WAIT AOF: deny if AOF is off. Trigger sync in proper place.

AE_BARRIER was implemented like:

    - Fire the readable event.
    - Do not fire the writabel event if the readable fired.

However this may lead to the writable event to never be called if the
readable event is always fired. There is an alterantive, we can just
invert the sequence of the calls in case AE_BARRIER is set. This commit
does that.

README.md

@@ -1,4 +1,4 @@
-This README is just a fast *quick start* document. You can find more detailed documentation at http://redis.io.
+This README is just a fast *quick start* document. You can find more detailed documentation at [redis.io](https://redis.io).
 
 What is Redis?
 --------------

redis.conf

@@ -59,7 +59,7 @@
 # internet, binding to all the interfaces is dangerous and will expose the
 # instance to everybody on the internet. So by default we uncomment the
 # following bind directive, that will force Redis to listen only into
-# the IPv4 lookback interface address (this means Redis will be able to
+# the IPv4 loopback interface address (this means Redis will be able to
 # accept connections only from clients running into the same computer it
 # is running).
 #
@@ -296,7 +296,9 @@ dir ./
 #
 # 2) if slave-serve-stale-data is set to 'no' the slave will reply with
 #    an error "SYNC with master in progress" to all the kind of commands
-#    but to INFO and SLAVEOF.
+#    but to INFO, SLAVEOF, AUTH, PING, SHUTDOWN, REPLCONF, ROLE, CONFIG,
+#    SUBSCRIBE, UNSUBSCRIBE, PSUBSCRIBE, PUNSUBSCRIBE, PUBLISH, PUBSUB,
+#    COMMAND, POST, HOST: and LATENCY.
 #
 slave-serve-stale-data yes
 
@@ -606,7 +608,7 @@ slave-priority 100
 # deletion of the object. It means that the server stops processing new commands
 # in order to reclaim all the memory associated with an object in a synchronous
 # way. If the key deleted is associated with a small object, the time needed
-# in order to execute th DEL command is very small and comparable to most other
+# in order to execute the DEL command is very small and comparable to most other
 # O(1) or O(log_N) commands in Redis. However if the key is associated with an
 # aggregated value containing millions of elements, the server can block for
 # a long time (even seconds) in order to complete the operation.
@@ -621,7 +623,7 @@ slave-priority 100
 # It's up to the design of the application to understand when it is a good
 # idea to use one or the other. However the Redis server sometimes has to
 # delete keys or flush the whole database as a side effect of other operations.
-# Specifically Redis deletes objects independently of an user call in the
+# Specifically Redis deletes objects independently of a user call in the
 # following scenarios:
 #
 # 1) On eviction, because of the maxmemory and maxmemory policy configurations,
@@ -914,7 +916,7 @@ lua-time-limit 5000
 # Docker and other containers).
 #
 # In order to make Redis Cluster working in such environments, a static
-# configuration where each node known its public address is needed. The
+# configuration where each node knows its public address is needed. The
 # following two options are used for this scope, and are:
 #
 # * cluster-announce-ip
@@ -1154,6 +1156,20 @@ client-output-buffer-limit normal 0 0 0
 client-output-buffer-limit slave 256mb 64mb 60
 client-output-buffer-limit pubsub 32mb 8mb 60
 
+# Client query buffers accumulate new commands. They are limited to a fixed
+# amount by default in order to avoid that a protocol desynchronization (for
+# instance due to a bug in the client) will lead to unbound memory usage in
+# the query buffer. However you can configure it here if you have very special
+# needs, such us huge multi/exec requests or alike.
+#
+# client-query-buffer-limit 1gb
+
+# In the Redis protocol, bulk requests, that are, elements representing single
+# strings, are normally limited ot 512 mb. However you can change this limit
+# here.
+#
+# proto-max-bulk-len 512mb
+
 # Redis calls an internal function to perform many background tasks, like
 # closing connections of clients in timeout, purging expired keys that are
 # never requested, and so forth.

src/Makefile

@@ -144,7 +144,7 @@ endif
 
 REDIS_SERVER_NAME=redis-server
 REDIS_SENTINEL_NAME=redis-sentinel
-REDIS_SERVER_OBJ=adlist.o quicklist.o ae.o anet.o dict.o server.o sds.o zmalloc.o lzf_c.o lzf_d.o pqsort.o zipmap.o sha1.o ziplist.o release.o networking.o util.o object.o db.o replication.o rdb.o t_string.o t_list.o t_set.o t_zset.o t_hash.o config.o aof.o pubsub.o multi.o debug.o sort.o intset.o syncio.o cluster.o crc16.o endianconv.o slowlog.o scripting.o bio.o rio.o rand.o memtest.o crc64.o bitops.o sentinel.o notify.o setproctitle.o blocked.o hyperloglog.o latency.o sparkline.o redis-check-rdb.o redis-check-aof.o geo.o lazyfree.o module.o evict.o expire.o geohash.o geohash_helper.o childinfo.o defrag.o siphash.o rax.o
+REDIS_SERVER_OBJ=adlist.o quicklist.o ae.o anet.o dict.o server.o sds.o zmalloc.o lzf_c.o lzf_d.o pqsort.o zipmap.o sha1.o ziplist.o release.o networking.o util.o object.o db.o replication.o rdb.o t_string.o t_list.o t_set.o t_zset.o t_hash.o config.o aof.o pubsub.o multi.o debug.o sort.o intset.o syncio.o cluster.o crc16.o endianconv.o slowlog.o scripting.o bio.o rio.o rand.o memtest.o crc64.o bitops.o sentinel.o notify.o setproctitle.o blocked.o hyperloglog.o latency.o sparkline.o redis-check-rdb.o redis-check-aof.o geo.o lazyfree.o module.o evict.o expire.o geohash.o geohash_helper.o childinfo.o defrag.o siphash.o rax.o t_stream.o listpack.c
 REDIS_CLI_NAME=redis-cli
 REDIS_CLI_OBJ=anet.o adlist.o redis-cli.o zmalloc.o release.o anet.o ae.o crc64.o
 REDIS_BENCHMARK_NAME=redis-benchmark

src/adlist.c

@@ -353,7 +353,7 @@ void listJoin(list *l, list *o) {
     else
         l->head = o->head;
 
-    l->tail = o->tail;
+    if (o->tail) l->tail = o->tail;
     l->len += o->len;
 
     /* Setup other as an empty list. */

src/ae.c

@@ -159,6 +159,10 @@ void aeDeleteFileEvent(aeEventLoop *eventLoop, int fd, int mask)
     aeFileEvent *fe = &eventLoop->events[fd];
     if (fe->mask == AE_NONE) return;
 
+    /* We want to always remove AE_BARRIER if set when AE_WRITABLE
+     * is removed. */
+    if (mask & AE_WRITABLE) mask |= AE_BARRIER;
+
     aeApiDelEvent(eventLoop, fd, mask);
     fe->mask = fe->mask & (~mask);
     if (fd == eventLoop->maxfd && fe->mask == AE_NONE) {
@@ -411,19 +415,49 @@ int aeProcessEvents(aeEventLoop *eventLoop, int flags)
             aeFileEvent *fe = &eventLoop->events[eventLoop->fired[j].fd];
             int mask = eventLoop->fired[j].mask;
             int fd = eventLoop->fired[j].fd;
-            int rfired = 0;
+            int fired = 0; /* Number of events fired for current fd. */
 
-	    /* note the fe->mask & mask & ... code: maybe an already processed
-             * event removed an element that fired and we still didn't
-             * processed, so we check if the event is still valid. */
-            if (fe->mask & mask & AE_READABLE) {
-                rfired = 1;
+            /* Normally we execute the readable event first, and the writable
+             * event laster. This is useful as sometimes we may be able
+             * to serve the reply of a query immediately after processing the
+             * query.
+             *
+             * However if AE_BARRIER is set in the mask, our application is
+             * asking us to do the reverse: never fire the writable event
+             * after the readable. In such a case, we invert the calls.
+             * This is useful when, for instance, we want to do things
+             * in the beforeSleep() hook, like fsynching a file to disk,
+             * before replying to a client. */
+            int invert = fe->mask & AE_BARRIER;
+
+	    /* Note the "fe->mask & mask & ..." code: maybe an already
+             * processed event removed an element that fired and we still
+             * didn't processed, so we check if the event is still valid.
+             *
+             * Fire the readable event if the call sequence is not
+             * inverted. */
+            if (!invert && fe->mask & mask & AE_READABLE) {
                 fe->rfileProc(eventLoop,fd,fe->clientData,mask);
+                fired++;
             }
+
+            /* Fire the writable event. */
             if (fe->mask & mask & AE_WRITABLE) {
-                if (!rfired || fe->wfileProc != fe->rfileProc)
+                if (!fired || fe->wfileProc != fe->rfileProc) {
                     fe->wfileProc(eventLoop,fd,fe->clientData,mask);
+                    fired++;
+                }
             }
+
+            /* If we have to invert the call, fire the readable event now
+             * after the writable one. */
+            if (invert && fe->mask & mask & AE_READABLE) {
+                if (!fired || fe->wfileProc != fe->rfileProc) {
+                    fe->rfileProc(eventLoop,fd,fe->clientData,mask);
+                    fired++;
+                }
+            }
+
             processed++;
         }
     }

src/ae.h

@@ -38,9 +38,14 @@
 #define AE_OK 0
 #define AE_ERR -1
 
-#define AE_NONE 0
-#define AE_READABLE 1
-#define AE_WRITABLE 2
+#define AE_NONE 0       /* No events registered. */
+#define AE_READABLE 1   /* Fire when descriptor is readable. */
+#define AE_WRITABLE 2   /* Fire when descriptor is writable. */
+#define AE_BARRIER 4    /* With WRITABLE, never fire the event if the
+                           READABLE event already fired in the same event
+                           loop iteration. Useful when you want to persist
+                           things to disk before sending replies, and want
+                           to do that in a group fashion. */
 
 #define AE_FILE_EVENTS 1
 #define AE_TIME_EVENTS 2
@@ -64,7 +69,7 @@ typedef void aeBeforeSleepProc(struct aeEventLoop *eventLoop);
 
 /* File event structure */
 typedef struct aeFileEvent {
-    int mask; /* one of AE_(READABLE|WRITABLE) */
+    int mask; /* one of AE_(READABLE|WRITABLE|BARRIER) */
     aeFileProc *rfileProc;
     aeFileProc *wfileProc;
     void *clientData;

src/anet.c

@@ -237,7 +237,7 @@ int anetResolveIP(char *err, char *host, char *ipbuf, size_t ipbuf_len) {
 
 static int anetSetReuseAddr(char *err, int fd) {
     int yes = 1;
-    /* Make sure connection-intensive things like the redis benckmark
+    /* Make sure connection-intensive things like the redis benchmark
      * will be able to close/open sockets a zillion of times */
     if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof(yes)) == -1) {
         anetSetError(err, "setsockopt SO_REUSEADDR: %s", strerror(errno));

src/aof.c

@@ -197,10 +197,60 @@ ssize_t aofRewriteBufferWrite(int fd) {
  * AOF file implementation
  * ------------------------------------------------------------------------- */
 
+/* Return true if an AOf fsync is currently already in progress in a
+ * BIO thread. */
+int aofFsyncInProgress(void) {
+    return bioPendingJobsOfType(BIO_AOF_FSYNC) != 0;
+}
+
 /* Starts a background task that performs fsync() against the specified
  * file descriptor (the one of the AOF file) in another thread. */
-void aof_background_fsync(int fd) {
-    bioCreateBackgroundJob(BIO_AOF_FSYNC,(void*)(long)fd,NULL,NULL);
+void aofStartBackgroundFsync(void) {
+    if (aofFsyncInProgress()) {
+        /* We never want to start another fsync if one is in progress. */
+        return;
+    }
+
+    /* Before starting a new fsync, we need to flush the AOF buffer to
+     * disk if there are clients blocked in WAIT AOF, otherwise we are
+     * not going to sync their data. */
+    if (server.blocked_clients_by_type[BLOCKED_AOF])
+        flushAppendOnlyFile(0);
+
+    /* No fsync is in progress. If there was one, the new epoch is stored
+     * in server.aof_fsync_in_progress_epoch. So update the current fsync
+     * epoch with the one of the fsync in progress. */
+    server.aof_fsync_epoch = server.aof_fsync_in_progress_epoch;
+    bioCreateBackgroundJob(BIO_AOF_FSYNC,(void*)(long)server.aof_fd,NULL,NULL);
+    server.aof_fsync_in_progress_epoch++;
+}
+
+/* Returns an AOF epoch so that, when such epoch is reached by
+ * server.aof_fsync_epoch, it means that everything that was written in the
+ * AOF up to the moment this function returned the epoch, is now flushed
+ * on disk. */
+uint64_t aofNextEpoch(void) {
+    return server.aof_fsync_in_progress_epoch + 1;
+}
+
+/* Kills an AOFRW child process if exists */
+static void killAppendOnlyChild(void) {
+    int statloc;
+    /* No AOFRW child? return. */
+    if (server.aof_child_pid == -1) return;
+    /* Kill AOFRW child, wait for child exit. */
+    serverLog(LL_NOTICE,"Killing running AOF rewrite child: %ld",
+        (long) server.aof_child_pid);
+    if (kill(server.aof_child_pid,SIGUSR1) != -1) {
+        while(wait3(&statloc,0,NULL) != server.aof_child_pid);
+    }
+    /* Reset the buffer accumulating changes while the child saves. */
+    aofRewriteBufferReset();
+    aofRemoveTempFile(server.aof_child_pid);
+    server.aof_child_pid = -1;
+    server.aof_rewrite_time_start = -1;
+    /* Close pipes used for IPC between the two processes. */
+    aofClosePipes();
 }
 
 /* Called when the user switches from "appendonly yes" to "appendonly no"
@@ -214,34 +264,18 @@ void stopAppendOnly(void) {
     server.aof_fd = -1;
     server.aof_selected_db = -1;
     server.aof_state = AOF_OFF;
-    /* rewrite operation in progress? kill it, wait child exit */
-    if (server.aof_child_pid != -1) {
-        int statloc;
-
-        serverLog(LL_NOTICE,"Killing running AOF rewrite child: %ld",
-            (long) server.aof_child_pid);
-        if (kill(server.aof_child_pid,SIGUSR1) != -1) {
-            while(wait3(&statloc,0,NULL) != server.aof_child_pid);
-        }
-        /* reset the buffer accumulating changes while the child saves */
-        aofRewriteBufferReset();
-        aofRemoveTempFile(server.aof_child_pid);
-        server.aof_child_pid = -1;
-        server.aof_rewrite_time_start = -1;
-        /* close pipes used for IPC between the two processes. */
-        aofClosePipes();
-    }
+    killAppendOnlyChild();
 }
 
 /* Called when the user switches from "appendonly no" to "appendonly yes"
  * at runtime using the CONFIG command. */
 int startAppendOnly(void) {
     char cwd[MAXPATHLEN]; /* Current working dir path for error messages. */
+    int newfd;
 
-    server.aof_last_fsync = server.unixtime;
-    server.aof_fd = open(server.aof_filename,O_WRONLY|O_APPEND|O_CREAT,0644);
+    newfd = open(server.aof_filename,O_WRONLY|O_APPEND|O_CREAT,0644);
     serverAssert(server.aof_state == AOF_OFF);
-    if (server.aof_fd == -1) {
+    if (newfd == -1) {
         char *cwdp = getcwd(cwd,MAXPATHLEN);
 
         serverLog(LL_WARNING,
@@ -255,17 +289,56 @@ int startAppendOnly(void) {
     if (server.rdb_child_pid != -1) {
         server.aof_rewrite_scheduled = 1;
         serverLog(LL_WARNING,"AOF was enabled but there is already a child process saving an RDB file on disk. An AOF background was scheduled to start when possible.");
-    } else if (rewriteAppendOnlyFileBackground() == C_ERR) {
-        close(server.aof_fd);
-        serverLog(LL_WARNING,"Redis needs to enable the AOF but can't trigger a background AOF rewrite operation. Check the above logs for more info about the error.");
-        return C_ERR;
+    } else {
+        /* If there is a pending AOF rewrite, we need to switch it off and
+         * start a new one: the old one cannot be reused becuase it is not
+         * accumulating the AOF buffer. */
+        if (server.aof_child_pid != -1) {
+            serverLog(LL_WARNING,"AOF was enabled but there is already an AOF rewriting in background. Stopping background AOF and starting a rewrite now.");
+            killAppendOnlyChild();
+        }
+        if (rewriteAppendOnlyFileBackground() == C_ERR) {
+            close(newfd);
+            serverLog(LL_WARNING,"Redis needs to enable the AOF but can't trigger a background AOF rewrite operation. Check the above logs for more info about the error.");
+            return C_ERR;
+        }
     }
     /* We correctly switched on AOF, now wait for the rewrite to be complete
      * in order to append data on disk. */
     server.aof_state = AOF_WAIT_REWRITE;
+    server.aof_last_fsync = server.unixtime;
+    server.aof_fd = newfd;
     return C_OK;
 }
 
+/* This is a wrapper to the write syscall in order to retry on short writes
+ * or if the syscall gets interrupted. It could look strange that we retry
+ * on short writes given that we are writing to a block device: normally if
+ * the first call is short, there is a end-of-space condition, so the next
+ * is likely to fail. However apparently in modern systems this is no longer
+ * true, and in general it looks just more resilient to retry the write. If
+ * there is an actual error condition we'll get it at the next try. */
+ssize_t aofWrite(int fd, const char *buf, size_t len) {
+    ssize_t nwritten = 0, totwritten = 0;
+
+    while(len) {
+        nwritten = write(fd, buf, len);
+
+        if (nwritten < 0) {
+            if (errno == EINTR) {
+                continue;
+            }
+            return totwritten ? totwritten : -1;
+        }
+
+        len -= nwritten;
+        buf += nwritten;
+        totwritten += nwritten;
+    }
+
+    return totwritten;
+}
+
 /* Write the append only file buffer on disk.
  *
  * Since we are required to write the AOF before replying to the client,
@@ -293,7 +366,7 @@ void flushAppendOnlyFile(int force) {
     if (sdslen(server.aof_buf) == 0) return;
 
     if (server.aof_fsync == AOF_FSYNC_EVERYSEC)
-        sync_in_progress = bioPendingJobsOfType(BIO_AOF_FSYNC) != 0;
+        sync_in_progress = aofFsyncInProgress();
 
     if (server.aof_fsync == AOF_FSYNC_EVERYSEC && !force) {
         /* With this append fsync policy we do background fsyncing.
@@ -323,7 +396,7 @@ void flushAppendOnlyFile(int force) {
      * or alike */
 
     latencyStartMonitor(latency);
-    nwritten = write(server.aof_fd,server.aof_buf,sdslen(server.aof_buf));
+    nwritten = aofWrite(server.aof_fd,server.aof_buf,sdslen(server.aof_buf));
     latencyEndMonitor(latency);
     /* We want to capture different events for delayed writes:
      * when the delay happens with a pending fsync, or with a saving child
@@ -342,7 +415,7 @@ void flushAppendOnlyFile(int force) {
     /* We performed the write so reset the postponed flush sentinel to zero. */
     server.aof_flush_postponed_start = 0;
 
-    if (nwritten != (signed)sdslen(server.aof_buf)) {
+    if (nwritten != (ssize_t)sdslen(server.aof_buf)) {
         static time_t last_write_error_log = 0;
         int can_log = 0;
 
@@ -442,7 +515,11 @@ void flushAppendOnlyFile(int force) {
         server.aof_last_fsync = server.unixtime;
     } else if ((server.aof_fsync == AOF_FSYNC_EVERYSEC &&
                 server.unixtime > server.aof_last_fsync)) {
-        if (!sync_in_progress) aof_background_fsync(server.aof_fd);
+        aofStartBackgroundFsync();
+        /* Note that we update this time regardless of the fact fsync
+         * was actually started or nmot. server.aof_last_fsync is not
+         * used to really know when the latest fsync succeeded, but just
+         * as a timer to *try* to fsync once every second. */
         server.aof_last_fsync = server.unixtime;
     }
 }
@@ -716,7 +793,7 @@ int loadAppendOnlyFile(char *filename) {
             }
             if (buf[0] != '$') goto fmterr;
             len = strtol(buf+1,NULL,10);
-            argsds = sdsnewlen(NULL,len);
+            argsds = sdsnewlen(SDS_NOINIT,len);
             if (len && fread(argsds,len,1,fp) == 0) {
                 sdsfree(argsds);
                 fakeClient->argc = j; /* Free up to j-1. */
@@ -1031,6 +1108,37 @@ int rewriteHashObject(rio *r, robj *key, robj *o) {
     return 1;
 }
 
+/* Emit the commands needed to rebuild a stream object.
+ * The function returns 0 on error, 1 on success. */
+int rewriteStreamObject(rio *r, robj *key, robj *o) {
+    streamIterator si;
+    streamIteratorStart(&si,o->ptr,NULL,NULL,0);
+    streamID id;
+    int64_t numfields;
+
+    while(streamIteratorGetID(&si,&id,&numfields)) {
+        /* Emit a two elements array for each item. The first is
+         * the ID, the second is an array of field-value pairs. */
+
+        /* Emit the XADD <key> <id> ...fields... command. */
+        if (rioWriteBulkCount(r,'*',3+numfields*2) == 0) return 0;
+        if (rioWriteBulkString(r,"XADD",4) == 0) return 0;
+        if (rioWriteBulkObject(r,key) == 0) return 0;
+        sds replyid = sdscatfmt(sdsempty(),"%U.%U",id.ms,id.seq);
+        if (rioWriteBulkString(r,replyid,sdslen(replyid)) == 0) return 0;
+        sdsfree(replyid);
+        while(numfields--) {
+            unsigned char *field, *value;
+            int64_t field_len, value_len;
+            streamIteratorGetField(&si,&field,&value,&field_len,&value_len);
+            if (rioWriteBulkString(r,(char*)field,field_len) == 0) return 0;
+            if (rioWriteBulkString(r,(char*)value,value_len) == 0) return 0;
+        }
+    }
+    streamIteratorStop(&si);
+    return 1;
+}
+
 /* Call the module type callback in order to rewrite a data type
  * that is exported by a module and is not handled by Redis itself.
  * The function returns 0 on error, 1 on success. */
@@ -1111,6 +1219,8 @@ int rewriteAppendOnlyFileRio(rio *aof) {
                 if (rewriteSortedSetObject(aof,&key,o) == 0) goto werr;
             } else if (o->type == OBJ_HASH) {
                 if (rewriteHashObject(aof,&key,o) == 0) goto werr;
+            } else if (o->type == OBJ_STREAM) {
+                if (rewriteStreamObject(aof,&key,o) == 0) goto werr;
             } else if (o->type == OBJ_MODULE) {
                 if (rewriteModuleObject(aof,&key,o) == 0) goto werr;
             } else {
@@ -1285,7 +1395,7 @@ int aofCreatePipes(void) {
 
     if (pipe(fds) == -1) goto error; /* parent -> children data. */
     if (pipe(fds+2) == -1) goto error; /* children -> parent ack. */
-    if (pipe(fds+4) == -1) goto error; /* children -> parent ack. */
+    if (pipe(fds+4) == -1) goto error; /* parent -> children ack. */
     /* Parent -> children data is non blocking. */
     if (anetNonBlock(NULL,fds[0]) != ANET_OK) goto error;
     if (anetNonBlock(NULL,fds[1]) != ANET_OK) goto error;
@@ -1499,10 +1609,10 @@ void backgroundRewriteDoneHandler(int exitcode, int bysignal) {
         if (server.aof_fd == -1) {
             /* AOF disabled */
 
-             /* Don't care if this fails: oldfd will be -1 and we handle that.
-              * One notable case of -1 return is if the old file does
-              * not exist. */
-             oldfd = open(server.aof_filename,O_RDONLY|O_NONBLOCK);
+            /* Don't care if this fails: oldfd will be -1 and we handle that.
+             * One notable case of -1 return is if the old file does
+             * not exist. */
+            oldfd = open(server.aof_filename,O_RDONLY|O_NONBLOCK);
         } else {
             /* AOF enabled */
             oldfd = -1; /* We'll set this to the current AOF filedes later. */
@@ -1532,10 +1642,11 @@ void backgroundRewriteDoneHandler(int exitcode, int bysignal) {
             /* AOF enabled, replace the old fd with the new one. */
             oldfd = server.aof_fd;
             server.aof_fd = newfd;
-            if (server.aof_fsync == AOF_FSYNC_ALWAYS)
+            if (server.aof_fsync == AOF_FSYNC_ALWAYS) {
                 aof_fsync(newfd);
-            else if (server.aof_fsync == AOF_FSYNC_EVERYSEC)
-                aof_background_fsync(newfd);
+            } else if (server.aof_fsync == AOF_FSYNC_EVERYSEC) {
+                aofStartBackgroundFsync();
+            }
             server.aof_selected_db = -1; /* Make sure SELECT is re-issued */
             aofUpdateCurrentSize();
             server.aof_rewrite_base_size = server.aof_current_size;

src/blocked.c

@@ -65,6 +65,8 @@
 
 #include "server.h"
 
+int serveClientBlockedOnList(client *receiver, robj *key, robj *dstkey, redisDb *db, robj *value, int where);
+
 /* Get a timeout value from an object and store it into 'timeout'.
  * The final timeout is always stored as milliseconds as a time where the
  * timeout will expire, however the parsing is performed according to
@@ -100,7 +102,8 @@ int getTimeoutFromObjectOrReply(client *c, robj *object, mstime_t *timeout, int
 void blockClient(client *c, int btype) {
     c->flags |= CLIENT_BLOCKED;
     c->btype = btype;
-    server.bpop_blocked_clients++;
+    server.blocked_clients++;
+    server.blocked_clients_by_type[btype]++;
 }
 
 /* This function is called in the beforeSleep() function of the event loop
@@ -132,10 +135,12 @@ void processUnblockedClients(void) {
 /* Unblock a client calling the right function depending on the kind
  * of operation the client is blocking for. */
 void unblockClient(client *c) {
-    if (c->btype == BLOCKED_LIST) {
+    if (c->btype == BLOCKED_LIST || c->btype == BLOCKED_STREAM) {
         unblockClientWaitingData(c);
     } else if (c->btype == BLOCKED_WAIT) {
         unblockClientWaitingReplicas(c);
+    } else if (c->btype == BLOCKED_AOF) {
+        unblockClientWaitingReplicas(c);
     } else if (c->btype == BLOCKED_MODULE) {
         unblockClientFromModule(c);
     } else {
@@ -143,9 +148,10 @@ void unblockClient(client *c) {
     }
     /* Clear the flags, and put the client in the unblocked list so that
      * we'll process new commands in its query buffer ASAP. */
+    server.blocked_clients--;
+    server.blocked_clients_by_type[c->btype]--;
     c->flags &= ~CLIENT_BLOCKED;
     c->btype = BLOCKED_NONE;
-    server.bpop_blocked_clients--;
     /* The client may already be into the unblocked list because of a previous
      * blocking operation, don't add back it into the list multiple times. */
     if (!(c->flags & CLIENT_UNBLOCKED)) {
@@ -158,10 +164,12 @@ void unblockClient(client *c) {
  * send it a reply of some kind. After this function is called,
  * unblockClient() will be called with the same client as argument. */
 void replyToBlockedClientTimedOut(client *c) {
-    if (c->btype == BLOCKED_LIST) {
+    if (c->btype == BLOCKED_LIST || c->btype == BLOCKED_STREAM) {
         addReply(c,shared.nullmultibulk);
     } else if (c->btype == BLOCKED_WAIT) {
         addReplyLongLong(c,replicationCountAcksByOffset(c->bpop.reploffset));
+    } else if (c->btype == BLOCKED_AOF) {
+        addReply(c,shared.czero);
     } else if (c->btype == BLOCKED_MODULE) {
         moduleBlockedClientTimedOut(c);
     } else {
@@ -193,3 +201,283 @@ void disconnectAllBlockedClients(void) {
         }
     }
 }
+
+/* This function should be called by Redis every time a single command,
+ * a MULTI/EXEC block, or a Lua script, terminated its execution after
+ * being called by a client.
+ *
+ * All the keys with at least one client blocked that received at least
+ * one new element via some PUSH/XADD operation are accumulated into
+ * the server.ready_keys list. This function will run the list and will
+ * serve clients accordingly. Note that the function will iterate again and
+ * again as a result of serving BRPOPLPUSH we can have new blocking clients
+ * to serve because of the PUSH side of BRPOPLPUSH. */
+void handleClientsBlockedOnKeys(void) {
+    while(listLength(server.ready_keys) != 0) {
+        list *l;
+
+        /* Point server.ready_keys to a fresh list and save the current one
+         * locally. This way as we run the old list we are free to call
+         * signalKeyAsReady() that may push new elements in server.ready_keys
+         * when handling clients blocked into BRPOPLPUSH. */
+        l = server.ready_keys;
+        server.ready_keys = listCreate();
+
+        while(listLength(l) != 0) {
+            listNode *ln = listFirst(l);
+            readyList *rl = ln->value;
+
+            /* First of all remove this key from db->ready_keys so that
+             * we can safely call signalKeyAsReady() against this key. */
+            dictDelete(rl->db->ready_keys,rl->key);
+
+            /* Serve clients blocked on list key. */
+            robj *o = lookupKeyWrite(rl->db,rl->key);
+            if (o != NULL && o->type == OBJ_LIST) {
+                dictEntry *de;
+
+                /* We serve clients in the same order they blocked for
+                 * this key, from the first blocked to the last. */
+                de = dictFind(rl->db->blocking_keys,rl->key);
+                if (de) {
+                    list *clients = dictGetVal(de);
+                    int numclients = listLength(clients);
+
+                    while(numclients--) {
+                        listNode *clientnode = listFirst(clients);
+                        client *receiver = clientnode->value;
+
+                        if (receiver->btype != BLOCKED_LIST) {
+                            /* Put on the tail, so that at the next call
+                             * we'll not run into it again. */
+                            listDelNode(clients,clientnode);
+                            listAddNodeTail(clients,receiver);
+                            continue;
+                        }
+
+                        robj *dstkey = receiver->bpop.target;
+                        int where = (receiver->lastcmd &&
+                                     receiver->lastcmd->proc == blpopCommand) ?
+                                    LIST_HEAD : LIST_TAIL;
+                        robj *value = listTypePop(o,where);
+
+                        if (value) {
+                            /* Protect receiver->bpop.target, that will be
+                             * freed by the next unblockClient()
+                             * call. */
+                            if (dstkey) incrRefCount(dstkey);
+                            unblockClient(receiver);
+
+                            if (serveClientBlockedOnList(receiver,
+                                rl->key,dstkey,rl->db,value,
+                                where) == C_ERR)
+                            {
+                                /* If we failed serving the client we need
+                                 * to also undo the POP operation. */
+                                    listTypePush(o,value,where);
+                            }
+
+                            if (dstkey) decrRefCount(dstkey);
+                            decrRefCount(value);
+                        } else {
+                            break;
+                        }
+                    }
+                }
+
+                if (listTypeLength(o) == 0) {
+                    dbDelete(rl->db,rl->key);
+                }
+                /* We don't call signalModifiedKey() as it was already called
+                 * when an element was pushed on the list. */
+            }
+
+            /* Serve clients blocked on stream key. */
+            else if (o != NULL && o->type == OBJ_STREAM) {
+                dictEntry *de = dictFind(rl->db->blocking_keys,rl->key);
+                stream *s = o->ptr;
+
+                /* We need to provide the new data arrived on the stream
+                 * to all the clients that are waiting for an offset smaller
+                 * than the current top item. */
+                if (de) {
+                    list *clients = dictGetVal(de);
+                    listNode *ln;
+                    listIter li;
+                    listRewind(clients,&li);
+
+                    while((ln = listNext(&li))) {
+                        client *receiver = listNodeValue(ln);
+                        if (receiver->btype != BLOCKED_STREAM) continue;
+                        streamID *gt = dictFetchValue(receiver->bpop.keys,
+                                                      rl->key);
+                        if (s->last_id.ms > gt->ms ||
+                            (s->last_id.ms == gt->ms &&
+                             s->last_id.seq > gt->seq))
+                        {
+                            streamID start = *gt;
+                            start.seq++; /* Can't overflow, it's an uint64_t */
+                            /* Note that after we unblock the client, 'gt'
+                             * is no longer valid, so we must do it after
+                             * we copied the ID into the 'start' variable. */
+                            unblockClient(receiver);
+
+                            /* Emit the two elements sub-array consisting of
+                             * the name of the stream and the data we
+                             * extracted from it. Wrapped in a single-item
+                             * array, since we have just one key. */
+                            addReplyMultiBulkLen(receiver,1);
+                            addReplyMultiBulkLen(receiver,2);
+                            addReplyBulk(receiver,rl->key);
+                            streamReplyWithRange(receiver,s,&start,NULL,
+                                                 receiver->bpop.xread_count,0);
+                        }
+                    }
+                }
+            }
+
+            /* Free this item. */
+            decrRefCount(rl->key);
+            zfree(rl);
+            listDelNode(l,ln);
+        }
+        listRelease(l); /* We have the new list on place at this point. */
+    }
+}
+
+/* This is how the current blocking lists/streams work, we use BLPOP as
+ * example, but the concept is the same for other list ops and XREAD.
+ * - If the user calls BLPOP and the key exists and contains a non empty list
+ *   then LPOP is called instead. So BLPOP is semantically the same as LPOP
+ *   if blocking is not required.
+ * - If instead BLPOP is called and the key does not exists or the list is
+ *   empty we need to block. In order to do so we remove the notification for
+ *   new data to read in the client socket (so that we'll not serve new
+ *   requests if the blocking request is not served). Also we put the client
+ *   in a dictionary (db->blocking_keys) mapping keys to a list of clients
+ *   blocking for this keys.
+ * - If a PUSH operation against a key with blocked clients waiting is
+ *   performed, we mark this key as "ready", and after the current command,
+ *   MULTI/EXEC block, or script, is executed, we serve all the clients waiting
+ *   for this list, from the one that blocked first, to the last, accordingly
+ *   to the number of elements we have in the ready list.
+ */
+
+/* Set a client in blocking mode for the specified key (list or stream), with
+ * the specified timeout. The 'type' argument is BLOCKED_LIST or BLOCKED_STREAM
+ * depending on the kind of operation we are waiting for an empty key in
+ * order to awake the client. The client is blocked for all the 'numkeys'
+ * keys as in the 'keys' argument. When we block for stream keys, we also
+ * provide an array of streamID structures: clients will be unblocked only
+ * when items with an ID greater or equal to the specified one is appended
+ * to the stream. */
+void blockForKeys(client *c, int btype, robj **keys, int numkeys, mstime_t timeout, robj *target, streamID *ids) {
+    dictEntry *de;
+    list *l;
+    int j;
+
+    c->bpop.timeout = timeout;
+    c->bpop.target = target;
+
+    if (target != NULL) incrRefCount(target);
+
+    for (j = 0; j < numkeys; j++) {
+        /* The value associated with the key name in the bpop.keys dictionary
+         * is NULL for lists, or the stream ID for streams. */
+        void *key_data = NULL;
+        if (btype == BLOCKED_STREAM) {
+            key_data = zmalloc(sizeof(streamID));
+            memcpy(key_data,ids+j,sizeof(streamID));
+        }
+
+        /* If the key already exists in the dictionary ignore it. */
+        if (dictAdd(c->bpop.keys,keys[j],key_data) != DICT_OK) {
+            zfree(key_data);
+            continue;
+        }
+        incrRefCount(keys[j]);
+
+        /* And in the other "side", to map keys -> clients */
+        de = dictFind(c->db->blocking_keys,keys[j]);
+        if (de == NULL) {
+            int retval;
+
+            /* For every key we take a list of clients blocked for it */
+            l = listCreate();
+            retval = dictAdd(c->db->blocking_keys,keys[j],l);
+            incrRefCount(keys[j]);
+            serverAssertWithInfo(c,keys[j],retval == DICT_OK);
+        } else {
+            l = dictGetVal(de);
+        }
+        listAddNodeTail(l,c);
+    }
+    blockClient(c,btype);
+}
+
+/* Unblock a client that's waiting in a blocking operation such as BLPOP.
+ * You should never call this function directly, but unblockClient() instead. */
+void unblockClientWaitingData(client *c) {
+    dictEntry *de;
+    dictIterator *di;
+    list *l;
+
+    serverAssertWithInfo(c,NULL,dictSize(c->bpop.keys) != 0);
+    di = dictGetIterator(c->bpop.keys);
+    /* The client may wait for multiple keys, so unblock it for every key. */
+    while((de = dictNext(di)) != NULL) {
+        robj *key = dictGetKey(de);
+
+        /* Remove this client from the list of clients waiting for this key. */
+        l = dictFetchValue(c->db->blocking_keys,key);
+        serverAssertWithInfo(c,key,l != NULL);
+        listDelNode(l,listSearchKey(l,c));
+        /* If the list is empty we need to remove it to avoid wasting memory */
+        if (listLength(l) == 0)
+            dictDelete(c->db->blocking_keys,key);
+    }
+    dictReleaseIterator(di);
+
+    /* Cleanup the client structure */
+    dictEmpty(c->bpop.keys,NULL);
+    if (c->bpop.target) {
+        decrRefCount(c->bpop.target);
+        c->bpop.target = NULL;
+    }
+    if (c->bpop.xread_group) {
+        decrRefCount(c->bpop.xread_group);
+        c->bpop.xread_group = NULL;
+    }
+}
+
+/* If the specified key has clients blocked waiting for list pushes, this
+ * function will put the key reference into the server.ready_keys list.
+ * Note that db->ready_keys is a hash table that allows us to avoid putting
+ * the same key again and again in the list in case of multiple pushes
+ * made by a script or in the context of MULTI/EXEC.
+ *
+ * The list will be finally processed by handleClientsBlockedOnLists() */
+void signalKeyAsReady(redisDb *db, robj *key) {
+    readyList *rl;
+
+    /* No clients blocking for this key? No need to queue it. */
+    if (dictFind(db->blocking_keys,key) == NULL) return;
+
+    /* Key was already signaled? No need to queue it again. */
+    if (dictFind(db->ready_keys,key) != NULL) return;
+
+    /* Ok, we need to queue this key into server.ready_keys. */
+    rl = zmalloc(sizeof(*rl));
+    rl->key = key;
+    rl->db = db;
+    incrRefCount(key);
+    listAddNodeTail(server.ready_keys,rl);
+
+    /* We also add the key in the db->ready_keys dictionary in order
+     * to avoid adding it multiple times into a list with a simple O(1)
+     * check. */
+    incrRefCount(key);
+    serverAssert(dictAdd(db->ready_keys,key,NULL) == DICT_OK);
+}
+
+

src/cluster.c

@@ -653,7 +653,7 @@ unsigned int keyHashSlot(char *key, int keylen) {
     for (e = s+1; e < keylen; e++)
         if (key[e] == '}') break;
 
-    /* No '}' or nothing betweeen {} ? Hash the whole key. */
+    /* No '}' or nothing between {} ? Hash the whole key. */
     if (e == keylen || e == s+1) return crc16(key,keylen) & 0x3FFF;
 
     /* If we are here there is both a { and a } on its right. Hash
@@ -2156,7 +2156,7 @@ void clusterReadHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
  * from event handlers that will do stuff with the same link later. */
 void clusterSendMessage(clusterLink *link, unsigned char *msg, size_t msglen) {
     if (sdslen(link->sndbuf) == 0 && msglen != 0)
-        aeCreateFileEvent(server.el,link->fd,AE_WRITABLE,
+        aeCreateFileEvent(server.el,link->fd,AE_WRITABLE|AE_BARRIER,
                     clusterWriteHandler,link);
 
     link->sndbuf = sdscatlen(link->sndbuf, msg, msglen);
@@ -2691,9 +2691,10 @@ void clusterSendFailoverAuthIfNeeded(clusterNode *node, clusterMsg *request) {
     }
 
     /* We can vote for this slave. */
-    clusterSendFailoverAuth(node);
     server.cluster->lastVoteEpoch = server.cluster->currentEpoch;
     node->slaveof->voted_time = mstime();
+    clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|CLUSTER_TODO_FSYNC_CONFIG);
+    clusterSendFailoverAuth(node);
     serverLog(LL_WARNING, "Failover auth granted to %.40s for epoch %llu",
         node->name, (unsigned long long) server.cluster->currentEpoch);
 }
@@ -4065,7 +4066,34 @@ void clusterCommand(client *c) {
         return;
     }
 
