fixed a few problems with redis-trib resharding.

[redis.git] / src / networking.c

diff --git a/src/networking.c b/src/networking.c
index 89613cedf5c8183c23caec05f5023e5e5bd57f48..3979ab6225c4ca24ca64f557db46e03a0edc6f9f 100644 (file)
--- a/src/networking.c
+++ b/src/networking.c
@@ -11,29 +11,34 @@ int listMatchObjects(void *a, void *b) {
 }
 
 redisClient *createClient(int fd) {
 }
 
 redisClient *createClient(int fd) {

-    redisClient *c;
-
-    /* Make sure to allocate a multiple of the page size to prevent wasting
-     * memory. A page size of 4096 is assumed here. We need to compensate
-     * for the zmalloc overhead of sizeof(size_t) bytes. */
-    size_t size = 8192-sizeof(size_t);
-    redisAssert(size > sizeof(redisClient));
-    c = zmalloc(size);
-    c->buflen = size-sizeof(redisClient);
+    redisClient *c = zmalloc(sizeof(redisClient));

     c->bufpos = 0;
 
     c->bufpos = 0;
 

-    anetNonBlock(NULL,fd);
-    anetTcpNoDelay(NULL,fd);
-    if (!c) return NULL;
+    /* passing -1 as fd it is possible to create a non connected client.
+     * This is useful since all the Redis commands needs to be executed
+     * in the context of a client. When commands are executed in other
+     * contexts (for instance a Lua script) we need a non connected client. */
+    if (fd != -1) {
+        anetNonBlock(NULL,fd);
+        anetTcpNoDelay(NULL,fd);
+        if (aeCreateFileEvent(server.el,fd,AE_READABLE,
+            readQueryFromClient, c) == AE_ERR)
+        {
+            close(fd);
+            zfree(c);
+            return NULL;
+        }
+    }
+

     selectDb(c,0);
     c->fd = fd;
     c->querybuf = sdsempty();
     selectDb(c,0);
     c->fd = fd;
     c->querybuf = sdsempty();

+    c->reqtype = 0;

     c->argc = 0;
     c->argv = NULL;
     c->argc = 0;
     c->argv = NULL;

+    c->cmd = NULL;
+    c->multibulklen = 0;

     c->bulklen = -1;
     c->bulklen = -1;

-    c->multibulk = 0;
-    c->mbargc = 0;
-    c->mbargv = NULL;

     c->sentlen = 0;
     c->flags = 0;
     c->lastinteraction = time(NULL);
     c->sentlen = 0;
     c->flags = 0;
     c->lastinteraction = time(NULL);


@@ -42,8 +47,10 @@ redisClient *createClient(int fd) {
     c->reply = listCreate();
     listSetFreeMethod(c->reply,decrRefCount);
     listSetDupMethod(c->reply,dupClientReplyValue);
     c->reply = listCreate();
     listSetFreeMethod(c->reply,decrRefCount);
     listSetDupMethod(c->reply,dupClientReplyValue);

-    c->blocking_keys = NULL;
-    c->blocking_keys_num = 0;
+    c->bpop.keys = NULL;
+    c->bpop.count = 0;
+    c->bpop.timeout = 0;
+    c->bpop.target = NULL;

     c->io_keys = listCreate();
     c->watched_keys = listCreate();
     listSetFreeMethod(c->io_keys,decrRefCount);
     c->io_keys = listCreate();
     c->watched_keys = listCreate();
     listSetFreeMethod(c->io_keys,decrRefCount);


@@ -51,17 +58,15 @@ redisClient *createClient(int fd) {
     c->pubsub_patterns = listCreate();
     listSetFreeMethod(c->pubsub_patterns,decrRefCount);
     listSetMatchMethod(c->pubsub_patterns,listMatchObjects);
     c->pubsub_patterns = listCreate();
     listSetFreeMethod(c->pubsub_patterns,decrRefCount);
     listSetMatchMethod(c->pubsub_patterns,listMatchObjects);

-    if (aeCreateFileEvent(server.el, c->fd, AE_READABLE,
-        readQueryFromClient, c) == AE_ERR) {
-        freeClient(c);
-        return NULL;
-    }
-    listAddNodeTail(server.clients,c);
+    if (fd != -1) listAddNodeTail(server.clients,c);

     initClientMultiState(c);
     return c;
 }
 
     initClientMultiState(c);
     return c;
 }
 

-int _ensureFileEvent(redisClient *c) {
+/* Set the event loop to listen for write events on the client's socket.
+ * Typically gets called every time a reply is built. */
+int _installWriteEvent(redisClient *c) {
+    if (c->flags & REDIS_LUA_CLIENT) return REDIS_OK;

     if (c->fd <= 0) return REDIS_ERR;
     if (c->bufpos == 0 && listLength(c->reply) == 0 &&
         (c->replstate == REDIS_REPL_NONE ||
     if (c->fd <= 0) return REDIS_ERR;
     if (c->bufpos == 0 && listLength(c->reply) == 0 &&
         (c->replstate == REDIS_REPL_NONE ||


@@ -71,90 +76,219 @@ int _ensureFileEvent(redisClient *c) {
     return REDIS_OK;
 }
 
     return REDIS_OK;
 }
 

-void _addReplyObjectToList(redisClient *c, robj *obj) {
-    redisAssert(obj->type == REDIS_STRING &&
-                obj->encoding == REDIS_ENCODING_RAW);
-    listAddNodeTail(c->reply,obj);
+/* Create a duplicate of the last object in the reply list when
+ * it is not exclusively owned by the reply list. */
+robj *dupLastObjectIfNeeded(list *reply) {
+    robj *new, *cur;
+    listNode *ln;
+    redisAssert(listLength(reply) > 0);
+    ln = listLast(reply);
+    cur = listNodeValue(ln);
+    if (cur->refcount > 1) {
+        new = dupStringObject(cur);
+        decrRefCount(cur);
+        listNodeValue(ln) = new;
+    }
+    return listNodeValue(ln);

