update deb changelog

Backport the time thread code from the enterprise version of KeyDB.

Also fix an issue with clustering where connections are potentially written to after being closed.

.github/workflows/ci.yml

@@ -8,52 +8,41 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v2
-      with:
-        submodules: recursive
     - name: make
       run: |
           sudo apt-get update
-          sudo apt-get -y remove libzstd || true
-          sudo apt-get -y install uuid-dev libcurl4-openssl-dev libbz2-dev zlib1g-dev libsnappy-dev liblz4-dev libzstd-dev libgflags-dev
-          make BUILD_TLS=yes -j2 KEYDB_CFLAGS='-Werror' KEYDB_CXXFLAGS='-Werror'
+          sudo apt-get -y install uuid-dev libcurl4-openssl-dev
+          make KEYDB_CFLAGS='-Werror' KEYDB_CXXFLAGS='-Werror' BUILD_TLS=yes -j2
     - name: gen-cert
       run: ./utils/gen-test-certs.sh
     - name: test-tls
       run: |
         sudo apt-get -y install tcl tcl-tls
-        ./runtest --clients 1 --verbose --tls --config server-threads 3
+        ./runtest --clients 2 --verbose --tls
     - name: cluster-test
       run: |
-        ./runtest-cluster --tls --config server-threads 3
+        ./runtest-cluster --tls
     - name: sentinel test
       run: |
           ./runtest-sentinel
     - name: module tests
       run: |
           ./runtest-moduleapi
-    - name: rotation test
-      run: |
-          ./runtest-rotation
         
   build-ubuntu-old:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-16.04
     steps:
-    - uses: actions/checkout@v2
-      with:
-        submodules: recursive
+    - uses: actions/checkout@v1
     - name: make -j2
       run: | 
         sudo apt-get update
-        sudo apt-get -y remove libzstd || true
-        sudo apt-get -y install uuid-dev libcurl4-openssl-dev libbz2-dev zlib1g-dev libsnappy-dev liblz4-dev libzstd-dev libgflags-dev
+        sudo apt-get -y install uuid-dev libcurl4-openssl-dev
         make -j2
 
   build-macos-latest:
     runs-on: macos-latest
     steps:
     - uses: actions/checkout@v2
-      with:
-        submodules: recursive
     - name: make
       run: make KEYDB_CFLAGS='-Werror' KEYDB_CXXFLAGS='-Werror' -j2
 
@@ -61,11 +50,20 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v2
-      with:
-        submodules: recursive
     - name: make
       run: |
           sudo apt-get update
-          sudo apt-get -y remove libzstd || true
-          sudo apt-get -y install uuid-dev libcurl4-openssl-dev libbz2-dev zlib1g-dev libsnappy-dev liblz4-dev libzstd-dev libgflags-dev
+          sudo apt-get -y install uuid-dev libcurl4-openssl-dev
           make KEYDB_CFLAGS='-Werror' KEYDB_CXXFLAGS='-Werror' MALLOC=libc -j2
+
+  build-ubuntu-32bit:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v2
+    - name: make
+      run: |
+          sudo dpkg --add-architecture i386
+          sudo apt-get update
+          sudo apt-get -y install gcc-multilib g++-multilib libc6-dev-i386 lib32z1 uuid-dev:i386 libcurl4-openssl-dev:i386
+          make KEYDB_CFLAGS='-Werror' KEYDB_CXXFLAGS='-Werror' 32bit -j2
+

.github/workflows/endurance.yml

@@ -0,0 +1,22 @@
+name: Endurance
+
+on:
+  schedule:
+    - cron: '0 * * * *'
+
+jobs:
+
+  test-endurance:
+    runs-on: octocore
+    timeout-minutes: 60
+    steps:
+    - uses: actions/checkout@v1
+    - name: make
+      run: |
+          sudo apt-get -y install uuid-dev libcurl4-openssl-dev
+          make -j8
+    - name: test-multithread (5X)
+      run: |
+        sudo apt-get install -y tcl tcl-tls
+        ./runtest --loopn 5 --config server-threads 3 --clients 5 --endurance
+

.gitignore

@@ -15,6 +15,7 @@ src/keydb-server
 **/app/keydb-server
 *.deb
 *.rpm
+keydb-pro-server
 src/keydb-cli
 **/bin/keydb-cli
 **/app/keydb-cli
@@ -48,7 +49,7 @@ src/nodes.conf
 deps/lua/src/lua
 deps/lua/src/luac
 deps/lua/src/liblua.a
-tests/tls*/*
+tests/tls/*
 .make-*
 .prerequisites
 *.dSYM

.gitmodules

@@ -1,6 +1,3 @@
-[submodule "deps/rocksdb"]
-	path = deps/rocksdb
-	url = https://github.com/facebook/rocksdb.git
 [submodule "deps/depot_tools"]
 	path = deps/depot_tools
 	url = https://chromium.googlesource.com/chromium/tools/depot_tools.git

COPYING

@@ -1,7 +1,6 @@
 Copyright (c) 2006-2020, Salvatore Sanfilippo
 Copyright (C) 2019-2021, John Sully
 Copyright (C) 2020-2021, EQ Alpha Technology Ltd.
-Copyright (C) 2022 Snap Inc.
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

Makefile

@@ -1,4 +1,4 @@
-# Top level makefile, this just calls into src/Makefile where the real work is done. Changes should be made there.
+# Top level makefile, the real shit is at src/Makefile
 
 default: all
 

README.md

@@ -2,20 +2,16 @@
 ![CI](https://github.com/JohnSully/KeyDB/workflows/CI/badge.svg?branch=unstable)
 [![StackShare](http://img.shields.io/badge/tech-stack-0690fa.svg?style=flat)](https://stackshare.io/eq-alpha-technology-inc/eq-alpha-technology-inc)
 
-##### KeyDB is now a part of Snap Inc! Check out the announcement [here](https://docs.keydb.dev/news/2022/05/12/keydb-joins-snap) 
-
-##### [Release v6.3.0](https://github.com/EQ-Alpha/KeyDB/releases/tag/v6.3.0) is here with major improvements as we consolodate our Open Source and Enterprise offerings into a single BSD-3 licensed project. See our [roadmap](https://docs.keydb.dev/docs/coming-soon) for details. 
-
 ##### Want to extend KeyDB with Javascript?  Try [ModJS](https://github.com/JohnSully/ModJS)
 
 ##### Need Help? Check out our extensive [documentation](https://docs.keydb.dev).
 
-##### KeyDB is on Slack. Click [here](https://docs.keydb.dev/slack/) to learn more and join the KeyDB Community Slack workspace.
+##### NEW!!! KeyDB now has a Slack Community Workspace. Click [here](https://docs.keydb.dev/slack/) to learn more and join the KeyDB Community Slack workspace.
 
 What is KeyDB?
 --------------
 
-KeyDB is a high performance fork of Redis with a focus on multithreading, memory efficiency, and high throughput. In addition to performance improvements, KeyDB offers features such as Active Replication, FLASH Storage and Subkey Expires. KeyDB has a MVCC architecture that allows you to execute queries such as KEYS and SCAN without blocking the database and degrading performance.
+KeyDB is a high performance fork of Redis with a focus on multithreading, memory efficiency, and high throughput.  In addition to multithreading, KeyDB also has features only available in Redis Enterprise such as [Active Replication](https://github.com/JohnSully/KeyDB/wiki/Active-Replication), [FLASH storage](https://github.com/JohnSully/KeyDB/wiki/FLASH-Storage) support, and some not available at all such as direct backup to AWS S3. 
 
 KeyDB maintains full compatibility with the Redis protocol, modules, and scripts.  This includes the atomicity guarantees for scripts and transactions.  Because KeyDB keeps in sync with Redis development KeyDB is a superset of Redis functionality, making KeyDB a drop in replacement for existing Redis deployments.
 
@@ -34,75 +30,54 @@ KeyDB has a different philosophy on how the codebase should evolve.  We feel tha
 
 Because of this difference of opinion features which are right for KeyDB may not be appropriate for Redis.  A fork allows us to explore this new development path and implement features which may never be a part of Redis.  KeyDB keeps in sync with upstream Redis changes, and where applicable we upstream bug fixes and changes. It is our hope that the two projects can continue to grow and learn from each other.
 
-Project Support
--------------------
-
-The KeyDB team maintains this project as part of Snap Inc. KeyDB is used by Snap as part of its caching infrastructure and is fully open sourced. There is no separate commercial product and no paid support options available. We really value collaborating with the open source community and welcome PRs, bug reports, and open discussion. For community support or to get involved further with the project check out our community support options [here](https://docs.keydb.dev/docs/support) (slack, forum, meetup, github issues). Our team monitors these channlels regularly.
-
-
 Additional Resources
 --------------------
 
-Try the KeyDB [Docker Image](https://hub.docker.com/r/eqalpha/keydb)
+Try our docker container: https://hub.docker.com/r/eqalpha/keydb
 
-Join us on [Slack](https://docs.keydb.dev/slack/)
+Talk on Gitter: https://gitter.im/KeyDB
 
-Learn more using KeyDB's extensive [documentation](https://docs.keydb.dev)
+Visit our Website: https://keydb.dev
 
-Post to our [Community Forum](https://community.keydb.dev)
+See options for channel partners and support contracts: https://keydb.dev/support.html
 
-See the [KeyDB Roadmap](https://docs.keydb.dev/docs/coming-soon) to see what's in store
+Learn with KeyDB’s official documentation site: https://docs.keydb.dev
+
+[Subscribe to the KeyDB mailing list](https://eqalpha.us20.list-manage.com/subscribe/post?u=978f486c2f95589b24591a9cc&id=4ab9220500)
+
+Management GUI: We recommend [FastoNoSQL](https://fastonosql.com/) which has official KeyDB support.
 
 
 Benchmarking KeyDB
 ------------------
 
-Please note keydb-benchmark and redis-benchmark are currently single threaded and too slow to properly benchmark KeyDB.  We recommend using a redis cluster benchmark tool such as [memtier](https://github.com/RedisLabs/memtier_benchmark).  Please ensure your machine has enough cores for both KeyDB and memtier if testing locally.  KeyDB expects exclusive use of any cores assigned to it.
+Please note keydb-benchmark and redis-benchmark are currently single threaded and too slow to properly benchmark KeyDB.  We recommend using a redis cluster benchmark tool such as [memtier](https://github.com/RedisLabs/memtier_benchmark).  Please ensure your machine has enough cores for both KeyDB and memteir if testing locally.  KeyDB expects exclusive use of any cores assigned to it.
 
+For more details on how we benchmarked KeyDB along with performance numbers check out our blog post: [Redis Should Be Multithreaded](https://medium.com/@john_63123/redis-should-be-multi-threaded-e28319cab744?source=friends_link&sk=7ce8e9fe3ec8224a4d27ef075d085457)
 
 New Configuration Options
 -------------------------
 
-With new features comes new options. All other configuration options behave as you'd expect.  Your existing configuration files should continue to work unchanged.
+With new features comes new options:
 
-```
     server-threads N
     server-thread-affinity [true/false]
-```
-The number of threads used to serve requests.  This should be related to the number of queues available in your network hardware, *not* the number of cores on your
-machine.  Because KeyDB uses spinlocks to reduce latency; making this too high will reduce performance.  We recommend using 4 here.  By default this is set to two.
 
-```
-min-clients-per-thread 50
-```
-The minimum number of clients on a thread before KeyDB assigns new connections to a different thread. Tuning this parameter is a tradeoff between locking overhead and distributing the workload over multiple cores
+The number of threads used to serve requests.  This should be related to the number of queues available in your network hardware, *not* the number of cores on your machine.  Because KeyDB uses spinlocks to reduce latency; making this too high will reduce performance.  We recommend using 4 here.  By default this is set to one.
 
-```
-replica-weighting-factor 2
-```
-KeyDB will attempt to balance clients across threads evenly; However, replica clients are usually much more expensive than a normal client, and so KeyDB will try to assign fewer clients to threads with a replica.  The weighting factor below is intented to help tune this behavior.  A replica weighting factor of 2 means we treat a replica as the equivalent of two normal clients.  Adjusting this value may improve performance when replication is used.  The best weighting is workload specific - e.g. read heavy workloads should set this to 1.  Very write heavy workloads may benefit from higher numbers.
+    scratch-file-path /path
 
-```
-active-client-balancing yes
-```
-Should KeyDB make active attempts at balancing clients across threads?  This can impact performance accepting new clients.  By default this is enabled.  If disabled there is still a best effort from the kernel to distribute across threads with SO_REUSEPORT but it will not be as fair. By default this is enabled
-
-```
-    active-replica yes
-```
-If you are using active-active replication set `active-replica` option to “yes”. This will enable both instances to accept reads and writes while remaining synced. [Click here](https://docs.keydb.dev/docs/active-rep/) to see more on active-rep in our docs section. There are also [docker examples]( https://docs.keydb.dev/docs/docker-active-rep/) on docs.
-
-```
-multi-master-no-forward no
-```
-Avoid forwarding RREPLAY messages to other masters? WARNING: This setting is dangerous! You must be certain all masters are connected to eachother in a true mesh topology or data loss will occur! This command can be used to reduce multimaster bus traffic
-
-
-```
+If you would like to use the [FLASH backed](https://github.com/JohnSully/KeyDB/wiki/FLASH-Storage) storage this option configures the directory for KeyDB's temporary files.  This feature relies on snapshotting to work so must be used on a BTRFS filesystem.  ZFS may also work but is untested.  With this feature KeyDB will use RAM as a cache and page to disk as necessary.  NOTE: This requires special compilation options, see Building KeyDB below.
+    
     db-s3-object /path/to/bucket
-```
+
 If you would like KeyDB to dump and load directly to AWS S3 this option specifies the bucket.  Using this option with the traditional RDB options will result in KeyDB backing up twice to both locations.  If both are specified KeyDB will first attempt to load from the local dump file and if that fails load from S3.  This requires the AWS CLI tools to be installed and configured which are used under the hood to transfer the data.
 
+    active-replica yes
+
+If you are using active-active replication set `active-replica` option to “yes”. This will enable both instances to accept reads and writes while remaining synced. [Click here](https://docs.keydb.dev/docs/active-rep/) to see more on active-rep in our docs section. There are also [docker examples]( https://docs.keydb.dev/docs/docker-active-rep/) on docs.
+
+All other configuration options behave as you'd expect.  Your existing configuration files should continue to work unchanged.
 
 Building KeyDB
 --------------
@@ -111,18 +86,19 @@ KeyDB can be compiled and is tested for use on Linux.  KeyDB currently relies on
 
 More on CentOS/Archlinux/Alpine/Debian/Ubuntu dependencies and builds can be found here: https://docs.keydb.dev/docs/build/
 
-Init and clone submodule dependencies:
-
-    % git submodule init && git submodule update
-
 Install dependencies:
 
-    % sudo apt install build-essential nasm autotools-dev autoconf libjemalloc-dev tcl tcl-dev uuid-dev libcurl4-openssl-dev libbz2-dev libzstd-dev liblz4-dev libsnappy-dev libssl-dev
+    % sudo apt install build-essential nasm autotools-dev autoconf libjemalloc-dev tcl tcl-dev uuid-dev libcurl4-openssl-dev
 
 Compiling is as simple as:
 
     % make
 
+To build with TLS support, you'll need OpenSSL development libraries (e.g.
+libssl-dev on Debian/Ubuntu) and run:
+
+    % make BUILD_TLS=yes
+
 To build with systemd support, you'll need systemd development libraries (such 
 as libsystemd-dev on Debian/Ubuntu or systemd-devel on CentOS) and run:
 
@@ -135,11 +111,14 @@ To append a suffix to KeyDB program names, use:
 ***Note that the following dependencies may be needed: 
     % sudo apt-get install autoconf autotools-dev libnuma-dev libtool
 
-To buik=ld with TLS support, use:
+If TLS is built, running the tests with TLS enabled (you will need `tcl-tls`
+installed):
 
-    % make BUILD_TLS=yes
+    % ./utils/gen-test-certs.sh
+    % ./runtest --tls
 
-Running the tests with TLS enabled (you will need `tcl-tls`
+
+If TLS is built, running the tests with TLS enabled (you will need `tcl-tls`
 installed):
 
     % ./utils/gen-test-certs.sh
@@ -241,11 +220,11 @@ All the options in keydb.conf are also supported as options using the command
 line, with exactly the same name.
 
 
-Running KeyDB with TLS:
+Running Redis with TLS:
 ------------------
 
 Please consult the [TLS.md](TLS.md) file for more information on
-how to use KeyDB with TLS.
+how to use Redis with TLS.
 
 
 Playing with KeyDB
@@ -305,6 +284,24 @@ KeyDB works by running the normal Redis event loop on multiple threads.  Network
 
 Unlike most databases the core data structure is the fastest part of the system.  Most of the query time comes from parsing the REPL protocol and copying data to/from the network.
 
+Future work:
+ - Allow rebalancing of connections to different threads after the connection
+ - Allow multiple readers access to the hashtable concurrently
+
+Docker Build
+------------
+Build the latest binaries from the github unstable branch within a docker container. Note this is built for Ubuntu 18.04.
+Simply make a directory you would like to have the latest binaries dumped in, then run the following commmand with your updated path:
+```
+$ docker run -it --rm -v /path-to-dump-binaries:/keydb_bin eqalpha/keydb-build-bin
+```
+You should receive the following files: keydb-benchmark,  keydb-check-aof,  keydb-check-rdb,  keydb-cli,  keydb-sentinel,  keydb-server
+
+If you are looking to enable flash support with the build (make MALLOC=memkind) then use the following command:
+```
+$ docker run -it --rm -v /path-to-dump-binaries:/keydb_bin eqalpha/keydb-build-bin:flash
+```
+Please note that you will need libcurl4-openssl-dev in order to run keydb. With flash version you may need libnuma-dev and libtool installed in order to run the binaries. Keep this in mind especially when running in a container. For a copy of all our Dockerfiles, please see them on [docs]( https://docs.keydb.dev/docs/dockerfiles/).
 
 Code contributions
 -----------------

TLS.md

@@ -6,7 +6,10 @@ Getting Started
 
 ### Building
 
-TLS support is enabled in the default build. To build without TLS, run `make BUILD_TLS=no`.
+To build with TLS support you'll need OpenSSL development libraries (e.g.
+libssl-dev on Debian/Ubuntu).
+
+Run `make BUILD_TLS=yes`.
 
 ### Tests
 

deps/Makefile

@@ -1,7 +1,6 @@
 # Redis dependency Makefile
 
 uname_S:= $(shell sh -c 'uname -s 2>/dev/null || echo not')
-uname_M := $(shell sh -c 'uname -m 2>/dev/null || echo not')
 
 CCCOLOR="\033[34m"
 LINKCOLOR="\033[34;1m"
@@ -38,7 +37,6 @@ distclean:
 	-(cd linenoise && $(MAKE) clean) > /dev/null || true
 	-(cd lua && $(MAKE) clean) > /dev/null || true
 	-(cd jemalloc && [ -f Makefile ] && $(MAKE) distclean) > /dev/null || true
-	-(cd rocksdb && $(MAKE) clean) > /dev/null || true
 	-(cd hdr_histogram && $(MAKE) clean) > /dev/null || true
 	-(rm -f .make-*)
 
@@ -95,17 +93,7 @@ JEMALLOC_LDFLAGS= $(LDFLAGS)
 
 jemalloc: .make-prerequisites
 	@printf '%b %b\n' $(MAKECOLOR)MAKE$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR)
-	cd jemalloc && ./configure --with-version=5.2.1-0-g0 --with-lg-quantum=3 --with-jemalloc-prefix=je_ --disable-cxx CFLAGS="$(JEMALLOC_CFLAGS)" LDFLAGS="$(JEMALLOC_LDFLAGS)"
+	cd jemalloc && ./configure --with-version=5.2.1-0-g0 --with-lg-quantum=3 --with-jemalloc-prefix=je_ CFLAGS="$(JEMALLOC_CFLAGS)" LDFLAGS="$(JEMALLOC_LDFLAGS)"
 	cd jemalloc && $(MAKE) CFLAGS="$(JEMALLOC_CFLAGS)" LDFLAGS="$(JEMALLOC_LDFLAGS)" lib/libjemalloc.a
 
 .PHONY: jemalloc
-
-rocksdb: .make-prerequisites
-	@printf '%b %b\n' $(MAKECOLOR)MAKE$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR)
-ifeq ($(uname_M),x86_64)
-	cd rocksdb && CFLAGS=-Wno-error PORTABLE=1 USE_SSE=1 FORCE_SSE42=1 $(MAKE) static_lib
-else
-	cd rocksdb && PORTABLE=1 $(MAKE) static_lib
-endif
-
-.PHONY: rocksdb

deps/concurrentqueue/blockingconcurrentqueue.h

@@ -1,583 +0,0 @@
-// Provides an efficient blocking version of moodycamel::ConcurrentQueue.
-// ©2015-2020 Cameron Desrochers. Distributed under the terms of the simplified
-// BSD license, available at the top of concurrentqueue.h.
-// Also dual-licensed under the Boost Software License (see LICENSE.md)
-// Uses Jeff Preshing's semaphore implementation (under the terms of its
-// separate zlib license, see lightweightsemaphore.h).
-
-#pragma once
-
-#include "concurrentqueue.h"
-#include "lightweightsemaphore.h"
-
-#include <type_traits>
-#include <cerrno>
-#include <memory>
-#include <chrono>
-#include <ctime>
-
-namespace moodycamel
-{
-// This is a blocking version of the queue. It has an almost identical interface to
-// the normal non-blocking version, with the addition of various wait_dequeue() methods
-// and the removal of producer-specific dequeue methods.
-template<typename T, typename Traits = ConcurrentQueueDefaultTraits>
-class BlockingConcurrentQueue
-{
-private:
-	typedef ::moodycamel::ConcurrentQueue<T, Traits> ConcurrentQueue;
-	typedef ::moodycamel::LightweightSemaphore LightweightSemaphore;
-
-public:
-	typedef typename ConcurrentQueue::producer_token_t producer_token_t;
-	typedef typename ConcurrentQueue::consumer_token_t consumer_token_t;
-	
-	typedef typename ConcurrentQueue::index_t index_t;
-	typedef typename ConcurrentQueue::size_t size_t;
-	typedef typename std::make_signed<size_t>::type ssize_t;
-	
-	static const size_t BLOCK_SIZE = ConcurrentQueue::BLOCK_SIZE;
-	static const size_t EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD = ConcurrentQueue::EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD;
-	static const size_t EXPLICIT_INITIAL_INDEX_SIZE = ConcurrentQueue::EXPLICIT_INITIAL_INDEX_SIZE;
-	static const size_t IMPLICIT_INITIAL_INDEX_SIZE = ConcurrentQueue::IMPLICIT_INITIAL_INDEX_SIZE;
-	static const size_t INITIAL_IMPLICIT_PRODUCER_HASH_SIZE = ConcurrentQueue::INITIAL_IMPLICIT_PRODUCER_HASH_SIZE;
-	static const std::uint32_t EXPLICIT_CONSUMER_CONSUMPTION_QUOTA_BEFORE_ROTATE = ConcurrentQueue::EXPLICIT_CONSUMER_CONSUMPTION_QUOTA_BEFORE_ROTATE;
-	static const size_t MAX_SUBQUEUE_SIZE = ConcurrentQueue::MAX_SUBQUEUE_SIZE;
-	
-public:
-	// Creates a queue with at least `capacity` element slots; note that the
-	// actual number of elements that can be inserted without additional memory
-	// allocation depends on the number of producers and the block size (e.g. if
-	// the block size is equal to `capacity`, only a single block will be allocated
-	// up-front, which means only a single producer will be able to enqueue elements
-	// without an extra allocation -- blocks aren't shared between producers).
-	// This method is not thread safe -- it is up to the user to ensure that the
-	// queue is fully constructed before it starts being used by other threads (this
-	// includes making the memory effects of construction visible, possibly with a
-	// memory barrier).
-	explicit BlockingConcurrentQueue(size_t capacity = 6 * BLOCK_SIZE)
-		: inner(capacity), sema(create<LightweightSemaphore, ssize_t, int>(0, (int)Traits::MAX_SEMA_SPINS), &BlockingConcurrentQueue::template destroy<LightweightSemaphore>)
-	{
-		assert(reinterpret_cast<ConcurrentQueue*>((BlockingConcurrentQueue*)1) == &((BlockingConcurrentQueue*)1)->inner && "BlockingConcurrentQueue must have ConcurrentQueue as its first member");
-		if (!sema) {
-			MOODYCAMEL_THROW(std::bad_alloc());
-		}
-	}
-	
-	BlockingConcurrentQueue(size_t minCapacity, size_t maxExplicitProducers, size_t maxImplicitProducers)
-		: inner(minCapacity, maxExplicitProducers, maxImplicitProducers), sema(create<LightweightSemaphore, ssize_t, int>(0, (int)Traits::MAX_SEMA_SPINS), &BlockingConcurrentQueue::template destroy<LightweightSemaphore>)
-	{
-		assert(reinterpret_cast<ConcurrentQueue*>((BlockingConcurrentQueue*)1) == &((BlockingConcurrentQueue*)1)->inner && "BlockingConcurrentQueue must have ConcurrentQueue as its first member");
-		if (!sema) {
-			MOODYCAMEL_THROW(std::bad_alloc());
-		}
-	}
-	
-	// Disable copying and copy assignment
-	BlockingConcurrentQueue(BlockingConcurrentQueue const&) MOODYCAMEL_DELETE_FUNCTION;
-	BlockingConcurrentQueue& operator=(BlockingConcurrentQueue const&) MOODYCAMEL_DELETE_FUNCTION;
-	
-	// Moving is supported, but note that it is *not* a thread-safe operation.
-	// Nobody can use the queue while it's being moved, and the memory effects
-	// of that move must be propagated to other threads before they can use it.
-	// Note: When a queue is moved, its tokens are still valid but can only be
-	// used with the destination queue (i.e. semantically they are moved along
-	// with the queue itself).
-	BlockingConcurrentQueue(BlockingConcurrentQueue&& other) MOODYCAMEL_NOEXCEPT
-		: inner(std::move(other.inner)), sema(std::move(other.sema))
-	{ }
-	
-	inline BlockingConcurrentQueue& operator=(BlockingConcurrentQueue&& other) MOODYCAMEL_NOEXCEPT
-	{
-		return swap_internal(other);
-	}
-	
-	// Swaps this queue's state with the other's. Not thread-safe.
-	// Swapping two queues does not invalidate their tokens, however
-	// the tokens that were created for one queue must be used with
-	// only the swapped queue (i.e. the tokens are tied to the
-	// queue's movable state, not the object itself).
-	inline void swap(BlockingConcurrentQueue& other) MOODYCAMEL_NOEXCEPT
-	{
-		swap_internal(other);
-	}
-	
-private:
-	BlockingConcurrentQueue& swap_internal(BlockingConcurrentQueue& other)
-	{
-		if (this == &other) {
-			return *this;
-		}
-		
-		inner.swap(other.inner);
-		sema.swap(other.sema);
-		return *this;
-	}
-	
-public:
-	// Enqueues a single item (by copying it).
-	// Allocates memory if required. Only fails if memory allocation fails (or implicit
-	// production is disabled because Traits::INITIAL_IMPLICIT_PRODUCER_HASH_SIZE is 0,
-	// or Traits::MAX_SUBQUEUE_SIZE has been defined and would be surpassed).
-	// Thread-safe.
-	inline bool enqueue(T const& item)
-	{
-		if ((details::likely)(inner.enqueue(item))) {
-			sema->signal();
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues a single item (by moving it, if possible).
-	// Allocates memory if required. Only fails if memory allocation fails (or implicit
-	// production is disabled because Traits::INITIAL_IMPLICIT_PRODUCER_HASH_SIZE is 0,
-	// or Traits::MAX_SUBQUEUE_SIZE has been defined and would be surpassed).
-	// Thread-safe.
-	inline bool enqueue(T&& item)
-	{
-		if ((details::likely)(inner.enqueue(std::move(item)))) {
-			sema->signal();
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues a single item (by copying it) using an explicit producer token.
-	// Allocates memory if required. Only fails if memory allocation fails (or
-	// Traits::MAX_SUBQUEUE_SIZE has been defined and would be surpassed).
-	// Thread-safe.
-	inline bool enqueue(producer_token_t const& token, T const& item)
-	{
-		if ((details::likely)(inner.enqueue(token, item))) {
-			sema->signal();
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues a single item (by moving it, if possible) using an explicit producer token.
-	// Allocates memory if required. Only fails if memory allocation fails (or
-	// Traits::MAX_SUBQUEUE_SIZE has been defined and would be surpassed).
-	// Thread-safe.
-	inline bool enqueue(producer_token_t const& token, T&& item)
-	{
-		if ((details::likely)(inner.enqueue(token, std::move(item)))) {
-			sema->signal();
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues several items.
-	// Allocates memory if required. Only fails if memory allocation fails (or
-	// implicit production is disabled because Traits::INITIAL_IMPLICIT_PRODUCER_HASH_SIZE
-	// is 0, or Traits::MAX_SUBQUEUE_SIZE has been defined and would be surpassed).
-	// Note: Use std::make_move_iterator if the elements should be moved instead of copied.
-	// Thread-safe.
-	template<typename It>
-	inline bool enqueue_bulk(It itemFirst, size_t count)
-	{
-		if ((details::likely)(inner.enqueue_bulk(std::forward<It>(itemFirst), count))) {
-			sema->signal((LightweightSemaphore::ssize_t)(ssize_t)count);
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues several items using an explicit producer token.
-	// Allocates memory if required. Only fails if memory allocation fails
-	// (or Traits::MAX_SUBQUEUE_SIZE has been defined and would be surpassed).
-	// Note: Use std::make_move_iterator if the elements should be moved
-	// instead of copied.
-	// Thread-safe.
-	template<typename It>
-	inline bool enqueue_bulk(producer_token_t const& token, It itemFirst, size_t count)
-	{
-		if ((details::likely)(inner.enqueue_bulk(token, std::forward<It>(itemFirst), count))) {
-			sema->signal((LightweightSemaphore::ssize_t)(ssize_t)count);
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues a single item (by copying it).
-	// Does not allocate memory. Fails if not enough room to enqueue (or implicit
-	// production is disabled because Traits::INITIAL_IMPLICIT_PRODUCER_HASH_SIZE
-	// is 0).
-	// Thread-safe.
-	inline bool try_enqueue(T const& item)
-	{
-		if (inner.try_enqueue(item)) {
-			sema->signal();
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues a single item (by moving it, if possible).
-	// Does not allocate memory (except for one-time implicit producer).
-	// Fails if not enough room to enqueue (or implicit production is
-	// disabled because Traits::INITIAL_IMPLICIT_PRODUCER_HASH_SIZE is 0).
-	// Thread-safe.
-	inline bool try_enqueue(T&& item)
-	{
-		if (inner.try_enqueue(std::move(item))) {
-			sema->signal();
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues a single item (by copying it) using an explicit producer token.
-	// Does not allocate memory. Fails if not enough room to enqueue.
-	// Thread-safe.
-	inline bool try_enqueue(producer_token_t const& token, T const& item)
-	{
-		if (inner.try_enqueue(token, item)) {
-			sema->signal();
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues a single item (by moving it, if possible) using an explicit producer token.
-	// Does not allocate memory. Fails if not enough room to enqueue.
-	// Thread-safe.
-	inline bool try_enqueue(producer_token_t const& token, T&& item)
-	{
-		if (inner.try_enqueue(token, std::move(item))) {
-			sema->signal();
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues several items.
-	// Does not allocate memory (except for one-time implicit producer).
-	// Fails if not enough room to enqueue (or implicit production is
-	// disabled because Traits::INITIAL_IMPLICIT_PRODUCER_HASH_SIZE is 0).
-	// Note: Use std::make_move_iterator if the elements should be moved
-	// instead of copied.
-	// Thread-safe.
-	template<typename It>
-	inline bool try_enqueue_bulk(It itemFirst, size_t count)
-	{
-		if (inner.try_enqueue_bulk(std::forward<It>(itemFirst), count)) {
-			sema->signal((LightweightSemaphore::ssize_t)(ssize_t)count);
-			return true;
-		}
-		return false;
-	}
-	
-	// Enqueues several items using an explicit producer token.
-	// Does not allocate memory. Fails if not enough room to enqueue.
-	// Note: Use std::make_move_iterator if the elements should be moved
-	// instead of copied.
-	// Thread-safe.
-	template<typename It>
-	inline bool try_enqueue_bulk(producer_token_t const& token, It itemFirst, size_t count)
-	{
-		if (inner.try_enqueue_bulk(token, std::forward<It>(itemFirst), count)) {
-			sema->signal((LightweightSemaphore::ssize_t)(ssize_t)count);
-			return true;
-		}
-		return false;
-	}
-	
-	
-	// Attempts to dequeue from the queue.
-	// Returns false if all producer streams appeared empty at the time they
-	// were checked (so, the queue is likely but not guaranteed to be empty).
-	// Never allocates. Thread-safe.
-	template<typename U>
-	inline bool try_dequeue(U& item)
-	{
-		if (sema->tryWait()) {
-			while (!inner.try_dequeue(item)) {
-				continue;
-			}
-			return true;
-		}
-		return false;
-	}
-	
-	// Attempts to dequeue from the queue using an explicit consumer token.
-	// Returns false if all producer streams appeared empty at the time they
-	// were checked (so, the queue is likely but not guaranteed to be empty).
-	// Never allocates. Thread-safe.
-	template<typename U>
-	inline bool try_dequeue(consumer_token_t& token, U& item)
-	{
-		if (sema->tryWait()) {
-			while (!inner.try_dequeue(token, item)) {
-				continue;
-			}
-			return true;
-		}
-		return false;
-	}
-	
-	// Attempts to dequeue several elements from the queue.
-	// Returns the number of items actually dequeued.
-	// Returns 0 if all producer streams appeared empty at the time they
-	// were checked (so, the queue is likely but not guaranteed to be empty).
-	// Never allocates. Thread-safe.
-	template<typename It>
-	inline size_t try_dequeue_bulk(It itemFirst, size_t max)
-	{
-		size_t count = 0;
-		max = (size_t)sema->tryWaitMany((LightweightSemaphore::ssize_t)(ssize_t)max);
-		while (count != max) {
-			count += inner.template try_dequeue_bulk<It&>(itemFirst, max - count);
-		}
-		return count;
-	}
-	
-	// Attempts to dequeue several elements from the queue using an explicit consumer token.
-	// Returns the number of items actually dequeued.
-	// Returns 0 if all producer streams appeared empty at the time they
-	// were checked (so, the queue is likely but not guaranteed to be empty).
-	// Never allocates. Thread-safe.
-	template<typename It>
-	inline size_t try_dequeue_bulk(consumer_token_t& token, It itemFirst, size_t max)
-	{
-		size_t count = 0;
-		max = (size_t)sema->tryWaitMany((LightweightSemaphore::ssize_t)(ssize_t)max);
-		while (count != max) {
-			count += inner.template try_dequeue_bulk<It&>(token, itemFirst, max - count);
-		}
-		return count;
-	}
-	
-	
-	
-	// Blocks the current thread until there's something to dequeue, then
-	// dequeues it.
-	// Never allocates. Thread-safe.
-	template<typename U>
-	inline void wait_dequeue(U& item)
-	{
-		while (!sema->wait()) {
-			continue;
-		}
-		while (!inner.try_dequeue(item)) {
-			continue;
-		}
-	}
-
-	// Blocks the current thread until either there's something to dequeue
-	// or the timeout (specified in microseconds) expires. Returns false
-	// without setting `item` if the timeout expires, otherwise assigns
-	// to `item` and returns true.
-	// Using a negative timeout indicates an indefinite timeout,
-	// and is thus functionally equivalent to calling wait_dequeue.
-	// Never allocates. Thread-safe.
-	template<typename U>
-	inline bool wait_dequeue_timed(U& item, std::int64_t timeout_usecs)
-	{
-		if (!sema->wait(timeout_usecs)) {
-			return false;
-		}
-		while (!inner.try_dequeue(item)) {
-			continue;
-		}
-		return true;
-	}
-    
-    // Blocks the current thread until either there's something to dequeue
-	// or the timeout expires. Returns false without setting `item` if the
-    // timeout expires, otherwise assigns to `item` and returns true.
-	// Never allocates. Thread-safe.
-	template<typename U, typename Rep, typename Period>
-	inline bool wait_dequeue_timed(U& item, std::chrono::duration<Rep, Period> const& timeout)
-    {
-        return wait_dequeue_timed(item, std::chrono::duration_cast<std::chrono::microseconds>(timeout).count());
-    }
-	
-	// Blocks the current thread until there's something to dequeue, then
-	// dequeues it using an explicit consumer token.
-	// Never allocates. Thread-safe.
-	template<typename U>
-	inline void wait_dequeue(consumer_token_t& token, U& item)
-	{
-		while (!sema->wait()) {
-			continue;
-		}
-		while (!inner.try_dequeue(token, item)) {
-			continue;
-		}
-	}
-	
-	// Blocks the current thread until either there's something to dequeue
-	// or the timeout (specified in microseconds) expires. Returns false
-	// without setting `item` if the timeout expires, otherwise assigns
-	// to `item` and returns true.
-	// Using a negative timeout indicates an indefinite timeout,
-	// and is thus functionally equivalent to calling wait_dequeue.
-	// Never allocates. Thread-safe.
-	template<typename U>
-	inline bool wait_dequeue_timed(consumer_token_t& token, U& item, std::int64_t timeout_usecs)
-	{
-		if (!sema->wait(timeout_usecs)) {
-			return false;
-		}
-		while (!inner.try_dequeue(token, item)) {
-			continue;
-		}
-		return true;
-	}
-    
-    // Blocks the current thread until either there's something to dequeue
-	// or the timeout expires. Returns false without setting `item` if the
-    // timeout expires, otherwise assigns to `item` and returns true.
-	// Never allocates. Thread-safe.
-	template<typename U, typename Rep, typename Period>
-	inline bool wait_dequeue_timed(consumer_token_t& token, U& item, std::chrono::duration<Rep, Period> const& timeout)
-    {
-        return wait_dequeue_timed(token, item, std::chrono::duration_cast<std::chrono::microseconds>(timeout).count());
-    }
-	
-	// Attempts to dequeue several elements from the queue.
-	// Returns the number of items actually dequeued, which will
-	// always be at least one (this method blocks until the queue
-	// is non-empty) and at most max.
-	// Never allocates. Thread-safe.
-	template<typename It>
-	inline size_t wait_dequeue_bulk(It itemFirst, size_t max)
-	{
-		size_t count = 0;
-		max = (size_t)sema->waitMany((LightweightSemaphore::ssize_t)(ssize_t)max);
-		while (count != max) {
-			count += inner.template try_dequeue_bulk<It&>(itemFirst, max - count);
-		}
-		return count;
-	}
-	
-	// Attempts to dequeue several elements from the queue.
-	// Returns the number of items actually dequeued, which can
-	// be 0 if the timeout expires while waiting for elements,
-	// and at most max.
-	// Using a negative timeout indicates an indefinite timeout,
-	// and is thus functionally equivalent to calling wait_dequeue_bulk.
-	// Never allocates. Thread-safe.
-	template<typename It>
-	inline size_t wait_dequeue_bulk_timed(It itemFirst, size_t max, std::int64_t timeout_usecs)
-	{
-		size_t count = 0;
-		max = (size_t)sema->waitMany((LightweightSemaphore::ssize_t)(ssize_t)max, timeout_usecs);
-		while (count != max) {
-			count += inner.template try_dequeue_bulk<It&>(itemFirst, max - count);
-		}
-		return count;
-	}
-    
-    // Attempts to dequeue several elements from the queue.
-	// Returns the number of items actually dequeued, which can
-	// be 0 if the timeout expires while waiting for elements,
-	// and at most max.
-	// Never allocates. Thread-safe.
-	template<typename It, typename Rep, typename Period>
-	inline size_t wait_dequeue_bulk_timed(It itemFirst, size_t max, std::chrono::duration<Rep, Period> const& timeout)
-    {
-        return wait_dequeue_bulk_timed<It&>(itemFirst, max, std::chrono::duration_cast<std::chrono::microseconds>(timeout).count());
-    }
-	
-	// Attempts to dequeue several elements from the queue using an explicit consumer token.
-	// Returns the number of items actually dequeued, which will
-	// always be at least one (this method blocks until the queue
-	// is non-empty) and at most max.
-	// Never allocates. Thread-safe.
-	template<typename It>
-	inline size_t wait_dequeue_bulk(consumer_token_t& token, It itemFirst, size_t max)
-	{
-		size_t count = 0;
-		max = (size_t)sema->waitMany((LightweightSemaphore::ssize_t)(ssize_t)max);
-		while (count != max) {
-			count += inner.template try_dequeue_bulk<It&>(token, itemFirst, max - count);
-		}
-		return count;
-	}
-	
-	// Attempts to dequeue several elements from the queue using an explicit consumer token.
-	// Returns the number of items actually dequeued, which can
-	// be 0 if the timeout expires while waiting for elements,
-	// and at most max.
-	// Using a negative timeout indicates an indefinite timeout,
-	// and is thus functionally equivalent to calling wait_dequeue_bulk.
-	// Never allocates. Thread-safe.
-	template<typename It>
-	inline size_t wait_dequeue_bulk_timed(consumer_token_t& token, It itemFirst, size_t max, std::int64_t timeout_usecs)
-	{
-		size_t count = 0;
-		max = (size_t)sema->waitMany((LightweightSemaphore::ssize_t)(ssize_t)max, timeout_usecs);
-		while (count != max) {
-			count += inner.template try_dequeue_bulk<It&>(token, itemFirst, max - count);
-		}
-		return count;
-	}
-	
-	// Attempts to dequeue several elements from the queue using an explicit consumer token.
-	// Returns the number of items actually dequeued, which can
-	// be 0 if the timeout expires while waiting for elements,
-	// and at most max.
-	// Never allocates. Thread-safe.
-	template<typename It, typename Rep, typename Period>
-	inline size_t wait_dequeue_bulk_timed(consumer_token_t& token, It itemFirst, size_t max, std::chrono::duration<Rep, Period> const& timeout)
-    {
-        return wait_dequeue_bulk_timed<It&>(token, itemFirst, max, std::chrono::duration_cast<std::chrono::microseconds>(timeout).count());
-    }
-	
-	
-	// Returns an estimate of the total number of elements currently in the queue. This
-	// estimate is only accurate if the queue has completely stabilized before it is called
-	// (i.e. all enqueue and dequeue operations have completed and their memory effects are
-	// visible on the calling thread, and no further operations start while this method is
-	// being called).
-	// Thread-safe.
-	inline size_t size_approx() const
-	{
-		return (size_t)sema->availableApprox();
-	}
-	
-	
-	// Returns true if the underlying atomic variables used by
-	// the queue are lock-free (they should be on most platforms).
-	// Thread-safe.
-	static bool is_lock_free()
-	{
-		return ConcurrentQueue::is_lock_free();
-	}
-	
-
-private:
-	template<typename U, typename A1, typename A2>
-	static inline U* create(A1&& a1, A2&& a2)
-	{
-		void* p = (Traits::malloc)(sizeof(U));
-		return p != nullptr ? new (p) U(std::forward<A1>(a1), std::forward<A2>(a2)) : nullptr;
-	}
-	
-	template<typename U>
-	static inline void destroy(U* p)
-	{
-		if (p != nullptr) {
-			p->~U();
-		}
-		(Traits::free)(p);
-	}
-	
-private:
-	ConcurrentQueue inner;
-	std::unique_ptr<LightweightSemaphore, void (*)(LightweightSemaphore*)> sema;
-};
-
-
-template<typename T, typename Traits>
-inline void swap(BlockingConcurrentQueue<T, Traits>& a, BlockingConcurrentQueue<T, Traits>& b) MOODYCAMEL_NOEXCEPT
-{
-	a.swap(b);
-}
-
-}	// end namespace moodycamel
-

