf(info,
+ "master_host:%s\r\n"
+ "master_port:%d\r\n"
+ "master_link_status:%s\r\n"
+ "master_last_io_seconds_ago:%d\r\n"
+ ,server.masterhost,
+ server.masterport,
+ (server.replstate == REDIS_REPL_CONNECTED) ?
+ "up" : "down",
+ server.master ? ((int)(time(NULL)-server.master->lastinteraction)) : -1
+ );
+ }
+ for (j = 0; j < server.dbnum; j++) {
+ long long keys, vkeys;
+
+ keys = dictSize(server.db[j].dict);
+ vkeys = dictSize(server.db[j].expires);
+ if (keys || vkeys) {
+ info = sdscatprintf(info, "db%d: keys=%lld,expires=%lld\r\n",
+ j, keys, vkeys);
+ }
+ }
addReplySds(c,sdscatprintf(sdsempty(),"$%d\r\n",sdslen(info)));
addReplySds(c,info);
addReply(c,shared.crlf);
c->flags |= (REDIS_SLAVE|REDIS_MONITOR);
c->slaveseldb = 0;
- if (!listAddNodeTail(server.monitors,c)) oom("listAddNodeTail");
+ listAddNodeTail(server.monitors,c);
addReply(c,shared.ok);
}
return dictDelete(db->dict,key) == DICT_OK;
}
-static void expireCommand(redisClient *c) {
+static void expireGenericCommand(redisClient *c, robj *key, time_t seconds) {
dictEntry *de;
- int seconds = atoi(c->argv[2]->ptr);
- de = dictFind(c->db->dict,c->argv[1]);
+ de = dictFind(c->db->dict,key);
if (de == NULL) {
addReply(c,shared.czero);
return;
}
- if (seconds <= 0) {
- addReply(c, shared.czero);
+ if (seconds < 0) {
+ if (deleteKey(c->db,key)) server.dirty++;
+ addReply(c, shared.cone);
return;
} else {
time_t when = time(NULL)+seconds;
- if (setExpire(c->db,c->argv[1],when))
+ if (setExpire(c->db,key,when)) {
addReply(c,shared.cone);
- else
+ server.dirty++;
+ } else {
addReply(c,shared.czero);
+ }
return;
}
}
+static void expireCommand(redisClient *c) {
+ expireGenericCommand(c,c->argv[1],strtol(c->argv[2]->ptr,NULL,10));
+}
+
+static void expireatCommand(redisClient *c) {
+ expireGenericCommand(c,c->argv[1],strtol(c->argv[2]->ptr,NULL,10)-time(NULL));
+}
+
+static void ttlCommand(redisClient *c) {
+ time_t expire;
+ int ttl = -1;
+
+ expire = getExpire(c->db,c->argv[1]);
+ if (expire != -1) {
+ ttl = (int) (expire-time(NULL));
+ if (ttl < 0) ttl = -1;
+ }
+ addReplySds(c,sdscatprintf(sdsempty(),":%d\r\n",ttl));
+}
+
+static void msetGenericCommand(redisClient *c, int nx) {
+ int j;
+
+ if ((c->argc % 2) == 0) {
+ addReplySds(c,sdsnew("-ERR wrong number of arguments\r\n"));
+ return;
+ }
+ /* Handle the NX flag. The MSETNX semantic is to return zero and don't
+ * set nothing at all if at least one already key exists. */
+ if (nx) {
+ for (j = 1; j < c->argc; j += 2) {
+ if (dictFind(c->db->dict,c->argv[j]) != NULL) {
+ addReply(c, shared.czero);
+ return;
+ }
+ }
+ }
+
+ for (j = 1; j < c->argc; j += 2) {
+ int retval;
+
+ retval = dictAdd(c->db->dict,c->argv[j],c->argv[j+1]);
+ if (retval == DICT_ERR) {
+ dictReplace(c->db->dict,c->argv[j],c->argv[j+1]);
+ incrRefCount(c->argv[j+1]);
+ } else {
+ incrRefCount(c->argv[j]);
+ incrRefCount(c->argv[j+1]);
+ }
+ removeExpire(c->db,c->argv[j]);
+ }
+ server.dirty += (c->argc-1)/2;
+ addReply(c, nx ? shared.cone : shared.ok);
+}
+
+static void msetCommand(redisClient *c) {