 }
 
 }
 

-void _ensureBufferInReplyList(redisClient *c) {
-    sds buffer = sdsnewlen(NULL,REDIS_REPLY_CHUNK_SIZE);
-    sdsupdatelen(buffer); /* sdsnewlen expects non-empty string */
-    listAddNodeTail(c->reply,createObject(REDIS_REPLY_NODE,buffer));
+/* -----------------------------------------------------------------------------
+ * Low level functions to add more data to output buffers.
+ * -------------------------------------------------------------------------- */
+
+int _addReplyToBuffer(redisClient *c, char *s, size_t len) {
+    size_t available = sizeof(c->buf)-c->bufpos;
+
+    if (c->flags & REDIS_CLOSE_AFTER_REPLY) return REDIS_OK;
+
+    /* If there already are entries in the reply list, we cannot
+     * add anything more to the static buffer. */
+    if (listLength(c->reply) > 0) return REDIS_ERR;
+
+    /* Check that the buffer has enough space available for this string. */
+    if (len > available) return REDIS_ERR;
+
+    memcpy(c->buf+c->bufpos,s,len);
+    c->bufpos+=len;
+    return REDIS_OK;

 }
 
 }
 

-void _addReplyStringToBuffer(redisClient *c, char *s, size_t len) {
-    size_t available = 0;
-    redisAssert(len < REDIS_REPLY_CHUNK_THRESHOLD);
-    if (listLength(c->reply) > 0) {
-        robj *o = listNodeValue(listLast(c->reply));
+void _addReplyObjectToList(redisClient *c, robj *o) {
+    robj *tail;
+
+    if (c->flags & REDIS_CLOSE_AFTER_REPLY) return;
+
+    if (listLength(c->reply) == 0) {
+        incrRefCount(o);
+        listAddNodeTail(c->reply,o);
+    } else {
+        tail = listNodeValue(listLast(c->reply));

 
 

-        /* Make sure to append to a reply node with enough bytes available. */
-        if (o->type == REDIS_REPLY_NODE) available = sdsavail(o->ptr);
-        if (o->type != REDIS_REPLY_NODE || len > available) {
-            _ensureBufferInReplyList(c);
-            _addReplyStringToBuffer(c,s,len);
+        /* Append to this object when possible. */
+        if (tail->ptr != NULL &&
+            sdslen(tail->ptr)+sdslen(o->ptr) <= REDIS_REPLY_CHUNK_BYTES)
+        {
+            tail = dupLastObjectIfNeeded(c->reply);
+            tail->ptr = sdscatlen(tail->ptr,o->ptr,sdslen(o->ptr));

         } else {
         } else {

-            o->ptr = sdscatlen(o->ptr,s,len);
+            incrRefCount(o);
+            listAddNodeTail(c->reply,o);

         }
         }

+    }
+}
+
+/* This method takes responsibility over the sds. When it is no longer
+ * needed it will be free'd, otherwise it ends up in a robj. */
+void _addReplySdsToList(redisClient *c, sds s) {
+    robj *tail;
+
+    if (c->flags & REDIS_CLOSE_AFTER_REPLY) {
+        sdsfree(s);
+        return;
+    }
+
+    if (listLength(c->reply) == 0) {
+        listAddNodeTail(c->reply,createObject(REDIS_STRING,s));

     } else {
     } else {

-        available = c->buflen-c->bufpos;
-        if (len > available) {
-            _ensureBufferInReplyList(c);
-            _addReplyStringToBuffer(c,s,len);
+        tail = listNodeValue(listLast(c->reply));
+
+        /* Append to this object when possible. */
+        if (tail->ptr != NULL &&
+            sdslen(tail->ptr)+sdslen(s) <= REDIS_REPLY_CHUNK_BYTES)
+        {
+            tail = dupLastObjectIfNeeded(c->reply);
+            tail->ptr = sdscatlen(tail->ptr,s,sdslen(s));
+            sdsfree(s);

         } else {
         } else {

-            memcpy(c->buf+c->bufpos,s,len);
-            c->bufpos += len;
+            listAddNodeTail(c->reply,createObject(REDIS_STRING,s));

         }
     }
 }
 
         }
     }
 }
 

-void addReply(redisClient *c, robj *obj) {
-    if (_ensureFileEvent(c) != REDIS_OK) return;
-    if (server.vm_enabled && obj->storage != REDIS_VM_MEMORY) {
-        /* Returns a new object with refcount 1 */
-        obj = dupStringObject(obj);
+void _addReplyStringToList(redisClient *c, char *s, size_t len) {
+    robj *tail;
+
+    if (c->flags & REDIS_CLOSE_AFTER_REPLY) return;
+
+    if (listLength(c->reply) == 0) {
+        listAddNodeTail(c->reply,createStringObject(s,len));

     } else {
     } else {

-        /* This increments the refcount. */
-        obj = getDecodedObject(obj);
+        tail = listNodeValue(listLast(c->reply));
+
+        /* Append to this object when possible. */
+        if (tail->ptr != NULL &&
+            sdslen(tail->ptr)+len <= REDIS_REPLY_CHUNK_BYTES)
+        {
+            tail = dupLastObjectIfNeeded(c->reply);
+            tail->ptr = sdscatlen(tail->ptr,s,len);
+        } else {
+            listAddNodeTail(c->reply,createStringObject(s,len));
+        }

     }
     }

+}

 
 

-    if (sdslen(obj->ptr) < REDIS_REPLY_CHUNK_THRESHOLD) {
-        _addReplyStringToBuffer(c,obj->ptr,sdslen(obj->ptr));
-        decrRefCount(obj);
+/* -----------------------------------------------------------------------------
+ * Higher level functions to queue data on the client output buffer.
+ * The following functions are the ones that commands implementations will call.
+ * -------------------------------------------------------------------------- */
+
+void addReply(redisClient *c, robj *obj) {
+    if (_installWriteEvent(c) != REDIS_OK) return;
+
+    /* This is an important place where we can avoid copy-on-write
+     * when there is a saving child running, avoiding touching the
+     * refcount field of the object if it's not needed.
+     *
+     * If the encoding is RAW and there is room in the static buffer
+     * we'll be able to send the object to the client without
+     * messing with its page. */
+    if (obj->encoding == REDIS_ENCODING_RAW) {
+        if (_addReplyToBuffer(c,obj->ptr,sdslen(obj->ptr)) != REDIS_OK)
+            _addReplyObjectToList(c,obj);