-    if (!strcasecmp(c->argv[1]->ptr,"meet") && (c->argc == 4 || c->argc == 5)) {
+    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"help")) {
+        const char *help[] = {
+"addslots <slot> [slot ...] -- Assign slots to current node.",
+"bumpepoch -- Advance the cluster config epoch.",
+"count-failure-reports <node-id> -- Return number of failure reports for <node-id>.",
+"countkeysinslot <slot> - Return the number of keys in <slot>.",
+"delslots <slot> [slot ...] -- Delete slots information from current node.",
+"failover [force|takeover] -- Promote current slave node to being a master.",
+"forget <node-id> -- Remove a node from the cluster.",
+"getkeysinslot <slot> <count> -- Return key names stored by current node in a slot.",
+"flushslots -- Delete current node own slots information.",
+"info - Return onformation about the cluster.",
+"keyslot <key> -- Return the hash slot for <key>.",
+"meet <ip> <port> [bus-port] -- Connect nodes into a working cluster.",
+"myid -- Return the node id.",
+"nodes -- Return cluster configuration seen by node. Output format:",
+"    <id> <ip:port> <flags> <master> <pings> <pongs> <epoch> <link> <slot> ... <slot>",
+"replicate <node-id> -- Configure current node as slave to <node-id>.",
+"reset [hard|soft] -- Reset current node (default: soft).",
+"set-config-epoch <epoch> - Set config epoch of current node.",
+"setslot <slot> (importing|migrating|stable|node <node-id>) -- Set slot state.",
+"slaves <node-id> -- Return <node-id> slaves.",
+"slots -- Return information about slots range mappings. Each range is made of:",
+"    start, end, master and replicas IP addresses, ports and ids",
+NULL
+        };
+        addReplyHelp(c, help);
+    } else if (!strcasecmp(c->argv[1]->ptr,"meet") && (c->argc == 4 || c->argc == 5)) {
         /* CLUSTER MEET <ip> <port> [cport] */
         long long port, cport;
 
@@ -4258,7 +4286,7 @@ void clusterCommand(client *c) {
             clusterAddSlot(n,slot);
         } else {
             addReplyError(c,
-                "Invalid CLUSTER SETSLOT action or number of arguments");
+                "Invalid CLUSTER SETSLOT action or number of arguments. Try CLUSTER HELP");
             return;
         }
         clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|CLUSTER_TODO_UPDATE_STATE);
@@ -4608,7 +4636,9 @@ void clusterCommand(client *c) {
         clusterReset(hard);
         addReply(c,shared.ok);
     } else {
-        addReplyError(c,"Wrong CLUSTER subcommand or number of arguments");
+         addReplyErrorFormat(c, "Unknown subcommand or wrong number of arguments for '%s'. Try CLUSTER HELP",
+            (char*)c->argv[1]->ptr);
+        return;
     }
 }
 
@@ -4867,14 +4897,16 @@ void migrateCloseTimedoutSockets(void) {
     dictReleaseIterator(di);
 }
 
-/* MIGRATE host port key dbid timeout [COPY | REPLACE]
+/* MIGRATE host port key dbid timeout [COPY | REPLACE | AUTH password]
  *
  * On in the multiple keys form:
  *
- * MIGRATE host port "" dbid timeout [COPY | REPLACE] KEYS key1 key2 ... keyN */
+ * MIGRATE host port "" dbid timeout [COPY | REPLACE | AUTH password] KEYS key1
+ * key2 ... keyN */
 void migrateCommand(client *c) {
     migrateCachedSocket *cs;
-    int copy, replace, j;
+    int copy = 0, replace = 0, j;
+    char *password = NULL;
     long timeout;
     long dbid;
     robj **ov = NULL; /* Objects to migrate. */
@@ -4889,16 +4921,20 @@ void migrateCommand(client *c) {
     int first_key = 3; /* Argument index of the first key. */
     int num_keys = 1;  /* By default only migrate the 'key' argument. */
 
-    /* Initialization */
-    copy = 0;
-    replace = 0;
-
     /* Parse additional options */
     for (j = 6; j < c->argc; j++) {
+        int moreargs = j < c->argc-1;
         if (!strcasecmp(c->argv[j]->ptr,"copy")) {
             copy = 1;
         } else if (!strcasecmp(c->argv[j]->ptr,"replace")) {
             replace = 1;
+        } else if (!strcasecmp(c->argv[j]->ptr,"auth")) {
+            if (!moreargs) {
+                addReply(c,shared.syntaxerr);
+                return;
+            }
+            j++;
+            password = c->argv[j]->ptr;
         } else if (!strcasecmp(c->argv[j]->ptr,"keys")) {
             if (sdslen(c->argv[3]->ptr) != 0) {
                 addReplyError(c,
@@ -4957,6 +4993,14 @@ try_again:
 
     rioInitWithBuffer(&cmd,sdsempty());
 
+    /* Authentication */
+    if (password) {
+        serverAssertWithInfo(c,NULL,rioWriteBulkCount(&cmd,'*',2));
+        serverAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,"AUTH",4));
+        serverAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,password,
+            sdslen(password)));
+    }
+
     /* Send the SELECT command if the current DB is not already selected. */
     int select = cs->last_dbid != dbid; /* Should we emit SELECT? */
     if (select) {
@@ -4974,7 +5018,9 @@ try_again:
             ttl = expireat-mstime();
             if (ttl < 1) ttl = 1;
         }
-        serverAssertWithInfo(c,NULL,rioWriteBulkCount(&cmd,'*',replace ? 5 : 4));
+        serverAssertWithInfo(c,NULL,
+            rioWriteBulkCount(&cmd,'*',replace ? 5 : 4));
+
         if (server.cluster_enabled)
             serverAssertWithInfo(c,NULL,
                 rioWriteBulkString(&cmd,"RESTORE-ASKING",14));
@@ -5017,9 +5063,14 @@ try_again:
         }
     }
 
+    char buf0[1024]; /* Auth reply. */
     char buf1[1024]; /* Select reply. */
     char buf2[1024]; /* Restore reply. */
 
+    /* Read the AUTH reply if needed. */
+    if (password && syncReadLine(cs->fd, buf0, sizeof(buf0), timeout) <= 0)
+        goto socket_err;
+
     /* Read the SELECT reply if needed. */
     if (select && syncReadLine(cs->fd, buf1, sizeof(buf1), timeout) <= 0)
         goto socket_err;
@@ -5036,13 +5087,21 @@ try_again:
             socket_error = 1;
             break;
         }
-        if ((select && buf1[0] == '-') || buf2[0] == '-') {
+        if ((password && buf0[0] == '-') ||
+            (select && buf1[0] == '-') ||
+            buf2[0] == '-')
+        {
             /* On error assume that last_dbid is no longer valid. */
             if (!error_from_target) {
                 cs->last_dbid = -1;
-                addReplyErrorFormat(c,"Target instance replied with error: %s",
-                    (select && buf1[0] == '-') ? buf1+1 : buf2+1);
+                char *errbuf;
+                if (password && buf0[0] == '-') errbuf = buf0;
+                else if (select && buf1[0] == '-') errbuf = buf1;
+                else errbuf = buf2;
+
                 error_from_target = 1;
+                addReplyErrorFormat(c,"Target instance replied with error: %s",
+                    errbuf+1);
             }
         } else {
             if (!copy) {
@@ -5107,7 +5166,7 @@ try_again:
         addReply(c,shared.ok);
     } else {
         /* On error we already sent it in the for loop above, and set
-         * the curretly selected socket to -1 to force SELECT the next time. */
+         * the currently selected socket to -1 to force SELECT the next time. */
     }
 
     sdsfree(cmd.io.buffer.ptr);
@@ -5363,7 +5422,8 @@ clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, in
      * node is a slave and the request is about an hash slot our master
      * is serving, we can reply without redirection. */
     if (c->flags & CLIENT_READONLY &&
-        cmd->flags & CMD_READONLY &&
+        (cmd->flags & CMD_READONLY || cmd->proc == evalCommand ||
+         cmd->proc == evalShaCommand) &&
         nodeIsSlave(myself) &&
         myself->slaveof == n)
     {

src/config.c

@@ -328,15 +328,19 @@ void loadServerConfigFromString(char *config) {
                 err = "maxmemory-samples must be 1 or greater";
                 goto loaderr;
             }
+        } else if ((!strcasecmp(argv[0],"proto-max-bulk-len")) && argc == 2) {
+            server.proto_max_bulk_len = memtoll(argv[1],NULL);
+        } else if ((!strcasecmp(argv[0],"client-query-buffer-limit")) && argc == 2) {
+            server.client_max_querybuf_len = memtoll(argv[1],NULL);
         } else if (!strcasecmp(argv[0],"lfu-log-factor") && argc == 2) {
             server.lfu_log_factor = atoi(argv[1]);
-            if (server.maxmemory_samples < 0) {
+            if (server.lfu_log_factor < 0) {
                 err = "lfu-log-factor must be 0 or greater";
                 goto loaderr;
             }
         } else if (!strcasecmp(argv[0],"lfu-decay-time") && argc == 2) {
             server.lfu_decay_time = atoi(argv[1]);
-            if (server.maxmemory_samples < 1) {
+            if (server.lfu_decay_time < 0) {
                 err = "lfu-decay-time must be 0 or greater";
                 goto loaderr;
             }
@@ -1132,6 +1136,10 @@ void configSetCommand(client *c) {
             }
             freeMemoryIfNeeded();
         }
+    } config_set_memory_field(
+      "proto-max-bulk-len",server.proto_max_bulk_len) {
+    } config_set_memory_field(
+      "client-query-buffer-limit",server.client_max_querybuf_len) {
     } config_set_memory_field("repl-backlog-size",ll) {
         resizeReplicationBacklog(ll);
     } config_set_memory_field("auto-aof-rewrite-min-size",ll) {
@@ -1220,7 +1228,11 @@ void configGetCommand(client *c) {
 
     /* Numerical values */
     config_get_numerical_field("maxmemory",server.maxmemory);
+    config_get_numerical_field("proto-max-bulk-len",server.proto_max_bulk_len);
+    config_get_numerical_field("client-query-buffer-limit",server.client_max_querybuf_len);
     config_get_numerical_field("maxmemory-samples",server.maxmemory_samples);
+    config_get_numerical_field("lfu-log-factor",server.lfu_log_factor);
+    config_get_numerical_field("lfu-decay-time",server.lfu_decay_time);
     config_get_numerical_field("timeout",server.maxidletime);
     config_get_numerical_field("active-defrag-threshold-lower",server.active_defrag_threshold_lower);
     config_get_numerical_field("active-defrag-threshold-upper",server.active_defrag_threshold_upper);
@@ -1990,8 +2002,12 @@ int rewriteConfig(char *path) {
     rewriteConfigStringOption(state,"requirepass",server.requirepass,NULL);
     rewriteConfigNumericalOption(state,"maxclients",server.maxclients,CONFIG_DEFAULT_MAX_CLIENTS);
     rewriteConfigBytesOption(state,"maxmemory",server.maxmemory,CONFIG_DEFAULT_MAXMEMORY);
+    rewriteConfigBytesOption(state,"proto-max-bulk-len",server.proto_max_bulk_len,CONFIG_DEFAULT_PROTO_MAX_BULK_LEN);
+    rewriteConfigBytesOption(state,"client-query-buffer-limit",server.client_max_querybuf_len,PROTO_MAX_QUERYBUF_LEN);
     rewriteConfigEnumOption(state,"maxmemory-policy",server.maxmemory_policy,maxmemory_policy_enum,CONFIG_DEFAULT_MAXMEMORY_POLICY);
     rewriteConfigNumericalOption(state,"maxmemory-samples",server.maxmemory_samples,CONFIG_DEFAULT_MAXMEMORY_SAMPLES);
+    rewriteConfigNumericalOption(state,"lfu-log-factor",server.lfu_log_factor,CONFIG_DEFAULT_LFU_LOG_FACTOR);
+    rewriteConfigNumericalOption(state,"lfu-decay-time",server.lfu_decay_time,CONFIG_DEFAULT_LFU_DECAY_TIME);
     rewriteConfigNumericalOption(state,"active-defrag-threshold-lower",server.active_defrag_threshold_lower,CONFIG_DEFAULT_DEFRAG_THRESHOLD_LOWER);
     rewriteConfigNumericalOption(state,"active-defrag-threshold-upper",server.active_defrag_threshold_upper,CONFIG_DEFAULT_DEFRAG_THRESHOLD_UPPER);
     rewriteConfigBytesOption(state,"active-defrag-ignore-bytes",server.active_defrag_ignore_bytes,CONFIG_DEFAULT_DEFRAG_IGNORE_BYTES);
@@ -2065,19 +2081,24 @@ void configCommand(client *c) {
         return;
     }
 
-    if (!strcasecmp(c->argv[1]->ptr,"set")) {
-        if (c->argc != 4) goto badarity;
+    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"help")) {
+        const char *help[] = {
+"get <pattern> -- Return parameters matching the glob-like <pattern> and their values.",
+"set <parameter> <value> -- Set parameter to value.",
+"resetstat -- Reset statistics reported by INFO.",
+"rewrite -- Rewrite the configuration file.",
+NULL
+        };
+        addReplyHelp(c, help);
+    } else if (!strcasecmp(c->argv[1]->ptr,"set") && c->argc == 4) {
         configSetCommand(c);
-    } else if (!strcasecmp(c->argv[1]->ptr,"get")) {
-        if (c->argc != 3) goto badarity;
+    } else if (!strcasecmp(c->argv[1]->ptr,"get") && c->argc == 3) {
         configGetCommand(c);
-    } else if (!strcasecmp(c->argv[1]->ptr,"resetstat")) {
-        if (c->argc != 2) goto badarity;
+    } else if (!strcasecmp(c->argv[1]->ptr,"resetstat") && c->argc == 2) {
         resetServerStats();
         resetCommandTableStats();
         addReply(c,shared.ok);
-    } else if (!strcasecmp(c->argv[1]->ptr,"rewrite")) {
-        if (c->argc != 2) goto badarity;
+    } else if (!strcasecmp(c->argv[1]->ptr,"rewrite") && c->argc == 2) {
         if (server.configfile == NULL) {
             addReplyError(c,"The server is running without a config file");
             return;
@@ -2090,12 +2111,8 @@ void configCommand(client *c) {
             addReply(c,shared.ok);
         }
     } else {
-        addReplyError(c,
-            "CONFIG subcommand must be one of GET, SET, RESETSTAT, REWRITE");
+         addReplyErrorFormat(c, "Unknown subcommand or wrong number of arguments for '%s'. Try CONFIG HELP",
+            (char*)c->argv[1]->ptr);
+        return;
     }
-    return;
-
-badarity:
-    addReplyErrorFormat(c,"Wrong number of arguments for CONFIG %s",
-        (char*) c->argv[1]->ptr);
 }

src/db.c

@@ -38,6 +38,15 @@
  * C-level DB API
  *----------------------------------------------------------------------------*/
 
+/* Update LFU when an object is accessed.
+ * Firstly, decrement the counter if the decrement time is reached.
+ * Then logarithmically increment the counter, and update the access time. */
+void updateLFU(robj *val) {
+    unsigned long counter = LFUDecrAndReturn(val);
+    counter = LFULogIncr(counter);
+    val->lru = (LFUGetTimeInMinutes()<<8) | counter;
+}
+
 /* Low level key lookup API, not actually called directly from commands
  * implementations that should instead rely on lookupKeyRead(),
  * lookupKeyWrite() and lookupKeyReadWithFlags(). */
@@ -54,9 +63,7 @@ robj *lookupKey(redisDb *db, robj *key, int flags) {
             !(flags & LOOKUP_NOTOUCH))
         {
             if (server.maxmemory_policy & MAXMEMORY_FLAG_LFU) {
-                unsigned long ldt = val->lru >> 8;
-                unsigned long counter = LFULogIncr(val->lru & 255);
-                val->lru = (ldt << 8) | counter;
+                updateLFU(val);
             } else {
                 val->lru = LRU_CLOCK();
             }
@@ -162,9 +169,9 @@ void dbAdd(redisDb *db, robj *key, robj *val) {
     int retval = dictAdd(db->dict, copy, val);
 
     serverAssertWithInfo(NULL,key,retval == DICT_OK);
-    if (val->type == OBJ_LIST) signalListAsReady(db, key);
+    if (val->type == OBJ_LIST) signalKeyAsReady(db, key);
     if (server.cluster_enabled) slotToKeyAdd(key);
- }
+}
 
 /* Overwrite an existing key with a new value. Incrementing the reference
  * count of the new value is up to the caller.
@@ -180,6 +187,9 @@ void dbOverwrite(redisDb *db, robj *key, robj *val) {
         int saved_lru = old->lru;
         dictReplace(db->dict, key->ptr, val);
         val->lru = saved_lru;
+        /* LFU should be not only copied but also updated
+         * when a key is overwritten. */
+        updateLFU(val);
     } else {
         dictReplace(db->dict, key->ptr, val);
     }
@@ -788,6 +798,7 @@ void typeCommand(client *c) {
         case OBJ_SET: type = "set"; break;
         case OBJ_ZSET: type = "zset"; break;
         case OBJ_HASH: type = "hash"; break;
+        case OBJ_STREAM: type = "stream"; break;
         case OBJ_MODULE: {
             moduleValue *mv = o->ptr;
             type = mv->type->name;
@@ -941,8 +952,8 @@ void scanDatabaseForReadyLists(redisDb *db) {
     while((de = dictNext(di)) != NULL) {
         robj *key = dictGetKey(de);
         robj *value = lookupKey(db,key,LOOKUP_NOTOUCH);
-        if (value && value->type == OBJ_LIST)
-            signalListAsReady(db, key);
+        if (value && (value->type == OBJ_LIST || value->type == OBJ_STREAM))
+            signalKeyAsReady(db, key);
     }
     dictReleaseIterator(di);
 }
@@ -1084,6 +1095,25 @@ void propagateExpire(redisDb *db, robj *key, int lazy) {
     decrRefCount(argv[1]);
 }
 
+/* This function is called when we are going to perform some operation
+ * in a given key, but such key may be already logically expired even if
+ * it still exists in the database. The main way this function is called
+ * is via lookupKey*() family of functions.
+ *
+ * The behavior of the function depends on the replication role of the
+ * instance, because slave instances do not expire keys, they wait
+ * for DELs from the master for consistency matters. However even
+ * slaves will try to have a coherent return value for the function,
+ * so that read commands executed in the slave side will be able to
+ * behave like if the key is expired even if still present (because the
+ * master has yet to propagate the DEL).
+ *
+ * In masters as a side effect of finding a key which is expired, such
+ * key will be evicted from the database. Also this may trigger the
+ * propagation of a DEL/UNLINK command in AOF / replication stream.
+ *
+ * The return value of the function is 0 if the key is still valid,
+ * otherwise the function returns 1 if the key is expired. */
 int expireIfNeeded(redisDb *db, robj *key) {
     mstime_t when = getExpire(db,key);
     mstime_t now;
@@ -1093,7 +1123,7 @@ int expireIfNeeded(redisDb *db, robj *key) {
     /* Don't expire anything while loading. It will be done later. */
     if (server.loading) return 0;
 
-    /* If we are in the context of a Lua script, we claim that time is
+    /* If we are in the context of a Lua script, we pretend that time is
      * blocked to when the Lua script started. This way a key can expire
      * only the first time it is accessed and not in the middle of the
      * script execution, making propagation to slaves / AOF consistent.
@@ -1141,11 +1171,13 @@ int *getKeysUsingCommandTable(struct redisCommand *cmd,robj **argv, int argc, in
     keys = zmalloc(sizeof(int)*((last - cmd->firstkey)+1));
     for (j = cmd->firstkey; j <= last; j += cmd->keystep) {
         if (j >= argc) {
-            /* Modules command do not have dispatch time arity checks, so
-             * we need to handle the case where the user passed an invalid
-             * number of arguments here. In this case we return no keys
-             * and expect the module command to report an arity error. */
-            if (cmd->flags & CMD_MODULE) {
+            /* Modules commands, and standard commands with a not fixed number
+             * of arugments (negative arity parameter) do not have dispatch
+             * time arity checks, so we need to handle the case where the user
+             * passed an invalid number of arguments here. In this case we
+             * return no keys and expect the command implementation to report
+             * an arity or syntax error. */
+            if (cmd->flags & CMD_MODULE || cmd->arity < 0) {
                 zfree(keys);
                 *numkeys = 0;
                 return NULL;
@@ -1352,6 +1384,36 @@ int *georadiusGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numk
     return keys;
 }
 
+/* XREAD [BLOCK <milliseconds>] [COUNT <count>] [GROUP <groupname> <ttl>]
+ *       [RETRY <milliseconds> <ttl>] STREAMS key_1 key_2 ... key_N
+ *       ID_1 ID_2 ... ID_N */
+int *xreadGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {
+    int i, num, *keys;
+    UNUSED(cmd);
+
+    /* We need to seek the last argument that contains "STREAMS", because other
+     * arguments before may contain it (for example the group name). */
+    int streams_pos = -1;
+    for (i = 1; i < argc; i++) {
+        char *arg = argv[i]->ptr;
+        if (!strcasecmp(arg, "streams")) streams_pos = i;
+    }
+    if (streams_pos != -1) num = argc - streams_pos - 1;
+
+    /* Syntax error. */
+    if (streams_pos == -1 || num % 2 != 0) {
+        *numkeys = 0;
+        return NULL;
+    }
+    num /= 2; /* We have half the keys as there are arguments because
+                 there are also the IDs, one per key. */
+
+    keys = zmalloc(sizeof(int) * num);
+    for (i = streams_pos+1; i < argc; i++) keys[i-streams_pos-1] = i;
+    *numkeys = num;
+    return keys;
+}
+
 /* Slot to Key API. This is used by Redis Cluster in order to obtain in
  * a fast way a key that belongs to a specified hash slot. This is useful
  * while rehashing the cluster and in other conditions when we need to

src/debug.c

@@ -239,6 +239,27 @@ void computeDatasetDigest(unsigned char *final) {
                     xorDigest(digest,eledigest,20);
                 }
                 hashTypeReleaseIterator(hi);
+            } else if (o->type == OBJ_STREAM) {
+                streamIterator si;
+                streamIteratorStart(&si,o->ptr,NULL,NULL,0);
+                streamID id;
+                int64_t numfields;
+
+                while(streamIteratorGetID(&si,&id,&numfields)) {
+                    sds itemid = sdscatfmt(sdsempty(),"%U.%U",id.ms,id.seq);
+                    mixDigest(digest,itemid,sdslen(itemid));
+                    sdsfree(itemid);
+
+                    while(numfields--) {
+                        unsigned char *field, *value;
+                        int64_t field_len, value_len;
+                        streamIteratorGetField(&si,&field,&value,
+                                                   &field_len,&value_len);
+                        mixDigest(digest,field,field_len);
+                        mixDigest(digest,value,value_len);
+                    }
+                }
+                streamIteratorStop(&si);
             } else if (o->type == OBJ_MODULE) {
                 RedisModuleDigest md;
                 moduleValue *mv = o->ptr;
@@ -262,53 +283,30 @@ void computeDatasetDigest(unsigned char *final) {
 }
 
 void debugCommand(client *c) {
-    if (c->argc == 1) {
-        addReplyError(c,"You must specify a subcommand for DEBUG. Try DEBUG HELP for info.");
-        return;
-    }
-
-    if (!strcasecmp(c->argv[1]->ptr,"help")) {
-        void *blenp = addDeferredMultiBulkLength(c);
-        int blen = 0;
-        blen++; addReplyStatus(c,
-        "DEBUG <subcommand> arg arg ... arg. Subcommands:");
-        blen++; addReplyStatus(c,
-        "segfault -- Crash the server with sigsegv.");
-        blen++; addReplyStatus(c,
-        "panic -- Crash the server simulating a panic.");
-        blen++; addReplyStatus(c,
-        "restart  -- Graceful restart: save config, db, restart.");
-        blen++; addReplyStatus(c,
-        "crash-and-recovery <milliseconds> -- Hard crash and restart after <milliseconds> delay.");
-        blen++; addReplyStatus(c,
-        "assert   -- Crash by assertion failed.");
-        blen++; addReplyStatus(c,
-        "reload   -- Save the RDB on disk and reload it back in memory.");
-        blen++; addReplyStatus(c,
-        "loadaof  -- Flush the AOF buffers on disk and reload the AOF in memory.");
-        blen++; addReplyStatus(c,
-        "object <key> -- Show low level info about key and associated value.");
-        blen++; addReplyStatus(c,
-        "sdslen <key> -- Show low level SDS string info representing key and value.");
-        blen++; addReplyStatus(c,
-        "ziplist <key> -- Show low level info about the ziplist encoding.");
-        blen++; addReplyStatus(c,
-        "populate <count> [prefix] [size] -- Create <count> string keys named key:<num>. If a prefix is specified is used instead of the 'key' prefix.");
-        blen++; addReplyStatus(c,
-        "digest   -- Outputs an hex signature representing the current DB content.");
-        blen++; addReplyStatus(c,
-        "sleep <seconds> -- Stop the server for <seconds>. Decimals allowed.");
-        blen++; addReplyStatus(c,
-        "set-active-expire (0|1) -- Setting it to 0 disables expiring keys in background when they are not accessed (otherwise the Redis behavior). Setting it to 1 reenables back the default.");
-        blen++; addReplyStatus(c,
-        "lua-always-replicate-commands (0|1) -- Setting it to 1 makes Lua replication defaulting to replicating single commands, without the script having to enable effects replication.");
-        blen++; addReplyStatus(c,
-        "error <string> -- Return a Redis protocol error with <string> as message. Useful for clients unit tests to simulate Redis errors.");
-        blen++; addReplyStatus(c,
-        "structsize -- Return the size of different Redis core C structures.");
-        blen++; addReplyStatus(c,
-        "htstats <dbid> -- Return hash table statistics of the specified Redis database.");
-        setDeferredMultiBulkLength(c,blenp,blen);
+    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"help")) {
+        const char *help[] = {
+"assert -- Crash by assertion failed.",
+"change-repl-id -- Change the replication IDs of the instance. Dangerous, should be used only for testing the replication subsystem.",
+"crash-and-recovery <milliseconds> -- Hard crash and restart after <milliseconds> delay.",
+"digest -- Outputs an hex signature representing the current DB content.",
+"htstats <dbid> -- Return hash table statistics of the specified Redis database.",
+"loadaof -- Flush the AOF buffers on disk and reload the AOF in memory.",
+"lua-always-replicate-commands (0|1) -- Setting it to 1 makes Lua replication defaulting to replicating single commands, without the script having to enable effects replication.",
+"object <key> -- Show low level info about key and associated value.",
+"panic -- Crash the server simulating a panic.",
+"populate <count> [prefix] [size] -- Create <count> string keys named key:<num>. If a prefix is specified is used instead of the 'key' prefix.",
+"reload -- Save the RDB on disk and reload it back in memory.",
+"restart -- Graceful restart: save config, db, restart.",
+"sdslen <key> -- Show low level SDS string info representing key and value.",
+"segfault -- Crash the server with sigsegv.",
+"set-active-expire (0|1) -- Setting it to 0 disables expiring keys in background when they are not accessed (otherwise the Redis behavior). Setting it to 1 reenables back the default.",
+"sleep <seconds> -- Stop the server for <seconds>. Decimals allowed.",
+"structsize -- Return the size of different Redis core C structures.",
+"ziplist <key> -- Show low level info about the ziplist encoding.",
+"error <string> -- Return a Redis protocol error with <string> as message. Useful for clients unit tests to simulate Redis errors.",
+NULL
+        };
+        addReplyHelp(c, help);
     } else if (!strcasecmp(c->argv[1]->ptr,"segfault")) {
         *((char*)-1) = 'x';
     } else if (!strcasecmp(c->argv[1]->ptr,"panic")) {
@@ -370,13 +368,13 @@ void debugCommand(client *c) {
         val = dictGetVal(de);
         strenc = strEncoding(val->encoding);
 
-        char extra[128] = {0};
+        char extra[138] = {0};
         if (val->encoding == OBJ_ENCODING_QUICKLIST) {
             char *nextra = extra;
             int remaining = sizeof(extra);
             quicklist *ql = val->ptr;
             /* Add number of quicklist nodes */
-            int used = snprintf(nextra, remaining, " ql_nodes:%u", ql->len);
+            int used = snprintf(nextra, remaining, " ql_nodes:%lu", ql->len);
             nextra += used;
             remaining -= used;
             /* Add average quicklist fill factor */
@@ -549,9 +547,15 @@ void debugCommand(client *c) {
         stats = sdscat(stats,buf);
 
         addReplyBulkSds(c,stats);
+    } else if (!strcasecmp(c->argv[1]->ptr,"change-repl-id") && c->argc == 2) {
+        serverLog(LL_WARNING,"Changing replication IDs after receiving DEBUG change-repl-id");
+        changeReplicationId();
+        clearReplicationId2();
+        addReply(c,shared.ok);
     } else {
-        addReplyErrorFormat(c, "Unknown DEBUG subcommand or wrong number of arguments for '%s'",
+        addReplyErrorFormat(c, "Unknown subcommand or wrong number of arguments for '%s'. Try DEBUG HELP",
             (char*)c->argv[1]->ptr);
+        return;
     }
 }
 
@@ -1023,7 +1027,7 @@ void sigsegvHandler(int sig, siginfo_t *info, void *secret) {
         "Redis %s crashed by signal: %d", REDIS_VERSION, sig);
     if (eip != NULL) {
         serverLog(LL_WARNING,
-        "Crashed running the instuction at: %p", eip);
+        "Crashed running the instruction at: %p", eip);
     }
     if (sig == SIGSEGV || sig == SIGBUS) {
         serverLog(LL_WARNING,

src/defrag.c

@@ -289,7 +289,7 @@ int defragKey(redisDb *db, dictEntry *de) {
          /* Dirty code:
           * I can't search in db->expires for that key after i already released
           * the pointer it holds it won't be able to do the string compare */
-        unsigned int hash = dictGetHash(db->dict, de->key);
+        uint64_t hash = dictGetHash(db->dict, de->key);
         replaceSateliteDictKeyPtrAndOrDefragDictEntry(db->expires, keysds, newsds, hash, &defragged);
     }
 

src/dict.c

@@ -66,7 +66,7 @@ static unsigned int dict_force_resize_ratio = 5;
 
 static int _dictExpandIfNeeded(dict *ht);
 static unsigned long _dictNextPower(unsigned long size);
-static int _dictKeyIndex(dict *ht, const void *key, unsigned int hash, dictEntry **existing);
+static long _dictKeyIndex(dict *ht, const void *key, uint64_t hash, dictEntry **existing);
 static int _dictInit(dict *ht, dictType *type, void *privDataPtr);
 
 /* -------------------------- hash functions -------------------------------- */
@@ -202,7 +202,7 @@ int dictRehash(dict *d, int n) {
         de = d->ht[0].table[d->rehashidx];
         /* Move all the keys in this bucket from the old to the new hash HT */
         while(de) {
-            unsigned int h;
+            uint64_t h;
 
             nextde = de->next;
             /* Get the index in the new hash table */
@@ -291,7 +291,7 @@ int dictAdd(dict *d, void *key, void *val)
  */
 dictEntry *dictAddRaw(dict *d, void *key, dictEntry **existing)
 {
-    int index;
+    long index;
     dictEntry *entry;
     dictht *ht;
 
@@ -362,7 +362,7 @@ dictEntry *dictAddOrFind(dict *d, void *key) {
  * dictDelete() and dictUnlink(), please check the top comment
  * of those functions. */
 static dictEntry *dictGenericDelete(dict *d, const void *key, int nofree) {
-    unsigned int h, idx;
+    uint64_t h, idx;
     dictEntry *he, *prevHe;
     int table;
 
@@ -476,7 +476,7 @@ void dictRelease(dict *d)
 dictEntry *dictFind(dict *d, const void *key)
 {
     dictEntry *he;
-    unsigned int h, idx, table;
+    uint64_t h, idx, table;
 
     if (d->ht[0].used + d->ht[1].used == 0) return NULL; /* dict is empty */
     if (dictIsRehashing(d)) _dictRehashStep(d);
@@ -610,7 +610,7 @@ void dictReleaseIterator(dictIterator *iter)
 dictEntry *dictGetRandomKey(dict *d)
 {
     dictEntry *he, *orighe;
-    unsigned int h;
+    unsigned long h;
     int listlen, listele;
 
     if (dictSize(d) == 0) return NULL;
@@ -955,9 +955,9 @@ static unsigned long _dictNextPower(unsigned long size)
  *
  * Note that if we are in the process of rehashing the hash table, the
  * index is always returned in the context of the second (new) hash table. */
-static int _dictKeyIndex(dict *d, const void *key, unsigned int hash, dictEntry **existing)
+static long _dictKeyIndex(dict *d, const void *key, uint64_t hash, dictEntry **existing)
 {
-    unsigned int idx, table;
+    unsigned long idx, table;
     dictEntry *he;
     if (existing) *existing = NULL;
 
@@ -995,7 +995,7 @@ void dictDisableResize(void) {
     dict_can_resize = 0;
 }
 
-unsigned int dictGetHash(dict *d, const void *key) {
+uint64_t dictGetHash(dict *d, const void *key) {
     return dictHashKey(d, key);
 }
 
@@ -1004,9 +1004,9 @@ unsigned int dictGetHash(dict *d, const void *key) {
  * the hash value should be provided using dictGetHash.
  * no string / key comparison is performed.
  * return value is the reference to the dictEntry if found, or NULL if not found. */
-dictEntry **dictFindEntryRefByPtrAndHash(dict *d, const void *oldptr, unsigned int hash) {
+dictEntry **dictFindEntryRefByPtrAndHash(dict *d, const void *oldptr, uint64_t hash) {
     dictEntry *he, **heref;
-    unsigned int idx, table;
+    unsigned long idx, table;
 
     if (d->ht[0].used + d->ht[1].used == 0) return NULL; /* dict is empty */
     for (table = 0; table <= 1; table++) {

src/dict.h

@@ -178,8 +178,8 @@ int dictRehashMilliseconds(dict *d, int ms);
 void dictSetHashFunctionSeed(uint8_t *seed);
 uint8_t *dictGetHashFunctionSeed(void);
 unsigned long dictScan(dict *d, unsigned long v, dictScanFunction *fn, dictScanBucketFunction *bucketfn, void *privdata);
-unsigned int dictGetHash(dict *d, const void *key);
-dictEntry **dictFindEntryRefByPtrAndHash(dict *d, const void *oldptr, unsigned int hash);
+uint64_t dictGetHash(dict *d, const void *key);
+dictEntry **dictFindEntryRefByPtrAndHash(dict *d, const void *oldptr, uint64_t hash);
 
 /* Hash table types */
 extern dictType dictTypeHeapStringCopyKey;

src/evict.c

@@ -60,8 +60,6 @@ struct evictionPoolEntry {
 
 static struct evictionPoolEntry *EvictionPoolLRU;
 
-unsigned long LFUDecrAndReturn(robj *o);
-
 /* ----------------------------------------------------------------------------
  * Implementation of eviction, aging and LRU
  * --------------------------------------------------------------------------*/
@@ -302,8 +300,8 @@ unsigned long LFUGetTimeInMinutes(void) {
     return (server.unixtime/60) & 65535;
 }
 
-/* Given an object last decrement time, compute the minimum number of minutes
- * that elapsed since the last decrement. Handle overflow (ldt greater than
+/* Given an object last access time, compute the minimum number of minutes
+ * that elapsed since the last access. Handle overflow (ldt greater than
  * the current 16 bits minutes time) considering the time as wrapping
  * exactly once. */
 unsigned long LFUTimeElapsed(unsigned long ldt) {
@@ -324,25 +322,22 @@ uint8_t LFULogIncr(uint8_t counter) {
     return counter;
 }
 
-/* If the object decrement time is reached, decrement the LFU counter and
- * update the decrement time field. Return the object frequency counter.
+/* If the object decrement time is reached decrement the LFU counter but
+ * do not update LFU fields of the object, we update the access time
+ * and counter in an explicit way when the object is really accessed.
+ * And we will times halve the counter according to the times of
+ * elapsed time than server.lfu_decay_time.
+ * Return the object frequency counter.
  *
  * This function is used in order to scan the dataset for the best object
  * to fit: as we check for the candidate, we incrementally decrement the
  * counter of the scanned objects if needed. */
-#define LFU_DECR_INTERVAL 1
 unsigned long LFUDecrAndReturn(robj *o) {
     unsigned long ldt = o->lru >> 8;
     unsigned long counter = o->lru & 255;
-    if (LFUTimeElapsed(ldt) >= server.lfu_decay_time && counter) {
-        if (counter > LFU_INIT_VAL*2) {
-            counter /= 2;
-            if (counter < LFU_INIT_VAL*2) counter = LFU_INIT_VAL*2;
-        } else {
-            counter--;
-        }
-        o->lru = (LFUGetTimeInMinutes()<<8) | counter;
-    }
+    unsigned long num_periods = server.lfu_decay_time ? LFUTimeElapsed(ldt) / server.lfu_decay_time : 0;
+    if (num_periods)
+        counter = (num_periods > counter) ? 0 : counter - num_periods;
     return counter;
 }
 

src/expire.c

@@ -103,7 +103,7 @@ void activeExpireCycle(int type) {
 
     int j, iteration = 0;
     int dbs_per_call = CRON_DBS_PER_CALL;
-    long long start = ustime(), timelimit;
+    long long start = ustime(), timelimit, elapsed;
 
     /* When clients are paused the dataset should be static not just from the
      * POV of clients not being able to write, but also from the POV of
@@ -111,7 +111,7 @@ void activeExpireCycle(int type) {
     if (clientsArePaused()) return;
 
     if (type == ACTIVE_EXPIRE_CYCLE_FAST) {
-        /* Don't start a fast cycle if the previous cycle did not exited
+        /* Don't start a fast cycle if the previous cycle did not exit
          * for time limt. Also don't repeat a fast cycle for the same period
          * as the fast cycle total duration itself. */
         if (!timelimit_exit) return;
@@ -140,7 +140,13 @@ void activeExpireCycle(int type) {
     if (type == ACTIVE_EXPIRE_CYCLE_FAST)
         timelimit = ACTIVE_EXPIRE_CYCLE_FAST_DURATION; /* in microseconds. */
 
-    for (j = 0; j < dbs_per_call; j++) {
+    /* Accumulate some global stats as we expire keys, to have some idea
+     * about the number of keys that are already logically expired, but still
+     * existing inside the database. */
+    long total_sampled = 0;
+    long total_expired = 0;
+
+    for (j = 0; j < dbs_per_call && timelimit_exit == 0; j++) {
         int expired;
         redisDb *db = server.db+(current_db % server.dbnum);
 
@@ -155,6 +161,7 @@ void activeExpireCycle(int type) {
             unsigned long num, slots;
             long long now, ttl_sum;
             int ttl_samples;
+            iteration++;
 
             /* If there is nothing to expire try next DB ASAP. */
             if ((num = dictSize(db->expires)) == 0) {
@@ -191,7 +198,9 @@ void activeExpireCycle(int type) {
                     ttl_sum += ttl;
                     ttl_samples++;
                 }
+                total_sampled++;
             }
+            total_expired += expired;
 
             /* Update the average TTL stats for this database. */
             if (ttl_samples) {
@@ -207,18 +216,31 @@ void activeExpireCycle(int type) {
             /* We can't block forever here even if there are many keys to
              * expire. So after a given amount of milliseconds return to the
              * caller waiting for the other active expire cycle. */
-            iteration++;
             if ((iteration & 0xf) == 0) { /* check once every 16 iterations. */
-                long long elapsed = ustime()-start;
-
-                latencyAddSampleIfNeeded("expire-cycle",elapsed/1000);
-                if (elapsed > timelimit) timelimit_exit = 1;
+                elapsed = ustime()-start;
+                if (elapsed > timelimit) {
+                    timelimit_exit = 1;
+                    server.stat_expired_time_cap_reached_count++;
+                    break;
+                }
             }
-            if (timelimit_exit) return;
             /* We don't repeat the cycle if there are less than 25% of keys
              * found expired in the current DB. */
         } while (expired > ACTIVE_EXPIRE_CYCLE_LOOKUPS_PER_LOOP/4);
     }
+
+    elapsed = ustime()-start;
+    latencyAddSampleIfNeeded("expire-cycle",elapsed/1000);
+
+    /* Update our estimate of keys existing but yet to be expired.
+     * Running average with this sample accounting for 5%. */
+    double current_perc;
+    if (total_sampled) {
+        current_perc = (double)total_expired/total_sampled;
+    } else
+        current_perc = 0;
+    server.stat_expired_stale_perc = (current_perc*0.05)+
+                                     (server.stat_expired_stale_perc*0.95);
 }
 