deps/concurrentqueue/lightweightsemaphore.h

@@ -1,412 +0,0 @@
-// Provides an efficient implementation of a semaphore (LightweightSemaphore).
-// This is an extension of Jeff Preshing's sempahore implementation (licensed 
-// under the terms of its separate zlib license) that has been adapted and
-// extended by Cameron Desrochers.
-
-#pragma once
-
-#include <cstddef> // For std::size_t
-#include <atomic>
-#include <type_traits> // For std::make_signed<T>
-
-#if defined(_WIN32)
-// Avoid including windows.h in a header; we only need a handful of
-// items, so we'll redeclare them here (this is relatively safe since
-// the API generally has to remain stable between Windows versions).
-// I know this is an ugly hack but it still beats polluting the global
-// namespace with thousands of generic names or adding a .cpp for nothing.
-extern "C" {
-	struct _SECURITY_ATTRIBUTES;
-	__declspec(dllimport) void* __stdcall CreateSemaphoreW(_SECURITY_ATTRIBUTES* lpSemaphoreAttributes, long lInitialCount, long lMaximumCount, const wchar_t* lpName);
-	__declspec(dllimport) int __stdcall CloseHandle(void* hObject);
-	__declspec(dllimport) unsigned long __stdcall WaitForSingleObject(void* hHandle, unsigned long dwMilliseconds);
-	__declspec(dllimport) int __stdcall ReleaseSemaphore(void* hSemaphore, long lReleaseCount, long* lpPreviousCount);
-}
-#elif defined(__MACH__)
-#include <mach/mach.h>
-#elif defined(__unix__)
-#include <semaphore.h>
-#endif
-
-namespace moodycamel
-{
-namespace details
-{
-
-// Code in the mpmc_sema namespace below is an adaptation of Jeff Preshing's
-// portable + lightweight semaphore implementations, originally from
-// https://github.com/preshing/cpp11-on-multicore/blob/master/common/sema.h
-// LICENSE:
-// Copyright (c) 2015 Jeff Preshing
-//
-// This software is provided 'as-is', without any express or implied
-// warranty. In no event will the authors be held liable for any damages
-// arising from the use of this software.
-//
-// Permission is granted to anyone to use this software for any purpose,
-// including commercial applications, and to alter it and redistribute it
-// freely, subject to the following restrictions:
-//
-// 1. The origin of this software must not be misrepresented; you must not
-//	claim that you wrote the original software. If you use this software
-//	in a product, an acknowledgement in the product documentation would be
-//	appreciated but is not required.
-// 2. Altered source versions must be plainly marked as such, and must not be
-//	misrepresented as being the original software.
-// 3. This notice may not be removed or altered from any source distribution.
-#if defined(_WIN32)
-class Semaphore
-{
-private:
-	void* m_hSema;
-	
-	Semaphore(const Semaphore& other) MOODYCAMEL_DELETE_FUNCTION;
-	Semaphore& operator=(const Semaphore& other) MOODYCAMEL_DELETE_FUNCTION;
-
-public:
-	Semaphore(int initialCount = 0)
-	{
-		assert(initialCount >= 0);
-		const long maxLong = 0x7fffffff;
-		m_hSema = CreateSemaphoreW(nullptr, initialCount, maxLong, nullptr);
-		assert(m_hSema);
-	}
-
-	~Semaphore()
-	{
-		CloseHandle(m_hSema);
-	}
-
-	bool wait()
-	{
-		const unsigned long infinite = 0xffffffff;
-		return WaitForSingleObject(m_hSema, infinite) == 0;
-	}
-	
-	bool try_wait()
-	{
-		return WaitForSingleObject(m_hSema, 0) == 0;
-	}
-	
-	bool timed_wait(std::uint64_t usecs)
-	{
-		return WaitForSingleObject(m_hSema, (unsigned long)(usecs / 1000)) == 0;
-	}
-
-	void signal(int count = 1)
-	{
-		while (!ReleaseSemaphore(m_hSema, count, nullptr));
-	}
-};
-#elif defined(__MACH__)
-//---------------------------------------------------------
-// Semaphore (Apple iOS and OSX)
-// Can't use POSIX semaphores due to http://lists.apple.com/archives/darwin-kernel/2009/Apr/msg00010.html
-//---------------------------------------------------------
-class Semaphore
-{
-private:
-	semaphore_t m_sema;
-
-	Semaphore(const Semaphore& other) MOODYCAMEL_DELETE_FUNCTION;
-	Semaphore& operator=(const Semaphore& other) MOODYCAMEL_DELETE_FUNCTION;
-
-public:
-	Semaphore(int initialCount = 0)
-	{
-		assert(initialCount >= 0);
-		kern_return_t rc = semaphore_create(mach_task_self(), &m_sema, SYNC_POLICY_FIFO, initialCount);
-		assert(rc == KERN_SUCCESS);
-		(void)rc;
-	}
-
-	~Semaphore()
-	{
-		semaphore_destroy(mach_task_self(), m_sema);
-	}
-
-	bool wait()
-	{
-		return semaphore_wait(m_sema) == KERN_SUCCESS;
-	}
-	
-	bool try_wait()
-	{
-		return timed_wait(0);
-	}
-	
-	bool timed_wait(std::uint64_t timeout_usecs)
-	{
-		mach_timespec_t ts;
-		ts.tv_sec = static_cast<unsigned int>(timeout_usecs / 1000000);
-		ts.tv_nsec = static_cast<int>((timeout_usecs % 1000000) * 1000);
-
-		// added in OSX 10.10: https://developer.apple.com/library/prerelease/mac/documentation/General/Reference/APIDiffsMacOSX10_10SeedDiff/modules/Darwin.html
-		kern_return_t rc = semaphore_timedwait(m_sema, ts);
-		return rc == KERN_SUCCESS;
-	}
-
-	void signal()
-	{
-		while (semaphore_signal(m_sema) != KERN_SUCCESS);
-	}
-
-	void signal(int count)
-	{
-		while (count-- > 0)
-		{
-			while (semaphore_signal(m_sema) != KERN_SUCCESS);
-		}
-	}
-};
-#elif defined(__unix__)
-//---------------------------------------------------------
-// Semaphore (POSIX, Linux)
-//---------------------------------------------------------
-class Semaphore
-{
-private:
-	sem_t m_sema;
-
-	Semaphore(const Semaphore& other) MOODYCAMEL_DELETE_FUNCTION;
-	Semaphore& operator=(const Semaphore& other) MOODYCAMEL_DELETE_FUNCTION;
-
-public:
-	Semaphore(int initialCount = 0)
-	{
-		assert(initialCount >= 0);
-		int rc = sem_init(&m_sema, 0, static_cast<unsigned int>(initialCount));
-		assert(rc == 0);
-		(void)rc;
-	}
-
-	~Semaphore()
-	{
-		sem_destroy(&m_sema);
-	}
-
-	bool wait()
-	{
-		// http://stackoverflow.com/questions/2013181/gdb-causes-sem-wait-to-fail-with-eintr-error
-		int rc;
-		do {
-			rc = sem_wait(&m_sema);
-		} while (rc == -1 && errno == EINTR);
-		return rc == 0;
-	}
-
-	bool try_wait()
-	{
-		int rc;
-		do {
-			rc = sem_trywait(&m_sema);
-		} while (rc == -1 && errno == EINTR);
-		return rc == 0;
-	}
-
-	bool timed_wait(std::uint64_t usecs)
-	{
-		struct timespec ts;
-		const int usecs_in_1_sec = 1000000;
-		const int nsecs_in_1_sec = 1000000000;
-		clock_gettime(CLOCK_REALTIME, &ts);
-		ts.tv_sec += (time_t)(usecs / usecs_in_1_sec);
-		ts.tv_nsec += (long)(usecs % usecs_in_1_sec) * 1000;
-		// sem_timedwait bombs if you have more than 1e9 in tv_nsec
-		// so we have to clean things up before passing it in
-		if (ts.tv_nsec >= nsecs_in_1_sec) {
-			ts.tv_nsec -= nsecs_in_1_sec;
-			++ts.tv_sec;
-		}
-
-		int rc;
-		do {
-			rc = sem_timedwait(&m_sema, &ts);
-		} while (rc == -1 && errno == EINTR);
-		return rc == 0;
-	}
-
-	void signal()
-	{
-		while (sem_post(&m_sema) == -1);
-	}
-
-	void signal(int count)
-	{
-		while (count-- > 0)
-		{
-			while (sem_post(&m_sema) == -1);
-		}
-	}
-};
-#else
-#error Unsupported platform! (No semaphore wrapper available)
-#endif
-
-}	// end namespace details
-
-
-//---------------------------------------------------------
-// LightweightSemaphore
-//---------------------------------------------------------
-class LightweightSemaphore
-{
-public:
-	typedef std::make_signed<std::size_t>::type ssize_t;
-
-private:
-	std::atomic<ssize_t> m_count;
-	details::Semaphore m_sema;
-	int m_maxSpins;
-
-	bool waitWithPartialSpinning(std::int64_t timeout_usecs = -1)
-	{
-		ssize_t oldCount;
-		int spin = m_maxSpins;
-		while (--spin >= 0)
-		{
-			oldCount = m_count.load(std::memory_order_relaxed);
-			if ((oldCount > 0) && m_count.compare_exchange_strong(oldCount, oldCount - 1, std::memory_order_acquire, std::memory_order_relaxed))
-				return true;
-			std::atomic_signal_fence(std::memory_order_acquire);	 // Prevent the compiler from collapsing the loop.
-		}
-		oldCount = m_count.fetch_sub(1, std::memory_order_acquire);
-		if (oldCount > 0)
-			return true;
-		if (timeout_usecs < 0)
-		{
-			if (m_sema.wait())
-				return true;
-		}
-		if (timeout_usecs > 0 && m_sema.timed_wait((std::uint64_t)timeout_usecs))
-			return true;
-		// At this point, we've timed out waiting for the semaphore, but the
-		// count is still decremented indicating we may still be waiting on
-		// it. So we have to re-adjust the count, but only if the semaphore
-		// wasn't signaled enough times for us too since then. If it was, we
-		// need to release the semaphore too.
-		while (true)
-		{
-			oldCount = m_count.load(std::memory_order_acquire);
-			if (oldCount >= 0 && m_sema.try_wait())
-				return true;
-			if (oldCount < 0 && m_count.compare_exchange_strong(oldCount, oldCount + 1, std::memory_order_relaxed, std::memory_order_relaxed))
-				return false;
-		}
-	}
-
-	ssize_t waitManyWithPartialSpinning(ssize_t max, std::int64_t timeout_usecs = -1)
-	{
-		assert(max > 0);
-		ssize_t oldCount;
-		int spin = m_maxSpins;
-		while (--spin >= 0)
-		{
-			oldCount = m_count.load(std::memory_order_relaxed);
-			if (oldCount > 0)
-			{
-				ssize_t newCount = oldCount > max ? oldCount - max : 0;
-				if (m_count.compare_exchange_strong(oldCount, newCount, std::memory_order_acquire, std::memory_order_relaxed))
-					return oldCount - newCount;
-			}
-			std::atomic_signal_fence(std::memory_order_acquire);
-		}
-		oldCount = m_count.fetch_sub(1, std::memory_order_acquire);
-		if (oldCount <= 0)
-		{
-			if ((timeout_usecs == 0) || (timeout_usecs < 0 && !m_sema.wait()) || (timeout_usecs > 0 && !m_sema.timed_wait((std::uint64_t)timeout_usecs)))
-			{
-				while (true)
-				{
-					oldCount = m_count.load(std::memory_order_acquire);
-					if (oldCount >= 0 && m_sema.try_wait())
-						break;
-					if (oldCount < 0 && m_count.compare_exchange_strong(oldCount, oldCount + 1, std::memory_order_relaxed, std::memory_order_relaxed))
-						return 0;
-				}
-			}
-		}
-		if (max > 1)
-			return 1 + tryWaitMany(max - 1);
-		return 1;
-	}
-
-public:
-	LightweightSemaphore(ssize_t initialCount = 0, int maxSpins = 10000) : m_count(initialCount), m_maxSpins(maxSpins)
-	{
-		assert(initialCount >= 0);
-		assert(maxSpins >= 0);
-	}
-
-	bool tryWait()
-	{
-		ssize_t oldCount = m_count.load(std::memory_order_relaxed);
-		while (oldCount > 0)
-		{
-			if (m_count.compare_exchange_weak(oldCount, oldCount - 1, std::memory_order_acquire, std::memory_order_relaxed))
-				return true;
-		}
-		return false;
-	}
-
-	bool wait()
-	{
-		return tryWait() || waitWithPartialSpinning();
-	}
-
-	bool wait(std::int64_t timeout_usecs)
-	{
-		return tryWait() || waitWithPartialSpinning(timeout_usecs);
-	}
-
-	// Acquires between 0 and (greedily) max, inclusive
-	ssize_t tryWaitMany(ssize_t max)
-	{
-		assert(max >= 0);
-		ssize_t oldCount = m_count.load(std::memory_order_relaxed);
-		while (oldCount > 0)
-		{
-			ssize_t newCount = oldCount > max ? oldCount - max : 0;
-			if (m_count.compare_exchange_weak(oldCount, newCount, std::memory_order_acquire, std::memory_order_relaxed))
-				return oldCount - newCount;
-		}
-		return 0;
-	}
-
-	// Acquires at least one, and (greedily) at most max
-	ssize_t waitMany(ssize_t max, std::int64_t timeout_usecs)
-	{
-		assert(max >= 0);
-		ssize_t result = tryWaitMany(max);
-		if (result == 0 && max > 0)
-			result = waitManyWithPartialSpinning(max, timeout_usecs);
-		return result;
-	}
-	
-	ssize_t waitMany(ssize_t max)
-	{
-		ssize_t result = waitMany(max, -1);
-		assert(result > 0);
-		return result;
-	}
-
-	void signal(ssize_t count = 1)
-	{
-		assert(count >= 0);
-		ssize_t oldCount = m_count.fetch_add(count, std::memory_order_release);
-		ssize_t toRelease = -oldCount < count ? -oldCount : count;
-		if (toRelease > 0)
-		{
-			m_sema.signal((int)toRelease);
-		}
-	}
-	
-	std::size_t availableApprox() const
-	{
-		ssize_t count = m_count.load(std::memory_order_relaxed);
-		return count > 0 ? static_cast<std::size_t>(count) : 0;
-	}
-};
-
-}   // end namespace moodycamel
-

deps/license/keycheck.h

@@ -1,3 +0,0 @@
-#pragma once
-
-bool FValidKey(const char *key, size_t cch);

keydb.conf

@@ -184,9 +184,7 @@ tcp-keepalive 300
 #
 # tls-client-key-file-pass secret
 
-# Configure a DH parameters file to enable Diffie-Hellman (DH) key exchange,
-# required by older versions of OpenSSL (<3.0). Newer versions do not require
-# this configuration and recommend against it.
+# Configure a DH parameters file to enable Diffie-Hellman (DH) key exchange:
 #
 # tls-dh-params-file keydb.dh
 
@@ -259,24 +257,6 @@ tcp-keepalive 300
 #
 # tls-session-cache-timeout 60
 
-# Allow the server to monitor the filesystem and rotate out TLS certificates if 
-# they change on disk, defaults to no.
-# 
-# tls-rotation no
-
-# Setup a allowlist of allowed Common Names (CNs)/Subject Alternative Names (SANs)
-# that are allowed to connect to this server. This includes both normal clients as
-# well as other servers connected for replication/clustering purposes. If nothing is
-# specified, then no allowlist is used and all certificates are accepted. 
-# Supports IPv4, DNS, RFC822, and URI SAN types.
-# You can put multiple names on one line as follows:
-#
-# tls-allowlist <dns1> <dns2> <dns3> ...
-# 
-#
-# This configuration also allows for wildcard characters with glob style formatting
-# i.e. "*.com" would allow all clients to connect with a CN/SAN that ends with ".com"
-
 ################################# GENERAL #####################################
 
 # By default KeyDB does not run as a daemon. Use 'yes' if you need it.
@@ -2055,22 +2035,9 @@ server-threads 2
 # bringing up replicas can result in data loss (the first master will win).
 # active-replica yes
 
-# KeyDB will attempt to balance clients across threads evenly; However, replica clients
-# are usually much more expensive than a normal client, and so KeyDB will try to assign
-# fewer clients to threads with a replica.  The weighting factor below is intented to help tune
-# this behavior.  A replica weighting factor of 2 means we treat a replica as the equivalent
-# of two normal clients.  Adjusting this value may improve performance when replication is
-# used.  The best weighting is workload specific - e.g. read heavy workloads should set
-# this to 1.  Very write heavy workloads may benefit from higher numbers.
-#
-# By default KeyDB sets this to 2.
-replica-weighting-factor 2
+# Enable Pro?  KeyDB pro provides support for pro only features
+#   note: you may omit the license key to demo pro features for a limited time
+# enable-pro [License Key]
 
-# Should KeyDB make active attempts at balancing clients across threads?  This can impact
-# performance accepting new clients.  By default this is enabled.  If disabled there is still
-# a best effort from the kernel to distribute across threads with SO_REUSEPORT but it will not
-# be as fair.
-#
-# By default this is enabled
-#
-active-client-balancing yes
+# Enable FLASH support? (Pro Only)
+# storage-provider flash /path/to/flash/db

monkey/monkey.py

@@ -1,283 +0,0 @@
-import random
-import time
-import socket
-import asyncore
-import threading
-import argparse
-from pprint import pprint
-
-# Globals
-ops = {}
-numclients = 0
-numkeys = 0
-runtime = 0
-clients = []
-
-def _buildResp(*args):
-    result = "*" + str(len(args)) + "\r\n"
-    for v in args:
-        result = result + "$" + str(len(v)) + "\r\n"
-        result = result + v + "\r\n"
-    return result.encode('utf-8')
-
-class Client(asyncore.dispatcher):
-    def __init__(self, host, port):
-        asyncore.dispatcher.__init__(self)
-        self.create_socket(socket.AF_INET, socket.SOCK_STREAM)
-        self.connect((host, port))
-        self.buf = b''
-        self.inbuf = b''
-        self.callbacks = list()
-        self.client_id = 0
-        self.get_client_id()
-
-    def handle_connect(self):
-        pass
-
-    def handle_read(self):
-        self.inbuf += self.recv(8192)
-        self.parse_response()
-
-    def handle_write(self):
-        sent = self.send(self.buf)
-        self.buf = self.buf[sent:]
-
-    def handle_close(self):
-        self.close()
-
-    def writable(self):
-        return len(self.buf) > 0
-
-    def parse_array(self, startpos):
-        assert(self.inbuf[startpos] == ord('*'))
-        endrange = self.inbuf[startpos+1:].find(ord('\r')) + 1 + startpos
-        assert(endrange > 0)
-        numargs = int(self.inbuf[startpos+1:endrange])
-        if numargs == -1: # Nil array, used in some returns
-            startpos = endrange + 2
-            return startpos, []
-        assert(numargs > 0)
-        args = list()
-        startpos = endrange + 2 # plus 1 gets us to the '\n' and the next gets us to the start char
-
-        while len(args) < numargs:
-            # We're parsing entries of the form "$N\r\nnnnnnn\r\n"
-            if startpos >= len(self.inbuf):
-                return # Not the full response
-            if self.inbuf[startpos] == ord('*'):
-                startpos, arr = self.parse_array(startpos)
-                args.append(arr)
-            else:
-                assert(self.inbuf[startpos] == ord('$'))
-                startpos = startpos + 1
-                endrange = self.inbuf[startpos:].find(b'\r')
-                if endrange < 0:
-                    return
-                endrange += startpos
-                assert(endrange <= len(self.inbuf))
-                length = int(self.inbuf[startpos:endrange])
-                if length < 0:
-                    return
-                startpos = endrange + 2
-                assert((startpos + length) <= len(self.inbuf))
-                assert(self.inbuf[startpos+length] == ord('\r'))
-                assert(self.inbuf[startpos+length+1] == ord('\n'))
-                args.append(self.inbuf[startpos:(startpos+length)])
-                startpos += length + 2
-        assert(len(args) == numargs)
-        return startpos, args
-
-    def parse_response(self):
-        if len(self.inbuf) == 0:
-            return
-        
-        while len(self.inbuf) > 0:
-            if self.inbuf[0] == ord('+') or self.inbuf[0] == ord('-') or self.inbuf[0] == ord(':'):
-                # This is a single line response
-                endpos = self.inbuf.find(b'\n')
-                if endpos < 0:
-                    return  # incomplete response
-                self.callbacks[0](self, self.inbuf[0:endpos-1])
-                self.callbacks.pop(0)
-                self.inbuf = self.inbuf[endpos+1:]
-
-            elif self.inbuf[0] == ord('*'):
-                #RESP response
-                try:
-                    startpos, args = self.parse_array(0)
-                except:
-                    return # Not all data here yet
-                self.callbacks[0](self, args)
-                self.callbacks.pop(0)
-                self.inbuf = self.inbuf[startpos:]
-            else:
-                print("ERROR: Unknown response:")
-                pprint(self.inbuf)
-                assert(False)
-
-
-    def default_result_handler(self, result):
-        pprint(result)
-
-    # Public Methods
-    def set(self, key, val, callback = default_result_handler):
-        self.buf += _buildResp("set", key, val)
-        self.callbacks.append(callback)
-
-    def lpush(self, key, val, callback = default_result_handler):
-        self.buf += _buildResp("lpush", key, val)
-        self.callbacks.append(callback)
-
-    def blpop(self, *keys, timeout=0, callback=default_result_handler):
-        self.buf += _buildResp("blpop", *keys, str(timeout))
-        self.callbacks.append(callback)
-
-    def delete(self, key, callback = default_result_handler):
-        self.buf += _buildResp("del", key)
-        self.callbacks.append(callback)
-
-    def unblock(self, client_id, callback=default_result_handler):
-        self.buf += _buildResp("client", "unblock", str(client_id))
-        self.callbacks.append(callback)
-
-    def scan(self, iter, match=None, count=None, callback = default_result_handler):
-        args = ["scan", str(iter)]
-        if match != None:
-            args.append("MATCH")
-            args.append(match)
-        if count != None:
-            args.append("COUNT")
-            args.append(str(count))
-        self.buf += _buildResp(*args)
-        self.callbacks.append(callback)
-
-    def get_client_id(self):
-        self.buf += _buildResp("client", "id")
-        self.callbacks.append(self.store_client_id)
-
-    def store_client_id(self, c, resp):
-        assert(resp[0] == ord(':'))
-        self.client_id = int(resp[1:])
-        assert(self.client_id == c.client_id)
-
-    def get(self, key, callback = None):
-       return 
-
-def getrandomkey():
-    return str(random.randrange(0, numkeys))
-
-def handle_lpush_response(c, resp=None):
-    global ops
-    if resp != None:
-        ops['lpush'] += 1
-        assert(resp[0] == ord(':'))
-    c.lpush("list_" + getrandomkey(), 'bardsklfjkldsjfdlsjflksdfjklsdjflksd kldsjflksd jlkdsjf lksdjklds jrfklsdjfklsdjfkl', handle_lpush_response)
-
-def handle_blpop_response(c, resp=None):
-    global ops
-    if resp != None:
-        ops['blpop'] += 1
-    c.blpop("list_" + getrandomkey(), callback=handle_blpop_response)
-
-def handle_set_response(c, resp=None):
-    global ops
-    if resp != None:
-        ops['set'] += 1
-        assert(resp[0] == ord('+'))
-    c.set("str_" + getrandomkey(), 'bardsklfjkldsjfdlsjflksdfjklsdjflksd kldsjflksd jlkdsjf lksdjklds jrfklsdjfklsdjfkl', handle_set_response)
-
-def handle_del_response(c, resp=None):
-    global ops
-    if resp != None:
-        ops['del'] += 1
-    c.delete("list_" + getrandomkey(), handle_del_response)
-
-def scan_callback(c, resp=None):
-    global ops
-    nextstart = int(resp[0])
-    c.scan(nextstart, count=500, callback=scan_callback)
-    ops['scan'] += 1
-
-def clear_ops():
-    global ops
-    ops = {'lpush': 0, 'blpop': 0, 'del': 0, 'scan': 0, 'set': 0, 'get': 0}
-
-def stats_thread():
-    global ops
-    global runtime
-    i = 0
-    while i < runtime or not runtime:
-        time.sleep(1)
-        print("Ops per second: " + str({k:v for (k,v) in ops.items() if v}))
-        clear_ops()
-        i += 1
-    asyncore.close_all()
-
-def flush_db_sync():
-    server = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
-    server.connect(('127.0.0.1', 6379))
-    server.send(_buildResp("flushdb"))
-    resp = server.recv(8192)
-    assert(resp[:3] == "+OK".encode('utf-8'))
-
-def init_blocking():
-    global clients
-    if numkeys > 100 * numclients:
-        print("WARNING: High ratio of keys to clients. Most lpushes will not be popped and unblocking will take a long time!")
-    for i in range(numclients):
-        clients.append(Client('127.0.0.1', 6379))
-        if i % 2:
-            handle_blpop_response(clients[-1])
-        else:
-            handle_lpush_response(clients[-1])
-
-def init_lpush():
-    global clients
-    for i in range(numclients):
-        clients.append(Client('127.0.0.1', 6379))
-        for i in range (10):
-            handle_lpush_response(clients[-1])
-        #handle_set_response(clients[-1], None)
-
-    scan_client = Client('127.0.0.1', 6379)
-    scan_client.scan(0, count=500, callback=scan_callback)
-
-    del_client = Client('127.0.0.1', 6379)
-    handle_del_response(del_client)
-
-def main(test, flush):
-    clear_ops()
-
-    if flush:
-        flush_db_sync()
-
-    try:
-        globals()[f"init_{test}"]()
-    except KeyError:
-        print(f"Test \"{test}\" not found. Exiting...")
-        exit()
-    except ConnectionRefusedError:
-        print("Could not connect to server. Is it running?")
-        print("Exiting...")
-        exit()
-
-    threading.Thread(target=stats_thread).start()
-    asyncore.loop()
-    print("Done.")
-
-parser = argparse.ArgumentParser(description="Test use cases for KeyDB.")
-parser.add_argument('test', choices=[x[5:] for x in filter(lambda name: name.startswith("init_"), globals().keys())], help="which test to run")
-parser.add_argument('-c', '--clients', type=int, default=50, help="number of running clients to use")
-parser.add_argument('-k', '--keys', type=int, default=100000, help="number of keys to choose from for random tests")
-parser.add_argument('-t', '--runtime', type=int, default=0, help="how long to run the test for (default: 0 for infinite)")
-parser.add_argument('-f', '--flush', action="store_true", help="flush the db before running the test")
-
-if __name__ == "__main__":
-    try:
-        args = parser.parse_args()
-    except:
-        exit()
-    numclients = args.clients
-    numkeys = args.keys
-    runtime = args.runtime
-    main(args.test, args.flush)

pkg/README.md

@@ -1,21 +1,3 @@
-### KeyDB DEB Package Source Builds
+### KeyDB Packaging
 
-This directory contains scripts and components needed to generate debian packages on different distributions/architectures from source
-
-The 'debian' directory contains debian source code for bionic, buster and later distributions as it uses functions only available with debhelper11+. 'debian_dh9' is used for xenial, stretch and earlier distributions using debhelper9.
-
-You will need to install pbuilder `sudo apt install pbuilder` along with other distribution specific dependencies
-
-Generate deb packages with the following script command run from this directory:
-
-```
-$ ./deb-buildsource.sh
-```
-
-This generates a directory structure, .dsc file, original.tar.gz, and new changelog for the distribution and architecture installed.
-
-When complete the produced debian packages will be located in deb_files_generated directory.
-
-Arguments for the deb-buildsource.sh script that can be passed are either 'None' or '"your custom comments"'. 'None' assumes a permanent changelog entry has been made listing the correct build for the version. If you quote and enter your own comment string, it will be appended to the changelog for that deb package. If no argument is passed a default comment is generated saying that this deb package was generated by this script.
-
-This script has been tested on xenial/bionic/focal/stretch/buster/bullseye
+The following directories contain packaging code used to create rpm and deb packages as well as docker images.

pkg/deb/README.md

@@ -1,17 +1,21 @@
 ### KeyDB DEB Package Source Builds
 
-This directory contains scripts and components needed to generate debian packages on different distributions/architectures using source
+This directory contains scripts and components needed to generate debian packages on different distributions/architectures from source
 
-You will need to install pbuilder `sudo apt install pbuilder`
+The 'debian' directory contains debian source code for bionic, buster and later distributions as it uses functions only available with debhelper11+. 'debian_dh9' is used for xenial, stretch and earlier distributions using debhelper9.
+
+You will need to install pbuilder `sudo apt install pbuilder` along with other distribution specific dependancies
 
 Generate deb packages with the following script command run from this directory:
+
 ```
 $ ./deb-buildsource.sh
 ```
-This generates a directory structure, .dsc file, original.tar.gz, .changes files and new changelog for the distribution and architecture installed.
+
+This generates a directory structure, .dsc file, original.tar.gz, and new changelog for the distribution and architecture installed.
 
 When complete the produced debian packages will be located in deb_files_generated directory.
 
 Arguments for the deb-buildsource.sh script that can be passed are either 'None' or '"your custom comments"'. 'None' assumes a permanent changelog entry has been made listing the correct build for the version. If you quote and enter your own comment string, it will be appended to the changelog for that deb package. If no argument is passed a default comment is generated saying that this deb package was generated by this script.
 
-This script has been tested on xenial/bionic/focal/stretch/buster/bullseye
+This script has been tested on xenial/bionic/stretch/buster

pkg/deb/conf/keydb.conf

@@ -257,24 +257,6 @@ tcp-keepalive 300
 #
 # tls-session-cache-timeout 60
 
-# Allow the server to monitor the filesystem and rotate out TLS certificates if 
-# they change on disk, defaults to no.
-# 
-# tls-rotation no
-
-# Setup a allowlist of allowed Common Names (CNs)/Subject Alternative Names (SANs)
-# that are allowed to connect to this server. This includes both normal clients as
-# well as other servers connected for replication/clustering purposes. If nothing is
-# specified, then no allowlist is used and all certificates are accepted. 
-# Supports IPv4, DNS, RFC822, and URI SAN types.
-# You can put multiple names on one line as follows:
-#
-# tls-allowlist <dns1> <dns2> <dns3> ...
-# 
-#
-# This configuration also allows for wildcard characters with glob style formatting
-# i.e. "*.com" would allow all clients to connect with a CN/SAN that ends with ".com"
-
 ################################# GENERAL #####################################
 
 # By default KeyDB does not run as a daemon. Use 'yes' if you need it.
@@ -2053,13 +2035,9 @@ server-threads 2
 # bringing up replicas can result in data loss (the first master will win).
 # active-replica yes
 
-# KeyDB will attempt to balance clients across threads evenly; However, replica clients
-# are usually much more expensive than a normal client, and so KeyDB will try to assign
-# fewer clients to threads with a replica.  The weighting factor below is intented to help tune
-# this behavior.  A replica weighting factor of 2 means we treat a replica as the equivalent
-# of two normal clients.  Adjusting this value may improve performance when replication is
-# used.  The best weighting is workload specific - e.g. read heavy workloads should set
-# this to 1.  Very write heavy workloads may benefit from higher numbers.
-#
-# By default KeyDB sets this to 2.
-replica-weighting-factor 2
+# Enable Pro?  KeyDB pro provides support for pro only features
+#   note: you may omit the license key to demo pro features for a limited time
+# enable-pro [License Key]
+
+# Enable FLASH support? (Pro Only)
+# storage-provider flash /path/to/flash/db

pkg/deb/deb-buildsource.sh

@@ -53,9 +53,8 @@ if [ "$codename" == "xenial" ]; then
 else
         sudo pbuilder create --distribution $codename
 fi
-
 sudo pbuilder --update
-sudo pbuilder --build *.dsc  --logfile /mnt/pbuilderlog.log
+sudo pbuilder --build *.dsc
 
 # move new packages to deb_files_generated and clean up
 cp /var/cache/pbuilder/result/*$version*.deb ../deb_files_generated

pkg/deb/debian/control

@@ -8,7 +8,6 @@ Build-Depends:
  systemd,
  libsystemd-dev <!nocheck>,
  procps <!nocheck>,
- pkg-config <!nocheck>,
  build-essential <!nocheck>,
  tcl <!nocheck>,
  tcl-dev <!nocheck>,
@@ -18,16 +17,11 @@ Build-Depends:
  autotools-dev <!nocheck>,
  autoconf <!nocheck>,
  libjemalloc-dev <!nocheck>,
- libssl-dev <!nocheck>,
- libsnappy-dev <!nocheck>,
- zlib1g-dev <!nocheck>,
- libbz2-dev <!nocheck>,
- liblz4-dev <!nocheck>,
- libzstd-dev <!nocheck>
+ libssl-dev <!nocheck>
 Standards-Version: 4.2.1
 Homepage: https://docs.keydb.dev/
-Vcs-Git: https://github.com/Snapchat/KeyDB.git
-Vcs-Browser: https://github.com/Snapchat/KeyDB
+Vcs-Git: https://github.com/JohnSully/KeyDB.git
+Vcs-Browser: https://github.com/JohnSully/KeyDB
 
 Package: keydb
 Architecture: all

pkg/deb/debian/copyright

@@ -6,9 +6,8 @@ Source: https://github.com/EQ-Alpha/KeyDB
 Files: *
 Copyright: 
  © 2006-2014 Salvatore Sanfilippo <antirez@gmail.com>
- © 2019 John Sully
+ © 2019-2021 John Sully
  © 2019-2021 EQ Alpha Technology Ltd.
- © 2021 Snap Inc.
 
 License: BSD-3-clause
 
@@ -90,7 +89,6 @@ Files: pkg/deb/debian/*
 Copyright:
  © 2009 Chris Lamb <lamby@debian.org>
  © 2020-2021 EQ Alpha Technology Ltd. <support@eqalpha.com>
- © 2021 Snap Inc.
 License: BSD-3-clause
 
 License: BSD-2-clause

pkg/deb/debian/watch

@@ -1,3 +1,3 @@
 version=6
 opts=uversionmangle=s/-?(alpha|beta|rc)/~$1/ \
-	https://github.com/EQ-Alpha/KeyDB/releases .*/archive/(.*).tar.gz
+	https://github.com/JohnSully/KeyDB-Pro/releases .*/archive/(.*).tar.gz

pkg/deb/debian_dh9/control

@@ -17,16 +17,11 @@ Build-Depends:
  autotools-dev <!nocheck>,
  autoconf <!nocheck>,
  libjemalloc-dev <!nocheck>,
- libssl-dev <!nocheck>,
- libsnappy-dev <!nocheck>,
- zlib1g-dev <!nocheck>,
- libbz2-dev <!nocheck>,
- liblz4-dev <!nocheck>,
- libzstd-dev <!nocheck>
+ libssl-dev <!nocheck>
 Standards-Version: 4.2.1
 Homepage: https://docs.keydb.dev/
-Vcs-Git: https://github.com/Snapchat/KeyDB.git
-Vcs-Browser: https://github.com/Snapchat/KeyDB
+Vcs-Git: https://github.com/JohnSully/KeyDB.git
+Vcs-Browser: https://github.com/JohnSully/KeyDB
 
 Package: keydb
 Architecture: all

pkg/deb/debian_dh9/copyright

@@ -6,7 +6,7 @@ Source: https://github.com/EQ-Alpha/KeyDB
 Files: *
 Copyright: 
  © 2006-2014 Salvatore Sanfilippo <antirez@gmail.com>
- © 2019 John Sully
+ © 2019-2021 John Sully
  © 2019-2021 EQ Alpha Technology Ltd.
 
 License: BSD-3-clause

pkg/deb/debian_dh9/watch

@@ -1,3 +1,3 @@
 version=4
 opts=uversionmangle=s/-?(alpha|beta|rc)/~$1/ \
-	https://github.com/EQ-Alpha/KeyDB/releases .*/archive/(.*).tar.gz
+	https://github.com/JohnSully/KeyDB/releases .*/archive/(.*).tar.gz

pkg/deb/master_changelog

@@ -1,35 +1,3 @@
-keydb (6:6.3.1-1distribution_placeholder) codename_placeholder; urgency=medium
-
-  * This point release contains fixes to bugs related to expires, active-rep, and rdb saving
-  * Issues fixed: #419, #422, #428
-  * PRs: #426, #429, #431, #433
-
--- Ben Schermel <ben@eqalpha.com>  Mon, 23 May 2022 20:00:37 +0000
-
-keydb (6:6.3.0-1distribution_placeholder) codename_placeholder; urgency=medium
-
-  * This release open sources KeyDB Enterprise features into the open source project along with PSYNC for active replication
-  * Partial synchronization for active replication is introduced
-  * MVCC introduced into codebase from KeyDB Enterprise
-  * Async commands added: GET, MGET. These will see perf improvements
-  * KEYS and SCAN commands will no longer be blocking calls
-  * Async Rehash implemented for additional stability to perf
-  * IStorage interface added
-  * In-process background saving (forkless) to comply with maxmemory setting
-  * See v6.3.0 tagged release notes on github for a detailed explanation of these changes
-
--- Ben Schermel <ben@eqalpha.com>  Wed, 11 May 2022 20:00:37 +0000
-
-keydb (6:6.2.2-1distribution_placeholder) codename_placeholder; urgency=medium
-
-  * Acquire lock in module.cpp to fix module test break
-  * Fix usage of nanosleep() on Apple hardware
-  * Fix networking deadlock (#183)
-  * Remove redundant zfree() in keydb-benchmark
-  * Issues resolved: #378, #384
-
--- Ben Schermel <ben@eqalpha.com>  Sat, 15 Jan 2022 20:00:37 +0000
-
 keydb (6:6.2.1-1distribution_placeholder) codename_placeholder; urgency=medium
 
   * This release of KeyDB is at parity with Redis 6.2.6. In addition, the following changes were included:
@@ -99,7 +67,7 @@ keydb (6:6.0.8-1distribution_placeholder) codename_placeholder; urgency=medium
 