     } else {
     } else {

-        _addReplyObjectToList(c,obj);
+        /* FIXME: convert the long into string and use _addReplyToBuffer()
+         * instead of calling getDecodedObject. As this place in the
+         * code is too performance critical. */
+        obj = getDecodedObject(obj);
+        if (_addReplyToBuffer(c,obj->ptr,sdslen(obj->ptr)) != REDIS_OK)
+            _addReplyObjectToList(c,obj);
+        decrRefCount(obj);

     }
 }
 
 void addReplySds(redisClient *c, sds s) {
     }
 }
 
 void addReplySds(redisClient *c, sds s) {

-    if (_ensureFileEvent(c) != REDIS_OK) {
+    if (_installWriteEvent(c) != REDIS_OK) {

         /* The caller expects the sds to be free'd. */
         sdsfree(s);
         return;
     }
         /* The caller expects the sds to be free'd. */
         sdsfree(s);
         return;
     }

-    if (sdslen(s) < REDIS_REPLY_CHUNK_THRESHOLD) {
-        _addReplyStringToBuffer(c,s,sdslen(s));
+    if (_addReplyToBuffer(c,s,sdslen(s)) == REDIS_OK) {

         sdsfree(s);
     } else {
         sdsfree(s);
     } else {

-        _addReplyObjectToList(c,createObject(REDIS_STRING,s));
+        /* This method free's the sds when it is no longer needed. */
+        _addReplySdsToList(c,s);

     }
 }
 
 void addReplyString(redisClient *c, char *s, size_t len) {
     }
 }
 
 void addReplyString(redisClient *c, char *s, size_t len) {

-    if (_ensureFileEvent(c) != REDIS_OK) return;
-    if (len < REDIS_REPLY_CHUNK_THRESHOLD) {
-        _addReplyStringToBuffer(c,s,len);
-    } else {
-        _addReplyObjectToList(c,createStringObject(s,len));
+    if (_installWriteEvent(c) != REDIS_OK) return;
+    if (_addReplyToBuffer(c,s,len) != REDIS_OK)
+        _addReplyStringToList(c,s,len);
+}
+
+void _addReplyError(redisClient *c, char *s, size_t len) {
+    addReplyString(c,"-ERR ",5);
+    addReplyString(c,s,len);
+    addReplyString(c,"\r\n",2);
+}
+
+void addReplyError(redisClient *c, char *err) {
+    _addReplyError(c,err,strlen(err));
+}
+
+void addReplyErrorFormat(redisClient *c, const char *fmt, ...) {
+    size_t l, j;
+    va_list ap;
+    va_start(ap,fmt);
+    sds s = sdscatvprintf(sdsempty(),fmt,ap);
+    va_end(ap);
+    /* Make sure there are no newlines in the string, otherwise invalid protocol
+     * is emitted. */
+    l = sdslen(s);
+    for (j = 0; j < l; j++) {
+        if (s[j] == '\r' || s[j] == '\n') s[j] = ' ';

     }
     }

+    _addReplyError(c,s,sdslen(s));
+    sdsfree(s);
+}
+
+void _addReplyStatus(redisClient *c, char *s, size_t len) {
+    addReplyString(c,"+",1);
+    addReplyString(c,s,len);
+    addReplyString(c,"\r\n",2);
+}
+
+void addReplyStatus(redisClient *c, char *status) {
+    _addReplyStatus(c,status,strlen(status));
+}
+
+void addReplyStatusFormat(redisClient *c, const char *fmt, ...) {
+    va_list ap;
+    va_start(ap,fmt);
+    sds s = sdscatvprintf(sdsempty(),fmt,ap);
+    va_end(ap);
+    _addReplyStatus(c,s,sdslen(s));
+    sdsfree(s);

 }
 
 /* Adds an empty object to the reply list that will contain the multi bulk
  * length, which is not known when this function is called. */
 void *addDeferredMultiBulkLength(redisClient *c) {
 }
 
 /* Adds an empty object to the reply list that will contain the multi bulk
  * length, which is not known when this function is called. */
 void *addDeferredMultiBulkLength(redisClient *c) {

-    if (_ensureFileEvent(c) != REDIS_OK) return NULL;
-    _addReplyObjectToList(c,createObject(REDIS_STRING,NULL));
+    /* Note that we install the write event here even if the object is not
+     * ready to be sent, since we are sure that before returning to the
+     * event loop setDeferredMultiBulkLength() will be called. */
+    if (_installWriteEvent(c) != REDIS_OK) return NULL;
+    listAddNodeTail(c->reply,createObject(REDIS_STRING,NULL));

     return listLast(c->reply);
 }
 
     return listLast(c->reply);
 }
 


@@ -170,14 +304,16 @@ void setDeferredMultiBulkLength(redisClient *c, void *node, long length) {
     len->ptr = sdscatprintf(sdsempty(),"*%ld\r\n",length);
     if (ln->next != NULL) {
         next = listNodeValue(ln->next);
     len->ptr = sdscatprintf(sdsempty(),"*%ld\r\n",length);
     if (ln->next != NULL) {
         next = listNodeValue(ln->next);

-        /* Only glue when the next node is a reply chunk. */
-        if (next->type == REDIS_REPLY_NODE) {
+
+        /* Only glue when the next node is non-NULL (an sds in this case) */
+        if (next->ptr != NULL) {

             len->ptr = sdscatlen(len->ptr,next->ptr,sdslen(next->ptr));
             listDelNode(c->reply,ln->next);
         }
     }
 }
 
             len->ptr = sdscatlen(len->ptr,next->ptr,sdslen(next->ptr));
             listDelNode(c->reply,ln->next);
         }
     }
 }
 