+ msetGenericCommand(c,0);
+}
+
+static void msetnxCommand(redisClient *c) {
+ msetGenericCommand(c,1);
+}
+
/* =============================== Replication ============================= */
static int syncWrite(int fd, char *ptr, ssize_t size, int timeout) {
* another slave. Set the right state, and copy the buffer. */
listRelease(c->reply);
c->reply = listDup(slave->reply);
- if (!c->reply) oom("listDup copying slave reply list");
c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
redisLog(REDIS_NOTICE,"Waiting for end of BGSAVE for SYNC");
} else {
c->repldbfd = -1;
c->flags |= REDIS_SLAVE;
c->slaveseldb = 0;
- if (!listAddNodeTail(server.slaves,c)) oom("listAddNodeTail");
+ listAddNodeTail(server.slaves,c);
return;
}
}
}
-static void updateSalvesWaitingBgsave(int bgsaveerr) {
+/* This function is called at the end of every backgrond saving.
+ * The argument bgsaveerr is REDIS_OK if the background saving succeeded
+ * otherwise REDIS_ERR is passed to the function.
+ *
+ * The goal of this function is to handle slaves waiting for a successful
+ * background saving in order to perform non-blocking synchronization. */
+static void updateSlavesWaitingBgsave(int bgsaveerr) {
listNode *ln;
int startbgsave = 0;
startbgsave = 1;
slave->replstate = REDIS_REPL_WAIT_BGSAVE_END;
} else if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) {
- struct stat buf;
+ struct redis_stat buf;
if (bgsaveerr != REDIS_OK) {
freeClient(slave);
continue;
}
if ((slave->repldbfd = open(server.dbfilename,O_RDONLY)) == -1 ||
- fstat(slave->repldbfd,&buf) == -1) {
+ redis_fstat(slave->repldbfd,&buf) == -1) {
freeClient(slave);
redisLog(REDIS_WARNING,"SYNC failed. Can't open/stat DB after BGSAVE: %s", strerror(errno));
continue;
}
static int syncWithMaster(void) {
- char buf[1024], tmpfile[256];
+ char buf[1024], tmpfile[256], authcmd[1024];
int dumpsize;
int fd = anetTcpConnect(NULL,server.masterhost,server.masterport);
int dfd;
strerror(errno));
return REDIS_ERR;
}
+
+ /* AUTH with the master if required. */
+ if(server.masterauth) {
+ snprintf(authcmd, 1024, "AUTH %s\r\n", server.masterauth);
+ if (syncWrite(fd, authcmd, strlen(server.masterauth)+7, 5) == -1) {
+ close(fd);
+ redisLog(REDIS_WARNING,"Unable to AUTH to MASTER: %s",
+ strerror(errno));
+ return REDIS_ERR;
+ }
+ /* Read the AUTH result. */
+ if (syncReadLine(fd,buf,1024,3600) == -1) {
+ close(fd);
+ redisLog(REDIS_WARNING,"I/O error reading auth result from MASTER: %s",
+ strerror(errno));
+ return REDIS_ERR;
+ }
+ if (buf[0] != '+') {
+ close(fd);
+ redisLog(REDIS_WARNING,"Cannot AUTH to MASTER, is the masterauth password correct?");
+ return REDIS_ERR;
+ }
+ }
+
/* Issue the SYNC command */
if (syncWrite(fd,"SYNC \r\n",7,5) == -1) {
close(fd);
return REDIS_ERR;
}
/* Read the bulk write count */
- if (syncReadLine(fd,buf,1024,5) == -1) {
+ if (syncReadLine(fd,buf,1024,3600) == -1) {
close(fd);
redisLog(REDIS_WARNING,"I/O error reading bulk count from MASTER: %s",
strerror(errno));
return REDIS_ERR;
}
+ if (buf[0] != '$') {