   * This is our first release to fully support all Redis 6.0.4 features, including but not limited to: TLS Support (fully supports multithreading!), Client side caching, RESP 3 Support
   * KeyDB Has also added the following new features: Improved memory efficiency for short strings, Fastlock autotuning, KeyDB.HRENAME (rename a member of a hash)In addition we've spent a lot of time focussing on stability.
-  * The following bug fixes are resolved: 150 - Deadlock in ReplicationFeedMonitors, 170 - KeyDB dying via SIGABORT, 180 - crash after setting maxclients via cmd line, 169 - Write performance of a master dropped sharply when a replica is added
+  * The following bug fixes are resolved: 150 - Deadlock in ReplicationFeedMonitors, 170 - KeyDB dying via SIGABORT, 180 - crash after setting maxclients via cmd line, 169 - Write performance of a master dropped sharply when a slave is added
   * In addition untracked issues were resolved: Potential deadlock when entering futex sleep, Pub/Sub Async messages may not be sent in a timely manner when load is low, Excessive logging during failed RREPLAY, KeyDB unresponsive handling clients on different threads during RDB load
   * Updated deb package builds to build from source and phasing out init.d. dh_installsystemd from deb helper 11 used where applicable
   * Naming conventions are now updated [ keydbpackage_version-build~distribution_architecture] and master changelog included and maintained

pkg/docker/Dockerfile

@@ -1,112 +1,91 @@
-FROM ubuntu:20.04
-SHELL ["/bin/bash","-c"]
+FROM ubuntu:18.04
+# add our user and group first to make sure their IDs get assigned consistently, regardless of whatever dependencies get added
 RUN groupadd -r keydb && useradd -r -g keydb keydb
-# use gosu for easy step-down from root: https://github.com/tianon/gosu/releases
-ENV GOSU_VERSION 1.14
+
+# grab gosu for easy step-down from root
+# https://github.com/tianon/gosu/releases
+ENV GOSU_VERSION 1.11
 RUN set -eux; \
-        savedAptMark="$(apt-mark showmanual)"; \
-        apt-get update; \
-        apt-get install -y --no-install-recommends ca-certificates dirmngr gnupg wget; \
-        rm -rf /var/lib/apt/lists/*; \
-        dpkgArch="$(dpkg --print-architecture | awk -F- '{ print $NF }')"; \
-        wget -O /usr/local/bin/gosu "https://github.com/tianon/gosu/releases/download/$GOSU_VERSION/gosu-$dpkgArch"; \
-        wget -O /usr/local/bin/gosu.asc "https://github.com/tianon/gosu/releases/download/$GOSU_VERSION/gosu-$dpkgArch.asc"; \
-        export GNUPGHOME="$(mktemp -d)"; \
-        gpg --batch --keyserver hkps://keys.openpgp.org --recv-keys B42F6819007F00F88E364FD4036A9C25BF357DD4; \
-        gpg --batch --verify /usr/local/bin/gosu.asc /usr/local/bin/gosu; \
-        gpgconf --kill all; \
-        rm -rf "$GNUPGHOME" /usr/local/bin/gosu.asc; \
-        apt-mark auto '.*' > /dev/null; \
-        [ -z "$savedAptMark" ] || apt-mark manual $savedAptMark > /dev/null; \
-        apt-get purge -y --auto-remove -o APT::AutoRemove::RecommendsImportant=false; \
-        chmod +x /usr/local/bin/gosu; \
-        gosu --version; \
-        gosu nobody true
-# build KeyDB
-ARG BRANCH
-RUN set -eux; \
-        \
-        savedAptMark="$(apt-mark showmanual)"; \
-        apt-get update; \
-        DEBIAN_FRONTEND=noninteractive apt-get install -qqy --no-install-recommends \
-                dpkg-dev \
-                pkg-config \
-                ca-certificates \
-                build-essential \
-                nasm \
-                autotools-dev \
-                autoconf \
-                libjemalloc-dev \
-                tcl \
-                tcl-dev \
-                uuid-dev \
-                libcurl4-openssl-dev \
-                libbz2-dev \
-                libzstd-dev \
-                liblz4-dev \
-                libsnappy-dev \
-                libssl-dev \
-                git; \
-        cd /tmp && git clone --branch $BRANCH https://github.com/Snapchat/KeyDB.git --recursive; \
-        cd /tmp/KeyDB; \
-        # disable protected mode as it relates to docker
-        grep -E '^ *createBoolConfig[(]"protected-mode",.*, *1 *,.*[)],$' ./src/config.cpp; \
-        sed -ri 's!^( *createBoolConfig[(]"protected-mode",.*, *)1( *,.*[)],)$!\10\2!' ./src/config.cpp; \
-        grep -E '^ *createBoolConfig[(]"protected-mode",.*, *0 *,.*[)],$' ./src/config.cpp; \
-        make -j$(nproc) BUILD_TLS=yes; \
-        cd src; \
-        strip keydb-cli keydb-benchmark keydb-check-rdb keydb-check-aof keydb-diagnostic-tool keydb-sentinel keydb-server; \
-        mv keydb-server keydb-cli keydb-benchmark keydb-check-rdb keydb-check-aof keydb-diagnostic-tool keydb-sentinel /usr/local/bin/; \
-        # clean up unused dependencies
-        echo $savedAptMark; \
-        apt-mark auto '.*' > /dev/null; \
-        [ -z "$savedAptMark" ] || apt-mark manual $savedAptMark > /dev/null; \
-        find /usr/local -type f -executable -exec ldd '{}' ';' \
-               | awk '/=>/ { print $(NF-1) }' \
-               | sed 's:.*/::' \
-               | sort -u \
-               | xargs -r dpkg-query --search \
-               | cut -d: -f1 \
-               | sort -u \
-               | xargs -r apt-mark manual \
-        ; \
-        apt-get purge -y --auto-remove -o APT::AutoRemove::RecommendsImportant=false; \
-        rm -rf /var/lib/apt/lists/*; \
-# create working directories and organize files
+# save list of currently installed packages for later so we can clean up
+	savedAptMark="$(apt-mark showmanual)"; \
+	apt-get update; \
+	apt-get install -y --no-install-recommends \
+		ca-certificates \
+		dirmngr \
+		gnupg \
+		wget \
+	; \
+	\
+	dpkgArch="$(dpkg --print-architecture | awk -F- '{ print $NF }')"; \
+	wget -O /usr/local/bin/gosu "https://github.com/tianon/gosu/releases/download/$GOSU_VERSION/gosu-$dpkgArch"; \
+	wget -O /usr/local/bin/gosu.asc "https://github.com/tianon/gosu/releases/download/$GOSU_VERSION/gosu-$dpkgArch.asc"; \
+	\
+# verify the signature
+	export GNUPGHOME="$(mktemp -d)"; \
+	gpg --batch --keyserver hkps://keys.openpgp.org --recv-keys B42F6819007F00F88E364FD4036A9C25BF357DD4; \
+	gpg --batch --verify /usr/local/bin/gosu.asc /usr/local/bin/gosu; \
+	gpgconf --kill all; \
+	rm -rf "$GNUPGHOME" /usr/local/bin/gosu.asc; \
+	\
+# clean up fetch dependencies
+	apt-mark auto '.*' > /dev/null; \
+	[ -z "$savedAptMark" ] || apt-mark manual $savedAptMark > /dev/null; \
+	apt-get purge -y --auto-remove -o APT::AutoRemove::RecommendsImportant=false; \
+	\
+	chmod +x /usr/local/bin/gosu; \
+# verify that the binary works
+	gosu --version; \
+	gosu nobody true
+
+# Load pre-generated equivalent binaries to image (reduces deployment build times)
 RUN \
-        mkdir /data && chown keydb:keydb /data; \
-        mkdir /flash && chown keydb:keydb /flash; \
-        mkdir -p /etc/keydb; \
-        cp /tmp/KeyDB/keydb.conf /etc/keydb/; \
-        sed -i 's/^\(daemonize .*\)$/# \1/' /etc/keydb/keydb.conf; \
-        sed -i 's/^\(dir .*\)$/# \1\ndir \/data/' /etc/keydb/keydb.conf; \
-        sed -i 's/^\(logfile .*\)$/# \1/' /etc/keydb/keydb.conf; \
-        sed -i 's/protected-mode yes/protected-mode no/g' /etc/keydb/keydb.conf; \
-        sed -i 's/^\(bind .*\)$/# \1/' /etc/keydb/keydb.conf; \
-        ln -s keydb-cli redis-cli; \
-        cd /etc/keydb; \
-        ln -s keydb.conf redis.conf; \
-        rm -rf /tmp/*
-# generate entrypoint script
+  mkdir -p /etc/keydb
+ADD ./app/keydb-* /usr/local/bin/
+ADD ./app/docker-entrypoint.sh /usr/local/bin/
+ADD ./app/*.conf /etc/keydb/
+# Set up config and binaries
+RUN \
+  cd /usr/local/bin && \
+  sed -i 's/^\(bind .*\)$/# \1/' /etc/keydb/keydb.conf && \
+  sed -i 's/^\(daemonize .*\)$/# \1/' /etc/keydb/keydb.conf && \
+  sed -i 's/^\(dir .*\)$/# \1\ndir \/data/' /etc/keydb/keydb.conf && \
+  sed -i 's/^\(logfile .*\)$/# \1/' /etc/keydb/keydb.conf && \
+  sed -i 's/protected-mode yes/protected-mode no/g' /etc/keydb/keydb.conf && \
+  ln -s keydb-cli redis-cli && \
+  cd /etc/keydb && \
+  ln -s keydb.conf redis.conf
+
+# Ensure deps installed for binaries
 RUN set -eux; \
-        echo '#!/bin/sh' > /usr/local/bin/docker-entrypoint.sh; \
-        echo 'set -e' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "# first arg is '-f' or '--some-option'" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "# or first arg is `something.conf`" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'if [ "${1#-}" != "$1" ] || [ "${1%.conf}" != "$1" ]; then' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo '        set -- keydb-server "$@"' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'fi' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "# allow the container to be started with `--user`" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'if [ "$1" = "keydb-server" -a "$(id -u)" = "0" ]; then' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "        find . \! -user keydb -exec chown keydb '{}' +" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo '        exec gosu keydb "$0" "$@"' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'fi' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'exec "$@"' >> /usr/local/bin/docker-entrypoint.sh; \
-        chmod +x /usr/local/bin/docker-entrypoint.sh
-# set remaining image properties
+	\
+	savedAptMark="$(apt-mark showmanual)"; \
+	apt-get install -y --no-install-recommends \
+		libcurl4 \
+		libc6 \
+		libssl1.1 \
+		libuuid1 \
+		libstdc++6 \
+		libgcc1 \
+		zlib1g \
+		libbz2-1.0 \
+		liblz4-1 \
+		libsnappy1v5 \		
+		libzstd1 \
+	; \
+	rm -rf /var/lib/apt/lists/*
+
+# create working directories
+RUN \
+  mkdir /data && chown keydb:keydb /data && \
+  mkdir /flash && chown keydb:keydb /flash
+
 VOLUME /data
 WORKDIR /data
 ENV KEYDB_PRO_DIRECTORY=/usr/local/bin/
+
+#COPY docker-entrypoint.sh /usr/local/bin/
 ENTRYPOINT ["docker-entrypoint.sh"]
+
+# Expose container port and start KeyDB by default on running container
 EXPOSE 6379
-CMD ["keydb-server","/etc/keydb/keydb.conf"]
+CMD ["keydb-server", "/etc/keydb/keydb.conf"]

pkg/docker/Dockerfile_Alpine

@@ -1,81 +0,0 @@
-FROM alpine:3.12
-# add our user and group first to make sure their IDs get assigned consistently, regardless of whatever dependencies get added
-RUN addgroup -S -g 1000 keydb && adduser -S -G keydb -u 999 keydb
-RUN mkdir -p /etc/keydb
-ARG BRANCH
-RUN set -eux; \
-        \
-        apk add --no-cache su-exec; \
-        apk add --no-cache --virtual .build-deps \
-                coreutils \
-                gcc \
-                linux-headers \
-                make \
-                musl-dev \
-                openssl-dev \
-                git \
-                util-linux-dev \
-                curl-dev \
-                g++ \
-                libunwind-dev \
-                bash \
-                perl \
-                git \
-                bzip2-dev \
-                zstd-dev \
-                lz4-dev \
-                snappy-dev \
-        ; \
-        cd /tmp && git clone --branch $BRANCH https://github.com/Snapchat/KeyDB.git --recursive; \
-        cd /tmp/KeyDB; \
-        # disable protected mode as it relates to docker
-        grep -E '^ *createBoolConfig[(]"protected-mode",.*, *1 *,.*[)],$' ./src/config.cpp; \
-        sed -ri 's!^( *createBoolConfig[(]"protected-mode",.*, *)1( *,.*[)],)$!\10\2!' ./src/config.cpp; \
-        grep -E '^ *createBoolConfig[(]"protected-mode",.*, *0 *,.*[)],$' ./src/config.cpp; \
-        make -j$(nproc) BUILD_TLS=yes; \
-        cd src; \
-        strip keydb-cli keydb-benchmark keydb-check-rdb keydb-check-aof keydb-diagnostic-tool keydb-sentinel keydb-server; \
-        mv keydb-server keydb-cli keydb-benchmark keydb-check-rdb keydb-check-aof keydb-diagnostic-tool keydb-sentinel /usr/local/bin/; \
-        runDeps="$( \
-                scanelf --needed --nobanner --format '%n#p' --recursive /usr/local \
-                        | tr ',' '\n' \
-                        | sort -u \
-                        | awk 'system("[ -e /usr/local/lib/" $1 " ]") == 0 { next } { print "so:" $1 }' \
-        )"; \
-        apk add --no-network --virtual .keydb-rundeps $runDeps; \
-        apk del --no-network .build-deps; \
-        # create working directories and organize files
-        mkdir /data && chown keydb:keydb /data; \
-        mkdir /flash && chown keydb:keydb /flash; \
-        mkdir -p /etc/keydb; \
-        cp /tmp/KeyDB/keydb.conf /etc/keydb/; \
-        sed -i 's/^\(daemonize .*\)$/# \1/' /etc/keydb/keydb.conf; \
-        sed -i 's/^\(dir .*\)$/# \1\ndir \/data/' /etc/keydb/keydb.conf; \
-        sed -i 's/^\(logfile .*\)$/# \1/' /etc/keydb/keydb.conf; \
-        sed -i 's/protected-mode yes/protected-mode no/g' /etc/keydb/keydb.conf; \
-        sed -i 's/^\(bind .*\)$/# \1/' /etc/keydb/keydb.conf; \
-        ln -s keydb-cli redis-cli; \
-        cd /etc/keydb; \
-        ln -s keydb.conf redis.conf; \
-        rm -rf /tmp/*
-# generate entrypoint script
-RUN set -eux; \
-        echo '#!/bin/sh' > /usr/local/bin/docker-entrypoint.sh; \
-        echo 'set -e' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "# first arg is '-f' or '--some-option'" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "# or first arg is `something.conf`" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'if [ "${1#-}" != "$1" ] || [ "${1%.conf}" != "$1" ]; then' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo '        set -- keydb-server "$@"' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'fi' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "# allow the container to be started with `--user`" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'if [ "$1" = "keydb-server" -a "$(id -u)" = "0" ]; then' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "        find . \! -user keydb -exec chown keydb '{}' +" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo '        exec su-exec keydb "$0" "$@"' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'fi' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'exec "$@"' >> /usr/local/bin/docker-entrypoint.sh; \
-        chmod +x /usr/local/bin/docker-entrypoint.sh
-VOLUME /data
-WORKDIR /data
-ENTRYPOINT ["docker-entrypoint.sh"]
-EXPOSE 6379
-CMD ["keydb-server", "/etc/keydb/keydb.conf"]

pkg/docker/README.md

@@ -1,7 +1,5 @@
-This Dockerfile will clone the KeyDB repo, build, and generate a Docker image you can use
-
-To build, use experimental mode to enable use of build args. Tag the build and specify branch name. The command below will generate your docker image:
+In order to create a docker image, generate the keydb binaries, copy them to the app directory, copy keydb.conf and sentinel.conf to the app directory as well, then run the following command:
 
 ```
-DOCKER_CLI_EXPERIMENTAL=enabled docker build --build-arg BRANCH=<keydbBranch> -t <yourImageName>
+$ sudo docker build . -t <yourimagename>
 ```

pkg/docker/app/docker-entrypoint.sh

@@ -0,0 +1,16 @@
+#!/bin/sh
+set -e
+
+# first arg is `-f` or `--some-option`
+# or first arg is `something.conf`
+if [ "${1#-}" != "$1" ] || [ "${1%.conf}" != "$1" ]; then
+	set -- keydb-server "$@"
+fi
+
+# allow the container to be started with `--user`
+if [ "$1" = 'keydb-server' -a "$(id -u)" = '0' ]; then
+	find . \! -user keydb -exec chown keydb '{}' +
+	exec gosu keydb "$0" "$@"
+fi
+
+exec "$@"

pkg/rpm/README.md

@@ -1,6 +1,6 @@
 ### Generate RPM files for the generated binaries
 
-After making the binaries you can run the following script
+After running make to produce keydb binaries you can run the following script to create rpm package
 