+/* Add a duble as a bulk reply */

 void addReplyDouble(redisClient *c, double d) {
     char dbuf[128], sbuf[128];
     int dlen, slen;
 void addReplyDouble(redisClient *c, double d) {
     char dbuf[128], sbuf[128];
     int dlen, slen;


@@ -186,6 +322,8 @@ void addReplyDouble(redisClient *c, double d) {
     addReplyString(c,sbuf,slen);
 }
 
     addReplyString(c,sbuf,slen);
 }
 

+/* Add a long long as integer reply or bulk len / multi bulk count.
+ * Basically this is used to output <prefix><long long><crlf>. */

 void _addReplyLongLong(redisClient *c, long long ll, char prefix) {
     char buf[128];
     int len;
 void _addReplyLongLong(redisClient *c, long long ll, char prefix) {
     char buf[128];
     int len;


@@ -197,17 +335,19 @@ void _addReplyLongLong(redisClient *c, long long ll, char prefix) {
 }
 
 void addReplyLongLong(redisClient *c, long long ll) {
 }
 
 void addReplyLongLong(redisClient *c, long long ll) {

-    _addReplyLongLong(c,ll,':');
-}
-
-void addReplyUlong(redisClient *c, unsigned long ul) {
-    _addReplyLongLong(c,(long long)ul,':');
+    if (ll == 0)
+        addReply(c,shared.czero);
+    else if (ll == 1)
+        addReply(c,shared.cone);
+    else
+        _addReplyLongLong(c,ll,':');

 }
 
 void addReplyMultiBulkLen(redisClient *c, long length) {
     _addReplyLongLong(c,length,'*');
 }
 
 }
 
 void addReplyMultiBulkLen(redisClient *c, long length) {
     _addReplyLongLong(c,length,'*');
 }
 

+/* Create the length prefix of a bulk reply, example: $2234 */

 void addReplyBulkLen(redisClient *c, robj *obj) {
     size_t len;
 
 void addReplyBulkLen(redisClient *c, robj *obj) {
     size_t len;
 


@@ -229,40 +369,43 @@ void addReplyBulkLen(redisClient *c, robj *obj) {
     _addReplyLongLong(c,len,'$');
 }
 
     _addReplyLongLong(c,len,'$');
 }
 

+/* Add a Redis Object as a bulk reply */

 void addReplyBulk(redisClient *c, robj *obj) {
     addReplyBulkLen(c,obj);
     addReply(c,obj);
     addReply(c,shared.crlf);
 }
 
 void addReplyBulk(redisClient *c, robj *obj) {
     addReplyBulkLen(c,obj);
     addReply(c,obj);
     addReply(c,shared.crlf);
 }
 

-/* In the CONFIG command we need to add vanilla C string as bulk replies */
+/* Add a C buffer as bulk reply */
+void addReplyBulkCBuffer(redisClient *c, void *p, size_t len) {
+    _addReplyLongLong(c,len,'$');
+    addReplyString(c,p,len);
+    addReply(c,shared.crlf);
+}
+
+/* Add a C nul term string as bulk reply */

 void addReplyBulkCString(redisClient *c, char *s) {
     if (s == NULL) {
         addReply(c,shared.nullbulk);
     } else {
 void addReplyBulkCString(redisClient *c, char *s) {
     if (s == NULL) {
         addReply(c,shared.nullbulk);
     } else {

-        robj *o = createStringObject(s,strlen(s));
-        addReplyBulk(c,o);
-        decrRefCount(o);
+        addReplyBulkCBuffer(c,s,strlen(s));

     }
 }
 
     }
 }
 

-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);
+/* Add a long long as a bulk reply */
+void addReplyBulkLongLong(redisClient *c, long long ll) {
+    char buf[64];
+    int len;

 
 

-    cfd = anetAccept(server.neterr, fd, cip, &cport);
-    if (cfd == AE_ERR) {
-        redisLog(REDIS_VERBOSE,"Accepting client connection: %s", server.neterr);
-        return;
-    }
-    redisLog(REDIS_VERBOSE,"Accepted %s:%d", cip, cport);
-    if ((c = createClient(cfd)) == NULL) {
+    len = ll2string(buf,64,ll);
+    addReplyBulkCBuffer(c,buf,len);
+}
+
+static void acceptCommonHandler(int fd) {
+    redisClient *c;
+    if ((c = createClient(fd)) == NULL) {

         redisLog(REDIS_WARNING,"Error allocating resoures for the client");
         redisLog(REDIS_WARNING,"Error allocating resoures for the client");

-        close(cfd); /* May be already closed, just ingore errors */
+        close(fd); /* May be already closed, just ingore errors */

         return;
     }
     /* If maxclient directive is set and this is one client more... close the
         return;
     }
     /* If maxclient directive is set and this is one client more... close the


@@ -282,15 +425,44 @@ void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
     server.stat_numconnections++;
 }
 
     server.stat_numconnections++;
 }
 

+void acceptTcpHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
+    int cport, cfd;
+    char cip[128];
+    REDIS_NOTUSED(el);
+    REDIS_NOTUSED(mask);
+    REDIS_NOTUSED(privdata);
+
+    cfd = anetTcpAccept(server.neterr, fd, cip, &cport);
+    if (cfd == AE_ERR) {
+        redisLog(REDIS_WARNING,"Accepting client connection: %s", server.neterr);
+        return;
+    }
+    redisLog(REDIS_VERBOSE,"Accepted %s:%d", cip, cport);
+    acceptCommonHandler(cfd);
+}
+
+void acceptUnixHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
+    int cfd;
+    REDIS_NOTUSED(el);
+    REDIS_NOTUSED(mask);
+    REDIS_NOTUSED(privdata);
+
+    cfd = anetUnixAccept(server.neterr, fd);
+    if (cfd == AE_ERR) {
+        redisLog(REDIS_WARNING,"Accepting client connection: %s", server.neterr);
+        return;
+    }
+    redisLog(REDIS_VERBOSE,"Accepted connection to %s", server.unixsocket);
+    acceptCommonHandler(cfd);
+}
+
+

 static void freeClientArgv(redisClient *c) {
     int j;
 static void freeClientArgv(redisClient *c) {
     int j;