 /*-----------------------------------------------------------------------------

src/hyperloglog.c

@@ -475,9 +475,8 @@ int hllPatLen(unsigned char *ele, size_t elesize, long *regp) {
 
 /* ================== Dense representation implementation  ================== */
 
-/* "Add" the element in the dense hyperloglog data structure.
- * Actually nothing is added, but the max 0 pattern counter of the subset
- * the element belongs to is incremented if needed.
+/* Low level function to set the dense HLL register at 'index' to the
+ * specified value if the current value is smaller than 'count'.
  *
  * 'registers' is expected to have room for HLL_REGISTERS plus an
  * additional byte on the right. This requirement is met by sds strings
@@ -486,12 +485,9 @@ int hllPatLen(unsigned char *ele, size_t elesize, long *regp) {
  * The function always succeed, however if as a result of the operation
  * the approximated cardinality changed, 1 is returned. Otherwise 0
  * is returned. */
-int hllDenseAdd(uint8_t *registers, unsigned char *ele, size_t elesize) {
-    uint8_t oldcount, count;
-    long index;
+int hllDenseSet(uint8_t *registers, long index, uint8_t count) {
+    uint8_t oldcount;
 
-    /* Update the register if this element produced a longer run of zeroes. */
-    count = hllPatLen(ele,elesize,&index);
     HLL_DENSE_GET_REGISTER(oldcount,registers,index);
     if (count > oldcount) {
         HLL_DENSE_SET_REGISTER(registers,index,count);
@@ -501,6 +497,19 @@ int hllDenseAdd(uint8_t *registers, unsigned char *ele, size_t elesize) {
     }
 }
 
+/* "Add" the element in the dense hyperloglog data structure.
+ * Actually nothing is added, but the max 0 pattern counter of the subset
+ * the element belongs to is incremented if needed.
+ *
+ * This is just a wrapper to hllDenseSet(), performing the hashing of the
+ * element in order to retrieve the index and zero-run count. */
+int hllDenseAdd(uint8_t *registers, unsigned char *ele, size_t elesize) {
+    long index;
+    uint8_t count = hllPatLen(ele,elesize,&index);
+    /* Update the register if this element produced a longer run of zeroes. */
+    return hllDenseSet(registers,index,count);
+}
+
 /* Compute SUM(2^-reg) in the dense representation.
  * PE is an array with a pre-computer table of values 2^-reg indexed by reg.
  * As a side effect the integer pointed by 'ezp' is set to the number
@@ -623,9 +632,8 @@ int hllSparseToDense(robj *o) {
     return C_OK;
 }
 
-/* "Add" the element in the sparse hyperloglog data structure.
- * Actually nothing is added, but the max 0 pattern counter of the subset
- * the element belongs to is incremented if needed.
+/* Low level function to set the sparse HLL register at 'index' to the
+ * specified value if the current value is smaller than 'count'.
  *
  * The object 'o' is the String object holding the HLL. The function requires
  * a reference to the object in order to be able to enlarge the string if
@@ -639,15 +647,12 @@ int hllSparseToDense(robj *o) {
  * sparse to dense: this happens when a register requires to be set to a value
  * not representable with the sparse representation, or when the resulting
  * size would be greater than server.hll_sparse_max_bytes. */
-int hllSparseAdd(robj *o, unsigned char *ele, size_t elesize) {
+int hllSparseSet(robj *o, long index, uint8_t count) {
     struct hllhdr *hdr;
-    uint8_t oldcount, count, *sparse, *end, *p, *prev, *next;
-    long index, first, span;
+    uint8_t oldcount, *sparse, *end, *p, *prev, *next;
+    long first, span;
     long is_zero = 0, is_xzero = 0, is_val = 0, runlen = 0;
 
-    /* Update the register if this element produced a longer run of zeroes. */
-    count = hllPatLen(ele,elesize,&index);
-
     /* If the count is too big to be representable by the sparse representation
      * switch to dense representation. */
     if (count > HLL_SPARSE_VAL_MAX_VALUE) goto promote;
@@ -880,11 +885,24 @@ promote: /* Promote to dense representation. */
      * Note that this in turn means that PFADD will make sure the command
      * is propagated to slaves / AOF, so if there is a sparse -> dense
      * convertion, it will be performed in all the slaves as well. */
-    int dense_retval = hllDenseAdd(hdr->registers, ele, elesize);
+    int dense_retval = hllDenseSet(hdr->registers,index,count);
     serverAssert(dense_retval == 1);
     return dense_retval;
 }
 
+/* "Add" the element in the sparse hyperloglog data structure.
+ * Actually nothing is added, but the max 0 pattern counter of the subset
+ * the element belongs to is incremented if needed.
+ *
+ * This function is actually a wrapper for hllSparseSet(), it only performs
+ * the hashshing of the elmenet to obtain the index and zeros run length. */
+int hllSparseAdd(robj *o, unsigned char *ele, size_t elesize) {
+    long index;
+    uint8_t count = hllPatLen(ele,elesize,&index);
+    /* Update the register if this element produced a longer run of zeroes. */
+    return hllSparseSet(o,index,count);
+}
+
 /* Compute SUM(2^-reg) in the sparse representation.
  * PE is an array with a pre-computer table of values 2^-reg indexed by reg.
  * As a side effect the integer pointed by 'ezp' is set to the number
@@ -1280,9 +1298,10 @@ void pfmergeCommand(client *c) {
     uint8_t max[HLL_REGISTERS];
     struct hllhdr *hdr;
     int j;
+    int use_dense = 0; /* Use dense representation as target? */
 
     /* Compute an HLL with M[i] = MAX(M[i]_j).
-     * We we the maximum into the max array of registers. We'll write
+     * We store the maximum into the max array of registers. We'll write
      * it to the target variable later. */
     memset(max,0,sizeof(max));
     for (j = 1; j < c->argc; j++) {
@@ -1291,6 +1310,11 @@ void pfmergeCommand(client *c) {
         if (o == NULL) continue; /* Assume empty HLL for non existing var. */
         if (isHLLObjectOrReply(c,o) != C_OK) return;
 
+        /* If at least one involved HLL is dense, use the dense representation
+         * as target ASAP to save time and avoid the conversion step. */
+        hdr = o->ptr;
+        if (hdr->encoding == HLL_DENSE) use_dense = 1;
+
         /* Merge with this HLL with our 'max' HHL by setting max[i]
          * to MAX(max[i],hll[i]). */
         if (hllMerge(max,o) == C_ERR) {
@@ -1314,22 +1338,29 @@ void pfmergeCommand(client *c) {
         o = dbUnshareStringValue(c->db,c->argv[1],o);
     }
 
-    /* Only support dense objects as destination. */
-    if (hllSparseToDense(o) == C_ERR) {
+    /* Convert the destination object to dense representation if at least
+     * one of the inputs was dense. */
+    if (use_dense && hllSparseToDense(o) == C_ERR) {
         addReplySds(c,sdsnew(invalid_hll_err));
         return;
     }
 
     /* Write the resulting HLL to the destination HLL registers and
      * invalidate the cached value. */
-    hdr = o->ptr;
     for (j = 0; j < HLL_REGISTERS; j++) {
-        HLL_DENSE_SET_REGISTER(hdr->registers,j,max[j]);
+        if (max[j] == 0) continue;
+        hdr = o->ptr;
+        switch(hdr->encoding) {
+        case HLL_DENSE: hllDenseSet(hdr->registers,j,max[j]); break;
+        case HLL_SPARSE: hllSparseSet(o,j,max[j]); break;
+        }
     }
+    hdr = o->ptr; /* o->ptr may be different now, as a side effect of
+                     last hllSparseSet() call. */
     HLL_INVALIDATE_CACHE(hdr);
 
     signalModifiedKey(c->db,c->argv[1]);
-    /* We generate an PFADD event for PFMERGE for semantical simplicity
+    /* We generate a PFADD event for PFMERGE for semantical simplicity
      * since in theory this is a mass-add of elements. */
     notifyKeyspaceEvent(NOTIFY_STRING,"pfadd",c->argv[1],c->db->id);
     server.dirty++;

src/latency.c

@@ -109,6 +109,8 @@ void latencyAddSample(char *event, mstime_t latency) {
         dictAdd(server.latency_events,zstrdup(event),ts);
     }
 
+    if (latency > ts->max) ts->max = latency;
+
     /* If the previous sample is in the same second, we update our old sample
      * if this latency is > of the old one, or just return. */
     prev = (ts->idx + LATENCY_TS_LEN - 1) % LATENCY_TS_LEN;
@@ -120,7 +122,6 @@ void latencyAddSample(char *event, mstime_t latency) {
 
     ts->samples[ts->idx].time = time(NULL);
     ts->samples[ts->idx].latency = latency;
-    if (latency > ts->max) ts->max = latency;
 
     ts->idx++;
     if (ts->idx == LATENCY_TS_LEN) ts->idx = 0;

src/lazyfree.c

@@ -64,9 +64,15 @@ int dbAsyncDelete(redisDb *db, robj *key) {
         robj *val = dictGetVal(de);
         size_t free_effort = lazyfreeGetFreeEffort(val);
 
-        /* If releasing the object is too much work, let's put it into the
-         * lazy free list. */
-        if (free_effort > LAZYFREE_THRESHOLD) {
+        /* If releasing the object is too much work, do it in the background
+         * by adding the object to the lazy free list.
+         * Note that if the object is shared, to reclaim it now it is not
+         * possible. This rarely happens, however sometimes the implementation
+         * of parts of the Redis core may call incrRefCount() to protect
+         * objects, and then call dbDelete(). In this case we'll fall
+         * through and reach the dictFreeUnlinkedEntry() call, that will be
+         * equivalent to just calling decrRefCount(). */
+        if (free_effort > LAZYFREE_THRESHOLD && val->refcount == 1) {
             atomicIncr(lazyfree_objects,1);
             bioCreateBackgroundJob(BIO_LAZY_FREE,val,NULL,NULL);
             dictSetVal(db->dict,de,NULL);

src/listpack.c

@@ -0,0 +1,783 @@
+/* Listpack -- A lists of strings serialization format
+ *
+ * This file implements the specification you can find at:
+ *
+ *  https://github.com/antirez/listpack
+ *
+ * Copyright (c) 2017, Salvatore Sanfilippo <antirez at gmail dot com>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ *   * Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *   * Redistributions in binary form must reproduce the above copyright
+ *     notice, this list of conditions and the following disclaimer in the
+ *     documentation and/or other materials provided with the distribution.
+ *   * Neither the name of Redis nor the names of its contributors may be used
+ *     to endorse or promote products derived from this software without
+ *     specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdint.h>
+#include <limits.h>
+#include <sys/types.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+
+#include "listpack.h"
+#include "listpack_malloc.h"
+
+#define LP_HDR_SIZE 6       /* 32 bit total len + 16 bit number of elements. */
+#define LP_HDR_NUMELE_UNKNOWN UINT16_MAX
+#define LP_MAX_INT_ENCODING_LEN 9
+#define LP_MAX_BACKLEN_SIZE 5
+#define LP_MAX_ENTRY_BACKLEN 34359738367ULL
+#define LP_ENCODING_INT 0
+#define LP_ENCODING_STRING 1
+
+#define LP_ENCODING_7BIT_UINT 0
+#define LP_ENCODING_7BIT_UINT_MASK 0x80
+#define LP_ENCODING_IS_7BIT_UINT(byte) (((byte)&LP_ENCODING_7BIT_UINT_MASK)==LP_ENCODING_7BIT_UINT)
+
+#define LP_ENCODING_6BIT_STR 0x80
+#define LP_ENCODING_6BIT_STR_MASK 0xC0
+#define LP_ENCODING_IS_6BIT_STR(byte) (((byte)&LP_ENCODING_6BIT_STR_MASK)==LP_ENCODING_6BIT_STR)
+
+#define LP_ENCODING_13BIT_INT 0xC0
+#define LP_ENCODING_13BIT_INT_MASK 0xE0
+#define LP_ENCODING_IS_13BIT_INT(byte) (((byte)&LP_ENCODING_13BIT_INT_MASK)==LP_ENCODING_13BIT_INT)
+
+#define LP_ENCODING_12BIT_STR 0xE0
+#define LP_ENCODING_12BIT_STR_MASK 0xF0
+#define LP_ENCODING_IS_12BIT_STR(byte) (((byte)&LP_ENCODING_12BIT_STR_MASK)==LP_ENCODING_12BIT_STR)
+
+#define LP_ENCODING_16BIT_INT 0xF1
+#define LP_ENCODING_16BIT_INT_MASK 0xFF
+#define LP_ENCODING_IS_16BIT_INT(byte) (((byte)&LP_ENCODING_16BIT_INT_MASK)==LP_ENCODING_16BIT_INT)
+
+#define LP_ENCODING_24BIT_INT 0xF2
+#define LP_ENCODING_24BIT_INT_MASK 0xFF
+#define LP_ENCODING_IS_24BIT_INT(byte) (((byte)&LP_ENCODING_24BIT_INT_MASK)==LP_ENCODING_24BIT_INT)
+
+#define LP_ENCODING_32BIT_INT 0xF3
+#define LP_ENCODING_32BIT_INT_MASK 0xFF
+#define LP_ENCODING_IS_32BIT_INT(byte) (((byte)&LP_ENCODING_32BIT_INT_MASK)==LP_ENCODING_32BIT_INT)
+
+#define LP_ENCODING_64BIT_INT 0xF4
+#define LP_ENCODING_64BIT_INT_MASK 0xFF
+#define LP_ENCODING_IS_64BIT_INT(byte) (((byte)&LP_ENCODING_64BIT_INT_MASK)==LP_ENCODING_64BIT_INT)
+
+#define LP_ENCODING_32BIT_STR 0xF0
+#define LP_ENCODING_32BIT_STR_MASK 0xFF
+#define LP_ENCODING_IS_32BIT_STR(byte) (((byte)&LP_ENCODING_32BIT_STR_MASK)==LP_ENCODING_32BIT_STR)
+
+#define LP_EOF 0xFF
+
+#define LP_ENCODING_6BIT_STR_LEN(p) ((p)[0] & 0x3F)
+#define LP_ENCODING_12BIT_STR_LEN(p) ((((p)[0] & 0xF) << 8) | (p)[1])
+#define LP_ENCODING_32BIT_STR_LEN(p) (((uint32_t)(p)[1]<<0) | \
+                                      ((uint32_t)(p)[2]<<8) | \
+                                      ((uint32_t)(p)[3]<<16) | \
+                                      ((uint32_t)(p)[4]<<24))
+
+#define lpGetTotalBytes(p)           (((uint32_t)(p)[0]<<0) | \
+                                      ((uint32_t)(p)[1]<<8) | \
+                                      ((uint32_t)(p)[2]<<16) | \
+                                      ((uint32_t)(p)[3]<<24))
+
+#define lpGetNumElements(p)          (((uint32_t)(p)[4]<<0) | \
+                                      ((uint32_t)(p)[5]<<8))
+#define lpSetTotalBytes(p,v) do { \
+    (p)[0] = (v)&0xff; \
+    (p)[1] = ((v)>>8)&0xff; \
+    (p)[2] = ((v)>>16)&0xff; \
+    (p)[3] = ((v)>>24)&0xff; \
+} while(0)
+
+#define lpSetNumElements(p,v) do { \
+    (p)[4] = (v)&0xff; \
+    (p)[5] = ((v)>>8)&0xff; \
+} while(0)
+
+/* Convert a string into a signed 64 bit integer.
+ * The function returns 1 if the string could be parsed into a (non-overflowing)
+ * signed 64 bit int, 0 otherwise. The 'value' will be set to the parsed value
+ * when the function returns success.
+ *
+ * Note that this function demands that the string strictly represents
+ * a int64 value: no spaces or other characters before or after the string
+ * representing the number are accepted, nor zeroes at the start if not
+ * for the string "0" representing the zero number.
+ *
+ * Because of its strictness, it is safe to use this function to check if
+ * you can convert a string into a long long, and obtain back the string
+ * from the number without any loss in the string representation. *
+ *
+ * -----------------------------------------------------------------------------
+ *
+ * Credits: this function was adapted from the Redis source code, file
+ * "utils.c", function string2ll(), and is copyright:
+ *
+ * Copyright(C) 2011, Pieter Noordhuis
+ * Copyright(C) 2011, Salvatore Sanfilippo
+ *
+ * The function is released under the BSD 3-clause license.
+ */
+int lpStringToInt64(const char *s, unsigned long slen, int64_t *value) {
+    const char *p = s;
+    unsigned long plen = 0;
+    int negative = 0;
+    uint64_t v;
+
+    if (plen == slen)
+        return 0;
+
+    /* Special case: first and only digit is 0. */
+    if (slen == 1 && p[0] == '0') {
+        if (value != NULL) *value = 0;
+        return 1;
+    }
+
+    if (p[0] == '-') {
+        negative = 1;
+        p++; plen++;
+
+        /* Abort on only a negative sign. */
+        if (plen == slen)
+            return 0;
+    }
+
+    /* First digit should be 1-9, otherwise the string should just be 0. */
+    if (p[0] >= '1' && p[0] <= '9') {
+        v = p[0]-'0';
+        p++; plen++;
+    } else if (p[0] == '0' && slen == 1) {
+        *value = 0;
+        return 1;
+    } else {
+        return 0;
+    }
+
+    while (plen < slen && p[0] >= '0' && p[0] <= '9') {
+        if (v > (UINT64_MAX / 10)) /* Overflow. */
+            return 0;
+        v *= 10;
+
+        if (v > (UINT64_MAX - (p[0]-'0'))) /* Overflow. */
+            return 0;
+        v += p[0]-'0';
+
+        p++; plen++;
+    }
+
+    /* Return if not all bytes were used. */
+    if (plen < slen)
+        return 0;
+
+    if (negative) {
+        if (v > ((uint64_t)(-(INT64_MIN+1))+1)) /* Overflow. */
+            return 0;
+        if (value != NULL) *value = -v;
+    } else {
+        if (v > INT64_MAX) /* Overflow. */
+            return 0;
+        if (value != NULL) *value = v;
+    }
+    return 1;
+}
+
+/* Create a new, empty listpack.
+ * On success the new listpack is returned, otherwise an error is returned. */
+unsigned char *lpNew(void) {
+    unsigned char *lp = lp_malloc(LP_HDR_SIZE+1);
+    if (lp == NULL) return NULL;
+    lpSetTotalBytes(lp,LP_HDR_SIZE+1);
+    lpSetNumElements(lp,0);
+    lp[LP_HDR_SIZE] = LP_EOF;
+    return lp;
+}
+
+/* Free the specified listpack. */
+void lpFree(unsigned char *lp) {
+    lp_free(lp);
+}
+
+/* Given an element 'ele' of size 'size', determine if the element can be
+ * represented inside the listpack encoded as integer, and returns
+ * LP_ENCODING_INT if so. Otherwise returns LP_ENCODING_STR if no integer
+ * encoding is possible.
+ *
+ * If the LP_ENCODING_INT is returned, the function stores the integer encoded
+ * representation of the element in the 'intenc' buffer.
+ *
+ * Regardless of the returned encoding, 'enclen' is populated by reference to
+ * the number of bytes that the string or integer encoded element will require
+ * in order to be represented. */
+int lpEncodeGetType(unsigned char *ele, uint32_t size, unsigned char *intenc, uint64_t *enclen) {
+    int64_t v;
+    if (lpStringToInt64((const char*)ele, size, &v)) {
+        if (v >= 0 && v <= 127) {
+            /* Single byte 0-127 integer. */
+            intenc[0] = v;
+            *enclen = 1;
+        } else if (v >= -4096 && v <= 4095) {
+            /* 13 bit integer. */
+            if (v < 0) v = ((int64_t)1<<13)+v;
+            intenc[0] = (v>>8)|LP_ENCODING_13BIT_INT;
+            intenc[1] = v&0xff;
+            *enclen = 2;
+        } else if (v >= -32768 && v <= 32767) {
+            /* 16 bit integer. */
+            if (v < 0) v = ((int64_t)1<<16)+v;
+            intenc[0] = LP_ENCODING_16BIT_INT;
+            intenc[1] = v&0xff;
+            intenc[2] = v>>8;
+            *enclen = 3;
+        } else if (v >= -8388608 && v <= 8388607) {
+            /* 24 bit integer. */
+            if (v < 0) v = ((int64_t)1<<24)+v;
+            intenc[0] = LP_ENCODING_24BIT_INT;
+            intenc[1] = v&0xff;
+            intenc[2] = (v>>8)&0xff;
+            intenc[3] = v>>16;
+            *enclen = 4;
+        } else if (v >= -2147483648 && v <= 2147483647) {
+            /* 32 bit integer. */
+            if (v < 0) v = ((int64_t)1<<32)+v;
+            intenc[0] = LP_ENCODING_32BIT_INT;
+            intenc[1] = v&0xff;
+            intenc[2] = (v>>8)&0xff;
+            intenc[3] = (v>>16)&0xff;
+            intenc[4] = v>>24;
+            *enclen = 5;
+        } else {
+            /* 64 bit integer. */
+            uint64_t uv = v;
+            intenc[0] = LP_ENCODING_64BIT_INT;
+            intenc[1] = uv&0xff;
+            intenc[2] = (uv>>8)&0xff;
+            intenc[3] = (uv>>16)&0xff;
+            intenc[4] = (uv>>24)&0xff;
+            intenc[5] = (uv>>32)&0xff;
+            intenc[6] = (uv>>40)&0xff;
+            intenc[7] = (uv>>48)&0xff;
+            intenc[8] = uv>>56;
+            *enclen = 9;
+        }
+        return LP_ENCODING_INT;
+    } else {
+        if (size < 64) *enclen = 1+size;
+        else if (size < 4096) *enclen = 2+size;
+        else *enclen = 5+size;
+        return LP_ENCODING_STRING;
+    }
+}
+
+/* Store a reverse-encoded variable length field, representing the length
+ * of the previous element of size 'l', in the target buffer 'buf'.
+ * The function returns the number of bytes used to encode it, from
+ * 1 to 5. If 'buf' is NULL the funciton just returns the number of bytes
+ * needed in order to encode the backlen. */
+unsigned long lpEncodeBacklen(unsigned char *buf, uint64_t l) {
+    if (l <= 127) {
+        if (buf) buf[0] = l;
+        return 1;
+    } else if (l < 16383) {
+        if (buf) {
+            buf[0] = l>>7;
+            buf[1] = (l&127)|128;
+        }
+        return 2;
+    } else if (l < 2097151) {
+        if (buf) {
+            buf[0] = l>>14;
+            buf[1] = ((l>>7)&127)|128;
+            buf[2] = (l&127)|128;
+        }
+        return 3;
+    } else if (l < 268435455) {
+        if (buf) {
+            buf[0] = l>>21;
+            buf[1] = ((l>>14)&127)|128;
+            buf[2] = ((l>>7)&127)|128;
+            buf[3] = (l&127)|128;
+        }
+        return 4;
+    } else {
+        if (buf) {
+            buf[0] = l>>28;
+            buf[1] = ((l>>21)&127)|128;
+            buf[2] = ((l>>14)&127)|128;
+            buf[3] = ((l>>7)&127)|128;
+            buf[4] = (l&127)|128;
+        }
+        return 5;
+    }
+}
+
+/* Decode the backlen and returns it. If the encoding looks invalid (more than
+ * 5 bytes are used), UINT64_MAX is returned to report the problem. */
+uint64_t lpDecodeBacklen(unsigned char *p) {
+    uint64_t val = 0;
+    uint64_t shift = 0;
+    do {
+        val |= (uint64_t)(p[0] & 127) << shift;
+        if (!(p[0] & 128)) break;
+        shift += 7;
+        p--;
+        if (shift > 28) return UINT64_MAX;
+    } while(1);
+    return val;
+}
+
+/* Encode the string element pointed by 's' of size 'len' in the target
+ * buffer 's'. The function should be called with 'buf' having always enough
+ * space for encoding the string. This is done by calling lpEncodeGetType()
+ * before calling this function. */
+void lpEncodeString(unsigned char *buf, unsigned char *s, uint32_t len) {
+    if (len < 64) {
+        buf[0] = len | LP_ENCODING_6BIT_STR;
+        memcpy(buf+1,s,len);
+    } else if (len < 4096) {
+        buf[0] = (len >> 8) | LP_ENCODING_12BIT_STR;
+        buf[1] = len & 0xff;
+        memcpy(buf+2,s,len);
+    } else {
+        buf[0] = LP_ENCODING_32BIT_STR;
+        buf[1] = len & 0xff;
+        buf[2] = (len >> 8) & 0xff;
+        buf[3] = (len >> 16) & 0xff;
+        buf[4] = (len >> 24) & 0xff;
+        memcpy(buf+5,s,len);
+    }
+}
+
+/* Return the encoded length of the listpack element pointed by 'p'. If the
+ * element encoding is wrong then 0 is returned. */
+uint32_t lpCurrentEncodedSize(unsigned char *p) {
+    if (LP_ENCODING_IS_7BIT_UINT(p[0])) return 1;
+    if (LP_ENCODING_IS_6BIT_STR(p[0])) return 1+LP_ENCODING_6BIT_STR_LEN(p);
+    if (LP_ENCODING_IS_13BIT_INT(p[0])) return 2;
+    if (LP_ENCODING_IS_16BIT_INT(p[0])) return 3;
+    if (LP_ENCODING_IS_24BIT_INT(p[0])) return 4;
+    if (LP_ENCODING_IS_32BIT_INT(p[0])) return 5;
+    if (LP_ENCODING_IS_64BIT_INT(p[0])) return 9;
+    if (LP_ENCODING_IS_12BIT_STR(p[0])) return 2+LP_ENCODING_12BIT_STR_LEN(p);
+    if (LP_ENCODING_IS_32BIT_STR(p[0])) return 5+LP_ENCODING_32BIT_STR_LEN(p);
+    if (p[0] == LP_EOF) return 1;
+    return 0;
+}
+
+/* Skip the current entry returning the next. It is invalid to call this
+ * function if the current element is the EOF element at the end of the
+ * listpack, however, while this function is used to implement lpNext(),
+ * it does not return NULL when the EOF element is encountered. */
+unsigned char *lpSkip(unsigned char *p) {
+    unsigned long entrylen = lpCurrentEncodedSize(p);
+    entrylen += lpEncodeBacklen(NULL,entrylen);
+    p += entrylen;
+    return p;
+}
+
+/* If 'p' points to an element of the listpack, calling lpNext() will return
+ * the pointer to the next element (the one on the right), or NULL if 'p'
+ * already pointed to the last element of the listpack. */
+unsigned char *lpNext(unsigned char *lp, unsigned char *p) {
+    ((void) lp); /* lp is not used for now. However lpPrev() uses it. */
+    p = lpSkip(p);
+    if (p[0] == LP_EOF) return NULL;
+    return p;
+}
+
+/* If 'p' points to an element of the listpack, calling lpPrev() will return
+ * the pointer to the preivous element (the one on the left), or NULL if 'p'
+ * already pointed to the first element of the listpack. */
+unsigned char *lpPrev(unsigned char *lp, unsigned char *p) {
+    if (p-lp == LP_HDR_SIZE) return NULL;
+    p--; /* Seek the first backlen byte of the last element. */
+    uint64_t prevlen = lpDecodeBacklen(p);
+    prevlen += lpEncodeBacklen(NULL,prevlen);
+    return p-prevlen+1; /* Seek the first byte of the previous entry. */
+}
+
+/* Return a pointer to the first element of the listpack, or NULL if the
+ * listpack has no elements. */
+unsigned char *lpFirst(unsigned char *lp) {
+    lp += LP_HDR_SIZE; /* Skip the header. */
+    if (lp[0] == LP_EOF) return NULL;
+    return lp;
+}
+
+/* Return a pointer to the last element of the listpack, or NULL if the
+ * listpack has no elements. */
+unsigned char *lpLast(unsigned char *lp) {
+    unsigned char *p = lp+lpGetTotalBytes(lp)-1; /* Seek EOF element. */
+    return lpPrev(lp,p); /* Will return NULL if EOF is the only element. */
+}
+
+/* Return the number of elements inside the listpack. This function attempts
+ * to use the cached value when within range, otherwise a full scan is
+ * needed. As a side effect of calling this function, the listpack header
+ * could be modified, because if the count is found to be already within
+ * the 'numele' header field range, the new value is set. */
+uint32_t lpLength(unsigned char *lp) {
+    uint32_t numele = lpGetNumElements(lp);
+    if (numele != LP_HDR_NUMELE_UNKNOWN) return numele;
+
+    /* Too many elements inside the listpack. We need to scan in order
+     * to get the total number. */
+    uint32_t count = 0;
+    unsigned char *p = lpFirst(lp);
+    while(p) {
+        count++;
+        p = lpNext(lp,p);
+    }
+
+    /* If the count is again within range of the header numele field,
+     * set it. */
+    if (count < LP_HDR_NUMELE_UNKNOWN) lpSetNumElements(lp,count);
+    return count;
+}
+
+/* Return the listpack element pointed by 'p'.
+ *
+ * The function changes behavior depending on the passed 'intbuf' value.
+ * Specifically, if 'intbuf' is NULL:
+ *
+ * If the element is internally encoded as an integer, the function returns
+ * NULL and populates the integer value by reference in 'count'. Otherwise if
+ * the element is encoded as a string a pointer to the string (pointing inside
+ * the listpack itself) is returned, and 'count' is set to the length of the
+ * string.
+ *
+ * If instead 'intbuf' points to a buffer passed by the caller, that must be
+ * at least LP_INTBUF_SIZE bytes, the function always returns the element as
+ * it was a string (returning the pointer to the string and setting the
+ * 'count' argument to the string length by reference). However if the element
+ * is encoded as an integer, the 'intbuf' buffer is used in order to store
+ * the string representation.
+ *
+ * The user should use one or the other form depending on what the value will
+ * be used for. If there is immediate usage for an integer value returned
+ * by the function, than to pass a buffer (and convert it back to a number)
+ * is of course useless.
+ *
+ * If the function is called against a badly encoded ziplist, so that there
+ * is no valid way to parse it, the function returns like if there was an
+ * integer encoded with value 12345678900000000 + <unrecognized byte>, this may
+ * be an hint to understand that something is wrong. To crash in this case is
+ * not sensible because of the different requirements of the application using
+ * this lib.
+ *
+ * Similarly, there is no error returned since the listpack normally can be
+ * assumed to be valid, so that would be a very high API cost. However a function
+ * in order to check the integrity of the listpack at load time is provided,
+ * check lpIsValid(). */
+unsigned char *lpGet(unsigned char *p, int64_t *count, unsigned char *intbuf) {
+    int64_t val;
+    uint64_t uval, negstart, negmax;
+
+    if (LP_ENCODING_IS_7BIT_UINT(p[0])) {
+        negstart = UINT64_MAX; /* 7 bit ints are always positive. */
+        negmax = 0;
+        uval = p[0] & 0x7f;
+    } else if (LP_ENCODING_IS_6BIT_STR(p[0])) {
+        *count = LP_ENCODING_6BIT_STR_LEN(p);
+        return p+1;
+    } else if (LP_ENCODING_IS_13BIT_INT(p[0])) {
+        uval = ((p[0]&0x1f)<<8) | p[1];
+        negstart = (uint64_t)1<<12;
+        negmax = 8191;
+    } else if (LP_ENCODING_IS_16BIT_INT(p[0])) {
+        uval = (uint64_t)p[1] |
+               (uint64_t)p[2]<<8;
+        negstart = (uint64_t)1<<15;
+        negmax = UINT16_MAX;
+    } else if (LP_ENCODING_IS_24BIT_INT(p[0])) {
+        uval = (uint64_t)p[1] |
+               (uint64_t)p[2]<<8 |
+               (uint64_t)p[3]<<16;
+        negstart = (uint64_t)1<<23;
+        negmax = UINT32_MAX>>8;
+    } else if (LP_ENCODING_IS_32BIT_INT(p[0])) {
+        uval = (uint64_t)p[1] |
+               (uint64_t)p[2]<<8 |
+               (uint64_t)p[3]<<16 |
+               (uint64_t)p[4]<<24;
+        negstart = (uint64_t)1<<31;
+        negmax = UINT32_MAX;
+    } else if (LP_ENCODING_IS_64BIT_INT(p[0])) {
+        uval = (uint64_t)p[1] |
+               (uint64_t)p[2]<<8 |
+               (uint64_t)p[3]<<16 |
+               (uint64_t)p[4]<<24 |
+               (uint64_t)p[5]<<32 |
+               (uint64_t)p[6]<<40 |
+               (uint64_t)p[7]<<48 |
+               (uint64_t)p[8]<<56;
+        negstart = (uint64_t)1<<63;
+        negmax = UINT64_MAX;
+    } else if (LP_ENCODING_IS_12BIT_STR(p[0])) {
+        *count = LP_ENCODING_12BIT_STR_LEN(p);
+        return p+2;
+    } else if (LP_ENCODING_IS_32BIT_STR(p[0])) {
+        *count = LP_ENCODING_32BIT_STR_LEN(p);
+        return p+5;
+    } else {
+        uval = 12345678900000000ULL + p[0];
+        negstart = UINT64_MAX;
+        negmax = 0;
+    }
+
+    /* We reach this code path only for integer encodings.
+     * Convert the unsigned value to the signed one using two's complement
+     * rule. */
+    if (uval >= negstart) {
+        /* This three steps conversion should avoid undefined behaviors
+         * in the unsigned -> signed conversion. */
+        uval = negmax-uval;
+        val = uval;
+        val = -val-1;
+    } else {
+        val = uval;
+    }
+
+    /* Return the string representation of the integer or the value itself
+     * depending on intbuf being NULL or not. */
+    if (intbuf) {
+        *count = snprintf((char*)intbuf,LP_INTBUF_SIZE,"%lld",val);
+        return intbuf;
+    } else {
+        *count = val;
+        return NULL;
+    }
+}
+
+/* Insert, delete or replace the specified element 'ele' of lenght 'len' at
+ * the specified position 'p', with 'p' being a listpack element pointer
+ * obtained with lpFirst(), lpLast(), lpIndex(), lpNext(), lpPrev() or
+ * lpSeek().
+ *
+ * The element is inserted before, after, or replaces the element pointed
+ * by 'p' depending on the 'where' argument, that can be LP_BEFORE, LP_AFTER
+ * or LP_REPLACE.
+ *
+ * If 'ele' is set to NULL, the function removes the element pointed by 'p'
+ * instead of inserting one.
+ *
+ * Returns NULL on out of memory or when the listpack total length would exceed
+ * the max allowed size of 2^32-1, otherwise the new pointer to the listpack
+ * holding the new element is returned (and the old pointer passed is no longer
+ * considered valid)
+ *
+ * If 'newp' is not NULL, at the end of a successful call '*newp' will be set
+ * to the address of the element just added, so that it will be possible to
+ * continue an interation with lpNext() and lpPrev().
+ *
+ * For deletion operations ('ele' set to NULL) 'newp' is set to the next
+ * element, on the right of the deleted one, or to NULL if the deleted element
+ * was the last one. */
+unsigned char *lpInsert(unsigned char *lp, unsigned char *ele, uint32_t size, unsigned char *p, int where, unsigned char **newp) {
+    unsigned char intenc[LP_MAX_INT_ENCODING_LEN];
+    unsigned char backlen[LP_MAX_BACKLEN_SIZE];
+
+    uint64_t enclen; /* The length of the encoded element. */
+
+    /* An element pointer set to NULL means deletion, which is conceptually
+     * replacing the element with a zero-length element. So whatever we
+     * get passed as 'where', set it to LP_REPLACE. */
+    if (ele == NULL) where = LP_REPLACE;
+
+    /* If we need to insert after the current element, we just jump to the
+     * next element (that could be the EOF one) and handle the case of
+     * inserting before. So the function will actually deal with just two
+     * cases: LP_BEFORE and LP_REPLACE. */
+    if (where == LP_AFTER) {
+        p = lpSkip(p);
+        where = LP_BEFORE;
+    }
+
+    /* Store the offset of the element 'p', so that we can obtain its
+     * address again after a reallocation. */
+    unsigned long poff = p-lp;
+
+    /* Calling lpEncodeGetType() results into the encoded version of the
+     * element to be stored into 'intenc' in case it is representable as
+     * an integer: in that case, the function returns LP_ENCODING_INT.
+     * Otherwise if LP_ENCODING_STR is returned, we'll have to call
+     * lpEncodeString() to actually write the encoded string on place later.
+     *
+     * Whatever the returned encoding is, 'enclen' is populated with the
+     * length of the encoded element. */
+    int enctype;
+    if (ele) {
+        enctype = lpEncodeGetType(ele,size,intenc,&enclen);
+    } else {
+        enctype = -1;
+        enclen = 0;
+    }
+
+    /* We need to also encode the backward-parsable length of the element
+     * and append it to the end: this allows to traverse the listpack from
+     * the end to the start. */
+    unsigned long backlen_size = ele ? lpEncodeBacklen(backlen,enclen) : 0;
+    uint64_t old_listpack_bytes = lpGetTotalBytes(lp);
+    uint32_t replaced_len  = 0;
+    if (where == LP_REPLACE) {
+        replaced_len = lpCurrentEncodedSize(p);
+        replaced_len += lpEncodeBacklen(NULL,replaced_len);
+    }
+
+    uint64_t new_listpack_bytes = old_listpack_bytes + enclen + backlen_size
+                                  - replaced_len;
+    if (new_listpack_bytes > UINT32_MAX) return NULL;
+
+    /* We now need to reallocate in order to make space or shrink the
+     * allocation (in case 'when' value is LP_REPLACE and the new element is
+     * smaller). However we do that before memmoving the memory to
+     * make room for the new element if the final allocation will get
+     * larger, or we do it after if the final allocation will get smaller. */
+
+    unsigned char *dst = lp + poff; /* May be updated after reallocation. */
+
+    /* Realloc before: we need more room. */
+    if (new_listpack_bytes > old_listpack_bytes) {
+        if ((lp = lp_realloc(lp,new_listpack_bytes)) == NULL) return NULL;
+        dst = lp + poff;
+    }
+
+    /* Setup the listpack relocating the elements to make the exact room
+     * we need to store the new one. */
+    if (where == LP_BEFORE) {
+        memmove(dst+enclen+backlen_size,dst,old_listpack_bytes-poff);
+    } else { /* LP_REPLACE. */
+        long lendiff = (enclen+backlen_size)-replaced_len;
+        memmove(dst+replaced_len+lendiff,
+                dst+replaced_len,
+                old_listpack_bytes-poff-replaced_len);
+    }
+
+    /* Realloc after: we need to free space. */
+    if (new_listpack_bytes < old_listpack_bytes) {
+        if ((lp = lp_realloc(lp,new_listpack_bytes)) == NULL) return NULL;
+        dst = lp + poff;
+    }
+
+    /* Store the entry. */
+    if (newp) {
+        *newp = dst;
+        /* In case of deletion, set 'newp' to NULL if the next element is
+         * the EOF element. */
+        if (!ele && dst[0] == LP_EOF) *newp = NULL;
+    }
+    if (ele) {
+        if (enctype == LP_ENCODING_INT) {
+            memcpy(dst,intenc,enclen);
+        } else {
+            lpEncodeString(dst,ele,size);
+        }
+        dst += enclen;
+        memcpy(dst,backlen,backlen_size);
+        dst += backlen_size;
+    }
+
+    /* Update header. */
+    if (where != LP_REPLACE || ele == NULL) {
+        uint32_t num_elements = lpGetNumElements(lp);
+        if (num_elements != LP_HDR_NUMELE_UNKNOWN) {
+            if (ele)
+                lpSetNumElements(lp,num_elements+1);
+            else
+                lpSetNumElements(lp,num_elements-1);
+        }
+    }
+    lpSetTotalBytes(lp,new_listpack_bytes);
+    return lp;
+}
+
+/* Append the specified element 'ele' of lenght 'len' at the end of the
+ * listpack. It is implemented in terms of lpInsert(), so the return value is
+ * the same as lpInsert(). */
+unsigned char *lpAppend(unsigned char *lp, unsigned char *ele, uint32_t size) {
+    uint64_t listpack_bytes = lpGetTotalBytes(lp);
+    unsigned char *eofptr = lp + listpack_bytes - 1;
+    return lpInsert(lp,ele,size,eofptr,LP_BEFORE,NULL);
+}
+
+/* Remove the element pointed by 'p', and return the resulting listpack.
+ * If 'newp' is not NULL, the next element pointer (to the right of the
+ * deleted one) is returned by reference. If the deleted element was the
+ * last one, '*newp' is set to NULL. */
+unsigned char *lpDelete(unsigned char *lp, unsigned char *p, unsigned char **newp) {
+    return lpInsert(lp,NULL,0,p,LP_REPLACE,newp);
+}
+
+/* Return the total number of bytes the listpack is composed of. */
+uint32_t lpBytes(unsigned char *lp) {
+    return lpGetTotalBytes(lp);
+}
+
+/* Seek the specified element and returns the pointer to the seeked element.
+ * Positive indexes specify the zero-based element to seek from the head to
+ * the tail, negative indexes specify elements starting from the tail, where
+ * -1 means the last element, -2 the penultimate and so forth. If the index
+ * is out of range, NULL is returned. */
+unsigned char *lpSeek(unsigned char *lp, long index) {
+    int forward = 1; /* Seek forward by default. */
+
+    /* We want to seek from left to right or the other way around
+     * depending on the listpack length and the element position.
+     * However if the listpack length cannot be obtained in constant time,
+     * we always seek from left to right. */
+    uint32_t numele = lpGetNumElements(lp);
+    if (numele != LP_HDR_NUMELE_UNKNOWN) {
+        if (index < 0) index = (long)numele+index;
+        if (index < 0) return NULL; /* Index still < 0 means out of range. */
+        if (index >= numele) return NULL; /* Out of range the other side. */
+        /* We want to scan right-to-left if the element we are looking for
+         * is past the half of the listpack. */
+        if (index > numele/2) {
+            forward = 0;
+            /* Left to right scanning always expects a negative index. Convert
+             * our index to negative form. */
+            index -= numele;
+        }
+    } else {
+        /* If the listpack length is unspecified, for negative indexes we
+         * want to always scan left-to-right. */
+        if (index < 0) forward = 0;
+    }
+
+    /* Forward and backward scanning is trivially based on lpNext()/lpPrev(). */
+    if (forward) {
+        unsigned char *ele = lpFirst(lp);
+        while (index > 0 && ele) {
+            ele = lpNext(lp,ele);
+            index--;
+        }
+        return ele;
+    } else {
+        unsigned char *ele = lpLast(lp);
+        while (index < -1 && ele) {
+            ele = lpPrev(lp,ele);
+            index++;
+        }
+        return ele;
+    }
+}
+