+ close(fd);
+ redisLog(REDIS_WARNING,"Bad protocol from MASTER, the first byte is not '$', are you sure the host and port are right?");
+ return REDIS_ERR;
+ }
dumpsize = atoi(buf+1);
redisLog(REDIS_NOTICE,"Receiving %d bytes data dump from MASTER",dumpsize);
/* Read the bulk write data on a temp file */
return REDIS_OK;
}
+static void slaveofCommand(redisClient *c) {
+ if (!strcasecmp(c->argv[1]->ptr,"no") &&
+ !strcasecmp(c->argv[2]->ptr,"one")) {
+ if (server.masterhost) {
+ sdsfree(server.masterhost);
+ server.masterhost = NULL;
+ if (server.master) freeClient(server.master);
+ server.replstate = REDIS_REPL_NONE;
+ redisLog(REDIS_NOTICE,"MASTER MODE enabled (user request)");
+ }
+ } else {
+ sdsfree(server.masterhost);
+ server.masterhost = sdsdup(c->argv[1]->ptr);
+ server.masterport = atoi(c->argv[2]->ptr);
+ if (server.master) freeClient(server.master);
+ server.replstate = REDIS_REPL_CONNECT;
+ redisLog(REDIS_NOTICE,"SLAVE OF %s:%d enabled (user request)",
+ server.masterhost, server.masterport);
+ }
+ addReply(c,shared.ok);
+}
+
+/* ============================ Maxmemory directive ======================== */
+
+/* This function gets called when 'maxmemory' is set on the config file to limit
+ * the max memory used by the server, and we are out of memory.
+ * This function will try to, in order:
+ *
+ * - Free objects from the free list
+ * - Try to remove keys with an EXPIRE set
+ *
+ * It is not possible to free enough memory to reach used-memory < maxmemory
+ * the server will start refusing commands that will enlarge even more the
+ * memory usage.
+ */
+static void freeMemoryIfNeeded(void) {
+ while (server.maxmemory && zmalloc_used_memory() > server.maxmemory) {
+ if (listLength(server.objfreelist)) {
+ robj *o;
+
+ listNode *head = listFirst(server.objfreelist);
+ o = listNodeValue(head);
+ listDelNode(server.objfreelist,head);
+ zfree(o);
+ } else {
+ int j, k, freed = 0;
+
+ for (j = 0; j < server.dbnum; j++) {
+ int minttl = -1;
+ robj *minkey = NULL;
+ struct dictEntry *de;
+
+ if (dictSize(server.db[j].expires)) {
+ freed = 1;
+ /* From a sample of three keys drop the one nearest to
+ * the natural expire */
+ for (k = 0; k < 3; k++) {
+ time_t t;
+
+ de = dictGetRandomKey(server.db[j].expires);
+ t = (time_t) dictGetEntryVal(de);
+ if (minttl == -1 || t < minttl) {
+ minkey = dictGetEntryKey(de);
+ minttl = t;
+ }
+ }
+ deleteKey(server.db+j,minkey);
+ }
+ }
+ if (!freed) return; /* nothing to free... */
+ }
+ }
+}
+
+/* ============================== Append Only file ========================== */
+
+static void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc) {
+ sds buf = sdsempty();
+ int j;
+ ssize_t nwritten;
+ time_t now;
+ robj *tmpargv[3];
+
+ /* The DB this command was targetting is not the same as the last command
+ * we appendend. To issue a SELECT command is needed. */
+ if (dictid != server.appendseldb) {
+ char seldb[64];
+
+ snprintf(seldb,sizeof(seldb),"%d",dictid);
+ buf = sdscatprintf(buf,"*2\r\n$6\r\nSELECT\r\n$%d\r\n%s\r\n",