-

     for (j = 0; j < c->argc; j++)
         decrRefCount(c->argv[j]);
     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->argc = 0;

-    c->mbargc = 0;
+    c->cmd = NULL;

 }
 
 void freeClient(redisClient *c) {
 }
 
 void freeClient(redisClient *c) {


@@ -324,24 +496,12 @@ void freeClient(redisClient *c) {
     ln = listSearchKey(server.clients,c);
     redisAssert(ln != NULL);
     listDelNode(server.clients,ln);
     ln = listSearchKey(server.clients,c);
     redisAssert(ln != NULL);
     listDelNode(server.clients,ln);

-    /* Remove from the list of clients waiting for swapped keys, or ready
-     * to be restarted, but not yet woken up again. */
-    if (c->flags & REDIS_IO_WAIT) {
-        redisAssert(server.vm_enabled);
-        if (listLength(c->io_keys) == 0) {
-            ln = listSearchKey(server.io_ready_clients,c);
-
-            /* When this client is waiting to be woken up (REDIS_IO_WAIT),
-             * it should be present in the list io_ready_clients */
-            redisAssert(ln != NULL);
-            listDelNode(server.io_ready_clients,ln);
-        } else {
-            while (listLength(c->io_keys)) {
-                ln = listFirst(c->io_keys);
-                dontWaitForSwappedKey(c,ln->value);
-            }
-        }
-        server.vm_blocked_clients--;
+    /* When client was just unblocked because of a blocking operation,
+     * remove it from the list with unblocked clients. */
+    if (c->flags & REDIS_UNBLOCKED) {
+        ln = listSearchKey(server.unblocked_clients,c);
+        redisAssert(ln != NULL);
+        listDelNode(server.unblocked_clients,ln);

     }
     listRelease(c->io_keys);
     /* Master/slave cleanup.
     }
     listRelease(c->io_keys);
     /* Master/slave cleanup.


@@ -359,19 +519,25 @@ void freeClient(redisClient *c) {
     if (c->flags & REDIS_MASTER) {
         server.master = NULL;
         server.replstate = REDIS_REPL_CONNECT;
     if (c->flags & REDIS_MASTER) {
         server.master = NULL;
         server.replstate = REDIS_REPL_CONNECT;

+        server.repl_down_since = time(NULL);

         /* Since we lost the connection with the master, we should also
          * close the connection with all our slaves if we have any, so
          * when we'll resync with the master the other slaves will sync again
          * with us as well. Note that also when the slave is not connected
         /* Since we lost the connection with the master, we should also
          * close the connection with all our slaves if we have any, so
          * when we'll resync with the master the other slaves will sync again
          * with us as well. Note that also when the slave is not connected

-         * to the master it will keep refusing connections by other slaves. */
-        while (listLength(server.slaves)) {
-            ln = listFirst(server.slaves);
-            freeClient((redisClient*)ln->value);
+         * to the master it will keep refusing connections by other slaves.
+         *
+         * We do this only if server.masterhost != NULL. If it is NULL this
+         * means the user called SLAVEOF NO ONE and we are freeing our
+         * link with the master, so no need to close link with slaves. */
+        if (server.masterhost != NULL) {
+            while (listLength(server.slaves)) {
+                ln = listFirst(server.slaves);
+                freeClient((redisClient*)ln->value);
+            }

         }
     }
     /* Release memory */
     zfree(c->argv);
         }
     }
     /* Release memory */
     zfree(c->argv);

-    zfree(c->mbargv);

     freeClientMultiState(c);
     zfree(c);
 }
     freeClientMultiState(c);
     zfree(c);
 }


@@ -383,15 +549,6 @@ void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) {
     REDIS_NOTUSED(el);
     REDIS_NOTUSED(mask);
 
     REDIS_NOTUSED(el);
     REDIS_NOTUSED(mask);
 

-    /* Use writev() if we have enough buffers to send */
-    if (!server.glueoutputbuf &&
-        listLength(c->reply) > REDIS_WRITEV_THRESHOLD &&
-        !(c->flags & REDIS_MASTER))
-    {
-        sendReplyToClientWritev(el, fd, privdata, mask);
-        return;
-    }
-

     while(c->bufpos > 0 || listLength(c->reply)) {
         if (c->bufpos > 0) {
             if (c->flags & REDIS_MASTER) {
     while(c->bufpos > 0 || listLength(c->reply)) {
         if (c->bufpos > 0) {
             if (c->flags & REDIS_MASTER) {


@@ -453,95 +610,21 @@ void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask) {
         }
     }
     if (totwritten > 0) c->lastinteraction = time(NULL);
         }
     }
     if (totwritten > 0) c->lastinteraction = time(NULL);

-    if (listLength(c->reply) == 0) {
+    if (c->bufpos == 0 && listLength(c->reply) == 0) {

         c->sentlen = 0;
         aeDeleteFileEvent(server.el,c->fd,AE_WRITABLE);
         c->sentlen = 0;
         aeDeleteFileEvent(server.el,c->fd,AE_WRITABLE);