 Usage: 
 ```

pkg/rpm/keydb_build/keydb.spec

@@ -1,13 +1,14 @@
 Name        : keydb
-Version     : 6.0.5
+Version     : 5.3.3
 Release     : 1%{?dist}
 Group       : Unspecified
 License     : BSD
-Packager    : Snap Inc.
+Packager    : EQ Alpha Technology
 
 URL         : https://keydb.dev
 Summary     : A persistent key-value database
 
+
 Requires:          /bin/awk
 Requires:          logrotate
 Requires(pre):     shadow-utils

pkg/rpm/keydb_build/keydb_rpm/etc/keydb/keydb.conf

@@ -257,24 +257,6 @@ tcp-keepalive 300
 #
 # tls-session-cache-timeout 60
 
-# Allow the server to monitor the filesystem and rotate out TLS certificates if 
-# they change on disk, defaults to no.
-# 
-# tls-rotation no
-
-# Setup a allowlist of allowed Common Names (CNs)/Subject Alternative Names (SANs)
-# that are allowed to connect to this server. This includes both normal clients as
-# well as other servers connected for replication/clustering purposes. If nothing is
-# specified, then no allowlist is used and all certificates are accepted. 
-# Supports IPv4, DNS, RFC822, and URI SAN types.
-# You can put multiple names on one line as follows:
-#
-# tls-allowlist <dns1> <dns2> <dns3> ...
-# 
-#
-# This configuration also allows for wildcard characters with glob style formatting
-# i.e. "*.com" would allow all clients to connect with a CN/SAN that ends with ".com"
-
 ################################# GENERAL #####################################
 
 # By default KeyDB does not run as a daemon. Use 'yes' if you need it.
@@ -2053,13 +2035,9 @@ server-threads 2
 # bringing up replicas can result in data loss (the first master will win).
 # active-replica yes
 
-# KeyDB will attempt to balance clients across threads evenly; However, replica clients
-# are usually much more expensive than a normal client, and so KeyDB will try to assign
-# fewer clients to threads with a replica.  The weighting factor below is intented to help tune
-# this behavior.  A replica weighting factor of 2 means we treat a replica as the equivalent
-# of two normal clients.  Adjusting this value may improve performance when replication is
-# used.  The best weighting is workload specific - e.g. read heavy workloads should set
-# this to 1.  Very write heavy workloads may benefit from higher numbers.
-#
-# By default KeyDB sets this to 2.
-replica-weighting-factor 2
+# Enable Pro?  KeyDB pro provides support for pro only features
+#   note: you may omit the license key to demo pro features for a limited time
+# enable-pro [License Key]
+
+# Enable FLASH support? (Pro Only)
+# storage-provider flash /path/to/flash/db

pkg/rpm/keydb_build/keydb_rpm/usr/lib/systemd/system/keydb-sentinel.service

@@ -1,7 +1,7 @@
 [Unit]
 Description=Advanced key-value store
 After=network.target
-Documentation=http://keydb.io/documentation, man:keydb-sentinel(1)
+Documentation=https://docs.keydb.dev, man:keydb-sentinel(1)
 
 [Service]
 Type=forking

pkg/rpm/keydb_build/keydb_rpm/usr/share/licenses/keydb/COPYING

@@ -1,8 +1,5 @@
-Copyright (c) 2006-2020, Salvatore Sanfilippo
+Copyright (c) 2006-2015, Salvatore Sanfilippo
 Copyright (C) 2019, John Sully
-Copyright (C) 2020-2021, EQ Alpha Technology Ltd.
-Copyright (C) 2021, Snap Inc.
-
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

runtest-rotation

@@ -1,31 +0,0 @@
-#!/bin/sh
-TCL_VERSIONS="8.5 8.6"
-TCLSH=""
-
-export ASAN_OPTIONS=allocator_may_return_null=1  $ASAN_OPTIONS
-
-for VERSION in $TCL_VERSIONS; do
-	TCL=`which tclsh$VERSION 2>/dev/null` && TCLSH=$TCL
-done
-
-if [ -z $TCLSH ]
-then
-    echo "You need tcl 8.5 or newer in order to run the KeyDB test"
-    exit 1
-fi
-if [ ! -r tests/tls ] || [ ! -r tests/tls_1 ] || [ ! -r tests/tls_2 ];
-then
-    echo "Generating neccessary certificates for TLS rotation testing."
-    rm -rf tests/tls tests/tls_1 tests/tls_2
-
-    utils/gen-test-certs.sh
-    mv tests/tls tests/tls_1
-    utils/gen-test-certs.sh
-    mv tests/tls tests/tls_2
-    utils/gen-test-certs.sh
-fi
-$TCLSH tests/test_helper.tcl \
---single unit/tls-rotation \
---tls \
---config server-threads 3 \
-"${@}"

src/AsyncWorkQueue.cpp

@@ -1,131 +0,0 @@
-#include "AsyncWorkQueue.h"
-#include "server.h"
-
-AsyncWorkQueue::AsyncWorkQueue(int nthreads)
-{
-    for (int i = 0; i < nthreads; ++i)
-    {
-        m_vecthreads.emplace_back([&]{
-            WorkerThreadMain();
-        });
-    }
-}
-
-void AsyncWorkQueue::WorkerThreadMain()
-{
-    redisServerThreadVars vars;
-    serverTL = &vars;
-
-    vars.clients_pending_asyncwrite = listCreate();
-
-    m_mutex.lock();
-    m_vecpthreadVars.push_back(&vars);
-    m_mutex.unlock();
-
-    while (!m_fQuitting)
-    {
-        std::unique_lock<std::mutex> lock(m_mutex);
-        if (m_workqueue.empty())
-            m_cvWakeup.wait(lock);
-
-        aeThreadOnline();
-        while (!m_workqueue.empty())
-        {
-            WorkItem task = std::move(m_workqueue.front());
-            m_workqueue.pop_front();
-
-            lock.unlock();
-            serverTL->gcEpoch = g_pserver->garbageCollector.startEpoch();
-            task.fnAsync();
-            g_pserver->garbageCollector.endEpoch(serverTL->gcEpoch);
-            lock.lock();
-        }
-
-        lock.unlock();
-        serverTL->gcEpoch = g_pserver->garbageCollector.startEpoch();
-        if (listLength(serverTL->clients_pending_asyncwrite)) {
-            aeAcquireLock();
-            ProcessPendingAsyncWrites();
-            aeReleaseLock();
-        }
-        g_pserver->garbageCollector.endEpoch(serverTL->gcEpoch);
-        serverTL->gcEpoch.reset();
-        aeThreadOffline();
-    }
-
-    listRelease(vars.clients_pending_asyncwrite);
-
-    std::unique_lock<fastlock> lockf(serverTL->lockPendingWrite);
-    serverTL->vecclientsProcess.clear();
-    serverTL->clients_pending_write.clear();
-    std::atomic_thread_fence(std::memory_order_seq_cst);
-}
-
-bool AsyncWorkQueue::removeClientAsyncWrites(client *c)
-{
-    bool fFound = false;
-    aeAcquireLock();
-    m_mutex.lock();
-    for (auto pvars : m_vecpthreadVars)
-    {
-        listIter li;
-        listNode *ln;
-        listRewind(pvars->clients_pending_asyncwrite, &li);
-        while ((ln = listNext(&li)) != nullptr)
-        {
-            if (c == listNodeValue(ln))
-            {
-                listDelNode(pvars->clients_pending_asyncwrite, ln);
-                fFound = true;
-            }
-        }
-    }
-    m_mutex.unlock();
-    aeReleaseLock();
-    return fFound;
-}
-
-void AsyncWorkQueue::shutdown()
-{
-    std::unique_lock<std::mutex> lock(m_mutex);
-    serverAssert(!GlobalLocksAcquired());
-    m_fQuitting = true;
-    m_cvWakeup.notify_all();
-    lock.unlock();
-
-    for (auto &thread : m_vecthreads)
-        thread.join();
-}
-
-void AsyncWorkQueue::abandonThreads()
-{
-    std::unique_lock<std::mutex> lock(m_mutex);
-    m_fQuitting = true;
-    m_cvWakeup.notify_all();
-    for (auto &thread : m_vecthreads)
-    {
-        thread.detach();
-    }
-    m_vecthreads.clear();
-}
-
-AsyncWorkQueue::~AsyncWorkQueue()
-{
-    serverAssert(!GlobalLocksAcquired() || m_vecthreads.empty());
-    std::unique_lock<std::mutex> lock(m_mutex);
-    m_fQuitting = true;
-    m_cvWakeup.notify_all();
-    lock.unlock();
-    
-    abandonThreads();
-}
-
-void AsyncWorkQueue::AddWorkFunction(std::function<void()> &&fnAsync, bool fHiPri)
-{
-    std::unique_lock<std::mutex> lock(m_mutex);
-    if (fHiPri)
-        m_workqueue.emplace_front(std::move(fnAsync));
-    else
-        m_workqueue.emplace_back(std::move(fnAsync));
-    m_cvWakeup.notify_one();
-}

src/AsyncWorkQueue.h

@@ -1,40 +0,0 @@
-#pragma once
-#include "fastlock.h"
-#include <vector>
-#include <deque>
-#include <mutex>
-#include <thread>
-#include <condition_variable>
-#include <atomic>
-#include <functional>
-
-class AsyncWorkQueue
-{
-    struct WorkItem
-    {
-        WorkItem(std::function<void()> &&fnAsync)
-            : fnAsync(std::move(fnAsync))
-            {}
-
-        WorkItem(WorkItem&&) = default;
-        std::function<void()> fnAsync;
-    };
-    std::vector<std::thread> m_vecthreads;
-    std::vector<struct redisServerThreadVars*> m_vecpthreadVars;
-    std::deque<WorkItem> m_workqueue;
-    std::mutex m_mutex;
-    std::condition_variable m_cvWakeup;
-    std::atomic<bool> m_fQuitting { false };
-
-    void WorkerThreadMain();
-public:
-    AsyncWorkQueue(int nthreads);
-    ~AsyncWorkQueue();
-
-    void AddWorkFunction(std::function<void()> &&fnAsync, bool fHiPri = false);
-    bool removeClientAsyncWrites(struct client *c);
-
-    void shutdown();
-
-    void abandonThreads();
-};

src/IStorage.h

@@ -1,54 +0,0 @@
-#pragma once
-#include <functional>
-#include "sds.h"
-
-#define METADATA_DB_IDENTIFIER "c299fde0-6d42-4ec4-b939-34f680ffe39f"
-
-class IStorageFactory
-{
-public:
-    typedef void (*key_load_iterator)(const char *rgchKey, size_t cchKey, void *privdata);
-
-    virtual ~IStorageFactory() {}
-    virtual class IStorage *create(int db, key_load_iterator itr, void *privdata) = 0;
-    virtual class IStorage *createMetadataDb() = 0;
-    virtual const char *name() const = 0;
-    virtual size_t totalDiskspaceUsed() const = 0;
-    virtual bool FSlow() const = 0;
-    virtual size_t filedsRequired() const { return 0; }
-};
-
-class IStorage
-{
-public:
-    typedef std::function<bool(const char *, size_t, const void *, size_t)> callback;
-    typedef std::function<void(const char *, size_t, const void *, size_t)> callbackSingle;
-
-    virtual ~IStorage();
-
-    virtual void insert(const char *key, size_t cchKey, void *data, size_t cb, bool fOverwire) = 0;
-    virtual bool erase(const char *key, size_t cchKey) = 0;
-    virtual void retrieve(const char *key, size_t cchKey, callbackSingle fn) const = 0;
-    virtual size_t clear() = 0;
-    virtual bool enumerate(callback fn) const = 0;
-    virtual size_t count() const = 0;
-
-    virtual void bulkInsert(char **rgkeys, size_t *rgcbkeys, char **rgvals, size_t *rgcbvals, size_t celem) {
-        beginWriteBatch();
-        for (size_t ielem = 0; ielem < celem; ++ielem) {
-            insert(rgkeys[ielem], rgcbkeys[ielem], rgvals[ielem], rgcbvals[ielem], false);
-        }
-        endWriteBatch();
-    }
-
-    virtual void beginWriteBatch() {} // NOP
-    virtual void endWriteBatch() {} // NOP
-
-    virtual void batch_lock() {} // NOP
-    virtual void batch_unlock() {} // NOP
-
-    virtual void flush() = 0;
-
-    /* This is permitted to be a shallow clone */
-    virtual const IStorage *clone() const = 0;
-};

src/Makefile

@@ -12,15 +12,11 @@
 # Dependencies are stored in the Makefile.dep file. To rebuild this file
 # Just use 'make dep', but this is only needed by developers.
 
-# we ship KeyDB with TLS by default
-# export it here as both this file and deps/Makefile uses it
-export BUILD_TLS ?= yes
-
 release_hdr := $(shell sh -c './mkreleasehdr.sh')
 uname_S := $(shell sh -c 'uname -s 2>/dev/null || echo not')
 uname_M := $(shell sh -c 'uname -m 2>/dev/null || echo not')
-OPTIMIZATION?=-O2 -flto
-DEPENDENCY_TARGETS=hiredis linenoise lua hdr_histogram rocksdb
+OPTIMIZATION?=-O2
+DEPENDENCY_TARGETS=hiredis linenoise lua hdr_histogram
 NODEPS:=clean distclean
 
 # Default settings
@@ -62,7 +58,6 @@ endif
 endif
 
 USEASM?=true
-ENABLE_FLASH?=no
 
 ifneq ($(strip $(SANITIZE)),)
 	CFLAGS+= -fsanitize=$(SANITIZE) -DSANITIZE  -fno-omit-frame-pointer
@@ -72,19 +67,6 @@ ifneq ($(strip $(SANITIZE)),)
 	USEASM=false
 endif
 
-ifeq ($(ENABLE_FLASH),yes)
-	FINAL_LIBS+= -lz -lcrypto -lbz2 -lzstd -llz4 -lsnappy
-	CXXFLAGS+= -I../deps/rocksdb/include/ -DENABLE_ROCKSDB
-	STORAGE_OBJ+= storage/rocksdb.o storage/rocksdbfactory.o
-	FINAL_LIBS+= ../deps/rocksdb/librocksdb.a 
-	DEPENDENCY_TARGETS+= rocksdb
-endif
-
-
-ifeq ($(CHECKED),true)
-	CXXFLAGS+= -DCHECKED_BUILD
-endif
-
 # Do we use our assembly spinlock? X64 only
 ifeq ($(uname_S),Linux)
 ifeq ($(uname_M),x86_64)
@@ -100,6 +82,7 @@ endif
 
 ifeq ($(COMPILER_NAME),clang)
 	CXXFLAGS+= -stdlib=libc++ 
+	LDFLAGS+= -latomic
 endif
 
 # To get ARM stack traces if KeyDB crashes we need a special C flag.
@@ -130,19 +113,14 @@ ifeq ($(USE_JEMALLOC),no)
 	MALLOC=libc
 endif
 
-
-ifeq ($(NO_LICENSE_CHECK),yes)
-	CXXFLAGS+=-DNO_LICENSE_CHECK=1
-endif	
-
 # Override default settings if possible
 -include .make-settings
 
-DEBUG=-g -ggdb
 FINAL_CFLAGS=$(STD) $(WARN) $(OPT) $(DEBUG) $(CFLAGS) $(KEYDB_CFLAGS) $(REDIS_CFLAGS)
 FINAL_CXXFLAGS=$(CXX_STD) $(WARN) $(OPT) $(DEBUG) $(CXXFLAGS) $(KEYDB_CFLAGS) $(REDIS_CFLAGS)
 FINAL_LDFLAGS=$(LDFLAGS) $(KEYDB_LDFLAGS) $(DEBUG)
-FINAL_LIBS+=-lm -lz -lcrypto -lbz2 -lzstd -llz4 -lsnappy
+FINAL_LIBS+=-lm
+DEBUG=-g -ggdb
 
 ifneq ($(uname_S),Darwin)
 	FINAL_LIBS+=-latomic
@@ -252,7 +230,6 @@ endif
 
 ifdef OPENSSL_PREFIX
 	OPENSSL_CFLAGS=-I$(OPENSSL_PREFIX)/include
-	OPENSSL_CXXFLAGS=-I$(OPENSSL_PREFIX)/include
 	OPENSSL_LDFLAGS=-L$(OPENSSL_PREFIX)/lib
 	# Also export OPENSSL_PREFIX so it ends up in deps sub-Makefiles
 	export OPENSSL_PREFIX
@@ -260,7 +237,7 @@ endif
 
 # Include paths to dependencies
 FINAL_CFLAGS+= -I../deps/hiredis -I../deps/linenoise -I../deps/lua/src -I../deps/hdr_histogram
-FINAL_CXXFLAGS+= -I../deps/hiredis -I../deps/linenoise -I../deps/lua/src -I../deps/hdr_histogram -I../deps/rocksdb/include/ -I../deps/concurrentqueue
+FINAL_CXXFLAGS+= -I../deps/hiredis -I../deps/linenoise -I../deps/lua/src -I../deps/hdr_histogram
 
 # Determine systemd support and/or build preference (defaulting to auto-detection)
 BUILD_WITH_SYSTEMD=no
@@ -369,17 +346,15 @@ ENDCOLOR="\033[0m"
 
 ifndef V
 QUIET_CC = @printf '    %b %b\n' $(CCCOLOR)CC$(ENDCOLOR) $(SRCCOLOR)$@$(ENDCOLOR);
-QUIET_CP = @printf '    %b %b\n' $(CCCOLOR)COPY$(ENDCOLOR) $(SRCCOLOR)$@$(ENDCOLOR);
 QUIET_LINK = @printf '    %b %b\n' $(LINKCOLOR)LINK$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR);
 QUIET_INSTALL = @printf '    %b %b\n' $(LINKCOLOR)INSTALL$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR);
 endif
 
 REDIS_SERVER_NAME=keydb-server$(PROG_SUFFIX)
 REDIS_SENTINEL_NAME=keydb-sentinel$(PROG_SUFFIX)
-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 t_nhash.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 crcspeed.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.o localtime.o acl.o storage.o rdb-s3.o fastlock.o new.o tracking.o cron.o connection.o tls.o sha256.o motd_server.o timeout.o setcpuaffinity.o AsyncWorkQueue.o snapshot.o storage/rocksdb.o storage/rocksdbfactory.o storage/teststorageprovider.o keydbutils.o StorageCache.o monotonic.o cli_common.o mt19937-64.o $(ASM_OBJ)
-KEYDB_SERVER_OBJ=SnapshotPayloadParseState.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 t_nhash.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 crcspeed.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.o localtime.o acl.o storage.o rdb-s3.o fastlock.o new.o tracking.o cron.o connection.o tls.o sha256.o motd.o timeout.o setcpuaffinity.o monotonic.o mt19937-64.o $(ASM_OBJ)
 REDIS_CLI_NAME=keydb-cli$(PROG_SUFFIX)
-REDIS_CLI_OBJ=anet.o adlist.o dict.o redis-cli.o redis-cli-cpphelper.o zmalloc.o release.o anet.o ae.o crcspeed.o crc64.o siphash.o crc16.o storage-lite.o fastlock.o motd_client.o monotonic.o cli_common.o mt19937-64.o $(ASM_OBJ)
+REDIS_CLI_OBJ=anet.o adlist.o dict.o redis-cli.o redis-cli-cpphelper.o zmalloc.o release.o anet.o ae.o crcspeed.o crc64.o siphash.o crc16.o storage-lite.o fastlock.o new.o motd.o monotonic.o cli_common.o mt19937-64.o $(ASM_OBJ)
 REDIS_BENCHMARK_NAME=keydb-benchmark$(PROG_SUFFIX)
 REDIS_BENCHMARK_OBJ=ae.o anet.o redis-benchmark.o adlist.o dict.o zmalloc.o release.o crcspeed.o crc64.o siphash.o redis-benchmark.o storage-lite.o fastlock.o new.o monotonic.o cli_common.o mt19937-64.o $(ASM_OBJ)
 REDIS_CHECK_RDB_NAME=keydb-check-rdb$(PROG_SUFFIX)
@@ -436,8 +411,8 @@ endif
 	@touch $@
 
 # keydb-server
-$(REDIS_SERVER_NAME): $(REDIS_SERVER_OBJ) $(KEYDB_SERVER_OBJ)
-	$(REDIS_LD) -o $@ $^ ../deps/hiredis/libhiredis.a ../deps/lua/src/liblua.a ../deps/rocksdb/librocksdb.a $(FINAL_LIBS)
+$(REDIS_SERVER_NAME): $(REDIS_SERVER_OBJ)
+	$(REDIS_LD) -o $@ $^ ../deps/hiredis/libhiredis.a ../deps/lua/src/liblua.a $(FINAL_LIBS)
 
 # keydb-sentinel
 $(REDIS_SENTINEL_NAME): $(REDIS_SERVER_NAME)
@@ -463,18 +438,12 @@ $(REDIS_BENCHMARK_NAME): $(REDIS_BENCHMARK_OBJ)
 $(KEYDB_DIAGNOSTIC_NAME): $(KEYDB_DIAGNOSTIC_OBJ)
 	$(REDIS_LD) -o $@ $^ ../deps/hiredis/libhiredis.a $(FINAL_LIBS)
 
-DEP = $(REDIS_SERVER_OBJ:%.o=%.d) $(KEYDB_SERVER_OBJ:%.o=%.d) $(REDIS_CLI_OBJ:%.o=%.d) $(REDIS_BENCHMARK_OBJ:%.o=%.d)
+DEP = $(REDIS_SERVER_OBJ:%.o=%.d) $(REDIS_CLI_OBJ:%.o=%.d) $(REDIS_BENCHMARK_OBJ:%.o=%.d)
 -include $(DEP)
 
 # Because the jemalloc.h header is generated as a part of the jemalloc build,
 # building it should complete before building any other object. Instead of
 # depending on a single artifact, build all dependencies first.
-motd_client.o: motd.cpp .make-prerequisites
-	$(REDIS_CXX) -MMD -o motd_client.o -c $< -DCLIENT -fno-lto
-
-motd_server.o: motd.cpp .make-prerequisites
-	$(REDIS_CXX) -MMD -o motd_server.o -c $< -DSERVER
-
 %.o: %.c .make-prerequisites
 	$(REDIS_CC) -MMD -o $@ -c $<
 
@@ -486,8 +455,6 @@ motd_server.o: motd.cpp .make-prerequisites
 
 clean:
 	rm -rf $(REDIS_SERVER_NAME) $(REDIS_SENTINEL_NAME) $(REDIS_CLI_NAME) $(REDIS_BENCHMARK_NAME) $(REDIS_CHECK_RDB_NAME) $(REDIS_CHECK_AOF_NAME) $(KEYDB_DIAGNOSTIC_NAME) *.o *.gcda *.gcno *.gcov KeyDB.info lcov-html Makefile.dep
-	rm -rf storage/*.o
-	rm -rf keydb-server
 	rm -f $(DEP)
 
 .PHONY: clean

src/SnapshotPayloadParseState.cpp

@@ -1,344 +0,0 @@
-#include "server.h"
-#include "SnapshotPayloadParseState.h"
-#include <sys/mman.h>
-
-static const size_t SLAB_SIZE = 2*1024*1024;   // Note 64 MB because we try to give 64MB to a block
-class SlabAllocator
-{
-    class Slab {
-        char *m_pv = nullptr;
-        size_t m_cb = 0;
-        size_t m_cbAllocated = 0;
-
-    public:
-        Slab(size_t cbAllocate) {
-            m_pv = (char*)zmalloc(cbAllocate);
-            m_cb = cbAllocate;
-        }
-
-        ~Slab() {
-            zfree(m_pv);
-        }
-
-        Slab(Slab &&other) {
-            m_pv = other.m_pv;
-            m_cb = other.m_cb;
-            m_cbAllocated = other.m_cbAllocated;
-            other.m_pv = nullptr;
-            other.m_cb = 0;
-            other.m_cbAllocated = 0;
-        }
-
-        bool canAllocate(size_t cb) const {
-            return (m_cbAllocated + cb) <= m_cb;
-        }
-
-        void *allocate(size_t cb) {
-            if ((m_cbAllocated + cb) > m_cb)
-                return nullptr;
-            void *pvret = m_pv + m_cbAllocated;
-            m_cbAllocated += cb;
-            return pvret;
-        }
-    };
-    std::vector<Slab> m_vecslabs;
-public:
-
-    void *allocate(size_t cb) {
-        if (m_vecslabs.empty() || !m_vecslabs.back().canAllocate(cb)) {
-            m_vecslabs.emplace_back(std::max(cb, SLAB_SIZE));
-        }
-        return m_vecslabs.back().allocate(cb);
-    }
-};
-
-static uint64_t dictCStringHash(const void *key) {
-    return dictGenHashFunction((unsigned char*)key, strlen((char*)key));
-}
-
-
-static void dictKeyDestructor(void *privdata, void *key)
-{
-    DICT_NOTUSED(privdata);
-    sdsfree((sds)key);
-}
-
-static int dictCStringCompare(void *, const void *key1, const void *key2)
-{
-    int l1,l2;
-
-    l1 = strlen((sds)key1);
-    l2 = strlen((sds)key2);
-    if (l1 != l2) return 0;
-    return memcmp(key1, key2, l1) == 0;
-}
-
-dictType metadataLongLongDictType = {
-    dictCStringHash,            /* hash function */
-    NULL,                       /* key dup */
-    NULL,                       /* val dup */
-    dictCStringCompare,         /* key compare */
-    dictKeyDestructor,          /* key destructor */
-    nullptr,                    /* val destructor */
-    nullptr,                    /* allow to expand */
-    nullptr                     /* async free destructor */
-};
-
-dictType metadataDictType = {
-    dictSdsHash,            /* hash function */
-    NULL,                       /* key dup */
-    NULL,                       /* val dup */
-    dictSdsKeyCompare,         /* key compare */
-    dictSdsDestructor,          /* key destructor */
-    dictSdsDestructor,          /* val destructor */
-    nullptr,                    /* allow to expand */
-    nullptr                     /* async free destructor */
-};
-
-
-SnapshotPayloadParseState::ParseStageName SnapshotPayloadParseState::getNextStage() {
-    if (stackParse.empty())
-        return ParseStageName::Global;
-
-    switch (stackParse.back().name)
-    {
-        case ParseStageName::None:
-            return ParseStageName::Global;
-
-        case ParseStageName::Global:
-            if (stackParse.back().arraycur == 0)
-                return ParseStageName::MetaData;
-            else if (stackParse.back().arraycur == 1)
-                return ParseStageName::Databases;
-            break;
-
-        case ParseStageName::MetaData:
-            return ParseStageName::KeyValuePair;
-
-        case ParseStageName::Databases:
-            return ParseStageName::Dataset;
-
-        case ParseStageName::Dataset:
-            return ParseStageName::KeyValuePair;
-
-        default:
-            break;
-    }
-    throw "Bad protocol: corrupt state";
-}
-
-void SnapshotPayloadParseState::flushQueuedKeys() {
-    if (vecqueuedKeys.empty())
-        return;
-    serverAssert(current_database >= 0);
-
-    // TODO: We can't finish parse until all the work functions finish
-    int idb = current_database;
-    serverAssert(vecqueuedKeys.size() == vecqueuedVals.size());
-    auto sizePrev = vecqueuedKeys.size();
-    ++insertsInFlight;
-    auto &insertsInFlightTmp = insertsInFlight; // C++ GRRRRRRRRRRRRRRRR, we don't want to capute "this" because that's dangerous
-    if (current_database < cserver.dbnum) {
-        g_pserver->asyncworkqueue->AddWorkFunction([idb, vecqueuedKeys = std::move(this->vecqueuedKeys), vecqueuedKeysCb = std::move(this->vecqueuedKeysCb), vecqueuedVals = std::move(this->vecqueuedVals), vecqueuedValsCb = std::move(this->vecqueuedValsCb), &insertsInFlightTmp, pallocator = m_spallocator.release()]() mutable {
-            g_pserver->db[idb]->bulkStorageInsert(vecqueuedKeys.data(), vecqueuedKeysCb.data(), vecqueuedVals.data(), vecqueuedValsCb.data(), vecqueuedKeys.size());
-            --insertsInFlightTmp;
-            delete pallocator;
-        });
-    } else {
-        // else drop the data
-        vecqueuedKeys.clear();
-        vecqueuedKeysCb.clear();
-        vecqueuedVals.clear();
-        vecqueuedValsCb.clear();
-        // Note: m_spallocator will get free'd when overwritten below
-    }
-    m_spallocator = std::make_unique<SlabAllocator>();
-    cbQueued = 0;
-    vecqueuedKeys.reserve(sizePrev);
-    vecqueuedKeysCb.reserve(sizePrev);
-    vecqueuedVals.reserve(sizePrev);
-    vecqueuedValsCb.reserve(sizePrev);
-    serverAssert(vecqueuedKeys.empty());
-    serverAssert(vecqueuedVals.empty());
-}
-
-
-SnapshotPayloadParseState::SnapshotPayloadParseState() {
-    // This is to represent the first array the client is intended to send us
-    ParseStage stage;
-    stage.name = ParseStageName::None;
-    stage.arraylen = 1;
-    stackParse.push_back(stage);
-
-    dictLongLongMetaData = dictCreate(&metadataLongLongDictType, nullptr);
-    dictMetaData = dictCreate(&metadataDictType, nullptr);
-    insertsInFlight = 0;
-    m_spallocator = std::make_unique<SlabAllocator>();
-}
-
-SnapshotPayloadParseState::~SnapshotPayloadParseState() {
-    dictRelease(dictLongLongMetaData);
-    dictRelease(dictMetaData);
-}
-
-const char *SnapshotPayloadParseState::getStateDebugName(ParseStage stage) {
-    switch (stage.name) {
-        case ParseStageName::None:
-            return "None";
-
-        case ParseStageName::Global:
-            return "Global";
-
-        case ParseStageName::MetaData:
-            return "MetaData";
-
-        case ParseStageName::Databases:
-            return "Databases";
-
-        case ParseStageName::KeyValuePair:
-            return "KeyValuePair";
-
-        case ParseStageName::Dataset:
-            return "Dataset";
-
-        default:
-            return "Unknown";
-    }
-}
-
-size_t SnapshotPayloadParseState::depth() const { return stackParse.size(); }
-
-void SnapshotPayloadParseState::trimState() {
-    while (!stackParse.empty() && (stackParse.back().arraycur == stackParse.back().arraylen))
-        stackParse.pop_back();
-    
-    if (stackParse.empty()) {
-        flushQueuedKeys();
-        while (insertsInFlight > 0) {
-            // TODO: ProcessEventsWhileBlocked
-            aeReleaseLock();
-            aeAcquireLock();
-        }
-    }
-}
-
-void SnapshotPayloadParseState::pushArray(long long size) {
-    if (stackParse.empty())
-        throw "Bad Protocol: unexpected trailing data";
-
-    if (size == 0) {
-        stackParse.back().arraycur++;
-        return;
-    }
-    
-    if (stackParse.back().name == ParseStageName::Databases) {
-        flushQueuedKeys();
-        current_database = stackParse.back().arraycur;
-    }
-
-    ParseStage stage;
-    stage.name = getNextStage();
-    stage.arraylen = size;
-
-    if (stage.name == ParseStageName::Dataset) {
-        g_pserver->db[current_database]->expand(stage.arraylen);
-    }
-
-    // Note: This affects getNextStage so ensure its after
-    stackParse.back().arraycur++;
-    stackParse.push_back(stage);
-}
-
-void SnapshotPayloadParseState::pushValue(const char *rgch, long long cch) {
-    if (stackParse.empty())
-        throw "Bad Protocol: unexpected trailing bulk string";
-
-    if (stackParse.back().arraycur >= static_cast<int>(stackParse.back().arrvalues.size()))
-        throw "Bad protocol: Unexpected value";
-
-    auto &stage = stackParse.back();
-    stage.arrvalues[stackParse.back().arraycur].first = (char*)m_spallocator->allocate(cch);
-    stage.arrvalues[stackParse.back().arraycur].second = cch;
-    memcpy(stage.arrvalues[stackParse.back().arraycur].first, rgch, cch);
-    stage.arraycur++;
-    switch (stage.name) {
-        case ParseStageName::KeyValuePair:
-            if (stackParse.size() < 2)
-                throw "Bad Protocol: unexpected bulk string";
-            if (stackParse[stackParse.size()-2].name == ParseStageName::MetaData) {
-                if (stage.arraycur == 2) {
-                    // We loaded both pairs
-                    if (stage.arrvalues[0].first == nullptr || stage.arrvalues[1].first == nullptr)
-                        throw "Bad Protocol: Got array when expecing a string"; // A baddy could make us derefence the vector when its too small
-
-                    if (!strcasecmp(stage.arrvalues[0].first, "lua")) {
-                        /* Load the script back in memory. */
-                        robj *auxval = createStringObject(stage.arrvalues[1].first, stage.arrvalues[1].second);
-                        if (luaCreateFunction(NULL,g_pserver->lua,auxval) == NULL) {
-                            throw "Can't load Lua script";
-                        }
-                    } else {
-                        dictAdd(dictMetaData, sdsnewlen(stage.arrvalues[0].first, stage.arrvalues[0].second), sdsnewlen(stage.arrvalues[1].first, stage.arrvalues[1].second));
-                    }
-                }
-            } else if (stackParse[stackParse.size()-2].name == ParseStageName::Dataset) {
-                if (stage.arraycur == 2) {
-                    // We loaded both pairs
-                    if (stage.arrvalues[0].first == nullptr || stage.arrvalues[1].first == nullptr)
-                        throw "Bad Protocol: Got array when expecing a string"; // A baddy could make us derefence the vector when its too small
-                    vecqueuedKeys.push_back(stage.arrvalues[0].first);
-                    vecqueuedKeysCb.push_back(stage.arrvalues[0].second);
-                    vecqueuedVals.push_back(stage.arrvalues[1].first);
-                    vecqueuedValsCb.push_back(stage.arrvalues[1].second);
-                    stage.arrvalues[0].first = nullptr;
-                    stage.arrvalues[1].first = nullptr;
-                    cbQueued += vecqueuedKeysCb.back();
-                    cbQueued += vecqueuedValsCb.back();
-                    if (cbQueued >= queuedBatchLimit)
-                        flushQueuedKeys();
-                }
-            } else {
-                throw "Bad Protocol: unexpected bulk string";
-            }
-            break;
-
-        default:
-            throw "Bad Protocol: unexpected bulk string out of KV pair";
-    }
-}
-
-void SnapshotPayloadParseState::pushValue(long long value) {
-    if (stackParse.empty())
-        throw "Bad Protocol: unexpected integer value";
-    
-    stackParse.back().arraycur++;
-
-    if (stackParse.back().arraycur != 2 || stackParse.back().arrvalues[0].first == nullptr)
-        throw "Bad Protocol: unexpected integer value";
-
-    dictEntry *de = dictAddRaw(dictLongLongMetaData, sdsnewlen(stackParse.back().arrvalues[0].first, stackParse.back().arrvalues[0].second), nullptr);
-    if (de == nullptr)
-        throw "Bad Protocol: metadata field sent twice";
-    de->v.s64 = value;
-}
-
-long long SnapshotPayloadParseState::getMetaDataLongLong(const char *szField) const {
-    dictEntry *de = dictFind(dictLongLongMetaData, szField);
-
-    if (de == nullptr) {
-        serverLog(LL_WARNING, "Master did not send field: %s", szField);
-        throw false;
-    }
-    return de->v.s64;
-}
-
-sds SnapshotPayloadParseState::getMetaDataStr(const char *szField) const {
-    sdsstring str(szField, strlen(szField));
-
-    dictEntry *de = dictFind(dictMetaData, str.get());
-    if (de == nullptr) {
-        serverLog(LL_WARNING, "Master did not send field: %s", szField);
-        throw false;
-    }
-    return (sds)de->v.val;
-}

src/SnapshotPayloadParseState.h

@@ -1,66 +0,0 @@
-#pragma once
-#include <stack>
-#include <vector>
-
-class SlabAllocator;
-
-class SnapshotPayloadParseState {
-    enum class ParseStageName {
-        None,
-        Global,
-        MetaData,
-        Databases,
-        Dataset,
-        KeyValuePair,
-    };
-    struct ParseStage {
-        ParseStageName name;
-        long long arraylen = 0;
-        long long arraycur = 0;
-        std::array<std::pair<char*, size_t>, 2> arrvalues;
-
-        ParseStage() {
-            for (auto &pair : arrvalues) {
-                pair.first = nullptr;
-                pair.second = 0;
-            }
-        }
-    };
-
-    std::vector<ParseStage> stackParse;
-    
-    
-    std::vector<char*> vecqueuedKeys;
-    std::vector<size_t> vecqueuedKeysCb;
-    std::vector<char*> vecqueuedVals;
-    std::vector<size_t> vecqueuedValsCb;
-    
-    
-    std::atomic<int> insertsInFlight;
-    std::unique_ptr<SlabAllocator> m_spallocator;
-    dict *dictLongLongMetaData = nullptr;
-    dict *dictMetaData = nullptr;
-    size_t cbQueued = 0;
-    static const size_t queuedBatchLimit = 64*1024*1024;    // 64 MB
-    int current_database = -1;
-
-    ParseStageName getNextStage();
-    void flushQueuedKeys();
-
-public:
-    SnapshotPayloadParseState();
-    ~SnapshotPayloadParseState();
-
-    long long getMetaDataLongLong(const char *field) const;
-    sds getMetaDataStr(const char *szField) const;
-
-    static const char *getStateDebugName(ParseStage stage);
-
-    size_t depth() const;
-
-    void trimState();
-    void pushArray(long long size);
-    void pushValue(const char *rgch, long long cch);
-    void pushValue(long long value);
-    bool shouldThrottle() const { return insertsInFlight > (cserver.cthreads*4); }
-};

src/StorageCache.cpp

@@ -1,216 +0,0 @@
-#include "server.h"
-
-uint64_t hashPassthrough(const void *hash) {
-    return static_cast<uint64_t>(reinterpret_cast<uintptr_t>(hash));
-}
-
-int hashCompare(void *, const void *key1, const void *key2) {
-    auto diff = (reinterpret_cast<uintptr_t>(key1) - reinterpret_cast<uintptr_t>(key2));
-    return !diff;
-}
-
-dictType dbStorageCacheType = {
-    hashPassthrough,            /* hash function */
-    NULL,                       /* key dup */
-    NULL,                       /* val dup */
-    hashCompare,                /* key compare */
-    NULL,                       /* key destructor */
-    NULL                        /* val destructor */
-};
-
-StorageCache::StorageCache(IStorage *storage, bool fCache)
-        : m_spstorage(storage)
-{
-    if (fCache)
-        m_pdict = dictCreate(&dbStorageCacheType, nullptr);
-}
-
-StorageCache::~StorageCache()
-{
-    if (m_pdict != nullptr)
-        dictRelease(m_pdict);
-}
-
-void StorageCache::clear(void(callback)(void*))
-{
-    std::unique_lock<fastlock> ul(m_lock);
-    if (m_pdict != nullptr)
-        dictEmpty(m_pdict, callback);
-    m_spstorage->clear();
-    m_collisionCount = 0;
-}
-
-void StorageCache::clearAsync()
-{
-    std::unique_lock<fastlock> ul(m_lock);
-    if (count() == 0)
-        return;
-    if (m_pdict != nullptr) {
-        dict *dSav = m_pdict;
-        m_pdict = dictCreate(&dbStorageCacheType, nullptr);
-        g_pserver->asyncworkqueue->AddWorkFunction([dSav]{
-            dictEmpty(dSav, nullptr);
-        });
-    }
-    m_spstorage->clear();
-    m_collisionCount = 0;
-}
-
-void StorageCache::cacheKey(sds key)
-{
-    if (m_pdict == nullptr)
-        return;
-    uintptr_t hash = dictSdsHash(key);
-    if (dictAdd(m_pdict, reinterpret_cast<void*>(hash), (void*)1) != DICT_OK) {
-        dictEntry *de = dictFind(m_pdict, reinterpret_cast<void*>(hash));
-        serverAssert(de != nullptr);
-        de->v.s64++;
-        m_collisionCount++;
-    }
-}
-
-void StorageCache::cacheKey(const char *rgch, size_t cch)
-{
-    if (m_pdict == nullptr)
-        return;
-    uintptr_t hash = dictGenHashFunction(rgch, (int)cch);
-    if (dictAdd(m_pdict, reinterpret_cast<void*>(hash), (void*)1) != DICT_OK) {
-        dictEntry *de = dictFind(m_pdict, reinterpret_cast<void*>(hash));
-        serverAssert(de != nullptr);
-        de->v.s64++;
-        m_collisionCount++;
-    }
-}
-
-bool StorageCache::erase(sds key)
-{
-    bool result = m_spstorage->erase(key, sdslen(key));
-    std::unique_lock<fastlock> ul(m_lock);
-    if (result && m_pdict != nullptr)
-    {
-        uint64_t hash = dictSdsHash(key);
-        dictEntry *de = dictFind(m_pdict, reinterpret_cast<void*>(hash));
-        serverAssert(de != nullptr);
-        de->v.s64--;
-        serverAssert(de->v.s64 >= 0);
-        if (de->v.s64 == 0) {
-            dictDelete(m_pdict, reinterpret_cast<void*>(hash));
-        } else {
-            m_collisionCount--;
-        }
-    }
-    return result;
-}
-
-void StorageCache::insert(sds key, const void *data, size_t cbdata, bool fOverwrite)
-{
-    std::unique_lock<fastlock> ul(m_lock);
-    if (!fOverwrite && m_pdict != nullptr)
-    {
-        cacheKey(key);
-    }
-    ul.unlock();
-    m_spstorage->insert(key, sdslen(key), (void*)data, cbdata, fOverwrite);
-}
-
-long _dictKeyIndex(dict *d, const void *key, uint64_t hash, dictEntry **existing);
-void StorageCache::bulkInsert(char **rgkeys, size_t *rgcbkeys, char **rgvals, size_t *rgcbvals, size_t celem)
-{
-    std::vector<dictEntry*> vechashes;
-    if (m_pdict != nullptr) {
-        vechashes.reserve(celem);
-
-        for (size_t ielem = 0; ielem < celem; ++ielem) {
-            dictEntry *de = (dictEntry*)zmalloc(sizeof(dictEntry));
-            de->key = (void*)dictGenHashFunction(rgkeys[ielem], (int)rgcbkeys[ielem]);
-            de->v.u64 = 1;
-            vechashes.push_back(de);
-        }
-    }
-
-    std::unique_lock<fastlock> ul(m_lock);
-    bulkInsertsInProgress++;
-    if (m_pdict != nullptr) {
-        for (dictEntry *de : vechashes) {
-            if (dictIsRehashing(m_pdict)) dictRehash(m_pdict,1);
-            /* Get the index of the new element, or -1 if
-                * the element already exists. */
-            long index;
-            if ((index = _dictKeyIndex(m_pdict, de->key, (uint64_t)de->key, nullptr)) == -1) {
-                dictEntry *deLocal = dictFind(m_pdict, de->key);
-                serverAssert(deLocal != nullptr);
-                deLocal->v.s64++;
-                m_collisionCount++;
-                zfree(de);
-            } else {
-                int iht = dictIsRehashing(m_pdict) ? 1 : 0;
-                de->next = m_pdict->ht[iht].table[index];
-                m_pdict->ht[iht].table[index] = de;
-                m_pdict->ht[iht].used++;
-            }
-        }
-    }
-    ul.unlock();
-
-    m_spstorage->bulkInsert(rgkeys, rgcbkeys, rgvals, rgcbvals, celem);
-
-    bulkInsertsInProgress--;
-}
-
-const StorageCache *StorageCache::clone()
-{
-    std::unique_lock<fastlock> ul(m_lock);
-    // Clones never clone the cache
-    StorageCache *cacheNew = new StorageCache(const_cast<IStorage*>(m_spstorage->clone()), false /*fCache*/);
-    return cacheNew;
-}
-
-void StorageCache::expand(uint64_t slots)
-{
-    std::unique_lock<fastlock> ul(m_lock);
-    if (m_pdict) {
-        dictExpand(m_pdict, slots);
-    }
-}
-
-void StorageCache::retrieve(sds key, IStorage::callbackSingle fn) const
-{
-    std::unique_lock<fastlock> ul(m_lock);
-    if (m_pdict != nullptr)
-    {
-        uint64_t hash = dictSdsHash(key);
-        dictEntry *de = dictFind(m_pdict, reinterpret_cast<void*>(hash));
-        
-        if (de == nullptr)
-            return; // Not found
-    }
-    ul.unlock();
-    m_spstorage->retrieve(key, sdslen(key), fn);
-}
-
-size_t StorageCache::count() const
-{
-    std::unique_lock<fastlock> ul(m_lock, std::defer_lock);
-    bool fLocked = ul.try_lock();
-    size_t count = m_spstorage->count();
-    if (m_pdict != nullptr && fLocked) {
-        serverAssert(bulkInsertsInProgress.load(std::memory_order_seq_cst) || count == (dictSize(m_pdict) + m_collisionCount));
-    }
-    return count;
-}
-
-void StorageCache::beginWriteBatch() { 
-    serverAssert(GlobalLocksAcquired());    // Otherwise we deadlock
-    m_spstorage->beginWriteBatch(); 
-}
-
-void StorageCache::emergencyFreeCache() {
-    std::unique_lock<fastlock> ul(m_lock);
-    dict *d = m_pdict;
-    m_pdict = nullptr;
-    if (d != nullptr) {
-        g_pserver->asyncworkqueue->AddWorkFunction([d]{
-            dictRelease(d);
-        });
-    }
-}

src/StorageCache.h

@@ -1,61 +0,0 @@
-#pragma once
-#include "sds.h"
-
-class StorageCache
-{
-    std::shared_ptr<IStorage> m_spstorage;
-    dict *m_pdict = nullptr;
-    int m_collisionCount = 0;
-    mutable fastlock m_lock {"StorageCache"};
-    std::atomic<int> bulkInsertsInProgress;
-
-    StorageCache(IStorage *storage, bool fNoCache);
-
-    void cacheKey(sds key);
-    void cacheKey(const char *rgchKey, size_t cchKey);
-
-    struct load_iter_data
-    {
-        StorageCache *cache;
-        IStorageFactory::key_load_iterator itrParent;
-        void *privdataParent;
-    };
-    static void key_load_itr(const char *rgchKey, size_t cchKey, void *privdata)
-    {
-        load_iter_data *data = (load_iter_data*)privdata;
-        data->cache->cacheKey(rgchKey, cchKey);
-        if (data->itrParent)
-            data->itrParent(rgchKey, cchKey, data->privdataParent);
-    }
-
-public:
-    ~StorageCache();
-
-    static StorageCache *create(IStorageFactory *pfactory, int db, IStorageFactory::key_load_iterator fn, void *privdata) {
-        StorageCache *cache = new StorageCache(nullptr, pfactory->FSlow() /*fCache*/);
-        load_iter_data data = {cache, fn, privdata};
-        cache->m_spstorage = std::shared_ptr<IStorage>(pfactory->create(db, key_load_itr, (void*)&data));
-        return cache;
-    }
-
-    void clear(void(callback)(void*));
-    void clearAsync();
-    void insert(sds key, const void *data, size_t cbdata, bool fOverwrite);
-    void bulkInsert(char **rgkeys, size_t *rgcbkeys, char **rgvals, size_t *rgcbvals, size_t celem);
-    void retrieve(sds key, IStorage::callbackSingle fn) const;
-    bool erase(sds key);
-    void emergencyFreeCache();
-    bool keycacheIsEnabled() const { return m_pdict != nullptr; }
-    void expand(uint64_t slots);
-
-    bool enumerate(IStorage::callback fn) const { return m_spstorage->enumerate(fn); }
-
-    void beginWriteBatch();
-    void endWriteBatch() { m_spstorage->endWriteBatch(); }
-    void batch_lock() { return m_spstorage->batch_lock(); }
-    void batch_unlock() { return m_spstorage->batch_unlock(); }
-
-    size_t count() const;
-
-    const StorageCache *clone();
-};

src/ae.cpp

@@ -53,7 +53,6 @@
 #include "zmalloc.h"
 #include "config.h"
 #include "serverassert.h"
-#include "readwritelock.h"
 
 #ifdef USE_MUTEX
 thread_local int cOwnLock = 0;
@@ -88,7 +87,6 @@ mutex_wrapper g_lock;
 #else
 fastlock g_lock("AE (global)");
 #endif
-readWriteLock g_forkLock("Fork (global)");
 thread_local aeEventLoop *g_eventLoopThisThread = NULL;
 
 /* Include the best multiplexing layer supported by this system.
@@ -128,13 +126,9 @@ struct aeCommand
     };
     void *clientData;
 };
-#ifdef PIPE_BUF
-static_assert(sizeof(aeCommand) <= PIPE_BUF, "aeCommand must be small enough to send atomically through a pipe");
-#endif
 
 void aeProcessCmd(aeEventLoop *eventLoop, int fd, void *, int )
 {
-    std::unique_lock<decltype(g_lock)> ulock(g_lock, std::defer_lock);
     aeCommand cmd;
     for (;;)
     {
@@ -156,22 +150,18 @@ void aeProcessCmd(aeEventLoop *eventLoop, int fd, void *, int )
 
         case AE_ASYNC_OP::PostFunction:
             {
-            if (cmd.fLock && !ulock.owns_lock()) {
-                g_forkLock.releaseRead();
+            std::unique_lock<decltype(g_lock)> ulock(g_lock, std::defer_lock);
+            if (cmd.fLock)
                 ulock.lock();
-                g_forkLock.acquireRead();
-            }
             ((aePostFunctionProc*)cmd.proc)(cmd.clientData);
             break;
             }
 
         case AE_ASYNC_OP::PostCppFunction:
         {
-            if (cmd.fLock && !ulock.owns_lock()) {
-                g_forkLock.releaseRead();
+            std::unique_lock<decltype(g_lock)> ulock(g_lock, std::defer_lock);
+            if (cmd.fLock)
                 ulock.lock();
-                g_forkLock.acquireRead();
-            }
             (*cmd.pfn)();
 
             delete cmd.pfn;
@@ -266,7 +256,7 @@ int aePostFunction(aeEventLoop *eventLoop, std::function<void()> fn, bool fLock,
 
     aeCommand cmd = {};
     cmd.op = AE_ASYNC_OP::PostCppFunction;
-    cmd.pfn = new std::function<void()>(fn);
+    cmd.pfn = new (MALLOC_LOCAL) std::function<void()>(fn);
     cmd.fLock = fLock;
 
     auto size = write(eventLoop->fdCmdWrite, &cmd, sizeof(cmd));
@@ -555,11 +545,7 @@ static int processTimeEvents(aeEventLoop *eventLoop) {
             if (te->next)
                 te->next->prev = te->prev;
             if (te->finalizerProc) {
-                if (!ulock.owns_lock()) {
-                    g_forkLock.releaseRead();
-                    ulock.lock();
-                    g_forkLock.acquireRead();
-                }
+                if (!ulock.owns_lock()) ulock.lock();
                 te->finalizerProc(eventLoop, te->clientData);
                 now = getMonotonicUs();
             }
@@ -579,11 +565,7 @@ static int processTimeEvents(aeEventLoop *eventLoop) {
         }
 
         if (te->when <= now) {
-            if (!ulock.owns_lock()) {
-                g_forkLock.releaseRead();
-                ulock.lock();
-                g_forkLock.acquireRead();
-            }
+            if (!ulock.owns_lock()) ulock.lock();
             int retval;
 
             id = te->id;
@@ -607,11 +589,8 @@ extern "C" void ProcessEventCore(aeEventLoop *eventLoop, aeFileEvent *fe, int ma
 {
 #define LOCK_IF_NECESSARY(fe, tsmask) \
     std::unique_lock<decltype(g_lock)> ulock(g_lock, std::defer_lock); \
-    if (!(fe->mask & tsmask)) { \
-        g_forkLock.releaseRead(); \
-        ulock.lock(); \
-        g_forkLock.acquireRead(); \
-    }
+    if (!(fe->mask & tsmask)) \
+        ulock.lock()
 
     int fired = 0; /* Number of events fired for current fd. */
 
@@ -723,11 +702,8 @@ int aeProcessEvents(aeEventLoop *eventLoop, int flags)
 
         if (eventLoop->beforesleep != NULL && flags & AE_CALL_BEFORE_SLEEP) {
             std::unique_lock<decltype(g_lock)> ulock(g_lock, std::defer_lock);
-            if (!(eventLoop->beforesleepFlags & AE_SLEEP_THREADSAFE)) {
-                g_forkLock.releaseRead();
+            if (!(eventLoop->beforesleepFlags & AE_SLEEP_THREADSAFE))
                 ulock.lock();
-                g_forkLock.acquireRead();
-            }