+ strlen(seldb),seldb);
+ server.appendseldb = dictid;
+ }
+
+ /* "Fix" the argv vector if the command is EXPIRE. We want to translate
+ * EXPIREs into EXPIREATs calls */
+ if (cmd->proc == expireCommand) {
+ long when;
+
+ tmpargv[0] = createStringObject("EXPIREAT",8);
+ tmpargv[1] = argv[1];
+ incrRefCount(argv[1]);
+ when = time(NULL)+strtol(argv[2]->ptr,NULL,10);
+ tmpargv[2] = createObject(REDIS_STRING,
+ sdscatprintf(sdsempty(),"%ld",when));
+ argv = tmpargv;
+ }
+
+ /* Append the actual command */
+ buf = sdscatprintf(buf,"*%d\r\n",argc);
+ for (j = 0; j < argc; j++) {
+ robj *o = argv[j];
+
+ if (o->encoding != REDIS_ENCODING_RAW)
+ o = getDecodedObject(o);
+ buf = sdscatprintf(buf,"$%d\r\n",sdslen(o->ptr));
+ buf = sdscatlen(buf,o->ptr,sdslen(o->ptr));
+ buf = sdscatlen(buf,"\r\n",2);
+ if (o != argv[j])
+ decrRefCount(o);
+ }
+
+ /* Free the objects from the modified argv for EXPIREAT */
+ if (cmd->proc == expireCommand) {
+ for (j = 0; j < 3; j++)
+ decrRefCount(argv[j]);
+ }
+
+ /* We want to perform a single write. This should be guaranteed atomic
+ * at least if the filesystem we are writing is a real physical one.
+ * While this will save us against the server being killed I don't think
+ * there is much to do about the whole server stopping for power problems
+ * or alike */
+ nwritten = write(server.appendfd,buf,sdslen(buf));
+ if (nwritten != (signed)sdslen(buf)) {
+ /* Ooops, we are in troubles. The best thing to do for now is
+ * to simply exit instead to give the illusion that everything is
+ * working as expected. */
+ if (nwritten == -1) {
+ redisLog(REDIS_WARNING,"Exiting on error writing to the append-only file: %s",strerror(errno));
+ } else {
+ redisLog(REDIS_WARNING,"Exiting on short write while writing to the append-only file: %s",strerror(errno));
+ }
+ exit(1);
+ }
+ now = time(NULL);
+ if (server.appendfsync == APPENDFSYNC_ALWAYS ||
+ (server.appendfsync == APPENDFSYNC_EVERYSEC &&
+ now-server.lastfsync > 1))
+ {
+ fsync(server.appendfd); /* Let's try to get this data on the disk */
+ server.lastfsync = now;
+ }
+}
+
+/* In Redis commands are always executed in the context of a client, so in
+ * order to load the append only file we need to create a fake client. */
+static struct redisClient *createFakeClient(void) {
+ struct redisClient *c = zmalloc(sizeof(*c));
+
+ selectDb(c,0);
+ c->fd = -1;
+ c->querybuf = sdsempty();
+ c->argc = 0;
+ c->argv = NULL;
+ c->flags = 0;
+ /* We set the fake client as a slave waiting for the synchronization
+ * so that Redis will not try to send replies to this client. */
+ c->replstate = REDIS_REPL_WAIT_BGSAVE_START;
+ c->reply = listCreate();
+ listSetFreeMethod(c->reply,decrRefCount);
+ listSetDupMethod(c->reply,dupClientReplyValue);
+ return c;
+}
+
+static void freeFakeClient(struct redisClient *c) {
+ sdsfree(c->querybuf);
+ listRelease(c->reply);
+ zfree(c);
+}
+
+/* Replay the append log file. On error REDIS_OK is returned. On non fatal
+ * error (the append only file is zero-length) REDIS_ERR is returned. On