-    }
-}
-
-void sendReplyToClientWritev(aeEventLoop *el, int fd, void *privdata, int mask)
-{
-    redisClient *c = privdata;
-    int nwritten = 0, totwritten = 0, objlen, willwrite;
-    robj *o;
-    struct iovec iov[REDIS_WRITEV_IOVEC_COUNT];
-    int offset, ion = 0;
-    REDIS_NOTUSED(el);
-    REDIS_NOTUSED(mask);
-
-    listNode *node;
-    while (listLength(c->reply)) {
-        offset = c->sentlen;
-        ion = 0;
-        willwrite = 0;
-
-        /* fill-in the iov[] array */
-        for(node = listFirst(c->reply); node; node = listNextNode(node)) {
-            o = listNodeValue(node);
-            objlen = sdslen(o->ptr);
-
-            if (totwritten + objlen - offset > REDIS_MAX_WRITE_PER_EVENT)
-                break;
-
-            if(ion == REDIS_WRITEV_IOVEC_COUNT)
-                break; /* no more iovecs */
-
-            iov[ion].iov_base = ((char*)o->ptr) + offset;
-            iov[ion].iov_len = objlen - offset;
-            willwrite += objlen - offset;
-            offset = 0; /* just for the first item */
-            ion++;
-        }
-
-        if(willwrite == 0)
-            break;
-
-        /* write all collected blocks at once */
-        if((nwritten = writev(fd, iov, ion)) < 0) {
-            if (errno != EAGAIN) {
-                redisLog(REDIS_VERBOSE,
-                         "Error writing to client: %s", strerror(errno));
-                freeClient(c);
-                return;
-            }
-            break;
-        }
-
-        totwritten += nwritten;
-        offset = c->sentlen;

 
 

-        /* remove written robjs from c->reply */
-        while (nwritten && listLength(c->reply)) {
-            o = listNodeValue(listFirst(c->reply));
-            objlen = sdslen(o->ptr);
-
-            if(nwritten >= objlen - offset) {
-                listDelNode(c->reply, listFirst(c->reply));
-                nwritten -= objlen - offset;
-                c->sentlen = 0;
-            } else {
-                /* partial write */
-                c->sentlen += nwritten;
-                break;
-            }
-            offset = 0;
-        }
-    }
-
-    if (totwritten > 0)
-        c->lastinteraction = time(NULL);
-
-    if (listLength(c->reply) == 0) {
-        c->sentlen = 0;
-        aeDeleteFileEvent(server.el,c->fd,AE_WRITABLE);
+        /* Close connection after entire reply has been sent. */
+        if (c->flags & REDIS_CLOSE_AFTER_REPLY) freeClient(c);

     }
 }
 
 /* resetClient prepare the client to process the next command */
 void resetClient(redisClient *c) {
     freeClientArgv(c);
     }
 }
 
 /* resetClient prepare the client to process the next command */
 void resetClient(redisClient *c) {
     freeClientArgv(c);

+    c->reqtype = 0;
+    c->multibulklen = 0;

     c->bulklen = -1;
     c->bulklen = -1;

-    c->multibulk = 0;

 }
 
 void closeTimedoutClients(void) {
 }
 
 void closeTimedoutClients(void) {


@@ -564,7 +647,7 @@ void closeTimedoutClients(void) {
             redisLog(REDIS_VERBOSE,"Closing idle client");
             freeClient(c);
         } else if (c->flags & REDIS_BLOCKED) {
             redisLog(REDIS_VERBOSE,"Closing idle client");
             freeClient(c);
         } else if (c->flags & REDIS_BLOCKED) {

-            if (c->blockingto != 0 && c->blockingto < now) {
+            if (c->bpop.timeout != 0 && c->bpop.timeout < now) {

                 addReply(c,shared.nullmultibulk);
                 unblockClientWaitingData(c);
             }
                 addReply(c,shared.nullmultibulk);
                 unblockClientWaitingData(c);
             }


@@ -572,84 +655,176 @@ void closeTimedoutClients(void) {
     }
 }
 
     }
 }
 

-void processInputBuffer(redisClient *c) {
-again:
-    /* Before to process the input buffer, make sure the client is not
-     * waitig for a blocking operation such as BLPOP. Note that the first
-     * iteration the client is never blocked, otherwise the processInputBuffer
-     * would not be called at all, but after the execution of the first commands
-     * in the input buffer the client may be blocked, and the "goto again"
-     * will try to reiterate. The following line will make it return asap. */
-    if (c->flags & REDIS_BLOCKED || c->flags & REDIS_IO_WAIT) return;
-    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;
-
-            query = c->querybuf;
-            c->querybuf = sdsempty();
-            querylen = 1+(p-(query));
-            if (sdslen(query) > querylen) {
-                /* leave data after the first line of the query in the buffer */
-                c->querybuf = sdscatlen(c->querybuf,query+querylen,sdslen(query)-querylen);
-            }
-            *p = '\0'; /* remove "\n" */
-            if (*(p-1) == '\r') *(p-1) = '\0'; /* and "\r" if any */
-            sdsupdatelen(query);
+int processInlineBuffer(redisClient *c) {
+    char *newline = strstr(c->querybuf,"\r\n");
+    int argc, j;
+    sds *argv;
+    size_t querylen;

 
 

-            /* Now we can split the query in arguments */
-            argv = sdssplitlen(query,sdslen(query)," ",1,&argc);
-            sdsfree(query);
+    /* Nothing to do without a \r\n */
+    if (newline == NULL)
+        return REDIS_ERR;

 
 

-            if (c->argv) zfree(c->argv);
-            c->argv = zmalloc(sizeof(robj*)*argc);
+    /* Split the input buffer up to the \r\n */
+    querylen = newline-(c->querybuf);
+    argv = sdssplitlen(c->querybuf,querylen," ",1,&argc);

 
 