src/listpack.h

@@ -0,0 +1,61 @@
+/* Listpack -- A lists of strings serialization format
+ *
+ * This file implements the specification you can find at:
+ *
+ *  https://github.com/antirez/listpack
+ *
+ * Copyright (c) 2017, Salvatore Sanfilippo <antirez at gmail dot com>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ *   * Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *   * Redistributions in binary form must reproduce the above copyright
+ *     notice, this list of conditions and the following disclaimer in the
+ *     documentation and/or other materials provided with the distribution.
+ *   * Neither the name of Redis nor the names of its contributors may be used
+ *     to endorse or promote products derived from this software without
+ *     specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __LISTPACK_H
+#define __LISTPACK_H
+
+#include <stdint.h>
+
+#define LP_INTBUF_SIZE 21 /* 20 digits of -2^63 + 1 null term = 21. */
+
+/* lpInsert() where argument possible values: */
+#define LP_BEFORE 0
+#define LP_AFTER 1
+#define LP_REPLACE 2
+
+unsigned char *lpNew(void);
+void lpFree(unsigned char *lp);
+unsigned char *lpInsert(unsigned char *lp, unsigned char *ele, uint32_t size, unsigned char *p, int where, unsigned char **newp);
+unsigned char *lpAppend(unsigned char *lp, unsigned char *ele, uint32_t size);
+unsigned char *lpDelete(unsigned char *lp, unsigned char *p, unsigned char **newp);
+uint32_t lpLength(unsigned char *lp);
+unsigned char *lpGet(unsigned char *p, int64_t *count, unsigned char *intbuf);
+unsigned char *lpFirst(unsigned char *lp);
+unsigned char *lpLast(unsigned char *lp);
+unsigned char *lpNext(unsigned char *lp, unsigned char *p);
+unsigned char *lpPrev(unsigned char *lp, unsigned char *p);
+uint32_t lpBytes(unsigned char *lp);
+unsigned char *lpSeek(unsigned char *lp, long index);
+
+#endif

src/listpack_malloc.h

@@ -0,0 +1,45 @@
+/* Listpack -- A lists of strings serialization format
+ * https://github.com/antirez/listpack
+ *
+ * Copyright (c) 2017, Salvatore Sanfilippo <antirez at gmail dot com>
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ *   * Redistributions of source code must retain the above copyright notice,
+ *     this list of conditions and the following disclaimer.
+ *   * Redistributions in binary form must reproduce the above copyright
+ *     notice, this list of conditions and the following disclaimer in the
+ *     documentation and/or other materials provided with the distribution.
+ *   * Neither the name of Redis nor the names of its contributors may be used
+ *     to endorse or promote products derived from this software without
+ *     specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* Allocator selection.
+ *
+ * This file is used in order to change the Rax allocator at compile time.
+ * Just define the following defines to what you want to use. Also add
+ * the include of your alternate allocator if needed (not needed in order
+ * to use the default libc allocator). */
+
+#ifndef LISTPACK_ALLOC_H
+#define LISTPACK_ALLOC_H
+#include "zmalloc.h"
+#define lp_malloc zmalloc
+#define lp_realloc zrealloc
+#define lp_free zfree
+#endif

src/lzfP.h

@@ -79,7 +79,11 @@
  * Unconditionally aligning does not cost very much, so do it if unsure
  */
 #ifndef STRICT_ALIGN
-# define STRICT_ALIGN !(defined(__i386) || defined (__amd64))
+# if !(defined(__i386) || defined (__amd64))
+#  define STRICT_ALIGN 1
+# else
+#  define STRICT_ALIGN 0
+# endif
 #endif
 
 /*

src/module.c

@@ -216,6 +216,31 @@ static list *moduleUnblockedClients;
  * allow thread safe contexts to execute commands at a safe moment. */
 static pthread_mutex_t moduleGIL = PTHREAD_MUTEX_INITIALIZER;
 
+
+/* Function pointer type for keyspace event notification subscriptions from modules. */
+typedef int (*RedisModuleNotificationFunc) (RedisModuleCtx *ctx, int type, const char *event, RedisModuleString *key);
+
+/* Keyspace notification subscriber information.
+ * See RM_SubscribeToKeyspaceEvents() for more information. */
+typedef struct RedisModuleKeyspaceSubscriber {
+    /* The module subscribed to the event */
+    RedisModule *module;
+    /* Notification callback in the module*/
+    RedisModuleNotificationFunc notify_callback;
+    /* A bit mask of the events the module is interested in */
+    int event_mask;
+    /* Active flag set on entry, to avoid reentrant subscribers
+     * calling themselves */
+    int active;
+} RedisModuleKeyspaceSubscriber;
+
+/* The module keyspace notification subscribers list */
+static list *moduleKeyspaceSubscribers;
+
+/* Static client recycled for all notification clients, to avoid allocating
+ * per round. */
+static client *moduleKeyspaceSubscribersClient;
+
 /* --------------------------------------------------------------------------
  * Prototypes
  * -------------------------------------------------------------------------- */
@@ -442,9 +467,7 @@ void moduleFreeContext(RedisModuleCtx *ctx) {
 void moduleHandlePropagationAfterCommandCallback(RedisModuleCtx *ctx) {
     client *c = ctx->client;
 
-    /* We don't want any automatic propagation here since in modules we handle
-     * replication / AOF propagation in explicit ways. */
-    preventCommandPropagation(c);
+    if (c->flags & CLIENT_LUA) return;
 
     /* Handle the replication of the final EXEC, since whatever a command
      * emits is always wrappered around MULTI/EXEC. */
@@ -615,7 +638,7 @@ int RM_CreateCommand(RedisModuleCtx *ctx, const char *name, RedisModuleCmdFunc c
     sds cmdname = sdsnew(name);
 
     /* Check if the command name is busy. */
-    if (lookupCommand((char*)name) != NULL) {
+    if (lookupCommand(cmdname) != NULL) {
         sdsfree(cmdname);
         return REDISMODULE_ERR;
     }
@@ -650,7 +673,7 @@ int RM_CreateCommand(RedisModuleCtx *ctx, const char *name, RedisModuleCmdFunc c
  *
  * This is an internal function, Redis modules developers don't need
  * to use it. */
-void RM_SetModuleAttribs(RedisModuleCtx *ctx, const char *name, int ver, int apiver){
+void RM_SetModuleAttribs(RedisModuleCtx *ctx, const char *name, int ver, int apiver) {
     RedisModule *module;
 
     if (ctx->module != NULL) return;
@@ -662,6 +685,15 @@ void RM_SetModuleAttribs(RedisModuleCtx *ctx, const char *name, int ver, int api
     ctx->module = module;
 }
 
+/* Return non-zero if the module name is busy.
+ * Otherwise zero is returned. */
+int RM_IsModuleNameBusy(const char *name) {
+    sds modulename = sdsnew(name);
+    dictEntry *de = dictFind(modules,modulename);
+    sdsfree(modulename);
+    return de != NULL;
+}
+
 /* Return the current UNIX time in milliseconds. */
 long long RM_Milliseconds(void) {
     return mstime();
@@ -1164,6 +1196,9 @@ int RM_ReplyWithDouble(RedisModuleCtx *ctx, double d) {
  * in the context of a command execution. EXEC will be handled by the
  * RedisModuleCommandDispatcher() function. */
 void moduleReplicateMultiIfNeeded(RedisModuleCtx *ctx) {
+    /* Skip this if client explicitly wrap the command with MULTI, or if
+     * the module command was called by a script. */
+    if (ctx->client->flags & (CLIENT_MULTI|CLIENT_LUA)) return;
     /* If we already emitted MULTI return ASAP. */
     if (ctx->flags & REDISMODULE_CTX_MULTI_EMITTED) return;
     /* If this is a thread safe context, we do not want to wrap commands
@@ -1216,6 +1251,7 @@ int RM_Replicate(RedisModuleCtx *ctx, const char *cmdname, const char *fmt, ...)
     /* Release the argv. */
     for (j = 0; j < argc; j++) decrRefCount(argv[j]);
     zfree(argv);
+    server.dirty++;
     return REDISMODULE_OK;
 }
 
@@ -1234,6 +1270,7 @@ int RM_ReplicateVerbatim(RedisModuleCtx *ctx) {
     alsoPropagate(ctx->client->cmd,ctx->client->db->id,
         ctx->client->argv,ctx->client->argc,
         PROPAGATE_AOF|PROPAGATE_REPL);
+    server.dirty++;
     return REDISMODULE_OK;
 }
 
@@ -1444,6 +1481,20 @@ int RM_DeleteKey(RedisModuleKey *key) {
     return REDISMODULE_OK;
 }
 
+/* If the key is open for writing, unlink it (that is delete it in a 
+ * non-blocking way, not reclaiming memory immediately) and setup the key to
+ * accept new writes as an empty key (that will be created on demand).
+ * On success REDISMODULE_OK is returned. If the key is not open for
+ * writing REDISMODULE_ERR is returned. */
+int RM_UnlinkKey(RedisModuleKey *key) {
+    if (!(key->mode & REDISMODULE_WRITE)) return REDISMODULE_ERR;
+    if (key->value) {
+        dbAsyncDelete(key->db,key->key);
+        key->value = NULL;
+    }
+    return REDISMODULE_OK;
+}
+
 /* Return the key expire value, as milliseconds of remaining TTL.
  * If no TTL is associated with the key or if the key is empty,
  * REDISMODULE_NO_EXPIRE is returned. */
@@ -3012,7 +3063,7 @@ int64_t RM_LoadSigned(RedisModuleIO *io) {
 void RM_SaveString(RedisModuleIO *io, RedisModuleString *s) {
     if (io->error) return;
     /* Save opcode. */
-    int retval = rdbSaveLen(io->rio, RDB_MODULE_OPCODE_STRING);
+    ssize_t retval = rdbSaveLen(io->rio, RDB_MODULE_OPCODE_STRING);
     if (retval == -1) goto saveerr;
     io->bytes += retval;
     /* Save value. */
@@ -3030,7 +3081,7 @@ saveerr:
 void RM_SaveStringBuffer(RedisModuleIO *io, const char *str, size_t len) {
     if (io->error) return;
     /* Save opcode. */
-    int retval = rdbSaveLen(io->rio, RDB_MODULE_OPCODE_STRING);
+    ssize_t retval = rdbSaveLen(io->rio, RDB_MODULE_OPCODE_STRING);
     if (retval == -1) goto saveerr;
     io->bytes += retval;
     /* Save value. */
@@ -3643,6 +3694,120 @@ void moduleReleaseGIL(void) {
     pthread_mutex_unlock(&moduleGIL);
 }
 
+
+/* --------------------------------------------------------------------------
+ * Module Keyspace Notifications API
+ * -------------------------------------------------------------------------- */
+
+/* Subscribe to keyspace notifications. This is a low-level version of the
+ * keyspace-notifications API. A module cand register callbacks to be notified
+ * when keyspce events occur.
+ *
+ * Notification events are filtered by their type (string events, set events,
+ * etc), and the subsriber callback receives only events that match a specific
+ * mask of event types.
+ *
+ * When subscribing to notifications with RedisModule_SubscribeToKeyspaceEvents 
+ * the module must provide an event type-mask, denoting the events the subscriber
+ * is interested in. This can be an ORed mask of any of the following flags:
+ *
+ *  - REDISMODULE_NOTIFY_GENERIC: Generic commands like DEL, EXPIRE, RENAME
+ *  - REDISMODULE_NOTIFY_STRING: String events
+ *  - REDISMODULE_NOTIFY_LIST: List events
+ *  - REDISMODULE_NOTIFY_SET: Set events
+ *  - REDISMODULE_NOTIFY_HASH: Hash events
+ *  - REDISMODULE_NOTIFY_ZSET: Sorted Set events
+ *  - REDISMODULE_NOTIFY_EXPIRED: Expiration events
+ *  - REDISMODULE_NOTIFY_EVICTED: Eviction events
+ *  - REDISMODULE_NOTIFY_STREAM: Stream events
+ *  - REDISMODULE_NOTIFY_ALL: All events
+ *
+ * We do not distinguish between key events and keyspace events, and it is up
+ * to the module to filter the actions taken based on the key.
+ *
+ * The subscriber signature is:
+ *
+ *   int (*RedisModuleNotificationFunc) (RedisModuleCtx *ctx, int type,
+ *                                       const char *event,
+ *                                       RedisModuleString *key);
+ *
+ * `type` is the event type bit, that must match the mask given at registration
+ * time. The event string is the actual command being executed, and key is the
+ * relevant Redis key.
+ *
+ * Notification callback gets executed with a redis context that can not be
+ * used to send anything to the client, and has the db number where the event
+ * occured as its selected db number.
+ *
+ * Notice that it is not necessary to enable norifications in redis.conf for
+ * module notifications to work.
+ *
+ * Warning: the notification callbacks are performed in a synchronous manner,
+ * so notification callbacks must to be fast, or they would slow Redis down.
+ * If you need to take long actions, use threads to offload them.
+ *
+ * See https://redis.io/topics/notifications for more information.
+ */
+int RM_SubscribeToKeyspaceEvents(RedisModuleCtx *ctx, int types, RedisModuleNotificationFunc callback) {
+    RedisModuleKeyspaceSubscriber *sub = zmalloc(sizeof(*sub));
+    sub->module = ctx->module;
+    sub->event_mask = types;
+    sub->notify_callback = callback;
+    sub->active = 0;
+
+    listAddNodeTail(moduleKeyspaceSubscribers, sub);
+    return REDISMODULE_OK;
+}
+
+/* Dispatcher for keyspace notifications to module subscriber functions.
+ * This gets called  only if at least one module requested to be notified on
+ * keyspace notifications */
+void moduleNotifyKeyspaceEvent(int type, const char *event, robj *key, int dbid) {
+    /* Don't do anything if there aren't any subscribers */
+    if (listLength(moduleKeyspaceSubscribers) == 0) return;
+
+    listIter li;
+    listNode *ln;
+    listRewind(moduleKeyspaceSubscribers,&li);
+
+    /* Remove irrelevant flags from the type mask */
+    type &= ~(NOTIFY_KEYEVENT | NOTIFY_KEYSPACE);
+
+    while((ln = listNext(&li))) {
+        RedisModuleKeyspaceSubscriber *sub = ln->value;
+        /* Only notify subscribers on events matching they registration,
+         * and avoid subscribers triggering themselves */
+        if ((sub->event_mask & type) && sub->active == 0) {
+            RedisModuleCtx ctx = REDISMODULE_CTX_INIT;
+            ctx.module = sub->module;
+            ctx.client = moduleKeyspaceSubscribersClient;
+            selectDb(ctx.client, dbid);
+
+            /* mark the handler as activer to avoid reentrant loops.
+             * If the subscriber performs an action triggering itself,
+             * it will not be notified about it. */
+            sub->active = 1;
+            sub->notify_callback(&ctx, type, event, key);
+            sub->active = 0;
+            moduleFreeContext(&ctx);
+        }
+    }
+}
+
+/* Unsubscribe any notification subscirbers this module has upon unloading */
+void moduleUnsubscribeNotifications(RedisModule *module) {
+    listIter li;
+    listNode *ln;
+    listRewind(moduleKeyspaceSubscribers,&li);
+    while((ln = listNext(&li))) {
+        RedisModuleKeyspaceSubscriber *sub = ln->value;
+        if (sub->module == module) {
+            listDelNode(moduleKeyspaceSubscribers, ln);
+            zfree(sub);
+        }
+    }
+}
+
 /* --------------------------------------------------------------------------
  * Modules API internals
  * -------------------------------------------------------------------------- */
@@ -3680,9 +3845,14 @@ void moduleRegisterCoreAPI(void);
 
 void moduleInitModulesSystem(void) {
     moduleUnblockedClients = listCreate();
-
     server.loadmodule_queue = listCreate();
     modules = dictCreate(&modulesDictType,NULL);
+
+    /* Set up the keyspace notification susbscriber list and static client */
+    moduleKeyspaceSubscribers = listCreate();
+    moduleKeyspaceSubscribersClient = createClient(-1);
+    moduleKeyspaceSubscribersClient->flags |= CLIENT_MODULE;
+
     moduleRegisterCoreAPI();
     if (pipe(server.module_blocked_pipe) == -1) {
         serverLog(LL_WARNING,
@@ -3733,6 +3903,28 @@ void moduleFreeModuleStructure(struct RedisModule *module) {
     zfree(module);
 }
 
+void moduleUnregisterCommands(struct RedisModule *module) {
+    /* Unregister all the commands registered by this module. */
+    dictIterator *di = dictGetSafeIterator(server.commands);
+    dictEntry *de;
+    while ((de = dictNext(di)) != NULL) {
+        struct redisCommand *cmd = dictGetVal(de);
+        if (cmd->proc == RedisModuleCommandDispatcher) {
+            RedisModuleCommandProxy *cp =
+                (void*)(unsigned long)cmd->getkeys_proc;
+            sds cmdname = cp->rediscmd->name;
+            if (cp->module == module) {
+                dictDelete(server.commands,cmdname);
+                dictDelete(server.orig_commands,cmdname);
+                sdsfree(cmdname);
+                zfree(cp->rediscmd);
+                zfree(cp);
+            }
+        }
+    }
+    dictReleaseIterator(di);
+}
+
 /* Load a module and initialize it. On success C_OK is returned, otherwise
  * C_ERR is returned. */
 int moduleLoad(const char *path, void **module_argv, int module_argc) {
@@ -3753,7 +3945,10 @@ int moduleLoad(const char *path, void **module_argv, int module_argc) {
         return C_ERR;
     }
     if (onload((void*)&ctx,module_argv,module_argc) == REDISMODULE_ERR) {
-        if (ctx.module) moduleFreeModuleStructure(ctx.module);
+        if (ctx.module) {
+            moduleUnregisterCommands(ctx.module);
+            moduleFreeModuleStructure(ctx.module);
+        }
         dlclose(handle);
         serverLog(LL_WARNING,
             "Module %s initialization failed. Module not loaded",path);
@@ -3768,6 +3963,7 @@ int moduleLoad(const char *path, void **module_argv, int module_argc) {
     return C_OK;
 }
 
+
 /* Unload the module registered with the specified name. On success
  * C_OK is returned, otherwise C_ERR is returned and errno is set
  * to the following values depending on the type of error:
@@ -3787,25 +3983,10 @@ int moduleUnload(sds name) {
         return REDISMODULE_ERR;
     }
 
-    /* Unregister all the commands registered by this module. */
-    dictIterator *di = dictGetSafeIterator(server.commands);
-    dictEntry *de;
-    while ((de = dictNext(di)) != NULL) {
-        struct redisCommand *cmd = dictGetVal(de);
-        if (cmd->proc == RedisModuleCommandDispatcher) {
-            RedisModuleCommandProxy *cp =
-                (void*)(unsigned long)cmd->getkeys_proc;
-            sds cmdname = cp->rediscmd->name;
-            if (cp->module == module) {
-                dictDelete(server.commands,cmdname);
-                dictDelete(server.orig_commands,cmdname);
-                sdsfree(cmdname);
-                zfree(cp->rediscmd);
-                zfree(cp);
-            }
-        }
-    }
-    dictReleaseIterator(di);
+    moduleUnregisterCommands(module);
+
+    /* Remvoe any noification subscribers this module might have */
+    moduleUnsubscribeNotifications(module);
 
     /* Unregister all the hooks. TODO: Yet no hooks support here. */
 
@@ -3900,6 +4081,7 @@ void moduleRegisterCoreAPI(void) {
     REGISTER_API(Strdup);
     REGISTER_API(CreateCommand);
     REGISTER_API(SetModuleAttribs);
+    REGISTER_API(IsModuleNameBusy);
     REGISTER_API(WrongArity);
     REGISTER_API(ReplyWithLongLong);
     REGISTER_API(ReplyWithError);
@@ -3940,6 +4122,7 @@ void moduleRegisterCoreAPI(void) {
     REGISTER_API(Replicate);
     REGISTER_API(ReplicateVerbatim);
     REGISTER_API(DeleteKey);
+    REGISTER_API(UnlinkKey);
     REGISTER_API(StringSet);
     REGISTER_API(StringDMA);
     REGISTER_API(StringTruncate);
@@ -4002,4 +4185,5 @@ void moduleRegisterCoreAPI(void) {
     REGISTER_API(DigestAddStringBuffer);
     REGISTER_API(DigestAddLongLong);
     REGISTER_API(DigestEndSequence);
+    REGISTER_API(SubscribeToKeyspaceEvents);
 }

src/modules/testmodule.c

@@ -30,6 +30,7 @@
  * POSSIBILITY OF SUCH DAMAGE.
  */
 
+#define REDISMODULE_EXPERIMENTAL_API
 #include "../redismodule.h"
 #include <string.h>
 
@@ -120,81 +121,187 @@ int TestStringPrintf(RedisModuleCtx *ctx, RedisModuleString **argv, int argc) {
     return REDISMODULE_OK;
 }
 
+int failTest(RedisModuleCtx *ctx, const char *msg) {
+    RedisModule_ReplyWithError(ctx, msg);
+    return REDISMODULE_ERR;
+}
+
+int TestUnlink(RedisModuleCtx *ctx, RedisModuleString **argv, int argc) {
+    RedisModule_AutoMemory(ctx);
+    REDISMODULE_NOT_USED(argv);
+    REDISMODULE_NOT_USED(argc);
+
+    RedisModuleKey *k = RedisModule_OpenKey(ctx, RedisModule_CreateStringPrintf(ctx, "unlinked"), REDISMODULE_WRITE | REDISMODULE_READ);
+    if (!k) return failTest(ctx, "Could not create key");
+    
+    if (REDISMODULE_ERR == RedisModule_StringSet(k, RedisModule_CreateStringPrintf(ctx, "Foobar"))) {
+        return failTest(ctx, "Could not set string value");
+    }
+
+    RedisModuleCallReply *rep = RedisModule_Call(ctx, "EXISTS", "c", "unlinked");
+    if (!rep || RedisModule_CallReplyInteger(rep) != 1) {
+        return failTest(ctx, "Key does not exist before unlink");
+    }
+
+    if (REDISMODULE_ERR == RedisModule_UnlinkKey(k)) {
+        return failTest(ctx, "Could not unlink key");
+    }
+
+    rep = RedisModule_Call(ctx, "EXISTS", "c", "unlinked");
+    if (!rep || RedisModule_CallReplyInteger(rep) != 0) {
+        return failTest(ctx, "Could not verify key to be unlinked");
+    }
+    return RedisModule_ReplyWithSimpleString(ctx, "OK");
+    
+}
+
+int NotifyCallback(RedisModuleCtx *ctx, int type, const char *event,
+                   RedisModuleString *key) {
+  /* Increment a counter on the notifications: for each key notified we
+   * increment a counter */
+  RedisModule_Log(ctx, "notice", "Got event type %d, event %s, key %s", type,
+                  event, RedisModule_StringPtrLen(key, NULL));
+
+  RedisModule_Call(ctx, "HINCRBY", "csc", "notifications", key, "1");
+  return REDISMODULE_OK;
+}
+
+/* TEST.NOTIFICATIONS -- Test Keyspace Notifications. */
+int TestNotifications(RedisModuleCtx *ctx, RedisModuleString **argv, int argc) {
+    REDISMODULE_NOT_USED(argv);
+    REDISMODULE_NOT_USED(argc);
+
+#define FAIL(msg, ...)                                                                       \
+    {                                                                                        \
+        RedisModule_Log(ctx, "warning", "Failed NOTIFY Test. Reason: " #msg, ##__VA_ARGS__); \
+        goto err;                                                                            \
+    }
+    RedisModule_Call(ctx, "FLUSHDB", "");
+
+    RedisModule_Call(ctx, "SET", "cc", "foo", "bar");
+    RedisModule_Call(ctx, "SET", "cc", "foo", "baz");
+    RedisModule_Call(ctx, "SADD", "cc", "bar", "x");
+    RedisModule_Call(ctx, "SADD", "cc", "bar", "y");
+
+    RedisModule_Call(ctx, "HSET", "ccc", "baz", "x", "y");
+    /* LPUSH should be ignored and not increment any counters */
+    RedisModule_Call(ctx, "LPUSH", "cc", "l", "y");
+    RedisModule_Call(ctx, "LPUSH", "cc", "l", "y");
+
+    size_t sz;
+    const char *rep;
+    RedisModuleCallReply *r = RedisModule_Call(ctx, "HGET", "cc", "notifications", "foo");
+    if (r == NULL || RedisModule_CallReplyType(r) != REDISMODULE_REPLY_STRING) {
+        FAIL("Wrong or no reply for foo");
+    } else {
+        rep = RedisModule_CallReplyStringPtr(r, &sz);
+        if (sz != 1 || *rep != '2') {
+            FAIL("Got reply '%s'. expected '2'", RedisModule_CallReplyStringPtr(r, NULL));
+        }
+    }
+
+    r = RedisModule_Call(ctx, "HGET", "cc", "notifications", "bar");
+    if (r == NULL || RedisModule_CallReplyType(r) != REDISMODULE_REPLY_STRING) {
+        FAIL("Wrong or no reply for bar");
+    } else {
+        rep = RedisModule_CallReplyStringPtr(r, &sz);
+        if (sz != 1 || *rep != '2') {
+            FAIL("Got reply '%s'. expected '2'", rep);
+        }
+    }
+
+    r = RedisModule_Call(ctx, "HGET", "cc", "notifications", "baz");
+    if (r == NULL || RedisModule_CallReplyType(r) != REDISMODULE_REPLY_STRING) {
+        FAIL("Wrong or no reply for baz");
+    } else {
+        rep = RedisModule_CallReplyStringPtr(r, &sz);
+        if (sz != 1 || *rep != '1') {
+            FAIL("Got reply '%.*s'. expected '1'", sz, rep);
+        }
+    }
+    /* For l we expect nothing since we didn't subscribe to list events */
+    r = RedisModule_Call(ctx, "HGET", "cc", "notifications", "l");
+    if (r == NULL || RedisModule_CallReplyType(r) != REDISMODULE_REPLY_NULL) {
+        FAIL("Wrong reply for l");
+    }
+
+    RedisModule_Call(ctx, "FLUSHDB", "");
+
+    return RedisModule_ReplyWithSimpleString(ctx, "OK");
+err:
+    RedisModule_Call(ctx, "FLUSHDB", "");
+
+    return RedisModule_ReplyWithSimpleString(ctx, "ERR");
+}
 
 /* TEST.CTXFLAGS -- Test GetContextFlags. */
 int TestCtxFlags(RedisModuleCtx *ctx, RedisModuleString **argv, int argc) {
     REDISMODULE_NOT_USED(argc);
     REDISMODULE_NOT_USED(argv);
-  
+
     RedisModule_AutoMemory(ctx);
-  
+
     int ok = 1;
     const char *errString = NULL;
-  
-  #define FAIL(msg)    \
-    {                  \
-      ok = 0;          \
-      errString = msg; \
-      goto end;        \
+#undef FAIL
+#define FAIL(msg)        \
+    {                    \
+        ok = 0;          \
+        errString = msg; \
+        goto end;        \
     }
-  
+
     int flags = RedisModule_GetContextFlags(ctx);
     if (flags == 0) {
-      FAIL("Got no flags");
+        FAIL("Got no flags");
     }
-  
+
     if (flags & REDISMODULE_CTX_FLAGS_LUA) FAIL("Lua flag was set");
     if (flags & REDISMODULE_CTX_FLAGS_MULTI) FAIL("Multi flag was set");
-  
+
     if (flags & REDISMODULE_CTX_FLAGS_AOF) FAIL("AOF Flag was set")
     /* Enable AOF to test AOF flags */
     RedisModule_Call(ctx, "config", "ccc", "set", "appendonly", "yes");
     flags = RedisModule_GetContextFlags(ctx);
-    if (!(flags & REDISMODULE_CTX_FLAGS_AOF))
-      FAIL("AOF Flag not set after config set");
-  
+    if (!(flags & REDISMODULE_CTX_FLAGS_AOF)) FAIL("AOF Flag not set after config set");
+
     if (flags & REDISMODULE_CTX_FLAGS_RDB) FAIL("RDB Flag was set");
     /* Enable RDB to test RDB flags */
     RedisModule_Call(ctx, "config", "ccc", "set", "save", "900 1");
     flags = RedisModule_GetContextFlags(ctx);
-    if (!(flags & REDISMODULE_CTX_FLAGS_RDB))
-      FAIL("RDB Flag was not set after config set");
-  
+    if (!(flags & REDISMODULE_CTX_FLAGS_RDB)) FAIL("RDB Flag was not set after config set");
+
     if (!(flags & REDISMODULE_CTX_FLAGS_MASTER)) FAIL("Master flag was not set");
     if (flags & REDISMODULE_CTX_FLAGS_SLAVE) FAIL("Slave flag was set");
     if (flags & REDISMODULE_CTX_FLAGS_READONLY) FAIL("Read-only flag was set");
     if (flags & REDISMODULE_CTX_FLAGS_CLUSTER) FAIL("Cluster flag was set");
-  
+
     if (flags & REDISMODULE_CTX_FLAGS_MAXMEMORY) FAIL("Maxmemory flag was set");
-    ;
+
     RedisModule_Call(ctx, "config", "ccc", "set", "maxmemory", "100000000");
     flags = RedisModule_GetContextFlags(ctx);
     if (!(flags & REDISMODULE_CTX_FLAGS_MAXMEMORY))
-      FAIL("Maxmemory flag was not set after config set");
-  
+        FAIL("Maxmemory flag was not set after config set");
+
     if (flags & REDISMODULE_CTX_FLAGS_EVICT) FAIL("Eviction flag was set");
-    RedisModule_Call(ctx, "config", "ccc", "set", "maxmemory-policy",
-                     "allkeys-lru");
+    RedisModule_Call(ctx, "config", "ccc", "set", "maxmemory-policy", "allkeys-lru");
     flags = RedisModule_GetContextFlags(ctx);
-    if (!(flags & REDISMODULE_CTX_FLAGS_EVICT))
-      FAIL("Eviction flag was not set after config set");
-  
-  end:
+    if (!(flags & REDISMODULE_CTX_FLAGS_EVICT)) FAIL("Eviction flag was not set after config set");
+
+end:
     /* Revert config changes */
     RedisModule_Call(ctx, "config", "ccc", "set", "appendonly", "no");
     RedisModule_Call(ctx, "config", "ccc", "set", "save", "");
     RedisModule_Call(ctx, "config", "ccc", "set", "maxmemory", "0");
     RedisModule_Call(ctx, "config", "ccc", "set", "maxmemory-policy", "noeviction");
-  
-    if (!ok) {
-      RedisModule_Log(ctx, "warning", "Failed CTXFLAGS Test. Reason: %s",
-                      errString);
-      return RedisModule_ReplyWithSimpleString(ctx, "ERR");
-    }
-  
-    return RedisModule_ReplyWithSimpleString(ctx, "OK");
-  }
 
+    if (!ok) {
+        RedisModule_Log(ctx, "warning", "Failed CTXFLAGS Test. Reason: %s", errString);
+        return RedisModule_ReplyWithSimpleString(ctx, "ERR");
+    }
+
+    return RedisModule_ReplyWithSimpleString(ctx, "OK");
+}
 
 /* ----------------------------- Test framework ----------------------------- */
 
@@ -269,12 +376,18 @@ int TestIt(RedisModuleCtx *ctx, RedisModuleString **argv, int argc) {
     T("test.string.append","");
     if (!TestAssertStringReply(ctx,reply,"foobar",6)) goto fail;
 
+    T("test.unlink","");
+    if (!TestAssertStringReply(ctx,reply,"OK",2)) goto fail;
+
     T("test.string.append.am","");
     if (!TestAssertStringReply(ctx,reply,"foobar",6)) goto fail;
 
     T("test.string.printf", "cc", "foo", "bar");
     if (!TestAssertStringReply(ctx,reply,"Got 3 args. argv[1]: foo, argv[2]: bar",38)) goto fail;
 
+    T("test.notify", "");
+    if (!TestAssertStringReply(ctx,reply,"OK",2)) goto fail;
+
     RedisModule_ReplyWithSimpleString(ctx,"ALL TESTS PASSED");
     return REDISMODULE_OK;
 
@@ -310,10 +423,23 @@ int RedisModule_OnLoad(RedisModuleCtx *ctx, RedisModuleString **argv, int argc)
     if (RedisModule_CreateCommand(ctx,"test.ctxflags",
         TestCtxFlags,"readonly",1,1,1) == REDISMODULE_ERR)
         return REDISMODULE_ERR;
+    
+    if (RedisModule_CreateCommand(ctx,"test.unlink",
+        TestUnlink,"write deny-oom",1,1,1) == REDISMODULE_ERR)
+        return REDISMODULE_ERR;
 
     if (RedisModule_CreateCommand(ctx,"test.it",
         TestIt,"readonly",1,1,1) == REDISMODULE_ERR)
         return REDISMODULE_ERR;
 
+    RedisModule_SubscribeToKeyspaceEvents(ctx,
+                                            REDISMODULE_NOTIFY_HASH |
+                                            REDISMODULE_NOTIFY_SET |
+                                            REDISMODULE_NOTIFY_STRING,
+                                        NotifyCallback);
+    if (RedisModule_CreateCommand(ctx,"test.notify",
+        TestNotifications,"write deny-oom",1,1,1) == REDISMODULE_ERR)
+        return REDISMODULE_ERR;
+
     return REDISMODULE_OK;
 }

src/networking.c

@@ -33,7 +33,7 @@
 #include <math.h>
 #include <ctype.h>
 
-static void setProtocolError(const char *errstr, client *c, int pos);
+static void setProtocolError(const char *errstr, client *c, long pos);
 
 /* Return the size consumed from the allocator, for the specified SDS string,
  * including internal fragmentation. This function is used in order to compute
@@ -67,6 +67,16 @@ int listMatchObjects(void *a, void *b) {
     return equalStringObjects(a,b);
 }
 
+/* This function links the client to the global linked list of clients.
+ * unlinkClient() does the opposite, among other things. */
+void linkClient(client *c) {
+    listAddNodeTail(server.clients,c);
+    /* Note that we remember the linked list node where the client is stored,
+     * this way removing the client in unlinkClient() will not require
+     * a linear scan, but just a constant time operation. */
+    c->client_list_node = listLast(server.clients);
+}
+
 client *createClient(int fd) {
     client *c = zmalloc(sizeof(client));
 
@@ -124,18 +134,21 @@ client *createClient(int fd) {
     listSetDupMethod(c->reply,dupClientReplyValue);
     c->btype = BLOCKED_NONE;
     c->bpop.timeout = 0;
-    c->bpop.keys = dictCreate(&objectKeyPointerValueDictType,NULL);
+    c->bpop.keys = dictCreate(&objectKeyHeapPointerValueDictType,NULL);
     c->bpop.target = NULL;
+    c->bpop.xread_group = NULL;
     c->bpop.numreplicas = 0;
+    c->bpop.aofepoch = 0;
     c->bpop.reploffset = 0;
     c->woff = 0;
     c->watched_keys = listCreate();
     c->pubsub_channels = dictCreate(&objectKeyPointerValueDictType,NULL);
     c->pubsub_patterns = listCreate();
     c->peerid = NULL;
+    c->client_list_node = NULL;
     listSetFreeMethod(c->pubsub_patterns,decrRefCountVoid);
     listSetMatchMethod(c->pubsub_patterns,listMatchObjects);
-    if (fd != -1) listAddNodeTail(server.clients,c);
+    if (fd != -1) linkClient(c);
     initClientMultiState(c);
     return c;
 }
@@ -376,6 +389,12 @@ void addReplyErrorLength(client *c, const char *s, size_t len) {
     addReplyString(c,"-ERR ",5);
     addReplyString(c,s,len);
     addReplyString(c,"\r\n",2);
+    if (c->flags & CLIENT_MASTER) {
+        char *cmdname = c->lastcmd ? c->lastcmd->name : "<unknown>";
+        serverLog(LL_WARNING,"== CRITICAL == This slave is sending an error "
+                             "to its master: '%s' after processing the command "
+                             "'%s'", s, cmdname);
+    }
 }
 
 void addReplyError(client *c, const char *err) {
@@ -566,7 +585,7 @@ void addReplyBulkSds(client *c, sds s)  {
     addReply(c,shared.crlf);
 }
 
-/* Add a C nul term string as bulk reply */
+/* Add a C null term string as bulk reply */
 void addReplyBulkCString(client *c, const char *s) {
     if (s == NULL) {
         addReply(c,shared.nullbulk);
@@ -584,6 +603,26 @@ void addReplyBulkLongLong(client *c, long long ll) {
     addReplyBulkCBuffer(c,buf,len);
 }
 
+/* Add an array of C strings as status replies with a heading.
+ * This function is typically invoked by from commands that support
+ * subcommands in response to the 'help' subcommand. The help array
+ * is terminated by NULL sentinel. */
+void addReplyHelp(client *c, const char **help) {
+    sds cmd = sdsnew((char*) c->argv[0]->ptr);
+    void *blenp = addDeferredMultiBulkLength(c);
+    int blen = 0;
+
+    sdstoupper(cmd);
+    addReplyStatusFormat(c,
+        "%s <subcommand> arg arg ... arg. Subcommands are:",cmd);
+    sdsfree(cmd);
+
+    while (help[blen]) addReplyStatus(c,help[blen++]);
+
+    blen++;  /* Account for the header line(s). */
+    setDeferredMultiBulkLength(c,blenp,blen);
+}
+
 /* Copy 'src' client output buffers into 'dst' client output buffers.
  * The function takes care of freeing the old output buffers of the
  * destination client. */
@@ -743,9 +782,10 @@ void unlinkClient(client *c) {
      * fd is already set to -1. */
     if (c->fd != -1) {
         /* Remove from the list of active clients. */
-        ln = listSearchKey(server.clients,c);
-        serverAssert(ln != NULL);
-        listDelNode(server.clients,ln);
+        if (c->client_list_node) {
+            listDelNode(server.clients,c->client_list_node);
+            c->client_list_node = NULL;
+        }
 