             eventLoop->beforesleep(eventLoop);
         }
 
@@ -738,11 +714,8 @@ int aeProcessEvents(aeEventLoop *eventLoop, int flags)
         /* After sleep callback. */
         if (eventLoop->aftersleep != NULL && flags & AE_CALL_AFTER_SLEEP) {
             std::unique_lock<decltype(g_lock)> ulock(g_lock, std::defer_lock);
-            if (!(eventLoop->aftersleepFlags & AE_SLEEP_THREADSAFE)) {
-                g_forkLock.releaseRead();
+            if (!(eventLoop->aftersleepFlags & AE_SLEEP_THREADSAFE))
                 ulock.lock();
-                g_forkLock.acquireRead();
-            }
             eventLoop->aftersleep(eventLoop);
         }
 
@@ -811,27 +784,14 @@ void aeSetAfterSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *aftersleep,
 }
 
 thread_local spin_worker tl_worker = nullptr;
-thread_local bool fOwnLockOverride = false;
 void setAeLockSetThreadSpinWorker(spin_worker worker)
 {
     tl_worker = worker;
 }
 
-void aeThreadOnline()
-{
-    g_forkLock.acquireRead();
-}
-
 void aeAcquireLock()
 {
-    g_forkLock.releaseRead();
     g_lock.lock(tl_worker);
-    g_forkLock.acquireRead();
-}
-
-void aeAcquireForkLock()
-{
-    g_forkLock.upgradeWrite();
 }
 
 int aeTryAcquireLock(int fWeak)
@@ -839,47 +799,14 @@ int aeTryAcquireLock(int fWeak)
     return g_lock.try_lock(!!fWeak);
 }
 
-void aeThreadOffline()
-{
-    g_forkLock.releaseRead();
-}
-
 void aeReleaseLock()
 {
     g_lock.unlock();
 }
 
-void aeSetThreadOwnsLockOverride(int fOverride)
-{
-    fOwnLockOverride = fOverride;
-}
-
-void aeReleaseForkLock()
-{
-    g_forkLock.downgradeWrite();
-}
-
 int aeThreadOwnsLock()
 {
-    return fOwnLockOverride || g_lock.fOwnLock();
-}
-
-int aeLockContested(int threshold)
-{
-    ticket ticketT;
-    __atomic_load(&g_lock.m_ticket.u, &ticketT.u, __ATOMIC_RELAXED);
-    return ticketT.m_active < static_cast<uint16_t>(ticketT.m_avail - threshold);
-}
-
-int aeLockContention()
-{
-    ticket ticketT;
-    __atomic_load(&g_lock.m_ticket.u, &ticketT.u, __ATOMIC_RELAXED);
-    int32_t avail = ticketT.m_avail;
-    int32_t active = ticketT.m_active;
-    if (avail < active)
-        avail += 0x10000;
-    return avail - active;
+    return g_lock.fOwnLock();
 }
 
 void aeClosePipesForForkChild(aeEventLoop *el)
@@ -888,4 +815,4 @@ void aeClosePipesForForkChild(aeEventLoop *el)
     el->fdCmdRead = -1;
     close(el->fdCmdWrite);
     el->fdCmdWrite = -1;
-}
+}

src/ae.h

@@ -164,17 +164,10 @@ void aeSetDontWait(aeEventLoop *eventLoop, int noWait);
 void aeClosePipesForForkChild(aeEventLoop *eventLoop);
 
 void setAeLockSetThreadSpinWorker(spin_worker worker);
-void aeThreadOnline();
 void aeAcquireLock();
-void aeAcquireForkLock();
 int aeTryAcquireLock(int fWeak);
-void aeThreadOffline();
 void aeReleaseLock();
-void aeReleaseForkLock();
 int aeThreadOwnsLock();
-void aeSetThreadOwnsLockOverride(int fOverride);
-int aeLockContested(int threshold);
-int aeLockContention(); // returns the number of instantaneous threads waiting on the lock
 
 #ifdef __cplusplus
 }

src/aelocker.h

@@ -21,6 +21,7 @@ public:
 
         if (c != nullptr)
         {
+            serverAssert(!m_fArmed);
             serverAssert(c->lock.fOwnLock());
 
             if (!aeTryAcquireLock(true /*fWeak*/))    // avoid locking the client if we can

src/aof.cpp

@@ -765,7 +765,7 @@ void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int a
 /* In Redis commands are always executed in the context of a client, so in
  * order to load the append only file we need to create a fake client. */
 struct client *createAOFClient(void) {
-    struct client *c = new client();
+    struct client *c =(client*) zmalloc(sizeof(*c), MALLOC_LOCAL);
 
     selectDb(c,0);
     c->id = CLIENT_ID_AOF; /* So modules can identify it's the AOF client. */
@@ -778,6 +778,7 @@ struct client *createAOFClient(void) {
     c->argv = NULL;
     c->original_argc = 0;
     c->original_argv = NULL;
+    c->argv_len_sum = 0;
     c->bufpos = 0;
     c->fPendingAsyncWrite = FALSE;
     c->fPendingAsyncWriteHandler = FALSE;
@@ -805,7 +806,6 @@ struct client *createAOFClient(void) {
     c->sockname = NULL;
     c->resp = 2;
     c->user = NULL;
-    c->mvccCheckpoint = 0;
     listSetFreeMethod(c->reply,freeClientReplyValue);
     listSetDupMethod(c->reply,dupClientReplyValue);
     fastlock_init(&c->lock, "fake client");
@@ -820,7 +820,7 @@ void freeFakeClientArgv(struct client *c) {
     for (j = 0; j < c->argc; j++)
         decrRefCount(c->argv[j]);
     zfree(c->argv);
-    c->argv_len_sumActive = 0;
+    c->argv_len_sum = 0;
 }
 
 void freeFakeClient(struct client *c) {
@@ -831,7 +831,7 @@ void freeFakeClient(struct client *c) {
     freeClientOriginalArgv(c);
     fastlock_unlock(&c->lock);
     fastlock_free(&c->lock);
-    delete c;
+    zfree(c);
 }
 
 /* Replay the append log file. On success C_OK is returned. On non fatal
@@ -870,11 +870,6 @@ int loadAppendOnlyFile(char *filename) {
     fakeClient = createAOFClient();
     startLoadingFile(fp, filename, RDBFLAGS_AOF_PREAMBLE);
 
-    for (int idb = 0; idb < cserver.dbnum; ++idb)
-    {
-        g_pserver->db[idb]->trackChanges(true);
-    }
-
     /* Check if this AOF file has an RDB preamble. In that case we need to
      * load the RDB file and later continue loading the AOF tail. */
     char sig[5]; /* "REDIS" */
@@ -884,7 +879,7 @@ int loadAppendOnlyFile(char *filename) {
     } else {
         /* RDB preamble. Pass loading the RDB functions. */
         rio rdb;
-        rdbSaveInfo rsi;
+        rdbSaveInfo rsi = RDB_SAVE_INFO_INIT;
 
         serverLog(LL_NOTICE,"Reading RDB preamble from AOF file...");
         if (fseek(fp,0,SEEK_SET) == -1) goto readerr;
@@ -1010,11 +1005,6 @@ int loadAppendOnlyFile(char *filename) {
     }
 
 loaded_ok: /* DB loaded, cleanup and return C_OK to the caller. */
-    for (int idb = 0; idb < cserver.dbnum; ++idb)
-    {
-        if (g_pserver->db[idb]->processChanges(false))
-            g_pserver->db[idb]->commitChanges();
-    }
     fclose(fp);
     freeFakeClient(fakeClient);
     g_pserver->aof_state = old_aof_state;
@@ -1548,6 +1538,8 @@ ssize_t aofReadDiffFromParent(void) {
 }
 
 int rewriteAppendOnlyFileRio(rio *aof) {
+    dictIterator *di = NULL;
+    dictEntry *de;
     size_t processed = 0;
     int j;
     long key_count = 0;
@@ -1555,63 +1547,69 @@ int rewriteAppendOnlyFileRio(rio *aof) {
 
     for (j = 0; j < cserver.dbnum; j++) {
         char selectcmd[] = "*2\r\n$6\r\nSELECT\r\n";
-        redisDb *db = g_pserver->db[j];
-        if (db->size() == 0) continue;
+        redisDb *db = g_pserver->db+j;
+        dict *d = db->dict;
+        if (dictSize(d) == 0) continue;
+        di = dictGetSafeIterator(d);
 
         /* SELECT the new DB */
         if (rioWrite(aof,selectcmd,sizeof(selectcmd)-1) == 0) goto werr;
         if (rioWriteBulkLongLong(aof,j) == 0) goto werr;
 
         /* Iterate this DB writing every entry */
-        bool fComplete = db->iterate([&](const char *keystr, robj *o)->bool{
+        while((de = dictNext(di)) != NULL) {
+            sds keystr;
             redisObjectStack key;
-            initStaticStringObject(key,(sds)keystr);
+            robj *o = nullptr;
+
+            keystr = (sds)dictGetKey(de);
+            o = (robj*)dictGetVal(de);
+            initStaticStringObject(key,keystr);
+
+            expireEntry *pexpire = getExpire(db,&key);
 
             /* Save the key and associated value */
             if (o->type == OBJ_STRING) {
                 /* Emit a SET command */
                 char cmd[]="*3\r\n$3\r\nSET\r\n";
-                if (rioWrite(aof,cmd,sizeof(cmd)-1) == 0) return false;
+                if (rioWrite(aof,cmd,sizeof(cmd)-1) == 0) goto werr;
                 /* Key and value */
-                if (rioWriteBulkObject(aof,&key) == 0) return false;
-                if (rioWriteBulkObject(aof,o) == 0) return false;
+                if (rioWriteBulkObject(aof,&key) == 0) goto werr;
+                if (rioWriteBulkObject(aof,o) == 0) goto werr;
             } else if (o->type == OBJ_LIST) {
-                if (rewriteListObject(aof,&key,o) == 0) return false;
+                if (rewriteListObject(aof,&key,o) == 0) goto werr;
             } else if (o->type == OBJ_SET) {
-                if (rewriteSetObject(aof,&key,o) == 0) return false;
+                if (rewriteSetObject(aof,&key,o) == 0) goto werr;
             } else if (o->type == OBJ_ZSET) {
-                if (rewriteSortedSetObject(aof,&key,o) == 0) return false;
+                if (rewriteSortedSetObject(aof,&key,o) == 0) goto werr;
             } else if (o->type == OBJ_HASH) {
-                if (rewriteHashObject(aof,&key,o) == 0) return false;
+                if (rewriteHashObject(aof,&key,o) == 0) goto werr;
             } else if (o->type == OBJ_STREAM) {
-                if (rewriteStreamObject(aof,&key,o) == 0) return false;
+                if (rewriteStreamObject(aof,&key,o) == 0) goto werr;
             } else if (o->type == OBJ_MODULE) {
-                if (rewriteModuleObject(aof,&key,o) == 0) return false;
+                if (rewriteModuleObject(aof,&key,o) == 0) goto werr;
             } else {
                 serverPanic("Unknown object type");
             }
             /* Save the expire time */
-            if (o->FExpires()) {
-                std::unique_lock<fastlock> ul(g_expireLock);
-                expireEntry *pexpire = db->getExpire(&key);
+            if (pexpire != nullptr) {
                 for (auto &subExpire : *pexpire) {
                     if (subExpire.subkey() == nullptr)
                     {
                         char cmd[]="*3\r\n$9\r\nPEXPIREAT\r\n";
-                        if (rioWrite(aof,cmd,sizeof(cmd)-1) == 0) return false;
-                        if (rioWriteBulkObject(aof,&key) == 0) return false;
+                        if (rioWrite(aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                        if (rioWriteBulkObject(aof,&key) == 0) goto werr;
                     }
                     else
                     {
                         char cmd[]="*4\r\n$12\r\nPEXPIREMEMBERAT\r\n";
-                        if (rioWrite(aof,cmd,sizeof(cmd)-1) == 0) return false;
-                        if (rioWriteBulkObject(aof,&key) == 0) return false;
-                        if (rioWrite(aof,subExpire.subkey(),sdslen(subExpire.subkey())) == 0) return false;
+                        if (rioWrite(aof,cmd,sizeof(cmd)-1) == 0) goto werr;
+                        if (rioWriteBulkObject(aof,&key) == 0) goto werr;
+                        if (rioWrite(aof,subExpire.subkey(),sdslen(subExpire.subkey())) == 0) goto werr;
                     }
-                    if (rioWriteBulkLongLong(aof,subExpire.when()) == 0) return false; // common
+                    if (rioWriteBulkLongLong(aof,subExpire.when()) == 0) goto werr; // common
                 }
             }
-            
             /* Read some diff from the parent process from time to time. */
             if (aof->processed_bytes > processed+AOF_READ_DIFF_INTERVAL_BYTES) {
                 processed = aof->processed_bytes;
@@ -1628,15 +1626,14 @@ int rewriteAppendOnlyFileRio(rio *aof) {
                     updated_time = now;
                 }
             }
-
-            return true;
-        });
-        if (!fComplete)
-            goto werr;
+        }
+        dictReleaseIterator(di);
+        di = NULL;
     }
     return C_OK;
 
 werr:
+    if (di) dictReleaseIterator(di);
     return C_ERR;
 }
 
@@ -1675,12 +1672,7 @@ int rewriteAppendOnlyFile(char *filename) {
 
     if (g_pserver->aof_use_rdb_preamble) {
         int error;
-        std::vector<const redisDbPersistentDataSnapshot*> vecpdb;
-        for (int idb = 0; idb < cserver.dbnum; ++idb)
-        {
-            vecpdb.push_back(g_pserver->db[idb]);
-        }
-        if (rdbSaveRio(&aof,vecpdb.data(),&error,RDBFLAGS_AOF_PREAMBLE,NULL) == C_ERR) {
+        if (rdbSaveRio(&aof,&error,RDBFLAGS_AOF_PREAMBLE,NULL) == C_ERR) {
             errno = error;
             goto werr;
         }

src/asciilogo.h

@@ -32,7 +32,7 @@ const char *ascii_logo =
 "                  _                                                   \n"
 "               _-(+)-_                                                \n"
 "            _-- /   \\ --_                                            \n"
-"         _--   /     \\   --_            KeyDB  %s (%s/%d) %s bit     \n"
+"         _--   /     \\   --_            KeyDB %s (%s/%d) %s bit      \n"
 "     __--     /       \\     --__                                     \n"
 "    (+) _    /         \\    _ (+)       Running in %s mode\n"
 "     |   -- /           \\ --   |        Port: %d\n"

src/blocked.cpp

@@ -96,8 +96,6 @@ void blockClient(client *c, int btype) {
 
     c->flags |= CLIENT_BLOCKED;
     c->btype = btype;
-    if (btype == BLOCKED_ASYNC)
-        c->casyncOpsPending++;
     g_pserver->blocked_clients++;
     g_pserver->blocked_clients_by_type[btype]++;
     addClientToTimeoutTable(c);
@@ -154,7 +152,7 @@ void processUnblockedClients(int iel) {
             }
             /* Then process client if it has more data in it's buffer. */
             if (c->querybuf && sdslen(c->querybuf) > 0) {
-                processInputBuffer(c, true /*fParse*/, CMD_CALL_FULL);
+                processInputBuffer(c, CMD_CALL_FULL);
             }
         }
     }
@@ -201,9 +199,6 @@ void unblockClient(client *c) {
     } else if (c->btype == BLOCKED_MODULE) {
         if (moduleClientIsBlockedOnKeys(c)) unblockClientWaitingData(c);
         unblockClientFromModule(c);
-    } else if (c->btype == BLOCKED_ASYNC) {
-        serverAssert(c->casyncOpsPending > 0);
-        c->casyncOpsPending--;
     } else if (c->btype == BLOCKED_PAUSE) {
         listDelNode(g_pserver->paused_clients,c->paused_list_node);
         c->paused_list_node = NULL;

src/childinfo.cpp

@@ -69,7 +69,6 @@ void closeChildInfoPipe(void) {
 /* Send save data to parent. */
 void sendChildInfoGeneric(childInfoType info_type, size_t keys, double progress, const char *pname) {
     if (g_pserver->child_info_pipe[1] == -1) return;
-    if (g_pserver->rdbThreadVars.fRdbThreadActive && g_pserver->rdbThreadVars.fRdbThreadCancel) return;
 
     static monotime cow_updated = 0;
     static uint64_t cow_update_cost = 0;

src/cluster.cpp

@@ -771,7 +771,7 @@ unsigned long getClusterConnectionsCount(void) {
     /* We decrement the number of nodes by one, since there is the
      * "myself" node too in the list. Each node uses two file descriptors,
      * one incoming and one outgoing, thus the multiplication by 2. */
-    return g_pserver->cluster_enabled && g_pserver->cluster != nullptr ?
+    return g_pserver->cluster_enabled ?
            ((dictSize(g_pserver->cluster->nodes)-1)*2) : 0;
 }
 
@@ -785,7 +785,7 @@ unsigned long getClusterConnectionsCount(void) {
  * However if the key contains the {...} pattern, only the part between
  * { and } is hashed. This may be useful in the future to force certain
  * keys to be in the same node (assuming no resharding is in progress). */
-unsigned int keyHashSlot(const char *key, int keylen) {
+unsigned int keyHashSlot(char *key, int keylen) {
     int s, e; /* start-end indexes of { and } */
 
     for (s = 0; s < keylen; s++)
@@ -1525,9 +1525,7 @@ void clusterProcessGossipSection(clusterMsg *hdr, clusterLink *link) {
                  * it's greater than our view but is not in the future
                  * (with 500 milliseconds tolerance) from the POV of our
                  * clock. */
-                mstime_t mstime;
-                __atomic_load(&g_pserver->mstime, &mstime, __ATOMIC_RELAXED);
-                if (pongtime <= (mstime+500) &&
+                if (pongtime <= (g_pserver->mstime+500) &&
                     pongtime > node->pong_received)
                 {
                     node->pong_received = pongtime;
@@ -4129,7 +4127,7 @@ int verifyClusterConfigWithData(void) {
 
     /* Make sure we only have keys in DB0. */
     for (j = 1; j < cserver.dbnum; j++) {
-        if (g_pserver->db[j]->size()) return C_ERR;
+        if (dictSize(g_pserver->db[j].dict)) return C_ERR;
     }
 
     /* Check that all the slots we see populated memory have a corresponding
@@ -4579,7 +4577,7 @@ NULL
         clusterReplyMultiBulkSlots(c);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"flushslots") && c->argc == 2) {
         /* CLUSTER FLUSHSLOTS */
-        if (g_pserver->db[0]->size() != 0) {
+        if (dictSize(g_pserver->db[0].dict) != 0) {
             addReplyError(c,"DB must be empty to perform CLUSTER FLUSHSLOTS.");
             return;
         }
@@ -4916,7 +4914,7 @@ NULL
          * slots nor keys to accept to replicate some other node.
          * Slaves can switch to another master without issues. */
         if (nodeIsMaster(myself) &&
-            (myself->numslots != 0 || g_pserver->db[0]->size() != 0)) {
+            (myself->numslots != 0 || dictSize(g_pserver->db[0].dict) != 0)) {
             addReplyError(c,
                 "To set a master the node must be empty and "
                 "without assigned slots.");
@@ -5076,7 +5074,7 @@ NULL
 
         /* Slaves can be reset while containing data, but not master nodes
          * that must be empty. */
-        if (nodeIsMaster(myself) && c->db->size() != 0) {
+        if (nodeIsMaster(myself) && dictSize(c->db->dict) != 0) {
             addReplyError(c,"CLUSTER RESET can't be called with "
                             "master nodes containing keys");
             return;
@@ -5092,7 +5090,6 @@ NULL
 /* -----------------------------------------------------------------------------
  * DUMP, RESTORE and MIGRATE commands
  * -------------------------------------------------------------------------- */
-ssize_t rdbSaveAuxFieldStrStr(rio *rdb, const char *key, const char *val);
 
 /* Generates a DUMP-format representation of the object 'o', adding it to the
  * io stream pointed by 'rio'. This function can't fail. */
@@ -5105,10 +5102,6 @@ void createDumpPayload(rio *payload, robj_roptr o, robj *key) {
     rioInitWithBuffer(payload,sdsempty());
     serverAssert(rdbSaveObjectType(payload,o));
     serverAssert(rdbSaveObject(payload,o,key));
-    char szT[32];
-    uint64_t mvcc = mvccFromObj(o);
-    snprintf(szT, 32, "%" PRIu64, mvcc);
-    serverAssert(rdbSaveAuxFieldStrStr(payload,"mvcc-tstamp", szT) != -1);
 
     /* Write the footer, this is how it looks like:
      * ----------------+---------------------+---------------+
@@ -5292,21 +5285,6 @@ void restoreCommand(client *c) {
         addReplyError(c,"Bad data format");
         return;
     }
-    if (rdbLoadType(&payload) == RDB_OPCODE_AUX)
-    {
-        robj *auxkey, *auxval;
-        if ((auxkey = rdbLoadStringObject(&payload)) == NULL) goto eoferr;
-        if ((auxval = rdbLoadStringObject(&payload)) == NULL) {
-            decrRefCount(auxkey);
-            goto eoferr;
-        }
-        if (strcasecmp(szFromObj(auxkey), "mvcc-tstamp") == 0) {
-            setMvccTstamp(obj, strtoull(szFromObj(auxval), nullptr, 10));
-        }
-        decrRefCount(auxkey);
-        decrRefCount(auxval);
-    }
-eoferr:
 
     /* Remove the old key if needed. */
     int deleted = 0;
@@ -5584,8 +5562,7 @@ try_again:
     /* Create RESTORE payload and generate the protocol to call the command. */
     for (j = 0; j < num_keys; j++) {
         long long ttl = 0;
-        std::unique_lock<fastlock> ul(g_expireLock);
-        expireEntry *pexpire = c->db->getExpire(kv[j]);
+        expireEntry *pexpire = getExpire(c->db,kv[j]);
         long long expireat = INVALID_EXPIRE;
         if (pexpire != nullptr)
             pexpire->FGetPrimaryExpire(&expireat);
@@ -5876,7 +5853,7 @@ clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, in
     multiState *ms, _ms;
     multiCmd mc;
     int i, slot = 0, migrating_slot = 0, importing_slot = 0, missing_keys = 0;
-    serverAssert((c->cmd->flags & CMD_ASYNC_OK) || GlobalLocksAcquired());
+    serverAssert(GlobalLocksAcquired());
 
     /* Allow any key to be set if a module disabled cluster redirections. */
     if (g_pserver->cluster_module_flags & CLUSTER_MODULE_FLAG_NO_REDIRECTION)
@@ -5982,7 +5959,7 @@ clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, in
 
             /* Migrating / Importing slot? Count keys we don't have. */
             if ((migrating_slot || importing_slot) &&
-                lookupKeyRead(g_pserver->db[0],thiskey) == nullptr)
+                lookupKeyRead(&g_pserver->db[0],thiskey) == nullptr)
             {
                 missing_keys++;
             }

src/compactvector.h

@@ -30,16 +30,7 @@ public:
         zfree(m_data);
     }
 
-    compactvector(const compactvector &src)
-    {
-        m_celem = src.m_celem;
-        m_max = src.m_max;
-        m_data = (T*)zmalloc(sizeof(T) * m_max, MALLOC_LOCAL);
-        for (size_t ielem = 0; ielem < m_celem; ++ielem)
-        {
-            new (m_data+ielem) T(src[ielem]);
-        }
-    }
+    compactvector(compactvector &) noexcept = delete;
 
     compactvector(compactvector &&src) noexcept
     {
@@ -51,6 +42,7 @@ public:
         src.m_max = 0;
     }
 
+    compactvector &operator=(const compactvector&) noexcept = delete;
     compactvector &operator=(compactvector &&src) noexcept
     {
         zfree(m_data);
@@ -69,7 +61,7 @@ public:
     inline T* end() { return m_data + m_celem; }
     inline const T* end() const { return m_data + m_celem; }
 
-    T* insert(T* where, const T &val)
+    T* insert(T* where, T &val)
     {
         assert(where >= m_data);
         size_t idx = where - m_data;

src/config.cpp

@@ -29,15 +29,10 @@
  */
 
 #include "server.h"
-#include "storage/rocksdbfactory.h"
-#include "storage/teststorageprovider.h"
 #include "cluster.h"
 
 #include <fcntl.h>
 #include <sys/stat.h>
-#ifdef __linux__
-#include <sys/sysinfo.h>
-#endif
 
 const char *KEYDB_SET_VERSION = KEYDB_REAL_VERSION;
 
@@ -105,11 +100,6 @@ configEnum repl_diskless_load_enum[] = {
     {NULL, 0}
 };
 
-configEnum storage_memory_model_enum[] = {
-    {"writeback", STORAGE_WRITEBACK},
-    {"writethrough", STORAGE_WRITETHROUGH},
-};
-
 configEnum tls_auth_clients_enum[] = {
     {"no", TLS_CLIENT_AUTH_NO},
     {"yes", TLS_CLIENT_AUTH_YES},
@@ -343,72 +333,6 @@ void queueLoadModule(sds path, sds *argv, int argc) {
     listAddNodeTail(g_pserver->loadmodule_queue,loadmod);
 }
 
-sds g_sdsProvider = nullptr;
-sds g_sdsArgs = nullptr;
-
-bool initializeStorageProvider(const char **err)
-{
-    try
-    {
-        bool fTest = false;
-        if (g_sdsProvider == nullptr)
-            return true;
-        if (!strcasecmp(g_sdsProvider, "flash") && g_sdsArgs != nullptr)
-        {
-#ifdef ENABLE_ROCKSDB
-            // Create The Storage Factory (if necessary)
-            serverLog(LL_NOTICE, "Initializing FLASH storage provider (this may take a long time)");
-            adjustOpenFilesLimit();
-            g_pserver->m_pstorageFactory = CreateRocksDBStorageFactory(g_sdsArgs, cserver.dbnum, cserver.storage_conf, cserver.storage_conf ? strlen(cserver.storage_conf) : 0);
-#else
-            serverLog(LL_WARNING, "To use the flash storage provider please compile KeyDB with ENABLE_FLASH=yes");
-            serverLog(LL_WARNING, "Exiting due to the use of an unsupported storage provider");
-            exit(EXIT_FAILURE);
-#endif
-	    }
-        else if (!strcasecmp(g_sdsProvider, "test") && g_sdsArgs == nullptr)
-        {
-            g_pserver->m_pstorageFactory = new (MALLOC_LOCAL) TestStorageFactory();
-            fTest = true;
-        }
-
-        if (g_pserver->m_pstorageFactory != nullptr && !fTest)
-        {
-            // We need to set max memory to a sane default so keys are actually evicted properly
-            if (g_pserver->maxmemory == 0 && g_pserver->maxmemory_policy == MAXMEMORY_NO_EVICTION)
-            {
-#ifdef __linux__
-                struct sysinfo sys;
-                if (sysinfo(&sys) == 0)
-                {
-                    // By default it's a little under half the memory.  This gives sufficient room for background saving
-                    g_pserver->maxmemory = static_cast<unsigned long long>(sys.totalram / 2.2);
-                    g_pserver->maxmemory_policy = MAXMEMORY_ALLKEYS_LRU;
-                }
-#else
-                serverLog(LL_WARNING, "Unable to dynamically set maxmemory, please set maxmemory and maxmemory-policy if you are using a storage provier");
-#endif
-            }
-            else if (g_pserver->maxmemory_policy == MAXMEMORY_NO_EVICTION)
-            {
-                g_pserver->maxmemory_policy = MAXMEMORY_ALLKEYS_LRU;
-            }
-        }
-        else
-        {
-            *err = "Unknown storage provider";
-        }
-        return g_pserver->m_pstorageFactory != nullptr;
-    }
-    catch(std::string str)
-    {
-        serverLog(LL_WARNING, "ERROR: Failed to initialize %s storage provider. Details to follow below.", g_sdsProvider);
-        serverLog(LL_WARNING, "\t%s", str.c_str());
-        serverLog(LL_WARNING, "KeyDB cannot start. Exiting.");
-        exit(EXIT_FAILURE);
-    }
-}
-
 /* Parse an array of CONFIG_OOM_COUNT sds strings, validate and populate
  * g_pserver->oom_score_adj_values if valid.
  */
@@ -747,15 +671,6 @@ void loadServerConfigFromString(char *config) {
                 g_pserver->fActiveReplica = CONFIG_DEFAULT_ACTIVE_REPLICA;
                 err = "argument must be 'yes' or 'no'"; goto loaderr;
             }
-        } else if (!strcasecmp(argv[0], "tls-allowlist")) {
-            if (argc < 2) {
-                err = "must supply at least one element in the allow list"; goto loaderr;
-            }
-            if (!g_pserver->tls_allowlist.empty()) {
-                err = "tls-allowlist may only be set once"; goto loaderr;
-            }
-            for (int i = 1; i < argc; i++)
-                g_pserver->tls_allowlist.emplace(argv[i], strlen(argv[i]));
         } else if (!strcasecmp(argv[0], "version-override") && argc == 2) {
             KEYDB_SET_VERSION = zstrdup(argv[1]);
             serverLog(LL_WARNING, "Warning version is overriden to: %s\n", KEYDB_SET_VERSION);
@@ -763,16 +678,9 @@ void loadServerConfigFromString(char *config) {
             g_fTestMode = yesnotoi(argv[1]);
         } else if (!strcasecmp(argv[0],"rdbfuzz-mode")) {
             // NOP, handled in main
-        } else if (!strcasecmp(argv[0],"storage-provider") && argc >= 2) {
-            g_sdsProvider = sdsdup(argv[1]);
-            if (argc > 2)
-                g_sdsArgs = sdsdup(argv[2]);
-	    } else if (!strcasecmp(argv[0],"is-flash-enabled") && argc == 1) {
-#ifdef ENABLE_ROCKSDB
-            exit(EXIT_SUCCESS);
-#else
-            exit(EXIT_FAILURE);
-#endif
+        } else if (!strcasecmp(argv[0],"enable-pro")) {
+            cserver.fUsePro = true;
+            break;
         } else {
             err = "Bad directive or wrong number of arguments"; goto loaderr;
         }
@@ -882,18 +790,8 @@ void configSetCommand(client *c) {
     int err;
     const char *errstr = NULL;
     serverAssertWithInfo(c,c->argv[2],sdsEncodedObject(c->argv[2]));
-
-    if (c->argc < 4 || c->argc > 4) {
-        o = nullptr;
-        // Variadic set is only supported for tls-allowlist
-        if (strcasecmp(szFromObj(c->argv[2]), "tls-allowlist")) {
-            addReplySubcommandSyntaxError(c);
-            return;
-        }
-    } else {
-        o = c->argv[3];
-        serverAssertWithInfo(c,c->argv[3],sdsEncodedObject(c->argv[3]));
-    }
+    serverAssertWithInfo(c,c->argv[3],sdsEncodedObject(c->argv[3]));
+    o = c->argv[3];
 
     /* Iterate the configs that are standard */
     for (standardConfig *config = configs; config->name != NULL; config++) {
@@ -1055,12 +953,6 @@ void configSetCommand(client *c) {
             enableWatchdog(ll);
         else
             disableWatchdog();
-    } config_set_special_field("tls-allowlist") {
-        g_pserver->tls_allowlist.clear();
-        for (int i = 3; i < c->argc; ++i) {
-            robj *val = c->argv[i];
-            g_pserver->tls_allowlist.emplace(szFromObj(val), sdslen(szFromObj(val)));
-        }
     /* Everything else is an error... */
     } config_set_else {
         addReplyErrorFormat(c,"Unsupported CONFIG parameter: %s",
@@ -1264,14 +1156,6 @@ void configGetCommand(client *c) {
         addReplyBulkCString(c, g_pserver->fActiveReplica ? "yes" : "no");
         matches++;
     }
-    if (stringmatch(pattern, "tls-allowlist", 1)) {
-        addReplyBulkCString(c,"tls-allowlist");
-        addReplyArrayLen(c, (long)g_pserver->tls_allowlist.size());
-        for (auto &elem : g_pserver->tls_allowlist) {
-            addReplyBulkCBuffer(c, elem.get(), elem.size()); // addReplyBulkSds will free which we absolutely don't want
-        }
-        matches++;
-    }
 
     setDeferredMapLen(c,replylen,matches);
 }
@@ -1941,20 +1825,6 @@ int rewriteConfig(char *path, int force_all) {
     rewriteConfigStringOption(state, "version-override",KEYDB_SET_VERSION,KEYDB_REAL_VERSION);
     rewriteConfigOOMScoreAdjValuesOption(state);
 
-    if (!g_pserver->tls_allowlist.empty()) {
-        sds conf = sdsnew("tls-allowlist ");
-        for (auto &elem : g_pserver->tls_allowlist) {
-            conf = sdscatsds(conf, (sds)elem.get());
-            conf = sdscat(conf, " ");
-        }
-        // trim the trailing space
-        sdsrange(conf, 0, -1);
-        rewriteConfigRewriteLine(state,"tls-allowlist",conf,1 /*force*/);
-        // note: conf is owned by rewriteConfigRewriteLine - no need to free
-    } else {
-        rewriteConfigMarkAsProcessed(state, "tls-allowlist"); // ensure the line is removed if it existed
-    }
-
     /* Rewrite Sentinel config if in Sentinel mode. */
     if (g_pserver->sentinel_mode) rewriteConfigSentinelOption(state);
 
@@ -2522,12 +2392,8 @@ static int updateJemallocBgThread(int val, int prev, const char **err) {
 static int updateReplBacklogSize(long long val, long long prev, const char **err) {
     /* resizeReplicationBacklog sets g_pserver->repl_backlog_size, and relies on
      * being able to tell when the size changes, so restore prev before calling it. */
-    if (cserver.repl_backlog_disk_size) {
-        *err = "Unable to dynamically resize the backlog because disk backlog is enabled";
-        return 0;
-    }
+    UNUSED(err);
     g_pserver->repl_backlog_size = prev;
-    g_pserver->repl_backlog_config_size = val;
     resizeReplicationBacklog(val);
     return 1;
 }
@@ -2540,7 +2406,7 @@ static int updateMaxmemory(long long val, long long prev, const char **err) {
         if ((unsigned long long)val < used) {
             serverLog(LL_WARNING,"WARNING: the new maxmemory value set via CONFIG SET (%llu) is smaller than the current memory usage (%zu). This will result in key eviction and/or the inability to accept new write commands depending on the maxmemory-policy.", g_pserver->maxmemory, used);
         }
-        performEvictions(false /*fPreSnapshot*/);
+        performEvictions();
     }
     return 1;
 }
@@ -2587,7 +2453,7 @@ static int updateMaxclients(long long val, long long prev, const char **err) {
         adjustOpenFilesLimit();
         if (g_pserver->maxclients != val) {
             static char msg[128];
-            snprintf(msg, 128, "The operating system is not able to handle the specified number of clients, try with %d", g_pserver->maxclients);
+            sprintf(msg, "The operating system is not able to handle the specified number of clients, try with %d", g_pserver->maxclients);
             *err = msg;
             if (g_pserver->maxclients > prev) {
                 g_pserver->maxclients = prev;
@@ -2626,7 +2492,7 @@ static int updateMaxclients(long long val, long long prev, const char **err) {
 
                 if (res != AE_OK){
                     static char msg[128];
-                    snprintf(msg, 128,"Failed to post the request to change setsize for Thread %d", iel);
+                    sprintf(msg, "Failed to post the request to change setsize for Thread %d", iel);
                     *err = msg;
                     return 0;
                 }
@@ -2777,21 +2643,13 @@ standardConfig configs[] = {
     createBoolConfig("cluster-enabled", NULL, IMMUTABLE_CONFIG, g_pserver->cluster_enabled, 0, NULL, NULL),
     createBoolConfig("appendonly", NULL, MODIFIABLE_CONFIG, g_pserver->aof_enabled, 0, NULL, updateAppendonly),
     createBoolConfig("cluster-allow-reads-when-down", NULL, MODIFIABLE_CONFIG, g_pserver->cluster_allow_reads_when_down, 0, NULL, NULL),
-    createBoolConfig("delete-on-evict", NULL, MODIFIABLE_CONFIG, cserver.delete_on_evict, 0, NULL, NULL),
-    createBoolConfig("use-fork", NULL, IMMUTABLE_CONFIG, cserver.fForkBgSave, 0, NULL, NULL),
     createBoolConfig("io-threads-do-reads", NULL, IMMUTABLE_CONFIG, fDummy, 0, NULL, NULL),
-    createBoolConfig("time-thread-priority", NULL, IMMUTABLE_CONFIG, cserver.time_thread_priority, 0, NULL, NULL),
-    createBoolConfig("prefetch-enabled", NULL, MODIFIABLE_CONFIG, g_pserver->prefetch_enabled, 1, NULL, NULL),
-    createBoolConfig("allow-rdb-resize-op", NULL, MODIFIABLE_CONFIG, g_pserver->allowRdbResizeOp, 1, NULL, NULL),
     createBoolConfig("crash-log-enabled", NULL, MODIFIABLE_CONFIG, g_pserver->crashlog_enabled, 1, NULL, updateSighandlerEnabled),
     createBoolConfig("crash-memcheck-enabled", NULL, MODIFIABLE_CONFIG, g_pserver->memcheck_enabled, 1, NULL, NULL),
     createBoolConfig("use-exit-on-panic", NULL, MODIFIABLE_CONFIG, g_pserver->use_exit_on_panic, 0, NULL, NULL),
     createBoolConfig("disable-thp", NULL, MODIFIABLE_CONFIG, g_pserver->disable_thp, 1, NULL, NULL),
     createBoolConfig("cluster-allow-replica-migration", NULL, MODIFIABLE_CONFIG, g_pserver->cluster_allow_replica_migration, 1, NULL, NULL),
     createBoolConfig("replica-announced", NULL, MODIFIABLE_CONFIG, g_pserver->replica_announced, 1, NULL, NULL),
-    createBoolConfig("enable-async-commands", NULL, MODIFIABLE_CONFIG, g_pserver->enable_async_commands, 0, NULL, NULL),
-    createBoolConfig("multithread-load-enabled", NULL, MODIFIABLE_CONFIG, g_pserver->multithread_load_enabled, 0, NULL, NULL),
-    createBoolConfig("active-client-balancing", NULL, MODIFIABLE_CONFIG, g_pserver->active_client_balancing, 1, NULL, NULL),
 
     /* String Configs */
     createStringConfig("aclfile", NULL, IMMUTABLE_CONFIG, ALLOW_EMPTY_STRING, g_pserver->acl_filename, "", NULL, NULL),
@@ -2807,7 +2665,6 @@ standardConfig configs[] = {
     createStringConfig("bio_cpulist", NULL, IMMUTABLE_CONFIG, EMPTY_STRING_IS_NULL, g_pserver->bio_cpulist, NULL, NULL, NULL),
     createStringConfig("aof_rewrite_cpulist", NULL, IMMUTABLE_CONFIG, EMPTY_STRING_IS_NULL, g_pserver->aof_rewrite_cpulist, NULL, NULL, NULL),
     createStringConfig("bgsave_cpulist", NULL, IMMUTABLE_CONFIG, EMPTY_STRING_IS_NULL, g_pserver->bgsave_cpulist, NULL, NULL, NULL),
-    createStringConfig("storage-provider-options", NULL, IMMUTABLE_CONFIG, EMPTY_STRING_IS_NULL, cserver.storage_conf, NULL, NULL, NULL),
     createStringConfig("ignore-warnings", NULL, MODIFIABLE_CONFIG, ALLOW_EMPTY_STRING, g_pserver->ignore_warnings, "", NULL, NULL),
     createStringConfig("proc-title-template", NULL, MODIFIABLE_CONFIG, ALLOW_EMPTY_STRING, cserver.proc_title_template, CONFIG_DEFAULT_PROC_TITLE_TEMPLATE, isValidProcTitleTemplate, updateProcTitleTemplate),
 
@@ -2822,7 +2679,6 @@ standardConfig configs[] = {
     createEnumConfig("loglevel", NULL, MODIFIABLE_CONFIG, loglevel_enum, cserver.verbosity, LL_NOTICE, NULL, NULL),
     createEnumConfig("maxmemory-policy", NULL, MODIFIABLE_CONFIG, maxmemory_policy_enum, g_pserver->maxmemory_policy, MAXMEMORY_NO_EVICTION, NULL, NULL),
     createEnumConfig("appendfsync", NULL, MODIFIABLE_CONFIG, aof_fsync_enum, g_pserver->aof_fsync, AOF_FSYNC_EVERYSEC, NULL, NULL),
-    createEnumConfig("storage-cache-mode", NULL, IMMUTABLE_CONFIG, storage_memory_model_enum, cserver.storage_memory_model, STORAGE_WRITETHROUGH, NULL, NULL),
     createEnumConfig("oom-score-adj", NULL, MODIFIABLE_CONFIG, oom_score_adj_enum, g_pserver->oom_score_adj, OOM_SCORE_ADJ_NO, NULL, updateOOMScoreAdj),
     createEnumConfig("acl-pubsub-default", NULL, MODIFIABLE_CONFIG, acl_pubsub_default_enum, g_pserver->acl_pubsub_default, USER_FLAG_ALLCHANNELS, NULL, NULL),
     createEnumConfig("sanitize-dump-payload", NULL, MODIFIABLE_CONFIG, sanitize_dump_payload_enum, cserver.sanitize_dump_payload, SANITIZE_DUMP_NO, NULL, NULL),
@@ -2843,7 +2699,7 @@ standardConfig configs[] = {
     createIntConfig("lfu-decay-time", NULL, MODIFIABLE_CONFIG, 0, INT_MAX, g_pserver->lfu_decay_time, 1, INTEGER_CONFIG, NULL, NULL),
     createIntConfig("replica-priority", "slave-priority", MODIFIABLE_CONFIG, 0, INT_MAX, g_pserver->slave_priority, 100, INTEGER_CONFIG, NULL, NULL),
     createIntConfig("repl-diskless-sync-delay", NULL, MODIFIABLE_CONFIG, 0, INT_MAX, g_pserver->repl_diskless_sync_delay, 5, INTEGER_CONFIG, NULL, NULL),
-    createIntConfig("maxmemory-samples", NULL, MODIFIABLE_CONFIG, 1, INT_MAX, g_pserver->maxmemory_samples, 16, INTEGER_CONFIG, NULL, NULL),
+    createIntConfig("maxmemory-samples", NULL, MODIFIABLE_CONFIG, 1, INT_MAX, g_pserver->maxmemory_samples, 5, INTEGER_CONFIG, NULL, NULL),
     createIntConfig("maxmemory-eviction-tenacity", NULL, MODIFIABLE_CONFIG, 0, 100, g_pserver->maxmemory_eviction_tenacity, 10, INTEGER_CONFIG, NULL, NULL),
     createIntConfig("timeout", NULL, MODIFIABLE_CONFIG, 0, INT_MAX, cserver.maxidletime, 0, INTEGER_CONFIG, NULL, NULL), /* Default client timeout: infinite */
     createIntConfig("replica-announce-port", "slave-announce-port", MODIFIABLE_CONFIG, 0, 65535, g_pserver->slave_announce_port, 0, INTEGER_CONFIG, NULL, NULL),
@@ -2861,9 +2717,7 @@ standardConfig configs[] = {
     createIntConfig("min-replicas-to-write", "min-slaves-to-write", MODIFIABLE_CONFIG, 0, INT_MAX, g_pserver->repl_min_slaves_to_write, 0, INTEGER_CONFIG, NULL, updateGoodSlaves),
     createIntConfig("min-replicas-max-lag", "min-slaves-max-lag", MODIFIABLE_CONFIG, 0, INT_MAX, g_pserver->repl_min_slaves_max_lag, 10, INTEGER_CONFIG, NULL, updateGoodSlaves),
     createIntConfig("min-clients-per-thread", NULL, MODIFIABLE_CONFIG, 0, 400, cserver.thread_min_client_threshold, 20, INTEGER_CONFIG, NULL, NULL),
-    createIntConfig("storage-flush-period", NULL, MODIFIABLE_CONFIG, 1, 10000, g_pserver->storage_flush_period, 500, INTEGER_CONFIG, NULL, NULL),
     createIntConfig("replica-quorum", NULL, MODIFIABLE_CONFIG, -1, INT_MAX, g_pserver->repl_quorum, -1, INTEGER_CONFIG, NULL, NULL),
-    createIntConfig("replica-weighting-factor", NULL, MODIFIABLE_CONFIG, 1, INT_MAX, g_pserver->replicaIsolationFactor, 2, INTEGER_CONFIG, NULL, NULL),
     /* Unsigned int configs */
     createUIntConfig("maxclients", NULL, MODIFIABLE_CONFIG, 1, UINT_MAX, g_pserver->maxclients, 10000, INTEGER_CONFIG, NULL, updateMaxclients),
     createUIntConfig("loading-process-events-interval-keys", NULL, MODIFIABLE_CONFIG, 0, LONG_MAX, g_pserver->loading_process_events_interval_keys, 8192, MEMORY_CONFIG, NULL, NULL),
@@ -2881,12 +2735,9 @@ standardConfig configs[] = {
     createLongLongConfig("proto-max-bulk-len", NULL, MODIFIABLE_CONFIG, 1024*1024, LLONG_MAX, g_pserver->proto_max_bulk_len, 512ll*1024*1024, MEMORY_CONFIG, NULL, NULL), /* Bulk request max size */
     createLongLongConfig("stream-node-max-entries", NULL, MODIFIABLE_CONFIG, 0, LLONG_MAX, g_pserver->stream_node_max_entries, 100, INTEGER_CONFIG, NULL, NULL),
     createLongLongConfig("repl-backlog-size", NULL, MODIFIABLE_CONFIG, 1, LLONG_MAX, g_pserver->repl_backlog_size, 1024*1024, MEMORY_CONFIG, NULL, updateReplBacklogSize), /* Default: 1mb */
-    createLongLongConfig("repl-backlog-disk-reserve", NULL, IMMUTABLE_CONFIG, 0, LLONG_MAX, cserver.repl_backlog_disk_size, 0, MEMORY_CONFIG, NULL, NULL),
-    createLongLongConfig("max-snapshot-slip", NULL, MODIFIABLE_CONFIG, 0, 5000, g_pserver->snapshot_slip, 400, 0, NULL, NULL),
 
     /* Unsigned Long Long configs */
     createULongLongConfig("maxmemory", NULL, MODIFIABLE_CONFIG, 0, LLONG_MAX, g_pserver->maxmemory, 0, MEMORY_CONFIG, NULL, updateMaxmemory),
-    createULongLongConfig("maxstorage", NULL, MODIFIABLE_CONFIG, 0, LLONG_MAX, g_pserver->maxstorage, 0, MEMORY_CONFIG, NULL, NULL),
 
     /* Size_t configs */
     createSizeTConfig("hash-max-ziplist-entries", NULL, MODIFIABLE_CONFIG, 0, LONG_MAX, g_pserver->hash_max_ziplist_entries, 512, INTEGER_CONFIG, NULL, NULL),
@@ -2908,8 +2759,6 @@ standardConfig configs[] = {
     createULongConfig("loading-process-events-interval-bytes", NULL, MODIFIABLE_CONFIG, 0, LONG_MAX, g_pserver->loading_process_events_interval_bytes, 2*1024*1024, MEMORY_CONFIG, NULL, NULL),
     createBoolConfig("multi-master-no-forward", NULL, MODIFIABLE_CONFIG, cserver.multimaster_no_forward, 0, validateMultiMasterNoForward, NULL),
     createBoolConfig("allow-write-during-load", NULL, MODIFIABLE_CONFIG, g_pserver->fWriteDuringActiveLoad, 0, NULL, NULL),
-    createBoolConfig("force-backlog-disk-reserve", NULL, MODIFIABLE_CONFIG, cserver.force_backlog_disk, 0, NULL, NULL),
-    createBoolConfig("soft-shutdown", NULL, MODIFIABLE_CONFIG, g_pserver->config_soft_shutdown, 0, NULL, NULL),
 
 #ifdef USE_OPENSSL
     createIntConfig("tls-port", NULL, MODIFIABLE_CONFIG, 0, 65535, g_pserver->tls_port, 0, INTEGER_CONFIG, NULL, updateTLSPort), /* TCP port. */
@@ -2920,7 +2769,6 @@ standardConfig configs[] = {
     createEnumConfig("tls-auth-clients", NULL, MODIFIABLE_CONFIG, tls_auth_clients_enum, g_pserver->tls_auth_clients, TLS_CLIENT_AUTH_YES, NULL, NULL),
     createBoolConfig("tls-prefer-server-ciphers", NULL, MODIFIABLE_CONFIG, g_pserver->tls_ctx_config.prefer_server_ciphers, 0, NULL, updateTlsCfgBool),
     createBoolConfig("tls-session-caching", NULL, MODIFIABLE_CONFIG, g_pserver->tls_ctx_config.session_caching, 1, NULL, updateTlsCfgBool),
-    createBoolConfig("tls-rotation", NULL, MODIFIABLE_CONFIG, g_pserver->tls_rotation, 0, NULL, updateTlsCfgBool),
     createStringConfig("tls-cert-file", NULL, MODIFIABLE_CONFIG, EMPTY_STRING_IS_NULL, g_pserver->tls_ctx_config.cert_file, NULL, NULL, updateTlsCfg),
     createStringConfig("tls-key-file", NULL, MODIFIABLE_CONFIG, EMPTY_STRING_IS_NULL, g_pserver->tls_ctx_config.key_file, NULL, NULL, updateTlsCfg),
     createStringConfig("tls-key-file-pass", NULL, MODIFIABLE_CONFIG, EMPTY_STRING_IS_NULL, g_pserver->tls_ctx_config.key_file_pass, NULL, NULL, updateTlsCfg),
@@ -2964,7 +2812,7 @@ NULL
         };
 
         addReplyHelp(c, help);
-    } else if (!strcasecmp(szFromObj(c->argv[1]),"set") && c->argc >= 3) {
+    } else if (!strcasecmp(szFromObj(c->argv[1]),"set") && c->argc == 4) {
         configSetCommand(c);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"get") && c->argc == 3) {
         configGetCommand(c);

src/connection.cpp

@@ -29,7 +29,6 @@
 
 #include "server.h"
 #include "connhelpers.h"
-#include "aelocker.h"
 
 /* The connections module provides a lean abstraction of network connections
  * to avoid direct socket and async event management across the Redis code base.
@@ -113,7 +112,7 @@ static int connSocketConnect(connection *conn, const char *addr, int port, const
     conn->state = CONN_STATE_CONNECTING;
 
     conn->conn_handler = connect_handler;
-    aeCreateFileEvent(serverTL->el, conn->fd, AE_WRITABLE|AE_WRITE_THREADSAFE,
+    aeCreateFileEvent(serverTL->el, conn->fd, AE_WRITABLE,
             conn->type->ae_handler, conn);
 
     return C_OK;
@@ -226,15 +225,14 @@ static int connSocketSetWriteHandler(connection *conn, ConnectionCallbackFunc fu
     else
         conn->flags &= ~CONN_FLAG_WRITE_BARRIER;
 
+    int flags = AE_WRITABLE;
     if (fThreadSafe)
-        conn->flags |= CONN_FLAG_WRITE_THREADSAFE;
-    else
-        conn->flags &= ~CONN_FLAG_WRITE_THREADSAFE;
+        flags |= AE_WRITE_THREADSAFE;
 
     if (!conn->write_handler)
         aeDeleteFileEvent(serverTL->el,conn->fd,AE_WRITABLE);
     else
-        if (aeCreateFileEvent(serverTL->el,conn->fd,AE_WRITABLE|AE_WRITE_THREADSAFE,
+        if (aeCreateFileEvent(serverTL->el,conn->fd,flags,
                     conn->type->ae_handler,conn) == AE_ERR) return C_ERR;
     return C_OK;
 }
@@ -244,18 +242,17 @@ static int connSocketSetWriteHandler(connection *conn, ConnectionCallbackFunc fu
  */
 static int connSocketSetReadHandler(connection *conn, ConnectionCallbackFunc func, bool fThreadSafe) {
     if (func == conn->read_handler) return C_OK;
-    
+