-            for (j = 0; j < argc; j++) {
-                if (sdslen(argv[j])) {
-                    c->argv[c->argc] = createObject(REDIS_STRING,argv[j]);
-                    c->argc++;
-                } else {
-                    sdsfree(argv[j]);
-                }
+    /* Leave data after the first line of the query in the buffer */
+    c->querybuf = sdsrange(c->querybuf,querylen+2,-1);
+
+    /* Setup argv array on client structure */
+    if (c->argv) zfree(c->argv);
+    c->argv = zmalloc(sizeof(robj*)*argc);
+
+    /* Create redis objects for all arguments. */
+    for (c->argc = 0, j = 0; j < argc; j++) {
+        if (sdslen(argv[j])) {
+            c->argv[c->argc] = createObject(REDIS_STRING,argv[j]);
+            c->argc++;
+        } else {
+            sdsfree(argv[j]);
+        }
+    }
+    zfree(argv);
+    return REDIS_OK;
+}
+
+/* Helper function. Trims query buffer to make the function that processes
+ * multi bulk requests idempotent. */
+static void setProtocolError(redisClient *c, int pos) {
+    c->flags |= REDIS_CLOSE_AFTER_REPLY;
+    c->querybuf = sdsrange(c->querybuf,pos,-1);
+}
+
+int processMultibulkBuffer(redisClient *c) {
+    char *newline = NULL;
+    int pos = 0, ok;
+    long long ll;
+
+    if (c->multibulklen == 0) {
+        /* The client should have been reset */
+        redisAssert(c->argc == 0);
+
+        /* Multi bulk length cannot be read without a \r\n */
+        newline = strchr(c->querybuf,'\r');
+        if (newline == NULL)
+            return REDIS_ERR;
+
+        /* Buffer should also contain \n */
+        if (newline-(c->querybuf) > ((signed)sdslen(c->querybuf)-2))
+            return REDIS_ERR;
+
+        /* We know for sure there is a whole line since newline != NULL,
+         * so go ahead and find out the multi bulk length. */
+        redisAssert(c->querybuf[0] == '*');
+        ok = string2ll(c->querybuf+1,newline-(c->querybuf+1),&ll);
+        if (!ok || ll > 1024*1024) {
+            addReplyError(c,"Protocol error: invalid multibulk length");
+            setProtocolError(c,pos);
+            return REDIS_ERR;
+        }
+
+        pos = (newline-c->querybuf)+2;
+        if (ll <= 0) {
+            c->querybuf = sdsrange(c->querybuf,pos,-1);
+            return REDIS_OK;
+        }
+
+        c->multibulklen = ll;
+
+        /* Setup argv array on client structure */
+        if (c->argv) zfree(c->argv);
+        c->argv = zmalloc(sizeof(robj*)*c->multibulklen);
+    }
+
+    redisAssert(c->multibulklen > 0);
+    while(c->multibulklen) {
+        /* Read bulk length if unknown */
+        if (c->bulklen == -1) {
+            newline = strchr(c->querybuf+pos,'\r');
+            if (newline == NULL)
+                break;
+
+            /* Buffer should also contain \n */
+            if (newline-(c->querybuf) > ((signed)sdslen(c->querybuf)-2))
+                break;
+
+            if (c->querybuf[pos] != '$') {
+                addReplyErrorFormat(c,
+                    "Protocol error: expected '$', got '%c'",
+                    c->querybuf[pos]);
+                setProtocolError(c,pos);
+                return REDIS_ERR;
+            }
+
+            ok = string2ll(c->querybuf+pos+1,newline-(c->querybuf+pos+1),&ll);
+            if (!ok || ll < 0 || ll > 512*1024*1024) {
+                addReplyError(c,"Protocol error: invalid bulk length");
+                setProtocolError(c,pos);
+                return REDIS_ERR;

             }
             }

-            zfree(argv);
-            if (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;
+
+            pos += newline-(c->querybuf+pos)+2;
+            c->bulklen = ll;
+        }
+
+        /* Read bulk argument */
+        if (sdslen(c->querybuf)-pos < (unsigned)(c->bulklen+2)) {
+            /* Not enough data (+2 == trailing \r\n) */
+            break;
+        } else {
+            c->argv[c->argc++] = createStringObject(c->querybuf+pos,c->bulklen);
+            pos += c->bulklen+2;
+            c->bulklen = -1;
+            c->multibulklen--;
+        }
+    }
+
+    /* Trim to pos */
+    c->querybuf = sdsrange(c->querybuf,pos,-1);
+
+    /* We're done when c->multibulk == 0 */
+    if (c->multibulklen == 0) {
+        return REDIS_OK;
+    }
+    return REDIS_ERR;
+}
+
+void processInputBuffer(redisClient *c) {
+    /* Keep processing while there is something in the input buffer */
+    while(sdslen(c->querybuf)) {
+        /* Immediately abort if the client is in the middle of something. */
+        if (c->flags & REDIS_BLOCKED) return;
+
+        /* REDIS_CLOSE_AFTER_REPLY closes the connection once the reply is
+         * written to the client. Make sure to not let the reply grow after
+         * this flag has been set (i.e. don't process more commands). */
+        if (c->flags & REDIS_CLOSE_AFTER_REPLY) return;
+
+        /* Determine request type when unknown. */
+        if (!c->reqtype) {
+            if (c->querybuf[0] == '*') {
+                c->reqtype = REDIS_REQ_MULTIBULK;

             } else {
             } else {

-                /* Nothing to process, argc == 0. Just process the query
-                 * buffer if it's not empty or return to the caller */
-                if (sdslen(c->querybuf)) goto again;
+                c->reqtype = REDIS_REQ_INLINE;

             }
             }

-            return;
-        } else if (sdslen(c->querybuf) >= REDIS_REQUEST_MAX_SIZE) {
-            redisLog(REDIS_VERBOSE, "Client protocol error");
-            freeClient(c);
-            return;

         }
         }

-    } else {
-        /* Bulk read handling. Note that if we are at this point
-           the client already sent a command terminated with a newline,
-           we are reading the bulk data that is actually the last
-           argument of the command. */
-        int qbl = sdslen(c->querybuf);
-
-        if (c->bulklen <= qbl) {
-            /* Copy everything but the final CRLF as final argument */
-            c->argv[c->argc] = createStringObject(c->querybuf,c->bulklen-2);
-            c->argc++;
-            c->querybuf = sdsrange(c->querybuf,c->bulklen,-1);
-            /* 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;
+
+        if (c->reqtype == REDIS_REQ_INLINE) {
+            if (processInlineBuffer(c) != REDIS_OK) break;
+        } else if (c->reqtype == REDIS_REQ_MULTIBULK) {
+            if (processMultibulkBuffer(c) != REDIS_OK) break;
+        } else {
+            redisPanic("Unknown request type");
+        }
+
+        /* Multibulk processing could see a <= 0 length. */
+        if (c->argc == 0) {
+            resetClient(c);
+        } else {
+            /* Only reset the client when the command was executed. */
+            if (processCommand(c) == REDIS_OK)
+                resetClient(c);