         /* Unregister async I/O handlers and close the socket. */
         aeDeleteFileEvent(server.el,c->fd,AE_READABLE);
@@ -939,10 +979,15 @@ int writeToClient(int fd, client *c, int handler_installed) {
          * scenario think about 'KEYS *' against the loopback interface).
          *
          * However if we are over the maxmemory limit we ignore that and
-         * just deliver as much data as it is possible to deliver. */
+         * just deliver as much data as it is possible to deliver.
+         *
+         * Moreover, we also send as much as possible if the client is
+         * a slave (otherwise, on high-speed traffic, the replication
+         * buffer will grow indefinitely) */
         if (totwritten > NET_MAX_WRITES_PER_EVENT &&
             (server.maxmemory == 0 ||
-             zmalloc_used_memory() < server.maxmemory)) break;
+             zmalloc_used_memory() < server.maxmemory) &&
+            !(c->flags & CLIENT_SLAVE)) break;
     }
     server.stat_net_output_bytes += totwritten;
     if (nwritten == -1) {
@@ -1000,13 +1045,25 @@ int handleClientsWithPendingWrites(void) {
         /* Try to write buffers to the client socket. */
         if (writeToClient(c->fd,c,0) == C_ERR) continue;
 
-        /* If there is nothing left, do nothing. Otherwise install
-         * the write handler. */
-        if (clientHasPendingReplies(c) &&
-            aeCreateFileEvent(server.el, c->fd, AE_WRITABLE,
+        /* If after the synchronous writes above we still have data to
+         * output to the client, we need to install the writable handler. */
+        if (clientHasPendingReplies(c)) {
+            int ae_flags = AE_WRITABLE;
+            /* For the fsync=always policy, we want that a given FD is never
+             * served for reading and writing in the same event loop iteration,
+             * so that in the middle of receiving the query, and serving it
+             * to the client, we'll call beforeSleep() that will do the
+             * actual fsync of AOF to disk. AE_BARRIER ensures that. */
+            if (server.aof_state == AOF_ON &&
+                server.aof_fsync == AOF_FSYNC_ALWAYS)
+            {
+                ae_flags |= AE_BARRIER;
+            }
+            if (aeCreateFileEvent(server.el, c->fd, AE_WRITABLE,
                 sendReplyToClient, c) == AE_ERR)
-        {
-            freeClientAsync(c);
+            {
+                    freeClientAsync(c);
+            }
         }
     }
     return processed;
@@ -1107,7 +1164,7 @@ int processInlineBuffer(client *c) {
 /* Helper function. Trims query buffer to make the function that processes
  * multi bulk requests idempotent. */
 #define PROTO_DUMP_LEN 128
-static void setProtocolError(const char *errstr, client *c, int pos) {
+static void setProtocolError(const char *errstr, client *c, long pos) {
     if (server.verbosity <= LL_VERBOSE) {
         sds client = catClientInfoString(sdsempty(),c);
 
@@ -1148,7 +1205,8 @@ static void setProtocolError(const char *errstr, client *c, int pos) {
  * to be '*'. Otherwise for inline commands processInlineBuffer() is called. */
 int processMultibulkBuffer(client *c) {
     char *newline = NULL;
-    int pos = 0, ok;
+    long pos = 0;
+    int ok;
     long long ll;
 
     if (c->multibulklen == 0) {
@@ -1220,7 +1278,7 @@ int processMultibulkBuffer(client *c) {
             }
 
             ok = string2ll(c->querybuf+pos+1,newline-(c->querybuf+pos+1),&ll);
-            if (!ok || ll < 0 || ll > 512*1024*1024) {
+            if (!ok || ll < 0 || ll > server.proto_max_bulk_len) {
                 addReplyError(c,"Protocol error: invalid bulk length");
                 setProtocolError("invalid bulk length",c,pos);
                 return C_ERR;
@@ -1246,7 +1304,7 @@ int processMultibulkBuffer(client *c) {
         }
 
         /* Read bulk argument */
-        if (sdslen(c->querybuf)-pos < (unsigned)(c->bulklen+2)) {
+        if (sdslen(c->querybuf)-pos < (size_t)(c->bulklen+2)) {
             /* Not enough data (+2 == trailing \r\n) */
             break;
         } else {
@@ -1255,13 +1313,13 @@ int processMultibulkBuffer(client *c) {
              * just use the current sds string. */
             if (pos == 0 &&
                 c->bulklen >= PROTO_MBULK_BIG_ARG &&
-                (signed) sdslen(c->querybuf) == c->bulklen+2)
+                sdslen(c->querybuf) == (size_t)(c->bulklen+2))
             {
                 c->argv[c->argc++] = createObject(OBJ_STRING,c->querybuf);
                 sdsIncrLen(c->querybuf,-2); /* remove CRLF */
                 /* Assume that if we saw a fat argument we'll see another one
                  * likely... */
-                c->querybuf = sdsnewlen(NULL,c->bulklen+2);
+                c->querybuf = sdsnewlen(SDS_NOINIT,c->bulklen+2);
                 sdsclear(c->querybuf);
                 pos = 0;
             } else {
@@ -1366,7 +1424,7 @@ void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask) {
     if (c->reqtype == PROTO_REQ_MULTIBULK && c->multibulklen && c->bulklen != -1
         && c->bulklen >= PROTO_MBULK_BIG_ARG)
     {
-        int remaining = (unsigned)(c->bulklen+2)-sdslen(c->querybuf);
+        ssize_t remaining = (size_t)(c->bulklen+2)-sdslen(c->querybuf);
 
         if (remaining < readlen) readlen = remaining;
     }
@@ -1540,7 +1598,7 @@ sds getAllClientsInfoString(void) {
     listNode *ln;
     listIter li;
     client *client;
-    sds o = sdsnewlen(NULL,200*listLength(server.clients));
+    sds o = sdsnewlen(SDS_NOINIT,200*listLength(server.clients));
     sdsclear(o);
     listRewind(server.clients,&li);
     while ((ln = listNext(&li)) != NULL) {
@@ -1556,7 +1614,22 @@ void clientCommand(client *c) {
     listIter li;
     client *client;
 
-    if (!strcasecmp(c->argv[1]->ptr,"list") && c->argc == 2) {
+    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"help")) {
+        const char *help[] = {
+"getname -- Return the name of the current connection.",
+"kill <ip:port> -- Kill connection made from <ip:port>.",
+"kill <option> <value> [option value ...] -- Kill connections. Options are:",
+"     addr <ip:port> -- Kill connection made from <ip:port>.",
+"     type (normal|master|slave|pubsub) -- Kill connections by type.",
+"     skipme (yes|no) -- Skip killing current connection (default: yes).",
+"list -- Return information about client connections.",
+"pause <timeout> -- Suspend all Redis clients for <timout> milliseconds.",
+"reply (on|off|skip) -- Control the replies sent to the current connection.",
+"setname <name> -- Assign the name <name> to the current connection.",
+NULL
+        };
+        addReplyHelp(c, help);
+    } else if (!strcasecmp(c->argv[1]->ptr,"list") && c->argc == 2) {
         /* CLIENT LIST */
         sds o = getAllClientsInfoString();
         addReplyBulkCBuffer(c,o,sdslen(o));
@@ -1702,7 +1775,7 @@ void clientCommand(client *c) {
         pauseClients(duration);
         addReply(c,shared.ok);
     } else {
-        addReplyError(c, "Syntax error, try CLIENT (LIST | KILL | GETNAME | SETNAME | PAUSE | REPLY)");
+        addReplyErrorFormat(c, "Unknown subcommand or wrong number of arguments for '%s'. Try CLIENT HELP", (char*)c->argv[1]->ptr);
     }
 }
 

src/notify.c

@@ -54,6 +54,7 @@ int keyspaceEventsStringToFlags(char *classes) {
         case 'e': flags |= NOTIFY_EVICTED; break;
         case 'K': flags |= NOTIFY_KEYSPACE; break;
         case 'E': flags |= NOTIFY_KEYEVENT; break;
+        case 't': flags |= NOTIFY_STREAM; break;
         default: return -1;
         }
     }
@@ -79,6 +80,7 @@ sds keyspaceEventsFlagsToString(int flags) {
         if (flags & NOTIFY_ZSET) res = sdscatlen(res,"z",1);
         if (flags & NOTIFY_EXPIRED) res = sdscatlen(res,"x",1);
         if (flags & NOTIFY_EVICTED) res = sdscatlen(res,"e",1);
+        if (flags & NOTIFY_STREAM) res = sdscatlen(res,"t",1);
     }
     if (flags & NOTIFY_KEYSPACE) res = sdscatlen(res,"K",1);
     if (flags & NOTIFY_KEYEVENT) res = sdscatlen(res,"E",1);
@@ -98,6 +100,12 @@ void notifyKeyspaceEvent(int type, char *event, robj *key, int dbid) {
     int len = -1;
     char buf[24];
 
+    /* If any modules are interested in events, notify the module system now. 
+     * This bypasses the notifications configuration, but the module engine
+     * will only call event subscribers if the event type matches the types
+     * they are interested in. */
+     moduleNotifyKeyspaceEvent(type, event, key, dbid);
+    
     /* If notifications for this class of events are off, return ASAP. */
     if (!(server.notify_keyspace_events & type)) return;
 
@@ -115,7 +123,7 @@ void notifyKeyspaceEvent(int type, char *event, robj *key, int dbid) {
         decrRefCount(chanobj);
     }
 
-    /* __keyevente@<db>__:<event> <key> notifications. */
+    /* __keyevent@<db>__:<event> <key> notifications. */
     if (server.notify_keyspace_events & NOTIFY_KEYEVENT) {
         chan = sdsnewlen("__keyevent@",11);
         if (len == -1) len = ll2string(buf,sizeof(buf),dbid);

src/object.c

@@ -98,7 +98,9 @@ robj *createEmbeddedStringObject(const char *ptr, size_t len) {
     sh->len = len;
     sh->alloc = len;
     sh->flags = SDS_TYPE_8;
-    if (ptr) {
+    if (ptr == SDS_NOINIT)
+        sh->buf[len] = '\0';
+    else if (ptr) {
         memcpy(sh->buf,ptr,len);
         sh->buf[len] = '\0';
     } else {
@@ -145,7 +147,7 @@ robj *createStringObjectFromLongLong(long long value) {
  *
  * The 'humanfriendly' option is used for INCRBYFLOAT and HINCRBYFLOAT. */
 robj *createStringObjectFromLongDouble(long double value, int humanfriendly) {
-    char buf[256];
+    char buf[MAX_LONG_DOUBLE_CHARS];
     int len = ld2string(buf,sizeof(buf),value,humanfriendly);
     return createStringObject(buf,len);
 }
@@ -232,6 +234,13 @@ robj *createZsetZiplistObject(void) {
     return o;
 }
 
+robj *createStreamObject(void) {
+    stream *s = streamNew();
+    robj *o = createObject(OBJ_STREAM,s);
+    o->encoding = OBJ_ENCODING_STREAM;
+    return o;
+}
+
 robj *createModuleObject(moduleType *mt, void *value) {
     moduleValue *mv = zmalloc(sizeof(*mv));
     mv->type = mt;
@@ -303,6 +312,10 @@ void freeModuleObject(robj *o) {
     zfree(mv);
 }
 
+void freeStreamObject(robj *o) {
+    freeStream(o->ptr);
+}
+
 void incrRefCount(robj *o) {
     if (o->refcount != OBJ_SHARED_REFCOUNT) o->refcount++;
 }
@@ -316,6 +329,7 @@ void decrRefCount(robj *o) {
         case OBJ_ZSET: freeZsetObject(o); break;
         case OBJ_HASH: freeHashObject(o); break;
         case OBJ_MODULE: freeModuleObject(o); break;
+        case OBJ_STREAM: freeStreamObject(o); break;
         default: serverPanic("Unknown object type"); break;
         }
         zfree(o);
@@ -560,7 +574,7 @@ int getDoubleFromObject(const robj *o, double *target) {
             value = strtod(o->ptr, &eptr);
             if (sdslen(o->ptr) == 0 ||
                 isspace(((const char*)o->ptr)[0]) ||
-                eptr[0] != '\0' ||
+                (size_t)(eptr-(char*)o->ptr) != sdslen(o->ptr) ||
                 (errno == ERANGE &&
                     (value == HUGE_VAL || value == -HUGE_VAL || value == 0)) ||
                 isnan(value))
@@ -602,7 +616,7 @@ int getLongDoubleFromObject(robj *o, long double *target) {
             value = strtold(o->ptr, &eptr);
             if (sdslen(o->ptr) == 0 ||
                 isspace(((const char*)o->ptr)[0]) ||
-                eptr[0] != '\0' ||
+                (size_t)(eptr-(char*)o->ptr) != sdslen(o->ptr) ||
                 (errno == ERANGE &&
                     (value == HUGE_VAL || value == -HUGE_VAL || value == 0)) ||
                 isnan(value))
@@ -727,7 +741,7 @@ size_t objectComputeSize(robj *o, size_t sample_size) {
                 elesize += sizeof(quicklistNode)+ziplistBlobLen(node->zl);
                 samples++;
             } while ((node = node->next) && samples < sample_size);
-            asize += (double)elesize/samples*listTypeLength(o);
+            asize += (double)elesize/samples*ql->len;
         } else if (o->encoding == OBJ_ENCODING_ZIPLIST) {
             asize = sizeof(*o)+ziplistBlobLen(o->ptr);
         } else {
@@ -788,6 +802,49 @@ size_t objectComputeSize(robj *o, size_t sample_size) {
         } else {
             serverPanic("Unknown hash encoding");
         }
+    } else if (o->type == OBJ_STREAM) {
+        stream *s = o->ptr;
+        /* Note: to guess the size of the radix tree is not trivial, so we
+         * approximate it considering 64 bytes of data overhead for each
+         * key (the ID), and then adding the number of bare nodes, plus some
+         * overhead due by the data and child pointers. This secret recipe
+         * was obtained by checking the average radix tree created by real
+         * workloads, and then adjusting the constants to get numbers that
+         * more or less match the real memory usage.
+         *
+         * Actually the number of nodes and keys may be different depending
+         * on the insertion speed and thus the ability of the radix tree
+         * to compress prefixes. */
+        asize = sizeof(*o);
+        asize += s->rax->numele * 64;
+        asize += s->rax->numnodes * sizeof(raxNode);
+        asize += s->rax->numnodes * 32*7; /* Add a few child pointers... */
+
+        /* Now we have to add the listpacks. The last listpack is often non
+         * complete, so we estimate the size of the first N listpacks, and
+         * use the average to compute the size of the first N-1 listpacks, and
+         * finally add the real size of the last node. */
+        raxIterator ri;
+        raxStart(&ri,s->rax);
+        raxSeek(&ri,"^",NULL,0);
+        size_t lpsize = 0, samples = 0;
+        while(samples < sample_size && raxNext(&ri)) {
+            unsigned char *lp = ri.data;
+            lpsize += lpBytes(lp);
+            samples++;
+        }
+        if (s->rax->numele <= samples) {
+            asize += lpsize;
+        } else {
+            if (samples) lpsize /= samples; /* Compute the average. */
+            asize += lpsize * (s->rax->numele-1);
+            /* No need to check if seek succeeded, we enter this branch only
+             * if there are a few elements in the radix tree. */
+            raxSeek(&ri,"$",NULL,0);
+            raxNext(&ri);
+            asize += lpBytes(ri.data);
+        }
+        raxStop(&ri);
     } else if (o->type == OBJ_MODULE) {
         moduleValue *mv = o->ptr;
         moduleType *mt = mv->type;
@@ -1012,11 +1069,20 @@ robj *objectCommandLookupOrReply(client *c, robj *key, robj *reply) {
 }
 
 /* Object command allows to inspect the internals of an Redis Object.
- * Usage: OBJECT <refcount|encoding|idletime> <key> */
+ * Usage: OBJECT <refcount|encoding|idletime|freq> <key> */
 void objectCommand(client *c) {
     robj *o;
 
-    if (!strcasecmp(c->argv[1]->ptr,"refcount") && c->argc == 3) {
+    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"help")) {
+        const char *help[] = {
+"encoding <key> -- Return the kind of internal representation used in order to store the value associated with a key.",
+"freq <key> -- Return the access frequency index of the key. The returned integer is proportional to the logarithm of the recent access frequency of the key.",
+"idletime <key> -- Return the idle time of the key, that is the approximated number of seconds elapsed since the last access to the key.",
+"refcount <key> -- Return the number of references of the value associated with the specified key.",
+NULL
+        };
+        addReplyHelp(c, help);
+    } else if (!strcasecmp(c->argv[1]->ptr,"refcount") && c->argc == 3) {
         if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.nullbulk))
                 == NULL) return;
         addReplyLongLong(c,o->refcount);
@@ -1035,13 +1101,17 @@ void objectCommand(client *c) {
     } else if (!strcasecmp(c->argv[1]->ptr,"freq") && c->argc == 3) {
         if ((o = objectCommandLookupOrReply(c,c->argv[2],shared.nullbulk))
                 == NULL) return;
-        if (server.maxmemory_policy & MAXMEMORY_FLAG_LRU) {
-            addReplyError(c,"An LRU maxmemory policy is selected, access frequency not tracked. Please note that when switching between policies at runtime LRU and LFU data will take some time to adjust.");
+        if (!(server.maxmemory_policy & MAXMEMORY_FLAG_LFU)) {
+            addReplyError(c,"An LFU maxmemory policy is not selected, access frequency not tracked. Please note that when switching between policies at runtime LRU and LFU data will take some time to adjust.");
             return;
         }
-        addReplyLongLong(c,o->lru&255);
+        /* LFUDecrAndReturn should be called
+         * in case of the key has not been accessed for a long time,
+         * because we update the access time only
+         * when the key is read or overwritten. */
+        addReplyLongLong(c,LFUDecrAndReturn(o));
     } else {
-        addReplyError(c,"Syntax error. Try OBJECT (refcount|encoding|idletime|freq)");
+        addReplyErrorFormat(c, "Unknown subcommand or wrong number of arguments for '%s'. Try OBJECT help", (char *)c->argv[1]->ptr);
     }
 }
 

src/pubsub.c

@@ -325,8 +325,16 @@ void publishCommand(client *c) {
 
 /* PUBSUB command for Pub/Sub introspection. */
 void pubsubCommand(client *c) {
-    if (!strcasecmp(c->argv[1]->ptr,"channels") &&
-        (c->argc == 2 || c->argc ==3))
+    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"help")) {
+        const char *help[] = {
+"channels [<pattern>] -- Return the currently active channels matching a pattern (default: all).",
+"numpat -- Return number of subscriptions to patterns.",
+"numsub [channel-1 .. channel-N] -- Returns the number of subscribers for the specified channels (excluding patterns, default: none).",
+NULL
+        };
+        addReplyHelp(c, help);
+    } else if (!strcasecmp(c->argv[1]->ptr,"channels") &&
+        (c->argc == 2 || c->argc == 3))
     {
         /* PUBSUB CHANNELS [<pattern>] */
         sds pat = (c->argc == 2) ? NULL : c->argv[2]->ptr;
@@ -364,8 +372,7 @@ void pubsubCommand(client *c) {
         /* PUBSUB NUMPAT */
         addReplyLongLong(c,listLength(server.pubsub_patterns));
     } else {
-        addReplyErrorFormat(c,
-            "Unknown PUBSUB subcommand or wrong number of arguments for '%s'",
+        addReplyErrorFormat(c, "Unknown subcommand or wrong number of arguments for '%s'. Try PUBSUB HELP",
             (char*)c->argv[1]->ptr);
     }
 }

src/quicklist.c

@@ -149,7 +149,7 @@ REDIS_STATIC quicklistNode *quicklistCreateNode(void) {
 }
 
 /* Return cached quicklist count */
-unsigned int quicklistCount(const quicklist *ql) { return ql->count; }
+unsigned long quicklistCount(const quicklist *ql) { return ql->count; }
 
 /* Free entire quicklist. */
 void quicklistRelease(quicklist *quicklist) {

src/quicklist.h

@@ -64,7 +64,7 @@ typedef struct quicklistLZF {
     char compressed[];
 } quicklistLZF;
 
-/* quicklist is a 32 byte struct (on 64-bit systems) describing a quicklist.
+/* quicklist is a 40 byte struct (on 64-bit systems) describing a quicklist.
  * 'count' is the number of total entries.
  * 'len' is the number of quicklist nodes.
  * 'compress' is: -1 if compression disabled, otherwise it's the number
@@ -74,7 +74,7 @@ typedef struct quicklist {
     quicklistNode *head;
     quicklistNode *tail;
     unsigned long count;        /* total count of all entries in all ziplists */
-    unsigned int len;           /* number of quicklistNodes */
+    unsigned long len;          /* number of quicklistNodes */
     int fill : 16;              /* fill factor for individual nodes */
     unsigned int compress : 16; /* depth of end nodes not to compress;0=off */
 } quicklist;
@@ -154,7 +154,7 @@ int quicklistPopCustom(quicklist *quicklist, int where, unsigned char **data,
                        void *(*saver)(unsigned char *data, unsigned int sz));
 int quicklistPop(quicklist *quicklist, int where, unsigned char **data,
                  unsigned int *sz, long long *slong);
-unsigned int quicklistCount(const quicklist *ql);
+unsigned long quicklistCount(const quicklist *ql);
 int quicklistCompare(unsigned char *p1, unsigned char *p2, int p2_len);
 size_t quicklistGetLzf(const quicklistNode *node, void **data);
 

src/rax.c

@@ -131,7 +131,7 @@ static inline void raxStackFree(raxStack *ts) {
 }
 
 /* ----------------------------------------------------------------------------
- * Radis tree implementation
+ * Radix tree implementation
  * --------------------------------------------------------------------------*/
 
 /* Allocate a new non compressed node with the specified number of children.
@@ -873,7 +873,8 @@ raxNode *raxRemoveChild(raxNode *parent, raxNode *child) {
     memmove(((char*)cp)-1,cp,(parent->size-taillen-1)*sizeof(raxNode**));
 
     /* Move the remaining "tail" pointer at the right position as well. */
-    memmove(((char*)c)-1,c+1,taillen*sizeof(raxNode**)+parent->iskey*sizeof(void*));
+    size_t valuelen = (parent->iskey && !parent->isnull) ? sizeof(void*) : 0;
+    memmove(((char*)c)-1,c+1,taillen*sizeof(raxNode**)+valuelen);
 
     /* 4. Update size. */
     parent->size--;
@@ -1092,28 +1093,36 @@ int raxRemove(rax *rax, unsigned char *s, size_t len, void **old) {
 
 /* This is the core of raxFree(): performs a depth-first scan of the
  * tree and releases all the nodes found. */
-void raxRecursiveFree(rax *rax, raxNode *n) {
+void raxRecursiveFree(rax *rax, raxNode *n, void (*free_callback)(void*)) {
     debugnode("free traversing",n);
     int numchildren = n->iscompr ? 1 : n->size;
     raxNode **cp = raxNodeLastChildPtr(n);
     while(numchildren--) {
         raxNode *child;
         memcpy(&child,cp,sizeof(child));
-        raxRecursiveFree(rax,child);
+        raxRecursiveFree(rax,child,free_callback);
         cp--;
     }
     debugnode("free depth-first",n);
+    if (free_callback && n->iskey && !n->isnull)
+        free_callback(raxGetData(n));
     rax_free(n);
     rax->numnodes--;
 }
 
-/* Free a whole radix tree. */
-void raxFree(rax *rax) {
-    raxRecursiveFree(rax,rax->head);
+/* Free a whole radix tree, calling the specified callback in order to
+ * free the auxiliary data. */
+void raxFreeWithCallback(rax *rax, void (*free_callback)(void*)) {
+    raxRecursiveFree(rax,rax->head,free_callback);
     assert(rax->numnodes == 0);
     rax_free(rax);
 }
 
+/* Free a whole radix tree. */
+void raxFree(rax *rax) {
+    raxFreeWithCallback(rax,NULL);
+}
+
 /* ------------------------------- Iterator --------------------------------- */
 
 /* Initialize a Rax iterator. This call should be performed a single time
@@ -1175,7 +1184,7 @@ void raxIteratorDelChars(raxIterator *it, size_t count) {
  * The function returns 1 on success or 0 on out of memory. */
 int raxIteratorNextStep(raxIterator *it, int noup) {
     if (it->flags & RAX_ITER_EOF) {
-        return 0;
+        return 1;
     } else if (it->flags & RAX_ITER_JUST_SEEKED) {
         it->flags &= ~RAX_ITER_JUST_SEEKED;
         return 1;
@@ -1187,10 +1196,6 @@ int raxIteratorNextStep(raxIterator *it, int noup) {
     size_t orig_stack_items = it->stack.items;
     raxNode *orig_node = it->node;
 
-    /* Clear the EOF flag: it will be set again if the EOF condition
-     * is still valid. */
-    it->flags &= ~RAX_ITER_EOF;
-
     while(1) {
         int children = it->node->iscompr ? 1 : it->node->size;
         if (!noup && children) {
@@ -1291,7 +1296,7 @@ int raxSeekGreatest(raxIterator *it) {
  * effect to the one of raxIteratorPrevSte(). */
 int raxIteratorPrevStep(raxIterator *it, int noup) {
     if (it->flags & RAX_ITER_EOF) {
-        return 0;
+        return 1;
     } else if (it->flags & RAX_ITER_JUST_SEEKED) {
         it->flags &= ~RAX_ITER_JUST_SEEKED;
         return 1;
@@ -1412,6 +1417,7 @@ int raxSeek(raxIterator *it, const char *op, unsigned char *ele, size_t len) {
         it->node = it->rt->head;
         if (!raxSeekGreatest(it)) return 0;
         assert(it->node->iskey);
+        it->data = raxGetData(it->node);
         return 1;
     }
 
@@ -1430,6 +1436,7 @@ int raxSeek(raxIterator *it, const char *op, unsigned char *ele, size_t len) {
         /* We found our node, since the key matches and we have an
          * "equal" condition. */
         if (!raxIteratorAddChars(it,ele,len)) return 0; /* OOM. */
+        it->data = raxGetData(it->node);
     } else if (lt || gt) {
         /* Exact key not found or eq flag not set. We have to set as current
          * key the one represented by the node we stopped at, and perform
@@ -1502,6 +1509,7 @@ int raxSeek(raxIterator *it, const char *op, unsigned char *ele, size_t len) {
                  * the previous sub-tree. */
                 if (nodechar < keychar) {
                     if (!raxSeekGreatest(it)) return 0;
+                    it->data = raxGetData(it->node);
                 } else {
                     if (!raxIteratorAddChars(it,it->node->data,it->node->size))
                         return 0;
@@ -1618,8 +1626,8 @@ int raxCompare(raxIterator *iter, const char *op, unsigned char *key, size_t key
     int eq = 0, lt = 0, gt = 0;
 
     if (op[0] == '=' || op[1] == '=') eq = 1;
-    if (op[1] == '>') gt = 1;
-    else if (op[1] == '<') lt = 1;
+    if (op[0] == '>') gt = 1;
+    else if (op[0] == '<') lt = 1;
     else if (op[1] != '=') return 0; /* Syntax error. */
 
     size_t minlen = key_len < iter->key_len ? key_len : iter->key_len;
@@ -1647,6 +1655,19 @@ void raxStop(raxIterator *it) {
     raxStackFree(&it->stack);
 }
 
+/* Return if the iterator is in an EOF state. This happens when raxSeek()
+ * failed to seek an appropriate element, so that raxNext() or raxPrev()
+ * will return zero, or when an EOF condition was reached while iterating
+ * with raxNext() and raxPrev(). */
+int raxEOF(raxIterator *it) {
+    return it->flags & RAX_ITER_EOF;
+}
+
+/* Return the number of elements inside the radix tree. */
+uint64_t raxSize(rax *rax) {
+    return rax->numele;
+}
+
 /* ----------------------------- Introspection ------------------------------ */
 
 /* This function is mostly used for debugging and learning purposes.

src/rax.h

@@ -148,6 +148,7 @@ int raxInsert(rax *rax, unsigned char *s, size_t len, void *data, void **old);
 int raxRemove(rax *rax, unsigned char *s, size_t len, void **old);
 void *raxFind(rax *rax, unsigned char *s, size_t len);
 void raxFree(rax *rax);
+void raxFreeWithCallback(rax *rax, void (*free_callback)(void*));
 void raxStart(raxIterator *it, rax *rt);
 int raxSeek(raxIterator *it, const char *op, unsigned char *ele, size_t len);
 int raxNext(raxIterator *it);
@@ -155,6 +156,8 @@ int raxPrev(raxIterator *it);
 int raxRandomWalk(raxIterator *it, size_t steps);
 int raxCompare(raxIterator *iter, const char *op, unsigned char *key, size_t key_len);
 void raxStop(raxIterator *it);
+int raxEOF(raxIterator *it);
 void raxShow(rax *rax);
+uint64_t raxSize(rax *rax);
 
 #endif

src/rdb.c

@@ -31,6 +31,7 @@
 #include "lzf.h"    /* LZF compression library */
 #include "zipmap.h"
 #include "endianconv.h"
+#include "stream.h"
 
 #include <math.h>
 #include <sys/types.h>
@@ -254,7 +255,7 @@ void *rdbLoadIntegerObject(rio *rdb, int enctype, int flags, size_t *lenptr) {
         char buf[LONG_STR_SIZE], *p;
         int len = ll2string(buf,sizeof(buf),val);
         if (lenptr) *lenptr = len;
-        p = plain ? zmalloc(len) : sdsnewlen(NULL,len);
+        p = plain ? zmalloc(len) : sdsnewlen(SDS_NOINIT,len);
         memcpy(p,buf,len);
         return p;
     } else if (encode) {
@@ -343,10 +344,10 @@ void *rdbLoadLzfStringObject(rio *rdb, int flags, size_t *lenptr) {
     /* Allocate our target according to the uncompressed size. */
     if (plain) {
         val = zmalloc(len);
-        if (lenptr) *lenptr = len;
     } else {
-        val = sdsnewlen(NULL,len);
+        val = sdsnewlen(SDS_NOINIT,len);
     }
+    if (lenptr) *lenptr = len;
 
     /* Load the compressed representation and uncompress it to target. */
     if (rioRead(rdb,c,clen) == 0) goto err;
@@ -424,7 +425,7 @@ ssize_t rdbSaveLongLongAsStringObject(rio *rdb, long long value) {
 }
 
 /* Like rdbSaveRawString() gets a Redis object instead. */
-int rdbSaveStringObject(rio *rdb, robj *obj) {
+ssize_t rdbSaveStringObject(rio *rdb, robj *obj) {
     /* Avoid to decode the object, then encode it again, if the
      * object is already integer encoded. */
     if (obj->encoding == OBJ_ENCODING_INT) {
@@ -471,7 +472,7 @@ void *rdbGenericLoadStringObject(rio *rdb, int flags, size_t *lenptr) {
 
     if (len == RDB_LENERR) return NULL;
     if (plain || sds) {
-        void *buf = plain ? zmalloc(len) : sdsnewlen(NULL,len);
+        void *buf = plain ? zmalloc(len) : sdsnewlen(SDS_NOINIT,len);
         if (lenptr) *lenptr = len;
         if (len && rioRead(rdb,buf,len) == 0) {
             if (plain)
@@ -482,8 +483,8 @@ void *rdbGenericLoadStringObject(rio *rdb, int flags, size_t *lenptr) {
         }
         return buf;
     } else {
-        robj *o = encode ? createStringObject(NULL,len) :
-                           createRawStringObject(NULL,len);
+        robj *o = encode ? createStringObject(SDS_NOINIT,len) :
+                           createRawStringObject(SDS_NOINIT,len);
         if (len && rioRead(rdb,o->ptr,len) == 0) {
             decrRefCount(o);
             return NULL;
@@ -622,6 +623,8 @@ int rdbSaveObjectType(rio *rdb, robj *o) {
             return rdbSaveType(rdb,RDB_TYPE_HASH);
         else
             serverPanic("Unknown hash encoding");
+    case OBJ_STREAM:
+        return rdbSaveType(rdb,RDB_TYPE_STREAM_LISTPACKS);
     case OBJ_MODULE:
         return rdbSaveType(rdb,RDB_TYPE_MODULE_2);
     default:
@@ -762,7 +765,39 @@ ssize_t rdbSaveObject(rio *rdb, robj *o) {
         } else {
             serverPanic("Unknown hash encoding");
         }
+    } else if (o->type == OBJ_STREAM) {
+        /* Store how many listpacks we have inside the radix tree. */
+        stream *s = o->ptr;
+        rax *rax = s->rax;
+        if ((n = rdbSaveLen(rdb,raxSize(rax))) == -1) return -1;
+        nwritten += n;
 
+        /* Serialize all the listpacks inside the radix tree as they are,
+         * when loading back, we'll use the first entry of each listpack
+         * to insert it back into the radix tree. */
+        raxIterator ri;
+        raxStart(&ri,rax);
+        raxSeek(&ri,"^",NULL,0);
+        while (raxNext(&ri)) {
+            unsigned char *lp = ri.data;
+            size_t lp_bytes = lpBytes(lp);
+            if ((n = rdbSaveRawString(rdb,ri.key,ri.key_len)) == -1) return -1;
+            nwritten += n;
+            if ((n = rdbSaveRawString(rdb,lp,lp_bytes)) == -1) return -1;
+            nwritten += n;
+        }
+        raxStop(&ri);
+
+        /* Save the number of elements inside the stream. We cannot obtain
+         * this easily later, since our macro nodes should be checked for
+         * number of items: not a great CPU / space tradeoff. */
+        if ((n = rdbSaveLen(rdb,s->length)) == -1) return -1;
+        nwritten += n;
+        /* Save the last entry ID. */
+        if ((n = rdbSaveLen(rdb,s->last_id.ms)) == -1) return -1;
+        nwritten += n;
+        if ((n = rdbSaveLen(rdb,s->last_id.seq)) == -1) return -1;
+        nwritten += n;
     } else if (o->type == OBJ_MODULE) {
         /* Save a module-specific value. */
         RedisModuleIO io;
@@ -826,21 +861,25 @@ int rdbSaveKeyValuePair(rio *rdb, robj *key, robj *val,
 }
 
 /* Save an AUX field. */
-int rdbSaveAuxField(rio *rdb, void *key, size_t keylen, void *val, size_t vallen) {
-    if (rdbSaveType(rdb,RDB_OPCODE_AUX) == -1) return -1;
-    if (rdbSaveRawString(rdb,key,keylen) == -1) return -1;
-    if (rdbSaveRawString(rdb,val,vallen) == -1) return -1;
-    return 1;
+ssize_t rdbSaveAuxField(rio *rdb, void *key, size_t keylen, void *val, size_t vallen) {
+    ssize_t ret, len = 0;
+    if ((ret = rdbSaveType(rdb,RDB_OPCODE_AUX)) == -1) return -1;
+    len += ret;
+    if ((ret = rdbSaveRawString(rdb,key,keylen) == -1)) return -1;
+    len += ret;
+    if ((ret = rdbSaveRawString(rdb,val,vallen) == -1)) return -1;
+    len += ret;
+    return len;
 }
 
 /* Wrapper for rdbSaveAuxField() used when key/val length can be obtained
  * with strlen(). */
-int rdbSaveAuxFieldStrStr(rio *rdb, char *key, char *val) {
+ssize_t rdbSaveAuxFieldStrStr(rio *rdb, char *key, char *val) {
     return rdbSaveAuxField(rdb,key,strlen(key),val,strlen(val));
 }
 
 /* Wrapper for strlen(key) + integer type (up to long long range). */
-int rdbSaveAuxFieldStrInt(rio *rdb, char *key, long long val) {
+ssize_t rdbSaveAuxFieldStrInt(rio *rdb, char *key, long long val) {
     char buf[LONG_STR_SIZE];
     int vlen = ll2string(buf,sizeof(buf),val);
     return rdbSaveAuxField(rdb,key,strlen(key),buf,vlen);
@@ -943,6 +982,20 @@ int rdbSaveRio(rio *rdb, int *error, int flags, rdbSaveInfo *rsi) {
     }
     di = NULL; /* So that we don't release it again on error. */
 
+    /* If we are storing the replication information on disk, persist
+     * the script cache as well: on successful PSYNC after a restart, we need
+     * to be able to process any EVALSHA inside the replication backlog the
+     * master will send us. */
+    if (rsi && dictSize(server.lua_scripts)) {
+        di = dictGetIterator(server.lua_scripts);
+        while((de = dictNext(di)) != NULL) {
+            robj *body = dictGetVal(de);
+            if (rdbSaveAuxField(rdb,"lua",3,body->ptr,sdslen(body->ptr)) == -1)
+                goto werr;
+        }
+        dictReleaseIterator(di);
+    }
+
     /* EOF opcode */
     if (rdbSaveType(rdb,RDB_OPCODE_EOF) == -1) goto werr;
 
@@ -1395,6 +1448,45 @@ robj *rdbLoadObject(int rdbtype, rio *rdb) {
                 rdbExitReportCorruptRDB("Unknown RDB encoding type %d",rdbtype);
                 break;
         }
+    } else if (rdbtype == RDB_TYPE_STREAM_LISTPACKS) {
+        o = createStreamObject();
+        stream *s = o->ptr;
+        uint64_t listpacks = rdbLoadLen(rdb,NULL);
+
+        while(listpacks--) {
+            /* Get the master ID, the one we'll use as key of the radix tree
+             * node: the entries inside the listpack itself are delta-encoded
+             * relatively to this ID. */
+            sds nodekey = rdbGenericLoadStringObject(rdb,RDB_LOAD_SDS,NULL);
+            if (sdslen(nodekey) != sizeof(streamID)) {
+                rdbExitReportCorruptRDB("Stream node key entry is not the "
+                                        "size of a stream ID");
+            }
+
+            /* Load the listpack. */
+            unsigned char *lp =
+                rdbGenericLoadStringObject(rdb,RDB_LOAD_PLAIN,NULL);
+            if (lp == NULL) return NULL;
+            unsigned char *first = lpFirst(lp);
+            if (first == NULL) {
+                /* Serialized listpacks should never be empty, since on
+                 * deletion we should remove the radix tree key if the
+                 * resulting listpack is emtpy. */
+                rdbExitReportCorruptRDB("Empty listpack inside stream");
+            }
+
+            /* Insert the key in the radix tree. */
+            int retval = raxInsert(s->rax,
+                (unsigned char*)nodekey,sizeof(streamID),lp,NULL);
+            sdsfree(nodekey);
+            if (!retval)
+                rdbExitReportCorruptRDB("Listpack re-added with existing key");
+        }
+        /* Load total number of items inside the stream. */
+        s->length = rdbLoadLen(rdb,NULL);
+        /* Load the last entry ID. */
+        s->last_id.ms = rdbLoadLen(rdb,NULL);
+        s->last_id.seq = rdbLoadLen(rdb,NULL);
     } else if (rdbtype == RDB_TYPE_MODULE || rdbtype == RDB_TYPE_MODULE_2) {
         uint64_t moduleid = rdbLoadLen(rdb,NULL);
         moduleType *mt = moduleTypeLookupModuleByID(moduleid);
@@ -1589,6 +1681,13 @@ int rdbLoadRio(rio *rdb, rdbSaveInfo *rsi) {
                 }
             } else if (!strcasecmp(auxkey->ptr,"repl-offset")) {
                 if (rsi) rsi->repl_offset = strtoll(auxval->ptr,NULL,10);
+            } else if (!strcasecmp(auxkey->ptr,"lua")) {
+                /* Load the script back in memory. */
+                if (luaCreateFunction(NULL,server.lua,auxval) == NULL) {
+                    rdbExitReportCorruptRDB(
+                        "Can't load Lua script from RDB file! "
+                        "BODY: %s", auxval->ptr);
+                }
             } else {
                 /* We ignore fields we don't understand, as by AUX field
                  * contract. */
@@ -2000,6 +2099,9 @@ void bgsaveCommand(client *c) {
         }
     }
 