+    int flags = AE_READABLE;
     if (fThreadSafe)
-        conn->flags |= CONN_FLAG_READ_THREADSAFE;
-    else
-        conn->flags &= ~CONN_FLAG_READ_THREADSAFE;
+        flags |= AE_READ_THREADSAFE;
 
     conn->read_handler = func;
     if (!conn->read_handler)
         aeDeleteFileEvent(serverTL->el,conn->fd,AE_READABLE);
     else
         if (aeCreateFileEvent(serverTL->el,conn->fd,
-                    AE_READABLE|AE_READ_THREADSAFE,conn->type->ae_handler,conn) == AE_ERR) return C_ERR;
+                    flags,conn->type->ae_handler,conn) == AE_ERR) return C_ERR;
     return C_OK;
 }
 
@@ -263,7 +260,7 @@ static const char *connSocketGetLastError(connection *conn) {
     return strerror(conn->last_errno);
 }
 
-void connSocketEventHandler(struct aeEventLoop *el, int fd, void *clientData, int mask)
+static void connSocketEventHandler(struct aeEventLoop *el, int fd, void *clientData, int mask)
 {
     UNUSED(el);
     UNUSED(fd);
@@ -282,11 +279,7 @@ void connSocketEventHandler(struct aeEventLoop *el, int fd, void *clientData, in
 
         if (!conn->write_handler) aeDeleteFileEvent(serverTL->el,conn->fd,AE_WRITABLE);
 
-        {
-        AeLocker locker;
-        locker.arm(nullptr);
         if (!callHandler(conn, conn->conn_handler)) return;
-        }
         conn->conn_handler = NULL;
     }
 
@@ -308,27 +301,15 @@ void connSocketEventHandler(struct aeEventLoop *el, int fd, void *clientData, in
 
     /* Handle normal I/O flows */
     if (!invert && call_read) {
-        AeLocker lock;
-        if (!(conn->flags & CONN_FLAG_READ_THREADSAFE))
-            lock.arm(nullptr);
-        
         if (!callHandler(conn, conn->read_handler)) return;
     }
     /* Fire the writable event. */
     if (call_write) {
-        AeLocker lock;
-        if (!(conn->flags & CONN_FLAG_WRITE_THREADSAFE))
-            lock.arm(nullptr);
-
         if (!callHandler(conn, conn->write_handler)) return;
     }
     /* If we have to invert the call, fire the readable event now
      * after the writable one. */
     if (invert && call_read) {
-        AeLocker lock;
-        if (!(conn->flags & CONN_FLAG_READ_THREADSAFE))
-            lock.arm(nullptr);
-        
         if (!callHandler(conn, conn->read_handler)) return;
     }
 }
@@ -471,3 +452,4 @@ const char *connGetInfo(connection *conn, char *buf, size_t buf_len) {
     snprintf(buf, buf_len-1, "fd=%i", conn == NULL ? -1 : conn->fd);
     return buf;
 }
+

src/cowptr.h

@@ -1,55 +0,0 @@
-#pragma once
-
-template <class T>
-class CowPtr
-{
-    public:
-        typedef std::shared_ptr<T> RefPtr;
-
-    private:
-        mutable RefPtr m_sp;
-
-        void detach()
-        {
-            if( !( m_sp == nullptr || m_sp.unique() ) ) {
-                m_sp = std::make_shared<T>(*m_sp);
-            }
-        }
-
-    public:
-        CowPtr() = default;
-
-        CowPtr(const RefPtr& refptr)
-            :   m_sp(refptr)
-        {
-        }
-
-        bool operator==(std::nullptr_t) const
-        {
-            return m_sp == nullptr;
-        }
-
-        bool operator!=(std::nullptr_t) const
-        {
-            return m_sp != nullptr;
-        }
-        
-        const T& operator*() const
-        {
-            return *m_sp;
-        }
-        T& operator*()
-        {
-            detach();
-            return *m_sp;
-        }
-        const T* operator->() const
-        {
-            return m_sp.operator->();
-        }
-        T* operator->()
-        {
-            detach();
-            return m_sp.operator->();
-        }
-};

src/cron.cpp

@@ -31,8 +31,7 @@ void cronCommand(client *c)
     if (getLongLongFromObjectOrReply(c, c->argv[ARG_EXPIRE], &interval, "missing expire time") != C_OK)
         return;
 
-    long long base;
-    __atomic_load(&g_pserver->mstime, &base, __ATOMIC_ACQUIRE);
+    long long base = g_pserver->mstime;
     if (getLongLongFromObject(c->argv[ARG_EXPIRE+1], &base) == C_OK) {
         arg_offset++;
         std::swap(base, interval);
@@ -61,7 +60,7 @@ void cronCommand(client *c)
     spjob->interval = (uint64_t)interval;
     spjob->startTime = (uint64_t)base;
     spjob->fSingleShot = fSingleShot;
-    spjob->dbNum = dbnumFromDb(c->db);
+    spjob->dbNum = c->db - g_pserver->db;
     for (long i = 0; i < numkeys; ++i)
         spjob->veckeys.emplace_back(sdsdup(szFromObj(c->argv[ARG_KEYSTART+i])));
     for (long i = ARG_KEYSTART + numkeys; i < c->argc; ++i)
@@ -117,25 +116,23 @@ void executeCronJobExpireHook(const char *key, robj *o)
     int dbId = job->dbNum;
     if (job->fSingleShot)
     {
-        serverAssert(dbSyncDelete(cFake->db, keyobj));
+        dbSyncDelete(cFake->db, keyobj);
     }
     else
     {
         job->startTime += job->interval;
-        mstime_t mstime;
-        __atomic_load(&g_pserver->mstime, &mstime, __ATOMIC_ACQUIRE);
-        if (job->startTime < (uint64_t)mstime)
+        if (job->startTime < (uint64_t)g_pserver->mstime)
         {
             // If we are more than one interval in the past then fast forward to
             //  the first interval still in the future.  If startTime wasn't zero align
             //  this to the original startTime, if it was zero align to now
             if (job->startTime == job->interval)
             {   // startTime was 0
-                job->startTime = mstime + job->interval;
+                job->startTime = g_pserver->mstime + job->interval;
             }
             else
             {
-                auto delta = mstime - job->startTime;
+                auto delta = g_pserver->mstime - job->startTime;
                 auto multiple = (delta / job->interval)+1;
                 job->startTime += job->interval * multiple;
             }

src/debug.cpp

@@ -78,7 +78,6 @@ void bugReportStart(void);
 void printCrashReport(void);
 void bugReportEnd(int killViaSignal, int sig);
 void logStackTrace(void *eip, int uplevel);
-void getTempFileName(char tmpfile[], int tmpfileNum);
 
 /* ================================= Debugging ============================== */
 
@@ -149,8 +148,7 @@ void mixStringObjectDigest(unsigned char *digest, robj_roptr o) {
 void xorObjectDigest(redisDb *db, robj_roptr keyobj, unsigned char *digest, robj_roptr o) {
     uint32_t aux = htonl(o->type);
     mixDigest(digest,&aux,sizeof(aux));
-    std::unique_lock<fastlock> ul(g_expireLock);
-    expireEntry *pexpire = db->getExpire(keyobj);
+    expireEntry *pexpire = getExpire(db,keyobj);
     long long expiretime = INVALID_EXPIRE;
     char buf[128];
 
@@ -293,15 +291,19 @@ void xorObjectDigest(redisDb *db, robj_roptr keyobj, unsigned char *digest, robj
  * as input in order to ensure that a different ordered list will result in
  * a different digest. */
 void computeDatasetDigest(unsigned char *final) {
+    unsigned char digest[20];
+    dictIterator *di = NULL;
+    dictEntry *de;
     int j;
     uint32_t aux;
 
     memset(final,0,20); /* Start with a clean result */
 
     for (j = 0; j < cserver.dbnum; j++) {
-        redisDb *db = g_pserver->db[j];
+        redisDb *db = g_pserver->db+j;
 
-        if (db->size() == 0) continue;
+        if (dictSize(db->dict) == 0) continue;
+        di = dictGetSafeIterator(db->dict);
 
         /* hash the DB id, so the same dataset moved in a different
          * DB will lead to a different digest */
@@ -309,22 +311,24 @@ void computeDatasetDigest(unsigned char *final) {
         mixDigest(final,&aux,sizeof(aux));
 
         /* Iterate this DB writing every entry */
-        db->iterate_threadsafe([final, db](const char *key, robj_roptr o)->bool {
-            unsigned char digest[20];
-            robj *keyobj;
+        while((de = dictNext(di)) != NULL) {
+            sds key;
+            robj *keyobj, *o;
 
             memset(digest,0,20); /* This key-val digest */
+            key = (sds)dictGetKey(de);
             keyobj = createStringObject(key,sdslen(key));
 
             mixDigest(digest,key,sdslen(key));
 
+            o = (robj*)dictGetVal(de);
             xorObjectDigest(db,keyobj,digest,o);
 
             /* We can finally xor the key-val digest to the final digest */
             xorDigest(final,digest,20);
             decrRefCount(keyobj);
-            return true;
-        });
+        }
+        dictReleaseIterator(di);
     }
 }
 
@@ -542,7 +546,7 @@ NULL
         if (save) {
             rdbSaveInfo rsi, *rsiptr;
             rsiptr = rdbPopulateSaveInfo(&rsi);
-            if (rdbSave(nullptr, rsiptr) != C_OK) {
+            if (rdbSaveFile(g_pserver->rdb_filename,rsiptr) != C_OK) {
                 addReply(c,shared.err);
                 return;
             }
@@ -576,14 +580,15 @@ NULL
         serverLog(LL_WARNING,"Append Only File loaded by DEBUG LOADAOF");
         addReply(c,shared.ok);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"object") && c->argc == 3) {
+        dictEntry *de;
         robj *val;
         const char *strenc;
 
-        val = c->db->find(c->argv[2]);
-        if (val == NULL) {
+        if ((de = dictFind(c->db->dict,ptrFromObj(c->argv[2]))) == NULL) {
             addReplyErrorObject(c,shared.nokeyerr);
             return;
         }
+        val = (robj*)dictGetVal(de);
         strenc = strEncoding(val->encoding);
 
         char extra[138] = {0};
@@ -627,14 +632,16 @@ NULL
             strenc, rdbSavedObjectLen(val, c->argv[2]),
             val->lru, estimateObjectIdleTime(val)/1000, extra);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"sdslen") && c->argc == 3) {
-        auto itr = c->db->find(c->argv[2]);
-        robj *val = itr.val();
-        const char *key = itr.key();
+        dictEntry *de;
+        robj *val;
+        sds key;
 
-        if (val == NULL) {
+        if ((de = dictFind(c->db->dict,ptrFromObj(c->argv[2]))) == NULL) {
             addReplyErrorObject(c,shared.nokeyerr);
             return;
         }
+        val = (robj*)dictGetVal(de);
+        key = (sds)dictGetKey(de);
 
         if (val->type != OBJ_STRING || !sdsEncodedObject(val)) {
             addReplyError(c,"Not an sds encoded string.");
@@ -644,7 +651,7 @@ NULL
                 "val_sds_len:%lld, val_sds_avail:%lld, val_zmalloc: %lld",
                 (long long) sdslen(key),
                 (long long) sdsavail(key),
-                (long long) sdsZmallocSize((sds)key),
+                (long long) sdsZmallocSize(key),
                 (long long) sdslen(szFromObj(val)),
                 (long long) sdsavail(szFromObj(val)),
                 (long long) getStringObjectSdsUsedMemory(val));
@@ -669,8 +676,8 @@ NULL
 
         if (getPositiveLongFromObjectOrReply(c, c->argv[2], &keys, NULL) != C_OK)
             return;
-        
-        c->db->expand(keys);
+
+        dictExpand(c->db->dict,keys);
         long valsize = 0;
         if ( c->argc == 5 && getPositiveLongFromObjectOrReply(c, c->argv[4], &valsize, NULL) != C_OK ) 
             return;
@@ -714,8 +721,8 @@ NULL
 
             /* We don't use lookupKey because a debug command should
              * work on logically expired keys */
-            auto itr = c->db->find(c->argv[j]);
-            robj* o = (robj*)(itr == NULL ? NULL : itr.val());
+            dictEntry *de;
+            robj *o = (robj*)((de = dictFind(c->db->dict,ptrFromObj(c->argv[j]))) == NULL) ? NULL : (robj*)dictGetVal(de);
             if (o) xorObjectDigest(c->db,c->argv[j],digest,o);
 
             sds d = sdsempty();
@@ -840,11 +847,11 @@ NULL
         }
 
         stats = sdscatprintf(stats,"[Dictionary HT]\n");
-        g_pserver->db[dbid]->getStats(buf,sizeof(buf));
+        dictGetStats(buf,sizeof(buf),g_pserver->db[dbid].dict);
         stats = sdscat(stats,buf);
 
         stats = sdscatprintf(stats,"[Expires set]\n");
-        g_pserver->db[dbid]->getExpireStats(buf, sizeof(buf));
+        g_pserver->db[dbid].setexpire->getstats(buf, sizeof(buf));
         stats = sdscat(stats, buf);
 
         addReplyVerbatim(c,stats,sdslen(stats),"txt");
@@ -906,10 +913,6 @@ NULL
             c->flags &= ~(CLIENT_MASTER | CLIENT_MASTER_FORCE_REPLY);
         }
         addReply(c, shared.ok);
-    } else if (!strcasecmp(szFromObj(c->argv[1]), "get-temp-file") && c->argc == 2) {
-        char tmpfile[256];
-        getTempFileName(tmpfile, g_pserver->rdbThreadVars.tmpfileNum);
-        addReplyBulkCString(c, tmpfile);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"truncate-repl-backlog") && c->argc == 2) {
         g_pserver->repl_backlog_idx = 0;
         g_pserver->repl_backlog_off = g_pserver->master_repl_offset+1;
@@ -1790,10 +1793,12 @@ void logCurrentClient(void) {
      * selected DB, and if so print info about the associated object. */
     if (cc->argc > 1) {
         robj *val, *key;
+        dictEntry *de;
 
         key = getDecodedObject(cc->argv[1]);
-        val = cc->db->find(key);
-        if (val) {
+        de = dictFind(cc->db->dict, ptrFromObj(key));
+        if (de) {
+            val = (robj*)dictGetVal(de);
             serverLog(LL_WARNING,"key '%s' found in DB containing the following object:", (char*)ptrFromObj(key));
             serverLogObjectDebugInfo(val);
         }

src/defrag.cpp

@@ -851,7 +851,6 @@ long defragModule(redisDb *db, dictEntry *kde) {
  * all the various pointers it has. Returns a stat of how many pointers were
  * moved. */
 long defragKey(redisDb *db, dictEntry *de) {
-    std::unique_lock<fastlock> ul(g_expireLock);
     sds keysds = (sds)dictGetKey(de);
     robj *newob, *ob;
     unsigned char *newzl;
@@ -865,14 +864,15 @@ long defragKey(redisDb *db, dictEntry *de) {
     if (newsds)
     {
         defragged++, de->key = newsds;
-        if (!db->setexpire()->empty()) {
-            bool fReplaced = replaceSatelliteOSetKeyPtr(*const_cast<expireset*>(db->setexpire()), keysds, newsds);
+        if (!db->setexpire->empty()) {
+            bool fReplaced = replaceSatelliteOSetKeyPtr(*db->setexpire, keysds, newsds);
             serverAssert(fReplaced == ob->FExpires());
         } else {
             serverAssert(!ob->FExpires());
         }
     }
 
+    /* Try to defrag robj and / or string value. */
     if ((newob = activeDefragStringOb(ob, &defragged))) {
         de->v.val = newob;
         ob = newob;
@@ -989,8 +989,9 @@ long defragOtherGlobals() {
 
 /* returns 0 more work may or may not be needed (see non-zero cursor),
  * and 1 if time is up and more work is needed. */
-int defragLaterItem(sds key, robj *ob, unsigned long *cursor, long long endtime) {
-    if (ob) {
+int defragLaterItem(dictEntry *de, unsigned long *cursor, long long endtime) {
+    if (de) {
+        robj *ob = (robj*)dictGetVal(de);
         if (ob->type == OBJ_LIST) {
             return scanLaterList(ob, cursor, endtime, &g_pserver->stat_active_defrag_hits);
         } else if (ob->type == OBJ_SET) {
@@ -1002,9 +1003,7 @@ int defragLaterItem(sds key, robj *ob, unsigned long *cursor, long long endtime)
         } else if (ob->type == OBJ_STREAM) {
             return scanLaterStreamListpacks(ob, cursor, endtime, &g_pserver->stat_active_defrag_hits);
         } else if (ob->type == OBJ_MODULE) {
-            redisObjectStack oT;
-            initStaticStringObject(oT, key);
-            return moduleLateDefrag(&oT, ob, cursor, endtime, &g_pserver->stat_active_defrag_hits);
+            return moduleLateDefrag((robj*)dictGetKey(de), ob, cursor, endtime, &g_pserver->stat_active_defrag_hits);
         } else {
             *cursor = 0; /* object type may have changed since we schedule it for later */
         }
@@ -1049,11 +1048,11 @@ int defragLaterStep(redisDb *db, long long endtime) {
         }
 
         /* each time we enter this function we need to fetch the key from the dict again (if it still exists) */
-        robj *o = db->find(defrag_later_current_key);
+        dictEntry *de = dictFind(db->dict, defrag_later_current_key);
         key_defragged = g_pserver->stat_active_defrag_hits;
         do {
             int quit = 0;
-            if (defragLaterItem(defrag_later_current_key, o, &defrag_later_cursor, endtime))
+            if (defragLaterItem(de, &defrag_later_cursor, endtime))
                 quit = 1; /* time is up, we didn't finish all the work */
 
             /* Once in 16 scan iterations, 512 pointer reallocations, or 64 fields
@@ -1201,7 +1200,7 @@ void activeDefragCycle(void) {
                 start_stat = g_pserver->stat_active_defrag_hits;
             }
 
-            db = g_pserver->db[current_db];
+            db = &g_pserver->db[current_db];
             cursor = 0;
         }
 
@@ -1212,8 +1211,7 @@ void activeDefragCycle(void) {
                 break; /* this will exit the function and we'll continue on the next cycle */
             }
 
-            // we actually look at the objects too but defragScanCallback can handle missing values
-            cursor = dictScan(db->dictUnsafeKeyOnly(), cursor, defragScanCallback, defragDictBucketCallback, db);
+            cursor = dictScan(db->dict, cursor, defragScanCallback, defragDictBucketCallback, db);
 
             /* Once in 16 scan iterations, 512 pointer reallocations. or 64 keys
              * (if we have a lot of pointers in one hash bucket or rehasing),

src/dict.cpp

@@ -42,7 +42,6 @@
 #include <stdarg.h>
 #include <limits.h>
 #include <sys/time.h>
-#include <algorithm>
 
 #include "dict.h"
 #include "zmalloc.h"
@@ -63,15 +62,13 @@ static unsigned int dict_force_resize_ratio = 5;
 
 static int _dictExpandIfNeeded(dict *ht);
 static unsigned long _dictNextPower(unsigned long size);
-long _dictKeyIndex(dict *ht, const void *key, uint64_t 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 -------------------------------- */
 
 static uint8_t dict_hash_function_seed[16];
 
-extern "C" void asyncFreeDictTable(dictEntry **de);
-
 void dictSetHashFunctionSeed(uint8_t *seed) {
     memcpy(dict_hash_function_seed,seed,sizeof(dict_hash_function_seed));
 }
@@ -126,22 +123,32 @@ int _dictInit(dict *d, dictType *type,
     d->privdata = privDataPtr;
     d->rehashidx = -1;
     d->pauserehash = 0;
-    d->asyncdata = nullptr;
-    d->refcount = 1;
-    d->noshrink = false;
     return DICT_OK;
 }
 
+/* Resize the table to the minimal size that contains all the elements,
+ * but with the invariant of a USED/BUCKETS ratio near to <= 1 */
+int dictResize(dict *d)
+{
+    unsigned long minimal;
+
+    if (!dict_can_resize || dictIsRehashing(d)) return DICT_ERR;
+    minimal = d->ht[0].used;
+    if (minimal < DICT_HT_INITIAL_SIZE)
+        minimal = DICT_HT_INITIAL_SIZE;
+    return dictExpand(d, minimal);
+}
+
 /* Expand or create the hash table,
  * when malloc_failed is non-NULL, it'll avoid panic if malloc fails (in which case it'll be set to 1).
  * Returns DICT_OK if expand was performed, and DICT_ERR if skipped. */
-int _dictExpand(dict *d, unsigned long size, bool fShrink, int* malloc_failed)
+int _dictExpand(dict *d, unsigned long size, int* malloc_failed)
 {
     if (malloc_failed) *malloc_failed = 0;
 
     /* the size is invalid if it is smaller than the number of
      * elements already inside the hash table */
-    if (dictIsRehashing(d) || (d->ht[0].used > size && !fShrink) || size == 0)
+    if (dictIsRehashing(d) || d->ht[0].used > size)
         return DICT_ERR;
 
     dictht n; /* the new hash table */
@@ -181,169 +188,15 @@ int _dictExpand(dict *d, unsigned long size, bool fShrink, int* malloc_failed)
     return DICT_OK;
 }
 
-/* Resize the table to the minimal size that contains all the elements,
- * but with the invariant of a USED/BUCKETS ratio near to <= 1 */
-int dictResize(dict *d)
-{
-    unsigned long minimal;
-
-    if (!dict_can_resize || dictIsRehashing(d)) return DICT_ERR;
-    minimal = d->ht[0].used;
-    if (minimal < DICT_HT_INITIAL_SIZE)
-        minimal = DICT_HT_INITIAL_SIZE;
-    return _dictExpand(d, minimal, false /*fShirnk*/, nullptr);
-}
-
-int dictMerge(dict *dst, dict *src)
-{
-#define MERGE_BLOCK_SIZE 4
-    dictEntry *rgdeT[MERGE_BLOCK_SIZE];
-
-    assert(dst != src);
-    if (dictSize(src) == 0)
-        return DICT_OK;
-    
-    if (dictSize(dst) == 0)
-    {
-        dict::swap(*dst, *src);
-        std::swap(dst->pauserehash, src->pauserehash);
-        return DICT_OK;
-    }
-
-    size_t expectedSize = dictSize(src) + dictSize(dst);
-    if (dictSize(src) > dictSize(dst) && src->asyncdata == nullptr && dst->asyncdata == nullptr)
-    {
-        dict::swap(*dst, *src);
-        std::swap(dst->pauserehash, src->pauserehash);
-    }
-
-    if (!dictIsRehashing(dst) && !dictIsRehashing(src))
-    {
-        if (dst->ht[0].size >= src->ht[0].size)
-        {
-            dst->ht[1] = dst->ht[0];
-            dst->ht[0] = src->ht[0];
-        }
-        else
-        {
-            dst->ht[1] = src->ht[0];
-        }
-        _dictReset(&src->ht[0]);
-        dst->rehashidx = 0;
-        assert(dictIsRehashing(dst));
-        assert((dictSize(src)+dictSize(dst)) == expectedSize);
-        return DICT_OK;
-    }
-
-    if (!dictIsRehashing(src) && dictSize(src) > 0 &&
-        (src->ht[0].size == dst->ht[0].size || src->ht[0].size == dst->ht[1].size))
-    {
-        auto &htDst = (src->ht[0].size == dst->ht[0].size) ? dst->ht[0] : dst->ht[1];
-
-        assert(src->ht[0].size == htDst.size);
-        for (size_t ide = 0; ide < src->ht[0].size; ide += MERGE_BLOCK_SIZE)
-        {
-            if (src->ht[0].used == 0)
-                break;
-
-            for (int dde = 0; dde < MERGE_BLOCK_SIZE; ++dde) {
-                rgdeT[dde] = src->ht[0].table[ide + dde];
-                src->ht[0].table[ide + dde] = nullptr;
-            }
-
-            for (;;) {
-                bool fAnyFound = false;
-                for (int dde = 0; dde < MERGE_BLOCK_SIZE; ++dde) {
-                    if (rgdeT[dde] == nullptr)
-                        continue;
-                    dictEntry *deNext = rgdeT[dde]->next;
-                    rgdeT[dde]->next = htDst.table[ide+dde];
-                    htDst.table[ide+dde] = rgdeT[dde];
-                    rgdeT[dde] = deNext;
-                    htDst.used++;
-                    src->ht[0].used--;
-
-                    fAnyFound = fAnyFound || (deNext != nullptr);
-                }
-
-                if (!fAnyFound)
-                    break;
-            }
-        }
-        // If we copied to the base hash table of a rehashing dst, reset the rehash
-        if (dictIsRehashing(dst) && src->ht[0].size == dst->ht[0].size)
-            dst->rehashidx = 0;
-        assert(dictSize(src) == 0);
-        assert((dictSize(src)+dictSize(dst)) == expectedSize);
-        return DICT_OK;
-    }
-
-    _dictExpand(dst, dictSize(dst)+dictSize(src), false /* fShrink */, nullptr);    // start dst rehashing if necessary
-    auto &htDst = dictIsRehashing(dst) ? dst->ht[1] : dst->ht[0];
-    for (int iht = 0; iht < 2; ++iht)
-    {
-        for (size_t ide = 0; ide < src->ht[iht].size; ide += MERGE_BLOCK_SIZE)
-        {
-            if (src->ht[iht].used == 0)
-                break;
-
-            for (int dde = 0; dde < MERGE_BLOCK_SIZE; ++dde) {
-                rgdeT[dde] = src->ht[iht].table[ide + dde];
-                src->ht[iht].table[ide + dde] = nullptr;
-            }
-
-            for (;;) {
-                bool fAnyFound = false;
-                for (int dde = 0; dde < MERGE_BLOCK_SIZE; ++dde) {
-                    if (rgdeT[dde] == nullptr)
-                        continue;
-                    uint64_t h = dictHashKey(dst, rgdeT[dde]->key) & htDst.sizemask;
-                    dictEntry *deNext = rgdeT[dde]->next;
-                    rgdeT[dde]->next = htDst.table[h];
-                    htDst.table[h] = rgdeT[dde];
-                    rgdeT[dde] = deNext;
-                    htDst.used++;
-                    src->ht[iht].used--;
-                    fAnyFound = fAnyFound || (deNext != nullptr);
-                }
-                if (!fAnyFound)
-                    break;
-            }
-#if 0
-            dictEntry *de = src->ht[iht].table[ide];
-            src->ht[iht].table[ide] = nullptr;
-            while (de != nullptr)
-            {
-                dictEntry *deNext = de->next;
-                
-                uint64_t h = dictHashKey(dst, de->key) & htDst.sizemask;
-                de->next = htDst.table[h];
-                htDst.table[h] = de;
-                htDst.used++;
-
-                de = deNext;
-                src->ht[iht].used--;
-            }
-#endif
-        }
-    }
-    assert((dictSize(src)+dictSize(dst)) == expectedSize);
-    return DICT_OK;
-}
-
 /* return DICT_ERR if expand was not performed */
-int dictExpand(dict *d, unsigned long size, bool fShrink) {
-    // External expand likely means mass insertion, and we don't want to shrink during that
-    d->noshrink = true;
-    return _dictExpand(d, size, fShrink, NULL);
+int dictExpand(dict *d, unsigned long size) {
+    return _dictExpand(d, size, NULL);
 }
 
 /* return DICT_ERR if expand failed due to memory allocation failure */
-int dictTryExpand(dict *d, unsigned long size, bool fShrink) {
+int dictTryExpand(dict *d, unsigned long size) {
     int malloc_failed;
-    // External expand likely means mass insertion, and we don't want to shrink during that
-    d->noshrink = true;
-    _dictExpand(d, size, fShrink, &malloc_failed);
+    _dictExpand(d, size, &malloc_failed);
     return malloc_failed? DICT_ERR : DICT_OK;
 }
 
@@ -359,7 +212,6 @@ int dictTryExpand(dict *d, unsigned long size, bool fShrink) {
 int dictRehash(dict *d, int n) {
     int empty_visits = n*10; /* Max number of empty buckets to visit. */
     if (!dictIsRehashing(d)) return 0;
-    if (d->asyncdata) return 0;
 
     while(n-- && d->ht[0].used != 0) {
         dictEntry *de, *nextde;
@@ -390,7 +242,7 @@ int dictRehash(dict *d, int n) {
 
     /* Check if we already rehashed the whole table... */
     if (d->ht[0].used == 0) {
-        asyncFreeDictTable(d->ht[0].table);
+        zfree(d->ht[0].table);
         d->ht[0] = d->ht[1];
         _dictReset(&d->ht[1]);
         d->rehashidx = -1;
@@ -401,158 +253,6 @@ int dictRehash(dict *d, int n) {
     return 1;
 }
 
-dictAsyncRehashCtl::dictAsyncRehashCtl(struct dict *d, dictAsyncRehashCtl *next) : dict(d), next(next) {
-    queue.reserve(c_targetQueueSize);
-    __atomic_fetch_add(&d->refcount, 1, __ATOMIC_ACQ_REL);
-    this->rehashIdxBase = d->rehashidx;
-}
-
-dictAsyncRehashCtl *dictRehashAsyncStart(dict *d, int buckets) {
-    assert(d->type->asyncfree != nullptr);
-    if (!dictIsRehashing(d) || d->pauserehash != 0) return nullptr;
-
-    d->asyncdata = new dictAsyncRehashCtl(d, d->asyncdata);
-
-    int empty_visits = buckets*10;
-
-    while (d->asyncdata->queue.size() < (size_t)buckets && (size_t)d->rehashidx < d->ht[0].size) {
-        dictEntry *de;
-
-        /* Note that rehashidx can't overflow as we are sure there are more
-         * elements because ht[0].used != 0 */
-        while(d->ht[0].table[d->rehashidx] == NULL) {
-            d->rehashidx++;
-            if (--empty_visits == 0) goto LDone;
-            if ((size_t)d->rehashidx >= d->ht[0].size) goto LDone;
-        }
-
-        de = d->ht[0].table[d->rehashidx];
-        // We have to queue every node in the bucket, even if we go over our target size
-        while (de != nullptr) {
-            d->asyncdata->queue.emplace_back(de);
-            de = de->next;
-        }
-        d->rehashidx++;
-    }
-
-LDone:
-    if (d->asyncdata->queue.empty()) {
-        // We didn't find any work to do (can happen if there is a large gap in the hash table)
-        auto asyncT = d->asyncdata;
-        d->asyncdata = d->asyncdata->next;
-        delete asyncT;
-        return nullptr;
-    }
-    return d->asyncdata;
-}
-
-void dictRehashAsync(dictAsyncRehashCtl *ctl) {
-    if (ctl->abondon.load(std::memory_order_acquire)) {
-        ctl->hashIdx = ctl->queue.size();
-    }
-    for (size_t idx = ctl->hashIdx; idx < ctl->queue.size(); ++idx) {
-        auto &wi = ctl->queue[idx];
-        wi.hash = dictHashKey(ctl->dict, dictGetKey(wi.de));
-    }
-    ctl->hashIdx = ctl->queue.size();
-    ctl->done = true;
-}
-
-bool dictRehashSomeAsync(dictAsyncRehashCtl *ctl, size_t hashes) {
-    if (ctl->abondon.load(std::memory_order_acquire)) {
-        ctl->hashIdx = ctl->queue.size();
-    }
-    size_t max = std::min(ctl->hashIdx + hashes, ctl->queue.size());
-    for (; ctl->hashIdx < max; ++ctl->hashIdx) {
-        auto &wi = ctl->queue[ctl->hashIdx];
-        wi.hash = dictHashKey(ctl->dict, dictGetKey(wi.de));
-    }
-
-    if (ctl->hashIdx == ctl->queue.size()) ctl->done = true;
-    return ctl->hashIdx < ctl->queue.size();
-}
-
-
-void discontinueAsyncRehash(dict *d) {
-    // We inform our async rehashers and the completion function the results are to be
-    //  abandoned.  We keep the asyncdata linked in so that dictEntry's are still added
-    //  to the GC list.  This is because we can't gurantee when the other threads will
-    //  stop looking at them.
-    if (d->asyncdata != nullptr) {
-        auto adata = d->asyncdata;
-        while (adata != nullptr && !adata->abondon.load(std::memory_order_relaxed)) {
-            adata->abondon = true;
-            adata = adata->next;
-        }
-        if (dictIsRehashing(d))
-            d->rehashidx = 0;
-    }
-}
-
-void dictCompleteRehashAsync(dictAsyncRehashCtl *ctl, bool fFree) {
-    dict *d = ctl->dict;
-    assert(ctl->done);
-    
-    // Unlink ourselves
-    bool fUnlinked = false;
-    dictAsyncRehashCtl **pprev = &d->asyncdata;
-
-    while (*pprev != nullptr) {
-        if (*pprev == ctl) {
-            *pprev = ctl->next;
-            fUnlinked = true;
-            break;
-        }
-        pprev = &((*pprev)->next);
-    }
-
-    if (fUnlinked) {
-        if (ctl->next != nullptr && ctl->deGCList != nullptr) {
-            // An older work item may depend on our free list, so pass our free list to them
-            dictEntry **deGC = &ctl->next->deGCList;
-            while (*deGC != nullptr) deGC = &((*deGC)->next);
-            *deGC = ctl->deGCList;
-            ctl->deGCList = nullptr;
-        }
-    }
-
-    if (fUnlinked && !ctl->abondon) {
-        if (d->ht[0].table != nullptr) {    // can be null if we're cleared during the rehash
-            for (auto &wi : ctl->queue) {
-                // We need to remove it from the source hash table, and store it in the dest.
-                //  Note that it may have been deleted in the meantime and therefore not exist.
-                //  In this case it will be in the garbage collection list
-                
-                dictEntry **pdePrev = &d->ht[0].table[wi.hash & d->ht[0].sizemask];
-                while (*pdePrev != nullptr && *pdePrev != wi.de) {
-                    pdePrev = &((*pdePrev)->next);
-                }
-                if (*pdePrev != nullptr) {  // The element may be NULL if its in the GC list
-                    assert(*pdePrev == wi.de);
-                    *pdePrev = wi.de->next;
-                    // Now link it to the dest hash table
-                    wi.de->next = d->ht[1].table[wi.hash & d->ht[1].sizemask];
-                    d->ht[1].table[wi.hash & d->ht[1].sizemask] = wi.de;
-                    d->ht[0].used--;
-                    d->ht[1].used++;
-                }
-            }
-        }
-
-        /* Check if we already rehashed the whole table... */
-        if (d->ht[0].used == 0 && d->asyncdata == nullptr) {
-            asyncFreeDictTable(d->ht[0].table);
-            d->ht[0] = d->ht[1];
-            _dictReset(&d->ht[1]);
-            d->rehashidx = -1;
-        }
-    }
-
-    if (fFree) {
-        d->type->asyncfree(ctl);
-    }
-}
-
 long long timeInMilliseconds(void) {
     struct timeval tv;
 
@@ -560,18 +260,6 @@ long long timeInMilliseconds(void) {
     return (((long long)tv.tv_sec)*1000)+(tv.tv_usec/1000);
 }
 
-dictAsyncRehashCtl::~dictAsyncRehashCtl() {
-    while (deGCList != nullptr) {
-        auto next = deGCList->next;
-        dictFreeKey(dict, deGCList);
-        if (deGCList->v.val != nullptr)
-            dictFreeVal(dict, deGCList);
-        zfree(deGCList);
-        deGCList = next;
-    }
-    dictRelease(this->dict);
-}
-
 /* Rehash in ms+"delta" milliseconds. The value of "delta" is larger 
  * than 0, and is smaller than 1 in most cases. The exact upper bound 
  * depends on the running time of dictRehash(d,100).*/
@@ -581,8 +269,8 @@ int dictRehashMilliseconds(dict *d, int ms) {
     long long start = timeInMilliseconds();
     int rehashes = 0;
 
-    while(dictRehash(d,1000)) {
-        rehashes += 1000;
+    while(dictRehash(d,100)) {
+        rehashes += 100;
         if (timeInMilliseconds()-start > ms) break;
     }
     return rehashes;
@@ -597,15 +285,13 @@ int dictRehashMilliseconds(dict *d, int ms) {
  * dictionary so that the hash table automatically migrates from H1 to H2
  * while it is actively used. */
 static void _dictRehashStep(dict *d) {
-    int16_t pauserehash;
-    __atomic_load(&d->pauserehash, &pauserehash, __ATOMIC_RELAXED);
-    if (pauserehash == 0) dictRehash(d,1);
+    if (d->pauserehash == 0) dictRehash(d,1);
 }
 
 /* Add an element to the target hash table */
-int dictAdd(dict *d, void *key, void *val, dictEntry **existing)
+int dictAdd(dict *d, void *key, void *val)
 {
-    dictEntry *entry = dictAddRaw(d,key,existing);
+    dictEntry *entry = dictAddRaw(d,key,NULL);
 
     if (!entry) return DICT_ERR;
     dictSetVal(d, entry, val);
@@ -707,9 +393,6 @@ static dictEntry *dictGenericDelete(dict *d, const void *key, int nofree) {
     dictEntry *he, *prevHe;
     int table;
 
-    // if we are deleting elements we probably aren't mass inserting anymore and it is safe to shrink
-    d->noshrink = false;
-
     if (d->ht[0].used == 0 && d->ht[1].used == 0) return NULL;
 
     if (dictIsRehashing(d)) _dictRehashStep(d);
@@ -727,19 +410,11 @@ static dictEntry *dictGenericDelete(dict *d, const void *key, int nofree) {
                 else
                     d->ht[table].table[idx] = he->next;
                 if (!nofree) {
-                    if (table == 0 && d->asyncdata != nullptr && (ssize_t)idx < d->rehashidx) {
-                        dictFreeVal(d, he);
-                        he->v.val = nullptr;
-                        he->next = d->asyncdata->deGCList;
-                        d->asyncdata->deGCList = he;
-                    } else {
-                        dictFreeKey(d, he);
-                        dictFreeVal(d, he);
-                        zfree(he);
-                    }
+                    dictFreeKey(d, he);
+                    dictFreeVal(d, he);
+                    zfree(he);
                 }
                 d->ht[table].used--;
-                _dictExpandIfNeeded(d);
                 return he;
             }
             prevHe = he;
@@ -747,7 +422,6 @@ static dictEntry *dictGenericDelete(dict *d, const void *key, int nofree) {
         }
         if (!dictIsRehashing(d)) break;
     }
-    _dictExpandIfNeeded(d);
     return NULL; /* not found */
 }
 
@@ -787,16 +461,9 @@ dictEntry *dictUnlink(dict *ht, const void *key) {
 void dictFreeUnlinkedEntry(dict *d, dictEntry *he) {
     if (he == NULL) return;
 
-    if (d->asyncdata) {
-        dictFreeVal(d, he);
-        he->v.val = nullptr;
-        he->next = d->asyncdata->deGCList;
-        d->asyncdata->deGCList = he;
-    } else { 
-        dictFreeKey(d, he);
-        dictFreeVal(d, he);
-        zfree(he);
-    }
+    dictFreeKey(d, he);
+    dictFreeVal(d, he);
+    zfree(he);
 }
 
 /* Destroy an entire dictionary */
@@ -810,26 +477,17 @@ int _dictClear(dict *d, dictht *ht, void(callback)(void *)) {
         if (callback && (i & 65535) == 0) callback(d->privdata);
 
         if ((he = ht->table[i]) == NULL) continue;
-        dictEntry *deNull = nullptr;
-        __atomic_store(&ht->table[i], &deNull, __ATOMIC_RELEASE);
         while(he) {
             nextHe = he->next;
-            if (d->asyncdata && (ssize_t)i < d->rehashidx) {
-                dictFreeVal(d, he);
-                he->v.val = nullptr;
-                he->next = d->asyncdata->deGCList;
-                d->asyncdata->deGCList = he;
-            } else {
-                dictFreeKey(d, he);
-                dictFreeVal(d, he);
-                zfree(he);
-            }
+            dictFreeKey(d, he);
+            dictFreeVal(d, he);
+            zfree(he);
             ht->used--;
             he = nextHe;
         }
     }
     /* Free the table and the allocated cache structure */
-    asyncFreeDictTable(ht->table);
+    zfree(ht->table);
     /* Re-initialize the table */
     _dictReset(ht);
     return DICT_OK; /* never fails */
@@ -838,30 +496,25 @@ int _dictClear(dict *d, dictht *ht, void(callback)(void *)) {
 /* Clear & Release the hash table */
 void dictRelease(dict *d)
 {
-    if (__atomic_sub_fetch(&d->refcount, 1, __ATOMIC_ACQ_REL) == 0) {
-        _dictClear(d,&d->ht[0],NULL);
-        _dictClear(d,&d->ht[1],NULL);
-        zfree(d);
-    }
+    _dictClear(d,&d->ht[0],NULL);
+    _dictClear(d,&d->ht[1],NULL);
+    zfree(d);
 }
 
-dictEntry *dictFindWithPrev(dict *d, const void *key, uint64_t h, dictEntry ***dePrevPtr, dictht **pht, bool fShallowCompare)
+dictEntry *dictFind(dict *d, const void *key)
 {
     dictEntry *he;
-    uint64_t idx, table;
+    uint64_t h, idx, table;
 
     if (dictSize(d) == 0) return NULL; /* dict is empty */
     if (dictIsRehashing(d)) _dictRehashStep(d);
+    h = dictHashKey(d, key);
     for (table = 0; table <= 1; table++) {
-        *pht = d->ht + table;
         idx = h & d->ht[table].sizemask;
         he = d->ht[table].table[idx];
-        *dePrevPtr = &d->ht[table].table[idx];
         while(he) {
-            if (key==he->key || (!fShallowCompare && dictCompareKeys(d, key, he->key))) {
+            if (key==he->key || dictCompareKeys(d, key, he->key))
                 return he;
-            }
-            *dePrevPtr = &he->next;
             he = he->next;
         }
         if (!dictIsRehashing(d)) return NULL;
@@ -869,14 +522,6 @@ dictEntry *dictFindWithPrev(dict *d, const void *key, uint64_t h, dictEntry ***d
     return NULL;
 }
 
-dictEntry *dictFind(dict *d, const void *key)
-{
-    dictEntry **deT;
-    dictht *ht;
-    uint64_t h = dictHashKey(d, key);
-    return dictFindWithPrev(d, key, h, &deT, &ht);
-}
-
 void *dictFetchValue(dict *d, const void *key) {
     dictEntry *he;
 
@@ -999,16 +644,10 @@ dictEntry *dictGetRandomKey(dict *d)
     if (dictSize(d) == 0) return NULL;
     if (dictIsRehashing(d)) _dictRehashStep(d);
     if (dictIsRehashing(d)) {
-        long rehashidx = d->rehashidx;
-        auto async = d->asyncdata;
-        while (async != nullptr) {
-            rehashidx = std::min((long)async->rehashIdxBase, rehashidx);
-            async = async->next;
-        }
         do {
             /* We are sure there are no elements in indexes from 0
              * to rehashidx-1 */
-            h = rehashidx + (randomULong() % (dictSlots(d) - rehashidx));
+            h = d->rehashidx + (randomULong() % (dictSlots(d) - d->rehashidx));
             he = (h >= d->ht[0].size) ? d->ht[1].table[h - d->ht[0].size] :
                                       d->ht[0].table[h];
         } while(he == NULL);
@@ -1347,12 +986,11 @@ static int dictTypeExpandAllowed(dict *d) {
 /* Expand the hash table if needed */
 static int _dictExpandIfNeeded(dict *d)
 {
-    static const size_t SHRINK_FACTOR = 4;
     /* Incremental rehashing already in progress. Return. */
     if (dictIsRehashing(d)) return DICT_OK;
 
     /* If the hash table is empty expand it to the initial size. */
-    if (d->ht[0].size == 0) return _dictExpand(d, DICT_HT_INITIAL_SIZE, false /*fShrink*/, nullptr);
+    if (d->ht[0].size == 0) return dictExpand(d, DICT_HT_INITIAL_SIZE);
 
     /* If we reached the 1:1 ratio, and we are allowed to resize the hash
      * table (global setting) or we should avoid it but the ratio between
@@ -1363,12 +1001,7 @@ static int _dictExpandIfNeeded(dict *d)
          d->ht[0].used/d->ht[0].size > dict_force_resize_ratio) &&
         dictTypeExpandAllowed(d))
     {
-        return _dictExpand(d, d->ht[0].used + 1, false /*fShrink*/, nullptr);
-    }
-    else if (d->ht[0].used > 0 && d->ht[0].size >= (1024*SHRINK_FACTOR) && (d->ht[0].used * 16) < d->ht[0].size && dict_can_resize && !d->noshrink)
-    {
-        // If the dictionary has shurnk a lot we'll need to shrink the hash table instead
-        return _dictExpand(d, d->ht[0].size/SHRINK_FACTOR, true /*fShrink*/, nullptr);
+        return dictExpand(d, d->ht[0].used + 1);
     }
     return DICT_OK;
 }
@@ -1393,7 +1026,7 @@ 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. */
-long _dictKeyIndex(dict *d, const void *key, uint64_t hash, dictEntry **existing)
+static long _dictKeyIndex(dict *d, const void *key, uint64_t hash, dictEntry **existing)
 {
     unsigned long idx, table;
     dictEntry *he;
@@ -1541,13 +1174,6 @@ void dictGetStats(char *buf, size_t bufsize, dict *d) {
     if (orig_bufsize) orig_buf[orig_bufsize-1] = '\0';
 }
 
-void dictForceRehash(dict *d)
-{
-    int16_t pauserehash;
-    __atomic_load(&d->pauserehash, &pauserehash, __ATOMIC_RELAXED);
-    while (pauserehash == 0 && dictIsRehashing(d)) _dictRehashStep(d);
-}
-
 /* ------------------------------- Benchmark ---------------------------------*/
 
 #ifdef REDIS_TEST
@@ -1577,7 +1203,7 @@ char *stringFromLongLong(long long value) {
     int len;
     char *s;
 
-    len = snprintf(buf,32,"%lld",value);
+    len = sprintf(buf,"%lld",value);
     s = zmalloc(len+1);
     memcpy(s, buf, len);
     s[len] = '\0';

src/dict.h

@@ -56,7 +56,6 @@ extern "C" {
 
 /* Unused arguments generate annoying warnings... */
 #define DICT_NOTUSED(V) ((void) V)
-struct dictAsyncRehashCtl;
 
 typedef struct dictEntry {
     void *key;
@@ -77,7 +76,6 @@ typedef struct dictType {
     void (*keyDestructor)(void *privdata, void *key);
     void (*valDestructor)(void *privdata, void *obj);
     int (*expandAllowed)(size_t moreMem, double usedRatio);
-    void (*asyncfree)(dictAsyncRehashCtl *);
 } dictType;
 
 /* This is our hash table structure. Every dictionary has two of this as we
@@ -104,49 +102,24 @@ struct dictAsyncRehashCtl {
     struct dict *dict = nullptr;
     std::vector<workItem> queue;
     size_t hashIdx = 0;
-    long rehashIdxBase;
+    bool release = false;
     dictAsyncRehashCtl *next = nullptr;
     std::atomic<bool> done { false };
-    std::atomic<bool> abondon { false };
 
-    dictAsyncRehashCtl(struct dict *d, dictAsyncRehashCtl *next);
-    dictAsyncRehashCtl(const dictAsyncRehashCtl&) = delete;
-    dictAsyncRehashCtl(dictAsyncRehashCtl&&) = delete;
-    ~dictAsyncRehashCtl();
+    dictAsyncRehashCtl(struct dict *d, dictAsyncRehashCtl *next) : dict(d), next(next) {
+        queue.reserve(c_targetQueueSize);
+    }
 };
 #else
 struct  dictAsyncRehashCtl;
 #endif
 
-void discontinueAsyncRehash(dict *d);
-
 typedef struct dict {
     dictType *type;
     void *privdata;
     dictht ht[2];
     long rehashidx; /* rehashing not in progress if rehashidx == -1 */
-    unsigned refcount;
-    dictAsyncRehashCtl *asyncdata;
     int16_t pauserehash; /* If >0 rehashing is paused (<0 indicates coding error) */
-    uint8_t noshrink = false;
-
-#ifdef __cplusplus
-    dict() = default;
-    dict(dict &) = delete;  // No Copy Ctor
-
-    static void swap(dict& a, dict& b) {
-        discontinueAsyncRehash(&a);
-        discontinueAsyncRehash(&b);
-        std::swap(a.type, b.type);
-        std::swap(a.privdata, b.privdata);
-        std::swap(a.ht[0], b.ht[0]);
-        std::swap(a.ht[1], b.ht[1]);
-        std::swap(a.rehashidx, b.rehashidx);
-        // Never swap refcount - they are attached to the specific dict obj
-        std::swap(a.pauserehash, b.pauserehash);
-        std::swap(a.noshrink, b.noshrink);
-    }
-#endif
 } dict;
 
 /* If safe is set to 1 this is a safe iterator, that means, you can call
@@ -214,8 +187,8 @@ typedef void (dictScanBucketFunction)(void *privdata, dictEntry **bucketref);
 #define dictSlots(d) ((d)->ht[0].size+(d)->ht[1].size)
 #define dictSize(d) ((d)->ht[0].used+(d)->ht[1].used)
 #define dictIsRehashing(d) ((d)->rehashidx != -1)
-#define dictPauseRehashing(d) __atomic_fetch_add(&(d)->pauserehash, 1, __ATOMIC_SEQ_CST)
-#define dictResumeRehashing(d) __atomic_fetch_sub(&(d)->pauserehash, 1, __ATOMIC_SEQ_CST)
+#define dictPauseRehashing(d) (d)->pauserehash++
+#define dictResumeRehashing(d) (d)->pauserehash--
 