+ * fatal error an error message is logged and the program exists. */
+int loadAppendOnlyFile(char *filename) {
+ struct redisClient *fakeClient;
+ FILE *fp = fopen(filename,"r");
+ struct redis_stat sb;
+
+ if (redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0)
+ return REDIS_ERR;
+
+ if (fp == NULL) {
+ redisLog(REDIS_WARNING,"Fatal error: can't open the append log file for reading: %s",strerror(errno));
+ exit(1);
+ }
+
+ fakeClient = createFakeClient();
+ while(1) {
+ int argc, j;
+ unsigned long len;
+ robj **argv;
+ char buf[128];
+ sds argsds;
+ struct redisCommand *cmd;
+
+ if (fgets(buf,sizeof(buf),fp) == NULL) {
+ if (feof(fp))
+ break;
+ else
+ goto readerr;
+ }
+ if (buf[0] != '*') goto fmterr;
+ argc = atoi(buf+1);
+ argv = zmalloc(sizeof(robj*)*argc);
+ for (j = 0; j < argc; j++) {
+ if (fgets(buf,sizeof(buf),fp) == NULL) goto readerr;
+ if (buf[0] != '$') goto fmterr;
+ len = strtol(buf+1,NULL,10);
+ argsds = sdsnewlen(NULL,len);
+ if (fread(argsds,len,1,fp) == 0) goto fmterr;
+ argv[j] = createObject(REDIS_STRING,argsds);
+ if (fread(buf,2,1,fp) == 0) goto fmterr; /* discard CRLF */
+ }
+
+ /* Command lookup */
+ cmd = lookupCommand(argv[0]->ptr);
+ if (!cmd) {
+ redisLog(REDIS_WARNING,"Unknown command '%s' reading the append only file", argv[0]->ptr);
+ exit(1);
+ }
+ /* Try object sharing and encoding */
+ if (server.shareobjects) {
+ int j;
+ for(j = 1; j < argc; j++)
+ argv[j] = tryObjectSharing(argv[j]);
+ }
+ if (cmd->flags & REDIS_CMD_BULK)
+ tryObjectEncoding(argv[argc-1]);
+ /* Run the command in the context of a fake client */
+ fakeClient->argc = argc;
+ fakeClient->argv = argv;
+ cmd->proc(fakeClient);
+ /* Discard the reply objects list from the fake client */
+ while(listLength(fakeClient->reply))
+ listDelNode(fakeClient->reply,listFirst(fakeClient->reply));
+ /* Clean up, ready for the next command */
+ for (j = 0; j < argc; j++) decrRefCount(argv[j]);
+ zfree(argv);
+ }
+ fclose(fp);
+ freeFakeClient(fakeClient);
+ return REDIS_OK;
+
+readerr:
+ if (feof(fp)) {
+ redisLog(REDIS_WARNING,"Unexpected end of file reading the append only file");
+ } else {
+ redisLog(REDIS_WARNING,"Unrecoverable error reading the append only file: %s", strerror(errno));
+ }
+ exit(1);
+fmterr:
+ redisLog(REDIS_WARNING,"Bad file format reading the append only file");
+ exit(1);
+}
+
+/* ================================= Debugging ============================== */
+
+static void debugCommand(redisClient *c) {
+ if (!strcasecmp(c->argv[1]->ptr,"segfault")) {
+ *((char*)-1) = 'x';
+ } else if (!strcasecmp(c->argv[1]->ptr,"object") && c->argc == 3) {
+ dictEntry *de = dictFind(c->db->dict,c->argv[2]);
+ robj *key, *val;
+
+ if (!de) {
+ addReply(c,shared.nokeyerr);
+ return;
+ }
+ key = dictGetEntryKey(de);
+ val = dictGetEntryVal(de);
+ addReplySds(c,sdscatprintf(sdsempty(),
+ "+Key at:%p refcount:%d, value at:%p refcount:%d encoding:%d\r\n",
+ key, key->refcount, val, val->refcount, val->encoding));
+ } else {
+ addReplySds(c,sdsnew(