+    rdbSaveInfo rsi, *rsiptr;
+    rsiptr = rdbPopulateSaveInfo(&rsi);
+
     if (server.rdb_child_pid != -1) {
         addReplyError(c,"Background save already in progress");
     } else if (server.aof_child_pid != -1) {
@@ -2012,7 +2114,7 @@ void bgsaveCommand(client *c) {
                 "Use BGSAVE SCHEDULE in order to schedule a BGSAVE whenever "
                 "possible.");
         }
-    } else if (rdbSaveBackground(server.rdb_filename,NULL) == C_OK) {
+    } else if (rdbSaveBackground(server.rdb_filename,rsiptr) == C_OK) {
         addReplyStatus(c,"Background saving started");
     } else {
         addReply(c,shared.err);
@@ -2033,22 +2135,37 @@ rdbSaveInfo *rdbPopulateSaveInfo(rdbSaveInfo *rsi) {
     *rsi = rsi_init;
 
     /* If the instance is a master, we can populate the replication info
-     * in all the cases, even if sometimes in incomplete (but safe) form. */
-    if (!server.masterhost) {
-        if (server.repl_backlog) rsi->repl_stream_db = server.slaveseldb;
-        /* Note that if repl_backlog is NULL, it means that histories
-         * following from this point will trigger a full synchronization
-         * generating a SELECT statement, so we can leave the currently
-         * selected DB set to -1. This allows a restarted master to reload
-         * its replication ID/offset when there are no connected slaves. */
+     * only when repl_backlog is not NULL. If the repl_backlog is NULL,
+     * it means that the instance isn't in any replication chains. In this
+     * scenario the replication info is useless, because when a slave
+     * connects to us, the NULL repl_backlog will trigger a full
+     * synchronization, at the same time we will use a new replid and clear
+     * replid2. */
+    if (!server.masterhost && server.repl_backlog) {
+        /* Note that when server.slaveseldb is -1, it means that this master
+         * didn't apply any write commands after a full synchronization.
+         * So we can let repl_stream_db be 0, this allows a restarted slave
+         * to reload replication ID/offset, it's safe because the next write
+         * command must generate a SELECT statement. */
+        rsi->repl_stream_db = server.slaveseldb == -1 ? 0 : server.slaveseldb;
         return rsi;
     }
 
-    /* If the instance is a slave we need a connected master in order to
-     * fetch the currently selected DB. */
+    /* If the instance is a slave we need a connected master
+     * in order to fetch the currently selected DB. */
     if (server.master) {
         rsi->repl_stream_db = server.master->db->id;
         return rsi;
     }
+
+    /* If we have a cached master we can use it in order to populate the
+     * replication selected DB info inside the RDB file: the slave can
+     * increment the master_repl_offset only from data arriving from the
+     * master, so if we are disconnected the offset in the cached master
+     * is valid. */
+    if (server.cached_master) {
+        rsi->repl_stream_db = server.cached_master->db->id;
+        return rsi;
+    }
     return NULL;
 }

src/rdb.h

@@ -69,8 +69,9 @@
 #define RDB_ENC_INT32 2       /* 32 bit signed integer */
 #define RDB_ENC_LZF 3         /* string compressed with FASTLZ */
 
-/* Dup object types to RDB object types. Only reason is readability (are we
- * dealing with RDB types or with in-memory object types?). */
+/* Map object types to RDB object types. Macros starting with OBJ_ are for
+ * memory storage and may change. Instead RDB types must be fixed because
+ * we store them on disk. */
 #define RDB_TYPE_STRING 0
 #define RDB_TYPE_LIST   1
 #define RDB_TYPE_SET    2
@@ -89,10 +90,11 @@
 #define RDB_TYPE_ZSET_ZIPLIST  12
 #define RDB_TYPE_HASH_ZIPLIST  13
 #define RDB_TYPE_LIST_QUICKLIST 14
+#define RDB_TYPE_STREAM_LISTPACKS 15
 /* NOTE: WHEN ADDING NEW RDB TYPE, UPDATE rdbIsObjectType() BELOW */
 
 /* Test if a type is an object type. */
-#define rdbIsObjectType(t) ((t >= 0 && t <= 7) || (t >= 9 && t <= 14))
+#define rdbIsObjectType(t) ((t >= 0 && t <= 7) || (t >= 9 && t <= 15))
 
 /* Special RDB opcodes (saved/loaded with rdbSaveType/rdbLoadType). */
 #define RDB_OPCODE_AUX        250
@@ -139,7 +141,7 @@ robj *rdbLoadObject(int type, rio *rdb);
 void backgroundSaveDoneHandler(int exitcode, int bysignal);
 int rdbSaveKeyValuePair(rio *rdb, robj *key, robj *val, long long expiretime, long long now);
 robj *rdbLoadStringObject(rio *rdb);
-int rdbSaveStringObject(rio *rdb, robj *obj);
+ssize_t rdbSaveStringObject(rio *rdb, robj *obj);
 ssize_t rdbSaveRawString(rio *rdb, unsigned char *s, size_t len);
 void *rdbGenericLoadStringObject(rio *rdb, int flags, size_t *lenptr);
 int rdbSaveBinaryDoubleValue(rio *rdb, double val);

src/redis-benchmark.c

@@ -614,7 +614,7 @@ int showThroughput(struct aeEventLoop *eventLoop, long long id, void *clientData
     UNUSED(id);
     UNUSED(clientData);
 
-    if (config.liveclients == 0) {
+    if (config.liveclients == 0 && config.requests_finished != config.requests) {
         fprintf(stderr,"All clients disconnected... aborting.\n");
         exit(1);
     }

src/redis-check-rdb.c

@@ -193,12 +193,12 @@ int redis_check_rdb(char *rdbfilename, FILE *fp) {
     buf[9] = '\0';
     if (memcmp(buf,"REDIS",5) != 0) {
         rdbCheckError("Wrong signature trying to load DB from file");
-        return 1;
+        goto err;
     }
     rdbver = atoi(buf+5);
     if (rdbver < 1 || rdbver > RDB_VERSION) {
         rdbCheckError("Can't handle RDB format version %d",rdbver);
-        return 1;
+        goto err;
     }
 
     startLoading(fp);
@@ -270,7 +270,7 @@ int redis_check_rdb(char *rdbfilename, FILE *fp) {
         } else {
             if (!rdbIsObjectType(type)) {
                 rdbCheckError("Invalid object type: %d", type);
-                return 1;
+                goto err;
             }
             rdbstate.key_type = type;
         }
@@ -307,6 +307,7 @@ int redis_check_rdb(char *rdbfilename, FILE *fp) {
             rdbCheckInfo("RDB file was saved with checksum disabled: no check performed.");
         } else if (cksum != expected) {
             rdbCheckError("RDB CRC error");
+            goto err;
         } else {
             rdbCheckInfo("Checksum OK");
         }
@@ -321,6 +322,8 @@ eoferr: /* unexpected end of file is handled here with a fatal exit */
     } else {
         rdbCheckError("Unexpected EOF reading RDB file");
     }
+err:
+    if (closefile) fclose(fp);
     return 1;
 }
 

src/redis-cli.c

@@ -107,6 +107,7 @@ static struct config {
     char *pattern;
     char *rdb_filename;
     int bigkeys;
+    int hotkeys;
     int stdinarg; /* get last arg from stdin. (-x option) */
     char *auth;
     int output; /* output mode, see OUTPUT_* defines */
@@ -198,6 +199,92 @@ static sds getDotfilePath(char *envoverride, char *dotfilename) {
     return dotPath;
 }
 
+/* URL-style percent decoding. */
+#define isHexChar(c) (isdigit(c) || (c >= 'a' && c <= 'f'))
+#define decodeHexChar(c) (isdigit(c) ? c - '0' : c - 'a' + 10)
+#define decodeHex(h, l) ((decodeHexChar(h) << 4) + decodeHexChar(l))
+
+static sds percentDecode(const char *pe, size_t len) {
+    const char *end = pe + len;
+    sds ret = sdsempty();
+    const char *curr = pe;
+
+    while (curr < end) {
+        if (*curr == '%') {
+            if ((end - curr) < 2) {
+                fprintf(stderr, "Incomplete URI encoding\n");
+                exit(1);
+            }
+
+            char h = tolower(*(++curr));
+            char l = tolower(*(++curr));
+            if (!isHexChar(h) || !isHexChar(l)) {
+                fprintf(stderr, "Illegal character in URI encoding\n");
+                exit(1);
+            }
+            char c = decodeHex(h, l);
+            ret = sdscatlen(ret, &c, 1);
+            curr++;
+        } else {
+            ret = sdscatlen(ret, curr++, 1);
+        }
+    }
+
+    return ret;
+}
+
+/* Parse a URI and extract the server connection information.
+ * URI scheme is based on the the provisional specification[1] excluding support
+ * for query parameters. Valid URIs are:
+ *   scheme:    "redis://"
+ *   authority: [<username> ":"] <password> "@"] [<hostname> [":" <port>]]
+ *   path:      ["/" [<db>]]
+ *
+ *  [1]: https://www.iana.org/assignments/uri-schemes/prov/redis */
+static void parseRedisUri(const char *uri) {
+
+    const char *scheme = "redis://";
+    const char *curr = uri;
+    const char *end = uri + strlen(uri);
+    const char *userinfo, *username, *port, *host, *path;
+
+    /* URI must start with a valid scheme. */
+    if (strncasecmp(scheme, curr, strlen(scheme))) {
+        fprintf(stderr,"Invalid URI scheme\n");
+        exit(1);
+    }
+    curr += strlen(scheme);
+    if (curr == end) return;
+
+    /* Extract user info. */
+    if ((userinfo = strchr(curr,'@'))) {
+        if ((username = strchr(curr, ':')) && username < userinfo) {
+            /* If provided, username is ignored. */
+            curr = username + 1;
+        }
+
+        config.auth = percentDecode(curr, userinfo - curr);
+        curr = userinfo + 1;
+    }
+    if (curr == end) return;
+
+    /* Extract host and port. */
+    path = strchr(curr, '/');
+    if (*curr != '/') {
+        host = path ? path - 1 : end;
+        if ((port = strchr(curr, ':'))) {
+            config.hostport = atoi(port + 1);
+            host = port - 1;
+        }
+        config.hostip = sdsnewlen(curr, host - curr + 1);
+    }
+    curr = path ? path + 1 : end;
+    if (curr == end) return;
+
+    /* Extract database number. */
+    config.dbnum = atoi(curr);
+}
+
 /*------------------------------------------------------------------------------
  * Help functions
  *--------------------------------------------------------------------------- */
@@ -624,7 +711,7 @@ int isColorTerm(void) {
     return t != NULL && strstr(t,"xterm") != NULL;
 }
 
-/* Helpe  function for sdsCatColorizedLdbReply() appending colorize strings
+/* Helper  function for sdsCatColorizedLdbReply() appending colorize strings
  * to an SDS string. */
 sds sdscatcolor(sds o, char *s, size_t len, char *color) {
     if (!isColorTerm()) return sdscatlen(o,s,len);
@@ -1002,6 +1089,8 @@ static int parseOptions(int argc, char **argv) {
             config.dbnum = atoi(argv[++i]);
         } else if (!strcmp(argv[i],"-a") && !lastarg) {
             config.auth = argv[++i];
+        } else if (!strcmp(argv[i],"-u") && !lastarg) {
+            parseRedisUri(argv[++i]);
         } else if (!strcmp(argv[i],"--raw")) {
             config.output = OUTPUT_RAW;
         } else if (!strcmp(argv[i],"--no-raw")) {
@@ -1041,6 +1130,8 @@ static int parseOptions(int argc, char **argv) {
             config.pipe_timeout = atoi(argv[++i]);
         } else if (!strcmp(argv[i],"--bigkeys")) {
             config.bigkeys = 1;
+        } else if (!strcmp(argv[i],"--hotkeys")) {
+            config.hotkeys = 1;
         } else if (!strcmp(argv[i],"--eval") && !lastarg) {
             config.eval = argv[++i];
         } else if (!strcmp(argv[i],"--ldb")) {
@@ -1109,6 +1200,7 @@ static void usage(void) {
 "  -p <port>          Server port (default: 6379).\n"
 "  -s <socket>        Server socket (overrides hostname and port).\n"
 "  -a <password>      Password to use when connecting to the server.\n"
+"  -u <uri>           Server URI.\n"
 "  -r <repeat>        Execute specified command N times.\n"
 "  -i <interval>      When -r is used, waits <interval> seconds per command.\n"
 "                     It is possible to specify sub-second times like -i 0.1.\n"
@@ -1140,6 +1232,8 @@ static void usage(void) {
 "                     no reply is received within <n> seconds.\n"
 "                     Default timeout: %d. Use 0 to wait forever.\n"
 "  --bigkeys          Sample Redis keys looking for big keys.\n"
+"  --hotkeys          Sample Redis keys looking for hot keys.\n"
+"                     only works when maxmemory-policy is *lfu.\n"
 "  --scan             List all keys using the SCAN command.\n"
 "  --pattern <pat>    Useful with --scan to specify a SCAN pattern.\n"
 "  --intrinsic-latency <sec> Run a test to measure intrinsic system latency.\n"
@@ -1292,8 +1386,9 @@ static void repl(void) {
     /* Only use history and load the rc file when stdin is a tty. */
     if (isatty(fileno(stdin))) {
         historyfile = getDotfilePath(REDIS_CLI_HISTFILE_ENV,REDIS_CLI_HISTFILE_DEFAULT);
+        //keep in-memory history always regardless if history file can be determined
+        history = 1;
         if (historyfile != NULL) {
-            history = 1;
             linenoiseHistoryLoad(historyfile);
         }
         cliLoadPreferences();
@@ -1980,7 +2075,8 @@ static void pipeMode(void) {
 #define TYPE_SET    2
 #define TYPE_HASH   3
 #define TYPE_ZSET   4
-#define TYPE_NONE   5
+#define TYPE_STREAM 5
+#define TYPE_NONE   6
 
 static redisReply *sendScan(unsigned long long *it) {
     redisReply *reply = redisCommand(context, "SCAN %llu", *it);
@@ -2039,6 +2135,8 @@ static int toIntType(char *key, char *type) {
         return TYPE_HASH;
     } else if(!strcmp(type, "zset")) {
         return TYPE_ZSET;
+    } else if(!strcmp(type, "stream")) {
+        return TYPE_STREAM;
     } else if(!strcmp(type, "none")) {
         return TYPE_NONE;
     } else {
@@ -2127,7 +2225,7 @@ static void findBigKeys(void) {
     unsigned long long biggest[5] = {0}, counts[5] = {0}, totalsize[5] = {0};
     unsigned long long sampled = 0, total_keys, totlen=0, *sizes=NULL, it=0;
     sds maxkeys[5] = {0};
-    char *typename[] = {"string","list","set","hash","zset"};
+    char *typename[] = {"string","list","set","hash","zset","stream"};
     char *typeunit[] = {"bytes","items","members","fields","members"};
     redisReply *reply, *keys;
     unsigned int arrsize=0, i;
@@ -2254,6 +2352,129 @@ static void findBigKeys(void) {
     exit(0);
 }
 
+static void getKeyFreqs(redisReply *keys, unsigned long long *freqs) {
+    redisReply *reply;
+    unsigned int i;
+
+    /* Pipeline OBJECT freq commands */
+    for(i=0;i<keys->elements;i++) {
+        redisAppendCommand(context, "OBJECT freq %s", keys->element[i]->str);
+    }
+
+    /* Retrieve freqs */
+    for(i=0;i<keys->elements;i++) {
+        if(redisGetReply(context, (void**)&reply)!=REDIS_OK) {
+            fprintf(stderr, "Error getting freq for key '%s' (%d: %s)\n",
+                keys->element[i]->str, context->err, context->errstr);
+            exit(1);
+        } else if(reply->type != REDIS_REPLY_INTEGER) {
+            if(reply->type == REDIS_REPLY_ERROR) {
+                fprintf(stderr, "Error: %s\n", reply->str);
+                exit(1);
+            } else {
+                fprintf(stderr, "Warning: OBJECT freq on '%s' failed (may have been deleted)\n", keys->element[i]->str);
+                freqs[i] = 0;
+            }
+        } else {
+            freqs[i] = reply->integer;
+        }
+        freeReplyObject(reply);
+    }
+}
+
+#define HOTKEYS_SAMPLE 16
+static void findHotKeys(void) {
+    redisReply *keys, *reply;
+    unsigned long long counters[HOTKEYS_SAMPLE] = {0};
+    sds hotkeys[HOTKEYS_SAMPLE] = {NULL};
+    unsigned long long sampled = 0, total_keys, *freqs = NULL, it = 0;
+    unsigned int arrsize = 0, i, k;
+    double pct;
+
+    /* Total keys pre scanning */
+    total_keys = getDbSize();
+
+    /* Status message */
+    printf("\n# Scanning the entire keyspace to find hot keys as well as\n");
+    printf("# average sizes per key type.  You can use -i 0.1 to sleep 0.1 sec\n");
+    printf("# per 100 SCAN commands (not usually needed).\n\n");
+
+    /* SCAN loop */
+    do {
+        /* Calculate approximate percentage completion */
+        pct = 100 * (double)sampled/total_keys;
+
+        /* Grab some keys and point to the keys array */
+        reply = sendScan(&it);
+        keys  = reply->element[1];
+
+        /* Reallocate our freqs array if we need to */
+        if(keys->elements > arrsize) {
+            freqs = zrealloc(freqs, sizeof(unsigned long long)*keys->elements);
+
+            if(!freqs) {
+                fprintf(stderr, "Failed to allocate storage for keys!\n");
+                exit(1);
+            }
+
+            arrsize = keys->elements;
+        }
+
+        getKeyFreqs(keys, freqs);
+
+        /* Now update our stats */
+        for(i=0;i<keys->elements;i++) {
+            sampled++;
+            /* Update overall progress */
+            if(sampled % 1000000 == 0) {
+                printf("[%05.2f%%] Sampled %llu keys so far\n", pct, sampled);
+            }
+
+            /* Use eviction pool here */
+            k = 0;
+            while (k < HOTKEYS_SAMPLE && freqs[i] > counters[k]) k++;
+            if (k == 0) continue;
+            k--;
+            if (k == 0 || counters[k] == 0) {
+                sdsfree(hotkeys[k]);
+            } else {
+                sdsfree(hotkeys[0]);
+                memmove(counters,counters+1,sizeof(counters[0])*k);
+                memmove(hotkeys,hotkeys+1,sizeof(hotkeys[0])*k);
+            }
+            counters[k] = freqs[i];
+            hotkeys[k] = sdsnew(keys->element[i]->str);
+            printf(
+               "[%05.2f%%] Hot key '%s' found so far with counter %llu\n",
+               pct, keys->element[i]->str, freqs[i]);
+        }
+
+        /* Sleep if we've been directed to do so */
+        if(sampled && (sampled %100) == 0 && config.interval) {
+            usleep(config.interval);
+        }
+
+        freeReplyObject(reply);
+    } while(it != 0);
+
+    if (freqs) zfree(freqs);
+
+    /* We're done */
+    printf("\n-------- summary -------\n\n");
+
+    printf("Sampled %llu keys in the keyspace!\n", sampled);
+
+    for (i=1; i<= HOTKEYS_SAMPLE; i++) {
+        k = HOTKEYS_SAMPLE - i;
+        if(counters[k]>0) {
+            printf("hot key found with counter: %llu\tkeyname: %s\n", counters[k], hotkeys[k]);
+            sdsfree(hotkeys[k]);
+        }
+    }
+
+    exit(0);
+}
+
 /*------------------------------------------------------------------------------
  * Stats mode
  *--------------------------------------------------------------------------- */
@@ -2364,7 +2585,7 @@ static void statMode(void) {
         sprintf(buf,"%ld",aux);
         printf("%-8s",buf);
 
-        /* Requets */
+        /* Requests */
         aux = getLongInfoField(reply->str,"total_commands_processed");
         sprintf(buf,"%ld (+%ld)",aux,requests == 0 ? 0 : aux-requests);
         printf("%-19s",buf);
@@ -2631,6 +2852,7 @@ int main(int argc, char **argv) {
     config.pipe_mode = 0;
     config.pipe_timeout = REDIS_CLI_DEFAULT_PIPE_TIMEOUT;
     config.bigkeys = 0;
+    config.hotkeys = 0;
     config.stdinarg = 0;
     config.auth = NULL;
     config.eval = NULL;
@@ -2691,6 +2913,12 @@ int main(int argc, char **argv) {
         findBigKeys();
     }
 
+    /* Find hot keys */
+    if (config.hotkeys) {
+        if (cliConnect(0) == REDIS_ERR) exit(1);
+        findHotKeys();
+    }
+
     /* Stat mode */
     if (config.stat_mode) {
         if (cliConnect(0) == REDIS_ERR) exit(1);

src/redis-trib.rb

@@ -701,7 +701,12 @@ class RedisTrib
 
         masters.each{|m| puts m}
 
-        # Alloc slots on masters
+        # Rotating the list sometimes helps to get better initial
+        # anti-affinity before the optimizer runs.
+        interleaved.push interleaved.shift
+
+        # Alloc slots on masters. After interleaving to get just the first N
+        # should be optimal. With slaves is more complex, see later...
         slots_per_node = ClusterHashSlots.to_f / masters_count
         first = 0
         cursor = 0.0
@@ -769,6 +774,131 @@ class RedisTrib
                 end
             end
         end
+
+        optimize_anti_affinity
+    end
+
+    def optimize_anti_affinity
+        score,aux = get_anti_affinity_score
+        return if score == 0
+
+        xputs ">>> Trying to optimize slaves allocation for anti-affinity"
+
+        maxiter = 500*@nodes.length # Effort is proportional to cluster size...
+        while maxiter > 0
+            score,offenders = get_anti_affinity_score
+            break if score == 0 # Optimal anti affinity reached
+
+            # We'll try to randomly swap a slave's assigned master causing
+            # an affinity problem with another random slave, to see if we
+            # can improve the affinity.
+            first = offenders.shuffle.first
+            nodes = @nodes.select{|n| n != first && n.info[:replicate]}
+            break if nodes.length == 0
+            second = nodes.shuffle.first
+
+            first_master = first.info[:replicate]
+            second_master = second.info[:replicate]
+            first.set_as_replica(second_master)
+            second.set_as_replica(first_master)
+
+            new_score,aux = get_anti_affinity_score
+            # If the change actually makes thing worse, revert. Otherwise
+            # leave as it is becuase the best solution may need a few
+            # combined swaps.
+            if new_score > score
+                first.set_as_replica(first_master)
+                second.set_as_replica(second_master)
+            end
+
+            maxiter -= 1
+        end
+
+        score,aux = get_anti_affinity_score
+        if score == 0
+            xputs "[OK] Perfect anti-affinity obtained!"
+        elsif score >= 10000
+            puts "[WARNING] Some slaves are in the same host as their master"
+        else
+            puts "[WARNING] Some slaves of the same master are in the same host"
+        end
+    end
+
+    # Return the anti-affinity score, which is a measure of the amount of
+    # violations of anti-affinity in the current cluster layout, that is, how
+    # badly the masters and slaves are distributed in the different IP
+    # addresses so that slaves of the same master are not in the master
+    # host and are also in different hosts.
+    #
+    # The score is calculated as follows:
+    #
+    # SAME_AS_MASTER = 10000 * each slave in the same IP of its master.
+    # SAME_AS_SLAVE  = 1 * each slave having the same IP as another slave
+    #                      of the same master.
+    # FINAL_SCORE = SAME_AS_MASTER + SAME_AS_SLAVE
+    #
+    # So a greater score means a worse anti-affinity level, while zero
+    # means perfect anti-affinity.
+    #
+    # The anti affinity optimizator will try to get a score as low as
+    # possible. Since we do not want to sacrifice the fact that slaves should
+    # not be in the same host as the master, we assign 10000 times the score
+    # to this violation, so that we'll optimize for the second factor only
+    # if it does not impact the first one.
+    #
+    # The function returns two things: the above score, and the list of
+    # offending slaves, so that the optimizer can try changing the
+    # configuration of the slaves violating the anti-affinity goals.
+    def get_anti_affinity_score
+        score = 0
+        offending = [] # List of offending slaves to return to the caller
+
+        # First, split nodes by host
+        host_to_node = {}
+        @nodes.each{|n|
+            host = n.info[:host]
+            host_to_node[host] = [] if host_to_node[host] == nil
+            host_to_node[host] << n
+        }
+
+        # Then, for each set of nodes in the same host, split by
+        # related nodes (masters and slaves which are involved in
+        # replication of each other)
+        host_to_node.each{|host,nodes|
+            related = {}
+            nodes.each{|n|
+                if !n.info[:replicate]
+                    name = n.info[:name]
+                    related[name] = [] if related[name] == nil
+                    related[name] << :m
+                else
+                    name = n.info[:replicate]
+                    related[name] = [] if related[name] == nil
+                    related[name] << :s
+                end
+            }
+
+            # Now it's trivial to check, for each related group having the
+            # same host, what is their local score.
+            related.each{|id,types|
+                next if types.length < 2
+                types.sort! # Make sure :m if the first if any
+                if types[0] == :m
+                    score += 10000 * (types.length-1)
+                else
+                    score += 1 * types.length
+                end
+
+                # Populate the list of offending nodes
+                @nodes.each{|n|
+                    if n.info[:replicate] == id &&
+                       n.info[:host] == host
+                       offending << n
+                    end
+                }
+            }
+        }
+        return score,offending
     end
 
     def flush_nodes_config

src/redismodule.h

@@ -82,6 +82,18 @@
 #define REDISMODULE_CTX_FLAGS_EVICT 0x0200 
 
 
+#define REDISMODULE_NOTIFY_GENERIC (1<<2)     /* g */
+#define REDISMODULE_NOTIFY_STRING (1<<3)      /* $ */
+#define REDISMODULE_NOTIFY_LIST (1<<4)        /* l */
+#define REDISMODULE_NOTIFY_SET (1<<5)         /* s */
+#define REDISMODULE_NOTIFY_HASH (1<<6)        /* h */
+#define REDISMODULE_NOTIFY_ZSET (1<<7)        /* z */
+#define REDISMODULE_NOTIFY_EXPIRED (1<<8)     /* x */
+#define REDISMODULE_NOTIFY_EVICTED (1<<9)     /* e */
+#define REDISMODULE_NOTIFY_STREAM (1<<10)     /* t */ 
+#define REDISMODULE_NOTIFY_ALL (REDISMODULE_NOTIFY_GENERIC | REDISMODULE_NOTIFY_STRING | REDISMODULE_NOTIFY_LIST | REDISMODULE_NOTIFY_SET | REDISMODULE_NOTIFY_HASH | REDISMODULE_NOTIFY_ZSET | REDISMODULE_NOTIFY_EXPIRED | REDISMODULE_NOTIFY_EVICTED | REDISMODULE_NOTIFY_STREAM)      /* A */
+
+
 /* A special pointer that we can use between the core and the module to signal
  * field deletion, and that is impossible to be a valid pointer. */
 #define REDISMODULE_HASH_DELETE ((RedisModuleString*)(long)1)
@@ -112,6 +124,7 @@ typedef struct RedisModuleBlockedClient RedisModuleBlockedClient;
 
 typedef int (*RedisModuleCmdFunc) (RedisModuleCtx *ctx, RedisModuleString **argv, int argc);
 
+typedef int (*RedisModuleNotificationFunc) (RedisModuleCtx *ctx, int type, const char *event, RedisModuleString *key);
 typedef void *(*RedisModuleTypeLoadFunc)(RedisModuleIO *rdb, int encver);
 typedef void (*RedisModuleTypeSaveFunc)(RedisModuleIO *rdb, void *value);
 typedef void (*RedisModuleTypeRewriteFunc)(RedisModuleIO *aof, RedisModuleString *key, void *value);
@@ -143,7 +156,8 @@ void *REDISMODULE_API_FUNC(RedisModule_Calloc)(size_t nmemb, size_t size);
 char *REDISMODULE_API_FUNC(RedisModule_Strdup)(const char *str);
 int REDISMODULE_API_FUNC(RedisModule_GetApi)(const char *, void *);
 int REDISMODULE_API_FUNC(RedisModule_CreateCommand)(RedisModuleCtx *ctx, const char *name, RedisModuleCmdFunc cmdfunc, const char *strflags, int firstkey, int lastkey, int keystep);
-int REDISMODULE_API_FUNC(RedisModule_SetModuleAttribs)(RedisModuleCtx *ctx, const char *name, int ver, int apiver);
+void REDISMODULE_API_FUNC(RedisModule_SetModuleAttribs)(RedisModuleCtx *ctx, const char *name, int ver, int apiver);
+int REDISMODULE_API_FUNC(RedisModule_IsModuleNameBusy)(const char *name);
 int REDISMODULE_API_FUNC(RedisModule_WrongArity)(RedisModuleCtx *ctx);
 int REDISMODULE_API_FUNC(RedisModule_ReplyWithLongLong)(RedisModuleCtx *ctx, long long ll);
 int REDISMODULE_API_FUNC(RedisModule_GetSelectedDb)(RedisModuleCtx *ctx);
@@ -184,6 +198,7 @@ int REDISMODULE_API_FUNC(RedisModule_ReplicateVerbatim)(RedisModuleCtx *ctx);
 const char *REDISMODULE_API_FUNC(RedisModule_CallReplyStringPtr)(RedisModuleCallReply *reply, size_t *len);
 RedisModuleString *REDISMODULE_API_FUNC(RedisModule_CreateStringFromCallReply)(RedisModuleCallReply *reply);
 int REDISMODULE_API_FUNC(RedisModule_DeleteKey)(RedisModuleKey *key);
+int REDISMODULE_API_FUNC(RedisModule_UnlinkKey)(RedisModuleKey *key);
 int REDISMODULE_API_FUNC(RedisModule_StringSet)(RedisModuleKey *key, RedisModuleString *str);
 char *REDISMODULE_API_FUNC(RedisModule_StringDMA)(RedisModuleKey *key, size_t *len, int mode);
 int REDISMODULE_API_FUNC(RedisModule_StringTruncate)(RedisModuleKey *key, size_t newlen);
@@ -249,6 +264,8 @@ RedisModuleCtx *REDISMODULE_API_FUNC(RedisModule_GetThreadSafeContext)(RedisModu
 void REDISMODULE_API_FUNC(RedisModule_FreeThreadSafeContext)(RedisModuleCtx *ctx);
 void REDISMODULE_API_FUNC(RedisModule_ThreadSafeContextLock)(RedisModuleCtx *ctx);
 void REDISMODULE_API_FUNC(RedisModule_ThreadSafeContextUnlock)(RedisModuleCtx *ctx);
+int REDISMODULE_API_FUNC(RedisModule_SubscribeToKeyspaceEvents)(RedisModuleCtx *ctx, int types, RedisModuleNotificationFunc cb);
+
 #endif
 
 /* This is included inline inside each Redis module. */
@@ -263,6 +280,7 @@ static int RedisModule_Init(RedisModuleCtx *ctx, const char *name, int ver, int
     REDISMODULE_GET_API(Strdup);
     REDISMODULE_GET_API(CreateCommand);
     REDISMODULE_GET_API(SetModuleAttribs);
+    REDISMODULE_GET_API(IsModuleNameBusy);
     REDISMODULE_GET_API(WrongArity);
     REDISMODULE_GET_API(ReplyWithLongLong);
     REDISMODULE_GET_API(ReplyWithError);
@@ -304,6 +322,7 @@ static int RedisModule_Init(RedisModuleCtx *ctx, const char *name, int ver, int
     REDISMODULE_GET_API(Replicate);
     REDISMODULE_GET_API(ReplicateVerbatim);
     REDISMODULE_GET_API(DeleteKey);
+    REDISMODULE_GET_API(UnlinkKey);
     REDISMODULE_GET_API(StringSet);
     REDISMODULE_GET_API(StringDMA);
     REDISMODULE_GET_API(StringTruncate);
@@ -368,8 +387,11 @@ static int RedisModule_Init(RedisModuleCtx *ctx, const char *name, int ver, int
     REDISMODULE_GET_API(IsBlockedTimeoutRequest);
     REDISMODULE_GET_API(GetBlockedClientPrivateData);
     REDISMODULE_GET_API(AbortBlock);
+    REDISMODULE_GET_API(SubscribeToKeyspaceEvents);
+
 #endif
 
+    if (RedisModule_IsModuleNameBusy && RedisModule_IsModuleNameBusy(name)) return REDISMODULE_ERR;
     RedisModule_SetModuleAttribs(ctx,name,ver,apiver);
     return REDISMODULE_OK;
 }

src/replication.c

@@ -1330,7 +1330,8 @@ char *sendSynchronousCommand(int flags, int fd, ...) {
             cmd = sdscat(cmd,arg);
         }
         cmd = sdscatlen(cmd,"\r\n",2);
-
+        va_end(ap);
+        
         /* Transfer command to the server. */
         if (syncWrite(fd,cmd,sdslen(cmd),server.repl_syncio_timeout*1000)
             == -1)
@@ -1340,7 +1341,6 @@ char *sendSynchronousCommand(int flags, int fd, ...) {
                     strerror(errno));
         }
         sdsfree(cmd);
-        va_end(ap);
     }
 
     /* Read the reply from the server. */
@@ -1970,6 +1970,12 @@ void replicationUnsetMaster(void) {
      * with PSYNC version 2, there is no need for full resync after a
      * master switch. */
     server.slaveseldb = -1;
+
+    /* Once we turn from slave to master, we consider the starting time without
+     * slaves (that is used to count the replication backlog time to live) as
+     * starting from now. Otherwise the backlog will be freed after a
+     * failover if slaves do not connect immediately. */
+    server.repl_no_slaves_since = server.unixtime;
 }
 
 /* This function is called when the slave lose the connection with the
@@ -2205,7 +2211,7 @@ void replicationResurrectCachedMaster(int newfd) {
     server.repl_state = REPL_STATE_CONNECTED;
 
     /* Re-add to the list of clients. */
-    listAddNodeTail(server.clients,server.master);
+    linkClient(server.master);
     if (aeCreateFileEvent(server.el, newfd, AE_READABLE,
                           readQueryFromClient, server.master)) {
         serverLog(LL_WARNING,"Error resurrecting the cached master, impossible to add the readable handler: %s", strerror(errno));
@@ -2388,15 +2394,26 @@ void waitCommand(client *c) {
     mstime_t timeout;
     long numreplicas, ackreplicas;
     long long offset = c->woff;
+    int waitaof = 0; /* True if the user requested to wait for AOF sync. */
 
     if (server.masterhost) {
         addReplyError(c,"WAIT cannot be used with slave instances. Please also note that since Redis 4.0 if a slave is configured to be writable (which is not the default) writes to slaves are just local and are not propagated.");
         return;
     }
 
-    /* Argument parsing. */
-    if (getLongFromObjectOrReply(c,c->argv[1],&numreplicas,NULL) != C_OK)
-        return;
+    /* AOF or number of replicas argument parsing. */
+    if (!strcasecmp(c->argv[1]->ptr,"AOF")) {
+        waitaof = 1;
+        if (server.aof_state != AOF_ON) {
+            addReplyError(c,"WAIT AOF is only allowed when AOF is enabled");
+            return;
+        }
+    } else {
+        if (getLongFromObjectOrReply(c,c->argv[1],&numreplicas,NULL) != C_OK)
+            return;
+    }
+
+    /* Timeout parsing. */
     if (getTimeoutFromObjectOrReply(c,c->argv[2],&timeout,UNIT_MILLISECONDS)
         != C_OK) return;
 
@@ -2409,11 +2426,17 @@ void waitCommand(client *c) {
 
     /* Otherwise block the client and put it into our list of clients
      * waiting for ack from slaves. */
-    c->bpop.timeout = timeout;
-    c->bpop.reploffset = offset;
-    c->bpop.numreplicas = numreplicas;
-    listAddNodeTail(server.clients_waiting_acks,c);
-    blockClient(c,BLOCKED_WAIT);
+    if (waitaof) {
+        c->bpop.aofepoch = aofNextEpoch();
+        listAddNodeTail(server.clients_waiting_acks,c);
+        blockClient(c,BLOCKED_AOF);
+    } else {
+        c->bpop.timeout = timeout;
+        c->bpop.reploffset = offset;
+        c->bpop.numreplicas = numreplicas;
+        listAddNodeTail(server.clients_waiting_acks,c);
+        blockClient(c,BLOCKED_WAIT);
+    }
 
     /* Make sure that the server will send an ACK request to all the slaves
      * before returning to the event loop. */
@@ -2432,7 +2455,7 @@ void unblockClientWaitingReplicas(client *c) {
 
 /* Check if there are clients blocked in WAIT that can be unblocked since
  * we received enough ACKs from slaves. */
-void processClientsWaitingReplicas(void) {
+void processClientsBlockedInWait(void) {
     long long last_offset = 0;
     int last_numreplicas = 0;
 
@@ -2443,23 +2466,36 @@ void processClientsWaitingReplicas(void) {
     while((ln = listNext(&li))) {
         client *c = ln->value;
 
-        /* Every time we find a client that is satisfied for a given
-         * offset and number of replicas, we remember it so the next client
-         * may be unblocked without calling replicationCountAcksByOffset()
-         * if the requested offset / replicas were equal or less. */
-        if (last_offset && last_offset > c->bpop.reploffset &&
-                           last_numreplicas > c->bpop.numreplicas)
-        {
-            unblockClient(c);
-            addReplyLongLong(c,last_numreplicas);
-        } else {
-            int numreplicas = replicationCountAcksByOffset(c->bpop.reploffset);
-
-            if (numreplicas >= c->bpop.numreplicas) {
-                last_offset = c->bpop.reploffset;
-                last_numreplicas = numreplicas;
+        if (c->btype == BLOCKED_AOF) {
+            /* Handle WAIT AOF. */
+            if (server.aof_fsync == AOF_FSYNC_ALWAYS ||
+                c->bpop.aofepoch <= server.aof_fsync_epoch)
+            {
                 unblockClient(c);
-                addReplyLongLong(c,numreplicas);
+                addReply(c,shared.cone);
+            }
+        } else {
+            /* Handle WAIT for slaves to ACK.
+             *
+             * Every time we find a client that is satisfied for a given
+             * offset and number of replicas, we remember it so the next client
+             * may be unblocked without calling replicationCountAcksByOffset()
+             * if the requested offset / replicas were equal or less. */
+            if (last_offset && last_offset > c->bpop.reploffset &&
+                               last_numreplicas > c->bpop.numreplicas)
+            {
+                unblockClient(c);
+                addReplyLongLong(c,last_numreplicas);
+            } else {
+                int numreplicas =
+                    replicationCountAcksByOffset(c->bpop.reploffset);
+
+                if (numreplicas >= c->bpop.numreplicas) {
+                    last_offset = c->bpop.reploffset;
+                    last_numreplicas = numreplicas;
+                    unblockClient(c);
+                    addReplyLongLong(c,numreplicas);
+                }
             }
         }
     }
@@ -2613,6 +2649,23 @@ void replicationCron(void) {
         time_t idle = server.unixtime - server.repl_no_slaves_since;
 
         if (idle > server.repl_backlog_time_limit) {
+            /* When we free the backlog, we always use a new
+             * replication ID and clear the ID2. This is needed
+             * because when there is no backlog, the master_repl_offset
+             * is not updated, but we would still retain our replication
+             * ID, leading to the following problem:
+             *
+             * 1. We are a master instance.
+             * 2. Our slave is promoted to master. It's repl-id-2 will
+             *    be the same as our repl-id.
+             * 3. We, yet as master, receive some updates, that will not
+             *    increment the master_repl_offset.
+             * 4. Later we are turned into a slave, connecto to the new
+             *    master that will accept our PSYNC request by second
+             *    replication ID, but there will be data inconsistency
+             *    because we received writes. */
+            changeReplicationId();
+            clearReplicationId2();
             freeReplicationBacklog();
             serverLog(LL_NOTICE,
                 "Replication backlog freed after %d seconds "

src/rio.c

@@ -310,7 +310,7 @@ void rioSetAutoSync(rio *r, off_t bytes) {
  * generating the Redis protocol for the Append Only File. */
 