 /* If our unsigned long type can store a 64 bit number, use a 64 bit PRNG. */
 #if ULONG_MAX >= 0xffffffffffffffff
@@ -226,9 +199,9 @@ typedef void (dictScanBucketFunction)(void *privdata, dictEntry **bucketref);
 
 /* API */
 dict *dictCreate(dictType *type, void *privDataPtr);
-int dictExpand(dict *d, unsigned long size, bool fShrink = false);
-int dictTryExpand(dict *d, unsigned long size, bool fShrink);
-int dictAdd(dict *d, void *key, void *val, dictEntry **existing = nullptr);
+int dictExpand(dict *d, unsigned long size);
+int dictTryExpand(dict *d, unsigned long size);
+int dictAdd(dict *d, void *key, void *val);
 dictEntry *dictAddRaw(dict *d, void *key, dictEntry **existing);
 dictEntry *dictAddOrFind(dict *d, void *key);
 int dictReplace(dict *d, void *key, void *val);
@@ -237,7 +210,6 @@ dictEntry *dictUnlink(dict *ht, const void *key);
 void dictFreeUnlinkedEntry(dict *d, dictEntry *he);
 void dictRelease(dict *d);
 dictEntry * dictFind(dict *d, const void *key);
-dictEntry * dictFindWithPrev(dict *d, const void *key, uint64_t h, dictEntry ***dePrevPtr, dictht **ht, bool fShallowCompare = false);
 void *dictFetchValue(dict *d, const void *key);
 int dictResize(dict *d);
 dictIterator *dictGetIterator(dict *d);
@@ -260,14 +232,6 @@ uint8_t *dictGetHashFunctionSeed(void);
 unsigned long dictScan(dict *d, unsigned long v, dictScanFunction *fn, dictScanBucketFunction *bucketfn, void *privdata);
 uint64_t dictGetHash(dict *d, const void *key);
 dictEntry **dictFindEntryRefByPtrAndHash(dict *d, const void *oldptr, uint64_t hash);
-void dictForceRehash(dict *d);
-int dictMerge(dict *dst, dict *src);
-
-/* Async API */
-dictAsyncRehashCtl *dictRehashAsyncStart(dict *d, int buckets);
-bool dictRehashSomeAsync(dictAsyncRehashCtl *ctl, size_t hashes);
-void dictCompleteRehashAsync(dictAsyncRehashCtl *ctl, bool fFree);
-void dictRehashAsync(dictAsyncRehashCtl *ctl);
 
 /* Hash table types */
 extern dictType dictTypeHeapStringCopyKey;

src/endianconv.c

@@ -112,15 +112,15 @@ int endianconvTest(int argc, char *argv[], int accurate) {
     UNUSED(argv);
     UNUSED(accurate);
 
-    snprintf(buf,32,"ciaoroma");
+    sprintf(buf,"ciaoroma");
     memrev16(buf);
     printf("%s\n", buf);
 
-    snprintf(buf,32,"ciaoroma");
+    sprintf(buf,"ciaoroma");
     memrev32(buf);
     printf("%s\n", buf);
 
-    snprintf(buf,32,"ciaoroma");
+    sprintf(buf,"ciaoroma");
     memrev64(buf);
     printf("%s\n", buf);
 

src/evict.cpp

@@ -34,7 +34,6 @@
 #include "bio.h"
 #include "atomicvar.h"
 #include <mutex>
-#include <map>
 #include <math.h>
 
 /* ----------------------------------------------------------------------------
@@ -227,47 +226,32 @@ struct visitFunctor
     int dbid;
     dict *dbdict;
     struct evictionPoolEntry *pool;
-    int count = 0;
-    int tries = 0;
+    int count;
 
     bool operator()(const expireEntry &e)
     {
         dictEntry *de = dictFind(dbdict, e.key());
-        if (de != nullptr)
-        {
-            processEvictionCandidate(dbid, (sds)dictGetKey(de), (robj*)dictGetVal(de), &e, pool);
-            ++count;
-        }
-        ++tries;
-        return tries < g_pserver->maxmemory_samples;
+        processEvictionCandidate(dbid, (sds)dictGetKey(de), (robj*)dictGetVal(de), &e, pool);
+        ++count;
+        return count < g_pserver->maxmemory_samples;
     }
 };
-int evictionPoolPopulate(int dbid, redisDb *db, expireset *setexpire, struct evictionPoolEntry *pool)
+void evictionPoolPopulate(int dbid, dict *dbdict, expireset *setexpire, struct evictionPoolEntry *pool)
 {
     if (setexpire != nullptr)
     {
-        std::unique_lock<fastlock> ul(g_expireLock);
-        visitFunctor visitor { dbid, db->dictUnsafeKeyOnly(), pool, 0 };
+        visitFunctor visitor { dbid, dbdict, pool, 0 };
         setexpire->random_visit(visitor);
-        return visitor.count;
     }
     else
     {
-        int returnCount = 0;
         dictEntry **samples = (dictEntry**)alloca(g_pserver->maxmemory_samples * sizeof(dictEntry*));
-        int count = dictGetSomeKeys(db->dictUnsafeKeyOnly(),samples,g_pserver->maxmemory_samples);
+        int count = dictGetSomeKeys(dbdict,samples,g_pserver->maxmemory_samples);
         for (int j = 0; j < count; j++) {
             robj *o = (robj*)dictGetVal(samples[j]);
-            // If the object is in second tier storage we don't need to evict it (since it alrady is)
-            if (o != nullptr)
-            {
-                processEvictionCandidate(dbid, (sds)dictGetKey(samples[j]), o, nullptr, pool);
-                ++returnCount;
-            }
+            processEvictionCandidate(dbid, (sds)dictGetKey(samples[j]), o, nullptr, pool);
         }
-        return returnCount;
     }
-    return 0;
 }
 
 /* ----------------------------------------------------------------------------
@@ -346,7 +330,7 @@ uint8_t LFULogIncr(uint8_t 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. */
-unsigned long LFUDecrAndReturn(robj_roptr o) {
+unsigned long LFUDecrAndReturn(robj *o) {
     unsigned long ldt = o->lru >> 8;
     unsigned long counter = o->lru & 255;
     unsigned long num_periods = g_pserver->lfu_decay_time ? LFUTimeElapsed(ldt) / g_pserver->lfu_decay_time : 0;
@@ -355,8 +339,6 @@ unsigned long LFUDecrAndReturn(robj_roptr o) {
     return counter;
 }
 
-unsigned long getClientReplicationBacklogSharedUsage(client *c);
-
 /* We don't want to count AOF buffers and slaves output buffers as
  * used memory: the eviction should use mostly data size. This function
  * returns the sum of AOF and slaves buffer. */
@@ -373,15 +355,9 @@ size_t freeMemoryGetNotCountedMemory(void) {
         while((ln = listNext(&li))) {
             client *replica = (client*)listNodeValue(ln);
             std::unique_lock<fastlock>(replica->lock);
-            /* we don't wish to multiple count the replication backlog shared usage */
-            overhead += (getClientOutputBufferMemoryUsage(replica) - getClientReplicationBacklogSharedUsage(replica));
+            overhead += getClientOutputBufferMemoryUsage(replica);
         }
     }
-
-    /* also don't count the replication backlog memory
-     * that's where the replication clients get their memory from */
-    overhead += g_pserver->repl_backlog_size;
-
     if (g_pserver->aof_state != AOF_OFF) {
         overhead += sdsalloc(g_pserver->aof_buf)+aofRewriteBufferSize();
     }
@@ -412,21 +388,16 @@ size_t freeMemoryGetNotCountedMemory(void) {
  *              limit.
  *              (Populated both for C_ERR and C_OK)
  */
-int getMaxmemoryState(size_t *total, size_t *logical, size_t *tofree, float *level, bool fQuickCycle, bool fPreSnapshot) {
+int getMaxmemoryState(size_t *total, size_t *logical, size_t *tofree, float *level) {
     size_t mem_reported, mem_used, mem_tofree;
 
     /* Check if we are over the memory usage limit. If we are not, no need
      * to subtract the slaves output buffers. We can just return ASAP. */
     mem_reported = zmalloc_used_memory();
     if (total) *total = mem_reported;
-    size_t maxmemory = g_pserver->maxmemory;
-    if (fPreSnapshot)
-        maxmemory = static_cast<size_t>(maxmemory * 0.9);   // derate memory by 10% since we won't be able to free during snapshot
-    if (g_pserver->FRdbSaveInProgress())
-        maxmemory = static_cast<size_t>(maxmemory*1.2);
 
     /* We may return ASAP if there is no need to compute the level. */
-    int return_ok_asap = !maxmemory || mem_reported <= maxmemory;
+    int return_ok_asap = !g_pserver->maxmemory || mem_reported <= g_pserver->maxmemory;
     if (return_ok_asap && !level) return C_OK;
 
     /* Remove the size of slaves output buffers and AOF buffer from the
@@ -437,24 +408,20 @@ int getMaxmemoryState(size_t *total, size_t *logical, size_t *tofree, float *lev
 
     /* Compute the ratio of memory usage. */
     if (level) {
-        if (!maxmemory) {
+        if (!g_pserver->maxmemory) {
             *level = 0;
         } else {
-            *level = (float)mem_used / (float)maxmemory;
+            *level = (float)mem_used / (float)g_pserver->maxmemory;
         }
     }
 
     if (return_ok_asap) return C_OK;
 
     /* Check if we are still over the memory limit. */
-    if (mem_used <= maxmemory) return C_OK;
+    if (mem_used <= g_pserver->maxmemory) return C_OK;
 
     /* Compute how much memory we need to free. */
-    mem_tofree = mem_used - maxmemory;
-    if (g_pserver->m_pstorageFactory && !fQuickCycle)
-    {
-        mem_tofree += static_cast<size_t>(maxmemory * 0.05); // if we have a storage provider be much more aggressive
-    }
+    mem_tofree = mem_used - g_pserver->maxmemory;
 
     if (logical) *logical = mem_used;
     if (tofree) *tofree = mem_tofree;
@@ -462,69 +429,6 @@ int getMaxmemoryState(size_t *total, size_t *logical, size_t *tofree, float *lev
     return C_ERR;
 }
 
-class FreeMemoryLazyFree : public ICollectable
-{    
-    ssize_t m_cb = 0;
-    std::vector<std::pair<dict*, std::vector<dictEntry*>>> vecdictvecde;
-
-public:
-    static std::atomic<int> s_clazyFreesInProgress;
-
-    FreeMemoryLazyFree() {
-        s_clazyFreesInProgress++;
-    }
-
-    FreeMemoryLazyFree(const FreeMemoryLazyFree&) = delete;
-    FreeMemoryLazyFree(FreeMemoryLazyFree&&) = default;
-
-    ~FreeMemoryLazyFree() {
-        aeAcquireLock();
-        for (auto &pair : vecdictvecde) {
-            for (auto de : pair.second) {
-                dictFreeUnlinkedEntry(pair.first, de);
-            }
-        }
-        aeReleaseLock();
-        --s_clazyFreesInProgress;
-    }
-
-    ssize_t addEntry(dict *d, dictEntry *de) {
-        ssize_t cbFreedNow = 0;
-        ssize_t cb = sizeof(dictEntry);
-        cb += sdsAllocSize((sds)dictGetKey(de));
-        robj *o = (robj*)dictGetVal(de);
-        switch (o->type) {
-        case OBJ_STRING:
-            cb += getStringObjectSdsUsedMemory(o)+sizeof(robj);
-            break;
-
-        default:
-            // If we don't know about it we can't accurately track the memory so free now
-            cbFreedNow = zmalloc_used_memory();
-            decrRefCount(o);
-            cbFreedNow -= zmalloc_used_memory();
-            de->v.val = nullptr;
-        }
-
-        auto itr = std::lower_bound(vecdictvecde.begin(), vecdictvecde.end(), d, 
-            [](const std::pair<dict*, std::vector<dictEntry*>> &a, dict *d) -> bool {
-                return a.first < d;
-            }
-        );
-        if (itr == vecdictvecde.end() || itr->first != d) {
-            itr = vecdictvecde.insert(itr, std::make_pair(d, std::vector<dictEntry*>()));
-        }
-        serverAssert(itr->first == d);
-        itr->second.push_back(de);
-        m_cb += cb;
-        return cb + cbFreedNow;
-    }
-
-    size_t memory_queued() { return m_cb; }
-};
-
-std::atomic<int> FreeMemoryLazyFree::s_clazyFreesInProgress {0};
-
 /* Return 1 if used memory is more than maxmemory after allocating more memory,
  * return 0 if not. Redis may reject user's requests or evict some keys if used
  * memory exceeds maxmemory, especially, when we allocate huge memory at once. */
@@ -550,9 +454,8 @@ static int evictionTimeProc(
     UNUSED(eventLoop);
     UNUSED(id);
     UNUSED(clientData);
-    serverAssert(GlobalLocksAcquired());
 
-    if (performEvictions(false) == EVICT_RUNNING) return 0;  /* keep evicting */
+    if (performEvictions() == EVICT_RUNNING) return 0;  /* keep evicting */
 
     /* For EVICT_OK - things are good, no need to keep evicting.
      * For EVICT_FAIL - there is nothing left to evict.  */
@@ -567,14 +470,11 @@ static int evictionTimeProc(
 static int isSafeToPerformEvictions(void) {
     /* - There must be no script in timeout condition.
      * - Nor we are loading data right now.  */
-    if (g_pserver->shutdown_asap || g_pserver->lua_timedout || g_pserver->loading) return 0;
+    if (g_pserver->lua_timedout || g_pserver->loading) return 0;
 
     /* By default replicas should ignore maxmemory
      * and just be masters exact copies. */
-    if (g_pserver->m_pstorageFactory == nullptr && listLength(g_pserver->masters) && g_pserver->repl_slave_ignore_maxmemory && !g_pserver->fActiveReplica) return 0;
-
-    /* If we have a lazy free obj pending, our amounts will be off, wait for it to go away */
-    if (FreeMemoryLazyFree::s_clazyFreesInProgress > 0) return 0;
+    if (listLength(g_pserver->masters) && g_pserver->repl_slave_ignore_maxmemory && !g_pserver->fActiveReplica) return 0;
 
     /* 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
@@ -626,7 +526,7 @@ static unsigned long evictionTimeLimitUs() {
  *   EVICT_RUNNING  - memory is over the limit, but eviction is still processing
  *   EVICT_FAIL     - memory is over the limit, and there's nothing to evict
  * */
-int performEvictions(bool fPreSnapshot) {
+int performEvictions(void) {
     if (!isSafeToPerformEvictions()) return EVICT_OK;
     serverAssert(GlobalLocksAcquired());
 
@@ -636,13 +536,9 @@ int performEvictions(bool fPreSnapshot) {
     mstime_t latency, eviction_latency;
     long long delta;
     int slaves = listLength(g_pserver->slaves);
-    const bool fEvictToStorage = !cserver.delete_on_evict && g_pserver->db[0]->FStorageProvider();
     int result = EVICT_FAIL;
-    int ckeysFailed = 0;
 
-    std::unique_ptr<FreeMemoryLazyFree> splazy = std::make_unique<FreeMemoryLazyFree>();
-
-    if (getMaxmemoryState(&mem_reported,NULL,&mem_tofree,NULL,false,fPreSnapshot) == C_OK)
+    if (getMaxmemoryState(&mem_reported,NULL,&mem_tofree,NULL) == C_OK)
         return EVICT_OK;
 
     if (g_pserver->maxmemory_policy == MAXMEMORY_NO_EVICTION)
@@ -657,17 +553,13 @@ int performEvictions(bool fPreSnapshot) {
     monotime evictionTimer;
     elapsedStart(&evictionTimer);
 
-    if (g_pserver->maxstorage && g_pserver->m_pstorageFactory != nullptr && g_pserver->m_pstorageFactory->totalDiskspaceUsed() >= g_pserver->maxstorage)
-        goto cant_free_storage;
-
     while (mem_freed < (long long)mem_tofree) {
         int j, k, i;
         static unsigned int next_db = 0;
         sds bestkey = NULL;
         int bestdbid;
         redisDb *db;
-        bool fFallback = false;
-        
+
         if (g_pserver->maxmemory_policy & (MAXMEMORY_FLAG_LRU|MAXMEMORY_FLAG_LFU) ||
             g_pserver->maxmemory_policy == MAXMEMORY_VOLATILE_TTL)
         {
@@ -680,33 +572,33 @@ int performEvictions(bool fPreSnapshot) {
                  * so to start populate the eviction pool sampling keys from
                  * every DB. */
                 for (i = 0; i < cserver.dbnum; i++) {
-                    db = g_pserver->db[i];
+                    db = g_pserver->db+i;
                     if (g_pserver->maxmemory_policy & MAXMEMORY_FLAG_ALLKEYS)
                     {
-                        if ((keys = db->size()) != 0) {
-                            total_keys += evictionPoolPopulate(i, db, nullptr, pool);
+                        if ((keys = dictSize(db->dict)) != 0) {
+                            evictionPoolPopulate(i, db->dict, nullptr, pool);
+                            total_keys += keys;
                         }
                     }
                     else
                     {
-                        keys = db->expireSize();
+                        keys = db->setexpire->size();
                         if (keys != 0)
-                            total_keys += evictionPoolPopulate(i, db, db->setexpireUnsafe(), pool);
+                            evictionPoolPopulate(i, db->dict, db->setexpire, pool);
+                        total_keys += keys;
                     }
                 }
                 if (!total_keys) break; /* No keys to evict. */
 
                 /* Go backward from best to worst element to evict. */
                 for (k = EVPOOL_SIZE-1; k >= 0; k--) {
-                    if (pool[k].key == NULL) {
-                        continue;
-                    } 
+                    if (pool[k].key == NULL) continue;
                     bestdbid = pool[k].dbid;
                     sds key = nullptr;
 
-                    auto itr = g_pserver->db[pool[k].dbid]->find(pool[k].key);
-                    if (itr != nullptr && (g_pserver->maxmemory_policy & MAXMEMORY_FLAG_ALLKEYS || itr.val()->FExpires()))
-                        key = itr.key();
+                    dictEntry *de = dictFind(g_pserver->db[pool[k].dbid].dict,pool[k].key);
+                    if (de != nullptr && (g_pserver->maxmemory_policy & MAXMEMORY_FLAG_ALLKEYS || ((robj*)dictGetVal(de))->FExpires()))
+                        key = (sds)dictGetKey(de);
 
                     /* Remove the entry from the pool. */
                     if (pool[k].key != pool[k].cached)
@@ -724,35 +616,32 @@ int performEvictions(bool fPreSnapshot) {
                     }
                 }
             }
-            if (bestkey == nullptr && fEvictToStorage)
-                fFallback = true;
         }
 
         /* volatile-random and allkeys-random policy */
-        if (g_pserver->maxmemory_policy == MAXMEMORY_ALLKEYS_RANDOM ||
-                 g_pserver->maxmemory_policy == MAXMEMORY_VOLATILE_RANDOM
-                 || fFallback)
+        else if (g_pserver->maxmemory_policy == MAXMEMORY_ALLKEYS_RANDOM ||
+                 g_pserver->maxmemory_policy == MAXMEMORY_VOLATILE_RANDOM)
         {
             /* When evicting a random key, we try to evict a key for
              * each DB, so we use the static 'next_db' variable to
              * incrementally visit all DBs. */
             for (i = 0; i < cserver.dbnum; i++) {
                 j = (++next_db) % cserver.dbnum;
-                db = g_pserver->db[j];
-                if (g_pserver->maxmemory_policy == MAXMEMORY_ALLKEYS_RANDOM || fFallback)
+                db = g_pserver->db+j;
+                if (g_pserver->maxmemory_policy == MAXMEMORY_ALLKEYS_RANDOM)
                 {
-                    if (db->size() != 0) {
-                        auto itr = db->random_cache_threadsafe(true /*fPrimaryOnly*/);  // primary only because we can't evict a snapshot key
-                        bestkey = itr.key();
+                    if (dictSize(db->dict) != 0) {
+                        dictEntry *de = dictGetRandomKey(db->dict);
+                        bestkey = (sds)dictGetKey(de);
                         bestdbid = j;
                         break;
                     }
                 }
                 else
                 {
-                    if (db->expireSize())
+                    if (!db->setexpire->empty())
                     {
-                        bestkey = (sds)db->random_expire().key();
+                        bestkey = (sds)db->setexpire->random_value().key();
                         bestdbid = j;
                         break;
                     }
@@ -762,51 +651,34 @@ int performEvictions(bool fPreSnapshot) {
 
         /* Finally remove the selected key. */
         if (bestkey) {
-            db = g_pserver->db[bestdbid];
-
-            if (fEvictToStorage)
-            {
-                // This key is in the storage so we only need to free the object
-                dictEntry *deT;
-                if (db->removeCachedValue(bestkey, &deT)) {
-                    mem_freed += splazy->addEntry(db->dictUnsafeKeyOnly(), deT);
-                    ckeysFailed = 0;
-                }
-                else {
-                    delta = 0;
-                    ckeysFailed++;
-                    if (ckeysFailed > 1024)
-                        goto cant_free;
-                }
-            }
+            db = g_pserver->db+bestdbid;
+            robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
+            propagateExpire(db,keyobj,g_pserver->lazyfree_lazy_eviction);
+            /* We compute the amount of memory freed by db*Delete() alone.
+             * It is possible that actually the memory needed to propagate
+             * the DEL in AOF and replication link is greater than the one
+             * we are freeing removing the key, but we can't account for
+             * that otherwise we would never exit the loop.
+             *
+             * Same for CSC invalidation messages generated by signalModifiedKey.
+             *
+             * AOF and Output buffer memory will be freed eventually so
+             * we only care about memory used by the key space. */
+            delta = (long long) zmalloc_used_memory();
+            latencyStartMonitor(eviction_latency);
+            if (g_pserver->lazyfree_lazy_eviction)
+                dbAsyncDelete(db,keyobj);
             else
-            {
-                robj *keyobj = createStringObject(bestkey,sdslen(bestkey));
-                propagateExpire(db,keyobj,g_pserver->lazyfree_lazy_eviction);
-                /* We compute the amount of memory freed by db*Delete() alone.
-                * It is possible that actually the memory needed to propagate
-                * the DEL in AOF and replication link is greater than the one
-                * we are freeing removing the key, but we can't account for
-                * that otherwise we would never exit the loop.
-                *
-                * AOF and Output buffer memory will be freed eventually so
-                * we only care about memory used by the key space. */
-                delta = (long long) zmalloc_used_memory();
-                latencyStartMonitor(eviction_latency);
-                if (g_pserver->lazyfree_lazy_eviction)
-                    dbAsyncDelete(db,keyobj);
-                else
-                    dbSyncDelete(db,keyobj);
-                latencyEndMonitor(eviction_latency);
-                latencyAddSampleIfNeeded("eviction-del",eviction_latency);
-                delta -= (long long) zmalloc_used_memory();
-                mem_freed += delta;
-                g_pserver->stat_evictedkeys++;
-                signalModifiedKey(NULL,db,keyobj);
-                notifyKeyspaceEvent(NOTIFY_EVICTED, "evicted",
-                    keyobj, db->id);
-                decrRefCount(keyobj);
-            }
+                dbSyncDelete(db,keyobj);
+            latencyEndMonitor(eviction_latency);
+            latencyAddSampleIfNeeded("eviction-del",eviction_latency);
+            delta -= (long long) zmalloc_used_memory();
+            mem_freed += delta;
+            g_pserver->stat_evictedkeys++;
+            signalModifiedKey(NULL,db,keyobj);
+            notifyKeyspaceEvent(NOTIFY_EVICTED, "evicted",
+                keyobj, db->id);
+            decrRefCount(keyobj);
             keys_freed++;
 
             if (keys_freed % 16 == 0) {
@@ -832,9 +704,9 @@ int performEvictions(bool fPreSnapshot) {
                 /* After some time, exit the loop early - even if memory limit
                  * hasn't been reached.  If we suddenly need to free a lot of
                  * memory, don't want to spend too much time here.  */
-                if (g_pserver->m_pstorageFactory == nullptr && elapsedUs(evictionTimer) > eviction_time_limit_us) {
+                if (elapsedUs(evictionTimer) > eviction_time_limit_us) {
                     // We still need to free memory - start eviction timer proc
-                    if (!isEvictionProcRunning && serverTL->el != nullptr) {
+                    if (!isEvictionProcRunning) {
                         isEvictionProcRunning = 1;
                         aeCreateTimeEvent(serverTL->el, 0,
                                 evictionTimeProc, NULL, NULL);
@@ -849,38 +721,7 @@ int performEvictions(bool fPreSnapshot) {
     /* at this point, the memory is OK, or we have reached the time limit */
     result = (isEvictionProcRunning) ? EVICT_RUNNING : EVICT_OK;
 
-    if (splazy != nullptr && splazy->memory_queued() > 0 && !serverTL->gcEpoch.isReset()) {
-        g_pserver->garbageCollector.enqueue(serverTL->gcEpoch, std::move(splazy));
-    }
-
 cant_free:
-    if (g_pserver->m_pstorageFactory)
-    {
-        if (mem_reported < g_pserver->maxmemory*1.2) {
-            return EVICT_OK;    // Allow us to temporarily go over without OOMing
-        }
-    }
-
-    if (!cserver.delete_on_evict && result == EVICT_FAIL)
-    {
-        for (int idb = 0; idb < cserver.dbnum; ++idb)
-        {
-            redisDb *db = g_pserver->db[idb];
-            if (db->FStorageProvider())
-            {
-                if (db->size() != 0 && db->size(true /*fcachedOnly*/) == 0 && db->keycacheIsEnabled()) {
-                    serverLog(LL_WARNING, "Key cache exceeds maxmemory, freeing - performance may be affected increase maxmemory if possible");
-                    db->disableKeyCache();
-                } else if (db->size(true /*fCachedOnly*/)) {
-                    serverLog(LL_WARNING, "Failed to evict keys, falling back to flushing entire cache.  Consider increasing maxmemory-samples.");
-                    db->removeAllCachedValues();
-                    if (((mem_reported - zmalloc_used_memory()) + mem_freed) >= mem_tofree)
-                        result = EVICT_OK;
-                }
-            }
-        }
-    }
-
     if (result == EVICT_FAIL) {
         /* At this point, we have run out of evictable items.  It's possible
          * that some items are being freed in the lazyfree thread.  Perform a
@@ -895,7 +736,6 @@ cant_free:
 
     latencyEndMonitor(latency);
     latencyAddSampleIfNeeded("eviction-cycle",latency);
-cant_free_storage:
     return result;
 }
 

src/expire.cpp

@@ -33,8 +33,6 @@
 #include "server.h"
 #include "cron.h"
 
-fastlock g_expireLock {"Expire"};
-
 /* Helper function for the activeExpireCycle() function.
  * This function will try to expire the key that is stored in the hash table
  * entry 'de' of the 'expires' hash table of a Redis database.
@@ -83,7 +81,8 @@ int activeExpireCycleExpire(redisDb *db, expireEntry &e, long long now, size_t &
     }
 
     expireEntryFat *pfat = e.pfatentry();
-    robj *val = db->find(e.key());
+    dictEntry *de = dictFind(db->dict, e.key());
+    robj *val = (robj*)dictGetVal(de);
     int deleted = 0;
 
     redisObjectStack objKey;
@@ -163,14 +162,13 @@ int activeExpireCycleExpire(redisDb *db, expireEntry &e, long long now, size_t &
         }
     }
 
-    if (pfat->FEmpty())
-    {
-        removeExpire(db, &objKey);
-    }
-    else if (!pfat->FEmpty() && fTtlChanged)
+    if (!pfat->FEmpty() && fTtlChanged)
     {
         // We need to resort the expire entry since it may no longer be in the correct position
-        db->resortExpire(e);
+        auto itr = db->setexpire->find(e.key());
+        expireEntry eT = std::move(e);
+        db->setexpire->erase(itr);
+        db->setexpire->insert(eT);
     }
 
     if (deleted)
@@ -184,6 +182,10 @@ int activeExpireCycleExpire(redisDb *db, expireEntry &e, long long now, size_t &
         }
     }
 
+    if (pfat->FEmpty())
+    {
+        removeExpire(db, &objKey);
+    }
     return deleted;
 }
 
@@ -214,7 +216,7 @@ void expireMemberCore(client *c, robj *key, robj *subkey, long long basetime, lo
 
     /* No key, return zero. */
     robj *val = lookupKeyWriteOrReply(c, key, shared.czero);
-    if (val == nullptr) {
+    if (val == NULL) {
         return;
     }
 
@@ -371,7 +373,7 @@ void activeExpireCycleCore(int type) {
     long total_expired = 0;
 
     for (j = 0; j < dbs_per_call && timelimit_exit == 0; j++) {
-        redisDb *db = g_pserver->db[(current_db % cserver.dbnum)];
+        redisDb *db = g_pserver->db+(current_db % cserver.dbnum);
 
         /* Increment the DB now so we are sure if we run out of time
          * in the current DB we'll restart from the next. This allows to
@@ -383,18 +385,17 @@ void activeExpireCycleCore(int type) {
         now = mstime();
 
         /* If there is nothing to expire try next DB ASAP. */
-        if (db->setexpireUnsafe()->empty())
+        if (db->setexpire->empty())
         {
             db->avg_ttl = 0;
             db->last_expire_set = now;
             continue;
         }
         
-        std::unique_lock<fastlock> ul(g_expireLock);
         size_t expired = 0;
         size_t tried = 0;
         long long check = ACTIVE_EXPIRE_CYCLE_FAST_DURATION;    // assume a check is roughly 1us.  It isn't but good enough
-        db->expireitr = db->setexpireUnsafe()->enumerate(db->expireitr, now, [&](expireEntry &e) __attribute__((always_inline)) {
+        db->expireitr = db->setexpire->enumerate(db->expireitr, now, [&](expireEntry &e) __attribute__((always_inline)) {
             if (e.when() < now)
             {
                 expired += activeExpireCycleExpire(db, e, now, tried);
@@ -494,20 +495,20 @@ void expireSlaveKeys(void) {
         int dbid = 0;
         while(dbids && dbid < cserver.dbnum) {
             if ((dbids & 1) != 0) {
-                redisDb *db = g_pserver->db[dbid];
+                redisDb *db = g_pserver->db+dbid;
 
                 // the expire is hashed based on the key pointer, so we need the point in the main db
-                auto itrDB = db->find(keyname);
-                auto itrExpire = db->setexpire()->end();
-                if (itrDB != nullptr)
-                    itrExpire = db->setexpireUnsafe()->find(itrDB.key());
+                dictEntry *deMain = dictFind(db->dict, keyname);
+                auto itr = db->setexpire->end();
+                if (deMain != nullptr)
+                    itr = db->setexpire->find((sds)dictGetKey(deMain));
                 int expired = 0;
 
-                if (itrExpire != db->setexpire()->end())
+                if (itr != db->setexpire->end())
                 {
-                    if (itrExpire->when() < start) {
+                    if (itr->when() < start) {
                         size_t tried = 0;
-                        expired = activeExpireCycleExpire(g_pserver->db[dbid],*itrExpire,start,tried);
+                        expired = activeExpireCycleExpire(g_pserver->db+dbid,*itr,start,tried);
                     }
                 }
 
@@ -515,7 +516,7 @@ void expireSlaveKeys(void) {
                  * corresponding bit in the new bitmap we set as value.
                  * At the end of the loop if the bitmap is zero, it means we
                  * no longer need to keep track of this key. */
-                if (itrExpire != db->setexpire()->end() && !expired) {
+                if (itr != db->setexpire->end() && !expired) {
                     noexpire++;
                     new_dbids |= (uint64_t)1 << dbid;
                 }
@@ -693,8 +694,7 @@ void ttlGenericCommand(client *c, int output_ms) {
 
     /* The key exists. Return -1 if it has no expire, or the actual
         * TTL value otherwise. */
-    std::unique_lock<fastlock> ul(g_expireLock);
-    expireEntry *pexpire = c->db->getExpire(c->argv[1]);
+    expireEntry *pexpire = getExpire(c->db,c->argv[1]);
 
     if (c->argc == 2) {
         // primary expire    
@@ -752,7 +752,7 @@ void persistCommand(client *c) {
                 addReply(c,shared.czero);
             }
         } else if (c->argc == 3) {
-            if (c->db->removeSubkeyExpire(c->argv[1], c->argv[2])) {
+            if (removeSubkeyExpire(c->db, c->argv[1], c->argv[2])) {
                 signalModifiedKey(c,c->db,c->argv[1]);
                 notifyKeyspaceEvent(NOTIFY_GENERIC,"persist",c->argv[1],c->db->id);
                 addReply(c,shared.cone);
@@ -776,25 +776,19 @@ void touchCommand(client *c) {
     addReplyLongLong(c,touched);
 }
 
+expireEntryFat::expireEntryFat(const expireEntryFat &src)
+{
+    m_keyPrimary = sdsdupshared(src.m_keyPrimary);
+    m_vecexpireEntries = src.m_vecexpireEntries;
+    m_dictIndex = nullptr;
+}
+
 expireEntryFat::~expireEntryFat()
 {
     if (m_dictIndex != nullptr)
         dictRelease(m_dictIndex);
 }
 
-expireEntryFat::expireEntryFat(const expireEntryFat &e)
-    : m_keyPrimary(e.m_keyPrimary), m_vecexpireEntries(e.m_vecexpireEntries)
-{
-    // Note: dictExpires is not copied
-}
-
-expireEntryFat::expireEntryFat(expireEntryFat &&e)
-    : m_keyPrimary(std::move(e.m_keyPrimary)), m_vecexpireEntries(std::move(e.m_vecexpireEntries))
-{
-    m_dictIndex = e.m_dictIndex;
-    e.m_dictIndex = nullptr;
-}
-
 void expireEntryFat::createIndex()
 {
     serverAssert(m_dictIndex == nullptr);

src/expire.h

@@ -41,22 +41,21 @@ public:
     };
 
 private:
-    sdsimmutablestring m_keyPrimary;
+    sds m_keyPrimary;
     std::vector<subexpireEntry> m_vecexpireEntries;  // Note a NULL for the sds portion means the expire is for the primary key
     dict *m_dictIndex = nullptr;
 
     void createIndex();
 public:
-    expireEntryFat(const sdsimmutablestring &keyPrimary)
+    expireEntryFat(sds keyPrimary)
         : m_keyPrimary(keyPrimary)
         {}
+    expireEntryFat(const expireEntryFat &);
     ~expireEntryFat();
 
-    expireEntryFat(const expireEntryFat &e);
-    expireEntryFat(expireEntryFat &&e);
-
     long long when() const noexcept { return m_vecexpireEntries.front().when; }
-    const char *key() const noexcept { return static_cast<const char*>(m_keyPrimary); }
+
+    const char *key() const noexcept { return m_keyPrimary; }
 
     bool operator<(long long when) const noexcept { return this->when() <  when; }
 
@@ -65,15 +64,15 @@ public:
     bool FEmpty() const noexcept { return m_vecexpireEntries.empty(); }
     const subexpireEntry &nextExpireEntry() const noexcept { return m_vecexpireEntries.front(); }
     void popfrontExpireEntry();
-    const subexpireEntry &operator[](size_t idx) const { return m_vecexpireEntries[idx]; }
+    const subexpireEntry &operator[](size_t idx) { return m_vecexpireEntries[idx]; }
     size_t size() const noexcept { return m_vecexpireEntries.size(); }
 };
 
 class expireEntry {
-    struct
+    union
     {
-        sdsimmutablestring m_key;
-        expireEntryFat *m_pfatentry = nullptr;
+        sds m_key;
+        expireEntryFat *m_pfatentry;
     } u;
     long long m_when;   // bit wise and with FFatMask means this is a fat entry and we should use the pointer
 
@@ -87,11 +86,11 @@ public:
     class iter
     {
         friend class expireEntry;
-        const expireEntry *m_pentry = nullptr;
+        expireEntry *m_pentry = nullptr;
         size_t m_idx = 0;
 
     public:
-        iter(const expireEntry *pentry, size_t idx)
+        iter(expireEntry *pentry, size_t idx)
             : m_pentry(pentry), m_idx(idx)
         {}
 
@@ -123,29 +122,16 @@ public:
         if (subkey != nullptr)
         {
             m_when = FFatMask() | INVALID_EXPIRE;
-            u.m_pfatentry = new (MALLOC_LOCAL) expireEntryFat(sdsimmutablestring(sdsdupshared(key)));
+            u.m_pfatentry = new (MALLOC_LOCAL) expireEntryFat(key);
             u.m_pfatentry->expireSubKey(subkey, when);
         }
         else
         {
-            u.m_key = sdsimmutablestring(sdsdupshared(key));
+            u.m_key = key;
             m_when = when;
         }
     }
 
-    expireEntry(const expireEntry &e)
-    {
-        *this = e;
-    }
-    expireEntry(expireEntry &&e)
-    {
-        u.m_key = std::move(e.u.m_key);
-        u.m_pfatentry = std::move(e.u.m_pfatentry);
-        m_when = e.m_when;
-        e.m_when = 0;
-        e.u.m_pfatentry = nullptr;
-    }
-
     expireEntry(expireEntryFat *pfatentry)
     {
         u.m_pfatentry = pfatentry;
@@ -160,6 +146,14 @@ public:
         }
     }
 
+    expireEntry(expireEntry &&e)
+    {
+        u.m_key = e.u.m_key;
+        m_when = e.m_when;
+        e.u.m_key = (char*)key();  // we do this so it can still be found in the set
+        e.m_when = 0;
+    }
+
     // Duplicate the expire, note this is intended to be passed directly to setExpire
     expireEntry duplicate() const {
         expireEntry dst;
@@ -178,31 +172,21 @@ public:
             delete u.m_pfatentry;
     }
 
-    expireEntry &operator=(const expireEntry &e)
-    {
-        u.m_key = e.u.m_key;
-        m_when = e.m_when;
-        if (e.FFat())
-            u.m_pfatentry = new (MALLOC_LOCAL) expireEntryFat(*e.u.m_pfatentry);
-        return *this;
-    }
-
     void setKeyUnsafe(sds key)
     {
         if (FFat())
-            u.m_pfatentry->m_keyPrimary = sdsimmutablestring(sdsdupshared(key));
+            u.m_pfatentry->m_keyPrimary = key;
         else
-            u.m_key = sdsimmutablestring(sdsdupshared(key));
+            u.m_key = key;
     }
 
     inline bool FFat() const noexcept { return m_when & FFatMask(); }
     expireEntryFat *pfatentry() { assert(FFat()); return u.m_pfatentry; }
-    const expireEntryFat *pfatentry() const { assert(FFat()); return u.m_pfatentry; }
 
 
-    bool operator==(const sdsview &key) const noexcept
+    bool operator==(const char *key) const noexcept
     { 
-        return key == this->key(); 
+        return this->key() == key; 
     }
 
     bool operator<(const expireEntry &e) const noexcept
@@ -218,7 +202,7 @@ public:
     { 
         if (FFat())
             return u.m_pfatentry->key();
-        return static_cast<const char*>(u.m_key);
+        return u.m_key;
     }
     long long when() const noexcept
     { 
@@ -240,7 +224,7 @@ public:
             {
                 // we have to upgrade to a fat entry
                 long long whenT = m_when;
-                sdsimmutablestring keyPrimary = u.m_key;
+                sds keyPrimary = u.m_key;
                 m_when |= FFatMask();
                 u.m_pfatentry = new (MALLOC_LOCAL) expireEntryFat(keyPrimary);
                 u.m_pfatentry->expireSubKey(nullptr, whenT);
@@ -252,8 +236,8 @@ public:
         u.m_pfatentry->expireSubKey(subkey, when);
     }
     
-    iter begin() const { return iter(this, 0); }
-    iter end() const
+    iter begin() { return iter(this, 0); }
+    iter end()
     {
         if (FFat())
             return iter(this, u.m_pfatentry->size());
@@ -268,13 +252,6 @@ public:
             pfatentry()->m_vecexpireEntries.begin() + itr.m_idx);
     }
 
-    size_t size() const
-    {
-        if (FFat())
-            return u.m_pfatentry->size();
-        return 1;
-    }
-
     long long FGetPrimaryExpire() const noexcept
     { 
         return m_when & (~FFatMask()); 
@@ -286,8 +263,7 @@ public:
         return *pwhen != INVALID_EXPIRE;
     }
 
-    explicit operator sdsview() const noexcept { return key(); }
+    explicit operator const char*() const noexcept { return key(); }
     explicit operator long long() const noexcept { return when(); }
 };
-typedef semiorderedset<expireEntry, sdsview, true /*expireEntry can be memmoved*/> expireset;
-extern fastlock g_expireLock;
+typedef semiorderedset<expireEntry, const char *, true /*expireEntry can be memmoved*/> expireset;

src/fastlock.cpp

@@ -437,7 +437,7 @@ extern "C" void fastlock_unlock(struct fastlock *lock)
         serverAssert(pidT >= 0);  // unlock after free
         int t = -1;
         __atomic_store(&lock->m_pidOwner, &t, __ATOMIC_RELEASE);
-        std::atomic_thread_fence(std::memory_order_acq_rel);
+        std::atomic_thread_fence(std::memory_order_release);
         ANNOTATE_RWLOCK_RELEASED(lock, true);
         uint16_t activeNew = __atomic_add_fetch(&lock->m_ticket.m_active, 1, __ATOMIC_RELEASE);  // on x86 the atomic is not required here, but ASM handles that case
 #ifdef __linux__

src/fastlock_x64.asm

@@ -190,11 +190,10 @@ fastlock_unlock:
 	mov dword ptr [rdi], -1      # pidOwner = -1 (we don't own it anymore)
 	mov esi, [rdi+64]            # get current active (this one)
 	inc esi                      # bump it to the next thread
-	sfence                       # ensure whatever was written in the lock is visible
 	mov word ptr [rdi+64], si    # give up our ticket (note: lock is not required here because the spinlock itself guards this variable)
+	mfence                       # sync other threads
 	# At this point the lock is removed, however we must wake up any pending futexs
-	mov rdx, [rdi+64]            # load the futex mask, note we intentionally also read the ticket we just wrote to ensure this is ordered with the above mov
-	shr rdx, 32                  # isolate the mask
+	mov edx, [rdi+64+4]          # load the futex mask
 	bt edx, esi                  # is the next thread waiting on a futex?
 	jc unlock_futex              # unlock the futex if necessary
 	ret                          # if not we're done.

src/gc.h

@@ -1,136 +0,0 @@
-#pragma once
-#include <vector>
-#include <assert.h>
-#include <unordered_set>
-#include <list>
-
-struct ICollectable
-{
-    virtual ~ICollectable() {}
-    bool FWillFreeChildDebug() { return false; }
-};
-
-template<typename T>
-class GarbageCollector
-{
-    struct EpochHolder
-    {
-        uint64_t tstamp;
-        std::unique_ptr<std::vector<std::unique_ptr<T>>> m_spvecObjs;
-
-        EpochHolder() {
-            m_spvecObjs = std::make_unique<std::vector<std::unique_ptr<T>>>();
-        }
-
-        // Support move operators
-        EpochHolder(EpochHolder &&other) = default;
-        EpochHolder &operator=(EpochHolder &&) = default;
-
-
-
-        bool operator<(uint64_t tstamp) const
-        {
-            return this->tstamp < tstamp;
-        }
-
-        bool operator==(uint64_t tstamp) const
-        {
-            return this->tstamp == tstamp;
-        }
-    };
-
-public:
-    ~GarbageCollector() {
-        // Silence TSAN errors
-        m_lock.lock();
-    }
-
-    uint64_t startEpoch()
-    {
-        std::unique_lock<fastlock> lock(m_lock);
-        ++m_epochNext;
-        m_setepochOutstanding.insert(m_epochNext);
-        return m_epochNext;
-    }
-
-    void shutdown()
-    {
-        std::unique_lock<fastlock> lock(m_lock);
-        m_listepochs.clear();
-        m_setepochOutstanding.clear();
-    }
-
-    bool empty() const
-    {
-        std::unique_lock<fastlock> lock(m_lock);
-        return m_listepochs.empty();
-    }
-
-    void endEpoch(uint64_t epoch, bool fNoFree = false)
-    {
-        std::unique_lock<fastlock> lock(m_lock);
-        serverAssert(m_setepochOutstanding.find(epoch) != m_setepochOutstanding.end());
-        bool fMinElement = *std::min_element(m_setepochOutstanding.begin(), m_setepochOutstanding.end());
-        m_setepochOutstanding.erase(epoch);
-        if (fNoFree)
-            return;
-        std::list<EpochHolder> listclean;
-        
-        // No outstanding epochs?
-        if (m_setepochOutstanding.empty())
-        {
-            listclean = std::move(m_listepochs);  // Everything goes!
-        }
-        else
-        {
-            uint64_t minepoch = *std::min_element(m_setepochOutstanding.begin(), m_setepochOutstanding.end());
-            if (minepoch == 0)
-                return; // No available epochs to free
-
-            // Clean any epochs available (after the lock)
-            for (auto itr = m_listepochs.begin(); itr != m_listepochs.end(); /* itr incremented in loop*/)
-            {
-                auto &e = *itr;
-                auto itrNext = itr;
-                ++itrNext;
-                if (e < minepoch)
-                {
-                    listclean.emplace_back(std::move(e));
-                    m_listepochs.erase(itr);
-                }
-                itr = itrNext;
-            }
-
-            assert(listclean.empty() || fMinElement);
-        }
-
-        lock.unlock();  // don't hold it for the potentially long delete of vecclean
-    }
-
-    void enqueue(uint64_t epoch, std::unique_ptr<T> &&sp)
-    {
-        std::unique_lock<fastlock> lock(m_lock);
-        serverAssert(m_setepochOutstanding.find(epoch) != m_setepochOutstanding.end());
-        serverAssert(sp->FWillFreeChildDebug() == false);
-
-        auto itr = std::find(m_listepochs.begin(), m_listepochs.end(), m_epochNext+1);
-        if (itr == m_listepochs.end())
-        {
-            EpochHolder e;
-            e.tstamp = m_epochNext+1;
-            e.m_spvecObjs->push_back(std::move(sp));
-            m_listepochs.emplace_back(std::move(e));
-        }
-        else
-        {
-            itr->m_spvecObjs->push_back(std::move(sp));
-        }
-    }
-
-private:
-    mutable fastlock m_lock { "Garbage Collector"};
-
-    std::list<EpochHolder> m_listepochs;
-    std::unordered_set<uint64_t> m_setepochOutstanding;
-    uint64_t m_epochNext = 0;
-};

src/help.h

@@ -1071,7 +1071,7 @@ struct commandHelp {
     1,
     "2.2.0" },
     { "SHUTDOWN",
-    "[NOSAVE|SAVE|SOFT]",
+    "[NOSAVE|SAVE]",
     "Synchronously save the dataset to disk and then shut down the server",
     9,
     "1.0.0" },

src/keydb-diagnostic-tool.cpp

@@ -211,11 +211,6 @@ extern "C" void _serverAssert(const char *estr, const char *file, int line) {
     *((char*)-1) = 'x';
 }
 
-/* asyncFreeDictTable is needed by dict */