+ "-ERR Syntax error, try DEBUG [SEGFAULT|OBJECT <key>]\r\n"));
+ }
+}
+
/* =================================== Main! ================================ */
+#ifdef __linux__
+int linuxOvercommitMemoryValue(void) {
+ FILE *fp = fopen("/proc/sys/vm/overcommit_memory","r");
+ char buf[64];
+
+ if (!fp) return -1;
+ if (fgets(buf,64,fp) == NULL) {
+ fclose(fp);
+ return -1;
+ }
+ fclose(fp);
+
+ return atoi(buf);
+}
+
+void linuxOvercommitMemoryWarning(void) {
+ if (linuxOvercommitMemoryValue() == 0) {
+ redisLog(REDIS_WARNING,"WARNING overcommit_memory is set to 0! Background save may fail under low condition memory. To fix this issue add 'vm.overcommit_memory = 1' to /etc/sysctl.conf and then reboot or run the command 'sysctl vm.overcommit_memory=1' for this to take effect.");
+ }
+}
+#endif /* __linux__ */
+
static void daemonize(void) {
int fd;
FILE *fp;
int main(int argc, char **argv) {
initServerConfig();
if (argc == 2) {
- ResetServerSaveParams();
+ resetServerSaveParams();
loadServerConfig(argv[1]);
} else if (argc > 2) {
fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf]\n");
exit(1);
+ } else {
+ redisLog(REDIS_WARNING,"Warning: no config file specified, using the default config. In order to specify a config file use 'redis-server /path/to/redis.conf'");
}
initServer();
if (server.daemonize) daemonize();
redisLog(REDIS_NOTICE,"Server started, Redis version " REDIS_VERSION);
- if (rdbLoad(server.dbfilename) == REDIS_OK)
- redisLog(REDIS_NOTICE,"DB loaded from disk");
+#ifdef __linux__
+ linuxOvercommitMemoryWarning();
+#endif
+ if (server.appendonly) {
+ if (loadAppendOnlyFile(server.appendfilename) == REDIS_OK)
+ redisLog(REDIS_NOTICE,"DB loaded from append only file");
+ } else {
+ if (rdbLoad(server.dbfilename) == REDIS_OK)
+ redisLog(REDIS_NOTICE,"DB loaded from disk");
+ }
if (aeCreateFileEvent(server.el, server.fd, AE_READABLE,
acceptHandler, NULL, NULL) == AE_ERR) oom("creating file event");
redisLog(REDIS_NOTICE,"The server is now ready to accept connections on port %d", server.port);
aeDeleteEventLoop(server.el);
return 0;
}
+
+/* ============================= Backtrace support ========================= */
+
+#ifdef HAVE_BACKTRACE
+static char *findFuncName(void *pointer, unsigned long *offset);
+
+static void *getMcontextEip(ucontext_t *uc) {
+#if defined(__FreeBSD__)
+ return (void*) uc->uc_mcontext.mc_eip;
+#elif defined(__dietlibc__)
+ return (void*) uc->uc_mcontext.eip;
+#elif defined(__APPLE__) && !defined(MAC_OS_X_VERSION_10_6)
+ return (void*) uc->uc_mcontext->__ss.__eip;
+#elif defined(__APPLE__) && defined(MAC_OS_X_VERSION_10_6)
+ #if defined(_STRUCT_X86_THREAD_STATE64) && !defined(__i386__)
+ return (void*) uc->uc_mcontext->__ss.__rip;
+ #else
+ return (void*) uc->uc_mcontext->__ss.__eip;
+ #endif
+#elif defined(__i386__) || defined(__X86_64__) /* Linux x86 */
+ return (void*) uc->uc_mcontext.gregs[REG_EIP];
+#elif defined(__ia64__) /* Linux IA64 */
+ return (void*) uc->uc_mcontext.sc_ip;
+#else
+ return NULL;
+#endif
+}
+
+static void segvHandler(int sig, siginfo_t *info, void *secret) {