 /* Write multi bulk count in the format: "*<count>\r\n". */
-size_t rioWriteBulkCount(rio *r, char prefix, int count) {
+size_t rioWriteBulkCount(rio *r, char prefix, long count) {
     char cbuf[128];
     int clen;
 

src/rio.h

@@ -130,7 +130,7 @@ void rioInitWithFdset(rio *r, int *fds, int numfds);
 
 void rioFreeFdset(rio *r);
 
-size_t rioWriteBulkCount(rio *r, char prefix, int count);
+size_t rioWriteBulkCount(rio *r, char prefix, long count);
 size_t rioWriteBulkString(rio *r, const char *buf, size_t len);
 size_t rioWriteBulkLongLong(rio *r, long long l);
 size_t rioWriteBulkDouble(rio *r, double d);

src/scripting.c

@@ -358,6 +358,13 @@ int luaRedisGenericCommand(lua_State *lua, int raise_error) {
     static size_t cached_objects_len[LUA_CMD_OBJCACHE_SIZE];
     static int inuse = 0;   /* Recursive calls detection. */
 
+    /* Reflect MULTI state */
+    if (server.lua_multi_emitted || (server.lua_caller->flags & CLIENT_MULTI)) {
+        c->flags |= CLIENT_MULTI;
+    } else {
+        c->flags &= ~CLIENT_MULTI;
+    }
+
     /* By using Lua debug hooks it is possible to trigger a recursive call
      * to luaRedisGenericCommand(), which normally should never happen.
      * To make this function reentrant is futile and makes it slower, but
@@ -535,6 +542,7 @@ int luaRedisGenericCommand(lua_State *lua, int raise_error) {
      * a Lua script in the context of AOF and slaves. */
     if (server.lua_replicate_commands &&
         !server.lua_multi_emitted &&
+        !(server.lua_caller->flags & CLIENT_MULTI) &&
         server.lua_write_dirty &&
         server.lua_repl != PROPAGATE_NONE)
     {
@@ -1133,18 +1141,38 @@ int redis_math_randomseed (lua_State *L) {
  * EVAL and SCRIPT commands implementation
  * ------------------------------------------------------------------------- */
 
-/* Define a lua function with the specified function name and body.
- * The function name musts be a 42 characters long string, since all the
- * functions we defined in the Lua context are in the form:
+/* Define a Lua function with the specified body.
+ * The function name will be generated in the following form:
  *
  *   f_<hex sha1 sum>
  *
- * On success C_OK is returned, and nothing is left on the Lua stack.
- * On error C_ERR is returned and an appropriate error is set in the
- * client context. */
-int luaCreateFunction(client *c, lua_State *lua, char *funcname, robj *body) {
-    sds funcdef = sdsempty();
+ * The function increments the reference count of the 'body' object as a
+ * side effect of a successful call.
+ *
+ * On success a pointer to an SDS string representing the function SHA1 of the
+ * just added function is returned (and will be valid until the next call
+ * to scriptingReset() function), otherwise NULL is returned.
+ *
+ * The function handles the fact of being called with a script that already
+ * exists, and in such a case, it behaves like in the success case.
+ *
+ * If 'c' is not NULL, on error the client is informed with an appropriate
+ * error describing the nature of the problem and the Lua interpreter error. */
+sds luaCreateFunction(client *c, lua_State *lua, robj *body) {
+    char funcname[43];
+    dictEntry *de;
 
+    funcname[0] = 'f';
+    funcname[1] = '_';
+    sha1hex(funcname+2,body->ptr,sdslen(body->ptr));
+
+    sds sha = sdsnewlen(funcname+2,40);
+    if ((de = dictFind(server.lua_scripts,sha)) != NULL) {
+        sdsfree(sha);
+        return dictGetKey(de);
+    }
+
+    sds funcdef = sdsempty();
     funcdef = sdscat(funcdef,"function ");
     funcdef = sdscatlen(funcdef,funcname,42);
     funcdef = sdscatlen(funcdef,"() ",3);
@@ -1152,30 +1180,35 @@ int luaCreateFunction(client *c, lua_State *lua, char *funcname, robj *body) {
     funcdef = sdscatlen(funcdef,"\nend",4);
 
     if (luaL_loadbuffer(lua,funcdef,sdslen(funcdef),"@user_script")) {
-        addReplyErrorFormat(c,"Error compiling script (new function): %s\n",
-            lua_tostring(lua,-1));
+        if (c != NULL) {
+            addReplyErrorFormat(c,
+                "Error compiling script (new function): %s\n",
+                lua_tostring(lua,-1));
+        }
         lua_pop(lua,1);
+        sdsfree(sha);
         sdsfree(funcdef);
-        return C_ERR;
+        return NULL;
     }
     sdsfree(funcdef);
+
     if (lua_pcall(lua,0,0,0)) {
-        addReplyErrorFormat(c,"Error running script (new function): %s\n",
-            lua_tostring(lua,-1));
+        if (c != NULL) {
+            addReplyErrorFormat(c,"Error running script (new function): %s\n",
+                lua_tostring(lua,-1));
+        }
         lua_pop(lua,1);
-        return C_ERR;
+        sdsfree(sha);
+        return NULL;
     }
 
     /* We also save a SHA1 -> Original script map in a dictionary
      * so that we can replicate / write in the AOF all the
      * EVALSHA commands as EVAL using the original script. */
-    {
-        int retval = dictAdd(server.lua_scripts,
-                             sdsnewlen(funcname+2,40),body);
-        serverAssertWithInfo(c,NULL,retval == DICT_OK);
-        incrRefCount(body);
-    }
-    return C_OK;
+    int retval = dictAdd(server.lua_scripts,sha,body);
+    serverAssertWithInfo(c ? c : server.lua_client,NULL,retval == DICT_OK);
+    incrRefCount(body);
+    return sha;
 }
 
 /* This is the Lua script "count" hook that we use to detect scripts timeout. */
@@ -1274,10 +1307,10 @@ void evalGenericCommand(client *c, int evalsha) {
             addReply(c, shared.noscripterr);
             return;
         }
-        if (luaCreateFunction(c,lua,funcname,c->argv[1]) == C_ERR) {
+        if (luaCreateFunction(c,lua,c->argv[1]) == NULL) {
             lua_pop(lua,1); /* remove the error handler from the stack. */
             /* The error is sent to the client by luaCreateFunction()
-             * itself when it returns C_ERR. */
+             * itself when it returns NULL. */
             return;
         }
         /* Now the following is guaranteed to return non nil */
@@ -1422,7 +1455,17 @@ void evalShaCommand(client *c) {
 }
 
 void scriptCommand(client *c) {
-    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"flush")) {
+    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"help")) {
+        const char *help[] = {
+"debug (yes|sync|no) -- Set the debug mode for subsequent scripts executed.",
+"exists <sha1> [<sha1> ...] -- Return information about the existence of the scripts in the script cache.",
+"flush -- Flush the Lua scripts cache. Very dangerous on slaves.",
+"kill -- Kill the currently executing Lua script.",
+"load <script> -- Load a script into the scripts cache, without executing it.",
+NULL
+        };
+        addReplyHelp(c, help);
+    } else if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"flush")) {
         scriptingReset();
         addReply(c,shared.ok);
         replicationScriptCacheFlush();
@@ -1438,22 +1481,9 @@ void scriptCommand(client *c) {
                 addReply(c,shared.czero);
         }
     } else if (c->argc == 3 && !strcasecmp(c->argv[1]->ptr,"load")) {
-        char funcname[43];
-        sds sha;
-
-        funcname[0] = 'f';
-        funcname[1] = '_';
-        sha1hex(funcname+2,c->argv[2]->ptr,sdslen(c->argv[2]->ptr));
-        sha = sdsnewlen(funcname+2,40);
-        if (dictFind(server.lua_scripts,sha) == NULL) {
-            if (luaCreateFunction(c,server.lua,funcname,c->argv[2])
-                    == C_ERR) {
-                sdsfree(sha);
-                return;
-            }
-        }
-        addReplyBulkCBuffer(c,funcname+2,40);
-        sdsfree(sha);
+        sds sha = luaCreateFunction(c,server.lua,c->argv[2]);
+        if (sha == NULL) return; /* The error was sent by luaCreateFunction(). */
+        addReplyBulkCBuffer(c,sha,40);
         forceCommandPropagation(c,PROPAGATE_REPL|PROPAGATE_AOF);
     } else if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"kill")) {
         if (server.lua_caller == NULL) {
@@ -1481,9 +1511,10 @@ void scriptCommand(client *c) {
             c->flags |= CLIENT_LUA_DEBUG_SYNC;
         } else {
             addReplyError(c,"Use SCRIPT DEBUG yes/sync/no");
+            return;
         }
     } else {
-        addReplyError(c, "Unknown SCRIPT subcommand or wrong # of args.");
+        addReplyErrorFormat(c, "Unknown subcommand or wrong number of arguments for '%s'. Try SCRIPT HELP", (char*)c->argv[1]->ptr);
     }
 }
 

src/sds.c

@@ -39,6 +39,8 @@
 #include "sds.h"
 #include "sdsalloc.h"
 
+const char *SDS_NOINIT = "SDS_NOINIT";
+
 static inline int sdsHdrSize(char type) {
     switch(type&SDS_TYPE_MASK) {
         case SDS_TYPE_5:
@@ -72,6 +74,7 @@ static inline char sdsReqType(size_t string_size) {
 /* Create a new sds string with the content specified by the 'init' pointer
  * and 'initlen'.
  * If NULL is used for 'init' the string is initialized with zero bytes.
+ * If SDS_NOINIT is used, the buffer is left uninitialized;
  *
  * The string is always null-termined (all the sds strings are, always) so
  * even if you create an sds string with:
@@ -92,7 +95,9 @@ sds sdsnewlen(const void *init, size_t initlen) {
     unsigned char *fp; /* flags pointer. */
 
     sh = s_malloc(hdrlen+initlen+1);
-    if (!init)
+    if (init==SDS_NOINIT)
+        init = NULL;
+    else if (!init)
         memset(sh, 0, hdrlen+initlen+1);
     if (sh == NULL) return NULL;
     s = (char*)sh+hdrlen;
@@ -175,7 +180,7 @@ void sdsfree(sds s) {
  * the output will be "6" as the string was modified but the logical length
  * remains 6 bytes. */
 void sdsupdatelen(sds s) {
-    int reallen = strlen(s);
+    size_t reallen = strlen(s);
     sdssetlen(s, reallen);
 }
 
@@ -319,7 +324,7 @@ void *sdsAllocPtr(sds s) {
  * ... check for nread <= 0 and handle it ...
  * sdsIncrLen(s, nread);
  */
-void sdsIncrLen(sds s, int incr) {
+void sdsIncrLen(sds s, ssize_t incr) {
     unsigned char flags = s[-1];
     size_t len;
     switch(flags&SDS_TYPE_MASK) {
@@ -589,7 +594,7 @@ sds sdscatprintf(sds s, const char *fmt, ...) {
 sds sdscatfmt(sds s, char const *fmt, ...) {
     size_t initlen = sdslen(s);
     const char *f = fmt;
-    int i;
+    long i;
     va_list ap;
 
     va_start(ap,fmt);
@@ -721,7 +726,7 @@ sds sdstrim(sds s, const char *cset) {
  * s = sdsnew("Hello World");
  * sdsrange(s,1,-1); => "ello World"
  */
-void sdsrange(sds s, int start, int end) {
+void sdsrange(sds s, ssize_t start, ssize_t end) {
     size_t newlen, len = sdslen(s);
 
     if (len == 0) return;
@@ -735,9 +740,9 @@ void sdsrange(sds s, int start, int end) {
     }
     newlen = (start > end) ? 0 : (end-start)+1;
     if (newlen != 0) {
-        if (start >= (signed)len) {
+        if (start >= (ssize_t)len) {
             newlen = 0;
-        } else if (end >= (signed)len) {
+        } else if (end >= (ssize_t)len) {
             end = len-1;
             newlen = (start > end) ? 0 : (end-start)+1;
         }
@@ -751,14 +756,14 @@ void sdsrange(sds s, int start, int end) {
 
 /* Apply tolower() to every character of the sds string 's'. */
 void sdstolower(sds s) {
-    int len = sdslen(s), j;
+    size_t len = sdslen(s), j;
 
     for (j = 0; j < len; j++) s[j] = tolower(s[j]);
 }
 
 /* Apply toupper() to every character of the sds string 's'. */
 void sdstoupper(sds s) {
-    int len = sdslen(s), j;
+    size_t len = sdslen(s), j;
 
     for (j = 0; j < len; j++) s[j] = toupper(s[j]);
 }
@@ -782,7 +787,7 @@ int sdscmp(const sds s1, const sds s2) {
     l2 = sdslen(s2);
     minlen = (l1 < l2) ? l1 : l2;
     cmp = memcmp(s1,s2,minlen);
-    if (cmp == 0) return l1-l2;
+    if (cmp == 0) return l1>l2? 1: (l1<l2? -1: 0);
     return cmp;
 }
 
@@ -802,8 +807,9 @@ int sdscmp(const sds s1, const sds s2) {
  * requires length arguments. sdssplit() is just the
  * same function but for zero-terminated strings.
  */
-sds *sdssplitlen(const char *s, int len, const char *sep, int seplen, int *count) {
-    int elements = 0, slots = 5, start = 0, j;
+sds *sdssplitlen(const char *s, ssize_t len, const char *sep, int seplen, int *count) {
+    int elements = 0, slots = 5;
+    long start = 0, j;
     sds *tokens;
 
     if (seplen < 1 || len < 0) return NULL;

src/sds.h

@@ -34,6 +34,7 @@
 #define __SDS_H
 
 #define SDS_MAX_PREALLOC (1024*1024)
+const char *SDS_NOINIT;
 
 #include <sys/types.h>
 #include <stdarg.h>
@@ -236,11 +237,11 @@ sds sdscatprintf(sds s, const char *fmt, ...);
 
 sds sdscatfmt(sds s, char const *fmt, ...);
 sds sdstrim(sds s, const char *cset);
-void sdsrange(sds s, int start, int end);
+void sdsrange(sds s, ssize_t start, ssize_t end);
 void sdsupdatelen(sds s);
 void sdsclear(sds s);
 int sdscmp(const sds s1, const sds s2);
-sds *sdssplitlen(const char *s, int len, const char *sep, int seplen, int *count);
+sds *sdssplitlen(const char *s, ssize_t len, const char *sep, int seplen, int *count);
 void sdsfreesplitres(sds *tokens, int count);
 void sdstolower(sds s);
 void sdstoupper(sds s);
@@ -253,7 +254,7 @@ sds sdsjoinsds(sds *argv, int argc, const char *sep, size_t seplen);
 
 /* Low level functions exposed to the user API */
 sds sdsMakeRoomFor(sds s, size_t addlen);
-void sdsIncrLen(sds s, int incr);
+void sdsIncrLen(sds s, ssize_t incr);
 sds sdsRemoveFreeSpace(sds s);
 size_t sdsAllocSize(sds s);
 void *sdsAllocPtr(sds s);

src/server.c

@@ -125,7 +125,7 @@ volatile unsigned long lru_clock; /* Server global current LRU time. */
  *    are not fast commands.
  */
 struct redisCommand redisCommandTable[] = {
-    {"module",moduleCommand,-2,"as",0,NULL,1,1,1,0,0},
+    {"module",moduleCommand,-2,"as",0,NULL,0,0,0,0,0},
     {"get",getCommand,2,"rF",0,NULL,1,1,1,0,0},
     {"set",setCommand,-3,"wm",0,NULL,1,1,1,0,0},
     {"setnx",setnxCommand,3,"wmF",0,NULL,1,1,1,0,0},
@@ -258,7 +258,7 @@ struct redisCommand redisCommandTable[] = {
     {"persist",persistCommand,2,"wF",0,NULL,1,1,1,0,0},
     {"slaveof",slaveofCommand,3,"ast",0,NULL,0,0,0,0,0},
     {"role",roleCommand,1,"lst",0,NULL,0,0,0,0,0},
-    {"debug",debugCommand,-1,"as",0,NULL,0,0,0,0,0},
+    {"debug",debugCommand,-2,"as",0,NULL,0,0,0,0,0},
     {"config",configCommand,-2,"lat",0,NULL,0,0,0,0,0},
     {"subscribe",subscribeCommand,-2,"pslt",0,NULL,0,0,0,0,0},
     {"unsubscribe",unsubscribeCommand,-1,"pslt",0,NULL,0,0,0,0,0},
@@ -276,7 +276,7 @@ struct redisCommand redisCommandTable[] = {
     {"readonly",readonlyCommand,1,"F",0,NULL,0,0,0,0,0},
     {"readwrite",readwriteCommand,1,"F",0,NULL,0,0,0,0,0},
     {"dump",dumpCommand,2,"r",0,NULL,1,1,1,0,0},
-    {"object",objectCommand,3,"r",0,NULL,2,2,2,0,0},
+    {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
     {"memory",memoryCommand,-2,"r",0,NULL,0,0,0,0,0},
     {"client",clientCommand,-2,"as",0,NULL,0,0,0,0,0},
     {"eval",evalCommand,-3,"s",0,evalGetKeys,0,0,0,0,0},
@@ -302,6 +302,11 @@ struct redisCommand redisCommandTable[] = {
     {"pfcount",pfcountCommand,-2,"r",0,NULL,1,-1,1,0,0},
     {"pfmerge",pfmergeCommand,-2,"wm",0,NULL,1,-1,1,0,0},
     {"pfdebug",pfdebugCommand,-3,"w",0,NULL,0,0,0,0,0},
+    {"xadd",xaddCommand,-5,"wmF",0,NULL,1,1,1,0,0},
+    {"xrange",xrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
+    {"xrevrange",xrevrangeCommand,-4,"r",0,NULL,1,1,1,0,0},
+    {"xlen",xlenCommand,2,"rF",0,NULL,1,1,1,0,0},
+    {"xread",xreadCommand,-3,"rs",0,xreadGetKeys,1,1,1,0,0},
     {"post",securityWarningCommand,-1,"lt",0,NULL,0,0,0,0,0},
     {"host:",securityWarningCommand,-1,"lt",0,NULL,0,0,0,0,0},
     {"latency",latencyCommand,-2,"aslt",0,NULL,0,0,0,0,0}
@@ -547,10 +552,21 @@ dictType objectKeyPointerValueDictType = {
     NULL,                      /* key dup */
     NULL,                      /* val dup */
     dictEncObjKeyCompare,      /* key compare */
-    dictObjectDestructor, /* key destructor */
+    dictObjectDestructor,      /* key destructor */
     NULL                       /* val destructor */
 };
 
+/* Like objectKeyPointerValueDictType(), but values can be destroyed, if
+ * not NULL, calling zfree(). */
+dictType objectKeyHeapPointerValueDictType = {
+    dictEncObjHash,            /* hash function */
+    NULL,                      /* key dup */
+    NULL,                      /* val dup */
+    dictEncObjKeyCompare,      /* key compare */
+    dictObjectDestructor,      /* key destructor */
+    dictVanillaFree            /* val destructor */
+};
+
 /* Set dictionary type. Keys are SDS strings, values are ot used. */
 dictType setDictType = {
     dictSdsHash,               /* hash function */
@@ -1079,7 +1095,7 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
         }
     } else {
         /* If there is not a background saving/rewrite in progress check if
-         * we have to save/rewrite now */
+         * we have to save/rewrite now. */
          for (j = 0; j < server.saveparamslen; j++) {
             struct saveparam *sp = server.saveparams+j;
 
@@ -1102,8 +1118,9 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
             }
          }
 
-         /* Trigger an AOF rewrite if needed */
-         if (server.rdb_child_pid == -1 &&
+         /* Trigger an AOF rewrite if needed. */
+         if (server.aof_state == AOF_ON &&
+             server.rdb_child_pid == -1 &&
              server.aof_child_pid == -1 &&
              server.aof_rewrite_perc &&
              server.aof_current_size > server.aof_rewrite_min_size)
@@ -1212,9 +1229,19 @@ void beforeSleep(struct aeEventLoop *eventLoop) {
     }
 
     /* Unblock all the clients blocked for synchronous replication
-     * in WAIT. */
-    if (listLength(server.clients_waiting_acks))
-        processClientsWaitingReplicas();
+     * or AOF sync in WAIT. */
+    if (listLength(server.clients_waiting_acks)) {
+        processClientsBlockedInWait();
+
+        /* If after this cycle we have still clients blocked waiting for
+         * AOF fsync, try to start a new sync cycle. Note that if one is
+         * already in progress, the call does nothing. */
+        if (server.blocked_clients_by_type[BLOCKED_AOF] &&
+            server.aof_state == AOF_ON)
+        {
+            aofStartBackgroundFsync();
+        }
+    }
 
     /* Check if there are clients unblocked by modules that implement
      * blocking commands. */
@@ -1372,6 +1399,7 @@ void initServerConfig(void) {
     server.active_defrag_threshold_upper = CONFIG_DEFAULT_DEFRAG_THRESHOLD_UPPER;
     server.active_defrag_cycle_min = CONFIG_DEFAULT_DEFRAG_CYCLE_MIN;
     server.active_defrag_cycle_max = CONFIG_DEFAULT_DEFRAG_CYCLE_MAX;
+    server.proto_max_bulk_len = CONFIG_DEFAULT_PROTO_MAX_BULK_LEN;
     server.client_max_querybuf_len = PROTO_MAX_QUERYBUF_LEN;
     server.saveparams = NULL;
     server.loading = 0;
@@ -1400,6 +1428,8 @@ void initServerConfig(void) {
     server.aof_rewrite_incremental_fsync = CONFIG_DEFAULT_AOF_REWRITE_INCREMENTAL_FSYNC;
     server.aof_load_truncated = CONFIG_DEFAULT_AOF_LOAD_TRUNCATED;
     server.aof_use_rdb_preamble = CONFIG_DEFAULT_AOF_USE_RDB_PREAMBLE;
+    server.aof_fsync_in_progress_epoch = 0;
+    server.aof_fsync_epoch = 0;
     server.pidfile = NULL;
     server.rdb_filename = zstrdup(CONFIG_DEFAULT_RDB_FILENAME);
     server.aof_filename = zstrdup(CONFIG_DEFAULT_AOF_FILENAME);
@@ -1411,7 +1441,9 @@ void initServerConfig(void) {
     server.active_defrag_running = 0;
     server.notify_keyspace_events = 0;
     server.maxclients = CONFIG_DEFAULT_MAX_CLIENTS;
-    server.bpop_blocked_clients = 0;
+    server.blocked_clients = 0;
+    memset(server.blocked_clients_by_type,0,
+           sizeof(server.blocked_clients_by_type));
     server.maxmemory = CONFIG_DEFAULT_MAXMEMORY;
     server.maxmemory_policy = CONFIG_DEFAULT_MAXMEMORY_POLICY;
     server.maxmemory_samples = CONFIG_DEFAULT_MAXMEMORY_SAMPLES;
@@ -1549,16 +1581,29 @@ int restartServer(int flags, mstime_t delay) {
 
     /* Check if we still have accesses to the executable that started this
      * server instance. */
-    if (access(server.executable,X_OK) == -1) return C_ERR;
+    if (access(server.executable,X_OK) == -1) {
+        serverLog(LL_WARNING,"Can't restart: this process has no "
+                             "permissions to execute %s", server.executable);
+        return C_ERR;
+    }
 
     /* Config rewriting. */
     if (flags & RESTART_SERVER_CONFIG_REWRITE &&
         server.configfile &&
-        rewriteConfig(server.configfile) == -1) return C_ERR;
+        rewriteConfig(server.configfile) == -1)
+    {
+        serverLog(LL_WARNING,"Can't restart: configuration rewrite process "
+                             "failed");
+        return C_ERR;
+    }
 
     /* Perform a proper shutdown. */
     if (flags & RESTART_SERVER_GRACEFULLY &&
-        prepareForShutdown(SHUTDOWN_NOFLAGS) != C_OK) return C_ERR;
+        prepareForShutdown(SHUTDOWN_NOFLAGS) != C_OK)
+    {
+        serverLog(LL_WARNING,"Can't restart: error preparing for shutdown");
+        return C_ERR;
+    }
 
     /* Close all file descriptors, with the exception of stdin, stdout, strerr
      * which are useful if we restart a Redis server which is not daemonized. */
@@ -1570,6 +1615,8 @@ int restartServer(int flags, mstime_t delay) {
 
     /* Execute the server with the original command line. */
     if (delay) usleep(delay*1000);
+    zfree(server.exec_argv[0]);
+    server.exec_argv[0] = zstrdup(server.executable);
     execve(server.executable,server.exec_argv,environ);
 
     /* If an error occurred here, there is nothing we can do, but exit. */
@@ -1764,6 +1811,8 @@ void resetServerStats(void) {
     server.stat_numcommands = 0;
     server.stat_numconnections = 0;
     server.stat_expiredkeys = 0;
+    server.stat_expired_stale_perc = 0;
+    server.stat_expired_time_cap_reached_count = 0;
     server.stat_evictedkeys = 0;
     server.stat_keyspace_misses = 0;
     server.stat_keyspace_hits = 0;
@@ -2265,8 +2314,9 @@ void call(client *c, int flags) {
                 propagate_flags &= ~PROPAGATE_AOF;
 
         /* Call propagate() only if at least one of AOF / replication
-         * propagation is needed. */
-        if (propagate_flags != PROPAGATE_NONE)
+         * propagation is needed. Note that modules commands handle replication
+         * in an explicit way, so we never replicate them automatically. */
+        if (propagate_flags != PROPAGATE_NONE && !(c->cmd->flags & CMD_MODULE))
             propagate(c->cmd,c->db->id,c->argv,c->argc,propagate_flags);
     }
 
@@ -2444,8 +2494,9 @@ int processCommand(client *c) {
         return C_OK;
     }
 
-    /* Only allow INFO and SLAVEOF when slave-serve-stale-data is no and
-     * we are a slave with a broken link with master. */
+    /* Only allow commands with flag "t", such as INFO, SLAVEOF and so on,
+     * when slave-serve-stale-data is no and we are a slave with a broken
+     * link with master. */
     if (server.masterhost && server.repl_state != REPL_STATE_CONNECTED &&
         server.repl_serve_stale_data == 0 &&
         !(c->cmd->flags & CMD_STALE))
@@ -2489,7 +2540,7 @@ int processCommand(client *c) {
         call(c,CMD_CALL_FULL);
         c->woff = server.master_repl_offset;
         if (listLength(server.ready_keys))
-            handleClientsBlockedOnLists();
+            handleClientsBlockedOnKeys();
     }
     return C_OK;
 }
@@ -2732,7 +2783,16 @@ void commandCommand(client *c) {
     dictIterator *di;
     dictEntry *de;
 
-    if (c->argc == 1) {
+    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"help")) {
+        const char *help[] = {
+"(no subcommand) -- Return details about all Redis commands.",
+"count -- Return the total number of commands in this Redis server.",
+"getkeys <full-command> -- Return the keys from a full Redis command.",
+"info [command-name ...] -- Return details about multiple Redis commands.",
+NULL
+        };
+        addReplyHelp(c, help);
+    } else if (c->argc == 1) {
         addReplyMultiBulkLen(c, dictSize(server.commands));
         di = dictGetIterator(server.commands);
         while ((de = dictNext(di)) != NULL) {
@@ -2766,8 +2826,7 @@ void commandCommand(client *c) {
         for (j = 0; j < numkeys; j++) addReplyBulk(c,c->argv[keys[j]+2]);
         getKeysFreeResult(keys);
     } else {
-        addReplyError(c, "Unknown subcommand or wrong number of arguments.");
-        return;
+        addReplyErrorFormat(c, "Unknown subcommand or wrong number of arguments for '%s'. Try COMMAND HELP", (char*)c->argv[1]->ptr);
     }
 }
 
@@ -2898,7 +2957,7 @@ sds genRedisInfoString(char *section) {
             "blocked_clients:%d\r\n",
             listLength(server.clients)-listLength(server.slaves),
             lol, bib,
-            server.bpop_blocked_clients);
+            server.blocked_clients);
     }
 
     /* Memory */
@@ -3087,6 +3146,8 @@ sds genRedisInfoString(char *section) {
             "sync_partial_ok:%lld\r\n"
             "sync_partial_err:%lld\r\n"
             "expired_keys:%lld\r\n"
+            "expired_stale_perc:%.2f\r\n"
+            "expired_time_cap_reached_count:%lld\r\n"
             "evicted_keys:%lld\r\n"
             "keyspace_hits:%lld\r\n"
             "keyspace_misses:%lld\r\n"
@@ -3111,6 +3172,8 @@ sds genRedisInfoString(char *section) {
             server.stat_sync_partial_ok,
             server.stat_sync_partial_err,
             server.stat_expiredkeys,
+            server.stat_expired_stale_perc*100,
+            server.stat_expired_time_cap_reached_count,
             server.stat_evictedkeys,
             server.stat_keyspace_hits,
             server.stat_keyspace_misses,
@@ -3539,7 +3602,8 @@ void loadDataFromDisk(void) {
                 rsi.repl_id_is_set &&
                 rsi.repl_offset != -1 &&
                 /* Note that older implementations may save a repl_stream_db
-                 * of -1 inside the RDB file. */
+                 * of -1 inside the RDB file in a wrong way, see more information
+                 * in function rdbPopulateSaveInfo. */
                 rsi.repl_stream_db != -1)
             {
                 memcpy(server.replid,rsi.repl_id,sizeof(server.replid));

src/server.h

@@ -59,6 +59,7 @@ typedef long long mstime_t; /* millisecond time type. */
 #include "anet.h"    /* Networking the easy way */
 #include "ziplist.h" /* Compact list data structure */
 #include "intset.h"  /* Compact integer set structure */
+#include "stream.h"  /* Stream data type header file. */
 #include "version.h" /* Version macro */
 #include "util.h"    /* Misc functions useful in many places */
 #include "latency.h" /* Latency monitor API */
@@ -160,6 +161,7 @@ typedef long long mstime_t; /* millisecond time type. */
 #define CONFIG_DEFAULT_DEFRAG_IGNORE_BYTES (100<<20) /* don't defrag if frag overhead is below 100mb */
 #define CONFIG_DEFAULT_DEFRAG_CYCLE_MIN 25 /* 25% CPU min (at lower threshold) */
 #define CONFIG_DEFAULT_DEFRAG_CYCLE_MAX 75 /* 75% CPU max (at upper threshold) */
+#define CONFIG_DEFAULT_PROTO_MAX_BULK_LEN (512ll*1024*1024) /* Bulk request max size */
 
 #define ACTIVE_EXPIRE_CYCLE_LOOKUPS_PER_LOOP 20 /* Loopkups per loop. */
 #define ACTIVE_EXPIRE_CYCLE_FAST_DURATION 1000 /* Microseconds */
@@ -255,6 +257,9 @@ typedef long long mstime_t; /* millisecond time type. */
 #define BLOCKED_LIST 1    /* BLPOP & co. */
 #define BLOCKED_WAIT 2    /* WAIT for synchronous replication. */
 #define BLOCKED_MODULE 3  /* Blocked by a loadable module. */
+#define BLOCKED_STREAM 4  /* XREAD. */
+#define BLOCKED_AOF 5     /* WAIT AOF waiting for AOF sync. */
+#define BLOCKED_NUM 6     /* Number of blocked states. */
 
 /* Client request types */
 #define PROTO_REQ_INLINE 1
@@ -424,7 +429,8 @@ typedef long long mstime_t; /* millisecond time type. */
 #define NOTIFY_ZSET (1<<7)        /* z */
 #define NOTIFY_EXPIRED (1<<8)     /* x */
 #define NOTIFY_EVICTED (1<<9)     /* e */
-#define NOTIFY_ALL (NOTIFY_GENERIC | NOTIFY_STRING | NOTIFY_LIST | NOTIFY_SET | NOTIFY_HASH | NOTIFY_ZSET | NOTIFY_EXPIRED | NOTIFY_EVICTED)      /* A */
+#define NOTIFY_STREAM (1<<10)     /* t */
+#define NOTIFY_ALL (NOTIFY_GENERIC | NOTIFY_STRING | NOTIFY_LIST | NOTIFY_SET | NOTIFY_HASH | NOTIFY_ZSET | NOTIFY_EXPIRED | NOTIFY_EVICTED | NOTIFY_STREAM) /* A flag */
 
 /* Get the first bind addr or NULL */
 #define NET_FIRST_BIND_ADDR (server.bindaddr_count ? server.bindaddr[0] : NULL)
@@ -446,11 +452,11 @@ typedef long long mstime_t; /* millisecond time type. */
 /* A redis object, that is a type able to hold a string / list / set */
 
 /* The actual Redis Object */
-#define OBJ_STRING 0
-#define OBJ_LIST 1
-#define OBJ_SET 2
-#define OBJ_ZSET 3
-#define OBJ_HASH 4
+#define OBJ_STRING 0    /* String object. */
+#define OBJ_LIST 1      /* List object. */
+#define OBJ_SET 2       /* Set object. */
+#define OBJ_ZSET 3      /* Sorted set object. */
+#define OBJ_HASH 4      /* Hash object. */
 
 /* The "module" object type is a special one that signals that the object
  * is one directly managed by a Redis module. In this case the value points
@@ -463,7 +469,8 @@ typedef long long mstime_t; /* millisecond time type. */
  * by a 64 bit module type ID, which has a 54 bits module-specific signature
  * in order to dispatch the loading to the right module, plus a 10 bits
  * encoding version. */
-#define OBJ_MODULE 5
+#define OBJ_MODULE 5    /* Module object. */
+#define OBJ_STREAM 6    /* Stream object. */
 
 /* Extract encver / signature from a module type ID. */
 #define REDISMODULE_TYPE_ENCVER_BITS 10
@@ -575,6 +582,7 @@ typedef struct RedisModuleDigest {
 #define OBJ_ENCODING_SKIPLIST 7  /* Encoded as skiplist */
 #define OBJ_ENCODING_EMBSTR 8  /* Embedded sds string encoding */
 #define OBJ_ENCODING_QUICKLIST 9 /* Encoded as linked list of ziplists */
+#define OBJ_ENCODING_STREAM 10 /* Encoded as a radix tree of listpacks */
 
 #define LRU_BITS 24
 #define LRU_CLOCK_MAX ((1<<LRU_BITS)-1) /* Max value of obj->lru */
@@ -586,7 +594,7 @@ typedef struct redisObject {
     unsigned encoding:4;
     unsigned lru:LRU_BITS; /* LRU time (relative to global lru_clock) or
                             * LFU data (least significant 8 bits frequency
-                            * and most significant 16 bits decreas time). */
+                            * and most significant 16 bits access time). */
     int refcount;
     void *ptr;
 } robj;
@@ -638,16 +646,24 @@ typedef struct blockingState {
     mstime_t timeout;       /* Blocking operation timeout. If UNIX current time
                              * is > timeout then the operation timed out. */
 
-    /* BLOCKED_LIST */
+    /* BLOCKED_LIST and BLOCKED_STREAM */
     dict *keys;             /* The keys we are waiting to terminate a blocking
-                             * operation such as BLPOP. Otherwise NULL. */
+                             * operation such as BLPOP or XREAD. Or NULL. */
     robj *target;           /* The key that should receive the element,
                              * for BRPOPLPUSH. */
 
+    /* BLOCK_STREAM */
+    size_t xread_count;     /* XREAD COUNT option. */
+    robj *xread_group;      /* XREAD group name. */
+    mstime_t xread_retry_time, xread_retry_ttl;
+
     /* BLOCKED_WAIT */
     int numreplicas;        /* Number of replicas we are waiting for ACK. */
     long long reploffset;   /* Replication offset to reach. */
 
+    /* BLOCKED_AOF */
+    uint64_t aofepoch;      /* WAIT AOF: the sync epoch we are waiting for. */
+
     /* BLOCKED_MODULE */
     void *module_blocked_handle; /* RedisModuleBlockedClient structure.
                                     which is opaque for the Redis core, only
@@ -722,6 +738,7 @@ typedef struct client {
     dict *pubsub_channels;  /* channels a client is interested in (SUBSCRIBE) */
     list *pubsub_patterns;  /* patterns a client is interested in (SUBSCRIBE) */
     sds peerid;             /* Cached peer ID. */
+    listNode *client_list_node; /* list node in client list */
 
     /* Response buffer */
     int bufpos;
@@ -937,6 +954,8 @@ struct redisServer {
     long long stat_numcommands;     /* Number of processed commands */
     long long stat_numconnections;  /* Number of connections received */
     long long stat_expiredkeys;     /* Number of expired keys */
+    double stat_expired_stale_perc; /* Percentage of keys probably expired */
+    long long stat_expired_time_cap_reached_count; /* Early expire cylce stops.*/
     long long stat_evictedkeys;     /* Number of evicted keys (maxmemory) */
     long long stat_keyspace_hits;   /* Number of successful lookups of keys */
     long long stat_keyspace_misses; /* Number of failed lookups of keys */
@@ -1001,7 +1020,7 @@ struct redisServer {
     int aof_fd;       /* File descriptor of currently selected AOF file */
     int aof_selected_db; /* Currently selected DB in AOF */
     time_t aof_flush_postponed_start; /* UNIX time of postponed AOF flush */
-    time_t aof_last_fsync;            /* UNIX time of last fsync() */
+    time_t aof_last_fsync;            /* UNIX time of last fsync() *attempt* */
     time_t aof_rewrite_time_last;   /* Time used by last AOF rewrite run. */
     time_t aof_rewrite_time_start;  /* Current AOF rewrite start time. */
     int aof_lastbgrewrite_status;   /* C_OK or C_ERR */
@@ -1011,6 +1030,8 @@ struct redisServer {
     int aof_last_write_errno;       /* Valid if aof_last_write_status is ERR */
     int aof_load_truncated;         /* Don't stop on unexpected AOF EOF. */
     int aof_use_rdb_preamble;       /* Use RDB preamble on AOF rewrites. */
+    uint64_t aof_fsync_epoch;       /* AOF sync epoch used for WAIT AOF. */
+    uint64_t aof_fsync_in_progress_epoch; /* Current ongoing AOF sync epoch. */
     /* AOF pipes used to communicate between parent and child during rewrite. */
     int aof_pipe_write_data_to_child;
     int aof_pipe_read_data_from_parent;
@@ -1118,10 +1139,12 @@ struct redisServer {
     unsigned long long maxmemory;   /* Max number of memory bytes to use */
     int maxmemory_policy;           /* Policy for key eviction */
     int maxmemory_samples;          /* Pricision of random sampling */
-    unsigned int lfu_log_factor;    /* LFU logarithmic counter factor. */
-    unsigned int lfu_decay_time;    /* LFU counter decay factor. */
+    int lfu_log_factor;             /* LFU logarithmic counter factor. */
+    int lfu_decay_time;             /* LFU counter decay factor. */
+    long long proto_max_bulk_len;   /* Protocol bulk length maximum size. */
     /* Blocked clients */
-    unsigned int bpop_blocked_clients; /* Number of clients blocked by lists */
+    unsigned int blocked_clients;   /* # of clients executing a blocking cmd.*/
+    unsigned int blocked_clients_by_type[BLOCKED_NUM];
     list *unblocked_clients; /* list of clients to unblock before next loop */
     list *ready_keys;        /* List of readyList structures for BLPOP & co */
     /* Sort parameters - qsort_r() is only available under BSD so we
@@ -1288,6 +1311,7 @@ typedef struct {
 extern struct redisServer server;
 extern struct sharedObjectsStruct shared;
 extern dictType objectKeyPointerValueDictType;
+extern dictType objectKeyHeapPointerValueDictType;
 extern dictType setDictType;
 extern dictType zsetDictType;
 extern dictType clusterNodesDictType;
@@ -1319,6 +1343,8 @@ void moduleBlockedClientPipeReadable(aeEventLoop *el, int fd, void *privdata, in
 size_t moduleCount(void);
 void moduleAcquireGIL(void);
 void moduleReleaseGIL(void);
+void moduleNotifyKeyspaceEvent(int type, const char *event, robj *key, int dbid);
+
 
 /* Utils */
 long long ustime(void);
@@ -1357,6 +1383,7 @@ void addReplyDouble(client *c, double d);
 void addReplyHumanLongDouble(client *c, long double d);
 void addReplyLongLong(client *c, long long ll);
 void addReplyMultiBulkLen(client *c, long length);
+void addReplyHelp(client *c, const char **help);
 void copyClientOutputBuffer(client *dst, client *src);
 size_t sdsZmallocSize(sds s);
 size_t getStringObjectSdsUsedMemory(robj *o);
@@ -1385,6 +1412,7 @@ int handleClientsWithPendingWrites(void);
 int clientHasPendingReplies(client *c);
 void unlinkClient(client *c);
 int writeToClient(int fd, client *c, int handler_installed);
+void linkClient(client *c);
 