-extern "C" void asyncFreeDictTable(struct dictEntry **de) {
-    zfree(de);
-}
-
 static redisContext *getRedisContext(const char *ip, int port,
                                      const char *hostsocket)
 {
@@ -904,7 +899,7 @@ int main(int argc, const char **argv) {
 
     while (self_threads < config.max_threads) {
         for (int i = 0; i < config.numclients; i++) {
-            snprintf(command, 63, "SET %d %s\r\n", self_threads * config.numclients + i, set_value);
+            sprintf(command, "SET %d %s\r\n", self_threads * config.numclients + i, set_value);
             createClient(command, strlen(command), NULL,self_threads);
         }
 

src/keydbutils.cpp

@@ -1,10 +0,0 @@
-#include "server.h"
-
-namespace keydbutils
-{
-    template<>
-    size_t hash(const sdsview& t)
-    {
-        return (size_t)dictGenHashFunction(static_cast<const char*>(t), t.size());
-    }
-}

src/latency.cpp

@@ -230,7 +230,7 @@ void analyzeLatencyForEvent(char *event, struct latencyStats *ls) {
     if (ls->samples) ls->mad = sum / ls->samples;
 }
 
-/* Create a human readable report of latency events for this KeyDB instance. */
+/* Create a human readable report of latency events for this Redis instance. */
 sds createLatencyReport(void) {
     sds report = sdsempty();
     int advise_better_vm = 0;       /* Better virtual machines. */
@@ -256,7 +256,7 @@ sds createLatencyReport(void) {
     if (dictSize(g_pserver->latency_events) == 0 &&
         g_pserver->latency_monitor_threshold == 0)
     {
-        report = sdscat(report,"I'm sorry, Dave, I can't do that. Latency monitoring is disabled in this KeyDB instance. You may use \"CONFIG SET latency-monitor-threshold <milliseconds>.\" in order to enable it.\n");
+        report = sdscat(report,"I'm sorry, Dave, I can't do that. Latency monitoring is disabled in this KeyDB instance. You may use \"CONFIG SET latency-monitor-threshold <milliseconds>.\" in order to enable it. If we weren't in a deep space mission I'd suggest to take a look at https://redis.io/topics/latency-monitor.\n");
         return report;
     }
 
@@ -398,7 +398,7 @@ sds createLatencyReport(void) {
     dictReleaseIterator(di);
 
     /* Add non event based advices. */
-    if (g_pserver->aof_enabled && THPGetAnonHugePagesSize() > 0) {
+    if (THPGetAnonHugePagesSize() > 0) {
         advise_disable_thp = 1;
         advices++;
     }

src/lazyfree.cpp

@@ -143,28 +143,19 @@ size_t lazyfreeGetFreeEffort(robj *key, robj *obj) {
  * a lazy free list instead of being freed synchronously. The lazy free list
  * will be reclaimed in a different bio.c thread. */
 #define LAZYFREE_THRESHOLD 64
-bool redisDbPersistentData::asyncDelete(robj *key) {
+int dbAsyncDelete(redisDb *db, robj *key) {
     /* If the value is composed of a few allocations, to free in a lazy way
      * is actually just slower... So under a certain limit we just free
      * the object synchronously. */
-
-    if (m_spstorage != nullptr)
-        return syncDelete(key); // async delte never makes sense with a storage provider
-
-    dictEntry *de = dictUnlink(m_pdict,ptrFromObj(key));
+    dictEntry *de = dictUnlink(db->dict,ptrFromObj(key));
     if (de) {
-        if (m_pdbSnapshot != nullptr && m_pdbSnapshot->find_cached_threadsafe(szFromObj(key)) != nullptr)
-        {
-            uint64_t hash = dictGetHash(m_pdict, szFromObj(key));
-            dictAdd(m_pdictTombstone, sdsdup((sds)dictGetKey(de)), (void*)hash);
-        }
-
         robj *val = (robj*)dictGetVal(de);
+
         if (val->FExpires())
         {
             /* Deleting an entry from the expires dict will not free the sds of
              * the key, because it is shared with the main dictionary. */
-            removeExpire(key,dict_iter(m_pdict, de));
+            removeExpireCore(db,key,de);
         }
 
         /* Tells the module that the key has been unlinked from the database. */
@@ -183,25 +174,21 @@ bool redisDbPersistentData::asyncDelete(robj *key) {
         if (free_effort > LAZYFREE_THRESHOLD && val->getrefcount(std::memory_order_relaxed) == 1) {
             atomicIncr(lazyfree_objects,1);
             bioCreateLazyFreeJob(lazyfreeFreeObject,1, val);
-            dictSetVal(m_pdict,de,NULL);
+            dictSetVal(db->dict,de,NULL);
         }
     }
 
     /* Release the key-val pair, or just the key if we set the val
      * field to NULL in order to lazy free it later. */
     if (de) {
-        dictFreeUnlinkedEntry(m_pdict,de);
+        dictFreeUnlinkedEntry(db->dict,de);
         if (g_pserver->cluster_enabled) slotToKeyDel(szFromObj(key));
-        return true;
+        return 1;
     } else {
-        return false;
+        return 0;
     }
 }
 
-int dbAsyncDelete(redisDb *db, robj *key) {
-    return db->asyncDelete(key);
-}
-
 /* Free an object, if the object is huge enough, free it in async way. */
 void freeObjAsync(robj *key, robj *obj) {
     size_t free_effort = lazyfreeGetFreeEffort(key,obj);
@@ -216,16 +203,12 @@ void freeObjAsync(robj *key, robj *obj) {
 /* Empty a Redis DB asynchronously. What the function does actually is to
  * create a new empty set of hash tables and scheduling the old ones for
  * lazy freeing. */
-void redisDbPersistentData::emptyDbAsync() {
-    std::unique_lock<fastlock> ul(g_expireLock);
-    dict *oldht1 = m_pdict;
-    auto *set = m_setexpire;
-    m_setexpire = new (MALLOC_LOCAL) expireset();
-    m_pdict = dictCreate(&dbDictType,this);
-    if (m_spstorage != nullptr)
-        m_spstorage->clearAsync();
-    if (m_fTrackingChanges)
-        m_fAllChanged = true;
+void emptyDbAsync(redisDb *db) {
+    dict *oldht1 = db->dict;
+    auto *set = db->setexpire;
+    db->setexpire = new (MALLOC_LOCAL) expireset();
+    db->expireitr = db->setexpire->end();
+    db->dict = dictCreate(&dbDictType,NULL);
     atomicIncr(lazyfree_objects,dictSize(oldht1));
     bioCreateLazyFreeJob(lazyfreeFreeDatabase,2,oldht1,set);
 }

src/module.cpp

@@ -365,7 +365,11 @@ typedef struct RedisModuleCommandFilter {
 static list *moduleCommandFilters;
 
 /* Module GIL Variables */
-static readWriteLock s_moduleGIL("Module GIL");
+static int s_cAcquisitionsServer = 0;
+static int s_cAcquisitionsModule = 0;
+static std::mutex s_mutex;
+static std::condition_variable s_cv;
+static std::recursive_mutex s_mutexModule;
 thread_local bool g_fModuleThread = false;
 
 typedef void (*RedisModuleForkDoneHandler) (int exitcode, int bysignal, void *user_data);
@@ -835,7 +839,6 @@ int64_t commandFlagsFromString(char *s) {
         else if (!strcasecmp(t,"may-replicate")) flags |= CMD_MAY_REPLICATE;
         else if (!strcasecmp(t,"getkeys-api")) flags |= CMD_MODULE_GETKEYS;
         else if (!strcasecmp(t,"no-cluster")) flags |= CMD_MODULE_NO_CLUSTER;
-        else if (!strcasecmp(t,"async")) flags |= CMD_ASYNC_OK;
         else break;
     }
     sdsfreesplitres(tokens,count);
@@ -2442,8 +2445,7 @@ int RM_UnlinkKey(RedisModuleKey *key) {
  * If no TTL is associated with the key or if the key is empty,
  * REDISMODULE_NO_EXPIRE is returned. */
 mstime_t RM_GetExpire(RedisModuleKey *key) {
-    std::unique_lock<fastlock> ul(g_expireLock);
-    expireEntry *pexpire = key->db->getExpire(key->key);
+    expireEntry *pexpire = getExpire(key->db,key->key);
     mstime_t expire = INVALID_EXPIRE;
     if (pexpire != nullptr)
         pexpire->FGetPrimaryExpire(&expire);
@@ -2479,7 +2481,7 @@ int RM_SetExpire(RedisModuleKey *key, mstime_t expire) {
  * If no TTL is associated with the key or if the key is empty,
  * REDISMODULE_NO_EXPIRE is returned. */
 mstime_t RM_GetAbsExpire(RedisModuleKey *key) {
-    auto expire = key->db->getExpire(key->key);
+    auto expire = getExpire(key->db,key->key);
     if (expire == nullptr || key->value == NULL) 
         return REDISMODULE_NO_EXPIRE;
     return expire->when();
@@ -2517,7 +2519,7 @@ void RM_ResetDataset(int restart_aof, int async) {
 
 /* Returns the number of keys in the current db. */
 unsigned long long RM_DbSize(RedisModuleCtx *ctx) {
-    return ctx->client->db->size();
+    return dictSize(ctx->client->db->dict);
 }
 
 /* Returns a name of a random key, or NULL if current db is empty. */
@@ -4287,13 +4289,7 @@ RedisModuleCallReply *RM_Call(RedisModuleCtx *ctx, const char *cmdname, const ch
         if (!(flags & REDISMODULE_ARGV_NO_REPLICAS))
             call_flags |= CMD_CALL_PROPAGATE_REPL;
     }
-
-    {
-    AeLocker locker; locker.arm(nullptr);
-    std::unique_lock<fastlock> ul(c->lock);
     call(c,call_flags);
-    }
-
     g_pserver->replication_allowed = prev_replication_allowed;
 
     serverAssert((c->flags & CLIENT_BLOCKED) == 0);
@@ -5966,58 +5962,95 @@ void RM_ThreadSafeContextUnlock(RedisModuleCtx *ctx) {
 //  as the server thread acquisition is sufficient.  If we did try to lock we would deadlock
 static bool FModuleCallBackLock(bool fServerThread)
 {
-    return !fServerThread && aeThreadOwnsLock() && !g_fModuleThread && s_moduleGIL.hasReader();
+    return !fServerThread && aeThreadOwnsLock() && !g_fModuleThread && s_cAcquisitionsServer > 0;
 }
 void moduleAcquireGIL(int fServerThread, int fExclusive) {
+    std::unique_lock<std::mutex> lock(s_mutex);
+    int *pcheck = fServerThread ? &s_cAcquisitionsModule : &s_cAcquisitionsServer;
+
     if (FModuleCallBackLock(fServerThread)) {
         return;
     }
 
+    while (*pcheck > 0)
+        s_cv.wait(lock);
+
     if (fServerThread)
     {
-        s_moduleGIL.acquireRead();
+        ++s_cAcquisitionsServer;
     }
     else
     {
-        s_moduleGIL.acquireWrite(fExclusive);
+        // only try to acquire the mutexModule in exclusive mode 
+        if (fExclusive){
+            // It is possible that another module thread holds the GIL (and s_mutexModule as a result). 
+            // When said thread goes to release the GIL, it will wait for s_mutex, which this thread owns. 
+            // This thread is however waiting for the GIL (and s_mutexModule) that the other thread owns.
+            // As a result, a deadlock has occured. 
+            // We release the lock on s_mutex and wait until we are able to safely acquire the GIL 
+            // in order to prevent this deadlock from occuring. 
+            while (!s_mutexModule.try_lock())
+                s_cv.wait(lock);     
+        }    
+        ++s_cAcquisitionsModule;
+        fModuleGILWlocked++;
     }
 }
 
 int moduleTryAcquireGIL(bool fServerThread, int fExclusive) {
+    std::unique_lock<std::mutex> lock(s_mutex, std::defer_lock);
+    if (!lock.try_lock())
+        return 1;
+    int *pcheck = fServerThread ? &s_cAcquisitionsModule : &s_cAcquisitionsServer;
+
     if (FModuleCallBackLock(fServerThread)) {
         return 0;
     }
 
+    if (*pcheck > 0)
+        return 1;
+
     if (fServerThread)
     {
-        if (!s_moduleGIL.tryAcquireRead())
-            return 1;
+        ++s_cAcquisitionsServer;
     }
     else
     {
-        if (!s_moduleGIL.tryAcquireWrite(fExclusive))
-            return 1;
+        // only try to acquire the mutexModule in exclusive mode 
+        if (fExclusive){
+            if (!s_mutexModule.try_lock())
+                return 1;
+        }
+        ++s_cAcquisitionsModule;
+        fModuleGILWlocked++;
     }
     return 0;
 }
 
 void moduleReleaseGIL(int fServerThread, int fExclusive) {
+    std::unique_lock<std::mutex> lock(s_mutex);
+
     if (FModuleCallBackLock(fServerThread)) {
         return;
     }
-    
+
     if (fServerThread)
     {
-        s_moduleGIL.releaseRead();
+        --s_cAcquisitionsServer;
     }
     else
     {
-        s_moduleGIL.releaseWrite(fExclusive);
+        // only try to release the mutexModule in exclusive mode
+        if (fExclusive)
+            s_mutexModule.unlock();
+        --s_cAcquisitionsModule;
+        fModuleGILWlocked--;
     }
+    s_cv.notify_all();
 }
 
 int moduleGILAcquiredByModule(void) {
-    return s_moduleGIL.hasWriter();
+    return fModuleGILWlocked > 0;
 }
 
 
@@ -7881,7 +7914,7 @@ int RM_Scan(RedisModuleCtx *ctx, RedisModuleScanCursor *cursor, RedisModuleScanC
     }
     int ret = 1;
     ScanCBData data = { ctx, privdata, fn };
-    cursor->cursor = dictScan(ctx->client->db->dictUnsafeKeyOnly(), cursor->cursor, moduleScanCallback, NULL, &data);
+    cursor->cursor = dictScan(ctx->client->db->dict, cursor->cursor, moduleScanCallback, NULL, &data);
     if (cursor->cursor == 0) {
         cursor->done = 1;
         ret = 0;
@@ -8460,8 +8493,6 @@ void moduleFireServerEvent(uint64_t eid, int subid, void *data) {
      * cheap if there are no registered modules. */
     if (listLength(RedisModule_EventListeners) == 0) return;
 
-    aeAcquireLock();
-
     int real_client_used = 0;
     listIter li;
     listNode *ln;
@@ -8537,8 +8568,6 @@ void moduleFireServerEvent(uint64_t eid, int subid, void *data) {
             moduleFreeContext(&ctx);
         }
     }
-
-    aeReleaseLock();
 }
 
 /* Remove all the listeners for this module: this is used before unloading

src/motd.cpp

@@ -1,11 +1,7 @@
-#ifdef CLIENT
 extern "C" {
 #include <sdscompat.h>
 #include <sds.h>
 }
-#else
-#include "sds.h"
-#endif
 #include <cstring>
 #include <unistd.h>
 #include <sys/types.h>
@@ -19,7 +15,6 @@ extern "C" {
 #ifdef MOTD
 #include <curl/curl.h> 
 
-#ifdef CLIENT
 extern "C" {
 __attribute__ ((weak)) hisds hi_sdscatlen(hisds s, const void *t, size_t len) {
     return sdscatlen(s, t, len);
@@ -28,7 +23,6 @@ __attribute__ ((weak)) hisds hi_sdscat(hisds s, const char *t) {
     return sdscat(s, t);
 }
 }
-#endif
 
 static const char *szMotdCachePath()
 {

src/multi.cpp

@@ -413,15 +413,15 @@ void touchAllWatchedKeysInDb(redisDb *emptied, redisDb *replaced_with) {
 
     dictIterator *di = dictGetSafeIterator(emptied->watched_keys);
     while((de = dictNext(di)) != NULL) {
-        robj *key = (robj*)de->key;
-        list *clients = (list*)de->v.val;
+        robj *key = (robj*)dictGetKey(de);
+        list *clients = (list*)dictGetVal(de);
         if (!clients) continue;
         listRewind(clients,&li);
         while((ln = listNext(&li))) {
             client *c = (client*)listNodeValue(ln);
-            if (emptied->find(key)) {
+            if (dictFind(emptied->dict, ptrFromObj(key))) {
                 c->flags |= CLIENT_DIRTY_CAS;
-            } else if (replaced_with && replaced_with->find(key)) {
+            } else if (replaced_with && dictFind(replaced_with->dict, ptrFromObj(key))) {
                 c->flags |= CLIENT_DIRTY_CAS;
             }
         }

src/new.cpp

@@ -41,10 +41,3 @@ void operator delete(void *p, std::size_t) noexcept
 }
 
 #endif
-
-#if defined(USE_JEMALLOC)
-extern "C" size_t malloc_usable_size(void *ptr)
-{
-    return zmalloc_usable_size(ptr);
-}
-#endif

src/new.h

@@ -1,6 +1,5 @@
 #pragma once
 #include <cstddef>  // std::size_t
-#include "storage.h"
 
 void *operator new(size_t size, enum MALLOC_CLASS mclass);
 

src/object.cpp

@@ -96,18 +96,15 @@ robj *createRawStringObject(const char *ptr, size_t len) {
  * an object where the sds string is actually an unmodifiable string
  * allocated in the same chunk as the object itself. */
 robj *createEmbeddedStringObject(const char *ptr, size_t len) {
-    serverAssert(len <= UINT8_MAX);
-    // Note: If the size changes update serializeStoredStringObject
     size_t allocsize = sizeof(struct sdshdr8)+len+1;
     if (allocsize < sizeof(void*))
         allocsize = sizeof(void*);
 
     size_t mvccExtraBytes = g_pserver->fActiveReplica ? sizeof(redisObjectExtended) : 0;
-    char *oB = (char*)zmalloc(sizeof(robj)+allocsize-sizeof(redisObject::m_ptr)+mvccExtraBytes, MALLOC_SHARED);
+    char *oB = (char*)zcalloc(sizeof(robj)+allocsize-sizeof(redisObject::m_ptr)+mvccExtraBytes, MALLOC_SHARED);
     robj *o = reinterpret_cast<robj*>(oB + mvccExtraBytes);
     struct sdshdr8 *sh = (sdshdr8*)(&o->m_ptr);
 
-    new (o) redisObject;
     o->type = OBJ_STRING;
     o->encoding = OBJ_ENCODING_EMBSTR;
     o->setrefcount(1);
@@ -139,8 +136,9 @@ robj *createEmbeddedStringObject(const char *ptr, size_t len) {
  *
  * The current limit of 52 is chosen so that the biggest string object
  * we allocate as EMBSTR will still fit into the 64 byte arena of jemalloc. */
-size_t OBJ_ENCODING_EMBSTR_SIZE_LIMIT = 52;
+#define OBJ_ENCODING_EMBSTR_SIZE_LIMIT 52
 
+//static_assert((sizeof(redisObject)+OBJ_ENCODING_EMBSTR_SIZE_LIMIT-8) == 64, "Max EMBSTR obj should be 64 bytes total");
 robj *createStringObject(const char *ptr, size_t len) {
     if (len <= OBJ_ENCODING_EMBSTR_SIZE_LIMIT)
         return createEmbeddedStringObject(ptr,len);
@@ -887,7 +885,7 @@ size_t streamRadixTreeMemoryUsage(rax *rax) {
  * case of aggregated data types where only "sample_size" elements
  * are checked and averaged to estimate the total size. */
 #define OBJ_COMPUTE_SIZE_DEF_SAMPLES 5 /* Default sample size. */
-size_t objectComputeSize(robj_roptr o, size_t sample_size) {
+size_t objectComputeSize(robj *o, size_t sample_size) {
     sds ele, ele2;
     dict *d;
     dictIterator *di;
@@ -898,7 +896,7 @@ size_t objectComputeSize(robj_roptr o, size_t sample_size) {
         if(o->encoding == OBJ_ENCODING_INT) {
             asize = sizeof(*o);
         } else if(o->encoding == OBJ_ENCODING_RAW) {
-            asize = sdsZmallocSize((sds)szFromObj(o))+sizeof(*o);
+            asize = sdsZmallocSize(szFromObj(o))+sizeof(*o);
         } else if(o->encoding == OBJ_ENCODING_EMBSTR) {
             asize = sdslen(szFromObj(o))+2+sizeof(*o);
         } else {
@@ -1091,7 +1089,7 @@ struct redisMemOverhead *getMemoryOverheadData(void) {
     mem_total += g_pserver->initial_memory_usage;
 
     mem = 0;
-    if (g_pserver->repl_backlog && g_pserver->repl_backlog != g_pserver->repl_backlog_disk)
+    if (g_pserver->repl_backlog)
         mem += zmalloc_size(g_pserver->repl_backlog);
     mh->repl_backlog = mem;
     mem_total += mem;
@@ -1127,22 +1125,21 @@ struct redisMemOverhead *getMemoryOverheadData(void) {
     mem_total+=mem;
 
     for (j = 0; j < cserver.dbnum; j++) {
-        redisDb *db = g_pserver->db[j];
-        long long keyscount = db->size();
+        redisDb *db = g_pserver->db+j;
+        long long keyscount = dictSize(db->dict);
         if (keyscount==0) continue;
 
         mh->total_keys += keyscount;
         mh->db = (decltype(mh->db))zrealloc(mh->db,sizeof(mh->db[0])*(mh->num_dbs+1), MALLOC_LOCAL);
         mh->db[mh->num_dbs].dbid = j;
 
-        mem = db->size() * sizeof(dictEntry) +
-              db->slots() * sizeof(dictEntry*) +
-              db->size() * sizeof(robj);
+        mem = dictSize(db->dict) * sizeof(dictEntry) +
+              dictSlots(db->dict) * sizeof(dictEntry*) +
+              dictSize(db->dict) * sizeof(robj);
         mh->db[mh->num_dbs].overhead_ht_main = mem;
         mem_total+=mem;
-        
-        std::unique_lock<fastlock> ul(g_expireLock);
-        mem = db->setexpire()->estimated_bytes_used();
+
+        mem = db->setexpire->bytes_used();
         mh->db[mh->num_dbs].overhead_ht_expires = mem;
         mem_total+=mem;
 
@@ -1325,7 +1322,7 @@ int objectSetLRUOrLFU(robj *val, long long lfu_freq, long long lru_idle,
 /* This is a helper function for the OBJECT command. We need to lookup keys
  * without any modification of LRU or other parameters. */
 robj_roptr objectCommandLookup(client *c, robj *key) {
-    return c->db->find(key);
+    return lookupKeyReadWithFlags(c->db,key,LOOKUP_NOTOUCH|LOOKUP_NONOTIFY);
 }
 
 robj_roptr objectCommandLookupOrReply(client *c, robj *key, robj *reply) {
@@ -1416,6 +1413,7 @@ void memoryCommand(client *c) {
         };
         addReplyHelp(c, help);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"usage") && c->argc >= 3) {
+        dictEntry *de;
         long long samples = OBJ_COMPUTE_SIZE_DEF_SAMPLES;
         for (int j = 3; j < c->argc; j++) {
             if (!strcasecmp(szFromObj(c->argv[j]),"samples") &&
@@ -1434,14 +1432,12 @@ void memoryCommand(client *c) {
                 return;
             }
         }
-
-        auto itr = c->db->find(c->argv[2]);
-        if (itr == nullptr) {
+        if ((de = dictFind(c->db->dict,ptrFromObj(c->argv[2]))) == NULL) {
             addReplyNull(c);
             return;
         }
-        size_t usage = objectComputeSize(itr.val(),samples);
-        usage += sdsZmallocSize(itr.key());
+        size_t usage = objectComputeSize((robj*)dictGetVal(de),samples);
+        usage += sdsZmallocSize((sds)dictGetKey(de));
         usage += sizeof(dictEntry);
         addReplyLongLong(c,usage);
     } else if (!strcasecmp(szFromObj(c->argv[1]),"stats") && c->argc == 2) {
@@ -1563,7 +1559,7 @@ void memoryCommand(client *c) {
 
 void redisObject::SetFExpires(bool fExpire)
 {
-    serverAssert(this->refcount != OBJ_SHARED_REFCOUNT || fExpire == FExpires());
+    serverAssert(this->refcount != OBJ_SHARED_REFCOUNT);
     if (fExpire)
         this->refcount.fetch_or(1U << 31, std::memory_order_relaxed);
     else
@@ -1576,131 +1572,6 @@ void redisObject::setrefcount(unsigned ref)
     refcount.store(ref, std::memory_order_relaxed); 
 }
 
-sds serializeStoredStringObject(sds str, robj_roptr o)
-{
-    uint64_t mvcc;
-    mvcc = mvccFromObj(o);
-    str = sdscatlen(str, &mvcc, sizeof(mvcc));
-    str = sdscatlen(str, &(*o), sizeof(robj));
-    static_assert((sizeof(robj) + sizeof(mvcc)) == sizeof(redisObjectStack), "");
-    switch (o->encoding)
-    {
-    case OBJ_ENCODING_EMBSTR:
-    case OBJ_ENCODING_RAW:
-        str = sdscatsds(str, (sds)szFromObj(o));
-        break;
-
-    case OBJ_ENCODING_INT:
-        break;  //nop
-    }
-        
-    return str;
-}
-
-robj *deserializeStoredStringObject(const char *data, size_t cb)
-{
-    uint64_t mvcc = *reinterpret_cast<const uint64_t*>(data);
-    const robj *oT = (const robj*)(data+sizeof(uint64_t));
-    robj *newObject = nullptr;
-    switch (oT->encoding)
-    {
-    case OBJ_ENCODING_INT:
-        newObject = createObject(OBJ_STRING, nullptr);
-        newObject->encoding = oT->encoding;
-        newObject->m_ptr = oT->m_ptr;
-        break;
-
-    case OBJ_ENCODING_EMBSTR:
-        newObject = createEmbeddedStringObject(data+sizeof(robj)+sizeof(mvcc), cb-sizeof(robj)-sizeof(uint64_t));
-        break;
-
-    case OBJ_ENCODING_RAW:
-        newObject = createObject(OBJ_STRING, sdsnewlen(SDS_NOINIT,cb-sizeof(robj)-sizeof(uint64_t)));
-        newObject->lru = oT->lru;
-        memcpy(newObject->m_ptr, data+sizeof(robj)+sizeof(mvcc), cb-sizeof(robj)-sizeof(mvcc));
-        break;
-
-    default:
-        serverPanic("Unknown string object encoding from storage");
-    }
-    setMvccTstamp(newObject, mvcc);
-
-    return newObject;
-}
-
-robj *deserializeStoredObjectCore(const void *data, size_t cb)
-{
-    switch (((char*)data)[0])
-    {
-        case RDB_TYPE_STRING:
-            return deserializeStoredStringObject(((char*)data)+1, cb-1);
-
-        default:
-            rio payload;
-            int type;
-            robj *obj;
-            rioInitWithConstBuffer(&payload,data,cb);
-            if (((type = rdbLoadObjectType(&payload)) == -1) ||
-                ((obj = rdbLoadObject(type,&payload,nullptr,NULL,OBJ_MVCC_INVALID)) == nullptr))
-            {
-                serverPanic("Bad data format");
-            }
-            if (rdbLoadType(&payload) == RDB_OPCODE_AUX)
-            {
-                robj *auxkey, *auxval;
-                if ((auxkey = rdbLoadStringObject(&payload)) == NULL) goto eoferr;
-                if ((auxval = rdbLoadStringObject(&payload)) == NULL) {
-                    decrRefCount(auxkey);
-                    goto eoferr;
-                }
-                if (strcasecmp(szFromObj(auxkey), "mvcc-tstamp") == 0) {
-                    setMvccTstamp(obj, strtoull(szFromObj(auxval), nullptr, 10));
-                }
-                decrRefCount(auxkey);
-                decrRefCount(auxval);
-            }
-        eoferr:
-            return obj;
-    }
-    serverPanic("Unknown object type loading from storage");
-}
-
-robj *deserializeStoredObject(const redisDbPersistentData *db, const char *key, const void *data, size_t cb)
-{
-    robj *o = deserializeStoredObjectCore(data, cb);
-    o->SetFExpires(db->setexpire()->exists(key));
-    return o;
-}
-
-sds serializeStoredObject(robj_roptr o, sds sdsPrefix)
-{
-    switch (o->type)
-    {
-        case OBJ_STRING:
-        {
-            sds sdsT = nullptr;
-            if (sdsPrefix)
-                sdsT = sdsgrowzero(sdsPrefix, sdslen(sdsPrefix)+1);
-            else
-                sdsT = sdsnewlen(nullptr, 1);
-            sdsT[sdslen(sdsT)-1] = RDB_TYPE_STRING;
-            return serializeStoredStringObject(sdsT, o);
-        }
-            
-        default:
-            rio rdb;
-            createDumpPayload(&rdb,o,nullptr);
-            if (sdsPrefix)
-            {
-                sds rval = sdscatsds(sdsPrefix, (sds)rdb.io.buffer.ptr);
-                sdsfree((sds)rdb.io.buffer.ptr);
-                return rval;
-            }
-            return (sds)rdb.io.buffer.ptr;
-    }
-    serverPanic("Attempting to store unknown object type");
-}
-
 redisObjectStack::redisObjectStack()
 {
     // We need to ensure the Extended Object is first in the class layout
@@ -1735,4 +1606,4 @@ void setMvccTstamp(robj *o, uint64_t mvcc)
         return;
     redisObjectExtended *oe = reinterpret_cast<redisObjectExtended*>(o) - 1;
     oe->mvcc_tstamp = mvcc;
-}
+}

src/pqsort.c

@@ -41,7 +41,6 @@
 
 #include <errno.h>
 #include <stdlib.h>
-#include <stdint.h>
 
 static inline char	*med3 (char *, char *, char *,
     int (*)(const void *, const void *));
@@ -63,7 +62,7 @@ static inline void	 swapfunc (char *, char *, size_t, int);
         } while (--i > 0);				\
 }
 
-#define SWAPINIT(a, es) swaptype = (uintptr_t)a % sizeof(long) || \
+#define SWAPINIT(a, es) swaptype = ((char *)a - (char *)0) % sizeof(long) || \
 	es % sizeof(long) ? 2 : es == sizeof(long)? 0 : 1;
 
 static inline void

src/rax.c

@@ -707,7 +707,7 @@ int raxGenericInsert(rax *rax, unsigned char *s, size_t len, void *data, void **
             nodesize = sizeof(raxNode)+trimmedlen+raxPadding(trimmedlen)+
                        sizeof(raxNode*);
             if (h->iskey && !h->isnull) nodesize += sizeof(void*);
-            trimmed = zcalloc(nodesize, MALLOC_LOCAL);//rax_malloc(nodesize);
+            trimmed = rax_malloc(nodesize);
         }
 
         if (postfixlen) {

src/rdb-s3.cpp

@@ -4,7 +4,7 @@
 #include <sys/wait.h>
 
 /* Save the DB on disk. Return C_ERR on error, C_OK on success. */
-int rdbSaveS3(char *s3bucket, const redisDbPersistentDataSnapshot **rgpdb, rdbSaveInfo *rsi)
+int rdbSaveS3(char *s3bucket, rdbSaveInfo *rsi)
 {
     int status = EXIT_FAILURE;
     int fd[2];
@@ -38,7 +38,7 @@ int rdbSaveS3(char *s3bucket, const redisDbPersistentDataSnapshot **rgpdb, rdbSa
             return C_ERR;
         }
 
-        if (rdbSaveFp(fp, rgpdb, rsi) != C_OK)
+        if (rdbSaveFp(fp, rsi) != C_OK)
         {
             fclose(fp);
             return C_ERR;

src/rdb.h

@@ -119,11 +119,10 @@
 #define RDB_MODULE_OPCODE_STRING 5  /* String. */
 
 /* rdbLoad...() functions flags. */
-#define RDB_LOAD_NONE       0
-#define RDB_LOAD_ENC        (1<<0)
-#define RDB_LOAD_PLAIN      (1<<1)
-#define RDB_LOAD_SDS        (1<<2)
-#define RDB_LOAD_SDS_SHARED ((1 << 3) | RDB_LOAD_SDS)
+#define RDB_LOAD_NONE   0
+#define RDB_LOAD_ENC    (1<<0)
+#define RDB_LOAD_PLAIN  (1<<1)
+#define RDB_LOAD_SDS    (1<<2)
 
 /* flags on the purpose of rdb save or load */
 #define RDBFLAGS_NONE 0                 /* No special RDB loading. */
@@ -152,15 +151,15 @@ int rdbLoadFile(const char *filename, rdbSaveInfo *rsi, int rdbflags);
 int rdbSaveBackground(rdbSaveInfo *rsi);
 int rdbSaveToSlavesSockets(rdbSaveInfo *rsi);
 void rdbRemoveTempFile(pid_t childpid, int from_signal);
-int rdbSave(const redisDbPersistentDataSnapshot **rgpdb, rdbSaveInfo *rsi);
-int rdbSaveFile(char *filename, const redisDbPersistentDataSnapshot **rgpdb, rdbSaveInfo *rsi);
-int rdbSaveFp(FILE *pf, const redisDbPersistentDataSnapshot **rgpdb, rdbSaveInfo *rsi);
-int rdbSaveS3(char *path, const redisDbPersistentDataSnapshot **rgpdb, rdbSaveInfo *rsi);
+int rdbSave(rdbSaveInfo *rsi);
+int rdbSaveFile(char *filename, rdbSaveInfo *rsi);
+int rdbSaveFp(FILE *pf, rdbSaveInfo *rsi);
+int rdbSaveS3(char *path, rdbSaveInfo *rsi);
 int rdbLoadS3(char *path, rdbSaveInfo *rsi, int rdbflags);
-ssize_t rdbSaveObject(rio *rdb, robj_roptr o, robj_roptr key);
+ssize_t rdbSaveObject(rio *rdb, robj_roptr o, robj *key);
 size_t rdbSavedObjectLen(robj *o, robj *key);
 robj *rdbLoadObject(int type, rio *rdb, sds key, int *error, uint64_t mvcc_tstamp);
-void backgroundSaveDoneHandler(int exitcode, bool fCancelled);
+void backgroundSaveDoneHandler(int exitcode, int bysignal);
 int rdbSaveKeyValuePair(rio *rdb, robj *key, robj *val, long long expiretime);
 ssize_t rdbSaveSingleModuleAux(rio *rdb, int when, moduleType *mt);
 robj *rdbLoadCheckModuleValue(rio *rdb, char *modulename);
@@ -173,7 +172,7 @@ int rdbLoadBinaryDoubleValue(rio *rdb, double *val);
 int rdbSaveBinaryFloatValue(rio *rdb, float val);
 int rdbLoadBinaryFloatValue(rio *rdb, float *val);
 int rdbLoadRio(rio *rdb, int rdbflags, rdbSaveInfo *rsi);
-int rdbSaveRio(rio *rdb, const redisDbPersistentDataSnapshot **rgpdb, int *error, int flags, rdbSaveInfo *rsi);
+int rdbSaveRio(rio *rdb, int *error, int rdbflags, rdbSaveInfo *rsi);
 rdbSaveInfo *rdbPopulateSaveInfo(rdbSaveInfo *rsi);
 
 #endif

src/readwritelock.h

@@ -1,96 +0,0 @@
-#pragma once
-#include <condition_variable>
-
-class readWriteLock {
-    fastlock m_readLock;
-    fastlock m_writeLock;
-    std::condition_variable_any m_cv;
-    int m_readCount = 0;
-    int m_writeCount = 0;
-    bool m_writeWaiting = false;
-public:
-    readWriteLock(const char *name) : m_readLock(name), m_writeLock(name) {}
-
-    void acquireRead() {
-        std::unique_lock<fastlock> rm(m_readLock);
-        while (m_writeCount > 0 || m_writeWaiting)
-            m_cv.wait(rm);
-        m_readCount++;
-    }
-    
-    bool tryAcquireRead() {
-        std::unique_lock<fastlock> rm(m_readLock, std::defer_lock);
-        if (!rm.try_lock())
-            return false;
-        if (m_writeCount > 0 || m_writeWaiting)
-            return false;
-        m_readCount++;
-        return true;
-    }
-
-    void acquireWrite(bool exclusive = true) {
-        std::unique_lock<fastlock> rm(m_readLock);
-        m_writeWaiting = true;
-        while (m_readCount > 0)
-            m_cv.wait(rm);
-        if (exclusive) {
-            /* Another thread might have the write lock while we have the read lock
-               but won't be able to release it until they can acquire the read lock
-               so release the read lock and try again instead of waiting to avoid deadlock */
-            while(!m_writeLock.try_lock())
-                m_cv.wait(rm);
-        }
-        m_writeCount++;
-        m_writeWaiting = false;
-    }
-
-    void upgradeWrite(bool exclusive = true) {
-        releaseRead();
-        acquireWrite(exclusive);
-    }
-
-    bool tryAcquireWrite(bool exclusive = true) {
-        std::unique_lock<fastlock> rm(m_readLock, std::defer_lock);
-        if (!rm.try_lock())
-            return false;
-        if (m_readCount > 0)
-            return false;
-        if (exclusive)
-            if (!m_writeLock.try_lock())
-                return false;
-        m_writeCount++;
-        return true;
-    }
-
-    void releaseRead() {
-        std::unique_lock<fastlock> rm(m_readLock);
-        m_readCount--;
-        m_cv.notify_all();
-    }
-
-    void releaseWrite(bool exclusive = true) {
-        std::unique_lock<fastlock> rm(m_readLock);
-        serverAssert(m_writeCount > 0);
-        if (exclusive)
-            m_writeLock.unlock();
-        m_writeCount--;
-        m_cv.notify_all();
-    }
-
-    void downgradeWrite(bool exclusive = true) {
-        releaseWrite(exclusive);
-        acquireRead();
-    }
-
-    bool hasReader() {
-        return m_readCount > 0;
-    }
-
-    bool hasWriter() {
-        return m_writeCount > 0;
-    }
-
-    bool writeWaiting() {
-        return m_writeWaiting;
-    }
-};

src/redis-benchmark.cpp

@@ -42,7 +42,6 @@
 #include <assert.h>
 #include <math.h>
 #include <pthread.h>
-#include <string>
 extern "C" {
 #include <sdscompat.h> /* Use hiredis' sds compat header that maps sds calls to their hi_ variants */
 #include <sds.h> /* Use hiredis sds. */
@@ -871,10 +870,6 @@ static void createMissingClients(client c) {
     }
 }
 
-extern "C" void asyncFreeDictTable(dictEntry **de) {
-    zfree(de);
-}
-
 static void showLatencyReport(void) {
 
     const float reqpersec = (float)config.requests_finished/((float)config.totlatency/1000.0f);
@@ -1017,11 +1012,7 @@ static void benchmark(const char *title, const char *cmd, int len) {
     createMissingClients(c);
 
     config.start = mstime();
-    if (!config.num_threads) {
-        aeThreadOnline();
-        aeMain(config.el);
-        aeThreadOffline();
-    }
+    if (!config.num_threads) aeMain(config.el);
     else startBenchmarkThreads();
     config.totlatency = mstime()-config.start;
 
@@ -1061,9 +1052,7 @@ static void freeBenchmarkThreads() {
 
 static void *execBenchmarkThread(void *ptr) {
     benchmarkThread *thread = (benchmarkThread *) ptr;
-    aeThreadOnline();
     aeMain(thread->el);
-    aeThreadOffline();
     return NULL;
 }
 
@@ -1702,7 +1691,7 @@ int main(int argc, const char **argv) {
     int len;
 
     client c;
-    aeThreadOnline();
+
     storage_init(NULL, 0);
 
     srandom(time(NULL) ^ getpid());
@@ -1755,7 +1744,6 @@ int main(int argc, const char **argv) {
         cliSecureInit();
     }
 #endif
-    aeThreadOffline();
 
     if (config.cluster_mode) {
         // We only include the slot placeholder {tag} if cluster mode is enabled
@@ -1992,58 +1980,11 @@ int main(int argc, const char **argv) {
             sdsfree(key_placeholder);
         }
 
-        if (test_is_selected("mget")) {
-            const char *cmd_argv[1002];
-            cmd_argv[0] = "MGET";
-            sds key_placeholder = sdscatprintf(sdsnew(""),"key%s:__rand_int__",tag);
-            for (int keys = 1; keys < 1002; keys += 100) {
-                for (i = 1; i < keys + 1; i++) {
-                    cmd_argv[i] = key_placeholder;
-                }
-                len = redisFormatCommandArgv(&cmd,keys+1,cmd_argv,NULL);
-                std::string title = "MGET (" + std::to_string(keys) + " keys)";
-                benchmark(title.data(),cmd,len);
-                free(cmd);
-            }
-            sdsfree(key_placeholder);
-        }
-
-        if (test_is_selected("hmset")) {
-            const char *cmd_argv[22];
-            cmd_argv[0] = "HMSET";
-            cmd_argv[1] = "testhash";
-            sds key_placeholder = sdscatprintf(sdsnew(""),"key%s:__rand_int__",tag);
-            for (i = 2; i < 22; i += 2) {
-                cmd_argv[i] = key_placeholder;
-                cmd_argv[i+1] = data;
-            }
-            len = redisFormatCommandArgv(&cmd,22,cmd_argv,NULL);
-            benchmark("MSET (10 keys)",cmd,len);
-            free(cmd);
-            sdsfree(key_placeholder);
-        }
-
-        if (test_is_selected("hmget")) {
-            const char *cmd_argv[1003];
-            cmd_argv[0] = "HMGET";
-            cmd_argv[1] = "testhash";
-            sds key_placeholder = sdscatprintf(sdsnew(""),"key%s:__rand_int__",tag);
-            for (int keys = 1; keys < 1002; keys += 100) {
-                for (i = 2; i < keys + 2; i++) {
-                    cmd_argv[i] = key_placeholder;
-                }
-                len = redisFormatCommandArgv(&cmd,keys+2,cmd_argv,NULL);
-                std::string title = "HMGET (" + std::to_string(keys) + " keys)";
-                benchmark(title.data(),cmd,len);
-                free(cmd);
-            }
-            sdsfree(key_placeholder);
-        }
-
         if (!config.csv) printf("\n");
     } while(config.loop);
     zfree(data);
 
+    zfree(data);
     if (config.redis_config != NULL) freeRedisConfig(config.redis_config);
 
     return 0;

src/redis-cli-cpphelper.cpp

@@ -431,7 +431,7 @@ int clusterManagerGetAntiAffinityScore(clusterManagerNodeArray *ipnodes,
                 }
             }
         }
-        if (offending_len != NULL) *offending_len = offending_p - *offending;
+        //if (offending_len != NULL) *offending_len = offending_p - *offending;
         dictReleaseIterator(iter);
         dictRelease(related);
     }

src/redis-cli.c

@@ -153,11 +153,6 @@ static void cliRefreshPrompt(void) {
     sdsfree(prompt);
 }
 
-struct dictEntry;
-void asyncFreeDictTable(struct dictEntry **de) {
-    zfree(de);
-}
-
 /* Return the name of the dotfile for the specified 'dotfilename'.
  * Normally it just concatenates user $HOME to the file specified
  * in 'dotfilename'. However if the environment variable 'envoverride'
@@ -5485,8 +5480,8 @@ static int clusterManagerCommandRebalance(int argc, char **argv) {
             listAddNodeTail(lsrc, src);
             table = clusterManagerComputeReshardTable(lsrc, numslots);
             listRelease(lsrc);
-            int table_len = 0;
-            if (!table || (table_len = (int) listLength(table)) != numslots) {
+            int table_len = (int) listLength(table);
+            if (!table || table_len != numslots) {
                 clusterManagerLogErr("*** Assertion failed: Reshard table "
                                      "!= number of slots");
                 result = 0;
@@ -6835,17 +6830,17 @@ void bytesToHuman(char *s, long long n) {
     }
     if (n < 1024) {
         /* Bytes */
-        snprintf(s,256,"%lldB",n);
+        sprintf(s,"%lldB",n);
         return;
     } else if (n < (1024*1024)) {
         d = (double)n/(1024);
-        snprintf(s,256,"%.2fK",d);
+        sprintf(s,"%.2fK",d);
     } else if (n < (1024LL*1024*1024)) {
         d = (double)n/(1024*1024);
-        snprintf(s,256,"%.2fM",d);
+        sprintf(s,"%.2fM",d);
     } else if (n < (1024LL*1024*1024*1024)) {
         d = (double)n/(1024LL*1024*1024);
-        snprintf(s,256,"%.2fG",d);
+        sprintf(s,"%.2fG",d);
     }
 }
 
@@ -6875,12 +6870,12 @@ static void statMode(void) {
         for (j = 0; j < 20; j++) {
             long k;
 
-            snprintf(buf,64,"db%d:keys",j);
+            sprintf(buf,"db%d:keys",j);
             k = getLongInfoField(reply->str,buf);
             if (k == LONG_MIN) continue;
             aux += k;
         }
-        snprintf(buf,64,"%ld",aux);
+        sprintf(buf,"%ld",aux);
         printf("%-11s",buf);
 
         /* Used memory */
@@ -6890,23 +6885,23 @@ static void statMode(void) {
 
         /* Clients */
         aux = getLongInfoField(reply->str,"connected_clients");
-        snprintf(buf,64,"%ld",aux);
+        sprintf(buf,"%ld",aux);
         printf(" %-8s",buf);
 
         /* Blocked (BLPOPPING) Clients */
         aux = getLongInfoField(reply->str,"blocked_clients");
-        snprintf(buf,64,"%ld",aux);
+        sprintf(buf,"%ld",aux);
         printf("%-8s",buf);
 
         /* Requests */
         aux = getLongInfoField(reply->str,"total_commands_processed");
-        snprintf(buf,64,"%ld (+%ld)",aux,requests == 0 ? 0 : aux-requests);
+        sprintf(buf,"%ld (+%ld)",aux,requests == 0 ? 0 : aux-requests);
         printf("%-19s",buf);
         requests = aux;
 
         /* Connections */
         aux = getLongInfoField(reply->str,"total_connections_received");
-        snprintf(buf,64,"%ld",aux);
+        sprintf(buf,"%ld",aux);
         printf(" %-12s",buf);
 
         /* Children */

src/redis-cli.h

@@ -97,6 +97,9 @@ extern "C" {
 #define CC_FORCE (1<<0)         /* Re-connect if already connected. */
 #define CC_QUIET (1<<1)         /* Don't log connecting errors. */
 
+struct clusterManagerLink;
+typedef struct clusterManagerLink clusterManagerLink;
+
 /* Dict Helpers */
 
 uint64_t dictSdsHash(const void *key);

src/rio.cpp

@@ -73,14 +73,6 @@ static size_t rioBufferRead(rio *r, void *buf, size_t len) {
     return 1;
 }
 
-static size_t rioConstBufferRead(rio *r, void *buf, size_t len) {
-     if (r->io.buffer.len-r->io.buffer.pos < (off_t)len)
-        return 0; /* not enough buffer to return len bytes. */
-    memcpy(buf,r->io.buffer.ptr+r->io.buffer.pos,len);
-    r->io.buffer.pos += len;
-    return 1;
-}
-
 /* Returns read/write position in buffer. */
 static off_t rioBufferTell(rio *r) {
     return r->io.buffer.pos;
@@ -99,29 +91,11 @@ static const rio rioBufferIO = {
     rioBufferTell,
     rioBufferFlush,
     NULL,           /* update_checksum */
-    NULL,           /* update checksum arg */
     0,              /* current checksum */
     0,              /* flags */
     0,              /* bytes read or written */
     0,              /* keys since last callback */
     0,              /* read/write chunk size */
-    0,              /* last update time */
-    { { NULL, 0 } } /* union for io-specific vars */
-};
-
-static const rio rioConstBufferIO = {
-    rioConstBufferRead,
-    nullptr,
-    rioBufferTell,
-    rioBufferFlush,
-    NULL,           /* update_checksum */
-    NULL,           /* update checksum arg */
-    0,              /* current checksum */
-    0,              /* flags */
-    0,              /* bytes read or written */
-    0,              /* keys since last callback */
-    0,              /* read/write chunk size */
-    0,              /* last update time */
     { { NULL, 0 } } /* union for io-specific vars */
 };
 
@@ -131,14 +105,6 @@ void rioInitWithBuffer(rio *r, sds s) {
     r->io.buffer.pos = 0;
 }
 
-void rioInitWithConstBuffer(rio *r, const void *buf, size_t cb)
-{
-    *r = rioConstBufferIO;
-    r->io.buffer.ptr = (sds)buf;
-    r->io.buffer.pos = 0;
-    r->io.buffer.len = cb;
-}
-
 /* --------------------- Stdio file pointer implementation ------------------- */
 
 /* Returns 1 or 0 for success/failure. */
@@ -180,13 +146,11 @@ static const rio rioFileIO = {
     rioFileTell,
     rioFileFlush,
     NULL,           /* update_checksum */
-    NULL,           /* update checksum arg */
     0,              /* current checksum */
     0,              /* flags */
     0,              /* bytes read or written */
     0,              /* keys since last callback */
     0,              /* read/write chunk size */
-    0,              /* last update time */
     { { NULL, 0 } } /* union for io-specific vars */
 };
 
@@ -281,13 +245,11 @@ static const rio rioConnIO = {
     rioConnTell,
     rioConnFlush,
     NULL,           /* update_checksum */
-    NULL,           /* update checksum arg */
     0,              /* current checksum */
     0,              /* flags */
     0,              /* bytes read or written */
     0,              /* keys since last callback */
     0,              /* read/write chunk size */
-    0,              /* last update time */
     { { NULL, 0 } } /* union for io-specific vars */
 };
 
@@ -402,13 +364,11 @@ static const rio rioFdIO = {
     rioFdTell,
     rioFdFlush,
     NULL,           /* update_checksum */
-    NULL,           /* update checksum arg */
     0,              /* current checksum */
     0,              /* flags */
     0,              /* bytes read or written */
     0,              /* keys since last callback */
     0,              /* read/write chunk size */
-    0,              /* last update time */
     { { NULL, 0 } } /* union for io-specific vars */
 };
 

src/rio.h

@@ -58,7 +58,6 @@ struct _rio {
      * and len fields pointing to the new block of data to add to the checksum
      * computation. */
     void (*update_cksum)(struct _rio *, const void *buf, size_t len);
-    void *chksum_arg;
 
     /* The current checksum and flags (see RIO_FLAG_*) */
     uint64_t cksum, flags;
@@ -72,16 +71,12 @@ struct _rio {
     /* maximum single read or write chunk size */
     size_t max_processing_chunk;
 
-    /* last update time */
-    long long last_update;
-
     /* Backend-specific vars. */
     union {
         /* In-memory buffer target. */
         struct {
             sds ptr;
             off_t pos;
-            off_t len;  // For const buffers only
         } buffer;
         /* Stdio file pointer target. */
         struct {
@@ -170,7 +165,6 @@ static inline void rioClearErrors(rio *r) {
 
 void rioInitWithFile(rio *r, FILE *fp);
 void rioInitWithBuffer(rio *r, sds s);
-void rioInitWithConstBuffer(rio *r, const void *rgch, size_t cch);
 void rioInitWithConn(rio *r, connection *conn, size_t read_limit);
 void rioInitWithFd(rio *r, int fd);