+ void *trace[100];
+ char **messages = NULL;
+ int i, trace_size = 0;
+ unsigned long offset=0;
+ time_t uptime = time(NULL)-server.stat_starttime;
+ ucontext_t *uc = (ucontext_t*) secret;
+ REDIS_NOTUSED(info);
+
+ redisLog(REDIS_WARNING,
+ "======= Ooops! Redis %s got signal: -%d- =======", REDIS_VERSION, sig);
+ redisLog(REDIS_WARNING, "%s", sdscatprintf(sdsempty(),
+ "redis_version:%s; "
+ "uptime_in_seconds:%d; "
+ "connected_clients:%d; "
+ "connected_slaves:%d; "
+ "used_memory:%zu; "
+ "changes_since_last_save:%lld; "
+ "bgsave_in_progress:%d; "
+ "last_save_time:%d; "
+ "total_connections_received:%lld; "
+ "total_commands_processed:%lld; "
+ "role:%s;"
+ ,REDIS_VERSION,
+ uptime,
+ listLength(server.clients)-listLength(server.slaves),
+ listLength(server.slaves),
+ server.usedmemory,
+ server.dirty,
+ server.bgsaveinprogress,
+ server.lastsave,
+ server.stat_numconnections,
+ server.stat_numcommands,
+ server.masterhost == NULL ? "master" : "slave"
+ ));
+
+ trace_size = backtrace(trace, 100);
+ /* overwrite sigaction with caller's address */
+ if (getMcontextEip(uc) != NULL) {
+ trace[1] = getMcontextEip(uc);
+ }
+ messages = backtrace_symbols(trace, trace_size);
+
+ for (i=1; i<trace_size; ++i) {
+ char *fn = findFuncName(trace[i], &offset), *p;
+
+ p = strchr(messages[i],'+');
+ if (!fn || (p && ((unsigned long)strtol(p+1,NULL,10)) < offset)) {
+ redisLog(REDIS_WARNING,"%s", messages[i]);
+ } else {
+ redisLog(REDIS_WARNING,"%d redis-server %p %s + %d", i, trace[i], fn, (unsigned int)offset);
+ }
+ }
+ free(messages);
+ exit(0);
+}
+
+static void setupSigSegvAction(void) {
+ struct sigaction act;
+
+ sigemptyset (&act.sa_mask);
+ /* When the SA_SIGINFO flag is set in sa_flags then sa_sigaction
+ * is used. Otherwise, sa_handler is used */
+ act.sa_flags = SA_NODEFER | SA_ONSTACK | SA_RESETHAND | SA_SIGINFO;
+ act.sa_sigaction = segvHandler;
+ sigaction (SIGSEGV, &act, NULL);
+ sigaction (SIGBUS, &act, NULL);
+ sigaction (SIGFPE, &act, NULL);
+ sigaction (SIGILL, &act, NULL);
+ sigaction (SIGBUS, &act, NULL);
+ return;
+}
+
+#include "staticsymbols.h"
+/* This function try to convert a pointer into a function name. It's used in
+ * oreder to provide a backtrace under segmentation fault that's able to
+ * display functions declared as static (otherwise the backtrace is useless). */
+static char *findFuncName(void *pointer, unsigned long *offset){
+ int i, ret = -1;
+ unsigned long off, minoff = 0;
+
+ /* Try to match against the Symbol with the smallest offset */
+ for (i=0; symsTable[i].pointer; i++) {
+ unsigned long lp = (unsigned long) pointer;
+
+ if (lp != (unsigned long)-1 && lp >= symsTable[i].pointer) {
+ off=lp-symsTable[i].pointer;
+ if (ret < 0 || off < minoff) {
+ minoff=off;
+ ret=i;
+ }
+ }
+ }
+ if (ret == -1) return NULL;
+ *offset = minoff;
+ return symsTable[ret].name;
+}
+#else /* HAVE_BACKTRACE */
+static void setupSigSegvAction(void) {
+}
+#endif /* HAVE_BACKTRACE */
+
+
+
+/* The End */
+
+
+
projects / redis.git / blobdiff