         }
     }
 }
         }
     }
 }


@@ -676,10 +851,143 @@ void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask) {
         return;
     }
     if (nread) {
         return;
     }
     if (nread) {

-        c->querybuf = sdscatlen(c->querybuf, buf, nread);
+        c->querybuf = sdscatlen(c->querybuf,buf,nread);

         c->lastinteraction = time(NULL);
     } else {
         return;
     }
     processInputBuffer(c);
 }
         c->lastinteraction = time(NULL);
     } else {
         return;
     }
     processInputBuffer(c);
 }

+
+void getClientsMaxBuffers(unsigned long *longest_output_list,
+                          unsigned long *biggest_input_buffer) {
+    redisClient *c;
+    listNode *ln;
+    listIter li;
+    unsigned long lol = 0, bib = 0;
+
+    listRewind(server.clients,&li);
+    while ((ln = listNext(&li)) != NULL) {
+        c = listNodeValue(ln);
+
+        if (listLength(c->reply) > lol) lol = listLength(c->reply);
+        if (sdslen(c->querybuf) > bib) bib = sdslen(c->querybuf);
+    }
+    *longest_output_list = lol;
+    *biggest_input_buffer = bib;
+}
+
+void clientCommand(redisClient *c) {
+    listNode *ln;
+    listIter li;
+    redisClient *client;
+
+    if (!strcasecmp(c->argv[1]->ptr,"list") && c->argc == 2) {
+        sds o = sdsempty();
+        time_t now = time(NULL);
+
+        listRewind(server.clients,&li);
+        while ((ln = listNext(&li)) != NULL) {
+            char ip[32], flags[16], *p;
+            int port;
+
+            client = listNodeValue(ln);
+            if (anetPeerToString(client->fd,ip,&port) == -1) continue;
+            p = flags;
+            if (client->flags & REDIS_SLAVE) {
+                if (client->flags & REDIS_MONITOR)
+                    *p++ = 'O';
+                else
+                    *p++ = 'S';
+            }
+            if (client->flags & REDIS_MASTER) *p++ = 'M';
+            if (p == flags) *p++ = 'N';
+            if (client->flags & REDIS_MULTI) *p++ = 'x';
+            if (client->flags & REDIS_BLOCKED) *p++ = 'b';
+            if (client->flags & REDIS_DIRTY_CAS) *p++ = 'd';
+            if (client->flags & REDIS_CLOSE_AFTER_REPLY) *p++ = 'c';
+            if (client->flags & REDIS_UNBLOCKED) *p++ = 'u';
+            *p++ = '\0';
+            o = sdscatprintf(o,
+                "addr=%s:%d fd=%d idle=%ld flags=%s db=%d sub=%d psub=%d\n",
+                ip,port,client->fd,
+                (long)(now - client->lastinteraction),
+                flags,
+                client->db->id,
+                (int) dictSize(client->pubsub_channels),
+                (int) listLength(client->pubsub_patterns));
+        }
+        addReplyBulkCBuffer(c,o,sdslen(o));
+        sdsfree(o);
+    } else if (!strcasecmp(c->argv[1]->ptr,"kill") && c->argc == 3) {
+        listRewind(server.clients,&li);
+        while ((ln = listNext(&li)) != NULL) {
+            char ip[32], addr[64];
+            int port;
+
+            client = listNodeValue(ln);
+            if (anetPeerToString(client->fd,ip,&port) == -1) continue;
+            snprintf(addr,sizeof(addr),"%s:%d",ip,port);
+            if (strcmp(addr,c->argv[2]->ptr) == 0) {
+                addReply(c,shared.ok);
+                if (c == client) {
+                    client->flags |= REDIS_CLOSE_AFTER_REPLY;
+                } else {
+                    freeClient(client);
+                }
+                return;
+            }
+        }
+        addReplyError(c,"No such client");
+    } else {
+        addReplyError(c, "Syntax error, try CLIENT (LIST | KILL ip:port)");
+    }
+}
+
+/* Rewrite the command vector of the client. All the new objects ref count
+ * is incremented. The old command vector is freed, and the old objects
+ * ref count is decremented. */
+void rewriteClientCommandVector(redisClient *c, int argc, ...) {
+    va_list ap;
+    int j;
+    robj **argv; /* The new argument vector */
+
+    argv = zmalloc(sizeof(robj*)*argc);
+    va_start(ap,argc);
+    for (j = 0; j < argc; j++) {
+        robj *a;
+        
+        a = va_arg(ap, robj*);
+        argv[j] = a;
+        incrRefCount(a);
+    }
+    /* We free the objects in the original vector at the end, so we are
+     * sure that if the same objects are reused in the new vector the
+     * refcount gets incremented before it gets decremented. */
+    for (j = 0; j < c->argc; j++) decrRefCount(c->argv[j]);
+    zfree(c->argv);
+    /* Replace argv and argc with our new versions. */
+    c->argv = argv;
+    c->argc = argc;
+    c->cmd = lookupCommand(c->argv[0]->ptr);
+    redisAssert(c->cmd != NULL);
+    va_end(ap);
+}
+
+/* Rewrite a single item in the command vector.
+ * The new val ref count is incremented, and the old decremented. */
+void rewriteClientCommandArgument(redisClient *c, int i, robj *newval) {
+    robj *oldval;
+   
+    redisAssert(i < c->argc);
+    oldval = c->argv[i];
+    c->argv[i] = newval;
+    incrRefCount(newval);
+    decrRefCount(oldval);
+
+    /* If this is the command name make sure to fix c->cmd. */
+    if (i == 0) {
+        c->cmd = lookupCommand(c->argv[0]->ptr);
+        redisAssert(c->cmd != NULL);
+    }
+}