 #ifdef __GNUC__
 void addReplyErrorFormat(client *c, const char *fmt, ...)
@@ -1410,9 +1438,7 @@ int listTypeEqual(listTypeEntry *entry, robj *o);
 void listTypeDelete(listTypeIterator *iter, listTypeEntry *entry);
 void listTypeConvert(robj *subject, int enc);
 void unblockClientWaitingData(client *c);
-void handleClientsBlockedOnLists(void);
 void popGenericCommand(client *c, int where);
-void signalListAsReady(redisDb *db, robj *key);
 
 /* MULTI/EXEC/WATCH... */
 void unwatchAllKeys(client *c);
@@ -1455,6 +1481,7 @@ robj *createIntsetObject(void);
 robj *createHashObject(void);
 robj *createZsetObject(void);
 robj *createZsetZiplistObject(void);
+robj *createStreamObject(void);
 robj *createModuleObject(moduleType *mt, void *value);
 int getLongFromObjectOrReply(client *c, robj *o, long *target, const char *msg);
 int checkType(client *c, robj *o, int type);
@@ -1492,7 +1519,7 @@ void replicationScriptCacheInit(void);
 void replicationScriptCacheFlush(void);
 void replicationScriptCacheAdd(sds sha1);
 int replicationScriptCacheExists(sds sha1);
-void processClientsWaitingReplicas(void);
+void processClientsBlockedInWait(void);
 void unblockClientWaitingReplicas(client *c);
 int replicationCountAcksByOffset(long long offset);
 void replicationSendNewlineToMaster(void);
@@ -1527,6 +1554,8 @@ void backgroundRewriteDoneHandler(int exitcode, int bysignal);
 void aofRewriteBufferReset(void);
 unsigned long aofRewriteBufferSize(void);
 ssize_t aofReadDiffFromParent(void);
+void aofStartBackgroundFsync(void);
+uint64_t aofNextEpoch(void);
 
 /* Child info */
 void openChildInfoPipe(void);
@@ -1754,6 +1783,7 @@ int *evalGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
 int *sortGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
 int *migrateGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
 int *georadiusGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
+int *xreadGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys);
 
 /* Cluster */
 void clusterInit(void);
@@ -1781,6 +1811,7 @@ void scriptingInit(int setup);
 int ldbRemoveChild(pid_t pid);
 void ldbKillForkedSessions(void);
 int ldbPendingChildren(void);
+sds luaCreateFunction(client *c, lua_State *lua, robj *body);
 
 /* Blocked clients */
 void processUnblockedClients(void);
@@ -1789,6 +1820,9 @@ void unblockClient(client *c);
 void replyToBlockedClientTimedOut(client *c);
 int getTimeoutFromObjectOrReply(client *c, robj *object, mstime_t *timeout, int unit);
 void disconnectAllBlockedClients(void);
+void handleClientsBlockedOnKeys(void);
+void signalKeyAsReady(redisDb *db, robj *key);
+void blockForKeys(client *c, int btype, robj **keys, int numkeys, mstime_t timeout, robj *target, streamID *ids);
 
 /* expire.c -- Handling of expired keys */
 void activeExpireCycle(int type);
@@ -1802,6 +1836,7 @@ void evictionPoolAlloc(void);
 #define LFU_INIT_VAL 5
 unsigned long LFUGetTimeInMinutes(void);
 uint8_t LFULogIncr(uint8_t value);
+unsigned long LFUDecrAndReturn(robj *o);
 
 /* Keys hashing / comparison functions for dict.c hash tables. */
 uint64_t dictSdsHash(const void *key);
@@ -1990,6 +2025,11 @@ void pfdebugCommand(client *c);
 void latencyCommand(client *c);
 void moduleCommand(client *c);
 void securityWarningCommand(client *c);
+void xaddCommand(client *c);
+void xrangeCommand(client *c);
+void xrevrangeCommand(client *c);
+void xlenCommand(client *c);
+void xreadCommand(client *c);
 
 #if defined(__GNUC__)
 void *calloc(size_t count, size_t size) __attribute__ ((deprecated));

src/setproctitle.c

@@ -39,7 +39,11 @@
 #include <errno.h>	/* errno program_invocation_name program_invocation_short_name */
 
 #if !defined(HAVE_SETPROCTITLE)
-#define HAVE_SETPROCTITLE (defined __NetBSD__ || defined __FreeBSD__ || defined __OpenBSD__)
+#if (defined __NetBSD__ || defined __FreeBSD__ || defined __OpenBSD__)
+#define HAVE_SETPROCTITLE 1
+#else
+#define HAVE_SETPROCTITLE 0
+#endif
 #endif
 
 

src/slowlog.c

@@ -140,7 +140,17 @@ void slowlogReset(void) {
 /* The SLOWLOG command. Implements all the subcommands needed to handle the
  * Redis slow log. */
 void slowlogCommand(client *c) {
-    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"reset")) {
+    if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"help")) {
+        const char *help[] = {
+"get [count] -- Return top entries from the slowlog (default: 10)."
+"    Entries are made of:",
+"    id, timestamp, time in microseconds, arguments array, client IP and port, client name",
+"len -- Return the length of the slowlog.",
+"reset -- Reset the slowlog.",
+NULL
+        };
+        addReplyHelp(c, help);
+    } else if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"reset")) {
         slowlogReset();
         addReply(c,shared.ok);
     } else if (c->argc == 2 && !strcasecmp(c->argv[1]->ptr,"len")) {
@@ -177,7 +187,6 @@ void slowlogCommand(client *c) {
         }
         setDeferredMultiBulkLength(c,totentries,sent);
     } else {
-        addReplyError(c,
-            "Unknown SLOWLOG subcommand or wrong # of args. Try GET, RESET, LEN.");
+         addReplyErrorFormat(c, "Unknown subcommand or wrong number of arguments for '%s'. Try SLOWLOG HELP", (char*)c->argv[1]->ptr);
     }
 }

src/stream.h

@@ -0,0 +1,59 @@
+#ifndef STREAM_H
+#define STREAM_H
+
+#include "rax.h"
+#include "listpack.h"
+
+/* Stream item ID: a 128 bit number composed of a milliseconds time and
+ * a sequence counter. IDs generated in the same millisecond (or in a past
+ * millisecond if the clock jumped backward) will use the millisecond time
+ * of the latest generated ID and an incremented sequence. */
+typedef struct streamID {
+    uint64_t ms;        /* Unix time in milliseconds. */
+    uint64_t seq;       /* Sequence number. */
+} streamID;
+
+typedef struct stream {
+    rax *rax;               /* The radix tree holding the stream. */
+    uint64_t length;        /* Number of elements inside this stream. */
+    streamID last_id;       /* Zero if there are yet no items. */
+} stream;
+
+/* We define an iterator to iterate stream items in an abstract way, without
+ * caring about the radix tree + listpack representation. Technically speaking
+ * the iterator is only used inside streamReplyWithRange(), so could just
+ * be implemented inside the function, but practically there is the AOF
+ * rewriting code that also needs to iterate the stream to emit the XADD
+ * commands. */
+typedef struct streamIterator {
+    streamID master_id;     /* ID of the master entry at listpack head. */
+    uint64_t master_fields_count;       /* Master entries # of fields. */
+    unsigned char *master_fields_start; /* Master entries start in listpack. */
+    unsigned char *master_fields_ptr;   /* Master field to emit next. */
+    int entry_flags;                    /* Flags of entry we are emitting. */
+    int rev;                /* True if iterating end to start (reverse). */
+    uint64_t start_key[2];  /* Start key as 128 bit big endian. */
+    uint64_t end_key[2];    /* End key as 128 bit big endian. */
+    raxIterator ri;         /* Rax iterator. */
+    unsigned char *lp;      /* Current listpack. */
+    unsigned char *lp_ele;  /* Current listpack cursor. */
+    /* Buffers used to hold the string of lpGet() when the element is
+     * integer encoded, so that there is no string representation of the
+     * element inside the listpack itself. */
+    unsigned char field_buf[LP_INTBUF_SIZE];
+    unsigned char value_buf[LP_INTBUF_SIZE];
+} streamIterator;
+
+/* Prototypes of exported APIs. */
+
+struct client;
+
+stream *streamNew(void);
+void freeStream(stream *s);
+size_t streamReplyWithRange(struct client *c, stream *s, streamID *start, streamID *end, size_t count, int rev);
+void streamIteratorStart(streamIterator *si, stream *s, streamID *start, streamID *end, int rev);
+int streamIteratorGetID(streamIterator *si, streamID *id, int64_t *numfields);
+void streamIteratorGetField(streamIterator *si, unsigned char **fieldptr, unsigned char **valueptr, int64_t *fieldlen, int64_t *valuelen);
+void streamIteratorStop(streamIterator *si);
+
+#endif

src/t_hash.c

@@ -287,8 +287,8 @@ int hashTypeDelete(robj *o, sds field) {
         if (fptr != NULL) {
             fptr = ziplistFind(fptr, (unsigned char*)field, sdslen(field), 1);
             if (fptr != NULL) {
-                zl = ziplistDelete(zl,&fptr);
-                zl = ziplistDelete(zl,&fptr);
+                zl = ziplistDelete(zl,&fptr); /* Delete the key. */
+                zl = ziplistDelete(zl,&fptr); /* Delete the value. */
                 o->ptr = zl;
                 deleted = 1;
             }
@@ -616,7 +616,7 @@ void hincrbyfloatCommand(client *c) {
 
     value += incr;
 
-    char buf[256];
+    char buf[MAX_LONG_DOUBLE_CHARS];
     int len = ld2string(buf,sizeof(buf),value,1);
     new = sdsnewlen(buf,len);
     hashTypeSet(o,c->argv[2]->ptr,new,HASH_SET_TAKE_VALUE);

src/t_list.c

@@ -603,119 +603,6 @@ void rpoplpushCommand(client *c) {
  * Blocking POP operations
  *----------------------------------------------------------------------------*/
 
-/* This is how the current blocking POP works, we use BLPOP as example:
- * - If the user calls BLPOP and the key exists and contains a non empty list
- *   then LPOP is called instead. So BLPOP is semantically the same as LPOP
- *   if blocking is not required.
- * - If instead BLPOP is called and the key does not exists or the list is
- *   empty we need to block. In order to do so we remove the notification for
- *   new data to read in the client socket (so that we'll not serve new
- *   requests if the blocking request is not served). Also we put the client
- *   in a dictionary (db->blocking_keys) mapping keys to a list of clients
- *   blocking for this keys.
- * - If a PUSH operation against a key with blocked clients waiting is
- *   performed, we mark this key as "ready", and after the current command,
- *   MULTI/EXEC block, or script, is executed, we serve all the clients waiting
- *   for this list, from the one that blocked first, to the last, accordingly
- *   to the number of elements we have in the ready list.
- */
-
-/* Set a client in blocking mode for the specified key, with the specified
- * timeout */
-void blockForKeys(client *c, robj **keys, int numkeys, mstime_t timeout, robj *target) {
-    dictEntry *de;
-    list *l;
-    int j;
-
-    c->bpop.timeout = timeout;
-    c->bpop.target = target;
-
-    if (target != NULL) incrRefCount(target);
-
-    for (j = 0; j < numkeys; j++) {
-        /* If the key already exists in the dict ignore it. */
-        if (dictAdd(c->bpop.keys,keys[j],NULL) != DICT_OK) continue;
-        incrRefCount(keys[j]);
-
-        /* And in the other "side", to map keys -> clients */
-        de = dictFind(c->db->blocking_keys,keys[j]);
-        if (de == NULL) {
-            int retval;
-
-            /* For every key we take a list of clients blocked for it */
-            l = listCreate();
-            retval = dictAdd(c->db->blocking_keys,keys[j],l);
-            incrRefCount(keys[j]);
-            serverAssertWithInfo(c,keys[j],retval == DICT_OK);
-        } else {
-            l = dictGetVal(de);
-        }
-        listAddNodeTail(l,c);
-    }
-    blockClient(c,BLOCKED_LIST);
-}
-
-/* Unblock a client that's waiting in a blocking operation such as BLPOP.
- * You should never call this function directly, but unblockClient() instead. */
-void unblockClientWaitingData(client *c) {
-    dictEntry *de;
-    dictIterator *di;
-    list *l;
-
-    serverAssertWithInfo(c,NULL,dictSize(c->bpop.keys) != 0);
-    di = dictGetIterator(c->bpop.keys);
-    /* The client may wait for multiple keys, so unblock it for every key. */
-    while((de = dictNext(di)) != NULL) {
-        robj *key = dictGetKey(de);
-
-        /* Remove this client from the list of clients waiting for this key. */
-        l = dictFetchValue(c->db->blocking_keys,key);
-        serverAssertWithInfo(c,key,l != NULL);
-        listDelNode(l,listSearchKey(l,c));
-        /* If the list is empty we need to remove it to avoid wasting memory */
-        if (listLength(l) == 0)
-            dictDelete(c->db->blocking_keys,key);
-    }
-    dictReleaseIterator(di);
-
-    /* Cleanup the client structure */
-    dictEmpty(c->bpop.keys,NULL);
-    if (c->bpop.target) {
-        decrRefCount(c->bpop.target);
-        c->bpop.target = NULL;
-    }
-}
-
-/* If the specified key has clients blocked waiting for list pushes, this
- * function will put the key reference into the server.ready_keys list.
- * Note that db->ready_keys is a hash table that allows us to avoid putting
- * the same key again and again in the list in case of multiple pushes
- * made by a script or in the context of MULTI/EXEC.
- *
- * The list will be finally processed by handleClientsBlockedOnLists() */
-void signalListAsReady(redisDb *db, robj *key) {
-    readyList *rl;
-
-    /* No clients blocking for this key? No need to queue it. */
-    if (dictFind(db->blocking_keys,key) == NULL) return;
-
-    /* Key was already signaled? No need to queue it again. */
-    if (dictFind(db->ready_keys,key) != NULL) return;
-
-    /* Ok, we need to queue this key into server.ready_keys. */
-    rl = zmalloc(sizeof(*rl));
-    rl->key = key;
-    rl->db = db;
-    incrRefCount(key);
-    listAddNodeTail(server.ready_keys,rl);
-
-    /* We also add the key in the db->ready_keys dictionary in order
-     * to avoid adding it multiple times into a list with a simple O(1)
-     * check. */
-    incrRefCount(key);
-    serverAssert(dictAdd(db->ready_keys,key,NULL) == DICT_OK);
-}
-
 /* This is a helper function for handleClientsBlockedOnLists(). It's work
  * is to serve a specific client (receiver) that is blocked on 'key'
  * in the context of the specified 'db', doing the following:
@@ -785,97 +672,6 @@ int serveClientBlockedOnList(client *receiver, robj *key, robj *dstkey, redisDb
     return C_OK;
 }
 
-/* This function should be called by Redis every time a single command,
- * a MULTI/EXEC block, or a Lua script, terminated its execution after
- * being called by a client.
- *
- * All the keys with at least one client blocked that received at least
- * one new element via some PUSH operation are accumulated into
- * the server.ready_keys list. This function will run the list and will
- * serve clients accordingly. Note that the function will iterate again and
- * again as a result of serving BRPOPLPUSH we can have new blocking clients
- * to serve because of the PUSH side of BRPOPLPUSH. */
-void handleClientsBlockedOnLists(void) {
-    while(listLength(server.ready_keys) != 0) {
-        list *l;
-
-        /* Point server.ready_keys to a fresh list and save the current one
-         * locally. This way as we run the old list we are free to call
-         * signalListAsReady() that may push new elements in server.ready_keys
-         * when handling clients blocked into BRPOPLPUSH. */
-        l = server.ready_keys;
-        server.ready_keys = listCreate();
-
-        while(listLength(l) != 0) {
-            listNode *ln = listFirst(l);
-            readyList *rl = ln->value;
-
-            /* First of all remove this key from db->ready_keys so that
-             * we can safely call signalListAsReady() against this key. */
-            dictDelete(rl->db->ready_keys,rl->key);
-
-            /* If the key exists and it's a list, serve blocked clients
-             * with data. */
-            robj *o = lookupKeyWrite(rl->db,rl->key);
-            if (o != NULL && o->type == OBJ_LIST) {
-                dictEntry *de;
-
-                /* We serve clients in the same order they blocked for
-                 * this key, from the first blocked to the last. */
-                de = dictFind(rl->db->blocking_keys,rl->key);
-                if (de) {
-                    list *clients = dictGetVal(de);
-                    int numclients = listLength(clients);
-
-                    while(numclients--) {
-                        listNode *clientnode = listFirst(clients);
-                        client *receiver = clientnode->value;
-                        robj *dstkey = receiver->bpop.target;
-                        int where = (receiver->lastcmd &&
-                                     receiver->lastcmd->proc == blpopCommand) ?
-                                    LIST_HEAD : LIST_TAIL;
-                        robj *value = listTypePop(o,where);
-
-                        if (value) {
-                            /* Protect receiver->bpop.target, that will be
-                             * freed by the next unblockClient()
-                             * call. */
-                            if (dstkey) incrRefCount(dstkey);
-                            unblockClient(receiver);
-
-                            if (serveClientBlockedOnList(receiver,
-                                rl->key,dstkey,rl->db,value,
-                                where) == C_ERR)
-                            {
-                                /* If we failed serving the client we need
-                                 * to also undo the POP operation. */
-                                    listTypePush(o,value,where);
-                            }
-
-                            if (dstkey) decrRefCount(dstkey);
-                            decrRefCount(value);
-                        } else {
-                            break;
-                        }
-                    }
-                }
-
-                if (listTypeLength(o) == 0) {
-                    dbDelete(rl->db,rl->key);
-                }
-                /* We don't call signalModifiedKey() as it was already called
-                 * when an element was pushed on the list. */
-            }
-
-            /* Free this item. */
-            decrRefCount(rl->key);
-            zfree(rl);
-            listDelNode(l,ln);
-        }
-        listRelease(l); /* We have the new list on place at this point. */
-    }
-}
-
 /* Blocking RPOP/LPOP */
 void blockingPopGenericCommand(client *c, int where) {
     robj *o;
@@ -930,7 +726,7 @@ void blockingPopGenericCommand(client *c, int where) {
     }
 
     /* If the list is empty or the key does not exists we must block */
-    blockForKeys(c, c->argv + 1, c->argc - 2, timeout, NULL);
+    blockForKeys(c,BLOCKED_LIST,c->argv + 1,c->argc - 2,timeout,NULL,NULL);
 }
 
 void blpopCommand(client *c) {
@@ -956,7 +752,7 @@ void brpoplpushCommand(client *c) {
             addReply(c, shared.nullbulk);
         } else {
             /* The list is empty and the client blocks. */
-            blockForKeys(c, c->argv + 1, 1, timeout, c->argv[2]);
+            blockForKeys(c,BLOCKED_LIST,c->argv + 1,1,timeout,c->argv[2],NULL);
         }
     } else {
         if (key->type != OBJ_LIST) {

src/t_set.c

@@ -407,7 +407,7 @@ void spopWithCountCommand(client *c) {
     /* Get the count argument */
     if (getLongFromObjectOrReply(c,c->argv[2],&l,NULL) != C_OK) return;
     if (l >= 0) {
-        count = (unsigned) l;
+        count = (unsigned long) l;
     } else {
         addReply(c,shared.outofrangeerr);
         return;
@@ -626,7 +626,7 @@ void srandmemberWithCountCommand(client *c) {
 
     if (getLongFromObjectOrReply(c,c->argv[2],&l,NULL) != C_OK) return;
     if (l >= 0) {
-        count = (unsigned) l;
+        count = (unsigned long) l;
     } else {
         /* A negative count means: return the same elements multiple times
          * (i.e. don't remove the extracted element after every extraction). */
@@ -774,15 +774,21 @@ void srandmemberCommand(client *c) {
 }
 
 int qsortCompareSetsByCardinality(const void *s1, const void *s2) {
-    return setTypeSize(*(robj**)s1)-setTypeSize(*(robj**)s2);
+    if (setTypeSize(*(robj**)s1) > setTypeSize(*(robj**)s2)) return 1;
+    if (setTypeSize(*(robj**)s1) < setTypeSize(*(robj**)s2)) return -1;
+    return 0;
 }
 
 /* This is used by SDIFF and in this case we can receive NULL that should
  * be handled as empty sets. */
 int qsortCompareSetsByRevCardinality(const void *s1, const void *s2) {
     robj *o1 = *(robj**)s1, *o2 = *(robj**)s2;
+    unsigned long first = o1 ? setTypeSize(o1) : 0;
+    unsigned long second = o2 ? setTypeSize(o2) : 0;
 
-    return  (o2 ? setTypeSize(o2) : 0) - (o1 ? setTypeSize(o1) : 0);
+    if (first < second) return 1;
+    if (first > second) return -1;
+    return 0;
 }
 
 void sinterGenericCommand(client *c, robj **setkeys,

src/util.c

@@ -84,7 +84,7 @@ int stringmatchlen(const char *pattern, int patternLen,
             }
             match = 0;
             while(1) {
-                if (pattern[0] == '\\') {
+                if (pattern[0] == '\\' && patternLen >= 2) {
                     pattern++;
                     patternLen--;
                     if (pattern[0] == string[0])

src/util.h

@@ -33,6 +33,11 @@
 #include <stdint.h>
 #include "sds.h"
 
+/* The maximum number of characters needed to represent a long double
+ * as a string (long double has a huge range).
+ * This should be the size of the buffer given to ld2string */
+#define MAX_LONG_DOUBLE_CHARS 5*1024
+
 int stringmatchlen(const char *p, int plen, const char *s, int slen, int nocase);
 int stringmatch(const char *p, const char *s, int nocase);
 long long memtoll(const char *p, int *err);

src/ziplist.c

@@ -53,20 +53,20 @@
  * <prevlen> <encoding>
  *
  * The length of the previous entry, <prevlen>, is encoded in the following way:
- * If this length is smaller than 255 bytes, it will only consume a single
+ * If this length is smaller than 254 bytes, it will only consume a single
  * byte representing the length as an unsinged 8 bit integer. When the length
- * is greater than or equal to 255, it will consume 5 bytes. The first byte is
- * set to 255 (FF) to indicate a larger value is following. The remaining 4
+ * is greater than or equal to 254, it will consume 5 bytes. The first byte is
+ * set to 254 (FE) to indicate a larger value is following. The remaining 4
  * bytes take the length of the previous entry as value.
  *
  * So practically an entry is encoded in the following way:
  *
- * <prevlen from 0 to 254> <encoding> <entry>
+ * <prevlen from 0 to 253> <encoding> <entry>
  *
- * Or alternatively if the previous entry length is greater than 254 bytes
+ * Or alternatively if the previous entry length is greater than 253 bytes
  * the following encoding is used:
  *
- * 0xFF <4 bytes unsigned little endian prevlen> <encoding> <entry>
+ * 0xFE <4 bytes unsigned little endian prevlen> <encoding> <entry>
  *
  * The encoding field of the entry depends on the content of the
  * entry. When the entry is a string, the first 2 bits of the encoding first
@@ -440,7 +440,7 @@ unsigned int zipStorePrevEntryLength(unsigned char *p, unsigned int len) {
     if ((prevlensize) == 1) {                                                  \
         (prevlen) = (ptr)[0];                                                  \
     } else if ((prevlensize) == 5) {                                           \
-        assert(sizeof((prevlensize)) == 4);                                    \
+        assert(sizeof((prevlen)) == 4);                                    \
         memcpy(&(prevlen), ((char*)(ptr)) + 1, 4);                             \
         memrev32ifbe(&prevlen);                                                \
     }                                                                          \

tests/instances.tcl

@@ -318,7 +318,7 @@ proc end_tests {} {
         puts "GOOD! No errors."
         exit 0
     } else {
-        puts "WARNING $::failed tests faield."
+        puts "WARNING $::failed test(s) failed."
         exit 1
     }
 }

tests/integration/psync2.tcl

@@ -10,7 +10,7 @@ start_server {} {
     # Config
     set debug_msg 0                 ; # Enable additional debug messages
 
-    set no_exit 0;                  ; # Do not exit at end of the test
+    set no_exit 0                   ; # Do not exit at end of the test
 
     set duration 20                 ; # Total test seconds
 
@@ -175,6 +175,69 @@ start_server {} {
         assert {$sync_count == $new_sync_count}
     }
 
+    test "PSYNC2: Slave RDB restart with EVALSHA in backlog issue #4483" {
+        # Pick a random slave
+        set slave_id [expr {($master_id+1)%5}]
+        set sync_count [status $R($master_id) sync_full]
+
+        # Make sure to replicate the first EVAL while the salve is online
+        # so that it's part of the scripts the master believes it's safe
+        # to propagate as EVALSHA.
+        $R($master_id) EVAL {return redis.call("incr","__mycounter")} 0
+        $R($master_id) EVALSHA e6e0b547500efcec21eddb619ac3724081afee89 0
+
+        # Wait for the two to sync
+        wait_for_condition 50 1000 {
+            [$R($master_id) debug digest] == [$R($slave_id) debug digest]
+        } else {
+            fail "Slave not reconnecting"
+        }
+
+        # Prevent the slave from receiving master updates, and at
+        # the same time send a new script several times to the
+        # master, so that we'll end with EVALSHA into the backlog.
+        $R($slave_id) slaveof 127.0.0.1 0
+
+        $R($master_id) EVALSHA e6e0b547500efcec21eddb619ac3724081afee89 0
+        $R($master_id) EVALSHA e6e0b547500efcec21eddb619ac3724081afee89 0
+        $R($master_id) EVALSHA e6e0b547500efcec21eddb619ac3724081afee89 0
+
+        catch {
+            $R($slave_id) config rewrite
+            $R($slave_id) debug restart
+        }
+
+        # Reconfigure the slave correctly again, when it's back online.
+        set retry 50
+        while {$retry} {
+            if {[catch {
+                $R($slave_id) slaveof $master_host $master_port
+            }]} {
+                after 1000
+            } else {
+                break
+            }
+            incr retry -1
+        }
+
+        # The master should be back at 4 slaves eventually
+        wait_for_condition 50 1000 {
+            [status $R($master_id) connected_slaves] == 4
+        } else {
+            fail "Slave not reconnecting"
+        }
+        set new_sync_count [status $R($master_id) sync_full]
+        assert {$sync_count == $new_sync_count}
+
+        # However if the slave started with the full state of the
+        # scripting engine, we should now have the same digest.
+        wait_for_condition 50 1000 {
+            [$R($master_id) debug digest] == [$R($slave_id) debug digest]
+        } else {
+            fail "Debug digest mismatch between master and slave in post-restart handshake"
+        }
+    }
+
     if {$no_exit} {
         while 1 { puts -nonewline .; flush stdout; after 1000}
     }

tests/integration/replication-2.tcl

@@ -2,9 +2,12 @@ start_server {tags {"repl"}} {
     start_server {} {
         test {First server should have role slave after SLAVEOF} {
             r -1 slaveof [srv 0 host] [srv 0 port]
-            after 1000
-            s -1 role
-        } {slave}
+            wait_for_condition 50 100 {
+                [s -1 master_link_status] eq {up}
+            } else {
+                fail "Replication not started."
+            }
+        }
 
         test {If min-slaves-to-write is honored, write is accepted} {
             r config set min-slaves-to-write 1

tests/integration/replication-3.tcl

@@ -100,7 +100,6 @@ start_server {tags {"repl"}} {
                 close $fd
                 puts "Master - Slave inconsistency"
                 puts "Run diff -u against /tmp/repldump*.txt for more info"
-
             }
 
             set old_digest [r debug digest]
@@ -109,5 +108,27 @@ start_server {tags {"repl"}} {
             set new_digest [r debug digest]
             assert {$old_digest eq $new_digest}
         }
+
+        test {SLAVE can reload "lua" AUX RDB fields of duplicated scripts} {
+            # Force a Slave full resynchronization
+            r debug change-repl-id
+            r -1 client kill type master
+
+            # Check that after a full resync the slave can still load
+            # correctly the RDB file: such file will contain "lua" AUX
+            # sections with scripts already in the memory of the master.
+
+            wait_for_condition 50 100 {
+                [s -1 master_link_status] eq {up}
+            } else {
+                fail "Replication not started."
+            }
+
+            wait_for_condition 50 100 {
+                [r debug digest] eq [r -1 debug digest]
+            } else {
+                fail "DEBUG DIGEST mismatch after full SYNC with many scripts"
+            }
+        }
     }
 }

tests/test_helper.tcl

@@ -26,6 +26,7 @@ set ::all_tests {
     unit/type/set
     unit/type/zset
     unit/type/hash
+    unit/type/stream
     unit/sort
     unit/expire
     unit/other

tests/unit/dump.tcl

@@ -308,4 +308,28 @@ start_server {tags {"dump"}} {
         }
     }
 
+    test {MIGRATE AUTH: correct and wrong password cases} {
+        set first [srv 0 client]
+        r del list
+        r lpush list a b c d
+        start_server {tags {"repl"}} {
+            set second [srv 0 client]
+            set second_host [srv 0 host]
+            set second_port [srv 0 port]
+            $second config set requirepass foobar
+            $second auth foobar
+
+            assert {[$first exists list] == 1}
+            assert {[$second exists list] == 0}
+            set ret [r -1 migrate $second_host $second_port list 9 5000 AUTH foobar]
+            assert {$ret eq {OK}}
+            assert {[$second exists list] == 1}
+            assert {[$second lrange list 0 -1] eq {d c b a}}
+
+            r -1 lpush list a b c d
+            $second config set requirepass foobar2
+            catch {r -1 migrate $second_host $second_port list 9 5000 AUTH foobar} err
+            assert_match {*invalid password*} $err
+        }
+    }
 }

tests/unit/latency-monitor.tcl

@@ -47,4 +47,18 @@ start_server {tags {"latency-monitor"}} {
         assert {[r latency reset] > 0}
         assert {[r latency latest] eq {}}
     }
+
+    test {LATENCY of expire events are correctly collected} {
+        r config set latency-monitor-threshold 20
+        r eval {
+            local i = 0
+            while (i < 1000000) do
+                redis.call('sadd','mybigkey',i)
+                i = i+1
+             end
+        } 0
+        r pexpire mybigkey 1
+        after 500
+        assert_match {*expire-cycle*} [r latency latest]
+    }
 }

tests/unit/type/incr.tcl

@@ -144,4 +144,11 @@ start_server {tags {"incr"}} {
         r set foo 1
         roundFloat [r incrbyfloat foo -1.1]
     } {-0.1}
+
+    test {string to double with null terminator} {
+        r set foo 1
+        r setrange foo 2 2
+        catch {r incrbyfloat foo 1} err
+        format $err
+    } {ERR*valid*}
 }

tests/unit/type/stream.tcl

@@ -0,0 +1,256 @@
+# return value is like strcmp() and similar.
+proc streamCompareID {a b} {
+    if {$a eq $b} {return 0}
+    lassign [split $a -] a_ms a_seq
+    lassign [split $b -] b_ms b_seq
+    if {$a_ms > $b_ms} {return 1}
+    if {$a_ms < $b_ms} {return -1}
+    # Same ms case, compare seq.
+    if {$a_seq > $b_seq} {return 1}
+    if {$a_seq < $b_seq} {return -1}
+}
+
+# return the ID immediately greater than the specified one.
+# Note that this function does not care to handle 'seq' overflow
+# since it's a 64 bit value.
+proc streamNextID {id} {
+    lassign [split $id -] ms seq
+    incr seq
+    join [list $ms $seq] -
+}
+
+# Generate a random stream entry ID with the ms part between min and max
+# and a low sequence number (0 - 999 range), in order to stress test
+# XRANGE against a Tcl implementation implementing the same concept
+# with Tcl-only code in a linear array.
+proc streamRandomID {min_id max_id} {
+    lassign [split $min_id -] min_ms min_seq
+    lassign [split $max_id -] max_ms max_seq
+    set delta [expr {$max_ms-$min_ms+1}]
+    set ms [expr {$min_ms+[randomInt $delta]}]
+    set seq [randomInt 1000]
+    return $ms-$seq
+}
+
+# Tcl-side implementation of XRANGE to perform fuzz testing in the Redis
+# XRANGE implementation.
+proc streamSimulateXRANGE {items start end} {
+    set res {}
+    foreach i $items  {
+        set this_id [lindex $i 0]
+        if {[streamCompareID $this_id $start] >= 0} {
+            if {[streamCompareID $this_id $end] <= 0} {
+                lappend res $i
+            }
+        }
+    }
+    return $res
+}
+
+set content {} ;# Will be populated with Tcl side copy of the stream content.
+
+start_server {
+    tags {"stream"}
+} {
+    test {XADD can add entries into a stream that XRANGE can fetch} {
+        r XADD mystream * item 1 value a
+        r XADD mystream * item 2 value b
+        assert_equal 2 [r XLEN mystream]
+        set items [r XRANGE mystream - +]
+        assert_equal [lindex $items 0 1] {item 1 value a}
+        assert_equal [lindex $items 1 1] {item 2 value b}
+    }
+
+    test {XADD IDs are incremental} {
+        set id1 [r XADD mystream * item 1 value a]
+        set id2 [r XADD mystream * item 2 value b]
+        set id3 [r XADD mystream * item 3 value c]
+        assert {[streamCompareID $id1 $id2] == -1}
+        assert {[streamCompareID $id2 $id3] == -1}
+    }
+
+    test {XADD IDs are incremental when ms is the same as well} {
+        r multi
+        r XADD mystream * item 1 value a
+        r XADD mystream * item 2 value b
+        r XADD mystream * item 3 value c
+        lassign [r exec] id1 id2 id3
+        assert {[streamCompareID $id1 $id2] == -1}
+        assert {[streamCompareID $id2 $id3] == -1}
+    }
+
+    test {XADD with MAXLEN option} {
+        r DEL mystream
+        for {set j 0} {$j < 1000} {incr j} {
+            if {rand() < 0.9} {
+                r XADD mystream MAXLEN 5 * xitem $j
+            } else {
+                r XADD mystream MAXLEN 5 * yitem $j
+            }
+        }
+        set res [r xrange mystream - +]
+        set expected 995
+        foreach r $res {
+            assert {[lindex $r 1 1] == $expected}
+            incr expected
+        }
+    }
+
+    test {XADD mass insertion and XLEN} {
+        r DEL mystream
+        r multi
+        for {set j 0} {$j < 10000} {incr j} {
+            # From time to time insert a field with a different set
+            # of fields in order to stress the stream compression code.
+            if {rand() < 0.9} {
+                r XADD mystream * item $j
+            } else {
+                r XADD mystream * item $j otherfield foo
+            }
+        }
+        r exec
+
+        set items [r XRANGE mystream - +]
+        for {set j 0} {$j < 10000} {incr j} {
+            assert {[lrange [lindex $items $j 1] 0 1] eq [list item $j]}
+        }
+        assert {[r xlen mystream] == $j}
+    }
+
+    test {XRANGE COUNT works as expected} {
+        assert {[llength [r xrange mystream - + COUNT 10]] == 10}
+    }
+
+    test {XREVRANGE COUNT works as expected} {
+        assert {[llength [r xrevrange mystream + - COUNT 10]] == 10}
+    }
+
+    test {XRANGE can be used to iterate the whole stream} {
+        set last_id "-"
+        set j 0
+        while 1 {
+            set elements [r xrange mystream $last_id + COUNT 100]
+            if {[llength $elements] == 0} break
+            foreach e $elements {
+                assert {[lrange [lindex $e 1] 0 1] eq [list item $j]}
+                incr j;
+            }
+            set last_id [streamNextID [lindex $elements end 0]]
+        }
+        assert {$j == 10000}
+    }
+
+    test {XREVRANGE returns the reverse of XRANGE} {
+        assert {[r xrange mystream - +] == [lreverse [r xrevrange mystream + -]]}
+    }
+
+    test {XREAD with non empty stream} {
+        set res [r XREAD COUNT 1 STREAMS mystream 0.0]
+        assert {[lrange [lindex $res 0 1 0 1] 0 1] eq {item 0}}
+    }
+
+    test {Non blocking XREAD with empty streams} {
+        set res [r XREAD STREAMS s1 s2 0.0 0.0]
+        assert {$res eq {}}
+    }
+
+    test {XREAD with non empty second stream} {
+        set res [r XREAD COUNT 1 STREAMS nostream mystream 0.0 0.0]
+        assert {[lindex $res 0 0] eq {mystream}}
+        assert {[lrange [lindex $res 0 1 0 1] 0 1] eq {item 0}}
+    }
+
+    test {Blocking XREAD waiting new data} {
+        r XADD s2 * old abcd1234
+        set rd [redis_deferring_client]
+        $rd XREAD BLOCK 20000 STREAMS s1 s2 s3 $ $ $
+        r XADD s2 * new abcd1234
+        set res [$rd read]
+        assert {[lindex $res 0 0] eq {s2}}
+        assert {[lindex $res 0 1 0 1] eq {new abcd1234}}
+    }
+
+    test {Blocking XREAD waiting old data} {
+        set rd [redis_deferring_client]
+        $rd XREAD BLOCK 20000 STREAMS s1 s2 s3 $ 0.0 $
+        r XADD s2 * foo abcd1234
+        set res [$rd read]
+        assert {[lindex $res 0 0] eq {s2}}
+        assert {[lindex $res 0 1 0 1] eq {old abcd1234}}
+    }
+
+    test "XREAD: XADD + DEL should not awake client" {
+        set rd [redis_deferring_client]
+        r del s1
+        $rd XREAD BLOCK 20000 STREAMS s1 $
+        r multi
+        r XADD s1 * old abcd1234
+        r DEL s1
+        r exec
+        r XADD s1 * new abcd1234
+        set res [$rd read]
+        assert {[lindex $res 0 0] eq {s1}}
+        assert {[lindex $res 0 1 0 1] eq {new abcd1234}}
+    }
+
+    test "XREAD: XADD + DEL + LPUSH should not awake client" {
+        set rd [redis_deferring_client]
+        r del s1
+        $rd XREAD BLOCK 20000 STREAMS s1 $
+        r multi
+        r XADD s1 * old abcd1234
+        r DEL s1
+        r LPUSH s1 foo bar
+        r exec
+        r DEL s1
+        r XADD s1 * new abcd1234
+        set res [$rd read]
+        assert {[lindex $res 0 0] eq {s1}}
+        assert {[lindex $res 0 1 0 1] eq {new abcd1234}}
+    }
+
+    test {XREAD with same stream name multiple times should work} {
+        r XADD s2 * old abcd1234
+        set rd [redis_deferring_client]
+        $rd XREAD BLOCK 20000 STREAMS s2 s2 s2 $ $ $
+        r XADD s2 * new abcd1234
+        set res [$rd read]
+        assert {[lindex $res 0 0] eq {s2}}
+        assert {[lindex $res 0 1 0 1] eq {new abcd1234}}
+    }
+
+    test {XREAD + multiple XADD inside transaction} {
+        r XADD s2 * old abcd1234
+        set rd [redis_deferring_client]
+        $rd XREAD BLOCK 20000 STREAMS s2 s2 s2 $ $ $
+        r MULTI
+        r XADD s2 * field one
+        r XADD s2 * field two
+        r XADD s2 * field three
+        r EXEC
+        set res [$rd read]
+        assert {[lindex $res 0 0] eq {s2}}
+        assert {[lindex $res 0 1 0 1] eq {field one}}
+        assert {[lindex $res 0 1 1 1] eq {field two}}
+    }
+
+    test {XRANGE fuzzing} {
+        set low_id [lindex $items 0 0]
+        set high_id [lindex $items end 0]
+        for {set j 0} {$j < 100} {incr j} {
+            set start [streamRandomID $low_id $high_id]
+            set end [streamRandomID $low_id $high_id]
+            set range [r xrange mystream $start $end]
+            set tcl_range [streamSimulateXRANGE $items $start $end]
+            if {$range ne $tcl_range} {
+                puts "*** WARNING *** - XRANGE fuzzing mismatch: $start - $end"
+                puts "---"
+                puts "XRANGE: '$range'"
+                puts "---"
+                puts "TCL: '$tcl_range'"
+                puts "---"
+                fail "XRANGE fuzzing failed, check logs for details"
+            }
+        }
+    }
+}